bpipes?

doehps
jeakyptacioreectpeepsecitedag pints
WILEY SERIES IN PURE AND APPLIED OPTICS

Founded by Stanley S. Ballard, University of Florida

ADVISORY EDITOR: Joseph W. Goodman, Stanford University

ALLEN AND EBERLY ® Optical Resonance and Two-Level Atoms

BABCOCK ©° Silicate Glass Technology Methods
BOND ® Crystal Technology
CATHEY ® Optical Information Processing and Holography
DE VANY ®* Master Optical Techniques
DUFFIEUX ® The Fourier Transform and Its Applications to Optics, Second Edition
EBERLY AND LAMBROPOULOS ° Multiphoton Processes
ELMER ° The Optical Design of Reflectors, Second Edition
GASKILL °* Linear Systems, Fourier Transforms, and Optics
GERRARD AND BURCH ® Introduction to Matrix Methods in Optics
GOLDIN ° Waves and Photons: An Introduction to Quantum Optics
HUDSON ° Infrared System Engineering
JUDD AND WYSZECKI ® Color in Business, Science, and Industry, Third Edition
KNITTL ® Optics of Thin Films
LENGYEL ®* Lasers, Second Edition
LEVI * Applied Optics, A Guide to Optical System Design, Volume I
Volume II
LOUISELL * Quantum Statistical Properties of Radiation
MALACARA ° Optical Shop Testing
MARATHAY ©® Elements of Optical Coherence Theory
McCARTNEY ® Absorption and Emission by Atmospheric Gases: The Physical Processes
McCARTNEY ® Optics of the Atmosphere: Scattering by Molecules and Particles
MIDWINTER °® Optical Fibers for Transmission
SUEMATSU AND IGA ® Jntroduction to Optical Fiber Communications
VEST ° Holographic Interferometry
WILLIAMS AND BECKLUND ©° Optics
WYSZECKI AND STILES ° Color Science: Concepts and Methods, Quantitative Data
and Formulae, Second Edition
ZERNIKE AND MIDWINTER ® Applied Nonlinear Optics
| About the authors
Gunter Wyszecki is currently Assis-
tant Director of the Division of Physics,
National Research Council of Canada,
and Head of its Optics Section. Prior to
this, he held positions with the Nation-
al Bureau of Standards (as a Fulbright
Scholar, 1953-54) and the Berlin Fed-
eral Bureau of Testing Materials
| (1954-55). In 1979, Dr. Wyszecki re-
_ ceived the Gold Medal of the Associa-
| tion Internationale de la Couleur, the
Inter-Society Color Council’s Godlove
| Award,and the Federation of Societies
| of Coating Technologies’ Bruning
| Award. He is a Fellow of the Royal
Society of Canada, the Optical Society
of America, the Illuminating Engineer-
ing Society, and Vice-President of the
Commission Internationale de
l’Eclairage. Dr. Wyszecki is the co-au-
| thor (with D.B. Judd) of Color in Busi-
ness, Science, and Industry, 3rd Edi-
tion (Wiley). He received his Dr.-Ing.
from the Technical University of Berlin.
W.S. Stiles was Deputy Chief Scien-
tific Officer, National Physical Labora-
| tories, Teddington, England, until his
retirement in 1961. The joint discover-
er of the Stiles-Crawford Effect, Dr.
Stiles recently published his selected
papers in a volume entitled Mecha-
nisms of Colour Vision. Among his
many honors and awards, he received
the 1944 Carpenter Medal of London
University and the 1964 Tillyer Medal
of the Optical Society of America. Dr.
Stiles is a past president of the Illumi-
nating Engineering Society of Great
Britain, past chairman of the Colour
Group of Great Britain, a Fellow of the
Royal Society, and the Optical Society
of America, and an Officer of the
British Empire. He received his Ph.D.
and D.Sc. from University College,
London.
 acaiae
i

aM- a id.

ra
0 ;

. wee”ra Wy)

=
COLOR SCIENCE
COLOR SCIENCE
Concepts and Methods,
Quantitative Data and Formulae,
2nd Edition

GUNTER WYSZECKI
National Research Council,
Ottawa, Ontario, Canada

W. S. STILES
Richmond, Surrey, England

YEAR»
9 O

1807 1982
“OR I aan

A Wiley-Interscience Publication

John Wiley & Sons

New York °: Chichester - Brisbane - Toronto Singapore
 PACKARD LIBRARY
JUN 18 1984

OF-THE COLEMBUS COLLEGE OF ART ANB DESIGH

Copyright © 1982 by John Wiley & Sons, Inc.

All rights reserved. Published simultaneously in Canada.

Reproduction or translation of any part of this work

beyond that permitted by Section 107 or 108 of the
1976 United States Copyright Act without the permission
of the copyright owner is unlawful. Requests for
permission or further information should be addressed to
the Permissions Department, John Wiley & Sons, Inc.

Library of Congress Cataloging in Publication Data:

Wyszecki, Ginter.
Color Science.

(The Wiley series in pure and applied optics)

“A Wiley-Interscience publication.”
Bibliography: p.
Includes index.
1. Color. I. Stiles, W. S. (Walter Stanley),
1901- ; II. Title. III. Series.

QC495.W88 1982 535.6 82-2794

ISBN 0-471-02106-7 AACR?

Printed in the United States of America

10h. «8* 7 KET

3 #2. o
PREFACE

The apparent success of the first edition of this book, published in 1967,
and the persistent encouragement of many of our colleagues was necessary
and sufficient persuasion for us to write the second edition. Our basic
approach to color science has not changed. The second edition is again a
collection of concepts and methods, quantitative data, and formulae bearing
on color science and directed to the advanced student and research worker
in color—physicist, psychologist, and physiologist—and those actively en-
gaged in color engineering projects—illuminating engineers, designers, and
industrial consultants. Our aim has again been to assemble the quantitative
tools for work on color: details of light sources, color filters, monochroma-
tors, and photon detectors; in depth accounts of the working concepts in
color matching, color discrimination, and chromatic adaptation; logical
presentations of the mathematical algorithms used in the development of
these concepts; international standards and procedures of the measurement
of light and color.
Much of the material selected for the book is organized in tabular or
mathematical form, stressing data and formulae that have gained accep-
tance as standard or hard data applicable to color calculations or further
developments in color-vision research. Some of the contents are presented in
considerable detail so as to make the book as much as possible a unique
source of reference material for the research worker and serious student of
color science. Again, trichromatic principles and their application loom
large in this book and, in fact, have been expanded somewhat to include
Maxwell’s method of trichromatic matching. Color matching is here deployed
from first principles and in later sections has been linked to issues of
color-vision mechanisms.
A great deal of new and expanded material has been included in the
second edition, reflecting the considerable growth of knowledge that has
taken place since the 1960s. This has particularly been the~¢as¢“i color
matching, increment-threshold work, chromatic adaptation, and the theoret-
ical modeling of color discrimination and other color-vision phenomena. We
have attempted to keep pace with the new developments, but this has not
been an easy task. For the most part, descriptive and qualitative material on
color phenomena which would properly find a place in a textbook or
introductory treatise on color has not been included. We have also refrained
from delving too deeply into highly speculative issues of color-vision model-
ing, which have attracted a number of research workers, particularly in
vi PREFACE

recent years. Our account on theories and models of color vision is only a
brief review of the basic principles and approaches that have emerged.
We wish to acknowledge with thanks our indebtedness to our many
associates.

In Canada and the United States. At the National Research Council of

Canada: W. Budde, W. B. Cowan, G. H. Fielder, A. R. Robertson, C. Ware;
at York University, Toronto: P. K. Kaiser; at Carleton University, Ottawa:
B. Tansley; at the University of California, San Diego: R. M. Boynton and
D. I. A. MacLeod; at the University of Chicago: J. Pokorny and V. C.
Smith; at the Ohio State University: C. Ingling; at the University of
Michigan: M. Alpern; at the University of Pennsylvania: L. M. Hurvich, D.
Jameson, and E. N. Pugh; at Indiana University: S. L. Guth; at Eastman
Kodak Research Laboratories: C. J. Bartleson, F. Grum; at Rochester:
D. L. MacAdam, and at Alexandria, Va.: D. Nickerson.

In Europe. At the University of Cambridge: J. Mollen, W. A. H. Rushton

(recently deceased); at Kodak Ltd.: R. W. G. Hunt; at City University:
P. W. Trezona; at the Institute of Ophthalmology at London: H. J. A.
Dartnall, D. A. Palmer; at the National Physical Laboratory: F. J. J. Clarke,
O. C. Jones; at London: B. H. Crawford and W. D. Wright; at the
University of Karlsruhe: H. W. Bodmann; at Berlin: M. Richter; at the
Institute of Perception at Soesterberg, Netherlands: J. J. Vos and P. L.
Walraven; at the University of Amsterdam: O. Estévez.

In Japan At the Electrotechnical Laboratory: T. Nayatani. N. Ooba; at

Fuji Photo Research Laboratories: N. Ohta.

In Australia. At the National Measurement Laboratory: W. R. Blevin.

It is also with great pleasure that we acknowledge the fine work of Miss
Lorna Silcox; Mrs. Josie Graham, who painstakingly typed the lengthy
manuscript; and the production staff of John Wiley and Sons, who, once
again, produced the book in expert fashion.

GUNTER WYSZECKI
W.S. STILES

August 1982
ACKNOWLEDGMENT

Numerous illustrations and tables in this book are taken from, or based on,
the work of many different authors published in a variety of journals and
books. In each instance we have cited the names of these authors in the
captions of the illustrations and legends of the tables, and the complete
reference to the original publication is given in the Reference Section. The
illustrations and tables listed below have been reproduced in detail, and
permission to do so has been received from the following publishing bodies
who are the holders of the copyrights. We wish to express our grateful
acknowledgment to all the authors and publishing bodies for their kind
cooperation.

British Medical: Fig. 1(2.4.5) and Tables 2(2.4.5) and 3(2.4.5)

Cambridge University Press: Figs. 1(5.6.5) and 4(7.3.1)
Illuminating Engineering Society: Fig. 6(1.2.3)
Instituto de Biofisica, Rio de Janeiro: Fig. 1(2.4.7)
John Wiley & Sons, Inc.: Figs. 1(1.4), 2(1.4), 1(3.3.9), 2(3.3.9), 1(3.3.10),
8(3.8.2), 1 1061836) 91d (3-8:6),, 12.8.6), ..2(3..12.2),” 3(3,12.3),, -4(3:12.3),
2(3.12:3)) 20325) 4Q. 12.5), 3(5.4.1),- 7(52:2).2 8(5.12.2),, .9(5:12.2),
VO shoec ya tele mor oes 83O123) a0 23) (5,1 225) O(5212 5):
2(5.14.1), 1(5.14.2), 3(5.15), 4(6.3), 5(6.3), 1(6.6.1), 2(6.6.1), 3(6.6.1),
1(7.10.5), Tables 1(3.3.5), 1(3.3.8), 11(3.3.8), I1(3.3.8), [V(3.3.8), 1(3.3.8),
2(3.3.8), 3(3-3.8), 1(3.3.10), 2(3.3.10), 13:9); 11(6.5. 1A; B, C), 1(5:5.2)
Junk BW Publishers: Figs. 2(7.3.2), 1(7.4.1) and 2(7.4.1)
S. Karger AG: Fig. 2(2.4.6)
Kimpton Publishers: Figs. 3(5.14.1), 4(5.14.2), and 5(5.14.2)
National Physical Laboratories, Teddington, Middlesex: Figs. 4(5.4.1),
1(5.5.4), 1(5.10), 5(5.12.2)
Optical Society of America: Figs. 2(1.2.1), 3(1.3.7), 7(1.3.7), 3(1.4), 1(1.4.1),
11(1.6.2), 1(2.4.3), 7(2.4.6), 1(3.6), 1(3.8.6), 2(3.8.6), 3(3.8.6), 4(3.8.6),
5(3.8.6), 6(3.8.6), 7(3.8.6), 8(3.8.6), 1(3.11), 2(3.11), 3(4.3.5), 7(4.3.5),
5(4.3.6), 1(5.4.3), 2(5.4.3), 1(5.4.4), 1(5.4.5), 2(5.4.5), 3(5.4.5), 4(5.4.5),
Meee ele LN 2.0.1 Je Ol 0. yw Mo L222 ).01(5, 13), 205313);
PG Ay 1G 511) ol 99), Ol.3), 19.5), aU at o)y) LOC 1.9.5),) 2(1.10.2),
3(7.10.2), 1(7.10.6), 2(7.10.6), 3(7.10.6), 4(7.10.6), 5(7.10.6), 2(8.3.3),
3(8.3.3), 1(8.3.4), 2(8.3.4), 3(8.3.4), 4(8.3.4), 5(8.3.4), 6(8.3.4), 7(8.3.4),
8(8.3.4), Tables 1(3.8.2), II(3.8.2), IV(3.8.2), 2(5.4.1), 1(5.4.2), I1(5.4.2),
1(5.4.3), 11(5.4.4), 1(6.6.1), 1(6.6.4)
Vii
vii ACKNOWLEDGMENT

Pergamon Press, Inc.: Figs. 2(5.6.5), 3(5.6.5), 4(5.6.5), 5(5.6.5), 11(5.7.2),

3(5.12.2), 2(7.4.4), 1(7.9.1), 1(8.2.4), 2(8.2.4), 3(8.2.4), 4(8.2.4), 5(8.2.4),
2(8.2.6), 3(8.2.6), 5(8.4.4), 1(8.4.6), 2(8.4.6)
Physical Society: Fig. 2(5.5.6)
Mrs. Stephen Polyak: Figs. 1(2.2.4) and 1(2.2.6)
The Royal Society (London): Figs. 2(2.2.4), 1(5.11), 1(7.3.2)
Science (American Association for the Advancement of Science): Figs.
11(7.9.3) and 1(8.3.5)
Taylor & Francis Ltd.: Figs. 6(5.5.6), 7(5.5.6), 8(5.6.2), 1(5.6.3), 2(5.11),
1(7.5)
CONTENTS

CHAPTER 1 PHYSICAL DATA

ieee basicis aciometiic Guanes ANG WITILS sn. yes ee eee oe l

D2) moources' Of Radianiseneroy oo) Nee Ok ON hn ee we GOR. Ls ]

eee Peerhe Sureandaviight4 4 42 tes oc 2.4 toe SER, yee ae 4
Ae a CTUIAN INAUIALOIS Ce en core coon ois cee cee ee Se, Bok 11
122 ee ECC Disclrarge Laps ett es ets ee ets ee te Sas 19
[i eee CICCLOIUIIIE SCENE SOUT CES ae ce Re Me Ree teen; oe ks 23
(OS We uisht Pattie Diodes .4%.4044..0444
OS FBe ee, o. 25
ey re et oe eae Ee ee an oe ene a 4 2S
l Dal soources op Ultraviolet Radiant Enersy..s ....5% 9.20. «6 as es 27
1.2.8 Some Photometric and Colorimetric Characteristics of
Sources of Radiant Eneroyiae? toasted a: ansad.. LE E..-- a4|

USS MeO DICAlIE Mects ere te ere ce eaten ee ee oie ee ee ee rere soa ee 30
L3i, . -AbsorpionFilterpasat a8s beste) AT atk actus eietoet > * fo 30
RES PCE REISS TTT(ee kale OR reer tyar eR Se]PORE ee eae Se aR REN: 34
Rs ae CoeLatiibstel steer er onote a ee ee ee ee eg aed +p
Ree Pd EAT ENTANG goUnTor ne oiealliapety 5 caalbelic}, othe: ye Ratetctialal dg r i P Ieae 35
1.3.5 Absorption Filters for Special Applications ............... 35
133.64) aMiscellanGous Absorption Filters os at tenes in eects oe es 40
Rem CELEN COP tel sete ee et on seston ts ntem fe co) ater st: 3 40
Lee RRP CrerencesFitcr WEURES aie re eanete eter CaM. eee eo, 46
Ses EPL STE TT ETE aspS ornae ha, a Ree retapet pts Maule ae dt hl oR hg 48
Uy Ome SMCCE OLOIIZETS ote ta, Oe ari Ot cage Rice ee aie aoe « 49
13-11 eMviscellaneéus‘Optical Pilters.«.. oe 4 oh ae ee hee oe a1

ted) geReflecting Materials 0 iy tics Rite uiine ies», Ueerayeiriiaars

«« 2. 51
We eee TONE SUT aCe Mell TOle ne ert se Boe een On: re Ri ae a)
P42 seenile Retiectance olanGalus 2 2. oe ee oe es ee es oe 55
143 4, Golored- Reflectance Standards west} .ctasevidel fF-- 1-08... - 58
1.4.4. .Black Surfaces) 20): 4....,<,-. +... Saba Seen? ~~ LGR. oo 60
14:5 so.Buslding Materials ... steer orsersiiety } signet -. $0.8 - > 60
1.4.6. .Natural Objects |....:...,.. dherntt 2 tnla’) Hiskiieen) AR oes 60

1X
X CONTENTS

1.5 Monochromators sxc. «dseighaey eet ek ees eee 63

L5.1 Basic Designsiol Monochtomalorsys... cer .,.0e < he wie: 63
1.5.2 Resolving Power and Transmitted Radiant Flux............ 66
18523 tee SHitaw idth @aorrection tice ose ete ts rae ve 69
LSA, BRolanization 6g Pee eee se a toma ie teen ae 70
135;5, 4 “Wavelensth Calibration cake: Sum. oh ete ee tl
1:5,6.~ Stray Laght ca. 3 totes eke eee ee ere i (p

1.6 Physical Detectors of Radiant Energ Vink oe oe ee ee 72

l6s¥-. ‘ThermahDetectors 144, 0). case a ee 73
1.6:2-° >:Photon Detectors... 0 o.2 cata eS ees 76

CHAPTER 2 THE EYE

2.1 Preamblew ogi s-sc capateetarts he eke eae ae ee ee ene ee 83

2.2 ‘The Structure‘of therhuman yeoman: (ec c an eeeeee 83

2.2.1¢°-Gomea #32200 08S shin nk. eee ae. SS 84
Ddedy MENS, | creas hate ocOia ee ae a eee 84
2.2.3 Aqueous Humor and Vitreous Body, 2s. scneaustt = eee: -- 84
2.2.4 ~The Fine-Structure of the Retina, 6. .c8¥ uccstel. ee Pe ao 85
2.2.5 | Main Topographical Features of the Retina............... 88
22.6 “The Fovea’ 222 2 Ane.) sae ee ae ee eee 90
22,8. Lhe Photorecept6ts 32. a ss. oe de ee 90

2.3 Specification of the External Stimulus ge. ee ks ee i of

2.3:1-'. -Position in External Ficld 9) eee Oe, 93
25.2 RadiomeinicSpecications: 7... 2... - ee ee 93
2.3:3 ,, Photometric Specifications ===. 2...) - seeee ee 95

2.4 Factors in the Eye That Control the Internal Stimulus ............. Js)
2.4.1. Image Formation by the Theoretical Eye ................. 98
2.4.2) “Hye Axes and yesAnpies 00.2 6... cae er a Bey 100
2.43 . Chromatic§tAbercation ol the hyve oe ao ae ee 101
2.4.4 The Troland Values of Retinal luminance .............. 102
DAD. Pupil Size) ies, canon. ek vie aaa ne es a ee 105
2.4.6 Light Losses inthe Eye eset es a ee oy 107
24:1 Rluorescent Light inthe’ Hye... > eee ee ee 116

CHAPTER 3 COLORIMETRY

3.1 Préamble +.» 0500.0.-.tce0 dss hatin eh rae tatese ees SOURS, DEDIROROOL cm Ly

3.2 Basic Colorimetric Concepts), 74. eees

be ee ie 117
3.2.1 - Trichromatic.Generalization™, Peres PaO! ... LE. |. 117
3.2;2.)* Inistimulus-Space +..<.04.:.0.05 5 ee TO, 8, 119
3.2.3 - Basie Colorimetric Equations .; Sema BAe, .. Oe... 121
3:2.4. Imaginary Color Stimuli...) ... Pee SRAM. GA 127
3.2.5. Colorimetric Transiormations +30) eee 129
 Contents Xi

93. The ClE-Colonmetric Systems. cancsa. oiostatadt beahaay?-oith ». 130

3.3.1 The CIE 1931 Standard Colorimetric
fODSErver un ele. ars ersacil? swbenewon? bupkmarl «CK. 131
3.3.2 The CIE 1964 Supplementary Standard Colorimetric Observer. 132
3.3.3 Development of the Two CIE Standard Observers ......... 133
3.3.4. sGlbiStandard Tuminants nein! 4 fevneeals .. 5 143
335. .ibStandarthSourcesit. diesen a ee. AGsb « . 146
Oo: Oma antial dOp Cue ance aClOlere yay ate eet. eatery , 155
3.3.7. Standard Illuminating and Viewing Conditions............ 155
3.3.8 Calculation of CIE Tristimulus Values and Chromaticity
SSDOLUIUGLCS Srna ch ate, eT Rd, Gt oeered rite AWer puree eee 156
3.3.9 CIE Uniform Color Spaces and Color-Difference Formulae... 164
3.3.10 CIE Metamerism Index for Change in Illuminant .......... 169
a3. lige LE OlOr- Renuering (nde kas, (yaks os aM ead oi Aanadre ee.. « le

3.4 Dominant Wavelength, Excitation Purity, and Colorimetric Purity .... 175

S.5 <Gomplementary. Color Stimuli tae nb oe a nn Fos os BR ey Dee 176

3.6 Maximum Attainable Luminous Efficiency of Color Stimuli of

Diferen Gnome yaa te oe ee ee ee Lay

b.7), “Optimal Obicct Golo sum lig os ot eel oe ke eS es 179

See WICLAMmeriO COlOG SUDO ets. tees Be ki Me te elon eek eee, 183
a baa | Delinitionron Mietamiensiniwe, £608. oo ee oe tage oop oedyenss 184
3.823 MethodsoiGenérating-Metamerss gio, i fae. on 2 > - 185 —
3.8.3. Intersections of Spectral Reflectance Curves of Metamers .... 194
Bo.Si4amaGounting Metamersags «5 ake eter cust) inctates © 8 3 oe ew 196
3.8.5 Boundaries of Mismatches of Metamers................. 200
3.8.6 Application of Linear Programming to Miscellaneous
Colorimetric Rroblem gt. ie eo hog iE co Ree: os Bho» Zi

3.92 Colorant Formulations. cre ae et ok netsh Rt oe ES su ees - 221

3.10. ‘Specification .of Color Tolerances «...... , ES ME.LI. Zan

3.11 Distribution Temperature, Color Temperature, and Correlated

Color Teiiperature ies Bie Feat ROT... PS. 224

3.12- Colonmeéeticinsiramentatione |: Wil-7 29 POPE.

6 Gees. 228
SD Te SDECUOLACIOMCICLS ate oe ea ae gees a cao: 2 229
On EOE CIIODMOLOMICLCE Ssrete erect ee ae TB nus, 0 teOa
3.12.3 Spectrophotometry of Fluorescent Materials ............. 230
3.12.4 Propagation of Random Spectrophotometric Errors ........ 240
B12 See CISUIMULUS Etter OLOTINCLELS im hs eee a os an a 243

CHAPTER 4 PHOTOMETRY

41 Basic Photometric.Quantitiesand Units ot ccsarSnatd’s

PRs + ee b's 249

Ae ne DNOtnetic OTT ples. nine. eas peidcaa Saloeastas ol: oss © « 249
xii CONTENTS

43. - The Standard ‘Photometric System™ =~ 72276 PU!

PA See © -
4.3.1 Historical Notes hte ee 2a eer eee EF
4.3.2 Standard Photometric Observers’... 35... STORMONT.
4.3.3 Phétometrie Methods “arSaiatiye ners Aga. aE
4.3.4 Measurement of Total Luminous Flux ..................
4.3.5 Measurement of Luminous Intensity and Illuminance .......
4.3.6 Measurementiof Luminance’ . 2706, Wb rais Se. Se Lt

4.4 Calculation of Illuminance Produced by Lambert Sources of

Differént'Shapes (UA 20 BASEN th 3G MOURNE... ehh |

CHAPTER 5 VISUAL EQUIVALENCE AND VISUAL MATCHING

5.1 Preamble? oy 6 ees eeeee

5.2". Classification of MatchingtPinced utes (78.u,9.0. erate eae ee

ask Visual Equivalence and Visual Match by Strict Substitution . . .
okra Asymmetric Comparison and Matching;
COUAST=SYMINe LIC WLATCIIINIR = fette tete ea eters eee ee ree
SP ie Limited Groups of Asymmetric Matching Procedures .......
5.2.4 Matching Crmtera 0 ain eat neni gied tenetos reer ora
Ses Some Particular Matching or Equivalence Procedures .......

53% Maxwells Method obG olor Matchine w= 1, cee eee ene

erePa Historical Notre. 5 Sc ce eer cae ae eee eee on eee
313.2 Basis of! Maxwellimg (rictiramacy i. 4 sssgs ne eng as doa eee
nyeRe Maxwell Trichromacy and Full Trichromacy in
Quasi-Symimetiic IMAtenin eS re ee a ieee eee
5.3.4 Maxwell’s Method in Nontrichromatic Matching ..........
S.3.0) Maxwellian Matching as Correlation of Stimulus Spaces .....
3.3.6 Maxwellian Matching in the Weaker Sense...............

5.4 Precision of Color Matching for Normal Trichromats .............

5.4.1 MacAdam Ellipses.."- .... oaamewihea P aches? tee aden ieee:
5.4.2 Brown—MacAdam EBlInscids 0. ce © ars) oe ee
5.4.3 Wyszecks;;Fielder, Ellipsoids. ic.) naemecearnes EYaneierineeit’ - «i
5.4.4 Repeatability of Color-Matching Ellipsoids for the
Mate CIDSERVET a oa ca, ches hse ena eae en ee a ee ee
5.4.5 Intercomparison of Color-Matching Ellipses for
Different Observers: o. 10" ogee oo en
5.4.6 Propagation of Random Errors in Colorimetric
J ransformaugos 7.097 ice ee rent to 9 een eer

5.5 Color-Matching Functions of Normal Trichromats...............

Seoul Two-Degree Data of Guild and Wright .................
ast Judd Modification: vies: 20h ewer.) gr ay ert nee ae
yote Stiles Two-Degree Pilot Data . 2a on. ee,
5.5.4 Stiles Borch)en-Degrce Lala... csc icat ate le Gee ee
BROS: ETL *Ten-DesreeDatar™.. . eee
5.9.0 Variations of Color-Matching Functions of Different
Normal ‘Trichromats( ="".-.1.°11, 22a oe ee ee
 Contents

5.6 Pactots Modifying, Color, Matchingensarsensst’

be: cater isl] «-4 *- 347
Ore peice tepsncitis Ii le IAVc nt)" renee, meee, See 347
Dice SOE at ICA ULOEL. ies ot eee Ey Sere eg meet 354
2.0 Came SOCAUOI ORY ASUAl ICI Ge ih. aries he err ats Wel bd eae « BY ps
3.0.4) Oe OL Visual Picid “fens 2 ct es fe eee, eee eee =
30.0. .Hig Cuouuance Level... <mealacgtonters Hnscatdt « «doe «- 374
5.6.6 Maxwell Method versus Maximum-Saturation Method ...... 379

».7 Luminous Efficiency Functions of Normal Trichromats ........... 392

5.7.1. Matching or Equivalence Criteria and Experimental Procedures 392
De EA DEDIMENal alas ete ee Ty sca py use rene shee eee 394

5.8 Heterochromatic Brightness Matching of Complex Stimuli ......... 410

5.8.1 | Luminance of Equally Bright Color Stimuli .............. 411
moose © Addiivitydalluresis te sere the Coanricngt seek? Dicneats: 413

5.9 Abney arid Bezold=BrickeiEffectss. VN era ipel Me tt wt tiene 420

5.10 Hue Reversals: Brindley PsochrOmes ea) ce Sn aes eee 424

5.11 Stles=Crawiordilffect¢? ysbyitiosn® Airdoaelk teeteredeeedh

+. 1 Aen 424

5.12 Chromatic Adaptationn sents pitt dn vasyivaat Minnie et. PRS 429
5.12.1 Asymmetric Matching—Basic Concepts ................ 429
ai2a7 Expenmental Procedures*and' Data (Pre re. a ee es 432
5.12.3 A Comparison of Chromatic-Adaptation Transforms ....... 449

5.13 Chromatic-Response Functions.d:ai4 tyhovienstD-qele}

ough) «bee. 451

5.14 Color-Matching Properties of Color-Defective Observers........... 458

5.14.1. Normal and Anomalous Trichromats................... 459
Pen ee eMC ATOIMNAIS eres ce ete wdes nels Win ae ian alee eee he 463
lest Nionoctromitoven: 35 245 44 se 8S te 2 RING MONE |. : 471

5.15 Instrumentation‘for Color Vision Research’ + 2%... Sit

Se See... 472
YP aaitePINAR WCHIALT VICW Stet tera tree Ema en ee ES bereer 5 478
5.15.2 Measurement of Directional Sensitivity and
focrement] hresholds «+ > <> > oghiatt nated? peed © ° 483
oe Gabe OUR orytherCP Vedale repea ital g Manners oa pone ae eRe oe 484

CHAPTER 6 UNIFORM COLOR SCALES

6.1 SVANEDCoe eae ik 5 PE Ge ak Sing ee SET gMPPRR Ie OXRR nt Ee Fa EL 486

6.2 PV DeScOliora Osun Scan eyIVIELNOUS 0 58 io sg. Kn toe ckaas = “uot ont-* 488

6.3 Pike otas ce ness CAS en Want eee ca ee oped ak cs 493

6.4 Color Scales of Constant Lightness -soansisii mmeoinid. L045... 500

6.5 Three- Dimensional Color Scales... <.. @30N9I UE Yin. COLT... 503
CONTENTS

65:1 @Printiples'of Gonstractignrr rer. wes ae tee 2) eta

6.5.2 Color-Difference Formolaéoe 24) .ncycyaath
eet «fee 2
6.5:3. “White. Whiteness Formulae: >: . - ysteasviniacull
duekt «+ foe >

66° .Color-Order Systemsig gas cence

Mee, ee ae
6/6.) -*Munsell’ Color Systema aren eee ee en. 5 eee
6:62. “DEINI Color Systeme enn ete ee 5 re
6.6.3 Swedish Natural Color System ........... dir ee ey PR
6.6.44 N'OSAKCOlOMSYSICHI 8 Tiere: eee nt ee ee ee, S|

CHAPTER 7 VISUAL THRESHOLDS

Ls =Preamble seine ie ec ye BU. co ee Ree ee ea cone

‘l2”* General ‘Concepts® $2 ae een ee ste SAC ee ee ee ee

7.2: +.Basic Terms and) Definitions. yum oa: bare a. oho te
7.2.2 Quantum Fluctuations and Visual Stimuli ...............

7.3. Dark Adaptation and Absolute Thresholds ....................

7.3.1 Recovery of Threshold Sensitivity; Dark-Adaptation Curves . .
7.3.2 Threshold Variation Over the Visual Field ...............
7.3.3 Threshold Sensitivity of Fully Dark-Adapted Eye ..........
7.3.4 Absolute Threshold Values for Different Conditions
of Measurement os 68 tessa ueshea ele Pas eatvenat foe oes

7.4 Chromatic Adaptation and Increment Thresholds................

7.4.1 Tsvo-Color:threshold Method ey 7 See 2 OU siiatt.7 | &
T:42. -Basic-Rormiulagg ie) Wh ob) eee ln eo Ea ae
7.4.3 9 Stiles Meat Datagas <)2 is eas SS Se eh?
7.4.4 Specific Aspects of 7 Mechanisms and Later Developments . . .

15: Rod Saturationix,: téeseine fay. areal Teele ee ee ese:

7.6. ‘Cone Saturation’). ocean See ie ne ee

1.1 * *Rod and Gone'Interactions’ .; ¢) =" 254. feet, Renter tere

718 Uniform Equivalent Pields# <5 41 eer ene ate ee

LS." . Basic Pormulicae 90. tie) ane eee, «ee ee

1.9: “Spatial and “Veaiporal e actors iy. nan eae eee

Tile \Distinetiess of Bordettareas Tecchenee 2... ee ee
EO2) =Mach Bands 23550. ee a eas Re, nD ees
Peeda ® EOMCKEL Wet sce aie yasseven 4 ic eter ainsi ee eae oe

710° Discrimimation_ Phresholds’.. a cs eee ee

710-1. \Euminance;Differences: seitial 2 pentiin ) Ie eeleee wot). . 4
710.2) Wavelength, Differenceswt.0 Se ceeeae 2 SH ity on eee
7.10.3 Purity Differences 50: 10. 50) ye fu. holt Rey Get Ge Ce ER Ce «(Oefe] ‘w Jewie, ‘e: eh ie fe) 6 “Ss Velle & 6 © &
 Contents XV

PAL ee OLGA OUIPerasuire Lyi ferences concn eer ey ae eg ae 574

Wd Oe LCOMIALICLEVE LyLISELENICOS ar, 7 (avn erm, Mea key bee: mip.
HALLO PRSOLOTLILLCTCDCE NIGtCHES <2) eemers aert ene Sie pen! s ini ae, or 576

CHAPTER 8 THEORIES AND MODELS OF COLOR VISION

Side LCAINIOCMMtttEe OEE Gere ee NR eee rn ees. 2 Passe MN re dca cas 582

8.2 Visual Response Functions and the Spectral Properties of

ASMA Es OICIIS Bre Oey ety eee ee een ae as. 2 agpee yesh twee te a 586
BZ Up eer NCI DIG OL OMY arlaliCe «a feiata atcy Sean eel eee eee 586
Oiled OCR ISUAl-Piginent LAVversemares., A. loaeranr a teituc. aac: mans 588
8.2.3. Dartnall’s Standard Shape of Visual Pigment-Absorption
CCOSTTICICT (pita peat ese teen oer cele hy trate mt) yout UY ante 591
8.2.4 Color-Matching Data and the Spectral Absorption Curves of
ViSUal PIs men tse ug ee eres eee Me tea nena Sones ce Ee ai 591
8.2.5 Dichromatism and the Fundamental Spectral Sensitivities .... 604
8.2.6 | Color-Matching Data and the Pigment-Bleaching Model ..... 619

B:) MENCULADIMIOCE Satter Ure oF: Aa ey iat Aw NEY eT Es 633

PeBUY COOSA TTANTSogUVa Sa
FG Ih. oR ok re ak Raed de hy elt 634
a NULLS Mae RS) IRS os ls mate ace OR anit ee en ace ee 639
ase memerA UEVIC OPAC FAINCSON De ryone 20 ho wear cao any ane eee 642
5 CTL WP eit ah OE 5 Cera keane eas ts ete cad, At aac 646
Reh SMM EYSUISOS Or ee eRe eet. GS scatvis Susie Caged yee RN a 652

SA eee ine eacMenits OL. COLOM SPACE ie.¢,her- nc he. ne tude cian a aera ie 654
BA mre SAS LONE CDS Weta ec. calys mee oes coke elnee meet PRO eee 654
BAL) memeLOLIMNONLZ een ee ele en egg hee eee ames Ra he tose ace ee 658
am OC OT OCISCRs te hie aa casi wd, eM eis, eh ea 659
Be SLLCOMEN eT ete) ne Pi odin 2,Jicae ae MeN. cnt, ee ey, 660
BAe DEOL ter ee ee co nscy EN ae eee e 672
S 4 OrmeOS ATION Wd ITAV EDN tn aan Cee cu wha Meee oe tates tee 673
8.4.7 General Construction of Inductive Line Elements .......... 677

APPENDIX OF EXTENDED TABLES AND ILLUSTRATIONS ........ 690

RE PERENGES cr to oe es eee Ra ae ata ome ahve 885

ESRC) RGU) Erect ere ere aos welighign hn 5 cnc Ss aca ears Lai We a ony mes ta 925
 nse 34

5 key!
;
2
2,BALA ad sworn easy
“ sy bed » . oe ,

AOE? ROI? 49, “T a

ver aonbe. eeLio} edswescba 42) aes

a.
\ J

:' Pe BIS
~
; .

\
. ayy =
+ ee .) yan
what?
/ Vel s 4
ecg?! se) bas enpiadyT binges lac” SH
a 3 ees a Fo eae foueiy © ;
fi 2 o

pct (ure ee ee re ea he a) eee UC) HORTA

wn | ; _

EG vs.T. PRA

vad eeene ar
pie
 A tw sa 5 7 ? : 7 ;
ai A a ' a Be § : y ae ‘s*
\ae BJ . ae

- COLOR SCIENCE |
a
a) as
a3 L A.

vw

eyea law 0
eA af
re all
Secu ~~ , tp 4a 4

eo. OF: Measireiient (>i - erm ; - won

J) dUnités)
a CA c. ’ ” ’ ' ve a ae

7 ROS
ify gryCa ID
' 4 ; eeye
ve 4 PS Porn «sr ¢ . —* te
=
psi il fin Sa} a ‘
4

ite 264) Mustrstes: ti cept tt Bolih ang | =

eo
a 7
as wht;ee
>
etert ta

a lg ts

im 00fGS Of cadian! cong. thot or the Gectes ire

; ro wa il ee, od
if be > tievest to the colori ‘ VANCIEO RA. FOU) Le
 ee
wae AoA a
ae
>
 CHAPTER 1
PHYSICAL DATA
1.1 BASIC RADIOMETRIC QUANTITIES
AND UNITS
A selection has been made of the radiometric
quantities that are regarded as basic and of most
general value. These quantities are displayed in
Table I(1.1) of the Appendix. The definitions
used here are substantially those adopted in the
International Lighting Vocabulary (third edition,
and an early draft of the fourth edition) of the
Commission Internationale de lEclairage (CIE,
1970). This vocabulary may also be referred to for
equivalent definitions in French, German, and
Russian, and the translations of the terms (without
the definitions) in five additional languages
(Spanish, Italian, Dutch, Polish, and Swedish).
The reader is also referred to the Self-Study
Manual on Optical Radiation Measurements pub-
lished by the U.S. National Bureau of Standards
(NBS, 1976 to 1979) which includes a rather
detailed and advanced treatise of the radiometric
concepts. Also recommended is the introductory
booklet by Bauer (1965) and the monograph by
Grum and Becherer (1979).
Table 1(1.1) is a supplement to Table I(1.1). It
contains the defining equations of the basic radio-
metric quantities and the corresponding SI units
of measurement (SI = Systéme Internationale
d’Unités). Figure 1(1.1) illustrates the explanatory
notes given in Table 1(1.1) and will be found
useful in interpreting the defining equations. Fig-
ure 2(1.1) illustrates the concept of solid angle
and its unit, the steradian.
1.2 SOURCES OF RADIANT ENERGY
The data on sources of radiant energy that are
considered of basic interest to the colorimetrist
and visual research worker are relative spectral
radiant power distributions. The spectral range
of these distributions is usually confined to the
visible. However, some distributions extend to
the ultraviolet and near infrared regions of the
spectrum to allow for applications to the colorim-
etry of fluorescent materials and to assess the
potential effects of nonvisible radiant power on
radiometric, photometric, and colorimetric para-
meters. For some sources, indications are given of
the radiant power emitted in terms of absolute
radiometric quantities, for example, irradiance (W
men):
The spectral distribution of a radiometric
quantity is commonly given in terms of that
quantity per unit wavelength interval. This prac-
tice is followed in this book. However, occasion-
ally it is advantageous to consider the radiometric
quantity per unit frequency interval, frequency
being in some respects the more fundamental
parameter of radiant energy. The conversion of a
spectral radiant power distribution based on
wavelength to one based on frequency is carried
out as follows:
By definition [see Table 1(1.1)] it is
ANZ=cr? [1(1.2)]
thus
dX = —cv “dv [2(1.2)]
Over a spectral region dA the radiant power may
be given by P,dA, which on the frequency basis
must be equal to —P,dv. The minus sign allows
for the decrease in frequency with an increase in
wavelength. From Egs. 1(1.2) and 2(1.2), it fol-
1]
 Pe
1 (TT) siseg sjoworpey sonnuend
pue syrup yuswejddng]
0} F148 [DI
Arojeueldxy
saj0Npue Iv[NULIO,
wa foquidg Surmyjaq uonenby 2as] “s8ryz (111pue [Dz sya)
JUeIPeY AZiaua ¢ (aqnof)
Seseeldeh
ul eg VE pp soo!
l2 2 so0¢
3 Wes
DHE
A

JULIPeY J0UL}IXS W3 7
W a> LVR 'pp — JoRJINs JUDUTITA
JO J0INOS 7_Ul-

doUvIPeII] oy —°"F 5 “pp— dORJINs JUoWIZ[a

JO -o1 7M
ie IaA1a0
jURIPeYy AjIsuaqutai ="] Te —!mp juauajajo pros g[sue MA ;—1S-
lop yim xade(,) je sovjins
d<P JO JOINOS
sourIpey i —*7 . '3 — aur usamjoquaat3 -samp 'pyp='3so00!pp Ayfeuosoyiio ;18-;_W-M
yp soo!
'2 oop uon (z)-({)
pue ewou
wv =! paysaford
uo sueyd -norpuadiad
a Jo yp se] 07 uaard uondamp(Z)-(1)
= 7 ¢3 — a[8ue UsamiaqUdAl3 -dalIp =—'3sootpp
pp Ajfeuogoyyio
soatyp
Ta tmp uon (Z)-({)
pure yeutsou
“uv= paysaford
uo sued -noipuadiad
)P (Ca jo “pp Ie] Oo} uaats uonsamp(Z)-(1)
= ate cmp— juaWwaya
Jo prjos g[sue aP
2 yim xode(7) ye aoejans ="(°7)p —.——
JO JOAIIN0I Ps
pros a[zue ” ” ¢ — uoniod
Jo a1oyds soejins aor Spon IS (UeIPeId}s)
aee I =poy se
uv — snipes
jo ‘arayds
osye -SIp as
sour} uaamjaq
(1) jo 'pp =‘mp rent
pure (Z) jo “yp 4
Aouanbal,y a a8
AA, pUSUITIAR XY == 9 = AyIOOTAA
JO JURIPeI ASIOUD wi
ul
ondva
AA plIQUINUDAR
wu =u a

uy, AT[enqIA
[ye smjawitojoo ‘suoneoydde
oy} TUN JOJ yIUaTIAeM
st ay) JoJaWIOUeU
|) WU = ,_O[ “(WE UT aJOW! UOISIA Ps}UIIO“YIOM Joyita
YIZUsZTIAeM
UT SUIID}
JO WU JO JOqUINUDAeM
Ul SuJa}
Jo Wd | ore ‘pasn
:210N
J] ay) [eNDNdS uoNesUs0U0N
Jo& omjaworpes Aynuenb
7y st ‘paraprsuos
11 st Aqfensn poyeusisap
Aq ay) sureu
Jo ay} Ayuenb pepsseid =
Aq ayy aanoalpe ‘;oujoads
pure Aq oy} sures joquiAs
Joy ay} Ajnuenb
WIM oy) yduosqns
y Jo)‘a Jo :(w =*’x “yp/?vp
Jog ‘ojdwrexa
Jt °x st
jUeIpel ‘1amodoy) jenoads uoMeUaouoo
Jo juerpes Jamod
st Ajdurts posojar
0} se [enoads juerpes Jomodpuest pajousp
AQ =“’d"dP ‘\P/
 ne

r
Fig. 1(1.1). Illustration of explanatory notes given in Table 1(1.1) concerning basic radiomet-
ric quantities and units.

S
w = —>— (steradian)
r

Sphere
(radius r)
Fig. 2(1.1). IMlustration of a solid angle w and its measurement in terms of the unit of solid
angle, the steradian (sr). The apex of the solid angle is located at C. The solid angle cuts off an
area S on the surface of a sphere centered at C and of radius r. The size of the solid angle «@ is
then given by the quotient of S over r”. In the case illustrated, «w is approximately equal to one
steradian. The concept of solid angle is not confined to right-circular cones of the kind depicted
in the illustration. Almost any shape of cone, generated by the straight lines emerging from the
apex to the points of a closed curve, can represent a solid angle. If the closed curve is a polygon
(e.g., a square), the cone and thus the solid angle takes on the shape of a pyramid.
4 PHYSICAL DATA

100 -20 p20

25

400 500 WAVELENGTH [rm 600 700

750 600 500 429x102

FREQUENCY [I/s]
Fig. 1(1.2). Number of photons N, emitted in unit time per unit frequency interval for an
equal-energy spectrum on the wavelength basis (P, = const = 1.0).

lows that scattered by the atmosphere are the components

of daylight that play a key role in color science.
l
Pee Pye atAUS [3(1.2)] Many spectroradiometric measurements of
daylight have been made and reported in the
If it is required to determine the number of literature. Many of these measurements were made
photons (quanta), N,, emitted in unit time per with applications in mind not directly connected
unit frequency interval, then, with the energy of a with color science. The spectral radiant power
single photon being equal to hv [see also Section distributions of daylight are basic data in astro-
1.2.2(1)], we have physics, materials research, architecture, photo-
biology, illuminating engineering, solar energy
N, = <2 [4(1.2)] engineering, space research, and other fields. In
the last two decades, considerable progress has
been made in the development of measurement
N= ma [5(1.2)] techniques that facilitate the collection of large
volumes of data and their analysis at ground-level
Figure 1(1.2) illustrates Eqs. 3(1.2) and 4(1.2) for altitudes in different locations of the earth as well
the hypothetical case of P, = const.= 1.0; that is as outside the earth’s atmosphere.
for an equal-energy spectrum on the wavelength There are several books available that discuss
basis. The values of the constants c and h are daylight in detail and from different points of
those given in the Section 1.2.2(ii) on Planck’s view. The reader may find the following of partic-
formula. ular interest: Coulson (1975), Henderson (1977),
and Meinel and Meinel (1977).
In specifying certain representative spectral
1.2.1 The Sun and Daylight
radiant power distributions of daylight, it is nec-
The most important natural source of radiant essary to state whether they refer to direct sunlight
energy is the sun. Direct sunlight and sunlight only, to sunlight plus scattered light from the
 Sources of Radiant Energy 5

atmosphere, or to scattered light only. The condi-
tions of the atmosphere and the position of the
sun are clearly important factors.
Table 1(1.2.1) contains spectral irradiance data
for the whole solar disk obtained by means of
different instruments installed in high-flying
aircraft and rockets. The data, illustrated in Fig-
ure 1(1.2.1), are representative of the solar spec-
tral irradiance outside the earth’s atmosphere.
They are referred to as the NASA (U.S. National
Aeronautical Space Agency) Standard and are
based on work reported by Thekaekara (1973).
They supersede the earlier and often quoted data
reported by Moon (1940) and Johnson (1954).
Table 1(1.2.1) also contains interim NASA
standard data of spectral irradiance for the solar
disk at the earth’s surface for air mass m = 2
(Brandhorst et al., 1975). These data were derived
by calculation from the NASA standard data for
the solar disk outside the earth’s atmosphere by
assuming an atmosphere with a water content of
20 mm and ozone of 3.4 mm. The air mass m is

Table 1(1.2.1) NASA Standards of Solar Irradiance*

Irradiance W-m?-nm"! Irradiance W-m ?-ym!

Outside at Earth’s Outside at Earth’s
Wavelength, Earth’s Surface for Wavelength, Earth’s Surface for
nm Atmosphere m=2 nm Atmosphere m=2

300 514 0 550 25 1120

10 689 1 60 1695 1111
20 830 50 70 HAZ 1134
30 1059 105 80 V715 1147
40 1074 177 90 1700 1148
350 1093 217 600 1666 I
60 1068 255 10 1635 1137
70 1181 320 20 1602 1137
80 1120 344 30 1570 1137
90 1098 371 40 1544 1141
400 1429 530 650 1511 1140
10 17S] 693 60 1486 1136
20 1747 737 70 1456 1127
30 1639 157 80 1427 1119
40 1810 867 90 1402 1114
450 2006 1022 700 1369 1102
60 2066 1089 10 1344 1089
70 2033 1107 20 1314 1071
80 2074 1167 30 1290 1058
90 1950 1133 40 1260 1040
500 1942 1165 750 zor 1026
10 1882 1147 60 1211 660°
20 1833 1136 70 1185 997
30 1842 1159 80 1159 981
40 1783 1140 90 1134 965
800 1109 950
10 1085 850?
20 1060 900°
30 1036 898

“Abridged data from Thekaekara (1973) and Brandhorst et al. (1975).

Interpolated from original table by Brandhorst et al. (1975).
6 PHYSICAL DATA

2200

2000

1500
outside atmosphere
1600

1400

1200

1000
earth's surface (m=2)

(W-m-*-um-—!)
IRRADIANCE
400

200

300 400 500 600 700 800

WAVELENGTH A(nm)
Fig. 1(1.2.1). NASA standard data of spectral irradiance (W-m ?-pm ') for the solar disk
measured outside the atmosphere (solid dots) and at the earth’s surface at air mass 2 (open
circles). Data points are those given in Table 1(1.2.1). Neighboring data points have been
connected by straight lines for ilustrative purposes only.

defined as the ratio: length of path within the
atmosphere for rays from the sun to the observa-
tion point on the earth, to the corresponding
length of path for a hypothetical position of the
sun at the zenith. On the basis of this definition,
air mass 2 means a sun altitude of 30° (measured
from the horizon).
In a more recent document (NASA, 1977), a
somewhat revised spectral distribution of ter-
restrial irradiance is given that is considered
more appropriate distribution with regard to ter-
restrial photovoltaic measurement procedures. The
revised data refer to an air mass of m = 1.5 and
are calculated from extraterrestrial data of the
sun disk obtained by Labs and Neckel (1968).
The CIE (Commission Internationale de
YEclairage) has issued two recommendations con-
cerning daylight:
(1) Recommendations for the integrated irradi-
ance and spectral distribution of simulated solar
radiation for testing purposes (CIE, 1972).
(ii) Recommendations on standard illuminants
for colorimetry (CIE, 1964, 1968, 1971).
CIE recommendations (i) specify a spectral
distribution of solar irradiance outside the atmo-
sphere and a spectral distribution of global radia-
tion (direct sunlight plus skylight) at the earth’s |
surface and for air mass m = 1 (90° altitude). The
spectral distribution outside the atmosphere is
a identical to that of the NASA standard discussed
earlier and given in Table 1(1.2.1) and Figure
1(1.2.1). The spectral distribution at the earth’s
surface is based on calculations made by Schulze
(1970) for rather broad bands of wavelengths. For
details, the interested reader is referred to CIE
(1972).
CIE recommendations (ii) specify relative
spectral radiant power distributions of typical
phases of daylight at the earth’s surface for corre-
lated color temperatures ranging from 4,000 K to
25,000 K. For a given correlated color tempera-
ture T. within this range, or equivalently, for a
 : fe) Rea 3500K
5500 K + 0°
wis
g

O
4000K

bis

.30

ery, i3333kK— \
+ 1 ty 2

e
20000 K
ico
+ V,

~ o£tsf 40000 K

OC mK
+

hes xo, siere) 40

Fig. 2(1.2.1). Portion of CIE 1931 (x, y)-chromaticity diagram showing chromaticity points
of daylight in comparison with locus of chromaticity points implied by Planck’s radiation law
(large open circles connected by solid line). The temperature of the blackbody (Planckian)
radiator corresponding to the points at the centers of these open circles are indicated in kelvin
(K). The straight lines intersecting the Planckian locus at 4,800 K, 5,500 K, 6,500 K, 7,500 K,
and 10,000 K are examples of isotemperature lines defining loci of constant correlated color
temperature. The chromaticity points computed from the 622 measured relative spectral power
distributions of daylight are identified by solid dots for the Budde (1963) data, by small open
circles for the Condit and Grum (1964) data, and by crosses for the Henderson and Hodgkiss
(1963) data. These 622 chromaticity points suggest that the different phases of daylight can be
represented by a “daylight’locus” in the chromaticity diagram defined by the equation
Vp = —3.000x}, +2.870x —0.275 shown as the dashed curve connecting small open squares.
This curve compares well with a curve obtained by Nayatani and Wyszecki (1963), shown as a
solid line, and with a curve obtained by Chamberlin et al. (1963), shown as a dashed curve
located close to the Planckian locus from about 6,500 K to ooK. (From Judd, MacAdam, and
Wyszecki, 1964).
 8 PHYSICAL DATA

given chromaticity of a particular phase of
daylight, the relative spectral radiant power distri-
bution can be calculated for wavelengths from
300 to 830 nm at 10-nm intervals. The method of
calculation is based on work reported by Judd,
MacAdam, and Wyszecki (1964) who subjected
622 spectral radiant power distributions measured
by Budde (1963), Henderson and Hodgkiss (1963),
and Condit and Grum (1964), to an eigenvector
analysis. Details of the CIE method of calculation
are given in Section 3.3.4.
It was found by Judd et al. (1964) that the
chromaticity coordinates x, y computed from
the measured spectral radiant power distribu-
tions agreed well with those determined visually
by Nayatani and Wyszecki (1963) and by Cham-
berlin, Lawrence, and Belbin (1963), and that
both sets of chromaticities cluster about the curve
Vp = —3.000x;, + 2.810x, — 0.295
[1(1.2.1)]
as shown in Figure 2(1.2.1). This curve defines the
“daylight locus” in the CIE 1931 x, y-chromatic-
ity diagram and runs slightly above and ap-
proximately parallel to the “Planckian locus”;
that is, the curve generated by the chromaticities
of Planckian (blackbody) radiators operating at

Table 2(1.2.1) Relative Spectral Radiant Power Distributions of Daylight of Different Correlated Color
Temperatures Calculated in Accordance with CIE Method

A(mm) 5,000K 5,500K 6,000K 6,500K 7,000K 7,500K 8,000K 10,000K 15,000K 20,000 K

300 0.02 0.02 0.03 0.03 0.04 0.04 0.05 0.06 0.08 0.09
310 2.05 E07 De 3.29 4.17 5.13 6.10 9.77 16.27 20.03
320 tot 11.20 Lee 20.21 25.01 29.78 34.38 50.58 AP 9212
330 14.73 20.61 28.36 37.00 45.94 54.87 63.54 94.28 145.16 173a%6
340 17.93 23.85 31.48 39.90 48.57 57.20 65.57 95°13 143.92 171.30
350 20.99 27.78 36.01 44.86 53.83 62.68 71.20 101.04 149.80 L702
360 23.92 30.58 SHR 46.59 54.84 62.93 70.68 97.65 141.40 165.74
370 26.94 34.27 42.89 52.04 61.23 70.25 78.91 109.06 158.06 185.33
380 24.46 32.54 41.19 49.93 58.45 66.65 74.41 100.90 142.96 166.09
390 29.84 38.04 46.41 54.60 62.46 69.91 76.90 100.45 137.24 157.28
400 49.26 60.89 [pate 82.70 92.61 101.87. = 110.43 138.72 181.86 205.04
410 56.46 68.49 80.23 91.42 101.96 LERS3ee 412099 151.41 198.08 223.24
420 59.98 Treo Pa Se FSO 03.38 112.74 =121.42 150.18 194.21 292
430 ad lk, 67.86 TOL 86.63 ports 103.04 110.35 134.44 171.07 190.72
440 74.78 85.55 OST A Oe ee 1332 12st 128132 151.66 186.55 205.07
450 87.20 D194 2 ¢107.850, 116,96 bd 30 132.702 Gals 2O 162.67 196.45 214.34
460 90.5.7 4 100.4 esl 09.47 oe 127 775.0 325,35 13223... 2 138.66 159.22 189.74 205.88
470 O1e33 99. B71 tlOE69.5 114.824 121.33 |P94, ae SS 2a | 150.23 176.16 189.84
480 DO Tite tOD.10 sod) 09.0200 Le 89a 21 58 126.76 131.48 146.62 168.89 180.58
490 91.93 98.05, 10103.65. 0108278. 13.46 Viv.1 cel? 1207 134.32 [RB IK 162.97
500 DSTO TONGS IOS TOF EID Saar SN | MG 7 ior 129.90 144.96 152,91
510 96°59 100.67" 2910440 “1LO7i787 8110187 PIS68S 16225 124.51 136.71 143.14
320 OTN OO GTOMIN MST es TOF SO OCs 108.65 110.30 113}57 123.18 127.14
530 102009 10420" 4106.05 10T637> “109213 110.44 =-:111.61 115.29 120.54 123.25
540 LlOOHS. “10290 1033 lereelOsAte 210539 106.28 107.10 109.70 113.50 115.48
550 102.31 102.96 103.54 104.04 104.49 104.90 105.27 106.42 108.06 108.92
560 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00 100.00
570 97.74 122 96.75 96.34 95.96 Jo.02 aoe 94.32 92.88 92.123
20 8 96.92 PIPE ie 96.72 oye 94.96 94.22 93.54 91.41 88.32 86.72
590 93.51 91.44 89.89 88.69 S719 87.00 86.39 84.82 83.26 82.68
 Sources of Radiant Energy 9

Table 2(1.2.1)— Continued

A(nm) 5,000K 5,500K 6,000K 6,500K 7,000K 7,500K 8,000K 10,000K 15,000K 20,000 K

600 Ft) 94.44 eAlee) 90.02 88.48 87.24 86.21 83.50 80.65 HOS,
610 99.29 95.16 92.03 89.61 87.70 86.15 84.89 81.57 78.15 76.83
620 99.07 94.24 ey 87.71 85.44 83.59 82.08 78.04 So) 72.07
630 ee 90.47 86.45 83.30 80.80 78.76 77.08 T29 67.79 65.85
640 98.89 92.36 87.47 83.72 80.79 78.44 76.55 PteiG 67.00 65.32
650 95.71 88.89 83.85 80.05 47 A2 74.82 Pe 68.48 64.55 63.37
660 98.23 90.36 84.57 80.24 76.92 74.34 12:3) 67.40 63.39 62.33
670 103.05 93:99 87.32 82.30 78.45 75.44 73.06 67.25 62333 60.96
680 99.18 90.00 83.30 78.31 74.52 71.60 O91 63.90 ae ie) 58.88
690 87.42 gOSh 1 74.03 69.74 66.45 63.87 61.82 56.81 By aa| 51.26
700 91.65 82.88 76.45 71.63 67.96 65.10 62.84 57.42 53.04 as
710 O23 84.88 78.91 74.37 70.86 68.09 65.87 60.33 55. 53.69
720 76.89 70.27 65.35 61.62 58.74 56.46 54.64 50.09 45.98 44.66
730 86.55 79.34 73.98 69.91 66.75 64.26 62.26 aiebie Dod | 51.23
740 92.62 85.03 Wo2 got th mies eee 67.07 61.82 57.03 55.48
750 78.27 Tales 67.19 63.61 60.84 58.64 56.89 52.50 48.51 47.22
760 Dhsh2 52.82 49.18 46.43 44.31 42.63 41.29 37.96 34.98 34.04
770 82.96 75.96 70.77 66.83 63.78 61.37 59.44 54.64 50.30 48.91
780 WGvou 71.85 67.05 63.40 60.58 58.34 56.55 52.06 47.97 46.63
790 a9D9 LIF. 68.05 64.32 61.44 59.16 57.34 52579 48.68 47.36
800 73.44 67.38 62.88 59.47 56.83 54.74 53.08 48.91 45.15 43.93
810 63.95 58.75 54.90 Eo | 49.71 47.93 46.50 42.95 Se Alp) 38.73
820 70.81 65.02 60.72 57.46 54.93 PRC otFee 47.32 43.68 42.49
830 74.48 68.34 63.78 60.33 57.66 32.0) 53.87 49.68 45.90 44.70

different absolute temperatures 7, measured in
kelvin (K).
It was further found that the 622 measured
spectral radiant power distributions could be re-
constituted remarkably well by appropriate linear
combinations of the calculated mean spectral dis-
tribution {S)(A) dA}, and the spectral distribu-
tions {$,(A) dA} and {S,(A) dA} of the two most
important eigenvectors of the set of 622 distribu-
tions.
Table 2(1.2.1) contains a set of relative spec-
tral radiant power distributions {S(A) dA} de-
rived by means of the CIE method given in detail
in Section 3.3.4. Figure 3(1.2.1) illustrates this set
of functions. Note that the relative spectral radiant
power distributions so derived have the common
value 100 at A = 560 nm.
The CIE relative spectral radiant power distri-
butions are defined at 10-nm intervals, and each
value given in Table 2(1.2.1) is considered to
represent the average radiant power measured in
a band of wavelengths 10 nm in width centered at
the specified wavelength. The extremely fine
structure of the daylight spectrum is thus lost as
may be seen by comparing Figure 3(1.2.1) with
Figure 4(1.2.1). In Figure 4(1.2.1), two spectral
radiant power distributions are shown in the form
in which they are obtained with a monochromator
whose slits are opened to give spectral band-
widths ranging between 0.25 nm in the ultraviolet
and 2.0 nm in the red (Budde, 1963). However,
for the colorimetrist, the very fine structure ex-
hibited here is of academic interest rather than
practical importance.
The CIE method of calculating daylight il-
luminants makes use of the remarkably close
correspondence between the chromaticity (or cor-
related color temperature) of a given phase of
daylight and its relative spectral radiant power
distribution. Once the chromaticity of a given
daylight irradiating a test surface is known, a
unique relative spectral radiant power distribu-
tion corresponds to that chromaticity. There do
not seem to occur in nature any other (metameric)
 20000K

SOOOK

POWER
RADIANT
RELATIVE

=i
300 400 500 600 700 800
WAVELENGTH A (nm)
Fig. 3(1.2.1). Relative radiant power distributions of 10 different phases of daylight obtained
in accordance with CIE method of calculating daylight illuminants for colorimetry.

Se ie ell a rh lah

200

I5O

100
RELATIVE
POWER
RADIANT

50
: Ha

300 400 500 600 5 i7GG

WAVELENGTH X(nm)

_. Fig. 4(1.2.1). Relative spectral power distributions of two phases of daylight, measured at
spectral bandwidths ranging from 0.25 nm in the ultraviolet to 2.0 nm in the red. Some of the
major Fraunhofer lines are indicated by their conventional symbols (Budde, 1963).

spectral radiant power distributions of daylight of
identical chromaticity, at least over the visible
range of the spectral distribution from about 400
to 700 nm.
Whereas the chromaticity of daylight fixes
rather uniquely the relative spectral radiant power
distribution of that daylight, the sky conditions
offer only rather vague clues as to the spectral
distribution that may prevail. However, as a rule
of thumb, the following relationships seem to
exist:
(a) Daylight from the sun plus the total sky
incident on a horizontal surface ranges in corre-
lated color temperature from about 5000 K to
7000 K regardless of cloud coverage, (1.e., clear to
overcast sky). A good representative spectral
power distribution for this case of “global radia-
tion” is the one shown in Figure 3(1.2.1) for 6000
K.
(b) Daylight from the north sky or, equivalently,
daylight from the total sky but with the sun
occluded, has correlated color temperatures above
7000 K. The amount of cloud coverage de-
termines the correlated color temperature most
applicable. For a clear north sky, very high corre-
lated color temperatures of 40,000 K and higher
can be reached. The overcast sky falls into cate-
gory (a).
(c) Daylight from the sun disk alone or daylight
from the sun disk at low altitudes plus sky gener-
ally results in spectral power distributions corre-
sponding to correlated color temperatures below
5000 K.
In the ultraviolet (300 to 400 nm) and infrared
(above 700 nm) regions of the spectrum, seasonal
and time-of-day variations are known to occur.
Thus, in these spectral regions, the CIE distribu-
tions may, at times, deviate significantly from the
actual distribution occurring.
Studies made since 1964, when the CIE recom-
mendation was formulated on the basis of the
work reported by Judd et al. (1964), have, on the
whole, confirmed the CIE spectral distributions.
However, some deviations have been noted and
have been attributed to the particular location at
which the measurements were taken, as well as to
seasonal parameters (Das and Sastri, 1965; Winch
et al., 1966; Sastri and Das, 1966, 1968; Nayatani
et al., 1967; Tarrant, 1968; Boshoff and Kok,
1969; Kok, 1972; Tarrant and Brock, 1975;
Dixon, 1978).
Sources of Radiant Energy 1]
1.2.2 Thermal Radiators
Thermal radiators are hot bodies. Their spectral
power distributions are continuous functions of
wavelength and depend both on the nature of the
body and its temperature. However, a thermal
radiator may be imagined capable of providing a
spectral power distribution that depends on the
temperature alone. Such a radiator is called a full
radiator, Planckian radiator, blackbody radiator,
or cavity radiator. In the following discussion, the
term blackbody radiator is used. Because of its
fundamental importance in radiometry and, con-
sequently, in colorimetry and photometry, some
of the properties of blackbody radiators are given
here.
(i) Blackbody Radiator. The radiant energy in
an enclosed space or cavity, the walls of which are
maintained at a uniform temperature 7, is depen-
dent both in regard to density and spectral distri-
bution on the wall temperature only, irrespective
of the wall material. The introduction of a small
viewing aperture in the enclosure wall modifies
slightly the radiant energy in a way that depends
on the size and shape of the aperture and cavity
and on the emissivity of the cavity walls. This
small modification aside, the radiant power
emitted through the viewing aperture is depen-
dent solely on the wall temperature, and the
aperture itself constitutes the ideal blackbody
radiator. Any surface that emits radiant energy
identical in all respects with that from a small
aperture in a constant temperature enclosure is
also described as a blackbody radiator and is
characterized similarly by one parameter, the
temperature T.
In the above discussion and in the formulae
that follow, the interior of the cavity and the
ambient medium of emitting surfaces are assumed
to be empty space.
(ii) Planck’s Formula. The spectral density
(u,,) of radiant energy per unit wavelength inter-
val in the constant temperature enclosure of the
blackbody is given by Planck’s formula:
uo, = Saher
(ee ** — 1) ‘(J Site)
[1(1.2.2)]
in which c, h, and k are constants and T is the
temperature taken on the absolute scale in kelvin
(kelvin = degrees Celsius + 273.15). The cur-
12 PHYSICAL DATA

rently recommended values of c,h, k are as fol- radiance of a blackbody radiator is independent
lows: of the direction of emission for all wavelengths;
that is, Lambert’s law is obeyed.
C= 2.907 99458 x 10" ms Equations 2(1.2.2) and 3(1.2.2) apply to non-
(velocity of light) polarized radiant energy. For polarized radiant
energy, the first radiation constant c, must be
=6626 116 10° es multiplied by the factor 3.
(Planck constant) The radiant exitance M,, has a maximum value
M,,,, at a wavelength X,,. The product 4,,7 is
k= 1380662 10°70 JR constant and given by
(Boltzmann constant)
= ne
ar: S aGhS WIE al 2.897791 X 10° —3 (m- K)
For all the constants the values adopted as best m

values are those specified by CODATA [Commit-

tee on Data for Science and Technology of the
International Council of Scientific Unions
(ICSU)] as compiled by Cohen and Taylor (1973).
The spectral radiant exitance (M,,) of a
blackbody at temperature T per unit wavelength
interval is given by
[4(1.2.2)]
that is, the maximum of the spectral distribution
curve occurs at A,,, which varies inversely with T
(Wien’s displacement law). The value at the maxi-
mum is given by

M, = [uy = cA $(e2/7 — 1)! (Wm) M.y,T° = 1.286673 X 10°° (W-m-3- K~°)

[2(1.2.2)] [5(1.2.2)]

which, written in terms of the first and second Theoretically, the spectraldistribution of
radiation constants c, and c,, is perhaps the most radiant exitance extends from A = 0 to A= 0,
familiar form of Planck’s formula. but M,, soon becomes very small on either side of
The two radiation constants c, and c, are its maximum at A,,. One percent of its maximum
obtained from value M,, is reached at approximately 0.33 4X,
and 6.5 A,,, independent of the value of T.
C= qorhe = 2ahc? Figure 1(1.2.2) shows three examples of spec-
tral distributions of radiant exitance for tempera-
tures T = 2500, 3000, and 3500 K and their cor-
= 3.741 832 X 107'6 (W - m*)
responding maxima.
The total radiant exitance (M,) is given by the
and integral

ues
; = 1.438786 x 10-* (m- K)
3°
M,={ M,dd=T*(W-m?) [6(1.2.2)]
0)

In pyrometry, colorimetry, and photometry, the

rounded value a relation often referred to as the Stefan—
Boltzmann law, which states that the total radiant
exitance of a blackbody radiator is proportional
Cy = 1.4388 X 1072 (m- K)
to the fourth power of the absolute temperature.
is currently recommended. The proportionality factor (Stefan—Boltzmann
The spectral radiance (L,,) 1s given by constant) is given by

51.4
Len = a pe (Wiemg sr [32 oa A Knes seri Ome Weemine® Kin
TW 15h3c*

where w, is the solid angle of magnitude one The radiant exitance from A = 0 toA = X,, is
steradian. This relationship results because the very closely equal to one quarter of the total
 Sources of Radiant Energy 13

6000

5000

=
4000

3000

2000

d(W-mr2-nm-!)
SPECTRAL
EXCITANCE
RADIANT
Me

1000

1000 1200 1400

WAVELENGTH A (nm)

Fig. 1(1.2.2). Spectral distributions of radiant exitance of a blackbody radiator at three

different temperatures, T = 2500 K, 3000 K, and 3500 K. At these temperatures, the blackbody
radiator has a maximum radiant exitance M,) of approximately 1260, 3130, and 6760
W-m ?-nm "|, respectively, at wavelengths A, equal to 1160, 965, and 830 nm (approx.).

radiant exitance; that is, approximately, The radiant exitance per unit frequency interval
is then given by
fXm, dd = 0.25 iPMSUR
c
[7Gie22)] ov jUey == 2m¢ Z
hv 3

The approximate form of Planck’s formula for

x (eT 1) (T- m=?) [9(1.2.2)]
M.»

M!, = c,A~Se°2/™ (W - m7?) where pv is the frequency and

[8(1.2.2)]
Mae *h= AG32351X108-o (sm +)
which is valid when e°?/7* is large compared with
unity, is the earlier Wien’s formula. te
rs 4.799275 X 10 -11°' (s- K)
Planck’s formula is also often expressed with
respect to a uniform scale of frequencies rather
than wavelengths. The quantity M,, has a maximum value M,, at
 14 PHYSICAL DATA

frequency v,,, where (iii) Tabulation of Planck’s Formula. For pho-

tometric and colorimetric purposes, it is generally
py
To) = SB 78887 x 16 se KD preferable to use a table that gives the relative
spectral distribution of radiant exitance of a
M.,,,T7= 5.956646 107"? blackbody at various temperatures, expressed on
a wavelength basis, as a function of wavelength.
(stem pK Table I(1.2.2) has been prepared for this book
with the following specification:
It may be noted that the wavelength c/v,, does
1. Spectral radiant exitance given by Planck’s
not coincide with the wavelength »y,, obtained
formula (see Eq. 2(1.2.2)):
previously (see Wien’s displacement law). In fact,

= = 1959781 A, MA =cA "(ear = [ae

m

with
The number of photons emitted in unit time per
unit frequency interval and in all directions in the Cp O41 8 10 Wee
solid angle 27 by unit area of a blackbody at
temperature T is given by C5 = 14388 oc le Wea

[10(1.2.2)] 2. The wavelength A ranges from 300 to 830 nm

at intervals of 10 nm
3. The temperature T ranges
The maximum value N,, occurs at frequency »,,
where
from 1000 to 2000 K at intervals of 200 K,

fe AR ahs
AUin Pir, iLU K yin, i 9,
from 2000 to 3400 K at intervals of 50 K,

NL 19636270 TO” Gass «KR from 3400 K to 8000 K at intervals of 200 K

The wavelength c/v, does not coincide with either 4. Additional temperatures
v,, OF C/V,,. In fact,
T = 9000, 10,000, and infinity K,
+ a ERSGL2 me:
Py T = 2042, 2790, and 2856 K

The total number of photons of all frequencies
is given by the integral
N Sib ve dv(m-?-s-') [11(1.2.2)]
where
NT? = 4.808228 (F)e>
= 12520434. 10? (me? he- "> K-*)
For additional information on the blackbody
radiator and Planck’s formula see papers by
Forsythe (1920), Holladay (1928), Ribaud (1931),
Benford (1939, 1943), Worthing (1939), DeVos
(1954), Nakaji (1957), and Kaye and Laby (1973).
The computed values of M,,, given to five
significant figures followed by the power of 10,
are normalized at each temperature so that M,, at
A = 560 nm becomes unity. The absolute value at
A = 560 nm is given at the bottom of each col-
umn; the unit is watts per square centimeter per
10-nm wavelength interval [W - m 7 - (10 nm) ‘J.
The absolute values at the other wavelengths may
be obtained by multiplying the quoted (relative)
values by the absolute value for \ = 560 nm. If a
change in the value of c, to c5, = ac, is required,
the tabulation may be corrected by changing the
temperature T to T’ = aT. A change in the value
of c, to c; = bc, may be corrected by changing
the absolute value of M,, to M;, = bM,,. Figure
2(1.2.2) illustrates some of the tabulated relative
spectral distributions of radiant exitance of the
blackbody at different temperatures.

(Me)
RELATIVE
EXITANCE
RADIANT
400 500
WAVELENGTH
Fig. 2(1.2.2). Relative spectral distributions of radiant exitance of a blackbody radiator at
different temperatures 7(K). All spectral distributions have been normalized such that the
radiant exitance M,, is unity at A = 560 nm.
Many other tabulations of Planck’s formula
can be found in the literature. Recent ones are
those by Czerny and Walther (1961), Pivovonsky
and Nagel (1961), Hahn et al. (1964), and Levi
(1974).
(iv) Nonfull Radiators. The radiant energy
emitted by an incandescent solid is always less
than that emitted by a blackbody radiator (full
radiator) at the same temperature. The ratio of
the radiant exitance of a thermal radiator to that
of a blackbody radiator at the same temperature
is called the emissivity (e) of the thermal radiator.
In general, the emissivity is a function of wave-
length, and the thermal radiator is then said to be
a selective radiator. Numerically, the spectral
emissivity e(A) of the matte surface of a thermal
radiator is the same as the spectral absorbance
a(A) of the surface. Kirchhoff’s law states that
the ratio of the radiant power emitted by the
matte surface of a thermal radiator to the ab-
sorbance of the surface is the same for all thermal
radiators at the same temperature. If such a
surface receives the radiant power P), the portion
a(A)P, is absorbed-and converted into heat and
the rest p(A)P, is reflected diffusely. The quan-
tity p(A) is called the reflectance and is related
Sources of Radiant Energy 15
600 700
A(nm)
with a(A) [= e(A)] as follows:
p(A)=1—a(A) [12(1.2.2)]
For a blackbody radiator we have p(A) = 0 and
a(A) = e(A) = | at all wavelengths. For a perfect
reflecting diffuser we have p(A) = 1 and a(A) =
e(A) = 0 at all wavelengths.
(v) Tungsten. In the visible range, the spectral
radiant exitance distributions of some selective
bodies such as carbon, platinum, and tungsten are
approximations to the spectral radiant exitance
distribution of a blackbody. Tungsten is the most
preferred material for making filaments of in-
candescent lamps. It has a high melting point,
above 3650 K, and can be operated at high tem-
perature either in vacuo or in an inert gas for long
periods before loss by evaporation brings about
failure of the filament. Tungsten filaments can be
made with high mechanical strength to produce
favorable vapor pressure and resistance character-
istics.
The emissivity of tungsten is not quite con-
stant but decreases with increase in wavelength
and increase in temperature (Ornstein, 1936;
 ee ee ee ee ee eee ee eee ee ee
=
00r'°0 vor'o 607 0 vivo 0cr'0 Scr 0 lev'0 TIVO ely0 9170 6170 cctv 0 008
por'0 800 €Ivo 8Iv0 vcr'0 6cr 0 pero CIV 0 SIV0 8Ir'0 Icv0 vcr'0 O8Z
80¢'0 clv 0 LIv0 Icv0 Lev 0 ctv 0 Lev 0 eIv'0 9IP'0 Ocr'0 tcv 0 Lev 0 O09L
cIvV'0 9IP 0 0cr'0 Scv'0 Ofr'0 sev 0 6fv 0 SIv0 617 0 ccv 0 9cr'0 Ofr 0 Obl
9Ir'0 O0cr'0 vcr 0 6cv 0 eer 0 8tr 0 chr 0 LIV 0 Icv0 Scv 0 6cr 0 vero OCcL
617 0 €cv'0 Lev 0 cer 0 9¢r'0 Orr'0 vrr 0 61 0 vcr 0 8cr 0 ecr'0 Lev 0 OOL
cer 0 9cv'0 O£r'0 pero stv 0 cry 0 9rr 0 Icv0 9cr 0 Ofr 0 ser 0 Orr 0 089
9cr'0 6cr 0 ctv 0 Lev 0 Ivrv0 Spr 0 8rr 0 pcr 0 8cr 0 ctv 0 9€7 0 Ivy 0 099
8cv 0 cero 9€r7'0 6£r 0 trv 0 Lvr'0 Osr'0 9cr'0 Ofr 0 per 0 stv 0 cry 0 Ov9
Tero sev 0 str 0 chr 0 Spr 0 6rr 0 csr 0 Ofr 0 €tv 0 Lev 0 Ivy 0 srr 0 0c9
rer 0 8tr 0 Ipro Srv 0 8rr 0 csv 0 ssv0 pero Lev 0 Orr 0 very 0 Lrv 0 009
8tr'0 Ipv0 S¢r'0 8rr 0 cSv'0 ssVv'0 6Sr 0 Ler'o Orr 0 ery 0 Lvrv0 Osr'0 08S
chy 0 srr 0 8rr 0 Isv'0 csr0 67 0 cov 0 Ipv0 try 0 9br 0 6rr 0 cSPr'0 09S
perro 8rr'0 Isv0 psv 0 8S7'0 19¢'0 69v'0 err 0 9br 0 8rr 0 ls7 0. we USVe Ovs
9br'0 6¢r'0 esp 0 9S7'0 0970 £970 L9V'0 9rr 0 8rr 0 OSr'0 esv'0 ssv0 OcS
8rr'0 Isp0 ssv0 8sr0 c9r'0 Sor 0 697 0 6rr 0 Isv0 esv 0 SsVv0 LSv 0 00S
OSr'0 esv0 LSv'0 09¢'0 y9r 0 L9V'0 ILv'0 csv 0 psy 0 9S7 0 LSv 0 6S7 0 O87
csv 0 ssv 0 6S7'0 c9P'0 9970 69r 0 tLv 0 9S7'0 LSv0 6S 0 0970 c9v'0 09P
ssv'0 8Sv'0 c9P'0 c9or'0 890 clr 0 9Lv 0 6S7'0 19¢'0 cor 0 £97 0 S9v'0 OvY
8Sr'0 19v'0 S9r'0 L9V'0 ILV'0 SLv'0 8Lv 0 €9r'0 y9r0 9970 Lov'0 69h 0 OCP
19¢'0 p9v'0 L9V'0 ILv'0 ply 0 LLV'0 1870 9970 8970 6970 ILvV 0 ely 0 O00V
cov 0 s9r'0 8970 cLV'0 SLvV 0 8Lr 0 c8Pr 0 L9V'0 690 ILv'0 Lv 0 SLv0 06¢
c9or'0 9970 69¢ 0 cLv'0 9LV 0 6Lv'0 c8Pr 0 69r'0 ILv'0 eLv0 SLv'0 LLV'0 O8€
€9r'0 99r'0 OLY'0 eLv0 9Lr 0 6Lv 0 C80 OLr'0 CLV 0 ply 0 9Lv'0 6Lr'0 OLE
€9r'0 9970 6970 cLY'0 SLv'0 8Ly 0 I8v'0 ILv0 eLyv 0 SLv'0 8Lv 0 0870 09€
190 y9r 0 L9v'0 OLtv'0 €Lv0 9LV'0 6Lv 0 cLY'0 ply 0 9LV'0 6Lv'0 I8r'0 OSE
6Sv'0 970 9970 6970 cLV'0 SLV0 8Lb'0 cLyV'0 Ply 0 LLV'0 6LV 0 c8P 0 Ove
8Sb'0 c9oV'0 S9r'0 8970 cLv 0 SLv'0 8Lv'0 cLvV 0 plro LLv'0 08r'0 c8P 0 Oct
LSv'0 19P'0 g9r'0 897 0 cLv 0 9LV'0 6Lv 0 ILv'0 €Lv 0 9LV'0 6Lv'0 c8P 0 OcE
9S7'0 19¢'0 9970 OLv 0 ely 0 LLvV'0 180 897 0 ILv'0 pLrv'o 9Lv'0 6LV 0 Ole
9S7°0 1970 $970 OLYv'0 ply 0 8Lv'0 c8r 0 O0£
0087 009¢ O0VC 00? 000¢ 0081 0091 O0rC 00¢¢ 00072 0081 0091 (wu)
(x) einyeraduia 5 (1) ainjesroduia | v
(pS61)SOA oq Aq s}usweinseayy (661) aeqeiseT Aq s}UsWoINseayy
uajssuny, Jo Aytatssturg peysedg =(7'7'T)T AGeL

16

Forsythe and Adams, 1945; Nottingham and
Mutter, 1948; DeVos, 1954; Larrabee, 1959). Ta-
ble 1(1.2.2) gives the spectral emissivity of tungs-
ten as measured by DeVos (1954; values read
from published graphs) and Larrabee (1959).
Small but systematic deviations between the two
independent sets of measurements may be ob-
served. These deviations are attributable on the
one hand to experimental errors and on the other
to different amounts of impurities and different
surface conditions of the tungsten samples used
by the two experimenters. Figure 3(1.2.2) il-
lustrates the spectral distribution of emissivity of
tungsten for two temperatures (T = 1600 K and
T = 2400 K). The solid curves refer to measure-
ments by Larrabee, the dashed curves to corre-
sponding measurements by DeVos.
In most practical cases, the spectral transmit-
tance of the glass envelope of a tungsten filament
lamp can be assumed uniform, especially when
only the visible region of the spectrum is consid-
ered. However, in critical work, and particularly
when the ultraviolet region of the spectrum plays
a part in the investigation, account must be taken
of the spectral transmittance of the envelope. A
typical spectral transmittance curve of a glass
TUNGSTEN
OF
EMISSIVITY
300 400 500
WAVELENGTH
Fig. 3(1.2.2). Spectral emissivity of tungsten at temperatures T = 1600 K and T= 2400 K
(after measurements by Larrabee, 1959 and DeVos, 1954).
Sources of Radiant Energy 17
envelope of a 500-W lamp is shown in Figure
4(1.2.2).
Figure 5(1.2.2) shows the spectral irradiance
distribution of a tungsten filament lamp operat-
ing at a color temperature of 2850 K.
(vi) Tungsten—Halogen. Tungsten—halogen lamps
show a number of important advantages over
ordinary tungsten lamps (IES, 1972). They are
more compact and provide a considerably better
light output. By the addition of a small quantity
of halogen to the filling gas of a tungsten filament
lamp, a regenerative halogen cycle is set up.
(Halogen is the name given to a family of nega-
tive elements; i.e., bromide, chlorine, fluorine,
and iodine.)
The tungsten filament, when operating at in-
candescence, loses some tungsten through
evaporation. In an ordinary tungsten lamp, the
evaporated tungsten would deposit on the lamp
bulb, resulting in bulb blackening, common with
a tungsten lamp and most evident near the end of
the lamp’s life. However, in a tungsten—halogen
lamp, the temperature of the quartz bulb, en-
veloping the filament in close proximity, is suffi-
ciently high (above 260°C) so that the evaporated
—— LARRABEE (1959)
DE VOS (1954)
600 700 800
A(nm)
 O:oO

O: S
TRANSMITTANCE

0.2

300 400 500 600 700 800

WAVELENGTH )(nm)
Fig. 4(1.2.2). Typical spectral transmittance curve of a glass envelope of a 500 W tungsten
filament lamp.

ari >: D4 ae ak

O95- Tungsten Lamp

200 W, Osram Type Wi 41 /G
aes Measured at a distance of |0Ocm,
5.78 A, 316 Vdd i
—s Color Temp. Te =2850 K |
EWOr
Cc

oy
=
2506. = 4
=
=

: |
pO Bee!ox

804+
S3) 03 +

D
02+
OL

| ee ee |

300 400 500 600 700 800

Wavelength (nm)
Fig. 5(1.2.2). Spectral irradiance distribution of a tungsten filament lamp operating at a
color-temperature of 2850 K. The lamp measured serves as a spectral irradiance working
standard.

18

tungsten combines with the halogen to a gaseous
tungsten—halogen compound. When this com-
pound comes in contact with the hotter filament,
it breaks up into its original components of
tungsten and halogen. The tungsten is redeposited
on the filament and the halogen continues its role
in the regenerative tungsten—halogen cycle. De-
spite the fact that the filament does not lose
appreciable amounts of tungsten, the life of a
tungsten—halogen lamp is not unlimited. The re-
deposition of tungsten does not necessarily occur
at the same spots from where the tungsten came,
thus gradually weakening the filament.
Tungsten—halogen lamps can operate at higher
filament temperatures than ordinary tungsten
lamps, thus providing spectral power distri-
butions with correspondingly higher color tem-
peratures. Figure 6(1.2.2) shows the spectral
irradiance distribution of a tungsten—halogen
lamp operating at a color temperature of ap-
proximately 3100 K.
(vii) Filtered Tungsten Filament Lamps. Tung-
sten filament lamps are often combined with col-
ored glass filters to duplicate approximately the
spectral radiant power distribution of certain de-
sired phases of natural daylight. Such “daylight
simulators” are of importance for visual inspec-
tion and grading of colored materials and visual
color matching in a variety of industrial produc-
tion control tasks. The problem of assessing the
Tungsten - Halogen Lamp
2OOW, GE Type Q6.6A/T4/1CL
Measured at a distance
6.50 A, 28.9 V (dc)
Color Temp. Tc, = 3I00K
(uW:-cm-2-nm-!)
Irradiance
Spectral
300 400 500
Wavelength d (nm)
Fig. 6(1.2.2). Spectral irradiance distribution of a tungsten—halogen lamp operating at a
color temperature of 3100 K. The lamp measured serves as a spectral irradiance standard.
Sources of Radiant Energy 19
quality of daylight simulators of any make, in-
cluding filtered tungsten filament lamps, has been
studied by the CIE Colorimetry Committee and is
treated in detail in Section 3.3.5 dealing with CIE
standard sources.
1.2.3 Electric Discharge Lamps
In most electric discharge lamps, radiant energy is
produced by passing an electric current through a
gas or vapor. Electrons, accelerated by an electric
field applied between two electrodes, collide with
the gas molecules and give up their kinetic energy
in processes yielding electrically excited molecules
and atoms, ions, and further electrons. The
charged particles produced are themselves accel-
erated by the field and contribute to building up
the discharge process. When the particles whose
energy levels have been raised return to their
normal state, the excitation energy is emitted in
the form of quanta of radiant energy. In contrast
to the thermal radiators described above, electric
discharge lamps emit /ine spectra, in addition,
frequently to a more or less intense continuous
spectrum. A line spectrum is composed of a num-
ber of monochromatic lines of certain wave-
lengths, separated by regions in which no radiant
energy is emitted. The spectral radiant power
distribution of an electric discharge lamp and the
amount of radiant power emitted depend on many
factors but primarily on the vapor present in the
of 43cm,
600 700 800
20 PHYSICAL DATA
e)
00
ine)
units)
(arbitrary
Power
Radiant
500
Wavelength d (nm)
Fig. 1(1.2.3). Relative spectral radiant power distribution of a “white” carbon arc lamp.
lamp, the vapor pressure, the density of electric
current passing through the vapor, and the elec-
trode material.
There are many types of electric discharge
lamps (IES, 1972); the following are of particular
importance to colorimetry and its associated
fields.
(i) Carbon Arc Lamps. Carbon arc lamps emit
considerable radiant power in the ultraviolet, visi-
ble, and infrared regions of the spectrum. The
radiation is a combined effect of the incandes-
cence of the electrodes and the luminescence of
vaporized material and other chemical com-
pounds in the surrounding gaseous atmosphere.
The spectral radiant power distribution of a
carbon arc lamp is a complex function of the
electrode materials. It consists of a continuous
distribution on which are superimposed spectral
lines and bands. By means of properly chosen
electrodes and core material, the spectral lines
and bands can be broadened to give, together
with the continuum, a radiant power distribution
effectively continuous throughout the visible
spectrum (Coblentz et al., 1926). An example of a
spectral radiant power distribution of a carbon
arc lamp is shown in Figure 1(1.2.3).
(ii) Mercury Vapor Lamps. The principle lines
of the ultraviolet and visible spectrum of a
mercury vapor lamp are at approximately 254,
313, 365, 405, 436, 546, and 578 nm, and for the
ultraviolet lines to be available outside the lamp,
the envelope of the latter must be of quartz. The
radiant power emitted at the various wavelengths
varies mainly with the vapor pressure and the
density of electric current passing through the
vapor. Increasing operating pressure tends to shift
600 700
a larger proportion of the radiant power into
longer wavelengths. The spectral lines then be-
come broader and a continuous spectrum is
superimposed upon the line spectrum. About one
half of the total radiant power emitted by the
lamp is in the ultraviolet region. By the use of
coatings of “fluorescent” materials (phosphors)
applied to the inside of the lamp envelope, the
output of visible radiant power can be increased.
The phosphor layer absorbs ultraviolet radiant
power and emits, as a result, a continuous spec-
trum of longer wavelengths. The spectrum of the
fluorescent light depends upon the phosphor or
mixture of phosphors used in the envelope (Barnes
and Forsythe, 1937; IES, 1972). Figure 2(1.2.3)
shows the relative spectral radiant power distribu-
tion of a typical mercury vapor lamp.
(iii) Fluorescent Lamps. Closely related to the
mercury vapor lamp is the fluorescent lamp. The
fluorescent lamp is also an electric discharge lamp
usually in the form of a long glass tube, the inside
of which is coated with phosphors with the elec-
trodes at the two ends. The tube is filled with a
rare gas that carries the electric discharge until
the drop of liquid mercury in the tube has been
vaporized. The discharge conditions are such as
to give a high yield of the ultraviolet components
of the mercury spectrum. The ultraviolet radiant
power, mainly that of wavelength 253.7 nm,
excites the phosphor to radiate energy of longer
wavelengths.
The phosphors used in improved-color mercury
lamps and in fluorescent lamps are crystalline
inorganic compounds of high chemical purity.
Frequently, one or more activators are added to
increase the efficiency of the phosphor and con-
trol the spectral emittance to some extent. Table
 Sources of Radiant Energy 21

units)
(arbitrary
POWER
RADIANT

400 500 600 700

WAVELENGTH A (nm)

Fig. 2(1.2.3). Relative spectral radiant power distribution of a typical mercury vapor lamp.

1(1.2.3) lists some of the phosphors and activators
used together with the wavelength at which maxi-
mum emittance occurs (IES, 1972). Figures
3(1.2.3) and 4(1.2.3) show examples of relative
spectral radiant power distributions of common
fluorescent lamps.
Fluorescent lamps are used in a variety of
visual tasks. However, when they are used as
“daylight simulators” in critical color matching
tasks, they are often found to be unsatisfactory.
As can be seen by comparing Figure 3(1.2.1) with
Figures 3(1.2.3) and 4(1.2.3), the spectral radiant
power distributions of fluorescent lamps do not
approximate those of natural daylight. The large
spectral differences can lead to unacceptable mis-
matches of colored materials that are supposed to
color match under daylight (see Section 3.3.5).
Fluorescent lamps are also used in tachisto-
scopes; that is, instruments that provide visual
stimuli of brief duration for studies such as back-
ward masking, apparent movement, visual laten-
cies, and information processing. The time
constants of electronic tachistoscopes incorporat-
ing fluorescent lamps have been found to be
much poorer than has usually been assumed. Mol-
len and Polden (1978) have published a critical
review of such instruments and discuss the possi-
ble adverse effects the long instrumental time

Table 1(1.2.3) Examples of Phosphors and Activators Used in Fluorescent Lamps (IES, 1972)
Wavelength of
Phosphor Activator Maximum Emittance, nm Lamp Use

Calcium tungstate Lead 440 Blue

Magnesium tungstate — 480 Daylight
Barium titanium phosphate Titanium 490 Various whites
Calcium halophosphate Antimony and 590 Cool white, white,
manganese Warm white, daylight
Calcium silicate Lead and manganese 610
Strontium magnesium Tin 620 Color-improved cool
phosphate and warm white
Calcium-strontium Tin 640
phosphate
Magnesium fluoro- Manganese 660 Color-improved whites
germanate
 200

100

(arbitrary
units)
POWER
RADIANT

400 500 600 790

WAVELENGTH (nm)
Fig. 3(1.2.3). Relative spectral radiant power distributions of common fluorescent lamps: (1)
standard warm white; (2) white; (3) standard cool white; and (4) daylight. The distribution
curves have been scaled by appropriate constant factors to provide a common value of 100 at
A = 560 nm.

200

100

units)
(arbitrary
POWER
RADIANT

400 500 600 700

WAVELENGTH XA(nm)
Fig. 4(1.2.3). Relative spectral radiant power distributions of common fluorescent lamps: (5)
warm white de luxe; (6) soft white; (7) cool white de luxe. The distribution curves have been |
scaled by appropriate constant factors to provide a common value of 100 at A = 560 nm.

pH)
 Sources of Radiant Energy A

constants have on certain visual experimental
data.
(iv) High-pressure Xenon Arc Lamps. High-
pressure xenon arc lamps are used with consider-
able advantage in colorimetry; they can be excel-
lent “daylight simulators” (Section 3.3.5). These
electric discharge lamps have an extremely high
output of radiant power throughout the ultra-
violet, visible, and infrared regions of the spec-
trum. In the ultraviolet and visible regions of the
spectrum, the spectral distribution is continuous
but superimposed by a series of closely packed
and pronounced emission bands between 450 and
500 nm. In the near infrared region of the spec-
trum (from 800 to 1200 nm), there are numerous
emission bands of considerable power making the
spectrum of the arc lamp rather jagged. Figure
5(1.2.3) shows an example of a spectroradiometric
record of a high-pressure xenon arc lamp.
(v) Zirconium Concentrated-Arc Lamps. The
radiant power from a zirconium concentrated-arc
lamp shows a strong continuous spectrum from
the cathode surface, and a line spectrum and
weak continuous spectrum from the excited inert
gas (usually argon) and zirconium vapor in the
arc stream. A valuable feature of the low wattage
discharge lamps of this type is that they provide
an extremely small, very intense arc forming a
luminous spot of approximately 0.12 to 0.25 mm
in diameter (depending on wattage) which is a
practical approximation to a point source. Figure
6(1.2.3) illustrates the spectral radiant power dis-
tribution of a typical zirconium concentrated-arc
lamp.
(vi) Flashtubes. Flashtubes are electric dis-
charge lamps that emit radiant power of large
amounts during an extremely short period of time.
The spectral distribution provided by a flashtube
mainly depends on its inert gas. The most com-
mon gases used in flashtubes are xenon, argon,
hydrogen, and krypton. Figure 7(1.2.3) illustrates
the spectral radiant energy distribution of a
xenon-filled flashtube (IES, 1972).
1.2.4 Electroluminescent Sources
Some special phosphors convert energy of an
alternating electric current directly into radiant
energy without the intermediate step of producing

10

09 High-Pressure
Xenon Lamp
08
Anode

i
O6 S point of
Arc Te : ot a measurement

O05 &
04 Cathode =

0.3
(arbitrary
unit)
Power
Radiant

O72

0.

300 400 500 600 700 800 900

Wavelength (nm)
Fig. 5(1.2.3). Relative spectral radiant power distribution of a high-pressure (2 kW) xenon arc lamp.
24 PHYSICAL DATA

60
=
=
fe
q
oc
=
& 40 54
ac
Ss

uJ
=
fe)
a
2az 20
ra
<
ac

400 500 600 700

WAVELENGTH X(nm)
Fig. 6(1.2.3). Relative spectral radiant power distribution of a zirconium concentrated-arc
lamp.

ultraviolet radiant energy by an arc discharge
through a vapor. This phenomenon is called elec-
troluminescence. The electroluminescent phos-
phor, such as cubic zinc sulfide, is mixed with
one or two activators (copper, lead) and im-
bedded into the dielectric of a thin-plate capaci-
tor. The capacitor consists of a transparent
conductive film on a sheet of glass, clear ceramic,
or plastic. The spectral radiant power distribution
obtained from an electroluminescent source de-
pends on the phosphor, but to some extent also
on the frequency of the current. For a given

I6r T T T pes ae T

XENON-FILLED
Pe FLASHTUBE
‘o
7
5 l2
2
Z
-|8
‘|S
aI

3|8
Se 8;
>

ra
ui 6F
Prd
Lid

B 4
z
(a
ee

a Eh i ies ees) ee ey eee ae See ih

300 400 500 600 700 800 3900
WAVELENGTH )(nm)
Fig. 7(1.2.3). Spectral radiant energy distribution of a xenon-filled flashtube operating at a
current density of 1700 A-cm? (IES, 1972).

(cd-m-2)
Luminance
Volt (V)
Fig. 1(1.2.4)._ Luminance of green ceramic (Curve
1) and plastic (Curve 2) electroluminescent sources
operated at 400 cycles as a function of voltage
(IES, 1972).
design, the luminance is mainly a function of
both voltage and frequency, as illustrated in Fig-
ure 1(1.2.4).
The application of electroluminescent sources
is limited to locations where the general illumina-
tion is low. They are used in instrument panels,
switchplates, clockfaces, and the like. In certain
visual experiments, they may be useful as sur-
round stimuli of low luminance. They are com-
mercially available in green, blue, yellow, and
white, and can be fabricated into simple or com-
plex shapes (IES, 1972).
1.2.5 Light-Emitting Diodes
Light-emitting diodes (LED) are solid-state
sources of radiant energy. The radiant energy
producing material in a light-emitting diode is a
specially prepared semiconductor material of high
purity to which have been added small amounts
of other chemical compounds as controlled “im-
purities.” Two kinds of impurities are used. One
yields a semiconductor material, called n-type
material, having an excess of electrons. The other
yields a p-type material having a shortage of
electrons or “holes.” The two kinds of impurities
are diffused into the same piece of semiconductor
material so that an interface or “junction” is
produced between the n-type and the p-type
materials. When a sufficiently high dc potential is
applied to the semiconductor material with polar-
ity such that the n-type is negative and the p-type
is positive, the electrons and the holes are forced
Sources of Radiant Energy 22
units)
Radiant
(arbitrary
Power
600 620 640 660 680 700 720
Wavelength X (nm)
Fig. 1(1.2.5). Relative spectral radiant power dis-
tribution of a typical light emitting diode (LED).
to meet at the junction where they combine to
produce photons (Hall, 1969).
The spectral radiant power distribution of an
LED is confined to a fairly narrow spectral re-
gion. The wavelength of its peak emission is de-
pendent on the semiconductor material and the
added impurities. Most LEDs presently available
emit in either the yellow, red, or infrared region
of the spectrum. Special phosphors have been
incorporated in some LEDs to produce emission
in the green. There are no shortwave-emitting
LEDs on the market. Figure 1(1.2.5) shows a
relative spectral radiant power distribution of a
typical LED.
Light-emitting diodes are physically very small.
The p-n junction semiconductor “chip” generally
has a size ranging from 0.2 to 1 mm square; the
assembled source varies in size from 2 to 5 mm
diameter depending on its design and construc-
tion. In general, LEDs operate at low voltages
(one to three volts) and currents below 10 mA to
about 200 mA. In the visible spectrum, luminous
intensities of the order of 0.015 candelas are
produced.
Apart from their wide application in computer
systems, LEDs are also used in some color-vision
instrumentation as point-fixation lights.
1.2.6 Lasers
The radiant energy emitted by a laser (Laser =
light amplification by stimulated emission of
radiation) differs in several ways from that emitted
by other sources. A laser emits coherent and
highly monochromatic radiant energy of very high
26 PHYSICAL DATA

intensity. Laser radiation is a process in which a
photon stimulates a molecule or ion, already in an
excited state, to emit another photon in phase
with the stimulating photon. The lasing material
can be a gas, a liquid, or a solid. It is brought into
an excited state by absorption of electromagnetic
energy at wavelengths other than that of stimu-
lated emission. When a sufficient “population” of
lasing material is in the excited state, an external
source of photons of exactly the same energy as
that to be emitted will stimulate the population of
excited-state material to give up its excitation
energy in phase, thus greatly amplifying the num-
ber of photons over those that stimulated the

300
=
[=9
ro

2
(a) (b) “— 200
oO

=
ef
cm-*
3nm
#W
per

100

coe
300 350 250 300 350 400
WAVELENGTH (nm) WAVELENGTH X(nm)

5000
3
Cc
Mm

4000
d@) -
=
oO

3
oS ~S oe

cm-?
3nm
#W
per

1000

250 300 350 400 250 3005) £360) 400

WAVELENGTH Xinm) WAVELENGTH Xnm)

Fig. 1(1.2.7). Spectral radiant power distributions in the ultraviolet of 20 different sources
measured by Hirt et al. (1960) in terms of pW-em ?- (3 nm) !.(a). Direct Sunlight at noon
at Stamford, Connecticut, U.S.A.: Curve 1: Feb. 11, 1957; Curve 2: May 3, 1957; Curve 3: July
17, 1957; Curve 4: Dec. 27, 1957. (6b). Xenon Arc Lamps: Curve 5: Hanovia 507Cl, 1000 W,
25 cm distance; Curve 6: same as 5, but with Corning filter 7740; Curve 7: Hanau Xenotest W
1500 W, 6.25 cm distance. (c). Fade-Ometer: Curve 8: Atlas OA-2, 2700 W, 25 cm distance.
(d). Sunlamps: Curve 9: GE S-1, 400 W, 21.15 cm distance; Curve 10: GE RS, 275 W, 21.25
cm distance. (e). Mercury Arc Lamps: Curve 11: GE AH3, 85 W, 25 cm distance; Curve 12:
GE AH6, 1000 W, 25 cm distance. (f). Fluorescent Lamps: Curve 13: GE Blacklight
FIST8/BL, 15 W, 7.5 cm distance; Curve 14: GE Blacklight F1ST8/BLB, 15W, 7.5 cm
distance; Curve 15: GE Daylight F1S5T8/D, 15 W, 7.5 cm distance; Curve 16: Westinghouse
Sunlamp F20T12/Sun, 20 W, 7.5 cm distance. (g). Projection Lamp: Curve 17: GE 100T8,
100 W, 7.5 cm distance; Blacklight: Curve 18: GE Purple-X 250 A21/60, 250 W, 7.5 cm
distance. (4). Hydrogen Arc Lamps: Curve 19: Beckmann Instruments, air-cooled, 7.5 cm
distance; Curve 20: A.G. Nester, water-cooled, 50W, 7.5 cm distance.
 Sources of Radiant Energy 27

emission. Flashtubes are often used as external
sources supplying the electromagnetic energy re-
quired for “pumping.” However, other methods
of pumping are also used.
A large variety of lasers are commercially
available, designed to serve many different appli-
cations in science and engineering. Their radiant
power output may be a few milliwatts or several
hundred kilowatts; they may operate continu-
ously or in a pulsed mode; they may emit radiant
energy at one wavelength, or simultaneously at a
few discrete wavelengths, or may be “tunable” to
emit any desired wavelength in a certain spectral
range. Useful lists of present-day lasers, their
classification, wavelengths, and performance
characteristics can readily be found in the litera-
ture (for example, Grum and Becherer, 1979).
In color science, the main application of lasers
is in the field of spectroradiometry where they are
sometimes used in place of conventional mono-
chromators. They have also proven very useful
tools for aligning optical components in a variety
of instruments. Their use as sources of visual
stimuli is limited because of the “speckle” pattern
laser beams produce in the field of view (Dainty,
1975). However, visual scientists continue to be
interested in lasers as a tool and, at times, find
them useful (Ingelstam and Ragnarsson, 1972).
1.2.7 Sources of Ultraviolet Radiant Energy
To specify the color of fluorescent materials, or to
deal with problems of fading or bleaching, it is
necessary to consider the spectral distribution as
well as the quantity of ultraviolet radiant energy
received by the material. An informative set of 20
different spectral radiant power distributions of
various sources emitting (incoherent) ultraviolet
radiant energy has been published by Hirt et al.
(1960). Their main results are illustrated in Figure
1(1.2.7). Note that all data are given uniformly in
terms of wW- cm” - (3nm)~! for wavelength A
from 266 to 410 nm.
1.2.8 Some Photometric and Colorimetric
Characteristics of Sources of Radiant Energy
Table 1(1.2.8) gives representative data for the
luminance, chromaticity coordinates, and corre-
lated color temperature of some of the sources of
radiant energy discussed above. In connection
with this table, the following notes may be of
interest.
(i) Daylight. Data on natural daylight varies
greatly with the time of day, the season of the
year, and the weather conditions. Some investiga-
tors have studied daylight over many years, and

400

€
c fe
c
rm
Mm

S 4
Cy (uae
£
oO
E
o

=
>
a
1

20

(g) E
fe) taal +
3 20

cm-*
3nm
pW
per (MF
h a, 19 a
oO

=
a
250 350 250 300 350 400
WAVELENGTH X(nm) WAVELENGTH X(nm)

Fig. 1(1.2.7).. Continued

Table 1(1.2.8) Representative Photometric and Colorimetric Data of Different Sources of
Radiant Energy
get oe eS Se ee ear ee
Chromaticity
Coordinates Correlated
Luminance e eee
2 Color Temperature
Source (cd-cm ”) x y (K)
pi
eebalai'n tm feceeeer oiMES see ener, scl nar af a
Sun disk“
Measured above atmosphere
(m= 0) 200,000 0.318 0.330 6,200
Measured near sea level
for air masses?
m=1 150,000 0.331 0.344 5,600
m=2 125,000 0.343 0.357 5,100
m=3 100,000 0.356 0.369 4,700
m=4 80,000 - 0.368 0.379 4,400
m=5 65,000 0.380 0.388 4,100
Sky‘
Clear blue sky 0.06 to 0.4 0.262 0.270 15,000
0.247 0.251 30,000
Partly cloudy sky 0.1 to 0.4 0.294 0.309 8,000
0.279 0.291 10,000
Overcast sky 0.2 to 0.5 0.313 0.329 6,500

Blackbody“
Gold point 0.11 0.607 0.381 1,326
Platinum point 60 0.522 0.414 2,042
Rhodium point 170 0.502 0.415 2,293
Iridium point 1,250 0.458 0.410 2,716
Tungsten filament*®
Vacuum lamp (10 Im-W') 200 0.477 0.414 2,400 to 2,600
Gas-filled lamp 1,200 0.444 0.406 2,700 to 3,100
(20 lm-W!)
750 W projection lamp 2,400 0.430 0.402 3,000 to 3,200
(26 Im: W~')
1200 W projection lamp 3,300 0.417 0.396 3,200 to 3,400
(31.5 Im-W~')
60 W inside frosted bulb 12 0.452 0.409 2,800
(141m-W~')
60 W “white” bulb 3 0.452 0.409 2,800
(141m-W~!)
Carbon arcs (DC) 15,000 to 4,100 to 6,500
145,000
Mercury arc (high-pressure) 13,000
(1000 W; 65 Im/W)
Fluorescent lamps (T-12; 40W; 430 mA)
Std. Warm White 0.85 0.431 0.406 3,100
White 0.85 0.402 0.394 3,600
Std. Cool White 0.83 0.367 0.375 4,370
Daylight 0.69 0.313 0.332 6,500
Deluxe Cool White 0.58 0.376 0.366 4,050
Deluxe Warm White 0.57 0.434 0.400 3,020
Xenon (high-pressure, XBO
1600 W DC) 65,000 07322 0.320 6,000
Zirconium (300 W) 4,500 0.423 399 3,200

“Chromaticity coordinates after Moon (1940).

’The air masses m=1,2,...,5 are equivalent to the secants of the following angles between the zenith
and the sun: 0°(m=1), 60°(m = 2), 70.5°(m = 3), 75.5°(m = 4), 78.5°(m = 5).
“For clear blue and partly cloudy skies, the range of minimum luminance is given (Paix, 1963).
Chromaticity coordinates and correlated color temperature are estimated from results by Judd,
MacAdam, and Wyszecki (1964).
“After Pivovonski and Nagel (1961).
“After IES Lighting Handbook (IES, 1972).

on the basis of their records it is possible to
establish averages that have proven to be useful
for the architect, the photobiologist, and others
(Walsh, 1961; Paix, 1963; Henderson, 1977). For
the spatial distribution of the luminance of the
overcast sky and the blue sky, formulae have been
developed that agree with actual experimental
data to a fair approximation. For the overcast
sky, the following formula by Moon and Spencer
(1942) has been recommended by the CIE (1955):
Ly = 3L,(1 + 2sin8) [1(1.2.8)]
where L, is the luminance of the sky at an angle 0
above the horizon and L, is the luminance at the
zenith. In accordance with this formula, the lumi-
nance at the horizon (@ = 0) is one-third of that
at the zenith. For the blue sky, Pokrowski’s (1929)
theoretically derived formula may be used:
PA 1 + cos? a ae
aH 1 — cosa [ exp (=)
[2(1.2.8)]
where L,, is the luminance at an angle @ above
the horizon and an angle a from the sun, P is a
“scattering coefficient” usually set equal to 0.32,
and L,, is the luminance at two points on the
horizon whose angular distance from the sun is
90°.
A modified formula which has been found in
better agreement with experimental data in cer-
tain cases is as follows:
Joppaaieba apne
|1 —cosa
se + x]
x[1-e(S9)| [3(1.2.8)]
where the additional constant K is introduced
empirically to take account of scattering by par-
ticles larger than those responsible for the
Rayleigh scattering. The value found for K is
approximately 5.
A more recent relationship of the luminance
distribution of the blue sky has been given by the
CIE (1973).
The gamut of chromaticity coordinates corre-
sponding to daylight of all phases is illustrated in
Figure 2(1.2.1). The graph is based on many
hundreds of measurement made either by direct
colorimetry or by spectroradiometry and compu-
tation (Judd, MacAdam, and Wyszecki, 1964).
Sources of Radiant Energy 2a
= 60 cd-cm2
2042 K Platinum Point
(Primary Standard
of Light)
(cd-cm-2)
of
Logarithm
Luminance
1000 2000 3000 4000
Temperature (K)
Fig. 1(1.2.8). Logarithm of the luminance of
blackbody radiators as a function of temperature
(kelvin). The blackbody radiator at the tempera-
ture (T = 2042 K) of freezing platinum has a
luminance of 60 cd-cm7. (See Chapter 4 on
Photometry.)
The mean daylight locus passing through this
gamut is also shown. It lies somewhat above the
Planckian locus in the CIE 1931 (x, y)-chro-
maticity diagram.
(ii) Blackbody Radiator. For the blackbody
radiator, the spectral radiant emittance is given
by Planck’s formula [Eq. 2(1.2.1)], and luminance
and chromaticity coordinates can be computed
readily in accordance with the CIE colorimetric
system (Section 3.3.8). The luminance of black-
body radiation increases rapidly with increasing
temperature. Figure 1(1.2.8) shows the relation-
ship for a range of temperatures of practical
blackbody radiators. For related discussions, see
Section 3.11 on distribution temperature, color
temperature, and correlated color temperature.
(iii) Incandescent Filament Lamps and Electric-
Discharge Lamps. From the enormous variety of
lamps commercially available, only very few have
been chosen for their particular interest to the
colorimetrist and these are listed in Table 1(1.2.8).
For more detailed and comprehensive data, the
catalogues of the lamp manufacturers should be
consulted. The JES Lighting Handbook also con-
tains a valuable collection of data (IES, 1972).
 30 PHYSICAL DATA
The correlated color temperatures given in Ta-
ble 1(1.2.8) have been determined graphically in
accordance with the procedure described in Sec-
tion 3.11.
13 OPTICAL FILTERS
In the broadest sense, an optical filter is a device
or material that changes selectively or nonselec-
tively the spectral distribution of the incident
radiant flux. Typical examples are colored glass
filters and interference filters. Prism and grating
monochromators and reflecting materials such as
mirrors and paint surfaces could also be classified
as optical filters, but these are more conveniently
treated separately. Many different types of optical
filters have been described in the literature, but
only a few are of direct interest to colorimetrists
and research workers in vision. For additional
information the reader is especially referred to the
papers by Mohler and Loofbourow (1952), Geff-
ken (1957), Dobrowolski et al. (1977), and
Dobrowolski (1978).
1.3.1. Absorption Filters
The most common types of absorption filters are
made of glass, gelatin, or liquids in which color-
ing agents are dissolved or suspended. Radiant
flux P(A) incident on the first surface of the
filter propagates through the filter medium and
emerges from the second surface as radiant flux
P(A), see Figure 1(1.3.1). Portions of the radiant
flux arriving at the first and second surfaces are
lost by reflection, whereas the remaining portions
P()
ABSORPTION
FILTER
Fig. 1(1.3.1). Schematic diagram showing radiant
flux P)(A) passing through an absorption filter.
The flux P(A) emerging from the filter is reduced
due-to reflection losses on the two filter surfaces
and absorption within the filter medium.
are transmitted but reduced because of absorp-
tion within the filter medium.
A precise mathematical formulation of the
transmittance of an absorption filter is rather
involved and depends on a number of parameters
including the optical constants of the medium
given by its complex index of refraction, the
thickness of the medium, the angle of incidence,
polarization, wavelength, and temporal and spa-
tial coherence of the radiant energy. All these
parameters also play a part in the measurement of
spectral transmittances when the highest precision
and accuracy is the goal of the measurement. A
detailed account of such measurements has been
given by Mielenz et al. (1973). However, in most
practical situations, many of the parameters in-
volved in the precise definition of transmittance
and its measurement are of little or no impor-
tance, and formulae related to the transmittance
of absorption filters become rather simple. A few
of these formulae have been selected for general
use.
For homogeneous isotropic absorption filters,
Bouguer’s law (Bouguer, 1729) is valid
o(A) = 10°4"™ [1(1.3.1)]
where d is the path length of the beam of radiant
flux in the filter medium and m(A) is the spectral
absorptivity of the medium. At normal incidence,
d is equal to the thickness of the filter. The
quantity #(A) is called the internal spectral trans-
mittance of the filter and is defined as the ratio of
the radiant flux reaching the back surface of the
filter to the flux that enters the filter at its front
surface.
Changes in concentration of a liquid solution
have the same effect as changes in path length.
Thus with
m(r) = cee ez
we obtain
O(A) = 10°40 [3(1.3.1)]
which is usually referred to as Beer’s law (Beer,
1852). The quantity c is called the concentration of
the absorbing solute; it is measured, for example,
in grams per liter. The quantity e(A) is the extinc-
tion coefficient of the solute material at wave-
length A.
Beer’s law has been found generally valid at
low and moderate concentrations. At higher con-
centrations, many solutions do not obey Beer’s

law, but the deviations are associated with the
formation or breakup of molecular aggregates
and other modifications of the absorbing mole-
cules and the solvent. Apparent deviations from
Beer’s law are also observed as a result of inho-
mogeneities in the liquid or because scattering
particles are present (Hardy and Young, 1948).
If dy and d are the original and new thick-
nesses, respectively, of an absorption filter obey-
ing Bouguer’s or Beer’s law, and if (A) and
o(A) are the corresponding internal spectral
transmittances, then
H(A) =[9Q)]%* — [4.3.19]
a relation often referred to as Lambert’s law
(Lambert, 1760). For the practical handling of
(A) it is convenient to use the internal optical
density (A) defined as
5(A) = —logig 8(A) = losi057R)
dce(A) = dm(A) [5(1.3.1)]
It follows that
8(A) = 4-80() [6(1.3.1)]
Losses of radiant flux by reflection from the
front and back surface of the filter may be com-
puted from Fresnel’s law of reflection (see Section
1.4). For normal incidence, the fraction p(A) of
radiant flux reflected from a surface separating
two media is given by
Foe (2.— | 7(1.3.1)]
where n(A) is the ratio of refractive indices of the
two media. If interreflections occurring between
the two plane-parallel surfaces of a filter are
taken into account, the spectral transmittance T(A)
is given by
mihi [1 ie p(A)] 9(X)
cad Bo 1— [pA)A)/
[8(1.3.1)]
This relation assumes that the only discontinuities
of refractive index in the filter are those at the
front and back surfaces.
Optical Filters 31
Equation 8(1.3.1) and the similar ones below
ignore any modifications produced by optical in-
terference among the various transmitted and re-
flected beams generated at the filter surfaces and
derived from a common incident beam. Because
the path differences of the interfering beams are
normally large, the effects of interference on
transmission are averaged out for any incident
beam containing a considerable band of wave-
lengths, such as a band of 10 nm width from a
monochromator. However, in determining the
transmittance of a train of very accurately aligned
plane-parallel optical components (filters, cubes,
etc.) for a parallel beam of more strictly mono-
chromatic light (for example, a spectral line), the
possibility of interference effects has to be kept in
mind. They are usually detected by the ap-
pearance of light and dark fringes in the field.
In many instances, interreflections contribute
little to the transmittance of the filter, and the
denominator in Eq. 8(1.3.1) may be ignored; thus
r(A) = [1 eAJPOQA) [9013.1]
Absorption filters may be combined in series
in order to obtain other and more desirable spec-
tral transmittance characteristics. The spectral
transmittance 7,.(A) of a filter combination is dif-
ficult to predict if interreflections between surfaces
are to be taken into account. Formulae become
complicated and a numerical prediction impracti-
cable (Stokes, 1862; Benford, 1923; Smith, 1926;
Tuckerman, 1947). Direct spectrophotometric
measurement is to be preferred. However, if inter-
reflections can be ignored, the formulae for 7,(A)
are simple. The following are of particular inter-
est:
(i) Combination of Filters in Optical Contact.
The spectral transmittance 7,(.) of a combination
of k filters in optical contact [see Figure 2(1.3.1)]
is given (approximately) by
TAA) = [1 = po A) HAIL! — 1A)
] 0)
x [1 Vv p(A)] i [1 x px—(A)]
<9, (A)[1 — o,(A)] [10(1.3.1)]
where p,(A) is the reflectance occurring at the jth
surface (j = 0,1,2,...,k) separating two neigh-
boring media, and #,(A) (i = | to k) is the inter-
nal spectral transmittance of the ith component
filter.
32 PHYSICAL DATA

(O) (1) (2) (3) (k-1) (k)

Fig. 2(1.3.1). Schematic diagram of a combination of k absorption filters in optical contact.
P(A) = incident radiant flux; P(A) = transmitted radiant flux.

It may often be assumed that the refractive given (approximately) by

indices of the component filters are the same.
Thus, according to Fresnel’s law, TAA) = [1 = oA) H(A) = 02(A)]?
p(A) = p2(A) = +--+ pz_-(A) = 0 x 3,(A) wes [1 a p.(A)] 9, (A)

and [12(1.3.1)]
P(A) = px(A) = P(A)
TAA) = 7)(A) (A) ++ (A)
and Eq. 10(1.3.1) reduces to
[13(1.3.1)]

TAA) = [1 = p(A) P(A)

(A) «+ 8,0)
[11(1.3.1)]
This formula describes the important practical
combination of filters of glass fastened together
with a cement of the same refractive index as that
of the component filters. The thickness and ab-
sorption of the cement layers are assumed to be
negligible.
This means that the spectral transmittance of a
combination of filters, all of which are separated
by air, is given by the product of the spectral
transmittances of the component filters.
If all filters have the same refractive index, so
that
pi(A) = pa(A) = + p.(A) = pA)
Eq. 12(1.3.1) becomes

(ii) Combination of Filters Separated by Air.

TA) = [1 = p(A)]*
(A) (A) --- 8,0)
The spectral transmittance 7,() of a combination
of k filters separated by air [see Figure 3(1.3.1)] is [14(1.3.1)]

Y
P(A)

Fig. 3(1.3.1).
li
Schematic diagram of a combination of k absorption filters separated by air.
-Po(A) = incident radiant flux; P(A )= transmitted radiant flux.
 Optical Filters 33

1.0

0.8

0.6

T(A)

0.4

0.2

400 500 600 700

WAVELENGTH A (nm)
Fig. 4(1.3.1). Spectral transmittances 7(A ) of component filters 1(2 mm), 1*(5 mm), and 2(5
mm), a combination of filters 1 and 1* in air, and a combination of filters 1 and 1* cemented
together.

From this equation it is readily seen that the single wavelength A = 550 nm; but the complete
reflection loss in a multifilter combination can be results are shown in Figure 4(1.3.1).
considerable if the filters are separated by air. ROCUTIONe

(iii) Numerical Example. Case (a)

Given: Spectral transmittance 7,(A) and
T(A) of two glass filters as shown in Figure EEOC CMU toenail aii
4(1.3.1); Filter 1 has a thickness of 2 mm, Filter 2 Tes meee Nea
Internal transmittance #,(A) from
has a thickness of 5 mm. :
peta: be serch ‘d f inverse of Eq. 9(1.3.1): 0.560
EEE Sal fae rnep ae ponder Internal optical density 6,(A) from
|ra ee] ibk ealBe 0.2518
(a) Filter 1 when its thickness is increased from
Internal optical density 6,(A) of desired
2 to 5 mm; the new filter is denoted by 1*;
Filter 1* from Eq. 6(1.3.1) with
(b) Uncemented combination of Filters 1* and
0. d/dy) = 5/2 = 2:5: 0.6295
‘ ay Internal transmittance 3*(A) from
(c) Cemented combination of Filters 1** and 2. invee of Bq esclael): 1 0.235
“ie 28 3 ¥
The calculations are made for normal inci- fens aeS RiPArof desires baltee |
ratits al from inverse of Eq. 9(1.3.1): 0.215
dence and a constant refractive index of n = 1.53
for both the given glasses and the cement; thus Gaceth)

5 Transmittance T/*(A) of Filter 1* as

[1 — p(A)]}" = const = 0.9142 calculated in Case (a): 0.215
Transmittance T,(A) of Filter 2 from
The computational procedure is illustrated for the Figure 4(1.3.1): 0.625
34 PHYSICAL DATA
Transmittance 7,(A) of desired uncemented
combination from Eq. 13(1.3.1): 0.134
Case (c)
Transmittance 7;*(A) of Filter 1* as
calculated in Case (a): 0.215
Transmittance 7,(A) of Filter 2 from
Figure 4(1.3.1): 0.625
Internal transmittance #*(A) of Filter 1*
as calculated in Case (a): 0.235
Internal transmittance 3,(A) of Filter 2
from inverse of Eq. 9(1.3.1): 0.684
Transmittance 7,(A) of desired cemented
combination from Eg. 14(1.3.1): 0.147
NOTE: It is important to recall that the for-
mulae used above apply only to completely ho-
mogeneous filter media with all surfaces perfectly
flat, polished, and parallel, and with the radiant
flux propagating in parallel rays. Any deviations
from these conditions will change the results to
some extent. In precision photometry, it is wise
not to rely too heavily on computational means,
but wherever possible to make direct measure-
ments. Filter combinations are preferably cali-
brated in exactly the same instrumental arrange-
ment as that in which they are to be used in
practice.
1.3.2 Glass Filters
A large variety of colored glass filters are com-
mercially available from several manufacturers in
different countries. A rather comprehensive col-
TRANSMITTANCE
——©>
WAVELENGTH
_Fig. 1(1.3.2). Effect of a temperature increase AT on the spectral transmittance of a glass
filter.
lection and an intercomparison of about 800 dif-
ferent glasses offered by 13 manufacturers has
recently been made by Dobrowolski et al. (1977).
Such a comprehensive collection cannot be pre-
sented here. Instead, we have limited our collec-
tion to the major types of colored glass filters
with their representative spectral transmittance
curves. We hope that this collection, given in the
Appendix as Table I(1.3.2), will provide the reader
with a useful overview of the domain of commer-
cially available colored glass filters.
The colored glass filters offered by the manu-
facturers in their catalogues carry an identifica-
tion and are usually characterized by their respec-
tive spectral transmittance curves over a wide
wavelength range. The given spectral transmit-
tance curve of a particular filter type characterizes
the type for a given glass thickness. Specific speci-
mens of that same type may vary somewhat in
their spectral transmittance curves from the nomi-
nal curve given in the catalogue, even if their
thicknesses are the same. In particular, if the
specimens come from different melts, slightly dif-
ferent spectral transmittance curves are not un-
common.
As a general rule, for precision photometry
and colorimetry, individual calibration of each
specimen is strongly recommended. It is further
recommended that each specimen be checked for
uniformity in transmittance across its area. Bub-
bles and striae within the glass medium occur
occasionally and can, at times, affect seriously the
uniformity of the filter.
The spectral transmittance of all filters is a
function of temperature (Gibson, 1916; Meyer,
\——*>

1936; Holland and Turner, 1941; Angstrom and
Drummond, 1959; Dodd and West, 1961). For
certain glasses, notably those containing selenium
oxide as a coloring agent, even small variations in
room temperature affect the spectral transmit-
tance significantly. For other glasses, the effect
may be small and measurable only over larger
ranges of temperature. Some general rules, il-
lustrated in Figure 1(1.3.2), include the following:
A spectral transmittance that increases with
increasing wavelength usually decreases with in-
creasing temperature. The greater the slope of the
spectral transmittance curve, the greater the tem-
perature effect, as a rule. A spectral transmittance
that decreases with increasing wavelength is, in
general, little affected, but if it is, it often in-
creases with increasing temperature. Spectral
transmittance curves varying only slightly with
wavelength are usually little affected by variations
in temperature. Transmittance changes due to
temperature are reversible; that is, the filter will
always return to its original transmittance at a
specified temperature.
Many glass filters fluoresce, particularly when
irradiated by ultraviolet radiant energy. Most
manufacturers identify them in their catalogues.
The fluorescence can, at times, be troublesome. If
a similar filter without fluorescence cannot be
found, it is sometimes possible to select an aux-
iliary filter that either absorbs the exciting radiant
energy or the fluorescence itself.
The surfaces of some glass filters tarnish when
exposed to the atmosphere, high humidity, or
high temperature for prolonged periods of time.
Some manufacturers identify them in their cata-
logues and provide useful information concerning
the stability of their glasses. Sensitive glasses can
sometimes be hardened by the manufacturer, or,
alternatively, be protected by cementing them
between two stable glasses. It is also now possible
to protect these glasses by thin-film coatings which
make them virtually impervious to the atmo-
sphere.
1.3.3 Gelatin Filters
Gelatin filters are made by mixing organic dyes in
gelatin and coating the mixture on glass plates.
After the coatings are dry, the gelatin films are
removed from the plates. Gelatin filters are avail-
able either as lacquered gelatin films or as gelatin
films cemented between glass plates. Manufac-
turers such as the Eastman Kodak Company offer
a large variety of gelatin filters with spectral
Optical Filters BD
transmittance characteristics often similar to those
of colored glass filters.
Gelatin filters are generally less expensive than
glass filters and they can be cut readily to the
desired size and shape. However, many of the
gelatin filters are less stable than glass filters with
regard to prolonged exposure to radiant energy,
high humidity, and temperature. In precision ra-
diometry, photometry, and colorimetry, they are
rarely used.
1.3.4 Liquid Filters
A wide variety of spectral transmittance curves
are provided by colored liquids and solutions.
Their use as optical filters for photometry and
colorimetry is, however, mostly confined to spe-
cial applications. Glass filters are usually pre-
ferred because of their greater convenience.
Rather comprehensive listings of chemical
solutions and their respective absorption char-
acteristics are given, for example, in the Landolt-
Bornstein books of tables (Halban, 1923; 1927;
1931; 1935; Dreisch, 1935; Pestemer et al., 1951;
Geffken, 1957), and the International Critical Ta-
bles (Becquerel and Rossignol, 1929; Carr and
Sherill, 1929; Henri, 1929, W. C. Holmes, 1930).
In these publications, extensive references to the
literature can also be found.
1.3.5 Absorption Filters for
Special Applications
(i) Heat-Absorbing Glasses. Heat-absorbing
glasses transmit near-ultraviolet and visible
radiant energy but absorb the infrared that causes
heating. Heat-absorbing glasses have a low ther-
mal expansion reducing the possibility of frac-
ture. The optical quality of heat-absorbing glasses
is usually not as good as that of most other glass
filters. Striae and particularly scattered small
bubbles may be found in the glass. Typical spec-
tral transmittance curves of commercially avail-
able heat-absorbing glasses are shown in Table
TC1..2) UNGer s sl ype, tz iter. .
(ii) Color Temperature Conversion Filters.
Color temperature conversion filters change, by
selective transmission, the spectral radiant power
distribution of a blackbody radiator at tempera-
ture 7, to a distribution approximating in the
visible to a blackbody radiator at another temper-
ature T,. Glasses that are especially suitable for
raising the color temperature are those given as
“Type 7 filters” in Table (1.3.2); glasses that
36 PHYSICAL DATA

Table 1(1.3.5) Color-Temperature Conversion Filters (Davis and Gibson, 1931)

Two-compartment cell consisting of three borosilicate crown glass plates (each 2.5 mm thick, refractive
index, np =1.51) separated by two glass frames, each 10 mm thick, for containing the respective A and B
solutions.

Solution A
Copper sulfate (CuSO, -5H,O) 8; (grams)
Mannite (C;H,(OH),) > (grams)
Pyridine (C;H,N) 30.0 (cubic centimeters)
Water (distilled) to make 1000.0 (cubic centimeters)

Solution B
Copper sulfate (CuSO, -5H,O) 8; (grams)
Cobalt ammonium sulphate
(CoSO,:(NH,).SO,:6H,O) > (grams)
Sulfuric acid (specific gravity, 1.84) 10.0 (cubic centimeters)
Water (distilled) to make 1000.0 (cubic centimeters)

The values g, and g,, weights in grams per liter of solution, vary with the purpose of the filter and are
given in the table below. Comprehensive listings are given by Davis and Gibson (1931).

Color Temperature* Solution

A B
Incident Transmitted
Light Light 81 8» 81 8
l 2856 K 3500 K 1.113 1.113 7.197 9.780
2 2856 K 5000 K 2.599 2.599 17.090 23.080
3 2856 K 6000 K 3.188 3.188 21.000 28.480
4 2856 K 7000 K 3.605 3.605 23.800 32.400
5 2856 K 8000 K 3.916 3.916 25.800 355302

“Color temperatures given in this Table refer to those of a blackbody radiator expressed by Planck’s
formula with c, =1.4388. The original data of Davis and Gibson (1931) are based on c, =1.4350.
Although the color temperature of the incident light has been corrected from the earlier value of 2848 K
to 2856 K, the color temperature of the transmitted light has been left unchanged. The new values differ
only little and insignificantly from the old ones. (For example, 3507 K instead of 3500 K, and 8017 K
instead of 8000 K.)

lower the color temperature are of “Type 2.” The
magnitude and direction of color temperature
change that can be accomplished depend on the
particular glass type and its thickness. Gelatin
filters are also available with similar properties.
The Davis—Gibson filters (Davis and Gibson,
1931) are liquid filters designed to convert CIE
standard source A (2856 K) to a series of sources
of higher color temperatures. Table 1(1.3.5) lists
the chemical compositions of some of them. For a
definition of color temperature, see Section 3.11.
The calculations required in the design of a
color temperature conversion filter are greatly
simplified if Wien’s formula [Eq. 8(1.2.2)] is used
instead of Planck’s [Eq. 2(1.2.2)] to describe the
spectral radiant power distribution of a blackbody
radiator (Fabry, 1913; Langmuir and Orange,
1913; Gage, 1933). The spectral transmittance
7(A) of the filter that changes the color tempera-
ture of a blackbody radiator from T, to 7, is
found from the following formula:
l i) 0.00624 Sa) dOSron 108
[1(1.3.5)]

The constant b takes account of the fact that only
relative spectral radiant power distributions are
being considered; A is measured in micrometers
(um). The formula defines a straight line when
logiol1/7(A)] is plotted against 1/A. The slope
tan ¢ of this line is given by
6 6
tan} = -0.006245| ie
TT| [2(1.3.5)]
It follows that the difference in reciprocal color
temperatures when measured in reciprocal mega-
kelvins (MK~'; see Section 3.11) is given by the
expression
Let see,
Ti Wee = — 160.1 tan¢ [3(1.3.5)]
A positive slope (tan ¢ > 0) indicates a filter that
reduces the color temperature; a negative slope
(tan @ < 0) indicates a filter that increases the
color temperature. If 7, is the original tempera-
ture, then T, is given by
=2 _
1+ 0.0001601 7, tan¢
[4(1.3.5)]
ate
A change in thickness of the filter changes the
difference in reciprocal megakelvins proportion-
ally, and we have
[5(1.3.5)]
where d, is the original, and d, the new, thickness
of the filter. The value of [(10°/7,) — (10°/T7,)]
for a filter of one millimeter thickness is some-
times called the color temperature altering power
of the material.
Practical color temperature conversion filters
always deviate to some extent from the ideal filter
(Estey, 1936; Gage and Macbeth, 1936; Harding,
1944; 1948). A plot of log, 9[1/7(A)] versus 1/A
gives a curve which is only approximately a
straight line, see Figure 1(1.3.5). Such a plot pro-
vides a good way of checking qualitatively whether
a given filter may be used as a color temperature
filter (Weaver, 1936). Usually, however, the plot
is not suitable for determining the actual color
temperature altering power provided by the filter.
Drawing the “best” straight line through the curve
is subject to considerable variation, and different
Optical Filters Dy
individuals may obtain different results
(Middleton, 1954). A more satisfactory method is
to determine the CIE 1931 (x, y)-chromaticity
coordinates of the filter with respect to the
blackbody radiator at temperature T, and to ob-
tain for those chromaticity coordinates the corre-
lated color-temperature T, by means of one of the
procedures outlined in Section 3.11.
An example of a practical color temperature
conversion filter is given in Figure 1(1.3.5). The
curves log; [1/7(A)] of a suitable type of glass at
thickness 1 mm (Curve 1) and at 6.5 mm (Curve
2) are drawn against 1 /A. Neither curve is straight,
but both resemble straight lines with sufficient
approximation for the glass to be considered rea-
sonably suitable as a color temperature conver-
sion filter. The color temperature altering power
is estimated to be approximately 35 MK _!. This
means that the l-mm filter changes the color
temperature of a tungsten lamp operating at 2856
K to approximately 3200 K. The same filter at 6.5
mm changes the color temperature from 2856 K
to approximately 8200 K.
The use of Wien’s formula instead of Planck’s
is only an approximation. Above 4000 K, dis-
crepancies between the formulae become increas-
ingly important. Estey (1938) has worked out a
procedure to take account of the discrepancies.
Color temperature conversion filters can be
improved by combining them with appropri-
ately designed thin-film interference filters
(Dobrowolski, 1970).
(iii) Narrow Passband Filters. Individual filters
or filter combinations may be designed to select a
narrow region of the spectrum within which a
portion of the incident radiant flux is transmitted,
whereas at all other regions of the spectrum, the
incident flux is absorbed. These narrow passband
filters can be used to isolate particular emission
lines of a discharge lamp or to replace a mono-
chromator in abridged spectrophotometry. The
literature dealing with narrow passband filters of
the absorption type is extensive and only a short
bibliography can be given here (Holmes, 1928;
Jones, 1928; Bowen, 1932; Gibson, 1935; Winther,
1937; Naumann and Kaltenbach, 1938; Staats,
1938; Biicher and Kaspers, 1946; Kasha, 1948;
Frihling, 1953; Kortiim, 1955; Angerer, 1957;
Wyszecki, 1960c).
In recent years, narrow passband filters of the
absorption type have commonly been replaced by
narrow passband interference filters which will be
discussed in Section 1.3.7. However, whatever the
38 PHYSICAL DATA

0.400 0.450 0.500 0.550 0.600 0.650 0.700 0.750

Fig. 1(1.3.5). Plot of log, 9[1/7(A)] against 1/2 for a practical color-temperature conversion
filter of spectral transmittance 7(A). Straight line corresponds to ideal filter. Curve 1 refers to
the selected glass filter at 1-mm thickness; Curve 2 refers to the same glass filter but at 6.5 mm
thickness. The slope of Curve 1 gives the color-temperature altering power of the glass
material, approximately equal to 35 MK ' in the given case.

type, the effectiveness of a narrow passband filter
in abridged spectrophotometry is sometimes ex-
pressed in terms of its maximum transmittance
and its half-width, the latter being the wavelength
interval within which the transmittance exceeds
half the maximum. However, such a specification
is incomplete, for it ignores the often important
“tails” of the transmittance curve.
The representative wavelength of a narrow
passband filter is given by the spectral centroid X .
defined by
[orA)SARA)A dd
ina apy TORTTON AT [6(1.3.5)]
where 1(A)=spectral transmittance of narrow
passband filter
S(A) dA=relative spectral radiant power dis-
tribution of light source
R(A)=relative spectral responsivity of de-
tector, such as eye or photocell
The spectral centroid defined above must not
be confused with the ‘‘effective wavelength” of
red filters used in optical pyrometry. For a defini-
tion of this very special concept see, for example,
Lovejoy (1962).
For details on some theoretical and practical
aspects of abridged spectrophotometry and re-
lated types of measurement, the reader is referred
to the original literature, for example, Wyszecki
(1960c), N. Mori (1961), Fleming and Wark (1965,
1966), and Chickering (1970).
(iv) Cutoff Filters. Filters that show a spectral
transmittance curve with zero (or near zero)
transmittance at shorter wavelength and high con-
stant (or near constant) transmittance at longer
 Optical Filters 39

wavelength are called cutoff filters. Examples of Keegan, 1938; Keegan and Gibson, 1944; Keegan,
spectral transmittance curves of such filters are 1960). Holmium is a rare earth, whereas didy-
displayed in Table 1(1.3.2) as ‘‘Type 1” filters. mium is a technical mixture of the rare earths
Cutoff filters are effectively used to reduce stray neodymium and praseodymium, with some
light in monochromators. They can also be used samarium added at times. The spectral transmit-
for abridged spectrophotometric measurements tance curves of these special glasses show a
(Mori, 1961). number of distinct minima and maxima at wave-
lengths characteristic of the glass. Figure 2(1.3.5)
(v) Filters for Wavelength Calibration of Spec- [also see ‘‘Type 10” filters in Table 1(1.3.2)] il-
trophotometers. The initial wavelength calibra- lustrates the spectral transmittance curves of di-
tion of a spectrophotometer is usually carried out dymium and holmium glasses.
by direct use of an electric discharge source emit- In an initial calibration, the location on the
ting spectral lines that are of suitable intensity wavelength scale of the minima and maxima can
and adequately spaced over the spectral range of be determined accurately. However, most absorp-
interest. Table 2(1.3.5) provides a short list of tion bands that can be used for this purpose
such lines. Although periodical checks of the consist of very narrow bands grouped closely
wavelength scale may be made in the same way, it together, and the wavelength of maximum ab-
is often more convenient to employ suit- sorption within such a group depends, to some
ably calibrated absorption filters for the purpose. degree, on the width of the spectral passband of
Didymium and holmium glasses have been the spectrophotometer. Generally, different initial
used for such checks for some time (Gibson and calibrations of the glasses are required for differ-
ent types of spectrophotometers.
In the ultraviolet region (200 to 300 nm),
benzene vapor has proven to be a useful medium
Table 2(1.3.5) Sources of Wavelength for wavelength calibrations (ASTM, 1966). In the
Suitable for Calibration of Monochroma- infrared region (2.0 to 15 pm), polystyrene or
tors (Meggers et al., 1961; Gibson, 1949) indene are used (Tables, 1961).
Helium Cadmium
Mercury Arc Discharge tube Discharge tube (vi) Filters for Checking the Photometric Scale
(in quartz) (in glass) (in glass)
of Spectrophotometers. Carefully calibrated
A (nm) A (nm) A (nm) neutral glass filters of different optical densities
253.7 318.8 298.1
are usually used to check the photometric scale of
27553 361.4 313.3 a spectrophotometer (Mavrodineanu, 1973). Their
289.4 363.4 326.1 spectral transmittance curves are approximately
296.7 370.5 340.4 horizontal in the visible spectrum as shown in
302.2 382.0 346.6 Table 1(1.3.2) as ‘‘Type 3” filters. Other glass
filters of somewhat greater spectral selectivity may
312.6 388.9 361.1
S132 396.5 467.8 also be used for this purpose if a careful calibra-
334.1 402.6 480.0 tion of their spectral transmittances is ensured
365.0 412.1 508.6 (Keegan et al., 1962). These filters are somewhat
365.5 414.4 643.8 inferior to strictly neutral filters, because devi-
ations from the true transmittance may arise from
366.3 438.8
causes, notably wavelength errors, unconnected
404.7 443.8
407.8 447.2 with the inaccuracy of the photometric scale. They
435.8 471.3 may, however, be used with advantage to detect
546.1 492.2 stray light, slit width, and wavelength errors (Gib-
son et al., 1934; Gibson and Balcom, 1947).
577.0 501.6
Liquid filters are also used as suitable stan-
579.1 504.8
623.4 587.6
dards for checking the photometric scale of spec-
671.6 667.8 trophotometers (Gibson, 1949; Haupt, 1952;
690.7 706.5 Menis and Schultz, 1971). These filters are quite
stable and precisely reproducible from specifica-
1014.0 728.1 tion. Table 3(1.3.5) contains the pertinent in-
1128.7 1083.0
formation for three such filters.
40 PHYSICAL DATA

1.0

0.8
SV
06

—DIDYMIUM
04
—HOLMIUM
TRANSMIT
TANCE
0.2

400 500 600 700 800

WAVELENGTH X(nm)
Fig. 2(1.3.5). Spectral transmittance curve of a holmium and a didymium glass.

(vii) Lovibond Glasses. Lovibond red, yellow, 1.3.7 Interference Filters

and blue glasses are used in conjunction with the
Lovibond Tintometer to measure the color of
light-transmitting objects such as lubricating oil,
sugar solutions, and beer, and of light-reflecting
materials such as oleomargarine (Judd and
Wyszecki, 1975). The coloring agents of Lovibond
glasses are gold, chromium, and cobalt, respec-
tively, and are confined to a thin layer of glass
flashed onto a clear glass support. The three
Lovibond scales, each represented by a series of
red, yellow, and blue glasses, respectively, are
calibrated in arbitrary units, but the three units
are related in such a way that for daylight il-
lumination a subtractive combination of one unit
of each of the red, the yellow, and the blue scales
results in an approximately neutral filter. Each
glass is identified by the number of unit glasses to
which it is equivalent. The Lovibond numeral of a
combination of two or more glasses of the same
kind is found by adding the numbers of units for
each individual glass of the combination. By
means of graphs, Lovibond numerals may be
converted to CIE color specifications and vice
versa (Judd et al., 1962).
1.3.6 Miscellaneous Absorption Filters
Occasionally, spectral transmittance data are re-
quired for noncolored glasses, quartz, and water
cells used as window materials. A large variety
exists with different spectral transmittance char-
acteristics and only a few typical examples can be
given here. Their spectral transmittance curves
are illustrated in Figure 1(1.3.6).
When a beam of radiant energy passes through an
assembly of layers of different materials, multiple
reflections occur within the assembly. If the dis-
tances between the boundaries of the layers are
sufficiently short, the reflected beams will be
coherent with one another and will interfere. This
effect, known_as optical interference and used in
Fabry—Peérot interferometers, has been utilized to
construct optical filters with diverse spectral
transmittance or spectral reflectance properties
(Heavens, 1955; 1960; Wolter, 1956; VaAsicék,
1960; Berning, 1963; Anders, 1967; Thelen, 1969;
Dobrowolski, 1978).
(i) Fabry-Perot Passband Filters. The essential
components of the simplest Fabry—Pérot inter-
ference filter are two partially reflecting thin-film
layers separated by a dielectric spacer layer, as
shown in Figure 1(1.3.7). Other more complex.
designs of Fabry—Pérot type interference filters
are shown in Figure 2(1.3.7).
The theory of Fabry—Pérot type interference
filters is, in essence, the same as that of low-order
Fabry—Pérot interferometers, as given, for exam-
ple by Born and Wolf (1959). Our presentation of
the important formulae follows essentially that
given by Dobrowolski (1978).
The transmittance, 7, of a simple Fabry—Pérot
filter [Figure 1(1.3.7)] is given by the relation
a TR [1(1.3.7)]
(1—R) + 4Rsin’s
 4]
6580 8°Sfr 3H Z0Z'0 06¢ 086'0 8°L99 2H $960 06S O8r'0 8°L99 2H 883°0 06S
tae eyeee Ek Feeceale NBIGEN eS wcrc one Hee deSette ia) ae
09S
£
696'0 lors 3H 1560
‘
09S
¢ af 3
Z68'0
F
ie Kaeo ay Ore €18°0 [obs 3H
0¢9°0 cret 9H 8r1°0 O9t 789°0 9°10$ 9H L€8'0 oss +660 9°10S 9H 6960 oss
$060 TEE 3H 9170 Ose
80L'0 9°16b 3H pLL'0 Ors 9660 9°16 3H ¢L6°0 Ors
ee E +06'0 8'Stp 3H 9IL'0 o¢s L660 8°Stp 3H 786'0 Os
ILL0 ccOt 8H t8h0 Ort L960 L’bOb 3H 8190 ozs 6660 L'v0b 3H L860 ozs
8650 L967 3H OIL'0 Of 0L9°0 O1S 166°0 O1S
6L7'0 9°€S7 3H 98'0 OZE 7660 OSL 989°0 OOS CSTD OSL 7660 00S
p10 Ott £660 Or cra 06r IL1°0 Or 9660 06r
000°! 00S 60L°0 00t £660 OL €£L'0 O8r €61°0 O€L L660 O8r
166°0 OZL 9SL'0 OLY 1720 OZL L660 OLY
= 686'0 OIL 68L'0 09r 9670 OIL L660 09r
L660; 06r €LE0 0672
OOL
a
L€3°0 OSr L670 OOL L660 OSr
i ace ieee abd 8860
086'0 OLY O81‘0 OLZ $860 069 L88°0 Orr 8re'0 069 L660 Orr
096'0 09 €€Z7'0 097 £860 O89 6760 OLP 9070 089 L660 OP
* 086'0 OL9 0S6'0 O@r 9960 OL9 966'0 O@r
0 8160 099 796'0 Ol zes'0 099 9660 OI
Ne Nice pat Ore
9160 0s9 ZL6°0 00r L650 0S9 $66'0 00r
Z88°0 Orr LOS‘0 Orz
eo Me as Of mo aa oe ee aR ne Bae aR
ae a gobi ue bL6'0 0z9 686'0 OLE ZLL'0 0z9 686'0 OLE
ct9°0 OlP OI? 760 019 1660 09¢ 918°0 O19 9860 09¢€
ZOr'0 OOF 007 6960 009 1660 Ose $S$3'0 009 6L6°0 OSE
(Ye (wu) 7 (Ye (uiu)yY (Y)é (wu) 7 (Ye (wu) 7 (Ye (wu) ¥ (ME (wu) ¥
99Ue}} IW YISUITIAR AA, 99UP}}IW YIBUITIARAA d9UR) IU YIBUITIARM 99URI}IU YIBUITIAR AA, 99UP}} IW YIBUITIAC AA 90UR}IUWI YIBUITOARM
-SUPIL -SUBI -SUPIL -SUBI I] -SUPI | -SUPIL
[BusoUT [Puso]Uy [BusojUy [PUIO}U] ]eusajuy [Buss Uy
POLGG
ainyesadua | “IOJEM Pe] [Sip SUIUIP}JUOS [[99 AP IWIS Ouse
pos] [stp JUSIOYNS ul 12 (oul dAl}PlOI poinsesf] “YOyy wu Ol JaAP| pinbry oinjesod wd [ “TOTEM Pe[HSip SUIUIJUODS [99 AB[IUWIS
JO}I] if oyeu O} JOJEM
apixoipAy wnis jul 0° 0001 oy eu Oj (poa][Hsip) JOE AN U.OF dAI} [Ol poinsesj] “Oryy wu Ol JOAP] pinbry
Aytyenb jud3 RAI jo (HOM 5 C8) SYS
-sejod jo 3 ¢°¢ BUIAJOSSIP Aq poasedaid aq ued 1911P| eemoney| (Seg | AuAess oyloads) plow diunj[ns [4 00001 DYPU O} (PI|[NSIP) JOIV AM
sy, ‘uoNjos spixoipAy winissejod |eusOoU ¢0'0 3 18P'rl (0°H9-"OS*(FHN)-"OSOD) jw 00! (Seg ] AwWavis sytoads) plow ounying
ur ("O1D*M) aeWOIYS WnIssej}od jo 41931]/3 OOPO'O “II a1ej[NS WNIUOWWE IPGOD ‘]] 3 000'0Z (O°HS-FOSND) areyns JaddoD *]
e
eee ee eS
e ..__ —e —
(ZS61 qdnepy ‘6P61 ‘UOSgIx)) siajaWIOJOYdo.NIadG Jo adURUIOPIag ay) BSuTypeYyD JO} S19ITLY pmbry (SECEDE AGEL
 42 PHYSICAL DATA

WINDOW GLASS (3.00mm)

CROWN GLASS

DENSE FLINT GLASS (486mm)

FUSED QUARTZ (2.69 mm)

DISTILLED WATER (in quartz

TRANSMITTANCE cell, Imm walls ,

water column |Omm)

0.2 0.3 0.4 0.5 0.6 0.7 LO J5- 20° 25ma83¢

WAVELENGTH (um)
Fig. 1(1.3.6). Spectral transmittance curves of miscellaneous absorption filters.

in which absorption and multiple reflections in thickness, and angle of reflection of the spacer,
the substrate are ignored. In the above relation, and ¢€,, €, are the phase changes that occur when
we have radiant energy of wavelength A is reflected by the
1/2 1/2 two reflector surfaces in contact with the surfaces
Le (T,T, ) ; and R= (R,R2) : of the spacer layer.
The maxima of the transmittance T, as given
[2(1.3.7)] by Eq. 1(1.3.7), occur at wavelengths
where 7,, R, and T,, R, are the transmittances
and reflectances of the first and second reflectors, 2ntcos o
aN a ith k=0,1,2,-
respectively, as determined from within the di- mPa e/a) |oo:
electric spacer medium. The quantity 5, a phase-
difference parameter, is defined by [4(1.3.7)]
s=
Qn1 Nicos te [3(1.3.7)]
These maxima have values determined from
—

with p alee”?
5 a:

where n, t, and ¢ are the refractive index, physical

T= (— = [5(1.3.7)]

SUBSTRATE aoa METAL REFLE CTOR

DIELECTRIC SPACER
SUBSTRATE peoee METAL REFLECTOR

Fig. 1(1.3.7). Essential components of the simplest form of Fabry—Pérot interference filter.
 Optical Filters 43

(LMM LTE ELE

Fig. 2(1.3.7). Fabry-Perot type interference filters of different designs. S = spacer; H =

dielectric reflector with high refractive index; L = dielectric reflector with low refractive index;
M = metal reflector.

The minima of the transmittance T occur at wave- by

lengths
di
2nt cos ; 1x3
=) ae Os ea Di 2
[9(1.3.7)]
[6(1.3.7)]
This function defines a Lorentz line shape. Its
and have the values
(1/10)-width (AA)o,; and its (1/100)-width
(AA) oo; are both related to its half-width (AA)9;

Taoee = aOR(TR)
as follows:
[7(1.3.7)]
(Ad Jo. = 3(AA Jos
For reflectors with reflectances of R > 0.7, the
half-width (AA); at A max, that is, the wavelength and
interval, centered at A,,,,, Within which the trans-
mittance exceeds half the maximum, is given ap- (AA Joo ~10(AA)o; [10(1.3.7)]
proximately by
The first interference filter ever made
= 2 (Geffcken, 1942; 1948) was of the simple Fabry—
ayes
R'/?_—-
oeee
eee
2antcos — A,4,(de/9A)
Pérot type shown in Figure 1(1.3.7) with the
reflecting thin-film layers made of silver.
Aluminum is another suitable metallic reflector.
[8(1.3.7)] For filters with metal reflectors, the mean phase
change e has a small but finite value at A,,,,, and
The quantity de/0A is the dispersion of the mean other wavelengths, and thus, the positions A ,,,, Of
phase change e. The values of R and de/0A in Eq. the transmittance maxima T,,,,,, aS given by Eq.
8(1.3.7) are taken at A = A jax: 4(1.3.7), are somewhat affected by e. However,
With small values of de/0A in the neighbor- the dispersion de/dA can often be neglected in
hood of A jax» the spectral transmittance T(A) of these cases, making the half-width (AA)5, de-
a single Fabry—Pérot filter is approximately given fined by Eq. 8(1.3.7), a function of only the
44 PHYSICAL DATA

reflectance, the position of the passband, and the
order k of interference; that is, for filters with
metal reflectors, we have
Aa CRD [11(1.3.7)]
(AA )os = akR'/?
The maximum transmittance T,,,, of filters with
metal reflectors is rather limited. Filters with rela-
tive half-width 100(AA)95/A max Of 1 to 8% can
have maximum transmittances as high as 40%,
but for narrower relative half-widths, values of
T.,ax Of only 20% may be achieved. Further limi-
tations occur when A ,,,, falls into the short wave-
length region of the spectrum.
The metal reflectors in a Fabry—Pérot inter-
ference filter can be replaced by two all-dielectric
“quarter-wave stacks,” as shown on the left of
Figure 2(1.3.7). Each reflecting stack consists of a
set of quarter-wavelength thick dielectric layers of
alternating high (H) and low (L) refractive in-
dices. The phase change e caused by such a stack
at the reflector /spacer boundary is zero at A pax,
and thus the wavelength position of the maximum
transmittance is not ‘affected by «. However, the
dispersion of de/dA of the phase change is finite
and its value depends on the dielectric materials
used in the stack. The half-width (AA)5 of the
passband of the filter can be affected consider-
ably by the nonvanishing dispersion. In fact,
suitable dielectric materials can be chosen to ob-
tain very narrow relative half-widths of between
0.1 and 5% in the visible part of the spectrum.
The maximum transmittance of an all-dielec-
tric Fabry—Pérot interference filter can be quite
high. In the visible spectrum, values of 80% are
fairly readily obtained with relative half-widths of
1%. The maximum transmittance reduces to lower
values when A, approaches 0.2 or 20 wm.
In Figure 3(1.3.7), the passband of a simple
Fabry—Pérot interference filter with metal reflec-
tors is compared with that of a Fabry—Pérot filter
with all-dielectric reflecting stacks.
The reflecting stacks in a Fabry—Pérot inter-
ference filter are sometimes made of both dielec-
tric and metal layers as shown on the right of
Figure 2(1.3.7). The transmission properties of
such designs are intermediate to those using only
metal reflectors or all-dielectric reflectors.
(ii) Square-Top Miulticavity Passband Filters.
Interference filters can be made whose spectral
passband curves have steeper slopes than those
found in a Fabry—Pérot passband filter. This
feature often leads to a “square”-shaped top of
the passband. The layer structure of such filters
consists of a Fabry—Pérot-type structure repeated
one or more times. Figure 4(1.3.7) shows two
examples, each consisting of two Fabry—Peérot
type structures in series. The half-wavelength
spacer layers (S) are called cavities; thus the name
square-top multicavity passband filter.
Except for the different shape of the passband
of such filters, the other characteristics, such as
achievable relative half-width and peak transmit-
tance, are generally similar to those of the Fabry—
Pérot type filter.

0.8 ALL-DIELECTRIC

SPACER i REFLECTING

Pd
LAYERS
0.6

METAL
0.4
SPACER oh REFLECTING
TRANSMITTANCE
LAYERS
0.2

400 500 600

WAVELENGTH (nm)
Fig. 3(1.3.7). Spectral passbands of two different designs of Fabry-Perot type interference
filters (Polster, 1952): solid curve = passband of all-dielectric filter consisting of 15 layers;
dashed curve = passband of metal filter with dielectric spacer.
 Optical Filters 45

WMA hh bb

Fig. 4(1.3.7). Examples of different designs of square-top multicavity passband interference

filters. S = substrate; H = dielectric reflector with high refractive index; L = dielectric reflector
with low refractive index; M = metal reflector.

(iii) Sidebands of Passband Interference Filters.
The complete spectral transmittance curve of a
passband interference filter represents a periodic
function consisting of a series of clearly separated
passbands across the spectrum. The passbands
correspond to the orders (k = 0,1,2,---) of in-
terference that are generated by the filter. The
formulae describing the Fabry—Pérot-type inter-
ference filter given as Eqs. 1(1.3.7) to 11(1.3.7)
describe the spectral transmittance characteristics
in full. Figure 5(1.3.7) illustrates a passband inter-
ference filter with its second, third, and fourth-
order passbands in the visible and ultraviolet
spectrum. In the illustrated case, the filter was
designed to have its third-order passband located
at A wax = 475 nm with a half-width of (AA)5 =
10 nm (approximately). All other passbands are
considered sidebands and normally an undesira-
ble feature of the interference filter. To eliminate
the sidebands, blocking filters are usually ce-
mented to the interference filter. Blocking filters
are either absorption-type filters (see Section 1.3.5)

0.6 T Tee
Thre 3rd Order
0.5

re 4th Order
by! Pa 2nd Order
5
= 03 \ (AXo5
= 0.5 Twax
ec
re wes, 8ia
=
=

Or

il

300 400 Dae 600 700

Wavelength A (nm)
Fig. 5(1.3.7). Complete spectral transmittance curve of a Fabry-Perot type interference filter
in ultraviolet and visible spectrum. The passband corresponding to the third-order interference
is, in the illustrated case, the desired passband with its peak at 2
,,,,, = 475 nm and a half-width
of (AA) ,;=10 nm. The other two passbands shown in the spectral range correspond to
second- and fourth-order inference, respectively, and are undesirable sidebands.
46 PHYSICAL DATA

0.8 0.8

0.6 0.6

0.4 0.4

TANCE
TRANSMIT TRANSMITTANCE

0.2 0.2

500 550 600 500 550 600

WAVELENGTH (nm) WAVELENGTH _ (nm)

0.8 0.8

0.6 0.6

0.4 0.4

TRANSMITTANCE TRANSMITTANCE

0.2 0.2

500 550 600 500 550 600

WAVELENGTH ~ )(nm) WAVELENGTH = A(nm)

Fig. 6(1.3.7). Examples of spectral transmittances of fully blocked interference filters; top:
All-dielectric Fabry—Pérot filters; bottom: All-dielectric square-top multicavity filters.

or other interference filters with spectral transmit-
tance characteristics that pass only the desired
passband of the given interference filter, and not
any of the sidebands. The blocking cannot avoid
some loss of the peak transmittance of the desired
passband. Figure 6(1.3.7) shows a few examples
of the passbands of blocked interference filters of
different designs commercially available.
(iv) Angular Properties of Interference Filters.
When the angle of incidence (#) of the radiant
energy is increased, the transmittance maximum
(Tmax) Of a typical passband interference filter
moves toward shorter wavelengths, as indicated
by Eq. 4(1.3.7), and T,,,, decreases. For moderate
to large angles of incidence (¢ > 20°), the pass-
band becomes asymmetric and eventually splits
into two perpendicularly polarized components.
Figure 7(1.3.7) illustrates the angular properties
of a typical interference filter for angles of inci-
dence up to 20°.
The angular properties of interference filters
are considerably more pronounced than those of
absorption filters and can cause serious calibra-
tion problems when they are used in optical in-
struments. In most cases, the incident beam does
not consist of strictly parallel bundles of rays and
a divergence from parallelism can have a mea-
surable effect on the transmittance characteristics
of the interference filter. In high-accuracy work, it
is advantageous to calibrate the interference filter
in situ; that is, in the instrument in which it is
actually used.
1.3.8 Interference Filter Wedges
An interference filter wedge is a passband inter-
ference filter in which the thicknesses of all the
 O27 peer

0.6

0.5

0.4

0.3

Transmittance
Or2

0.1

470 480 490 500 510

Wavelength d (nm)

Fig. 7(1.3.7). Spectral transmittance of a typical passband interference filter as a function of

angle of incidence of radiant energy (after Dobrowolski, 1978).

1000

Spectral Passband Wedge Dispersion

at 550 nm 900
E
- iS
cob)
Oo
800 <
So
~ =
Ss
4
es
=
700 2
:
-_

c
c®
S 600 a
2 5
=
500

500 520 540, / 560 580 600 0 50 100

Xmax* 550nm Slit Position (mm)

/ Wavelength Scale (nm)

400 500 /'600 700 800 900 1000

ie) 20 40 ) 60 80 100 120 140 160 180 200

siti Wedge Scale (mm)
position

Fig. 1(1.3.8). Typical interference filter wedge (straight version) with its performance char-
acteristics. The wedge transmits narrow spectral passbands of somewhat varying peak transmit-
tance 7,,,, and half-width (AA) 5 when a slit of constant width (up to 2 mm) moves across the
wedge.

47
48 PHYSICAL DATA

layers vary in proportion across the surface of the
substrate, causing the transmittance maximum of
the passband to vary its wavelength position in a
corresponding way. Interference filter wedges are
commercially available in versions in which the
wavelength of the transmittance maximum varies
either along a straight line or a circle. The circular
wedge has practical advantages when it is used as
a robust rapid-scan monochromator of moderate
spectral resolution and purity.
The maximum transmittances and half-widths
achieved for fully blocked interference filter
wedges are generally somewhat less than for indi-
vidual passband filters. Figure 1(1.3.8) illustrates
the performance characteristics of a typical inter-
ference filter wedge of the straight version.
1.3.9 Beamsplitters
A beamsplitter divides the incident beam of
radiant energy into two beams emerging from the
beamsplitter in two different directions. Figure
1(1.3.9) illustrates four of the most common de-
signs of beamsplitters.
The transmission and reflection properties of a
beamsplitter depend, in general, upon the state of
polarization of the incident beam of radiant en-
ergy. However, in achromatic beamsplitters, that
is, beamsplitters in which the relative spectral
radiant power distributions of the two emerging
beams remain approximately the same, the quan-
tity
1
a(R a Rives pec
is approximately equal to the reflectance R at
normal incidence, 6 = 0°. In the above expres-
sion, R, is the reflectance (at the angle of inci-
dence @ = 45°) of a plane-polarized beam with its
plane of vibration so oriented to the beamsplit-
ting surface that its reflection is most facilitated,
and R, is the reflectance (at 9 = 45°) of a plane-
polarized beam with its plane of vibration so
oriented that its reflection is most hindered. The
reflectance R of absorbing, uncemented beam-
splitters depends also on the angle of incidence 6
in accordance with Fresnel’s law of reflection (see,
for example, Judd and Wyszecki, 1975).
Figure 2(1.3.9) illustrates the spectral transmit-
tance and reflectance properties of typical achro-
matic beamsplitters. More details on beamsplitters
are given, for example, by Anders (1967).
Beamsplitters can also be made to divide the
incident beam into two component beams of dif-
ferent spectral compositions. The coatings used in
such beamsplitters are essentially coatings of

Substrate Plate Substrate Cube

(1)

Coating Coating

Substrate Plates

Pellicle

PN
(3) Coating (4) Mount

Fig. 1(1.3.9). Four common beamsplitters. (1) Beamsplitter consisting of a single plane-paral-
lel glass plate with a dielectric or metal (silver) coating. Transmitted beam is laterally displaced
from incident beam. (2) Beamsplitter (glass) cube with coating deposited in the diagonal plane
of the cube. Transmitted beam is not displaced laterally. (3) Beamsplitter in which coating is
sandwiched between two plane-parallel glass plates. Transmitted beam is laterally displaced but
both the transmitted and reflected beam traverse the beamsplitter in paths of identical lengths.
(4) Pellicle beamsplitter. Coating is deposited on pellicle of (approximately) 1-pm thickness.
The device comes in a special mount and may introduce an interference pattern into its spectral
transmittance and reflectance characteristics.
 Optical Filters 49

(T)
(R),
Transmittance
Reflectance
400 500 600 700
Wavelength d (nm)

Inconel Beam - Splitting Plate

(T
)Reflectance(R),
Transmittance
400 500 600 700
Wavelength d (nm)
Fig. 2(1.3.9). Spectral reflectance and spectral transmittance curves of two typical achromatic _
beamsplitters.

‘cutoff’ interference filters designed for use at 1.3.10 Sheet Polarizers

45° incidence. While one part of the spectrum is
Typical sheet polarizers, such as Polaroid films
transmitted, the other part is reflected. Only small
(Land, 1951) and Zeiss polarizers (Haase, 1961),
losses of radiant energy occur due to absorption.
consist of thin sheets of plastic film packed with
Figure 3(1.3.9) illustrates the spectral transmit-
ultramicroscopic dichroic crystals with their opti-
tance of a beamsplitter of this kind, sometimes
cal axes all parallel and in the plane of the sheet.
called a dichroic or color-selective beamsplitter.
The transmittance 7 of a sheet polarizer for a
100% linearly polarized beam of incident radiant
energy is given by the following formula:
1.0

0.8 Tu=sloet (1; —“T, )cos7é [1(1:3.10)]

@
oO
c
506
€ where 7, and 7; are the principal transmittances;
c 04 that is, the maximum and minimum transmit-
12)
Ss
tances of the polarizer, respectively. In general,
0.2
T, > T,. The principal transmittances occur at
the ‘‘parallel” and ‘‘crossed” positions of the
400 500. gat. 600 700 polarizer, with respect to the plane of vibration of
Wavelength A (nm) the incident beam. The quantity @ is the angle of
Fig. 3(1.3.9). Spectral transmittance of a dio- rotation measured from the parallel position (6 =
chroic or color-selective beamsplitter. 0) of the polarizer.
50 PHYSICAL DATA

The transmittance T of a polarizer in un- another, is given by

polarized light is simply
T= 5(T; +7) — [201.3.10)]
It follows that a perfect polarizer would transmit
50% of the unpolarized incident beam.
When two identical polarizers are combined
such that their principal transmission directions
are parallel, the transmittance of the combination
is given by
l
5
Time (Peet) [3(1.3.10)]
When, on the other hand, the two polarizers
are combined in crossed positions, the transmit-
tance of the combination is given by
T,=T,T, [4(1.3.10)]
The transmittance of 7, of the two polarizers
combined such that the directions of principal
transmittance are inclined at an angle @ to one
a 5(7? + T?)cos?6@ + T;T> sin’6
[5(1.3.10)]
The polarizing properties of a polarizer are
usually characterized either by the quantity
P $.0le~
Te 5
tp
[6(1.3.10)]
which is called the degree of polarization, or by
the quantity
T.
= = [7(1.3.10)]
1
which is called the extinction ratio.
Figure 1(1.3.10) illustrates the principal trans-
mittance 7, and extinction ratio p, of two
commercial sheet polarizers as a function of wave-
length. Further details on sheet polarizers and
polarizers of different design and materials are
given, for example, by Bennett and Bennett (1978).

1.0
ey

a eee
ce)

—E 06
”
Cc
o

EK 04 HN-22
re)
Qa
So 0.2
i=
ie
= ult
400 500 600 700
Wavelength A (nm)
107!

ae
AS
(eo)
= -3
RS 10 HN-38

8 10-4
a)
—
it ne A aa oh ea
-6
A 400 500 600 700
Wavelength (nm)
Fig. 1(1.3.10). Extinction ratio p, = 7, — T, and principal transmittance 7, of two commercial
sheet polarizers (Polaroid HN-22 and HN-38) as functions of wavelength.

1.3.11 Miscellaneous Optical Filters
There are several other types of optical filters that
are not described in the previous sections. These
are considered of minor importance in applica-
tions to color science and we will thus confine our
description to a brief listing only. A more detailed
and very informative account of these filters has
been given by Dobrowolski (1978).
(i) Polarization Filters. A polarization filter
consists of a series of birefringent plates that are
separated by polarizers (Lyot, 1933; Oehman,
1938; Evans, 1949).
The spectrum of the radiant energy passing
through such a device shows a series of dark
bands and for this reason is usually referred to as
a “‘channel spectrum.” Filters have been desig-
ned, primarily for the study of solar prominences,
that consist of a series of such combinations
where the birefringent plates have thicknesses in
the relation 1:2:4:8, and so on. The passband
transmitted by the complete assembly is de-
termined largely by the thickness of the thickest
plate; the spacing between the bands is given by
the spacing between the maxima of transmittance
of the thinnest plate.
Sheet polarizers are a special kind of polariza-
tion filters and of some importance to color scien-
tists. They have been described in Section 1.3.10.
(ii) Christiansen Filters. A Christiansen filter
(Christiansen, 1885; Weigert and Staude, 1927;
Elvegard et al., 1929) consists of a transparent cell
containing a closely packed coarse powder of a
homogeneous isotropic transparent substance sus-
pended in a liquid. Because of their different
dispersions, the powder and the liquid have the
same refractive index at one wavelength only, d..
Thus, radiant flux of a narrow passband at and
around A. will be transmitted by the cell without
deviation. Radiant flux of all other wavelengths
will be scattered. At the back of the cell, a center
spot of a rather pure spectral color (A) is ob-
served surrounded by a halo of complementary
color. With proper optics, the center spot can be
isolated, and a narrow passband filter of remark-
able purity is obtained. Usually the refractive
index of the liquid changes rapidly with tempera-
ture, whereas the refractive index of the powder
varies only slightly. Hence the wavelength A.
varies with temperature. With a combination of
glass powder and ethyl salicylate, the wavelength
\.. varies from approximately 430 to 570 nm for a
temperature range of 22 to 57°C (Denmark and
Reflecting Materials 51
Cady, 1935). In this instance, the half-width of
the passband is approximately 1 nm at around
550 nm. The peak transmittance at A. reaches
40%. Other powder-liquid combinations provide
different ranges for A. and different half-widths
and peak transmittances. It is possible to obtain
Christiansen filters for the ultraviolet, the visible,
as well as the infrared regions of the spectrum
(von Fragstein, 1933; Mohler and Loofbourow,
1952; Auerbach, 1957; Dobrowolski, 1978).
(iii) Gratings, Metallic Grids, and Meshes. Dif-
fraction gratings, metallic grids, and metallic
meshes, suitably designed, can produce spectral
transmittance or spectral reflectance curves of the
cutoff as well as the passband type. However,
these devices are usually employed in filtering the
far infrared region of the spectrum, typically in
the range of 10 to 400 pm.
1.4 REFLECTING MATERIALS
Reflection is a physical process in which mono-
chromatic radiant energy incident on a material
(the medium) is at least partially returned by the
material without change of wavelength.
The reflection can be regular, diffuse, or mixed
(CIB 1O27),
Regular reflection (or specular, or mirror reflec-
tion) follows the laws of optical reflection without
diffusion (scattering) as exemplified by mirrors.
Diffuse reflection is reflection, on a macro-
scopic scale, without any regular reflection. The
diffuse reflection is called isotropic when the spa-
tial distribution of the reflected radiant energy,
measured in terms of radiance [see Table I(1.1)],
is the same in all directions of the hemisphere of
incidence.
Mixed reflection is partly regular and partly
diffuse reflection.
Two special concepts are of interest: retro-
reflection and gloss.
Retro-reflection is said to occur when the
radiant energy is returned in directions close to
the direction from which it came, and it is main-
tained over wide variations of the direction of the
incident radiant energy. Virtually all materials
that show mixed reflection also show retro-reflec-
tion to some extent. However, only those that
retro-reflect most of the incident radiant energy
are called retro-reflectors or retro-reflective materi-
als. Special measurement techniques are required
to assess their properties (Moerman, 1977).
52 PHYSICAL DATA

Gloss is defined as the property of a material
surface that involves mixed reflection and is re-
sponsible for a mirrorlike appearance of the
surface (Hunter, 1975). The term glossiness is used
to describe the visual appearance of a surface
attributable to gloss. There are five kinds of glos-
siness, each having its characteristic scale of de-
grees of glossiness: (a) specular; (b) sheen; (c)
contrast; (d) distinctness of image; (e) absence of
bloom (Hunter, 1975; Judd and Wyszecki, 1975).
For details of the measurement of the different
kinds of glossiness, see Hunter (1975), and with
regard to specular gloss see also Budde (1980).
The basic quantity that characterizes the pro-
cess of reflection is called reflectance. It is defined
as the ratio of the reflected radiant flux (or power)
to incident radian flux (or power), and has been
given the symbol p. Spectral reflectance is thus
defined by
p(A) = 5Py (114)
where P, is the spectral concentration of the
radiant power reflected by the medium, and Pp), is
the spectral concentration of the radiant power
incident on the medium.
When mixed reflection occurs, the reflectance
can be expressed as the sum of two parts: regular
reflectance (p,) and diffuse reflectance (p,), that
is,
p= peepi [2(1.4)]
The reflectance of materials with only regular
reflection can be expressed by Fresnel’s equations
For the case considered here, Fresnel’s equa-
tions are as follows:
2
COS € —[(ns/n,y ie sin?e]
Pier te Tt ln ow, hue
SOnomy a
cos e + [(n3/m1)° = sin’e|
sin?”
(n,/n,) cos e +[(n3/n,)°
[3(1.4)]
where p, is the reflectance of the nonabsorbing
nonmetallic medium for a plane-polarized beam
of radiant energy so oriented that the plane of
vibration of the electric vector is parallel to the
plane of incidence defined by the central ray of
the beam and the normal to the surface of the
medium; p, is the reflectance for a _plane-
polarized beam whose plane of vibration is per-
pendicular to the plane of incidence. The symbols
p,, and p, are sometimes replaced by p, and p,,
respectively. Figure 1(1.4) illustrates the depen-
dence of Fresnel reflectance on angle of incidence
on a surface for which the refractive index ratio is
1.5 (representative of an air—glass boundary).
For unpolarized radiant energy, the reflectance
is simply
p=
p\ + p
[4(1.4)]
Figure 2(1.4) illustrates the dependence of Fresnel
reflectance on angle of incidence for various val-
ues of refractive index ratio.
For normal incidence (¢ = 0), the reflectance,
defined by Eq. 3(1.4), becomes

(58)
(see, for example, Bennett and Bennett, 1978). A
case of particular interest concerns nonmetallic
media, such as transparent glass, where one wishes P No + Nn, [5(1.4)]
to determine the loss of incident radiant energy
due to regular reflection from the front and back
With the medium in air, one can put n, ~ 1.0;
surfaces of the medium. As there is no diffuse
thus
reflection, the regular reflectance p, of such
surfaces is equal to the reflectance p. The reflec-
tance, also called Fresnel reflectance, depends on = Lied ‘

the angle of incidence e on the medium, the index ey Mo tl [6(1.4)]

of refraction n, of the medium, the index of
refraction n, of the medium (usually air) in con- This equation compares with Eq. 7(1.3.1) used
tact with the medium under test, and the state of in the development of simple formulae for the
polarization of the incident radiant energy. transmittance of absorption filters.
 0.11

0.10 1.0
0.09 °°
0.08 0.8

0.07 0.7
S 0.06 0.6
8
8 0.05 0.5
@®
e
0.04 0.4

0.03 0.3

0.02 0.2
0.01 0.1
[Ss A

10° 20° 30° 40° 50° 60° 70° 80° 90°

Angle of incidence

Fig. 1(1.4). Fresnel reflectance as a function of angle of incidence € on a boundary for which
the refractive index ratio is 1.5 (representative of an air-to-glass boundary). Note that the
reflectance, p, for unpolarized radiant energy is the average of the reflectances for radiant
energy plane-polarized parallel (||) to and perpendicular (p + ) to the plane of incidence (Judd
and Wyszecki, 1975).

Angle of incidence
Fig. 2(1.4). Fresnel reflectance for unpolarized radiant energy as a function of the angle of
incidence ¢ for various values of refractive index ratio (Judd and Wyszecki, 1975).

ae
54 PHYSICAL DATA
Incident Radiant Energy
di VeSls or.
Hemispherical
Reflected Radiant Energy
Fig. 3(1.4). (a) Examples of hemispherical, conical, and unidirectional incidence of radiant
energy. (b) Examples of hemispherical, conical, and unidirectional collection and viewing of
reflected radiant energy (Judd, 1967).
In practice, the need arises for making mea-
surements of the reflecting properties of materials
under different illuminating and viewing condi-
tions; that is, different angular (or geometrical)
configurations of the incident radiant energy and
of the reflected radiant energy. As illustrated in
Figure 3(1.4), the illuminating conditions may be
described as hemispherical, conical, or unidirec-
tional; and similarly, the viewing conditions may
be described as hemispherical, conical, or uni-
directional. This classification leads to nine dif-
ferent illuminating conditions (Judd, 1967a).
However, most practical needs can be satisfied by
choosing just one of these possible conditions.
The quantity measured under any of these
conditions is given the generic term reflectance
factor, denoted by 8. Its definition is given as
follows (CIE, 1977):
The reflectance factor of an element of a given
surface is the ratio of the radiant flux reflected in
the directions delimited by a given cone, with
apex at the given surface element, to the radiant
flux reflected in the same directions by the perfect
reflecting diffuser identically illuminated.
In accordance with this definition, the re-
flectance factor B becomes identical to the reflec-
tance p, defined in Eq. 1(1.4), when the illuminat-
ing and viewing conditions are either hemispheri-
cal—hemispherical or directional—hemispherical.
When the conditions are hemispherical—
unidirectional, conical—unidirectional, or
directional—unidirectional (bidirectional), the re-
flectance factor B is usually given the name radi-
ance factor, in accordance with the radiometric
quantity radiance that is measured for both the
| |
as — Sample
|
Conical Unidirectional
1
|'
‘ Ss
y =.
be
‘
‘ *
x
hy eee
given surface element and the perfect reflecting
diffuser.
Hemispherical illumination and hemispherical
collection are usually accomplished by means of
an integrating sphere (Ulbricht, 1920; Walsh, 1958;
Budde, 1958a,b; CIE, 1977). An integrating
sphere is a hollow sphere coated inside with a
spectrally nonselective (white) material (e.g.,
barium sulphate paint) of high diffuse reflectance.
The sphere wall has apertures of appropriate size
and location to admit a beam of radiant energy
into the sphere, to serve as ports for mounting the
test sample and a reference sample of known
reflecting properties, and to permit viewing by
means of a detector of the test sample, reference
sample, and/or a portion of the inside wall.
The CIE (1971) recommends four specific il-
luminating and viewing conditions as standard
conditions. These are described in Section 3.3.7
and illustrated in Figure 1(3.3.7). Further notes
on the definition and measurement of spectral
reflectance factor are given in Section 3.12.2. The
perfect reflecting diffuser as the primary standard
of reflectance is discussed in Section 3.3.6. Trans-
fer standards and working standards of reflec-
tance are discussed in Section 1.4.2.
Reflecting materials that scatter radiant en-
ergy, such as those consisting of dyes and pig-
ments suspended in an appropriate substrate, are
of particular interest to industrial colorimetrists
uni- when the objective is to make these materials of
specified color. Colorant formulation is practiced
widely and theories exist to aid in this practice.
Section 3.9 reviews the essential aspects of col-
orant formulation.
 Reflecting Materials 55

Fo 7 P, 1.4.1 Front-Surface Mirrors

The majority of front-surface mirrors used in

oun \ee [ees — Sample holder precision optics are produced by evaporating
with apertures metals onto glass or quartz in a high vacuum. The
spectral reflectance of a good vacuum-deposited
metal coating is always higher than that of an
A
electroplated or a polished surface of the metal.
Auxiliary mirror i
in position The most widely used metal for mirrors is
aluminum (Al), which shows a high reflectance
B
Po Po throughout the ultraviolet, visible, and infrared,
but silver (Ag), gold (Au), copper (Cu), and
rhodium (Rh) are also frequently used. Figure
Sample holder 1(1.4.1) shows the spectral reflectance curves of
with apertures freshly deposited coatings of Al, Ag, Au, Cu, and
Mirror under test Rh over the spectral region from approximately
p=(P,/P,) 2 0.2 to 5.0 wm (Hass, 1955). These curves refer to
coatings prepared under optimal conditions. Many
Fig. 4(1.4). Method of measuring the reflectance
factors, such as speed of deposition, pressure
of a mirror.
during evaporation, temperature of the substrate,
and aging in air, can significantly influence the
optical properties, particularly the spectral reflec-
Materials with only regular reflectance, such as tance, of the coating.
mirrors, are often measured by a procedure il- For many applications, the metal surfaces are
lustrated in Figure 4(1.4) (Strong et al., 1944). In protected by hard and transparent films. Mag-
the first step, top of the figure, the radiant flux P, nesium fluoride (MgF,) is usually used, but for
is measured which is proportional to the incident aluminum mirrors, a film of silicon monoxide
radiant flux Pp. In the second step, bottom of the (SiO) has been found especially suitable. The
figure, the auxiliary mirror is moved from posi- protective coatings change to some degree the
tion A to B and the mirror under test is placed at spectral reflectance characteristics of the mirror.
the sample holder. The radiant flux P, is propor- By the use of suitable thicknesses of the protec-
tional to the radiant flux P) and to the square of tive coatings, the reflectance may be increased
the reflectance p of the mirror under test. From over parts of the spectrum. A significant increase
the two steps of measurement, it follows that in reflectance in the visible is obtained for Al-mir-
rors coated with superposed quarter-wave layers
of Al,O, + TiO, or MgF, + CeO, (Hass, 1955)
In an optical instrument that includes a light
[7(1.4)] source, heat is often an undesirable factor. Heat
reflectors and cold mirrors can sometimes be em-
ployed effectively to reduce the problem. Heat
The concepts and terms that are used when the reflectors, or hot mirrors, are special long-wave-
radiance factors of fluorescent materials are mea- length cutoff filters, with a cutoff at about 700
sured are discussed in Section 3.12.3. It suffices nm, that transmit the visible part of the spectrum.
here to note that for fluorescent materials, the Cold mirrors reflect the visible part of the spec-
radiance factor is the sum of two radiance factors, trum and transmit the infrared (Dobrowolski,
B; and B,, which are due to reflection only and 1978). Figure 2(1.4.1) illustrates the spectral
due to fluorescence only, respectively: transmittance curves of typical hot and cold mir-
rors.
Bo Boag, [8(1.4)]
1.4.2 White Reflectance Standards

Of the many (nonfluorescent) reflecting White surfaces are of special interest to the pho-
materials that may interest the colorimetrist, spec- tometrist and colorimetrist as they are used as
tral reflectance curves of only a few can be given reference surfaces in reflectance measurements.
here. For many years, smoked magnesium oxide (MgO)
56 PHYSICAL DATA

REFLECTANCE

1. eae SSS rn
WAVELENGTH A (um)
Fig. 1(1.4.1). Spectral reflectance curves of freshly deposited coatings of aluminum (AI),
silver (Ag), gold (Au), copper (Cu), and rhodium (Rh) for use as front-surface mirrors (from
Haas, 1955).

was adopted as the white standard (i.e., standard
of reflectance), and its spectral reflectance was set
equal to 1.0 for all wavelengths within the visible
spectrum. Now the CIE recommends that reflec-
tance measurements be related to the primary
standard of reflectance defined as a perfectly
reflecting, perfectly diffusing surface, as described
in Section 3.3.6. When an MgO surface is
measured in absolute units of reflectance, it is
found that its spectral reflectance comes close to
that of the perfect reflecting diffuser but does not
coincide with it exactly. In the visible spectrum,

Hot Mirror Cold Mirror

Oo‘ SS wl
i.

Transmittance

0.4
ae 0.6
Se,
0.8 1.0 Pre
Wavelength d (um)
Fig. 2(1.4.1). Spectral transmittance curves of typical hot and cold mirrors used for heat
control in optical instruments.
 Reflecting Materials x

1.00

SMOKED

MgO

0.98

W
Oo
ea i

O
W
=)
w
uJ
a

5 aa

; —

:
0.90 a
0.2 0.4 0.6 0.8 FOsehO 1.5 ae) 2D
WAVELENGTH A (ym)
Fig. 1(1.4.2). Absolute spectral reflectance curve of a well prepared surface of smoked
magnesium oxide (Sanders and Middleton, 1951; 1953).

reflectances of up to 0.98 may be reached. An ing properties:

average spectral reflectance of a good surface of
smoked MgO for the spectral range from ap-
proximately 0.2 to 2.4 wm is given in Figure
1(1.4.2) (Middleton and Sanders, 1951; Sanders
and Middleton, 1953). The reflectance values de-
pend to some extent on the method of preparing
the MgO surface, its thickness, age, and pre-ex-
posure to ultraviolet radiant energy, as well as on
possible impurities in the magnesium used. These
difficulties are considered highly undesirable in
precision measurements and, for a number of
years, a considerable effort has been made by
several workers to find materials that would be
more suitable as secondary standards of reflec-
tance. The CIE (1979; see also Erb and Budde,
1979; Grum and Becherer, 1979) has issued a
detailed review of the publications on the proper-
ties and reflectances of such materials. In particu-
lar, the CIE review report deals with: (a) smoked
magnesium oxide; (b) pressed powder of mag-
nesium oxide; (c) pressed powder of barium
sulfate; and (d) various glasses, tiles, and plastics.
A material is considered suitable as a sec-
ondary standard of reflectance if it has the follow-
(1) Mechanically robust surface; transport-
able.
(1) Stable against common ambient condi-
tions.
(111) Uniform and flat surface.
(iv) Easy to clean.
(v) Opaque (no optical interference with the
background of the specimen; no edge-losses, that
is, lateral diffusion of radiant energy penetrating
specimen is small).
(vi) Capable of being prepared with a matte or
a polished surface.
(vii) Near uniform spectral reflectance across
ultraviolet and visible spectrum.
(viii) No luminescence excited by radiant energy
of wavelengths greater than 300 nm.
There is some leeway as to the extent that any
or all of the above conditions must be met by a
particular material to be used as a secondary
standard of reflectance for a particular applica-
tion. Considerable judgment is required of the
worker who has to make the selection.
58 PHYSICAL DATA

BoS0O,4
(pressed powder)
Ces
BaSO,4
(paint)
@®
oO
Cc
ie)
—_—
(©)
eb)
=
®
ac (pressed — \\—/-_-
powder) ‘\\j{,’
AS Vitrolite Glass~
‘ ‘
mt

=
(eo) BaSO4q
w
A
@}
(paint)
<I

0.80
“300 400 500 600 700 g0008 10 12 14 16 18 20 22 2.4 26
Wavelength \(nm) Wavelength (um)
Fig. 2(1.4.2). Absolute spectral reflectance curves of different materials used as transfer or
working standards of reflectance (private communication from F. Grum, 1981).

Two classes of application are distinguished.
In the first class, the actual spectral reflectance
values are of primary importance, particularly
when the material is to be used as a transfer
standard; that is, its known spectral reflectance
values are used to bring the calibrations of differ-
ent instruments in line with one another.
In the second class of application, the material
is an integral part of a particular instrument and
used in it as a working standard. Its calibration is
usually ascertained by means of a transfer stan-
dard, and its spectral reflectances are considered
valid only in conjunction with the particular in-
strument and its peculiar measurement geome-
tries.
In order to meet the requirements for a trans-
fer standard, the material should approximate the
properties of the perfect reflecting diffuser as
closely as possible; that is, it should be matte,
highly opaque, and have high spectral reflectance
values everywhere in the spectral range of inter-
est. The CIE Committee on Materials (TC-2.3)
names the following materials suitable as transfer
standards:
1. Pressed tablets of barium sulfate powder.
2. Pressed tablets of Halon (an organic fluorin-
ated polymer of microcrystalline structure formed
under high pressure).
3. Matte surfaces of opal glass MS 14 and MS
20 (a Russian development of glasses doped with
lead fluoride and other compounds).
Materials suitable as working standards are the
following:
1. Vitrolite glass (a special opal glass of Ameri-
can origin).
2. Ceramic plates.
3. White enamel plates.
4. Polished surfaces of opal glass MS 14 and MS
20 (see transfer standards No. 3).
Figure 2(1.4.2) shows typical spectral reflec-
tance curves obtained for some of the ‘‘white”
surfaces used as either transfer or working stan-
dards of reflectance.
The measurement of absolute reflectance
values (i.e., the reflectance of a given material
relative to that of the primary standard of the
reflectance, the perfect reflecting diffuser) re-
quires special techniques. The interested reader is .
referred to the publications of the CIE (1979) and
Budde (1976).
1.4.3 Colored Reflectance Standards
Of the many colored surfaces of special interest to
the colorimetrist are a set of ceramic tiles avail-
able from the National Physical Laboratory, Ted-
dington, England (Clarke and Samways, 1968).
This set of tiles consists of 12 colored and gray
specimens suitable as reference surfaces for
checking routinely the performance of colorime-
ters and spectrophotometers. Figure 1(1.4.3) il-
lustrates the spectral reflectance curves of these
tiles.
 59
"S19}9I0}OYdO.QIIdS PUL S.19JOUILIO[OI JO BdUBULIOJI0d VY} YOY 0} ddULJII[Jo1 JO
SPALPUL]S Po1O[OD S¥ PasN Ud}JO SaT IUIv.19I TIN JO SOAIND souRJIaTJoI [RIEdS “(Ep [)T “Sl
(WU) X¥ HLONATSAVM (wu) X HLONSTZAVM . (wu) X HLONA TIAVM
002 009 00s O0v 002 009 0Os 0O0v ool 009 00S 00
T T T
uae 440g

J3ONVL93143u

JOINVLI3171453y
SINVL931434

M0}|2\
(WU) X HLONZTSAVM (WU) YX HLON3ZTSAVM
002 009 OOS 00+ 002 009 OOS OOb

uoouDW
T

Ae)
fos winipey;

ps]
.
vO
=

m
=I
a
AaSe

m im
m 6 D>S
MM.5 24

44619

90
Aoag
80
Ol
 60 PHYSICAL DATA

1.4.4 Black Surfaces 1.4.5 Building Materials

There are several materials that absorb most of
the incident radiant energy and reflect only a
small portion of it. These materials are perceived
as black under normal conditions of viewing and
have useful applications as baffles in optical in-
struments to reduce the effects of stray energy,
and as backing materials in reflectance measure-
ments of semitranslucent objects. Some special
materials, such as gold blacks, are used as coat-
ings of the elements receiving radiant energy in
radiation detectors (Section 1.6).
Gold black, cloth materials (e.g., Norzon cloth),
flat black paints (3M Company), and black glasses
(Carrara glass) have spectral reflectance values
between 0.5 and 1.0% in the visible spectrum, but
often show relatively high reflectances in the in-
frared (Grum and Becherer, 1979).
The spectral reflectance curves of a number of
representative building materials (bricks, shingles,
sheet metals, rocks, and enamel paints) are given
in Figures 1(1.4.5) to 5(1.4.5) over a wavelength
range from 0.3 to 1.8 wm. These results all refer to
samples diffusely illuminated and viewed per-
pendicularly.
1.4.6 Natural Objects
Krinov (1947) has published a rather comprehen-
sive atlas of spectral reflectance properties of
natural formations found within the USSR. They
include forests and shrubs, grass, mosses, field
and garden crops, soils, roads, water surfaces, and
snow. Figure 1(1.4.6) shows spectral reflectance
curves of samples of black rich soil, green grass,

ae
a

REFLECTANCE

5c mee
eT
0.6 0.7 O68. 2UO0 ie 41 4e 2 oes
WAVELENGTH A (ym)
Fig. 1(1.4.5). Spectral reflectance curves of different brick samples: (1) extruded brick,
medium light brown (Milton, Ontario); (2) extruded brick, red-brown (Cooksville Ontario); (3)
dry-press brick, light reddish brown (Cooksville, Ontario); (4) extruded brick, light yellowish
brown (Hanley Co., U.S.A.); (5) extruded brick, dark red-brown (Ottawa, Ontario); (6)
dry-press brick, yellow (Cooksville, Ontario); (7) soft mud brick, reddish brown (Toronto,
Ontario); (8) dry-press brick, brownish yellow (Toronto, Ontario); (9) sand lime brick, light
- gray (Cooksville, Ontario).
REFLECTANCE

Eft
Se
es
ee
Es
iTAG
|
WAVELENGTH Ad (um)
Fig. 2(1.4.5). Spectral reflectance curves of six differently colored shingles: (1) black; (2) dark
brown; (3) dark red-brown; (4) dark green; (5) blue; (6) light gray.

2
te
REFLECTANCE

eS
ie
a
oi
FNS
SAUNT
0.3 0.4 0:5 0.6 O76 (OC Re 1.4 16 16
WAVELENGTH A (ym)
Fig. 3(1.4.5). Spectral reflectance curves of various sheet metals used as building materials:
(1) aluminum sheet—3S; (2) brass (Muntz metal); (3) sheet copper; (4) copper bearing steel;
(5) hard lead, (6) magnesium alloy; (7) stainless steel; (8) galvanized steel (hot dipped); (9)
rolled zinc.
 eos
Neue
REFLECTANCE

PeiiG4
——)
ele
Na
LAA
0.3 0.4 0.5 0.6 6? O08 LOhOl2- 1.4 @6 hs
WAVELENGTH A (ym)

Fig. 4(1.4.5). Spectral reflectance curves of various rocks used as building materials: (1) black
granite (polished); (2) rose granite (polished), (3) black marble (polished); (4) veined marble
(unpolished); (5) white marble (polished); (6) travertine (filled, polished); (7) sandstone (Nova
Scotia); (8) Indiana limestone; (9) concrete; (10) asbestos sheet-transite; (11) asbestos cement
siding.

y)
iz

REFLECTANCE

2)
ee
ieee
alee
incurs
ee
-
Z BS!eh
cae,
Lal
tfJ

0.3 0.4 0.5 0.6 Oskar O.Baw LOmeol 2 ch4ach6 cke

WAVELENGTH A(uem)
Fig. 5(1.4.5). Spectral reflectance curves of a set of enamel paints: (1) medium gray; (2)
_ white; (3) deep blue; (4) yellow; (5) bright red; (6) brown; (7) medium green; (8) light blue; (9)
light green; (10) peach; (11) ivory; (12) black.

62
 Monochromators 63

FRESH SNOW

SNOW WITH
ICE FILM

REFLECTANCE

WAVELENGTH A (nm)
Fig. 1(1.4.6). Spectral reflectance curves of some typical natural objects (after Krinov, 1947).

dry sand, and snow selected from Krinov’s data. 15 MONOCHROMATORS

From the nature of the samples, it is clear that the
spectral reflectance curves given can only serve as
a very rough guide in assessing the reflectance
properties of other samples.
The spectral reflectance curves of typical
specimens of living human skin are shown in
Figures 2(1.4.6) and 3(1.4.6). The data have been
selected from detailed studies made by Edwards
and Duntley (1939), Buck and Froelich (1948),
and Stimson and Fee (1953).
A monochromator is an apparatus designed to
disperse the incident radiant flux into its spec-
trum from which any desired narrow band of
wavelengths can be isolated with a slit aperture.
1.5.1 Basic Designs of Monochromators
In its simplest form, a monochromator comprises
a dispersion prism, two imaging lenses, and an

1.0 T 1 1oe T i Pie SSI cae i ear

Human Skin of Different Races 0.8b Caucasian Skin light a

o @

: |
ol oe
o 5 ils average
aoe
fo)LS @
Hindu
L Mulatto
sae hi dark 4
@
Reflectance a
aD
Oo ~ 0.2

gl gi
a
I | 1 4 ! |
400 500 600 700 400 500 600 700
Wavelength d (nm) Wavelength d (nm)
Fig. 2(1.4.6). Spectral reflectance curves of living Fig. 3(1.4.6). Spectral reflectance curves of living
human skin of different races (after Edwards and human skin of average, light, and dark Caucasians
Duntley, 1939). (after Buck and Froelich, 1948).
 64 PHYSICAL DATA
~
Entrance Slit
Fig. 1(1.5.1). Schematic diagram of a single-prism monochromator.
entrance and exit slit, as illustrated in Figure
1(1.5.1). The radiant flux passing through the
entrance slit is collimated by the first lens, then
dispersed by the prism into its spectrum which is
brought to a focus in the plane of the exit slit by
means of the second lens. The exit slit may be
placed at any position within the spectrum so that
the emergent radiant flux contains, in principle,
only the desired narrow band of wavelengths. In
fact, there will be mixed with it some flux of all
wavelengths entering the monochromator. This
additional flux, usually referred to as stray radiant
energy or stray light, arises from inter-reflections
between components and scattering by dust,
scratches, and other imperfections in the dispers-
ing element and imaging components of the
monochromator.
An effective way of reducing stray energy is to
combine two monochromators in series in such a
way that the exit slit of the first becomes the
entrance slit of the second. Such an apparatus is
called a double monochromator [Figure 2(1.5.1)], in
contrast to a single monochromator with just one
dispersing system.
-There are two ways of combining the two
dispersive systems in a double monochromator.
The two systems can be combined either to give a
subtractive or additive dispersion; that is, the
dispersion of each half of the double monochro-
mator may either annul the other or combine with
it to give double dispersion. With subtractive
dispersion, the losses experienced by rays in dif-
ferent parts of the instrument field (by prism
Limiting
Aperture
absorption, for example) are more nearly identi-
cal.
A diffraction grating is an alternative dispers-
ing element to be used in a monochromator; and
the imaging lens optics can be replaced by imag-
ing mirrors, usually spherical or parabolic. Often
used designs of single-grating monochromators
are illustrated in Figures 3(1,.5.1) to 5(1.5.1).
As with prism monochromators, grating mono-
chromators can be combined to form double
monochromators to reduce stray energy, particu-
larly higher-order spectra generated by the grat-
ing, and to increase the purity of the emerging
monochromatic radiant flux. However, it is often
advantageous to combine a grating monochroma-
tor with a prism monochromator in such a way
that the prism serves as the predispersing element
of the system. The prism—grating arrangement, in.
that order, preserves approximately the often de-
sired linear dispersion provided by a grating. Fig-
ure 6(1.5.1) illustrates a double monochromator
of that type.
The usable spectral range of a prism mono-
chromator is limited by absorption of the dispers-
ing prism and the lenses, and is confined to the
visible and near infrared (0.38 to 2.0 »m) when
the optical components are of glass. With quartz
prisms and lenses, the spectral range is extended
into the ultraviolet down to approximately 0.2
pm. In monochromators going beyond the visible,
the lenses are commonly replaced by mirrors that
have a high reflectance over a wide spectral range
and, in addition, do not introduce chromatic
 First Second M
Monochromoator Monochromoator BL Exit
(additive
ar rangement) Pan

Field
Lenses

Entrance / Exit
Slits Slit
Sy Pe
/* “\
a First Second :
/ Monochromator Monochromator \
(subtractive
arrangement)
Fig. 2(1.5.1). Schematic diagram of two single-prism monochromators combined in series,
forming a van Cittert-type double monochromator. In the lower part of the figure, the second
monochromator is coupled with the first so as to give subtractive dispersion. When the second
monochromator is coupled to the first as shown in the top part of the figure, additive dispersion
is obtained.

Entrance Slit

Plane Mirror

Off-Axis
Parabolic Mirror Grating

Plane Mirror

ewiwm (Exit Slit

Fig. 4(1.5.1). Schematic diagram of an Ebert-type monochromator with a grating as the

dispersive element and a spherical mirror as the imaging optics.
66 PHYSICAL DATA

Entrance Slit

Spherical Exit Slit

Mirror

Fig. 3(1.5.1). Schematic diagram of a Czerny—Turner type monochromator with a grating as

the dispersive element and imaging optics consisting of mirrors.

aberrations in the optical imagery. Fluorite,
rocksalt, and potassium bromide are used as prism
materials to extend the spectral range into the
infrared. Fluorite gives a usable range of ap-
proximately 0.5 to 9.0 wm, rocksalt of 0.5 to 16.0
pm, and potassium bromide of 0.5 to 25.0 wm.
There are other materials (potassium iodide, thal-
lium bromide) that allow a still further extension
of the spectral range, but beyond 40 um, a grating
must be used as the dispersing element.
Interference-filter wedges of the kind described
in Section 1.3.8 are sometimes used as monochro-
mators when a robust rapid wavelength-scan
instrument of moderate spectral resolution and
purity is demanded.
1.5.2 Resolving Power and Transmitted
Radiant Flux
The fundamental quantities that characterize ev-
ery monochromator are the resolving power and
the ratio of the radiant flux (in the selected wave-
length band) emerging from the exit slit to that
received through the entrance slit. The resolving
power R of a monochromator is defined as the
ratio of the mean wavelength A of two spectral
lines of wavelengths A, and A,, that are just
resolved, to their wavelength difference 6A; that
is,
R= [1(1.5.2)]
For a prism, the resolving power R, can be
expressed in terms of the angular dispersion
d@/dX or the material dispersion dn/dX of the
prism; thus
[2(1.5.2)]

Entrance
Slit

Grating

Fig. 5(1.5.1). Schematic diagram of a monochromator with a plane reflection grating in a

Littrow mount with a single lens serving as the imaging optics.
 Monochromators 67

Entrance
Slit

Predispersing
Prism

Central
Slit

Grating
y

Exit Slit

Fig. 6(1.5.1). Schematic diagram of a prism-grating double monochromator with the prism
serving as the predispersing element.

and beam; thus

r=0(8 = [3(1.5.2)]

The quantity a, is the width of the rectangular

R, = él x) [4(1.5.2)]
The angular dispersion of a grating is given by
beam leaving the prism, and b is the base length
of the prism, as illustrated in Figure 1(1.5.2). It is
assumed that the incident beam passes through (5
Sah
ZI MS OME
Wcos@ a,
5e837
the prism at minimum deviation and completely
fills the prism face. If the beam does not fill the where m is the order of the spectrum, N the
prism, b is to be replaced by the difference of the number of rulings on the grating, and W the
two extreme paths through the prism. When there width of the grating. The resolving power then
are two or more prisms in tandem, b is taken as reduces to
the sum of the bases of the individual prisms.
These formulae give the theoretical resolving R,=mN [6(1.5.2)]
power only; that is, for the case when the slits are
of infinitesimal width. In practice, the effective
resolving power is always smaller than the theo- Between the angle of incidence i and the angle
retical because the slits must have an appreciable 6 at which the beam leaves the grating, there is
width to transmit a useful quantity of flux. the following relation:
For a diffraction grating, as for the prism, the
ae oy ee mAN
resolving power is equal to the product of the sin i + sin a (761.52)]
angular dispersion and the width of the emerging
which is often called the grating equation.
Many modern gratings are blazed; that is; the
groove surfaces are sloped at an angle 9¢, as
shown in Figure 2(1.5.2), in order to concentrate
most of the dispersed radiant flux in the mth
order. Taking the blaze angle ¢ into account, one
Ss Q
has i=¢—a,0=¢+4, and thus, sini + sin@
= 2sin@cos a, which becomes a maximum for
Fig. 1(1.5.2). Diagram showing basic quantities a = 0; that is, when i = 0 = 9. This optimal con-
used to evaluate the resolving power of a prism at dition can be achieved with a Littrow mounting of
minimum deviation. the grating [see Figure 5(1.5.1)].
68 PHYSICAL DATA
Fig. 2(1.5.2). Diagram showing basic quantities
associated with a reflecting diffraction grating. The
magnified section on the right refers to a blazed
grating.
The relative spectral distribution of radiant
flux P emerging from the exit slit of a single
monochromator set to a particular wavelength is
a function of the widths of the entrance and exit
slits. This is illustrated in Figure 3(1.5.2) which
ignores any optical aberrations and assumes that
the effects of diffraction at the slits can be ne-
glected, which is true if the slits are sufficiently
wide. In Figure 3(1.5.2), 5, represents the width of
the entrance slit and s,, the width of the exit slit,
both measured on a wavelength scale. The flux P
emerging at the exit slit is given by the ordinate.
In the first case illustrated, s, is different from s,;
in the second case, s, is the same as s,. The
maximum flux for given total bandwidth is trans-
mitted when both slits have the same width; the
spectral distribution of P is of triangular shape
and the conventional spectral bandwidth or half-
value width 6A =s, = 5, = is the half-base of
the triangle (Réssler, 1949, Gibson, 1949).
The simple geometrical profiles of Figure
3(1.5.2) correspond to the assumption that, over
the spectral passbands considered, the triple
es, + nm)
S\+S2=2 NS)
S,=0.5 Sp S,=S,=S
Fig. 3(1.5.2). Radiant flux P as a function of
wavelength within the spectral band emerging from
the exit slit of a single monochromator (s,, 5. =
width of entrance and exit slit, respectively).
product of (a) the spectral distribution function
of the flux entering the monochromator, (b) the
spectral transmittance of the optical pathway, and
(c) the angular dispersion of the prism or grating,
has a value independent of wavelength. Small
deviations from constancy of the triple product
will produce distortions. For a prism monochro-
mator, the triangle is not exactly symmetrical
because of the variable dispersion, but the dif-
ferences are usually unimportant with reasonable
slit widths. However, for a critical study of the
problem as it affects the use of monochromators
of the highest accuracy, the reader is referred to
the work of Kostkowski (1979) and Shumaker
(1979).
The total radiant flux P,,, passed by a mono-
chromator, with slit widths sufficiently large for
the effects of optical aberrations and diffraction
at the slits to be negligible, is given approximately
by the following formulae.
(a) Prism monochromator set for minimum de-
viation at wavelength 2... with its incidence surface
fully irradiated, and for w,/f; < w2/f>:
Iavioria2
Prop = Ly (A) (AG
fasUh(42 — |a,u( dd
GetPhalioees 2)
e hwy (2w.
_wi)2H (dd
= Lyr(0) | h fi b (aa)
[8(1.5.2)]
where L,= the spectral concentration of the radi-
ance of the source imaged on the en-
trance slit
T(A)=the spectral transmittance of the opti-
cal train
w,,h,, f; =the width, height, and collimator focal
length on the input side of the prism.
W,,h>, f,=the width, height, and collimator focal
length on the output side of the prism
(the precise value of h, is immaterial
if, as assumed, h,/f, =h,/f))
H,b,a,=the prism height, its baselength, and
the width of the projection of its inci-
dence surface in the direction of inci-
dence
Ay, =the angular dispersion of the prism
di
oi =the refractive-index dispersion of the
material of the prism
If A, and A, are the extremes of wavelength

passed by the monochromator, the difference |),
—X,|, defining a half-bandwidth AA = +|A, —
A, |, is given by
w, dr w, dX
or
2AA=s,+ 5,
(where s, and s, are the slit widths expressed on a
wavelength scale), or
For given AA, P,,, is maximal when s, = s,, and
the total flux reduces to
(ees tnr(a)(} (2) (ary
[9(1.5.2)]
where A = Hb is the area of the base of the
prism. If w,/f; 2 w,/f,, reverse roles of w,/f;
and w,/f, in all formulae in (a).
(b) Reflection-grating monochromator used in
the Littrow arrangement [see Figure 5(1.5.1)] and
set to select A. in the mth order spectrum:
Pe tnr(a)( 5)a( 2) ba( AA)?
[10(1.5.2)]
where A’= the area of the fully irradiated rectan-
gular grating
e=the separation between adjacent rul-
ings of the grating
=the fraction of the radiant flux re-
flected by the grating, which is con-
centrated in the mth order of the spec-
trum (Ww, = W2,fi= fo, hy =h2)
In Eqs. 9(1.5.2) and 10(1.5.2), (A./Rer) may
be substituted for AA, where R,,, is defined as the
effective resolving power that allows for slit widths
but ignores diffraction at the slits.
Comparing a prism monochromator with a
grating monochromator, Jacquinot (1954) has
shown that a prism monochromator transmits
considerably less flux than a grating monochro-
Monochromators 69
mator if the two dispersing elements have the
same resolving power and the base of the prism
has the same area as the grating.
1.5.3. Slit Width Correction
The accuracy of spectrophotometric measure-
ments is limited by the fact that the slits of the
monochromator must be wide enough to provide
sufficient radiant flux for the measurement. The
problem of correcting measurements with finite
slits to give values corresponding to zero slit width
has been studied by many investigators (Paschen,
1897; Runge, 1897; Hyde, 1912; Koana et al.,
1940; Hardy and Young, 1949; Eberhardt, 1950).
Suppose the spectral distribution of a source il-
luminating the entrance slit of the monochroma-
tor is {P(A) dA} and the spectral sensitivity of
the detector system receiving the radiant flux
from the exit slit is s(A). With the monochroma-
tor set at wavelength A, using finite slits, the
indication of the detector system is given by
the product P(A)s(A)u(A — A.) integrated over
the total wavelength band passed by the mono-
chromator. In this expression u(A — A,), the pro-
file transmission [see Figure 1(1.5.2)] defines, for
every wavelength within the band passed, the
proportion of the radiant flux that enters the
entrance slit and emerges from the exit slit. With
tip.oeg) = [P(A)s(A)w(r = walya
the following expression enables the product
P(A..)s(A,.) to be computed:
P(X,)s(A,) = a[J(A,) — KATA.)
+1A4J(A.) — ---|
[1(1.5.3)]
xi lity (s,/s2)
6[1 a (s,/s)]}°
| Ss 6 95(5;/52) + 10(s\/s>)° i Ouse aes
180[1 +(s,/s2)]°
70 PHYSICAL DATA

a
It is of interest to note that in Eq. 1(1.5.3) only
A27(2,) =5(X, bi“22 ) even-order differences of the observed quantity
J(A) occur. This shows that spectrophotometric
+I (A.Spt ; Sy
FON) errors due to finite slit widths increase with the
curvature of the product curve J(A). No slit width
+- corrections are necessary in regions of the spec-
A47(X,.) = ws(2. i 122) trum where this curvature is zero; that is, in
straight portions of the curve.
+0(2, “| — 2A4(X.)
=- *

1.5.4 Polarization

The quantities s,; and s, denote the widths
(measured on a wavelength scale) of the entrance
and exit slit, respectively. For equal slit width
5; = 5S; that is, for a triangular shape of the
profile transmission function of the monochroma-
tor, one has k = 1/12 and /= 1/90 [see Figure
3(1.5.2)].
If two different spectral radiant power distri-
butions P’(A) and P’(A) are compared with
the monochromator using the same detector sys-
tem, the ratio of the products P’(A)s(A) and
P’’(X)s(A) derived from Eq. 1(1.5.3) gives di-
rectly the ratio P’(A)/P’(A). Similarly, with the
same spectral distribution of the source, the ratio
of the spectral sensitivities of different detector
systems can be evaluated free from slit width
error.
Monochromators affect the state of polarization
of the incident radiant energy, and the effect is
usually a function of wavelength. Figure 1(1.5.4)
illustrates the polarizing properties of different
commercially available monochromators as mea-
sured by Budde (1980). The radiant energy enter-
ing the monochromator under test is unpolarized.
This is accomplished by placing a small lamp
inside an integrating sphere and imaging the exit
port of the sphere on the entrance slit of the
monochromator. A polarization analyzer, that is,
a polarizer that transmits radiant energy vibrating
in only one plane (linear polarizer), is placed
between the exit slit of the monochromator and a
photon detector insensitive to changes in the state
of polarization of the radiant energy incident on
its cathode. When, for a given wavelength setting

o N“N

©o

om
6 >(6)

POLARIZATION
OF
DEGREE

6
WAVELENGTH ) (nm)
Fig. 1(1.5.4).Degree of polarization and its dependence on wavelength for a few commercial
monochromators: (1) Single monochromator with quartz prism and quartz-lens optics. (2)
Double monochromator with two glass prisms and glass-lens optics. (3) Double monochromator
with two glass prisms and mirror optics, arranged in additive dispersion. (3’) Same as 3, but
arranged in subtractive dispersion. (4) Single monochromator with reflection grating and mirror
optics. [From Budde (1980), private communication. ]

(A) of the monochromator, the analyzer is rotated,
one finds two mutually perpendicular positions at
which the detector gives a maximum (J/,,,,) and a
minimum (J,,;,) response to the incident radiant
energy, respectively. From these readings, the de-
gree of polarization (P) of the radiant energy
leaving the monochromator’s exit slit can be
calculated:
oe Ss it
This equation is equivalent to Eq. 6(1.3.10) given
earlier in connection with sheet polarizers (Sec-
tion 1.3.10) and characterizes the polarizing prop-
erties of the monochromator set at wavelength A.
When the wavelength is changed, the positions
of the analyzer change at which /,,,, and Ji, are
recorded, as well as the values of /,,,, and J,i,;
and thus, the degree of polarization. A typical
wavelength dependence of the azimuth of the
plane of vibration of the radiant energy emerging
from the exit slit of the monochromator is shown
in Figure 2(1.5.4). A strong dependence on wave-
length of the degree of polarization caused by the
grating monochromator is noted, which seems
typical of reflection gratings. It also occurs in
double monochromators with a prism—grating
combination as the dispersive system.
The polarizing properties of a monochromator
can cause problems when the monochromator is
used in a spectroradiometer or spectrophotometer
to measure radiometric quantities of light sources
or spectral reflectances and spectral transmit-
tances of materials. The radiant energy emitted
of Monochromator
150°
no (e)°
AZIMUTH
90°
$9 400 500
WAVELENGTH
Fig. 2(1.5.4). Typical wavelength dependence of the azimuth of the plane of vibration of
radiant energy emerging from the exit slit of a monochromator whose entrance slit is irradiated
by unpolarized radiant energy [from Budde (1980), private communication].
Monochromators 71
from most light sources is partially polarized and
so is the radiant energy reflected or transmitted
by materials. For example, the radiant energy
from a coiled filament imaged at the entrance slit
of the monochromator may have a degree of
polarization of the order of 0.10 depending on the
design of the filament and the portion of its
image that falls within the slit aperture. A 250-W
high-pressure xenon arc lamp was found to emit
radiant energy with a degree of polarization of
approximately 0.07. Strip filaments, on the other
hand, emit radiant energy with little or no polar-
ization, particularly if care is taken to admit into
the entrance slit only radiant energy originating
from the central area of the strip.
A virtually perfect depolarizer is the integrat-
ing sphere. Polarized radiant energy entering an
integrating sphere will generally emerge from it
completely depolarized. A plane diffuser (opal
glass, ground glass) will reduce the degree of
polarization of the incident radiant energy to
some extent, but generally will not reduce it to
zero.
1.5.5 Wavelength Calibration
The wavelength scale of a monochromator is nor-
mally calibrated by setting on particular lines of
known wavelength provided by a standard emis-
sion-line source. Among the many sources availa-
ble, probably the most useful are the low-pressure
mercury arc and the helium and cadmium dis-
charge tubes. Table 2(1.3.5) lists a number of
wavelengths in the ultraviolet, visible, and near
infrared regions of the spectrum at which these
Exit Slit O°
Analyzer
600 700
A(nm)
Te PHYSICAL DATA

sources have lines of sufficient strength for most and it follows

wavelength calibrations (see, for example,
Gibson, 1949; Meggers et al., 1961; Habell and P R5Trax — Ry
$=eet —_ [21.5.6
Jackson, 1963). Mercury and cadmium or mercury,
zinc, and cadmium can also be combined in single
tubes and are sold by the manufacturer as The accuracy of this method of determining stray
mercury—cadmium tubes and mercury—zinc-— light depends to a large extent on the quality of
cadmium tubes, respectively. the interference filter and its calibration obtained
independently.
1.5.6 Stray Light
Stray light or stray radiant energy in a monochro-
1.6 PHYSICAL DETECTORS OF
mator arises from inter-reflections between the
RADIANT ENERGY
optical components in the instrument and scatter-
ing by dust, scratches, and other imperfections in
In many investigations in color science, the need
the dispersing element(s) and imaging optics. Stray
arises to make physical measurements of the
light that emerges from the exit slit of the mono-
radiant power involved. Such measurements are
chromator systematically distorts the geometrical
made by radiometers or photometers of various
profile of the spectral radiant flux passed through
designs, all of which employ a physical detector
the instrument [see Figure 3(1.5.2)] and can seri-
of radiant energy. There are several types of de-
ously affect the performance characteristics of the
tector commercially available, but many separate
monochromator. Stray light is one of the major
performance characteristics must be considered
concerns when it comes to selecting a monochro-
when selecting one that is most appropriate for a
mator for a particular application. In many ap-
particular application. The important perfor-
plications, stray light will be most critical in
mance characteristics of a physical detector relate
wavelength regions where the monochromator has
to the following properties of usual concern:
a low transmittance and where the detector, for
example, a photocell or the eye, has a low respon-
(i) Spectral range of response to incident
sivity.
radiant power,
An accurate measurement of stray light pre-
(1) Magnitude of response, distinguishable from
sents some difficulty, but the following procedure
noise, to a given level of incident radiant power,
is reasonably simple and satisfactory in most cases
(iii) Capability of absorbing the incident radiant
(CIE, 1979a).
power without damage or severe aging effects,
A well-blocked interference filter of high peak
(iv) Functional relation between response and
transmittance is used, usually of the square-top
incident radiant power (linearity),
multicavity passband type [see Section 1.3.7(ii)].
(v) Speed of response.
It is necessary to know its peak transmittance
Tmax and the wavelength A,,,, at which T,,,, oc-
Other considerations of importance are those
curs. With the monochromator set to transmit at
of cost and the kind of ancillary apparatus (optics,
\ max» twO measurements of transmitted radiant
electronic circuitry, cooling) needed to operate a’
energy are made, one with and one without the
particular detector.
interference filter over the exit slit. The light
There are two main groups of physical detec-
source should be the one normally used with the
tors: thermal detectors and photon (or quantum)
monochromator, and the detector should be one
detectors. Within each group there are several
with a spectrally uniform (flat) response, at least
types; those of some interest to photometrists,
over the total spectral range within which the
colorimetrists, and visual scientists are briefly de-
monochromator is to be used. Otherwise, it should
scribed below. In describing the performance
be the detector normally used with the monochro-
characteristics of these detectors, certain technical
mator. If R, and R, are the measurements made
terms are commonly used (see, for example, Reed,
with and without the interference filter, P(A jax)
1971; Grum and Becherer, 1979).
is the true radiant power transmitted by the
monochromator, and P, is the stray light, then (a) Spectral Responsivity. The spectral respon-
(approximately) sivity, denoted by R(A), of a physical detector is
the ratio of the detector output, the signal, mea-
R, i keleeeP (Ngee. )
sured in amperes or volts, to the incident spectral
Ry =kP(Xmmax) + KP, [1(1.5.6)] radiant power, measured in watts. If the incident

radiant power is chopped (modulated), R(A) is
measured in terms of the rms amplitudes of the
fundamental frequency components. It is not af-
fected by the bandwidth, nor by the chopping
frequency, if the latter is low.
The term ‘‘spectral sensitivity” has often been
used for the same quantity, but is no longer
recommended.
(b) Quantum Efficiency. The quantum ef-
ficiency, denoted by n, of a photon detector is the
number of electrons released per incident photon.
Quantum efficiencies for different photon detec-
tors vary between 0.001 and 0.35 (approximately)
and depend upon wavelength.
(c) Signal-to-Noise Ratio. The signal-to-noise
ratio is defined as the ratio of the detector output,
the rms signal, measured in amperes or volts, to
the rms value of the noise, also measured in
amperes or volts. The signal-to-noise ratio is a
quantity without a dimension. Noise in a detector
expresses itself as detector signals whose small
values distribute randomly within a narrow range.
Noise originates at one or more points in the
chain of events that govern the detection process
and can be caused, for example, by the incident
radiant energy to be measured, the detector tem-
perature, the detector bias voltage (if applicable),
and the signal-modulation frequency (if applica-
ble).
(d) Noise Equivalent Power (NEP). The noise
equivalent power is the rms value of the sinusoid-
ally modulated incident radiant power, measured
in watts, required to produce a signal-to-noise
ratio equal to unity; that is, SNR = 1. The quan-
tity is a function of wavelength and its determina-
tion depends on the modulation frequency and
the noise bandwidth. A greater NEP means a
poorer detector performance.
(e) Detectivity (D, D*). The detectivity D of a
physical detector is defined as the reciprocal of
the noise equivalent power; that is, D = 1/NEP.
A more commonly used quantity is the normalized
detectivity D*, which is defined as the detectivity
D multiplied by the square root of the active area
of the detector and the square root of the noise
bandwidth. This normalization factor was chosen
because the NEP of many detectors proves to be
proportional to the square roots of both the active
area and the noise bandwidth. Both D and D* are
functions of wavelength and their determinations
depend on the conditions of measurement, such
as the modulation frequency used.
Physical Detectors of Radiant Energy B
(f) Response Time. The response time (or time
constant), denoted by 7, is the time, measured in
seconds, required for the detector output to rise
from a stated low value to a stated high value
when a steady level of radiant energy is instanta-
neously incident on the active area of the detec-
tor. For detectors whose output rises exponen-
tially with time, the response time is the time at
which the output has reached a value equal to
(1 — 1/e) = 0.632 of its final value. After a time
of 57, the detector output will be approximately
99% of its final value. At high chopping frequency
f, the responsivity R, of the detector decreases
significantly in comparison with the dc value Ry
of the responsivity. The following relation holds:
Ry= Ro|1 + (2nfr) |”
1.6.1 Thermal Detectors
A thermal detector detects the heat produced in
its receiving element caused by the absorption of
the incident radiant energy. The receiving element
has some suitable temperature dependent electri-
cal property which is used to measure the temper-
ature increase. Depending on the design of the
detector, the temperature increase may
(i) result in a change of the detector resistance,
(ii) produce a thermoelectric voltage, or
(ii1) alter the detector capacitance.
Theoretically, the response of a thermal detec-
tor is independent of the relative spectral distri-
bution of the incident radiant energy. This feature
gives the thermal detector an important ad-
vantage over a photon detector whose response
will always be wavelength dependent. However,
in practice, there are certain limitations to the
wavelength independence of the thermal detector.
The limitations are introduced by the spectrally
nonselective absorptance of the receiving element
of the thermal detector and the spectral transmit-
tance of window materials used in front of the
detector element, such as air, pyrex glass, fluorite,
fused quartz, or sapphire. A marked disadvantage
of thermal detectors is their usually long response
time which is the result of the fact that the signal
production in these detectors is an indirect pro-
cess. An exception to this rule is the pyroelectric
detector. Table 1(1.6.1) gives the performance
characteristics of some typical thermal detectors.
The theory and design of thermal detectors has
been studied in considerable detail. The interested
74 PHYSICAL DATA

Table 1(1.6.1) Performance Characteristics of Typical Thermal Detectors (According to Reed, 1971)

Responsivity Detectivity Response Time Active Area

R= R(A) D* (max) T Dimensions Impedance
Detector (V-w ') (mm: Hz!/?-W~') (s) (mm) (Q)
Bolometer
Thermistor ec ty) 2.0 10? 0.016 0.50.5 2.0 10°
Metal (Pt) 4.0 Beant: 0.016 7.00.3 40.0
Thermopile
Schwarz vacuum 30.0 13.010? 0.10 9.00.5 pS ait
Moll 0.25 0.1110? LS 10 10 60.0
Pyroelectric DIGI 2.0 10° 2.010 4 6.02.5

reader may find the following publications partic- semiconductor (thermistor) strip as shown in Fig-
ularly useful: Simon and Suhrmann (1958); Conn ure 1(1.6.1). Most of the radiant energy incident
and Avery (1960); Kruse, McGlauchlin, and on the strip is absorbed and then causes its tem-
McQuistan (1962); Smith, Jones and Chasmar perature to rise which, in turn, changes the resis-
(1968); Putley (1970), and Stevens (1970). tance R, of the strip. The resistance change then
leads to a change in the electric current i flowing
(i) Bolometers. In a bolometer, the detector through the strip due to the bias voltage V. The
element consists of a thin blackened metal or same current i flows through the resistor R, placed

DETECTOR
ELEMENT
FRONT VIEW OF
CHOPPER
(reduced scale)

RADIANT g
ENERGY

Fig. 1(1.6.1). Schematic diagram of a bolometer. Detector element is blackened metal or

semiconductor strip whose resistance R, changes when incident radiant energy raises its
temperature. The resulting change in current i leads to a change in the voltage V, across the
external resistor R,. This voltage change is the detector output to be measured. A chopper is
used to facilitate the measurement.
 Physical Detectors of Radiant Energy 75

:
BLACK RADIATION
SHIELD
ABSORBER LS
ae
mk
is
RADIANT Gavi
/K &
ts
Lite
WU
(OX.

ENERGY
Stewie

=<
a.
i
Fig. 2(1.6.1). Schematic diagram of a thermocouple. The detector element is the first junction
of two thin ribbons of dissimilar metals covered by a black absorber. The incident radiant
energy raises the temperature of the junction to 7. The second junction is maintained at the
ambient temperature 7). A voltage AV is measured across the second junction which is
proportional to the temperature differences AT = T — T, between the two junctions.

parallel to the strip. The voltage V, across R;
changes with a change in i, and this change is
measured and recorded as the dector output in
response to the radiant energy incident on the
detector element, the blackened strip. When mea-
suring a steady beam of radiant energy, a chopper
is used to modulate the detector output, thus
facilitating the measurement of the small voltage
changes.
(iii) Pyroelectric Detectors. A pyroelectric de-
tector is a more recently developed thermal detec-
tor. Certain acentric crystals, such a triglycine
sulphate, show permanent electrical polarization
at temperatures below the Curie point. A temper-
ature change of the crystal changes its lattice
Electrode

(ii) Thermocouples, Thermopiles. A thermo-

couple consists of two junctions in an electric Pyroelectric
current made up of two dissimilar materials (thin Axis
and narrow ribbons of metals or semiconductors),
Radiant
as illustrated in Figure 2(1.6.1). Typical materials
Energy
used in thermocouples are bismuth—alloy and an-
timony—alloy combinations. One of the two junc-
tions is covered by a black absorber and receives
the radiant energy to be measured, whereas the
other is screened and remains at the ambient
temperature. An electric current is generated in
the circuit proportional to the difference of tem- Electrode

perature of the two junctions, and the voltage

measured across the unblackened junction repre-
sents the detector output.
A thermopile consists of a number of thermo-
couples connected in series with their blackened
junctions clustered to form a target for the inci-
dent radiant energy.
In most photometric work, thermocouples and
thermopiles are preferred to bolometers because
they are less bulky and more convenient to use.
Fig. 3(1.6.1). Schematic diagram of a pyroelec-
tric detector consisting of a crystal slice acting as a
capacitor to which electrodes are attached. When
chopped incident radiant energy increases the tem-
perature of the crystal from 7T to 7+ AT, an
electric current flows between the electrodes and a
low external impedance (not shown). The electric
current is representative of, and in phase with, the
chopped incident radiant energy.
 76 PHYSICAL DATA

dimensions, and thus shows a change in electrical 1.6.2 Photon Detectors

polarization. This change can be observed as an
electrical signal if electrodes are placed at the
opposite faces of a crystal slice, forming a capaci-
tor, as shown in Figure 3(1.6.1). If the radiant
energy incident on the crystal is chopped, an
electric current representative of, and in phase
with, the incident radiant energy will flow be-
tween the electrodes and a sufficiently low exter-
nal impedance. The chopping frequency must be
greater than the reciprocal of the thermal re-
sponse time of the device.
Pyroelectric detectors have a responsivity that
is essentially independent of wavelength. Their
main advantages over other thermal detectors,
such as bolometers, thermocouples, and thermo-
piles, are that their absolute responsivity is much
higher because higher modulation frequencies can
be used. They also have a shorter response time,
and for optimal performances they do not require
cooling. However, pyroelectric detectors are sensi-
tive to vibration including sound, which increases
their electric noise signals.
Photon (or quantum) detectors utilize either an
external photoelectric effect or an internal photo-
electric effect.
An external photoelectric effect (or photoemis-
sion) occurs when the energy of a photon is
sufficiently high to free an electron from the
surface of the photosensitive material; that is, the
photon energy is greater than the work function
of the surface.
As the photon energy is inversely proportional
to wavelength and the range of suitable photosen-
sitive materials with a low work function is limited,
the spectral responsivity of photoemissive detec-
tors is mainly restricted to the ultraviolet and
visible part of the spectrum. Figure 1(1.6.2) il-
lustrates a number of relative spectral responsiv-
ity curves (normalized so that the maximum re-
sponse reads unity) of common detectors of this
type.
An internal photoelectric effect occurs when the
photon energy is too small to free an electron, but

ane te
YER
pte
analgea
pce
andehe tel

0.8

A AZ a
a @TN
RESPONSIVITY
RELATIVE

LS es
TNS LARBde 500 600
WAVELENGTH
700
d (nm)
800

Fig. 1(1.6.2). Spectral responsivity curves of typical photon detectors operating by the
external photoelectric effect; that is, photoemission. All curves have been normalized so that
their respective maxima read unity. The curves correspond to photosensitive materials (cathodes)
made of the following materials: (S1) cesium on silver oxide (Cs—Ag,O); (S4) cesium—anti-
mony on nickel base (Cs—Sb); (S8) cesium—bismuth on nickel base (Cs—Bi); (S10) cesium on
bismuth and silver oxide (Cs—Bi-Ag,O); (S11) cesium on antimony and oxygen (Cs—Sb-O);
(S13) same as (S11) but in quartz instead of glass; (S20) cesium—antimony on sodium and
potassium (Cs—Sb—Na-K); (SX1) gallium-arsenide (Ga—As); (SX2) gallium-indium—arsenide
(Ga-In-As).
 Physical Detectors of Radiant Energy 77

RESPONSIVITY
RELATIVE
300 400 500 600 700 800 900 1000 1100 1200

WAVELENGTH A (nm)
Fig. 2(1.6.2). Relative spectral responsivity curves of some typical photon detectors operating
by an internal photoelectric effect. (1) all-semiconductor (silicon) photodiode (“diffused”
series); (2) Schottky barrier silicon photodiode (metal—semiconductor device); (3) same as 2 but
uv-enhanced type.

is sufficiently high to raise a charge carrier in a
semiconductor from the valence band or an im-
purity level to the conduction band. Two specific
internal photoelectric effects of importance are
the photoconductive effect and the photovoltaic
effect.
The photoconductive effect is observed when
the mobilized charge carriers increase the conduc-
tivity of the semiconductor.
The photovoltaic effect is observed when the
charge carriers are generated to a point where a
potential barrier exists, for example, at a p—n
junction or metal—semiconductor junction, and a
voltage is produced by charge separation.
There are several materials available that pro-
vide spectral responsivity curves from the near-ul-
traviolet, through the visible, into the infrared
region of the spectrum. Some materials are useful
for detection in the far infrared. Figure 2(1.6.2)
illustrates typical spectral responsivity curves of
photon detectors operating by an internal photo-
electric effect.
Table 1(1.6.2) gives the performance character-
istics of some typical photon detectors of general
interest to those who make photometric measure-
ments.
The theory and design of photon detectors has
been studied in considerable detail. The interested
reader may find particularly useful the books by
Gibbons (1966) and Sommer (1968). Also recom-
mended are the brochures of some of the major
manufacturers of photon detectors which contain
a wealth of useful information.
(i) Photomultipliers. Of the various types of
photon detectors operating by the mechanism of

Table 1(1.6.2) Performance Characteristics of Typical Photon Detectors (According to Reed, 1971)

Responsitivity Detectivity Response Time Active Area

R(A) D* (max) T Dimensions Impedance
Detector (A-W') (mm: Hz!/?-W~') (s) (mm) (Q)
Photomultiplier
Sl cathode 0.50 10° Lo x10" LO-<10 36 (dia) 1.3107
S11 cathode 0.12 10° 5.9 10!° POX 50 (dia) 0x 10°
$20 cathode 0.40x 10° 9.2 10!° LOIO. 50 (dia) Pt?
Photodiode
Si (photoconductive 0.45 ho) 64 Uj Dl Usee 7.5 (dia) LO<10"
mode)
Si (photovoltaic 0.50 Sos 10 5.0) SLD 2 10 (dia) 0.5 10°
mode)
78 PHYSICAL DATA
Rr (Feedback resistor)
ia (Current)
Anode
nth Dynode‘} -.----
High-Voltage Power Supply
(1 to2 kV)
Zener diode
Fig. 3(1.6.2). Schematic diagram of a photomulti-
plier and its electric circuit (drawing kindly sup-
plied by W. Budde of the National Research
Council of Canada, 1980).
photoemission (external photoelectric effect), the
photomultiplier is the most commonly used type
in photometry. Figure 3(1.6.2) illustrated sche-
matically a photomultiplier tube and its internal
design. Radiant energy incident on the cathode,
coated with an appropriate photosensitive
material, frees electrons. Each of these electrons is
accelerated by a fixed potential difference of 100
V, say, to the first dynode, which is an electrode
having a surface treated to give a high secondary
emission ratio for electron impact. The electron
incident on the first dynode causes the release of
a few (e.g., three) secondary electrons that are
accelerated by a further potential difference of
100 V, say, to the second dynode, and so on.
Usually, between 9 and 14 dynodes are placed
between the photocathode and the final anode.
The dynode chain generates an anode-current
gain of several orders of magnitude. The gain
depends on such factors as the number of dy-
nodes used, and the potential differences between
the dynodes. A gain of 10’ is not uncommon in
modern photomultipliers.
Photomultipliers are very suitable for photo-
metric calibrations in color science, particularly
when low levels of radiant energy are to be mea-
sured accurately. However, considerable care must
be taken with regard to the operation of these
devices. A few notes may be appropriate here.
(a) The high-voltage power supply must
highly stabilized (0.01% or better).
(b) The cathode-to-first dynode voltage must be
stabilized if relative spectral responsivity is to
remain constant. This is accomplished by intro-
ducing an appropriate zenerdiode in the circuit.
(c) The current in the voltage-divider chain of
the dynodes must be kept at a much higher level
than that of the maximum anode current gener-
ated by the photomultiplier. A ratio of 1000: 1 is
considered satisfactory.
(d) The anode current in most photomultipliers
should not exceed | »A. This restriction helps to
reduce fatigue effects.
(e) Shielding against earth or other external
magnetic fields is necessary. This is accomplished
by wrapping “mu sheet metal” around the photo-
multiplier tube.
(f) Wherever feasible, an appropriate diffuser
should be mounted in front of the cathode to
avoid effects of nonuniformities in the responsiv-
ity of the cathode surface.
The last note (f) deserves some further consid-
eration (Budde, 1969). Figure 4(1.6.2) shows the
results of uniformity measurements made on a
photomultiplier tube with a plane end-on
semitransparent cathode of the S10-type spectral
responsivity as shown in Figure 1(1.6.2). The
image, somewhat reduced in size, of the exit slit
of a monochromator, set at a fixed wavelength,
was moved diagonally across the cathode in two
mutually perpendicular directions. The responsiv-
ity relative to that recorded at an arbitrarily
selected spot of the cathode varied with the posi-
tion of the image. The results shown in Figure
4(1.6.2) indicate large spatial nonuniformities of
responsivity, which unfortunately are quite typi-
cal of all photon detectors of the photoemissive
type.
Figure 5(1.6.2) illustrates for the same photo-
multiplier tested in Figure 4(1.6.2) a number of
spectral responsivity curves obtained at different
spots of the cathode. The different curves differ
by factors that are wavelength dependent.
If a diffuser of the kind illustrated in Figure
6(1.6.2) is placed in front of the cathode, spatial
uniformity of the spectral responsivity is ob-
tained. The device also eliminates any effect
polarized incident radiant energy may have on the
responsivity of the photomultiplier.
Photomultipliers, properly operated, will re-
spond linearly to incident radiant energy within a
range of approximately three to four decades;
be that is, the proportionality between voltage out-
put and radiant power input is maintained if the
 PHOTOCATHODE

LIGHTSPOT

(ARBITRARY
RESPONSIVITY
UNITS)

30 20 10 0 10 20 30
DISTANCE FROM CENTER OF CATHODE (mm)
Fig. 4(1.6.2). Responsivity of a photomultiplier as a function of position of spot on cathode on
which constant monochromatic radiant energy is incident (after Budde, 1969).

Distance from
center of cathode

20mm

RESPONSIVITY
SPECTRAL
RELATIVE

300 400 500 600 700 800

WAVELENGTH A(nm)
Fig. 5(1.6.2). Spectral responsivity curves of a photomultiplier as a function of position of
spot on cathode on which radiant energy is incident (after Budde, 1969).

79
80 PHYSICAL DATA

Housing of 12 photomultipliers are shown, all belonging to

Alumina cylinder
the same type (S-10 response), and all manufac-
tured by the same company (Budde, 1969; 1973).
I SALI
///1/11111111
al
1 1.1 .® ' The spectral responsivity of the photocathode
Radiant
of a photomultiplier can change with age. A typi-
Photocathode oe cal example of the magnitude of change is shown
Energy
SWIMM
\ hsWt in Figure 8(1.6.2) for a photomultiplier whose
WLLL
Yh Ny(Dob) MIM IILTSS

spectral responsivity was measured over a period

TEE, S
Se

Opal glass or ground quartz

Fig. 6(1.6.2). Cross section of photomultiplier
tube with end-on cathode to which is attached a
diffuser cavity to obtain spatially uniform respon-
sivity of the photomultiplier (Budde, 1969).
radiant power varies over a range of approxi-
mately three to four decades (see, for example,
Pitz, 1979).
The nominal spectral responsivity curve of a
given type of photomultiplier, such as the curve
denoted by S10 in Figure 1(1.6.2), provides only a
crude indicator of the actual spectral responsivity
curve of a particular photomultiplier of that type.
In Figure 7(1.6.2), the spectral responsivity curves
of 18 months (Budde, 1969).
(ii) Silicon Photodiodes. The silicon photodi-
ode is at present the most commonly used photon
detector operating by an internal photoelectric
effect; that is, either the photoconductive or the
photovoltaic effect. The essential component of a
silicon photodiode is a p—n junction, that is, an
interface between a p-type material and an n-type
material. The p-type material is a material having
a shortage of electrons or ‘‘holes” and is usually
a semiconductor (silicon) appropriately doped
with impurities. The n-type material is a material
having an excess of electrons and is either a thin
metal film, such as gold, or a semiconductor
(silicon or selenium) appropriately doped with
impurities.

RELATIVE
SPECTRAL
RESPONSIVITY

=
300 400 500 600 700
WAVELENGTH (nm)

Fig. 7(1.6.2). Spectral responsivity curves of 12 different photomultipliers of the same type
(S-10 response) manufactured by the same company. The curves have been normalized so that
the area under each is the same within the spectral range from 450 to 700 nm (after Budde,
1969).
 Physical Detectors of Radiant Energy 81

2.0 |
a

A. Sept. 66
B. Jan. 67
C. Jun. 67
D. Feb. 68

0.5

RESPONSIVITY
SPECTRAL
RELATIVE

300 400 SOO 600 700 800

WAVELENGTH A(nm)

Fig. 8(1.6.2). Spectral responsivity of a photocathode changing with age (Budde, 1969).

An all-semiconductor photodiode is usually

Vig=iceR
Signal
out Is Ry constructed by diffusing through one surface of a
Ghcranty is }
p-type silicon wafer an appropriate impurity to
Photodiode R, produce a very thin surface layer having n-type
( Load
Resistor
)
(A)
properties. Incident radiant energy that reaches
the depletion layer at the p—n junction causes the
generation of electron-hole pairs. With an ap-
ee
propriate external electric circuit, a current is
ay

(Signal Current)
is Ry
(Feedback) (B)
Photodiode Resistor

Vout = is Re eiah Wir

Responsivity
Relative

=i tO) o=5 0 ) 10 IS
Fig. 9(1.6.2). Electric circuit diagrams required
to operate a silicon photodiode in the photocon- Distance from Center of Photodiode (mm)

ductive mode (A) or in the photovoltaic mode (B).
Typical bias voltage used in the photoconductive
mode operation is 50 V. The load resistor (R, ) is
chosen so that the voltage drop across it remains
small as compared to the bias voltage.
Fig. 10(1.6.2). Spatial uniformity of responsivity
of silicon photodiode measured by moving a small
light spot across the active area of the photodiode
(insert). Scans parallel to the scan shown gave
similar results (Budde and Dittmann, 1973).
 82 PHYSICAL DATA
SIGNAL CURRENT (A)
Fig. 11(1.6.2). Linearity of six different silicon photodiodes. The ordinate represents the
factor f of deviation from linearity. The abscissa gives the measured signal current in amperes
(A) on a logarithmic scale (Budde, 1979).
generated that is proportional to the incident
radiant power.
When the depletion layer, often called the
intrinsic or i-region separating the p-type and
n-type layers in the photodiode, is clearly dis-
tinguished in the design of the photodiode, the
manufacturer refers to PIN photodiodes as
opposed to PN photodiodes. PIN photodiodes
generally have an improved response time, re-
sponsivity, and linearity over PN photodiodes.
Photodiodes of the types described above can
be operated in either the photoconductive or the
photovoltaic mode. Figure 9(1.6.2) illustrates the
essential components of the electric circuits used
in the two modes of operation. The photoconduc-
tive mode is mainly used when a fast response
time is required such as in the measurement of
short light pulses. In virtually all other photomet-
ric applications, the photovoltaic mode is pre-
SIGNAL CURRENT (A)
ferred because it leads to lower noise signals and
is less sensitive to changes in ambient tempera-
ture.
The spatial uniformity of the relative as well as
absolute spectral responsivity of silicon photodi-
odes is generally good, as shown in Figure
10(1.6.2).
Polarized incident radiant energy has only a
very small, if not negligible, effect on the respon-
sivity of the photodiode provided the incident
beam is perpendicular to the active surface of the
photodiode (zero degree incidence; Budde and
Dittmann, 1973).
The linearity of silicon photodiodes can be
very good over eight decades, but there are cases
where some deviations from strict linearity have
been observed (Budde, 1979), as shown in Figure
11(1.6.2).
 CHAPTER 2
THE EYE
2.1 PREAMBLE
The visible radiant flux incident on the eye pro-
vides the stimulus to vision and is conveniently
termed the external stimulus. No properties of the
observing eye (except its position and orientation
in space) intervene in the specification of the
external stimulus. The initial stages of the visual
process, that is, the modifications of the radiant
flux when it enters the cornea, traverses the optic
media, and penetrates the retina, involve the
properties of the eye structure and tissues. How-
ever, these initial stages are still essentially physi-
cal in character and proceed in accordance with
the laws of physical optics. The critical point (the
X-point) is reached when radiant flux, originating
from the external stimulus, is absorbed by the
light-sensitive visual pigments of the retinal end-
organs (the rods and cones) so as to set going the
further processes that culminate in vision. Beyond
the X-point, the visual process is no longer opti-
cal. Different external stimuli that lead to the
same absorptions at the X-point stage over the
whole retina are visually equivalent. If it were
possible to specify, for a given external stimulus,
the quality and intensity of the radiant flux at
each point occupied by visual pigment, the speci-
fication would define a kind of internal stimulus
in which account would already have been taken
of all the optical modifications undergone by the
external stimulus before the X-point. Since no
further radiational specification would be possi-
ble, this might be termed the ultimate internal
stimulus.
Knowledge of the ultimate internal stimulus
could simplify empirical laws of visual response
and make easier the analysis of the subsequent
nonoptical stages in the visual process. Although
it is not yet feasible to reach even an approximate
specification of the ultimate internal stimulus,
nevertheless, it is possible to correct the external
stimulus for some of the simpler pre-X-point vari-
ables, notably, pupil size and light losses in the
optic media, to obtain intermediate internal stimuli.
The data and formulae of this chapter are mainly
concerned with the optical stages of the visual
process involved in deriving intermediate internal
stimuli.
2.2 THE STRUCTURE OF THE HUMAN EYE
The human eye is an organ of approximately
spherical shape. The sphere has a radius of about
12 mm with a transparent protuberance towards
the front. In the protective envelope, which pre-
serves the eye’s shape, are inserted three pairs of
extrinsic muscles that move the eye in its bony
orbit (the eye socket), a cone-shaped cavity point-
ing slightly outward relative to the central plane
of the head. A change of the point of fixation in
the external field is followed by a corresponding
change of the position of the eye caused by a
well-coordinated contraction and relaxation of the
extrinsic muscles. Fixation of the external point is
maintained also when the head is moved. The
intricate mechanisms of eye movements have been
studied extensively, and the interested reader is
referred to Alpern (1962; 1972); Robinson (1968),
and Monty and Senders (1976).
Figure 1(2.2) shows the general plan of the
human eye with the names of its principal parts.
Brief descriptive notes on the optic media (cornea,
lens, aqueous humor, and vitreous body) and on
83
84 TARIBYE
TEMPORAL
2 The /ens is a biconvex multilayered structure,
x
a
i
a
oO
he,
a AXIS
VISUAL
(e)
Fig. 1(2.2). Schematic near-horizontal median
section of the right eye seen from above. (1)
cornea; (2) aqueous humor; (3) eye lens or crystal-
line lens; (4) vitreous body (2 to 4 = optic media);
(5) retina (layer containing photosensitive end-
organs); (6) choroid (vascular and pigmented mem-
brane); (7) sclera (outer coat of the eyeball); (8)
optic nerve; (9) fovea; (10) optic disk (papilla); (11)
ora serrata (front edge of the retina); (12) ciliary
muscle; (13) zonule fibers; (14) iris; (15) ocular
conjunctiva.
the retina are given below. Detailed accounts of
the structure of the human eye, as well as of other
eyes, are given, for example, by Walls (1942),
Smelser (Ed., 1961), Davson (Ed., 1962), and
Davson (1972).
2.2.1 Cornea
The cornea is the outermost layer through which
the external stimulus enters the eye. Although of
a complex lamellar structure, the cornea is trans-
parent and void of blood vessels. It protrudes
toward the front with a convexity of a radius of
about 8 mm; that is, a somewhat greater curva-
ture than the main body of the eye. The Jacrimal
gland, located on the outside of the eye, secretes
tear and mucus solutions to help maintain a nor-
mal exchange of oxygen and the water balance
within the cornea. This is a continuous process
keeping the cornea transparent. The blinking
eyelids distribute the tears across the cornea. Pain
receptors in the cornea help, together with the
eyelids, to protect the eye from injury.
The cornea is contiguous with the white sclera,
the tough protecting envelope of the eye in which
are inserted the extrinsic muscles that move the
eye in its orbit. The sclera has many blood vessels
but few pain receptors.
2.2.2 Lens
enclosed in a sheath which is connected to the
ciliary muscle by means of the zonule fibers. The
shape of the lens changes during accommodation,
the process that permits bringing the image of an
object in the external field to a sharp focus in the
fovea of the retina.
The change of shape in the lens during accom-
modation occurs mostly at its anterior surface
which touches the iris. When the eye is focused
for an object “at infinity”, the zonule fibers,
attached to the sheath of the lens, exert their
maximum pull and hold the lens in a flattened
position. In that position, the cornea provides
most of the refraction needed to bring the beam
of parallel rays to a sharp focus in the fovea.
When the object is nearer, the ciliary muscle
contracts to release some of the tension exerted
by the zonule fibers, thus allowing the anterior
surface of the lens to bulge. This moves the image
of the near object into the fovea from its position
behind where it would be otherwise.
Some eyes, regardless of age, have corneas and
lenses that do not provide a sufficient range of
refractive power to allow perfect focusing for
distances ranging from near to infinity. Such
deficiencies, as well as certain others that lead to
blurred or distorted images on the retina, can be
corrected by spectacles or contact lenses. In the
aging eye, the lens usually loses its capacity to
change its shape, making near vision of fine detail
impossible without appropriate corrections. The
design and solution of correcting spectacles
and contact lenses is the domain of optometry
(Emsley, 1957; Bennett and Francis, 1962; Borish,
1970).
2.2.3 Aqueous Humor and Vitreous Body
Between the cornea and the lens lies the anterior
chamber of the eye which is filled with a clear
liquid, the aqueous humor. This liquid, continu-
ously generated and absorbed, controls the in-
traocular pressure, which is greater than the
 The Structure of the Human Eye 85

Ms A4 fy By

Mgt} bby etHf AE er a a ad a eS Geer '

\

Fig. 1(2.2.4). Schematic diagram of the structures of the primate retina that can be observed
through a light microscope (reprinted from The Retina by S. L. Polyak by permission of Mrs.
Stephen Polyak, 1981). The numbered layers are those enumerated in the text. The letters
show Polyak’s designations of the nerve cells: (a) rods; (b) cones; (c) horizontal cells;
(d, e, f, h) bipolar cells; (7,/) amacrine cells; (7m, n, 0, p, s) ganglion cells; (uw) “radial fibers” of
Miiller. Polyak comments: “In this scheme the nervous elements are reduced to their
essentials, with, however, the characteristic features of each variety preserved—the location of
the bodies, the size, shape, and the spreading of the dendrites and of the axis cylinders— and
with the synaptical contacts presented accurately.”

atmospheric pressure, to maintain the structural
integrity of the eye.
In the space between the lens and the retina,
about two-thirds of the volume of the eye, is the
vitreous body which consists of a transparent jelly
interlaced with fibers.
2.2.4 The Fine-Structure of the Retina
The retina is a complex and multilayered struc-
ture lining most of the choroid, the vascular and
pigmented layer attached to the sclera, the pro-
tecting envelope of the eye. A schematic cross
section of the retina is given in Figure 1(2.2.4)
which illustrates its structure as revealed through
a light microscope. The illustration is taken from
Polyak’s well-known treatise, The Retina (1941).
Polyak’s intensive study of the retinae of primates
covers and extends much of the earlier work.
More recent preparations of primate retinae, in-
cluding those from human eyes, have confirmed
the main features of the retinal fine-structure
illustrated in Figure 1(2.2.4) (see, for example,
Cohen, 1963; Rohen, 1964; Boycott and Dowling,
1969; Rodieck, 1973).
In accordance with Figure 1(2.2.4), the follow-
ing are the main retinal layers containing the
different kinds of nerve cells (or neurons) and
fibers that can be distinguished by means of a
light microscope:
(1) Pigment epithelium. Cells with processes in-
terdigitating to a small degree the outer segments
86 THESEYE
(2), and containing nonphotosensitive melanin
pigment in which most of the light traversing the
retina is finally absorbed.
(2) Rod and cone layer (bacillary layer).
(a) Outer segments of the rods and cones
containing the actual light-sensitive pigments of
vision.
(b) Inner segments.
(3) Outer limiting membrane. Thin sheet made
up of fibers of the cells (Miller’s fibers) sustain-
ing the framework of the retina.
(4) Outer nuclear layer. Mainly cell bodies
(nuclei) of the rod and cone cells.
(5) Outer plexiform layer. Inner fibers of rod
and cone cells and their synaptic contacts with
outer fiber expansions of the cells of layer (6).
(6) Inner nuclear layer. Mainly nuclei of various
types of bipolar cells and horizontal and amacrine
cells.
(7) Inner plexiform layer. Inner fiber expansions
of bipolar cells and dendritic expansions of gan-
glion cells of all kinds.
(8) Layer of ganglion cells. The cytoplasmic body
of a ganglion cell has an axon that becomes a
fiber of the optic nerve.
(9) Layer of optic nerve fibers. Fibers from the
ganglion cells proceeding across the retina to leave
the eyeball at the optic disk.
(10) Inner limiting membrane. Thin sheet made
up of the inner terminations of Miller’s fibers.
A more recent subdivision of the retina into
layers takes account of the functional properties
of the neural network (Stell, 1972). Between the
pigment epithelium and the inner limiting mem-
brane, the neurons are segregated into three layers:
(1) Layer of photoreceptor cells (rods and cones)
(includes layers of outer and inner segments and
layer of photoreceptor cell bodies; that is, the
outer nuclear layer).
(II) Layer of intermediate neurons (or inner
nuclear layer, containing bipolar cells, horizontal
cells, and amacrine cells).
(III) Layer of ganglion cells.
(I/II) First synaptic layer (or outer plexiform
layer): interconnects the processes of photorecep-
tors and intermediate neurons.
(I/II) Second synaptic layer (or inner plexi-
form layer): interconnects the processes of inter-
mediate neurons and ganglion cells.
~ Refined techniques of tissue preparations and,
most significantly, the use of electron microscopy
have helped to bring out a great deal of the
ultrafine structure of the retina that cannot be
revealed by light microscopy. In particular, the
intricate network of neural connections has been
traced in considerable detail. Rodieck’s (1973)
extensive treatise on the vertebrate retina pro-
vides a comprehensive insight into the subject
matter.
Figure 2(2.2.4) is a reproduction of the
summary diagram assembled by Dowling and
Boycott (1966) which, in a highly schematic
manner, illustrates the main nerve cells and their
interconnections. The subdivision into layers, de-
scribed above, is also indicated.
The human retina covers an area inside the eye
of about 1100 mm” and its average thickness is
about 250 um, giving it a volume of tissue of
about 27.5 mm’. Within this rather small volume
are estimated to be about 200 million nerve cells
of different kinds that are directly involved with
the early stages of the processing of the visual
stimulus reaching the retina.
The stream of photons that constitutes the
internal visual stimulus enters the retina at the
inner limiting membrane (layer 10) and proceeds
through the various other retinal layers before it
reaches the outer segments of the photoreceptors.
The photons that are absorbed by the visual
pigments in the outer segments elicit signals from
the photoreceptors that trigger a chain of events
through the neural network of the retina cul-
minating in coded signals emerging from the
ganglion cells into the optic nerve fibers which
transmit these signals to the brain where they are
interpreted in terms of visual information. Most
of the photons that are not absorbed by the visual
pigments in the outer segments are absorbed by
the melanotic pigment contained in cells of the
pigment epithelium and in the melanophores of
the outer coat of the eye. Some photons back-
scatter through the photoreceptors, the outer
layers of the retina, and out of the eye.
The various processes of signal generation,
transmission, and coding that take place in the
retina when a visual stimulus enters are not well
understood. All explanations given in the litera-
ture are considered to be of a more or less specu-
lative nature; but they are based on structural
data of the retina, such as those summarized in
Figure 2(2.2.4), on electrophysiological data col-
lected inside single cells, though mostly of retinae
of nonprimate and cold-blooded creatures, and
on a host of psychophysical data, some of which
are the subject matter of Chapters 5 to 7.
 The Structure of the Human Eye 87

2a
<=
jo)
_
Qa
®
oO
® Tt
2b Pas
®
c
3°
Oo rod
receptor receptor
cone

iA =
od
(op)
~
midget : Q
bipolar horizontal fa
cell °
a
rm >
Oo
a
=
dp)
Subdivision
by
Polyak

I/II

ganglion
9 cell

10

Fig. 2(2.2.4). Schematic diagram assembled by Dowling and Boycott (1966) of the ultrafine
structure of the primate retina that can be observed through an electron microscope. The
subdivisions of the retina into layers proposed by Polyak (1941) and Stell (1972) are indicated
in the margin of the illustration.

A closer inspection of Figure 2(2.2.4) suggests
a well-organized hierarchy of signal generation,
transmission, and coding in the retina, and it is
interesting to follow the steps that seem to con-
stitute the retinal processes.
Suppose one or more photons have been ab-
sorbed in the outer segment of a rod receptor and,
as a result of the absorption, a signal is elicited
from that receptor. This signal is transmitted down
the length of the rod to its synaptic body, the
spherule. The spherule contains junctions, called
synapses, between the rod receptor and other nerve
cells in the first synaptic layer itself and in the
layer of intermediate neurons. In particular, we
note synapses at which the signal is transferred to
a rod-bipolar cell, other synapses at which the
signal is transferred to a horizontal cell, and still
other synapses that connect directly with the syn-
aptic body, the pedicle, of a cone receptor. The
most important transmission appears to be the
one that feeds directly into the rod-bipolar cell,
because the signal emerging from it is transmitted
through a synaptic body in the second synaptic
layer to a diffuse ganglion cell. The signal will
undergo final retinal processing, together with
signals arriving at the same time from other cells
of the neural network. The output signal gener-
ated by the ganglion cell in response to the input
signals it receives is transmitted through the optic
nerve fibers to the brain.
88 THERYL
It appears that several neighboring rod recep-
tors pool their signals in a common rod-bipolar
cell. The same rod receptors are also connected to
a common horizontal cell, which in turn is con-
nected with a number of rods in other areas of the
retina as well as with cone receptors. The pres-
ence of horizontal cells and their interconnections
with the first synaptic layer suggest lateral
processes to take place in this retinal layer. Hori-
zontal cells are estimated to cover relatively long
distances (of the order of 1000 pm) laterally in
the layer of intermediate neurons. They do not
interconnect with other cells in the layer of inter-
mediate neurons nor do they connect with gan-
glion cells (Kolb, 1970).
When photons are absorbed in the outer seg-
ment of a cone receptor, the transmission and
processing of the resulting signal are, in the main,
similar to that observed for a signal originating at
the outer segment of a rod receptor; but there are
some important differences in detail. The cone
pedicle, considerably larger than the rod spherule,
has between 15 and 30 indentations (invagina-
tions) at its base, within which synapses are found.
Through one such synapse the signal is trans-
ferred to a cone-bipolar cell, always of the same
type and referred to as a midget bipolar. Through
another synapse, the signal is transferred to a
horizontal cell that is central to the lateral processes
that take place in that retinal layer, as described
above. Another type of synapse is located at the
flat portion of the base of the pedicle, between
the invaginations. Through a synapse of that type,
the signal is transferred to another cone-bipolar
cell, called a flat bipolar that generally receives
signals from several (7 to 12) cones.
The signals processed in the midget and flat
bipolars proceed to the second synaptic layer
through which they are transferred to ganglion
cells. Midget bipolars seem to synapse preferen-
tially with midget ganglion cells, whereas flat bi-
polars synapse with both midget and diffuse gan-
glion cells.
In the layer of intermediate neurons, amacrine
cells are found which appear to govern lateral
processes in the second synaptic layer. These
processes run parallel to the lateral processes in
the first synaptic layer governed by the horizonal
cells.
Color vision has been thought to be mediated
entirely by the cones. However, the “wiring dia-
gram”, shown in Figure 2(2.2.4) and discussed
above, strongly supports certain other experi-
mental evidence that rods can, under favorable
circumstances such as at appropriate levels of
stimulation and location in the retina, interact
with cones. The potential interactions are indi-
cated by the existence of
(i) direct contacts between rod spherules and
cone pedicles,
(ii) indirect connections between rods and cones
through horizontal cells, and
(iii) indirect connections between rod-bipolars
and cone-bipolars through amacrine and ganglion
cells.
2.2.5 Main Topographical Features of the
Retina
Inspection of the whole surface area of the retina
reveals a number of easily identifiable features
and major regions. These are the following:
(1) Optic disk, papilla. The gap in the retina
proper, occupied by the optic disk which is not
light sensitive and corresponds to the blind spot in
the visual field. The optic disk is roughly oval
with its longer diameter vertical.
Approximate dimensions: Vertical, 2 to 2.4 mm (7
to 8°); horizontal, 1.5 to 1.8 mm (5 to 6°).
Approximate position of center: 4.8 mm to nasal
side of visual pole and 0.47 mm above horizontal
meridian (16° temporal and 1.6° below in exter-
nal field).
(2) Ora serrata. The abrupt boundary of the
retina toward the front of the eye. Vision is
impossible for stimuli beyond the ora serrata, but
in fact may not reach as far as this.
(3) Fovea. Small region about the visual pole
where vision is most acute.
(4) Rod-free area. Small area centered on the
visual pole that is blind to weak stimuli in dark
adaptation.
(5) Central avascular region. This area is also
centered on the visual pole (it probably corre-
sponds to the region of a very bright external
field, particularly of shortwave light, in which
“dancing dots” of light are not observable).
(6) Yellow spot (macula lutea). A central area
extending beyond the fovea in which a yellow
pigmentation can be seen in retinal preparations
and associated with the Maxwell spot (observable
as a shaded spot in the direction of vision, on
intermittent observation of a uniform field of
deep blue light, and also as a more complicated
pattern with various other colored stimuli).
Table 1(2.2.5) Subdivision of the Retina into Regions (from Polyak, 1941)

Corresponding
Approximate Angular
Outer Diameter
Diameter on in External
Retinal Region Retina Field

Central Area
I. Fovea This corresponds to a pit or 1,500 wm neh
depression in the retina-
vitreous surface, the inner
limiting membrane. [See
Fig. 1(2.2.6)]
The nearly flat central area 400 »m 1.4°
of the foveal pit is called
the foveola. There are no
blood vessels here.
In a central island within 50-75 pm 0.17-0.24°
the foveola, the cones have
maximum length. There are
no rods here.

II. Parafovea Roughly circular belt of 2,500 pm 8.6°

width about 500 wm. Here
the thickness of layer 2 has
dropped to the value it re-
tains throughout the re-
mainder of the retina.

III. Perifovea Circular belt of approxi- 5,500 wm 19.0°

mate width 1500 pm
marked by the progressive
reduction in thickness of
the ganglion-cell layer from
about 4 cells to about | cell
thick.

Peripheral Area

IV. Near periphery Approximate width 1,500 »m 8,500 4m 29.0°

V. Middle periphery Approximate width 3,000 4m 14,500 »m 50.0°

VI. Far periphery Approximate width 10,000 40,000 wm (horizontal)

pm (temporal side)
Approximate width 16,000
um (nasal side)
Approximate width 2,100
jum (temporal side)
Approximate width 700-
800 wm (nasal side)
VII. Ora serrata [For the special characteris- 40,000 pm (horizontal)
(extreme periphery) tics distinguishing regions
IV to VII, see Polyak
(1941).]

89
90 TRE EYE

Table 1(2.2.5)— Continued

Corresponding
Approximate Angular
Outer Diameter
Diameter on in External
Retinal Region Retina Field

VIII. Yellow spot The yellow pigment per- 3,000 pm 10°

meates diffusely all layers (intense)
from 4 to 9. Thus, pigmen- 5,000 p 17°
tation is very slight in the (total)
foveola, intense on the
slopes and margin of the
fovea, and gradually fades
out beyond. However, it is
visible in some preparations
nearly up to the papilla
(Polyak, 1941). The yellow
spot is more extended in
the horizontal than in the
vertical meridian.

IX. Rod-free area [Oesterberg’s data (1935) Diameter 500-600 pm _ 1.7—2.0°

correspond to a_ smaller (Polyak)
value] 10°
(Oesterberg)

X. Avascular area Diameter 400-600 pm _—_-1.4—2.3°

(7) Parafoveal area. Retinal region immediately
surrounding the fovea.
(8) Peripheral area. Region between the para-
fovea and the ora serrata.
Polyak’s more precise subdivision into retinal
regions, based on easily identifiable features in
the morphology of the retina and not on its
functional properties, is given in Table 1(2.2.5).
There is an arbitrary element in assigning some of
the boundaries.
2.2.6 The Fovea
The fovea in the retina is of particular interest as
this is the retinal area where vision is most acute.
The structure of the fovea is illustrated in Figure
1(2.2.6). A pit or depression is seen in the retina—
vitreous surface, the inner limiting membrane.
The retinal layers 5 to 9, redrawn as bands from
Figure 1(2.2.4) and shown at the top of Figure
1(2.2.6) are considerably thinner near the center,
while at the same time, the retinal layers 2 to 4
are somewhat thicker than in the outer region.
The nearly flat central area of the foveal pit,
where layers 5 to 9 are almost absent, is called the
foveola. There are no blood vessels in the foveola.
It is the high density of cones in the fovea
which gives this retinal region its exceptional
capacity to resolve fine detail in an optical image
focused there. Between 100,000 and 150,000 cones
are estimated to be packed in the fovea. The
foveal cones have a somewhat different structure
than the cones in other retinal regions, and the
relative number of pathways from the fovea to the
brain is also greater than from other regions.
These features are believed to be responsible for
the higher sensitivity of the fovea to visual stimuli
as compared to that of other retinal regions, when
the eye is light adapted.
2.2.7 The Photoreceptors
There are two types of photoreceptors in the
human retina, the rods and cones. As repeatedly
 The Structure of the Human Eye 91

pigment epithelium-| ES a EE 010101

Gist ods and n cones
outer nuclear layer—
—— ee ee 5— outer plexiform layer
inner plexiform layer-7 a 6— inner nuclear layer
optic nerve fibers-9 ee
8— ganglion cells

O 500 1000 1500

Distance (um)

=ESS OES SS ISS ° ‘ comm oaGOS <

ini “en Hi|()\!"miewn

Bi go®

O 100 200 300 400 500 600

Distance (um)
Fig. 1(2.2.6). Structure of the human fovea (Reprinted from The Retina by S. L. Polyak by
permission of Mrs. Stephen Polyak, 1981). The retinal layers drawn as bands in the upper part
of the illustration are numbered and identified in accordance with Polyak’s subdivision of the
retina into layers shown in Figure 1(2.2.4) and described in the text. In the figure, the visual
stimulus enters the retina from below.

noted in the previous sections and shown in Fig-

ures 1(2.2.4) and 2(2.2.4), as well as Figure
1(2.2.6), the rods and cones are found in the
photoreceptor layer of the retina. Figure 1(2.2.7)
shows the main structural features of a typical raat ne aa
rod and cone. The names of the two photorecep-
tors are derived from their shapes (Miiller, 1851;
Schultze, 1866). A typical rod has a cylindrical aet
3 : , Ellipsoid
outer segment and a cylindrical inner segment, ROD CONE
consisting of the ellipsoid and myoid regions, both Ahn 2m
of similar diameters. A typical cone, on the other Mold
hand, is tapered and has its largest diameter at
the bottom of the inner segment. Variations in
shape and size are often noted for both rods and
cones, and the distinction by shape between the ucleus
two types is not always possible, particularly in
certain nonprimate eyes (Cohen, 1972).
Table 1(2.2.7) contains data on the distribu-
tion and dimensions of rods and cones compiled
from Polyak (1941) and Oesterberg (1935; see eck
also Pirenne, 1962a). (Spherule)
(Pedicle)
The photoreceptors are optically inhomoge-
neous and it is difficult, if not impossible, to Fig. 1(2.2.7). Main structure features of a typical
assign representative single values for refractive human rod and cone (Rodieck, 1973).
92 THREYE

Table 1(2.2.7) Distribution and Dimensions of Rods and Cones

(a) Total Number in Human Retina

Rods 110-125-10° (Oesterberg)

Cones 6.8-10° (Oesterberg)
more than 4-10° (Polyak)
Nerve fibers 0.8-1.0-10° (Polyak)
Number of cones in foveola (about 400-um dia.): 25,000
Number of cones in fovea (about 1500-m dia.): 110,000 to 115,000

(b) Length and Diameter (in 4m) (from Polyak 1941)

Retinal Region Length (i.e., Diameter of Cones

as Defined in Thickness of Rod Inner Segment — Outer Segment
Pablest(2.275) and Cone Layer) (Tapering) Diameter of Rods

I foveola 70 1.5-1.0 > 1.0 Absent

outer edge (reducing) 3.5-4.0 > 1.3 1.0
of fovea
II 40-45 4.5 > 1.5-2.0 1.0-1.5
Ill 42-45 5:0 > 2.0 |
IV as for III 5.0-6.0 > 2.5 1.5-2.0
V Little thinner 6.0-7.5 > 2.5 1.5-2.0
than for IV
VI Little thinner 8.0-9.0 > 3.0 1.5-2.5
than for IV

index, absorptance, scattering, and so on to a
cone or a rod photoreceptor. Only few optical
data are available on photoreceptors, of which
those on refractive index reported by Barer (1957)
and Sidman (1957) are noteworthy and sum-
marized in Figure 2(2.2.7).
The outer segment of a photoreceptor is of
special interest because it is generally accepted
that this is the part of the photoreceptor wherein
occurs the interaction of visual pigment and light
that initiates the visual process. Electron micros-
copy has revealed a lamellar fine-structure of the
outer segments of both the rods and the cones.
Ever since the first records of such a structure
were produced by Sjostrand (1948, 1953), many
more preparations have been made by numerous
investigators that provide a detailed picture of the
architecture of the outer segments. Reviews are
given, for example, by Sjostrand (1961), Cohen
(1972), and Rodieck (1973). Figure 3(2.2.7) il-
lustrates schematically the fine-structure of the
outer segments of a typical rod and cone.
In the outer segment of the rod, the lamina-
tions have the shape of double-membrane disks
that are not connected with the outer membrane
and seem to float within it. An exception to this
arrangement occurs at the lower end of the outer
segment where the discs are actually infoldings or
ingrowths of the outer membranes [not shown in
Figure 3(2.2.7)]. In time, these infoldings are
pinched off the outer membrane and new disks
are thus generated. The replenishing process
causes the stack of disks to move upwards gradu-
ally. It has been estimated that the stack moves at
a rate of about 10 wm per day. The disks pushed
off the top are absorbed in the pigment epi-
thelium, into which the upper portion of the outer
segment protrudes (Young, 1976).
Throughout the outer segment of a cone, usu-
ally infoldings of the outer membrane occur as
illustrated in Figure 3(2.2.7). A replenishing pro-
cess of cone disks also has been reported recently
(Young, 1978).
The visual pigments that have been extracted
from primate retinae as well as from retinae of
many other species are combinations of the chro-
mophores, |1-cis retinal,, or 1l-cis retinal,, and
various proteins, called opsins. The pigments ex-
tracted from retinae of land vertebrates are all of
the same chemical structure and are called

rod cone
outer
1.4) |Segment), 38
oil
droplet
Fig. 2(2.2.7). Schematic outline of a rod and
cone photoreceptor with average refractive indices
indicated in the various segments. Values are rep-
resentative of rods and cones of various animal
eyes. They were determined by refractometry and
interference microscopy by Barer (1957) and Sid-
man (1957).
rhodopsin (for a review see, e.g., Rodieck, 1973).
Visual pigments prove to be insoluble in saline
solutions, but may be brought into colloidal sus-
pension with detergents. These properties suggest
that the pigment molecules are built into the
membranes of the disks in the outer segments of
the photoreceptors. A detailed model of the struc-
ture of the disk membrane and the arrangement
of pigment molecules in it has been proposed by
Vanderkooi and Sundaralingam (1970).
It is interesting to note that the visual pig-
ments extracted so far from primate retinae all
come from the outer segments of rod photorecep-
tors. No extractions of cone pigments have been
reported. However, physical techniques of mea-
surement, such as microspectrophotometry, show
that three types of cone pigment must exist with
absorptance maxima in the short-, middle-, and
long-wavelength region, respectively (Liebmann,
1972; Bowmaker et al., 1978; Bowmaker and
Dartnall, 1980). The three types of cone pigment
reside in different cones. No fundamental dif-
ference between the visual pigments of rods and
cones is expected to be found. The difficulty
experienced in attempts to extract cone pigments
Specification of the External Stimulus 93
Fig. 3(2.2.7). Schematic illustration of fine-struc-
ture of outer segments of a typical rod and cone.
The inside of most of the outer segment of the rod
(left) contains a stack of double-membrane disks
not connected with the outer membrane. The outer
segment of the cone (right) consists of infoldings
of the outer membrane. (Illustration by Joana
Wyszecki).
is thought to be connected with differences in the
fine-structure between rods and cones (Rodieck,
173).
2.3 SPECIFICATION OF THE EXTERNAL
STIMULUS
2.3.1 Position in External Field
Figure 1(2.3.1) illustrates the coordinate system
used to specify the position of a point P in the
external field of the left eye. Suppose, at a given
moment, the observing left eye is located with its
nodal point (or with its pupil center) at O, and
the eye’s visual axis, produced forward, lies along
OQ. Then, the position of a given point P in the
external field is specified by the polar coordinates
r, 9, 6, where the azimuth ¢ is the angle by which
a half plane bounded by OQ must be rotated
from a conventional reference position to a posi-
tion in which it contains the point P. A suitable
convention is to choose the reference position so
that the half-plane contains the nodal point of the
observer’s other eye (the nasal position), the rota-
tion being taken in the direction nasal-—
upper—temporal—lower—nasal.
2.3.2 Radiometric Specification
Under usual conditions of vision, the stimulus
from a given direction (@, ¢) in the external field
is provided by the radiant power emitted diffusely
94. THE EYE

Plane containing point P of external field

Pir,O,p)

visual axis
of left eye
left eye —_ -——_——

visual axis
of right eye
—_—-

reference plane

Fig. 1(2.3.1). Specification of the position of a given point P in the external field of the left
eye.

by a solid surface at distance r. Let the spectral
concentration of radiance at the point P(r, 0, $)
of the emitting surface in the direction of the eye,
be represented by L,,(@,) and be measured in
terms of watts per unit wavelength interval per
unit area per unit solid angle of emission. Then,
by the definition of spectral radiance (see Section
1.1), an element of area dA, of the surface at P
acts as a source of radiant intensity towards the
eye, which for the wavelength interval dA has the
value
dI,, dX = L,,(0, 6) dd- dAscos es (W- sr_')
measured in terms of watts per unit solid angle of
emission. The angle e, is between the normal to
the surface element dA, at P and the line OP.
At a plane surface dA, containing the nodal
point O of the left eye and turned normal to the
direction (@,), the source dA, produces the
irradiance
dA
GE 3 dr = L.(6, o) dX - a
sas L. (9, o) dn : dw p (W
where
ys dA, COS Es
doy =F (81)
is the solid angle subtended at the eye O by
the surface element dA, of the source. Thus
[L..(8, 6) dA] provides the appropriate measure
of the external stimulus, being equal both to the
external radiance in the direction (@,) and to
the irradiance per unit solid angle (steradian) of
external field on a plane surface at the eye turned
to face the oncoming radiant flux.
Figure 1(2.3.2) illustrates the geometrical
quantities that are involved in Eqs. 1(2.3.2) to
3(2.3.2). The reader will notice that the above
specification of the external stimulus is a radio-
metric procedure governed by principles, quanti-
ties, and units described more fully in Chapter 1,
particularly Section 1.1.
[1(2.3.2)] The steradian is a large unit of solid angle and
it is often more convenient in specifying external
stimuli to use the square degree, that is,
2
1 sq. deg. = Ge steradian [4(2.3.2)]
It follows that the value of the irradiance
E.y,(9,) produced at the eye by one square
degree of the source in the external field is ob-
tained by multiplying the value of the radiance
L.,(9, &) by the factor (1/57.3)7, thus,
. m7)
E.y,(9, 6) = 3.05 - 10-*L,,(0, 6) (W- m7)
[2(2.3.2)]
[5(2.3.2)]
In some cases, the state of polarization of the
radiant flux incident at the plane of the eye is
[3(2.3.2)]
important and the specification L,,(@, @) must be
supplemented by this information. Also, it is im-
plied that the radiant flux reaching O from differ-
 Factors in the Eye that Control the Internal Stimulus
dWp
Fig. 1(2.3.2). Geometrical quantities involved in the specification of the external stimulus.
ent points P in dA, is incoherent, so that no
interference effects can be produced.
The above method of specifying the external
stimulus is primarily applicable when the irradi-
ance at O from a given direction in the field is
uniform over an area covering the apparent area
of the natural eye pupil in that direction. The
most important exception to this occurs when
either by a material artificial pupil placed in front
of the eye or by means of an optical system (as in
the method of Maxwellian view), the irradiance at
O from a given direction in the field is confined
to a small area lying within the apparent area of
the natural pupil. The external stimulus is then
best specified by the total incident radiant flux in
this small area. For example, if the artificial pupil
has the projected area p(0@,¢) in the direction
(9, $), the quantity to be specified is the product
[L..(6, >)p(0, >) dA]. This product is the radio-
metric analogue of the photometric quantity
known as the troland value to be discussed in
Section 2.4.4.
2.3.3 Photometric Specification
External stimuli are also specifiable in photomet-
ric units when the spectral radiant power distribu-
tions are immaterial or are indicated separately.
The quantity specified is the luminance L(@, $)
or the scotopic luminance L’(6, ) in the direc-
tion of the eye at the point P(r,6,¢) of the
emitting surface in the external field. These equal
the corresponding illuminances (E, and E,) at
the eye per steradian. Then
L(8,6) =Ky [L.(8,¢)V(A) da
(Im-m~*-sr_') [1(2.3.3)]
95
Source
with K,, = 683 Im - W7'; and
L'(8,¢) = Ki, ifL.,(8,6)V(A) da
(inerm eres) 412(245-3)]
with K’, = 1700 Im- W!.
The functions V(A) and V’(A) are the stan-
dard luminous efficiency functions for photopic
and scotopic vision, respectively. A complete
treatment of photometry is given in Chapter 4.
2.4 FACTORS IN THE EYE THAT CONTROL THE
INTERNAL STIMULUS
The role of the optical system of the eye is to
image the external scene on the retina; but even
aside from diffraction effects, the imaging is, in
general, imperfect both because of optical aberra-
tions and because the object distance may be
unsuited to the state of accommodation of the
eye’s optical system at the time. Only for objects
on or fairly near the optic axis of an emmetropic
(optically normal) eye and at the distance to
which that eye is accommodated is the image
sharply focused on the retina (to the approxima-
tion of geometrical optics); even then, the radiant
flux from the object must be confined to a limited
spectral region (avoidance of chromatic aberra-
tion), and the eye pupil must not be fully dilated
(avoidance of spherical aberration).
In such cases, a small object surface emitting
radiant flux in the wavelength interval A toA + dA
and subtending the solid angle dw at the eye
produces an image on the retina of area dA pro-
portional to dw; that is,
dA =mdw [1(2.4)]
96 THE\SBYE
where the factor m, apart from its possible depen-
dence on 9, ¢, and A, will vary somewhat from
one eye to another. For example, in eyes optically
identical except for size, m will be proportional to
the square of the linear dimensions (see Section
2.4.1).
If the object surface has radiance (L,, dd)
and the apparent area of the pupil seen from the
direction (8,¢) is p(@,¢), the total spectral
radiant flux P,, dA reaching the retinal image will
be
P,, dX =[L,, dAp(4, ¢)|[1 — 1(8, 6, A)] dw
[2(2.4)]
where /(8,¢,A) is the fraction of the incident
radiant flux lost from the imaging beam in pass-
ing through the eye and reaching the visual pig-
ment in the retina. The retinal irradiance E,, dA
in the direct image at the pigment level is then
E.y d= —[Len ddp(6, ¢) [1 a (0, >, )]
[3(2.4)]
measured in (W- m 7”).
The validity of this equation depends, as indi-
cated in its derivation, on several limiting as-
sumptions, but even when not strictly applicable,
it often provides a useful if crude approximation
to the retinal irradiance corresponding to a given
external stimulus, suitable average values being
inserted for the quantities p(6, ¢), /(@, ¢, A), and
m (see Sections 2.4.5 and 2.4.6, respectively).
In addition to the direct image, the visual
pigment in the retina receives other radiant flux
Originating in the external stimulus but produced
by:
(a) double reflection at the various interfaces in
the eye; that is, cornea—air, lens—aqueous humor,
lens—vitreous body, vitreous body—inner limiting
membrane, and so on,
(b) scattering in the cornea, lens, and optic
humors,
(c) scattering within the retina or reflection from
one part of the retina to another,
(d) fluorescence of the eye lens or the retina.
The nonfluorescent contributions are not gen-
erally significant at a given retinal position unless
the stimulus at the corresponding region of the
external field is small relative to that in other
regions. Some limited data on the nature and
extent of internal stimuli extraneous to the direct
image are available (see Section 2.4.6).
Allowance for diffraction effects in the deriva-
tion of internal stimuli falls under two heads.
First, the image of a monochromatic point ob-
ject on the photosensitive pigment layer of the
retina—regarded as a simple surface—is a dif-
fraction pattern; that is, an extended distribution
of irradiance of size and shape depending on the
wavelength and the form and dimensions of the
apparent pupil. Given a circular pupil and assum-
ing the image of geometrical optics to be perfect,
that is, a point, we find that this distribution is
the familiar Airy disk (Born and Wolf, 1959).
More realistic predictions of the distribution of
irradiance in the image of a point source (line
source or any object with fine detail) must take
account of the following factors:
(a) The geometrical focus may be displaced by
small amounts in front of or behind the visual
pigment layer, and the distribution will be sensi-
tive to the precise accommodation of the eye.
Because of the chromatic aberration of the eye,
such displacements are inevitable for some wave-
lengths if the point source is not monochromatic
(e.g., white light). The distribution in photometric
units is then obtainable by summing the distribu-
tions of different wavelengths, weighted by the
factor V(A) or V’(A), but the spectral composi-
tion of the irradiance will vary from point to
point of the pattern.
(b) For object points on the axis, the geometri-
cal image is subject to optical aberration and also
to astigmatism (as the object point is moved off
the axis); that is, to an increasing separation of
the focus in the plane containing the axis and the
plane normal to this plane.
(c) Scattered light coupled with possible irregu-
lar defects in the eye’s optical system and small
involuntary fluctuations in accommodation have
been observed to produce a spread of the retinal
distribution not attributable to diffraction or reg-
ular aberrations (Flamant, 1955). The tentative
determination of the retinal distributions for
stimulus objects with fine detail presents a special
problem for each experimental situation and sim-
ple general formulae are not obtainable. For a
discussion of work on these lines, see, for exam-
ple, Fry, 1955; LeGrand, 1956; Pirenne, 1962b;
Gubish, 1967; and Westheimer, 1972).
 Factors in the Eye that Control the Internal Stimulus
If the point object is expanded to a surface
patch in the external field over which the radiance
L,, 4X in the direction of the eye is uniform, all
the complications of diffraction, exact accom-
modation, astigmatism, and other aberrations will
affect the retinal irradiance at the edge of the
image but will be of diminishing importance in
moving in to the central area. Equation 3(2.4)
may be regarded as specifying, in the limiting
case of large object patches, the approximate rela-
tion between the external stimulus and the retinal
irradiance in the central area of the direct image.
Many of the studies of color science are con-
cerned with uniform stimuli of one degree or
more in angular subtense to which this approxi-
mation can be applied.
Diffraction phenomena arise also from the fact
that the visual pigment is not an infinitely thin
surface bounding an optically homogeneous
medium. Radiant flux from the vitreous body
incident on the inner limiting membrane passes
initially through the layers between the inner and
outer limiting membranes (layers of total thick-
ness varying from about 50 pm at the fovea to
200 »m in the parafovea and thinning towards
the peripheral retina). These layers, comprising
nerve fibers, neurones of various kinds, and the
parts of the rod and cone receptor cells contain-
ing their nuclei, may be regarded optically as
fairly homogeneous, perhaps as very slightly
turbid. However, the optical situation is altered as
radiant flux enters the inner segments of the rods
and cones which have refractive indices a little
larger than that of the surrounding tissue. A
stratification of slightly different indices parallel
to the outer limiting membrane would be unim-
portant. But as the radiant flux is incident at
quite small angles (not more than about 10°) with
the normal to the outer limiting membrane, that
is, at nearly grazing incidence with the side walls
of the inner segments, effects corresponding to
total internal reflection may occur. The inner
segments, however, have diameters comparable to
the wavelengths of visible light, so that the
determination of the distribution of radiant inten-
sity within them and within their continuations,
the outer segments, becomes a problem of diffrac-
tion and one of considerable difficulty. Of prin-
cipal importance is the relation between the exter-
nal stimulus in the appropriate region of the field
and the distribution of radiant intensity in the
outer segment, which is the part of the end-organ
containing the actual photosensitive pigment.
97
Both theory and observation (Enoch, 1960;
1967) indicate that the propagation of radiant
energy in the visible range by the elongated struc-
tures of the end-organs is basically a problem
of wave-mode propagation, analogous to the trans-
fer of microwave energy along dielectric wave
guides. Wave-mode propagation in long circular
cylinders, of refractive index and diameter—wave-
length ratio similar to those in the end-organs, is
soluble theoretically (Snitzer, 1961; Snitzer and
Osterberg, 1961; Snyder and Hamer, 1972; Miller,
1974; Snyder, 1975). The complexities of the reti-
nal situation (irradiation by a diffraction pattern,
with phase variation across it, of a closely packed
array of inner segments tapering to outer seg-
ments of smaller diameter with refractive index
variations in the transitional region) have until
now prevented a satisfactory approximation to
the key relation just mentioned.
Two general conclusions relevant to the speci-
fication of internal stimulus that also follow by
the uncritical application of ray optics to sim-
plified models of end-organs, are the following:
(a) The outer segments, despite their smaller
diameter, probably receive nearly all the radiant
flux incident (at normal incidence) on the inner
extremities of the inner segments. This light-trap-
ping or funneling effect may be estimated from
the approximate diameters of the segments to
produce a ratio of the average radiant intensity in
the outer segment compared with that in the inner
segment of some 4:1 for foveal cones and more
for cones in the surrounding retina.
(b) The radiant intensity in the outer segments
may vary strongly according to the direction of
incidence of the radiant flux on the outer limiting
membrane, even in the limited range of angles of
incidence normally accessible in the living eye (up
to about 10° for radiant flux entering the eye near
the edge of the fully dilated pupil).
The known directional properties of visual re-
sponse (Stiles—Crawford effect; see Section 5.11)
provide evidence that the variation noted under
(b) is occurring, at least in the case of cone
end-organs; in fact, one attempt to explain the
directional properties led to the first suggestion of
the wave-guide analogy for the end-organ (Toraldo
di Francia, 1949).
Unlike the previously considered effects of
aberrations and diffraction on the irradiance in
the central area of the retinal image of a uniform
object, the dependence on retinal angle of inci-
98 (Me lzeN Ge
dence of the radiant flux reaching the pigment in
the outer segments is not eliminated by increasing
the angular size of the object. Thus, in specifying
internal stimulus by the retinal irradiance, it is
necessary, in principle, to indicate the angle or
angles of incidence on the retina. This require-
ment is met effectively by stating the size of the
natural pupil, if that is being used, or the size and
position in the natural pupil of the artificial pupil
(real or, in Maxwellian view, virtual).
2.4.1 Image Formation by the Theoretical Eye
Table 1(2.4.1) gives the optical specification of a
representative or theoretical eye as derived by
LeGrand (1946). It applies primarily to the adult
European eye and is based on more recent statis-
tical data than the earlier theoretical eyes of
Helmholtz and Gullstrand. The table also gives
LeGrand’s specification of a simplified eye in
which the refractive indices of the aqueous humor
and vitreous body and the cornea are taken as
identical, and the eye lens is replaced by a thin
lens (zero thickness) in such a way that the prin-
cipal points and the image points are the same as
for the theoretical eye. It may be assumed that the
theoretical eye applies either for white light or for
monochromatic light in the yellow. Figure 1(2.4.1)
illustrates the optical specification of LeGrand’s
full theoretical eye (“unaccommodated complete
eye”) in accordance with the data given in Table
1(2.4.1).
Where differences between individual eyes are
not critical, the theoretical eye can be used to
make approximate calculations by Gaussian optics
of the image, size, position, and so forth of ob-
jects not too far from the direction of vision. In
particular, it gives the value of the quantity m in
Eq. 3(2.4) for retinal irradiance.
If the unit of length is taken as the millimeter
throughout the equation, we obtain
a OS Ry mal Use
ea ean ES
[1(2.4.1)]
for the unaccommodated theoretical eye; that is,
the eye accommodated on a point infinitely dis-
tant, and
m = (14.7762) = 218.34
for the theoretical eye accommodated on a point
distant 143.6 mm. Here N.D. is the distance in
millimeters of the image nodal point from the
retina. With increase in the angle @ that the
direction of the object makes with the visual axis,
the approximations of the Gaussian paraxial the-
ory fail. For the theoretical eye (LeGrand, 1956)
and probably for most actual eyes, the two
astigmatic images of a small object straddle the
retina as they separate with increasing 9. How-
ever, the actual retinal illuminance, for example,
in the image of an object subtending 5 min at the
eye, may be very much [perhaps a tenth at 6 = 60°
(Stiles, 1952)] below the amount to be anticipated
from Eg. 2(2.4) with the above value of m. On the
other hand, for extensive object surfaces subtend-
ing several degrees at the eye, the equation can
still give a crude approximation of the retinal
illuminance.
The optical specification of the theoretical eye
can be used to calculate the size of the retinal
image of an object in the external field.
Take, for example, the simple case of a cir-
cular source whose normal at its center coincides
with the optical axis of the unaccommodated eye,
as illustrated in Figure 2(2.4.1). If o is the angular
subtense or apparent diameter of the source, mea-
sured in radians, the length h’ of the diameter of
the retinal image of the source, located at the
image focal plane at F’ in the retina, is given by
h’ = 16.68320 [2(2.4.1)]
The solid angle w subtended by the circular source
at the eye (nodal point NV) is obtained from
re 4msin? 7 [3(2.4.1)]
For a small angular subtense o, Eq. 3(2.4.1) be-
comes
vents
wd) ~—T70
4 [4(2.4.1)]
The circular area A of the retinal image is given
by
Table 1(2.4.1) LeGrand’s Theoretical Eye

Full theoretical eye Simplified eye

Unaccommodated Accommodated Unaccommodated Accommodated

Refractive index
Cornea L370 1377 1.336 1.336
Aqueous humor 1.3374 1.3374 1.336 1.336
Crystalline lens (total index) 1.42 1.427 1.4208 1.4260
Vitreous humor 1.336 1.336 1.336 1.336

Distance along axis from

corneal pole (mm)
Posterior surface of the
cornea 0.55 0.55 — —
Anterior surface of the lens 3.6 8.2 6.3740 D763
Posterior surface of the lens 7.6 Ts 6.3740 5.7763

Radius of curvature (mm)

Anterior surface of the cornea 7.8 7.8 8 8
Posterior surface of the cornea LS 6.5 — —
Anterior surface of the lens 10.2 6.0 10.2 6
Posterior surface of the lens —6 —5.5 —6 —5.5

Dioptric power
Anterior surface of the cornea 48.3462 48.3462 42 42
Posterior surface of the cornea —6.1077 —6.1077 — —
Anterior surface of the lens 8.0980 14.9333 8.3097 15.0049
Posterior surface of the lens 14 16.5455 14.1265 16.3690

Cornea
Power 42.3564 42.3564 42 42
Position of principal
points : object —0.0576 —0.0576 0) 0)
: image —0.0597 —0.0597 0 0
Focal length: object — 23.6092 — 23.6092 — 23.8095 —23.8095
: image 31.5749 31.5749 31.8095 31.8095

Crystalline lens
Power 21.7787 30.6996 22.4362 31.3739
Position of principal
points : object 6.0218 5.4730 6.3740 5.7763
: image 6.2007 5.6506 6.3740 5.7763
Focal length: object — 61.4087 — 43.5641 -- 59.5466 — 42.5832
: image 61.3444 43.5185 59.5466 42.5832

Complete eye
Power 59.9404 67.6767 59.9404 67.6767
Position of principal
points : object (P) 1.5946 1.8190 1.7858 2.0043
: image (P’) 1.9078 2.1915 1.9078 2.1915
Position of focal
points : object (F) — 15.0887 — 12.9571 — 14.8974 —12.7718
: image (F') 24.1965 21.9325 24.1965 21.9325
Focal length: object (f) — 16.6832 — 14.7761 — 16.6832 — 14.7761
: image (f’) 22.2888 19.7409 22.2888 19.7409
Position of nodal
points : object (N) 7.2001 6.7838 7.3914 6.9691
: image (N’) ess 7.1563 7.5133 7.1563
Accommodation (diopters) 0 6.9633 0 6.9633
Distance along
axis : image nodal
point to image focal point
(at retina), (mm) 16.6832 14.7762 16.6832 14.7762

99
100 THE EYE

Cornea

S20) arto reo ees 0

t
Distance from corneal pole in mm
Fig. 1(2.4.1). Optical specification of LeGrand’s full theoretical eye [see Table 1(2.4.1),
“unaccommodated complete eye”): F = object focal point; F’= image focal point; P = object
principal point; P’= image principal point; N = object nodal point; N’= image nodal point;
f = object focal length; f’ = image focal length.

which, in terms of the angular subtense o, be- subtense or solid angle of external sources in the
comes shape of a line, circle, and square of different
Sizes.
ave 37(16.6832)° 0”

or
A = 218.607
In terms of the small solid angle w, we obtain
A =27896
where w is measured in steradian and A in square
millimeter.
Table 2(2.4.1) gives the relations between the
lengths or areas on the retina and the angular
2.4.2 Eye Axes and Eye Angles
The visual axis of the eye is defined as the line
[5(2.4.1)] joining the point on which the eye is fixed—the
fixation point—to the object nodal point N. The
optical axis—the approximate axis of symmetry
of the refracting surfaces—and the visual axis are
[6(2.4.1)] not generally coincident; they are inclined at an
angle termed the angle a. When the eyes are fixed
on a distant point ahead, with their visual axes
parallel and horizontal, the optical axes are in-
clined outwards by an angle that in the normal
adult eye has an average value of 5° (range about

Source w= 47 sin 224 Retinal Image

F!

4
N

ey =16 6832
Fig. 2(2.4.1). Parameters involved in calculating diameter h’ and area A of the retinal image
of a circular source in the external field. The “infinitely distant” source subtends an angle o
(solid angle w) at the eye (nodal point VV). For the unaccommodated eye, the object focal
- length f is equal to the distance between the image nodal point NV’ and the image focal point F’.
 Factors in the Eye that Control the Internal Stimulus 101

Table 2(2.4.1) Relations between Lengths or Areas on the Retina and Angular Subtense or Solid
Angle of External Sources of Different Shapes and Sizes“

Source Angular Retinal Image

Shape Subtense Solid Angle Length Area

Straight 1.0 rad _ 16.683 mm =

Line 0.05994 rad — LOagemnr
1.0 min — 4.853 wm _
0.2061 min — 10 pm —
1.0 deg — 0.2912 mm —
3.434 deg — 10 mm —
Circle 1.0 min dia 6.65-10 °s — 18.50 (um)?
1.0 deg dia 2.39-10°-*s — 6.65-10° (mm)?
2.0 deg dia 9.57-10° 4s — 2.66-10! (mm)?
10.0 deg dia 2.39-10-? sr — 6.65 (mm)?
64.7 deg dia 1.0 sr a 278.3 (mm)?
3.9 deg dia 3.59-10° 7s — 1.0 (mm)?
Square 1.0 min side 8.46-10 *s — 23.55 (um)*
1.0 deg side 3.05-10 4s ut 84.78 - 10° (um)?
3.44 deg side 3.59-10 3s — 1.0 (mm)?
0.206 min side 3.59-10°?s = 1.0 (um)?

“The numerical values apply to the theoretical eye accommodated on a point at infinity, with the image
nodal point 16.683 mm from the retina. For the theoretical eye fully accommodated (6.963 diopters of
accommodation) with the image nodal point 14.776 mm from the retina, the diameters of the retinal
images given above must be decreased by a factor 0.8857, and the corresponding areas by a factor
0.7845.

4—8°) but which may be as high as 10° or as low
as 0° (or even be slightly negative, corresponding
to inward inclination) for longsighted and short-
sighted eyes, respectively. The optical axes are
also inclined slightly downwards.
The center of curvature of the corneal surface
lies very close to the object nodal point N, so that
lines passing normally through the cornea, in
directions normal to its surface and at points not
too far from the optical axis, also pass through N.
One such line passes through the center of the
apparent pupil—the pupil as seen from outside
the eye by refraction at the cornea—and is termed
the pupillary axis or central pupillary line. Its angle
with the visual axis is called angle x. Angles a and
k are nearly the same and would be identical but
for a slight displacement of the pupil center to the
nasal side of the optical axis.
The eyeball rotates about a point situated ap-
proximately at its center. The line joining the
center of rotation to the fixation point is defined
as the fixation axis, and its angle with the optic
axis as the angle y. For a distant fixation point, y
equals a but diminishes progressively by a small
amount as the fixation point is brought nearer the
eyes.
2.4.3. Chromatic Aberration of the Eye
The refractive indices of the eye media, and hence
the optical power of the eye system, depend on
wavelength. If the image of a distant point source
emitting all wavelengths and located on or near
the optical axis produces a focused image on the
retina for a reference wavelength A, (usually taken
as in the yellow), the image for shorter wave-
lengths lies slightly behind the retina. This axial
chromatic aberration is conveniently specified by
the power (in diopters) of a thin correcting lens
placed at the eye to bring the image of wave-
length A into focus on the retina.
102 THR EYE

Table 1(2.4.3) Axial Chromatic Aberra- Table 1(2.4.3) gives the mean result of two
tion of the Human eye. Mean Results of experimental studies (Wald and Griffin, 1947;
Wald and Griffin (1947) for 14 Ob- Bedford and Wyszecki, 1957). They are in good
servers and Bedford and Wyszecki (1957) agreement both with each other and with the
for 12 Observers results of other investigations. Figure 1(2.4.3)
shows the construction data of a lens designed to
Power (diopters) of Eye-Correction correct the aberration. With this lens in front of
Lens (to bring eye/lens combi- the eye, there remains less than 0.25 diopter aber-
nation for 4 to the same power
ration over the visible spectrum (Bedford and
as for 4) = 578 nm)
Waveleneth ee Wyszecki, 1957).
A (nm) Wald-Griffin Bedford-Wyszecki For object points off the optical axis, chro-
matic aberration causes both defocusing and a
365 —2.63 small displacement in apparent position. Ideally,
389 —1.93
an off-axis point of white light should appear
405 —1.70 —1.63
436 —1.19 —1.20 spread out as a spectrum on a radial line from the
457 —0.97 optical axis, with red on the outside, and of
480 —0.72 length proportional to the angular separation from
492 —0.59 the axis. For a 5° separation, a spectrum length
509 —0.45
(430 to 770 nm) of approximately 3 min of arc is
546 —0.16 —0.20
578 0.00 0.00
computed (LeGrand, 1956). However, for any
588 0.05 actual eye, deviations from axial symmetry mod-
621 0.18 ify the net effect.
636 0.26
644 0.30
668 0.38 2.4.4 The Troland Values of
691 0.40 Retinal Illuminance
700 0.47
750 0.62 In visual investigations, the actual retinal illumi-
nance or irradiance produced by an external
stimulus cannot be directly determined. Instead, a
conventional retinal illuminance of a particular ret-
inal area is defined by taking the product of

(a) the luminance L(0, ¢) in the corresponding

direction (6, ¢) of the external field; and
(b) the apparent area p(0, ¢) of the pupil (natu-
DF 613369 ral or artificial) seen from that direction.

For an eye free of all light losses in the optic

media, the actual retinal illuminance E(6, >)
would be given by |

E(,)=—[L(0,¢)p(9,)] [1(2.4.4)]
where, for an eye approximating the theoretical
eye and for not too large angles 6, the quantity m
has the value 278.3 (unit of length is one millime-
ter), as calculated in Section 2.4.1. For actual
DBC 614599 eyes, these assumptions are not generally valid,
Fig. 1(2.4.3). Design data of a lens to correct but the simple product L - p is still useful as a
axial chromatic aberration of the human eye (Bed- measure of the internal stimulus from which the
ford and Wyszecki, 1957). main effects of pupil variations have been
 Factors in the Eye that Control the Internal Stimulus 103

eliminated. In practice, the unit adopted for the tion 1.2) that
product L-p is always the troland (after its
originator, L. Troland).
A troland is defined as the (conventional) reti- [4(2.4.4)]
nal illuminance when a surface of luminance one
candela per square meter is viewed through a
pupil at the eye (natural or artificial) of area one
where fh and c are Planck’s constant and the
square millimeter. (For “troland,” the now ob-
velocity of light, respectively (see Section 1.2.2.
solete terms “photon” and “luxon” have been
ii).
used.)
Useful relations between P, and troland val-
In addition to the ordinary troland (or pho- ues, and similarly between N, and troland values,
topic troland ), in the definition of which the lumi- are the following:
nance is photopic luminance based on the CIE
V(A) curve, a second troland—the scotopic tro-
land—is sometimes used in which the luminance De ap=eP VAN SOS 2.242 2104
in question is based on the scotopic curve V’(A).
Thus, the troland values T and T’ of a stimulus
with photopic luminance L (cd-m ’), and [5(2.4.4)]
scotopic luminance L’ (cd - m7), respectively,
seen through a pupil of p (mm”) are TAP Pen) 81 10

[2(2.4.4)] [6(2.4.4)]

and
T, =Ly,- p= SAV(A) x 4.454- 107"

ae [3(2.4.4)]
The term “troland value” is preferable to “ret-
inal illuminance,” particularly when the actual
retinal illuminance is also under consideration.
Supplementary notes concerning retinal illumi-
nance and troland value are given in Section
SEN &
Table 1(2.4.4) lists factors to convert photopic
troland values T to scotopic troland values 7’ for
monochromatic stimuli of different wavelengths
as well as for stimuli with an extended spectrum.
Monochromatic stimuli from an emitting sur-
face in a given direction in the field of view are
frequently specified (a) by the radiant flux (or
power) in watts (W) entering the eye per square
degree of external field, or (b) by the number of
quanta per second entering the eye per square
degree of field. Here, radiant flux entering the eye
means flux that would pass the eye pupil in the
absence of any losses by reflection, absorption, or
scattering in the cornea and aqueous humor. If we
denote the spectral concentration of the radiant
flux by P, and the corresponding number of
quanta per second by Nj, we first note (see Sec-
[7(2.4.4)]
N
(an ee a) $1109, 410
[8(2.4.4)]
where p=area (in mm”) of the apparent pupil
(natural or artificial) seen from the given
field direction
L, L’=photopic and scotopic luminances, re-
spectively, of the source in the given
direction (expressed in cd-m™ *)
A =wavelength (in m)
If the stimulus has an extended spectrum of
absolute spectral radiant power distribution
{P, dX}, measured in terms of watts per square
degree of surface area of source (W - deg °),
and of spectral quantum distribution {N, dA},
measured in terms of number of quanta per sec-
ond per square degree of surface area of source
(s_'! - deg 7), the products P,V(A), P,V’(A),
NV(A)/A, NV'(A)/A, in Eqs. 5(2.4.4) to 8(2.4.4)
104 PREY

Table 1(2.4.4) Conversion of Photopic Troland Values T to Scotopic Troland Values 7’ for Monochro-
matic Stimuli of Different Wavelengths A and for a Few Selected Stimuli with an Extended Spectrum

Blackbody Radiators of
Monochromatic Stimuli Different Temperatures

d (nm) fTiiTe) d (nm) Stare} T (K) Fhecadbos

400 58.39 550 1.20 1000 0.265
10 71.67 60 0.822 1250 0.435
20 60.11 70 0.543 1500 0.610
30 42.87 80 0.347 1750 0.781
40 35.51 90 0.215 2000 0.943
450 29.80 600 0.131 2250 1.093
60 2BA2 10 0.0788 2500 esc)
70 18.49 20 0.0481 2750 1.363
80 14.20 30 0.0313 3000 1.481
90 10.82 40 0.0213 3500 1.689
500 7.57 650 0.0157 6000 2.358
10 4.93 60 0.0128 7000 2.521
20 3.28 70 0.0115
30 2.34 80 0.0105 2042 1.000
40 1.70 90 0.0107 2856 1.411
700 0.0108 Daylight D5 2.464

°T’= f(T TT, where f(T> T’) =2.489 V’(A)/V(A) for monochromatic stimuli and f(T> T’)=
2.489/S(A)V(A) dA /fS(A)V(A) dA for stimuli with an extended spectrum, defined by the relative
spectral radiant power distribution {.S(A) dA}. The factor 2.489 is the quotient K/, /K,,, with K’, =1700
Im-W! and K,, = 683 Im-W_! (see Chapter 4).

are replaced by the corresponding integrals €,=Sscotopic retinal illuminance in lumens per
square millimeter (Im-mm7”).
[PVO) dX, [PV'A) dy,
The above quantities are related to one another as
follows:

[(P)vara. [(P)vaya, Ny= Neverrd.034e 1024 [9(2.4.4)]

and 7, and Tx become T and T’ respectively.

The actual retinal irradiance or illuminance in
the image of a field surface emitting monochro-
ef = eV'(A) X 1700 [11(2.4.4)]
matic radiant flux of wavelength A(m) may be
specified by or a | mar x 2.489 [12(2.4.4)]
€,= the retinal irradiance, measured in watts per
square millimeter (W-mm 7”),
Their relations to the other stimulus quantities,
n,=retinal quantum irradiance, measured in
P,, Ny, T,, and Tx of Eqs. 4(2.2.4) to 8(2.4.4), are
number of quanta per second per square mil-
given by
limeter (s_'-mm
”),
€,=photopic retinal illuminance in lumens per
square millimeter (Im- mm”), €, = t(X)P, X 11.80 [13(2.4.4)]
 Factors in the Eye that Control the Internal Stimulus —_ 105

ny, = t(A)N, X 11.80 [14(2.4.4)] directional sensitivity of the retina (Moon and
Spencer, 1944; LeGrand, 1946, 1957; De Groot
€, =1(A)T X 3.593 - 10-9 [15(2.4.4)] and Gebhard, 1952). As directional effects de-
pend on the luminance level and other factors,
e, = t(A)T, X 3.593- 107? ~—-[16(2.4.4)] they are best dealt with separately in the next
section.

&, = ae x §.261- 107 [17(2.4.4)]

& = ae x 2.114-107 [18(2.4.4)]
ny = oe x 2.649-10" —-[19(2.4.4)]
eS wee x 1.064-10'° = [20(2.4.4)]
The eye is assumed to be accommodated on infin-
ity, see Table 2(2.4.1). The factor #(A) represents
the effective transmittance of the eye for wave-
length A, allowing for all preretinal light losses as
given in Table III(2.4.6).
An effective troland that has been considered
takes account of the effect on the internal stimu-
lus of both the variations in pupil size and the
resulting changes in response produced by the
2.4.5 Pupil Size
The natural pupil responds by contraction and
dilation to changes in the field luminance or in
the distance from the object on which the eyes
accommodate and converge, to variations in many
other sensory stimuli (not only visual), and to
psychical influences. Exposed to a steady lumi-
nance level in the field, the pupil tends to fluctuate
in diameter about a value that decreases as the
luminance rises. Although the relation between
pupil diameter and the luminance of a large steady
adapting field shows large individual variations,
as shown in Figure 1(2.4.5), various approximate
mean curves of pupil size against luminance level
have been put forward, mainly to enable visual
observations with natural pupils or with artificial
pupils to be compared through their respective
troland values. A complicating factor is that pupil
size under given conditions is generally lower as
age increases. The following formulae have been

Subject

+>
Ww
oa

(mm)
Diameter
Horizontal
ine)

=—co —2

Log (Trolands)

Fig. 1(2.4.5). Pupil diameter in viewing a 52° uniform field of different troland values. Curves
for 12 subjects under identical conditions (Spring and Stiles, 1948; reproduced with permission
of the editor of the Brit. J. Ophthalmol.).
 106 THE-EYE

LOG (T)

LOG (L)

Fig. 2(2.4.5). Graphs for obtaining log (trolands, 7) from log (luminance, L in cd-m7) and
vice versa, based on data of DeGroot and Gebhard (1952). CurveA: log T = logL+ 6; Curve B:
log L = logT— 6. (No allowance has been made for the Stiles—Crawford effect.)

proposed: high value (about 320 cd - m 7) compared with

dilatation following the reverse transition [Table
d= 4.9 — 3 tanh[0.4(log L + 1.0)] 1(2.4.5)].

(Moon and Spencer, 1944)

[1(2.4.5)] Table 1(2.4.5) Speed of Opening and

Closing of Eye Pupil. Mean Data of
log d = 0.8558 — 4.01 - 10~4(log L + 8.6)” Reeves (1920) for Six Observers

(De Groot and Gebhard, 1952) Opening in the dark

after steady adaptation Closing on exposure
([2(2.4.5)] to a large field of lumi- _ of adark-adapted eye
nance 320 cd-m-? to the same bright
(white light) field
In the above formulae, the diameter d is given in
mm, and the field luminance L in cd - m7”. Mean Pupil Mean Pupil
Figure 2(2.4.5) gives a graph for the rapid Time Diameter Time Diameter
conversion of log (external luminance L) to log (seconds) (mm) (seconds) (mm)
(photopic troland value T), and vice versa, in the
0 2.9 0 8.0
case of an average eye whose pupil varies with
0.5 34 0.1 7.8
luminance level according to the mean results of 1.0 ao 0.2 7.6
various investigations compiled by De Groot and PS a 0.3 Cit|
Gebhard (1952). The data do not take into account 3 4.7 0.4 6.8
the Stiles—Crawford effect. > 5.6 0.5 6.4
The speed of opening or closing of the pupil 9 6.4 1.0 a3
15 6.6 AD 4.4
following an abrupt change in the luminance level
29 6.9 2.0 3.2
to which the observer is adapted is sometimes 60 me rie ai
relevant in visual observations. Average data for 180 7.6 3.0 ond
six observers (Reeves, 1920) illustrate the greater 2.5 2
rapidity of pupil contraction on raising the lumi- 4.0 32
nance level from darkness to a moderately 4.5 Sel
 Factors in the Eye that Control the Internal Stimulus 107

Table 2(2.4.5) Apparent Pupil Viewed from Different Angles. Mean Data of Spring and
Stiles (1948) for 13 Observers

Large pupil Small pupil

6 Mean Horizontal Ratio Horizontal/Vertical Apparent Area Ratio Horizontal/Vertical
(degrees) Diameter (mm) Diameters (mm)? Diameters

20 nasal 7:13 0.90 44.4 0.91

0 7.69 0.98 47.3 1.00
35 temporal 7.01 0.90 43.1 0.90
55 temporal Sede 0.72 35.9 0.73
70 temporal 4.43 0.52 28.3 0.53
80 temporal eps 0.40 20.9 0.40
90 temporal 1.88 0.23 12.0 0.28
100 temporal 0.75 0.10 4.53 0.15
105 temporal 0.25 0.03 1.46 0.07

6 = angle of view, with respect to visual axis, in a horizontal plane through the eye.

In a steady but nonuniform field, such as an
isolated light patch, the contraction of the pupil
below its maximum size becomes less as the light
patch moves from the center to the periphery of
the visual field. Formulae for the effect are given
by Crawford (1936).
The efficiency of monochromatic luminances
of different wavelengths in producing a given
(steady-state) contraction of the pupil corre-
sponds approximately to the scotopic luminous
efficiency curve V’(A). However, in pupil re-
sponse to flashes, both rod and cone response
systems are involved (Bouma, 1965). Bouma’s the-
sis gives a great deal of experimental material on
the subject covered in this section and includes a
good bibliography.
The dependence of the shape and size of the
apparent pupil for stimuli in the peripheral field
are summarized in the mean experimental data of
Tables 2(2.4.5) and 3(2.4.5).
2.4.6 Light Losses in the Eye
In Section 2.4, the fraction of incident radiant
flux was introduced that reaches photosensitive
visual pigment in the outer segments of the retinal
end-organs. The quantity [1 — /(@, ¢, A)] in Eq.
2(2.4) is the product of two factors, t, the fraction
reaching the external limiting membrane, and g,
the proportion of this flux actually reaching the
visual pigment.
The factor g depends on the amount of radiant
flux getting through to the pigment epithelium at
the back of the eye without traversing visual
pigment. This in turn is bound up with the precise
wave-mode distribution of radiant energy in and
around the end-organs. Variation of g with the
retinal angle of incidence of the radiant flux is
almost certainly responsible for the directional
properties of the visual response. For normal
incidence, g is commonly taken as independent of

Table 3(2.4.5) Factor by Which Area of the Apparent Pupil is reduced in moving away
from the Optical Axis. Mean Data of Spring and Stiles (1948) for 13 Observers

0’ Apparent Pupil Diameter (mm) Viewed

(degrees) Along Optical Axis = 8 6 4 2
1.00 1.00 1.00 1.00
0.94 0.94 0.94 0.94
0.76 0.76 0.76 0.76
0.44 0.44 0.44 0.44
0.25 0.27 0.29 0.30

6’ = angle between object direction and optical axis.

108 THEEYE

wavelength and equal to a constant that is less

than, but comparable with, unity.
The factor ¢ is less complex, and approximate
estimates of its magnitude and wavelength depen- Wald (1945)
|Rod vision} Aphakic
Wright (1951 )
dence can be made. It depends on the following Wald (1945) -Cone vision
Comparisons

sources of loss: Ludvigh and McCarthy Direct

(1938) (Densities include |Measurements
all optic media) on Dead

(a) absorption and scattering in the optic media, x Weale (1954) Material

and reflections at surfaces between the media,

(b) absorption and scattering in the retinal layers
preceding the outer limiting membrane.

Scattered light from the radiant flux forming an

image on a particular small spot of the retina may
reach parts of the retina outside the image spot.
Because it is better to consider this phenomenon
separately (see below), all radiant flux not form-
ing the regular image is included under (a) or (b)
as lost flux. Reflections at interfaces (air—cornea,
cornea—aqueous humor, aqueous humor-lens, and 700)
[LOG
"a%,)-
LOG
DIFFERENCE
DENSITY
OPTICAL
so forth) are very little dependent on wavelength
and are certainly very small, amounting in all to a 300 400 500 600 700 800
loss of some 3 to 4%. WAVELENGTH (nm)
Single reflections give rise to images (Purkinje Fig. 1(2.4.6). Various determinations of the opti-
images) visible outside the eye, and are invaluable cal density of human eye lenses relative to that at
for probing eye properties (see, for example,
A = 700 nm. The solid curve is an attempt at a
Bennett and Francis, 1962).
compromise curve in which reliance is placed on
Light reflected at two interfaces, and hence
the aphakic data for short wavelengths (365 to 436
very much weaker than single reflections, reaches
nm) and on the excised eye data for the rest of the
the retina but is only detectable under special
spectrum. Numerical values for the curve are given
conditions (see Le Grand, 1946).
in Table 1(2.4.6).

(i) Light Losses in the Lens. The eye lens is

principally responsible for losses under (a). The
light losses are markedly dependent on the age of 1S,
the eye, and at very short wavelengths, they be-
come very large. The spectral transmittance ¢,(A) A ,(A, 700) = bel 5 |a oe| |
of the human eye lens has been estimated by:

(1) measurements on excised human eyes,
(2) comparison of the visual spectral sensitivities
of normals, with those of aphakics (persons with
the eye lens removed by operation),
(3) an objective method (Said and Weale, 1959)
in which the light reflected at the lens—vitreous
surface (after passing twice through the eye lens)
is compared in amount with the light reflected at
the aqueous—lens surface (which has not traversed
the eye lens).
The relative spectral transmittance is the quantity
of most interest in correcting spectral response
curves to make them refer to retinal response. The
curve of Figure 1(2.4.6) is an attempt at a com-
promise curve in which reliance is placed on the
aphakic data for short wavelengths (365 to 436
nm relative to 700 nm) and on the excised eye
data for the rest of the spectrum. Numerical
values are listed in Table 1(2.4.6). The following
limitations must be noted:

The principal data, as they were available up (a) Sensory data show at short wavelengths, large
to about 1960 by methods (1) and (2), are sum- variations (of the order of 0.5 log units at 400 nm)
marized in Figure 1(2.4.6) which shows the dif- normally attributed to differing light losses in the
ference of lens density at \ = 700 nm and A; that lens which occur even in the same age group. A
 Factors in the Eye that Control the Internal Stimulus 109

Table 1(2.4.6) Optical Density Differences of the Young Human Eye Lens (Completely Open Pupil)
as a Function of Wavelength’

\(nm) A (A, 700)? A’,(A,700)° A(nm) A (A, 700)? A’,.(A,700)°

350 550 0.070 0.070
60 BT 60 0.060 0.060
70 2.70 70 0.050 0.050
80 215 2.61 80 0.040 0.040
90 1.66 2.02 90 0.031 0.031
400 1.20 1.45 600 0.024 0.024
10 0.82 0.96 10 0.018 0.018
20 0.55 0.59 20 0.012 0.012
30 0.37 0.37 30 0.008 0.008
40 0.27 0.27 40 0.004 0.004
450 0.225 0.225 650 0.002 0.002
60 0.195 0.195 60 0.000 0.000
70 0.175 0.175 70 0.000 0.000
80 0.155 0.155 80 0.000 0.000
90 0.140 0.140 90 0.000 0.000
500 0.125 0.125 700 0.000 0.000
10 0.110 0.110 10 0.000 0.000
20 0.100 0.100 20 0.000 0.000
30 0.090 0.090 30 0.000 0.000
40 0.080 0.080 40 0.000 0.000

“To convert the tabulated density differences to actual densities, add 0.15 to all given values. To make
the tabulated density differences applicable to a small pupil size, multiply all values by the factor 1.16.
"A (A, 700) = difference of lens density at X= 700 nm and A; that is, A,;(A,700) = log[{1/r,(A)]—
log[1/t,(700)], where ¢,(A) is the spectral transmittance of the lens. The values of 4,(A,700) are those
taken from the curve of Fig. 1(2.4.6) which represents a compromise curve placed through various
experimental data points, and which is proposed as a standard for the selective light losses in the young
human eye lens.
‘A’, (A, 700) has the same meaning as A,(A,700) but has been deduced by Norren and Vos (1974) from
the CIE 1951 scotopic luminous efficiency function V’(A) and Dartnall’s standard shape of the spectral
absorptance of visual pigments. The difference between A’, and A, is in the spectral region below 430
nm.

single representative correction curve is therefore
applicable, at best, only to mean data for many
observers.
(b) It is not clear to what age group the repre-
sentative curve of Figure 1(2.4.6) should apply.
From the ages of the normal observers used in the
aphakic comparisons that determine the big cor-
rections at very short wavelengths, an age range
of 20 to 30 could be assumed, but the variation at
longer wavelengths is derived in part from data
for older eyes.
(c) The objective method (3), which in many
respects may be considered the most satisfactory,
gives results considerably at variance with those
of Figure 1(2.4.6), notably in the density at 400
nm (relative to 700), which comes out much lower
for all observer ages. This can be seen from the
data of Said and Weale (1959) shown in Figure
2(2.4.6).
Norren and Vos (1974) critically reviewed
our proposed compromise curve of A,(A, 700)
given in Figure 1(2.4.6) by comparing it with
more recent experimental data, notably those of
Geeraets et al. (1960), Boettner and Wolter (1962),
Alpern et al. (1965), Cooper and Robson (1969),
Mellerio (1971), and Tan (1971). They also derived
additional data on spectral transmittances of the
FO -nSHEEYE

0.7

0.6

0.5

oO in

ow
Density
Optical

0.2

0.1

400 500 600 700

Wavelength, nm

Fig. 2(2.4.6). Spectral density curves of the human eye lens determined for the living eye by
an objective method. Data for English observers of different ages (ages are shown against the
curves; from Said and Weale, 1959). The crosses refer to mean data for two eyes (ages 48 and
53) obtained, after their removal in operations, by a different method (Weale, 1954).

ocular media from the CIE standard scotopic
luminous efficiency curve V’(A) and the spectral
absorptance curve of rhodopsin as defined by
Dartnall’s standard shape of spectral absorptance
of visual pigment (see Section 8.2.3).
Norren and Vos (1974) conclude that the more
recent experimental data agree well with the earlier
data for wavelengths longer than 420 nm; but for
wavelengths of 420 nm and below, somewhat
higher optical densities are observed. The spectral
optical densities derived by Norren and Vos from
the V’(A) function tend to agree better with some
of the more recent data.
The discrepancies between the old data and
most of the new in the short-wavelength region
are not large when one considers the difficulties
associated with measurements of this kind. In
fact, in view of the uncertainties inherent in lens
transmittance measurements and population sam-
pling, we propose to retain the compromise curve
established in the first edition (1967) of this book
and which is shown here in Figure 1(2.4.6) and
tabulated in Table 1(2.4.6).
However, in Table 1(2.4.6) we also give the
data A‘,(A,700) that Norren and Vos (1974)
deduced from the CIE 1951 scotopic luminous
efficiency function V’(A) and Dartnall’s standard
shape of the spectral absorptance of visual pig-
ment in the wavelength region from 380 to 430
nm. The remaining spectral data above 430 nm
are, according to Norren and Vos, those given as
A (A, 700). Figure 3(2.4.6) illustrates the results
of Norren and Vos’ calculations of A’,(A, 700)
together with the range of variation (95%) from
the average they found for the individual curves
of 50 young observers that took part in Crawford’s
(1949) experimental determination of the V’(A)
function. These variations amount to approxi-
mately +25% from the average and are indicative
of the extent to which individual eye lenses of the
17 to 30 year age group may differ from the
average of that group.
The values given in Table 1(2.4.6) and the
corresponding curves plotted in Figures 1(2.4.6)
and 3(2.4.6) are all given relative to the optical
density at A = 700 nm. The optical density in
absolute terms is difficult to determine with cer-
tainty and mainly depends on intraocular scatter-
ing and pupil size. Norren and Vos (1974) report
that the experimental data in the literature sug-
 Factors in the Eye that Control the Internal Stimulus 111

A@,
(X00)

300 400 500

Wavelength (nm)
Fig. 3(2.4.6). Optical density differences A; (A, 700) of human eye lens deduced by Norren
and Vos (1974) from the CIE 1951 scotopic luminous efficiency function V’(2) and Dartnall’s
standard shape of spectral absorptance of visual pigment. The hatched area indicates the extent
to which individual eye lenses of the 17- to 30-year age group may differ from the average of
that group.

gest that a density value of 0.15 should be added
to all the values given in Table 1(2.4.6) to convert
the tabulated relative values to absolute
values.
The values in Table 1(2.4.6) are probably most
applicable to the completely open pupil. For small
pupil sizes, the spectral densities will change
somewhat because of the varying properties (in
particular, thickness) of the lens across the pupil
opening (Weale, 1961). Norren and Vos (1974)
estimate that for a small pupil, the values given in
Table 1(2.4.6) can still be used with good ap-
proximation at all wavelengths, provided they are
multiplied by the factor 1.16.

0.8

0.6

0.4

Density
Optical

350 400 450 500

Wavelength d (nm)
Fig. 4(2.4.6). Spectral absorptance (or optical density) of crystalline lutein or leaf xanthophyll
in chloroform (continuous line) compared with the spectral absorptance of human macular
pigment (in chloroform) obtained from nine human maculae (Wald, 1945b; 1949).
112 THREYE

The effect of age on the optical density of the Table 2(2.4.6) Optical Density of the Macular
human eye is indicated in Figure 2(2.4.6) illustrat- Pigment Covering the Fovea as a Function
ing the data obtained by Said and Weale (1959). of Wavelength?’
These data together with more recent data pro-
vided by Tan (1971) and Coren and Girgus (1972) A(nm) Au(A) A(nm) Ay(A)
suggest that only minor changes occur from age
400 0.085 475 0.410
5 to 30 years; but after 30 years, the process
05 0.120 80 0.415
of yellowing becomes increasingly more pro-
10 0.160 85 0.420
nounced.
15 0.225 90 0.410
(ii) Light Losses in the Macular Pigment. The 20 0.300 LB 0.360
nonphotosensitive yellow pigment present in the
425 0.345 500 0.275
central area of the retina, known as the macula
30 0.365 05 0.195
lutea (yellow spot) and principally responsible for
33 0.380 10 0.130
the absorption of radiant energy on its way to the
40 0.400 15 0.085
outer limiting membrane, has been identified as a
45 0.425 20 0.050
carotenoid pigment, xanthophyll. In chloroform
solution, this has the density spectrum shown in 450 0.460 525 0.025
Figure 4(2.4.6), with which is compared the spec- 55 0.490 30 0.010
trum of a similar solution of yellow pigment 60 0.495 35 0.000
extracted from human maculae (Wald 1945b). 65 0.470 40 0.000
Psychophysical measurements showing that 70 0.445 45 0.000
such a yellow pigment exists in the living retina
are of two main kinds: “The optical density A,,(A) of the macular
pigment has been taken from the continuous
(1) A spectral sensitivity, for example, the re- curve of Fig. 5(2.4.6) assuring that the representa-
ciprocal of the absolute threshold, is determined tive maximum density, occurring at A = 458 nm,
at the fovea and at an extrafoveal point where the has a value of 0.5.
amount of pigment present is much less and

0.75
wald (1949)
Brown and Wald (1963)
Stiles (1953)
De Vries, Spoor and Jielof (1953)
Naylor and Stanworth (1954)
pexi)
oo Ruddock (1963)

0.50

DENSITY
OPTICAL
e

\ O°
°
g rs
0.00
400 450 500 550 600

WAVELENGTH A (nm)
Fig. 5(2.4.6). Variations of the relative optical density of macular pigment with wavelength
according to determinations by different psychophysical methods: (a) Relative visual cone
sensitivities in the foveal and extrafoveal retina (Wald, 1949; Stiles, 1953; Brown and Wald,
1963); (b) Color matching in the foveal and extrafoveal retina (Ruddock, 1963); (c) Haidinger
brushes (DeVries et al., 1953; Naylor and Stanworth, 1954). For comparison, the densities are
adjusted to have maximum values of approximately 0.5. A preferred “mean curve” is drawn
giving greatest weight to the less difficult measurements.
 Factors in the Eye that Control the Internal Stimulus 113

possibly nil. If the relative spectral sensitivities
underlying the pigment-containing layer are as-
sumed to be the same, the ratio of the thresholds
(extrafoveal /foveal) should correspond to the rel-
ative spectral transmittance of the macular pig-
ment at the fovea (Wald, 1949b; Stiles, 1953).
Color-matching functions determined in foveal
and extrafoveal retinal areas can also be used
effectively to deduce the spectral transmittance
curve of the macular pigment (Ruddock, 1963).
(2) In this method, use is made of the fact that
an observer viewing a uniform white-light field in
which the radiant energy is plane-polarized sees,
in the area imaged on the fovea, a faint “cross”
(about 3° diameter) made up of two opposed
yellowish “brushes,” the direction defined by the
incident light. This phenomenon, referred to as
Haidinger’s brushes, is explained if (a) the upper
retinal layer contains dichroic blue-absorbing pig-
ment molecules that absorb maximally when the
electric vector is parallel to their long axes, and if
(b) the molecules are oriented at each point P of
the retina with their long axes perpendicular to
the line joining P to the foveal center. By measur-
ing the relative apparent brightness of two stimuli
with polarizations at right angles, imaged in a
narrow elongated field through the fovea, ap-
proximate determinations can be made of the
spectral density of the oriented pigment (De Vries,
Spoor, and Jielof, 1953; Naylor and Stanworth,
1954). Not all the pigment molecules may be
oriented, or the orientation may not be complete
so that this method gives in the first instance only
the shape of the spectral absorption curve of the
pigment.
Examples of the data obtained by the methods
(1) and (2), described above, are shown in Figure
5(2.4.6).
There is good agreement that the form of the
spectral absorptance (or optical density) of macu-
lar pigment is of carotenoid type, has maximal
absorption at about 460 nm, and is not certainly
distinguishable from that of xanthophyll, which
has nill absorption for A greater than about 520
nm. If this is accepted, though there are some
data, notably those obtained by Ruddock (1963),
that suggest deviations from the xanthophyll ab-

LOG(ife
)

0.5 1.0 Re)

LOG(@°)
Fig. 6(2.4.6). Variations of the log ratio of scattered luminance L to eye illuminance E with
angular distance 0° from the source of the scattered light. Curve A: direct measurements
(LeGrand, 1937); straight line B: indirect measurements assuming that the increase in the
increment threshold near the image of the source arises solely from scattered light (L /E =
k /07).
114 tHEEYE

sorptance curve, particularly at the longer wave- affect vision is that of the eye viewing a small
lengths, absolute pigment densities are obtainable bright source against a completely dark back-
from psychophysical measurement on the follow- ground. The rod and cone receptors at a retinal
ing lines. location P lying well outside the image of the
Suppose the relative spectral sensitivity (re- source may receive light scattered by the rays
ciprocal of absolute threshold, for example) would forming the image in their passage through the
be the same at the fovea O and at an extrafoveal cornea, lens, aqueous and vitreous humors, or
point P but for the presence of macular pigment. light diffused from the intensely illuminated spot
The difference (Ay, 9— Ay, p) of pigment den- of the retina (and underlying pigment epithelium,
sity at O and P for wavelength 460 nm is then etc.) where the image is formed. The resultant
given by total distribution of scattered light produces ob-
servable visual effects from which the approxi-
mate magnitude of the scattered light (insofar as
R520
Au.o vy Awu,P = log R, [1(2.4.6)]
60 it reaches visual receptors) can be estimated. Ob-
jective measurements of scattered light on excised
where R4¢. and R59 represent, respectively, the animal eyes are difficult to make and interpret
ratio (observed sensitivity at O/observed sensitiv- (Boynton, Enoch, and Bush, 1954; De Mott and
ity at P) taken at wavelengths 460 nm and 520
nm (or for 520 nm, some other wavelength > 520
nm). It is not assumed that the absolute spectral
sensitivities in the absence of pigment would be
the same at the foveal and extrafoveal points. If P
lies far enough from the fovea for Ay p to be
negligibly small, the method gives the absolute
density Ay,9 of the foveal pigment for an un- o Flamant
polarized stimulus of wavelength 460 nm. From e JK (8mm)
estimated values of Ay 9, mainly credited to
Wald and his collaborators, 0.5 may be taken as a
representative figure, although wide individual
variations occur from more than 1.0 to negligibly
small. There is little, if any, systematic increase of
macular pigmentation with age (Weale, 1963).
Table 2(2.4.6) gives the numerical values of the
mean optical densities plotted as a curve in Fig-
ure 5(2.4.6).
The assumption that the relative spectral sensi-
tivity is the same at the fovea and at an extrafo- (percent)
Illuminance
Peak
veal point, but for the presence of the macular
pigment, can be only a crude one at best. Several
investigations concerning the sensitivities of the
cone mechanisms for foveal and peripheral re-
gions of the retina strongly suggest that these
sensitivities change with retinal region (Weale,
1951a,b; 1953a; 1956; Stiles, 1955b; Wooten and
Wald, 1973; Wooten, Fuld, and Spillman, 1975;
Stabell and Stabell, 1980). However, it is difficult
2 4
to extract from the many and often diverse data External Angle (minutes)
the information needed to improve, at this time,
Fig. 7(2.4.6). Variation of illuminance in the reti-
on the numerical values of the mean optical den-
nal image of a fine line source (thickness 0.5 min)
sities of the macula pigment given in Table 2(2.4.6)
with distance in minutes (external angle) from the
and illustrated in Figure 5(2.4.6).
central line of the image with pupil diameter of 3
(iii) Scattered Light in the Eye. The typical mm. The line represents the variation e °7*! (from
situation in which light scattered in the eye may Westheimer and Campbell, 1962).
 Factors in the Eye that Control the Internal Stimulus 115

Boynton, 1958). Salient points are the following:
(1) Measurements by a special photometric
method of apparent brightness in the scattered-
light distribution around a small source, produc-
ing an illuminance F at the plane of the eye’s
pupil, show that at an angular distance @ (external
angle = 1°) from the source, the retinal illumi-
nance by scattered light corresponds to an exter-
nal luminance L where the log ratio, log (L/E), is
plotted against log 0 in Figure 6(2.4.6), curve A
(Le Grand, 1937; 1956). By integration it is esti-
mated that some 10% of the light incident on the
eye appears as scattered light outside the main
image. (The foregoing applies to young eyes.)
(2) If the equivalent background luminance
around the source, as determined by threshold
methods (see Section 7.8), is attributed wholly to
scattered light, the ratio L/E would equal k/0”,
with k = 10, n = 2 as representative values (young
eyes) [straight line B in Figure 6(2.4.6)]. Methods
1 an 2 agree in fixing at least the order of magni-
tude of scattered light in young eyes.
(3) There is no evidence of an appreciable col-
oration of the scattered light; that is, the amount
is about the same fraction of the source intensity
for different wavelengths in the visible.
(4) The contributions of preretinal media (par-
ticularly cornea and lens) and of the fundus to the
total scattering are probably of the same order of
magnitude (Vos, 1963a; Vos and Boogaard, 1963;
Vos and Bouman, 1964; Boynton and Clarke,
1964).
As the angle @ becomes very small, the
scattered-light distribution must merge finally into
the source image proper which itself is spread into
a distribution by diffraction, aberrations, and im-
perfections in the optical imagery of the eye. The
latter distribution across the image of a fine slit
has been measured by a photographic method
(Flamant, 1955) and later by photoelectric meth-
ods (Westheimer and Campbell, 1962; Krauskopf,
1962). All the results are substantialy concordant.
The line-spread function f(x), for an observer using
a 3-mm pupil, as measured by Westheimer and
Campbell, is shown in Figure 7(2.4.6). It corre-
sponds well to a retinal illuminance falling off as
e °l*| where x is expressed in minutes of exter-
nal angle.

1.00

0.75

0.50

0.25

units)
(arbitrary
Power
Radiant

Be
500 550 600

Wavelength d (nm)
Fig. 1(2.4.7). Spectral distribution of fluorescent light from the excised eye lens of the rabbit
and ox under ultraviolet irradiation (365 nm) (from LeGrand, 1948). The dots represent mean
results obtained by a photographic method; the crosses represent mean results obtained by a
visual method. The latter method is more precise in the region of longer wavelengths, the
former is more precise in the region of shorter wavelengths. The curve is considered the best
estimate of the true mean results.
ll6: THE:BYE
In more recent years, the modulation transfer
function has been used as an alternative way of
specifying image/object relationships in the eye.
The modulation transfer function is based on the
proposition that an object in the form of a
sinusoidal grating (i.e., a grating with luminance
variations that are sinusoidal with distance at
right angles to the grating lines) is always imaged
as a sinusoidal grating, but with reduced contrast
(Westheimer, 1972). The specification of the eye’s
imaging properties can then be given in a graph
that shows the demodulation of sine-wave grat-
ings as a function of their spatial frequency.
(iv) Summary of Light Losses in the Eye. The
mean data on light losses in the eye described in
subsections (i) to (iii) of this section have been
pooled in Tables 1(2.4.6) to IV(2.4.6) in the Ap-
pendix for easy reference and convenience in
estimating the internal stimulus or correcting ob-
served visual spectral sensitivity curves to curves
corresponding to no light losses in the ocular
media.
2.4.7 Fluorescent Light in the Eye
The fluorescence of the human eye lens is caused
by ultraviolet radiant energy (A < 400 nm). It
produces a general veil of bluish light over the
retina. The spectral composition of the fluo-
rescent light probably agrees with the spectral
distributions— all similar—determined for the eye
lenses of sheep, rabbit, and ox, by LeGrand
(1948); see Figure 1(2.4.7).
The retina is also fluorescent to ultraviolet and
possibly to short-wavelength visible light, but in
the normal eye, absorption in the eye lens removes
most of the exciting radiant energy before it can
reach the retina. Aphakic observers (eye lens
removed) are able to see stimuli of wavelengths
much deeper in the ultraviolet than are normals,
the color seen in the range 360 to 310 nm being
blue, not violet (Gaydon, 1938). This change of
color may be the visual response to a mixture of
bluish-green fluorescent radiant flux with the
original ultraviolet stimulus; however, this is con-
troversial.
 CHAPTER 3
COLORIMETRY
3.1 PREAMBLE
Colorimetry is the branch of color science con-
cerned, in the first instance, with specifying
numerically the color of a physically defined vi-
sual stimulus in such a manner that:
(a) when viewed by an observer with normal
color vision, under the same observing conditions,
stimuli with the same specification look alike,
(i.e., are in complete color-match),
(b) stimuli that look alike have the same specifi-
cation, and
(c) the numbers comprising the specification are
continuous functions of the physical parameters
defining the spectral radiant power distribution of
the stimulus.
The experimental laws of color matching as
summed up in an empirical generalization, which
we will refer to as the trichromatic generalization,
provide the foundation for any system of colorim-
etry meeting these requirements. The concepts
and terms currently employed in colorimetry are,
in fact, to a large-extent bound up with the
trichromatic generalization.
Colorimetry is also concerned with the specifi-
cation of small color differences that an observer
may perceive when the differences in the spectral
radiant power distributions of the given visual
stimuli are such that a complete color match is
not observed. For this purpose, color-difference
formulae are used, which in current colorimetric
practice are derived from a variety of different
blocks of experimental data.
The CIE Colorimetric System comprises the
essential standards and procedures of measure-
ment that are necessary to make colorimetry a
useful tool in science and technology.
3.2 BASIC COLORIMETRIC CONCEPTS
The colorimetric concepts considered to be most
basic and of most general use are discussed in this
section. They provide the framework for the CIE
Colorimetric System to be discussed in Section
3.3. For the convenience of the reader, the colori-
metric terms and their definitions, as they are
commonly used to describe succinctly the basic
concepts, are collected in Table I(3.2) of the Ap-
pendix. This table is not complete; other more
comprehensive listings may be found in the cur-
rent third edition of the International Lighting
Vocabulary (CIE, 1970). The CIE has a fourth
edition in preparation.
The terms given in Table 1(3.2) are classified
as psychophysical terms of color and as such refer
to color matching of two visual stimuli typically
presented in the two halves of a bipartite visual
field, and to judgments of similarities and degree
of difference between the two stimuli. The psy-
chophysical terms are distinguished from psycho-
logical terms of color, such as hue, saturation, and
brightness, which apply to visual concepts that
enable the individual observer to describe color
perceptions. Chapter 6 on Uniform Color Scales
deals with such concepts.
3.2.1 Trichromatic Generalization
The experimental laws of color matching are
summed up in the trichromatic generalization. This
states that over a wide range of conditions of
observation, many color stimuli can be matched
in color completely by additive mixtures of three
fixed primary stimuli whose radiant powers have
been suitably adjusted. Other color stimuli have
to be mixed with one of the primary stimuli
before a complete color match with a mixture of
Liv
118 COLORIMETRY
the other two primary stimuli can be obtained.
For some sets of primary stimuli, there remain yet
other color stimuli that have to be mixed with two
of the three primary stimuli before a color match
between this mixture and the third primary
stimulus can be obtained. However, every color
stimulus can be completely matched in color in
one or another of these ways in terms of three
fixed primary stimuli whose radiant powers can
be adjusted by the observer to suitable levels. The
choice of three primary stimuli, though very wide,
is not entirely arbitrary. Any set that is such that
none of the primary stimuli can be color matched
by a mixture of the other two may be used. The
different dispositions of the primary stimuli and
their relevance to trichromatic matching will be
discussed later [see Figures 5(3.2.3) and 1(3.2.4)].
The precise meaning of the term additive mix-
ture in the above statement on trichromacy is
made explicit by the following definition.
Additive mixture means a color stimulus for
which the radiant power in any wavelength inter-
val, small or large, in any part of the spectrum is
equal to the sum of the powers in the same
interval of the constituents of the mixture, con-
stituents that are assumed to be optically incoher-
ent.
If, in addition to the weaker or qualitative
trichromacy principle enunciated in the foregoing
paragraph, the results of color matching are also
assumed to obey certain linearity laws, a stronger
quantitative form of the trichromatic generaliza-
tion is obtained. To these linearity laws of pro-
portionality and additivity may be added two
others, the laws of symmetry and transitivity,
which, although they are generally tacitly as-
sumed, it is desirable to make explicit. The four
laws may be stated as follows:
(i) Symmetry Law.
If color stimulus A matches color stimulus B,
then color stimulus B matches color stimulus A.
(1) Transitivity Law.
If A matches B and B matches C, then A
matches C.
(i) Proportionality Law.
_If A matches B, then aA matches aB, where a
is any positive factor by which the radiant power
of the color stimulus is increased or reduced,
while its relative spectral distribution is kept the
same.
(iv) Additivity Law.
If A, B, C, D are any four color stimuli, then if
any two of the following three conceivable color
matches
A matches B, C matches D,
and (A + C) matches (B + D)
holds good, then so does the remaining match
(A + D) matches (B + C)
where (A + C), (B+ D), (A+ D), (B+ ©) de-
note, respectively, additive mixtures of A and C,
B and D, A and D, and B and C.
The stronger form of trichromatic generaliza-
tion of color matching as formulated above is a
concise statement of what is implied in Grassman’s
three laws of additive color mixture (Grassman,
1853). Different formulations of Grassman’s laws
can be found in the literature. An example of a
traditional formulation has been given by Judd
and Wyszecki (1975). A mathematically formal
and detailed exposition of Grassman’s laws has
been presented by Krantz (1975a).
Three considerations, ignored in the above ex-
position of the trichromatic generalization, are:
(a) the dependence of a match on the observa-
tional conditions under which the two color
stimuli are compared,
(b) the possible effects on a match of different
previous exposures of the eyes to light,
(c) differences in the color matches made by
different observers.
All three factors as well as other aspects of
color matching will be discussed in some detail in
Chapter 5. However, a brief account of the main
factors may be appropriate in this section.
(i) Observational Conditions. In matching, the
two stimuli are normally presented as contiguous
light patches of similar shape and area. However,
a color match of two stimuli of different spectral
radiant power distributions, valid for a given ob-
server when looking directly at the center of the
matching field, will not generally remain valid if
the observer looks to the side; and changing the
area of the matching field may also upset the
match. These viewing conditions must befixed if
critical color measurements are to be made.
In practical colorimetry, direct viewing of the
matching field is of principal interest, and can be
assumed unless it is expressly indicated otherwise.
However, the angular size of the matching fields
used varies widely, and this factor has been taken

into account in recommendations made by the
CIE, as will be seen in Section 3.3.2.
(ii) Previous Exposure of the Eye to Light. The
effect of previous exposure of the eye to light is
dealt with in an extension of the trichromatic
generalization known as the persistence of color
matches. According to this, a color match is unaf-
fected by the previous exposure to white or col-
ored light or by varying light adaptations of the
eye making the match provided these potentially
disturbing factors are not calculated to affect
differently the areas of the retina on which the
two color stimuli are imaged. It is not implied,
and it is not true, that the color perception of the
two matched stimuli remains unchanged or that
the precision of matching is unaffected. The per-
sistence property holds for pre-exposures to mod-
erate or fairly high brightnesses, but it certainly
breaks down for very high brightnesses.
There is a further limitation if the image of the
matching fields extends beyond the rod-free area
of the retina. In that case, color matches of mod-
erately intense stimuli with different spectral
radiant power distributions may be upset if both
color stimuli are reduced by the same fractional
factor, or if, by other changes in conditions, the
level of adaptation is lowered so that the rod
mechanism of vision can make an effective contri-
bution towards determining the match.
(iii) Different Observers. Not all observers
need three independent primary stimuli to make
additive color matches with an arbitrary color
stimulus. For a small but important proportion
(about 2.6% of males, less than 0.05% of females),
two primary stimuli are sufficient. Such observers
are called dichromats and are classed as color
defectives. A much smaller proportion (0.003%),
known as monochromats, have no color dis-
crimination and their matching is, in effect,
brightness matching, which can be carried out
with one fixed primary stimulus. Among the re-
maining observers, all requiring three primary
stimuli for color matching, a small minority of
about 5.5% males and 0.4% females are classed as
anomalous trichromats. Their color matches show
pronounced deviations from those of the main
group, deviations usually accompanied by a
lowered sensitivity to color differences, and some-
times by other differences in visual response. The
distinction between anomalous trichromats and
normal trichromats is not clearcut because varia-
tions in the color-matching properties of those
classed as normal are appreciable, and the fixing
Basic Colorimetric Concepts 119
of a point where the deviation from the average
constitutes an anomaly is, in some degree, arbi-
trary.
With the exclusion of all color defectives
(monochromats, dichromats, and anomalous tri-
chromats) there remain about 95% of all ob-
servers (male and female) who are regarded as
color-normal and whose color-matching proper-
ties can be thought of as approximations to those
of a hypothetical ideal or average normal ob-
server. Such an ideal observer was established in
colorimetry in 1931 by the Commission Interna-
tionale de l’Eclairage under the name C/E 193]
Standard Colorimetric Observer. It has played a
dominating role in colorimetry since that time.
The specifications of this standard observer and
of a second standard observer, the CIE 1964
Supplementary Standard Colorimetric Observer,
appropriate for large matching fields are fully
treated in Sections 3.3.1 and 3.3.2.
3.2.2 Tristimulus Space
Given the validity of the stronger form of the
trichromatic generalization, it is possible and con-
venient to represent color stimuli by vectors in a
three-dimensional space, called the tristimulus
space. |
In the description of tristimulus space and its
properties, color stimuli are appropriately de-
noted by boldface letters, such as Q, R, G, and B.
Whereas Q represents an arbitrary color stimulus,
the letters R, G, B are reserved for the fixed
primary stimuli chosen for the color-matching
experiment that is assumed to be at the basis of
the discussion.
Each color stimulus is assumed to be produced
by the imaging on the retina of a surface in the
external field uniformly emitting radiant power.
Strictly, in the complete specification of each
color stimulus, it is necessary to know not only
the absolute spectral distribution of the radiant
power emitted by the external surface, but also
the factors that characterize the display situation
of the stimulus, including, among many other
factors, the angular size of the external surface
(for more details, see Section 5.2). In the present
context, however, only the absolute spectral power
distributions of the stimuli involved in the color-
matching experiment are of concern, and all other
factors are kept unchanged. Thus, the color stimuli
are assumed to be uniquely defined by their re-
spective absolute spectral radiant power distribu-
tions {P, dX}. A color-matching experiment that
120 COLORIMETRY
usually meets these requirements is one that em-
ploys a bipartite visual field of 2° angular sub-
tense, whose image is centered at the fovea of the
retina [see Section 5.2.2 on quasi-symmetric
matching and Section 5.2.5(a) on some particular
matching or equivalence procedures].
The spectral distribution of a specific color
stimulus such as Q is denoted by {P) dA}o.
Similar notations are used for the primary
stimuli. However, one simple yet significant fea-
ture is added to them, such that R, G, B can be
regarded as primary stimuli of unit amounts. In
particular, the absolute spectral radiant power
distributions of the unit amounts of the primary
stimuli are denoted by
{oP, dX}, defining R
{yP, dX}, defining G
{BP, dd}, defining B
where p, y, B are positive constant factors whose
values are generally determined solely for con-
venience in a particular investigation. It is
customary to choose as unit amounts those
amounts of the primary stimuli that when addi-
tively mixed yield a complete color match with a
specified achromatic stimulus, usually one having
an equal-energy spectrum on the wavelength ba-
SiS.
With the above notations, a color match be-
tween a given stimulus Q and the additive mix-
ture in suitable amounts of the three fixed primary
stimuli R, G, B can then be expressed by the
following vector equation:
Q = RoR+ GG + BoB [1(3.2.2)]
where the scalar multipliers Rp, Gg, By measured
in terms of the assigned respective units of the
given primary stimuli R, G, B are called the
tristimulus values of Q.
In the first instance, Eq. 1(3.2.2) applies only
if the given stimulus Q can be matched by a
mixture of all three primary stimuli in positive or
nil amounts. In the general case, when Q may
have to be mixed with suitable amounts of one or
two primary stimuli before a color match can be
made with the remaining primaries or primary,
Eq. 1(3.2.2) still holds if the amount of a primary
mixed with Q is assigned a negative sign. For
example, if an amount Ro of R has to be mixed
with Q to yield a match with the mixture (GgG +
BoB), the equation for the match becomes
Q + RoR = GG + BoB
which may be written as
Q= —-RoR+ GG + BoB
or, with Rp = —Ro,
Q= RoR + GoG + BoB
The vector equation, Eq. 1(3.2.2), has an obvi-
ous geometrical interpretation in linear three-
dimensional space, as shown in Figure 1(3.2.2).
The unit vectors R, G, B, representing the primary
stimuli, define the space. They have a common
origin and point in three different, but otherwise
arbitrary, directions. Their lengths correspond to
the units of length in terms of which distances are
measured along the three directions. The vector Q
has the same origin as R, G, and B. Its three
components are located along the axes defined by
R, G, and B and have lengths respectively equal
to Ro, Go, Bo, the tristimulus values of Q. The
direction and length of Q is obtained by simple
vector addition as defined by Eq. 1(3.2.2). The
space defined by R, G, B is called the (R, G, B)-tri-
stimulus space. In that space, a color stimulus Q
appears as an (Ro, Go, Bo )-tristimulus vector.
The representation of color stimuli in tn-
stimulus space as vectors is informative, but usu-
ally not convenient in colorimetric practice. A
two-dimensional representation is commonly pre-
ferred, though it is recognized that this necessarily
reduces the information contained in the tn-
stimulus vector of the stimulus. A useful two-
dimensional representation is obtained in the unit
plane, R+ G+ B= 1, of the tristimulus space,
as shown in Figure 1(3.2.2). All tristimulus vec-
tors, such as Q, or their extensions must intersect |
the unit plane in a point Q. The location of that
point uniquely determines the direction of the
tristimulus vector Q, but it carries no information
as to the vector’s length which characterizes the
“intensity” of the stimulus Q.
Convenient coordinate axes in the unit plane
are the lines of intersection of the unit plane with
the three planes defined by pairs of the three
vectors representing the primary stimuli. These
lines of intersection form a triangle which defines
the (7, g, b)-chromaticity diagram redrawn in the
upper part of Figure 1(3.2.2). The chromaticity
coordinates of Q are related to the tristimulus
values R= Ro, G= Go, B=Bo of Q by the

Fig. 1(3.2.2). (R,G, B)-tristimulus space and
(r, g, b)-chromaticity diagram. The primary stimuli
R, G, B are represented by vectors of unit length
with a common origin 0. A color stimulus Q is
represented by the tristimulus vector Q whose
components have lengths (tristimulus values)
R 9, Gg, Bg along the directions defined by R, G, B,
respectively. The tristimulus vector intersects the
unit plane (R + G+ B= 1) of the tristimulus
space in the point Q. The unit plane contains the
(r, g, b)-chromaticity diagram, in which Q repre-
sents the chromaticity point of Q. The chromatic-
ity coordinates, rp, £9, bg of Q are related to the
tristimulus values R 9, Gg, By by simple equations
given in the text.
following equations:
r= R
R+G+B
b G
Sener [2(3.2.2)]
ye B
Rok GB
It follows that
Basic Colorimetric Concepts 121
The chromaticity diagram shown in Figure
1(3.2.2) is in the shape of an equilateral triangle,
used in the early days of colorimetric practice,
and sometimes referred to as the Maxwell color
triangle. A more convenient version is a triangle
in which the r and g coordinate axes are per-
pendicular to one another. Such a right-angled
triangle is now almost exclusively used when plot-
ting chromaticity points. In making appropriate
allowances in marking the units along the direc-
tions defined by the vectors representing the
primary stimuli in tristimulus space, it is easy to
obtain a unit plane that embodies a right-angled
triangle. Such an operation does not alter the
above defining equations of the chromaticity co-
ordinates.
3.2.3. Basic Colorimetric Equations
A given color stimulus Q, with spectral radiant
power distribution {P, dA}9,may be thought of
as an additive mixture of a set of stimuli Q, with
spectral power distributions {P, dA}g confined
to n wavelength intervals of width (AA); (i = 1 to
n). The n wavelength intervals, not necessarily of
equal width, completely subdivide the visible
spectrum, considered to be a closed interval rang-
ing, for example, from A, = 360 to A, = 830 nm.
Figure 1(3.2.3) illustrates such a case where the
visible spectrum has been subdivided into n = 12
parts of different widths (AA).
For each stimulus Q,, Eq. 1(3.2.2) must be
fulfilled, that is
Q, = Ro R+ GoG + BoB [1(3.2.3)]
where Ro, Go, Bo, are the tristimulus values of
Q,, one or two of which may be negative.
The additive mixture of all stimuli Q; (i = 1 to
n) must then be the original stimulus Q, that is
= $0-(Sne)a+(
i=]
$0.) i=1 i=]
+ |) 2o,)B [2(3.2.3)]
i=1
with tristimulus values
n n n
ede a ee Le)
i—1 i= i=]
[3(3.2.3)]
122 COLORIMETRY

ny o

Bey

Radiant
Power
P

—>{(Ar);
dp, = 830 nm
jd
Wavelength, 2X»
Visible Spectrum

Fig. 1(3.2.3). Example of the spectral radiant power distribution {P) dX}, of a color stimulus
Q. The visible spectrum is a closed interval of wavelengths ranging from A , = 360 to A,, = 830
nm. It is subdivided into n = 12 wavelengths intervals i= 1,2,...,12 of different width (AA).
Within each wavelength interval i, the spectral radiant power distribution {P, dA}, defines a
color stimulus Q;. The additive mixture of all Q; ({=1 to 12) constitutes the total color
stimulus Q.

The argument of the additive mixture can be e >0O there exists a 5>0, such that 0<
carried to the limiting case in which the number n (AA), =,11< 8, which makes
of the wavelength intervals is increased while
their widths (AA); are decreased to the point
where the largest wavelength interval approaches > Py, (4a); — Pi<e
zero width. This limiting case generates the fol- [=]
I=

lowing sum:
n regardless of where the wavelength A, is chosen to
lim P, (AA), =P be within the wavelength interval of width (AA),.
(AA) j=l >0 2 (Ar) Equation 4(3.2.3) defines the definite integral
of the spectral concentration P, of radiant power
[4(3.2.3)] in the visible spectrum [A ,, A,], that is,
In this expression, ||(AA);—,|| denotes the width
of the largest wavelength interval (i =k); P_ is dA
‘Pedvs lim » Py (AX);
the radiant power at wavelength A, located any- io Il(AA);=,|| 70 js]

where within the wavelength interval (AA);. The

limit P of the sum is the area under the curve
defined by the spectral concentration P, of radiant
power from A, to A,, representing the boundaries
of the visible spectrum. Figure 2(3.2.3) illustrates
the situation prior to increasing the number n of
the wavelength intervals while making their widths
smaller and smaller until the largest interval ap-
proaches zero.
The limit of the above sum, Eq. 4(3.2.3), means
that, in the closed interval [A,,A,], for every
[5(3.2.3)]
provided the limit exists. It can be shown that P,
is always integrable if P, is a continuous function
in the visible spectrum [A,, Aj].
Normally, the subdivision of the visible spec-
trum is in wavelength intervals i of equal width
(AA); = AA, which simplifies somewhat the limit-
ing process and Eq. 5(3.2.3) can be written as
 Basic Colorimetric Concepts 123

Visible Spectrum

Radiant
Power

Wavelength, A
Fig. 2(3.2.3)._ An example of the spectral concentration P, of the radiant power of a color
stimulus Q. The visible spectrum ranging from A, = 360 to A, = 830 nm is subdivided into
n = 12 wavelength intervals i of different widths (AX);. Within each wavelength interval i, a
wavelength A ; is chosen at which the spectral concentration is P, .The sum of the rectangular
areas P, (AX); from i to n= 12 approximates the area under the curve P, from A, to A,.
When the number of the intervals is increased while decreasing their widths such that the
largest interval (i= =11 in the figure) approaches zero width, the sum of the rectangular
areas becomes, in the limit, equal to the area under the curve P).

follows: power (P, = E, =const= 1) at every wave-

length A within the visible spectrum. Such a
i - stimulus is called an equal-energy stimulus and
P={ P,dA= 1 P, AX
is ‘ Ar—0 2 x denoted by E. Its monochromatic constituents are
denoted by E,. The spectral distribution {£, dA}
[6(3.2.3)] of the equal-energy stimulus is uniform across the
visible spectrum and often referred to as an
where the wavelengths 4, may be chosen as the equal-energy spectrum on the wavelength basis [see
central wavelengths in the wavelength intervals of Figure 3(3.2.3)].
width AX. The equation for a color match involving a
The quantity P, dA represents the radiant monochromatic constituent E, of the equal-
power in the wavelength interval of width dA energy stimulus E is as follows:
centered at the wavelength A, and it is said to
define a monochromatic stimulus of wavelength X, E, =7(A)R+ 2(A)G+ b(A)B
denoted by Q).
For each monochromatic stimulus Q,, Eq. [8(3.2.3)]
1(3.2.2) must hold, that is,
where 7(A), (A), B(A) are the (spectral) tri-
O-= KR + GG + B,B [7(3!2.3)] stimulus values of E,. The lowercase letters with
an overbar are standard notation for this special
where R,, G,, By are the tristimulus values of set of tristimulus values.
Q,, often called spectral tristimulus values. Figure 4(3.2.3) shows the results of a color-
A particularly important set of spectral tri- matching experiment of a typical observer with
stimulus values is obtained when all the mono- normal trichromatic vision viewing a 2°-bipartite
chromatic stimuli Q, contained in the spectrum visual field. The monochromatic stimuli E),
of a given color stimulus Q have unit radiant of unit radiant power vary in wavelength from
124 COLORIMETRY

chromaticity point E of the equal-energy stimulus

Visible Spectrum E at the center of the (7, g)-chromaticity dia-
gram; that is, at rg = gp = 0.333.
The sets of spectral tristimulus values r(A),
P, =E, =const =| Z(A), b(A) of monochromatic stimuli E, of unit
radiant power embody the color-matching prop-
erties of the observing eye in the particular system
of primary stimuli R, G, B. They are ap-
Power
Radiant
propriately called color-matching functions.
To illustrate the use of Figure 4(3.2.3), con-
sider the monochromatic stimulus E, at A = 475.
hg = 360 dp=830nm
At that wavelength we read off the values
7(475) = —0.045, 2(475) = 0.032, and b(475) =
Wavelength, X 0.186 which are the tristimulus values of E,,; and
Fig. 3(3.2.3). Spectral radiant power distribution thus, Eq. 8(3.2.3) becomes, in this example,
{E, dX} of a color stimulus E having an equal-
energy spectrum on the wavelength basis. E475 = —0.045R + 0.032G + 0.186B

\,, = 380 to A, = 760 nm. The fixed primary
stimuli R, G, B are also monochromatic
and of wavelengths Ap = 700, A, = 546.1, and
A, = 435.8 nm, respectively. The units of these
primaries were chosen in the radiant power ratios
72.1: 1.4: 1.0 (approximately) which places the
The negative amount of the “red” primary
stimuli stimulus R signifies that in the actual color match,
0.045R had to be added to E47, in order to
“desaturate” E,,, sufficiently to obtain a com-
plete color match with the mixture 0.032G +
0.186B. Thus, the equation for the actual color

0.4

0.2

Tristimulus
values

0.0
4

40 —S(2)oO

Wavelength (A), nm

435.8
—
nm
r(A)
css
546.1
nm
—
osba 700.0
nm
—»

Fig. 4(3.2.3). Spectral tristimulus values 7(A), (A), (A) of monochromatic stimuli E, of
unit radiant power, with respect to the three fixed primary stimuli R,G,B, also monochromatic
Stimuli and of wavelengths A p = 700, Ag = 546.1, A , = 435.8 nm. The units of the primary
stimuli are of radiant power ratios 72.1: 1.4: 1.0.
 Basic Colorimetric Concepts 125

Fig. 5(3.2.3). (r, g)-Chromaticity diagram in the form of a right-angled triangle showing the
spectrum locus of monochromatic stimuli from A = 380 to 700 nm. The straight line joining
the two extremes of the spectrum locus is the purple line. The chromaticity points R,G, B of
the monochromatic primary stimuli R,G,B are represented by the corners of the triangle. The
chromaticity point E of the equal-energy stimulus E is at the center of the triangle at
re = £¢ — 9.333.

match reads as follows: with

E475 + 0.045R = 0.032G + 0.186B r(d) + 2(A) + B(A) = 1

In applying Eq. 2(3.2.2) to the spectral tri-
stimulus values r(A), g(A), b(A) illustrated in
Figure 4(3.2.3), the chromaticity coordinates r(A),
g(A), b(A) of the monochromatic stimuli are ob- .
tained:
=f yes FTEa BOE)
The locus of chromaticity points so determined is
called the spectrum locus in the (r, g)-chromatic-
ity diagram. Figure 5(3.2.3) shows the (r, g)-
chromaticity diagram in the preferred form of a
right-angled triangle. The wavelengths of some of
the monochromatic stimuli are marked along the
spectrum locus. The straight line joining the two
end points of the spectrum locus is called the
purple line, representing the locus of chromatici-

ee ties of additive mixtures of the extreme short-wave

stimulus (A, = 380) with the extreme long-wave
Fr ee nbOD stimulus (A, = 700).
With the introduction of the color-matching
Ce b(a)
ee een [9(3.2.3)] functions, it readily follows how to determine the
r(r) + B(A) + BA) tristimulus values of a complex color stimulus Q
126 COLORIMETRY X

defined by a spectral radiant power distribution {P,, dX} and {P,, dd}, are in complete color
{P, dX}o that is not restricted to narrow bands of match if the following three equations are simul-
wavelengths. Multiplication of both sides of Eq. taneously satisfied:
8(3.2.3) by P, dA results in
[rPnFA)
rv
dd = fP, F(A) da
Q, =(Py dA)E, = (Py, dA)r(A)R ue te

+ (P, dd) Z(A)G + (Py, dd) b(A)B [Pn

[10(3.2.3)]
which represents the equation for a color match
involving a monochromatic stimulus Q, of radiant
power P, dd, where P, can be any non-negative
finite value. If it is further assumed that P, is a
continuous function in the visible spectrum
[A,,,A,], Eq. 10(3.2.3) is integrable; that is, the
tristimulus values R= Ro, G= Go, B= Bo of
the given stimulus Q are given by
R=["PFA) Ad, G= [P,B(A) AA,
" a
B=[''P,b(A)dd [113.2.3)]
Ne
Two color stimuli, Q, and Q,, defined respec-
tively by spectral radiant power distributions
x,
ZA) dd = [PP
a
B(A) ad
['Pinb (A) dd= [Pb (A) dd
[12(3.2.3)]
The integrals on the left-hand side of these equa-
tions are the tristimulus values R,, G,, B, of
stimulus Q,, and the integrals on the right-hand
side are the tristimulus values R,, Gy, B, of Q».
At match, R,; = R», G; = Go, B, = By.
When the color-match conditions expressed by
Eg. 12(3.2.3) are satisfied for two color stimuli Q,
and Q, with different spectral radiant power dis-
tributions, {P,}, dA} # {P,, dA}, the two stimuli
are called metameric stimuli. Figure 6(3.2.3) shows
an example of a pair of spectral radiant power
;
distributions defining two metameric color stimuli.
The color-match conditions remain unchanged
if the respective color-matching functions r(A),

POWER
RADIANT

600 70Onm
WAVELENGTH, A

Fig. 6(3.2.3). Example of a pair of spectral radiant power distributions defining two metameric
- color stimuli.

Z(A), b(A) are multiplied by any three (nonzero)
factors. In fact, any three functions that are lin-
early independent combinations of r(A), g(A),
b(A) will serve equally well with 7(A), Z(A), b(X)
in stating the conditions for a color match.
3.2.4 Imaginary Color Stimuli
Although a color stimulus represented by its tri-
stimulus values with respect to a particular set of
primaly stimuli may have one, or perhaps two,
of these tristimulus values negative, not every
ordered set of three numbers, each of which may
be positive or negative, can be identified as the
tristimulus values, in the given primary system, of
an actual color stimulus. In colorimetry and color
theory it is sometimes convenient to admit so-
called imaginary color stimuli with sets of tri-
stimulus values that do not correspond, in the
primary system used, to actual color stimuli. Sub-
ject to a qualification to be introduced im-
mediately, it is postulated that such imaginary
stimuli participate in “color matches” specified
solely by relations between the tristimulus values
of the participant color stimuli in the same way as
do actual color stimuli. These color matches con-
form to the four matching laws required for strong
trichromacy (see Section 3.2.1). However, this
postulate requires a somewhat wider interpreta-
tion of the terms used in the proportionality and
additivity laws when the imaginary stimuli are in
question. As the imaginary stimulus is solely de-
fined by its tristimulus values, increasing or re-
ducing the stimulus by a factor a means simply
reducing it by another stimulus whose tristimulus
values are a times the original ones. In fact,
applied to imaginary stimuli, the proportionality
law is trivial. Adding an imaginary stimulus to
another stimulus, real or imaginary, means simply
defining an additive mixture (e.g., those used in
the additivity law), as the real or imaginary
stimulus, obtained by adding the respective tni-
stimulus values of the constituents.
It is convenient to place a restriction on imag-
inary stimuli. When it is admitted that imaginary
- stimuli leave intact the additivity principle, it will
be required that no additive mixture of a real
color stimulus with a second stimulus can match
the null color stimulus; that is, the stimulus with
all zero tristimulus values. This restriction means
that in tristimulus space (see Section 3.2.2),
whatever the primary stimuli used, the vectors
representing real stimuli all lie in one of the
halves into which the space is divided by some
Basic Colorimetric Concepts ey
plane through the origin, but not in the plane
itself. It follows, that if R, G, B are the tristimu-
lus values of a real color stimulus, the all negative
tristimulus values —R, —G, —B are not admitted
to specify an imaginary color stimulus.
Unfortunately, if neither (R,G, B) nor
(—R, —G, —B) specify a real stimulus, then both
sets of tristimulus values qualify to specify imag-
inary stimuli and their mixture matches the null
stimulus. By the following stronger restriction on
imaginary stimuli, it can be assumed that no
mixture of two stimuli, real or imaginary, will
match the null stimulus unless both are also null
stimuli. In tristimulus space, a plane through the
origin, aR + bG + cB = 0, is selected such that
all vectors representing real non-null stimuli lie in
one of the half-spaces so formed, but not in the
dividing plane. All other vectors in the same
half-space and not in the plane are accepted as
imaginary stimuli, and no others. As there is a
wide range of planes that might be selected, it is
necessary in any particular context to specify the
one adopted (e.g., by giving the relative values of
the numbers a, b, c). With the stronger restriction
on imaginary stimuli, a factor a applied to a real
or imaginary stimulus yields another stimulus,
which will be respectively real or imaginary, only
if a is positive. |
Associated with any set of tristimulus values
R, G, B are the chromaticity coordinates r, g, b
defined by Eq. 2(3.2.2). The chromaticity, that is,
the set of chromaticity coordinates, in a given
system of primary stimuli is the same for all color
stimuli having the same relative spectral power
distribution. In this sense, chromaticity specifies
color stimulus independent of “intensity”, as il-
lustrated earlier in Figure 1(3.2.2). Of course, in
general, color stimuli of many different spectral
distributions are represented by the same chro-
maticity. This follows immediately from the con-
cept of metameric stimuli discussed in Section
a)eS
The outstanding property of a chromaticity
diagram is that the representative point of an
additive mixture of two color stimuli lies on the
straight line passing through the chromaticity
points corresponding to the constituents of the
mixture. Furthermore, empirical properties of
normal human color vision ensure that:
(i) the units of any set of independent primary
stimuli can always be chosen so that the sum
(R+G+B) of the tristimulus values of any
color stimulus exceeds zero, and
128 COLORIMETRY

(ii) with units so chosen, the resultant chro-
maticities of all real color stimuli plot in the
chromaticity diagram as points of a finite domain
bounded by the spectrum locus and the purple line,
as shown in Figure 5(3.2.3).
The closed boundary of the gamut of real
chromaticities is everywhere straight or concave
inwards, and the representative point of any addi-
tive mixture of real stimuli lies between the
chromaticity points of the constituents on the
straight line through them. Points lying outside
the domain of real chromaticities in such a dia-
gram represent imaginary stimuli. If the stronger
restriction is imposed on the latter (ie., these
stimuli together with all real stimuli are repre-
sented by points in one of the halves of tristimu-
lus space generated by the dividing plane aR +
bB + cB=0), then there is a corresponding
straight line in the chromaticity diagram, defined
by (a — c)r + (b—c)g + c= 0, which separates
points representing stimuli for which R+ G+ B
> 0 including the chromaticities of all real and
some imaginary stimuli (region A), from points
representing stimuli for which R+ G+ B<0O,
comprising all the remaining imaginary stimuli
(region B). Points on the dividing line represent
neither real nor imaginary stimuli. Additive mix-
tures of stimuli represented by points lying in
different regions have chromaticity points that lie
on the straight line through the constituent points
but not on the segment lying between these points.
In the special case where a=b=c#0O, the

<

X—————_P

Fig. 1(3.2.4). The points (U),(V),(W) represent the chromaticities of three color stimuli in
the CIE 1931 (x, y)-chromaticity diagram proposed for use as independent primary stimuli
_U,V,W in color-matching experiments. They are necessarily noncollinear, so that, taken in
pairs, they define three different straight lines that, provided none of the points (U),(V),(W)
lies on the boundary (spectrum locus and purple line) of the domain of real stimuli, divide the
domain into seven regions, UVW,UV/W,VW/U,WU/V,U/VW,V/WU,W/UV. A color
stimulus in region UV W will match with positive amounts of the primary stimuli U,V,W; a
color stimulus in any other region will require for the match a positive contribution from the
one or two primaries specified before the solidus (/), and desaturating contributions from the
two or one primaries specified after the solidus.

straight line is removed to infinity and the whole
diagram constitutes region A. Any two stimuli
represented by the same point of the diagram
have tristimulus values that differ only by a posi-
tive factor.
An example of a chromaticity diagram with
real (spectral) primaries is shown in Figure
5(3.2.3), and a second example, Figure 1(3.2.4),
based on the important set of imaginary stimuli
adopted by the CIE as primary stimuli, is used to
illustrate the various possible desaturation re-
quirements when the primary stimuli may be any
three independent stimuli whose chromaticities
do not lie on the boundary of the domain of real
stimuli.
3.2.5 Colorimetric Transformations
The limitation on the choice of primary stimuli in
color matching, namely that none of the primary -
stimuli can be matched by an additive mixture of
the other two, means in the vector representation
that the primary vectors must be linearly indepen-
dent; that is, they must be such that no set of
values R, G, B makes
RR+
GG-+ BB=0,
except the trivial set R= G= B= 0.
For practical reasons it is convenient to use
sometimes one set of primary stimuli, R, G, B,
sometimes a different set, R’,G’, B’, but the vec-
tors of any two sets are linearly related. Since
each primary stimulus of the one set can be
matched with a mixture of the primaries of the
other set, the following homogeneous linear rela-
tions between the primaries will hold good:
R’ =a,,R+a,,G+ a;,B
G =a,,R+ aGt a35.B H1t3. 205)
B’ = a,3R + a,,G + a;;B
Both sets of primary stimuli are linearly indepen-
dent sets, and the matrix A* formed by the trans-
formation coefficients a,; (k,i = 1,2,3) must
have a determinant different from zero.
When employing the notation used in Section
3.2.3, the matrix A* with elements a,,; is identical
to the matrix
Rp Gp: By
[2(3.2.5)]
Basic Colorimetric Concepts 129
where the three elements in each of the three rows
are the tristimulus values of the new primary
stimuli R’, G’, B’, respectively, measured in terms
of the original primary stimuli R, G, B.
The tristimulus values R, G, B of a color
stimulus measured in terms of the original set of
primary stimuli (R, G, B) and the tnstimulus
values R’, G’, B’ measured in terms of the new set
of primary stimuli (R’,G’, B’) are related by the
homogeneous linear transformation
R= a,,R’' + a,,G’ + a,3B’
G = a,R’ + anG’ + ay3B' BiCre shy
B= a,,R’ + a3G’ + a3,B’
where the transformation matrix A is the trans-
pose of matrix A* of Eq. 2(3.2.5); that is, A is
obtained from A* by interchanging rows and
columns.
The solution of Eq. 2(3.2.5) for R’, G’, B’ gives
the inverse transformation
Re= bi R+b GbR
Go= by Ret bynG byB? 1436.25)
where the coefficients b;, (k,i = 1,2,3) are the
elements of a matrix which is the inverse of
matrix A and is written A’.
In the particular case of the monochromatic
constituents of the equal-energy stimulus E, whose
spectral tristimulus values are the color-matching
functions 7(A), g(A), b(A) in the R, G, B system
of primary stimuli, similar transformation equa-
tions apply
7’(A) = by F(A) + by B(A) + 63 (A)
2(A) = by F(A) + by B(A) + B35 (A)
b’(A) = by, 7(A) + B38 (A) + b33b (A)
[5(3.2.5)]
The new color-matching functions 7’(A), 3’(A),
b’(A) are in terms of the new system of primary
stimuli R’, G’, B’.
If R’, G’, B’ and R, G, B are, respectively, the
tristimulus values of the same color stimulus with
respect to two sets of primary stimuli, and if they
are related as in Eq. 4(3.2.5), then it is readily
130 COLORIMETRY

seen that the corresponding chromaticity coordi- Explicit equations inverse to Eq. 7(3.2.5) can
nates are related by the following equations: be derived from the following expressions:

Tet Boe »Bas By ilorteBag

,
a a
8’ Bo Bog Bo, 8’ =Bas
(bd) + by, + b3,)r Hf (bi + by, + b37) 8
l B32 B33 B31 B33
+ (b13 + bo3 + b33)b hee See ere ae
Bi itt oath By oho ee

ai byr + byg + b53b By, Bu g’ Bo, By 8g’

B3, Bs. 1 B3, Bs, 1
3 (bi, + by + b3,)r + (by. + by + by) 8
ur (43 + by3 + b33)b
[8(3.2.5)]
b’ b3ir + by2g + 6336 All that is needed is to expand the determinants
ss (dy, + by) + b3,)r + (bi + by + b37) 8
in these expressions in the normal manner and
the appropriate equations for r and g will evolve.
+(b,3 + by; + b33)b
When the tristimulus values (or chromaticity
[6(3.2.5)] coordinates) obtained in one system of primary
stimuli are transformed into another system, there
Using the identities will generally be random errors associated with
the original tristimulus values (or chromaticity
r-e+g+b=1 andvirs:2'tbt= coordinates), and other random errors associated
with the transformation coefficients, themselves
and Eq. 6(3.2.5), the following relation between obtained by color matching or by spectrophotom-
r’, g’ and r, g is obtained: etry. All these errors will propagate into the new
tristimulus values (or chromaticity coordinates).
a Burt Big + Big By application of the general theory of propaga-
Bir + B38 + B33 tion of random errors, it is possible to determine
the resulting uncertainty in the final tristimulus
pes Bar + Bag + Bo; values (or chromaticity coordinates). Section 5.4.6
. Bir + B38 + B33 ee gives the basic equations used in such calcula-
tions.
The coefficients B;, (i, k = 1,2,3) in terms of the
coefficients b;, of Eq. 4(3.2.5) are
3.3. THE CIE COLORIMETRIC SYSTEM
By = 5), — 53, By = by. — 513, Big
= 443
In colorimetric practice, the main objective is to
Bo, = by) — a3, Bap = bay — 23, Bo3 = 23 obtain results valid for the group of normal tri-
chromats. To this end, the color-matching proper-
B3, = (bi, — by3) + (by, Zi by3) + (b3, — b33) ties of an ideal trichromatic observer are defined
by specifying three independent functions of
B32 = (bi a bi3) a (by) iu) by3) “i (35 - b33) wavelength that are identified with the ideal ob-
server's color-matching functions. By definition,
B33 = by3 + by3 + b33
the ideal observer makes matches in strict accord
with the stronger trichromatic generalization, and
Transformations of the form of Eq. 7(3.2.5) are these matches are subject to the principle of per-
called projective transformations. They are a sistence of color matches, all without restriction on
special class of nonlinear inhomogeneous trans- the range of stimuli involved. The values of the
formations of the (r, g)-plane that play an im- color-matching functions of the ideal observer, as
portant role in the search for so-called uniform- functions of wavelength, are fixed so that the
chromaticity-scale diagrams (UCS-Diagrams) to metameric matches predicted with their aid using -
be discussed in Section 3.3.9. Eq. 12(3.2.3) represent, in some sense, typical or

average matches for the group of normal trichro-
mats. This requirement can, of course, be met
only for matches where all the conditions lie
within the range where the matching of actual
observers conforms, at least approximately, with
the trichromatic generalization.
By general consent, in all countries the specifi-
cation of standard (i.e., ideal) observers for use in
colorimetry and photometry is the province of the
Commission Internationale de l’Eclairage (or CIE),
a body that in 1913 took over the functions of the
earlier Commission Internationale de Photomeétrie.
The CIE specifications of the standard observers
of colorimetry are dealt with in this chapter,
together with CIE recommendations on related
questions, the whole forming what may be called
the CIE colorimetric system. The link with the
CIE standard photometric observer is explained
in the following, but this observer is considered
on its own merits in Chapter 4. The CIE official
recommendations on colorimetry are spelled out
in detail in the CIE colorimetry document and in
two supplementary publications (CIE, 1971;
1972a; 1978a).
3.3.1 The CIE 1931 Standard
Colorimetric Observer
Since 1931 the ideal observer on which the whole
of applied colorimetry has been based is the CIE
1931 Standard Colorimetric Observer, laid down
by the CIE in that year. The color-matching prop-
erties of the CIE 1931 standard observer are
defined by the color-matching functions x(A),
yA), 2(A).
In 1931 the color-matching functions x(A),
y(A), Z(A) were defined in the wavelength range
A = 380 to 780 nm at wavelength intervals of
AX = 5 nm. Their values were given to four deci-
mal places. In 1971, the CIE recommended a new
and extended table of the color-matching func-
tions to define the CIE 1931 standard colorimet-
ric observer. The new table, given as Table 1(3.3.1)
in the Appendix, is essentially the same as the
original 1931 Table, but contains interpolated
values at one nanometer intervals and covers an
extended wavelength range from A = 360 and 830
nm. The original values were also smoothed
slightly to eliminate small irregularities.
The color-matching functions x(A), y(A), Z(A)
in Table 1(3.3.1) are given in seven significant
figures, the corresponding chromaticity coor-
dinates x(A), y(A), z(A) given in the same table,
The CIE Colorimetric System 131
have five significant figures. The large number of
significant figures used to define the color-match-
ing functions must not be construed as being
proof that the experimental color-matching
data, on which the tabulated values are based,
have a comparable inherent precision. Visual color
matching is far from being that precise (see Sec-
tions 5.4 and 5.5). However, the use of the CIE
1931 color-matching functions in a large variety
of colorimetric calculations and specifications
makes it very desirable, from the practical point
of view, to define these functions precisely and at
small intervals (AA = 1 nm), and to make them
and the corresponding chromaticity coordinates
smooth functions of wavelength.
When the values of Table I(3.3.1) are rounded
off, as given in Table II(3.3.1), to four decimals,
the original 1931 values are obtained in all but a
few instances. The few minor discrepancies that
do occur (see note at bottom of Table II(3.3.1))
are of no consequence in most practical colori-
metric calculations.
The experimental data for actual observers
that provided the basis for the standard x(A),
y(A), Z(A) values of Tables 1(3.3.1) and II(3.3.1),
were obtained with matching fields of two degrees
angular subtense directly viewed so that any par-
ticipation of rod vision in the measurements was
avoided. However, the CIE did not associate the
1931 standard observer specifically with any par-
ticular range of field sizes. Following a more
recent series of studies of the adequacy of alterna-
tive sets of color-matching functions in predicting
metameric matches in fields of different sizes, the
CIE has confirmed its recommendation of the
CIE 1931 standard observer with the further in-
dication that it is suitable for application to
matching fields of between one and four degrees
angular subtense.
In 1964, the CIE recommended an alternative
set of standard color-matching functions, denoted
by Xio(A), Vio(A), Zio(A), aS a supplement to
those of the 1931 standard observer for use
whenever more accurate correlation with visual
color-matching in fields of large angular subtense
(more than four degrees) is desired. The color-
matching functions X,9(A), Vio(A), Zio(A) define a
new ideal observer which is referred to as the CIE
1964 Supplementary Standard Colorimetric Ob-
server and which is specified below.
The CIE 1931 standard observer embodies
another visual property of the average or typi-
cal normal observer that goes beyond the
132 COLORIMETRY

color-matching properties and is not a necessary ing functions X19(A), Vio(A), Z10(A) given in Table
consequence of obedience to the trichromatic gen- I(3.3.2). It is intended for use when the colors of
eralization. The property in question is expressed large angular areas are in question and is based
by the basic principle of photometry, the additiv- on matching data obtained for centrally viewed
ity law for brightness, according to which the matching fields of 10 degrees angular subtense.
condition for a match in brightness, but not nec- The values of Table I(3.3.2) are given for wave-
essarily in color, of any two stimuli of spectral lengths ranging from A = 360 to 830 nm at
radiant power distribution {P,, dA} and {P,, dA} wavelength intervals of AX = 1 nm. The color-
takes the form matching functions are defined by six significant
figures and the corresponding chromaticity coor-
JP,V(A) dA = ifP,V(r) da dinates by five decimals.
The large number of significant figures used to
[1(3.3.1)] define the X,9(A), Vio(A), Zio(A) must, as in the
case of the CIE 1931 color-matching functions,
where V(A) is a function of wavelength known as
not be taken as an indication that visual color
the photopic luminous efficiency function, and
matching has a comparable precision. Consider-
the integral extends over the visible spectrum.
ations concerning the practical application of
Because of the difficulties of heterochromatic
these functions to colorimetric calculations pro-
photometry by direct comparison, indirect meth-
vide the overriding argument in favor of their
ods such as the flicker method are commonly
definition to such a high degree of precision and
used. To be strict, therefore, the validity of the
smoothness.
additivity law for brightness and the values of
However, for most practical applications, an
V(A) are relative to the particular method of
abridged set of color-matching functions and cor-
heterochromatic photometry employed in the
responding chromaticity coordinates is adequate;
measurements. For an observer who by a particu-
such a set is given in Table II(3.3.2).
lar method makes brightness matches that con-
The precision of large-field color matching is
form to the additivity law for brightness, and who
generally somewhat greater than that of small-field
also makes color matches that are trichromatic in
color matching. For a 10° as compared with a 2°
the stronger sense, it is readily shown that the
field, color matching is estimated to be two or
function V(A) must be a linear combination of
three times as precise. This is an obvious ad-
the color-matching functions, provided all pairs
vantage in practical situations where large match-
of color-matched (metameric) stimuli are also in
ing fields are often available and relevant to the
brightness match. For such an observer it is possi-
application at hand. In such situations, the CIE
ble to select from the infinitely many equivalent
1964 supplementary standard observer data enter
sets of color-matching functions, some sets for
into the colorimetric calculations that are in-
which one of the color-matching functions, by
tended to predict the color matching. However,
convention taken to be the central or y(A) func-
large-field color matching has limitations that are
tion, coincides with the V(A) function. The CIE
directly related to the use of a large field and do
1931 standard observer is endowed by definition
not apply to small-field (foveal) color matching.
with these ideal brightness-matching properties,
By definition, the large-field standard ob-
and the color-matching function y(A) of Table
server, like the small-field (1931) observer, makes
(3.3.1) has been chosen so that it coincides with
matches in strict accord with the stronger tri-
the standard V(A) function laid down earlier, in
chromatic generalization, but with different
1924, by the CIE to provide the quantitative basis
color-matching functions. The range of stimulus
for photometry under photopic conditions (see
intensities within which the metameric matches
Chapter 4). Thus, the 1931 standard observer
predicted by a standard observer are representa-
combines both color-matching and heterochro-
tive of those of actual observers must be more
matic brightness-matching properties in a single
restricted for the large-field case. The retinal image
quantitative scheme.
of the large field inevitably covers areas that
contain rod receptors. When the stimuli involved
3.3.2 The CIE 1964 Supplementary Standard
(i.e., the matching stimuli themselves and other
Colorimetric Observer
stimuli that may help to determine the condition
The supplementary standard observer introduced of adaptation) are of sufficiently low intensity,
by the CIE in 1964 is defined by the color-match- the scotopic mechanism may contribute to the

match, and in such a way as to upset the predict-
ions of the standard observer. Upsets of this kind
are described as arising from rod _ intrusion.
Large-field color matching under conditions of
low-level adaptation is therefore not in general
correctly predicted by the large-field standard
observer, and the CIE 1931 standard observer is
equally inapplicable. Precise indications of the
stimulus conditions under which rod intrusion
will upset the predictions of the standard observer
have not yet been given, but some work related to
the problem is discussed in Section 5.6.2.
For certain color stimuli filling uniformly a
large field, the appearance to an observer is clearly
nonuniform. The nonuniformity consists of a spot,
the so-called Maxwell spot. The Maxwell spot has
an ill-defined boundary and a diameter of the
order of four degrees, which differs in color from
the apparently uniform main area of the field and
which is centered on the observer's momentary
point of regard within the field. As the point of
regard is shifted from place to place in the field,
the Maxwell spot moves with it. The differences
in the color-response properties of the central
(foveal) area of the retina and of the surrounding
area that are made evident by the Maxwell spot
are also responsible for the deviating color-match-
ing functions required for small- and large-field
color matching. However, they are not related, or
not directly related, to the presence of rod recep-
tors in the retina. The Maxwell spot is seen most
clearly at moderate luminances of the field. Its
occurrence is attributable, in the main, to a yellow
pigment, the macular pigment, which covers all
the retinal receptors in an area covering the fovea
and beyond and whose optical density diminishes
sharply in moving out from the fovea into the
surrounding area (see Section 2.2.5).
For many color stimuli, including most of
those met in everyday life, the Maxwell spot is
not noticeable, but for color stimuli of more
extreme spectral radiant power distributions, such
as the mixtures of two or three monochromatic
stimuli used in the basic determinations of the
color-matching functions, it can be a striking
feature that complicates the task of large-field
color matching.
In the studies of large-field color matching
from which the standard color-matching func-
tions Xjo(A), Vio(A), Z19(A) Of Table 1(3.3.2) were
derived, the observers’ matches were made for the
color stimuli occupying the main area of the field
excluding or ignoring the Maxwell spot. The pre-
dictions of the large-field standard observer must
The CIE Colorimetric System 133
be regarded therefore as applying to the color
stimuli in the main area of the extended field and
not to the Maxwell spot.
There is no officially agreed luminous ef-
ficiency function for large-field brightness match-
ing under photopic conditions, analogous to the
V(A) function for the small-field case. The CIE
has, in fact, standardized a second luminous ef-
ficiency function V’(A) but this refers to very low
levels of adaptation and to conditions where the
scotopic mechanism completely dominates bright-
ness matching (see Section 4.3.2). It is not rele-
vant here and thus, the question of choosing
large-field color-matching functions so that y)9(A)
coincides with an agreed luminous efficiency
function does not arise. Nevertheless, from the
infinitely many equivalent sets of color-matching
functions, the particular set that was adopted by
the CIE has a y,.(A) function whose values at
three wavelengths were made to agree. This is
always possible with the mean luminous efficien-
cies of these three wavelengths as measured in a
10° field by a group of observers using the flicker
method of heterochromatic brightness matching.
The three wavelengths were those of the primary
stimuli used in the determination of the color-
matching functions. The flicker measurements
were not, at that time, extended to other wave-
lengths covering the spectrum. However, on the
basis of more recent work, it is generally accepted
that a luminous efficiency function directly mea-
sured for all wavelengths in a 10° field, agrees
well with the y,.(A) function (CIE, 1978b).
3.3.3 Development of the Two CIE Standard
Colorimetric Observers
Because of the important role of the standard
observers, some further details of the experimen-
tal studies and other considerations leading up to
them may be noted.
(i) The CIE 1931 Standard Observer. The
color-matching functions of the 1931 standard
observer are based of the work of Guild (1931)
and Wright (1928-1929) and on the investiga-
tions mainly by Coblentz and Emerson (1918)
and Gibson and Tyndall (1923) that led to the
adoption of the V(A) function in 1924. Both
Guild and Wright determined in the first in-
stance, the chromaticity coordinates of a set of
monochromatic stimuli in the wavelength range
’ = 400 to 700 nm using a 2° bipartite matching
field. Guild obtained these data for seven
134 COLORIMETRY
COORDINATES
CHROMATICITY
WAVELENGTH,A
Fig. 1(3.3.3). Wright’s mean color-matching results for the r(A), g(A), b(A) chromaticity
coordinates of monochromatic stimuli.
observers by measurements on the Guild trichro-
matic colorimeter. In this instrument, primary
stimuli were generated by light from an in-
candescent tungsten lamp (color temperature 2900
K), limited to relatively broad wavelength bands
in the red, green, and blue, respectively, by suit-
able colored filters. The units of these primary
stimuli were chosen to make equal the three chro-
maticity coordinates of a particular white-light
stimulus, the NPL white, of known spectral
radiant power distribution, which corresponded
to a correlated color temperature of about 4800
K. Guild’s mean results for the chromaticity coor-
dinates of the monochromatic stimuli are shown
in Figure 1(3.3.3). , known, after Wright, as the W.D.W. method (for
The primary stimuli of Wright’s trichromatic
colorimeter, which he used for his measurements,
are themselves monochromatic stimuli, located at
wavelengths A = 650, 530, and 460 nm, respec-
tively.
Wright introduced a novel method of defining
the units of his primary stimuli, which were cho-
sen so that the r and g chromaticity coordinates
were equal for a monochromatic stimulus of
wavelength 582.5 nm (first normalizing wave-
length) and also equal for a second normalizing
wavelength of 494.0 nm. This method of fixing
the units of monochromatic primary stimuli leads
to chromaticity coordinates for the monochro-
matic stimuli that are identical for observers whose
color-matching properties can be regarded as the
same, except that for different observers, different
colored filters are interposed between the stimuli
and the photoreceptor system of the retina. Thus,
observers differing only in the (visually inert)
pigmentation of the optic media or of the prere-
ceptor layers of the retina (see Section 2.4.6)
yield, on this system, the same chromaticities for
the monochromatic stimuli. Of course, this is not
generally true for stimuli containing more than
one wavelength. This valuable method of fixing
the units of monochromatic primary stimuli is
the detailed calculation procedure, see Section
5.6.1). .
The mean chromaticity coordinates for mono-
chromatic stimuli, obtained by Wright from the
results of 10 observers, are shown in Figure
2(3.3.3). For a larger group of 36 observers, in-
cluding the 10 just mentioned, Wright determined
the chromaticity coordinates of the NPL-white
stimulus, these data being required in the further
calculations.
For a given trichromatic observer, the chro-
maticity coordinates of all stimuli expressed in
one system of primary stimuli can be transformed
 The CIE Colorimetric System 135

ad o

= (oe)
COORDINATES
CHROMATICITY

600 700 nm
WAVELENGTH,
Fig. 2(3.3.3). Wright’s (1928-1929) mean color-matching results for the r(A), g(A), b(A)
chromaticity coordinates of monochromatic stimuli, normalized at A, = 494.0 and A, = 582.5
nm in accordance with the WDW method of normalization.

into chromaticity coordinates in another system,
if the chromaticities in the two systems of four
specified color stimuli are known. These four
stimuli, sometimes called the cardinal stimuli, need
be specified only by their relative, not their abso-
lute, spectral radiant power distributions, but they
must be such that none of them can be matched
by a mixture containing only two (or one) of the
remaining three stimuli. The measurements de-
scribed enabled Guild (1931) to transform both
his own and Wright’s chromaticities of the mono-
chromatic stimuli to a common system in which
the primary stimuli were monochromatic and of
wavelengths Ap = 700 nm, A, = 546.1 nm, and
A, = 435.8 nm. Their units were such that the
NPL-white, the fourth cardinal stimulus, had
_ chromaticity coordinates rxypy = Znpr = Onpy =
|/3. The transformed chromaticities of the mono-
chromatic stimuli from the two independent in-
vestigations were found to be in good agreement,
and the mean results were used in defining the
CIE 1931 standard colorimetric observer.
Nevertheless, the chromaticities of the mono-
chromatic stimuli and the specified cardinal
stimuli are not sufficient to define the color-
matching properties of an observer; the color-
matching functions are not determinable from
these data alone. Those functions would have
been obtained directly if the quantities of the
primary stimuli in a color match with a measured
radiant power of each of the monochromatic
stimuli had been determined. Instead, a different
method of completing the definition of the stan-
dard observer was followed. It was assumed that
heterochromatic brightness matching by the
flicker method is an additive process and that the
luminous efficiency function V(A) is a linear
combination of the color-matching functions 7(A),
Z(A), b(A). This assumption implies the relation
V(A) = Lar(A) + Leg(A) + Lgb(A)
[1(3.3.3)]
where Lr, Lo, and Lz are readily seen to be
constants proportional to the (flicker) bright-
nesses of unit quantities of the primary stimuli in
the trichromatic system in which 7(A), g(A), b(A)
are expressed.
With the chromaticity coordinates r(A), g(A),
b(A) replacing the color-matching functions r(A),
g(A), b(A) in Eq. 1(3.3.3), the following equation
136 COLORIMETRY

is obtained: Thus, if the chromaticity coordinates r, g, b of a

stimulus of known relative spectral radiant power
V(A) = s(A)[Lar(A) + Leg(r) + Lpb(A)] distribution {P, dA} are measured, they may be
expressed by way of the tristimulus values R, G, B
[2(3.3.3)] as

with R l

s(A) =7(A) + B(A) + BA)

> REGHEB ~ REGHBLPA) a
Both Guild and Wright have made direct mea- Se ee See
surements of the relative brightness of unit quan-
8=REGEB REGEB APA) a
tities of their primary stimuli, using the flicker
rf B = l .
method of heterochromatic photometry, for their b
respective groups of observers. Thus, in each case,
PREG TIERCE Oe
by identifying V(A) with the function already [4(3.3.3)]
standardized by the CIE in 1924, the quantity
s(A) could be calculated to a constant factor from In substituting 7(A), Z(A), D(A) by Eg. 3(3.3.3),
the measured spectral chromaticities r(A), g(A), one obtains
b(A) and relative values of Lr, Le, and Lz.
When s(A) is known, the color-matching func- Pyr(X)V(A) dad
tions are immediately derivable. This, in princi-
r= —_— — oes

ple, was the method used in determining the

color-matching functions of the CIE 1931 stan- ne (OA | Py,g(A)V(A) da
dard observer from the mean data of Guild and
Wright. However, because of the different primary
ST R+G+ Bs Lar(A) + Log(A) + Lgb(A)
stimuli used, the details are a little more elaborate.
An interesting alternative method of deriving
ae I P,b(A)V(A)da
the color-matching functions, which still assumes hy Sates BAT AN) + LeeA iol ae
V(A) but does not involve measurements of the
relative brightnesses of the units of the primary [5(3.3.3)]
stimuli, was explored by Wright. The method
rests on the connection between the color-match- Apart from the constant 1/(R + G+ B), the
ing functions and the chromaticities of the mono- unknowns in these three equations are Lp, Lg,
chromatic stimuli in terms of V(A), Lp, Lg, and Lz. Provided {P, dA} is a complex distribution
L». One has (i.e., not confined to a single wavelength), a solu-
tion for the relative values of Lr, Lo, and Lz
r(A) = r(A)s(A) will be obtainable by a method of successive
approximation. Given this solution, the color-
r(A)V(A) matching functions 7(A), Z(A), b(A) are de-
Lrar(A) + Leg(A) + Lgd(A) termined from Eq. 3(3.3.3). Wright’s application
of this method to his data gave results that checked
and similarly satisfactorily the earlier determinations.
The CIE defined the properties of the CIE
(A) = g(A)s(A) 1931 standard observer initially by two different
but equivalent (linearly related) sets of color-
¥ g(A)V(A) matching functions. The R, G, B system again
Dert\ jer Leet) + 1 DtA,) employs monochromatic stimuli at 700.0, 546.1,
and 435.8 nm as primary stimuli, but with the
b(A) = b(A)s(A) units adjusted so that the chromaticity coordi-
nates, not of the NPL-white stimulus but of the
z b(A)V(A) equal-energy stimulus (wavelength basis), are all
= LattN ut dee (Nd) shel (AN) equal. The luminous efficiencies Lp, Lo, Lp
of unit quantities of the primary stimuli in this
[3(3.3.3)] system are in the ratios 1 : 4.5907 : 0.0601.
 The CIE Colorimetric System 137

CIE 1931 Standard Colorimetric System

(R,G,B) - System (X,Y,Z)- System

|
Sb Mapes %(A), ¥CA),2 (A)
r(A),Q(A), b(A) color -matching
color- matching functions
functions

0.5

Values
Tristimulus
Tristimulus
Values
0.0 700 nm
0.0
t | Wavelérigth r f 500 600
: E ; Wavelength XA
: : :
'
Ae Ag da

(r,g)-chromaticity (x,y) - chromaticity

diagram diagram

0.4

0.2

(Z) 02 0.6 0.8

Fig. 3(3.3.3). Color-matching functions and corresponding chromaticity diagrams of CIE 1931
standard colorimetric observer in the system of real primary stimuli R,G,B, and in the
transformed system of imaginary primary stimuli X, Y,Z. The (X, Y, Z )-system constitutes the
official CIE 1931 system for practical colorimetry involving color-matching fields of angular
subtense between one and four degrees. The units of the primary stimuli in either system have
been fixed so that the chromaticity point E of the equal-energy stimulus plots in the centroid of
the respective chromaticity diagram, that is, 7, = g,;= 1/3, and x;= yp = 1/73.

This corresponds to radiant power ratios
72-0962 1.3191 le
Both the chromaticity coordinates r(A), g(A),
b(A) and the color-matching functions r(A), g(A),
b(A) are tabulated in Table 1(3.3.3) for wave-
length A ranging from 380 to 780 nm at constant
wavelength intervals of 5 nm. Figure 3(3.3.3)
shows the color-matching functions and the spec-
trum locus in the (r, g)-chromaticity diagram.
The chromaticity E of the equal-energy stimulus
lies at the centroid of the chromaticity diagram;
that is, at re = g, = 1/73.
As a direct consequence of fixing the units of
the primary stimuli in this way, the areas under
138 COLORIMETRY

the curves 7(A), Z(A), b(A) are equal; that is,
fFQ) dd = [3(A) dd = [5(A) ar
The color-matching functions 7(A), 2(A), b(A)
and the corresponding chromaticity coordinates
r(A), g(A), b(A) [Table I(3.3.3)] include negative
values, which is inconvenient when tristimulus
values of a given stimulus are to be evaluated
according to Eq. 11(3.2.3) by computational
methods from the spectral radiant power distribu-
tion of the stimulus and the standard color-
matching functions. The calculation involves, first,
summing separately the products with negative
and positive signs, and second, a final differenc-
ing of the sums. The development of direct-read-
ing photoelectric colorimeters is also made more
difficult by the sign change. Further, the de-
termination of photometric quantities such as
luminance for a given stimulus requires either
another direct integration using the V(A) func-
tion or the evaluation of a linear combination,
Eq. 1(3.3.3), of the previously computed tristimu-
lus values.
Reasons such as these led the CIE to adopt a
transformation of the trichromatic system based
on the primary stimuli R,G,B to one based on
new primary stimuli X, Y, Z. The X, Y, Z primary
stimuli are nonreal or imaginary; that is, they
cannot be realized by actual color stimuli (see
Section 3.2.4). In tristimulus space [see Figure
4(3.3.3)], they are represented by vectors outside
the domain of vectors representing real stimuli; in
the chromaticity diagram, their representative
points are not included among the chromaticities
of real stimuli; they lie outside the chromaticity
region bounded by the spectrum locus and
the purple line connecting the two ends of the
spectrum locus. The triangle formed by the chro-
maticity points of X, Y, Z completely encloses the
spectrum locus and the purple line. This ensures
that the chromaticity coordinates x, y, z and the
corresponding tristimulus values X,Y, Z of any
real color stimulus (monochromatic or otherwise)
are never negative.
Table 1(3.3.3) summarizes the relationships be-
tween the CIE 1931 (R, G, B)-system and the CIE
1931 (X, Y, Z)-system. It also gives transformation
equations that relate the chromaticity coordinates

Spectrum locus
Chromaticity
S (A) diagram
S (600)
S (700)

Locus of colors 5(A)

of constant radiance

Fig. 4(3.3.3). Tristimulus space based on the CIE 1931 imaginary primary stimuli X, Y,Z,
showing the (x, y )-chromaticity diagram with spectrum locus and purple line. Also shown is the
locus of all monochromatic stimuli E, of the equal-energy stimulus E. The vector E represent-
ing the equal-energy stimulus intersects the chromaticity diagram at E with coordinates
x=y=1/3.
 The CIE Colorimetric System ep

Table 1(3.3.3) Relationship between the CIE 1931 (R,G,B) Primary System and the CIE
1931 (X,Y, Z) Primary System of Color Specification

R, G, B System X, Y, Z System
(Chromaticity Coordinates) (Chromaticity Coordinates)

Stimulus r g b x y Z

(R) 700.0 nm ] 0 0 0.73467 0.26533 0.00000

(G) 546.1 nm 0 l 0 0.27376 0.71741 0.00883
(B) 435.8 nm 0 0 ] 0.16658 0.00886 0.82456
Illuminant E 4 4 4 4 + 4
Illuminant A 0.55255 0.32126 0.12619 0.44757 0.40745 0.14498
Illuminant B 0.36230 0.34305 0.29465 0.34842 0.35161 0.29997
Illuminant C 0.28226 0.33326 0.38448 0.31006 0.31616 0.37378

The relationship between the chromaticity coordinates r,g,b and x,y,z of any given stimulus are
expressed by the following projective transformation:

_ 0.490007 +0.31000 g +.0.200006

~~ 0.66697r + 1.13240 ¢ + 1.200635
_ 0.17697r +0.81240g +0.010636
»~ 0.66697r + 1.13240 g +1.20063b
_ 0.000007 + 0.01000g + 0.990005
7 ~ 0.66697r + 1.13240 ¢ + 1.20063b
From the chromaticity coordinates x, y, z the tristimulus values X, Y, Z may be obtained as follows:

where Vis the luminance of the stimulus computed in accordance with the luminous efficiency function
V(A).
The transformation of the chromaticity coordinates r(A), g(A), b(A) of the monochromatic stimuli is
carried out by using the same equations as above, and from it one obtains the x(A), y(A), Z(A)
color-matching functions as follows:

a tA)
x(A) jays y(A) V(A), y(A)=V(A),
teats z(a}
z(A) — y(A) V(A)

‘and tristimulus values of one system to those of
the other.
The particular choice of the X, Y,Z primary
stimuli gave the (X,Y, Z)-system a further im-
portant property. Following a suggestion by Judd
(1930a), the X and Z primary stimuli are located
on the alychne which, in the chromaticity dia-
gram, is a straight line on which are located the
chromaticity points of all stimuli (they are all
nonreal) having zero luminance. In the (7, g)-
chromaticity diagram, the equation of the alychne
is given by
Dart Lee t+ Lo = 0" [6(33:33)|
where Lp = 1, Lo = 4.5907, Lz = 0.0601 (see
above).
With the primary stimuli X and Z placed on
the alychne, their luminous efficiencies Ly and
Lz must equal zero, and the luminous efficiency
140 COLORIMETRY
Ly of the remaining Y primary stimulus may be
set equal to unity. This makes the color-matching
function y(A) identical to the luminous efficiency
function V(A); that is,
V(A) =)
and facilitates the computation of photometric
quantities from color-matching data.
In Figure 3(3.3.3) the CIE 1931 color-match-
ing functions x(A), y(A), Z(A) based on the
X, Y,Z primary stimuli are plotted side by side
with the color-matching functions r(A), g(A),
b(A) based on the R,G,B primary stimuli. The
same figure also shows the corresponding (x, y)-
chromaticity diagram together with the (7, g)-
chromaticity diagram. In both diagrams, the
chromaticity point E of the equal-energy stimulus
is at the centroid of the chromaticity triangle;
that is, x; =y,=74 and r;—g,—+4, respec-
tively.
In the CIE 1931 colorimetric system, the
color-match conditions, defined earlier in Eq.
12(3.2.3), are expressed as follows:
[PrzQ) dr = [PaxX(A)dn
[PrnvO)dr= JParva)
[PnZ(r)aa = [Panz(A)da [7(3.3.3)]
where, as before, {P,,dA} and {P,,dA} define
the spectral radiant power distributions of two
metameric color stimuli Q, and Q,, respectively.
The integration is taken over the visible spectrum
from A, = 360 nm to A, = 830 nm. In practice,
the integration is replaced by a summation; the
procedures used to carry out the summations are
given in Section 3.3.8.
(ii) The CIE 1964 Standard Observer. The
color-matching functions of the CIE 1964 supple-
mentary standard colorimetric observer are based
on experimental investigations made by Stiles and
Burch (1959) and Speranskaya (1959).
Stiles and Burch used a trichromatic colorime-
ter with monochromatic primary stimuli to de-
termine for 49 observers the tristimulus values of
monochromatic stimuli of equal energy and of
wavelengths from approximately 390 to 830 nm.
In the course of the investigation, different sets of
primary stimuli (all monochromatic stimuli) were
employed, but the final mean results were all
transformed to refer to the primary stimuli R
(645.2 nm), G (526.3 nm), and B (444.4 nm). As
Stiles and Burch measured the radiant power
{P,dX} of each monochromatic test stimulus
presented to the observer, the color-matching
functions were obtainable directly from the ob-
servations, and no appeal to heterochromatic
brightness measurements or to any luminous ef-
ficiency function was required. In this respect, the
work differed from earlier investigations on color
matching such as those by Guild and Wright with
a 2° visual field.
To reduce possible rod intrusion, the lumi-
nance of the matching field was kept high. At the
ends of the spectrum, it was kept as high as
possible within the limits of the apparatus and
light sources. (A xenon arc was used in the long-
wave region.) A correction for rod intrusion was
applied in arriving at the final mean data, but it
was very small and would make a negligible dif-
ference in the calculation of the tristimulus values
for almost all cases of practical interest. Figure
5(3.3.3) shows the mean 10° color-matching func-
tions of the 49 observers used by Stiles and Burch.
Speranskaya used 18 observers (later increased
to 27) who determined 10° color-matching func-
tions with results essentially similar to those of
Stiles and Burch, although the experimental con-
ditions of the two studies differed in several re-
spects. Speranskaya used a 10° visual field with
the central area (2°) masked off to eliminate the
Maxwell spot. The luminance of the visual field
was in general 30 to 40 times lower than in the
measurements of Stiles and Burch and was at a
level where the rod mechanism systematically af-
fected the results, particularly the values of b(A)
in the yellow, orange, and red regions of the
spectrum. The primary stimuli in Speranskaya’s
instrument were obtained from a tungsten fila-
ment lamp in conjunction with colored glass filters
giving bands of dominant wavelength of ap-
proximately 640, 545, and 465 nm, respectively.
Speranskaya transformed her data to the same
system of monochromatic primary stimuli as used
by Stiles and Burch, that is, R (645.2 nm), G
(526.3 nm), B (444.4 nm). Figure 5(3.3.3) shows
the mean values in this form in comparison with
those of Stiles and Burch.
Judd, acting for the CIE, averaged the color-
matching functions obtained by Stiles and Burch
and Speranskaya, after correcting the Speranskaya
data for rod intrusion (CIE, 1959). In taking this
average, different weights were assigned to the
 The CIE Colorimetric System 14]

3.0

STILES-BURCH
XXxxx SPERANSKAYA

7p)
WW
=)
See
<q
>
no
=)
=)
=
=

2) 1.0
ac
Ke

0.0

WAVELENGTH ) (nm)

Fig. 5(3.3.3). Mean 10° color-matching functions r49(A), Z19(A),5 1o(A) of 49 observers used
by Stiles and Burch (1959), and of 18 observers used by Speranskaya (1959) in the system of
monochromatic primary stimuli R at A 2 = 645.2 nm, G at A, = 525.3 nm, B at A = 444.4 nm.
The units of the primary stimuli are fixed at unit radiant power.

two sets of data, major weights being given to the
Stiles and Burch results. Throughout most of the
spectrum, the ratio of weighting was set at three
to one, approximately in the ratio of the number
of observers in the two studies. At the two ends of
the spectrum, this ratio was progressively in-
creased so that the mean color-matching func-
tions progressed smoothly into the Stiles—Burch
data that extended over the greater spectral range.
The smoothed and extrapolated average results of
the color-matching functions and corresponding
chromaticity coordinates are given in Table
II(3.3.3). Note that in this table, the monochro-
matic stimuli are identified in terms of wavenum-
bers m=(1/A)- 10’ as well as wavelengths A;
with A given in nanometers, m measured in cm !.
The visible spectrum is covered from m = 27,750
cm! (A = 360.4 nm) to m = 12,250 cm | (A=
816.3 nm) at constant wavenumber intervals of
Am = 250 cm™~'. Figure 6(3.3.3) shows the de-
rived average color-matching functions 7,9(A),
Z10(A), Dio(A) on the wavelength instead of wave-
number basis.
The CIE elected to use for the large-field
standard observer a coordinate system con-
structed as far as possible on the same general
plan as the 1931 CIE (X, Y, Z)-system. The trans-
formation equations finally agreed for converting
the 10° color-matching functions 7,9(A), 219(A),
b,o(A) to new standard functions X,9(A), Vio(A),
Zi9(A) are as follows:
X19(A) = 0.3410807,9(A) + 0.1891452,.(A)
+ 0.387529b
(A)
Jio(A) = 0.1390587,9(A) + 0.837460 219(A)
+ 0.073316b,.(A)
Z19(A) = 0.0000007,.(A) + 0.0395532,.(A)
+2.0262006,(A) [8(3.3.3)]
The transformed data were interpolated and ex-
trapolated for every nanometer from A = 360 to
830 nm, and corresponding chromaticity coordi-
nates were also computed. Minor arbitrary adjust-
ments were made to smooth the final values and
142 COLORIMETRY

CIE 1964 Supplementary Standard Colorimetric System

(R,G,B) - System (X,Y,Z) - System

2.0

KilA), Hig A)s Zig (A)

3.0
Fig(A)s Dol), DigA) color - matching
functions
color - matching
functions

[e)

Psy

Tristimulus
Values Values
Tristimulus
0.5
0.0
Wavelength A

526.3
nm—— 645.2
—
nm
0.0

— ~—.
a
444.4
>
—
nm i>) 2

4908
08

(119+ G19) - chromaticity (Xi91 Yio ) - chromaticity

diagram diagram

0.2

-2.0
(Z) 0.8

Fig. 6(3.3.3). Color-matching functions and corresponding chromaticity diagrams of the CIE
1964 supplementary standard colorimetric observer in the system of real primary stimuli
R,G,B, and in the transformed system of imaginary primary stimuli X, Y,Z. The (X, Y, Z)-sys-
tem constitutes the official CIE 1964 system for practical colorimetry involving color-matching
fields of angular subtense greater than four degrees. The units of the R,G,B primary stimuli
are of unit radiant power. The units of the X,Y,Z primary stimuli have been fixed so that the
chromaticity point E of the equal-energy stimulus plots at the centroid of the CIE 1964
(195 ¥i9)-chromaticity diagram, that is x,9(E) = y49(£) =3-

eliminate slight and probably insignificant con-
cavities of the spectrum locus. The tristimulus
values X19(A), Vio(A), Zi9(A) Of the monochro-
matic constituents of the equal-energy stimulus
and the chromaticity coordinates X19, Yi95 Z10
so obtained are those given in Table [(3.3.2).
They define the CIE 1964 supplementary colori-
metric observer. Figure 6(3.3.3) illustrates the
Xio(A)s Vio(A)s Ziop(A) functions and the corre-
sponding CIE 1964 (x9, ¥j9)-chromaticity dia-
gram together with the 7,9(A), 210(A), bio(A)
color-matching functions and corresponding
(Ti0> Z19)-chromaticity diagram.
For most practical purposes, an abridged set
of the color-matching functions is adequate; such
a set Xi9(A), Vio(A), Zi9(A) is given in Table

II(3.3.2) together with corresponding chromaticity
coordinates.
In the CIE 1964 colorimetric system, the
color-match conditions are similar to those given
as Eq. 7(3.3.3) for the CIE 1931 colorimetric
system. The following equations are valid:
[Pirxio(A)dd = [Prrxo(A)ar
Xr rv
[Pirvo(r)ar = [Poy Iro(A) da
r r
if
Py,Z,9(A)dA = [Py,Z(A)ddA [9(3.3.3)]
To these equations, the same comments apply as
those given to Eq. 7(3.3.3).
3.3.4 CIE Standard Illuminants
In industrial colorimetry, the main concern is
with the measurement of the color of objects.
Light emitted from a source of radiant power,
such as the sun or a tungsten filament lamp,
strikes a given object and thus, makes the object
visible to the observer. The color stimulus Q
emerging from the object and entering the ob-
server's eye produces a sensation of color that the
observer identifies as the color of the object. The
spectral radiant power distribution {P,dA} of the
object-color stimulus Q must clearly be the spec-
tral distribution {B(A)S,dA}, where S, denotes
the spectral concentration of the radiant power
incident on the object and B(A) denotes the spec-
tral reflectance properties of the object de-
termined in accordance with given geometrical
conditions of illumination and viewing. The ex-
pression
{P,dd} = {B(A)S,dA} [13.3.4)]
applies only to nonfluorescent objects which will
be assumed here. The case of fluorescent objects
will be treated in Section 3.3.8.
The color-match conditions of Eq. 7(3.3.3) and
9(3.3.3) for two object-color stimuli Q, and Q,
involving two objects of different spectral reflec-
tance properties B,(A) and £,(A), and a given
light source providing the incident radiant power
The CIE Colorimetric System 143
distribution {S,dA}, are as follows:
[BASHA = [Bo(A)S\X(A)AD
[BAS FADIA = [BAYS H(A)
[BiA)Sz(A)Ar= [B(A)S\z(A)Ad
[2(3.3.4)]
In these equations it is assumed that the color
match between the two metameric color stimuli
is established by the CIE 1931 standard observer
defined by the color-matching functions
X(A), y(A), Z(A).
Similar equations exist that involve the CIE
1964 supplementary standard observer defined by
the color-matching functions Xj (A), Vio(A),
Zio(A). However, for Eq. 2(3.3.4) to be satisfied
with these color-matching functions, a different
pair of objects, defined by B,(A) and B,(A), is
implied.
Whenever object-color stimuli are of main
concern, it is desirable to restrict colorimetric
measurements and calculations to a few specific
and well-defined spectral distributions {S,dA} of
radiant power incident on the objects under study.
The CIE recommends a set of such spectral radiant
power distributions called CIE standard illumi-
nants.
The spectral radiant power distributions of the
CIE standard illuminants are given in terms of an
arbitrary unit of radiant power and thus are re-
ferred to as relative spectral radiant power distri-
butions {S(A)dA}. For colorimetric calculations
that aim at testing whether the color-match condi-
tions defined by Eq. 2(3.3.4) are satisfied, absolute
spectral radiant power distributions, given, for
example, in terms of W - m ’, are obviously not
required, because relative and absolute spectral
distributions differ only by a constant and wave-
length independent factor, that is
{S,dA} =c{S(A)dA} [3(3.3.4)]
which cancels when inserted in Eq. 2(3.3.4).
(i) CIE Standard Iluminant A. CIE standard
illuminant A represents light from the full radiator
at absolute temperature 2856 K according to The
International Practical Temperature Scale, 1968
144 COLORIMETRY

200 |—

RELATIVE
RADIANT
POWER
50+

=I
300 400 500 600 700 800

WAVELENGTH AX (nm)

Fig. 1(3.3.4). Relative spectral radiant power distributions of CIE illuminants A, B, C, and
D¢s °

(CGPM, 1969). Its relative spectral power distri- as 1.4380 10°? m- K which has led the CIE to
bution {S,(A)dA} has been calculated by means assign to illuminant A the temperature T = 2854
of Planck’s radiation formula, given in Section K, so as again to retain the 1931 agreed upon
1.2.2, and is given in Table I(3.3.4) for wave- relative spectral radiant power distribution.
lengths A from 360 to 830 nm at one nm intervals.
(ii) CIE Standard Iluminant D,;. CIE stan-
An abridged set of values for {S,(A)dA} is given
dard illuminant D,,; represents a phase of natural
in Table 1(1.2.2) and Table II(3.3.4). Figure
daylight with a correlated color temperature of
1(3.3.4) illustrates CIE standard illuminant A.
approximately 6504 K. Its relative spectral radiant
The relative spectral radiant power distribu-
power distribution {Sp..(A)dA} is given in Table
tion of illuminant A has not changed since it was
(3.3.4) for wavelengths A from 360 to 830 nm at
first recommended by the CIE in 1931. However,
one nm intervals. An abridged set of values for
its temperature T = 2856 K, assigned by the CIE
{Sp,(A)dA} is given in Table II(3.3.4). Figure
in 1971, has changed. Originally T was set at 2848
1(3.3.4) illustrates CIE standard illuminant D,;.
K, but at that time the radiation constant c,,
The experimental work on which illuminant.
which in Planck’s radiation formula is the essen-
Ds is based has been referred to in Section 1.2.1.
tial parameter for the relative spectral radiant
A detailed account of the concept of correlated
power distribution, was assumed to be equal to
color temperature is given in Section 3.11. It
1.4350 10°? m- K, in accordance with The Inter-
suffices to note here that the assignment of a
national Practical Temperature Scale, 1927. The
correlated color temperature to D,;, and in fact,
current value is c, = 1.4388 10°? m- K. By ad-
any illuminant whose relative spectral radiant
justing T from 2848 to 2856 K and adopting the
power distribution is fixed, depends on the speci-
current value of c,, the relative spectral radiant
fication of isotemperature lines in the CIE 1931
power distribution is kept unchanged. The rela-
(x, y)-chromaticity diagram and the adopted val-
tion that was used in 1971 to set the new tempera-
ues of Planck’s radiation constant c,.
ture for illuminant A is
_ 1.4388 (iii) Other CIE Standard Illuminants. In 1931
~ 1.4350
2848 K = 2856K [4(3.3.4)] the CIE recommended, in addition to illuminant
A, standard illuminants B and C. CIE standard
For a number of years prior to 1971, c, was given illuminant B was intended to represent direct
 The CIE Colorimetric System 145

sunlight with a correlated color temperature of cluding direct sunlight, skylight, and sunlight with
approximately 4874 K; CIE standard illuminant C different cloud coverage, overcast skylight, and
was intended to represent average daylight with a light from the blue sky.
correlated color temperature of approximately In the interest of standardization however, the
6774 K. CIE recommends that D,; be used whenever pos-
The relative spectral radiant power distribu- sible. When D,; cannot be used, it is recom-
tions {Sp(A)dA} and {Sc(A)dA} of illuminants mended that either D;, or D7, be considered for
B and C are given in Table II(3.3.4). They are the application at hand. The illuminants D,, and
illustrated in Figure 1(3.3.4) together with those D,, have correlated color temperatures of ap-
of A and Ds. proximately 5503 K and 7504 K, respectively.
Both illuminant B and illuminant C are con- Table III(3.3.4) gives the values of {Sp.(A)d\}
sidered inadequate in representing the intended and {Sp_(A)dA} and Figure 2(3.3.4) illustrates
phases of natural daylight. In particular, both are these spectral distributions in comparison with
seriously deficient in their spectral distribution in that of Ds.
the ultraviolet region that plays an important role (iv) CIE Method of Calculating D-Mluminants.
in the colorimetry of fluorescent materials. The The correlated color temperature 7, of a D-illumi-
CIE has indicated that at some future date that is nant is related to the chromaticity coordinates
yet to be decided, illuminants B and C will be (Xp; Yp) of that D-illuminant by the following
dropped from the list of recommended standard equations: For correlated color temperatures from
illuminants (CIE, 1971). In practice, illuminants approximately 4000 K to 7000 K:
B and C have already fallen into disuse in most
9 6
applications. Instead, CIE standard illuminant D,; Xp = ~4.607012- a 2.967812
ie Ts
is now widely used as the representative of aver-
age daylight for colorimetry. 3
In 1964, when the CIE recommended TD; as +0.09911 7 + 0.244063
the main standard daylight illuminant, it also
recommended a method of calculating the relative For correlated color temperatures from ap-
spectral radiant power distributions of any day- proximately 7000 K to 25,000 K:
light illuminant D with correlated color tempera- na B 10° 10°
Xp = SS aut + | ON
tures T. ranging from 4,000 to 25,000 K. The
c (&
method, given below, provides a set of spectral
distributions {S,(A)dA} that covers a wide range 103
0.24748
+0. a + 0.237040 [5(3.3.4)]
of phases of daylight encountered in nature, in- c

160

_—ms(jo)

oype)oO

100

COro)

(op)oO

> (a,
units)
(arbitrary
Radiant
power

pe)co)

300 400 500 600 700 800 900

Wavelength (A), nm

Fig. 2(3.3.4). Relative spectral radiant power distributions of CIE illuminants D,;, De;, and
D,s.
146 COLORIMETRY

5500 4800 K
6500

0.300

0.250

0.250 0.300 0.350 0.400

x
Fig. 3(3.3.4). Portion of CIE 1931 (x, y)-chromaticity diagram illustrating CIE daylight locus
D in comparison with Planckian locus P. Also shown are isotemperature lines 4,800, 5,500,
6,500, 7,500, and 10,000 K defining straightline loci of constant correlated color temperature
(see also Section 3.11).

The yp-chromaticity coordinate is related to xp are defined by

as follows:
M, < 13515 = 1.7703xp + 5.9114yp
| 0.0241 + 0.2562x, — 0.7341 yp
Vp = —3.000xf + 2.870xp — 0.275
_ 0.0300 — 31.4424x, + 30.0717yp
[6(3.3.4)]
2 0.0241 + 0.2562x, — 0.7341 yp

which defines the “CIE daylight locus” in the [8(3.3.4)]

CIE 1931 (x, y)-chromaticity diagram as shown
in Figure 3(3.3.4).
The relative spectral radiant power distri-
bution {Sp)(A)dA} of a D-illuminant is given by
a linear combination of a mean distribution
{So(A)dA} and two distributions {5,(A)dA} and
{S,(A)dA} representing the two most important
“eigenvectors” of the set of all daylight distribu-
tions; thus,
Sp(A) = So(A) + M,S\(A) + MZ S,(A)
[7(3.3.4)]
Table IV(3.3.4) gives the chromaticity coordi-
nates (Xp, Yp) and scalar multipliers (M,, M,)
for the entire range of daylight phases from 4000
K to 25,000 K at close intervals. Table V(3.3.4) |
contains the mean and two eigenvector functions
[So(A), S\(A), S3(A)] which are also illustrated in
Figure 4(3.3.4).
Examples of calculations made by the above
method are given in Table 2(1.2.1) and illustrated
in Figure 3(1.2.1). In Section 1.2.1 are also given
some other relevant data concerning spectral
radiant power distributions of daylight.

3.3.5 CIE Standard Sources

where M, and M, are scalar multipliers that are
functions of the chromaticity coordinates The CIE distinguishes between sources and illumi-
(Xp, Yp) of the D-illuminant. These multipliers nants. Whereas a source refers to a physical emitter

WAVELENGTH
Fig. 4(3.3.4). Mean spectral radiant power distribution S)(A) of all phases of daylight and the
two first eigenvector functions S,(A) and S,(A) used in the CIE method of calculating daylight
illuminants.
of radiant power, such as a lamp or the sun and
sky, an illuminant refers to a specific spectral
radiant power distribution incident on the object
viewed by the observer. The spectral radiant power
distribution of an illuminant may not necessarily
be exactly realizable by a source. Illuminants,
particularly those recommended by the CIE as
standard illuminants, find their primary use in
colorimetric calculations. Sources are needed to
conduct visual colorimetry or color measurements
by means of physical colorimeters.
The CIE recommends the following sources as
standard sources whenever it is desired to realize
the standard illuminants, defined in Section 3.3.4,
for actual laboratory inspection of colored
materials.
(i) CIE Standard Source A. CIE standard il-
luminant A is to be realized by a gas-filled
coiled-tungsten filament lamp operating at a cor-
related color temperature of 2856 K. A lamp with
a fused-quartz envelope or window is recom-
mended if the spectral distribution of radiant
power in the ultraviolet region of illuminant A is
to be realized more accurately.
(ii) CIE Standard Sources B and C. CIE stan-
dard illuminants B and C are to be realized by
The CIE Colorimetric System 147
600 700 800
X (nm)
CIE standard source A, combined with a filter
consisting of a layer, 1 cm thick of each of two
solutions B, and B,, and C, and C), respectively,
contained in a double cell made of colorless opti-
cal glass. The solutions are made up in accor-
dance with Table 1(3.3.5). The liquid filters
required to change source A to B and C are
commonly called Davis—Gibson filters (Davis and
Gibson, 1931).
CIE standard sources B and C are not used
much in practice. The liquid filters are balky and
require periodic refilling. The illuminance levels
that normally can be achieved with such sources
are relatively low and confined to small areas of
illumination, making them unsuitable as sources
for visual inspection. Their relative spectral
radiant power distributions, closely correspond-
ing to CIE standard illuminants B and C, do not
represent adequately spectral distributions of
daylight, particularly in the ultraviolet region of
the spectrum (see Section 3.3.4).
(iii) CIE Standard Sources D, Daylight Simula-
tors. At present, the CIE has made no recom-
mendations of artificial sources to realize any of
the CIE illuminants D. The difficulty lies in the
unique and rather jagged spectral distribution of
148 COLORIMETRY

Table 1(3.3.5) Davis—Gibson Filters to these samples match in color when irradiated by a
Convert CIE Standard Source A to CIE given CIE standard daylight illuminant D and
Standard Source B and CIE Standard when viewed by the CIE 1964 supplementary
Source C standard colorimetric observer. A test source of
known spectral radiant power distribution, differ-
Source B filter
ent from that of the given standard daylight
illuminant D, will generally cause color mis-
Solution B, matches between the sample pairs that originally
Copper sulfate (CuSO,:5H2O) 2.452 g
were complete color matches. The magnitudes of
Mannite [C,H,(OH)¢.] 2.452 g the mismatches, calculated by means of a CIE
Pyridine (C;H;N) 30.0 ml color-difference formula (see Section 3.3.9), lead
Distilled water to make 1000.0 ml
to two indices, M/,,, and MI,,,. These indices are
Solution By used to categorize the quality of the test source in
the visible part of the spectrum and in the ultra-
Cobalt ammonium sulfate Dilareliee 2 violet part, respectively. Table 2(3.3.5) defines the
[Co SO,°(NH,4).SO,°6 H,0]

Copper sulfate (CuSO,:5H,O) 16.11 g

categories. For each index, five categories (A to
Sulfuric acid (density 1.835 10.0 ml E) are defined. The correlation of these categories
g-ml~') for various practical applications must be
Distilled water to make 1000.0 ml
determined by experience which, at this time, has
Source C filter not been fully developed. However, a daylight
simulator categorized as a “BC”-daylight simula-
Solution C,
tor would be considered useful for many colori-
Copper sulfate (CuSO,:5H,O) 3.412 g metric applications. According to Table 2(3.3.5),
Mannite [C,H,(OH)¢] 3.412 g its relative spectral radiant power distribution
Pyridine (C;H;N) 30.0 ml would yield an index MI,,, (CIE L*u*v*) be-
Distilled water to make 1000.0 ml
tween 0.32 to 0.65, resulting in category B for the
Solution C2 visible part of the spectrum, and an index MI,,
(CIE L*u*v*) between 0.65 to 1.30, resulting in
Cobalt ammonium sulfate 30.58 g
[CoSO,:(NH,4).SO,:6H,O)]
category C for the ultraviolet part of the spec-
Copper sulfate (CuSO,-5H,O) Ap SP trum.
Sulfuric acid (density 1.835 10.0 ml An additional requirement imposed on a day-
g-ml~') light simulator is that the CIE 1964 (xj, yi9)-
Distilled water to make 1000.0 ml
chromaticity point calculated from the spectral
radiant power distribution of the simulator must
fall within a specified small domain of chromatic-
daylight as shown in Figure 2(3.3.4). No artificial ity centered at the chromaticity point of the given
sources with such spectral distributions are known, CIE standard daylight illuminant D. Figure
and modifying the spectral distributions of exist- 1(3.3.5) illustrates the permissible chromaticity
ing sources by placing filters in front of them or domains for sources simulating standard illumi-
using other means has been only partially success- nants D,;, D,;, and Ds. |
ful (McLaren, 1967; Henderson, 1968; 1977; The CIE method for assessing the quality of
Grum et. al., 1970; Wyszecki, 1970). daylight simulators for colorimetry (CIE, 1980) is
In recognizing the unlikelihood of ever obtain- too elaborate to be reproduced here in detail.
ing an artificial source with a relative spectral However, because of its importance to colorime-
radiant power distribution identical to that try, an extract of the material is given which is
defined as either D¢;, D;5, or D7;, the CIE Col- confined to the assessment of the quality of
orimetry Committee has developed a method for sources simulating CIE standard illuminant D,;.
assessing the usefulness of existing artificial The main steps of the procedure are briefly as
sources for colorimetric purposes (CIE, 1980). follows:
Sources that qualify are referred to as daylight
simulators for colorimetry. (a) Spectral radiant power distribution of test
The CIE method employs a set of fluorescent source. The relative spectral radiant power dis-
as well as nonfluorescent samples specified in tribution of the test source is determined by spec-
terms of their spectral radiance factors. Pairs of troradiometry from A = 300 to 700 nm at 5-nm
 The CIE Colorimetric System 149

Table 2(3.3.5) Categories of Daylight Simulators intervals. At these wavelengths, the data are given
for Colorimetry (CIE, 1980) in terms of spectral irradiance at the sample
surface for bands of wavelengths of 5-nm width.
Indices M/,,, and MTI,,, They are denoted by S(A).
Determined by Color-Difference Formula Only relative values of spectral irradiance are
required. The following normalization of the
CAE fab GIE Li 6* Category
spectral data S(A) is made:
< 0.25 < 0,32 A
0.25 to 0.50 0.32 to 0.65 B
100S(A) [1(3.3.5)]
0.5 to 1.0 0.65 to 1.30 Cc SO) = S00 S(d) Fold)
1.0 to 2.0 1.3 to 2.6 D
= LA) > 2.6 Es (b) Chromaticity of test source. For the spec-
tral data SA), the CIE 1964 (x9, yj9)-chro-
“Two indices MJ,,, and MI, are calculated in accor- maticity coordinates are computed in accordance
dance with a procedure outlined in the text. MJ,;, with procedures given in Section 3.3.8 using 5-nm
specifies the category of a given daylight simulator with interval summations. For the test source to qualify
regard to its quality in the visible spectrum; M/,, as a daylight simulator of D,;, the calculated
specifies the category of the same daylight simulator (X19 Yio)-chromaticity point must fall within the
with regard to its quality in the ultraviolet region of the ellipse shown in Figure 1(3.3.5) centered at the
spectrum. A daylight simulator is completely cate-
chromaticity point of D,;. In the CIE 1976
gorized by two letters, for example, “BD”. The first
letter (“B”, in the example) identifies the category
(Ujo, Vio)-uniform chromaticity scale diagram,
applicable to the visible spectrum. The second letter Section 3.3.9, the ellipse is a circle of radius 0.015,
(“D”, in the example) identifies the category applicable exactly.
to the ultraviolet region of the spectrum of the given (c) Calculation of index MI,,, for visible part of
daylight simulator.
spectrum. CIE 1964 (Xjo, Yio, Zj9)-tristimulus
values are calculated for five pairs of samples
irradiated by the test source whose spectral distri-
bution of irradiance S,(A) was determined under

0.400

0.300

0.250 0.300 0.350

X10
Fig. 13.3.5). Portion of CIE 1964 (xj, y19)-chromaticity diagram showing elliptical chro-
maticity domains centered at the chromaticity points of CIE standard illuminants D,,,D¢5,D-5,
within which the chromaticity points of daylight stimulators must fall. The ellipses are circles of
equal radius (r = 0.015) in the CIE 1976 (uo, v9)-Uniform Chromaticity Scale Diagram.
150 COLORIMETRY

(a) above. The X;9(A), Vio(A), Z19(A) color-match-

ing functions are used. The summations leading
to the (Xj9, Yio, Z19)-tristimulus values are again
carried from 400 to 700 nm at 5-nm intervals.
Each pair of samples, whose spectral radiance
factors are given in Table 1(3.3.5), provides a
metameric pair under CIE standard illuminant
D,; when viewed by the CIE 1964 supplementary
standard observer. Figure 2(3.3.5) illustrates the
spectral radiance factor curves of the five sample
“400 500 600 700
pairs. The degree of metamerism exhibited by
these five sample pairs is small or moderate, but
they are considered representative of actual sam-
ples found frequently in practice, particularly in
the textile industry.
The color differences AE; (i = 1 to 5) for the
five sample pairs irradiated by the test source are
calculated by either the CIE 1976 L*a*b* or the
CIE 1976 L*u*v* formula as given in Section
Aide9:
The index MI,,,, also referred to as the
“400 500 600 700 metamerism index for the visible part of the spec-
trum, is the mean of the five color differences,
that is

s Ap eon

(d) Calculation of index MI, for ultraviolet part

of spectrum. CIE 1964 (Xo, Yio, Zi9)-tristimu-
lus values are calculated for three pairs of samples
irradiated by the test source whose spectral distri-
bution of irradiance S,(A) was determined under
(a) above. The X;o(A), Vio(A), Z19(A) color-match-
ing functions are used. The summations leading
to the (X19, Yio, Zi9)-tristimulus values are car-
ried out from 400 to 700 nm at 5-nm intervals, in
the same manner as under (c) above.
Each pair of samples consists of a nonfluo- —
rescent and a fluorescent sample. Table II(3.3.5)
Factor
Radiance
Radiance
Factor
Radiance gives the spectral radiance factors, B,(A) (j = 1
to 3), for the nonfluorescent samples as well as
the spectral data, Q,(A’), B;,(A), F(A), required
400 500 600 700 to calculate the total spectral radiance factors
B,(A) of the fluorescent samples for any test
source of known spectral radiant power distribu-
tion S,(A).
The following equations are used to calculate
the total spectral radiance factor B,(A) for a test

Fig. 2(3.3.5). Spectral radiance factors of five

Factor
Radiance
pairs of nonfluorescent samples that provide
metameric pairs under CIE standard illuminant
0.0
400 500 600 T00 Des when viewed by the CIE 1964 supplementary
standard observer.
 The CIE Colorimetric System tay

source with given S,(A): The different symbols in the above equations
denote the following quantities:
Br(A) = Bs(A) + BCA) [3(3.3.5)]
B,(A)= Total spectral radiance factor
where B;(A)= Spectral radiance factor (reflection)
B,(A)=Spectral radiance factor (lumines-
BH AAO)
QeSi(A)
ALE cence)
F(A) =Relative spectral radiant power
generated by luminescence [2 F(A)
and = 1.0]
bee Q(A’)=External spectral radiant efficiency
N= oD SA(N)O(X) [5(3.3.5)] of luminescence
d’ =360 N=Effectiveness of excitation

ke
®
=
fe)
ou
- fo) B
Cc
i.e]
ao]
5 fe) §
a
@®
>
=

=
@
a )
luminescence
by
(generated

Oo

Factor
Radiance
)
only
(reflection

(@)On

Luminescence
of
Q)(7) Q2(")

Efficiency
Radiant
External

Fig. 3(3.3.5). Spectral distributions of the radiometric quantities of three fluorescent samples
that are required in the calculation of the total spectral radiance factors of these samples when
they are irradiated by a source of known spectral radiant power distribution.
152 COLOR IMETRY
S,(A or A\’}=Normalized spectral radiant power
(irradiance) of test source
A\=Wavelength of incident, reflected,
or emitted radiant power
\’= Wavelength of incident radiant pow-
er causing excitation of lumines-
cence
Figure 3(3.3.5) illustrates the spectral data
given in Table II(3.3.5) which are needed in the
calculation of B;,(A) for each of the three fluo-
rescent samples.
Figure 4(3.3.5) shows the spectral radiance fac-
tors BA) of the three nonfluorescent samples
listed in Table II(3.3.5). These curves coincide
with the curves of the total spectral radiance
factors B;,(A) of the corresponding fluorescent
samples when these are calculated with S,(A)
being that of CIE standard illuminant D,;.
In other words, under D,; the spectral radiance
factors of the two members of each sample pair
j are identical at every wavelength from 400 to
700 nm. This automatically yields identical
(Xo, Yio, Zio)-tristimulus values for the two
members of the sample pair (i.e., a complete color
match is ensured under D,;). In the strict sense,
the pair does not constitute a metameric pair
because the spectral radiance factors of its two
members do not differ. However, when D,; is
replaced by another source, that is, the test source
with a different S,(A), the fluorescent member of
the sample pair will yield different total spectral
radiance factors B;,(A) in accordance with Eqs.
3(3.3.5) to 5(3.3.5). The spectral differences will
be confined to the short-wavelength region of the
visible spectrum where the fluorescence [i.e.,
luminescence F;(A)] occurs. These spectral dif-
ferences lead to a color mismatch between the
two members of the sample pair, which, in the
main, is caused by the differences in the ultra-
violet part of the spectrum between the spectral
radiant power distributions of the test source and
Dgs.
The color mismatches are specified by the
color differences AE; (j = 1 to 3) for the three
sample pairs irradiated by the test source, using
again either the CIE 1976 L*a*b* or the CIE
1976 L*u*v* formula. The index MI/,, is then
defined as the mean of the three color differences,
that is,
MI., == s AE, —[6(3.3.5)]
j=
Factor
Radiance
d (nm)
Fig. 4(3.3.5). Spectral radiance factors of pairs
of samples that are metameric pairs when irradia-
ted by CIE standard illuminant D,, and viewed by
the CIE 1964 supplementary standard observer.
(e) Interpretation of the indices MI,,, and MI,,,,.
The indices MI/,,. and MI,,, calculated by means
of Eqs. 2(3.3.5) and 6(3.3.5), respectively, are en-
tered into Table 2(3.3.5) to find for each index the
corresponding category (A to E£) that is used to
indicate the quality of the test source as a daylight
simulator in the ultraviolet and in the visible part
of the spectrum. A double-letter specification of
the source is the result.
Figure 5(3.3.5) shows three examples of rela-
tive spectral radiant power distributions of pro-
posed D,;-daylight simulators (Wyszecki, 1970).
Their qualities have been assessed by the CIE
method described above. Table 3(3.3.5) gives the
result of the assessment. The quality categories
found for the three D,.-daylight simulators ap-
pear to correlate well with what one might expect
by comparing their relative spectral radiant power
distributions with that of D,, in Figure 5(3.3.5).
The filtered high-pressure xenon arc lamp (Xe 5)
is of very high quality in both the visible part and
the ultraviolet part of the spectrum, indicated by
the letters B and A, respectively.
The usefulness of the CIE method of assessing
the quality of daylight simulators for practical
colorimetric purposes has not been established as
yet. To a large extent, it will depend on whether
the samples selected for the method are indeed
representative of all or most practical situations.
It is conceivable that, in some situations, the
degree of metamerism exhibited by the five pairs
of samples shown in Figure 2(3.3.5) is too low
and thus does not represent the actual situations.
In such cases, daylight simulators of higher qual-
ity would be required than would be determined
by the CIE method.
 200

180
= +i an
Filtered high-pressure
©2
°o es,

(o>)
x< o jes

(sun Asesjique) 4amod jueipey

oO

aa 4 ' i i 1 a
jo)
|
et
Con
Lol
mn

oO

es
ace
ek
Be ey
jo)

al y
bye
oO
N
o
=
Ve)
oO
fo)

300 400 500 600 700 800 900

Wavelength (A), nm
N ro)
(=)

a ioe)
(je) Cyex
iss
:: eyes elo:2
Site =
© ro) fey)wn
a)Xe)
a)
(sjiun
Ajesjique)

a + je)

a)oe)
(@)

& i ; ‘
jueipey samod

a)o)
(2)
:i - }
ioe)
(2)
Fi
x
oOjo) ¥
f 5
# ' ‘i

+ (=)

N (@)

4
iO! i i Te)fo) o a]ro) Ss oO 00 S)
+ Oo(oe) fo) ;(=) ro) ron)(S)oO
Wavelength (A), nm

NO
(Sie)
[oe

WOO’
poe)
ee
Leer eS Ow
63
O Ks) te 2 is) ® wnoO ® c ~ As) £ for
ce Ae ee
ice)
oO

Re eee
oO ON a
We)
ee

ee
ee
eww poss
+ oO ee
women

eeaoe
N 2)
eS 2a
eee

ees
ee
eae
oO2)
a
Sa ‘)
:
co
onl
a)
NMNNN
a]
N
a = = 7slag,
t..i rt
Lt aa
|tt
as
ew

omg a=
eoi=)
ee
Asesjique)
jueipey
(syiun
samod —oee-k,

«!

300 400 500 600 700 800 900

Wavelength (A), nm

Fig. 5(3.3.5). Three examples of relative spectral power distributions of D,.-daylight simula-
tors.

ee)
 = ‘N
‘S97 10} OI9Z 0} [enba “yuauISsasse JO poyjou ay} Jo UsISap Aq ‘ore NW pue “yyy ‘sny) pue sanyea 7V IV 69°01
‘69 L = “N 679 ='N SI UOT}E}OXA JO SSOUBATIAJJO SH ‘TIEE¢0 —Ol ‘LETE'O = Olx ore sayeUTpIOON AjloryeMOIYO Si] “poye[Nuris
poyepnoyeo ay}
aq 01 1y3yAep ay} syuosordor $°qq juURUTUNTI prepueis AID (cee)e aqeuy, Wor pear pue “yw pue “yp saotput
porary = $9X »
UO paseq SI juaUISsasse A1099}"9 ay “durey yUsosoIONTJ = LA ‘durey uais8un} posayfyy = EL {ore wousx sinssaid-ysty
ee eee ee ce F-aRar=Et ann SRS ERS Gn ae
SE
@Zbr 997% BEPL LI'EL 666 PII Srl SPI STI 6€0 €7I ISckO OSTEO LA
AD 16Z Sl
See OTS Shoe Ske, CLI OVO SO LS0 S10 ePO. PEO: tte 0 “UrItO €L
4a fe mol
6€0 910 610 SOIL €8L Lr9 €€0 OFO @HO 610 670 EO OLZEO I9IED g9X
Va cv70
Riise) CIW aw ay ‘VSN ieRiNS. Ee (Fue IY fave eve ay ee Oly — JOYRTNUTTS
woTje}IOXY JO sireg ojdures Ayoneuomy) iwsqyAeq
sireg ojdures jusosarIonyy
SSOUDATINIJ JA JUIISIIONTJUON JO S9oUdIIFJIC JO[OD -59q
JO SdoUdIIIJIG IO[OD
ate" %T 9L61 AIO #%41%T 9L61 ATO
e
ae SS SS SS Se
T
(O86 AID) POW AIO Aq saoyemung yysyAvq-Pq eer], Jo AyTEnd Jo JUsUIssassy (SECOE AIL

154

3.3.6 Standard of Reflectance Factor. The CIE
recommends the perfect reflecting diffuser as the
reference for making measurements of reflectance
factor (CIE, 1971).
The perfect reflecting diffuser is defined as the
ideal uniform diffuser with a reflectance equal to
unity.
There is no material surface that has the prop-
erties of the perfect reflecting diffuser, but by
means of special techniques, it is possible to
calibrate suitable secondary standards, also called
white standards, in terms of the ideal uniform
diffuser. Measurements of that kind are often
referred to as absolute measurements of reflec-
tance factor. Detailed accounts of the various
methods used for such measurements can be found
in the literature (Budde, 1976; CIE, 1979b).
There is a fair variety of materials available
that are useful as secondary standards. Some of
these have been discussed in Section 1.4.2.
3.3.7 Standard Illuminating and Viewing
Conditions
The CIE recommends that the reflectance-factor
measurements be made under one of the follow-
ing illuminating and viewing conditions:
1. (45/0)-CONDITION
The sample is illuminated by one or more beams
whose axes are at an angle of 45 + 5° from the
normal to the sample surface. The angle between
the direction of viewing and the normal to the
sample should not exceed 10°. The angle between
the axis and any ray of the illuminating beam
should not exceed 5°. The same restriction should
be observed in the viewing beam.
2. (0/45)-CONDITION
The sample is illuminated by a beam whose axis
is at an angle not exceeding 10° from the normal
to the sample. The sample is viewed at an angle of
45 + 5° from the normal. The angle between the
axis and any ray of the illuminating beam should
not exceed 5°. The same restriction should be
observed in the viewing beam.
3. (d/0)-CONDITION
The sample is illuminated diffusely by an in-
tegrating sphere. The angle between the normal to
the specimen and the axis of the viewing beam
should not exceed 10°. The integrating sphere
may be of any diameter provided the total area of
the ports does not exceed 10% of the internal
The CIE Colorimetric System 155
reflecting sphere area. The angle between the axis
and any ray of the viewing beam should not
exceed 5°.
4. (0/d)-CONDITION
The sample is illuminated by a beam whose axis
is at an angle not exceeding 10° from the normal
to the sample. The reflected radiant power is
collected by means of an integrating sphere. The
angle between the axis and any ray of the il-
luminating beam should not exceed 5°. The in-
tegrating sphere may be of any diameter provided
the total area of the ports does not exceed 10% of
the internal reflecting sphere area.
These four illuminating and viewing condi-
tions are illustrated schematically in Figure
1(3.3.7). For the conditions (d/0) and (0/d),
provision is made to reduce the specular compo-
nent of samples that are not completely matte: a
gloss trap is incorporated in the design of the
integrating sphere. Its effectiveness depends on its
size, shape, and position.
In the (0/45) and (45/0) conditions, it is
important to note that strictly normal, that is, 0°
illumination, is not recommended. The slight
deviation from strict normal illumination reduces
the possibility of introducing systematic errors in
the measurement caused by interreflections be-
tween samples that are not completely matte and
the optical components that might be used to
illuminate the sample. In the case of the (0/d)-
condition, it also facilitates the measurement with
either the specular component of the sample
excluded (gloss trap inserted) or specular compo-
nent included (gloss trap removed).
Figure 1(3.3.7) shows a small baffle in the
integrating sphere placed between the sample and
the spot of the sphere wall illuminated or viewed.
This baffle reduces the possibility of superimpos-
ing on the sample or the spot on the sphere wall a
component of directly reflected radiant power
which would upset, to some extent, the desired
diffuse illumination or viewing condition.
When the solid angle of the viewed reflected
beam is very small, as could be the case in the
conditions (0/45), (45/0), and (d/0), the mea-
sured reflectance factors, denoted by Bp /45, Bas /o>
and B7/o, are often referred to as radiance factors
with the same notations. When the solid angle of
the viewed reflected beam is equal to or ap-
proaches that of the hemisphere (= 27 stera-
dians), as is usually the case in condition (0/d),
the measured reflectance factor is often referred
to as reflectance and denoted by p. A more de-
156 COLORIMETRY

LLL
OLLI LLG,
0/45

White or black
sphere cap

Integrating
sphere

Baffle

OOOO

O/diff. diff./0

Fig. 1(3.3.7). Schematic diagram showing the four CIE standard illuminating and viewing
geometries for reflectance-factor measurements.

tailed discussion of reflectance-factor measure- Ze k [P,z(A)dd Zo= Kio fPrZio(A)A

ments is given in Section 3.12.2 (see also Section r A
1.4).
X & Xo
3.3.8 Calculation of CIE Tristimulus Values vem Gans
and Chromaticity Coordinates
ye Yio = Yio
(i) General Equations. The general equations
2 Xo + Yio + Zio
that define the CIE tristimulus values and chro-
maticity coordinates follow directly from the dis-
cussions in Sections 3.3.1 to 3.3.3. A distinction [1(3.3.8)]
must be made between those that apply to the
CIE 1931 colorimetric system and those that apply where WX, Y,Z,'x;*y; and® Xp; Yio; Zijp Mor ep
to the CIE 1964 colorimetric system. denote the tristimulus values in the CIE 1931 and
CIE 1964 colorimetric systems, respectively. The
In the CIE 1931 System In the CIE 1964 System sets of color-matching functions x(A), y(A), Z(A),
and X19(A), Vio(A), Z19(A) define the CIE 1931
x= k[ Pyx(A)dd Xn = Kio {PrXi(A)dA standard colorimetric observer (for viewing of
A A matching fields of angular subtense from 1 to 4°)
and the CIE 1964 supplementary standard colori-
metric observer (for viewing of matching fields of
Y=k[PH
A
AAA Yio = kro frv PrFro( AAA angular subtense greater than 4°). The quantity
 The CIE Colorimetric System 157

P, denotes the monochromatic component of The tristimulus values defined in Eq. 1(3.3.8)
wavelength A of the given color stimulus whose then become
spectral radiant power distribution is {P,dA}.
The factors k and k,,) are normalizing factors to In the CTE 1931 System
which values may be assigned arbitrarily, pro-
vided these values are kept constant throughout
any particular discussion involving Eq. 1(3.3.8).
x= kf BOA)S(A)X(A) AA
Special applications of Eq. 1(3.3.8) will be identi-
fied immediately below in which the factors k and
kj, are explicitly defined. The integrations in Eq. Views tfB(A)S(A)
H(A) A
1(3.3.8) are carried out over the visible spectrum
within fixed boundary wavelengths, that, in the
extreme, are given by A, = 360 nm to A, = 830 Z= k [BOA)S(A)Z(A) AA
nm. In practice, the integrations are replaced by
summations which are defined later in this sec-
tion. In the CIE 1964 System

(ii) Special Equations.

(a) Application to Photometry. If P, in Eq. a

1(3.3.8) represents the spectral concentration of
radiance, that is P, = L,,, measured in terms of
radiant power per unit solid angle per unit area,
for example, (W-sr_'- m7), the Y-tristimulus
value in the CIE 1931 system becomes the lumi-
nance L, (or simply L) of the color stimulus,
measured in terms of luminous flux per unit solid
angle per unit area, for example, (Im- sr! -
m ”), provided that the constant factor k is set
equal to K,,= 683 Im-W', the maximum
luminous efficacy of radiant power (see Chapter 4
on photometry). Thus,
Y=L= Ky fLexv(A)an [2(3.3.8)]
where, by definition, y(A) is identical to V(A), the
luminous efficiency function for photopic vision.
An equivalent equation for the CIE 1964 sys-
tem is sometimes used, by defining y,.(A) =
V\9(A), a luminous efficiency function for large-
field photopic vision. However, such an equation
has not been officially agreed upon.
(b) Application to Object-Color Stimuli. In Sec-
tion 3.3.4, object-color stimuli were defined by
P, = B(A)S)
where B(A) is the spectral reflectance of the il-
luminated and viewed object, and S, is the spec-
tral concentration of the radiant power of the
source illuminating the object. In most practical
situations, only relative spectral radiant power
distributions of the given light sources are re-
quired. The use of relative values of S, is indi-
cated by using the symbol S(A).
erepccad am ipBOA)S(A)X1o(A)dA
Syseeel thBOA)S(A)Fio( AAA
Zia ifBA)S(A)Z
(A) AA
[3(3.3.8)]
where the factors k and kj, are usually defined as
follows:
100 100
k > ke, = ——
ifS(A) F(A) dd thS(A)Fio(A) dA
This definition of k and k, )makes the Y and
Y,o-tristimulus values of the object-color stimulus
involving the perfect reflecting diffuser [B(A) = 1.0
at all wavelengths A] equal to 100 exactly.
The Y-tristimulus value of a given object-color
stimulus, calculated by means of Eq. 3(3.3.8),
defines the /uminance factor of the object-color
stimulus. The Yj9-tristimulus value has no such
photometric significance.
The same equations apply when the object is
transmitting instead of reflecting radiant power.
The spectral transmittance 7(A) of the object is
then used in Eq. 3(3.3.8) instead of the reflectance
factor B(A). The perfect reflecting diffuser is re-
placed by the perfect transmitting diffuser with
7(A) = 1.0 for all wavelengths A. Of particular
interest are objects that transmit in the regular
sense, that is, do not diffuse any part of the
radiant power. For such objects, typical examples
158 COLORIMETRY
are the optical filters described in Section 1.3. The
spectral transmittance 7(A) is simply the ratio of
the regularly transmitted radiant power to the
incident radiant power; no recourse to the perfect
transmitting diffuser is required here.
Equations 3(3.3.8) also apply to luminescent
objects, of which fluorescent objects are of partic-
ular interest in colorimetry. The spectral reflec-
tance factor B(A) in Eq. 3(3.3.8) is then replaced
by the total spectral radiance factor B;(A, 0) of
the fluorescent material irradiated by a source o
of known spectral radiant power distribution
{S,(A)dA}. In Section 3.12.3 two methods are
described in detail that can be used to obtain
B;(A, 0). The first method requires the given
source o to be physically available in the spectro-
photometric analysis as shown in Figure 2(3.12.3).
The second method, schematically illustrated in
Figure 4(3.12.3), provides a complete spectropho-
tometric record in matrix form from which
B,(A, o) can be obtained readily by computation
for any given source o of known spectral radiant
power distribution.
(iii) Integration Replaced by Summation. In
virtually all practical calculations of tristimulus
values defined by Eq. 1(3.3.8) to 3(3.3.8), the
integrals are replaced by sums. The argument that
justifies this approximation is straightforward and
drawn from basic principles of elementary calcu-
lus. A brief discussion of the principles involved
has already been given in Section 3.2.3. However,
some additional points must be raised which,
though not of any conceptual consequence, are
important when calculating tristimulus values by
means of summations.
The method most commonly used is the
weighted-ordinate method. The following equa-
tions apply
In the CIE 1931 System
Ap
X—k >, -BIAISAILIAIOA
A=A,,
rp,
Y=k D BA)S(A)A)AA
NEN:
A»,
Z=k > BA)S(A)Z(A)AA
A=Ay
In the CIE 1964 System
Ap
Xo = kyo 2) B(A)S(A)X1o(A)AA
A=
a
Xp
Yio = kio p B(A)S(A)
Yio(A)AA
A=A
a
Ap
Zio = kio y B(A)S(A)Z,0(A)AA
A=A,
[4(3.3.8)]
The wavelength intervals into which the visible
spectrum from A, to A, is divided are of equal
width AA centered at wavelength A. The relative
concentration of radiant power S(A) at wave-
length A is assumed constant for all wavelengths
within the wavelength interval of width AA;
fay S(A) dA = S(A)AA. The reflectance factor
B(A) at wavelength A is measured by standard
spectrophotometric techniques. The products
B(A)S(A)AX specified at the discrete but uni-
formly spaced wavelengths A are then used as
weighting factors of the color-matching functions
X(A), V(A), Z(A), and X1o(A), Vio(A), Z10(A)
which have been defined in tabular form at all
wavelengths within the visible spectrum (see Ta-
bles 1(3.3.1), T1(3.3.1), 1(3.3.2), and II(3.3.2)).
The visible spectrum over which the summa-
tions in Eq. 4(3.3.8) are made is, in the extreme,
bounded by the wavelengths A, = 360 nm and
A, = 830 nm. The finest recommended subdivi-
sion of the visible spectrum is in wavelength
intervals of width AA = 1 nm. However, in most
practical colorimetric applications, wavelength in-
tervals of width AA = 5 nm or even AA = 10 nm
give satisfactory approximations to calculations
made at l-nm intervals. Also, the visible spectrum
is usually taken as less broad and bounded at
A,
= 380 nm and A, = 780 nm. |
The application of Eq. 4(3.3.8) to two practical
cases is demonstrated in Tables 1(3.3.8) and
2(3.3.8). In the first example, Table 1(3.3.8), an
object-color stimulus is selected which involves an
opaque plastic sample with given spectral reflec-
tance factors B(A) and daylight as the illuminant
with a relative spectral radiant power distribution
identical to that of CIE standard illuminant D,..
The color-matching functions are those of the
CIE 1964 supplementary standard observer. The

spectral data are given from A, = 380 nm to
A, = 780 nm at 10-nm intervals. The values for
S(A) and X1(A), Vio(A), Zip(A) are taken from
Tables I1(3.3.4) and I1(3.3.2), respectively, for the
required wavelengths A. The first phase of the
calculations consists of taking at each wavelength
d the products B(A)S(A)Xio(A), BOA)S(A) Prof),
and B(A)S(A)Z;o(A). In the second phase of the
calculations the sums (2,) of the three columns of
products are formed and entered into the equa-
tions for k,9, X19, Yio, Zio Shown at the bottom
of the table. The (x,9, yjo)-chromaticity coordi-
nates are then readily derived from the calculated
tristimulus values.
In the second example of calculating tristimu-
lus values, given in Table 2(3.3.8), a self-luminous
color stimulus is selected that is produced by a
fluorescent lamp. The light emitted by the fluo-
rescent lamp has been measured by a spectroradi-
ometer providing values S(A) of relative radiant
power at wavelengths A in intervals of AA = 10
nm. The relative spectral radiant power distribu-
tion of the fluorescent lamp light is generally
smooth, except at the four wavelengths at which
the mercury emission occurs, that is, at 404.7,
435.8, 546.1, and 577.8 nm. At these wavelengths,
the spectral radiant power distribution has pro-
nounced “spikes,” the mercury emission lines.
The radiant power measured at each of these
wavelengths is composed of two parts: the radiant
power of the continuum originating with the ex-
cited phosphors in the lamp, and the radiant
power of the emission line. Again a wavelength
interval of width AA = 10 is used over which the
measured radiant power is uniformly spread.
For the purpose of calculating the tristimulus
values, the measured values of S(A) are given for
the continuum from A, = 380 nm to A, = 780
nm at 10-nm intervals, followed by the four
values of s(A) of the mercury emission lines,
which are in excess of the values of the continuum
at the four specified wavelengths. The actual
calculation then proceeds in a similar manner as
in the first example given in Table 1(3.3.8). How-
ever, this time the color-matching functions
x(A), y(A), Z(A) of the CIE 1931 standard ob-
server have been used. The values of x(A),
y(A), Z(A) needed at the four wavelengths of the
mercury emission lines have been found by linear
interpolation from Table [(3.3.1).
Reverting briefly to the calculation of tn-
stimulus values of object-color stimuli involving
one of the CIE standard illuminants defined in
The CIE Colorimetric System 159
Section 3.3.4, we note that such a calculation
can be simplified somewhat by making use of
special tables in which the color-matching func-
tions have already been weighted by the relative
spectral power distributions of these illuminants.
Moreover, the resultant products S(A)x(A),
S(A)y(A), S(A)Z(A) have been normalized so that
the sum %,S(A)y(A), required to determine k
becomes equal to 100. A similar normalization is
applied to the products S(A)X,9(A),
S(A)Vio(A), S(A)Z10(A) so that 2, S(A)V1o(A), re-
quired to determine k,), becomes 100. Tables
I(3.3.8) and II(3.3.8) are such special tables in
which the color-matching functions
x(A), y(A), Z(A) are weighted by the relative
spectral radiant power distributions of the CIE
standard illuminants A,B,C, and D,;,D¢5,D7s.
The table entries are given at 10-nm intervals.
Tables III(3.3.8) and IV(3.3.8) are similar tables
but involve the color-matching functions
X19(A), Vio(A), Z10(A)-
An alternative method to the weighted-ordinate
method of calculating tristimulus value is the
selected-ordinate method. In the selected-ordinate
method, which is not used much anymore, the
spectrum is again broken up into a finite number
of wavelength intervals. However, instead of in-
tervals uniform in width, intervals of the spec-
trum are selected that vary in width. The width
AX of each interval of the spectrum depends on
the relative spectral radiant power distribution of
the illuminant and the color-matching functions
of the observer. In particular, AA is chosen such
that S(A)x(A)AA, for the median wavelength A in
each interval AX, has a constant value throughout
the spectrum. Another set of intervals AA is cho-
sen to give constant values for S(A) y(A)AA, and
still another for S(A)Z(A)AA.
The tristimulus values X,Y, Z of an object-
color stimulus are then obtained simply by
adding the spectral reflectance factors B(A) at the
median wavelength A of each of the appropriate
wavelength intervals. The three sums obtained
this way are multiplied by factors proportional to
the constant values of S(A)xX(A)AA, S(A)y
(A)AA, S(A)Z(A)AA, respectively, and whose ab-
solute values are such that the factor of the sum Y
equals 100 times the reciprocal of the number of
intervals.
Table 3(3.3.8) gives the three sets of 30 selected
ordinates (median wavelengths in nanometers) for
computing the CIE 1931 tristimulus values X, Y, Z
of objects illuminated by CIE standard illuminant
160
FIGeL =
(g'¢'E)T jo
ajdurexq ALD Pog]
suyensyey puv
sanfeA
SN|NWYSUY Jo Uv
AWOHVUIOIYD SoJVUIPIOOD SN|NUINS
10[0D-elqO
INSeIq) y9efqoO pajyeurumnyy
Aq QysyAeg

piepuris
-JARM jueUIWNy|I I9A19SgO piepueys ‘[ddns po6] AIO sjonpold
y)3u2] 199fGO °q
(wiu)Y (vd (WS (y)otx (y)ora —
(x)Z SWI
(W™MVSOOS (WAMU SW (WAV (WAYS

O8t COTO 00S 70000 0000°0 L000°0 0°0 00 00 00

06£ 0 Src 9'PS ¥700°0 £000°0* 0 SOl0 00 00 10 0°0

OOV 0 Src 8°C8 16100 0700°0 0980°0 9°0 10 C20 70

Ov 0 6Ir 16¢ Lv80'°0 8800°0 0 v6st es £0 vl 6 80
Or 0 09r V'£6 0 SPOc v1c0'0 0 ScL6 8°8 60 S17 0?
OLY LLr'0 L98 0 Lele 0 L8t0 cess Oe! 9'l v9 c ve
OY CLv'0 vO!6 Lese'0 1790°0 £L96'1 16! a 0°86 ¢‘9

OSP OLY'0 O'LIT 0 LOLt ¢680°0 I 8766 v'0¢ 61 L601 COl

09P c9r'0 SLIT 0 €cO0L 0 C8C1 vSPrl'l Ol¢ OL 0°S6 Isl
OLY 0 csr VII6 0 9S61 cS81°0 0 9LIt Olc L6 6°89 Ele
O8Y psr'0 6
SII ¢080°0 0 967 6 ree (64 etl 9'0P V'6C
067 0 6S¢ 80l8 c910°0 0 16t¢ 0 estv 8°0 91 6 L'02 69£

00S 0 8Lv 601 8£00'0 80970 0 S8Ic7 70 £6 1 vlt OS bv

c9
y'S9
Bee
Ors
OcII'O
Vc
LIS‘0
LOl
8°
¢Lt00
L909°0
y9°'0
LLIIT
8°6L
ES
9°¢
OcS
81920
L090°0
OL
8°r0l
0
T8s¢
€v6
O¢s
0°?
919°0
0LLOl
$9€7
cSL8°0
S0£0'0
Lt
Ors £99°0 v0!vy 0 89L¢ 0 0796 Le10°0 1°92 9°99 60 00!vy

OSs 0 69 0'r0l 0 86cS 81660 0v00'0 18 VIL £0 ¢cOlI

096 169°0 0001 CSOL'0 £L66'0 0000°0 L’8P 6°89 0°0 L’66
OLS 089°0 £96 L8L8'0 0 9666 0000°0 CLS 9°C9 0°0 0°C6
O8S £69°0 8°S6 | cvIO 0 6898 0000°0 £149 Lal 00 £8Cc
06$ 0 IPL L’88 S8II'T vLLL'0 0000°0 Saec Ls 00 0°69
 009 062°0 0°06 Orcl I €869°0) 0000°0 6 6L 8°9OV 0°0 c6S
019 9780 9°68 SOE0'I 08750 0000'0 €'9L 1'6€ 0°0 €Lp
079 Cr8'0 L'L8 €958°0 186£°0 0000°0 S€9 4 0°0 6'bE
0€9 Z58°0 €€8 ¢Lr9°0 SE87°0 0000'0 0°9F 102 070 9°€Z
0r9 €58'0 L'€8 91€b'0 86L1°0 0000'0 Z'0€ 97 0°0 0ST
0s9 €58'0 0°08 €897°0 9L01°0 0000°0 ERI EL 00 9°8
099 €68'0 Z'78 97S1'0 €090°0 0000°0 L‘0l Zp 00 0's
0L9 €68°0 €°78 €180°0 81£0°0 0000°0 Ls OG 0°0 97
ZI
089
0°0
€68'0
Lt
€8L
6060'0
fal
6S10°0
0000°0
069 €68'0 L'69 6610'0 LLO0'O 0000°0 ZI $0 0°0 $0
00L 758'0 QL 9600'0 L€00°0 0000°0 90 Z'0 0°0 €'0
10
OIL
1S8°0
070
€pL
9600°0
8100°0
00000
€0
OZL
6r8°0
0°0
9°19
7Z00°0
8000°0
00000
[0
O€L 8780 6°69 0100°0 7000'0 0000'0 10 0°0 00 070
OPL
06L'0
0°0
ISL
5000°0
Z000°0
0000°0
070
OSL 0SL'0 9°€9 €000°0 1000°0 0000°0 00 0°0 0°0 0°0
092 Z1L'0 y9P 1000°0 0000°0 0000°0 00 00 00 0°0
OLL 089°0 8°99 1000°0 0000°0 0000°0 0°0 00 0°0 070
O8Z 679'0 r'€9 0000'0 0000°0 0000°0 00 0°0 0°0 0°0
—X% 9°81 8°09L 8085 LI9II

BUIZI[eUION
Joey %y = Z/O0L (v)""(Y)S
= 1980°0 FIO HOGI Kx) -01Y9-(9T
AVONeW So}eUIPIOOS
Oly = REY OY ae| 2H ue (oUF — ZI9E'0
KX= CYK SWE
MV (VW)= €°S9
¥ Org= ( Yaar
a LD =" DIF Je (oUF as €Z79€'0
FIO 961 SN|NWNSL
Son[BA 1g = xy SWE (V)"A(Y)
= 6°59
%Z= XY ("ASCOT
= 00S

161
 JO[OD (W)2W)S
10 9°0 €7 0°6 CSE8°€6Civ 6991 ler6°76Les IIe9°LI'8 bY Lee 6'0 b'0 70 70
9°PLI

snourumy-jjag

(WAY)S
sjonpold 0°0 0°0 0°0 10 Z'0 8°0 61 9°€ €9 1016'°SI0'rZ 6 0°9¢p9OL€°687°96 €'665°66796€°68
'9€

&
JO
sajyeuIps00D
(v) 0°0 10 50 6'1 ens761787 jaa pOLA601Le 9°0 01 8°9 et €°67 eer56SOLL16
(Vv)
S

AJOQRWIOIND

10Z0°06L90'°0
$900'0 9¢6¢9°0
pL07'0 9¢8e'l 9L87'1
7699"
ILpL IZLL'I 7S9P'007L7'0
0€18°0 78L0'0
Z8S1°0 7ZP0'0 6£00°0
L800°0
€020'°0 L100°0
1Z00'0
(x)2
I9AIBSGO
| |
pu

sonfeA plepueIs
0000°0
1000°0¢000°0 0+600'0
Z100°0 0€70°008£0°0
9110°0 0090°0
0160°0 0807'0O€ZE'0
06€1°0 07980
OOIL'0
0€0S'0 0rS6'005660
05660 00L8°0
07S6'0
(x)
snMMWISLIY
[E61

AIO
r100'0
Zr00'0€r10'0 prel6£87'0
S€r0'0 E8re'0Z9EE80670
rS61'0 0Z£0'06¢00°0
9660°0 70670peer0
SS9l'0
€£90'°0
€600°0 €916'0
Sr6S1Z9L'0
(x)x
oy)
0 ‘0 0
SuyR~NseD

jUSDSOION[

jo (WS
dw] 0°€2OLE pee9°€bossL'L9018 769°r01
LAUevcit 7 ruil 0101 8°660°001
9°€01
9°L0I 1101L201
Pll
sduexy
YY

(g'¢E'¢)Z YISUITOAC
(wu)y
O8€ 06€ 00rOIPr0? OEPOrr OSP 09P OLYPO8r 06 00S O1S 07S OES Ors Oss 09¢ OLS 08S
10
06
L701
ELL
€970'1
OLSL'0
1100°0
ps0l8000°0
10
009
101
c6c790
€9
LOI
¢
I01¢9°0
JIQUI, AA

snyNUyS
Jusdsai0ny,yq)
(duie’]

162
 9SOI=(V)Z(V)SZY=Z

163
OO0OI=(V)A(VISZY=A
=(V)4(Y)SZ/001

(VIX(V)SK

eZECO=(Z
+A+X)/A=4

snynwnsiy

BUIZTTPULION
LOIEO=(Z+A4X)/X AjTOTyeUTOIYS SayeUIpIOOD =X AIO [€61

sanyea
E6G=—

[€6]

AIO
= Yea
61800

JOJOVJ
¥
SUT] UOTSSTUID AINDIOUI JO YSUITIAR MA ,
0°0 6°S7 8ST L100°0 0168°0 67880 162 rg*ZLs
cu C 66 6 Lt c710'0 brs6 0 SSLe 0 8 001 a) 29
6 00€ (ons 8°09 L6v9'T 8L10°0 ceee 0 b°78I v8 Str
€'8 00 Per L901 °0 9600°0 47700 LLL oL vOP
00 0°0 0°0 0000°0 0000°0 0000°0 0°9 082
0°0 00 0°0 0000°0 0000°0 1000°0 LL OLL
0°0 0°0 00 0000°0 1000°0 70000 0°6 09L
0°0 0:0 0°0 0000°0 1000°0 £000°0 OTT OSL
00 0:0 0°0 0000°0 c000°0 L000°0 Stl OL
0°0 0°0 0°0 0000°0 $000°0 v1000 te OI Otel
0°0 00 10 0000°0 0100°0 6c00°0 $02 OcL
00 10 10 0000°0 1c00°0 8$00°0 > Ste OIL
00 10 v0 0000°0 Tv00°0 bIl00 Ble 00L
0°0 £0 60 0000°0 €800 0 Lcc0'0 T8e 069
0°0 8°0 CL 0000°0 0L10°0 89r0°0 CLV 089
0°0 0? 9°¢ 0000°0 0ct0'0 ¢L80°0 989 0L9
00 6S 8 SI 0000°0 0190°0 6r91 0 0°96 099
0:0 € Ol Pic 0000°0 OLOI 0 St87 0 $°96 0S9
0°0 Cc9l Sv 0000°0 OSZI'0 6LPy 0 LC Ov9
0°0 8°S7 9.29 0000°0 0$97°0 vcv9'0 b'L6 0¢9
0°0 Cle S'8 7000°0 O18¢ 0 vrss'0 6 86 079
00 00S 8°66 €000°0 0£0S 0 9200'T ¢°66 O19
164 COLORIMETRY

Table 3(3.3.8) Selected Ordinates for Computing The spectral reflectance-factor curve B(A) [or
CIE 1931 (X, Y, Z )-Tristimulus Values of spectral transmittance curve 7(A)] of a given ob-
Object-Color Stimuli Involving CIE ject is sampled differently for the calculation of
Standard Illuminant A each tristimulus value. The curve is sampled often
where S(A)x(A), S(A)y(A), and S(A)z(A) have
Illuminant A maximal values, but is sparsely sampled near the
Wavelengths (nm) to compute ends of the spectrum.
tristimulus values Further tables of selected ordinates for various
Ordinate
number (X) (Y) (Z)
illuminants and the CIE 1931 standard observer
may be found in The Science of Color (Committee
1 444.0 487.8 416.4 on Colorimetry, Optical Society of America, 1963)
29 516.9 507.7 424.9
and in Hardy’s Handbook of Colorimetry (1936).
3 544.0 517.3 429.4
The latter publication also contains tables for 100
4 554.2 524.1 432.9 selected ordinates. Sugiyama (1969) added to the
52 561.4 529.8 436.0
6 567.1 534.8 438.7
collection a number of tables of 90 and 30 selected
ordinates for the CIE standard daylight illumi-
7 572.0 539.4 441.3
82 576.3 543.7 443.7
nants D;;,D.;,D7;, and the CIE 1931 standard
9 580.2 547.8 446.0 observer, as well as similar tables for CIE stan-
10 583.9 S57 448.3
dard illuminants A,B,C,D<;,D,;,D,7;, and the
11¢ 587.2 555.4 450.5 CIE 1964 supplementary standard observer.
12 590.5 559.1 452.6 The weighted- as well as the selected-ordinate
13 593.5 562.7 454.7 method to calculate tristimulus values are ap-
142 596.5 566.3 456.8 proximate methods only. In accordance with
15 599.4 569.8 458.8 studies concerned with the question of numerical
16 602.3 573.3 460.8 accuracy (see, e.g., De Kerf, 1958) the weighted-
1724 605.2 576.9 462.9 ordinate method using AA = 5 nm is accurate for
18 608.0 580.5 464.9
most practical cases. The use of AA = 10 nm also
19 610.9 584.1 467.0 provides accurate results for many practical cases
202 613.8 587.9 469.2
21 616.9 591.8 471.6
except those special cases where the spectral re-
flectance factors B(A), the spectral transmittances
2 620.0 595.9 474.1
232 623.3 600.1 476.8
T(A), or the spectral radiant power distribution
24 626.9 604.7 479.9 {S(A)dA} are highly irregular (Ohta and
Wyszecki, 1975b). The selected-ordinate method
25 630.8 609.7 483.4
262 635.3 615.2 487.5 is usually inferior to the weighted-ordinate method
27, 640.5 621.5 492.7 unless many ordinates (i.e., 100 or more) are used.
28 646.9 629.2 499.3 Digital computing devices, often coupled with the
292 655.9 639.7 508.4 readout of spectrophotometers, almost exclusively
30 673.5 659.0 526.7 make use of the weighted-ordinate method. In
virtually all modern colorimetry, the selected-
Multiplying
factors: ordinate method is no longer employed.
30 ordinates 3.661 3.333 1.185 Graphical methods for the determination of
10 ordinates 10.984 10.000 3.555 tristimulus values have rather limited accuracy
2 For abridged method use only these ordinates. and precision and are not used in modern colori-
Note. The multiplying factors are chosen to make the Y metric practice. An example of a graphical method
value of the perfect reflecting diffuser or the perfect is that given by Bodmann and Jantzen (1961).
transmitting diffuser or filter equal to 100.000.

3.3.9 CIE Uniform Color Spaces and

A. By taking the second, fifth, eighth, and so on,
ordinate, indicated in Table 3(3.3.8) by asterisks,
the 30-selected-ordinate method can be abridged
to a 10-selected-ordinate method. The corre-
sponding multiplying factors are given at the bot-
tom of the table.
Color-Difference Formulae
The CIE colorimetric system includes computa-
tional methods designed to aid in the prediction
of the magnitude of the perceived color difference
between two given object-color stimuli that are
found to mismatch in color. The development of
 The CIE Colorimetric System 165

a satisfactory color-difference formula has been Supplementary Standard Colorimetric Observer

recognized as an urgent industrial problem [for a and Coordinate System; in the latter case, the
review, see Judd and Wyszecki (1975)]. However, appropriate tristimulus values and chromaticity
none of the many color-difference formulae that coordinates must be used.
have been proposed in the literature over the past
several decades is considered a sufficiently ade- (ij) CIE 1976 (L*u*v*)-Space and Color-
quate solution of the problem. Difference Formula. The first approximately
The determination of a quantity that suitably uniform color space, illustrated in Figure 1(3.3.9),
describes the color difference an observer may is produced by plotting in rectangular coordinates
perceive between two given color stimuli rests on the quantities L*, u*, v* defined by:
the ability of the observer to judge the relative
magnitude of two color differences possibly per-
ceivable when viewing two pairs of stimuli. The 1/3
observer’s judgment varies greatly with the condi- Lx = 116{ +| a ts
tions of observation and the kind of stimuli pre-
sented. Sizes, shapes, luminances, and relative ior (an — 1°.)
spectral radiant power distributions of the test
stimuli and the stimuli surrounding them and oe 13L* (0 = 0,) [1(3.3.9)]
making up the display situation, are important
factors affecting the observer’s judgment. There
with the constraint that Y/Y, > 0.01. If values of
are a great deal of experimental data available on
Y/Y,, less than 0.01 occur, a somewhat modified
color discrimination and uniform color scaling,
procedure is recommended for calculating L”*.
that help to appreciate the complexity of the
For values of Y/Y,, equal to or less than 0.008856,
problem, and that have led many investigators to
the following L* formula is used:
propose empirical formulae designed to predict
observed color differences under certain experi-
mental conditions. Attempts have also been made mn Ue 26 for es< 0.008856
to develop appropriate mathematical models of Y,n

the visual processes that govern color discrimina-

tion, and then predict, by means of such models, [2(3.3.9)]
a variety of experimental data. A model of this
kind takes on the form of a line element, a mea-
In Eq. 1(3.3.9), the quantities u’, v’ and u,, 0,
sure of distance in a postulated space in which
are calculated from:
perceived colors are represented by points or vec-
tors. = 4X Pad WyecOWY wh

Both empirical color difference formulae and ot ane Pia

Lh Z X= bY 467
line elements of color space will be discussed in
more detail in Sections 6.5.2 and 8.4 where the 4X, OY,
fa ea a One
relevant experimental material is collected and ArgiXe lS 3Zeewo Xe lS Ye 8Z;
related topics are discussed.
In this section, we confine our presentation to
the current recommendations made by the CIE
[3(3.3.9)]
with regard to the evaluation of color differences
as required in colorimetric practice (CIE, 1978a). The tristimulus values X,, Y,, Z,, are those of
The CIE recommends the use of two ap- the nominally white object-color stimulus. Usu-
proximately uniform color spaces and associated ally, the white object-color stimulus is given by
color-difference formulae that were chosen from the spectral radiant power of one of the CIE
amongst several of similar merit to promote uni- standard illuminants, for example, D,; or A,
formity of practice, pending the development of a reflected into the observer’s eye by the perfect
space and formula giving substantially better cor- reflecting diffuser. Under these conditions,
relation with visual judgments. The following rec- X,,, Y,, Z, are the tristimulus values of the stan-
ommendations are all given in terms of the CIE dard illuminant with Y,, equal to 100.
1931 Standard Colorimetric Observer and Coor- The total color difference AE*, between two
dinate System, but also apply to the CIE 1964 color stimuli, each given in terms of L*, u*, v*, is
 166 COLORIMETRY

Fig. 1(3.3.9). Sketch of CIE 1976 (L*u*v*) color space with outer boundary generated by
optimal color stimuli with respect to CIE standard illuminant D,; and the CIE 1964 supplemen-
tary standard observer. The colors of all object-color stimuli fall within this boundary. This is
also the gamut within which the CIE 1976 color-difference formula A E(L*u*v*) is intended to
be valid. Note that the spectrum locus of monochromatic stimuli is generally well outside the
boundary of object-color stimuli (from Judd and Wyszecki, 1975).

calculated from: tive transformation of the CIE 1931 (x, y)-chro-

maticity diagram. Straight lines in the (x, y)-
AE*, =[(AL*)’ + (Au*)’ + (Av*)’]'” chromaticity diagram remain straight in the
[4(3.3.9)] (u’, v’)-diagram. This feature is considered to be
of importance in cases where color stimuli are
The values of the coefficients of the equations
mixed additively. The chromaticity diagram de-
defining the CIE 1976 (L*u*v*)-space and its
fined by Eq. 3(3.3.9) is called the CIE 1976 UCS
associated color-difference formula are intended
diagram.
to apply to the observing conditions normally
If v* is plotted against u*, the points in the
found in practice. However, there is evidence
resulting (u*, v*)-diagram are not uniquely re- |
that, in certain situations, different values for the
lated to chromaticity because their position de-
coefficients may be more appropriate. In particu-
pends on the value of L*. However, if L* is
lar, the weight of the perceived lightness dif-
constant, straight lines in the CIE 1931 (x, y)-
ference between two given color stimuli relative to
chromaticity diagram or in the CIE 1976 UCS
the perceived chromatic difference between the
diagram remain straight in the (u*, v*)-diagram.
same two stimuli may, under certain observing
The CIE 1976 (L*u*v*)-space and its associ-
conditions, require different coefficients.
ated color-difference formula are intended to be
The color space defined by Eq. 1(3.3.9) is
improvements over and supersede the original
called the CIE 1976 (L*u*v*)-space. The color-
CIE 1964 (U*V*W*)-space and its associated
difference formula, defined by Eq. 4(3.3.9), is
color-difference formula (see Section 6.5.2).
called the CIE 1976 (L*u*v*) color-difference
formula. The letters CIELUV are used as an
abbreviation. (ii) CIE 1976 (L*a*b*)-Space and Color-
As indicated by Eq. 3(3.3.9), the CIE 1976 Difference Formula. The second approximately
(L*u*v*)-space incorporates, for constant L*, a uniform color space, illustrated in Figure 2(3.3.9),
(u’, v’)-chromaticity diagram which is a projec- is produced by plotting in rectangular coordinates
 The CIE Colorimetric System 167

Fig. 2(3.3.9). Sketch of CIE 1976 (L*a*b*) color space with outer boundary generated by
optimal color stimuli with respect to CIE standard illuminant D,, and the CIE 1964 supplemen-
tary standard observer. The colors of all object-color stimuli fall within this boundary. This is
also the gamut within which the CIE 1976 color-difference formula AE(L*a*b*) is intended to
be valid. Note that the spectrum locus of the monochromatic stimuli is generally well outside
the boundary of object-color stimuli (from Judd and Wyszecki, 1975).

the quantities L*, a*, b* defined by:

|
y\'2 1/3
ftwe 116bray
Y = 16 eS|x|
~~ 9.008856 xX,n
n

| x
x,
=e 0.008856

(x) -(Z)
y\'2 7\'3
b* = 200 [5(3.3.9)]
ta Caled aise
with the constraint that X/X,,Y/Y,, Z/Z, >
0.01.
In calculating L*, a*,
X/X,,Y/Y,, Z/Z, less than 0.01 may be in-
and b*, values of | a ~<|~
< 0.008856

|= |2)
1/3
cluded if the normal formulae are replaced by the
Z ~ 9.008856
following modified formulae: N
Y
Lx* —=903.35
|
bein for = < 0.008856 Z
n Z,
— =
0.00856
and

[6(3.3.9)}
The tristimulus values X,,, Y,,, Z,, are those of
the nominally white object-color stimulus. Usu-
ally, the white object-color stimulus is given by
168 COLORIMETRY

the spectral radiant power of one of the CIE
standard illuminants, for example, D,; or A,
reflected into the observer’s eye by the perfect
reflecting diffuser. Under these conditions,
X,,,Y,5Z, are the tristimulus values of the stan-
dard illuminant with Y, equal to 100.
The total color difference AE*, between two
color stimuli, each given in terms of L*, a*, b*, is
calculated from:
AE*, =[(AL*)’ + (Aa*)’ + (Abt) |”
to identify the components of color differences in
terms of correlates of the perceived lightness,
chroma, and hue. It is also often desirable to
express color specifications in terms of such cor-
relates. Either one of the two CIE 1976 uniform
color spaces and associated color-difference for-
mulae can be used to define appropriate cor-
relates.
The quantity L*, given in Eq. 1(3.3.9) and
5(3.3.9) serves as the correlate of lightness.
The quantities C*, and C¥, defined by

[7(3.3.9)]
The values of the coefficients of the equations
defining the CIE 1976 (L*a*b*)-space and its
associated color-difference formula are intended
to apply to the observing conditions normally
found in practice. However, there is evidence
that, in certain situations, different values for the
coefficients may be more appropriate. In particu-
lar, the weight of the perceived lightness dif-
ference between two given color stimuli relative to
the perceived chromatic difference between the
same two stimuli may, under certain observing
conditions, require different coefficients.
The color space defined by Eg. 5(3.3.9) is
called the CIE 1976 (L*a*b*)-space. The color-
difference formula, defined by Eq. 7(3.3.9), is
called the CIE 1976 (L*a*b*) color-difference for-
mula. The letters CIELAB are used as an abbrevi-
ation.
The L* function defining the lightness corre-
late in the CIE 1976 (L*a*b*)-space is identical
to the L* function of the CIE 1976 (L*u*v*)-
space. No simple relation exists between the scales
a*, b* of the CIE 1976 (L*a*b*)-space and u*, v*
of the CIE 1976 (L*u*v*)-space. If b* is plotted
against a*, the points in the resulting (a*, b*)-
ct,=[(uty
+ (vt) ]%,
Cs =[(a*’ + (o*)7] 7 [863.9)]
serve as correlates of chroma.
The quantity
ce
eee [9(3.3.9)]
derived from C*, and L* in the CIE 1976
(L*u*v*)-space can be used as a correlate for
saturation. In a series of object-color stimuli of
constant chromaticity, but increasing (or decreas-
ing) luminance factor, s*, remains constant with
corresponding increases (or decreases) in C*,. An
equivalent relation for s*, is the following:
sto= 13[(a!— ws) + (0"= wf")?
[10(3.3.9)]
A similar correlate of saturation cannot be given
for the CIE 1976 (L*a*b*)-space.
The quantities h,,, and h,,, defined by
* *

diagram are not uniquely related to chromaticity hiot arctan( *>), hap= arctan( 7 |
because their position depends on the value of
L*. If L* is constant, straight lines in the CIE [11(3.3.9)]
1931 (x, y)-chromaticity diagram or in the CIE
1976 UCS diagram become, in general, curved define hue angles which are useful quantities in
lines in the (a*, b*)-diagram. specifying hue numerically. The angles are given
The CIE 1976 (L*a*b*) color-difference for- in degrees using the following conventions:
mula is a cube-root version of the AE(AN40)
Uae alle aU Iw ee at eae
Adams-Nickerson color-difference formula (see
Section 6.5.2). Both formulae make similar pre- Oe = = LOU. 4 uf wu 0, ote
dictions, particularly when AE(AN4O0) is multi-
plied by the constant factor 1.10. [SOS Ue NM Tad
rae et
(iii) CIE 1976 Correlation of Perceived Light-
PMNpuis D thoes SoBlahg. TE ON
ness, Chroma, and Hue. When color-difference
formulae are used in practice, it is often desirable and similarly for h,, using a*, b*.

The quantities AH*, and A H*, defined by
AHS, =|(AE*,) — (AL*) — (act)|”
and
AH, =|(AE3,) — (AL*) — (Acz)']'”
[12(3.3.9)]
specify hue differences. A quantity of hue dif-
ference is useful to describe a total color dif-
ference AE* in terms of its three components:
RU AAC*, and WH?
For a small color difference between two color
stimuli both of which are sufficiently different in
color from the achromatic stimulus, the hue dif-
ference can be calculated from
* =—
CAuv h uv oem.
ae 180
C*Ah
A eae [13(3.3.9)]
The hue difference is given a positive sign when
the hue angle A increases and a negative sign
when h decreases.
3.3.10 CIE Metamerism Index for
Change in Illuminant
Object-color stimuli of different spectral radiant
power distributions {B“(A)S(A) dA} and
{B® (A)S(A) dA} are metameric stimuli with re-
spect to a given observer if they yield identical
tristimulus values (see Section 3.3.4). The given
observer judges the two stimuli to be in complete
color match. However, it is clear that the color
match cannot be expected to remain a color match
if the illuminant, defined by {S(A) dA}, is
changed to another one of different spectral
radiant power distribution. Similarly, changing
the observer to one characterized by a different
set of color-matching functions will, in general,
result in a color mismatch between the two given
stimuli. The magnitude of the color mismatch is
intuitively expected to be directly related to the
sizes of the spectral differences between the re-
flectance-factor curves {B‘?(A)} and {B(A)} of
the two given objects. The larger the spectral
differences, the larger the color mismatch when
either the illuminant or the observer is changed.
The CIE Colorimetric System 169
These qualitative observations have led to the
concept of the degree of metamerism of colored
objects; that is, objects of different spectral re-
flectance-factor curves colorimetrically evaluated
with respect to a given observer and illuminant.
The concept is of considerable practical impor-
tance, particularly to those industries in which
materials are colored to close specifications, such
as the textile and paint industries. Colorants are
formulated and mixed to reproduce the color of a
given standard material with the constraint that
the degree of metamerism between the reproduc-
tion and the standard is minimal (see Section 3.9).
A low degree of metamerism generally ensures
that the color match between reproduction and
standard will essentially be invariant with regard
to changes in the illuminant and/or observer.
The derivation of a suitable measure of the
degree of metamerism of colored objects is a
complex colorimetric problem. Considerations
relevant to this problem will be discussed in Sec-
tion 3.8 which deals with the subject matter of
metameric color stimuli in somewhat greater
depth. In the meantime, a recommendation made
by the CIE (1972) is given below that serves as a
practical method of quantifying the degree of
metamerism of any pair of objects with given
spectral radiance-factor curves.
The CIE metamerism index is referred to as a
special metamerism index because it is limited to
the case in which the illuminant is changed. The
observer is kept the same throughout the discus-
sion. The following definition is used:
The degree of metamerism M of two objects
{B\?(A)} and {B® (A)} that match in color un-
der a given illuminant {S(A) dA} for a given
observer [x(A), y(A), Z(A)], is measured in terms
of the color difference AE observed between the
two objects but under another given illuminant
{S’(A) dA}. The CIE procedure is then as fol-
lows:
For two objects whose corresponding tri-
stimulus values (X, = X,, Y; = Y;, Z, = Z,) are
identical with respect to a reference illuminant
and reference observer, the metamerism index,
M,, is set equal to the index of color difference
A E between the two objects computed for the test
illuminant f.
The preferred reference illuminant is CIE
standard illuminant D,;. The preferred test illu-
minant is CIE standard illuminant A or one of
the illuminants defined in Table 1(3.3.10). The
F-illuminants given in Table 1(3.3.10) represent
typical fluorescent lamps having reasonably high
170 COLORIMETRY

Table 1(3.3.10) Relative Spectral Radiant more appropriate to choose an illuminant differ-
Power Distributions of [lluminants Recom- ent from any of the three F-illuminants given in
mended for Calculating CIE Index of Metamer- Table 1(3.3.10). In some cases, it may also be
ism: change of Iuminant. Illuminant A is useful to determine the metamerism index with
Identical to CIE Standard Iluminant A and respect to several test illuminants. The specific
Illuminants F1, F2, F3 are Representative of test illuminant used must be identified as a sub-
Fluorescent Lamps (CIE, 1972) script to M, for example, M, or My.
The reference observer should be either the
A Fl 2 F3 CIE 1931 standard colorimetric observer or the
d(nm) S(A) S(A) S(A) S(A)
CIE 1964 supplementary standard colorimetric
380 9.8 5.4 10.7 23.0 observer, according to the field size involved, and
90 12.1 5.6 12.0 27.3
it should be stated which observer has been used.
400 14.7 5.8 1359, 33.4
The color difference A E is computed from the
10 17.7 6.1 16.8 43.6
20 21.0 7.4 20.8 5520 tristimulus values Xj,, Y1,, Z,, Of object-color
30 24.7 10.6 28.0 6757 stimulus 1 and X;,, Y;,, Z,, of object-color
40 28.7 ie? SiHlose. 81.0
stimulus 2 involving test illuminant ¢. At present,
450 331 26.5 48.8 94.2
60 37.8 33.6 58.5 104.6
the CIE recommends two color-difference for-
70 42.9 38.3 64.4 111.1 mulae as defined in Section 3.3.9, and either one
80 48.2 397 66.5 114.3 can be used for the computation of AE.
90 5-9 39.8 67.0 Li5;5
We have worked out examples to demonstrate
500 599 40.7 66.6 114.2
10 66.1 43.6 67.7 111.4
the procedure recommended by the CIE. Table
20 TPR) 49.9 69.9 107.6 2(3.3.10) and Figure 1(3.3.10) give the spectral
30
40
19
85.9
57.4
67.8
1B?
78.7
103.6
101.0
reflectance factors BA), BA), BO(A) of
three objects that, when illuminated by CIE
550 92.9 82.3 88.4 99.8
60 100.0 100.0 100.0 100.0 standard illuminant D,;, provide for the CIE
70 107.2 1132 110.4 101.1 1931 standard observer a set of three metameric
80 114.4 12537 116.0 102.7
90 121.7 iM Wee) Lis 102.7
object-color stimuli. Their CIE 1931 (X,Y, Z)-
tristimulus values are identical:
600 129.0 103.2 111.2 101.2
10 136.3 93.3 104.6 99.5
20 143.6 109.8 104.0 98.9
30
40
150.8
158.0
145.7
143.6
104.9
103.6
97.4
yaad
Xp =X Sage25 we = eee
650 165.0 DILL 116.9 96.5
60
70
172.0
178.8
296.5
86.9
147.7
62.3
96.0
63.6
Zire? een onl
80 185.4 35.6 40.5 47.2
90 191¢9 21.2 30.2 38.1

700 198.3 12.4 23.6 31.4 It follows that the color difference AE between
10 204.4 8.0 18.0 2553
any of the three pairs of stimuli (0, 1), (0, 2), (1, 2)
20 210.4 2 14.0 20.5
30 216.1 35 10.8 16.7 is Zero.
40 2217 2.0 933 BS When the illuminant is changed from D,; to
750 227.0 1.0 6.6 11.0 A, F,, F,, and F, [see Table 1(3.3.10)], the result-
60 232-4 0.2 ayy? 9.0
70 237.0 0.0 4.0 US:
ing object-color stimuli involving the three objects
80 241.7 0.0 3.1 6.0 given in Table 2(3.3.10) have different tristimulus
values and the color differences between the pairs
404.7° Dee 42.3 et |
435.8 84.0 2 182.4 are different from zero. Table 3(3.3.10) give the
546.1 Pied gcd, 100.8 numerical results in terms of chromaticity coordi-
577.8 23r7 23.0 Pao)
nates x, y, and luminance factor Y. The color
“ Power at principal mercury lines taken above continuum and differences AE were calculated by means of the
spread evenly over a 10-nm band.
CIE 1976 (L*u*v*) color-difference formula given
in Section 3.3.9. Only the differences for pairs
CIE general color-rendering indices and cor- (0, 1) and (0, 2) are recorded in Table 3(3.3.10).
related color temperatures of about 3000 K for In accordance with the CIE method of de-
F,, 4000 K for F,, 6500 K for F,. The most termining the metamerism index M,, outlined
appropriate choice of test illuminant depends on above, it follows from Table 3(3.3.10) that the
application, and in some instances, it may be object pairs (0,1) and (0,2) have the following
 The CIE Colorimetric System 171

Table 2(3.3.10) Spectral Reflectance Factors metamerism indices:

of Three Objects Which for the CIE 1931
Standard Observer Provide Metameric Stimuli
when Illuminated by CIE Standard Iluminant Pair (0, 1) Pair (0, 2)
Des Ma 2 11.2
A(nm)
My, 4.7 13.8
Me, 3.3 5.1
BOC) BO) B(A)

380 10.50 15.92 11.74

M- 2.1 0.2
12.00 14.85 13.44

13.61 9.80 15.48 These examples demonstrate that the degree of

14.27 5.90 16.14 metamerism can vary strongly with the test illu-
14.28 5.42 16.05 minant chosen to assess it. For both pairs the
14.09 6.82 15.60 largest index is obtained when the test illuminant
13.94 9.32 15.13
is F,, that is, the 3000 K fluorescent lamp; and
450 13.86 12.39 14.64 the smallest index is obtained when the test illu-
13.74 15.54 13.90 minant is F,, that is, the 6500 K fluorescent lamp.
13.68 19.07 13.03 With D,; being the reference illuminant, this re-
13.67 22.00 11.92 sult is to be expected for most practical situa-
13.60 23.01 10.37
tions.
500 13.56 21.86 8.78 It is also noted that pair (0,2) has a larger
13:77 19.26 7.84 index than pair (0,1) in three of the four test
14.17 15.79 7.84 cases. This result is difficult to deduce by mere
14.67 peo 8.73 inspection of Figure 1(3.3.10).
16.06 9.85 iBieee) If one considers object 0 to be the original and
pakote9: 13193 19.18 objects 1 and 2 two reproductions, each made to
27.78 24.47 Slei2 provide the same color as the original under il-
37.47 38.27 46.97 luminant D,,, the calculations indicate that object
48.48 52.58 62.26 1 would be the preferred reproduction because it
aE 61.00 70 98 has a lower degree of metamerism with the origi-
62.59 63.87 70.20 nal object 0 with respect to an incandescent lamp
65.68 65.64 63.10 and with respect to two of three fluorescent lamps.
67517 66.90 5595 However, even for object | the resulting degree of
68.18 68.27 51.09 metamerism is probably not negligible with re-
68.76 69.27 48.46 spect to fluorescent lamps of low correlated color
69.31 70.28 47.63
temperature. Metamerism indices greater than two
69.80 71.20 47.24 or three indicate that the mismatch between the
70.40 pee, 47.41 original and the reproduction may be objectiona-
cA tee q3237 47.59 bly large under those illuminants.
71.86 74.01 47.81 The question of how many different test illu-
minants ¢ one should use to arrive at an adequate
72.61 75.06 47.82
73.14 76.15 47.54
assessment of the degree of metamerism has been
73.69 77.43 47.41 studied by Brockes (1969, 1970). For many practi-
74.45 79.07 47.71 cal purposes it suffices to restrict the number of
74.62 80.09 47.13 test illuminants to two, one being illuminant A,
the other a warm white de luxe fluorescent lamp,
750 74.82 81.40 46.52
such as F,.
75.20 83.13 46.37
The CIE calls its index of metamerism M,
404.7 13:92 1-97 1534/9 “special metamerism index: change in illuminant.”
435.8 14.00 8.27 15.33 The index is special in the sense that the degree of
546.1 18.66 12.34 16.20 metamerism is evaluated in terms of a limited
377,8 46.06 49.43 58.90
number of specific illuminants. Other special in-
dices are conceivable (Allen, 1969; Wyszecki,
1969; Judd and Wyszecki, 1975). In fact, the CIE
 1.0

0.8

=) Oo

factor
Reflectance
oS pp

0.2

400 500 600 700

Wavelength (A), nm

Fig. 1(3.3.10). Spectral reflectance-factor curves of three objects that when illuminated by
CIE standard illuminant D,; provide metameric object-color stimuli with respect to the CIE
1931 standard observer. For numerical data, see Table 2(3.3.10).

Table 3(3.3.10) Chromaticity Coordinates x, y and Luminance Factor Y of Three Objects (0, 1,2)
Viewed by the CIE 1931 Standard Observer under Illuminants D,<, A, F1, F2, F3°

Object Ds A Fl F2 F3
Xo 0.4691 0.5680 0.5580 0.5188 0.4695
0 Yo 0.3643 0.3847 0.3877 0.3872 0.3680
Yy 33.0 40.25 39.73 38.86 33.07
x, 0.4691 0.5683 0.5631 0.5233 0.4743
I y 0.3643 0.3810 0.3847 0.3856 0.3699
Y, 33.0 40.23 39.36 36.55 32.76
x5 0.4691 0.5592 0.5458 0.5140 0.4691
2 yy 0.3643 0.3941 0.3991 0.3928 0.3678
Y; 33.0 40.36 40.34 37.88 33.27
A E(0,1) 0.0 2.7 4.7 3.3 2.1
A E(0,2) 0.0 11.2 13.8 5.1 0.2

“Also given are the color differences AE between objects 0 and 1 and between objects 0 and 2 under
each of the five illuminants. Under illuminant D,; the three objects are, by design, metameric
[A E(0, 1) = AE(O,2) = AE(1,2) = 0}.

172

colorimetry committee is considering the develop-
ment of a “special metamerism index: change in
observer.”
3.3.11 CIE Color-Rendering Index
Color rendering of a light source is the effect the
source has on the color appearance of objects in
comparison with their appearance under a refer-
ence source (CIE, 1974). In this context it is
appropriate to define a color-rendering index of a
light source as a measure of the degree to which
the perceived colors of objects illuminated by the
source conform to those of the same objects il-
luminanted by a standard source, for specified
conditions. The usual conditions are that the ob-
server shall have normal color vision and be
adapted to the environment illuminated by each
source in turn.
In order to derive a color-rendering index that
is in accord with the above definition, one must
know how to specify object-color perceptions and
differences between them, and one must agree
upon a standard source with which to compare
the given test source.
The problem of specifying object-color percep-
tions, that is, colors perceived as belonging to a
non-self-luminous body, has not yet been solved
for the general case in which the observer views a
complicated scene composed of a large variety of
objects. Various visual phenomena, such as
simultaneous contrast, successive contrast, color
constancy, memory color, size, and shape of the
objects, come into play and contribute signifi-
cantly to the resultant color perception of the
complicated scene; but the science of color has
not advanced far enough to deal with this prob-
lem quantitatively (Judd and Wyszecki, 1975).
However, a simplified version of the problem can
be dealt with by restricting the viewing conditions
to such scenes as those in which the adaptive state
of the observer’s eyes is governed predominantly
by the quality of the controlled illumination and
is only negligibly affected by objects in the field
of view. Adequate predictions of object-color per-
ceptions can then be made by using the CIE
colorimetric system and a model capable of pre-
dicting the effect of the adaptive state of the
observer’s eyes on the color perceptions under the
given simplified viewing conditions. A simple
model of chromatic adaptation that has been used
with some success is that of von Kries (1904) (see
Section 5.12).
The CIE Colorimetric System 173
The choice of a standard source with which to
compare the test source must be governed by
whatever is considered to be the original object-
color perception; that is, the perceived color of an
object in the light in which it is usually observed.
In most practical cases, this light will be in-
candescent lamp light or some phase of natural
daylight.
The degree to which the perceived colors of
objects illuminated by the test source conform to
those of the same objects illuminated by the
standard source is measured in the perceptually
uniform color space.
Crude approximations to the perceptually uni-
form color space are the CIE 1976 (L*v*v*)-space
and the CIE 1976 (L*a*b*)-space presented in
Section 3.3.9.
The CIE (1974) has recommended for measur-
ing and specifying the color rendering properties
of light sources, the CIE color-rendering index
which satisfies reasonably well the requirements
outlined above. The CIE color-rendering index is
based on a color-shift method, because it evaluates
for a number of test objects the average colori-
metric shift in a color space, that perceptually is
approximately uniform, when the standard source
is changed to the test source.
The essentials of the CIE (1974) method of
calculating a color-rendering index are as follows:
(i) Reference Iluminant. The reference illumi-
nant for light sources with a correlated color
temperature below 5000 K shall be a Planckian
radiator (Section 1.2.2) and from 5000 K, one of a
series of spectral radiant power distributions of
phases of CIE daylight D [Section 3.3.4(iv)].
The reference illuminant is intended to be of
the same or nearly the same chromaticity as the
light source to be tested. The acceptable chro-
maticity difference between the test source and
the reference illuminant corresponds to 15 mireds
along the Planckian locus in the chromaticity
diagram. The term mired has now been replaced
by “reciprocal megakelvin”; 1 mired = 1 MK7™!
= 10° K“! (see Section 3.11).
(ii) Test-Color Samples. For calculating the
CIE General Color-Rendering Index, a set of eight
test-color samples is specified in terms of spectral
radiance factors. Under daylight illumination,
these samples represent colors around the hue
circuit of moderate saturation and equal lightness.
Their approximate Munsell notations and color
174 COLORIMETRY

appearance under daylight are as follows: (i)], the state of chromatic adaptation of the ob-
Approximate server’s eye is assumed to change slightly when
Sample Munsell Color Appearance the reference illuminant is changed to the test
No. Notation Under Daylight source. The change in chromatic adaptation causes
an adaptive color shift for each test sample which
] 7.5R6/4 Light grayish red can be estimated by making use of a von Kries-
2 5 Y6/4 Dark grayish yellow type linear transformation of the CIE tristimulus
3 5GY6/8 Strong yellow-green values (see Section 5.12).
- 2.5G6/6 Moderate yellowish Once the adaptive color shift has been applied,
green the color difference between the test-color sample
5 10 BG 6/4 — Light bluish green illuminated by the test source and the same
6 5PB6/8 Light blue test-color sample illuminated by the reference il-
7 2.5P6/8 Light violet luminant is calculated making use of the CIE
8 10P6/8 Light reddish purple standard color-difference formula. In 1974, when
the CIE method of calculating the color-rendering
For calculating CJE Special Color-Rendering index was issued, the CIE 1964 color-difference
Indices, any one of the above eight test-color formula was in force and recommended for use in
samples may be used in addition to any one of these calculations. However, as outlined in Sec-
the following six color samples whose spectral tion 3.3.9, in 1976 the CIE modified the CIE 1964
radiance factors are also specified by the CIE: color-difference formula slightly and, in addition,
recommended another formula of equal merit for
Approximate practical applications. This development of 1976
Sample Munsell Color Appearance makes the CIE 1974 method of calculating a
No. Notation Under Daylight color-rendering index out of date. A correspond-
ing revision of the 1974 method is required and
9 45R4/13 Strong red the CIE Color-Rendering Committee has begun
10 5 Y 8/10 Strong yellow to prepare such a revision.
1] 45G5/8 Strong green In replacing the CIE 1964 color-difference for-
12 3 PB3/11 Strong blue mula by, for example, the CIE 1976 (L*u*v*)
13 5 YR8/4 Light yellowish pink color-difference formula, only relatively minor
(Caucasian changes in the calculation procedure would occur.
complexion) The principle of the CIE method of calculating a
14 5GY4/4 Moderate olive green color-rendering index would remain the same. In
(leaf green) particular, the calculated color difference (AE),
continues to characterize the resultant color shift
(iii) Determination of CIE 1931 Tristimulus Val-
that the observer perceives when the ith test-color
ues of the Test-Color Samples. The CIE 1931
sample (i = 1 to 14) is first illuminated by the
(X, Y, Z)-tristimulus values of the test-color sam-
reference illuminant and then by the test source.
ples are calculated in accordance with the estab-
lished method given in Section 3.3.8 making use (v) Calculation of CIE Special Color-Rendering
of the given spectral radiance factors of the sam- Indices. A CIE Special Color-Rendering Index
ples, the appropriate spectral radiant power dis- R, is calculated on the basis of the color dif-
tribution of the reference illuminant, and the ference (AE), determined in Step (iv) above for
spectroradiometrically measured radiant power test-color sample i, by means of the following
distribution of the light source to be tested. formula:
(iv) Determination of the Resultant Color Shift.
In Step (iii) above, for each test-color sample, two R, = 100 — 4.6(AE), [1(3.3.11)]
sets of CIE 1931 tristimulus values are obtained.
One set applies to the case where the sample is The index R; has been scaled so that R; = 100 is
illuminated by the light source to be tested. The obtained when the CIE 1931 tristimulus values of
other set applies to the case where the sample is the test-color sample i under the test source are
illuminated by the reference illuminant. the same as those calculated with respect to the
- As the test source and reference illuminant can reference illuminant. In other words, R; = 100
have somewhat different chromaticities [see Step means perfect color rendition for that test-color
 Dominant Wavelength, Excitation Purity, and Colorimetric Purity
sample. The factor 4.6 in Eq. 1(3.3.11) has been
chosen so that the CIE General Color-Rendering
Index [see Step (vi) below] becomes equal to 50
when a “standard warm white” fluorescent lamp
is used as a test source in comparison with an
incandescent lamp as a reference illuminant.
(vi) Calculation of CIE General Color-Rendering
Index. The CIE General Color-Rendering Index
R_, is defined as the arithmetic mean of the eight
CIE Special Color-Rendering Indices R, for the
eight test-color samples i= 1 to 8 described in
Step (11) above:
R.=5 : R, —-[2(3.3.11)]
The development of the CIE General
Color-Rendering Index can be traced through the
literature over a period of many years. Some of
the more important contributions are those of
Jerome and Judd, 1953; Barnes, 1957; Nickerson,
1958; 1960; 1962; Crawford, 1959; Wyszecki,
1959b; Azuma and Mori, 1960; Hennicke, 1960;
Minch, 1960; Ouweltjes, 1960; 1969; Minch and
Schultz, 1963; 1970; Nickerson and Jerome, 1965;
Nayatani et al., 1972a,b; Halstead, 1978.
The concept of the color-rendering properties
of light sources is taken up again in Sections 3.8.5
and 3.8.6 where methods are described that do
not rely on the use of a specific and very limited
set of test-color samples such as those employed
in the CIE method.
3.4 DOMINANT WAVELENGTH, EXCITATION
PURITY, AND COLORIMERIC PURITY
The dominant wavelength of a color stimulus is the
wavelength of the monochromatic stimulus that,
when mixed with some specified achromatic
stimulus, matches the given stimulus in color. Not
all stimuli have a dominant wavelength. The
achromatic stimulus is usually provided by a
standard illuminant such as CIE standard illumi-
nant A or D,;, either of which may be considered
in this connection as having a neutral (or
achromatic) color, devoid of hue. The dominant
wavelength of a color stimulus correlates in an
approximate way with what would be called in
ordinary language the hue of the color stimulus as
observed under everyday conditions. Thus, in
general, color stimuli of constant dominant wave-
175
length perceived under similar conditions would
be said to have the same hue.
For color stimuli not possessing a dominant
wavelength, it is always possible to match the
specified achromatic stimulus with a mixture of
the given stimulus and a monochromatic stimu-
lus. The wavelength of the latter is termed the
complementary wavelength.
Figure 1(3.4) illustrates the definitions of
dominant wavelength and complementary wave-
length and their graphical determination in the
CIE 1931 (x, y)-chromaticity diagram. Tables
have been prepared by Judd (1933b) that allow
the computation of dominant wavelength with
respect to CIE standard illuminants A, B, C in
the CIE 1931 (x, y)-chromaticity diagram.
The excitation purity of a given color stimulus
is an exactly defined ratio of distances in the
chromaticity diagram, indicating how far the
chromaticity point of the given stimulus is dis-
placed from the chromaticity point of the given
achromatic stimulus towards the spectrum locus
or the purple line. Excitation purity correlates
loosely with saturation of the color perceived
under ordinary observing conditions.
The following formulae are used to calculate
excitation purity:
=XT Xy Or = J. ow
a Xp Xy - Vo Yw
[1(3.4)]
The formula giving the least computational rejec-
tion error is preferred. The chromaticity coordi-
nates (x, y) refer to the chromaticity of the given
color stimulus, such as $,(x, y) and S,(x, y) in
Figure 1(3.4); x,,, y, are the chromaticity coordi-
nates of the achromatic stimulus, such as D,;,
and x,, y, are the chromaticity coordinates of the
boundary color stimulus; that is, the monochro-
matic stimulus of wavelength A, in the case of S,,
or the pure purple P in the case of S,. For the
cases illustrated in Figure 1(3.4), p, comes out to
about 60% for S,, and about 35% for S,.
Another scale of purity is provided by colori-
metric purity p, which is defined by the following
sata
formulae:
¥ eM Vier.Be
or _ (2x) ( X= %w
Pe = (7 \(= = ad [2(3.4)
176 COLORIMETRY

CIE 1931
Chromaticity diagram

580
2d (xb, Yb)

Des (xw Yw)

YP(xp,¥b)

0.4 0.6 0.8

x

Fig. 103.4). CIE 1931 (x, y)-chromaticity diagram showing the dominant wavelength
A a(Xp55 ¥,) Of a color stimulus S, with a chromaticity point S,(x, y), determined with respect
to CIE standard illuminant D¢;, whose chromaticity coordinates are denoted by (x,,, y,,). Also
shown is the complementary wavelength A. of a color stimulus S, with chromaticity point
S,(x, y), again with respect to D,;. The intersection P(x,, y,) of the line running through the
chromaticity points D¢<(x,,, y,,) and S,(x, y), with the purple line is needed to calculate the
excitation purity of S,, as explained in the text.

The various symbols in these formulae denote the
same quantities as in the previous formulae for
excitation purity p,, but with one important ex-
ception: in the case of color stimuli that do not
possess a dominant wavelength A, but instead a
complementary wavelength A. [as for S,(x, y) in
Figure 1(3.4)], the chromaticity coordinates x,, y,
in the formulae for p. denote the chromaticity
coordinates of \.., the complementary wavelength
of S,. MacAdam (1938; see also Committee on
Colorimetry, OSA, 1963) has derived formulae
and tables for converting p, to p. and vice versa.
It used to be fairly common practice to con-
vert the chromaticity of a given color stimulus
into the corresponding dominant or complemen-
tary wavelength and the excitation purity or
colorimetric purity, because these terms were con-
sidered more easily related to descriptions of color
in ordinary language. In recent years, however,
this practice has not been much used, and color
stimulus specifications are now expressed pre-
dominantly in terms of chromaticity coordinates.
3.55 COMPLEMENTARY COLOR STIMULI
Pairs of color stimuli, that, when mixed in suit-
able proportions, color match some agreed achro-
matic color stimulus, are known as complementary
color stimuli (Committee on Colorimetry, OSA,
1963). Of special interest are monochromatic
complementary stimuli. Tables 1(3.5) and 2(3.5)
give important examples of pairs of monochro-
matic complementary stimuli involving CIE
 Maximum Attainable Luminous Efficiency of Color Stimuli of Different Chromaticity 177

Table 1(3.5) Pairs of Monochromatic Complementary Color Stimuli with Respect to CIE Standard
Illuminants D,, and A, and the CIE 1931 Standard Observer”

Achromatic Stimulus D,; Achromatic Stimulus A

Wavelengths Power Ratio Wavelengths Power Ratio
A; (nm) A, (nm) Pe Ted A; (nm) A, (nm) PER Dp

380 566.5 0.608-2 380 579.0 0.227-1

390 566.6 0.189-1 390 579.0 0.642-1
400 566.6 0.639-1 400 579.0 0.217
410 566.6 0.195 410 579.0 0.665
420 566.7 0.609 420 S79 0.207 1
430 567.0 0.130 1 430 579.1 0.444 |
440 S615 0.164 1 440 579.3 0.560 1
450 568.3 0.165 1 450 579.5 0.568 1
460 569.6 0215541 460 579.9 0.535 1
470 Size OVI 470 580.7 0.411 1
480 578.4 0.688 480 582.5 0.256 1
485 585.4 0.489 490 587.5 0.140 1
490 600.7 0,353 495 593.3 0.103 1
491 607.0 0.345 500 606.4 0.807
492 617.2 0.375 501 612.0 0.811
493 642.6 0.804 502 619.9 0.886
* * *
503 635.8 0 |
567 430.2 0.761 * * *

568 447.2 0.600 579 381.1 0.437 2

569 456.0 0.620 580 462.1 0.193
570 462.1 0.667 581 472.7 0272
575 475.7 0.113 585 486.2 0.565
580 481.4 0.160 1 590 492.6 0.836
590 487.1 0.23441 600 498.2 0.116 1
600 489.9 0.282 1 620 502.0 0.113 1
620 492.2 0.254 1 640 503.2 0.625
640 492.9 0.138 1 660 503.6 0:235
660 493.2 0.516 680 503.9), 0.670-1
680 493.3 0.147

“Each pair of complementary stimuli (A, and A,), additively mixed in the radiant power ratio
(P,,/Px,): 10”, matches for the CIE 1931 standard observer the chromaticity of the achromatic stimulus
(Des or A).
Achromatic stimulus Dgs: xp,, = 0.3127 yp,, = 0.3290
Achromatic stimulus A: x,=0.4476 yx =0.4074

standard illuminants A and D,, as achromatic 3.6 MAXIMUM ATTAINABLE LUMINOUS

stimuli and the CIE 1931 and 1964 standard
observers. Each pair of monochromatic stimuli is
defined by a pair of wavelengths A, and A, and
the ratio P,,/P,, of radiant powers required for
their mixture to provide the desired chromaticity
match with the achromatic stimulus. Figure 1(3.5)
illustrates a few examples in the CIE 1931
(x, y)-chromaticity diagram with the data taken
from Table 1(3.5).
EFFICIENCY OF COLOR STIMULI
OF DIFFERENT CHROMATICITY
In general, there is an infinite number of color
stimuli with different spectral radiant power dis-
tributions that have in common some given chro-
maticity. But for these color stimuli, the ratio of
the luminous power to the radiant power, that is,
the luminous efficiency expressed in lumen per
178 COLORIMETRY

Table 2(3.5) Pairs of Monochromatic Complementary Color Stimuli with Respect to CIE Standard
Illuminants D,. and A, and the CIE 1964 Standard Observer“

Achromatic Stimulus D,; Achromatic Stimulus A

Wavelengths Power Ratio Wavelengths Power Ratio
d, (am) d> (nm) Py Px P A, (am) \> (nm) P,,/Pr, P
380 560.9 0.642-3 380 ayes, 0.217-2
390 560.9 0.955-2 390 75:1 0.323-1
400 561.1 0.785-1 400 575.1} 0.265
410 561.3 0.356 410 575.4 0.120 1
420 561.7 0.891 420 at ey 0.300 1
430 562.2 0.142 1 430 575.4 0.480 1
440 562.9 0.179 1 440 ATS6 0.607 1
450 564.0 0.179 1 450 D159 0.615 1
460 565.9 0.15358 460 576.6 0.536 |
470 570.4 0.109 1 470 578.0 0.400 1
475 5156 0.812 480 581.8 0.227 |
480 584.6 0.562 485 586.1 0.162 1
482 591.1 0.482 490 593.7 0.114 1
484 602.1 0.440 495 614.5 0.989
485 611.3 0.457 496 627.1 0.125 1
486 629.6 0.668 497 655.6 0.422 |
* * * * * *

561 394.5 O37 2

562 425.8 0.822 576 451.3 0.164
563 440.8 0.554 577 464.4 0.207
565 456.4 0.604 580 476.3 0.351
570 469.5 0.891 590 487.9 0.760
580 477.9 0.1501 600 492.3 0.100 1
600 483.7 22014 620 495.6 0.934
620 485.6 0.191 1 640 496.6 0.504
640 486.3 0.100 1

“Each pair of complementary stimuli (A, and A,), additively mixed in the radiant power ratio
(P,,/P,)° 10”, matches for the CIE 1964 standard observer the chromaticity of the achromatic stimulus
(Des or A).
Achromatic stimulus Dg: xp,. = 0.3138 yp,, = 0.3310
Achromatic stimulus A: X= 0.4512 y, = 0.4059

watt, varies widely. It turns out that for human
color vision, the luminous efficiency for any given
chromaticity is maximal for a color stimulus
consisting of a mixture in the correct proportions
of a particular pair of monochromatic stimuli
(MacAdam, 1950a). The luminous efficiencies for
the various monochromatic stimuli themselves are
given by the luminous efficiency function V(A) =
y(A) multiplied by K,,, where K,, = 683 Im-
W' is the maximum luminous efficacy for a
monochromatic stimulus that occurs at wave-
length 555 nm, where V(A) = 1. Figure 1(3.6)
shows the CIE 1931 chromaticity diagram with
lines of constant maximum attainable luminous
efficacy as computed by MacAdam (1950a). From
this diagram one finds that the chromaticity x =
0.313, y = 0.329 (the chromaticity of illuminant
D,;) can be produced with a maximum luminous
efficacy of approximately 425 Im:W !'. To
achieve this efficacy, a source giving a mixture of
two wavelengths only is required, one wavelength
at about 450 nm, the other at about 570 nm. Such
a source is of little use in most applications. The
maximum attainable luminous efficacies from
given chromaticities are of interest mainly in defi-
ning upper limits with which efficacies of practi-
 Optimal Object-Color Stimuli 179

CIE I93I
Chromaticity Diagram
‘\
‘y550
tS

0.2

d,=470 Pd
a
0.0 4 0.4 0.6 0.8
d, = 38Onm 0.2

Fig. 1(3.5). Three pairs of monochromatic complementary stimuli that, when additively mixed
in the given radiant power ratios, match the chromaticity of the achromatic stimulus, given by
CIE standard illuminant Des (xp = 9.313, yp = 0.329). There are no monochromatic comple-
mentary stimuli of wavelength A, for given stimuli of wavelength A, between 493.3 and 566.5
nm (dotted part of spectrum locus). The power ratios of the complementary pairs of stimuli of
wavelengths illustrated in the graph are as follows: P399 /Ps66.5 = 9.00608; P4549 / P5722 = 1.17;
P493,/ P42.6 = 9.804.

cal sources can be compared and the scope for
improvement assessed.
3.7 OPTIMAL OBJECT-COLOR STIMULI
Object-color stimuli are color stimuli that arise by
the reflection (or transmission) of incident radiant
power by objects (see Section 3.3.4). Suppose the
incident radiant power has a fixed spectral distri-
bution {P, dA}. The object-color stimulus is then
governed by its spectral reflectance function p(A)
throughout the spectrum, with 0 < p(A) <1 for
all A (or for transmitting objects by 7(A), 0 <
T(A) < 1 for all A). The tristimulus values X, Y, Z
for any particular object-color stimulus, that is,
any particular p(A) [or 7(A)] function derived for
a standard observer, for example, the CIE 1931
standard observer, define a point in (X, Y, Z)-
tristimulus space. The points obtained by taking
all possible object-color stimuli, that is, all suita-
ble functions p(A) [or T(A); the same functions
can serve in either capacity] define a solid in the
tristimulus space that is called the object-color
solid. A change in the absolute value of the inci-
dent radiant power that leaves its relative spectral
distribution the same, gives an object-color solid
of the same shape and position but scaled up or
down (i.e., the X, Y, Z values of any object-color
stimulus are all multiplied by a factor, the same
- for all object-color stimuli). It is convenient,
therefore, to regard the object-color solid as a
property of the relative spectral distribution of
the incident radiant power and to define it for an
absolute radiant power distribution normalized to
make .
[PIO dd = 100
A
180 COLORIMETRY
Fig. 1(3.6). CIE 1931 (x, y)-chromaticity with lines of constant maximum attainable luminous
efficacy (after MacAdam, 1950a).
The Y value for an object-color stimulus is then
its luminous reflectance (or transmittance).
The object-color solid so defined is, in general,
an asymmetric “torpedo”-shaped body, with ver-
tices at the origin and at the point W with tri-
stimulus values Xj, Yj, = 100, Z,,, correspond-
ing to the object-color stimulus involving the per-
fect reflecting diffuser as the object with p(A) = 1
at all wavelengths. All other points X, Y, Z of the
solid satisfy the conditions
0<X<X,; 0< Y< 100; OZ= Zy
It is readily shown that all points on the straight
line joining two points of the solid also belong;
that is, the solid is convex or at least not concave.
For a monochromatic incident radiant power, the
object-color solid degenerates to a line joining
Xw,Yw = 100, Zy to the origin; for an incident
radiant power comprising just two wavelengths, it
degenerates to a diamond-shaped plane figure
with OW as one diagonal. It is truly a solid if the
incident radiant power contains three or more
wavelengths. In most applications, the incident
radiant power possesses an everywhere-positive
continuous visible spectrum.
A straight line drawn from the origin through
any other point P( Xp, Yp, Zp) of the surface of
the object-color solid contains all the points rep-
resenting object-color stimuli of some fixed chro-
maticity (xp, yp) defined by the direction of the
line. Two situations develop:
(a) There may be no points of the object-color
solid that are on
the line and that lie beyond the
initial segment OP; and
(b) There may be points of the line beyond OP,
up to a farthest point P’, that belong to the solid,
in which case OP’ must lie on the surface of the
solid. :
In case (a) it is clear that P is the tristimulus
point corresponding to object-color stimuli that
for the given chromaticity (x, y) have the greatest
value of Y and hence, the greatest luminous re-
flectance or transmittance.
In either case (a) or case (b) the plane Y =
constant drawn through P intersects the color
solid in a closed (nonconcave) plane figure. Any
point Q, as shown in Figure 1(3.7), within this
plane figure may be arbitrarily chosen as the

Y= const= 50
120
100
80
60
40
20
20 40 60 80 100
Fig. 1(3.7). Cross section Y = const = 50.0
through object-color solid in (X, Y, Z )-tristimulus
space.
point representing the achromatic color stimulus
with respect to which dominant or complemen-
tary wavelength and excitation purity can be
defined (see Section 3.4). Object-color stimuli cor-
responding to all points on the segment QP have
the same luminous reflectance or transmittance
Y, and they have either the same dominant wave-
length or the same complementary wavelength. It
will be the dominant wavelength that is held in
common if QP produced beyond P passes through
a point representing a monochromatic color
stimulus; that is, a point which will not, of course,
belong to the object-color solid. Thus, P repre-
sents object-color stimuli that have the maximum
possible excitation purity for given luminous
reflectance and given dominant (or complemen-
tary) wavelength. Object-color stimuli possessing
the maximal property are termed optimal color
stimuli, and it is clear that every point on the
surface of the object-color solid represents one or
more optimal color stimuli.
Optimal color stimuli are sometimes described
as object-color stimuli that for a given chromatic-
ity have the greatest luminous reflectance or
transmittance. The representative point of every
such stimulus must lie on the surface of the
object-color solid, but a point P of the surface
will be representative of color stimuli that are
optimal according to this second definition, only
if the above case (a) applies. The first definition
Optimal Object-Color Stimuli 181
on which all points of the surface of the
object-color solid represent optimal color stimuli
is adopted here.
For incident radiant power with an every-
where-positive continuous spectrum, optimal color
stimuli have, as a rule, spectral reflectance curves
of a very simple type. The spectral reflectance
p(A) is either zero or unity, and in moving through
the visible spectrum, there are generally not more
than two transitions between these values. Opti-
mal color stimuli are imaginary stimuli in the
sense that no actual object surfaces have reflec-
tance curves with abrupt transitions of this kind.
However, they are of considerable interest be-
cause they represent limiting cases of all (nonfluo-
rescent) object-color stimuli.
The theory of optimal color stimuli was
developed first by Schrédinger (1920a) and later
by Rédsch (1929) and MacAdam (1935a, b).
MacAdam computed chromaticity loci of optimal
color stimuli as a function of Y, on the basis of
the CIE 1931 standard observer and for incident
radiant power from standard illuminants A and
C, respectively. MacAdam’s computations have
been repeated for illuminants A and D,,; with
updated input data. The results of these computa-
tions are given in Table I(3.7) and IJ(3.7) which
400
REFLECTANCE
TANCE
TRANSMIT
OR A, 500 2 600 700
WAVELENGTH (nm)
400
REFLECTANCE
TRANSMITTANCE
OR 500 ™ 600 - 2 700
WAVELENGTH A (nm)
Fig. 2(3.7). Spectral reflectance (or transmit-
tance) curves of the two types of optimal color
stimuli.
182 COLORIMETRY

0.8

0.7

0.6

5009

0.5

0.4

0.3
n650
770nmMm

0.2

0.3 0.5 0.6 0.7 0.8

Fig. 3(3.7). Chromaticity loci of optimal color stimuli as a function of Y, on the basis of the
CIE 1931 standard observer and for incident radiant power from the CIE standard illuminant A.

include the transition wavelengths (A,,A,) for
each optimal color stimulus listed. From these
wavelengths the spectral reflectance (or transmit-
tance) curve corresponding to a given optimal
color stimulus can readily be drawn.
‘Two types of curves must be distinguished; the
first has zero reflectance (or transmittance) at
wavelengths A<A, and A>A,, the second at
wavelengths A, <A <A,, as illustrated by the
examples shown in Figure 2(3.7). In Tables I(3.7)
and I1(3.7), the optimal color stimuli with Type 1
spectral reflectance (or transmittance) curves are
listed in the upper half of each group of columns
(Y = 95.0, Y = 90.0,..., Y = 10.0), those of Type
2 in the lower half separated from the upper half
by short lines.
The chromaticity loci specified in Tables I(3.7)
and II(3.7) are plotted in Figures 3(3.7) and 4(3.7).
In the American literature, these loci are often
called the MacAdam limits. An oblique projection
of the (x, y, Y)-space, showing the boundary
surface representing the optimal color stimuli for
illuminant D,, is given in Figure 5(3.7). In the
German literature, the (x, y, Y) object-color solid
 Metameric Color Stimuli 183

0.8

0.6

5009

0.4

650
770 nm

0.2

0.6 0.8
X
Fig. 4(3.7).Chromaticity loci of optimal color stimuli as a function of Y, on the basis of the
CIE 1931 standard observer and for incident radiant power from the CIE standard illuminant
Da

is known as the Résch Farbkérper (Rosch color
solid). Still another representation of the object-
color solid is given in Figure 6(3.7). In this
Luther—Nyberg color solid (Nyberg, 1928), the co-
ordinates M,, M,, S, and the tristimulus values
X, Y, Z are related as follows:
M,=—-X+Y — (First color moment)
M,=-—Z-—Y _ _ (Second color moment)
S=X+Y+Z _— (Color weight) —
3.8 METAMERIC COLOR STIMULI
Metameric color stimuli have been encountered in
several previous sections, and in each one they
have played a key role in the discussions of
different aspects of colorimetry. Because of their
importance to many basic colorimetric concepts,
we have prepared the following in which we offer
a more comprehensive treatment of the concept
of metamerism than was considered necessary in
the previous sections.
184 COLORIMETRY

08

Fig. 5(3.7). Oblique projection of the (x, y,Y) object-color solid showing the boundary
surface representing the optimal color stimuli (Résch color solid).

3.8.1 Definition of Metamerism spectral radiant power distributions but match in

color for a given observer.
Metameric color stimuli are color stimuli with the
Metameric color stimuli, as defined above, are
same tristimulus values but different spectral
often referred to as metamers, and the underlying
radiant power distributions.
concept is referred to as metamerism.
An equivalent definition states that metameric
In quantitative colorimetric terms, two meta-
color stimuli are color stimuli that have different
meric color stimuli whose different spectral
radiant power distributions are denoted by
{P,, dX} and {P,, dX}, respectively, must satisfy
the following equations:

[Pn7Q) dn = [Parr) dn

Fig. 6(3.7). Representation
Nyberg color solid.
fPnzQ) dd = [Parg(A) dd
JPnbO) dd = [Paxb ) dn
[1(3.8.1)]
These equations were presented in Section 3.2.3
as the necessary conditions for a complete color
match between the two given stimuli of spec-
of the Luther- tral radiant power distributions {P,, dA} and
{P,, dd}. The quantities 7(A), Z(A), b(A) define
 Metameric Color Stimuli 185

the color-matching properties of the observer per-
ceiving the stimuli; they are referred to as the
color-matching functions.
3.8.2 Methods of Generating Metamers
Metamers are generated in virtually all color re-
production processes; that is, in processes de-
signed to duplicate the color of a given light or
the color of a given object by means of an ap-
propriate mixture of a limited number of
“primary” lights or colorants.
The following examples may serve to illustrate
the kinds of metamers that are commonly gener-
ated in the color vision research laboratory and in
industry.
Figure 1(3.8.2) shows the spectral radiant
power distributions of a pair of metameric stimuli
Q, and Q,, each consisting of an additive mix-
ture of two monochromatic stimuli. Such meta-
meric color stimuli are generated in experiments
designed to determine the color-matching func-
tions of an observer with trichromatic vision.
Details of such experiments and their results are
given in Sections 3.2.1 and 5.5.
Figure 2(3.8.2) shows the spectral radiant
power distributions of CIE standard illuminant
Ds; (representative of average daylight) and the
light emitted by a fluorescent lamp designed to
duplicate the color of average daylight. The two
spectral distributions define a pair of metameric
color stimuli.
Figure 3(3.8.2) shows the spectral radiant
power distributions of daylight reflected from a
Caucasian face and light emitted from a color
television screen tuned to duplicate the color of
the face. The light from the color television screen
is an additive mixture of three primary lights
generated by three phosphors deposited on the
screen in the form of numerous small adjacent
dots (Judd and Wyszecki, 1975; Hunt, 1967;
Wright, 1969).
Figure 4(3.8.2) shows a pair of spectral reflec-
tance curves of two objects that, under CIE
standard illuminant D,;, have the same color for
the CIE 1931 standard observer. The objects are
textile swatches colored by mixtures of different
dyestuffs. More details on formulating mixtures
of dyestuffs (or colorants) are given in Section
379)

On

10° bipartite
ba visual field

vo)
Q,=1.00T +0.40R

¢ Q, =0.33B+0.37G
y
7

fe) (6,

y
)
units
(arbitrary
Radiant
Power
400 X8 Ar 500 Xe
600 700

444.4 4800
|: 526.|
B G
Wavelength A (nm ) —
Fig. 1(3.8.2). Example of two color stimuli Q, and Q, generated in the 10° bipartite field of a
visual colorimeter, that are in complete color match for observer GW. Color stimulus Q, is an
additive mixture of the two monochromatic stimuli B and G of wavelengths A , = 444.4 nm and
A g = 526.1 nm, and of relative radiant powers 0.33 and 0.37, respectively. Color stimulus Q, is
an additive mixture of the two monochromatic stimuli T and R of wavelength A, = 480.0 nm
and A p = 645.1 nm and of relative radiant powers 1.0 and 0.4, respectively. The stimuli Q, and
Q, are called a metameric pair of color stimuli.
186 COLORIMETRY

340

320 x

300

280
se
d
see
ewe
|
5
oaks
Reaiah
aie

260 a 0 ees
eo FLUORESCENT LAMP
200 { )

ie
_
‘et 1
Gre i
z 7 ! |
!

> 160 1

::
FE !40
:
t
|| ::
|
'

1
Ves f

r
<q
Sie he
S
or a
= fe
1 ot

oO 100 feat
a !
= i
2 +280 te
<q
—_—
ie
! 1

afa ac0
|i

40

20

300 400 500 600 700 800 900

WAVELENGTH i (nm)

Fig. 2(3.8.2). Spectral radiant power distributions of two metameric color stimuli: average
daylight (CIE standard illuminant D,,;) and light from a fluorescent lamp.

The examples of metameric color stimuli given
in Figures 1(3.8.2) to 4(3.8.2) illustrate the do-
main of metamers encountered in color vision
research and in industry. There is a vast supply of
metamers from such sources and indeed one could
establish a large collection of them without too
much difficulty. However, to the theorist inter-
ested in exploring its consequences to different
aspects of colorimetry, such a collection is often
not very attractive. The theorist prefers a collec-
tion in which the variability of the spectral radiant
power distributions defining the metamers is not
unduly constrained by the particular processes
used in generating the metamers. The theorist
prefers to generate a suitable collection of
metamers by numerical means.
Several numerical methods of generating meta-
mers have been reported in the literature. In
presenting some of these methods, we will choose
as our observer, unless stated otherwise, the CIE
1931 standard observer defined by the color-
matching functions x(A), y(A), Z(A). This choice
is for convenience only and will not alter the
general conclusions drawn. In fact, this choice
will often help to interpret the results in terms of
a familiar colorimetric system. Furthermore, we
will confine our discussions to object-color stimuli
as these are of prime interest in practical colorim-
etry. Finally, we will replace, in the usual way
(Section 3.3.8), all integrations by summations,
and use as illuminants those recommended by the
CIE that are defined in terms of relative spectral
radiant power distributions {S(A) AA}. Equa-
tions 1(3.8.1) of the previous section then take on
the following form:
DLei(A)Si(A)x(A) AA = Sip2(A)S2(A)x(A) AA
A
LUe(A)S(APA) AA = J p2(A)S(A)
(A) AA
A
2 i(A)Si(A)z(A) AA = i po(A)S(A)z(A) AA
A
[1(3.8.2)]
 Metameric Color Stimuli 187

illuminant is used to irradiate the two given ob-

60
jects; thus $,(A) = S,(A). In such situations, the
he
spectral differences between the two metameric
$2 50
a. object-color stimuli are given solely by the dif-
ferences between the two spectral reflectance
5 40
functions p,(A) and p,(A) defining the two ob-
me)
5 jects.
© 30
@o
<Z (i) Metameric Black Method. In the metameric
° 20 Caucasian Face black method, developed by Wyszecki (1953;
td in Daylight
1958), spectral reflectance functions p,(A) are
10 generated, which for a given illuminant S(A) and
a given observer x(A), y(A), Z(A), fulfill the fol-
lowing conditions:
400 500 600 700
Wavelength dX (nm)
Fig. 3(3.8.2). Spectral radiant power distributions
De(A)S(A)E(A) AA =0
of daylight reflected from Caucasian face and
emitted by color television screen tuned to match TeA)SAFA)AA=0 [23.82)]
the color of the face. The light from the screen is
an additive mixture of light emitted by the three
(red, green, blue) phosphors which are deposited as Bey(A)S(A)z(A)AA=0
small adjacent dots on the screen.
The tristimulus values X, Y, Z calculated for each
In these equations, the spectral radiant power dis- stimulus {p,(A)S(A) AA} are equal to zero; hence
tributions {p,(A)S\(A) AA} and {p,(A)S(A) the name metameric black. As in the case of
AX}, given in arbitrary radiometric units, define a ordinary metameric stimuli, there are infinitely
pair of metameric object-color stimuli with p,(A) many metameric blacks that fulfill Eq. 2(3.8.2).
and p,(A) denoting the spectral reflectance func- All metameric blacks, more specifically all spec-
tions of the two given objects and S$,(A) and tral reflectance functions p,(A), have the
S,(A) denoting the relative radiant powers at important feature of being positive at some wave-
wavelength A of the two given illuminants lengths and negative at others, and thus, they
irradiating objects (1) and (2), respectively. Cases cannot be realized physically. Nevertheless, they
are of interest only in which can be used very effectively for constructing sets
of all positive spectral reflectance functions that
{p(A)Si(A) AA} * {p2(A)S2(A) AA} are metameric and match some color other than
black. The procedure is to add a spectral reflec-
Most practical situations are concerned with tance function p,(A) of a metameric black to a
metameric object-color stimuli in which the same given spectral reflectance function p,(A) of non-
vanishing tristimulus values X,, Yo, Z >. One ob-
tains a new spectral reflectance function

p(A) = po(A) + p,(A) [3(3.8.2)]

which has the tristimulus values Xp, Yo, Zp. The

procedure is useful only if the new function p(A)
fulfills the condition
Factor
Radiance
0<p(A) <1.0 [4(3.8.2)]

at all wavelengths A. Obviously, this is a neces-

(nm) sary condition that must be satisfied by all spec-
Fig. 4(3.8.2). Spectral reflectance curves of two tral reflectance functions of real (nonfluorescent)
textile swatches that, under CIE standard illumi- objects. In general, this condition can readily be
nant D,<, match in color for the CIE 1931 stan- met by multiplying p,(A) by an appropriate scal-
dard observer. ing factor before adding it to po(A).
188 COLORIMETRY a
ro)
\
ro) NY)
ro)
REFLECTANCE
Fig. 5(3.8.2). Example of the spectral reflectance
function of an elementary metameric black.
Table I(3.8.2) contains 27 linearly independent
spectral reflectance functions of black objects
which under CIE standard illuminant C have zero
CIE 1931 (X,Y, Z)-tristimulus values. The 27
functions p‘)(A) (i = 1 to 27) were calculated in
accordance with Eq. 2(3.8.2) with the visible spec-
trum taken from A = 400 to 690 nm, and divided
into wavelength intervals AA = 10 nm. They rep-
resent samples of the most elementary spectral
reflectance functions of metameric blacks that
can be constructed. Each reflectance function has
only four values different from zero. Three are
located always at 450 nm, 520 nm, and 600 nm,
respectively; the fourth, which is always equal to
unity, varies its position in the spectrum from one
reflectance function to the other. Figure 5(3.8.2)
illustrates function i = 14 of Table 1(3.8.2).
Any linear combination of the 27 spectral re-
flectance functions will provide new functions of
metameric blacks. Because the 27 functions of the
given basic set are linearly independent, they can
be combined in suitable linear combinations to
reproduce the spectral reflectance function of any
metameric black defined by its values (supposed
constant) in 30 wavelength intervals AA = 10 nm
centered on the 30 specified wavelengths (A, =
400, A, = 410, A; = 420,...,A3) = 690) of the
visible spectrum.
From the values of Table 1(3.8.2), a mathe-
matically more elegant basic set of spectral reflec-
tance functions of metameric blacks may be de-
rived by orthogonalizing the functions. The 27
orthogonal spectral reflectance functions of the
new set are given in Table II(3.8.2). Each pair of
functions fulfills the following condition:
DPP(A)PYP(A) =O — [5(3.8.2)]
A
with i, 7= 1,2,...,27 andi #/.
Figure 6(3.8.2) illustrates two spectral reflec-
tance curves of a metameric pair of object-color
stimuli with tristimulus values X = 24.51, Y=
25.00, Z = 29.53 and corresponding chromaticity
coordinates x = 0.310, y = 0.316. They represent
objects that, when irradiated by illuminant C,
would appear gray under normal viewing condi-
tions. Whereas Curve 1 represents a spectrally
nonselective object, Curve 2 is a spectrally highly
selective gray. Curve 2 was found by adding to
Curve 1 a metameric black curve p,(A) that is
0.4342 times the sum of curves i = 19, 20, and 21
given in Table II(3.8.2).
The metameric black method has been applied
to the study of the color-rendering properties of
light sources (Wyszecki, 1960a) and to the assess-
ment of overall differences of alternative sets
of color-matching functions (Wyszecki, 1959b).
These applications will be touched upon briefly in
Section 3.8.3 dealing with the general problem of
counting metamers.
(ii) Linear Combination Method. This method
requires a collection of sets of three different
spectral reflectance (or transmittance) functions
PPA) pH) 253 yd dy 284, eng) erties
made up artificially or determined from spectro-
photometric measurements of real materials, such
that the following n sets of three simultaneous
linear equations have unique solutions:
ai 2Pr(A)S(A)¥(A) AX
peizion (SC) A0) Ar
agrees (ASO) (2.)AA = X,
Bit Quy CEST)
Fain S81 WAS (Aga) Ar
+a Xp(A)S(A)
(A) Ad = ¥,
r
 Metameric Color Stimuli 189

0.5

REFLECTANCE

400 500 600 700

WAVELENGTH \)
(nm)

Fig. 6(3.8.2). Spectral reflectance functions of a pair of (hypothetical) metameric object-color

stimuli of tristimulus values X = 24.51, Y = 25.00,Z = 29.53, with respect to the CIE 1931
standard observer and CIE standard illuminant C. Curve 1 represents a nonselective gray
(horizontal straight line) of 25% reflectance; Curve 2 represents a metameric gray constructed
by the metameric black method described in the text.

aj, Ye; (A)S(A)Z(A)AA that yields the given tristimulus values with re-
A spect to the same illuminant and observer.
In keeping the tristimulus values the same,
Fain d pi(A)S(A)Z(A) AA that is, X, = X, Y, = Y, Z, = Z, Eqs. 6(3.8.2) and
nN 7(3.8.2) provide the spectral reflectance functions
of n metameric object-color stimuli of spectral
+32
p) (A)S(A)zZ(A) AA = Z, radiant power distribution {p;(A)S(A) AA} (with
r i= 1 ton).
Up to this point, the procedure of generating
[6(3.8.2)] metamers follows the one suggested by Richter
(1958). Occasionally, however, the resulting spec-
For the ith set of three spectral reflectance func- tral reflectance function p,(A) may exceed the
tions p$?(A), pA), pO(A) drawn from the col- limits
lection of n such sets, and for given tristimulus
values X,, Y,, Z;, given illuminant S(A) and given 0<p,(A) <1.0 [8(3.8.2)]
observer x(A), y(A), Z(A), one can calculate the
factors dj), 4;7,4;3. It readily follows that the
linear combination
that must hold for any real (nonfluorescent) ob-
ject. Whenever this occurs, a scaling procedure
must be applied to bring p,;(A) within the spectral
a;p\)(A) a ainp(A) a aj3p)(A) con p(A)
range while preserving the original tristimulus
[7(3.8.2) values. In general, this can be done readily.
Suppose, for i = k, the computed p,(A) has at
defines a new spectral reflectance function p;(A) some wavelengths values greater than unity. To
 190 COLORIMETRY

0.4
REFLECTANCE

0.2 Neares)

0.8

4
0.4
REFLECTANCE

0.0

WAVELENGTH 2 (nm)

Fig. 7(3.8.2). Twelve spectral reflectance functions of objects yielding (hypothetical gray)
object-color stimuli with respect to CIE standard illuminant C and the CIE 1931 standard
observer.

modify p,(A), one first subtracts p,(A) from
another spectral reflectance function of the same
set, for example, p,(A) obtained for i = /. The
difference [p,(A) — p,(A)] yields the spectral
reflectance function p,(A) of a metameric black
as defined earlier by Eq. 2(3.8.2). The function
p,(A) can be multiplied by any positive or nega-
tive factor without changing its character as a
metameric black. A suitable factor a can be found
such as to ensure that the new function
p, =p(A) + ap,(A) —[9(3.8.2)]
falls within the limits
0<p,(A) < 1.0
Table III(3.8.2) and Figure 7(3.8.2) give 12
spectral reflectance functions p;(A) (with i = 1 to
12) constructed by the method just described. The

functions define 12 hypothetical objects that, un-
der CIE standard illuminant C, match in color for
the CIE 1931 standard observer. The object-color
stimuli obtained in this way form a metameric set
of stimuli with the following common tristimulus
values and chromaticity coordinates:
X = 29.41, Y = 30.00, Z = 35.43
x = 0.310, y = 0.316
The metamers generated above refer to a par-
ticular illuminant S(A) and a particular observer
x(A), y(A), Z(A). If a change is made in either the
illuminant, by replacing S(A) with S’(A), or in
the observer, by replacing x(A), y(A), Z(A) with
x'(A), y'(A), Z’(A), the object-color stimuli that
originally matched in color will generally do so no
longer.
However, it is possible to extend the method
based on Eqs. 6(3.8.2) and 7(3.8.2) to generate
spectral reflectance functions p,(A) that yield ob-
ject-color stimuli that are metameric with respect
to more than one illuminant and/or observer.
For example, the extended method can be used to
generate p,(A) functions that under illuminant
S(A) will have tristimulus values X,Y, Z, and
simultaneously tristimulus values X’, Y’, Z’ under
illuminant S’(A). Thus, the given observer will see
the different objects match in color under S(A) as
well as under S’(A). However, under each illumi-
nant, a different color may be perceived.
The extended method uses the following equa-
tions when two illuminants [.$(A) and S’(A)], and
one observer [x(A), y(A), Z(A)], are considered:
J
S [a 200)S0)F0) AN=X,=X
rv
fai,BoAIS(A) TO) AD Siar
3 fa,BISA) jr= Zils
fa,B4A)S'O)FO) Ait N CX?
faSeA)S'O)FO) A= Y=
> fa,20)9')z0) AN=Z/=Z’
= r
[10(3.8.2)]
Metameric Color Stimuli 19]
These equations form a set of 3 X 2 = 6 simulta-
neous linear equations into which is inserted the
ith set (i= 1 to n) of six given spectral reflec-
tance functions p\/)(A) (with j = 1 to 6). Assum-
ing that the functions p\/)(A) have been chosen
such that the determinant of the system of equa-
tions does not vanish, a unique solution is ob-
tained for the six factors a; ;(j = 1 to 6,i = fixed).
The desired spectral reflectance function p;(A)
then follows from the equation
6
p;(A) = 2 a; ;p\(A) [11(3.8.2)]
If p,(A) does not meet the condition expressed by
Eq. 8(3.8.2), appropriate adjustments can be made
in accordance with the procedure outlined above
[Eq. 9(3.8.2)].
Figure 8(3.8.2) shows an example of two spec-
tral reflectance curves calculated by means of
Eqs. 10(3.8.2) and 11(3.8.2) with S(A) being the
spectral radiant power distribution of CIE illumi-
nant C and S’(A) that of CIE illuminant A. For
the CIE 1931 standard observer, the two objects
defined by the two curves match in color when
viewed under either illuminant C or illuminant A.
It is apparent that the method given by Eqs.
10(3.8.2) and 11(3.8.2) can be extended further to
generate metamers with respect to more than two
illuminants and to more than one observer. Such
an extension leads to correspondingly larger sys-
tems of simultaneous linear equations involving
correspondingly increased numbers of initial
spectral reflectance functions p‘/)(A). Such com-
plex cases will not be pursued here.
(iii) Method Based on Variation Calculus.
Takahama and Nayatani (1972) developed an in-
teresting method of generating metamers by mak-
ing use of variation calculus.
The method requires a collection of initial
spectral reflectance functions p,(A) (with a =
1,2,...,n) each leading to tristimulus values
X,, Yy, Z, with respect to a given illuminant S(A)
and given observer such as x(A), y(A), Z(A). As
the tristimulus values X,, Y,, Z, do not need to
be the same for different a, it is relatively easy to
acquire a suitable collection of p,(A) functions.
For example, one may use the spectral reflec-
tance functions obtained by spectrophotometry of
a set of paint samples selected from the Munsell
Book of Color (Section 6.6.1).
Alternatively, one may produce p,(A) func-
tions by a random number technique. The visible
Ps COLORIMETRY

a3

oO ron)

Reflectance
factor

0.2

0
400 500
Wavelength (A), nm
Fig. 8(3.8.2). Example of spectral reflectance curves of two (hypothetical) objects that color
match for the CIE 1931 standard observer when these objects are illuminated by CIE standard
illuminant C or by CIE standard illuminant A (Judd and Wyszecki, 1975).

spectrum is divided into an appropriate number and

of intervals of equal width AA. At the central
wavelength A of each such interval, the value of
p,(A) is chosen at random with the constraint
0 <p,(A) < 1.0. With the visible spectrum taken,
for example, from A = 400 to 700 nm at intervals
of width AA = 10 nm, the p,(A) functions are
defined by 31 mutually independent values. Func-
tions produced in this way are jagged and repre-
sent objects usually not found in practice.
The method of Takahama and Nayatani then
consists of changing each given p,(A) function of
the initial collection of spectral reflectance func-
tions to a new function, denoted by p*(A), that
meets the following conditions:
2 Ler(r) i p(A)]° = minimum
[13(3.8.2)]
The main steps of the numerical procedure
used to determine the desired metamers defined
by Eqs. 12(3.8.2) and 13(3.8.2) are more conveni-
ently illustrated by employing the following nota-
tions:
The tristimulus values X,, Y,, Z, of the color
stimulus involving p,(A) of the given initial col- .
lection of spectral reflectance functions are
denoted by:

cr Xa ’ ‘e a Anas Log = X03

as (ASE
AuC CAR) NA =k ta eX
2 Pa(A)S(A)Z(A) Ad pre
The products S(A)x(A), S(A)y(A), S(A)Z(A),
containing the given illuminant and observer, are
Nags: respectively denoted by x,(A), x,(A), x3(A).
Summations (replacing integrations) of cross

Dpz(A)S(A)zZ(A)AA= ZE=Z
A
products of functions f(A) and g(A), taken over
the visible spectrum are denoted by

[12(3.8.2)] (f-g) = 2f(d) (A) AA

 Metameric Color Stimuli 193

With the above equations we have (with i = 1, 2, 3) By introducing Eq. 21(3.8.2) into Eq. 16(3.8.2)
and noting Eqs. 22(3.8.2) and 23(3.8.2) an explicit
Xvi = (Py i x;) [14(3.8.2)] expression is obtained for p*(A):

Ke =~ Dre. , x;) [15(3.8.2)] 8) 3

pz(X) = p,(A) + > yy AX,;B;;x;(A)

by substituting
[24(3.8.2)]
pa(A) = pr) + e(r) [16(3.8.2)] The p*(A) functions derived by means of Eq.
into Eq. 15(3.8.2): 24(3.8.2) do not necessarily satisfy the condition

0<p*(A) <1.0
aM Dart fo ee)
but, if the case arises, appropriate adjustments
[17(3.8.2)] can be made in accordance with the method
described earlier in connection with Eq. 9(3.8.2),
and subtracting Eq. 14(3.8.2) from Eq. 17(3.8.2), or by means of a procedure suggested by Ohta
we obtain
(1975).
Takahama and Nayatani demonstrate the ef-
AX ge (rir) [18(3.8.2)] fectiveness of their method on a number of ex-
with
amples. They show that metameric blacks are
calculated readily from the same initial collection
AX Ae ke [19(3.8.2)] of p,(A) functions, but set Eq. 15(3.8.2) equal to
zero; that is, X*. = 0.
The condition expressed by Eq. 13(3.8.2) with They also show that the method can be used to
regard to p*(A) is fulfilled when the value of change a given set of p*(A) functions, which
yields metamers with respect to a given illuminant
and observer, into a new set of p**(A) functions,
ben 269) =2 [pz(A) — pa(A)] AA which yields metamers with respect to another
[20(3.8.2)] illuminant and/or observer. Table IV(3.8.2) re-
produces the results obtained by Takahama and
becomes a minimum. The determination of this Nayatani for a set of 12 spectral reflectance func-
minimum is referred to as an isoperimetric prob- tions px*(A) (a= 1 to 12) that yield “gray”
lem in variation calculus (see, for example, metamers with respect to CIE standard illumi-
Smirnov, 1964) with the constraint, in our case, nant D,; and the CIE 1931 standard observer.
given by Eq. 18(3.8.2). From Euler’s theorem it These functions were derived from the 12 spectral
follows that the required function ¢,(A) is given reflectance functions given in the previous Table
III(3.8.2) that yield “gray” metamers with respect
by
to CIE standard illuminant C and the CIE 1931
3
standard observer. When the data given in Table
AWM ee 2 Haj%/) [21(3.8.2)] IV(3.8.2) are plotted, similar spectral reflectance
curves are obtained as those shown in Figure
The constant but unknown factors p,,; are de- IB82)
termined from Finally, Takahama and Nayatani show that
their method can also be extended without diffi-
3 culty to the more general case in which p,(A)
Boj — > AX,; Bi, [22(3.8.2)] functions are generated that yield metamers with
i=]
respect to more than one illuminant and/or more
than one observer. For details, the reader is re-
where B;; are the elements of the symmetrical
matrix B, which is the inverse of the 3 X 3 matrix ferred to the original paper by Takahama and
A whose elements are given by Nayatani (1972).
(iv) Monte Carlo Method. A large number of
Ny We od Pro aaae PSL NY metamers can be generated relatively quickly on
194 COLORIMETRY
the computer by making use of a Monte Carlo
method. In this method, every spectral reflectance
function p/(A) (j = 1 to N) is obtained in the
form of a step function by assigning to a fixed set
of wavelengths A, (i = | to M) random numbers
o/(A,) that satisfy the condition
0<p(A,)<1.0 [25(3.8.2)]
A typical set of wavelengths A, divides the visible
spectrum into M equal parts of width AX. If the
visible spectrum is taken from A, = 400 nm
to A, = 700 nm and AA = 10 nm, each spec-
tral reflectance function is defined by M = 31
random numbers p/(A,). By inserting the p/(A;)
function into the familiar equations for cal-
culating tristimulus values with respect to a given
illuminant {S(A,;) AA} and observer [x(A,),
y(A,), Z(A;)] one obtains
Ngege ike 5 p/(A,)S(A,)x(A;) AA
i=]
Yk > e7(A)S(A,)F(A;) AA
i=]
£, =k 5 p/(X;)S(A;)z(A,) AA
i=1
[26(3.8.2)]
with A being a convenient normalizing constant
such as k = 100/3™,S(A,)p(A,;) AA (see Section
3.3.8).
The computer collects all those p/(A;) func-
tions that provide tristimulus values X;; Ys Z; that
coincide or nearly coincide with a set of given
tristimulus values X,Y), Z); that is, all those
that satisfy the conditions
AX AX
( Sealias besesel
Wie AY,
oe ae
AZ AZ
(40 -) <%= (20+)
[27(3.8.2)]
This provides a set of near metamers represented
in tristimulus space by a set of points (X;, ¥;, Z;)
all of which fall into a small box of sides
AX, AY, AZ. The time required for the computer
to find a desired number n of p/(A;) functions
that fulfill Eqs. 25(3.8.2) to 27(3.8.2) depends on
the number M of wavelengths A, used in each
function, and the size as well as the location of
the box in tristimulus space. The size and location
of the box are of particular importance in that
regard. Obviously, the smaller the size of the box,
the longer it will generally take to generate the
desired n metamers. For a given box size, the time
will be a minimum if the center of the box coin-
cides with the center of the object-color solid,
located at
k M
Ain = > > S(A;)x(A;) AA
isl
a M
Xn =£ SSA) FA) Ad
i=1
ke M
i=]
[28(3.8.2)]
The time required to generate n near metamers
will increase rapidly with the length of the
displacement of the box from the centerX,,,
Y,,, Zm- Lhis finding is directly linked with the
statistical problem of determining the proportion
of metameric object-color stimuli that belong to
the total set of p/(A;) functions and are ap-
proximately metameric with a given object-color
stimulus. Section 3.8.4 will deal with this problem
further. As a rule, no metamers will be found
when the box is on or beyond the boundary of the
object-color solid (optimal color stimuli, Section
3.0).
Once the desired n near metamers are found, a
fine tuning can be applied to each p/(A,) function
so as to force it to yield the given tristimulus
values Xp, Yo, Zp), precisely. The method of gener-
ating metamers by variation calculus, developed
-by Takahama and Nayatani (1972) and described
above, may be used for the fine tuning.
3.8.3 Intersections of Spectral Reflectance
Curves of Metamers
Two reflecting objects that yield metameric ob-
ject-color stimuli under a particular illuminant
and for a particular observer must have spectral
reflectance curves that intersect. Intuitively or
from experience, we expect the two spectral re-
flectance curves to intersect at a minimum of

three wavelengths within the visible spectrum:
one intersection at the short end, one
middle, and one at the long end of the spectrum.
More specifically, we expect them to fall ap-
proximately at the wavelengths of the peaks of
the three color-matching functions of the given
observer. Surprisingly, proving our expectation is
not straightforward. Stiles and Wyszecki (1968)
have worked out the exact conditions under which
a minimum of three or fewer than three intersec-
tions is required for ordinary trichromatic vision.
They have also derived the relevant rules that
apply to extended cases when metamerism rests
on more than three matching conditions; for
example, metamerism for both photopic and
scotopic vision, or metamerism with respect to
more than one illuminant. A detailed account of
the theoretical thesis of Stiles and Wyszecki is
beyond the scope of this book. However, it is of
interest to note that the rules developed for one
case, such as the one involving the CIE 1931
color-matching functions, do not necessarily hold
for other (hypothetical) trichromatic systems that
lead to different (but physically feasible) spec-
trum loci in their corresponding chromaticity dia-
grams. The empirical fact that the spectrum locus
of the CIE trichromatic system is nowhere con-
cave, turns out to be an essential parameter in the
development of the rules concerning the number
of intersections of the spectral reflectance curves
of two metamers based on such a trichromatic
system.
Rules for the precise location of the intersec-
tions of the spectral reflectance curves of meta-
mers are even more difficult to develop than rules
400
300
>
oO
cS
(>) fe)
& 200
Lo
QO
100
ory el
co.
O
400
Fig. 1(3.8.3). Frequency of occurrence of intersections between pairs of metamers as a
function of their locations in the visible spectrum. The solid line with open circles refers to a
set of 990 pairs of metamers generated by one technique; the dashed line with solid dots refers
to a set of 990 pairs of metamers generated by another technique (Ohta and Wyszecki, 1977).
Metameric Color Stimuli 195
concerning the number of such intersections. In
in the fact, no such rules of general validity have been
developed as yet.
In a study by Thornton (1973), the locations
of the intersections are pinpointed at wavelengths
448 + 4 nm, 537 + 3 nm, and 612 + 8 nm. These
locations are claimed to be quite firmly fixed for
virtually all metamers found in practice and are
closely connected with the functioning of the
mechanisms of trichromatic vision.
Thornton’s results have been critically re-
viewed by Ohta and Wyszecki (1977) who were
able to show that the situation concerning the
location of the intersections is far more complex
than Thornton’s work implies. In particular, Ohta
and Wyszecki demonstrated that the locations of
the intersections and the frequency of their occur-
rence at specific locations depend strongly on the
method of generating metamers. For the two sets
of metamers, each containing 990 pairs of rela-
tively smooth spectral reflectance curves gener-
ated on the computer by different methods, Ohta
and Wyszecki obtained the frequency distri-
butions shown in Figure 1(3.8.3). One set of
metamers obtained yields four peak frequencies
in the wavelength bands (430 to 440 nm), (465 to
475 nm), (520 to 530 nm), and (610 to 620 nm).
The other set of metamers has three peak fre-
quencies in the wavelength bands (455 to 475
nm), (520 to 535 nm), and (610 to 620 nm).
The number of peak frequencies found for
each of the two sets of metamers does not imply
that the total number of intersections for each
pair of metamers is found to be always four or
three, respectively. In fact, for both sets of
9
Y es
I I
|
¢ Q
pay 4 I
M i d
bed
500 600
Wavelength X (nm)
196 COLORIMETRY
metamers an abundance of five intersections is
found, and sometimes more than five; but, as
expected, less than three intersections were never
found.
A further analysis of the frequency distri-
bution for the second set of metamers showed
that for three intersections to occur frequently
(about 80% of the 990 pairs) and simultaneously
in three different bands of wavelength, it is neces-
sary that these bands are centered at approxi-
mately 460 nm, 530 nm, and 610 nm, and cover
the rather broad spectral ranges (440 to 485 nm),
(510 to 555 nm), and (590 to 640 nm), respec-
tively.
3.8.4 Counting Metamers
Consider the object-color solid belonging to a
given illuminant and to a given observer (see
Section 3.7). As a rule, a point on the surface of
the solid will correspond to just one object-color
stimulus, that is, to a unique spectral reflectance
curve, and there are no metamers. On the other
hand, for any point within the solid, there are
infinitely many metameric object-color stimuli,
each with its proper spectral reflectance curve. To
study further the variation in the multiplicity of
metamers at different points within the object-
color solid, a more refined method of counting
metamers is needed. Elementary statistical meth-
ods have been applied to this problem by Stiles
and Wyszecki (1962b).
The initial objective in counting metameric
object-color stimuli may be formulated as fol-
lows: How many or what proportion of the ob-
ject-color stimuli that belong to a given collection
are approximately metameric with a given color
stimulus of tristimulus values Xo, Y%, Z)? Ap-
proximately metameric means that the tristimulus
values of the selected object-color stimuli should
lie in a three-dimensional interval (AX AY AZ)
centered on Xp, Y%, Zo, the tolerances AX,
AY, AZ defining the closeness of metamerism
demanded.
Only those collections of object-color stimuli
are considered that contain an indefinitely large
number of object-color stimuli with spectral
reflectance functions p(A) extending over a con-
tinuous wavelength range. Such collections of ob-
ject-color stimuli are called basic collections. For
such collections, the counting problem has a
meaning only if the frequency with which differ-
ent p(A) functions occur in the basic collection
can be specified. For an explicit formulation of
this frequency distribution, it is necessary to have
some systematic scheme for subdividing the set of
p(A) functions (i.e., all functions of A with values
lying in the interval 0 to 1 throughout the visible
spectrum) into a large number of subsets. Given
such a scheme, the frequency distribution of the
basic collection is specified by the probability
that a p(A) function randomly selected from the
basic collection will lie in each of the subsets. Of
the unlimited number of schemes of subdivisions
that might be used, the only ones of interest here
are those in which all the p(A) functions belong-
ing to a particular subset have approximately the
same tristimulus values; that is, are nearly
metameric with respect to a given illuminant
{S(A) dA} and a given set of color-matching
functions [x(A), y(A), Z(A)].
The simplest way of grouping all p(A) func-
tions into suitable subsets is the following. The
visible spectrum is divided into M equal wave-
length intervals of width 6. All functions p(A) are
classified by their mean values in these intervals,
Pp; tO py. Because the weighting functions
S(A)x(A), SCA) (A), S(A)Z(A) used to deter-
mine the tristimulus values X, Y, Z are continu-
ous functions, then, provided 6 is fairly small, the
tristimulus values of any function p(A) are closely
determined by the corresponding values of p, to
Py in the M intervals. These M numbers char-
acterize the p(A) functions.
The ‘tristimulus values “X = X,Y XZ =
X, for any function p(A) are set equal to
M
(k = 1,2,3)
Ay AG > P;4q,;
i=1
[1(3.8.4)]
with
ay, = {S(A)X(A) dd
ay, = JS(A)F(A) dd
ay, = [S(A)z(A) dd
and the normalizing constant
100
 Metameric Color Stimuli 197

where the integrals are taken over the respective
wavelength intervals i.
Adopting this scheme of classification of p(A)
functions, the statistical makeup of any basic
collection is specified by F(p,,...,,,), where
F(p1,---,0m) 401 +++ dows
is the chance that a p(A) function selected at
random from the basic collection will have mean
values in the respective wavelength intervals i,
lying between (p; + dp;) and p;. Obviously,
l l
2 TA Ot
J sf El Basten Dy, oO Pa
Pi 0 Pu—90
[2(3.8.4)]
The very general form of the above func-
tion F(p,,...,P,,) allows for the possibility that
in the basic collection considered, the chance
that a randomly chosen p(A) function has a par-
ticular value p, in the pth interval will depend on
the values in neighboring intervals, (—,~1),
Pr p+1)> P(p—2)» and so forth. In collections of p(A)
curves corresponding to actual reflecting (or
transmitting) objects, some such interdependence
at least over a limited wavelength range must be
operative, because abrupt variations of p(A) of
wide amplitude are not met. However, if 6 is not
too small, the assumption that the variations of
p(A) in the different intervals occur indepen-
dently leads to useful approximations. The inter-
val 6 must be taken just sufficiently large to
warrant the assumption of independence for the
type of p(A) curve included in the basic collec-
tion. As 6 should be kept small for other reasons,
the choice of 6 is necessarily a compromise. No
essentially different conclusions emerge when
more refined methods are used to allow for the
limited sharpness of fluctuation of p(A) curves.
The assumption that the variations of p; of the
ith wavelength interval are independent of the
variations of the reflectances in all other wave-
length intervals implies that the frequency func-
tion splits into the product of M functions, each
depending on one reflectance value only; that is,
A wide range of basic collections can be gener-
ated by adopting different forms for the frequency
functions F;(p;) dp;.
Given the basic collection of spectral reflec-
tance functions as well as its controlling frequency
function F(p,,...,,,), the problem reduces to
deriving the frequency function of the tristimulus
values X,Y, Z. This frequency function de-
termines the proportion of the object-color stimuli
that have X, Y, Z values in any defined region of
the tristimulus space. In particular, the pro-
portion of those object-color stimuli may be
determined that have (approximately) the same
soul
tristimulus values as a given color stimulus speci-
fied by X,Y, Zo; that is, which are (nearly)
metameric with a given color stimulus with re-
spect to a given illuminant and observer.
The required frequency function is
o(X,, xX), X;) dx, dx, dx,
which represents the chance that a member of the
basic collection will have its respective tristimulus
values between (X, + dX,) and X, for k = 1,2,3.
A general form of the Central Limit Theorem
(Cramér, 1937) can be applied to the present
problem. This theorem shows that the tristimulus
vector whose components (Xj, X,, X3) are the
sums of the M random variables £, = ca, jp;,
(i = 1,2,...,M), for example,
M
Ai » g;
=
will approximate for fairly large M to the normal
distribution
O(X,, Xo, X;) dX#dX, dX,
dX, dX, dx, ( Ps]
______ ooo x —
3/2 2 3
(277) |o,;|'/7
[4(3.8.4)]
in which

q gis
F(p1,...,P1) = F\(p;) --- F(a) Py = » a oX'( X;,aT Xn, nC X, iT es)
k=11=1
where
[5(3.8.4)]
fo Flo) dp,=1 — [3(3.8.4)] In the above expression the mean values X,, , of
198 COLORIMETRY

the sums X, are given by
M
] The conditions of calculation were as follows:
Xn k=OD axif pF,(p) dp
[6(3.8.4)]
and o*! (k,/ = 1,2,3) is a matrix determined by
the variances and covariances of these sums X,.
The variances and the covariances are given by
M fe
Ox) — CS > ay.i41,;( = } [7(3.8.4)]
i=]
with
wi 12|['PAC) ip
-|force) ao]|
The variances correspond to k = /, the covari-
ances to k #/. The matrix {o,,} is called the
variance-covariance matrix with the determinant
|o,,|. The matrix {o*'} occurring in Eq. 5(3.8.4)
is the inverse of {o,,}. It has the determinant
1/| o,;|
If one makes the assumption that for every
wavelength interval i, all reflectance values be-
tween zero and unity are equally probable, that is,
validity of the normal distribution for certain
conditions.
{S(A) dA} =CIE standard illuminant C
x(A), y(A), Z(A)=CIE 1931 standard observer
AX =6= 20 nm=constant wavelength interval
M=20=number of wavelength inter-
vals in the visible spectrum
from 370 to 770 nm, inclu-
sive.
Approximately 10,000 sets of tristimulus val-
ues X, Y, Z were computed by using an indepen-
dent random choice of p; in each wavelength
interval for each set. The number of object-color
stimuli with tristimulus values X, Y, Z, contained
within a cubic cell of dimensions AX = AY = AZ
= 5.0 units, was counted as well as the number of
object-color stimuli contained in a “slice” AY =
5.0 units wide, perpendicular to the Y-axis. The
frequency distribution of object-color stimuli ob-
tained in the slice 50.0 < Y < 55.0 with midpoint
Y = 52.5 is illustrated in Figure 1(3.8.4) as an
example. In each cell of the slice the number of
50.0<Y< 55.0

Cp yeaa [8(3.8.4)] 120

one obtains
100
s,=s=1]

fou
Xmk = 2 Oki [9(3.8.4)]
Z
2 M 60

Ox; — ae i ay i 41, j [10(3.8.4)]

i=}
40}
All three tristimulus values (X,) of any
object-color stimulus have finite maximum and 20
minimum values. Thus, for all points in tristimu-
lus space outside the range of these maximum
and minimum values, the normal distribution
20 40 60 80 100
$(X,) of Eq. 4(3.8.4) must obviously be incorrect.
In fact, it deviates appreciably from the true X
distribution as soon as one or more of the three Fig. 1(3.8.4). Slice of the object-color solid cut
differences (X, — X,, ,) exceed a certain limit. perpendicular to the Y-axis. Each cell of the*slice
As the true frequency distribution of the tri- contains the number of object-color stimuli found
stimulus values is extremely difficult to derive, in a Monte Carlo calculation of (approximate)
Monte Carlo calculations were made to check the metamers.
 Metameric Color Stimuli 199

0.8

0.7

0.6

50.0)
0.5 m

=
y an
0.4 is)
S
=
E
=
0.3 Oo 640
,700 nm

0.2

0.1

460
C5 O

TK ene 0.3 GS 0.6 0.7 0.8

Fig. 2(3.8.4). Frequency distribution of (appréximate) metamers found in a Monte Carlo

calculation in a slice through tristimulus space (Y= const = 52.5) is plotted in the CIE 1931
(x, y)-chromaticity diagram.

near metamers found is given. Also shown is the
locus of optimal color stimuli for Y = 50.0.
_ Figure 2(3.8.4) illustrates the same slice but
transformed into the CIE 1931 (x, y)-chromatic-
ity diagram. The number of near metamers found
in each cell of the slice is plotted vertically. The
peak of the frequency distribution is above the
chromaticity point of illuminant C (x, = 0.310,
y. = 0.316). The rapid decrease in the relative
number of near metamers found in cells displaced
from the central cell is clearly evident. At the
boundary of the object-color stimuli for the given
Y-value (= 50.0), the frequency distribution has
zero values.
Table 1(3.8.4) gives the frequency distribution
of all the Y values obtained. It was found that for
values of
1
(oe Ga a 05] Sa ns
the results obtained lay closely on a normal distri-
bution. The number of sets computed was insuffi-
cient to check the distribution near the boundary
200 COLORIMETRY

Table 1(3.8.4) Frequency Distribution of Y-Values Obtained from Monte Carlo Calculations

Midpoint Frequency Cumulative

J Yj Fj Relative Frequency

2: 125 es 0.02
3 17.5 Kl? 0.13
4 22.) 70 0.82
5 2hS 229 3.09
6 32.5 a12 8.16
a 315 1021 18.28
8 42.5 1477 32.91
9 47.5 1729 50.04
10 52.5 1734 67.22
11 a Bs 1477 81.86
12 62.5 997 92.74
13 67.5 538 97.07
14 a} 202 99.07
15 TIS 74 99.80
16 82.5 20° 100.00
10,093

“These values at the tails of the distribution function deviate significantly from the normal distribution.

of the object-color stimuli determined by the opti- This value compared well with the value
mal color stimuli. The effective value of the con-
stant 1/207 may readily be deduced from a plot dhe = 4. Mes
of the cumulative frequency distribution as shown nN

in Figure 3(3.8.4). It was found that

obtained by direct calculation using Eq. 10(3.8.4).
The mean Y,, obtained from the plot agrees al-
50 = 43 ie = most exactly with the mean directly calculated
from Eq. 9(3.8.4), which gives Y,, = 50.0.

3.8.5 Boundaries of Mismatches of Metamers

© wo ee) Perhaps the most important application of a set
wo © of metamers generated with respect to a given
illuminant and observer is to the determination of
© (2) the magnitude of the color mismatches that will
occur when the illuminant and/or the observer is
50 changed. Such determinations have a direct bear-
ing on the assessment of the color-rendering
properties of light sources, the adequacy of
daylight simulators for colorimetry, and the total
difference of two sets of color-matching func-
OF
FREQUENCY
REL.
CUMULATIVE
Y;
tions.
125 225 32.5 42.5 52.5 62.5 72.5 82.5 Our discussion will again center on object-
color stimuli; for a given illuminant and observer,
4 their metamerism is governed entirely by the dif-
Fig. 3(3.8.4). Cumulative relative frequency dis- ferences between the spectral reflectance func-
tribution of Y-values of approximately 10,000 tions of the different objects involved. If p/(A)
object-color stimuli obtained in a Monte Carlo (j = 1,2,...,m) represents such a set of spectral
calculation. reflectance functions, the tristimulus values
 Metameric Color Stimuli 201

X;, Y;, Z; obtained for each object illuminated by The given test illuminant {S’(A) AA} is then
the given illuminant {S(A) 4A} and viewed by used to replace in Eq. 1(3.8.5) the reference illu-
the given observer [x(A), y(A), Z(A)] are the same, minant {S(A) AA}. The resulting tristimulus val-
that is, ues, denoted by X;, Y/, Z;, yield the expected
cloud of points in tristimulus space. These points
are often transformed into the more convenient
X= X= a IE ae, An (x, y, Y)-space where the luminous reflectance Y
is plotted perpendicular to the (x, y)-chromatic-
ity plane. The following transformation is used:
H=Y,= k Me A)SA)I(A) AX
X;
Z)= Z, = PUIG AX IX
Fs /
EYP AZ!
, /

ye
[1(3.8.5)]

with
The calculation of a boundary in (x, y, Y)-
a 100 space that contains a given percentage of all
ASU A) AL points (x;, y/, Y/) of the cloud is done by a
familiar multivariate statistical technique. De-
note:

The geometrical interpretation of metamerism,

as defined by Eq. 1(3.8.5), is that every object-
color stimulus {p/(A)S(A) AA} is represented by
the same point (Xo, Yo, Z,) in tristimulus space.
When the illuminant and/or observer is
where
changed in Eq. 1(3.8.5), metamerism will gener-
ally no longer hold. The original single point
(Xo, Yo, Z)) in tristimulus space will “explode”
into a cloud of points whose size, shape, and
location in tristimulus space depends on the spec-
tral radiant power distribution of the new illumi- The variance-covariance matrix of (x’, y’, Y’) is
nant and the color-matching functions of the new determined by the matrix Q with elements Q,,
observer. given by
Different investigators have studied the prob-
lem of determining the boundary of the cloud of N
points in (X,Y, Z)-tristimulus space or in the 1
0,,=—— > XX, Uh) 1h 203)
related (x, y, Y)-space. Some of these studies will NH
be discussed below. In each case we will confine
our discussions to a change in the illuminant from
From the theory of normal trivariate distribu-
the original {S(A) AA}, the reference illuminant,
tions, it follows that
to a new illuminant {S’(A) AA}, the test illumi-
nant. The observer will not be changed. Calcula-
3
tions involving a change in the observer are, in
principle, similar to those involving a change in y OMX, X,
i=
the illuminant.
(i) Monte Carlo Calculations. The first step is distributed as x7 (chi-square) with three degrees
consists of generating by the Monte Carlo method, of freedom. The quantity Q*’ denotes the ele-
described in Section 3.8.2(iv), a number of p/(A) ments of the matrix Q~! inverse to Q.
functions of metamers with common tristimulus Since x? with three degrees of freedom is less
values X), Yo, Zp for a given illuminant {S(A) AA} than 7.81 in 95% of all cases, it follows that 95%
and given observer [x(A), y(A), Z(A)]. of all points (x;, y/, Y/) will he within (or on) the
202 COLORIMETRY
0.430
Monte Carlo
0.420
y tive calculation is based on a set of metamers
0.410
0.390
0.430 0.440 0.450
x
Fig. 1(3.8.5). Chromaticity points of 100 objects
viewed under CIE standard illuminant A, that when
illuminated by CIE standard illuminant D,, are all
metameric with identical coordinates (x9 = 313,
Yo = 0.329, Y) = 50.0). The spectral reflectance
curves of the 100 objects were generated by a
Monte Carlo method. The ellipse drawn around
the chromaticity points represents the 95% statisti-
cal boundary of the chromaticity mismatches ob-
served under illuminant A. The center of the el-
lipse (cross) coincides with the mean of the 100
calculated chromaticity points and lies close to the
chromaticity point (open circle) of illuminant A.
closed boundary defined by
3
>, OFX,
=H=71 8k, M25)
i
k,l=1
The closed boundary is an ellipsoid in (x, y, Y)-
space with its center at (x’,, y,,, Y,,), the mean of
points (x;, y;, Y;).
Figure 1(3.8.5) illustrates an example. This and
some of the other examples that follow in this
section were calculated by N. Ohta when he was a
post-doctorate fellow with the National Research
Council of Canada in 1975-1976. One hundred
o/(A) functions were generated by the Monte
Carlo technique that, with respect to CIE stan-
dard illuminant D,; and the CIE 1931 standard
observer, provided 100 metamers with the com-
mon coordinates x, = 0.313, yo = 0.329, and
Y) = 50.0. When the illuminant is changed to
CIE standard illuminant A, a cloud of points
(x;, yj,¥/, with 7 = 1 to 100) is obtained. The
chromaticity points (x;, y/) are shown in Figure
Method
1(3.8.5) together with the projection of the 95%-
ellipsoid that provides the statistical boundary of
_ the chromaticity mismatches that would be ob-
served for all object-color stimuli that belong to
the same class as the calculated sample of 100.
(ii) Metameric Black Calculations. An alterna-
generated by the metameric black method de-
scribed in Section 3.8.2(i) (Wyszecki, 1958). The
set of spectral reflectance functions p/(A) that
provides metamers with respect to CIE standard
illuminant D,; and the CIE 1931 standard ob-
server is of the form
p/(A) = po(A) + a 2 ah py, (A)
0.460
[3(3.8.5)]
where p,(A) is an initial spectral reflectance func-
tion that leads to the given coordinates (Xo, yo, Yo)
under Dg, and p, ,(A) (with k = 1 to B) is a set
of B metameric blacks under D,;. The metameric
blacks are combined linearly by assigning random
numbers to the coefficients a/. An overall scaling
factor a/ is helpful to force the resulting p/(A)
function to fulfill the necessary condition
0<p/(A) <1
for all wavelengths A.
Figure 2(3.8.5) illustrates an example of such
calculations. As in the calculations described un-
der 3.8.5(i), 100 p/(A) functions were generated
by the metameric black method leading to 100
(gray) metamers with respect to D,; and the CIE
1931 standard observer with the common coordi-
nates X9 = 0.313, yy = 0.329, and Y) = 50.0. For
the initial spectral reflectance function p)(A) the
nonselective gray object was chosen, that is,
P(A) = 0.5
for all wavelengths A.
The illuminant was then changed from D,; to
A resulting in the cloud of points shown in Figure
2(3.8.5). The statistical 95% boundary of the chro-
maticity mismatches is an ellipse which is very
similar to the one shown in Figure 1(3.8.5).
The basic collection (see Section 3.8.4) of spec-
tral reflectance functions p(A) that governs the
metameric black method differs, in principle, from
 Metameric Color Stimuli 203

the basic collection that governs the Monte Carlo

method. To generate a basic collection by means
Metameric Black Method
( p,(4)=0.5 )
of the metameric black method, an initial pp(A) is
required for every point in (X, Y, Z)-tristimulus
space (or (x, y, Y)-space). There is a great deal of
0.420
freedom in choosing an appropriate p,(A) func-
tion for each point in space, and its effect on
statistical inferences, such as those concerning the
y boundary of chromaticity mismatches discussed
0.410 here, is not fully understood. However, we must
expect the different choices of p (A) lead to
somewhat different 95%-boundary ellipses. The
following numerical examples indicate that widely
0.400 different choices of p,(A) for a given point in
space may have only a relatively small effect on
the size, shape, and location of the 95%-boundary
ellipse in the chromaticity diagram.
0.390 The calculations resulting in Figure 2(3.8.5)
0.430 0.440 0.450 0.460
x were repeated with two initial p (A) functions as
Fig. 2(3.8.5). Chromaticity mismatches and 95% shown in Figure 3(3.8.5) and denoted by p§(A)
statistical boundary derived from 100 object-color and pj*(A). The two functions are metameric
stimuli calculated by the metameric black method. with the original nonselective function p,)(A) =
The object-color stimuli are metamers with regard 0.5, also shown in Figure 3(3.8.5); that is, both
to CIE standard illuminant D,; and mismatch with lead to the same coordinates x) = 0.313, yy =
regard to CIE standard illuminant A. The results 0.329, and Y) = 50.0 under D,;. Figure 4(3.8.5)
agree closely with those shown in Figure 1(3.8.5) compares the resulting 95%-boundary ellipses, one
obtained by a Monte Carlo method. based on pj(A) and the other based on p(*(A),
with the one obtained earlier on the basis of
Po(A) = 0.5. The three ellipses resemble each other
fairly well, but their centers do not coincide ex-
actly and some differences in size are also noted.
The orientations of their major axes are nearly the
same.

Reflectance

400 500 600 TOO

Wavelength (nm)

Fig. 3(3.8.5). Three initial spectral reflectance functions po(A), p9(A), 09 *(A) used in the
calculation of three sets of metameric object-color stimuli by the metameric black method. The
initial functions are themselves metameric with respect to the CIE 1931 standard observer and
CIE standard illuminant D,s.
204 COLORIMETRY

length interval i, all reflectance values p; between

0.430
zero and unity are equally probable and that the
Metameric Black Method
(0) : p, (4) #0.5
variations of p; of the ith interval are independent
of the variations of the reflectance values in all
(1) = p*OA)
other wavelength intervals. These assumptions are
0.420 (2p (3)
those one normally makes in generating spectral
reflectance functions by the Monte Carlo tech-
nique, which leads to jagged step functions p,.
The frequency function then splits into the prod-
y
0.410 uct of M functions, each depending on one reflec-
(O)°x (O) tance value only, and each being equal to unity;
x (1)
(2)x o(1)
thus,

0.400
F( prst- -s0m) = F\(p1) +>: Fu(em)

[4(3.8.5)]
0.390 where

0.430 0.440 0.450 0.460 (i-oF:(p;) = 1 withi=1toM

X

Fig. 4(3.8.5). 95%-boundary ellipses calculated on Given the basic collection of spectral reflec-
the basis of three different sets of object-color tance functions as well as its controlling frequency
stimuli obtained by means of the metameric black function, as defined by Eg. 4(3.8.5), the next step
method. Each set of 100 object-color stimuli was is then to determine the joint frequency function
generated with a different initial spectral reflec- of the tristimulus values. Actually, the problem at
tance function, po(A), p9(A), 09 *(A) shown in hand involves the tristimulus values of the p,;
Figure 3(3.8.5). The resulting 95%-boundary el- functions with respect to two given illuminants (a
lipses deviate from one another in minor ways. reference and a test illuminant) and a given ob-
Note that the boundary based on p,(A) is that server. Thus, it is necessary to determine the
shown earlier in Figure 2(3.8.5). The centers of the frequency distribution of six-stimulus values,
three ellipses are indicated by crosses, and the which is of the form
chromaticity points of the objects defined by
the initial spectral reflectance functions po(A), OU Xho Aa Ke Aa ake oa
po (A), po *(A) are indicated by open circles.
and represents the chance that a member of the
basic collection of spectral reflectance functions
(iii) Deterministic Method. In pursuing further will have its respective six-stimulus values be-
their statistical approach to counting metamers, tween (X, + dX,) and X, for k = 1 to 6.
as described in Section 3.8.4, Stiles and Wyszecki The six-stimulus values, denoted by X, (k = 1:
(1962b) developed a deterministic approach to to 6), are defined by
predict 95%-boundary ellipsoids. The intriguing
M
feature of this method is that it does not require
the actual production of spectral reflectance func- = ” PiX
Kj [5(3.8.5)]
i=1
tions p/(A). Instead, a basic collection of spectral
reflectance function is assumed, in which each
member function is thought of as a set of reflec- where, explicitly,
tance values p,;, with 0 < p; < 1, where i denotes
one of M wavelength intervals of constant width
AX = 6, which divide the visible spectrum uni-
7. _ 100S(A)F(A)AA
5SREY ENO SG
formly. The basic collection is controlled by its
xX) = ¥, nile FR
iBfere
iat SSCA VAM
frequency function F(p,,...,0,,) for which dif-
ferent forms can be adopted. The case studied by o Ce = xe
a = S;
“ 2
100s(ay¥(A)AN
Se ee
Stiles and Wyszecki assumes that for every wave- : he eK OReime)
 Metameric Color Stimuli 205

4 _ _ 100S(A)z(A)AX familiar methods and, moreover, can be con-

X,=Z, X%3=S87=———
: me Sy S(A) (AJAX verted to the lower triangular matrix

%=¥, %.=Si5= 100S’(A) (A) AA PG

#7) 0
pt ES(X) FOQOAA ~
= ~
NN
i) =)

eek i ee _ _ 100S’(A)X(A)AA tah &

eo
eo

ee EE SHCA) (0) oe
ea
=
eS

ler feo, tes. tea. esvorntee

See USA ZLAJAA
TT a la nm S,S’(A) p(A)AA that is needed in the present case. The new six-
stimulus values U, (k = 1 to 6) are then defined
The subscript i denotes the ith wavelength in- by
terval of constant width AA = 6, centered at
M M
wavelength A. S; (or S(A)AA) defines the refer-
ence illuminant and S/ (or S’(A)AA) defines U,.= > PiU, i= 2 t,,X, [8(3.8.5)]
i=1 i=1
the test illuminant. The observer is defined by the
color-matching functions <x;, y;, Z; [or Their joint frequency-distribution function is the
x(A), y(A), Z(A)]. Note that the luminous reflec- normal distribution of the form
tances X, = Y and X,= Y’ of the object-color
stimuli p;S,; and p,S; are given first position in W(U,,...,Us) dU, --- dU,
each set of tristimulus values. The weighting func-
tions x, , have been normalized in the usual ets tne? ete
manner as indicated on the far right of the above
notation table (see also Section 3.3.8). This nor-
cag iegttaen Ay De
malization makes the luminous reflectance of the [9(3.8.5)]
perfect reflecting diffuser, the ideal “white” ob-
ject, given by p; = 1 for all i, equal to 100 for where the quadratic Q, in the exponent is given
both illuminants, S; and S/. by
The detailed procedure for obtaining the joint
frequency-distribution function of the six-stimu-
lus values X, has been outlined in the paper by
ah = 4B) 5 (U, — Uz m) [10(3.8.5)]
k=1
Stiles and Wyszecki (1962b) and only the major
steps are given here bearing on the particular with U, ,,, being the mean six-stimulus values.
basic collection of spectral reflectance functions In the next step, all tristimulus values of those
and its controlling frequency distribution [Eq. object-color stimuli are considered that, governed
4(3.8.5)] used in the present discussion. by the given basic collection of p; functions,
The first step is to replace the x, ,; weighting are metameric with respect to the reference
functions by new weighting functions u, ; derived weighting functions u, ;(kK = 1,2,3). The meta-
by the linear transformation mers have tristimulus values U, (k = 1,2, 3) fall-
ing within a very small volume element
6

tei =Dtu%,,l=]
(k= 106) [6(3.8.5)] (dU,, dU,, dU;) centered on the point corre-
sponding to the tristimulus values U, = U,9 (k
= 1,2,3). In (U,, U5, U;)-space, the metamers dis-
with the transformation matrix {t,,} so chosen tribute according to the following normal distri-
that the new functions are both orthogonal and bution:
normal; that is, they fulfill the conditions
0 4(Us, Us, Us)dU,dUsdU,
3a a = for kes [ [7(3.8.5)]
PAWS 0 fork +l dU,dU;dUg 1
72e-s)
i=]

Provided the x, ; functions are linearly indepen-

~ Qa/i2y? exp(—
7

dent, the transformation {t,,} is determinable by [11(3.8.5)]

206 COLORIMETRY

with the quadratic and the ellipsoid center is located at

3
]
— S (UU. VS (12G.85)] X40 ad X4 mm adh > tas (X/0 i X), a)
k=4 Af f=1

The quadratic Q,_; has the x? distribution with

three degrees of freedom. Thus, a sphere in the DCP eae, tied , 5 faiAy ogee
44 55 [=]
(U,, Us, Us)-space that contains a fraction h of the
metamers with tristimulus values (U,9, Uz,9, U3. 9) hike
is defined by ae » ts)( X).0 =e Xm)
55 /=1
2 2
(U, rs} Us, m) a (U; Sy ty
leg
X60 oe at
2 ] teala4
ig (ug Ue. m) See

—bestsa—
[13(3.8.5)] ae 5 ta (X90 — Xm)
44°55°66 [=]

For h = 95%, the value of x’ is equal to 7.81.

The sphere in (U,, U;, U,)-space corresponds
nay 5 ts)( X19 — Xm)
to an ellipsoid in (X,, X;, X,)-space because the 55/66 [=]
two spaces are related to each other by the linear
transformation given by Eq. 8(3.8.5). The hA-
ellipsoid in (X,, X5, X,)-space has the following PEA
Ly te (X10 — Xm)
66 J=]
equation:
% 2 [17(3.8.5)]
T44( X4 7x Xo,4) a T55( Xs as Xo.5)
with
2 . te
+ 166 ( X¢ a Xo.6)

Xi i DE (/ [2p
+2145(X, i, Xa) Xs = Xo.5)
i=1

+2155 ( X5 al Xo,5)( Xe = Xo 6) [18(3.8.5)]

l The location of the center of the h-ellipsoid as
+2 %64( X6 in Xo,6)(X% ~ Xo,4) a, mee
given by Eq. 17(3.8.5) depends on the given tri-
stimulus values X, 9, X29, X39 of the metamers
[14(3.8.5)]
in (X,, X,, X3)-space, and of course, on the coef-
where ficients of the {t,,} matrix. However, the shape,
orientation, and volume of the ellipsoid are the
6
same whatever the reference tristimulus values
w= Dd tetn [15(3.8.5)]
k=4 X1,0> X2,0) X3,0-
In connection with this at first slightly surpris-
which, explicitly becomes ing result, it should be noted that
TAcs betes re
(a) The h-ellipsoid concerns only the distribu-
tion in (X4, X;, X;)-space of the representative
155 = t55 + t6s
points of the subset of object-color stimuli that
ee ps
are metameric with respect to the given reference
Os [16(3.8.5)] weighting functions; the actual “number” of the
metameric object-color stimuli in the subset de-
5 — tsatss 1 lates
pends strongly on X; 9, X29, X3,9 aS discussed in
T56 — lestee Section 3.8.4.
(b) The normal distribution underlying the
764 — leat whole discussion is only an approximation that

decreases in validity as the distance of the point
(X10, X20, X3,.9, X4, X5, Xe) increases from the
pomt.GXy,,, stg.)
The derivation of a corresponding h-ellipsoid
in (x, y, Y)-space, instead of the (X, Y, Z)-space
as outlined above, is, strictly speaking, not possi-
ble. The transformation of the tristimulus values
X,Y, Z to chromaticity coordinates x, y is non-
linear, which, when applied to the normal joint
frequency-distribution function of the tristimulus
values, leads to an exponent Q in the distribution
that is not of second degree. The contours of
equal probability density of the distribution be-
come nonellipsoidal.
However, the nonellipsoidal surfaces can be
approximated by ellipsoids if one ignores in the
expression for Q all terms of total degree higher
than two. Usually, such an approximation is quite
acceptable.
In practice, the numerical procedure is first to
linearize the given nonlinear transformation of
the X,Y, Z values to x, y,Y values and then
to use the linear approximation to transform
Q,(X, Y, Z) to Q3(x, y, Y) which defines an el-
lipsoid in (x, y, Y)-space. Whereas the original
ellipsoid in (X, Y, Z)-space had the same shape
orientation and volume regardless of the reference
tristimulus values (Xp, Yo, Zo), the new ellipsoid
in (x, y, Y)-space varies in shape, orientation,
and volume as a function of the location of its
center, that is (Xo, ¥%, Yo). For details of the
numerical procedure, see Section 5.4.2.
Figure 5(3.8.5) illustrates in the CIE (x, y)-
chromaticity diagram nine cross sections (Y =
const, ranging from 49.2 to 50.8) of a 95%-
boundary ellipsoid. The input data used to derive
this ellipsoid by means of the deterministic
method described above are the same as those
used to illustrate the Monte Carlo method and
the metameric black method. Thus, the elliptical
envelope shown in Figure 5(3.8.5) can be directly
compared with the ellipses shown in Figures
1(3.8.5), 2(3.8.5), and 4(3.8.5). The 95%-boundary
ellipse obtained by the deterministic method is
very similar to those obtained by the other two
methods, except its length is slightly greater. The
differences that are noted are most likely statisti-
cally not significant.
The spectral reflectance functions that make
up the basic collection in the deterministic method
are “jagged” step functions in the visible spec-
trum. Other basic collections are conceivable, in
particular, those whose members are more repre-
Metameric Color Stimuli 207
Deterministic Method
(jagged p(X) functions)
0.420
0.400
0.390
0.430 0.440 0.450 0.460
x
Fig. 5(3.8.5). Nine cross-sections of 95%-
boundary ellipsoid (Y = const ranging from 49.2
to 50.8) calculated by means of the deterministic
method. The basic collection of spectral reflec-
tance functions consists of “jagged” step func-
tions. The reference illuminant is CIE standard
illuminant D,., the test illuminant is CIE standard
illuminant A. For the CIE 1931 standard observer,
objects that under D¢; are metameric grays (x =
0.313, yo = 0.329, Y) = 50.0), do not match under
illuminant A, but 95% of their (x, y, Y) points fall
within the ellipsoid shown, which is centered at
A(xo = 0.448, yo = 0.407, Yo = 50.0).
sentative of real objects, that is, spectral reflec-
tance functions that are smooth.
In a follow-up paper, Stiles, Wyszecki, and
Ohta (1977) study a basic collection that consists
of frequency-limited spectral reflectance func-
tions that, by an appropriate choice of parame-
ters, can be made quite smooth and representative
of real objects.
The particular set of frequency-limited spec-
tral reflectance functions p(A) adopted as the
basic collection was defined by
p(A)=24+726(A) — [19(.8.5)]
with
RO oy, sin? [a(A — A,)o — in72]
i sai issige
pat ose
pens eof [7(A — Aq) w — ia /2]
[20(3.8.5)]
208 COLORIMETRY
[e) a
Reflectance
os 400 500 600
Wavelength i (nm)
Fig. 6(3.8.5). Example of a frequency-limited
spectral reflectance function p(A) with a limiting
frequency of w = 1/50.
where y; is a random number satisfying the condi-
tion
| IN ce IN a U=0,--1, = 2, etc.)
w is a constant representing the limiting frequency;
Xo is a “zero-adjustment” constant normally cho-
sen to coincide (approximately) with the wave-
length in the middle of the visible spectrum.
The frequency-limited spectral reflectance
function p(A) defined by Eqs. 19(3.8.5) to
21(3.8.5) satisfies the necessary condition
0<p(A) <1
for all wavelengths A. It also ensures that the
frequency spectrum of p(A) with respect to A has
no component at a frequency greater than w
cycles per unit change in A. The choice of w
provides control over the shape and smoothness
of the spectral reflectance curve.
The deterministic method employing the above
basic collection of frequency-limited spectral re-
flectance functions follows statistical procedures
that are very similar to those described above.
For details, the paper by Stiles et al. (1977) is
recommended.
Figure 6(3.8.5) shows an example of a fre-
quency-limited spectral reflectance function p(A)
for w = 1/50. Figure 7(3.8.5) illustrates in the
CIE (x, y)-chromaticity diagram seven cross sec-
tions (Y = const, ranging from 49.4 to 50.6) of
the 95%-boundary ellipsoid that was obtained by
the deterministic method with a basic collection
of frequency-limited spectral reflectance func-
tions (w = 1/50) leaving all other input data the
same as in the case illustrated in Figure 5(3.8.5).
The use of a basic collection of smooth spectral
reflectance functions of the limited-frequency type
with w = 1/50 leads to a 95%-boundary ellipsoid
which is surprisingly similar to the one based on a
basic collection of jagged spectral reflectance
functions.
Calculations with different choices of w lead to
similar 95%-boundary ellipsoids with regard
to orientation and shape, but the volume is found
to vary noticeably with w. The ellipsoid calculated
with w = 1/100 is considerably smaller than that
with w = 1/50.
700
(iv) Linear Programming Method. The Monte
Carlo method, the metameric black method,
and the deterministic method all use statistical
procedures to estimate the boundary of the cloud
of points in (X, Y, Z)-space or in (x, y, Y)-space
that is generated when objects matching under
one illuminant are observed under another illumi-
nant. The statistical procedures lead to boundary
ellipsoids that contain a given percentage (e.g.,
95%) of the original metamers. The 95%-boundary
[21(3.8.5)] ellipsoids that are computed by the different
methods are, in general, very similar in shape,
orientation, and volume. The differences that do
occur are, in the main, caused by differences in
the basic collections of the spectral reflectance
functions that govern the different methods. None
of the methods is capable of deducing the optimal
boundary within which ail of the original
metamers must lie. Such a boundary must exist
[22(3.8.5)] and standard procedures of linear programming
can be used to determine it.
The application of linear programming to the
problem was demonstrated first by Allen (1969)
and developed further by Ohta and Wyszecki
(1975a). A related method is that recently used by
Schmitt (1976).
The numerical procedure of finding the opti-
mal boundary is summarized by the following
equations taken from the paper by Ohta and
Wyszecki (1975a). |
The defining equations of the tristimulus val-
ues X,Y, Z© of all object-color stimuli
{p/S‘?} that are metameric for the CIE 1931
standard observer [x;, y,, Z;] are written in the
form
n iS (Ds
DiS oy,
D"_ piSMy.
YY = 190—=" += [23(3.8.5)]
7-8!” i
n IS (Ms
ZO) = 100 aes Zi
Tre SPY,
 Metameric Color Stimuli 209

0.430
Deterministic Method
(smooth p(X) functions)

0.420

0.410
y

0.400

0.390

0.430 0.440 0.450 0.460

x

Fig. 7(3.8.5). Seven cross-sections of a 95% boundary ellipsoid (Y= const ranging from 49.4
to 50.6) calculated by means of the deterministic method. The basic collection of spectral
reflectance functions consists of frequency-limited functions («= 1/50). The reference illumi-
nant is D,;, the test illuminant is A. For the CIE 1931 standard observer, objects which under
Des are metameric grays (x9 = 0.313, yo = 0.329, Yy = 50.0), do not match under illuminant A,
but 95% of these (x,y,Y) points fall within the ellipsoid shown, which is centered at
A(xq = 0.448, yy = 0.407, Y) = 50.0).

with the constraint boundary. When the points (X,Y, Z) are

(for all i andj)
[24(3.8.5)]
Equations 23(3.8.5) compare with Eq. 1(3.8.5) of
this section except that the subscript i has been
used here to refer to wavelengths A selected at
equal intervals of width AA across the visible
spectrum. If the visible spectrum is taken from
400 to 700 nm and divided uniformly in steps of
AX = 10 nm, i runs from 1| to n = 42. The super-
script j attached to the spectral reflectance func-
tion p/ of the object indicates a particular
transformed to the (x, y, Y)-space, a similar cloud
of points is obtained with a finite boundary in
that space. Because the (x, y, Y)-space is gener-
ally considered the colorimetrically more infor-
mative space, the numerical procedure described
below is confined to finding the optimal boundary
in that space.
Cross sections of the optimal boundary in
(x, y, Y)-space for Y = const = Y, are closed and
convex loci in the (x, y)-chromaticity diagram.
These chromaticity loci are generated by the fol-
lowing equations:
D"_ pJSy,
Y,= 100 i=1Pi LS

member of the set of metamers all of which are yi! Y,

seen under the same illuminant S{”’, the reference Or
illuminant.
When the reference illuminant S$‘ is replaced 30/5] 5,(2 i |by |
by a test illuminant, denoted by S$, the object- | ay, Vie So Se ae
&
aE ae ee
color stimuli {p/S@} now at hand will have tri- - BS Py,
stimulus values X,Y, Z. In tristimulus
space a cloud of points is generated with a finite [25(3.8.5)]
210 COLORIMETRY
In these equations the chromaticity coordinate x
has been assigned the value x,. The maximum
and minimum value of y, when Y = Y, and x =
Xg, can be calculated from
Meme: eRe as
Gers 10027_ pjSz Weiss SPY;
[26(3.8.5)]
with Eqs. 23(3.8.5) to 25(3.8.5) being the con-
straints to be fulfilled simultaneously.
The complete cross section of the optimal
boundary for Y = Y, is obtained by making ap-
propriate changes in the value of x,. Different
cross sections are obtained by changing Y,.
The optimization of Eq. 26(3.8.5) with the
constraints given by Eqs. 23(3.8.5) to 25(3.8.5) is
0.400 0.420 0.440 0.460 0.480
x
Fig. 8(3.8.5). Three cross sections of an optimal
boundary surface (Y = const = Y, = 49.0,
50.0,51.0) calculated by the linear programming
method. The reference illuminant is D,<; the test
illuminant is A. For the CIE 1931 standard ob-
server objects, which under D,; are metameric
grays (X9 = 0.313, yp = 0.329, Y) = 50.0), do not
match under illuminant A, but all of them are
represented by points in (x, y, Y)-space that fall
within the optimal boundary shown. The chro-
maticity point of the nonselective gray is indicated
by A.
referred to as linear programming (for example,
Beveridge and Schechter, 1970).
Figure 8(3.8.5) shows in the (x, y)-chromatic-
ity diagram three cross sections of the optimal
boundary (Y= Y, = 49.0, 50.0, 51.0) that is ob-
tained when the reference illuminant is D,; and
the test illuminant is A. The optimal boundary is
not an ellipsoid and embraces a considerably
larger domain in the (x, y, Y)-space than the
95%-boundary ellipsoids determined by statistical
procedures described earlier. Figure 9(3.8.5) com-
pares the optimal boundary with the 95%-
boundary ellipsoid from Figure 5(3.8.5).
At a point P of the optimal boundary, where y
of Eq. 25(3.8.5) takes on an extreme value under
the five conditions of constraint defined by Eqs.
23(3.8.5) to 25(3.8.5), the spectral reflectance
function p/ that corresponds to that boundary
point has values of either zero or unity at
all wavelengths i except at five of them, at which
values intermediate to zero and unity are found.
Figure 10(3.8.5) illustrates the p/ function that
leads to the point P at the boundary shown in
Figure 8(3.8.5). In this example, the intermediate
reflectance values occur at i = 4, 12, 17, 22, and
SY¢
The shape of the optimal boundary surface
and its volume in the object-color solid depend
optimal boundary
95% PRE?
0.400 0.420 0.440 0.460 0.480
x
Fig. 9(3.8.5). Comparison of optimal boundary
with 95%-boundary ellipsoid, both projected into
CIE(x, y)-chromaticity diagram.
 Metameric Color Stimuli 2iAt

caused by a change in the location of the refer-

ence point, attention is drawn to Figures 11(3.8.5)
and 12(3.8.5). Figure 11(3.8.5) shows the cross
section of the (X, Y, Z)-object-color solid at Y =
50.0 bounded by the locus of optimal object-color
° on
stimuli (see Section 3.7). Inside that boundary are
Reflectance marked three points (p = 1,2,3). Each of these
points represents a set of object-color stimuli
which with respect to illuminant D,; and the CIE
1931 standard observer are metameric. The refer-
i=l 5 10 15 20 25 30 ence tristimulus values for the three sets of
400 500 600 700
Wavelength i (nm)
metamers are given by

Fig. 10(3.8.5). Example of an optimal spectral

reflectance function leading to the chromaticity
point P of Figure 8(3.8.5) located at the optimal
boundary of chromaticity mismatches of metamers.
on the reference tristimulus values X“, Y, Z™
of Eq. 23(3.8.5) of the metamers characterized by
the p/ functions. The examples given in this sec-
tion all refer to reference tristimulus values that
correspond to a reference point in the middle of
the object-color solid. To illustrate the change in
shape and volume of the optimal boundary surface
120
100
80
z 60
40
20
O 20 40 60
x metamers at points of the boundary of the solid.
Fig. 11(3.8.5). Cross section (Y = 50.0) of ob-
ject-color solid that refers to CIE standard illumi-
nant D,; and the CIE 1931 standard observer.
Points p = 1,2,3 represent reference points of
three sets of metamers in the cross section that are
used in the calculation of boundaries of mis-
matches of metamers shown in Figure 12(3.8.5).
0 0e eeY ee Sia 21 50.0
x S100, 8 46¥9 =500, ZS — 70.0
Ay ==80.0, > _YA0"="50Ope Zs" — 80.0
Whereas the first set is represented by a point
near the center of the object-color solid with
respect to illuminant D,, and the CIE 1931 stan-
dard observer, the third set is represented by a
point near the boundary of the object-color solid.
When the reference illuminant D,; is replaced
by the test illuminant A, the three sets of metamers
will be represented by three clouds of points in
three different locations of the object-color solid
with respect to illuminant A and the same ob-
server. The cross sections (Y = 50.0) of the opti-
mal boundaries of these three clouds of points are
shown in Figure 12(3.8.5). The optimal boundary
corresponding to the first set of metamers [ p = 1
in Figure 11(3.8.5)] covers the largest area in the
(X, Z)-plane, whereas the optimal boundary cor-
responding to the third set of metamers (p = 3)
has become quite small.
Obviously, the cloud of points shrinks when
the corresponding reference point of metamers
moves toward the boundary of the object-color
solid. This result corresponds to the relative num-
ber of metamers that can occur at a given refer-
ence point in the object-color solid. As shown in
Section 3.8.5 [see particularly Figure 2(3.8.4)], the
maximum number of metamers occurs in the
80 100 center of the solid; in general, there are no
Figure 12(3.8.5) also shows, for each case, the
cross section of the corresponding 95%-boundary
ellipsoid obtained by the Monte Carlo method
(solid line) and the cross section of the 95%-
boundary ellipsoid obtained by the deterministic
method. The 95%-boundary ellipsoid obtained by
the Monte Carlo method follows the decrease in
212 COLORIMETRY

p=l
x50, ¥("s50, 2/"=50
Monte Carlo 95% — boundary

7
Deterministic
95% — boundary
optimal boundary

pz2 optimal boundary

x ero: 1602504 rage270

Monte Carlo
95% - boundary

Deterministic
95 % — boundary

p=3
(1) (1) (1) ~ — Deterministic
Xz =80,Y; 750,23 =80
95 % - boundary

Monte Carlo OR boundary

95 % — boundary

86 90 94 102 106 110

x

Fig. 12(3.8.5). Cross sections (Y=50.0) of boundaries of mismatches of three sets of

metamers shown in the object-color solid that refers to CIE standard illuminant A and the CIE
1931 standard observer. The upper, center, and lower diagrams contain boundaries correspond-
ing to the first, second, and third set of metamers, identified in Figure 11(3.8.5) by their
respective reference points p = 1,2,3.

size of the optimal boundary, but the 95%-
boundary ellipsoid obtained by the deterministic
method does not. In fact, the latter retains its
shape and size regardless of the location of the
reference point in the object-color solid. This
feature, already indicated in Section 3.8.5(iii), is a
direct consequence of the assumption made in the
deterministic method that the frequency distri-
bution of the tristimulus values is a normal dis-
tribution everywhere within the object-color solid.
Evidently, this assumption cannot be upheld when
the reference point deviates appreciably from the
midpoint of the object-color solid.
3.8.6 Application of Linear Programming to
Miscellaneous Colorimetric Problems
The linear programming technique described in
Section 3.8.5(iv) has been applied to a number of
other colorimetric problems. Some of these are
sketched below to illustrate the power of the
technique.
(i) Designing Illuminants with Special Proper-
ties. Linear programming has been used to
calculate spectral radiant power distributions of
illuminants that, when irradiating a given (limited)

set of objects, render these objects in prescribed
colors (Ohta and Wyszecki, 1976a).
The basic equations needed in this application
are similar to those given earlier in Section
3.8.5(iv). The tristimulus values X,, Y;,Z; of a
given object p/ irradiated by an iiluminant of
spectral radiant power distribution {5*} are given
by
x, — (jeer
Ds DAS
Dads: Y;
>" piSky.
¥,=10—="F™ —11(3.8.6)]
PNY: Jj
Z, —_ 100
>ie:AY
a The p% function
j=19; Ji
where again i refers to wavelengths A selected at
equal intervals of width AX across the visible
spectrum. If the visible spectrum is taken from
400 to 700 nm and divided uniformly in steps of
AX = 10 nm, i runs from | to n = 31. The super-
script j refers to a particular object of a given set
of objects with spectral reflectance functions p/
fulfilling the necessary condition
[2(3.8.6)]
The superscript k refers to a particular illuminant
of a set of illuminants with spectral radiant power
distributions {S*}. Two constraints are imposed
on each illuminant k:
Skips ancrmeilly: [3(3.8.6)]
> sky, = 100
i=1
[4(3.8.6)]
The first constraint is an obvious lower bound
necessary for real illuminants. The second con-
straint is a convenient one and does not affect the
generality of the problem. From Eq. 1(3.8.6) it
follows immediately that any spectral distribution
S/* that is proportional to S*, that is, S/*= aS*,
where a is a constant factor, leads to identical
tristimulus values. The constraint of Eq. 4(3.8.6)
simplifies Eq. 1(3.8.6).
The constraints imposed on the spectral reflec-
tance functions p/ [Eq. 2(3.8.6)] and the spectral
radiant power distributions S* govern, to a con-
siderable extent, the permissible range within
which one can assign specific values to X;, Y;, Z;
Metameric Color Stimuli 213
for each given p/ function to ensure solutions for
Sa
If solutions S* exist, a set of corresponding
tristimulus values X¥, Ys, Zf can be calculated
for the set of object-color stimuli {p2S*} that
defines, in tristimulus space, a domain bounded
by a nonconcave locus containing all possible
points (Xz,F555 Y;,5 Z;). These tristimulus values are
given by
n n
ee » 07'S; Xi» ee » o2S;y;,
a) i=1
n
= > ofS\z; [5(3.8.6)]
=!
characterizes an arbitrarily
selected object with g not necessarily equal to any
j. In most cases, p2 can conveniently be set equal
to unity for all 7.
The domain in the tristimulus space con-
taining all possible points (X;, Y,, ZS), depends
on the input data, particularly the set “of given p/
functions and tristimulus values X;, Y;, Z;. Obvi-
ously, given an appropriate ae of of func-
tions with associated (X,, Y;, Z;)-values and
meeting conditions of linear independence, a
unique solution can be found for S* that would
shrink the domain of solutions to a single point.
However, this special case is only of little practi-
cal interest and will not be pursued here.
The optimal boundary of the domain in
(X, Y, Z)-space, generated by the solution set of
S*, is obtainable by the linear programming tech-
nique as outlined in Section 3.8.5(iv). A useful
variation of the earlier procedure determines the
boundary in the (u,v, Y)-space, where the
(u, v)-coordinates define the CIE 1960 uniform
chromaticity-scale diagram obtained from
4X _ 6Y
ita yo tla 6 X 4 16y a7
u—
[6(3.8.6)]
In 1976, the CIE 1960 (u, v)-UCS diagram has
been replaced by the CIE 1976 (u’, v’)-UCS dia-
gram (see Section 3.3.9), where u’ = u and v’ =
1.5v. However, in the examples given below, the
1960 version has still been used.
In the (u,v, Y)-space the nce boundary
of the domain of points (ug, vs, Y;) of the ob-
ject-color stimuli {p%S*} is then obtained by
214 COLORIMETRY

means of the following equations:
n Rendering Index (Section 3.3.11). Their tristimu-
»; p2Si‘y, = Y,
i=1
E l
o2S*| |1 —— |x, + 15y,
l I
+ 3z,| =0
Uu 1 Jj
i=1 5
5 UsYs Ds PES; X= Un,om
3 ; SOK =
i=1
with the constraints already given by Eqs. 1(3.8.6)
to 4(3.8.6).
By keeping Y, fixed and assigning different
values to u,, one obtains pairs of points of the
cross section es = const = Y,) of the optimal
boundary located at (u,, v,,) and (u,, va,). Dif-
ferent values of Y, lead to different cross sections,
which eventually provide a complete picture of
the three-dimensional domain
points. The coordinates v,, and v,, are minimum
and maximum values, respectively, of the v-chro-
maticity coordinate defined by Eq. 7(3.8.6) and
can be found by the linear programming tech-
nique known as the mutual primal dual method
(Llewellyn, 1964).
EXAMPLE |
The first example has been selected merely to
demonstrate the theory and its basic equations
outlined above. No immediate practical applica-
tions are envisaged for this particular example.
The input data are given in Table 1(3.8.6). We
have selected eight objects (j = 1 to 8) whose
spectral reflectance curves p/ are shown in Figure
1(3.8.6). These objects are identical to those used
in the method of calculating the CIE Color-
lus values X;p,., ¥jp,<> Z;,p,,» When irradiated
by CIE standard illuminant D,;, and viewed by
the CIE 1931 standard colorimetric observer
(X;, ¥;,Z;) are given in Table 1(3.8.6) (Columns
I
2—4). We wish to determine the color gamut in
(u, v, Y)-space of all illuminants {5*} that yield
tristimulus values X;, Y;, Z; for the eight objects
[7(3.8.6)] prescribed in the table (Columns 5-7). We also
wish to obtain the actual spectral radiant power
distribution {S*} of one of the possible illumi-
nants.
Figure 2(3.8.6) illustrates the results. The up-
per part of the figure is a portion of the CIE 1960
UCS diagram showing the chromaticity points
(solid circles) u;, v; of the eight objects (j = 1 to
8) under D,; and the corresponding prescribed
chromaticity points (tips of arrows) of the same
of (uf, vg, Y;) objects under any of the possible illuminants {5/}.
The chromaticity gamut generated by these illu-
minants is very small and located within the small
square next to the chromaticity point for D,;. In
the upper right-hand corner of Figure 2(3.8.6),
this gamut is shown again but much enlarged.
The boundary of this gamut is the optimal
boundary. No illuminants S* are possible that
have chromaticities outside this boundary. The
illuminant 57“ given in the lower part of Figure
2(3.8.6) has the chromaticity point M on the
boundary. It presents an example of the set of
possible solutions.
The chromaticity gamut [shown as the insert in
the upper part of Figure 2(3.8.6)] generated by the
possible illuminants S* is completely described in
the (u,v) diagram. Its third dimension is zero

Table 1(3.8.6) Input Data for Numerical Example Number 1

Object Tristimulus Values Under D,; Prescribed Tristimulus Values

J Xi,D6s YiDes Zi,Des Xj Y; Z;

l a28 29.77 24.51 30.40 29035 25s3

ys 27.48 28.91 14.92 feb 2a 91lot 14.85
3 Pike N 30.46 9.90 23.01 Sie 9.63
4 20.44 29.50 2425 ON eee 29.89 20.02
5 24.97 30.85 40.31 a) 30.89 40.33
6 28.19 29.81 57.78 27.88 29.01 59.32
7 fe We Janpea) 53.07 31.44 28.37 56.21
8 37.60 31.36 45.37 fee P 29.67 48.04
 Reflectance
ce)
oO
\e)ie)
‘NN
Oo
oo
OnW
pu
—N

!
< WAND
on
ao
N

6 ee ee ee Se ee ee
i=l GS 10 15 20 25 30
=400 500 600; 7OOnm
Wavelength

Fig. 1(3.8.6). Spectral reflectance curves p/(j =1 to 8 ) of eight objects used in numerical
examples.

Radiant
Power

Wavelength

Fig. 2(3.8.6). Upper part: Portion of CIE 1960 UCS diagram. Solid circles represent
chromaticity points of the eight objects defined in Figure 1(3.8.6) under reference illuminant
Dgs- Tips of arrows represent prescribed chromaticity points of the same eight objects under
new illuminants. The new illuminants have chromaticities generating a gamut falling into the
small square. In the upper right-hand corner, the exact boundary of this gamut is shown much
enlarged. Lower part: Example of spectral radiant power distribution S;” of illuminant that
fulfills conditions illustrated in upper part. Its chromaticity point M lies on the boundary shown
in the insert above (Ohta and Wyszecki, 1976a).

215
216 COLORIMETRY

because of the fact that we have chosen p£ to be
equal to unity for all i.
EXAMPLE 2
The second example, as well as those that follow,
are special cases of the first example with practi-
cal applications in mind.
Again we use the eight objects p/ (j = 1 to 8)
shown in Figure 1(3.8.6) with tristimulus values
Xj, Des? 1),Des? Zj,Des 28 Biven in Table 1(3.8.6)
(columns 2-4). However, this time we make the
prescribed tristimulus values (X,, Y;, Z;) identical
to (Xj D65> Vj,D652 27, Des) _Puat is, we look for
illuminants {5} that render the eight objects in
the same colors as the reference illuminant D,;.
Because the eight objects are those recommended
by the CIE for determining the CIE general
color-rendering index of light sources, the illumi-
nants S* we wish to find have, by design, a
general color-rendering index R, = 100; that is,
they have “perfect” color-rendering properties in
accordance with the CIE method.
The chromaticity gamut generated by these
illuminants {5*} for any given object p% is very
small indeed, and nearly of negligible size. The
chromaticity coordinate u ranges from 0.19795 to
0.19770 and v from 0.31231 to 0.31233. On the
other hand, the spectral radiant power distribu-
tions {S*} differ greatly from one another. These
results are in agreement with those by Nayatani
et al. (1971) who studied the same case but ap-
proached its solution differently.
Figure 3(3.8.6) gives two examples of illumi-
nants {S*} (k =1 and 2) with a CIE general
color-rendering index R,= 100. The spectral
radiant power distribution of D,; is shown for
comparison purposes. The (u, v)-coordinates of
the three illuminants are
Up,, = 0.19768, vp, = 0.31232
dp 2|ON9TIO Moeap=0.31231
Up=a = 0.19767, vy= = 0.31232

rr Yy

7: Zj
::o jy
Z
Z *
Z
Z

®
:
FS
j
Z

: yy
a

s
:r lJ
i Z
4
ra Wavelength
@
3
(e}

Taio )

ee YU CV AMAL ALA ALALY A 1 )V’.’!Y)UL:

i= | 5 10 5 20 25 30
d= 400 500 600 700nm
Wavelength

Fig. 3(3.8.6). Two examples of illuminants (5; and S) that have perfect color-rendering
- properties (R = 100) in accordance with the CIE method with respect to reference illuminant
Dg; shown at the bottom.

Fig. 4(3.8.6). Chromaticity gamuts in the CIE 1960 UCS diagram illustrating magnitude of
optimal mismatches of metameric grays when the illuminant D,; is replaced in turn by the
illuminants S$} and S$? shown in Figure 3(3.8.6) (Ohta and Wyszecki, 1976a).
When the two illuminants S} and S? are used
to replace illuminant D,; in object-color stimuli
that are metameric under D,;, quite large chro-
maticity mismatches for these stimuli can be
found. Figure 4(3.8.6) shows the chromaticity
gamut of mismatched metameric grays (Y = 10).
The optimal boundary of this gamut was obtained
by the method described in Section 3.8.5(iv).
The large size of the gamut is a confirmation of
what has already been known; that is, that illumi-
nants having “perfect” CIE color-rendering prop-
erties (R, = 100) do not necessarily also have
good color-matching properties (Wyszecki, 1970;
Nayatani et al., 1972a,b; Ohta and Wyszecki,
1975a).
EXAMPLE 3
In our third example, we choose again the eight
objects p/ shown in Figure 1(3.8.6) to which we
add a ninth object whose spectral reflectance p? is
unity for all i. The prescribed tristimulus values
for the ninth object when irradiated by {S*} are
identical to the tristimulus values of that object
when it is irradiated by the reference illuminant
D,;. The prescribed tristimulus values for the
eight objects p/ (j = 1 to 8) under {S*} are to
correspond to higher excitation purities. In other
words, we are looking for illuminants {S*} that
have the same chromaticity as D,;, but that in-
crease the excitation purity of each of the eight
given object-color stimuli. Excitation purity corre-
Metameric Color Stimuli oA |
0.4 0.6
lates roughly with the perception of saturation
(Section 3.4). Thus, in replacing D,, by any of the
illuminants {S*} we expect to perceive the
object-color stimuli with increased saturation of
certain fixed amounts. For some object-color
stimuli, such as those involving human complex-
ion and certain foods, an increase in the satura-
tion has a pleasing effect on the observer and the
object is considered to have a preferred color
appearance. Illuminants that generate pleasing or
preferred object-color perceptions are said to have
a high “flattery index” (Judd, 1967b).
The desired simultaneous increases in the exci-
tation purities of the eight object-color stimuli
can be only of very modest magnitudes. Figure
5(3.8.6) shows what we were able to obtain. The
increase in the excitation purity of objects j = 6
and 8 had to be kept negligibly small to obtain
small to moderate increases for the remaining six.
Note that in the illustration the arrow lengths
represent five times the actual increases in excita-
tion purities.
Figure 6(3.8.6) shows an example of {S*} that
produces the above increases in excitation purities
for the eight objects, and which has chromaticities
identical to D,;. The spectral radiant power dis-
tribution consists of essentially five narrow bands
of wavelength centered at 400, 455, 525, 610, and
700 nm, approximately.
When we reduce the number of objects p/ for
which prescribed tristimulus values are to be pro-
duced, the increases in excitation purity can be
218 COLORIMETRY

0.38

0.34 |- =

Vv —

Vv
O32 \—

0.30

0.30 |e ore | | | it) Ps |e

0.18 0.20 0.22 0.24 0.26

0.26
0.12 0.16 0.20 0.24
u showing maximum possible changes in chromatic-
Fig. 5(3.8.6). Portion of CIE 1960 UCS diagram
showing chromaticity points of eight objects under
Ds (solid circles) and under new illuminants SK
(tips of arrows). The new illuminants have been
designed to shift the original chromaticity points to
higher excitation purities. The lengths of the arrows
are enlarged five times. The chromaticity points of
the illuminants are all coincident in the center
(open circle).
Fig. 7(3.8.6). Portion of CIE 1960 UCS diagram
ity of Caucasian complexion. The point Cy repre-
sents the chromaticity of Caucasian complexion
under D,;. The points C, and C, represent the
maximum and minimum excitation purities of
Caucasian complexion that can be obtained under
new illuminants (S;) and (57) which are il-
lustrated in Figure 8(3.8.6). These illuminants have
the same chromaticity as Dg.

considerably larger, because fewer constraints
have to be satisfied in the calculations. Figure
7(3.8.6) shows the result of calculations involving
only one object, that of Caucasian complexion.
The locus in Figure 7(3.8.6) shows the chromatic-
ity change of Caucasian complexion when irradia-
ted by illuminants {S*}. These illuminants have
the same chromaticities as D,;, and render the
Caucasian complexion with the same luminance
factor (Y = 57.13). The particular illuminants
{S!} and {57}, whose chromaticity points are C,
and C, on the boundary, respectively increase and
decrease the excitation purity of the Caucasian
complexion. Thus, from the point of view of
generating a preferred color appearance, we would
choose the illuminant {S/}. It increases the exci-

Power
Radiant

Wavelength

Fig. 6(3.8.6). Example of spectral radiant power distribution (.S*) of illuminants that increases
the excitation purities of eight objects as shown in Figure 5(3.8.6).
 Metameric Color Stimuli 219

Fluorescent lamps with properties resembling

those discussed in Examples | to 3 above, are
now commercially available. A typical spectral
te)
Shas radiant power distribution of such a fluorescent
lamp is shown in Figure 9(3.8.6). Such lamps can
72) =
=

be useful when the special color-rendering effects

are indeed desired in a particular application.
However, in many other applications, the color-
rendering properties of the same lamps may be
Radiant
Power quite unacceptable, for example, in situations
where objects must be rendered in closely the
same colors as they would appear under natural
daylight.
AAAI
MAA
7

(ii) Color Changes Caused by Specified Changes

NG
Wavelength
Fig. 8(3.8.6). Two illuminants S} and S$? that
give the maximum and minimum excitation puri-
ties [(C, and C, in Figure 7(3.8.6)] for Caucasian
complexion.
tation purity of the Caucasian complexion by a
factor of 1.44.
Figure 8(3.8.6) shows the illuminants {5S}}
(solid line) and {57} (dotted line) that provide
extreme chromaticities C, and C, for Caucasian
complexion. The illuminant that increases the
excitation purity of Caucasian complexion has a
spectral radiant power distribution {5/} consist-
ing essentially of three bands of wavelength
centered approximately at 480, 565, and 700 nm.
in the Illuminant. Linear programming can also
be used to calculate the optimal boundaries of
color changes that occur when the spectral radiant
power distribution of a given test illuminant is
permitted to deviate from the given reference
illuminant by specified amounts. The problem
has a direct bearing on the problem of specifying
tolerances for illuminants relative to a given refer-
ence illuminant (Section 3.3.5 and British Stan-
dards Institution, 1967).
Suppose, for example, the reference illuminant
is CIE standard illuminant D,,; and all those test
illuminants are acceptable D,,-simulators whose
spectral radiant power distributions do not de-
viate from the reference distribution by more
than a specified percentage value at any wave-
length A within the visible spectrum. Figure
10(3.8.6) illustrates an example where S(A) is that

& 100 =
4
°
a
c
a4
no]
o
Power
Radiant
x 50 065
AS(A)=-15%

400 500 600 700 a het wheSo LE NOE ameli Ne en ee La en

Wavelength dX (nm) 400 500 600 700
Wavelength A (nm)
Fig. 9(3.8.6). Example of relative spectral radiant
power distribution of commercially available Fig. 10(3.8.6). Relative spectral radiant power
“cool-white” fluorescent lamp providing special ef- distribution of CIE standard illuminant Dg, (solid
fects resembling those calculated in Examples 1 to line) with upper bound AS(A) = 15% and lower
3 [Figures 2(3.8.6) to 8(3.8.6)]. bound AS(A) = — 15% (broken lines).
220 COLORIMETRY

SOT asin Hliw fae

0.360} |

0.340 a
M:AS(\)=+10%
Power
Radiant
y ‘|
0.320 4
(2 ee ee
400 500 600 700
Wavelength d (nm)

0.300
AS WItI5S% _—i|

| ye es eens|
0.280 0.300 0.320 0.340
X

Fig. 11(3.8.6). Optimal boundaries of (x, y)-

chromaticity of acceptable illuminants $’(A )which m:AS(A)=+10%
deviate from the standard illuminant S(A) by Radiant
Power

AS(A) = £5%, +10%, and +15%.

of D,; and the range 400 500 600 700

Wavelength d (nm)

—AS(A) < S(A) < +AS(A) [8(3.8.6)]
with AS(A) = 15% of S(A), defines acceptable
D,s5-simulators.
In addition to the limiting spectral range
defined by Eq. 8(3.8.6), it is often also required
that the chromaticity points of the acceptable test
illuminants fall within a specified chromaticity
boundary centered at the chromaticity point of
S(A). However, the relation between the limiting
spectral range and the chromaticity boundary is
not immediately obvious, but linear programming
techniques can be used to establish it.
Figure 11(3.8.6) shows the results of applying
a linear programming technique, similar to the
one used above, to the problem at hand (Ohta
and Wyszecki, 1976b). The optimal (x, y)-
chromaticity boundaries are shown of acceptable
D,;-simulators with spectral radiant power distri-
butions that deviate from that of the reference
Dos by +5%, £10%, and +15%, respectively.
These optimal boundaries are nonelliptical loci
centered approximately at the chromaticity point
of Dgs.
For the specific chromaticity points M and m
in Figure 11(3.8.6), lying on the boundary corre-
sponding to AS(A) = 10%, the illuminants $;,(A)
Fig. 12(3.8.6). Relative spectral radiant power
distributions of illuminants giving upper chromatic-
ity limit M and lower chromaticity limit m in
Figure 11(3.8.6). The two distributions (heavy
broken lines) coincide partly with the upper bound
and partly with the lower bound [A.S(A) = +10%]
(thin broken lines) within which D,.-simulators are
allowed to vary.
and S’ (A) are obtained, respectively, as shown in
Figure 12(3.8.6). Note that the spectral radiant
power distributions coincide partly with the up-
per bound [AS(A) = +10%] and partly with the
lower bound [AS(A) = —10%] of the specified
range.
(iii) Other Applications. Other applications of
linear programming to colorimetric problems have
been noted in the literature.
Allen (1966b, 1969) and Ohta and Wyszecki
(1975a) draw attention to the possibility of calcu-
lating by a linear programming technique, the loci
of optimal object-color stimuli (Section 3.7). The
corners of these loci, usually difficult to pinpoint
by traditional methods, can be determined rather
easily with high precision by the linear program-
ming technique.
 Colorant Formulation 221

Ohta and Wyszecki (1975b) have used the then belong to the upward- instead of the down-
technique to estimate the extreme errors of ward-proceeding flux. The quantity K is the frac-
numerical integrations that can occur in colori- tion of the downward-proceeding flux lost by
metric calculations when wavelength intervals AX absorption in the elementary layer, and the quan-
of different width are employed. tity S is the fraction lost by having its direction
Ohta and Wyszecki (1975d) have applied the reversed. The quantities K and S are called the
technique also to a study of the colorimetric absorption coefficient and the scattering coefficient,
significance of the mercury-emission lines in fluo- respectively, of the material forming the colorant
rescent lamps. layer. It must be noted here that the absorption
and scattering coefficients are being used here in
a special sense and are not identical with the
3.9 COLORANT FORMULATION normally defined coefficients for a parallel beam
traversing a material medium.
An important industrial problem is how to pre- During its passage through the same elemen-
dict the proportions of the three or more col- tary layer, the upward-proceeding flux j is reduced
orants (dyes or pigments) required to produce a by the amounts Kj dx and Sj dx by absorption
color match for an object showing the same color and scattering in exactly the same way as is the
as a given reference object or standard. The prob- downward-proceeding flux. As the flux Si dx,
lem is a difficult one because the light-scattering reversed in direction by scattering from the down-
properties of colorant layers deposited on a sub- ward-proceeding flux, is added to the upward-
strate such as paper or textile materials are dif- proceeding flux, the total changes dj of the
ficult to describe in quantitative terms. upward-proceeding flux becomes
There are several theories of varying complexi-
ties designed to explain the light-scattering prop- Cia (iS 4K) dx Si du 1103.9) |
erties of colorant layers. Among these theories, and, similarly, the total changes di of the down-
perhaps the simplest one is the Kubelka—Munk ward-proceeding flux becomes
theory which succeeds reasonably well in a num-
ber of applications and provides a quantitative sodh= si GSiotek )idortSy-dxdi[2(3-9)]
description of the properties of light-scattering
Different algebraic signs appear in front of the
colorant layers, embracing their reflectance, trans-
differentials dj and di in the above equations,
mittance, absorptance, hiding power, and other
because upward-proceeding flux and distance x
characteristics.
are measured in the same direction, whereas
The theoretical model of Kubelka and Munk
downward-proceeding flux and x are measured in
(1931; see also Judd and Wyszecki, 1975) assumes
opposite ‘directions. The thickness dx of the ele-
that the colorant layer can be divided into a large
mentary layer refers to an increase of X, where
number of elementary layers, with boundaries
X = 0 corresponds to the unilluminated side of
parallel to those of the complete layer, which have
the colorant layers. Figure 1(3.9) is a cross section
identical optical properties. The thickness dx of
through a colorant layer.
an elementary layer is small compared with the
thickness X of the colorant layer as a whole, but
is large compared with the diameters of the pig-
ment particles comprising the layer. Thus, the
actions of the individual particles need not be
considered but only their average effect in pro-
ducing an elementary layer of particular absorb-
ing and scattering properties. The model further
envisages just two completely diffuse light fluxes,
one proceeding downward throughout the layer,
the other simultaneously proceeding upward. The
downward-proceeding flux i in its passage through
an elementary layer is decreased by absorption by
an amount Ki dx, and is also decreased by an
amount (— Si dx) by scattering which reverses the Fig. 1(3.9). Schematic diagram of a cross section
direction of some of the light rays. These rays through a colorant layer.
Doz COLORIMETRY
The above differential equations can be in-
tegrated to give explicit expressions for the reflec-
tance R and for other properties of the colorant
layer. The formulae reproduced in Table 1(3.9)
were obtained by Kubelka and Munk (1931),
Kubelka (1948) and other workers. They form the
main part of a wider collection of Kubelka—Munk
formulae assembled by Judd and Wyszecki (1975)
who also go into details of a number of practical
applications of these formulae.
The Kubelka—Munk theory is the most fre-
quently used in predicting color matches. The
literature on the use of the Kubelka—Munk the-
ory and derivatives of it is vast. For the reader
who wishes to study the subject more deeply, we
have selected the following references: Allen
(1966a, 1971), Beasly et al. (1967), Stearns (1969),
Strocka (1970), Mudgett and Richards (1971,
1972), Billmeyer and Abrams (1973), Billmeyer
and Richards (1973), Gall (1973), Judd and
Wyszecki (1975), Kuehni (1975).
Gall (1973) has described a typical procedure
widely followed in industry to make adequate
color-match predictions on the basis of the
Kubelka—Munk theory (see also Judd and
Wyszecki, 1975). A great deal of technical skill is
required concerning the preparation and calibra-
tion of the component colorants, without which
adequate color-match predictions cannot be made.
Furthermore, certain additional assumptions must
be made with regard to the absorbing and scatter-
ing properties of the component colorants. In
particular, sometimes it can be assumed that ad-
ditivity of the scattering and absorption coeffi-
cients of the component colorants holds and that
both absorption and scattering increases in direct
proportion to the concentration of the component
colorants. These assumptions lead to greatly sim-
plified computational procedures, but unfor-
tunately, they are not applicable to many other
practical situations. In general, it is found that the
K/S ratio of a component colorant is a nonlinear
function of concentration, and to take this into
account, more elaborate calibration data must be
inserted into the computation procedure to obtain
adequate color-match predictions.
3.10 SPECIFICATION OF COLOR TOLERANCES
The setting-up of color tolerances is an important
colorimetric task in industrial applications. The
main methods used in specifying color tolerances
(i) by provision of special material color stan-
dards with which the product is compared visu-
ally;
(ii) as in (i), but using as color standards the
paint chips of one of the available color atlases
(see Section 6.6),
(ii) by assigning acceptable values for quanti-
ties computed from measurements made on the
actual product using visual or photoelectric col-
orimeters, or spectrophotometers.
In method (i), the special material color stan-
dards, preferably made in the same material as
the product, reproduce the desired color and the
limiting colors that are considered just acceptable.
The standards for limit colors can be chosen so
that two of them show the tolerable variation in
hue; two others, the tolerable variation in satura-
tion; and a further two, the tolerable variation in
lightness, making six limit standards in addition
to the central standard.
In method (ii), which is a variation of method
(i), the required material standards are chosen
directly from the chips of a color atlas represent-
ing a systematic organization of the object-color
space such as the Munsell Book of Color (Section
6.6.1). .
It is sometimes useful to increase the number
of limit standards to define the tolerance range
more precisely. With twelve chips grouped around
the central standard color chip, it may be ad-
vantageous to aim at uniform tolerances in all
directions and to adopt an arrangement corre-
sponding to the cubo-octahedron of Fig. 1(6.5.1)
(Wyszecki, 1954; 1955; Richter, 1955). The color
atlas of the OSA Uniform Color Scales (Section.
6.6.4) provides an excellent collection of color
standards arranged in that way. However, in other
applications, it is desirable to have very strict
tolerances in some directions, usually hue and
saturation, and less strict tolerances in other di-
rections, usually lightness.
In method (iii), the CIE system of color speci-
fication is commonly used. Measurements made
on the chosen material standard yield CIE
tristimulus values and CIE chromaticity coordi-
nates, and the acceptable range of these quantities
of materials made to reproduce the standard color
can be laid down in terms of these quantities. An
important example of such a tolerance specifica-

Fig. 1(3.10). CIE 1931 (x, y)-chromaticity diagram showing CIE recommended chromaticity
domains for signal lights (CIE 1975).
tion is given in Figure 1(3.10) which shows in the
CIE 1931 (x, y)-chromaticity diagram the accep-
table chromaticity domains recommended by the
CIE (1975) for signal lights for all forms of trans-
port and other uses. The chromaticity domains
are set up for red, yellow, green, blue, violet, and
white signals. Blue and violet signals are rarely
used for long-range signaling, because often
blue(or violet)-signal glassware transmits some of
the long-wave (red) radiant power of the lamp.
Because of the chromatic aberration of the eye
(Section 2.4.3), the blue (or violet) signal then
appears as a red dot surrounded by out of focus
blue (or violet). For this reason, the use of purple
signals is also avoided for long-range signaling.
The measured CIE tristimulus values or CIE
chromaticity coordinates can be entered into one
of the two CIE color-difference formulae (Section
3.3.9) and a value of AE can be specified to
Specification of Color Tolerances 223
define a color tolerance domain centered at the
color of the given standard. This approach of
setting color tolerances is frequently used, partic-
ularly when the tolerances are to be very small,
that is, of the order of one or two just-perceptible
color differences.
Color-difference formulae are derived to pre-
dict perceived color differences between pairs of
color stimuli. Yet, in industry, it is often required
to specify the acceptability (or objectionableness)
of color variations between a given standard and
its reproduction. Whereas the perceptibility of a
color difference is strictly a visual judgment, the
acceptability can be biased by considerations in-
volving the intended application of the product
and its price. Nonetheless, color-difference for-
mulae can be used to some advantage even in
such complex situations (Judd and Wyszecki,
1975).
224 COLORIMETRY
3.11 DISTRIBUTION TEMPERATURE, COLOR
TEMPERATURE, AND CORRELATED
COLOR TEMPERATURE
Distribution temperature, color temperature, and
correlated color temperature are terms used to
describe certain properties of light sources.
The distribution temperature of a source is de-
fined as the absolute temperature (in kelvin K) of
the blackbody radiator for which the spectral
radiant power distribution, at every wavelength in
the visible spectrum, is proportional (or ap-
proximately so) to that of the source considered.
It follows that both spectral radiant power distri-
butions provide the same chromaticity coordi-
nates.
The term color temperature is applied to highly
selective radiators, such as electric discharge
lamps, when the light of this radiator has the
same (or nearly the same) chromaticity coordi-
nates as a blackbody radiator at a certain temper-
ature; this temperature is then called the color
temperature of the selective radiator.
The above distinction between distribution and
color temperatures has not been consistently ob-
served in the literature. In many publications
(especially American) the term color temperature
is used to cover both cases, whereas, in some
British work, color temperature is reserved for
sources that have a distribution temperature as
here defined.
The relative spectral radiant power distribu-
tions of tungsten filament sources approximate
those of blackbody radiators at specific absolute
temperatures. For these sources, it is irrelevant
whether the term color temperature or distribu-
tion temperature is used.
It is of interest, however, to note that the true
filament temperature of the source is not the
same as the color temperature when the tungsten
source and the blackbody radiator emit light of
the same chromaticity. This is because the emis-
sivity of the tungsten filament varies with
wavelength. The filament temperature is always
somewhat lower than the color temperature; for
the range of color temperatures around 2850 K,
the filament temperature is approximately 40 K
lower.
For tungsten filament lamps, there is a simple
empirical relationship between the color tempera-
ture 7, and the voltage V applied to the lamp.
According to Wensel, Judd, and Roeser (1934),
the relation takes the form
TL =a; Has [1(3.11)]
where a, and a, are constants depending to some
extent on the lamp. Typical values are a, = 620
and a, = 245. This formula is used in the first
instance to interpolate and extrapolate a color
temperature—voltage scale, which has been estab-
lished experimentally for a given lamp at two
points.
The relationship between the color tempera-
ture T, and the logarithm of the luminous power
P of a tungsten lamp is given approximately by
log P= ay
_ 11350 [2(3.11)]
FE €
where dy is a constant, the value of which de-
pends on the size of the filament (Weaver and
Hussong, 1939).
Approximate relations between luminous pow-
er P, potential E, and current J have also been
suggested (Walsh, 1958). These take the general
form
2 dP dE
P= mE OT tpiae W epeea
ieRN OE (on aPy vu dl,
ies leak Fer [3(3.11)]
Values for the exponents 1, and n, vary with the
type of lamp and its luminous efficacy; but, when
using the formulae for rough estimates, the values
n, = 3.5 and n, = 6.5 are satisfactory. A practical
example is as follows: the voltage across a lamp
may change from 90 to 92 volts; the correspond-
ing change in luminous power is then computed
as dP/P = 3.5(2/90) = 0.078, which corresponds
to a change of 7.8%.
The color temperature T.. is often expressed on
a reciprocal scale (10°/T,). In the past, the unit
on this scale has been “microreciprocal degree”
(urd) or, in short, “mired.” However, in accor-
dance with present terminology, the unit is now
called “reciprocal megakelvin” and denoted by
MK! (= 10° K~'). As pointed out by Priest
(1933), the use of reciprocal color temperature
instead of color temperature itself may be an
advantage in at least two important cases (see
also Judd, 1933a). First, a given small interval in
reciprocal color temperature is approximately
equally perceptible as a difference in chromatic-
ity, regardless of the color temperature; for exam-
ple, a difference of 100 K at 20,000 K is about
equally perceptible, as a difference in chromatic-
ity, as 1 K at 2000 K. Second, a color temperature
conversion filter (see Section 1.3.5) changes the
 Distribution Temperature, Color Temperature, and Correlated Color Temperature 225

hia ie FE Hee WUE? es a

HY ==Ly Wy ye} |+
hl} a 4
Ak Hy WH) WLU SIUCGLOCC ES
| iH ie iit
cH]
NIMES
HiWEE A,yy foo

; A= A i HiTie WUUUY,LGGUY)Gh Seg

aa HTpeeHY WUUGD Ya.LV
| MMMM OIAGES
H+ Hee?ata VNU Wy Y
Ye Hd
y i LAT My MELA pega
y
af +N \ tle AT Bee
=< \ A |
Sas a \| NE o Hie |
0.300 DERN A N WV AWE i] HE B Be bak “1
i NSN \ VX WHE, eeBar it e mie
NAAN IN an NAV if adie

SSE
SSN
SISA NN
AVAVU ITT 35,0 | |
TE Oeaed ee

RNs nn sisal alae

a | rfEEE
fii He i a FEEEEE
0.200 0.300 0. He 0.500 0.600

Fig. 1(3.11). CIE 1931 (x, y )-chromaticity Fr showing isotemperature lines as computed
by Kelly (1963).

reciprocal color temperature by nearly the same
amount anywhere on the scale.
The term correlated color temperature is intro-
duced when the chromaticity of a selective radia-
tor (e.g., a fluorescent lamp) is not exactly equal
to any of the chromaticities of a blackbody radia-
tor. The correlated color temperature is defined as
the temperature of the blackbody radiator whose
perceived color most closely resembles that of the
given selective radiator at the same brightness
and under specified viewing conditions.
Judd (1936b) was the first to propose the term
and compute so-called isotemperature lines for the
evaluation of correlated color temperatures for
selective radiators. If chromaticities are plotted in
a uniform-chromaticity-scale diagram (see Section
6.4), isotemperature lines are short straight lines
crossing the Planckian locus perpendicularly. In
the CIE 1931 (x, y)-chromaticity diagram these
lines are, of course, no longer perpendicular to
the Planckian locus; nevertheless, they facilitate
the estimation of the correlated color temperature
for any selective radiator with a chromaticity
point somewhat off the Planckian locus.
Figure 1(3.11) shows Judd’s isotemperature
lines in the CIE 1931 (x, y)-chromaticity diagram
as computed by Kelly (1963) using the then valid
value of Planck’s radiation constant c, = 1.4380
x 10°-* m- K and the CIE 1960 (u, v)-diagram
as the uniform-chromaticity-scale diagram.
Both Planck’s radiation constant and the
uniform-chromaticity-scale diagram have been
changed since that time. The present value of c,
equals 1.4388 X 10°-* m- K (see Section 1.2.2),
and the CIE 1960 (u, v)-diagram has been super-
seded by the CIE 1976 (u’, v’)-diagram. These
changes forced the CIE to reconsider the specifi-
cation of the isotemperature lines. It was decided
that in calculating isotemperature lines, the pre-
sent value of c, be used, but the CIE 1960 (u, v)-
diagram be retained as the uniform-chromaticity-
scale diagram. This decision has the effect that all
correlated color temperature determinations of
light sources made since 1968, that is, the year the
present value of c, was adopted, remain the same.
A change from the CIE 1960 (u, v)-diagram to
the 1976 CIE (v’, v’)-diagram was considered un-
desirable as the resulting change in the correlated
226 COLORIMETRY

color temperature scale would offer no advantage Equations 6(3.11) define the tristimulus space in
in practical applications. In this instance, continu- which the chromaticity diagram is the CIE 1960
ity of practice was considered more important. UCS diagram required for the calculation of iso-
The change of c, from 1.4380 X 10°*m- K to temperature lines.
1.4388 X 10°? m- K has only a small effect on From the (U;, V,, W;)-tristimulus values, ob-
the positioning of the isotemperature lines in Fig- tained by Eq. 5(3.11), the (u;, v;)-chromaticity
ure 1(3.11). In fact, on the scale of that figure, the coordinates are given by
change is of little or no consequence if the figure
is used for the graphical determination of corre-
lated color temperature. es Ur
In some practical applications, correlated color iL Ute Vet We
temperatures are required as part of a sequence
of colorimetric calculations, for example, in F
WEE Vr
ENG aa (7(3.11)]
the calculation of CIE daylight illuminants D
(Section 3.3.4), and in the calculation of CIE
color-rendering indices (Section 3.3.11). For such The (uz, v;)-chromaticity points define the
applications, numerical methods of determining Planckian locus in the (u, v)-chromaticity dia-
correlated color temperatures are preferred over gram. The slope m, of the tangent to the Planckian
graphical methods (Mori et al., 1964; Robertson, locus at the point (u;, v7) is given by the first
1968; Mahr et al., 1969). derivative dv;/du,, that is,
Mori et al. (1964) propose the following
method by means of which the slope of the iso-
temperature lines can be calculated. m _ der
=

Fie.
The spectral radiant power S;(A) of a
blackbody radiator at temperature 7 (in kelvin K)
is obtained by an equation of the following form which is obtained from

S7(A) = oA [exp( <2) = 1 dor _ dv,/dT

duy du,/dT
[8(3.11)]
[4(3.11)]
which has been discussed in detail in Section 1.2.2 By using dots to denote derivatives with regard to
[see Eq. 6(1.2.2)]. The spectral radiant power dis- temperature T, such as (dx;/dT)=xX,7, Eq.
tribution {S;(A)dA} leads to the tristimulus val- 8(3.11) then reads
ues
dvr _ er
Ur = {Sp(A)a(A)ar dup ur

ae [SrQ)oA)aa [5(3.11)] which can be expressed in terms of the tristimulus

values U;, V;, Wand their derivatives U,, V;, W,
as follows:
W, = [Sp(A)w(A) dd
per
where u(A), o(A), w(A) are linear combinations Wapeee
5, y.

of the CIE 1931 x(A), y(A), Z(A) standard color-

matching functions; that is, Lows V>(Ur + Vert W,) a V-(U; + Vr+ W,)

u(A) = 35(A) U;(U; + Vp t+ Wr) 3 U; (Uy + Vr+ Wr)

[9(3.11)]
(A) = 7A) [6(3.11)]
w(A) = —4e(A) + 49(A) + 420A) Whereas U;,V7;,W, are determined by Eq.
 Distribution Temperature, Color Temperature, and Correlated Color Temperature 227

5(3.11), Up, Vr, Wy follow from Robertson’s (1968) method of determining the
correlated color temperature of a source of given
5) chromaticity is the numerical equivalent to the
ligt feveaT 2A Sa(A)exp| =)
graphical method normally used, and involves an
interpolation between two adjacent members of
the set of 31 isotemperature lines defined in Table
1(3.11). With each line (i = 1 to 31) is associated
the correlated color temperature T.;. A computer
program is written that permits the search through
the set of lines to find the two adjacent lines
between which the chromaticity point of the given
source lies. This is done by calculating in the
(u, v)-chromaticity diagram the distance of the
given chromaticity point from each of the 31
lines. If the coordinates of the chromaticity point
Pos [evesT ?A*(d)ex| <=) of the given source are denoted by (u,, v,), and
(u;,v;) denotes the chromaticity point through
which the ith isotemperature line of slope ft, passes,
the required distance d; is obtained from
Jeo(<] ~ i)a [10(3.11)]

The isotemperature line passing through

= (0, a v;) ‘iP t;(u, ae u;)

G+ ee
(uy, vr) is perpendicular to the tangent at that >

point of the Planckian locus; thus its slope [13(3.11)]

t,(u, v) is given by
The chromaticity point (u,, v,) lies between the
11
tr(u, v) Bibi [11(3.11)] adjacent lines i=j and i=; +1 if the ratio
d,/d;+, is negative.
The correlated color temperature T.. that corre-
The integrations required in Eqs. 5(3.11) and
sponds to (u,, v,) is found by assuming that (a)
10(3.11) are, in practice, replaced by summations
the Planckian locus between 7; and 7;,, can be
using the weighted-ordinate method from A, =
replaced by the arc of a circle whose center is at
360 nm to A, = 830 nm with wavelength interval
the intersection of the two adjacent isotempera-
AX of constant 1-nm width (see Section 3.3.8).
ture lines j and j + | [see Figure 2(3.11)], and (b)
The chromaticity points (u;,v,) of the
the reciprocal color temperature is a linear func-
Planckian locus and the slopes t;(u,v) of the
tion of distance along this arc. It follows that
isotemperature lines passing through (u7, v7) can
be transformed to the CIE 1931 (x, y)-chromatic-
ity diagram by means of the following equations:
RohiNGe s WONT
1 Cc vlogs 0, |I -4|

motes
elo thit ps
ae OTe rade [14(3.11)]

a 5 aa
OT
omc where 9, and 6, are the angles between the two
isotemperature lines (7; and 7;,,, respectively)
and the line joining (u,, v,) to their intersection.
and For adjacent lines, 6, and 8, are small, and thus
one can set 0,/0,= sin@,/sin@,, so that Eq.
vxeoe
|4trQuse)iek| Hi2t-(aye) 14(3.11) becomes
iu,v0). =
aoe aera) Stl [483

Table 1(3.11) shows the results of the calcula-

tions made in accordance with Eqs. 4(3.11) to
12(3.11). [15(3.11)]
 228 COLORIMETRY

Table 1(3.11) Isotemperature Lines (c, = 1.438810? m-K) Calculated in Accordance with the Method
Proposed by Mori et al. (1968)

Chromaticity Coordinates Chromaticity Coordinates

Reciprocal in CIE 1960 UCS Diagram in CIE 1931 Diagram
Megakelvin Temperature ——————__—_______—. Slope. —————————_ Slope
i (MK ') T(K) ur Or ty(u, v) Xr JF tr(x,y)
l 0 00 0.18006 0.26352 —0.24341 0.23987 0.23404 —0.68705
2 10 100,000 0.18066 0.26589 —0.25479 0.24258 0.23802. —On21g7
3 20 50,000 0.18133 0.26846 —0.26876 0.24560 0.24240 —0.77926
4 30 33433 0.18208 0.27119 —0.28539 0.24890 0.24714 —0.84248
>, 40 25,000 0.18293 0.27407 —0.30470 0.25251 0.252227 55a
6 50 20,000 0.18388 027709 0.32675, 025645 0.25763 —1.01403
7 60 16,667 0.18494 0.28021 —=0:35156-- 0.26070 (26833 = ia
8 70 14,286 0.18611 0.28342 —0.37915 0.26526 0.26930 "== 152695"
9 80 12,500 0.18740 0.28668 —0.40955 0.27011 0.27547 —1.44313
10 90 1 Leh 0.18880 0.28997 —0.44278 0.27524 0:28182 «> Goons
11 100 10,000 0.19032 0.29326 —0.47888 0.28063 0.28828 Tada so0 t
12 125 8,000 0.19462 0.30141 —0.53204..— 0.29513 0.30477 —3.08425
13 150 6,667 0.19962 0.30921 —0.70471 0.31101 0.32116 —6.18336
14 175 5,714 0.20525 0.31647 —0.84901 0.32775 0.33690 —39.34888
15 200 5,000 0.21142 0.3231 2am OL 0.34510 0.35162 11.17883
16 225 4,444 0.21807 0.32909) att apt eek OS 0.36276 0.36496 5.34398
17 250 4,000 0.22511 0.33439 —1.4512 0.38045 0.37676 3.68730
18 275 3,636 0.23247 0.33904 —1.7298 0.39792 0.38690 2.90309
19 300 3333 0.24010 0.34308 —2.0637 0.41502 0.39535 2.44455
20 325 S07 0.24702 0.34655 —2.4681 0.43156 0.40216 2.14300
21 350 oY 0.25591 0.34951 — 2.9641 0.47464 0.40742 1.92863
22 375 2011 0.26400 0.35200 —3.5814 0.46262 0.41121 1.76811
23 400 2,500 0.27218 0.35407 —4.3633 0.47701 0.41368 1.64291
24 425 L353 0.28039 3557 Teoria 27162 0.49059 0.41498 1.54240
25 450 2222 0.28863 0.35714 —6.7262 0.50338 0.41525 1.45962
26 475 2,105 0.29685 0.3582 3inwra8 5955 0.51541 0.41465 1.39021
2 500 2,000 0.30505 0.359072 rellB24 0.52669 0.41331 1.33101
28 D259 Tou 0.31320 0.35968 —15.628 0:537328 0.41131 1.27989
29 550 1,818 0.32129 0.3601 lasi4232325 0.54712 0.40882 1.23522
30 So 1,739 0.32931 0.36038 —40.770 0.55640 0.40593 1.19579
31 600 1,667 0.33724 0.36051 —116.45 0.56508 0.40271 1.16074

“Data supplied in part by A. R. Robertson, private communication, 1980.

The success of the method depends on choosing a
set of isotemperature lines with sufficiently nar-
row spacing so that the errors introduced by the
above assumptions are negligible. Robertson
(1968) has shown that the set given in Table
1(3.11) leads to very satisfactory results.
3.12 COLORIMETRIC INSTRUMENTATION
Modern colorimetric instrumentation is designed
to provide automatically the tristimulus values
and chromaticity coordinates of a given color
stimulus without making use of an actual human
eye aS a measuring component of the instrument.
There are three kinds of colorimetric instrumenta-
tion: the spectroradiometer, the spectrophotometer,
and the tristimulus-filter colorimeter. Each kind is
commercially available in a variety of makes,
degrees of sophistication, and specializations to
suit specific colorimetric applications. It is not the
purpose of this book to treat in detail colorimetric
instrumentation and to describe details of their
design and operation. Such information is readily

Fig. 2(3.11). Method of interpolation to find the
correlated color temperature of a source of given
(u,, v,)-chromaticity (Robertson, 1968).
available from the well-established manufacturers
of such instruments. However, we have selected a
few topics that we consider relevant and of some
general importance to the subject matter.
3.12.1 Spectroradiometers
A spectroradiometer is an apparatus designed to
measure radiometric quantities as a function of
wavelength.
Figure 1(3.12.1) illustrates schematically a
spectroradiometric system. The radiant power
emitted by the test source in a preselected direc-
tion enters the entrance slit of the monochroma-
tor through a coupler which optically connects
the test source with the monochromator. In the
particular case shown in the figure, the optical
coupler is an integrating sphere, possibly with
(not shown) external lenses, mirrors, apertures,
and baffles. The monochromator (see Section 1.5)
disperses the incoming radiant power and trans-
mits it via a narrow band of wavelengths through
the exit slit which is optically coupled with the
detector (see Section 1.6). The detector’s photo-
electric response is analyzed .by the computer
which, in turn, commands an output display. The
computer is also interfaced with various other
components of the spectroradiometric system to
control their automatic operation.
In practice, the measurement of a spectral
radiometric quantity involves the comparison of
the test source with a suitable reference source of
known spectral radiant power distribution. The
arrangement shown in Figure 1(3.12.1) includes a
Colorimetric Instrumentation 229
reference source. The comparison of the test
source with the reference source is accomplished
by rotating the integrating sphere by 180° to
admit light from the reference source to the
monochromator.
Reference sources, most often in the form of
tungsten filament lamps, are usually available
from National Standardizing Laboratories [see
Table 1(3.12.1)]. The spectral radiant power dis-
tributions of these sources have been determined,
directly or indirectly, by a spectroradiometric
comparison with a blackbody (Planckian) radia-
tor (see Section 1.2.2) operating at a known tem-
perature, or with some other primary standard
source, such as one providing synchrotron radia-
tion (Key and Ward, 1978).
A survey of sources used for accurate measure-
ment of visible, near-infrared, and near ultraviolet
radiation conducted by Steiner (1976) contains
useful information for the practicing spectrora-
diometrist.
A typical output of a spectroradiometric sys-
tem is:
(a) spectral radiant power distribution of the
test source;
(b) tristimulus values and chromaticity with re-
spect to either the CIE 1931, or CIE 1964 colori-
metric system, or both;
(c) CIE color-rendering indices;
(d) correlated color temperature;
(e) photometric quantities, such as luminance.
Whereas the spectral radiant power distribution is
determined by actual physical measurement, the
other output data are derived from it by digital
computation using the appropriate formulae given
in the previous sections of this chapter.
The term spectral radiant power distribution
[see Table I(1.1)], when applied to a source of
radiant energy or a color stimulus, is defined as
the spectral concentration of radiant power as a
function of wavelength. More specifically, the
spectral concentration of radiant power is the
quotient of radiant power taken over an infinitesi-
mal wavelength range on either side of a given
wavelength, by that wavelength range. This defi-
nition implies that the spectral concentration P)
of radiant power should be strictly determined for
an infinitely narrow band of wavelength. How-
ever, spectral measurements are virtually always
required, not as ends in themselves, but for appli-
cation to the calculation of radiometric, photo-
metric, and colorimetric quantities.
230 COLORIMETRY

4 REFERENCE
ME SOURCE

DETECTOR

re
re

aya En
MONOCHROMATOR
He} BE
i ea

ES
8 TEST
3 SOURCE
nes

OUTPUT DISPLAY

Fig. 1(3.12.1). Schematic diagram showing principal components of a spectroradiometric

system; PS & ME= power supply and measuring equipment; OP = optical coupler; IF=
electronic interface; E n= entrance slit of monochromator; E, = exit slit of monochromator.

These quantities are usually derived by first value

applying spectral weighting functions [e.g.,
x(A), y(A), Z(A)], to the measured spectral
radiant power distribution, and then processing
the product functions further by means of in-
tegration or summation across the spectrum. This
procedure treats the individual measured data
points as a finite set of discrete points, although,
in reality, they may well be points belonging to a
continuum. In other words, an actual spectral
radiant power distribution curve is usually treated
as a step-function, similar in appearance to a
histogram.
Thus, what is required of the spectroradiomet-
ric system is not a set of values for the spectral
concentration P, of radiant power at specific
wavelengths, but rather measurements of the mean
spectral concentration of radiant power over a set
of contiguous bands of wavelength of small but
finite width, covering the spectral range of inter-
est, for example, the visible spectrum from A, =
360 nm to A, = 830 nm at intervals of AA = 5
nm. The spectral concentration P, of radiant
power is therefore approximated by the mean
ee a eee ee
MOEA SDN Ar
The appropriate choice of the width of the
wavelength interval at which the step-function of
spectral radiant power is to be determined de-
pends on the complexity of the spectral radiant
power distribution and on the nature of the spec-
tral weighting functions to be employed. For in-
candescent filament sources, such as tungsten or
tungsten—halogen lamps (see Section 1.2.2), a rel-
atively broad wavelength interval (e.g., AA = 10
nm) can be chosen across the spectrum because
such sources have a smooth and continuous spec-
tral radiant power distribution varying only slowly
with wavelength. In the spectroradiometric mea-
surement, at each wavelength setting of the
monochromator, the ratio of the spectral radiant
power of the test source to that of the reference
source is given by the ratio of the detector
responses. The size of the wavelength band
 Colorimetric Instrumentation 231

Table 1(3.12.1) Major National Standards Laboratories Which Offer

Radiometric and Photometric Calibrations of Sources of Radiant Energy
for Reference Purposes in Colorimetry

Country Name and Address

Australia CSIRO Division of Applied Physics

P.O. Box 218
Lindfield, NSW 2070
Canada National Research Council (NRC)
Ottawa, Canada KIA OR6
Federal Republic Physikalisch Technische Bundesanstalt (PTB)
of Germany Bundesallee 100
3300 Braunschweig
France Institut National de Métrologie (INM)
292, rue Saint Martin
75003 Paris
German Democratic Amt fiir Standardisierung, Messwesen
Republic und Warenpriifung (ASMW)
Wallstrasse 16, DDR-102 Berlin
Great Britain National Physical Laboratories (NPL)
Teddington, Middlesex TW11 OLW
Japan Electrotechnical Laboratory (ETL)
1-1-4, Umezono, Sakura Mura,
Niihari-Gun, Ibaraki-ken
Japan 305
National Bureau of Standards (NBS)
Washington, DC 20234
USSR Institut de Métrologie D.I.
Mendeéléev (IMM)
Leningrad
Republic of National Physical Research Laboratory (NPRL)
South Africa P.O. Box 395
Pretoria 0001

transmitted by the monochromator is of little
importance in the case of incandescent sources
provided that this wavelength band is not appre-
ciably greater than the interval at which the mea-
surements are taken.
For electric discharge sources, such as fluo-
rescent lamps (see Section 1.2.2), the spectral
radiant power distributions are mixtures of multi-
line and spectrally continuous radiation. For such
sources spectral measurements at closer inter-
vals are usually required to obtain a sufficiently
accurate spectral distribution and consequently,
sufficiently accurate tristimulus values and chro-
maticity coordinates. In most cases, spectral data
taken at 2-nm intervals, with comparable wave-
length bands transmitted by the monochromator,
are considered adequate.
Some fluorescent lamps, however, can be mea-
sured by a more economical hybrid technique
consisting of measurements at the rather few
principal mercury emission lines, and separately,
of measurements made of the rather smooth and
continuous phosphor emissions. Examples of such
measurement records are given in Figure 3(1.2.3)
Zo COLORIMETRY
and in Table 2(3.3.8). In these cases, the con-
tinuum can often be represented adequately by
spectral data taken at 10-nm intervals to which
are added the emission-line values each spread
uniformly over a 10-nm wavelength band and
centered at its respective emission-line wave-
length.
The wavelength band transmitted by the
monochromator is usually chosen to be of the
same width as the wavelength interval used in
the calculation of tristimulus values. Its de-
termination and effect on the measured spectral
radiant power distribution is a problem of consid-
erable concern in spectroradiometry, but its full
treatment is beyond the scope of this book. The
interested reader is referred to Kostkowski (1971)
and Shumaker (1979).
The more recent developments in spectroradi-
ometry and its application to photometry and
colorimetry are exemplified in studies by Merrill
and Layton (1966), Sanders and Gaw (1967), Eby
(1970), Nonaka et al. (1972), Bauer (1973), Sanders
and Jerome (1973), Kok and Hengstberger (1974),
de Groot et al. (1975), Morren (1975), Suzuki and
Ooba (1976), and Grum and Becherer (1979).
The CIE Technical Committee on Photometry
and Radiometry has under preparation a Report
on the Spectroradiometric Measurement of Light
Sources that treats the subject matter of spec-
troradiometry in considerable detail, including
methods of measurement and error analysis. Much
of this section has been guided by an early draft
of the CIE Report.
3.12.2 Spectrophotometers
A spectrophotometer is an apparatus designed to
measure the spectral transmittance and spectral
reflectance factors of objects. It is similar in de-
sign to the spectroradiometer described in Section
3.12.1, but has a rather different objective, namely,
to compare at each wavelength the radiant power
leaving the object with that incident on it.
Figure 1(3.12.2) illustrates schematically a
spectrophotometric system used for the measure-
ment of nonfluorescent objects. Radiant power is
emitted by the light source and passes through the
coupler that optically connects the source with
the monochromator. The monochromator (see
Section 1.5) disperses the incoming radiant power
and transmits it via a narrow band of wavelengths
through the exit slit which is optically coupled
with the illuminating and viewing chamber hold-
ing the object under test and, in the case of
reflectance-factor measurements, a standard of
reflectance factor. The detector system receives
the spectral radiant power reflected from the ob-
ject and the standard in close succession and
generates a ratio signal that is transmitted to the
computer for analysis and display. The computer
is interfaced with various components of the spec-
trophotometric system and controls its automatic
operation.
Figure 1(3.12.2) also shows schematic dia-
grams of alternative illuminating and viewing
chambers that are commonly used in spectropho-
tometry.
With regard to reflectance-factor measure-
ments the CIE recommends four different geome-
tries of illuminating and viewing the object; these
have been described in Section 3.3.7 and are
shown in Figure 1(3.3.7). The standard of reflec-
tance factor is the perfect reflecting diffuser, an
ideal surface that does not exist in reality, but
that is used as a basis for calibrating secondary
standards or white standards for practical spec-
trophotometry. A discussion of such standards
has been given in Section 3.3.6.
It is worth noting that in the measurement of
nonfluorescent objects, the actual spectral radiant
power distribution of the light source of the spec-
trophotometric system is not critical, because only
radiant power ratios are measured at the various
wavelengths. The main requirement is that the
light source emits adequate radiant power at all
wavelengths of interest. The situation is quite
different when fluorescent objects are to be mea-
sured. This case will be discussed in Section 3.12.3.
The spectral transmittance 7(A) of a trans-
parent nonfluorescent object is the ratio of trans-
mitted radiant power P,AA, to incident radiant
power P,,AX, that is,
ts Py
ON Fe Py, AA een
where A is the central wavelength of the narrow
band AX.
The definition of spectral reflectance factor
B(A) is more complex because it takes account of
the fact that the reflecting properties of an opaque
object depend much on the way the object is
illuminated and viewed. All diffusing materials
fail to be perfectly matte; that is, they exhibit a
certain amount of gloss. Part of the radiant power
incident on the surface is reflected in some more
or less good approximation of an image-forming
state; the remainder penetrates the surface and,
 LIGHT SOURCE

En

MONOCHROMATOR

Ex
DETECTOR

ILLUMINATING AND
VIEWING CHAMBER
CLLLZA
(0/45 GEOMETRY)
BJ.

OUTPUT DISPLAY

ALTERNATIVE ILLUMINATING AND VIEWING CHAMBERS

Ex EX
as

INTEGRATING SPHERE
(DIFF./O GEOMETRY )
FOR REFLECTANCE -FACTOR ATTACHMENT FOR
MEASUREMENT TRANSMIT TANCE
MEASUREMENT
Fig. 1(3.12.2). Schematic diagram showing principal components of a spectrophotometric
system for measurement of spectral reflectance factors and spectral transmittance. PS &
ME = power supply and measuring equipment; OP = optical coupler; IF = electronic interface;
M = mirror; RM = rotating half-mirror; En = entrance slit; Ex = exit slit; Obj = object under
test; Std = standard of reflectance factor. In the lower part of the figure are shown two
examples of alternative illuminating and viewing chambers that can be used in the spectropho-
tometric system. One shows an integrating sphere for the measurement of reflectance factor
with diffuse illumination and normal viewing; the other shows the arrangement of mirrors used
to make measurements of transmittance.

233
234 COLORIMETRY
Test object
Fig. 2(3.12.2). Schematic diagram showing principles involved in measuring the spectral
reflectance factor B(A) of an opaque nonfluorescent test object.
after suffering absorption, scattering, and multi-
ple reflection beneath the surface, it is re-emitted
in a nearly diffuse state.
It is appropriate, and in accordance with the
CIE recommendation (Section 3.3.6), to assess the
reflecting properties of an opaque object relative
to the perfect reflecting diffuser identically il-
luminated and viewed and a formal definition of
spectral reflectance factor may then be given as
follows:
The spectral reflectance factor B(A) of a non-
fluorescent object is the ratio of the spectral
radiant power reflected in the directions delimited
by a given cone whose apex is at a given point on
the surface of the object, to the spectral. radiant
power reflected in the same directions by the
perfect reflecting diffuser identically illuminated.
Figure 2(3.12.2) illustrates the definition of
spectral reflectance factor. The spectral radiant
power incident on the perfect reflecting diffuser is
denoted by P),AA and is identical in every re-
spect to that incident on the test object. Depend-
ing on whether the perfect reflecting diffuser or
the test object is in position, the spectral radiant
power P,,AXA or P,AA passes through the aper-
ture A of given shape and size. The spectral
reflectance factor B(A) of the test object is then
given by
PAD
B(A) = PAN [2(3.12.2)]
and from Figure 2(3.12.2) it can be readily de-
duced that its value depends to a large extent
upon the direction of illumination e, the direction
of viewing ¢,, and solid angle w of the cone
determined by the size and distance of aperture
A, particularly when the test object has a glossy
surface. For example, if €, is equal to e and w is
small, the spectral reflectance factor B(A) may
have a value much larger than unity because the
cone may contain the mirror image of the source.

There are two limiting cases for the spectral
reflectance factor with regard to the size of the
cone delimiting the reflected spectral radiant
powers. If the solid angle w of the cone ap-
proaches zero, the spectral reflectance factor is
more appropriately called spectral radiance factor.
If the solid angle of the cone approaches 27
(hemisphere above the test object), the spectral
reflectance factor is called spectral reflectance and
denoted by p(A). By definition, the perfect re-
flecting diffuser reflects all the incident radiant
flux Pp,AA in all directions uniformly without
loss. It then follows that the spectral reflectance
p(A) of a test object is simply the ratio of the
spectral radiant flux P,AA reflected into the
hemisphere above the test object to the incident
spectral radiant flux P),AA; thus,
Pre Ay
p(A) = BDA [3(3.12.2)]
There is no spectrophotometric system on the
market that is versatile enough to serve equally
well on every colorimetric problem. To select the
spectrophotometric system that will serve a par-
ticular purpose best and economically, the re-
quirements must be clearly understood. The main
requirements concern the kinds of specimen that
must be measured, its number per day, the spec-
tral range, and the precision and accuracy.
General guidelines for selecting an appropriate
spectrophotometric system have been prepared,
for example, by Judd and Wyszecki (1975) and
Johnston (1971).
3.12.3. Spectrophotometry of Fluorescent
Materials
The phenomenon of fluorescence can be observed
in many materials and consequently is often of
concern to colorimetrists wishing to determine
colorimetric data for objects made of such
materials. Fluorescence belongs to a class of phe-
nomena with the generic name of /uminescence.
Luminescence refers to the process by which
radiant energy is emitted by molecules or atoms
that have been excited by the absorption of radiant
energy. The concepts of quantum theory are used
to construct models of the luminescence phenom-
ena.
A system of molecules or atoms takes up
radiant energy only in certain definite steps, and
is stable only in discrete, more or less sharply
defined, energy levels. The lowest of the levels is
Colorimetric Instrumentation 235
N'
N
oS lS calls)ec a mae)
Fig. 1(3.12.3). Schematic diagram of energy
levels for the explanation of models of different
cases of luminescence: (1) resonance radiant en-
ergy; (2) fluorescence; (3) and (4) anti-Stokes
fluorescence; (5) phosphorescence.
the ground level, or the ground state, of the
system. For the atoms and for many diatomic
molecules found in gasses, the energy levels are
well known. For polyatomic molecules and for
systems found in crystals and other solids, the
energy levels are difficult to determine and are
mostly unknown. However, the assumption of the
existence of such energy levels generally has led
to important advances in the understanding of
the luminescence phenomena (Pringsheim, 1948;
Becker, 1969).
Figure 1(3.12.3) shows a typical schematic dia-
gram of several energy levels of a simple system
of molecules. The energy levels are represented by
horizontal lines. The separation between any two
of these lines is proportional to the corresponding
difference in energy. At the ground state, repre-
sented by line N, the system has the energy Qy;
at level F, it has the energy Q;. The difference
between the two energies is proportional to the
frequency v;-, of the radiant energy that is emitted
or absorbed by the transition between the two
levels F and N. The basic law is
hvry = Or — On
whereh is Planck’s constant (~ 6.63 X 10° *4J- s).
The quantity Av;, constitutes the elementary
quantity of radiant energy, called quantum or
photon.
236 COLORIMETRY
When radiant energy of frequency vpy is ab-
sorbed, the energy of the system is raised from
level N to level F. If no other energy levels exist
between WN and F, the system can return to N only
by re-emitting radiant energy of the same
frequency v;-,. Theoretically, this is the simplest
case of luminescence; it is called resonance radia-
tion [see Case | in Figure 1(3.12.3)].
Generally, however, there are several other
intermediate levels (C, D,---) between N and F,
and thus, other transitions from F to C, D,--:-
can occur. These transitions result in fluores-
cence; that is, the re-emission of radiant energies
of frequencies ypc,V¥rp,---, all of which are
smaller than ypy; see Case 2 of Figure 1(3.12.3).
The fact that fluorescence occurs at a lower
frequency (which means a longer wavelength, be-
cause of vy= c/A) than the frequency of the ab-
sorbed energy was first found empirically by
Stokes (1852) and is often referred to as Stokes’
law. However, deviations from Stokes’ law are
possible if other energy levels N’ and F’ are
located immediately above levels N or F, respec-
tively. Thermal energy can raise the system from
N to N’ before the absorbed radiant energy of
frequency vf raises it to F. The fluorescence is
of frequency vy > Vyr. Thermal energy can also
raise the system to F’ during its lifetime in the
excited state F. The fluorescence is of frequency
Ven > Vyp- Both these cases are referred to as
anti-Stokes fluorescence; see Cases 3 and 4 in
Figure 1(3.12.3).
Case 5 of Figure 1(3.12.3) illustrates phos-
phorescence. The energy of the molecules is first
raised by absorption of radiant energy from level
N to F; it falls back to a nearby metastable level
M, only to be returned again to F by thermal
energy before it changes to a lower state, such as
D, through emitting radiant energy.
When the incident radiant flux is cut off,
luminescence persists for a certain period of time.
The time for decay to a given fraction of the
initial radiant power re-emitted depends on the
kind of luminescence (e.g., fluorescence or phos-
phorescence), the nature of the substance, and
other factors (e.g., temperature, in the case of
phosphorescence). The decay times of fluores-
cence are generally considerably shorter than those
of phosphorescence. For fluorescence the time
during which the emitted radiant power decays to
e | times its initial value is of the order of 10°
sec; whereas for phosphorescence, this may be
from 10~° sec to many days (e = base of natural
logarithms = 2.718 ---).
The luminescence excited by the absorption of
radiant energy has a certain period of growth or
induction. With a sufficiently long period of con-
stant irradiance, the re-emitted radiant power
reaches an equilibrium value when the number of
molecules excited per second by absorption is
equal to the number of molecules deactivated by
emission. For practically all fluorescent materials,
the periods of growth and decay are equal be-
cause the relative number of excited molecules
always remains small compared to the total num-
ber of unexcited molecules, even if the level of
irradiance is very high. It follows that the fluores-
cence produces a radiant power that is strictly
proportional to the radiant power incident on the
substance (Pringsheim, 1948; Ooba, 1963). The
linear relationship between the radiant power re-
emitted and the one incident on the substance is
of considerable importance when the colorimetry
of fluorescent materials is done by way of
spectrophotometry. The same linear relationship
usually does not hold for phosphorescence
(Pringsheim, 1948).
For complex systems of molecules, such as
those found in solids, the energy levels of a fluo-
rescent molecule are not as sharply defined as was
implied in Figure 1(3.12.3) by the horizontal lines.
The location of each energy level is influenced by
the continuously fluctuating interaction of the
surrounding field of molecules whose configura-
tion changes due to thermal vibration. The mole-
cules oscillate around stable positions within
broadened bands of energy, and it follows that
both the absorption and the emission of radiant
energy takes place in bands covering a range of
wavelengths rather than lines.
The relative spectral radiant power distribu-
tion within the emission band describes the spec-
tral character of the fluorescence. Generally, it
does not change with the wavelength of the inci-
dent radiant power. It often resembles a bell-
shaped curve with its peak located near the
centroid wavelength of the emission band. How-
ever, the absolute value of the radiant power
emitted by fluorescence at wavelength A can vary
greatly with the incident wavelength p. Its value is
a function of the radiant power absorbed at p and
the excitation efficiency (or quantum efficiency) of
the material at that wavelength. If the material
absorbs gj quanta at a given wavelength p, and
re-emits a total of q quanta within the entire
emission band, the excitation efficiency of the
material is defined as q/qo. The absorbed quanta
that are not re-emitted by fluorescence are con-

verted into thermal energy. Many fluorescent
solids have an excitation efficiency that is nearly
constant over a considerable range of incident
wavelengths nw. However, the excitation efficiency
drops sharply when the incident wavelength falls
within the emission band.
The absorbed radiant power is that part of the
incident power that is not reflected and trans-
mitted. The sum of the absorbed power plus the
reflected power plus the transmitted power al-
ways equals the incident power. The reflectance p
of a material surface of area dA is defined as the
ratio of the radiant power reflected from dA into
the entire hemisphere ({2= 27 sr) above the
surface to the power incident on dA. Similarly,
the transmittance + of the same material is de-
fined as the ratio of the power transmitted through
the material to the power incident on it. Thus, the
absorbance a of the material is given by
a=l1-p-T
All three quantities, a, p, and + are wavelength
dependent. It is important to recall here the dif-
ference between reflectance p and radiance factor
B (see Section 3.12.2). In order to determine a
from the above equation, it is essential that p be
used and not B.
The spatial distribution of the radiant power
re-emitted by fluorescence is uniform in all direc-
tions above the surface of the fluorescing object.
This means that the radiance of the fluorescence
component measured in different directions from
the surface is constant; such a surface is called a
Lambertian emitter of radiant energy.
The objective of a spectrophotometric analysis
of a fluorescent material is to determine the total
spectral radiance factor of the material across the
visible spectrum so that this can be used in the
calculation of colorimetric data appropriate for
the material when this is irradiated by a given
light source.
The total spectral radiance factor B;(A, 0) has
two additive components, that is
| Br(A,o) = Bs(A) + BL(A,¢) [1(3.12.3)]
In this expression, B,(A) is the spectral radiance
factor due to reflection only (the subscript S stands
for “substrate,” because in many industrial appli-
cations, fluorescent materials are made of a non-
fluorescent material, the substrate, to which is
added one or more fluorescent agents). The quan-
tity B,(A, o) is the spectral radiance factor due to
fluorescence only (the subscript L stands for
Colorimetric Instrumentation 234
Source
uid
Photon
Detector
gy;
CStandard Object
a
Fig. 2(3.12.3). Schematic diagram illustrating the
principal components of a spectrophotometer de-
signed to measure the total spectral radiance factor
of a fluorescent object.
luminescence). The spectral radiance factor
B,(A, o) is not only a function of wavelength A,
but depends also on the spectral radiant power
distribution {S(A)dA} of the source o irradiating
the material. Consequently, the total spectral
radiance factor B;(A, @) is a function of o as well
as X.
There are two principal ways of determining
the total spectral radiance factor of a fluorescent
material for colorimetric applications.
The first method is schematically illustrated in
Figure 2(3.12.3). The source irradiates the given
fluorescent object and, in turn, the (nonfluo-
rescent white) working standard of known spec-
tral radiance factor. The radiant powers reflected
from the object and the standard in a fixed direc-
tion are spectrally analyzed by the monochroma-
tor—detector system, so that appropriate ratios
can be taken that define the total spectral radi-
ance factors of the fluorescent object.
When the fluorescent object is placed in the
beam, a photocurrent is generated in the detector
which is directly proportional to the radiance
L,(Obj)dA of the object at the wavelength A»
transmitted by the monochromator. When the
test object is replaced by the standard of reflec-
tance, another photocurrent is generated that is
directly proportional to the radiance L,(Std)daA
of the standard at that wavelength. The ratio of
the two photocurrents is equal to the ratio of the
two radiances L,(Obj)dA and L,(Std)dA. With
known spectral radiance factor Bg,g(A) of the
standard, the spectral radiance factor of the fluo-
rescent object simply follows from:
ae L (Obj) [2(3.12.3)]
Bale OC lae Boal A)
238 COLORIMETRY
The spectral radiance factor Bg,g(A) of the white
working standard is assumed to be given in terms
of the perfect reflecting diffuser identically irradi-
ated; that is,
L,(Std)
Bsta(A) = Fen) [3(3.12.3)]
where L,(PRD) is the spectral radiance of the
perfect reflecting diffuser. By inserting this
expression into Eq. 2(3.12.3) it follows:
L,(Obj)
aS [4(3.12.3)]
Br(A, o) ae
which defines the total spectral radiance factor of
the fluorescent object as a ratio of the spectral
radiances measured for the object and perfect
reflecting diffuser identically irradiated by the
source o and viewed by the monochromator-—
detector system.
It is important to note that the resultant total
spectral radiance factors are not only characteris-
tic of the particular object under study, but also
of the particular source o used to irradiate the
object. Figure 3(3.12.3) illustrates the spectral dif-
ferences that can occur when a given fluorescent
object is irradiated first by daylight (o = D,;)
and then by incandescent lamplight (o = A). This
phenomenon makes it mandatory that the spec-
trophotometric procedure is such that the source
factor
Radiance
300 400 500
Wavelength (A), nm
Fig. 3(3.12.3). Spectral radiance factors of a fluorescent object measured under a source with
the relative spectral power distribution S(A) of CIE standard illuminant D,;, and under a
source with $(A) of CIE standard illuminant A. The instrumental setup used in the measure-
ments is that illustrated in Figure 2(3.12.3).
irradiates the object directly as shown in Figure
2(3.12.3). The more conventional arrangement in
spectrophotometers, in which the object is irradia-
ted through a monochromator as shown in Figure
1(3.12.3), must not be used for the measurement
of total spectral radiance factor of fluorescent
objects. It further follows that, if the fluorescent
object is to be analyzed colorimetrically [e.g., CIE
tristimulus values are to be calculated with
B,(A, o)], the colorimetric data can apply only to
the particular source o used in the measurement
of B;(A, o). Or, stating it differently, it follows
that if the colorimetric data of the fluorescent
object are to apply to a particular source 9, it is
important that this source o be used in the spec-
trophotometric determination of B;(A,o). This
seemingly trivial point is somewhat belabored
here, because in practice, confusion has been often
noted. Presumably this confusion originates from
the fact that most objects used in practice are
nonfluorescent and therefore do not require the
special attention with regard to the source used in
the spectrophotometric analysis. Moreover, many
spectrophotometers are not readily converted so
as to be capable of measuring fluorescent objects.
The application of the method of measuring
total spectral radiance factors of fluorescent ob-
jects, described above, to the colorimetry of such
objects can have an important practical problem.
In colorimetry (see Section 3.3 and 3.4), standard
illuminants are to be used in the calculation of
B(A, Dés)
600 700 800

the CIE tristimulus values of object-color stimuli.
These standard illuminants, defined by tables
of relative spectral radiant power distributions
{S(A)dA}, cannot exactly be materialized by a
source and its associated optics in a spectropho-
tometer set up to measure the total spectral radi-
ance factor of a given fluorescent object. In fact,
the spectral differences between the radiant power
distributions of the instrument source and the
desired standard illuminant may be sufficiently
large so as to render the colorimetric calculations
unacceptable.
A number of investigators have attempted to
quantify the colorimetric errors that may result
from a spectral mismatch between the instrument
source and the standard illuminant, and indeed
have proposed schemes by means of which ap-
propriate corrections can be applied to the erro-
neous measurements (Eitle and Ganz, 1968;
Allen, 1972; Alman and Billmeyer, 1977; Grum
and Costa, 1977). All these schemes must make
certain assumptions concerning the fluorescence
properties of the object and the instrument used
to measure it, and it is not always a priori evident
that these assumptions hold strictly in each case
considered. The CIE Materials Committee (TC-
2.3) is studying the proposed schemes and it is
anticipated that a quantitative assessment of their
usefulness will be forthcoming in the not too
distant future. Schemes such as those proposed in
the literature are in demand in industry because
they are relatively simple to use and in many
=owt
Ow
ec
oD
Source veSic
o &
°o @®
c>
oo
==
eg vA
— Standard object nea
<_—__>>
Fig. 4(3.12.3). Schematic diagram illustrating the arrangement of the principal components of
a special spectrophotometer designed to make a detailed spectral analysis of fluorescent
objects.
Colorimetric Instrumentation 239
instances provide results with acceptable accu-
racy.
The second method of measuring the total
spectral radiance factors of a given fluorescent
object is the most comprehensive method that can
be devised. It provides a detailed spectrophoto-
metric analysis of the fluorescent object, on the
basis of which CIE tristimulus values can be
calculated with respect to any illuminant of known
spectral radiant power distribution (Donaldson,
1954; Fukuda and Sugiyama, 1961; Grum, 1972;
Baba and Sengoku, 1973; Baba et al., 1974/75).
However, the instrumental setup for such an anal-
ysis is more elaborate and the volume of data that
must be taken for each complete spectrophoto-
metric record is considerably larger than that
required in the first method.
Figure 4(3.12.3) schematically illustrates the
instrumental setup required in the second method.
Instead of exposing the object to radiant power of
all wavelengths simultaneously and then measur-
ing its spectral radiance factor, as indicated in
Figure 2(3.12.3), the object is now exposed to
monochromatic radiant power of different wave-
lengths p. At each wavelength p (first monochro-
mator) a spectral scan of wavelengths A (second
monochromator) is made to obtain a record of
the spectral radiant power reflected and emitted
from the object.
With the white (nonfluorescent) working stan-
dard in the beam, the measured spectral radiances
L,(Std) are equal to zero at all wavelengths A
Photon
we detector
240 COLORIMETRY
except when A = p». When the fluorescent object
is substituted for the working standard, spectral
radiances Ly ,(Obj) are measured which are dif-
ferent from zero not only at A = p but also within
the entire emission band characteristic of the flu-
orescent object. The radiance factor B(A, ») of
the object irradiated by monochromatic radiant
power of wavelength p is then given by
Ly. (Obj
B(A,m) = Boal) a the case where the incident radiant power has a
[5(3.12.3)]
where Bs,4(A) is the spectral radiance factor of
the standard assumed to be known in terms of the
perfect reflecting diffuser (see Eq. 3(3.12.3)).
The complete spectrophotometric record for
the fluorescent object takes on a matrix of the
form in which the rows are labeled by the wave-
lengths A of the reflected or emitted radiant power
and the columns are labeled by the wavelengths p
of the incident radiant power. All entries B(A, 1)
have either positive or zero values and, if Stokes’
law applies, as it usually does, the matrix is of
triangular form; that is, all its positive entries are
on and below its diagonal:
WavelengthA
of Reflected
or Emitted Wavelength pu (nm) of Incident Radiant Power
Radiant Power 300 310
300 BAS) ei
320 BA,p) BCA,w) BCA,B) ~«
780 B(A,Hw) BOA,pw) BOA,B)
790 BOA,w) BOB) BOA,BH)
800 B(A,pw) BA,BH) BOA,2)
Figure 5(3.12.3) illustrates a small part of the
complete matrix record obtained for a fluorescent
red ink. The four parts of the illustration record
the spectrophotometric data for incident radiant
power held at wavelength p = 340, 460, 580, and
760 nm respectively, whereas wavelength A is
scanned through the spectrum. Thus, these re-
cords represent the entries of the matrix columns
bw = 340, 460, 580, and 760 nm, respectively. The
first and second examples, » = 340 and 460 nm,
illustrate cases where p is smaller than the short-
est wavelength of the fixed emission band of the
fluorescent object. In these cases, a radiance fac-
tor B,(A, ») due to reflection only is recorded at
\ = p and a set of radiance factors B;(A, w) due
to fluorescence at wavelengths A of the entire
emission band. In the third example, » = 580 nm,
the wavelength of the incident radiant power falls
just inside the emission band. Fluorescence still
occurs in this case; however, when p is increased
to or beyond the centroid wavelength (~ 610 nm)
of the emission band, fluorescence generally
ceases. The fourth example, »p = 760 nm, shows
wavelength near the long wavelength end of the
emission band and a radiance factor due to reflec-
tion only is recorded at A = p.
The spectrophotometric matrix record of the
fluorescent object can be used to calculate
the total spectral radiance factors B;(A, 0) for
the object irradiated by any source o of known
spectral radiant power distribution {S,(A)dA}.
The procedure follows directly from Eq. 5(3.12.3)
and by inspection of the matrix format. The fol-
lowing relation emerges:
B,(A, 0)S,(A)dA = Ih
B(A, #)S,(m) du
[6(3.12.3)]
780 790 800
7 os a
= a a
BN) == ae
BA,nw) BAB) —
BA,pw) BA,pH) BOA,B)
The resulting 8;(A, 0) corresponds with its direct
measurement by means of the first method dis-
cussed above and illustrated in Figure 2(3.12.3).
Figure 3(3.12.3) is also recalled to show how the
resulting total radiance-factor curve changes when
the irradiating source o is changed from daylight
(D,;) to incandescent lamp light (A).
3.12.4 Propagation of Random
Spectrophotometric Errors
In the determination of tristimulus values and
chromaticity coordinates by means of spectropho-
 Colorimetric Instrumentation 241

S —o

0.10 ba Emission band oH

0.05
factor
Radiance

300 400 500 600 700 800

0.15

0.10

0.05
factor
Radiance

300 400. 500 600 700 800

0.10

factor
Radiance

300 400 500 600 700 800

on Emission band 4

N
factor
Radiance
\

LLL
300 400 500 600 700 800
Wavelength (A), nm

Fig. 5(3.12.3). Typical examples of a detailed spectrophotometric analysis of a fluorescent red

ink irradiated by spectral radiant flux of different wavelengths 1. (Adapted from measurements
reported by Donaldson, 1954).

tometry and computation, it is of interest to The chromaticity coordinates (x, y) are com-
estimate the error in the computed results that puted by means of the familiar equations
originate from uncertainties in the spectrophoto- rl
metric data. The uncertainties in the spectropho- PES x is 2 p(A)S(A)X(A)AA
tometric data in question are random fluctuations Kk Kits Z >,p(A)w(A)AA
in the photometric-scale readings of the instru-
ment when repeated observations are made. A yi _ Yypl(A)S(A)y(A)AA
Nimeroff (1953) and MacAdam (1953, 1954, 1957) Shoes >, p(A)w(A)AA
have dealt with the problem, using elementary but
effective statistical techniques. [1(3.12.4)]
242 COLORIMETRY

with and a frequency function expressed by

w(A) = S(A)x(A) + S(A)V(A) + S(A)Z() ®[x;, Vis Xm> Ym» M |dx; dy,

The errors in the (x, y)-chromaticity coordinates l l

= IaM'/2 exp(eA 5 O;dx,dy,
are assumed to arise entirely from random un-
certainties in p(A), the spectrophotometrically
determined spectral reflectance function of the [4(3.12.4)]
given object. The weighting functions S(A)x(A),
S(A) (A), S(A)Z(A) are not subject to uncer- Here M is the determinant of M,
tainty if the CIE recommended color-matching
functions x(A), y(A), Z(A) and a standard illumi- M = var(x)var(y) — [cov(xy)]?
nant {S(A)AA} are used.
The variances of x and y, denoted by var(x) and Q, is the quadratic form
and var(y) and the covariance, denoted by
cov(xy), are then obtained from Ooea (ihe ee

2, [WS(A)x(A)
warx) i= = Xw(A)Jvar [ (AJAX 4 2(2%,= yp) =n) ee
W4

5 LWS(A) F(A) + (yp) sia)

var(y) = = Q, has ax? (chi-square) distribution with two
degrees of freedom.
3, [WS(A)x(A) — Xw(A)] With appropriate choice of Q,, the ellipse de-

cov(xy A)
) LWS(= re WvarloY] AA
Yn)F(A) fined by Eq. 5(3.12.4) can be computed which
should contain a specified fraction h of all points
(x;, y;). In particular, the ellipse expected to con-
tain 95% of all chromaticity points computed
[2(3.12.4)] from the n spectral reflectance functions p,(A)
with and used in deriving the constants in Eq. 5(3.12.4),
has QO, = 6.0 (from tabulated x? distribution for
W=X+Y+X
two degrees of freedom).
and The theory outlined above can be readily ap-

varfo(A)] = 27 Lei) ~ oT
[3(3.12.4)]
where p;(A) denotes the ith individual measure-
ment of the spectral reflectance at wavelength A,
and p,,(A) is the mean value of n measurements.
It is assumed that there is no correlation between
the random errors in the spectral reflectance mea-
surements at different wavelengths.
The chromaticity coordinates x,, y, derived
from the ith set of individual measurements of
p(A) are assumed to have a joint normal distribu-
tion about their mean values, x,,, y,,, With a
variance—covariance matrix
plied to practical cases. In general, the random
errors of spectral reflectance measurements are
not constant throughout the spectrum, and they
may also depend on the actual values of p(A).
Nevertheless, calculations made on the assump-
tion of a specified and constant standard devia-
tion for p(A), namely
o, = var[ p(A)] = const [6(3.12.4)]
are informative. For this case, the variances and
covariances of x and y take on simpler forms;
that is,
aaprebaants: amS [WS(A)z(A) — Xw(A)PAA
A

var( x) cov( xy)

cov( xy) var(y) var(y) = = S[WS(A) (A) — Yw(A)PAA
rv

2
poe i SLWS(A)E(A) = Xw(A))
x [WS(A)y(A) — Yw(A)]
AA
[7(3.12.4)]
The uncertainty ellipse (h = 0.95) becomes
bs re ee
+ 2x — x0)(y — yp)
pane, a
ay vi 6.0
[8(3.12.4)]
with the following interpretation. The center
(Xo, Yo) Of the ellipse represents the chromaticity
computed by Eq. 1(3.12.4) for a single set of
individual measurements of the spectral reflec-
tance p(A). Thus, on 95% of those occasions
when such a single set is measured, the true
chromaticity point (x,,,, y,,) Will, in fact, lie within
the ellipse. The size, shape, and orientation of
the ellipse depend on the tristimulus values
(X>, Y, Z>) and also, of course, on the assumed
standard deviation o,. MacAdam (1954) has
shown that the ellipses all cannot be transformed
into equal size circles of radius o, by any plane
transformation of chromaticities, whether linear
or nonlinear.
The more general case in which the weighting
functions S(A)x(A), S(A) y(A), S(A)Z(A), as well
as the spectral reflectance function p(A) are sub-
ject to random error is, of course, more complex
but can be studied by the same methods [see
Nimeroff (1957) and Section 5.4.6.].
There are numerous types of systematic errors
that can occur in spectrophotometric measure-
ments, such as errors in the wavelength scale and
photometric scale, and errors due to dispersed
and undispersed radiant power, slit width, and
temperature changes. Such errors cannot be
analyzed by normal statistical techniques. Their
effect on the calculated (x, y)-chromaticity coor-
dinates is estimated in a different way [see Sec-
tion 1.5.3 and also Robertson (1967)].
3.12.5 Tristimulus-Filter Colorimeters
A tristimulus-filter colorimeter is an instrument
with spectral response functions directly propor-
Colorimetric Instrumentation 243
tional to the color-matching functions of one or
the other CIE standard colorimetric observers. Its
importance as a tool for production control of
colored objects is well established. Most commer-
cially available instruments of this kind are satis-
factorily precise and easy to operate, and they are
fast enough to measure many specimens a day.
However, many of them seriously lack accuracy;
that is, tristimulus values read by such instru-
ments do not generally agree with tristimulus
values obtained by spectrophotometry and com-
putation. Fortunately, this defect does not neces-
sarily render such an instrument entirely useless.
There are numerous important practical applica-
tions for which less accurate but precise instru-
ments can still be quite useful.
The fundamental problem in building a tri-
stimulus-filter colorimeter is to adjust three
photocells or photomultipliers so that (to a suffi-
cient approximation) their responses are propor-
tional throughout the visible spectrum to, for
example, the CIE 1931 color-matching functions
x(A), y(A), Z(A), or to some other linear combi-
nation of them. With an exact duplication of the
CIE color-matching functions, the response of
the first photocell gives the X-tristimulus value of
the radiant power incident on the photocell, the
response of the second gives the Y-tristimulus
value, and the response of the third, the Z. In a
sense, the tristimulus-filter colorimeter is a photo-
electric analogue computer with an output corre-
sponding to the three integrals defining the
tristimulus values given in Eq. 3(3.3.8).
The proper spectral response adjustment of
the photocells can be achieved, at least approxi-
mately, by one of two principal methods. The
first method employs a monochromator with three
alternately used spectrum templates, each of which
modifies the spectral response function of a single
photocell to one of the desired x(A), y(A), Z(A)
functions of the standard observer. The second
method modifies the spectral response function of
the photocell by means of colored glass or gelatin
filters that are placed in front of the photosensi-
tive surface of the photocell. The first method
results in a template colorimeter; the second
method leads to a filter-photocell colorimeter.
Figure 1(3.12.5) is a schematic diagram of a
template colorimeter. Radiant power leaving the
specimen and passing through the entrance slit of
a monochromator is dispersed into its spectrum.
One of three templates is placed in the plane of
the spectrum—the X-template, Y-template, or Z-
template. Each template is designed to modify
244 COLORIMETRY
PRISM
<-f}---+—
ier er
a, Lx
ace ere antes BS)
SPECIMEN
(X)-TEMPLATE (Y)-TEMPLATE
Fig. 1(3.12.5). Schematic diagram of a template colorimeter.
selectively an equal-energy spectrum in such a
way that the spectral response curve of the photo-
cell takes the shape of one of the three color-
matching curves of the standard observer. With
the X-template in position, the response of the
receiver will correspond to the X-tristimulus value
of the source—specimen combination no matter
what the spectral composition of this combina-
tion is. Similarly, with the Y-template and Z-
template in position, the corresponding Y and
Z-tristimulus values are obtained.
Several template colorimeters have been built
and are used for determining the tristimulus val-
ues of light sources, particularly fluorescent lamps,
and of reflecting and transmitting objects of all
kinds (Mader, 1945; Winch, 1951; Harrison, 1952;
Mahr, 1958 and Ives, 1915). The accuracy of
template colorimeters can be very high, especially
when the spectrum into which the templates are
placed is of large dimension (typically 10 by 5
cm). In these cases, the cutting and calibration of
the templates is relatively easy and can be carried
out accurately. Instruments that employ small
spectrum templates (3 by 1 cm or less) usually
suffer in accuracy because of the great difficulties
encountered in the cutting and calibration of the
templates.
A good template colorimeter is an elaborate
and complicated instrument. The complexity of
its design is comparable with that of a spectro-
photometer, and it is necessarily an expensive
instrument. It is thus not surprising that only few
template colorimeters have been used in practice.
Instead, the spectroradiometer or the spectro-
photometer coupled with appropriate digital
computer facilities is employed as discussed in
Sections 3.12.1 and 3.12.2.
PHOTOMULTIPLIER
3] 1)fSa
all
=
1X
O
LJ
2 '
TEMPLATE READ OUT
HOLDER rulyl 3g
(Z)-TEMPLATE
The cheaper but usually less accurate filter-
photocell colorimeters are still in demand and
Figure 2(3.12.5) is a schematic diagram of such an
instrument. Radiant power leaving the object
passes through one of the three tristimulus filters
X, Y, Z and falls onto the photocell causing it to
given a response proportional to the correspond-
ing tristimulus value of the object-source com-
bination. Each tristimulus filter is usually a
combination of colored glass filters so chosen that
the resultant spectral transmittance function of
the combination corrects the spectral response
function of the photocell to one of the CIE color-
matching functions. If the spectral response func-
tion of the photocell is to match the CIE color-
matching function x(A), the spectral transmit-
tance function T(A, X) of the ideal X-tristimulus
filter has to be
ra.)
=20
where s(A) is the uncorrected spectral response
function of the photocell expressed in arbitrary
units. Similarly, the spectral transmittance func-
tions of the ideal Y and Z-tristimulus filters are
found from
Fg pes(A) £0)
and 7T(A,Z)= AUN)
respectively.
Sometimes it is more desirable to duplicate the
CIE color-matching functions weighted by the
relative spectral radiant power distribution of one
of the standard illuminants. Thus, instead of using
x(A), y(A), Z(A) as the target response functions,
 Colorimetric Instrumentation 245

Source

Test Object &

SI
Tristimulus
filters

Possible arrangements of
glass filter combinations
for tristimulus filters

YMMV 4, Kha

UU
Yl
SSS

VLLLLLLLLLLL 2
I
Fig. 2(3.12.5). Schematic diagram of a filter-photocell colorimeter. Also shown are two (I and
II) principal ways of combining glass filters to give a tristimulus filter.

the functions in series [see lower part of Figure 2(3.12.5)]. In

Ssa(A) — Ssa(A) -
Ssta(A) -
are chosen, where S(A) is the relative spectral
radiant power distribution of the actual source
built into the instrument, and S¢,4(A) is the rela-
tive radiant power distribution of a CIE standard
illuminant, for example, that of D,;. The conver-
sion from the actual source [S(A)] to [Sg,g(A)] is
permanently built into the instrument, and the
tristimulus values of a reflecting (nonfluorescent)
specimen are directly read by the instrument with
respect to the standard illuminant even though
the actual source illuminating the specimen may
be quite different, such as an incandescent lamp.
There are two main methods of making a
tristimulus filter. One is to select colored filters of
suitable spectral transmittance and combine them
this arrangement (I) the light is filtered suc-
cessively before it falls on the photocell.
Unfortunately, the range of different spectral
transmittance curves available from glass filters is
rather restricted and makes it impossible to ob-
tain a tristimulus filter with the ideal spectral
transmittance curve. This is particularly evident
in the case of the X-tristimulus filter, which
requires a double-peaked curve of spectral trans-
mittance, and only two separate filter-photocell
combinations can help to avoid this difficulty.
The tristimulus filter actually obtainable by
arranging its components in series is usually a
rather crude approximation of the ideal filter. The
use of many components may improve the ap-
proximation, but normally makes the combina-
tion increasingly dense and eventually renders the
combination useless because it transmits practi-
cally no light.
Figure 3(3.12.5) shows what can at best be
expected from a filter-photocell colorimeter with
tristimulus filters built up by combinations of
246 COLORIMETRY

2.0
— CIE 193] standard observer
---- Photocell with tristimulus
filters

oO

functions
Color-matching
and
functions
filter-
photocell
response
©o ea)

600 700
Wavelength \ (nm)

Fig. 3(3.12.5). Examples of spectral response functions showing the degree to which a
photocell combined with tristimulus filters may best duplicate the color-matching functions of
the CIE 1931 standard observer. The tristimulus filters are made up of combinations of colored
glasses arranged in series [see arrangement I in Figure 2(3.12.5)] (after Robertson et al., 1972).

colored glasses arranged in series (Robertson
et al., 1972). The agreement with the target color-
matching functions is close but not perfect. In the
case of this specially designed (noncommercial)
colorimeter, the discrepancies are generally very
small and may even be considered negligible in
most practical situations. However, this is a spe-
cial case and does not apply to commercial instru-
ments. Commercial instruments show consider-
ably larger discrepancies between the spectral
response functions of the filter-photocell combi-
nations and the target color-matching functions,
and the chromaticities and luminance factors
measured by these instruments are usually in
noticeable disagreement with the target values
obtained by spectrophotometry and computa-
tions. Discrepancies of 0.020 in the chromaticity
coordinates x and y, and discrepancies of 1.5 in
the luminance factor Y of object-color stimuli of
medium to high lightness are not uncommon; and
such discrepancies are about 10 times the toler-
ance required for most colorimetric work.
The other method of combining filters of
suitable spectral transmittances is given by
arrangement II [see lower part of Figure 2(3.12.5)].
In this arrangement, sometimes referred to as the
Dresler principle (Dresler, 1933; Geutler, 1958),
some filter components are placed side by side.
Different portions of the beam of light are filtered
differently before reaching the photocell. The re-
sultant spectral transmittance curve of the combi-
nation can be effectively controlled by varying
the relative sizes of the filter components. By
using this principle the ideal tristimulus filters can
be approximated with a high degree of accuracy
and relatively high peak transmittance.
Figure 4(3.12.5) gives an example of what can
be expected from a filter-photocell colorimeter
with tristimulus filters carefully built up by com-
binations of colored glass arranged in series and
side by side (Fruhling and Krempel, 1955). How-
ever, because of the technical difficulties involved
in making these rather elaborate stacks of filter
combinations, few colorimeters employing the
Dresler principle have been built commercially.
The selection of suitable filters for either
method I or method II [see Figure 2(3.12.5)]
is generally made empirically (Hunter, 1938;
Corduan, 1957), but also numerical methods can
be applied to facilitate the selection (Davies and
Wyszecki, 1962; Wright, Sanders, and Gignac,
1969).
 Colorimetric Instrumentation 247

CIE 1931 standard observer

1.6 © © o 0 o Photocell with tristimulus filter:

w
=
le
5
Cc
2
c
os
Cc
ro)
argrre
es
> Qa
rT)
oO wo
c =

S ®
-_- oO
oo 08
Es
Teo
= (Sy
Tol
oO2
=

0.4

0
700
Wavelength » (nm)

Fig. 4(3.12.5). Example of spectral response functions showing the degree to which a
photocell combined with tristimulus filters may duplicate the color-matching functions of the
CIE 1931 standard observer. The tristimulus filters are made up of combinations of colored
glasses arranged in series as well as side by side [see arrangement II in Figure 2(3.12.5)]
(Frihling and Krempel, 1955).

The degree to which the spectral transmit-
tances of the tristimulus filters match the ideal is
perhaps the most important indicator of the accu-
racy to be expected from the tristimulus-filter
colorimeter. To give accurate color specifications,
the instrument must give tristimulus values (or
chromaticity coordinates and luminance factor)
exactly in accordance with the standards recom-
mended by the CIE. This requirement goes be-
yond the fitting of the spectral response functions
of the instrument to the standard color-matching
functions. It also demands that the source used to
illuminate the specimen be carefully considered in
the design of the instrument and ideally be one
whose spectral radiant power distribution dupli-
cates that of a CIE standard illuminant such as
D,;. Furthermore, reflecting specimens must be
measured in a CIE standard illuminating and
viewing geometry [see Figure 1(3.3.7)] and the
perfect reflecting diffuser must be the reference
for the measurement of reflecting specimens (Sec-
tion 3.3.7).
However, there are many, perhaps most, col-
orimetric measurements in practice that do not
require color specifications of each specimen
in absolute terms. Colorimetry by difference
(Judd and Wyszecki, 1975) is a very common
technique in color production control which re-
quires good precision but rarely high accuracy. If
the accuracy requirements can be relaxed, low-cost
tristimulus-filter colorimeters can be built that are
useful tools in colorimetric practice.
In particular, a tristimulus-filter colorimeter
designed to measure small color differences be-
tween objects whose spectral reflectance-factor
distributions are essentially the same can be per-
mitted to have large discrepancies between the
spectral response functions of the instrument and
the color-matching functions of the CIE standard
observer. The color difference is measured with
high accuracy. The spectral differences between
the two specimens must be small to ensure a
satisfactory measurement of the color difference
between the objects. When highly metameric pairs
of objects are compared, the measured color dif-
ference may significantly deviate from the true
difference.
In practice, a standard (or reference) object is
often available which is made of the same material
and colored with similar colorants as the dupli-
248 COLORIMETRY
cate objects. Colorimetry by difference of the
duplicate objects against the standard is the most
accurate way of finding the degree to which the
duplicate differs in color from the standard.
When the standard is made of a different
material and/or colored with different colorants
than the duplicate objects, high degrees of
metamerism may occur and the operator of a
tristimulus-filter colorimeter must be on guard
lest the standard used differs too drastically in
spectral composition from the duplicates to per-
mit a reliable color-difference measurement.
 CHAPTER 4
PHOTOMETRY
4.1 BASIC PHOTOMETRIC QUANTITIES AND
UNITS
The photometric quantities regarded as basic and
of most general use are displayed in Table I(4.1)
in the Appendix. The definitions used here are
substantially those adopted in the International
Lighting Vocabulary (third edition, and an early
draft of the fourth edition) of the Commission
Internationale de lEclairage (CIE) 1970). The
CIE Vocabulary may also be referred to for
equivalent definitions in French, German, and
Russian, and the translations of the terms (without
the definitions) in five additional languages
(Spanish, Italian, Dutch, Polish, and Swedish).
Table 1(4.1) is a supplement to Table I(4.1).
It contains the defining equations of the basic
photometric quantities and corresponding SI units
of measurement (SI = Systeme Internationale
d’Unités).
Although the SI units of measurement are now
widely used, some of the literature still refers to
nonSI units, particularly with regard to length
and area units. Table 2(4.1) provides appropriate
conversion factors for different units of illumi-
nance and different units of luminance.
4.2 THE PHOTOMETRIC PRINCIPLE
The matching of brightness is the fundamental
operation in photometry. In direct visual photom-
etry, the visual stimuli provided by two juxta-
posed uniform light patches of similar shape and
angular area in the observer’s field of view are
adjusted so that the patches appear equally bright.
If, at match, the relative spectral radiant power
distributions of the two stimuli in the wavelength
range comprising the visible spectrum are the
same, then the distributions will also be the same
in absolute terms. The observer’s eye functions
are little more than a sensitive null instrument
that could be replaced by wholly physical light-
sensitive devices with response properties deviat-
ing widely from those of the human eye, including
a different and unrelated spectral responsivity.
The essentially visual character of brightness
matching enters where matches are made between
visual stimuli of different relative spectral radiant
power distribution. The basic postulate of pho-
tometry is then that whatever the visual stimulus,
fixed in all respects, of one patch, and whatever
the fixed relative spectral distribution of the
stimulus of the second patch, a brightness match
can always be reached by varying the absolute
value of the second stimulus.
Superficially, brightness matching appears
more elementary than trichromatic color match-
ing as only one variable is involved. However,
from a psychophysical standpoint, brightness
matching is a more recondite process. It involves
the recognition, in the visual sensations produced
by the two patches, not of complete similarity but
of similarity of one quality (brightness) in the
presence, in general, of differences in other quali-
ties (hue and saturation). Heterochromatic bright-
ness matching (i.e., brightness matching of two
stimuli having different relative spectral radiant
power distributions, when carried out by the di-
rect method), is difficult for the observer and is
attended by a wide spread in the values of re-
peated matches. For this reason, indirect visual
methods, notably the flicker method and the step-
by-step method, have been developed that are
easier and give more precise results; these purport
249
 250 PHOTOMETRY

Table 1(4.1) Basic Photometric Quantities, Their Defining Equations and Units.“ :

Term Symbol Defining Equation Explanatory Notes Units

Luminous flux F (or P,,) F, = KP, P,: Radiant flux (W) Im

(or power)
Fi= Kofta K: Luminous efficacy (Im-W~')
K,,,. Max. luminous efficacy
(683 Im-W_')
dF, :
Luminous he I,= 7 dw,: Element of solid angle with cd
intensity mf apex (1) at surface of source (Im-sr_')
d°F.
Luminance L L,=>— dA,: Surface element of source cd-m~?
_ ° dA,cos €,da, Be Ss
(im-m™“-sr -)

e,: Angle between given direction

(1)-(2) and normal n, of dA,
dl
= + — dA cos €;: dA orthogonally projected
dA cos &,
on plane perpendicular to
given direction (1)—(2)
dE,
TPES dE,, ,,: Illuminance on dA, normal
2008 £242 to the direction (1)—(2)
dE
a dw,: Element of solid angle with
dw,
apex (2) at surface of receiver
dF
I!luminance E, E=— vi dA,: Surface element of Im-m~?
‘ receiver
dF. we
Luminous M M= Im-m
i dA,
exitance

Luminous K(A) K(A)=K,,V(A) V(A): Relative photopic luminous

efficacy efficiency function
function

“Supplement to Table I(4.1) in the Appendix. [Figures 1(1.1) and 2(1.1) of Chapter 1 illustrate the
explanatory notes of this table.]
If photometric quantities are considered over small intervals of the spectrum, their respective symbols are
correspondingly designated by the index A or v. For example: L,, ,= dL, , /dA is the spectral concentration
of luminance per unit wavelength interval at A. With regard to the definition of the solid angle w and its
element dw see Fig. 2(1.1) in Chapter 1. For the conversion of luminance and illuminance expressed in the
above units to the values in terms of other common units, metric and nonmetric, see Table 2(4.1).

to equate the same quality of the sensation as the
direct method. Whether or not this is so, it must
be recognized that, strictly speaking, the results
and experimental laws of brightness matching are
relative to the particular method of brightness
matching being used, as well as, of course, to the
particular observational conditions, such as field
size and level of stimulation.
In Sections 5.7 and 5.8, we will discuss in more
detail the experimental data that have been
gathered on heterochromatic brightness matching
and how these depend on the method and ob-
servational conditions used. In the present section
we will confine our discussion, in the main, to the
principles that govern photometry as it is prac-
ticed today by international agreement.
 251
“SqINS OI X OTHE
Aq Aydryynur ‘a]qei oy JO pesy 94}
x Y = swoquiryjooj yx ‘ojdurexa Joy ‘Wun yeuy oy Ysnory) UNOS oY} pue BUN [eur oY} ySnosY} MOI 94} JO JOasSIO}UT 94} 1B Joquinu oy}
94} UO ULUNOS 94} Ul PoUBU HUN B JO SUT} UT possoidxa Ayiquenb & JO onjeA 94} WAAUOD OT, ‘aJON
ye MOI OY} Ul poweU JUN B JO SWIIO} UL ON]RA S$} OF 9[QE} OY} JO Opis puBY-1J9]
JIOQUIR}IOOJ [| = a]puRd}00J JUETBAINbe |

tt
qinsode | = jepuolq | xn] Juayeainbo |

ol
JIIQUIP] | yoyd yuayeambea 7 :sjrun (juayeAmnba) 1394109
Sn - e e k .
a i
gul/po | = I sO X Pl Z0l X b@S"P -OL X 698°h 1-OI X 698° -cOl X 698'h 1-O1 X OSST Ol X OSST = gUl/eJEpUuRD I
[,—UIp9] — 7/ePPpueD I
z/P9l = = ¢_OI X Pr6'9 I Trl'€ Z8ee eeOl X ZBE'E OI X TREE c-O1 X OLOT OL X 9401
[YP] = Woquievyj}00-4 |
23/Po (4/1) = = s-Ol X OIT'Z~=s-OT X EBT'E I 9L0:1 ~~e20 EX 92021 WOE X9L0 TH 120m APOCE 97P'E€
ie)
«OI X pS0'% 1-OI X LS6°T 1-OI X 0676 I e-Ol Ol s-OI X E8I'€ ESE = Moquelyiw |
[qu] ,W9d/po (4/1) --Ol = woque’y |
S07. .z0OleXESG6c <0 X 06¢6 01 I 701 1-O] X €8TE eOl X EST E=
(1 zW9/po (4/1) =
;O1 Xrsot% <Ol X L567 -<-Ol X 0676 1-0l r-Ol I c-O] X €8I'E 1-Ol X ESTE = qinsody |
[qse] gw/po (24/1) =
Tsp9 <Ol X 0676 OI X 6167 ecOl X CHIE THEE 2401 CRTC I 01 = qins I
[qs] gwis/po [| =
«01-01 X 616° 1-OISX CHILEw oe0EX erie CEI Ge 7207 I = IN I
(3) i/o) [aa OL %eTSe 9m 3 Ol 40676 ee
ig 2 fie. @ 2 2 bee Bee oe Ee Ree See eee ee ee
(voie Jo g_Ul BJopueD=s_g-YeRjepuey = Woquieyjooj = Woquiey iy] Waquey] qinsody qIns 1IN
UOTVIADIQGY
yiun asenbs Jad)
Rjapuey ee ee
Sa aes ete me S S Se eRe eS Se Se SEW Uhl SS SS
S10}OV-] UOISIOAUOD BdURUIWUINT *Z
2 8° 2 oe eS eR SR Se SW WS Se hm ee ee
Peeras~ Be ater
[p93] 2/wy | = I 9L0°1 e-Ol X 9LOT OL X 9LO'T = 91pued}004 J
[ydw] =,WId/W , Ol = 1-01 X 0676 I e-Ol Ol = wydiIW I
[yd] guid/tuy | = 201 X 0676 201 I 101 = 10Ud I
[xq] gui/wy | = z-Ol X 0676 1-01 r-Ol I = xn |
8s kb eS 2s BR eee ee ee eee a. ee ee Ee
uoneiAsiqqy (vale jo yun gjpuesjoo7 rYdi[W 10Ud xn]
aienbs Jad)
uouwinTy
Oe fC ee eS eee ee ee ee
Poe ei 2 he Se 2 ee Re
$10}0¥- UOISIDAUOD dURUIWN]] *]
aL es SU OU
ee ee ee ee ee
a =Ee a ee ee
ee ee ee SSS
Be ES ee
(T"H)Z PIGEL
SS a eB
OULU] pue sdULUTUINATT JO SUE JUaAAFJIC] 1OJ S10}IV UOISIAAUOD
252 PHOTOMETRY
The basic laws of brightness matching are
assumed to be simple and equivalent to those
assumed to be valid for trichromatic color match-
ing (see Sections 3.2.1 and 5.3.1). In particular, it
is assumed that the following laws hold strictly:
(i) Symmetry Law:
If stimulus A matches stimulus B, then stimulus B
matches stimulus A.
(u) Transitivity Law:
If A matches B and B matches C, then A matches
C.
(ii) Proportionality Law:
If A matches B, then aA matches aB, where a is
any positive factor by which the radiant power of
the stimulus is increased or reduced, while its
relative spectral distribution is kept the same.
(iv) Additivity Law:
If A, B,C,D are any four stimuli, then if any two
of the following three conceivable brightness
matches
A matches B, C matches D,
and
(A + C) matches (B + D)
holds good, then so does the remaining match
(A + D) matches (B + C)
where (A + C), (B+ D), (A+ D), (B+ OC) de-
note, respectively, additive mixtures of A and C,
B and D, A and D, and B and C.
For a brightness-matching process obeying
these laws, the necessary and sufficient condition
for a brightness match between two visual stimuli
of radiant power distributions {P,dA} and
{PxdX}, respectively, takes the form
[PBA
AA=[PABA [1(4.2)]
where B(A) is a fixed function of wavelength
characterizing the brightness-matching process.
The brightness matches, for specified method
and observational conditions, made by actual ob-
servers deviate in various ways from the above
laws (Sections 5.7 and 5.8). For example, if the
matching field is viewed eccentrically or has an
angular subtense substantially in excess of 2°,
there is a clearcut breakdown in the proportional-
ity law when the absolute levels of the stimuli
are reduced from moderate to very low values.
For the same observational conditions, the other
laws are also unreliable and Eq. 1(4.2) is no
longer generally valid. For the most part, the
breakdown arises from a change in stimulus level
in the effective spectral sensitivity controlling the
matching process. This shift in spectral sensitivity
(the Purkinje shift) is associated with a transition
from cone to rod vision.
The approximate validity of the brightness-
matching laws, and thus, of Eq. 1(4.2), can be
restored if limitations are placed on the stimuli to
which they are applied. For some pairs of stimuli
in the brightness match, the match is determined,
or nearly so, by the response of the cone receptors
of the retina. Such matches are described as pho-
topic matches. If all the matches involved in an
application of the matching laws are photopic, the
laws hold at least to a tolerable approximation.
Similarly, Eq. 1(4.2) is applicable as an approxi-
mation to all photopic matches, and B(A) is a
fixed function of wavelength that characterizes
photopic brightness matching.
Scotopic matches, on the other hand, are those
predominantly determined by the rod receptors of
the retina, and the matching laws and Eq. 1(4.2)
are approximately true if all the matches involved
are scotopic; B(A) is now a different fixed func-
tion characterizing scotopic brightness matching.
Whether a particular brightness match is pho-
topic or scotopic depends on the spectral radiant
power distributions, relative and absolute, of the
matched stimuli and on the observational condi-
tions, such as method, field size, eccentricity in
the field of view, and state of adaptation as
modified by previous or surrounding stimuli. Pre-
cise rules for determining whether matches are
photopic or scotopic are difficult to lay down.
Moreover, some matches belong to neither class
as both rod and cone receptors are significantly
involved. The description mesopic is applied to
these hybrid matches.
For the particular case of small centrally
fixated matching fields (two degrees or less), com-
plications in brightness matching arising from rod
vision and the Purkinje shift are absent. All
matches are photopic. However, just as in color
matching, the color-matching functions are differ-
ent for small and large fields, so the fixed func-
tion B(A) for brightness matching with small
centrally fixated fields shows some differences
from the function appropriate to photopic matches
with large fields. Macular pigmentation is again
the principal cause of the differences.

By a procedure similar to that used in colorim-
etry (Chapter 3), the complications of actual
brightness matching are set aside by the introduc-
tion of ideal photometric observers. By definition,
the brightness matches of these observers obey
strictly the laws of brightness matching given
above. Two such ideal or standard photometric
observers have been defined by the CIE by the
specification of two fixed functions [B(A) = V(A)
and B(A) = V’(A)] appropriate, respectively, to
photopic and scotopic matching.
The matches predicted for the standard photo-
metric observer, with B(A) = V(A), are typical of
the small-field photopic matches of actual ob-
servers. This standard observer is much the more
important, as all ordinary photometric mea-
surements of luminous flux, luminous intensity,
luminance, illuminance, and so forth, rest, in
principle, on matches made by that observer.
The scotopic standard observer, with B(A) =
V’(A), is designed to give matches typical of the
scotopic matches of actual observers. The photo-
metric quantities, luminous flux, and so on, can
all be derived on the basis of the matches made
by the standard scotopic observer, but it is then
necessary to distinguish them from the corre-
sponding quantities for the standard photopic
observer. This is done by inserting the adjective
“scotopic” in the name of the quantity, for exam-
ple, scotopic luminous flux. It is not necessary,
but is sometimes useful, to qualify as photopic all
those photometric quantities that are based on
the standard photopic observer.
Similar uniform light patches, producing visual
stimuli defined by {P,dA} and {PxdA} respec-
tively, are in brightness match for the standard
photopic observer if, and only if,
JPxv(ryan = [Pry ar [2(4.2)]
This condition is also expressed by the statement
that photopic luminous flux entering the eye per
unit solid angle must be the same for the two
patches. For this to be an equivalent statement of
the condition for a photopic match, any defini-
tion of the photopic luminous flux corresponding
to a radiant flux {P, ,dA} that makes it a steadily
increasing function of the integral {P,V(A)d4,
will serve. It is most generally used to employ the
simplest possible function, and to define the pho-
topic luminous flux P,, as
PL=KyfP.V(A)dd — [3(4.2)]
The Standard Photometric System 253
where K,, is a constant. This is the definition
adopted in photometry, and once luminous flux is
defined, the definitions of all the other photomet-
ric quantities are determined [see Table 1(4.1)].
We note that the subscript e is used with P to
signify radiant flux (or radiant power) and
the subscript v is used with P to signify /uminous
flux (or luminous power). The use of the sub-
scripts is recommended when both “radiant” and
“Juminous” quantities are involved in the same
discussion.
The constant K,,, can be assigned an arbitrary
numerical value, for example, unity. However, for
historical and practical reasons, photometrists
have agreed to fix the numerical value of K,, at
683. We will discuss this agreement further in
Section 4.3.
The scotopic luminous flux P; corresponding
to a radiant flux {P, ,dA} is similarly defined as
the quantity
pea ifP,.V(A)dd —‘[4(4.2)]
where Kj, has the numerical value 1700.
43 THE STANDARD PHOTOMETRIC SYSTEM
4.3.1 Historical Note
The standard photometric system has evolved over
many decades along with our understanding of
human vision and with the technological develop-
ments in the measurement of radiant energy. Fur-
ther improvements in the standard photometric
system are expected to be made in the years to
come, particularly with regard to the system’s
capability of predicting the brightness of visual
stimuli of different colors perceived by actual
human observers with normal color vision (see
also Section 5.7).
There are two international organizations that
carry the responsibility of developing the stan-
dard photometric system: the Commission Inter-
nationale de l’Eclairage (CIE) and the Conférence
Générale des Poids et Mesures (CGPM).
The CIE was founded in 1913 by taking over
the functions of the earlier Commission Interna-
tionale de Photométrie, which was established in
1903. Since that time, the CIE has grown into the
international body that, by general consent in all
countries, develops standards and standard proce-
dures of measurement in virtually all fields of
illuminating engineering and those fields that are
254 PHOTOMETRY
directly relevant to it. The outstanding contribu-
tions of the CIE are the specifications of standard
observers, that is, ideal observers for use in col-
orimetry and photometry. The CIE Technical
Committees concerned specifically with the devel-
opment of colorimetry and photometry are, at
present, TC-1.2 (Photometry and Radiometry),
TC-1.3 (Colorimetry), TC-1.4 (Vision), TC-2.2
(Detectors), and TC-2.3 (Materials).
The CGPM and its executive committee, the
Comité International des Poids et Mesures
(CIPM) operates under the Convention du Metre
signed in 1875. It defines the base units (meter,
kilogram, second, ampere, kelvin, mole, candela)
of the fundamental quantities (length, mass, time,
electric current, thermodynamic temperature,
amount of substance, luminous intensity) of
physical measurement that, in virtually all coun-
tries around the world, become units of measure-
ment of legal status. In 1930, the CIPM became
actively involved in photometry, establishing in
1933 the Comité Consultatif de Photométrie
(CCP), which in 1971 became the Comité Con-
sultatif de Photométrie et Radiométrie (CCPR).
The members of the CCPR are representatives of
the major national standardizing laboratories [see
Table 1(3.12.1)], whose mandate it is to develop
and maintain the base unit of luminous intensity,
the candela, and to advise the CIPM on any
improvements that should be introduced interna-
tionally with regard to the definition of the
candela or to related fundamental aspects of the
standard photometric system. The CCPR works
in close cooperation with the appropriate commit-
tees of the CIE, particularly TC-1.2 (Photometry
and Radiometry).
The mechanism by means of which the CCPR
and CIE TC-1.2 maintain a measure of uniform-
ity of photometric and radiometric practice
throughout the world involves the occasional ex-
ecution of international intercomparisons of mea-
surements of appropriate quantities. The results
of such intercomparisons are generally published
for the benefit of all those concerned with photo-
metric and radiometric standards. A review of the
more recent international intercomparisons of
measurements of this kind is given in the second
draft of the revision of CIE Publication No. 18,
The Basis of Physical Photometry (CIE, 1979).
The evolution of the definition of the base unit
of photometry, the candela, best reflects the
evolution of the standard photometric system. In
the early period of photometric practice, the eye
served as the only detector of light and was used
to judge the equality of brightness between ‘lights
emitted by different lamps. During this period
when visual photometry played the dominant role
in measuring light, it was most appropriate to use
a specified light source as the basic photometric
standard. This light source became known as the
primary standard of light.
In actual photometric practice, the inverse-
square law of illumination was used very widely to
intercompare the brightness of different lamps.
This led to the early adoption of the primary
standard of light as the standard of /uminous
intensity rather than some other photometric
quantity, such as luminous flux. The unit of
luminous intensity, now called the candela, be-
came the photometric base unit, which is one of
the seven base units in the SJ system of units
(Systeme International d’Unités). Recommenda-
tions made by both the CCPR (1975) and the CIE
(1979) to the CIPM to adopt the /umen, the unit
of luminous flux, as the photometric base unit, so
far have not been successful. Although it is gener-
ally agreed that in modern photometric practice
luminous flux is the most convenient quantity to
be considered first, maintaining the tradition of
the present SI system is valued higher than practi-
cal convenience.
The first primary standards of light were
actual candles whose design, construction, and
operation were specified in detail to ensure rea-
sonably satisfactory performance characteristics
(see, for example, Walsh, 1958). However, in the
second half of the last century, these standards
were replaced by several types of oil lamps. The
search for an improved primary standard of light
continued into this century, but it was not until
1931 when Wensel et al. (1931) of the U.S. Na-
tional Bureau of Standards followed up a
suggestion by Waidner and Burgess (1908) and
succeeded in setting up a blackbody radiator op-
erating at the freezing point of platinum. Na-
tional standardizing laboratories of several other
countries followed the move of the U.S. National
Bureau of Standards and realized similar
blackbody radiators. In 1948, the ninth CGPM
formally adopted the blackbody radiator at the
freezing point of platinum as the new primary
standard of light and redefined the unit of
luminous intensity and changed its name from
bougie nouvelle to candela (CGPM 1948). In 1967,
the thirteenth CGPM improved the wording of
the 1948 definition of the candela by adopting the

following form (CGPM, 1967):
The candela is the luminous intensity, in the perpendicu-
lar direction, of a surface of 1/600,000 square metre of
a blackbody at the temperature of freezing platinum
under a pressure of 101,325 newtons per square metre.
The surface area of 1/600,000 m* was chosen
to ensure reasonable continuity with the units in
use in different countries prior to 1948. The defi-
nition translates to the equivalent and more fre-
quently quoted statement that the blackbody
radiator at the freezing point of platinum has a
luminance of 60 candelas per square centimeter.
The blackbody radiator at the freezing point
of platinum proved to be a useful primary stan-
dard of light for many years. Its major advantages
over its predecessors, the candles and oil lamps,
were more reproducible performance characteris-
tics, higher luminous intensity, and a calculable
spectral radiant power distribution by means of
Planck’s radiation formula (see Section 1.2.2) and
the thermodynamic temperature of freezing
platinum.
The operation of a blackbody radiator at the
freezing point of platinum is a major and complex
laboratory undertaking, despite technological ad-
vancements in high-temperature physics (see, for
example, Sanders and Jones, 1962). The develop-
ment of an alternative, less cumbersome, and
possibly still more accurate method of realizing
the candela remained a goal of several national
standardizing laboratories. A suggestion made by
Preston (1963) and Gillham (1964), both of the
National Physical Laboratory of Great Britain,
proved to be a noteworthy alternative and was
taken up by other national standardizing labora-
tories. The proposed new method formulated in
specific terms by Blevin and Steiner (1975) of the
Australian National Measurement Laboratory and
the U.S. National Bureau of Standards, respec-
tively, related the photometric base unit to the
watt in an exactly known numerical relationship.
This was considered a very appropriate approach
in view of the rapid increase in the very satisfac-
tory application of radiometric techniques to pho-
tometry, particularly those techniques involving
electrically calibrated radiometers (Gillham, 1964;
Blevin and Brown, 1967; Bischoff, 1968; Geist
and Blevin, 1973; Brown, 1975; Katsuyama and
Kihuchi, 1975; Hengstberger, 1977; Boivin and
Smith, 1978; Ono, 1979). In 1979 the sixteenth
CGPM approved the recommendations made by
The Standard Photometric System 255
the CCPR through the CIPM and adopted the
following new definition of the candela (CGPM,
1999):
The candela is the luminous intensity, in a given direc-
tion, of a source which is emitting monochromatic
radiant energy of frequency 540 < 10'* hertz and whose
radiant intensity in that direction is 1/683 watt per
steradian.
In this new definition, the candela is applicable to
photopic quantities, as well as to scotopic photo-
metric quantities. If it ever becomes desirable to
define mesopic photometric quantities, the above
definition could similarily be applied to those as
well. The specific value of 540 < 10! Hz for the
frequency of radiant energy required and the
specific value of 1/683 W - sr! of the intensity
of that radiant energy were chosen so as to ensure
reasonable continuity with the previous defini-
tion.
The specification of frequency instead of
wavelength for the monochromatic radiant energy
has the advantage that it is independent of the
optical medium through which the radiant energy
passes. However, the relation between frequency
and wavelength is well established, particularly
for the most common medium, air. If the
frequency is denoted by v and the wavelength by
A, we have
[1(4.3.1)]
with the velocity of light in vacuum given by
c = 2.99792458 x 10°m-s_!
[see Section 1.2.2(ii)] and the refractive index of
standard air given approximately by
n, = 1.00028
throughout the visible spectrum (Edlén, 1966).
Standard air is defined as dry air at 15°C and
101,325 Pa, containing 0.03% by volume of carbon
dioxide. It follows that v = 540 x 10!” Hz corre-
sponds to A = 555.016 nm in standard air.
In comparing the new definition of the candela
with those that preceded it, we note that it offers
more freedom to a laboratory wishing to realize
the candela. For example, it is now possible to
choose a method that either uses a blackbody
256 PHOTOMETRY

radiator operating at a given temperature or that Table 1(4.3.2) Abridged Values of

employs an electrically calibrated radiometer. Standard Luminous Efficiency Func-
The unit of luminous flux is the /umen (1m), tions V(A) and V’(A) for Photopic and
which is derived from the candela (cd), the photo- Scotopic Vision, Respectively
metric base unit by the following relation:
The lumen is the luminous flux emitted within Wavelength Photopic Scotopic
unit solid angle fone steradian (sr)] by a point 4(nm) V(A) V"(A)
source having a uniform luminous intensity of 380 0.0000 0.0006
one candela, thus 390 0001 0022

460 0004 0093

Im = cd - sr! [2(4.3.1)] 410 0012 0348
420 0040 0966
An equivalent definition of the lumen, which 430 0116 1998
440 0230 3281
avoids direct reference to the candela, is as fol-
lows: 450 0380 4550
The lumen is the luminous flux of monochro- 460 0600 5670
470 0910 6760
matic radiant energy whose radiant flux is 1/683
480 1390 7930
W and whose frequency is 540 X 10’? Hz. 490 2080 9040
This definition would be used if the lumen
were to become the photometric base unit. 500 3230 9820
510 5030 9970
520 7100 9350
4.3.2 Standard Photometric Observers 530 8620 8110
540 9540 6500
In accordance with the photometric principle out-
lined in Section 4.2, two standard photometric 550 9950 4810
observers have been established, one char- 560 9950 3288
570 9520 2076
acterizing photopic vision, the other, scotopic vi- 580 8700 1212
sion. 590 7570 0655
The standard photopic observer was estab-
600 6310 0332
lished by the CIE in 1924 and is defined by the 610 5030 0159
spectral luminous efficiency function V(A) for 620 3810 0074
photopic vision (CIE, 1924). The standard 630 2650 0033
scotopic observer was established by the CIE in 640 1750 0015
1951 and is defined by the spectral luminous 650 1070 0007
efficiency function V’(A) for scotopic vision (CIE, 660 0610 0003
1951). 670 0320 0001
680 0170 0001
Table 1(4.3.2) gives abridged values of V(A) 690 0082 0000
and V’(A) for the wavelength range from A = 380
to 780 nm at 10-nm intervals. 700 0041 0000
The official values of V(A) are given for the 710 0021 0000
720 0010 0000
wavelength range from 360 to 830 nm at one-nm 730 0005 0000
intervals. These values are identical to those defi- 740 0003 00C0
ning the CIE 1931 color-matching function y(A),
750 0001 0000
as discussed in Section 3.3.1, and given in Table 760 0001 0000
(3.3.1). The official values of V’(A) are given for 770 0000 0000
the wavelength range from 380 to 780 nm at 780 0000 0000
one-nm intervals as displayed in Table I(4.3.2) in
the Appendix. Figure 1(4.3.2) illustrates the two
spectral luminous efficiency functions defining nm for V(A) and at A = Xj, = 507 nm for V’(A).
the standard photopic and standard scotopic ob- The two functions are scalar or weighting func-
server, respectively. tions; that is, they have no dimensions.
The spectral luminous efficiency functions Dimensions are introduced through the basic
V(A) and V’(A) are relative functions of wave- equations of photometry given as Eqs. 3(4.2) and
length, that is, functions normalized to have max- 4(4.2) which relate photopic luminous flux P, to
imal values of unity. This occurs at A = A,, = 555 radiant flux P,, and scotopic luminous flux P’ to
 The Standard Photometric System — 257

0.8 AYR
eae al a
soe a Ea
e/a
EFFICIENCY
LUMINOUS
0.4

Breae WAVELENGTH \ (nm)

Fig. 1(4.3.2). Spectral luminous efficiency functions V(A) and V’(A) defining the standard
photopic and standard scotopic observer, respectively. Each function is normalized to have a
maximum value of unity which occurs at A,,=555 nm for V(A) and at X/,,=507 nm for
VA).

radiant flux P., respectively. In accordance with and

these equations, the ratios
’ = 1700 lumen per watt (Im - W~')
a aa
K(A) = P. » [3(4.3.2)]

and The normalizations used for V(A) and V’(A) make

K,, and K’, the maximum photopic and scotopic
Boe luminous efficacies, respectively. Figure 2(4.3.2)
K’(A) ==> [1(4.3.2)]
View
illustrates K(A) and K’(A).
As noted in Section 4.3.1, the lumen can be
are equal to defined as the luminous flux of a monochromatic
radiant energy whose radiant flux is 1/683 W and
Ro eh oi) whose frequency is 540 X 10'* Hz. This frequency
corresponds to a wavelength A, = 555.016 nm, in
and standard air. It follows, by definition, that the
luminous efficacy of radiant flux of this wave-
K'(A) = Ki V'(A) [2(4.3.2)] length equals 683 Im - W' for both photopic and
scotopic vision, that is,
which define the photopic luminous efficacy and
scotopic luminous efficacy, respectively, for mono- K(A,).=K'(A,) = 683 lm W
chromatic radiant flux P, , of wavelength A. The
factors K,, and K/, are fixed in accordance with with
the new definition of the photometric base unit
and given as Az = 555.016 nm

K,,, = 683 lumen per watt (Im - W~') At the wavelength, A,, the photopic and scotopic
258 PHOTOMETRY

1800
a— Kin 1700 Im: W
1600

1400

8 _— (scotopic ) K(x) = KimV (aA)

Km= 683 Im: Ww!

ee

(photopic ) K(A)= KmV(\)

(Im-W7!)
Efficacy
Luminous

400 500 600 700

Wavelength A (nm)
Fig. 2(4.3.2). Spectral luminous efficacy functions K(A) and K’(A) for photopic and scotopic
vision, respectively.

efficiencies are subtense [see Gibson and Tyndall, (1923) for a

V(Aq) = 0.999998, and V'(Aq) = 0.401750
found by linear interpolations in Tables 1(3.3.1)
and 1(4.3.2), respectively. From Eq. 2(4.3.2) it
follows:
K(A
ie ChE) a 683.002 Im - W!
V(Aq)
K’(X
Pa f t) = 1700.06 1m - W~!
V'(Aq)
These precise values of K,, and Kj, derived in
strict correspondence with the definition of the
lumen (or candela) and the V(A) and V’(A) func-
tions, can safely be rounded to 683 and 1700,
respectively, to be adequate for all practical appli-
cations (CIE, 1982).
The V(A) function was derived from the mean
results of several experimental determinations of
the relative luminous efficiencies of spectral
stimuli mainly by the flicker and step-by-step
methods of heterochromatic photometry (see
Section 5.7) using large and separate groups of
observers. The observations were made with
matching fields of two to three degrees angular
review of the earlier work and of the derivation of
a composite V(A) function]. In the light of later
measurements, a standard V(A) curve differing
somewhat from the present curve in the short-
wavelength end of the spectrum would probably
have been adopted today. But any minor im-
provement at this stage would be outweighed by
the very considerable practical inconvenience of a
change in the basic function on which all photo-
pic photometry has been based for more than 50
years.
The V’(A) function rests on measurements
of the scotopic luminous efficiency of monochro-
matic stimuli by the method of direct visual com-
parison in a large matching field of low brightness
(Crawford, 1949) and on visual thresholds for
monochromatic stimuli viewed eccentrically
(Wald, 1945b). [See LeGrand (CIE, 1951) for a
review of the work culminating in the agreed
V’(A) function.] The V’(A) function is intended
to represent as nearly as possible the mean spec-
tral sensitivity of the rod mechanism for observers
whose age does not exceed 30, when observing at
angles of not less than 5° from the foveal center
under conditions of complete dark adaptation.
Observer age is more significant in relation to
VA) than V(A), because the maximum of the

former function is shifted nearer to the blue and
to the region of greatest absorption of the yellow
pigments, which in the aging eye are increasingly
present in the prereceptor path (see Section 2.4.6).
However, the limited practical applications of
V(A) make it unlikely that a standard function
representative of an age range more comprehen-
sive than the present standard function will be
required.
4.3.3 Photometric Methods
In accordance with Table I(4.1) and its supple-
ment, Table 1(4.1), the photopic photometric
quantities, whose measurements are most often
required in practice, are luminous flux P,,
luminous intensity 7,, luminance L,,, and illumi-
nance E,,. All four quantities have similar defi-
ning equations, that is,
Luminous flux:
P= Ky,fPe V(A)an [1(4.3.3)]
Luminous intensity:
ES Ky fTe.VOA)ad [2(4.3.3)]
Luminance:
Lt Ky fLe V(d) ar [3(4.3.3)]
Tlluminance:
jr ey, ifE, ,V(A)dd_ [4(4.3.3)]
in which the radiometric quantities P, ,, I, ,
L..,, E.,., are, respectively, the spectral con-
centrations of radiant flux (or radiant power),
radiant intensity, radiance, and irradiance, which
have all been defined in Table I(1.1) and its
supplement, Table 1(1.1).
The defining equations for the corresponding
scotopic photometric quantities are, in accordance
with Eq. 4(4.2), readily obtained by adding primes
to the appropriate symbols. However, our re-
maining discussions will be confined to photopic
photometry only.
There are three main ways of determining the
above photopic photometric quantities in terms of
their respective units. A brief description of each
is given below (CIE 1982).
The Standard Photometric System 259
(i) Broad-Band Photometry. Broad-band pho-
tometry is the most widely used method. It
involves as its main hardware component a pho-
tometer which consists of either a thermal detec-
tor or (more commonly) a photon detector (see
Section 1.6) whose relative spectral responsivity
has been modified, often referred to as “cor-
rected,” to approximate the V(A) function. Typi-
cally, this approximation is accomplished by
means of a specially designed combination of
absorption filters with appropriate spectral trans-
mittance characteristics. The procedure of desig-
ning a V(A)-filter is similar to that used in desig-
ning tristimulus-filters for colorimeters, described
in Section 3.12.5.
In addition to the V(A)-corrected detector,
suitable optics, apertures, and possibly an in-
tegrating sphere are needed to provide the
geometrical configuration associated with the par-
ticular photometric quantity to be measured.
The detector output signals are usually cali-
brated by means of lamps that have been stan-
dardized in terms of the quantity the photometer
is designed to measure. Modern photometers can
be very precise and reproducible, but their accu-
racy is often limited when lights of dissimilar
spectral radiant power distributions are intercom-
pared. The deviations from accurate measure-
ments are usually caused by residual differences
between the actual spectral responsivity of the
photometer and the standard V(A) function.
(ii) Spectroradiometric Photometry. Spectrora-
diometric photometry is the basic method of pho-
tometry in which the photometric quantity (e.g.,
luminance of the light to be measured) is
determined by first measuring the spectral con-
centration of the appropriate radiometric quan-
tity (e.g., radiance, of the given light), and then
calculating the desired photometric quantity
(luminance) in accordance with the appropriate
equation [Eq. 3(4.3.3)] involving the standard
V(A) function. In principle, this procedure corre-
sponds to that used in the determination of the
colorimetric quantities of a given color stimulus,
as described in Section 3.12.1. However, in col-
orimetry, it is rarely necessary to know the spec-
tral distribution of the radiometric quantity in
absolute terms, such as [W-m *- sr ']. Only
relative spectral distributions are required in most
applications, simplifying somewhat the spectrora-
diometric task. In photometry, on the other hand,
absolute spectroradiometric measurements are
260 PHOTOMETRY
virtually always needed. This is usually accom-
plished by making use of standards of spectral
radiance and standards of spectral irradiance.
In general, standards of spectral radiance are
gas-filled, tungsten-ribbon-filament lamps having
a plane glass or silica window and operated base
downwards at a color temperature of about 2700
K. The calibration of such a lamp relates to a
specified area of the filament and a specified
direction, usually horizontal approximately nor-
mal to the filament. The actual calibration proce-
dure employed to obtain a standard of spectral
radiance is rather involved and normally the do-
main of national standardizing laboratories (see,
for example, Stair et al., 1960, Sanders and Gaw,
1967, Kostkowski et al., 1970).
A blackbody radiator operating at the freezing
point of platinum, or at an even higher tempera-
ture, is often used as the reference source, because
its spectral radiance distribution can be calcu-
lated with satisfactory accuracy by means of
Planck’s radiation formula (see Section 1.2.2).
(iii) Visual Photometry. Visual photometry is
rarely used in routine photometric measurements.
It suffers from low precision, and the visual judg-
ments made by individual observers differ some-
what from one observer to the next and from
those that the (hypothetical) standard photomet-
ric observer, characterized by V(A) and V’(A),
would make.
However, the techniques of visual photometry
are still of importance to the researcher who is
interested in the basic concepts of light measure-
ments as it relates to human vision. Research
in this field has gained momentum in recent
years, particularly with regard to heterochromatic
brightness matching (see Section 5.7). Therefore,
we considered it appropriate to include in this
book some notes on visual photometry.
If it is given that the measurement to be made
is reduced to a brightness match between two
stimuli of similar relative spectral radiant power
distributions, the visual photometer comprises a
device for presenting to the eye the two lght
patches as symmetrically disposed areas in juxta-
position and a means for varying in a known way
the luminance of at least one patch. In the classic
method of comparing the luminous intensities of
lamps, the two sources can be moved along a
photometric bench, and their distances d, and d,
from two similar, diffusely reflecting surfaces S,
and S, contained in the photometer head and
illuminated by the respective sources, can be ad-
justed until the observer judges the surfaces to
match in brightness. Figure 1(4.3.3) illustrates the
Lummer-Brodhun Photometer Head (probably
the best type) and Figure 2(4.3.3) the bench
arrangements.
By the inverse-square law of illumination (see
also Section 4.3.5), the illuminance of the two
surfaces S, and S, will be I,/d? and I,/d3,
respectively, where 7, and J, are the luminous
intensities of the sources. At match, the corre-
sponding luminances are equal, so that
al et CE) [5(4.3.3)]
P17) 1, P2771,
where p, and p, are the reflectances of S, and S),
and 7, and 7, are the transmittances of any optical
parts (other than S, and S,) intervening between
the light source and the eye. For a perfect pho-
tometer p, =p, and 7, =7, exactly, and the
match condition is then
Fade 6
ary [6(4.3.3)]
The construction of the Lummer-—Brodhun
head enables the roles of S,; and S, to be reversed,
and, by taking means of duplicate readings for
the two arrangements, residual differences in p,
and p, are eliminated. Left—-right asymmetries in
the observing eye are similarly removed by a
second reversal procedure. It is essential that any
light other than the direct light from the ap-
propriate source is prevented by suitable baffles
from reaching the surfaces S, and S,. In the
accurate determination of the distances d, and
d,, allowance must be made for the reduction of
optical path length if a length e of the beam is
through glass of index of refraction n (e.g., if a
glass filter is present). The corrected d,, for exam-
plepis*thend)= (e—“by/nye.
The eye can detect a luminance difference of
about 1% between two light patches of the same
relative spectral radiant power distribution pre-
sented in a well-designed matching field, provided
the luminance level is adequate (above about 100
td). When asymmetries in the photometer head
and other systematic errors are eliminated,
luminous intensities obtained as means of sets of
repeated settings (e.g., 10) may reach an accuracy
of 0.2% (Walsh, 1958).
The use of the inverse-square law of illumina-
tion in making matches in a photometer may be
replaced or supplemented by the insertion in the
 The Standard Photometric System 261

WHITE SURFACES

FROM TEST SOURCE Ss;

FROM COMPARISON SOURCE
La 4

FIELD OF VIEW

EYE PIECE

LUMME
R- BRODHUN
CUBE

EVE

Fig. 1(4.3.3). Schematic diagram of a Lummer-Brodhun photometer head to judge equality of

luminance between two surfaces $, and S,. The Lummer-Brodhun cube inside the photometer
head consists of two right-angled prisms. One of the prisms has a principal surface polished into
spherical shape, except for a small central portion which is flat and in optical contact with the
central portion of the flat principal surface of the second prism. From the light entering the
prism face AB, only the part falling on the portion of AC, not in optical contact with EF,
reflects totally at AC and emerges at BC. From the light entering the prism face ED, only the ©
part falling on the portion of EF, in optical contact with AC, passes through the cube
undeviated and emerges at BC. The geometrical configuration of the visual field as seen by the
observer’s eye depends on the shape and exact location of the portion of the prism surface EF
that is in optical contact with AC. The visual field produced by the cube illustrated in this
figure is shown as an insert, but a centrally divided bipartite field has certain advantages.

light path of neutral (nonselective for wavelength)
filters or wedges, or sector disks run at speeds
high enough to eliminate visible flicker. The sec-
ond device has the merits of being (a) strictly
neutral and (b) capable of calibration to a fair
accuracy by length measurements on the disk that
determine the relative angular widths of the clear
and occulting sectors.
In visual measurements of luminance and il-
luminance, a photometer that employs the in-
verse-square law, provides, in effect, a variable
luminance in one-half of the matching field, and
the other half is occupied by the light of the
surface whose luminance, or illuminance, is being
measured and at which the instrument is pointed.
The luminance calibration will generally be made

PHOTOMETER
HE AD

SOURCE| SOURCE 2
eee eR BE 2 See Ee ee ee we
s N :

Fig: 2(4.3.3). Schematic diagram of photometer bench and photometer head used in visual
photometric measurements.
262 PHOTOMETRY

TEST PLATE
LUMMER-BRODHUN CUBE
(DEPOLISHED OPAL GLASS)

COMPARISON SURFACE
(OPAL GLASS)

LAMP IN
LAMP HOUSE

RACK AND
PINION

CALIBRATED
SCALE

Fig. 3(4.3.3). Schematic diagram of the Macbeth illuminometer. A lamp, movable along the
tube by means of rack and pinion, illuminates an opal glass serving as a comparison surface.
The test plate, consisting of a depolished opal glass, is placed at a convenient distance outside
the instrument. Its luminance is compared with that of the comparison surface by means of the
Lummer-Brodhun cube viewed through the eyepiece. A luminance match is obtained by
moving the lamp. The scale is calibrated to read the illuminance of the test plate directly, for
example, Im-cm7. The range of measurement can be increased by inserting calibrated neutral
density filters either at F, or F,. The lamp is operated at a fixed current, or better at a fixed
resistance using a Wheatstone bridge circuit, battery and meter being kept in a separate case.

by matching on a diffusely reflecting test plate of a satisfactory diffusely reflecting test plate is im-
known reflectance p,, illuminated by a standard portant for accuracy and presents some difficulty
lamp of luminous intensity J, at a measured dis- (see Section 1.4.2). The Macbeth Illuminometer
tance d, (m). The luminance is given by [see Figure 3(4.3.3) for diagram and description]
is an example of a portable instrument for visual
dink Reayik
Boric (cd - m~*) (7(4.3.3)] measurements of luminance and illuminance.
When the two lights to be compared are differ-
If illuminance in a given plane is to be measured, ent in color, brightness matches may not corre-
the test plate is placed in position in the plane spond to equality of luminance. Also the basic
and a luminance measurement L is made. The laws of brightness matching that are assumed to
illuminance E is then given by govern the standard standard photometric system
(see Section 4.2) may no longer hold. These diffi-
culties can be greatly reduced by making use of a
E=—L [8(4.3.3)]
p, flicker photometer.
In the flicker photometer, the criterion of
For both kinds of measurement, the provision of equality of luminance of two geometrically simi-
 The Standard Photometric System 263

Channel |

Synchronized
Shutters

ff
Channel 2 Beamsplitter
Cube Y Field of View

—— | | — Open
channel! f[ \ [ \_,
! : — Closed
| | |
|
Time Sequence of Shutters
Channel 2 \ |
Closed

Fig. 4(4.3.3). Schematic diagram showing a two-channel photometer, equipped with two
synchronized shutters or rotating sectors, which provides in the field of view two alternating
patches of light. The two slits may be the exit slits of two monochromators through which
monochromatic radiant flux of wavelength A , and A, emerge, respectively. At the bottom of the
figure, the time sequence is given at which the two light patches are presented to the eye. If
spectral luminous efficiencies are measured by this technique, minimum flicker rates are
typically between 10 and 40 Hz, depending on the wavelengths of the lights and other observing
conditions.

lar light patches is the disappearance of the flicker
produced by presenting them alternately to the
eye at a certain minimum frequency. The mini-
mum frequency is defined as the lowest frequency
that will just cause flicker to disappear over the
smallest possible range of variation of either lumi-
nance alone.
The usefulness of the method of flicker pho-
tometry rests on the visual phenomenon that a
chromatic difference between the two alternating
light patches disappears at a lower frequency than
a brightness difference.
For alternating lights of the same chromatic
appearance, flicker disappears when they are
equally bright as judged by the steady comparison
in a bipartite field. However, when there is a
chromatic difference, flicker may disappear when
photometric laboratories, some of these have been
described by Walsh (1958). Figure 4(4.3.3) il-
lustrates a system that provides flickering light
patches by means of synchronized shutters or
synchronized rotating sectors inserted in a two-
channel visual photometer.
4.3.4 Measurement of Total Luminous Flux
The fundamental method of determining the total
luminous flux P,, emitted by a source is by measur-
ing its /uminous intensity I,(0, >) in a large num-
ber of directions (9, ¢) uniformly distributed in
space about the source, and then integrating the
measured luminous intensities over the total solid
angle of 47 sr. The defining equation is
7 f20 ‘

the two light patches are unequal in brightness as

judged by the steady (or direct) comparison
method. The magnitude of the discrepancy be-
tween the results of the two methods depends to
some extent upon the conditions of observation.
In Section 5.7, experimental data are given and
discussed.
There are several different designs of flicker
photometers that have been used in the past in
eG Ea 1,(8,)|sin@| d0 de
[1(4.3.4)]
where @ is the polar angle and ¢ is the azimuth
angle in a spherical coordinate system whose
origin is at the lamp.
Alternatively, the illuminance E,(8,) pro-
vided by the source is measured at points (6, ¢)
 264 PHOTOMETRY
on the surface of an imagined sphere of radius r
centered at the source. The defining equation is
then
ante E(0,$)|sin0| d0 do
7 fla 3
[2(4.3.4)]
The measurements are usually made by a
broad-band photometer [V(A)-corrected] ap-
propriately mounted so that it can be rotated
about the lamp. Such an assembly is called a
goniophotometer and is illustrated in Figure
1(4.3.4). There are different goniophotometer de-
signs and methods of sampling the points of
measurement about the source (Walsh, 1958;
Krochmann and Marx, 1969; Jones and Berry,
1970; Kurioka, 1971; Schaefer and Mohan, 1974;
Brown, 1979).
Goniophotometers are usually large instru-
ments, particularly when the luminous flux is
determined by spatial summation of /uminous in-
tensity measurements as given by Eq. 1(4.3.4). The
oe Photometer
\
Fig. 1(4.3.4). Schematic diagram of a gonio-
photometer to measure the total luminous flux
emitted by a lamp.
tungsten filament lamp normally used in such
measurements must effectively become a “point
source” when viewed by the photometer; thus, a
considerable distance is required between the lamp
and the photometer head. However, when using
the approach characterized by Eq. 2(4.3.4), where
illuminance measurements are made about the
lamp, a significantly shorter distance between the
lamp and the photometer head can be used, mak-
ing the goniophotometer more compact (Brown,
1979).
The described goniophotometer determination
of the total luminous flux emitted by a lamp is
the approach taken by national standardizing
laboratories to establish national scales of total
luminous flux. The lamps used for this purpose
are tungsten-coil-filament lamps carefully selected
for stability of output. Vacuum lamps are used at
color temperatures up to 2400 K (approximately),
and gas-filled lamps at higher color temperatures.
The bulbs may be clear or diffusing, and the
lamps are usually operated base upwards. The
main application of these standards of total
luminous flux is for calibrating other lamps by
means of a substitution method involving an in-
tegrating sphere.
The successful application of an integrating
sphere in such a measurement is based on the
following principle: If a source is placed inside a
hollow sphere coated internally with a perfectly
diffusing material, the illuminance on any part of
the surface due to light reflected from the re-
mainder of the sphere is the same and propor-
tional to the total luminous flux emitted by the
source. One has
| he aap
oP, er [3(4.3.4)]
where
E,,=illuminance; that is, total luminous flux re-
ceived per unit area of sphere wall due to
reflection and multiple interreflections only
P,=total luminous flux emitted by source inside
sphere |
r=radius of sphere
p=reflectance of sphere wall
Equation 3(4.3.4) refers to an ideal integrating
sphere and is only an approximation for any
actual sphere. The reflectance p of the sphere wall
is neither strictly constant nor perfectly diffusing,
and it is also wavelength dependent even for the

best coating materials available, such as barium
sulphate paint. Further, the reflected flux is not
completely undisturbed; the sphere contains ob-
jects such as the source itself and baffles, and
usually there are one or more windows in the
sphere wall that produce further deviations from
the ideal. Nevertheless, by applying suitable
methods and corrections, photometrists obtain
satisfactory results with the integrating sphere
(Walsh, 1958; Keitz, 1971). This is particularly so
if the substitution method is used [Figure 2(4.3.4)].
A standard lamp of known luminous flux is placed
in the sphere, and the flux incident at some small
part of the sphere wall is measured by a broad-
band photometer. The measurement is repeated
with the test lamp substituted for the standard.
Proper baffling inside the sphere is important so
that no direct light but only the reflected light
reaches the specified part of the sphere wall where
the measurement is made. The ratio of the two
measurements gives the ratio of the luminous
fluxes.
Formulae for the effects of baffles, windows,
and other sphere details are given, for example, in
the books by Ulbricht (1920), Walsh (1958), and
Keitz (1971), and a general treatment of the the-
Integrating
Sphere
SSS
Baffle Photometer
Fig. 2(4.3.4). Measurement of total luminous flux
by means of the substitution method in an integrat-
ing sphere. The small window, shielded by a baffle
from direct illumination by the lamp, may be a
translucent diffusing window at which the photom-
eter is pointed. Two measurements are made, one
with the test lamp in place, the other with the
standard lamp of known total luminous flux. From
the two measurements, the total luminous flux of
the test lamp is readily deduced.
The Standard Photometric System 265
ory of the integrating sphere in the papers by
Jacquez and Kuppenheim (1955) and Fussel
(1974). (Note that in the German literature the
integrating sphere is usually referred to as the
“Ulbricht-Kugel.”’)
4.3.5 Measurement of Luminous Intensity
and Illuminance
In accordance with Tables I(4.1) and 1(4.1),
luminous intensity /,, refers to that portion dP,, of
the total luminous flux P, emitted by a point
source, which propagates, uniformly distributed,
through a small cone of solid angle dw whose
apex is at the source and whose axis is in a given
direction. Figure 1(4.3.5) illustrates the concept.
The luminous intensity J is then defined as the
quotient of the luminous flux dP (measured in
Im) over the solid angle dw (measured in sr); thus
(omitting the subscript v):
Tiea alsa (ims =f 3} [1(4.3.5)]
The unit Im - sr' is identical to the candela, the
base unit of luminous intensity:
Ped =) im sp™"
An equivalent definition of luminous intensity
makes use of the quantity i//uminance E,, which in
the context of Figure 1(4.3.5), is simply defined as
the quotient of the luminous flux dP over the
elementary area dA illuminated by the source. In
this definition, we assume that the area dA is
perpendicular to the direction (axis of cone) of
illumination; thus,
dP
E mrniLsS ae [2(4.3.5)]
We now recall that the solid angle dw is ob-
tained from the quotient dA over the square of
the distance r between the source and dA. The
solid angle is measured in terms of the steradian,
srof2 ol
Point Source
Fig. 1(4.3.5). Concept of luminous intensity.
266 PHOTOMETRY

a unit that clearly does not follow from the opera-
tional definition of the solid angle; that is, dw =
dA/r*. In order to use the quotient d4/r? in
mathematical operations and retain its imposed
dimension of steradian (sr), we introduce the di-
mensionality factor Q whose scalar value is unity
but whose dimension is the steradian. Thus, we
write
do=4Q(sr) — [3(4.3.5)]
By inserting Eq. 3(4.3.5) into Eq. 2(4.3.5), we
obtain
dP T
~ Q=—2 [4(4.3.5)]
r? dw r-
whose dimension is cd - m7: sr =1Im- m ’, as
it should be in accordance with Eq. 2(4.3.5). In
practice, the dimensionality factor Q is often
ignored and we then find Eq. 4(4.3.5) written
simply as
n
tel =?
[5(4.3.5)]
The correct dimension (or unit) of E, is not
immediately obvious; in fact, the wrong dimen-
sion of (cd - m *) is deduced from Eg. 5(4.3.5).
Equation 4(4.3.5), or Eq. 5(4.3.5) properly in-
terpreted, states that the illuminance at a surface
element, perpendicular to the line joining the
point source and the point at the center of
the surface, is equal to the luminous intensity of
the source in the direction of the surface element,
divided by the square of the distance between the
source and the surface element.
An alternative, somewhat simplified, statement
is as follows: The illuminance at a surface ele-
ment is proportional to the luminous intensity of
the illuminating source, and inversely propor-
tional to the square of the distance from the
source. It is for this reason that Eq. 5(4.3.5) is
often referred to as the inverse-square law of il-
lumination for normal incidence.
The lux (Ix) = (Ix - m ”) has been chosen as
the SI unit of illuminance and is defined, in
accordance with Eq. 2(4.3.5), as the illuminance
produced by one lumen of luminous flux on one
square meter of surface.
On the other hand, the important inverse-
square law of illumination, expressed by Eq.
5(4.3.5), leads to a unit, sometimes called the
meter-candle (or foot-candle, when the British
nonmetric system is used), which is defined as the
illuminance produced by one candela of luminous
intensity on a surface that is one meter away from
the point source. The meter-candle (or foot-can-
dle) is still used by some illuminating engineers,
and at times, this leads to confusion. It is recom-
mended that, for the sake of uniformity of prac-
tice, the SI unit, the lux (or lm- m7”) be used
exclusively as the unit for illuminance.
Equations 2(4.3.5) and 5(4.3.5) apply to nor-
mal incidence of radiant flux on the surface ele-
ment dA, emitted by a point source from a dis-
tance r, leading to the illuminance E,,. The illumi-
nance F for oblique incidence is given by
E=E,cose [6(4.3.5)]
where ¢ is the angle of incidence measured from
the normal to the surface element dA, as shown in
Figure 2(4.3.5). Equation 6(4.3.5) follows readily
when one considers illuminance as a surface den-
sity of luminous flux, which at normal incidence
has its maximum value of E, =dP/dA. The
surface density decreases with the cosine of the
angle of incidence because the surface element,
delineated by the fixed cone of luminous flux,
increases in area by the same cosine value; thus,
dP /dA’ = (dP /dA)cos e.
Equation 6(4.3.5) expressed in terms of
luminous intensity becomes
aed
E= Sor [7(4.3.5)]
dP
oo
rom
point source
'
dA’=dA / cose
Fig. 2(4.3.5). Oblique angle of incidence («) of
luminous flux dP on surface element dA’.
 The Standard Photometric System 267

In this form it is often referred to as the inverse-
square law of illumination for oblique incidence.
Lamps usually used as standards of luminous
intensity are tungsten filament lamps, operated
base downwards, and with the
made in a specified direction. The filament is
usually of the uniplanar type using welded con-
nections between the ends of the filament and the
lead-in wires, but omitting intermediate filament
supports. The bulb is clear, and the specified
direction of measurement is normal to the fila-
ment plane. Vacuum lamps are used for color
temperatures up to about 2400 K, and gas-filled
lamps at higher color temperatures.
These lamps can also be used as standards of
illuminance at a specified distance from the plane
of the filament.
The calibration of these lamps is performed
measurements either with reference to the blackbody radiator at
the freezing point of platinum, which was the
primary standard of light prior to 1979 (see Sec-
tion 4.3.1), or by means of an electrically
calibrated radiator equipped with a V(A)-filter.
The method of using the blackbody radiator to
calibrate standards of luminous intensity is il-
lustrated in Figure 3(4.3.5). The lower part of this
figure shows a cross section of a conventional
type of blackbody radiator operating at the freez-

| D, STANDARD LAMP
PHOTOMETER BG |

LENS
z

| D,
PHOTOMETER ha wo- ---- - - - ee - - = ee ee ee eee eee er
PRISM

SIGHT TUBE
LENS - seinen oa APERTURE
! j ieee

PORCELAIN CRUCIBLE

THORIA POWDER
THORIA CRUCIBLE
PLATINUM INGOT
(4.5CM LONG x 2.0 CM DIA.)

SIGHT TUBE /3
FULL OF THORIA
POWDER

INDUCTION HEATER COIL ener

Nor
Fig. 3(4.3.5). Schematic diagram showing method of using a blackbody radiator at the
freezing point of platinum (formerly the primary standard of light) to calibrate a standard lamp
of luminous intensity (from Sanders and Jones, 1962).
268 PHOTOMETRY
ing point of platinum. In this arrangement, the
illuminance per unit reading of the photometer is
obtained at the aperture plane of the photometer.
In the upper part of the figure, the blackbody
radiator is replaced by the lamp to be calibrated
as a Standard of luminous intensity, and again the
illuminance per unit reading of the photometer is
obtained at the photometer aperture. By equating
the two illuminances, one arrives at the luminous
intensity J of the lamp as follows:
nage | 45 [8(4.3.5)]
Be,
where the various symbols represent the following
quantities:
A =area of lens diaphragm in square centimeters
T=transmittance of lens and prism (measured
in an auxiliary experiment)
R,=photometer reading obtained while the
platinum around the blackbody is freezing
D,=distance from photocell aperture to lens
diaphragm
R,=photometer reading for standard lamp
D,=distance from lamp filament to photocell
aperture
The number 60 is the number of candelas per
square centimeter emitted by the blackbody radi-
ator (see Section 4.3.1).
In view of the recent change of the definition
of the candela (Section 4.3.1), it is unlikely that
the above method of calibrating standards of
luminous intensity will continue to be used by the
national standardizing laboratories. Instead, the
now well developed electrically calibrated radi-
espa ap Trap
Fig. 4(4.3.5). Schematic diagram showing method of using an electrically calibrated radiome-
ter to calibrate a standard lamp of illuminance (Brown, 1975; Boivin and Smith, 1978).
ometers, equipped with a V(A)-filter, are ex-
pected to be used for such a basic calibration.
The method of using an electrically calibrated
radiometer, equipped with a V(A)-filter, is il-
lustrated in Figure 4(4.3.5). The apparatus is set
up to measure the illuminance E,, produced by
the lamp at the aperture of the radiometer. The
essential steps in the procedure are the following
(Brown, 1975):
First, it is recalled that the illuminance E,, is
defined by
ES Ke fE, ,V(A) dd [9(4.3.5)]
where K,,, is, by definition, equal to 683 lm - W |,
and E,, is the spectral concentration of the
irradiance produced by the lamp at the radiome-
ter aperture without the V(A)-filter in place.
When the V(A)-filter is in place, the spectral
irradiance at the aperture is given by
Enea pe, ray( 2) [10(4.3.5)]
where 7(A) is the spectral transmittance of the
filter, d is the distance from the lamp filament to
the aperture plane, and A is a small correction
dependent on the refractive index and thickness
of the filter. The total irradiance E’, given by
e=(Z4) fen 1(A) dd 111(43.5)]
which is the quantity measured with the V(A)-
filter in place.
V(A) ~ Filter
Thermal Insulation
|ecteeqmn!)
X To Electrical
Measuring
Equipment
Electrically Calibrated
Radiometer

It then follows that the illuminance E,, is re-
lated to the measured irradiance E/ by
r= nx,(4=4)
he 2 [Bc aV(A) ar
On a Se
[Bcat(A) da
[12(4.3.5)]
The ratio of the two integrals in the above equa-
tion allows the use of the relative spectral distri-
bution of irradiance {S(A) dA} instead of the
absolute distribution {E, ,dA}. This proves to
be an important feature of this method, as it is
less difficult to measure the relative spectral dis-
tribution of irradiance produced by the tungsten
filament lamp. Moreover, in the design of the
required V(A)-filter, the spectral transmittance
curve T(A) can be optimized so as to minimize the
effect of errors in {S(A) dA}.
The standard of illuminance calibrated by the
above method can also be used as a standard of
luminous intensity, provided the inverse-square
law of illumination can be applied to it accu-
rately.
The inverse-square law holds strictly when the
lamp filament constitutes a point source. In prac-
tice, this is never the case. However, it is possible
to apply the inverse-square law provided the dis-
tance between the filament and the surface receiv-
ing luminous flux from the filament is large com-
pared to the physical dimensions of the filament.
This will make the filament effectively a point
source as seen from the distant receiving surface.
This condition is sometimes expressed by the
relations
T= lim Er [13(4.3.5)]
r=
7 OO
Or
ead MP
ey a dw
recalling that dw =dA/r* [and ignoring the
mathematical difficulty of dimensionality dis-
cussed in connection with Eq. 4(4.3.5)].
In practice, the distance r is limited to finite
values, and this introduces a systematic error in
the value of J derived from E. However, in most
practical cases arising in a photometric labora-
tory, an appropriate distance can be found that
The Standard Photometric System 269
leads to very small or negligible deviations from
the inverse-square law. The obvious experimental
test as to whether a chosen distance r is suffi-
ciently long, is to determine J from E measured at
r, and J’ from E’ measured at r’ > r. If J and J’
agree with one another within the precision of
measurement, or within the tolerance demanded
by the application, the distance r is satisfactory.
A precise theoretical estimate of what con-
stitutes a satisfactory distance r in a practical
situation is difficult to make (Walsh, 1958; Keitz,
1971; Nicodemus, 1978). In making such an
estimate, one must consider the shape and size of
the source, the angular distribution of radiance
(or luminance) from each surface element of the
source, and similarly, the size and shape of the
detector surface and its angular and spatial re-
sponsivity. Only relatively simple and somewhat
idealized designs of source and detector can be
treated in a mathematically useful way (see also
Section 4.3.7). However, there is a common rule
of thumb in photometry that states that the in-
verse-square law of illumination will be applica-
ble to a source of area A, and a detector of area
A p, if the distance r between A, and A, is at least
10 times the largest of the maximum transverse
dimensions of A, and Ap. The error that results
from such an approximation to the inverse-square
law will then be less than 1%.
There are many broad-band photometers com-
mercially available that measure illuminance. The
essential component of an illuminance meter, also
called a /ux meter, is a photon detector, often a
silicon photodiode operating in the photovoltaic
mode (see Section 1.6.2), whose spectral respon-
sivity has been corrected by means of a suitable
V(A)-filter. The output signals of the detector can
be read from a scale that has been calibrated, for
example, in terms of lux, by means of either a
standard of illuminance or a standard of luminous
intensity used together with the inverse-square
law relation of Eq. 5(4.3.5). In some illuminance
meters, the detector surface covered by the V(A)-
filter and bounded by the rim of its casing, de-
fines the surface element at which the incident
luminous flux produces the illuminance to be
measured. An instrument of this design involves
the risk of being wrongly used, because, except
for normal and near normal incidence of the
luminous flux, the output signals of the meter
usually depart from the desired cosine law, as
expressed by Eq. 6(4.3.5). The greater the angle of
incidence, usually the greater is the departure
from the cosine law. A typical performance of
270 PHOTOMETRY

Illuminance
Meter

Ideal: cos €
---- Well corrected
Uncorrected

Relative
Illuminance
of
Readout
Meter

O 30 60 90
Angle of Incidence € (degrees)

Fig. 5(4.3.5). Angular responsivity of illuminance meters. Solid curve. ideal meter follows
cos e; dashed curve: example of well corrected meter; dotted curve. example of uncorrected
meter.

such an uncorrected illuminance meter is il- meter with a silicon photodiode may provide an
lustrated in Figure 5(4.3.5). angular responsivity similar to that shown as the
It is difficult to design and construct an illumi- dashed curve in Figure 5(4.3.5). The design of
nance meter that strictly obeys the cosine law. such a meter is illustrated in Figure 6(4.3.5). In
All so-called cosine-corrected illuminance meters order to achieve an even closer fit with cos e, it is
show some deviation from cos e. A well corrected necessary to mount an integrating sphere in front
of the sensitive surface of the detector with several
baffles inserted at appropriate places in between.
| A good example of such a design is that given by
| Opal Glass
Entrance
Budde (1964), shown in Figure 7(4.3.5). However,
| viAperture Rim the improvement in the cosine response of such a
Part of photometer has been gained at the expense of its
Enclosure detectivity. The incorporation of a photomulti-
plier tube instead of a silicon photodiode has then
advantages.

V(A)- Filter
Silicon~ (5-Component
Photodiode 4.3.6 Measurement of Luminance
Filter- Combination)
The photometric quantity /uminance (L,, or sim-
ply L) is of particular interest because it corre-
Fig. 6(4.3.5). Cross section of an illuminance me- lates in some way (Section 4.2) with the perceived
ter that has been cosine-corrected by making use quantity brightness. However, from the con-
of an opal glass and carefully designed enclosure ceptual point of view, luminance is the more
with an outer rim around the entrance aperture of complex quantity of the major photometric quan-
the instrument. The detector is a silicon photodi- tities usually used; the others being luminous
ode. flux, luminous intensity, and illuminance.
 The Standard Photometric System 271

Circular Baffle luminous intensity is, in accordance with Eq.

(held in place by 3 wires)
1(4.3.5), given by
bee Integrating Sphere
_ ae
ae dP (Im-sr-') [1(4.3.6)]

The apex of the solid angle dw, can be placed at

S as Ar can be made negligibly small compared to
r; that is, Ar <r, and

dA dA
UR ee (sr)
(r= Ar) r

— Baffles
[2(4.3.6)]

V(X)
With dA, placed close to S, the luminous inten-
Filter sity of dA, is clearly that of a small area of the
source itself.
1
| Cathode a The limiting quotient dI/dA, is called the
of
luminance L at the point S of a surface element
Photomultiplier
dA, of the source perpendicular to a specified
Fig. 7(4.3.5). Cross section of an illuminance me- direction SD, that is,
ter especially designed to provide an angular re-
sponsivity closely approximating the cosine law.
The detector is a photomultiplier tube (after Budde,
ere
inating ee
(lm: srg a
amin) -
xdond) (cde-amg4)
1964).
[3(4.3.6)]
Luminance, as defined in Table I(4.1) and The inverse-square law of illumination for nor-
Table 1(4.1), is a quantity used to describe the mal incidence, as given by Eq. 4(4.3.5), allows us
luminous flux in a specified direction and at a to write
specified point on a surface element of the source
or on a surface element of a detector placed in the
specified direction at some distance away from (Im-m~*) [4(4.3.6)]
the source.
The derivation of useful mathematical re-
lations between luminance and the other major where dE,, is the illuminance produced by dA, on
photometric quantities rests on the assumption the detector element dA, at normal incidence.
that the concept of luminous intensity can be ap- With Eq. 3(4.3.6), we obtain a useful alternative
plied, within certain limits, to non-point sources; expression for the luminance; that is,
that is, sources of finite physical dimensions.
Figure 1(4.3.6) illustrates an extended source _ dE, —2.o-1
eel Fe (ime amnesty)
with a given point S, a specified direction SD,
and two opaque baffles perpendicular to SD with
small apertures AA, and AAp around SD. The [5(4.3.6)]
specified direction SD is at an angle e, from the
normal n to the surface of the source at S. Whereas where
the aperture AA, is a small distance Ar away
dA
from the point S, the aperture AA, (the detector) dwp=— (sr) [6(4.3.6)]
is placed at a considerable distance r away from r
AAs.
In making the apertures AA, and AAp, suffi- is the solid angle that dA, subtends at the point D
ciently small compared to the distance r, the in the detector element dA p.
aperture A A, effectively becomes a “point source” The defining Eq. 3(4.3.6) of luminance can be
with respect to the detector aperture AA p, and its rewritten in terms of luminous flux by recalling
272 PHOTOMETRY

Fig. 1(4.3.6). Ilustration of geometrical configuration used in the development of the concept
of luminance. Apertures AA, and 44 > are perpendicular to the specified direction SD of the
luminous flux 4 P emerging from a point S of the extended source.

Eq. 1(4.3.6); that is emitted luminous flux. Thus,

File [7(4.3.6)]
_ dP [10(4.3.6)]
dA, dA, dug aS Fe
which, in the notation of differentials used here, From Eq. 10(4.3.6), the statement can be derived
means that the luminance is defined as the second that luminous flux emitted by a source is propor-
derivative of the luminous flux with respect to the tional to its geometric extent, the proportionality
area of the source aperture and the solid angle factor being the luminance of the source.
into which the luminous flux is emitted, assuming Figure 1(4.3.6) and the various expressions
that the specified direction of emission is per- derived above that are related to it, are valid only
pendicular to the aperture. for the geometrical condition that the two aper-
An alternative expression for Eq. 7(4.3.6) is tures AA, and AA, (or dA, and dAp, respec-
obtained with reference to Eq. 2(4.3.6); that is, tively) are perpendicular to the specified direction
SD. When this condition is removed, the geomet-
d*P
er 2 heel rical layout is that shown in Figure 2(4.3.6). Note
L=r [8(4.3.6)]
that the aperture AA, is parallel to the tangent
plane of the source whose normal is shown as nx
The quantity from which is measured the angle e, of the speci-
fied direction SD. When AA, is moved close to
_ dAsdAp ‘ the source, it acts as the source element itself. The
G ri RATE (m*) [9(4.3.6)]
other aperture 4A,, defining the detector, has
been tilted by an angle ¢,, measured between its
is often referred to as the geometric extent of the normal np and the specified direction SD.

Tangent

poe Wa |
Fig. 2(4.3.6). Illustration of geometrical configuration used in the development of the concept
of luminance. In contrast to Figure 1(4.3.6), the apertures AA, and AA > are now tilted with
respect to the specified direction SD.
 The Standard Photometric System 2n3

As a result of the tilt of the source (and thus, stant luminance; that is,
AA,) and the detector aperture AA p relative to
the specified direction, the apparent areas of AA, L,=L=const. (for alle)
and AA, as seen from D and S, respectively,
change to AA,cose, and AApcos ep [compare [19(4.3.6)]
with Figure 2(4.3.5)]. Thus, the equations devel-
oped above for luminance are now as follows: which means that a Lambert surface appears
equally bright regardless of the direction from
“a dI which it is viewed. Equation 18(4.3.6) is often
~ TSG [11(4.3.6)]
referred to as Lambert’s law or the cosine law of
emission.
As an example of a photometric calculation,
eee 2(4 we have selected the classical case of determining
the total luminous flux dP emitted by an element
dA ; of a Lambert surface in the entire hemisphere
; SOU eee
dA , COS €5 dws
[13(4.3.6)]
is above the surface element. Figure 3(4.3.6) il-
lustrates the case.
2 From Eq. 13(4.3.6) it follows:
nee aT
dA; cos &€, dA pCOS Ep
d*P = L dA,cos eg dws [20(4.3.6)]
[14(4.3.6)]
The small solid angle dw, in the direction eg is
etdP given by
L 1G [15(4.3.6)]
foe [21(4.3.6)]
with

Hite een fa TRS pale Es where dA, is a narrow circular strip on the hemi-
sphere whose area is obtained from
r r

[16(4.3.6)] dAp = 27e do [22(4.3.6) ]

and with

e=rsine, [23(4.3.6)]
ee dA 5 COS Es feels En [17(4.3.6)]
r and

Equation 13(4.3.6) is usually given as the basic o=TFrés [24(4.3.6)]

equation defining luminance. In fact, in the Ger-
do=rde,
man literature this equation is often referred to as
Photometrisches Grundgesetz (basic photometric
We then have
law). Most photometric calculations use Eq.
13(4.3.6) as the point of departure. dA, =2ar*sine de, [25(4.3.6)]
An important concept in photometric calcula-
tions is the Lambert surface. A Lambert surface is and thus,
a uniformly diffusing surface for which the
luminous intensity 7, in any given direction varies dw,=2msine,de, [26(4.3.6)]
as the cosine of the angle e between that direction
and the normal (e = 0) to that surface, that is,
Inserting this into Eq. 20(4.3.6), we obtain
I, = I,=9cos € [18(4.3.6)]
d*P = L dA, cos e,2msin es des
This relationship follows readily from Eq.
11(4.3.6) by assuming that the surface has con- [27(4.3.6)]
 274 PHOTOMETRY

Fig. 3(4.3.6). Illustration used in the calculation of the total luminous flux emitted per unit
area of a Lambert surface.

The integration of this equation over the range of which states that the total luminous flux emitted
és from 0 to 7/2 is equivalent to an integration of by a Lambert surface is equal to 7 times (sr) the
the luminous flux over the whole hemisphere: luminous intensity of the surface measured in the
direction of its normal (¢ = 0).
dP = [d?P = L dA, f°" 2msin e,cos es des A small aperture in a constant temperature
enclosure (emitting blackbody radiant energy) is
e=0
an almost perfect Lambert surface having the
[28(4.3.6)] cosine law of emission, but material surfaces are
at best only good approximations.
dP =aLdA,
The perfect reflecting diffuser (Sections 1.4.2
and 3.3.6) clearly becomes a Lambert surface
In accordance with Table I(4.1) and Table
when it is uniformly illuminated by a source.
1(4.1), the quotient dP/dA, defines the /uminous
Moreover, if the uniform illumination is obtained
exitance M of the surface element of the source;
by placing this ideal surface in a parallel beam of
thus, we have the simple relation
luminous flux incident at an angle e, with the
dP normal to the surface [see Figure 4(4.3.6)], the
MA ied a aL (Im - m~*) luminance Lp of that surface is then given by

[29(4.3.6) 2 E,, COS Ep

ip T
[32(4.3.6)]
It follows that the total luminous flux emitted per
where E,, is the luminous flux per unit area in the
unit area of a Lambert surface is 7 (sr) times its
incident beam; that is, the illuminance on a surface
luminance.
normal to the direction of the beam. Thus, a
With Eq. 11(4.3.6) for e = 0; that is,
necessary condition for a surface to be perfectly
diffusing is that the luminance observed in some
Lawns dl,
[30(4.3.6)] fixed direction ¢, varies with the cosine of the
angle of incidence ep of a parallel incident beam
inserted into Eq. 29(4.3.6), we obtain - of luminous flux.
This must be true whether the directions of
P= aI,—6 (Im) [31(4.3.6)] incidence and of view (€p and ¢,) lie in the same

Reflected Incident
Luminous Flux Luminous Flux
Perfect Reflecting
Diffuser
Fig. 4(4.3.6). The perfect reflecting diffuser be-
comes a Lambert surface when illuminated uni-
formly by a source. The luminance observed in
some fixed direction (¢,) varies with the cosine of
the angle of incidence (¢)) of the incident luminous
flux.
or different half-planes bounded by the normal to
the surface, and it gives a test often used to see
how closely material surfaces approach the ideal.
For an imperfect isotopic surface, viewed at an
angle e,, with the illuminating beam incident at
angle ep, the luminance Lp is given by
Ly = Blep, €s; "ee [33(4.3.6)]
The quantity B(ep, es, W) is the /uminance factor
of the material surface which is defined as the
ratio of the luminance of the surface to that of the
perfect reflecting diffuser identically illuminated
(compare radiance factor, discussed in Section
1.4.2). The luminance factor B(ep,e5, ~) de-
pends, in general, on ¢;,€p, and the angle y
between the half-planes through the normal to the
surface that contain these directions. Isotopic
means that B(es, ep,¥) is unchanged by any
rotation of the surface in its own plane. For the
perfect reflecting diffuser, we have B(ex,, ep, W)
= 1 for all e,, ep, and y. According to what is
known as the principle of reciprocity, the reflec-
tance is the same if the directions of incidence
and view are interchanged, that is,
Blep, €s,¥) = Bles, ep, v) [34(4.3.6)]
This principle is easily established for some espe-
cially simple cases, for example, when the dif-
fusing surface is a very thin layer of scattering
particles; but for more general cases, where the
polarization of the luminous flux is a factor, the
proof is not straightforward (see Chandrasekhar,
1960). However, the result is held to be true in all
cases. Thus, if for the fixed angle of view ég,
The Standard Photometric System 275
B(ep, &s, ¥) is found to show small deviations
from constancy as €p is varied, similar deviations
will occur for a fixed angle of incidence e, when
the angle of view is varied. The degree of ap-
proximation to the cosine law of emission can
therefore be studied in this indirect way.
A material surface that obeys, or nearly obeys,
the cosine law of emission for all directions of
incidence with an approximately constant value
of B(ep, €s,¥) for all ep, es, p, is a close ap-
proach to a perfect diffuser, but the common
value of B(ep, &5, Y) may be less than unity, so
that the surface is not necessarily a perfect reflec-
tor.
Close approximations to the perfect reflecting
diffuser are of particular interest to photometrists
and colorimetrists as working standards of reflec-
tance. There are several materials well suited for
the purpose and some of their properties have
been discussed in Sections 1.4.2 and 3.3.6.
Figure 5(4.3.6) compares results obtained by
Budde (1964) for an MgO surface and a BaSO,
surface with the values for an ideal surface. The
insert in Figure 5(4.3.6) shows schematically the
experimental arrangement. After setting the read-
ing of the photometer equal to unity when the
incident flux is normal (€, = 0) to the test surface,
the subsequent readings should vary exactly with
the cosine of the angle ¢, of incident flux if the
test surface were a perfect diffuser. The observed
deviations for MgO and BaSO, surfaces are shown
in Figure 5(4.3.6). Positive deviations indicate that
the photometer reading for the ideal surface would
have been higher.
Measurements of luminance, in practice, or
the calibration of /uminance meters, are often
made with the aid of standards of luminance. For
high values of luminance, tungsten-ribbon-fila-
ment lamps are usually used as standards that
have been calibrated by a national standardizing
laboratory. These lamps operate base downwards,
with the measurements made in a specified hori-
zontal direction, perpendicular to the plane of the
filament. The calibration data of these lamps refer
only to a specified small area of the filament
because the temperature and thus, the luminance,
varies over the surface of the ribbon filament. A
lamp operating at a color temperature of about
2000 K provides a luminance of its filament of
about 2 X 10° cd - m 7”. An increase in the color
temperature to 3000 K raises the luminance to
about 1 X 10’ cd- m 7.
Luminance standards for lower values of lumi-
nance usually consist of a diffuser of known
276 PHOTOMETRY

+0.1
MgO

j—___80S04

0.0 _

«
SI
>
o
iS)

-0.1 ~a
Ep
DEVIATION
COS
FROM
| Test Surface SY

o° 10° 20° 30° 40° 50° 60° paee 80° 30°

ANGLE OF INCIDENCE e, (DEGREES FROM NORMAL)

Fig. 5(4.3.6). Deviations from the cosine law of emission for practical surfaces used in
photometry and colorimetry. Insert shows experimental arrangement (from Budde, 1964).

luminance factor £, illuminated by a lamp illuminated by a source, is facilitated by means of

calibrated for illuminance. If the illuminance pro-
duced by the lamp on the diffuser is E,, at normal
incidence, the luminance Lp of the diffuser in a
specified direction is obtained from Eq. 33(4.3.6),
that is
Lp = B=
Reflecting diffusers prepared by pressing BaSO,
powder into tablets have been found particularly
useful for the purpose. When such tablets are
illuminated at normal incidence (ep = 0°) and
viewed at e, = 45°, their luminance factors 6 are
very close to unity (Kartachevskaia et al., 1975).
The measurement of the luminance of small
sources, small areas on extended sources, or spots
an optical imaging system. The basic optical sys-
tem employed is illustrated in Figure 6(4.3.6). The
lens forms an image, if necessary, enlarged, of the
small source at the aperture behind which a
broad-band photometer is located. The image
must be large enough to fill the aperture. The
[35(4.3.6)] optical arrangement is the same as that used to
provide Maxwellian view in visual instruments,
where the observer’s eye is placed directly behind
the aperture (Section 5.15.1). The lens is then seen
uniformly filled with ight coming from the small
source; the luminance of the lens being tL, where
L is the luminance of the small source and T is the
luminous transmittance of the lens.
In many practical measurements of the lumi-
nance of small sources, it is important to define

Aperture

Source

Photometer

To Viewfinder

Fig. 6(4.3.6). Optical system used to facilitate measurement of luminance of small sources.
Modification at the bottom employs a Pritchard aperture mirror with a viewfinder to define the
spot on the source actually being measured.
 Calculation of Illuminance Produced by Lambert Sources of Different Shapes
the actual spot measured. The simple optical
arrangement shown in Figure 6(4.3.6) can be
modified by a variety of means to permit the
operator of the luminance meter to locate the spot —
and define it within the area considered. In port-
able luminance meters used in practice, the
Pritchard aperture mirror system has advantages
over other systems (Walker, 1972). The lower part
of Figure 6(4.3.6) illustrates the principle of the
Pritchard aperture mirror. The front-surface mir-
ror with an elliptical aperture through the entire
mirror body is placed in the image plane of the
lens, centered, and then oriented at 45° to the
optical axis. The luminous flux passing through
the aperture to the photon detector originates at a
circular area of the source whose image coincides
with the tilted mirror aperture. The rest of the
field imaged by the lens onto the mirror is viewed
by the operator through a viewfinder ap-
propriately adjusted so that both the image of the
field and the edge of the mirror aperture appear
in sharp focus. The operator sees the field, for
example, the entire source and its surrounding,
with a black circular spot in its center that defines
the actual spot of the source whose luminance is
being measured. Mirrors with different size aper-
tures can be inserted to change the size of the
spot in the field under test.
Zit
4.4 Calculation of Iluminance Produced by
Lambert Sources of Different Shapes
The illuminance from some idealized sources can
be calculated in accordance with the defining
equations given in Table 1(4.1) and in Sections
4.3.5 and 4.3.6.
The basic assumption is that the source always
obeys the cosine law of emission as defined by
Eq. 18(4.3.6). Formulae for the illuminance E on
specified areas in terms of the luminance L or
luminous intensity 7, have been derived for a
number of sources of different shapes (see, for
example, Zijl, 1951; Mahan and Malmberg, 1954;
Reeb, 1962; Keitz, 1971). Some of the simpler
formulae are listed in Table I(4.4) without proof
in the Appendix. They are only of limited value
to the illuminating engineer whose more intricate
problems usually require a combination of graph-
ical and numerical methods (see, for example,
Illuminating Engineering Society, IES, 1972). The
direct illumination of surfaces by light sources is,
in any case, only half the practical problem. The
increase in illuminance by inter-reflections be-
tween all the surfaces present has also to be
evaluated. Simplified methods of doing this ap-
proximately for rooms of different shapes have
been developed by illuminating engineers.
 CHAPTER 5
VISUAL EQUIVALENCE AND
VISUAL MATCHING
5.1 PREAMBLE
Chapters 3 and 4 specify the simple and rigid
systems of trichromatic colorimetry and photome-
try required in practical applications. The pro-
nouncements of the systems are exact only for
one or another of the hypothetical colorimetric
and photometric standard observers. In colorime-
try, the CIE 1931 standard colorimetric observer
conforms perfectly with the trichromatic generali-
zation in its stronger form including the per-
sistence of full color matches. Its set of three
independent color-matching functions is uniquely
determined, subject only to an arbitrary nonsin-
gular linear transformation. Similarly in photome-
try, the CIE 1924 standard photometric observer
has a unique luminous efficiency function, which
in fact is a particular linear combination of the
color-matching functions of the CIE 1931 stan-
dard colorimetric observer.
The various factors that make the assumed
properties of the CIE colorimetric and photomet-
ric standard observers, at best, approximations
for actual visual matching are taken into account
in the formal CIE scheme only by introducing the
following additional standard observers and con-
straints to the color-matching conditions:
(a) An alternative set of color-matching func-
tions is standardized for use when the matching
field exceeds a certain angular size (4° diameter)
and when the matching conditions are otherwise
such that rod vision may be assumed to be pro-
ducing only an insignificant distortion of the
match that would be obtained in the absence of a
278
rod response. This alternative set of color-match-
ing functions defines the CIE 1964 supplementary
colorimetric standard observer.
(b) A second luminous efficiency function is
standardized that is not related to either of the
two standard sets of color-matching functions, for
use in photometry when the scotopic or rod
mechanism is judged to be the only, or at least
the greatly predominating, factor determining a
brightness match. This luminous efficiency func-
tion V’(A) defines the CIE 1951 standard ob-
server for scotopic vision.
In this chapter are collected supplementary
notes on the methods, results, and laws of visual
matching, both for normal and color-defective
eyes, which aim at providing a more general pic-
ture going beyond the minimal requirements for
the setting up of classical colorimetry as em-
bodied in the CIE systems. As the diverse topics
covered need to be treated from somewhat differ-
ent standpoints, the several notes are kept, as far
as possible, self-contained.
5.2 CLASSIFICATION OF MATCHING
PROCEDURES
In a visual match, a determination is made of two
physical stimuli that, in some sense, produce the
same visual response. The following descriptive
classification of various matching procedures
shows the main differences in their objectives and
implications; it is not a complete logical scheme
of definitions.

5.2.1 Visual Equivalence and Visual Match by
Strict Substitution
If, in a visual experiment of observation whose
result is determined by the observer’s judgment of
what is seen, one physical stimulus (test stimulus
A) can be substituted for another (test stimulus
B) without affecting the prescribed result, then,
with respect to the particular experiment, pre-
scribed result, and observer, the two stimuli are
said to be visually equivalent by strict substitution.
The substitution must be exact; that is, the alter-
native test stimulus must be imaged on the same
retinal area, and for the same duration. All other
stimuli, which constitute the conditioning stimuli,
must be kept the same. They may be applied
anywhere in the field, including the possibility
that they extend over the retinal area on which
the test stimulus is imaged. If they are not con-
stant in time and if the retina is not fully adapted
to them when the test stimulus is applied, the
time of application of the latter in the time se-
quence of the observation must be invariable.
Furthermore, if the effective ocular entrance pupil
for the conditioning stimuli and hence, the corre-
sponding retinal illuminance might be altered as a
result of the interchange of test stimuli, this must
be obviated by some modification of the experi-
ment (e.g., by the use of an artificial pupil).
All the various factors that determine the vis-
ual environment in which the test stimulus is
displayed to the observer constitute the display
situation of the test stimulus. These factors in-
clude the angular size and position of the test
area, the duration and epoch in the time sequence
of the observation during which the test stimulus
is exposed, and, in fact, all factors that must be
kept the same when the test stimulus is sub-
stituted by another one.
If not already implied by the foregoing, another
condition is that the observations with one test
stimulus must be entirely distinct from those with
the other. This last condition means that the
observer’s judgment cannot be a direct compara-
tive judgment of the visual effects of the two test
stimuli presented in the same experiment, such
as: A appears brighter than B. Instead, it must be
an absolute judgment, for example, the naming of
an apparent color, or a decision as to the visible
presence or absence of an object, not necessarily
the test stimulus, in the field. Alternatively, there
can be an “internal” comparison such as a judg-
ment of equality in the appearance of the test
stimulus and the appearance of a second stimulus
Classification of Matching Procedures 279
forming one of the conditioning stimuli and re-
maining, therefore, unchanged when test stimulus
B is substituted for test stimulus A.
To define a particular visual equivalence by
strict substitution, both the display situation and
the observer’s visual judgment must be specified.
The latter is conveniently regarded as cast in the
form of a proposition about some feature of what
the observer sees, which can be affirmed or de-
nied. If, for a given test stimulus, the observer
affirms the proposition, the test stimulus is
accepted as yielding the prescribed result. In
practice, repeat observations will normally be
made enabling acceptance to be based on a more
refined statistical condition such as “affirmation
in more than 50% of repeated observations.”
Many of the characteristics of the test stimu-
lus, for example, its size, duration, and position in
the visual field, are already laid down in the
display situation. In most cases, the test stimulus
will be produced by the imaging on the retina of a
surface in the external field uniformly emitting
radiant power. Thus, the principal way in which
such a test stimulus can be varied is, in effect, by
changing the absolute spectral distribution of the
radiant power emitted by the surface. However,
in certain cases, the state of polarization of the
test stimulus may be varied, or changes may be
brought about in the effective ocular entrance
pupil (for this stimulus) so as to modify the
angular distribution of the stimulus and the reti-
nal illuminance it produces.
In the complete specification of any test
stimulus, in any part of the visual field at any
time, it would be necessary to know (a) the abso-
lute spectral radiant power distribution of the
corresponding external emitting surface; (b) the
size and shape of the effective ocular entrance
pupil for the stimulus concerned, and its position
of entry in the natural pupil of the eye; and (c)
the state of polarization of the radiant energy
incident on the cornea. Where the effects of
changes in the spectral composition are the main
interest, as in the present context, the tacit as-
sumption is normally made that the other factors
are kept unchanged in any particular investiga-
tion, and stimuli are discussed as though they
were uniquely defined by the spectral composi-
tion of the corresponding external emitting
surface. Where these assumptions are not justified
[as in Example (e) of the various matching proce-
dures cited in Section 5.2.5], this must be ex-
plicitly indicated.
280 VISUAL EQUIVALENCE AND VISUAL MATCHING
Although the commonest form of test stimulus
is the single uniform light patch, less simple forms
may be used for which many of the possible
propositions about the appearance of a uniform
light patch could never be affirmed by the ob-
server. Other propositions, however, that are never
true of a uniform patch may be appropriate for
various special groups of nonuniform test stimuli.
For example, the proposition may assert that “the
two similar halves of the test stimulus appear of
the same hue although different, possibly, in
brightness and saturation.” This could be used in
establishing a visual equivalence for test stimuli
the two halves of which differ in spectral radiant
power distributions. Another example would be
the assertion that the apparent color difference
between the two halves of the test stimulus is
equal to that presented in a neighboring retinal
area by a similarly divided comparison stimulus
forming part of the unchanged conditioning
stimuli. Also, a temporal variation in the color
appearance of the test stimulus may be embodied
in a proposition such as the uniform test patch in
the first half of its period of exposure appears of
the same brightness although, possibly, of differ-
ent hue and saturation as compared with its ap-
pearance in the second half-period. This again
could generate an equivalence relation among sui-
table test stimuli.
These more complex visual judgments, that are
appropriate to types of test stimuli that are not
uniform light patches but constant during their
period of exposure, are equally admissible with
the simpler judgments. When they are applied,
the physical specification of the test stimulus be-
comes correspondingly more complicated and will
vary from one application to another. Where nec-
essary, test stimuli that are uniform and constant
during their exposure periods will be described as
simple, whereas the term complex will be applied
to other test stimuli, namely those with internal
structure, spatial or temporal.
The classification of equivalence and matching
procedures deployed here has primarily in mind
cases where only simple test stimuli are used. But
nothing essential is altered when the visual judg-
ment necessarily demands a complex test stimu-
lus. (Two of the examples of matching procedure
given in Section 5.2.5 employ complex test
stimul..)
All the test stimuli that, for a given equiva-
lence experiment and observer, yield the pre-
scribed result, constitute an equivalence set in
that any member A of the set is visually equiva-
lent by strict substitution to any member B: (A
equiv. B). Clearly this equivalence relation is
(i) reflexive: (A equiv. A),
(ii) symmetric: (A equiv. B) implies (B equiv.
A), and
(iii) Transitive: (A equiv. B) and (B equiv. C)
imply (A equiv. C).
It is also
(iv) transitive in the wider sense: if (A equiv. B)
and (C equiv. D) then any two of A, B, C, and D
are equivalent.
However, whether the members of the equiva-
lence set satisfy any form of linearity law can be
decided only empirically for the equivalence
experiment and observer in question. A compre-
hensive linearity property that may hold in a
particular case is defined by the following condi-
tions:
(i) if (A equiv. B), then (aA equiv. B), where a
is any positive factor, and
(ii) either of the two equivalences (A equiv. B)
and ({A + B] equiv. B) implies the other.
This would mean, for example, that if the re-
quired result of the equivalence experiment was
that the test stimulus had the same hue as a
juxtaposed yellow stimulus (one of the condition-
ing stimuli), then adding any two stimuli that
match yellow in hue must again yield a stimulus
that matches yellow. This is nearly true. There
are, however, certain deviations, notably the
Bezold—Briicke effect. On the other hand, if the
required result is that the test stimulus has a
different hue from that of the fixed yellow com-
parison patch, additivity does not hold, because a
red and a green test stimulus each differ in hue
from the comparison patch, but an appropriate
additive mixture of the two will be yellow.
A more restricted form of linearity law holds if
the equivalence satisfies the condition that for
any three test stimuli, A, B, and [A + (1 — a)B],
where 0 < a < 1, the equivalence of any two im-
plies that all three are equivalent. This form is
important in the discussion of Maxwellian tri-
chromacy (see Section 5.3.2).
It is important to note that not all procedures
defining a visual equivalence by strict substitution
would be described as visual matching. Consider,
for example, two alternative test stimuli A and B
applied at a particular location in the extrafoveal

retina. These stimuli may be equivalent by strict
substitution in an experiment where the visual
judgment concerns the appearance of a fixed light
patch imaged on the fovea and forming one of the
conditioning stimuli. However, these test stimuli
would not normally be said to be in visual match.
If the visual judgment relates specifically to the
appearance of the test stimulus itself, the resulting
equivalence represents a visual match by strict
substitution. If, for example, the observer has to
judge identity of color between each of the two
test stimuli A and B and another stimulus C,
where C forms one of the conditioning stimul,
the relation of equivalence between A and B is
also one of visual match.
In the expression visual match by strict substitu-
tion, a specific meaning is being given to the term
visual match. This specific meaning differs from
the usual sense of identity, complete or in one or
more qualities, in the appearance of two test
stimuli presented in the same observation. The
context will usually indicate which kind of visual
match is intended; but, where necessary, a visual
match by strict substitution will be referred to as
a substitution match or an indirect match, as com-
pared with a direct match. Thus, in the example at
the end of the previous paragraph, the stimuli A
and B are in substitution or indirect match if the
stimuli A and C, and B and C, are, respectively,
in direct match.
5.2.2 Asymmetric Comparison and Matching;
Quasi-Symmetric Matching
A comparison of the visual effects of two test
stimuli in which the display situations of the two
stimuli are not in all respects the same, is quali-
fied as asymmetric. Asymmetric comparison is
certainly involved if the alternative test stimuli
are not imaged on identical areas of the same
retina; if their size, shape, duration, and so forth,
differ; if the dispositions in time and space of the
conditioning stimuli with respect to the test stimuli
are different; or if the observations with the alter-
native test stimuli are not independent. Thus, the
only comparisons of test stimuli not in some
respect asymmetric are determinations of equiva-
lence by strict substitution.
A particular kind of asymmetric equivalence is
asymmetric matching which is obtained if the
following conditions are fulfilled:
(i) The two stimuli are presented to the observer
in different display situations in completely in-
dependent observations;
Classification of Matching Procedures 281
(ii) the common proposition to be satisfied by
each stimulus is an absolute judgment of its ap-
pearance as a patch of color, but not excluding
complex stimuli, irrespective of overall shape, size,
or duration.
For example, the observer may have to assert
that the patch has the hue yellow, or that it is
perceptibly different from black, or that it is
achromatic, or that its two contiguous halves show
no color difference, and so on. If the two test
stimuli in their respective display situations satisfy
the same prescribed proposition, they are deemed
to be matched with respect to the appearance
quality judged, and represent what may be called
an indirect asymmetric match. Such matches are
clearly closely related to matches by strict sub-
stitution. They represent an extension of the latter,
because the display situations of the two test
stimuli are not the same. They also represent a
limitation, because the criterion for match must
be an absolute judgment of the appearance of the
test stimulus with no explicit reference to the
appearance of any other stimulus in the visual
field functioning as a comparison stimulus. Any
such stimulus, even though it were present in both
display situations, would have an appearance sub-
ject to possible modification by the other condi-
tioning stimuli in the display, which would not
generally be the same in the two cases. However,
an internal comparison stimulus might be used if
there were good reason for thinking its ap-
pearance would be the same in both display situa-
tions.
If the proposition embodying the appearance
criterion is kept fixed and any number of differ-
ent display situations are considered, then every
combination of test stimulus and its proper dis-
play situation that satisfies the prescribed pro-
position is a member of an equivalence class of
(stimulus /display)-combinations. The test stimuli
of any two members of such an equivalence class
represent a pair in indirect asymmetric match, a
match that can be represented symbolically by:
(A/ds: x) — iam: k — (B/ds: y)
where
A = stimulus A
B = stimulus B
ds: x = display situation x
ds: y = display situation y
iam: k = indirect asymmetric match with ap-
pearance criterion k
282 VISUAL EQUIVALENCE AND VISUAL MATCHING
Obviously, the relation denoted by the link sym-
bol “iam: &” is reflexive, symmetrical, and tran-
sitive, as well as transitive in the wider sense (see
Section 5.2.1). But it is not symmetrical to an
interchange of test stimuli only. The relation de-
noted by (A/ds:x) — iam: k — (B/ds: y) does
not imply (B/ds: x) — iam: k — (A/ds: y),
nor, in general, does any other law relating to
combinations hold that involves separating a test
stimulus from its proper display situation.
The equivalence class just defined is made up
of subclasses in each of which the display situa-
tion is the same and only the test stimulus varies
from one combination to another. Each such sub-
class is the equivalence set of matches by strict
substitution for the given display situation and
matching criterion. If there is no restriction on
the nature or the number of different display
situations taken into account, the full asymmetric
equivalence class is characteristic of the proposi-
tion defining the matching criterion; it can be
regarded, in philosophical parlance, as defining in
extension and in a way that can be communicated
to others what the proposition means for the
person making the observations.
By contrast with indirect asymmetric matches,
direct asymmetric comparisons are derived from
an observer’s judgments about the appearances of
two test stimuli (A and B) presented in the same
observation. This means that the stimulus distri-
bution, in visual field space and in time, during
the course of the observation, but with the omis-
sion of the test stimuli, will be common to the
display situations of both test stimuli. However,
the positions of the test stimuli in this distribu-
tion will generally be different. In addition, the
display situation of test stimulus A will include
test stimulus B, and conversely. When direct
matches are in question, the term total display
situation will be used to cover the display situa-
tions of both test stimuli.
With indirect asymmetric matching, it is un-
necessary to place any restriction on the absolute
appearance criterion employed. However, not ev-
ery proposition about the appearances of the two
test stimuli, to be asserted or denied by the ob-
server in a direct asymmetric comparison, leads to
a matching relation between the test stimuli. The
assertion must mean, in effect, that one
stimulus, A for example, in its display situation
(ds : x), possesses some appearance quality equal
to that of the other test stimulus B in its display
situation (ds: y). If “dam: k” is used as an
abbreviation for matches in direct asymmetric com-
parison with criterion k, the match is expressed
symbolically by:
(A/ds: x) — dam :.k — @B7ds* y)
where k must be an equality criterion. The equal-
ity criterion must be of such a kind that an
interchange of the stimulus /display combinations
leaves the truth of the above relation unaffected.
In that case, the symmetrical relation is implied:
(B/ds:y) — dam: k — (A/ds:x)
A criterion expressing an inequality does not
by itself define a matching relation. For example,
a criterion with the assertion “(A/ds: x) is
brighter than (B/ds: y),” is no longer true on
interchanging the stimulus /display combinations.
Also, not all equality criteria are acceptable.
For example, affirming the proposition “(D/
ds: x) and (B/ds: y) both have the hue yellow,”
corresponds certainly to an equality criterion, but
it approaches closely the assertion of two in-
dependent propositions, “(A /ds: x) is yellow”
and “(B/ds: y) is yellow,” so that the ap-
pearances of the two test stimuli are not really
being compared. The necessary condition can be
put as follows: for direct asymmetric matching,
the criterion must not only be an equality crite-
rion; it also should be a strictly comparative
judgment of the appearances of the two test
stimuli, not separable into two judgments each
having reference to the appearance of one test
stimulus only. In this sense, the criterion of equal-
ity must be nonseparable.
Most investigations of matching properties
employ direct asymmetric matching. The asym-
metric procedures fall into two main groups. In
the first group, the asymmetry enters incidentally
when the two test stimuli are brought together in
the same experiment. This facilitates much more
precise comparisons of their visual appearance
than can be obtained when absolute judgments
are used in visual matching by strict substitution.
The outstanding example is matching in a bipar-
tite field. The two similarly placed, juxtaposed
halves of the bipartite field are occupied respec-
test tively by simple test stimuli. The procedure is
clearly asymmetric as the test stimuli are imaged
on different, even though closely adjacent, retinal
areas. Their positions with respect to other stimuli
in the field are not precisely the same. Also the
display situation for each test stimulus necessarily

includes the other test stimulus. Thus, the two
display situations differ in certain particulars.
Nevertheless, it is reasonable to assume in
many cases that two test stimuli matched asym-
metrically in this way would also match if sub-
stantially the same comparison could be made,
with comparable precision, by a strictly symmetri-
cal substitution procedure. Where this is so, the
asymmetric procedure may be described as quasi-
symmetric.
In the above example of the bipartite field, it
is not difficult to make specific tests on a selec-
tion of the whole range of matches under study to
determine whether the procedure qualifies as
quasi-symmetric. In particular, one would try to
ascertain:
(a) whether identical test stimuli applied respec-
tively in the two half-fields yield a match,
(b) whether interchanging any two different but
matched test stimuli occupying the respective
half-fields, leaves the match intact,
(c) [given that (b) holds] whether simple transi-
tivity is valid among sets of three matches: if A
matches B and B matches C, then A matches C.
Affirmative results from these tests would in-
dicate that the particular direct asymmetric
matching procedure possessed the three character-
istic properties of matching by strict substitution,
namely, reflexivity, symmetry, and transitivity.
The procedure would then qualify as quasi-sym-
metric. However, it must be noted that the set of
all stimuli that are members of pairs in quasi-
symmetric matching do not form an equivalence
class. Transitivity in the wider sense, defined
earlier, and valid for matching by strict substitu-
tion, does not hold. We expect quasi-symmetric
matching if the following conditions are fulfilled:
(a) the areas of the retina responding to the two
test stimuli have the same response properties,
(b) the other stimuli contributing to the display
situation are so disposed as to affect equally the
appearance quality being assessed for the two test
stimuli,
(c) any possible effect of one test stimulus on
the appearance of the other is at least the same
for both when match is reached.
In the second main group of asymmetric com-
parisons, the emphasis is on the asymmetry. The
objective is to determine how differences in the
retinal area used in the conditioning stimuli and
Classification of Matching Procedures 283
resulting adaptation conditions, and so forth, may
affect the similarity of response to different test
stimuli. The asymmetry is then an essential ele-
ment in the comparison, and such matching
procedures may be described as specifically asym-
metric.
As previously explained, direct asymmetric
matches will be obtained from the comparison if
the observer’s criterion is one of equality and is
also nonseparable. But interchange of test stimuli,
leaving the total display situation otherwise un-
changed, will generally upset the match. The
failure of this symmetry feature for direct specifi-
cally asymmetric matches prevents even the fram-
ing of the simple transitivity principle relating the
matches, in pairs, of any three test stimuli, such
as holds for quasi-symmetrical matching.
However, three modified forms of transitivity
principle may hold in direct asymmetric match-
ing. The first, asymmetric transitivity for stimuli in
direct asymmetric matches, states that if any
stimulus A, in display situation (ds: x) matches
the two different stimuli B, and B, both in dis-
play situation (ds: y), and if B, in (ds: y) also
matches A, in (ds: x), then B, does the same. In
the adopted symbolic language, this statement
reads as follows:
if bedZalSeots Jaca,
Kay 4 Ba es 5 J.)
and: WU Bs/ds sy) dam: ke GA5/ds ix)
then (B,/ds:y) — dam: k — (A,/ds:
x)
If this holds, it follows that the corresponding
relation with the roles of the two display situa-
tions interchanged will also be true. The other
modified transitivity laws concern limited groups
of different asymmetric matching procedures to
be discussed in the next section.
5.2.3 Limited Groups of Asymmetric
Matching Procedures
Particular examples of direct asymmetric match-
ing procedures are briefly described in Section
5.2.5. However, first some consideration will be
given to limited groups of such procedures, all
using the same matching criterion but with differ-
ent total display situations. It would be a useful
concept in relating results obtained with the dif-
ferent procedures of such a group if it were per-
missible to think of the same (test stimulus /dis-
play situation)-combination participating in dif-
ferent procedures. It might then be found, for
284 VISUAL EQUIVALENCE AND VISUAL MATCHING
example, that there is simple transitivity in the
matching of different combinations, just as there
is for similar groups of indirect asymmetric
matching procedures with a common matching
criterion. The difficulty is that in a direct and
specifically asymmetric match, the display situa-
tions of the two test stimuli are not independent.
One display situation cannot be modified without,
at least formally, modifying the other. However,
suppose that in a particular matching procedure
of the group, all the factors in the display situa-
tion of one test stimulus (A, for example) that
were responsible for influencing As appearance,
could be separated from the other factors. This
would then enable one to define an effective dis-
play situation, abbreviated to (eds: x), but ex-
pressly excluding the other test stimulus as a
possible influence. Suppose further that this could
be done for both display situations in each match-
ing procedure of the group. Again, one would
assume that no factor in the effective display
situation of one test stimulus would have an
appreciable effect on the appearance of the other
test stimulus and would not therefore belong to
the effective display situation of the latter. Be-
sides the test stimuli themselves, there will no
doubt be other factors in the total display situa-
tion of any particular procedure that do not be-
long to either of the effective display situations
of the two test stimuli. It is implied in the as-
sumptions made that these factors are indifferent
with regard to the appearances of the two test
stimuli and have a negligible effect in modifying
the matches made. These fairly demanding as-
sumptions would justify the concept of indepen-
dent combinations (test stimulus/effective dis-
play situation) or, in abbreviation, (ts/eds). They
would allow for the same combination to occur in
a number of different matching procedures.
However, the test stimuli themselves present
certain problems. They have been excluded from
the effective display situations on the tacit as-
sumption that the effect on the appearance of any
particular test stimulus resulting from the pres-
ence of any other test stimulus that it matches is
likely to be slight. At least, the effect is not of
such a kind as to vitiate the idea of the test
stimulus in its effective display situation having a
constant appearance in whatever match of the
group it is concerned. This assumption is most
plausible if the matching criterion calls for a
complete color match.
An overall test of the validity of the foregoing
concepts and assumptions for a particular group
is provided by the second modified transitivity
principle for direct asymmetric matches. This as-
serts transitivity among the (test stimulus /effec-
tive display)-combinations of a group:
if (A /eds: x) — dam: k — (B/eds: y)
and (B/eds: y) — dam: k — (C/eds: z)
then (C/eds: z) — dam: k — (A/eds: x)
A hypothetical example of a simple group of
asymmetric matching procedures is illustrated in
Figure 1(5.2.3), caption (a).
More practical and interesting groups of
matching procedures are obtained if one does not
insist that every pair of different (ts /eds)-combi-
nations can be coupled together to form a match-
ing procedure of the group. Instead, suppose that
the (ts /eds)-combinations fall into two classes (I
and II) and that the procedures of the group
consist entirely of all the possible couplings of
two combinations, one from each class. Then, the
second modified transitivity principle that treats
all (ts /eds)-combinations on an equal footing can
no longer be formulated, but the third modified
transitivity principle for direct asymmetric matching
may be valid for the group. This states that
if
(A /eds: x) and (B/eds: y) belong toclass I
(C/eds: z) and (D/eds: w) belong to class II
and (A,fedsiix,)
adam :.Angn i G/eds acs
(.B/eds.:.y.) dams dersol C7 edsicezes
andif: .(.A/eds:'
2) dam skis (D/eds: w)
then (B/eds: y) — dam: k — (D/eds: w)
The above third modified transitivity principle for
direct asymmetric matching may be referred to as
the asymmetric transitivity for the (test stimulus/
effective display )-combinations of a two-class group
of matching procedures. Figure 1(5.2.3), Caption
(b), illustrates the kind of two-class group that
might meet the present conditions.
Embodied in all the transitivity principles ap-
plicable to direct matching, is the following idea:
When each of two different test stimuli in the
same display situation matches a particular com-
 Classification of Matching Procedures 285

| i FOVEAL
10° "2 @Q x BY
| AREA
Q

ES

as
Fig. 1(5.2.3). (a) Hypothetical single-class group of asymmetric matching procedures. The
diagram shows 13 possible positions in the visual field for a 1° diameter test stimulus, each at
the center of a uniform nonvariable illuminated surround of 10° diameter, the surrounds being
all nonoverlapping. The (13-12) /1-2) = 78 asymmetric matching procedures of the group are
obtained by using, in turn, every possible pair of test stimulus positions with their associated
surrounds, the rest of the visual field being kept dark. Subject to the assumption that in each of
the 78 matching procedures the appearance of either test stimulus is influenced solely by the
absolute spectral radiant power distribution of its own surround stimulus, the group exemplifies
the concept of a single class group with independent (ts /eds)-combinations. (b) Hypothetical
two-class group of asymmetric matching procedures. By contrast with the group under (a), just
one pair of test stimulus positions (X and Y) with their associated surrounds is used for all
matching procedures of the group. But in different procedures the absolute spectral radiant
power distributions of the two surround stimuli are varied, independently, in any way desired.
All the various surrounds about X and Y, respectively, that are included in the group correspond
to the (ts /eds)-combinations of classes I and II.

parison stimulus in its display situation, the 5.2.4 Matching Criteria

appearance of the comparison stimulus is not
The criteria embodied in the visual judgments on
different in the two matches because the test
which matching procedures are based vary widely
stimuli have different spectral compositions.
in complexity. The simplest and most important
From what is known of the processes of vision
is matching two light patches to complete color
through the initial absorption in visual pigments,
match, that is, in brightness, hue, and saturation.
we should certainly expect this to hold in most
More complex are matching judgments that re-
cases when complete color match is the matching
quire the observer to mentally isolate a particular
criterion used. Where the criterion requires
quality in the appearances of the two patches
matching of a single appearance quality of the
when these may differ in other respects. The
test stimuli, which may differ widely in other
principal judgments of this kind are the follow-
respects, it is more necessary to check whether
ing:
any transitivity principle used in interpreting re-
sults does in fact hold.
Asymmetric matching is considered further in (a) Matching brightness when the patches may
Sections 5.3, 5.4.12, and 5.12.1. differ in hue or saturation or both.
286 VISUAL EQUIVALENCE AND VISUAL MATCHING
(b) Matching hue when brightness and satura-
tion may differ.
(c) Matching saturation when brightness and hue
may differ.
(d) Matching brightness and hue when the
saturations of the patches may differ.
(e) Matching hue and saturation when the
brightness may differ.
(f) Matching saturation and brightness when the
hues may differ.
All these judgments of what may be called incom-
plete color matches have in fact been used to some
extent, but judgment (a) is outstanding because it
forms the basis of heterochromatic photometry by
the direct comparison method. It is tacitly assumed
in describing the above criteria that each of the
light patches is homogeneous in appearance over
its whole area. But this may not be the case even
when the external stimuli applied to the respec-
tive test areas are uniform and steadily exposed.
In some circumstances, notably with large (more
than 2° diameter) foveally centered bipartite
fields, inhomogeneities in patch appearance may
necessitate the introduction of supplementary
conditions in the matching criteria to enable the
observer to make unambiguous judgments [see
Examples (b) and (c) of Section 5.2.5]. An addi-
tional practical problem is that on first looking at
the bipartite field, the color appearances may not
be the same as after continued viewing, and a
regular observational routine may have to be
stipulated. With certain complex test stimuli,
the light patches may have internal spatial or
temporal structure that is perceived, but that is
sufficiently fine for the observer to make mental
averages of the appropriate qualities and apply
the uniform-patch criteria. This complication is
not unrelated to the everyday problem of match-
ing the color of material surfaces of different
textures.
The uniform-patch matching criteria have been
commonly applied to symmetric bipartite fields.
In the two halves of a bipartite field, simple test
stimuli are imaged (in addition, possibly to a
uniform conditioning stimulus) in total display
situations such that a quasi-symmetric matching
procedure is almost certainly assured. But they
are also applicable in specifically asymmetric
matching procedures in which the two test stimuli
are in quite different display situations [see Ex-
amples (g), (h), and (i) in Section 5.2.5].
As well as matching the qualities of simple
light patches treated as uniform, the eye can also
assess the similarity of contrasts contained in the
two patches to be matched between contiguous
areas, particularly when the contrasts are weak.
The criterion of equal contrast is applied to in-
crease the precision of brightness matching in a
bipartite field. By a suitable optical device, the
radiant flux irradiating each half of the field also
irradiates a patch (usually trapezoidal in shape)
forming an enclave in the other half but to a
stimulus level some 10% lower, although of the
same relative spectral distribution. Figure 1(5.2.4)
shows schematically the appearance of such a
field introduced by 0. Lummer and E. Brodhun in
the latter part of the last century (Walsh, 1958).
The observer makes the match by adjusting the
controls to equalize as closely as possible in all
respects the contrasts presented by the trapezoids
with their immediate surrounds.
The above photometric application of match-
ing to equal contrasts employs, in a particular
form, a more general matching criterion, the
matching of color differences, contained respec-
tively in two light patches on which are imaged
complex stimuli. In the simplest case, each patch
contains two similar uniform areas of different
color, the two color differences being equalized in
the match [see also Section 5.2.5(j)].
A matching judgment of a quite different kind
is obtained when two geometrically similar light
patches imaged on the same retinal area are
presented in rapid alternation. By repeated ob-
servations and adjustments of the frequency of
aB B
A aA
Fig. 1(5.2.4). Lummer-Brodhun photometric
field of the contrast type. In the trapezoidal
patches, the stimuli A and B are reduced by a
factor a (typically, a = 0.9).

alternation and the radiance of one of the light
patches, a condition is reached for which the
matching field appears quiescent but begins to
flicker if the radiance of the variable light patch is
raised or lowered by a small amount. Although
elaborate, the visual judgment or set of judgments
involved here is in practice easy to apply and is
effective whether or not the two light patches,
when they are viewed in turn under steady condi-
tions, have similar or widely different hues and
saturations. The quality matched is commonly
described as the “brightness”—perhaps better
known as the flicker brightness—and provides the
basis for the flicker method of heterochromatic
photometry (Walsh, 1958).
Different in another way, but bearing some
analogy with the minimal flicker criterion, is the
use of the distinctness of the straight-line border
between two contiguous simple test stimuli of
different spectral compositions. This matching
criterion stipulates that the distinctness of the
border is minimal with respect to variations in
either sense of a physical parameter controlling
one of the test stimuli. If the relative spectral
radiant power distributions of both test stimuli
are kept constant, and the absolute radiant power
(or luminance) of one of them is varied, the
criterion provides a method of heterochromatic
photometry yielding results similar to those ob-
tained by the flicker method, but different from
those obtained by the direct comparison method
(Boynton and Kaiser, 1968). Distinctness of the
border is used as matching criterion in a different
way in the matching procedure employing two
adjacent complex test stimuli, developed by
Kaiser, Herzberg, and Boynton (1971), and
Wagner and Boynton (1972). The two halves of
each complex test stimulus are contiguous along a
straight line and are occupied by unform stimuli
of different relative or absolute spectral composi-
tion. The matching criterion is that the distinct-
ness of the internal straight-line border is the
same in the two test stimuli.
- Any of the above criteria appropriate to the
direct matching of two test stimuli can, of course,
be used for indirect matching by the method of
strict substitution. One of the two test stimuli is
then kept fixed and forms part of the invariant
display situation of the other, variable, test
stimulus. However, in indirect matching by strict
substitution, a very much wider range of match-
ing criteria is available. In that case, the proposi-
tion defining the criterion does not involve the
actual equating of some appearance quality of the
Classification of Matching Procedures 287
test stimulus with that of a more or less similar
stimulus in the invariant display situation. Two
interesting visual judgments in this category were
developed by Hurvich and Jameson (1955). One
test stimulus only is entailed, and the observer
judges whether its appearance (1) possesses neither
the qualities of blueness nor yellowness, or (2)
possesses neither the qualities of redness nor
greenness. These criteria are absolute judgments
that rest on the existence of the so-called unique
hues: blue and yellow, red and green, and on the
observer’s ability to recognize for each of them
whether the appearance of any given test stimulus
partakes in some degree of the hue in question.
Empirically, no stimulus appearance partakes of
both blueness and yellowness, or of both redness
and greenness. (The derivation of chromatic-
response or chromatic-valence curves for spectral
stimuli, using these criteria, is summarized in
Section 5.4.12.)
Propositions defining criteria for indirect
matching by strict substitution may correspond
rather closely to those that lie at the root of the
extensive development of visual scaling. For ex-
ample, the observer may be required to assert that
the test stimulus appears midway in color be-
tween a fixed red and a fixed white reference
stimulus, all three being presented simultaneously
in the visual field. Or, the observer may have to
make the judgment that the test stimulus is, say,
five times the brightness of a fixed reference
stimulus. The numerical assessment of “sensation
magnitudes,” which seems to be involved in vari-
ous judgments of this kind, presents certain logi-
cal difficulties, particularly to some observers.
However, in practice, the degree of consistency of
results obtained with visual judgments in these
terms has enabled them to be applied successfully
in applications and theoretical developments. (See
Chapter 6 for material on scaling in color vision,
not tied to the limited application to color match-
ing procedures.)
One concluding remark may be made about
matching criteria. Almost any statement about
the appearance of a test stimulus could serve as a
basis for a special kind of matching procedure.
However, to be worthwhile, the procedure must
yield results bearing on practical problems or be
susceptible of interesting theoretical interpre-
tation. Contrived matching procedures such as
that mentioned earlier based on the visual judg-
ment: “The test stimulus differs in hue from a
given fixed comparison stimulus,” can be of use
only in illustrating extreme cases.
288 VISUAL EQUIVALENCE AND VISUAL MATCHING
5.2.5 Some Particular Matching or
Equivalence Procedures
The display situations and test stimuli for several
matching or equivalence procedures employed in
particular investigations are illustrated diagram-
matically in Figure 1(5.2.5). The different proce-
dures are labeled (a) to (m).
(a) Matching in a symmetric bipartite field small
enough to be imaged in the rod-free foveal area of
the retina. Wright (1929-1930) used monochro-
matic primary stimuli R, G, B to make complete
color matches of monochromatic stimuli A mixed
with one of the primaries as desaturating stimulus
D. His results, and closely similar measurements
by Guild (1931) using primaries of wide-band
spectral distributions, provided the spectral chro-
maticity data on which the CIE 1931 Standard
Colorimetric Observer is based [Sections 3.3.3(a)
and 5.15(ii)].
(b) Matching in a large, foveally centered sym-
metric bipartite field. With such large fields, the
two halves of a central area of some 1° to 2°
diameter may show a well defined color dif-
ference, both differing in color from the sur-
rounding half-fields, when the latter are in good
color match with each other. This occurs particu-
larly when the stimuli being matched have widely
different spectral compositions (Phenomenon of
Maxwell Spot). Stiles (1955b) instructed observers
to ignore the small central area and from their
matches with spectral stimuli A, suitably de-
saturated by D, derived complete color-matching
functions. To reduce to probably insignificant
amounts any intrusion of rod vision in the match,
the primaries R,G,B were changed from red,
green, and blue to red, yellow, and blue for the
longer wavelength test stimuli, the data being
finally transformed to a common reference set of
primaries. A 14° diameter guard-ring surround z
of the same spectral composition as the compari-
son stimulus C, separated from the matching field
by a circular grey line, countered any disturbance
of the match attributable to the high contrast
edge between matching field and dark surround.
This feature was probably of more importance in
small field (2°) matching which was investigated
concurrently [Sections 5.5.3, 5.5.4, and 5.15(11)].
(c) Matching in a large foveally centered sym-
metric annular bipartite field. Speranskaya (1958,
1959) occluded a 2° diameter central area to
eliminate match disturbance by the Maxwell Spot.
She used a single set of primaries R,G, B
throughout. Her data, after some correction for
rod intrusion, were combined by Judd (CIE, 1960)
with those of Stiles and Burch (1959) in the
specification of the CIE 1964 Supplementary
Standard Colorimetric Observer [Section 3.3.3(b)].
(d) Heterochromatic brightness matching by the
minimal flicker method. Ives (1912) specified the
field pattern (that shown) and other conditions
best suited for this matching procedure. As the
frequency of alternation of test T and comparison
stimuli C of different chromaticities is increased,
chromatic flicker disappears first. At a higher
frequency, the optimal condition for application
of the brightness criterion is reached. This
frequency, selected by the observer, depends on
the luminance level of the stimuli and on their
respective chromaticities [Sections 5.7.1(i) and
eR eaeleayp
(e) Matching by slow alternation of test and
comparison stimuli in a single small test area.
Alternation at a rate slow enough for the observer
to assess separately the complete color ap-
pearances of test JTand comparison stimuli C was
used by Enoch and Stiles (1961) in determining
the effect on the color of a monochromatic
stimulus A of changes in its angle of incidence on
the retina. The main constituent of both test and
comparison stimuli was monochromatic and of
the same wavelength. Just sufficient mixtures of
the primary stimuli R, G, B were added to each A
to enable satisfactory complete color matches to
be made. The monochromatic component of each
stimulus entered the eye pupil as a narrow pencil.
The retinal angle of incidence obtained by dis-
placing the point of entry in the pupil varied from
normal to about 12° for the fully dilated pupil, a ©
mydriatic being used, if necessary (Section 5.11).
(f) Matching in a symmetric bipartite field
imaged on the extrafoveal retina. Clarke’s inves-
tigations (1960a,b, 1963), in which the procedure
illustrated was used (among others), was mainly
directed to the breakdown of the additivity law
for complete color matching in the extrafoveal
retina and to the effects of rod participation in
the match. The main complication is the tendency
of the matching field to fade on steady fixation
FP (Troxler’s Effect), which is particularly marked
when the field is imaged well outside the foveal
area (Clarke, 1960a). The fading is largely avoided
by exposing test 7 and comparison stimuli C
 Classification of Matching Procedures 289

simultaneously in brief observation periods that (i) Matching with differential conditioning of the
must, however, be long enough for color judg- two halves of a foveally centered symmetric bipar-
ments to be made (Section 5.6.3). tite field. In the investigation by MacAdam
(1956), accurate fixation on the center of the
matching field was maintained throughout, and
(g) Matching of test and comparison stimuli
different conditioning stimuli (X and Y) were
imaged on small areas of widely different parts of
exposed in the respective halves of the matching
the retina. In this procedure, the display situa-
field in intervals between the actual presentation
tions of the stimuli being matched are clearly
and matching of the test T and comparison stimuli
different because of the test locations on retinal
C. As chromatic adaptation was the main objec-
areas with different properties. In addition, the
tive of the work, the luminances of the members
separation @ of the areas makes it possible to
of each of the six pairs of differently colored
introduce conditioning stimuli that can modify
conditioning stimuli were approximately the same.
the observer’s state of adaptation differentially.
Matching by MacAdam’s method for each pair of
In Moreland and Cruz’s work (1959), the pattern
conditioning stimuli may be regarded as a partic-
shown was used with azimuthal angles ~ of 0°
ular matching procedure. The use of numerous
and 90° and radial angles 8 up to 50°. The
different pairs then represents a limited group of
comparison stimulus C, and R,G, B primary
asymmetric matching procedures (Section 5.2.3).
mixture, was imaged on the fovea. The quantities
Whether the concept of effective display situation
of these primaries in a match with a monochro-
is applicable for this group and whether the sec-
matic test stimulus A were used to calculate foveal
ond modified transitivity principle for direct asym-
chromaticities specifying the apparent color of
metric matches holds good is questionable. The
the extrafoveal stimulus A. Plotted in a diagram
answers turn on whether the conditioning of one
showing also the foveal spectrum locus for the
half-field also affects the state of the other half-
same observer, these chromaticities display the
field so as to modify significantly the match.
progressive modification of color perception and
MacAdam’s subsequent analysis of his data (1961,
its dimensionality with increasing extrafoveal an-
1963) suggests by its form the validity of the
gle 8 (Section 5.6.3).
effective display concept and the transitivity
principle mentioned for the conditions used
(h) Dichoptic matching with differential adapta- without perhaps establishing them (Stiles, 1967)
tion of the two retinal test areas in left and right [see also Section 5.12.2(v)].
eye, respectively. This method, pioneered by
(j) Matching the color difference between two
Wright (1934), was applied by him to compare
contiguous test stimuli with the corresponding dif-
the effects of chromatic adaptation on color
ference for a second copresent pair. In the work
matches during the recovery from different condi-
of Wyszecki (1965) and Wyszecki and Fielder
tioning stimuli (X and Y) of the two eyes. In
(1971), the procedure, based on this matching
Hunt’s procedure (1953), illustrated, a cyclic pre-
concept, involved just three uniform hexagonal
sentation (Period A followed by Period B) of
stimuli J,, J>, J, surrounded by a large white sur-
conditioning and test stimuli was used. The re-
round z. Each hexagon was filled by a variable
sults refer, in effect, to a constant adaptation at a
mixture of the three instrumental primaries
particular time in the cycle. To facilitate the lock-
R,G, B. The observer has to assess three color
ing of the half-fields presented to the respective
differences in the stimulus array provided by
eyes in a dichoptic bipartite matching field, four
the contiguous hexagon pairs (Jj, J>), (J), J3),
fusion points (FuP) were constantly present in
(J,, J;). The final match is reached when all three
the fields of both eyes. Large and small (10° and
hexagons appear equally bright and when their
2° diameter) symmetric matching fields were used.
three color differences are judged to be equal.
The possibility of stimuli applied to one eye mod-
They then define in trichromatic color space a
ifying the display situation of the test stimulus
perceptually equilateral triangle on a surface of
applied to the other eye arises. However, some
constant brightness (Section 7.10.6).
experimenters, using comparable methods, believe
such effects to be slight. Frequently, the dichoptic (k) Matching the distinctness of the straight-line
matching method is also referred to as haploscopic borders contained in two adjacent complex stimuli.
matching or as binocular matching [Section The simplest application of the distinctness of
5.12.2(i)]. border concept as a method of heterochromatic
 LO Ate

N
Zen
ec a
ee
See: '
sa

EL
[=
Ee
.
>
»
t na 2 m > N
|
co

(f) (mono)

Yas 2 boaE,
ye OG
al )
ve / eka
‘ \‘ / A
A
4
4
y Y))
(25)
hd RIGHT DICHOPTIC if)
PERIOD A B A B A
t —
ee
Several seconds

(h) (dichopt)
Fig. 1(5.2.5)._ Diagrams (a) to (m) illustrating the display situations and matching stimuli for
various equivalence and matching procedures. Symbols used: 7: test stimulus; C: comparison
stimulus; R,G, B: set of primary stimuli; A: monochromatic stimulus; D: desaturating stimulus;
I: increment stimulus; J,, J,,J;,K ,,K 2,K3,K,: components in complex stimuli producing
perceptual differences; M: variable stimulus in equivalence determinations; X, Y, Z: condition-
ing stimuli; FP, OP, FuP: small “light” points, and fusion points, all steadily exposed; hatched
area: dark surround or dark field; ¢: time variable during observation; where no time sequence is
shown, the observer is adapted to a steady stimulus pattern, apart from the test stimulus
 acer
X(M) orX(M)+I(A)
———____
s

(m) (mono )
adjustments to reach a match. An exception is made in the case of flicker matching, where the
rapid alternation in the test area is not illustrated; mono: the stimuli concerned in the match are
exposed to one eye only, the other eye being kept in darkness or dim light (monocular viewing);
bino: both eyes view the same stimulus pattern (binocular viewing); dichopt: the two eyes view
somewhat different patterns, the match depending on what is seen in the combined binocular
field (dichoptic viewing). Most of the diagrams are not to scale. In the time-sequence diagrams,
the ordinate defines, in a limited sense, the magnitude of the indicated stimulus, showing when
it is exposed and for how long.
292 VISUAL EQUIVALENCE AND VISUAL MATCHING
matching involves only two contiguous simple
test stimuli that are set to minimal distinctness of
border (Boynton and Kaiser, 1968). The first step
in a more searching study of border distinctness
(Kaiser, Herzberg, and Boynton, 1971), illustrated
in the diagram, was to vary the luminance of the
monochromatic field K, to obtain minimal dis-
tinctness of the border with the fixed white field
K,. In further measurements, the contiguous
stimuli K,; and Ky, constituted one complex
stimulus K, being set at various luminance levels
above and below the value for minimal distinct-
ness of the border with K,. The second complex
stimulus was provided by white-light stimuli K,
and K,, again presenting a straight-line border,
the luminance of K, being made equal to that of
K,. Matching was carried out by varying the
luminance of K, to equalize the distinctness of
the vertical borders in the upper and lower semi-
circular fields. This could be done satisfactorily
provided the resultant achromatic contrast | K, —
K, |/(K, + K,) was less than about 0.3 and above
threshold. In this work, the observer must be
equipped with an achromatizing lens to minimize
the effects of chromatic aberrations of the eye. To
reduce fading of the borders, the observer shifted
fixation up and down between the middle and the
upper and lower borders. In some experiments,
the whole field was occluded for about 0.25 s in
every 4s. The results of the investigations led to a
theory of the contribution of achromatic and
chromatic differences to border distinctness.
(1) Determination of equivalent fields with re-
spect to their effect on the detection of a fixed
increment stimulus. In the procedure used by
Stiles (1939), the relative spectral compositions of
the variable field stimulus M and the fixed incre-
ment stimulus J are, in general, different (two-
color threshold technique). The radiant power of
the field stimulus M is determined at which / is at
the threshold of detection. In practice, best results
are obtained not by allowing the observer to raise
and lower the field level to which the observer is
supposed to be adapted, but by measuring incre-
ment thresholds at field levels giving values in the
neighborhood of J, and by then interpolating to
find the value corresponding to J precisely. The
reciprocal powers of equivalent monochromatic
fields define field sensitivities that, for ap-
propriate choices of the fixed increment stimulus,
may apply to different retinal response mecha-
nisms (Sections 5.12.2 and 7.4).
(m) Determination of equivalent contrast pulses.
As in the previous example, the criterion of equiv-
alence is the effect of the detectability of a fixed
increment stimulus J of a variable stimulus which,
in this case, is an annular contrast pulse M. But
the display situation of the increment pulse J is
much more complex. Pulse J is presented for a
brief period of time before the contrast pulse M
which is intended to bring J to the threshold of
detection. Also, additional conditioning stimuli Y
and X, both steady, are applied respectively to
the contrast pulse and increment stimulus areas.
In this work, Alpern, Rushton, and Torti (1970)
applied their technique to establishing the satura-
tion, or upper limit, of the inhibitory effect of a
contrast pulse on the detectability of an incre-
ment stimulus as the stimulus level in the annular
field was raised by increasing Y. (The bracketed
Greek letters in the diagram identify the various
different-colored stimuli in the notation of Alpern
et al.) The saturation effect was demonstrated for
various sets of conditioning stimuli, designed re-
spectively to confine the threshold response to the
rod or one of several cone mechanisms. In doing
this, the authors employed a modified form of
equivalence criterion. The contrast pulse M on
the steady field X had to raise the threshold of the
increment stimulus to a fixed multiple of the
value obtained when no contrast pulse was ap-
plied and when the only stimulus in the annular
area was the steady field X. This procedure was
devised to allow for the effect on the threshold of
light scattered from X onto the central area (Sec-
tion 7.6).
Six of the foregoing examples of matching
procedures, namely (a) to (f), must be expected to
satisfy the three conditions for quasi-symmetric
matching. This will continue to hold good if the
total display situation is changed by including
additional or modified conditioning stimuli, either
pre-exposed or steadily present, provided they are
symmetrically arranged with respect to, and are
likely to affect equally, the two halves of the
bipartite field. Examples (g), (h), and (i) represent
specifically asymmetric matching procedures. In
Examples (j) and (k), the test stimuli are complex.
The equivalence procedures (1) and (m) contrast a
simple and a fairly elaborate display situation.
Among the many matching procedures unrep-
resented in Figure 1(5.2.5) the following may be
mentioned:
(1) Dichoptic procedures in which the observer’s
viewing conditions are not so rigidly defined as in
Example (h). A central septum may be used to
separate two scenes containing simple test stimuli

in different display situations, viewed respectively
by the left and the right eye of the observer. In
making the match, the observer’s eyes are free to
wander, the left eye scanning one scene, the right
eye the other (Winch and Young, 1951).
(2) Viewing conditions employing binocular vi-
sion and approximating very closely those of
every day life. In work by Helson, Judd, and
Warren (1952), observers were given preliminary
training in which they learned to identify colored
test samples, in a standard reference situation, by
their Munsell specifications. Subsequently, they
viewed colored test stimuli in different display
situations and reported their apparent colors using
the Munsell descriptions. This is a refinement of
color-naming procedures in which nonspecially
trained observers employ for description their
ordinary equipment of color nomenclature. Proce-
dures on these lines may be designated matching
by long-term memory [see also Section 5.12.2(vii)].
(3) Newhall, Burnham, and Clark (1957) used a
form of asymmetric matching designed to test
how well colors are remembered over very brief
periods. A color is presented alone for a limited
period (5-10 s). Some five seconds later, a vari-
able test color is set up by the subject (or selected
from a range of samples) which is judged by
memory to be in match with the first test color.
The conditions for observation of the two test
stimuli are closely similar, but the results show
that the subject makes matches in which the
saturation and luminance of the second test
stimulus are appreciably greater than when the
normal method of matching with simultaneous
presentation is used. Newhall et al. were able to
show that the asymmetric factor to which these
differences are in the main attributable, is the
matching of a stimulus actually being perceived
with the short-term memory of a stimulus ex-
posed a short time earlier.
(4) Complex test stimuli, produced by superim-
posing a resolvable dark bar grating over a uni-
form and minimally desaturated monochromatic
test field, were used by Nunn (1977) to make
complete color matches with a simple comparison
test stimulus consisting of an adjustable R,G, B
mixture. Even with identical display situations of
the simple and complex test stimuli, large color
differences were recorded, in some cases. Mea-
surements were also made with different display
situations that included internally structured or
pre-exposed conditioning stimuli.
Any direct color-matching procedure (1.e., one
in which test and comparison stimuli are pre-
sented to the observer in the same observation),
Maxwell’s Method of Color Matching 293
reduces to matching by strict substitution if
the comparison stimulus is kept constantly the
same. The range of possible matches is thereby
drastically limited, but this limitation may be
useful in attempting to elucidate some of the
causes for deviations from the ideal laws of tri-
chromatic matching. The approach to color
matching through such a simplifying restraint is
dealt with in Section 5.3, Maxwell’s method.
5.3. MAXWELL’S METHOD OF COLOR
MATCHING
5.3.1 Historical Note
Maxwell in his original determination of his
“mixture curves,’ which would now be called
trichromatic color-matching functions, derived
them from a set of matches in each of which a
mixture of up to three different monochromatic
stimuli was adjusted to match a fixed white-light
stimulus (Maxwell, 1860a,b). By contrast, the
principal later determinations of the trichromatic
matching data for the average normal eye (Guild,
1931; Wright, 1928-1929; 1929-1930; Stiles,
1955b; 1958; 1963; Stiles and Burch, 1955; 1959;
Speranskaya, 1958; 1959) have used mainly sets
of color matches of the monochromatic stimuli,
each desaturated with the minimal amount of one
fixed primary stimulus necessary to make possible
a match with a mixture of the remaining two
fixed primary stimuli. If trichromatic color match-
ing conformed completely to the linear laws of
proportionality and additivity, deemed to hold
for the CIE standard colorimetric observers,
Maxwell’s method and the second method, ap-
propriately described by Crawford (1965) and
Lozano and Palmer (1968) as the maximum
saturation method, would give the same color-
matching functions, although it is realized that in
Maxwell’s method the effect of random matching
errors on the values of the derived functions will
be magnified. However, if the validity of the
linear laws in general trichromatic color matching
is questionable, then the Maxwell method has the
merit that all the matches used are made in a field
of constant luminance and chromaticity and any
nonlinearities caused by changes in the luminance
and chromaticity of the matching field will be
eliminated. Obviously, if linearity fails in some
degree, no single set of color-matching functions
will predict correctly all matches. The Maxwellian
set might tend to give better predictions for
294 VISUAL EQUIVALENCE AND VISUAL MATCHING
matches of colors in a domain around the refer-
ence white used in their determination, whereas
for predictions of matches of a range of highly
saturated colors, the maximum saturation set
might have advantages. However, these are no
more than general tendencies. Difficulties in defi-
ning a rigorous and generally applicable system of
colorimetry on the basis of the strong trichro-
matic generalization led Blottiau to propose, in
1947, a “Colorimétre trichromatique 4 visibilité
invariable.” Blottiau’s proposal rests, in effect, on
an extension of Maxwell’s method to color mea-
surement generally. In his colorimeter, an adapta-
tion of the Donaldson instrument (Donaldson,
1935), to any stimulus to be measured for color,
an adjustable mixture of three primary stimuli
had to be added so that the resulting total mix-
ture matched a reference white. The color specifi-
cation of the stimulus under test could then be
calculated from the amounts of primary stimuli in
the total mixture, and the amounts required in a
calibrating match with the reference white of the
trichromatic mixture by itself (Blottiau, 1947).
Blottiau was much concerned with the consider-
able defects from the linear laws of color match-
ing that he found in his own measurements on
this issue. Although Blottiau’s ideas are certainly
soundly based, later work on the additivity of
color matches by Trezona (1953, 1954), using a
1.33° bipartite matching field, although confirm-
ing the general character and magnitude of
Blottiau’s observed defects from linearity for indi-
vidual observers, suggested that, on the average,
the deviations were not large compared with just
discriminable color differences (see also Section
5.6.6).
When the data for large field color matching
were determined and standard mean color-match-
ing functions derived, some tests of additivity
were made and some defects were found, particu-
larly for individual observers. The defects in addi-
tivity when the data for the whole group of
observers were averaged, although still not nil,
were regarded as probably insufficient to stand in
the way of the adoption of a large field standard
observer (Stiles 1958, 1963). Since then, various
field trials of the large field observer and more
especially the work of Crawford (1965) on color-
matching functions obtained by Maxwell’s
method, coupled with more recent work on the
same lines by Lozano and Palmer (1968), suggest
that defects in additivity in large field color
matching may be more serious than previously
thought. This is particularly evident for the ob-
servations of a single observer. Defects of the
same character are observed when a small or large
matching field (2° or 10°) is used. Although the
effects are greater in the case of a large matching
field, it cannot be simply rod intrusion in match-
ing that is responsible (Crawford, 1965). A fea-
ture of the work of Lozano and Palmer was their
inclusion of matches by both Maxwell’s and max-
imum-saturation methods on color stimuli that
were neither near-white nor highly saturated, with
chromaticities spread over the whole range of the
chromaticity diagram.
Another phase in the current interest in
Maxwell’s method is its use in studies of the color
matching of color defectives (Alpern and Pugh,
1977; Alpern and Moeller, 1977) and in studies
of color matching at levels of retinal illuminances
causing bleaching of the photopigments
(Wyszecki, 1978; Alpern, 1979; Wyszecki and
Stiles, 1980).
5.3.2 Basis of Maxwellian Trichromacy
Maxwell was able to derive his set of color-match-
ing functions solely from direct matches to a
single white comparison stimulus. As these func-
tions can be applied, at least in the first instance,
only to predicting other matches of the same
kind, it is clear that less strong demands on the
properties of the observer’s color vision are being
made than those needed for full trichromacy in
the normal sense. It is appropriate therefore to
develop ab initio the minimal assumptions un-
derlying a restricted form of trichromatic princi-
ple embodying Maxwell’s concept, which will be
called “Maxwellian trichromacy/C” where C
specifies the particular fixed comparison stimulus
in use. Maxwell’s. method has usually been ap-
plied to a bipartite matching field with exactly
similar juxtaposed halves for simple test and com-
parison stimuli, the latter being white, imaged on
the foveal region of the retina. However, the
method is also applicable to many specifically
asymmetric matching procedures (see Section
5.2.2). In these procedures, dissimilar test and
comparison stimuli are used that are imaged on
more or less widely separated retinal areas. Any
set of conditioning stimuli may be employed,
provided they are not such as to prevent the
observer’s applying the matching criteria ap-
propriate to the matching of a simple test and
comparison stimuli enumerated in Section 5.2.4.
In what follows, only simple test and compari-
son stimuli will be considered. It is questionable

whether any method comparable to Maxwell’s
could be usefully developed for complex test and
comparison stimuli, and these are specifically ex-
cluded. but the comparison stimulus need not be
white. It may have any spectral composition, sub-
ject to the applicability of the matching criteria
just mentioned and to a minor formal limitation
discussed and effectively removed below. It
is essential that for a given comparison stimulus
and total display situation, and a fixed matching
criterion, the only parameter that may be varied
from one observation to another is the absolute
spectral radiant power distribution of the test
stimulus. This does not, of course, exclude the
possibility that any of the conditioning stimuli
may change according to a fixed time schedule.
The exposure durations of the simple test and
comparison stimuli may range from steadily ex-
posed to brief pulses that are not simultaneous.
The matching criterion will usually be taken to
demand complete color match. In fact, this re-
striction will be assumed throughout until in Sec-
tion 5.3.4 a generalized form of Maxwell’s method
is deployed. All considerations in the present
context apply to the matches made by some par-
ticular observer.
For a given total display situation, matching
criterion and comparison stimulus C, the retinal
response in the area on which the test stimulus is
imaged will conform to Maxwellian trichromacy /
C if certain conditions are satisfied which may be
put in the following form:
CONDITION I In using it later, the suffix 6 will be understood,
It is possible to find some set of three different
wavelengths A,,A,,A,, called the primary set,
such that a test stimulus consisting of a mixture
of radiant powers P, of A,, P, of A,, P, of A,,
where P., P,, P, are all greater than zero, matches
the comparison stimulus C. The mixture is unique,
that is, no other mixture of these three wave-
lengths in different amounts gives a complete
color match with C.
CONDITION II
For any primary set of wavelengths A,,A,, A, for
which Condition I is valid, and for each wave-
length A,, a unique mixture of appropriate (posi-
tive) radiant powers of A, and of at most two of
the wavelengths in the primary set, is determin-
able that also matches C. The particular wave-
lengths of the primary set that must be used with
A; depend on A,.
Maxwell’s Method of Color Matching 29S
CONDITION III
(a) Ifa test stimulus that matches C is increased
by adding power at some or all parts of the
spectrum, the modified stimulus no longer
matches C. More precisely, if for two test stimuli
P’ and P” with spectral power distributions {P’}
and {P,’} respectively, P/’ = P’ for all wave-
lengths A, and P’’ > P’ for some wavelengths X,,
then the two stimuli will not both match C.
(b) Suppose, for any three test stimuli P, P’,
and P”, the spectral power distribution of P is
related to those of P’ and P” by the “mixture”
equation: P,;=aP; +(1—a)P,” for all wave-
lengths A, where 0 < a < 1. Then, if any two of
P,P’, P” match C, so does the third. This is
referred to as the additivity law for Maxwellian
matching.
It will be noted that in Condition III, spectral
power distributions of stimuli are being repre-
sented by n-dimensional vectors P. The compo-
nents of these vectors are denoted by P. represent-
ing the total power in a wavelength interval
centered on the wavelength A;. The wavelength
interval has a length less than a small positive
quantity 6. The whole of the relevant spectral
range (the visible spectrum, taken as a suitably
wide but finite band of wavelengths) is divided
into ns such intervals where ns is necessarily a
large number. This device, which substitutes sum-
mations for integrals over the spectrum, has some
mathematical conveniences in the present context.
and n will stand for ns. Our concern will be
mainly with real stimuli (1.e., those for which
P,=0 for all 7), and this will be assumed unless
the contrary is indicated.
If Conditions I and II are satisfied, but not
necessarily III, matching in the stated circum-
stances is described as “Maxwellian trichro-
matic/C in the weaker sense.” However, if all
three conditions are obeyed, we refer to the
matching simply as “Maxwellian trichromatic/
C.” Even when only the weaker form holds, it is
possible to define three functions of A;, denoted
by 7(A,), Z(A,), b(A;), that play the role of color-
matching functions in a set of matching relations
of familiar form. When, in addition, Condition
III holds, it can be shown that the same color-
matching functions and matching relations pro-
vide the necessary and sufficient requirements for
any real test stimulus to match C, whatever its
spectral power distribution.
296 VISUAL EQUIVALENCE AND VISUAL MATCHING

In the matches for each A,, postulated in Con- for a match to C for a test stimulus of any
dition II, we suppose that the experimentally de- absolute spectral power distribution {P;}, we need
termined radiant powers of the components in the to assume Condition III or some equivalent lin-
test mixture matching C are: P,(A;) of A,; P(A;) earity laws. The proof presents no special diffi-
of A,; P,(A;) of A,; P(A;) of A;. However, by culty and will not be given. The result may be
hypothesis, at least one of P,(A;), P, (A;), P,(A;) simply restated as follows: if Maxwellian trichro-
is zero. The functions 7(A,), Z(A;), b(A;) are then macy/C holds good, then, any test stimulus which
defined as follows: matches C and, at the same time, satisfies Eq.
3(5.3.2), exhibits two equivalent properties; each
property implies the other. Obviously, two stimuli
Be aneEON
= a |r P(A;)
that both match C are in match by strict substitu-
tion. In fact, Maxwellian trichromacy/C provides
= Pan P(A;)
the most important example of this concept. The
g(A;) aig’s (ig [1(5.3.2)] validity of Maxwellian trichromacy/C, and the
particular color-matching functions associated
with this trichromatic principle are primarily two
b(A;) = a wal properties of the visual response of the retinal
area covered by the test stimulus in the state to
It is clear by putting A; equal in turn toA,, A,, Az, which it is brought by the conditioning stimuli of
the functions are normalized in the usual way for the display situation. There is thus also a possible
color-matching functions appropriate to mono- dependence on the properties of other parts of the
chromatic primaries of wavelengths A,, A,, Aj. retina on which the comparison and other stimuli
Thus, the normalizing relations are: forming part of the total display situation are
imaged.
It is apparent from their normalization proper-
F(A;) g(A,) 5 (A,) ties that the color-matching functions are linearly
0 0
Nee Ne
independent functions of wavelength. The match-
Se Gove l 0 ing relations expressed in Eq. 3(5.3.2) may be
kai ar) 0 equally well formulated in terms of any new set of
color-matching functions that are independent
[2(5.3.2)] linear combinations of 7(A;), Z(A;), b(A;). The
new set, denoted by ,(A;), $2(A;), ¥3(A;), 1S Ob-
If the test stimulus in any of the mixtures of tained by applying to the original set a 3 x 3
Condition II—say for A; = A,—is represented by nonsingular matrix transformation T with ele-
P, then the spectral power distribution for this ments ¢,,. It follows:
test stimulus satisfies the following three match-
ing relations for a match to the comparison
stimulus C. The functions 7(A,), Z(A;), b(A;) as 3 ¥(A,)P
i=]
=¥(P)
defined in Eq. 1(5.3.2) do, in fact, make this true.
Thus, with P, = P, pscl (okt thes ee ae tin Pp ae sa
n
[4(5.3.2)]
> PAL) PF,— 2,
i=]
where ¥,(A;) = ¢r(A,;) + ti28(A;) + £;3b(A;)
n
and j = 1,2,3. The summations on the left-hand
sy BA) P= P, [3(5.3.2)] side of the equality sign in these three equations
i=1
can be evaluated for a test stimulus P of any
absolute spectral distribution {P,}. The resulting
5 5(A,)P, =P, sums ¥(P) are termed the tristimulus values, or
i=] components of the tristimulus vector ¥(P) of the
test stimulus, with respect to the particular set of
The match used in Condition I also clearly satis- color-matching functions, ~,(A;). On the right-
fies these relations. To prove that the above rela- hand side of the equality sign stand the compo-
tions are the necessary and sufficient conditions nents ¥., of the tristimulus vector ¥. which is
 Maxwell’s Method of Color Matching —_297

common to all test stimuli that match the com- formed sets of color-matching functions,
parison stimulus C. These components are ex-
pressed in terms of the basic primary mixture 8

(P,, P,, P,) postulated in Condition I. Wi(A;) = » tv, (A;)

The chromaticity coordinates or components k=1

of the chromaticity vector »(P) for any test

and
stimulus P are derived in the usual way from the
tristimulus values: 3

Hd) = D P(A) [6(5.3.2)]

YP)
wj(P) = [5(5.3.2)] yields identical and linearly independent tri-
Pelle) stimulus vectors for P“, P®, and P®. Denoting
k=1 the components of the vectors by X(”, XO), XO),
the following relations must hold:
3
itp:
/ ]) =
—_ 0/(P™)
, =
Aes xi 1 )
Lz 2 pj (P)

¥/(P®) = 0(P®) = x [7(5.3.2)]

The tristimulus vector WY. and the corresponding
EP)
/ 3 =
== 0/(P”)
/ 3 =—
a xX?
3
chromaticity vector 4, are termed, respectively,
the tristimulus vector and the chromaticity vector
associated with the comparison stimulus C, for the for j = 1,2,3.
particular Maxwellian matching procedure under There is considerable freedom in choosing
consideration. WY. and p- are obviously changed if XM, X, X®. The only constraint is that they
a new set of color-matching functions is adopted represent linearly independent vectors. As arule,
that is related to the original set by a nonsingular the three columns of the 3 X3 unit matrix
transformation. are adopted. But the set of reference stimuli,
It is readily shown that if one starts from an P‘), P®) P®), must be such that the matrices M,
alternative primary set of wavelengths, for exam- and M, formed from their calculated tristimulus
ple, A,4,Aga> Aba» for which the Maxwellian Con- vectors in the two systems are nonsingular:
ditions I, II, and III are valid, the resulting new N
set of color-matching functions and the corre- (M,);, =>, POANLOS* ef 23
sponding matching relations are obtainable from =
those found for the original primaries A,, A,, Aj,
N
by applying to the latter a particular nonsingular
(My, rr 2 p(A,) PP, Base ppags,
3 x 3 matrix transformation T in the way il- i=1
lustrated above.
The set of color-matching functions in the [8(5.3.2)]
matching equations defining Maxwellian match-
ing to a given comparison stimulus is, to the The unique solution of the following nine
extent of a nonsingular 3 X 3 matrix transforma- equations for the nine elements of the matrix T“”?
tion, indeterminant. Thus, given two sets purport- provides the required transformation:
ing to represent the results of perhaps the same or
3
perhaps different matching procedures, it is nec-
essary to apply to them a common normalization 2 7 (My), = XP
7— 1
LoS ple ly 3

before they become comparable.

The most general normalizing procedure for
comparing two sets of color-matching functions,
W(A,) and $,(A;), 7= 1 to 3, is to find for each a
suitable nonsingular 3 X 3 matrix transformation,
T™) and T‘®), respectively. These matrices are
chosen such that for a selected set of three test
stimuli P“, P™, P®, each of the two trans-
[9(5.3.2)]
The nine elements of the matrix T‘® are similarly
calculated.
The common practice is to take for the refer-
ence stimuli, monochromatic stimuli of unit power
at three wavelengths, ,,A,,A,, so chosen that
298 VISUAL EQUIVALENCE AND VISUAL MATCHING
both sets of three column vectors, ¥,(A,), ¥;(A;),
YA.) and (Aq), $A), (As = 1, 2,3)
are linearly independent sets. If and ¢ refer to
the results of the same matching procedure, the
choice of reference wavelengths to suit y will also
suit @. On the other hand, if they refer to differ-
ent matching procedures, it may not be possible
in certain extreme cases to find wavelengths
A,» A, A- Meeting the requirements for both sets
of color-matching functions. Other power distri-
butions, not exclusively monochromatic, must
then be adopted for the stimuli P“, P®, P® to
bring the two sets of color-matching functions to
a common normalization. This is always possible.
5.3.3. Maxwell Trichromacy and Full
Trichromacy in Quasi-Symmetric Matching
Tests of Maxwellian trichromacy for a number of
different comparison stimuli C, but with other-
wise the same total display situation and with
complete color matching as the criterion, may
show that, although Maxwellian trichromacy is
valid for each C, the sets of color-matching func-
tions, when all suitably normalized (see Section
5.3.2), are not the same. For example, this may
occur in studying the effect of high stimulus levels.
In such an experiment, it appears probable that a
different set of color-matching functions holds
good than that valid when the comparison stimu-
lus has the same relative spectral composition but
is of only moderate or low level. The analysis of
such results can assist in the identification of the
factors in matching that modify the simple tn-
chromatic model (see Section 5.6.5).
Another especially interesting case arises if the
total display situation is such that the matches
with many different comparison stimuli C, re-
garded as examples of direct asymmetric match-
ing, satisfy the conditions for a quasi-symmetric
matching procedure in a bipartite field. In that
case, it may be quite possible for Maxwellian
trichromacy/C to hold whatever the comparison
stimulus used, subject perhaps to some upper
limit on stimulus level. Also, the matching rela-
tions may be expressible always in terms of the
same set of color-matching functions y,(A,),
W>(A;), ¥3(A;). Under those conditions, the tri-
chromacy of the retinal area on which the bipar-
tite field is imaged might be appropriately
described as full. The conditions for an indirect or
substitution match of two test stimuli P and P’
via some acceptable (not of excessively high level)
comparison stimulus C, would then be expressible
in the familiar form:
SADR = SA)
i=] i=]
SA) = Sv(A)P [1653.3]
i=1 i=1
y ¥3(A;)P; = s W3(A;)P/
i=1 i=1
These relations would be independent of the com-
parison stimulus used as intermediary. Exactly
the same relations would hold for a direct match
in the bipartite field between a test stimulus P of
spectral composition {P;} and a comparison
stimulus C whose spectral composition is denoted
by (P’}.
However, it is not in fact possible for Maxwel-
lian trichromacy/C, as defined by the Conditions
I, II, and III (see Section 5.3.2), to hold for all
comparison stimuli, quite apart from any upper
limit on the level. For certain comparison stimuli,
there is always a technical failure of Condition I.
This is so despite the fact that the color-matching
properties of the observer’s retinal test area in the
prescribed display situation might qualify as fully
trichromatic, with the matching relations ex-
pressed by Eq. 1(5.3.3) continuing to apply.
To clarify this point, consider quasi-symmetric
matching by an imaginary observer for whom Eq.
1(5.3.3) holds without restriction. The observer’s
set of all positive color-matching functions are
assumed to be those plotted in Figure 1(5.3.3)A.
These may be thought of as corresponding to the
spectral absorption factors of three (hypothetical)
visual pigments responsible for vision in the bi-
partite field area. The color-matching functions
have been modified by applying the factor A, to
allow for the fact that in Eq. 1(5.3.3) the spectral
radiant powers P, and P’ of the two stimuli P and
P’ are expressed in watts and not in quanta per
second. The assumed pigment-absorption curves
are not consistent with the color-matching func-
tions of any known form of human color vision,
but they are feasible as representing the proper-
ties of some biological response system. They
have been chosen so that the tristimulus values of
the different test stimuli lead to a chromaticity
diagram with a spectrum locus presenting various
special features, only some of which are paralleled
in the usual chromaticity diagrams for normal
color vision.
 Maxwell’s Method of Color Matching 299

(nm)
600 700 800

25.0 22.5 20.0 I/d(em™') 175 15.0 12.5x10>

(R,G,B)
of
Logarithms
Hypothetical
Functions
-Matching
Color
@Q
~~~ \_800-680nm
(r,g)-Chromaticity Diagram

Fig. 1(5.3.3). (A): R,G, B are the color-matching functions of a trichromatic system based on
a hypothetical set of visual pigment spectral absorption curves. (B): The corresponding
(r, g)-chromaticity diagram shows the spectrum locus and the chromaticity point W of the
equal-energy spectrum calculated on the wavelength basis. The chromaticity points of all real
colors correspond in this system to points in or on the boundary of the closed locus formed by
segments of the spectrum locus and the broken tangent lines.

Figure 1(5.3.3)B shows this special chromatic-
ity diagram with its odd spectrum locus. The
chromaticities of all test stimuli are represented
by points of the smallest convex set that contains
the spectrum locus. (A convex set is one con-
taining all points of the straight line joining any
two points of the set.) The boundary of this set,
which is included in it, is indicated in the figure.
It contains points and segments of the spectrum
locus as well as segments of those straight lines
that can be drawn through at least two distinct
points of the spectrum locus and which have no
points of the locus in one of the half-planes into
which the straight line, excluding the line itself,
300 VISUAL EQUIVALENCE AND VISUAL MATCHING
divides the diagram. It is apparent from the dia-
gram that to any interior point Q of the convex
set, three wavelengths may be chosen whose chro-
maticity points on the spectrum locus define a
triangle that includes Q in its interior. Since
matching is quasi-symmetric, there will certainly
be some comparison stimuli C that are matched
by test stimuli P, whose chromaticities coincide
with Q. It follows that for any such comparison
stimulus, matching will satisfy Condition I for
Maxwellian trichromacy/C. It is also similarly
clear that Condition II will be satisfied. However,
for no point Q on the boundary of the convex set
can Conditions I and II be satisfied. Four
boundary points, Q,,Q,,Q3;,Q, marked in Fig-
ure 1(5.3.3)B, illustrate various causes of failures.
Boundary point Q, represents the chromaticity
common to a range of wavelengths at the end of
the visible spectrum. A mixture of three different
wavelengths from within that range, all having
necessarily chromaticity Q,, will serve. For any
choice of these wavelengths, infinitely more mix-
tures could be used, so that the uniqueness re-
quired by Condition I fails. Similarly, for Q, on a
straight section of the spectrum locus, any set of
three different wavelengths on this section that
straddle Q, will provide a mixture required by
Condition I, except that again it will not be
unique. For the boundary point Q,, a unique
mixture of the two wavelengths 444 and 513 nm
would give the desired match, but with the obliga-
tory addition of a contribution from a third wave-
length, the match must break down. Finally, a
boundary point like Q, is considered which corre-
sponds to a single wavelength on a positively
curved part of the spectrum locus. Because no
mixture of two or more wavelengths can yield the
chromaticity Q,, Conditions I and II fail.
In the chromaticity diagram for normal color
vision, for example, the CIE 1931 (x, y)-chro-
maticity diagram [see Figure 3(3.3.3)], the convex
set of real chromaticity points is bounded by the
complete locus and the line of purples. All wave-
lengths from 700 to 830 nm have a common
chromaticity and resemble point Q, of Figure
1(5.3.3)B. For wavelengths from 700 to 650 nm,
the spectrum locus is a straight line and all points
in it have the same drawback as point Q,. All
remaining wavelengths of the spectrum locus are
of the type Q,. The points on the line of purples,
on. the other hand, fail in the same way as point
Q;. That for the CIE diagram no wavelengths
have chromaticities in the interior of the convex
set, and that there is no internal loop in the
spectrum locus, are empirical properties of normal
human color vision and are not necessary for
trichromatic systems in general.
This formal difficulty in relating Maxwellian
and full trichromacy can be met by reformulating
the definition of full trichromacy. Full trichro-
macy of a retinal area, evaluated by a particular
quasi-symmetric matching procedure in a bipar-
tite field involving any given comparison stimulus
C, means one or the other of the following:
(a) Maxwellian trichromacy/C holds with a
common set of color-matching functions.
(b) A sequence of comparison stimuli C,,
Cy, sn » Cysn%eycan, bet specified, “for, caches
which (a) applies.
The absolute spectral power distribution func-
tions..of C,, C,,...,C,»:+s CONVeRSe, abmeaam™
wavelength uniformly onto that of the given
stimulus C. Continuity then ensures that the
matching relations, expressed by Eg. 1(5.3.3),
apply in full trichromacy without any reservation
on the implied comparison stimulus.
An immediate consequence is that in full
trichromacy the linearity laws of matching will
apply in a form wider than that required
by Condition III for Maxwellian trichromacy.
For any four stimuli P,, P,, P;, P, whose absolute
spectral power distributions are given by
{Pi;}, {Po:}, (Psi}, (Pa;}, respectively, the follow-
ing linearity laws apply:
(i) If {P,;} color matches {P,,;}, then {aP,;}
color matches {aP,;} for any positive a.
(ii) Of the three possible color matches, {P);}
color matches {P;}, {P3;} color matches {P,;},
{P,; + P3;} color matches {P,;} + {P,;}, if any
two hold, so does the third.
For matching procedures that are not quasi-
symmetric, the relationship of Maxwellian and
full trichromacy is more complicated than that
described above. Most important are the facts
that, in specifically asymmetric matching, some
comparison stimuli may be unmatchable by any
test stimulus and some test stimuli may not figure
in a match with any comparison stimulus. The
general case is discussed in Section 5.3.5.
5.3.4 Maxwell’s Method in Nontrichromatic
Matching
In Maxwell’s method, the matching criterion need
not be limited, as it was in Sections 5.3.2 and

5.3.3, to complete color match. Any of the cri-
teria for matching simple test and comparison
stimuli discussed in Section 5.2.4 may be allowed.
Also, “dichromatic” or even “monochromatic”
Maxwellian matching may be admitted as alterna-
tive possibilities. In these cases, the mixtures of
three wavelengths in Conditions I and II for
Maxwellian trichromacy are replaced, respec-
tively, by mixtures of two wavelengths or by a
single wavelength. Clearly, such cases arise with
color defective observers, or with color normal
observers when the test stimulus is imaged on the
far extrafoveal retina, or when the matching crite-
rion is one of incomplete color match.
There is no evidence that the postulation of
more than three wavelengths would enable Condi-
tions I and II to be satisfied in any particular
case. This is not necessarily at variance with the
existence of more than three retinal response sys-
tems with linearly independent spectral sensitivity
curves in the retinal area on which the test stimu-
lus is imaged. For example, in the near extrafo-
veal region there are certainly four such response
systems, three cone systems and the rod system.
In assessing the appearance of a simple test
stimulus, which is perceived as a uniform patch,
the eye seems able to employ, in effect, only three
independent qualities, even though these may re-
sult from the summation in the visual process of
the responses of more than three systems with
linearly independent spectral sensitivities. This
would make possible obedience to Conditions I
and II (Maxwellian trichromacy/C in the weaker
sense). However, Condition II could be met only
if the summation in the visual process were such
as to relate the appearance qualities linearly to
responses of the component retinal systems. It
must be emphasized that, in Maxwell’s method,
the total display situation and the comparison
stimuli are kept fixed. This limits any variations
in the condition of the test area to possible effects
of the different spectral compositions of the test
stimuli matching the comparison stimulus. If the
factors mentioned are not held constant, then
more than three mixture primaries may be re-
quired; but this is not Maxwellian matching. An
example is Trezona’s method of tetrachromatic
color matching (Section 5.4.13) in which, in fact,
two matches under different conditions are made.
It will be convenient to restate, for the general
case, the conditions for Maxwellian N-chro-
macy/C, for given total display situation, match-
ing criterion and nonzero comparison stimulus C,
where N, the number of primaries required, is left
Maxwell’s Method of Color Matching 301
open. The restated conditions are as follows:
CONDITION I’:
It is possible to find some set of N different
wavelengths, A,,...,A,, called the primary set,
such that a mixture of radiant powers P, of
Xe. Po X,, where P,,...)/P, are all greater
than zero, matches the comparison stimulus C.
The mixture is unique; that is, no other mixture
of these N wavelengths in different amounts gives
a complete color match with C. If N= 1, by
“unique mixture” is to be understood a specific
amount of A, only.
CONDITION II’:
For any primary set of wavelengths (A,,...,A,)
for which Condition I’ is valid, and for each
wavelength A,;, a unique mixture of appropriate
(positive) radiant powers, of A, and of at most
(N — 1) of the wavelengths of the primary set, is
determinable that also matches C. The particular
wavelengths of the primary set that must be used
with A, depend upon A,. If N = 1, by “unique
mixture” is to be understood a specific amount of
A; only.
CONDITION III’:
This condition is identical with Condition III,
given in Section 5.3.2.
The derivation of N normalized color-match-
ing functions follows exactly the same course as
the derivation of 7(A,), Z(A,), b(A;) for N = 3. If
Maxwellian N-chromacy/C is assumed to hold,
matching C and satisfying the generalized match-
ing relations are equivalent statements about a
test stimulus P of absolute spectral power distri-
bution {P,}. Each of these statements implies the
other. The generalized matching relations may be
expressed by the equation
Dw)? = ¥(P) =¥%o (i= 1,2,...,N)
[1(5.3.4)]
where ¥(A;) are N linearly independent color-
matching functions, and ¥,(P) are the compo-
nents of the corresponding N-stimulus vector
(P) for test stimulus P; ¥.; are the components
of the N-stimulus vector Y, associated with the
comparison stimulus C.
To go over completely to matrix notation, the
N color-matching functions ¥,(A;) are made
the rows of an (N X n) color-matching matrix, the
302 VISUAL EQUIVALENCE AND VISUAL MATCHING
test stimulus vector P is an (n X 1) column ma-
trix, and W(P) and WV. are (N X 1) column
matrices. Equation 1(5.3.4) is then stated alterna-
tively in the single matrix equation
Wo PoP) We 1s A)
Two color-matching matrices y and y’ are said to
be equivalent if they are related by a nonsingular
N X N matrix T such that
yy
The original matching relations restated in terms
of the color-matching matrix ~’ are then given by
y- P=W(P)=%
where
W(P)=T-¥(P) and YO=T-Y
Wé and VY, are equivalent if they are related by the
same transformation matrix T as that connecting
the color-matching matrices of the matching rela-
tions in which they appear. The effect of the
transformation T on the N-chromatic chromatic-
ity vector u(P) is less simple. It follows
¥/(P) 2 UMP)
pi(P) = 5 Tae naa
SeeyGP) Ded eek)
k=1 k=1[=1
[3(5.3.4)]
In the special case that the matching proce-
dure is quasi-symmetric in a bipartite field and
the criterion is for complete matching, full N-chro-
macy is defined like full trichromacy.
For every comparison stimulus C, either
matching is Maxwellian N-chromatic/C with
matching relations expressible in terms of a com-
mon set of color-matching functions (A), or C
is the limit of a sequence of comparison stimuli
for which this is true.
5.3.5 Maxwellian Matching as Correlation of
Stimulus Spaces
Simple test and comparison stimuli T and C may
be represented as points in two n-dimensional
stimulus spaces, 7-space and C-space. The coor-
dinates of a point T or C are the components P,
of the n X 1 matrix (or column vector) P defining
the stimulus. In special cases, the spaces may be
enlarged by additional coordinates for parameters
specifying the polarization and retinal angle of
incidence conditions of the stimuli. However, this
complication will not be allowed for in the pre-
sent discussion. The parameters in question will
be assumed to remain unchanged throughout.
These spaces provide a useful way of thinking of
the results of matching when matches are made
for a wide range of test and comparison stimuli,
but under the condition that the matching crite-
rion and the respective display situations of test
and comparison stimuli are kept unchanged. Two
principal cases arise:
(i) The two display situations may differ, in
particular, when different colored conditioning
stimuli control the states of test and comparison
retinal areas. These are the circumstances of much
work on chromatic adaptation.
(ii) Matching may be quasi-symmetric with the
two display situations virtually the same. The
effects of changes in the luminance or chromatic-
ity of the comparison stimulus are then of prime
interest. In this category are the match changes
resulting from the bleaching of the visual pig-
ments at high stimulus levels.
The following notes on match correlations be-
tween stimulus spaces provide a preliminary to
the more detailed consideration of chromatic
adaptation and pigment bleaching in Sections
5.12, 5.6.5, and 8.2.6.
Not every comparison stimulus, that is, not
every point in C-space, will necessarily have a
correlate in 7-space representing a test stimulus
that matches it, and conversely. But, in general,
where there is a correlated pair of points, denoted
by T°? and C?, there will be an extended set of
points Met;[C] in T-space, including T®,
that match C“?, and a set of points Met-[T]
in C-space that match T“?. In general, these sets
are (n — 3)-dimensional, but may reduce to a
single point in special cases. The points of each
set are metameric. They constitute metameric do-
mains. If matching to C‘) is Maxwellian N-chro-
matic/C’? with color-matching matrix y,
Met,[C?] will consist of the points common to
N (n— 1)-dimensional linear manifolds (or hy-
perplanes) in 7-space. Each such hyperplane is
defined as the set of points whose coordinates P,
satisfy one of the matching relations expressed by
Eq. 1(5.3.4). There is, however, the constraint
injected that only points are included with all

non-negative coordinates, that is, points corre-
sponding to real stimuli. The test stimuli in these
matches can then be specified adequately for many
purposes in color analysis by the N-stimulus vec-
tor associated with C“. This is the assumption
made in interpreting most experimental work.
Matching to C“) may, however, be Maxwellian
N-chromatic/C“? in the weaker sense only. One
or more of the matching relations are then nonlin-
ear and generate nonlinear subspaces in T-space.
A mathematical model for this case is given in
Section 5.3.6.
In the present context, there is no real distinc-
tion between the roles played by test and com-
parison stimuli. There is equally a possibility that
the domain of points in C-space that match T?
may be Maxwellian N’-chromatic/T“?. The di-
mensionalities N and N’ are not necessarily the
same. Even if they are the same, the color-match-
ing matrices will not necessarily be equivalent.
This is the kind of situation that can arise if the
comparison stimulus is imaged on the foveal ret-
ina where trichromacy operates, whereas the test
stimulus is located in a dichromatic area of the
extrafoveal retina.
The metameric domains associated respec-
tively with T°) and C exhibit an important
simplifying property, if the principle of asymmet-
ric transitivity for test stimuli (see Section 5.2.2) is
valid for the matching procedure. It then follows
that every point in the domain about TJ? in
T-space correlates (matches) with every point in
the domain about C? in C-space, and con-
versely. This follows from the principle of asym-
metric transitivity whether or not matching is
Maxwellian in either sense. However if, in addi-
tion, for comparison stimulus C@? Maxwellian
N-chromacy/C“? does hold, it holds also for
every other point of Met [T“]. Furthermore, all
the test stimuli in the T-space domain Met,[C?]
then satisfy a set of N matching relations with the
same N color-matching functions y,(A;) and the
same N-stimulus vector Y*?. A similar result ap-
plies with the roles of test and comparison stimuli
interchanged. In what follows, the principle of
asymmetric transitivity for test stimuli will be
assumed without further comment.
For any particular asymmetric matching pro-
cedure, Maxwellian N-chromacy/C may hold for
a group of comparison stimuli C for each of
which the matching relations can be expressed in
terms of a common color-matching matrix p‘),
but with various values of the corresponding asso-
ciated N-stimulus vector ¥{“). The group will be
Maxwell’s Method of Color Matching 303
enlarged by including in it any comparison
stimulus whose spectral composition is the limit
of a sequence of comparison stimuli all satisfying
the conditions just stated. Physical continuity
considerations ensure that for each such addi-
tional comparison stimulus, the matching rela-
tions will apply with y‘*) as color-matching
matrix and an associated N-stimulus vector ¥{“)
that is the limit of the associated vectors of the
defining sequence. The values of ¥{*) can be
plotted in an N-dimensional (e.g., trichromatic)
space to generate a set J‘4) which is associated
specifically with the color-matching matrix p“.
To each point of the set J{*) will correspond a
metameric domain in 7-space. The aggregate of
all these domains defines, in 7-space, the total
range R& of test stimuli that match some com-
parison stimulus ¥{“) in accordance with the
color-matching matrix y~‘*), Thus, these test
stimuli satisfy Eq. 2(5.3.4), and ¥& is a member
of the set J‘*). The range R¢&) may include all
real test stimuli. It will then occupy the whole
region of 7-space for which every point has all
non-negative coordinates. In that event, the ret-
inal area, on which the test stimulus is imaged,
will have a property that is the nearest approach,
in a given specifically asymmetric matching pro-
cedure, to full N-chromacy as defined for quasi-
symmetric matching with the complete color-
match criterion in Section 5.3.3. Generally how-
ever, the R¢) range will occupy only a part of the
region of T-space representing real test stimuli.
T-space may well contain other similarly defined
ranges R?), R&, and so on, for different non-
equivalent color-matching matrices y7), py, and
the like.
An interesting example is presented by matches
made with comparison stimuli under two separate
conditions:
(a) The comparison stimuli are of such low
power levels that they produce no change, by
bleaching, in the spectral absorptance function of
the visual pigments in the receptors.
(b) The comparison stimuli are of such high
levels that the pigments are bleached to densities
where the following “dilute solution” approxima-
tion to their spectral absorptance function holds:
1 —e Fe) = cha(A)
where a(A) is the absorption coefficient and c, h
are constants.
304 VISUAL EQUIVALENCE AND VISUAL MATCHING
Matches under condition (a) may be expected
to yield a range R&), and under condition (b), a
range RY). If the Conditions I, II, and III for
Maxwellian matches are also satisfied for com-
parison stimuli of intermediate power levels, rea-
sons of physical continuity indicate that the
color-matching matrix ‘*) must change gradu-
ally to the nonequivalent matrix y‘*). The
metameric domain Met,(C,) for each such inter-
mediate comparison stimulus C, must belong to
some intermediate range R{) and may in fact be
the sole constituent of this range (see Section
8.2.6 for specific data and interpretations of ef-
fects of pigment bleaching).
The geometrical picture of what is happening
in T-space in cases such as the foregoing is rather
simple. If Maxwellian N-chromacy holds, each
hyperplane, defined by one of the matching
relations expressed by Eq. 1(5.3.4), will be
merely displaced parallel to itself for comparison
stimuli that generate different metameric domains
Met,(C;) all belonging to the same range, for
example, R&). On the other hand, if the compari-
son stimulus is changed to one generating a
metameric domain belonging to a different range,
RY), the orientation, and probably the distance
from the origin, of at least one of the N hyper-
planes will be altered. In an extreme case, it is
conceivable that with a progressive change in the
comparison stimulus, two of the hyperplanes may
coalesce, with a consequent change in the dimen-
sionality N of the matching process. However, the
artificial dichromacy and monochromacy in nor-
mally trichromatic areas of the retina, described
by Brindley (1953), probably doesn’t arise in this
way. Instead, a very great reduction occurs in the
sensitivity of one or two of the response mecha-
nisms on which matching depends.
The above discussion of different matching
ranges in 7-space applies equally to different
ranges RW) in C-space, each with its own
color-matching matrix $). However, although
for any asymmetric matching procedure there is a
one-to-one correlation of metameric domains in
the two spaces, this is not necessarily true of
ranges. Two domains belonging to the same range
in T-space may correspond to domains in differ-
ent ranges of C-space. As already observed, the
dimensionalities of Maxwellian matching in T-
and C-space may differ. However, in the special
case when all real test stimuli belong to a single
range in 7-space, and similarly, all comparison
stimuli belong to a single range in C-space, the
dimensionalities must be the same. Even in
this case, the color-matching matrices are not
necessarily equivalent. For example, consider an
asymmetric matching procedure that would be
quasi-symmetric, except that all the receptors in
the test area, but not in the comparison area, are
covered by an inert, selectively transmitting pig-
ment layer (possibly the macular pigment). Every
test stimulus can be matched by some comparison
stimulus, and conversely. However, the two
color-matching matrices operating, respectively,
in T-space and in C-space are clearly nonequiva-
lent.
Whether or not this is so, the matching results
of the asymmetric matching procedure are com-
pletely specified if the following conditions are
fulfilled:
(a) In each space T and C, a single range in-
cludes all real stimuli.
(b) The two color-matching matrices, y and 9,
are known.
(c) The necessarily one-to-one correlation is
known between the values of the N-stimulus vec-
tors ¥. and W,, belonging, respectively, to the
metameric domains that match. In other words,
the correlation is known between the points of
sets J; and Jc in their respective N-dimensional
spaces.
When the matching results reveal more than
one range in either or both spaces, a similar
correlation can be made between the associated
N-stimulus (test) and N’-stimulus (comparison)
vectors at match. However, in that case, this is
only possible for matches in which the test stimuli
belong to metameric domains generated by a
common matrix Y‘*, and the comparison stimuli
belong to metameric domains generated by a
common matrix ¢°”). In the further case, when
one of the color-matching matrices, for example,
Wy, changes progressively and continuously as
the comparison stimulus moves along some suita-
bly chosen path in C-space, the correlation pro-
duced by matching is confined to that between
single metameric domains in the respective spaces.
The set J; then reduces to a single point that
correlates with one point of Jc.
5.3.6. Maxwellian Matching in the Weaker
Sense
The T and C spaces are also useful in visualizing
a certain kind of deviation from linearity that
may appear in the results of some asymmetric

matching procedures and that illustrates Maxwel-
lian N-chromacy in the weaker sense (see Section
5.3.2). Consider first a quasi-symmetric matching
procedure for which Maxwellian N-chromacy
(with all three Conditions I’, II’, and III’ satisfied)
holds throughout for both variable test stimulus
and variable comparison stimulus, and with al-
ways the same color-matching matrix VW. The
matching relations for matching a real test stimu-
lus T,; with a real comparison stimulus C; may be
written as follows:
Q ¥(Ai)T; a V,(T) = Vo
i=1
>, ¥(Ai)C; ci) ¥,(C) =. ie APR
i=]
[1(5.3.6)]
with ¥%. = VW; at match. The restriction to real
stimuli may now be removed. Solutions to the
above equations may be admitted to which one or
more of the components 7; or C; become negative.
At the same time, the associated N-stimulus vec-
tors YW. and WV; will be allowed to assume any
values, not only those for which at least one real
test or comparison stimulus satisfies the equa-
tions. For this enlarged system, in each space the
metameric domains corresponding to different
N-stimulus vectors will all be nonoverlapping (as
before). However, collectively, they will now fill
the whole space. If ¥. = V;, the test and com-
parison stimuli, T and C, will be deemed to
match, setting up in this way the correlation
between test and comparison metameric domains,
denoted by Met7;(C) and Met (7), respectively.
Suppose now that 7-space and C-space are
each filled completely with a jelly-like solid in
which are embedded the hyperplanes concerned
in defining the correlated domains representing
matches. Suppose further that independently in
each space, the jelly is elastically and continu-
ously deformed in any desired fashion, subject to
the condition that the jelly at the origin remain
stationary. All the N hyperplanes defining
metameric domains, and hence, the domains
themselves, are assumed to remain embedded and
to move with the jelly, the coordinate axes, of
course, remaining fixed. At every stage and at the
completion of the deformation, there will be a
one-to-one correspondence between the initial and
displaced positions of every jelly-point, that is,
Maxwell’s Method of Color Matching 305
one moving with the jelly. Thus, in T-space, the
initial and final coordinates, T;l and 7)I respec-
tively, will be related by the following equations:
Teele eT = I) = tor
[2(5.3.6)]
where the functions w;(7’) are all single-valued
continuous functions of their arguments. No two
different sets of 7; values yield the same T, val-
ues, and if 7’ is at the origin, so is 7;. Similarly
for the initial and final coordinates, C; and C/ of
a jelly-point in C-space, the following equations
are valid:
C;= ¥(Cy, CoB FOS) = y;(C’)
[3(5.3.6) |
where the functions y;(C’) satisfy similar condi-
tions to the functions w,(7T’). The solutions T/
and C; of the sets of the N matching relations
DYQ)e(T) = YT) = %,
[4(5.3.6)]
and
2 ¥(A;) @;(C’) ay Vie) = V7;, J=I1toN
[5(5.3.6)]
define, respectively, the points of the metameric
domains, {Met,(C)}’ and {Met-(T)}’, in the
distorted system. These domains are correlated as
matching pairs, if Yo = V,.
The restriction to real stimuli may be restored
by accepting only solutions of Eq. 4(5.3.6) and
5(5.3.6) such that all the 7/ and C; values are
non-negative. These solutions could then repre-
sent the results of an asymmetric matching proce-
dure that is nonlinear and would not conform to
Condition III’ for Maxwellian matching. Also,
not every real comparison stimulus will be
matched by some real test stimulus. On the other
hand, it can be shown that Conditions I’ and II’
for Maxwellian matching in the weaker sense
would hold for matching of test stimuli to almost
all fixed (real) comparison stimuli, matched by
some real test stimulus. The exceptions arise in
306 VISUAL EQUIVALENCE AND VISUAL MATCHING
the same way as those discussed in Section 5.3.5
and are dealt with similarly. Maxwellian matching
in the weaker sense obviously applies equally to
matching variable comparison stimuli to fixed test
stimuli.
5.4 PRECISION OF COLOR MATCHING
FOR NORMAL TRICHROMATS
5.4.1 MacAdam Ellipses
Visual sensitivity to small color differences is
the essential factor determining the precision of
color matching. Early data on color-difference
thresholds provide indirect evidence on how this
precision will vary with the color of the matching
field and with other factors that affect sensitivity.
The first systematic studies of matching precision
in different parts of tristimulus space were made
by MacAdam (1942). The work was confined to
matches in which the variable test and fixed com-
parison stimuli were constrained always to have
constant luminance (15 millilambert or ~ 48 cd -
m ~), the same throughout this investigation.
MacAdam used an ingeniously designed col-
orimeter. His observer (PGN) turned a single
control knob to vary the color of one half of the
2° visual field along a straight line in the CIE
1931 (x, y)-chromaticity diagram, while the lumi-
nance of that field was automatically held con-
stant. The 2° matching field was surrounded by a
42° diameter surround field having a chromaticity
similar to that of CIE source C and a luminance
of 7.5 millilambert (~ 24 cd-m 7). Both the
variable and the fixed stimulus were mixtures of
the same set of red, green, and blue primaries so
that for an ideal match, the two stimuli had the
same spectral radiant power distribution. Re-
peated color matches were made between the
variable stimulus and a fixed stimulus of the same
luminance occupying the second half of the visual
field. The chromaticity of the fixed stimulus was
located on the given straight line in the CIE
chromaticity diagram. The standard deviations of
the settings were determined, a setting being
specified by the distance in the chromaticity dia-
gram between the variable and the fixed chro-
maticity points. MacAdam’s auxiliary experi-
ments on just-noticeable color differences in-
dicated that for the same observer these are ap-
proximately proportional to the corresponding
standard deviations of color matching. The just-
noticeable difference was found to be about three
Table 1(5.4.1) Example of Results Obtained by
MacAdam’s Observer (PGN) in a
Color-Matching Experiment
Direction of Line
through Standard Deviation
Xg = 0.305, yp = 0.323 of Color Matching
Ay/Ax As
52159 +0.00101
1.03 + 0.00203
1.28 + 0.00228
1.76 + 0.00213
0.27 + 0.00095
—0.84 + 0.00085
2.09 + 0.00215
times as large as the corresponding standard devi-
ation.
For the variable stimulus, several straight lines
intersecting in a common fixed chromaticity were
used in turn. The resulting standard deviations of
color matching were plotted as distances As along
these lines, measured from the central fixed chro-
maticity point (x9, Yo). The end points fell closely
on an ellipse whose size, shape, and orientation
changed when the whole procedure was repeated
with a new fixed chromaticity point. Table 1(5.4.1)
gives an example of the experimental results ob-
tained by MacAdam’s observer for a chromaticity
represented by the point x, = 0.305, yy = 0.323.
These results are plotted in Figure 1(5.4.1), which
also shows the best fitting ellipse. A total of 25
ellipses was determined for different locations of
the fixed chromaticity (x9, yo). Figure 2(5.4.1)
shows all these ellipses in the CIE 1931 (x, y)-
chromaticity diagram but drawn to ten times their
correct linear dimensions in relation to the coor-
dinate scales. The lengths a and b of the major
and minor semi-axes of the ellipses and the angle
of inclination 6 of the major axis to the axis of the
x-coordinates are given in Columns 3 to 5 of
Table 2(5.4.1). An illustration of the ellipse
parameters (Xo, Y), a, b, and @ is added to the
table.
Each of the 25 ellipses representing the known
standard deviations of color matching can be
expressed by the equation
2 2
81 (dx) + 21) dx dy + gy (dy) zy
[1(5.4.1)]
 Precision of Color Matching for Normal Trichromats 307

07-525

0.324

OSsi2i2

Oy 20

0. 3°@2 0.304 x O70 Ore 0.308

Fig. 1(5.4.1). Example of experimental results, obtained by MacAdam’s observer PGN,
showing how standard deviations of matching a fixed comparison chromaticity (x9, yo) along
different directions in the CIE 1931 (x, y)-chromaticity diagram lead to a MacAdam (1942)
ellipse.

where dx is the difference of the x-coordinates of where

the center (x9, yo) of the ellipse and any point on
6<90° when g,,<0
the ellipse; dy is the difference of the y-coordi-
nates of the same pair of points; and g)), 215, 209 6>90° when g,,>0
are constants for each ellipse. The values of the
g;,s can be computed from the lengths a and b of l
the major and minor semi-axes of each ellipse and 5 = 822 t 812 Cot 8
a
the angle of inclination 6 of the major axis to the
axis of the x-coordinates (MacAdam, 1943). The
_ = $11 — 812 Cot 6
be
following relations exist:

The values of the g;,’s differ from one ellipse

Eo eating.
kera co. 22 serps
Liao 6 to another in a systematic manner, and it is
possible by a process of interpolation to plot
smooth contour diagrams showing the variations
gis (+. dl sinfcos@ [2(5.4.1)] of these coefficients as functions of the chro-
a maticity of the central color of the ellipses
(MacAdam, 1942). In deriving smooth contour
Be AOR So3 ESE epee:
eee toe 6 lines, MacAdam had to depart somewhat from
the computed coefficients for the 25 ellipses, but
by allowable amounts in relation to the 15%
Sometimes, the inverse relations are needed. They uncertainties associated with the observed stan-
are as follows: dard deviations of color matching. With the
limited number (25) of observed ellipses and the
complicated dependence of the coefficients on
2212 chromaticity, construction of the contour dia-
tan26 = ee 3(5.4.1
811 ‘$12 ( )| grams involves some arbitrary element.
 308 VISUAL EQUIVALENCE AND VISUAL MATCHING

0.4 0.6
x

Fig. 2(5.4.1). MacAdam (1942) ellipses (observer PGN) plotted in CIE 1931 (x, y)-chro-
maticity diagram. The axes of the plotted ellipses are 10 times their actual lengths.

The distribution of repeated color matches distribution belonging to the chromaticity (x9, Yo)
around a color center in tristimulus space has is uniquely determined by the standard deviations
been studied in a slightly different way by Silber- (o,., 0,) and the correlation coefficient p of the x
stein and MacAdam (1945). If it is assumed that and y values in repeated matches. These coeffi-
color matches distribute normally about a color cients vary from chromaticity to chromaticity and
center in tristimulus space, it can be shown that are directly related to the metric coefficients g;,
loci of constant standard deviation of color as follows:
‘matching are ellipsoids. In the special case of
colors of equal luminance, the loci reduce to
l
ellipses when plotted in the CIE 1931 (x, y)-chro- §11 — o2(1 - p?)
maticity diagram. Silberstein and MacAdam
analyzed the observations made by MacAdam’s
observer PGN (MacAdam, 1942) and concluded §12
Ut0,9,(
GRE aa
1 — p*) [4(5.4.1)]
that, in the main, these data are consistent with a
normal (Gaussian) frequency distribution of the
color matches about the fixed comparison chro- l
S2222 a, —
o7(1 ae
p?)
maticity (X9, Yo). The particular normal frequency
Table 2(5.4.1) MacAdam Ellipses (Observer PGN) Observed and Calculated on
the Basis of a Normal Distribution of Color Matches about a Color Center
(Silberstein and MacAdam, 1945)

Color Center Observed Calculated

Xo Yo 10°b 10°a 1025

0.160 0.057 0.35 0.94 0.30

0.187 0.118 0.55 ie 0.44
0.253 0.125 0.50 2.49 0.49
0.150 0.680 25 | 9.09 ous
0.131 0.521 2.0 ] 4.67 2.10

0.212 0.550 2.3 1 5.63 2.30

0.258 0.450 2.0 4.54 2.08
0.152 0.365 1-9 I 3.81 1.86
0.280 0.385 Wee) 4.26 1.46
0.380 0.498 LZ 4.23 132

0.160 0.200 0.95 1 2.08 0.94

0.228 0.250 0.90 3.09 0.82
0.305 0.323 0.90 DES) 0.68
0.385 0.393 1.6 3.70 1.48
0.472 0.399 CIA | 1.30

opal 0.350 2.56 1.27

0.475 0.300 2.89 0.99
0.510 0.236 29.5 2.40 1.15
0.596 0.283 13.0 2.49 T5
0.344 0.284 60.0 2.24 0.97

0.390 0.237 47.0 2.43 0.98

0.441 0.198 34.5 pamg) 0.90
0.278 0.223 51,9 2.34 0.61
0.300 0.163 54.0 St 0.60
0.365 0.153 40.0 4.12 0.90
Note. Values of the observed ellipses were read off graphs of MacAdam’s 1942 paper.

Illustration of Ellipse Parameters.

Xo, Vo = coordinates of center

a = major semiaxis
b = minor semiaxis
6 = angle in degrees

309
310 VISUAL EQUIVALENCE AND VISUAL MATCHING
The calculation of the standard deviations (9,, 9,)
and the correlation coefficient p from observed
chromaticity coordinates x;, y, (i= 1,2,...,n)
whose mean values are (X, yo), 18 carried out by
means of the following formulae:
oy = var (x) == (4 — x6)"
0} = var(y) = n— |
l
yibran Van
i
cov(x, y) = man p>(x; — Xo)(% — Yo)
] l
Xo = 7 LX Vnm een: (i = 1,2,. ,n)
2_ 2
cov'(x,
y) 5(5.4.1
° var(x) var(y) DO Au
The notations var (x), var(y), cov(x, y) stand for
the variances and the covariance of x and y.
The agreement between the MacAdam ellipses
derived by the original method and those calcu-
lated on the assumption of a normal distribution
is good. Table 2(5.4.1) compares the two sets of
ellipses. The basic assumption that color matches
distribute normally about a given color center has
been checked experimentally by Brown (1952)
and found to agree well with the experimental
data.
If, in Eq. 1(5.4.1) which defines a MacAdam
ellipse, the expression on the left is put equal to a
quantity ds in place of unity, it can be regarded as
the distance between two points with coordinates
(x, y) and (x + dx, y + dy) in a nonlinear space
for which ds is the so-called line element:
2 2
(ds) cst (dx) + 281) (dx dy) + g) (dy)
[6(5.4.1)]
In using MacAdam’s approximate relation, we
should put ds = 3 to define the so-called dis-
crimination ellipse. The distance ds =3 from
the center to any given point of the ellipse
then represents a just-perceptible chromaticity
difference under the observing conditions used by
MacAdam. In practical applications of MacA-
dam’s ellipses, ds = 3 has been used only infre-
quently. Almost always, ds = 1 has been used
making the standard deviation of color matches
equal to the unit of perceived chromaticity dif-
ference, sometimes referred to as the MacAdam
unit of color difference.
From the properties of line elements, which
are more fully treated in Section 8.4, it follows
that, given the coefficients g,,, 217, 29) as func-
tions of x and y, a surface in an ordinary Euclidean
three-dimensional space can be defined on which
all the stimuli of constant luminance considered
by MacAdam plot as points. If the distance apart
in the Euclidean space of any two points on the
surface has a certain small constant value, the
corresponding stimuli are just-perceptibly differ-
ent. Also, on this surface, the MacAdam ellipses
plot as circles of identical size, so that the surface
may be described as a uniform-chromaticity-scale
surface. The complicated involuted form of this
surface as derived from the MacAdam line ele-
ment can be seen from a photograph [Figure
5(8.4.4)] of a model of the surface, constructed by
MacAdam (1944).
The CIE 1976 (u*, v*) diagram (Section 3.3.9)
is a plane diagram that may be used instead of
the true uniform-chromaticity-scale surface de-
rived from MacAdam’s data. Transformed to this
diagram, MacAdam’s ellipses approach circles of
equal size in some instances, but there remain
considerable discrepancies as may be seen in Fig-
ure 3(5.4.1).
Farnsworth (1958) has developed a nonlinear
transformation of the CIE 1931 (x, y)-chromatic-
ity diagram in which the transformed MacAdam
ellipses are very close to circles of equal size. The
deviations from circles of equal size are in fact
never greater than 20% with respect to the radii,
and in most cases, are less than 10% [Figure
4(5.4.1)]. Because such a close approximation is
possible and the experimental uncertainties of the
MacAdam data are of the order of 10 to 20%, the
question arises whether, after all, a uniform-chro-
maticity-scale surface of zero Gaussian curvature
(Section 8.4.4) would not be entirely consistent
with the measurements. However, other experi-
mental data for different observers (Brown and
MacAdam, 1949; Brown, 1957; and Wyszecki and
Fielder, 1971; see Sections 5.4.2 and 5.4.3) strongly
indicate that a nonzero curvature occurs per-
sistently. It may also be noted that the actually
observed points in Figure 4(5.4.1) are in many
cases better fitted by ellipses than by the circles
shown.
It is possible to approximate the uniform chro-
maticity surface by a number of plane diagrams
each of which is designed to represent only a
 100

50

fe)
ye

-50

-100
COLORS

-150
-150 -100 -50 e) * 50 100 150 200

Fig. 3(5.4.1). MacAdam (1942) ellipses plotted in CIE 1976 (u*,v*)-diagram (Robertson,
1977).

560 ri al LOCus

eo EOE ES SS Nee 600

EGE ISOS

tho = Sone
Bie eBilewe shoo 2 --- 480

Fig. 4(5.4.1). Nonlinear transformation of CIE 1931 (x, y)-chromaticity diagram designed to
make MacAdam’s (1942) ellipses approximately equal to circles of identical size (from
Farnsworth, 1958).

311
ae VISUAL EQUIVALENCE AND VISUAL MATCHING

restricted area of the curved surface (MacAdam,
1943). Within such plane diagrams, a discrimina-
tion ellipse can be transformed into a circle of
unit radius (= 1 MacAdam unit of color dif-
ference), if x and y are plotted as oblique coordi-
nates with modified scale units. The angle w
between the axes of x and y to accomplish this, is
determined by the values of the metric coeffi-
cients applicable in the portion of the surface
under consideration, and is given by
The quantity As becomes the line element ds
which is constant if (x, y) and (x + Ax,x + Ay)
are separated by the standard deviation of color
matching. The example of Figure 5(5.4.1) refers to
the region of the chromaticity diagram that con-
tains the MacAdam ellipse having its center at
Xo = 0.305, yo = 0.323 [see Figure 1(5.4.1)]. In
this region, the metric coefficients are
= 102% 10°, ge, — 0 ee

COS w = aN a (7654.1) (Foun k bese AU

(211822) .
Thus
At the same time, the scale units for the oblique
x and y coordinates are changed by factors COs wW = —— = —0.679
h(g,,) ‘77, and h(g,,)_'/*, where h is a con- (102 X 53)
stant. By ordinary coordinate geometry, the dis-
tance As apart of points in the oblique coordinate and
diagram representing chromaticities (x, y) and
OX Oa Vee y)ris then w = 132.8°

prises Ax

h(gi1) i
: With h = 1, the displacements along the oblique
axes corresponding to Ax = 0.001 and to Ay =
0.001 are 0.001(102 X 107)'7", — 1 Oia
0.001(53 X 104)'/* = 0.73, respectively.
A A
se ee . weer W The oblique coordinate system applies within
h(g11) h( 83) the region over which any variations of the coeffi-
cients 21), 81>, 82) are confined to a quite small
Ay 3 range that may be taken as some 25%. Within this
a —1/2 restricted region, the (Euclidean) metrical proper-
h( 83) ties of plane geometry are applicable to the chro-
maticity points represented. Thus, the straight-line
or distance between two points in this plane is di-
rectly proportional to the smallest number of
df 2 just-noticeable chromaticity steps between the two
ae *[g1:(Ax) + 2gpAxdy + gy(Ay)’| chromaticities represented by the:two points. This
method of determining small color differences
[8(5.4.1)] measured in terms of MacAdam units has been

Ne NEN es
OS NSS GAGStas
Aaa
Not es Naan
0.520 CaN NGOTAO BS
0:13 0:2 01304 5X2 07306 0.308
Fig. 5(5.4.1). Example of small portion of CIE 1931 (x, y)-chromaticity diagram, centered at
Xq = 0.305, yo = 0.323, that has been transformed linearly such that the MacAdam (1942)
ellipse with center at (x9, yo) becomes a circle of radius As =1. The (x, y)-coordinates are
now measured along oblique axes intersecting at the angle w = 132.8°.
 Precision of Color Matching for Normal Trichromats 313

developed by Davidson and Hanlon (1955) to the tion

point where it can be used in practical color
control problems. ®(R,G, B)A(dR)A(dG)A (dB)
Several attempts have been made to derive
approximate but explicit functions for the g,,- = Kexp|—40(dR,
dG, dB)|
coefficients of MacAdam’s ellipses in terms of the
(x, y)-chromaticity coordinates. The most suc- <A (dR)A(dG)A
(dB)
cessful attempts have been those of Chickering
(1967; 1971) who, in developing further the work
[1(5.4.2)]
of Friele (1961; 1966) and MacAdam (1964a,b;
1965), derived two formulae with computer opti-
where
mized parameters, either one of which can be
used to calculate MacAdam’s ellipses with close Q(dR, dG, dB) =c,,(dR)° + 2c,, dR dG
approximation (see also Section 5.4.5).
+¢y)(dG) + 2c,, dG dB

5.4.2 Brown-—MacAdam Ellipsoids

+¢3;(dB)° + 2c, dB dR
The restriction to colors of equal luminance was and
removed in determinations of trichromatic color
matching by Brown and MacAdam (1949). They
used a method in which the observer could vary os |Cik eas
K 3
the amounts of all three mixture primaries in the (277)
variable stimulus, the fixed stimulus being a mix-
ture of similar primaries so that again, for an
The quantity |c,,| is the determinant of the coef-
ideal match, the spectral compositions of the
ficients c;, (with c;, = c,;). The coefficients c;,
matched stimuli were the same. The specially
are determinable from the variances and covari-
designed colorimeter employed a bipartite match-
ances of tristimulus values obtained in repeated
ing field of 2° diameter in a dark surround,
color matches with the given color center. The
viewed monocularly with the natural pupil. Each
variances are computed from the following for-
primary was provided by a tungsten-filament lamp
mulae:
with suitable color filters. Results for 38 fixed
comparison stimuli were reported for two ob-
servers (WRJB and DLM). The luminances of the
comparison stimuli, all save two of different chro-
var(R) = ——— 3(R, — Ro)”
maticities, ranged from approximately 0.3 to 10
footlamberts (~ 1 to 34 cd - m *). Separate mea-
surements for a blue, a green, and a white
var(G) = 3,(G,— GJ’
comparison stimulus were made of the standard
deviations of the test primaries in matches over a
wide range of luminance. The approximate CIE var(B) = — J (B, — By)”
1931 (x, y)-chromaticities of the primaries used
were: Red = (0.699, 0.301); Blue = (0.152, 0.113); [2(5.4.2)]
Green = (0.176, 0.766) or (0.317, 0.624).
If R=(Ro + dR), G= (Gy + dG), B= (Bo
where R,, G;, B; are the tristimulus values of the
+ dB) are tristimulus values, with respect to any
ith match (i = 1,2,3,...,n) and Ro, Go, Bo are
given set of primaries, of a color that has been set
the tristimulus values of the central color assumed
up by an observer as a match with a central color
to be identical with the mean color match; that is,
Ro, Go, By, normal distribution of three-dimen-
sional color matches means that the chance, de-
noted by ®(R,G, B)A(dR)A(dG)A(dB), that
dR lies between dR and [dR + A(dR)], dG be-
tween dG and [dG + A(dG)], dB between dB and
[dB + A(dB)] is given by the exponential func- [3(5.4.2)]
314 VISUAL EQUIVALENCE AND VISUAL MATCHING

The formulae for the covariances are uniform cylinder with the line as axis) passing
through (Ro, Gp, By) in a fixed direction, the
standard deviations of the matches, measured as
cov(R,G) = —_. d(R, — Ro)(G; — Gy)
distances along the line, define two points on
either side of (Ro, Go, By), which coincide with
= cov(G,R) the intersections of the line with the surface of the
ellipsoid. The ellipsoid for (ds)* = 1 is sometimes
cov(G, B) = ——— B(G,~ G)(B,~ By) called the standard deviation ellipsoid, color-match-
ing ellipsoid, or color-discrimination ellipsoid. The
intercept of the plane of constant luminance
= cov (B,G) [4(5.4.2)] through (Roy, Go, By) with the ellipsoid is an el-
lipse as derived by MacAdam. If just-noticeable
cov(B, R) =~ us 2B, Bi iChro nis differences are about three times standard devi-
ations, the ellipsoid corresponding to (ds)* = 37,
would define the locus of colors just-noticeably
= cov(R,B) different from (Ro, Go, Bo).
An alternative concept may be used in assign-
Variances and covariances together form the so- ing a constant value of (ds). It may be stipulated
called variance—covariance matrix that the corresponding ellipsoid should contain,
on the average, a certain percentage a of any
var (R) cov(R,G) cov(R, B) random set of color matches. The necessary con-
M = | cov(G,R)_ var(G) cov(G,B) stant value for (ds)* = h(a), let us say, is ob-
cov(B,R) cov(B,G) var(B) tained from standard statistical tables noting that
the quadratic form Q(dR, dG, dB) is distributed
[5(5.4.2)] as x* (chi-square) with 3 degrees of freedom.
Thus, h(a) is the x* value for 3 degrees of free-
The elements of the matrix M_', inverse to M, dom, and is given in Table 1(5.4.2) for a series of
are the required coefficients c;,. values a.
The quadratic form Q(dR, dG, dB), or rather It is important to note that if the same concept
its square root plays the role of the line element is applied to fix the size of a chromaticity ellipse
ds. The equation for matches in the constant luminance plane or of
a critical segment for matches along a fixed line,
Q (dR, dG, dB) = const |= (ds)’| the constant value to be assumed for (ds)? = h(a)
is not the same as for determining the ellipsoid of
[6(5.4.2)]
defines an ellipsoid, and Eq. 1(5.4.2) shows that a Table 1(5.4.2) Values h(a) of the x?
random match (R,G,B) on the central color (Chi-Square) Distribution (from Hald, 1952) -
(Ro, Go, Bo) is more likely to lie in any assigned
space element [A(dR)A(dG)A(dB)] within the Probability h(a) h(a) h(a)
ellipsoid, than in any assigned element of in percent (1 degree (2degrees (3 degrees
ro of freedom) of freedom) of freedom)
the same volume lying outside. The surface of the
ellipsoid is the locus of “threshold” colors, sep- 1.0 0.000157 0.0201 0.115
arating colors that are difficult to distinguish or 5.0 0.00393 0.103 0.352
cannot be distinguished from (Ro, Gp, By) and 10.0 0.0158 0.211 - 0.584
colors that can be distinguished easily. Different 20.0 0.0642 0.446 1.000
30.0 0.148 0.713 1.42
choices of the constant (ds)? generate ellipsoids 40.0 0.275 1.02 1.87
corresponding to “threshold criteria” of different 50.0 0.455 1.39 2:37
degrees of severity. To agree with the standard 60.0 0.708 1.83 2.95
70.0 1.07 2.41 3.67
deviation criterion used by MacAdam in deriving 80.0 1.64 ASs 4.64
his chromaticity ellipses, the constant (ds)* must 90.0 rae A 4.61 6.25
be taken as unity. If, as in MacAdam’s experi- 95.0 3.84 309) 7.81
ment, attention is confined to matches made on a 99.0 6.63 9.21 Li.3
be BB 12.1 15.2 Liat
line (or strictly, to matches made in a very thin
 Precision of Color Matching for Normal Trichromats 315

equal a in the three-dimensional case. The ap- where, for example, 0x,/0x; = a,,;. We note that
propriate values of h(a) for two and one-dimen- Co) = C19, C32 = C93, aNd C)3 = C3; this reduces
sional cases are also listed in Table 1(5.4.2). the expression for c}, to
If, for example, a is taken as 95%, the corre-
sponding ellipsoid is obtained with h(a) = 7.81, yRe A ed Z
and the corresponding ellipse with h(a) = 5.99. Chy = AYCy, F 244147)
Cyq F 5, Co2 + 2471 43)Cr3
Conversely, when the same constant value for 2
(ds)* is used, irrespective of the number of di- FA31C33 + 24)1431C13
mensions, the corresponding values of a for el-
lipsoid, ellipse, and linear range are different. For Similar expressions are obtained for the other
example, with (ds)? = h(a) = 1 corresponding to coefficients cj. The transformed line element is
standard deviation loci in the three, two, or one- then
dimensional cases, the fractions a are, respec-
tively, 20% (ellipsoid), 40% (ellipse), and 68%
(critical linear range). (ds) =c',(dX) + 2ci, dX dY
To convert the expression for the line element,
Q(dR, dG, dB) = (ds)? given in Eq. 1(5.4.2) in +¢5,(dY) +25, dYdZ
terms of some set of instrumental primaries, to
the expression in terms of the standard CIE tri- tHe,3( GZ)a
/ 2
eve, vd ed Xi
f

stimulus values X, Y, Z, use may be made of the

following transformation of the coefficients c;, to [10(5.4.2)]
new coefficients cj:
3 A further transformation of the line element,
Ox; 0x of considerable value in applications, replaces
Cit = > ik ax’ = [7(5.4.2)]
i,k=1 J X, Y, Z by the two chromaticity coordinates x, y,
together with the luminance Y. Equation 7(5.4.2)
Here x,, Xj, X3 represent the original tristimulus is now used in the form
values R, G, B, whereas x}, x5, x; denote the tri-
stimulus values X, Y, Z. The two sets of tristimu-
lus values are related to each other by the linear
transformation

x, = 5 A; ;X; [8(5.4.2)]
with x; — xX, X%> = ys.%3 — Y, x, = X, X=
Y, x3 = Z, to determine the new coefficients c’;.
or in explicit form : ;
The relations connecting (xj’, x3, x3) and
R=a,,X + ayY + a,3Z (X}, X32, X3) are
Ga, (Xa Pd F [9(5642)| PP 2IE

ae — “1 ee = a
1 oes p) y

B= a3,X+ a3,Y + 4332

where the a;,’s are given constants derived from suey=. ”

Caen,
ast

the calibration of the colorimeter. Application of

Eq. 7(5.4.2) gives [thensieee9cisrn) edbate4
ae oy. be 0X Oi leacsl, OX 1 Ox 2 wae 0x, 1 Ox 3
. 1 Qx1 OX; ¢ Ox} Ox; a Ox; OX} [11(5.4.2)]
OX» Ox, OX OX 0X7 0X; The rather complicated explicit expressions for c;;
eon ee c
OK OX, ee eX, “sb in terms of c/, (i, / = 1 to 3) will not be written
out. They are obtainable by exactly the same
0x3 0X, 0x3 0X, q 0x3 0X3 procedure as used for deriving c/, in terms of c,,
+ C3] 2 os

Ox; OX; oe Ox xd sh date OX; (i,/= 1 to 3). The transformed line element is
316 VISUAL EQUIVALENCE AND VISUAL MATCHING

then Brown and MacAdam (1949) give transforma-

(ds) = cl, (dx) + 2c, dx dy
5 (dy) Th Cy. Gy a).
+c%,(dY) + 2c’ dx dY
[12(5.4.2)]
For ds = const, this equation defines an ellipsoid
in the x, y, Y space.
To obtain ellipsoids of comparable dimensions
in the directions of the three variables, one fur-
ther change from the variable Y to a new one,
(1/k) log, Y, has advantages. In each term of Eq.
13(5.4.2) containing dY, the latter is replaced by
dY/kY, and at the same time the factor kY is
introduced so that there is no net effect. This
modification is written out as follows:
tion equations that convert directly the coeffi-
cients c;, of Eq. 1(5.4.2) to the coefficients g;, of
Eq. 16(5.4.2) without going first to c;, and then to
cj. This reduces the chance rejection errors in
computing two successive transformations.
When dY, and thus, d/, is zero, Eq. 16(5.4.2)
reduces to the equation of the chromaticity ellipse
in the plane of constant luminance. Thus,
the coefficients g)), 215, 2, that Brown and
MacAdam derived are directly comparable with
those reported by MacAdam (1942) for his ob-
server PGN which have been described earlier.
On the other hand, if variations in luminance
only are considered and chromaticity is kept the
same (dx = dy=0), the corresponding radius
vector of the ellipsoid is closely related to the
Weber fraction AY/Y. When ds = 1 (standard
deviation criterion for the line element), the radius
in question, d/, satisfies

(ds) = cl, (dx) +2c%, dx dy + c% (dy) (ds) = 1=2,,(aDeotG4en

ey
y Cia,
kY dy kY is ik?¥?| =|
dy \’ The quantity d/ in this case represents the stan-
dard deviation of / = 0.2 1log,,. Y or approximately
(0.2/2.3) (oy/Y), where oy is the standard devi-
+26}kY dx [14(5.4.2)] ation of Y in luminance matching. Thus, o,/Y =
11.53/(g33)'/7. The quantity AY in the Weber
fraction is the just-noticeable difference of lumi-
However, dY/kY equals, approximately, nance that has been found experimentally to equal
about 30). Thus, on this assumption, we’ have
dlog,. Y 2.3
(“Ben )= (2) atogio¥

“r= —s_[186.42)]
(233)
Thus, if we put k = 5 X 2.3 (a suitable scaling
factor) and introduce the new variable / =
The results obtained by Brown and MacAdam
(1/5) logj) Y in place of Y and new coefficients
(1949), expressed in terms of Eq. 16(5.4.2) are
g;, defined by given in Table 1(5.4.2) of the Appendix. This
table contains the metric coefficients g,, for (ds)?
21 Cab Sioa 612) Bn
= CH = const = 1.0 with the corrections suggested by
Brown (1952). Also given are the cross sections of
S74 Goa Ca (KY )°c435 813 — kYci3 the ellipsoids for A/ = 0; these are compared with
corresponding ellipses of observer PGN deduced
[15(5.4.2)] from the smoothed summary of PGN’s observa-
tions of MacAdam (1943).
the final form of the equation of the ellipsoid is The cross sections of all ellipsoids (number 33
given by omitted because of its low luminance) obtained
by Brown and MacAdam for observer WRJB are
illustrated in Figure 1(5.4.2). An overall resemb-
(ds) = const = gi,(dx) steed iam OXHay
lance is noted between these ellipses and the
2 earlier data for observer PGN [MacAdam ellipses
+8y(dy)° + 285; dy dl shown in Figure 2(5.4.1)]. There is also an overall
resemblance between the ellipses for observer
+g33(dl)” + 2g), dx dl [16(5.4.2)] DLM and those for PGN and WRJB as indicated
 Precision of Color Matching for Normal Trichromats
0.8
0.6
500 9
0.4
0.2
e O
3899 5
0.4
Fig. 1(5.4.2). Cross sections of Brown—MacAdam
plotted in CIE 1931 (x, y)-chromaticity diagram.
by the data compiled in Table (5.4.2). Howeve1,
when we inspect the data more critically for a
given color center, the ellipses of different ob-
servers disagree with one another markedly in
perhaps half the cases. Rather large discrepancies
are observable, particularly in the ratio of the
major and minor axes of the ellipse. The varia-
tions in orientation also appear significant. A
discussion of the differences noted between el-
lipses obtained by different observers is given in
Section 5.4.4.
In later measurements, Brown (1957) em-
ployed 12 observers to obtain mean discrimina-
aig
CROSS SECTIONS OF 1949 BROWN-MacADAM
ELLIPSOIDS FOR OBSERVER WRJB.
(10 TIMES ENLARGED)
0.6 0.8
x
(1949) ellipsoids for observer WRJB
tion ellipsoids at 22 color centers extending over a
wide range of chromaticities with luminances be-
tween 3 and 6 footlamberts (~ 10.3 to 20.6 cd -
m ~*). Brown used the MacAdam binocular col-
orimeter (Section 5.15) with a 10° diameter
matching field and a large near-white surround
[x = 0.357, y = 0.344, Y = 2.6 footlamberts (~ 9
cd-m*)]. The specifications of the ellipsoids
representing straight means or weighted means of
the results of the 12 observers are reproduced in
Table I1(5.4.2). Figure 2(5.4.2) shows the cross
sections of the ellipsoids (those based on weighted
means) by the plane AY = 0. These ellipses, al-
318 VISUAL EQUIVALENCE AND VISUAL MATCHING
0.4
Fig. 2(5.4.2). Cross sections of mean ellipsoids determined by 12 observers (Brown, 1957).
though generally similar to those shown in Fig-
ures 2(5.4.1) and 1(5.4.2), have orientations that
appear to vary less regularly with the chromaticity
than in the results of a single observer.
Straight-average results for each ellipsoid were
obtained by averaging the elements of the vari-
ance—covariance matrix for each observer at each
color and then inverting the mean variance—
covariance matrix to obtain average metric coeffi-
cients g,, (it is incorrect to average directly the
g;,’s of the individual observers).
-Brown justified his use of weighted instead of
straight averages by arguing that whenever an
observer’s results correspond to an ellipsoid of
exceptionally large size, which occurs particularly
0.6 0.8
x
for the less skilled and more erratic observers, the
effect on the shape of the mean ellipsoid is un-
duly great. The weighting factors, designed to
reduce this distortion of the mean, were derived
from the actual experimental results and should
provide mean ellipsoids of more representative
shape. (For details of the weighting method, see
Brown, 1957.) Brown’s paper also brings out the
pronounced learning effect in the results obtained
by an observer in protracted investigations of this
kind.
The effects of field size, chromatic surround-
ings, and luminance level on color-matching el-
lipsoids have also been studied by Brown (1951;
1952a, b). His results on luminance level obtained
 Precision of Color Matching for Normal Trichromats
for two observers using the MacAdam binocular
colorimeter with a two-degree matching field and
dark surround, indicate that the precision of chro-
maticity matching remains fairly constant for a
considerable range of luminance above about 3
and up to 30 cd-m ~. Below the level of 3
cd-m °, the elliptical cross sections of the el-
lipsoids rapidly become larger and are rotated to
point with their major axes towards the “blue
corner” of the chromaticity diagram. This result
correlates with color discrimination which one
expects of an anomalous observer of the tritan
type (Hartridge, 1947; Judd et al., 1950).
In general, large fields (10°) allow better preci-
sion of color matching than fields covering only
the foveal (2°) region (Brown, 1952b). Little im-
provement in precision is expected for fields larger
than 12 degrees. This result is in agreement with
observations by other investigators (Judd, 1930b;
Lobanova and Rautian, 1949). In general, the
color-matching ellipsoids obtained for large-field
and small-field viewing are similar in orientation
and shape; only their sizes differ. However, in
some cases, slight but significant changes of
orientation do occur when the field size is changed.
This phenomenon has not been fully explored.
As a rule, the color of the field surrounding
the matching field has a quite noticeable effect on
the precision of color matching when the match-
ing field is small (Brown, 1952b). The observer
usually finds it more difficult to detect small color
differences in the presence of a surrounding field
differing in chromaticity. Best precision is ob-
tained when the surround has the same chro-
maticity as the matching field. This agrees with
results reported by Schénfelder (1933). The orien-
tation and shape of the color-matching ellipsoids
(2° matching field) are also affected by the color
of the surround. In some cases, a rotation and
elongation of the major axis of the (x, y)-cross
section of the ellipsoid toward the chromaticity of
the surround has been observed.
The effect of the color of the surrounding field
is considerably less pronounced for large-field
than for small-field matching, but an exception
has been noted for matches in the red where the
color-matching ellipsoids are modified for both
field sizes.
5.4.3. Wyszecki-Fielder Ellipsoids
In more recent measurements of color-matching
ellipsoids made for three observers by Wyszecki
and Fielder (1971), Wyszecki’s 7-field binocular
319
colorimeter (Section 5.15) was used with all but
two of the juxtaposed 3° diameter hexagonal fields
replaced by a white surround (Figure 1(5.2.5)j) of
luminance 6 cd - m ? and of diameter 40° with
the bipartite field at its center. As in Brown’s
work with the MacAdam binocular colorimeter,
in each half-field was presented a mixture of the
same three primary stimuli, the comparison mix-
ture being fixed while the observer adjusted all
three components of the test mixture to obtain a
match. A feature of the work was that both eyes
were used and could wander over the field with
no strict fixation although, of course, central
vision was used for the match. These conditions,
shared with Brown’s binocular measurements, ap-
proximated most closely those of ordinary view-
ing. The data obtained determined the color-
matching ellipsoid for each of 28 comparison
stimuli all of luminance 12 cd - m? and of chro-
maticities covering the triangle in the diagram
defined by the chromaticities of the matching
primaries. The statistical methods for handling
the blocks of matches were as explained earlier in
Sections 5.4.1 and 5.4.2. The three-dimensional
space in which the ellipsoids were finaly ex-
pressed have the coordinates x, y, and /, where
(x, y) are CIE 1931 chromaticity coordinates and
/ is the log luminance variable / = 0.2log Y. In
addition, the cross section of each ellipsoid was
determined; that is, the ellipse of intersection of
each ellipsoid with the plane of constant lumi-
nance through its center. These cross sections
facilitated comparison with MacAdam’s and other
earlier data.
An example of an ellipsoid determined by
observer GW is shown in Figure 1(5.4.3). Ob-
server GW made 29 matches with the given com-
parison stimulus whose coordinates were x9 =
0.3251, y = 0.3325, and /) = 0.2156. These 29
matches result in 29 points (x,, y;, /;, with i= 1
to 29) which cluster about (X9, yo, /9). A com-
plete picture of their distribution in the three-
dimensional space is obtained by plotting them as
projections in the three coordinate planes (x, y),
(x,/), and (/, y). The color-matching ellipsoid
calculated on the basis of the 29 matches, with
(ds)? = const = 7.81, is also shown in its three
projections, which, of course, are ellipses in the
coordinate planes centered at (X9, Yo), (Xo; /o),
and (/, yo), respectively. As expected, the three
projections of the ellipsoid envelop the three cor-
responding sets of observed points. Figure 1(5.4.3)
also shows the cross section (dashed line) of the
ellipsoid. In this example, this cross section nearly
320. VISUAL EQUIVALENCE AND VISUAL MATCHING

eeu

0.210
0320 r 0.330

A
0.335

0.320 : 0.330 O.2I0 ’ 0.220

Fig. 1(5.4.3). Portion of (x, y,/)-space, in its three projections, showing the distribution of
points (solid dots) representing 29 color matches made by observer GW with fixed comparison
color (open circle). The ellipses (solid lines) are projections of the resulting color-matching
ellipsoid, with (ds)? = 7.81, into the three respective coordinate planes. The ellipse (dashed
line) in the (x, y)-plane is the cross section through the center of the ellipsoid when the
luminance is constant, that is, /= 0.2158.

coincides with the projection of the ellipsoid in
the (x, y)-plane.
Table 1(5.4.3) gives the complete specifications
of the ellipsoids and their cross sections for all
three observers. The cross sections of the el-
lipsoids are illustrated in Figure 2(5.4.3).
It should be noted that these specifications are
derived on the assumption that the constant (ds)?
in the ellipsoid equation has the value 7.81, ap-
propriate to an ellipsoid embracing on the aver-
age 95% of any random set of matches. Constant
standard deviation ellipsoids on the other hand
correspond to (ds)? = 1, and are therefore
smaller.
5.4.4 Repeatability of Color-Matching
Ellipsoids for the Same Observer
The work of Wyszecki and Fielder (1971) de-
scribed in Section 5.4.3 also included tests of the
repeatability of the color-matching ellipsoids for
each of the three observers (GF, AR, and GW) on
different occasions.
Table (5.4.4) gives several examples of the
results of these tests. Each observer went through
two or three different sessions to produce two or
three sets of color matches with each of the
comparison colors numbered 9, 11, 13, and 15. In
each session, each observer made up to 30
matches; the exact number is recorded in the last
column of Table I(5.4.4). The observational data
of each session were then analyzed separately,
leading to a color-matching ellipsoid that can be
directly compared with one obtained from data of
another session but for the same observer and
comparison color.
Table 1(5.4.4) also gives the mean results for
each comparison color matched by the same ob-
server in the different sessions. The mean results
(Rows M in the table) were obtained by lumping
together all the color matches recorded in the
different sessions into a single set, as if they were
obtained in a single session. This leads, for each
comparison color and observer, to a mean color-
matching ellipsoid, but one whose g;, coefficients
are not necessarily the arithmetic mean of the g,,

+
T
w
--
z2
 02

Fig. 2(5.4.3). Cross sections of Wyszecki-Fielder (1971) ellipsoids plotted in CIE 1931
(x, y)-chromaticity diagram for observers GF, AR, and GW. The open circle (S) represents the
chromaticity of the white surround. The triangle (dashed lines) represents the chromaticity
gamut provided by the primary stimuli (R, G, B).

321
a ee
ZO re)
ome) vO ZO 210 vO 20
vO

£0

re)
“Sl “(p'P'S)T sso suoNIes
JO J°P[etY-PPezsKAA
(TL6T) sprosdypapayojd
Ul AID T€61
) ‘x4 ApIOQeUIOIYI-(
WeIseIpP
10} pajyvadal
SjasJO 10[0I Soyd}VUI
apvul
Aq Sidarasqo
‘yD YYV pue
AAD)UO JUDIOIJIP ‘SUOISEIIO
JNq ISIA.IBYJO [BITJUSPI SUTAIBSGO “SUOTJIPUOD
 Precision of Color Matching for Normal Trichromats
coefficients of the ellipsoids corresponding to the
individual sessions. The mean color-matching el-
lipsoids given in Table I(5.4.4) are identical to
those given in Table 1(5.4.3) for the comparison
colors 9, 11, 13, and 15.
Figure 1(5.4.4) illustrates the results of the
repeatability tests for the cross sections of the
ellipsoids. In several cases, considerable dis-
crepancies are noted between the ellipses ob-
tained by the same observer at different occasions
(separated by weeks or months) under otherwise
identical observing conditions. The discrepancies
are remarkable because statistical estimates of the
standard deviations of the major and minor
semi-axes a and b and the orientation @ of the
ellipses suggest much smaller discrepancies.
The statistical estimates of the standard devia-
tions o(a), o(b), and o(@) of the ellipse parame-
ters a, b, and 6, respectively, can be calculated for
a given set of n color matches by means of the
following formulae:
a
RO) =D:
BG
[1(5.4.4)]
< orientation, size, and shape of the ellipses. No
oe Snr GaD
where @ is measured in radians. For n = 30, the
standard deviations of a and b are approximately
+0.13a and +0.13b, respectively. This means
that for 30 color matches, one would expect an
uncertainty of +13% in the major and minor
semi-axes of the cross section ellipse. The uncer-
tainty of orientation @ of the same ellipse is a
function of the shape of the ellipse, characterized
by the ratio a/b; the greater the ratio, the smaller
o(@).
The obvious disagreement between the statisti-
cal estimates of the uncertainties of the ellipse
parameters and those indicated by the repeatabil-
ity tests casts doubts on the applicability of classi-
cal statistical techniques to visual experimental
data of the kind discussed here. However, no
quantitative explanations have been offered as to
possible reasons for the occurrence of the dis-
agreement.
323
5.4.5 Intercomparison of Color-Matching
Ellipses for Different Observers
In their critical intercomparison of color-match-
ing ellipses, Wyszecki and Fielder (1971) included
the following seven observers:
(a) Observer PGN from the work of MacAdam
(1942) described in Section 5.4.1. The relevant
data are given in Table 2(5.4.1) and illustrated in
Figure 2(5.4.1).
(b) Observers WRJB and DLM from the work
of Brown and MacAdam (1949) described in Sec-
tion 5.4.2. The relevant data are given in Table
I(5.4.2) and illustrated in Figure 1(5.4.2).
(c) Mean of 12 observers, taken as one observer
and denoted by 12B, from the work of Brown
(1957) described in Section 5.4.2. The relevant
data are given in Table II(5.4.2) and illustrated in
Figure 2(5.4.2).
(d) Observers GF, AR, and GW from the work
of Wyszecki and Fielder (1971) described in Sec-
tion 5.4.3. The relevant data are given in Table
1(5.4.3) and illustrated in Figure 1(5.4.3).
There is an overall resemblance between the
seven sets of ellipses illustrated in the figures
listed above. However, when one confines the
intercomparison to a particular location in the
chromaticity diagram, the ellipses of different ob-
servers are quite often not in close agreement.
Rather, larger discrepancies are noted in the
doubt, some of these discrepancies arise from the
inconsistencies inherent in the ellipses of the indi-
vidual observer as discussed in Section 5.4.4 and
illustrated in Figure 1(5.4.4). The remainder of
the discrepancies, particularly those which appear
to be systematic, must be attributed to other
causes, such as differences in the underlying vis-
ual mechanisms that characterize the different
observers, and differences in the observing condi-
tions used in the different studies.
The ellipses of Observer 12B (weighted means
of 12 observers) shown in Figure 2(5.4.2) stand
out as a somewhat odd set. Although generally
similar to those of the other sets of ellipses, their
orientations change less regularly with chromatic-
ity than those of a single observer. This suggests
that averaging the data of several observers in a
way it would normally be done and, in fact, was
done by Brown (1957), does not necessarily yield
a set of ellipses that could be treated as repre-
sentative of an actual observer intermediate to a
group of observers.
324 VISUAL EQUIVALENCE AND VISUAL MATCHING

@(deg),
Obs.
GF

20 40 60 80
@(deg), Obs. GW

-4.0

GF
Obs.
A,
109i
=5.0

Obs.
a/b,
GF Obs.
a/b,
AR

a/b, Obs. GW a/b, Obs. GW

Fig. 1(5.4.5). Diagrams showing the correlation of the orientations (8, in degrees), sizes
(log ;94, with A = wab), and shapes (a/b) between corresponding color-matching ellipses of
observers (GF,GW), and of observers (AR, GW) (from Wyszecki-Fielder, 1971).

Although inspection of the different sets of
ellipses drawn in the chromaticity diagram reveals
some general trends of agreement and disagree-
ment between them, a more informative inter-
comparison may be made by resorting to ap-
propriate correlation diagrams for the orienta-
tions, shapes, and sizes of the corresponding el-
lipses of a given pair of sets. Two ellipses from
different sets are called corresponding ellipses if
they have a common center; that is, they refer to
the same comparison color.
Correlation diagrams are readily plotted for
the sets of ellipses of observers GF, AR, and GW,
because the set of comparison colors involved is
identical for each of the three observers. Simi-
larly, correlation diagrams can be plotted for the
sets of ellipses of observers WRJB and DLM.
Figures 1(5.4.5) and 2(5.4.5) show the results ob-
tained when correlating the ellipses of GW with
those of GF and AR, and DLM with WRJB,
respectively.
The agreement between the orientation (0) of
the corresponding ellipses for each of the given
pairs of observers is generally quite good. The
points distribute well along the 45°-line represent-
ing perfect correlation. However, there is a
suggestion that most of the ellipses of GF are less
steeply sloped than the corresponding ellipses of
GW and AR. This is indicated by the fact that
the points in the 6-diagram for GF and GW are
 Precision of Color Matching for Normal Trichromats 325

points along the 45°-lines in the (log), A)-dia-

grams.
However, the agreement between the shape,
characterized by the (a/b)-ratio, of the corre-
sponding ellipses is not as good. In fact, few if
any inferences can be made in this case. At least
in part, the poor correlation may be attributed to
the fact that the ratios of the semi-axes are inter-
compared, each one of which, that is a and b,
being subject to a considerable uncertainty.
Correlation diagrams of the kind shown in
Figures 1(5.4.5) and 2(5.4.5) cannot be drawn
readily for observers whose ellipses do not refer
to the same comparison colors. However, by using
a formula derived by Chickering (1967), a link
can be established between such observers.
@ e 7 Chickering optimized the parameters of his
a rs formula such that the ellipse he calculated by
> > °”” e
2 750 - Le e
means of his formula for comparison chromatici-
ro)
2 m Le ®
3 e ties (Xo, Yo) identical to the 25 chosen by
a
ro)
7,
7 MacAdam (see Section 5.4.1), agree very closely
oe e /’e
° with the ellipses of observer PGN in orientation,
— 7
6.07
6, size, A, and shape, a/b. The good agreement is
7
illustrated in Figure 3(5.4.5). Because of the good
-6.0 -5.0 -4.0
109,94, Obs. DLM agreement, it is reasonable to assume that the
calculated ellipses, denoted by (PGN)c, are
@a 5 e Zz
iA representative of the observed (PGN)-ellipses.
4 of Furthermore, it is reasonable to assume that
=! 7 °
a
Oo ES eo. pice (PGN) -ellipses similarly calculated for other
Sb chromaticity centers are representative of ellipses
— e %
= e e ig °
that observer PGN would have obtained if asked
2 Sins to do color matches at these other centers. With
e* vas
7 these assumptions, it is then possible to calculate
I 2 3 4 5 6 (PGN),-ellipses at centers exactly coinciding with
a/b, Obs. DLM
those at which other observers actually had ob-
Fig. 2(5.4.5). Diagrams showing the correlation tained color-matching ellipses. Appropriate corre-
of the orientations (0, in degrees), sizes (log 19 A, lation diagrams for observer (PGN) = (PGN)c
with A = mab), and shapes (a/b) between corre- paired with any of the other observers are then as
sponding color-matching ellipses of observers readily drawn as before.
(DLM, WRJB) (from Wyszecki—Fielder, 1971). Figure 3(5.4.5) also shows the correlation dia-
grams of the observer pair [(PGN)., GW], while
Figure 4(5.4.5) shows similar diagrams of the
below the 45°-line, whereas for AR and GW, the observer pairs [(PGN)c, WRJB] and [(PGN)<,
points distribute fairly closely along the 45°-line, 12B]. All these diagrams reveal important sys-
as shown at the top of Figure 1(5.4.5). A similar, tematic discrepancies between the ellipses of PGN
though in opposite direction, displacement of [i.e., (PGN)-] and those of GW, WRJB, and 12B.
points is indicated in the 6-diagram for WRJB The most apparent discrepancy occurs in the
and DLM. In this case, almost all of the ellipses orientation @. In the @-correlation diagrams, the
of DLM are less steeply sloped than those of points distribute along a line of much lesser slope
WRIJB. than the 45°-line. In referring back to Figures
The agreement between the size (A = mab) of 2(5.4.1), 1(5.4.2), and 2(5.4.2), and 2(5.4.3), the
the corresponding ellipses for each of the given discrepancy can be traced to the ellipses in the
pairs of observers is generally also quite good. “purple” and “red’ region of the chromaticity
This is again indicated by the distribution of the diagram. In these chromaticity regions, the
326 VISUAL EQUIVALENCE AND VISUAL MATCHING

100 7
/
ve 7
7 7
£100 zee 80 . fe
z a e ee Z Sf °
1) ez 2
a 80 i pes s ie 60 ay & l

$s: . PPsis =
3 a oe rn e ® Kis

© 60 ae © 40;-, ° 7
33 bb, Le, aa
Pc 3
3S /
a
be=) eo @ °
e IK ,
20 7 ) ee
20 40 60 80 100 (@) 20 40 60 80 100
6@(deg), Obs. PGN @(deg),Obs. GW
-4.0 aoe
— Ad >
2 ny oO
© a
O. os ~—- -40
ont e a
FS as =
S 5.0 of 3
<x e Ke e °
° a =
$ 7 2
-5.0

-5.0 —4. O
10919 A, Obs. GW

Obs.(PGN),
a/b, Obs.(PGN),
a/b,

2 3 4 2 3 4 5
a/b,Ob s.PGN a/b, Obs. GW
Fig. 3(5.4.5). Diagrams showing the correlation of the orientations (0, in degrees), sizes
(log 94, with A = mab), and shapes (a/b) between corresponding color-matching ellipses of
observers [PGN,(PGN)-], and between observers [GW,(PGN) ] (from Wyszecki-Fielder,
1971).

(PGN)-ellipses are systematically more
sloped than those of observers GW, WRJB, and
12B. Because observer GW correlates well with
(GF and AR, and observer WRJB correlates well
with DLM, it is concluded that PGN shows simi-
lar discrepancies with GF, AR, and DLM.
Systematic discrepancies between PGN and
GW are also indicated for the sizes of correspond-
ing ellipses. The GW-ellipses range in size over a
full log unit, approximately; those for PGN range
over only 0.4 log units. The discrepancies between
PGN and WRJB, and between PGN and 12B are
not as pronounced.
steeply Detailed explanations for the noted dis-
crepancies between corresponding ellipses of dif-
ferent observers cannot be offered. However, not
all the discrepancies should be attributed to dif-
ferences in the visual mechanisms of the different
_ observers. No doubt, differences in the observing
conditions and methodology used in the different
studies also play an important part. Perhaps the
most important parameters of the observing con-
ditions are the size of the matching field, lumi-
nance of the comparison stimuli, and chromatic-
ity and luminance of the surround. All three
parameters changed from one study to the next,
 Precision of Color Matching for Normal Trichromats 327

“do é
loo 7° 100 ie A
79 = rg
i
Zz
< 80 /. /~e& 280 e /4 as a
a eg . z @ ods e e
=: ) © Co Rial @
# 60 ee a 60-e0e 7
2 e @ 7 = es 7
o e ; 0&7 ja S re

o> 40r , se ° 40 A
= * “d 7 . - ° yi
520 fap bd S20 ° 7,
7 @ 7
7 7
fe) dh Ch ee ee oe eee eere ees
160 y, fe) 20 40 60 80 100 160 740 20 40 60 80 100 120
a 6(deg), Obs. WRUB 7/ 9(deg), Obs. 128
' > oO

Obs.
(PGN),
A,
10dig

510 -4.0
lOgig A, Obs.12B

(PGN)¢
Obs.
a/b,

2 3 4 5
a/b, Obs. WRJB a/b, Obs. 12B

Fig. 4(5.4.5). Diagrams showing the correlation of orientations (0, in degrees), sizes (log 1) A,
with A = wab), and shapes (a /b) between corresponding color-matching ellipses of observers
[WRJB,(PGN)], and between observers [12B, (PGN)] (from Wyszecki—Fielder, 1971).

at least to some extent. The effect of a difference
in field size is perhaps the one most easily identi-
fied in some of the correlation diagrams. Other
studies (for example, Judd,
1952b; see also Section 5.4.2) have shown that a
larger field size tends to improve the observer’s
precision of making color matches, thus reducing
the sizes of the color-matching ellipses. Such an
effect is indicated when intercomparing the el-
lipses of PGN and GW. Whereas PGN used a 2°
field, GW used a 3° X 6° field. This field size
increase, no doubt, contributes to the apparent
average decrease by a factor of 0.4 in ellipse size
for GW relative to the ellipse sizes of PGN.
The methodology used in the studies by Brown
and MacAdam (1949), Brown (1957), and
Wyszecki and Fielder (1971) are similar to one
another. All use color-matching techniques in-
volving three independent controls of the three
given primary stimuli to produce the desired color
1930b; and Brown, mixture that matches the color of a given com-
parison stimulus. By contrast, the methodology
used in MacAdam’s (1942) study involved a
single-knob control to vary the test color along a
given straight line in the chromaticity diagram,
keeping the luminance automatically constant.
MacAdam’s one-variable method is much less
demanding on the observer than the three-varia-
ble method used by the other investigators. How-
ever, whether the noted discrepancies, particularly
evident in the orientations of some of the
PGN-ellipses relative to those of all other ob-
servers, are caused by the different methods used,
cannot be established with certainty.
 AL

328 VISUAL EQUIVALENCE AND VISUAL MATCHING

5.4.6 Propagation of Random Errors in Variances

Colorimetric Transformations
The transformation of the color-matching el-
lipsoids from the instrumental primaries to p=)
another set was explained in Section 5.4.2, in
which it was tacitly assumed that the coefficients (Uz, » 7a
of Eq. 9(5.4.2) were known precisely. The fact var (U, ) c* > n= 1
p=1
that these coefficients are not free from error is,
in general, unimportant in this application be- [2(5.4.6) ]
cause of the approximate character of the original
ellipsoids. Different considerations apply when Cowariances
tristimulus values obtained in one primary system
are transformed into another system, as there will S (5 ee)
cov (7,7;) =
be random errors associated with the original
p=!
tristimulus values, and other random errors asso-
ciated with the transformation coefficients, them-
selves obtained by color matching or by spectro-
photometry. All these errors will propagate into cov (U,U;) = s (
the new tristimulus values. By application of the eee
general theory of propagation of random errors,
the resulting uncertainty in the final tristimulus (kK#l) [3(5.4.6)]
values can be determined. For fuller explanation
of the general theory and proof of the formulae to The means, variances, and covariances of the
follow, see, for example, Scarborough (1962). measured vy; and derived U, quantities are related
In general, a set of known functions U, = according to the general theory of error propaga-
U,(y;) with k = 1,2,3,--- of the independent tion in the following way:
variables v,(i = 1,2,3,---) is presupposed, and Mean Values
the uncertainty in the U,’s has to be calculated
from the random errors of the v,’s. Although the U, = U,(7;) [4(5.4.6)]
variables v; are supposed independent, the errors
in determining them are not necessarily unrelated. Variances
For example, if v,,v,,v,; are the R,G, B tr-
stimulus values of a match, errors in R may be
correlated with errors in G or B. It is necessary, var(Uy) = (SE) var)
; I
therefore, to define a combined single measure-
ment of the whole set of variables v; in which a dU, \ | dU,
single value of each variable is obtained, and, in +25) |= (4 cov (7;, »;)
Ov,
the case of any subgroup of the variables p,, oy
whose random errors may conceivably be corre- (Gi, 7=1,2,3 ---;i As) [5(5.4.6)]
lated, all these single values must be obtained in
the same multiple observation. From such a com- Covwariances
bined single measurement of the v,’s, a single set
of values of the U,’s is computed, using the
known functional relations between U,’s and p,’s.
i
From n combined single measurements, the mean
values, variances, and covariances of the v,’s and
U,’s are derivable from the standard statistical +2ij aUOD, |(aUdv;
expressions. The suffix p identifies values refer-

“(3 ](52)
ring to the respective combined single measure-
dU, dU;
ments. cov (7,7; )
Ov; Ov;
Mean Values

Vp. Zs a U,.
(Gi, j= 1,2,3,---
31 Fy)
sabe [1(5.4.6)]
(k, f= 1,2;3; 3k ~D: [664.60]
 Precision of Color Matching for Normal Trichromats 329

The important colorimetric application is to obtained for the variances and covariances of the
the transformation from measured tristimulus val- b., and subsequently for the variances and covari-
ues R, G, B to derived values R’, G’, B’ in a new ances of the tristimulus values R’, G’, B’. Even so,
primary system the formulae are extremely elaborate, and the
principle of their derivation having been estab-
(Tee fat elon iy ee ie lished, their explicit statement will be omitted.
A much simpler case develops if one assumes
G’ = bu Rt boG + b,8.17(5,4.6)] that the variables (Rp, Gr, Br), (Rg, Gg, Be),
(Rg, Gp, Bz’), or the coefficients b,, are known
precisely, that is, if they are free from random
errors. Such an assumption may be made if the
Here the coefficients b,, are the elements of a
relations between the old and new primaries are
matrix that is the inverse of the matrix
based on very precise spectrophotometry or if
Rr Ro Rp
they are specifically assigned on some theoretical
or conventional basis. Then the only variables
AmitGi, GYeiG. mts(5.4.6)| with random errors are the visually determined
Bre Bratt, tristimulus values R, G, B. The formulae for the
variances and covariances of R’, G’, B’ then re-
whose respective columns are the (R, G, B)-tri- duce to the following:
stimulus values of unit quantities of the primaries
var (R’) = bj, var(R) + bi, var (G)
R’, G’, B’ with respect to the R, G, B primary sys-
tem.
+ b?; var (B) + b,,b,;,
cov (RG)
It will be supposed initially that these sets of
tristimulus values are, in fact, determined by di- +b,,b,,cov(
RB) + b,.b,3cov(GB)
rect matching. Thus, the coefficients b,, are all
known functions of the nine measured quantities
var (G’) = b3, var (R) + b5, var (G)
Rr, Gr, Br, Rg, Gg, Bg, and Rp, Gp, Bg; for
example,
+ b3, var (B) + b,,b,, cov( RG)
Es (By Go — Gp Bg)
bi |A | ’ [9(5.4.6)] +b,,b5,cov( RB) + b,,b,, cov (GB)

var (B’) = bj, var (R) + bj, var (G)

where |A |is the determinant of A.
For the colorimetric application, the variables + b3; var (B) + b3,b3)
cov (RG)
in the general formulae are identified as follows:
+b3,b3;cov(RB) + b3,b33 cov (GB)
U, = R’, vy,—R, v,= Rp,
U,= G’, ¥,—G, Vs = Gp, cov (R’G’) = b,, by, var (R) + b7b)
var (G)
U; = B’, v,—B, Yo = Br,
+ b,3b>;
var (B)
vz= Rg, Vio
— Rp
Vg = Go, Vin— Gp + (1b. + b,2b,)
cov (RG)
Vo = Bg, Vin = Bp
+ (1b; + b13b,)
cov (RB)
The general formulae then give the required vari-
ances and covariances of R’,G’, B’ in terms of + (b12by3 + b 13622)
cov (GB)
the experimentally determined variances and co- cov (R’B’) = b,,b3,
var (R) + 6,253)
var (G)
variances of R,G, B, Rp, Gr, Br, Ro, Go,
Bo, Rp, Gg’, Bp. Since the random errors in each + b,3b33
var (B)
of the groups of measured variables (R, G, B),
(Re, Gr’, Br), (Ro, Go's Bg), (Rp, Gp, Bz’) GD; Disats DioPay) COVAGRG,)
are uncorrelated with those of any other group,
the only covariances not equal to zero are those (Bj pag te Dy503;),
COV CRB)
between individual variables in any one group.
This simplifies somewhat the formulae that are es (b12b33 ot b 13532) cov (GB)
 330 VISUAL EQUIVALENCE AND VISUAL MATCHING
cov (B’G’) = b3,b5, var (R) + b32b>,
var (G)
+ b3,b>;
var (B)
+ (b31b2 + b32b,) cov (RG)
+ (b3,b5; + b33b>,)
cov (RB)
+ (b32b3 + b33b22) cov (GB)
[10(5.4.6)]
5.5 COLOR-MATCHING FUNCTIONS OF
NORMAL TRICHROMATS
The matching procedures usually used in the de-
termination of color-matching functions are those
illustrated in Figure 1(5.2.5) labeled (a), (b), and
(c). In the sections below, we have selected a few
examples of different sets of color-matching func-
tions obtained by different investigators for a
variety of observing conditions.
5.5.1 Two-Degree Data of Guild and Wright
The color-matching data of Guild (1931) and
Wright (1928-1929), together with the CIE 1924
luminous efficiency function V(A), form the ex-
perimental basis of the color-matching functions
of the CIE 1931 standard colorimetric observer.
The essential features of the work of Guild and
Wright have already been described in Section
3.3.3(a) dealing with the development of the CIE
1931 standard observer and thus they do not need
repeating here. The actual experimental data re-
ported by Guild and Wright are now of only
historical value and the reader interested in them
is referred to the original papers of Guild and
Wright. However, their combined spectral chro-
maticity data which ultimately emerged and
which, together with the CIE 1924 luminous ef-
ficiency function V(A), led to the definition of
the CIE 1931 standard observer in terms of the
color-matching functions 7(A), Z(A), b(A), are
still of some importance. The resulting color-
matching functions are given in Table I(3.3.3) and
illustrated in Figure 3(3.3.3). They refer to a
system of monochromatic primary stimuli of
wavelengths Ap = 700.0, A, = 546.1, and A, =
435.8 nm, whose units are chosen so that the
chromaticity coordinates r;, g;, b; of the equal-
energy stimulus (wavelength basis) are all equal;
that is, 7g = gp = bb; = 1/73.
5.5.2 Judd Modification
In reviewing the experimental data on which the
CIE 1924 luminous efficiency function V(A) is
based, Judd (1951) concluded that the CIE ap-
pears to have given weight to early measurements
which led to average values at short wavelengths
that must now be considered excessively low.
Judd proposed a modification of the CIE V(A)-
function for wavelengths below 460 nm.
This modification is shown graphically by the
continuous line in Figure 1(5.5.2) which gives the
logarithm of the ratio of Judd’s modified values
to the original CIE values. The same figure shows
the results of subsequent experimental studies of
the values of V(A) in the blue end of the spec-
trum (Ishak and Teele, 1955; Stiles, 1955). These
substantially confirm Judd’s modification, at least
for observers in an age group centered on 30.
If the Judd-modified V(A)-function had been
used instead of the CIE 1924 V(A)-function, the
CIE 1931 color-matching functions derived from
oO oO T
ce]
[V(A)/V(A)
CONST
LOG
+
27500
|
25000
!
22500
l
20000 cm!
| i |
400 450 500nm
WAVENUMBER/ WAVELENGTH
Fig. 1(5.5.2). Comparison of measurements of
V(A) in the blue end of the spectrum with the CIE
1924 standard V(XA)-curve and Judd’s modified
V(A)-curve (Stiles, 1955). Curve: Judd’s (1951)
V(A)-curve; open circles: Mean of direct-compari-
son measurements with 1/A = 22,500 cm ' for 28
observers, mean age 30.4 (Stiles, 1955); open tri-
angles: Mean flicker comparison measurements
with 1/A = 21,750 cm! for 104 observers, re-
duced to age 30 (Ishak and Teele, 1955). The
constant value added to the ordinate was chosen so
as to make log[V(A)/V(A) cp] + const. equal to
zero at 21,750 cm.
 Color-Matching Functions of Normal Trichromats 331

the Guild—Wright spectral chromaticities and the and z(A) to preserve the spectral chromaticities
V(A)-function would have been assigned the val- are less simple [see Section 3.3.3(a)]. Although,
ues shown in Table 1(5.5.2) which were computed for various reasons, mainly practical, the CIE has
by Judd (1951). The new y(A) is simply Judd’s decided to make no change in the standard
modified V(A)-function, but the changes in x(A) V(A)-function (or the CIE 1931 color-matching
functions), the Judd-modified data are of interest
in visual studies.
Table 1(5.5.2) Color-Matching Functions Vos (1978) has refined Judd’s modified color-
of CIE 1931 Standard Observer, Modified matching functions x’(A), y’(A), Z’(A) by:
by Judd (1951)
(a) making use of more precise computational
Color-Matching Functions procedures than those available to Judd in 1951;
Wavelength = ——————_______—_- (b) extending the data further into the far red
A (nm) x’(A) yA) 2'(X)
by making use of data consistent with the reversal
370 0.0008 0.0001 0.0046 phenomenon at long wavelengths, that is, Brind-
380 0.0045 0.0004 0.0224 ley’s isochromes described in Section 5.10;
390 0.0201 0.0015 0.0925
(c) reducing by 0.2 log units the values of Judd’s
400 0.0611 0.0045 0.2799 modified V(A)-function for wavelengths below
410 0.1267 0.0093 0.5835 410 nm, making them follow more closely the
420 0.2285 0.0175 1.0622 values obtained by Stiles (1955) shown in Figure
430 0.3081 0.0273 1.4526
440 0.3312 0.0379 1.6064 TS. e2y,
(d) smoothing slightly the CIE 1931 color-
450 0.2888 0.0468 1.4717 matching functions in the region from 380 to 400
460 0.2323 0.0600 1.2880
470 0.1745 0.0910 1.1133
nm and eliminating a minor concavity of the CIE
480 0.0920 0.1390 0.7552 1931 spectrum locus between 435 and 445 nm;
490 0.0318 0.2080 0.4461 and
(e) truncating the color-matching functions at
500 0.0048 0.3230 0.2644
510 0.0093 0.5030 0.1541 380 nm, because the CIE 1931 data below that
520 0.0636 0.7100 0.0763 wavelength are not based on experimental data
530 0.1668 0.8620 0.0412 but are only the result of numerical extrapola-
540 0.2926 0.9540 0.0200
tions.
550 0.4364 0.9950 0.0088
560 0.5970 0.9950 0.0039 Table 1(5.5.2), given in the Appendix, contains
570 0.7642 0.9520 0.0020
the color-matching functions x’(A), y’(A), Z’(A)
580 0.9159 0.8700 0.0016
590 1.0225 0.7570 0.0011 and corresponding chromaticity coordinates
x'(A), vy(A) derived by Vos (1978) for wave-
600 1.0544 0.6310 0.0007 lengths from 380 to 825 nm at 5-nm intervals.
610 0.9922 0.5030 0.0003
620 0.8432 0.3810 0.0002 The Judd (1951) modification of the CIE 1931
630 0.6327 0.2650 0.0001 standard colorimetric observer, or its refinement
640 0.4404 0.1750 0.0000 by Vos (1978), is generally considered an im-
650 0.2787 0.1070 0.0000
provement over the original 1931 data that are
660 0.1619 0.0610 0.0000 still the official data. However, from the apparent
670 0.0858 0.0320 0.0000 improvement that has been accomplished, it does
680 0.0459 0.0170 0.0000 not necessarily follow that an accurately de-
690 0.0222 0.0082 0.0000
termined photopic luminous efficiency function
700 0.0113 0.0041 0.0000 must be a linear combination of the color-match-
710 0.0057 0.0021 0.0000 ing functions. As already pointed out in Section
720 0.0028 0.0011 0.0000
730 0.0015 0.0005 0.0000
3.3.1, the color-matching properties of a typical
740 0.0005 0.0002 0.0000 normal trichromat are a direct consequence of
obedience to the trichromatic generalization (Sec-
750 0.0003 0.0001 0.0000
tion 3.2.1), and the brightness-matching property
760 0.0002 0.0001 0.0000
770 0.0001 0.0000 0.0000 of the same observer does not necessarily follow
it. In order to merge meaningfully the two visual
Totals 10.7533 10.7526 10.7531
properties into a single system, the additivity law
332 VISUAL EQUIVALENCE AND VISUAL MATCHING

for brightness must hold, according to which the better fit is obtained for the V(A)-function ob-
condition for a match in brightness, but not nec- tained by the flicker method, except in the wave-
essarily in color, of any two stimuli of spectral length range below 510 nm. Figure 2(5.5.2) il-
radiant power distribution {P,}, dA} and {P,, dA} lustrates the results of Sperling (1958) together
takes the form with results obtained by Stiles (1955) using the
color-matching functions of his 2° pilot group of
[PaVQ) dd = [Pa) dr 10 observers (see Section 5.5.3) and Judd’s mod-
ified CIE V(A)-function. If allowance is made for
[1(5.5.2)] the somewhat different primary systems used in
the studies of Stiles and Sperling as well as dif-
Direct measurements of the color-matching func- ferences in the normalization procedure, Stiles’
tions and of the luminous efficiency function (by data resemble those of Sperling involving the
the flicker and direct comparison methods) made flicker V(A)-function. Sperling’s data involving
by Sperling (1958) on six observers show that the the direct comparison V(A)-function, on the other
linear combination of the color-matching func- hand, take a quite different course.
tions does not adequately fit the V(A)-function A useful critical review of the basic photomet-
obtained by direct comparison; but a somewhat ric and colorimetric data governing the CIE 1931

Wavelength \ (nm)
400 450 500 5}0) 8) 600 (S19 112OL)
ls ames I a ate T aeaiesa

< LS
ipa [ Sperling (1958) i
SI (Flicker) \
pad ES _= Sperling (1958) i!
— ? . . 1
S / (Direct Comparison)
fies It L in
x :

ot
+ 1.0
«
ips O99 iz
al
2408 =
< “Stiles (1955)
= 10,7 r (Judd modification)

EI | Bre le | | | all i | a ae |
25000 20000 15000

Wavenumber m (cm7!)
Fig. 2(5.5.2). Ratio of an experimentally determined V(A)-function and a linear combination
of independently determined color-matching functions; that is, V(A)/[Lpyr(A)+Le6g(A)+
L3b(A)] where the coefficients Lz, L¢, Lz are the luminous efficiencies of unit amounts of the
primary stimuli R,G,B used in the color-matching experiment. Solid dots: mean data of six
observers obtained by Sperling (1958). Each of the six observers measured the V(A)-function
by the flicker method. The primary stimuli in the color-matching experiment were: R at
Ap = 650 nm, G at A, = 520 nm, B at A,= 460 nm, with Lp =V(AR) = 0.125, Lg =V(Ag)
= 0.649, L, =V(Ag)= 90.590. Crosses: mean data of the same six observers as above.
However, this time, each of the six observers measured the V(X )-function by the direct -com-
parison method, giving coefficients L p = V(A pz) = 0.122, Lg = V(AG¢) = 0.654, Ly =V(Ag)=
0.074. Open circles: the V(X)- Forction is Judd’s modification of the CIE 1924 standard
V(X )-function. The color-matching functions are the mean of the 10 observers of Stiles’ (1955)
2° pilot group, based on the primary stimuli R at A,=15,500 cm | (~ 654 nm), G at
se = 19,000 cm! (~526 nm), B at dp = 22,500 cm”! (~ 444 nm). The coefficients
rR—V(AR) = 9.137, Lg = V(AG) = 0.813, Lz = V(Ag) = 9.043, follow from Judd’s modified
a (V(A)-function. The ratio-curve has been normalized to have unity value at A = 18,000
em! (~ 556 nm).
 Color-Matching Functions of Normal Trichromats
standard observer has been given by Estévez
(1979).
5.5.3. Stiles Two-Degree Pilot Data
In the 1950s, Stiles conducted a major re-
examination of the CIE standard photometric
and colorimetric data for the average normal tri-
chromatic observer. His investigation was encour-
aged by the CIE Colorimetry Committee which
had expressed concern about the correctness of
the CIE 1931 standard observer data (CIE 1948,
1951) with regard to
(a) the luminous efficiency function V(A) =
y(A) in the ends of the spectrum, particularly the
blue end;
(b) the small discrepancies between the color
differences of certain titanium pigments, as ob-
served and as computed from the CIE 1931 data;
and
(c) their applicability to color matching in large
visual fields that are used in industrial color pro-
duction control.
d It was also noted, as already expressed by
Guild (1931), that the method of determining the
CIE 1931 color-matching functions by combining
measurements of the chromaticity coordinates for
monochromatic stimuli and a standard white, with
measurements of the CIE 1924 V(A)-function, is
not an ideal procedure when one considers that
color-matching functions are directly measureable
without introducing any heterochromatic bright-
ness matching (flicker or direct comparison).
Stiles’ investigation began with a pilot study
involving a group of 10 observers that made color
matches and heterochromatic brightness matching
in 2° and 10° visual bipartite fields on the spe-
cially designed colorimeter, known as the Stiles
(or NPL) trichromator (Stiles, 1955; see also Sec-
tion 5.15). The interim report by Stiles (1955;
with an appendix by Stiles and Burch) on the
results of the pilot study were considered by the
CIE Colorimetry Committee at the CIE 1955
Session at Zurich. There it was decided that the
discrepancies between the CIE 1931 standard ob-
server data and the 2° pilot data obtained by
Stiles, though significant from the experimental
point of view, were not large enough to warrant a
change in the standard data for practical col-
orimetry. On the other hand, the work on large-
field (10°) color matching was to continue with
the view of establishing a supplementary standard
—_ 333
observer for large-field viewing. In following the
direction recommended by the CIE Colorimetry
Committee, Stiles proceeded with his work by
concentrating on large-field color matching ex-
periments. These have been discussed in Section
3.3.3(b); further notes are given in Section 5.5.4.
Although no further work was done by Stiles
on 2° color-matching, his 2° pilot data provide a
valuable set of color-matching functions that are
obtained without appeal to any form of hetero-
chromatic brightness matching.
Table 1(5.5.3) gives the 2° pilot color-match-
ing data as obtained by Stiles and Burch (1955) as
functions of wavenumber m at intervals of 250
cm '. However, small changes have been made to
the original data by first applying to them correc-
tion factors given by Stiles and Burch (1959), and
then renormalizing them so that the newly tabu-
lated color-matching functions refer to primary
stimuli located precisely at mp = 15,500, me =
19,000, mz = 22,500 cm! instead of their origi-
nal locations at m’p = 15,418, mo = 18,997, m,
= 22,456 cm”'. The unit quantities of the three
primary stimuli are of equal radiance measured in
terms of [W-m 2: sr_'].
Table 1(5.5.3) in the Appendix gives inter-
polated 2° pilot color-matching data as functions
of wavelength 4 at intervals of 1 nm.
5.5.4 Stiles-Burch Ten-Degree Data
Stiles’ re-examination of the CIE standard photo-
metric and colorimetric data, already referred to
in Section 5.5.3 culminated in the 1958 final
report of the N.P.L. color-matching investigation
by Stiles and Burch (1959). These data, together
with those obtained by Speranskaya (1959),
formed the basis for the CIE 1964 supplementary
standard observer presented in Section 3.3.3(b).
Table 1(5.5.4) gives the numerical values of the
Stiles-Burch mean 10° color-matching functions
for 49 observers at wavenumbers m from 14,000
to 25,500 cm™! and intervals of 250 cm'. The
primary stimuli are located at mpg = 15,500, m¢
= 19,000, and m, = 22,500 cm~'. The retinal
illuminances (photopic trolands) provided by the
monochromatic test stimuli and their spectral
bandwidths are listed in Table 2(5.4.4) together
with other pertinent details concerning the
Stiles—Burch investigation.
Table 1(5.5.4) in the Appendix gives inter-
polated color-matching data as functions of wave-
length A at intervals of 5 nm.
334 VISUAL EQUIVALENCE AND VISUAL MATCHING

In an auxiliary experiment (Part III) con-
ducted by Stiles and Burch, the color-matching
functions at four wavenumbers in the deep red
were determined for nine observers in terms of
three instrumental primary stimuli at 14,250,
17,000, and 21,250 cm~'. After transforming the
measured data to the primary stimuli at mp =
15,500, mg = 19,000, and mz = 22,500 cm™' and
making appropriate interpolations at intervals of
500 cm ', the color-matching functions given in
Table 3(5.5.4) were obtained.
Data on the differences between color-match-
ing functions obtained by the different observers
that participated in the experiment will be dis-
cussed in Section 5.5.6.
Large-field color matches are subject to rod
intrusion (see also Section 5.6.2), the extent of
which depends on the spectral compositions of

Table 1(5.5.3) Mean Color-Matching Functions and Spectral Chromaticity Coordinates of Stiles
2° Pilot Group of 10 Observers (Stiles and Burch, 1955)”

Wavenumber
m Color-Matching Functions Chromaticity Coordinates
(cm) r(m) g(m) b(m) r(m) g(m)
13,000 0.22584(—3) —0.15247(— 5) 0.10263(—5) 1.0022 — 0.0068
13,250 0.5817 1( 3) — 0.42384(— 5) 0.16896(—5) 1.0044 — 0.0073
13,500 0.14968(— 2) —0.11768(—4) 0.26913(-5) 1.0061 — 0.0079
13,750 0.38554(—2) — 0.32707(— 4) 0.46069( — 5) 1.0073 — 0.0085
14,000 0.99201(—2) — 0.90802(— 4) 0.76490( — 5) 1.0085 — 0.0092
14,250 0.24514(— 1) =—0,22699(— 3) 0.14193(— 4) 1.0088 — 0.0093
14,500 0.61782(— 1) —0.58137(— 3) 0.31514(—4) 1.0090 — 0.0095
14,750 0.13860 —0.12144(—2) 0.65460(— 4) 1.0084 — 0.0088 -
15,000 0.29121 —0.21620(—2) 0.20189(— 3) 1.0068 — 0.0075
15,250 0.57162 — 0.26607(—2) 0.19995(—3) 1.0043 — 0.0047
15,500 0.10000(1) 0.00000 0.00000 1.0000 0.0000
15,750 0.15291(1) 0.12894(— 1) —0.31980(— 3) 0.9918 0.0084
16,000 0.21054(1) 0.46504(— 1) — 1.95522(3) 0.9788 0.0216
16,250 0.25837(1) 0.11027 —0.19698(— 2) 0.9598 0.0410
16,500 0.28388(1) 0.21447 — (2 108 — 2) 0.9306 0.0703
16,750 0.28447(1) 0.34935 Ay SPSS — 2) 0.8921 0.1096
17,000 0.26118(1) 0.50834 — 0.74860(— 2) 0.8391 0.1633
17250 0.22587(1) 0.65470 —0.10018¢— 1) 0.7780 0.2255
17,500 0.18272(1) 0.80895 =-0:12293¢-)) 0.6964 0.3083
17,750 0.13891(1) 0.89702 —0.14092(— 1) 0.6114 0.3948
18,000 0.10321(1) 0.99238 —0.14941(— 1) 0.5136 0.4938
18,250 0.70528 0.10450(1) —0.14461(— 1) 0.4063 0.6020
18,500 0.43766 0.10509(1) swt Ae PET eri y, 0.2965 0.7120
18,750 0.20801 0.10509(1) — 0.68923(— 2) 0.1661 0.8394
19,000 0.00000 0.10000(1) 0.00000 0.0000 1.0000
19,250 — 0.15687 0.90132 0.14714(—1) —0.2066 1.1873
19,500 — 0.23984 0.76091 0.37289(—1) —0.4295 1.3628
19,750 0.29321 0.62551 0.69418(—1) —0.7299 1.5571
20,000 — 0.29500 0.49060 0.10749 i BLES 1.6186
20,250 — 0.28935 0.38750 0.15779 1 00 1.5140
20,500 — 0.26819 0.30712 PARI) et 235 1.1719
20,750 — 0.24494 0.25529 0.32272 — 0.7354 0.7665
 Color-Matching Functions of Normal Trichromats 335

Table 1(5.5.3) — Continued

21,000 —0.22016 0.20892 0.45409 —0.4972 0.4718

21,250 —0.17931 0.15607 0.60188 —0.3099 0.2697
21,500 —0.13851 0.11116 0.66513 =0 2172 0.1743
21,750 —0.94407(— 1) 0.69402(— 1) 0.78943 =0.1235 0.0908
22,000 —0.57710(—1) 0.41276(—1) 0.82933 —0.0710 0.0508
22,250 —0.25781(—1) 0.17136(—1) 0.93143 —0.0279 0.0186
22,500 0.00000 0.00000 0.10000(1) —0.0000 0.0000
22,750 0.16196(—1) —0.80375(—2) 0.95813 0.0168 — 0.0083
23,000 0.32896(— 1) —0.16395(—1) 0.89094 0.0363 —0.0181
23,250 0.44063(—1) —0.19878(—1) 0.79823 0.0536 — 0.0242
23,500 0.52281(—1) —0.21437(—1) 0.69981 0.0716 —0.0293
23,750 0.52901(—1) —0.17656(—1) 0.56137 0.0887 — 0.0296
24,000 0.45690(— 1) —0.13934(— 1) 0.40841 0.1038 —0.0317
24,250 0.36885(—1) —0.10023(—1) 0.28893 0.1168 —0.0317
24,500 0.26022(—1) —0.54649(—2) 0.18538 0.1264 — 0.0265
24,750 0.17025(—1) —0.41916(—2) 0.11742 0.1307 — 0.0322
25,000 0.96264(—2) —0.21689(—2) 0.62371(—1) 0.1379 —0.0311
25,250 0.54459(—2) —0.12055(—2) 0.36362(—1) 0.1341 —0.0297
25,500 0.27211(—2) —0.64566(—3) 0.18043(—1) 0.1353 —0,0321
25,750 0.13596(—2) —0.34574(—3) 0.89530(—2) 0.1364 — 0.0347
26,000 0.67930(—3) —0.18510(—3) 0.44424(—2) 0.1376 —0.0375
26,250 0.33940(— 3) —0.99080(—4) 0.22042(—2) 0.1388 —0.0405

“The above data have been corrected in accordance with instructions given by Stiles and Burch (1959)
and renormalized to primary stimuli located precisely at mp=15,500, m,- =19,000, and mz = 22,500
cm!.
’The numbers in parentheses denote powers of 10. Thus, 0.22584(—3) reads 0.22584 10-7.

the two color stimuli to be matched in color and equation

their radiances in absolute radiometric terms. In
their investigation, Stiles and Burch (1959) con-
sidered the phenomenon in great detail and at-
tempted to account for it in their data. Color
matches were done in two separate runs, one
involving a “green” middle primary stimulus at
mg = 19,000 cm! (referred to as the “green
run’), the other involving a “yellow” middle
primary stimulus at my = 17,000 cm“! (referred
to as the “yellow run”). The yellow run was
introduced to reduce rod intrusion in the mea-
surements for m < 17,250 cm |.
The way rod intrusion comes about is prob-
ably as follows (Stiles, 1958; Stiles and Wyszecki,
1973). Suppose that, using for the middle primary
a “green” at 19,000 cm!, color matches are
made in a retina whose rod mechanism has in
some way been prevented from responding. A
particular match is presented by the (vector)
QO(m)Q(m) + R(m)R = G(m)G + B(m)B
[1(5.5.4)]
where Q(m) is the test stimulus of wavenumber
m, and R, G, B are the “red,” “green,” and “blue”
monochromatic primary stimuli located at mp,
mg, and mz. The quantities O(m), R(m), G(m),
and B(m) are the amounts of the test and primary
stimuli, measured in radiometric (energy) units
such as [W-m ~~ sr '], required to obtain a
color match between the two-component mixture
on the left-hand side of the equation (presented in
the upper half of the 10° bipartite visual field)
and the two-component mixture on the right-hand
side of the equation (presented in the lower half
of the visual field). The quantitative effect that
the two halves of the visual field would have on
336
"MA‘C' AA put ‘saj}euIps007D AVDOQVUIOIYD [e.ydedg ‘suOTIUNY SUTYDJVIA[-10JOD OT UBIFAL §(H'S'S)T FIQUL
pue ,_ Wd 0OS‘SZ 9} NOO'PT Woly SISquINUDAL AA JB SIDAIISAO GF (6S61) YOING-S2[NS JO Sajyeurps00+)
p 1 Wd OSZ JO sfeasojuy

JOQUINUDAR
ul suotoun,y BuTyo}eJA[-IO[OD Sd}PUIPIOOD AjyOTyeUTOIYD) Sd}VUIPIOOD “A\'C' AA
(,_w9) (uw) (w)s (m)q (ui )a (um )s (YG (Wyk (UW )yB (W)ed
000'rI 26000 10000— 00000 LLOO OS eeEL0081 70000 60£0'1 eve00— 7e00°0
OST 0700 cOOOO— 0000°0 1800 0— 82001 £000°0 eee rr0 0£00°0
00S‘r1 10900 S000 0— 00000 8L000— 9L00'1 £000°0 fcc) | erie 97000
OSL'vI LOPIO O1000— 00000 SLOOO0— £001 7000 0 SiO | fer cC00 0
9100°0
096c0
81000—
1000°0
0900'1
c900:0—
70000
LECO.02-=-.09C0
I000°SI
OS7'SI
98950
OcOO'0—
1000°0
Set00'1
9€000—
10000
ISIOT
O9I00—_
6000°0
00000
0000
0000°0
100S°SI
I0000'I
OSL‘SI I Secs vrI100 c0000—0 8766 ¥L00°0 cOO00— 66960 s1t00 LIODOO—
9£86
ILt00
60000—
OOLI
89100
yoooo0—
pet6
00L00
0yeo000—
000°91
OST7'9I 'C OOOL 01600 Sl000—0 6L96 97£0°0 S0000—0 ISL8 £6cl0 yro00—
00S‘91 0080'€ O06LI00— 9700 0 8576 0sS0°0
0— 8000 0 L108 £70c-0
= = 190G.0
OSL‘9I OO9T'€ OS0E0 6£000—
0 0£16 18800 I1000—0 I80L 0 L667 8L000—
000°L1 00CO'E 0 Och €S000— L1L8°0 v0ElO S1000— $609°0 10070 96000—
OS7LI 009°C 0209°00— $900 6S18°0 19810 0cOOO0— 96060 9505-0
es EIG'OS
00S‘LI OOVC'C 0620 8L000—
0 68P7L at} SES MS 2COO 0 vLov €$09°0 LClOO—
OSL‘LI OOSL'I 06280 68000— 6L99'0 TSSEEd VrUO=—O LITE EL69:0) OS
Vy
000°81 | 000¢€ 08960 96000— 9SLS'°0 98c¢r70 er000— 18€c0 SILLL-0
ss iGOPS
OS7'81 00060 06101 $6000—
0 tILy 0 =9ttS OSOO0—0 COLT 8Sr80 I9100—
00S°81 0 O19S ObSO'l c8000— 16ve0 09S9°0 ISOO0— 86010 96060 vrIOO—
OSL'8I 0 Ovrc O€fO'l €S000— 61610 0 ECB cvo00— 91500 8560 OOIOO—
000°61 0000°0 00001 00000 0000°0 0000'I 0000°0 0000°0'I 0000 0000°0
S761 O8610— 01760 LOL000—* Loc 1 S8vc +=cHIO0
0— c6v0 psco'l 8£70°0
05°61 OzzeO0— 0 OLE8 8Lc0°0
= O— ce6S I Ocrs cIsdood S6800—I €0¢0 €690°0
OSL6I
= OLOVO— O&eL0 OvS0°0 IILOT—I 68¢6 lo Acr Oy Evo. 89L6°0 69710
 L OSer0—
0$8L8°0
01600
09790
(LOC)
COcrs
6610
LO9T
COE
LO
000'02
===
0702 OrrrO— O1ES 0 OFF LD eco Ls L867 C vt790 66e10— LeeL 0 c90P 0
05°07 O€th0— Oly 0 0910 L0G 1S 9788 1 126.0 scclLO— vi9soO p19 0
OSL‘0Z Or6E0— 069¢ 0 OSTEO LPP eae v6Sc | eS80'l 89600— SL6£ 0 £669°0
(een
O0€6C
0O€PEO0—
OSL80—
Ocrrd
00017
OULO
bLvl
€C9T
LLO8
OLE
st
= OST'IZ O16TO— OLCC 0 09690 OLbSO— L9ICY 0 EUGI I COMO Ores 8910 VE88 0
6l€eO—
OO0PLO
P1601
pore
OC9T
91e00—
OS7TO—
O00S°IZ
c1£6
0SOOT
= OSL'IZ 9SS10— 6S01°0 Oves0 p86l0— ISel 0 Eveo0d OGCOO = L6S00 £096 0
0
0cc6
¥L90°0
0090
veOLO—=
09€0'l
601I00—
c1t00
L6L6
000°
007600—
OST'7
ss 97K0'0— 0Sc0'0 | 96000—s Ctr €¢C0'0 S8Z10:1 LYOUD= Icl00 0 9766
00S°7Z 0000°0 0000°0 0000'I 0000°0 0000°0 0000'1 0000°0 0000°0'I 0000
OSL‘CZ 6r£00 I8100—0 7696 vSt00 p8I00— 0 086 O0F000 c6000— cS00'1
OLS00
O8c00—
O€L8
cEe90°0
O1<0O0—
$800
0000°€Z
8L96
+2000
8S100—
I
vbrldood
06re00—
699L
cc600
cevO0—
seco
SITIO
OS7'EZ
01660
GOLIO0
0
00S*E7 09L00 Ore0O0— 00£9°0 EU see 205000 CLE6 SLEl00 0— S900 Ielol
OSL*ET ECLOIO O=-*2 COED 0 C660 oO creel 6SS00— L160 $9100 L0c0O0—| 10 9E
000°rZ 0) 0090
== Secu 0 0$9¢ 0 Sé6rl 0— 8850 0 £606 298100=0 beeO Selol
OST'P7 OtPr00 vSIOO0— EVETO er9l00— 9850 0 £768 +=8070°0
0— Scte0 LILOs1
00S‘ 06c00 0— $600 O Orr PLLI0 I8s00—0 L088 U Ccuwee Ue PGcg 6600'1
OSL*PZ 9100 0S0O00— 18L0°0 0 Stst O—— [9S0 0 SCL8 U0 eeeOr Bee 1800'I

0
I

0 Oss
Pre
BOLU
8S00'I

1c98°0

68000
00700

000°SZ
—05C0'0
cS00

ScOOO—
6eS00—

eV Oe
9000
$9580

86100
8sI6l9161-0

OS7'STZ
20570:0

SLV8
£700
€l000—

E000
CVSU0—

IL070
ILLEO
cc000
06000
00S°S7
90000—
09pS00—
JO
JY}
=
=

dAvY
useq

I[Qe}

Ie
DAO”
%™

UT ose
7m

BIOS,
SdtIjUd

sy,
=?

Ul
0}
"wo

ON
JOJ
oy}
sey
POI
00¢‘77

LEP
‘00061

TYNUINS
‘O0S‘SI

usoq
poyeoo]

‘eiep
Jqu],
dy} AreutId

UdSATS
‘AyTeouaunu

pojepfodisjur

poydde
UOISNIJUT
pd}dd1ION
UOTIDIIION
ae (r's'S)E
UOISNI}UI-poy

337
Table 2(5.5.4) Photopic Trolands and Total Spectral Bandwidth of Monochromatic
Test Stimuli Used in 10° Color-Matching Experiment by Stiles and Burch (1959)
Wavenumber Log (Photopic Trolands) Total
m Approximate Values Bandwidth
(cm ') Part I Part II Part III (cm~')
Yellow Run
14,000 (Zr Source) 22 Bee, 188
500 3.0 2.9 182
14,500 (W Source) 2.4 263 182
15,000 3.0 2.9 176
500 pa) 2.9 169
16,000 2.4 ie 162
250 2.6 3.4 158
500 2.6 3.4 155
750 Ze] She) 152
17,000 2.7 ode 149
Green Run
17,000 (W Source) Pa ah 149
250 2.8 35 146
500 2.8 an 143
750 2.8 335 140
18,000 2.8 a 138
250 2.7 3.4 135
500 21 3.4 132
750 2.9 no 129
19,000 2.8 a2 126
250 2.8 3.1 123
500 ail 3.0 121
750 2.9 2.8 118
20,000 2.8 vie | 116
250 2.6 2.6 Le
500 V7) 2.4 110
750 2.4 2.3 108
21,000 2.4 2.2 132
250 oat 2.0 129
500 24 1.9 126
750 — 1.9 123
22,000 2.0 1.9 =sre BL)
500 1.9 1.8 114
23,000 1.9 2.0 88
500 1.6 1.8 83
24,000 1.6 - ed) 131
500 1.4 | Ba) 226
24,500 (Zr Source) 1.8 19 124
25,000 je, 1.6 210
500 1.3 1.4 529
Deep Red Experiment
15,000 (Xe Source) Je): 774
15,000 2.2 176
14,000 2g 830
14,000 22 188
13,081 22 (890)?
133 ies (950)?
Table 2(5.5.4) — Continued

Visual Field: Circular, horizontally divided bipartite field. Ten-degree angular subtense with a
narrow surround of the same spectral composition and luminance as the mixture
half-field. With the surround the extended total field diameter is 14°. Maxwellian
viewing technique is used through the center of the natural pupil.

Instructions to In making color matches, observers are instructed (a) not to stare fixedly at the
Observers: center of the field while approaching the color match but, on the other hand, to base
their final judgment on the field appearance with the gaze directed to the center; (b)
to disregard in the final judgment the central region of the field (1° to 2° diameter)
in those matches, where the Maxwell spot made it impossible to make the half-fields
similar over the whole area for any setting of the primary stimuli.

Yellow Run: Instrumental primary stimuli in

Part I: MR; = 15,420 my, =17,000 min 22450cm ©

Part IT: M pr; = 15,420 my, = 17,000 mis 27150

Twenty-four observers participated in Part I experiment, 29 observers in Part II.

Four observers were common to both parts.

Green Run: Instrumental primary stimuli in

Part I: Mp, — 15,420 Mo, = 19,000 M9 = 22,450 ema

Part IT: MR; — 15,420 Mo, = 19,000 Hine 22 oU.cme

Twenty-four observers participated in Part I experiment, 29 observers in Part II.

Four observers were common to both parts. (Groups of observers are the same as in
the yellow run.)
Deep Red
Experiment: Instrumental primary stimuli in

Part LIT: Wiese — 14,250 my, = 17,000 ee Oe

Nine observers, selected from Parts I and II, participated.

Observer Age and 16 to 23 24 to 32 33 to 42 43 to 55

Sex distribution: Years Years Years Years
M F M F M F M F

(Total of Parts I to III) 6 10 17 5 8 l 13 2

The color-matching data obtained in Parts I to III with different sets of primary
stimuli were transformed by conventional methods to color-matching functions
referring to a common set of primary stimuli located at mp =15,500, mg = 19,000,
m g = 22,500 cm™', whose unit quantities were of equal radiances [W-m *-sr_ '].

“Band contains prominent Xe-lines.

339
340 VISUAL EQUIVALENCE AND VISUAL MATCHING

Table 3(5.5.4) Mean 10° Color-Matching Functions, Spectral Chromaticity Coordinates, and W.D.W.
Coordinates for Nine Observers Determined by Stiles and Burch (1959) in the Deep Red of the Spectrum.
These Data Extend the Data Given in Table 1(5.5.4) from 14,000 to 12,000 cm‘. No Correction for Rod
Intrusion Has Been Applied to the Data [see Table 4(5.5.4)].

m Color-Matching Functions Chromaticity Coordinates W.D.W. Coordinates

(cm~') r(m) g(m) b(m) r(m) g(m) b(my "(my 4a(t)
12,000 2857506) SOISS2(S9 2 OO ae? L019 —0.0014 —0.0005 1.0110 —0.0062 —0.0048
1250052 2665) 29= 8, 9638) 25G=10)a3h0032 —0.0032 +0.0000 1.0142 —0.0143 0.0001
13,000 2.0(—4) — 1.00(—6) 5.9(—8) 1.0047 —0.0050 0.0003 1.0197 —0.0224 0.0027
13500) © “4 B39) =9.2(6-6) 6 Tee LOG! —0.0065 0.0004 1.0258 —0.0294 0.0036
14,000 9.74(-3) —7.4(—5) 3.6(—6) 1.0073 —0.0077 — 0.0004_-1.0311 —0.0345 0.0034

“The negative digits in parentheses following the values of the color-matching functions denote powers of 10. Thus,
the entry 3.75(—6) reads 3.75 X 10~©, etc.

the rod mechanism can be estimated by multiply-
ing the quantity in energy units of each stimulus
by the scotopic (rod) luminous efficiency function
V’(m) and the constant factor k’ required to
express the effect on the rod mechanism in
scotopic trolands T’. It follows:
(a) Total scotopic effect in the upper half of the
visual field:
T, = k'Q(m)V'(m) + k'R(m)V'(mg)
[2(5.5.4)]
(b) Total scotopic effect in the lower half of the
visual field:
Ti = k’G(m)V'(mg) + k’B(m)V'"(mz)
[3(5.5.4)]
If the scotopic effects T/ and T/ in the two
halves come out equal, then we may assume that
even if the rod mechanism were still to respond,
the color match would be unaffected. Otherwise,
the possible disturbance of the match may be
expected to depend to a first approximation on a
quantity S defined as the difference AT’ = T/ —
T; of scotopic effect, divided by the smallest
perceptible or threshold difference 7, for the
scotopic mechanism, adapted to a level some-
where between T/ and T/, a level for which the
geometric mean T = (T/T/)'/* is probably a good
estimate. Thus,
a AT’
S [4(5.5.4)]
Ts
An approximate calculation of the ratio S for
color matches made with test stimuli varying in
wavenumber through the visible spectrum was
made by Stiles (1958) for his 10° pilot group data
using the stimulus levels employed in that experi-
ment and the threshold data for rod vision ob-
tained by Aguilar and Stiles (1954) (see Section
7.5). The results of these calculations, shown in
Figure 1(5.5.4), indicate that over most of the
spectrum, the ratio S is below unity. This is
largely because the scotopic level is high enough
to lift the rod mechanism to the first stages of
saturation. However, in the orange and red region
of the spectrum, on either side of thé “red”
primary stimulus (mz = 15,500 cm7'), the ratio
S reaches a rather high value of nearly 12.
The scotopic imbalance specified by the ratio
S is not the only factor determining the dis-
turbance of the color match. Other factors may be
considered as defining the extent to which the eye
will allow an imbalance of the cone mechanisms
in the compromise match it is forced to make in a
three-control color-matching system. Neverthe-
less, the results of the calculation of S given in
Figure 1(5.5.4) indicate the spectral regions where
rod intrusion is potentially important.
The ratio S is high in the orange and red
spectral regions because the only stimulus in either
half of the visual field of high scotopic value is
the “green” primary stimulus and, at the same
time, the mean scotopic level T’ is not high. The
main impact of the “green” primary stimulus is
on the “green” cone mechanism, which has maxi-
mal sensitivity at about 18,500 cm '!. To produce
the same effect on this cone mechanism with a
“yellow” stimulus at 17,000 cm~! and a “green”
stimulus at 19,000 cm!, the radiance of the
 Color-Matching Functions of Normal Trichromats 341

low” primary stimulus at 17,000 cm~! has a

17000
primary eth
primary stimulus ‘
25.0 220 20.0 17.5
Wavenumber m (cm')
Fig. 1(5.5.4). Results of calculations of ratios
S = AT’/z, for 10° color matches with test stimuli
of different wavenumbers and (photopic) troland
values used in Stiles 10° pilot-group investigation.
Curve through open circles refers to color matches
where the middle primary stimulus is a “green” at
19,000 cm '. Curve through solid dots refers to
color matches where the middle primary stimulus
is a “yellow” at 17,000 cm '. For detailes see text
(after Stiles, 1958).
former stimulus must be about twice that of the
latter. On the other hand, for the same effect on
the rod mechanism, the radiance of the stimulus
at 17,000 cm! must be about 13 times that for
the stimulus at 19,000 cm~!. Thus, for the same
effect on the “green” cone mechanism, the “yel-
smaller effect on the rod mechanism, correspond-
ing to a radiance ratio of 1:6.5. A calculation of
the S ratio for color matches made with a middle
primary stimulus at 17,000 cm~! leads to the
curve through the solid dots in Figure 1(5.5.4), the
(photopic) troland values of the test stimuli as-
sumed being approximately the same as for the
curve through the open circles. Potential intrusion
by the rod mechanism is materially reduced.
The modification produced by the use of the
15.0X103 “yellow” middle primary stimulus appears prin-
cipally in the b(m) color-matching function for
wavenumbers m < 17,250 cm'. Other devices,
the addition of equal amounts of “green” primary
stimulus to both halves of the visual field, and an
increase in the luminance of the visual field also
reduce rod intrusion in a way conforming with
the general picture given above of the rod intru-
sion phenomenon.
Any one of the devices or combinations of
them may still leave a small residual effect of rod
intrusion in the 10° color-matching functions for
m < 17,250 cm '. However, corrections can be
calculated that lead to rod-free color-matching
functions. The corrections are made on the as-
sumptions that:
(a) for these rod-free color-matching functions,
the spectrum locus in the chromaticity diagram

Table 4(5.5.4) Mean Rod-Free 10° Color-Matching Functions, Spectral Chromaticity Coordinates, and
W.D.W. Coordinates Determined by Stiles—Burch (1959). The Data Are Derived from Those Given in
Tables 1(5.5.4) and 3(5.5.4) by Applying Corrections for Rod Intrusion as Outlined in the Text.*

m Color-Matching Functions Chromaticity Coordinates W.D.W. Coordinates

(cm~') r(m) g(m) b(m) rim) g(m) b(m) r*(m) g*(m) b*(m)
12,000 3.75(-6) —1.64(—9) 1.8(—11) 1.0004 —0.0004 0.0000 1.0019 —0.0019 0.0000
00) S257 665-5)! He 88C= 8)0095(— 10) 1.0032" —0,0032. 0.0000. 1.0138 —0'0141 0.0003
13,000 2.00(—4) —1.08(—6) 1.17(—8) 1.0054 —0.0054 0.0001 1.0239 —0.0244 0.0005
Perea 4150-3) Ge 1.0165) 910907), al0071. > 0.0072. 0.0001 ..1.03.15, 0.0322, 0.0007
14,000 9.74(-3) —7.9(-—5) 8.6(—7) 1.0081 -—0.0082 0.0001 1.0361 -—0.0370 0.0008
14,500 6.01(—2) —4.90(-—4) 5.3(—6) 1.0081 —0.0082 0.0001 1.0362 —0.0370 0.0008
15,000 2.96(-1) —1.89(—3) 2.0(—5) 1.0064 —0.0064 0.0001 1.0280 —0.0287 0.0006
15,500 1.000 0 0 1.0000 0.0000 0.0000 1.0000 0.0000 0.0000
16,000 2.17 3.79( —2)—4.1(—4) + 0.9830 =0.0172—0.0002 0.9303 0.0713 — 0.0016
16,250. 2:70 9.2(—2) —9.9(-4) 0.9674 0.0330—0.0004 0.8722 0.1306 — 0.0029
16,500 3.08 1.80(— 1)—1.95(—3) 0.9454 — 0.0552—0.0006 0.7991 0.2054 — 0.0045
16,750 3.16 BOs!) 333 (423) 000.9126 0.0884—0.0010 0.7071 0.2994 — 0.0066
17,000 3.02 4.53(—1)—4.9(—3) 0.8708 0.1306—0.0014 0.6088 0.4000 — 0.0088

“The negative digits in parentheses following the values of the color-matching functions denote powers of
10. Thus, the entry 3.75(—6) reads 3.75X 10°, etc.
Table 1(5.5.5) Mean 10° Color-Matching Functions Obtained by Katori and Fuwa (1979) for 10 Observers.
The Coefficients of Deviation are the Ratios of the Standard Deviations over the Mean Values, Expressed in
Percent. The Mean Photopic Trolands are Derived from the Luminance of the Visual Field at the Mean Color
Match.”
Wavelength Color-Matching Functions? Coefficients of Deviation (%) Photopic Trolands, Bandwidth
\ (nm) r(A) g(A) b(A) Ar(A) Ag(A) Ab(A) T(A) (nm)

395 WS8(E-2) a Poor) 86 61 64 3 6.0

400 3.929(—2). —5,8136—A) 5.093(—2) 70 61 59 13 6.8
410 ROS TGs ty) 291662) 2.054(— 1) 38 36 30 96 7.8
420 HAIN HT) BF 2) 4.541(—1) 25 a2 15 157 8.7
430 D306C= 1)” 8:3 9TE—2) 7.761(— 1) 19 16 12 161 9:7
440 570356 1) VS ey SS 18 18 11 181 5.4
450 3:624(—T)y' '~6.174(~—2) 1.099 2h 17 33 187 6.0
460 0.000 0.000 1.000 — — os 178 6.5
470 —5.990(— 1) biS6¢s 2) 7.398(— 1) 16 2 5.0 pets 7.2
480 — 1.245 251561) 4.356(— 1) 16 8.8 4.9 221 22
490 = 1.695 396361) 2.349(— 1) 14 8.9 8.3 248 32
500 — 1.818 5.654(— 1) 1.245(— 1) 9.1 6.6 9.2 257 3.4
510 — 1.801 7.399(— 1) 5.631(—2) re 4.8 8.9 265 5:7
520 —-0226 8.942(— 1) 2:362¢=.2) 17 2.9 14 255 4.0
530 0.000 1.000 0.000 — —_ 229 4.3
540 1.298 1.047 I ed K=3) 8.5 1.1 13 201 4.6
550 3.000 1.018 py halON( 2) 9.5 1.3 iy, 214 4.9
560 5.135 9.450(= 1). .— 1.470(—2) Bl 2.0 14 218 5.2
570 7.274 SOON Ir eer tet2 6.4 2.8 13 213 me
580 9.267 GASI(-1)y lat 2) ia 4.7 13 Paw 39
590 1.081(1) 4.624(—1) —1.616(—2) 3.6 6.1 14 203 6.3
600 1.146(1) 2.940(—1) —1.483(—2) 33 8.2 14 193 6.6
610 1.088(1) 1694-13" = Te 2) aol 7.6 16 182 7.0
620 9.444 DOL 2) 8 2Ie oo) 23 8.7 20 200 oe
630 7.456 3.906(—2) —4.350¢—3) 4.4 9.6 20 199 7.8
640 5.166 1.543(—2) —2.173(—3) 2.6 1.2 30 199 8.2
650 3.318 593 180473) a989.47F( 34) 4.5 9.6 20 7 1 8.6
660 1.897 1.206(—3) —3.942(—4) 32 19 26 213 9.1
670 1.000 0.000 0.000 oo -- _ 204 pe
680 P2351) = LS Ze) 5.386(—5) Sul 38 67 255 10.0
690 23553(— bai F=9.450645) 3.410(—5) 8.3 29 66 206 10.5
700 1.146(— 1): ,.-5-9T1(55) 1.249(—5) 2.6 25 80 195 29.0
720 2911 (2)e, 243065) 3.262(—6) 6.4 139 55 yy ayAL
740 3.34 1(=33).. = 2173(—6) 1.181(—6) 6.3 33 46 75 35.0

“Visual Field: Circular horizontally divided bipartite field. Ten-degree angular subtense with a narrow
surround of the same spectral composition and luminance as the mixture half-field. With the surround, the
extended total field diameter is 14°. Maxwellian viewing technique is used through an artificial pupil of
3-mm diameter.
Primary Stimuli: Xp = 670 nm, Ag = 530 nm, Az = 460 nm with bandwidths AA p = 26 nm, AA; = 6 nm,
AA , = 12 nm.
>The numbers in parentheses denote powers of 10. Thus, 1.588(—2) reads 1.588 X 1077.

342
 Color-Matching Functions of Normal Trichromats
[g*(m) versus r*(m) W.D.W. coordinates] coin-
cides for m< 17,250 cm~' exactly with the
straight line joining the chromaticity points at
m = 17,250 and 15,500 cm!; and
(b) the deviations from these color-matching
functions produced by the rod mechanism corre-
spond to the addition to the stimulus with the
greater scotopic value of a small amount of
equal-energy white (wavelength basis). It is not
necessary to assume in advance how much equal-
energy white must be added to represent rod
intrusion under any given observational condi-
tions.
From the above assumptions, rod-free color-
matching functions can be derived for m < 17,250
cm ' from the mean observed functions given in
Tables 1(5.5.4) and 3(5.5.4). They are listed in
Table 4(5.5.4).
5.5.5 ETL Ten-Degree Data
Katori and Fuwa (1979) of the Electrotechnical
Laboratory (E.T.L.) of Japan determined 10°
color-matching functions for 10 Japanese normal
trichromats (eight males, two females) 24 to 44
years old (35.7 years average age), on an instru-
ment constructed along the design features of the
N.P.L. trichromator. The observers used the
method of adjustment to arrive at a color match.
The study aimed at the derivation of a 10°
luminous efficiency function from the measured
color-matching functions and from heterochro-
matic brightness matches made between the three
primary stimuli by the flicker technique.
Further reference to the derived luminous ef-
ficiency function will be made in Section 5.7.2.
The mean color-matching functions determined
for the 10 observers are listed in Table 1(5.5.5)
together with other pertinent information con-
cerning the color-matching experiment.
The luminance of the bipartite visual field at
color match was kept approximately at 200 td
throughout the spectrum except at the ends where
lower troland values had to be used. The lumi-
nance levels used by Stiles and Burch (1959),
given in Table 2(5.5.4), were generally higher in
the long-wavelength part of the spectrum. Katori
and Fuwa estimated that these level differences
were the main reason for their color-matching
functions, particularly 7(A) and b(A), to deviate
somewhat from those of Stiles and Burch at longer
wavelengths. Rod intrusion was considered to
have been a factor.
343
In a supplementary report by Katori, Fuwa,
and Ooba (1980), the same 10 observers made
color matches by the method of constant stimuli.
The mean color-matching functions of the 10
observers closely agree with the mean data ob-
tained by the same observers by the method of
adjustment.
5.5.6 Variations of Color-Matching Functions
of Different Normal Trichromats
The extent to which the color-matching properties
of normal trichromats differ from one observer to
another is not readily specified. The data on
which the CIE standard observers were based
provide some information. Guild (1931) and
Wright (1928-1929) both give diagrams showing
the individual variations in the spectral chro-
maticity curves of their groups of observers. Fig-
ure 1(5.5.6) shows the range of those variations
deduced from the graphs of Guild and Wright.
Coordinates
Chromaticity
!
400 450 500 550 600 650 700
Wavelength A (nm)
Coordinates
D.
W.
450
Wavelength A (nm)
Fig. 1(5.5.6). Two-degree spectral chromaticity
coordinates determined by Guild (1931) and
W.D.W. coordinates determined by Wright (1928-
1929) with ranges of variation for the different
groups of observers involved. Guild’s data (upper
graph) are based on measurements made by seven
observers, those of Wright (lower graph) are based
on measurements made by 10 observers.
 —

344 VISUAL EQUIVALENCE AND VISUAL MATCHING

520

-0.2
r

Fig. 2(5.5.6). Chromaticity points of matches on a standard white of 36 observers plotted in a

W.D.W. diagram (normalizing wavelengths 582.4 and 494.0 nm), with the spectrum locus (mean
of 10 observers). The white stimulus has a correlated color temperature of approximately 4800
K (from Wright 1928-1929).

Figure 2(5.5.6) shows the variations in the match
on a standard white of 36 observers included in
Wright’s basic study. The variability of the other
essential constituent of the CIE 1931 standard
observer, the V(A) curve, is illustrated in Figure
2(5.14.1). | malization. For example, the blue color-matching
Individual variations in the basic data of the
CIE 1964 supplementary standard observer for
colorimetry are illustrated (for the data of Stiles
and Burch, 1959) by Figures 3(5.5.6) to 5(5.5.6).
Figures 6(5.5.6) and 7(5.5.6) illustrate (a) the
standard deviation of each of the color-matching
functions expressed as a percentage of the mean
values of the function, and (b) the standard devia-
tions of the W.D.W. coordinates, computed in
each case from the results of the whole group of
observers. Table 1(5.5.6) illustrates in addition the
correlation coefficients between the color-match-
ing functions at each wavenumber.
Variations in the match on a standard white
are given in Figure 8(5.5.6).
The singularities in the curves in Figures
6(5.5.6) and 7(5.5.6) that appear at the wavenum-
bers of the primary stimuli result from the nor-
function b(m) has the value unity at the blue
primary (m, = 22,500 cm') and the value zero
at the green and red primaries (m, = 19,000 and
mp = 15,500 cm~'), and because this is exactly
true for each observer, the standard deviations
must be zero.
The information on individual variability con-
tained in these figures is incomplete for no account
is taken of the fact that for a given observer, the
deviations from the mean of the values at differ-
ent wavenumbers may be correlated.
The deviations of different observers in mak-
ing any particular metameric match—not only
 Color-Matching Functions of Normal Trichromats 345

4.0

10° Color Matches

3.5 49 Observers
Stiles - Burch (1959)

3.0

2.5

Na (6)
(@)

°
Tristimulus
Values

0.5

=0.5

25 24 23 ee 2\ 20 19 T:) 17 16 15 14-10°
Wavenumber m (cm’!)
Fig. 3(5.5.6). 10° color-matching functions of 49 observers participating in the Stiles—Burch
(1959) experiment. All functions refer to primary stimuli at mp = 15,500, m, = 19,000, and
m pz= 22,500 cm'.

those involving four monochromatic components
—would be the most useful information. For the
special case of a white match with three mono-
chromatic primaries, the diagrams of Figures
2(5.5.6) and 8(5.5.6) already provide this, except
that the W.D.W. method of normalization in this
application complicates the interpretation. It is
probably simplest to consider the color-matching
functions of all observers expressed in a system
based on three monochromatic primaries, the re-
spective functions being normalized to have the
value unity at the corresponding primary wave-
Suppose two stimuli of spectral distributions
{P, dX} and {Px dX} are approximately meta-
meric (metameric for one observer, say), and the
color-matching functions of a representative group
of N observers equal 7,(A), 2;(A), 5,(A), (i = 1 to
N), producing tristimulus values R,;, G;, B; and
R’,, G;, B; respectively, for the two stimuli. The
differences of tristimulus values AR; = R; — R 1?
AG;, AB; or of chromaticities
RAGMARY Gch Gi;
ér, =
8g; a ip ps

length.
 346 VISUAL EQUIVALENCE AND VISUAL MATCHING

5.0

4.0

3.0

an

JS
PS x
>

48 ~~ S
Spectrum Loci of 49 Observers
Stiles
- Burch (1949)

(R)
-6.0 =7.0 -4.0 -3.0

-1.0
Fig. 4(5.5.6). (r, g)-chromaticity diagram of Stiles—Burch (1959) 10° color-matching investi-
gation showing the spectrum loci of 49 observers.

where tigation of Stiles and Burch (1959). The selection

> =R,+G,+B, and >’ = Ri + G+ B
1 i
then provide, when suitably plotted, a quick ap-
praisal of the kind of variations to be expected in
this particular metameric match from individual
differences in color-matching properties. With a
computer, this procedure is readily applicable to
any match, given a representative set of individ-
ual color-matching functions.
For large-field color matching, Table I(5.5.6)
in the Appendix gives such a set of color-match-
ing functions obtained for 20 observers of which
the first 18 were used in the color-matching inves-
from the data of the whole group was made by
taking the observers with the greatest reliability
and experience in trichromatic matching, without
reference to their actual results. The data of the
last two observers, GW (58) and DBJ (59) in
Table 1(5.5.6), were added after completion of the
main color-matching investigation of Stiles and
Burch.
The individual variations obtained in the
R, G, B-system are readily transformed to a CIE-
type system corresponding closely to that in which
the 1964 standard observer system is expressed.
An appropriate transformation to be applied to
R;, G;, B;, (R;, Gj, B; or AR;, AG;, AB;) is the
 Factors Modifying Color Matching 347

1.50

49 Observers
1.25 Stiles - Burch (1959)

0.75

"4 je)

Coordinates
W.D.W.
0.25

=0.25

-0.50
25 24 23 22 2 20 19 18 17 16 15 14-10°
Wavenumber m (cm’')
Fig. 5(5.5.6). W.D.W. coordinates of 49 observers of the Stiles—Burch (1959) 10° color-
matching investigation.

following: 5.6 FACTORS MODIFYING COLOR MATCHING

The color-matching properties of individual ob-

_ X, = 0.3411R, + 0.1891G, + 0.3875B,
Y, = 0.1391R, + 0.8375G, + 0.0733B,
Z, = 0.0000R, + 0.0396G, + 2.0262.B,
[1(5.5.6)]
A similar test set of color-matching functions
for small-field matching is not available.
servers—all normal trichromats or members of
the same class of color defectives—differ from
those of the ideal representative observer by more
than can be ascribed to random variations in
actual matching measurement. The causes of some
systematic deviations are recognized and dis-
cussed below.
5.6.1 Filter Pigments in the Eye
The yellow pigments of the ocular media, pre-
dominantly of the eye lens and of the macular
348 VISUAL EQUIVALENCE AND VISUAL MATCHING

&,/b(m)
100
Log

&,/9(m)
100
Log

(m)
&,/r
100
Log

25000 22500 20000 17500 15000

Wavenumber (cm7!)
Fig. 6(5.5.6). Logarithms of the standard deviations of 10° color-matching functions (primaries
at 15,500, 19,000, and 22,500 cm” ') from measurements on a group of 49 observers (Stiles and
Burch, 1959). Open triangles: variations of b(m) for 49 observers; oblique crosses: variations of
g(m) for 49 observers; open circles: variations of r(m) for 49 observers; open diamond:
variations of b(m) for 2 experienced observers. One observer measured the color-matching
functions at selected wavenumbers four times, the other five times. Mean variations for the two
observers are plotted. Right- angled crosses: same as open diamonds but of g(m) for the same
two experienced observers; open squares: same as open diamonds but of r(m) for the same two
experienced observers.

area of the retina (see Section 2.4.6), are believed
to be photochemically inert and to act simply as
filters interposed between the external stimulus
and the visual receptors. Light that fails to reach
the receptors in the direct image of a test stimulus
because of scattering in the path through the eye
may, for the present purpose, be lumped in with
losses by absorption in filter pigments.
Suppose that, for a group of observers, the
properties of the visual system from the visual
receptors and beyond are identical and would
correspond, in the absence of all prereceptor light
losses, to color-matching functions 7(A), go(A),
bo(A) based on a set of monochromatic primary
stimuli. The actual color-matching functions of
each observer, based on the same primaries, have
the form t(A)7(A), f(A) 8o(A), H(A) bo(A), where
[1 — t(A)] is the fraction of light of wavelength A
lost in passage through the particular observer’s
eye. In making use of a special normalizing proce-
dure developed by Wright (1928-1929), already
referred to in Section 3.3.3(a), W.D.W. color-
matching functions r*(A), g*(A), b*(A) are ob-
tained for each observer of the group. The W.D.W.
 Factors Modifying Color Matching 349

2.0

3.0

2.0

Log oq

3.0

2.0

Log o,

3.0

25000 22500 . 20000 17500 15000 cm7!

Wavenumber (cm-!)
Fig. 7(5.5.6). Logarithms of the standard deviations of the W.D.W. coordinates (normalizing
wavenumbers 17,250 and 20,500 cm') corresponding to the variations in color-matching
functions shown in Figure 6(5.5.6). The open circles represent, for each coordinate, the
standard deviation corresponding to variations among observers; the crosses represent the
standard deviations of a single observer in repeat measurements, averaged for two observers
(Stiles and Burch, 1959).

color-matching functions are defined simply by W.D.W. system:

RUEELCOLAC®
r*(X) W, The
:
: (A) ‘
PoCA ie" (A) GsGA.)
B*(A) = H(A) B(A) © [165.6.1)]
: = eee?
B*(A) (A) bo(A) 7 rol) -
ue | 2D 4 gory + BOD
where the constants

ya 22]
B()
wt ma[B@2]
B(A)
dap EO
P(A) + g*(A) + b*(A)
are determined at the normalizing wavelengths A y
BoA)
and A pc. Le Sor Eee or
The chromaticity coordinates r*(A), g*(A), La + g(r) + ea
b*(A) of monochromatic stimuli become, in the W, W,
Table 1(5.5.6) Standard Deviations o,, o,, 0, of Color-Matching Functions Expressed as
Percentages of Mean Values.“ Correlation Coefficients C,,C,,,C,, Expressed as
Percentages.”

m 1006, 1000, 1006,

(cm~!) =a 100C,, ae 100C;,. ate 100C,,

14,000 4.5 = 0 15 “ii 51 ole

14,500 3.9 ys 14 + 43 51 =
eed)!
15,000 3.1 —4] 16 +29 88 1G
15,500 (0) — — — — —
16,000 3.6 22 10 31 54 +31
16,250 at — 43 8.8 “24 35 +22
16,500 6.7 729 8.2 13 31 +34
16,750 7.9 28 ae +) 24 a5
17,000 9.1 ae 6.5 =32 22 +39
17,250 9.4 e335 5.6 ==35 22 sea
17,500 10 =26 3.9 ==34 20 + 30
17,750 12 —4] 4.1 — 46 24 +4
18,000 11 7 20 2 —49 19 +26
18,250 12 =6 sel = i| 22 +10
18,500 12 15 3.0 92 21 +10
18,750 14 Pale re == 39 25 + 36
19,000 o (0) —- — —
19,250 Li —14 2.8 == Si) 18 + 44
19,500 14 — 40 3.9 — 43 IZ +40
19,750 13 —44 apee) — 44 14 + 38
20,000 14 —54 7.9 — 44 11 +49
20,250 14 — 60 10 —33 1] 253
20,500 5 — 64 9.0 ~~16 12 +49
20,750 Is si 2 10 +226 9.4 +47
21,000 15 aot] 11 — 34 95 + 48
21,250 16 — 80 11 238 11 + 36
21,500 20 = on 14 — 48 9.9 AS 2
22,000 20 = $83 24 == 38 Tal + 52
22,500 — _- —- — (0) _
23,000 43 = 31 aed) 4 7.5 +54
23,500 26 5 26 332 8.9 +35
24,000 ei =—§2 24 ame ie 17 + 68
24,500 44 PO 44 =+1$3 be. 87
25,000 66 =935 58 coef! 54 + 93
25,500 95 —94 87 95 719 + 96

“Standard deviation:

ays 2.1 we
_> __ x(x xe

na

of color-matching function x in n separate determinations under identical conditions of

observation.
Correlation coefficient:

= D(X X wean Ve vee

[s(x m Cee X(y iy 4 ans) te

of color-matching functions x and y in n separate determinations under identical

conditions of observation.
 Factors Modifying Color Matching 35]

will be different for different functions (A) in a

way that depends on {P, dA}. This important
relation is illustrated in Figure 2(5.5.6). The chro-
maticity points shown in the diagram are those
measured by Wright’s (1928-1929) 36 observers,
of a stimulus of color temperature 4800 K in a
system with primary stimuli at Ap = 650 nm,
Ag = 530 nm, and A, = 460 nm and normalizing
wavelengths at Ay = 582.5 and Ap, = 484 nm.
Subject to the assumption that the chromatici-
ties of the monochromatic stimuli are substan-
0.100 0.150 0.200 0.250
tially the same for the different observers, the
distribution of the chromaticity points in Figure
Fig. 8(5.5.6). Chromaticity points of matches on 2(5.5.6) embodies the variations in their f(A)
a white stimulus made by 55 observers, plotted in a functions. Thus, the observer with chromaticity at
portion of a W.D.W. diagram (primaries of wave- point A is identified as the one with the least
lengths 645.2, 526.3, 444.4 nm; normalizing wave- prereceptor pigment, the chromaticity points dis-
lengths at 579.7 and 487.8 nm). The chromaticity placed towards the yellow region of the spectrum
of the stimulus of an equal-energy spectrum com- representing observers with increasing amounts of
puted from the average color-matching functions yellow pigment (Wright 1928-1929).
of the 55 observers is shown as a square (from Another example is given in Figure 8(5.5.6)
Stiles and Burch, 1959). showing the chromaticity points of 49 observers
obtained in the large-field color-matching investi-
gation of Stiles and Burch (1959). Note that the
spectral primaries, normalizing wavelengths, and
b*(A) = b*(A) the spectral distribution of the “white” stimulus
r*(X) + g*(A) + B*(A) differed from those relevant to Figure 2(5.5.6).
There is again a drawing-out of the distribution of
bo(A) the chromaticity points toward the yellow region
of the spectrum.
ro(A) bo(A)
W
W
AIKOS banemerrne
W,
A study by Ruddock (1963) goes into further
details, particularly of the effect of macular pig-
mentation on color matching when the field size
[2(5.6.1)] and field location are varied. In the same study,
Ruddock uses color matches to predict the den-
They are independent of ¢(A), and therefore sity of the macular pigment both with regard to
the same for all observers. However, the chro- its spectral variation and its variation across the
maticity of any mixed stimulus of spectral radiant fovea.
power distribution {P, dA}, that is, The approximate relative spectral density
curves of the macular pigment and the lens pig-
_ Jt(A)r(A)Pada ment are known (Section 2.4.6), so that the effect
? > on the chromaticity of the “white” stimulus of
increasing or reducing the amount of these pig-
ree [3(5.6.1)] ments in an observer whose color-matching func-
tions are known can be calculated. The results of
such a calculation are shown in Figure 1(5.6.1).
Jt) b(A)PA AA The square point represents the chromaticity of
SEP
YAS Mehoe the equal-energy “white” stimulus, {P, dA} =
const, for the average observer of the Stiles—Burch
investigation in the same W.D.W. system used in
where Figure 8(5.5.6). The assumed relative spectral
density curves for lens and macular pigment are
ya fro] 2 4 2(A) +
bol) those shown in the inset to the figure, the macular
: P, dX
pigment density being specified by its value at
 @563.4nm

@570nm
\.571.4.1m
a pontie) J Sesz2nm
v
y!
@579.7nm

/x
[+1.136 (L)}#o \+2696 (my)
2

0.696} O-=—AVERAGE OBSERVER]

ene

wherh wee

acy 0.2 0.3 04 05 0.6

Fig. 1(5.6.1). Computed effects on chromaticity in the W.D.W. diagram of increasing the
density of macular and lens pigments in an average eye viewing an equal-energy stimulus. The
assumed spectral variations of the density of lens and macular pigments are shown in the inset
figure (see also Section 2.4.6). The increase (or decrease) of lens or macular density, specified
at A = 455 nm for macular pigment, and at A = 403 nm for lens pigment, is shown for each
point. The calculations are based on the average large-field color-matching functions (Stiles
and Burch, 1959) for which a section of the spectrum locus is shown. The chromaticity diagram
is the same as that used for Figure 8(5.5.6).

352
 ols!
(wd)
saquInuaADM
td
9I
di

We
-2ABM SUOT|IUNY
Iv[MUISsey

(a)
sua]
JO
Jopun
ooro

UOVIUNY oAa
SUTYIICUI-10JOI
O0SO'O

oy)
6/6]
Jo
SAN|DA SNiNWIjsi4] jD4j0edS &
JUoUIsId
Sv Ul
(W)g
EG
JO
Ide

(|.WUd)
SJ9Say)
(4)

JaquINUSADM
“(w)s
Je
OA}‘sivad
PUL
(U)4 JY}
M9:890

W
626] (w)6 -.—.—.-

QCE] [7
UIBMJI
(9)

SUOTIUN Jo
UI Ueds
ZE
ade SIDUDAIJJIP
dy)
SUIYDJUI-1OJOI
Je IDAO
(8)

PoUIvIJqO
‘Jey
IY}
ool1o

JO SuTWUNSse
0SOO

MA JSOJA]
00011 00¢c'0 daisep-usy,
;
S000
BSE!
(Ww)
B-——
S000

SONIDA SNiNuySi4] |D4y99edS JOAIaSGgO

“SUOIPUOD
Aq
poure[dxa
626) (W) 4 --—-——
BS6I (W) 4° —°-

Jo
(uy)

*(1°9°S)Z
w SUTAIASGO
(;-Wud) WW JOqUINUaADM

Jaquinuoq
“BY ued
agi
cOlS!
eOlsi
SZ
ve
¢2
2OI
[2
Se
ze
le
6O2
Sl
ZI
Of
I
Sl
ZI ATTeQULISqNS
UT
“AjISUOp
pasevaIDUI
(9)

O02
I2 22 £2 be

(a)
Se
0Sso'0
0001

00SsO oolo
SAN|DA SNinwysij j04j99dS

353
354 VISUAL EQUIVALENCE AND VISUAL MATCHING

455 nm and the lens pigment density by its value

at 403 nm.
The cross and circle points represent, respec-
tively, the modified chromaticities if the density
of macular pigment or lens pigment is increased
or reduced by the amount shown, and the triangle —-— b(m)7o/b(m)g,
points show examples of simultaneous changes in ———— kp t)!7(m)/ t_(m)
both pigments. The chromaticity shifts for this
“white” stimulus are very similar for macular and
lens pigments (although this is not always so) and
are in the direction of a spectral yellow at about
572 nm for macular and 570 nm for lens pigment.
The density of the lens pigment increases
materially with age (Section 4.2.6). Figure 2(5.6.1)
illustrates the aging effect on the color-matching
functions of observer GW who measured his first —eee. a(m)z9 /3(M)sg

set of 10° color-matching functions 7(m)se, = kg t(m)/t_(m)

2(m)sg, b(m)sg at age 32 in 1958 at the NPL
(Stiles) trichromator, and his second set r(m)79,
Z(m), b(A)7q at age 53 in 1979 at the NRC
trichromator of similar design.
Figure 3(5.6.1) shows the ratios of the two sets
of color-matching functions displayed in Figure ——— F(m)75/° (Ms
2(5.6.1). The three spectral distributions of these —— k,t,'(m)/t,(m)
ratios are somewhat jagged, but all three suggest
that aging of the eye lens of observer GW must be
a major factor contributing to the change of his 25 24 23 22 2l 20 (G) 18 I7 16 15-105
color-matching functions of 1979 compared with (B) (R)
those of 1958. Wavenumber m (cm!)
Suppose (a) that in 1958, observer GW had a Fig. 3(5.6.1). Ratios of color-matching functions
lens of relative spectral transmittance ¢,(m) iden- obtained by observer GW in 1979 and in 1958
tical to that of the average young eye defined in under similar observing conditions [see Figure
Section 2.4.6, and (b) that aging of the lens can be 2(5.5.1)] compared with ratios of spectral transmit-
characterized completely by increasing its pig- tances t}>(m) and t,(m) of eye lens. It is as-
ment density by a factor a whose actual value sumed that in 1958, the lens of observer GW had
depends on the age increment. Then, assuming spectral transmittances ¢,(m) identical to that of
that aging of the eye lens is the only cause of the the average young eye (see Section 2.4.6), and in
change in the color-matching functions, the fol- 1979, the density of the lens pigment had in-
lowing relations must hold: creased by the factor 1.75, thus giving spectral
transmittances t}’°(m).
im(m)79 _ Keti(m)
rin( ™ )58 t;(m)
When a is set equal to 1.75 (or there about), a

ACE TCD amu oy

Bm M )79 EL kgtr(m) satisfactory fit of the ratios of GW’s two sets of
color-matching functions is obtained. This is il-
lustrated in Figure 3(5.6.1).
b,,(m)79 re k gtr(m)
5.6.2 Rod Participation
b,,(m)sg t,(m)
The scotopic or rod mechanism of vision must
where k,, k g» ky are constants, each normalizing play some part in matches made in a bipartite
the ratio ¢7(m)/t;(m) to unity at the wavenum- field that is imaged wholly in the extrafoveal
ber of the appropriate primary stimulus used in retina or that is large enough to extend beyond
the color-matching experiment. the rod-free area. But most investigators agree

that satisfactory matches can be obtained in rod-
containing areas by using, for the comparison
stimulus, appropriate mixtures of not more than
three fixed primary stimuli. Thus, the addition of
the rod mechanism does not in practice entail
that, for two stimuli to match, their spectral
radiant power distributions {P,, dA} and
{P,, dX}, respectively, must satisfy four, instead
of the usual three, matching conditions. On the
other hand, the matches made are subject to
variation with changes in the adaptation of the
matching field and the principle of persistence of
color matches cannot be assumed (Section 3.2).
For retinal areas not too far from the fovea
which are kept adequately light-adapted by the
matching stimuli assisted possibly by surrounding
or pre-exposed adapting stimuli, three primary
stimuli and three matching conditions are in fact
necessary, and the matching process is trichro-
matic, at least in the weaker sense. As the level of
adaptation is allowed to fall, the matching pro-
cess finally degenerates to become less than
three-dimensional and may approach “mono-
chromatic” scotopic matching with the scotopic
luminous efficiency function V’(A) as the single
color-matching function.
There is some experimental evidence on the
quantitative properties of color-matching in rod-
containing retinal areas. Also, some tentative
quantitative schemes have been proposed to pre-
dict color matches under such conditions. Some
of the more recent developments are reviewed
below.
The question as to whether, and to what ex-
tent, the rods contribute to the color matches
made in a large visual field, such as a 10° bipar-
tite field, has been studied by Stiles and Wyszecki
(1973).
(i) Rod Participation in Large-Field Color
Matching. Suppose that the two stimuli occupy-
ing the respective halves of the visual field have
spectral radiant power distributions {P,, dA} and
{P,, dX} and would be in complete color match
if the rod receptors could in some way be pre-
vented from responding to the stimuli.
Such a hypothetical match may be called a
rod-suppressed match. It will be mediated by the
cones only, and we may suppose that the ordinary
linear laws of trichromatic color matching, which
hold for the purely cone system in the small
foveal area of the retina, are also valid for the
area concerned in large-field matching in the
rod-suppressed state. The conditions for a match
Factors Modifying Color Matching 355
will then be expressible by the equations:
[Pna) dd = [Para) dd
[Pinb(A) dd = [Pb Q) dd
JPre) dd = [Parc) dd
[1(5.6.2)]
where a(A), b(A), G(A) are the color-matching
functions appropriate to matching by cones in the
rod-suppressed condition of the retinal area
covered by the image of the large visual field.
A match made in the rod-suppressed condition
will not be expected to remain valid when, as in
practice, the rods as well as the cones are contrib-
uting their responses. This expected failure would
not occur if the spectral sensitivity curve for rod
vision were a linear combination of the color-
matching functions @(A), b(A), C(A), because in
that case, any pair of stimuli {P,, dA} and
{P,, dX} satisfying Eq. 1(5.6.2) would necessarily
satisfy also .
ifPU Nee ifP,,V'(A) dd
[2(5.6.2)]
where V’(A) represents the spectral sensitivity
curve of the rods and may be identified with the
scotopic luminous efficiency function. From Eq.
2(5.6.2), it follows that the rods would be stimu-
lated equally on the two sides of the visual field.
However, V’(A) is certainly not a linear combina-
tion of the foveal color-matching functions
x(A), V(A), Z(A) that define the CIE 1931 stan-
dard colorimetric observer (Section 3.3.1). The
color-matching functions a(A), b(A), c(A) may
differ somewhat from x(A), y(A), Z(A) because of
a difference in the effective density of macular
pigment for large and small fields and, probably,
also because of minor differences in the proper-
ties of the underlying cones in the foveal and
surrounding retinal areas (see Section 5.7.2).
Enough is known of these sources of difference to
make it certain that no linear combination of
a(X), b(A), (A) would duplicate V’(A).
Nevertheless, for certain pairs of stimuli
{P,, dA}, {P,, dA}, that match in the rod-sup-
pressed condition, it may be that Eq. 2(5.6.2) 1s,
356 VISUAL EQUIVALENCE AND VISUAL MATCHING
in fact, satisfied. On allowing the rods to function
in their normal way, we should not then expect
the match to be disturbed, because the rod re-
sponses in the two halves of the field as well as
the cone responses would be equal. Only on the
assumption of some complex interaction between
rods and cones which made the shapes of the
cone spectral sensitivities depend on the degree of
response of the rods, would there be any upset in
the rod-suppressed match. We note that for the
special pairs of stimuli that satisfy both Eggs.
1(5.6.2) and Eq. 2(5.6.2), the match would remain
valid if the spectral radiant powers in both half-
fields were reduced or increased by the same
factor independent of wavelength; the match
would be independent of stimulus level.
In general, for a rod-suppressed match, Eq.
2(5.6.2) will not hold, but there is still a possibil-
ity that the rods will produce only an insignificant
upset of the match. Broadly, this possibility arises
first because the smallest perceptible difference of
rod response in the two half-fields, expressed as a
fraction of the mean response, already has a
rather high value around 20% in the wide range of
low and moderate stimulus levels over which it is
nearly constant (Aguilar and Stiles, 1954; see also
Section 7.5). In the second place, when a certain
level is exceeded, the smallest perceptible dif-
ference rises at an increasing rate until, at the
so-called “saturation” level, the rods are incapa-
ble of detecting a difference. At and above the
saturation level, the rod system may perhaps be
thought of as responding to its maximum capac-
ity in both half-fields so that at some higher
neural stage, where the discrimination process
operates, there is no difference to detect.
We put
k’ [Piav(r) accal,
k’ [Pav") AN 22OTREG TION
AT! SAGs Fea es ippais
where the constant k’ is adjusted so that the rod
or scotopic values 77, T;, 75, AT’ are obtained in
scotopic trolands (see Section 2.4.4). The smallest
perceptible value of AT’ is termed the threshold
difference t; for the scotopic mechanism whose
value depends on the mean level 7}.
‘The variation of log7; with log75 can be
deduced from the work of Aguilar and Stiles
(1954), described in Section 7.5 of which Figure
1(7.5) is of particular relevance. The conditions
used in their measurements were rather different
from those one would have in a bipartite visual
field; but in default of data for precisely the
present case, the results may be accepted as in-
dicative of what will occur. In the Aguilar and
Stiles results plotted in Figure 1(7.5), the abscissa
represents the scotopic troland value of the back-
ground stimulus on which an increment stimulus
is superposed. The effective scotopic level in the
bipartite visual field, therefore, must be between
T; and 7;, and the geometric mean 75 of Eq.
3(5.6.2) is a tentative estimate and used as the
value on the abscissa of Figure 1(7.5).
If AT’ and 7) are calculated for an actual
match and AT” is smaller than 7,, the rod contri-
bution can be regarded as producing, at most, a
barely perceptible shift from the rod-suppressed
match. In the Stiles—Burch (1959) investigation of
the large-field color-matching functions, condi-
tions were chosen so that in most cases, the rods
could be assumed to be producing a negligible
shift from the rod-suppressed matches. As dis-
cussed in Section 5.5.4, the ratio AT’/t, was
calculated for each of the matches made between
a spectrum color and the monochromatic primary
stimuli and was shown to be less than unity
except for monochromatic stimuli of wavelengths
longer than about 590 nm. For these wavelengths,
auxiliary experiments were made to find how
changes in the matched stimuli designed to reduce
the ratio AT’ /7, would affect the observed match.
These changes were principally an increase in
stimulus level and desaturation of both halves of
the field by addition of the same amount of green
or blue primary stimuli. It was found that as
AT’ /t,; was reduced, the point in the chromaticity
diagram corresponding to the observed match
moved along a line directed towards the extension
of the straight line representing the observed
matches at wavelengths below 590 nm for which
AT’/t; had values smaller than unity. The mod-
ifications in stimulus conditions enable the ob-
served chromaticity point to be moved very near
but not quite up to this straight line extension. It
was assumed that, if the effect of the rods could
have been completely eliminated, the observed
point would have moved right onto the straight
line portion of the spectrum locus and a small
correction of the data was made to achieve this.
As can be seen from the published work of Stiles
and Burch (1959) and the account given in Sec-
tion 5.5.4, the necessary correction was, in fact,
very small.

Thus, the color-matching functions embodied
in the CIE 1964 supplementary standard colori-
metric observer aim to define the matching prop-
erties of the rod-suppressed retina for a large
visual field; that is, they are believed to represent
the functions a(A), b(A), c(A) used in Eq. 1(5.6.2).
Large-field color matches can be grouped un-
der four headings:
(a) Those for which the rod contributions of the
two stimuli are equal (AT’ = 0) and which should
therefore be identical with their evaluation from
the rod-suppressed color-matching functions. Such
matches should hold, irrespective of the level of
stimulation.
(b) Matches for which the difference of rod
response AT’ is smaller than the smallest percep-
tible difference 1; appropriate to the mean level of
rod response TJ, = (T;T;)'/*. From these, the
matches predicted by the rod-suppressed color-
matching functions should be at least just accep-
table in practice.
(c) Matches for which the ratio AT’/t, exceeds
unity but for which by an increase in stimulus
level, within practical bounds, this ratio could be
reduced to an acceptable value.
(d) Matches for which the ratio AT’/7, cannot
be reduced below unity at any reasonable stimu-
lus level.
The numerical procedure proposed for esti-
mating whether, and to what extent, the rods
contribute to color matching in a large visual field
is straightforward.
The (photopic) luminance L (cd- m7”) of a
stimulus {P, dA}, where P, is given in terms of
the spectral concentration of radiance L,, (W -
m ~*-sr_'), is defined by
_ (rod) vision.
bm ibPV(rX) dd [4(5.6.2)]
In this expression, V(A) denotes the spectral
luminous efficiency function and K,,, (= 683 Im -
W_') the maximum luminous efficacy of the CIE
standard photometric observer for photopic and
foveal vision (see Section 4.3).
The V(A) function is identical to the y(A)
function of the CIE 1931 standard colorimetric
observer that applies to foveal color matching. In
our case, we are concerned with large-field color
matching and thus, require the color-matching
functions X,9(A), Vio(A), Zip(A) Of the CIE 1964
Factors Modifying Color Matching ne)
supplementary standard colorimetric observer.
Two stimuli of different spectral distributions,
{P,, dA} # {P,, dX}, but identical tristimulus
values, X{q) = X{0, Yio = Yio, Zio = Zio,
match in color when viewed in a large field but
may have different luminances, L,~ Lj.
There is no official luminous efficiency func-
tion for large-field viewing, but we may stipulate
that the y,,.(A) function of the CIE 1964 supple-
mentary standard colorimetric observer defines
the desired luminous efficiency function applica-
ble to large-field viewing (see also Section 5.7.2).
Consequently, for the purpose of this exercise, we
define the (photopic) luminances of our stimuli
{P, dX} by
L* = Ki [PV) dd [5(5.6.2)]
where V*(A) = y,.(A). The value of the maxi-
mum luminous efficacy K* is not the same as
that of K,, in Eq. 4(5.6.2), but is not expected to
deviate much from it. Without serious loss of
accuracy we chose K* = K,, = 683 (Im- W_').
The modified definition of luminance, as given
by Eq. 5(5.6.2), assures that two metameric stimuli
{P,, dA} and {P,, dA}, viewed in a large field,
will have the same luminance.
The scotopic luminance L’ (cd-m *) of a
stimulus {P, dA} is defined by
ie Kn fPV) dd [6(5.6.2)]
In this expression, V’(A) denotes the spectral
luminous efficiency function and K,, (= 1700
Im - W_!) the maximum luminous efficacy of the
CIE standard photometric observer for scotopic
The photopic troland value 7* of a stimulus is
directly proportional to the photopic luminance
(cd-m 7) of the stimulus, and accordingly we
set
T* = pL* [7(5.6.2)]
where p is the pupil size at the eye given in mm”
(see Section 2.4.4).
Similarly, the scotopic troland value 7” is given
by
T’ = pL’ [8(5.6.2)]
The natural pupil of the eye varies with the
luminance of the stimulus. For an average eye,
358 VISUAL EQUIVALENCE AND VISUAL MATCHING
the relation between p and L is fairly well known
(see Section 2.4.5), and we employ the existing
relation by assuming it to be identical to that
between p and L*.
From Eqs. 5(5.6.2) to 8(5.6.2), it follows
_K’,[PV(A) da
Tar [9(5.6.2)]
K* [P,V*(X) da
In our analysis, we will either assume specific
values for T* or derive them from given values of
L* using Figure 1(2.4.5). The corresponding value
of T’ can then be calculated from Eq. 9(5.6.2)
provided the spectral distribution {P, dA} of the
stimulus is known in terms of some arbitrary
radiometric unit. Note that absolute radiometric
quantities for P, dA are not required, because
only the ratio of the two integrals in Eq. 9(5.6.2)
is needed. In accordance with international prac-
tice of expressing relative spectral distributions of
radiometric quantities, we may write P(A) instead
of P).
In the case of viewing an object-color stimulus,
it is often more convenient to replace P(A) by
B(A)S(A). The quantity B(A) is the spectral re-
flectance factor of the object measured under
specified illuminating and viewing conditions
(Section 3.3.7). The quantity S(A) denotes the
relative spectral radiant power distribution of the
source illuminating the object (Section 3.3.5).
Equation 9(5.6.2) can then be rewritten as
p= pe KnIBO)SO)V'O) dd
KE IB(AYS(A)V*(A)da
[10(5.6.2)]
which reduces to the convenient form of
[PERE
Yy’ K’fS(A)V'(A) da
Yio K*{S(X)V*(X) da
[11(5.6.2)]
where Y’ and Yj, are the scotopic and photopic
luminance factors, respectively, of the object-color
stimulus. They are defined by
Y= 1007FSV)dd
ay
— IB(A)SCQ)V’(A)
and
4A
¥9 = 109 BAISAIFio(A)
[S(A)Vi(A) ar
where Y,,. is obtained routinely when the tri-
stimulus values of the object-color stimulus are
calculated. The quotient
K! [S(A)V(A) da
K* [S(A)V*(A) dA
may be calculated for a series of different illumi-
nants S(A) commonly used in colorimetry and
vision research (Section 3.3.4). For S(A) of the
CIE standard illuminant D,;, the above quotient
is equal to 2.243, approximately.
The numerical examples given below refer to
pairs of stimuli each of which provides a rod-sup-
pressed color match for the CIE 1964 supplemen-
tary standard observer. The metameric spectral
distributions P,\(A) and P,(A) of each pair are
hypothetical functions. They were constructed for
the purpose of demonstrating the different types
of large-field color matches described earlier. An
attempt was made to construct stimulus functions
that resemble those found in visual colorimeters
employing mixtures of two or three monochro-
matic stimuli, and in colored objects illuminated
by daylight.
The relative spectral radiant power distribu-
tions of the metameric stimuli composed of two
or three monochromatic stimuli are given in Ta-
ble 1(5.6.2). The corresponding chromaticity coor-
dinates x9, ¥j9, and the ratios Y’/Y,, for each
pair are also given. Table 2(5.6.2) contains the
spectral reflectances of the hypothetical ' objects
that, when illuminated by CIE standard illumi-
nant D,;, provide metameric object-color stimuli
for the CIE 1964 supplementary standard ob-
server. The corresponding chromaticity coordi-
nates X19, ¥jo, and luminance factors Y,, and Y’
are also given.
The analysis of the pairs of metameric stimuli
is carried further in accordance with the numeri-
cal procedure outlined above. Table 3(5.6.2)
contains the results obtained for each pair. The
photopic troland values T assumed for each pair
ranged from a low value of 10 td to an optimum
value at which the threshold difference 1, be-
comes infinitely large. In accordance with Figure
1(7.5), this optimum value occurs when Tj reaches
the value of approximately 3500 scotopic tro-
lands.
We note that both the luminance (photopic
troland value T) of a given pair of metameric
stimuli as well as their relative spectral radiant
power distributions {P,(A) dA}, {P,(A) dA} are
the key parameters that decide whether the rods
 Factors Modifying Color Matching 359

Table 1(5.6.2) Pairs of Metameric Stimuli Composed of Two or Three Monochromatic Stimuli
d Pair 1 Pair 2 Pair 3 Pair4 Pair 5
(mm) PA) PA) PA) P(A) PA) PAA) PA) PA) P(A) P(A)
435 0.15510
445 0.00208 0.37376 0.02051 0.00001
480 1.00000
482 0.42308
525 0.09804 0.33449 0.99111 0.00025
533 0.47274
555 1.60000
588 0.827909
615 1.00000
625 1.00000
645 2.71928 0.35464 1.31679 0.02796
700 1.00000
X10 0.4690 0.4690 0.6712 0.6712 0.1584 0.1584 0.3742 0.3742 0.7097 0.7097
ao 0.3863 0.3863 0.3259 0.3259 0.2374 0.2374 0.6002 0.6002 0.2895 0.2895
Y/Y, 0.5255 0.5255 0.0253. 0.1927 2.6157 1.4548 +=0.4098 -~—«0..8730 (0.0608 0.0073

may contribute to the color match made in a large
field.
Pairs of stimuli can be found for which any
rod contribution has no effect on the color match.
Pair 1 is an example. The relative spectral radiant
power distributions {P,(A) dA} and {P,(A) dA}
of the two stimuli are such that the difference AT’
of the two scotopic values T; and T; is zero for all
values of T. In other words, Pair 1 is not only a
metameric match with respect to the cone mecha-
nism [Xj (A), VCoA)s Z10(A)], but also with re-
spect to the rod mechanism [V’(A)].
In general, the rod mechanism can be sup-
pressed by increasing the illuminance of the
stimuli to an appropriate level. When AT’ /r, falls
substantially below unity, we may assume that we
have reached the level at which the rod mecha-
nism can have only a negligible effect on the color
match. No contribution of the rod mechanism is
expected when AT’ /t, approaches zero because T,;
becomes infinitely large. This occurs at approxi-
mately 3500 scotopic trolands; the corresponding
photopic troland value T depends on the relative
spectral distributions of the stimuli involved. In
some cases, for example, Pair 3, the optimum T
value may be relatively low (~ 700 td). In other
cases, for example, Pair 5, the T value must be
extremely high (~ 60,000 td ~ 30,000 cd - m*)
which indeed may be prohibitive for safe viewing
of the stimuli. Figure 1(5.6.2) illustrates the spec-
tral radiant power distributions of the two stimuli
of Pair 5.
It is clear from the foregoing that the lumi-
nance and spectral distributions of the stimuli
enter in a far from simple manner in the evalua-
tion of potential rod intrusion in large-field color
matching. Whereas a value for the ratio AT’/t,
(which is appreciably below unity) indicates that
rod intrusion will not occur significantly, the pre-
cise nature and extent of the distortion of a color
match when the ratio exceeds unity are unde-
termined. Some attempts have been made to show
how the effect on specific matches might be
evaluated, but no safe and simple practical rule
has yet emerged.
Rod intrusion is more likely to occur where
the mixtures involved are made up of two or three
spectral lines or fairly narrow spectral bands. The
differences in the spectral distributions of
metameric stimuli of this kind are extreme as
compared with rather moderate differences ex-
hibited by metameric object-color stimuli com-
monly met with in practical color problems.
If we confine our consideration to metameric
object-color stimuli with small to moderate spec-
tral differences, it is conceivable that a luminance
level of not less than 1000 photopic trolands
(~ 125 cd - m *) may be sufficient to assure that
rod intrusion will be negligible or small.
It is worth noting that in “severe” cases, where
the ratio AT’/t, exceeds unity even when the
photopic level is 1000 photopic trolands or more,
the scotopic level will be comparatively low, as
for Pair 5 in the above examples. If such a pair
Table 2(5.6.2) Spectral Reflectances of Pairs of Objects That, when Illuminated by Daylight (D,;;),
Provide Metameric Object-Color Stimuli for the CIE 1964 Supplementary Standard Observer

rN Pair 6 Pair 7 Pair 8 Pair 9 Pair1l0

(nm) BA) BA(A) BA) — By(A) BA) B(A) BA) BA(A) BA) —Bo(A)
380 0.083 0.176 0.195 0.159 0.199 0.042 0.132 0.800 0.000 0.001
390 0.124 0.127 0.146 0.157 0.189 0.061 0.150 0.341 0.000 0.001
400 0.164 0.078 0.111 0.166 0.182 0.088 0.175 0.118 0.000 0.001
410 0.186 0.054 0.105 0.153 0.172 0.106 0.181 0.041 0.000 0.001
420 0.190 0.055 0.108 0.144 0.163 0.111 U2 pe email 1A Ue 0.000 0.001
430 0.195 0.057 OPTS moa OF122 0.148 0.110 0.170 0.060 0.000 0.001
440 O97 <O.072 0.134 0.096 ON32 0.109 0.161 0.092 0.000 0.001
450 0.203 0.090 0.138 0.086 0.117 0.108 0.150 0.134 0.000 0.001
460 0.199 0.141 0.132 0.086 0.100 0.109 ats Aefaso hg ig 0.000 0.001
470 0.176 0.250 0.113 0.098 0.086 0.112 0.117 0.204 0.000 0.000
480 0.126 0.432 0.082 0.134 0.068 0.116 0.094 0.232 0.001 0.000
490 0.048 0.692 0.042 0.201 0.045 0.121 0.062 0.248 0.001 0.000
500 0.000 0.918 0.011 0.309 0.018 0.128 OB 0r 0.256 0.004 0.000
510 0.093 0.894 0.040 0.377 0.001 0.141 0.011 0.240 0.020 0.000
520 0.293 0.628 0.121 0.388 0.000 0.158 0.009 0.210 0.043 0.000
530 0.459 0.343 0.158 0.474 0.019 0.164 0.025 0.148 0.042 0.000
540 0.549 0.144 0.220 0.465 0.064 0.160 0.074 0.084 0.032 0.005
0.583 0.028 0.370 G5 O27 O80 0:t2 U6" OT 0.020 0.027
0.563 0.000 0.555 0.056 0.356 0.153 0.390 0.143 0.012 0.043
0.506 0.400 0.695 0.000 0.531 0.207 Obl ="0.295 0.006 0.048
0.417 0.138 0.745 0.176 0.653 0.341 0.824 0.500 0.003 0.050
0.306 0.273 0.712 0.402 0.690 0.501 0.900 0.640 0.002 0.051
0.199 0.405 UDS9ms 0577 0.658 0.633 Lie elem LUMeteee, 0.001 0.052
0.128 0.494 0.403 0.697 0.591 0.731 0s aU 0.001 0.053
0.084 0.553 0.254 0.766 0.528 0.794 ps pede | Ne as OOS aes
0.061 0.585 0.147 — 0.817 0.485 0.830 0.221 0.728 0.007 0.054
0.043 0.614 0.091 0.849 0.455 0.853 0.146 0.732 0.077 0.054
0.031 0.632 0.069 0.869 0.346 0.870 0.116 0.736 0.489 0.055
0.033 0.633 0.053 0.891 0.417 0.887 0.099 = 0.724 0.822 0.055
0.059 0.602 0.042 0.906 0.411 0.900 0.094 0.714 0.880 0.055
0.092 0.562 0.030 0.917 0.406 0.916 0.087 0.699 0.815 0.055
0.132 0.514 0.019 0.929 0.397 0.934 0.080 0.679 0.741 0.055
0.154 0.490 0.012 0.925 0.389 0.948 0.069 0.679 0.669 0.055 -
0.154 0.499 0.009 0.925 D370 0.963 0.054 0.677 0.595 0.055
O32. ae.U3e 0.010 0.916 0.353 0.976 0.043 0.683 0.537 0.049
0.120 0.599 0.014 0.921 0.355 0.986 0.040 0.717 0.472: 0.054
0.148 0.537 0.018 0.889 0.318 0.992 0.024 0.734 0.418 0.054
0.206 0.473 0.006 0.888 0.297 0.997 C007 "0776 0.367 0.054
760 0.250 0.436 0.000 0.887 0.272 0.999 0.000 0.862 0.317 0.054
— X19 0.334 0.334 0.442 0.442 0.494 0.494 0.468 0.468 0.565 0.565
Y10 0.420 0.420 0.408 0.408 0.358 0.358 0.362 0.362 0.422 0.422
Yio 33,3 33.3 34.3 34.3 28.9 28.9 31.7 Sie! PES. 2.59
y’
23.1 42.9 15.0 25.4 9.0 14.2 Tie LoS 1.38 0.52

360
 Factors Modifying Color Matching 361

Table 3(5.6.2) Threshold Differences 7, and and forth between metamers and surround will be
Ratios AT’ /z, for Different Photopic Troland higher than the steady-state value appropriate to
Values T of the Metameric Pairs of Stimuli the metamers because of the time taken for the
Defined in Tables 1(5.6.2) and 2(5.6.2) scotopic mechanism to follow the changes. As a
result, the computed value of the ratio AT’/t1,
Pair TS 1, AT’ /1, will tend to overestimate the chance of rod intru-
sion. This mitigating factor will not, of course,
Any — 0.0
operate when, as in some colorimeters, the match-
2 10 0.41 10.5 ing field has a surround that is dark or of about
107 0.36 X 10 11.8 the same scotopic luminance as the matching
103 0.47 X 102 9.2 field.
tO 0.15 RiP 0.3
(ii) Maximal Rod Participation. The above dis-
2-104 00 0.0 cussion on rod intrusion in color matching may
3 10 1.00 X 10 3.0 be expanded by considering in a more general
107 3.63. 4710- 0.8 way all possible pairs of stimuli of given chro-
7-10 00 0.0 maticity and photopic troland value T that match
4 10 3.16 3.8 in the rod-suppressed condition, and by providing
102 372 a0 3.2 a quick method for estimating, in any particular
103 ean Te 0.1 case, the maximal rod intrusion as specified by
2-103 00 0.0 the ratio AT’/t, which may be referred to as the
index p of rod intrusion. The paper by Rodieck
5 10 0.13 10.5
(1976) answers the closely related question of
10? 17 11.6
“Which two lights that match for cones show the
103 1.1210 cm
greatest ratio for rods?”
104 0b xX 107 2.7 The analysis requires the calculation of the
6-104 00 0.0 ratio D[S(A), X19, Yio, Zi9] Of scotopic to pho-
6 10 4.47 3.0 topic troland values T’ and T of the stimulus with
102 6.31 < 10 2.1 a spectral power distribution {S(A) dA} and
103 3.63 X 104 0.0 (rod-suppressed) large-field tristimulus values
2 #10? 00 0.0 X05 Yio, Z10:
7 10 2.82 2.5 If ¢(Q) is the set of all spectral power distri-
10? 3.02 x 10 2.4 butions {S(A) dA} that yield the same point Q in
tristimulus space (X19, Yio, Zio), then any two
103 550><103 0.1
members of $(Q) observed through the same
a0" 00 0.0
pupil area will be in rod-suppressed match, but
8 10 1.91 22 will usually have different scotopic values. The
10? 1.82 X 10 223 difference is greatest for the two spectral power
10? 6 * 10° 0.3 distributions having, respectively, the maximal
4-10? 00 0.0 and minimal scotopic values. These extremes are
9 10 2.34 2.6 also extremes of the ratio D[S(A), X19, Yio, Zio]
10? 2.19 X 10 2.8 since T is the same for all members of $(Q).
103 2.88 X 10° 0.2 Thus, the stimuli of maximal and minimal 7” will
3 -10° 00 0.0 have the same relative spectral power distribu-
10 10 1.66 4.6 tions {S(A) dA} for all those points Q in tn-
stimulus space that correspond to stimuli with the
107 U5 XAG 54
same chromaticity point (Xj9, Vio), where
10? 107 x10" 0.7
5 -10° 00 0.0 Xo
X19—
X10 a Yio 25 Z\0

is in an achromatic surround (scotopic and photo- and

pic troland values about equal) of the same pho-
topic luminance as the metamers, the effective
scotopic adaptation level for an eye moving back
362 VISUAL EQUIVALENCE AND VISUAL MATCHING

Z 1.0
Z
Z
Stimulus 1 Ppa
ec
co
Z 0.1

2
Z 0.01
Stimulus 2 ————_~ Z
Z
Z
Z 0.001
g SQ)
]
S(A)
log

gg Z
Z
0.0001

gg Z
g Z
Z
0.00001

100 500
q 600: as
Z
Z
22
700
0.000001
Wavelength (A) nm

Fig. 1(5.6.2). Logarithms of relative spectral power distributions S(A) of two color stimuli
[Pair 5 of Table 1(5.6.2)] that match in color with respect to the CIE 1964 supplementary
standard observer when it is assumed that the responses of rod mechanism are suppressed. In
reality, however, there will be rod intrusion unless the stimuli are presented above a certain
luminance level at which the rods become saturated and incapable of detecting the scotopic
difference between the two stimuli. In the illustrated case, the stimuli must be at a luminance
level of approximately 30,000 cd-m7 (equivalent to approx. 60,000 photopic td) before rod
intrusion can be expected to be negligible.

For any trichromatic system, whatever the
shapes of the color-matching and scotopic func-
tions as continuous functions of wavelength, it
can be shown that the maximal and minimal
values of the ratio D for stimuli of a given chro-
maticity are obtained for relative spectral power
distributions consisting in each case of a mixture
of not more than three monochromatic compo-
nents. For the actual determination of the maxi-
mal and minimal spectral power distributions and
of the corresponding values of D, our calculations
have been made by the method of /inear program-
ming.'
distribution {S(A) dA} are calculated in accor-
dance with standard practice by using sums
instead of integrals (Section 3.3.8), after dividing
the visible spectrum into n wavelength intervals 6;
(i = | ton) of constant length AA. The problem
then reduces to finding sets of n values for
{S(A) dA}, that are solutions of three linear equa-
tions of order n for X19, Yi9, Z,9 and that simul-
taneously make the quantity
D[S(A), Xo Yio» Z10] = Ti hn ae
Tih K.
m

This method of calculation, already applied to

Ky SVOOL[SQ) 4],
n

other colorimetric problems discussed in Section

3.8.6, gives the desired values of D in a more l=

direct way than Rodieck’s (1976) method which,

in part, involved the construction and use of a
metal model of the (x19, yi9, v)-space where v =
V’/(X10 + Yio + Zjo).
To apply the method of linear programming,
the tristimulus values Xj , Yio, Zj9 and the
scotopic value V’ of a stimulus of spectral power
‘The cooperation of Dr. N. Ohta is acknowledged
here, who actually carried out the calculations for us in
1975 while he was a Research Associate with one of us
(GW) at NRC.
Km 3LFS) 4],
[12(5.6.2)]
maximal or minimal. As expected from the gen-
eral theoretical result, the solutions obtained make
[S(A)@A]; equal to zero except for, at most, three
intervals of 6;. In the calculation of the maximal
values (and for the minimal values corresponding
to chromaticities located in most regions of the
chromaticity diagram), two of the nonzero values

of [S(A) dA]; occur at adjacent intervals 5,. This
means that the solution represents a distribution
for which S(A) is zero except in two intervals, one
of length 6, the other of length 26. On repeating
the calculation with smaller and smaller intervals
5, (with consequent increase in n), in order to
improve the “resolution,” again two of the non-
zero intervals come out to be adjacent, and the
spectral power distribution remains a mixture of
two spectral bands with progressively diminishing
widths. This behavior is consistent with the con-
clusion (see also Rodieck, 1976) that the shapes of
the color-matching functions Xj9(A), Vio(A),
Zi9(A) and the scotopic luminous efficiency func-
tion V’(A) defining human visual properties are
such that in the limit, the maximal and minimal
spectral power distributions reduce to mixtures of
just two monochromatic components. However,
for some other conceivable examples of trichrom-
atism in which the four functions deviate in shape
in various ways from those valid for human vision
[see Figure 4(5.3.3)], the theoretical maximum of
three separated spectral components in a mixture
to give the maximal (or minimal) value of the
ratio D[S(A), X19, Yio, Z19], would be necessary.
For the present purpose, the maximal and
minimal values of the ratio V’/Y,) were calcu-
0.8
0.2
O
O O2
Fig. 2(5.6.2). Loci of constant maximal values of the ratio V’/Yj9 as function of (x49, ¥19)-
chromaticity coordinates.
Factors Modifying Color Matching 363
lated at a network of points in the (xj9, Vj9)-
chromaticity diagram, covering the domain of real
color stimuli. The sums used were based on a
subdivision of the spectrum into intervals of length
6 = 10 nm, except for some checks using smaller
intervals. Interpolation was then used to derive
families of loci of constant maximal and minimal
values of the ratio V’/Y,,) which are shown in
Figures 2(5.6.2) and 3(5.6.2). The constant factor
K,,,/Km = 2.489 [see Table 1(2.4.4)] must be
applied to the value of each locus to find the
corresponding constant value of the ratio
D[S(A), X105 Yio, Z19], Maximum or minimum.
Figure 4(5.6.2) gives similar loci of constant
values of the ratio
(Vis Yio) max
nt eee tae)
(V'/ Yio ete
a quantity that is also useful to know in certain
visual experiments. It is of interest to note that
the greatest possible value for this ratio occurs for
rod-suppressed matches at a chromaticity located
in the diagram at approximately x,) = 0.685 and
Yio = 0.299. For that chromaticity point, the ratio
Venax/ Vin Teaches the value 15.14. The corre-
sponding stimuli providing the maximal and
0.4 5406 0.8
364 VISUAL EQUIVALENCE AND VISUAL MATCHING
540
O 0.2
Fig. 3(5.6.2). Loci of constant minimal values of the ratio V’/Yj9 as function of (x19, ¥19)-
chromaticity coordinates.
minimal scotopic values are, respectively, the mix-
tures of two monochromatic components (500 nm
and 700 nm) for V;,,, and (410 nm and 630 nm)
for V,.,,. The above data as well as the loci shown
in Figures 2(5.6.2) to 4(5.6.2) agree very closely
with those obtained by Rodieck (1976).
The index of rod intrusion for the rod-sup-
pressed match between the two stimuli of the set
6(Q), mand M say, having respectively the maxi-
mal and minimal scotopic value, is given by p =
Pum> Where
ey OY elec
with 7, being the threshold difference for the
scotopic mechanism whose value depends on the
effective scotopic level
Te = (Tu Tn) = TDi Dy)!
(ee L/D) t/2
[14(5.6.2)]
as discussed above in connection with Eq. 3(5.6.2).
The quantity T refers to the photopic troland
value of the color match, dependent on the pupil
area as well as the point Q in (Xj, Yio, Zi9)-
560
580
600
620
& 650
SS $% 770 nm
eS 0.01
\ 0.03
yee
04 O06 0.8
Xo
tristimulus space, and Dy and D,, are abbrevia-
tions for the extreme values of the ratio
DISCHEMP IE
If the chromaticity point, defined by the point
Q(Xi0; Yio, Zi90) 18 known, Dy, and D,, can be
estimated by interpolation between the loci of
Figures 2(5.6.2) and 3(5.6.2). The value of 7, can
be read off the Aguilar—Stiles plot of log 7; against
Tp [see Figure 1(7.5)].
For any other pair of stimuli in the set $(Q),
say stimuli 1 and 2, with the same photopic
troland value 7, the index p,, will be given by the
identical formula given in Eq. 13(5.6.2), with Dy
and D,, replaced by D, and D,, it being assumed
that stimuli 1 and 2 are assigned so that
Dy aD, 2D, = De
It would be a convenient general rule if pj)
never exceeded p,,,, for a given chromaticity and
given photopic troland value 7. However, this
rule breaks down in some cases. The reduction in
the first term on the right-hand side of Eq.
13(5.6.2) may be more than counterbalanced by a
change in 7, due to a shift in the effective scotopic
level 75. It is readily shown that the maximum
value of the index p,, is either equal to py,,,

O 0.2
Fig. 4(5.6.2). Loci of constant values of the ratio V,,,, /Vein aS function of (x19, ¥49)-chro-
maticity coordinates.
or will be the value p,, = p,,, for a match bet-
ween the minimal spectral power distribution
{S(A) dA} and a distribution {S(A) dA} such
that D, satisfies the equation
|
dlog T. Z sured in scotopic trolands as defined in Eq.
dlog T Ti=T(D, Diy)? 1 = D,,/D,
[15(5.6.2)]
where
gO aie
Ph Rees
lest Ds /D; Lie Dy 37
Equation 15(5.6.2) presumes that the solution for
D, is not greater than D,,. If the solution exceeds
Dy; then py, is the maximum index. The actual
determination of the maximum index for any case
is best done by means of a graph displaying the
relationship expressed by Eq. 15(5.6.2).
(iii) Rod Participation and the Notion of the
Line Element. A rather different approach to
rod participation in color matching is to adopt
the ideas and notation introduced in the discus-
sion of inductive line elements presented in Sec-
tion 8.4.7. For the rod mechanism, the response
Factors Modifying Color Matching 365
0.4 0.6 0.8
Kio
difference to be detected is
AS=S,-—S, [16(5.6.2)]
which is proportional to the difference AT’ mea-
3(5.6.2). The sensitivity of detection is measured
by a difference threshold T;, equivalent to 7;
defined earlier, which depends on the rod re-
sponses S$, and S, and on surrounding or pre-
exposed adapting stimuli. The ratio AS/T¢,
equivalent to AT’/7,, determines the strength of
imbalance of the rod mechanism. Similarly, the
imbalance of the three cone mechanisms may be
represented by the ratios AF,/7,, AF,/Th,
AF;/T;, where the differences of response,
AF,, AF,, AF, are all linear combinations of AR
= R, — R,,AG=G, — G,, AB= B, — B, with
R,,G,, B, and R,, G, B, being the tristimulus
values of the two given stimuli determined with
respect to an appropriate set of primary stimuli.
The total mismatch Q is defined by the sums-
of-squares relation
BANS A ee el Aer Ar 4
o=(52| ACA lea ills
[17(5.6.2)]
 366 VISUAL EQUIVALENCE AND VISUAL MATCHING
which can also be regarded as an expression for
the inductive line element in the four-dimensional
(R, G, B, S)-space.
The crucial hypothesis is that making an ex-
trafoveal match corresponds to reducing the mis-
match Q to the minimum value possible with the
controls available to the observer. Clearly, if the
controls enable the matching stimuli to be ad-
justed so that the tristimulus and scotopic values
are equal in the two half-fields, which leads to
AR=AG=AB=AS=0, then Q can be re-
duced to its absolute minimum value, zero, and
the match is obviously the correct one. However,
in general, with a fixed test stimulus in one half,
and a variable mixture of three fixed primary
stimuli in the other, Q can be reduced only to a
minimum not equal to zero, and the hypothesis
asserts that this will be the match made. This
leaves unexplained how, if the minimum value of
QO exceeds about unity, the eye can accept the
minimum setting as a tolerable match.
Despite this difficulty, the theory has been
applied with partial success to explain the varia-
tion with stimulus magnitude of color matches in
a 10° matching field, variations which are cer-
tainly to be attributed to the participation of rod
vision (Stiles, 1955b). In this application, the
quantities g,, q2, q3 of three spectral primaries in
a match on a monochromatic stimulus of fixed
magnitude are the variables controlled by the
observer. The match values of g,, g, g3 are com-
puted by making Q a minimum; that is, by mak-
ing
eOnea
gs) tue)
== == O
0q, 04> 043
The result must depend on the forms assumed for
the terms in the line element defined by Eq.
17(5.6.2). The cone contributions (AF,/T,)’,
(A F,/T,)’,(AF,/T;)* were taken to be the three
terms of the Stiles line element (Section 8.4.4).
For the rod term (AS/T,)*, use was made of the
data by Aguilar and Stiles (1954) in the same
“manner as the term AT’/zt, was used above. For
the case of a monochromatic test stimulus in the
yellow region of the spectrum, the calculations
show that the main effect of reducing the quantity
gq, of the test stimulus from about 200 to 1
photopic trolands is to increase by about 30 times
the ratio q3/q4, where q; is the originally small
quantity of the blue primary in the match. This is
substantially the experimental result.
(iv) Tetrachromatic Color Matches. A number
of experimental studies on rod participation in
color matching specifically aim at balancing not
only the three cone responses but also the rod
responses evoked by the two stimuli {P,, dA}
and {P,, dA} presented in a large bipartite visual
field, such that the color match remains com-
pletely stable regardless of the level of stimula-
tion. The color-matching process then clearly de-
fines a four-dimensional system, and the color
matches are obtained by using four instead of
three primary stimuli. The results are ap-
propriately referred to as tetrachromatic color
matches.
Bongard, Smirnov, and Friedrich (1958) are
believed to be the first who studied tetrachro-
matic color matching. More recent and more sys-
tematic studies are those of Palmer (1970, 1972b,
1978, 1980), Trezona (1970, 1972, 1973a,b, 1974,
1976), Clarke (1973), and Clarke and Trezona
(1976).
The experimental method developed by
Trezona (1972) is of particular interest. Trezona
used a modified version of the NPL (Stiles) tri-
chromator which provided, in a 10°-bipartite field,
a test stimulus Q of (variable) wavelength A and
four primary stimuli R, Y,C,B of (fixed) wave-
lengths Ap = 644 nm, Ay = 588 nm, A, = 509
nm, and A, = 468 nm, respectively. All five
stimuli could be varied in radiant power from
zero to a level giving moderate retinal illumi-
nances. Although the test stimulus Q always ap-
peared in the upper half of the bipartite field,
only three or two of the four primary stimuli
appeared in the lower half, and the remaining one
or two were added to the upper half to “de-
saturate” the test stimulus. As in trichromatic
color matching (Section 3.2.1), desaturation of the
monochromatic test stimulus is necessary in order
to obtain a complete color match.
An iterative procedure is used to establish the
tetrachromatic color match for a given test
stimulus Q. First, an ordinary trichromatic color
match is made for Q set at a moderate luminance
level, the experimental stimulus level, by means of
the primary stimuli R, Y, and B. One of these
primaries serves to desaturate the test stimulus.
The luminance of the entire bipartite field is then
lowered by means of a rotating sector disk to a
level below the cone threshold so as to evoke only
rod responses. At that level, the bipartite field
appears colorless except for test stimuli in the far
red end of the spectrum which continue to exhibit
a red tinge even near the absolute threshold of
 10° visual field
in trichromatic
= color match
y)7,.p)g

Q 10° visual field

R“R.BB in tetrachromatic
color match

Photopic
Trolands
Primary
of
Stimuli
Ifor
Level

Level of Test Stimulus Q

(decreasing —= )
Fig. 5(5.6.2). Trichromatic and tetrachromatic color matches at four luminance levels. Test
stimulus Q of wavelength A 9 = 530 nm provides a retinal illuminance of 400 photopic td at
experimental level I. The low level of Q is 1/2160 of level I. The trichromatic and tetrachro-
matic color matches made at level I of the test stimulus Q require settings of the primary
stimuli that are given in the ordinate in terms of photopic trolands. When the level of Q is
reduced to 1/6, 1/67, and I/6°, the settings of the primary stimuli would reduce by
corresponding factors, provided that the proportionality law in color matching holds. Deviations
from the proportionality law are illustrated conveniently by plotting the experimentally de-
termined settings after they have been multiplied by the factor 6, 67, or 6°, depending on the
level of Q. Strict obedience to the proportionality law would place the plotted points on
horizontal lines. The trichrometric color matches whose settings are denoted by R®, Y¥®, B®
indicate failure of the proportionality law because the blue settings [B®] deviate substantially
from the horizontal line. At level I/6° the blue primary stimulus must be increased by more
than a factor of 20 as compared to that predicted by the proportionality law. The tetrachromatic
color matches, R®, Y,C®, B®, on the other hand, appear to obey the proportionality law
much more closely. The dip of B™ at the level 1/67 has been explained by Trezona to have an
“operational cause” and is not considered to be significant with regard to the validity of the
proportionality law. The inset in the graph shows the visual field configuration and the way the
test stimulus Q and primary stimuli R, Y,C,B are displayed (after Trezona, 1975).

367
368 VISUAL EQUIVALENCE AND VISUAL MATCHING

vision. Normally, the two halves of the visual
field, though generally colorless in appearance,
will not match in brightness indicating an imbal-
ance in the responses of the rods. The primary
stimulus C, whose wavelength A; is located near
the peak of the scotopic luminous efficiency func-
tion V’(A), is then introduced in the appropriate
half of the visual field to enable the observer to
produce a brightness match between the two
halves. When the original experimental stimulus
level is restored, the two halves generally show a
color difference that the observer eliminates by
making appropriate trichromatic adjustments on
the R, Y,B primary stimuli. The luminance level
of the field is then again lowered to the scotopic
level in order to check and, if necessary, readjust
the brightness of the two halves of the field by
means of C. The procedure of alternating between
the experimental luminance level and the low
level, and restoring at each level in turn the match
between the two halves of the field, eventually
leads to radiant power settings R®, ¥?, C%, B®
of the respective primary stimuli R, Y, C, B, which,
for the given test stimulus Q, are unique at both
levels. Typically, four iterations are sufficient to
reach the unique settings. With these settings, the
observer finds the two halves in complete match
at both the experimental and low level. In fact,
this tetrachromatic color match appears to persist
also at levels intermediate to the two levels at
which it was obtained. Figure 5(5.6.2) illustrates
tetrachromatic color matches in comparison with
trichromatic color matches at four luminance
levels for a test stimulus Q of wavelength A = 530
nm.
In repeating the procedure outlined above for
test stimuli of different wavelengths, Trezona de-
termined a complete set of tetrachromatic color-
matching functions for one observer. Figure
6(5.6.2) illustrates the results. Table 4(5.6.2) pro-
vides numerical values of smoothed tetrachro-
matic color-matching functions obtained by
Trezona (private communication, 1980) for two
observers and their mean.
For a specified level of stimulation, trichro-
matic color-matching functions can be deduced
from the tetrachromatic color-matching functions.
The tetrachromatic color-match equation may be
written as
O0Q= ROR + YOY + COC + BOB
[18(5.6.2)]
where Q is the given test stimulus of (variable)

Tetrachromatic color-matching functions Observer PMG

{ R(644nm) g

a Y (588nm)

sn
$

700

B(468nm)

Wavelength dX (nm) Wavelength A (nm)

Fig. 6(5.6.2). Tetrachromatic color-matching functions r(A), p(A), €(A), BA) de-
termined by Trezona (1976) for one observer (PGM). The values of these functions are given
in terms of the radiant power (arbitrary units) of the primary stimuli (R, Y,C,B) required to
match unit radiant power of the test stimulus of variable wavelength 2. In each instance, the
mean of three color matches is shown (solid dot) with a +2 standard-error bar of that mean.
 Factors Modifying Color Matching 369

Table 4(5.6.2) Smoothed Tetrachromatic Color-Matching Functions for Two Observers (PGM and WHM)
and Their Mean as Determined by Trezona. The Primary Stimuli R,Y,C,B Are Located at 2, = 664,
Ay = 588, Ac = 509, A, = 468 nm, Respectively. The Values in Italics are Numerically Interpolated between
Observed Values. (Kindly Supplied by P. W. Trezona in a Private Communication, 1980.)

d FOX) FA) e@(A) dO (A)

(nm) PGM WHM Mean PGM WHM Mean PGM WHM Mean PGM WHM Mean

400 — 0.34 0.12 =—O:13 Loe

Sinem Ub he 0.8300 5-0,50 Osmo te 0.20050. 05." 0397 —022" 7013 © -0.66 0.39
220a7 +0.36°0—1.31) :+0.84 Ds129 058 0337 O99" 0.66%" 0:42a0" 0.65 22" 9095
425 —0.44 0.14 — 0.24 . 0.80
ee O49 = 1.61 — 1.05 UG) S02 0449-82029 0 859" —0.58" 9 1203 1.56 1.30
440 —0.54 —1.53 —1.03 OMS OF ee O.Ave a 019 0855 — O59 F132 LTT 155
ates 039 es 09065 0.13 064 0.39 —0.26 —067 —0.46 1.43 Leys (bey
460 —0.02 0.08 0.03 0.01 0.33 O19 S002 et 0 25m Yi hot 24 1.54 ie39
470 TORS E04 OF) tite OL ON OR 0.15 0.16 O27 ECO)? BOOS EL Oye geOl
475 0.76 1.40 O08 0.208 26:37" 0,30 0.30 O51 0.40 0.86 0.76 0.81
480 O19, 4 01262 1 SOQ OZ2. = 0. 52 — 0.41 0.44 0.70 Olas O60 220 Dire 62
483 bi 1.68 M408 =0/33=0:58%— 0.46 0.53 0.80 0.665 057" 046 0.52
490 1.38 1.62 1.50 —0.44 —0.56 —0.50 0.73 0.97 0.35% 05335 }5 0.26 0.30
495 1.54 1.4] 148 —0.50 —0.49 —0.49 0.91 L.11 LOPES LO. £0016 ely
500 1.54 1.11 1.32 —0.50 —0.38 —0.44 1.07 Lg HIses010: ©" 008 0.09
505 1.26 0.70 OOS er 0i4 le 0 2382 — 0.32 1.17 1223 1.20 0.04 0.03 0.03
510 0.70 0.21 O60 = Or228= O02 0.12 APU 28) Le SO POOL SaREO!]
peomre— O10 =-0.83 (2=0:77 O88 @O0:465" (0.37 1.11 12 M620 05-0. 0592-00)
Soe 158 1.62 —1.60 0.64 0.85 0.74 1.00 1.08 1.04 —0.07 —0.05- =-0.06
Saree ae 212 2 5 Booey 1 Ole e 0.97, 0.86 0.83 0.84 —0.06 —0.05 —0.05
poveeer 215° 2.39 2:29 Rider 1.25 1.17 0.69 0.62 0.66;--0:05: =0.04 -—0.05
poem 21990-2379 2.28 1.29 8 1.44 1.36 0.46 0.42 0.44 —0.04 —0.03 —0.03
Bray 1.69 tlees ¥ — 176 Waves alkes, 1.42 0.27 0123 0 2 0.02 2-081 -002
eodn 0.66" —0.78 70.72 116 1.36 1.26 0.11 0.07 0:09 SAG, es]
590 57 S00. oO) J 075 S090 10 0.96 0.01 0 0.01 O 0 0
59) 3 1.20 1.16 O75 084) 0.80 20.027 s-0.0l 0.01 0 0 0
600 1.58 L772 1.65 0.60 0.68 0:64 0.03008; 0:02 fx0.02, 0.01 0 0
610 210 Zk 2.2) O33 039 "-0. 3691-2002 70.01 —O0l 0 ) 0
620 Ds: Be" 233 2.24 0.15 0.16 Ose O01) 20 = 0). 4) 0 0
630 1.89 2.02 195 0.03 0.01 0.02 0.01 0.01 0 le 0 ar) 0
640 1251 P39 LOOM 003 oe O06 — 0.05 0.02 702 a0 0 ai) sail
650 1.08 ting 5) Deel eet 06 0108 0.07 0.03 0.03 0:03, —0 Sh) ail!
660 0.67 0.69 0.68 —0.06 —0.06 —0.06 0.03 0.03 0:03 °—0 <4, A)
670 Oe lee 39) 2 38 A 00.05 0,035 0.02 0.02 O02 ae 0 seg f ale
680 0.18 0.17 Cy pee ee 2a 0.02 0.01 0.01 0.01 —O =a) =)
690 0.06 0.06 (E06 0.01 0 1)1 eed 0 0 ON ea 4) =)
700: 0.01 0.01 0.01 0 0 0 eat) nit an) 0 0 0)

wavelength A and R,Y,C,B are the primary
stimuli of fixed wavelengths Ap, Ay,Ac, Ag, Fe-
spectively. The factors Q, R®, ¥®, C, B® de-
note the amounts of the test stimulus and primary
stimuli, given, for example, in terms of radiance
(W-m ’-sr_'), determined by an appropriate
radiometric technique after the observer has
established the color match. Depending on the
wavelength A of the test stimulus, one or two of
the primary stimuli are required to “desaturate”
the test stimulus; their amounts are entered in Eq.
18(5.6.2) with a negative sign.
370 VISUAL EQUIVALENCE AND VISUAL MATCHING

For unit radiance of the test stimulus, that is, For unit radiance of C, we define
Q=1, the amounts of the primary stimuli re-
quired in the color match become RO Gras yO =
"aia ap he). Bates Daan
R® = yo fe
Folie lookin wee
(3) LS
=
etises b™ (Xe) [22(5.6.2)]
BO £
cE == b(r)
and obtain
[19(5.6.2)]
C= FO(AC)R + PO(AC)Y + 6O(AC)B
and Eq. 18(5.6.2) appears in the form
[23(5.6.2)]
Q = FO(AYR + PO(AYY + CO(AYC + BO(A)B
Inserting Eq. 23(5.6.2) into Eq. 20(5.6.2) results in
[20(5.6.2)]
a trichromatic color-match equation for the test
The quantities F(A), pO (A), EA), BMA) de- stimulus Q of unit radiance:
fine the tetrachromatic color-matching functions.
For a specified level of luminance, one of the Q =[F(A) + 2(A)FOAC)IR
four primary stimuli, typically C, is then color
matched in terms of the other three. The trichro- + [FA + OA)FOAC)IY
matic color-match equation may be written
+[ BO A) + (A)b (A) |B
CC = ROR + YOY + BOB
[21(5.6.2)] [24(5.6.2)]

Derived trichromatic color-matching functions

3r-
from CIE 1964
standard data
R (644nm) Si “in Y (588nm)

vy, aN
y(3) - y; <

7 (3) (@) @) “
be” 500 600 700
sl Wavelength dA (nm)

2 B (468nm)

? 500 600
Wavelength (nm)

-/~ from tetrachromatic

matches
= 7

Fig. 7(5.6.2). Trichromatic color-matching functions 7 (A), p©(A), b©(A), derived by

Trezona (1976) from her tetrachromatic color-matching functions shown in Figure 6(5.6.2) for
observer PMG. The trichromatic reduction equation that expresses the C primary stimulus in
terms of the primary stimuli R, Y, B for a specified luminance level was found experimentally to
be

C= —5.00R+ 1.43Y
+ 0.087B

The dashed curves are the trichromatic color-matching functions derived, by an appropriate
linear transformation, from the CIE 1964 standard observer data x 19(A), V19(A ),Z 19(A)-
 Factors Modifying Color Matching 371

deduced from the tetrachromatic color-matching
functions 7d), pA), EA), BMA). The
validity of the trichromatic color-matching func-
tions is limited to the extent to which changes in
the luminance level and the particular spectral
compositions of the stimuli displayed in the visual
field will not upset the balance of the responses of
the rod mechanism.
The trichromatic color-matching functions de-
duced by Trezona from her set of tetrachromatic
color-matching functions are illustrated in Figure
7(5.6.2) together with corresponding trichro-
matic color-matching functions obtained by an
appropriate linear transformation of the CIE 1964
standard observer data X;9(A), Vio(A), Zi9(A).-
In her investigation on tetrachromatic color
matching, Trezona (1973a,b) reported on two
rather surprising effects concerning (i) the
Maxwell spot when matching in large fields, and
(ii) the persistence of the color matching when
changing the field size. According to her observa-
tion (see also Trezona, 1970), the Maxwell spot,
which is quite vivid in a 10° bipartite field when
trichromatic matching is done for certain mono-
chromatic test stimuli (Section 3.3.2), is not visi-
ble in tetrachromatic color matches involving the
same test stimuli. Trezona further observed that a
tetrachromatic color match obtained in a 10°
field remained an acceptable color match when
the field was reduced to 1°20”.
Palmer (1978, 1980) disagrees with Trezona’s
(1974) observations and finds that the Maxwell
spot does not disappear completely when making
tetrachromatic color matches in a 10° field. His
experiments also indicate that the Maxwell spot
is, at least in part, responsible for residual failures
of the additivity law he finds in tetrachromatic
color matching.
(v) Extrafoveal Color Matches. Rod participa-
tion in color matching has also been investigated
in the extrafoveal region of the retina (Clarke,
1960b; 1963). Figure 8(5.6.2) from the work on
Clarke (1963) shows how the observed chromatic-
ities (W.D.W. system) for mixtures of two spec-
tral stimuli (the curved line) differ from those
predicted from the chromaticities of the pure
components assuming strict validity of the addi-
tivity law (straight line). The matches were made
with small test and comparison stimuli (each 40

B 0.2 0.4 0.6 0.8 1.0 B 0.2 0.4 0.6 0.8 1.0
450nm r 450nm r

Fig. 8(5.6.2). Defect of additivity in extrafoveal color-matching in measurements by Clarke

(1963). Chromaticities plotted in a modified W.D.W. system with monochromatic stimuli at
wavelengths 450, 500, and 650 nm, and normalizing wavelengths at 570 and 480 nm. The
measured chromaticities of mixtures in various proportions of wavelengths 430 and 530 nm
(lefthand diagram), and 480 and 650 nm (right-hand diagram), are shown as open circles. The
points marked on the straight lines joining the observed values for two extreme mixtures are
the chromaticities of intermediate mixtures derived on the assumption of additivity. The point
S represents the chromaticity of a “rod color,” or “scotopic white,” which enters in a possible
explanation of the failure of additivity in terms of rod participation.
372 VISUAL EQUIVALENCE AND VISUAL MATCHING
>< 80 minutes) exposed at 10° from the fovea in a
dark field. With large stimuli in a white surround,
the discrepancies are much reduced, the observed
points falling on the straight line but with small
displacements from the predicted positions.
Whether rod participation accounts for all the
observed additivity failures is not clear.
5.6.3 Location of Visual Field
Most color-matching experiments are made with
the visual test field, usually a bipartite field,
viewed by the central region of the retina covering
the fovea when the field has an angular subtense
not exceeding 2°, and including the parafoveal
region for fields of 10°. The CIE 1931 and 1964
standard colorimetric observers provide im-
portant examples of color-matching functions for
these viewing conditions. When the test field is
moved away from the central region of the retina,
neither of the two sets of standard color-matching
functions applies. Peripheral color vision, and in
particular, peripheral color matching, deviates sig-
nificantly from foveal and parafoveal color vision.
The deviations generally increase with the dis-
tance of the visual field from the center of the
retina.
Apart from effects of the ocular media (macu-
lar and lens pigments) and refractive defects,
which influence the periphery and the center dif-
ferently, it is mainly the variations in the neuro-
nal topography, both in the retina and in the
visual pathways, that govern the differing proper-
ties of central and peripheral color matching (see
Section 2.2). A useful recent review of peripheral
color vision has been given by Moreland (1972a).
There are three factors of particular impor-
tance in peripheral color-matching experiments.
Color matching in a specific peripheral retinal
location requires the use of a fixation target for
foveal viewing, whereas the test field itself is
imaged at the required peripheral location. Care-
ful adherence to fixation is important because
even small shifts in the location of the test field in
the periphery can result in marked upsets in the
color matches. The requirement of simultaneous
foveal fixation and peripheral color matching is a
task of divided attention which a naive observer
must practice in order to master.
With small test fields viewed steadily by the
peripheral retina, local adaptation occurs that is
often referred to as the Troxler effect (Troxler,
1804; Clarke, 1960a). This effect, though also
noticeable in the fovea (Clarke, 1957), is generally
more pronounced in the extrafoveal and periph-
eral regions of the retina. However, under certain
observing conditions, all involving voluntary but
strict fixation, strong foveal local adaptation ef-
fects can also be observed (McCree, 1960b;
Kaneko, 1965). The use of an appropriate pulse
and recovery cycle is often a necessary experi-
mental technique to obtain a reasonable precision
in peripheral color vision tasks (Moreland,
1955a,b; Clarke, 1960b).
In general, juxtaposed bipartite visual fields
cannot be used in peripheral color matching be-
cause visual acuity is rather poor in the periphery.
This is particularly so when the test field is small,
that is, less than 10°. The two halves of the test
field must then be separated sufficiently to pro-
vide a clear resolution between them [see part (f)
of Figure 1(5.2.5); Weale, 1951b; Moreland,
1955a,b; Clarke, 1960b].
Cruz and Moreland (1955) and Moreland and
Cruz (1959) conducted an extensive experimental
study of color matching involving different loca-
tions of the visual field in the retina. Their results
demonstrate the large variations that occur in
color matches made in foveal, extrafoveal, and
peripheral regions of the retina. They also demon-
strate that differences between observers can be
rather large for peripheral color matches.
In the study of Cruz and Moreland, the visual
field consisted of a test and comparison field each
subtending 80’ < 40’ and presented simulta-
neously for 0.5 sec every 2 sec to reduce and
control the Troxler effect. The comparison field
was centered at the fovea; the test field was
imaged on a specified location in the peripheral
retina. The surrounding field was kept dark. Color
matches between the two fields were made while
viewing a sequence of some 20 to 30 exposures to
the pulse /recovery cycle. ;
The main results, when plotted in a W.D.W.
type chromaticity diagram, showed a progressive
reduction in size and shape of the test field from
the fovea. Figure 1(5.6.3) illustrates the data for
observers JM and AC when the test field was
located 25° nasal from the fovea.
The experimental data of Cruz and Moreland
suggest that trichromatic vision in the central
region of the retina gradually reduces to dichro-
matic vision for a test field of moderate size and
given luminances and exposure times. These find-
ings are supported by other more recent experi-
mental data obtained by Boynton, Schafer, and
Neun (1964) and by Weitzman and Kinney (1969)
who used color-naming techniques to explore color
vision in the peripheral region of the retina.
 Factors Modifying Color Matching 373

Visual Field

540 ‘Test Comparison

550
A| AEN
|
1 A ' fixation
a CON aun dots

490

Fig. 1(5.6.3). Asymmetric color matches of peripheral test stimuli of wavelength A, located at
25° nasal from the fovea, expressed in a W.D.W. type chromaticity diagram for observers JM
and AC. The line through the corners (R),(G),(B) of the triangle represents the foveal
spectrum locus for each of the two observers and provides the reference frame for the
experiment. The point W represents the “white” stimulus (CIE standard illuminant C) in
foveal matching. When the test stimuli are presented in the peripheral test field, as indicated in
the inset, and matched by appropriate mixtures of the given primary stimuli R, G, B, chromatic-
ity points are obtained that result in the loci falling completely inside the triangles: open circles
for observer JM, solid dots for AC. The primary stimuli used in the experiment are at
wavelengths 650,530, 460 nm; the normalizing wavelengths used to obtain W.D.W. chromatic-
ity coordinates are at 494.0 and 582.5 nm. The peripheral test field was kept at constant
brightness for most wavelengths, corresponding to a foveal retinal illuminance of approximately
30 td for the mixture of the primary stimuli. (Moreland and Cruz, 1959).

5.6.4 Size of Visual Field functions of average normal trichromats making

The importance of field size to color matching
has already become evident in several previous
sections, notably those dealing with the develop-
ment of the CIE 1931 and 1964 standard colori-
metric observers that define the color-matching
color matches in small fields of 1° to 4° angular
subtense and large fields of greater than 4° angu-
lar subtense, respectively, when the visual field is
a juxtaposed bipartite field centered at the fovea
(Section 3.3.3). The differences between the two
standard observers are, to a large extent, caused
374 VISUAL EQUIVALENCE AND VISUAL MATCHING
by a decrease in the effective optical density of
the ocular media, mainly the macular pigment,
which occurs when going from the small field to
the large. Considerably less macular pigmentation
covers the parafoveal region of the retina. Rod
intrusion in the large-field data may also be a
contributing factor, but it is believed that this has
been virtually eliminated in the CIE 1964 stan-
dard data by correction procedures along the
lines discussed in Section 5.5.4.
When differences in the effective optical den-
sity of the ocular media and possible rod intru-
sion have been taken into account, small but
experimentally significant discrepancies between
2° and 10° color-matching functions remain. In
his pilot investigation of 2° and 10° color-match-
ing functions (Section 5.5.3), Stiles (1955b) made
a careful comparison of the two sets of data he
obtained for a group of 10 normal trichromats on
the NPL (Stiles) trichromator. He concluded that,
in the wavelength range 570 to 700 nm, the dif-
ferences between his two sets of data show fea-
tures not attributable to an ocular pigment effect.
In that region of the spectrum, the ocular media
are virtually nonselective (Section 2.4.6). Stiles
suggested that in going from 570 to 625 nm, the
“blue” mechanism in the parafovea (10° field)
might become relatively more sensitive compared
with the foveal “blue” mechanism, whereas for
the “green” and “red” mechanisms, a reverse
effect occurs.
Changes in the sensitivity of the cone mecha-
nisms with retinal location of the image of the
visual field have been observed by several other
investigators studying the spectral sensitivity or
the luminous efficiency of the peripheral and
foveal regions of the retina by a variety of differ-
ent techniques [see Sections 2.4.6 and 5.7.2(v)].
However, the agreement between the results ob-
tained by the different investigators is not entirely
satisfactory and it is difficult to extract from
them information particularly relevant to the case
at hand, that is, to changes in the size of centrally
viewed visual fields.
_ The field size has also an effect on the preci-
sion of color matching. Color-matching ellipsoids
(Section 5.4.2) obtained in large fields (10°) are
usually smaller, that is, the observer can repeat
color matches more precisely, than color-match-
ing ellipsoids obtained in small fields (2°) but
otherwise identical observing conditions. For field
sizes greater than 12°, little or no gain in preci-
sion is noted. For field sizes smaller than 2°, the
precision deteriorates rapidly and dichromatic
color-vision characteristics begin to become evi-
dent for very small fields of 20’ angular subtense
and less, when observations are made by strict
fixation.
Field-size effects have been studied a great
deal in connection with color discrimination, par-
ticularly wavelength discriminations, which will
be discussed in Section 7.10.2.
5.6.5 High Luminance Level
It is generally accepted that foveal trichromatic
matching obeys, at least to a close approximation,
the proportionality law (Sections 3.2.1 and 5.2)
over a wide range of photopic luminance levels.
Similarly, extrafoveal or large-field trichromatic
matching obeys the proportionality law if the rod
receptors do not contribute to the match, their
responses being either suppressed by a suffi-
ciently high luminance level or being equal for the
two spectrally different stimuli to be color
matched (Section 5.6.2).
In this section, we will discuss the failure of
the proportionality law for trichromatic color
matching when the color stimuli presented to the
observer reach high luminance levels. The discus-
sion is relevant to the von Knies law of the
persistence of color matches (Section 3.2.1).
Early work by Wright (1936; 1946) was con-
firmed and extended by the results of Brindley
(1953) who showed that the amounts of red (680
nm) and green (550 nm) required to match a
monochromatic test stimulus in the range 480 to
620 nm are changed materially following a 30-sec
exposure to a yellow adapting stimulus of about
30,000 photopic trolands [Figure 1(5.6.5)]. More-
over, intense adapting stimuli of other wave-
lengths all produced similar changes in this match,
namely a considerable increase of up to three
times in the amount of red and a rather small
decrease up to about 25% in the amount of green
[Figure 1(5.6.5)]. Terstiege’s (1967) experimental
data show similar trends.
More recent work by Wyszecki (1978), Alpern
(1979), and Wyszecki and Stiles (1980) provide
further data and in some ways, more comprehen-
sive quantitative evidence of the failure of the
proportionality law for trichromatic color match-
ing. The experimental technique used by these
investigators differs from the one employed in the
earlier work. The retinal illuminance, measured in
trolands, which controls the adaptation of the
observer’s eye, is provided by the luminance of
the bipartite visual field itself. Thus, a (metameric)
 Factors Modifying Color Matching —_ 375

R
and
G
Alog

Slog
Gand
R

—0.2 70 500 520 540 560 580 600 620 640 660 £680 700 0.5 1.0 15) 2.0
Test wavelength (nm) log (adapting luminance)

Fig. 1(5.6.5). Left-hand diagram: changes AlogR and AlogG in the logarithms of the
amounts of red (A = 680 nm) and green (A = 550 nm) in a match on test stimuli of various
wavelengths, following adaptation to a yellow stimulus (A = 578 nm) of high luminance (about
30,000 photopic td). Right-hand diagram: similar changes in a match on a yellow stimulus
(A = 580 nm), following adaptation to stimuli of various filter colors [deep red (DR), red(R),
orange (QO), yellow (Y), yellow-green (YG), blue-green (BG), blue (B), violet (V), and deep
violet (DV)], and various luminances. In both diagrams the heavy points with experimental
spread refer to Alog R, the light points with experimental spread to Alog G. The open circles in
the left-hand diagram are computed from Brindley’s theory of the phenomenon (Brindley,
1953).

color match is made first at a moderate level of
retinal illuminance, say 1000 td, with the eye
adapted to that level, and then the luminance
level of the color-matched bipartite field is raised
(anywhere between 1,000 and 100,000 td) without
requiring the observer to look elsewhere to adapt
to the higher level. Compared to the method used
in the earlier studies, in which adaptation to
higher levels was achieved by a 30-sec exposure
to a secondary stimulus, this method and the
restoration of the color match at higher levels
leads to more precise measurements of the
breakdown of the color matches with the eye’s
adaptation under better control.
In the experimental work reported by Wyszecki
(1978) and Wyszecki and Stiles (1980), the NRC
trichromator was used (an instrument whose basic
design is that of the NPL (Stiles) trichromator).
Both a large (9°) and a small (2°) bipartite field
were available and used, and all trichromatic color
matches were made against a fixed “white” (D;;)
reference stimulus presented in the upper half of
the bipartite field. This method, known as the
Maxwell method of color matching (Sections 5.3
and 5.6.6) has advantages over the more conven-
tional maximum saturation method because, at
color match, the visual field will always appear
“white” regardless of the wavelength of the test
stimulus that is presented as one of the three
already mixed primary stimuli in the lower half of
the bipartite field. The observer’s state of chro-
matic adpatation is thus kept constant throughout
the experimental study.
Examples of the results of Wyszecki’s (1978)
pilot study are given in Figures 2(5.6.5), 3(5.6.5),
and 4(5.6.5), which conveys a similar message of
breakdowns in color matches as that obtained
from Figure 1(5.6.5).
Figure 2(5.6.5) refers to large-field color
matches (9° field) where the test stimulus was
composed of monochromatic stimuli (L) =
17,500 cm~'!, (G) = 19,000 cm™', and (B)=
22,500 cm '. The color matches were made at
retinal illuminances of 1,000, 2,000, 4,000, 6,000,
8,000, 10,000, 12,000, 15,000, 25,000, 50,000, and
100,000 td. The radiant powers, measured in pW,
376 VISUAL EQUIVALENCE AND VISUAL MATCHING

RETINAL ILLUMINANCE ( TROLANDS)

| 4 4 6 10.) 155-25 50 100X10°
“O12
ees

=
=o
_| 2-04
wo
=
=
ica
L=17500 cm”!

1'

fe)
>fe)(e)
e))
@
G= 19000 cm!

' Nn

L,G,B,
stimulus
of
(»W,
rad.
power
, )(rad.
field
reference
in
stimulus
white
of
w»W,
power
,
b
LOG

-1.0 -0.5 0.0 0.5 1.0

LOG ( rad. power, »W, of white stimulus in reference field )

Fig. 2(5.6.5). Color matching with 9° bipartite field at retinal illuminances from 1,000 to
100,000 td. The test stimulus is composed of wavebands (L),(G),(B) located at 17,500,
19,000, and 22,500 cm ', respectively, and color matched with a white reference stimulus
composed of wavebands (R),(G),(B) located at 15,500, 19,000 and 22,500 cm ', respectively.
The radiant power, ».W, requirements of the mixed wavebands (L),(G),(B) in the test field,
do not all obey the proportionality law. Although the L and G curves deviate drastically from
the predicted horizontal curve between 8,000 and 50,000 td, the B curve remains horizontal
over the entire range of levels (Wyszecki, 1978).

required for the (L), (G), (B) wavebands
obtain, when mixed in the lower half of the field,
a color match with the white reference stimulus in
the upper half of the field, are found not to be
proportional to the retinal illuminance. To il-
lustrate the failure of proportionality, the loga-
rithm of the ratio of the radiant power of (L),
(G), and (B) in the test field to the radiant power
of the white stimulus in the reference field is
plotted against the logarithm of the radiant power
of the white stimulus in the reference field.
If proportionality would hold, the L,G, B
curves would be strictly horizontal in Figure
2(5.6.5). Although the B curve is virtually hori-
to zontal within 5%, the L and G curves are clearly
not. While the observer increases the radiant
power of the waveband (G) in the test field
beyond what the proportionality law would re-
quire, the radiant power of the test waveband (L)
is decreased. The breakdown of the proportional-
ity law sets in at approximately 8,000 td and
appears to be completed at approximately 50,000
td. Above 50,000 td, the proportionality law ap-
pears to hold again.
The magnitude of the breakdown is quite large.
Above 50,000 td, the radiant power (G) must be
approximately three times higher than strict pro-
portionality would predict on the basis of the
 Factors Modifying Color Matching 377

RETINAL ILLUMINANCE ( TROLANDS)

| 2 3 5 10 18 30 45 65 100 200x103

|
.——*x
LS
)field
in
test L=17500 cm°!

B=22500 cm"!

G=19000 cm"!
stimulus
of
L,G,B,

)white
field
reference
in
stimulus
of
nW,

(rad.
pW,
power,
(rad.
power
,

|LOG

wee ee
-2.5 -2.0 =f25 =a -0.5 0.0
LOG ( rad. power, wnW, of white stimulus in reference field )

Fig. 3(5.6.5). Color matching with 2° bipartite field at retinal illuminances from 1,000 to
200,000 td. Similar, but somewhat more pronounced failures of proportionality in the radiant
power requirements of the (Z) and (G) wavebands in the test field are noticed here as
compared to color matching with a 9° bipartite field [Figure 2(5.6.5)].

matches made below 8,000 td. Simultaneously,
the radiant power of (L) must be approximately
1.5 times lower than strict proportionality would
predict.
Figure 3(5.6.5) shows that similar results are
obtained for small-field color matching (2°).
Again, the breakdown of color matches occurs at
approximately 8,000 td, and after approximately
50,000 td, no further breakdown is observed. The
magnitude of the breakdown appears somewhat
larger than for the large-field case. The radiant
power requirements for the waveband (G) in-
creases by a factor of four above the prediction
by strict proportionality. (L) decreases by a fac-
tor of three, whereas (B) is again virtually pro-
portional to the level of stimulus.
When the (variable) waveband (L) in the test
stimulus is located in the middle to short-wave
region of the spectrum, breakdowns of color
matches with increasing levels of retinal illumi-
nance are much less pronounced. In fact, it ap-
pears as if breakdowns do not occur even at levels
as high as 100,000 td.
Figure 4(5.6.5) is an example of such a case.
The test waveband (L) is at 20,750 cm|. All
other conditions are identical to those given for
Figure 2(5.6.5). Only a slight suggestion of a
breakdown is indicated in the G curve which
bulges slightly in the 8,000 to 100,000 td region. It
is not certain whether the slight upswings in the L
curve at the 75,000 and 100,000 td levels are
significant. When ignoring these last two levels,
the curve is, as before, horizontal.
The experimental procedure described above
was used to determine complete sets of Maxwell
spectral distribution functions, at different levels
of retinal illuminance, by varying (L) over the
entire range of the visible spectrum. Figure 5(5.6.5)
illustrates the results obtained by observer GW in
a 2° bipartite visual field for two levels of retinal
illuminance: 1,000 td and 100,000 td. The numeri-
cal values on which Figure 5(5.6.5) is based are
given in Table 1(5.6.5).
The results of the color-matching experiment
by means of the Maxwell method, as illustrated in
Figure 5(5.6.5) and tabulated in Table 1(5.6.5),
378 VISUAL EQUIVALENCE AND VISUAL MATCHING

RETINAL ILLUMINANCE ( TROLANDS)

6 10 (5 25 50 100x10°
0.2
=e]

so)
@
-
= 0.0
”
en
=
—_ os aoe X X=—X —— x —— KX x — X—
' ° a)
R=15500 cm"!

fe) BS ante ee ee oe
L=20750 cm’!

)field
,nW,(in
reference
rad.
of
stimulus
pW,
white
R,G,L,
rad.
power power,
(stimulus
of
G=19000 cm"!
}
|LoG

-1.0 -0.5 0.0 0.5 1.0

LOG ( rad. power, pW, of white stimulus in reference field )

Fig. 4(5.6.5). Color matching with 9° bipartite field at retinal illuminances form 1,000 to
100,000 td. The test stimulus is composed of wavebands (R),(G),(L) located at 15,500,
19,000, and 20,750 cm', respectively, and color matched with a white reference stimulus
composed of wavebands (R ),(G),(B) located at 15,500, 19,000, and 22,500 cm ', respectively.
Only a slight failure of proportionality in the radiant power requirement of the (G) waveband
in the test stimulus is indicated between 8,000 and 100,000 td. The other two wavebands (R))
and (B) follow proportionality law predictions closely.

can be converted to the familiar color-matching respectively, presented in the upper half of the
functions r(A), g(A), b(A). The following equa- bipartite field providing the fixed “white” refer-
tions are used: ence stimulus W. The quantities L(A), R(A),
G(A), B(A) are the radiant powers of L, R,G,B

Dares eC
= as Ray R(A) stimuli presented in the lower half of the bipartite
field and required by the observer to obtain a
color match with the given juxtaposed “white”
reference stimulus. The stimulus L varies in wave-
2(A) oe [1(5.6.5)]
length A and is referred to as the test stimulus; its
amount L(A) never becomes zero, whereas
R(A), G(A), B(A) are respectively zero in the
long-, middle-, and short-wavelength regions of
the spectrum. Thus, for any given test stimulus L,
In the above equations, Ry, Gy, By are the only two of the available three stimuli R,G,B
radiant powers of the primary stimuli R,G,B, need to be added to L to obtain the desired
 Factors Modifying Color Matching 379

WAVELENGTH \ (nm)
400 450 500 550 600 650 700
dre) R 250
1000 Trolands
—-0-0-0-0- | QO00 Trolands

2.0 200

100

IOOO
FOR
NANOWATTS
TROLANDS
IN
VALUES
TRISTIMULUS
MAXWELL
\/
25.24 23 22 21 20 i9 18 I7 16 45 14-103 TROLANDS
|OOO0O0
FOR
NANOWATTS
IN
VALUES
TRISTIMULUS
MAXWELL
WAVENUMBER (cm’)
Fig. 5(5.6.5). Maxwell spectral distribution functions L(A), R(A), G(A), and B(A) obtained
by observer GW in a 2° bipartite visual field for two levels of retinal illuminance: lines refer to
data obtained at 1,000 td, thin lines through open circles refer to data obtained at 100,000 td.
The reference “white” stimulus consists of an appropriate mixture of the three instrumental
primary stimuli R,G,B to give the chromaticity of CIE standard illuminant D,, (average
daylight). At the 1,000 td level the radiant powers of R,G,B are R ,, = 1.0856, G,, = 0.5567, and
B,, = 0.3516 nW, respectively. At the 100,000 td level these radiant powers are increased by a
factor of 100, precisely.

trichromatic color match with W. When L has a
wavelength identical to R, G, B, the two halves of
the visual field become physically identical.
Table 2(5.6.5) gives the color-matching func-
tions 7(A), Z(A), (A) derived by means of Eq.
1(5.6.5) from the values of Table 1(5.6.5). Also
given are the corresponding chromaticity coordi-
nates. Figures 6(5.6.5) and 7(5.6.5) illustrate the
data.
The widely favored physiological interpre-
tation of the breakdowns of color matches at high
luminance levels, as demonstrated in Figures
1(5.6.5) to 7(5.6.5), is that of bleaching of the
photopigments in the cone receptors (Wright,
1936; 1946; Brindley, 1953; 1960; Terstiege, 1967;
Alpern, 1979; Wyszecki and Stiles, 1980). The
bleaching hypothesis will be discussed in Section
8.2.6.
5.6.6 Maxwell Method versus
Maximum Saturation Method
There are two main methods that have been used
in color-matching experiments: the maximum
saturation method and the Maxwell method. Fig-
ure 1(5.6.6) gives the parameters that are essential
to the two methods. A bipartite field is usually
used in either method. In the illustrated case, the
field has a horizontal dividing line. There are
three fixed primary stimuli R, G, B which are usu-
ally monochromatic stimuli consisting of narrow
bands of wavelength A, one in the red, one in the
green, and one in the blue part of the spectrum.
Then there is a variable test stimulus L, also
monochromatic but of variable wavelength A, and
finally, a fixed reference stimulus W, usually of a
“white” chromaticity.
Table 1(5.6.5) Maxwell Spectral Distribution Functions for Retinal Illuminances of
1,000 and 100,000 td, Obtained by Observer GW in a 2° Bipartite Visual Field with a
Fixed “White” (D,;) Reference Stimulus’
m Maxwell Values for 1,000 td Maxwell Values for 100,000 td
(cm ') Lim) Rim) Gm) Bim) Lyoolm) Riool™) Groom) Bioo(m)

14,500 12.616 0:5 73enei3 360608568 56.6 35.8

14,750 6.412 O:512 7 0856 665.2 56.4 25.2
1D,000R TS 3t215 209 2 E0205 348.1 56.4 35.4
15,250 1.749 0.566 0.353 169.4 56.1 BInpe)
15,500 1.086 52) caetela DA 108.6 Sauk 35.2
15, 750° | 02730 0.542 0.348 67.6 55.0 35.0
16,000 0.525 0.521 0.345 47.4 54.0 34.7
16,250 0.435 0.494 0.344 36.6 52.4 34.5
16,5008 7 03399 0.455 0.344 30.8 50.1 34.3
16,750 0.406 0.400 0.347 All 47.4 34.2
17,000 0.437 0.32225 U.to2 24.3 441] 34.1
17,2503 40501 0210 — 0/358 28.8 39.6 34.1
17,500 0.616 0.000 0.044 0.365 33.0 33,1 34.4
17,7504 (0579, 0.2397 S060" G5) 41.3 Pa ap 35.0
18,000 0.530 0.541 0.366 55.0 0.0 4.8 35.6
18,2507, "0500; 10,717 0.364 55.0 30:5 0.0 35.9
18,500 0.499 0.860 0.361 52.0 63.8 35.9
18,7503. O17" *0.986 0.358 53.4 89.2 Sa)
19,000 0.557 1.086 0.352 55.7. 108.6 eh
T9200 wee UOLae nial Oo 0.341 OU ata 34.2
BS Os Us)Sale 2 0.319 JU ete Lo gt
19,750 0.885 1.334 0.285 85.9. 142.2 28.0
20,000 bA2Z3mwli4t9 02 16 116.5 154.1 20.0
20,250 1.449 1.508 0.000 0.086 158.8 167.6 0.0 @)5,
20,500 1.260 1.427 0.150 0.000 124.2 150.8 20.4 0.0
20750- “O25 5 lore Usa! aise pau)selaje'd 34.0
21,000 0.628 1.188 0.408 6i.dag AZS8 43.4
21,250. 0)¢.0:520, 4lil4deu 0.463 58Hae lil927, 48.7
21,500 0.465 1.122 0.500 46.7 116.1 51.4
DATS AVAZ Ze be 413 ., 113-5 we
22000 70390 OO IPG 539 39 lesarldt5 54.2
22,250 0.364 1.090 0.549 867778010939 55.0
225900 OSS 204071 OS6RUIO SST, 39.2) 921086 ee
225750 "7203674 SNO820920563 B7.3:¢ 81070 559
23,000 0.401 1.079 0.568 40.7 105.6 56.0
23,250 0.474 1.076 0.570 49.0 104.0 56.0
23,500 — 625 4 1 O71 SO 57) 63:5... 1027.8 56.0
23,750 0.846 1.064 0.570 86.5 101.6 oy)
24,000 1.374 1.058 0.569 141.1 100.2 55.8
2425027 12.235 , A105 020569 230.0 98.8 DSi7
24,500 5.709 1.045 0.568 585.0 ov 55.8

“m=10'/X when A is given in nanometers (nm). All Maxwell values are given in
nanowatts (nW). The “white” (D,;) reference stimulus has tristimulus values R ,,= 1.086
nW, Gy = 0.557 nW, By, = 0.352 nW to provide 1,000 td; the tristimulus values of the
“white” (D,;) reference stimulus are increased by the factor 100 for measurements at
100,000 td.
 (wu $800°0—
1L00°0— 0000°0
7900°0—
S€00'0— 88Z0'0 60L1°0
pS€70861€'0
ESIr'0 6LtL'0
0000'I pS£9'l
SdCUIPIOOD AyONeUIOIYD
1870'0 L906'I SIOZ
uOnNnCLOsSIg
aouasajoy (°C) .AWUM», POX & WIM Plaly [ensia aywaedig 7 & Ul MAD) J2AIasqQ Aq poutesqgQ (S°9'S)T AIqBL Jo suopoung

ls
(P1 000‘001)

(1 E101
SOlo'|0600'I 0000'I
€S00'I 67960
Epr6 9€78'0 0£89°0
E£9L'0 L98S°0 96S7'0
0000°0ErVeLO—
CEST L6TTT—
jl
[eyseds

0
I—
+
(UL 7000'0—
90000—
8000 0000°0
11000— P8700
76100 rr00'o0
ZZ00'0 0000°06LP0O0
8710 SOIOO—
€800':0— E300 Z88I'O
]JJOMXBYA,

Oe
suoTOUN,| BUTYDeP[-IOJOD

0— 0—
~=—r
WO1f

oon L000'0—OzZ000—
I1000— 0000d'0
7Z2000— 9880°0
96L1°0 O€I0'l OZPrO'loddd'l €76L'0
€0I18°08PrZ6'0
£€689°0
(Pp? 000‘00T)

1879°0 LOS
E'0
PsALlIog]

mal
1 0080°0
ZE9I‘0
6IIE0 0000'I
LOP9‘0 8796'7
elaaid's 69797OSL6'I
EL6CT'E ELT’ LI9E00000°0 gcseo—
O%6E0—OTLEO—
YOO
So}VUIPIOO-)

rP} 00000T PUL PI OOO'T :2eouBUTUMITT FEUEY Jo sjaaoy] OM], ye snnUINS

—-SayeUTPIOOD AyONeWOIyD

(um 0Ssl0'0— 0000°0

L110$800'0—
Ivl0'0— 7S80°0
Irs0'0 9ETE0 71S1019°0 09680
rLirv'0 0000’ LSIE'I
8L8r'l 8b6E'l
)'3
AWIYVUOIYD

0— 0 I
(P} 0001)

(w)'4 $6101 OOI0'I

O810'IISIO'I 0000'I 06£6'0
7806'0 9€0r'0pesto
1v6S°0Z10S°0
84890 0000°0 E%6EO—
6b990—Z6SOI—
pue
suonpouny

(u)'qro0d0'0— 00000
60000—
9000 I1000— r8100
E8100 SIZOO— pL7@00—
S9~O'O0O—8rc0O— 00000 bbb00
8II00— 6PrL0'0 O€81'0
suoTounN, 8uTYyo}ePl-JO[OD

0—
SUTYIIVJA[-10]OD

+¥z000—
(m)'3 €1000—6€000—o00ddo0
7zS000— 8€rl'0
LOS 07960OISO'l
O€E80 Y9EIT'I 69L0'10000'I IZ9L0
16790 Zr8E0
(P} 000‘1)

05
= §
(w)'4 09800 O00001
90790
£6910E€LEEO OL6h7%
6b €69€1P8Z0'l
O€9L'I 99€EL'0O€61°0
0000°0 Z7L7Z'O—
§=©©ZI8Z'0— 816Z70—
7e
(S°9°S)Z

ce
(,;_wH9)
00s‘rI
OSL‘rIO00'SI OOos‘SsI
OSZ‘SI 0G]
OSZ‘9I OSL‘LI00081
OOS‘LI OSL‘6I 0SZ‘0Z
OSZ'8I OSL‘8I000‘61 00S‘6I
age

wu

~§=3=—-
9¢609'I
ZOL0'0
L700'0
9€10°0
€£86'0
17660
€900'0
OSL‘SI
#£SL8rI
90200
Lv00'0 LO086'0
S%6TC
1S€0'0
10100
1Z£0'0
€796'0
OS10'0
000‘9I
¥890°7%
06900
OZ7100 ICIG'E
01670
p9600
090L8'0
8SZ1
C1760
66900
OSL‘9I
6I1L97
O98€0
O11I00
6LLI‘0
0086'€
r6£00
1ZZ8'0
09680
1$60°0 607ZL'0
17980
SILOT
97100—
p8ss'0
76rr'0
L167'0
O00S‘8I1
91Sr'0
P9II'l
r6l00—
=6-—dabBES'O——«C‘LSB'Z~—Ss«OOO'LI
0000;
977%r'0
S99OL'E
pSe0'0
O€b7'0
ZI9L‘0
17980
8/Z1'0
OSZLI
O891'%
07690
SII00—
ZLSS0 ec
(ecco
SEI60
99100
lLol.t
0SZ‘6I
£9€1'0—
0ELIEO—
6£67
ZTELLO—
09060
L100 OCZIO01€870
O16EO0—
O8Zr'0
86Z1'0
9€08I—
8P07'7
8/7%0—
OLbS'I
000‘0Z
7L67O—
SS6r'0 OOPEO—
EL6vpI—
£5260
Ir6l80—
80ST
00S‘0Z
60L70—
v2Z7E0
I16L70
OPgz0
9S6€0
€66T0—
OI880—
OSL‘0Z
991L'0
709°
OL870O
OOLpO—
€I1IP'O0
vEPrz'0
€€77'0—
L

381
 panuyuod — (S'9'°S)T AGL

382
suoTOUN, BUTYDeP[-1O[OD SoJCUIPIOOD AyIONeWOIyD
(Pp?
000001) (P}
000001)
sajeuIpIoOD AyONeWIOIyD
(P?
000‘1) (P
OOO)
suotjoUN,{ SUTyd}ePl-IO[OD
ul
(,
und) (w)'a )'8 (1 (u)'q (uw)! oe
(ula OESCt>2 Ue af GS IC hi ee
jOOls COOL )lZ(wu
00017 O£€91'0— 99€70 96550 pLSTO— 9ELEO 09STO— E€E€8I'0 7ZETSO €g9S0— 69060
OSZ‘IZ OPII'O— LOSIO p9L90 reslO— ZEb70 08070— £6710 8rS90 II9E0— tb77'0
00S‘IZ O8L00— O7CI'O0 €9SL0 SL600— v¥zst0 O1I910— £7600 ZESLO TSETO— 8rel‘0
OSL‘IZ ZOSO'O-— LrLoo LZE8O 985c¢00— Z7LE0O OSIIT'O0— £5900 POrs80 86r10— 6780°0
2060
Ltrb00
6670°0—
€8680
IL800—
LEv0'0
97E00—-
L800
OSLOO—
00077
9ZLE00
ZZ OS7 O€I00— OIZ@O0 11960 Es100— SIzZ00 O9E00— ZLIO0 68560 E8E00— E8100
00S‘? 00000 000070 I 0000’ 0000°0 0000°0 0000°0 0000°0 0000'I 0000°0 0000°0
OSL‘7Z 98000 8ZLI00— 16560 0600°0 0— L810 7700 LSOOO— cr60 I€h00 6S00°0—
89180
LSIO'O
ILZOO—
8/L00
¢800'0—
78100
O00°EZ
O€980
€1E0'O—
77200
6L000—
€IpL'0
10700
vlL7@O0—
OS
SLILO
LvIl‘0
7800°0—
€L70'0
ELE0'0—
EZ
17260'0
89000—
pcz~O0—
p79S0
6E700
96660
‘0
Trl
7800'0—
E7PO'O
L6£0°0—
71600
O00S*EZ
7S000—
OSL‘EZ 8SZ00 7ZSIOO0— SSIvO 66S0'0 9S€0'0— 800OL L™OO'O—_——
+= L90v'0 OL9I‘0 ¢S00'0—
0007 £0700 €6000— 09670 09L0°0 6r£0'0— S'0HH 0—_—s’—«sO10§«©=6 T6p>T70 Z£61°0 Z£00'0—
PZ O0S7 €S100 95000— ELSI'O 61600 8€£0'0— ¥7Zr0'0 10000— 67S1°0 ILIZ0 — 9000°0
00S‘rZ ZLO000 Oz000— 91900 SLOI‘O 00£0°0— 76100 7zoodo0— 10900 6770 — ¢Z00'0

dL, ZUTYD}CU-IO[OO
SUOTIOUNS
(u)4 ‘(uw)3
q (uw) ore poute}go
:WIOIJ

=(“)- —“u] (“)7/[(“)u

=(“)8 (“)7/[(“)9—“9]
49 —“a]=(%) (w)T/[(“)a
aay “(w)T “(w)y)D “(WwW)g (Wi are oy} [[amxeJ[ fejoads UOTNQLYsSIp‘suoT}oUNJ
purey ““ ““D “g are ay} sn[NUIT}st)
Son[eA
JO
aq} ..d1TyM,,
(59) sdUaIaTaI “sNyNUINs
dU, AjLOT}eUTOIYS SaxJeUTPIOOD
“(W)4
)s (Wmoe PouTe}gO
‘WOIJ

(HM)
= 9+(H)84+(H)4]/(H)4
[(#)
3 (wu)
= 8 +(H)4]/(m)
9+(4)8
[(“)
 Factors Modifying Color Matching 383

i are tall oe
35 Observer GW
2°-Bipartite Field
Maxwell Method
30 (Des - White Reference)

1000 Trolands --——--

25 100000 Trolands --——--

N (e)

LS

Values
Tristimulus

fe)

05

24
oOee
23 (B) 22
ee
2l (G)
Wavenumber
is]
m(cm-!)
7
an 0
I6 (R) 15:10

Fig. 6(5.6.5). Color-matching functions 7(A),Z(A),5(A) derived from Maxwell spectral

distribution functions obtained by observer GW with a 2° bipartite field at retinal illuminances
of 1,000 and 100,000 td [see Figure 5(5.6.5)].

The maximum saturation method shown on
the left of Figure 1(5.6.6) is the most commonly
used method. The test stimulus L, of magnitude
L(A), presented in the lower half of the bipartite
field, is “desaturated” by the minimum amount
required of one of the three primary stimuli
R, G, B such that the mixture of the test stimulus
and the desaturating primary can be color
matched by an appropriate mixture of the remain-
ing two primary stimuli appearing in the upper
half of the bipartite field. The choice of the
desaturating primary depends solely on the wave-
length of the test stimulus relative to the wave-
lengths of the primaries. In the case illustrated,
the test stimulus L is desaturated by the B primary
so that the mixture [L(A)L + B(A)B] color
matches the mixture [R(A)R + G(A)G]. when the
amount of the desaturating B primary is very
small, it is sometimes helpful to add a small
amount of the B primary to the upper half of the
bipartite field and then take as the match the
difference in the amounts of the B primaries in
the two halves.
The amounts R(A), G(A), B(A) of the three
primary stimuli required in the match are called
the tristimulus values of the test stimulus L(A)L.
The amount of the desaturating primary stimulus
is given a negative sign.
Apart from differences at test stimuli of short
wavelengths consistent with different absorptions
in the eye lens and macular pigment, the respec-
tive color matches of normal trichromatic ob-
servers agree rather closely, leading to similar
tristimulus values R(A), G(A), and B(A) for a
given stimulus L(A)L. From such data, the
familiar color-matching functions r(A), g(A),
384 VISUAL EQUIVALENCE AND VISUAL MATCHING

Observer
:GW (1977)
AGE ie]
(20 000) 1000 TWolands —«—«—e—
° 100000 Trolands
2° Bipartite Field
Maxwell Method (Dgs5-White)

20000
dee SS
eo:
(19500)
G 1.5 |
g

os
19500
(20 500) °o

20 500°

18 500
ie
a =f (18 500)

(21000 SS
)

(21 oe
(22 000)-S.\4

; 0.5 22500\"="%uP 500

(24500) 0.5 (16 500) 10

Fig. 7(5.6.5). Spectrum loci derived from Maxwell spectral distribution functions obtained by
observer GW with a 2° bipartite field at retinal illuminances of 1,000 and 100,000 td.

b(A) emerge:
: alta)
Ao Say)
g(A) = a [1(5.6.6)]
LI) + BAB aS

The Maxwell method shown on the right of

MAXIMUM SATURATION MAXWELL METHOD Figure 1(5.6.6) employs a fixed reference stimulus
METHOD W, usually of a “white” chromaticity, appearing
Fig. 1(5.6.6). Color matching in a bipartite visual in the upper half of the bipartite field. In the
field using the maximum-saturation method or the lower half, a mixture of the test stimulus L and
Maxwell method. R,G, B = Primary stimuli (of two of the available three primary stimuli (R, G,
fixed wavelengths Ap, AG, Ag); R(A), GA), and B) is presented, which when appropriate
B(A) = amounts of primary stimuli (tristimulus amounts of them are chosen, color matches the
values (e.g., in terms of radiance: W - m~? - sr~!); reference stimulus. The wavelength of the test
L = variable test stimulus of wavelength A; L(A) stimulus determines which two of the three
= amount of variable test stimulus (e.g., in terms primaries must be added to the test stimulus. In
of radiance: W-m 7- sr_'); W = fixed refer- the example illustrated, the stimulus WW color
ence stimulus (usually of “white” chromaticity); matches the mixture [R(A)R + G(A)G + L(A)L].
W = amount of fixed reference stimulus (e.g., in As in the maximum saturation method, normal
cd - m ” or troland, td). trichromatic observers agree closely with their

respective color matches, leading to similar set-
tings of L(A), R(A), G(A), BCA).
The results of the color-matching experiment
by means of the Maxwell method can be con-
verted to the familiar color-matching functions
r(A), Z(A), b(A) by equations given in the previ-
ous section as Eq. 1(5.6.5).
It is clear that if the basic laws of trichromatic
color matching, particularly the proportionality
and additivity laws, hold strictly, the color-match-
ing functions r(A), g(A), D(A) derived from
Maxwell matches in accordance with Eq. 1(5.6.5)
must (within the experimental uncertainties) be
identical to those derived more directly by means
of the maximum saturation method, that is, Eq.
1(5.6.6).
Early tests of the additivity law by Blottiau
(1947) in foveal rod-free areas revealed failures;
but then, in a more critical investigation, Trezona
(1953, 1954) found that the failures of the additiv-
ity law are not considerable in comparison with
just-discriminable color differences. In fact,
Trezona concluded that the observed discrepan-
cies are, in part, associated with the technique
used by the observer for arriving at the actual
match settings within the range of visual uncer-
tainty.
aii nen (10°)
Maxwell Method
-oe-oes Maximum -
Ke) eT Saturation
Method
00
~ -05 00 0.5 1.0
{—_
Fig. 2(5.6.6). Spectrum loci derived from color
matches made in a 10° bipartite field by observer
BHC by means of the Maxwell method against a
“white” (CIE standard illuminant C) reference
stimulus and by means of the maximum saturation
method. The data points are means of three com-
plete sets of readings. The primary stimuli
(R),(G),(B) have wavelengths 650, 530, and 460
nm, respectively (Crawford, 1965).
Factors Modifying Color Matching 385
However, more recent experimental work by
Crawford (1965), Lozano and Palmer (1968), and
Wyszecki (given below) points at visually small
but systematic differences between the two sets of
color-matching functions obtained by the Maxwell
method and the maximum saturation method.
An example of Crawford’s (1965) data is given
in Figure 2(5.6.6) which shows the spectrum loci
obtained from color matches made by the two
methods. The two loci deviate from one another
particularly in the blue-green region of the spec-
trum. Five other observers engaged by Crawford
in the same investigation gave similar results,
though the degree of deviation varied somewhat
from observer to observer. Crawford also studied
the effect of field size, by making corresponding
color matches for 2° and 1° bipartite fields. The
results were similar to those obtained for the 10°
field, but, in general, somewhat reduced in the
magnitude of the deviation.
Results of another experiment by Lozano and
Palmer (1968) are shown in Figure 3(5.6.6). Al-
though the observing conditions and the system
Ne.
Obs.: RDL (10°)
{20000 -o— Maxwell Method
—x—Maximum-Saturation
Method
\,, 18500
*s, 18000
\ ®,_!7000
Fig. 3(5.6.6). Spectrum loci derived from color
matches made in a 10° bipartite field by observer
1.5 RDL by means of the Maxwell method against a
“white” (NPL Bluish White; Stiles, 1955) refer-
ence stimulus and by means of the maximum-
saturation method. The retinal illuminance of the
visual field was of the order of 160 td. For the
purpose of this illustration, the original data, sup-
plied by Dr. D. A. Palmer in a private communica-
tion, have been smoothed somewhat and then
transformed to a _ set of primary stimuli
(R),(G),(B) of wavenumbers 15,500, 19,000,
22,500 cm', respectively.
386 VISUAL EQUIVALENCE AND VISUAL MATCHING

Blue Corner

_ 500

20 500

0.5 18 OOO
Observer GW (1978)
9° Bipartite Field (IOOOtd)
Maxwell Method (D¢s) --—o—o=
Des 17500
Max-Saturation Method --—e—e=

17000

16 500
22 500
16 000
15 500
Sp pp fp $$
-45 -.0 -0.5 . 0.5 (R) 1LO\.
2300034/e—
500 24 500 gp oeies 14 500

Fig. 4(5.6.6). Spectrum loci derived from color matches made in a 9° bipartite field by
observer GW by means of the Maxwell method against a “white” (CIE standard illuminant
Ds) reference stimulus and by the maximum-saturation method. The retinal illuminance of the
visual field was of the order of 1000 td. The primary stimuli (R),(G),(B) are of wavenumbers
15,500, 19,000, and 22,500 cm‘, respectively.

of primary stimuli used by these investigators
differ somewhat from those used by Crawford,
similar deviations are noted between the data
obtained by the Maxwell method and those ob-
tained by the maximum saturation method. The
magnitude of the effect appears larger for ob-
server RDL than BHC.
Figures 4(5.6.6) and 5(5.6.6) are pilot data
obtained by Wyszecki in 1977 and 1978 on the
NRC trichromator for 9° and 2° bipartite visual
fields at an illuminance level of 1000 td. The data
show similar features to those illustrated in the
previous Figures 2(5.5.6) and 3(5.5.6). However,
Crawford’s (1965) finding, that the effect becomes
less pronounced when the field size is reduced, is
not confirmed throughout the spectrum. In fact,
Wyszecki’s observations indicate that the effect
can be somewhat larger at some wavelengths and
smaller at others.
All investigators, whose results are illustrated
in Figures 2(5.6.6) to 5(5.6.6) report that the
points of a spectrum locus derived from color
matches made by the Maxwell method against a
“white” reference stimulus are, in general, consid-
erably less certain than those derived from color
matches made by the maximum saturation
method. The uncertainty in a “white” match is
unfavorably amplified when the match is trans-
formed to predict the corresponding point on the
spectrum locus. Although an appropriate analysis
 0.04
Blue Corner

9+0,02
22 500
/ 002 004 006 008 O10 O12 O14
i

Observer GW (1978)
2°-Bipartite Field (1000 td)
Maxwell Method (Des) -o==o=mom
Max.-Saturation Method=e=-e==e=

2
22 500
wa
(B) >24 500

Fig. 5(5.6.6). Spectrum loci derived in a similar manner as those shown in Figure 4(5.6.6),
but the observations were now made in a 2° bipartite field. The numerical values for the plotted
points are given in Table 1(5.6.6).

2.0 (-2,-2,-2)

19 Se abe 4 O22)

18 =e
17 /”'5(-2,-2,-2)
r 7(-2,-2,-2) | go 2-9)
16
6(-2,-2,-2)
15

1.4

1.3 Max.- Sat. I9500“.

.2
-|.4 -|.2 -|.0 -0.8 -06 -0.4
1 ame

Fig. 6(5.6.6). Portion of (r,g)-chromaticity diagram showing spectrum loci derived from
Maxwell color matches against a D,.-“ white” reference stimulus and from color matches made
by the maximum-saturation method. For the monochromatic test stimulus of wavenumber
m = 20,000 cm~! (=500 nm) chromaticity uncertainty gamuts are shown that result from
+2% perturbations of the observed mean color matches. In each case, eight match perturba-
tions are made which, in tristimulus space, correspond to the eight corner points (1 to 8) of a
cube centered at the point (0) representing the observed mean color match. For details, see
text.

387
388
FIGUT,
| (9°9°S) SuTyd}VP-10][0D
sUOnIUNY
pue AjIOHeUIOIYD soajeuIp100DpoulejqQG
Aq JeAIaSqC
MAXUl & ,Z aynsedig
yensi, ppalyAq 9y) uMUTTxey, uUOHemeSpos pue oy) [JOMxepA, powoyl(%q) ye v juvjsuoy[eda]
Jo [eunay
souvulUInyyy
jo Ajajeurxoiddy
COOOL»P}

uNnuITxeyy UoneInyespoyoypy
wu BUTyOyeP-IJO[OD
suoTOUN. AjIONeWOIYD sajyeUIpIOOD SUTYDeP[-IOJOD
suOoT}OUN, AJONVUIOIYD SdJCUIPIOOD
WD) (,_ (m)4
= (m)8 (m)q )a (wu )s (uw (u)a (mje[axe
pompoyy (w)q (um) (u)s
00S‘rI £6400 60000— 0000°0 PeSulgl Ce SLO OLL00 60000— 00000 Fcl0t VA00—
OSL'rI SE9I'0 LIO00— 00000 SOl0l SOlIOO— £€9S1'0 ZLIO00— 00000 U—eeUllut
Vite
00ST SIEEO €€000— 0000°0 OOI0T OOIOO— €0I€0 Of£000— 00000 L600'1 L6000—
O0SZ'SI 90L50 8Z000— 00000 66001 6v000— 9L8S°0 ctOo00— 00000 $S00'1 $s000—
O0ds’sI 00001 0000°0 0000°0 I 0000 0000°0 0000'1 0000°0 00000 I 0000 0000°0
OSL'SI S6EST 7000'0—ZL710°0
0 0¢66 €800°0 610S'I £€800°0 00000 0 $766 ¢S00°0
00091 8r60'7 1100'0—0870'0
0 1826 0 v7C0 S100°¢ ILt0°0 00000 81860 100c8
9I OST L86€7 £9010 7c00'0— $8560 ¢cr0'0 7c v6IS O9IT°0 00000 09°60 Ovr0'0
00S‘9I
+ 9099°7 €v00'0—L907'°0
0 £6¢6 0 CCLO C6ELC 0 L607 00000 0 6876 I1Z0°0
OSL'9I Lys97 0 Leve c€900°0— 0 L838 oO LvIl Chole 0 ILS 00000 LS88°0 evi10
O00LI €6L57 0 0—Iv7cS 1600 9£€8°0 76910 6S19°C 0—7c7S'0ve00 0 Sves 99910
LI OST 98E1'7 0 0—€£89 vcl0 CIE) 0 CEPT LOT? 85900600'0— 0 6cLL 0 COET
O0S‘LI
1 6PSL' 0 0—67¢b8S10 96L9'°0 0 v9ITE 6618'1 C66L'0
0— cELO 78690 0 L90E
OSL‘LI 690E1 0 0—8966 9810 0 Lv8s 0 9¢cr | [cre 0—¥%768'°0L910 1$09°0 0 vcor
00081 £7860 L0cO'0—L870'1Ov6r'0 €91S0 I CEPO 0—1£00'1L810 0 SrIs Lv6v'0
O0S7'8I 6LL90 0—S60'I
170 0 CLBE 1$c9'0 €999'0 0—cO0v0'l
$810: I16€0 66190
00s‘81 rivr'O I 0—LSOIL
L610 06870 0 6€cL 0 6SCcV 1 €9100—PrS0
0 606¢ 0 cOCL
SOSL'81 LWICO LCl00—IPL0'16S91°0 [pvs'0 80170 0—9990'I
6010 16910 L680
 389
‘76LT70 = "8 pure Cpps 0=“ are poyjour
[JOMxePY 9y} Aq opeUr soydieUT IO[OS dy} UL pasn snnuINs soUaIOJOI .,OTYM,,-°°q IY} JO SayeuIpIOOS AyIONeUIOIYD JUL,
8660.04 O1Sb0 09900 6c000— 9600°0 UITO0 ss Bt Sel 0 16€0'0 LOOO'O— sSs000 00S'rz
SEN Op +69L1-0 VEEl 0 CLS00 0] =PL100 L8100— S8010 E~OI'0 +I17000— 77100 SZ PZ
PIL0 0 sc cul.0 68Sc0 I0100— +8700 c61IO0— 76800 p6L10 LEOO'O— LINO 0002
Iee00— cL3800 9IIv0 prlIOO— 6Lt00 OLIOO— _ €TILO0O S8LE0 OS00'0— OIZ00 OSL'EZ
00C0O0— 61400 cO8S 0 c8l00— 9€h00 IvI0OO0— 6€S00 6L0P'0 0900'0— 67200 00S°EZ
LS~wOO— LSS00 cE9L 0 cOCOO0— 8th00 SOIOO— Se0'0 99550 0900°0— ZOO SZ EZ
86100— £600 0995.07, SLIOO0— LPrr00 69000— 98100 OreL'0 IS000— 8E€10'0 000°E7
OUTO'D=. 390700 8cS60 90I00— 86100 LcOOO— £8000 79880 z7000— vL000 OSL‘?
0000°0 00000 0000'I 0000°0 0000°0 0000°0 0000°0 00001 0000°0 0000°0 00S°7Z
0ve0 0 L8100— 9£960 €€c00 1810 0— L1t00 Lt GOR L8€60 96700 8rE0'O— OSC
L8S00 6170 0— £9680 Stso0 c8t0 0— 9€90°0 9SLO00= v7680 19500 L990'0— 000°7Z
ceOl 0 CLLO) = C9180 12800 1090: 0— OSOI'0 LOL Or 8L78°0 LL800 0001'0— OSL*IZ
LS9T 0 vee Lo) — 8LPL0 OLEI0 0880 0— 98S1°0 ELLE OR O18L0 sIzlo Z9EL‘O— OOS*IZ
cese0 9661 0— 6cv9'0 CELIO POEL = 69L7 0 UES Os poeovo SELIO eLELO—.087 I
etpr 0 868e 0— vlOSO 6vec0 $907 0— ccS¢'0 8S0rv 0— cSOS'0 + 96EC0 OSIZ'O— 00017
6769 0 DoE 9.0 = 6L9EO Y9O0LTO O87c 0— 0cLI9'0 bcs Ose 008£0 66670 9S€7O— OSL'0Z
Ors 0 LE38 0=> 85970 Ove 0 py3c O— 9L96 0 YLLO 0— tS9CD LIVE0 ZESTO— 00S°07
ISIV I cOcOT— Pee LOLy 086c 0 vEcc | 19SL 0— CL810 66¢70 LS9TO— OS7'0Z
6£89 | 900 T= LyOlO L80¢°0 cei Os 8Ler | Lire Oe O€clO L6cS0 Ev870— 000°0Z
€8£9 | EE6L0 — c8S00 L6190 666¢ 0— 6Lcr | VL8S 0— 6£L00 v199°0 IZL70— OSL‘6I
161v I ply 0— 66000 OOLL0O tes UF 980€ | EL EEO 90700 ISI180 8ZETO— 00°61
PICGH 86£7 0 — 9€100 88680 SOLE LSvl OS9T O— OSIO00 L168°0 p871'0— 0S7'61
0000 I 0000°0 00000 00001 0000°0 0000 I 0000°0 00000 0000'1 0000°0 00°61
390 VISUAL EQUIVALENCE AND VISUAL MATCHING
of the error propagation has not as yet been
made, the problem is appreciated readily by con-
sidering the following numerical example.
Let us consider color matches involving the
test stimulus L of wavenumber m = 20,000 cm !.
The mean Maxwell values obtained by observer
GW against the D,;-“white” reference stimulus
W are, in this case, as follows:
Ry = 1.427, Ly = 1.095,
By perturbing the above mean observed values
by +2%, we generate a set of eight points
(R,, L;, B;), with i= 1 to 8, which represent the
eight corners of a cube centered at the point
(Ry, Lo, By) in (R, L, B) space, as shown in
Figure 6(5.6.6). These +2% perturbations are
consistent, in a rough approximation, with the
uncertainties found in the settings made by the
observer in repeat color matches. Of course, in
reality, actual repeat color matches would lead to
an ellipsoid and not to a cube (see Section 5.4) in
(R, L, B)-space, but for the purpose of this dis-
cussion, we will dispense with the refinement.
The mean and the eight perturbed color
matches (R,, L;, B;) (i& O0§1,!
transformed to the tristimulus values 7{™),
gz’, bp” of the given test stimulus L (20,000
cm') by means of
Re -eR:
i; D is changed, the resulting perturbation gamuts in
9 G,
Bew a8; 1
b;
(M) we
Z The fact that the (r, g)-chromaticity diagrams,
which, in fact, are the values of the color-match-
ing functions for m = 20,000 cm™! in terms of
the instrumental primary stimuli R, G, B of fixed
wavenumbers mp = 15,500, mg = 19,000, mz =
22,500 cm~!, respectively, as given in Table
1(5.6.6). Equation 2(5.6.6) corresponds to Eq.
1(5.6.5) introduced earlier.
The calculated values 7,, 2,,b, are then con-
verted to the chromaticity coordinates r,, g; by the
familiar equations
(M) ae
FO + BOO + BE
=(M ~(M M
5(M)
(M) — : different wavelengths is difficult to assess quanti-
FOO
BO + 5m
The results are plotted in Figure 6(5.6.6) which
shows an elongated convex four-sided polygon
approximately centered at the observed mean
chromaticity point (0). The cube corners whose
(R,, L;, B;)-coordinates differ only in L; have
identical chromaticity coordinates (1,, g;).
The mean maximum saturation values ob-
tained by observer GW for the same test stimulus
L (20,000 cm~') in terms of the primary stimuli
By = 0.2374 R, G, B are, in accordance with Table 1(5.6.6), as
follows:
7S) = 0.2843, 6S) = 0.5297,
BSS) = 0.1230
If the above mean observed values are perturbed
by +2%, a similar perturbation cube is obtained
as for the observed Maxwell values. When trans-
forming the mean and the eight perturbed color
matches (75), gz), b{)(i = 0,1,...,8) to corre-
sponding chromaticity coordinates r{5), g‘°), the
results illustrated in Figure 6(5.6.6) are obtained.
The emerging polygon is, in this case, consider-
4,8) are then ably smaller than the one based on perturbed
Maxwell matches. The cube corners (1) and (7)
have identical chromaticity points and coincide
with the chromaticity point of the mean (0).
When the wavenumber of the test stimulus L
the chromaticity diagram for the Maxwell matches
and the maximum saturation matches exhibit, in
general, similar features to those shown in Figure
6(5.6.6).
in which the spectrum loci in Figures 2(5.6.6) to
5(5.6.6) are plotted, are perceptually nonuniform,
adds to the problem of assessing the visual signifi-
cance of the deviations between corresponding
loci. A given linear distance Ad plotted in differ-
ent parts of the chromaticity diagram will most
certainly correspond to quite different perceived
color differences Ac. Unfortunately, the relation-
ship between Ad and Ac as a function of the
location of Ad is not immediately obvious (Sec-
tions 3.3.9, 7.10.5, and 8.4).
When the color-matching functions are plotted
instead of the spectrum loci, the deviations be-
tween corresponding sets of data reveal them-
selves in a different perspective, though again, the
visual significance of the deviations noted for
tatively. Figure 7(5.6.6) shows the color-matching
functions obtained by observer GW of the
[3(5.6.6)] Maxwell method and the maximum saturation
 a

(w)®
oot

001

oso
oso

(+)
Y
of
4,

oro
oro

oro
soo
SON|DA sNinwiysiay

SONIOA sNinWysiaL
SONIA sNNWIYsIIL
(-)
100 100
100
M9
4eAs08q0 (8261)
so00
sooo 2 4001Q-
O41,
4 PIO!
OOO!
Spudjol,
\ (-) —e—e—e— WNWIKDW UOUDINJOSPOUL
(99g)
poyew
jemxDy)
—o—o—o—
02
(9)

(8) 02 (9) ¢Ol-Sl aod 8) 02 (9) ¢Ol-Sl o) iso'0 (8)

(4)

100'0
4:aford)
¢Ol-Si

(j-Wd) W sequINUeADM (j-Wd) W sequnueADM

“SI *(9°9°S)L suTypyeul-10]0F)
()-W9)

suoyouny pourejqo
Aq JoArasqo
AA)Ul & 7 aysediq
ppelyAq ayy
[JAMxEyA] poyjoUrSUISN
& .,2}TYA,,(£9q) sdUeI0Jo1 snus pue Aq oy) uonvINjes-UMUTxeul
seQuNUeADM

‘poyjour
sy], [eUye! sdULUTUINTT
JO dy} [VNSIA Poly Sv JO ay) Jopso
JO OOOT‘P} sy] ArewLid
RUINS (g7)‘(.D)‘(Yy)
oavy StoqunuasaRM
‘OOS’ST ‘O00’6T
PUL QOS‘ZZ_UWd
‘, ‘Ajaaysedsar
pue
Jay) stunoie pojsn{pe
0} JURIpes S1aMO0d
JO J “AAU dU], [VOLIOUNUSonyeA
105 ay) pajjojd sjutod
ore UaAIS
Ul F1qGR], ‘(9'9O'S)T

39]
ays VISUAL EQUIVALENCE AND VISUAL MATCHING
method. From these functions, the spectrum loci
shown in Figure 5(5.6.6) were calculated. Table
1(5.6.6) gives the corresponding numerical values.
Taking into account the uncertainties in the
observations between corresponding color
matches made by the two different methods, we
are inclined to consider the deviations small from
the experimental point of view but not insignifi-
cant. The fact that all investigators independently
obtain similar deviations, particularly with regard
to their direction in the (r, g)-chromaticity dia-
gram, is a strong argument in favor of considering
the deviations to be significant. However, to our
knowledge, no definitive physiological model has
emerged as yet which would predict the devia-
tions quantitatively.
The deviations are appropriately termed fail-
ures of the additivity law of color matching in a
bipartite field (see Section 5.2.2 on asymmetric
matching procedures). Suggestions have been
made with regard to possible causes of these
failures.- They include chromatic adaptation
(Crawford, 1965), Maxwell spot (Palmer, 1980),
and interactions or linkages between different
cone mechanisms (Ingling and Drum, 1973), but
further work is obviously needed to resolve the
conundrum.
5.7 LUMINOUS EFFICIENCY FUNCTIONS OF
NORMAL TRICHROMATS
Luminous efficiency functions, which have also
been called /uminosity functions or sensitivity func-
tions, form the basis of our present photometric
procedures (Chapter 4). They also play a funda-
mental role in our efforts to understand the visual
mechanism that governs our perception of visual
stimuli and our ability to discriminate between
stimuli of different spectral radiant power distri-
butions presented in a variety of display situa-
tions.
A luminous efficiency function gives the ratio
of the radiant power of a monochromatic stimu-
lus of wavelength A,,, to that of a monochromatic
stimulus of wavelength A, when the two mono-
chromatic stimuli produce luminous sensations
that under given observing conditions and in
accordance with some specified criterion are
judged to be equivalent or the same, A,, being
chosen so that the maximum value of this ratio is
unity.
The criteria embodied in the visual judgments
on which luminous efficiency functions are based
vary widely in complexity. Some of these criteria
and the associated matching or equivalence pro-
cedures have already been described and il-
lustrated in Sections 5.2.4 and 5.2.5. These and
several others are compiled below for quick refer-
ence purposes.
The results obtained by the different measure-
ment procedures, each with its own particular
criterion, usually differ systematically from one
another. These differences offer clues toward a
better understanding of the functioning of the
visual mechanism, but they are also disturbing as
they put constraints on the validity of the basic
principle of photometry (Section 4.2).
Below we have compiled the major measure-
ment procedures and criteria used in determining
luminous efficiency functions. This is followed by
a selection of luminous efficiency functions ob-
tained by different investigators. Our selection of
the material presented below has been guided to
some extent by the recent review report “Light as
a True Visual Quantity: Principles of Measure-
ment” prepared by the CIE Technical Committee
on Vision (CIE, 1978b), and by Le Grand’s (1972)
review paper.
5.7.1 Matching or Equivalence Criteria and
Experimental Procedures
(i) Minimum Flicker. Two geometrically simi-
lar light patches imaged on the same retinal area
are presented in rapid alternation. By repeated
observations and adjustments of the frequency of
alternation and the radiance of one of the light
patches, a condition is reached for which the
matching field appears quiescent (or at minimum
flicker) but begins to flicker if the radiance of the
variable light patch is raised or lowered by a
small amount. Although elaborate, the visual
judgment or set of judgments involved is, in prac-
tice, easy to apply and is effective whether or not
the two light patches, when they are viewed in
turn under steady conditions, have the same chro-
matic appearance or not. What is matched is
commonly described as brightness, but a more
appropriate term is, perhaps, flicker brightness.
The frequency of alternation at which the
chromatic appearance of one light patch fuses
with the chromatic appearance of the other must
be selected carefully so as to be optimal with
regard to the flicker sensitivity of the visual mech-
anism. In practice, the frequency is set just above
the one at which fusion of chromatic appearance
occurs.
 Luminous Efficiency Functions of Normal Trichromats
The variable light patch usually represents the
monochromatic test stimulus of wavelength A,
whereas the other light patch represents the refer-
ence stimulus which is either monochromatic and
of fixed wavelength A, or a “white” stimulus of
given spectral radiant power distribution. The
reciprocal of the radiance required for minimum
flicker is then plotted as a function of wavelength
of the test stimulus. The resulting luminous ef-
ficiency curve is commonly normalized to unity at
its maximum value occurring at wavelength 4,,.
Useful publications concerning the experimen-
tal procedure are those of Ives (1912), Walsh
(1958), LeGrand (1957), and Kaiser (1979).
(ii) Direct Heterochromatic Brightness Match-
ing. The two light patches of different chromatic
appearance are typically presented at a steady
state in a foveally centered bipartite field. One
patch, making up one half of the field, represents
the reference stimulus which is of fixed spectral
radiant power distribution. The reference stimu-
lus is usually chosen to be of “white” appearance
or to be monochromatic of a fixed wavelength 4.,,.
The second patch, making up the other half of the
field, represents the monochromatic test stimulus
of wavelength A. While fixating at the center of
the dividing line of the bipartite field, the ob-
server adjusts the radiance of the test stimulus
until the test stimulus is perceived to be of equal
brightness to the reference stimulus. The judg-
ment requires the observer to mentally isolate the
brightness quality in the appearances of the two
stimuli when these may differ in hue and satura-
tion. The task is difficult and results in consider-
able variability of the brightness matches made
by the same observer and even more so between
different observers.
In some circumstances, notably with large bi-
partite fields, the color difference between the
two stimuli may necessitate the introduction of
supplementary conditions in the matching crite-
rion and procedure to enable the observer to
make unambiguous judgments. An appropriate
surround of fixed color may be provided and
strict fixation may be relaxed and replaced by a
regular observational routine that involves look-
ing alternately at the two halves of the field and
at the surround. An automatic alternating presen-
tation of a conditioning stimulus for three sec-
onds, say, followed by the presentation of the
bipartite field for one second, and so on, is another
controlling procedure.
The derivation of the luminous efficiency
function from data obtained by direct heterochro-
393
matic brightness matching is similar to that de-
scribed earlier under 5.7.1(i).
(iii) Step-By-Step Brightness Matching. This
method is a special case of heterochromatic
brightness matching described above. Both the
reference and the test stimulus are monochro-
matic and, at each step of the experiment, their
wavelengths are set just a few nanometers apart.
This keeps the color difference between the two
stimuli small and thus makes the task of bright-
ness matching less difficult to perform. The pro-
cedure leads, in fact, to the derivative dV(A)/daA
for a larger number (20 to 30) of wavelengths A,
and the luminous efficiency function V(A) is then
derived by integration (Gibson and Tyndall, 1923;
LeGrand, 1957; 1972).
(iv) Minimally Distinct Border. The two stimuli
are presented in a precisely juxtaposed bipartite
field. One half represents the reference stimulus
of fixed spectral radiant power distribution, usu-
ally of a “white” appearance. The other half
represents the monochromatic test stimulus of
variable radiance. The observer is required to
minimize the distinctness of the border between
the two halves of the field by adjusting the radi-
ance of the test stimulus. This task is made easier
by the use of a lens which compensates for the
chromatic aberration of the eye. The variability of
the settings of minimally distinct border is com-
parable to that obtained by the minimum-flicker
method (Boynton and Kaiser, 1968; Wagner and
Boynton 1972).
(v) Color Matching. Complete trichromatic
color matches are made in a bipartite field involv-
ing the monochromatic test stimulus of wave-
length A and three fixed monochromatic primary
stimuli of wavelengths A zp, AG, Ag. At match, there
is perfect visual identity in hue, saturation, and
brightness between the two halves of the bipartite
field. In varying A of the test stimulus and keep-
ing its radiance constant, the radiances required
for the primary stimuli to provide, for each test
stimulus, a complete color match between the two
halves of the field define three functions of wave-
length A, known as the color-matching functions
r(A), g(A), b(A) (see Section 3.2.1). In a second
experiment, the luminous efficiencies Lp =
V(Ar), Lg = V(Ag), and Lpg= V(Ag) of the
primary stimuli are determined, for example, by
the minimum-flicker method. It then follows the
luminous efficiency function
V(A) = Lar(A) + Leb (A) + Lgb (A)
394 VISUAL EQUIVALENCE AND VISUAL MATCHING
The validity of this relation rests on the validity
of the additivity law for brightness matching (Sec-
tions 3.3.1 and 5.5.2).
(vi) Absolute Threshold. The monochromatic
test stimulus of wavelength A is presented in brief
and repeated pulses as a patch of light of given
dimensions, carefully imaged at a specified lo-
cation of the retina. The minimum radiance (or
number of quanta per second) required by the
observer to “just see” the patch of light defines
the absolute threshold (see Sections 7.2.1 and
7.2.2) for that wavelength and for the given condi-
tions of observation. The determination of foveal
absolute thresholds demands a light patch suffi-
ciently small, less than 1° angular subtense, and
well fixated by the observer, so as to avoid rod
intrusion. The reciprocal values of the minimum
radiance are plotted as a function of wavelength
to arrive at a luminous efficiency curve.
It has been argued by LeGrand (1972) that
foveal absolute threshold data do not measure a
luminous efficiency function if one takes this
function to be the result of the cooperative re-
sponses of the three cone mechanisms. At
threshold, there is no summation of the cone
responses because only one of the cone mecha-
nisms is active, that is, the one most sensitive for
the test stimulus used. A foveal absolute threshold
curve nearly coincides with the “envelope” of the
sensitivities of the individual cone mechanisms
acted upon by the test stimulus (Pirenne, 1962c).
(vii) Increment Threshold. The monochromatic
test stimulus of wavelength A is presented, in
either the steady-state or in brief pulses, against a
fixed conditioning stimulus of given spectral
radiant power distribution. The minimum radi-
ance required by the observer to “just see” the
test stimulus against the conditioning stimulus
defines the increment threshold for that wave-
length A and for the given observing conditions
(retinal location, angular subtense, characteristics
of the conditioning stimulus, and others). This
method has a wide range of applications to isolat-
ing visual mechanisms and exploring their proper-
ties (Sections 7.2.1 and 7.4), but under certain
conditions of observation, which, for example,
involve a conditioning stimulus of a “white” ap-
pearance, spectral sensitivity functions are ob-
tained that may be classed luminous efficiency
functions (Sperling and Harwerth, 1971; Verriest
and Kandemir, 1974).
The absolute threshold method, discussed
above in 5.7.1(vi), may be regarded as a special
case of the increment threshold method by letting
the radiance of the conditioning stimulus in the
increment threshold method go to zero, thus
creating a dark surround.
(viii) Visual Acuity. A spatial visual acuity
target, such as a grating or Landolt C, is placed
on a patch of monochromatic light of wavelength
A whose radiance is adjusted by the observer until
he or she can identify correctly the distinguishing
feature (grating or gap in the C) in the target. The
reciprocal of the radiance of A is used to plot a
luminous efficiency function (Ives, 1912; Graham
and Guth, 1970; Myers, Ingling, and Drum, 1972;
Guth and Graham, 1975).
(ix) Critical Flicker Frequency. A single mono-
chromatic test stimulus of wavelength A is pre-
sented against a dark background at a constant
flicker frequency. The observer adjusts the radi-
ance of the test stimulus for “just-detectable”
flicker. The results depend to some extent on the
given flicker frequency method 5.7.1(i) (Ives, 1912;
De Lange Dzn, 1957; Kelly, 1961; Bornstein and
Marks, 1972; Roufs, 1972; Bornstein, 1975; see
also Section 7.9.3).
(x) Other Methods. Several other matching or
equivalence criteria and procedures have been
used in an effort to arrive at a luminous efficiency
function. In some of these cases, it is debatable
whether the response functions obtained can be
classed luminous efficiency functions. We will not
pursue the matter further except to note some
relevant publications. For studies dealing with the
pupil diameter as a criterion, see, for example,
Laurens (1923), Alpern and Campbell (1962),
Bouma (1965), Adrian (1970), Alexandridis and
Koeppe (1969). Reaction time as a function of
wavelength and radiance has been studied by Lit,
Young, and Shaffer (1971). Finally, there are
numerous papers exploring the luminous ef-
ficiency of the visual mechanism by electrophysio-
logical techniques (see, for example, Regan 1970,
and Riggs and Wooten, 1972).
5.7.2 Experimental Data
(i) The CIE Standard Luminous Efficiency Func-
tions V(A) and V’(A). In 1924, the CIE adopted
the standard photopic luminous efficiency func-
tion V(A) and in 1951, the standard scotopic
luminous efficiency function V’(A) (CIE, 1926;
1951). Both these functions have already been
introduced in our Chapter 4, Photometry, and
 Luminous Efficiency Functions of Normal Trichromats 395

have been used in a number of other sections
throughout the book. The official numerical val-
ues of V(A) and V’(A) have been given in Tables
1(4.3.2) and 1(4.3.2). Figure 1(5.7.2) illustrates the
two functions.
The V(A)-function was derived from the re-
sults of several independent experimental de-
terminations of the luminous efficiencies of
monochromatic stimuli, as summarized below:
(a) Minimum flicker:
125 observers (Coblentz and Emerson, 1918)
(b) Step-by-step:
29 observers (Hyde, Forsythe, and Cady, 1918)
(500 to 660 nm)
52 observers (Gibson and Tyndall, 1923)
(c) Direct brightness matching:
9 observers (Hyde, Forsythe, and Cady, 1918)
(red end of spectrum)
20 observers (Hartman, 1918) (blue end of spec-
trum)
The values adopted in 1924 were those sug-
gested by Gibson and Tyndall (1923) who com-
posed a smooth and symmetric V(A)-curve from
the data cited above. The final result was not an
average of the experimental data, but a weighted
assembly of the different sets of data. From 400
to 490 nm, the V(A)-curve represents roughly the
results of Hartman (1918); from 490 to 540 nm,
those of Coblentz and Emerson (1918); from 540
to 650 nm, those of Gibson and Tyndall; and
above 650 nm, those of Coblentz and Emerson
(1918).
In all experiments, a 2° test field was used,
often with a large surround of the same bright-
ness. In the minimum-flicker measurements, the
brightness of the field was kept constant, corre-
sponding to a luminance of about 1.5 cd - m
seen through the natural pupil. In the step-by-step
method, the brightness of the field varied some-
what, corresponding to a luminance range of about

V3 CIE 1924 V(A)

0.5
»
~
“C
_CIE 1951 oN
°

oO re)re)on

Luminous
Efficiency

O00O00l
400 450 500 550 600 650 700
Wavelength \ (nm)
Fig. 1(5.7.2). Luminous efficiency functions of CIE 1924 standard observer V(A) for photo-
pic vision, CIE 1951 standard observer V’(A) for scotopic vision, and Judd’s 1951 modification
of the CIE 1924 V(A)-function in the blue end of the spectrum.
 396 VISUAL EQUIVALENCE AND VISUAL MATCHING
52 Observers of
Gibson and
Tyndall (1923)
0.8
Variation
Efficiencies
Luminous
of
0.2
400 500° 600 700
Wavelength A (nm)
Fig. 2(5.7.2). Range within which the luminous
efficiency functions of the 52 observers of Gibson
and Tyndall (1923) fall after each function was
normalized to unity at its maximum value. The
luminous efficiency functions were determined by
the step-by-step method.
3 cd- m ” in the middle of the spectrum to 0.5
cd-m * at the ends (LeGrand, 1972).
There are considerable differences between
the luminous efficiency functions of individual
normal trichromats. Figure 2(5.7.2) illustrates the
variability found by Gibson and Tyndall (1923)
for 52 observers using the step-by-step method.
The variability of data obtained by the
minimum-flicker method is, in general, somewhat
less pronounced, but for the 125 observers of
Coblentz and Emerson (1918), the individual
maximum varied between 549 and 570 nm, with
the average being at 557.6 nm.
In the light of more recent measurements, a
standard V(A)-curve differing somewhat from the
“present curve in the blue end of the spectrum
would probably have been adopted today. How-
ever, any such improvements at this stage would
most likely be outweighed by the very consider-
able practical inconvenience of a change in the
basic function on which all photopic photometry
has been based for over 50 years. Moreover, even
if an improved V(A)-curve were chosen for pho-
tometry, the problem that photometry faces today
would not be resolved completely. Failures of the
additivity law for brightness, actually known for
some time, have begun to concern those pho-
tometrists who wish to measure and specify the
brightness of lights of different colors in terms
that relate closely to the visual experience of
actual observers (Section 5.7.3).
The CIE 1951 standard scotopic luminous ef-
ficiency function V’(A) is based on measurements
made by Wald (1945a) and Crawford (1949).
Crawford’s data are for 50 observers, all under 30
years of age, who made direct comparisons be-
tween a “white” reference stimulus of a lumi-
nance of about 3 X 10° cd- m ? seen through
the natural pupil, and a monochromatic test
stimulus of (variable) wavelength A in a 20° field
with a vertical dividing line. The observers fixated
the top of the dividing line, and were dark adapted
for one hour before the start of the experiment.
Wald’s data are for 22 observers, averaging 20
years of age, who determined absolute thresholds
for the monochromatic test stimuli. For further
details of the derivation of the standard V’(A)-
curve;.see CIE (1951).
The standard scotopic luminous efficiencies
apply to completely dark-adapted observers un-
der the age of 30. For stimuli of wavelengths
below 500 nm, V’(A) decreases with age A ap-
proximately in accordance with Crawford’s for-
mula
A [logioV’(A)] = 10~4(500 — A)(A — 30)
(Seren
The decrease in V’(A) is attributed to the progres-
sive yellowing of the eye’s lens with age (Section
2.4.6).
(ii) Judd Modification of CIE V(A). Con-
cerned about inadequacies of the CIE 1931 stan-
dard color-matching functions in the shortwave
region of the spectrum, and recognizing that the
problem may have been caused by the CIE 1924
V(A)-function being too low in the blue end of
the spectrum, Judd (1951) proposed a modifica-
tion to the standard V(A)-function and, resulting
from this modification, a revised set of color-
matching functions. The modification of the
V(A)-function is in the wavelength region below
460 nm as illustrated in Figure 1(5.7.2). The
numerical values have already been given in Table
1(5.5.2). A somewhat refined modification pro-
posed by Vos (1978) is given in Table 1(5.5.2) in
the Appendix. Reference is also made to the
 Luminous Efficiency Functions of Normal Trichromats 397

0.5
CIE 1964 0
ef ye
H(A) =YolA) of

Ol fe)
/
|/

0.05 WY ee 1951
Mod. V (A)

(e)(e)oO on

Efficiency
Luminous

0.0001
400 450 S00 550 600 650 700
Wavelength A (nm)

Fig. 3(5.7.2). Luminous efficiency function V,9(A) for large (10°) centrally viewed fields at
photopic levels derived from large-field color-matching data. V,9(A) is identical to the CIE
1964 y,.(A) color-matching function. For the purpose of comparison, the Judd 1951 modified
V(A)-function of Figure 1(5.7.2) applicable to small (2°) field viewing is also shown.

earlier Figure 1(5.5.2) in which Judd’s modified
V(A)-function is compared with subsequent ex-
perimental data by Ishak and Teele (1955) and
Stiles (1955b).
Although for various reasons, mainly practical,
the CIE has decided to make no change in the
standard V(A)-function, the Judd modification
has found applications in several visual studies.
(iii) 10°-Field Luminous Efficiency Function
Vio(A). The luminous efficiency function which,
at present, is considered most representative for
large centrally viewed fields at photopic levels is
the y,.(A) color-matching function of the CIE
1964 supplementary standard colorimetric ob-
server. Denoted by V,9(A), this function is il-
lustrated in Figure 3(5.7.2) in comparison with
the small-field (2°) Judd 1951 modified V(A)-
function. It has been derived from a linear combi-
nation of the mean experimentally determined
color-matching functions 7,9(A), Zio(A), Bio(A)
that form the basis for the CIE 1964 supplemen-
tary standard observer data:
Vio(A) = Lat) an LG 810(A) wi lend OS
[2(5.7.2)]
The luminous efficiencies Lp, Lg, Lz, of the
primary stimuli, in terms of which the color-
matching functions Fi(A), Zi9(A), Dio(A) are
given, are based on minimum-flicker measure-
ments made by 26 observers of the Stiles and
Burch (1959) investigation. Further details con-
cerning the large-field color-matching functions
have already been given in Sections 3.3.2 and
5.5.4. In particular, reference is made to Tables
398 VISUAL EQUIVALENCE AND VISUAL MATCHING

Table 1(5.7.2) Mean Luminous Efficiency Func- 1(3.3.2) and II(3.3.2) which contain the numerical
tion V,9(A) rn. for 10 Observers of ETL (Japan) values of ¥i9(A) = Vio(A).
Derived from 10° Color-Matching Functions and More recently, Katori and Fuwa (1979) of the
Minimum Flicker Determinations of the Luminous Electrotechnical Laboratory (ETL) of Japan have
Efficiencies of the Primary Stimuli. Also Given determined V,.(A)-functions using a_ similar
Are the Coefficients of Variation c(A), Defined as method of trichromatic color matching for 10
the Ratios of the Standard Deviations of the Japanese observers (see Section 5.5.5). Their mean
Luminous Efficiencies Divided by the Mean Val- results are given in Table 1(5.7.2).
ues for the 10 Observers, Expressed in Percent In their experiment, Katori and Fuwa (1979)
(Katori and Fuwa, 1979) had their observers make color matches by the
method of adjustment as well as by the method of
A (nm) Vio(A eri c(A)
constant stimuli. The method of adjustment is the
395 1.154(—3)4 9.25(1) traditional method used in color matching and
allows the observer to have full manual control
400 2.073(—3) 7.66(1) over the settings of the radiances required for the
410 6.975(—3) 5.50(1) primary stimuli to achieve a complete color match
420 1.699(—2) 3.05(1) in the bipartite visual field. In the method of
430 3.359%(—2) 1.95(1) constant stimuli, on the other hand, the observer
440 5.925(—2) 1.49(1) has no direct control over the mixtures of the
450 8.765(—2) a2
stimuli in the two halves of the field. Instead,
460 1.209(—1) 3.76 these mixtures are generated by automatic means
470 1.762(—1) 6.36 and presented to the observer at short regular
480 2.432(—1) 6.96 intervals. Every presentation involves mixtures
490 3.363(— 1) 8.03
with slightly different radiances of the primary
500 4.718(—1) 6.78 stimuli, all randomly selected from ranges that
510 6.211(—1) 5.40 are expected to include those radiances at which
520 7.723(—1) 2.76 the particular observer would see the two halves
530 9.002(— 1) 1.38 of the field in a complete color match. To each
540 9.788(—1) 1.00 presentation, the observer responds with a “yes”
550 1.000 5.12(—1) when a match is seen or with a “no” when it is
560 9.965(—1) 9.61(—1) not. An appropriate statistical procedure is used
570 9.357(—1) 1.13 to arrive at the settings of the observer’s complete
580 8.458(— 1) 2.19 color match.
590 7.251(—1) 2.47 Katori and Fuwa (1979) found that the two
methods of color matching gave essentially the
600 5.924(—1) 3.73 same results and thus, we have reproduced here
610 4.639(—1) 2.10
only their results based on the method of adjust-
620 3.519(—1) 2.89 ment.
630 2.491(—1) 4.98 The “coefficient of variation” c(A), calculated
640 1.623¢—)) 4.00
by Katori and Fuwa for the values of the mean
650 1.002(— 1) 5262
luminous efficiency function derived from the
660 5.564(—2) 4.90 measurements of their 10 observers, are of inter-
670 2.878(—2) 3.83 est as they relate to the observer variability of
680 1.493(—2) 5.84 luminous efficiencies derived from trichromatic
690 *1:365(23) 8.79
color matches coupled with minimum-flicker set-
700 3.245(—3) 4.31 tings between the primary stimuli. The coefficient
720 7.007(—4) 9.04 c(A) is defined by
740 1.519(—4) 5.90

s [YiolA); a Vio(A) mI] vs

“The numbers in parentheses following the entries de- md 100 ra
note a power of 10. Thus, 1.154(—3) reads 1.154 X 1073. oO aor c mor 17 cee |

[3(5.7.2)]
 Luminous Efficiency Functions of Normal Trichromats 399

fop)ro)

f Oo

ine)je)

Efficiencies
Luminous
of
Variation
% Efficiencies
Luminous
of
Variation
%
400 500 600 700
Wavelength A (nm)
Fig. 4(5.7.2).Variation of luminous efficiencies obtained by Katori and Fuwa (1979) for 10
observers who determined their luminous efficiency functions by linear combinations of their
10° color-matching functions with minimum-flicker measurements between the primary stimuli
(A zp = 670 nm, A, = 530 nm, A z = 460 nm). The variations plotted at each wavelength A, are
the ratio of the calculated standard deviation divided by the mean for the 10 observers,
expressed in percent. Note the changes of scale made at A = 450 nm.

where V,,(A); is the luminous efficiency function defined by

for the ith observer, with i = 1 ton, and V,)(A),,
is the mean luminous efficiency function for n = 100
rel = ———|V,,(A
Ven [ 10 ett — Poth
10( Jere |
10 observers. Figure 4(5.7.2) illustrates c(A) as
obtained by Katori and Fuwa (1979). The varia-
bility between the observers is less than 10% of [4(5.7.2)]
the mean data for wavelengths 450 nm and above,
and becomes rapidly poorer below 450 nm. The agreement between the two sets of data is
The agreement between the mean values ob- considered good, particularly if one takes into
tained by Katori and Fuwa and the CIE 1964 account a possible difference in the effective den-
values is illustrated in Figure 5(5.7.2) in which is sity of the lens pigment for the two groups of
plotted the “relative deviation” D,., (in percent), observers, and a potential rod participation in the

1979
1964
ETL
CIE
and
(in%)
Efficiencies
Luminous
Deviation
Relative
Between

400 500 600 700

Wavelength A (nm)
Fig. 5(5.7.2). Relative deviations (in percent between the luminous efficiencies obtained by
Katori and Fuwa (1979) at ETL (Japan) and those defined by the CIE 1964 V,,.(A)
[ = y(A)]-function.
400 VISUAL EQUIVALENCE AND VISUAL MATCHING
ETL data which were obtained at a retinal illumi-
nance of approximately 200 td.
(iv) 2°-Field Luminous Efficiency Functions
Based on Heterochromatic Brightness Matching.
The CIE Vision Committee (CIE, 1978b) has
produced a weighted average luminous efficiency
function for 31 observers making direct hetero-
chromatic brightness matches in centrally viewed
fields of 1° to 2° angular subtense. The average
curve is derived from the experimental data re-
ported in seven different studies (Bedford and
Wyszecki, 1958; Sperling, 1958; Sperling and
Lewis, 1959; Kinney, 1964; Wagner and Boynton,
1972; Guth and Lodge, 1973; Comerford and
Kaiser, 1975). The data taken from the individual
studies are entered into the calculation of the
mean with weights corresponding to the number
of the observers participating in these studies.
Table 2(5.7.2) contains the numerical values of
the seven individual studies as well as their
weighted mean, and Figure 6(5.7.2) illustrates
them.
Some of the variations between the data of the
seven different studies are no doubt caused by
differences in the observing conditions sum-
marized at the bottom of Table 2(5.7.2): Dif-
ferences in field size, color of reference stimulus,
and level of stimulation all have measurable ef-
fects on the results, but these effects are difficult
to extract and correct from the given data, and no
attempt has been made to do so. Furthermore,
the observational procedures used were generally
different from one study to the next, and
those differences may have enhanced the varia-
tions between the different sets of data. In some
studies, the observers were presented with pulsed
stimuli, in others with steady-state fields which
were viewed with casual fixation and frequent
re-adaptation to the reference stimulus. Again, no
attempt has been made to correct for differences
in the observational procedures used by the dif-
ferent investigators.
When considering the obvious experimental
shortcomings of the weighted mean luminous ef-
ficiency function given in Table 2(5.7.2) and Fig-
ure 6(5.7.2), it is somewhat surprising that
the deviations of the individual curves from the
mean are not larger. If we then accept the mean
curve as representative of luminous efficiencies
obtained by direct heterochromatic brightness
matching in, say, a 2° bipartite field with a
“white” (daylight type) reference stimulus provid-
ing a retinal illuminance of 200 td, we may be
permitted to compare it with Judd’s modified
V(A)-curve, given in Figure 1(5.7.2), which is
considered representative of luminous efficiencies
obtained by both the method of minimal flicker
and the method of minimally distinct border
(Wagner and Boynton, 1972). Figure 7(5.7.2)
shows the ratio between the two luminous ef-
ficiency curves. From the shape of the ratio curve,
it is apparent that the luminous efficiency curve
based on heterochromatic brightness matching is
broader than Judd’s modified V(A)-curve. It also
exhibits more pronounced humps. These features
have led several investigators to suggest that het-
erochromatic brightness matching is mediated by
an achromatic mechanism as well as by two chro-
matic mechanisms, whereas the V(A)-curve based
on minimum flicker (or minimally distinct border)
is essentially the output of only the achromatic
mechanism (Section 8.3).
(v) Foveal Luminous Efficiency Function Based
on Minimally Distinct Border Judgments. Figure
8(5.7.2) shows the mean luminous efficiency func-
tion obtained by Wagner and Boynton (1972) for
four observers who used as the equivalence crite-
rion that of minimally distinct border. A centrally
viewed bipartite circular field of 1°40’ angular
subtense was used in a dark surround. The retinal
illuminance provided by the “white” reference
stimulus was approximately 80 td.
The same four observers made also -measure-
ments by using the criteria of minimum flicker,
step-by-step heterochromatic brightness match-
ing, and direct heterochromatic brightness match-
ing. The resulting mean luminous efficiency func-
tion for each of these cases is also given in Figure
8(5.7.2).
The luminous efficiency function based on
minimally distinct border judgments agrees well
with that based on minimum flicker (see also
Tansley and Glushko, 1978). Only relatively small
systematic deviations are noted between the two
functions, making the one based on minimally
distinct border marginally broader in shape. Both
functions deviate in a significant manner from
those determined by the step-by-step and direct
heterochromatic brightness matching; but they
agree well with Judd’s modified V(A)-function in
Figure 1(5.7.2).
The luminous efficiency function based on the
step-by-step method agrees quite well with the
one based on direct heterochromatic brightness
matching in the short to middle region of the
spectrum, but deviates from it significantly for
longer wavelengths.
Table 2(5.7.2) Luminous Efficiencies Obtained by Direct Heterochromatic Brightness Matching in
Seven Different Studies and Their Weighted Mean. Each Set of Data Has Been Normalized To Be
Unity at 570 nm (CIE, 1978b).

rN Weighted
(nm) 1 2 3 4 3 6 7 Mean

400 0.003 0.034 0.019

10 0.008 0.025 0.064 0.032
20 0.020 0.052 0.037 0.036 0.012 0.094 0.042
30 0.038 0.124 0.057 0.064 0.064 0.104 0.076
40 0.050 0.161 0.078 0.080 0.076 0.167 0.103
450 0.074 0.187 0.10 O19 0.088 0.080 0.229 0.135
60 0.104 0.210 0.13 0.21 tz 0.092 0.312 0.166
70 1.480 0.258 0.18 0.23 OA72 0.130 0.375 0.214
80 0.250 0.378 0.23 0.29 0.206 0.168 0.542 0.294
90 0.300 0.464 0.38 O37 0.231 0.166 0.583 0.359

500 0.420 0.603 0.60 0.56 0.368 0.412 0.709 0.523

10 0.560 0.809 0.82 0.74 0.542 0.58 0.823 0.698
20 0.880 1.090 ed byl 0.775 0.847 0.979 0.978
30 1.02 1.136 Tt 1.25 0.961 0.947 1.07 1.103
40 tt 1.238 1.44 1.24 1.15 1.06 1.08 1.200

550 1.08 1.308 1.24 23 1.186 1.07 1.06 is Wk

60 1.02 1.118 1.11 1.06 eB 1.05 1.01 1.075
70 1.00 1.000 1.00 1.00 1.00 1.00 1.00 1.000
80 1.06 0.962 0.90 1.02 0.907 0.947 1.01 0.967
90 0.94 0.951 0.89 1.06 0.945 0.962 1.04 0.965
600 0.86 0.941 0.89 0.89 0.814 0.886 1.01 0.895
10 0.68 0.910 0.77 0.72 0.658 0.801 0.969 0.782
20 0.50 0.792 0.64 0.56 0.511 0.702 0.904 0.651
30 0.38 0.556 0.48 0.45 0.346 0.495 0.813 0.494
40 0.23 0.439 0.36 0.35 0.235 0.298 0.614 O35;
650 0.136 0.294 0.20 0.27 0.145 0.113 0.099 0.184
60 0.076 0.159 0.070 0.084 0.070 0.093
70 0.038 0.085 0.039 0.024 0.049
80 0.024 0.045 0.021 0.018 0.028
90 0.013 0.021 0.021 0.005 0.014

700 0.005 0.011 0.004 0.007

Data Number of
Set Investigators Visual Field Reference Stimulus Level Observers

] Bedford and Wyszecki (1958) 1° bipart. “white” (xenon) 50 td 4

2 Comerford and Kaiser (1975) —_1° bipart. “white” 150 to 400 td 5
x = 0.34 for diff.
y = 0.34 observers
we, Guth and Lodge (1973) 45’ bipart. “white” 190 td 5,
2430 K
4 Kinney (1964) 20 .2ire: “white” 0.34 4
test in 120° x410:33 cd: m7?
ref. y = 0.33
5 Sperling (1958) 2° bipart. monochr. 520 nm 500 td 6
6 Sperling and Lewis (1959) 2° bipart. monochr. 520 nm 500 td 3
L Wagner and Boynton (1972) 1°40’ bipart. “white” 10 td at 4
x = 0.29 445 nm, 570 td
y= 0.31 at 560 nm,
1 td at 690 nm

401
 Efficiency
Luminous Direct Heterochromatic
Brightness Matching
Weighted Mean (CIE 1978)
Bedford & Wyszecki (I958)
Comerford & Kaiser (1975)
—x—x—- Guth & Lodge (1973)
Kinney (1964)
Sperling (1958)
—e—e—e— Sperling & Lewis (1959)
—a—s—a— Wagner & Boynton (1972)

0.00!
400 450 500 550 600 650 700

Wavelength A (nm)
Fig. 6(5.7.2). Luminous efficiency functions obtained in seven different investigations by
means of heterochromatic brightness matching. The heavy line through the open circles is the
weighted mean of the seven sets of data (CIE, 1978b).

Luminous
of
Ratio
Efficiencies

400 450 500 950 600 650 700

Wavelength A (nm)
Fig. 7(5.7.2). Ratio of mean luminous efficiency function obtained by heterochromatic
brightness matching over Judd’s modified luminous efficiency function.

402
 Luminous Efficiency Functions of Normal Trichromats 403

Efficiency
Luminous

eee Minimum Flicker

ee Minimally Distinct Border
meres Step by Step
=—O=0=0= Direct Heterochromatic Brightness Matching

Wagner & Boynton (1972)

0.001
4 00 450 500 550 600 650 700

Wavelength A (nm)
Fig. 8(5.7.2). Mean luminous efficiency functions obtained by Wagner and Boynton (1972)
for four observers, each making measurements by means of the equivalence criteria of
minimum flicker, minimally distinct border, step-by-step heterochromatic brightness matching,
and direct heterochromatic brightness matching:

(vi) 1°-Field Luminous Efficiency Functions
Based on Absolute Threshold Determinations.
Absolute threshold data and the interpretation of
such data is a complex issue that will be discussed
in Sections 7.2 and 7.3. What is presented here
are data on absolute threshold luminous efficien-
cies for centrally viewed fields of 42’ to 1.0°
angular subtense. These data have been selected
and compiled by the CIE Vision Committee (CIE,
1978b) as representative data of that kind. Table
3(5.7.2) contains the numerical values taken from
five different investigations and their mean
weighted by the number of observers involved in
each of these sets of data. Figure 9(5.7.2) il-
lustrates the data. The mean curve resembles fairly
closely the mean curve obtained by heterochro-
matic brightness matching shown in Figure
6(5.7.2).
Maximum luminous efficiency (i.e., minimum
radiant power) occurs for A = 540 nm, approxi-
mately. For fixed observing conditions, different
observers agree in their settings of minimum
radiant power within a range of 0.3 log units,
whereas an individual observer varies in his or her
own settings made on different days within the
range of 0.1 to 0.2 log units (Hsia and Graham,
1957).
(vii) Photopic Luminous Efficiency Functions in
the Periphery of the Retina. The determination
of photopic luminous efficiency functions in pe-
ripheral regions of the retina requires special tech-
nical precautions, some of which are of the same
nature as those discussed in Section 5.6.3 dealing
with the effects of the location of the visual field
on color matching.
Changes in the properties of peripheral vision
caused by prolonged exposure to light (Troxler
effect) can adversely influence the experimentally
determined luminous efficiency data both with
regard to their mean values and their precision.
To reduce or control the Troxler effect, the color
stimuli are presented as brief pulses followed by
appropriate off-periods. '
Table 3(5.7.2) Luminous Efficiencies for Centrally Viewed Visual Fields (42’ to 1.0°) Obtained by
the Absolute Threshold Method in Five Different Studies. Their Weighted Mean Is Also Given
(CIE, 1978b).

r Weighted
(nm) l 2 3 4 5 Mean
400 0.013 0.011 0.012
10 0.026 0.025 0.089 0.045
20 0.045 0.032 0.031 0.021 0.089 0.043
30 0.070 0.052 0.050 0.038 0.107 0.062
40 0.094 0.058 0.055 0.051 0.109 0.074
450 0.103 0.045 0.062 0.065 0.114 0.077
60 0.123 0.091 0.081 0.087 0.118 0.102
70 0.186 0.129 0.106 0.142 0.134 0.145
80 0.247 0.204 0.168 0.194 0.330 0.229
90 0.381 0.363 0.255 0.282 0.432 0.350
500 0.598 0.513 0.373 0.468 0.573 0.520
10 0.825 0.724 0.540 0.709 0.855 0.749
20 13 0.912 0.733 0.934 1.02 0.972
30 1.24 1.15 0.845 0.892 Lit 1.08
40 1.34 1.29 1.03 1.29 1.14 i
550 1.19 1.28 1.06 1.12 1.11 ea,
60 1.03 1.07 1.03 0.934 1.09 1.03
70 1.00 1.00 1.00 1.00 1.00 1.00
80 0.928 0.813 0.975 0.794 0.932 0.875
90 0.825 0.646 0.870 0.676 1.05 0.787
600 0.732 0.513 0.733 0.709 1.05 0.720
10 0.690 0.576 0.565 0.709 0.886 0.676
20 0.567 0.457 0.429 0.550 0.739 0.504
30 0.410 0.323 0.360 0.380 0.682 0.419
40 0.289 0.204 0.217 0.194 0.522 0.276
650 0.129 0.118 0.148 0.182 0.144
60 0.058 0.062 0.063 0.105 0.071
70 0.036 0.031 0.042 0.061 0.042
80 0.018 0.012 0.021 0.034 0.018
90 0.008 0.006 0.007 0.007
700 0.004 0.004 0.004 0.004

Data Number of
Set Investigations Visual Field Presentation Observers

] Guth and Lodge (1973) 45’ cire. steady-state 5

method of adj.
2 Hsia and Graham (1952) 42’ 4-msec flashes 5
method of limits
a Hurvich and Jameson (1953) 1° x 48’ 45-msec flashes, 2
ellipt. method of limits
4 Sperling and Hsia (1957) 42’ 4-msec flashes, 2 to4
method of limits
5 Sperling and Lewis (1959) 45’ 4 ms flashes, :
method of limits
 Luminous Efficiency Functions of Normal Trichromats 405

Efficiency
Luminous

Absolute Thresholds
0.01 |
—o==o==o= Weighted Mean (CIE 1978)
xx Guth & Lodge (1973)
0.005 — — Hsia & Graham (1952)
Hurvich & Jameson (1953)
—4—s—a— Sperling & Hsia (1957)
—e—e—e— Sperling & Lewis (1959)

0.001
400 450 500 550 600 650 700

Wavelength A (nm)
Fig. 9(5.7.2). Luminous efficiency functions obtained in five different investigations by means.
of the absolute-threshold method. The heavy line through the open circles is the weighted mean
of the five sets of data (CIE, 1978b).

Another difficulty is to eliminate effectively
rod participation from the visual response to given
test stimuli. Appropriate conditioning (or adapta-
tion) stimuli are presented to suppress the re-
sponse of the rod mechanism without selectively
adapting the different cone mechanisms.
There are many studies reported in the litera-
ture that pursue the difficult task of measuring
photopic luminous efficiency functions in differ-
ent parts of the parafoveal and peripheral retina,
of which the following are a selection from more
recent such efforts: Weale (195la,b, 1953a, b);
Sperling and Hsia (1957); Kinney (1958); More-
land and Cruz (1959); Wooten and Wald (1973);
Wooten, Fuld, and Spillman (1975); Stabell and
Stabell (1976, 1980) and Abramov and Gordon
(1977). The results obtained by the different in-
vestigators often differ considerably from one
another. The deviations are attributed to observer
differences as well as differences in the observing
conditions, such as size and retinal location of the
test stimulus, the state of adaptation of the eye,
and the matching or equivalence criterion used by
the observers.
The study by Wooten, Fuld, and Spillman
(1975) is particularly informative. These investiga-
tors determined photopic luminous efficiency
functions (they call them spectral sensitivity func-
tions) in the mid- and far-peripheral retina by two
methods. The first method consisted of measuring
increment thresholds on a background stimulus of
a spectral radiant power distribution similar to
that of CIE standard illuminant A (representative
of tungsten-lamp light). The resulting luminous
efficiency function showed a pronounced hump at
about 440 nm, in agreement with previous work
done by Weale (1953a). The second method con-
sisted of measuring, for monochromatic test
stimuli of different wavelengths, dark-adaptation
curves (see also Section 7.3.1) after abrupt extinc-
tion of the high-luminance, “white”-background
stimulus. From these dark-adaptation curves,
luminous efficiency functions were derived for
different times in the dark. The results showed
that the maximum at 440 nm diminished quickly.
During the period at which the cone plateau was
observed in the dark-adaptation curves, the corre-
sponding luminous efficiency functions peaked at
406 VISUAL EQUIVALENCE AND VISUAL MATCHING
approximately 550 nm and had much the same
shape for different locations in the retina, from
the parafovea to the far periphery. Wooten et al.
(1975) suggested that the previous finding by
Weale (1953) of maximum photopic luminous
efficiency in the short-wavelength region of the
spectrum resulted from chromatic adaptation in-
duced by a background stimulus (such as CIE
illuminant A) that selectively weights the cone
mechanisms. Figure 10(5.7.2) illustrates some of
the results obtained by Wooten et al. (1975) for
observer BW at 30° retinal eccentricity.
A comparison of parafoveal luminous ef-
ficiency functions obtained from the cone-plateau
values of the dark-adaptation curves of Wooten
et al. (1975) with foveal luminous efficiency curves
indicated that the deviations in the short-
wavelength region of the spectrum could, for the
most part, be attributed to the macular pigmenta-
tion in the fovea. However, some small but sys-
tematic deviations remained after appropriate
corrections for macular pigmentation had been
applied to the foveal data. These deviations were
thought to originate from small changes in the
sensitivities of the different cone mechanisms
when going from the parafoveal to the foveal
region of the retina (see also Section 2.4.6).
(viii) Mesopic Luminous Efficiency Functions.
Mesopic vision is mediated by the rod mechanism
and the cone mechanisms responding simul-
taneously to stimuli of appropriate spectral
compositions and luminances. Whereas mesopic
vision begins at luminances of approximately 107°
cd-m 7, the luminance level at which it ends
depends to a large extent on the spectral composi-
tion of the stimuli viewed, their sizes, and loca-
tion on the retina. The level of 3 cd - m ? for the
upper luminance level often quoted (for example,
CIE, 1978b), is at best a crude measure and
applies only to certain experimental conditions.
According to Johnson (1937), the upper limit is
about 5 cd: m * (~ 60 td) for a 3° centrally
viewed visual field and at least 15 cd-m 7”)
(~ 140 td) for a 25° field. Stiles and Wyszecki
(1973), studying rod intrusion in large-field color
matching, estimate that in some cases, luminance
levels of 30,000 cd - m ° are required to saturate
the rod mechanism (see Section 5.6.2).
There are many experimental studies in which
mesopic luminous efficiency functions have been
determined for a variety of observing conditions,
and several other studies in which attempts have
been made to develop a photometric procedure of
measuring lights in the mesopic range of vision. A
selected list of references is the following: Walters
and Wright (1943), Wright (1946), Kinney (1955,
1958, 1964), CIE (1964), Palmer (1966, 1967, 1968,
1972a), Hough (1968), Hough and Ruddock
(1969), Kokoschka (1972), Clarke and Trezona
(1976), and Kokoschka and Bodmann (1976).
The experimental data show that a luminous
efficiency curve in a mesopic range has a maxi-
mum whose location is intermediate to the maxi-
mum of the scotopic luminous efficiency curve
near 505 nm and to the maximum of the photopic
luminous efficiency curve near 555 nm. The shift
of the luminous efficiency function from V’(A) to
V(A) as the luminance level is raised, or from
V(A) to V’(A) as the luminance level is lowered,
has been called the Purkinje shift; it is a direct
consequence of the transition from rod to cone
vision, or vice versa. The phenomenon was first
described by Purkinje (1823), who observed a
dramatic change in the relative brightness of blue
and red objects when the illumination was
changed from daylight to twilight levels.
The experimental evidence also shows that
mesopic luminous efficiency curves cannot be ob-
tained by linear combinations of scotopic and
photopic luminous efficiency curves, Complex
rod—cone interactions in the mesopic range pre-
clude such a simple mathematical model for
mesopic photometry (Hough, 1968; Hough and
Ruddock, 1969; Kokoschka, 1972; Clarke and
Trezona, 1976).
From the experimental studies listed above,
those by Palmer (1967) and by Kokoschka (1972)
are particularly noteworthy and will be dealt with
briefly below. Another important contribution is
that of Clarke and Trezona (1976) which has
already been discussed at some length in Section
3.612) .
Palmer (1967) presented to his 18 observers a
15° bipartite field for central viewing. In one half
of the field, the reference stimulus was displayed
with a spectral radiant power distribution ap-
proximately that of a blackbody radiator of 2042
K (freezing point of platinum; see Section 1.2.2
and 4.3.1). In the other half of the field, quasi-
monochromatic test stimuli were displayed with
the aid of interference filters of approximately 30
nm half-width and peaking at wavelengths be-
tween 410 and 680 nm. The observers made direct
heterochromatic brightness matches between the
test and reference stimuli by adjusting the radi-
ances of the test stimuli, for fixed luminances L,..
of the reference stimulus from 0.316 to 0.000316
 10000
Cone Plateau
5000

1000

500

100

50

1.0 —_
Increment

(on
basis)
Efficiency
Luminous
quantum

0.

0.05

0.0l
400 450 500 550 600 650 700
Wavelength A (nm)
Fig. 10(5.7.2). Luminous efficiency functions obtained by Wooten, Fuld, and Spillmann
(1975) for observer BW at 30° retinal eccentricity. The lower curve was obtained from
increment-threshold measurements of 1° monochromatic test stimuli of different wavelengths,
presented in flashes of 190 ms every 2 s, on a steady 12° background stimulus of spectral
radiant power distribution similar to CIE illuminant A providing a retinal illuminance of
300,000 td. The background stimulus was on for 4 min; increment thresholds were measured
during the first two minutes. The remaining upper curves were derived from recovery dark
adaptation curves measured after the abrupt extinction of the background stimulus. For each
monochromatic test stimulus, a new dark-adaptation curve was measured. From these curves,
luminous efficiency curves were derived for different times in the dark, as indicated next to
each curve in the figure. The top curve is derived from the cone-plateaus of the dark-adaptation
curves. The cone-plateaus begin after about 4 min in the dark; they terminate at the rod—cone
break which occurs at times that depend on the wavelength of the test stimulus. For
wavelengths longer than 610 nm, there is no rod—cone break.

407
408 VISUAL EQUIVALENCE AND VISUAL MATCHING
cd-m ? at intervals of 0.5 log units. For each
brightness match at a given L,.,, the photopic and
scotopic luminances L and L’ of the test stimulus
were determined in accordance with the equations
Ls Ky, {LeViolA) dx
A filters for wavelengths A = 650 nm and A = 470
Leake ifESV (Ae Meals Oday
An “equivalent luminance” L.,yiy of the test
stimulus is then defined as an empirical function
of the luminances L and L’.
Figure 11(5.7.2) shows examples of mean
curves of L.quiy (L, L’) obtained by Palmer (1967)
for four different luminances L,.; of the reference
stimulus (2042 K). In the ideal case, the loci of
constant equivalent luminance would be straight
lines parallel to the L-axis at low luminances L,,.r
and parallel to the L’-axis at high luminances
L.-¢- However, the ideal is not reached in the
experiment. Instead, the limiting loci are rather
650* oo
0 70 470
L(cd-m®)
Luminance
Photopic
(0: ee:
Scotopic Luminance L’ (cd-m@)
Fig. 11(5.7.2). Mean curves of constant equiva-
lent luminance L,,,;, aS a function of the photopic
and scotopic luminances P and S of quasi-mono-
chromatic test stimuli matched in brightness with a
2042 K reference stimulus whose luminance L_., is
set at four different levels. Curves 1 to 4 refer to
luminances L,., = 3.16 X 10-4, 3.16 < 1073,
3.16 X 10-7, and 3.16 < 107! cd- m ”, respec-
tively. Eighteen observers took part in the experi-
ment making heterochromatic brightness matches
between the test and reference stimuli in a 15°
bipartite field viewed centrally. (Redrawn from
Palmer, 1967.)
curved. In comparing the four loci drawn in Fig-
ure 11(5.7.2), it is helpful to consider that test
stimuli of the same wavelength must lie on the
same 45° line in the diagram; this is an inherent
property of the method of presenting the data.
For the high level (Curve 4), slightly different
nm were used, resulting in slight shifts of their
actual spectral designations, as indicated by
crosses in the diagram.
In a subsequent paper, Palmer (1968) derives
an empirical formula for the observed curves L equiv
(L, L’). This formula is given as
ML’ + L?
|Beal biBe L’) = M+L [6(5.7.2)]
in which all quantities are given in cd - m 7. The
quantities L and L’ are calculated by means of
Eq. 5(5.7.2) involving the V,.(A) = yi (A) and
V’(A)-functions representing the spectral
luminous efficiencies for (large-field) photopic and
scotopic vision, respectively. The quantity M is an
optimized value of 6 X 10°? cd - m~?.
The validity of Eq. 6(5.7.2) rests on the as-
sumption that both the mechanism for photopic
and the mechanism for scotopic vision each con-
tribute to the equivalent luminance judgment in
the mesopic range as single-parameter functions.
This assumption may be well founded for the
scotopic mechanism as only the rods are mediat-
ing the scotopic response. However, for the pho-
topic mechanism, the assumption may be too
simple to account for other evidence concerning
that mechanism. The interactions between the
three cone mechanisms and the rod mechanism
are likely to vary in a rather complex manner as. a
function of luminance level, making it difficult to
| express the response of the photopic mechanism
as a single-parameter quantity. It is.more likely
that equivalent luminance L,,,;, in the mesopic
range is a function of three cone responses and
the rod response whose inter-relations are depen-
dent on the level of stimulation. Such a model
has, in fact, been proposed by Kokoschka (1972)
and Kokoschka and Bodmann (1976).
Kokoschka’s (1972) extensive study on mesopic
photometry contains, among other items, a useful
set of mesopic luminous efficiency functions.
Three observers made direct heterochromatic
brightness matches between quasi-monochromat-
ic test stimuli, produced with the aid of inter-
ference filters with peaks at 400, 430, 460, 480,
 Luminous Efficiency Functions of Normal Trichromats 409

Scotopic

Mesopic

—o—o—o — 10% cd-m? (0.003 td)

—x—x—x- 10? cd-m™@ (0.03 td)
es 10°! cd-m™? (03 td)
—A—A—A— | cd-m™? (3 td)
—e—e—e— |0 cd-m7~2 (30 td)

Efficiency
Luminous Kokoschka (1972)

0.0005

0.0001
4 00 450 500 550 600 650 700

Wavelength A (nm)
Fig. 12(5.7.2). Mean mesopic luminous efficiency functions obtained by Kokoschka (1972)
for three observers (aged 27 to 32) in a 9.5° bipartite field centrally viewed. The observers
made direct heterochromatic brightness matches between quasi-monochromatic test stimuli and
a 530 nm reference stimulus whose luminance was set at 10°, 10-7, 10°', 1, and 10 cd-m7
seen through a 3 mm” artificial pupil. The heavy lines represent the standard scotopic luminous
efficiency function V’(A) and the photopic V,9(A )-function.

500, 550, 575, 600, 620, 660, and 700 nm, and
half-widths between 9 and 12 nm, and a quasi-
monochromatic reference stimulus with a peak
wavelength at 530 nm. Test and reference stimuli
were presented in a bipartite centrally viewed
circular field. The main experimental results refer
to three different field sizes (3°, 9.5°, 64°), and to
five different luminance levels of the (530 nm)-
reference stimulus: 107°7,10~7,107',1,10 cd-
m * which correspond to approximately 0.003,
0.03, 0.3, 3, and 30 td, respectively. Table 1(5.7.2)
in the Appendix gives Kokoschka’s (1972)
numerical values of his luminous efficiency func-
tions from 400 to 700 nm at 10-nm intervals.
Figure 12(5.7.2) illustrates the set obtained for the
9.5° bipartite field at the five given luminance
c
levels of the reference stimulus. For comparison
purposes, the standard V’(A)-function and the
Vio(A) [= ¥i0(A)]-function have been added to
the figure.
Kokoschka (1972) made a factor analysis of
his experimental data to determine the number of
visual response mechanisms that might be re-
quired to explain mesopic heterochromatic bright-
ness matching. His analysis suggests that his data
can be explained by a physiological model that is
based on four component mechanisms, three cone
and one rod mechanism. The responses of these
mechanisms are weighted independently by dif-
ferent functions of the luminance of the reference
stimulus before they are added linearly to give the
total response corresponding to the observed
410 VISUAL EQUIVALENCE AND VISUAL MATCHING
equivalent luminance of the test stimulus. With
decreasing luminance of the reference stimulus,
the model predicts the following physiological
processes to take place:
(a) Rods are increasingly activated.
(b) “Red” cones are increasingly inhibited.
(c) “Green” cones are, at first, activated but are
inhibited in the middle of the mescopic range.
(d) “Blue” cones are increasingly activated at
rates greater than the “green” cones, but they
cease to respond when the luminance level be-
comes very low. The contribution of the “blue”
cones to the total equivalent-luminance response
is, however, small. The latter finding seems to
contradict Hough and Ruddock (1969) who pro-
posed that in mesopic vision, the “blue” cones
play a dominant role.
5.8 HETEROCHROMATIC BRIGHTNESS
MATCHING OF COMPLEX STIMULI
In Section 5.7, heterochromatic brightness match-
ing was one of several matching criteria used to
arrive at luminous efficiency functions. In a typi-
cal experimental procedure, the monochromatic
test stimulus of wavelength A was matched in
brightness against a fixed reference stimulus usu-
ally of a “white” appearance. Test and reference
stimuli were presented in a centrally viewed bi-
partite field of given angular subtense. Such an
experiment is a special case of the general visual
task of making direct heterochromatic brightness
matches of test stimuli of arbitrary spectral radiant
power distributions (not restricted to a single
wavelength) against a given reference stimulus.
The general case is of particular interest to practi-
cal photometry of sources of colored lights. It also
offers further insights into the functioning of the
visual mechanism.
When a test stimulus of complex spectral
radiant power distribution has been matched in
brightness against a fixed “white” reference
stimulus, it is generally found that the luminance
of the chromatic test stimulus is lower than the
luminance of the reference stimulus. Or, stating it
another way, a chromatic stimulus, of the same
luminance as the “white” reference stimulus will,
in general, appear brighter than the reference
stimulus. To many observers, the chromatic
stimulus appears to “glow” (or appears to exhibit
“Farbenglut”). The effect is sometimes referred
to as the Helmholtz—Kohlrausch effect (see, for
example, Kohlrausch, 1935; H. Konig, 1947; and
Judd, 1958).
If the luminance of the reference stimulus is
denoted by Lp = B and that of the test stimulus
of L;= L, the ratio B/L is, in general, greater
than unity and increases with increasing satura-
tion (or excitation purity) of the test stimulus
when the reference stimulus has zero saturation,
that is, has an achromatic appearance. An excep-
tion to this rule is often reported for yellow
stimuli for which B/L remains close to unity even
at maximum saturation.
In deriving the B/L ratio for the brightness-
matched reference and test stimuli, the basic
equation of photometry (Section 4.2) is used, that
is,
L (or B) = Ky,[LeV(A) dr [1(5.8)]
where L,, is the spectral concentration of radi-
ance of the stimulus (either test or reference),
V(A) is the CIE 1924 standard photopic luminous
efficiency at wavelength A, and K,,, is a constant
fixed at 683 Im - W_'. This equation implies that
the basic laws of brightness matching, particularly
the proportionality and additivity laws, often re-
ferred to as Abney’s laws, hold strictly.
Deviations from unity of the B/L ratios for
brightness-matched reference and test stimulus
can be caused by:
(a) use of the wrong V(A)-function in Eq. 1(5.8),
(b) failure of Abney’s laws, or
(c) both (a) and (b) simultaneously.
The experimental evidence is that another
V(A)-function, such as Judd’s modified V(A)-
function or one based on direct heterochromatic
brightness matching, if used in place of the CIE
standard V(A)-function in Eq. 1(5.8), does not
yield B/L ratios of unity for chromatic test stimuli
each of which brightness-matched against a given
“white” reference stimulus. Abney’s additivity law
is not valid for the general case of direct hetero-
chromatic brightness matching.
Below, we present some selected experimental
data on heterochromatic brightness matching of
complex stimuli and results of investigations that
aimed specifically at testing Abney’s additivity
law.
 Heterochromatic Brightness Matching of Complex Stimuli
5.8.1. Luminance of Equally Bright
Color Stimuli
There are several experimental studies on hetero-
chromatic brightness matching of stimuli of com-
plex spectral radiant power distributions. Of
particular interest are the studies of Chapanis and
Halsey (1955), Sanders and Wyszecki (1957, 1958,
1964), Wyszecki and Sanders (1957), Breneman
(1958), Wyszecki (1967), Kaiser and Smith (1972),
and Alman (1977). The paper by Dresler (1953)
summarizes the more important studies on the
subject matter available up to 1952.
We have selected two sets of experimental
data that illustrate the general trend found by
most investigators.
The first set has been taken from the paper by
Sanders and Wyszecki (1964). Twenty observers
with normal color vision made direct heterochro-
matic brightness matches of 95 test stimuli
of different chromaticities against a “white”
reference stimulus of a fixed chromaticity. A
MacAdam-type binocular colorimeter [see Figure
4(5.15)] was used providing a 10° bipartite visual
field surrounded by a “white” field of approxi-
mately 40° X 40° angular subtense. The CIE
specifications of the stimuli presented to the ob-
servers are given in Table 1(5.8.1) and illustrated
in Figure 1(5.8.1).
Each observer viewed the visual field with
both eyes; strict fixation was avoided, encourag-
ing frequent direct views of the “white” surround
for the purpose of controlling to some extent the
state of adaptation of the observer’s eyes. The test
stimuli presented in one half of the bipartite field
were kept at a constant luminance L = 20 cd -
m 7, while the luminance B of the “white” refer-
ence stimulus was changed by the observer until
equality of brightness was perceived between the
two halves. The luminances L and B were calcu-
lated in accordance with Eq. 1(5.8), but using the
Yio(A)-function as the luminous efficiency func-
tion V(A). In the last column of Table 1(5.8.1),
the mean B/L ratios are given for the group of 20
observers.
Curved lines of constant values of the B/L-
ratio can be plotted in the chromaticity diagram
to illustrate how brightness changes with the
chromaticity of test stimuli of constant (20 cd -
m ~*) luminance [Figure 2(5.8.1)]. The further the
chromaticity point of the test stimulus is from
that of the “white” reference stimulus, the brighter
the test stimulus appears against the “white”
reference stimulus of the same luminance. In order
411
to make the “white” reference stimulus appear as
equally bright as the test stimulus of 20 cd - m~’,
its luminance must be raised by a factor equal to
the B/L ratio determined for that test stimulus.
Similarly, it follows that for a test stimulus to be
equally bright to a “white” reference stimulus of
constant luminance (20 cd - m~”), the luminance
of the test stimulus must be lowered by the factor
L/B.
The variation of B/L ratios between different
observers is fairly large. An example is given in
Figure 3(5.8.1).
The variation of B/L ratios for an individual
observer making brightness matches on different
occasions is not as large as that between different
observers. The range within which the B/L ratios
varied for observer GW with respect to test stimuli
601/1 to 7, who made brightness matches with
each stimulus on six different occasions, is indi-
cated by vertical lines in Figure 4(5.8.1). As the
chromaticity point of the test stimulus moves
farther away from the chromaticity of the “white”
(W) reference stimulus, the variability of the B/L
ratio increases. This is consistent with the general
report of the observers who find it more difficult
to make brightness matches when the chromatic
difference between test and reference stimuli be-
comes larger.
The second set of experimental data has been
taken from a paper by Wyszecki (1967) which
reports on an experiment conducted by the Com-
mittee on Uniform Color Scales of the Optical
Society of America (see also Section 6.6.4). In this
experiment, object-color stimuli were used made
up of colored ceramic tiles illuminated by either
natural daylight or an artificial daylight source.
The CIE 1931 (x, y)-chromaticity coordinates and
the luminance factor Y of each of the 43
object-color stimuli are specified, with respect to
CIE standard illuminant C, in Table 2(5.8.1).
Figure 5(5.8.1) shows the 43 chromaticity points.
A total of 76 observers located in different
laboratories in the United States and Canada
took part in the experiment. Thirty-nine observers
used 23 tiles, the other 37 observers used the
remaining 20 tiles plus three of the subset of 23 to
provide an overlap of experimental data. Each
colored tile was compared in lightness with a set
of 10 gray tiles which provided a scale of known
luminous reflectances ranging from Yy, = 28.12
to Yy,, = 52.46 in perceptual intervals that were
approximately equal in size. The observer was
asked to find, for each chromatic tile, the gray tile
that he or she perceived as being of the same
Table 1(5.8.1) Specification of Test Stimuli and Mean B/L -Ratios for a Group of 20 Observers
Who Made Direct Heterochromatic Brightness Matches Against a “ White” (W ) Reference Stimulus by
Viewing a 10° Bipartite Field Displayed in a “White” (S) Surround of 40° < 40° Angular Subtense
(after Sanders and Wyszecki, 1964)°

Stimulus Chromaticity Mean Stimulus Chromaticity Mean

Number X10 Yi0 B/L Number X10 Yio B/L

601/1 0.382 0.365 1.14 482/1 0.304 0.351 1.05

0.422 0.365 1.17 2 0.266 0.338 1:12
3 0.462 0.365 122 3 0.229 0.324 1.21
4 0.502 0.365 1.28 “ 0.191 0.310 1.30
5 0.542 0.365 1.32 5 0.167 0.301 133
6 0.582 0.365 1.36 475/1 0.326 0.352 1.06
7 0.613 0.365 1.35 2 0.296 0.326 1.09
583/1 0.360 0.372 1.10 3 0.265 0.301 jh
0.398 0.386 1.11 4 0.234 0.275 1.23
3 0.436 0.399 1.13 5 0.204 0.249 1.31
4 0.473 0.413 1.15 6 0.173 0.223 ay
5 0.510 0.427 1.14 450/1 0.322 0.330 1.16
574/1 0.372 0.391 1.07 2 0.302 0.296 1.20
2 0.403 0.416 1.07 3 0.282 0.261 1.26
3 0.433 0.442 1.08 4 0.262 0.277 1.31
4 0.464 0.468 1.14 5 0.242 0.192 1.37
565/1 0.352 0.382 1.05 6 0.222 0.158 bse
2 0.372 0.417 1.04 yi 0.202 0.123 1.39
he 0.392 0.452 1.05 SOF F 0.338 0.345 1.10
4 0.412 0.486 1.09 2 0.331 0.306 1.17
5 0.427 0.513 1.09 3 0.324 0.267 1.24
556/1 0.348 0.404 1.01 A 0.317 0.227
2 0.355 0.444 1.05 5 0.310 0.188 1.51
3 0.362 0.483 1.10 6 0.305 0.155 1.61
4 0.369 0.523 1.14 —541/1 0.348 0.326 1.16
5 0.374 0.553 Lt 2 0.355 0.286 1.28
541/1 0.338 0.385 1.03 3 0.362 0.247 1.42
2 0.331 0.424 1.06 - 0.369 0.208 1.61
* 0.324 0.463 1.11 5 0.374 0.177 1.83
4 0.317 0.503 1.15 matey Tt 0.352 0.348 1.14
5 0.310 0.542 1.20 2 0.372 0.313 1.21
S17/1 0.322 0.400 1.04 3 0.392 0.279 1.36
2 0.302 0.434 1.13 - 0.412 0.244 1.53
3 0.282 0.469 aE 5 0.432 0.209 bi3
+ 0.262 0.504 1.25 —500/1 0.372 0.339 1.16
500/1 0.326 0.378 1.03 2 0.403 0.314 1.24
2 0.296 0.404 1.10 3 0.433 0.288 1.38
3 0.265 0.429 1.16 “ 0.464 0.262 1.55
4 0.234 0.455 1.19 5 0.495 0.237 1.76
492/1 0.304 0.379 1.07 —492/1 0.360 0.358 ed tat
2 0.266 0.392 1.17 p2 0.398 0.345 1.18
3 0.229 0.406 1.19 3 0.436 0.331 1.26
* 0.206 0.415 1.21 4 0.473 0.317 1.32
487/1 0.322 0.365 1.04 5 0.511 0.303 1.46
2 0.282 0.365 1.10 6 0.548 0.290 1.61
3 0.242 0.365 1.18 7 0.586 0.276 1.71
4 0.202 0.365 1.24 8 0.620 0.264 1.75
5 0.174 0.365 1-27 Ref. (W) 0.342 0.365 1.07
Sur. (S) 0.294 0.335 —

“The B/L ratio is the ratio of the luminance B of the “white” reference stimulus and the luminance L of the test
stimulus when the two stimuli are at brightness match for the group of 20 observers. In arriving at the match, the
observers vary the B while L is kept constant at 20 cd - m~ *. The luminance of the surround (S) was also fixed at 20
oi ta v1ile

412
 Heterochromatic Brightness Matching of Complex Stimuli
are)
O Ole Oe 03 04 O05 O06 O07,
*10
Fig. 1(5.8.1). CIE 1964 (x49, ¥19)-chromaticity
diagram showing locations (solid dots) of test
stimuli used by Wyszecki and Sanders (1964) in
their study on heterochromatic brightness match-
ing. The chromaticity point of the “white” refer-
ence stimulus is given by an open circle at x1) =
0.342, y49 = 0.365, that of the surround is indi-
cated by a cross (8). Numerical data are given in
Table 1(5.8.1).
lightness as the given chromatic tile. If necessary,
the observer interpolated between two neighbor-
ing gray tiles. The luminance factor Y, of the
gray tile that matched in lightness the given chro-
matic tile of luminance factor Y was then used to
form the ratio Y,/Y. The mean values of the
Y,/Y-ratios for the two groups of observers are
given in Table 2(5.8.1).
As for the previous set of experimental data
given in Table 1(5.8.1), the data of Table 2(5.8.1)
can be illustrated effectively by means of loci of
constant values of Y,/Y ratio in the CIE 1931
(x, y)-chromaticity diagram. A family of such
loci is shown in Figure 6(5.8.1). The chromaticity
gamut provided by the daylight-illuminated col-
ored tiles is somewhat smaller than that which
was available on the colorimeter, but in the region
where the two gamuts overlap, the two families of
loci shown in Figures 2(5.8.1) and 6(5.8.1), re-
spectively, exhibit similar features. A quantitative
comparison between the two sets of data is, how-
ever, difficult and has not been attempted be-
cause of the marked differences in the observing
conditions used in the two experiments which
cannot be easily reconciled.
413
5.8.2 Additivity Failures
The additivity law in brightness matching (see
Section 4.2 on the principle of photometry, and
Section 5.2 on the classification of matching pro-
cedures) has been the subject of several investiga-
tions. Most of these investigations were motivated
by a concern over the validity of the present
standard system of photometry, governed by Eq.
3(4.2). A summary of the earlier work has been
given by Dresler (1953). More recent work in-
cludes the studies already listed in Section 5.8.1
and those of Boynton and Kaiser (1968); Guth,
Donley, and Marrocco (1969); Guth (1970);
Wagner and Boynton (1972); Guth and Lodge
(1973); and Kaiser and Wyszecki (1978).
108 In direct heterochromatic brightness matching,
the additivity law, also referred to as Abney’s
additivity law, fails substantially. The experimen-
tal evidence for its failure is conclusive and has
been demonstrated by a number of investigators.
An example will be given below. It is largely
because of this additivity failure that the photo-
metric principle governing our present standard
photometry is being reviewed by the CIE (1978b).
In order to comply with the basic equation of
photometry [see Eq. 3(4.2)], the photopic luminous
efficiency function V(A) must be determined by a
visual equivalence or matching criterion that leads
to visual responses obeying the additivity law.
Direct heterochromatic brightness matching does
not fulfill that condition, but minimum flicker
judgments and judgments of minimally distinct
border both do, at least to a close approximation
(Section 5.7.2 and Wagner and Boynton, 1972).
However, in practical photometry, both mini-
mum-flicker and minimally distinct border criteria
are unrealistic when the brightnesses of differ-
ently colored lights are to be compared and
specified so as to relate as closely as possible to
actual visual experience. In such cases, hetero-
chromatic brightness matching is the most
relevant procedure, and a mathematical model,
possibly of considerable complexity, that predicts
the responses of the visual mechanism to het-
erochromatic brightness comparisons must be
preferred over the simple (linear) model, char-
acterized by Eq. 3(4.2), which has been shown to
fail significantly in its prediction of heterochro-
matic brightness matches.
Experiments designed to test the additivity law
in brightness matching usually involve a bipartite
field, one half of which contains a fixed “white”
reference stimulus W, the other half a chromatic
stimulus C,. The observer is asked to equate the
414 VISUAL EQUIVALENCE AND VISUAL MATCHING

oh

0.7
500¢

0.4

0.3

0.2

0.|

0:3 0.4 0.5 0.6 0.7 08

Fig. 2(5.8.1). CIE 1964 (x19, 19)-chromaticity diagram showing loci of constant values of the
B/L_-ratio calculated for the group of 20 observers making heterochromatic brightness matches
of test stimuli defined in Figure 1(5.8.1) against a “white” (W) reference stimulus. The loci
were drawn freehand through the experimental data points given in Table 1(5.8.1).

brightness between the two stimuli in a direct
comparison by adjusting the radiance of C,. The
experiment is repeated with another stimulus C,
of a different color than C,. With C, brightness
matching W, and C, brightness matching W, it is
generally assumed that C, then brightness matches
C, (transitivity law). The crucial step in the ex-
periment is to ascertain whether the additive mix-
ture of C, and C,, produced in one half of the
bipartite field, will brightness match the “white”
reference stimulus whose radiance has been dou-
bled, that is, 2W.
When C, is a “blue” stimulus and C, a “yel-
low” stimulus, each of the same brightness as a
“white” reference stimulus W, the additive mix-
ture of the “blue” and “yellow” stimuli is found
to be Jess bright than the “white” reference
stimulus of twice its original radiance (Boynton
and Kaiser, 1968).
A variation of the above experimental proce-
dure has been applied by Guth and his coworkers
(Guth, Donley, and Marrocco, 1969; Guth and
Lodge, 1973). The test stimulus C is monochro-
matic and of wavelength A and is matched in
 Heterochromatic Brightness Matching of Complex Stimuli 415

2.5 hi doe Saat , wea lege ec 2.5

| x

ie
20 Soro. 4 3/3 cy2- 7k a 20
|
A
:
ate 15
x ou Oo

Og 10

(Oe)

“0
Fig. 3(5.8.1). Variation of B/L -ratios between different observers. Diagram shows B/L -ratios
for observers YN, GW, and WB in comparison with the mean ratios of the group of 20
observers recorded in Table 1(5.8.1). YN represents an observer with very high ratios, GW is
one with average ratios, and WB is one with unusually low ratios. The test stimuli are those of
the series 601/1 to 7 and 487/1 to 5 given in Table 1(5.8.1) and Figure 1(5.8.1). The
B/L-ratios are plotted as a function of the x,9-chromaticity coordinates of the test stimuli.

brightness against a “white” reference stimulus
W, again using a bipartite field. This match
establishes “unit amounts” of brightness both for
C(A) and W. Then, C(A) and 0.5W is additively
mixed in one half of the field and compared with
W remaining in the other half. If the two halves
do not match in brightness, the observer adjusts
the radiance of C(A) to produce a stimulus aC(A)
which in the mixture with 0.5W brightness
matches W; thus,
B If additivity would hold, the data points in
[aC(A) + 0.5W] =W
The main results obtained by Guth (1970) are
shown in Figure 1(5.8.2). The monochromatic test
stimulus C(A), added to 0.5W (a 2800 K “white”
stimulus), varies in wavelength A from approxi-
mately 430 to 630 nm as shown on the abscissa of
Figure 1(5.8.2). The number a of brightness units
of C(A) required to make [aC(A) + 0.5W] bright-
ness match with W is plotted as the ordinate. The
solid dots in Figure 1(5.8.2) are the mean values,
the solid lines indicate the range of a obtained for
nine observers.
Figure 1(5.8.2) would all coincide with the hori-

03 0.4 0.5 0.6

X10
Fig. 4(5.8.1). Range of variation of B/L-ratios for test stimuli 601/1 to 7 obtained for
observer GW making brightness matches on six different occasions. Open circles are the mean
ratios of the six individual determinations.
 {
C

Table 2(5.8.1) Specification of Colored Tiles and Equally Light Gray Tiles in Terms of CIE (x, y)-
Chromaticity Coordinates and Luminance Factor Y with Respect to CIE Standard Iluminant C.
Also Given are Mean Y,, / Y-Ratios (Wyszecki, 1967).”

Chromaticity Luminance Factor

Tile Coordinates Colored Tile Gray Tile Ratio
Number x y 4 Yn Yy/¥
| 0.2955 0.5321 30.99 42.20 1.36
2t 0.2816 0.4652 30.24 37.92 1.25
3° 0.3341 0.4813 30.08 36.20 1.20
4 0.3777 0.4976 29.08 36.46 1.25
$3 0.2520 0.4038 29.22 37.38 1.28
6 0.2922 0.4238 28.25 35.04 1.24
ic 0.3439 0.4416 28.14 33.16 1.18
8* 0.3838 0.4466 26.20 30.87 1.18
OTT 0.4379 0.4599 28.53 33.29 1.17
10* 0.2299 0.3383 28.59 36.20 127
11 0.2661 0.3603 28.16 35.04 1.24
122 0.3075 0.3801 28.00 33.41 11g
13 0.3544 0.3982 29.59 33.66 1.14
14* 0.3959 0.4100 28.75 31.83 1.11
ees 0.4415 0.4193 28.91 32.92 1.14
16 0.4764 0.4238 28.54 35.81 1.26
17 0.2258 0.2926 28.71 38.18 133
18* 0.2474 0.3082 28.58 35.81 1.25
19 0.2804 0.3248 28.35 33.29 1.17
20* 0.3196 0.3448 29.26 30.99 1.06
21 0.3581 0.3599 29.58 33;29 T3
22% 0.4026 0.3765 29.42 33.29 1.13
23 0.4426 0.3827 29.74 35.94 1.21
24* 0.4888 0.3934 28.12 35.68 127
25f 0.2342 0.2595 28.28 36.59 1.29
26 0.2632 0.2796 29.21 37.52 1.28
pb: 0.2978 0.3023 30.24 3474 1.05
28 0.3310 0.3211 28.42 30.64 1.08
29: 0.3713 0.3355 29.68 33.66 ls
30 0.4068 0.3488 29.68 35.68 1.20
31* 0.4580 0.3571 29.19 36.46 1.25
32 0.4864 0.3602 30.09 41.07 1.36
33 0.2512 0.2400 29.66 39:43 132
34* 0.2790 0.2626 30.25 35.68 1.18
35* 0.3058 0.2796 29.09 34.41 1.18
36 0.3398 0.2994 29.54 34.16 1.16
ad 0.3785 0.3157 28.86 34.66 1.20
38 0.4208 0.3240 29.60 31.18 1.28
39* 0.4580 0.3288 29.40 37.92 1.29
40 0.2960 0.2432 29.50 39.00 1.32
41* 0.3258 0.2576 28.74 36.07 1.26
42 0.3642 0.2708 28.84 38.05 ese
43** 0.3919 0.2790 28.34 37.92 1.34
N 0.315 0.322 ~ 28.12 to —
52.46

“Colored tiles marked with an asterisk (*) were matched in brightness with gray tiles by 39
observers, the other tiles by 37 different observers. Those marked with a double asterisk (**) were
observed by both groups (76 observers). The tiles (hexagonal in shape, 5 cm inside diameter) were
displayed on a table covered with a matte gray sheet of Munsell Value 6/(Y = 30) illuminated by
artificial or natural daylight providing an illuminance on the table surface of at least 500 Im-m 7’.
The observers viewed the tiles at a distance of 40 to 50 cm at an angle between 20 and 45° from the
vertical.
 Heterochromatic Brightness Matching of Complex Stimuli 417

.4 i 6 sh .8

J 450°

| se 5 3 4 5 6 yt 8
x
Fig. 5(5.8.1). CIE 1931 (x, y)-chromaticity diagram showing the chromaticity points of 43
colored tiles, illustrated by CIE standard illuminant C.

zontal dotted line of a = 0.5. This is obviously
not the case. Two types of additivity failure are
noted. For some wavelengths a value of a > 0.5 is
found, for others we have a < 0.5. Only in the
regions around A = 580 nm, and A = 490 nm, is
the value of a at or near 0.5. Values of a greater
and smaller than 0.5 suggest the presence of in-
hibitory and excitatory visual processes when a
heterochromatic brightness match is obtained by
the observer. The two types of additivity failure
have been called additivity failure of the cancella-
tion type for a > 0.5, and additivity failure of the
enhancement type for a< 0.5 (Kaiser and
Wyszecki, 1978). Guth (1970) referred to them as
subadditivity and superadditivity, respectively.
The experimental data described in Section
5.8.1, though obtained by a different procedure,
are to a large extent consistent with the data
obtained by Guth (1970). A reanalysis of the
earlier data, particularly those of Sanders and
Wyszecki (1964) and Kaiser and Smith (1972),
made by Kaiser and Wyszecki (1978), brings out
their similarities and the existence of the two
types of additivity failure.
Guth, Donley, and Marrocco (1969), in another
study on additivity of brightness matching, used
418 VISUAL EQUIVALENCE AND VISUAL MATCHING

| 2 3 4 > 6 We 8

.8 + .8

7 : a!

6 | 6

5006

nS BS)

4 .4

3 3
2 650
"7 70nm

2 WE

A |

O 380 4 me) mS if 8
x
Fig. 6(5.8.1). CIE 1931 (x, y)-chromaticity diagram showing loci of constant values of the
Y, / Y-ratio calculated for two groups of observers [Table 2(5.8.1)] making lightness matches
between colored tiles of luminance factors Y and gray tiles of luminance factors Y,. The
chromaticity points of the colored tiles are shown in Figure 5(5.8.1); the chromaticity point
common to the gray tiles is indicated by the solid dot at x, = 0.315 and y, = 0.322. The loci
are drawn freehand through the data points of Table 2(5.8.1; Wyszecki, 1967).

two equivalence or matching criteria: threshold
and direct heterochromatic comparison. In the
threshold experiments, the general procedure was
first to establish in a centrally viewed 0.75° field,
a subthreshold stimulus of wavelength A; and
then determine the radiance (or some other con-
venient quantity) of another stimulus of wave-
length A ; which, when added to the first stimulus,
brings the mixture to threshold. Both A; and 4,
were varied, and for any given A,, the set of A,’s
covered the visible spectrum and included A ;. For
each wavelength A,, the unit of brightness is de-
fined by the radiance of the stimulus of wave-
length A; required to elicit a threshold response.
With this definition, it is then possible to check
whether the threshold responses obtained for the
bichromatic mixtures of test stimuli (A;, A;) obey
the additivity law. As expected, the results re-
ported by Guth et al. (1969) clearly show pro-
nounced additivity failures when A; #A,. Figure
 Heterochromatic Brightness Matching of Complex Stimuli 419

et al. (1969) and their vector model predictions.

400 500 600 700
Wavelength \ (nm)
Fig. 1(5.8.2). Additivity failures in heterochro-
matic brightness matching measured by Guth
(1970) for nine observers. Solid dots: mean values;
solid lines: overall range. If data points would have
coincided with dashed line, additivity law would be
valid. For details, see text.
2(5.8.2) gives an example of the experimental data
obtained.
In following up earlier suggestions made by
Swets (1960) and Cohen and Gibson (1962), Guth
et al. (1969) applied a vector model to the experi-
mental data in an effort to predict their results. In
such a model, any monochromatic test stimulus is
represented by a vector whose length and direc-
tion characterize its brightness and wavelength,
respectively. The resultant brightness of addi-
tively mixed stimuli (A; and 4,) is characterized
by the vector sum. In general, there is fairly good
agreement between the experimental data of Guth
However, the vector model itself does not offer
any immediate clues as to the physiological na-
ture of the complex interactions between the cone
mechanisms which appear to play an important
part in the threshold determinations and which
manifest themselves in the additivity failures.
Guth et al. (1969) explain the additivity failures
by a physiological model designed within the
framework of the opponent-color theory (Section
8.3.4). ;
Increment threshold determinations constitute
the more general case of the absolute threshold
determinations discussed above, and additivity
failures for bichromatic mixtures set at increment
thresholds have provided further evidence for
interactions between the cone mechanisms
(Boynton, Ikeda, and Stiles, 1964; Kranda and
King-Smith, 1979). The increment threshold
method, widely considered a tool of major impor-
tance to the exploration of the visual mechanisms
by psychophysical means, will be treated more
fully in Section 7.4.
In the experiment using direct heterochromatic
brightness matching, Guth et al. (1969) used a
centrally viewed 1.5° bipartite field in a dark
surround. One half of the field displayed the
“white” reference stimulus (tungsten-lamp light)
at one of four luminance levels (0.31, 15.3, 31.8,
and 153 cd-m *”, respectively). Each of several
monochromatic test stimuli of wavelength A, was
displayed in the other half and brightness matched
against the reference stimulus by adjusting the
radiance of A;. The measured radiance L,, of the
brightness-matched test stimulus defined the

Wavelength A (nm)
Fig. 2(5.8.2). Mean results for three observers of an absolute threshold experiment conducted
by Guth et al. (1969) to check the additivity law. If C(525) is a monochromatic stimulus of
wavelength A = 525 nm, with a radiance that makes the stimulus just visible in the 0.75° field,
and C(A) is a monochromatic stimulus of wavelength A (selected from a set of 10), with a
radiance that makes that stimulus just visible, then the additive mixture aC(A)+90.5 C(525)
with a= 0.5 will be just visible if additivity holds. The experimental data indicate strong
failures of additivity (a > 0.5) for long wavelengths A.
420 VISUAL EQUIVALENCE AND VISUAL MATCHING

= | |

oO GBs

, % C (A)

i ve
Wee —— aa

05 a a aes

ens |
ee

400 450 500 550 600 650 700

Wavelength A (nm)

Fig. 3(5.8.2). Mean results for nine observers of a direct heterochromatic brightness experi-
ment conducted by Guth et al. (1969) to check the additivity law. If W is a fixed “white”
reference stimulus, C(521) a monochromatic test stimulus of wavelength A = 521 nm, C(A) a
monochromatic test stimulus of wavelength A (selected from a set of 10), and if C(521)
B B
brightness matches W and C(A) also brightness matches W, that is, C(521) = W, C(A) = W;
B
then aC(A )+0.5C(521) = W where a is equal to 0.5 if additivity holds. The experimental data
indicate strong failure of additivity except for the trivial case of C(A) = C(521).

brightness unit for A,;. In the next step of the
experiment, the observer was presented with an
additive mixture that consisted of A; and A;
(another member of the same set of test stimuli),
fixing the radiance of A; to half its unit value
(L,,.). The radiance of A; was adjusted by the
observer until an equality of brightness between
the mixture and the “white” reference stimulus
was seen. The similarity between this experiment
and the one described in connection with Figure
1(5.8.2) is apparent.
An example of the results obtained by Guth
et al. (1969) -is illustrated in Figure 3(5.8.2). It
refers to the case where monochromatic stimuli of
various wavelengths A, were added to a mono-
chromatic stimulus of wavelength A; = 521 nm.
Strict obedience to the additivity law would have
resulted in data points lying on a horizontal line
of a = 0.5.
A statistical analysis applied by Guth et al.
(1969) to their data yielded no significant effect
of the luminance level of the “white” reference
stimulus on the additivity failure. The observed
additivity failures were of the two types described
earlier: additivity failure of the cancellation type
and additivity failure of the enhancement type.
The latter type is, in general, not observed in the
threshold experiment, as illustrated in Figure
2(5.8.2).
5.9 ABNEY AND BEZOLD-BRUCKE EFFECTS
The apparent hue of a color stimulus is, in gen-
eral, modified if, starting from a monochromatic
stimulus or a purple, an achromatic stimulus is
added to it to decrease progressively the colori-
metric purity (Section 3.4). The effect, called the
Abney effect (Abney, 1910), is manifest in the
choice of samples (chips) in color-appearance sys-
tems, particularly the Munsell System (Section
6.6.1). In a CIE chromaticity diagram showing the
chromaticities of the Munsell color chips of con-
stant Munsell value, the chromaticity loci of chips
to which the system assigns the same hue are (for
most hues) curved lines joining the achromatic
point to a point on the spectrum locus or the line
of purples [see Figures I(6.6.1) to [X(6.6.1)]. For a
particular yellow at about 570 nm and a particu-
lar bluish purple (x = 0.240, y = 0.035, ap-
proximately), the hue locus is a straight line,
indicating no effect of chroma or saturation on
hue.
Other examples of experimentally determined
loci of constant hue in the chromaticity diagram
have been provided by MacAdam (1950b, 1951),
Wilson and Brocklebank (1955), and Robertson
(1970).
Robertson (1970) used three methods of hue
matching with the aid of Wyszecki’s Seven-Field
Colorimeter (Section 5.15). In the first method,
three of the seven fields were used, whose centers
formed the corners of an equilateral triangle, with
each 3° hexagonal field having one side adjacent
to each of the other two. The three-field array had
a large white surround of 6 cd - m ”.
In two of the fields, the observer was shown
color stimuli of the same hue and brightness
(luminance of 12 cd - m~*) with high but slightly
different saturations. The task was to use the
 Abney and Bezold— Briicke Effects 421

controls of the third field to adjust its stimulus so
that the hue and brightness were the same as for
the two given stimuli, and the saturation a little
less. After recording the settings, the observer
arbitrarily altered the stimulus in the first field of
highest saturation and replaced it by a fourth of
lower saturation than the third, while still main-
taining the same hue and brightness. This
sequence was continued with stimuli of progres-
sively lower saturation until one was reached
which appeared to the observer to be achromatic.
Eight different hue lines were determined in this
step-by-step way; these are identified by the ap-
proximate dominant wavelength (in nm) of the
most saturated starting stimuli, which were 435,
485, 510, 565, 575, 585, 600, and the complemen-
tary wavelength of 505 nm.
The second method was the same as the first,
except that a fourth field was added to the stimu-
lus array. The chromaticity of the stimulus in the
fourth field was fixed and the same as that of the
white surround; its luminance was set equal to
that of the saturated starting stimulus of 12 cd -
m ”.
The addition of the white fourth stimulus
helped the observer in moving along a constant
hue line that ended near the target achromatic
point. When using the first method, the user had
difficulties reaching a common achromatic point
for the different hue lines. However, neither of
the two methods proved to be completely satisfac-
tory, as simultaneous contrast effects, chromatic
adaptation, and gradual shifts away from the
starting hue as a line is determined, appeared to
affect adversely the precision of the experimental
data, and observer differences were also fairly
large. Figure 1(5.9) illustrates the mean results for
three different observers.
In order to avoid particularly the problem of
gradual hue shift, a third method was tried. In
this method, only two adjacent fields were used.
The first field contained a fixed stimulus of the
starting hue and of high saturation, and in the
other, the observer was shown a series of stimuli
of different saturations and of various hues close
to that of the fixed stimulus. These stimuli were
presented in random order and the observer was
required to state, for each one, whether it fell to
one side or the other of the hue line through the
given stimulus. The observer did not adjust the
controls of the instrument. A stimulus was taken
to lie on the hue line if the observer placed it
sometimes on one side and sometimes on the
other in repeated viewings. This method proved
to be considerably better than the previous two
step-by-step methods. Both the repeatability of

0.350

0.300

0.250

0.150 0.200 0.250 0.300 0.350

Fig. 1(5.9). Mean constant-hue lines derived by Robertson (1970) for three observers
(ARR, HW, GW), plotted in the (now obsolete) CIE 1960 UCS diagram whose coordinates are
defined in terms of CIE 1931 (x, y)-chromaticity coordinates by u= 4x /(—2x +12 y +3) and
v=6y/(—2x+12y +3). A step-by-step hue-matching technique was used, starting in each
case with a stimulus of high saturation, a luminance of 12 cd-m ~?. and known dominant
wavelength (600, 585 nm, etc.) and then progressing towards an achromatic stimulus. In moving
along a line of constant hue, brightness was kept constant. The cross in the diagram indicates
2
the chromaticity of the “white” surround used in the experiment, kept at 6 cd-m' “.
422 VISUAL EQUIVALENCE AND VISUAL MATCHING

one observer and the variability between different
observers was much improved. However, the third
method was not used extensively to generate many
data, because the time required to determine a
single hue line was rather long.
The Abney effect is of importance also in
connection with chromatic response functions, to
be discussed in Section 5.13.
The effect of luminance on hue, known as the
Bezold—Briicke effect, is not so clearly shown in
the current diagrams of color appearance systems,
but direct measurements by Purdy (193la) make
clear the direction and approximate magnitude of
the effect for monochromatic stimuli. Purdy’s
measurements were hue matches (in a 3° bipartite
field) between a monochromatic stimulus of fixed
wavelength and luminance, and a second mono-
chromatic stimulus whose wavelength was varied,
the luminance being kept at a constant value
considerably lower than for the first stimulus. The
excess of the wavelength A,, of the higher lumi-
nance field over that of the lower luminance field
A,, when the perceived hues were as nearly as
possible the same, defined the Bezold—Briicke hue
shift (Ay —A,). The values obtained for the
luminance pair, yielding a ratio of retinal illumi-
nance of 1000 to 100 td, are shown in Figure
2(5.9). In addition to the three spectral hues
showing no Bezold—Briicke shift, Purdy de-
termined a fourth reddish-purple hue with the
same property.
Boynton and Gordon (1965) repeated Purdy’s
(1931) experiment of direct hue matching in a
steady-state bipartite field and added two other
experimental procedures. The first of their addi-
tional experiments was a modification of Purdy’s
procedure in which the observer was asked to
make hue matches on the basis of 300-msec pulses
of the bipartite field with 9-sec dark intervals. In
the second experiment, a color naming method
was used employing a 3° circular field centrally
fixated, superimposed upon a tiny fixation point.
The monochromatic stimuli were presented in
300-msec pulses providing either 1000 or 100 td,
with dark intervals of 12 sec. The observer named
the hues in terms of four color names: red (R),
yellow (Y), green (G), and blue (B), and was
allowed to assign also intermediate hue names
such as yellow-red (YR) or red-yellow (RY) with
Y being the predominant cardinal hue in (YR)
and R the predominant cardinal hue in (RY).
Thus, the observer had 12 possible choices of
naming a hue: BR, B, BG, GB, G, GY, YG, Y,
YR, RY, R, and RB. Each judgment made by the
observer was assigned an arbitrary score value of
3. When the judgment was described by a cardi-
nal hue name, such as red (R), the entire score
value of 3 was assigned to it. When intermediate

eee 1000 td

100 td
3°

Bezold-Brucke
(nm)
(A,-A,)
shift
hue
450 500 600 650

Wavelength i, (nm)
Fig. 2(5.9). Change of apparent hue of monochromatic stimuli corresponding to a reduction in
retinal illuminance 1000 to 100 td. Extract of the data of Purdy (1931b) who made hue matches
in a 3° bipartite field between a stimulus of fixed wavelength A ,, and of a luminance providing
1000 td, and a second stimulus of variable wavelength A, and of a luminance providing 100 td.
The Bezold—Briicke shift (A ;, — A) is plotted as a function of A,,.
 Abney and Bezold - Briicke Effects 423

Table 1(5.9) Example of Scoring Procedure Used in Color Naming Experiment by

Boynton and Gordon (1965) for Observer JG, Monochromatic Stimulus at 520 nm,
1000 td’
Score Value Assigned to Color Names
Blue Green Yellow Red
Judgment Frequency (B) (G) (Y) (R)
BG 5 10 a 0 0
GB 9 9 18 0 0
G 6 0 18 0 0
GY 5 0 10 - 0
Total 25 19 51 5 0

“The 520 nm stimulus was never judged to be BR, B, YG, Y, YR, RY, R, RB.

hue names were used, such as red-yellow (RY), a considered possible. The same authors also con-
value of 2 was assigned to the first (R) and a cluded from their experimental data that mono-
value of | to the second (Y) name. For each level chromatic “unique yellow” (approximately 578
(1000 and 100 td), every monochromatic stimulus nm) and monochromatic “unique blue” (ap-
of wavelength A was judged a total of 25 times in proximately 478 nm) showed either no variation
a randomized presentation cycle, providing a total
score value of 75 distributed among the four
cardinal hue names. Table 1(5.9) gives a typical
example of the scoring procedure. ya ae |S ee OTe Se | ee eT

The mean results obtained by Boynton and 40 —o—o—o— hue matching in steady-state viz
Benge hue matching of pulses y Aas
Gordon (1965) for three observers determining = 30 —.—x—»— hue naming of pulses A ike
Bezold—Briicke hue shifts by three different ex- THESES 4 i]
perimental methods are illustrated in Figure o ©
3
ylog /

3(5.9). TIO we ign:

< 0. gis GaN
The data obtained by the steady-state hue Sz
< 0 A arMt
oa OL
pea wh reheat
matching method are directly comparable with 7 ae Lae See a ae
ok eA Re ye
those obtained by Purdy (1931), shown in Figure Oo JL
@ -20} : wie
2(5.9). The agreement between the two sets of a ii
-30F °
data is quite good. When the experimental proce-
dure is changed, differences in the results are -40

noted which Boynton and Gordon (1965) attri-
bute to simultaneous contrast effects whose im-
portance varies with the presentation time of the
stimuli. The authors also discuss possible connec-
tions between stimuli for which zero shifts of hue
are observed, that is, (A;, — A,,) = 0, and stimuli
that have been found in other studies to exhibit
the psychologically “unique hues” red, yellow,
green, and blue. Some correspondences are noted,
except for the red.
In a later study by Larimer, Krantz, and
Cicerone (1974), in which the Bezold—Briicke ef-
fect was explored further, it was suggested that
the hue-naming data of Boynton and Gordon are
likely to have been biased somewhat by the scor-
ing method used. In particular, a bias toward a
shorter wavelength of the “unique blue” was
meSieme A 88A hse Mt bena CE Ss et Se
450 500 ate 600 650
Wavelength A,, (nm)
Fig. 3(5.9). Change of apparent hue of mono-
chromatic stimuli corresponding to a reduction in
retinal illuminance 1000 to 100 td, determined by
Boynton and Gordon (1965) for three observers
and three different experimental procedures. The
mean of the three observers for the Bezold—Briicke
hue shift (A,, — A,) is plotted as a function of
Ay, the wavelength of the monochromatic stimu-
lus held at a retinal illuminance of 1000 td. The
second stimulus of wavelength A, was held at 100
td. The data obtained by the steady-state hue-
matching method are directly comparable with
those of Purdy (1931b) shown in Figure 2(5.9).
424 VISUAL EQUIVALENCE AND VISUAL MATCHING
or virtually insignificant variation over a lumi-
nance range of one to two log units; within that
range, these stimuli yield invariant hues relative
to the Bezold-Briicke shift. In a subsequent paper,
the authors (Larimer et al., 1975) also found that
the monochromatic “unique green” (approxi-
mately 500 nm) stimulus yields an invariant hue.
“Unique red” was obtained for nonmonochro-
matic stimuli consisting of additive mixtures of
the shortwave and longwave ends of the visible
spectrum. The additive mixture required for the
stimulus to remain “unique red” must be varied
with luminance; otherwise, a stimulus perceived
to be “unique red” at a low luminance level,
turns bluish-red when the luminance is increased.
The study by Nagy and Zacks (1977) dealt
with the zero shifts of hue and the measurement
procedure used to assess Bezold—Briicke hue
shifts. These authors concluded that the stimulus
presentation time was the more important vari-
able in accounting for the differences between
different procedures.
In accordance with the opponent-colors theory
(see Sections 5.13 and 8.3), the independent
“red-green” and “yellow-blue” chromatic re-
sponse systems (Section 5.13) are functions of
luminance and exhibit differential response rates
that are consistent with the observed Bezold-
Briicke hue shifts (Hurvich and Jameson, 1955).
In particular, it is hypothesized that blueness and
yellowness responses increase with luminance
more rapidly than redness and greenness re-
sponses. However, this explanation does not ap-
pear to be entirely satisfactory with regard to
recent observations of hue shifts for very brief
pulses (17 ms) of stimuli varying in luminance
(Nagy, 1980). Under such conditions, the
“unique” hues no longer exhibit zero shifts, and
in some instances, a reversal in hue shift direction
occurs at high luminance levels.
There are several other studies of the Bezold—
Bricke effect, of which the following are of par-
ticular interest: Walraven (1961), Coren and Keith
(1970), Yager and Taylor (1970), Cohen (1971),
Savoie (1973).
5.10 HUE REVERSALS: BRINDLEY
ISOCHROMES
The CIE 1964 large-field chromaticity coordinates
[see Table I(3.3.2)] which extend to the wave-
length 830 nm show that beyond 700 nm the
Table 1(5.10) Brindley Isochromes
Wavelengths in nm
Shortwave member 641 652 674 679 688
Longwave member 887 850 786 749 7I1
changes in x,9(A), Yio(A), [Z19(A) is zero] reverse
their direction, which means that the monochro-
matic stimuli become slightly yellower. This result
corresponds to an observation first made by
Brindley (1955b) using a 4° field and is of consid-
erable theoretical interest although unimportant
in practical applications. Brindley’s isochromes,
that is, pairs of wavelengths in the red end of the
spectrum that have the same chromaticity, are
given in Table 1(5.10) and are plotted in Figure
1(5.10) together with the longwave chromaticities
from the 10° matching results of Stiles and Burch
(1959), see also Section 5.5.4, which substantially
confirm them. One implication of Brindley’s re-
sult is that in a plot of the logarithm of the
spectral sensitivity (expressed on a quantum ba-
sis) against wavenumber, the gradients of the
curves for the red and green mechanisms do not
tend to the same constant value as the wavelength
is moved deep into the red. A theoretical interpre-
tation of this deviation from the simpler behavior
commonly assumed has been advanced by Lewis
(1955).
5.11 STILES-CRAWFORD EFFECT
The term directional sensitivity is now commonly
used, in a visual context, to denote variations in
visual sensitivity that are observed, not when
stimuli arrive from different directions in the
visual field to be imaged on different retinal areas,
but when stimuli enter the eye through different
parts of the eye pupil and are imaged on the same
retinal area. Both brightness-matching and incre-
ment threshold methods yield quantitative mea-
surements of directional sensitivity. The value of
the directional sensitivity at a pupil point of entry
P may be defined as the ratio
Le —[1(5.11)]
where Sp and Sp are the same magnitudes produc-
ing the same visual effect on the given retinal area
when entry is, respectively, at P and at a fixed
reference point of entry P). Experiment shows
 Stiles — Crawford Effect 425

+.10

+.05

o
TAN
——
ie
8 Oo

Bo TAO.
oe
Ses
oom
alee
O

—.05 Brindley lsochromes

Short wave Long wave

E=i10
16000 15000 14000 13000 12000 cm —}
Wavelength or Wavenumber
Fig. 1(5.10). Observations of hue reversal in the deep red. The curves are the mean g(A) and
b(A) 10° chromaticity coordinates referring to primaries at 15,500, 19,000, 22,500 cm! in a
W.D.W. system, with normalizing wavenumbers at 17,250 and 20,500 cm! (Stiles and Burch,
1959). In making the comparison with Brindley’s isochromes, the very small contribution of the
b(A)-chromaticity coordinate is ignored, so that the Brindley isochromes may be taken as
having the same g(A)-chromaticity coordinate. The short-wave member of each pair is placed
on the g(A)-curve at its correct wavelength, and the long-wave member is then plotted to have
the same g(A )-value [see Table 1(5.10)].

that: the variation of 7 with the distance r in millime-

(a) For most eyes studied, the response of a
foveal rod-free area is maximal when P is at or
near the pupil center and is reduced to about a
quarter when P approaches the pupil edge.
(b) For extrafoveal (parafoveal) retinal areas,
the directional sensitivity resembles that for foveal
vision, provided the conditions are such that the
visual response measured is a cone response.
However, when the measured response is attribu-
table to the rod mechanism, the directional
changes are very much reduced and the response
may be regarded as nearly nondirectional.
The following empirical formula was intro-
duced by Stiles and Crawford (1933) to represent
ters between P and Py (taken to be the point of
entry for maximal 7)
n= 10?” [2(5.11)]
where the constant p is a convenient measure of
the directional effect. Figure 1(5.11) illustrates the
agreement between predicted and observed values
of n. The effect characterized by 7 is commonly
referred to as the Stiles—Crawford effect of the first
kind and sometimes abbreviated by SCE.
For foveal vision, p has a value in the neigh-
borhood of 0.05. Few data on the variation of p
with the wavelength of the stimulus are available,
but p appears to be rather larger in the blue and
red than in the green, yellow, or orange. Table
426 VISUAL EQUIVALENCE AND VISUAL MATCHING

—0.25

Loe,

—0.5

—0.75

—1.0

NASAL TEMPORAL

Fig. 1(5.11). Mean directional sensitivity curve (fovea) showing its approximate representa-
tion by the formula n = 10 ”” . Observed values are often higher at the extremities than the
computed curve fitted in the central region, as is the case here (Stiles, 1937a).

1(5.11) contains a set of measured values, ob-
tained by the threshold method for one observer,
illustrating the range of variation of p with A that
have been observed (Stiles, 1939). Alternate em-
pirical formulae for n have been proposed but
offer little, if any, advantage over Eq. 2(5.11) (see,
for example, LeGrand, 1957). Representative
mean data based on a large group of observers
have yet to be obtained.
If the visual effects of rays incident through
different parts of the pupil and imaged on the
same foveal area of the retina are additive, a
 Stiles— Crawford Effect 427

Table 1(5.11) Variation of the Value of p with Wavelength X in the

Expression n = 10 ” "* for the Directional Sensitivity of the Retina
(Stiles, 1939)

Wavelength 4 (nm) 430 500 540 580 620 660

Coefficient p 0.067 0.060 0.048 0.046 0.052 0.056

directional sensitivity factor o can be obtained by
integration of the expression 2(5.11). For a uni-
form beam entering the eye through a circular
pupil with its center at P) and of radius R in
millimeters, the total radiant flux S(R, 7) ad-
mitted by the pupil and appropriately weighted,
at each elementary area (r drd¢) of the pupil, by
the directional sensitivity, 7 must be proportional
to
a(1 — 107?”")
legs nr drdo=
r=0 Plog,10
with 7 given as in Eq. 2(5.11). The corresponding
total radiant flux S(R) admitted by the pupil but
not weighted by n, is proportional to the area of
the pupil, that is, 7R?. The directional sensitivity
factor o is then defined as the ratio of the two
fluxes, S(R, 7) and S(R), and equals, for differ-
ent values of p,
= 1 — 107?®” [3(5.11)]
PR‘ log, 10
and is computed in Table 2(5.11).
Stimuli entering the pupil at points displaced
from the center (strictly, displaced from point
Py), appear of slightly different color and not
merely of reduced brightness. This is commonly
referred to as the Stiles—Crawford effect of the
second kind and abbreviated SCE2. The magni-
tude of the color changes is illustrated by the data
of Figure 2(5.11) which show the changes in
stimuli when the point of entry is shifted from Pp
to a point 3.5 mm temporal (Enoch and Stiles,
1961). There is a hue shift with a characteristic
variation through the spectrum (inset figure) and
an increase in saturation in the blue-green.
Directional sensitivity is predominantly retinal
in origin; that is, it corresponds to changes in the
visual response of a retinal area as the retinal
angle of incidence of the stimulus is altered; but
differential light losses in the optic media experi-
enced by rays entering through different points of
the pupil may modify the effect observed in the
intact eye. As emphasized by Weale (1961), the
lens with its diminishing thickness from center to
edge and its high absorption (or scattering) at
short wavelengths must be expected to have a
considerable effect on the observed directional
sensitivity for spectral stimuli in the blue and
violet. The modification is in the sense of a smaller
observed variation of sensitivity than would be
expected from the purely retinal effect.

Table 2(5.11) Directional Sensitivity Factors o for Centered Circular Pupils

Pp 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10

R Values of o

0.5 0.997 0.994 0,991 0,988 0.986 0.983 0.980 0.977 0.974 0.972
1.0 0.988 0.977 0.966 0.955 0.945 0.934 0.924 0.913 0.903 0.893
1.5 0.974 0.949 0.926 0.903 0.881 0.859 0.839 0.819 0.799 = 0.781
2.0 0.955 0.913 0.874 0.836 0.801 O:768 st-0-737 O70 T7ie® 0.680." 205653
2.5 0.931 0.869 0.812 0.760 O713 40 O670-41.0.630) 60,594 .- 50561 0.530
3.0 0.903 0.819 0.745 0.680 0.623 0.572 0.528 0,488 0.453 0.422
3.5 0.871 0.764 0.675 0.600 0536 0.4825 .0:436., 0.398. . 0:363 . 0333
4.0 0.837 0.708 0.605 0.523 0.457 0.403 0.358 0.321 0.291 0.265
4.5 0.799 0.650 0.538 0.453 O367 #10 33665 0.205 5 262; 0235 0.212
5.0 0.760 0.594 0.476 0.391 0.328 “O20 e0244@" 215 0.192 ~. 0.473

R = radius of pupil in mm,

p= directional sensitivity parameter in 7 = io-*",
o =(pR* log, 10)(1 — 10-?*").,
 ‘(196T ‘S9INS pue ysoug wWory) “WAL 1eAsasqOC “WU g°eGp YIsuajaawm 1O¥
Pays] Sv pouluLiajop si JTS any JUsIedde sy] ‘a.MBIy JoSUT dy) UT UMOYS A]IVa][D d10UI Je
pure ‘sns0] umnydeds ay} SuoTe AjAvaU ore S}JTYS 1O[OD dy) ‘UOBINJeS Sasva.1DUI sdUIpTOUT anbI[qo
d19YM “Ud0I3-9N]q oY) UT Jdodxq *(a[SUe OQ] JNOqe) ddUapIOUI aNnbITqO 10J JyIYS AJIOWeUIOIYD
ay} MOYS sjutod sso1d 0} sjop sutuIof souT]T oy) pu JdUaPIOUT [eULIOU JO} SaTOHeUIOIYD
dy} JUdSIIda1 SJOP POS sy} ‘WIBASLIP APIOQVUIOIYS Sy) UT ‘BUSI [BdAOJ JY} UO SN[NUIS ay) Jo
QOUIPIOUI JO J[SuL dy} UT BSULYD YIM TPNUIYS IVUIOIIOUOU JO 1O[OI JO asuKYyD =*({]'S)Z “BY
Ayioipowosyos - 4
2.0-—
20
v0
9°0
80
Ol
oO
wu OOL
F ed ne eg tee a ae OP fe)
oso Xx GeSp
09b
sz9
029
NY
\
OID NG
re)
(909 \
NS
009 \y
SS
NS
3.LIHM
LNIOd
\ (SISV@ HLON313AVM)
: ADYINZINOA

6
O6S
vO
:
2788S \\ oO .

—-
\ \

Ayiotpyowosyo
NS NN
\
\
~
X
\
XY x
N
\ Ss 9°0
ILS WN og
OLS NS .
‘
);-W2 YAIGWANSAVM ~
ooos! ooo!
g9<gscs NN < 8°0
oss SS wug ¢6b LV
1J3IHS 30H S
IN3uVdd¥
‘\ 00S
S°OvsS »s
NX
>
~
(awe) O£S \e> "1
afte rere L4IHS 3NH LN3YVdd¥ 92s Fols

N co
SEIS “QAO

zw
—_
wu 009 Oss 00S OSb 00

Reviews of work on directional sensitivity and
attempts to explain the phenomenon by means of
an appropriate physiological model have been
published by Vos (1960), Stiles (1962), Enoch
(1963), Crawford (1972), Snyder and Pask (1973),
Starr (1977), Fuld, Wooten, and Katz (1979),
Bour and Verhoosel (1979), and Kinnear et al.
(1979); they also contain most of the other rele-
vant references.
A directional effect of a quite different kind
was discovered by Campbell (1958), who showed
that visual acuity for light incident obliquely on
the retina is considerably lower than for normally
incident light, provided the object details to be
resolved are separated in the plane of incidence.
The effect is, in part, of retinal origin (Campbell
and Gregory, 1960).
5.12 CHROMATIC ADAPTATION
The term chromatic adaptation is applied in rela-
tion to modifications of visual response, particu-
larly the response to chromatic test stimuli,
brought about by chromatic conditioning (adapt-
ing) stimuli that are surrounding or pre-exposed.
The response may be studied by determining
discrimination or increment thresholds (see Sec-
tion 7.4), but quantitative investigations have been
concerned predominantly with measurements of
luminance and chromaticity by asymmetric
matching. Despite much observational material
obtained under diverse conditions, few general
principles of chromatic adaptation have emerged.
The following notes sketch the preliminaries to
any unified scheme based on asymmetric match-
ing and discuss the salient points of the work
done by a number of investigators.
5.12.1 Asymmetric Matching—
Basic Concepts
The phenomena of chromatic adaptation are in
full play when the test stimuli are imaged on the
foveal, rod-free area of the retina. It is a simplifi-
cation, adopted in most experimental work, to
deal only with this case. The basic concepts that
underlie such experiments have been formulated
by Stiles (1961).
Two stimuli {P, dA} and {Px dA} applied to
foveal areas O and O’, whose luminance and
chromaticity are to be compared under different
adapting conditions, are then specified by their
Chromatic Adaptation 429
tristimulus values
X= [P,x(A) dd
Y= [P,y(A) dd
Z = [P,z(A) da
and
X’ = [PKx'(A) dd
y’ = [Pxy'(A) da
Z’ = fPxz(A)dd [1(5.12.1)]
where the sets of color-matching functions x(A),
y(A), Z(A) and x’(A), yA), z’(A), which in bi-
nocular asymmetric matching refer to stimuli
imaged on foveal areas of the left and right eyes,
respectively, may be different. In fact, they may
be different even if only as a result of prereceptor
pigment differences, and thus they are not
transformable one into the other by a linear
transformation. Although in some applications
the differences may be unimportant, the possible
distinction between the sets of color-matching
functions is retained.
The various investigations of chromatic adap-
tation indicate that the matching of the stimuli
in differently adapted retinal areas is three-
dimensional in the sense that for a great many
stimuli {Px dA}, whose tristimulus values corre-
spond to points filling a considerable domain of
the (X’, Y’, Z’)-space, a satisfactory match can be
made with a comparison stimulus {P, dA} con-
sisting of a mixture in suitable proportions of
three, well-chosen fixed primaries of high satura-
tion, whereas the use of just two primaries will
not serve.
It may be assumed therefore, that there are
three matching conditions which may be written
in the following functional forms:
il Kh, Zed eel Xoo Zt As)
UaleX ke Ci Wee a Ah 2 A)
Welk li ead slay 2 sad)
[2(5.12.1)]
430 VISUAL EQUIVALENCE AND VISUAL MATCHING
where A and J’ represent, respectively, for O and
O’, the values of all the variables that may con-
tribute to determining the adaptation of the
stimulus area. However, it must be emphasized
that, with widely different adaptations A and 4’,
the number of stimuli that cannot be matched
with a given three-primary mixture may be
very great; their tristimulus points may occupy an
extensive domain in the (X’, Y’, Z’)-space. Super-
ficially, this resembles the impossibility of match-
ing certain too saturated stimuli in ordinary sym-
metric matching. However, the unmatchable
stimuli in asymmetric matching are not neces-
sarily well described by the expression “too
saturated,” and the indirect method of deriving a
match by desaturating the test color with quanti-
ties of one or two of the matching primaries is
questionable because it assumes that matching is
additive. Moreover, an important difference from
symmetric matching is that some test stimuli
{Px dX} may be not only unmatchable by any
comparison mixture of three fixed primaries, but
also by a comparison stimulus {P, dA} of any
spectral composition whatever. Striking examples
of these points have been brought out in the work
of Hunt (1953) and Bartleson (1977, 1978a, b).
The conditioning variables (A or A’) for a test
area (O or O’) must include the tristimulus values
for that test area (X,Y, Z or X’, Y’, Z’) and all
the parameters of surround and pre-exposed
stimuli applied to the retina in which the test area
is located. They may also have to include the
tristimulus values and adaptation parameters for
the other test area. At this point, the different
methods of asymmetric matching must be dis-
tinguished.
The crudest procedure, a form of color naming
or memory method, was applied in studies of
chromatic adaptation by Bouma and Kruithof
(1947) and Helson, Judd, and Warren (1952). The
observers, in preliminary training, familiarized
themselves with the appearance of the samples of
a color atlas, exposed under adaptation condi-
tions A. In the main experiments, the observers
‘had to name, by long-term memory, the sample
corresponding to the perceived color of the test
object exposed under the test conditions A’. This
method is a more systematic application of
the purely descriptive procedure used in many
demonstrations of chromatic adaptation. In
asymmetric matching of the kind just described,
the stimuli {P, dA} and {Px dA} are applied to
the same retinal area (O = O’) but on entirely
distinct occasions; A and J’ are the two sets of
values assumed on these occasions by a single set
of conditioning variables. In these circumstances,
asymmetric matching is transitive; that is, if
stimulus (X, Y, Z) presented under conditions A
matches (X’, Y’, Z’) presented under conditions
Aipvand «ify (4, Y..Z) Wunder) Aemmatcies
(X”, Y”, Z”) under A”, then (X”, Y”, Z”) under
A” matches (X, Y, Z) under A.
In binocular color matching, on the other hand,
the response to (X, Y, Z) in area O in the left eye
could depend not only on the adapting stimuli
applied to the left eye, but also on all the stimuli
applied to the right eye; in effect, A would repre-
sent a double set of parameters. Fortunately for
this method, the evidence is that such interference
of stimuli in one eye on the response in the other
eye is slight. However, in monocular matching in
a bipartite field with differential adaptation of the
two halves, as used by MacAdam (1956), it is
questionable whether a pre-exposed adapting
stimulus applied to one half (retinal area O) has a
negligible effect on the condition of the juxta-
posed half (retinal area O’). To the extent that
this occurs, the matches made represent the hy-
brid effects of two adapting stimuli on perception
in each of the half-fields, and further interpreta-
tion is made more difficult. In any study of
chromatic adaptation by asymmetric matching,
tests of the transitivity principle should show
whether there is transfer to one matching area of
the effects of stimuli used to adapt the other.
The important linearity laws (proportionality
and additivity) are not generally true for asym-
metric matching, although within the limits of
some investigations confined to certain pairs of
conditioning stimuli they have been found to be
approximately valid. A full discussion of asym-
metric comparison and matching and associated
basic laws has been given, in more general terms,
in Section 5.2.2.
The scheme of Table 1(5.12.1) shows the rela-
tion of the linearity laws for symmetric color
matching, which are presumed to hold for matches
between stimuli exposed under common adapta-
tion conditions A (or A’), to the linearity laws for
asymmetric matching when the stimuli are ex-
posed under different adaptation conditions, A
and A’, respectively.
The implications of obedience to the linearity
laws for asymmetric matching have played a vital
part in the study of chromatic adaptation. If they
hold for adaptations A and A’, the tristimulus
values X, Y, Z and X’, Y’, Z’ at match are related
by a constant linear transformation with the 3 < 3
 Chromatic Adaptation 43]

Table 1(5.12.1) Linearity Laws

Stimuli Matched under

Asymmetric Adaptation
Stimuli Matched under Conditions Stimuli Matched under
Adaptation Conditions left-hand right-hand Adaptation Conditions
A A’ A’

If A, => By By => Gy C; = D,

A, = By By => Cy ce => Do,

then (i) | aA, = «B, aC, =aD,

Fisldedeariaed sy deen, tical?) (Cy a C2) = (D, ste Dz)
Linearity Laws for Symmetric Matching
and i) aC, = “D, aA, = «B,
a Co Cee Dre Ds (A, aF Ap) — (B, =F B,)

Persistence-of-Color Matches under Different Symmetric Adaptations

but aby
(ill)
aB, = «aC,
not necessarily,
Linearity Laws for Asymmetric Matching

— . The symbols 4, As, B, and so forth, represent stimuli generally of different spectral compositions.
2. « is any positive number.
3. “=” denotes “‘matches under symmetric condtiions.”’
‘‘=" denotes ‘“‘matches under asymmetric conditions.”

matrix T, 4’: terminant) to the color-matching functions

x(A), y(A), Z(A). The latter are supposed ex-
X\ - xe pressed with respect to any convenient set of
Y) =Tal Y’ [3(5.12.1)] primaries without reference to visual theory. The
Z, ze fundamental tristimulus values R,G,B of a
stimulus {P, dA} are then related to X,Y, Z by
The matrix 7, ,, can be determined in princi-
the same matrix M; that is
ple by making just three asymmetric matches
between independent pairs of stimuli. In more R x
recent work, a larger group of matching pairs has G|=Ml| y [5(5.12.1)]
been used, and the best values of the matrix B Zi
elements of 7, ,, are determined by a least-squares
method. If the matching process with different If the two test areas O and O’ in an asym-
pairs of adapting conditions is transitive, the cor- metric matching process have the same color-
responding matrices satisfy the matrix equation matching functions (x(A) = x’(A), and so forth)
and the same fundamental color mechanisms, the
Tya0 = Tha Tyran [4(5.12.1)] same matrix M will be applicable to both, and
this is commonly assumed. According to the hy-
A celebrated hypothesis on chromatic adap- pothesis, the asymmetric match between {P, dA}
tation originating in a /aw of coefficients put and {P; dX} corresponds to equating the abso-
forward by von Kries (1904) would lead under lute responses of the respective fundamentals to
certain conditions to asymmetric matching which give
is linear and transitive. The von Kries hypothesis
assumes that different adaptations of a particular K ypR = KypR’
retinal area modify the overall sensitivities of
three fundamental color-response mechanisms, K4¢gG = KygG’ [6(5.12.1)]
whose relative spectral sensitivities r(A), g(A),
b(X) are a particular fixed set of the color-match-
ing functions applicable to the area in question.
Thus, 7(A), Z(A), b(A) are related by a linear where the coefficients K,r, K4g, K4g and
transformation with matrix M (of nonzero de- K yr» Kag, Karp are supposed determined by the
432 VISUAL EQUIVALENCE AND VISUAL MATCHING

surround and pre-exposed stimuli contributing to can be identified with the fundamental spectral
the adaptation of areas O and O’, respectively, sensitivities. This has been perhaps the principal
but are to be independent of the actual test application of the von Kries hypothesis: the
stimuli {P, dA} and {Px dA}. determination of the spectral sensitivities
There are obvious objections to conceding that F(A), Z(A), B(A) of the fundamental color-
adaptation is unaffected by the test stimuli, but response mechanisms. If valid, the hypothesis
accepting the hypothesis as a whole, Eq. 6(5.12.1) would reduce the problem of chromatic adapta-
implies that tion to determining the way the coefficients
Kar: Kac, Kaz depend on the surround and
WK ak X pre-exposure stimuli in various adaptation situa-
‘ Kg M\ Y tions. Several quantitative studies have shown the
. Kap Z von Kries scheme to be imperfect in one or more
ways, as indicated in Section 5.12.2. However,
K yr aa
often the adaptational changes of perceived color
— : Kuc M y’ are qualitatively of the kind it would predict for
fundamentals r(A), g(A), b(A), which, according
‘ Kap Zz to other evidence (properties of color defectives,
or two-color threshold data, objectively determined
absorptions of visual pigments), could be reason-
X XxX’
able (see Section 8.2).
YY} =| Y’ [7(5.12.1)]
Z vA
5.12.2 Experimental Procedures and Data
where In this section we present the salient points of a
number of studies on chromatic adaptation. The
T,4°— M~ 'DM reader is also encouraged to consult the very
or useful reviews by Jameson and Hurvich (1972)
and by Bartleson (1978a, b).
MT,,,M' =D
(i) Wright (1934, 1937, 1946). Wright pio-
and D is the diagonal matrix neered the use of binocular matching by the full
trichromatic method; that is, the monochromatic
K gr/Kar test stimulus in one eye is matched with a mixture
D= ; Kyc¢/Kacg of three monochromatic primary stimuli in the
other eye which is differently adapted [see Section
j Kp/K ap
5.2.5(h)]. The measurements were mainly con-
cerned with the luminance and chromaticity of a
whose three elements are often referred to as the
test stimulus applied during the recovery of the
von Kries coefficients. If the matrix T,,, is ob-
fovea from highly saturated pre-exposed condi-
tained empirically from an asymmetric matching
tioning stimuli (see also Section 5.6.5). The
process obeying the linearity laws, at least one
quantitative results include, with some tentative
matrix M can, in general, be found by standard
laws on the time course of the recovery process, a
mathematical methods which makes MT, ,,M-'a
determination of the spectral sensitivities of the
diagonal matrix D. Provided the nonzero ele-
fundamental response mechanisms by the von
ments of D are all different, the matrix M is
Knes method. These were in general agreement
uniquely determined except that the rows may be
with what might be expected if each mechanism
taken in any order. If M, ; are the elements of M
were equipped with a single visual pigment, the
(and if they are real quantities) the expressions
spectral maxima being located at wavelengths 445,
540, and 575 nm.
Mx(A) + Myy(A) + Mj3Z(A) = 7(A)
(ii) Walters (1942). Binocular matching was
Mx(A) + My (A) + M3Z(A) = g(a) employed as in 5.12.2(i), but with test stimuli
confined to medium and long wavelengths, so
M3,X(A) + M3, y(A) + M332(A) = b(A) that the experiments were reduced effectively to a
study of dichromatic vision. Matrices T,,,, M, D,
[8(5.12.1)] defined in Section 5.12.1, are then of second
 Chromatic Adaptation 433

instead of third order. The application of the von
Kries procedure to some of the sets of data led to
complex (nonreal) values for the elements of M
and D. This has proved a recurrent difficulty in
applying the von Kries scheme. Mathematically,
although M can, in general, be determined from
any experimental matrix T, ,,, there is no guaran-
tee that the solutions will be real. Walters related
his complex solutions to nonlinearity in his asym-
metric matches. Deviations from linearity in-
creased with the luminance of the test stimulus
and by extrapolating back to test stimuli of zero
luminance, he obtained real solutions correspond-
ing to fundamental spectral sensitivities (two only)
in fair agreement with Wright’s [5.12.2(i)]. This
procedure evades one of the weak points in the
von Kries hypothesis, the assumption that the test
stimulus will not modify the adaptation.
(iii) Burnham, Evans, and Newhall (1952, 1957),
Brewer (1954), and Wassef (1958, 1959). Binoc-
ular matching was used but with steady adapta-
tion of the two test areas by different surround
fields generally of rather low saturation (e.g.,
daylight, tungsten light). These studies use the
more sophisticated computational method which
assumes in effect, that “random errors” are re-
sponsible for any deviations of the observations
from some von Kries scheme. The appropriate
best solution, the matrix M, is then obtained by a
least-squares procedure applied to a large number
of matches. The results indicated that, even with
the comparatively mild adaptation differences
used, different optimal fundamental response
functions are obtained from different pairs of
adaptation conditions (failure of the transitivity
law), and complex solutions M still occur. Table
1(5.12.2) taken from Wassef’s results, illustrates
these difficulties. The columns under (1), (2), and
(3) defined the chromaticities in the CIE system
of the fundamental vectors (i.e., three hypotheti-
cal stimuli, each of which would stimulate one
color-response mechanism and one only). The
fundamental vectors, when all three are known
and are real, determine the fundamental response
functions.
The appearance of different fundamental vec-
tors in experiments using different conditioning
stimuli, suggests that the way in which receptors
containing different photopigments are associated
in the quasi-independent mechanisms whose re-
sponses are equated in asymmetric matching, may
vary with the conditioning stimulus. Physiologi-
cally, this would not be hard to accept. Threshold
studies on summation in rod vision suggest that
the lateral neural connections of receptors are
changed as the adaptation level is raised, and, if it
is granted that the effective color mechanisms in
asymmetric matching need not be associations of
receptors all containing the same photosensitive
pigment, the possibility of some change in the
neural association under different conditioning is
not unreasonable. However, if this is admitted,
the method of handling the data of asymmetric
matching must be changed. The transformation
for converting tristimulus values to a system in

Table 1(5.12.2) CIE Chromaticity Coordinates of Fundamental Vectors Derived from

Chromatic Adaptation by Wassef (1958, 1959)

(1) (2) (3)

Pairs of
Conditioning Stimuli =x y z x y z 2 y Z Reference

Tungsten color (A):

Daylight color (C) 0.700 0.375 —0.074 0.924 —0.664 0.740 0.149 0.041 0.810 Wassef (1958)

Tungsten (A)
Daylight (C)
(Black background) 0.702 0.345 —0.047 <..... ENE Sil complex
Po Te 6 aap oped tes > Wassef (1958)
values
Tungsten (A) :
Daylight (C) 0.641 0.342 COViAee SG.- 3s eee complex es... > Wassef (1959)
values

Magenta ;
Daylight (C) 0.555 0.200 0.245 0.156 0.359 0.485 0.132 0.014 0.852 Wassef (1959)

Yellow :
Daylight (C) 0.684 0,335 —0.019 <..... Ne ORCS complex
Oe Se eee ea > Wassef (1959)
values
Green
Daylight (C) complex values 2.481 —0.856» —-0.625 complex values _Wassef (1959)
434 VISUAL EQUIVALENCE AND VISUAL MATCHING
which each component stimulates just one mecha-
nism, will be different for the two adaptations A
and A’, and the matching equations [see Eq.
3(5.12.1)] must now be written
X x
M,\| Y| =M,| Y’
Z we
Or
whe /
Yilc= IM lay We [its ))
2, Za
The experimental transformation 7, ,, has merely
to be of the form
eas Faeire Bene)
There is no way of determining M, and M,, from
T,,° indefinitely many pairs of transformation
M, and M,, will satisfy. Naturally, the matrix
T, 4, contains some information on the relation-
ship of the possible pairs of transformations M,
and M,,, but the information takes no simple or
striking form. The physiological implication in
the case of a matrix T, ,,, which leads to complex
values when processed on the von Kries assump-
tion, may be that the two adaptations have
modified differently the way in which receptors
containing different pigments are organized into
the systems whose responses are equated in mak-
ing asymmetric matches. However, it then follows
that even when the matrix T,,, leads to all real
solutions on the von Kries method, this may be
regarded in a way as a fortunate accident. The
two matrix transformations M, and M,. just hap-
pen to yield a product M,M,,. for which a third
matrix M can be found, such that
MT,,M~'= M(M;,'M,)M~!
[3(5.12.2)]
is a diagonal matrix. Stated briefly, the continual
appearance of complex solutions makes the inter-
pretation in the von Kries way of even the all-real
solutions doubtful.
The above remarks apply strictly to the situa-
tion when the asymmetric matching conforms
closely with the linearity laws embodied in Eq.
3(5.12.2), which was given as
i xt
y p=Tjpley
Z Va:
Burnham, Evans, and Newhall (1957) obtained
a better representation of their asymmetric
matches by means of the nonhomogeneous rela-
tions
X = ty,
X' + tyY’ +932’ + tog
X = t3,X! + tyoV! + t33Z! + bay
[4(5.12.2)]
The von Kries method is then no longer strictly
applicable.
(iv) Hunt (1950, 1952, 1953, 1958). Binocular
matching was used with steady surrounds control-
ling the adaptation. The principal objective was
to establish the effects of conditioning stimuli
(surrounds) of widely different luminances, the
chromaticity being in most cases the same (white).
Hunt’s main results were expressed by plotting in
the CIE chromaticity diagram, the chromaticity
of the comparison stimulus, seen in a condition-
ing surround of fixed luminance (yielding a reti-
nal illuminance of 75 td), that matched a test
stimulus of fixed chromaticity, seen under a wide
range of surround luminances (0, 0.7 to 6300 td).
The approximately radial lines in Figure 1(5.12.2)
show the change of chromaticity of eight test
colors, as the surround luminance was changed.
The circle point on each line represents the chro-
maticity when the surrounds of test and compari-
son stimuli were the same; this is the symmetrical
situation. In these measurements, the test stimu-
lus was always given the same luminance as the
surround. The very large increase in excitation
purity of all the test stimuli, as the surround
luminance was increased from the lowest to the
highest value (marked Z to A on one of the lines),
is evident. The test stimulus and comparison
stimulus for the data of Figure 1(5.12.2) were of
1° linear extent, and foveal vision was used.
Hunt’s conclusion that no participation of the rod
mechanism could explain the observed large
changes in purity (saturation) can safely be
accepted.
 Chromatic Adaptation 435

0.6

500

0.5
580

0.4
O
600

D
C
S @ R
M A ‘650

0.2

480

0.1

“7
: 0.1 0.2 0.3 0.4 0.5 0.6 0.7
X

Fig. 1(5.12.2). Change in apparent color of eight stimuli (R,O,Y,G,C,B,V,M) of fixed

CIE chromaticity, as the equal luminances of test stimulus and surround are decreased together
in steps from A (6300 td) to F (0.7 td). An additional observation with zero luminance of the
surroundings and a low luminance (0.7 td) of the test stimulus is also included (point Z ). For
the stimulus R, the chromaticities for the different luminance levels are labeled in order of
decreasing luminance 4, B, C, D, E, F, and Z (Hunt, 1953).

Hunt (1950) discusses a possible explanation
of the change of saturation with luminance which
rests on essentially the same idea as the interpre-
tation just given of the complex imaginary solu-
tions obtained in applying the von Kries principle
in studies of differential chromatic adaptation.
He supposes, in effect, that in the dark-adapted
state, the receptors (cone receptors only), contain-
ing. among them all three kinds of pigments, are
neurally grouped into three mechanisms that de-
termine the brightness and chromatic appearance
of a test stimulus; that is, the quantities equated
in the asymmetric match. The three groupings are
not identical so that stimuli of different spectral
composition can produce different perceived col-
ors, but because each mechanism incorporates
some receptors of all kinds, there is extensive
overlapping of their relative spectral sensitivity
curves, and the range of possible color percep-
tions is limited. Adaptation to higher levels, by
breaking down some of the neural links, tends
progressively to limit the receptors of a particular
mechanism to those containing just one pigment,
a different one for each mechanism. The effective
spectral sensitivity curves are then much nar-
rower, and highly saturated as well as unsaturated
perceived colors can be produced by suitable test
stimuli.
Other data of Hunt show that his asymmetric
matches do not obey the linearity laws. When the
surrounds of test and comparison stimuli are kept
fixed, a match is, in general, upset by changing
the luminances of the two stimuli by the same
factor.
(v) MacAdam (1956, 1958, 1961, 1963). The
method employs a monocular bipartite field with
differential adaptation of the retina in the two
436 VISUAL EQUIVALENCE AND VISUAL MATCHING

wae Gras

G
Yt | Tungsten* ——> \,
light a
e
Daylight

0.6 0.8

Fig. 2(5.12.2). Some typical chromaticity shifts obtained by MacAdam (1956) for observer
DLM using a method of differential adaptation of the retina. Each vector shows the change in
chromaticity that was required to maintain a visual match between the two halves of a
monocularly viewed bipartite field when one part of the central area of the retina was adapted
to daylight and an adjacent part of the same retina was adapted to tungsten light.

halves [see Section 5.2.5(1)]. Extensive asymmetric
data were collected with pairs of adapting stimuli
selected from the series: tungsten light, daylight,
green, pink, red, greenish-yellow, blue; the mem-
bers of the selected pair having, in nearly every
case, about the same luminance. Figure 2(5.12.2)
gives an example of the set of data collected with
tungsten light and daylight as the adapting stimuli.
MacAdam’s analysis showed considerable sys-
tematic deviations from the predictions of any
von Kries scheme. These deviations were found to
be of such a kind that if the true linearity of
asymmetric matching was assumed, the existence
of more than three independent fundamental
spectral sensitivities would be indicated. The
concept of just three fundamentals is restored by
admitting nonlinearity and, in later work,
MacAdam (1961, 1963) put forward a nonlinear
hypothesis which replaced the von Kries match-
ing equations [Eq. 6(5.12.1)] by the following
(MacAdam’s symbols have been modified to suit
the present context):
dp + by(R)"* = az + be R’)*
ag + bg(G)"? = ag + b6(G’)”*
az + b,(B)’? =a‘ + b5(B’)””
[5(5.12.2)]
The nine quantities dp, dc, ap, br, bc, bp,
Pr: Pc» Pp are dependent on the adaptation con-
ditions A of one half-field, and the similar primed
quantities depend on the adaptation conditions A’
of the other half-field. These 18 constants replace
the six adaptation constants in the von Kries
equations. However, as the equations are unal-
tered by multiplying both sides by a constant
factor, the number of independent constants is
 Chromatic Adaptation 437

reduced by three. As for the von Kries scheme,
MacAdam’s scheme also requires the matrix M
which transforms x(A), y(A), Z(A) to fundamen-
tals r(A), (A), b(A). MacAdam’s calculations are
then directed to finding optimal fundamentals as
well as best values of all the assignable constants.
For each pair of adaptations studied, the ap-
propriate criterion throughout was the minimiza-
tion of the differences between the matches
predicted by the nonlinear hypothesis and the
matches actually observed. For the most satisfac-
tory fundamentals, the chromaticities of the
fundamental vectors were obtained as follows:
R-Funda- G-Funda- B-Funda-
mental mental mental
9 0.747 Kis 0.1785
y 0.253 =A) p35 0.0
The indices pr, Pr (OF Pg, Pg) were found to
have values, for different adaptations, in the range
0.33 to 1.0, with a rather smaller range, 0.5 to 1.0
for Pz, Pp If the transitivity law holds in the
present application, the indices pr, pcg, pg would
be expected to depend only on the adaptation in
the one half-field, pp, pG, pz only on the adap-
tation of the other half. Thus, pp and pp would be
the values assumed by a single function of the
adaptation variables of a half-field and similarly
for pg and pG, pg and pz. This appears to be
approximately true as far as the dependence of
the value of the index on the chromaticity of the
adapting stimulus is concerned. For example,
contours of constant pp values in the chromaticity
diagram of the adapting stimulus as derived by
MacAdam (1961) are shown in Figure 3(5.12.2),
and similar figures were obtained from pg, and pp.
The analysis was also applied to the independent
data of Burnham, Evans, and Newhall (1957) and
Wassef (1958, 1959) with considerable success.
It may be noted that MacAdam (1961) also
examined another nonlinear response relation, the

0 0.2 0.4 0.6 0.8

Fig. 3(5.12.2). Contours in the CIE (x, y)-chromaticity diagram of an adapting stimulus, on
which the index p(R) of R in the power-law form of the asymmetric matching conditions has
the constant value indicated. The encircled numbers represent the results from the independent
data of Burnham, Evans, and Newhall (1952) (from MacAdam, 1961).
438 VISUAL EQUIVALENCE AND VISUAL MATCHING
logarithmic expression crlog(dpR +t ep), as an
alternative to the power law expression dp +
bpR?®, for use in Eqs. 5(5.12.2) with similar
changes in the terms involving G, B, R’, G’, B’.
This also proved greatly superior to the linear
(von Kries) form and was comparable with the
power law in representing the experimental data.
Whether these nonlinear laws would explain the
effects of adapting stimuli of widely different
luminances (Hunt 1953) for which a different
kind of explanation (adaptational changes in the
matrix M) was noted above is not clear. A further
discussion of this point has been given by Stiles
(1967).
(vi) Hunt (1950, 1952). A binocular matching
method was used with large matching fields (20°
diameter), steady and interrupted surrounds. In
adaptation by means of a surround, there is a
question whether the total adaptive effect can be
resolved into a contribution ascribable to simulta-
neous contrast between test stimulus and sur-
round and a contribution of a more persistent
kind that does not depend on the actual presence
of the surround during the match. Hunt investi-
gated how the color of the test stimulus was
modified if the adapting surround was tempor-
arily extinguished while the match was being
made. It was found that when the effect of simul-
taneous contrast was eliminated, the adaptive
color shifts were certainly smaller but generally of
the same kind, that is, in the same direction in the
chromaticity diagram.
Hunt also measured the time taken by the
surround to produce the full adaptive color shifts
in a test eye originally dark adapted, and the time
for the eye to recover fully after the removal of
the surround. He obtained values of the order of
100 sec. The smaller but quite measurable effects
of simultaneous contrast on the other hand were
very rapid, taking less than a second to operate
after the abrupt removal of the surround. That a
very rapid process and a much slower process
enter into photopic adaptation is important and
has been elaborated by Schouten and Ornstein
(1939) in their work on brightness.
(vii) Helson, Judd, and Warren (1952). A same table, if necessary, by graphical interpola-
memory method was used. The observer first
learned how to describe the colors of painted
chips, presented in daylight, in terms of hue,
lightness, and chroma, conveniently making use
of Munsell color chips and their notations,
Munsell hue, Munsell value, and Munsell chroma
(Section 6.6.1). After sufficient practice, which
might take eight hours or so, the observer has
memorized hue, lightness, and chroma scales so
as to be able to describe with reasonable accuracy
and precision the color of any object he or she
may perceive. The trained observer is then re-
quested to look at the objects in chromatic light,
such as tungsten light and, after the observer’s
eyes have adapted to the new illumination, de-
scribe the color perceptions in terms of the mem-
orized color scales.
Figure 4(5.12.2) illustrates some typical results
obtained by the authors for six observers judging
Munsell samples under CIE standard sources C
(average daylight) and A (tungsten light) against a
white background. Only the resultant shifts in hue
and chroma are shown in the figure, and these are
indicated by arrows. Each arrow starts at the
point corresponding to the hue and chroma judged
to belong to the Munsell sample in daylight
(source C) and ends at the point corresponding to
the hue and chroma judged to belong to the same
sample in tungsten light. The judgment in daylight
was made after complete adaptation to daylight;
similarly, the judgment in tungsten light was made
after complete adaptation to tungsten light. The
authors noted that the luminance factor (CIE
tristimulus value Y) of the background affects the
results greatly, as found earlier by Helson (1938)
who showed that the perceived hue and chroma
of an object-color stimulus depend on the relation
of its luminance to adaptation luminance. Adap-
tation luminance depends on a weighted mean of
the luminance factors of all object-color stimuli in
the field of view and is largely determined by the
background.
The analysis that Helson et al. (1952) apply to
their data is consistent with the von Kries hy-
pothesis and the transformation equations
governing the hypothesis as outlined in Section
5.12.2. The predictions are in fair agreement with
the experimental data.
The Munsell notations for the samples viewed
by the observers under daylight can be converted
readily to CIE (X,Y, Z)-tristimulus values by
means of Table 1(6.6.1). The Munsell notations
estimated by the observers for the samples under
tungsten light can be converted similarly from the
tion in Figures 1(6.6.1) to IX(6.6.1), to
(X’, Y’, Z’)-tristimulus values. The test validity of
the von Kries hypothesis then consists of finding
the matrix T,,,of Eq. 7(5.12.1), such that
4¢ ,
a anyhe ae
[6(5.12.2)]
vs Dis
 Chromatic Adaptation 439

Fig. 4(5.12.2). Changes in hue and chroma in passing from CIE source C (daylight) to CIE
source A (tungsten light), according to data reported by Helson, et al. (1952) using the memory
method. Observations shown here refer to Munsell samples of medium Munsell value in
daylight, displayed on a white background.

The first step is to determine a suitable matrix M definition,

which transfers the (X, Y, Z)-tristimulus values
to the fundamental tristimulus values R, G, B by 6 ey eh ear, Gayals
means of the equation
and that, in addition, both the (X, Y, Z) and the
R x (R, G, B)-system are normalized such that the tri-
G|=M| yY’ [7(5.12.2)]
stimulus values of the equal-energy stimulus are
B ze
Helson et al. (1952) assumed the fundamental
primaries to have the following CIE (x, y,z)-
chromaticity coordinates
and

Fundamental Primaries
haw ey aay
R G B
It follows that
x 0.7471 1 0.1803
y 0.2529 0 0
z 0 0 0.8197 HEYA PG;
Mote ey wey Ys
which satisfy the familiar constraint of x + y +z ZR ZG Zp

= 1 (see Sections 3.2 and 3.3). Corresponding

tristimulus values (Xp, Yr, Zz), and so on, are 2.9541 —2.1741 0.2200
obtained from the above chromaticity coordinates Sale. MY 0 0
(Xr» Yr» Zr) and the like by recalling that, by 0 0 I
440 VISUAL EQUIVALENCE AND VISUAL MATCHING

and resulting in

0 l 0 1.1511 —0.4464 0.4786

M= | —0.4510 1.3588 0.1012 (Bip pea 1 0
0 0 1 0 0 3.3267
[8(5.12.2)] In an experiment in which different pairs of
adapting stimuli (A, A’) are studied and the same
The second step is to calculate by means of Eq. matrix M~!' is used in the analysis, a useful
1(5.12.2) the (R, G, B)-tristimulus values of the expression for the matrix 7, ,, is the following:
stimuli that control the state of adaptation of the
observer in the daylight and tungsten-light experi- B 2.9541(a—B) 0.2200(y — B)
ments, respectively. These stimuli are given as Ty 4 a 0 a 0

CIE standard source C (identified as condition A) 0 0 Y

and as CIE standard source A (identified as con-
dition A’). Possible changes in adaptation caused
[11(5.12.2)]
by the colored samples themselves are being
ignored. The tristimulus values of source C and where the von Kries coefficients have been de-
source A are known to be noted by a, B, y, that is

X- = 98.0, No 100. cae LER Qa

= Kar
B ae
Kac
a ¥,
K 4

Kar K yg K yp

SOROS Meee eyed ON pee Oeil S

(viii) Land (1959a,b, 1962, 1964). An absolute
which transform into color naming method is used. Land’s experiments
on chromatic adaptation are distinguished from
Révct 100 Ke eeGroeiyl 036s thar LLB most earlier work by the use of a complex stimu-
lus pattern containing a wide range of different
Rien eee 00 nue chromatic and achromatic stimuli with sharp or
graded boundaries and often representing actual
In the notation used in Section 5.12.1, the meaningful scenes. The observer’s eyes are al-
above (R, G, B)-tristimulus values of the adapt- lowed to range over the pattern, the perceived
ing stimuli are given as colors of the various parts being reported on in
the color nomenclature of everyday life. Thus,
there is the possibility that information sent by
Kyug =Ge, Kyp
= Be the eyes to the brain about the nature of the scene
being viewed may feed back so as to modify the
Kyg=G4, Kyp=B,
visual, including the color, perceptions by whic
it is conveyed. :
and thus, the diagonal matrix D becomes
The original striking demonstrations (Land,
1959a, b) were pictures, mainly elaborate still-lifes,
Rc/R 4 produced by projecting, in exact register, two
dS | Gc a Gy,
black and white positives obtained originally by
using, in turn, red and green filters over the
Bo/By
camera lens. The two projectors emit lights of
1 ; different color (white and red, for example), and
ep elle) [9(5.12.2)] Land established, among other things, that pic-
3.3267 | tures of complex scenes could be produced in this
way, showing a remarkable variety and richness
of color, comparable in some cases with three-
In the final step, the matrix T, ,, is obtained
color projections. This could be done even if
from Eq. 7(5.12.1), that is,
monochromatic projector sources were used, pro-
vided their wavelengths were not too close; in the
T44,=M"'DM [10(5.12.2)] yellow region, the separation need be no more

than about 20 nm. Land rejects explanations of
the colors seen in terms of the accepted concepts
of trichromatic theory and chromatic adaptation
(but he appears to have in mind only the slow
component) and elaborates a set of principles of
color perception in complex scenes coordinating
his observations. For these principles, difficult to
summarize, reference must be made to Land’s
own explanations.
Some of Land’s phenomena are covered in
earlier work (Judd, 1940; Evans, 1943). Judd
(1960) gave more traditional explanations of many
of Land’s observations, using his own fairly
elaborate formulae for perceived color, developed
in 1940. The main element in these explanations
is contained in a simple discussion by Woolfson
(1959) who used approximate methods to
determine, in effect, the possible range (three-
dimensional) of the tristimulus vectors corre-
sponding to small areas in the Land projection of
a complex picture produced by a particular pair
of monochromatic projector sources. The tri-
stimulus values were determined in a fundamental
system close to those to which most color re-
search leads. The particular choice of the system
is not very critical. He then applied a simple von
Kries-type transformation by dividing the respec-
tive components of the tristimulus vectors by
three constants. The transformed vectors, treated
as stimuli viewed in isolation or in a standard
white surround, correspond to perceived colors
which Woolfson accepts as sufficiently well de-
fined by the positions of their chromaticities in
the standard chromaticity chart. The perceived
colors in the actual Land projection are supposed
to coincide with those derived in this way.
Woolfson’s basic assumption corresponds very
nearly to a conception going back to Helmholtz;
that is, in judging the color of objects in a real
scene, we make an allowance for the color of the
illuminant. An interpretation in terms of meaning
would postulate that the viewer knows certain
objects in the picture ought to appear white under
“normal” conditions of illumination, and the eye
and brain effect an inevitable and unconscious
color transformation that makes this true or nearly
true. Against this is the fact that when the picture
consists of a meaningless assembly of patches in a
sufficiently random arrangement, the range of
colors perceived is at least comparable with that
in a meaningful picture, the parts of which pro-
vide a set of stimulus vectors (untransformed)
similar in range to that of the patches. It may be
proposed that involuntary color transformation is
Chromatic Adaptation 44]
provided not by the internal content of the pic-
ture, but by side clues—for example, light
scattered from the projector beams, visible to the
observer and informing the observer of the illumi-
nant color that must be allowed for.
The other extreme of interpretation would
make no appeal to “meaning” and would attri-
bute the wide range of colors perceived in a
random arrangement of patches or a complex
(short-period chromatic adaptation) generated by
the inevitable juxtaposition of areas with markedly
different tristimulus vectors. It is certainly true, as
shown by Land and others, that in an ordered
arrangement of patches in which the tristimulus
vectors (untransformed) vary in a simple progres-
sive way by small steps on any line across the
picture, and especially in a picture where there is
a similar continuous variation on any line, the
range of colors perceived is greatly curtailed.
Whereas simultaneous contrast would be the
principal element in the explanation of the wide
differences of perceived color, the long-period
component of chromatic adaptation would con-
tribute to determining the average perceived color
of the picture. It is, however, difficult to maintain
the extreme interpretation that “meaning” is not
a factor. In some experiments (see, for example,
Judd, 1940), observers report that they can change
the color they perceive by consciously altering
their conception of what it is they are viewing.
Land has introduced the term retinex to char-
acterize the proposed retinal—cerebral system be-
lieved to underlie the phenomena that he has
described (Land 1964; 1974; Land and McCann,
1971; McCann, McKee, and Taylor, 1976). Fur-
ther developments along the lines of Land’s ap-
proach have been reported, notably by McCann
(1972) and McKee, McCann, and Benton (1977).
(ix) Hurvich and Jameson (1958, 1960), and
Jameson and Hurvich (1956, 1959, 1972). From
the extensive studies of Hurvich and Jameson, we
have selected for this section the experimental
work reported by them in 1958. They employed
binocular matching to determine chromaticity
shifts for a single 2° test stimulus at two lumi-
nance levels (32 cd-m ? and 3.2 cd-m ’,
approximately) for three pairs of adaptation con-
ditions (A, A’). Figure 5(5.12.2) illustrates the
results which show that the proportionality law,
as defined in Table 1(5.12.1), fails under asym-
metric adaptation conditions, making the von
Kries hypothesis inadequate to predict such ex-
perimental data.
442 VISUAL EQUIVALENCE AND VISUAL MATCHING
Sr jaqtes ies a eG
0.8
0.6 |i-
5004
O4-r
0.2
Fig. 5(5.12.2). Chromaticity shifts for a single 2°
test stimulus at two luminance levels (32 and 3.2
cd - m 7) for three pairs of adaptation conditions.
The open circle in the CIE 1931 (x, y)-chromatic-
ity diagram at x = 0.23 and y = 0.20 represents
the chromaticity of the variable 2°- test stimulus in
the left eye for both the high and low luminance
level of the test stimulus of fixed relative spectral
radiant power distribution in the right eye when no
surround stimulus is present in either eye. When
the left eye is adapted to a near-achromatic stimu-
lus (32 cd - m ”) and the right eye to a 450-nm
monochromatic stimulus of the same luminance,
the binocular match for the high luminance test
stimulus is the chromaticity indicated by the arrow
tip at H, on the diagram, and for the same adapt-
ing and surround conditions, the match to the low
luminance stimulus is at the arrow tip at L ,. When
the right eye is adapted to a 650-nm monochro-
matic stimulus, and all other conditions are kept
the same, the comparable matches for the two
luminance levels are at H, and L., respectively.
For adaptation of the right eye to 550 nm, the
matches plot at H, and L.,, respectively. (Hurvich
and Jameson, 1958 and Jameson and Hurvich,
1972).
Jameson and Hurvich (1959) suggested that
chromatic adaptation may be explained more
satisfactorily by postulating two processes. In ad-
dition to the “multiplicative process,” governed
by the von Kries coefficients and located at the
cone receptor level, there exists an “incremental
process” located in post-receptoral neural units or
systems in which the opponent, presumably
color-coded, neural responses are detected (see
also Jameson and Hurvich, 1972; and Sections
5.13 and 8.3.3).
(x) Sobagaki, Yamanaka, Takahama, and
Nayatani (1974). A scaling method, known as
magnitude estimation, was used along the lines
studied in an earlier paper by Nayatani,
Yamanaka, and Sobagaki (1972). The method
aims at assessing the strength of an observer’s
color perceptions in terms of numbers selected by
the observer in proportion to the magnitudes of
those perceptions (Section 6.2). Hue, saturation,
and lightness were scaled in that way for 95
colored chips selected from the Munsell Book of
Color (Section 6.6.1), which the observer viewed
one at a time and in random order under a
daylight source (a fluorescent lamp simulating
CIE standard illuminant D,;) and under a tung-
sten source (simulating CIE standard illuminant
A). Thirteen observers took part in the investiga-
tion.
Hues were estimated in terms of the “unique
hues”: red (R), yellow (Y), green (G), and blue
(B). Between neighboring “unique hues,” the ob-
server had 100 steps available to assign an ap-
propriate scale value to the hue of a given colored
chip. For example, to an orange chip, the ob-
server might assign the value (80R/20Y) which
corresponded to the judgment that the chip ap-
peared to have 80% redness and 20% yellowness.
A chip of achromatic appearance was assigned
the value zero.
Saturation was estimated on a scale from 0 to
100, zero being an achromatic perception, 100
being maximum saturation. For studies of the
effect of using different methods of estimating
saturation, see Sobagaki, Takahama, and Nayatani
(1976) and Takahama, Sobagaki, and Nayatani
(1978). .
Lightness was similarly estimated on a scale
from 0 to 100, zero being “no lightness” and 100
being maximum lightness.
For details of the numerical procedure used to
convert the observed scale values to (X, Y, Z)
and (X’, Y’, Z’)-tristimulus values, the original
papers of Sobagaki et al. (1974) and Nayatani
et al. (1972) should be consulted, as well as the
paper by Mori, Kambe, and Hanayama (1972).
The numerical procedure was essentially a least-
squares technique yielding, for each of the 13
observers, a transformation matrix 7, ,, such that
the expression
oY X
Vd ees PV Ue
ZL vA

was Satisfied as closely possible. The fundamental
vectors of each matrix provide the fundamental
primaries that govern the von Kries-type transfor-
mation. These are obtained by standard numeri-
cal techniques by solving the characteristic equa-
tion of T, 4. For five of the 13 observers, complex
solutions of the characteristic equations were ob-
tained. This, as discussed above in Section
5.12.2(iii), makes the interpretation of the experi-
mental data of all 13 observers in the von Kries
way a doubtful exercise.
In subsequent studies by Takahama, Sobagaki,
and Nayatani (1977) and Nayatani, Takahama,
and Sobagaki (1981), these authors have devel-
oped a nonlinear model of chromatic adaptation
designed to predict more adequately various ex-
perimental data. The proposed model is a two-
stage model. Its first stage operates in accordance
with a modified von Kries-type transformation,
and its second stage in accordance with a nonlin-
ear transformation (power function) that de-
scribes a response compression assumed to take
place in each receptor mechanism (Boynton and
Whitten, 1970). The model awaits experimental
testing.
(xi) Pointer, Ensell, and Bullock (1977) and
Pointer (1980). In the first study, direct scaling
by a magnitude estimation method was used to
assess the hue and saturation of each of 60
object-color stimuli of constant luminance of 60
cd-m *. Each object-color stimulus was gener-
ated by a lamp-filter combination and displayed
in a 2° field surrounded by an adapting stimulus
of 18° angular subtense, representing either CIE
standard illuminant A (2856 K) or CIE standard
illuminant D,; (6504 K). The luminance of the
adapting stimulus was fixed at 120 cd - m’, that
is, twice the level of the test stimuli, making the
latter to appear in the object (or surface) mode.
Five observers took part in the experiment using
monocular viewing.
Hue scaling was done in a similar way to that
described in Section 5.12.2(x), but saturation scal-
ing was approached differently. A gray scale was
provided in the field of view to help the observer
with estimating saturation. The observer was first
required to find the gray chip that most closely
matched the test stimulus in lightness, and then to
estimate the magnitude of the difference between
that gray chip and the chromatic test stimulus.
Thus, the saturation scale became open-ended
and not limited to a fixed upper value, such as
100. This arrangement was considered important
to allow for the fact that maximum saturations
Chromatic Adaptation 443
that can be experienced for stimuli of different
hues are not equal in magnitude. For example,
the most saturated yellow is generally seen as less
saturated than the most saturated blue.
The test stimuli were calibrated in terms of the
(u’, v’)-chromaticity coordinates that define the
CIE 1976 UCS diagram (Section 3.3.9). To each
chromaticity point was assigned the mean ob-
served hue scale value which then permitted the
drawing of lines of constant hue by simple free-
hand procedure. The observed saturation-scale
values obtained for each test stimulus by the
different observers had to be normalized first
before a mean value could be calculated, because
the scale was open-ended and each observer used
a different scale. Lines of constant saturation
were then drawn freehand through the chromatic-
ity points. The above analysis of the data yielded
two diagrams, Figure 6(5.12.2), of grids of con-
stant hue and constant saturation; one diagram
for daylight (D,;) adaptation, the other for
tungsten-light (A) adaptation.
The two grids shown in Figure 6(5.12.2) pro-
vide a graphical method of estimating the hue and
saturation of any object color stimulus that we
may wish to view under conditions similar to
those of the above experiment. Suppose a colored
chip is given whose spectral reflectance character-
istics are known. Its (u’, v’)-chromaticity coordi-
nates can then be calculated both with respect to
CIE standard illuminant D,; and standard illumi-
nant A. The resulting chromaticity points are
plotted in the “daylight (D,;) adaptation dia-
gram” and “tungsten-light (A) adaptation dia-
gram,” respectively. The corresponding hue- and
saturation-scale values can be read off for each
point by suitable interpolation between neighbor-
ing grid lines. The result are hue and saturation
values for the colored chip illuminated by D;,
and corresponding hue and saturation values for
the same chip illuminated by A; the observer
being adapted to D,, in the first instance and to
A in the second. The differences between the
corresponding hue and saturation values provide
a measure of the change in appearance of the
colored chip by going from the daylight to the
tungsten-light condition.
In the second study (Pointer, 1980), a similar,
but somewhat more extensive and in various ways
improved, experiment was conducted that gave
essentially comparable results to those discussed
above. An interesting feature of the second study
was the introduction of the concept of colorful-
ness, aS Opposed to chroma or saturation, that
had been proposed by Hunt (1977) as a new
Ade VISUAL EQUIVALENCE AND VISUAL MATCHING

Daylight (Dés) Adaptation

0,000 O0JOO 0.200 i“ 0.300 0.400 0.500

0.300
0,000 0.100 0.200 U i)
0.300 0400 0500

Fig. 6(5.12.2). Portion of CIE 1976 (u’,v’)-chromaticity diagram showing lines of constant
hue and constant saturation for adaptation to daylight (D,,) and for adaptation to tungsten light
(A). The lines are mean data obtained by Pointer, Ensell, and Bullock (1977) for three
observers taking part in a direct-scaling experiment of object-color stimuli by a magnitude
estimation technique. The 2° stimuli had a constant luminance of 6 cd-m 7 and were viewed in
an 18° surround (D,; or A) of constant luminance of 12 cd-m 2 “*.

appearance attribute. Observers are said to grasp
the new concept quickly and are able to scale it
consistently with a minimum of training. Hunt’s
and Pointer’s observers found scaling of satura-
tion more difficult. In this connection, the paper
by Takahama, Sobagaki, and Nayatani (1978) is
also of interest.
(xii) Bartleson (1977, 1978a,b, 1979a, b).
Bartleson’s extensive investigation on chromatic
adaptation provides, apart from a detailed review
of earlier work, new experimental data, an inter-
comparison of data from different investigators,
and a proposal of predicting the effect of chro-
matic adaptation.
_ Bartleson’s experimental work involved meth-
ods of direct scaling (for the main part, magni-
tude estimation) applied to a set of 24 test stimuli
of different chromaticities seen under different
states of adaptation. Each test stimulus (1° diam-
eter) was viewed centrally in a surround (15.2°
diameter) of high (near nonselective) reflectance
illuminated by one of three sources that provided
(x, y)-chromaticities of the surround closely rep-
resenting those of CIE standard illuminant A
(2856 K), CIE standard illuminant D,,; (6504 K),
and of an illuminant D,, with an intermediate
correlated color temperature 5003 K. The test
stimulus was presented for periods of 2 sec at
intervals of 10 sec. This was accomplished by a
shutter made of the same material as the surround
and placed right behind the opening in the sur-
round. When the shutter was in its closed posi-
tion, the entire visual field appeared virtually
uniform.
Each of the seven observers, of ages 20 to 53
who took part in the experiment, was adapted for
20 min before the start of an experimental session
to the uniform visual field with a luminance not

less than 200 cd - m~*. This ensured sufficiently
complete adaptation to the conditioning stimulus
provided by the surround, making the surround
appear perfectly “white” (i.e., a hueless area with
no trace of gray), regardless of which source was
used to illuminate it. The luminance of 200 cd -
m * was found to be the minimum level that
could be used to obtain the appearance of “white”
for the surround.
The experimental conditions were varied not
only with regard to the correlated color temper-
ature of the sources used to illuminate the sur-
round, but also with regard to the luminance of
the surround and the luminance of the test stimuli.
At this point, it is useful to recall two lumi-
nance-related terms, /uminance factor and illumi-
nance, which figure in the experiment. Luminance
factor is defined as the ratio of the luminance of
the test stimulus to the luminance of the sur-
round. It is a term commonly used for object-color
stimuli as found in everyday life (Section 3.3.8)
and as simulated in Bartleson’s experiment. [//u-
minance is defined as the luminous flux per unit
area incident on a surface, measured in terms of
Im - m ? (Section 4.3.5). In the present case, this
term applies to the field of view as a whole
(surround and test area). By varying both the
luminance of the surround and the luminance of
the test stimulus by the same factor, the illumi-
nance of the field of view changes by the same
factor but the /uminance factor of the test stimulus
remains constant. With real objects, such as col-
ored chips from the Munsell Book of Color
presented on a white background, varying the
illuminance means varying the incident radiant
flux emitted by the light source without altering
its relative spectral distribution.
Bartleson’s results lead him to draw the follow-
ing general conclusions:
(a) Perceived hue of a given test stimulus, ex-
pressed as proportions of neighboring “unique”
(or “unitary’’) hues (red, yellow, green, and blue)
varies with the chromaticity (correlated color tem-
perature) of the adapting stimulus (surround), but
does not vary significantly with the luminance
factor of the test stimulus or the illuminance of
the field of view.
(b) Colorfulness of a given test stimulus, that is,
the attribute of a visual sensation according to
which the test stimulus appears more or less chro-
matic, varies with the chromaticity (correlated
color temperature) of the adapting stimulus (sur-
round), the luminance factor of the test stimulus,
Chromatic Adaptation 445
as well as the illuminance of the field of view. The
attributes of chroma and saturation, which are
related to colorfulness, do not vary significantly
with illuminance. Chroma is judged in relation to
the brightness of the surround, whereas saturation
is judged in relation to the brightness of the test
stimulus itself. Chroma and saturation are de-
scribed as two types of “relative” colorfulness,
whereas colorfulness is considered by some inves-
tigators an “absolute” attribute of color percep-
tion (Hunt, 1977).
(c) Lightness of a given test stimulus, that is, the
attribute of a visual sensation according to which
the test stimulus appears brighter or darker in
proportion to the brightness of the “white” sur-
round, varies with the luminance factor of the test
stimulus, but is essentially invariant to changes in
the chromaticity (correlated color temperature) of
the adapting stimulus (surround) and the illumi-
nance of the field of view.
(d) Color appearance data obtained for daylight
adaptation (D,;) are highly correlated with the
Munsell renotation specifications (Section 6.6.1).
An extract of Bartleson’s main experimental
results is summarized in Figures 7(5.12.2) to
10(5.12.2). Figure 7(5.12.2) shows (u’, v’)-chro-
maticity points of the 24 test stimuli used in the
experiment. Figure 8(5.12.2) shows the shifts in
color appearances of the 24 test stimuli when
adaptation is changed from daylight (D,;) to
tungsten light (A) for a surround luminance of
1000 cd - m-? and a luminance of the test stimuli
fixed at 200 cd-m 7, making the luminance
factor B of the test stimuli equal to 0.2. The data
given in Figure 8(5.12.2) allow the construction of
a polar-coordinate system of color appearance for
adaptation to A in terms of color appearances for
adaptation to D,;. Figure 9(5.12.2) shows the
A-adaptation diagram superimposed on the D,;-
adaptation diagram. For the benefit of colorimet-
ric applications, it is useful to present the data in
a CIE standard chromaticity diagram. This has
been done in Figure 10(5.12.2). The numerical
values for the grid points are given in Table
2(5.12.2). One application of Figure 10(5.12.2) is
to determine colorimetric specifications for corre-
sponding color stimuli as obtained, for example,
by an asymmetric color match involving daylight
and tungsten-light adaptations. First, the chro-
maticity point for one adaptation condition, D,;,
say, 1s located in the appropriate diagram in
which its color appearance specification is de-
termined by interpolation in the grid-overlay.
 0.300

0.000 0.100 0.200 0.300 0.400 0.500

u'

Fig. 7(5.12.2). Chromaticity points of 24 test stimuli plotted in the CIE 1976 (u’,v’,)-chro-
maticity diagram. These stimuli were used in Bartleson’s (1977, 1979a,b) study on chromatic
adaptation.

Fig. 8(5.12.2). Color-appearance diagram in which lines of constant hue are radii originating
at the center and lines of constant colorfulness (chroma, saturation) are concentric circles; the
greater the radius of the circle, the more colorful (the higher the chroma or saturation) of the
test stimulus. Neighboring “unique” (or “unitary”) hues (R,Y,G,B) are at right angles.
Shown in this diagram are the observed shifts in color appearances of the 24 test stimuli used
by Bartleson (1977, 1979a,b) when the observers’ state of adaptation was changed from
daylight (D,,)-adaptation to tungsten-light (A) adaptation. Direct scaling by the method of
magnitude estimation was employed. The “white” surround that controlled the adaptation had a
luminance of 1000 cd-m~?; the test stimuli were kept at a luminance of 200 cd-m7, giving
them a luminance factor B = 0.2. Arrow heads are for adaptation to A, tails for D,<.

446

e
_=
 Chromatic Adaptation 447

Dayeaty (Des)
ior Adaptation

0.500
y'

0400 |5¢

0.300 ‘ <BR
0.000 Ol1I00 O200 0300 O400 0500 O600
u'

= i apo es go ee ees
0.600 Soar Tungsten-light
(A)
tis SUR Adaptation ;
\ alee a5 -

o500- \ eae
oe L 100 7”
o400- +h; et ee a Nil
\ = iM near , bra

F Ca ee Be ]
0.3001. H act L |
O:000 O100 0200 “0300 0400 0500 0600
u!
—e—e@— Des ~Adaptation
-A-—-A4— A -Adaptation
Fig. 10(5.12.1). Portion of CIE 1976 (u’,v’)-
Fig. 9(5.12.2). Superimposed color-appearance chromaticity diagram showing lines of constant
diagrams for D,,-adaptation (solid dots) and for hue and constant colorfulness (chroma, saturation)
A-adaptation (open triangles), derived from data for adaptation to daylight (D,;) and for adaptation
given in Figure 8(5.12.2). A triangle in the A-adap- to tungsten light (A). The lines are mean and
tation diagram indicates the color appearance of a smoothed data obtained by Bartleson (1977, 1979a,
test stimulus viewed under D,.-adaptation that, b) for seven observers taking part in a direct-scal-
under A-adaptation, would have the same color ing experiment with 24 test stimuli specified in
appearance as the corresponding solid dot in the Figure 7(5.12.2). The 1° test stimuli had a lumi-
D,;-adaptation diagram (Bartleson 1977, 1979a, b). nance of 200 cd-m 7” and were presented in a
15.2° “white” surround with a luminance of 1000
cd-m 7”, The numerical values of the grid points
Then, the just determined color appearance are given in Table 1(5.12.2).
specification is transferred to the grid for the
other (A) adaptation condition given in the other
diagram. This provides the chromaticity point of (see Section 8.2.5). The same matrix converts the
the desired corresponding color stimulus. (X, Y, Z)-CIE tristimulus values to (R, G, B)-
In an effort to predict his experimental data fundamental tristimulus values for any given
by means of a suitable numerical expression, stimulus of known spectral radiant power distri-
Bartleson (1979a,b) arrived at an empirical bution, that is,
scheme that is based on the notion of the
von Kries hypothesis to which is added a nonlin- R Xe
ear compression of the response of the “blue” Gl=MILY [13(5.12.2)]
fundamental. In some ways, Bartleson’s scheme is B X
a simplified version of that proposed by Takahama
et al. (1977) and Nayatani et al. (1981) mentioned Stimuli of spectral radiant power distributions
in Section 5.12.2(x) above. In particular, Bartle- {P, dA} and {Px dA} that elicit identical color
son chooses K6nig-type fundamentals 7(A), perceptions, equate the absolute responses of the
2(A), b(A) which are linearly related to the CIE respective fundamentals as follows:
1931 standard color-matching functions
R’=aR
xX(A), y(A), Z(A) by the matrix

0.0713 0.9625 —0.0147

Foasyie [14(5.12.2)]
Mi=}\ =0,3952* 11668 0.0815
0 0 0.5610 B’ = k(yB)’
[12(5.12.2)] where a, B, y are von Kries coefficients which are
448 VISUAL EQUIVALENCE AND VISUAL MATCHING

Table 2(5.12.2) CIE 1976 (u’,v’)-Chromaticity Coordinates of Corresponding Color Stimuli for
Daylight (D,;) and Tungsten-Light (A) Adaptation Conditions. The 1° Test Stimuli Are at Constant
Luminance of 200 cd-m~7, Presented in a 15.2° “White” Surround at a Luminance of 1000 ed-m?
(Bartleson, 1977, 1979a,b). The Data Are Illustrated in Figure 10(5.12.2).°

G H=R YR ne YG G BG B BR Gray

Daylight (D,;) Adaptation

20 u’= 0.218 0.215 0.208 0.199 0.185 0.182 0.189 0.207 0.198
v’ = 0.474 0.492 0.495 0.493 0.480 0.465 0.453 0.457 0.468
40 u’ = 0.267 0.249 0.226 0.196 0.151 0.144 0.171 0.226
v’ = 0.486 0.534 0.543 0.543 0.498 0.454 0.412 0.427
60 u’ = 0.321 273 0.233 0.191 0.120 0.111 0.157 0.243
v’ = 0.497 0.555 0.566 0.570 0.512 0.442 0.371 0.401
80 u’ = 0.405 — oa -- 0.090 0.082 0.140 0.270
v’ = 0.507 — — ~- 0.522 0.430 0.295 0.355
100 u’ = 0.484 — — — 0.045 0.045 — 0.286
v’=0.511 — — — 0.534 0.408 — 0.326

Tungsten-Light (A) Adaptation

20 u’=0.278 0.281 0.269 0.258 0.243 0.235 (2232 0.267 0.256

v’ = 0.528 0.541 0.543 0.539 0.531 0.523 0.500 0.516 0.524
40 u’ = 0.316 0.315 0.281 0.258 0.227 0.211 0.205 0.286
v’= 0.529 0.554 0.560 0.555 0.538 0.520 0.468 0.491
60 u’ = 0.366 — — — 0.195 0.166 0.179 0.303
v’ = 0.530 — — — 0.548 0.508 0.424 0.452
80 u’ = 0.413 — — — 0.150 0.117 — —
vo’ = 0.528 — — — 0.556 0.486 — —

“H = hue; C = colorfulness (chroma, saturation).

completely determinable from the fundamental the stimulus providing adaptation condition A
tristimulus values of the two conditioning stimuli [= daylight (D,;) adaptation, in Bartleson’s case].
that control the two states of adaptation A and A’ From the calculated fundamental tristimulus
[A = daylight (D,;), A’ = tungsten-light (A), in values (R’,G’, B’) , the CIE tristimulus values
Bartleson’s case]. A detailed exercise of a de- (X’, Y’, Z’) are obtained from
termination of the von Kries coefficients was
given in Section 5.12.2(vii). The exponent p in Eq. x’ ‘
14(5.12.2) that compresses the response of the ¥- |ay ere |[17(5.12.2)]
“blue” fundamental, interpretable as a_post- Zz B’
receptor neural activity, is obtained from the em-
pirical expression with

py = 0.32602745 + 0.3258739! + 0.340779 ht |

2SM0 a 2.07G3 O.|
MeV B20 0.1538 0
[15(5.12.2)] 0 0 1.7825

The coefficient k is given by Figure 11(5.12.2) illustrates the agreement

Bartleson obtains between his observed and pre-
ERG
yB,
[16(5.12.2)] dicted data. The diagram shows observed (solid)
lines and predicted (dashed) lines of constant hue
and colorfulness (chroma, saturation) for the mean
where B, is the fundamental tristimulus value of observer whose state of adaptation was changed
 Chromatic Adaptation 449

0.600 Tungsten-light (A) Adaptation

Observed
Predicted

0400

0.300
0.000 0100 0.200 0.300 0.400 0500
u'

Fig. 11(5.12.2). Observed and predicted lines of constant hue and constant colorfulness
(chroma, saturation) for the mean of the observers whose state of adaptation was changed from
daylight (D,,) to tungsten light (A) in Bartleson’s (1977, 1979a, b) experiment, summarized in
Figure 10(5.12.2). The predictions were made by means of a simple nonlinear model whose
coefficients were determined empirically from the bulk of the experimental data (Bartleson,
1979).

from daylight (D,;) to tungsten light (A). The
agreement is quite reasonable.
Bartleson (1979a,b) applied his formulae also
to experimental data gathered by other investiga-
tors, notably those of Burnham et al. (1957) and
Wassef (1959) [see Section 5.12.2(111)], MacAdam
(1963) [see Section 5.12.2(v)], and Helson et al.
(1952) [see Section 5.12.2(vii)]. The agreement
between his predictions and the observed data
varies from one data set to the other, from fair to
quite reasonable and is, in general, comparable
with the agreement the other investigators had
obtained by applying their own formulae to their
own data.
5.12.3. A Comparison of Chromatic
Adaptation Transforms
Bartleson (1978a, b) made a comparison of differ-
ent methods of predicting observed chromatic
adaptation data. The following eight methods were
selected from the literature, some of which have
already been discussed in Sections 5.12.1 and
wi.
(a) von Kries (linear) transformation as dis-
cussed in Sections 5.12.1 and 5.12.2(vii). The work
by Helson et al. (1952) provides a typical example
for a von Kries transformation.
(b) and (c) Two inhomogeneous, linear transfor-
mations used by Burnham et al. (1957) and by
Sobagaki et al. (1974), as discussed in Section
5.12.2(iii), which, because of their additional con-
stants, are not of the von Kries type.
(d) Nonlinear transformation, given in terms of
power functions, used by MacAdam (1956), as
discussed in Section 5.12.2(v).
(e) Nonlinear transformation, based on a con-
cept of level dependent neural interaction pro-
posed by Hunt (1950), and formulated by
Takahama et al. (1977) [see also Sections 5.12.2(vi)
and 5.12.2(x)].
(f) Two-stage nonlinear transformation based on
proposals by Jameson and Hurvich (1972), as
formulated by K. Richter (1976).
(g) and (h) Two graphical transformations,
which may be considered equivalent to (unspeci-
fied) nonlinear transformations, as used by Pointer
et al. (1977) [see Section 5.12.2(xi)] and by Bartle-
son (1977) [see Section 5.12.2(xii)].
The comparison of the different methods of
prediction was made on a common set of 120
paint samples selected from the Munsell Book of
Color. For each of these samples, the CIE 1976
(u’, v’)-chromaticity coordinates were determined
with respect to CIE standard illuminant D,;,
450 VISUAL EQUIVALENCE AND VISUAL MATCHING

0.6
HELSON - JUDD-WARREN (1952)

(a)

Fig. 1(5.12.3). Portion of CIE 1976 (u’, v’)-chro-

maticity diagram showing lines of constant Munsell
hue and chroma for Munsell value 5, with regard
B
to daylight (D,,;)-adaptation (Bartleson, 1978a, b). BURNHAM - EVANS -NEWHALL (1957)
0.4
0.2 1 0.3 0.4

which was taken as the reference state of adapta-

tion of the standard observer. The fact that the
Munsell notations for the samples are only ap-
plicable for CIE standard illuminant C, and not
D,5, presented only a minor difficulty, which was Pie

overcome by using a von Kries transformation to =

shift C into D,; with corresponding adaptation cm

shifts for the chromaticity points of the 120 sam-
ples. The correction is small because C (6770 K)
and D,; (6504 K) have similar chromaticities.
Figure 1(5.12.3) illustrates the chromaticity points
of 30 of the total set of 120 Munsell samples, all
of which are of Munsell value 5; that is, a con-
stant luminance factor B ~ 0.2. Their Munsell
chromas are 2, 6, or 10, distributed over 10 differ-
ent Munsell hues.
Each of the eight methods of transformation
referred to above under (a) to (h) were then used
to predict the (u’, v’)-chromaticity points for the
same samples when these were illuminated by
CIE standard illuminant A, with the standard
observer completely adapted to illuminant A. Fig-
ures 2(5.12.3), 3(5.12.3), and 4(5.12.3) illustrate
the results of these predictions, indicating im-
portant differences between them.
Another demonstration of inherent differences
between the eight methods of transformation is
given in Figure 5(5.12.3) where eight predicted
adaptation shifts are plotted for each of five
samples of the primary Munsell hues (5R, 5Y,5G,
5B,5P) and Munsell chroma 10 (except for
Munsell hue 5Y for which the maximum chroma
—
--~ SP BARTLESON (1977)
=—-
0.2 u’ 0.3 0.4
Fig. 2(5.12.3). Portion of CIE 1976 (u’, v’)-chro-
maticity diagram showing lines of constant Munsell
hue and chroma that are predicted by different
methods for a change from daylight (D,. )-adapta-
tion to tungsten-light (A) adaptation. The different
prediction methods are those referred to in the
text: (a) Helson et al. (1952), (b) Burnham et al.
(1957), (h) Bartleson (1977). (From Bartleson,
1978a, b)
attainable for all eight methods was 6). The largest
discrepancies occur in the blue and purple region
(Munsell hues 5B and 5P).
In comparing the predictions made by the
eight transformations, as illustrated particular-
ly in Figures 2(5.12.3) to 4(5.12.3), Bartleson
(1978a, b) divided them into two groups which he
identified as Type I and Type II predictions.
Type I predictions are obtained by transforma-
tions that accord in whole or in part with the von
Kries hypothesis. There is obviously some degree
of arbitrariness in differentiating between all those
 Chromatic Response Functions 451

0.6
8 Se See
5GY (d)
5G — = ee a =o 5Y
FE 2 SYR
v’ 5BG -* Po
; 3 25R
—— qe < = a ees

0.5} ae ee FORE
5B me <a, ae Gn

st ea SP
| ~
MAC ADAM (1963)
0.1 OF2 03 0.4

—— a B~x0.2
5p
SOBAGAKI ET. AL. (1974)
0.1 0.2

Fig. 3(5.12.3). Same as Figure 2(5.12.3) but for prediction methods (d) MacAdam (1963), (e)
Takahama et al. (1977), and (c) Sobagaki et al. (1974). (From Bartleson, 1978.)

transformations that are not strictly von Kries which the different investigators developed their
transformations, but a pattern seems to emerge methods of prediction. However, these systematic
when one considers the differences in the shapes differences are not as yet fully understood and
of the loci of constant Munsell chroma predicted need further exploration. For some tentative
by the different transformations as shown in Fig- explanations, see Takahama et al. (1979) and
ures 2(5.12.3) to 4(5.12.3). Type I predictions are Bartleson (1979a, b).
those of Burnham et al. (1957), Bartleson (1977),
Takahama et al. (1977), K. Richter (1977), and
Pointer et al. (1977). Type II predictions are those
of Helson et al. (1952), MacAdam (1963), and 5.13 CHROMATIC RESPONSE FUNCTIONS
Sobagaki et al. (1974). When the data correspond-
ing to Type I and Type II predictions are aver- The opponent-process (or opponent-colors) theory
aged separately, the results shown in Figure has its origin in Hering’s (1878) proposal that any
6(5.12.3) are obtained. hue can be described in terms of its redness or
The differences between Type I and Type II greenness and its yellowness or blueness. Red and
data are considered quite large and indicative of green are opposite hues because they cannot be
the systematic differences actually found between elicited simultaneously by a single color stimulus;
the various experimental data on the basis of and the same is true for yellow and blue. The two
 0.6

5PB B=0.2

K. RICHTER (1977)

0.1 O.2 i 0.3 0.4

Ol 0.2 ess 0.4

Fig. 4(5.12.3). Same as Figure 2(5.12.3) but for prediction methods (f) K. Richter (1977), and
(g) Pointer et al. (1977). (From Bartleson, 1978a, b.)

0.6

SY 5/6
LG
05 5G 5/10
5R 5/10
y’

0.4
5B 5/10 JP 5/710

0.1 0.2 1 0.3

Uu
Fig. 5(5.12.3). Adaptation shifts predicted by eight different methods for five Munsell samples
first viewed under daylight (D,;) (common points) then under tungsten light (A). (Bartleson,
1978a,b.)

452
 Chromatic Response Functions 453

0.6 Type I
ae nee a? GY

G SEO SSS O ==, Y

= ie Se = oi SS YR
BG ee al oc FO = ay
vy a 7 She : or
o.5Lf
at era
fe oO ee aT
as~-50<------pad Os eRe
AC HS pet) Drie Le = = be == Sade 10

PB P

0.4
0.1 0.2 4 0.3 0.4

Fig. 6(5.12.3). Lines of constant Munsell hue and Munsell chroma, at Munsell value 5, as
predicted for tungsten-light (A) adaptation by methods classified as Type I and by methods
classified as Type II (Bartleson, 1978a, b).

pairs of opponent hue responses (redness /green-
ness and yellowness /blueness) are independent of
one another, thus allowing, for example, red to be
seen in a single color stimulus simultaneously
with either yellow or blue; or, similarly, green to
be seen simultaneously with either yellow or blue.
Jameson and Hurvich (1955) put the oppo-
nent-process theory on a more quantitative basis
by determining experimentally, using a hue cancel-
lation technique, the spectral distributions of the
opponent hue responses which are often referred
to as the chromatic response functions (or chro-
matic valence functions). The data Jameson and
Hurvich (1955) obtained for their own vision are
illustrated in Figure 1(5.13).
The observer who viewed a 1° test stimulus
exposed repeatedly for several seconds in a 37°-
diameter “white” surround (luminance 32 cd -
m ~*), consisting of a mixture of a monochro-
matic test stimulus of wavelength A (luminance 32
cd-m 7”) with an adjustable amount of one of
four so-called cancellation stimuli of fixed wave-
length: 467, 490, 588, and 700 nm (observer H);
475, 500, 580, and 700 nm (observer J). The first
three of these were chosen as the monochromatic
stimuli evoking for each observer “perceptually
unique” blue, green, and yellow hues, whereas the
fourth (700 nm) evokes a predominantly red hue,
the “unique” red being normally “extraspectral.”
Jameson and Hurvich remark that there is
nothing critical about the use of perceptually
unique stimuli but that the analysis is simplified
by doing so. By repeated adjustments of the radi-
ance L(A.) of the cancellation stimulus, the
condition is reached when its mixture with the
monochromatic test stimulus of radiance L,(A)
454 VISUAL EQUIVALENCE AND VISUAL MATCHING

(b) For A < 490 nm and A. = 588 nm:

OBSERVER H LAA) + L588) satisfies (YB);

—19(A) = ony [2(5.13)]

(c) For A > 588 nm or A < 467 nm, and A, =

490 nm;
Le(A) + L,(490) satisfies (RG);

VALENCE
CHROMATIC
nal) = ony [3(5.13)]

(d) For 467 nm <A < 588 nm and A. = 700

nm;
LAA) + L,(700) satisfies (RG);
400 500 600 700
WAVELENGTH-nm
—1g(A) = 9 “aw [4(5.13)]
OBSERVER J

Here ny(A) and 7,(A) define, in their respective

spectral regions, the chromatic response function
relevant to the opponent-hue pair yellow-blue, oy
Lu
oO
a
and o, being suitably chosen constant factors.
Lud
aad
<@
Similarly, nr(A) and n,(A) define the other
=

Ss)
chromatic response function. The relative values
=
i¢ of oy and o, are determined by postulating that
=
(2)
a
when two monochromatic stimuli of wavelengths
am
Oo A}, A>, respectively greater than and less than 490
nm, are mixed in such radiance proportions
LAA,): L(A2) that the (YB)-criterion is satis-
fied, the yellowness ny(A,)L.(A,) of the first
stimulus equals with opposite sign the blueness
Np(A7)L(A>2) of the second stimulus. By applica-
400 500 600 700 tion of this postulate to the special case, A, = 588
WAVELENGTH-nm nm, A, = 467 nm, it is readily shown that
Fig. 1(5.13). The chromatic-response functions
determined experimentally by a hue-cancellation Cale L (588)
technique for two observers (Jameson and Hurvich,
[5(5.13)]
0p L,(467)
1955).
Thus, each chromatic response curve is com-
pletely determined in its relative spectral values.
satisfies the criterion (YB): neutral as between It can also be inferred from the postulate that for
blueness and yellowness, or the criterion (RG): a mixture of two wavelengths A, and A, with
neutral as between redness and greenness. The set-
A, >490nm and A, < 490 nm
ting made (by observer H) may be expressed
formally as follows: A, # 588 nm and A, ~ 467 nm

(a) For A < 490 nm and A. = 467 nm: to be neutral-(YB), the radiances L,(A,) and
LAA) + L(467) satisfies (YB); L(A) must be given by

[1(5.13)] L.(Ai) rs —np(Ar)

ny(A) = wee
ACU EOO ee

the right-hand side being determined completely
once 0;/60, has been fixed. Because this informa-
tion is not part of the actual experimental data
contained in Figure 1(5.13), the postulate goes
beyond a mere normalizing convention. The ratio
0,/Op is similarly obtained as
0, _ L.(490)
[7(5.13)]
Gp L,(700)
where the mixture of the two stimuli of wave-
lengths 490 and 700 nm is neutral-( RG).
The values of the factors (oy,o,) for one
chromatic response curve relative to those of the
other factors (o¢, 6p) are determined by subjec-
tive estimates of the monochromatic stimuli for
which the hue sensation appears to partake equally
of two qualities belonging to different opponent
pairs, for example, equal redness and yellowness,
equal greenness and yellowness. When the
ordinates of the two chromatic response curves
are made equal (in absolute magnitude) or nearly
so, at such estimated wavelengths, the curves are
completely determined except for an arbitrary
overall factor applied to both.
Romeskie (1976) repeated, for two observers,
the hue cancellation experiment and her results
agree closely with those obtained by Jameson and
Hurvich shown in Figure 1(5.13).
The generality and hence the usefulness of the
chromatic response functions as a database for
making quantitative predictions of other experi-
mental data, such as wavelength discrimination
and chromatic adaptation, rests to a large extent
on the assumption that the /inearity laws for
opponent-hue cancellation, which comprise the
linear operations of proportionality and additivity,
are strictly valid. The work by Jameson and
Hurvich (1955, 1956a, 1959, 1964, 1972), Hurvich
and Jameson (1955, 1958, 1960, 1961, 1974), and
Hurvich (1978) appears to support that assump-
tion. However, direct experimental tests of the
linearity laws, made by Larimer, Krantz, and
Cicerone (1974, 1975) and Cicerone, Krantz, and
Larimer (1975), have brought out certain failures
of these laws.
The work of Krantz and his colleagues centers
on proportionality and additivity tests of stimuli
that are perceived to be in (red/green)- and in
(yellow /blue)-equilibrium. A (red/green)-equi-
librium stimulus is defined as a stimulus that
appears neither reddish nor greenish; that is, one
that is perceived as having either the “unique”
hue of yellow, or the “unique” hue of blue, or is
Chromatic Response Functions 455
perceived as being achromatic (neutral). Similarly,
a (yellow /blue)-equilibrium stimulus is a stimulus
that appears neither yellowish nor bluish; that is,
one that is perceived as having either the “ unique”
hue of red, or the “unique” hue of green, or is
perceived as being achromatic (neutral).
In denoting the set of all stimuli (in
(red /green)-equilibrium by S, and the set in (yel-
low/blue)-equilibrium by S,, the linearity laws
may be formulated as follows:
(a) Proportionality Law:
If the stimulus A is in S;, then the stimulus aA is
also in S; (i = 1,2), where a is any positive factor
by which the radiance of the stimulus is increased
or reduced, while its relative spectral distribution
is kept the same.
(b) Additivity Law:
If stimulus A is in S,, then stimulus B is also in S,
if, and only if, the additive mixture of the two
stimuli, that is, A + B, is in S; (i = 1,2).
Krantz (1975b) has presented a mathemati-
cally formal and detailed exposition of the linear-
ity laws in the context of the opponent-process
theory. He shows that the two linear operations of
proportionality and additivity hold only if the
corresponding chromatic response function is a
linear function of the color-matching functions
7(A), 2(A), b(A) derived from matches made be-
tween appropriate additive mixtures of mono-
chromatic test stimuli of wavelength A and three
primary stimuli (Sections 3.2 and 3.3). More
specifically, the proportionality law (a) and the
additivity law (b) hold for stimuli of set S; only if
there exists a real-valued linear function ¢; such
that
B is in S, if, and only if, ¢,(B) = 0
The function ¢; is linear with respect to the
operations of trichromatic color matching; that is,
for any stimuli A and B, and any positive factors
of a, B, the following condition is met:
¢;(aA + BB) = a¢,(A) + B4,(B)
[8(5.13)]
It follows that the function ¢, is a linear function
of any set of primary stimuli. It can be de-
termined experimentally by the hue cancellation
method which results in ¢; being a chromatic
response function for monochromatic stimuli of
456 VISUAL EQUIVALENCE AND VISUAL MATCHING
wavelength A and of constant radiance (equal-
energy spectrum on the wavelength basis), which
is independent, except for a constant scale factor,
of both the radiance level (or luminance level) at
which the determination is made and the choice
of cancellation stimulus.
The tests of the linearity laws conducted by
Larimer et al. (1974, 1975) and Cicerone et al.
(1975) consisted first of the determination, at
several luminance levels, of the monochromatic
equilibrium stimuli (“unique” blue, green, and
yellow) and the extraspectral equilibrium stimu-
lus, “unique” red. Then, the results of adding the
equilibrium stimuli yellow and blue, or the equi-
librium stimuli red and green, were studied. In
accordance with Krantz (1975b), these specific
tests of the linearity laws (proportionality and
additivity) are exhaustive, since the equiliibnum
stimuli act as primary stimuli with respect to
cancellation-equivalence. Two stimuli are defined
as cancellation-equivalent when their hues are
canceled by the same stimulus. If proportionality
and additivity hold, then multiplying each stimu-
lus of a cancellation-equivalent pair by a constant
factor (a, for example), or adding a third stimulus
to each will not upset the cancellation-equiva-
lence. It is important to note that cancellation-
equivalent stimuli are not necessarily perceived to
be visually equivalent in every respect, that is, in
complete color match. For example, desaturating
a red stimulus with an equilibrium “white”
(achromatic or neutral) stimulus makes the de-
saturated stimulus cancellation-equivalent to the
original stimulus, even though its perceived red-
ness decreases.
In tristimulus space, based on the particular
observer’s color-matching functions, strict validity
of the linearity laws, (a) and (b), means that a set
of equilibrium stimuli must be represented by
points located on a straight line. Thus, it is suffi-
cient to test whether the four equilibrium
(“unique”) stimuli represented by four points in
tristimulus space are invariant with luminance
and whether all additive mixtures of these stimuli
are, in turn, equilibrium stimuli.
The Bezold—Briicke and Abney effects (Sec-
tion 5.9) are obviously related to the question of
the validity of the linearity laws. The proportion-
ality law demands that the equilibrium stimuli are
invariant and thus, show no Bezold—Briicke hue
shift with increasing or decreasing luminance. The
additivity law demands that the equilibrium
stimuli are invariant and thus, show no Abney
hue shift when they are desaturated by means of
other equilibrium stimuli, in particular, by means
of an equilibrium “white” stimulus.
The results obtained by Larimer et al. (1974,
1975) in their linearity tests may be summarized
as follows:
(i) The wavelengths of the monochromatic equi-
librium stimuli perceived as “unique” yellow (~
578 nm), “unique” green (~ 500 nm), and
“unique” blue (~ 478 nm) were invariant, or
virtually so, with luminance changes ranging over
two log units. Zero Bezold—Briicke hue shifts
were observed for these three stimuli.
(ii) For the equilibrium stimulus perceived as
“unique” red, which is an extraspectral stimulus,
that is, one that cannot be represented by a point
on the spectrum locus, invariance with luminance
did not hold. When the equilibrium stimulus was
produced by an appropriate mixture of a mono-
chromatic stimulus of long wavelength and a
monochromatic stimulus of short wavelength, the
red equilibrium stimulus turned bluish-red when
the luminance was increased while keeping its
wavelength composition constant.
(iii) The additive mixture of the monochromatic
equilibrium stimuli perceived as “unique” yellow
and blue, in any luminance ratio, is again an
equilibrium stimulus. “Unique” yellow and
“unique” blue are found to be complementary
hues relative to an equilibrium “white”. No Ab-
ney hue shift was observed when a yellow equi-
librium stimulus or a blue equilibrium stimulus
was desaturated with the equilibrium “white”
stimulus.
Additive mixtures of red and green equi-
librium stimuli are again equilibrium stimuli, that
is, neither yellowish nor bluish stimuli.
From the above experimental results, Larimer
et al. (1974, 1975) draw the following conclusions.
(1) The set of (red/green)-equilibria, that 1s,
stimuli that appear neither reddish nor greenish,
obeys the linearity laws (proportionality and
additivity). Thus, the (red/green) chromatic
response function, such as that measured by
Jameson and Hurvich by the method of hue
cancellation [Figure 1(5.13)], is a linear function
of the observer’s color-matching functions.
(II) The set of (yellow/blue)-equilibria, that is,
stimuli that appear neither yellowish nor bluish,
does not strictly obey the linearity laws. The
observed nonlinearities are assumed to be associ-
ated only with the responses of the long-wave-
 Chromatic Response Functions 457

length (“red”) cone mechanism which contributes
to the perception of yellowness, but via a “com-
pressive function of luminance.” Thus, the rela-
tive values of the (yellow/blue)-chromatic-
response function of Figure 1(5.13), determined
for an equal-energy spectrum, must vary some-
what with the radiance level.
In a subsequent paper by the same authors
(Cicerone, Krantz, and Larimer, 1975), the linear-
ity laws are studied for moderate changes in the
observer’s state of chromatic adaptation caused
by conditioning stimuli of “unique” blue, green,
yellow, and by a (nonunique) monochromatic red
stimulus of 650 nm, varying in luminances that
provided retinal illuminances from 40 to 2000 td.
The main results of this study indicate that adap-
tation to blue or green stimuli did not introduce
significant failures of additivity of (red /green)-
equilibria; but adaptation to yellow stimuli of 900
td or more resulted in failures of proportionality
and thus, additivity. A small proportionality
failure was also observed for red-conditioning
stimuli at 40 td. The (yellow /blue)-equilibria ex-
hibited nonlinearities under chromatic adaptation
that were of similar kind to those observed in the
previous experiments, described above, which ap-
plied to conditions of dark adaptation.
In the application of chromatic response func-
tions to a wide range of color-vision problems,
Hurvich and Jameson (1955) assume for their
own experimentally determined functions that the
linearity laws hold strictly. This assumption per-
mits them to substitute their chromatic response
functions by others constructed from appropriate
linear combinations of the CIE 1931 standard
color-matching functions x(A), y(A), Z(A). The
new chromatic response functions are then re-
garded as applicable to the CIE 1931 standard
observer. Figure 2(5.13) illustrates the chromatic
response functions postulated for the CIE 1931
standard observer by Hurvich and Jameson (1955)
(see also Judd, 1951b). For the luminance level
(32 cd- m *) of the stimuli used in their experi-
ment, they put
[y(A) — b(A)] = 0.4y(A) — 0.427(A)
[r(A) — g(A)] = 1.0x(A) — 1.0~()
[9(5.13)]

WHITE ADAPTATION
0.75

0.50
a

O
Pa 0.25
Lu
=|
=
Ss
O 0.00
at

<x
=
©
+ 0,25
O

0.50

O75

400 500 600 700

WAVELENGTH-nm

Fig. 2(5.13). Chromatic-response functions postulated for the CIE standard observer (Hurvich
and Jameson, 1955).
458 VISUAL EQUIVALENCE AND VISUAL MATCHING
where the quantities [y(A) — b(A)] and [r(A) —
g(A)] are the (yellow/blue)- and (red/green)-
chromatic-response functions represented in Eqs.
1(5.13) to 4(5.13) by [mny(A), ng(A)] and
[np(A), ng(A)], respectively, and which define also
the functions ¢, (i= 1,2) referred to in Eq.
8(5.13).
The chromatic response functions characteriz-
ing two opponent processes in the visual system
are supplemented by an achromatic response
function that characterizes a nonopponent pro-
cess. The achromatic response function is
commonly identified as the luminous efficiency
function obtained by the minimum-flicker method
(Section 5.7). There is considerable evidence from
electrophysiological research on the visual mecha-
nisms of different species that the response func-
tions of the above three types are being generated
in “color-coding sites” located in one or more of
the different parts of the visual tract starting in
the retina and ending in the visual cortex. An
informative review of the present state of our
knowledge on this aspect of vision research can
be found in Boynton’s (1979) book entitled Hu-
man Color Vision (pp. 207 to 250). However, the
precise relation between the responses of the cone
receptor mechanisms and the responses generated
in the color-coding sites is not known. Hurvich
and Jameson (see Jameson, 1972) postulate the
following general form
@ (A) = —en pA) + C2 p2(A) + €1373(A)
@.(A) = P(A) — C22 pa(A) + €23p3(A)
$3(A) = 3, p(A) + ¢32 p2(A) + C33)3(A)
[10(5.13)]
in which ¢$,(A), $,(A) denote the chromatic re-
sponse functions (yellow/blue) and (red/green),
respectively, and ¢$,(A) denotes the achromatic
response function (light/dark or white/black).
The functions p,(A) (i = 1 to 3) are the spectral
response functions of the three cone receptor
mechanisms, also referred to as the fundamentals;
PA), pA), p3(A) representing, respectively, the
(never negative) spectral response of short-wave
(“blue”) receptors, middle-wave (“green”) recep-
tors, and long-wave (“red”) receptors. The
coefficients c;, (i,k = 1,2,3) are non-negative
constant factors. Attention is drawn to the two
negative signs in Eq. 10(5.13) which signify the
possibility of having either positive or negative
spectral responses emerging from the opponent-
process mechanisms characterized by the (yel-
low/blue) [¢,(A)] chromatic response function
and the (red/green) [$,(A)] chromatic response
function.
When assuming KO6nig-type fundamentals
(Section 8.2.5) for p;(A), Jameson (1972) arrives
at the following explicit equations:
6 (A) = —Odlp Xd) + 0.06p,(A) + 0.34p,(A)
$(A) = 0.37p (A) — 2.23p.(A) + 1.66p3(A)
$3(A) = 0.15p,(A) + 0.85p3(A)
[11(5.13)]
5.14 COLOR-MATCHING PROPERTIES OF
COLOR-DEFECTIVE OBSERVERS
There is a vast literature dealing with numerous
aspects of color-defective vision of congenital as
well as acquired types. It is beyond the scope of
this book to even attempt a comprehensive trea-
tise on the subject matter. However, certain
aspects of color-defective vision are of consider-
able interest to color science in general, and some
of these, notably those related to color matching,
we have selected for inclusion in this section. The
reader interested in a more comprehensive treat-
ment of a broader range of topics of color-defec-
tive vision is encouraged to consult the recent
book entitled Congenital and Acquired Color Vi-
sion Defects edited by Pokorny, Smith, Verriest,
and Pinckers (1979). Other valuable reviews on
color-vision deficiencies are those by Hurvich
(1972), Jaeger (1972), and Griitzner (1972).
Variations of color-matching properties among
actual observers fall under the following head-
ings:
(a) Variations among observers classed all as
normal trichromats, or as some particular type of
color defective
(b) More radical deviations between normals
and the several varieties of color defective
Differences in color-matching properties, includ-
ing the related properties of color discrimination,
provide, in fact, the basis for the scientific classi-
fication of observers into normal and color defec-
tive. In the classification now to be summarized,
it may be taken, unless otherwise indicated, that
matching is carried out in a small field imaged on
 Color-Matching Properties of Color-Defective Observers
the foveal rod-free area of the retina. Where het-
erochromatic brightness-matching is quantita-
tively critical, the use of flicker photometry may
be presumed.
5.14.1 Normal and Anomalous Trichromats
An observer requiring mixtures of three fixed and
independent primary stimuli to color match all
color stimuli is a trichromat. For the observer to
be color-normal, the observer’s full color and
heterochromatic brightness matching (minimum-
flicker method) must obey at least approximately
the linearity laws (proportionality and additivity,
see Section 3.2). Given this, a trichromat is classed
as a normal if the quantitative properties of the
matches are sufficiently close to the average for
the population. These quantitative properties for
an individual trichromat are specifiable by the
observer’s:
(a) color-matching functions 7(A), 2(A), b(A)
for some set of independent primary stimuli
(R, G, B);
(b) luminous efficiency function V(A) (mini-
mum-flicker method); and
(c) brightness and chromatic-difference thresh-
olds (e.g., the least perceptible wavelength dif-
ference through the spectrum).
In testing for color normality, it is convenient
to consider first the observer’s color-matching
functions 7(A), Z(A), b(A) referred to a set of
monochromatic primary stimuli located, respec-
tively, at wavelengths Ap,A¢,A, and expressed
in radiometric units. These functions are con-
verted into the corresponding W.D.W. color-
matching functions
Wi
g*(A) = 8(A) [1(5.14.1)]
b*(A) = i
by means of the constant ratios W, and W, so
chosen that r*(A) = g*(A) at a fixed “normal-
izing” wavelength A y in the yellow, 2*(A) = b*(A)
at a second normalizing wavelength A», in the
459
blue-green (see Section 5.6.1). Thus,
_[ ra) _[ba)
m= [EO ae m= [EQ]
[2(5.14.1)]
The red-green ratio W, provides one of the
main criteria by which the normal trichromat is
separated from the two groups of anomalous tri-
chromats, known as protanomalous when the value
of W, is abnormally high and as deuteranomalous —
when it is abnormally low.
In practice, advantage is taken of the fact that
the blue contribution b*(Ay) (which is negative,
i.e., it has to be added to the yellow to complete
the match) is very small and can be omitted
without seriously upsetting the match or altering
the red and green contributions. The red-green
ratio, which was first used by Rayleigh when he
discovered protanomaly and deuteranomaly, is
measured on an instrument known as an anoma-
loscope, the Nagel anomaloscope being particu-
larly well known. In a large randomly selected
group of observers, the measured values of the
R/G ratio are mostly clustered about a central
value, but on either side there occurs a second,
much smaller, concentration of measured values
for deuteranomalous observers and for pro-
tanomalous observers. Figure 1(5.14.1) illustrates
the results of tests made on 864 males by Schmidt
(1955). Earlier, somewhat less distinct data have
been published by Trendelenburg and Schmidt
(1935) and Nelson (1938).
The assignment of a limiting R/G ratio sep-
arating normal and deuteranomalous observers is
straightforward, but there is a more arbitrary
element in the choice of a limiting value separat-
ing normals from protanomalous observers. By
determining R/G values relative to the central or
average value for a large group, which is taken as
the representative normal value, the need for an
absolute calibration of the results given by an
anomaloscope is avoided. The relative R/G val-
ues so obtained are not critically dependent on
the exact wavelength used for Az, Ag, and Az,
nor on the use of monochromatic light; a
narrow-band filter anomaloscope may be used.
Although no agreed limits have been laid down,
reasonable limiting values for the relative R/G
values are
R/G greater than 1.5 for protanomaly
R/G less than 1/1.5 for deuteranomaly
460 VISUAL EQUIVALENCE AND VISUAL MATCHING

100

Normal

Deuteranomalous Protanomalous
20 --£a 7Scale—
(%)
Observers
of
Number rc mad
1
Scale—> (%)
Observers
of
Number
eKte ee
ar L ‘

Log (R/G)
Fig. 1(5.14.1). Distribution of the values of log(R /G) where R /G incorporates a factor that
makes its average value for the normal trichromat equal to unity. The number of observers,
given in percent, is shown on a scale from 0 to 100% for the midpoint values of log(R /G) =
— 0.1, 0.0, + 0.1. For all other midpoint values, the expanded scale from 0 to 10% on the right
applies. 63.6% of the observers have log(R /G) values in the interval —0.05<0< +0.05;
1.8% have log(R /G) values in the interval —0.55 < —0.50 < —0.45; 0.6% have log(R /G)
values in the interval 0.55 < 0.60 < 0.65. The remaining 34% observers are distributed as
indicated. The data are taken from Schmidt (1955) who tested 864 males.

An observer who measured the R/G ratio after
inserting between the eye and the instrument a
selectively transmitting filter would obtain in-
stead of the true value R/G, an apparent value
Rt;/Gtpr, where tg and ftp represent the filter
transmittances at A, and Ap, respectively. Simi-
larly, inert selectively transmitting pigments in
the path of the rays within the eye will influence
the observed R/G value. However, the inert pig-
ments known to be present in the eye lens and the
macular retina are believed to be nearly nonselec-
tive in the wavelength range green to red, and are
unlikely to falsify the diagnosis of deuteranomaly
or protanomaly from the relative R/G value. In
any event, these anomalies as a rule are also
marked by reduced color discrimination in the
green to red spectral range, and measurements of
color thresholds, such as least-perceptible dif-
ferences of wavelength at constant luminance,
provide independent evidence of the anomaly,
which is unaffected by possible inert eye pigmen-
tation. Diminished color discrimination has the
consequence that the yellow match on the
anomaloscope can be reached over a relatively
wide range of fixed R/G ratios simply by varying
the luminance of the yellow stimulus. Normally,
the setting accepted is assumed to correspond
approximately to the midpoint of the acceptable
range. A more refined application of the anoma-
loscope is to determine, for each observer, the
limits of R/G between which an acceptable match
can be made by luminance adjustment only. Pick-
ford (1964), using this procedure, further subdi-
vides anomalous observers as follows:
(i) Extreme protanomalous or extreme de-
uteranomalous: when the R/G range is very large
and includes the mean value for normals. The
protanomalous and deuteranomalous observers
are then distinguished by their different luminous
efficiency curves (see below).
(ii) Simple protanomalous or simple deuteranoma-
lous: when the R/G range is fairly small but is
widely displaced above or below the mean for
normals.
(iii) Deviants: when the R/G range shows some
enlargement (more than twice the modal range).
(See also Pickford, 1951; Pickford and Lakowski,
1961; and Willis and Farnsworth, 1952).
A much rarer anomalous trichromat, the fri-
tanomalous observer, is marked by poor color
discrimination in the green to blue spectral range,
whereas discrimination in the green to red range
and the R/G ratio are within the limits of the
normal. The use of the blue-green ratio W, =
[3(A)/g(A)],,,, to detect tritanomaly is com-
plicated by the fact that the yellow (shortwave
absorbing) inert pigments of the eye lens and
 Color-Matching Properties of Color-Defective Observers 461

Twelve protanomalous O O Average of six

0.9} Observers (Nelson, protanopes (Pitt)
McKeon and Wright)
x x Average of six
Six deuteranomalous deuteranopes (Pitt)
observers (Nelson) ——

efficiency
Luminous

©w | | Normal limits
AK (Gibson and Tyndall)©

= ho

0.1

an
0.0 vee
500 600 650
at Wavelength, nm

Fig. 2(5.14.1). Luminous efficiency for normal and color-defective observers, showing individ-
ual variations. For normals, the individual curves—all with maximum value unity—lie in the
tract defined by the arrowheads. The limits have been determined from the data of Gibson and
Tyndall (1923) for 52 observers (from Judd and Wyszecki, 1975).

retina have an important effect on the ratio
(Moreland, 1972b; Moreland and Kerr, 1978).
Ideally, the luminous efficiency curve V(A) of
a normal trichromat is a linear combination of
the color-matching functions 7(A), Z(A), b(A),
that is
V(A) = Laf(A) + Lez(A) + Lgbh (A)
[5(5.14.1)]
where only the relative values of the luminous
efficiency factors Lr, Lz, Lz are significant.
These factors are formally independent of
r(A), Z(A), b(A) in the sense that two trichromats
who agreed exactly on all their complete color
matches—and hence possessed identical color-
matching functions 7(A), Z(A), b(A)—could
conceivably show very different V(A) curves in
heterochromatic brightness matching, correspond-
ing to different sets of values of the factors
Lr, Lg, Lg. The considerable range variation in
the V(A) curves of normal trichromats [see Figure
2(5.14.1)] is attributable in part to individual dif-
ferences in Lp, Lg, Lz (see also Section 5.7).
Differences between the luminous efficiency
functions of anomalous and normal trichromats
must be attributable to the combined effect of
differences in color-matching functions and
luminous efficiency factors. However, for de-
uteranomalous observers, the resultant effect is
not sufficient to make the measured V(A)-curves
clearly distinguishable from those of the normal
group, allowing for the considerable spread in
these latter curves. On the other hand, the V(A)-
curves of protanomalous observers differ signifi-
cantly from the normal curves. In Figure 3(5.14.1)
the mean curves are plotted for comparison,
showing that the protanomalous curve is narrower
and displaced towards shorter wavelengths. The
salient difference is the very low luminous ef-
ficiency shown by the protanomalous in the red
end of the spectrum, and it is this difference
(“shortening of the spectrum in the red’’) that
provides a valuable additional diagnostic test dis-
tinguishing deuteranomalous and protanomalous
observers. It is pointed out, however, by Wright
(1946) (who with his collaborators provided the
quantitative basis for many current conclusions
on the properties of color defectives) that the
 462 VISUAL EQUIVALENCE AND VISUAL MATCHING

1.0 AIS
NORMAL
\ 7 pEUTERANOMALOUS
\_/ PROTANOMALOUS

s
Oo
0.8
r
za
a
9
te
w O06
wn
=)
ro)
2
S
=
04

0.2 ‘ye |

400 9 500 600 700

WAVELENGTH, nm

Fig. 3(5.14.1)._ Mean luminous efficiency curves for the normal trichromat, the protanomalous
and the deuteranomalous observer (from Wright, 1946).

method of selection of observers used for the
protanomalous V(A)-curve of Figure 3(5.14.1)
may have excluded some observers showing a less
marked difference in this respect from the nor-
mal. There is no evidence that tritanomalous ob-
servers have V(A)-curves differing appreciably
from the normal.
The chromaticity coordinates of the mono-
chromatic stimuli, expressed in the W.D.W. sys-
tem,
Payee)
yk
RE Oa)irtsh Cy) atebS CA)
and similar expressions for g*(A) and b*(A) (see
Section 5.6.1), would be the same through the
spectrum for anomalous and normal observers, if
their differences in color vision were equivalent to
the effect of different selective pigment filters in
the eyes of the two observer types. For pro-
tanomalous observers, the chromaticity-coordi-
nate curves differ little from those of the normal
(see Section 5.5.6), and in this respect their
anomaly corresponds predominantly to a filter
difference. Nevertheless, a filter difference would
not account for the reduced color discrimination,
that is, the high values of the hue threshold in the
red-green region. The chromaticity-coordinate
curves of deuteranomalous observers show more
pronounced differences from the normal curves,
and, in general, such observers have a similar
reduced color discrimination. For details of chro-
maticity-coordinate curves of anomalous ob-
servers, Wright’s (1946) book entitled Researches
on Normal and Defective Colour Vision 1s particu-
larly recommended.
Several hypotheses have been advanced to
account for congenital anomalous trichromatic
vision, but so far none appears to have found
general acceptance. Some investigators have
searched for clues in the shapes of the spectral
absorbance curves of the photopigments in the
cone mechanisms and the positions of their max-
ima in the spectrum (photopigment-alteration hy-
pothesis). Such studies consist mainly of analyses
of the means and the variability of color-match-
ing data obtained by anomalous trichromats in
comparison with normal trichromats (von Kries,
1899; De Vries, 1948; Wald, 1966; Alpern and
Torii, 1968; Vos and Walraven, 1971; Pokorny,
Smith, and Katz, 1973; MacLeod and Hayhoe,
1974; Piantanida, 1974; Ruddock and Nagh-
shineh, 1974; Pokorny, Moreland, and Smith,
1975; Hayhoe and MacLeod, 1976; Alpern and
Moeller, 1977; Alpern and Pugh, 1977; Alpern
and Wake, 1977).
Miller (1924) and Jameson and MHurvich
(1956b; see Hurvich, 1972 for a review) assume
 Color-Matching Properties of Color-Defective Observers
photopigment alteration as only partly responsi-
ble for anomalous trichromatic vision and
postulate, in addition, neural changes at post-
receptor levels, specifically at the color-coding
sites at which opponent-processes are being gen-
erated (Section 5.13).
5.14.2. Dichromats
Observers requiring mixtures of just two fixed
and independent primary stimuli to match all
color stimuli are called dichromats. They are fur-
ther subdivided into two main types—protanopes
and deuteranopes—and a third, much rarer, type
—tritanopes. The salient properties of the color
vision of the various kinds of dichromat are sum-
marized in Table 1(5.14.2). Sets of color-matching
functions and luminous efficiency functions de-
fining ideal or standard dichromats, comparable
with those of the CIE standard colorimetric
observers for normal trichromats, have not been
laid down. The current quantitative picture of
dichromatic vision is based, to a large extent, on
measurements by Pitt (1935) on protanopes and
deuteranopes, and by Wright (1952) on tri-
tanopes. However, numerous more recent studies
have added valuable information to the data bank
(Judd, 1945; 1948; 1951b; 1964; DeVries, 1948;
Judd, Plaza, and Farnsworth, 1950; Thomson and
Wright, 1953; Hurvich and Jameson, 1955; Walls
and Heath, 1956; Sperling, 1960; Rushton, 1963;
1965b; Hsia and Graham, 1965; Cole, Henry, and
Nathan, 1966; Cole and Watkins, 1966; Heath,
1968; Scheibner and Boynton, 1968; Nimeroff,
1970; Verriest, 1971; Vos and Walraven, 1971;
Vogt, 1974; Walraven, 1974; Smith and Pokorny,
1975; 1977; Massof and Bailey 1976; Alpern and
Moeller, 1977; Alpern and Pugh, 1977; Padmos,
van Norren, and Faijer, 1978; Tansley and
Glushko, 1978; Paulus, 1979; Nagy 1980).
The color-matching functions, luminous ef-
ficiency functions, and data on color discrimina-
tion for dichromats are usually determined with
small (1 to 2°) centrally viewed fields. When large
(~ 8°) centrally viewed fields are used, many
observers, diagnosed as dichromats under the
small-field condition, take on a form of trichro-
matic vision (Nagel, 1905; Scheibner and Boyn-
ton, 1968; Smith and Pokorny, 1977; Nagy, 1980).
This phenomenon, though of considerable im-
portance, is not fully understood. It is postulated
that responses of the parafoveal rod receptors add
the third dimension to dichromatic cone vision,
making the observer’s vision trichromatic. But
463
other factors may also enter, such as responses of
residual cone receptors that are not contributing
measurably in small field viewing.
Complete color-matching and (minimum-
flicker) heterochromatic brightness matching by
dichromats, made in a small and centrally viewed
field, are believed to conform to the linearity laws
(proportionality and additivity) although these
properties seem not to have been critically tested.
If it is assumed that they are valid, then the
color-matching properties of a dichromat are
specifiable by two linearly independent functions
of wavelength, the color-matching functions,
which are replaceable by any other pair of inde-
pendent functions related to the first by a nonsin-
gular linear transformation. For a dichromat who
will accept any complete color match made by a
particular trichromat, the dichromat’s two color-
matching functions [let us say, h(A) and j(A)
referred to some pair of fixed and independent
primary stimuli] must be independent linear com-
binations of the color-matching functions
7(A), Z(A), b(A) of the trichromat; thus,
h(a) = piyr(A) + prg(A) + p30 (A)
iQ) = put (A) + png (r) + pb (A)
[1(5.14.2)]
In this sense, the dichromat’s color vision is a
reduced form of the trichromat’s. If any given
dichromat will accept all the matches of some
normal trichromat, dichromatic vision may be
said to be a reduced form of normal trichromatic
vision, and this is often believed to be the case.
For a discussion of the possible difficulties associ-
ated with this assumption, the studies of Hurvich
and Jameson (see Hurvich, 1972) and Alpern and
his coworkers (see Alpern and Moeller, 1977;
Alpern and Pugh, 1977; Alpern and Wake, 1977)
are recommended.
It should be noted that if a dichromat will
accept any trichromatic match but only after ad-
justing the luminance (but not the relative spec-
tral composition) of one of the stimuli being
compared, then the dichromat’s color vision is not
a reduced form in the strict sense, and it cannot
be concluded that his or her color-matching func-
tions are expressible as linear combinations of
those of the trichromat.
As an approximation, the color-matching
properties of normal trichromats and of the three
464
Jo
AGEL
JOJO)
IYSLIIIVIVYD snojeuourjoig snojeuouvisjneg adouvrjO1g nog adouvie odour}, JCUIOIYIOUOJA[-PO

wayEg
(Z'pT'S)T
Sansodoig
SOAT}N9JA(I
IO[OD UONCUIULIOSIP ATTeLIA}eypsonpalWOT]pol0} yUasqyWo;94} pdI WOIJjUIsqYy
oy} pol JUasqy
UT oy} ON IO[OD-SIP
Ysnom}
ay} wingjdeds U9dI3-YSIMOTIOA
1NG0} & SUIAIVA 0} INogeQZS wu 0} NoeOES wu qG-YsTUse13
on] 0} UOT}CUTUITIO
VoIZap
Ul JUDIIJJIP
SdSED ONT Spp)0} QP (WU
jenayutod“a't) osuON suON ¢6¢v—067
wu ¢OS—S6P
wu 89¢ pue QL¢ wu [TV SyISUDTOARM
IEWUIOIYIOUOUT JO YSUITIAPM
B SoyoyeUw JeY} SNp[NUiT}s
p(SQINUINS ,.21TYM,, POX

JO
ON
ON
ON

dy}
pol
So
sox
SOX

JO
BuO]
paonpel
YISUITOAR
JO OY} UINUITXBUT Ops wu 09¢ wu Ops wu

3utua}10NS
09¢ wu ¢¢¢ wu /0¢ wu

snoulumny
Jo snouruny AQUITOIJa
dAIND

AOUdTOTIJJS
AID 1€61 Ajoneworyo
Jo si al

‘d°T)(syyZuaTaaeM
(LO *Px= 0801 =x IZ1'0 a
34} woIsnjuod
yutod k Cera =7?a Oso'o— o=°"
syeutoryoIp)
g(ATuo
a8e}U9019g Aouanboayy
jo
,90UdINIDO
Suoure soyeu 01 61 Ol Lh 700°0 €00°0
Suoure soyeutey c0'0 8€°0 c0'0 10°0 100°0 c00°0

10.7, sJ0UI payteiap

BIeP99S Iq ‘(SE61) syempure MeaH] ‘“(9¢6])
pue jossep
pue Ao[reg (9/61)
soy sadourjoid
pue ‘sadouvsajnap
pure BOD39 ‘Te (9961)
IOJ OM}
sadoueiL
ou) puUOdes [eNIU yUIOd UT dy} aAeMZIOYS
pusJo oy} uNIDeEds
sem punoyJOJ OY} ‘(OM YW UOTOUTISIP
St SOUITJOUMOS
IPeU UsaMjoq [ernou,,.jutod
pue
oneMOIYOR,,
,,“WuUIod
ay} 19};e] BuTaq PouTjapSe dy} YISUITIAeM
JO dy} oNeWOIYSOUOUT
snNUTs
yey) seadde ssapJopoo
0} & ‘JEUTOIYSIP
: UT, 19ex2 UONRIOT
JO ay} UOIsNJUOD
syuTod
SI j0U uMOUYpue sreodde
01 9q yoalqns
0} a[qQeJapIsuod
UOTeLeA UsIMJOq “S[ENPIAIPUTJUAIIJJIC| SUOTIESSIAUT
OARY
opeu JWOIOTJIP SdIeUIT}Sa
Sv 0} ITY} JSOW AJOAT] ‘sUOTLIOT
YW sAIsUaYyoIduIOD
UOTFE[NQe}
JO ay} suoTeoo]
Jo ay} uorsnyuoo
syutodpasnAq JUarayJIp sioyeStsaaut
dn 0}
196] sey usI2q opeur
Aq AaNsIA],(L961)
JO.J Ize] wep ‘vas Joj ‘o[durexa JJOIOWIN“(0/6])
SOA pueA Woaese“(1L61)
Ay Uoaese“(pL61) yITUgpue Ausoyog‘(SL61)
(9L61) wiodty pue |
10,7, [eUOMIppe
eiep ‘sas Joy ‘a_durexa ‘Ausoyog“YIUIg “sara
pue SOxUTA *6/6]) Typ SopqeL
['L pue (LL
 Color-Matching Properties of Color-Defective Observers 465

types of dichromat are expressed by typical or planes

average sets of color-matching functions. For nor-
mals, in particular, these are expressed by the set P, = pyyX + PY + py3Z
x(A), y(A), Z(A) defining the CIE 1931 standard
observer. To this approximation, the condition Py = PipX + PY + p73Z [6(5.14.2)]
that protanopia, deuteranopia, and tritanopia
should be reduced forms of normal trichromatic which cannot be parallel because the linear forms
vision means that their typical sets of on the right are independent. The corresponding
color-matching functions can be written as the lines for any other pairs of bistimulus values
following pairs of independent linear forms in Pi = kP,, P3 = kP, (k #0) which are in the
same ratio as P,, P, and correspond, for the pro-
X(A), y(A), 2(A).
For the protanope: tanope, to stimuli with the same protanopic chro-
maticity
PA) = puixX(A)prv(A)pi3Z()
Pi P;
P(X) = Pri X(A)PrV(A)p232(A) Pee Pep ap
[2(5.14.2)] all lie in the plane Ol through the line | and the
origin O. If P,; # 0, the plane Ol has the equation
For the deuteranope:
Q0= (pPi Spay)
d\(A) = dy, X(A) dy H(A) d,32(A) (pp — Pa)¥

d.(A) = dy,X(N)
doy y(A) dy32(A) + (P13 — P23)Z

where p = P,/P,. If P, #0, the equation is

[3(5.14.2)]
For the tritanope: O= (Pi — OP21)
X + (Pir — 9py)Y

t(A) = ty)X(A )ty (A) t132(A)
t,(A) = to) X(A )ty Y(A)to32(A)
[4(5.14.2)]
The necessary and sufficient conditions for two
color stimuli {P, dA} and {Px dX}, to be in
match for a particular type of dichromat—for
example, the protanope—are, in terms of the
color-matching functions, given by
P= r Pypi(d) dd = fr PKp(A) dd = PY
+ (P13 — 9P23)Z
where o = P,/P,. If neither P,; nor P, equals
zero, either equation defines the plane Ol. The
plane Ol and the plane 1 = X + Y + Z, in gen-
eral, intersect in a line m, all points of which
represent color stimuli with a common protanopic
chromaticity. Their coordinates X,Y, Z are the
trichromatic chromaticities
xX
x
a ts Ae DEA vf = Jy;
Cs
of the stimuli. The exceptional case in which
plane Ol is parallel to the plane 1 = X + Y + Z,
can arise only if the determinant

P, =[P.72() dd = [PXP2) dd = P5 l l 1
Do— PUA eeeLAG eee 2
[5(5.14.2)] slo Pratl 2

where P,, P, and P;, P5 define, respectively, the
protanopic bistimulus values of the two stimuli.
The tristimulus values X, Y, Z, of all stimuli hav-
ing the same fixed bistimulus values P,, P, (not
both equal to zero) lie on a straight line | in
(X, Y, Z,)-space defined by the intersection of the
equals zero and then only if p (or o) has a
particular value. There is in that case no line of
intersection (except at infinity). Excluding this
possibility, all the lines m corresponding to differ-
ent protanopic chromaticities (i.e., to all possible
values of p and o) intersect in a common point
466 VISUAL EQUIVALENCE AND VISUAL MATCHING

(X59 Yncs Zp-) Which satisfies the equations

CIE 1931
Chromaticity diagram

|, yenh) pe
pe
bps 7Afae

Omi Xpe | Puen | P15 oe

O = paXpe t+PrrVpc + P23 Ze

[7(5.14.2)]

This point, sometimes called the confusion point,

is determinable experimentally by plotting in the
trichromatic chromaticity diagram the loci
(straight lines) representing stimuli of the same
protanopic chromaticity (i.e, which the pro-
tanope can match by luminance adjustments only)
and by then noting that for different protanopic
chromaticities the lines converge, at least ap-
proximately, on a single point, as illustrated in CIE 1931
Chromaticity diagram
Figure 1(5.14.2).
Protanopes, dueteranopes, and tritanopes are
distinguished by leading to different confusion
points [see Table 1(5.14.2)]. Given that each of
these types of dichromatic vision is a reduced
form of trichromatic vision, a knowledge of the
trichromatic chromaticity (x. = X., y. = Y.) of
the appropriate confusion point, determines the
dichromatic color-matching functions completely
except for the arbitrary element corresponding to
the choice of primary stimuli. This arbitrariness
may be conveniently removed, say for pro-
tanopes, by choosing any two monochromatic
stimuli of wavelengths A, and A, whose pro-
tanopic chromaticities are different, and by
stipulating that the color-matching functions shall
have the values

CIE 1931
PUA), pi MeO Chromaticity diagram

Pi(Ar) = 0, — po(A) = 1
respectively, at these wavelengths. The six trans-
formation coefficients {p;;} for protanopic vision

Fig. 1(5.14.2). Lines of constant dichromatic

chromaticity drawn in the normal trichromat’s
chromaticity diagram (represented here by the CIE
1931 (x, y)-chromaticity diagram), and showing
convergence to (a) the protanope’s, (b) the de-
uteranope’s, and (c) the tritanope’s, confusion
points (P),(D),(T), respectively (from Judd and
Wyszecki, 1975).
 Color-Matching Properties of Color-Defective Observers 467

then satisfy the simultaneous equations In the exceptional case that the determinant
D,, is zero, the plane Ol corresponding to a partic-
US X(A)) Pu + y(Ai) Pi2 + Z(A1) P13
ular—and unique—value of p or o (that is, of
protanopic chromaticity) is parallel to the plane
0 = X(Az) Pi + ¥(A2)
Pin + Z(A2) Dis 1+ X+ Y + Z, and for all other values of p and
o, the two planes intersect in lines m that are
Wiss XpcPi1 S YieP12 aa LocP13 parallel and in a direction specifiable by the value
of dY/dX = a (or of dX /dY = B if a approaches
0 = x(A,) po + y(A,) Po» < Z(A;) p23 infinity). Thus, in this case, the confusion point is
a point at infinity whose coordinates may be
Ree X(A2) pr fe V(Ax) Px» + Z(A2) pos represented by |

OX, Par + WpcPrue i Pa Ae

[8(5.14.2)]
Yo. ak

which have a unique solution, provided the Z,c=1—(1t+a)Q

determinant of coefficients—which equals A’, with

where {2 is to be made indefinitely large. When

X(A;) VC) 2(AY) these values are inserted in the equations for the
ayhes X(A,) y(Az) Z(A2) coefficients {p;;} and & is made to tend to infin-
Xpe Yc ZLpc ity in the solutions, the expressions for p;; become

—is not zero. However, A p cannot be zero be- ae (1 + a)y(A2) + .a2z(A2)

cause A, and A, have different protanopic chro- 11 A’,

maticities, and the line joining their representative

points in the chromaticity diagram cannot pass
through the point (X,., Y,., Z,-). The solutions _ (+a)
they ime
x(a) +a7(A1) A’,
are readily obtained as

¥(A2)Z,- ad Z(AS YS) Py

_ 2(Az) +(1 + a) Xz)
A’,
Pie A
P

ar ¥(A,)Z,. sy Z(A,)¥%- _ Cr
= 20a) + 0A’,+ a)x(Ar)
nd(4 Tees A
P

Z(A2)X,. oa X(A,)Z,, ide ax(A,) — y(Az)

Pia A PBs A’
p Pp

ait Z(A\) Xp a x(A\)Z,- sa _ ax(A,) — yy)

jwLiebe Aa A |Le OG A’
Pp

f [10(5.14.2)]
a x(A2)Y,. — (Az) X,-
PIGS aS Jo te Raut, Uiceeng Tt
Pp
where

x(A,)¥,. — (Ay) Xp
T2305 souk me Ae
P
fare ee x(A1) YA) — 7))
A, =|*(A2) ¥(Aa) z(X3)
[9(5.14.2)] l a = (1 ton
468 VISUAL EQUIVALENCE AND VISUAL MATCHING

Table 2(5.14.2) Transformations from CIE Trichromatic to Dichromatic Color-Matching

Functions

Protanopic (Normalizing wavelengths: 4, = 460 nm, 4, = 530 nm)

Pi 0.010597, Pi2 = —0.031290, P13 — 0.59840,

P21 — 0.41968, P22 = 1.23912, Pog = 0.028599,

Le }hy, = 01150.
Deuteranopic (Normalizing wavelengths: 4, = 460 nm, A, = 650 nm)
ay = 0, dys — 0, dy, = 0.59909,
do} = 0.57870, doo = (RADA dys = —0.38164,

Lg /La, = 0.6914.

Tritanopic (Normalizing wavelengths: 2, = 530 nm, 4, = 650 nm)

hy J —0.4701 De hie —= 1.2456, hig == 0.096974,

toy = 3.7874, Too mS —0.68891, fog — —0.78123,

Ly[Ly, = 6.713.

These simplify to With these values of the coefficients, the

color-matching functions of the dichromats have

11
_ (Az) + 202) the values computed in Table 3(5.14.2). These
color-matching functions have not been directly
“+P 4

measured, and the values of Table 3(5.14.2) as-

: _ (Ay) + Z(A2) sume:
oh 1 im ”
ve

(a) The functions for the representative dichro-

ey _ X(A2) Pie Eom _ x(Aj) mat of each class are reduction forms of the
12 e 13 = ”
A’ ’ 22 23am ”
A’
functions for the representative normal trichro-
mat.
X(A,) ¥(Ay) (A) (b) The latter functions are adequately specified
by the functions of the CIE 1931 standard ob-
An =|x(A2) y(A2) 7A) server.
0 l aa
(c) The confusion points of the dichromats have
the locations given above.
if dY/dZ = a approaches infinity, that is, if the
lines m are parallel to the plane X = constant.
The numerical values of the coefficients Further. the dichromat color-matching func-
{Piz}, {4i;}, {t;;} by which the color-matching tions of Table 3(5.14.2) are expressed with respect
functions of a representative protanope, de- to convenient real primary stimuli, chosen without
uteranope, and tritanope are related to the color- reference to theoretical considerations about nor-
matching functions of the CIE 1931 standard mal and defective color vision.
observer are given in Table 2(5.14.2). They as- The above assumptions (b) and (c) can be
sume for the trichromatic chromaticities of the modified to take account of new and improved
respective confusion points the values estimates of color-matching functions that are
representative of a normal trichromat, such as
Apart, Py 0253 those proposed by Judd (195la, see also Section
5.5.2) and measured by Stiles (1955b, see also
Xie = 1.080, — y,.= —0.080 Section 5.5.3), and by choosing different locations
for the confusion points, such as those proposed
X,.=90.171, — y,, = 0.000 by Walraven (1974). However, if one adheres to
assumption (a), that is, dichromatism is a reduc-
quoted in Table 1(5.14.2), and for the normalizing tion form of trichromatism, the computational
wavelengths A, and ),, (the effective dichromatic procedure outlined here would not change, only
primary stimuli), the pairs of suitably chosen val- the numerical data would be different (see also
ues stated. Section 8.2.5).
 Color-Matching Properties of Color-Defective Observers 469

Table 3(5.14.2) Color-Matching Functions of Dichromats

Protanope Deuteranope Tritanope

Wavelength
(nm) P(A) P(X) d,(X) d() 4,(Q) Z( A)
400 0.0408 —0.0036 0.0407 —0.0145 0.0004 0.00084
10 0.125 —0.0108 0.124 —0.0446 0.0012 0.00190
20 0.388 —0.0330 0.387 —0.137 0.0044 0.00190
30 0.832 —0.0652 0.830 —0.274 0.0153 —0.0152
40 1.048 —0.0678 1.047 —0.286 0.0343 —0.062
50 1.063 —0.0434 1.062 —0.185 0.0611 —0.137
60 1.000 0 1.000 0 0.100 —0.244
70 0.770 0.0676 0.771 0.333 0.146 —0.329
80 0.483 0.155 0.487 0.831 0.207 —0.369
90 0.272 0.258 0.279 1.466 0.289 —0.386

500 0.153 0.406 0.163 2.422 0.426 —0.416

10 0.079 0.624 0.095 3.875 0.638 —0.435
20 0.0252 0.856 0.0469 5.554 0.862 —0.310
30 0 1.000 0.0253 6.814 1.000 0
40 —0.0147 1.061 0.0122 GOA 1.054 0.427
50 —0.0214 1.051 0.0052 8.021 1.036 0.949
60 —0.0225 0.984 0.00233 8.116 0.960 1.563
70 —0.0205 0.860 0.00126 7.878 0.828 2.229
80 —0.0165 0.694 0.00102 TESPA! 0.653 2.870
90 —0.0122 0.507 0.00066 6.508 0.461 3.365

600 —0.0080 0.336 0.00048 5.544 0.287 3.588

10 —0.0050 0.202 0.00018 4.510 0.155 3.450
20 —0.00277 0.114 0.00012 3.471 0.073 2.973
30 —0.00150 0.059 0.00000 2.442 0.0281 2.250
40 —0.00074 0.0289 0.00000 1.626 0.0074 1.576
50 —0.00035 0.0136 0.00000 1.000 0 1.000
60 —0.00016 0.0064 0.00000 0.572 —0.00154 0.582
70 —0.00008 0.00297 0.00000 0.301 —0.00123 0.309
80 —0.00004 0.00142 0.00000 0.160 —0.00083 0.166
90 —0.00002 0.00063 0.00000 0.077 —0.00046 0.0803

700 —0.00001 0.00030 0.00000 0.0386 —0.00025 0.0404

A neutral point for a dichromat with respect to
a particular achromatic stimulus {P, dA} is a
particular wavelength which has the same chro-
maticity as the achromatic stimulus. Thus, for the
respective protanope, the neutral point A,, is de-
fined by
5 Py p(X) dd
nee = JPL EE [11(5.14.2)]
0%) fp,pyr) dd
neutral point only, but for tritanopes there may
be two intercepts of the specified straight line
with the spectrum locus, one in the greenish-
yellow, the other in the blue or violet, and thus,
two neutral points.
For an equal-energy (achromatic) stimulus, P,
= const, and from the locations of the confusion
points given above, the following neutral points
have been derived:
Protanope: Ayn — 494 nm

The neutral point A,, can be determined from the Deuteranope: Nan = 499 nm
known color-matching functions p,(A), p(A). In
the trichromatic chromaticity diagram, A,, is at Tritanope: Ain, = 970.0m
the intersection with the spectrum locus of a
A.,.a0) = 400 nm
straight line through the confusion point and the
point representing the chromaticity of {P, dA}.
For protanopes and deuteranopes there is one These derived neutral points agree well with the
 470 VISUAL EQUIVALENCE AND VISUAL MATCHING

Table 4(5.14.2) Luminous Efficiency Functions of Dichromats

V(X) Vq(A) V, (A)
are Experimental Computed Experimental Computed Experimental Computed

400 — 0.0011 — 0.0017 — 0.0004

410 0.020 0.0033 — 0.0051 0.0096 0.0012
420 0.029 0.0109 0.041 0.0160 0.0154 0.0039
430 0.041 0.0288 0.056 0.0370 0.0208 0.0110
440 0.055 0.0498 0.067 0.0540 0.0254 0.0211
450 0.073 0.0744 0.080 0.0677 0.0276 0.0340
460 0.108 0.1086 0.096 0.0852 0.0374 0.0532
470 0.154 0.1474 0.119 0.1067 0.0670 0.0816
480 022.16 0.1991 0.156 0.1439 0.0951 0.1274
490 0.296 ce a p12 0.2043 0.1256 0.1941
500 0.444 0.3999 0.339 0.3123 2228 0.3053
510 0.668 0.5976 let 0.4854 0.4027 0.4799
520 0.871 0.8104 0.713 0.6881 0.6577 0.6837
530 0.971 0.9440 D850 0.8416 0.8337 0.8380
540 1.000 1.0000 O90 0.9389 0.9419 0.9364
550 0.965 0.9903 Doi2 0.9885 0.9908 0.9872
560 0.883 TAS Balaie 999 1.0000 0.9908 1.0000
570 0.748 0.8096 0.991 0.9706 0.9594 Oo ee
580 0.600 0.6530 0.948 0.9026 0.8913 0.9058
590 0.444 0.4777 0.873 0.8017 0.8035 0.8064
600 0.288 0.3165 0.770 0.6830 0.6966 0.6885
610 0.183 0.1907 0.633 0.5556 0.5715 0.5612
620 0.101 0.1069 0.480 0.4276 0.4150 0.4326
630 0.057 0.0553 0.338 0.3008 Spies 0.3047
640 0.025 UR ee Ce Ne 0.2004 i Sat 0.2031
650 0.013 0.0128 0.140 O23. 0.0910 Ost 2ou
660 0.004 0.0060 0.075 0.0705 0.0519 0.0715
670 — 0.0028 0.038 0.0370 OT. 0.0376
680 — 0.0013 0.014 0.0197 0.0140 0.0200
690 — 0.0006 0.009 0.0095 — 0.0097
700 — 0.0003 _- 0.0048 — 0.0048

spectral ranges that have been found experimen- Vi(A) = Lyt(A) + Lyt)(A)
tally by some investigators as quoted in Table
1(5.14.2). | (tritanopic vision)
Direct measurements made by Pitt (1935) and
Wright (1952) of the luminous efficiency func- [12(5.14.2)]
tions V,(A), V,(A), V(A) of dichromats yield the
experimental values of Table 4(5.14.2) and plots
of the logarithm of the mean against A are shown where (Lj, Lp2), (Lars La), (Ln, Liz) are ap-
propriate pairs of constant luminous efficiency |
in Figure 2(5.14.2). As for trichromatic vision, the
factors. If, for a given form of dichromatic vision, |
luminous efficiency function is assumed to be a
the color-matching functions are known, and also
linear combination of the color-matching func-
tions, so that
two stimuli, {P, dA} and {Px dA} which the pro-
tanope matches in luminance but not in chro- |
0) caebeee hn) al maticity, then the relative values of the luminous
pA) = Lp Pil) + Lyn PrlA) efficiency factors and the whole luminous ef-
(protanopic vision) ficiency function are determined. Thus, for pro-
a z tanopic vision, if
Va(A) = Lada) + Lan da(A)
(deuteranopic vision) /PyV,(A) dd = i}PV,(A) da
 Color-Matching Properties of Color-Defective Observers 471

the ratio L,,/L,, and the complete V,(A)-func-

tion can be computed.
“ TRITANOPE
A similar procedure is applicable for de-
uteranopic and tritanopic vision, and the derived
luminous efficiency ratios and luminous effi-
ciency functions for the three types, are given in
Tables 2(5.14.2) and 4(5.14.2). The experimental
™“DEUTERANOPE
and derived luminous efficiency functions are
compared in the logarithmic plot of Figure
3(5.14.2). The considerable deviations that appear
at wavelengths below 460 nm are probably related
to the excessively low values at short wavelengths
“S PROTANOPE of the CIE standard V(A)-function for normals
(see below) on which the CIE 1931 color-match-
ing functions [x(A), y(A), Z(A)] (Section 5.5.1)
and hence, the derived dichromatic functions
EFFICIENCY)
(LUMINOUS
LOG
[p (A), p(A), etc.], depend.
If it is not stipulated that the color-matching
functions of the representative dichromat of a
given type shall be exactly linear combinations of
the color-matching functions of the CIE 1931
400 500 600 700 standard observer, the dichromat’s functions can
WAVELENGTH A (nm) be derived from the experimental luminous ef-
ficiency function V,(A),Vj(A),V,(A) and from
Fig. 2(5.14.2). Logarithm of the mean luminous
the dichromat’s chromaticity coordinates in terms
efficiency curve for protanopes (Pitt, 1935), de-
of two primary stimuli units whose (dichromatic)
uteranopes (Pitt, 1935), and tritanopes (Wright,
luminous efficiency is known. The method is pre-
1952). The deuteranope and protanope curves have
cisely similar for two dimensions to that used
been displaced downwards by one and two log
originally in deriving the CIE 1931 color-match-
units, respectively.
ing functions for trichromatic vision (see Section
3.3.3). It has the advantage that no use is made of
then trichromatic data. Wright (1946) applied this
method to obtain the pairs of dichromatic color-
fam POR) ar matching functions plotted in Figures 4(5.14.2),
5(5.14.2), and 6(5.14.2).
[(Px— PRA) aa
5.14.3 Monochromats
[13(5.14.2)]
The principal category of the very small per-
In particular, if the two luminance-matched centage of observers devoid of color discrimina-
stimuli are monochromatic of wavelengths A = a tion is the group of rod-monochromats. These have
and A = b and radiant powers P,, and P,,, respec- generally very poor acuity in foveal vision, are
tively, the equation becomes commonly subject to photophobia and nystag-
mus, and are unable to make satisfactory observa-
Ly inte 6 P,P2(a) — P,pr(b) tions on small matching fields under onditicons
Ly2 P,p\(a) — P,p\(b) of foveal fixation. Under relaxed conditions,
matches can be made, and all color stimuli can be
[14(5.14.2)] matched by varying the luminance of a compari-
son stimulus of fixed spectral composition. The
If the experimental values of V,(A) at A=a, spectral luminous efficiency curve—the sole spec-
A = b are used to derive the relative values of P, tral response curve provided by the “color
and P,, that is, matches” of such observers—is maximal in the
neighborhood of 510 nm and otherwise resembles
the scotopic luminous efficiency curve of normals.
The rod-monochromats are believed to be
472 VISUAL EQUIVALENCE AND VISUAL MATCHING

0.6

0.4

DEUTERANOPE (b)
.
X< x x”
xe fas 0 Sime
x
Sa
x Xe xy

Xu ee x
N
Xsy x

TRITANOPE

EFFICIENCY)
LUMINOUS

COMPUTED
(RATIO
TO
MEASURED
OF
LOG

400 500 600 700

WAVELENGTH X (nm)
Fig. 3(5.14.2). Logarithm of the ratio of the experimental values to the computed values of
luminous efficiency [see Table 4(5.14.2)], plotted against wavelength. The curves show the
measure of agreement. However, the comparison is complicated by two facts: (i) the computed
values are derived from color-matching functions that rest on the CIE standard V(A)-function
which is believed to be considerably too low in the blue below 460 nm; (ii) given that the
agreement at longer wavelengths is not perfect, the nature of the discrepancies must depend on
the choice of the two wavelengths a and b (see text) at which the ratio of observed to computed
luminous efficiencies is made to have the value unity.

equipped with a single retinal response system,
the scotopic or rod system. In trichromats, as well
as dichromats, the rod system is absent from the
central foveal area of the retina, which is popu-
lated exclusively with cone receptors. The defects
of the rod-monochromat’s vision, apart from a
lack of color discrimination, are correlated with
the absence, particularly at the fovea, of cone
receptors suited to the mediation of vision at
daylight levels.
A still rarer type of color defective known as a
cone-monochromat is free of the visual defects
associated with complete absence of cone recep-
tors but is nevertheless able to match in foveal
vision all color stimuli by varying the luminance
of a fixed comparison stimulus. The single spec-
tral response curve may resemble most closely the
protanopic luminous efficiency function and the
“green” Konig or Fick fundamental (Section
8.2.5). For the results of a detailed investigation
of three cases of this type, see Weale (1953b).
More recently, cases have been described by
Blackwell and Blackwell (1961) in which, apart
from the rods, the retinal response is confined to
a cone system with maximum at about 440 nm
and approximating the K6énig’ or Fick “blue”
fundamental. The accommodation of cone-mono-
chromats in a single theoretical scheme with di-
chromats and trichromats is controversial.
The review by Hurvich (1972) is recommended
for a more comprehensive discussion of mono-
chromats and further literature references.
5.15 INSTRUMENTATION FOR COLOR-VISION
RESEARCH
There is a considerable variety of instrumentation
that has been developed by instrument makers
and individual workers for use in color-vision
 PROTANOPE

0.75
[Loe Bald)]

0.50

0.25

LUMINANCE
IN
UNITS
COLOR-MATCHING
FUNCTIONS
0.00

400 450 500 550 600 650 700

WAVELENGTH ) (nm )

Fig. 4(5.14.2). Protanope curves of [L,,;p,(A)] <5 and [L,.p2(A)] against A, as derived by
Wright (1946) from the average spectral chromaticities, the average luminous efficiency curve,
and the average luminous efficiencies of the primaries used in the chromaticity measurements.
A constant factor applied to both curves brings the maximum value L,,, p,(A) to unity. The
derivation of the curves is based on experimental data of Pitt (1935), and is independent of the
CIE standard V(A) luminous efficiency function.

WAVELENGTH (nm)
UNITS
LUMINANCE
IN
FUNCTIONS
COLOR-MATCHING
Fig. 5(5.14.2). Deuteranope curves of [L n4(A)] <5 and [L 224,(A)] against A as derived by
Wright (1946). See caption to Figure 4(5.14.2).

473
474 VISUAL EQUIVALENCE AND VISUAL MATCHING

100

TRITANOPE

80

ME
60

A
40

\
20

FUNCTIONSO
COLOR-—MATCHING
LUMINANCE
IN
FUNCTION
EFFICIENCY
LUMINOUS
AND 400
UNITS 500 60 700
WAVELENGTH A (nm)

Fig. 6(5.14.2). Tritanope curves of [L,,2,(A)] and [L,.7,(A)] against A, as derived by Wright
(1946), based on his own measurements using the method indicated in the caption to Figure
4(5.14.2). The V,(A)-curve corresponds to the experimental values in Table 4(5.14.2). Note:
The shapes of the color-matching functions are not, in this case, directly comparable to those of
the tritanopic color-matching functions tabulated in Table 3(5.14.2), because the normalizing
wavelengths are different: 480 and 650 nm for the present figure and 530 and 650 nm for the
table.

research. Most of the instruments are highly spe-
cialized to serve one or a limited number of
purposes. Researchers have demonstrated a great
deal of ingenuity in putting together the instru-
ments they need to carry out certain visual experi-
ments, often under heavy budgetary constraints.
Apart from auxiliary electronic equipment,
now frequently used to provide “computer-con-
trolled operation” of an instrument, automatic
“data acquisition,” and “data processing,” virtu-
ally all instruments of moderate to high sophisti-
cation include several of the following “optical”
components:
(a) Light Sources (tungsten-filament lamps,
high-pressure xenon arcs, and others; Section 1.2)
(b) Optical Filters (absorption filters, inter-
ference filters, beamsplitters, polarizers, and
others; Section 1.3)
(c) Reflecting Materials (mirrors, white diffusers,
and others; Section 1.4)
(d) Monochromators (prism and/or grating dis-
persive elements; Section 1.5)
(e) Physical Detectors (thermopiles, photodiodes,
photomultipliers, and others; Section 1.6)
Our Chapter | on physical data contains a collec-
tion of data on the above optical components
together with other relevant information consid-
ered of importance when an appropriate compo-
nent must be selected for a specific application.
The same chapter also discusses the basic radio-
metric quantities and their units that are of im-
portance when the visual stimulus generated by a
particular instrument is to be calibrated with re-
gard to its spectral composition and radiant
power.
Photometric calibrations of such quantities as
luminance and illuminance are treated in Chapter
4 on photometry, and the special case of retinal
illuminance is discussed in Section 2.2.4 of
Chapter 2 entitiled The Eye.
Colorimetric calibrations of the visual stimulus
in accordance with the CIE colorimetric system
are fully described in Chapter 3 on colorimetry,
particularly Sections 3.3, 3.4, 3.11, and 3.12. Sec-
 Instrumentation for Color-Vision Research 475

tion 3.12 deals with Colorimetric Instrumentation
as it is now practiced, including spectroradiome-
ters and spectrophotometers used to calibrate light
sources and reflecting or transmitting materials,
respectively.
From the large number of different instru-
ments that have been used or are being used by
different researchers in color vision, we have
selected a few to illustrate certain features of their
design. Useful additional information on instru-
mentation can be found, for example, in the
booklet entitled Techniques of Photostimulation in
Biology in which Crawford and Granger (1968)
give “some worked examples” and Ruddock
(1968) discusses the “control of the time pattern
of the stimulus.” Although our own selection is a
very limited one indeed, we hope that it provides
at least a general impression of the kind of opti-
cal-mechanical hardware that is often employed
in such instrumentation. However, no attempt has
been made to deal with the auxiliary electronic
equipment often used in conjunction with a visual
instrument for the purpose of automating stimu-
lus presentation, data gathering, and processing.

Source

Side View L,

Fig. 1(5.15). Schematic diagram of the Wright colorimeter. This instrument incorporates a
double monochromator whose two pairs of dispersing prisms (!/, — M,; M, — M,) are mounted
on top of each other. Two small totally reflecting prisms (7 and D), placed in the spectrum S,
formed by the upper dispersing system (7, — M,) serve as two slits for the lower dispersing
system (M, — M,). After reflection in the prism P, the two monochromatic components are
brought to a common focus in the pupil of the eye of the observer (Obs.). The prism R, splits
the beam from the upper monochromator so as to form a second spectrum at S,. Three small
totally reflecting prisms, R,G, B are positioned at S,. The three monochromatic components
are reflected by these prisms and redirected by the prism R, through the lower dispersing
system (M, — M,) and the prisms P and Q, and brought to a common focus in the pupil of the
eye of the observer (Obs.). The mixtures of the two sets of monochromatic components isolated
from the spectra S, and S, fill adjacent areas of the field prism Q, forming the field of view.
The monochromatic lights reflected from the prisms R,G, B serve as primary stimuli in a
color-matching project and may be fixed at wavelengths 650, 530, and 460 nm, respectively.
Calibrated neutral-density wedges in front of each of these prisms control the radiances of the
primary stimuli. The prism 7 at S, reflects the monochromatic test stimulus which can be
varied along S,. The prism D is placed at S, so as to reflect that primary stimulus which may be
necessary to desaturate the test stimulus before it can be matched by a mixture of the other two
primary stimuli (from Wright, 1927-1928).
476 VISUAL EQUIVALENCE AND VISUAL MATCHING

Second stage First stage

6
(= ay Upper tier

Exit slit
vy Re i Entrance slit
i Field lenses [i
3
[4 Source

Middle tier
|
| cA
: .: )‘ Entrance slit
Photometer IVAN
cube ~ ~
Source

ptt og Lower tier

See Lupe Cyne slit

MeN Source

Fig. 2(5.15). Schematic diagram of the Stiles trichromator. Three double monochromators of
van Cittert type with canceling (subtractive) dispersion are mounted vertically, one above the
other. In the central spectrum of the middle tier, a single movable narrow slit selects the
monochromatic test stimulus, which fills one half of the test field provided in the photometer
cube in front of the eye of the observer. Three fixed slits in the upper tier select the
monochromatic primary stimuli. These, after recombination in the second stage of the
monochromator, provide a mixture which fills the comparison field in the photometer cube. The
radiances of the primary stimuli are independently variable by neutral-density wedges inserted
immediately after the slits in the central spectrum. The lower tier provides a similar mixture, in
variable amounts, of three primary stimuli of the same wavelengths as those in the upper tier.
These primary stimuli can be added to the test stimulus to desaturate it. The exit slits of all
three tiers are imaged at a common point in space; the eye of the observer is placed so that the
pupil center coincides with this point, and the whole of the light in the slit images, all of which
lie within a square area of about two millimeters size, is collected by the pupil. The observer
sees the bipartite visual field by the method of Maxwellian view. Different photometer cubes
can be provided to give different field sizes and fields of different shape (from Stiles, 1955).

(i) Wright Colorimeter. Figure 1(5.15) il-
lustrates the basic design of the Wright (1927-
1928) colorimeter that has been employed exten-
sively at the Imperial College of London for
many years. Its first major use was in Wright’s
(1928-1929) color-matching experiment which,
together with Guild’s (1931) data, formed the
basis for the CIE 1931 standard observer data for
small visual fields (Sections 3.3.3 and 5.5.1).
Several Wright colorimeters have been built in
other laboratories for a variety of different color-
vision experiments.
(ii) Stiles Trichromator. Figure 2(5.15) shows
the basic optical arrangement of the Stiles (1955)
trichromator as it was designed initially at the
National Physical Laboratory at Teddington, En-
gland, to determine color-matching functions for
a large group of normal trichromats viewing
centrally a large 10° bipartite field. These data
formed the major part of the input data that were
used to develop the CIE 1964 supplementary
standard observer for large-field viewing (Sections
3.3.3 and 5.5.4). Instruments of the same basic
design are in use at the National Research Coun-
 Source
(ws) Front view of
filter frame
/ | \__ Six primary een orcs
v y filters
RRS OTL Willi tth SOOSESE SERRE aaa ean Cee CE ETS:

Shutters \ | \\> Shutters

|) _controlled |[ee Condenser
Source from front of lens
instrument \, J
\ /

] Integrating sphere \ \ 1 7A
7 where mixing of ox va
primary lights
takes place

Exit
(prism)

Photometric field as
seen by observer
m~

+2) Photometer cube

Six duplicate
Specimen primary filters Test Comparision
on 7-hole disk field (mixture) field

J Eyepiece

Fig. 3(5.15). Schematic diagram of Donaldson six-primary colorimeter. The six primary
stimuli (red, orange, yellow-green, green, blue-green, blue) are provided by incandescent-lamp
light combined separately with each of six colored filters. The additive mixture of the primary
stimuli taking place in the integrating sphere, fills one half of the photometric field seen by the
observer. The other half (test field) is filled by light coming from a second source and the
specimen mounted outside the main instrument. The instrument can be used as an ordinary
tristimulus colorimeter by employing three of the six primary filters only; that is, red, green,
and blue, or orange, yellow-green, and blue-green. In practical color-matching work it is,
however, advantageous to use all six primaries and in this way make an attempt to reduce the
metamerism between the matched color stimuli of the test and comparison field. Color
matching with six primaries is carried out with the aid of six duplicate primary filters mounted
near the eyepiece in a seven-hole disk. The observer looks through one of six duplicate filters
and adjusts the corresponding primary stimulus in the comparison field to an approximate
brightness match with the test field. This is repeated for the remaining five duplicate filters.
Then the observer looks at the photometric field without a duplicate primary in the beam. Since
test field and comparison field have been adjusted to approximate matches in brightness for
each of six parts of the spectrum separately, they will be found to be an approximate color
match. Final adjustments to obtain a perfect match are made by using three primary stimuli
only. The Donaldson colorimeter can be calibrated in terms of the 1931 or 1964 CIE standard
observer and coordinate system (from Judd and Wyszecki, 1975).

477
478 VISUAL EQUIVALENCE AND VISUAL MATCHING

cil (NRC) of Canada (Wyszecki, 1978; Section

5.6.5) and at the Electrotechnical Laboratory
(ETL) of Japan (Katori and Fuwa, 1979; Section
nape)?
(iii) Donaldson Colorimeter. Figure 3(5.15) il-
lustrates the design of the Donaldson six-primary
colorimeter. This instrument was initially de-
signed by Donaldson (1935, 1947) for color-
matching tasks in industry. However, its merits
were primarily exploited in research laboratories.
One such application was to field trials on the 2°
and 10° CIE color-matching functions (Stiles and
Wyszecki, 1962a; Wyszecki and Nayatani, 1964;
Wyszecki, 1964). The instrument proved to be
difficult to calibrate and provided only low to
moderate field luminances. It is not being used
much anymore, but some of its basic features
have been retained in other instrument designs,
notably those described below as examples 5.15
(iv) and 5.15(v).
(iv) MacAdam Binocular Colorimeter. Figure
4(5.15) shows the design of the MacAdam (1950b)
binocular colorimeter. Its main feature is that it
provides a large bipartite field that can be viewed
with both eyes simultaneously. It has been used in
a variety of color-vision experiments (MacAdam,
1950b; Brown, 1951; Breneman, 1958; Wright
and Wyszecki, 1960; Sanders and Wyszecki, 1964;
Cruz, 1965), related to color matching and hetero-
Fig. 4(5.15). Schematic diagram of the
chromatic brightness matching.
MacAdam binocular colorimeter. The various
(v) Wyszecki Seven-Field Colorimeter. Figure components are identified as follows: (1) source;
5(5.15) illustrates the design of the Wyszecki (2) and (5) condensing lenses (plastic Fresnel
(1965) colorimeter with an array of seven visual lenses); (3) controllable shutters; (4) primary filter;
fields viewed by the observer with both eyes. The (6) integrating sphere; (7) and (8) baffles; (9) field
color stimulus presented in each field is generated lens (Fresnel type). The large bipartite visual field
by a controllable mixture of three primary stimuli. provided by the instrument can be viewed with
The instrument has been used in color-difference both eyes simultaneously. The instrument consists
matching experiments (Wyszecki, 1965; 1972; of two symmetrical parts, each of which can be
Wyszecki and Fielder, 1971; Section 7.10.6), in a used to provide a spectrally identical match with
redetermination of color-matching ellipses the color stimulus provided by the other part (from
(Wyszecki and Fielder, 1971; Section 5.4.3) in Judd and Wyszecki, 1975).
hue-matching experiments (Robertson, 1970; Sec-
tion 5.9), as well as other studies. The availability
of more than two fields was an important prere- color-vision experiments (Burnham et al., 1952;
quisite for most of these studies. 1957; Bartleson, 1977), dealing mostly with
aspects of chromatic adaptation (Section 5.12.2).
(vi) Burnham Colorimeter. Figure 6(5.15)
shows the essential components of the Burnham
5.15.1 Maxwellian View
(1952) colorimeter that offers a relatively simple
way of generating color stimuli by means of addi- In many instruments, the visual field is presented
tive mixtures of primary stimuli made up of col- to the observer in Maxwellian view, an imaging
ored filters and a light source. Such a device, or technique first introduced by Maxwell (1860a, b).
variations of it, has been used in a number of Detailed accounts of Maxwellian view have been
 Instrumentation for Color-Vision Research 479

Sd

Obs. S
c)

Fig. 5(5.15). Schematic diagram of Wyszecki seven-field colorimeter. Only one of seven
similar channels, the one providing field No. 5, is shown in top view. Obs: observer’s two eyes;
Sd: surround, illuminated by two projectors (not shown) placed to the left and right of the
observer; FL: hexagonal field lens No. 5 imaging exit aperture of integrating sphere into plane
containing observer’s eyes; PM: plane mirror; Sp: integrating sphere; Sh: shutter; CL: con-
denser lenses; R,G, B: red, green, blue primary filters with controllable shutters; WC: heat-
absorbing water cell; S: source (1000 W); M: mirror. The insert at the top shows the field
configuration the observer sees when the surround exposes all seven fields (each about 3°
angular subtense).

given by LeGrand (1957), Walsh (1958), and By assuming that:

Westheimer (1966). The simplest optical arrange-
ment is illustrated in Figure 1(5.15.1). A small
(a) the eye is perfectly transparent (see, how-
source, that might be represented by the ribbon
ever, Section 2.4.6),
filament of a tungsten lamp or the exit slit of a
(b) the Stiles—Crawford effect is negligible (Sec-
monochromator, is imaged by a biconvex lens
tion 5.11),
into the pupil of the observer’s eye. The eye
(c) the eye’s nodal points are coincident and
focused on the lens, or an aperture placed next to
located in the center of the pupil (Section 2.4.1),
it, sees the lens uniformly filled with light. The
(d) the eye focuses at the Maxwellian lens, and
principal advantage of the Maxwellian-view tech-
(e) all relevant optical rays originating at the
nique is that it yields a retinal illuminance which
source and entering the eye’s pupil have angles «
can be as high as that for ordinary direct-viewing
with the optical axis such that cose = 1 is an
of the source, but with a larger area of the retina
adequate approximation (Gaussian approxima-
being illuminated. A disadvantage of the Maxwel-
tion),
lian-view technique arises when the source is ex-
tremely small (w, — 0) and the light from each
part of the Maxwellian lens passes through a very the basic photometric relation, developed in
small area within the pupil’s eye. The field of Chapter 4 on photometry and given by Eq.
view then becomes marred by the existing imper- 13(4.3.6), can be applied readily to the derivation
fections in the pupil, such as variations in trans- of a useful formula for the retinal illuminance
mission which cause patchiness, and specks or obtained by the Maxwellian-view technique il-
other opaque particles which give shadows mov- lustrated in Figure 1(5.15.1).
ing apparently across the field of view (Michel- In accordance with Eq. 13(4.3.6), the luminous
son, 1924). flux P that is passing through the lens and thus,
480 VISUAL EQUIVALENCE AND VISUAL MATCHING

Glass integrating Primary

-filter

= ©
bar ve frame

\ ys
Source

Observer — - —— -——--§

Opal glass
diffuser
Aperture plate

Front view of
primary-filter frame

Aperture plate

Fig. 6(5.15). Schematic diagram of the Burnham colorimeter. Light from an incandescent
lamp passes through a frame holding the primary filters (red, green, and blue), an aperture plate
with an opening one quarter of the filter area, and an integrating bar whose front and back
faces are covered with opal-glass plates. The back face of the bar appears uniformly filled with
the mixture of the primary stimuli. The mixture can be controlled by moving the filter frame in
both horizontal and vertical directions in front of the fixed aperture plate. A light attenuator
may be placed between the source and the filter frame to control the luminance of the visual
field. Other filter arrangements than that shown in the illustration may be used that incorporate
four or more primary filters. A second similar colorimeter assembly is usually used to provide
another visual field for color-matching purposes or for studying other colorimetric problems.
The two fields are brought into juxtaposition so as to provide a bipartite field of view.

through the pupil, is given by where Ap (m’) = area of illuminated part of ret-
ina R
P=LAsw, (Im) [1(5.15.1)] Inserting Eq. 1(5.15.1) in Eq. 2(5.15.1) we obtain

where L (cd-m~”)=luminance of the source S as LA sW>

A, (m’)=area of the source S Er 7p [3(5.15.1)]
W, (sr)=solid angle subtended by the
area Av of the lens Q as seen A more useful expression is obtained by introduc-
from the center of the source ing the following quantities that can be seen in
S Figure 1(5.15.1):

The illuminance Ep at the retina R must then be, apes

assay} Ae
Ao Ap
[4(5.15.1)]
by definition [see Eq. 2(4.3.5)],

(im atl oo ee [5(5.15.1)]

 Instrumentation for Color-Vision Research 481

Maxwellian View

Lens a Aperture

i
1
1
i}

Image S of Source S in Plane p

Cae nr "White" Standard
u ae u Baca D

Fig. 1(5.15.1). Schematic diagram of the essential components and parameters involved in a
Maxwellian-view arrangement. The lower-right part of the figure shows the arrangement used
to determine the retinal illuminance or the troland value when the area A ,, of the source image
is smaller than the area A, of the eye’s pupil. For details of procedure see text.

7
—_— = eae COED) 6(5.15.1
In the above calculations, it has been assumed
that the source image S’ is not larger than the
pupil area A, that is
where w, (st)=solid angle subtended by the
area Av of the lens Q, or As <A,
the area Ap illuminated at
the retina R, as seen from If this condition is not met, that is, if
the center of the pupil
Ay >A,
wy (sr)=solid angle subtended by the
area Ao of the lens Q as seen
the smaller pupil reduces the illuminance Ep at
from the center of the source
the retina by the factor A,/A,,, and we have
S
A, (m*)=area of the source S$
A,, (m’)=area of the image S’ of S ijBs ae rd
S r? Ay
u (m)=distance of the source S from
the lens Q
and thus
u’ (m)=distance of the image S’ from
the lens Q
r (m}distance of the retina from
the pupil
n= 3° (im-m~)
pert (1 AAPA [8(5.15.1)]
GE SM

Inserting Eqs. 4(5.15.1) to 6(5.15.1) in Eq. 3(5.15.1)
we find for the illuminance of the retina:
oe ast (Im-m~?) [7(5.15.1)]
where oS = Wp (sr) = solid angle subtended by
4 the area A,, of the source
image S’ as seen from the
retina R
This expression is equivalent to the one given as
Eq. 3(2.4) for the spectral concentration of the
retinal irradiance E,, dd if one considers negligi-
ble the radiant flux lost from the imaging beam
when it passes through the eye on its way to the
retina. In the earlier expression, the quantity m
corresponds to r? in Eg. 8(5.15.1) and, as shown
in Section 2.4.1 entitled Image Formation by the
Theoretical Eye, has the value of 267.33 for the
eye accommodated on a point at infinity, and the
 482 VISUAL EQUIVALENCE AND VISUAL MATCHING
value of 218.34 for the eye accommodated on a
point distant 143.6 mm (all distances in the eye
are measured in millimeters).
The generally accepted unit of retinal illumi-
nance is the troland (td), as discussed in Section
2.4.4, which is defined as the retinal illuminance
when the surface of a source of luminance
cd - m 7 is viewed through a pupil of area (A, or
A,) 1 mm’. Thus, the troland value is given by
T=L-Ay (td) fords <A, (mm’)
and
T=L-A, (td) fords >A, (mm’)
[9(5.15.1)]
when L refers to photopic luminance (cd - m ”),
and by
T’=L'-Ay (td) fordy <A, (mm?)
and
Wel’ ged Peenth)
ford, >A, (mm°)
[10(5.15.1)]
when L’ refers to scotopic luminance (cd - m°).
The measurement of retinal illuminance or
troland value follows directly from the above
considerations. With the entrance pupil of the eye
(natural or artificial) completely covered by the
image of the source (A, > A,), the retinal illumi-
nance Ep (or the troland value T) is the same as
if the eye with that particular pupil (A,) were
looking directly at the source, though allowance
must be made for luminous flux lost from the
imaging beam when it passes through the external
optical system (Maxwellian lens and possibly
other optical components) between the source
and the eye. What are required, in this case, are
measurements of the luminance L (cd - m 7) of
the source (for details, see Section 4.3.6), and of
the pupil area 4, (in mm”). Although there is no
problem in measuring the size of an artificial
pupil, the measurement of the size of the natural
pupil requires special techniques, such as high-
speed flash photography; for a discussion of a
number of techniques, see, for example, LeGrand
(1957).
When the image (A,,) of the source is smaller
than the pupil, an indirect method of measuring
the retinal illuminance or the troland value may
be used (Westheimer, 1966).
At an appropriate distance d (m) beyond the
source image S’, a “white” reflectance standard
D is placed [see Figure 1(5.15.1)] with reflecting
properties closely approximating those of the per-
1 fect reflecting diffuser (Sections 1.4.2 and 3.3.6).
Its radiance factor B is obtained by means of a
separate calibration; good “white’’ standards have
a radiance factor B ~ 0.98. From a measurement
of the luminance Lp, made by means of an
established technique (Section 4.3.6), of the
“white” standard, the retinal illuminance Ep or
the troland value T can then be calculated readily.
In accordance with Eq. 12(4.3.6), we have for
the luminance L., of the source image S’
Ly = — (cdo powst a)
D
[11(5.15.1)]
where E, (Im-m *)= illuminance at surface of
“white” standard D
placed at distance d (m)
from S’, normal to the
optical axis
Wp = ae (sr) = solid angle subtended by
d the area A,, of the source
image S’ as seen from the
center of the “white”
standard D
The measured luminance Ly of the “white”
standard D is related to the illuminance E, by the
relation given as Eq. 34(4.3.6), that is,
E
Eo pies [12(5.15.1)]
where we assume that our “white” standard has
the diffusing properties of a “Lambert surface.”
From Eqs. 11(5.15.1) and 12(5.15.1) we obtain
jp Bolo [13(5.15.1)]
Recalling that the product L.A, gives the
luminous intensity [,, of the source image S’ [see
Eq. 11(4.3.6)], we have
vie
ly = Lele 34 Lp [14(5.15.1)]
The illuminance Ep in the plane R (retina) is then
 Instrumentation for Color-Vision Research 483

given by bright patch of light which forms the central field

i nal Get sls onset
z ‘a r? Br? a aperture in the diaphragm 7, is partially reflected
[15(5.15.1)]
and the troland value T is obtained by the con-
vention that A, is measured in mm’ while other
distances are measured in m and the luminance
Lp in cd - m ’, thus
Tae 10° 1, (td) [16(5.15.1)]
5.15.2 Measurement of Directional
Sensitivity and Increment Thresholds
Many studies aim at the determination of the
spectral sensitivities of the visual mechanisms,
particularly the receptor mechanisms. In these
studies, the increment threshold technique is often
used (Stiles, 1978; Section 7.4) which requires an
optical arrangement not normally employed in
color-matching experiments. The essential part of
an instrument used for increment threshold mea-
surements is illustrated in Figure 1(5.15.2). The
same arrangement is also useful for the de-
termination of the directional sensitivity of the
retinal receptors (Section 5.11). The optical con-
figuration shown in Figure 1(5.15.2) is that used
by Stiles (1939).
A collimated beam of light provided by mono-
chromator M, is partially transmitted by the
beam-splitter cube C and is brought to a focus F,
in the plane of the observer’s pupil at 0 by lens L.
This beam is seen by the observer as a uniformly
of the conditioning stimulus. A collimated beam
from monochromator M, is delimited by a square
at the diagonal surface in the cube C, and is
finally brought to a focus F, in the plane of the
observer’s pupil by the lens L. The diaphragm T
is so placed that the observer sees a virtual image
of the square aperture at infinity. This image
forms the test stimulus.
It is important that in the area of the field of
view occupied by the test stimulus, the eye re-
ceives radiant flux from both the test stimulus and
the conditioning stimulus. Thus, in this sense, the
test stimulus is added to or superimposed on the
conditioning stimulus.
The complete visual field, as it is seen by the
observer, is shown as an insert in Figure 1(5.15.2).
The orientation points in the visual field are
introduced from the side by reflection in a thin
plain glass plate P inserted in the collimated
beam from monochromator M,. The rays forming
a particular orientation point are reflected by P to
give a real image in the plane K which is the focal
plane of the lens L. This real image, when seen
through the lens L, appears to the observer as a
virtual image at infinity.
By slight rotations of the cube C, the beam
providing the test stimulus can be sent into the
eye through different points of the pupil, while
the point of entry of the beam forming the central
field remains practically unchanged. After such a
rotation of cube C, the diaphragm 7 must be
readjusted by a displacement in its own plane to
restore the test stimulus to its original position in
the observer’s field of view.

Monochromator
Mo Pape een ays

Conditioning
Stimulus

From Auxiliary Test

Optical System = Stimulus
Providing Orientation Orientation
Points (x) in Visual ; ’ Point
Field. Visual Field
From Monochromator

Fig. 1(5.15.2). Schematic diagram showing the principal components of an optical arrange-
ment used to make increment-threshold measurements and measurements of directional
sensitivity (redrawn from Stiles, 1939).
484 VISUAL EQUIVALENCE AND VISUAL MATCHING

VAN be aadINUAYpa VIN WAY

@ vs\, @ e @ A @ @ @ e® e@

\/ \/ °

o'No" Response
e Yes' Response
Stimulus
of
Le¢)
Log
Test
0.80

20 25 30 35
Trial Number
Fig. 1(5.15.3). Example
Pp of “yes”
y and “no” responses
po to p pulsed test stimuli of radiances L er
(arbitrary units, with fixed A) presented in a single-staircase series containing 36 trials. The
radiance increments or decrements are constant and equal to 0.05 log units. A threshold level of
log L,, = 1.075 (approx.) is derived from the sequence of data points between the fifteenth and
thirty-sixth trial.

5.15.3. The Staircase Methods
Both the single-staircase method and _ the
double-staircase method are being used quite fre-
quently now in experiments in which a visual
threshold of a pulsed stimulus is to be determined
with high precision by a relatively small number
of trial pulses. A fine account of the two methods
has been given by Cornsweet (1962) who must be
given credit for inventing the double-staircase
method which is an improvement over the origi-
nal single-staircase method. Figures 1(5.15.3) and
2(5.15.3) illustrate the experimental procedures
used in the two methods.
In the single-staircase method, the first stimu-
lus presented to the observer, for example, an
incremental test pulse on a fixed conditioning
stimulus as shown in Figure 1(5.15.2) of the previ-
ous section, has arbitrary radiance, but usually
somewhere near the level at which threshold is to

Trial Number - Staircase A

lO I5

= Yes' Response
Staircase A u
o'No' Response

Neos ah
i Mp EX AQ
Sf-t ef ay
Log
of
Stimulus
Ley
Test o No’ Response
Staircase B u
e Yes’ Response

f°) lO IS
Trial Number - Staircase B
Fig. 2(5.15.3). Example of “yes” and “no” responses to pulsed test stimuli of radiances L,,
(arbitrary units; with fixed 1) presented in two concurrently run, but randomly mixed, staircase
series, A and B. Total number of trials is 36, 18 for A and 18 for B. The radiance increments or
decrements in each staircase are constant and equal to 0.05 log units. A threshold level of
log L, , — 1.075 (approx.) is derived from the second half of the sequence of data points.

be expected. The observer responds to the test
pulse with “Yes” when it is seen, or with “No”
when it is not seen. If the observer says “Yes,”
the second pulse is given a lower radiance de-
creasing the original radiance by a small step
whose actual and most appropriate magnitude is
usually predetermined in exploratory trial runs. If
the observer says “No,” the second pulse is given
at a higher radiance. If the observer responds with
“Yes” to the second pulse, the third pulse will be
given a lower radiance; if the response is “No,”
the radiance will be increased. This procedure is
continued until some predetermined criterion or a
certain number of trial pulses has been reached.
In the ideal situation, a particular radiance level
can soon be identified after a few trial pulses,
around which the observer’s responses alternate
between “Yes” and “No” from any pulse to the
next.
The statistical analysis of such data (see, for
example, Dixon and Massey (1969), assumes that
the response to a particular test pulse is indepen-
dent of the preceding pulses and the responses to
them. This assumption may not hold with cer-
Instrumentation for Color-Vision Research 485
tainty and a bias in the data may be intro-
duced.
The double-staircase method greatly reduces
the possibility of a bias in the experimental re-
sults. The experimental procedure consists essen-
tially of running two series of staircased test
stimuli concurrently. This may be done by operat-
ing staircase A on odd trial presentations of the
test stimulus, whereas staircase B is operated on
even trial presentations. Staircase A starts with a
test pulse whose radiance is set above the ex-
pected threshold level; staircase B starts with a
test pulse of radiance below threshold level. After
a number of trials, the two alternating staircase
series converge toward a common radiance level
that characterizes the threshold level.
An often preferred alternative procedure is to
randomize the operation of the two staircase series
A and B, which is the case illustrated in Figure
2(5.15.3). This refinement virtually eliminates any
bias that may be introduced to the data by possi-
ble interdependencies between the response to a
given stimulus and the responses to preceding
stimuli.
 CHAPTER 6
UNIFORM COLOR SCALES
6.1 PREAMBLE
The experimental data and theoretical considera-
tion of this chapter deal with the ability of the eye
to judge differences in attributes of color percep-
tion [see Table 1(6.1)], such as brightness, hue,
and chroma, with the resulting possibility of
establishing uniform scales of brightness, hue,
and chroma. In some instances, the data are di-
rectly related to the CIE tristimulus space or the
CIE chromaticity diagram and are in a form
suitable for application to practical colorimetry
and industrial color-control problems. In other
instances, experimental data and some notes on
methods are given which are of more specific
interest to the research worker in color vision.
There are two important preliminary observa-
tions.
(a) In the derivation and application of much of
the material on color spacing and the establish-
ment of uniform color scales, the typical experi-
mental situation is provided by a fixed illuminant
(daylight, for example) under which material color
samples (painted chips, colored papers, etc.) are
exposed against a background surface, the latter
being usually nonselectively reflecting but with an
assigned reflectance that may range from high to
low. All the color stimuli are then object-color
stimuli whose tristimulus values define repre-
sentative points contained in the object-color solid
appropriate to the relative spectral radiant power
distribution of the illuminant being used. Also,
points in the object-color solid that mark out the
graduations of any derived color scale are realiz-
able, at least in principle, by actual samples. On
the other hand, color scales can be constructed
and formulated in terms that do not imply the
486
foregoing object-color situation. The distinction is
generally clear from the context but needs keep-
ing in mind.
(b) The usual judgments being made by the
observer often go beyond mere matching. The
observer is assessing differences or equalities in
certain attributes of the perception of the color
stimuli presented, attributes which are isolated
subjectively. The attributes are initially undefined
in any quantitative or numerical sense, and their
scaling has to be constructed from the subjective
judgments of observers, obtained in extended
studies.
An observer’s judgment of the relative magni-
tude of two color differences rests on the subjec-
tive impressions experienced when perceiving the
two pairs of color stimuli. The judgment varies
greatly with the viewing conditions and the kind
of stimuli (samples) presented, and, in general, its
precision assessed in any reasonable way com-
pares poorly with that of complete color-match-
ing. The difficulty of making such judgments is
reduced if the object-color perceptions concerned
are closely similar in one or two of the three basic
attributes of color perception, such as hue,
chroma, and lightness (or brightness). Thus, judg-
ments of the relative sizes of lightness (or bright-
ness) differences are readily made and with good
precision if the stimuli viewed yield at least ap-
proximately the same hue and chroma. Similarly,
judgments of differences in either hue or chroma,
if the stimuli yield the same lightness, are not
generally found difficult.
As the lightness, hue, and chroma differences
to be judged become large, the observer finds it
increasingly hard to assess their magnitude. There
is also very little practical value in judging
extremely large color differences. Of principal
Table 1(6.1) Basic Terms of Color Perceptions and Their Definitions

Light Light is that aspect of radiant energy of which a human observer is aware through
the visual sensations that arise from the stimulation of the retina of the eye by the
radiant energy.
Color Color is that aspect of visual perception by which an observer may distinguish
differences between two structure-free fields of view of the same size and shape,
such as may be caused by differences in the spectral composition of the radiant
energy concerned in the observation. In this sense, the term color is sometimes
referred to as perceived color to distinguish it from color used in the sense of
psychophysical color. Basic terms of psychophysical color have been given in Table
I(3.2).
A related color is a color perceived to belong to an area seen in relation to other
colors.
An unrelated color is a color perceived to belong to an area seen in isolation from
other colors.
A chromatic color is a perceived color possessing hue.
An achromatic color is perceived color devoid of hue.
Hue Hue is the attribute of a color perception denoted by blue, green, yellow, red, purple,
and so on.
Unique Hues Unique hues are hues that cannot be further described by the use of the hue names
other than its own (also referred to as unitary hue). There are four unique hues, each
of which shows no perceptual similarity to any of the others; they are: red, green,
yellow, and blue. A light (color stimulus) perceived to be unique red is judged to be
neither yellow nor blue. Similarly, unique yellow is neither red nor green; unique
green is neither yellow nor blue; and unique blue is neither green nor red. The
hueness of a light (color stimulus) can be described as combinations of two unique
hues; for example, orange is yellowish-red or reddish-yellow. Sometimes, nonunique
hues such as orange are referred to as binary hues.
Brightness Brightness is the attribute of a visual sensation according to which a given visual
stimulus appears to be more or less intense; or, according to which the area in which
the visual stimulus is presented appears to emit more or less light.
Variations in brightness range from “bright” to “dim.”
Lightness Lightness is the attribute of a visual sensation according to which the area in which
the visual stimulus is presented appears to emit more or less light in proportion to
that emitted by a similarly illuminated area perceived as a “white” stimulus. In a
sense, lightness may be referred to as relative brightness. Variations in lightness
range from “light” to “dark.”
Chromaticness | Chromaticness is the attribute of a visual sensation according to which the
(perceived) color of an area appears to be more or less chromatic.
Chromaticness has also been referred to as colorfulness.
Chroma Chroma is the attribute of a visual sensation which permits a judgment to be made
of the degree to which a chromatic stimulus differs from an achromatic stimulus of
the same brightness.
Saturation Saturation is the attribute of a visual sensation which permits a judgment to be
made of the degree to which a chromatic stimulus differs from an achromatic
stimulus regardless of their brightness.

487
488 UNIFORM COLOR SCALES
interest is the study of small differences and, in
the limit, of differences which the observer finds
to be just-noticeable or to be on the threshold.
Just-noticeable or threshold differences may be
considered as the natural units for measuring
larger differences. However, the phenomena and
concepts associated with threshold judgments dif-
fer in various ways from those involved in the
assessment of small but clearly perceived color
differences and additional discussions and data
will be presented in Chapter 7.
6.2 TYPES OF SCALES AND SCALING
METHODS
In recent years, the fundamental concepts of scal-
ing psychological and psychophysical attributes
have been clarified greatly, and methods of scal-
ing have been developed which, when combined
with appropriate statistical techniques, provide
powerful tools for the psychologist and psycho-
physicist. The application of these methods to the
scaling of the attributes of color perception and
to quantifying color discrimination is well estab-
lished. Also, the theory of scaling has been put on
a sound foundation and is well covered in the
literature. Excellent recent accounts on the sub-
ject have been given by Torgerson (1958); Krantz,
Luce, Suppes, and Tversky (1971); Krantz (1972,
1974); Falmagne (1974); Stevens (1975); and
Gescheider (1976).
The scope of our book does not allow us to
treat the often elaborate theories and methods of
scaling in detail, but a few brief notes on some of
the basic concepts are included here which have
particular relevance to the data of this chapter.
A color scale is a series of ordered numbers
which represents observable gradations of a given
attribute or gradations of a combination of attri-
butes of color perception.
Different types of color scales can be estab-
lished in accordance with different operations
that can be performed on numbers. In following
the exposition of Torgerson (1958), the following
four types of scales may be used for one-dimen-
sional attributes:
I. Ordinal scale
II. Ordinal scale with natural origin
III. Interval scale
IV. Ratio scale
In the ordinal scale, the numbers, y say, are
assigned to the magnitudes x of the attribute, so
that the order of the numbers corresponds to the
order of these magnitudes. An ordinal scale is
thus represented by a monotonic function, y =
f(x). Any other function y = $(x) will also serve,
provided that $(x) is related to f(x) by a mono-
tonic, that is, order-preserving transformation.
With appropriately chosen sets of color stimuli,
ordinal scales can be established readily for the
perceptual attributes of lightness and of chroma.
Consider, for example, a set of colored paint
chips picked at random from the Munsell Book of
Color (Section 6.6.1), illuminated by daylight, and
displayed randomly in front of the observer. The
observer will have little or no difficulty in order-
ing the chips in accordance with what he or she
perceives their lightnesses to be under the given
observing conditions. In fact, the only difficulty
that the observer may have is one of conceptual
nature concerning the term lightness. The ob-
server's idea of what is meant by lightness may
differ from that of the experimenter particularly
when the colored chips do not only vary in light-
ness, but also in hue and chroma. However, in the
case of lightness, such conceptual difficulties are
rare and different observers agree remarkably well
with each other’s ordinal scales of lightness gener-
ated with a given set of colored paint chips.
Ordinal scales of chroma can also be estab-
lished fairly readily, though the conceptual diffi-
culties that are associated in this case are often
more pronounced than in the case of lightness. In
fact, other attributes have been identified that are
related to chroma or to a combination of chroma
and lightness. The attribute of saturation is per-
haps the most prominent example of such another
attribute. In more recent work, the attribute of
colorfulness has been identified by Hunt (1977) as
the attribute that may actually be the basic attri-
bute accounting for the variation in color percep-
tion that cannot be described as variations in
lightness or in hue. When taken in this sense, the
saturation and chroma of a given stimulus are
then special attributes derivable from assessments
of the colorfulness and lightness of the stimulus.
The interesting arguments put forward by Hunt
deserve further study (see also Section 5.12.2).
The attribute of hue perceived by the observer
when viewing color stimuli does not present, in
general, any conceptual difficulties as to what is
meant by the term hue. However, it is clear that
what is meant is ordering magnitudes of four
attributes of hueness: redness, yellowness, green-
ness, and blueness. The four attributes are directly
associated with the “unique hues” of red, yellow,
green, and blue (Sections 5.9 and 5.13).

Although ordinal scales exist that have no
natural origin, the most common attributes of
color perception all have a natural origin; that is,
there exists a point on the scale to which the
number zero is assigned when the magnitude of
the attribute considered is, in fact, perceived to be
nil. Lightness, brightness, saturation, chroma, col-
orfulness, redness, yellowness, greenness, and
blueness can each be characterized by an ordinal
scale with a natural origin.
The basic property of an ordinal scale, that is,
its monotonicity, is retained for ordinal scales
with a natural origin within order-preserving
(monotonic) transformations that include preserv-
ing the natural origin.
An interval scale is, in the first instance, an
ordinal scale and as such fulfills all the require-
ments (postulates or axioms) of an ordinal scale.
However, two additional requirements must be
added:
(a) Differences between different magnitudes
can indeed be determined.
(b) Numbers can be assigned to these magni-
tude differences so that the differences between
numbers characterize the sizes of the correspond-
ing observed magnitude differences of the attri-
bute.
In scaling of psychological attributes, it has
often been useful to introduce the notion of dis-
tance, that is, to identify an observed magnitude
difference of the attribute by the distance (inter-
val), say x, and then to represent x by a number
y. The number y increases or decreases in direct
proportion to an observed increase or decrease in
distance x. It follows that in an interval scale, the
numbers are determined to within a linear trans-
formation of the form
Vad (X) Sax [1(6.2)]
where m is any positive number, and y is any
finite number.
A ratio scale is an interval scale with a natural
origin. The invariance characteristics of an inter-
val scale given by Eq. 1(6.2) is retained in the
ratio scale except that now only one number,
instead of two, may be assigned arbitrarily and
the linear transformation within which the ratio
scale remains invariant takes on the reduced form
y = f(x) = mx [2(6.2)]
Types of Scales and Scaling Methods 489
which leaves the natural origin of the ratio scale
unchanged.
Both interval scales and ratio scales are by far
the most useful scales in efforts of quantifying
attributes of color perception. In the particular
case in which adjacent intervals x (distances
X1,X7,X3,°°:) are assigned the same number y
(¥1, ¥2, ¥3,°°°) throughout the domain of the
attribute, we have an interval scale of equal spac-
ing, which, in the context of this chapter, is
referred to as a (one-dimensional) uniform color
scale.
For a detailed formal treatment of the systems
of axioms that govern ordinal and interval scales,
the reader is referred to the book by Krantz et al.
(1971) entitled Foundations of Measurement.
Scaling methods are procedures that attempt to
give quantitative descriptions of perceptual attri-
butes. Different scaling methods have been used
(for summaries see, for example, Torgerson, 1958; .
Krantz, 1972; Stevens, 1975; Gescheider, 1976),
and the following have been selected from those
most applicable to scaling the attributes of color
perception.
(i) Confusability Scaling. Confusability scaling
refers to a procedure in which the observer is
required to discriminate between color stimuli
that evoke only slight differences in visual sensa-
tion. Typically, the stimulus confusions are de-
fined by the just-noticeable difference (jnd) [also
called the just-perceptible difference (jpd)], which
serves as a unit of sensation difference. The defi-
nition of the jnd is usually linked with the
numerical value of the probability of confusion, or
with some other discriminability index. The jnd
concept is directly associated with the concept of
threshold (Chapter 7). The number of jnds is then
counted to establish the desired scale, which is an
interval scale.
A good example is obtained by considering an
interval scale of brightness in which a stimulus of
fixed relative spectral radiant power distribution
is arbitrarily adjusted to luminance x providing
the zero point of the scale. If y, denotes a slightly
higher luminance of the same stimulus that is
perceived as just-noticeably brighter than x, then
y, is assigned a brightness value of 1; if the
luminance of the stimulus is raised to y5, per-
ceived as just-noticeably brighter than y,, and
thus, two jnds above x, y, is given the brightness
value 2; and so forth.
For the interval scale constructed in this way
to be useful, it is essential that the scale remain
invariant within a linear transform [Eq. 1(6.2)]
490 UNIFORM COLOR SCALES
when
(a) the method by which discrimination is tested
is changed, and
(b) another discriminability index is used to de-
fine the jnd.
A discussion of the invariance of confusability
(jnd) scales, when changes (a) and (b) are intro-
duced, has been given by Krantz (1972). A
mathematically formal exposition of a theory of
discrimination has been given by Falmagne (1974).
The Weber law and the Fechner law play key
roles in the basic concepts of confusability scal-
ing.
There is considerable evidence that shows when
jnds are determined for a number of stimuli that
differ with respect to the perceptual attribute
(e.g., brightness); the size of the jnd, measured in
physical or psychophysical units (e.g., radiance or
luminance) for the given attribute depends on the
magnitude of the stimuli involved. In general, it is
found that the greater the magnitude of the
stimuli, the greater the size of the jnd. This em-
pirical fact is mathematically stated in Weber’s
law.
=k [3(6.2)]
where Ax is the increment that must be added to
the magnitude x of the given stimulus to be just
noticeable in accordance with some specified value
of the discrimination index. The quantity x, is a
constant, independent of the specified value of
the discrimination index, and is often interpreted
as internal noise in the visual mechanism. The
proportionality constant k is called the Weber
fraction.
Postulating that jnds are equal increments As
in “sensation magnitude” at all stimulus magni-
tudes, and assuming that the empirical Weber law
is valid, it follows that
me On
pee sarc ok [4(6.2)]
with k’ being a constant factor specifying an
appropriate unit of the sensation-magnitude in-
crement. By treating As and Ax as differentials,
ds and dx, respectively, Eq. 4(6.2) can be in-
tegrated
s(x) = fdas=k'f ie
x + Xo
which leads to the expression of the form
s(x) =a+t blog(x + xo) [5(6.2)]
relating by a logarithmic function the stimulus
magnitude x, measured in terms of a physical or
psychophysical unit, to the sensation magnitude s.
The quantities a and b are constants. The relation
expressed by Eq. 5(6.2) is known as the Fechner
law.
Another important concept in confusability
scaling is the law of comparative judgment derived
by Thurstone (1927a, b).
Thurstone (1927a,b) postulates that each
stimulus, when presented to an observer, has a
variable effect on a hypothetical continuum of
sensation magnitude. The variability is described
by a normally distributed random variable, called
a discriminable process. The mean and standard
deviation of the normal distribution associated
with each given stimulus are taken as its scale
value and discriminal dispersion, respectively.
For example, if required to judge which of two
given stimuli, one of luminance x and the other of
luminance y, is perceived as being the brighter of
the two, the observer samples randomly from the
bivariate normal distribution defined by the two
discriminal processes, and chooses whichever
value is the higher. If we denote the means of the
two discriminal processes by S(x) and S(y) and
their standard deviations by o(x) and a(y), re-
spectively, with p(x, y) being the correlation
coefficient, the probability, P(y, x), that y is per-
ceived brighter than x is then simply the probabil-
ity that the difference between the sampled sensa-
tion magnitude evoked by y and the sampled
sensation magnitude evoked by x is positive. This
sample difference is normally distributed, has a
mean value equal to the difference [S(A) — S(x)]
of the means of the two-dimensional processes,
and a standard deviation o( y, x) given by
o(y,x) =[0?(y) + 67(x)
~2p(y, x)o(y)o(x)] ”
[6(6.2)]
It follows that
S(y) — S(x)
ro.x=r| ST adie
[7(6.2)]

where f' denotes the familiar cumulative normal
distribution function, that is,
jou tate aa I 2 =) aes
The special case in which both o(x)
p(y, x) are constant, and thus, in accordance
with Eq. 6(6.2),
o(y, x) = constant [8(6.2)]
is known as Thurstone’s Case V. Krantz (1972),
whose exposition we have followed here, notes
that, in this case, Fechner’s law of sensation would
be justified.
If Thurstone’s Case V does not apply, that is,
o(y, x) # constant
and, in particular, if nonconstancy of o(y, x) is
caused by p(y, x) being some function of the
discriminability defined by P(y,x), it can be
shown (Krantz, 1972) that the standard deviation
of the discriminal process is a linear function of
its mean, and that the sensation magnitude fol-
lows a power law. Thus,
S(x) = asb(xbhaxg)”
and
o(x) =c(x+x,)" [9(6.2)]
Or
c attribute over a certain range of magnitudes, those
ae ints pecayeetn
Stevens (1959) has referred to the above case as
Thurstone’s Case VI and the power function,
given in a slightly different form, is called the
Stevens law, though it is usually based on direct
estimation and not on confusability scaling
(Gescheider, 1976; see further on, this section).
The method of paired comparisons readily fol-
lows from the law of comparative judgment under
the condition characterizing Thurstone’s Case V.
From Egs. 7(6.2) and 8(6.2) we deduce
S(y) — S(x) = of [P(y, x)]
[10(6.2)]
where o = const can be set equal to unity. The
Types of Scales and Scaling Methods 49]
difference between the scale values S(y) and S(x)
of the two given stimuli (y and x) is then given as
the inverse normal transform f of the probability
FUY- x).
For n given stimuli, denoted by x,, x5,...,X,,
there exist n(n — 1)/2 pairs of stimuli. Each pair
and is presented to the observer who is required to
indicate which member of the pair appears greater
with respect to the perceptual attribute to be
scaled (for example, brightness). The observer
must designate one member of the pair as greater;
no equality judgments are allowed. In order to
make adequate estimates of the proportion of
greater, a large number of comparisons must be
made for each pair of stimuli. This may be
accomplished by (a) having a single observer judge
each stimulus pair many times, (b) each of many
observers judge each pair just once, and (c) each
of several observers judge each pair several times.
The particular method, (a), (b), or (c), to be used
depends on the aim of the experiment and the
type of stimuli involved. Details of the numerical
procedures commonly used in analyzing paired
comparisons are given, for example, by Torgerson
(1958).
(ii) Scaling by Direct Estimation. Methods of
direct estimation of the magnitude of a given
perceptual attribute may be grouped into partition
scaling and ratio scaling.
Partition scaling yields an equal-interval scale
of the perceptual attribute with boundaries identi-
fied by specific stimuli. In the equisection method,
the observer is required to choose from a large
number of stimuli, representing the perceptual
stimuli that result in a specified number of per-
ceptually equal differences. In the method of
equal-appearing intervals, a common form of cate-
gory scaling, the observer is required to assign all
stimuli of the large set of given stimuli to a
specified number of categories. In this technique,
it is assumed that the observer is capable of
keeping the intervals between the boundaries of
the categories perceptually equal as stimuli are
assigned to the various categories. The experi-
menter then treats the category value assigned to
a particular stimulus as a value on the interval
scale.
Ratio scaling yields an equal-interval scale of
the perceptual attribute which includes a natural
origin; that is, it yields a ratio scale. The specific
techniques used in ratio scaling are often referred
to as ratio production, ratio estimation, magnitude
492 UNIFORM COLOR SCALES
estimation, and magnitude production. In each case,
the observer is required to judge the ratio between
two magnitudes of the perceptual attribute under
study.
In ratio production, also known as fractiona-
tion, the observer adjusts a variable stimulus while
observing a reference stimulus. Prescribed magni-
tude (e.g., brightness) ratios between test and
reference stimuli are produced by the observer.
In ratio estimation, the observer estimates the
magnitude ratio between the given test and refer-
ence stimuli; no adjustment by the observer of
either the test or the reference stimulus is made.
In magnitude estimation, the observer makes
direct estimates of the perceived magnitudes of
given stimuli and assigns appropriate numbers to
those estimates. A commonly used form of in-
struction (Stevens, 1975) given to the observer is
the following:
You will be presented with a series of stimuli in irregu-
lar order. Your task is to tell how intense (bright,
saturated, etc.) they seem by assigning numbers to them.
Call the first stimulus any number that seems ap-
propriate to you. Then assign successive numbers in
such a way that they reflect your subjective impression.
There is no limit to the range of numbers that you may
use. You may use whole numbers, decimals, or fractions.
Try to make each number match the intensity (bright-
ness, saturation, etc.) as you perceive it.
In magnitude production, the observer adjusts a
variable stimulus to produce a perceived magni-
tude whose numerical value seems to match that
of the magnitude of some given stimulus.
Of the four techniques used in ratio scaling,
the one referred to as magnitude estimation is the
most commonly used in color-perception studies
(see, for example, Section 5.12).
It is important to note that all so-called “di-
rect estimation” methods of scaling imply that the
observer makes “comparative judgments” of at
least two pairs of stimuli. The number assigned to
the perceived magnitude of a given stimulus must
_be “in proportion” to the magnitude of at least
one previous stimulus, presented earlier in the
particular experiment or remembered by the ob-
server from his or her bank of experiences with
color stimuli. Thus, direct estimation of the mag-
nitude of a given perceptual attribute cannot,
strictly speaking, be an absolute measure of the
-attribute’s magnitude.
A great deal of experimental evidence exists
that for many one-dimensional continua, not only
those experienced in color perception, the mean
estimate (x) of the magnitude of a given attri-
bute increases approximately as a power function
of the “intensity” x of the stimulus exhibiting
that attribute; that is,
W(A) = bx? [11(6.2)]
In this power law, Stevens (1975) considers
the exponent p the parameter that characterizes
the perceptual continuum [loudness (p = 0.67),
brightness (p = 0.33), taste (salt; p = 1.4), smell
(p = 0.6), heaviness (p = 1.45)]. In each case, p
may vary somewhat with individuals and with
stimulating conditions (e.g., state of adaptation),
but nonetheless, a representative value of p seems
to exist for each of many perceptual continua.
The exponent p is generally estimated from the
slope of the power function in log—log coordi-
nates, that is, from
log ¥(x) = log b+ plogx [12(6.2)]
When the variable x has low values, a better
agreement between magnitude estimation and
power function is obtained by introducing a con-
stant in Eq. 11(6.2). Three different forms have
been used:
(ho) OC xg)" [13(6.2)]
W(x) = bx? — bh [14(6.2)]
W(x) = b[(x + x9)? — x8] [15(6.2)]
In Eq. 13(6.2), the constant x, is sometimes inter-
preted as the stimulus intensity at which “abso-
lute threshold’”’ is observed. Thus, the constant
shifts the stimulus scale so that its zero point
corresponds to the zero point on the perceptual
scale (Ekman, 1956, J.C. Stevens and S.S. Stevens,
1963). However, the connection between x, and
the concept of theshold in the conventional sense
of a statistical quantity is not made clear. In
Eq. 14(6.2), the subtractive constant Wp is used
which Jameson and Hurvich (1964) introduced to
account for simultaneous contrast effects. In Eq.
15(6.2), the constant x, is chosen to represent a
special case of the earlier expressions, given as
Eq. 9(6.2), which arose from the combination of
the Weber law and the law of comparative judg-
ment of Thurstone (Krantz, 1972).
The three variants of the power law given by
Eqs. 13(6.2), 14(6.2), and 15(6.2) exhibit an

asymptotic slope p when logy(x) is plotted
against log x. In the case of 0 < p < 1, the slopes
for small x-values are steeper than p; this corre-
lates well with most experimental data.
(iii) Multidimensional Scaling. The scales con-
sidered above are one-dimensional scales; that is,
they quantify attributes whose perceived magni-
tudes or magnitude differences form a one-dimen-
sional series of numbers. In that context, the
attribute itself is considered as a one-dimensional
continuum that can be represented geometrically
by a straight line. When taking “color” as the
attribute to be scaled, instead of just one of its
component attributes (e.g., lightness, or hue, or
chroma), it is immediately obvious that the scal-
ing problem is more complex and the magnitude
differences that one may wish to scale may form a
multidimensional series with its geometrical repre-
sentation in a multidimensional space. Each ob-
served magnitude difference is, in effect, assigned
a set of numbers which, in terms of a specified
geometrical system, characterizes the relations
between the various observed magnitude dif-
ferences, each of which is represented by a point
in the multidimensional space. The number of
numbers that is assigned to each observed magni-
tude difference specifies the dimensionality of the
space and thus, the dimensionality of the attri-
bute.
There are different spaces, each characterized
by a different metric, that can be employed as the
underlying geometrical model for multidimen-
sional scaling. The Euclidean metric is the one
most commonly used, but other metrics and even
nonmetric multidimensional scaling procedures
have been developed. An excellent introductory
booklet on multidimensional scaling has been
written by Kruskal and Wish (1978) in which the
reader will find many worked examples of multi-
dimensional scaling demonstrating different anal-
ysis procedures and data interpretations. Details
of computational procedures, all of which require
a computer, have been given, for example, by
Kruskal (1976). Also, the book entitled Multidi-
mensional Scaling (Vol. I; R.H. Shepard et al.,
Editors, 1972) contains useful information.
Multidimensional scaling is an indirect method
for determining the number of the component
attributes that are evoked by a given set of color
stimuli. It also offers the possibility of extracting
the internal structure of the given observational
data that may offer useful insights into the visual
mechanism itself. In following Shepard’s (1962)
Brightness and Lightness Scales 493
method, which is widely used in practice, for each
pair of stimuli some “proximity measure’’ is first
obtained. Then, one looks for a configuration of
points in Euclidean space, one point correspond-
ing to the observed magnitude of the attribute
associated with each stimulus, such that the order-
ing of magnitude proximities between stimuli is
reproduced as closely as possible by the ordering
of the inverse distances between the correspond-
ing points. The aim is to generate a reproduction
of the proximity ordering in as small a number of
dimensions of the Euclidean space as possible. In
general, the computational procedure provides a
good fit for a dimensionality => n, but the good-
ness of the fit seems to collapse abruptly for
dimensionality n — 1. When this occurs, it is con-
cluded that the proximities are based on varia-
tions in n independent perceptual component
attributes. Often an attempt is also made to draw
inferences on the nature of the component attri-
butes, but this is difficult because the “principal
axes” that the analysis provides are only unique
within an arbitrary rotation and translation in the
given space model. However, a kind of informal
principal-axes analysis can sometimes lead to a
useful insight into the nature of the component
attributes. |
In color perception, the dimensionality prob-
lem is, in general, of lesser concern to the re-
searcher than it is in other fields of behavioral or
social study. There is overwhelming evidence,
originating from psychophysical experiments, that
perceived color is a three-dimensional attribute,
whose most commonly isolated component attri-
butes are (1) hue, (2) chroma, and (3) lightness. In
fact, the three-dimensionality of perceived color is
considered so well founded that it is often used as
a criterion for the validity of a particular multidi-
mensional analysis.
Multidimensional color scaling has been used
in many studies. The following is a selection of
such studies: Torgerson (1958); Indow and
Kanazawa (1960); Shepard (1962); Indow (1963,
1980); Helm (1964); Carroll and Chang (1970);
Carroll and Wish (1974).
63 BRIGHTNESS AND LIGHTNESS SCALES
Brightness is defined as the attribute of a visual
sensation according to which a given visual
stimulus appears to be more or less intense, or
according to which the area in which the visual
stimulus is presented appears to emit more or less
494 UNIFORM COLOR SCALES
light. The magnitude of brightness can be esti-
mated for unrelated visual stimuli as well as for
related visual stimuli. Unrelated visual stimuli are
stimuli presented individually in isolation (dark-
ness). Related visual stimuli are presented in a
simple or complex display of other visual stimuli.
Brightness scales have been determined for a
variety of observing conditions by several investi-
gators, notably by S.S. Stevens and his collabora-
tors (J.C. Stevens and S.S. Stevens, 1963; Aiba
and S.S. Stevens, 1964; S.S. Stevens and
Diamond, 1965), Jameson and Hurvich, 1964;
Bartleson and Breneman, 1967a,b; Jameson,
1970; and Bodmann, Haubner, and Marsden,
1980. The power law, which is used either in the
form of Eq. 13(6.2) or in the form of Eq. 14(6.2),
is generally found to provide a satisfactory fit to
the experimental data. However, the more recent
studies, particularly those by Bodmann, Haubner,
and Marsden (1980) provide a convincing case
that the power law in the form of Eq. 14(6.2), that
is,
[1(6.3)]
is to be preferred over the one in the form of Eq.
13(6.2) that has been favored especially by S.S.
Stevens and his collaborators. In Eq. 1(6.3), B
denotes a measure of the brightness estimated
by an observer presented with a stimulus of
luminance L under specified conditions of ob-
servation. The exponent p has a value of ap-
proximately 1/3; Bodmann et al. (1980) find
p = 0.31 + 0.03. The values of a and B, depend
on the observing conditions (such as field size of
test stimulus, luminance of surround) and include
an arbitrary scaling factor. The term By has been
given a plausible physiological interpretation
by Jameson and Hurvich (1961, 1964); see also
Jameson (1970) and Hurvich and Jameson (1972)
who see it as a “response correction” caused by
the neural process of induction when the test
stimulus is viewed in a surround of nonzero lumi-
nance. Haubner (1977) has advanced this idea
further.
Figure 1(6.3) illustrates typical results of
brightness-scaling experiments. The two curves
were drawn from data obtained by Bodmann
et al. (1980). They refer to brightness scaling
made with a dark surround (0 cd - m~*) and with
a surround of luminance 300 cd- m 7”. In both
cases of observing conditions, interval scaling (bi-
section method) and ratio scaling (magnitude
estimation) were used to derive the corresponding
brightness scales. The test stimuli had an angular
subtense of 2° and the surround was 180°. Both
test and surround stimuli were achromatic. Con-
trary to work by S.S. Stevens and his collabora-
tors, Bodmann et al. find that brightness scales
obtained by bisection are /inearly related to those
obtained by magnitude estimation.
Lightness is defined as the attribute of a visual
sensation according to which the area in which
the visual stimulus is presented appears to emit
more or less light in proportion to that emitted by
a similarly illuminated area perceived as a “white”
stimulus. Lightness is thus an attribute of visual
sensation that has meaning only for related visual
stimuli (see above). As lightness is judged with
reference to the brightness of a “white” stimulus,
it may be considered a special form of brightness
measure that could be referred to as relative
brightness.
Perhaps the largest body of existing data on
uniform color scales is the one dealing with the
establishment of uniform lightness scales for a
variety of conditions of observation. These data
refer to samples that have approximately the same
hue and chroma, and differ only in lightness.
Most of the investigations are subject to the re-
striction that they deal with achromatic stimuli
only, and the lightness scales then developed are
often called gray scales.
One way of developing a gray scale is by the
method of bisection which leads to an interval
scale. The observer is given a black chip, a white
chip, and a number of different gray chips and is
asked to select a gray that appears about as
different from white as it does from black. Then
the interval between the black and mid-gray is
halved by the same procedure, and also that
between white and mid-gray, resulting in a gray
scale extending from black to white by four steps
that are perceived to be equal. The uniform sub-
division can be continued, following the same
procedure, and a multiple-step gray scale is then
obtained.
Another way to develop a gray scale is by
making use of the principles of confusability scal-
ing (Section 6.2) which again lead to an interval
scale. One starts with black, selects a gray that is
just-noticeably lighter, then a gray that is just-
noticeably lighter than the latter, and so on until
the white or near-white is reached.
Both procedures can also be used to develop
uniform lightness scales for colored chips, all of
approximately the same hue and chroma.

105
BBrightness
Measure
| 10 102
Luminance [cd-m-@] of Test Stimulus
Fig. 1(6.3). The functional relationship between brightness and luminance for two different
surround luminances (from data obtained by Bodmann et al., 1980).
2
The results of such experiments are conveni-
ently represented in a graph where the ordinate is
the number of lightness steps V between black
and each gray, and the abscissa is the luminance
factor Y of the corresponding chip. The relation-
ship between V and Y can be expressed by means
of an empirical formula, which in more recent
investigations has commonly taken the form of a
power function, in accordance with the notion of
the power law (Section 6.2). Table 1(6.3) in the
Appendix lists a number of formulae reported by
different investigators and Figure 2(6.3) shows
the corresponding graphical relations between V
and Y.
There are important differences between some
of the curves. These differences mean that the
spacing of the lightness scales, represented by the
curves, is different. For example, the luminance
factor of the middle gray (V = 5) for Curve 5
corresponds to that for the third step (V = 3) for
Curve 1. These differences are generally attri-
buted to the differences in the observing condi-
tions to which the different lightness scales are
intended to apply. In particular, the luminance
factor of the background is considered the most
important parameter influencing the spacing of a
lightness scale. The background essentially con-
trols the adaptive state of the observer’s eyes,
particularly when the scale samples are relatively
small.
Studies of the influence of the adaptive state
of the observer’s eyes on lightness (and bright-
Brightness and Lightness Scales 495
Surround luminance
‘He O cd-m~2
2: 300 cd-m-2
Tek 104 105
ness) and its uniform scaling have been made for
quite some time. Notable examples of early stud-
ies are those by Adams and Cobb (1922) and
Judd (1940). But it was only in recent years that a
much clearer picture has emerged. The studies by
Kaneko (1964), Takasaki (1966), and Semmelroth
(1970, 1971) are of particular interest to lightness
scaling.
Kaneko (1964) constructed experimentally
lightness scales for black, gray, and white back-
grounds and for samples of different sizes. As
expected, he found that the luminance factor of
the background has an important effect on the
spacing of the lightness scale. He also found that
the background effect was more pronounced for
smaller samples. But of particular interest is
Kaneko’s finding that a gray background en-
hances the sensitivity of the observer to lightness
differences between gray samples of about the
same luminance factor as that of the background.
This phenomenon has been observed before (for
example, by Adams and Cobb, 1922) but was
considered to apply only to the case involving
rather small differences, of about threshold size.
Kaneko’s experimental data show that the phe-
nomenon can also be observed when relatively
large lightness differences are present, such as
those between neighboring samples of a 10-step
lightness scale going from black to white.
In his attempt to quantify his experimental
data, Kaneko derived a lightness-index formula
based on the formula by Adams and Cobb (1922)
 496 UNIFORM COLOR SCALES

—x——_
3 and! 4
—— 5
—_Oo— 6

0 20 40 60 80 100
y
Fig. 2(6.3). Relationship between lightness-scale value V and luminance factor Y plotted in
accordance with different formulae [see Table 1(6.3)]. 1: Priest, Gibson, and McNicholas
(1920); 2: Munsell, Sloan, and Godlove (1933); Godlove (1933); 3: Newhall, Nickerson and
Judd (1943) (also known as the Munsell Renotation-Value Scale); 4: CIE (1976) lightness scale
(plotting V = L*/10); 5: Foss (1944) (gray scale of the Color Harmony Manual); 6: Richter
(1953) (gray scale of the German DIN Color Chart; plotting V = 10—D).

and modified by Judd (1940). Kaneko’s formula
consists of two simultaneous equations:
Fit ed
ae AS te
* + -Y,a — (1 — Y,
y a + Ysa = 1)a¥e [2(6.3)]
rq(Yz rt a)
where V denotes the lightness index and Y and Y,
the luminance factors of the sample and the back-
ground, respectively. The parameter r denotes
sample size; small values of r refer to small sam-
_ ples. The parameter g denotes the level of illumi-
nance at the plane of the viewing field containing
the background and sample. The parameter a
denotes the state of adaptation of the observer’s
eye.
Takasaki (1966) presented to his observers a
viewing situation as sketched in Figure 3(6.3).
‘There are two backgrounds, Bl and B2, side by
side with luminance factors Y,, and Y,,, respec-
tively. In the center of each background there are
presented gray samples S1 and $2 with lumi-
nance factors Y, and Y;, respectively. For any
given sample $1 and background B1 on the left
side, the observer was instructed to choose, for a
given background B2 on the right, a sample $2
appearing as light as that on the left.
A typical set of results is given in Figure
4(6.3), where the logarithm of the luminance fac-
tor Y, of sample $2 is plotted against the loga-
rithm of the luminance factor Y, of sample S1.
When the two backgrounds are the same, we
expect no difference between Y, and Y, and,
indeed, the observer does not report a difference.
The 45° line in Figure 4(6.3) illustrates this trivial
result. When the background B2 is lighter than
Bl, a sample $2 is always chosen with a lumi-
nance factor Y, higher than that of sample S1.
We also find that, when the background B2 is
darker than Bl, a sample $2 is always chosen
with a luminance factor Y, lower than that of
sample $1. These observational results are, of
course, in agreement with the well-known con-
trast effect which says that a relatively small gray

Fig. 3(6.3). Sketch of visual field presented to
observers in Takasaki’s (1966) experiment. B1 and
B2 are gray backgrounds of different luminance
factors: S1 and $2 are gray samples. The observer
is instructed to match the lightness of $2 with the
lightness of S1 by finding a sample $2 with the
appropriate luminance factor.
sample looks lighter (darker) when it is placed on
a darker (lighter) background.
Takasaki’s results, as illustrated in Figure
4(6.3), also show a hump in the lightness-change
curve, in addition to the expected parallel dis-
placement from the 45° line. This hump results
from the fact that the observer sees the lightness
of sample S2 change rapidly with change in Y,
when Y, is close to Y,, of the background B2.
N5-N7
Ye
0.1
0.01
0.01 0.05 0.1 05 1.0
Y4
Fig. 4(6.3). Typical results of an observer making
lightness matches of the kind sketched in Figure
3(6.3). The luminance factors Y, and Y, refer to
samples S, and S,, respectively. The backgrounds
B1 and B2 are Munsell grays of values N.../.
The results of three different background pairs
(B1 — B2) are shown: (N1/-N9/); (N3/-
N7/); (N5/-N7/). (After Takasaki, 1966.)
Brightness and Lightness Scales 497
Takasaki calls this phenomenon the crispening
effect. It is the same effect Kaneko (1964) re-
ported in his study, discussed above; a gray back-
ground enhances the sensitivity of the observer to
lightness differences between gray samples of
about the same luminance factor as that of the
background.
In analyzing his experimental data, Takasaki
arrives at an empirical lightness-index formula
which predicts fairly well both the general con-
trast effect and the crispening effect when light-
ness changes are observed for a wide range of
experimental conditions. His formula is of the
following form:
(V, se ra)
TOV
I fee OV ee ca]
C3
|
X exp [3(6.3)]
C3
When i = 1, we have the lightness index 7, for
sample $1 on background B1; similarly, when
i = 2, we have T, for $2 on B2. The formula does
not use the luminance factors Y, and Y;, and Yp,
and Y,,, but, instead, the Munsell renotation
values V,, V,, Vz,, and V,,5, respectively, which
are derived from the luminance factors by means
of Equation (3) in Table I(6.3). The quantity V, is
the average of the Munsell renotation values for
the two backgrounds, thus, Vg = (Vz; + Vp>)/2.
There are three constants c,,c,,c3; which vary
somewhat from observer to observer and depend
on observing conditions, such as sample size and
presumably also level of illuminance at the sam-
ple plane.
The general contrast effect is taken into
account by the first two terms of Eq. 3(6.3).
This is in agreement with the induction theory
(Jameson and Hurvich, 1961; 1964) which states
that a certain part of the visual excitation from
the background is inversely induced on the sam-
ple area and therefore changes its apparent light-
ness. Thus, the constant c, controls the amount of
induction.
The remaining term in Eq. 3(6.3), an exponen-
tial function, takes into account the crispening
effect. The factor c, controls the amount of cris-
pening and the factor c, adjusts the reciprocal
sharpness of the crispening.
Semmelroth (1970) proposes a model for the
functioning of the visual mechanism responsible
for the perception of brightness and lightness. He
 498 UNIFORM COLOR SCALES
postulates that the visual mechanism responds
simultaneously to two signals of the stimulus field.
First, there is a response a to the luminance of the
stimulus. A power law is assumed to relate the
magnitude of the response to the luminance level.
Then, there is a response B to the difference of
luminance between the sample and its back-
ground. Again, a power law is assumed to relate
the magnitude of this response to the luminance
difference. It is assumed that the f-response is
added to the a-response when the luminance of
the sample is greater than that of the background,
and it is subtracted from the a-response when the
luminance of the sample is lower than that of the
background. Thus, we have
Se OR ws (etnias
Se L” —kie—
el" for b=,
[4(6.3)]
where S denotes the lightness or brightness index,
L and L, are the luminances of the sample and
background, respectively, and k is a constant that
controls the proportional effect of the a and
B-responses on the total response S. The expo-
nents m and n govern the a-response and f-re-
sponse, respectively.
Semmelroth showed that Eq. 4(6.3) describes
Takasaki’s experimental data, illustrated in Fig-
ure 4(6.3), quite well when he chooses m= n=k
= 0.2. The luminances L and L, are replaced by
luminance factors Y and Y,. There is some flexi-
bility in the choice of the appropriate values for
m,n, and k. When other experimental data are to
be fitted in addition to those of Takasaki, the
values m = 0.4, n = 0.2, and k = 0.65 are found
to serve best.
The studies by Kaneko (1964), Takasaki (1966),
and Semmelroth (1970) draw attention to the
rather limited applicability of the earlier light-
ness-index functions compiled in Table 1(6.3) and
illustrated in Figure 2(6.3). For example, the
Munsell renotation value scale was intended to
apply best for a middle-gray background; yet, the
function given by Equation (3) in Table I(6.3)
does not exhibit anywhere the effect of crispen-
ing. If the Munsell gray scale would apply to a
fixed gray background, it ought to show a local
expansion near that fixed gray. Lacking such an
expansion, we may ask whether the Munsell gray
scale could be thought of as referring to appraisal
of lightness differences with a gray background of
Munsell value intermediate to those of the two
gray samples being compared. Such a background
of sliding luminance factor implies that, for exam-
ple, two samples of Munsell value 1/ and 2/
viewed on a background of Munsell value 1.5/
would appear just as different in lightness as two
other samples, say Munsell value 8/ and 9/,
viewed on a background of Munsell value 8.5/.
Semmelroth (1971) has tested this possibility
by using his Eq. 4(6.3) with m= 0.4, n= 0.2,
k = 0.65. First, he finds that for a background of
fixed luminance factor, the Munsell renotation
value scale yields nonuniformities in lightness
scaling by factors exceeding 3.0. When he intro-
duces a background of sliding luminance factor,
that is, Y, is maintained intermediate to the lumi-
nance factors of the two grays whose lightness
difference is to be judged, the situation does
improve. But there are still nonuniformities by
factors of about 1.5. A further improvement to
factors less than 1.3 is obtained when the lumi-
nance factor of the background is maintained
equal to, or slightly less than, the intermediate
luminance factor of the two gray samples. The
residual errors result from the fact that Eq. 4(6.3)
takes account of the observed crispening effect,
whereas Equation (3) in Table 1(6.3) does not.
Semmelroth (1971) has provided a table which
can be used to make adjustments to the Munsell
renotation value scale so as to make it apply more
accurately to uniform lightness steps for back-
grounds of different luminance factors. The same
table can be used to make equivalent adjustments
to the CIE (1976) L*-scale by recalling that
Munsell value V is approximately 0.1 L*.
Judd and Wyszecki (1975) converted Semmel-
roth’s table into a graph shown as Figure 5(6.3).
The abscissa V; denotes the Munsell value of gray
samples determined by the Munsell renotation
value formula and may be called the nominal
Munsell value cf the sample. The ordinate V,
denotes the background-adjusted Munsell value
so as to correspond more closely to Semmelroth’s
lightness index S[Eq. 4(6.3)]; in fact, Semmelroth
gives the relation
V, = 0.575421S — 0.658646 [5(6.3)]
The background Munsell value V, as a function
of the nominal Munsell value V; of the sample
depends, of course, on the background. This is
why Figure 5(6.3) shows a series of curves, one

Fig. 5(6.3). Background adjusted Munsell value V, plotted against nominal Munsell value V,
for backgrounds of nominal Munsell value 1/, 2/,...,9/. (Prepared by Judd and Wyszecki,
1975 from a table by Semmelroth, 1971.)
for each of nine different backgrounds. Each
background is specified by its nominal Munsell
value, denoted in Figure 5(6.3) by the numerals 1
to 9. The units of V,, are chosen so as to corre-
spond to units approximately equal to Munsell
value with respect to a background of sliding
luminance factor. Thus, we note that for Vo =
1,2,3,..., we read (approximately) V, =
1,2,3,..., respectively, for corresponding back-
erounds },2;3,. -./,.
The following examples, taken from Judd and
Wyszecki (1975), illustrate the use of Figure 5(6.3):
EXAMPLE |
The lightness difference between samples of
Munsell value V; = 2 and V;=3 on a back-
ground value 7/ is (approximately) AV, = 1.65 —
0.7 = 0.95 units. The corresponding difference
between the same two samples viewed on a back-
ground of value 3/ is AV, = 3.0 — 1.2 = 1.8 units,
Brightness and Lightness Scales 499
that is, almost twice as large as on the value 7/
background.
EXAMPLE 2
A sample of Munsell value V; = 3 viewed on a
value 5/ background has a lightness index V, = 2
(approximately). The same lightness index V, =
2 is obtained for a sample of Munsell value
Vs = 2 on a value 2/ background.
EXAMPLE 3
The crispening effect is quite evident in the region
where V; is about equal to the background value;
in that region, each curve has its inflection point.
For example, increasing V; by half a Munsell step
from 5/, for a background of 5/, increases V, by
6.35 — 5.0 = 1.35 steps. This is more than double
the increase caused by V; increased (9.9 — 9.4 =
0.5) by half a step from 8/, with the same back-
ground value 5/.
 500 UNIFORM COLOR SCALES

6.4 COLOR SCALES OF CONSTANT

LIGHTNESS

Uniform scaling of the lightness of object-color

stimuli perceived to have approximately the same
hue and chroma is a specially simple case because
a single variable only is involved. With more than
one variable, the possibility of uniform scaling
depends on the observer’s judgments satisfying a
further requirement. Consider all object-color
stimuli having a particular fixed lightness; their
representative points, defined in terms of tn-
stimulus values, in the object-color solid will gen-
erate a surface that is not necessarily contained in
a plane. Two pairs of stimuli belonging to this
surface that are judged by observers to present
the same color difference will not generally be Fig. 1(6.4). Triangular lattice used for uniform
represented by pairs of points separated by the scaling of colors of constant lightness.
same distance, whether this distance is taken as
the length of the straight line joining the two
tem as shown in Figure 2(6.4). The central point
points of the pair or as the length of the shortest
represents a gray; points on a circle represent
joining line lying wholly in the surface (geodesic).
object-color stimuli of constant chroma. All circles
The question is whether by a systematic distortion
are centered at the gray and evenly spaced corre-
of the tristimulus space, that is, by a one-one but
sponding to uniform steps in chroma. Points on a
not necessarily linear transformation, the separa-
radial line starting at the center represent stimuli
tion in the transformed space of all pairs of
of the same hue. Equal angles between these
points in the equilightness surface representing
radial lines correspond ideally to equal steps in
object-color stimuli judged to show a given fixed
hue. A complete color solid is constructed from
color difference can be made the same. If this can
color scales of this kind at different lightness
be done at all, then the transformed equilightness
surface can be made a plane covered by a net-
work of equilateral triangles, each lattice point of
the network representing a stimulus showing the
same apparent color difference with each of its six
nearest neighbors, as illustrated in Figure 1(6.4).
The set of lattice-point stimuli then provides an
ideal uniform scaling of the equilightness system.
Present evidence is that observers’ judgments of
the color difference do not satisfy accurately the
requirements for constructing such ideal uniform
scaling systems; the geometry of color-perception
space is not Euclidean. However, the failure of

Seip”
the Euclidean property is not so serious as to rule
out practically useful approximations of uniform
scaling. The color-order system developed by the
Committee on Uniform Color Scales of the Optical
Society of America (see Section 6.6.4) provides
examples of good approximations to uniform color
scales of object-color stimuli of constant light-
ness. Fig. 2(6.4). Polar-coordinate system. Points on a
Another representation of color scales of con- circle represent object-color stimuli of constant
stant lightness is in terms of loci of constant hue saturation, points on a radial line starting at the
and constant chroma on a surface of constant center represent object-color stimuli of the same
lightness; these loci form a polar coordinate sys- hue.
 Color Scales of Constant Lightness 501

levels ranging from black to white. An example of
such a color solid is provided by the Munsell color
system (see Section 6.6.1)
Given that the Euclidean property of color
perception is approximately valid, a color scale
based on a polar coordinate system represented
by uniformly graduated loci of constant hue and
chroma can be converted very simply to a uni-
form color scale defined by a regular triangular
lattice system. The triangular lattice is superim-
posed onto the polar coordinate system, and
points of the lattice are expressed in terms of the
polar coordinates by interpolation. Wyszecki
(1954) has carried out such a transformation for
the Munsell color system. Figure 3(6.4) is an
example of the triangular lattice sampling of col-
ors of Munsell value 5/. Each lattice point is
identified by the proper notations of the Munsell
system.
When an approximate uniform scaling of hue
and chroma has been set up by selecting, on the
basis of observer judgments or otherwise, the
particular material color samples that realize
the scale graduations, the CIE 1931 tristimulus
values of these samples, under the assigned
illuminant, can be determined. Analytical expres-
sions can then be sought that transform—nonlin-
ear transformations must be allowed—CIE tri-
stimulus values to three new variables which, used
as rectangular coordinates, define the required
distorted space. This is the space in which equi-
lightness samples occupy a common plane, and a
given color difference between any two samples
in an equilightness plane corresponds to the same
distance separation of their representative points.
With the Munsell system taken as basis, a number
of attempts at acceptable but fairly simple ana-
lytical transformations have been made (Adams,
1942; Moon and Spencer, 1943a; Saunderson and
Milner, 1944; 1946; Godlove, 1952; Nickerson,
Judd, and Wyszecki, 1955; Farnsworth, 1958;
Glasser, McKinney, Reilley, and Schnelle, 1958;
Taguti and Sato, 1962; Fukuda and Fuju, 1965).
Some of these transformations are given in Table
I(6.5.1) in connection with three-dimensional color
scales. Perhaps the most popular transformation

a i.
au ee ree! iM
Sei alate
SE ES aLe
Kor
a
oneal] renee |ese | eee
| 28872 |5386/53 |47618 | Tpev/f2 | 336/100
4 6.78/80 | 548/53 | Toca |BBOY/ES |330/80 |
| at 6.7B/60 | 8386/55 | 176/35 | 33GY /6.0 |
eR a ee a a | oe et
2 |20P8/106 | 05PB/7.2 | 6.78/40 | 106/20 | 33GY/£0 | 95Y/72
ST ene | ee ag Ta OO Ge a a |
| 3pB/i2.2 | 3.26/82 | 2.0PB/5.3 | 6.78/20 | 330/20 | gtv/sa |
ay | ee eles SAE PZ Sl Se) ee Sn “
0 |5.0PB/ 13.9 |5.0PB/10.4 | 5, 0PB/69 | 5 OPB/3.5 | N/0.0 | 50Y/35 | 50v/69 |
Pane ee ano Cure een. eg Sees SE ae) Pe. Ke
0PB/15.6 | 638/122 | 6. 8PB/8.7 | 8.0PB/5.3 | 33P/2.0 | 6.7YR/2.0 | 20v/53 | 32¥/87
a Eee ited AW Ay Oaks cane | EON ee bess Ue Cee oe eS
—? | 7.3PB/ 14.0 |8.0PB/10.6 | 9 5PB/72 | 33P/40 | 1ORP/Z0 | BIYR/40 | OBY/72 |
le BA ed eS hase |e Sh Da dy SD TE 1 OOM Oregtt
q 8.08P/18 9 |8.9BF /125 |OP |33P/60 |LRP AS. |BARS |BIRD |BRA
-4 | oat | 09P/1i1 | BaP 82 | ser/53 |-JorP/4.0 |LARS |BIRO |ee
i | 33P/100 | Deriie |AR 63 |SIR/SS |2/72 |BR/100
-6 23P/1Z0 |BMPS |TpRPIES |RP 6 | Sv6e | aps2 | >
a |BPEL | s9/6_| 6 )RP/T 2 Bua |oeréy_ |anya
a Seen |BBP 105 | 3.5RP/87 |_Tone780 |BRIT |1avR/10.6|i
| FarrrZs | uRe/Ga |rpweKe | s0rs2 | adenda |28/25
-10 ee arn a |a7RP106 |Towerida | S3R/166 |S8R/22 |
3.0RP 12.2 |7.5RP/TL.A | 2erni | 7R/162 |
=i? ve |5.5RP/12.5 |_JorP/20 | ASR/NAS | Se
a |> | 2:R/151 |SARIS
mas goo BRIE |

—9 —8 —J] —6.-5. —45—<3 —2 =-1 0 ] eS. 4 5

p

Fig. 3(6.4). Triangular lattice sampling of object-color stimuli of the Munsell color system for
Munsell value 5/ (from Wyszecki, 1954).
 0.2

0.1

CIE 1960 UCS

Diagram ea

0.4 0.5 0.6

CIE I976.UCS
Diagram

0.4 0.5 06

Fig. 4(6.4). CIE 1960 and CIE 1976 UCS diagrams.

502

of this kind is the one providing the Adams-chro-
matic-value-diagram (Adams, 1942; Nickerson,
1947; 1950; Vickerstaff and Tough, 1949;
Godlove, 1950; Stearns, 1951; Billmeyer, 1963;
Bridgeman, 1963). That particular transformation
played an important role in the development of
the CIE 1976 (L*a*b*)-space (Sections 3.3.9 and
6.5.1).
In a slightly different approach to uniform
scaling, all those color stimuli, not necessarily
object-color stimuli, are considered that lie in a
plane of constant luminance of the CIE 1931
tristimulus space. Only linear transformations of
the CIE space are allowed, corresponding to pro-
jective transformations of the CIE 1931 (x, y)-
chromaticity diagram, and it is sought to make
equal displacements in the transformed chro-
maticity diagram correspond to perceptually equal
color differences. This objective is known to be
impossible in any strict sense, but the simplicity
of linear or projective transformations makes even
a crude approximation of practical value. The
general form of the transformation is as follows:
, —_ 41x + 4iy + Qy3
a3)X + ayy + 33
[1(6.4)]
, 2 Ae + Ag y + Ay;
a3\X + Ax) + a33
where x, y are the CIE 1931 chromaticity coor-
dinates and x’, y’ the coordinates of the new
chromaticity diagram which is called a uniform-
chromaticity-scale diagram (= UCS diagram)
when the coefficients a;, are properly chosen.
In the first attempt to develop a UCS diagram
(Judd, 1935a), the choice of the coefficients a;,
was based on a variety of data on chromaticity
spacing published prior to the Munsell data,
including chromaticity spacing of the monochro-
matic stimuli, spacing of chromaticities just-
noticeably different from some central nearly
neutral chromaticity, spacing of chromaticities of
constant dominant wavelength, and spacing along
the Planckian locus.
The usefulness of a UCS diagram led several
other investigators to try either improvements on
Judd’s UCS diagram or new UCS diagrams based
on different experimental evidence (MacAdam,
1937; Breckenridge and Schaub, 1939; Hunter,
1942; Farnsworth, 1944; Neugebauer, 1949;
Sugiyama and Fukuda, 1959; Committee on Col-
orimetry, 1963). Table 1(6.4) lists a number of sets
of coefficients a;, of Eq. 1(6.4) proposed by vari-
Three-Dimensional Color Scales 503
ous investigators to represent approximately
uniform chromaticity spacing. MacAdam’s UCS
diagram [Number 5 in Table I(6.4)] was adopted
by the CIE in 1960 as a standard UCS diagram,
referred to as the CIE 1960 UCS diagram whose
coordinates were given the designations u and v.
However, on the basis of more recent experimen-
tal evidence, the CIE in 1976 modified its 1960
UCS diagram to a slightly improved version with
coordinates
us—wyand vo = 150
thus leading to the transformation coefficients a;,
given as Column (9) in Table 1(6.4) (see also
Section 3.3.9). Figure 4(6.4) illustrates the CIE
1960 as well as the CIE 1976 UCS diagram.
6.5 THREE-DIMENSIONAL COLOR SCALES
6.5.1 Principles of Construction
In many industrial color control problems, color
differences have to be judged that are a combina-
tion of differences in lightness, hue, and chroma;
to assess these, a uniform three-dimensional color
scale is needed. The logical extension of the regu-
lar triangular lattice arrangement in the plane,
shown in Figure 1(6.4), is to a regular rhombo-
hedral lattice arrangement in space (Wyszecki,
1954). In a regular rhombohedral lattice sampling
of the color space, each color is surrounded by 12
nearest neighbors, all equally distant. The 12
points representing these colors form a poly-
hedron called a cubo-octohedron for it is ob-
tained by cutting off the eight corners of the cube
by planes through the midpoints of each three
concurrent edges. Figure 1(6.5.1) illustrates a
cubo-octahedron.
The regular rhombohedral lattice arrangement
provides not merely a regular array of colors but
also the most closely packed arrangement. The
lattice can be resolved into a series of parallel
plane lattices in seven different ways. In four of
these, the meshes of the plane lattice are equi-
lateral triangles, and in the remaining three, they
are squares. Figure 2(6.5.2) shows the seven dif-
ferent ways of cutting the regular rhombohedral
lattice, the typical cell being the cubo-octahedron
illustrated.
A spatial arrangement of color samples such
as the Munsell color solid, that has, or aims to
have, the property of uniformity with respect to
504 UNIFORM COLOR SCALES
Fig. 1(6.5.1).
judgments of magnitude of color differences com-
bining lightness, hue, and chroma differences, can
be respecified as a regular rhombohedral lattice
system. In the case of the Munsell solid, position
in the uniform space is defined by a cylindrical
coordinate system, the coordinates being Munsell
hue, Munsell chroma, and Munsell value, respec-
tively. If a regular rhombohedral lattice is in-
serted in the Munsell space (a convenient lattice
spacing and lattice position and orientation can
be chosen), the cylindrical coordinates (the
Munsell notations) of the lattice points can be
determined by interpolation between the known
points of the Munsell system (Wyszecki, 1954).
With the proper choice of lattice orientation, each
plane of constant Munsell value will contain a
triangular lattice as discussed earlier [see Figure
3(6.4)]. A direct determination of a regular
rhombohedral lattice in color space has been made
by the Committee on Uniform Color Scales of the
Optical Society of America. The results of this
work will be discussed in Section 6.6.4.
_ The traditional way of deriving formulae for
three-dimensional color scales has been by choos-
ing a chromaticity diagram with approximately
uniform spacing for equilightness (or equilumi-
nance) color stimuli and by adding to it some
function of luminance factor which gives ap-
proximately uniform lightness spacing for color
stimuli of the same chromaticity. Suitable scaling
factors for both the chromaticity scale and the
lightness scale are selected to make the unit of
Cubo-octahedron.
chromaticity difference closely equivalent to the
unit of lightness difference. Table 1(6.5.1) in the
Appendix lists a number of three-dimensional
color scales that have been used in practice. Spe-
cial weight attaches to the two CIE uniform color
spaces, known as the CIE 1976 (L*u*v*)-space
and as the CIE 1976 (L*a*b*)-space. These two
spaces, recommended by the CIE for practical
applications have been discussed in Section 3.3.9.
6.5.2 Color-Difference Formulae
For each of the three-dimensional color scales of
Table 1(6.5.1), the formula for AE represents the
magnitude of the color difference perceived be-
tween two object-color stimuli which are specified
in terms of their CIE tristimulus values. The
difference AE is the quantity of importance in
industrial color-control problems and in the
establishment of color tolerances. Values of AE
calculated from the different formulae for the
same pair of color stimuli, differ because of dif-
ferences in the basic observational data and in the
approximations used by the authors of the differ-
ent scales. The numerical relation between AE
values given by the different formulae is very
complicated, and explicit and useful mathemati-
cal expressions that convert AE values obtained
by one formula to corresponding AE values ob-
tained by another formula can, in general, not be
produced. Noteworthy is a partly graphical, partly
numerical method developed by Ohta (1977) for
 cell being a
Fig. 2(6.5.1). Seven ways of cutting a regular rhombohedral lattice, the typical
cubo-octahedron.
505
 506 UNIFORM COLOR SCALES
converting from CIE 1976 A E( L*u*v*)-values to
CIE 1976 AE(L*a*b*)-values and vice-versa.
Line elements are commonly considered a sep-
arate class of color-difference formulae. We will
discuss them in Section 8.4.
6.5.3 White. Whiteness Formulae
White is the attribute of a visual sensation accord-
ing to which a given stimulus appears to be void
of any hue and grayness. In this sense, we also
speak of “white” stimuli. The gamut of stimuli
that give rise to the perception of “whites” or
“near-whites” in color space depends markedly
on the observing conditions, but is always very
small as compared to the whole gamut of color
perceptions (Hurvich and Jameson, 195la,b;
Jameson and Hurvich, 1951; Bartleson and Witzel,
1967; Honjo, Shimada, and Nonaka, 1967; and
Hunt and Winter, 1975).
The scaling of whiteness and, in particular, the
derivation of suitable formulae for the prediction
of whiteness is of considerable importance to a
variety of industries and their products (textiles,
paper, plastics, photography, television). With ap-
plications in mind, whiteness may be defined as
the attribute of a diffusing surface under an as-
signed illumination by which its similarity in color
to some preferred or standard white is judged.
Many formulae for whiteness so defined make
direct use of color-difference formulae.
Whiteness has also been defined as a simple
function of the luminance factor of the surface or
of its reflectances measured for one or more nar-
row regions of the visible spectrum.
Despite many attempts, no general agreement
has yet been reached on the most useful formula
for specifying whiteness for practical purposes.
Visual estimates of whiteness are highly subjec-
tive, and different observers disagree widely in
their estimates; also a single observer may change
an estimate on different occasions, quite fre-
quently by large amounts. Different observers
have, in addition, different ideas of the ideal
_white that should be taken as the standard for a
particular product. The widespread use of blue
fluorescent dyes has aggravated the practical
problem.
There is an extensive literature on whiteness
and its specification for practical purposes. For
the interested reader, the following is a selection
of references from a total of several hundred
papers that have been published in this field:
MacAdam (1934, 1955); Judd (1935b, 1936a);
Harrison (1938-1939, 1942); Asmussen and
Buchmann-Olsen (1949); Selling and Friele
(1950); E. Allen (1957a,b, 1959); Dearth (1957);
Hunter (1958, 1960, 1962, 1975); Berger (1959);
Croes (1959); Friele (1959); Fukuda (1959);
Dearth, Shillcox, and Van den Akker (1960);
Fukuda and Sugiyama (1962); Grum and Patek
(1965); Seve (1966); Vaeck (1966); Stensby (1967);
Mori (1969); Anders and Daul (1970); Ganz
(1972, 1976, 1979a,b); Thielert and Schliemann
(1973); Berglund and Stenius (1979); Fago de
Mattiello and Lozano (1979); Seve (1979); Stenius
(1979).
The work by Ganz (1972, 1976, 1979a,b) is
especially recommended as an in-depth treatment
of whiteness specifications as applied to industry.
Seve (1979) provides a comprehensive bibliogra-
phy on whiteness of about 300 references cover-
ing the period until 1976.
Table 1(6.5.3) in the Appendix lists a number
of whiteness formulae suggested by different in-
vestigators.
6.6 COLOR-ORDER SYSTEMS
A color-order system is a rational method or a
plan of ordering and specifying all object colors
or all within a limited domain by means of a set
of material standards selected and displayed so as
to represent adequately the whole set of object
colors under consideration.
Of the many conceivable color-order systems,
a number have been developed and are used.
Fairly detailed accounts of the more important
systems are given by Judd and Wyszecki (1975)
and Wyszecki (1960b).
All color-order systems fall broadly into three
major groups. Those in the first group are based
primarily on the principles of additive mixtures
of color stimuli. Systematic variations of the set-
tings of a Maxwell disk can be used to generate
color scales which in turn can be duplicated by
actual material standards. Examples include the
Ridgway Color System and the Ostwald Color Sys-
tem.
The principles of colorant mixtures provide
the basis for color-order systems of the second
group. The desired colored objects are developed
by compounding a limited number of pigments of
dyes in systematically varied proportions. Exam-
ples include the Plochere Color System and the
Martin—Senour Nu-Hue System. By application of
the screen-plate process of printing and by vary-
ing the screening systematically, color-order sys-

tems are obtained intermediate in character to the
two groups just mentioned. Examples include the
Maerz and Paul Dictionary of Color, the Villalobos
Colour Atlas, and the Hickethier System.
The third group of color-order systems is based
on principles of color perception. Material stan-
dards are selected to represent scales of constant
hue, chroma, and lightness, each one spaced uni-
formly in accordance with the perceptions of an
observer with normal color vision. Color-order
systems of this group are sometimes called color-
appearance systems. Examples include the Munsell
Color System.
Color-order systems, those mentioned above
and others, have usually been developed to serve
specific purposes and as such, have been used
with more or less success (see, for example, Judd
and Wyszecki, 1975). None of the existing color-
order systems can claim a universal usefulness.
However, the Munsell Color System has gained
remarkable popularity among research workers
and color technologists of many countries and
primarily for this reason has been selected for
more detailed treatment here. The DIN Color
System (German Standard Color Chart), of im-
portance particularly in central Europe, and the
more recent Swedish Natural Color System and
the OSA Color System have been added as further
examples of color-appearance systems.
6.6.1 Munsell Color System
The Munsell Color System has been realized in the
Munsell Book of Color, available through the
Munsell Color Company. There are different edi-
tions of the Munsell Book of Color exhibiting
different numbers of samples, different sample
sizes, and different finishes (glossy or matte). For
example, the glossy finish collection displays 1450
glossy color paint chips inserted in slots on 40
constant-hue charts arranged on facing pages. An
18-step gray scale is also included. Each color
chip may be removed from the chart. The chart
sizeis10 by 12 in.; the chip size 11/16 by 13/16
in.; the whole collection comes in two volumes
(13 by 12 by 2 in. each).
On each constant-hue chart, the chips are
arranged in rows and columns. It was intended
that the chips in any one row should be perceived
to have equally light colors under ordinary view-
ing conditions (daylight illumination, middle gray
to white surroundings) and that the chips in any
one column should be perceived to have colors of
equal chroma. The colors progress from very light
at the top of each chart to very dark at the
Color-Order Systems 507
bottom by steps that are intended to be perceptu-
ally equal; and they progress from achromatic
colors, black, gray, or white, at the inside edge of
the chart to chromatic colors at the outside edge
of the chart by steps that are also intended to be
perceptually equal.
Each chip is identified by a 3-part code. For
example, the notation 2.5 YR 5/10 describes the
chip with the Munsell hue 2.5 YR, Munsell value
5/, and Munsell chroma /\0. The notation of the
Munsell system, because it is based on nearly
uniformly spaced scales of color, is a useful tool
for the formulation and solution of many colori-
metric problems.
A Munsell constant-hue chart is laid out as
shown in Figure 1(6.6.1). The lightness scale is
represented by the Munsell value scale with black
denoted by 0/ and white by 10/. There are nine
grays placed uniformly in between. Colors of
constant chroma are placed on the vertical lines
parallel to the Munsell value scale. Munsell
chroma increases in steps of two (/2, /4,..., /10).
Colors placed on the horizontal lines have con-
stant Munsell value. The gamut of the colors in
the Munsell constant-hue chart is limited. Mix-
tures of chromatic pigments with black and white
pigments yield a color gamut whose boundary is
represented by the dashed curve in Figure 1(6.6.1).
Each row of colors of constant value is extended
until the next interval of two chroma steps would
extend beyond the boundary of the colorant mix-
ture gamut.
Figure 2(6.6.1) shows the organization of the
Munsell constant-value charts that can be built up
from the constant-hue charts. The 100-point
Munsell hue scale and notation are shown around
the outer circle. Colors of constant Munsell hue
are shown by the radial lines intersecting at the
center representing gray, or chroma /0. The hue
scale is built up to 10 segments of 10 hues each,
such as the segment from 1 YR to 10 YR, and the
fifth hue of each of these segments is alternatively
notated by the letters alone, thus, 5 YR is often
written simply YR. The Munsell chroma scale up
to chroma /10 is shown along any radial line in
Figure 2(6.6.1), and the colors of constant Munsell
chroma are shown by each of the five concentric
circles. It will be noted that the sampling of the
color solid near the black-white axis (chroma /0)
is, by this plan of development, much denser than
that of the high chroma colors represented far
from the axis. This is a necessary defect of any
collection of color chips organized on the radial
plan. The Munsell Book of Color does not show all
100 Munsell hues on the scale of Figure 2(6.6.1),
508 UNIFORM COLOR SCALES
value
Munsell
———~>
Fig. 1(6.6.1). Organization of the colors of constant Munsell hue in the Munsell Book of
Color.
but only 40 of them: 2.5 YR, 5 YR, 7.5 YR,.10
VALE IN OR el IY
Figure 3(6.6.1) is an oblique projection of the
Munsell color solid illustrating the arrangement
of planes of constant Munsell hue and constant
Munsell value; constant Munsell chroma is repre-
sented by cylinders with the gray scale as com-
mon axis.
The color chips in the Munsell Book of Color
have CIE tristimulus values in close agreement
‘with the specifications laid down in the “Final
Report of the OSA Subcommittee on the Spacing
of Munsell Colors” (Newhall, Nickerson, and
Judd, 1943). The 1943 report gives a revised spac-
ing of the Munsell system represented in
the earlier editions of the Munsell Book of Color.
This system is commonly referred to as the
Munsell Renotation System. The editions of
the Munsell Book of Color currently issued by the
Munsell Color Company or by the Japan Color
Munsell chroma ———>
Research Institute are intended to conform with
the Renotation System of 1943.
Table 1(6.6.1) in the Appendix is a reproduc-
tion of the table given by Newhall, Nickerson,
and Judd (1943), which contains the CIE (Y, x, y)
equivalents of the recommended Munsell renota-
tions for 40 hues and nine values (V) at every
second chroma (C) step from zero to the loci of
optical color stimuli (see Section 3.7). Figures
I(6.6.1) to [X(6.6.1) in the Appendix illustrate the
hue and chroma scales of the Munsell system for
Munsell renotation values 1/ to 9/, respectively,
in the CIE 1931 (x, y)-chromaticity diagram. The
luminance factor Y and chromaticity coordinates
x, y refer to CIE standard illuminant C. The
Munsell value scale is defined as follows:
Y = 1.2219V — 0.23111V? + 0.23951V?
—0.021009V* + 0.0008404V°
and has been tabulated in Table II(6.5.1).
=
et
ana
—

Fig. 2(6.6.1). Organization of the colors of constant Munsell value in the Munsell Book of
Color.
Munsell renotations intermediate to those
given in Table I(6.6.1) may be found by interpola-
tions from Figures I(6.6.1) to [X(6.6.1). Renota-
tions between value 9/ and 10/ (perfect white)
can be found reliably by extrapolation from the
graphs representing values 7/, 8/, and 9/. How-
ever, renotations between 1/ and 0/ are not as
easily found by extrapolation, and charts pre-
pared by Judd and Wyszecki (1956) for Munsell
renotation values 0.8/, 0.6/, 0.4/, and 0.2/ may
be used for that purpose. Digital computers have
also been employed to carry out conversions from
Munsell renotations to CIE color coordinates and
vice versa (Keegan et al., 1958).
The Munsell constant-hue lines in the CIE
1931 (x, y)-chromaticity diagram are generally
curved with the exceptions of one (10 Y) in the
yellow region and one (between 5 P and 7.5 P) in
the purple region. The curvature of the hue lines
has been documented in other experimental stud-
ies (see section 5.9) and is considered well estab-
lished, at least qualitatively. A closer inspection
of the hue lines at different Munsell-value levels
reveals that most hue lines do not retain their
position in the chromaticity diagram (Newhall,
Nickerson, and Judd, 1943). The Bezold—Briicke
phenomenon (see Section 5.9) is indicated, which
says that most color stimuli change their hue with
Color-Order Systems 509
luminance. Among the few notable exceptions, we
have again the 10 Y hue and the hue between 5 P
and 7.5 P; these hue lines remain in the same
position in the chromaticity diagram at all value
levels.
6.6.2 DIN Color System
The DIN Color System (Beuth-Vertrieb G.m.b.H.,
Berlin, Germany) is the official German Standard
Color System (DIN 6164) developed by M.
Richter and his associates (Richter, 1953; 1955;
1961). The color solid is organized in terms of hue
(DIN-Farbton), saturation (DIN-Sattigung), and
lightness (DIN-Dunkelstufe). Colors of constant
DIN-Farbton have the same dominant wave-
length or complementary wavelength with CIE
standard illuminant C as the achromatic color
stimulus. The hue circuit is divided into 24 per-
ceptually equal steps of DIN-Farbton. In the CIE
1931 chromaticity diagram, loci of constant DIN-
Sattigung are ovals around the chromaticity point
of illuminant C. These loci are the same for
all values of DIN-Dunkelstufe. The DIN-
Dunkelstufe is defined as a logarithmic function
of “relative lightness” h = Y/Y), where Y is the
luminance factor of the sample and Yo the lumi-
nance factor of the optimal color stimulus having
510 UNIFORM COLOR SCALES

Pigment
gamut

/10
of
Cylinder
chroma,
constant Plane of
constant hue
Se 5R

Chroma

Fig. 3(6.6.1). Schematic diagram of the Munsell color solid showing four planes of constant
Munsell hue (5R, 10Y, 5BG, 10PB) and a cylinder representing constant Munsell chroma
(/10). Planes of constant hue intersect each other at the Munsell value axis, which also is the
common axis of all cylinders of constant Munsell chroma. Planes of constant Munsell value
intersect the axis perpendicularly.

the same chromaticity as the sample. Constant
Dunkelstufe means constant relative lightness h
but usually not constant lightness in the sense of
perceived lightness (Wyszecki 1960b). The DIN-
Dunkelstufe function is defined as
D = 10 — 6.1723 log (40.7h + 1)
with
Be
Y
and is plotted in Figure 1(6.6.2). The network of
lines of constant DIN-Farbton and constant
DIN-S&ttigung is illustrated in Figure 2(6.6.2).
Table 1(6.6.2) in the Appendix gives values of D
as a function of h in accordance with the above
equation; Table II(6.6.2) gives the chromaticity
coordinates of the colors of the DIN-System il-
lustrated in Figure 2(6.6.2).
The introduction of relative lightness rather
than lightness itself is said to give the DIN-hue
circuit a “psychologically balanced” appearance.
It is thus not surprising that the loci of constant
DIN-Sattigung do not approach closely the corre-
sponding loci of constant Munsell chroma.
In fact, the relationship between the two color-
order systems is rather complicated (Kundt and
Wyszecki, 1955).
The DIN Color System has been realized in the
form of painted paper chips with a matte finish.
The chips are rectangular in shape and approxi-
mately 0.9 by 1.1 in. large. Samples of the same
DIN-Farbton are displayed on a single sheet;
there are 24 such sheets, one for each Farbton.
The samples progress from light at the top of each

5 o
ui 2
ae
ne
tS
“ec
=
5 <5
ie
z
Oo
0.0 0.2
DIN
Fig. 1(6.6.2). Relationship between DIN-Dunkelstufe
Relativ-Helligkeit h (= DIN relative lightness).
sheet to dark at the bottom by steps of equal
difference in Dunkelstufe (D = 1 to 8), and from
near achromatic colors (Sattigung S = 1) at the
left side of each sheet to saturated colors at the
right, the samples changing by steps of equal
difference in Sattigung until the gamut boundary
of the pigments used is approached. The samples
are held in small slots cut into each sheet and can
be removed for use in color comparisons. Each
sample is identified by DIN-notations (F': S: D)
and CIE (x, y, Y) coordinates with respect to the
1931 standard observer and source C. Also given
are dominant wavelength or complementary
wavelength and excitation and colorimetric pur-
ity, as well as corresponding Munsell renotations
(Budde, Kundt, and Wyszecki, 1955) and Ostwald
notations (Richter, 1957).
A review of the DIN Color System in its final
form issued by the Beuth—Vertrieb, Berlin as the
German Standard DIN 6164, has been given by
Nickerson and Granville (1963).
6.6.3 Swedish Natural Color System
The Swedish Natural Color System (Hard, 1970) is
an example of a color-appearance system based
Color-Order Systems 511
0.4 0.6
HreSSoe
eett
co
elece
ps
ceee
eee
ele
vee
era
Poses
RELATIV-HELLIGKEIT (h)
(Relative lightness)
D (= DIN darkness degree) and DIN
on the six psychologically unique color percep-
tions of black, white, red, green, yellow, and blue.
The German physiologist Ewald Hering is
given credit for developing the basis for this model
of color perception in the second half of the last
century and the Swedish physicist Tryggve
Johansson developed it further in the first half of
this century. The Swedish Color Center under the
direction of A. Hard has attempted to achieve a
practical realization of the Hering—Johansson
ideas of systemizing color perceptions in the form
of a color atlas (Hard, 1966).
A description of the quality of a color percep-
tion is the primary objective of this system; the
perceptually uniform spacing of colors as stressed,
for example, in the Munsell system, is given sec-
ondary consideration. The appropriate quality
attributes of a color perception are the psycho-
logically unique perceptions: black, white, red,
green, yellow, blue. A given color perception can
be specified by judging its resemblance to the
unique (also called “elementary’’) perceptions, and
not by judging differences from the unique per-
ceptions. The resemblance is expressed in degrees
of whiteness, blackness, yellowness, redness, blue-
ness, and greenness.
512 UNIFORM COLOR SCALES

\//
0.3 rity
TU

€ 4
{SS
\\ANS

a4
an aa 380 to 410
A

Fig. 2(6.6.2). CIE 1931 (x, y)-chromaticity diagram showing straight loci of DIN-Farbton
(hue) and oval loci of DIN-Sattigung (saturation). The network of lines is valid for all values of
DIN-Dunkelstufe (darkness degree). (After Richter, 1955.)

There appear to be functional relationships the object-color solid) is surrounded by 12

between degrees of whiteness, blackness, yellow-
ness, and so on, exemplified in the Swedish
Natural Color System and the color-perception
attributes of hue, chroma, and lightness as exem-
plified in the Munsell system (Judd and
Nickerson, 1975). But these functional relation-
ships are not as yet clearly understood and re-
quire further study, particularly when the color
atlas, based on the Swedish Natural Color System,
becomes more widely available.
6.6.4 OSA Color System
The OSA Color System is a color-appearance sys-
tem that exemplifies uniform color spacing by
means of the regular rhombohedral lattice
arrangement of color samples [see Section 6.5.1,
especially Figures 1(6.5.1) and 2(6.5.1)]. In this
system, each color (not lying on the boundary of
neighboring colors, all perceptually equally dis-
tant from the given color. The arrangement is
such that any given color appears in seven en-
tirely different series of equally spaced colors,
with each series traversing the color solid.
The OSA Color System developed by the Com-
mittee on Uniform Color Scales of the Optical
Society of America between the years of 1947 to
1974 is available in atlas form from the Optical
Society of America in Washington, DC. The paint
chips contained in the atlas closely rep-
resent, when viewed under daylight (D,;) il-
lumination and on a middle-gray surround (30%
reflectance), the colors of the OSA Color System
as specified by CIE 1964 color coordinates given
in Table 1(6.6.4). Figures 1(6.6.4) to XIII(6.6.4)
illustrate the plane square-lattices of the system
for constant OSA lightness L = —7 to +5 in the
CIE 1964 (x19, ¥j9)-chromaticity diagram.
 Color-Order Systems at

All colors of medium lightness are represented where

by points on the plane square-lattice L = 0. The
planes parallel to and above L = 0 are denoted ye
by L=1,2,---; those below are denoted by C= =
L = —I, —2,---. There are only a few points on 5.9(¥1v3 —2)
the planes L = 5 and L = —7 representing colors oo 1/3
that are producible by stable paint mixtures. —
=1+0.042 (Yio=1/3 —30)
10
2
In each plane of constant lightness (L), lattice 10 3
coordinates (j, g) are used to identify the lattice
points. Both j and g are zero for grays. Positive
values of j, with zero values of g, indicate yellow
or brownish colors. Negative values of 7, with zero
values of g indicate blues. Positive values of g,
with zero values of j, indicate greens. Negative
values of g, with zero values of j, indicate The (Rio, Gio, Bjo)-tristimulus values refer to a
reddish-purples. fundamental system of primary stimuli whose
The lattice points (L, 7, g) of the OSA Color (X19, Yio)-chromaticity coordinates are given as
System are related to the CIE 1964 tristimulus
values (X19, Yio, Zi9) by the following formulae:
x19(R) = 0.747 yg R) = 0.253
L =5.9| 4? — 3 + 0.042(¥9 — 30)'”)|
[1(6.6.4)]
Xjo(B) =0.171 — yio(B)= 0
This is a modified Semmelroth (1970) formula
that takes “crispening” of color differences into
(see also Section 8.2.5).
account (see Section 6.3) when the gray back- In the application of the above Eqs. 1(6.6.4) to
ground has a reflectance of 30%.
3(6.6.4) it is often useful to recall the following
The quantity Yj) in Eq. 1(6.6.4) is obtained relations:
from a modified Sanders—Wyszecki (1957) for-
mula (see also Section 5.8):
— thl.
pee 10%i.io
AU fia dle p=Br Ge a ae
ae
xX
Vio = Yio(4.4934xi) +.4.3034y%) — 4.276x10Vi0 Y10 Yi0

—1,3744x19 — 2.5643 yj) + 1.8103) [4(6.6.4)]

[2(6.6.4)]
where Yi9, X19, Yio are the CIE 1964 color specifi-
cations of a given object-color stimulus. The
calculated value of Y,, refers to the luminance
factor of the gray object-color stimulus that ap-
pears equally light to the given object-color
stimulus (Yj9, X19, Yio). The perfect reflecting dif-
fuser illuminated by CIE standard illuminant D,;
yields the “white” object-color stimulus with Y,,
= 100.
The coordinates j and g are defined by
f= C(H13.7R13 +:17.7G14?.— 4BiY3)
g = C(1.7R16? + 8GIg? — 9.7B16°)
———
i
ai
The OSA Committee on Uniform Color Scales
makes no claim as to the applicability of Eqs.
1(6.6.4) to 3(6.6.4) to the evaluation of color
differences other than those materialized in the
OSA Color System. In particular, the Committee’s
final report (MacAdam, 1974) stresses that Eqs.
1(6.6.4) to 3(6.6.4) “should not be used to con-
struct a formula for evaluation of small color
differences or tolerances, because the color dif-
ferences judged by the Committee were all more
than 20 times just noticeable”’.
More specific aspects of the OSA Color System
have been treated in detail in a number of recent
publications, a selection of which is the following:
MacAdam (1974, 1978); Nickerson
1978); Wyszecki (1975); Davidson (1978); Foss
(1975, 1977,

[3(6.6.4)] (1978).
 CHAPTER 7
VISUAL THRESHOLDS
7.1 PREAMBLE
Some data on visual thresholds have already been
given previously (see particularly, Chapter 5), but
the bulk of such data, especially those having a
direct bearing on color vision, will be summarized
in this chapter by highlighting basic concepts and
giving convenient (not necessarily the earliest or
“best”) results of experimental investigations. The
literature on visual thresholds, with the early work
dating back some 200 years, is too extensive to
permit a brief quantitative summary of the sub-
ject matter. For supplementary reading on visual
thresholds, the following review articles are re-
commended: Pirenne and Marriott (1959), Pirenne
(1962c), Bartlett (1965), Cornsweet (1970),
Baumgardt (1972), Blackwell (1972), Bouman and
Walraven (1972), and Nachmias (1972).
7.2 GENERAL CONCEPTS
7.2.1 Basic Terms and Definitions
A visual threshold is usually defined by the prob-
ability p of seeing a difference in brightness, or
hue, or saturation, or a difference in some other
perceptual attribute exhibited by a given set of
stimuli. The value of p of the probability of seeing
is usually taken as 0.5 (or 50%), but other values
of p between zero and unity are sometimes pre-
ferred.
The above definition includes the special case
of the absolute threshold when the probability p
refers to the absence or presence of visual sensa-
tion determined at an expected point in time and
in an expected region of the visual (monocular or
binocular) field.
514
A wide variety of difference threshold studies
can be usefully coordinated by introducing the
concept of increment-threshold sensitivity. This
concept envisages the retina as a complex receiv-
ing surface with a characteristic topography
(fovea, optic disk, etc.), and is primarily con-
cerned with monocular vision. If, as is usual, the
complications produced by blinking and involun-
tary eye motions are ignored, the retinal irradi-
ance at each point of this surface is determined
by the external stimulus (see Section 2.3); that is,
by the spatial and spectral radiant power distribu-
tion of the light in the field viewed and by the
effective entrance pupil at the eye. In a typical
determination of the increment threshold, the eye
views a stimulus of given spectral radiant power
distribution which may be variable with time in a
prescribed way. This distribution provides the
so-called conditioning stimulus. At a given mo-
ment, an additional stimulus—the fest stimulus—
is exposed, and the subject reports if it is detected.
By repeated trials, using test stimuli of different
radiances but keeping unchanged all the other
characteristics (retinal location, epoch in the time
course of the conditioning stimulus, angular size,
duration, relative spectral composition, etc.), the
threshold radiance corresponding to, for example,
a 50% probability of seeing, is determined. This
increment threshold, expressed in suitable units
of radiance, and its reciprocal, the increment-
threshold sensitivity, depend for a particular eye
on two distinct groups of variables, those de-
termining, respectively, the conditioning stimulus
and the test stimulus, together with a parameter
fixing the time of application (epoch) of the test
stimulus, which is convenient to reckon in with
the conditioning variables.
The increment threshold can be determined
for almost any proposed conditioning stimulus
r
_é

and for a test stimulus of arbitrary characteristics.
The latter may be a distribution filling the whole
visual field as in some work (see, for example,
D. H. Kelly 1961, 1962), but most commonly
it is confined to a small area centered on a
particular point of the retina. Also, the exposure
time is usually brief, a fraction of a second only.
By determining the variation of the increment
threshold of such a small pulsed test stimulus
with changes in the test and conditioning varia-
bles, inferences can be drawn about the properties
of the visual response mechanisms, including color
mechanisms, that operate in a localized retinal
area.
Steady-state measurements are made when the
test stimulus is applied only after the observer’s
vision has become fully adapted to a nonvarying
conditioning stimulus. When the conditioning
stimulus is a completely dark field, except per-
haps for fixation points of minimal intensity, the
increment threshold coincides with the absolute
threshold. If conditioning and test stimuli are
both uniform light patches of the same spectral
composition of radiant power, the test patch being
small and applied well within the more extended
conditioning patch, then the measurement reduces
to a form of /uminance discrimination with the
increment threshold equal to the /uminance-dif-
ference threshold. From this situation, the two-color
threshold method is developed by allowing differ-
ent spectral radiant power distributions for test
and conditioning stimuli; with a test stimulus of
given form, size, and pulse duration, the principal
variables determining the increment threshold are
then the two spectral radiant power distributions
and the radiance of the conditioning stimulus.
Time-variable measurements are made when
increment thresholds are determined at different
times when the conditioning stimulus is a steady
stimulus of given spectral radiant power distribu-
tion and usually high radiance (or luminance)
which drops abruptly to a completely dark (zero
radiance) field. Measurements of this kind lead to
dark-adaptation curves. The increment threshold
is also used to follow variations in retinal re-
sponse during and after abrupt changes in the
reverse sense, that is, low to high level condition-
ing. When very brief test pulses are used, varia-
tions with very short time period can be traced.
The obvious extension of the two-color threshold
method to such light-adaptation curves, has also
been used (Boynton, 1956).
For almost any conditioning stimulus, the
variation of the increment threshold with size,
General Concepts 515
shape, and length of pulse of the test stimulus
provides one of the principal means of studying
the spatial and temporal integrating properties of
the response systems in a particular retinal area in
the condition to which they are brought by the
given conditioning stimulus. A special case of
some interest arises when the test stimulus is
presented to the same retinal area as homoge-
neous double pulses which may be perceived as
two flashes separated by a definite dark interval
or, instead, as a single flash that appears to vary
from bright to dim and back to bright (Boynton,
1972).
The measurement of increment thresholds at
very low radiances of the conditioning stimulus,
and in the limit, the measurement of absolute
thresholds, demands particularly severe working
conditions. Dark adaptation for 30 min to 1 hr
are normal, but the time period must be extended
substantially if, just prior to the start of the
session, the observer’s eye has been exposed to a
high-luminance environment. An effective tech-
nique sometimes used (Pugh, 1976) to arrive at a
controlled dark adaptation level is to keep the
observer in the dark for 5 min, followed by a
30-sec exposure of his or her eye to a 6.25 log
troland “white” bleach; absolute thresholds are
then measured 5 to 10 min later.
The collection of data leading to absolute
thresholds or increment thresholds with a condi-
tioning stimulus at a very low radiance level is
done best by the constant-stimulus method with
randomized stimulus presentation. The observer,
after being dark adapted (or adapted to the steady
conditioning stimulus), is instructed to regard the
fixation point(s) and to report with “yes” if, after
a warning (auditory) signal, another light is seen
in the visual field, and to report with “no” if not.
The radiance (or luminance) of each test stimulus
pulse is selected randomly from a set of radiance
levels that have been predetermined so as to give
nearly 0% of yes responses at the lower extreme
of radiance level and nearly 100% of yes re-
sponses at the upper extreme. At least seven
radiance levels are usually used in the set.
So-called “blanks” (zero-radiance stimuli) are
presented intermittently to assess the observer’s
reliability of making the required yes—no judg-
ments.
The randomized presentation of the set of
pulsed test stimuli of different radiance levels
(including the blanks) is repeated a sufficient
number of times so that for each radiance level,
the frequency of occurrence can be determined of
516 VISUAL THRESHOLDS

one of the two permissible, contrasting responses 7.2.2 Quantum Fluctuations and Visual
(e.g., the yes response). A plot of the percentage Stimuli
of flashes seen (yes responses) against the loga-
When radiant flux is incident on the photosen-
rithm (log,,.) of the radiance of the corresponding
sitive pigments of the retina, the effective stimu-
stimulus pulses leads to the frequency-of-seeing
lus is the number of quanta absorbed, and this
curve. From such a plot, the radiance can be read
number is subject inevitably to quantum fluc-
off at which the observer would be estimated to
tuations, purely physical in origin. Since 1942 the
respond with “yes” with a frequency of p%. The
bearing of these fluctuations on vision, particu-
shape of the frequency-of-seeing curve depends
larly on visual thresholds, has been extensively
on the underlying statistical distribution. In many
studied (Hecht, Shlaer, and Pirenne, 1942; De
absolute threshold experiments, frequency-of-see-
Vries, 1943; van der Velden, 1944). For critical
ing is approximated by the Poisson sum, and the
reviews of the subject, with ample references, see,
sigmoid- or ogive-shaped curve is nonsymmetri-
for example, Pirenne and Marriott (1959) and
cal. In increment-threshold experiments involving
Pirenne (1962c). For an elementary introduction,
higher radiance levels of the test stimuli, the
see, for example, Cornsweet (1970). The following
frequency-of-seeing curve is approximated by the
formulae are basic.
symmetrical Gaussian cumulative distribution
(i) If a beam of monochromatic light of wave-
curve (see also Section 7.2.2).
length A, measured in meters (m), and of radiant
A popular variation of the constant stimulus
power F, measured in watts (W), is incident for a
method is the double-staircase method which has
period ¢, measured in seconds (s), on an absorb-
been described in Section 5.12.3.
ing material, such as a visual pigment layer, that
The visual thresholds discussed above under
absorbs a fraction p, the chance that exactly m
the general concept of increment threshold sensi-
quanta will be absorbed is given by Poisson’s
tivity exclude, in general, the class of color-
distribution
discrimination data that quantifies the just-percep-
tible difference in some perceptual attribute such fy: xe

as brightness or hue by presenting to the observer w(m,x)= ay [1(7.2.2)]

a bipartite visual field, the two parts of which are
adjusted by the observer to be just-perceptibly with
different. The results of such experiments are also
referred to as discrimination thresholds (Section n

dali) Diwhony xed

m=0
Instead of making settings of just-perceptible
differences between the two halves of the visual
field, in some experimental procedures,: the where x= it okt Ap,= average number of
observer is required to make complete color quanta absorbed in repeated trials
matches between the two halves of the visual field c=velocity of light = 2.997925 x 108
by keeping the stimulus in one half (the compari- (m-s_')
son stimulus) fixed and adjusting the stimulus in
the other half (the test stimulus). The match is y= * = frequency of light (s~')
repeated many times (about 30) and the recorded
settings are then used to compute the standard
deviation or some other measure of the variability h=Planck’s constant = 6.626176 X 107*4
(or precision) of the equality setting. This measure (J-s)
is taken to be proportional to the color-dis-
crimination threshold. The precision of trichro- The chance that n or more quanta are ab-
matic matching offers an example of such data sorbed equals
and has been dealt with in some detail in Section
5.4. Whether the variability of the equality set-
ore) it
tings is indeed strictly proportional to the just-
p(n,x) => wm,
x) =1— > wh, =x)
perceptible difference has not been established m=—n m=0
with certainty, though some experiments seem to
suggest that proportionality does hold. [2(7.2.2)]
 General Concepts 517

Table 1(7.2.2) Probability of Absorption p(n, x) of n or More Quanta when the

Average Number Absorbed Is x. The Quantity x) Gives the Value of
x for p(n,x)=05.

Plin,z) n=1 # 3 4 5 7 10 14 20 30

Values of log 2/29

0.975 Zor? 0.521 0.432 0.378 0.341 0.292 0.248 0.212 7 0.179 0.148
0.95 0.636 0.449 0.372 0.325 0.292 0.249 0.212 0.180 (152° 0.925
0.90 0.521 0:356>* 0.300° °2:260° 0.233 0.198 0.167 0.142 0.120 0.098
0.80 0.366 0.251 0.204 0.177 0.158 O34 2. O42" = 0:095 0.080 0.066
0.70 0.240 062,940.13 1 0.113 0.101 0.085 0.071 0.061 0.050 0.041
0.60 0.121 0.081 0.065 0.056 0.050 0.042 0.035 0.029 0.024 0.020
0.55 0.061 0.041 0.033 0.028 0.025 0.021 0.017 0.015 0.012 0.010
0.50 0 0 0 0 0 0) 0 0 0 0
0.45 1.936 1.958 1.967 1.971 1.975 1.979 1.982 1.985 1.988 1.990
0.40 1.868 1.914 1.932 1.942 1.948 1.957 1.964 1.970 1.975 1.980
0.35 1.794 1.867 1.895 1,910 (3,924 1.934 1.945 1.954. 1.962 1.969
0.30 E713 1.815 1.855 1.877 1.891 1.909 1.925 1.937 1.947 1.958
0.25 16ts” 1.758" “1-810 “i839, 1.858, oo1 B82 561.902 of 918. 212932 eal 945
0.20 1.508 1.691 £159 91.996. 9) 1,820 bel Lous 1.898 1.915 1.931
0.15 1.370 1.609 1.697 1.744 1.775 1.814 1.847 1.873 1.894 1.914
0.10 1ISZ0F 1-500 ™1.615 Leite L760” VRs0s ST 841 1.868 1.893
0.08 POSOn 11.442 el572. 11.641 1. 686ppti742 11789, F85-824— “h855 1.882
0.06 2.951 1.370 1.518 1.597 1.648 R7il 1.764 1.807 1.839 1.870
0.04 2.770 1.270 1.444 1.538 1.597 1.670 731 177 1347, 1.852
0.02 2.464 1.107 1.324 1.442 515 1.604 1.679 135 1.782 1.825

log xy 1.841 0.225 0.427 0.565 0.669 0.824 0.985 1.136 1.294 1.472

ag 0.693 1.679 2.67 3.67 4.67 6.67 9.67 13267 19.67 29.67

p(n, n) 0,63289'40,594 «80.577 0.567 0.560, 0.550) 04542:105,0.536 » 5.0:53014..01524

The principal application of Eq. 2(7.2.2) is to and

the representation of frequency-of-seeing (proba-
bility-of-detection) curves in determinations of
visual thresholds. On the simplest theory, a detec-
tion occurs if, and only if, n or more quanta are
absorbed in the visual pigment and p(n, x) then
defines the probability of this occurring as a
function of the radiance of the stimulus pulse.
Sets of values of p(n, x) in the range of main
visual interest are given in Table 1(7.2.2). Figure
1(7.2.2) illustrates a sample of curves of p(n, x)
plotted against log x/x,. Molina (1942) gives val-
ues covering a much wider range.
The gradient of the curve of p(n, x) against x
or log x is frequently required, particularly in the
neighborhood of n=x or of p(n, x) = 0.5.
Sometimes, p(n, x) = 0.60 or 0.55 are also used
to define threshold radiance. We have
doe = w[(n—1),x] [3(7.2.2)]
eee)
eae: = 2.3026xw[(n — 1), x]
(log x)
[4(7.2.2)]
Some numerical values derived from Eq. 4(7.2.2)
are given in Table 2(7.2.2). The shape of a curve
of p(n, x) against logx depends only on n; a
change of x to z=cx merely shifts the curve
along the axis of log x by the amount of log c.
The above formulae for w(n, x) and p(n, x)
also apply when the incident monochromatic
beam is replaced by a beam with spectral radiant
power distribution {FS(A) dA} and the fraction
absorbed is a function of wavelength, denoted by
p(A). In the expression {FS(A) dA}, the quantity
S(A) dX defines the relative spectral radiant power
distribution, and F is a factor independent of
wavelength, which converts the relative distribu-
tion into an absolute distribution, measured in
518 VISUAL THRESHOLDS

1.0

0.8

0.6
p(n,x) 05
0.4

0.2

GiF2 80 (Bid 122R The 18 O O02 04 06 O8 1.0

log, (x/Xo)
Fig. 1(7.2.2).Probability of absorption p(n, x) of n or more quanta when the average number
absorbed is x, plotted against log, 9(x/x9) where x, is the value of x for p(n,x)=0.5.
Numerical values of p(n, x) are given in Table 1(7.2.2).

watts (W). When F is varied, the radiance of the where

beam is changed, although its spectral composi-
tion remains the same. The quantity is then de-
fined as
x= 2 [s(a)rp(a) da
= Fy
Table 2(7.2.2) Gradient of Curve of
Probability of Absorption of n or
More Quanta: dp(n, x) /d(log x)
for x =n and x= x, ( p=9.5)
y =z [S(A)Ap(A)ad
(ii) If a common beam irradiates material con-
taining N different “elements” (e.g., different reti-
[5(7.2.2)] nal receptor groups or different pigments in a
layer of mixed pigments) each with its own p(A)
function, denoted by p,(A), the chance that the
ith element absozbs n; or more quanta is given as
before by p;(n;, x;), where

dp(n, x)/d(log x)
5 + [S(A)d0,(A) dd [6(7.2.2)]
n rc=n z= Xo

The chance that at least one element absorbs its

1 0.847 0.798 minimum number n;, or more, quanta is
2 1.246 Ast
3 1.548 1.518
4 1.799 1.774 PCF) Same O pede) 4: es ee
5 2.02 2.00
6 222 2.20 [7(7.2.2)]
7 2.40 2.38
8 257 2255 where
9 raat pea
10 2.88 2.86
11 3.02 P; = DAN; x;) = p(n;, Fy;)
12 3.16 (Remaining
13 3.29 values of this Equation 7(7.2.2) depends on the absorption
14 3.42 column differ processes in the different elements being indepen-
15 3.54 from those of
dent, which is already implied in the definition of
16 3.66 column 2 by
17 34d less than 1
the p,(A) functions. The principal application of
18 3.88 percent) Eq. 7(7.2.2) is to theories of visual thresholds,
19 3.99 which assume that a response “seen” to a test
20 4.09 stimulus can be triggered off by a minimal ab-
25 4.58 sorption in two or more retinal elements.
30 5.02
100 9.18
For fixed {S(A) dA}, the curve P(F) against
log F has a general similarity with the simpler

i)
at
 Dark Adaptation and Absolute Thresholds 519

curves of the type p(n, x) against log x, but its
shape is different in some degree whatever the
value for n. However, an effective n, written nr,
can be chosen that makes the gradients of P( F)
versus log F, and p(n, x) versus log x, agree in
the neighborhood of P = p = 0.5. In the special
case when the N elements are all similar in every
respect (1, = 1>,°°:,N; y; = y2,°°+,y), the P(F)
versus log F curve is always sharper than the
p,(n, Fy) versus log F curve, and ns, is greater
than n. For example, for N = 2 andn, =n, = 6,
Ng 18 about 7. The sharpness increases with the
number of elements N, and, as shown by Brindley
(1963), for very large N, the curve approaches the
limiting form
Py(F) = 1 — e NY"/*! 18(7.2.2)]
In the general situation, the value of F for
which P(F) equals 0.5 corresponds in the visual
application to the threshold radiance F;, for the
whole set of elements. Given y,,...,yy, and
N,,.--,Ny, F, can always be calculated from Eq.
M22),
Similarly, F,,;, such that p;(n,, F;,y;) = 0.5,
defines a corresponding threshold value for the
ith element acting alone. Threshold radiance F; is
always less than any of the F,,’s, and its relation
with the latter quantities defines the “probability
summation” occurring when WN _ independent
elements act together. For further points on prob-
ability summation see Section 8.4.7, where a
Gaussian form of the probability-of-seeing curve
is taken as basis.
7.3 >DARK ADAPTATION AND ABSOLUTE
THRESHOLDS
7.3.1. Recovery of Threshold Sensitivity;
Dark-Adaptation Curves
Following adaptation of the eye to a field of high
luminance which is then abruptly extinguished,
the threshold for an extrafoveal area falls from its
value after a second or two through a very wide
range (of the order of five log units in some
conditions), and takes a time of some 40 min to
approach (within, say, 0.2 log units) its value for
full dark adaptation. By contrast, the threshold
for a foveal area drops over a much reduced range
(generally not more than two log units) and ap-
proaches its final value in a shorter time in the
neighborhood of 5 min. Figures 1(7.3.1) to 3(7.3.1)
give examples. Moreover, with a high initial

Adapting stimulus
e o 400,000 trolands
4A 38,900 trolands
#0 19,500 trolands
© © 3800 trolands
V 263 trolands

threshold
Log
microtrolands)
(in

0 10 20 30 40
Time in dark (min)

Fig. 1(7.3.1). Recovery of threshold sensitivity for extrafoveal (30°) vision. Prior to making
the measurements the eye was exposed for 2 min to a “white” adapting stimulus providing a
fixed retinal illuminance in the range 263 to 400,000 td. The “violet” 3°- test stimulus was
presented in 0.2-sec pulses. For the test stimulus to be treated as visible, the observer had to
detect in it a broad black cross (Hecht, Haig, and Chase, 1937).
 threshold
Log
microtrolands)
(in

a 10 20 30 40
Time in dark (min)
Fig. 2(7.3.1). Recovery of threshold sensitivity for foveal vision. Prior to making the measure-
ments the eye was exposed for 2 min to a “white” adapting stimulus providing a retinal
illuminance of 3,800, 19,500, 38,900, and 400,000 td. The “red” 3°- test stimulus was presented
in 0.2-sec pulses. The red color was used to reduce interference from rod vision. For the test
stimulus to be treated as visible, the observer had to detect in it a broad black cross (Hecht,
Haig, and Chase, 1937).

10.0)

ae

Ww

threshold
Log
lamberts)
micro-micro
(in

0 5 10 15 20 25 30
Time in dark (min)

Fig. 3(7.3.1). Variation in extrafoveal (7°) dark-adaptation curves obtained by different

observers after adaptation to a “white” stimulus of approximately 51,000 cd-m ” viewed
through the natural pupil. The threshold values are given in terms of micro-microlamberts (1
lambert = 3.183 x10° cd-m~ 2). Within the two extreme curves obtained, the range covered
by the curves of 80% of the 110 observers is shown stippled (Hecht and Mandelbaum, 1939).

520

ee
ee
S
ae _

adapting luminance, the extrafoveal dark-adapta-
tion curve [Figure 1(7.3.1)] is clearly divided be-
tween an initial branch in which the extrafoveal
cones have the lower threshold and determine the
observed value and a final branch in which the
rods take over. However, with lower adapting
luminances, the following phenomena may be ob-
served: (a) the cone branch is not in evidence; it
may have been traversed within a second or two
before observations began; (b) the shape of the
rod branch is modified; that is, it cannot be
regarded as the later part of the rod-recovery
curve from a high adapting luminance.
For foveal vision [Figure 2(7.3.1)], one branch
only is normally observable, which must be at-
tributed to cone vision. However, by the use of
extremely luminous adapting fields, multiple
branches attributable to different cone-response
systems have been obtained (Auerbach and Wald,
1955; Du Croz and Rushton, 1963; 1966).
The range of individual variation in dark-
adaptation curves is considerable, as indicated in
Figure 3(7.3.1).
The forms of dark-adaptation curves (ex-
trafoveal or foveal) for a given test stimulus
depend not only on the level of the adapting
luminance but also on the time during which it
has been exposed. In fact, by using different
adaptations in which the product of luminance
and duration is kept constant, dark-adaptation
curves are obtainable that actually intersect
(Winsor and Clark, 1936; Crawford, 1946). It
must be concluded that the changing state of the
rod (or cone) system during dark adaptation re-
quires more than one parameter for its specifica-
tion.
The course and range of dark-adaptation are
modified by changing the area and duration of
the test stimulus; thus, the spatial and temporal
summation properties of the retina must change
during dark adaptation, and in the direction that
summation is greater in the more dark adapted
state. Figure 4(7.3.1) shows experimental data of
Arden and Weale (1954) which demonstrate the
spatial effect.
7.3.2 Threshold Variation over the Visual
Field
For a fully dark-adapted eye, the variation over
the whole visual field of the threshold for a small
test stimulus, achromatic or chromatic, but not
red, is characterized by (a) a pronounced maxi-
mum at the fovea (rod-free); (b) indefinitely large
Dark Adaptation and Absolute Thresholds By
values in the area of the optic disk and near large
blood vessels in its neighborhood; and (c) a drift
towards higher values as the periphery of the field
is approached. Figure 1(7.3.2) gives examples of
these variations for two observers.
The height of the foveal maximum and the
character of the threshold variation at and around
the fovea are critically dependent on the wave-
length of a monochromatic test stimulus. Figure
2(7.3.2) illustrates this wavelength dependence.
The variation in the foveal area is of special
interest. For A = 435 nm particularly, a narrow
maximum is shown which is probably related to
the foveal tritanopia present in eyes with “normal”
color vision (see also Willmer and Wright, 1945;
Wald, 1967; Wright, 1971).
7.3.3 Threshold Sensitivity of Fully
Dark-Adapted Eye
In fully dark-adapted extrafoveal areas, the rela-
tive values of the threshold sensitivity (reciprocal
of the threshold value in radiant energy units) of
monochromatic test stimuli agree substantially
with the CIE scotopic luminous efficiency func-
tion V’(A) given in Table I(4.3.2) and illustrated
in Figure 1(5.7.2). For foveal vision, the Judd
modification of the CIE V(A)-function [see Sec-
tion 5.7.2(11)] provides a rough approximation to
the threshold sensitivity. An additional limited set
of foveal threshold-sensitivity data is discussed in
Section 5.7.2(vi), tabulated in Table 3(5.7.2), and
illustrated in Figure 6(5.7.2).
7.3.4 Absolute Threshold Values for Different
Conditions of Measurement
To specify the “smallest amount” of light detect-
able by the fully dark-adapted eye, it is necessary
to distinguish different conditions of measure-
ment and to indicate what precisely is being de-
termined. Following Pirenne (1962c), three main
sets of conditions may be considered.
CONDITION (1):
A very large test stimulus is presented in long
exposures, usually of many seconds, and is ob-
served with the natural pupils, the eyes being free
to move as in ordinary vision. For incandescent
lamp light (2400 K), Pirenne found as the average
for 22 young subjects that the stimulus luminance
for 50% seeing chance was
POS) SS Oi7S< 10 89 (ede a2)
 nL)
(in
threshold
Log

0
Minutes in dark
Fig. 4(7.3.1). Thresholds during dark adaptation showing the effect of size of test stimulus
(Arden and Weale, 1954). Extrafoveal (8°) vision. Curves A obtained with 2.7’-dia field; Curves
C obtained with 7°-dia field. Foveal vision. Curves B obtained with 2.7’-dia field. The threshold
values are given in terms of nanolamberts (nL); 1 lambert = 3.183 < 10° cd-m_~”.

S22

(Nasal) x ¥ = 190°
\¢ VARIABLE.
(Lower) @ ¥ = 270°
Logigl
X xl#
(Nasal) x ¥ = 190°
(Lower) @ ¥ = kee ¢ VARIABLE
.
Logiol
Fig. 1(7.3.2). Variation of the increment threshold of the fully dark-adapted eye for two
observers (Stiles and Crawford, 1937). The large maximum in the threshold of observer BHC
= near 16° is associated with a vessel radiating from the optic disk. A “white” 5’- test stimulus
was used in 0.2 s pulses.
Allowing for an average pupil area of 50 mm’,
this is equivalent to a nominal retinal illuminance
of 4.4 < 10~° (scotopic) td. Individual values
ranged from about half to double this value.
Observations by other investigators agree fairly
closely with this result. A monochromatic test
stimulus of wavelength 507 nm [where V’(A) is
maximal], scotopically equivalent to 0.75 < 10°°
(cd - m~*) of incandescent lamp light (2400 K),
would have a radiance of 5.08 - 10° '°(W- sr! -
m ~*). Pirenne also estimates the actual retinal
irradiance (allowing for pupil size, light losses in
the optic media, and refraction) to which the last
figure would correspond; he obtains the result
0.45 x 10°'? (W-m-’). For the number of
quanta per second of light of wavelength 507 nm
entering the pupil of the eye per square degree of
the large-field stimulus, Pirenne finds, at
threshold, the value 19.6 quanta per second.
CONDITION (II):
The test stimulus is again exposed for long peri-
ods, but is of angular size sufficiently small for
Dark Adaptation and Absolute Thresholds 523
‘\ ol a
Lake (Temporal) x ¥ = 10°
@ VARIABLE.
(Upper) @ ¥ = 90°
y
(Temporal) x ¥ = io
(Upper) @ ¥=90° f ® eee
xX
xX
the threshold to be determined substantially by
the total flux entering the eye for the whole
stimulus. For these conditions, Marriott, Morris,
and Pirenne (1959) found that the 50% threshold
corresponded to a flux of 100 to 150 quanta of
507-nm light entering the eye per second during
the long exposure. The eye was free to move, and
this value certainly corresponds to the extrafoveal
rod threshold.
With free eye motion and long exposures, de-
termination of a cone (foveal) threshold is not
practicable. However, for many practical purpo-
ses (such as in navigational signals), it is desired
to fix the radiant intensities of the test stimuli so
that even if the stimulus is fixated and imaged on
the fovea, it shall still be just perceptible. Values
(summarized by Stiles, Green, and Bennett, 1937)
obtained in various early investigations, indicate
that, for this to be so, a monochromatic test
stimulus of wavelength 555 nm (maximum of the
CIE V(A)-curve) must be a radiant flux contain-
ing about 20,000 quanta per second entering the
eye during the period of exposure.
524 VISUAL THRESHOLDS

AT 12° NASAL

UNITS)
(ARBITRARY
[THRESHOLD]
LOG

10° 55 0) 5°
NASAL TEMPORAL
POSITION OF TEST STIMULUS
Fig. 2(7.3.2). Variation of increment threshold in traverses through the dark-adapted foveal
and parafoveal area with monochromatic test stimuli of different wavelengths. For all wave-
lengths except 700 nm there are sharply descending wings representing rod response (Stiles,
1949c).

The practical threshold for lighthouse signals,
and so forth, is frequently fixed at rather larger
values. For steady signals of this kind, a threshold
value of about 0.5 photopic mile-candle is often
given (1 mile-candle = illuminance produced by
a point source of unit photopic candle power at a
distance of 1 mile in a completely nonscattering
or nonabsorbing atmosphere).
CONDITION (III):
Angular size and exposure time are made suffi-
ciently small for the perception of the pulse of the
 Chromatic Adaptation and Increment Thresholds
test stimulus to be dependent only on the total
luminous flux (lumens) or the number of quanta
entering the effective pupil in each pulse. Under
these conditions, absolute thresholds are obtained
that provide the most precise visual significance.
The observer fixates so that the stimulus is
imaged on a particular retinal area. In general, an
artificial pupil rather than the natural pupil is
used. Under these conditions and with a mono-
chromatic stimulus of wavelength near the maxi-
mum for rod vision (507 nm), measurements of
extrafoveal thresholds at about 20° from the fovea
have been found to range for different subjects
from about 50 to 150 quanta (Hecht, Shlaer, and
Pirenne, 1942; Baumgardt, 1959). These appear to
be fairly representative values, but both lower and
higher results have been recorded. (It may be
noted that 20° extrafoveal vision is probably not
quite the most sensitive extrafoveal area.)
For foveal vision, few comparable measure-
ments have been made. Observations by Marriott
(1963) gave for nine observers values ranging
from about 500 to 900 quanta for the foveal
threshold of a one-minute stimulus, exposed in
1.2-msec pulses provided by a source emitting
radiant flux in a narrow spectral band centered
on 550 nm. These results agree fairly well with
earlier measurements reviewed by Marriott. For a
more recent review of such measurements, see, for
example, Baumgardt (1972).
A somewhat different experimental approach
to determining the “smallest amount” of light
detectable by the fully dark-adapted eye has been
reported by Sakitt (1972). Four observers were
required to rate both blanks and very dim flashes
of light at 7° in the extrafoveal region of the
retina by using a 7-point rating scale from 0 (“do
not see anything’) to 6 (“a very bright light’).
The pulsed (16 ms) “blue-green” test stimulus
contained on the average, at the cornea, either 66
or 55 quanta.
For one of Sakitt’s observers, the distributions
of ratings were approximately Poisson distribu-
tions and the data were consistent with each
rating being the actual number of effective quan-
tum absorptions plus the number of noise events.
It was concluded that the observer was presum-
ably able to count every rod signal, that is, effec-
tive absorption plus noise.
For two other observers, the experimental data
were consistent with the ratings being respectively
one less or two less than the number of effective
absorptions plus noise. These two observers were
presumably able to count every rod signal begin-
azo
ning with two and three respectively. The fourth
observer produced “erratic” data and those re-
sults were discarded.
Sakitt estimated the fraction of quanta inci-
dent at the cornea that resulted in a rod signal to
be approximately 0.03. This was considered con-
sistent with physical estimates of effective absorp-
tion for that retinal region obtained by other
investigators (Hagins, 1955; Rushton, 1956a,b;
Barlow, 1962). A simulated forced-choice experi-
ment resulted in an absolute threshold that was
approximately 0.4 log units below the absolute
threshold normally obtained by the “yes-no”
method. This and his other experimental results
led Sakitt to conclude that observers can utilize
the sensory information they receive only if one,
two, or three quanta are effectively absorbed,
depending on the individual. Thus, the human
visual system may be able to count every action
potential or every discreet burst of such potentials
in some critical neurone.
7.4. CHROMATIC ADAPTATION AND
INCREMENT THRESHOLDS
Increment thresholds provide one way of studying
the visual effects of conditioning the eye to color
(chromatic adaptation), although the measure-
ments made have not the same direct application
to practical questions as determinations of hue,
saturation, and brightness (Section 5.12). Also,
each observation gives just one measured quantity
(the increment threshold) instead of three (hue,
saturation, and brightness of a perceived color),
and the information content is therefore less. On
the other hand, threshold methods have the ad-
vantage that the near-threshold test stimulus pro-
duces a minimal disturbance of the adaptation
being measured; some special difficulties of ob-
servation and interpretation in binocular match-
ing and other asymmetric matching methods are
avoided.
7.4.1 Two-Color Threshold Method
One principal interest has been the analysis of
different color-response systems (mechanisms) on
the same line as the familiar separation of rod
and cone systems by their differential spectral
sensitivities at threshold. The concept that under-
lies most of this work and that is exploited in the
two-color threshold method is well summarized in a
number of publications of which the following
 526 VISUAL THRESHOLDS
provide a selected list: Stiles (1949b, 1978),
Marriott (1962), Alpern (1968), Ripps and Weale
(1969), Enoch (1972), Boynton (1979).
In the retinal area on which the test stimulus is
imaged, there are several component mechanisms,
any one of which may be the means of triggering
off the response “seen” when the test pulse is
applied. Each component mechanism is conceived
of as an association of a selected proportion of
the end-organs (rods and cones) in the area, the
association being effected somewhere in the neu-
ral system through which the neural activity, ini-
tiated by light absorption in the end-organs, is
transmitted to the brain and finally to the effector
nerves of the external response. It is not assumed
that the end-organs of a component mechanism
are necessarily all of the same kind or contain just
one photopigment, nor that end-organs belong
exclusively to one component mechanism, but the
mechanisms are assumed to have sufficient in-
dependence for the idea of a particular compo-
nent mechanism acting alone to have a meaning.
It is assumed, in fact, that the response of all but
one component mechanism could be blocked
(ideally, it may not be practicable to do it), leav-
ing the remaining mechanism to respond in its
normal way.
For a given test stimulus and given condition-
ing stimulus, the observed increment threshold in
the retinal area is assumed to depend only on the
respective increment thresholds of the component
mechanisms that would be observed under the
same conditions if they were acting alone. In
particular, it is assumed that the observed in-
crement threshold will not exceed any of the
component-increment thresholds and will gener-
ally approximate the smallest of these.
Finally, each component mechanism is as-
sumed to have a characteristic relative spectral
sensitivity function, so that the effect on the
mechanism of an irradiance E, of one wavelength
A,, forming the whole or part of the test or
conditioning stimulus, is just the same as that of
an irradiance E, of any other wavelength 2, that
may be substituted for it, provided that E, and E,
are in the inverse ratio of the relative spectral
sensitivity values at these wavelengths (Stiles,
1959).
In the most frequently studied situation, the
eye is fully adapted to a uniform conditioning
field (usually monochromatic, say of wavelength
‘w) which extends over the retinal area in which
the test stimulus is applied. The spectral char-
acteristics of the test stimulus (which, as a rule, is
also monochromatic, say of wavelength A), as
well as its angular size, exposure time, angle of
incidence on the retina, are kept fixed. For given
wavelengths A,u, and given field radiance M,, (ex-
pressed, preferably in quantum units), the fest
radiance for a 50% seeing chance-that is, the
increment threshold N, (preferably quantum units)
—is determined. The increment threshold N, is a
function of the three variables A, w, M,,.
The main objectives are to see whether its
observed variations with the above quantities
(A, #, M,,, N,) could arise in the way proposed in
the basic concept, and if so, to determine the
relative spectral sensitivities and other properties
of the component mechanisms that have to be
assumed.
The experimental results are displayed in
curves of three types:
(a) Threshold versus radiance, or t.v.r.-curves
These curves show the variation of log N, with
log M, for fixed A and wp.
(b) Threshold versus test wavelength, or t.v.X.-
curves
These curves show the variation of log N, with
A (or 1/X) for fixed p and M,,.
(c) Field radiance versus field wavelength, or
f.U..-curves
These curves show the variation of log M, with
., where A is fixed and M, is the field radiance
which experiment shows to be necessary to raise
the increment threshold to some fixed multiple of
its value on zero field (M, = 0).
The application of the basic concept to the
t.v.r.-curves for an extrafoveal area with two
response mechanisms [a rod mechanism with
the spectral sensitivity V’(A) and a cone mecha-
nism—here assumed to be simple—with the spec-
tral sensitivity V(A)] is illustrated in Figure
1(7.4.1). Curves A and B represent the supposed
t.v.r.-curves of rod and cone mechanisms acting
alone. Their positions in the diagram for any pair
of wavelengths, A and pw, are determined by the
rod and cone spectral sensitivities in the way
indicated. The resultant increment threshold at
any field radiance is to be taken as the Jower of
the values of Curves A and B, with perhaps some
small correction for summation effects when these
values are nearly equal.
Experimentally, extrafoveal t.v.r.-curves are in
general agreement with this scheme, at least as
regards the shape and positioning of the rod
branch A [see Figure 2(7.4.1)]. Some deviations
 Chromatic Adaptation and Increment Thresholds 527

LOG (FIELD RADIANCE) = LOG Mz

6 4 2 0

<

=
5 6
ro)
iol

I|

_o
(a ve
+
(7p)
Lid
ar
Se
op leek
7 ©
ro)
oe

(THRESHOLD
LOG
ON
FIELD)
ZERO

700 600 500 400 nm

WAVELENGTH a OF TEST STIMULUS
700 nm 6,
20
600 ae
OS
500 ui =

Lg
400 <9
6 4 2 O =
LOG (FIELD RADIANCE TO RAISE
THRESHOLD BY 1 LOG UNIT)

Fig. 1(7.4.1). Diagram illustrating how t.v.r.-curves like that of Figure 2(7.4.1) result from the
t.v.r.-curves of two-component mechanisms, A (rod) and B (cone), whose positions with respect
to the axes of log NV, and log M,,, for any test, and field wavelengths A and p are determined by
the respective spectral sensitivity curves of the mechanisms shown in the auxiliary diagrams.

for the cone branch B are attributable to the fact
that the cone mechanism is not simple, a conclu-
sion that is fully confirmed by the foveal (rod-free)
t.v.r.-curves. A very sharp separation of the rod
and cone branches as in the example of Figure
2(7.4.1) is obtainable only by the use of a suitable
pair of different wavelengths for test and condi-
tioning stimuli (hence the name: two-color
threshold method ).
‘Application of the same experimental proce-
dure to test stimuli imaged in the foveal area
(with additional tests distinguishing rod and cone
vision) has shown the following:
(a) In the t.v.r.-curves obtained, no part or
branch is attributable to the response of the rod
system.
(b) For certain wavelength combinations (par-
ticularly for A<510 nm, p> 550 nm), t.v.r.-
curves with two, in some cases three, clearly
distinguishable branches are obtained. Figure
3(7.4.1) shows an example with two branches.
This finding is the crucial observation. The differ-
ent branches are attributable to cone-response sys-
tems with different spectral sensitivities. The posi-
tions of the branches are controlled on the same
general lines as for the rod and cone systems of
Figure 1(7.4.1).
(c) The possibility that just three cone-response
mechanisms, identifiable with “red,” “green,” and
“blue” systems of color-matching theory, would
account for foveal increment thresholds over the
whole range of values of A, p», and M,, has not
been borne out. The actual situation is more
complicated. Although the results indicate that
the foveal mechanisms may be classed as blue-
sensitive (maximum sensitivity near 440 nm),
green-sensitive (max at 540 nm), and red-sensitive
528 VISUAL THRESHOLDS

Ny
LOG

(ZERO) 9 8 7
LOG M,,
Fig. 2(7.4.1). Rod and cone branches in the t.v.r.-curve for a 1°, 0.06 s test stimulus imaged
5° from the fovea. Wavelengths of test and field stimuli 580 and 500 nm, respectively (Stiles,
1939).

(max 570 to 590 nm), it has proved necessary to
admit the following:
(i) There are three blue-sensitive mechanisms
with different absolute thresholds and with
spectral sensitivites that diverge widely at
long wavelengths (A > 500 nm). This di-
vergence makes possible their separation
by the threshold method.
(ii) For the green- and red-sensitive mecha-
nisms, some modification (a narrowing) in
the shape of the spectral sensitivity curves
takes place in going from low to high
conditioning levels [as shown in Table
1(7.4.1)]. Another complication examined
by Boynton, Ikeda, and Stiles (1964) re-
lates to the derivation of resultant thresh-
olds from component thresholds when the
latter have nearly the same value. The re-
sultant log-increment threshold may differ
from the value to be expected from proba-
bility summation of independent mecha-
nisms (see Section 8.4.7) by perhaps 0.1 to
0.2 log units, and the discrepancy may be
in the direction of inhibition rather than
summation.
7.4.2 Basic Formulae
The following formulae express the theory of
increment threshold sensitivity deriving from the
basic concept discussed in Section 7.4.1, in a
precise form suitable for quantitative application
and for future modification to meet some of the
difficulties.
 a
Chromatic Adaptation and Increment Thresholds 529

FOVEAL VISION
AY)

475nm

= 550nm

a oy BLUE - SENSITIVE
MECHANISM

aa
LOG
THRESHOLD),
(INCREMENT
Ny

GREEN-SENSITIVE
MECHANISM Ni

6a
2
LOG (FIELD RADIANCE), LOG My
Fig. 3(7.4.1). Foveal t.v.r.-curve with A = 475 nm and » = 550 nm showing two branches
attributable to blue- and green-sensitive cone mechanism, respectively (Stiles, 1961).

For the ith component mechanism, the incre-
ment threshold N,, of a monochromatic (A) test
stimulus of fixed characteristics exposed on a
uniform monochromatic (1) field of radiance M,
is given by
ment threshold on zero field (absolute threshold)
and regarded as a function of A, it represents the
test spectral sensitivity. Similarly, if the field M,,
raises the increment threshold to 10 times its zero
field value, then from conditions defined by Eq.
2(7.4.2), it follows that M,II;, = 1. The quantity

Nt m,$(M,II,,) — [1(7.4.2)]

The function §,(x) defines the shape of the t.v.r.-

curve of the mechanism and can be taken without as a function of pw, represents the field spectral
loss of generality so that sensitivity of the mechanism.
According to the basic concept, 7; and II;,
Six) =| when x = 0 correspond to the same relative spectral sensi-
tivity so that the ratio (II;,/7;,),=, equals a
f(x) =0.1 when x = 1 constant, which is closely related to the Weber—
Fechner fraction. When M, = 1/II;,, the incre-
([2(7.4.2)] ment threshold N;, equals 10(.Na: where (N;)o
is the absolute threshold, and the Weber—Fechner
Thus, if M, = 0, 7;, is the reciprocal of the incre- fraction or the ratio of the increment threshold
530
Jo
Aq
ay)
AGEL
(6S61
YISUITIAR
JO JOQUINUDAR
JO JouyooF uoTjORIy

[eUEY
(TPL)
‘S2INS)
ewaypsg
paaliag]
[eUITXe] AJTATIISUAg [eUNTXRY, AWANISUAg(%)’m 998-791)

pjoysesyy,
anbruysay,
10J0D-om],
(,
joquiss
SyIeWIIY
wisTURyooJ|
(sn[nuIns
(wu)
7
wid)
1s9

SuistuvyoayA]
pow On 8 OY}
JUASQY BIAOJ €0S 08861 (0¢)
,INI_,,9u0d oo oney
jo anjq0} | ‘xoidde Orb OEL‘7Z L'8
MOTIOA AVIAT}ISUaS
} 9°Z BOT
0b»)
/ (06S sjrun
ub oneYy
Jo anTq0} | Japso
JO Orr)©} (O8r OEL‘77
01 A[Qeqosq)
se Joy
MOTIIA AIAT}ISUdS
4 | BOT O€8‘0Z 'y pue (f~
/
(06S
yan
0b»)
‘uy oney
jo onjq0} | ‘xoidde Orb OEL‘T on
4
MOTIIA
AVIATISUSS
30]
O'p
0b»)
/ (06S syrun
,UI2IN,,
QUOD "% Ors OzS‘81 61
7 = PayIpoyy AyIsus}UT-YsTY
ayes Ors O7S‘81
..P2Y,,209 Lb CLS 06€LI 8]
AOA)yeTy (‘xeur
jb PaTjIPOY| A}ISUS}UT-YBTY
97e4s L8S Or0'LI
 Chromatic Adaptation and Increment Thresholds
for A = p to the field radiance is given by
Nin 10(Mado
10(N, _=|== | (
M, M,, LEN A=p
[3(7.4.2)]
The threshold response of the ith mechanism to
the particular test stimulus is completely specified
by the function {(x), the test or field spectral
sensitivity, 7, or II;,, and the Weber—Fechner
fraction w; at M, = 1/II;,.
To construct the t.v.r.-curve corresponding to
a particular wavelength pair (A, ») for a retinal
area containing several mechanisms for which the
above properties are specified, a log N;, is plotted
against log M,,, using the formula for each mecha-
nism to give component curves similar to A and B
in Figure 1(7.4.1).
Similarly, for t.v.A.-curves appropriate to a
field M,, log N;, is plotted against A (or 1/A) for
each component, and at each wavelength, the
resultant is taken to the first approximation as the
lowest of the component values.
Finally, for the f.v.y.-curve, corresponding to
a particular increment threshold N, (A fixed), the
value of the product M,II;,, ip? which inserted in
Eq. 1(7.4.2) gives
m§;(M,Ii,) = + [4(7.4.2)]
is determined for each mechanism. If the required
value of the product for the ith mechanism is P,,
the corresponding component curve in the dia-
gram of log M, against pu is obtained by plotting
log M,, = log( P;/II;,,) = log P; — log II;, against
uw. The resultant curve is obtained at each value of
pw as the highest of the various component-curve
values. The component curve of a mechanism
disappears from the diagram if N, has been fixed
lower than the mechanism’s absolute threshold
1/1; .
It is convenient to assume definitely at this
point that threshold and field radiances (mono-
chromatic) are all expressed as numbers of quanta
per second entering the eye per square degree of
the field occupied by the stimulus in question.
When the test and conditioning stimuli, in-
stead of being monochromatic, have extended
spectral radiant power distributions, the assump-
tion that for an individual mechanism the effects
of different stimulus wavelengths are additive
leads to a generalization of Eq. 1(7.4.2).
531
The use of quantum and wavenumber units
entails some care in the integrations involved.
Suppose the absolute spectral distribution of the
field stimulus is specified by E,, where E,, dy is
the radiant power in ergs per second (1 erg-
sec_'= 10-7 W) entering the observer’s eye per
square degree of field, the wavelength p being
expressed in centimeters. The corresponding
number of quanta per degree square is
E, ee
aa a|
(ch rE|i{fb
1
where the wavenumber (1 /p,) is expressed in cm
units. If
the integral
formal?
takes the place of the product M_II,, in Eq.
1(7.4.2).
For the test stimulus, it is convenient to put
the absolute spectral radiant power distribution
equal to {pe, dA} where e, 1s the relative distri-
bution and p is a parameter that is varied to
arrive at the threshold. For the threshold of the
ith mechanism,
nf)
replaces N,,7;,. The formula for extended spectral
distributions may then be written
>= Aximd(x)§ l=ae i 1
[5(7.4.2)]
7.4.3 Stiles’ Mean Data
Tables 1(7.4.3) and 2(7.4.3) contain mean data in
the form sketched in Sections 7.4.1 and 7.4.2 for
the scheme of foveal mechanisms summarized in
Table 1(7.4.1). These data were based on results
for four observers obtained by Stiles (1953, 1959).
They are incomplete but sufficient for the predic-
tion, on the theory, of the various increment-
threshold curves for a considerable number of
532 VISUAL THRESHOLDS

Table 1(7.4.3) Function ¢(x) Representing pulsed (0.2 sec) test stimulus wavelength A had an
Approximately the Shape of a t.v.r.-Curve angular size of 1° (circular) and was presented on
of an Individual Foveal Mechanism* a 10° (circular) conditioning stimulus.

log log ¢(.v) log x log (x) (ii) Notes to Table 2(7.4.3). The data of Table
2(7.4.3) are illustrated in Figures 2(7.4.3), 3(7.4.3),
3.00 0 2.80 1.67
3.05 0 2.85 1.65
and 4(7.4.3).
3.10 0 2.90 1.63 The quantity tabulated for each mechanism
Bal 0 2.95 1.61 represents the logarithm of the reciprocal of the
20 1.99 T.00 1.59 field radiance of a monochromatic field that raises
3125 1.99 1.05 Pst the increment threshold for a given mechanism to
3.30 1.98 1.10 155
10 times its value on zero field, the retina being
3.35 1.98 fi 52
3.40 1.98 1.20 1.50
assumed always to be fully adapted to the field.
3.45 1.97 125 1.48 The radiance of the uniform monochromatic
8.50 1.96 1.30 1.45 adapting field is expressed in terms of the number
3:55 1.96 1.35 1.43 of quanta per second entering the observing eye
3.60 1.95 1.40 1.40 per square degree of field.
3.65 1.94 1.45 1.37
The values given are derived from the results
3.70 1.94 1.50 1.34
5.05 1.93 1.55 1.32 of experimental work described by Stiles (1953)
3.80 1.92 1.60 1.28 and given in his paper in graphical form. Values
3.85 1.92 1.65 1.26 in parentheses are interpolations; values enclosed
3.90 1.91 1.70 k22 in boxes were obtained by indirect methods in-
3.95 1.90 1.75 1.19 volving special assumptions and are correspond-
2.00 1.89 1.80 te fe
ingly speculative (see further below).
2.05 1.88 1.85 1.12
2.10 1.87 1.90 1.08 (a) Test Stimulus. 1° (circular), 0.2 sec pulse,
215 1.86 1.95 1.04
monochromatic of wavelength A.
2.20 1.85 0.00 1.00
2.25 1.84 0.05 2.96 (b) Field. 10° (circular), continuously exposed
2.30 1.83 0.10 2.92
with pulsed test stimulus superimposed at the
235 1.82 0.15 2.88
2.40 1.80 0.20 2.83
center; monochromatic of wavelength pw, or mix-
2.45 1.79 0.25 2.79 ture of a main monochromatic field of wavelength
2.50 hae 0.30 2.74 pw with an arbitrary field (filtered light providing a
2:55 1.76 0.35 2.70 narrow spectral band).
2.60 1.74 0.40 2.65
2.65 Te7Z 0.45 2.61 (c) Observation. Test stimulus centered at the
2.70 1.70 O05 ies 2156 fovea; equilibrium condition. Increment thresh-
OR 1.69 olds were derived from frequency-of-seeing curves
(50% thresholds), or from a reduced form of this
Note. For log x >0.5, log ¢(x)=1.06—log x.
method using a series of test stimulus exposures
of increasing and decreasing radiance, or by hav-
ing the observer set the test stimulus radiance to
cases. Several new studies by different investiga-
the threshold value. No essential difference in the
tors have been reported in more recent years;
results given by the different methods was ob-
some of these will be brought out in the next few
served except that threshold values set by the
sections dealing with specific aspects of the cone-
observer were generally rather higher.
response mechanisms.
The following notes supplement the data given (d) Observers. Three females, aged 20 to 30,
in Tables 1(7.4.3) and 2(7.4.3). and one male, aged 51.
(i) Notes to Table 1(7.4.3). The data of Table (e) I15,-Field Sensitivity. The values of II.,
1(7.4.3) are illustrated in Figure 1(7.4.3). It is were determined for each observer in two blocks
noted that small differences, probably not signifi- (A and B) of measurements. In Block A, a test
cant, in the shapes of mean experimental curves stimulus of wavenumber 15,000 cm~! (A = 666.7
of different mechanisms (Stiles, 1953) have been nm) was used throughout with monochromatic
averaged out to give a mean function §(x). The fields of wavenumbers 14,500 to 20,250 cm! in
 Chromatic Adaptation and Increment Thresholds 533

Table 2(7.4.3) Field Sensitivities of Component Visual Mechanisms at the Fovea

(Stiles 1953, 1959)

Wave number

(cm?) log IN,, log IIa, log Hs,

14,250 12.060
500 (12.380)
750 14.970 12.700
15,000 (13.235) (11.010)
250 13.500 11.620
500 (13.740) 11.840
750 13.980 (10 060)
16,000 (12.170) 10.270
250 12.360 (10.370
500 12.610 10.650
750 12.800 10.910
17,000 12.840 10.990
250 12.800 10.620 13.160
500 12.850 10.980 (13.420)
750 12.910 9.240 (13.670)
18,000 11.130 10.990 (13.920)
250 11.160 10.880 12.140
500 11.390 10.870 (12.370)
750 11.599 (12.670)
19,000 11.860 (12.950)
250 10.054 (11.230)
500 70.257 11.528
750 10.424 11.756
20,000 10.528 11.873
250 10.647 10.031
500 10.775 10.227
750 10.922 10.345
21,000 9.051 10.499
250 9.154 10.622
500 9.237 10.652
750 9.268 10.723
22,000 9.354 10.745
250 9.410 10.870
500 9.478 10.892
750 9.454 10.903
23,000 9.460 10.880
250 9.433 10.868
500 9.372 10.778
750 9.296 10.731
24,000 9.232 10.595
250 9.035 10.456
500 10.854
750 10.664

steps of 250 cm™'. In Block B, the monochro-
matic fields were of wavenumbers 19,250 to 24,500
cm! in steps of 250 cm7!.
In the first run of Block A, the field radiance
at each field wavenumber was set at a value that,
on the basis of an estimate from preliminary
measurements, would raise the log (threshold) by
a fixed amount (about one log unit or rather less)
above the zero field value. In that run, the
thresholds for all field wavenumbers were mea-
sured, and from the deviation of the value at each
field wavenumber from the mean for all field
wavenumbers, an estimate was made of the change
in the corresponding field radiance that would be
expected to produce a threshold exactly equal to
the mean. This correction was made using the
relation between threshold and field radiance de-
fined by the standard shape of a component
t.v.r.-curve given in Table 1(7.4.3) and illustrated
in Figure 1(7.4.3). The reciprocals of the field
radiances so corrected provided a first determina-
tion of the relative values of II, in the wavenum-
ber range of Block A measurements.
In the next run, the field radiances were preset
at the corrected values obtained in the first run,
and then proceeding as before, a second de-
termination of the relative values of II;, was
derived.
 log
x >
Fig. 1(7.4.3). Shape of a t.v.r.-curve of an individual foveal mechanism. Numerical data are
given in Table 1(7.4.3).

@!

ol

oO!

)(field
sensitivity
log

Qu tedi PS OP (e209 aIB) Mb ntlee IS lai.

-|
Wavenumber |/n (cm)
Fig. 2(7.4.3). Field sensitivities (II,,,,112,,,113,) of the three “blue”-sensitive component
mechanisms at the fovea (Stiles, 1953; 1959). Numerical data are given in Table 2(7.4.3).

534
 Chromatic Adaptation and Increment Thresholds
ol
oO!
)sensitivity
(field
log
Wavenumber
Fig. 3(7.4.3). Field sensitivities (II,, and I14,) of the two “green”-sensitive component
mechanisms at the fovea (Stiles, 1953; 1959). Numerical data are given in Table 2(7.4.3).
At least four runs on different days, traversing
the field wavenumbers in alternating order in
successive runs, were made for each observer, and
the means of the logarithms of the relative values
at each wavenumber were computed.
Absolute values of II,,, were found by making
further threshold measurements (Block C) for
fixed test and field wavenumbers both equal to
15,000 cm~! with:
(i) zero field radiance, and
(ii) a field radiance estimated to raise the
threshold by not more than about one log unit.
The standard t.v.r.-shape was again used to de-
termine from the mean measured threshold
values, the field radiance to raise the threshold
exactly one log unit above the zero field value.
This gave the absolute value of II5, at (1/) =
15,000 cm~' and by comparison with the relative
535
/u (cm™')
value found earlier for this wavenumber, the fac-
tor for converting all the relative values of Block
A to absolute values was derived. For converting
the mean relative values for the wavenumbers
covered by the Block B measurements, the factor
used was derived by comparing the mean relative
values obtained respectively in Blocks A and B in
their common wavenumber range. The final val-
ues of log (II;,,) in Table 2(7.4.3) are the means
of the log absolute values for all four observers.
(f) I4,-Field Sensitivity. The values of II,,
were obtained by a precisely similar procedure to
that described for II.,, except that the test
stimulus had a wavenumber of 20,000 cm™! (A =
500 nm), and the factor for conversion of relative
to absolute values was obtained for (1/p) =
20,000 cm~!.
(g) IL,,-Field Sensitivity. The values of II),
were determined with a test stimulus of wave-
536 VISUAL THRESHOLDS
)sensitivity
(field
log
Wavenumber
Fig. 4(7.4.3). Field sensitivities (II; and IIs) of the two “red’-sensitive
mechanisms at the fovea (Stiles, 1953; 1959). Numerical data are given in Table 2(7.4.3).
number 23,000 cm! (A = 434.8 nm). For ob-
taining relative values of II), in the shorter
wavelength range, a fixed yellowish-green aux-
iliary field (effective wavelength around 555 nm)
was used of sufficient radiance to raise the
threshold onto the lower part of the 7,-branch of
the t.v.r.-curve. The main (monochromatic) field
had a wavenumber 18,750 cm7', increasing by
250 cm! steps, to 24,750 cm |. The radiances of
the main fields of different wavenumbers required
to increase the log threshold by a constant amount
above the value for the auxiliary field only, were
determined by the same indirect procedure used
for finding relative values of II,,, and II,,,, ignor-
ing the presence of the auxiliary field.
In a second block of measurements, complete
t.v.r.-curves were determined for monochromatic
fields (no auxiliary field) of wavenumber 14,250
cm', increasing in 16 steps, to 20,250 cm !.
For each field wavelength, the 7,-branch of the
curve was fitted with the standard t.v.r.-shape,
from the position of which along the log (field
radiance)-axis the absolute value of II,, could at
once be derived. The relative values of II,,, ob-
tained in the shorter wavelength range were then
converted in absolute values by a factor chosen to
\/p (cm!)
component
give best agreement at the join with the absolute
values found at longer wavelengths by the direct
method. The final values in Table 2(7.4.3) are
again the means for all four observers. —
(h) I1;,-Field Sensitivity. The values of II;,
were determined with a test stimulus of wave-
number 23,000 cm~' (A = 434.8 nm) throughout.
For finding IT, in the shorter wavelength range,
an orange auxiliary field (effective wavelength
around 600 nm) was used of radiance just suffi-
cient to bring the threshold onto the initial flat
range of the 7,-branch of the t.v.r.-curve, so that
the observed threshold could be taken as the
absolute threshold for the 7,-mechanism. Thus,
after determining as for II,,, the relative values of
II,,, for main (monochromatic) fields of wave-
numbers 19,500 to 24,250 cm™', in steps of 250
cm ', the factor for converting to absolute values
was readily obtained from the observed increase
in threshold above the value for the auxiliary field
only.
For longer field wavelengths, complete t.v.r.-
curves showing a well-developed 73-branch were
determined for monochromatic fields (no aux-
iliary field) provided by the green and yellow
 Chromatic Adaptation and Increment Thresholds
emission lines of a high-intensity mercury arc,
and the corresponding absolute values of II;,
deduced by fitting with the standard t.v.r.-curve.
As before, Table 2(7.4.3) gives the final means for
four observers.
(i) I1,,-Field Sensitivity. The values of II,,
were determined with a test stimulus of wave-
number 23,000 cm~! (A = 434.8 nm) throughout.
One observer had to be omitted from these de-
terminations as this observer’s t.v.r.-curve showed
no, or at most an extremely shallow, 7,-branch.
The other three observers gave 7,-branches with
an absolute threshold on the average 0.25 log
units lower than the absolute threshold for 7,. All
determinations of II,, were made by fitting the
standard t.v.r.-shape to the rather small portion
of the z,-branch visible in t.v.r.-curves for differ-
ent field wavenumbers, and the fitting errors were
materially greater than in the determinations of
II,,, and II,,. Identifying and fitting the 7,-branch
proved most difficult for short-wavelength fields,
and satisfactory mean values of log II,,, by the
direct method were not obtained in this spectral
region. The boxed-in values in Table 2(7.4.3) be-
tween 22,750 and 24,500 cm! were derived on
the following assumptions:
(i) in this spectral region, the test sensitivity and
field sensitivity of 7, stand in a fixed ratio, and
(ii) the Weber fraction of 7, has the value of
0.087 which is the same as for the other two
blue-sensitive mechanisms 7, and 73.
The boxed-in values can then be derived from the
absolute threshold for test stimuli of different
wavelengths exposed on zero field, which in this
short-wavelength range and for the three ob-
servers concerned can be identified with 7,-
thresholds. The boxed-in values are, of course,
speculative only, and in Table 2(7.4.3), represent
the means for three observers.
(j) IIS, and I1%,-Field Sensitivities. True equi-
librium conditions of observation were not
achieved and thus, the direct measurements made
of these field sensitivities were unacceptable. The
boxed-in values given in Table 2(7.4.3) were ob-
tained by a speculative and indirect method
(Stiles, 1959), which reduces in practice to multi-
plying the II,, and II,, values by respective
wavelength-dependent factors derived from the
change in test sensitivity of the appropriate mech-
anism on raising the adaptation level.
Soe
For the 7,-mechanism, the factor was taken as
proportional to the ratio of the test sensitivity for
a blue-green field (20,000 cm ') providing a reti-
nal illuminance of 1552 td, to that for a field of
similar color but providing a retinal illuminance
of 43 td (see also Stiles, 1964).
For the 7,-mechanism, the factor was assumed
proportional to the ratio of the test sensitivity for
a red field (15,000 cm™') providing a retinal
illuminance of 2506 td, to that for a field of
similar color but providing a retinal illuminance
of 29 td.
The constants of proportionality involved
in the two factors have been taken so that the val-
ues of II¢,, and II,,, are approximately the same
at long wavelengths (ie., 1/u< 17,000 cm7'),
whereas II, and II, agree approximately in the
medium-wavelength region (1.e., 1/p < 18,000
cm '), but there is an arbitrary element in these
normalizations.
7.4.4 Specific Aspects of a-Mechanisms and
Later Developments
(i) The Blue-Sensitive Mechanism a,. The 7,
mechanism is the least well established of the
three (7), 7, 7) blue-sensitive mechanisms. As
already noted by Stiles (1953), it is not in evi-
dence for every observer. As shown, for example,
in Figure 1(7.4.4), 7), in an observer for whom it
appears, is represented by the lowest branch of
the t.v.r.-curve for very short wavelengths on a
long-wave background, a branch which covers a
range of threshold values of no more than 0.3 to
0.4 log units. This range, which corresponds to
the difference in absolute thresholds for 7, and
m,, varies with the observer and it is possible that
where an observer does not show 7, the 7, mech-
anism may exist but may have an absolute
threshold equal to or higher than that of 7,. For a
test stimulus of fixed wavelength in the extreme
blue (say, A = 420 nm), the initial part of the
t.v.r.-curve for any field wavelength would be
attributable to 7, for any observer who with a
long-wavelength field (say, 7 = 600 nm) showed a
clear 7,-branch. Thus, in theory, the field sensitiv-
ity curve of 7, throughout the whole spectrum
should be determinable. Stiles (1953) determined
only the long-wavelength region of the 7, field
sensitivity curve. One of the difficulties noted was
a certain instability leading to small changes in
the relative positions of the 7- and z,-branches.
Among several later workers who have con-
firmed the appearance of a 7, mechanism, Watkins
538 VISUAL THRESHOLDS
4 how associated with a main blue-sensitive mecha-
Test stimulus \A=430nm
Field stimulus »=600nm
5 ie}
os \ (c) The photoreceptors of the 7, mechanism
Observer A
= Observer B °
é e
)threshold
(increment
log °
oO! 4 2
log ( field radiance )
Fig. 1(7.4.4). Variation of the foveal increment
threshold with field radiance for two observers.
The lowest branch of the t.v.r.-curve for observer
B is attributed to the 7,-mechanism which is not in
evidence in the data of observer A. (Redrawn from
Stiles, 1953).
(1969a,b) studied a group of normal observers
which included one observer whose original
t.v.r.-curve showed no 7,-branch, but who in later
measurements obtained a “weak” 7, contribution
(shallow branch presumably). Mollen and Polden,
working at Cambridge, have found observers who
show a clear 7,-branch, but they also report some
instability in 7, (private communication, 1977).
Despite the difficulties, a full determination of
the II,,, field sensitivity curve would be of inter-
est. The result would be of particular value,
because in the only available but incomplete de-
termination, illustrated in Figure 2(7.4.3), a curi-
ous feature of the curve appears in the range
18,500 cm! to 16,400 cm~! (540 to 610 nm), a
feature which has some resemblance to the ab-
sorption bands of haemoglobin, oxyhaemoglobin,
and cytochromes. Although this feature may turn
out to be an artifact of the original measurements,
if confirmed, it would be an interesting novel type
of visual response curve.
(ii) The Blue-Sensitive Mechanism a,. The
field sensitivity curve II, of the 7, mechanism,
shown in Figure 2(7.4.3), has a characteristic sec-
ondary lobe low down on the descending limb on
the long-wave side of the peak, for which various
explanations have been proposed:
(a) The green-sensitive 7, mechanism, or more
probably, the red-sensitive 7; mechanism is some-
nism, 73 perhaps.
(b) The photoreceptors of the 7, mechanism
contain a mixture of two or perhaps three visual
pigments.
contain just one pigment but this pigment has a
secondary absorption band around 580 nm.
It will be noted that the secondary lobe corre-
sponds to a sensitivity no more than a small
percentage of the peak sensitivity in the blue, and
0) it would probably produce very little distortion of
color-matching data. It will also be noted that the
secondary lobe occurs in much the same spectral
region as the presumed special feature of the
II, ,-field sensitivity curve referred to above.
An important investigation by Pugh (1976) has
shown that the field sensitivity II), is not an
additive spectral sensitivity such that the adapting
effect of any field with a spectral power distribu-
tion {P, du} is determined by an integral
fI1,,P, du whose numerical value S, defines the
“r-stimulus value,” a term coined by analogy to
the familiar term of “tristimulus values.” Pugh
(1976) established that for fields made up of
mixtures of short wavelengths, p < 450 nm, say,
and long wavelengths, » < 550 nm, say, the mix-
ture produces effectively a higher S,-value than
would be expected from the S,-values appropriate
to the short- and long-wavelength components of
the field’s radiant power distribution used sep-
arately. Figure 2(7.4.4) gives an example of Pugh’s
results. A more rigorous statement of this result is
as follows:
If two monochromatic fields, one of short
wavelength », and radiant power P,, the other of
long wavelength », and power P,, each produce
when applied by itself the same effect in raising
the increment threshold of the 7, mechanism,
then a mixture of the two of power P = aP, + (1
— a)P,, where 0 < a <1, produces a greater ef-
fect. To obtain the same effect, a field mixture of
radiant power P’ = kP is required, where k < 1.
Obviously, Pugh’s result favors the explana-
tion (a) above, and Pugh elaborates this to a
two-stage model of the 7, mechanism whereby the
factor by which a mixed field increases the incre-
ment threshold for 7, above the zero field value,
is the product of two factors, one determined by
absorption in the visual pigment of the receptors
with peak sensitivity in the blue, the other by
absorption in the visual pigment of the receptors
with peak sensitivity at longer wavelengths. For
the blue peak sensitivity, the field sensitivity curve
 Chromatic Adaptation and Increment Thresholds
Ooa
)log oO
(threshold
oOo
)
(threshold
log
log (field radiance )
Fig. 2(7.4.4). Upper figure: A field-mixture ex-
periment for the a,-branch in which additivity
holds. One field component is of wavelength p, =
500 nm, the other component of wavelength p,. =
430 nm. The pulsed (50 ms) test stimulus is set at
A = 435 nm and presented foveally. The observer’s
threshold was allowed to reach steady-state on the
500-nm field of a given radiance [expressed in
terms of quanta - sec | - (deg)7], and then a
series of 430-nm fields of increasing radiance were
admixed and the threshold [expressed in terms of
quanta - sec _' - (deg) 2] was measured for each
mixture. The solid line is the standard t.v.r.-curve
[see Figure 1(7.4.3)] fitted through the data points
for each field component (yu, and p,) alone. The
dashed lines are the additivity predictions, which
are in obvious agreement with the experimental
data points (triangles and squares). Lower figure:
A field-mixture experiment for the 7,-branch in
which additivity does not hold. The experimental
conditions are similar to those used above, how-
ever, with the essential difference that the wave-
length , of the first field component has now
been shifted to p, = 590 nm (~ 17,000 cm‘)
which is in the spectral region in which the II,,,
field sensitivity [see Figure 2(7.4.3)]| has its sec-
ondary maximum. The solid line is again the
standard t.v.r.-curve fitted through the data points
for each component (u, and p,) alone, and the
dashed lines are the additivity predictions, which,
this time, miss the experimental data points (trian-
gles and squares) by substantial margins. [The
illustrations are based on Pugh’s (1976) work, and
redrawn from Pugh and Mollen (1979).]
II; of the 7, mechanism suggests itself as a
possibility and raises the interesting question
whether the II, curve complies with the additiv-
ity principle in the sense used above.
539
(iii) Adaptation-Recovery Anomaly of the Blue-
Sensitive (a,) Mechanism; Transient Tritanopia.
The reduction of different cone mechanisms by
measuring the increment threshold on zero field
during recovery from exposure to high-radiance
fields was not accomplished until many years
after their separation by the two-color threshold
technique using steady adapting fields. Auerbach
and Wald (1954) first produced a kind of double
branch in a foveal threshold-recovery curve indi-
cating the operation of a “violet” cone receptor
mechanism. This was obtained after adapting the
retina to a very intense conditioning field. Before
that, Wright (1934) and Stiles (1949b) had ob-
served that immediately after adaptation to long-
wavelength fields of moderate radiance, the fovea
becomes very insensitive to short-wavelength
stimuli. Stiles (1949b) measured increment thresh-
olds for test stimuli of various wavelengths when
the retina was adapted to a moderately high level
(20,000 td) of wavelength » = 640 nm. With test
wavelengths 4 < 500 nm, the test stimuli were
detected by the 7, mechanism. However, im-
mediately on the extinction of the field, the
threshold increased by a large factor of the order
of one log unit and it was established by the
variation of this increment threshold with test
wavelength that the test pulse was being seen by
means of the green-sensitive 7, and not the 7,
mechanism which remained temporarily “blind”
for many seconds. Das (1964) substantially con-
firmed this observation. .
A quite different result was obtained by Du
Croz and Rushton (1966) who by using very high
intensity adapting levels of different colors (nar-
row-bandpass interference filters were employed)
were able to show that the recovery curves [log
(increment threshold) against time] after extinc-
tion of the field showed branches corresponding
closely to independent recovery curves of the
7, 4, mechanisms previously derived under
equilibrium conditions. In particular, they found
that after a long-wavelength (orange) adaptation,
the initial branch of the recovery curve for a
short-wavelength (blue) test stimulus was attribu-
table to 7, and the anomaly described above was
not observed. A similar, if less striking, separation
in the recovery curves of branches corresponding
to 7, and 7; was also established. This remarkable
extension of the independent mechanism con-
cepts was obtained by adaptations to fields of the
order of 10° td exposed for 20 sec. Such adap-
tation levels were known by the fundus reflection
work of Rushton (1972) to bleach nearly all the
cone pigment.
 540 VISUAL THRESHOLDS
The apparent discrepancy between the Du Croz
and Rushton (1966) results concerning the re-
covery of the 7, mechanism and those of Stiles
(1949b,c), has been elucidated by work by
Mollen and Polden (1975, 1977a) adding new
insights into what may be happening. They found
that at high, but not too high, adapting field
radiances (providing up to 10° td) of yellow light
the extinction, or even reduction of the field
radiance to one tenth, produced the anomalous
rise of the increment threshold of a short-wave-
length test stimulus. However, when very high
adapting levels were used of the order of 10° td or
more, the anomalous rise did not occur and Du
Croz and Rushton type of recovery supervened.
Figure 3(7.4.4) gives an example of the results
obtained by Mollen and Polden (1977a).
Mollen and Polden’s initial interpretation of
the anomaly, which they refer to as transient
tritanopia, is that the recoverying 7, mechanism
exerts a strong inhibitory effect on the 7, mecha-
4 ine)
ne
)
'-deg~
in
radiance
of
(threshold
log
terms
erg:s
log (trolands a oe by ne )
Fig. 3(7.4.4). Adaptation-recovery anomaly of the
-blue-sensitive m,-mechanism (transient tritanopia)
as found by Mollen and Polden (1977). The open
circles represent increment-threshold data for ob-
server JDM taken by the conventional steady-state
method. Through the data points is drawn the
standard t.v.r.-curve in two branches characteristic
of the a, and 7;-mechanisms. The solid dots repre-
sent the threshold 400 msec after the field has
been extinguished.
nism. The problem then is why this ceases at very
high-level fields. Their further results and com-
ments suggest that the change may be connected
with the switch from 7, to 7; in the steady-state
t.v.r.-curves.
The more recent work by Augenstein and Pugh
(1977) both confirms and extends some of the
results noted above. In particular, these authors
find the following.
(a) The time courses of adaptation of the 7,
mechanism to (2,-equated) short-wavelength
(u < 500 nm) and long-wavelength (1 < 550 nm)
fields are very distinct. A large and relatively
long-lasting transient threshold elevation occurs
at the onset of the long-wavelength, but not of the
short-wavelength, fields.
(b) Similarly, the time courses of recovery of the
a, mechanism from (7,-equated) long- and short-
wavelength fields are very distinct. A large and
relatively long-lasting transient threshold eleva-
tion (transient tritanopia) occurs at the offset of
the long-wavelength, but not of the short-
wavelength, fields.
(c) The wavelengths of the fields which cause
the adaptation transients coincide with those
shown previously by Pugh (1976) to combine
nonadditively with » = 430-nm fields in effecting
the adaptation of the 7, mechanism. The failure
of the time course of adaptation to be indepen-
dent of wavelength combines with the failure of
field additivity to demonstrate that signals from
the red- and/or green-sensitive cones affect the
state of adaptation of the 7, mechanism [see also
discussion of Pugh’s (1976) paper in Section
7.4.4(41), above].
(d) The adaptation transients are not observed
in the 7, and 7, mechanisms. Thus, instantaneous
signals from the green- and/or red-sensitive cones
are not the cause of the transients. Instead, it
must be concluded that the cause lies either in the
path by which those cones transmit their signals
to the 7, mechanism, or in the 7, mechanism
itself.
(ec) The recovery transient can be diminished by
adding a short-wavelength field of adequate high
radiance to a long-wavelength field that can nor-
mally cause the effect. It is suggested that the 7,
mechanism receives chromatically opponent sig-
nals.
The dynamic property of the blue-sensitive
as compared with the green- and red-sensitive
cone mechanism has been studied in an earlier
 Chromatic Adaptation and Increment Thresholds
paper by Stiles (1949c). It was found that the
blue-sensitive mechanism takes a much longer
time to become adapted to a high-radiance chro-
matic field of long wavelength. Tests with fields
of different wavelengths suggest that the process
responsible for this delay in light adaptation has
maximal sensitivity, not at the short-wavelength
where the 7, mechanism is most sensitive, but at a
much longer wavelength.
The question of increment-threshold sensitiv-
ity during the early adaptation period on ex-
posing a high-radiance field was examined by
Boynton (1956) and Boynton, Kandel, and Onley
(1959) [see also Boynton’s (1979) review]. Their
t.v.A.-curves in these early stages of light adap-
tation were resolved into component spectral sen-
sitivities differing considerably from _ those
indicated for 7 mechanisms by steady-state mea-
surements. Just as in the earlier phases of dark
adaptation, there are complicating factors still
needing to be elucidated.
(iv) Relationship of Blue-Sensitive Mechanisms
(1, 7, 7). Much more is needed to un-
derstand the connection between the three
blue-sensitive mechanisms. Originally, all the 7
mechanisms were based on increment thresholds
measured on fields to which the retina was fully
adapted. If for a particular conditioning field of
given spectral composition and for a test stimulus
of particular wavelength, the t.v.r.-curve shows
two branches assigned to distinct mechanisms, A
and B, we would like to know that when the field
radiance is such that the observed threshold lies
on the A branch, then if the response to the
stimulus of the A mechanism could be in some
way blocked, we should observe a threshold on
the B branch.
Suppose A is 7, and B is 7,, and the field is at
p = 600 nm. We can then, in general, change the
test wavelength, say from A = 475 nm to 510 nm,
without disturbing in any way the adaptation
condition, so that the observed threshold for a
particular field radiance M, which was initially
the threshold of 7,, becomes for the new test
wavelength a threshold for 7,. On the basis of
arguments of this kind, we are justified in saying
that for a// radiances of the 600-nm field both
mechanisms are present and we can draw a com-
plete branch for each.
However, this test fails for 7,, 7, and 7;
because for the test and field wavelength combi-
nations such that distinct branches of the
t.v.r.-curve corresponding to these mechanisms
541
can be obtained, the test wavelength must always
be in the short-wavelength region below about
510 nm. There, it seems probable that the relative
test sensitivities of the 7,, 7), and 7, mechanisms
remain fixed; that is, the logarithms of the spec-
tral sensitivity curves have the same shape in this
spectral range. Changing test wavelength then
merely moves the 7,, 7, 7; branches up or down
as a whole without altering their relative posi-
tions. This being so, an alternative to treating
7\, 7, 73; as distinct mechanisms would be to say
that they correspond to a single mechanism which,
when it is fully adapted in turn to each of a series
of increasing levels of the adapting field of some
fixed wavelength, undergoes at a certain level,
dependent on the field wavelength, a fairly sharp
change in threshold response properties. This
change has been attributed previously [see Section
7.4.4(ii1)] to a switch from mechanism 7, to 7,. At
still higher levels, a further sharp change occurs,
previously attributed to a switch from mechanism
a, to 7;. We could speculate that in the transition
from 7 to 7,, the mechanism jettisons its associa-
tion with the photoreceptors responsible for the
rather high field sensitivity of 7, to medium and
long wavelengths. This process involves possibly
the special feature noted in Section 7.4.4(i). In the
other transition from the 7, to the 7, mechanism,
a further jettisoning of associations with medium-
and long-wavelength receptors is taking place.
This general conception is consistent with Pugh’s
(1976) work on the field sensitivity of the 7,
mechanism, because we do not expect light ab-
sorbed in the main receptors and light absorbed
in the associated receptors to yield a truly addi-
tive field-sensitivity curve.
In a recent study, Pugh and Mollen (1979)
propose a theory of the 7, and 7, mechanisms
that accounts for a number of features of the
experimental data concerning these two blue-
sensitive mechanisms. Among the visual phenom-
ena explained by the theory are additivity failures
and transient tritanopia described briefly in Sec-
tions 7.4.4(i1) and 7.4.4(11). Pugh and Mollen
(1979) base their theory on the assumption that
under the experimental conditions which isolate
the 7, and 7,-branches of the two-color threshold,
a signal originating with the absorption of pho-
tons in the pigment of the short-wavelength cones
passes through two “sites of attenuation.” The
gain of the first site is by the short-wavelength
cones alone; the gain of the second site is de-
termined by a net “blue-yellow” opponent signal.
The two sites are distinguished by spectral sensi-
 542 VISUAL THRESHOLDS
tivity, absolute field sensitivity, and dynamic
properties. Four equations are given which pro-
vide the formal mathematical description of the
theory and whose predictions are in reasonable
agreement with experimental data. However, the
authors emphasize that their proposed theory is
most likely an oversimplification of what actually
happens in the neural network. They consider
their theory, in the first instance, to be a guide for
future experimentation.
(v) The Green-Sensitive and Red-Sensitive
Mechanisms (74, 743 75,5). The difficulties of
separating the 7, and 7; mechanisms by identify-
ing branches in t.v.r.-curves have been discussed
in detail by Stiles (1953) who gives the first evi-
dence for distinguishing low and moderately high
radiance states for both 7, and 7,;. The higher
radiance states are referred to as 74 and 7 respec-
tively, whereas 7, and 7, represent the two mecha-
nisms when their respective conditioning stimuli
are low; that is, such as not to raise the absolute
threshold of the mechanism to more than about
10 times its zero field value. At first, it was
thought that the distinction between 7, and 74 or
am, and 7; mechanisms evident in the increment-
threshold data might arise because of the bleach-
ing of the pigment in the cones. This bleaching
can modify the shape of the t.v.r.-curve if the
pigment density is sufficiently high. Figure 4(7.4.4)
shows how the t.v.r.-curve of a particular mecha-
nism, 7,, which has the standard shape [see Table
1(7.4.3) and Figure 1(7.4.3)] when the test and
field wavelengths are the same, will be modified
when different test and field wavelengths are used
if the receptors of the mechanism contain high-
density pigment which is bleached as the field
radiance to which the retina is adapted is raised.
If the test wavelength lies at the peak of the
spectral absorption curve and the field wave-
length near the “tail” of the curve, then as the
pigment is bleached, the receptor’s sensitivity to
the test wavelength will increase with respect to
its sensitivity to the field. The t.v.r.-curve will
‘then develop a dip over a certain range of field
radiances that could be interpreted as the intru-
sion of a new mechanism, with a modified spec-
tral sensitivity curve. On the other hand, if the
spectral positions of test and field wavelengths
are reversed, the t.v.r.-curve will develop an eleva-
tion over the bleaching range, a distortion quite
unlike the expected effect of a new mechanism.
Stiles (1959) rejects the possibility that the
transitions from 7, to 74 and from 7, to 7{ might
t.v.r.- CURVE of #4 MECHANISM
1/N)]
[Loc
Fig. 4(7.4.4). Effect of pigment bleaching on the
t.v.r.-curve of the a,-mechanism. Curves A, B, C,
show the shape of the t.v.r.-curve of the 7,-mecha-
nism whose cone receptors are assumed activated
by absorption in the visual pigment chlorolabe with
an effective optical density, in the dark, of 1.0 at
the wavelength (540 nm) of maximum absorption
coefficient. The fourth curve (D) (Rushton, 1963a)
shows how the pigment density is reduced as the
field radiance M, to which the retina is adapted is
raised.
be attributable to this effect, because (a) the
transition seems to occur at levels where insuffi-
cient bleaching has occurred, and (b) in one case,
ms to m3, the bleaching would provide an elevation
and not a dip. However, if the presence of rather
high densities (e.g., greater than 0.5) of pigments
in cone receptors is confirmed, some distortion of
the shape of t.v.r.-curves for the same mechanism
for different combinations of test and field wave-
lengths would have to be admitted. So far (how-
ever, see Section 7.6), in the interpretation of
t.v.r.-curves, a constant shape of the branches
attributable to individual mechanisms has seemed
adequate. This may be so, in part because the
distortion effects may be rather small, and be-
cause it is not easily possible in many instances to
follow the t.v.r.-curves over the range when the
pigment bleaching would be expected to produce
distortion. However, a re-examination of this
 Chromatic Adaptation and Increment Thresholds
whole argument with respect especially to (74, 74)
and (7;,75) 1s required. As regards the blue-
sensitive mechanisms, we know virtually nothing
about the bleaching of the visual pigment in the
receptors. Rushton’s fundus reflection methods
have not, in fact, been successful in demonstrat-
ing the presence of a bleachable “blue-absorbing”
pigment (Rushton 1958, 1963a,b, 1972).
There are few increment-threshold studies pro-
viding evidence on the existence of properties of
the duality (74, 74) and (7,, 73) in the green- and
red-sensitive systems. Watkins (1969a,b) claims
that his measurements at a limited number of test
and field wavelengths and for groups of normal,
deutan, and protan observers are consistent with
m4 and 74, 7, and 73, but with certain identifiable
differences in the shapes of the spectral sensitivity
curves for the color-defective observers. As
Watkins used only a limited number of test and
field wavelengths, it is difficult to assess how
reliable his conclusions may be.
The most significant recent contribution on
the (7,, 74) and (7,, 75) systems is that of Whittle
and Challands (1969), Whittle (1973, 1974), and
Whittle and Swanston (1974) who measured, not
increment thresholds, but suprathreshold bright-
ness matches obtained by a haploscopic (binocu-
lar-matching) method in which two test pulses (40
msec, exposed simultaneously), seen respectively
by the left and nght eyes on backgrounds, also
exposed onto the respective eyes, were matched in
brightness. Generally, a white (background) field
of 1.76 log td and a white test pulse were used for
the left eye, these being kept constant. For the
right eye, different combinations of test and field
wavelengths, A and p respectively, were used in a
complete range of different field radiances. Thus,
the procedure was analogous to the two-color
increment-threshold method except that the test
pulse (seen by the right eye) was adjusted not to
threshold but to match the brightness of a con-
stant reference pulse on a constant background
seen by the left eye. The results were curves of log
(matching pulse radiance) versus log (background
radiance) for each (A, 1.) combination, one corre-
sponding to the actual t.v.r.-curve, the others to
progressively increasing pulse-brightness levels.
The results showed a close correspondence
with increment threshold analysis for 7, and 7,
and 7, and 7, even though the pulse-brightness
levels correspond to pulses of radiance from one
to three log units above the increment threshold.
For the blue-sensitive mechanism, however, the
a ,-branch shown by the increment-threshold curve
543
is not in evidence when pulse-brightness matches
are made at some one or more log units above
increment threshold. The authors conclude, in
conformity with well-established viewpoints, that
the blue-sensitive mechanism carries little weight
as a brightness-producing system (see, for exam-
ple, Marks, 1974; and Boynton, 1979). Their work
is perhaps most interesting for the persistent ap-
pearance of (74, 74) and (7,, 75) transition in the
suprathreshold measurements.
In a lmited application of the two-color
threshold method confined to the determination
of t.v.A.-curves, making no use of the more
powerful t.v.r. and f.v.u.-approaches, Wald (1964)
obtained spectral sensitivity results for normals
which are substantially in agreement with the
earlier t.v.A.-curves and the f.v.y.-curves of Stiles,
except that for the red-sensitive process (corre-
sponding to 7, or 75), the data of Wald suggest
considerably lower values than for 73 for wave-
lengths shorter than about 560 nm, provided the
curves are superimposed at longer wavelengths
where they have similar shape. This is in interest-
ing difference that has been discussed by Enoch
(1972). It raises the question whether in the
shorter-wavelength half of the spectrum (A < 560
nm), the test sensitivity (t.v.A.) of 7; (or 72) as
measured by Wald may not, in fact, deviate from
the field sensitivity as measured by Stiles. On the
other hand, Ingling (1969), Cavonius and Estévez
(1975), and Estévez and Cavonius (1977) obtained
field spectral sensitivities that closely agree with
ms rather than with the shape obtained by Wald
using the t.v.A.-approach. Ingling’s (1969) 2s,-
curve shows, however, a dip in the wavelength
region of 1 /u = 17,000 cm™', which Ingling sug-
gests may imply the existence of more than one
pigment in the red-sensitive cone mechanism.
(vi) a-Mechanisms and Color-Matching Func-
tions. A first comparison of the spectral sensi-
tivities of a-mechanisms with color-matching
functions made by Stiles (1953) indicated (a) that
the field sensitivities II,,, I1,,, and IIs, of the
mechanisms 7, 74, and 7; respectively, could be
represented to a very rough approximation by
linear combinations of CIE 1931 standard color-
matching functions, or rather better, in the short-
wavelength region, by the modified CIE 1931
color-matching functions of Judd (1951); and (b)
that the chromaticity coordinates of the mono-
chromatic stimuli in a W.D.W.-system, derived on
the hypothesis that the color-matching functions
are linear combinations of the IT,,,, I14,,, and IT,
544 VISUAL THRESHOLDS
field sensitivities, show certain deviations from
Wright’s directly measured chromaticities. (We
recall that the CIE 1931 data are largely based on
these chromaticities of Wright; see Section 3.3.3.)
The noted deviations were taken rather seriously,
particularly the largest one, namely, the too small
amount of negative red chromaticity in the spec-
tral region 19,000 to 22,000 cm ', approximately.
However, it was later realized that the modifica-
tion of the shapes of the field sensitivities, partic-
ularly I1,, and II,,, to eliminate the major
discrepancy would be very small. Nevertheless, at
that time the issue was not pursued further.
Estévez and Cavonius (1977) have made a
much more satisfactory comparison between the
field sensitivities II,,, I1,,, II5,, and the color-
matching functions for the same observers. They
found that there was no significant discrepancy.
They also noted that the field sensitivities of
Stiles given in Table 2(7.4.3) would yield good
agreement with Wright’s mean spectral chro-
maticities if the comparatively small distortion of
less than 0.1 log units were made in the II,,,-curve
in the blue-green spectral region. Estévez and
Cavonius (1977) found no evidence in any of their
work for the field sensitivities I14,, and IIs, and,
in particular, no support for the suggestion that
they would be needed rather than II,, and II., in
explaining their color-matching data which they
obtained at moderately high radiances. Their re-
sults emphasize the unsatisfactory position as re-
gards the existence or properties of the 7; and 7
mechanisms discussed earlier in Section 7.4.4(v).
Pugh and Sigel (1978) used a computer mini-
mization technique to obtain linear combinations
of color-matching functions that yield “best fits”
to the field sensitivities that are numerically de-
fined in Table 2(7.4.3). Three different sets of
color-matching functions were employed: Stiles’
2° pilot data (Section 5.5.3), CIE 1931 standard
colorimetric observer (Section 3.3.1), and Judd
modification of CIE 1931 standard data (Section
dade.)
The main results of the numerical exercise are
summarized in Table 1(7.4.4). Particular attention
is drawn to the rms errors of the best fits calcu-
lated by using the different sets of color-matching
functions. The Stiles 2° pilot data and the Judd
modification of the CIE 1931 data each yield
about equally good fits with the field sensitivities
IT,,, Il5,, U14,, and I,,. Remarkably good fits
are obtained for II,, and II, with very small rms
errors, indeed. The field sensitivities II), and
II,,, on the other hand, are less well fitted, and
the rms errors are about twice as large as those
for II,, and II.,. Pugh and Sigel note special
difficulties in obtaining good fits with II, and
II,,, when the sensitivity values in the long-wave-
length region (1/p < 18750 cm™') are included
in the fitting exercise. However, when the long-
wavelength region of II,, and II;, is truncated
and only the remaining parts of these field sensi-
tivities are fitted, much improved agreements, that
is, much smaller rms errors, are obtained. Pugh
and Sigel argue that such a truncation is justifi-
able on the basis of certain experimental evidence
in color matching which suggests that the contri-
bution of the blue-sensitive mechanism (either
IT,,, or II;,,) to color matches involving test stimuli
of wavenumbers 1/p < 18750 cm! (that is, p >
533 nm) is negligibly small or nonexisting. The
validity of such an argument may deserve further
study.
The results of the fitting exercise of Pugh and
Sigel (1978) with the CIE 1931 standard color-
matching functions are not as good as those ob-
tained with the other two sets of color-matching
functions. They confirm to a large extent the
results obtained earlier by Stiles (1953).
In the same study, Pugh and Sigel (1978) apply
their fitting technique also to pigment spectra
obtained from the Dartnall standard shape of
visual pigment absorption coefficient (see Section
8.2.3) and the iodopsin spectrum of Wald, Brown,
and Smith (1955). Each of these pigment spectra
was numerically varied by shifting (a) the wave-
length at which its absorption maximum occurs,
(b) the optical density, and (c) by applying lens
and macular pigment corrections (see Section
2.4.6). Good fits were obtained between II,, and
an appropriately corrected rhodopsin spectrum
(Dartnall’s standard shape extended by the
iodopsin spectrum), and between II, and an
appropriately corrected iodopsin spectrum. A rea-
sonable fit was also obtained between II, and an
appropriately corrected iodopsin spectrum. How-
ever, these best fits required, in each case, the
assumption of different lens and macular pigment
corrections.
7.55 ROD SATURATION
In extrafoveal areas, color matching is subject to
modification by the participation of the scotopic
or rod mechanism of vision (Section 5.6.2). Evi-
dence on the adaptation levels above which such
rod participation may be expected to be small or
 PIGEL.
T (pp)
= SJUDIDIFJa0D
JO ABOU] SUONLUIQUIOD
Jo suTIvPA[-10JOD
‘suoyouny poyndwio0;D
0} PPA Seg SIL] YIM SIDS PJP SODIAYISUIC
YY TT
TT “yy ure) synsay
Jo yngpur ‘281g (861

Ul
yy
PIs A\ UENTOAL
FOC gS SINS of IO p(SS61) HID 161 pPlepueys ppne UoneoyIpoy!
p61)

[erjoods
a3ue
pAHAT}ISUIS _w9)
(, ; sO]Wate 0.30)
3 20[ iD Soy=! “n8o, p30, 80' 80,_-~—s=n p30,
age OSL'P7—-00S'r1 el cer iy Sibel cl eLELC06: fl. Ob) Le tos erimoe
"or OST'PZ-OSTLI St ee 1) VAdt) el Boer PSIt €1 68S" 11) esi Pol
tl 00S‘P7—-00S‘rI OL 966° £966 OL 68S 7966 OL 38196° 8008 OL veo
Uu 00S*P7—-00S‘F1 6 CEC.
PM 698468 BOLE ESO 6 681 §\ 6 906 Grimul

pjaly, Aziazisuas
— jeooidiay
Jo platy aouKIpEI poInseaw)
UI SULID}
Jo Jaquinu
Jo eyueNb
sod puodses Sula}Ue
oy} ButArasqo
aha tad -(sap)]
Jo [Piety
yeu) SasTeray} JUSUIIIOUI P[OYSaIY}
JO & UDATZ UISTURYDIUT
‘'y)‘cu ‘tw ‘Pu (‘Su0} QT sawIty
st sinjosqe ‘Ploysery
joazadg,asuns
— jegJo winadeds
IJ9Ao YOryM Poly AWANISUSSBQu] [(E'p'L)Z
sem ‘pony
AV,= Joquinu
Jo syutod ‘poiiTy JoquInusARM
Suloeds
= O¢Z ,_Wod ‘ynoysnoIiy)
aie ‘cy ‘fn— sJUdTOIJJOOD
JO ay} IeaUT] UOTeUTGUIOD
Jo PayeoIpUI
19S JO SUTYDJeUT-IO[OD
sUOTIOUN
YOTYM praIA
& WnuITUTU
Jo ay} umsjo perenbs
saoudsajjIp
pouljap
:AQ

301]"z
"11
— 801 _['9 WIM "=" (sH)ala'o]
+ 8° (af)
+ g°un [(1)

SU], BuTYO\eUI-I0[OO
‘suonouNypajouap
Aq “(r1)u ‘(1)8 ‘(11)gore ay) sninuITsty
sonyeA
JO OT}eWIOIYDOUOU
1S9} TTNUITIS
JO YISUITIARM
7 (Wd) pue yrun
souKIpeI poInsvaul)
UI SULId}
JO AA _IS- “(,_W-,
oY] MTVUOIYIOUOU!
VUIIaJoITNUWIT}S
o1e PoJVOO]
-J@

Boccecl= "ay| Ie = 000 ble Tal PUD OOS tC=

3 ‘(WIJ-JO-ssoupoos JO dINsvau) JOIIa aTenbs Uva [eNpIsay = /OL/a SULA,

oy} YOTYyM JO surJo} Ur ‘s}TUN WINjUENb 0} s}TUN (ASIDUD) sOURIPeI WOIT suOT}OUN BUTYO}eUT-IO[OO ay} JaysueN "> ur fn ‘Cn ‘!a “at /] sIOVORT OY, :2/0N
‘POUTJop Ie SOTIIAT}ISUIS PIT]

545
546 VISUAL THRESHOLDS

OF
SATURATION

REGION

(=)
=
a§
=—
”
uJ

FRACTION
FECHNER

(INCREMENT
LOG
THRESHOLD,
TROLANDS)
SCOTOPIC
Nl
oo

oll

>!
& 2 0 2 4
LOG (FIELD RADIANCE, SCOTOPIC TROLANDS)
Fig. 1(7.5). Approach to saturation of rod mechanism in 9° extrafoveal vision by rapid rise in
increment threshold. Curve A: log (increment threshold); Curve B: ratio (increment threshold /
field radiance). (Aguilar and Stiles, 1954).

negligible is provided by determinations of the
increment threshold for rod vision as a function
of adaptation level. The adaptation of the two-
color threshold method with a red adapting field
and a green test stimulus enables the increment
threshold for (extrafoveal) rod vision to be de-
termined up to high levels at which the rod sys-
tem appears to saturate in the sense that it is
incapable of responding to the increment stimu-
lus, however intense.
The mean data obtained in this way by
Aguilar and Stiles (1954) are shown in Figure
1(7.5) as a plot (Curve A) of the increment
threshold against the field radiance, both ex-

Table 1(7.5) Variation of the Ratio (Increment Threshold) /(Field Radiance) as a Function of
Log (Field Radiance). See Curve B in Figure 1(7.5).4
Log (Field Radiance) 4.0 3.0 2.0
Increment Threshold/ 7.1 1.0 0.35
Field Radiance
“Both field radiance and increment threshold are expressed in scotopic trolands.
pressed in scotopic trolands. This t.v.r.-curve
should be applicable, whatever the spectral com-
positions of test and field stimuli to the ap-
proximate estimate of the thresholds for rod
response. In many cases with spectral composi-
tions different from those used in the experi-
ments, the cone mechanism will have a lower
threshold than the rod mechanism at the higher
levels, and the threshold of the latter will not be
observable. The curves of Figure 1(7.5) indicate
that the rod threshold is almost proportional to
the field luminance in the range 107! to 10?
scotopic trolands; it then begins to rise with in-
creasing rapidity at 100 scotopic trolands.
According to estimates by Aguilar and Stiles, the
rod mechanism is effectively saturated [the ratio
(increment threshold /field level) equals 100 times
its value in the 10-' to 10? scotopic troland
range] at a level between 2000 and 5000 scotopic
trolands. The variations of the ratio mentioned
(Fechner fraction) are given in Table 1(7.5).
According to these results, possible participation
of rod vision in color matches is likely to diminish
progressively above 100 scotopic trolands and to
be entirely absent at 2000 to 5000 scotopic tro-
lands.
It has been shown by Fuortes, Gunkel, and
Rushton (1961) that for an observer very deficient
in cone vision, a near rod-monochromat, a rod-in-
crement-threshold curve closely resembling that
of Figure 1(7.5) could be obtained without the
special devices needed with normal observers to
avoid the cones obscuring the approach to satura-
tion. However, the saturation level appeared in
this subject to be sensibly lower, at about 500 to
1000 scotopic trolands.
7.6 CONE SATURATION
The t.v.r.-curve for the extrafoveal retina of the
normal eye corresponds to rod vision of the test
stimulus up to high field radiances if the two-color
threshold method is used with a red field and a
Cone Saturation 547
1 .Cige0 Op 0s oe. 6L00 835
0.260502238 00.22.0020 60/324 0054 LO FR232,
green test stimulus. It then shows the rod-satura-
tion effect illustrated in Figure 1(7.5) of the previ-
ous section. However, if in the same figure there
is also shown the fraction of the rod rhodopsin
bleached at each field radiance, as derived from
Rushton’s work (see Rushton, 1972 for summary,
and Stiles, 1961), it is apparent that rod response
is already saturated when no more than a few
(about six) percent bleaching has occurred. For
foveal (cone) vision, on the other hand, the t.v.r.-
curve, at least for the 7, and 7,-branches, can be
followed to very high field radiances at which
nearly all the cone pigment is bleached, and yet
shows no upward turn characterizing saturation
(Stiles, 1961). ;
An outstanding series of papers by Alpern,
Rushton, and Torii has elucidated the saturation
property by using responses to brief intense con-
ditioning stimuli rather than to fields to which the
retina has become fully adapted.
In their initial researches, concerned with rod
vision, the authors used the so-called after-flash
effect (Alpern, 1965; Alpern and Rushton, 1965)
and found a new kind of saturation in which it is
the inhibitory effect of a pulsed stimulus added to
a Steady field stimulus, namely, its action in rais-
ing the increment threshold in a neighboring reti-
nal area, that is subject to saturation.
In their subsequent researches, Alpern et al.
(1970) applied a similar method to cone vision,
and for this they also established a saturation of
the inhibitory effect. Their work also included
measurements on cone vision by a method more
closely akin to the procedure used originally to
establish rod saturation (Section 7.5).
The work has, to a large extent, been con-
firmed and extended by King-Smith and Webb
(1974) and by Shevell (1977). In using a simpler
method, these authors showed that cone mecha-
nisms, notably the 7,- and 7,-mechanisms, saturate
in the sense that the increment threshold increases
indefinitely when the conditioning stimulus (a
background) is progressively raised towards a
(finite) saturation level, provided, however, that
548 VISUAL THRESHOLDS

the test stimulus is pulsed before the cone mecha-

nism has had time to adapt to the background. In
a typical experimental procedure, the condition-
ing (background) stimulus was exposed on the
previously dark field for just 200 msec; 40 msec
after the onset of the background, a 20-msec test
stimulus was added to the background stimulus. pulsed field

A plot of log (increment threshold) against log

(background radiance) then shows the character-
istic upturn from the 45° line in asymptote to a
vertical saturation line at a background of the
)(increment
threshold
log
order of 4.0 log trolands (red test stimulus, to
isolate 75) or 4.5 log trolands (green test stimulus,
to isolate 7,). A “white” conditioning stimulus
was used in both cases. Figure 1(7.6) illustrates log (field radiance )
schematically typical results obtained for a cone
Fig. 1(7.6). Schematic diagram showing the vari-
mechanism. The fact that cone saturation is ob-
ation of the logarithm of the increment threshold
served by using a pulsed conditioning field but
for a cone mechanism as a function of the loga-
not by a field to which the retina has become
rithm of the field (background) radiance for two
adapted is explained by noting that in the former
conditions of observation. In one case, the field is
situation, the pulsed field generates, by absorp-
exposed steadily; in the other case, the field is
tion in the visual pigment, a strong surge of
exposed as a pulse to the observer. Saturation of
neural responses before the stimulus has time to
the cone mechanism whose increment threshold is
bleach more than a very little pigment. Thus,
being measured occurs only under the pulsed-field
attributing the saturation to congestion in the
condition and not under the steady-field condition.
further transmission of the neural responses, this
The scale values on the abscissa and ordinate of
can occur when the number of quanta absorbed
the figure apply approximately to the experimental
approaches a certain limit. In the equilibrium
case studied by Shevell (1977) for the a5-mecha-
situation, however, this limit may never be
nism: Field (background): 4.1° “white” (tungsten);
reached, because the more intense the adapting
in the pulsed mode, the field was on for 200 msec;
field, the more the pigment is bleached and the
its radiance is converted to log (photopic td), rang-
smaller its capacity for absorbing quanta.
ing from — oo to 6. Test stimulus: 0.57°, red (~ 620
As noted above, the discovery of cone satura-
nm), 20-msec pulse presented 40 msec after onset
tion was first made by Alpern, Rushton, and
of field when pulsed-field condition was used. The
Torii (1970d) by a more involved method that
presentation rate of trials was one per sec. When
utilizes the after-flash effect. In that method, an
the steady-field condition was used, the observer
annular stimulus surrounding the foveal test
was first adapted to the given field radiance for 10
stimulus is exposed for 100 msec at a fixed time
min before beginning increment-threshold mea-
(50 msec) after the exposure (10 msec) of the test
surements. The test stimulus was then presented at
stimulus. The radiance of the annular stimulus is
a rate of 1 pulse (20 msec) per sec.
set so as to raise the threshold of the test stimu-
lus. The effect of the annular stimulus, seen as an
“after-flash,” is reduced if, in the annulus, a
steady field is exposed all the time on which the fixed factor above the value obtained with the
annular stimulus is superimposed. Measurements steady field acting alone. The assumption is made
can then be made of the way the radiance of the that, when this is done at all steady fields, the
annular stimulus, which produces a given effect annular stimulus itself will be producing always
on the threshold of the test stimulus, varies with the same inhibitory effect on the detectability of
the radiance of the steady field. Even if the pulsed the test stimulus. The experimental data so ob-
annular stimulus is not exposed, the steady an- tained show the characteristic saturation shape.
nular field will itself raise the threshold of the test The after-flash method of Alpern, Rushton,
stimulus. To take account of this, the radiance of and Torii (1970d) and the simpler pulsed-field
the annular stimulus is raised until it increases the method of King-Smith and Webb (1974) and
threshold of the test stimulus by some prescribed Shevell (1977), both methods designed to observe

cone saturation, are strictly concerned with differ-
ent visual effects, though, no doubt, they have a
common root cause. It is perhaps not surprising,
therefore, that the field radiance to produce
saturation differs widely, by some two log units,
in the two cases.
Mollen and Polden (1977b) have reported a
saturation behavior also for the blue-sensitive
(7, 7, 73)-mechanisms. By applying the aux-
iliary-field method (Section 7.4.3(g), IT,,-field
sensitivity) to high radiances, the t.v.r.-curve of
these mechanisms bends upwards from the straight
Weber-fraction line towards a vertical asymptote
in a way indicative of saturation. They used
a short-wavelength test stimulus and a short-
wavelength main field of progressively increased
radiance mixed with a constant long-wavelength
auxiliary field. Mollen and Poiden’s results indi-
cate that the blue-sensitive mechanisms not only
saturate, but also, unlike the green 7,- and red
m;-mechanisms, they do so in the equilibrium case
(steady field).
7.7 ROD AND CONE INTERACTIONS
The extent to which a visual response to light is
determined exclusively by the quanta absorbed in
the rod receptor system (absorption by the rod
pigment rhodopsin) or exclusively by those
absorbed in one of the cone receptor systems
(absorption by the cone pigments, erythrolabe,
chlorolabe, or cyanolabe) or, in some way, by
contributions from absorptions of more than one
of these kinds, is a wide-ranging and controversial
question. The following remarks deal with the
more limited question of the extent to which
increment thresholds (including thresholds on zero
fields or absolute thresholds) reveal interaction
between different component mechanisms.
In an experimental determination of an incre-
ment threshold, where the response may arise
through two different receptor mechanisms, an
apparent summation of the effects of the two
mechanisms may be produced by what is known
as probability summation, even though there is the
highest degree of independence of the mecha-
nisms, that is no interaction. This is now a familiar
concept and is discussed in Section 8.4.7 on line
elements. It leads to the general rule that if the
thresholds of the independent component mecha-
nisms, when acting alone, have infinitely sharp
values, there is no probability summation and the
observed threshold is the lower of the component
Rod and Cone Interactions 549
thresholds. However, if a threshold is simply the
value which gives a certain probability of detec-
tion, the lowering of the resultant threshold will
be determined by the steepness of the probability-
of-detection curves given by the component
mechanisms, acting alone. In recent years, the
more sophisticated approach of signal detection
theory (see, for example, Nachmias, 1972 and
Egan, 1975) may modify the details of the proba-
bility summation concept, without necessarily al-
tering the essential point.
Where the observed threshold cannot be iden-
tified as the resultant of the thresholds of inde-
pendent component mechanisms, the summation
is described as physiological summation and be-
comes evidence of some interaction in the
threshold responses of the mechanisms. This con-
cept is not, of course, limited to threshold re-
sponses of component color mechanisms for test
stimuli of fixed characteristics. In the case of a
large test stimulus, the resultant threshold may be
thought of as arising from similar mechanisms
operating in the various small areas of the retina
on which the test stimulus is imaged. The same
distinction between probability and physiological
summation can be drawn.
In the original increment-threshold analysis of
Stiles (for the collection of his relevant papers, see
Stiles, 1978), it was assumed that the component
mechanisms (rods and three cone mechanisms)
were independent so that only probability sum-
mation operated, or that any physiological sum-
mation effects were small enough to be ignored to
a first approximation. Although this is still proba-
bly acceptable as a crude approximation, small
but significant deviations have been found.
The first study of the summation of threshold
responses from different 7-mechanisms was made
by Stiles in 1959 at the National Research Coun-
cil of Canada (see Stiles, 1967). The two-color
threshold method was used to set up conditions in
which a test stimulus consisting of a mixture of
two wavelengths was such that for each of them
in the absence of the other, the threshold could be
attributed to a single mechanism, different for the
two wavelengths. With varying proportions in the
mixture, it was possible to determine the way in
which the responses from the two mechanisms
were summating.
For a mixture of two wavelengths that are
both on a section of the envelope provided by one
mechanism, for example, the 73-mechanism, the
mechanism integrates completely. It follows that
at the threshold of the mixture, the quantities of
550 VISUAL THRESHOLDS
the constituent wavelengths are fractions adding
up to unity if these quantities are expressed in
units equal, respectively, to the threshold for the
separate wavelengths. For the special case of an
equal mixture, these fractions are equal, and minus
the logarithm of their values, are defined as the
summation index, equal in this case to 0.301.
On the other hand, for two wavelengths in
sections belonging to different mechanisms, the
summation index will be smaller, as the mecha-
nisms are assumed independent and can only
collaborate in determining a resultant threshold
by probability summation.
Clearly, this arises because the threshold cor-
responds to a 50% chance of seeing the test
stimulus, and, if two completely independent
mechanisms each give a 50% chance, the resultant
chance is 75% (see also Section 8.4.7), correspond-
ing to a lower threshold. From the steepness of
experimental frequency-of-seeing curves, it is
found that because of probability summation, the
summation index cannot be less than 0.1 log units
or a little more.
In Figure 1(7.7) are plotted the 1959 results
Stiles (1967) obtained for a green (filtered light
with peak wavelength at 555 nm) field of high
(30,000 td) radiance. The summation index refers
to pairs of 1° (200 msec) test stimulus of wave-
0.5
—-—--—
04
is Observers:
os
£03
c
2
=)
e 0.2
3
P= | =—_———
i?) ee
oO
O
400
25.0 22.5
Fig. 1(7.7). Summation index for high-radiance green field, and for test stimuli of wave-
lengths A, = 640 nm (fixed) and A, (variable through the spectrum). Test stimulus: 1°,
200 msec (Stiles, 1967).
lengths A, and A, of which one is kept fixed at
640 nm on the section of the envelope provided
by the 7 mechanism while the other is varied
through the spectrum. Despite inevitable scatter,
the points are in tolerable accord with expecta-
tion when the second wavelength is also on the
m-section, or when it is on the 7-section. The
striking discrepancy occurs for intermediate posi-
tions around 530 nm, where the summation index
deviates by 0.1 to 0.2 log units from the predict-
ion of probability summation in the direction of
less summation, indicating, apparently, a physio-
logical inhibitive rather than summative effect.
Actually, around 530 nm the summation index
drops to the inexplicable value of zero, that is,
below the theoretical limit of 0.1 log units.
Repetition and extension of the work, using,
however, smaller (10 min) test stimuli and making
measurements for negative as well as positive
increments of the test stimuli, showed many simi-
lar and rather complicated summation or inhibi-
tion effects of the same order of magnitude
(Boynton, Ikeda, and Stiles, 1964; Ikeda 1963;
1964). Ikeda, for example, showed that with very
brief test pulses the inhibitive effect did not ap-
pear. Whereas perhaps none of the deviations
from strict probability summation is large enough
to affect the increment-threshold analysis re-
Curve computed for probability summation
Mean experimental curve
o WSS
Aa GHF
450 x 550 600 640 700
Wavelength \, (nm)
é 17.5 15.0-10°
Wavenumber m, 2-2 (cm')
t

garded as a first approximation, they are not yet
understood and need further study. An element in
any further work on the same lines that might be
undertaken, is the use of double-pulse increments
in which two stimuli are separated by some milli-
seconds but are sufficiently close to be appreci-
ated as a single presentation for detection. It has
been found that the white-light thresholds for
double pulses also show interesting inhibitive ef-
fects (for summary, see Boynton, 1972) and the
potentialities of this new temporal factor in
elucidating component color mechanisms seem
worth exploring.
In the special case of increment thresholds on
a dark field, Guth (1967) and his collaborators
have observed summation anomalies of much the
same kind as those discussed above. However,
Guth’s interpretation of his results aims primarily
at the problem of brightness additivity associated
with the difficult concept of heterochromatic
brightness matching. Some of his experimental
data have been given in Section 5.8.
The above interactions have been considered
from a position that assumes each mechanism to
be brought independently to a certain state by the
conditioning stimulus, the actual interaction oc-
curring in the processes by which the signals
produced by the pulsed test stimulus in the re-
spective mechanisms are combined at some higher
level to determine a detectable or not-detectable
resultant response. Interaction of a different kind
might fairly be said to occur if the condition of a
particular mechanism were dependent not only
on light absorbed in the “proper receptors” but
also on light absorbed in the “proper receptors”
of other mechanisms. To make precise this state-
ment which uses the undefined and vague term
“proper receptors,” it is useful to look again
briefly at the development of the concepts of
increment threshold mechanisms.
Originally (Stiles, 1939), it was postulated that
any radiant energy (that is, any conditioning
stimulus) modifying the condition of a particular
mechanism and any radiant energy (that is, any
test stimulus) producing a threshold response
would do so, as far as variations of wavelengths
were concerned, in accordance with a spectral
sensitivity curve unique to the mechanism and the
same for both conditioning and test stimuli. It
was also implied—and the formulae given in Sec-
tion 7.4.2 assumed this—that for nonmonochro-
matic stimuli of various spectral compositions,
their effects on the mechanism would be de-
termined by a simple integration over the spec-
Rod and Cone Interactions 551
trum of the product of its spectral sensitivity and
the spectral radiant power of the stimulus. Thus,
an initial /inear stage in the effect was assumed.
Most of the earlier studies employed monochro-
matic stimuli and for these it was concluded
(Stiles, 1953) that, in the spectral regions where
the relative spectral sensitivities both to test
stimuli and to conditioning stimuli could be de-
termined, they were the same. However, the unex-
pected lobe in the II,,-field sensitivity curve at
moderate and long wavelengths [see Figure
2(7.4.3)] led, among other suggested explanations,
to the proposal that a second mechanism was
contributing to the field sensitivity. As discussed
in Section 7.4.4(1), Pugh (1976) has now estab-
lished that the II,,-field sensitivity curve must be
composite and dependent on two (perhaps even
three) spectral sensitivities associated probably
with the “proper receptors” of other mechanisms.
If we think of two sensitivities for each of
which the additivity law holds, then the basic
formula for the increment threshold given in Sec-
tion 7.4.2 as Eq. 1(7.4.2), that is,
= 7 $;(M,II;)
=|— >
must be modified for mechanism i = | to
l
M,Mi,
,
],p?
[1(7.7)]
where the function Z,{ ---, --- } depends on two
arguments, and where II}, and II}, are two com-
ponent field sensitivities. Equation 1(7.7) applies
to monochromatic test and field stimuli. For non-
monochromatic field stimuli, the products M, 11},
and MII‘), must be replaced by integrals of the
form given earlier as Eq. 5(7.4.2). Pugh (1976) has
shown that the function Z,(x, y) does not reduce
simply to a function of (x+y), which would
correspond to the original theory. The theory
proposed by Pugh and Mollen (1979), briefly
referred to in Section 7.4.4(iv), may lead to fur-
ther experimental studies specifically designed to
explore more deeply the interactions between the
different blue-sensitive mechanisms.
However, the important modification in the
a-mechanism concept necessary to accommodate
the above interaction phenomena, must not ob-
scure the fact that in certain experimental situa-
tions, the mechanisms preserve a high degree of
392 VISUAL THRESHOLDS
independence. Alpern and Rushton (1965) have
shown that after-flash phenomena are consistent
with the view that the 7-mechanisms are behaving
independently, even though in these cases it is
frequently a stimulus applied not only at a differ-
ent time, but also in a part of the retina that
merely adjoins the area on which the test pulse is
imaged, which is producing a “field” increase in
test threshold (see also Section 7.6).
The possibility of a modification of the t.v.r.-
curve of the rod mechanism produced by a contri-
bution from the cone systems has been raised by
Makous and Boothe (1974) from the results of
increment-threshold studies that apparently con-
tradict the results of Flamant and Stiles (1948).
However, in later work, Makous and Peeples
(1979) find that using the same conditions they
find no conflict in the two sets of data, but that
when sufficiently high adapting fields (higher than
those of Flamant and Stiles) are used, the t.v.r.-
curve for fields of long wavelength differs from
that for fields of shorter wavelength, the former
apparently having more effect at high field levels
in raising the rod threshold even though the fields
are scotopically balanced. It certainly seems that
the results of Makous and his co-workers present
a problem.
From material summarized by Stiles (1939),
Figure 2(7.7) has been prepared showing the
t.v.r.-curve for 5° parafoveal vision for the combi-
nation A = 580 nm, » = 500 nm. From the test
spectral sensitivity data in the same paper, it
appears that changing A from 580 nm to 490 nm
would lower the rod branch by some 0.6 log units.
On the other hand, if parafoveal cones behaved
just like foveal cones, the same change of A would
raise the rod branch by an estimated 0.9 log units.
In fact, the increase is likely to be less for the
parafoveal cones than for the foveal cones. Proba-
bly, therefore, the net effect of changing A from
580 to 490 nm would be to raise the cone branch
with respect to the rod branch by not more than
1.5 log units, and the dashed curve in Figure
2(7.7) shows the new (relative) position. Clearly,
if this approximate prediction of the situation is
‘valid for the combination A = 490 nm, p = 500
nm, the cone branch will take over at a suffi-
ciently high scotopic field level and thereby mod-
ify the apparent t.v.r.-curve. With a deep-red field,
the scotopic level at which the takeover by the
cone branch occurs will certainly be higher, so
that the t.v.r.-curve for green (500 nm) and deep-
red fields will deviate in shape at a certain level
that will coincide with the takeover level for the
green field.
Pol
ou
BY)
DISPLACED—
CONE ras
on
oO
“Ni
ROD BRANCH
THRESHOLD)
(INCREMENT
LOG
@:
io 8 6 4 2
LOG (FIELD RADIANCE)
Fig. 2(7.7). Schematic illustration of the problem
raised by the results of Makous and his collabora-
tors on the possible effect of cone stimulation on
the form of the t.v.r.-curve of parafoveal rod vi-
sion.
In Figure 2(7.7), the line of heavy dots indi-
cates where the breakaway point may be when
one uses the results of Makous and Peeples (1979).
It is clearly at a lower level than that predicted
for cone intervention according to the usual the-
ory. Makous and Peeples determined the ap-
proximate position of the absolute cone threshold,
not in the way adopted here, but experimentally
from the break in dark-adaptation recovery curves
(Section 7.3.1), and established by a direct method
that there is a problem. It is not altogether clear
the kind of rod—cone interaction that could ex-
plain their results. If it is a contribution of cones
to a rod—cone pool (Sternheim and Glass, 1975),
through which the effect of the field in raising the
threshold is produced, one might expect a red
field with its relatively large cone effect to bend
an ideal straight line rod branch upwards at high
field radiances, rather than for short-wavelength
fields to bend the branch downwards at high field
radiances. Alternatively, the cone saturation might
be assumed to inhibit the effect of rod stimulation
in raising the threshold.
7.8 UNIFORM EQUIVALENT FIELDS
In the specification of adaptation, a method first
used by Holladay (1927) in expressing the results
of measurements of the rise in increment threshold
when the visual field contains a small “glare”
source, is relevant. The underlying idea was that
the glare source acts by throwing a veil of scattered

light (in the eye) over the retinal area where the
test stimulus is applied and that the level of this
veiling luminance would be determined by find-
ing the actual uniform field luminance producing
the same rise in threshold in the absence of the
glare source.
In the later development of this idea (Stiles
and Crawford, 1932), it was shown that the value
of the equivalent uniform field luminance, which
also became known as the equivalent background
luminance, is substantially independent of the
characteristics of the test stimulus used in its
determination. It was further shown that it pro-
vides a measure of the condition of the retina as
modified by the glare source irrespective of
whether this modification corresponds to a super-
posed veil of scattered light or to some deeper
physiological process by which stimulation in one
retinal area affects response in another. By apply-
ing the same methods to the increment threshold
measured after the abrupt removal of a condition-
ing stimulus of high luminance, the course of
dark-adaptation was expressed by a curve of
equivalent background luminance against time.
Such a curve was then independent of the char-
acteristics of the test stimulus and more accessible
to theoretical treatment than the original dark-
adaptation curves (see also Crawford, 1947).
The above studies apply when only achromatic -
stimuli are involved and when the condition of
the retina may (possibly) be specifiable by one
variable only. For the more complicated case,
when chromatic stimuli are used and trichromatic
principles must enter, Moon and Spencer (1943c)
have elaborated a three-dimensional method of
specifying foveal adaptation that rests on the
concept of equivalent background luminance.
In more recent years, a number of important
applications and extensions have been made of
the concept of equivalent background luminance.
Blakemore and Rushton (1965), using both uni-
form fields and gratings find comparable results
to those of Crawford (1937, 1947) and, in particu-
lar, confirm the equivalence to hold for different
field sizes.
Spillmann and Fuld (1979) tested the validity
of the equivalent-background concept for differ-
ent retinal locations on the nasal half of the
horizontal meridian. They obtained both dark-
adaptation and increment-threshold curves for
different test-stimulus conditions [A = 550 nm
for 2° (175 msec) and 0.5° (40 msec)]. The experi-
mental results showed that the equivalent back-
ground luminance (converted to trolands) follows
its own time course at each retinal location. For
Uniform Equivalent Fields 553
photopic conditions, the equivalent background
luminance for the inner periphery is lower than
for the outer periphery. For scotopic conditions,
on the other hand, this trend reverses. These
results compare reasonably well with those ob-
tained earlier by Crawford (1937). Spillmann and
Fuld (1979) suggest that the equivalence hypothe-
sis, also referred to as Crawford transformation,
obtains only for retinal locations that have com-
parable spatial and temporal properties.
The studies of Alpern and Campbell (1962) on
the pupil behavior during dark adaptation, and of
Barlow and Sparrock (1964) on the brightness of
afterimages have provided further support to the
equivalence concept.
Of interest is also the work of Blakemore and
Rushton (1965), Du Cruz and Rushton (1966),
Spillman, Wolf, and Nowlan (1971), Spillman,
Nowlan, and Bernholz (1972), and Geiser (1979)
who conducted detailed studies on foveal dark-
adaptation curves. The experimental data support
the hypothesis that the effects of a bleaching field
are interchangeable with those of a uniform field
and are, in fact, linearly additive.
There are, however, also a number of studies
that suggest limitations to the equivalence con-
cept. Some of these studies are concerned with
summation pools (Rushton, 1965c), spatial sensi-
tization (Westheimer, 1968), photopic spatial
and temporal summation (Rinalducci, 1968;
Rinalducci, Higgins, and Cramer, 1970; Stewart,
1972; Rinalducci, Lowenhaupt, and Martinez,
1973), and border contrast (Tachibana, 1977).
A useful review of the concept of uniform
equivalent fields and particularly its application
to studies of mechanisms of pigment-bleaching
recovery has been given by Barlow (1972).
7.8.1 Basic Formulae
For steady conditioning by achromatic stimuli,
the formula used to represent measurements of
the equivalent background luminance L in a field
containing an actual background luminance L
extending over the test-stimulus area together with
a “point” source producing an illuminance E at
the observer’s eye and located @ degrees from the
test stimulus has the form
b=L+ke [1(7.8.1)]
where k and n are constants (Holladay, 1927;
Stiles, 1929a,b). The values of the constants ob-
tained by different investigators using different
 554 VISUAL THRESHOLDS
ranges of the variables show some differences;
but for foveal vision of the test stimulus and with
OSPR Os2S S10 means placed
m ’, the values n= 2 and k= 10 are
fairly representative values. The experimental
data, directed to practical glare problems, were
determined with the natural eye pupil. The cor-
rection to give a relation in terms of nominal
retinal illuminances (7) is complicated (Stiles,
1929), but little error (perhaps 10%) is incurred by
assuming the pupil to have the same size under
the glare and the equivalent field stimuli. Equa-
tion 1(7.8.1) then becomes
abi 7 a [2(7.8.1)]
where 7, T, T; are, respectively, the troland val-
ues of the equivalent background, the actual
background, and the glare source, and we is the
solid angle in steradians subtended at the eye by
the glare source.
Equation 1(7.8.1) with n = 2 has been found
to apply approximately to the equivalent back-
ground luminance for vision of a test stimulus
imaged in the extrafoveal retina with @ put equal
to the angle subtended at the eye by the line
joining test stimulus and glare source (Stiles and
Crawford, 1937). The value of the constant k
increases from 10 to 16 as the test-stimulus image
is moved into the extrafoveal retina. The position
of the glare source with respect to the fovea is not
a significant factor. For one or more glare sources
copresent in the visual field, the resultant equiva-
lent background luminance is found to be ap-
proximately additive:
ee
Lark pegs.
[3(7.8.1)]
(Luckiesh, Taylor, and Holladay, 1925; Crawford,
1936).
Equation 3(7.8.1) enables the equivalent back-
ground within a small darker area in a large
bright surround (ideally filling the whole field) to
be calculated by integration of Eq. 1(7.8.1). At
the center of a circular darker area of diameter
6(0.5° <8 < 40°) and luminance L, with a sur-
round of luminance L,, L is given by
L = L + k(0.0087 — 0.0044 log 8)(L, — L)
[4(7.8.1)]
for n = 2.
The most important case is with L = 0 when
for foveal vision (k = 10), the surround has the
same effect as a uniform field of luminance 8.7%
(6 = 1°) or 4.3% (8 = 10°) of the surround value.
Direct measurements of 6 = 1° (Stiles and Craw-
ford, 1932) indicate values in fair agreement with
this estimate.
For other related studies to glare see, for ex-
ample, Fry and Alpern (1953, 1954); Boynton,
Bush and Enoch (1954); Vos and Bouman (1960);
Vos (1963b); and J. Walraven (1973).
7.9 SPATIAL AND TEMPORAL FACTORS
The response characteristics of the visual mecha-
nism are functions of not only the spectral prop-
erties of the test and conditioning stimuli, but are
also functions of the temporal modes and spatial
configurations in which these stimuli are pre-
sented to the observer’s eye(s). The previous sec-
tions give important examples of color vision and
its dependence on spectral, temporal, and spatial
characteristics of the visual stimuli.
The color-matching properties of the human eye
are obviously dependent on the spectral radiant
power distribution of the stimulus, its field size,
and location on the retina (Section 5.6).
The brightness-matching properties are similarly
dependent on the spectral radiant power distribu-
tion, field size, and location on the retina of the
stimulus, but also show a distinct functional rela-
tionship with the temporal mode of the stimulus.
Flicker photometry and heterochromatic bright-
ness matching are important examples (Section
Di).
In assessing the perceptual attributes, hue,
saturation, and brightness of a given test stimu-
lus, again all three physical characteristics (spec-
tral, temporal, spatial) of the test stimulus as well
as those of the conditioning stimuli are important
parameters (Section 6.3 to 6.5).
Then there are the visual thresholds, particu-
larly those discussed in Section 7.4 dealing with
chromatic adaptation and increment thresholds,
whose dependence on all three physical character-
istics of both test and conditioning stimuli often
play a major role in the experimental procedures
and interpretation of the results.
Many visual scientists have directed their at-
tention to visual phenomena that are mainly asso-
ciated with temporal and spatial variations of the
observed stimuli. In particular, the objective has
been to quantify the temporal and spatial re-

sponse characteristics of the visual system. The
tools of Fourier analysis have been applied effec-
tively to this task yielding temporal and spatial
modulation transfer functions of the visual system.
The experimental procedure used in studying
the temporal response characteristics of the visual
system typically involves a visual stimulus of some
specified size whose luminance is varied sinusoid-
ally as a function of time over a range of frequen-
cies and luminance amplitudes. For a given
frequency and mean luminance, the observer ad-
justs the luminance amplitude until the imposed
sinusoidal variation is just large enough that
the field does not appear steady in brightness
(de Lange Dzn, 1952; Kelly, 1961).
An analogous procedure is used in studying
the spatial characteristics of the visual system.
Typically, the visual stimulus is displayed in the
form of a sine-wave grating, that is, a regular
stripe pattern whose luminance across the pattern
varies sinusoidally. The observer is asked to de-
termine the threshold for detecting the pattern.
Presented with a sine-wave grating of given spa-
tial frequency, the observer adjusts the amplitude
of the luminance variation until he or she can just
see the presence of the grating, or just distinguish
it from a perfectly uniform field (Van Nes and
Bouman, 1967). A different procedure was intro-
duced by M. L. Davidson (1968), whose observers
matched the luminance contrast perceived for one
grating with that of another (reference) grating.
Fourier-analysis techniques are well-estab-
lished tools in electrical engineering, communica-
tion systems, and optical imaging. There is a host
of textbooks that provide the mathematical algo-
rithm of these techniques; for their applications
to vision, an excellent and well illustrated intro-
duction has been given by Cornsweet (1970).
From the numerous studies of temporal and
spatial response characteristics of the visual
mechanism that have been made, we have selected
a few of those that have a direct bearing on color
vision and that have not been treated in the
previous sections or have been referred to only
briefly in passing.
7.9.1 Distinctness of Border
The perception of contours is an important exam-
ple demonstrating the effect of the spatial re-
sponse characteristics of the visual system on the
perception of hue, saturation, and brightness.
Of particular interest is the pioneering work of
Boynton and his coworkers (Boynton and Kaiser,
Spatial and Temporal Factors 555
1968; 1978; Tansley and Boynton, 1976; 1978:
Valberg and Tansley, 1977; for reviews, see Boyn-
ton, 1973; 1978; 1979; see also Section 3.7):
These workers used three methods for estimat-
ing the perceived distinctness of the border be-
tween juxtaposed fields. In the first method, two
juxtaposed fields of uniform chromaticity are pre-
sented slightly above, but clearly separated from,
two other juxtaposed chromatic fields. The ob-
server adjusts the radiance of one of the two
upper fields until a border is perceived which is as
distinct as the border seen between the two fields
below.
In the second method, only one pair of juxta-
posed fields is used. The observer first establishes
the minimally distinct border between the two fields
by making an appropriate adjustment of the radi-
ance of one field, and then judges the distinctness
of the border in terms of a seven-point rating
scale (“zero” meaning no border at all, “seven”
meaning a very strong border).
In the third method, the observer steadily
fixates the border between two juxtaposed fields
for a certain period of time, five minutes, say, and
the tendency of the border to fade is then mea-
sured in terms of the percentage of time that a
border is not seen.
All three methods yield comparable results of
border distinctness and provide a measure of the
equivalent amount of achromatic contrast that
produces the same “strength of border” as that
produced by a purely chromatic difference. One
or the other method has been used to study a
number of visual response characteristics.
Luminous efficiency functions have been
determined with the minimally distinct border
criterion and were found to be in close agreement
with the V(A)-curve obtained by the conventional
minimum-flicker technique (Wagner and Boyn-
ton, 1972; see also Section 5.7.2).
When the equivalent achromatic contrast was
determined as a function of wavelength for mono-
chromatic stimuli paired with a “white”-stimulus,
the resultant curve closely resembles that of the
colorimetric purity (“saturation”) discrimination
curve (Kaiser, Herzberg, and Boynton, 1971;
Boynton, 1973; see also Section 7.10.3).
Another interesting study has been made by
Tansley and Boynton (1976), extended by
Valberg and Tansley (1977), dealing with tri-
tanopic purity, that is, colorimetric purity dis-
crimination along tritanopic confusion lines in
the chromaticity diagram. When two color stimuli
have chromaticities on a tritanopic confusion line
556 VISUAL THRESHOLDS

0.3 0.4 0.5 0.6 0.7 0.8

x

Fig. 1(7.9.1). Results of minimally distinct border experiment of Tansley and Boynton (1976).
Each monochromatic stimulus of wavelength A; ({=1 to 6) is compared with each of three
tristimulus mixtures [(a;,S;; + 4;25;.) where a;, and a;, are variable and j = 1, 2, or 3]. The
observer adjust the ratio a;, /a;, at constant luminance until a color stimulus (open circle) is
obtained which gives a minimally distinct border with the juxtaposed monochromatic stimulus
(solid dot). In most of the 18 cases, the border actually disappears entirely. The tristimulus
mixtures (open circles) that give zero (or nearly zero) border with a given monochromatic
stimulus (solid dot) lie on (or nearly on) a tritanopic confusion line with origin at (7). This
suggests that regardless of the activity of the blue-sensitive cones, the distinctness of a visually
distinct border may be related to the activity of the red- and green-sensitive cones only.

(see Section 5.14.2), although they stimulate the
blue-sensitive cones of a normal trichromat dif-
ferentially, they stimulate the red-and green-sensi-
tive cones to exactly the same extent at constant
luminance. Figure 1(7.9.1) illustrates some of the
results obtained by Tansley and Boynton (1976).
7.9.2 Mach Bands
Mach bands (Mach, 1865) are light or dark nar-
row bands that are perceived near the border of
two juxtaposed fields, one field being darker than
the other. Figure 1(7.9.2) illustrates a typical case
showing the spatial luminance distribution across
the border and the corresponding spatial distribu-
tion of brightness perceived by the observer.
Mach bands have been studied extensively and
the interested reader may find the reviews by
Ratliff (1965) and Fiorentini (1972) useful start-
ing points for more detailed information. The
phenomenon in all its complexities is not fully
understood but it is generally agreed that lateral
interactions in the neural network of the visual
system account for it. It is assumed that the
function that characterizes the response of each
element of the neural network has an excitatory
component, which depends on the luminance of
the stimulus eliciting a response in that element,
and an inhibitory component which depends on
the luminance of the stimulus (or stimuli) eliciting
responses in the elements surrounding the first
element. The magnitude of the inhibitory effect of
each surrounding element on the first element

Luminance
)units
(arbitrary
Brightness
)units
(arbitrary
O B
-2 =I ) 2
Distance from Center Line (O) of Field
(arbitrary units)
Fig. 1(7.9.2). Luminance and brightness profiles
of a visual field with a graded border. A light (A)
and dark (B) Mach band appears near the points
where the luminance of the border changes
abruptly.
depends on the distance between the two ele-
ments.
The experimental evidence of particular inter-
est to color vision is that Mach bands
general, not found near chromatic borders when
the two juxtaposed fields have no luminance vari-
ation but differ only in chromaticity, including
possibly a graded transition from the chromatic-
ity of one field to the chromaticity of the other.
7.9.3 Flicker
A rotating sector disk may be used to expose and
occlude a spot of light for equal intervals of time.
Depending on the speed of rotation of the disk,
the intermittent stimulation of the observer’s vis-
ual system results in the sensation of flicker. At a
very slow rotation of the disk, the spot appears to
flash on and off in a discreet but regular fashion.
A somewhat increased rotation causes the dis-
creetness of the on-off flashes to disappear. In
fact, the brightness of the “pulsing” light is greater
than the brightness perceived of the spot of light
when seen through the opening of the stationary
disk; this enhancement of brightness phenom-
Spatial and Temporal Factors 557
enon is known as the Briicke—Bartley effect in
accordance with observations made by Briicke
(1864) and Bartley (1938, 1941). When the rota-
tion of the disk is further increased, the spot of
light is almost steady but not quite; a sensation of
flicker remains and the Briicke—Bartley effect is
no longer observed. A still further increase of the
disk’s rotation eventually leads to an abrupt dis-
appearance of flicker and the observer perceives a
steady spot of light. The frequency, that is, the
number of cycles of exposures and occlusions of
the spot of light per second, at which flicker
disappears is called the flicker threshold or, more
commonly, the critical flicker frequency (CFF);
the phenomenon of the disappearance of flicker
at that frequency is called flicker fusion. The spot
_Of light then seen has a brightness, that is, for the
special case of the rotating disk used above, ex-
actly equal to the brightness of the steady spot of
light, exposed through the opening of the sta-
tionary disk, but at half its original radiance.
The special case of the rotating disk used to
demonstrate the flicker phenomenon can be ex-
tended to any technique that generates rapidly
repetitive visual stimuli with periodic variations
of the radiance (or luminance) that can take any
function of time, not restricted to on—off modes
of equal periods. When flicker fusion is reached,
the brightness of the physically alternating stimu-
lus is equal to the physically steady stimulus
whose radiance (or luminance) is set equal to the
mean of the time-dependent radiance (or lumi-
nance) of the physically alternating stimulus,
are, in
calculated by appropriate integration over the
time it takes to complete a full cycle of the
periodic variation. This relationship is known as
the Talbot law or Plateau-Talbot law (Talbot,
1834; Plateau, 1835).
Talbot’s law is of particular importance in
visual experiments in which stimuli are alternated
by a rotating sector disk whose transmittance can
be determined precisely by measuring the open
portion of the disk. The validity of Talbot’s law
has been tested and was found to be very accurate
over a wide range of conditions (see, for example,
Le Grand, 1957; Waish, 1958).
There are many papers that have been devoted
to the study of flicker with regard to a variety of
conditions of observation. Useful reviews on the
results of such studies, some of them containing
comprehensive bibliographies, are those of Landis
(1953, 1954), Pirenne (1962d), J. L. Brown (1965),
Cornsweet (1970), and Kelly (1972). Some of the
early flicker data are given in Figures 1(7.9.3) and
 558 VISUAL THRESHOLDS

50

40 fovea

5° above fovea
15° above fovea

0 | 2 3 4
Log [ retinal illuminance (trolands
)]
Fig. 1(7.9.3). Relationship between critical flicker frequency (CFF) and retinal illuminance
obtained by Hecht and Verrijp (1933) for a 2° “white” stimulus centered at three different
retinal locations: fovea (0°) and 5° and 15° above the fovea. A 10° “white” surround was used
of the same luminance as the test stimulus (above CFF). The observer looked through an
artificial pupil of 1.8 mm in diameter. The photopic retinal illuminance scale used as the
abscissa is the corrected scale in accordance with instructions given by Hecht and Shlaer
(1936).

2(7.9.3). Figure 1(7.9.3) is based on results ob-
tained by Hecht and Verrijp (1933) which show
the variation of CFF with retinal illuminance for
a 2° “white” stimulus presented at three different
retinal locations: at the fovea (0°) and at 5° and
15° above the fovea. A 10° “white” (nonflicker-
ing) surround was used whose luminance was set
equal to the luminance of the test stimulus when
its “on-off” flicker rate was above the CFF. For
the foveal (0°) test location, a single curve is
obtained that peaks at about 10,000 td with a
CFF of approximately 45 Hz. The result applies
to predominantly cone vision. For the two
parafoveal (5° and 15°) test locations, each of the
resulting CFF-curves shows two branches. The
branch for low retinal illuminances is ascribed to
' rod vision, whereas the other branch, which oc-
curs at high retinal illuminances and resembles
that for central foveal vision, is ascribed to
parafoveal cone vision.
Figure 2(7.9.3) is based on results obtained by
Hecht and Shlaer (1936) which show the variation
-of CFF with retinal illuminance for a centrally
viewed 19° monochromatic stimulus of varying
wavelength, presented in a 35° (nonflickering)
surround. For high retinal illuminances, the mea-
surements for different wavelengths can roughly
be approximated by a single curve attributed to
cone vision, though some wavelength dependence
has been noted in that region (Landis, 1954). For
low retinal illuminances, on the other hand, sep-
arate wavelength-dependent branches are ob-
tained that are attributed to rod vision; the most
separated branch being the one for the shortest
wavelength (A = 450 nm). In interpreting the re-
sults, it must be emphasized that the retinal illu-
minances given on the abscissa are based on
photopic measurements employing the technique
of heterochromatic brightness matching in the
fovea (see also Section 5.7.1).
The modern approach to studying flicker phe-
nomena is by employing the principles of
linear-systems analysis, well developed in the fields
of electrical engineering, communication, and
optical imaging. The mathematical tool is that
known as Fourier analysis which permits flicker
waveforms to be expressed in terms of sets of sine
waves. De Lange Dzn’s (1952, 1958a,b) studies
are generally considered the first major studies
that introduced the principles of linear systems
 Spatial and Temporal Factors 559

70

50

oS(e)

te)
(Hz)
CFF
20

*) |
670 nm
605 625

-4 =3 -2 a 0 | a 3 4 5 6
Log [retinal illuminance (trolands
)|
Fig. 2(7.9.3). Relationship between critical flicker frequency (CFF) and retinal illuminance
(photopic trolands) for monochromatic test stimuli of seven different wavelength A. The 19°
test stimulus is presented in a 35° (nonflickering) surround. (Based on work by Hecht and
Shlaer, 1936.)

analysis to visual flicker phenomena (G. Sperling,
1964). Important further contributions have been
made by Kelly and his coworkers and many other
investigators since (for excellent summaries, see
Cornsweet, 1970; Kelly, 1972).
The model underlying flicker responses is
sketched schematically in Figure 3(7.9.3). The
waveform of the visual stimulus, the “input” to
the visual system, is represented by the function
f(t), typically expressed in terms of retinal illumi-
nance (td) as a function of time (ft). The “output”
of the visual system has a waveform represented
by the function h(t). It is postulated that the
principal of superposition (additivity) is strictly
obeyed; that is, the output, corresponding to the
sum of two given inputs is equal to the sum of the
two corresponding outputs. The superposition
property permits the output, corresponding to
virtually any input, to be calculated from the
knowledge of the visual system’s transfer function
G(w) = Ae’”® [1(7.9.3) ]
where ¢ denotes time and A sin (wt + ¢) repre-
sents the output obtained when the input is given

Detector

Input Transfer Output

Fig. 3(7.9.3). Schematic diagram of the linear model applied to visual flicker responses.
 560 VISUAL THRESHOLDS

20/w

mTo

To

f(t)=To(l+m sin wt)

(trolands)
T
Illuminance
Retinal
Time t (seconds)
Fig. 4(7.9.3). Basic waveform of a visual stimulus used in flicker studies. The waveform is the
sum of a constant component 7, and a sinusoidal component of amplitude m7, with 0 < m <1.
The time-average retinal illuminance, 7), provided by the stimulus is often referred to as the
“adaptation level.” The quantity m is called the “modulation factor” (or simply “modulation’’).

by sinw?t. The amplitude A and the phase ¢ are
functions of the frequency w (expressed in radians)
of the input. As visual stimuli cannot have nega-
tive components of radiant power, the input func-
tion sin wt is replaced by a waveform f(t) consis-
ting of a steady-state component, providing a
retinal illuminance 7(t) = 7) =const., and a
sinusoidally modulated component with an am-
plitude corresponding to a retinal illuminance
mT, with 0 < m < 1. Thus,
f(t)=Th(1+msinwt) [2(7.9.3)]
Figure 4(7.9.3) illustrates the input waveform de-
scribed by Eq. 2(7.9.3).
The use of the threshold of modulation (m) as
a response criterion ensures that the visual system
is perturbed to the psychophysically smallest
measurable extent, a range within which it must
certainly behave as a linear system.
The bulk of the earlier flicker data, such as
those shown in Figures 1(7.9.3) and 2(7.9.3), are
based on the use of the technically easily gener-
ated on-off alternation of the stimulus that pro-
vides a square waveform. The use of the sine
waveform, though sometimes more difficult to
generate, has obvious analytical advantages over
the square waveform. By fixing the adaptation
level provided by the average retinal illuminance
T, and then measuring the modulation-threshold
curve, also known as frequency-response curve, is
obtained for the given adaptation level. Experi-
ments with square waves, on the other hand, yield
but a single point at the CFF for a given level of
adaptation.
Figures: 5(7.9.3) and 6(7.9.3) show modu-
lation-threshold curves obtained by de Lange Dzn
(1958) for 2° “white” stimuli flickering with dif-
ferent waveforms (one near-sinusoidal, three non-
sinusoidal), and by Kelly (1961) for a very large
(65°) “white” stimulus flickering with a sinusoidal
waveform. Kelly’s data are also given numerically
in Table 1(7.9.3). The large-field data of Kelly
differ from the small-field data of de Lange Dzn
mainly in the region of low frequencies, where
Kelly’s data show a decrease in the sensitivity to
low frequencies. This change in sensitivity at low
frequencies has been attributed to lateral inhibi-
tion which increases with increasing field size
(Kelly, 1969; 1972).
Kelly (1961) describes his results as follows:
(a) At every adaptation level, except the lowest
(To = 0.06 td), the modulation-amplitude sensitiv-
ity (m_'= 1/modulation-amplitude threshold)
first rises to a maximum as the frequency in-
creases, then falls off sharply at still higher fre-
quencies.
(b) The sensitivity peak shifts gradually toward
higher frequencies with increasing adaptation
level. At the three highest adaptation levels, the
peak is located between 10 and 20 Hz.
(c) The relative modulation-amplitude sensitiv-
ity increases with adaptation level at all but the
 Spatial and Temporal Factors 561

0.5

Waveform
| Gasae |

el caren .2
¥ at 3

)(in
percent
Threshold
Modulation
m PAR
+ 4

100

| a 3° 4°55 10 20
Frequency (Hz)
Fig. 5(7.9.3). Modulation-threshold curves obtained by de Lange Dzn (1958) for different
stimulus conditions and a wide range of frequencies. A 2° “white” test stimulus was presented
in a 60° (nonflickering) surround of the same luminance as the average luminance of the
flickering (above CFF) test stimulus. Different (nonsinusoidal) waveforms (one to three) of the
test stimulus (see inset) lead to virtually the same results for the three given adaptation levels
(430, 43 and 4.3 td) and frequencies above 10 Hz, if the amplitude of the first Fourier
component is the same for each waveform.

lowest frequencies; thus, each curve
6(7.9.3) tends to lie above those corresponding to
lower adaptation levels.
(d) The total frequency bandwidth increases with
adaptation level at roughly the same rate as the
frequency of maximum sensitivity, so that the
cutoff frequency, which corresponds to the CFF,
is about 4.5 times the frequency of maximum
sensitivity at all levels (except the lowest).
(e) The lowest slope of the high-frequency cutoff
in the logarithmic plot remains nearly constant,
increasing slightly with adaptation level.
(f) The curves in Figure 6(7.9.3) for every adap-
tation level except the lowest are practically su-
perimposed at frequencies below 4 Hz, apparently
approaching a common low-frequency asymptote
in Figure at m_' = 15 (approximately). This behavior is in
accord with the Weber—Fechner law found to be
valid for other threshold data involving steady-
state or pulsed stimuli applied within a certain
range of observing conditions. However, the ac-
tual value of the Weber—Fechner fraction varies
with field size.
The modulation-threshold curves for very low
adaptation levels, as exemplified by Kelly’s (1961)
curve for 0.06 td, do not fall into the same pattern
of curves that apply to moderate and high adapta-
tion levels. It has been suggested that the
frequency-response characteristics of the rod
mechanism play the dominant role at such low
levels.
562 VISUAL THRESHOLDS

0.005 200

foxe}| 100

850 td
0.02 50

.
na
-_
20
=
[
=
@®
(¢p)

lo 6
Cc
bp |
(m)
Threshold
Modulation 0.06 td wm
°o
s
0.2 5

0.5 2

|
2 5 10 20 50
Frequency (Hz)

Fig. 6(7.9.3). Modulation-threshold curves obtained by Kelly (1961) for a large (65°) “white”
stimulus sinusoidally modulated over a wide range of frequencies and for different adaptation
levels (td). The numerical values of a number of experimental data points on the basis of which
the curves have been plotted are given in Table 1(7.9.3).

By replotting the data given in Table 1(7.9.3)
and illustrated in Figure 6(7.9.3) so as to show
them in terms of absolute modulation-amplitudes
(mT,), expressed in trolands, or in terms of abso-
lute modulation sensitivity (1/m7)), additional
relationships are readily revealed. Figure 7(7.9.3)
is such a plot whose curves are of the same shape
as those shown in Figure 6(7.9.3) but each has
been shifted downward by an amount propor-
tional to log 7). The order in which the shifted
curves appear in the figures is now reversed; that
is, in absolute units, the modulation sensitivity is
greatest at the /owest adaptation level.
The important feature exhibited in Figure
7(7.9.3) is the common asymptote of all curves at
high frequencies. This asymptote represents the
high-frequency cutoff over nearly the entire pho-
topic range of retinal illuminance. The absolute
modulation amplitude (m7)) uniquely determines
the high-frequency threshold value. No matter
what the adaptation level, the visual mechanism
that controls adaptation is incapable of attenuat-
ing a threshold response to flicker that is mod-
ulated at 100% (or nearly so).
Modulation-amplitude threshold data such as
those shown in Figure 7(7.9.3) do not yield the
complete transfer function G(w) that would be
required in the linear model [Figure 3(7.9.3)] to
predict also transient visual phenomena, such as
thresholds obtained with pulsed incremental
stimuli (see, for example, Section 7.4). In addition
to the amplitude-threshold data provided by Fig-
ure 7(7.9.3), data on phase-shift sensitivity are
required. However, such data are difficult to be
obtained by psychophysical techniques (Kelly,
1972).
The phenomenon usually referred to as chro-
matic flicker has also been studied extensively.
Chromatic flicker may be generated by presenting
two stimuli of different colors in a periodically
alternating mode. The appearance of the visual
field will depend, apart from size, retinal location,
 Spatial and Temporal Factors 563

Table 1(7.9.3) Relative Modulation-Amplitude Sensitivity m~! (=1 / Modulation-Amplitude Threshold)

for a Large (65°) “White” Stimulus Flickering with a Sinusoidal Waveform at Different Frequencies,
for One Observer (DHK) Adapted to Different Levels 7, (Trolands) (from Kelly, 1961)
ee
Frequency Adaptation Level J) (Trolands)
(Hz) 9300 850 Wy 721 0.65 0.06
eee
1.6 ee 15.4 17.8 17.3 1:2 5.86
2.7 Eke: 18.4 Na) 19.8 Eid 6.11
4.0 26.7 33e2 32.9 ae 22.) 6.27
Del 33.8 54.1 45.3 40.0 222 6.29
6.8 54.9 81.6 DUS B27): 22.0 4.58
8.8 81.3 114.5 23 50.5 16.5 3.43
10.7 86.8 120.8 96.7 ayRe Pz 2.43
12.7 104.0 1328 101.5 41.1 dod 1.42
16.2 f22.5 1323 iD 24.7 2.92
19.8 114.0 119:3 78.8 14.9 1.24
24 117.0 105.8 59.8 6.19
28 913 82.2 29.9 3.08
32 88.5 13,3 23.0 2.09
40 48.0 38.1 5.89
54 195 8.7 E23
64 95 32
75 2.74

nonflickering conditioning stimuli, and luminance
level, on the two colors of the stimuli seen under
the steady-state condition and the rate of alterna-
tion. At a very low rate of alternation, the ob-
server will see a distinct repetitive change from
the first color to the second and back to the first.
With an increasing rate of alternation, the ob-
server begins to see not only the two original
colors, but also colors of intermediate hue, satura-
tion, and brightness; and when the rate is further
increased, the two colors will fuse to a single color
which is the color of the additive mixture of the
two component stimuli.
At the point at which chromatic fusion occurs,
that is, the critical chromatic flicker frequency
(CCFF) is reached, there may still remain a per-
ceptual brightness flicker (however, see also
below). For a given pair of color stimuli, the
observed CCFF is highest when the luminances of
the steady-state component stimuli are the same;
it decreases with increasing luminance difference
between the two stimuli (Truss, 1957). The CCFF
also depends on the magnitude of the chromatic
differences of the flickering component stimuli
(van der Horst, 1969).
The difference between CCFF and CFF can
be minimized or eliminated if the phase relation
of the two component stimuli is appropriately
altered. This interesting dependence between
CCFF and CFF was demonstrated by de Lange
Dzn (1958) who worked with component. stimuli
of different wavelengths alternated sinusoidally in
time such that their phase relation was initially
180°. In general, brightness flicker could not be
eliminated just above the CCFF by a mere adjust-
ment of the luminance of one of the component
stimuli. However, if the initial 180° phase relation
was altered by advancing the phase of the compo-
nent stimulus of the shorter wavelength, the per-
ception of flicker could be eliminated. Figure
8(7.9.3) illustrates de Lange Dzn’s results. The
phase shift required is interpreted as directly re-
lated to a phase shift (differential latency) in the
visual response systems mediating brightness
(luminance) and chromatic perceptions. For some
further work concerning the phase relation, see
Walraven and Leebeek (1964).
De Lange Dzn’s (1958) study also includes a
comparison of chromatic flicker and brightness
flicker, as illustrated in Figure 9(7.9.3). The chro-
matic-flicker data are obtained by alternating a
monochromatic stimulus of wavelength 615 nm
with another monochromatic stimulus of wave-
length 549 nm and introducing an appropriate
564 VISUAL THRESHOLDS

10

cra 0.06 td

0.02
8

0.05
— 6 Hz

26
2
mo
0.2 ®
no}

S
4
0.5 dq

14 Hz

0 :
400 500 600 700
Wavelength A (nm) 689

(1/mT,
Fig. 8(7.9.3). Phase shift Aq required for a
)Sensitivity
Modulation
20 0.05 monochromatic stimulus of shorter wavelength A
in order to eliminate flicker when that stimulus is
alternated with a monochromatic stimulus of wave-
)
J,
trolands
in
(mT
Threshold
Amplitude
Modulation
-
850 td length A = 689 nm. One set of data was taken at
an alternation rate of 14 Hz, the other at 6 Hz. In
both cases, the retinal illuminance was kept at 45
td (de Lange Dzn, 1958).
200 0.005

0.002 of wavelength 595 nm. At high frequencies, the

9300 td
| 2 5 10 20 50
Frequency (Hz)
Fig. 7(7.9.3). Modulation-threshold curves ob-
tained from Table 1(7.9.3) whose values have been
converted to absolute units. The curves are identi-
cal in shape to those given in Figure 6(7.9.3) but
have been shifted downward by amounts propor-
tional to their respective values of log 7p.
phase shift (A@) between the two stimuli to com-
pensate for the phase shift in the response sys-
tems mediating brightness and chromatic percep-
tions as discussed above in connection with Fig-
ure 8(7.9.3). The brightness-flicker data are ob-
tained by modulating the luminance of the two
monochromatic stimuli of 615 and 549 nm, hav-
ing an “orange” color corresponding approxi-
mately to the color of a monochromatic stimulus
response to chromatic flicker is inferior to the
response to brightness flicker. At 100% modula-
100 tion and an adaptation level of 285 td, the CCFF
is 28 Hz with a phase shift of Ag = 5.9°; the
corresponding CFF for brightness flicker is 48
Hz. At low frequencies, there is no attenuation of
sensitivity of the chromatic-flicker response. The
marked differences between the two sets of data,
particularly in the moderate to high frequency
range, are consistent with observations made in
flicker photometry which is used to elicit bright-
ness (luminance) responses from only the achro-
matic channel of visual perception with no, or
only a minimal, intrusion from the opponent
chromatic channels (Section 5.7).
At moderate to low frequencies (< 10 Hz), the
sensitivity of the visual system to the brightness
(luminance) sinusoidally modulated stimulus de-
pends on the spatial configuration (field pattern)
of the stimulus. When the slowly flickering
stimulus is displayed uniformly (no pattern) in a

Chromatic flicker
)(in
Threshold
Modulation
percent
m
aes Oe ee,
Frequency
Fig. 9(7.9.3). Modulation threshold curves obtained by de Lange Dzn (1958) for phase-shift-
corrected chromatic flicker (615 nm against 549 nm), and for brightness flicker (mixture of 615
nm and 549 nm).
large field with no surround, the sensitivity is
lowest; but it is significantly higher in a patterned
field whose different parts flicker in opposite
phase (Kelly, 1969). The sensitivity to brightness
flicker reaches maximal values when a rectilinear
counterphase grating is used with stripes of angu-
lar subtense of approximately 10 min. However,
even a single edge in the field, as found, for
example, in the familiar bipartite visual field, is
quite effective in raising the sensitivity signifi-
cantly when the two halves of the field flicker in
counterphase.
In extending his study on brightness flicker to
chromatic flicker, Kelly (1975) discovered that
counterphase chromatic flicker in a bipartite
visual field has opposite effects to counterphase
brightness flicker in the same field. Figure
10(7.9.3) illustrates the spatial configurations of
the stimuli used by Kelly and the sequences of
their presentation. Figure 11(7.9.3) illustrates the
results obtained for the four variations of Kelly’s
experiments. The sensitivity to brightness flicker
Spatial and Temporal Factors 565
Brightness
flicker
10 20 30 40 50 100
(Hz)
in the counterphased bipartite field is increased,
in the low-frequency range, by a factor of about
six compared to the sensitivity to brightness flicker
in the uniform field. The opposite effect occurs
for chromatic flicker; the sensitivity is decreased
by a factor of about three when the uniform field
is changed to the counterphased bipartite field.
Kelly and van Norren (1977) extended the
earlier chromatic flicker studies and offered a
“two-band model” explaining the main features
of the experimental results. Some of the results
obtained by Kelly and van Norren are sum-
marized in Figures 12(7.9.3) and 13(7.9.3).
The upper part of Figure 12(7.9.3) illustrates
the most general waveform configuration of
the chromatic stimuli available to Kelly and
van Norren (1977) in a 1.8° uniform field with no
surround viewed through a 2.3-mm artificial pupil.
The left (a) and nght (b) diagrams depict two
different modulation settings at the same average
luminance (dashed line), illustrating how the two
out-of-phase chromatic stimuli [S(R) and S(G)]
566 VISUAL THRESHOLDS

Brightness (Luminance L,<«+L5) Flicker

Experiment
| !

Experiment
2

Chromatic (R<*G) Flicker

10

! '

TONeTO)
DO'BE
SO
PO
Experiment S(R)
Be

Experiment
4 : :
Fig. 10(7.9.3). Field configurations of stimuli and sequence of presentation in Kelly’s (1975)
experiments on brightness (luminance) flicker and chromatic flicker. In experiments 1 and 3, a
uniform field is flickered. In experiments 2 and 4, the two halves of a bipartite field are
flickered in counterphase; as the left half is changed from L , to L, (or from R to G), the right
half is changed from L, to L, (or from G to R), and vice versa. In all cases, sinusoidal
modulation is used. The chromaticities of the chromatic stimuli presented in experiments 3 and
4 varied sinusoidally about the chromaticity of a yellow stimulus. At full modulation, the
extreme chromaticities were a saturated red (R) and a saturated green (G) stimulus, respec-
tively. Their luminance ratio was adjusted so as to minimize low-frequency brightness flicker. A
10° visual field providing a retinal illuminance of 1000 td was used in both the brightness and
chromatic flicker experiments, viewed through a 2.3 mm artificial pupil.

could be adjusted by the observer. The waveforms
are plotted cumulatively in order to show the
total luminance of the field. In the case il-
lustrated, the amplitude of the green stimulus
[S(G)] is greater than the amplitude of the red
stimulus [S(R)]. The modulation of the ampli-
tudes of the green and red stimuli can only be
changed simultaneously and proportionately,
whereas the nonflickering yellow background
stimulus [S(Y)] is appropriately changed to keep
the average luminance at a constant level (equiva-
lent to 860 td).
The lower part of Figure 12(7.9.3) illustrates
three special cases of the general waveform con-
figuration shown in the upper part. Case (1) il-
lustrates strictly chromaticity modulation at con-
stant luminance; the red stimulus [S(R)] replaces
the green stimulus [.$(G)] and vice-versa. Case (2)
illustrates strictly brightness modulation after
having changed the phase of one, for example, the
green, of the two chromatic stimuli into the same
phase with the other (red) stimulus. The resulting
flickering stimulus appears yellow. Case (3) is
obtained by replacing the green stimulus of Case
(2) by a red stimulus of equal luminance and
adding it to the existing red stimulus, thus pro-
ducing a flickering stimulus that includes both the
brightness-flicker component of Case (2) and the
chromatic component of Case (1), modulated on
the same steady-state yellow background.
Figure 13(7.9.3) shows the modulation thresh-
old curves Kelly and van Norren (1977) obtained
for the three experimental conditions illustrated
in the lower part of Figure 12(7.9.3). The sensitiv-

0.002
fe)ro)fe)
fo) re) re)
Modulation
Threshold
Oo nN3)
0.500
| 2 5 10 20
Frequency (Hz)
Fig. 11(7.9.3). Modulation-threshold curves ob-
tained by Kelly (1975) for brightness and chro-
matic flicker under the experimental conditions
summarized in Figure 10(7.9.3). Curve 1 (experi-
ment 1): brightness flicker in uniform field. Curve
2 (experiment 2): brightness flicker
terphased bipartite field. Curve 3 (experiment 3):
chromatic flicker in uniform field. Curve 4 (experi-
ment 4): chromatic flicker in counterphased bipar-
tite field.
ities of the visual system to brightness and chro-
matic flicker (Cases 1 and 2) govern two separate
frequency bands [recall also de Lange Dzn’s re-
sults illustrated in Figure 9(7.9.3) and Kelly’s
results shown in Figure 11(7.9.3)]. At low fre-
quencies, the sensitivity to chromatic flicker is
considerably greater than that to brightness
flicker; the reverse is true at high frequencies. The
maximum sensitivities measured by Kelly and van
Norren (1977) are about the same for both types
of flicker, but occur at distinctly different fre-
quencies. For the Case (3) condition which in-
cludes both the brightness-flicker component of
Case (2) and the chromatic-flicker component of
Case (1), the modulation-threshold curve, repre-
sented by solid dots in the right-hand part of
Figure 13(7.9.3), follows the upper envelope of
the Case (1) and Case (2) curves. It appears that
Discrimination Thresholds 567
the sensitivity for the complex Case (3) is governed
by the flicker component having the greater sensi-
tivity at each tested frequency: below 5 Hz, this is
the chromatic-flicker component, above 5 Hz, it is
the brightness-flicker component.
Additional useful experimental data on chro-
matic and brightness flicker and various theoreti-
cal explanations of such data have been given,
among others, by van der Horst (1969); van der
Horst and Bouman (1969); Estévez and Spekreijse
(1974); Boynton and Baron (1975); Estévez and
Cavonius (1975); Kelly, Boynton, and Baron
(1976); Boynton (1979), and Wisowati and Boyn-
ton (1980).
7.10 DISCRIMINATION THRESHOLDS
Color discrimination has been studied in various
special investigations, not using material samples,
as 1s usual in color-scaling experiments and in the
50 development of color-order systems (Chapter 6),
but by presenting to the observer a bipartite
visual field [see Figure 1(7.10)], the two parts of
which are adjusted to be just perceptibly different
in some special quality such as hue or brightness.
Some of the principal data of this kind are sum-
marized in the following. These data are of a
somewhat different type than those we have dis-
in coun-
cussed in the previous sections of this chapter.
They are concerned primarily with measuring an
observer’s powers of color discrimination and less
concerned with elucidating the properties of
color-vision mechanisms.
7.10.1 Luminance Differences
The classical luminance-difference measurements
by Konig and Brodhun (1889) and later similar
ones by Blanchard (1918), Lowry (1931), and
Steinhardt (1936) are of the type to be considered
in this section; a representative result obtained by
the later investigators is reproduced in Figure
1(7.10.1). The quantity log(AL/L) is plotted
against log L. The curve shown refers to observa-
tions with “white” stimuli and with field sizes
larger than the rod-free area. It exhibits a sharp
bend at a luminance L of roughly 5 X 10°? cd -
m ~*. This apparent break signifies that at lumi-
nance levels below 5 X 10-3 cd - m 7”,judgments
of luminance differences A L were effectively made
by using parafoveal vision and thus, strongly in-
troducing the rod mechanism. Above this level,
foveal vision was dominant. For luminance levels
 Retinal

(td)
Illuminance

Time —>

- S(G)
oe S(R) S(R)
~— | s(G) S(R)
— $
EE FRAY TN
<1CC
-—-

ae S(Y) S(Y) S(Y)

Time —~

(1) (2) (3)

Fig. 12(7.9.3). Upper part: General waveform configuration of flickering chromatic stimuli
[.S(R) and S(G)] superimposed on steady-state yellow stimulus [.S(Y )]. Left (a) and right (b)
diagrams show two different amplitude modulations keeping the average retinal illuminance
(dashed line) constant. Lower part: Three special cases of the general waveform configuration
given in the upper part (Kelly and van Norren, 1977).

0.005

0.01

0.02

0.05

0.1

0.2
Threshold
Modulation Modulation
Threshold
0.5

0.2 O85 2 5 10 2 50 0.2 0:3) cl 2 5 \OnieC 50

Frequency (Hz) Frequency (Hz)
Fig. 13(7.9.3)._ Modulation-threshold curves obtained by Kelly and van Norren (1977) for the
three cases (1), (2), (3) of waveform configurations shown in the lower part of Figure 12(7.9.3):
(1) strictly chromatic (red < green) flicker; (2) strictly brightness flicker of a yellow stimulus;
(3) combination of chromatic and brightness flicker. The results for cases (1) and (2) are given
in the left-hand part of the figure by crosses and open circles, respectively, through which |
smooth curves are drawn. These curves are redrawn in the right-hand part of the figure for a
direct comparison with the results for case (3) given by solid dots.

568
 Discrimination Thresholds 569

(a)
Ve
Fig. 1(7.10). Two typical visual fields: (a) circular bipartite field with vertical dividing line; (b)
two concentric circular fields. Hatched areas indicate surrounding field.

higher than approximately 100 cd - m~’, the ratio
AL/L (Weber fraction, sometimes called Fechner
fraction, or Weber—Fechner fraction) has its
lowest value and tends to be constant. The best
values obtained by the various investigators differ
somewhat depending on the observing condi-
tions; but a good practical estimate is AL/L =
0.01. Sometimes a slight rise in log(AL/L) is
observed for very high values of L. This increase
in just-noticeable luminance difference has been
interpreted as being the result of imperfect adap-
tation of the eye to the luminance level of the
field during the experiment.
Luminance-difference measurements with mon-
ochromatic stimuli have also been made. Curves
of log(AL/L) against log L are very similar to
those obtained with “white” or “achromatic”
stimuli, especially for high values of L. The ap-
parent breaks in the curves, marking the transi-
tion from cone to rod vision are observed, and the

(AL/L)
LOG

Fig. 1(7.10.1). Discrimination of luminance differences (AL) as observed for “white” stimuli
in bipartite fields of angular sizes larger than the rod-free area.
570 VISUAL THRESHOLDS
values of L at which these occur depend on
wavelength. For stimuli of long wavelengths
(above 670 nm), the break is absent. Such varia-
tions are the result of the different spectral sensi-
tivity functions of the rods and cones. With strictly
foveal vision, the influence of the rod mechanism
is practically removed and the curve of
log(A L/L) shows no break at any wavelength.
A useful and detailed review of luminance-dif-
ference thresholds has been given by Blackwell
(1972) discussing the effects of a variety of differ-
ent experimental conditions on such data, in-
cluding the effects of spatial and temporal stimu-
lus characteristics.
7.10.2 Wavelength Differences
Discrimination of wavelength differences has usu-
ally been studied by means of a monocular col-
orimeter providing Maxwellian view of a bipartite
visual field [Figure 1(7.10); Type a]. Typical re-
sults can be found in papers by Tyndall (1933),
Wright and Pitt (1934), Weale (1950, 195la,
1953a), Thomson and Trezona (1951), Bedford
and Wyszecki (1958), McCree (1960a, b), Hurvich
and Jameson (1961), Connors (1964), and Siegel
(1965, 1969). For general reviews of wavelength
discrimination, see Birch and Wright (1961) and
Bouman and Walraven (1972).
Figures 1(7.10.2) and 2(7.10.2) illustrate
average results obtained by Wnght and Pitt
(1934) and Bedford and Wyszecki (1958). In these
cases, the wavelength discrimination was mea-
sured by setting just-noticeable differences of
wavelengths. Starting with the two halves of the
visual field at the same wavelength and lumi-
nance, the observer was required to change the
WAVELENGTH
Fig. 1(7.10.2). Wavelength discrimination curve based on observations made by Wright and
Pitt (1934).
wavelength of one half, at the same time main-
taining a brightness match between the two halves,
until a just-noticeable hue difference was seen.
Just-noticeable differences to longer and to shorter
wavelengths than the comparison wavelength were
observed in succession and denoted by (+AA)
and (—AA), respectively. The average difference
AA = [(+ AA) + (—AA)]/2 was obtained from
several repeat measurements of this kind for each
comparison wavelength A and plotted against A.
All curves have the same general appearance.
Discrimination rapidly becomes poor at both ends
of the visible spectrum; two relative maxima of
AX are observed at approximately 460 and 530
nm, and three relative minima at approximately
440, 490, and 590 nm.
Wavelength discrimination depends on the
luminance level, surround, field size, portion of
the retina used in the measurement (fovea, para-
fovea, or periphery), as well as the technique of
observation (strict fixation or scanning). The ef-
fect of field size is illustrated in Figure 2(7.10.2)
for the case in which the observer was allowed to
scan the field. It is noted that decreasing the field
size generally results in poorer discrimination, as
is to be expected. If very small fields are used and
also strict foveal fixation, discrimination becomes
extremely poor, particularly in the blue-green re-
gion of the spectrum. This phenomenon has been
attributed to the fovea being essentially tritanopic
under small-field viewing conditions. Figure
3(7.10.2) illustrates this case in accordance with
measurements reported by Willmer and Wright
(1945). The data of Willmer and Wright were
replotted to conform to the plotting technique
used in Figures 1(7.10.2) and 2(7.10.2) (Bedford
and Wyszecki, 1958). A more detailed discussion
A (nm)
 Discrimination Thresholds 571

Observer Visual Field

R.B.
100 trolands
GQ |coe
—_@_@—9—
ime be
e* 15 minutes 2000 trolands
40 minutes»
pits e 1.5 minutes 5 eS 5 eS,
= At Ar
Ar 1 Xp—
12 minutes | 500 trolands
(nm) 24minutes >
12 minutes
Lei?

Observer
G.W.

Ad
(nm)

400 500 600

WAVELENGTH X (nm)
Fig. 2(7.10.2). Wavelength discrimination curves based on observations made by Bedford and
Wyszecki (1958) for different field sizes and retinal illuminances (trolands).

on the effect of viewing techniques, that is, scan-
ning as compared to strict fixation, has been
given by McCree (1960a, b).
7.10.3 Purity Differences
Discrimination data on differences of colorimet-
ric purity or excitation purity have been recorded
by several investigators (Jones and Lowry, 1926;
Purdy, 1931b; Martin et al., 1933; Wright and
Pitt, 1937; Priest and Brickwedde, 1938; Wright,
1941; MacAdam, 1942; Kaiser, Comerford, and
Bodinger, 1976).
Additive mixtures of a monochromatic stimu-
lus of a given wavelength A and a given achromatic
(“white”) stimulus are produced in a bipartite
visual field. The luminance of both the mono-
chromatic stimulus and the white stimulus can be
varied independently, thus providing a series of
colors ranging from an achromatic color to a pure
monochromatic color. The mixtures are restricted
so as to maintain constant luminance throughout
ae VISUAL THRESHOLDS

Visual Field

as minutes
as
Approx. 100 trolands

Ad
(nm)

400 500 600 700

WAVELENGTH )\ (nm)
Fig. 3(7.10.2). Wavelength discrimination curve based on observations made by Willmer and
Wright (1945), using a small field and strict foveal fixation (from Bedford and Wyszecki, 1958).

the series. The observer is required to determine
the just-noticeable differences or equally noticea-
ble differences of purity by changing the color of
one half of the visual field (for example, Martin
et al., 1933). An alternative method is to make
repeated color matches and deduce the just-
noticeable difference from the standard deviation
of the matches (MacAdam, 1942).
Two scales of purity have been used. One
refers to excitation purity, the other to colorimetric
or spectral luminance purity; both quantities have
been defined in Section 3.4. Figure 1(7.10.3) gives

A =650nm
"white’' =4800 K
2° Field, 30 trolands

)
(PERCENT
c

Ap

MEAN COLORIMETRIC PURITY p, (PERCENT )

Fig. 1(7.10.3). Discrimination of colorimetric purity p, as observed by Martin et al. (1933) for
stimuli of constant dominant wavelength A = 650 nm.

an example of results, obtained by Martin et al.
(1933), of just-noticeable differences (A p.) in col-
orimetric purity at various purity values p. be-
tween a “white” stimulus (correlated color tem-
perature = 4800 K) and a monochromatic stimu-
lus of wavelength A = 650 nm (see also Wright,
1946). The visual field subtended two degrees,
observations were made at approximately 30 td
with the eye pre-adapted to the “white” stimulus
of the same luminance. The difference A p. first
tends to increase in going from “white” towards
the spectral color, reaches a maximum at some
intermediate purity, and decreases somewhat at
higher purities. This tendency is reported to be
typical for practically all wavelengths. The preci-
sion of the measurements, not indicated in Figure
1(7.10.3), is of the order of +10 or 20%.
Data of this kind have been used to determine
the number of just-noticeable steps between a
“white” stimulus and a monochromatic stimulus
of varying wavelength. Figure 2(7.10.3) gives an
average derived from measurements made by
Martin et al. (1933) and Jones and Lowry (1926).
A very marked minimum occurs at about 570 nm;
the number of steps increases rapidly on either
side of this wavelength and reaches a constant
value at both ends of the spectrum.
20 ii
oa
JUST-NOTICEABLE
OF
NUMBER
(4800
kK)
STEPS
WHITE
BETWEEN
FIELD)
(2-DEGREE
COLOR
SPECTRAL
AND
500
WAVELENGTH
Fig. 2(7.10.3). Number of just-noticeable steps between a “white” stimulus (4800 K) and a
monochromatic stimulus as a function of wavelength, A. (From Martin et al., 1933, and Jones
and Lowry, 1926.)
Discrimination Thresholds a3
Some investigators have concerned themselves
with the determination of the size of the first
just-noticeable differences in colorimetric purity
from a given “white” in the direction of a given
spectral color of wavelength A. Figure 3(7.10.3)
gives average data for this first step from white,
derived by Wright (1946) on the basis of measure-
ments made by Priest and Brickwedde (1938),
Martin et al. (1933), and Wright and Pitt (1937).
This average may be considered representative for
a “white” of correlated color temperature 4800 K
observed in a 2° visual field. In Figure 3(7.10.3),
the logarithm of the inverse of Ap., for the first
step is plotted against the wavelength. A pro-
nounced minimum is found at approximately 570
nm. This implies that a monochromatic stimulus
of wavelength 570 nm is in some sense less
saturated than a stimulus of any other wave-
length.
The determination of the first just-noticeable
difference from a monochromatic stimulus of
wavelength A toward a specified “white” stimulus
appears to be a more difficult task than the
reverse task going from “white” toward a given A,
described above. Wright and Pitt (1937) find the
first just-noticeable difference from A toward
“white” approximately constant in size over most
600 700
A (nm)
574 VISUAL THRESHOLDS

og 8
x
ro)
12
tite
9
st
=
ae we20 A
PE on
o x
ce sf
be
oc Oo
OMe tS
~~ a
gw

uJ
G)
oO = 1.0
to 4g
WW
oO
KE
ro)
z
> wo
600 700
WAVELENGTH A (nm)
Fig. 3(7.10.3). Log (1/4 p,) of first just-noticeable step (A p,) from “white” measured in
different directions which are indicated by A (from Wright, 1946).

of the spectrum; but a more recent study by
Kaiser, Comerford, and Bodinger (1976) yields
significantly different results. Kaiser et al. (1976)
find that for most of their observers, the size of
the first just-noticeable step from A towards
“white,” when plotted against A gives a curve
resembling in shape the one obtained for the size
of the first just-noticeable step from “white” to-
ward A, shown in Figure 3(7.10.3), with a pro-
nounced minimum at about 570 nm. However,
the range of the size of the just-noticeable dif-
ference from A toward “white” is generally nar-
rower than that of the just-noticeable difference
from “white” toward 4X.
The data shown in Figure 3(7.10.3) can be
transformed to the CIE 1931 (x, y)-chromaticity
diagram and compared with the corresponding
color-matching ellipse centered at the chromatic-
ity point of the “white” stimulus (Section 5.4). If
one assumes that the radii of the color-matching
ellipse in the directions of the spectral colors are
proportional (by the same factor) to the just-
noticeable differences from “white”, one would
expect to find the chromaticity locus derived from
Figure 3(7.10.3) to agree with the color-matching
ellipse in all those directions (A = 450 to 670 nm)
covered by the data of Figure 3(7.10.3). However,
as shown in Figure 4(7.10.3), the agreement is not
very good and it is difficult to explain the dis-
crepancies. If one takes the discrepancies at their
face value, one could argue that the precision
(statistical standard deviations) of color match-
ing, on which the color-matching ellipse is based,
is not related in a simple way with just-noticeable
color differences (Parra, 1972). On the other hand,
if one takes into account the possible observa-
tional uncertainties that underly both the color-
matching ellipse (Sections 5.4.4 and 5.4.5) and the
locus of just-noticeable differences from “white”,
the discrepancies between the two sets of data
shown in Figure 4(7.10.3) may not be as signifi-
cant as they may appear at first glance.
7.10.4 Color-Temperature Differences
A change of the color temperature of an
incandescent lamp corresponds to a change in
chromaticity along the Planckian locus in the
chromaticity diagram (Section 3.11). Since color
temperature is used to specify incandescent lamps
for colorimetry, it is important to determine the
color-temperature interval AT.. which corresponds
to a just-noticeable chromaticity difference. The
most comprehensive study on this subject has
been made by Judd (1933a). Measurements were
made with a 2° bipartite field with dark surround.
The retinal illuminance was kept at approxi-
mately 30 td regardless of the color temperature.
Results were obtained for a range of color tem-
perature from 1,800 to 11,000 K. It was found
 Discrimination Thresholds 375

0.400

0.300

0.250
0.250 0.300 0.350 0.400 0.450
Xx
Fig. 4(7.10.3). Portion of CIE 1931 (x, y)-chromaticity diagram showing the locus of just-
noticeable differences from “white” (4800 K) derived from the data plotted in Figure 3(7.10.3)
and a corresponding color-matching ellipse centered at the same “white.” The color-matching
ellipse has been taken from the Wyszecki—Fielder (1971) study (observer AR, ellipse no. 16
shifted to the 4800-K chromaticity point and its axes enlarged by a factor of 5), described in
Section 5.4.3.

that under these observing conditions, A(10°/T.)
is practically independent of T. and has a value of
approximately 5.5 reciprocal megakelvin (see Sec-
tion 3.11). The difference of 5.5 MK! decreases
if more favorable observing conditions are em-
ployed, such as higher retinal illuminance and a
larger visual field.
7.10.5 Chromaticity Differences
Wright’s (1941) data constitute the main body of
experimental evidence collected on just-noticeable
chromaticity differences. The color-matching el-
lipses of MacAdam (1942), Brown and MacAdam
(1949), Brown (1951, 1952a,b, 1957), and
Wyszecki and Fielder (1971) are often considered
to be directly related to just-noticeable chromatic-
ity differences, but strictly speaking, the color-
matching ellipses are standard deviations derived
from repeated complete color matches and not
just-noticeable chromaticity differences. Thus, we
have treated them separately in Section 5.4 deal-
ing with the precision of color matching.
Some of the results obtained by Wright (1941)
are illustrated in Figure 1(7.10.5). Wright adhered
to the method in which an observer sets the two
halves of a bipartite field so that, in the observer’s
judgment, they differ in chromaticity by some
small constant step (but rather larger than the
just-noticeable step), while the two brightnesses
are kept the same. This measurement was made
along the spectrum and along a number of chro-
maticity paths generated by varying the propor-
tions in a mixture of a pair of strongly colored
stimuli. The chromaticity interval defined by each
dash in the dashed lines was found by Wright to
correspond to a chromaticity difference judged to
be the same as that for every other dash. From
the figure, it is quite evident that the CIE 1931
chromaticity diagram is far from being a uniform
chromaticity diagram; the “green” region (upper
portion of the diagram) is spaced out, compared
with the “bluish-purple” region (lower left por-
tion).
The measurements were made on Wright’s col-
orimeter (see Section 5.15), using a 2° visual field
576 VISUAL THRESHOLDS

0 0.2 0.6 0.8

x

Fig. 1(7.10.5). CIE 1931 chromaticity diagram showing chromaticity differences of perceptu-
ally equal size; Wright’s dashes (from Judd and Wyszecki, 1975).

with dark surround. Four observers participated

and the retinal illuminance was kept, on the aver-
age, at approximately 100 td.

7.10.6 Color-Difference Matches

The work of Wyszecki (1965, 1972) and Wyszecki

and Fielder (1971) on color-difference matching
differs in an important way from the conven-
tional studies on color discrimination discussed
previously. Instead of determining just-noticeable
color differences, the observer is now required to
discriminate between color differences and, in
particular, adjust them so as to make them appear
equal in magnitude. The color differences pre-
sented to and matched by the observer are not
necessarily small or of near-threshold size, but
can be many times threshold. The criterion of 40°
equality of color difference requires the observer’s Fig. 1(7.10.6). Configuration of visual field used
ability to discriminate between color differences by Wyszecki and Fielder (1971) in their experi-
of comparable size generated by pairs of color ment on color-difference matching. The color
stimuli of different chromaticities. stimuli presented in fields (7), (j), and (i) dif-
Figure 1(7.10.6) illustrates the visual-field con- fered in chromaticity but were of constant lumi-
figuration used by Wyszecki and Fielder (1971) nance (~ 12.cd - m 7”). The “white” surround had
obtained on Wyszecki’s seven-field colorimeter a luminance of approximately 6 cd - m~”.

0.360 0.380 0.400
0.360 0.380 0.400
x
Fig. 2(7.10.6). Portion of (x, y,/)-space shown
in its two main projections: the (x, y )-chromatic-
ity plane and the (x, /)-plane where / is related to
the luminance L of the test stimuli by /= 0.2
log ,9(0.2L). The color stimuli (3) and (4) are
given both in chromaticity and in luminance; color
stimulus (3,4) is found by the observer so as to
make the chromaticity differences between (3, 4)
and (3), and between (3,4) and (4) perceptually
equal to the given difference between (3) and (4).
The brightness of (3, 4) is set by the observer to be
equal to (3) and (4). Each cross indicates a setting
made by the observer; there are 30 such settings
whose mean is indicated by a solid dot. The el-
lipses shown in the two planes are projections of
the color-difference-matching ellipsoid calculated
by familiar statistical methods (Section 5.4) from
the 30 settings made by the observer.
described in Section 5.15. Three adjacent fields
(7), (/), ()] are used. In fields (7) and (7), two
color stimuli are presented of different chromatic-
ity but equal brightness (luminance). In field (i),
a variable color stimulus is presented whose color
the observer adjusts until it is perceived to have
the same brightness as (i) and (/) and a chro-
maticity such that the perceived color differences
between (ij) and (i) and between (ij) and (/) are
the same as the given difference between (i) and
(j). When the chromaticities of the three color
Discrimination Thresholds 571
(29)
0.450
(29, 30)
0.430
0.410
0.330 0.340 0.350 0.360
Fig. 3(7.10.6). Portion of CIE (x, y)-chromatic-
ity diagram showing cross sections (dashed lines)
of four color-difference matching ellipsoids ob-
tained by the same observer (GW) in four separate
sessions. In each session, 30 different settings
were made, each yielding a perceptually equilateral
triangle with corner points at (29), (30), (29,30).
The small solid dots represent the mean chromatic-
ity points obtained in the four sessions. The large
solid dot and the ellipse drawn as a solid line
represent the mean of the four sessions and associ-
ated composite ellipse derived from the data of the
four sessions.
stimuli (i), (7) and (ij) are plotted in the CIE
1931 (x, y)-chromaticity diagram, a triangle is
obtained whose sides represent chromaticity dif-
ferences of perceptually equal size. The setting of
(ij) is not unique as there exists another possibil-
ity of setting (ij) on the opposite side of the given
side [(),(/)] of the triangle. However, in general,
the observer was instructed to place the setting of
(ij) such that a “clockwise” change in chromatic-
ity was obtained when going from (i) to (/) to
(ij) and back to (i). This procedure, which was
not difficult to follow, ensured that the three
observers who took part in the experiment ob-
tained perceptually equilateral triangles in the
 578 VISUAL THRESHOLDS

(41,51)

29
(30,28)
// 52 6 3
/ qs, 82:42) 28
/
56 (29,30) “30
5 4
(45,56)
(5,6) (3,4)
46
57

(54,55)
(46,57)

60 ‘op ( 34.29) ie | SIG)

(60,61) 5 = B26
(35,34) 93 (13,14) (26,27)
347 (37,34) 8 (25,26)
(17,18) (39,34)
33
(34,35) (62,63) 62 (58,59)
58
27
SK ate
(33,31) (34,37)
37 20 Do ee —_— ~
J ae (21,22)
33 19 Z(\9,20)
36
(31,32) (20,36)
——

0.2 OS 0.4 x 0.5 0.6 0.7

Fig. 4(7.10.6). Portion of CIE 1931 (x, y)-chromaticity diagram showing perceptually equi-
lateral triangles obtained by observer GW. For each triangle, the open circles represent the
chromaticity points of the given color stimuli (7) and (/); the solid dot represents the observer’s |
mean chromaticity setting of color stimulus (i, /). Different triangles are not necessarily of i
equal size; for details, see text. The points (R), (G), (B) represent the chromaticities of the
primary stimuli of the visual colorimeter used in the study. Point S represents the chromaticity
of the surround.

chromaticity diagram that could be directly inter- servations (3,4). The colordifference-matching el-
compared. lipsoid is centered at the mean chromaticity and
Figure 2(7.10.6) illustrates the results obtained luminance setting and its projections into the
by observer GW in 30 trials of finding the percep- (x, y)-plane and (x, /)-plane of the (x, y, /)-space
tually equilateral triangle (3), (4), (3,4), located become ellipses. The calculation of the color-
near the center of the CIE (x, y)-chromaticity difference-matching ellipsoid follows the same
-diagram. Whereas the color stimuli (i = 3) and procedure used in calculating color-matching el-
(j =4) have fixed chromaticities [(x3, y;) and lipsoids described in Section 5.4. For conveni-
(x4, ¥4)] and the same luminance (L; = L, = 12 ence, the luminance dimension / has been scaled
cd-m 7”), the chromaticity (x34, y34) and lumi- in accordance with / = 0.2 log, (0.2 L).
nance L,, of the color stimulus (3,4) set by the Figure 3(7.10.6) illustrates the repeatability of
observer varied somewhat from observation to the mean settings (i, /) for i= 29, 7 = 30 and
observation but permitted the calculation of a the associated cross sections of the color-dif-
“color-difference-matching ellipsoid” containing, ference-matching ellipsoids obtained by the same
on the average, 95% of any random set of ob- observer (GW) in four separate sessions but un-
 Discrimination Thresholds 579

\
(1,12) 2S
N
S
SS
~
oS
’ SS ne
(1,2) (53,4 3)
Ss
N

(26,27)

0.2 0.3 0.4 0.5 0.6 0.7

X

Fig. 5(7.10.6). Portion of CIE 1931 (x, y)-chromaticity diagram showing the color-
difference-matching ellipses associated with each of the 33 perceptually equilateral triangles
shown for the same observer (GW) in Figure 4(7.10.6). The axes of each ellipse have been
drawn 2.5 times their actual sizes.

der identical observing conditions. The different
ellipses (dashed lines) show a fair degree of varia-
bility in size, orientation, and eccentricity, but
their centers are closely clustered. The ellipse
shown as a Solid line is a “composite” ellipse
derived from the data defining the four observed
ellipses; its center is indicated by the large solid
dot.
Figure 4(7.10.6) illustrates the set of 33 percep-
tually equilateral triangles obtained by Wyszecki
and Fielder (1971) for observer GW, plotted in
the CIE 1931 (x, y)-chromaticity diagram. Open
circles indicate the chromaticity points of given
color stimuli; solid dots are the mean chromatic-
ity points obtained by observer GW to produce
for every pair of given color stimuli (i) and (/), a
third stimulus (i, 7) yielding a perceptually equi-
lateral triangle. Details for the triangles (3), (4),
(3,4) and (29), (30), (29,30) have been given in
Figures 2(7.10.6) and 3(7.10.6), respectively.
Although each of the triangles shown in Figure
4(7.10.6) has sides that are perceptually equal in
length for observer GW, the sides of one triangle
are not necessarily equal to those of the other
triangles. The relative lengths are only very
roughly the same. The base side, defined by the
two given color stimuli (i) and (/), of each of the
33 triangles was made the same length in accor-
dance with the CIE 1964 color-difference formula
giving, in each case, a color difference AE = 6.7
(see Section 6.5.2). The CIE 1964 color-difference
formula, and to some extent also the more recent
CIE 1976 formula, does not predict with entire
satisfaction perceived color differences, and it is
580 VISUAL THRESHOLDS

uncertainty domain of
color matching (i)
uncertainty domain of
matching color
perceptually equilateral
differences between
triangle
Ci) Gide slaty.

6) 3 6 9
Number (T) of j.n.d.’s between (i) and (j)
Fig. 6(7.10.6). A small portion of the (x, y)-chromaticity diagram has been linearly trans-
formed to a new chromaticity diagram (x’, y’) in which a perceptually equilateral triangle
[((7)( j)(i, j)] becomes a geometrically equilateral triangle, and a color matching ellipse with
center at (7) and a color-difference-matching ellipse with center at (i, 7)each becomes a circle.
When the color difference betwen (i) and (/), measured in number of just-noticeable
differences (j.n.d.’s), is increased, the equilateral triangle increases and with it the radius of the
circle increases linearly characterizing the precision of color-difference matching (c.d.m.).

thus not strictly possible to relate the length of
the sides of one triangle with the length of the
sides of another. However, the set of 33 triangles
includes three pairs, each consisting of twin trian-
gles with a common base side:
(a) [(34)(35)(34, 35)] and [(35)(34)(35, 34)]
(b) [(34)(37)4, 37)] and [(37)(34)(37, 34)]
(c) [(34)(39)(34, 39)] and [(39)(34)(39, 34)]
The sides of each of the above three pairs of
triangles form a regular rhombus; they represent
color differences of identical size.
- To each solid dot in Figure 4(7.10.6), repre-
senting the observer’s mean chromaticity setting
of color stimulus (i, 7), there corresponds an el-
liptical domain of uncertainty of color-difference
matching as illustrated earlier in Figure 2(7.10.6)
for (i, 7) = (3,4). The orientation and shape of
the color-difference-matching ellipse appear to be
solely dependent on the location of the center
(i, j) of the ellipse in the chromaticity diagram.
The relative position and orientation of the given
base line, defined by (i) and (/) do not seem to
have a bearing on the orientation and shape of
the color-difference-matching ellipse. However,
the perceived color difference between (7) and (/)
does control the size of the color-difference-
matching ellipse, increasing it with increasing dif-
ference between (i) and (/).
Figure 5(7.10.6) shows all 33 color-difference-
matching ellipses whose centers are identical to
the mean chromaticity settings of (i, /)-color
stimuli shown as solid dots in Figure 4(7.10.6). A
general similarity is noted between the color-
difference-matching ellipses shown in Figure
5(7.10.6) and the color-matching ellipses obtained
by the same observer at an earlier experiment
described in Section 5.4.3. In fact, the correlation
between the precision of color-difference match-
ing and that of color matching is quite high,
particularly with regard to the orientation and
shape of the corresponding ellipses resulting from
such matches made at approximately the same

location in the chromaticity diagram (Wyszecki,
1972). On the other hand, the size of the color-
difference-matching ellipse is always larger than
the corresponding color-matching ellipse ‘and its
axes appear to increase linearly with the percep-
tual size of the color differences to be matched, at
least for color differences of sizes in the range
between (slightly above) 1 and 10 just-noticeable
differences. This interesting result is schematically
illustrated in Figure 6(7.10.6) in which a linear
transformation is postulated that converts the rel-
Discrimination Thresholds 581
atively small portion of the (x, y)-chromaticity
diagram to a new (x’, y’)-diagram. In the
(x’, y’)-diagram, the perceptually equilateral tri-
angle becomes a geometrically equilateral triangle
in the Euclidean sense, and both the color-match-
ing and color-difference-matching ellipses ob-
tained for different triangle sizes become circles.
When the triangle size increases, the radius of the
corresponding color-difference-matching circle in-
creases linearly as shown in the inset of Figure
6(7.10.6).
 CHAPTER 8
THEORIES AND MODELS
OF COLOR VISION
8.1 PREAMBLE
A theory of color vision explains or helps to
comprehend one group of visual phenomena and
makes predictions about some other quite differ-
ent visual phenomena. Usually, an analogy is
employed in characterizing satisfactory explana-
tions by relating the unknown color-vision system
to known or more familiar physical and chemical
systems. The development of a theory of color
vision involves, apart from recognizing a number
of specific visual phenomena, an element of intui-
tion.
A theory of color vision, as any other theory of
science, rests on a set of assumptions about the
visual system. These assumptions attribute an in-
ner structure or mechanism that manifests itself
in other properties exhibited by the visual system.
For many years, the three-component or trichro-
matic theory of color vision played a dominant role
in color science, and a “competing” theory of
color vision, known as the opponent-colors theory
was given relatively little attention. In more re-
cent years, zone theories of color vision have be-
come widely accepted which, put in general terms,
bring together the basic ideas of the trichromatic
theory and the opponent-colors theory by confin-
ing the underlying processes postulated by these
theories to two separate but sequential zones in
which the visual stimulus arriving at the retina is
processed, converted to electrical signals, coded,
and transmitted, possibly through other process-
ing or coding zones, to the brain for interpre-
tation.
Thomas Young is usually credited with being
the first to propose a trichromatic theory of color
582
vision, though there have been others before him
who foreshadowed such a theory. For an informa-
tive collection of such early sources of color sci-
ence, see MacAdam (1970) who quotes Young as
follows:
As it is almost impossible to conceive each sensitive
point of the retina to contain an infinite number of
particles, each capable of vibrating in perfect unison
with every possible modulation, it becomes necessary to
suppose the number limited; for instance to the three
principal colours, red, yellow, and blue, and that each of
the particles is capable of being put in motion more or
less forcibly by modulations differing less or more from
perfect unison. Each sensitive filament of the nerve may
consist of three portions, one for each principal colour.
Helmholtz (1896; see also English translation
published by the Optical Society of America in
1924) developed further the ideas of Young con-
tained in the above statement. Helmholtz says
(translated into English):
(i) The eye is provided with three distinct sets of
nervous fibres. Stimulation of the first excites the sensa-
tion of red, stimulation of the second the sensation of
green, and stimulation of the third the sensation of
violet.
(ii) Objective homogeneous light excites these three
kinds of fibres in various degrees, depending on its
wavelength. The red-sensitive fibres are stimulated most
by light of longest wavelength, and the violet-sensitive
fibres by light of shortest wavelength. But this does not
mean that each color of the spectrum does not stimulate
all three kinds of fibres, some feebly and others strongly;

on the contrary, in order to explain a series of phenom-
ena, it is necessary to assume that that is exactly what
does happen.
Helmholtz then proceeds by giving a graph
showing three sensitivity curves as functions of
wavelength, which overlap substantially with one
another, but which have peaks in the red-orange,
green, and blue-violet regions of the spectrum.
These curves “may be taken to indicate some-
thing like the degree of excitation of the three
kinds of fibres.”
The trichromatic theory of color vision, also
referred to as the Young—Helmholtz three-compo-
nent theory, assumes the existence of three inde-
pendent cone types (“fibres”) with different
spectral sensitivities, and the signals generated in
these cones are transmitted directly to the brain
where “color sensations” are experienced that
correlate in a simple and direct way to the three
cone signals. The theory accounts for the experi-
mental data of foveal color-matching by means of
additive mixtures of color stimuli (Sections 3.2
and 5.5) and, with some additional assumptions,
to a lesser degree for experimental data of color
discrimination in the threshold domain (Section
8.4.2). The theory is inadequate to account for the
way color stimuli appear to an observer. For
example, it cannot explain why, under the usual
observing conditions, an observer with normal
color vision sees “yellow” when a “red” stimulus
is additively (and in appropriate proportions)
mixed with a “green” stimulus; “yellow” clearly
being perceived as qualitatively different from
each of the two components in the mixture. Simi-
larly, the theory cannot explain why a “yellow”
stimulus, when mixed with a “blue”’ stimulus, can
lead to a “white” stimulus.
The appearance of color stimuli or color per-
ception is explained with considerable success by
the opponent-colors theory. The opponent-colors
theory, proposed by Ewald Hering in the last
century (see Hering’s Outlines of a Theory of
the Light Sense as translated by Hurvich and
Jameson in 1964), assumes the existences of visual
processes capable of generating neural signals of
two opposite kinds, depending on wavelength.
Such processes account for the visual experience
of seeing a variety of, but combination-limited,
hues of varying saturation and brightness. More
specifically, these processes account for the attri-
bute of hue that can be described by the names of
red, yellow, green, and blue, and by names ex-
pressing hues intermediate to red-yellow, yellow-
Preamble 583
green, green-blue, and blue-red. The opponent
processes also account for the empirical fact that
other intermediate hues, such as reddish-greens or
yellowish-blues, are never experienced.
When taken alone, neither of the original theo-
ries of Young—Helmholtz and Hering, though they
competed against one another for some time, was
capable of giving satisfactory explanations of
several important color-vision phenomena and of
making adequate predictions of other phenomena
of color vision. However, when merged into a
single theory, known as a zone theory, many
color-vision phenomena could be explained and
predicted that relate to color matching, color dis-
crimination, color appearance, chromatic adapta-
tion, and other experiments for both observers
with normal and observers with (congenital) de-
fective color vision.
G. E. Miller (1930), see also Judd (1949;
1951b), is usually credited with being the first to
introduce the zone-theory concept and to explore
plausible specific zone theories. Essentially, a zone
theory assumes that in the first zone there are
located three independent types of cone in which
color vision is initiated through the process of
absorption of light in the photopigments of the
cones and conversion into electrical signals. This
zone complies with the basic assumption of tri-
chromacy in the Young—Helmholtz theory and
accounts for the experimental data of color
matching. In the second zone, the cone signals are
coded in a neural network that generates three
new signals; one achromatic signal and two
antagonistic chromatic signals. This second zone
complies with the basic assumption of the ex-
istence of opponent processes in the Hering the-
ory and accounts for experimental data of color
appearance. Subsequent zones in the visual sys-
tem are most likely to exist in which further
processing (coding and encoding) of the signals of
the second zone takes place, but specific assump-
tions as to the internal structure and functioning
of these zones are yet to be developed. In the final
zone, located in the cortex, of the assumed
hierarchal structure of the visual system, the sig-
nals are interpreted in the context of other visual
information (mainly spatial and temporal) re-
ceived at the same time and in the context of
previously accumulated visual experience (mem-
ory).
There are other theories of color vision, among
which the retinex theory of Edwin Land (1974) is
a more recent development. However, we will not
pursue these theories in this section. With regard
584 THEORIES AND MODELS OF COLOR VISION
to the retinex theory, we refer the reader to Sec-
tion 5.12.2(viii) where, apart from a brief descrip-
tion of the experiments done by Land and his
collaborators, a number of references are given
that may be helpful to the reader interested in a
more detailed treatment.
A model of color vision, as opposed to the
concept of a theory of color vision, rests on the
assumptions made in the theory but treats these
assumptions as a simplified approximation useful
for certain purposes. Although the framework of
the more basic theory is believed to be true,
though perhaps only tentatively, the model must
certainly be false as it is built, by definition, on
simplified approximations to the assumptions
made in the theory. But this feature of the model
does not necessarily defeat its usefulness in color
science. The possibility exists, though remotely,
that the model complies exactly and in every
respect with the theory; then the model would
change its status and acquire the title of theory.
Conversely, what starts as a theory may later be
seen to be inadequate in the light of empirical
investigation and so be “demoted” to the status
of model.
Models, like theories, employ analogies in
characterizing satisfactory explanations by relat-
ing the visual system to physical or chemical
systems. In accordance with Redhead’s (1980)
informative treatment of models in physics, two
basic cases may arise:
(a) Suppose a theory has been proposed satisfy-
ing the condition that it consists of a set of
assumptions about the visual system and the con-
dition that these assumptions attribute an inner
structure or mechanism that manifests itself in
other properties exhibited by the visual system.
The theory under consideration may be too com-
plicated to provide an “exact solution” of the
functioning of the visual mechanism. In such a
case, a simplified model may be built which may
permit an exact solution, or at any rate present a
more tractable structure than the given theory.
The model then plays the role of an impoverished
theory, the important feature being that the model
and the given theory logically contradict one
another, so that we believe the model to be false
insofar as we believe the theory to be true.
(b) A theory may be given that, though meeting
the conditions stated above under (a), is incom-
pletely specified in the sense that certain general
constraints may be imposed but this may still
allow considerable latitude in selecting the de-
tailed structure of the theory. In such a case, a
model may be built that, in a sense, “fills in the
missing detail.” This model is said to enrich the
given theory. We do not believe that what
the model tells us is false in the sense that it
contradicts what might be proposed as a complete
theory, which, in fact, we are envisaging has not
been proposed. Again, as under (a), the guiding
principle in building the model is its tractability.
There are models of color vision that im-
poverish and models that enrich some of the
properties of the underlying theories, notably
those outlined above. Also, in general, models of
color vision display useful analogies in char-
acterizing the explanations they provide with re-
gard to the visual system. An important type of
model displays the visual system, or part of it, in
the form of a diagram showing schematically its
essential anatomical features and its neural con-
nections assumed to play a role in generating and
transmitting signals in response to color stimuli
received by the system. In the light of the known
anatomical structure of the neural network of the
visual system (Section 2.2), on the basis of which
a rather complicated theory might be formulated,
such models play most certainly the role of an
impoverished theory. Nevertheless, they are useful
in the sense that they are tractable and as such
often provide clues for building another type of
model, appropriately called a mathematical model
of color vision. A mathematical model arises when
we establish a formal analogy between a color-
vision theory and some mathematical structure.
Often, the mathematical structure of this kind of
model is embedded in some wider structure that
has led to the notion of “surplus” structure in the
formulation of a theory (Redhead, 1980). The
calculus associated with this wider structure is
partially interpreted by means of the original
theory, and often is used as a mechanism for
further developing the original theory or to pro-
pose new theories.
Among the various kinds of models that phil-
osophers of science consider (see again Redhead,
1980), are so-called “floating models” (a term
coined by H. R. Post) which we seem to encoun-
ter quite frequently in the field of color vision.
Essentially, a floating model is disconnected from
the basic theory and the empirical facts; that is, it
has no theoretical or empirical support and, quite
obviously, its usefulness is in question. According
to Redhead (1980), H. R. Post has parodied the
case of a floating model with an example referred
 Preamble 585

Fig. 1(8.1). Schematic diagram of a farm gate with strut of length /, rails of length x, and
stiles of length y. The farmer’s model of predicting / from x and y is /= e(x, y)(x + y), where
e(x, y) is introduced as a new effect, “the farm-gate contraction.”

to as the Farm-Gate Contraction. The case is We notice that the farmer’s model characterized
described along the following lines: by Eq. 1(8.1) has several properties analogous to
A farmer, interested in the construction of the model of Eq. 2(8.1) which completely accounts
farm gates, investigates the relation between the for the properties of the true theory. In particular,
length (/) of the diagonal strut and the lengths (x for x = 0 or y= 0, the farmer’s model predicts
and y) of the rails and stiles of a farm gate, such correct values for /, and also, / increases monoton-
as that shown in Figure 1(8.1). Although he is ically with x or y as in the perfect model. How-
familiar with the theory and practice of length ever, when the farmer carries out detailed and
measurement and with Euclid, the law of precise measurements, he finds discrepancies. In
Pythagoras is utterly beyond his deductive powers. order to explain the mismatch between the pre-
So he invents a model, a linear one, in which the dictions of his model and the experimental re-
lengths (/, x, y) are related by sults, the farmer introduces a new effect, the
Farm-Gate Contraction. It is possible that the
J=xt+y [1(8.1)] farmer, before he took up the career of farming
instead of some years ago, had some intimate connection
with color science and indeed his peers may have
[2(8.1)] been builders of models of color vision. In any

(x,y)
€

Contraction
Farm-Gate

“00 0.2 0.4 0.6 08 1.0 2

x

Fig. 2(8.1). The function e(x, y) defined in Figure 1(8.1) is plotted on the basis of many
experimental data points (open circles) for different values of x and y (private communication
of a model-building farmer).
586 THEORIES AND MODELS OF COLOR VISION

event, the farmer casts his model into a new form,
that is,
b= el xtteyi) [3(8.1)]
The farmer establishes that the quantity e is not
only smaller than unity in every case he encoun-
ters, hence the term farm-gate contraction, but is
also a surprisingly complicated function of x and
y. The farmer’s experimental data (private com-
munication) showing e as function of x with y
being constant are illustrated in Figure 2(8.1).
Quite obviously, we could further expand on
the case described above, but we prefer to leave
this to the reader. However, observations
general kind might be made. It is true that any
model can be matched with experiment if one
allows appropriate modifications of the ad hoc
designed model to be made in an effort to “cor-
rect” it. The status of this floating model is put
into serious question because has it not misled us
into investing a mathematical artifice with a
semblance of physical significance (the mythical
contraction)?
In recognizing the danger of using floating
models that are disconnected from the basic the-
ory and the empirical facts, we must be careful in
our approach to constructing useful models of
color vision. At least three different ways emerge
(see also Green and Luce, 1974):
is possible. Our selection given in the sections
that follow must necessarily be very limited in
scope and ignore some of the models that many
readers may consider most important. In keeping
with the general scope of the book, the psycho-
physical aspects involved in the models are
stressed; other aspects such as those of biochem-
istry and electrophysiology, if they are discussed
at all, do not go beyond a very rudimentary level
of explanation. For accounts of that kind, the
reader is referred to the relevant literature; the
Handbook of Sensory Physiology Volume VII/1 to
3 (Springer-Verlag, New York, 1972 and 1973)
provides an excellent start. Reference is also made
of a
to the previous chapters in our book where, in
several cases, brief descriptions of models have
been given.
8.2 VISUAL RESPONSE FUNCTIONS AND THE
SPECTRAL PROPERTIES OF VISUAL PIGMENTS
There are various visual response functions that
can be compared directly with objective proper-
ties of visual pigments. Such comparisons are
often helpful in the derivation of appropriate
models that explain the functioning of cone and
rod receptor mechanisms. This section makes a
few elementary points concerning such compari-
sons.

(1) We can begin with a few empirical generali-

zations of a group of experimentally established 8.2.1. Principle of Univariance
phenomena that, when taken as the basic theory
Spectral visual response curves obtained by psy-
in the form of a set of postulates or axioms, lead
chophysical methods are essentially action spec-
deductively, by means of a mathematical model,
tra, and represent the reciprocal radiant powers
to a variety of testable predictions of other visual
of monochromatic stimuli producing some con-
phenomena.
stant sensory effect. When the latter is the result
(ii) We can postulate a theory in terms of an
of absorption in a visual pigment, the condition
internal structure or mechanism mediating color
that two stimuli of wavelengths A, and A, pro-
vision and, after estimating the parameters of an
duce the same effect is that the numbers of ab-
appropriate mathematical model of that mecha-
sorbed quanta (or photons) of the two kinds of
nism, then predict a variety of visual phenomena.
radiant energy are the same. This assumes that
(iii) We can actually investigate the internal
the effectiveness of an absorbed quantum in pro-
functioning of the photoreceptors and the neural
ducing the visual result is independent of the
network of the visual system, describe the system
wavelength of the quantum (i.e., that the quantum
axiomatically, determine experimentally the
efficiency is constant). This can probably be as-
needed parameters, and then use a mathematical
sumed if A, and A, lie in a single broad ab-
model to deduce the visual phenomena.
sorption band of the pigment associated with a
particular electronic transition in the pigment
Most models of color vision seem to have been molecule.
constructed in one of the three ways described Rushton (1972) summarized the above proper-
above. No brief summary of the many useful ties of a visual pigment and their relation to
models of color vision that have been constructed vision in the principle of univariance. When ap-
 Visual Response Functions and the Spectral Properties of Visual Pigments
plied to the pigment in the rod receptor, rhodop-
sin, which has been studied extensively, Rushton
formulates the principle of univariance as follows:
Any two lights that are equally absorbed by rhodopsin
will be equally seen by rods.
It follows that the rhodopsin spectral absorp-
tion curve must coincide with the scotopic
luminous efficiency curve V’(A) (Section 5.7).
The experimental evidence confirms this relation-
ship [see, for example, Crescitelli and Dartnall,
1953; and also Figure 1(8.2.3) of Section 8.2.3].
The biochemical evidence also supports the prin-
ciple of univariance by showing that the initial
action of the incident quanta on the rhodopsin
pigment is to produce a conformational change in
the rhodopsin molecule, known as photo-isomeri-
zation. It is assumed that equal numbers of
isomerized molecules lead to indistinguishable
sensations (Wald, Brown, and Gibbons, 1963;
Wald, 1968).
The principle of univariance is also applied to
each of the photopigments that, in accordance
with the Young—Helmholtz theory, are assumed
to reside in three kinds of cone receptor. The
color-matching functions, denoted, for example,
by u(A), 0(A), w(A), derived from actual trichro-
matic color matches (Section 5.5), are the visual
spectral response functions that are related with
the spectral properties of the cone pigments.
Ordinary trichromatic matching can be
summed up in the equations for a complete match
between two stimuli of spectral distributions
{P, dA} and {PX dd}, expressed in quantum
units instead of energy units, viewed side by side
in a bipartite field:
U =| P,u(A) dd = [PiPa(A) dA =U
V =[Po(A) dd = [P(A) dA SV
W = [P\w(A) dd = [PW(A) dd = WO
[1(8.2.1)]
The color-matching functions, u(A), o(A), w(A),
refer to three given fixed monochromatic stimuli
(primary stimuli) of different wavelengths and are
uniquely determined. Equations 1(8.2.1) express
the strong form of the trichromatic principle (Sec-
587
tion 3.2), valid for foveal vision for not too high
stimulus levels (retinal illuminances). They are
exactly those that would hold if the visual effect
of a stimulus were produced by the light absorp-
tion of three photosensitive pigments contained in
the retinal end-organs, the cones, and if, for all
the cones containing a particular pigment, the
relative spectral absorption factor of that pigment
were the same and were to remain unchanged
whatever stimuli were applied. Here, by absorp-
tion factor of a particular pigment in a cone
containing possibly other pigments, we mean the
fraction of the radiant power incident on the cone
that is absorbed by the pigment in question. This
definition leaves open whether the pigment in the
cone is to be regarded as a uniform thin or thick
layer or as being nonuniformly distributed, and
whether because of wave—mode propagation in
the cone (Section 2.4) there is a nonuniform spa-
tial distribution of the stimulating radiant energy.
The conditioning for the spectral absorption fac-
tor may well not be satisfied when the matching
or adapting color stimuli reach very high levels
(Section 5.6.5), but, if it fails, it is almost certain
that Eq. 1(8.2.1) will also cease to hold; they
would survive only if some other factor were to
counterbalance exactly the deviations in relative
spectral absorption factor.
From the above it follows that, within the
range of validity of the trichromatic principle,
there will be a linear relation between the empiri-
cal color-matching functions and the relative
spectral absorption factors, denoted, for example,
by p(A), y(A), B(A), of the three cone pigments,
or, more strictly, of these factors multiplied by
the spectral transmittance 7(A) of the prereceptor
light path in the eye (Section 2.4.6); thus,
p(A)r(A) = a,,i(A) + 4,,0(A) + a), (A)
y(A)r(A) = a,,a(A) + a,,0(A) + a,,0(A)
B(A)1(A) = aputt(A) + ag.0 (A) + apyw(A)
[2(8.2.1)]
where the nine coefficients a,,, pu? and so on, are
constants.
The objective detection and measurement of
cone pigments in recent years (Section 2.2.7) has,
of course, strongly consolidated our belief in the
pigment absorption explanation of the trichro-
matic principle (however, see also Section 8.2.4).
588 THEORIES AND MODELS OF COLOR VISION
In this explanation, it is immaterial how the three
pigments are distributed among the cones, or how
the neural activity is elaborated by neural mecha-
nisms in the retina or brain. Conversely, ordinary
trichromatic color matching gives no information
relative to visual processes beyond the pigment
absorptions. However, asymmetric matching (Sec-
tion 5.2) can provide such information within the
limitations of psychophysical models of color vi-
sion. Studies of chromatic adaptation (Section
5.12) and of color discrimination including incre-
ment thresholds (Sections 7.4 and 7.10) fall into
this category.
8.2.2 The Visual-Pigment Layer
The visual pigments in the retina are contained in
the end-organs of the photoreceptors, the rods
and cones (Section 2.2.7). Every visual pigment is
regarded as, effectively, a layer of a particular
thickness h and concentration c. Its spectral
transmittance 7(A) is then given by
tl lpeie ea [1(8.2.2)]
where a(A) is the pigment’s spectral absorption
coefficient.
The fraction k(A) of the stimulus absorbed is:
K(A) Se eee 2 (82.2)
which can be approximated by
k(A) = cha(d) [3(8.2.2)]
if cha(A) is small compared to unity. The relative
spectral visual response can then be compared
directly with a(A), known, for example, from in
vitro measurements on pigment extracts. When
cha(A) is not small, the comparison should be
with oa(A) where the factor
1 e chat A)
EW [4(8.2.2)]
is given in Table 1(8.2.2) with corresponding val-
ues of cha(A) and of cha(A) - logj, e which is the
optical density of the pigment layer for wave-
length A. This correction is sometimes referred to
as a correction for self-screening.
Visual pigment molecules have highly direc-
tional absorption properties and only the compo-
nent of the electric vector of the incident radiant
flux which is parallel to the long axis of the
molecule is effective as far as the main absorption
band is concerned. If a’(A) is the absorption
coefficient when the absorbing molecules have
their long axes all parallel to the electric vector in
a beam of polarized light, the absorption coeffi-
cient for the case in which the directions of the
molecules are completely at random equals
a’(A)/3, and is the same for polarized and un-
polarized light. In the retina, the approximate
arrangement of the visual pigment molecules is
with their long axes at random in a plane normal
to the incident radiant flux, and the absorption
coefficient is a’(A)/2 for polarized and un-
polarized light. Thus, the effective absorption coef-
ficient of the pigment in the retina is 1.5 times its
value measured in vitro in a solution. This point is
generally relevant only when absolute values of
absorptance are in question. However, if the pig-
ment molecule has a second absorption band (a
cis-peak, say), for which the absorption direction
is not along the long axis, a different allowance
for pigment orientation may have to be made in
different parts of the spectrum.
In many investigations of photosensitive visual
pigments by objective methods, what is de-
termined is a difference spectrum, that is, the
difference in optical density [log (1/transmit-
tance)] of the pigment layer before and after a
partial or complete bleaching of the pigment, the
result being independent of any stable “impuri-
ties” present in the layer. If 6)(A) and 6(A) are
the optical densities before and after bleaching,
the difference spectrum is given by
ALi Peat 1&4 ieee :
log io e ei Src 8A)
= cha(A) — h[(c — x)a(r) + yB(A)]
=h[xa(A) — yB(A)] [5(8.2.2)]
where x is the reduction of the concentration of
the original pigment and y is the concentration of
any products of bleaching, assumed to have the
absorption coefficient B(A). Provided the prod-
ucts of bleaching are stable and not lost from the
layer, y may be put equal to x and
89(A) — 8(A) = xh[a(A) — B(A)]
[6(8.2.2)]
Thus, relative values of the difference spectrum
for different wavelengths are independent of.the
original concentration and the amount bleached,
however large or small these may be. These values
 Visual Response Functions and the Spectral Properties of Visual Pigments 589

Table 1(8.2.2) Ratio o of the Fraction Absorbed by a Pigment Layer to the

Product k(A.) of Spectral Absorption Coefficient a(A), Concentration (c),
and Layer Thickness (/), for Various Optical Densities 6(A) of the Layer“

5(A) k(X)

48(X) = k(A)-log joe (optical density of layer); k(A) = cha(A) (approximate

fraction of stimulus absorbed); o=k(A)/k(A) (correction factor for self-
screening); k(A) =1—e “"*) (fraction of stimulus absorbed).

are characteristic of the pigment or, strictly, of
the combination of pigment plus products of
bleaching; the latter may be different in different
chemical environments. In some instances,
for a particular wavelength, the isobestic wave-
length, the difference spectrum may drop to zero
when a(A) = B(A) and may also go negative.
The difference spectrum is not comparable,
without careful qualifications, to spectral re-
sponse curves. It must probably be assumed that,
in the effective pigment layer of the retina, ab-
sorption of radiant energy will yield pigment
products with a different spectral absorption
coefficient B(A) from the original pigment, and
after some bleaching has occurred, the fraction
k*(A) of the radiant energy absorbed by the
original pigment becomes
k*(A) = (c — x)a(A)
(c — x)a(A) + yB(A)
X [1 — eT Blea) —YABA)]
[7(8.2.2)]
If, for all wavelengths concerned, c, x, and y are
sufficiently small, Eq. 7(8.2.2) reduces to
and
k*(X) =h(c—x)a(A) [8(8.2.2)]
so that visual response curves may be compared
with a(A) just as though there were no products
of bleaching in the retina. Also, if very little
bleaching of the retinal pigment has occurred
when a visual response curve is determined, so
that x and y are compared with c, the quantity
k*(A), defined by Eq. 7(8.2.2), reduces to k(A),
defined by Eq. 2(8.2.2) above.
Again, products of bleaching play no part.
This is probably the situation for most visual
response curves determined at moderate or low
stimulus levels. However, after very strong bleach-
ing which makes (c — x) small compared with y,
there is a possible effect of the products of
bleaching on the shape of the visual response
curve as shown by the expansion of Eq. 7(8.2.2)
 590 THEORIES AND MODELS OF COLOR VISION
in the form
k*(A) =h(c— x)a(A)
x {1 5p[h(e~ x)a(A) + yABO)]
+ Lae x)a(d) + yaBOAYE —--- |
[9(8.2.2)]
However, the effect is different and much less
radical than that produced on the difference spec-
trum. The latter can really be used in a com-
parison with visual response curves only at
wavelengths for which B(A) is negligible com-
pared with a(A) and
83(X) — 8’(A) ~ hxa(A) [10(8.2.2)]
The kinetics of the visual pigments of rods and
cones have been studied extensively by retinal
densitometry of the living human eye (Campbell
and Rushton, 1955; Rushton, Campbell, Hagins,
and Brindley, 1955; Rushton, 1958; 1972; Weale,
1965). Although retinal densitometry, also known
as the fundus reflectometry, has its technical diffi-
culties (see, for example, Boynton, 1979), some of
the results obtained by the technique have signifi-
cantly contributed to our understanding of the
visual process that takes place in the receptor
stage, particularly with regard to the visual adap-
tation process at both scotopic and photopic levels
of retinal illuminance (Stiles, 1961; Rushton,
1965a).
The kinetics of visual pigment bleaching in
man were shown to conform to a very simple law
of a first-order photochemical reaction which
Rushton (1972) expressed in the following three
statements:
(a) Pigment is bleached at a rate proportional to
the quantum catch.
(b) Regeneration proceeds at a rate proportional
to the fraction of the pigment in the bleached
_ State.
(c) Process (a, bleaching by light) and process
(b, regeneration by chemical processes) are inde-
pendent of one another and therefore are simply
additive.
The mathematical model describing the above
three statements has the following form
SprpepTr
ee ret Very)
- [11(8.2.2)]
where T=retinal illuminance (measured in tro-
lands) obtained from the luminance
(cd-m~*) of the external stimulus (Sec-
tion 2.4.4.). In the case of rhodopsin
pigment, the scotopic retinal illumi-
nance JT’ measured in scotopic tro-
lands, is used.
t=time (measured in seconds, sec).
p=fraction of pigment unbleached at t.
Q=energy of the pulsed external stimulus
(measured in troland-seconds, td-sec)
that bleaches p from 1 to e!.
to=time constant of regeneration (mea-
sured in seconds, sec), that is, time it
takes to restore the bleached pigment.
Typical values for the two constants Q and fy in
Eq. 11(8.2.2) for the case of rhodopsin-pigment
bleaching and regeneration are
Q = 1.57 - 10’ [(scotopic) td - sec]
ty = 519 (sec)
Typical values for the two constants correspond-
ing to cone-pigment bleaching and regeneration
(chlorolabe and erythrolabe cones only and
“white”’-stimulus bleaching, see below) are
Q = 5.0 - 10° [(photopic) td - sec]
ty = 130 (sec)
Simple calculations with Eq. 11(8.2.2) show
that the changes in rhodopsin density that occur
while the effective rod sensitivity is rising during
dark adaptation by more than one log unit are
very small. For the experimental data of Arden
and Weale (1954), illustrated in Figure 4(7.3.1),
the rhodopsin density at the end of the pre-
liminary adaptation to 3000 td is still about 90%
of maximum. The density changes in the period
from the eight-minute point of the recovery curves
to the final steady values, from 0.97 to 1.00, or
3%. The conclusion established by the work of
Rushton and his associates that apparently in-
significant changes of rhodopsin density may
accompany quite large adaptive changes in effec-
tive rod sensitivity is in line with earlier evidence
of various kinds (see, particularly, Baumgardt,
1949; 1972).
The application of Eq. 11(8.2.2) to the phot-
opic visual pigments (cone pigments) enables us
to see how far changes in pigment concentration
might play a part in chromatic adaptation and
color matching. Although it is virtually certain
 Visual Response Functions and the Spectral Properties of Visual Pigments
that three cone pigments are concerned in an eye
possessing normal color vision, retinal densitome-
try has revealed only two. These have the maxima
in their difference spectra at approximately 540
nm and 590 nm, respectively. They have been
described as the “green-sensitive” or “green-
catching” (chlorolabe) and the “red-sensitive” or
“red-catching” (erythrolabe) pigments. The eva-
sive “blue-sensitive” or “blue-catching” (cyano-
labe) pigment with a maximum at approximately
440 nm, presumably has a very low optical den-
sity and/or only relatively few “blue cones” re-
side in the foveal retina, Thus, strictly speaking,
Eq. 11(8.2.2) applies only to bleaching and regen-
eration of two cone pigments. The regeneration
rate is found to be the same for both pigments,
that is, 1/130th (or 0.77%) of the bleached pig-
ment is restored per second. The rate of bleaching
with “white” stimuli also happens to be the same
for the two pigments. However, stimuli with other
spectral radiant power distributions will result in
different bleaching rates for the two pigments
depending on their different spectral absorption
curves.
In the (photopic) range of retinal illuminance
T from about 10? to 10° td, the bleaching of the
cone pigments increases from nearly zero to nearly
100%, yet there is no indication from increment
threshold measurements (Section 7.6) that any-
thing remarkable occurs that is related to the
incidence of the bleaching process. Over the whole
of the range of retinal illuminance mentioned, the
Weber fraction (ratio of increment threshold to
field radiance) is substantially constant (Stiles,
1961). It seems that the discriminating power of
the photopic mechanism defined by the Weber
fraction is just about as good when all but a few
percent of the pigment is bleached as when none
of it is bleached. Nevertheless, there are other
visual observations, particularly the color ap-
pearance of the field and the persistence of color
matches, that show changes around the region
where the bleaching is becoming marked (Sec-
tions 5.6.5 and 8.2.6).
8.2.3. Dartnall’s Standard Shape of Visual
Pigment Absorption Coefficient
According to Dartnall’s generalization (1953), the
visual pigments, a great variety of which have
been extracted from animal retinae and mea-
sured, all have relative spectral absorption coeffi-
cients that vary with wavenumber in the same
way and differ only in showing maximum absorp-
tance at different wavenumbers. Thus, a plot of
591
log a(A) against 1/A would be a curve of the
same shape for all the pigments, being merely
shifted for different pigments to different posi-
tions with respect to the axes. When a visual
response curve may possibly correspond to ab-
sorption in some one or other visual pigment, a
comparison of its shape (after correction for pre-
pigment light losses in the eye) with Dartnall’s
standard shape may be informative. The standard
shape is specified in Columns 5 and 6 of Table
1(8.2.3) which also gives (Columns 1 to 4) the
logarithm of the response curve of rod vision (the
CIE V’(A) curve) converted to a quantum basis
and with an average correction applied for light
losses on the eye lens. A comparison of this latter
curve with Dartnall’s standard curve shifted to
have its maximum at A = 498 nm to give the best
fit is shown in Figure 1(8.2.3). This turns out to
be substantially the same comparison as made by
Crescitelli and Dartnall (1953) but with an im-
portant qualification. In Figure 1(8.2.3), to il-
lustrate the method, the standard curve was shifted
along the wavenumber axis to suit best the visual
response data; Crescitelli and Dartnall, however,
determined the curve and its position by photo-
chemical measurements on the rod pigment
(rhodopsin) from an actual human eye. The
agreement in shape shown by Figure 1(8.2.3) sug-
gests that the rod curve, which covers a much
greater range on the long-wavelength side than
the standard curve, may be used in place of the
latter in testing other visual response curves. Two
examples of this application are shown in Figures
2(8.2.3) and 3(8.2.3). There is an incompatibility
in shape between the “red’’-response curves [(A)
in Figure 2(8.2.3), 7; in Figure 3(8.2.3)], and the
standard pigment curve. At wavenumbers above
20,000 cm~! (A < 500 nm), an increasing uncer-
tainty in the visual response curves must be ad-
mitted because of their dependence on the eye
lens correction.
A detailed comparison with color-matching
functions, to be discussed in Section 8.2.4, has
been made by Stiles and Wyszecki (1974).
8.2.4 Color-Matching Data and the Spectral
Absorption Curves of Visual Pigments
Color-matching data and other visual response
data depend on the way the absorptions of the
visual pigments vary with wavelength, and con-
versely, certain limitations on the possible shapes
of the spectral absorption curves are imposed by
the visual response data. We have seen in Section
8.2.3, that at least for the long-wavelength (“red’’)
ay THEORIES AND MODELS OF COLOR VISION

Table 1(8.2.3) Relative Rod Response and Standard Shape of Visual Pigment
Absorption Coefficient“
1 2 3 4 5 6
1/A 1/r
(cm7) log W(A) (cm) log W(A) 1/A—1/Ay, log @(A) — log aA) max
13,000 7.270 19,750 1.992
250 7.630 20,000 0.000
500 7.988 250 1.997 —4000 (3.70)
750 6.347 500 1.984 —3600 2.30
14,000 6.708 750 1.967 —3200 2.699
250 5.068 21,000 1.946 —2800 1.041
500 5.432 250 1.922 —2400 1.342
750 3.790 500 1.898 —2000 1.568
15,000 4.152 750 1.874 —1600 1.736
250 4.515 22,000 1.846 —1200 1.863
500 4.874 250 1.812 —800 1.944
750 3.232 500 1.779 —400 1.987
16,000 3.578 750 1.730 0 0
250 3.908 23,000 1.669 +400 1.987
500 2.215 250 1.602 800 1.954
750 2.500 500 1.538 1200 1.906
17,000 2.756 750 1.488 1600 1.848
250 2.983 24,000 1.433 2000 1.782
500 1.182 250 1.376 2400 1.708
750 1.354 500 1.308 2800 1.628
18,000 1.504 750 1.225 3200 1.550
250 1.631 25,000 1.154 3600 1.498
500 1.739 250 1.088 4000 1.447
750 1.824 500 1.036 4400 1.389
19,000 1.890 750 2.982 4800 1.352
250 1.940 26,000 2.950 5200 1.322
500 1.974

“Columns | to 4: Logarithm (relative rod response) = log W(A) as a function of.

1/(A); W(A) is the CIE V’(A) function corrected for lens absorption, using the lens
factors given in Section 2.4.6, converted to quantum units by multiplying by 1/A,
and finally normalized to a maximum value of unity.
Columns 5 and 7: [log a(A)—log a(A) nx] aS a function of (1/A —1/A,,,). This
relation defines the standard shape of a visual pigment absorption coefficient as a
function of 1/A (numerical values, kindly supplied by Dr. H. J. A. Dartnall, are
probably the most reliable values available). !
According to more recent work (private communication from Dr. Dartnall), there
is probably a slight systematic difference in the standard shape for visual pigments
employing respectively as chromophore retinene 1 and retinene 2. The values in
Column 6 of the table are based on data for retinene | pigments.

fundamental p(A), of Thomson and Wright [Fig-
‘ure 2(8.2.3)] or for the “red” mechanism 7, of
Stiles [Figure 3(8.2.3)], the spectral response could
not correspond to absorption by a single visual
pigment of rhodopsin type, whereas for the
“sreen”- and “blue’”’-sensitive mechanisms [y(A)
and B(A)], or 7, and 7, (or 7;), the spectral visual
response could arise in this way. More generally,
Brindley (1960) quotes in his book calculations he
had made showing that no set of three visual
pigments, each with a similar spectral absorption
to rhodopsin but displaced to different positions
in the spectrum, would be compatible with the
empirical data of color matching. Our own calcu-
lations (Stiles and Wyszecki, 1974) are in general
agreement with Brindley’s statement, but bring
out a number of additional points of interest
which are discussed below. The question of the
shapes of the spectral sensitivity curves of the
receptor mechanisms has also been emphasized
 Visual Response Functions and the Spectral Properties of Visual Pigments 593

[alamo]
WAIOR
LOG

25000 22500 20000 17500 15000

WAVENUMBER 1/7 (em)

Fig. 1(8.2.3). Comparison of the scotopic luminous efficiency curve W(A), expressed in
quantum units and corrected for average lens absorption, with the standard shape defined by
the cross points, of spectral absorption curves for visual pigments [a@(A)/a(A)
max |:

by several other investigators, notably LeGrand
(1969), Smith and Pokorny (1972, 1975),
Piantanida and Sperling (1973). These investiga-
tors discuss evidence on the shape question
derived from the color-matching properties of
observers with normal color vision and of
observers with defective color vision, from two-
color threshold studies, from spectral absorptions
of retinal extracts, from retinal densitometry, and
from the microspectrophotometry of photorecep-
tors.
The numerical examination made by Stiles and
Wyszecki (1974) is confined to color-matching
data of normal trichromats and rests on the fol-
lowing basic assumptions:
(i) A visual pigment is said to be of rhodopsin
type if its spectral absorption coefficient as a
function of wavenumber, m = 1/A, is of the form
of a, f(m — my) where my is the wavenumber at
which the coefficient has its greatest value and
f(---+) is a fixed function such that ag f(m — mg)
coincides with the empirically determined absorp-
tion coefficient of rhodopsin with the appropriate
choice of my, (~ 1/500 nm) and ao.
594 THEORIES AND MODELS OF COLOR VISION

—!)

Nl

Basis
Quantum

[p(A),7(X),8(A)]
[a()/a(X)inax]
Log
or >|

5
25000 22500 20 000 17500 15000 12500

WAVENUMBER 1/A (cm)

Fig. 2(8.2.3). Thomson-Wright [o(A, y(A),8(A)] fundamentals corrected for prereceptor
light losses and compared with the displaced rod curve (dashed lines) (from Stiles, 1959).

(ii) Vision is effected by absorption in three
photosensitive pigments of rhodopsin type with
maximal absorption coefficients at three wave-
numbers m,, mj, m,, respectively, for which most
suitable values are to be found.
(11) All external stimuli of radiant energy reach-
ing the photosensitive pigments do so, in effect,
after passing through an overall filter (eye lens
and macular pigment) of spectral transmittance
t(m), for which again the values are left open
initially.
(iv) In the absorption of light in the end-organ
of the photoreceptor, each visual pigment can be
regarded as acting like a pigment layer (Section
8.2.2) of effective concentration c and effective
thickness h, giving an absorption factor [1 —
e “9(™)). the quantities c and h are sufficiently
small, or at least the product ch is sufficiently
small for the approximation
e hm) ~ 1 — cha(m)
to be valid for all relevant wavenumbers.
The objective of the numerical examination is
to see how near to the data of color matching the
hypothesis embedded in the above assumptions
(11) to (iv) will lead.
The mathematical model that satisfies the
above basic assumptions follows readily. Two
stimuli of spectral distributions {N,, dm},
{N,,dm}, where N,,, N,, represent numbers of
quanta of wavenumber m per unit wavenumber
interval, are visually equivalent as far as absorp-
tion a,(m) in pigment 1 is concerned, if the
number of quanta absorbed is the same; that is, if
Dae — exp[—a,(m)ch]}t(m)
dm
= [n{1 — exp[ —a,(m)ch]}t(m) dm
and similarly for absorption a,(m), a3,(m) in
pigments 2 and 3.
On making the simplifying assumption (iv)
above, and then replacing the absorption coeffi-
cients by their specifications using the function
f(- ++), the conditions for a complete (metameric)
 WAVELENGTH » (nm)

(RELATIVE
LOG
SENSITIVITY)

8 15000
25000 22500 20000 17500
WAVENUMBER _ 1/p (cm?)
Fig. 3(8.2.3). Field spectral sensitivity curves of Stiles’ mechanisms 77, 73, 74, 7; (Section 7.4)
corrected for prereceptor light losses and compared in each case with the similarly corrected
sensitivity curve of rod vision displaced parallel to the two axes to give the best fit (Stiles,
1959).

595
596 THEORIES AND MODELS OF COLOR VISION

color match become: light) the empirical matching conditions reduce to

[Ny fm = m,)t(m) dim [NamFQ) dm = [Nj,mr(A) dm

= [Nj f(m — m,)t(m) dm [NnmB(A) dm = [Nj,mgy(r) dm

[Nn f(m — mz)t(m) dm [Nnmb (A) dm = [Njmb (X) dm

2(8.2.4
= [Nj f(m — mz)t(m) dm [2¢ )

The color-matching functions 7(A), Z(A), b(A)

/N,, f(m — m,)t(m) dm are indeterminate to a linear transformation (Sec-
tion 3.2) so that the predictions of the model will
agree with the empirical color-matching data, if
= [Nj f(m — m3)t(m)dm
mr(X) = pi, f(m — m,)t(m)
[1(8.2.4)]
+P. f(m— mz)t(m)
The empirical color-matching conditions are +pi3f(m — m3)t(m)
usually based on stimuli defined in terms of spec-
tral radiant power distributions {P, dA}, {Px dA} mg(A) = po, f(m — m,)t(m)
with P,, Px expressed in watts per unit wave-
length interval. Alternatively, they are defined in +P f(m — m,)t(m)
terms of distributions {F,, dm}, {F,dm} with
F.,, F’, in watts per unit wavenumber interval. If +p, f(m — m,;)t(m)
we denote the color-matching functions by b
7(A), Z(A), b(A), the empirical color-matching mb (A) = p3,f(m— m,)t(m) -
conditions are then given by the familiar expres-
sions (Section 3.2) +3 f(m — m,)t(m)

+p33f(m — m3)t(m)
fP7Q) dy = [Px7(A) dd
[3(8.2.4)]
or .
when the matrix of transformation coefficients p,,
ifF7(A) dm = /F’ r(A) dm has a nonzero determinant.
In terms of the chromaticity coordinates:

with similar equations involving g(A) and b(A), mr(X)

respectively, instead of r(A). The quantity F,, is OM )i spre
related to P, by the equation mr (A) mg (AD ae Gay

F ones y a 70 Ne = mg (A) = ante

metas mr(A) + mg(A) + mb(A)

However, with N,, = F,,/h’v = F,,/h'c'm (h' = c,(m) = oii Bun anh), stiriniaanse
L: :
Planck’s constant, v = frequency, c’ = velocity of mr(A) + mg(A) + mb (A)
 Visual Response Functions and the Spectral Properties of Visual Pigments 597

Eq. 3(8.2.4) entails the following relations: f(m — m,), f(m — m3), f(m — m3) may be sup-
posed known for any wavenumber m. Thus, by
Pif(m— _m,) + py f(m— mz) applying the normalizing conditions, the follow-
+ pi3f(m— m3) ing eight independent linear and homogeneous
c.(m) ¥ G(m) equations in the nine coefficients p,, are obtained
which fix their values apart from a common con-
Praf(m— m,) + po f(m — m3) stant factor:
oF + Px f(m— m3)
c,(m) = G(m) 0 = pr f(u— m,) + px» f(u— m2)

P3if(m— m,) + py f(m— mz) + P23 f(u— m3)

¢,(m) = + p33f(m — m3)
0 = p3,f(u— m) + px f(u— mz)
G(m)
+p33f(u— m3)
G(m) = (pi + Par + Psi) f(m — m)
0=pif(e— m) + Profle — m2)
+ (Pio + P22 + P32) f(m — m2)
+pi3f(v — m3)
+ (p13 + Pos + P33) f(m — m3)
0 = ps3 f(v — m) + px f(v — mz)
c,(m) + ¢,(m) + ¢,(m) = |
+p33f(v — m3)
[4(8.2.4)]
0=pif(w— m) + Pr f(w— mz)
For convenience in comparison, the indefinite-
ness in the p,, coefficients is removed, except
for an overall constant factor, by applying cer-
+pafw— ms)
tain normalizing conditions which can be intro- 0 = poi f(w— m) + Pr f(w— m2)
duced without loss of generality. The empirical
color-matching functions r(A), g(A), D(A) are + P23 f(w — m3)
supposed expressed in a system with three mono-
chromatic stimuli—at wavenumbers u = |/a, Puf(hy — m) + Prof ~ m2)
v = 1/b, w = 1/c, or wavelengths a, b, c, respec-
tively, as primaries, so that +pisf(h; — m3)

ed wlan h, avs.c.(
191 )ileaO 2,aiic,(iv) 30 = puf(hy — m) + pr f(hy — m2)

cide Derr Colt), alae bec(Ww) 0 +po3f(h; — m3)

c(u)=0 c¢,v)=0 c,(w)=1 Pau f(h2 — m) + pr f(h2 — m2)

Furthermore, the units of primary stimuli are +po3f(hz — m3)

taken so that at two fixed wavenumbers h, = 1/d,
= ps f(hy — m1) + P32 f(h2 — m2)
h,=1/e, respectively, the following equations
hold: +D33f(h2 — ms) [5(8.2.4)]
c.(h,) = c(h,)
As a ninth equation to fix absolute values we put
c(h) = ¢,(h)
Pip t+Px + P33 = 1 [6(8.2.4) |
This is the W. D. Wright system of normalization
(Section 5.6.1). With the p,,’s determined by Eqs. 3(8.2.4) and
For any fixed choice of the maximal absorp- 4(8.2.4), the chromaticity coordinates of the
tion wavenumbers m,,m,,m,, the values of monochromatic stimuli can be computed at all
 598 THEORIES AND MODELS OF COLOR VISION
wavenumbers from Eq. 4(8.2.4) for comparison
with the actual empirical values. This can be done
without knowing the spectral transmittance func-
tion t(m) which has been eliminated by the
W.D.W. method of normalization. The model does
not actually predict the relative spectral transmit-
tances t(m), but if it yields the correct spectral
chromaticities, then the relative values of t(m)
can be calculated from the experimentally de-
termined color-matching functions using one of
the Eqs. 3(8.2.4) or 4(8.2.4), giving in the latter
case
aye m| F(A) “£% + b(A)|
[7(8.2.4)]
where 7(A), Z(A), D(A) are the empirical color-
matching functions. The four equations give ex-
actly the same relative values for t(m) if and only
if the model predicts precisely the correct chro-
maticities. If the predictions approach closely the
correct values, the most reliable estimate of t(m)
will be obtained from Eq. 7(8.2.4). If the devia-
tions from the correct values become large, the
calculation of t(m) loses its meaning.
The normalizing procedure ensures that the
model predictions for the chromaticity coordi-
nates must be correct at wavenumbers u, v, w and
must be partially correct [that is c,(h,) must equal
c,(h,), and c,(h,) must equal c,(h,)] at wave-
numbers h, and h,. Tests of the performance of
the model must therefore be applied to one chro-
maticity coordinate only at wavenumbers h, and
h,, and otherwise to the two independent chro-
maticity coordinates at a selection of wavenum-
bers avoiding u, v, w, h,, andh,.
It remains to specify the function f(--- ) which
is to represent the shape of the spectral absorp-
tion curve of any rhodopsin-type pigment. As the
starting point, Stiles and Wyszecki (1974) took
Dartnall’s standard shape which is based on ob-
jective measurements of a range of visual pig-
ments extracted from various animal eyes, all
pigments that, like rhodopsin, incorporate retinal
as prosthetic group. This shape is specified in
Columns 5 and 6 of Table 1(8.2.3) of the previous
section. In the notation used in this section, the
quantity log) f(m — mg) is identical to [log a(A)
— log a(A) max} in. Table 1(8.2.3) and (1/A —
1/X,,) is identical to (m — mj). The continuous
section of Curve 1 of Figure 1(8.2.4) is plotted
from the values of Table 1(8.2.3). However, the
range of (m — m,) is too limited for the present
purpose. If chromaticity coordinates over the
range m= 13,000-25,000 cm™! (A = 770-400
nm) have to be explained, and if the pigments
with spectral maxima between 15,000 and 24,000
cm | (A = 667-416 nm) are to be covered, the
function f( ---) must be defined for (m — mp) in
the range +10,000 to —11,000 cm™'. Stiles and
Wyszecki (1974) extrapolated Dartnall’s standard
shape on the “blue limb”, (m — m,) > 0, and on
the “red limb”, (m— m,) <0, by procedures
outlined in their original paper.
The extrapolation of the blue limb is shown by
the dotted extension of Curve | in Figure 1(8.2.4)
and corresponds to average iodopsin data ob-
tained by Wald et al. (1955). In extrapolating the
red limb of the standard curve, use may be made
of the curve of the gradient Alog, )[a(A)]A(1/A)
against (m—my,). This gradient, to judge by
visual response data, is expected to tend to a
constant value as the extreme red of the spectrum
is approached, that is, as —(m— m,) becomes
large.
Although the gradient curve for Dartnall’s
standard curve does not reach into the extreme
red (in fact, at its extreme point at —3800 cm!
the available objective data become erratic),
Dartnall has re-examined some of his test data
(for frog rhodopsin; private communication to
authors). By applying a justifiable correction, he
concludes that the curve should flatten out much
as shown in Figure 2(8.2.4) (Curve 1). The limit-
ing gradient he obtained for the frog data has a
slightly lower value (0.95 < 107+) than the value
of 1.05 X 10-3 accepted here as a suitable ex-
trapolation for the standard curve. Using the con-
tinuous line gradient Curve 1 of Figure 2(8.2.4),
the complete red limb of the extrapolated stan-
dard curve is derivable to arbitrarily low values
of m.
An alternative function f(---) can be derived
from the rod spectral sensitivity curve expressed
in quantum units and corrected for lens absorp-
tion. This is plotted as the continuous part of
Curve 2 in Figure 1(8.2.4). On the blue side of the
maximum, (m — m,) > 0, the standard Curve 1
and the rod Curve 2 show a certain systematic
difference which grows into a major discrepancy
as the broken-line section of the curves is reached.
The form of the extrapolated (broken line) section
of the rod curve was allowed to take the course
shown as there is no substantial evidence that the
visual response curve rises in the way indicated by
the objective curve in the blue at large (m — my).
 Visual Response Functions and the Spectral Properties of Visual Pigments 599

O a

o% ¢
7

--@- -~9..
J ie —
ft ~seL_

hart

aS

ie ae
PS °
= Pe
1 2
= Pee
oe a“
8 : 70”
ae
00°

] a

10 OOO 7500 5 000 2 500 fe) -2 500 =5 000° =7506

M-Mo

Fig. 1(8.2.4). Rhodopsin-type curves in the form log,, f(m — m,). Curve 1: Dartnall’s stan-
dard curve. From (m — m,) = 5,400— 10,000 cm ! extrapolated in accordance with iodopsin
data obtained by Wald et al. (1955). Curve 2: Rod-sensitivity curve obtained from logjo
[t(A)V(A)/A] where ¢(2) is the spectral transmittance of the lens and V(A) the CIE standard
luminous efficiency function (quantum basis) for scotopic vision. The portion for (m— m,) >
5400 cm! is an arbitrary extrapolation (from Stiles and Wyszecki, 1974).

The form of the visual response curve in the blue
is, of course, subject to an increasing uncertainty
in the correction for lens absorption as (m — my)
increases. The discrepancy in the blue between
the objective pigment absorption curve and the
visual response curve of the rods has been a
puzzling anomaly but appears to be real (Denton,
1958).
Even on the red side of the maximum, the two
curves do not quite agree, the rod curve being
slightly steeper. The rod curve has been followed
experimentally far deeper in the red than the
objective curve, in fact, up to values of (m — mo)
approaching — 10,000 cm! (Griffin et al., 1947),
and is consistent with a constant gradient in the
extreme red equal approximately to 1.44 x 10-7
[see Curve 2, Figure 2(8.2.4)].
For cone vision, evidence on the shape of the
visual response curve in the extreme red is pro-
vided by the large-field color-matching data of
Stiles and Burch (1959; see Section 5.5.4), and
specifically by the variation of the color-matching
function r(A). For this function, the gradient in
the red rises to a value of approximately 1.726 X<
10-3 where it remains nearly constant from
m = 14,000 cm! to the limit of the measure-
ments at approximately 12,000 cm™'. This gradi-
ent is very close to the value of 1.75 X 10°?
derivable from the mean foveal cone threshold
data in the deep red obtained by Goodeve (1936)
and Wald (1945a). A modification of the rod
gradient Curve 2 of Figure 2(8.2.4) to yield a
limiting gradient of 1.726 < 107° is shown as the
broken line extension (Curve 3).
A difference of long-wavelength gradient of
the visual response curves of rods and cones has
been known for some time (Griffin, Hubbard,
and Wald, 1947). Reviewing all the then available
 600 THEORIES AND MODELS OF COLOR VISION

2.0

oO

ro)

10?
t(X)V'(
ZAC
IZA)
Alogio
IZA
(17d)
lO°Alog,,a(d)/A
Oo uo

O “1000 “2 000 -~3000 -4000 -5 000 -6 OOO

m-M,

Fig. 2(8.2.4). Gradient curves of the red limb. Curve 1: Gradient (left scale in figure) of
Dartnall’s standard curve in accordance with a refined analysis made by Dartnall in a private
communication. Limiting gradient slightly raised to 1.05 10~* as compared to 0.95 10>
obtained by Dartnall for his frog data. The A refers to a 400 cm‘ step in 1/A. Curve 2:
Gradient (right scale in figure) of rod-sensitivity curve obtained for the quantum form of V(A)
after correction for the spectral transmittance ¢(A) of the lens. The A refers to a 250 cm! step
in 1/A. Curve 3: Modification of the gradient of the rod-sensitivity curve (Curve 2) to give
limiting gradient of cone curve. The red limb of the cone curve is identical to the red limb of
the rod curve from (m— m,)=0 to —3500 cm! (from Stiles and Wyszecki, 1974).

visual response data, Stiles (1948) derived gradi- an objective blue limb with a red limb based on
ents 0.79 hv/kT = 1.59 X 1073 for the rod curve visual response data, and vice versa. |
and 0.87 X hv/kT = 1.75 X 107? for the cone As color-matching functions, four sets of data
mechanism dominant at long wavelengths, and were considered:
suggested that the fairly small difference might be
attributable to a difference of chromophore of Small-field data:
rod and cone pigments. Although we now know
this is almost certainly not the case, there is the (1) CIE 1931 Standard Observer (Section 3.3.1)
possibility that the difference is associated merely (2) Stiles 2° Pilot Data (Section 5.5.3)
with the position along the spectrum of the peak
absorption wavelength. If so and this affects the Large-field data:
three cone pigments, the basic assumption (i)
made above could not be strictly valid. (3) Stiles 10° Pilot Data (Section 5.5.3)
As the considerations affecting the red and (4) Stiles-Burch 10° Data (corrected for rod
blue limb of the function f(---) are different, intrusion; Section 5.5.4)
Stiles and Wyszecki (1974) made the calculations
for six different shapes including combinations of However, only the CIE 1931 and Stiles—Burch
 Visual Response Functions and the Spectral Properties of Visual Pigments
sets were retained for the main calculations as the
others gave insignificant differences. For each of
the chosen sets, the data were expressed as color-
matching functions 7(A), Z(A), b(A) referred to
monochromatic primaries u = 15,500 cm!, v =
19,000 cm~', w = 22,500 cm', respectively, and
normalized by the W.D.W. system to make
r(h,) = g(h,) and g(h,) = b(h,) at the wave-
numbers h, = 17,250 and h, = 22,500) cm™!.
From the Stiles—Burch (1959) investigation (Sec-
tion 5.5.4) were also taken the standard devia-
tions 9o,(m), 0,(m), o,(m), for an individual
observer, of the chromaticity coordinates at each
wavenumber, calculated from the results of the 49
observers. As shown in that investigation, these
deviations arise principally from differences in
the color-matching properties of different ob-
servers, and not from observational uncertainties.
The standard deviations of the mean chromaticity
coordinates of the whole group would be about
one-seventh of the values of o,(m), ,(m), 6,(m)
but the latter are more appropriate in assessing
the performance of the model.
With u, v, w, h,, h, set throughout at the above
values, a model for test is obtained by choosing
first three wavenumbers m,,m,,m, defining
the positions of the peak absorptions of the
three visual pigments, and second, one of the six
(i= 1 to 6) particular shapes selected for the
rhodopsin-type absorption curve. The transforma-
tion matrix P = {p,,} is then computed by means
of Eqs. 5(8.2.4) and 6(8.2.4). From this, the chro-
maticity coordinates c,(m), c,(m), c,(m), and the
function G(m) are calculated at 250 cm! inter-
vals through the spectrum.
To compare the model for the particular choice
(m,,m,,m,, shape i) with the CIE or Stiles—
Burch data, the differences of the chromaticity
coordinates— model value less empirical value—
are determined and also these differences divided
by the appropriate standard deviations:
cm) —r(X)=D(m); — 4,(m) = ah
; _ D,(m)
cg) — 8(A) = Dy(m)s_ ag) =
c,(m) — b(A) = D,(m);
qy(m) = en
[8(8.2.4)]
601
The transmission function t(m) could be
evaluated using G(m) and the sum of the empiri-
cal color-matching functions as required for
selected comparisons.
Computer programs for all the above opera-
tions enabled a wide variety of cases to be tested
covering appropriate ranges of the peak wave-
numbers. About 1500 cases were calculated that
provided the basis for a number of conclusions:
(a) Models using shapes of the f(---) function
with a blue limb provided by the visual response
curve of the rods can be fitted reasonably well to
the empirical chromaticity data. The f(---) func-
tion displayed as Curve 2 of Figure 1(8.2.4) with
its (arbitrary) extrapolation in the blue and the
gradient 1.44 X 107° [Curve 2 of Figure 2(8.2.4)]
in the red was found to be the most successful of
the six that were studied.
(b) The difference between the comparisons
made with the CIE 1931 color-matching data and
with those of Stiles—Burch was not great, but the
model suited slightly better the Stiles—-Burch 10°
data.
(c) With the best shape function f(---) de-
scribed under (a) and the Stiles—Burch 10° data,
the most suitable wavenumbers (m,,m,, m3) at
which the three pigment-absorption curves have
‘their peaks are
m, = 16,200 cm~! (~ 617 nm)
m, = 18,700 cm~! (~ 535 nm)
m; = 23,150 cm | (* 432 nm)
With these values, the best fit to the empirical
(Stiles—Burch) chromaticity data is obtained.
Figure 3(8.2.4) shows the chromaticities (con-
tinuous curves) obtained by the optimum model,
compared with the Stiles-Burch mean data
(broken curves), whereas Figure 4(8.2.4) gives the
variation through the spectrum of the quotients
g,(m) and so on [Eq. 8(8.2.4)]. In the extreme red,
the deviation of the model values from those
observed is no doubt associated with the slight
yellowing of long-wavelength monochromatic
stimuli first observed by Brindley and confirmed
in the empirical chromaticities (Section 5.10). Any
model employing a shape function f(---)
asymptoting to a constant gradient in the red
must fail to explain the yellowing, but a slight
modification in the red limb would probably do
so without disturbing the situation in the rest of
 a ee fp- et

0.8

O o

Coordinates
Chromaticity

0.0

25 000 22 500 20 000 17500 I5 OOO I2 500 cm

m (cm-!)
Fig. 3(8.2.4). Chromaticity coordinates predicted by optimum model (solid line) compared
with chromaticity coordinates found empirically by Stiles—Burch in 10° color-matching investi-
gation (crosses connected by broken line). Large solid dots are normalization points (from
Stiles and Wyszecki, 1974).

SF x,
Le °
2 i ’ ° pi bai y aL
pala *x xx Kieran pal

~ oo 0888 ed kyRs
wane pes
Pret
- sf
-2
°g _o
2 =A vay a

-3 dae og

25000 22 500 20000 17 500 15 000 12 500 cm"

m (cm)
Fig. 4(8.2.4). Quotients g,(m),q,(m),q,(m) illustrating agreement between chromaticity
coordinates predicted by optimum model and chromaticity coordinates found empirically,
relative to standard deviation of empirical color matching. Perfect agreement would be
horizontal line at which quotients are zero. Acceptable agreement is the domain of +3 for each
quotient, that is, the differences between predicted and measured chromaticity coordinates are
within +3 times the standard deviation of the measured coordinates (from Stiles and Wyszecki,
1974).

602
 Visual Response Functions and the Spectral Properties of Visual Pigments
the spectrum. Rather similar remarks apply to the
discrepancies in the extreme blue which could
probably be reduced by some change in the un-
certain extrapolated part of f(---) on the short-
wavelength side.
All in all, we regard the chromaticity perfor-
mance of the optimum model as unexpectedly
good; the predicted chromaticity coordinates, if
obtained for a real observer in the Stiles—Burch
measurements, would no doubt have been
accepted as a normal variation from the mean.
The disturbing feature is that the peak wavenum-
ber for the long-wavelength pigment in the opti-
mum model lies so far from the position found in
various psychophysical studies. If we take a model
with peak wavenumbers fairly typical of psycho-
physical fundamentals, say, m, = 17,200, m, =
18,500, m3 = 23,000, we find for the quotients
qm), 9.(m), 4,(m), very bad values in the red,
rising as high as 18, elsewhere between the red
and the blue primaries the quotients are several
times larger than those of the optimum model.
However, the decisive failure of the rhodopsin-
type model appears when we compute the spec-
tral transmittance function t(m) which the
optimum model demands if the empirical
color-matching functions—as distinct from the
chromaticity coordinates—are to be correctly re-
produced.
The good chromaticity predictions justify the
use of Eq. 7(8.2.4) in conjunction with the Stiles—
Burch color-matching functions to determine
t(m), and the resulting values are plotted as Curve
1 in Figure 5(8.2.4). The figure also gives a plot of
an empirical t(m) curve derived from estimates of
the average spectral transmittances of the eye lens
and macular pigment given in Section 2.4.6. As
the Stiles—Burch data refer to a large matching
field, the effective maximum density of macular
pigment was reduced to 0.6 of the nominal aver-
age value given in Section 2.4.6. This adjustment
was based on a comparison of the 2° and 10°
color-matching data in the Stiles (1955b) Pilot
Investigation (Section 5.5.3).
With the two curves of Figure 5(8.2.4) normal-
ized to agree at m = 19,000 cm!, we see on the
long-wavelength side, the model curve deviates
downwards very steeply from the empirical curve
in just the region of the spectrum where the latter
is most reliable. At m= 14,000 (714 nm), the
model value is less than 1% of the empirical
estimate. The deviations of the two t(m) curves
on the short-wavelength side are less serious,
bearing in mind the uncertainties in the empiri-
603
cally estimated t(m). We conclude that the
thodopsin-type models are ruled out by the im-
possible eye transmission properties implied.
In the model on which the calculations were
based, it is assumed, in effect, that the pigments
were present in “indefinite dilution” [see assump-
tion (iv) above]. This assumption is not consid-
ered unreasonable and it simplified the numerical
expressions considerably. However, because of the
failure of the model to provide proof that
thodopsin-type spectral absorption curves can ac-
tually serve as a basis for the color-matching
properties in man, a revised model was tried.
The revision consisted of dropping the as-
sumption for one of the pigments, namely the one
with maximum absorption at wavenumber m),.
For this pigment, it is still assumed that the
spectral absorption coefficient a,(m) is given by
Qo, f(m — my) where f(---) is one of the six
adopted standard curves. However, the compari-
son is now made with oa,(m) in accordance with
Eq. 4(8.2.2) of Section 8.2.2 which takes self-
screening into account.
The revised model was used to make a series
of trial runs with similar input data as before, but
the new results were not substantially different
from the previous results. Again, quite good ap-
proximations to the empirical chromaticity co-
ordinates were found when m,,m,,m, were
placed at about 16,200, 18,700, 23,150 cm™', re-
spectively, with f(---) given by Curve 2 in Figure
1(8.2.4) and the Stiles-Burch 10° data as the
other input. Also the spectral transmission curve
t(m) of the ocular media was of similar shape
and thus, unacceptable as before.
It has been advanced, for example by Stiles
(1953), that it is the long-wavelength fundamental
sensitivity (7, or 75 in the Stiles scheme, discussed
in Section 7.4) that implies a pigment absorption
not conforming to the rhodopsin type, whereas
the “green”- and “blue’-sensitive fundamentals
(74, 7, Of 7) are probably consistent with such a
shape. The limitations proposed by color-match-
ing data on the shape of any visual pigment arise
from the fact that its absorption coefficient a(m)
must be a linear combination of the color-match-
ing functions of the form
a(m) = em [Ar(A) + Bg(A) + Cb(A)]
[9(8.2.4)]
provided only that assumption (iv) above is valid.
 604 THEORIES AND MODELS OF COLOR VISION

0.0

1.5 15

1.0 Ke)

2.5} 2.5

Gae,
oy
SK :
2.0 2.0

3.5 35

30 30

45 45
25 000 22 500 20000 a 500 15 000 12 500
m (cm
Fig. 5(8.2.4). Curve 1: Logarithm of relative spectral transmittances of ocular media derived
from Stiles—Burch 10° color-matching functions and model [G(m)-function]. Curve 2: Loga-
rithm of relative spectral transmittances of ocular media expected in accordance with objective
measurements (Section 2.4.6). (From Stiles and Wyszecki, 1974.)

Here A, B, and C are constants independent of
m; and again the color-matching functions are
expressed in the W.D.W. system of normaliza-
tion.
For pigments absorbing maximally at long
wavelengths, greater, say, than 540 nm, this ex-
pression can be used to illustrate the failure of the
rhodopsin shape. In fact, Stiles and Wyszecki
(1974) show that no visual pigment consistent
with the color-matching data can have its maxi-
mal absorption coefficient at a wavelength greater
than 591 nm. They also show that if the absorp-
tion curve is of the rhodopsin type, its peak
wavelength must be less than 556 nm. On the
other hand, for the pigment associated with the
“green“-sensitive cone process and absorbing
maximally at about 540 nm, a rhodopsin shape is
not inconsistent with the color-matching data.
8.2.5 Dichromatism and Fundamental
Spectral Sensitivities
(i) K6nig Fundamentals. According to the tri-
chromatic theory of color vision, one particular
 Visual Response Functions and the Spectral Properties of Visual Pigments 605

set of color-matching functions defining the
color-matching properties of the normal trichro-
mat is singled out from the infinitely many
mathematically equivalent sets because the re-
spective functions correspond to the actual spec-
tral sensitivities of three distinct response systems
present in the retina of the trichromat’s eye and
mediating the trichromat’s color-matching prop-
erties. The second crucial step in the theory, as it
was developed particularly by A. Konig (1903), is
that in a dichromat (see Section 5.14.2) there are
just two distinct response systems that correspond
to two of those present in the trichromat and have
respectively the same relative spectral sensitivities
as the latter. On this theory, three types of dichro-
mat would be possible, all possessing reduced
forms of trichromatic vision but each lacking one
—a different one—of the three normal response
systems. If protanopia, deuteranopia, and tri-
tanopia are identified with the predicted three
types, the relative spectral sensitivities of the dis-
tinct response systems—the so-called fundamental
spectral sensitivities—are derivable from the
color-matching functions of the normal trichro-
mat together with the locations in the trichro-
matic chromaticity diagram of the confusion points
of the three varieties of dichromat.
Suppose the fundamentals lacking in the
protanope, deuteranope, and tritanope are the
relative spectral sensitivities 7(A), Z(A), (A), re-
spectively. For definiteness these will be supposed
normalized, so that the maximum sensitivity is in
each case unity. Any set of color-matching func-
tions representing the normal trichromat must be
linearly related to r(A), g(A), b(A). In particular,
the CIE 1931 standard functions x(A), y(A), Z(A)
must satisfy
are then expressed in terms of 7(A), 2(A), b(A) by
substituting for x(A), y(A), Z(A) using the above
relation. This gives
P(A) = (pif + Piha + Pishid?)
+ similar terms in g(A) and b (2d)
P(A) = (Pah + Phar + Poshi)7(r)
+ similar terms in (A) and b (A)
Because the protanope, by hypothesis, lacks the
fundamental 7(A), the coefficients of r(A) in the
expressions for p,(A) and p,(A) must be zero and
it follows that
0=piihi + Piha + Pishi
0 = pafir + Parfa + Proxha [3(8.2.5)]
Thus, f,;, 51, and fs; satisfy two of the equations,
given in Section 5.14.2 as Eq. 7(5.14.2), by which
the chromaticity X,., Y,.,Z,- of the confusion
point of protanopic vision was defined. More gen-
erally, the equations are solved by —
fir = DP C2 Se— BY,
po pe? fay = Ke Zae
[4(8.2.5)]
where K,, is a constant factor. Should the confu-
sion point be at infinity in the direction dY/dX
= a, the coordinates X,,., Y,., Zp- are replaced in
these expressions by 1, a, —(1 + a), respectively.
An exactly similar argument applied to the deu-
teranope and tritanope relates the other elements
of the matrix F to the deuteranopic and tritanopic
confusion points, giving finally

x(A) fi fro fis F(A) x(A) Ky x6 KyXqc K,Xj¢ r(A)

y(A) =[hi fo ha (A) y(A) aS BSA be Kg Yac K,Y,, g(r)
z(A) fa fro fs b(A) z(X) eR, K 77,41 Vb (A)
[1(8.2.5)]
[5(8.2.5)]
where the matrix F = {f,;} has a nonzero de-
provided the confusion points are not collinear
terminant. The protanopic color-matching func-
(in which case K6nig’s hypothesis would be un-
tions
tenable). Thus, the matrix

PA) = pux(A) + pry(A) + p32)

ne Xac Xr

PAA) = prx(A) + pr V(A) + pr37(A) ieee eles

[2(8.2.5)] Lys ne, wie,}
Table 1(8.2.5) Konig and Fick Fundamentals Derived by
Wyszecki and Stiles (1967, First Edition of this Book)“

g(r) b(r)
Wavelength 7(X) T(A) (K6nig and (K6nig and
A(nm) (Konig) (Fick) Fick) Fick)

400 0.0004 0.0004 0.0004 0.0381

10 0.0012 0.0010 0.0011 0.1164
20 0.0039 0.0028 0.0042 0.3622
30 0.0110 0.0048 0.0143 0.7773
40 0.0213 0.0046 0.0316 0.9801
50 0.0345 0.0021 0.0559 0.9941
60 0.0539 (—0.001) 0.0912 0.9364
70 0.0826 0.0050 0.1339 0.7223
80 0.1286 0.0286 0.1907 0.4561
90 0.1956 0.0690 0.2680 0.2610

500 0.3072 0.1359 0.3971 0.1526

10 0.4825 0.2452 0.5961 0.0888
20 0.6867 0.3938 0.8098 0.0439
30 0.8408 0.5375 0.9438 0.0237
40 0.9386 0.6620 1.0000 0.0114
50 0.9884 0.7698 0.9904 0.0049
60 1.0000 0.8663 0.9263 0.0022
70 0.9706 0.9467 0.8098 0.0012
80 0.9027 0.9940 0.6531 0.0010
90 0.8018 0.9916 0.4778 0.0006

600 0.6830 0.9580 0.3166 0.0004

10 0.5556 0.8566 0.1907 0.0002
20 0.4276 0.7054 0.1069 0.0001
30 0.3009 0.5194 0.0553 —
40 0.2004 0.3571 0.0272 —
50 G:1252 0.2241 0.0128 —
60 0.0705 0.1297 0.0060 —
70 0.0370 0.0686 0.0028 —
80 0.0197 0.0366 0.0013 —
90 0.0095 0.0178 0.0006 —

700 0.0048 0.0089 0.0003 —

10 0.0024 0.0045 0.0002 —
20 0.0012 0.0022 = ~-
30 0.0006 0.0011 — —
40 0.0003 0.0006 — —
50 0.0001 0.0002 ~ —
60 — — — —

“The fundamentals are obtained from the following transfor-

mations of the CIE 1931 standard color-matching functions,
X(A), y(A), 2(A):
Konig and Fick:
g(A) = — 0.3952 x(A)+ 1.1668 y(A) +0.0815Z(A)
b(A) = 0.0000x(A) + 0.0000 (A) +0.5610Z(A).
Konig:
r(A) = 0.0713x(A)
+0.9625 p(A ) —0.01472(A)
Fick:
r(A) = 0.5960x(A)+0.5151 9(A)—0.12292(A)
The confusion points (fundamental primaries) are located in
the CIE 1931 (x, y)-chromaticity diagram at
ep OHM Gees 0253
SUM oes rei
+= 017 Vie= 0.000
 Visual Response Functions and the Spectral Properties of Visual Pigments 607

has a nonzero determinant and can be inverted to can be assumed to have a maximum value of
give the matrix unity. The required values of K,, Kz, K, are then
those that make this true. The Kénig fundamen-
Tg {Zix} tals, computed from x(A), y(A), Z(A) and the
locations of the dichromat confusion points,
Multiplication of both sides of Eq. 5(8.2.5) by G specified in Section 5.14.2, are given in Table
gives 1(8.2.5) and are plotted in Figure 1(8.2.5).
If the locations of only two confusion points
F(X) = x lenzQ) + gn5(A) + 832(A)] are known, for example, (X,., Yj-, Z4-) and
(X,.5 Yies Z;¢)) Only one fundamental can be de-
rived, the one present in both the dichromatic
B(A) = ge[ea¥(A) + gn F(A) + g257(0)] types whose confusion points are known. In the
case cited, the fundamental determinable is r(A)
and is given by
b(A) 3 Klenk) + gy (A) + 8332(A)]
r(A) = constant [(¥,.Z,¢ — YieZac)X(A)
[6(8.2.5)] a (7. Xi sine XacLZic) ¥(A)

The appearance of assignable constants 4" (XyeNic - YicXie)Z(A)]

K,, Ky, K, as factors in the expressions for
r(A), g(A), b(A) makes possible a simple normal- the constant being adjusted to make unity the
izing procedure for these functions, each of which maximum of r(A).

Sensitivity
Spectral
Relative

Konig Fundamentals
Wyszecki & Stiles (1967)

0.000!
400 450 500 550 600 650 700
Wavelength A (nm)

Fig. 1(8.2.5). Fundamental spectral sensitivities of the Konig type. The values are taken from
Table 1(8.2.5) which are identical to those derived in the first (1967) edition of this book.
608 THEORIES AND MODELS OF COLOR VISION

The luminous efficiency function for normal lay = (Lardy + Lard) KX;
trichromatic vision is given by
+ (Lai diz + Landy )K,¥ jc
V(A) =¥(A) = far) + fog) + fozb (A)
+ (Laid)3 + Lady) K,Z,.
= (K,Y,.)7(A) + (KaYac)8(A) +(Ki%c)b
(A)
Ly = (Lath + Ligtr)KyXp¢
[7(8.2.5)]
+ (Lith + Lita )KpY,.
Thus, the luminous efficiency factors of the
fundamentals in normal trichromatic vision are + (Liyti3 + Lyte) KpZpe

Up Rd ac ly Rg kaes reat G9 a Le (Lith + Ltr)

Ka Xac

[8(8.2.5)] + (Lyty, + Lyte) KaYac

For protanopic vision, the luminous
function V,(A) is given by
V,(A) = LyPiA) : Ly2 D(A)
=1,,8(A) +1,,b(A)
where
Le a (LiPu ats Ly2Pn)KaXac
+(LpiPi2 =a Ly2 P22) KaYic
efficiency + (Lytis + Lites) KaZac
(ii) Fick Fundamentals. An _ alternative to
KO6nig’s theory—but a weaker one—was ad-
vanced by Fick. In its most general form, this
— [9(8.2.5)] assumes that for each type of dichromatic vision
there are two independent linear combinations of
the fundamentals 7(A), 2(A), b(A) of normal tri-
chromatic vision, which play the role of funda-
mentals for the dichromatic type in question.
These dichromatic fundamentals may be written,
for protanopia:

a (Lyi P 13 ay Ly2P23)KaZac P(A) = puiF(A) + 9128(A) + 0136 (A)

lees (LyaPu is Loo P21) Ki Xie P2(A) = paF(A) + px8(A) + P23b (A)
[12(8.2.5)]
a (Ly Pi2 4p Ly2 Px) K Lic
for deuteranopia:
+(LyiPi3 a Ly P23) KZ

For deuteranopic and tritanopic vision

MA) = mA) + WBA) + 1138 A)
MA) = yaiF(A) + 2A) + 1136 (A)
V(A) = Laid(A) + Lad(A)
[13(8.2.5)]
=1,,7(A) +1 ,b(A) [10(8.2.5)]
for tritanopia: |
VA) = LA) bat (A)
BA) = Bu F(A) + B8(A) + B13 (A)
= L,,F(A) + ligB(A) [11(8.2.5)]
B,(A) = By F(A) + Bo B(A) + Bo3b (A)
where
[14(8.2.5)]
ie (Edits Lyd) Ky Xp¢ There are limitations in the choice of values that
may be postulated for the coefficients p; ;, y;;, B ij?
+ (Laid), + Linda) KY, these limitations are indicated below.
Any pair of color-matching functions specify-
+ (Lay dy3 + Lydy3)K,Z,. ing a particular type of dichromatic vision must
 Visual Response Functions and the Spectral Properties of Visual Pigments 609

be expressible as independent linear combina-
tions of the corresponding dichromatic funda-
mentals and, in particular, for the dichromatic
color-matching functions previously specified in
Section 5.14.2. Thus,
PA) = m1P\(A) + M202)
P(A) = 110 \(A) + m202(A)
[15(8.2.5)]
d\(X) = 870A) + 8%(A)
d,(X) = 8 7(A) + 872(A)
[16(8.2.5)]
t() a 71:B,(A) i 717By(A)
t,(A) ae 7B\(A) a 77 Bo(A)
[17(8.2.5)]
where the matrices 7 = {7;;},6 = {6;,},7 = {,}
have nonzero determinants.
Various special forms of the generalized Fick
hypothesis are obtained by assigning specific val-
ues to the elements of the matrices
More generally, any assignment of the ele-
ments of p, y, B could be made as a hypothesis,
provided that
(a) in each matrix, the two rows are independent
vectors so that the color-matching properties rep-
resented are truly dichromatic;
(b) no two of the matrices are related by a
nonsingular linear transformation; for example,
p # Ay, where A is any (2 X 2) matrix of nonzero
determinant. This is essential if the color-match-
ing properties of the three types are to be all
distinct;
(c) all the dichromat fundamentals finally de-
termined (see below)—p,(A), p»(A), y,(A), and so
forth—come out positive at all wavelengths. The
difficulty of a physiological interpretation of
fundamentals that change signs imposes this re-
striction.
The outstanding question is whether, given p,
y, and B, the fundamentals r(A), g(A), b(A) of
the normal trichromat are uniquely determined;
that is, can the matrix F = {f;;} [Eq. 1(8.2.5)] be
calculated? With matrix notation and
Pals! Pi2 Bs)
Pai Pez Ys
i |d 11 d 12 d 13

= (50 P12 ay dy, dy dy;

Poy Ps “Px,

y and B. Fick’s special hypothesis corresponds to

the assumption that in deuteranopia one funda-
mental—y,(A), say—is a “fusion” of the “red” the matrix F must satisfy
and “green” fundamentals r(A) and g(A) of the
normal trichromat, whereas the other fundamen- x(A) 7(A) r(X)
tal—y,(A)— coincides with the normal’s “blue” P| y(A) | = PF] 8(A) | =| 8A)
fundamental b(A). Thus, the assumption is that
z(X) b(A) b(A)
ho}11 ee ‘| [18(8.2.5)]
u er ool
or, because 7(A), 2(A), and b(A) are linearly
independent functions,
where the ratio y,,;/y,, determines the relative
contributions of r(A) and g(A) in the “fusion.”
This assignment of y is commonly combined with
PF = 1 [19(8.2.5)]
the assumption that protanopia and tritanopia Similarly,
arise as in K6nig’s theory simply by the lack,
respectively, of the normal “red” and “blue” DF = by [20(8.2.5)]
fundamentals, so that
and |
TF = 7B [21(8.2.5)]
610 THEORIES AND MODELS OF COLOR VISION

For Eq. 19(8.2.5) to hold, it is readily shown that and

O = pyymy, + Pj2M, + Pi3™M3) Gy 90 -¥i2 923 avy 13¥22 > 913 = Bir B23 — Bi3h22

O = pom, + P22M, + P23M3) Go." Vis20 @ tise 923 = Bi3B2, — Bi Bos

[22(8.2.5)] G39 — Vivtsa S111 721: 933 = By:B22 — Bi2Bo

with and

3 Mio Gad ge, Mo = Caldes M3. = CyZac

LNs waa > fii
i=1 Misi CLA, Mo, — Cl, m3 = C,Zye

and The relation satisfied by the matrix F is then

Gi1 — Pi2P23 ~ P13P22 FO=M [28(8.2.5)]

921 — P13P21 ~ P11P23
where
931 — Pi1P22 ~ Pi2P21
Q = {q;;} and
M = {m;;}
Thus, m,,, ™ ,, m3, Satisfy two of the equations
that define the confusion point of protanopic We exclude initially the exceptional case in which
vision [see Eq. 4(8.2.5)], therefore, the matrix Q, and hence also M, has a zero
determinant. The transformation matrix is
my — CA, pee ihe ane ©, Zp Cc
uniquely determined as

[23(8.2.5)] F= MQ"! [29(8.2.5)]

where C,, is a constant factor. and the fundamentals of the normal trichromat
Similarly, from Eqs. 20(8.2.5) and 21(8.2.5), it are given explicitly by
follows, respectively,

3 7(X) x(A) x(A)

0= Ydjmj,. (i=1,2) z(A) |=F'| x(a) | = OM" 9A)
ja b(A) z(A) 2(A)
[24(8.2.5)] [30(8.2.5)]
3
O= Dtjmj (=1,2) This relation written out in the form
j=l
7 l 1 l
[25(8.2.5)] r(A) mealrel ahs Coe 03 AnG

where a £5 l l ]
g(A) — Ad, Gi + Ay) C) + Ap; Gi
3
M j2 dis » KiQi2 (j= E23)
r= b(A) =A ulnayy
31 G loidyll
32 G, 33 C,

[26(8.2.5)]

3 where
M 73 7 . fii (j= L, 243)
4 Ee = drs |81 X(A) ag g,2V(A) " 8532(A)]

[27(8.2.5)] (r,s= 1,2,3)

 Visual Response Functions and the Spectral Properties of Visual Pigments 611

G =fe9 Si pwith £,3Zny) = 0 only if the chromaticity N is in line

with two confusion points. Thus, the constants
Age Xae Xr¢ 1/G,,1/Cy, 1/G, are uniquely determined.
H = :6 Lae Ye
Note that the relative spectral sensitivities of
the fundamentals 7(A), Z(A), b(A), as derived
Lie Lac Lrc
above, are unaffected by a change of normaliza-
tion, provided the appropriate adjustments are
obviously involves the constants C,, C,, C, whose made to the postulated dichromat fundamentals.
values, yet to be determined, depend on the nor- Thus, if, by a different normalization, the funda-
malization of r(A), g(A), D(A). It is difficult to mentals become
determine the values of SN OF TeOFT 5i
r(A), Z(A), b(A) are supposed normalized to have r'(A) = k,r(A), 8(A) = kB(A),
maximum values unity. A suitable alternative is
and
to suppose 7(A), g(A), b(A) normalized so that
for a conveniently chosen spectral radiant power b’(A) = k,b(A)
distribution {P, dA} whose CIE chromaticity
(Xx; Yy, Zy) 1S not collinear with any two of the the expressions for p (A), p,(A), for example, must
three dichromatic confusion points, the following be changed to
equalities hold:
mir) =(S)ray + (S)
FO)
(ge)
[FOA)P. ad = [AJP ddA = Xy

[ZAP dd = [PA)P. ad = Vy

[oQ)P,dr= [2OQ)P, AREAS

m0) = (Sra + (B)F70)
[31(8.2.5)] + P23. JPQ)
|,

The equal-energy stimulus, P, = 1 for all A, is a

to preserve the postulated relation between the
satisfactory choice.
protanope and normal trichromat fundamentals.
On multiplying the above expressions for
Similar expressions are obtained for y,(A), y>(A),
F(A), B(A), (A) by {P, dX} and integrating over
B,(A), B(A). With these changes, the trichromat
the spectrum, the following equations for the
fundamentals derived in the way shown above
constants C,, Cz, C, are obtained:
correspond to the same relative spectral sensitivi-
1 1 ties as before.
t Cle Buc + Bae uF Buc The exceptional case, in which the determinant
Dp t
of B equals zero, implies that the determinant of
l l l M equals zero and therefore, that the confusion
i Mi Buc ( Bc st Baa points of protanopic, deuteranopic, and tritanopic
D t
vision are in a straight line in the chromaticity
diagram. This is certainly not the case in human
l 1 l
fA Bae + Bac a BG vision, although for other conceivable systems of
Pp Tt
trichromatic and dichromatic vision, the gener-
alized Fick hypothesis would admit the possibil-
[32(8.2.5)] ity.
The luminous efficiency factors of the funda-
where
mentals appropriate to the generalized Fick hy-
pothesis are readily obtained:
B,. - 9rs1 851 Xn re 852 Yn + g,3Zy |
(a) Normal trichromat.
The matrix B= {B,,} cannot have a zero de-
terminant, since |Q|#0 and (g,,Xy + g,.Yy + V(A) = yA) = fir) + fog) + frsb(A)
612 THEORIES AND MODELS OF COLOR VISION

Thus, /, = fo), 1, = fo, and /, = f,3, and all are Fick’s special hypothesis (see above), corre-
computable from Eq. 29(8.2.5). sponding to
(b) Protanope.

V(A) ot Ly P(A )L 2 D(A)

= (L,im1 + Lyom) P(A)

+ (Lym + Ly2t% )p2(A)

leads to the trichromatic fundamentals given in
= 1,10 (A) + 1,202(A)
Table 1(8.2.5). These are calculated from Eq.
30(8.2.5) and assume
where

(1) normalization of the trichromat fundamen-

Loy = Lym, + L271; Loa = Lyi 2 + Ly2M2
tals on integrals over the equal-energy spectrum
(fr(A) dA = {x(A) dA, etc.); and
In the matrix
(2) Yu = Yi:

= (3 P12 ad This is the usual quantitative basis for deriving

Por P29 P33 so-called Fick fundamentals. With these assump-
tions, it is only for the function 7(A) that the
the rows are linearly independent so that at least relative values through the spectrum differ from
one of (111022 — P12P21)» (P12P23 — P13P22): those of the corresponding K6énig fundamental.
(P11P23 — P13P21) is different from zero. Suppose Other choices for the ratio y,, : y,7 (with the same
(P11P22 — P12P2;) #9; then the two pairs of normalization) lead to different r(A) fundamen-
equations tals. In fact, however, values of y, greater than
y,; turn out to be inadmissible because they lead
3 to negative values of r(A) at some wavelengths.
a Pifiy = MP1 + 712021 Justification on physiological grounds of a partic-
i=]
ular choice of y,; : y;> 1s difficult. Once the funda-
a mentals 7(A), Z(A), b(A) have been found, they
> Piifiz = TM1P12 + T2P2
can be renormalized to have maximum values of
i=1 unity, and this has been done for the values of
Table 1(8.2.5) which have been plotted in Figure
and 2(8.2.5).
3
> Prifir = 7 1P 11 + M7 Pr (iii) Other Derivations of Kénig Fundamentals.
i=1 Careful consideration of the confusion points of
dichromats, of the known short-wave deficiencies
3
in the CIE 1931 standard color-matching func-
» Prfi2z = M1P12 + TP)
i=] tions, and of other relevant factors in trichromatic
and dichromatic vision, led Vos and Walraven to
a new derivation of fundamental spectral sensitiv-
contained in Eq. 19(8.2.5) yield, respectively,
ity functions of the Konig type (Vos and Walra-
solutions for (7,1, 712), (71, % 2) from which the
ven, 1971; Walraven, 1974; Vos, 1978). Their best
luminous efficiency factors /,, and /,, of the
estimate (Walraven, 1974; and Vos, 1978) yields
protanopic fundamentals are computed.
the following transformation equations:
(c) Deuteranope and tritanope. The pairs of
luminous efficiency factors (/,,,/,.) and (/g; /,>) F(A) = 0.1551646%'(A) + 0.54307637’(A)
are determined in the same way as for the pro-
tanope. —0.0370161z’(A)
 Visual Response Functions and the Spectral Properties of Visual Pigments 613

10
(A)

0.l

Ool

Relative
Sensitivity
Spectral

Fick Fundamentals
Wyszecki & Stiles (1967)

0.000!
400 450 500 550 600 650 700
Wavelength A (nm)

Fig. 2(8.2.5). Fundamental spectral sensitivities of the Fick type. The values are taken from
Table 1(8.2.5) which are identical to those derived in the first (1967) edition of this book.

g(A) = —0.1551646x’(A) + 0.4569237p(A) (x’, y’)-chromaticity diagram:

+ 0.02969462’(A) xe ae 077465ei2 y=" 0.2535

b(A) =0.0073215z’(A)
[33(8.2.5)]
The color-matching functions x’(A), y’(A), Z’(A)
are Judd’s (1951) modified CIE 1931 standard
functions as refined by Vos (1978). They are
described in more detail in Section 5.5.2 and
given in Table 1(5.5.2) in the Appendix. The
values of the fundamentals 7(A), 2(A), D(A)
(which Vos and Walraven denote by R,, Gy, By,
respectively), calculated by Eq. 33(8.2.5) are given
in Table 2(8.2.5) and illustrated in Figure 3(8.2.5).
The confusion points which provide the im-
portant constraints on the coefficients in Eq.
33(8.2.5) have the following coordinates in the
x’, = 1.4000 y5, = —0.4000
x;.= 0.1747 = y, = 0.0060
Further constraints on the coefficients in Eq.
33(8.2.5) are introduced by gauging the relative
spectral sensitivities of the fundamentals so as to
correspond to effective population densities of the
“red,” “green,” and “blue” foveal cone receptors
in the ratios R: G: B= 32:16:1. Vos and
Walraven postulate that these population density
ratios are related to the different Weber fractions
that have been found by Stiles (1946) (see Section
8.4.4) for the responses of the cone mechanisms at
high luminances, and to the visual phenomenon
614 THEORIES AND MODELS OF COLOR VISION

of invariant-hue perceptions when observing cer-
tain monochromatic stimuli at different lumi-
nance levels (Bezold—Briicke effect, Sections 5.9
and 5.13). The invariant hues, “blue” and “yel-
low,” are observed (Walraven, 1961) for stimuli of
wavelengths A, = 475.5 nm and Ay = 570 nm.
They are believed to be invariant because at these
wavelengths (A, and A,) the neural opponent-
process mechanisms (red/green) and
(yellow /blue) are thought to be in balance; that
is, the rates at which these mechanisms saturate
remain the same with increasing level of stimula-
tion. In the underlying visual model adopted by
Vos and Walraven, this balance is achieved when
b(A,) and [7(Az) + Z(Ag)] are at the ratio 1: 16
for \, = 475.5 nm and when r(Ay) and g(Ay)
are at the ratio 2:1 for Ay = 570 nm.
The sum of the three fundamentals defined by
Eq. 33(8.2.5) yields the Judd modification of the
CIE standard luminous efficiency function [Sec-
tion 5.7.2(11)]:
V(A)moa = F(A) + B(A) + b(A)
[34(8.2.5)]

Table 2(8.2.5) Konig Fundamentals Derived by Vos and Walraven (Walraven, 1974, Vos, 1978) and by
Smith and Pokorny (1972)

Wavelength Vos and Walraven Smith and Pokorny

A (nm) 7(X) g(A) b(A) 7(A) g(A) b(A)
400 0.00096 0.00056 0.00127 0.00273 0.00177 0.00224
10 0.00246 0.00157 0.00337 0.00553 0.00377 0.00467
20 0.00561 0.00409 0.00781 0.01005 0.00745 0.00851
30 0.00870 0.00786 0.01074 0.01489 0.01241 0.01164
40 0.01239 0.01367 0.01184 0.01918 0.01872 0.01287
50 0.01568 0.02034 0.01078 0.02186 0.02494 0.01179
60 0.02083 0.02972 0.00946 0.02630 0.03370 0.01032
70 0.03524 0.04758 0.00816 0.03991 0.05108 0.00892
80 0.06175 0.07174 0.00554 0.06495 0.07404 0.00605
90 0.1013 0.1034 0.00327 0.1032 0.1047 0.00357
500 0.1664 0.1547 0.00194 0.1675 0.1555 0.00212
10 0.2689 0.2330 0.00113 0.2696 0.2334 0.00123
20 0.3926 0.3168 0.00056 0.3930 0.3170 0.00061
30 0.4925 0.3692 0.00030 0.4926 0.3694 0.00033
40 0.5628 0.3911 0.00015 0.5629 0.3911 0.00016
50 0.6077 0.3872 0.00006 0.6078 0.3871 0.00007
60 0.6329 0.3620 0.00003 0.6329 0.3621 0.00003
70 0.6355 0.3165 0.00002 0.6355 0.3164 0.00002 ~
80 0.6146 0.2554 0.00001 0.6146 0.2554 0.00001
90 0.5698 0.1872 0.00001 0.5697 0.1872 0.00001
600 0.5064 0.1246 0.00001 0.5063 0.1247 0.00001
10 0.4271 0.07586 0.00000 0.4271 0.07586 0.00000
20 0.3378 0.04322 0.00000 0.3377 0.04322 0.00000
30 0.2421 0.02289 0.00000 0.2421 0.02289 0.00000
40 0.1634 0.01160 0.00000 0.1634 0.01160 0.00000
50 0.1013 0.00566 0.00000 0.1014 0.00564 0.00000
60 0.05820 0.00280 0.00000 0.05820 0.00280 0.00000
70 0.03068 0.00132 0.00000 0.03068 0.00132 0.00000
80 0.01634 0.00066 0.00000 0.01634 0.00065 0.00000
90 0.00790 0.00031 0.00000 0.00790 0.00031 0.00000
700 0.00395 0.00015 0.00000 0.00395 0.00015 0.00000
 Visual Response Functions and the Spectral Properties of Visual Pigments 615

Table 2(8.2.5) — Continued

The fundamentals are obtained from the following transformations:

Vos and Walraven

r(A) 0.1551646 0.5430763 —0.0370161 xs)

g(A) | =| —0.1551646 0.4569237 0.0296946 y (A)
b(A) ) 0 0.0073215 z'(X)

Smith and Pokorny

r(A) 0.15514 0.54312 —0.03286 x'(A)

2(A) |=| —0.15514 0.45684 0.03286 y(A)
b (A) 0 0 0.00801 z'(X)

In the case of Vos and Walraven, the color-matching functions x’(A), y’(A), Z’(A), are the Judd modified
CIE 1931 standard functions as further revised by Vos (1978) [see Table I(5.5.2)]. In the case of Smith
and Pokorny, the color-matching functions x’(A), y’(A), Z’(A), are the original Judd modified CIE 1931
standard functions [see Table 1(5.5.2)].

The following confusion points were used:

Vos and Walraven Smith and Pokorny

X pe = 0.7465 yee 0.259 Ker UACS pie hehbeh aRis

X 4e= 1.4000 Vie= —0.4000 X ae = 1.4000 Vae = — 0.4000
Xj¢= 0.1747 Y,.= 0.0060 X,.= 0.1748 Vie 9.0000

Smith and Pokorny (1975), in extending their
earlier study (Smith and Pokorny, 1972) on the
spectral sensitivities of protanopes and deu-
teranopes, arrived at a set of K6nig-type funda-
mentals that differed somewhat from those of Vos
and Walraven described above. The transforma-
tion equations are as follows:
7(A) = 0.15514x’(A) + 0.54312 y(A)
—0.032862’(A)
Z(A) = —0.15514x’(A) + 0.45684 (A)
+ 0.032862’(A)
b(A) =0.008012’(A)
[35(8.2.5)]
The color-matching functions x’(A)(A), y’(A),
z’(A) are Judd’s (1951) modified CIE 1931 stan-
dard functions given in Table 1(5.5.2). The values
of the fundamentals 7(A), Z(A), b(A) calculated
by Eq. 35(8.2.5) are given in Table 2(8.2.5) and
are illustrated in Figure 4(8.2.5). Pokorny and
Smith (1975) do not specify the coefficient by
which z’(A) is to be multiplied to obtain b(A).
We have chosen the value 0.00801 so as to make
the ratio [r(Ap) + g(Ag)|/b(Ag) = 16 at AZ =
475.5 nm, in agreement with the proposal of Vos
and Walraven discussed above. Boynton (1979)
put the coefficient at 0.01608 by scaling the
b(A)-function so that b(A)/[7(A) + g(A)] = 1.0
at A = 400 nm.
The Pokorny and Smith (1975) transformation
equations are based on confusion points which, in
the (x’, y’)-chromaticity diagram, are located at
nO ey ee
x, = 1.4000 ys, = —0.4000
x/,-=10.1748 yin= 90.0000
616 THEORIES AND MODELS OF COLOR VISION

10

Ol

0.0|

K6énig Fundamentals
Sensitivity
Spectral
Relative Vos & Walraven
(I971,1974 1978) —

OO 450 500 550 600 650 700

Wavelength a (nm)
Fig. 3(8.2.5).Relative spectral sensitivities of K6nig-type fundamentals derived by Vos and
Walraven (Vos and Walraven, 1971; Walraven, 1974; Vos, 1978). Numerical values are given in
Table 2(8.2.5).

The essential difference between these confusion
points and those adopted by Vos and Walraven is
in the location of the tritanopic confusion point
(Xe. Yi) Which Pokorny and Smith locate on the
alychne (chromaticity locus, y’ = 0, of zero lumi-
nance) in accordance with the idea, proposed by
other investigators as well, that the (minimum-
flicker) luminous efficiency function of the nor-
mal trichromat is mediated only by the “red”-
and “green’-cone mechanisms and that the
“blue’”-cone mechanism makes no contribution to
“Juminance signals.” In particular, they put
Vinoa(A ) a r(X) a g(A) [36(8.2.5)]
This constraint on the coefficients of Eq. 35(8.2.5)
helps to improve the agreement between pre-
dicted and observed color-matching functions of
tritanopes.
- In their comparison of predicted and observed
data of all three types of dichromat, Pokorny and
Smith note small but systematic discrepancies in
the spectral distribution of the “red” cone sensi-
tivity of both the deuteranope and tritanope. The
discrepancy is difficult to explain and Pokorny
and Smith speculate that it may be caused by a
“selective screening filter of optical density at 460
nm of 0.12 and spectral shape similar to macular
pigment.” .
More recently, in his critical and valuable re-
view of the fundamental database of normal and
dichromatic color vision, Estévez (1979; and
slightly amended data received by the authors in
a private communication, 1981) derives K6énig-
type fundamentals from the Stiles 2° pilot
color-matching functions 7(m), Z(m), b(m) (Sec-
tion 5.5.3)’and a set of confusion points which he
locates in the Stiles 2° (r, g)-chromaticity dia-
gram at
Yel 25 gie=— 0.025
ric = 1.600 Zac = —0.612
r,,= 0.090 — g,, = —0.085
 Visual Response Functions and the Spectral Properties of Visual Pi gments 617

10

Konig Fundamentals
Sensitivity
Spectral
Relative
Pokorny & Smith
0,00!
(1975)

0000!
400 450 500 550 600 650 700
Wavelength A (nm)
Fig. 4(8.2.5). Relative spectral sensitivities of K6nig-type fundamentals derived by Smith and
Pokorny (1975). Numerical values are given in Table 2(8.2.5).

The fundamentals p(m), y(m), B(m) are ob-
tained from
o(m) =0.38457(m) + 1.0062 2(m) + 0.05125(m)
y(m) = 0.05387(m) + 1.44122(m) + 0.1183b(m)
B(m) = 0.03052(m) + 1.551b(m)
[37(8.2.5)]
Tables 3(8.2.5) and 4(8.2.5) reproduce the data
obtained by Estévez (1979; slightly amended in
1981) including the original Stiles 2° pilot data
which served as input data. Figures 5(8.2.5) and
6(8.2.5) illustrate some of the data given in these
tables.
The K6nig- and Fick-type fundamentals dis-
cussed in this section are not the only types of
fundamentals that have been proposed by vision
scientists. When one abandons the condition that
dichromats are strictly reduction forms of nor-
mal trichromats, the confusion points no longer
play a unique role in the derivation of the trans-
formation coefficients that relate linearly the
color-matching fundamentals. Other visual mod-
els can then be used that are based on other
experimental data and that are not necessarily
linked with the color-matching properties and
heterochromatic brightness matching properties
of observers with color-defective vision. Examples
of such models are derived from the experimen-
tally determined field sensitivities of component
visual mechanisms in the fovea (Stiles’ 7-mecha-
nisms, Section 7.4), from Dartnall’s standard
shape of the visual pigment absorption coefficient
(Sections 8.2.3 and 8.2.4), and from color-match-
ing experiments made at moderate and at high
(pigment-bleaching) levels of retinal illuminance
(Wyszecki and Stiles, 1980, Section 8.2.6).
For the purpose of comparison, we have repro-
duced in Figure 7(8.2.5) and Table 5(8.2.5) data
on the spectral absorption characteristics of hu-
man cones obtained by Bowmaker and Dartnall
Table 3(8.2.5) Stiles 2° Pilot Color-Matching Functions and Spectral Sensitivities of Fundamentals Derived by
Estevez (1979, 1981) from the Color-Matching Functions [See also Table 4(8.2.5)]

Logarithms of Stiles 2° Pilot Logarithms of Spectral

Wavenumber Color-Matching Functions Sensitivities of Fundamentals
m (cm!) log?(m) log Bm) — log b(m) ~—logp(m) ~— Log y(m) ~—_log B(m)
25,500 2563 —#3:175N — 1.744 —2 8853 — 2.8804 — 1.5526
23,250 = 2.264 —2.903N — 1.439 2.51/08 — 2.5574 —2Si5
25,000 —2.014 —2.648N — 1.208 —2.3429 2345) eet Poe Ps
24,750 — 1.766 +2\362.N — 0.930 — 2.0974 — 2.0805 —O7505
24,500 == £528 —2.244N =(. oz — 1.8769 = 18379 — (Saag
24,250 — 1.430 — 1.983N =().539 1518 — 1.6953 — 0.3694
24,000 —e1.337 — 1.840N — 0.389 — 1.6451 = 1.5500 (ess
23,750 Fe f:2 13 —TIS0N =O.251 — 1.5430 — 1.3997 — 0.0904
23,500 ana PNG —1.561N =), 153 = 15084 — 1.3076 — 0.0010
23,200 men) Be) — 1.680N — 0.098 — 1.4718 = 1:2160 — 0.0534
23,000 — 1.474 — | 756N — 0.050 a AI32 = 1.1312 — 0.0968
22,750 =|: 71S —2.034N —0.019 =='1:3850 — 1.0460 —0:1234
22,500 == "5t162 —2.897N — (0.000 cael Ee — 0.9880 — 0.1374
22,250 — 1.599N = 1797 0.031 == |3264. — 0.9393 —0.1071
22,000 —1.242N — 1.396 — 0.081 — 1.2796 — 0.8794 — 0.0471
21,750 — 1.027N = 14165 —0.103 —= 172055 — 0.7976 — 0.0204
21,500 —0.859N = (957 —(u/y — 1:1105 OG — 0.0579
21,250 —0.747N — 0.809 +0221 — 1.0068 — 0.6240 — 0.1062
21,000 —0.657N — 0.681 — 0.343 (132) — 0.5500 = (29
20,750 —0.611N —0.594 — 0.491 — 0.8376 — 0.4962 — 0.3821
20,500 —(.571N —0.514 —=U.G52 — 0.7598 — 0.4385 — 0.5435
20,252 —0.538N — 0.412 — 0.803 — 0.6433 —0.3510 — (0.6909
20,000 —0.529N — 0.309 — 0.970 =O 193 — 0.2578 — 0.8464
19,750 =U552N — 0.204 Sti G2 —-().3953 — 0.1600 — 1.0101
19,500 —0.619N =—=() 119 — 1.435 =) 2875 — 0.0802 = Le
19,250 — 0.802 N — 0.045 = Oot. —0.1941 — 0.0107 — 1.4309
19,000 — 3.084N 0.000 —3.097N ectogWan 0.0309 — 1.6616
18,750 — 0.683 0.022 pawet ohWh — 0.0774 0.0501 — 1.8316
18,500 — 0.359 0.022 —1.874N — 0.0509 0.0476 — 2.0863
18,250 = ela. 0.019 OLN — 0.0243 0.0427 eh E
18,000 0.014 — 0.003 — 1.802N — 0.0065 0.0205 eat eo
17,750 0.143 — 0.047 —1.828N © 0.0000 —0.0218 — eG
17,500 0.262 —0.092 —1.886N 0.0173 — 0.0621
17,250 0.354 —0.184 —1.974N 0.0141 —0.1429
17,000 0.417 — 0.294 —2.098N 0.0044 = (0.23561 .
16,750 0.454 =) 4g —2.254N — 0.0227 —0.3659
16,500 0.454 — 0.669 PBF UO P21 —0.5249
16,250 0.413 =0.957 —2.667N ly eh reyA BS
16,000 0.324 SAME. —2.978N — 0.2692 — 0.9461
15,750 0.185 — 1.889 —3.414N — 0.4299 Seg $f.Vins
15,500 0.000 Sa (585 — 4.442 N cd OR veOs) — 1.4854
15,250 —0.242 2 IOV —3.741 — 0.8857 4 19T)
15,000 rah aye — 2O25IN as aR — 1.1890 — 2.1405
14,750 ==). G00 —2.916N est: WIN WZ ==) 5209 — 2.4802
14,500 — 1.209 3.20 =A 52] — 1.8801 — 2.8487
14,250 = OL —3.645N —=4 SA} — 2.2894 20
14,000 = 2 NS —4.042 N = eh — 2.6890 = 0d
13,750 = 2.414 —4.789N — 5.348 — 3.1021 — 4.0023

618
 Visual Response Functions and the Spectral Properties of Visual Pigments 619

Table 3(8.2.5) — Continued

a
Notes

1. The Stiles 2° pilot color-matching functions refer to monochromatic primary stimuli located at
wavenumbers

mp=15,418 mo=18,997 m,=22,456cm™!

The values of 7(m), Z(m), b(m) of this table deviate slightly from those given in Table 1(5.3.3)
where entries refer to primary stimuli located precisely at wavenumbers 15,500, 19,000, and 22,500
cm', respectively (see Section 5.3.3 for details).

The color-matching functions imply monochromatic test stimuli of unit radiance.

2. The spectral sensitivities p(m) y(m), B(m) of the fundamentals imply monochromatic test stimuli
expressed in quantum units. They are obtained from the color-matching functions 7(m), z(m), b(m)
by means of

p(m)= k,m[0.38457(m)+ 1.0062 g(m)+0.0512b (m)|

y(m) = k, m[0.05387(m)+ 1.4412¢(m)

+0.11835 (m)|

B(m) =kpm[ 0.0305g(m)+1.5551b

(m)]
k,,k,, kg are normalizing constant factors which make the sums of three functions very nearly equal
to one another.

Multiplication of the linear combinations of 7(m), g(m), b(m) by the wavenumber m yields
functions that are based on quantum units.

The values of 6(A) in the long-wavelength region, from 13,750 to 17,500 cm_', have been ignored.

No correction for light losses in the ocular media has been applied to the fundamentals.

(1980) by microspectrophotometry applied trans-
versely through the outer segments of individual
cones.
8.2.6 Color-Matching Data and the Pigment-
Bleaching Model
In Section 5.6.5 we described trichromatic color-
matching experiments made by Wyszecki and
Stiles (1980) at retinal illuminances ranging from
moderate (1,000 td) to high (100,000 td) levels.
The results of these experiments clearly demon-
strated breakdowns of color matches when the
luminance of the bipartite field reached a level
equivalent to approximately 8000 td. Although
the experimental technique used by Wyszecki and
Stiles (1980) differed in important ways from
those used by other investigators, their results
were quite consistent with the earlier ones, nota-
bly those of Wright (1936), Brindley (1953), and
Alpern (1979). In all these studies, bleaching of at
least one of the cone pigments was considered the
important process controlling the observed
breakdowns of the color matches.
Wyszecki and Stiles (1980) adopted the simple
and generally accepted model in which color
matching is determined by the absorptions in
three classes of cone receptors containing, respec-
tively, visual pigments designated “red” (erythro-
labe), “green” (chlorolabe), and “blue” (cyano-
labe). In Maxwell-type color matching (Section
5.3) at a given level of the “white” reference
stimulus, the total number of quanta absorbed in
the visual pigment of each of the receptors of a
particular class is determined by the receptor’s
visual pigment absorption factor as a function of
wavenumber and the absolute spectral radiant
 {

Table 4(8.2.5) Spectral Sensitivities of Fundamentals at Retinal Level and Corresponding Spectral Absorption
Coefficients of Human Cone Pigments as Derived by Estévez (1979; 1981) [See also Table 3(8.2.5)]

Optical
Logarithms of Spectral Density
Sensitivities of Fundamentals of Ocular Spectral Absorption Coefficients
Wavenumber at Retinal Level Media of Human Cone Pigments
m(cm~') —logp’(m) —logy’(m) —logB(m) K(m) —9,(m) —a,(m)_—_—_o(m)
25,500 0.72 0.78 O12 Os, 14.76 12.47 68.90
25,250 0.71 0.77 0.13 1.94 14.92 12.87 67.06
25,000 0.71 0.79 0.14 1.71 14.84 12.33 66.41
24,750 0.73 0.78 0.12 1.45 14.33 12.47 69.11
24,500 0.75 0.78 0.17 1.21 13.67 Laks 60.43
24,250 0.76 0.78 OZ 1.07 1349 12.62 69.55
24,000 0.79 0.76 0.10 0.94 12.47 13.10 72.83
23,750 0.79 0/72 0.08 0.83 12.23 14.44 78.17
23,500 0.81 0.68 0.04 0.78 11.80 15.99 88.89
23,250 0.81 0.63 0.03 0.74 11.69 18.14 92.48
23,000 0.81 0.58 0.02 0.70 11.66 20.26 93.49
22,750 0.79 0.52 0.02 0.68 12.47 23.82 95.43
22,500 0.75 0.46 0.00 0.68 13.45 29:51 100.00
22,250 0.71 0.39 0.02 0.70 15.08 32:81 94.99
22,000 0.68 0.35 0.09 0.68 16.09 36.46 75.25
21,750 0.63 0.29 0.14 0.66 18.37 42.94 65.56
21,500 0.56 0.23 0.25 0.63 21.56 49.99 28.34
21,250 0.50 0.19 0.34 0.59 25.28 57.08 38.16
21,000 0.44 0.15 0.50 0.55 28.94 62.91 24.96
20,750 0.38 0.11 0.66 0.54 34.26 Th 16.79
20,500 0.33 0.08 0.85 0.51 32.82 77.73 10.58
20,250 Ora) 0.05 1.06 0.45 45.14 84.61 6.47
20,000 0.23 0.04 1.30 O37 50.90 88.21 3.73
19,750 0.19 0.04 1.54 0.27 54.39 87.61 202
19,500 0.15 0.01 1.81 0.22 64.08 96.41 © oe
19,250 0.12 0.01 2.10 0.15 68.51 96.26 0.58
19,000 0.09 0.00 2.36 0.12 77.04 100.00 0.32
18,750 0.06 0.00 Dasa 0.10 83.91 99.74 0.20
18,500 0.04 0.01 2.2 0.09 88.39 D512 0.11
18,250 0.02 0.03 2.98 0.08 93.21 Dice 0.08
18,000 0.01 0.05 313 0.08 98.75 85.00 0.05
17,750 0.01 0.10 0.07 97.63 72.46
17,500 0.00 0.15 0.06 100.00 62.51
17,250 0.02 0.24 0.05 95.10 48.47
17,000 0.05 0.36 0.03 87.08 35.88
16,750 0.09 0.51 0.01 TI? 24.43
16,500 0.15 0.68 0.00 63.25 16.10
16,250 0.23 0.87 0.00 50.53 10.03
16,000 0.35 1.10 0.00 36.90 5.89
15,750 0.51 1.36 0.00 24.46 Ba
15,500 0.71 1.64 0.00 14.90 1.68
15,250 0.97 1.94 0.00 8.11 0.83
15,000 bead 2:29 0.00 3 OT 0:37
14,750 1.60 2.63 0.00 1.84 0.17
14,500 1.96 3.00 0.00 0.80 0.07
14,250 2.37 3.4] 0.00 0.31 0.03
14,000 page| 3.80 0.00 0.12 0.01
13,750 3.18 4.15 0.00 0.05 0.01

620
Table 4(8.2.5) — Continued
ee ee ee
Notes

i The logarithms of the spectral sensitivities of the fundamentals at the retinal level have been
obtained from those given in Table 3(8.2.6) by adding to them the optical density of the ocular
media given as K(m) which has been taken from data given in Section 2.4.6 assuming a peak density
of 0.46 of the macular pigment at A = 460 nm.

The three functions p’(m), y’(m), B’(m) have then been normalized to have unit values at their
respective maxima.

The values of 9,(m),o,(m),og(m) are obtained from the logarithms of the fundamentals
p’(m), y’(m), B’(m) given in the previous columns by assuming that the latter represent pigment-ab-
sorption functions with peak optical densities of 0.3. Thus:

o’(m) =1—exp he co,(m)|

and similar expressions for y’(m) and B’(m), with c= 0.499 at maximum.

The table gives o,(m), o,(m), og(m) in percent of the maximum (100).

0.0 He Ae sa 3 ,

log p(m)
log B (m) wane

> -1.0 | (ee ce

2
=
(e
@
”

Oo -20 | bas
=

es)
@®
Qa
/p) AS

S ok
= -3.0 it -+—\—\4

ha Konig Fundamentals
a Estévez (i979, 1981)
oS
-40 + |

-~5.0 | L I | L | | | | |
Bhs eee oe 20 oe bak. ITalGielllS sil 4510
Wavenumber m(cm7!)
Fig. 5(8.2.5). Logarithms of relative spectral sensitivities of K6nig-type fundamentals derived
by Estevez (1979; and private communication, 1981). Nummerical values are given in Table
3(8.2.5).

621
 50

0.5

Coefficient
Absorption
Spectral

0.1

0.05

0.01

Wavenumber m (cm~!)
Fig. 6(8.2.5). Spectral absorption coefficients derived by Estévez (1979; and private communi-
cation, 1981) from his Konig-type fundamentals shown in Figure 5(8.2.5). Numerical values are
given in Table 4(8.2.5).

622
 Visual Response Functions and the Spectral Properties of Visual Pigments 623

ee 420 nm 534 "564nm

"Red" Cones
ze
0.5
- = Cones~
2 L
oO
Cc " "
Ss Blue Cones
ofa
he
|
(o)
Ww
a
Keel
®
ie
5
@ 0.05
ar

0.01
400 450 500 550 600 650
Wavelength d (nm)
Fig. 7(8.2.5). Relative spectral absorptances of human cone pigments measured by microspec-
trophotometry by Bowmaker and Dartnall (1980). For details see Table 5(8.2.5).

power distribution of the stimulus reaching the
receptors.
In the color-matching experiments, Wyszecki
and Stiles (1980) determined first the Maxwell
spectral distribution functions L(m), R(m),
G(m), B(m) for 1,000-td and 100,000-td levels
from which Maxwell color-matching functions
r(m), g(m), b(m) can be derived as follows:
ieee (ie
= ee Ry = R(m)
zit) = Se
which has a “white” appearance to the observer.
At the 100,000 td level, the values of Ry, Gy, By
are 100 times larger.
It is clear that, if the basic laws of trichromatic
color matching, particularly the proportionality
and additivity laws, hold strictly, the color-match-
ing functions 7(m), Z(m), b(m) derived from
Maxwell matches in accordance with Eq. 1(8.2.6)
must (within the experimental uncertainties) be
identical to those derived more directly by means
of the maximum-saturation method (Section
5.6.5). However, there is some experimental evi-
dence of systematic differences between the two
[1(8.2.6)] sets of color-matching functions obtained by the

nyo Pe
The observed quantity L(m) never becomes zero
[see Figure 1(8.2.6)] whereas R(m), G(m), B(m)
are respectively zero in the long, middle, and
short-wavelength region of the spectrum. The
quantities Ry, Gy, By are fixed throughout the
experiment and define the color of the reference
stimulus, which, in the case of the 1000 td level,
are given by Ry = 1.0856, Gy = 0.5567, By =
0.3516 (measured in terms of nanowatts),. and
two different methods (Section 5.6.6).
In the bleaching experiment, Wyszecki and
Stiles assume that color matching done at a given
level of retinal illuminance and with a fixed rela-
tive spectral radiant power distribution of the
“white” reference stimulus strictly obeys the
color-matching laws. With these restrictions,
the laws reduce to asserting a more limited addi-
tivity law, namely, that if any two test stimuli A
and B match individually the fixed reference
stimulus, then so does any additive mixture of A
and B in amounts pA and (1 — p)B, respectively,
where p is less than unity.
 624 THEORIES AND MODELS OF COLOR VISION

Table 5(8.2.5) Relative Spectral Absorbtances Table 5(8.2.5)— Continued

of Human Cones Measured by Bowmaker and
Dartnall (1980) The tabulated data represent mean data obtained
from several individual human cones by micro-
Wavelength “Blue” “Green” “Red” spectrophotometry applied transversely through
\ (nm) Cones Cones Cones the isolated outer segments of the cones. Three
classes of cones were identified with the following
370 0.59
mean characteristics:
80 0.67
90 0.76 Red) “Green” “Blue”
400 0.88 0.35 0.36 Cones Cones Cones
10 0.96 0.35 0.36 Peak Absorb. 562.8 533.8 420.3
20 1.00 0.34 0.33 AtASe £4nm) =3.lom a7) im
30 0.96 0.34 0.30 Transverse 0.027 0.032 0.037
40 0.86 0.35 0.29 Absorb. +0.005 +0.007 +0.011
Number
450 0.68 0.38 0.28 of Cones 19 11 3
60 0.50 0.42 0.30 Measured
70 0.36 0.49 0.34
80 0.25 0.56 0.39
90 0.18 0.67 0.47
500 O12 0.78 0.55
10 0.08 0.88 0.63 It is assumed that two different test stimuli
20 0.05 0.95 0.73 match the fixed reference stimuli if the total num-
30 0.03 0.99 0.83 ber of quanta absorbed per receptor is the same
40 0.99 0.91 for both mixtures in receptors (assumed identical)
of each of the three classes. Thus, two test stimuli
550 0.93 0.96 with external spectral radiant power distributions
60 0.82 0.99 {P,\(m) dm} and {P,(m) dm}, given in terms of
70 0.67 0.98 nanowatts, will color match the reference stimu-
80 0.53 0.93 lus of
90 0.40 0.85
600 0.29 0.74 [ F2 e(m)an(o) dm
10 0.20 0.61
20 0.14 0.47
30 0.09 0.34 = [ P22 (man(m) dm
40 0.06 0.24
[2(8.2.6)|
650 0.04 0.16
60 0.105 and two similar equations with a,(m) and a 3(™),
70 0.068 respectively, replacing a,p(m). Here ap(m),
ag(m), &p(m) are, respectively, the pigment ab-
sorption factors for receptors belonging to the
three classes, and 7(m) is the spectral transmit-
The Maxwell method of determining color- tance of the ocular media accounting for light
‘matching functions is preferred over the maxi- losses that occur in the eye before the stimuli
mum-saturation method because it eliminates reach the receptors.
effects that may arise when, for different test By the characteristic property of Maxwell
wavelengths, simultaneous changes are made in color-matching functions, the two test stimuli also
both the composition of the test stimulus and the satisfy the equations
reference stimulus as is done in the maximum-
saturation method. Also it allows the stimulus J Pim)r(m) dm = JPa(m)r(m) dm
level to which the color-matching functions apply
to be uniquely specified. [3(8.2.6)]
 Visual Response Functions and the Spectral Properties of Visual Pigments 625

ep) WAVELENGTH (nm)

2 400 450 500 950 600 650 700
a
gies B 7 R 250
ve
- poe
Tso 1000 Trolands
oO
oO 0-0-0 |QQO00 Trolands
fe)
ac
O 2.0 200
Y
2
BS / Ulm)
2<Ikb Ln
L(A) 0
za

z
op)
uJ
=>
aN)

J io 100
4
2a
WJ
>
x<

5S 0.5 —=_|_
TROLANDS
lOO000
FOR
NANOWATTS
IN
VALUES
MAXWELL

25 24 23 22 21 20 i918 i615 14-103

WAVENUMBER Loh:
Fig. 1(8.2.6). Maxwell spectral distribution functions L(m), R(m),G(m), B(m) obtained by
observer GW in a 2° bipartite visual field for two levels of retinal illuminance: solid lines refer
to data obtained at 1000 td, thin lines through open circles refer to data obtained at 100,000 td.
A “white” (D,.;) reference stimulus was used whose spectral composition consisted of the three
monochromatic primary stimuli used in the experiment, with radiant powers R ,, = 1.0856,
G y = 0.5567, By, = 0.3516 nW when the retinal illuminance is set at 1,000 td, and with radiant
powers 100 times larger when the retinal illuminance is set at 100,000 td (for further details, see
Section 5.6.5).

OT pigment— the following reasoning justifies the as-

sumption made.

J [Erm
(m)7(m) [2] i Suppose, in a retina adapted to darkness, the
m 0 7(m) concentration of the completely unbleached vis-
ual pigment in the outer segment varies along the

Tiered
ese
and two similar equations with g(m) and b(m),
respectively, in place of r(m).
In Eg. 2(8.2.6), it is assumed that the absorp-
tion factors ap(m), a¢(m), a_(m) do not depend
on the test stimuli {P,(m) dm} and {P,(m) dm}
that match the reference stimulus. For the simple
model in which the outer segment of the cone
receptor is taken to contain only one absorbing
substance—the “red,” “green,” or “blue” visual
segment according to a function of the distance s
from the junction with the inner segment. This
function may be designated cro(s) in the “red”
receptors, where s lies between zero and S, the
length of the outer segment. If, in the steady state,
a test stimulus {P,(m) dm} color matches the
reference stimulus, the concentration of the un-
bleached pigment in the “red” outer segments of
the test area will be reduced to a different func-
tion Cr(s). For another test stimulus {P,(m) dm}
that also color matches the reference stimulus, the
corresponding concentration function cr>(s) will
626 THEORIES AND MODELS OF COLOR VISION

not in general be the same as Cp,(s). The absorp-
tion factors for the “red” receptors in the two
cases will be, respectively,
19~Kee(m) fcri(s) ds
Api(m) =l]-
[4(8.2.6)]
and
=|]- 10 kxo(m) f’caa(s) ds
Ap2(m)
[5(8.2.6)]
where kp is a constant independent of wavenum-
ber, and p(m) is the spectral absorption coeffi-
cient of the “red” visual pigment, normalized to
unity at the wavenumber m(R) where the coeffi-
cient has its greatest value.
The total number of quanta of all wavenum-
bers absorbed per unit time in the outer segment
must equal the rate of restoration of complete
(unbleached) molecules (Section 8.2.2), which the
work of Rushton (1972) has shown to be propor-
tional to the total number of bleached pigment
molecules in the receptor; that is, the number of
protein moieties of originally unbleached pig-
ment, which may be thought of as anchored in the
structure of the lamellae of the outer segment.
Thus, for the test stimulus {P,(m) dm} we have
f Pre am) ani
= const J Lenols) teas) ae
[6(8.2.6)]
A similar equation holds for the stimulus
{P,(m) dm}, and since, at color match, the ex-
pressions on the left are equal for the two equa-
tions, the same is true for the expressions on the
right. Therefore,
Jens) ds = [en2ls) ds =C
In actual receptors, the situation is com-
plicated by the possible presence in appreciable
amounts of absorbing substances other than the
visual pigment, and the effect of these on the
visual pigment absorption factor for test stimuli
color matching the same reference stimulus may
be different. However, the assumption that the
function ap(m) is independent of the relative
spectral radiant power distribution of the test
stimulus is adopted as a working approximation.
This is consistent with the assumed validity of the
limited additivity law for Maxwell matching (Sec-
tion 5.3.6).
Equations 2(8.2.6) and 3(8.2.6) then imply that
the absorption factors ap(m), ag(m), ap(m) are
linear combinations of the Maxwell color-match-
ing functions for the fixed reference stimulus when
these functions are expressed in quantum units
and corrected for prereceptor eye losses. It is
convenient for later comparisons to define what
may be called the “pigment fundamentals”
fr(m), fo(m), fg(m) as the spectral absorption
factors normalized to unity at their maximum
values at wavenumbers m(R), m(G), m(B), re-
spectively. Thus,
1 — 107 Pre(™)
fa(m) = 1— 10->
= ny Leut(m) + c12e(m) + cub (™)
1 — 107 Pav™)
= my Lea?(m) + ¢y)8(m) + cy3b(m)]
1 — 10 280(™)
fp(m) = — 1072s
= a [¢3,7(m)
C30 2(m) ts €336(m)|
[8(8.2.6)]
where

Dr = k Cp; Dg = kgCo; Dz = kpCz

[7(8.2.6)]
[9(8.2.6)]
It follows from the above formulae for a,,(m)
and a,p>(m) that these reduce to the same func- are the total visual pigment densities of the outer
tion ap(m) applicable to any test stimulus that segments at their maximal values in the receptors
color matches the reference stimulus. A similar of the three classes.
argument defines C, and C, for “green” and The pigment densities Dp, Dg, Dz, the trans-
“blue” receptors. formation coefficients c;,, the observed color-
 Visual Response Functions and the Spectral Properties of Visual Pigments 627

matching functions 7(m), 2(m), b(m), and the media; that is,
pigment fundamentals f,(m), f,(m), f,(m), all
depend, in general, to some degree on the refer-
ence stimulus, and, in particular, vary as the level
of the reference stimulus of given relative spectral
fea) Soe Milam)m

radiant power distribution is raised from mod-

erate to high values. The absorption-coefficient
fou(m) =
t(m) fou(m) [12(8.2.6)]
functions, on the other hand, given by m

p(m) = ~- log[1 ~ (1 ~ 10-?*) fam]

Fe) et) m

y(m) = ~ slog[1— (1 ~ 10°) f-(m)] _ In the second step, we transform feu(m),

fou(™), fey(m) to predicted color-matching
B(m) = ~ log [1 —(1 ~ 10-*)fa(m)] functions 7,,(m), &);(m), b,,(m) normalized at the
primaries such that at wavenumber mp = 15,500
[10(8.2.6)] em 2m — 19,000 cm mn, = 22;500 cm! b

these functions obtain values of

represent invariable intrinsic properties of the
three visual pigments.
With known absorption coefficients p(m), Fy(mp)=1 Fy(mg)=0 — Fy (mg) = 0
y(m), B(m) we can calculate the pigment funda-
mentals fp(m), f;(m), f,(m) for different pig- Bu(mr)=90 By(me)=1 — &y(mz) = 0
ment densities Dz, Dz, Dg. In particular, we can
try to predict the pigment fundamentals when by(mr)=0 by(mg)=0 by (mz) =1
bleaching one or more of the photopigments has
occurred as hypothesized in the high-level color- [13(8.2.6)]
matching experiments.
If the pigment densities Day, DoH, Dgy at the This is readily accomplished by making the trans-
high level (symbolized by the subscript H) of formation matrix {b;,} of
retinal illuminance (100,000 td) were known, we
could determine fp;(™), fou(™), fay(m) from
?y(m) = bi fan(m) + Pifen(m) + bisfen(m)
1 — 10 Prxe(™)
fru(m) = 1 — 10~Drx 81(m) = by fan(m) + bofou(m) + bos fan(m)
ve lee 107 Pexy(™)
By (m) = bs fen (m) + bs fou(m) + bssfan(m)
La Sl a enor epee tras [14(8.2.6)]
— 1 = 107 PanBor)
LOANS Spa
meneoe equal to the inverse of the matrix

[11(8.2.6)] feu(mp) fau(mc) fru(ms)

fou(mr) fou(mg) fou(ms)
By reversing the procedure used above, the
pigment fundamentals fpy(™), fey(™), fex(™) fou(mp) fou (mo) fan (ma)
at the high level are readily transformed to color-
matching functions and Maxwell spectral distri- _ The Maxwell spectral distribution functions
bution functions. L,,(m), Ry(m), Gy(m), By(m) are then ob-
In the first step, we revert to pigment funda- tained from the predicted color-matching func-
mentals expressed in energy units and modulated tions 7,,(m), ,(m), b,(m) by means of the
by the spectral transmittance t(m) of the ocular reverse transformation of Eq. 1(8.2.6) Three sub-
628 THEORIES AND MODELS OF COLOR VISION

sets of transformations are required: When inserted in Eqs. 1(8.2.6) to 17(8.2.6), the
appropriate values of the above parameters are
(a) Spectral region within which test stimulus L then expected to transform the experimentally
replaces red primary R [i.e., R;,(m) = 0}: determined Maxwell spectral distribution L(m),
we R
R(m), G(m), B(m), characterizing color matches
Ly(m) 2 ee with a given “white” reference stimulus at a reti-
F,(m) nal illuminance of 1,000 td, to new Maxwell spec-
tral distribution functions L,(m),
Gy(m) suture Ly(m)&y(m) R,,(m), Gj(m), By(m) which agree with the ex-
perimentally determined functions
By,(m) = Bywais Ly(m)by(m) Li(m), Ry(m), Gy(m), By(m)_ characterizing
[15(8.2.6)] color matches with the same given “white” refer-
ence stimulus but at a retinal illuminance of
(b) Spectral region within which test stimulus L 100,000 td.
replaces green primary G [i.e., G,(m) = 0]:
Because of the nature of the experiment and
its inherent uncertainties in both the observations
and instrumental calibrations, one cannot expect
Linkre) racy to obtain perfect agreement between predicted
and observed Maxwell spectral distribution func-
R,(m) Nae Ly(m)F,(m) tions. However, one should expect to find at least
one set of parameters that yields an agreement
B,,(m) = Beit Ly(m)by(m) within the range of the experimental uncertain-
[16(8.2.6)] ties. Such a set might be called a “best” set of
parameters.
(c) Spectral region within which test stimulus L Another important requirement is to constrain
replaces blue primary B [i.e., B,,(m) = 0]: the parameters within certain “acceptable” ranges.
The transformation coefficients c;, should pro-
Bwr vide relative spectral distributions of the pigment
Ly(m) i fundamentals which resemble those expected from
bz(m)
other studies, such as the K6nig-type fundamen-
tals. The optical densities Dr, Do, Dz of the
R,(m) — Rinne Lj (m)F_y(m)
unbleached photopigments should fall within a
reasonable range of values, somewhere between
G(m) Oy ie Ly(m)&,(m)
0.2 and 1.0. The spectral transmittances t(m) of
[17(8.2.6)] the ocular media should agree reasonably well
with those found in normal eyes.
If the hypothesis holds that bleaching of the There are different procedures one might fol-
photopigments accounts for the breakdowns in low to obtain an acceptable best set of parame-
the color matches observed at 100,000 td, it must ters, and Wyszecki and Stiles (1980) explored
be possible to assign appropriate values to the several possibilities. The most successful turned
following parameters of the pigment-bleaching out to be an iterative procedure coupled with
model as outlined above: “trial and error” calculations guided by a consid-
erable amount of human judgment. The “best”
(i) The transformation coefficients c;, in Eq. set of parameters ultimately emerging was the
8(8.2.6) which convert the 1000-td-level color- following:
matching functions r(m), g(m), b(m) to the rela-
tive spectral distributions fp(m), f,(m), fg(m) of (a) Coefficients c;,:
the pigment fundamentals.
(ii) The pigment densities Dp, D5, Dg at the [35710 = 399558 10>? 3.008 OC10
1,000-td level and Dey, Dox, Day at the 100,000-
td level.
(1) The spectral transmittance t(m) of the ocu- 2.098 X 10~° 3.954 x 10-5 4.035 x 10°
lar media. [For details, see Eq. 18(8.2.6), intro-
duced later.] TIT 0 1.813 x 10~° 1148 10m?
 \
/
Calm)
(m)
Ly 83
8 WN \

et yyy |v
= LT. 4
8 ii Y
oO io@

bm)
Ry,
/
Ry(m)
~

0.92 Z
0.90 ‘ ' ;

Be 5 G Gh
VOOM
QA R

oot WY YY UY Y
1.00 LL LIL OOD

0.98 GY a ae ‘YY
~/
Gylm)
(m)
Gy 0.96 Ui YY Y

SASK
SS

\
By,
/(m)
By

Rw W
st Poa 24 AS 128s 228 Aled.
MY Yu J
20. HI wlB PIT Jere 15) 14.107
WAVENUMBER m (cm”')
Fig. 2(8.2.6). Ratios of predicted over observed 100,000-td Maxwell values as function of
wavenumber m. Hatched area represents uncertainty range of ratios when assuming +2%
errors in the observed 1000-td Maxwell matches and +4% in the observed 100,000-td Maxwell
matches.

629
630 THEORIES AND MODELS OF COLOR VISION

(b) Pigment densities: The discrepancies are considered small and

most likely can be attributed to the uncertainties
Dp = 0.44 Dg = 0.38 D, = 0.45 inherent in the observations and instrumental
Daxe=00l DS O08 Day = 0.44 calibrations. Upper and lower bounds of the in-
herent uncertainties were estimated that provided
(c) Spectral transmittance 7(m) of ocular media: uncertainty ranges (hatched areas) for the ratios
The inert pigments in the eye’s lens and macula plotted in Figure 2(8.2.6). Except for a few iso-
are initially assumed to have spectral transmit- lated wavenumbers, the ratios of predicted over
tance 7,(m) and t,,(m), respectively, in accor- observed 100,000 td Maxwell values generally fall
dance with data usually applied to the normal eye well inside the estimated uncertainty range. This
(Section 2.4.6). However, as the densities of these indicates that the “best” set of parameters, which
pigments may differ from individual to individ- governs the bleaching model, predicts 100,000-td
ual, the parameters 6 and e were introduced to Maxwell matches within the uncertainty of the
allow for such density difference between the experimental data. However, what remains to be
normal eye and the observer used in the experi- considered is whether the parameters also meet
ment. Thus, the spectral transmittances t(m) of the requirements of being “‘acceptable” from other
the ocular media were considered to be composed points of view, as discussed above. Of particular
of spectral transmittances 7°(m) of the lens pig- interest are the pigment fundamentals that the
ment and spectral transmittances t;,(m) of the bleaching model yields with its “best” parame-
macular pigment; that is, ters.
The pigment fundamentals yielded by the
t(m) = t(m) - 1%,(m) [18(8.2.6)] bleaching model and given in Table 1(8.2.6) are
specifically defined by envisaging a single visual
The best case required pigment in each receptor. Other sets of “funda-
mentals” can be derived from color-matching data
6=1.00 and e=0.80 on the basis of K6nig’s hypothesis that the three
types of dichromat (protanopes, deuteranopes,
that is, normal density of the lens pigment, but
tritanopes) possess reduced forms of trichromatic
only 80% of the normal density of the macular
vision, but each lacking one—a different one—of
pigment.
three normal visual response systems (Section
The agreement between predicted and ob-
8.2.5). The spectral sensitivities of the. distinct
served 100,000 td Maxwell spectral distribution
response systems, the fundamentals, are derivable
functions was found to be quite good when al-
from the color-matching functions of the normal
lowances were made for the inherent uncertainties
trichromat together with the locations in the tr-
in the observational data both at the 1000-td and
chromatic chromaticity diagram of the confusion
the 100,000-td levels of retinal illuminance. In
points of the three types of dichromat.
Figure 2(8.2.6) are plotted the ratios of the pre-
A further possibility is that the spectral sensi-
dicted over the observed Maxwell spectral distri-
tivities of the 7-mechanisms (Section 7.4.3) might
bution functions, that is,
be identified as pigment fundamentals. These
sensitivities were determined by methods not in-
Ly(m) Rj(m) volving color matching and aim to represent the
Ly(m)° —— -Ry(m) spectral sensitivities of distinct retinal response
systems, each of which might involve more than
one visual pigment. Nevertheless, the field sensi-
Gy(m) By(m) tivities of 75, 7,, and 7, have been found to be
Gy(m)° By (m) expressible approximately as linear combinations
of the color-matching functions (Section 7.4.4).
Perfect agreement between the predicted and ob- Figure 3(8.2.6) compares the fundamentals de-
served data would result in ratios of 1.00 at all rived from the bleaching model with those based
wavenumbers. This, of course, does not occur, on the 7-mechanisms and those of the Kénig type
except at the wavenumbers of the primary stimuli suggested by Vos and Walraven (see Walraven,
R,G, B. However, at these wavenumbers the 1974; Vos, 1978; and Section 8.2.5) and by Smith
agreement has been forced to be perfect through et al. (1976). Only relative spectral distributions
the required normalization of the data. of the fundamentals are shown; that is, all have
Table 1(8.2.6) Relative Spectral Absorption Factors fr(m), fg(m), fg(m)
of Pigment Fundamentals as Predicted by Bleaching Hypothesis
[See Also Figure 3(8.2.5)] (Wyszecki and Stiles, 1980)¢
a a ae
Wavenumber,
m(cm_') fR(m) fg(m) fp(™m)

14,500 0.0163 0.00187 0.294 1078

14,750 0.0327 0.00383 02110
15,000 0.0665 0.00823 0.960104
15,250 0.125 0.0167 0.430 1073
15,500 D244 0.0327 0.00121
15,750 0.332 0.0630 0.00329

16,000 0.495 0.117 0.00694

16,250 0.651 0.186 0.0113
16,500 0.784 0.276 0.0154
16,750 0.866 0.372 0.0184

17,000 0.932 0.485 0.0214

17,250 0.977 0.603 0.0244
17,500 0.991 0.715 0.0274
17,750 1.000 0.825 0.0313

18,000 0.996 0.910 0.0349

18,250 0.986 0.980 0.0386
18,500 0.934 1.000 0.0408
18,750 0.859 0.990 0.0422

19,000 0.790 0.970 0.0445

19,250 0735 0.955 0.0491
19,500 0.654 0.903 0.0567
19,750 0.597 0.880 0.0708

20,000 0.545 0.861 0.101

20,250 0.474 0.838 0.152
20,500 0.462 0.814 0.230
20,750 0.474 0.796 0.332

21,000 0.463 O35 0.446

21,250 0.451 0.689 0.578
21,500 0.404 0.605 0.699
21,790 0.364 0.534 0.834

22,000 0.331 0.472 0.923

22,250 0.297 0.410 0.971
22,500 0.262 0.351 1.000
22,750 0.217 0.284 0.980

23,000 0.192 0.243 0.973

23,250 0.172 0.208 0.902
23,500 0.162 0.180 0.805
23,750 0.157 0.161 0.677

24,000 0.124 OAL 0.486

24,250 0.0980 0.0872 0.353
24,500 0.0542 0.0459 0.175

“The three pigment fundamentals are normalized to unity at their respec-

tive peaks, are given in quantum units, and are corrected for light losses in
the ocular media.
632
| Uses9
| S|OjueWOPUNY

pey

—enig
eeeececces

-
A
o-oo
—

SOA
i Yyiws
(S)°10
4@

('iL)
sejtas

,8|Ojuewopuns
,S|Ojuewopuny

( $y1un
f= of
— Buiyr0e\g siseysodAy
Buiys0e/g siseyjodAy

USAD4/|D
0) @A1jouUse4)
(

wnyuoNnb
u )
--------- ol USADIJDM-SOA
cofeXe) —oo—o S®1NS (*1L)
er —— YIWS
38 (W)°10
—e—e—— Buiysoe|g siseyjodAy

Ayaiyisues
SOA

om se1NS (8)
KK $8
YIWS 7°10 (1)
USADIIOM

—e——— Buiys09|g sisaysodAy

|04j9edS
B@AILJO/
ey
000 100°0
G2 ve €2 22 i2 O2 6l 8I ZI 9I eOl:si ae ve 2 22 Ie Oe 6I Bi 2! Oi eOl:si G2 2 £2 22 12
61
91
2!

si

O02

JEQWNUSADM
W (wo) sequinueAD@M
W (,-w2)
w

“SIA R)E “(97

aayepy [eyjoods SUOKNGLYsIP
JO .‘pal,, ,“Udeds,,
pue ,anjq,, ‘s[ejuauIepuny
(,-wo)

dAmsD ysnompros :sjop yuoursid sjeyuourepuny

‘(a)*f
°F (1) *f (iw) sv payoipaid
Aq oy)
Suryovaiq ‘sisayjodAYysAINZD YsnoIy)
usdo :sapidPPT} SAIAYISUdS
JO SIJVG SUISIUBYIOUI-L
sequinusADmM

) ‘Sw‘Puan *( dAinsz Ysnory) :sasso1d sdAj-stuOY sjeyuawepuny

pasodoad
Aq Wg 19 ‘Te ‘(OL61)
poyogsaul] od4j-stUOYy s[ejusuepuny
pasodoid
Aq soa pue UaavleA,
Ay) “Ueavape “(PL6T
IV
sUOTOUNY
aie pazIpeULIOU
0} AjIUN
ye Ley) aayoodsa1
‘syvad
oe uaars
ul umjuenb‘sjtunpue ore
Pd}II.1109
10} JUST] SASSO] UI JY} IV[NIO “BIPIUI

been normalized to have unit values at the wave-
numbers at which their respective peaks occur.
All functions are given in terms of quantum units
and corrected for light losses in the ocular media.
In the case of the bleaching model, t(m) is de-
fined by Eq. 18(8.2.6) with 6 = 1.00 and e = 0.80.
In the case of the 7-mechanisms and the Vos-—
Walraven fundamentals, it has been assumed that
6 = e = 1.0, and in the case of the Smith et al.
fundamentals, 6 = 1.33 and e = 0.50.
In a qualitative sense, there is reasonable
agreement between the various fundamentals de-
rived from the “red”-receptor mechanism and an
equally reasonable agreement between those de-
rived for the “green’”-receptor mechanism. The
prediction of the “red” and “green” funda-
mentals made by the bleaching model fall com-
fortably in the domain of the corresponding
fundamentals derived by the other investigators.
A great deal of the somewhat larger spread be-
tween corresponding fundamentals in the short-
wavelength region of the spectrum can most
certainly be attributed to differences in the densi-
ties of the ocular media assumed to apply to the
different data. However, to assess the significance
of these differences and those of a more subtle
nature in quantitative terms is difficult as we lack
knowledge of the true fundamentals.
A somewhat less reasonable agreement is noted
between the “blue” fundamental predicted by
the bleaching model and those of the other in-
vestigators. The case of the blue fundamentals is
intriguing because the prediction made by the
bleaching model demands a pronounced sec-
ondary peak in the long-wavelength region of the
spectrum. There is a great deal of flexibility in the
shape of this secondary peak, which is governed
by the coefficients c,,; and c3, in the transforma-
tion of Eq. 8(8.2.6). The upper (solid) curve corre-
sponds to the “best” case. The lower (dashed)
curve corresponds to coefficients c,, = 6.084 x
10°® and c3, = 3.012 X 10~’. These new coeffi-
cients, with all others kept unchanged, have only
a marginal effect on the predicted values of the
100,000-td-level Maxwell spectral distribution
functions, L,(m), Ry(m), Gy(m), By(m).
However, they have a drastic effect on the long-
wavelength part of the “blue” fundamental. There
is an indication (Section 7.4.3) that the 7,-mecha-
nism also has a secondary peak in that spectral
region, but it is much less pronounced and is only
barely caught in Figure 3(8.2.6). The good agree-
ment between the 7,-mechanism and the Smith et
al. “blue” fundamental is not surprising because
Neural Models 633
Smith et al. chose the 7, mechanism as their
model for the “blue” fundamental.
With regard to the densities of the photopig-
ments, we believe that the estimates made by the
bleaching model for the “red” and “green” pig-
ments fall within expected ranges. The estimate
for the “blue” pigment, however, is virtually un-
certain because that pigment did not seem to
bleach by any measurable amount in the experi-
ment. It follows that the given densities of Dz =
0.45 and Dz = 0.44 could well be given any
other values, provided that the near unity ratio of
Dpgy/Dp is kept approximately the same. For
example, the use of Dz = 0.100 and Dz;, = 0.098
in the “best” set of parameters gives essentially
identical results.
The reason for the somewhat unexpected be-
havior of the “blue” pigment is not clear. At this
time, we can only speculate. We suggest that the
“blue” pigment may be in a near diluted state
already at moderate levels of retinal illuminance,
making it difficult to detect its bleaching at higher
levels.
8.3. NEURAL MODELS
For several decades, considerable experimental
evidence has been gathered in support of models
of color vision in which the idea of trichromacy at
the initial receptor stage is combined with schemes
of coding the receptor signals into two signals
carrying chromatic information and one signal
carrying achromatic information. Numerous psy-
chophysical as well as electrophysiological experi-
ments have provided the basis for such models.
However, though considerable progress has no
doubt been made, none of the present models can
claim to make completely satisfactory predictions
of all or most of the known color-vision phenom-
ena. Refinements in the experimental techniques
and research strategy of psychophysicists and
electrophysiologists are needed to unravel the
complex neural network of the visual system and
its many processes.
We have discussed some of the psychophysical
evidence supporting neural models in previous
sections, notably Section 5.12 on chromatic adap-
tation and Section 5.13 on chromatic-response
functions. We have also described in Section 2.2,
though rather briefly, the structure of the human
eye, with some particular attention to the fine
structure of the retina. However, it is beyond the
634 THEORIES AND MODELS OF COLOR VISION
scope of this book to offer a survey of the neuro-
physiology of the visual system, though we must
admit that many models of color vision must take
account of the system’s neurophysiological prop-
erties. To the interested reader we recommend the
informative review given by Lennie (1980) on
visual pathways, which discusses the present state
of knowledge concerning the properties of a
variety of different neurons and cell types found
primarily in the cat and the macaque monkey.
The visual organization of the macaque monkey is
of particular relevance to human vision. Lennie’s
review contains many references to the original
literature in this field. The reader will also find
the review on the encoding of color in Chapter 7
of Boynton’s (1979) book most informative and
up to date.
8.3.1 Miller and Judd
The now widely accepted concept that color vi-
sion is best explained by a zone theory (Section
8.1) dates back more than 50 years with Miiller
(1930) playing a prominent role in its develop-
ment. Judd’s (1949, 1951b) formulation of the
concept provided the insight needed to develop
mathematical models that describe the processes
postulated to take place in the various zones (or
stages) of the visual system. In particular, Judd
showed how the signals of the channels of each
stage might be related to the CIE 1931
(X, Y, Z)-tristimulus values of the color stimulus;
the signals being derivable by means of linear and
simple nonlinear transformations of the
(X, Y, Z)-values.
In Miiller’s theory, the visual system is pos-
tulated to consist of three processing stages, which,
using Judd’s (1949) terminology, are identified as
(1) an initial photochemical stage, (ii) an inter-
mediate chemical stage relating to the chromatic
aspect, and (iii) a final stage of excitations of the
optic nerve fibers. Judd’s (1949) mathematical
model of Miller’s theory incorporates the color-
matching properties of normal trichromats, pro-
tanopes, and deuteranopes, experimental data
concerning the stimulus perceived as being achro-
matic, and the stimulus perceived as unique
yellow (i.e., neither red nor green). The model
expresses in quantitative terms the qualitative
ideas of Miller; in particular, it shows how the
trichromatic theory of Young, Helmholtz, and
K6nig (Miller’s first stage) and the opponent-col-
ors theory of Hering (Miller’s third stage) may
both be accepted, and the explaining power of
both be simultaneously utilized. Miiller’s inter-
mediate (second) stage is also found to be a useful
theoretical tool.
With somewhat different notation and ter-
minology, Judd’s model of Miiller’s theory may
be summarized as follows:
(i) First Stage (Cone-Receptor Stage). Three
types of cone are postulated, each containing a
photopigment with its own characteristic spectral
absorption properties. The spectral response func-
tion of each type of cone can be expressed as a
linear combination of the color-matching func-
tions of a normal trichromat; the normal trichro-
mat being represented by the CIE 1931 standard
colorimetric observer. Thus,
DMA)= _ 3.1956%(A)+2.44787(A) — 0.64342(A)
po(A) =—2.5455x(A) +7.0492 p(A) + 0.49632(A)
p3(A) = 5.0000Z(A)
[1(8.3.1)]
where p,(A), p2(A), p3(A) are the spectral re-
sponse functions of the three types of cone and
x(A), y(A), Z(A) are the color-matching functions
of the CIE 1931 standard observer, applicable to
monochromatic stimuli of wavelength A and con-
stant radiance (expressed in energy units) arriving
at the cornea (external stimuli). For color stimuli
of complex spectral radiant power distributions
{S, dA}, the basic laws of trichromatic color
matching are assumed to hold, leading to CIE
tristimulus values
X=k[S,x(A) dd
Y=kf{S,y(A) dA [2(8.3.1)]
Z =k S2(A)dd
where k is a normalizing constant (for example,
k = 683 Im- W_'). The CIE tristimulus values
X, Y, Z convert to tristimulus cone responses by
means of a linear transformation identical to Eq.
1(8.3.1), that is,
P, = 3.1956X + 2.4478Y — 0.6434Z
Py = —2.5455X + 7.0492Y + 0.4963Z
jee 5.0000Z
[3(8.3.1)]

The coefficients of the transformation matrix
are determined by a procedure similar to that
employed in the derivation of Kénig-type funda-
mentals (Section 8.2.5). This includes, among
other constraints, the following chromaticity co-
ordinates for the dichromatic confusion points:
WORST, onli 0-253
Myr= 1.0008 yp = 10.000
5 (O50 a eaa).000
(ii) Second Stage (Intermediate Coding of Cone
Signals).
[yellowish red (yR)—bluish green (bG)]-signal:
a (A) = p(A) — pr(A)
[greenish yellow (gY)—reddish blue (rB)]-signal:
a,(A) = 0.0151p,(A)
—0.3849p,(A)
—0.4000p,(A) [4(8.3.1)]
The spectral response functions of the cones are
denoted by p,(A), p>(A), p3(A) and are defined
by Eq. 1(8.3.1). These cone responses are con-
verted to two difference signals denoted by a,(A)
and a,(A), respectively. The coefficients in the
above equations for a,(A) and a,(A) are de-
termined, in part, from the chromaticity coordi-
nates of the dichromatic confusion points (see
above), the achromatic chromaticity point
(equal-energy stimulus), and the chromaticity
point of unique yellow (578.1 nm).
Equations 4(8.3.1) can be expressed in terms
of the color-matching functions by making use of
Eq. 1(8.3.1), that is,
a,(A) = 5.741x(A) — 4.601y(A) — 1.1402(A)
a (A) = —0.932x(A) + 2.750 (A) = 1.8192(A)
[5(8.3.1)]
(iii) Third Stage (Final Neural Coding). Six
neural signals are postulated, eliciting the sensa-
tions “white” (w), “black” (schwarz, s), “red”
(r), “green” (g), “yellow” (y), and “blue” (b).
The signals eliciting the chromatic sensations are
Neural Models 635
generated in the form of two opponent signals:
[red (r)-green (g¢)]-signal:
B(A) = a,(A) — 0.6265a,(A)
[yellow ( y )—blue (b)]-signal:
B(A) = a,(A) + 0.1622a,(A) [6(8.3.1)]
Note that the 6,(A) signal is a difference signal
generated from the two second stage difference
signals a,(A) and a,(A), with a,(A) contributing
in a major way and a,(A) in a minor way (factor
0.6265). Similarly, the B,(A) signal is a difference
signal with a,(A) being the major and a,(A) the
minor (factor 0.1622) contributor.
The signal eliciting the achromatic sensation is
generated in a major way by the signals p,(A),
P(A), p3(A) emerging from the first stage to
which are added minor contributions from the
signals emerging from the second stage:
[white (w )—black(s )]-signal:
B3(A) = 0.0075p,(A) + 0.:1912p,(A)
+0.0013p3(A) + 0.0810a,(A)
+0.0024a,(2) [7(8.3.1)]
The sensation of black is considered to come
chiefly by induction from a white surrounding
stimulus or from a white pre-exposure stimulus.
Equations 6(8.3.1) and 7(8.3.1) defining the
signals emerging from the third and final stage of
the neural network can also be expressed in terms
of the color-matching functions by making use of
the previous Eqs. 1(8.3.1) and 5(8.3.1); that is,
B (A) = 6.325[x(A) — y)]
B,(X) = 2.004[ y(A) — z(A)]
B3(X) = y(A) [8(8.3.1.)]
Equation 8(8.3.1) reveals the simple relation pos-
tulated to exist between the final neural signals
and the CIE 1931 color-matching functions. This
simplicity arises in part from the fact that the
X-primary of the CIE color-matching system cor-
responds to a (hypothetical) stimulus assumed to
elicit “unique red” and the Z-primary corre-
sponds to “unique blue.”
636 THEORIES AND MODELS OF COLOR VISION
The form of Eq. 8(8.3.1) resembles closely the
system derived by Schrodinger (1925) and also
that derived by Adams (1923) (see below). It
satisfies the principles of Hering’s opponent-col-
ors theory (Section 8.1).
Miiller’s zone theory explains protanopic vi-
sion by the failure of the [(yR) — (bG)]-channel
[a,(A) = 0] in the second stage, calling the
phenomenon “outer red-green blindness.”
Deuteranopic vision is explained by the failure of
the [(7) — (g)]-channel [8,(A) = 0] in the third
stage, called “inner red-green blindness.” Tri-
tanopic vision, like protanopic vision, is consid-
ered a retinal defect, corresponding to the failure
of the [(gY) — (rB)]-channel [a,(A) = 0] in the
second stage, and is called “outer yellow-blue
blindness.”
The luminous efficiency function Vp(A) of the
deuteranope is identical to B;(A) = Vy(A) of the
normal trichromat, that is,
Vp(A) = Vy (A) = yA) [9(8.3.1)]
The luminous efficiency function V,(A) of the
protanope is defined by
Vp(A) = 0.0075p,(A) + 0.1921p,(A)
+0.0003p;(A) [10(8.3.1)]
which is derived from B,(A) [Eq. 7(8.3.1)] by
setting a,(A) = 0. In terms of the CIE color-
matching functions, Vp(A) becomes
Vp(A) = —0.4650x%(A) + 1.3725p(X)
+0.09202(A) [11(8.3.1)]
With regard to the tritanopic luminous ef-
ficiency function V;(A), Miller’s zone theory re-
quires that this function also deviates from that of
the normal trichromat; it is given by
V-(A) = 0.0075p,(A) + 0.1912 p,(A)
+0.0013p93;(A) + 0.0810e,(A)
[12(8.3.1)]
OT
V-(A) = 0.0023%(A) + 0.9935y(A)
+0.0043z(A) [13(8.3.1)]
Table 1(8.3.1) gives the numerical values obtained
by Judd (1949) for the spectral response functions
which, according to him, satisfy the principles of
the Miiller zone theory expressed in terms of the
above equations. Figure 1(8.3.1) illustrates the
spectral response functions characterizing normal
color vision.
Judd (1949) tested his model of the Miller
zone theory by predicting the chromatic-threshold
data of Abney (1910) and Priest and Brickwedde
(1938) (see also Section 7.10.3). According to the
Miller theory, the ability of an observer to detect
a slight variation in chromaticity from a chro-
maticity of a given “white” stimulus depends on
the excitation of the chromatic sensory processes
(second stage). Good agreement between predict-
ion and observation was obtained when the a,(A)
and a,(A) signals, defined by Eq. 4(8.3.1) were
combined to give
he kVy(A)
Sa
[a?(A) + £7a3(A)]'7
[14(8.3.1)]
where V,,(A) is the luminance signal generated in
the third stage and & and f are constant factors to
account for the particular units used and to scale
the relative effectiveness of the a,(A) process
compared to the a,(A) process.
In a later study, Judd and Yonemura (1970)
show that a close relationship exists between the
second stage of the Miller theory and the ap-
proximately uniform chromaticity spacing ex-
hibited by the CIE 1960 (u, v)-UCS diagram (see
Section 3.3.9). By considering normal color vision
as a combination of protanopia and tritanopia as
suggested by the second-stage processes, Judd
and Yonemura develop a more general measure
of the perceived size of a chromaticity difference |
than the commonly used length of the line con-
necting the two chromaticity points. The general
measure is the square root of the sum of the
squares of the angles subtended by the line at the
convergence points of the chromaticity confusion
lines by protanopia and tritanopia, respectively.
By this measure, the chromatic-threshold sensitiv-
ity to wavelength changes of monochromatic
stimuli is accounted for quantitatively not only
for protanopic and tritanopic vision, but also for
normal vision, including the secondary maximum
of wavelength discrimination in the neighborhood
of 420 nm (see Section 7.10.2).
 wey VCO'O CLIO 7 i We 5Bokgn mm [Fd low 90'0 + Dol laa v600 098'1 OV

637
$970 8900 $970 Soe seresur Loto er Ore 69P' C+ ECC "ee CUDIG Of
O8€'0 6c1 0 I8€0 Lee 9 fee te YOO CScUs cst e+ 1000. OLSO = £99y 6
cOS'0 877 0 €0S°0 8 HRS 0 We oe Sb oa 6br 0+ Ipye + cOO0 660 StrY Ol
6790 9LE0 1€9°0 RASBg 2fia oe eh i aa byl O+ VOR cm po0o0 SrlLl 867 009
velo ISO 8 LSL'0 ESO" Ser lS tee rOL tar BGI T a 80h C+ O00 Us” VOL G Mrccits 06
oer OLLO "OLs0 UL8) VET t6c 07 DUSAL ST Nee 8000 108 LS0S 08
8760 €66 0 eco ; -cs0U suo t+. lCh— v06°1 + 800'0— OL0'0: = CLOY = Voy OL
0660 6801 $66 0 SOOU © 9sO atts BLTG ae el OCO Oe 08.5. Steer 09
0660 eo El $66 0 SG, sgeGl; Etystc— Ll eaCme UN ota prod SIl6S SI8e OSS
8760 8 l AACS DTV
1 aS 0 A CO A Ml BN CMM OCLC COLO §=—966S = SOTE Ov
LS8'0 LOI 7980 C930 Erol SOPPR— Ovlc+ FOO tae LPOG. SLES Brel ac Of
SUL oro 0e “OIL 0 @ O1L0 ~99Gl+ 20607— ES EMT ae £66 C= l6e Qo 2350 me O68el 0C
10S°0 669 0 €0S°0 EOS0 iO 9) 1-8-— Cc ee LS0a Lv Cc I6£0™ 0095 RAGS iI Ol
coo = “Por0 Ete 0 ee GO ee re =
valli 686 0+ BO Val ie OSC te .6GuCelle O30) 00S
807°0 CLE O 80¢ 0 SLOCORORS O59 Gay Wood Vote l= VO Cees OCE. oO ie elev 06
CCU CE Umer ool 2 Oc Um OSt le eSlC UE S8l1— far be $907 OVE T eee cl.0 08
-960'0 cS1 0 1600 160:0' = 366:C—. 2.099 0-1 AW Coe POL Oss Str 9 eotkids 6100 OL
LSD Oi OOO 0900) «0900s PEGE 6S IG Cle 60S 0— 9pe8 ISO codd 09
9600 6500 8£0°0 St Orr SE ee 988 1 Cr p9T 0— 0988 c6c0 L0 OSP
I1€00 1€0°0 €70 0 te OF a Voy tee L0G 6tr t— L60:0— SEL Sin nck lO ~SV00 Ov
18100 COO ein OF eee Se oe). CCL le CLC £00°0 — 8669 LOO vrr00 Of
0LO0'0 Ssc7000 Or000 Ov000 Y980I— PC80+ 88¢ 1— LIO0+ 8cc ee = 9000 = v0 0 0c
Peou 0 000 0 CIOO |cLO0 O- LIV O=— 2 LOC 0+ pivot 8000+ LeOl 1000 8000 Ol
90000 10000 70000 voodoo stlO— 8800+ 9tl Os €00°0 + Ore 0 £00°0 OOP
70000 10000 10000 Or00— 9200+ 0v0'0— 100°0+ 0010 100°0 06
510,01 S00 0-- C10 0— c£00 O8t
dee Ae A | (Ye a R(X) (\)°~ (y)!0 VEE AWE sie (uu) Y
YISUIIAL AA
syeuIorysIq JO (q)—(4) (8)—(4) (g4)—(48) (9q)—(ak) ——_—_-sasuodsay au0p
soloUdIolyyq snourmn’| 930}S-p1Ty 938}S-Ppuo0dag 930}S-}SI14
(6r61) ppng Aq payejnsjeD pue A104], 2U0Z JoTNIA| Ul paye[NJsog Sassed01g UOISIA -10[0D Jo suoyouNy asuodsay eyvadG (T"¢'8)TFIGRL
 (¥)pz00'0+(¥)'™0180°0+
7000
60000
1c00°0
EO 800°01v00°0
ct0'0
LOL 190°0
0 0 0

800 7000
9100 C000
100
°0 Gs

(v)&ZE100'0
01000
C800) 1co00 $0000
L100 Ivoo0o0
c£0'0
LOL L90°0
0

O1000
e800 [7000 sd000
L100 [7000
LoCcu
BUCO
Un Us Use
saey
CCl Gyou'
veo0+
910+ p000+
8000+coodo+
1000+
LO
Us Ute
0a
teed &

(¥)<4z161'0+(v)'45L00'0
=

4z00'0—(v)*9
ESO OSE0 060 900+
8810+ €700+ 9000+
1100+
Lett
+ 0:

(YS
(x) 20000's

(v)*Z000r'0—(v)*46r8¢'0+(X)'Z1510'0 = (Vv)
(V)0+

JaArlasqo prepueis [€6] AID =(V)zZ “(v4 “(yx

vl00+
0£0°0+
L000
£00°0+
100°01000+

(x)
+

+
ZpE

(v)'0180'0—(X)
(V)A8Lpr' r9'

pse 0610+
RS 999 yc01100+
£60 7U0 900°0

Spt
(v)'™zz9L'0—
(x)“coz9'0—(V)'9 =(VX)'9

bal ig)
A)
0+ CY
Ue 0 or 0+ Oe +
7+ (VY) xX9OS6T'E =(Y)'d

CY) Gratin)

"A (x)
=
(KY)
EA
(Vv) = (Vv)

Ff
=

EeOLOUS
£000
LOO
£9 (v)
Stn 0

=
~(y)&d

US US)

Syn
Essiel 8S£0 yc0'0
£6009700
GIGeU e100
900°0
panujuo> —(1'€8)1 Ge.

4=76p0'L+
2€96r'0+
(x)
(VY)
XSSp5'7T—
(VX)
(Vv)

ty

OS9 09 OL 08 06 OOLOl OC Of
€000°0
€00°0+
£0000
€00°0
OV£000°0

suoyonby

638
 Neural Models 639

where (u,, 0,) and (u,v) are the chromaticity

First Stage
(cones)
points of the two given stimuli whose chromatic-
ity difference Ac, is to be determined, (u,, v,) is
the dichromat’s confusion point in the (u, v)-dia-
gram, and K , = const.

For protanope (d = p):

Relative
Response
Spectral ui, 01058, v, = 0.334

For tritanope (d = f):

u, = 0.249, v, = —0.009

(c) General Measure of Perceived Size (Ac) of

Second Stage Chromaticity Difference
(Intermediate
coding)
400
Response
Spectral
Relative 500 600 700 Ac =|(K,A0,)° + (K,A8,)"]
+4 S\ B3(A) = Vy)
q scale’ =
0.8
[17(8.3.1)]
0.6
+2 0.4
where K,,A6, is given by Eq. 16(8.3.1) applied to
0.2
ro) 0
protanopic vision, and K,A@, is given by Eq.
16(8.3.1) applied to tritanopic vision.
Third Stage
In the above equation, it is assumed that
(final coding) whatever the protanope can discriminate, the nor-
Relative
Response
Spectral mal trichromat, endowed with both protanopic
400 500 600 700 and tritanopic discriminative ability, can see
Wavelength A(nm) better.
Figures 2(8.3.1), 3(8.3.1), and 4(8.3.1) illustrate
Fig. 1(8.3.1). Relative spectral response func-
the agreement between the predictions made by
tions of color-vision processes postulated in Miller
the Judd—Yonemura model and selected observed
zone theory, as calculated by Judd (1949) and
wavelength discrimination data for protanopes,
tabulated in Table 1(8.3.1).
tritanopes, and normal trichromats.

The basic formulae given by Judd and 8.3.2 Adams

Yonemura (1970) are the following:
Another zone theory of color vision dating back
(a) CIE 1960 (u, v)-diagram many years is that of E. Q. Adams (1923, 1942).
In that theory [see Figure 1(8.3.2)], it is assumed
= 4X = 4x
that there are rod receptors (S) and three kinds
Seis ya © xe Pty 3 of cone receptor (R, W, B) located in the retina.
6Y 6y The rods and the W-cones are directly connected
to a neural mechanism responsible for eliciting a
D
ex eo ele
“white” sensation. The R and B-cones modulate
[15(8.3.1)] the response of the W-cones by means of lateral
neural connections. Only an excess of R-cone
(b) Perceived Size (Ac,) of Chromaticity Dif- responses over W-cone responses is permitted by
ference for Dichromat the neural network to reach the (R, W)-neural
coding station to elicit a red sensation. If, on the
es 2 [tan OneOg |G ele | other hand, the W-cone response is in excess, the
4 : u; ~~ Uy uy Ug (R, W)-neural coding station produces a negative
signal that elicits a green sensation. A similar
[16(8.3.1)] mechanism is postulated to function with W-cones
 es
ee
rs
a

(nm)
Difference
AX
Wavelength
Noticeable
Just-

O
420 440 460 480 500 520 540
Wavelength A» (nm)
Fig. 2(8.3.1). Wavelength discrimination of six protanopes (Pitt, 1935) compared with predict-
ions made by Judd—Yonemura model of Miller theory (prediction is given by solid dots;
K,, = 0.47) (from Judd and Yonemura, 1970).

oO
re)
fo)
>
|je)
(e)
(©)
je)

fe)
ia’) 1e) °

fe) (e) +> >

e) °

{e)
400 450 500 550 600 650
Wavelength A(nm)
(nm)
Just-Noticeable
Difference
Wavelength
Ad
Fig. 3(8.3.1). Wavelength discrimination of four tritanopes (Wright, 1952) compared with |
predictions made by Judd—Yonemura model of Miiller theory (prediction is given by solid dots
connected by solid line; K, = 0.24) (from Judd and Yonemura, 1970).

640
 Neural Models 641

S)
oO

0D
——=4¢
5,—

kO tS) 450 500 550 600 650

Wavelength 2X (nm )
(nm)
Ad
Difference
Wavelength
Just-Noticeable
Fig. 4(8.3.1). Wavelength discrimination of two normal trichromats (Bedford and Wyszecki,
1958) compared with predictions made by Judd—Yonemura model of Miiller theory (prediction
is given by solid dots connected by solid line; K, = K, = K, = 0.47) (from Judd and Yonemura,
1970).

and B-cones which, through lateral neural con-
nections, feed into a (B,W)-neural coding sta-
tion.
The photon absorption properties of the pig-
ments in the cone receptors are directly linked
with the color-matching properties of the visual
system. In particular, a linear combination of, for
example, the CIE 1931 standard color-matching
functions is assumed to characterize the photon
absorption properties. Equations 1(8.3.1) of the
previous section may serve. Also Eqs. 2(8.3.1) and
3(8.3.1) concerning complex color stimuli of spec-
tral radiant power distributions {S, dA} and hav-
ing tristimulus values X,Y, Z are assumed to
apply to the functioning of the initial receptor
stage of the visual system.

{I\\ +
Cone <n ‘) ee} =
=
—

Vz Cz (Vz -W)

Fig. 1(8.3.2). Schematic diagram of neural network postulated in Adams’ (1923, 1942) zone
theory.
 642 THEORIES AND MODELS OF COLOR VISION
The initial receptor stage is followed by an
intermediate coding stage in which the cone sig-
nals are converted to nonlinear functions of the
X, Y, and Z tristimulus values. These functions
are assumed to be in the form of the Munsell-value
functions Vy,Vy,Vz (Sections 6.3, 6.5.1, and
6.6.1). A good cube-root representation of such
functions is provided by the following expres-
sions:
[1(8.3.2)]
1/3
Vz=az |= + bz
where ay, by, and so on are constants, and
X,»Y,> Z, refer to the tristimulus values of the
“white” object-color stimulus.
The algorithm embedded in Adams’ model of
color vision has been exploited with moderate
success in a number of color-difference formulae
listed in Table 1(6.5.1), particularly those devel-
oped by Adams (1942) himself, Saunderson and
Milner (1946), Hunter (1948, 1958), Glasser et al.
(1958), and the CIE (1978a).
In his analysis of the Adams theory, Judd
(1951b) replaces the nonlinear functions
Vy,Vy,Vz by linear approximations and pos-
tulates the following spectral response functions
for the final (third) stage of the visual system:
[red-green] = B,(A) = x(A) — y(A)
[blue—yellow] = B,(A) = 0.4[ y(A) — Z(A)]
[white—black] = B,(A) = y(A)
(2(8.3.2)]
These equations closely agree with those of Eq.
8(8.3.1) of Judd’s (1949, 1951b) model of Miiller’s
third stage processes.
8.3.3. Hurvich and Jameson
Over the last three decades, Hurvich and Jameson
have made major contributions to a better under-
standing of Hering’s opponent-colors theory and
to the development of a broad psychophysical
database in experimental support of opponent
processes. The work of Hurvich and Jameson is
documented in their numerous publications, the
articles by Jameson (1972) and Hurvich (1978)
providing informative reviews and references to
the original literature.
The psychophysical data that are basic to many
ideas put forward by Hurvich and Jameson are
the chromatic-response functions (or chromatic va-
lence functions) described in some detail in Sec-
tion 5.13. These functions are assumed to be
linear combinations of color-matching functions,
and the basic mathematical structure adopted by
Hurvich and Jameson in their models is similar to
that used by Judd (1951b) for the third stage of
the Miller and Adams zone theories [see Eqs.
8(8.3.1) and 2(8.3.2)]; that is,
[y(A) — B(A)] = 0.45(A) — 0.42(A)
[r(A) — g(A)] = 1.0€(A) — 1.05(A)
[1(8.3.3)]
where x(A), y(A), Z(A) are the color-matching
functions of the CIE 1931 standard observer.
Figure 1(8.3.3) illustrates the chromatic-response
functions derived from Eq. 1(8.3.3).
According to the opponent-colors theory in
the initial (1955) Hurvich and Jameson formula-
tion, the chromatic response functions represent
the differential effects of ight on four substances
“B”, “G,” “Y,” and “R” with spectral sensitivi-
ties B(A), G(A), Y(A), R(A) expressible as linear
combinations of the CIE color-matching func-
tions x(A), y(A), Z(A):
B(A) = 13.0682 (A) + 0.26722(A)
G(A) = —0.6736xX(X) + 14.00187(A)
+0.00402(A)
¥(A) = —0.0039¥(A) + 13.46807(A)
—0.1327z(X)
R(A) = 0.3329%(A) + 13.0012 7(X)
—0.00112(X) [2(8.3.3)]
These four functions which are not linearly inde-
pendent, as illustrated in Figure 2(8.3.3), all have
a single maximum near 555 nm, and are unlike
 Neural Models 643

0.50

WW
Dn.
2 025
O
a
Ww
uJ
a
oO 0.00
fk

aq
=
(e)
a -0.25
4B
O

=@:50

[fa
400 500 600 700
WAVELENGTH (nm)

Fig. 1(8.3.3).Chromatic-response functions postulated for the CIE 1931 standard observer
when “white” adapted. (Hurvich and Jameson, 1955).

the sets of three fundamentals yielded by most equal to

visual research (Section 8.2.5).
On the basis of more recent evidence, coming
particularly from fundamental discoveries and de- [y(A) — BAY] = kL YA) — BA)]
velopments in electrophysiology, the notion of
“photochemical decompositions of substances in [r(A) — gA)] = ki [RO) — GA)]
post-receptor sites of the visual system” has given
[w(A) — s(A)] = k,[0.5B(A) + 0.5G(A)
way to “neural coding stations” in which neural
signals are generated. These neural signals are +1.0¥(A) + 1.0R(A)]
wavelength dependent and indeed can be pos-
tulated to have spectral response characteristics of —k,[0.5B(A) + 0.5G(A) + 1.0¥(A)
R(A), G(A), Y(A), and B(A). In fact, in de-
scribing the Miller theory (Section 8.3.2), we +1.0R(A)] [3(8.3.3)]
already avoided reference to “substances” in the
opponent mechanisms, though Miller, Judd, and
These equations can be expressed in terms of the
Adams all speak of such substances in their origi-
CIE color-matching functions by making use of
nal work.
Eq. 2(8.3.3):
With the four “substances” or, more ap-
propriately, four “neural mechanisms” defined by
their spectral sensitivities R(A), G(A), Y(A), B(A) [y(A) — b(A)] = &,[—0.0039x
(A)
as given in Eq. 2(8.3.3), Hurvich and Jameson
(1955) then set the spectral response functions +0.3998)(A) — 0.39997(A)]
644 THEORIES AND MODELS OF COLOR VISION

SENSITIVITY
SPECTRAL
RELATIVE

400 500 600 700

WAVELENGTH A(nm)
Fig. 2(8.3.3). Spectral sensitivity curves R(A),G(A), Y(A), B(A) for the four neural mecha-
nisms postulated by Hurvich and Jameson (1955) as linear combinations of the CIE 1931
color-matching functions.

[r(A) — g(A)] = ky[1.0065x

(A) — 1.00067(A) binations of the four neural mechanisms, with the
same relative spectral sensitivities but different
—0.00512(A)] absolute sensitivities. This function is assumed to
play the same role for the achromatic sensations,
[w(A) — s(A)] = (kg — k4)[0.0078%
(A) white and black, as do the chromatic-response
functions for the opponent colors yellow and
) +40.0042 5(A) + 0.00182(A)] blue, and red and green. The factors k,, k,,
k3,k,4,k5 are supposed dependent on the lumi-
~ ksy(X) nance level of the stimuli, and for the special
conditions of the basic experiments
with k; = 40(k; — k,) [4(8.3.3)]
koe 0:9Sact beeen
The additional function [w(A) — s(A)] repre-
sents the differential activity of two similar com- By varying the values of these factors, the relative
 Neural Models 645

importance of yellow-blue hues as against red- 5.13 on the work of Krantz and his co-workers, as
green hues at different luminance levels can be well as Section 8.3.5). The factors n, /, and so
allowed for. The changes of the k; factors leave forth, are then defined by
the relative values through the spectrum of each
response function unaltered.
pt ce Mie © RS
Under chromatic adaptation in the restricted
sense of adaptation to non-neutral surrounds and
B,’ G,’ Yooper FE,
so forth, a further modification is admitted by esas yee 1A pee panels
which the actual shapes of the response curves are ifs m’ ne? fy tp

changed (Jameson and Hurvich, 1956a). The

modification consists in the introduction of new where B,,G,, Y,, R, represent the tetrastimulus
factors n,/, p,m,n’,l’, p’,m’ in the following values for the actual condition of chromatic adap-
way: tation and By, Go, Yo, Ro are these values for an
equal-energy spectrum.
(y — b) =k,(nY— IB) Hurvich and Jameson have compared a num-
ber of experimental data with predictions made
(r— g) =k,(pR— mG) by their model in its original form or variations of
(w—s) =k,[0.5(1B + mG) + nY + pR] it and, in general, they find reasonable agreement.
Figure 3(8.3.3) illustrates, as an example, the
—k,[0.5(B + m'G) + n’Y + p’R] agreement obtained between observed and pre-
dicted “saturation” discrimination data (see Sec-
[8(8.3.3)] tion 7.10.3 on purity differences). The observed
data were obtained by measuring the luminance
Here, integrated “tetrastimulus” values Y, B, AL(A) of the monochromatic stimulus that must
R,G replace the spectral values Y(A), R(A), be added to a “white” stimulus of luminance L,,
G(A), B(A) previously used. The implication is for an observer to detect a just-perceptible change
that the resultant chromatic and achromatic re- in chromaticity. For different monochromatic
sponses (y — b), (r — g), and (w — s) satisfy the stimuli (A) but the same “white” reference
additivity law when stimuli with an extended stimulus, the logarithm of the following quantity
spectrum are employed (see, however, Section is plotted in Figure 3(8.3.3) as a function of

2.0

4 Observed

1.5
aN
Q “‘n

a 1.0
di|~.
+ |
a dq
ub 05 Predicted
ro)
pal|

400 500 600 700

Wavelength dX (nm)
Fig. 3(8.3.3). Observed “saturation” (colorimetric purity) discrimination data compared with
corresponding data predicted by the Hurvich-Jameson (1955) model. The two sets of data are
arbitrarily displaced vertically. In predicting 1/p,=[L,,+ AL(A)]/AL(A), the model
constants were set at k, = 0.5, k, =k, =1.0, k, = 0.95 (from Hurvich and Jameson, 1955).
646 THEORIES AND MODELS OF COLOR VISION
wavelength:
1. De tAAD
[6(8.3.3)]
pA) — AL(A)
The corresponding predicted quantity 1/p.(A) is
computed by the following procedure (see also
Graham, 1965):
The variable luminance increment from zero,
AL(A), that must be added to a constant “white”
stimulus of luminance L,, to yield a fixed minimal
saturation increment A6(A) for a threshold re-
sponse, is calculated from
A (A) =
AL(A)[ yA) = 6QA)| H7Q) = 8Q) 1]
AL(A)[| w(A) — s(A) |] + L007 — 5)
[7(8.3.3)]
In Eq. 7(8.3.3), the quantities |y(A) — b(A)| ,
|7(A) — g(A)|, |w(A) — s(A)| are computed for
each wavelength A from Eq. 4(8.3.3) and taken in
terms of their absolute values (without regard to
sign). Of the similar quantities for the given
“white” stimulus, only the achromatic component
(w — s),, remains in Eq. 7(8.3.3), the two chro-
matic components (y — b),, and (r — g),, being
zero. With all but A L(A) given or calculated from
Eq. 4(8.3.3), and A6(A), at threshold, assigned
some appropriate small value (say 0.1), the lumi-
nance increment AL(A) can be determined and
inserted in the expression
Ru POaD ie
[8(8.3.3)]
BAA) AL(X)
to provide the desired comparison with the ob-
served quantity 1 /p.(A).
8.3.4 Guth
Guth and his coworkers (Guth, 1972; Guth and
Lodge, 1973; Guth, Massof, and Benzschawel,
1980) have developed what they refer to as a
vector model of color vision within the framework
of the Miller zone theory described in Section
8.3.1. The model is capable of predicting with
reasonable success a number of color vision data
obtained by normal as well as dichromatic ob-
servers. The latest, but presumably not the final,
version of the vector model (Guth, Massof, and
Benzschawel, 1980), contains several interesting
modeling ideas that reflect a great deal of the
present thinking among researchers interested in
explaining color vision.
Figure 1(8.3.4) illustrates the neural network
assumed to govern color vision. Neural signals
generated in the first stage (cone pigment) are
coded in the second stage consisting of an
achromatic channel (A = R + G) and two chro-
matic channels, (D = B — R) and (T= R — G),
before being forwarded to the “detector mecha-
nism” in the third stage where detection (chro-
matic and achromatic) and interpretation of the
transmitted signals is completed.
The relative spectral absogption characteristics
of the photopigments of the*three types of cone
are assumed to be linear combinations of the
Judd-modified CIE 1931 standard color-matching
functions, x’(A), y’(A), Z(A), as suggested by
Smith and Pokorny (1975) (see also Section 8.2.5),
but it is pointed out by Guth et al. (1980) that
other sets of pigment fundamentals, such as those
by Vos and Walraven (1971) could have been
used with similar success. The adopted transfor-
mation equations are as follows:
(R): F(A) = 0.2435x%"(A) + 0.8524 (A)
—0.05162'(A)
(G): BA) = —0.3954X(A) + 1.1642 pA)
+0.0837z’(A)
(B): b(A) =0.62252'(A) [1(8.3.4)]
Figure 2(8.3.4) illustrates the relative spectral ab-
sorption curves of the cone pigments calculated
from Eq. 1(8.3.4).
In the second stage, the signals in the nonop-
ponent channel (A) and the two chromatic-oppo-
nent channels (7, D) are linearly related to the
cone signals (R,G,B) provided that certain
stimulus properties, such as size, duration, and
radiance, are kept constant. At absolute threshold
conditions, the (A, T, D)-signals are defined by
(A): A(A) = m,[0.59677(A) + 0.36542(A)]
(T):T(A) = m,[0.95537(A) — 1.28362(A)]
(D): D(A) = m,[—0.02487(A) + 0.04835 (A)]
[2(8.3.4)]
 Neural Models 647

THIRD STAGE
(detection and
inter pretation)
| ome |
ic
2
3
E
oO
ms
= achromat

=
fo) c
cae @
ro) rT)
=n —
fo)
S ro}
S i

= ao}
ra} 2
[\ [\ white
black
to
[=
SECOND STAGE
(neural coding)

Relative
Absorptance
Log
FIRST STAGE
(cone pigments)

Fig. 1(8.3.4). Schematic drawing of the neural

network of the visual system postulated by Guth,
Massof, and Benzschawel (1980).

with the weighting factors (multipliers) m, = m7;
=m,=1.0. At nonthreshold conditions, the
same equations hold but the multipliers take on
values different from unity. The physiological sig-
nificance of the multipliers has not as yet been
established and for the time being there remains a
certain degree of arbitrariness in the model. In
each case of making predictions with the model,
“reasonable” values for the multipliers are cho-
sen. The system is then characterized by three
parameters a, b, t which are given by
M4 Mp
a= =
mM, OM, t My Hoy, oie
Hl yt
(3.3
N,ztmr+rmyp
4)
which, in the initial case of absolute threshold
conditions, are
ay = by = Cy =3
When making predictions with the model of
suprathreshold data, m; is usually kept at unity,
and mp is increased while m , is decreased.
The spectral response functions defined by Eq.
2(8.3.4), at threshold level, are illustrated in Fig-
Wavelength dX (nm)
Fig. 2(8.3.4). Relative spectral absorption curves
(log scale) of pigment fundamentals according to
Smith and Pokorny (1975) and adopted by Guth,
Massof, and Benzschawel (1980) in their color-
vision model.
ure 3(8.3.4). The contribution of D-responses is,
in general, small relative to A and 7-responses.
For a stimulus to appear achromatic (“ white’),
both chromatic channels (T and D) must signal
zero. The (x’, y’)-chromaticity coordinates of such
a stimulus are
x’ =0.37 and y’=0.33
The location of this chromaticity point depends
on the wavelengths of the crossover points of the
T(A)-function and the D(A)-function which, at
575 nm and 502 nm, correspond to “unique yel-
low” and “unique green,” respectively.
The basic model defined by Eqs. 1(8.3.4) to
3(8.3.4) has. been used together with appropriate
additional constraints to make predictions of a
variety of different experimental data. We have
selected a few of these exercises from Guth et al.
(1980) to illustrate the merits of the model as it
now stands.
648 THEORIES AND MODELS OF COLOR VISION

1.0

0.8

fo)
° pSo

oO io

ce)

Relative
Response
Spectral

© no

-0.6
400 450 500 550 600 650 700
Wavelength dX (nm)
Fig. 3(8.3.4). Relative spectral response functions A(A),7(A), D(A) of nonopponent
achromatic channel (4) and opponent chromatic channels (J = R — G and D=R + B) pos-
tulated in the model of Guth, Massof, and Benzschawel (1980).

(i) Apparent Hue of Monochromatic Stimuli.
To make predictions of the apparent hue of
monochromatic stimuli (see Section 5.9), the mul-
tiplier mp is set to 4.0 to simulate a visual field
condition of low level, but a level well above
threshold. The relative contributions of “red”
CET), greens (£) pDluc (1h) andes yer
low” (—D) are computed from Eq. 2(8.3.4) at
20-nm intervals across the visible spectrum. Fig-
ure 4(8.3.4) illustrates the results and compares
these with experimental data obtained by
Boynton and Gordon (1965). There is general but
not detailed agreement between prediction and
experiment. The model does not predict “violet”
(reddish blues) in the short end of the spectrum as
seen by actual observers. This deficiency of the
model is difficult to remove without jeopardizing
some of its desirable features.
(ii) Heterochromatic Brightness. The model
postulates, in accordance with experimental evi-
dence obtained earlier by Guth, Donley, and
Marrocco (1969) and Guth and Lodge (1973), see
also Sections 5.7 and 5.8, that visual responses
involving temporal resolution are, in the main,
mediated by the nonopponent channel, whereas
threshold-detection or brightness responses are
mediated by the opponent channels. In particular,
the model’s nonopponent channel (A) is given
spectral response characteristics corresponding to
the spectral luminous efficiency function obtained
by the method of minimum flicker, that is,
AEE) a)
For the prediction of detection responses, and, in
particular, for the prediction of equality of bright-
ness of stimuli of different chromaticity, a “vec-
tor sum” is used of the chromatic and achromatic
signals whose magnitude is obtained simply by
calculating the square root of the sum of -the
squares of the three component signals of the
(A, 7, D)-visual system. For example, at absolute
 Neural Models 649

re) fo) — Observed -? RY — Predicted

/
bh
rs) Green Yellow , Red

60

(%)
Chromatic
Score
(%)
Chromati
Contribu
500 500 600
Wavelength A (nm) Wavelength X (nm)
Fig. 4(8.3.4). Apparent hue of monochromatic stimuli as observed by Boynton and Gordon
(1965) and predicted by the model of Guth, Massof, and Benzschawel (1980).

threshold, the spectral luminous efficiency func- tained by various experimenters (see Section 5.8).
tion obtained by the method of direct (hetero- The quantities A,7, D are linearly related with
chromatic) comparison is given by the (X’, Y’, Z’)-tristimulus values in accordance

V*(A) =[4?2(A) + T2(A) + D2(A)]'”

[4(8.3.4)]
with A(A), T(A), D(A) as defined by Eq. 2(8.3.4). Ore:
Figure 5(8.3.4) illustrates the agreement obtained
between a (foveal threshold detection) mean ob- -0.4
served luminous efficiency function (CIE, 1978;
see also Section 5.7.2) and one calculated in
accordance with Eq. 4(8.3.4).
' o oO
The data presented in Figure 5(8.3.4) are
threshold data and thus, unit multipliers m,, o o fo)

Mr,Mp apply to the calculation of predicted

data. However, Guth et al. note that since
luminous efficiency functions obtained above O
threshold by direct heterochromatic brightness
matching differ little, if at all, from the threshold ' ©)
function, the model reveals a deficiency as it calls
)(luminous
efficiency
Log
for the use of nonunit multipliers when supra-
threshold conditions prevail. The use of nonunit
multipliers would make the agreement less accep-
table.
In applying the vector-sum principle to stimuli
of complex spectral radiant power distributions
{S, dA} with (x’, y’)-chromaticity coordinates,
Guth et al. calculated “brightness to luminance
2.0
ratios” 400 500 600 700
Wavelength » (nm)
2 2 2\!/2 Fig. 5(8.3.4). Mean observed spectral luminous
pote. [5(8.3.4)]
efficiency function (foveal threshold detection)
compared with function (solid dots) predicted by
for comparison with corresponding ratios ob- model of Guth, Massof, and Benzschawel (1980).
650 THEORIES AND MODELS OF COLOR VISION

with the basic Eqs. 1(8.3.4) and 2(8.3.4) of the Figure 6(8.3.4) illustrates, in the Judd (x’, y’)-
model, that is chromaticity diagram, a number of loci of con-
stant B-ratios [Eq. 5(8.3.4)] centered around the
A=m,-0.9341Y’ model’s achromatic chromaticity point (NV) [x’ =
0.37, y’ = 0.33]. The multipliers were set at m, =
T = m,[0.7401X’ — 0.6801Y’ — 0.1567Z’] 0.4, m, = 1.0, mp = 4.0. The loci may be com-
pared with those shown in Figures 2(5.8.1) and
D = mp[—0.0061X’
— 0.0212¥’ + 0.031427] 6(5.8.1), obtained experimentally.
[6(8.3.4)] (iii) Chromaticity Discrimination. In (ATD)-
where space, which according to Eq. 6(8.3.4) is a linear
transform of (X’, Y’, Z’)-tristimulus space,
X= k[S,x(A)ar (t,,d,)-chromaticity diagrams are defined by
r
plotting

y=k[Syp'(A) ar - == against d, =2 [7(8.3.4)]

r

as rectilinear coordinates. Figure 7(8.3.4) il-

Z’ = kf Sy2'(A) dd lustrates such a diagram with its spectrum locus
Av
and purple line. In the particular case illustrated,
the constant luminance (A) plane is based on
and
suprathreshold-level conditions with correspond-
Pua I€ $ ¥. »hi
ing multipliers set at m, = 0.75, mp = 1.0, mp =
a pncpa FU8entree pen ha fom 6.0 [which lead to parameters a = 0.10, ¢t= 0.13,

0.4 0.6 0.8

Fig. 6(8.3.4). Loci of constant brightness to luminance ratio (8) predicted by model of Guth,
Massof, and Benzschawel (1980) in (x’, y’)-chromaticity diagram.
 Neural Models 651

d = 0.77, in accordance with Eq. 3(8.3.4)]. For

different multipliers, the resulting spectrum locus
changes its size and shape.
In a given (t,, d,)-chromaticity diagram, dis-
tances between two given points are related to the
perceived color differences between the corre-
sponding stimuli of equal luminance (A). Accord-
ing to the model, the detectability 6 of a stimulus
is directly proportional to its vector sum (or vec-
tor length), that is,

§~(A27+T?+ D*)'” [8(8.3.4)]

measured from the achromatic point of the spatial
model. A small increment in the stimulus results
in a just-noticeable difference Ad in agreement
with Weber’s law; that is, the following relation is
assumed to be valid:

“ =const =.c [9(8.3.4)]

If the just-noticeable difference Ad is taken as a

radius vector of constant length r rotating about a
given point in (A7D)-space, a small sphere is
generated that defines a locus of constant dis-
criminability about the given point. Because of

Ad = cB = c(A? +7? +,D7)'”*

[10(8.3.4)]
it follows that the radius r= Ad of the small
sphere is proportional to the distance between the
center of the sphere and the (achromatic) point of
origin of (ATD)-space. In the two-dimensional
case of the (t,,d,)-chromaticity diagram, in
which spheres become circles, the radius r of a
small circle centered at a point (t,, d,) 1s given
by
pel tee ah {18 34)
with c, = const (Weber fraction). In Figure

Fig. 7(8.3.4). Example of (t,, d,)-chromaticity

diagram for stimuli of equal (achromatic, A ) lumi-
nance in (ATD)-space of Guth, Massof, and
Benzschawel (1980). The diagram’s model multi-
pliers are m, = 0.75, m; = 1.0, mp = 0.77. Just-
noticeable chromaticity differences around a given
point (t,, d,) define a small circle whose radius is
proportional to the distance of the center of the
circle from the achromatic (“white”) point of the
diagram (after Guth et al., 1980).
652 THEORIES AND MODELS OF COLOR VISION

0.8
Guth, Massof, Benzschawel (1980)
(10 X Enlarged )

0.4

0.2

0.2 0.4 0.6 0.8

Fig. 8(8.3.4). Chromaticity-discrimination ellipses predicted by the model of Guth, Massof,

and Benzschawel (1980) in the (x’, y’)-chromaticity diagram. The model multipliers are
m _, = 0.75; m; = 1.0, mp = 6.0; the Weber fraction is c, = 0.004.

7(8.3.4), a few such circles of radius r have been
drawn with centers (t,,d,) in different parts of
the chromaticity diagram.
Each of the circles drawn in Figure 7(8.3.4)
converts to ellipses in the (x’, y’)-chromaticity
diagram when applying inverse transformations
to the (t,, d,)-points of each circle. If one places
the centers of the circles into points correspond-
ing to the centers of MacAdam’s color-matching
ellipses (Section 5.4.1), the resulting ellipses in the
(x’, y’)-chromaticity diagram can be directly
compared with MacAdam’s data. Figure 8(8.3.4)
illustrates the results of the model, whereas Figure
2(5.4.1) shows MacAdam’s experimentally deter-
mined ellipses.
8.3.5 Ingling
The recent work of Ingling and his associates
(Ingling and Drum, 1973; Ingling, 1977; 1978;
Ingling and Tsou, 1977; Ingling, Russell, Rea,
and Tsou, 1978) concerned with various aspects
of opponent-colors processes has provided intri-
guing new possibilities of modeling color vision.
From this work we have selected the study on the
disparity between cancellation and direct-match-
ing methods (Ingling, Russell, Rea, and Tsou,
1978) which raises a fundamental issue concern-
ing the spectral response functions of opponent-
colors channels and which offers an intriguing
idea of modeling the observed phenomenon.
The usual method of measuring the spectral
response functions, also referred to as chromatic
response functions (Section 5.13), takes advantage
of the property of opponency in the (red-green)-
and (yellow-blue)-channels. For example, for a
given monochromatic stimulus of wavelength A,
the amount of redness is determined by -the
amount of the opponent hue, green, required to
cancel the perceived redness.
 Line Elements of Color Space 653

Oo ‘p

©©

O fe Oe
ay |) C) C) re
C) C) ©

' Oo ine)

O f

Green
of
Primaries
Amounts
Red
or
' Oo 0)

400 440 480 536 560 600 640 680

Wavelength dX (nm)
Fig. 1(8.3.5). Relative spectral response functions of (red-green)-opponent-colors channel
determined by the method of hue cancellation (open squares) and by the method of hue
matching (open circles). Dotted and solid lines represent response functions predicted by a
special model. The square symbols show the amounts of the green primary stimulus (above zero
line) and the red primary stimulus (below zero line) required to cancel the redness or greenness
perceived to be in a 100-td monochromatic test stimulus of wavelength A. The circle symbols
show the amounts of the same red and green primary stimuli required to match the redness or
greenness of the same monochromatic test stimuli (Ingling, Russell, Rea, and Tsou, 1978).

The authors question the validity of the hue-
cancellation method as it requires desaturation of
the test stimulus. They argue that a visual mecha-
nism may exist that is sensitive to the saturation
of the test stimulus and may contribute to the
resulting spectral response function.
To test the validity of the hue-cancellation
method, the authors determine the spectral re-
sponse functions of the opponent-colors channels
by a hue-matching method that does not de-
saturate the test stimuli. Figure 1(8.3.5) illustrates
the results of their determination of the spectral
response function of the (7 — g)-channel by the
traditional method of hue cancellation as well as
by the method of hue matching. The hue cancella-
tions and hue matches were obtained for each test
stimulus A by making use of appropriate pairs of
primary stimuli. Both methods yield closely the
same results except for the violet region of the
spectrum, from 400 to 480 nm. In that region, the
cancellation method indicates that there is about
30 times more redness than indicated by the
hue-matching method.
The effect of greatly increasing the sensitivity
of the (r — g)-opponent channel at short wave-
lengths is thought to be caused by desaturation of
the test stimulus in the process of cancelling its
hue. The effect is considered related to the Abney
effect (Section 5.9) in which a “white’’-stimulus
desaturant increases the redness of violet stimuli.
From Figure 1(8.3.5), it follows that a green
primary stimulus is as effective as a “white”
stimulus in producing additional redness. On the
other hand, because the green stimulus also
cancels redness, the resultant hue change is not
directly observable.
In a separate experiment, Ingling et al. (1978)
measured the amount of green stimulus required
to cancel the redness in a violet stimulus and also
to cancel the redness in a mixture of 680 and 480
nm that matches the redness of the violet stimu-
lus. They found that it takes at least 10 times
654 THEORIES AND MODELS OF COLOR VISION
more green stimulus to cancel the matching red-
ness of the comparison stimulus, regardless of the
retinal illuminance (0.5 to 500 td). Thus, a
cancelling green stimulus, added to a violet
stimulus of 420 nm, produces nine-tenths as much
redness as it cancels, and hence 10 times as much
green stimulus must be added to cancel a unit
amount of redness at 420 nm as must be added to
cancel a unit amount of redness in stimuli of long
wavelengths.
Ingling et al. (1978) develop a neural model
that is based on electrophysiological evidence of
the “silent surround” found in single-unit record-
ings of receptive fields of the lateral geniculate
nucleus in the rhesus monkey ( Wiesel and Hubel,
1966). In accordance with the electrophysiological
evidence, a surround is called silent, if, when
stimulated alone, it produces no change in the
firing rate of the ganglian cell, but nonetheless is
shown to have an effect by inhibiting an active
center. For a simple opponent cell, displaying
different spectral response characteristics in its
center and its surround, a desaturating stimulus is
needed from a spectral region to which the center
is sensitive before the presence of the surround
becomes apparent; a silent surround does not fire
unless there is an active center to be inhibited.
The mathematical model of the silent-sur-
round concept introduces an operator @ such that
aup— ap when
a> b
afb =0 when a < b [1(8.3.5)]
It is assumed that short-wavelength (“blue”, B)
cones lie in the surrounds of receptive fields of
middle (“green’, G)- and long (“red,” R)-wave-
length cones responding in the (G — R) mode.
The B-cones are silent unless G > R.
The transformation of the spectral response
functions r(A), g(A), b(A) of R-, G-, B-cones to
the spectral response function a,(A) of the [r —
g|-opponent-colors channel is given by
a,(A) = —[1.4812(A) — 0.4286 (A) — 7(A)]
, [2(8.3.5)]
when the hue-cancellation method is used, and
a,(A) = —[1.481¢(A)00.4285
(A) — F(A)]
[3(8.3.5)]
when the hue-matching method is used. The coef-
ficients of the above equations are entirely con-
strained by the positions of the “unique hues”
characterized by the fixed crossovers (zeros) of
the a,(A)-response function. The dotted and solid
lines drawn in Figure 1(8.3.5) show the results
obtained by means of Eqs. 2(8.3.5) and 3(8.3.5).
The model explains how adding a green
stimulus to a violet stimulus elicits two opposing
effects. The green stimulus cancels the redness in
the usual opponent sense, but it also activates the
silent “blue”-cone surround, which by opposing
the “green’’-cone necessarily signals redness.
Two differencing mechanisms within the
[r — g]-channel are postulated. One opposes the
“red’-cone to the “green”-cone; the other oppo-
ses the “blue’-cone to the “green’-cone. The
results of the two experiments, hue cancellation as
compared with hue matching, suggests that the
two differencing mechanisms have different prop-
erties. The difference obtained by the (G — R)-
mechanism is a true [r — g]-signal; that is, it
elicits redness or greenness, depending on the sign
of the difference. The G#B mechanism signals
only redness; the difference being zero unless the
output of the G-cones is greater than that of the
B-cones. .
8.4 LINE ELEMENTS OF COLOR SPACE
8.4.1 Basic Concepts
The introduction into color science of the notion
of a line element in tristimulus space originated
with Helmholtz (1891, 1896) in his work on a
quantitative theory of how perceived color dif-
ferences were determined by the properties of
three fundamental color-response systems. To be
more specific, line elements aim to specify pairs
of color stimuli that present a particular constant
perceived difference, for example, a just-percepti-
ble difference, by the positions and distance apart
in tristimulus space of the points representing
these stimuli.
In ordinary or Euclidean space, the distance ds
between two neighboring points P, and P, whose
rectangular coordinates (tristimulus values) are
(U,, U;, U3) and (U, + dU,, U, + dU,, U; + dU;)
is given by
(ds)° = (dU,)° + (dU,)° + (dU,)°
[1(8.4.1)]
Figure 1(8.4.1) illustrates the geometrical concept

U3
dU
P,(U,,U2,Us)
ds
P,(U,+dU, ,U2+dU,,U3+dUs)
U2
U,
Fig. 1(8.4.1). (U,, U,, U,)-tristimulus space show-
ing points P, and P, representing two color stimuli
perceived to exhibit a small color difference whose
size is characterized by the distance (line element)
ds between the neighboring points P, and P,.
involved. However, other quantities ds that are
more complicated functions of dU,, dU,, dU;,
and U,, U;, U; can be postulated as defining the
distance apart of P, and P, in some hypothetical
space with metrical properties different from those
of ordinary space. Suppose, for (ds), the follow-
ing quadratic form is assumed:
2 2 2 value for all such pairs. In the particular case in
(ds) ire 81,(dU,) + 2g) dU, dU, + $29( dU, )
2
+225; dU, dU; + 833( dU; ) + 223, dU; dU,
[2(8.4.1)]
where the coefficients g;, may be any continuous
functions of the coordinates U,, U,,U;, which
make the form positive and definite, that is, ds > 0
for any choice of dU,, dU,, dU, at all points
(U,, U,, U;). The space in which ds is the distance
element (i.e., the line element), is known as a
three-dimensional Riemannian space.
It may be possible to transform the coordi-
nates U,, U,,U; to new coordinates V,, V,, V3 in
terms of which (ds)* reduces to the simple form
of Eq. 1(8.4.1), namely,
(ds) =(dV,) + (dV,) + (dv)
[3(8.4.1)]
In such cases, the Riemannian space is also
described as Euclidean, but in general such a
reduction is not possible. The term Euclidean is
Line Elements of Color Space 655
here used in the wide sense to mean a space for
which, by a suitable and, in general, nonlinear
transformation of the coordinates, the line ele-
ment can be expressed in the form of Eq. 1(8.4.1).
In applications, a particular system of coordi-
nates U,,U,,U; may be of special physical or
psychophysical interest; for example, in models of
color vision, this is so if U,,U,,U, are the tri-
stimulus values with respect to a selected funda-
mental set of primaries, or if U,, U,,U; are the
output signals generated in two chromatic chan-
nels and the achromatic channel of the visual
mechanism postulated in the opponent-colors the-
ory. It has then to be kept in mind that a space
whose metrical properties are defined by the line
element ds may be Euclidean in the wide sense,
but ds may not be in the Euclidean form of Eq.
1(8.4.1) when the physically or psychophysically
important coordinate system is used.
The application of the line element to model-
ing certain aspects of color vision rests on the
following hypothesis: If (U,, U,,U;) and (U, +
dU,, U, + dU,, U; + dU;) are three-variable
specifications of two neighboring color stimuli
(for example, their respective tristimulus values),
then the necessary and sufficient condition for the
pair of color stimuli to be “just-noticeably” or
“just-perceptibly” different is that some suitably
defined line element ds has the same constant
which one color stimulus is kept fixed and the
other is changed relative to the first such that the
distance between the two corresponding points P,
and P, in tristimulus space always represents a
just-perceptible color difference, one obtains an
ellipsoid centered at P,, as illustrated in Figure
2(8.4.1). The ellipsoid is defined by Eq. 2(8.4.1)
with known g;,-coefficients. Only the Riemannian
form, Eg. 2(8.4.1), for the line element has so far
been used as color-vision models. Other possibili-
ties exist and these would yield loci of just-per-
ceptible color differences around a given center
(P,) that are of nonellipsoidal shape.
For two color stimuli P,; and P, that differ by
more than the smallest perceptible difference, the
line element can still be used to evaluate the
magnitude of their perceptual difference, by
adopting the following further hypothesis associ-
ated particularly with Schrédinger (1920b) in his
development of a similar concept foreshadowed
in the work of Helmholtz. The points P,, P, are
supposed joined by a line following any desired
path in space. The number of just-perceptible
steps (ds = constant) between P, and P, along
656 THEORIES AND MODELS OF COLOR VISION
ds = const
U, ve
Fig. 2(8.4.1). (U,, U,, U;)-tristimulus space show-
ing a locus of just-perceptible color differences
between a fixed color stimulus P, and a variable
color stimulus P,. The Riemannian form for the
line element ds (= const) makes the locus an
ellipsoid centered at P,.
this path is evaluated. For some path, the number
has a smallest value and this value according to
Schrédinger’s hypothesis can be taken as the mea-
sure of the perceptual difference between the two
color stimuli. Mathematically, the determination
of the minimal number of steps corresponds to
integrating the line element ds along the geodesic
line between P, and P, and dividing the result by
the constant value of ds representing a just-per-
ceptible difference. In ordinary space with the
line element defined by Eq. 1(8.4.1), the geodesic
paths are straight lines, but with the line element
of the Riemannian form of Eq. 2(8.4.1) they are
generally curved. Figure 3(8.4.1) illustrates the
geometrical concept.
The line element can also be used to trace in
tristimulus space the loci of constant perceptual
attributes, provided one can identify an ap-
propriate geometrical quantity correlating in some
way to the given perceptual attribute. To quantify
the perceptual attributes brightness, hue, and
chroma, it has been proposed to make use again
of the notion of the geodesic line described above.
Figures 4(8.4.1) to 7(8.4.1), taken in sequence,
illustrate the idea. An alternative approach, which
in fact may imply the notion of geodesic lines, is
to make U,, U,, U, the output signals of the two
chromatic channels and the achromatic channel
associated with a color-vision model involving
opponent processes, and then postulate an ap-
propriate functional relationship between these
output signals and the perceptual attribute under
consideration. The latter approach is the more
commonly used approach and lends itself to more
So=_f ds = Minimum
F,
Fig. 3(8.4.1). (U,, U,, U;)-tristimulus space show-
ing various paths drawn between two points P, and
P, representing two color stimuli that differ in
perceived color by more than a just-perceptible
difference. The shortest distance (least number of
just-perceptible color differences) between P, and
P, measured in terms of a line element of the
Riemannian form is the curve denoted by 5); this
is the geodesic line between P, and P).
plausible interpretations by means of physiologi-
cal models of color vision.
Of special interest in colorimetry is the ques-
tion of whether a_ three-dimensional
Riemannian space with a given line element ds
can be mapped into (or embedded in) a three-
dimensional Euclidean space in such a way that
the equality of distances is preserved.
This requirement means that three functions
of U,, U,, U;, denoted by V,(U,, U3,U;),
V,(U,, U;, U3), V3(U,,U;,U;) can be found in
terms of which Eq. 2(8.4.1) becomes
(ds) = (dV,) + (dV,)° + (dV;
This is precisely the same as the condition for the
Riemannian space to be Euclidean in the wider
sense. If the Riemannian space is not of this kind,
it can still be mapped into a Euclidean space but
one of more than three dimensions; that is, func-
tions) Vj(U,,. Us, 03), |U0, U2 eee
V,,,(U,, U>, U3) with m > 3 can be found such that
(ds)’ = (dV,)° + (dV3)° +--+ (dV,,)
[4(8.4.1)]

P.
fas = Minimum
P
U,
Fig. 4(8.4.1). Definition of brightness in the con-
text of the line element ds, as proposed by
Schrodinger (1920b). As the point P,, representing
a given color stimulus, is allowed to move away
from (or towards) the origin, the brightness of the
color stimulus increases (or decreases). When the
distance between a fixed point P,, representing
another given color stimulus, and the variable point
P, is at a minimum, measured along the geodesic
line between P, and P,, the two given color stimuli
are said to be equally bright.
It can be shown that, in the general case, in order
to map a Riemannian space of n dimensions into
a Euclidean space, the dimensions of the latter,
that is, the number of functions V; required, may
have to be as high as m with
n= Bees [5(8.4.1)]
It follows that a six-dimensional Euclidean space
may be required to embed a three-dimensional
Riemannian space (Silberstein, 1943). However, if
the. Riemannian space is two-dimensional, a
three-dimensional Euclidean space can certainly
be found in which to embed it. This means that if
a three-dimensional line element is laid down and
attention is confined to color stimuli that have the
same brightness (or luminance)—this is a two-
dimensional subspace— these color stimuli can be
represented as points of a surface contained in a
three-dimensional Euclidean space. The coordi-
nates of the latter are functions of the tristimulus
values of the color stimuli so chosen that if two
Line Elements of Color Space 657
Us F,
f ds = Minimum
Surfaces of
constant brightness
Us
U,
Fig. 5(8.4.1). In making use of the definition of
brightness given in Figure 4(8.4.1) surfaces of
constant brightness in (U,, U,, U;)-tristimulus
space can be traced. Keeping P, fixed, it is possi-
ble to determine along each radius vector, pointing
in different directions from the origin, a point P,
that has a minimum distance from P,. The mani-
fold of points P, so determined lie on a surface in
tristimulus space representing a surface of con-
stant brightness, that is, of brightness equal to that
of the given stimulus P,. The distance of P, from
the origin is then changed, and the procedure
repeated to obtain other surfaces of constant
brightness.
ot ara
Surface of
ON
Fig. 6(8.4.1). Lines of constant hue on a surface
of constant brightness traced in (Uj, U,,U,;)-
tristimulus space along geodesic lines originating
at P) representing the achromatic stimulus on the
given surface of constant brightness.
658 THEORIES AND MODELS OF COLOR VISION
U,
Fig. 7(8.4.1). Lines of constant chroma on a been assumed to have the Riemannian form de-
surface of constant brightness are readily de-
termined when lines of constant hue [see Figure
6(8.4.1)] are calculated as geodesic lines originat-
ing from point P) representing the achromatic
stimulus on the given surface of constant bright-
ness. Lines of constant chroma are concentric
Riemannian circles.
points are a distance ds apart they are just-per-
ceptibly different, and conversely.
The characteristic embedding properties of a
space specified by a line element are unaffected
by any continuous transformation of the coordi-
nates and are determined mathematically by the
quantity known as the Gaussian curvature which
is a function of the coefficients g;, in the line
element (see, for example, Eisenhart, 1949;
Sokolnikoff, 1951). In order to map one space
into another of the same dimensions, and at the
same time preserve distances, it is necessary and
sufficient that the two spaces have the same
Gaussian curvature at corresponding points. The
Euclidean space has zero Gaussian curvature
throughout. Thus, if the three-dimensional color
space is found to have nonzero curvature, it is
impossible to map this space into a three-dimen-
sional Euclidean space without ruptures or gaps.
Instead, a Euclidean space of more than three,
possibly six, dimensions must be used. Embed-
ding in a space of more than three dimensions
cannot be visualized. The concept of embedding
is then of limited practical use, and one must be
content to describe color space and measure in it
by means of the line element ds uniquely de-
termined by its coefficients g;,.
Line elements and the color-difference for-
mulae already described in Section 6.5.2 provide
related methods for assessing color differences,
but there are differences in the way they have
been determined and in the way they are used.
The color-difference formulae in nearly every case
assume that color-perception space is Euclidean,
or approximately so, but a line element can be
used even when this assumption is not valid. The
basic observations used in setting up line ele-
ments are threshold measurements and standard
deviations in color matching (Section 5.4). Some
observations underlying the color-difference for-
mulae relate to small but not actual threshold
differences, and the formulae proposed may be
applicable to larger or smaller differences.
All line elements put forward up to now have
fined by Eq. 2(8.4.1) and the outstanding problem
is to determine the metric coefficients g;,. Two
differing approaches have been followed; one,
based on theoretical considerations regarding the
functioning of the visual mechanism coupled with
certain experimental threshold data, may be de-
scribed as the inductive method. On the other
hand, in the empirical method, the metric coeffi-
cients g,, as functions of the tristimulus values
U,, U;, U; are derived by an empirical analysis of
blocks of measurements of threshold differences
(or of the related standard deviations of color
matching) obtained for a selection of color stimuli
covering an extended domain of tristimulus space.
MacAdam’s (1943) original contour diagrams for
his coefficients g;, belong to the latter category.
Some of the color-difference formulae used in
recent years result from a hybrid approach in
which the algebraic form of the line element is
derived by the inductive method but with several
parameters involved left open; the best values of
the latter are then arrived at by minimizing the
computed deviations from the empirical data on
the precision of color matching determined at a
necessarily limited number of points in tristimu-
lus (or chromaticity) space.
8.4.2 Helmholtz
Helmholtz (1896) was probably the first to derive
a line element ds of color space, and his de-
termination of the metric coefficients g,, is based
on his three-component theory of color vision and
Weber’s law.
Helmholtz argued that the visual process is
governed by three independent response mecha-
nisms, each with a characteristic spectral response
function, for example, 7(A), Z(A), b(A). The spec-
tral response functions determine the trichromatic

color-matching properties of the eye and must be
representable as linear combinations of three em-
pirically determined color-matching functions,
such as the CIE color-matching functions
X(A), y(A), Z(A); thus,
F(A) = ay, X(A) + ay y(A) + a,37(A)
2(A) = ay, X(A) + ay V(A) + a53Z(A)
b (A) = a4,X(A) + ayy 9(A) + ay7(A)
[1(8.4.2)]
When applied to color stimuli of spectral radiant
power distributions {P, dA}, the responses of the
three mechanisms are
R= [P,7(A) dd, G= [P,B(A) A,
B = [P,b (A) dd —— [2(8.4.2)]
For each mechanism acting alone, the sensitivity
to a difference of response (R’ — R) from two
similar adjacent displayed color stimuli of differ-
ing spectral power distributions is supposed to
satisfy Weber’s law (sometimes referred to as
Fechner’s or Weber—Fechner’s law); that is, the
smallest perceptible difference dR is proportional
to R (or R’); R and R’ being very nearly equal.
Helmholtz, in fact, allows for deviations from
Weber’s law arising from “glare” at very high
luminances of the stimulus and from “self-light”
in the eye at very low luminances. These refine-
ments, which are not material to the main argu-
ment, are omitted here.
Helmholtz then assumes that:
(a) The Weber fractions in the intermediate
luminance range are all equal; that is,
dR
R _ dG
Goa _ dB
BR = const [3(8.4.2) |
(b) With all three mechanisms in operation, the
smallest perceptible difference is obtained by
combining the fractional deviations for the three
mechanisms in the same way as independent “er-
rors” are compounded, using the square root of a
sum-of-squares relation. This gives as the line
element
2 2 21/2
() + (| +(¥)] = const = ds
Line Elements of Color Space 659
or,
© bral wie|W eala
[4(8.4.2)]
To make the predictions of this simple line
element fit data on color discrimination, Helm-
holtz had at his disposal one set of parameters:
the coefficients a,, in the transformation given by
Eq. 1(8.4.2). Helmholtz determined these coeffi-
cients so that the line element would correctly
predict the experimental data of Konig and
Dieterici on the just-perceptible difference of
wavelength for monochromatic stimuli of equal
brightness measured at different positions along
the spectrum. The resulting spectral response
curves 7(A), 2(A), b(A), all of which show two
pronounced maxima, are completely incompatible
with the spectral absorption curves that have now
been measured for cone pigments. The line ele-
ment cannot then be valid if the photoreceptors
of each of the three mechanisms contain predomi-
nantly one pigment. Another difficulty, pointed
out by Schrédinger (1920b), is that the Helmholtz
line element entails a luminous efficiency curve
quite different from the directly observed V(A)-
curve. In fact, the Helmholtz line element predicts
surfaces of constant brightness of the form
R-G-B=const [5(8.4.2)]
These and other difficulties make the Helmholtz
line element mainly of historical interest.
8.4.3 Schroédinger
Schrédinger (1920b), in an effort to eliminate the
shortcomings of the Helmholtz line element, pro-
posed the following modified version:
nt 1
UO igs IpR +166 + 1pB
Ig(dR) ay Ig(aGy” ay [y(4By”
R G B
[1(8.4.3)]
where, as in the Helmholtz line element, R, G, B
are the responses of the cone mechanisms, operat-
ing independently from one another (no cross-
products). However, the contribution of each
mechanism to the square of the perceived total
660 THEORIES AND MODELS OF COLOR VISION
difference (ds)* depends on all three cone re-
sponses through the factor
IpR +166 + I5B
In fact, brightness for this line element is propor-
tional to (/pR+/¢G +1/,B), making a surface
of constant brightness a plane in tristimulus space.
As in all line elements, the choice of the funda-
mental primaries, or of the corresponding funda-
mental spectral response functions, is vital, and
Schrédinger adopted for the latter, the “Grun-
dempfindungen” curves previously derived by
Konig and Dieterici. These are single-peak curves
with maxima in the orange, green, and blue parts
of the spectrum, not radically different from more
recent determinations of fundamental response
functions. With these fundamentals and a suit-
able choice of the relative values of the con-
stants /p,/¢,/,, brightnesses deduced from the
Schrodinger line element agree approximately with
luminances based on the experimentally de-
termined luminous efficiency function V(A). This
concurrence and the additivity of brightness were
considered by Schrédinger as the strong points of
his line element, which in fact he designed to
achieve these ends. Schrédinger’s line element has
not been intensively studied, and it seems incon-
sistent with some of the more recent threshold
data of color discrimination. In tests by Bouma
and Heller (1935) and Rosemann (1938), it was
found to fail to predict correctly experimental
data of various kinds.
8.4.4 Stiles
A less radical modification of Helmholtz’s line
element was suggested by Stiles (1946) based on
his extensive experimental data on two-color
thresholds (Stiles, 1939; 1978; see also Section
7.4). In these measurements, the observer viewed
a large (10°) patch of light of wavelength p and
radiance L,, at whose center a small (1°) square
of light of wavelength A and radiance L,, was
applied in short (0.063 sec) pulses as an addi-
tional stimulus. Within the 1° square during the
pulse, the radiance was L,, + L,, whereas the
10° surrounding field remained at L,,. The ob-
server fixated accurately the center field ensuring
cone (rod-free) vision of the additional stimulus.
By repeated pulses at different radiances L,,, L,,
being kept constant, the threshold radiance L,,
was determined at which the observer has a 50%
chance of seeing the one-degree square. Measure-
ments of L,, were made for a series of values L,,
and for various wavelength pairs A and yp, result-
ing in a large number of experimental curves of
log(1/L,,) against log L,,, for one observer. To a
certain approximation and with the exclusion of
an anomalous effect occuring at extremely high
field radiances, these curves could be regarded as
resultants of three component curves of a com-
mon fixed shape whose positions along the axis of
log L,, depended on A and yp. From the shifts
along the axes, the approximate relative spectral
sensitivities of the retinal response systems were
determined. The observational results showed that
different Weber fractions should be applied to the
three postulated cone-response systems, rather
than equal fractions as assumed by Helmholtz.
A tentative line element consistent with the
salient conclusions of this work, but with implica-
tions going beyond the experimental basis, was
proposed (Stiles, 1946) in the following form:
= [BD Ba
[82a
: [1(8.4.4)]
where the experimentally determined functions
¢(R), §(G), §(B) are given approximately by
SAS ger a (G) = —
$(B) -5 [2(8.4.4)]
The constants p,y,8 are proportional to the
limiting Weber fractions of the three cone re-
sponses at high luminances and, for the observer
used, had the following values:
p = 1.28, y = 1.65, B = 7.25
The cone responses (R,G, B) are related to the
CIE 1931 tristimulus values (X, Y, Z) by the fol-
lowing linear transformation:
R66. 0X 4 12600Y “91202
G = —438.0X + 1620.0Y+ 123.0Z
B=0.708X + 417.0Z
[3(8.4.4)]
 Line Elements of Color Space 661

They correspond to spectral response functions and g, b and g’, b’ are similarly defined (see
with single maxima at about 580, 540, and 445 Section 3.2). In terms of these chromaticity coor-
nm, respectively. dinates, the line element, defined by Eq. 1(8.4.4),
At high luminances, the line element, defined takes the form
by Eq. 1(8.4.4), reduces to
s_[ (sr 3= Sr) ¢(Sr)| 2
(ds) + two similar terms
ee aed dee [6(8.4.4)]
[4(8.4.4)]
By varying S’, keeping r’, g’, b’ fixed, the radi-
and in this form its relationship with the Helm- ance of stimulus P’ is varied without changing its
holtz line element, defined by Eq. 1(8.4.3), is chromaticity. The minimum value of (ds)* under
clear. such a variation is reached when the following
The line element was tested by applying it to condition is fulfilled:
the calculation of just-noticeable wavelength dif-
ferences, the luminous efficiency function, and — l / Dot 2 A
ere es (S’r’— Sr)§(Sr)]|° +two similar terms
wtic

chromaticity discrimination as exemplified by

MacAdam’s (1942) ellipses in the CIE 1931
(x, y)-chromaticity diagram. The line element is [7(8.4.4)]
an improvement over the previous line elements
because some experimental data can be predicted In terms of tristuumulus values R, G, B and retain-
with it fairly well. However, there remain experi- ing only first-order quantities, Eq. 7(8.4.4) be-
mental data that cannot be predicted adequately. comes
In the remainder of this section, we give a
number of detailed numerical exercises with the o- £[Aen] +2 [Sea]
Stiles line element. The reader may find these
exercises of interest as they show the kind of
analysis involved in applying a Riemannian line
element to predict color-discrimination and other [8(8.4.4)]
relevant data.
This is the quality-of-brightness condition that
(i) Calculation 1: Step-by-Step Luminous Ef- applies to stimuli of nearly the same chromaticity.
ficiency Function. In the step-by-step method of In the special case of monochromatic wave-
determining experimentally the luminous ef- lengths A and (A+ dA), and radiance L,, and
ficiency function (see Section 5.7), two juxtaposed (L,, + dL,,), the above condition reduces to
patches of monochromatic stimuli of slightly dif-
ferent wavelengths are viewed, and the radiance
of one is varied until the perceived total dif-
aber
ea
_gy{RimOAL
GARn ,GinedAw
Ge IG, ,(Bs
ChByON
ference between the two patches is a minimum.
The wavelength difference can be made so small [9(8.4.4)]
that the difference of color (mainly hue) at this where
minimum setting is barely perceptible. The proce-
dure is repeated step-by-step along the spectrum.
The luminous efficiency function can be derived
a)
R
GS -

[Fen] [Feo] +[Beco]

from the line element by a corresponding proce-
dure.
The chromaticity coordinates (r,g,b) and
r’, g’, b’) of the two stimuli (P and P’) are re- [10(8.4.4)]
quired; the coordinates r and r’ are given by
and C;, Cj are similarly defined.
Ks R ive By successive applications of Eq. 9(8.4.4), the
rig Gapee Bs? radiances L,, of a series of monochromatic stimuli
whose wavelengths increase in small steps from
,— few hice -met zi
So aes R’ + (7! ats B’ sk [5(8.4.4)] the blue to the red end of the spectrum and each
662 THEORIES AND MODELS OF COLOR VISION

of which matches in brightness its neighbors in
the series, can be determined. The step-by-step
luminous efficiency function V(A) is then ob-
tained by forming the reciprocals of the radiances
L., and by normalizing them to make V(A) unity
at A = 555 nm. At high radiance levels, the quan-
tities C?,C?,Cf tend to (1/p)’,(1/y)”,(1/B)’,
respectively, and the luminous efficiency function
approaches the limiting form
constant
AUN homey este ed
= constant R\I/ey” . Giusy’ Bisby
[11(8.4.4)]
with the constant adjusted to make V(A) = 1.0 at
A = 555 nm. Figure 1(8.4.4) shows V(A) com-
puted from Eq. 11(8.4.4) in comparison with the
step-by-step luminous efficiency function ob-
tained by Gibson and Tyndall (1923) as a mean
for 52 observers (32 observers on the end of the
spectrum). If we bear in mind the large spread of
the individual experimental values in the blue, the
agreement is good. Comparison with the CIE
1924 V(A)-function shows, in fact, the largest
discrepancies in the blue, the CIE values being
lower. It may be recalled that in this region the
CIE values are generally considered too low and
not representative of actual observational data
(see Section 5.7.2).
The shape of the V(A)-curve predicted by the
line element depends to some extent on the radi-
ance level. For a level corresponding to ap-
proximately 80 td (= 8 cd - m * seen through an
eye pupil of 10-mm” area), which is commonly
used in visual photometry, the approximation of
Eq. 11(8.4.4) is not quite adequate, and the
V(A)-curve must be derived from Eq. 9(8.4.4) by
successive approximation. The results of such a
calculation are also shown in Figure 1(8.4.4).
When the radiance is made very low, Eq. 9(8.4.4)
reduces to the limiting form given below as Eq.
12(8.4.4), which can also be derived directly by
accepting the principle that color stimuli (of fixed

V(X)
LOG

600 700
WAVELENGTH Xnm)
Fig. 1(8.4.4). Comparison of luminous efficiency functions V(A), plotted as log V(A_) against
A, of different origins: Dashed line: CIE 1924 standard V(A )-curve; continuous line: Gibson and
Tyndall (1923) step-by-step function; solid dots: derived from Stiles line element for high
luminance; open circles: derived from Stiles line element for medium luminance (8 cd-m_ ”~ 80
td); crosses: derived from Stiles line element for absolute threshold level.

size) at absolute threshold have the same bright-
ness whatever the wavelength:
) ig(2)
V(A) = cons (#8)ns (2
[12(8.4.4)]
Figure 1(8.4.4) includes values of V(A) obtained
from Eq. 12(8.4.4). The main effect of lowering
the radiance level is to increase slightly the rela-
tive luminous efficiencies in the orange and red
and in the blue regions of the spectrum. There are
insufficient experimental data on step-by-step
V(A) functions at different levels to test these
predictions.
The line element can also be used to derive the
relation between color stimuli that are not neces-
sarily monochromatic but have the same “small-
step” brightness; that is, the two stimuli can be
linked by a small-step series of intermediate
stimuli, each having the same brightness as its
neighbors in the series. The condition for equality
of small-step brightness between stimuli P, and
P, of fundamental tristimulus values (R,, G,, B,)
and (R,, G,, B,) respectively, reduces to
Rie)”. G/7)? . BA/BY= RV/e”. Gi/*. Bu/BY
[13(8.4.4)]
when for both sets of tristimulus values the ap-
proximate form, given by Eq. 4(8.4.4), can be
assumed valid. The same Eq. 13(8.4.4) is obtained
by applying Schrédinger’s geodesic definition of
equal brightness to the line element of Eq. 4(8.4.4).
It is clear from Eq. 13(8.4.4) that “small-step”
brightness is not additive. If the two stimuli
(R,,G,,B,) and (R,,G,, B,) have the same
small-step brightness, the equality-of-brightness
still holds if their radiances are changed by the
same factor, c, for instance. In that case,
2 2
2 1 2 a)
7 eteg he j (cG,)' ‘Hee (cB) ¢ of additivity of brightness (see Section 5.8). In
[14(8.4.4)]
However, if stimuli P, and P, each match P; in
brightness, the additive mixture of stimuli P, and
P, will not, in general, match P; increased to
Line Elements of Color Space 663
double its original radiance, for the relation
(Ry, +R.) (G6, + G,)V™ -(B, + BY
10 (2R,)°/" (26, ; (2B,)°/9”
[15(8.4.4)]
is not, in general, true. The largest discrepancy
may be expected to occur when the chromaticities
of stimuli P, and P, are most widely different.
For a particular case involving monochromatic
stimuli in the blue (470 nm) and in the red (680)
— the chromaticity of stimulus P; is immaterial—
one obtains
(Race Ro)” 2
ree eel (B, + B,)°/”
= (264Rt)fa .03G,)> 2(2.638,)0
[16(8.4.4)]
Thus, the breakdown in additivity corresponds in
this case to the difference between 2.0 and 2.63,
or about 27%. ;
For stimuli of any extended spectral radiance
distributions, {L,, dA}, additivity would demand
the following relation:
(1/p)? (1/y) 2
S [Rata an x forte an
(1/8)?
x |[Brte an [17(8.4.4)]
If {L., dd] is the spectral radiance distribution
of a blackbody radiator at 2060 K, the above
equation is found by calculation to be in error by
approximately 8%. Experimental data on small-
step brightness matching for a quantitative test of
this result are not available. However, many ex-
periments on direct heterochromatic brightness
matching involving stimuli with widely different
B)?
chromaticities have shown a pronounced failure
general, a mixture of two saturated complemen-
tary color stimuli giving a resultant “white” has a
lower brightness than the sum of the brightnesses
of the components. Such effects are not explained
by the deviations predicted by the line element.
In fact, they go in the opposite direction.
664 THEORIES AND MODELS OF COLOR VISION

(ii) Calculation 2: Wavelength Discrimination. (iii) Calculation 3: Chromaticity Discrimination.

The difference of wavelength of two just-noticea-
ble different monochromatic stimuli of equal
brightness can be calculated from the line element
in Eqs. 1(8.4.4) and 8(8.4.4) using the substitu-
tions
dR,
R= 1.Ry ands dR = Ral aa Lal dX
and similar ones for G, B, and dG and dB. After
eliminating L,,, the following equation is ob-
tained for the wavelength difference dA, produc-
ing a just-noticeable color difference ds:
The line element given in Eq. 1(8.4.4) can be
expressed in terms of CIE (x, y)-chromaticity
coordinates and luminance L = Y by making use
of Eq. 3(8.4.4) and the fact that the chromaticity
coordinates x, y,z and the tristimulus values
X,Y, Z satisfy, by definition, the relations
dn
x= eh and Z= (z)y- C=)
Sty
[20(8.4.4)]
Thus, Eq. 3(8.4.4) may be written

a
(ds)! = (dd) i[Ron
= WR
no
n= 192)+tay+ orl2)
11
xX
y 12 13
Zz
y

= tat(ste carat
21
xX
y 22 23
Z
y

$e8Gy OB TRCay
2

XxX Zz
B t 3] y(=)
y +: 32 Yu 33 y(=]
y

+ two similar com [18(8.4.4) |

At high radiances this reduces to
[21(8.4.4)]
The numerical values of ¢;, are given in Eq.
3(8.4.4). By considering only variations in chro-
maticity (x, y) and keeping luminance (Y) con-
stant, we find that

(ay = (any |(1/p)”

R,
ei dR,
JA dR = (a — ty3) dx
My

(4)
%
(i/yy 9G, , (1/B)° aB, ues AGE as te) = t13 | dy
G, dA B, 2a
and two similar expressions for dG and dB. The
complete line element of Eq. 1(8.4.4) then takes
+ two similar cr [19(8.4.4)]
the following form:

For a sequence of pairs of stimuli spanning the (as)? = (ax)*|[ Bsc] +| eco]

[pe]
spectrum of all the same small-step brightness as
determined from the line element and at an abso-
lute level corresponding to 8 cd - m * (~ 80 td)
at AX = 555 nm, the variation of dA with A calcu-
lated from Eq. 18(8.4.4) is compared with the
mean experimental data reported by Wright and +2 dx y{pupn|
(8) |
Pitt (1934) and Bedford and Wyszecki (1958) in

rare) nD)
Figure 2(8.4.4). The predicted values of dX are
proportional to ds, which is here put equal to 0.01
corresponding to the limiting Weber fraction for Y
the conditions of observation commonly used in
determining di. The agreement between experi-
+(ay'{ [22] +[2200@))
sue}
mental values and those predicted from the line
element is considered satisfactory except for
shorter wavelengths where the predicted values
appear to be somewhat low.
[22(8.4.4)]
 Line Elements of Color Space 665

NR ome als
2 Ok ee
te Se aes |
dA
(nm)
DIFFERENCE
i
WAVELENGTH
NOTICEABLE
JUST
40 50 00

WAVELENGTH AX (nm)
Fig. 2(8.4.4). Comparison of wavelength discrimination curves observed by Wright and Pitt
(1934), Bedford and Wyszecki (1958), and predicted from Stiles’ line element.

where illustrates the results; for comparison with

vg MacAdam’s ellipses, see Figure 2(5.4.1).
1 =—(t,—7
a 11 — f3) There is a general similarity between the two
sets of ellipses. The orientations of the ellipses
tf correspond well, and their areas also show a
ee elses fiy)lcates| reasonable correlation. The absolute areas of the
a
Stiles ellipses exceed those of MacAdam’s on the
He average by a factor of 20.7, that is, (4.5)*. This
P21 =—(t,,—t
at 21 23) arises, at least in part, from the fact that the
MacAdam ellipses refer to standard deviations
Ng of color matching (estimated by MacAdam to
Poa sahil tasatlon ant a3,| be approximately one third the corresponding
threshold differences). The ratios of the axes of
iy" the ellipses cover a range of 2.5 to 1.0 and have
P31 =—(t3,-t
mrNy 33) average values of 2.3 (Stiles) and 2.9 (MacAdam),
but they show no correlation.
4 The discrepancies between MacAdam’s el-
P33 penal ect Vie lipses and those of Stiles have been discussed by
Stiles (1946) and more recently by Wyszecki and
An interesting application is to compute loci Fielder (1971). A quantitative analysis of the dis-
of constant ds around given points of chromatic- crepancies between the two sets of data and, for
ity (x, y) for a constant level of luminance Y. that matter, between those of MacAdam and of
Such loci are, of course, ellipses in the chromatic- other investigators, is a complex problem (see
ity diagram. If the positions of the given chro- Section 5.4.5) because different observing condi-
maticity points are appropriately chosen, a direct tions and different observers are involved in the
comparison of the computed ellipses with those independent studies, and all observational data
directly measured, for example, by MacAdam must be accepted as being erratic to some degree.
(1942) can be made. Again taking ds = const = On the other hand, the model of the functioning
0.00124 (in the units used by Stiles, this corre- of the visual mechanism, on which the line ele-
sponds approximately to a luminance of 48 ment proposed by Stiles is based, is probably too
cd - m * seen through a pupil of 2.6-mm dia, ie., simple, and it may not be permissible to postulate
about 250 td), we compute 25 ellipses whose complete independence of the three-cone re-
centers are located at the chromaticities used by sponse systems. Whatever new line elements may
MacAdam (see Section 5.4.1). Figure 3(8.4.4) be proposed, they may have to include cross-
666 THEORIES AND MODELS OF COLOR VISION

520
Stiles Line Element
Ellipses plotted to
three times actual scale

0.4 0.6 0.8

x

Fig. 3(8.4.4). CIE 1931 chromaticity diagram showing discrimination ellipses derived from
Stiles’ line element.

product terms of the cone responses to account This two-dimensional space, mapped in any
for the interdependence of the cone systems. three-dimensional space, generates a surface with
the Gaussian curvature K calculable from the line
(iv) Calculation 4: Gaussian Curvature of Color
element using the formula (see Sokolnikoff, 1951,
Space. The inherent differences between the two
for example):
sets of data obtained by Stiles (1946) and
MacAdam (1942) are strikingly demonstrated by
considering the Gaussian curvatures calculated pomicsih 24|Puil Shien (Eig Vee
from Stiles’ line element and estimated from 2g!/2| Ox \ g.g!/? dy gif? Ox
MacAdam’s ellipses (Silberstein, 1943).
The line element of a _ two-dimensional
Riemannian space, such as that describing dis-
‘@ p79 Oyppeog Is,
dy \ gli/2 dx gi/? dy
crimination ellipses in terms of chromaticity coor-
dinates (x, y) at constant luminance Y has the
form $2 981
gigi’?
——
Ox )
4(8.4.4
[2(8 )]
$2 2
(ds) = g,,(dx)” + 2g) dx dy + go) (dy)
When this formula is applied to Stiles’ line ele-
[23(8.4.4)] ment as defined by Eq. 22(8.4.4), we find thatK
where varies with chromaticity (x, y) and is negative
ST BAIS Te Sip # 0 throughout the domain of real chromaticities.

x 0.4
Fig. 3(8.4.4). CIE 1931 chromaticity diagram showing discrimination ellipses derived from
Stiles’ line element.
Figure 4(8.4.4) shows the CIE chromaticity dia-
gram with loci of constant Gaussian curvature K.
Silberstein’s (1943) estimates of K based on
MacAdam’s ellipses indicate that the Gaussian
curvature is positive in some areas of the chro-
maticity diagram, negative in others, thus clearly
indicating that the two results of Stiles and
MacAdam are incompatible. No linear or nonlin-
ear transformation exists that would convert one
set of ellipses into the other. However, it is dif-
ficult to estimate, and in fact has not been tried,
how the Gaussian curvature would vary if small
perturbations, simulating observational uncertain-
ties of about 15%, were introduced to the lengths
and orientations of the major and minor axes
of the MacAdam’s ellipses. On the basis of
Farnsworth’s (1958) work illustrated in Figure
4(5.4.1), such perturbations may show that the
Gaussian curvature underlying MacAdam’s el-
lipses could be near or equal to zero in the
chromaticity domain covered by the ellipses.
The Riemannian surfaces corresponding to
Stiles’ line element (Y = const) and MacAdam’s
Line Elements of Color Space 667
0.6 0.8
results can be embedded in a three-dimensional
Euclidean space. On such a surface, each ellipse
becomes a circle of equal size, thus making it
a uniform-chromaticity-scale surface. MacAdam
(1944) has constructed the surface for his results,
and its curvature can clearly be seen on the
model. From the photograph of this model, shown
in Figure 5(8.4.4), the meaning of the concept of
Gaussian curvature is readily appreciated, as is
the fact that it is impossible to flatten this surface
to a plane (K = 0) without rupture.
The different uniform-chromaticity-scale sur-
face of the Stiles’ line element is shown in Figure
6(8.4.4) as a drawing in an oblique projection of
the three-dimensional Euclidean space, in which
case the equation of the surface is derivable alge-
braically as below.
Suppose X,, X,, X; are Cartesian coordinates
in the Euclidean three-dimensional space. The
line element, given as Eq. 23(8.4.4), then repre-
sents the square of the distance between two
adjacent points on a surface, the Cartesian coor-
dinates of any surface point being expressed as
668 THEORIES AND MODELS OF COLOR VISION

Fig. 5(8.4.4). Uniform-chromaticity-scale surface based on MacAdam’s discrimination ellipses

(from MacAdam, 1944).

follows: For the Stiles’ line element, the metric coefficients

g,, are the coefficients of (dx), 2 dx dy, and
X, = X(x, y), X= X,(x,y), X3 = X3(x, y) (dy)? in Eq. 22(8.4.4), and according to the ex-
pressions for {(R) = 9/(1 + 9R), and so forth,
Thus, they are readily expressible in the following form
[inf ~+ +4) logs i:

(ds)° = (dx,)° ae (dx,)° al (dX;)°

lice 4
=|(4) sibare) lhcen aie) 211 Palaeee +9R)

(8 2 aX,
Ox dy
aX, aX;
Ox Oy
aX
dx dy
dx dy
l 0 2
+ i Fee _ 9G)|

(38) +(e) +() Jeo + alaxind + 9B)

[25(8.4.4)] 2
 Line Elements of Color Space 669

2
420
9 400

BAY re) '

x< Ol j
'

esae
a

SPECTRUM
LOCUS
CO)os
at)

f—PURPLE
LINE

Fig. 6(8.4.4). Uniform-chromaticity-scale surface derived from Stiles’ line element.

gn = [gy la( +9R)| Faue c °R)| gn = =|ayia + on)

varies + 96)]] F1m( + 90)| +3 aint +96))

+ aay + 98)]] >1m( + 98)| +5 [pit + 9B)| [26(8.4.4)]

670 THEORIES AND MODELS OF COLOR VISION

When Eq. 26(8.4.4) is compared with Eq. 25(8.4.4), can be made to coincide with any particular point
the appropriate functions X,, X,, X; are, clearly, (Xo, ¥) of the chromaticity diagram by putting

X,= X(x, y)= 7in(1+ 9R) + a,

ee yey “In(1 +96) +4, = =In(1+ 9Gy)

X; = X3,(x%, y) = raat aOR \ as

[27(8.4.4)] where Ry, Go, By are the fundamental tristimulus

values R,G,B of a stimulus having the chro-
where a), a, a, are arbitrary constants indepen- maticity (Xo, yo). In the example illustrated in
dent of x, y. From Eq. 27(8.4.4) and Eq. 21(8.4.4) Figure 6(8.4.4), the values chosen for x9, yp are
it follows that 0.310, 0.316, respectively, which represent the
chromaticity of CIE standard illuminant C. The
ein) —1+9R luminance Y was set equal to 0.00124, the value
used in the calculation of the discrimination el-
OY lipses of Figure 3(8.4.4).
=k sy Lex Tet a Victetys\ ey)
(v) Calculation 5: Loci of Constant Hue and
eV 42) — 1 + 9G Chroma. Schrédinger (1920b), as discussed in
Section 8.4.1, put forward an interesting assump-
oY tion by which the distribution of hue and chroma
oad Ci SS Lt hae Yoeta x =y)] in a surface of constant brightness (luminance)
might be derived from the line element. On this
ePS~43) = 1+ 9B assumption, in a surface of constant brightness,
the loci of constant hue are the geodesic lines
oY
Tape pels ap leep ante Gcwc 11K starting at a specified point representing the
achromatic color stimulus and terminating on the
points of the spectral locus or the line of purples;
[28(8.4.4)] loci of constant chroma are assumed to be geo-
desic circles centered at the achromatic point. The
and when x and y are eliminated from these
computation of these geodesic lines and circles
equations, the equation of the uniform-chromatic-
from a line element given in the form of Eq.
ity-scale surface is obtained as
23(8.4.4) is not particularly simple, but it can be
Ayer no) a A eV *2~ 42) nie Agel?) = Ay
carried out economically with the aid of a com-
puter.

[29(8.4.4)] The general equation of a geodesic line in a ~

Riemannian space (x = x', y = x”) is given by
where
(1, j,k =1,2)
Ay = ta, 33 — t43¢3)

Ay = t13031 — tyyl33 [30(8.4.4)]

Ay = tyto3 — ti3tay
where ds denotes the line element and the quanti-
Ay = 9Y (tA, + ty An + 1573) ties { ;",} are Christoffel 3-index symbols of the
second kind (Sokolnikoff, 1951). The Christoffel
+A, tiAg tA; symbols stand for certain functions of the metric
coefficients and their partial derivatives (see be-
The origin of the (X,, X,, X;)-coordinate system low). (Note that in mathematical space theory,
 Line Elements of Color Space 671

both upper and lower suffixes are used to dis- 1 dgy5

12 2b
tinguish components. To prevent confusion with Hee 2 Ox}
power indices, any quantity raised to a power is
enclosed in parentheses.) 23,1] = 08x 1 98>
If 0x? 2 9x!

Seta at howe wae dy _ dx? 1 0255

7 ake Sas mea 2232) \= =

It follows from Eq. 23(8.4.4) that and the quantities g'', g'*, g7!, g as
2 2
gi(V) + 2gpAW + gy(V) = 1 PRIME oa weges Ba 9 ae 22 _ $11

[31(8.4.4)]
with
and from Eq. 30(8.4.4) that
i 2

ak iontsa[ jre[ssjari=a
§ — 811822 —(g12)

A complete family of geodesic lines radiates

from any particular point (xj, x¢) representing,
oy {? Vo?+a{? bars {2 |ory? Bs for example, the chromaticity (x9, y)) of the
achromatic color stimulus. The lines of this family
have different initial directions specifiable by an
[32(8.4.4)] angle 0 of tangent A%,/A’). An initial direction 6
having been chosen, A‘), A% are taken so that the
The Christoffel symbols are defined as follows: condition laid down in Eq. 31(8.4.4) is satisfied. A
small step ds along the tangent of the geodesic
{1} satu + g?[,2 line terminates in a second point (x}, x7) of the
geodesic line given by

on = g7'[11,1]
+g” [11,2] a as 5 ‘x24 4, 58

At this point, the new quantities \'), A; have val-

fest = gh [12,1] + g?[12,2] ues

dX
L2]
dy
= g71(12)1] 977112,2] int, Vile oF ie N= Ny + — 8s

ee = g''[22,1] + g'? [22,2] where dA',/ds and dd>/ds are found from Eq.
32(8.4.4) by substituting A, for A. This step-wise
procedure is repeated to obtain points
ions = g7![22,1] + g”[22,2] {ORO NE AT ty, i x2)
eee X3,X35°° n? n

where the Christoffel 3-index symbols of the first and corresponding tangent vectors
kind [ij, k| are defined as
Woe tat eet IS 112
(A',, 2 ited etaaare eg,
tal 028i
Phi} <5 =r
until the boundary of real chromaticities, that is,
the spectrum locus or the line of purples, is
[11,2] = OZ _ ah 0811 reached. Figure 7(8.4.4) shows the results of such
dx! 2 Ox? computations for several geodesic lines starting
from x, = 0.310, y = 0.316 (CIE standard illu-
Py
N07;
she we
minant C) in various directions. The computa-
tions (assuming Stiles’ line element) were made
672 THEORIES AND MODELS OF COLOR VISION

K ee
\\
WerE

Sess
SST

0.6 0.8

Fig. 7(8.4.4). CIE 1931 chromaticity diagram showing a network of geodesic lines and circles,
derived from Stiles’ line element, which resemble lines of constant hue and chroma, respec-
tively.

with ds = const = 0.005. Points on these lines
which are away from (Xo, yo) by the same num-
ber of steps 6s lie on a geodesic circle. Several
such circles are shown in Figure 7(8.4.4). These
derived geodesic lines and circles resemble the
lines of constant hue and chroma, respectively, set
up empirically in the Munsell system (Section
6.6.1). This remarkable resemblance supports, in
a general way, the assumption that hue and
chroma can actually be described by means of
geodesic lines and circles. However, quite large
discrepancies are noticed between the chroma
lines of the Munsell system and the geodesic
circles derived from Stiles’ line element, and simi-
larly for the hue lines. How far these discrepan-
cies are connected with the observing conditions
used or assumed in the Munsell data and the line
element, or with observer variations is not clear.
8.4.5 Trabka
Trabka (1968a, b) offers an interesting interpreta-
tion of Stiles’ line element by applying basic
concepts of signal-detection theory. Specifically,
Trabka proposes that each cone mechanism is a
set of similar nonideal radiation detectors whose
signal-to-noise ratio outputs can be derived from
standard methods of detection theory:
Signal-to-Noise Output
(Sea OS Ou We aba ee
aheae
where m is the average rate of absorption of
quanta by a given cone mechanism, and (dm)
corresponds to the difference of the stimuli to be
discriminated. The constants b and c are func-
tions of the “dark current” m, and the “dead-
time” 7, and given by
ler tie mg
Oe: po ang EO PO ee
The “dark current” is very much equivalent to the
“self-light” (Eigenlicht) observed in vision and
already considered by Helmholtz in his line-
element work. The “dead time” is also a reason-

able concept in vision, as it relates to neurophysi-
ological evidence that after the absorption of each
quantum producing a signal, there is a brief period
of time during which no further absorption will
contribute to the signal. Estimates for the dark
current and the dead time are
My = 5 spikes/sec and t= 7 msec
Trabka defines the line element for a collec-
tion of N such nonideal detectors by assuming
that the greater the signal-to-noise fatio, the
greater the sensation of change at the detector
outputs, or the greater the “distance” between the
color stimuli; thus,
N 2
F,(.dm,
j=1 m?+2bm,+c¢
[2(8.4.5)]
where a is a scaling constant. Equation 2(8.4.5)
defines the line element in terms of the sum of the
signal-to-noise ratios at the outputs of nonideal
detectors which are identical except that they may
absorb different average numbers of quanta [i.e.,
the functions m; = m,(A) may be different func-
tions of wavelength] and may have different
amplification mechanisms (i.e., weighting factors
F).
Trabka shows that the quantities (1/p)’,
(1/y)?, (1/B)? in Stiles’ line element correspond
to the weighting factors F., thus making the limit-
ing Weber fractions p, y, B inversely proportional
to the square root of the quantum detectivities of
different amplifying mechanisms. Trabka also
shows that the denominator of Eq. 2(8.4.5) corre-
sponds with Stiles’ standard-shape function §(x)
representing the threshold-versus-radiance (t.v.r.)-
curve of an individual foveal mechanism (x), as
given in Table 1(7.4.3) and illustrated in Figure
1(7.4.3). The experimentally determined (t.v.r.)-
curve of Stiles agrees very closely with the expres-
sion
l (c) The signals emerging from the second zone,
(x) =
9(ax? + 2bx +c)!”
[3(8.4.5)]
when its coefficients are set at a = 1, b = 0.3, and
c = 0.0125.
Trabka showed that his version of Stiles’ line
element gave virtually identical predictions of ex-
Line Elements of Color Space —_ 673
perimental data to those made by the original
Stiles’ line element. He also added predictions of
wavelength discrimination data and luminous ef-
ficiency functions of tritanopes, deuteranopes, and
protanopes. His predictions agree reasonably well
with experimental data.
Trabka considered Stiles’ line element, and his
own version of it, only as a first-approximation
model to what the visual mechanism actually
might be. The work by Boynton, Ikeda, and Stiles
(1964) on the interactions among chromatic
mechanisms as inferred from positive and nega-
tive increment thresholds has strongly indicated
that the three cone mechanisms do not operate
independently as assumed in Stiles’ line element.
A dependence between them can be allowed for
by introducing in the line element cross-products
of pairs of the three cone responses. According to
Trabka, such cross-products could be generated
from radiation detector theory, but this intriguing
possibility has not as yet been followed through
in detail.
8.4.6 Vos and Walraven
Perhaps the most elaborate line element that
has been developed so far is that by Vos and
Walraven (1972a, b). It is also one of the latest of
complete versions of line elements meeting the
criteria of a line element outlined earlier in Sec-
tion 8.4.1.
The basic assumptions underlying the Vos—
Walraven line element are the following:
(a) Color stimuli are processed in two stages,
occurring in a Helmholtz-type three-receptor zone,
followed by a Hering-type neural conversion zone
in which a luminance and two antagonistic chro-
matic signals are formed.
(b) Color discrimination is essentially pho-
ton-noise-limited. At higher luminances, “satura-
tion” and “super-saturation” (channel overload-
ing) processes occur that keep color discrimina-
tion below the limit that photon noise alone would
provide.
the neural conversion zone, combine in a positive
and definite (quadratic) form to generate an out-
put signal directly proportional to a just-notice-
able difference between two given input stimuli.
Vos and Walraven consider first the case in
which the visual mechanism is stimulated by color
stimuli of low luminances. In the cone-mechanism
674 THEORIES AND MODELS OF COLOR VISION

zone, the signal-to-noise output can then be be as follows:

treated in terms of standard Poisson statistics, that
1S, First chromatic signal:

Signal " (dN) F=R/G (red-green)

ouput /N [1(8.4.6)]
ee
Second chromatic signal:
where WN is the average number of quanta ab-
S=(R+G)/B=Y/B (yellow-blue)
sorbed per second. If one were to assume that the
collective response of the three cone mechanisms
would be equivalent to the square root of the sum Achromatic signal:
of the squares of the individual cone responses,
L=R+G+8 (luminance)
one would obtain the line element

wr -(
[3(8.4.6)]

dG +
2 2
wwe PLN ‘ |
+{—y] The signals, F, S, L emerging from the (red-green),
/G@ /B (yellow-blue), and luminance channels are then
assumed to combine in the positive and definite
[2(8.4.6)] form

where R,G,B are the responses of the three

(ds) =n,{ 2) tar( 2) +ns(S)
independently operating cone mechanisms. The
spectral response functions of the cone mecha-
nisms are linear transformatins of the Judd-mod- [4(8.4.6)]
ified CIE 1931 standard color-matching functions
to generate an output signal ds that is directly
x'(A), y(A), Z(A) (see Sections 5.5.2 and 5.14.4).
proportional to a just-noticeable color difference
The line element defined by Eq. 2(8.4.6) clearly
between two given input stimuli. In Eq. 4(8.4.6),
differs from that of Helmholtz [Eq. 4(8.4.2)] but is
the quantities o,,0,, 06; are signal uncertainties
similar to that of Schrodinger [Eq. 1(8.4.3)] except
and 1,;,%r,Ns5 are constant weighting or gain
for Schrodinger’s luminance dependent factors
factors in the L, F, S channels.
Tp/(lpR +16G+1,B), Ilg/(UrR + IoG + 1pB),
The line element in the form of Eq. 4(8.4.6)
and /,p/(lpR+/¢G+1,B). These factors pre-
can be converted to a corresponding line element
serve Weber’s law from low to high levels of
_ in which the variables are the cone responses
luminance and make “brightness” a linear combi-
R,G, B. This conversion, though elaborate in ex-
nation of the cone responses. The line element of
ecution, follows standard mathematical proce-
Eq. 2(8.4.6), on the other hand, does not comply
dures (see, for example, Section 5.4.2). Briefly, if
with Weber’s law: the Weber fraction increases
the original line element (Eq. 4(8.4.6)] is denoted
with the square root of the luminance instead of
being constant, in accordance with the now well- by
established transition (de Vries—Rose behavior) 3
from absolute-threshold behavior to Weber be- (ds )ovigee teGe Eka ee
Pret
havior (Bouman, 1961; Barlow, 1965; van Nes
and Bouman, 1967).
and similarly, the new line element is of the form
The important new concept introduced by Vos
and Walraven (1972a) to the notion of the line f 3
element is that the output signals of the individ- (ds new = dy Cy dx; dx;
ual cone mechanisms generated in the first stage ii
are not combined in the manner described by Eq.
2(8.4.6), but instead are first coded in the second where the original coordinates x, (i = 1 to 3) are
stage, forming two antagonistic chromatic signals related to the new coordinates x; (j = 1 to 3) by
and one achromatic (luminance) signal, before a means of given (linear or nonlinear) transforma-
difference signal is transmitted to the brain. The tions [Eq. 3(8.4.6)]
coding (transformation) of the R, G, B cone sig-
nals in the neural conversion zone is assumed to x; = x;(x/)
 Line Elements of Color Space 675

then, the new coefficients cj, are related to the

original coefficients c;, in accordance with +2 dR dB( KGB
=X __ 4.9
3 eS Ns
ql ik , , Ra Gass B
i eee LOSK OX

and the original differentials dx; relate to the new +2 dG dB( ea LeaGee aey
differentials dx; as follows:

3 Ox’
|
R+G+B 6(8.4.6
Clerks
dx;i eee= 2 yy J Xi
With R+G+B=L, R/L=r, G/L=g, and
In applying the above procedure to Eq. 4(8.4.6) B/L = b, Eq. 6(8.4.6) becomes
making use of the transformations given in Eq.

|
3(8.4.6), one obtains as the new line element
g b
(dR+dG+ dB) (ay = ; (ary'( + IF a

r(r+g) Isp +g
GaP
2 —

+ (dG ‘(> 7 esb

(GdR—RdG)
TF (4G) ne Pe(rt+g) “rt+g
RG(R+ G)
(BdR+BdG-—(R+G) dB) 2 ae
+n 5 B(R+G)(R+G+B)
[5(8.4.6)]
+2(dR)(aG)( 9 ’ eae
Fr+g
ee ee
ealiie
The above expression is readily expanded and the
terms regrouped to obtain the line element in the
familiar form: +2(dR dB + d6-4B)(m, ~ 9)
2 —=
2 >
Uhe + ——
npG
dias ieik) esas: R(R+G) It is important to note that the weighting factors

pion tials 4
(R+G)(R+G+B)
Nr» NF, N;,, mtroduced by Vos and Walraven in
Eq. 4(8.4.6) are crucial modeling quantities. For
the special case of n, = n-=7, = 1, the line
element of Eq. 6(8.4.6) reverts to that of Eq.
2 NL npR 2(8.4.6) which characterizes a discrimination pro-
apes, |Reba B G(R+ G) cess at absolute threshold in which the three cone

pecs mB 84
mechanisms respond independently of one
another; the cross-products of Eq. 6(8.4.6) vanish.
(R+G)(R+G+B) In an effort to extend the applicability of their
line element defined by Eq. 6(8.4.6) to stimuli
of moderate and high luminances, Vos and
7 (aB)*| KetyG
BE BD deteia( Walraven (1972b) introduce a “saturation” and a
“supersaturation” model along the lines proposed
earlier by Bouman, Vos, and Walraven (1963).
n,(R+G) |
“Saturation” occurs when the spike density of
B(R+G+B) the neural signals increases to a level beyond
which the incidence of new spikes is prevented by
NL a le the inherent refractory period (dead time) that
+2 dR ac aos R+G follows each spike (see also Trabka’s line element
in Section 8.4.5). A simple mathematical model of
too | the saturation process relates the number of spikes
(R+G)(R+G+B) vy with the number of incident events, such as
676 THEORIES AND MODELS OF COLOR VISION

quanta n, as follows: for which supersaturation occurs. The relation

between n,) and n, is independent of the cone
n
[7(8.4.6)] mechanism involved and is given by
"T+ n/no
2
Ane Oe)
where 7, denotes the number of incident events a3= z [10(8.4.6)]
1 0
(quanta) for which saturation occurs. In this
model, it is assumed that the refractory process For example, an inaccuracy in the refractory
has a fixed time interval t, throughout which the period fy of 35% (i€., 9,,/to = 0.35) is equal to
prevention of new spikes occurring is complete, No /n, = 0.5 (approximately).
but at the end of the dead time the possibility of The basic steps of derivation of the general
new spikes occurring is at once restored. When form of the Vos—Walraven (1972a) line element
the number of spikes approaches its maximum, are given as follows:
which is determined by ny, the congestion in the
neural channel increases, leading to a gradual (i) The cone signals R, G, B first add to
reduction of the “accuracy of information on n.”
Vos and Walraven give a formula that describes L=R+G+B8B8 and Y=R+G
the postulated “accuracy reduction” (11) The new signals saturate in accordance with

of = n(1 +7) [8(8.4.6)] gy ecg he

1+n/no
No

For high luminances, that is, high quanta- (iii) The accuracy of signal transmission is
incidence rates, the quantity o becomes propor- governed first by Poisson statistics, then limited
tional to n. by saturation and supersaturation in accordance
with
When they introduce their saturation model
into an extended form of their line element of Eq. 0, = yn(1 + n/ng + n>/n;)
6(8.4.6) and then use the extended form to predict
a variety of experimental data, Vos and Walraven
which provides what Vos and Walraven call the
find that there is still room for further improve-
“physical line element”
ment. This improvement, so they argue, may be
obtained through a refinement of their rather 2
simple and possibly not realistic saturation model (ds); = ee
defined by Eqs. 7(8.4.6) and 8(8.4.6). Vos and n(1 +n/ng + n>/n?)
Walraven postulate that the refractory period f, is
“loaded with an inaccuracy +o,,, which makes (iv) After completion of steps (i) to (iii), the
interpretation of small variations in spike density signals of the two antagonistic chromatic chan-
void of sense, once saturation occurs.” This nels are formed in accordance with
changes Eq. 8(8.4.6) to
VR
F=— and
3 VG VR
Da
=n +242 [9(8.4.6)]
Qo. -n; The general form of the line element derived
in accordance with steps (i) to (iv), but expressed
where n, denotes the number of incident quanta in terms of the initial cone signals R,G, B, is
then given by

2 2
G B
(ds)” = (dR) nifi+ nefe@(1 +E]
0
+n (1+ 2)
)

Y 2
2
R B :
+ (ac)? nife + nefeR?(1 +2 | +n f6(1+2 | z3 (dB) nift + nh(I i 7)
0 0 0
 Line Elements of Color Space 677

+2 dRdG a R= G Wf ck SEN
0

Yi
+2 dR dB
|,f,— 1 68 +e )r(i +F)| +246 aB|n,f,— 1/8 |+2) y(i+$)]
0 0 Bo Yo

[11(8.4.6)]

B x
1/4--B/B, +B? /B} Loy yw Way,
2
B(l a7 B,) Y(4+ Y/%)
-(1+ Y/¥% + ¥°/Y?) - (1 + B/B, + B?/B})YB
1 B/By eB /B; Rey /y,,-re /Y,

B R
ae L G/Gg+-G./G; TtR/R, + R°/R?
fe
2
G(1 — G/G) R(1 + R/R,)
-(1 + R/Ry + R?/R}) - (1+ G/G, + G?/G})RG
1+ G/G,+G?/GF 14+ R/R,+ R?/R?

and

]
tie
EAC
The derivation of the general form of the
Vos—Walraven line element given in Eq. 11(8.4.6)
is not as straightforward as the low-level line
element given in Eq. 6(8.4.6). Equation 11(8.4.6)
has been reproduced from the original publica-
tion of Vos and Walraven (1972b) and has not
been verified by us.
Vos and Walraven (19720) test their line ele-
ment by using it to predict a variety of experi-
mental color-discrimination data. In general, they
find reasonable agreement between prediction and
experiment. Figures 1(8.4.6) and 2(8.4.6) illustrate
some of the predictions made for wavelength dis-
crimination and for chromaticity discrimination.
In each of the two cases, three luminance levels
(L = 1,10, 100, in arbitrary units) were used, the
inaccuracy of the refractory period ft, was set at
6,,/to = 0.35 (= 35%), and the gain factors
Nr» Nr, 7, Of the achromatic and the two chro-
matic channels were adjusted so as to provide
complete equivalence between the (R/G)- and
(Y/B)-channel for low to medium values of L.
8.4.7 General Construction of Inductive Line
Elements
Insofar as empirical line elements, such as those
of MacAdam (1942, 1943), correctly reproduce
Lan)
the standard deviations from the mean of re-
peated color matches (Section 5.4), they make no
assumptions about the visual response mecha-
nisms responsible for color discrimination. On
the other hand, the inductive line elements of
Helmholtz, Schrédinger, Stiles, Trabka, and
Vos—Walraven (Sections 8.4.2 to 8.4.6), that make
very simple assumptions of this kind do not
account for important features of the experimen-
tal data embodied in the empirical line elements.
Various attempts to develop more satisfactory
inductive line elements have been made; the fol-
lowing notes indicate the direction of this work.
Inductive line elements derive from the follow-
ing main guiding principles:
(a) that the perception of a difference between
two stimuli of spectral radiant power distribu-
tions, {P, dA} and {Px dA}, respectively, de-
pends on the possibilities of this difference being
detected by one or more of several component
visual mechanisms that are in some sense distinct
although not necessarily independent in the
highest possible degree, and
(b) that if the quantitative relation between the
two stimuli at threshold is expressed in terms of
678 THEORIES AND MODELS OF COLOR VISION

w (o)
Deuteranopia Protanopia
L=l L=!0 L=| L=|00

fs)

)
units
(arbitrary fo}
)
units
(arbitrary

-
Just
Noticeable
Difference
Wavelength Difference
Wavelength
Noticeable
Just-
500 400 500 600
Wavelength \ (nm) Wavelength A( nm)

8 Tritanopia 100 Normal Trichromatism

3 )

8 8
3 fe)

)(arbitrary
units )(arbitrary
units
1%) 19.

500 500 600 700 400 500 600 700

Wavelength \ (nm) Noticeable
Difference
Wavelength
Just- Wavelength dA (nm)
Noticeable
Just-
Difference
Wavelength
Fig. 1(8.4.6). Predictions of Vos—Walraven line element of wavelength discrimination curves
for the three types of dichromat and for the normal trichromat at low, moderate, and high
luminance levels (Vos and Walraven, 1972a, b).

their tristimulus values, The fundamentals [7(A), g(A), b(A)] are nor-
mally identified with the relative spectral absorp-
tions (corrected for light losses in the optic media)
R=[P\r(A)dd, RP’ = [Pxr(A) dd of three photosensitive pigments contained in the
response mechanisms and mediating all visual
effects of light on the retina (Sections 8.2.5 and
G= [P,g(A) A, G’ = [Pxg(A) dd 8.2.6).
The quantitative response of a particular com-
ponent mechanism—say the ith—to the stimulus
{P, dX} may be represented, to take a very gen-
B= [P,b(A) da, B’ = [Pxb(A) ad eral case, by a function of the tristimulus values,
F;(R, G, B). According to the Taylor expansion,
[1(8.4.7)] the difference of response to slightly different
stimuli {P, dA} and {Px dX} is given to the first
approximation by
based on a certain unique set of color-matching
functions [7(A), Z(A), b(A)], known as the funda- AE(R, GIB) =ARMf CR GAB)
mentals, then the relation in question can be
displayed in a particularly significant form which +AG fco(R,G,B)+AB f,(R,G,B)
makes the roles of the component mechanisms
apparent. [2(8.4.7)]
 Line Elements of Color Space 679

0.8

Fig. 2(8.4.6). Predictions of Vos—Walraven line element of chromaticity-discrimination

ellipses for low, moderate, and high luminance levels (Vos and Walraven, 1972a, b).

where
Provided the component mechanisms are suffi-
ciently distinct, it will be meaningful to consider
pore alana 2)
the response to the stimulus difference when only
the ith mechanism is operating, and to take as its
= fig(R,G, B) quantitative expression the same quantity AF..
Using the threshold concept in its crudest form,
the stimulus difference will be detected by the ith
= fir(R,G,
B) [3(8.4.7)] mechanism if AF,, taken initially to be positive,
exceeds a threshold value 7,I which, in general,
AR=R’—-R, AG=G'—-G, AB=B’-—B will vary with the mean response level,
680 THEORIES AND MODELS OF COLOR VISION

F(R, G, B), that is, and, given the quantitative properties of the
component mechanisms specified by the func-
T, = T;{ F;} tions F(R,G,B), T(F) and p,(AF/T,), all
sets of values of AR, AG, AB that correspond
and perhaps also with the response levels of the to threshold perception are determinable. At
other mechanisms; that is, threshold, the values of

= GR
By he}

In the latter case, distinctness of the mechanisms

is taken to imply that it is possible, in principle,
to study the behavior of the ith mechanism acting
alone by blocking the differential responses AF,
(7 +i) of the other mechanisms without interfer-
ing with the possible effects of their main re-
sponses F; (j #i) in determining the threshold
TFs Ey)
On the more refined probability definition of a
threshold, the chance of detecting the stimulus
difference will be a function of AF,, that is,
pi(AF.), which is zero when AF,=0 and ap-
proaches unity when AF, is large in comparison
with the threshold value 7; [which is, by defini-
tion, such that p/(7;) = 0.5]. Provided the stimuli
{P, dX} and {Px dA} are being presented for
discrimination under symmetric (or quasi-sym-
metric) conditions of comparison (Section 5.2),
the probability function p/(AF,) must follow the
same course for negative and positive values of
AF., and the threshold 7, will have the same
absolute value (but with opposite signs) in the
two cases.
It is convenient to change the argument of the
probability function from AF, to AF,/T;,, to ob-
tain
Ww ee P
Pry =P AAR)
This ensures that for all response levels F,, one
has p;(1) = 0.5, although the exact shape and
particularly the steepness of the probability curve
may still depend on F..
If the probabilities of discrimination by differ-
ent component mechanisms combine indepen-
dently, the probability P that the stimulus
difference will be detected by at least one mecha-
nism when all are in operation is given by
Bs PAC py CE pant Oh a a)
[4(8.4.7)]
where WN is the number of mechanisms. The dis-
crimination threshold is reached when P = 0.5
satisfy
0-5 =" [la pie) ie oe
[1 —py(év)] — [5(8-4.7)]
If the €,’s are very small relative to one particular
€., for example, €98 then Eq. 5(8.4.7) will have the
solution
é=1-—a,, §=a, [6(8.4.7)]
(with i #j) where the positive quantities a; and
a, are all small compared with unity and become
vanishingly small as the ratios £;/; (i #j) tend
simultaneously to zero. This follows from the fact
that p,(€;) increases monotonically with €,; for
£,= 1 we have p;(1) = 0.5, and when &; tends to
zero, p;(§;) tends to zero.
The so-called sum-of-squares relation
OF)" (45\" (Abe
ie 1, Tat oe
or
ge 4+ e34---+62=1 [7(8.4.7)]
obviously has a solution that can be written in a
similar form to Eq. 6(8.4.7):
> aul sod. €,= a‘ [8(8.4.7)]
(i #7) where a; and a; become vanishingly small
as the ratios ;/€; (i #/) tend to zero. Clearly, as
any of the limiting cases just considered is ap-
proached, the solution of the sum-of-squares rela-
tion is an increasingly good approximation to the
solution of the strict threshold condition given by
Eq. 5(8.4.7). Theorists, following Helmholtz, have
commonly adopted the same sum-of-squares rela-
tions as an approximation valid at threshold when
no one of the ratios AF./T, has predominantly
the largest value. The divergencies between the

threshold values of the ratios A F,/T, derived using
the simplified relation [Eq. 7(8.4.7)] and the val-
ues obtained using the rigorous probability rela-
tion of Eq. 5(8.4.7) depend on the shapes of the
probability curves p,(--- ).
To relate the p,(--- ) curves with experimental
probability-of-detection curves, consider the de-
termination of discrimination thresholds of the
kind sometimes known as general color limens
(Wright, 1941; MacAdam, 1942) in which the
tristimulus values of one stimulus are fixed at
R,G, B while those of the other stimulus (R +
AR,G+ AG, B + AB) are varied along a straight
line in tristimulus space passing through the point
(R,G, B) [see Figures 1(5.4.1) and 1(7.10.5)].
Thus, AR, AG, AB are kept in the same ratio
although their actual values have changed; we
have
AR=ls, AG =ms, AB=ns
where the direction cosines /,m,n define the
fixed direction of the line and s the variable
distance along it measured from (R, G, B). The
probability of detection P in such a determination
is a function of s/S, that is,
eae
where S is the threshold value of s such that with
s=S
Pl = P(1) =0.5
The shape of the probability curve P(---) is
determinable experimentally for various tristimu-
lus points (R, G, B) and directions (/, m, n).
By reference to Eq. 4(8.4.7) for P in terms of
the p,’s, it can be shown that even though the
p;(---) curves for all component mechanisms
have a common shape that is independent of their
response levels, F;(R,G, B), the resultant proba-
bility-of-detection curve P(s/S) will have strictly
the same shape as the p;(---) curves only if one
of the ratios AF,/T, is large compared with the
others. When this is not so, the P(---) curve
would be somewhat steeper, although the dif-
ferences arising in this way are not large. On the
whole, experimental P(---) curves show little, if
any, systematic dependence on (R, G, B), and the
assumption of a common shape for p,(- - - ) curves
independent of response level is perhaps admissi-
ble as a working approximation.
Line Elements of Color Space 681
Regardless of any differences in shape of
p;(-+-) curves, the bearing of the actual mean
shape on the acceptability of the sum-of-squares
threshold condition given by Eq. 7(8.4.7) is il-
lustrated by considering just two component
mechanisms. The solutions of
7 2
= (= + (2 = §? + £2?
are conveniently summarized in the diagram of
Figure 1(8.4.7) which shows, as the continuous
curves, log , and log €, plotted against
log(€,/&,). The three pairs of broken curves rep-
resent similarly the solutions of
0.5 = [1 —p(é,)JL) — p(é)]
corresponding respectively to the three probabil-
ity-of-detection curves plotted in Figure 2(8.4.7).
All these p(&)-curves are of one of the types
commonly used to fit to experimental probability-
of-detection curves, and are plotted from the for-
mula
p(é)=5[1 + B(hlogé)] —forg>1
= 5[1 e(atos +) for & < 1
[9(8.4.7)]
with different values of the constant h which
controls the steepness of the curve. At p(€) = 0.5,
the gradient dp/dlog € equals approximately
0.565 h. Here E(x) is the error function; that is,
[10(8.4.7)]
The sum-of-squares solution [Curves A, Figure
1(8.4.7)] gives a rather close approximation to the
predictions of the more rigorous probability
method if the steepness of the probability-of-
detection curve, for a single mechanism, corre-
sponds to h = 2.56 [Curves C, Figures 1(8.4.7)
and 2(8.4.7)].
The shape of the probability-of-detection curve
P(s/S) in a determination of a general color
limen when both mechanisms are in operation is
not representable exactly over its whole range by
an expression of the form of Eq. 9(8.4.7), but the
 ‘5°pay “hn ears CURVES A
fi Ry B—--0o--
a 7 x N| 7 x \
"Sa 70-2 id a
NI N Ny c pee faery

2 aay exit” ior este me

Gd hn Pe eh toe\
7 \
re) / / ‘N \
a ? wey \ [e}
a> -03 Bh/ Me / \‘a hehy\
2 r4 ee Sf X\

/ 7 z
log <: ef if

A “iy / Fase
-04
‘7
iJ f
7
/
Jo / Vag
/Y / \\

0.5 Be ra
-08 =06 =0.4 -0.2 0 0.2 0.6 08

oo (&4/,
Fig. 1(8.4.7). Curves showing pairs of values of £,,&, that satisfy the sum-of-squares relation
1=£{+ € or the probability relation 0.5 = [1— p(€,)][1— p(é,)].

1.0

0.8 |

0. o
p(é)

0. BSS

0. nm

hoe
<0 -06 -04 -0.2 0 0.2 04 0.6 08
log €

Fig. 2(8.4.7). Probability-of-detection curves: Curve B, h = 3.84; Curve C, h = 2.56; Curve D,

h = 1.92; Curve E, hol:

682
 Line Elements of Color Space 683

Table 1(8.4.7) Values of Con- steeper, and in some cases, much steeper. How-
stant h of Eq. 9(8.4.7) Which Con- ever, even if the sums-of-squares relation is not a
trols the Steepness of the Proba- particularly good approximation, its use is un-
bility-of-Detection Curve p(é), for likely to obscure other salient features of pro-
Different Numbers (NV) of posed systems of component mechanisms, and, in
Component Mechanisms providing a Riemannian line element, it makes it
N A pe easier to handle the quantitative, if approximate,
theory. It is of interest to note that the above
rd 2.56 37 discussion could have been based on a Poisson
3 2.43 6 2 distribution function, instead of the Gaussian dis-
4 2.35 3.45 tribution function of Eq. 9(8.4.7), to represent the
5 Lae Sead
6 2.24 3.56
typical shape of probability-of-detection curves.
Substantially the same general conclusions would
) This value gives exact agreement follow.
between probability and sums-of-squares For the simplest class of component mecha-
solutions when &, = §, =+-* = &y.
nisms, the response function depends on one
(2) This is the maximum value of hp which
arises when £, = §& =--- = &y.
tristimulus value only. For example, if this tri-
stimulus value is R, we have

F(R,G,B) = F(R)
latter gives a good approximation in the central,
steeply ascending part of the curve if the constant and
h is suitably chosen. The appropriate value (hp) OF af OF
depends on the relative contributions of the two SEE yess =e)()
ar fik(R), BG >aR
mechanisms to the particular general color limen
under consideration. If one mechanism pre- If, also, the threshold 7; depends only on the
dominates, hp becomes equal to hy (the value for response of the same mechanism, then
a single mechanism), but is increased to a maxi-
mum hp= 1.24 hy when the two mechanisms
contribute equally. The probability-of-detection
T.LF(R)] = 4,(R)
curve for the latter case, with hy = 2.56, is plotted is a function of R only. The contribution of such
as Curve C in Figure 2(8.4.7). a mechanism, which may be named a Class I
(simple) mechanism, to the sums-of-squares ap-
If there are more than two component mecha-
proximation to the threshold condition is a term
nisms (N > 2) each with the same shape probabil-
ity-of-detection curve of the type defined by Eq.
a= (48) (ary?
2

9(8.4.7), the value of the constant h for which the

solutions of the probability-threshold condition
[Eq. 5(8.4.7)] accord best with the solutions of the
sums-of-squares condition [Eq. 7(8.4.7)], is shown
For the discrimination properties of the mecha-
nism, only the ratio
as h“!) in Table 1(8.4.7). The third column of the
table defines the steepness of the probability-of-
detection curve in a determination of the general
color limen when A for each of the N mechanisms
fs= 4,(R)
has the value of A = The minimum value, when
is important, and the same function 7,(R) may
one mechanism makes a predominating contribu-
result from:
tion, is simply hp = A“.
The application of the sums-of-squares thresh-
(a) A response function
old condition [Egq. 7(8.4.7)] in an inductive line
element of the kind here being considered is
strictly justifiable only if the steepness of the F(R) =Cf"¥,(R)dR (C= const)
probability-of-detection curves obtained experi-
mentally is consistent with the range of hp values together with a constant threshold
just derived. Some experimental curves lie in this
range, but more generally they are somewhat T(R)=C
684 THEORIES AND MODELS OF COLOR VISION

(b) A response function leading to the threshold condition

F(R) = CR 1 =|ere ray|(ORY

together with a threshold that varies with the
response F,(R) according to the law
B ; 2
T,(F) = (
¥;(R)
i:|
KG F by) |pete

(c) Response and threshold functions that parti-

tion the R variation of ¥,(R) in a more com- leet
plicated way, subject only to the condition that
[14(8.4.7)]
aF(R) 1
WR)= "OR TIR(R)] The quantities a, B, y, a, b, c, and do, bp, Co, are
all positive constants. If, in particular, aa, = |
These different possibilities cover the various (or bby)= 1, or cco= 1), the response becomes
conceivable physiological models of a Class I zero when R=0O (or G=0, or B=0). Apart
(simple) mechanism leading to the term from the symbols used for the constants, Eq.
[W,(R) AR]? in the line element. With three Class 14(8.4.7) is substantially the Stiles line element
I (simple) mechanism dependent respectively on discussed in Section 8.4.4.
the three tristimulus values, the line element has It may be noted, at this point, that the as-
the form sumption of an exact logarithmic form for the
response functions in this and other line elements
is not generally a critical feature. Over limited
1 = [¥,(R) (AR) + [¥.(@) (AG) ranges of the variables, other simple functions
behave in much the same way as the logarithm
+ [y,(B) (4B) [11(8.4.7)] and may, in fact, provide closer approximations
to empirical data. Figure 3(8.4.7) compares a
and the discrimination ellipsoids in R, G, B space logarithmic with a power-law response function.
corresponding to different sets of values of The respective terms in the Schrédinger line
(R, G, B) all have their principal axes parallel to element (Section 8.4.3)
the coordinate directions, the length of the prin-
cipal axis in the R direction being the same at all a(AR)
As,
points (R,G, B) with the same R value, and
similarly for the other principal axes. The un-
eo R(aR + BG + yB)
satisfactory line element of Helmholtz and its
modified version proposed by Stiles (Section 8.4.4) B(AG)°
are of this type and are derivable (omitting some G(aR + BG + yB)
unessential refinements) from a_ physiological
model with response functions
y(AB)’
B(aR + BG+ yB)
F(R) =aln[a(R+ ay)]

F,(G) = Bln[b(G + b))] [12(8.4.7)]

[15(8.4.7)]
are not attributable to any Class I (simple) com-
F,(B) = yln[c(B
+ co)]
ponent mechanisms. Each term corresponds to a
and with constant thresholds mechanism with a response function dependent
on one tristimulus value only, but with a threshold
LCE YR y, ete) aK ae ke dependent on both its own response levels and
those of other mechanisms. Such mechanisms may
[13(8.4.7)] be termed Class I (complex). Possible models
 Line Elements of Color Space 685

F(R) ——— 0.4156 R@>"5!

F(R)
log,, (R+1)

Fig. 3(8.4.7). Similarity of shape of a logarithmic and a power-law response function.

leading to Schrédinger’s line element are: A F, has to be detected above a noise level attrib-
utable mainly to the response in the ith mecha-
(i) Model A. nism but which is also raised by responses in
other mechanisms. Model B with different re-
MUR oR FENG) —pG, F,(B) = 7B sponse functions corresponds to a common shared
noise level. Schrédinger himself, it may be noted,
T, = T(F,,.5,5) derived his line element to yield a luminous ef-
ficiency curve that was a linear combination of
=[F(R){F(R) + F(G) + B(B)}]'7 the color-matching functions.
Component mechanisms of Class II or Class
= [aR(aR + BG + yB)]'” III, that is, those whose response depends on two
or three tristimulus values, are physiologically
]'”
T, = [BG(aR + BG + yB)
more sophisticated in the sense that they contem-
plate either retinal end-organs containing more
than one visual pigment or some interplay, usu-
T, =[yB(aR+ BG+yB)]'” — [16(8.4.7)] ally neural, between the end-organs with different
visual pigments, occurring before the response is
(ii) Model B. evaluated. In the second case, particularly, a
mechanism may incorporate in itself a differenc-
pedal Ram OR)61, wrle(G)ia AQ) iar ing process or some “inhibitory” action. The
Vos—Walraven line element (Section 8.4.6) is an
F,(B) = (yB)'” [17(8.4.7)] example of a line element incorporating differenc-
ing processes.
T, = T,=T,=4(aR + BG+yB)” A Class II or Class III component mechanism,
differing from a Class I (simple) mechanism in
The physiological concept underlying model A is, the replacement of a single visual pigment by a
in general terms, that the difference of response fixed mixture of two or all three visual pigments
686 THEORIES AND MODELS OF COLOR VISION
would have:
(a) A response function
F.(R,G, B) =F(L) [18(8.4.7) |
where
L=I,pR+1[oGt1,B
the quantities /p,/,, /, being constants, provided
very intense stimuli, capable of appreciably
changing the relative composition of the pigment
mixture by bleaching, are excluded, and
(b) A threshold function
T=T(L) —[19(8.4.7)]
With the appropriate choice of the constants
lp, /¢, 1g, the quantity L becomes the luminance
of the stimulus evaluated for the luminous ef-
ficiency function V(A) which is a linear combina-
tion of the color-matching functions. To the
extent that heterochromatic brightness matches
agree with predictions from V(A) (however, see
Section 5.8), the component mechanism would
produce a nonzero differential response A F, only
if the two stimuli differed in brightness. In this
sense, the mechanism might play the role of a
separate system for the appreciation of bright-
ness. For the functions F,(L) and T,(L) the forms
F,(L) = rin[7(L+ t9)]
and
T;,(L) =constant [20(8.4.7)]
are one possibility among many.
An example of a Class II component mecha-
nism with some physiological basis is obtained by
postulating a linear differencing process between
two logarithmic (or corresponding power func-
tion) outputs, each involving one only of the two
tristimulus values concerned (e.g., R and G). The
corresponding response function is
F(R, G) = aln{a(R + ay)
—Bln[ b(G + by)] [21(8.4.7)]
(a, B, a, b, do, by all positive constants)
with a threshold
_ 7} la(R +a)"
ay sheer tit
[22(8.4.7)]
equal, in the simplest case, to a constant K,, for
example. The contribution to the line element is
ae. (a\R — BAG )
T2\R+a, Gtbo
[23(8.4.7)]
A differencing mechanism of this kind can
provide the basic model for the opponent-color
theory if the responses F,(R, G) of different sign
are associated—when above threshold—with
the perception of the respective opponent colors.
However, these interpretations are not im-
mediately relevant to the application to threshold
discrimination.
A contribution to the line element similar to
the above, except that the sign between the two
terms is positive, is provided by a mechanism that
sums linearly two logarithmic outputs:
F.(R,G) =aln[a(R+a,)]
+ Bin[ b(G + by)]
T; = T,{[a(R + ay)]"[b(G + by)]*}
BAG |?
[24(8.4.7)]
The line element of Friele (1961, 1966) incorpo-
rates such terms.
The complete inductive line element—the sum
of the ? terms of all the component mechanisms
assumed to be operating— is necessarily homoge-
neous and of the second degree in AR, AG, AB.
To ensure that it leads to discrimination el-
lipsoids, there must be at least three mechanisms
for which the equations €, = 0, €, = 0, €, =0
(regarded as linear equations for AR, AG, AB
with constant coefficients) define three distinct
planes through the origin in (AR, AG, A B)-space.
With more than three mechanisms—certainly a
 Line Elements of Color Space 687

possibility—it will generally be easier to adjust
the constants to fit empirical data, and the signifi-
cance of a successful fit is necessarily reduced.
One of the most elaborate and, at present,
possibly most successful line elements based on
“opponent-color mechanisms” is that of Vos and
Walraven (1972a,b), details of which have been
given in Section 8.4.6. However, for the purpose
of bringing out certain structural features of some
such line elements, we have selected here a slightly
generalized form of an “opponent-color line ele-
ment” given by Shklover (1958). The Shklover
line element assumes two differencing mecha-
nisms and a brightness mediating mechanism,
with response functions
F(R, G) = a ln[a,(R + ayo)]
—B,In[b\(G + bio)]
F,(B, G) = ay|n[a,(B + ay9)]
Shklover successfully applied this scheme to some
of the simpler color-discrimination data and to
the determination of loci of constant hue and
saturation in the chromaticity diagram. Compari-
son with the then already available discrimination
data of MacAdam and Brown (color-matching
ellipses, Section 5.4) was not made, and in the
light of the later work by Friele (1961) and Vos
and Walraven (1972a,b), offers little promise.
Note also that the (B — G) differencing mecha-
nism postulated by Shklover is now usually re-
placed by a (Y— 8B) differencing mechanism,
where Y (yellow) is a (R + G) summing mecha-
nism.
In the above three-mechanism scheme with the
response functions defined by Eq. 25(8.4.7), put-
ting 8, equal to zero corresponds to placing the
second chromatic differencing mechanism by a
Class I (simple) mechanism dependent only on
the B tristimulus value. The resultant line element
becomes

— ByIn[ b,(G + by)] ni a AR B,AG |?

— K2| R+aq Gt bio
F,(R,G, B) = a;In[ a3(/gR + IgG
+ 1,B) + a3] [25(8.4.7)]
and constant threshold functions
[27(8.4.7) |
T, = K,, T, = K,, T, = K;
with
leading to the line element
L=1pR+1¢G+1,B [28(8.4.7) ]
| |a,AR ~.. BAG )
aK REO 6, hPa Ip, 1G, 1, being generally defined to make L the
luminance. In that case, because the contribution
Bp eet hs Nae of the B tristimulus value to luminance is always
Ki \ B+ an G + by very small, it is a good approximation to define L
by L=/p,R+4+1,G.
The application of this line element to the
1 | a,/pAR+ 05/1, AG + azl,AB
special case of MacAdam’s (Section 5.4.1) dis-
K3| (rR +1gG+ 1, Bi) aaaca crimination ellipses in a plane of constant lumi-
nance is illuminating. With AL = 0 and G= (L
[26(8.4.7)] — [pR)/lg, Eq. 27(8.4.7) reduces to

The constants a,, b,, a>, b,a,; do not figure in

eae) ee “(ARY?
the line element; they merely control constant Pete ol ena
Ke R+ ajo L—IpR+t Igbip
quantities added to the responses. With the fol-
lowing special choice: 2 o
gs I 2
re neal C2

Ag = dy = 4x9 = Dy = 43 2 2
AB ee 2) [29(8.4.7)]
a, = B,, a, = By, 37h Sk Sp
688 THEORIES AND MODELS OF COLOR VISION

0.0

EXPERIMENTAL

__ LOG(B*+0.3)
+ CONSTANT
LOG (B+ 0.47)
+ CONSTANT

0.5 1.0 1.5

LOG B
Fig. 4(8.4.7). MacAdam ellipses; semiaxis S, in its dependence on B when the tristimulus
values R,G, B are based on Fick-type fundamentals r(A),g(A),b(A) related to the CIE 1931
X(A), y(A),z(A) color-matching functions by the equations: x(A) = 1.350r(A )— 0.460g(A) +
0.195b(A); p(A) = 0.4887(A) + 0.622. 2(A); Z(A) = B(A).

which defines the ellipses projected on the plane than by

containing the R and B axes.
LeGrand (1949), in an instructive study of the
MacAdam ellipses, expressed them in terms of
tristimulus values based on two sets of fundamen-
tals, one of Fick type, the other of Konig type
(see Section 8.2.5). He concluded that for either
set of fundamentals,
(a) the experimental ellipses (projected on the
R-B plane) had semiaxes that were nearly parallel
to the R and B directions; and
(b) the length of the semiaxis parallel to the B
direction depended substantially on the B value
only.
These results are in accord with the predic-
tions of the above line element, that is, with Eq.
29(8.4.7). For the dependence of the semiaxis S,
on B, the data were represented better by
S, — const(B + a.)
However, this is an unessential difference, corre-
sponding to a change in the response function
F,(B) from
F,(B) = a,|n{a,(B als Ao) }
to
F,(B) = ain{ a, B a (B? +e) }
with e being a constant. Figure 4(8.4.7) illustrates
the results.
LeGrand’s most significant finding was that
the semiaxis Sp, plotted against the logarithm of
the ratio

S» = const(B? + 2)'/*
 Line Elements of Color Space 689

~x—x— EXPERIMENTAL

LINE ELEMENT
(a,
=8)
Ss,
LOG

20 T5 fe) 05 lO 15
Loc (2)
Fig. 5(8.4.7). MacAdam ellipses: semiaxis Sp in its dependence on R /G. Tristimulus values
as for Figure 4(8.4.7).

showed a well-defined minimum at a particular anism with the response function F,(R, G) of Eq.
value of log(R/G), on either side of which it 25(8.4.7), the assumption of a constant threshold
increased symmetrically. Equation 29(8.4.7) pre- T\(F,) = K, must be abandoned; that is, the
dicts on the contrary, that Sp should show a threshold 7\(F,) must be allowed to depend on
maximum, which should occur at the magnitude of the response F,.
For useful reviews of line elements as mathe-
l matical models of color discrimination, which fol-
log ]= tos(72 + 5lo $")
low a somewhat different approach to the one
given here, the reader is referred to articles by
when it is assumed that the luminance L is high Bouman and Walraven (1972) and Vos (1979).
enough for the small constants a,,. and b), to have The thesis of Buchsbaum (1978) on models of
negligible effect. Figure 5(8.4.7) compares the ex- central signal processing in color perception ex-
perimental and predicted variations of Sp for the ploits, in detail, basic concepts of the signal
Fick-type fundamentals when a, = f,. The detection theory developed in various fields of
minimum in the experimental and the maximum communication. In particular, Buchsbaum shows
in the predicted curve are fairly close that simple color discrimination tasks, such as
wavelength discrimination, are accounted for by
; R R
—log G = —0.16 applying decision rules described by likelihood
Me Gi at pred
ratio tests on statistics derived from a Poisson-type
and could be made to coincide by putting 8B, = process. Possible physiological interpretations of
2a,. A similar expression for the KO6nig-type the statistical model are avoided by Buchsbaum
fundamentals shows a much larger displacement and, in fact, the model does not have a unique
of opposite sign physiological correlate. Nonetheless, Buchsbaum
points out that the logic and strategy of the
R
log G one ar = +0.76
optimum rules of the probabilistic signal-detec-
tion algorithm that account for the performance
exp G pred
in different visual discrimination tasks provided
To obtain an explanation of the minimum in clues that might help to further our understand-
Sp while retaining a chromatic differencing mech- ing of the functioning of the visual mechanism.
APPENDIX OF EXTENDED TABLES
AND ILLUSTRATIONS

Table 1(1.1) Basic Radiometric Quantities?

Radiant Energy, Radiant energy is energy propagated in the form of electromagnetic

Radiation waves or streams of particles (quanta or photons).
Radiation is the process of emitting or transferring radiant energy.
Sometimes, however, radiation is also identified as radiant energy
itself.
Monochromatic radiant energy is radiant energy of a single frequency.
In practice, this term is extended to include radiant energy of a small
range of frequency or wavelength, which can be described by stating
a single frequency or wavelength.
The spectrum of radiant energy is the radiant energy when it is
regarded as an assembly of monochromatic components. Also the
term is frequently used for the image produced by the dispersion of
radiant energy into its monochromatic components.

Radiant Power, P, Radiant power (or radiant flux) is radiant energy emitted, trans-
Radiant Flux, F, ferred, or received through a surface, in unit time interval.

Radiant Exitance, M, Radiant exitance at a point of a surface is the quotient of the radiant
power emitted by an infinitesimal surface element containing the
point, by the area of that surface element.
Irradiance, E, Trradiance at a point of a surface is the quotient of the radiant power
incident on an infinitesimal surface element containing the point, by
the area of that surface element.

Radiant Intensity, J, Radiant intensity (of a source in a given direction) is the quotient of
the radiant power emitted by the source in an infinitesimal element
of solid angle containing the given direction by the element of solid
angle.
Radiance, L, Radiance in a given direction at a point on the surface of a source or
a receiver, or at a point on the path of a beam, is the quotient of the
radiant power leaving, arriving at, or passing through an element of
surface at this point and propagated in directions defined by an
elementary cone containing the given direction, by the product of
the solid angle of the cone and the area of the orthogonal projection
of the surface element on a plane perpendicular to the given
direction.
Period, ik Period is the time between successive occurrences of the same
characteristics in a periodic phenomenon.
Frequency, v Frequency is the reciprocal of period.

690
Table I(1.1)— Continued
—————————————————————————EEESEEEeESEESSSSSSSSSSSSSSSSFSFSFMFFfFfefsfsFseses

Wavelength, A Wavelength is the distance between two successive points of a

periodic wave in the direction of propagation in which the oscilla-
tion has the same phase. The wave propagates a distance equal to
one wavelength during every period. Thus, the product (Av) of
wavelength and frequency is equal to the velocity of the wave. In
vacuo, the velocity c of propagation of an electromagnetic wave is
constant and independent of the frequency and amplitude. The
velocity c decreases to c/n when the wave propagates through a
medium other than a vacuum; n is the index of refraction of the
medium.
Unless otherwise stated, values of wavelength are generally those in
air. The refractive index of standard air (15°C, 101 325 N-m“) lies
between 1.00027 and 1.00029 for visible radiant energies.
Wavenumber, m Wavenumber is the reciprocal of wavelength.
Photon, Quantum Photon (or quantum) is an elementary quantity of radiant energy of
one frequency. It is equal in value to hy, the product of Planck’s
constant h and the frequency of the electromagnetic radiation.
Spectral The spectral concentration at a given wavelength of a radiometric
Concentration quantity X, is given by the amount of the particular quantity, having
wavelengths in an infinitesimal interval containing the given wave-
length divided by the width of the interval. The usual notation is
PPX 3
Xen a dn

For example, if X, is radiant power P,, the spectral concentration of

Spectral Distribution
Function (Curve)
Relative Spectral
Distribution Function (Curve)
radiation power is denoted by P,,=dP,/dX and is simply referred
to as spectral radiant power.
Note: Frequencies v and wavenumbers m may also be used; if there
is a risk of ambiguity this should be avoided by means of an
appropriate wording. For example, P.,,,=dP,/dm would be de-
scribed as a spectral radiant power (wavenumber basis).
The spectral distribution function characterizes the variation of the
spectral concentration of a radiometric quantity with wavelength (or
frequency, or wavenumber). A corresponding graph is termed the
spectral distribution curve. A useful symbol is
{X,. dx}
which becomes for {P,, dA} the spectral distribution function (or
curve) of radiant power.
A relative spectral distribution function (or curve) gives the spectral
concentration in an arbitrary radiometric unit, (i.e., it specifies only
the relative values at different wavelengths). Its symbol is
{X,(A) da}
which becomes {P,(A) dA} for the relative spectral power distribu-
tion function (or curve). Note that in this case the wavelength A
follows the symbol for the quantity in parenthesis instead of as a
subscript.

“Note: Distinction between radiant and photometric quantities. Each of the radiometric quantities
radiant power P., radiant exitance M.,, irradiance E,, radiant intensity J,, and radiance L, corresponds to
a photometric quantity [defined in Table I(4.1)] which is derived from the radiometric quantity by means
of the spectral luminous efficacy for the CIE standard photometric observer [see Table 1(4.3.2)]. Both the
radiometric and the photometric quantities are denoted by the same principal symbol and the distinction
is made by adding the subscript e (energy) in the case of radiometric quantities and the subscript v
(visual) in the case of photometric quantities. Should there be no confusion, the subscripts may be
omitted.
691
 Table I(1.2.2) Spectral Radiant Exitance (M,, ) of a Blackbody Radiator at Different
Temperatures 7(K)*

r T(K)
(nm) 1000 1200 1400 1600 1800
300 0.48399(—8) 0.19796(—6) 0.28044(—5) 0.20477(— 4) 0.96125(—4)
310 0.19299(—7) 0.60995(—6) 0.71872(—5) 0.45709(— 4) 0.19271(—3)
320 0.70225(—7) 0.17429(—5) 0.17280(—4) 0.96550(—4) 0.36805(—3)
330 0.23518(—6) 0.46511(—5) 0.39209( —4) 0.19398(—3) 0.67271(—3)
340 0.73028(—6) 0.11664(—4) 0.84404(— 4) 0.37240(—3) 0.11814(—2)
350 0.21166(—5) 0.27635(—4) 0.17318(—3) 0.68588(—3) 0.20007(—2)
360 0.57596(—5) 0.62168(—4) 0.34005(—3) 0.12163(—2) 0.32773(—2)
370 0.14791(—4) 0.13335(—3) 0.64140(—3) 0.20832(—2) 0.52076(—2)
380 0.36019(—4) 0.27381(—3) 0.11659(—2) 0.34561(—2) 0.80470(—2)
390 0.83514(—4) 0.54001(—3) 0.20485(—2) 0.55681(—2) 0.12119(—1)
400 0.18507(—3) 0.10262(—2) 0.34879( —2) 0.87312(—2) 0.17825(—1)
410 0.39330(—3) 0.18841(—2) 0.57690(—2) 0.13353(— 1) 0.25650(—1)
420 0.80403(—3) 0.33510(—2) 0.92889(—2) 0.19955(— 1) 0.36170(— 1)
430 0.15855(—2) 0.57864(—2) 0.14588(—1) 0.29188(—1) 0.50058(—1)
440 0.30235(—2) 0.97209( —2) 0.22387(— 1) 0.41850(—1) 0.68082(— 1)
450 0.55886(—2) 0.15918(—1) 0.33621(—1) 0.58904(— 1) 0.91108(—1)
460 0.10033(—1) 0.25452(— 1) 0.49488(— 1) 0.81486(— 1) 0.12010
470 0.17529(— 1) 0.39799(— 1) 0.71490(— 1) 0.11092 0.15610
480 0.29855(—1) 0.60950(— 1) 0.10148 0.14874 0.20024
490 0.49650(—1) 0.91537(—1) 0.14170 0.19665 0.25374
500 0.80742(—1) 0.13498 0.19483 0.25658 0.31784
510 0.12857 0.19564 0.26406 0.33065 0.39386
520 0.20072 0.27903 0.35305 0.42118 0.48314
530 0.30758 0.39195 0.46603 0.53065 0.58705
540 0.46313 0.54273 0.60782 0.66173 0.70693
550 0.68588 0.74142 0.78381 0.81720 0.84414
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.14366(1) 0.13326(1) 0.12630(1) 0.12132(1) 0.11758(1)
580 0.20350(1) 0.17557(1) 0.15799(1) 0.14598(1) 0.13727(1)
590 0.28448(1) 0.22882(1) 0.19586(1) 0.17430(1) 0.15918(1)
600 0.39271(1) 0.29518(1) 0.24073(1) 0.20659(1) 0.18342(1)
610 0.53568(1) 0.37711(1) 0.29348(1) 0.24318(1) 0.21009(1)
620 0.72246(1) 0.47736(1) 0.35505(1) 0.28437(1) 0.23927(1)
630 0.96393(1) 0.59898(1) 0.42640(1) 0.33046(1) 0.27103(1)
640 0.12730(2) 0.74535(1) 0.50853(1) 0.38175(1) 0.30544(1)
650 0.16648(2) 0.92016(1) 0.60247(1) 0.43852(1) 0.34253(1)
660 0.21571(2) 0.11274(2) 0.70929(1) 0.50104(1) 0.38236(1)
670 0.27703(2) 0.13715(2) 0.83005(1) 0.56956(1) 0.42493(1)
680 0.35278(2) 0.16570(2) 0.96583(1) 0.64430(1) 0.47027(1)
690 0.44563(2) 0.19888(2) 0.11177(2) 0.72547(1) 0.51836(1)
700 0.55860(2) 0.23723(2) 0.12867(2) 0.81327(1) 0.56920(1)
710 0.69507(2) 0.28128(2) 0.14740(2) 0.90785(1) 0.62274(1)
720 0.85880(2) 0.33161(2) 0.16805(2) 0.10093(2) 0.67896(1)
730 0.10540(3) 0.38881(2) 0.19072(2) 0.11179(2) 0.73781(1)
740 0.12852(3) 0.45351(2) 0.21551(2) 0.12335(2) 0.79921(1)
750 0.15574(3) 0.52634(2) 0.24252(2) 0.13563(2) 0.86311(1)
760 0.18761(3) 0.60794(2) 0.27183(2) 0.14863(2) 0.92941(1)
710 0.22472(3) 0.69897(2) 0.30351(2) 0.16236(2) 0.99804(1)
780 0.26771(3) 0.80009(2) 0.33765(2) 0.17680(2) 0.10689(2) ©
790 0.31726(3) 0.91197(2) 0.37432(2) 0.19195(2) 0.11418(2)
800 0.37408(3) 0.10353(3) 0.41357(2) 0.20781(2) 0.12168(2)
810 0.43896(3) 0.11707(3) 0.45546(2) 0.22437(2) 0.12937(2)
820 0.51269(3) 0.13188(3) 0.50004(2) 0.24161(2) 0.13723(2)
830 0.59614(3) 0.14804(3) 0.54735(2) 0.25953(2) 0.14526(2)
560 0.47193(—3) 0.34165(— 1) 0.72770 0.72148(1) 0.42964(2)
692
Table I(1.2.2)— Continued

r T(K)
(nm) 2000 2050 2100 2150 2200
300 0.33120(-3) ——-:0.43453(—3) 0.56277(-3) -0.72015(—3) —0.91128(—3)
310 0.60930(—3) -0.78446(—3) ——-0.9978—-3) _—-0.12553(—2) 0.15626(—2)
320 0.10736(—2) -0.13580(—2) 0.16986(—2) -0.21027(—2) 0.25777(—2)
330 0.18192(—2) 0.22632(—2) 0.27864(—2) -0.33976(—2) —-0.41056(—2)
340 0.29752(—2) 0.36439(—2) 0.44200(-2) ——-0.53135(—2) —-0.63343(—2)
350 0.47111(—2) 0.56854(—2) 0.68002(—2) 0.80661(—2) 0.94936(—2)
360 0.72428(—2) -0.86200(—2) + ~—s-0.10174(—1)_—s«<0.11917(—1) 0.13857(—1)
370 0.10830 —1) ——-0.12731(—1) 0.14839—1) —0.17173(—1) 0.19743(—1)
380 0.15822(—1) —-0.18354(—1) 0.21141(-1) ——-:0.24191(—1) 0.27512(—1)
390 0.22578(—1) 0.25882(—1) 0.29478(—1) ——-0.33370(—1) 0.37564(— 1)
400 0.31549(—1) 0.35762(—1) 0.40295(— 1) 0.45152(—1) 0.50334(—1)
410 0.43239(— 1) 0.48491(-1) —-0.54084(—1) —s:0.60017(/—1)_——«0.66287(—1)
420 0.58208(— 1) 0.64616(— 1) 0.71374(—1) 0.78474(— 1) 0.85910(—1)
430 0.77069(— 1) 0.84727(— 1) 0.92726(—1) 0.10106 0.10970
440 0.10048 0.10945 0.11873 0.12831 0.13817
450 0.12915 0.13943 0.14998 0.16078 0.17182
460 0.16379 0.17534 0.18709 0.19903 0.21113
470 0.20517 0.21785 0.23066 0.24358 0.25658
480 0.25403 0.26764 0.28129 0.29495 0.30861
490 0.31113 0.32537 0.33954 0.35363 0.36762
500 0.37722 0.39167 0.40595 0.42005 0.43397
510 0.45302 0.46715 0.48102 0.49462 0.50797
520 0.53921 0.55236 0.56519 0.57770 0.58989
530 0.63644 0.64783 0.65887 0.66956 0.67994
540 0.74531 0.75401 0.76239 0.77046 0.77825
550 0.86634 0.87129 0.87603 0.88057 0.88493
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.11467(1) 0.11404(1) 0.11344(1) 0.11288(1) 0.11234(1)
580 0.13067(1) 0.12927(1) 0.12795(1) 0.12670(1) 0.12552(1)
590 0.14804(1) 0.14570(1) 0.14350(1) 0.14144(1) 0.13950(1)
600 0.16677(1) 0.16333(1) 0.16011(1) 0.15710(1) 0.15428(1)
610 0.18690(1) 0.18216(1) 0.17776(1) 0.17366(1) 0.16984(1)
620 0.20840(1) 0.20218(1) 0.19642(1) 0.19109(1) 0.18613(1)
630 0.23128(1) 0.22337(1) 0.21609(1) 0.20936(1) 0.20314(1)
640 0.255531) 0.24571(1) 0.23672(1) 0.22845(1) 0.22082(1)
650 0.28111(1) 0.26917(1) 0.25828(1) 0.24831(1) 0.23914(1)
660 0.30800(1) 0.29372(1) 0.28074(1) 0.26889(1) 0.25806(1)
670 0.33616(1) 0.31931(1) 0.30405(1) 0.29017(1) 0.27752(1)
680 0.36556(1) 0.34590(1) 0.32816(1) 0.31209(1) 0.29749(1)
690 0.39614(1) 0.37343(1) 0.35302(1) 0.33460(1) 0.31791(1)
700 0.42785(1) 0.40186(1) 0.37859(1) 0.35765(1) 0.33874(1)
710 0.46063(1) 0.43113(1) 0.40480(1) 0.38119(1) 0.35994(1)
720 0.49443(1) 0.46118(1) 0.43159(1) 0.40516(1) 0.38144(1)
730 0.52917(1) 0.49194(1) 0.45892(1) 0.42951(1) 0.40319(1)
740 0.56480(1) 0.52336(1) 0.48673(1) 0.45419(1) 0.42516(1)
750 0.60123(1) 0.55536(1) 0.51494(1) 0.47914(1) 0.44729(1)
760 0.63841(1) 0.58789(1) 0.54350(1) 0.50431(1) 0.46954(1)
770 0.67625(1) 0.62087(1) 0.57236(1) 0.52964(1) 0.491841)
780 0.71468(1) 0.65424(1) 0.60145(1) 0.55509(1) 0.51417(1)
790 0.75363(1) 0.68794(1) 0.63072(1) 0.58060(1) 0.53648(1)
800 0.79302(1) 0.72189(1) 0.66010(1) 0.606 12(1) 0.55871(1)
810 0.83278(1) 0.75604(1) 0.68954(1) 0.63160(1) 0.58084(1)
820 0.87283(1) 0.7903 1(1) 0.71900(1) 0.65700(1) 0.60283(1)
830 0.91311(1) 0.82466(1) 0.74840(1) 0.68228(1) 0.62462(1)
560 0.17906(3) 0.24495(3) 0.33013(3) 0.43879(3) 0.57571(3)
693
 Table 1(1.2.2)— Continued

r T(K)
(nm) 2250 2300 2350 2400 2450
300 0.11411(—2) —-0.14150(—2) _—0.17387(—2) 0.21181(—2) 0.25596(—2)
310 0.19264(—2) 0.23534(—2) —-0.28506(—2) ——-0.34253(—2) 0.40853(—2)
320 0.31316(—2) 0.37724(—2) —-0.45085(—2) ~—-0.53484(—2) 0.63006(—2)
330 0.49196(—2) 0.584892) _:0.69026(—2) 0.80901(—2) 0.94207(—2)
340 0.74925(—2) 0.87980(-2) —-0..10261(—1) 0.11890(— 1) 0.13695(— 1)
350 0.11093(—1) —-0.12875(—1) 0.14848(— 1) 0.17023(—1) 0.19407(—1)
360 0.16006(—1) —-0.18373(—1) 0.20967(— 1) 0.23795(—1) 0.26866(—1)
370 0.22558(—1) —-0.25625(—1) 0.28951(—1) —-0.32543(—1) 0.36406(— 1)
380 0.31111(—1) ——:0.34993(—1) 0.39162(—1) —-0.43624(—1) 0.48380(— 1)
390 0.42063(—1) —-0.46869(—1) 0.51986(—1) 0.57412(—1) 0.63147(— 1)
400 0.55840(—1) ——-0.61669(— 1) 0.67819(—1) 0.74288(— 1) 0.81072(— 1)
410 0.72889%—1) 0.79819(—1) 0.87070(— 1) 0.94636(— 1) 0.10251
420 0.93673(—1) 0.10175 0.11014 0.11883 0.12781
430 0.11866 0.12790 0.13743 0.14723 0.15728
440 0.14830 0.15869 0.16932 0.18017 0.19123
450 0.18307 0.19452 0.20616 0.21796 0.22992
460 0.22338 0.23577 0.24827 0.26087 0.27356
470 0.26965 0.28278 0.29595 0.30914 0.32235
480 0.32225 0.33586 0.34943 0.36295 0.37641
490 0.38150 0.39527 0.40892 0.42244 0.43583
500 0.44769 0.46123 0.47457 0.48772 0.50067
510 0.52106 0.53390 0.54648 0.55882 0.57092
520 0.60179 0.61339 0.62470 0.63574 0.64651
530 0.69000 0.69976 0.70924 0.71844 0.72738
540 0.78577 0.79303 0.80004 0.80681 0.81337
550 0.88911 0.89314 0.89700 0.90073 0.90431
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.11183(1) 0.11135(1) 0.11088(1) 0.11044(1) 0.11002(1)
580 0.12440(1) 0.12334(1) 0.12233(1) 0.12137(1) 0.12046(1)
590 0.13767(1) 0.13595(1) 0.13431(1) 0.13277(1) 0.13130(1)
600 0.15164(1) 0.14915(1) 0.14680(1) 0.14459(1) 0.14250(1)
610 0.16626(1) 0.16291(1) 0.15977(1) 0.15682(1) 0.15403(1)
620 0.18152(1) 0.17721(1) 0.17318(1) 0.16941(1) 0.16586(1)
630 0.19737(1) 0.19200(1) 0.18700(1) 0.18232(1) 0.17795(1)
640 0.21378(1) 0.20725(1) 0.20118(1) 0.19554(1) 0.19027(1)
650 0.23070(1) 0.22291(1) 0.21570(1) 0.20900(1) 0.20278(1)
660 0.24811(1) 0.23895(1) 0.23050(1) 0.22269(1) 0.21544(1)
670 0.26595(1) 0.25533(1) 0.24556(1) 0.23655(1) 0.22822(1)
680 0.28417(1) 0.27200(1) 0.26083(1) 0.25056(1) 0.24109(1)
690 0.30275(1) 0.28892(1) 0.27627(1) 0.26467(1) 0.25401(1)
700 0.32162(1) 0.30604(1) 0.29184(1) 0.27886(1) 0.26694(1)
710 0.34074(1) 0.32334(1) 0.30751(1) 0.29307(1) 0.27986(1)
720 0.36007(1) 0.34075(1) 0.32323(1) 0.30729(1) 0.29274(1)
730 0.37956(1) 0.35825(1) 0.33897(1) 0.32147(1) 0.30554(1)
740 0.39916(1) 0.37579(1) 0.354691) 0.33559(1) 0.318241)
750 0.41884(1) 0.39333(1) 0.37036(1) 0.34961(1) 0.33081(1)
760 0.43856(1) 0.41084(1) 0.38595(1) 0.36351(1) 0.34322(1)
710 0.45825(1) 0.42828(1) 0.40142(1) 0.37726(1) 0.35546(1)
780 0.47790(1) 0.44561(1) 0.41674(1) 0.39083(1) 0.36750(1)
790 0.49746(1) 0.46280(1) 0.43189(1) 0.40421(1) 0.37933(1)
800 0.51689(1) 0.47983(1) 0.44684(1) 0.41735(1) 0.39091(1)
810 0.53616(1) 0.49665(1) 0.46156(1) 0.43026(1) 0.40224(1)
820 0.55524(1) 0.51325(1) 0.47603(1) 0.44291(1) 0.41331(1)
830 0.57410(1) 0.52960(1) 0.49024(1) 0.45527(1) 0.42409(1)
560 0.74630(3) 0.95659(3) 0.12132(4) 0.15236(4) 0.18957(4)
694
Table 1(1.2.2)— Continued

r T(K)
(nm) 2500 2550 2600 2650 2700
eee eee eee ee eee es ee ee a
300 0.30699( — 2) 0.36556(—2) 0.43240(—2) 0.50823(—2) 0.59379 —2)
310 0.48381(—2) 0.56918(—2) 0.66544(— 2) 0.77340( —2) 0.89389(—2)
320 0.73739( —2) 0.85769(— 2) 0.99183(—2) 0.11407(—1) 0.13051(— 1)
330 0.10903(— 1) 0.12548(—1) 0.14362(— 1) 0.16354(—1) 0.18534(—1)
340 0.15686(— 1) 0.17870(— 1) 0.20257(—1) 0.22854(— 1) 0.25670(—1)
350 0.22009(— 1) 0.24837(— 1) 0.27899(— 1) 0.31201(— 1) 0.34750(—1)
360 0.30185(—1) 0.33761(— 1) 0.37598(—1) 0.41701(— 1) 0.46074(— 1)
370 0.40546(— 1) 0.44966(— 1) 0.49670(— 1) 0.54660(— 1) 0.59939(— 1)
380 0.53433(— 1) 0.58784(— 1) 0.64434(— 1) 0.70383(—1) 0.76630(— 1)
390 0.69192(— 1) 0.75545(— 1) 0.82202(—1) 0.89160(— 1) 0.96418(—1)
400 0.88166(— 1) 0.95567(— 1) 0.10327 0.11126 0.11955
410 0.11069 0.11915 0.12790 0.13693 0.14623
420 0.13706 0.14658 0.15636 0.16638 0.17664
430 0.16758 0.17810 0.18884 0.19979 0.21094
440 0.20249 0.21393 0.22553 0.23729 0.24920
450 0.24201 0.25423 0.26656 0.27898 0.29150
460 0.28632 0.29914 0.31201 0.32491 0.33784
470 0.33555 0.34875 0.36193 0.37509 0.38820
480 0.38980 0.40311 0.41633 0.42947 0.44251.
490 0.44969 0.46220 0.47517 0.48800 0.50067
500 0.51343 0.52600 0.53837 0.55054 0.56253
510 0.58277 0.59440 0.60580 0.61697 0.62792
520 0.65703 0.66729 0.67731 0.68709 0.69664
530 0.73606 0.74450 0.75271 0.76069 0.76846
540 0.81971 0.82585 0.83179 0.83755 0.84314
550 0.90776 0.91110 0.91431 0.91741 0.92041
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.10962(1) 0.10923(1) 0.10886(1) 0.10850(1) 0.10816(1)
580 0.11960(1) 0.11877(1) 0.11798(1) 0.11722(1) 0.11650(1)
590 0.12991(1) 0.12858(1) 0.12732(1) 0.12612(1) 0.12498(1)
600 0.14052(1) 0.13865(1) 0.13687(1) 0.13518(1) 0.13357(1)
610 0.15141(1) 0.14893(1) 0.14659(1) 0.14436(1) 0.14226(1)
620 0.16253(1) 0.15940(1) 0.15644(1) 0.15364(1) 0.15100(1)
630 0.17386(1) 0.17001(1) 0.16639(1) 0.16298(1) 0.15977(1)
640 0.18535(1) 0.18074(1) 0.17642(1) 0.17236(1) 0.16854(1)
650 0.19698(1) 0.19156(1) 0.18649(1) 0.18175(1) 0.17729(1)
660 0.20871(1) 0.20243(1) 0.19658(1) 0.19111(1) 0.18599(1)
670 0.22050(1) 0.21333(1) 0.20666(1) 0.20044(1) 0.19462(1)
680 0.23233(1) 0.22422(1) 0.21669(1) 0.20969(1) 0.203 16(1)
690 0.24417(1) 0.23508(1) 0.22667(1) 0.21885(1) 0.21158(1)
700 0.25598(1) 0.24588(1) 0.23655(1) 0.22790(1) 0.21987(1)
710 0.26774(1) 0.25659(1) 0.24631(1) 0.23681(1) 0.22801(1)
720 0.27942(1) 0.26720(1) 0.25595(1) 0.24558(1) 0.23599(1)
730 0.29099(1) 0.27766(1) 0.26543(1) 0.25417(1) 0.24378(1)
740 0.30243(1) 0.28798(1) 0.27474(1) 0.26258(1) 0.25138(1)
750 0.31371(1) 0.29812(1) 0.28386(1) 0.27078(1) 0.25877(1)
760 0.32481(1) 0.30806(1) 0.29277(1) 0.27878(1) 0.26594(1)
770 0.33572(1) 0.31780(1) 0.30147(1) 0.28655(1) 0.27289(1)
780 0.34642(1) 0.32731(1) 0.30993(1) 0.29409(1) 0.27960(1)
790 0.35689(1) 0.33658(1) 0.31816(1) 0.30138(1) 0.28607(1)
800 0.36711(1) 0.34561(1) 0.32613(1) 0.30843(1) 0.29230(1)
810 0.37707(1) 0.35437(1) 0.33384(1) 0.31522(1) 0.29827(1)
820 0.38676(1) 0.36286(1) 0.34129(1) 0.32175(1) 0.303991)
830 0.39617(1) 0.37108(1) 0.34846(1) 0.32801(1) 0.30945(1)
560 0.23380(4) 0.28600(4) 0.34715(4) 0.41831(4) 0.500594)
695
 Table I(1.2.2)— Continued

‘! T(K)
(nm) 2750 2800 2850 2900 2950
300 0.68984(—2) 0.79715(—2) 0.91648(—2) 0.10486(— 1) 0.11943(—1)
310 0.10277(—1) 0.11757(—1) 0.13387(— 1) 0.15174(—1) 0.17127(—1)
320 0.14859(—1) 0.16839(— 1) 0.19000(— 1) 0.21349(—1) 0.23893(—1)
330 0.20909(— 1) 0.23487(—1) 0.26276(—1) 0.29282(—1) 0.32511(—1)
340 0.28710(—1) —-0.31982(-1)_—S—s«0.35493(—1) 0.39247(—1) 0.43251(—1)
350 0.38551(—1) 0.42609(— 1) 0.46929(— 1) 0.51515(—1) 0.56370(— 1)
360 0.50722(— 1) 0.55647(— 1) 0.60852(— 1) 0.66339(— 1) 0.72109(—1)
370 0.65507(— 1) 0.71365(—1) 0.77514(—1) 0.83953(— 1) 0.90681(—1)
380 0.83174(— 1) 0.90012(—1) 0.97143(— 1) 0.10456 0.11227
390 0.10397 0.11181 0.11994 0.12834 0.13702
400 0.12811 0.13695 0.14605 0.15542 0.16504
410 0.15578 0.16558 0.17563 0.18590 0.19639
420 0.18713 0.19783 0.20873 0.21982 0.23110
430 0.22226 0.23375 0.24541 0.25721 0.26915
440 0.26124 0.27340 0.28566 0.29803 0.31048
450 0.30409 0.31674 0.32945 0.34220 0.35499
460 0.35079 0.36374 0.37669 0.38962 0.40255
470 0.40127 0.41430 0.42726 0.44016 0.45299
480 0.45546 0.46829 0.48102 0.49364 0.50614
490 0.51319 0.52557 0.53779 0.54985 0.56177
500 0.57433 0.58593 0.59736 0.60859 0.61965
510 0.63866 0.64918 0.65950 0.66962 0.67955
520 0.70597 0.71508 0.72398 0.73269 0.74119
530 0.77602 0.78338 0.79054 0.79752 0.80432
540 0.84856 0.85381 0.85891 0.86387 0.86868
550 0.92331 0.92611 0.92882 0.93145 0.93399
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.10783(1) 0.10752(1) 0.10722(1) 0.10693(1) 0.10664(1)
580 0.11581(1) 0.11514(1) 0.11451(1) 0.11389(1) 0.11331(1)
590 0.12389(1) 0.12284(1) 0.12184(1) 0.12088(1) 0.11996(1)
600 0.132041) 0.13059(1) 0.12919(1) 0.12786(1) 0.12659(1)
610 0.14026(1) 0.13835(1) 0.13654(1) 0.13482(1) 0.13317(1)
620 0.14849(1) 0.14612(1) 0.14386(1) 0.14172(1) 0.13968(1)
630 0.15673(1) 0.15385(1) 0.15113(1) 0.14855(1) 0.14609(1)
640 0.164941) 0.16154(1) 0.15832(1) 0.15528(1) 0.15240(1)
650 0.17310(1) 0.16915(1) 0.16543(1) 0.16191(1) 0.15859(1)
660 0.18118(1) 0.17667(1) 0.17242(1) 0.16841(1) 0.16463(1)
670 0.18918(1) 0.18407(1) 0.17928(1) 0.17477(1) 0.17053(1)
680 0.19706(1) 0.19135(1) 0.18600(1) 0.18098(1) 0.17626(1)
690 0.20481(1) 0.19848(1) 0.19256(1) 0.18702(1) 0.18182(1)
700 0.21241(1) 0.20545(1) 0.19895(1) 0.19288(1) 0.18719(1)
710 0.21984(1) 0.21224(1) 0.20516(1) 0.19855(1) 0.19237(1)
720 0.22710(1) 0.21885(1) 0.21118(1) 0.20403(1) 0.19736(1)
730 0.23417(1) 0.22527(1) 0.21700(1) 0.20931(1) 0.20214(1)
740 0.24104(1) 0.23147(1) 0.22261(1) 0.21437(1) 0.20671(1)
750 0.24769(1) 0.23747(1) 0.22801(1) 0.21923(1) 0.21108(1)
760 0.25413(1) 0.24324(1) 0.23318(1) 0.22387(1) 0.21523(1)
710 0.26034(1) 0.24879(1) 0.23814(1) 0.22829(1) 0.21916(1)
780 0.26632(1) 0.25411(1) 0.24287(1) 0.23249(1) 0.22288(1)
790 0.27206(1) 0.25920(1) 0.24737(1) 0.23647(1) 0.22639(1)
800 0.27755(1) 0.26405(1) 0.25164(1) 0.24022(1) 0.22969(1)
810 0.28281(1) 0.26866(1) 0.255691) 0.24376(1) 0.23277(1)
820 0.28781(1) 0.27303(1) 0.25950(1) 0.24708(1) 0.23564(1)
830 0.29257(1) 0.27717(1) 0.26309(1) 0.25018(1) 0.23831(1)
560 0.59515(4) 0.70322(4) 0.82606(4) 0.96499(4) 0.11214(5)
696
~ Table 1(1.2.2)— Continued

r T(K)
(nm) 3000 3050 3100 3150 3200
300 0.13544(— 1) 0.15296(— 1) 0.17207(— 1) 0.19285(—1) 0.21536(—1)
310 0.19254(—1) 0.21561(—1) 0.24057(— 1) 0.26749(— 1) 0.29643(—1)
320 0.26640(— 1) 0.29598(— 1) 0.32772(— 1) 0.36169(—1) 0.39795(— 1)
330 0.35972(— 1) 0.39669(— 1) 0.43607(— 1) 0.47793(— 1) 0.52231(—1)
340 0.47509(— 1) 0.52026(— 1) 0.56805(— 1) 0.61850(— 1) 0.67164(— 1)
350 0.61498(—1) 0.66902(— 1) 0.72582(— 1) 0.78540(— 1) 0.84778(— 1)
360 0.78163(—1) 0.84501(—1) 0.91123(—1) 0.98029(— 1) 0.10522
370 0.97696(— 1) 0.10500 0.11258 0.12044 0.12859
380 0.12026 0.12852 0.13706 0.14587 0.15494
390 0.14597 0.15517 0.16464 0.17435 0.18430
400 0.17490 0.18500 0.19533 0.20587 0.21663
410 0.20710 0.21801 0.22911 0.24040 0.25187
420 0.24255 0.25417 0.26594 0.27785 0.28990
430 0.28122 0.29341 0.30570 0.31809 0.33058
440 0.32301 0.33561 0.34827 0.36098 0.37373
450 0.36780 0.38063 0.39348 0.40632 0.41917
460 0.41544 0.42831 0.44114 0.45392 0.46666
470 0.46575 0.47843 0.49103 0.50354 0.51597
480 0.51852 0.53079 0.54293 0.55495 0.56685
490 0.57353 0.58514 0.59659 0.60790 0.61905
500 0.63053 0.64123 0.65176 0.66212 0.67231
510 0.68928 0.69882 0.70818 0.71736 0.72636
520 0.74950 0.75763 0.76558 0.77336 0.78096
530 0.81095 0.81742 0.82372 0.82986 0.83586
540 0.87336 0.87791 0.88233 0.88663 0.89082
550 0.93646 0.93885 0.94117 0.94342 0.94560
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.10637(1) 0.10611(1) 0.10586(1) 0.10562(1) 0.10538(1)
580 0.11274(1) 0.11220(1) 0.11167(1) 0.111171) 0.11068(1)
590 0.11908(1) 0.11824(1) 0.11742(1) 0.11664(1) 0.11589(1)
600 0.12537(1) 0.12421(1) 0.12309(1) 0.12202(1) 0.12099(1)
610 0.13160(1) 0.13009(1) 0.12866(1) 0.12728(1) 0.12596(1)
620 0.13773(1) 0.13588(1) 0.13411(1) 0.13241(1) 0.13080(1)
630 0.14376(1) 0.14154(1) 0.13942(1) 0.13741(1) 0.13548(1)
640 0.14967(1) 0.14707(1) 0.14460(1) 0.14225(1) 0.14001(1)
650 0.15544(1) 0.15245(1) 0.14962(1) 0.14693(1) 0.14437(1)
660 0.16106(1) 0.15768(1) 0.15448(1) 0.15144(1) 0.14855(1)
670 0.16652(1) 0.16274(1) 0.15916(1) 0.15577(1) 0.15256(1)
680 0.17181(1) 0.16762(1) 0.16366(1) 0.15992(1) 0.15638(1)
690 0.17692(1) 0.17232(1) 0.16798(1) 0.163891) 0.16001(1)
700 0.18185(1) 0.17683(1) 0.17211(1) 0.16766(1) 0.16346(1)
710 0.18658(1) 0.18115(1) 0.17604(1) 0.17124(1) 0.16671(1)
720 0.19111(1) 0.18526(1) 0.17978(1) 0.17462(1) 0.16977(1)
730 0.19544(1) 0.18918(1) 0.18331(1) 0.17781(1) 0.17264(1)
740 0.19957(1) 0.19289(1) 0.18665(1) 0.18080(1) 0.17531(1)
750 0.20348(1) 0.19640(1) 0.18979(1) 0.18360(1) 0.17780(1)
760 0.20719(1) 0.19971(1) 0.19273(1) 0.18621(1) 0.18010(1)
770 0.21069(1) 0.20281(1) 0.19548(1) 0.18862(1) 0.18222(1)
780 0.21398(1) 0.20572(1) 0.19802(1) 0.19085(1) 0.18416(1)
790 0.21707(1) 0.20842(1) 0.20038(1) 0.19290(1) 0.18592(1)
800 0.21995(1) 0.21092(1) 0.20255(1) 0.19476(1) 0.18750(1)
810 0.22262(1) 0.21323(1) 0.20453(1) 0.19645(1) 0.18892(1)
820 0.22510(1) 0.21535(1) 0.20633(1) 0.19796(1) 0.19018(1)
830 0.22738(1) 0.21729(1) 0.20795(1) 0.19930(1) 0.19127(1)
560 0.12966(5) 0.14921(5) 0.17093(5) 0.19497(5) 0.22147(5)

697
 Table 1(1.2.2)— Continued

N T(K)
(nm) 3250 3300 3350 3400 3600
300 0.23968(—1) 0.26589— 1) 0.29405(— 1) 0.32422(— 1) 0.46638(— 1)
310 0.32747(—1) 0.36066(— 1) 0.39608(— 1) 0.43377(— 1) 0.60838(—1)
320 0.43656(—1) 0.47757(—1) 0.52104(—1) 0.56700(— 1) 0.77662(— 1)
330 0.56924(— 1) 0.61877(—1) 0.67094(— 1) 0.72577(— 1) 0.97220(—1)
340 0.72750(—1) 0.78609(— 1) 0.84743(— 1) 0.91154(—1) 0.11957
350 0.91296(— 1) 0.98095(—1) 0.10517 0.11253 0.14473
360 0.11269 0.12044 0.12846 0.13676 0.17264
370 0.13700 0.14568 0.15464 0.16385 0.20321
380 0.16427 0.17385 0.18368 0.19375 0.23632
390 0.19449 0.20490 0.21554 0.22639 0.27178
400 0.22759 0.23875 0.25010 0.26163 0.30939
410 0.26350 0.27529 0.28723 0.29932 0.34892
420 0.30207 0.31436 0.32676 0.33926 0.39013
430 0.34314 0.35578 0.36849 0.38125 0.43274
440 0.38652 0.39934 0.41219 0.42504 0.47649
450 0.43200 0.44482 0.45762 0.47039 0.52111
460 0.47934 0.49197 0.50454 0.51704 0.56633
470 0.52830 0.54054 0.55268 0.56473 0.61188
480 0.57863 0.59027 0.60180 0.61319 0.65752
490 0.63005 0.64091 0.65162 0.66218 0.70299
500 0.68233 0.69219 0.70189 0.71143 0.74808
510 0.73520 0.74386 0.75237 0.76071 0.79258
520 0.78841 0.79569 0.80282 0.80979 0.83628
530 0.84171 0.84743 0.85301 0.85845 0.87903
540 0.89489 0.89886 0.90273 0.90650 0.92065
550 0.94772 0.94978 0.95179 0.95373 0.96101
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.10515(1) 0.10493(1) 0.10472(1) 0.10451(1) 0.10375(1)
580 0.11021(1) 0.10976(1) 0.10932(1) 0.10890(1) 0.10734(1)
590 0.11517(1) 0.11447(1) 0.11380(1) 0.11315(1) 0.11077(1)
600 0.12000(1) 0.11905(1) 0.11813(1) 0.11725(1) 0.11403(1)
610 0.12470(1) 0.12348(1) 0.12232(1) 0.12120(1) 0.11712(1)
620 0.12925(1) 0.12776(1) 0.12634(1) 0.12498(1) 0.12003(1)
630 0.13364(1) 0.13188(1) 0.13020(1) 0.12858(1) 0.12276(1)
640 0.13787(1) 0.13583(1) 0.13388(1) 0.13202(1) 0.12531(1)
650 0.14193(1) 0.13961(1) 0.13739(1) 0.13527(1) 0.12768(1)
660 0.14581(1) 0.14320(1) 0.14071(1) 0.13834(1) 0.12987(1)
670 0.14951(1) 0.14661(1) 0.14386(1) 0.14123(1) 0.13189(1)
680 0.15302(1) 0.14984(1) 0.14682(1) 0.14394(1) 0.13373(1)
690 0.15635(1) 0.15288(1) 0.14959(1) 0.14646(1) 0.13541(1)
700 0.15949(1) 0.15573(1) 0.15218(1) 0.14881(1) 0.13691(1)
710 0.16244(1) 0.15840(1) 0.15458(1) 0.15097(1) 0.13826(1)
720 0.16520(1) 0.16088(1) 0.15681(1) 0.15296(1) 0.13945(1)
730 0.16777(1) 0.16318(1) 0.15886(1) 0.15477(1) 0.14048(1)
740 0.17016(1) 0.16530(1) 0.16073(1) 0.15641(1) 0.14136(1)
750 0.17236(1) 0.16724(1) 0.16243(1) 0.15789(1) 0.14211(1)
760 0.17438(1) 0.16901(1) 0.16396(1) 0.15920(1) 0.14271(1)
710 0.17623(1) 0.17060(1) 0.16532(1) 0.16036(1) 0.14318(1)
780 0.17790(1) 0.17203(1) 0.16653(1) 0.16136(1) 0.14353(1)
790 0.17940(1) 0.17330(1) 0.16758(1) 0.16222(1) 0.14375(1)
800 0.18074(1) 0.17441(1) 0.16849(1) 0.16293(1) 0.14386(1)
810 0.18191(1) 0.17537(1) 0.16924(1) 0.16351(1) 0.14386(1)
820 0.18293(1) 0.17617(1) 0.16986(1) 0.16396(1) 0.14376(1)
830 0.18380(1) 0.17684(1) 0.17035(1) 0.16427(1) 0.14355(1)
560 0.25060(5) 0.28251(5) 0.31733(5) 0.35524(5) 0.54071(5)

698
Table 1(1.2.2)— Continued

A T(K)
(nm) 3800 4000 4200 4400 4600
ee ee a ee ee ee Re
300 0.64560(— 1) 0.86498(—1) 0.11269 0.14330 0.17842
310 0.82334(—1) 0.10809 0.13825 0.17288 0.21198
320 0.10290 0.13253 0.16661 0.20510 0.24791
330 0.12627 0.15975 0.19759 0.23968 0.28584
340 0.15242 0.18960 0.23096 0.27630 0.32537
350 0.18125 0.22191 0.26646 0.31462 0.36610
360 0.21262 0.25644 0.30377 0.35427 0.40762
370 0.24636 0.29294 0.34257 0.39489 0.44953
380 0.28224 0.33111 0.38253 0.43610 0.49146
390 0.32002 0.37067 0.42331 0.47756 0.53305
400 0.35943 0.41130 0.46459 0.51893 0.57398
410 0.40020 0.45270 0.50605 0.55990 0.61397
420 0.44203 0.49456 0.54737 0.60018 0.65275
430 0.48464 0.53658 0.58829 0.63953 0.69010
440 0.52774 0.57850 0.62854 0.67770 0.72584
450 0.57107 0.62005 0.66790 0.71451 0.75982
460 0.61435 0.66099 0.70615 0.74979 0.79191
470 0.65735 0.70110 0.74311 0.78341 0.82202
480 0.69984 0.74019 0.77863 0.81524 0.85008
490 0.74159 0.77808 0.81258 0.84520 0.87606
500 0.78243 0.81463 0.84486 0.87324 0.89993
510 0.82218 0.84971 0.87537 0.89932 0.92169
520 0.86068 0.88321 0.90406 0.92340 0.94137
530 0.89782 0.91505 0.93089 0.94549 0.95899
540 0.93348 0.94516 0.95583 0.96561 0.97460
550 0.96756 0.97349 0.97886 0.98376 0.98825
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000( 1) 0.10000(1)
570 0.10307(1) 0.10247(1) 0.101931) 0.10144(1) 0.100991)
580 0.10597(1) 0.10475(1) 0.10367(1) 0.1026% 1) 0.10181(1)
590 0.108691) 0.10686(1) 0.10522(1) 0.10377(1) 0.10246(1)
600 0.11123(1) 0.108781) 0.10661(1) 0.10468(1) 0.10295(1)
610 0.113591) 0.110521) 0.10782(1) 0.10542(1) 0.103291)
620 0.11577(1) 0.112091) 0.10886(1) 0.10601(1) 0.103491)
630 0.11778(1) 0.11348(1) 0.10974(1) 0.10646(1) 0.10355(1)
640 0.11961(1) 0.11471(1) 0.11047(1) 0.10676(1) 0.103491)
650 0.12126(1) 0.11578(1) 0.11105(1) 0.10693(1) 0.10331(1)
660 0.12275(1) 0.116691) 0.111491) 0.10697(1) 0.10302(1)
670 0.12408(1) 0.11746(1) 0.111791) 0.10690(1) 0.10263(1)
680 0.12524(1) 0.11808(1) 0.11197(1) 0.10671(1) 0.10215(1)
690 0.12625(1) 0.11856(1) 0.11203(1) 0.10642(1) 0.10157(1)
700 0.12711(1) 0.11891(1) 0.11197(1) 0.10604(1) 0.10092(1)
710 0.12782(1) 0.11913(1) 0.11181(1) 0.10556(1) 0.100191)
720 0.12840(1) 0.11924(1) 0.11154(1) 0.10501(1) 0.99395
730 0.12885(1) 0.11923(1) 0.111191) 0.10437(1) 0.98539
740 0.12917(1) 0.11912(1) 0.11074(1) 0.10367(1) 0.97628
750 0.12936(1) 0.11891(1) 0.110221) 0.10290( 1) 0.96666
760 0.12945(1) 0.11860(1) 0.1096 1(1) 0.10207(1) 0.95660
770 0.12942(1) 0.11821(1) 0.10894(1) 0.10118(1) 0.94613
780 0.129291) 0.11773(1) 0.10821(1) 0.10025(1) 0.93531
790 0.12907(1) 0.11718(1) 0.10741(1) 0.99274 0.92418
800 0.12875(1) 0.11655(1) 0.10656(1) 0.98257 0.91278
810 0.12834(1) 0.11586(1) 0.10566(1) 0.97205 0.90114
820 0.12786(1) 0.11511(1) 0.10471(1) 0.96121 0.88930
830 0.12730(1) 0.11430(1) 0.10373(1) 0.95010 0.87730
560 0.78751(5) 0.11048(6) 0.15010(6) 0.19835(6) 0.25590(6)

699
 Table I(1.2.2)— Continued

nN T(R)
(nm) 4800 5000 5200 5400 5600

300 0.21807 0.26225 0.31085 0.36376 0.42082

310 0.25550 0.30332 0.35528 0.41121 0.47088
320 0.2949] 0.34590 0.40067 0.45898 0.52057
330 0.33586 0.38949 0.44646 0.50650 0.56933
340 0.37789 0.43358 0.49214 0.55326 0.61665
350 0.42057 0.47772 0.53722 0.59877 0.66208
360 0.46345 0.52145 0.58128 0.64266 0.70528
370 0.50614 0.56438 0.62396 0.68457 0.74595
380 0.54825 0.60617 0.66491 0.72423 0.78387
390 0.58946 0.64649 0.70389 0.76143 0.81890
400 0.62946 0.68509 0.74067 0.79600 0.85091
410 0.66799 0.72177 0.77510 0.82785 0.87988
420 0.70485 0.75633 0.80705 0.85690 0.90579
430 0.73985 0.78867 0.83646 0.88314 0.92868
440 0.77286 0.81869 0.86327 0.90657 0.94859
450 0.80376 0.84632 0.88748 0.92724 0.96563
460 0.83248 0.87154 0.90911 0.94521 0.97990
470 0.85898 0.89436 0.92820 0.96056 0.99151
480 0.88323 0.91479 0.9448 1 0.97340 0.10006(1)
490 0.90524 0.93287 0.95903 0.98382 0.10073(1)
500 0.92503 0.94867 0.97095 0.99196 0.10118(1)
510 0.94263 0.96225 0.98065 0.99795 0.10142(1)
520 0.95809 0.97369 0.98826 0.10019(1) 0.10147(1)
530 0.97149 0.98309 0.99389 0.10039(1) 0.10133(1)
540 0.98288 0.99054 0.99764 0.10042(1) 0.10104(1)
550 0.99236 0.99615 0.99964 0.10029(1) 0.10059(1)
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.10059(1) 0.10022(1) 0.99884 0.99574 0.99288
580 0.10101(1) 0.10029(1) 0.99627 0.99020 0.98463
590 0.10128(1) 0.10021(1) 0.99239 0.98351 0.97537
600 0.10140(1) 0.10000(1) 0.98732 0.97577 0.96521
610 0.10139(1) 0.99668 0.98116 0.96708 0.95424
620 0.10124(1) 0.99221 0.97401 0.95754 0.94256
630 0.10097(1) 0.98668 0.96595 0.94724 0.93027
640 0.10060(1) 0.98109 0.95708 0.93627 0.91744
650 0.10012(1) 0.97283 0.94748 0.92470 0.90416
660 0.99547 0.96467 0.93722 0.91263 0.89048
670 0.98888 0.95580 0.92639 0.90010 0.87649
680 0.98149 0.94627 0.91505 0.88720 0.86223
690 0.97338 0.93617 0.90326 0.87397 0.84777
700 0.96462 0.92556 0.89108 0.86048 0.83315
710 0.95527 0.91449 0.87858 0.84677 0.81842
720 0.94539 0.90303 0.86580 0.83289 0.80362
730 0.93504 0.89122 0.85279 0.81889 0.78878
740 0.92427 0.87911 0.83960 0.80480 0.77395
750 0.91314 0.86676 0.82626 0.79066 0.75916
760 0.90168 0.85419 0.81282 0.77650 0.74442
770 0.88994 0.84146 0.79930 0.76235 0.72978
780 0.87797 0.82860 0.78573 0.74824 0.71523
790 0.86580 0.81563 0.77216 0.73420 0.70082
800 0.85347 0.80260 0.75859 0.72023 0.68655
810 0.84101 0.78953 0.74506 0.70637 0.67245
820 0.82844 0.77643 0.73159 0.69262 0.65851
830 0.81581 0.76335 0.71820 0.67901 0.64476

560 0.32327(6) 0.4009 1(6) 0.48914(6) 0.58822(6) 0.69829(6)

700
Table I(1.2.2)— Continued

A T(K)
(nm) 5800 6000 6200 6400 6600
300 0.48185 0.54663 0.61493 0.68651 0.76114
310 0.53406 0.60050 0.66995 0.74216 0.81686
320 0.58518 0.65254 0.72238 0.79443 0.86844
330 0.63466 0.70222 0.77172 0.84292 0.91555
340 0.68202 0.74910 0.81762 0.88733 0.95800
350 0.72686 0.79285 0.85980 0.92748 0.99569
360 0.76889 0.83324 0.89811 0.96330 0.10286(1)
370 0.80787 0.87010 0.93246 0.99476 0.10569(1)
380 0.84364 0.90335 0.96283 0.10219(1) 0.10806(1)
390 0.87612 0.93296 0.98927 0.10449(1) 0.10999(1)
400 0.90527 0.95896 0.10119(1) 0.10639(1) 0.11151(1)
410 0.93110 0.98141 0.10308(1) 0.10791(1) 0.11264(1)
420 0.95365 0.10004(1) 0.10461(1) 0.10906(1) 0.11340(1)
430 0.97302 0.10162(1) 0.10581(1) 0.10988(1) 0.11383(1)
440 0.98932 0.10288(1) 0.10669(1) 0.11038(1) 0.11395(1)
450 0.10027(1) 0.10384(1) 0.10728(1) 0.11059(1) 0.11379(1)
460 0.10132(1) 0.10452(1) 0.10759(1) 0.11054(1) 0.11337(1)
470 0.10211(1) 0.10494(1) 0.10765(1) 0.11024(1) 0.11272(1)
480 0.10265(1) 0.10512(1) 0.10747(1) 0.10972(1) 0.11186(1)
490 0.10296(1) 0.10508(1) 0.10709(1) 0.10900(1) 0.11082(1)
500 0.10306(1) 0.10483(1) 0.10651(1) 0.10810(1) 0.10961(1)
510 0.10295(1) 0.10440(1) 0.10576(1) 0.10705(1) 0.10826(1)
520 0.10267(1) 0.10379(1) 0.10485(1) 0.10585(1) 0.10679(1)
530 0.10221(1) 0.10304(1) 0.10381(1) 0.10453(1) 0.10522(1)
540 0.10161(1) 0.10214(1) 0.10264(1) 0.10311(1) 0.10355(1)
550 0.10087(1) 0.10112(1) 0.10137(1) 0.10159(1) 0.10181(1)
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.99024 0.98780 0.98553 0.98341 0.98144
580 0.97950 0.97476 0.97037 0.96629 0.96250
590 0.96789 0.96100 0.95463 0.94873 0.94325
600 0.95553 0.94663 0.93842 0.93084 0.92380
610 0.94250 0.93173 0.92183 0.91269 0.90423
620 0.92891 0.91641 0.90493 0.89436 0.88460
630 0.91483 0.90073 0.88781 0.87593 0.86498
640 0.90035 0.88477 0.87052 0.85745 0.84542
650 0.88554 0.86861 0.85315 0.83899 0.82598
660 0.87046 0.85229 0.83572 0.82058 0.80669
670 0.85518 0.83587 0.81831 0.80227 0.78759
680 0.83975 0.81941 0.80094 0.78411 0.76871
690 0.82422 0.80295 0.78367 0.76612 0.75009
700 0.80863 0.78652 0.76652 0.74833 0.73175
710 0.79302 0.77017 0.74952 0.73077 0.71370
720 0.77744 0.75392 0.73270 0.71346 0.69596
730 0.76191 0.73781 0.71608 0.69642 0.67856
740 0.74647 0.72185 0.69969 0.67966 0.66148
750 0.73113 0.70606 0.68353 0.66319 0.64476
760 0.71593 0.69047 0.66763 0.64703 0.62838
770 0.70088 0.67510 0.65199 0.63118 0.61237
780 0.68600 0.65995 0.63663 0.61566 0.59671
790 0.67130 0.64504 0.62156 0.60045 0.58141
800 0.65681 0.63037 0.60677 0.58558 0.56648
810 0.64252 0.61597 0.59227 0.57103 0.55190
820 0.62846 0.60182 0.57808 0.55681 0.53768
830 0.61462 0.58794 0.56418 0.54293 0.52382
560 0.81942(6) 0.95163(6) 0.10949(7) 0.12490(7) 0.14140(7)

701
 Table I(1.2.2)— Continued

r T(K)
(nm) 6800 7000 7200 7400 7600

300 0.83856 0.91852 0.10008(1) 0.10851(1) 0.11713(1)

310 0.89379 0.97272 0.10534(1) 0.11356(1) 0.12191(1)
320 0.94415 0.10213(1) 0.10997(1) 0.11792(1) 0.12595(1)
330 0.98939 0.10642(1) 0.11398(1) - 0.12160(1) 0.12926(1)
340 0.10294(1) 0.11014(1) 0.11737(1) 0.12462(1) 0.13188(1)
350 0.10642(1) 0.11329(1) 0.12016(1) 0.12702(1) 0.13384(1)
360 0.10939(1) 0.11590(1) 0.12238(1) 0.12882(1) 0.13521(1)
370 0.11186(1) 0.11799(1) 0.12407(1) 0.13007(1) 0.13601(1)
380 0.11386(1) 0.11959(1) 0.12525(1) 0.13082(1) 0.13631(1)
390 0.11540(1) 0.12073(1) 0.12597(1) 0.13110(1) 0.13614(1)
400 0.11653(1) 0.12144(1) 0.12626(1) 0.13097(1) 0.13557(1)
410 0.11726(1) 0.12177(1) 0.12617(1) 0.13046(1) 0.13463(1)
420 0.11763(1) 0.12173(1) 0.12573(1) 0.12961(1) 0.13338(1)
430 0.11766(1) 0.12138(1) 0.12498(1) 0.12846(1) 0.13184(1)
440 0.11740(1) 0.12073(1) 0.12395(1) 0.12706(1) 0.13006(1)
450 0.11686(1) 0.11983(1) 0.12268(1) 0.12543(1) 0.12808(1)
460 0.11608(1) 0.11869(1) 0.12120(1) 0.12361(1) 0.12592(1)
470 0.11509(1) 0.11736(1) 0.11953(1) 0.12162(1) 0.12362(1)
480 0.11309(1) 0.11585(1) 0.11771(1) 0.11949(1) 0.12120(1)
490 0.11254(1) 0.11419(1) 0.11576(1) 0.11726(1) 0.11869(1)
500 0.11104(1) 0.11240(1) 0.11370(1) 0.11493(1) 0.11610(1)
510 0.10942(1) 0.11051(1) 0.11154(1) 0.11253(1) 0.11346(1)
520 0.10768(1) 0.10852(1) 0.10932(1) 0.11007(1) 0.11079(1)
530 0.10586(1) 0.10647(1) 0.10704(1) 0.10758(1) 0.10809(1)
540 0.10396(1) 0.10435(1) 0.10472(1) 0.10506(1) 0.10539(1)
550 0.10200(1) 0.10219(1) 0.10237(1) 0.10253(1) 0.10269(1)
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.97960 0.97788 0.97627 0.97475 0.97333
580 0.95896 0.95565 0.95256 0.94966 0.94694
590 0.93815 0.93340 0.92896 0.92480 0.92090
600 0.91727 0.91119 0.90552 0.90022 0.89526
610 0.89639 0.88910 0.88231 0.87598 0.87006
620 0.87557 0.86718 0.85938 0.85212 0.84533
630 0.85486 0.84548 0.83678 0.82867 0.82111
640 0.83432 0.82405 0.81452 0.80567 0.79742
650 0.81399 0.80291 0.79265 0.78313 0.77427
660 0.79390 0.78211 0.77120 0.76108 0.75168
670 0.77409 0.76166 0.75017 0.73953 0.72965
680 0.75459 0.74159 0:72959 0.71849 0.70819
690 0.73541 0.72191 0.70947 0.69797 0.68731
700 0.71657 0.70264 0.68981 0.67797 0.66700
710 0.69810 0.68379 0.67063 0.65849 0.64726
720 0.67999 0.66536 0.65192 0.63953 0.62808
730 0.66227 0.64736 0.63368 0.62109 0.60946
740 0.64493 0.62980 0.61592 0.60316 0.59140
750 0.62798 0.61267 0.59864 0.58575 0.57387
760 0.61143 0.59597 0.58182 0.56883 0.55687
770 0.59528 0.57971 0.56547 0.55241 0.54039
780 0.57952 0.56387 0.54957 0.53647 0.52442
790 0.56415 0.54846 0.53413 0.52101 0.50895
0.54918 0.53346 0.51913 0.50601 0.49396
0.53459 0.51888 0.50456 0.49146 0.47945
0.52039 0.50470 0.49042 0.47736 0.46540
0.50656 0.49092 0.47669 0.46370 0.45179
SS

0.15895(7) 0.17754(7) 0.19715(7) 0.21775(7) 0.23932(7)

702
TS

Table (1.2.2)— Continued

d T(K)
(nm) 7800 8000 9000 10000 00
300 0.12590(1) 0.13482(1) 0.18080(1) 0.22766(1) 0.12141(2)
310 0.13037(1) 0.13892(1) 0.18241(1) 0.22588(1) 0.10649(2)
320 0.13404(1) 0.14217(1) 0.18303(1) 0.22314(1) 0.93789(1)
330 0.13694(1) 0.14463(1) 0.18279(1) 0.21961(1) 0.82927(1)
340 0.13912(1) 0.14634(1) 0.18177(1) 0.21542(1) 0.73593(1)
350 0.14063(1) 0.14738(1) 0.18009(1) 0.21071(1) 0.65536(1)
360 0.14153(1) 0.14779(1) 0.17785(1) 0.20558(1) 0.58552(1)
370 0.14187(1) 0.14764(1) 0.17512(1) 0.20013(1) 0.52474(1)
380 0.14170(1) 0.14700(1) 0.17199(1) 0.19446(1) 0.47165(1)
390 0.14108(1) 0.14592(1) 0.16854(1) 0.18863(1) 0.42510(1)
400 0.14007(1) 0.14446(1) 0.16482(1) 0.18271(1) 0.38416(1)
410 0.13870(1) 0.14266(1) 0.16091(1) 0.17676(1) 0.34803(1)
420 0.13703(1) 0.14059(1) 0.15683(1) 0.17081(1) 0.31605(1)
430 0.13511(1) 0.13827(1) 0.15265(1) 0.16490(1) 0.28766(1)
440 0.13296(1) 0.13576(1) 0.14840(1) 0.15907(1) 0.26239(1)
450 0.13063(1) 0.13308(1) 0.14411(1) 0.15333(1) 0.23983(1)
460 0.12814(1) 0.13028(1) 0.13981(1) 0.14772(1) 0.21964(1)
470 0.12553(1) 0.12737(1) 0.13552(1) 0.14223(1) 0.20154(1)
480 0.12283(1) 0.12439(1) 0.13127(1) 0.13690(1) 0.18526(1)
490 0.12005(1) 0.12135(1) 0.12707(1) 0.13171(1) 0.17060(1)
500 0.11722(1) 0.11828(1) 0.12294(1) 0.12669(1) 0.15735(1)
510 0.11435(1) 0.11520(1) 0.11888(1) 0.12183(1) 0.14537(1)
520 0.11147(1) 0.11211(1) 0.11491(1) 0.11713(1) 0.13451(1)
530 0.10858(1) 0.10904(1) 0.11103(1) 0.11260(1) 0.12464(1)
540 0.10570(1) 0.10599(1) 0.10725(1) 0.10824(1) 0.11566(1)
550 0.10284(1) 0.10298(1) 0.10357(1) 0.10404(1) 0.10747(1)
560 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1) 0.10000(1)
570 0.97199 0.97072 0.96534 0.96118 0.93165
580 0.94438 0.94197 0.93176 0.92390 0.86904
590 0.91724 0.91379 0.89926 0.88812 0.81160
600 0.89060 0.88623 0.86783 0.85380 0.75883
610 0.86451 0.85930 0.83746 0.82088 0.71029
620 0.83898 0.83303 0.80814 0.78932 0.66556
630 0.81404 0.80743 0.77985 0.75908 0.62430
640 0.78971 0.78251 0.75257 0.73010 0.58618
650 0.76600 0.75828 0.72626 0.70233 0.55093
660 0.74292 0.73474 0.70092 0.67573 0.51829
“670 0.72046 0.71188 0.67651 0.65026 0.48804
680 0.69862 0.68971 0.65301 0.62586 0.45996
690 0.67742 0.66821 0.63038 0.60249 0.43387
700 0.65683 0.64737 0.60860 0.58011 0.40960
710 0.63685 0.62718 0.58765 0.55868 0.38701
720 0.61748 0.60764 0.56748 0.53814 0.36595
730 0.59871 0.58873 0.54808 0.51848 0.34631
740 0.58051 0.57042 0.52942 0.49964 0.32796
750 0.56289 0.55272 0.51147 0.48159 0.31082
760 0.54583 0.53560 0.49420 0.46429 0.29478
710 0.52931 0.51905 0.47759 0.44772 0.27976
780 0.51332 0.50305 0.46162 0.43183 0.26569
790 0.49785 0.48759 0.44625 0.41660 0.25249
800 0.48288 0.47264 0.43147 0.40200 0.24010
810 0.46840 0.45820 0.41724 0.38800 0.22846
820 0.45439 0.44425 0.40356 0.37458 0.21752
830 0.44085 0.43077 0.39040 0.36170 0.20722
ee ee ees ee ere ee
560 0.26183(7) 0.28525(7) 0.41503(7) 0.56353(7) 00

703
 Table I(1.2.2)— Continued

r T(K)
(nm) 2042 2790 2856
300 0.41642(—3) 0.77475(—2) 0.93165(—2)
310 0.75400(—3) 0.11449(— 1) 0.13593(—1)
320 0.13089(—2) 0.16429(— 1) 0.19272(— 1)
330 0.21871(—2) 0.22955(—1) 0.26625(— 1)
340 0.35299(—2) 0.31309(—1) 0.35930(— 1)
350 0.55204(—2) 0.41776(— 1) 0.47465(—1)
360 0.83880(—2) 0.54640(— 1) 0.61495(—1)
370 0.12414(—1) 0.70170(—1) 0.78271(—1)
380 0.17932(— 1) 0.88621(—1) 0.98018(—1)
390 0.25334(— 1) 0.11022 0.12093
400 0.35066(— 1) 0.13516 0.14717
410 0.47627(—1) 0.16360 0.17685
420 0.63567(— 1) 0.19567 0.21005
430 0.83478(—1) 0.23144 0.24682
440 0.10799 0.27095 0.28714
450 0.13776 0.31420 0.33098
460 0.17348 0.36115 0.37824
470 0.21581 0.41170 0.42881
480 0.26546 0.46573 0.48254
490 0.32310 0.52310 0.53924
500 0.38937 0.58363 0.59871
510 0.46491 0.64709 0.66073
520 0.55028 0.71327 0.72504
530 0.64603 0.78192 0.79139
540 0.75264 0.85277 0.85952
550 0.87051 0.92556 0.92914
560 0.10000(1) 0.10000(1) 0.10000(1)
570 0.11414(1) 0.10758(1) 0.10718(1)
580 0.12949(1) 0.11527(1) 0.11443(1)
590 0.14606(1) 0.12305(1) 0.12172(1)
600 0.16386(1) 0.13087(1) 0.12903(1)
610 0.18289(1) 0.13873(1) 0.13633(1)
620 0.20314(1) 0.14658(1) 0.14360(1)
630 0.224591) 0.15442(1) 0.15081(1)
640 0.24723(1) 0.16220(1) 0.15795(1)
650 0.27101(1) 0.16992(1} 0.16499(1)
660 0.29591(1) 0.17755(1) 0.17193(1)
670 0.32189(1) 0.18507(1) 0.17873(1)
680 0.34891(1) 0.19246(1) 0.18538(1)
690 0.37690(1) 0.19971(1) 0.19188(1)
700 0.40583(1) 0.20680(1) 0.19820(1)
710 0.43563(1) 0.21372(1) 0.20435(1)
720 0.46624(1) 0.22046(1) 0.21030(1)
730 0.49760(1) 0.22699(1) 0.21605(1)
740 0.52965(1) 0.23333(1) 0.22159(1)
750 0.56231(1) 0.23945(1) 0.22692(1)
760 0.59553(1) 0.24535(1) 0.23203(1)
710 0.62924(1) 0.25103(1) 0.23691(1)
780 0.66337(1) 0.25647(1) 0.24158(1)
790 0.69784(1) 0.26168(1) 0.24601(1)
800 0.73260(1) 0.26665(1) 0.25022(1)
810 0.76758(1) 0.271391) 0.25420(1)
820 0.80271(1) 0.27589(1) 0.25795(1)
830 0.83793(1) 0.28014(1) 0.26148(1)
560 0.23322(3) 0.68046(4) 0.84186(4)
“Note: The computed values of spectral radiant exitance (M,,) are given to five
significant figures followed by the power of ten (in parentheses). The decimal
point is placed consistently after the first figure. All values are normalized so that
M,, at \=560 nm becomes unity, that is 0.10000< 10'. The absolute value at 560
nm is given at the bottom of each column; the unit is watt per square meter per 10
nanometer wavelength interval: W-m 7-(10 nm) |.
704
Table 1(1.3.2) Spectral Transmittance Curves of Major Types of Commercially Available Colored Glass
Filters. Each Type is Illustrated by Examples

Type 1: Sharp Cutoffs

Sharp cutoffs are glass filters with high spectral transmittances over a wide wavelength range with an
abrupt drop to zero transmittance at the short wavelength of its transmitting range. The cutoff
wavelength can be anywhere from below 200 nm to about 700 nm. Manufacturers offer sets of such
filters with cutoff wavelengths approximately uniformly spaced through the ultraviolet and visible region
of the spectrum. An example is the set offered by Schott (Jenaer Glasswork Schott & Gen., D-6500
Mainz, Germany) containing the series WG 230 to WG 395, the series GG 400 to GG 495, the series OG
515 to OG 590, and the series RG 610 to RG 715, with a total of 28 different cutoff glasses. Other
manufacturers offer sets with similar properties.

Type 2: Shallow Cutoffs

Shallow cutoffs are glass filters with spectral transmittance properties similar to those of the Type 1
glasses but their cutoff in transmittance at the short wavelength is less sharp.

Type 3: Neutrals
Neutrals are glass filters with spectral transmittance curves approximately horizontal across the
visible part of the spectrum. They are mainly used as “nonselective” attenuators of visible radiant
energy. Manufacturers offer sets of neutrals with average transmittances ranging from a fraction of 1%
to over 80%.

Type 4: Broad-Band Through Visible

Broad-band-through-visible glass filters transmit the visible spectrum in a single broad-band with a
relatively high peak transmittance. However, all of these filters also transmit in the infrared, sometimes
significantly.

Type 5: Broad-Band in Ultraviolet

Broad-band-in-ultraviolet glass filters transmit the ultraviolet region of the spectrum in a single band
and are nearly opaque through most of the visible spectrum. However, all of them also transmit by
varying amounts in the infrared.

Type 6: Sharp Broad-Band in Blue

Sharp broad-band-in-blue glass filters transmit in a single band the short-wavelength region of the
visible spectrum. Most of these filters also transmit part of the ultraviolet, deep red, and infrared region
of the spectrum.

Type 7: Bluish Broad-Band

Bluish broad-band glass filters transmit predominantly in the short-wavelength region of the visible
spectrum and less in other regions. However, all of them transmit also in the infrared, often strongly.

Type 8: Broad-Band in Green

Broad-band-in-green glass filters transmit the middle region of the visible spectrum in a single band.
All of them transmit also in the infrared.

Type 9: Purples
Purple glass filters transmit in the short and long-wavelength regions but little or not in the middle
region of the visible spectrum. Source filters of Type 5 (broad-band in ultraviolet) are similar to purples
in their spectral transmittance characteristics.

Type 10: Multiple Peaks

Multiple-peak glass filters have spectral transmittance curves with a number of distinct minima and
maxima distributed across the spectrum.

Type 11: Infrared Windows

Infrared window glass filters transmit mainly in the infrared region of the spectrum.

Type 12: Heat Absorbers

Heat absorbing glass filters transmit in the ultraviolet and visible parts of the spectrum but transmit
little or not at all in the infrared.
er

705
706
Ol
UCUULGDCOYUUGLOUULDYLUYUULGLDLUUUYDOLUULUGULUULGDCGUULCLUULUULUYUG Wa Ub Wt Us Wa. Wh Ut. Us Ws Wa Ws Wt Wt, Wn UG Wo UW Yo Ub Wah Wa Ua Walls We YUVA Uo Wa
OOO aa ee MMé@€ MalN]g<coéoyyl opp Om OP PPI IO OI OO YA AO V AAG
yy) yyw nT : sceaasseest YYYUGYLYY) WW UH HW” VH@HQ@H@ H@WWWUUUCGUWUUUMUMUUCUUUMHULL THA.
aaee ALI eaae oh | MAAAAAAAADAAGG UWUMUUUUVUUVUOMUCACUPREEREERS RS VYAAG4AYGT

A
NNN

80
i]
ree A AAA ABBA Waa
WU
Wa
Ut
Ws
Un
aWb

UU
GO
UWI YWUYWMI WWM WUMOULLTA- U0. 7 i

YY
YUWWIWOWa MW UWUUUWO WY, sy.
YW WWae@/@ Y@~@ WY). ane a

aS|
a=

90
Eee

:
:
eo
(eo

“i
YM

“A

OAM
]
a8xi
ee

OOOO
ae
|

alae

YOLZALLLZZZLY
ar
ae

WS

AAAO
OOOO
AMI UUUUYUGUYLYUYBYUYLOLELSY
YY
YYEYYYULYYYEY
VLYYYYY

cease
OIC.

CYY AAP
20 ¢'0 v'0 L0

SScscce
NCLCA OOO

SSeeceeueesa:
COUVULULGLUULODGUYAYGGDGUUEDOYLYGUUULUY

UUGUYGGYGYLGYLDGYLYYLYLYUEGGYYS
UUGYYYVULGYUYUYGLYEULS
Ht
UGUYLLYUYULECCUYULLULUULYULUULY

ts
Ut
Seas LALAAATZ

a
th
Wn
Ws
LUYULYEYEYYS

Us
Ue
Ub
Uh
Ya
4Wa
Un.
Ut,
YWUWiWWUCMUCMCOUCTULAY

Wa

AAA U4 VAMC
W@W

CUULOUULUDUULULLUULOLYS
YUYYUYGYYLYYGGGLYYGUGLLYGSGS:
ay

AMMA
Wan
VedPDI
AA

asma
YY
—VUULULBULYDUYLUOUBUALUYGUGULLGSLEUZA Pe

WMV
4

CULYUYLCUCULYYUYUCULLECLLZLLLLLLY
YEU cal ert

SANS
ts

Ut.
tu
Ys

UGUYYYYGGGYOLGBULEGEGES
UU04YYUUYYYLYGGOIGEYGYYELGGYLLYYAYS
Ws

WWW
MMU

OOO BBUYYYYYYYYYYLYGLYELYGGES
05S
MBI

YYUUULUYYYGLGY sate

WYUCYUYLYUGLYYULYYGYYYOUS
YUYYYYYYUGYYUYUYLYLUYYLYYY
GZ

WUWYUUULYLBGGEYLYIGLGYEYSES
ao09
WUUYLDDLCADOLEGUDYLODYYYOYUYLYLYGUA ++
|
OOM
OMIM
4
WWM

|OOOO
Wy@A
VA
VPA

UULULUUUULULULYDULUULULUUULULLS
P@

scc22
MOV

PU—-@C
e=-
e@

722222
UUOUGYEBLOVYEBLLAYLEYLAYELYYLEGS
\WYAUYGG4GaGy Mm
POJJOP—JJJJJI@_
DORI

“4GULGYLUGGZLZGZGGGY LYUYYUYYGYYYGYGYYYY a |I ===: sae

7 iN OOOO, MOO MM
UGUUUULYLYODDUYYYLYYYELYDL
1 YDYLBEY MMM UM WM MUUWM
WCOYWUCUU0Y

GYYUYYGYGGLYYELYYYELYLYELLYEYELY
2
VWWIWW

Y
ORIOLE

WUYYYYYYYLEYYLLYYYLLLOLLLESEEOE
| L
; Ms 4—-t Ff4 Se ttt
Ga ete t,t, MMII.
YYYYYGUGUYYGLYGYYEYLLYYEYLYGYLYYGEY
UUUUYUGLLYLYLOLYDYDYYYYYYEDYDYLYUY
OOULYYEOLYLOULOLLLEEA
GULLY UYUYYYY is
OM OOM
OY)

UUGYLUUGYYYGYGYLYUGUYGYLYGUGYGULLYGYGGLGUGGLS
IeIa |
20 £0 vO s'0
HHI©SSSSSSeeceesees
HAA FSSATRERSTEIEIS

SSeumeRnRESaeE
YYW
YYYWYWYYYYYUYYAY

(X¥)2 9OUD}}IlUSUDI|
WUGUGUGUGUUGUGYLUUUY
O'l
UCGUVUUGCGULUUULCLLUUGGOGUALUUUGGLULDGOYY LAL AACA AAALAC
ELYDALEY
GLAD YOYDE YOULL OLYYY YY YOUUYY YY YYYYY II “MOO
YYYGLYYYSYYYYUAY
IGGL LL4L4 et
WY OUDALAGA
8'0 BYYZUULYUYYYGLYYGLYLYYUGGYZGYGYLUGGGLLYGGGLGA
GYYYYYYYYYYOUDLYD ©YULYYYYLY
ALDUDAY
YYGYYOLYYSOLYYYYBYYLLLLLYYAGEYGAY
YYYLLLGLLLLLEY MOO.
YYGYYYUYGYLULYGYGUGGSY SYYYYGLYYYUYYOYYLOULULLYL IEA YYYYYLI
9'°0 EUEUYUYULYYUYUYUYYUYLYUYUYUBGYLY
BYUAZYYYYLYUGLUGLYLYGUGULLS
WIUHMHPWHUMPWHUWOUCUACUCUUCUUHOA

OUAUE OLLAYELODLDELOLD
ALYY COOL LDYGALDYY
SULLY
YYYLYYYYEYYYYYYEYYYYYOU
YLYGY YYYYG BELLY LYALIAYAA
OOOO OO
ae) YYUYUYYYYYYYGGYYYUGGYS YYBYYYYYYYYYOY
LY BZZ YYGUYGLY
OOO OOOO A
YULULAY YYGUYGYYUYYYYY
AAC LTLTAAA LAC
|
YYYYEYYYLOLGEY
ILC LUE UAE AGEL YULGUGae AaaWu WU Y)YUYJ
ALAAMA MAA MMA MMMM
AM YYYUGYUYYYGYYGYLYLGGLGY YYYWYwwwWHPaey
OOOO
OO OOOO OOOO MAA MMMM WyywwywWa
yyy

20
OLOLLOLOOL YGYGYGYUOLLYG
L004 OYOOLYLYLYYOLLOL
OOOO OGUYLGGLOLETU
GO LYY LLOLULE
o'0 YWYUYYYYYYGYGGUY
YYYUYYUYYYGYYUYGYGYYLYYYZGYLYS
2'0 £0 v0 sO 9'°0 20
YYYWMMYMWJWJTHMM@

(Ww!) YyjbuajannM
 WG OEM
W@ We WW Wt Wt We We We Wa Uta Wa Ui Wea Ws WeUe We
DVO
UVE DUA MUU
MA WM MMA MU UYUYU
ees ee Pe WOW YULUMWMY

hp
YP

LAA
ULLUEYLLLEL
a2ee

YY

30
UGLGLLLOUE UUUYULYULYYUGYDYLYLYG
LY
GYUGUA UZLLZUY4ULUYLUGLULUYUYYGUYGULULIYYGY = (mes

=
(0.2 404

4304ND
11 WsWMOa4d0i00
Ut#WI Wh
AULULYLUUGULLULULUGULLUUUYILULULUUGS
DY

OAUIEHS
La z= ma LA LIL YUWYYYIE

SS
YYYYEY
BOUYUZAZAZLUUCLLULULUUULUUUUGUUUst Raga es ea hs hh ha Ub Oh MUM UUM
VUZLULLLULUYLGUULLLYGUULULUYULGUYS fes|Ena eal (0). YUa@00@
WU I O“0“[10MM YUWM
BYuAULLLYLULULUULLLUYUUULULULUGUUS

GCUULULUUYGT
011.0 €U U0 UYU UU
BYAUULLYYGUULGGYUULUYULULUUYY sso ae LLL YYYYLIL
DYUAZULULUU4aUUUasguUuUUUuUugns ANN UMAUMUAYUMY

LYYYYYUYLYY
VUULYUGLUULULYULULULULUUYLLUULS Ht ns # WhI UUUUIMMU“U
ACALARCACL LAQOODLTILAARAACAAAC
| 07, UWUWWUUMUY

MOO
VAZLZYAZLUYLZLLLGYGUUGLULLIYSY UU“ UWUUMUAUWUMM

PAPEL
DYYLZLULULULUULULULULULULAL UMMUMMUT
DIED
GALE YOGLD (U3
GZ

LITT TTT TTT TTT

(exe)
oe

UUULULGULUULULULUDULULULUYYY
To
Oo
% C0CGGGCUGDGDLYLULLGGLGLLYGGLLULLDGGOLUGLYLGGDLULGLLULULOLLLEEE
YYUGGUGLYDLOLLYLLDLLYLYLOLY
Z

= mojjoug ELELYDELOLEEYIIIOE
HO
YYLYOLOLYYYOULYYYY UCYGLYUYUG ULYGYGLY
LLYYALOLY EE
0

80 Ya 4404ND WCUUUGUYYYCLULYUULYY

OOO
ZA GYGUYLUDGGYYGLCLGYYGGELGYLEYEGEE:
LULULYLLGO
LOLLLGEDLODOLDEDLLELLODS VA LLLEDLLLYD
HE
iA

SSSEDPLULYYLLLLLCLGLLYLYDLYYELLLDYEGYEDYEDLYYLLIDLYYEGEGYEGS
OLGLLYGL
GLY

OOOO
YGEGLEYL
ODDO
Bib OLELLGIOLGEIOLIY

Ae
YYYYYYYLY
GYGLYYOOGLYYYOOGLYGL Y
BOULOYYLOLYLLY ILIELIIEE
OMIM

EYG

OO
ard ULLLLLYLD BYLLLLLLL
YLLYLEOLD LL EIA
OOLLLELLOE
OODLE LYELL GYYYYYOYY
LLYLY E EBYZZY4 OLELLOEDO

YUYUYYYGUYYYYYUYYBYYYYLDLULYULGULYDEDS
UGGUYULULY
ELE LCL
L ULUUGLYUGL

CUUUGCLDUULUCLLUUCUGDGLUGLUGDLUGLUGDULGG
MeL YLDYLODE
UUGYUYYOLLYYYYLOGYYLYGGES
vo YUYYCLLL
4 to YLDDOLLOLOLLE
GUUUGEULLG
(U4 Ws L040 UH HL UUUUUM
LYLUYULLLLYLGS 00-0 YUUUUMUMUUU
OYGDLLLLU
GUULY Y LYG Y 0 AY YUUMUU
WUYUYGLYGULYYGDGYLGYLYLGYGS UW UMUUWO
WV
DOODLE.
20 LULLED ULYGGOLLL UDLLLLYY
CUGG YLYYUCLLOYODULLYY
UUUUULCLLUULYYGEGYLGA
© © LO A
20. £0 me) ome) A

(X)2 90uUd}JIlWSUDI|
Aa UUCUUCLULULULLDCUULD
Mignon YYUOYUYLYYYYOULYLELLGUULLULY UUUBDULUUU
aSBUOys, ULYLLLOLL
GU YL: LYYLLOLYLLELELELELUG
Ls

9°0
4 $404ND
YLGGOYGYOYULYYLYLLYGYLLYOOG LLY
OLLLLOLY:
YOULL LYYYLGYYL “YYYGUYYY
YYYLYLYYO S
LLYY
OO

LAER LYELL LYLOLLYLO

DELLE
L GLYLE GULLY YOYLYYYYYY OO
0008

Y’YUGELYLL
OILED S LOCLLOLLL OO OOP Mm
OOOO

LOLYLOLDY
YLULYGLYU
ELOLLOL
CHO

YOLELYYOLO DOLLY
GOLLYLLEE: LYGLOY “UMPUMUUUUY
WY

9'0 UYULLLOLO
GO LYYGYYYYGYOYLYY UUYLULGUYLYGGLGGLS Ae Ee LL OTOL WWII
LL
LEYLLYLLLLY YYYLYYGYL LLYOYUYY
GLY YY ULLYYLLY aa
YA OE DO E
LMU

LY, TLE LYYYYYYYLYLOYLLY UYYGGUYYUYDLGYLYYYYLYELY

ULYYLLYAYI
LL A
bot
A

YLUYGYLYYUGYYLUGLY YUYYYYOYUYYYLYY
WEG
thle

GYWYYGYUUYVOGLYIGGGY YYYYYGYLY
WILY
GGI4VERLS YYGGLGS
OGLLYULELY YLYYYYLLOYUULGULYLUL
LUD
ULLLULUULL DULLOLE
VULLYDOUYDOUS
ALA CT A
ae WUUULGLYYG
YYYGYODLYLEYOULYLOO4 “YYGLYYLL
CI YYYYYLOUY
H Cy e
AA AAAAAA AMMA MMM oan
IUIIIIOIE
III LILEO
N OLE GLUTGUL OLA BBY BG
GUYLYULOCLLLIDS GLLYYLGYYYYYYYYUYYY ei
ee 4 LYYYYOLLLY
YU YYYYLYLOYY
BBDBR TH =
20. <0 v'O roa)

XY
}Huajaann (wu
)

707
 JOudjaADM

(w77)X Y

U’GGUGLYYLGGS
GUGUUUUGYY

CCC O
2
OAS EE 777A

OLA
YUAVNUU

B
Ye |TT

oh

ago
Syy
"UUW GVGDG,

a.
SES GE RES BEEBE BBRBY00000777700707
rea ee CLYYLYUULYGYYLEEGES
7,77 a o Cl WHA
7,
a YY, BD >So>> Senn MOM OOOO.
CT OOO Ang LBUDOULLGUYUYGGULLLLA YUL U4 L%,
Waa Wit WY OLL
OLY
LLDIVEL
LILLE
TLE
LOOTED
VION
YGYUYL GODLY YGYLOGYLOLEODYOYELYODYGLEYGYLYYLOYIGLGYLEYYGYA LUT
ELLE
ELL
ORDO
GUYGGCLOYYYYLGOUYGYDEGYLYYGOGEYYYYODYLYYEGGDYEYEGGEYGGG4
90
a LULUUL
UULULLUULULYULULOULUUGULLLLULULULUGUELLGULUY
4

80
YUYYYYYUGYYOLYYYYOGYLYODGGEGYGYYGYGLLYGLGELLEY: = % UUULLUUULULELULOULUULELLUULUULDELLLULUOLULUELUL
YUUUYYYYUYYYLLYEGEYYYGYYDLELYYLYGEGLLGYLYYEEEYA ee
4
UDULUYUULLUUULULUDUDUULUULUULLULULLULULUULUULULLYUGLULULLLLULDGSL
WUUMUM UM Ua Uh Ub Ua Uh Ut Uh Ua Wt Ua Ua. Ua Wn Wt Us CUUUGGCAULUGCLLLYLUCGLULUGYGULULYUUDUOLUUUGCLUULCLGUUUUUGE, Ol

(X)2 90UD}fIWSUDI|
o¢ oz 0!

ro)
Ov

°
LOU YY ae 7 7 a a a HWY7FA BOG8 SRRERREBEBYYZZZZZZZ22
YYLYLOCYULL Fee a |MAG D
s -PULYULYLUGULUGYYGUES
11
Bre | OAC AAALAC
= ra

ate De MAO
|__| YG. | | BRP, AAA AOC OOOO OOO
TOO YY CRVOLUGULEY UUULDGYGYGYYELEYYGLEYYYEDDYYLYDDLEYGGS

S
MAE UU UULUUULUUCUCUUULDCLLUUYLUULLULUYUGUOUYUULLULUUUULUYLYUYUUYUUUYYDDLY
YLYYLYGYLGYGGYA
YYGUYGUYGYYLOYYYGEGOGGGEGLGGGGYGLEDGEGYYGELY GOO COCO OO
WOOO OO CCC COO
GGYGYUGYGYUYGYYYLEYYYYYLDDEYYYGYDDEEYEGDYEGGDDEYGGEEYEEGS
YUM MWY UUUUULUUUULUOCULUULUULLUDUOUULULUUULDULULYLUUUOLULUUULLULLYULULLUYYLULY
CYCCYUUUYYYYOLYY GOYGEYGDGYYODZDEGLEDELELYEDLYYYGYYYYYYGYY YEG
EGLILYYLEYLY
UYUGYYYYGOUYUUYUGLYDODLYGUDDGEDGLEGYYLOGLEELELLLELLEYY

SS
YULLLCLLYLDYUULIYLULYEYYYUYEDYEGEOUUYEGLYDGE

SS RSSASS
UUUUUUDULUULUUDUULULUULLULULULULUULEULUUYLUDLUELULUUULOULULULULEL %

80
YUUCEYUUGULGYYYYOYYYYYGOGDYLYLIDGGELEGODDEGEY:
UUUGOGYYOYDYYUODYYEYGGGUIEYEGGGYGEGGGLYGLGGLA
GULGLLULULYLYYGUYYYYYODYYELEDDDEYELELYYYYEDGGYYS
GGG CUDCCULLGULYLUGLUYUULOGUUUULDLUUULDGOUULE

YG,
Un. Ua a t.ta U

S
AV. O'l
YUUUYGUGDLL
YMMV
01

Y 2
Ov O€ Oo?
2 SN NY ev 2
mz
H E i th
YY YY
Gn ee
oma as
ares
led
Tl AOGOODOL LOLOL LL
tl HAGOLOOO LAOOL
OGG,
ee PRY OO UUULUAUMY | A) f WUULULULUYDE VLULULUYLULULYUYUUS
CU UUM MMMUMA EEE OO OB TAA |_| AAA AAAS
AAA GLEEE
YY YOUU CLIO, L AAO A OO YA Oh |
YGUUYEYUYEYDLL YLILEEEYEEYE YYYLGGDYGOOD BCLYGYGYLYIO EGE
CMM UUW UWMUUUOUUU
WCW UUM UM VM YYUYUYUYYY BYBIDGGEOS YGULYEVEGOYLYEYYEYLYEYYYILYLYE GYYYYYYEGEGGEGYYOGLEYYYEEGLGESS
WU WUC MOWUUOUUVUYOUUYU YUYYYYYOGGGLS: UUDLYLULUOCLLYYUU LLUUULULCDULLUDUCGYYLUYULYULOLYUULUUUCLYUDYY
VU VOW WOOO WMUM MMMM UMMM OU EUUM MG YYYOYGYYEY COLYLVEGY YGLYLGLGLYLYLEG
YEYYGY LYYLYLYGGEGA:
LYGL
Ui Wie WOW UVMUUOUD UUYYYYGYOUYUGLGY : GUDYYYGYYGYY
YS. EGGLEY
UYTY)?> ”YY
LEGGY GGGG
OM WOM MOU WUMUWMUVUU MMUUMUBDUYAYUYYYYOQD QGEDZEGCESGS: YUUGLUELECYYYGEDYYAYY OGDEYYLOCGLYGYYEDELGYOGGLE LE
UGLY?
MO UVUUVOUYUU YYYUYUYYYGILYZJg2~GLCEGY GUUEY LUGGAGE YGYYDUUEEYYYE
WU UUM OM CMM
COMMU MOMUMUUOUUYUUUYUUYUUDUMY GOWUUACLYOUEGUUYEVUCLYOUYDUYUYUYYBLUYOUDULLUUUDULULUUUULLULULYDES
UW UM
OOM HOUPUMMYUUMYMYBUDYYDBDYYDDB?—I‘RD’GDgDDGEE: OYYYYLY LYLOELO LLL GGG
YLLLLYY
LLL LLL YGL EGGS
Ve WM UM
GUY VMAUWMYPYUEMYYPMAYYYYEGDLGDZDUIDDIDDGEESG UUGUDELYYEDOOLEGLELEYEDYYYYDGQYYGLEDEEYEYYYYLELOGOEL
YGGELALGGGLGYY
YUGUYYODDOYYYYIYGELYYEGYEGYDODDLYYDGLYYGYELYLOY %
CUWOUCLBYEGUYYLOUVULULYUYUULUOUYUYUOULLUYUULULLY (rsnan:—s €6&F/
WOOO OO COO COCO OOOO OOOO Y
GUYYDGYYOBYEGYYUBYOGGGOEGUYG GEEYYYGYYEEGL GE?
GYUGLAYVUEVYBYYEYYYYYVEUYY YYYYYYYYLYYLOLYYYGYIL SS,
CUUGULEDOUOULULLOCLOGLYYGUYDLYYYUUCUUGUCLULLULCLOUUULULOCULYG

708
 709
(wi!) y yybuajannM
Seaeeee

ceGeas

[asymp

CLCULLE

WI

UUUGY

o¢e
Ov
aOz
Ries
es

lat
ES
ea

es

ever
bo

acct
vaas

YU
|

9 iesmel iS Cae a GUULLGGYL4LLP AG GODEEOLODS

iain | |
MTR | | | | | minh ’CLUUULUUALUUUAU? BRSOCOOCOOO OOOO OOS
+
WWUUUWUUde Hee BS Bre YUGUUYUGYGGGYGGIN |
YOGGOOGLPOOLCLLOGLLE
BPUWWUUUU Pee ee ae GUGULGYLYUYLGLGS YUOGULUGYYUGYUYGIYGUG

UCUUUULULULLULULUUG
WWW; WOU Ws Wit [age YUYULUGGGLGYGIYGA UG LUG
UWUUGYYYGYIYVYYG

SNANASN
UUUUWUU Uh UW
WULWOU{[“CDNI mee COM nes
UMAGA (UGG
ODUUUUGULDLULUUL
eee YYYUYUYYYYUGYYYYES WY
YYYYLUYYY (UU, 4

NSS
YULDULULULULULS

NN

COOGEE
WUWVUUUUWUM Zp Y, cates De
GUGi GUA VAL YY, ALAAC YO;
LASSIE ies
WUUUUAGULULULUULY
CAAA Gy
BAA
“YDULLBULULUYUYGULLLY YY GU
LUEULODLODID IOS

Ni

‘la
YULULLULYUYYGULGULGYSY CO

U1G54%,UU,

Nae
UUUULLULUGUDUULULULULGLY YUUULUULLGDGDA

ALTA
“LU,4 Y
YO ULUUUUULUUDUUEYDUAG OLE” a
WUUWUWUYYY YUUULUULULULUUGULG

N
WU
UUUUMMM@OU“L CYL Veanes meee 04 UL. WM VM@ #0 WU Wt Vase AOU AAG GLA
UULULULYYGT pooig SIA
YUULUUAUU 6
W201 UUW Us
U#@OJ0HY00U UW
M@WaWhW UNMM UMM OUMA MUM ULMUMMA LYLYOGYOG UUYG GU4s GYG4G GLY
UVa Ua Us WsUh Ut. Uh Ua Wa Ua Wa Uh Ua Ua Wa ha Wd
Usa Wad Ue Uae UU ZZ
G~UUUGUGUUGG Gi ULUUULUULLELY

(X)2 aduD}{IWSUDI|
eee
G&UUYYLULEULYG:
Ov o€ Oo? O'l

44%G
A4AA77 7 ASSESS Seesaw OY A?

414%GY,%

COU
UULGLGYLYEGES
UMW, Oo GUapse=@—” JPPOQP0- 7-0 AD HH

\\
AMMA | | te 1 Lea AC AOR AIO AO ROBE

CULLGUULLULLUY
WWUUUMUYUAUM U@ODT OL TULUM Wt VT OU WU“_W[M UL Wa Ua Ub WU Us Wt Ut. ht th. G4
BELLLOL

4G,
wae
WU UMW UM UMMM W UUM MMEMM MUM UMU“=UWUWU_—IMA“L<MU€CM“«™P{[UW= UAL UM MUMM[M W& 0&0

pipet
S S S NSS
WUUWUUUTAMVMMA MEH I@M@@VH#o@UM#@ WW WWM UMA UWL UUMLU MU@UtL&00@ 02” 04,

NY

sks
SNS
YU WAWUUM“ UM Wea LUM UMUM0EMXWUUAUUMUUUMUUL UUM MMUME|[M™ UqW_0]“U: NL Wa UZ

bea
hh

CGGGUUUUCULLUGUY
UUM UMMM MMMM MM UU UM@OM AMMO M@UMA6 UM Uh WWW Wb UW Uk W@W Uk Wt. W@W te

SSuns

Aw
Va Oey
YUWUUUYUYUYUYUYYBYEYEYBYYLEYGGEOLEGYEGEEGGEEEEY:

UU
WYUULUYLUEY
SSS
=

SSSI
OO LE
WMUYYYUYDUYYYYYYLLIGQO9§ODO$PDE EYL YL GL GLGSG:
Gy;
WUDGLGUUYY
UUM UM LM PEMBURY
GLELODLIOOLOLOD

UUGU
YUUUUUMUMUUMUUWUUWULUMOUWMWMOUWUUUUUUMUWUUMMLHMUMLNZ
ULULUULULADUULULULLUG&LD

SSS
SSS
ASSN
SSNSSSSSN
YUYVUYUYUUYB2YUYBEYEYLYYYEEGLEYGEGYEOSGGEGGSS:
LYYYY
YYCYYYYGULY

AWS y
WWMM W@W CLMC@ WWW WW WWW WMUWUA WH
UYGULUYGGGYYGLGES
Wi MMM UMW
YUUGLYGYYYLOLYLLY
WWW MMMM UWUUW“OM@UMMMUMOUWMUW UMMA ULM UMA ULMAUWUUVDAL4,
GUYUU
YELEY YYYLUY

SS SS
UCWUUUMMUUMMMMULTU MUM M@UMMA
UMUMUUMUALUWUWOAUWUUMLOUMEUM|WM_MEEA“0UL04,
UUUUYLYYGGYGYYYG
UUUMMULUMW UMMM OUMMMMM@@/WIWUMMMAWUW“_LWMAUWUMWaM@1«U{M“£ U DM Wha Us Ut GG
UYGULYLGGYGGLGGS

WULUULYLULUULULLUL
YWUMUMWUUUAVU Maa UUdUh Ua Wb aU Wa Ub Wah Ue Ua Wb Uh Uh Wh Ua Wa Waa
WGULYYLLYLYYYOLE

VOOCOLOOCCOLLLEE
CUUUDULUUUULUUG

UUGYYLOUYLLYLUGL
GUGY LYYYLLYITE
GYYYYYYULUYLYYY

CUULUULULULUEL
o
SEEEEESEEES =EaS”

YYGGYU4GGG
CLIT
ANSNI

YYYYYGYYYLY,
Hira pede
Wide

YWYYYGYYGY
UUULLGB

UCGUUULULULUEGA

LGYYGOGGGLY
UUUULE
UU
Uh soba

N
GLEE
GGGHS
WO ke

Lge
4U4

OUGYG!
No
E

YY
WYLLIE LLL LUCY OGULYUYLYEG LY LY YGOGY (WUGUGUU
WW We WW Weta WW WM Ue.Ws Wi We Wt Wath te a Wh Ute ete ttt. a Wt. Utets LU UG GOUELELLOD ,
710
UUUDULUUUUDUULULUOYGUEULYUULGUAYGU Ut Wt.
4 Wa Wt. ts.Wt.Ut Wt Wt Wt WtUUULUULUUGUGUUGUULLULULULLGULULLLI
WWWa Wh We Wt We WWWt,Wt. te Ut.Wt. te Ue.
LGELYL te.Ue We
Ol
ULUUUGUGG4
GE GLEE LLGL LLL COLELLO
GUULE
YGGUYGGYGY UMUU WA Wi
AN pooig AAAMAA AB ULULULULULGULLULULUGLLGULLULULUGUGULULUGUUGGUILLGZLGZ UMUWMUUWUOUMUUUEUUU
WOO UUM WU
AANAABA HAA AN SAAN ASAI
ANNANAA WG WW
ULUULULULUGULULLS WWWWW CW Ws Wt WoW WUMU“_UJWHM
WWI“AMMM WAM UU UW UU
80 GUCUGCLUUUGLLOYLBUDLUUGUBL
UUUUAUUUGuaGs
U2LG 0 UMMA
YLLULDLUI ULM UMMUM“_M“[_wtMHMUIULOJ0U0{“UWU
WM Uh W@W.
Ut Wt Wh. Uh UA. VU
YOGUYGDGOEUOLLULLLLLLLE OOO MO CMM WWW UMMA WU UUW WUWW WHOM MUU MUWMOW
WU
UUDUUULOULULULGULUULULGUULULUUGEA
UCUGLLEYYUYDGUYEDGELUULLLGLELELGYEEGS WUUWUUUUWU
WY’Wa
UGLULUULUUULUUULUGLGULUULULUULUGULL
WaWa W@W WWWW@ WW W@W
UULUUGULUULUGULULLGULLOREZLZZZZGLZ4 MUM
HMMUU@C GLLGUUGLS WMAUUMUYUUMVWU“LM
90 CL
GOLLELEOEES
Came
OL GUGLLELLELLOL ULUULUUULULULULULULULULULLZLLULGLS
LWUUULLULGEE
14+
| debts YLULLELLLE
tbh
h db babshibld Mpbobdddbh
YYUUUGLOUUG YLLLLDLELEEL
OOOO POLO
RO CC UULbCLLDL
2ZULUUAULULLUGLS
“UUDLELLYEYS OOO. OE A MM
GLOGS
LULU UULLUULULULGULUUUUGLULUGLUUGLGLULULGLLULUULULUGLLLSL
Ae) NAHAS
A HABA
AA AA ARAMAAA AAAAA AA AAA
Ht ER UUMLAULMA
UUUUUDUUDA
“LUUUUUG BBYZAZZAGGZGZZB: UULUWUWUUU
| AAAI YULUULUUGLELG ARIA
| VUUWUUUUUL
PY’ UUEE YYYYYYY
YY YY SERBZZZZZzZWs Ute Ut
Wt Yt. WAU“.
|W WU: OOOO UAV AAAAASPY
20 v0 s'0 L0 Ol or o€e Ov

Ol OOP.
OOD OOOO
OL OPEL OOCLE OLE COOL OOOO
% UUYGEYULUGYCUUUUYYEYEYBLEYGLYLYUEYLYEEBELEDBLELLED
ee WUGUGULYYGGGYYYGYODLYLULYYLDGGLYLYGYGGYYGLYLDYLGES
% pudg ALAA
VU YUUULDGUYUYUYYOULLUYUULLUGUULULUCLUULUUUYUULLLY
8'04 D OLLLEOLLLYLLODLDDGE
LOLOL DULYY
YD YUUULUYL
BUAZZZZ4ZZLZLZ4Z OOOO UEULUUUULUGULUUS
WUUGUUUUG POPOL
OCC@ WAAL
AAALAC
| YUYYUYUYUUYUYYLUUYULULUYUULUULGLYYLECLYLCGCABE
90 AAALAC
| UUCUYLUULYLULYCULYYYLUUOUUYLLLLCLUULLOLOLLAL
LAA
| aun GODDGLEIGEG¢-
UYLLOGDGYLY
CUDDLY YYYOUGYDLY
VULZUUU4UaLs | Aa AAA OOOO
ULEYZLL4
Hh YUCYUYGULGGUYYGOGULYYYODLYYLYGEOGEGLGGAME
CLULULLLLS >0>0,>->0075-4
> PO OOO AAP
A OAL
YULLULY 4UUGUCGUYLUYCDLGYYYGUYLLYYLEELGYGGG>CG?7E LLULLLUEE
“YUULLLLL OGG LOLLEYEOUALG
ELEY UMULULUBUUUMUEU “
YY4G4GLGDYYOOGY
ae
AS L4G YYUYLUYUYEUCLYLUYLUULLUYLYUYGUGYLUYYUEYGLUY
UUCCLUUGULULCUYLUCCLUUYULUULLLYOYUGUULGLOR
LOCUGDYYLGSOLYYYYYGLEU
GULLY YYUGYYYLYYG
E bo beokeddecdecherfodedonbeokod
| VATAZA GLYLGUGLOLE:
YYLYG GOLDY YYPOULLYLY poh
PMBEBPYZZZZUYZYZZZUYGGZLGL WUWUUUUUUUUUUMUWUAUYA
W@W
Ws WaWaaWs
UW A UYAWU
BPA P72 YYGZUGLGLLYG4GGGGLCE
PPO PCOORP
PB LL Ud OW JUIUWU
MUU YWUYUWUUUWOU
ae) s'0 9°0 20

(X¥)2 BOUD}JIWUSUDSI|
W6d
ts U6 Wb Wa UaWi WW Wh ss ts. WW Wht Wea WW
W,
Ws

Wa
WAC

ACACALALALA,
AAC OOOO OOOO
OO OO MMM
TS UYUYUUYUYU
A MATA UUYUUUUUOU
VOLES “YYMYYLYY
OCG OOOO OOOO AO OOOO” OLLI OO
MMMM HOO LUM
YYOLYYYYTEG
LOLOL OUOAG YUYYYYYYYYGOUYY

\\SSa
NI
LYAUAYOOAALALOOLOLOLOO
YYOOLLLYLG LOOE
WA | a) LA
OOOO. AAOAAA LMM |
OEOOOLOATOA
DOLL YLYYYYGUGUG
BU AYD
SO ia OO. A: AAA AAAACA HAHAH
A MAMAN
| TATATAT
UM AAO AMAA | UYUYYYMYUYU
OLLGD YOUGCGYYYGL O79").
OO PO OO OPO OO OP OO DP OOOL MIOMM BWM
teat YDGYYDLGLYL
ae Y
Y4YY4CGGLYY
| LDGELAALOLOLOLEE: UOUUUULYUYO
GAGE YYAEY YOULL YLYYYDCLUYYLY OOOO ca OOO OO
“EEL WUWWUYbOVae Wt, 0000 ™4100( X#0I_@{™0 WI WY
iammmiitn_™—
ot |__| aa0
Ett WYUt—WaweaWwewwwwue
YIWIWa
WW
s'0 9'0 oe 0'% O'€ O'v
0'|

XY
JHusjavADM
(
)wu71
 4
Ya

Vn
Ua

th
Ut
th
Ua
UA
AU5

Wn.
WnGo
Wo
GUGLECUDCGDYCUDADOLULDCGAUOUOGUYYYYDCCDUYUUCDDUUULUUYUUULUD

Ws

Waa

(0
UaWt

Wn
UU

hnaa
WOOO

UnWs
Waa
OE.

W@#
.

UaPawn
Wats
YOO

€eeu@@
0
OAC COCO COCO COCO OOO OAH pom ania
UUW

WAdWédMd

(a
7Ua#
WW

UdWMaW

WLUW
Wt 0s
VLYYYYYYUYYGLOYYYUYY GLEE LLLELLLLLES

Us
WWI
on

UA
a0,
Uta

80
YW
(#0UU
(@

UWA
Was
Wasa
BYYYYYUUUYLVUYYUYOLYYUYYYYLUDLYOLDCO—S

U#€U0U
@YsYew
ts#000004
BOYYLYYYYYYLYOLYGYYYOLELYLYYEYYLLLYIE UYUYYYYYY 4

(Jet
Jb
aswep@MJtwW
WULLYUYYYS

WVU
TQM.
9'0

#400
Wats
UYYYUZLLYL4

PW
WU

Ut, Pa ||
WIC
BYYYYYYYYCGYLEULYYGOLOLYGLYLLYEAS

20 tp
AAA
UMMM

GUGUUALOCLCLYLULOLDDUBODLULUUIUVE AD | AMAA

}(J WU

IJ 4.
WU

VA
PMX: YW
UW
| LAAAAAAGG

W@WA

—
WU

Ye
U4Y4YGY¢YGYY4YGYYGYYGYYYYGYYYGYYYGGGGS

UEY
Yea

YUUUYYUGY

J]|@Z@@J#¥@?”#!
“_ttt:
YAU Wasa O0U“M@/@U@T

BYYYYLYOUULYYUOLGLYLEGLYLLAGE
UM

OJJYJ€?#
V0

UW),
WW

UUUIMIWA

GUUYGYYYYGGYLYYYGGGGLGGGGGGGL GS

WA

YAY
V/A

GE VYYGYYYS
YLUYLUOECAYLE DOLLY LEEEL E77 oe AAA A AAG

Ae)
U#UUM_OUWU“$™U’_Vz"

@ePCC—@JV—

€éxv~Y@q]@]PDW0?“PWW@w
W040

“|
AA»
VY UGUYYYYLYYLOYOUUYLOLOL GGG

UMMJMOMMUW“MM«M“
BYLCG4U4
ALLA

YYYUYYUCLUYUYYYLYEOGLLELL Tee |
QO
| | 1
beeenide bdisebdeeteriecd
PYU@WHM@CWJH
BYZYYYYYGYYYYYOYGYYYOGOGOGG2eeee —YYYUGYY
ad see

20
| WHO

—
V4YUYU4

|
PeC@C@@@P—@@P@|
—YYGUYGYYYYYYYYYYYYUYGLYAGYGA WG

SS
BLYELLYLYOY YY Y LLY

oe
UYGYY

ES
AUUCG0C0OCCU
BBYZZIZYZZZZ YYYUUY GALLI TTTTTTLLTH gape

FSS
0'0 BYY!

(0)
re) £0 ae) sO 9'°0 hee

2~W™0Y-/JIIIPJJJY¥0U¥Y™Q_|_-
WaMt YUIMMaMU_MJ0LWAWAaY

ee
bihasan

ES SOY)
GU
WYUWUWUUYUUUYYYVY
UU

Ue

Wa
Ua

th
Ue
Uh
Wa

Ute
Ua
Ua

Waa

Ut.
Ula
CIWIPIYPI-WIMW@XMPysqwtW
We

t,t
Wt@a

Wao

lMWéwsYrjGo—|\—
UA
@e@W@@PCWW@CP/WlUWOWU[U>"01
Wa

WWI

ts WU
Me MMO OI MOO OOO LOO,
“UWA

WH

LYLE
WM

LEGAL
WHY
Ut
Wn

Wt
WoW

Ws
Ws
Ca Wa

Us

Uh
ae,
Wis
UUULULULUUDULUUUYLULUULULULUUUDUDULUULUYLUYUUS
a4

Ut

ode
bob
deb
ede
aed
dicde
dp
bo

sa
ob
Wawa
WV

awa

Wb
0a

Wt WtYa4
Ub
UUW

WWF
Wan
UYUUUDULUULUYLUGUDLGYLUUYUULULYLUYLYYYGUYLLUYLIGGG
UWAWUIWAa

Ua WW Waa

C@
MA WaWasa
WU

eas

Us WM
YUU

UUULULUUULULGYUUYLULUULLUYLUYUGUUGLYGUYUGLY

MUI
Vn YWasWawqWa

Uh, @PMes
UUUULUUUDUUAYULGULULUUYDULUUGLULAYLAGULULS eves
WCW

WCW
UUUAUAUUUWA

UWAWUUWMItWAa

90
UYULUULUUYDUUDUGUULLYGUULGYLULYUUGLGUGYUUYYUG"
+H

UULYUULGUGULOULGULULULYOULUUYLUYLULYLYYGYLLLY:

Y44GGYYYEY
H/F

UMMAVWI=WU
vO GYUYYYGLGYUYGYUGGYUGUYYGUGYYGYYGGYGA
|

il
MA

UUUULULUEDUULLUDUUOUUULUOUUUUAYDUUUY

MMMM
YUDLYDUYUUYLUYYDYUULULYYGULLYLYLYYULLGILY4GA
Ma
UL

Ua#
AAA

UUYULDGUUYLYYGULGYLYUULULGLUYLYGIYGU4LLA
YUVUUYUYYYYULLY
Uda

Wana

LOO

H|
YUUGUGGYGIYYOYLYYYOYYLYYYGYG4 salsa
i TE UMUUOC DOO
MMM

ULYUYGULYLUIGYY
AO

OEE
BAWUGUUMbAAeew?
LOE

OOOO
UGYLYYUYLYDYLYYYGLYALLDA
44 See TP F(A
7551 Rr)(AA FT
LOCAL A.
| a ‘WO
O11
AL,

20
OOOO nn
U#
YUU
BAUZU4444UU4u44y4s

HALLO
=

oy UUYUULGUYLLYGULGLYL
BULLOCLLOOLES
ae
=

NN
AC P+wovp@PW@H@<P/a

OGG AOA MAA

0° zy ED ae) @) Lom 9'0 210 0?
Wh OUULAR

AOU

OLYYAYOLOLYAAEEYIIALIOY

(X)2 9OUDJ}IWUSUDI|
A UUMUUM

oi
OO
OOOO
WOOO
OOOO
OOOO
OCLC eee
OMMIWMUWWWWUCUMCUUUT

Y) GUUUUYYOGYYGYYYYYEYULYYYYGYGGYYGYIYYYYYLEGGELYGE: HAE
OOOO MEMO
eZ A WGUUUGGGGYYYGGYYGYYEYYGGYYYGYGYGIYGGGYGIYOGG OO
80 yy OMAP
YA. UUYGUYYGGGYYYYOYYLDGYYYLGGYYUYYUYULYDLUOUULYYDDGOUA
YUUMUWMWUUUUMAaaSE

OYLLLLULLLLLY
YY YOY LY
YWUUUUWUNGAEE

WUGGY4YYUYYDYYLUYIDGYYYIIYYEYGYGAIEBRY
YGUYGGGYYYGGGGGYGGGYYGOGGGGGGGGRBEEEBYZZZZYZYZZGGZG44LGG4Z
YOM UYYUYUYYUY
YOYGLYGYLEYLLGGA:
LLY Yee LS YOUYLYLYYGLOLYL
GYYYYUYOLY UUWUUUWUOUYUU
COOLOOOO
OU CLOUTLAYBY RRR7G UYUULUYULUDLEYS
YLYOCYYYUY MMOMMAWOWBUOM
AOA PE UUUM
BUYYZYYYZYYGYGGGGGG4IF
age YYYYYYYGULUYYYLOGGYYYYOGEEY
U4 UWUUUUUWUUOU:
WW
BYYYYYYYLUYYYYYEGGS
. ra YOULL YLYYYYLYLYYLALYOLYEY
; WWW:WAU UWAWU
© WOOOQOGOOGOOOULLLA,
PEA WYYYYGGYUGYYYYGYGGLYUIGYGYGZGGGGGGA
Uy ts Y Ee Fa UUUUUU
YUUULUUULULUUDUUUYLUYYUYUY UYLCOUUUUYULUUUUYLULDUUDUUAUUYYDLAaan
| OOO
YYYUYLYYYGYYGGYLLIS UUULUYUYLUYGUULUGYLYOLULLIYYLYGGGLG
Wh YLT WUUUWUUUUUU
WUYUULYOYUYGYGYYYGYGGS YYUYGUYGYYUDGLGYYLLYGGLYGGGLGG:
yyy”. | UMYUYYUUYUOUY
|e YYYLGULUGGYYYGGYUYGYIGGGGGGIG
UOC YOO| UWUUUUUMNUOYee
UW
2'0

GUUYGGYGYYYODGGYYYGGYYGYGDYYGYGYYYYGGYLGYYGGGYY
BBYZZYZZYYYLYZYZA“YG4 YLUULULUUAAUAGULAUUGYYGUUGUA4444GA
A MAA Mam WeW000WM WU
ZZZZYGYAG4GIGI4
BOYi Mn. BEIYAAAZZAGH4 a FY 0 UMUUM
UUUGYYOYLYLYGYYIE by, G4YLIYOYYAYYOEILYAYAAS
ep | fc, (a Pane UUUWUWUUO
VWACEE§i«OO
| .: YUGUGYGUUUYGGGYYGYYUUGELT.
Aton
| | ee ee | TY TTT At prrrbprpre
2'0 ¢'0 9'0 LO 0"! 0% o'e O'v

CUCGLUDLDUUUUOLDDULYLULUDDDGUULGGDYLUGULUUGGUUULULDULULUULLUUULUULUG
UUUUUUWUUWUWUUWUUUWUUUUYUUYYUYUYYGGYYGAUGEGGG

J}OuajasADM

711
(w7)X
 (w7/)y yybuajannm

VULULLULUULULLUULULUOD
YULULULLLULULIULULLLLY
EMO
BYYULLUYGYGY

YAYLLYLYBYGY
<0
(e)
(o)

'9)

vt
o

Y YUG
i

44%
Y
yyTT TT

Ee
Hh

MMMYAZz
CLLLLLELOLE ye TH Y
ee BEBZZZAZAAZU2UCAZLALLLGLSs
|
“LUGDCDLOUYLGYDGAIILA G7
LOOT
|_|
UUUDULYULULULUDUDLEGLGACYCGEGAZAAPAGEEEE Ee BEYZZAAZAZZELLL4LLLLY4A
oh

ULUUULUOULUULUUUULUEULUUULYOADGCoS PEE BRE RBBEBYZAAZAAAZALALALLELLS

wi MooooooooO OOOO ree A WU WUVULUUUULGLVOLE
WUUUUDDULUGYLUYDGUULGLULGYGUGLULGG UG CLLLYULULLLYLYLEY
WUGULELLYGEEGLELLGIS
YY
GUUUUUYLIYLYYGLGLYGYGLGELELGELGSS:
OT| Tt | | Le YUU
MAO OOOODOOODOICELO Orr UULLUULULOLLUYULYCCLY Ee
| |
Gof _l_ 0H]
WJH(WEE UUMMUMMML#@
LEY YYYELYUUYYGLYYY
ULUUEULUMU
YYYYY OYUYYLOLYYGDGDOLLYLYLELLLGLGDUUEEY
PLY ee
UYYYY GUULYLYLLYYYLUYGLYY GODLY YYOOLGGPEECLLYYGULLELYOYYGLLLGSS
OOOO OYE | — |_| LLL
LLL) LLL
UUUGLGLYOUGLDLYLLLY OY LLLYYYYYEYYLYGORLYYYYYYGEYELYYLLLYLLLY
A, IAI EIGEN
CLGDLLLULELUGLUGLEDLALLUUDYLYEUEGUDULUULODDLEE usa'pooia AO
WUMUYLUUDUYYGYDUUDUYYGYYYDYGGYGYUGGUYEYGLECL ELS Fa hee?
WUULUGYUUYVGEGYLGIUGGUYGUYUGEGYLUGGUYELEGELYGGEGL&GLGA ysinig Y
WOU OLGOL YOO LE LEE ELLE LLL. GY
UU 0s Wi Wt Ve
OOO OOO CUCLOCOGUYOCLLLULLLDLLLLGCCOCLLGLLOLOLLDOLLLGLDLLLLLD Cm

(¥)2 a0uD4} IWSUDI|

ES 9°0 ome) ae)
(ote)
YOY
VOM MOO OOOOIAL
UULUULLULELLLLGG Ae GUOUEDEL OCDE 20
BAR AAM AMAA AAA, GULULUGUYULGULLY
YUULULLYGGGUGLLLGL&L“AS
UOTE
VAAL ELL
SSaeasae
YUUUGUGLLYGLLLYLGLZOS SHERRSEEEBYZZZZZZZZGZLZZZZZ
MMA WU
WU
LUULLLLEULLLUULLLLLDES CECCCCCCCCCC AAP v0
LLELL Lae WM LLL
“Le OO WI YULGYYGUGULGUGLLLLGLELLGA EYZZLZYZYGYZGLLLGGLS
GUUYYUGGUOYYGELGGYLGGUGG VUYULGLLYULUULGLULGL
YUULGULYLYUEYGUGLLLGGS YYYYLUGLYYYGYYYYLOEE
WWE Yo UW
UCULUULULUULDLUGULLUEUGULULUG CLLUULLLULUUULCDCLOS Ee
Qo LL :[$
WWWH™wwH7nh
MMA AAA AAA GUGUUULODULYUEULLYLS
YUUYYGYDUGLGELLYLGGGY GUGOLOL LUU4.4 GLGLLYGLDGLOEA
Wt Uh GUAT VU V4 V7 Li MOKED
WY
yp
OO
OQ UM Vn 4
eLiL UCUUUYLLYELEY POM UG,44% 444%, YUU
LLYLLULELDLGLLDLGLAY CUUCOULLA Baer eee
UU OE OE WI OO YLOU4
V LE PHOT
OPPEP@E Oe UGG
LU
ee eeeeee eee ee
Up,
GG
UU
WUMYWUUWA OWA
WMOWHUEA
WU
UUOUMMYE MMB YMA YYLUYULULUDULUULYLUULLLLLELYELLELLUUGUGLLUYLELLLLS ATL
GG
UULUULLUDLLLLSD
ee
UD VMMM UUM
AMAYAlVMYEUMAUUU AMUUDAUIYW_AHI“a YAMADA L044. LUYYUYYLLY
pe
CHUUGUUUUGCLUULGCDLULUGUGULLUGDGLLULDULLE:

(e)
00
OOOO
WF
UYU UWUM“U“MMA0002
Gpe
a
Hoo
CCC
AA
AA
WG
YUL
ee “UGG
q
rue)

OLLIE.
DEMEMEECCOCCOCOCCCLOCOOLD WEEE VO rave)
SEBEBEBBLZZZZZZZZ 21999747 AUGRG RR ERRESROEEREED! 4177777
BY LLAALLLLLLLLLLLLLGTS BYUUYLUULLS
DYU“LUGYLLYUGULLULGYYBGA BYYYLLGLL4
Ge
BEYYYULLLLLUUULULYLUUULLELULE @YUGL407707
BYUYZYUZLYGYULGGGYLYUGGLLGY: VUUYLLLULG
BYYZYLEYGYLULUGL“LEOGGO“COGGAREEEE BYYZLLLGYLLL
OLYLYYYLYELLLLLYYLYLDLAUOAA VLZULULLULLGL
BY UYYLUUCLLUYUULLULLLLGDEGO“OOCOREE Aso
YY UYYOY GUYYYULYYYYOYLLOLYOZCReee Tm =puog pooig 2
|_|VACA ACO OG LLL OG, |) BZ
WHO oaee_lt_ OO aaa OU
OOOO ‘“CULDDGUULULY =
MAW
MM#
WJM
MA
YUUUMMMA
ae
UMM
OPYMMMM@/YYHWDM@
MAYO
WGUCUGGLULULULLLUULUULDULULULULGCLUULLLODUUUUGLDLUUUGY GULL “GU UMAU
UAUMM
UV UMMA UA 27 a 0402 Ua Wt Ut Ua UUs UU MAU UUM UMA
Ol
AMAU
UA
Ws.
UUM
Us
Vo
Ws
UM
VAM
AMMA
Uta
4.
YU
Ut,

712
 GUULYLLLYYGULGYYGDYLGGYYUGEGLYLULGELEGGGBGEG
ULLULYUULGULLULUGLGLLZULGLGAEPLLEULLLLLEEUL
YLYULGGYYGULULGGLLLYGGUYLYYUGGUGLYYGELUEGGLLGLYY

\DuajaAD
UYYUULUYUYLLELEULLEYYLELYELELYLULEELLYYY?
UULCLLYLYDULLULLEDUDLYLLYULLLDDLES

YULYDYLYUULELELYELYYYLEYYYYLY
GUODLLOYLOLOOLELLOLLYLELIL

75
GYYYGYEGLLGYOLLYYLYEYLYLLLL
(w7t)X y

GLLOLLDLELOLLLYLOLLD
COOLED)
GCGLGOLYYGOGYGYGOGEY

YLLOLLLOOM
YLOULYLLLYYOOLL

Ov Sete rain Sad aeO

bbb!
GULGGI
VOLE
OO
COMO
eee

OD
LOO
LODE
PLUG
OLOOOOS
OEE
OOOO
LLL
ae

eee
a

LLLLLOOLOL
YUYYYYUYYYEYEYYYYELYYYEEYEYYES
BOCCOCO COE C COCO COO CE CC CCE COOLS
GLLOLLL

L AMAA MAA Rete doh hp AAA pa A

MMOLE
GOGO GGODOCIOLE: Pf betatbed dant berlenbthar bodesbecpenpetererd ets ben

LOLA
“UUULLCUULLADT
LWYEYYYGGYLLYYELLUGLAGLLEEGLLLSY

Hh
GYYULUYYGLYYYLYYGLGS PEL GG0494
OOOO

AZ
LLL

LEC
COO Son = Eas

LLL:
L
YUUMUUUUUWUAU = a — | Z
WUUUUMUUAU Wt 4 Ut U4 Uhh (al si LUA
ULLAL,
oo EEEeeeeeee

YY

NNNSNAN
WUUUUWUUU UY UWL |
MAALL@@0@00,

pooig
puodg
UUM 1 011 ULM“ULL Slee
io
i

uaa19
LUAU HO OA A
UOMO UU Uh UU WAU A
ECPOPLTC@@PO@@P@?@@?®@?

YUU LMAO AMMA W004MMM UMMM OO

OPPO
UUUMUUMUMUUUUUOUEA

SS
YUUUUU HU UMUDWWM
YYYGUOULYOU GLY LY OL OLL YELL LYLYLLLEELLLEZ GL UU2
UM@OUIUMP0U0WU“>
UMM VU COO
UMMAUUM UaeUa VA UMA UU
MAUMA WW Ut Us Wa Wa We Uae Wa Ula Ue UtaWs Wa Ut Wt.
ttt GUUULCULLULCDLUUDDLDULUULLLULDGULLGLLUUGULLUDULGELLUULUULLUULLUY

GLYG4C¢7C7ER
YY-@U@PCPEPOP@OPOPOPOPOPEPO@PPOPEOOOPEELL

(X)2 a0uD}4 IWSUDI|

LGLLLLODE:
YLYLGLGL2
Lys
£0 20

ome)
on ac 90
oy WLYLYYYYOCUOY YUL ee

EG
|AMZ

: |
HOM GMO GIOLLLEL.

|
|
|
|
|

|
mY,
AUGYYYYYYELYYLYLYLEYS

GUE
CA
4
GOGE: LLLLOLLOLOLLLE
CLYUYUUGCLYYYUCYYLLGDY LLL 20
Becta

Yep
“1
GUYYLLGYYYLEOLLGLA Ie

NSS
ee 77
ee MAME
GLYYLYYCLLYYOLEYLYEYELLLE vo

NNN
YLLYYYDLLULGLLYYGILIGE

pape
SSS
SSS
SSS
SS
GLOGLELLLO COYLE OLDIDY YUUUUUUUULUULOULDULULYDUDUELLULULLES
MM OOF FOFF | LL AAA ES
tt

OLLI PY LYY YY OY DOL YL LLL LLLLLCLELS

YLYYGYYYYUYLYLY

SN

XS

90
CUUGUUCLELOCDOUGLGLLYLYLLDUGLULYLUDLDUG
Ws

GUYLOLYLOY LOLLY LLL YY UY LOLLY OYLYY LOY CULL ECL YYY OU GUL LLL
GOLLLLEAUYUGGULULEGGGLLLAGLGGLLLIDGLY WAAR AA AN AAA AMAIA
GOLOLOLOLALOLLOL IDOE LLOLEYOELOYOELOL EL LLL EL LLL CUGGLGYY
UMMM

ANNNSS
ULULLCDLULLULULLYLYDODLLYDDDLLYLYOLLYLDDLDLDGO4 puog pooig
O0L01@0U“0Y71’iUMMMUM WUMUU

80
YGOLYOGLYOOGLYY ELL LY EGYOLYLLYY OLY OLY GUYL GEG
“Wu. Us 04 MAA UUW
YUYYGULOYYLELYOLDOLGLYGLYGYELYLYYYGLGE LYUYYLGGE: uvedd
SO OWMUWOUMW UUUMOMEMEMM@ML to OE OO
s
UUUWMUMM@OAMM@
OOO
UIP
POTVUMMAMMMMUWMMMEMMAWUWUUWUMUI

MOO:
UULGGGYGULULGULULUULGULULULULULDULULULULLDLGUULLLULLGLULUUGOULULUYLULLUULUDUUEBLULUL
WOOL

CPOE WML EEE LCL

O'l
teleWe UW Wile
WW.Watle Wt: th Wl: Wt We Wi te Ws Wt Wt Wt Wa We Wt: WW WW
UMW

GRO
OPO@PPP@OOW

UUWUMUMUMUO
LOL

PPO

sP@P@UPZ@#@#PG
UMMM
PPP@POP@OPO@POOH@PPOP@P

0?
sO vo <0 20

ee
Oe
st
PEP
UU
UMMWUMUMU

MO
WWW

UUMUUWUOOC
UMW

Aga titi ty ttt ttt tt Ae CUGGUUUbGOL

CCC OL

WU
PUPPPPOP@OPOP@OP

MM

YY
YY@-—#

YUYAO
MBESEBEEBYZZZZZZAZAZLLUCLZLL4
MUM

AY YY
WU WMUM

VOOM OAL Bae CUULULLULULYULULLLEG 2'0

UMMM
nae UUVGUYYUYUYUYLYLUULYYY

PP@P@PT@P@_™”?@#?
YULGLULULLGGLAZAZGEE

WWW NN
PPE@CPCH@@P@P_ WMWMO

WeEY
Seer LTT TT

PUPP
YULGbL YU ULGLLLELS 14:4, AS

OPP???

UM GY
NR
|_|
OMG ODO. CC WYGODLOGCCOCOLLLE

UUW
CUULYMYEYR WW

M0
PPO
WU

BPUZAZLZAULULULLULUUUUULLLYLGYS vo

OPCPPCP@?_
GULL EOCOOLCODLO OOM

UMMM MY
LEPC

GYUYGULULUYLULYGGLLLGGGA BYZYZZLYLUL&LULUULGLULULULYS

UU
U4OOO

MMU
OO

OUOU
WMO OOOOOOO OO LLL.
PTPOPO@

C_?P#U@I
WP
BPLZUZELLGLGLULYLGGLLULLIGLLS

WW
ULLULUULLULGULLALECGL“CACOCAS

WOW
OU
UMW
UW
UU
MUM
CH@Z™@P@UE@P@_
9:0

OI
BY ULLLLUULLLULOLULUDULLUDEG

Wi
VAN ACA ALAA ACL ALALAA ALAC AAA |

WMA
OW
Ws
P@P@@@#@P/L_

WM
PUPP
GUYUULYUUUUDELYYGCLOLAS WUYULGULUUUUEBLUYUYYLUYUYLUUYYLYY

POV
YMCEP

PY@_
E>@PP?_EPP

POH
UULLULUULULLLS WO yy ENEMAS AGAR OCOD

PCH@
PP
MU¢MUUAU GO OEY BOBBY A LLLLOLLLLLLLLLILLLLLILEL

MAUVE.
ULLUULLULULLUUEDLUULUULULUOULULUULELULUULULULDULUYLOULLUUUULLLULND :

AMEUULAoM

P@®@Pe

AULA
GOLEE
LIMA OOOO OOOO OOO FEC pudg pooig 80

GUL
4G
AWM
UULUULGULULGLUGLULULGLLULUGUULULULUYLYGLULLUGUYULLLULGLGS BESS uaal9

AVA

P@

% 4
UG
F
MV
CLOLLLLGDYDDLLIDELOYLYEYIGOELYIYE YULLULDOD
Ul

UU
Us
YM

Vt

Z
Z
Ol
 YRUUYUBYBEYLEYEYLSYYG

UUWUUWUUUUUMUOLU

)w7l

}HuajanoM
(

rar
Aad
AGB
ABABA
T_T
WW

L0 9:0 sO v'O <0 20.

POV

UUW.

Oe

WYUWUYY? LL
LL

Ov o¢€ 0?
BBBBRYLYYY YYYUU YYYYOUYYLY YOULL 147 OMY LOY Y LULU LLL GY ioe)
AE

LL AAA AAA AAA AAA AA | a LGULL ELLA YUGUYL:

WEE
WU

H EAPO OCP LM LUX<UJ_EI&EIBI#PHFF7™

WU

ZAUAZAAAULULLAALLAG4 20
Ws

B UUULLUGULLULULULULUULLALULUEGCLULULAUGULULAALAOOCAAABEREBY
Ws

UML
MEX POP
POOO
LOO POPOP CPOE OCC CCL PIP”
B YUGELLELOEEELYGELEELOGLELEGGLLLG | | LAA UL’UUUUZaGUGZZs
Wh, BRZZZLLZLLLUGGLLLUGUULLLGULLULLULLAGAZAAISGOCASEEYZ YYYYYYYYYY OYYLY
BY GOL YLOLLYL EGY LLL LL CLUE EOLLLY ULL BZ (‘UUULLDULUAULALG vo
OLLYYGLYL YL LLL LEGALS (YGUGUGGUCUGGUYGEA
CLYYLYYYLYYOYUG
MAUUMEWA

AAA ANA AA NAA AAI AO OOOO EE UCLUUULYELUDLEYUBUE

WOU

UEULLULULULLUULULULLLLUGULLLALULLLGULLLGLGLSZS WOO, OOOO YUYUGULLG

WM

OEE WMP EEE GUULLELUELLEDLEEDYDYGELELEE GLULEYEGULELLYELYYDBLULYLOULELULLYEA 90

ULUUUULULUUUGLULULULGULGLULUYLUGUGLULULUGUGULULULILYLUUGLULUGUUEULLYLULGLULALULUULLLLALLY
ULUGE lYY LLY YGYYLYL LOL LLLE LLOLLLLLELLOOLL LEELA
MUD TA
ALOELLMMEMM@@MA UUMUULU
MMMMEAMOMMMMMMAMUMHMAMMMMVECUV AU
UMUMLM1MLWHEM
COOL YULGLLLULULGULUGLULGLLLS
WUUUUMUUUUUU
COCO
GG
CULELLELLYLDLELIDDLELYLOBYYLLBUYULOLLULULDLUUE Z 80
UUWUUUOAAAULMLA TULL
WUUUUUUYA
ZA
WAU“ MLM UMMM UL@O
HAUT OMM# ZY UL@# WU» UAL“MWM 0X
UUUUYUUUUYUYYUYUUYYYUEUYYEYYLUYYYYLYYYYLYYYGY?
GOL OYE LLY GG
YLOYYYYLLOLY LOY YOY EY LOE OY OGLE Z
YWUYMUUUMMOMAMM 4 aan
WU

WU

ee
YUUUUMAUUWUA
UCLULEEEGEGELBELYCLEYALYYYLLYGIG U4, LY GY YLYYDY EYL ee
Ch!

UGUULGULULLEGGULEGGBGYULGLGULULULLLUELGOCUGGYY LU LUAULULULULUUGULUULD Ol
epee tires eA Nege MUI MAyn
WAG
eg ope NSS AIR iy eM A NtTay
aod
YUU UUM
UNMULUM W@W Us Uh WaaWa Wa Ua
UU Ut Uh. Uh.
1,

(X¥)2 BOUD}}IlUSUDI|
BYLLLLYLYLLIY

|Led.
“LLLP
He
Hin y

deste
Sep
0SSeeeeons
UMAUWUUULG
a YO) 20
aeWU ELEY ELE YODLII LID
WYAUULUULULUDULYULULEDULULUDLUYLYDLGLLEY

YY,
UUUULCLUYIGYLEDEYGYGYDGEYGIS
44

SSS
VA GLUGLUELOLDUDLLULOLLLYLDUDA
YELL LLL LLL
GYGLGYGL
seasscim cite
M1
YUUWUMMAMX0HHWPUUdACOe

GL
UWUAU th TL 0A UA
UUUMMEA
| AGEL ULULULLLUYLUGCULULLLD UYYUYYYYGY CMA
LY,
BS BZAZZ2ZEZLZL&LLLELLLLI WCUGUUUCLYGYLGIYGIGGYA LY GYYGOLY
| CLLLLLELLEADLYYYACIYYELLAALOBULLYLOULLDDLLDA 9:0
a ane BYZUZUZLLULULUUULULULUULUULLULUUUULULULULULULYULULULULULULULLULLUULDL
a | |OOOO ACCA YOUYYLUYYULYYUYYYYYLYYYYLYUYY Ys
| BBY YLLLLYYYLYYYYLLLYIDLEYYYYIAYYYBGYIGEA "UU UUU-ULUDLY
SEREEEBBYZZCAAAAUSLLLULUYLUYUULUULUYULOULULLAYULEULLALYDLL Y 80
ap AAA AA AANA AAA AAA AAA OOO Z
WO
Y OUY YYUYY YEYUUEYYYYEULEYUYELYUsy Z
Y 4 RAMA AMI OOOO? LU UUYUGDG Za
C
UW
WMI
U“M10L@X UUULUMMOLMMUMULU pe ea Toe)
UUs With GUGUUGCULYUULULDYU Z 4 Uy Ol
MUM W@MXM MVE UM@
OUWMOMUM@UHMHMUME DUM AMMA TATU LL UMW,
WV AAU UWA UUs Wa Ua UU Wh Ut ts Wt We Wt Us Wt, a

—YAUULUUUYYYUULILG
CUUGLUYYYLULGY:

YUUULLLUGUGG,
UULULUUGUDGIUGG
L0 90 rome) v0 £0 20.

UUULUUYLLDULUGD
OO a aaYa_a—>a_i3y$ 7 °°

GeGUUUGUGYGIGGGY
COO

BAO
BRSBBPZZZ22Z2ZZ4ZG78 GUO GUUGLUYGLEY

ts
WOOL.

HYE@#PHqHWMW@HWI@DO
ree AOL
Hae BEAUZUZUCLULLULAULALLLULAAGCoae LULEUULEEDULES re)

WGUUGUUUYLYG
SEREREBBZZZZZZZZZ2LLUZZLL4LZLU4GzAZ~"E

SIE

YU
~Aig SERS
“CCL AAP

OSG
BEBYZZUZZLZALLLEULLULULGUGLLYLYIULLLULYGESL&SE
MELLEL LAELIA OE LADO

<<,
Y4UUULYUGLULLULLGLYYGULGLYYUYGLLLLYYGGZGGAES

WUUIUMUWTMMV

XS AASV
Ee Scns MMM OOOO OODLE OLLIE.

ROR
ZUGG%

glad
Y
U0 YUUGDULUULUUGULUUULGULGULULULYUGLYYGYEDULUYLYAGYGAE

4,444
WUUWUUUMUU
Ut
UUs
UU
WU
noel ULUULGCDULLELLLLELLLLULDODLLLULLLLL AD

Pee
a Ub
th
0s aS a el NAIL BAS AN ASSES MAE SIDE OGIE, |
eC

eee

NSS
WUAUWUUMUU MA W@YUtM000 0 Ls UUUUDGDUGULUULULDLUYYLGLULULUYGYLULULYYGYLELG“LLYGZA

UV.

YU;
WYAWUU MUU VAAL ez ULLGEOLEYALEEOYIL ED AIIIYEYIIFOYYII

itts
Up
WU
UU

ee
ULULUULUDULUAULULUULLDULUYUULODULULUEULULUEUULULUUULUUUU

80
NAN
Ui.

N
UUULULUDULULULULUGLLYUUULDUDUYLUULLDULGULULYGULULLYDLS

oe
CLULLYGYLYYYYEYEYYYODDEEYOGEELLDUELGIS
WOE ODO, YUUDUBULYUUYYOYUUYUEULYUYLUUUUYYLYELULYY!
UUM UUUAV AUT AVA MMA Wala Ua aU UUs Ua a. Ub We tr We Wa Uta Uh ts UUs Wh ta UCUULLUULUDULULLULUOULUULUUDULULULUULUDULUDUUDUEUL, Ol

714
 715
(w77)\y ypbuajaAdM

UUGOYUVVGO
YMCLYOCLYLYYYELYEYYGGYS
GUGUDLUUGUGUUULGEUGLUYLUULGULULUUDULUY

UULUUCUDCDLULUDEE
ULYYYGYLYGYYYGELYY

COL @LYUULULYUUGULLYLUS
YOO.
ABBELZLELYYLYUGLYLLLLES
UUULYYYGYGLYOCUY

ULGYYYLYYGYG4LGZ
UYUGLYYYOEY
9°0

20
20

£0

YUU
877 eee
SNSSSEeP77

UUUG
Y“UUUULGULULLULULLLLYGDGA

YUUYLLYDGULGLULLLGLLLS
to

|Aa
aad

UYUGGYGYY

YG

LTIAAAATALA
BUZZZZZA |

WAT
GLLLLL

20
BYZ7727277eREee

COT
BYZZZZZZAGER

LEGLGLLELE
| \AOOL OT,

BZZZZZ

AOAC
7am
YLULLULLEGLLS
VW IO: AAA

OD)
|
AZ,

GOYLYY
|YOO LYU4.UG,
pene

UWVAUUUUUUWMAa Ue Us Ut 0 -F pt. Ut.Wt

vo

YY

GUY

|
YVAMU YUU. UA 04, MMA AOA A AAA A | LOOT. _| GU
NSS

CEU? ULGLEGLOCGOGAEEEZZE

GOYA
UWUUWU>="10 4 “Uz Ut Ut WWUUUUUWU AE YU.WM

we
OA LOE LB

LY
UU. Ut. Ut. ts Ut. tt. ts Wh WM MOO

EE
YD UW. tt . Ut Us WY WUWUWUUUUUME PEUUM UTR MY
9°0
YU WUWUUWWU 420.12 WUU WU WUWUUUWUUMUUUCLOLUU UA HH 1 A OO OOO
A ALLA
LY OBL LOCLLGELOS CORUM AC ACAGA GGGLGAULGLY

aiduinw
UWAUUYWUA U0 U4. OAM AAA AAAS AIMED I OMA ISA. MODE BYYLY LYALL GLAM
UY WU Ua UUW Ws U0 Wt Ut U4 Ut. tts Ut. Ut, ts st 4 ©YYYLAYYLLLLY AUC LLYIECLDLYLYLEYAYOLDLYEE

8'0
UMM UAV.
WUAUMUUM YU. Uz WU 0, "GWA UM UM WMA UAW DAUM U“U“|>=_’[|™M™ Ua Uh Us U@ 4 0. ts Us YUULEULLUULULGULGLULYLULLGLLLYELELGLLS

UUG
LG GUGULLLG YYUGLLULULULULLLY

ULOU
MAA ALEPPO PO JP0-P-IJJJJY0zK00-0)

Y%
WUWUUUUUWUUUUUUUUUUUUYOY LY VACA AC ALAC AACA ALA ALAA GG
i Wt, We Ute. UteUt.tt taUt Wt WU UUM Va Uh. Us ts Ws 0. ts Ut. tt ta 0. tts. tt, ts hh 4 AAC ALA ALLEL
VACALAL
UULULULULLLLD Ol
|LMuprtitiiiiieo,,_—|

(¥)2 dduDd}JlWSUDI|
WUUUUUMAUWT

<SSNN
SSSs
La S NSGS88
UUM

ee

Cae ee

7 SS
pts
<<
mb
SS
UAdMUMAd 7a
na

WROTE
eee
tt
WUUU

“Amma Ss
UL

AAG
Wt
0.
UM

Wb
0.

WWW
a

UGLY
oT
y,| |_|
SSSGGE8BYUZ77A

NN
Git
4UGG
ULOALDLOUUULGULL

SE
U’UUDCLULUUULUDLELD
HOLE
UY
WW. UW UW Wt Wt Wt Wt Wa Wn We

UZZZAEEZZZLAVYZZZAZCAS

\e
44%

LUG
GU
4%,

VAAL
UUL

NSS
GL,
| GLULEEE
UZ G4
VALALALAG
GGUUG
YOLLELEDOLODOEE

%
UUYYCESBCACAEEYZZ
on
GOV ABBY ZZAEBZZ7 ZABREBEBYZYZUZAZLULYGULLULLULG
GULAEEYUZZAEBZZZZZE || TAA, YGULLLYLLYEY

|HI
UAZACABEYVZZAEBZZZAZZ4 ULLLLD ULLLLYGYGLLLY

BE
42227248

OMA
VZCCCABBYZZAEE
UZZZABELZZZ |_|BOTT,|| ULZLLUELLLULULLLS WAZ

Z
VULZLZLLLUULUULY LUG

<Se
GODLYGYOGY
YUL
GY
WW UME AE WEE OUT OO aso +i oan ig | | WLLLLULUULDLAG te

90
“Lt,
“GU,

SS;
YUMAUUAW MMM MMMM MMMM OUMM WM@@Uweas Wea tt U1’ WW: 0 0 LOGDCEBYYLLUGLULLGOUCOEL ABBZYAUYAUELULULUULULUGUUNG

WS
YYGULGECLLEYLLYLGLEGEGEN ABEL?“4UUULDLULUYUGULELL

ALALL
YUUMMMMMAUGLUMMM MMUUMOMMMMMM UMM0VWHWUMUW WU UWU“EWM O00 06

UGG:

iM
AA CUM MAA A_|ZZ| GZ UUGLGELLOEY

AC
A

SN
GY YUU DED CUED Hh

S$yD8q

%
VALLE LY LOLYDCYLOCLLLYLY ODER
E@CXC@CPCPH#@PePCCPC@P???T@@ BY LLY ACAAAZ,

“LG!

80
YUVA UA UUUUMMU“L aa Us W@YUtUM UM M@0OU UMA WU Ua Ya Uta ts Uo U0 Wo TUG @PEP@OOO
AGG
His
| UUUDCLLUULAPERUZULLLUULUUGULOULULULUULULAGULUUUYULGULLUULY
ULULGULULULUULULULULULUULGULULUUULLULLGLLY YGLULLLOA

44%
WUUUUUU UI UM0UM[EU V# 60 Ua Ut@UMMMA@>MWMF7/« WU. U@™M@™“&400 U0 Ut, Uo Ut. tt 4

4UG4
CAA
UU MUVUUUUUUVAUAUULUMUM MM UMMA UU Uh. th Ut. U4 Uh Ua tu’ Ut Us to Y UCGUULDLUOUDYLLYYLELOYLLYDULLYLYYULLUCLYLYLCLLLD
UUW UCUUULUUDULLEUUUDCULUUUUL*LUUUULULUGUULUUUGL
AMA
AU
AMMA
UUM
AMAA
AV
UV
Aa
UV. Zi
AUWUAUM Ol
 (wi!) yyHuajaaoM
/ Ee) 90 cl ae) ¢0 ZO
PL .ekdi&sten |) ON MW MMittletwwJ]€@-_ PY™(WJ™0¥0.}70.-A | BEBYAUULLLEY: YALL LY, LWOUYYYUYYYYLYEYYYYYYYYUYUbULGb
RBWPAAAaAAUAUUUUUUCUUUWUUMUWUUMMUUWUUMO BEBZZZ “YYYYEEYYLELY GGG AEE ASA AIS RAEN,
UUM WMA
GT MMM@VUMMVWIMWLTM@ MMM@JMMU[“M 06 U0 Uo WW“. a BEY LLLCZLLLLOGLLGS WOE
UUUUUUMUMUUMA MM OMOM@MO@WM@MPPOO UU Ut WX 0s ERVZZZZZZZ LM 20
| OTOL YUULLY,
||VOOR ACCA AAALAC AACA LY YY YY YGYLY YOLGLYLYLLLEELL LLG
BBZZZ22ZU2224G44GLLS COEELELOLELYLOLA GUYUYYUYUUYYYYYYYYGYYY
BELULLULLEY ULUUULULUULUYUULULLULUYUULDULUUYLUULUULUYDELUUUYLOULULULLUYULEL4. vo
BYZZZZZYGZZYGGG4 UGYYYYYYYYYYYYOGOE YY OGY YGYZ WGUULUGYLYULLYGL

JOUDIJIWWUSUDI |
BYAZAZUGULLLLLYLULLULLLUYYZLGLLLYSU“GUYUYUYYEYYGYYYYY YLYLOLYOLLY
BZZZZZZZZ4Z GULGYGLYLYUEEGS LY VULVA VYOUGYYYLYYGULYYYLLYLULLILLZLLLLS
BLLLAGLLYY GOLLY YYOUD OL LDOOLLDY LLL DDOLLDLOLED COMTI TLITL ATLL’ 3 O)
ULYYYYYUUY OYUUYYOOOUUYUUL EUG EL UUUUU UUs yyyyyy “ULV UWULUYUYLGE
YUULULGULUULLGLGYUYEDLLYGYGYGUUGYLUYYGLLLLYGS LWUYYOYYYY LY YYYYYU4YUYbULLL
ULLYLOLLYYYELLYL YELLE LYLE LLY YY LLL LLL. YUL 4-DLOLLY Y
CGULDULULYEYLDULYLYLYODEYELLELEEGYEDLLULEES ULUUULDEG
WWW OOO YIU” 7
UGUULULUUDLULUULDLULULLLGULLULLLEGLGLLIGSG: UYUUYYYYLY GLY
WUUMUUUUUWUUWUWUHUUUUWUWUUUWUWCUHUOUWUWUUUH YYYGYYLYGYLOLLOYOELLYLYG YYUYUDLOLGGULUGDGELYGLEELS: 4
UWUWULUMUUCUOULUM@UMUOMBOMEMMEMMM MEUM MMMM MLOTUMOUA ULM ZY, YUYYUUBYULYLYYLLYY ZGYLY YY LLLL YL LOL LLL LLL
YUMUMUUM UML WAY MUM UMMM UMMA OUUMMMUMLUMMVU“='MU'"0. Ut@ WW Ua 4 a Ua Us Wt Wa Ut te Ut, WCUGLUUUCDDLUDLULULUDGDULUUUCULLULDCDLUUUDDLUUULS
O'v o€ 02 Lo 9°0 s'0 v0 €0 ZO.
YYGUY YYYYYY
LYYOYYYYYY YYYY YYYYY UYYUUUUGYY 00
YLUYYLYLYYDLLLOGEGYY UOYLULLOLEL UUUGGYYLYY
GOGGOGEEEEDDLELELOIDLEL YGUVUUYYY UYYUGLLYY
UUULUYLUUULUELLULUUDULUUUUODLULUUYLYEEL UCUULULUUULY 20
YUUGUYYLUGYYUYYGGGEGSLAMELLA E
GOLLY YLYEESULLLLULULOLYYILOLODS UU UUYGYYYYLY GYLLOLLYY
LLLLELELELLE Go GOLEEELILE

(X)2
UUMUMMUU Oe LYULECLELELLEED: YOY
UUW OO LLYYYLLLLGOLLLOCLYOLY LY GLY UY YUE EYYOO GG GG GUN
GLO v0
Yyvooy emmy, +H YGUGOUYGLYGLLGLYGLYLY LOLOL LYELL GGL IGG
ELEY LY GY OY YY GUY
YWUUMUWUUUUUP RE? Ae CoO OE
YUUUUWUUUUYUUYUO EE GOULD YLLLYLLYELO LLY YELL YLLLYELELLLLD: UUUYLUGYYYOGGLYGYYLL
YUUVULUL UWA MWA OL OO CLLLLLLLLY GLUE YYOYOY LDL DYYDYLY GD LLLU LLY LULL LLY Gays 90
WM UMUMUMUUMUMUMUOOOE UCUUUDGUGYEGYGCGYYLYDGYLUYLYGYGGYGYOLYOLLLYLEES LYY GUY YUL YUU GYY
WUWWUUUUWUWHUUU0 OMELET A
WWUUUWUUWUUUMHUUUOOE YOLOLIEEOA IYO OLOEL YZ UUU4
Wl UMMA WW Ae j CULL ULDEYLDYOLUY GLY ULLAL YOLLYULLLDYYDLGOYLYY 80
WUWUMUUWUYUYOUUMU MOA TT OPFo YYYYYGUYLOOLGGLLLYYGL YOY GY GLEE EGSEG.LYYY YY OY Ys
WUUVUUUUWUUMUUUVUO eee OYA UUULUDGUGYGYGGUYILUGYGLYGEGGYGLGGS eee
YM LM EPPO OCOD BEEBE BELO GLO LLY GOL YG LLG LG EGYGUUULUGUULYYLYYGY
Ul ts Vn. a WW WW WW Ws Wb Wt.Wn Wt Uta th Wt th ta tt ta UW We Ua tn hs GOGLOCLELYLLOYDDLOUUGUDUDAGYODLLLLLULDDOD O11

716
 TH
ypbuajannm
Wi WMUWUUUMUUUOUMUBLLLDLVDLVLUVLLLUALLTO

(w7/)y
UVYYUYUYYLYYY

Ov o'¢€ Oo? Ol rome) 20

WW WWW WLUW UO Ria — PH aa CUGUUGCUUVGG NS
Mo poo HO OO
SO OO oa it Bau HH SrEEErGAA OD
LULLLLLLLLLGI fv g
Seer SEG a ate OBE aan
@LLLULULULLGGGLGOEEE
MOLL. pia
PH ae CCGA, ee
: He
WUMUWUUWYUYYO
re ce te te ee as ee oe LOCO GOCODE
yy : itt
-
’CUUVUEYYLEY GLLLOLELOEOE
starr
©aa
WAM
WA
Eae
ULULLLLLLELOO
PMWM

(X¥)2 aduDdjJIWWSUDI|
UUM UCUMUUMUCUMMOUOLUMU
UOMO
WU ¥ MA
POOO m HHH AE MIA
EEE?” ODODE WHOL 27779 GLOLLDLOLOOE
WMUUUMUVUEU 4¢¢¢¢7euee ee ee GG UUYYGYYUG
WUUMUCMMUUU GG4¢4¢77RER08 MG
YEUWYEUWABYUWYEUWAWS
UUM WUUUUUUEUUBUBUU YO OALL GUGLL COL,
UMM DMM MUM CMU UULLLLLLDULLAA A
DUMMY WWW W Wits YGLULLUULYLLGLLY 80 A sequosqy
WUWUMOUUOU UULGULYYLDDLLYLS Y 03H
MMWR
WMD
WY
WCUMUUMUOUEUUUUULUUUUCMUOU
GULLULOCLYLELDOL
ZY GUL
UUUUYUUUWMUUYUAU YULMMMA Wah UsaMa Uhh aah Wa 4, tb
LULULLULGUAGULLLLS Ua Uh Wa a
Uh. Ol
OYLYYLYYUYY
DMM

MMWEE
WWE

Ov
oe
0?
0'l
WUE
WWW
YWCWWU WW
VME
WWW

MOWMO
WWE

YG4%

2-0 0
LLY

A
YY
Wi

YL

vO
Gyyyys
G4

G4

LG
L%

L%

MOOI OLE LLL OELOE™_ LY,

WMO EEE EEE.
UUW UdMMA UOUM@MMWIUIZMMMMMMM 0M0ULUWDWIOOLOWUAU“RMm“; NG
WU UMM UMLMMMWMA“ALMM#OJMLM MH#Ul#M[];}|UM UW UUM UMA UMA U6 U0 0&0
WU UMMM M/]]4X]}]VVTMY@MC UMMM MAU UMMM A ULUUMUMULHHI U0
WULMUMM#OP0MMMUUX@~{WItI>?0(0:WzWVAWMMMOMM“_0MWU(“LM“MNLOaMM“MJ“M|J M000 tn
WUMMUUUM UMMM MMMM UMMM UMD UUM MMM M@EWMUML ULV Un Wa Uta Us Wa Uta Uh 4, GOULL LOLOL OOM
Uh.
Ub
Wh
W@W
Ub
UU
UMM
Ut.
bh
WU
Ua
Ut
XU
Wb
Yt
tt:
tt
UUM
UWUUUAUUM
W#@
Ui
Ut: ULUULULLLUUULULUGLADAR a
YWUAUMMUEAMMAUVUAMMMMAU“LMDWaa’ee
Wa UWtWaWa Was WWW UW W@W 0 tu WU UULLUOLLUUDLELULUULUYLLLULLAZADADZALZE
UMUMUUOUWHUUUUDHOUWOROUUOBDUOOUCUOUUUWUS
WU ULGULLULULLLGLULLULULLLLULUGLULLLLS
UN”
ULM
-@WM
U1
Ub
0s
U4
U#0
0
U@0H@00®I90 UCGULULCDLUYLLULUYLYYLDUELLYLLG
UUUUUUUMUMMMA
UUUMUUM MUU VAM dh ah Vb Vn Uh Ua Wt. Ut Ut tp ts Ut Ht,
UCULUDULULUULUDUULULULDUUGULLLUUULELES
Table I(2.4.6) Summary of Data on Light Losses in the Eye Caused by Reflection, Absorption, and
Scattering“
Ocular Medium .
[See Fig. 1(2.2)] Absorption Scattering References

Cornea 3-4% at air-corneal in- Transmittance between 25-30% of total light Boettner and Wolter (1962)
terface 0.8 and 0.95, independent scatter in eye. Data on Boynton and Clarke (1964)
of wavelength wavelength dependence LeGrand (1937)
are contradictory. Vos (1963)
Vos and Boogaard (1963)
Aqueous Humor Negligible Boettner and Wolter (1962)
Lens See Table _ I1(2.4.6), Significant Alpern et al. (1965)
III(2.4.6), and IV(2.4.6) Boettner and Wolter (1962)
for details. To convert the Boynton and Clarke (1964)
relative optical densities Cooper and Robson (1969)
given in these tables to Coren and Girgus (1972)
absolute values, a density Gaydon (1938)
of 0.15 is added at all Geeraets et al. (1960)
wavelengths. To convert Ludvigh and McCarthy (1938)
the optical densities to Mellerio (1971)
values applicable to a Norren and Vos (1974)
small pupil, multiply the Ruddock (1964, 1972)
tabulated values by the Said and Weale (1959)
constant factor 1.16. Tan (1971)
Vos (1963)
Vos and Boogaard (1963)
Wald (1945, 1949)
Weale (1954, 1961)
Wolf and Gardiner (1965)
Wright (1952)
Vitreous Body Transmittance of 0.8 or Some
Boettner and Wolter (1962)
more, independent of
Wolf and Gardiner (1965)
wavelength
Macular Pigment Brown and Wald (1953)
DeVries et al. (1953)
Ishak et al. (1962)
See Tables II(2.4.6),
Naylor and Stanworth (1954)
II1(2.4.6), and IV(2.4.6)
Ruddock (1963, 1972)
for details.
Stiles (1953)
Wald (1945, 1949)
Weale (1963)
Retina 30% of the total light Kishto (1968)
scatter in the eye. Vos and Bouman (1964)

“The light losses given in the table are estimates extracted from the literature. There is considerable
variation between measurements made by different investigators, particularly with regard to losses due
to absorption and scattering. These variations are caused by differences between different eyes of the
same age or different age, and by differences between the techniques of measurement used. Some
measurements are made on excised eyes, others by direct psychophysical techniques.
Table II(2.4.6) Optical Densities of the Eye Lens and Macular Pigment as Functions of Wavelength

(nm) A;(A) AOA) Arsm(A) Dri m(A) (nm) A, (A) Au(A) Ar+m(A) Dry u(A)

400 1.200 0.085 1.285 1.528 Te 0.045 0.000 0.045 0.130

405 1.000 0.120 1.120 1.358 580 0.040 0.040 0.122
410 0.820 0.160 0.980 122 585 0.035 0.035 0.113
415 0.670 0.225 0.895 jEXeS3 590 0.031 0.031 0.105
420 0.550 0.300 0.850 1.072 ab 0.027 0.027 0.098
425 0.450 0.345 0.795 1.012 600 0.024 0.000 0.024 0.091
430 0.370 0.365 0.735 0.947 605 0.021 0.021 0.084
435 0.310 0.380 0.690 0.897 610 0.018 0.018 0.078
440 0.270 0.400 0.670 0.872 615 0.015 0.015 0.071
445 0.240 0.425 0.665 0.862 620 0.012 0.012 0.065
450 0.225 0.460 0.685 0.877 625 0.010 0.000 0.010 0.059
455 0.205 0.490 0.695 0.882 630 0.008 0.008 0.054
460 07195 0.495 0.690 0.872 635 0.006 0.006 0.048
465 0.185 0.470 0.655 0.833 640 0.004 0.004 0.043
470 0.175 0.445 0.620 0.995 645 0.003 0.003 0.039
475 0.165 0.410 0.575 0.743 650 0.002 0.000 0.002 0.034
480 0.155 0.415 0.570 0.734 655 0.001 0.001 0.030
485 0.145 0.420 0.565 0.724 660 0.000 0.000 0.026
490 0.140 0.410 0.550 0.705 665 0.000 0.000 0.022
495 0.130 0.360 0.490 0.640 670 0.000 0.000 0.019
500 0.125 0.275 0.400 0.546 675 0.000 0.000 0.000 0.016
505 0.115 0.195 0.310 0.452 680 0.013
510 0.110 0.130 0.240 0.378 685 0.009
515 0.105 0.085 0.190 0.323 690 0.006
520 0.100 0.050 0.150 0.279 695 0.003
az. 0.095 0.025 0.120 0.245 700 0.000 0.000 0.000 0.000
530 0.090 0.010 0.100 0.221 705 — 0.003
ah 0.085 0.000 0.085 0.202 710 — 0.006
540 0.080 0.000 0.080 0.193 715 — 0.009
545 0.075 0.000 0.075 0.184 720 =0.012
550 0.070 0.000 0.070 0.175 725 0.000 0.000 0.000 =0.015
p55 0.065 0.065 0.166 730 — 0.018
560 0.060 0.060 0.157 735 ald Ua
565 0.055 0.055 0.148 740 — 0.024
570 0.050 0.050 0.139 745 ==U.U27

719
 Table II(2.4.6)— Continued

A ,(A)=Optical density of the young human eye lens (completely open pupil) at wavelength A, relative
to that at wavelength A = 700 nm. Values taken from the curve of Fig. 1(2.4.6).
A ,,(A)=Optical density of the macular pigment at wavelength A, covering the fovea. Values taken
from the curve of Fig. 5(2.4.6). The maximum density, occurring at A = 458 nm, has been set
at 0.5.
A, +4(A)=Combined optical density of eye lens and macular pigment at wavelength A.

D, ~y(A)=4z + (A) 110g (700/A)

=Combined optical density of eye lens and macular pigment together with the quantity
log(700/ A).
Adding D, . ,,(A) to the logarithm of a spectral sensitivity function W,(A) given in reciprocal
energy units, corrects that spectral sensitivity function for light losses due to absorption in the
lens and macular pigment and, at the same time, converts it to the logarithm of the spectral
sensitivity function W,(A) in reciprocal quantum units. Thus

log W,(A) = log W.(A)+ Drs (A)

The choice of 700 nm as the wavelength at which D, . ,,(A) equals zero is entirely arbitrary.
The above relation is equivalent to

700 W,(2)
Cie PC)
where ft, . (A) is the spectral transmittance of the combined lens and macular pigments
obtained from

l
19 i Sumer ee
LiM

720
 Table I1I(2.4.6) Transmittances of the Eye Lens and the Macular Pigment as
Functions of Wavelength

A(nm) t7(A) ty(A) A(nm) t7(A) ty(A)

400 0.0631 0.8222 550 0.8511 1.0000

405 0.1000 0.7586 ser 0.8610
410 0.1514 0.6918 560 0.8710
415 0.2138 0.5957 565 0.8810
420 0.2818 0.5012 570 0.8913

425 0.3548 0.4519 575 0.9016 1.0000

430 0.4266 0.4315 580 0.9120
435 0.4898 0.4169 585 0.9226
440 0.5370 0.3981 590 0.9311
445 0.5754 0.3758 595 0.9397

450 0.5957 0.3467 600 0.9462 1.0000

455 0.6237 0.3236 605 0.9528
460 0.6383 0.3199 610 0.9594
465 0.6531 0.3388 615 0.9661
470 0.6683 0.3589 620 0.9727

475 0.6839 0.3890 625 0.9772 1.0000

480 0.6998 0.3846 630 0.9817
485 0.7161 0.3802 635 0.9863
490 0.7244 0.3890 640 0.9908
495 0.7413 0.4365 645 0.9931

500 0.7499 0.5309 650 0.9954 1.0000

505 0.7674 0.6383 655 0.9977
510 0.7762 0.7413 660 1.0000
515 0.7852 0.8222 665 1.0000
520 0.7943 0.8913 670 1.0000

325 0.8035 0.944] 675 1.0000 1.0000

530 0.8128 0.9772 680
535 0.8222 1.0000 685
540 0.8318 1.0000 690
545 0.8414 1.0000 695

t,(A)=Transmittance of the young human eye lens (completely open pupil) at wavelength A, relative to
that at wavelength A = 700 nm, [t,(700) = 1.0].
ty(A)=Transmittance of the macular pigment at wavelength A, covering the fovea. The minimum
transmittance occurring at A = 458 nm has been set at 0.3162.
The data are derived from Table II(2.4.6) in which the optical densities A,;(A) and 4,,(A) of the
eye lens and the macular pigment are given. The following relations have been used:

A (A) = log[1/t,(A)] and Ay,(A) = log[1/ty(A)].

If it is desired to change the total densities of the lens and the macular pigment by factors
a), &, respectively, the new spectral transmittances become

tri(A) and rx/(A)

721
 Table IV(2.4.6) Optical Densities and Transmittances of the Eye Lens and the Macular Pigment as
Functions of Wavenumber

m(cm ') A,(m) =Ay(m) tym) ty(m) | mcm!) Ay(m) — Ay(m) — ty (m) tym)
25,000 1.200 0.085 0.063 0.822 20,000 0.125 0.275 0.750 0.531
24,750 1.038 0.113 0.092 0.771 19,750 0.114 0.178 0.769 0.664
24,500 0.866 0.145 0.136 0.716 19,500 0.107 0.105 0.782 0.785
24,250 0.759 0.191 0.174 0.644 19,250 0.101 0.054 0.793 0.883
24,000 0.630 0.250 0.234 0.562 19,000 0.094 0.021 0.805 0.953
23,750 0.529 0.309 0.296 0.491 18,750 0.087 0.003 0.818 0.993
23,500 0.441 0.347 0.362 0.450 18,500 0.079 0.000 0.834 1.000
25,250 0.358 0.365 0.439 0.432 18,250 0.072 0.000 0.847 1.000
23,000 0.313 0.379 0.486 0.418 18,000 0.064 0.000 0.863 1.000
22,750 0.274 0.398 0.532 0.400 17,750 0.047 0.877
22,500 0.243 0.422 0.571 0.378 17,500 0.049 0.893
22,250 0.227 0.456 0.593 0.350 17,250 0.040 0.912
22,000 0.207 0.487 0.621 0.326 17,000 0.032 0.000 0.929 1.000
21,750 0.195 0.495 0.638 0.320 16,750 0.026 0.942
21,500 0.185 0.469 0.653 0.340 16,500 0.020 0.955
21,250 0.174 0.441 0.670 0.362 16,250 0.015 0.966
21,000 0.163 0.411 0.687 0.388 16,000 0.010 0.000 oye 1.000
20,750 0.151 0.417 0.706 0.383 15,750 0.006 0.986
20,500 0.142 0.414 0.721 0.385 15,500 0.003 0.993
20,250 0.132 0.372 0.738 0.425 15,250 0.001 0.998
15,000 0.000 0.000 1.000 1.000

A,(m)=Optical density of the young human eye lens (completely open pupil) at wavenumber m, relative to that at
wavenumber m = 10//700.
A y,(m)= Optical density of the macular pigment at wavenumber m, covering the fovea. The maximum density occurring at
m = 101/458, has been set at 0.5.
A,(m) and A,,(m) are derived from Table II(2.4.6) by linear interpolation.
t,;(m)=Transmittance of the eye lens at wavenumber m, obtained from

]
A, (m) = log
t,(m)

ty(m)=Transmittance of the macular pigment at wavenumber m, obtained from

A y(m) =loa

m=wavenumber in cm | is related to wavelength A in nm by

ded
Table 1(3.2) Basic Colorimetric Terms and Their Definitions“

Color Color (in the psychophysical sense) is that characteristic of a visible

radiant power by which an observer may distinguish differences
between two structure-free fields of view of the same size and shape,
such as may be caused by differences in the spectral composition of
the radiant power concerned in the observation.
Psychophysical color is specified by the tristimulus values of the
radiant power (color stimulus) entering the eye.
Color stimulus A color stimulus is radiant power of given magnitude and spectral
composition, entering the eye and producing a sensation of color.
Monochromatic A monochromatic stimulus is monochromatic radiant power of given
stimulus magnitude and wavelength (or frequency), entering the eye and
producing a sensation of light or color.
Achromatic An achromatic stimulus is the color stimulus chosen because it
stimulus usually yields a color perception which is devoid of hue under the
desired observing conditions.

Primary Color Primary color stimuli are color stimuli by whose additive mixture
stimuli nearly all other color stimuli may be completely matched in color.
Note 1: These color stimuli are often chosen to be either red, green,
and blue, or red, green, and violet.
Note 2: In accordance with the laws of additive mixture of color
stimuli, primary color stimuli can be defined that have the useful
property that any real color stimulus can be represented by an
additive mixture of positive amounts of the primary color stimuli
(linear combination with positive coefficients).
Tristimulus values The tristimulus values of a color stimulus are the amounts of the
three primary color stimuli required to give by additive mixture a
color match with the color stimulus considered.

Color-matching The color-matching functions are the tristimulus values, with respect
functions to three given primary color stimuli, of monochromatic stimuli of
equal radiance, regarded as functions of the wavelength.
Chromaticity The chromaticity coordinates of a color stimulus are the ratios of each
coordinates tristimulus value of the color stimulus to their sum.
Note: A diagram in which any one of the three chromaticity
coordinates is plotted against any other is called a chromaticity
diagram. In this diagram, the chromaticity of a color stimulus plots
as a point, the chromaticity point.
Dominant The dominant wavelength of a color stimulus is the wavelength of the
wavelength monochromatic stimulus that, when additively mixed in suitable
proportions with a specified achromatic stimulus, yields a color
match with the color stimulus considered.

fx
Table I(3.2)— Continued

Complementary The complementary wavelength of a color stimulus is the wavelength

wavelength of the monochromatic stimulus that, when additively mixed in
suitable proportions with the color stimulus considered, yields a
color match with a specified achromatic stimulus.
Note: Every color stimulus has either a complementary wavelength
or a dominant wavelength. Some, but not all, color stimuli have
both.
Excitation The excitation purity of a color stimulus is the ratio of two lengths on
purity a chromaticity diagram. The first length is the distance between the
point representing the chromaticity of a specified achromatic
stimulus and that representing the chromaticity of the color stimulus
considered; the second length is the distance along the same
direction and in the sense from the first point to the edge of the
chromaticity diagram.

Metameric Metameric color stimuli have identical tristimulus values but different
color stimuli spectral radiant power distributions.

“Colorimetric terms are psychophysical terms of color that refer to the color matching of one visual
half-field with another, and to judgments of similarities and degree of difference between two such
half-fields. Psychophysical terms of color are distinguished from psychological terms which refer to color
perceptions. Psychological terms of color apply to visual concepts that enable the individual observer to
describe color perceptions. Chapter 6 on Uniform Color Scales deals with such concepts.

724
Table I(3.3.1) Color-Matching Functions and Corresponding Chromaticity Coordinates
of the CIE 1931 Standard Colorimetric System, for A = 360 to 830 nm at 1-nm Intervals

Wavelength Color-Matching Functions Chromaticity Coordinates

nN
(nm) y(r) x (A) yr) z(X)

360 0.000 0.000 003 917 000 0.000 0.175 56 0.005 29 0.819 15
361 0.000 0.000 004 393 581 0.000 0.175 48 0.005 29 0.819 23
362 0.000 0.000 004 929 604 0.000 0.175 0.005 28 0.819 32
363 0.000 0.000 005 532 136 0.000 0.175 0.005 27 0.819 41
364 0.000 0.000 006 208 245 0.000 0.175 0.005 26 0.819 50

365 0.000 0.000 006 965 000 0.001 0.175 0.005 26 0.819
366 0.000 0.000 007 813 219 0.001 0.175 0.005 25 0.819
367 0.000 0.000 008 767 336 0.001 0.175 0.005 24 0.819
368 0.000 0.000 009 839 844 0.001 0.174 0.005 23 0.819
369 0.000 0.000 O11 043 23 0.001 0.174 0.005 22 0.819

370 0.000 0.000 012 390 00 0.001 0.174 0.005 22 0.819

371 0.000 0.000 013 886 41 0.002 0.174 0.005 23 0.820
372 0.000 0.000 O15 352, 28 0.002 0.174 0.005 24 0.820
373 0.000 0.000 017 442 96 0.002 0.174 0.005 24 0.820
374 0.000 0.000 019 583 75 0.003 0.174 0.005 22 G.820

370 0.000 0.000 022 020 00 0.003 0.174 0.005 18 0.820

376 0.000 0.000 024 839 65 0.003 0.174 0.005 13 0.820
377 0.000 0.000 028 041 26 0.004 0.174 0.005 07 0.820
378 0.001 0.000 031 531 04 0.005 0.174 0.005 02 0.820
379 0.001 0.000 035 215 21 0.005 0.174 0.004 98 0.820

380 0.001 0.000 039 000 00 0.006 0.174 0.004 96 0.820

381 0.001 0.000 042 826 40 0.007 0.174 0.004 96 0.820
382 0.001 0.000 046 914 60 0.007 0.174 0.004 97 0.820
383 0.001 0.000 051 589 60 0.008 0.174 0.004 98 0.820
384 0.001 0.000 057 176 40 0.009 0.174 0.004 98 0.820

385 0.002 0.000 064 000 00 0.010 0.174 0.004 98 0.821

386 0.002 0.000 072 344 21 0.011 0.173 0.004 97 0.821
387 0.002 0.000 082 212 24 0.013 0.173 0.004 94 0.821
388 0.003 0.000 093 508 16 0.015 0.173 0.004 93 0.821
389 0.003 0.000 106 136 1 0.017 0.473 0.004 92 0.821

390 0.004 0.000 120 000 0 0.020 0.173 0.004 92 0.821

391 0.004 0.000 134 984 0 0.022 0.173 0.004 92 0.821
392 0.005 0.000 151 492 0 0.025 O.1iS 0.004 94 0.821
393 0.005 0.000 170 208 0 0.028 0.173 0.004 94 0.821
394 0.006 0.000 191 816 0 0.031 0.173 0.004 94 0.821

395 0.007 0.000 217 000 0 0.036 0.173 0.004 92 0.821

396 0.008 0.000 246 906 7 0.041 0.173 0.004 90 0.821
a7 0.010 0.000 281 240 0 0.047 0.173 0.004 86 0.821
398 0.011 0.000 318 520 0 0.054 0.173 0.004 84 0.821
399 0.012 0.000 357 266 7 0.060 0.173 0.004 81 0.821

400 0.014 0.000 396 000 0 0.067 0.173 0.004 80 0.821

401 0.015 0.000 433 714 7 0.074 0.173 0.004 79 0.821
402 0.017 0.000 473 024 0 0.081 0.173 0.004 78 0.821
403 0.018 0.000 517 876 0 0.089 0.173 0.004 78 0.822
404 0.020 748 Ol 0.000 572 218 7 0.098 0.173 0.004 77 0.822

rps)
 Table 1(3.3.1)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r
(nm) x(A) yr) x(X) y(X) z(X)

405 0.023 190 00 640 000 0 0.173 02 0.004 78 0.822 20

406 0.026 207 36 0.000 724 560 0 0.172 93 0.004 78 0.822 29
0.029 782 48 825 500 0 0.172 84 0.822 37
0.033 880 92 941 160 0 0.172 qe 0.822 45
0.038 468 24 069 880 0.172 66 0.822 54

0.043 510 00 210 000 0.172 0.822 62

0.048 995 60 362 091 Or iZ 0.822 71
0.055 022 60 530 752 0.172 0.822 81
0.061 718 80 720 368 0.172 0.822 90
0.069 212 00 935 323 0.172 0.822 99

0.077 630 00 180 0.172 0.823 08

0.086 958 11 454 0.171 0.823 16
0.097 176 72 764 0.171 0.823 24
0.108 406 3 117 0.171 0.823 32
0.120 767 2 526 0.171 0.823 40

0.134 380 000 0.171 0.823 49

0.149 358 546 0.171 0.823 58
0.165 395 159 0.170 0.823 68
0.181 983 829 0.170 0.823 76
0.198 611 ih
Ot 546 0.170 0.823 84

0.214 770 300 0.170 0.823 91

0.230 186 086 0.170 0.823 98
0.244 879 908 0.169 0.824 04
0.258 777 767 0.169 0.824 10
0.271 807 o®
Own 664 0.169 0.824 16

0.283 900 600 0.168

0.294 943 573 0.168
0.304 896 582 0.168
0.313 787 629 0.167
0.321 645 ©
CO
hwWN JIS 0.167

0.328 500 840 0.166

0.334 351 007 0.166
0.339 210 214 0.165
0.343 121 453 0.165
0.346 129 WO
NWe 718 0.164

0.348 280 000 0.164

0.349 599 294 0.163
0.350 147 610 0.163
0.350 013 958 0.162
0.349 287 oofwooe 351 CO
\O
(nN
m=
SYwWUNN
©
won
RAW
OW
©
NW
ANA
o
A
oo
©
We 0.161

0.348 060 800 0.161

0.346 333 310 0.160
0.344 262 883 0.159
0.341 808 cof
WO 521 0.158 57
0.339 094 1 PAD m=
©
NON 0.157 63

726
Table 1(3.3.1)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

(nm) x(A) y(n) 2(A) x (X) yr) z()

450 0.336 200 0 0.038 000 00 1.772 110 0 0.156 64 0.017 71 0.825 65
451 0.333 197 7 0.039 846 67 1.768 258 9 0.155 60 0.018 61 0.825 79
452 0.330 041 1 0.041 768 00 1.764 039 0 0.154 52 0.019 56 0.825 92
453 0.326 635 7 0.043 766 00 1.758 943 8 0.153 40 0.020 55 0.826 05
454 0.322 886 8 0.045 842 67 1.752 466 3 0.152 22 0.021 61 0.826 17

455 0.318 700 0 0.048 000 00 1.744 100 0 0.150 99 0.022 74 0.826 27
456 0.314 025 1 0.050 243 68 1733153955 0.149 69 0.023 95 0.826 36
457 0.308 884 0 0.052 573 04 1.720 858 1 0.148 34 0.025 25 0.826 41
458 0.303 290 4 0.054 980 56 1.705 936 9 0.146 93 0.026 63 0.826 44
459 0.297 257 9 0.057 458 72 1.688 737 2 0.145 47 0.028 12 0.826 41

460 0.290 800 0 0.060 000 00 1.669 200 0 0.143 96 0.029 70 0.826 34
461 0.283 970 1 0.062 601 97 1.647 528 7 0.142 41 0.031 39 0.826 20
462 0.276 721 4 0.065 277 52 1.623 412 7 0.140 80 0.033 21 0.825 99
463 0.268 917 8 0.068 042 08 1.596 022 3 0.139 12 0.035 20 0.825 68
464 0.260 422 7 0.070 911 09 1.564 528 0 0.137 37 0.037 40 0.825 23

465 0.251 100 0 0.073 900 00 1.528 100 0 0.135 50 0.039 88 0.824 62
466 0.240 847 5 0.077 016 00 1.486 111 4 0.133 51 0.042 69 0.823 80
467 0.229 851 2 0.080 266 40 Sh ys Be) 0.131 37 0.045 88 0.822 75
468 0.218 407 2 0.083 666 80 1.389 879 9 0.129 09 0.049 45 0.821 46
469 0.206 811 5 0.087 232 80 1.338 736 2 0.126 66 0.053 43 0.819 91

470 0.195 360 0 0.090 980 00 1.287 640 0 0.124 12 0.057 80 0.818 08
471 0.184 213 6 0.094 917 55 1.237 422 3 0.121 47 0.062 59 0.815 94
472 0.173 327 3 0.099 045 84 1.187 824 3 0.118 70 0.067 83 0.813 47
473 0.162 688 1 0.103 367 4 1.138 761 1 0.115 81 0.073 58 0.810 61
474 0.152 283 3 0.107 884 6 1.090 148 0 0.112 78 0.079 89 0.807 33

475 0.142 100 0 0.112 600 0 1.041 900 0 0.109 60 0.086 84 0.803 56
476 0.132 178 6 0.117 532 0 0.994 197 6 0.106 26 - 0.094 49 0.799 25
477 0.122 569 6 0.122 674 4 0.947 347 3 0.102 78 0.102 86 0.794 36
478 0.113 275 2 0.127.992 8 0.901 453 1 0.099 13 0.112 01 0.788 86
479 0.104 297 9 0.133 452 8 0.856 619 3 0.095 31 0.121 94 0.782 75

480 0.095 640 00 0.139 020 0 0.812 950 1 0.091 29 0.132 70 0.776 O1
481 0.087 299 55 0.144 676 4 0.770 517 3 0.087 08 0.144 32 0.768 60
482 0.079 308 04 0.150 469 3 0.729 444 8 0.082 68 0.156 87 0.760 45
483 0.071 717 76 0.156 461 9 0.689 913 6 0.078 12 0.170 42 0.751 46
484 0.064 580 99 0.162 717 7 0.652 104 9 0.073 44 0.185 03 0.741 53

485 0.057 950 01 0.169 300 0 0.616 200 0 0.068 71 0.200 72 0.730 57
486 0.051 862 11 0.176 243 1 0.582 328 6 0.063 99 0.217 47 0.718 54
487 0.046 281 52 0.183 558 1 0.550 416 2 0.059 32 0.235 25 0.705 43
488 0.041 150 88 0.192735 0.520 337 6 0.054 67 0.254 09 0.691 24
489 0.036 412 83 0.199 418 0 0.491 967 3 0.050 03 0.274 00 0.675 97

490 0.032 010 00 0.208 020 0 0.465 180 0 0.045 39 0.294 98 0.659 63
491 0.027 917 20 02170199 0.439 924 6 0.040 76 0.316 98 0.642 26
492 0.024 144 40 0.226 734 5 0.416 183 6 0.036 20 0.339 90 0.623 90
493 0.020 687 00 0.236 857 1 0.393 882 2 0.031 76 0.363 60 0.604 64
494 0.017 540 40 0.247 481 2 0.372 945 9 0.027 49 0.387 92° 0.584 59

727
 Table I(3.3.1)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

A
(nm) x(A) y(r) x(A) y(r) 2(X)

0.014 700 00 0.258 600 0 0.023 46 0.412 70 0.563 84

0.012 161 79 0.270 184 9 0.437 76 0.542 54
0.009 919 960 0.282 293 0.462 95 0.520 78
0.007 967 240 0.295 050 0.488 21 0.498 61
0.006 296 346 0.308 578 CoM. 0.513 40 0.476 12

0.004 900 0.323 000 0.538 42 0.453 41

0.003 777 0.338 402 0.563 07 0.430 65
0.002 945 0.354 685 0.587 12
0.002 424 0.371 698 0.610 45
0.002 236 0.389 287 ANoORO 0.633 01

0.002 400 0.407 300 0 0.654

0.002 925 0.425 629 9 0.675
0.003 836 0.444 309 6 0.696
0.005 174 0.463 394 4 0.715
0.006 982 0.482 939 5 ©
aohAN 0.733

0.009 300 0.503 000 0.750

0.012 149 0.523 569 0.765
0.015 535 0.544 512 0.779
0.019 477 0.565 690 0.792
0.023 992 0.586 965 wWOOWwWS ©
NOW
m 0.802

0.029 100 0.608 200 0.812

0.034 814 0.629 345 0.819
0.041 120 0.650 306 0.825
0.047 985 0.670 875 0.829
0.055 378 0.690 842 NODS 0.832

0.063 270 0.710 000 0.833

0.071 635 0.728 185 0.834
0.080 462 0.745 463 0.833
0.089 739 0.761 969 0.831
0.099 456 0.777 836 ©
ohAN 0.829

0.109 600 0.793 200 0.826

0.120 167 0.808 110 0.822
0.131 114 0.822 496 0.818
0.142 367 0.836 306 0.814
0.153 854 0.849 491 Oo
HAwonrnkh 0.810

0.165 500 0.862 000 0.805

0.177 257 0.873 810 0.801
0.189 140 0.884 962 0.796
0.201 169 0.895 493 0.791
0.213 365 oro
oh
NOUNABRO 0.905 443 NAL
OO 0.786

0.225 749 0.914 850 0.781

0.238 320 0.923 734 0.776
0.251 066 0.932 092 0.771
0.263 992 0.939 922 0.765
0.277 101 ©
0
0O
“IN 0.947 225 Ore
NAA 0.760

728
Table I(3.3.1)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r 2. AAA ae
(nm) (A) yr) 2(A) x (A) y(r) z(A)
540 0.290 400 0 0.954 000 0 0.020 300 00 0.229 62 0.754 33 0.016 05
541 0.303 891 2 0.960 256 1 0.018 718 05 0.236 89 0.748 52 0.014 59
542 O:317 572 6 0.966 007 4 0.017 240 36 0.244 13 0.742 62 0.013 25
543 0.331 438 4 0.971 260 6 0.015 863 64 0.251 36 0.736 61 0.012 03
544 0.345 482 8 0.976 022 5 0.014 584 61 0.258 58 0.730 51 0.010 91

545 0.359 700 0 0.980 300 0 0.013 400 00 0.265 78 0.724 32 0.009 90
546 0.374 083 9 0.984 092 4 0.012 307 23 0.272 96 0.718 06 0.008 98
547 0.388 639 6 0.987 481 2 0.011 301 88 0.280 13, 0711 72. 40,008 “15
548 0.403 378 4 0.990 312 8 0.010 377 92 0.287 29 0.705 32 ,0,007 39
549 0.418 311 5 0:992 811 6 0.009 529 306 0.294 45 0.698 84 0.006 71

550 0.433 449 9 0.994 950 1 0.008 749 999 0.301 60 0.692 31 0.006 09
551 0.448 795 3 0.996 710 8 0.008 035 200 0.308 76 0.685 71 0.005 53
552 0.464 336 0 0.998 098 3 0.007 381 600 0.315 92- 0.679 06 0.005 02
553 0.480 064 0 0:099 112 0 0.006 785 400 0.323 06 0.672 37 0.004 57
554 0.495 971 3 0.999 748 2 0.006 242 800 0.330 21 0.665 63 0.004 16

555 0.512 050 1 1.000 000 0 0.005 749 999 0.337 36 0.658 85 0,003 79
556 0.528 295 9 0.999 856 7 0.005 303 600 0.344 51 0.652 03 0.003 46
DoT 0.544 691 6 0.999 304 6 0.004 899 800 0.351 67 0.645 17 0.003 16
558 0.561 209 4 0.998 325 5 0.004 534 200 0.358 81 0.638 29 0.002 90
559 OS77 821 6 0.996 898 7 0.004 202 400 0.365 96 0.631 38 0.002 66

560 0.594 500 0 0.995 000 0 0.003 900 000 0.373 10 0.624 45 0.002 45
561 0.611 220 9 0.992 600 5 0.003 623 200 0.380 24 0.617 50 0.002 26
562 0.627 975 8 0.989 742 6 0.003 370 600 0.387 38 0.610 54 0,002 08
563 0.644 760 2 0.986 444 4 0.003 141 400 0.394 51 0.603 57 0.001 92
564 0.661 569 7 0.982 724 1 0.002 934 800 0.401 63 0.596 59 0.001 78

565 0.678 400 0 0.978 600 0 0.002 749 999 0.408 73 0.589 61 0.001 66
566 01695 239 2 0.974 083 7 0.002 585 200 0.415 83 0.582 62 0.001 55
567 0.712 058 6 O:969 FF! 2 0.002 438 600 0.422 92. 0.575 63 0.001 45
568 0.728 828 4 0.963 856 8 0.002 309 400 0.429 99 0.568 65 0.001 36
569 0.745 518 8 0.958 134 9 0.002 196 800 0.437 04 0.561 67 0.001 29

570 0.762 100 0 0.952 000 0 0.002 100 000 0.444 06 0.554 72 0,001 22
571 0.778 543 2 0.945 450 4 0.002 017 733 0.451 06 0.547 77 0,001 17
Bie 0.794 825 6 0.938 499 2 0.001 948 200 0.458 04 0.540 84 0.001 12
573 0.810 926 4 0.931 162 8 0.001 889 800 0.464 99 0.533 93 0.001 08
574 0.826 824 8 0.923 457 6 0.001 840 933 0.471 90 0.52705 0.001 05

575 0.842 500 0 0.915 400 0 0.001 800 000 0.478 78 0.520 20 0.001 02
576 0.857 932 5 0.907 006 4 0.001 766 267 0.485 61 0.513 39 0.001 00
377 0.873 081 6 0.898 277 2 0.001 737 800 0.492 41 0.506 61 0.000 98
578 0.887 894 4 0.889 204 8 0.001 711 200 0.499 15 0.499 89 0.000 96
579 0.902 318 1 0.879 781 6 0.001 683 067 0.505 85 0.493 21 0.000 94

580 0.916 300 0 0.870 000 0 0.001 650 001 0.512 49 0.486 59 0.000 92
581 929 FOSeS 0.859 861 3 0.001 610 133 0.519 07 0.480 03 0.000 90
582 0.942 798 4 0.849 392 0 0.001 564 400 0.525 60 0.473 53 0.000 87
583 0.955 277 6 0.838 622 0 0.001 513 600 0.532 07 0.467 09 0,000 84
584 0.967 217 9 0.827 581 3 0.001 458 533 0.538 46 0.460 73. 0.000 81

qe
 Table I(3.3.1)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

dr
(nm) x(A) y(r) 2(A) x(X) y(r) z(X)

585 0.978 600 0 0.001 400 000 0.544 0.454 43 0.000 78

586 0.989 385 6 0.001 336 667 0.551 0.448 23 0.000 74
587 0.999 548 8 270 000 0.557 0.442 10 0.000 71
588 1.009 089 2 205 000 0.563 0.436 06 0.000 67
589 1.018 006 4 146 667 0.569 0.430 10 0.000 64

590 1.026 300 0 100 000 0.575 0.424 23 0.000 62

591 1.033 982 7 068 800 0.580 0.418 46 0.000 60
592 1.040 986 0 049 400 0.586 0.412 76 0.000 59
593 1.047 188 0 035 600 0.592 0.407 19 0.000 59
594 1.052 466 7 021 200 0.597 0.401 76 0.000 58

595 1.056 700 0 0.602 0.396 50 0.000 57

596 1.059 794 4 0.608 0.391 41 0.000 56
597 1.061 TIS 0.612 0.386 48 0.000 54
598 1.062 806 8 0.617 0.381 71 0.000 51
599 1.062 909 6 0.622 0.377 05 0.000 49

600 1.062 200 0 0.627 0.372 49 0.000 47

601 1.060 15252 0.631 0.368 03 0.000 45
602 1.058 443 0.635 0.363 67 0.000 43
603 1.055 224 0.640 0.359 43 0.000 41
604 1.050 976 ff
ON
oo 0.644 0.355 33 0.000 40

605 1.045 600 0.648 0.351 40

606 1.039 036 0.652 0.347 63
607 1.031 360 0.655 0.344 02
608 1.022 666 0.659 0.340 55
609 1.013 047 ©
NOOO 0.662 aS. 22

610 1.002 600 0.665 0.334 OL

611 0.991 367 0.668 0.330 92
612 0.979 331 0.671 0.327 95
613 0.966 491 0.674 0.325 09
614 0.952 847 OnDLUNO 0.677 0.322 36

615 0.938 400 0.680

616 0.923 194 0.682
617 0.907 244 -_wo
So
Ww Ww
o_fon
is)
~
poor)
& oO
aS
i)
~
— 0.684
618 0.890 502 0.687
619 0.872 920 }&
ooco 6.6
SES)
Nahono
ONRDRUO
WoNnNn-o
OS
6
onan
o
Re
An 0.689

620 0.854 449 0.691

621 0.835 084 0.693
622 0.814 946 ies)N ~ 0.695
623 0.794 186 0.697
624 0.772 954 6
©
oS
oro HO
HON So
SS) 0.698
HWHOOWO
O's
Wo
So
ows
JO

625 0.751 400 0.700

626 0.729 583 0.702
627 0.707 588 0.703
628 0.685 602 0.705
629 0.663 810 NM
©
&
OOD aSssae
0.706

730
Table 1(3.3.1)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r
(nm) x(A) y(n) 2(A) ra 0) yA) z(X)

630 0.642 400 0 0.265 000 0 0.000 049 992099 0.707 92 0.292 03 0.000 05
631 0.621 514 9 0.254 763 2 0.000 044 160 00 0.709 23 0.290 72 0.000 05
632 0.601 113 8 0.244 889 6 0.000 039 480 00 0.710 50 0.289 45 0.000 05
633 0.581 105 2 0.235 334 4 0.000 035 720 00 0.711 73 0.288 23 0.000 04
634 0.561 S9RT7 0.226 052 8 0.000 032 640 00 0.712 90 0.287 06 0.000 04

635 0.541 900 0 0.217 000 0.000 030 000 00 0.714 03 0.285 93 0.000 04
636 0.522 599 0.208 161 0.000 027 653 33 O.715012 0.284 84 0.000 04
637 0.503 546 0.199 548 0.000 025 560 00 0.716 16 0.283 80 0.000 04
638 0.484 743 O19) 153 0.000 023 640 00 0.717 16 0.282 81 0.000 03
639 0.466 193 OnHRN 0.182 974 ©
BhNOD 0.000 021 813 33 0.718 12 0.281 85 0.000 03

640 0.447 900 0.175 000 0.000 020 000 00 0.719 03 0.280 94 0.000 03
641 0.429 861 0.167 223 0.060 018 133935 0.719 91 0.280 06 0.000 03
642 0.412 098 0.159 646 0.000 016 200 00 0.720 75 0.279 22 0.000 03
643 0.394 644 0.152 277 0.000 014 200 00 0.724855 0.278 42 0.000 03
644 0.377 533 wNOoOwWO 9.145 125 ONnDKRMNO 0.000 012 133533 0.722 32 0.277 66 0.000 02

645 0.360 800 0.138 200 0.000 010 000 00 0.723 03 0.276 95 0.000 02
646 0.344 456 0.131 500 0.000 007 7332333 0.723 70 0.276 28 0.000 02
647 0.328 516 0.125 024 0.000 005 400 000 0.724 33 0.275 66 0.000 01
648 0.313 019 0.118 779 0.000 003 200 000 0.724 91 0.275 08 0.000 01
649 0.298 001 NO
—=
©
0OW 0.112 769 oo
NO
©
W— 0.000 001 3333333 0.725 47 0.274 53 0.000 00

650 0.283 500 0.107 000 0 0.000 000 000 000 0.725599 0.274 Ol 0.000 00
651 0.269 544 0.101 476 2 0.726 49 0.273 31
652 0.256 118 0.096 188 64 0.726 98 0.273 02
653 0.243 189 0.091 122 96 0.727 43 0.29257
654 0.230 UP.5) NAA
OO 0.086 264 85 0.727 86 0.272 14

655 0.218 700 0.081 600 00 0.728: 27 0278 73

656 0.207 097 0.077 120 64 0.728 66 0.271 34
657 0.195 923 O.0%2. BZD 32 0.729 02 0.270 98
658 0.185 170 0.068 710 08 0.729 36 0.270 64
659 0.174 832 ©
Won 0.064 769 76 0.729 68 0.270 32

660 0.164 900 0.061 000 00 O72) 27, 0.270 03

661 0.155 366 0.057 396 21 0.730 23 0.269 77
662 0.146 230 0.053 955 04 0.730 47 0.269 53
663 0.137 490 0.050 673 76 0.730 69 0.269 31
664 0.129 146 ?o
NOON 0.047 549 65 0.730 90 0.269 10

665 0.121 200 0 0.044 580 00 0.731 09 0.268 91

666 0.113 639) 7 0.041 758 72 0.731 28 0.268 72
667 0.106 465 0 0.039 084 96 0.731 47 0.268 53
668 0.099 690 4 0.036 563 84 O.73 0 65 0.268 35
669 0.093 330 6= 0.034 200 48 0.731 83 0.268 17

670 0.087 400 00 0.032 000 00 0.731) 99 0.268 O1

671 0.081 900 96 0.029 962 61 0.732 15 0.267 85
672 0.076 0.028 076 64 0,732.30 0.267 70
673 0.072 0.026 329 36 0.732 44 0.267 56
674 0.067 686 64 0.024 708 05 0.732 58 0.267 42

731
 Table 1(3.3.1)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

(nm) x(A) y(n) Z(A) x(X) yr) z(X)

675 0.063 600 00 0.023 200 00 0.000 000 0 0.732 72 0.267 28 0.000 00
676 0.059 806 85 0.021 800 77 0.732 86 0.267 14
677 0.056 282 16 0.020 501 12 0.733 00 0.267 00
678 0.052 971 04 0.019 281 08 0.733 14 0.266 86
679 0.049 818 61 0.018 120 69 0.733 28 0.266 72

680 0.046 770 00 0.017 000 00 0.733 42 0.266 58

681 0.043 784 05 0.015 903 79 0.733 55 0.266 45
682 0.040 875 36 0.014 837 18 0.733 68 0.266 32
683 0.038 072 64 0.013 810 68 0.733 81 0.266 19
684 0.035 404 61 0.012 834 78 0.733 94 0.266 06

685 0.032 900 00 0.011 920 00 0.734 05 0.265 95

686 0.030 564 19 0.011 068 31 0.734 14 0.265 86
687 0.028 380 56 0.010 273 39 0.734 22 0.265 78
688 0.026 344 84 0.009 533 311 0.734 29 0.265 71
689 0.024 452 75 0.008 846 157 0.734 34 0.265 66

690 0.022 700 00 0.008 210 000 0.734 39 0.265 61

691 0.021 084 29 0.007 623 781 0.734 44 0.265 56
692 0.019 599 88 0.007 085 424 0.734 48 0.265 52
693 0.018 237 32 0.006 591 476 0.734 52 0.265 48
694 0.016 987 17 0.006 138 485 0.734 56 0.265 44

695 0.015 840 00 0.005 723 000 0.734 59 0.265 41

696 0.014 790 64 0.005 343 059 0.734 62 0.265 38
697 0.013 831 32 0.004 995 796 0.734 65 0.265 35
698 0.012 948 68 0.004 676 404 0.734 67 0.265 33
699 0.012 129 20 0.004 380 075 0.734 69 0.265 31

700 0.011 359 16 0.004 102 000 0.734 69 0.265 31

701 0.010 629 35 0.003 838 453 ' 0.734 69 0.265 31
702 0.009 938 846 0.003 589 099 0.734 69 0.265 31
703 0.009 288 422 0.003 354 219 0.734 69 0.265 31
704 0.008 678 854 0.003 134 093 0.734 69 0.265 31

705 0.008 110 916 0.002 929 000 0.734 69 0.265 31

706 0.007 582 388 0.002 738 139 0.734 69 0.265 31
707 0.007 088 746 0.002 559 876 0.734 69 0.265 31
708 0.006 627 313 0.002 393 244 0.734 69 0.265 31
709 0.006 195 408 0.002 237 275 0.734 69 0.265 31

710 0.005 790 346 0.002 091 000 0.734 69 0.265 31

711 0.005 409 826 0.001 953 587 0.734 69 0.265 31
712 0.005 052 583 0.001 824 580 0.734 69 0.265 31
713 0.004 717 512 0.001 703 580 0.734 69 0.265 31
714 0.004 403 507 0.001 590 187 0.734 69 0.265 31

715 0.004 109 457 0.001 484 000 0.734 69 0.265 31

716 0.003 833 913 0.001 384 496 0.734 69 0.265 31
717 0.003 575 748 0.001 291 268 0.734 69 0.265 31
718 0.003 334 342 0.001 204 092 0.734 69 0.265 31
719 0.003 109 075 0.001 122 744 0.734 69 0.265 31

732
Table 1(3.3.1)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r
(nm) x(A) y(n) z(A) x (A) y(r) z(A)

720 0.002 899 327 0.001 047 000 0.000 000 0 0.734 69 0.265 0.000 00
721 0.002 704 348 0.000 976 589 0.265
722 0.002 523 020 0.000 911 0.265
723 354 168 0.000 850 0.265
724 196 616 0.000 793 0.265

049 190 0.000 740 0.265

910 960 0.000 690 0.265
781 438 0.000 643 0.265
660 110 0.000 599 0.265
546 459 0.000 558 0.265

439 971 0.000 520 0.265

340 042 0.000 483 0.265
246 275 0.000 450 0.265
158 471 0.000 418 0.265
076 430 0.000 388 0.265

0.000 361 0.265

0.000 335 0.265
0.000 311 0.265
0.265
0.265

0.265
0.265
0.265
0.265
sNoOOono
OOD
NM
NOD
Mn
ODN
©
O
&
AO 0.265

0.265
0.265
0.265
0.265
©
Anh 0.265

0.265
0.265
0.265
0.265
0.265

0.265
0.265
0.265
0.265
WwW
ION
OWN
Nee
pf
OW
&
Oe
ON
AO
MOO
m 0.265

0.265
0.265
0.265
0.265
NnNWN
Como 0.734 69 0.265

733
 Table I(3.3.1)— Continued

Wavelength -Color-Matching Functions Chromaticity Coordinates

A
(nm) x(A) yr) 2(A) x (A) y(r) z(X)

0.000 117 413 0 0.000 042 400 00 0.000 000 0 0.734 69 0.265 0.000 00
0.000 109 0.000 039 561 0.734 69 0.265
0.000 102 0.000 036 915 0.734 69 0.265
0.000 095 0.000 034 448 0.734 69 0.265
0.000 089 0.000 032 148 0.734 69 0.265

0.000 083 0.000 030 000 0.734 69 0.265

0.000 077 0.000 027 991 0.734 69 0.265
0.000 072 0.000 026 113 0.734 69 0.265
0.000 067 0.000 024 360 0.734 69 0.265
0.000 062 0.000 022 724 0.734 69 0.265

0.000 058 0.000 021 200 0.734 69 0.265

0.000 054 0.000 019 778 0.734 69 0.265
0.000 051 0.000 018 452 0.734 69 0.265
0.000 047 676 0.000 017 216 0.734 69 0.265
0.000 044 485 0.000 016 064 0.734 69 0.265

0.000 041 0.000 014 990 0.734 69 0.265

0.000 038 0.000 013 987 0.734 69 0.265
0.000 036 0.000 013 051 0.734 69 0.265
0.000 033 0.000 012 178 0.734 69 0.265
0.000 031 0.000 O11 362 0.734 69 0.265

0.000 029 0.000 010 600 0.734 69 0.265

0.000 027 0.000 009 885 0.734 69 0.265
0.000 025 0.000 009 217 0.734 69 0.265
0.000 023 0.000 008 592 0.734 69 0.265
0.000 022 0.000 008 009 0.734 69 0.265

0.000 020 0.000 007 465 0.734 69 0.265

0.000 019 0.000 006 ao 0.734 69 0.265
0.000 017 0.000 006 487 0.734 69 0.265
0.000 016 0.000 006 048 0.734 69 0.265
0.000 015 0.000 005 639 0.734 69 0.265

0.000 014 0.000 005 257 0.734 69 0.265

0.000 013 0.000 004 901 0.734 69 0.265
0.000 012 0.000 004 569 0.734 69 0.265
0.000 O11 0.000 004 260 0.734 69 0.265
0.000 010 0.000 003 971 0.734 69 0.265

0.000 010 0.000 003 702 0.734 69 0.265

0.000 009 0.000 003 452 0.734 69 0.265
0.000 008 0.000 003 218 0.734 69 0.265
0.000 008 0.000 003 000 0.734 69 0.265
0.000 007 0.000 002 737 0.734 69 0.265

0.000 007 yA 0.000 002 607 0.734 69 0.265

0.000 006 732 0.000 002 431 0.734 69 0.265
0.000 006 276 0.000 002 266 0.734 69 0.265
0.000 005 851 0.000 002 113 013 0.734 69 0.265
0.000 005 455 0.000 001 969 943 0.734 69 0.265

734
Table 1(3.3.1)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r
(nm) x(X) y(r) 2(X) x(r) yr) z(X)

810 0.000 005 085 868 0.000 001 836 600 0.000 000 0 0.734 69 0.265 31 0.000 00
811 0.000 004 741 466 0.000 001 TZ 230 0.265
812 0.000 004 420 236 0.000 001 596 228 0.265
813 0.000 004 120 783 0.000 001 488 090 0.265
814 0.000 003 841 716 0.000 001 387 314 0.265

815 0.000 003 581 652 0.000 001 293 400 0.265
816 0.000 003 339 127 0.000 001 205 820 0.265
817 0.000 003 112 949 0.000 001 124 143 0.265
818 0.000 002 902 ea 0.000 001 048 009 0.265
819 0.000 002 705 645 0.000 000 UE) 057 8 0.265

820 0.000 002 522 525 0.000 000 910 930 0.265
821 0.000 002 351 726 0.000 000 849 251 0.265
822 0.000 002 we 415 0.000 000 791 721 0.265
823 0.000 002 043 902 0.000 000 738 090 0.265
824 0.000 001 905 497 0.000 000 688 109 ofpN
WO 0.265

825 0.000 001 776 509 0.000 000 641 530 0.265
826 0.000 001 656 215 0.000 000 598 089 0.265
827 0.000 001 544 022 0.000 000 557 574 0.265
828 0.000 001 439 440 0.000 000 519 808 0.265
829 0.000 001 341 OTT 0.000 000 484 612 wnOoOnNO 0.265

830 0.000 001 251 141 0.000 000 45] 810 0 0.734 69 0.265 at

Note: If the color-matching functions are required at closer intervals than the l-nm
interval given in the above table, a linear interpolation should be used (CIE, 1971).

735
 Table 11(3.3.1) Abridged Values of Chromaticity Coordinates and Color-Matching
Functions of the CIE 1931 Standard Colorimetric System, for A = 380 to 780 nm at
5-nm Intervals

Chromaticity Coordinates rrave e Color-Matching Functions

x() y(X) 2(X) on) X(X) yr) z(N)

0.1741 0.0050 0.8209 380 0.0014 0.0000 0.0065
0.1740 0.0050 0.8210 385 0.0022 0.0001 0.0105
0.1738 0.0049 0.8213 390 0.0042 0.0001 0.0201
0.1736 0.0049 0.8215 395 0.0076 0.0002 0.0362

0.1733 0.0048 0.8219 400 0.0143 0.0004 0.0679

0.1730 0.0048 0.8222 405 0.0232 0.0006 0.1102
0.1726 0.0048 0.8226 410 0.0435 0.0012 0.2074
0.1721 0.0048 0.8231 415 0.0776 0.0022 0.3713
0.1714 0.0051 0.8235 420 0.1344 0.0040 0.6456

0.1703 0.0058 0.8239 425 0.2148 0.0073 1.0391

0.1689 0.0069 0.8242 430 0.2839 0.0116 1.3856
0.1669 0.0086 0.8245 435 0.3285 0.0168 1.6230
0.1644 0.0109 0.8247 440 0.3483 0.0230 1.7471
0.1611 0.0138 0.8251 445 0.3481 0.0298 1.7826

0.1566 0.0177 0.8257 450 0.3362 0.0380 1.7721

0.1510 0.0227 0.8263 455 0.3187 0.0480 1.7441
0.1440 0.0297 0.8263 460 0.2908 0.0600 1.6692
0.1355 0.0399 0.8246 465 0.2511 0.0739 1.5281
0.1241 0.0578 0.8181 470 0.1954 0.0910 1.2876

0.1096 0.0868 0.8036 475 0.1421 0.1126 1.0419

0.0913 0.1327 0.7760 480 0.0956 0.1390 0.8130
0.0687 0.2007 0.7306 485 0.0580 0.1693 0.6162
0.0454 0.2950 0.6596 490 0.0320 0.2080 0.4652
0.0235 0.4127 0.5638 495 0.0147 0.2586 0.3533

0.0082 0.5384 0.4534 500 0.0049 0.3230 0.2720

0.0039 0.6548 0.3413 SOS 0.0024 0.4073 0.2123
0.0139 0.7502 0.2359 510 0.0093 0.5030 0.1582
0.0389 0.8120 0.1491 515 0.0291 0.6082 0.1117
0.0743 0.8338 0.0919 520 0.0633 0.7100 0.0782

0.1142 0.8262 0.0596 p25 0.1096 0.7932 0.0573

0.1547 0.8059 0.0394 530 0.1655 0.8620 0.0422
0.1929 0.7816 0.0255 535 0.2257 0.9149 0.0298
0.2296 0.7543 0.0161 540 0.2904 0.9540 0.0203
0.2658 0.7243 0.0099 545 0.3597 0.9803 0.0134

0.3016 0.6923 0.0061 550 0.4334 0.9950 0.0087

0.3373 0.6589 0.0038 Dad 0.5121 1.0002 0.0057
0.3731 0.6245 0.0024 560 0.5945 0.9950 0.0039
0.4087 0.5896 0.0017 565 0.6784 0.9786 0.0027
0.4441 0.5547 0.0012 570 0.7621 0.9520 0.0021

0.4788 0.5202 0.0010 575 0.8425 0.9154 0.0018

736
Table II(3.3.1)— Continued
ee ee re NES Oni sinninla + behing?) zyn eure ied? BAOt- RY arte
Chromaticity Coordinates Wavelength Color-Matching Functions
r
x (A) y(X) z(X) (nm) X( 2) yr) Z(A)
a
ee
© WH125 0.4866 0.0009 580 0.9163 0.8700 0.0017
0.5448 0.4544 0.0008 585 0.9786 0.8163 0.0014
0.5752 0.4242 0.0006 590 1.0263 0.7570 0.0011
0.6029 0.3965 0.0006 595 1.0567 0.6949 0.0010

0.6270 0.3725 0.0005 600 1.0622 0.6310 0.0008

0.6482 0.3514 0.0004 605 1.0456 0.5668 0.0006
0.6658 0.3340 0.0002 610 1.0026 0.5030 0.0003
0.6801 0.3197 0.0002 615 0.9384 0.4412 0.0002
0.6915 0.3083 0.0002 620 0.8544 0.3810 0.0002

0.7006 0.2993 0.0001 625 0.7514 0.3210 0.0001

0.7079 0.2920 0.0001 630 0.6424 0.2650 0.0000
0.7140 0.2859 0.0001 635 0.5419 0.2170 0.0000
0.7190 0.2809 0.0001 640 0.4479 0.1750 0.0000
0.7230 0.2770 0.0000 645 0.3608 0.1382 0.0000

0.7260 0.2740 0.0000 650 0.2835 0.1070 0.0000

0.7283 0.2717 0.0000 655 0.2187 0.0816 0.0000
0.7300 0.2700 0.0000 660 0.1649 0.0610 0.0000
0.7311 0.2689 0.0000 665 0.1212 0.0446 0.0000
0.7320 0.2680 0.0000 670 0.0874 0.0320 0.0000

0.7327 0.2673 0.0000 675 0.0636 0.0232 0.0000

0.7334 0.2666 0.0000 680 0.0468 0.0170 0.0000
0.7340 0.2660 0.0000 685 0.0329 0.0119 0.0000
0.7344 0.2656 0.0000 690 0.0227 0.0082 0.0000
0.7346 0.2654 0.0000 695 0.0158 0.0057 0.0000

0.7347 0.2653 0.0000 700 0.0114 0.0041 0.0000

0.7347 0.2653 0.0000 705 0.0081 0.0029 0.0000
0.7347 0.2653 0.0000 710 0.0058 0.0021 0.0000
0.7347 0.2653 0.0000 iS 0.0041 0.0015 0.0000
0.7347 0.2653 0.0000 720 0.0029 0.0010 0.0000

0.7347 0.2653 0.0000 725 0.0020 0.0007 0.0000

0.7347 0.2653 0.0000 730 0.0014 0.0005 0.0000
0.7347 0.2653 0.0000 735 0.0010 0.0004 0.0000
0.7347 0.2653 0.0000 740 0.0007 0.0003 0.0000
0.7347 0.2653 0.0000 745 0.0005 0.0002 0.0000

0.7347 0.2653 0.0000 750 0.0003 0.0001 0.0000

0.7347 0.2653 0.0000 755 0.0002 0.0001 0.0000
0.7347 0.2653 0.0000 760 0.0002 0.0001 0.0000
0.7347 0.2653 0.0000 765 0.0001 0.0000 0.0000
0.7347 0.2653 0.0000 7710 0.0001 0.0000 0.0000

0.7347 0.2653 0.0000 775 0.0000 0.0000 0.0000

0.7347 0.2653 0.0000 780 0.0000 0.0000 0.0000

Totals on Fewa) 21.3714 21.3715

737
 Table 1(3.3.2) _Color-Matching Functions and Corresponding Chromaticity Coordinates
of the CIE 1964 Supplementary Standard Colorimetric System, for A = 360 to 830 nm at
1-nm Intervals

Wavelength Color-Matching Functions Chromaticity Coordinates

aN
(nm) X49( A) Nol) Big (A) X19(A) Nol) Z19(A)

360 0.000 000 122 200 0.000 000 013 398 0.000 000 535 027 0.182 22 0.019 98 0.797 80

0.000 000 185 138 0.000 000 020 294 0.000 000 810 720 0.182 0.797 83
0.000 000 278 83 0.000 000 030 56 0.000 001 221 0.182 0.797 86
0.000 000 417 0.000 000 045 74 0.000 001 828 0.182 0.797 89
0.000 000 621 0.000 000 068 05 0.000 002 0.182 0.797 93
0.000 000 919 0.000 000 100 65 0.000 004 0.182 0.797 96

0.000 001 0.000 000 147 98 0.000 005 0.182 0.798 00

0.000 001 0.000 000 216 27 0.000 008 0.182 0.798 04
0.000 002 0.000 000 314 2 0.000 012 0.182 0.798 09
0.000 004 0.000 000 453 7 0.000 018 0.181 0.798 14
0.000 005 0.000 000 651 1 0.000 026 0.181 0.798 20

0.000 008 0.000 000 928 8 0.000 037 0.181 0.798 26

0.000 012 0.000 001 S17 nN 0.000 052 0.181 0.798 32
0.000 017 0.000 001 S572 0.000 074 0.181 0.798 39
0.000 023 0.000 002 602 0.000 104 0.181 0.798 46
0.000 033 0.000 003 625 0.000 146 0.181 0.798 53

0.000 046 0.000 005 019 0.000 202 0.181 0.798 61

0.000 063 0.000 006 907 0.000 279 0.181 0.798 69
0.000 086 0.000 009 0.000 382 0.181 0.798 78
0.000 118 0.000 012 0.000 520 0.181 0.798 88
0.000 159 0.000 017 0.000 704 0.181 0.798 98

0.000 215 0.000 023 0.000 948 0.181 0.799 09

0.000 287 0.000 031 0.001 268 0.181 0.799 20
0.000 381 0.000 041 0.001 686 0.181 0.799 31
0.000 504 0.000 054 0.002 228 0.181 0.799 43
0.000 662 0.000 071 0.002 927 0.180 0.799 55

0.000 0.000 093 0.003 823 0.180 0.799 69

0.001 0.000 120 0.004 964 0.180 0.799 83
0.001 0.000 155 0.006 406 0.180 0.799 99
0.001 0.000 199 0.008 219 0.180 0.800 16
0.002 0.000 253 0.010 482 0.180 0.800 34

0.002 0.000 320 0.013 289 0.180 0.800 55

0.003 0.000 402 0.016 747 0.180 0.800 76
0.004 0.000 502 0.020 980 0.179 0.800 99
0.005 0.000 623 0.026 127 0.179. 0.801 24
0.007 WreNMN~AN 0.000 768 0.032 344 0.179 0.801 49

0.008 0.000 941 0.039 802 0.179 0.801 76

0.010 0.001 147 0.048 691 0.179 0.802 03
0.013 0.001 390 0.059 210 0.178 0.802 31
0.015 0.001 674 0.071 576 0.178 0.802 61
0.019 NnNwowoan 0.002 004 hOWONA 0.086 010 0.178 0.802 90

0.022 0.002 386 0.102 740 0.178 0.803 20

0.027 0.002 822 0.122 000 0.177 0.803 51
0.031 0.003 319 0.144 020 0.177 0.803 83
0.037 0.003 880 0.168 990 0.177 0.804 15
0.043 0.004 509 0.197 120 D.1FF

738
Table I(3.3.2)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

(nm) X49(A) V106A) Z19(A) X19(A) Wola) Z49(A)

406 0.050 223 0.005 209 0.228 570 0.176 84 0.018 34 0.804 82
407 0.057 764 0.005 985 0.263 470 0.176 53 0.018 29 0.805 18
408 0.066 038 0.006 833 0.301 900 0.176 21 0.018 23 0.805 56
409 0.075 033 0.007 757 0.343 870 0.175 86 0.018 18 0.805 96
410 0.084 736 0.008 756 0.389 366 0.175 49 0.018 13 0.806 38

411 0.095 041 0.009 816 0.437 970 0.175 09 0.018 08 0.806 83
412 0.105 836 0.010 918 0.489 220 0.174 65 0.018 02 0.807 33
413 0.117 066 0.012 058 0.542 900 0.174 20 0.017 94 0.807 86
414 0.128 682 0.013 237 0.598 810 0.173 72 0.017 87 0.808 41
415 0.140 638 0.014 456 0.656 760 0.173 23 0.017 81 0.808 96

416 0.152 893 0.015 717 0.716 580 0172) 725 001K 76: 0:809-52
417 0.165 416 0.017 025 0.778 120 0:172) 2iey O.017s 72. 0.810, 07
418 0.178 191 0.018 399 0.841 310 0.171 68 0.017 73 0.810 59
419 0.191 214 0.019 848 0.906 110 0:1715 165: 0.01 F =6O:311.07
420 0.204 492 0.021 391 0.972 542 0.170 63 0.017 85 0.811 52

421 0.217 650 0.022 992 1.038 90 0.170 10 0.017 97 0.811 93

422 0.230 267 0.024 598 1.103 10 0.169 57 0.018 11 0.812 32
423 0.242 311 0.026 213 1.165 10 0.169 02 0.018 28 0.812 70
424 0.253 793 0.027 841 1.224 90 0.168 46 0.018 48 0.813 06
425 0.264 737 0.029 497 1.282 50 0.167 90 0.018 71 0.813 39

426 0.275 195 0.031 195 1.338 20 0.167 33 0.018 97 0.813 70

427 0.285 301 0.032 927 1.392 60 0.166 76 0.019 25 0.813 99
428 0.295 143 0.034 738 1.446 10 0.166 19 0.019 56 0.814 25
429 0.304 869 0.036 654 1.499 40 0.165 61 0.019 91 0.814 48
430 0.314 679 0.038 676 1.553 48 0.165 03 0.020 28 0.814 69

431 0.324 355 0.040 792 1.607 20 0.164 45 0.020 68 0.814 87

432 0.333 570 0.042 946 1.658 90 0.163 88 0.021 10 0.815 02
433 0.342 243 0.045 114 1.708 20 0.163 32 0.021 53 0.815 15
434 0.350 312 0.047 333 1.754 80 0.162 75 90.021 99 0.815 26
435 0.357 719 0.049 602 1.798 50 0.162 17 0.022 49 0.815 34

436 0.364 482 0.051 934 1.839 20 0.161 59 0.023 02 0.815 39

437 0.370 493 0.054 337 1.876 60 0.160 98 0.023 61 0.815 41
438 0:375. 727 0.056 822 1.910 50 0.160 36 0.024 25 0.815 39
439 0.380 158 0.059 399 1.940 80 0.159 71 0.024 95 0.815 34
440 0.383 734 0.062 077 1.967 28 0.159.027 03025. 73. 815.25

441 0.386 327 0.064 737 1.989 10 0.158 32 0.026 53 0.815 15

442 0.387 858 0.067 285 2.005 70 0.157 61 0.027 34 0.815 05
443 0.388 396 0.069 764 2.017 40 0.156 89 0.028 18 0.814 93
444 0.387 978 0.072 218 2.024 40 0.156 15 0.029 07 0.814 78
445 0.386 726 0.074 704 2.027 30 0.155 39 0.030 02 0.814 59

446 0.384 696 0.077 272 2.026 40 0.154 60 0.031 05 0.814 35

447 0.382 006 0.079 979 2.022 30 0.153 77 0.032 19 0.814 04
448 0.378 709 0.082 874 2.015 30 0.152 90 0.033 46 0.813 64
449 0.374 915 0.086 000 2.006 00 0.151 98 0.034 86 0.813 16
450 0.370 702 0.089 456 1.994 80 0.151 00 0.036 44 0.812 56

451 0.366 089 0.092 947 1.981 40 0.150 01 0.038 09 0.811 90

452 0.361 045 0.096 275 1.965 30 0.149 03 0.039 74 0.811 23
453 0.355 518 0.099 535 1.946 40 0.148 04 0.041 45 0.810 SI
454 0.349 486 0.102 829 1.924 80 0.147 02 0.043 26 0.809 72
455 0.342 957 0.106 256 1.900 70 0.145 94 0.045 22 0.808 84

739
 Table 1(3.3.2)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

aN
(nm) Ero( XI Mo lA) Bo (N 19(A) Nolr) Z19(A)

456 0.335 0.109 901 0.144 0.047 0.807

457 0.328 0.113 835 0.143 0.049 0.806
0.320 0.118 167 0.142 0.052 0.805
0.311 0.122 932 0.140 0.055 0.803
0.302 0.128 201 0.138 0.058 0.802

0.292 0.133 457 0.137 0.062 0.800

0.283 0.138 323 0.135 0.066 0.798
0.274 0.143 042 0.133 0.069 0.796
0.264 0.147 787 0.131 0.073 0.794
0.254 0.152 761 0.129 0.077 0.792

0.243 0.158 0.127 0.082 0.790

0.232 0.163 0.124 0.087 0.787
0.220 0.170 0.121 0.094 0.784
0.208 O17 0.118 0.101 0.780
0.195 0.185 0.115 0.109 OnnS

0.183 0.193 0.111 0.117 0.770

0.170 0.200 0.108 0.127 0.764
0.157 0.207 0.104 0.137 0.758
0.144 0.213 0.100 0.147 0.752
0.132 0.219 0.095 0.159 0.745

0.120 0.226 0.091 0.171 0.737

0.109 0.232 0.086 0.184 0.728
0.099 0.239 0.082 0.198 0.719
0.089 0.245 0.077 0.213 0.709
0.080 0.253 0.072 0.229 0.697

0.072 0.261 0.067 0.246 0.685

0.063 0.270 0.062 0.265 0.672
0.055 0.279 0.056 0.285 0.658
0.048 0.288 0.050 0.305 0.643
0.041 0.297 0.045 0.327 0.627

0.034 0.306 0.349 0.610

0.028 0.315 0.372 0.593
0.023 TOF, 0.394 0.575
0.019 07331 0.417 0.557
0.016 Yia39 0.440 0.538

0.013 0.347 0.463 0.519

0.010 0.358 0.487 0.498
0.008 0.370 0.512 0.476
0.006 0.382 0.537 0.453
0.005 0.395 0.562 0.430

0.003 0.408 0.587 0.406

0.003 0.421 0.611 0.383
0.002 0.434 0.634 0.361
0.003 0.447 0.655 0.340
0.003 0.460 0.674 6.319

0.005 0.474 0.692 0.300

0.006 0.488 0.708 0.281
0.009 0.502 0.724 0.262
0.012 147 0.516 0.739 0.243
0.015 444 0.531 0.752 0.225

740
Table 1(3.3.2)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r
(nm) ViolA) 219(A) %49(A) NolA) 219 (A)
506 0.546 190 0.148 673 0.026 0.764 0.208 ee
507 0.561 180 0.138 609 0.032 0.776 0.191 10
0.576 290 0.129 096 0.037 0.786 0.176
0.591 500 0.120 215 0.043 0.794 0.161
0.606 741 0.112 044 0.049 0.802 0.148

0.622 150 0.104 710 0.055 0.808 0.136

0.637 830 0.098 196 0.062 0.812 0.125
0.653 710 0.092 361 0.069 0.815 0.115
0.669 680 0.087 088 0.077 0.816 0.106
0.685 660 0.082 248 0.085 0.816 0.098

0.701 550 0.077 744 0.093 0.816 0.090

0.717 230 0.073 456 0.101 0.815 0.083
O732 570 0.069 268 0.109 0.813 0.076
0.747 460 0.065 060 0.117 0.812 0.070
0.761 TST 0.060 709 0.125 0.810 0.064

0.775 340 0.056 457 0.133 0.807 0.058

0.788 220 0.052 609 0.141 0.804 0.053
0.800 460 0.049 £22 0.149 0.801 0.049
0.812 140 0.045 954 0.158 0.796 0.045
0.823 330 0.043 050 0.166 OS792 0.041

0.834 120 0.040 368 0.174 0.787 0.038

0.844 600 0.037 839 0.183 0.781 0.035
0.854 870 0.035 384 0.191 0.776 0.032
0.865 040 0.032 949 0.199 0.771 0.029
0.875 211 0.030 451 0.207 0.766 0.026

0.885 370 0.028 029 0.214 0.761 0.024

0.895 370 0.025 862 0.222 0.756 0.021
0.905 150 0.023 920 0.229 0.750 0.019
0.914 650 0.022 174 0.236 0.745 0.018
0.923 810 0.020 584 0.243 0.739 0.016

0.932 550 0.019 127 0.250 0.734 0.015

0.940 810 0.017 740 O25 0.728 0.013
0.948 520 0.016 403 0.264 0.722 0.012
0.955 600 0.015 064 0.271 0.717 0.011
0.961 988 0.013 676 0.278 0.711 0.010

0.967 540 0.012 308 0.285 0.705 0.008

0.972 230 0.011 056 0-292, 0.699 0.007
0.976 170 0.009 915 05299 0.693 0.007
0.979 460 0.008 872 0.306 0.687 0.006
0.982 200 0.007 918 0.313 0.681 0.005

0.984 520 0.007 030 0.320 0.675 0.004

0.986 520 0.006 223 0.326 0.668 0.004
0.988 320 0.005 453 0.333 0.662 0.003
0.990 020 0.004 714 0.340 0.656 0.003
0.991 761 0.003 988 0.347 0.650 0.002

0.993 530 0.003 289 0.354 0.643 0.002

0.995 230 0.002 646 0.360 0.637 0.001
0.996 770 0.002 063 0.367 0.631 0.001
0.598 290 0.998 090 0.001 533 0.374 0.624 0.000
0.616 053 0.999 110 0.001 091 0.381 0.618 0.000

741
 Table 1(3.3.2)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

N
(nm) Fol A) Mo lA) Big (A) X19(A) NolA) 219(A)

0.633 948 0.999 770 0.000 711 0.387 0.611 69 0.000 44

0.651 901 1.000 000 0.000 407 0.394 0.605 0.000 25
0.669 0.999 710 0.000 184 0.401 0.598 0.000 11
0.687 0.998 850 0.000 047 0.407 0.592 0.000 03
0.705 0.997 340 0.000 000 0.414 0.585 0.000 00

0.722 0.995 260 0.420 0.579

0.740 0.992 740 0.427 0.572
0.758 0.989 750 0.433 0.566
0.776 0.986 300 0.440 0.559
0.793 0.982 380 0.446 0.553

0.811 0.977 0.453 0.546

0.828 0.973 0.459 0.540
0.845 0.967 0.466 0.533
0.862 0.961 0.472 0.527
0.878 0.955 0.479 0.520

0.894 0.948 0.485 0.514

0.909 0.940 0.491 0.508
0.923 0.932 0.497 0.502
0.937 0.924 0.503 0.496
0.951 0.915 0.509 0.490

0.964 0.905 0.515 0.484

0.977 0.896 0.521 0.478
0.989 0.887 0.527 0.472
1.001 0.877 0.532 0.467
1.014 0.868 0.538 0.461

1.026 0.860 0.544 0.455

1.038 0.549 0.450
0.554 0.445
0.560 0.439
0.565 0.434

0.570 0.429
0.575 0.424
0.580 0.419
0.585 0.414
0.589 0.410

0.594 0.405
0.598 0.401
beakWw—— 0.603 0.396
0.607 0.392
viSee
ein 0.611 0.388

0.615 0.384
0.619 0.380
— Ww a 0.623 0.376
0.627 0.372
‘one
Snae 0.630 0.369

0.634 0.365
0.637 0.362
0.640 0.359
0.644 0.355
0.647 0.352

742
Table 1(3.3.2)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r
(nm) Vi9(A) Fo (A) X49(A) Vio) 249 (A)

606 0.580 781 0.000 000 0.650 0.349 87 0.000 00

607 0.567 653 0.653 0.346
608 0.554 490 0.655 0.344
609 0.541 228 0.658 0.341
610 0.527 963 0.661 0.338

611 0.514 634 0.663 0.336

612 0.501 363 0.666 0.333
613 0.488 124 0.668 0.331
614 0.474 935 0.670 0.329
615 0.461 834 0.673 0.326

0.448 823 0.675 0.324

0.435 917 0.677 0.322
0.423 153 0.679 0.320
0.410 526 0.680 0.319
0.398 057 0.682 0.317

0.385 835 0.684 0.315

0.373 951 0.685 0.314
0.362 311 0.687 0.312
0.350 863 0.688 0.311
0.339 554 0.689 0.310

0.328 309 0.690 0.309

0.317 118 0.692 0.307
0.305 936 0.693 0.306
0.294 737 0.694 0.305
0.283 493 0.695 0.304

0.272 gle 0.696 0.303

0.260 990 0.697 0.302
0.249 877 0.698 0.301
0.238 946 0.699 0.300
0.228 254 0.700 0.299

0.217 853 0.702 0.297

0.207 780 0.703 0.296
0.198 072 0.704 0.296
0.188 748 0.704 0.295
0.179 828 0.705 0.294

0.171 285 0.706 0.293

0.163 059 0.707 0.292
0.155 151 0.708 0.291
0.147 535 0.709 0.290
0.140 211 0.710 0.289

0.133 170 0.711 0.288

0.126 400 0.711 0.288
0.119 892 0,712 0.287
0.113 640 0.713 0.286
0.107 633 0.713 0.286

0.101 870 0.714 0.285

0.096 347 0.714 0.285
0.091 063 0.715 0.284
0.086 010 0.715 0.284
0.081 187 0.715 0.284

743
 Table I(3.3.2)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r
(nm) Erol A) Myo (A) 219 (A) *19(A) NiolA) ZA)

656 0.192 950 0.000 000 0.715 87 0.284 13 0.000 00

657 0.182 110
658 0.171 770
0.161 920
0.152 568

0.143 670
0.135 200
0.127
0.119
0.112

0.105

0.280 31
0.280 24
0.280 17
0.280 11
0.280 06

oococcoco
aooo°o°o
ooo°co
woorce
NOOO Mm
0coOWN
Ww
cofhOoNnwn
oo
\O
MN
&
o~
Iw
OW
We
W
—

oo°o°0oo at
SN
ea

744
Table 1(3.3.2)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r
(nm) Vi0(A) Bo(A) X49(A) Mola) Z10(A)

706 0.002 380 48 0.000

000 00 0.720 31 0.279 69 0.000 00
707 0.002 209 0.720 29 0.279 71
708 0.002 051 0.720 27
709 0.001 904 0.720 25
710 0.001 768 0.720 23

0.001 642 0.720 20

0.001 525 0.720 18
0.001 416 0.720 15
0.001 315 0.720 12
0.001 222 0.720 09
0.720 06
0.720 02
0.719 99
0.719 95
0.719 91

0.719 87
0.719 83
0.719 78
0.719 74
0.719 69

0.719 64
0.719 60
0.719 55
0.719 50
0.719 45

0.719 40
0.719 34
0.719 29
0.719 24
0.719 19

0.719 13
0.719 08
0.719 02
0.718 96
0.718 91

0.000 473 0.718 85

0.000 440 0.718 79
0.000 410 580 0.718 73
0.000 382 490 0.718 67
0.000 356 380 0.718 61

0.000 332 110 0.718 55

0.000 309 550 0.718 48
0.000 288 580 0.718 42
0.000 269 090 0.718 36
0.000 250 969 0.718 29

0.000 234 130 0.718 23

0.000 218 470 0.718 16
0.000 203 910 0.718 10
0.000 190 350 0.718 03
0.000 177 730 0.717 96

745
 Table 1(3.3.2)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r
(nm) X19( A) Mo (A) Bo (A) 19) i0\A) Z19(A)

756 0.000 065 2a2 0.000 000 000 0 0.717 89 0.000 00

757 0.000 060 939 0.717 82
0.000 056 DJL? IS
0.000 053 0.717 68
0.000 049 0.717 61

0.000 046 0.717 54

0.000 043 0.717 46
0.000 040 0.717 39
0.000 038 U.iie oe
oo 0.000 035 0.717 24

0.000 084 0.000 033 0.717 16

0.000 078 0.000 031 0.717 09
0.000 073 0.000 029 0.717 O1
0.000 068 0.000 027 0.716 94
0.000 064 0.000 025 0.716 86

0.000 060 0.000 023 0.716 78

0.000 056 0.000 022 0.716 70
0.000 052 0.000 020 0.716 62
0.000 049 0.000 019 0.716 54
0.000 046 0.000 018 0.716 46

0.000 043 0.000 017 0.716 38

0.000 040 0.000 016 0.716 30
0.000 038 0.000 015 i=)—oO
So
\o[o-e) 0.716 22
0.000 035 0.000 014 0.716 14
0.000 033 0.000 013 oOonoshN
hohne 0.716 06

0.000 031 0.000 012 0.715 97

0.000 029 0.000 O11 0.715 89
0.000 027 0.000 010 \oN | 0.715 81
0.000 025 0.000 010 Ott) 12
0.000 024 0.000 009 DWOoOnond 0.715 64

0.000 022 0.000 009 O:715 S35

0.000 021 0.000 008 0.715 47
0.000 019 0.000 007 0.715 38
0.000 018 0.000 007 0.715 29
0.000 017 0.000 007 0.715 21

0.000 016 0.000 006

0.000 015 0.000 006
0.000 014 0.000 005
0.000 013 0.000 005
0.000 012 orerc
al
a
So
Sod
Sel
Sf
(Se
SOS
SO 0.000
nAnoooceo
Se
oe
al
me
aooo°oeeo 005

0.000 012 0.000 004

0.000 O11 0.000 004
0.000 010 0.000 004
0.000 010 0.000 004
0.000 009 Ooo
Ao
© 0.000 003

0.000 008 0.000 003

0.000 008 0.000 003
0.000 007 0.000 003
0.000 007 0.000 002 0.713 96
0.000 006 0.000 002 0.713 87 0.286 13

746
Table 1(3.3.2)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates

r
(nm) X19(A) Viol) 219(A) X10(A) — MolA) —-B19(A)
806 0.000 006 501 50 0.000 002 607 05 0.000 000 0000 =0.713 78 0.286 22 0.000 00
807 0.000 006 115 30 0.000 002 453 29 0.713 69 0.286 31
808 0.000 00S 752 90 0.000 002 308 94 0.713 60 0.286 40
809 0.000 005 412 70 0.000 002 173 38 0.713 50 0.286 50
810 0.000 005 093 47 0.000 002 046 13 0.713 41 0.286 59

811 0.000 004 793 80 0.000 001 926 62 0.713 32 0.286 68

812 0.000 004 512 50 ~=0.000 001 814 40 0.713 22 0.286 78
813 0.000 004 248 30 0.000 001 708 95 0.713 13 0.286 87
814 0.000 004 000 20 0.000 001 609 88 0.713 04 0.286 96
815 0.000 003 767 10 0.000 001 516 77 0.712 94 0.287 06

816 0.000 003 548 00 0.000 001 429 21 O12) Sam 0287 15
817 0.000 003 342 10 0.000 001 346 86 0.712 76 0.287 24
818 0.000 003 148 50 0.000 001 269 45 0.712 66 0.287 34
819 0.000 002 966 50 ~=0.000 001 196 62 O712°57 9 O287 43
820 0.000 002 795 31 0.000 001 128 09 0.712 47 0.287 53

821 0.000 002 634 50 = 0.000 001 063 68 O12 Shee 257 62
822 0.000 002 483 40 0.000 001 003 13 O.712 28, 90287 72
823 0.000 002 341 40 0.000 000 946 22 0.712 19 0.287 81
824 0.000 002 207 80 0.000 000 892 63 0.712 09 0.287 91
825 0.000 002 082 00 0.000 000 842 16 0.712 00 0.288 00

826 0.000 001 963 60 0.000 000 794 64 O71 SE O26s 10

827 0.000 001 851 90 0.000 000 749 78 0.711 81 0.288 19
828 0.000 001 746 50 0.000 000 707 44 U711 7S GO288 29
829 0.000 001 647 10 0.000 000 667 48 0.711 62 0.288 38
830 0.000 001 553 14 0.000 000 629 70 0.711 52 0.288 48

Note: If the color-matching functions are required at closer intervals than the 1-nm
interval given in the above table, a linear interpolation should be used (CIE, 1971).

747
 Table II(3.3.2) Abridged Values of Color-Matching Functions and Chromaticity
Coordinates of the CIE 1964 Supplementary Standard Colorimetric System, from
A = 380 to 780 nm at 5-nm Intervals
Wavelength Color-Matching Functions Wavelength Color-Matching Functions
r ee r LL

380 0.0002 0.0000 0.0007 580 1.0142 0.8689 0.0000

385 0.0007 0.0001 0.0029 585 1.0743 0.8256
390 0.0024 0.0003 0.0105 590 1.1185 0.7774
395 0.0072 0.0008 0.0323 595 1.1343 0.7204

400 0.0191 0.0020 0.0860 600 1.1240 0.6583

405 0.0434 0.0045 0.1971 605 1.0891 0.5939
410 0.0847 0.0088 0.3894 610 1.0305 0.5280
415 0.1406 0.0145 0.6568 615 0.9507 0.4618
420 0.2045 0.0214 0.9725 620 0.8563 0.3981

425 0.2647 0.0295 1.2825 625 0.7549 0.3396

430 0.3147 0.0387 1.5535 630 0.6475 0.2835
435 0.3577 0.0496 1.7985 635 0.5351 0.2283
440 0.3837 0.0621 1.9673 640 0.4316 0.1798
445 0.3867 0.0747 2.0273 645 0.3437 0.1402

450 0.3707 0.0895 1.9948 650 0.2683 0.1076

455 0.3430 0.1063 1.9007 655 0.2043 0.0812
460 0.3023 0.1282 1.7454 660 0.1526 0.0603
465 0.2541 0.1528 1.5549 665 0.1122 0.0441
470 0.1956 0.1852 1.3176 670 0.0813 0.0318

475 0.1323 0.2199 1.0302 675 0.0579 0.0226

480 0.0805 0.2536 0.7721 680 0.0409 0.0159
485 0.0411 0.2977 0.5701 685 0.0286 0.0111
490 0.0162 0.3391 0.4153 690 0.0199 0.0077
495 0.0051 0.3954 0.3024 695 0.0138 0.0054

500 0.0038 0.4608 0.2185 700 0.0096 0.0037

505 0.0154 0.5314 0.1592 705 0.0066 0.0026
510 0.0375 0.6067 0.1120 710 0.0046 0.0018
515 0.0714 0.6857 0.0822 715 0.0031 0.0012
520 0.1177 0.7618 0.0607 720 0.0022 0.0008

525 0.1730 0.8233 0.0431 725 0.0015 0.0006

530 0.2365 0.8752 0.0305 730 0.0010 0.0004
535 0.3042 0.9238 0.0206 735 0.0007 0.0003
540 0.3768 0.9620 0.0137 740 0.0005 0.0002
545 0.4516 0.9822 0.0079 745 0.0004 0.0001

550 0.5298 0.9918 0.0040 750 0.0003 0.0001

555 0.6161 0.9991 0.001 1 E55 0.0002 0.0001
560 0.7052 0.9973 0.0000 760 0.0001 0.0000
565 0.7938 0.9824 765 0.0001
570 0.8787 0.9555 770 0.0001

575 0.9512 0.9152 775 0.0000

580 1.0142 0.8689 780 0.0000

Totals: 23.3294 23-3005 23.3343

748
Table II(3.3.2)— Continued

Wavelength Chromaticity Coordinates Wavelength Chromaticity Coordinates

r r
(nm) X19 ( A) Niolr) Z19(A) (nm) X49(A) NiolA) Z49(A)

380 0.1813 0.0197 0.7990 0.5386 0.4614 0.0000

385 0.1809 0.0195 0.7996 0.5654 0.4346
390 0.1803 0.0194 0.8003 0.5900 0.4100
395 0.1795 0.0190 0.8015 0.6116 0.3884

400 0.1784 0.0187 0.8029 0.6306 0.3694

405 0.1771 0.0184 0.8045 0.6471 0.3529
410 0.1755 0.0181 0.8064 0.6612 0.3388
415 0.1732 0.0178 0.8090 0.6731 0.3269
420 0.1706 0.0179 0.8115 0.6827 0.3173

425 0.1679 0.0187 0.8134 0.6898 0.3102

430 0.1650 0.0203 0.8147 0.6955 0.3045
435 0.1622 0.0225 0.8153 0.7010 0.2990
440 0.1590 0.0257 0.8153 0.7059 0.2941
445 0.1554 0.0300 0.8146 0.7103 0.2898

450 0.1510 0.0364 0.8126 0.7137 0.2863

455 0.1459 0.0452 0.8088 0.7156 0.2844
460 0.1389 0.0589 0.8022 0.7168 0.2832
465 0.1295 0.0779 0.7926 0.7179 0.2821
470 0.1152 0.1090 0.7758 0.7187 0.2813

475 0.0957 0.1591 0.7452 0.7193 0.2807

480 0.0728 0.2292 0.6980 0.7198 0.2802
485 0.0452 0.3275 0.6273 0.7200 0.2800
490 0.0210 0.4401 0.5389 0.7202 0.2798
495 0.0073 0.5625 0.4302 0.7203 0.2797

500 0.0056 0.6745 073199 0.7204 0.2796

505 0.0219 0.7526 0.2256 0.7203 0.2797
510 0.0495 0.8023 0.1482 0.7202 0.2798
515 0.0850 0.8170 0.0980 0.7201 0.2799
520 0.1252 0.8102 0.0646 0.7199 0.2801

x25 0.1664 0.7922 0.0414 0.7197 0.2803

530 0.2071 0.7663 0.0267 0.7195 0.2806
535 0.2436 0.7399 0.0165 0.7192 0.2808
540 0.2786 0.7113 0.0101 0.7189 0.2811
545 0.3132 0.6813 0.0055 0.7186 0.2814

550 0.3473 0.6501 0.0026 0.7183 0.2817

355 0.3812 0.6182 0.0007 0.7180 0.2820
560 0.4142 0.5858 0.0000 0.7176 0.2824
565 0.4469 0.5531 OItT2 0.2828
570 0.4790 0.5210 0.7169 0.2831

575 0.5096 0.4904 0.7165 0.2835

580 0.5386 0.4614 0.7161 0.2839
Sir
nn nnn nnnn ne eaten nnn nnn nnnnEnnnnEnEEnEE EEEIEIEIO

749
 Table I(3.3.3) Chromaticity Coordinates and Color-Matching Functions of the CIE
1931 Standard Colorimetric System with Respect to Real Primary Stimuli: R at
Ap = 700.0 nm; G at A, = 546.1 nm; B at Az = 435.8 nm

Chromaticity Coordinates Wavelength Color-Matching Functions

r
r(X) g(r) ZeN) GUS, F(X) B(A) B(d)
0.0272 —0.0115 0.9843 380 0.00003 —0.00001 0.00117
0.0268 —0.0114 0.9846 385 0.00005 —0.00002 0.00189
0.0263 —0.0114 0.9851 390 0.00010 —0.00004 0.00359
0.0256 —0.0113 0.9857 395 0.00017 — 0.00007 0.00647

0.0247 —0.0112 0.9865 400 0.00030 —0.00014 0.01214

0.0237 —0.0111 0.9874 405 0.00047 —0.00022 0.01969
0.0225 —0.0109 0.9884 410 0.00084 —0.00041 0.03707
0.0207 —0.0104 0.9897 415 0.00139 —0.00070 0.06637
0.0181 —0.0094 0.9913 420 0.00211 —0.00110 0.11541

0.0142 —0.0076 0.9934 425 0.00266 —0.00143 0.18575

0.0088 —0.0048 0.9960 430 0.00218 —0.00119 0.24769
0.0012 —0.0007 0.9995 435 0.00036 —0.00021 0.29012
—0.0084 0.0048 1.0036 440 —0.00261 0.00149 0.31228
—0.0213 0.0120 1.0093 445 —0.00673 0.00379 0.31860

—0.0390 0.0218 1.0172 450 —0.01213 0.00678 0.31670

—0.0618 0.0345 1.0273 455 —0.01874 0.01046 0.31166
—0.0909 0.0517 1.0392 460 —0.02608 0.01485 0.29821
—0.1281 0.0762 1.0519 465 —0.03324 0.01977 0.27295
—0.1821 0.1175 1.0646 470 — 0.03933 0.02538 0.22991

—0.2584 0.1840 1.0744 475 —0.04471 0.03183 0.18592

—0.3667 0.2906 1.0761 480 —0.04939 0.03914 0.14494
—0.5200 0.4568 1.0632 485 —0.05364 0.04713 0.10968
—0.7150 0.6996 1.0154 490 —0.05814 0.05689 0.08257
—0.9459 1.0247 0.9212 495 —0.06414 0.06948 0.06246

501 685 1.3905 0.7780 500 —0.07173 0.08536 0.04776

his 182 137195 0.5987 505 —0.08120 0.10593 0.03688
=i3074 1.9318 0.4053 510 —0.08901 0.12860 0.02698
—7 12076 1.9699 0.2377 215 —0.09356 0.15262 0.01842
— 0.9830 1.8534 0.1296 520 —0.09264 0.17468 0.01221

— 0.7386 1.6662 0.0724 a Pes —0.08473 0.19113 0.00830

Uo 159 1.4761 0.0398 530 —0.07101 0.20317 0.00549
— 0.3304 1.3105 0.0199 aes —0.05316 0.21083 0.00320
—0.1707 1.1628 0.0079 540 —0.03152 0.21466 0.00146
—0.0293 1.0282 0.0011 545 —0.00613 0.21487 0.00023

0.0974 0.9051 —0.0025 550 0.02279 0.21178 —0.00058

0.2121 0.7919 — 0.0040 555 0.05514 0.20588 —0.00105
0.3164 0.6881 —0.0045 560 0.09060 0.19702 —0.00130
0.4112 0.5932 —0.0044 565 0.12840 0.18522 —0.00138
0.4973 0.5067 —0.0040 570 0.16768 0.17087 —0.00135

0.5751 0.4283 —0.0034 oy Bs 0.20715 0.15429 —0.00123

750
Table 1(3.3.3)— Continued

Chromaticity Coordinates Wavelength Color-Matching Functions

r
r(d) g(r) b() (nm) 7(X) g(A) b(X)

0.6449 0.3579 0.24526 0.13610 —0.00108

0.2952 0.27989 0.11686 —0.00093
0.2402 0.30928 0.09754 —0.00079
0.1928 0.33184 0.07909 —0.00063

0.1537 0.34429 0.06246 —0.00049

0.1209 0.34756 0.04776 —0.00038
0.0949 0.33971 0.03557 —0.00030
0.0741 0.32265 0.02583 —0.00022
0.0580 0.29708 0.01828 —0.00015

0.0454 0.26348 0.01253 —0.00011

0.0354 0.22677 0.00833 —0.00008
0.0272 0.19233 0.00537 —0.00005
0.0205 0.15968 0.00334 —0.00003
0.0152 0.12905 0.00199 — 0.00002

0.0113 0.10167 0.00116 —0.00001

0.0083 0.07857 0.00066 —0.00001
0.0061 0.05932 0.00037 0.00000
0.0047 0.04366 0.00021 0.00000
0.0035 0.03149 0.00011 0.00000

0.0025 0.02294 0.00006 0.00000

0.0016 0.01687 0.00003 0.00000
0.0009 0.01187 0.00001 0.00000
0.0004 0.00819 0.00000 0.00000
0.0001 0.00572 0.00000 0.00000

0.0000 0.00410 0.00000 0.00000

0.0000 0.00291 0.00000 0.00000
0.0000 0.00210 0.00000 0.00000
0.0000 0.00148 0.00000 0.00000
0.0000 0.00105 0.00000 0.00000

0.0000 0.00074 0.00000 0.00000

0.0000 0.00052 0.00000 0.00000
0.0000 0.00036 0.00000 0.00000
0.0000 0.00025 0.00000 0.00000
0.0000 0.00017 0.00000 0.00000

0.0000 0.00012 0.00000 0.00000

0.0000 0.00008 0.00000 0.00000
0.0000 0.00006 0.00000 0.00000
0.0000 0.00004 0.00000 0.00000
0.0000 0.00003 0.00000 0.00000

0.0000 0.00001 0.00000 0.00000

0.0000 0.00000 0.00000 0.00000

751
 €6 Ivl'T €9 669°0 9¢ Iv8'0— 818 69b°0 978 L870 Ore 9PE'0— COLY 000°1Z
OT SII'l OS 88£'0 09 90S°0— 911 0€9°0 Zr6 8170 66h S87'0— 9'0LP OSz‘IZ
06 080'I II +170 10 S6z'0— 180 Z9L'0 656 OST‘0 - $66 LOZ’0— I's9r 00S‘IZ
8L €S0'l 68 1710 L9 SLI0— LES €98°0 988 660°0 656 €rl'0— 8'6Sb OSL‘IZ
86 €£0'l 79 790'0 Z9 960°0— rly 7r6'0 OOI LS0'0 €L0 880°0— SPSb 000‘7Z
rl LIOT 69 $z0'0 €8 Zr0'0— P77 L86'0 9€6 bZ0'0 OLS IP0'0— b'6rr 0SZ‘°@Z
00 000'I 00 000°0 00 000°0 000 000'I 000 000°0 000 000°0 r'bby 00S‘Zz
66 £86'0 99 LIOO— ~—-L9: €£0'0 ZL $960 Zee LIO'O— 60 ££0'0 9'6€b OSL‘ZZ
rr 6960 IL O€<00-— = LZ 1900 879 €88'0 966 LZ0'0— 8r8 S$S0'0 8'Peh 000‘€Z
rl $s6°0 ZI OPO'O— ~—Sss 86: F800 Z9Z OLL'O LSE ZEO'O— SES 890'0 l'0er OS7‘EZ
56 6£6'0 SS LyO0—- —-09- LOT0 90€ 8+9'0 €6L ZE0'0— ZIZ rL00 SStb 00S‘€Z
9L ¥76'0 €L 7S0'0— ~—_L6 LZI'0 0S6 P1S'0 Z9E 620'0— 19% IL0°0 Vt OSL‘EZ
6S 6060 8L SsOO— _— 6: 9I'0 ILZ PLe‘0 1S6 7Z0'0— PSI 090°0 L'9Ip 000‘rZ
16 768°0 Ol LS00—- ~—-6: Z9T0 Sb IS7'0 rr0 910°0— ILS S0'0 rcp OS7‘bZ
6b £880 70 8500- — Ss _PLI'0 Ish ZS1'0 S10 010'0— OZI 0£0'0 Z'80b 00S ‘rz
EI SL8'0 ZS 8S00— —-6£-E8T'0 Z OPO +80°0 0 029 S00°0— 9 119 L100 0'r0r OSL‘bZ
€Z 898'0 LL 8S00— —-S ‘0610 L SOS Z¥0'0 0 LL8 Z00°;0— € 8ZE 600'0 0°00r 000‘sZ
EI 798'0 LL 8S00—- 9 :961°0 S I€8 610°0 6 IS€ 100°;0— € €Z$ 700'0 0°96€ OSz‘sz
ZZ LS8°0 LL 8900—- $$ 1070 9L LLS 300°0 EI 88S 000°0— $8 910 700'0 Z'76E 00S‘Sz
S€ €S8°0 LL 8S00—- 7b S070 ZO LSP £00°0 LO 8£Z 000°0— 91 Z€8 000'0 €°88E OSL‘SZ
8S 0S3°0 LL 8S00— 61-8070 OS €0€ 100°0 90 060 000°0— SO 6I€ 000°0 9'P8E 000‘9Z
rr 8r8'0 LL 8s00—- ZE OIZ0 $50 79% 000°0 00 Z£0 000°0— IpS PIT 000'0 O'I8€ 0Sz‘9z
Z9 98'0 LL 8S00— — ST ZIZ‘0 ESS PST 000°0 8ZL O10 000°0— 8ZZL 8£0 000°0 y'LLE 00S‘9Z
61 St8°0 LL 8S0'0— _— 8S £170 Z 786 8h0 000'0 L S0b €00 000°0— 9 LLE Z10 000°0 SELE OSL‘9Z
90 rr8'0 LL 8S0'0- —iIL vIz'0 9 OSL P10 000'0 I LZ0 100 000°0— Z ZSL £00 000'0 v'OLE 000‘LZ
LI €b8°0 LL 8S00— 09 SIZ'0 €£ LEZ 700 000°0 €€ S6Z 000 000'0— 8h £80 100 000'0 O'L9E OS7‘LZ
SS Z8"0 LL 8S00-— = 7 -9IZ'0 SL P9T 100 000°0 SZ 180 000 000°0— 16 86Z 000 000°0 9°€9€ 00S‘LZ
€0 7Zb8'0 LL 8S00— —bL:9IT0 66Z LOE 000 000'0 Lyb 170 000 000°0— 00T 640 000 000'0 r'09€ OSL‘LZ
(x)°'q (y O18 (x)ol (v)°'9 (y )°'8 (v)Ote (wu) (,-l9)
——————————————— eee x u
sajeuIploog Ayonewoiy9 suonoun., SuIYyo}ePl-JO]OD YISUSTOACAA JOQUINUSARAA
(wu ppp =*V) ; WO NOS7Z = "Ve
q (uu 975 = Y) , WO NOD6T = 2m 38D SCHUM Z-Sp9 = *Y) , WD OOSST = %m Je Y "YNUINS AreuNg [way 0} Joadsay YIM UIa}sAg
ILOUILIOJOD prepurjys Arejuowatddng foG] ALD AY) JO Sayeurpsooy AyDOyeUIOID puke suoHoUny SUTIVI-10]OD (CE OUT AGL

IZ
 753
a
vO 000°0 91 700';O— TI 7O00'T ¢ 000 000 000°0 0 7720 000 000°0— € ZIZ O10 0000 £918 vara
LO 000°0 6¢ £00'0— cr £00'1 6 100 000 000°0 I 960 000 000':0— 9 0$9 LZ0 000°0 0°008 00s ‘71
60 000°0 69 700';0— 09 ¥00'l 8 900 000 000'0 7 Lvte 000 000°0— fc CSE PLO 0000 €'y8L OSL‘TZI
IT 000°0 9L $00';0— $9 SO0'T 7ZO 000 000°0 pvrl 100 000';0— O£L 661 0000 7 69L 000°€ I
£1 000°0 69 900°0— 9S 900'1 690 000 000°0 €vs £00 000°0— 691 £eS 0000 LVSL Osz'El
9¢ LOO';O— le LOOT IZ 000 000'0 IS 010 000:0— 80 6lv 1000 L‘OvL 00S "EI
SI 000°0
60 800°0— £6 LOOT 6S 000 000°0 TI O£0 000°0— 87 tSL £000 £ LCL OSL'€l
91 000°0
LI 000°0 gs 800°0— 8 800'1 Z9 100 000°0 LL 780 000':0— 61 €9L 600°0 evil 000‘rI
LI 000°0 9L 800°0— 6S 800'T c ¥00 000°0 cos ke, UO Us v CLL ¥70'0 8° 10L OS7'rl
LI 000°0 6S 800°0— cy 800'T 0 010 000°0 0 €1s 000;0— 6 O77 0900 L689 00S‘rI
Il 800°0— $6 LOO'| 770 000°0 911 100';0— 8eL 8tl0 0°8L9 OSL’
91 000°0
88 900°0— vL 900'°1 0r0 000°0 670 cO0';0— 796 967'0 L999 000°S I
I 000°0
6L v00';O0— OL vO0'l ps0 0000 . Lvl C0O0';0— 9SL SLS‘0 L’Ss9 OS7'SI
60 000°0
00 000°0 00 000° 000 000'0 000 000°0 000 000°! Cc SP9 00S‘SI
00 000°0
6r¢ 800°0 89 1660 797 000'0— LOV £100 798 99S" 6 vt£9 OSL'SI
LI 000°0—
89 8100 69 1860 $18 000°0— £69 1p0'0 9S1 1617 0'Sc9 000°91
Le 000°0—
c9-££00 vO L96°0 vrs 100°:0— ILp 7600 CLC LiL € ea OST7'9I
99 000°0—
€£°9S0'0 8¢€ brv6'0 £6S £00 0— LSO 810 PEC 790t 1909 00s‘9I
Ltaico. 0=
69 880°0 vO £160 066 S$00°;0— 698 90£°0 6c 6OSI'E 0'L6S OSL‘91
€L 1000—
9S 700';0— 92 TET0 oO€ IL8°0 678 800°0— OO cSPO 16c cOO'e T'88S 000°LI
OS 981°0 vi LI8°0 ten Lid 0= 710 €09°0 186 Ip9'c L’6LS OS7'LI
v9 £00°0—
9S vS70 Iv OSL'0 718 vloOOo— O8L 8SL°0 608 9ETT v'ILs 00S ‘LI
L6 v00 0—
Tceoee.0 St 0L9°0 StESCIO 0 S16 L880 CTE OLLI y'e9s OSL'LI
LS 900°0—
9L 670 bry 8Ls'0 bro 8100— 8£8 9160 £08 Piel 9°SSS 000°81
0z 800°0—
9S LES'‘0 cs ILv0 ye BING8fees 6te IkO'l L£9 ~060 6 LYS 0S7'81
80 600°0—
60 799°0 98 9Pt'0 Ive ploo— eve 1901 TO 9SS°0 SOs 00S‘81
$6 800°0—
8S 808°0 87 8610 vS8 800°0— LS6 CHO'l vSl SSz70 £ CEe OSL'8I
98 900°0—
00 000°0 00 000°I 00 000°0 000 000°0 000 000°! 000 000°0 €°97S 000°61
o€ LIO0O 86 EST $f 1L0.0— vL8 c10°0 Lt £t6'0 688 107 0— S6ls OS7'6I
18 9S0°0 O€ 89° tt $t9.0— LL8 670°0 £00 0€8°0 666 tte 0— 8 TIS 00S ‘6
8S pSlO 80 It6'l 99 S80 t— vs9 LSO0'0 ISZ O¢L'0 L76 vOv'O— £°90S OSL‘6I
67 O9T'T PL-cto [= Or8 S60°0 Le9 vI9'0 6rs 8th 0— 0'00S 000°0@
Sv TSE°0
80 199°0 9b 9NTC vs LO61— LzO ISTO 8IZ £1S0 68L Str 0— 8°£6r 0S7'07
IL €96°0 $8 9P8'l 9S O18 I— 090 L770 8eI stro L8S 97h 0— 8° L8v 00S ‘07
vO TOIT Ts 7801 SS tac I LLO ¢t¢°0 £CL L5e0 687 886 0— 6 18P OSL‘0Z
 Table 1(3.3.4) Relative Spectral Radiant Power Distributions of CIE Standard
Illuminants A and D,, from A = 300 to 830 nm at 1-nm Intervals (CIE 1971)

d A Dy; d A Ba
(nm) S(A) S(A) (nm) S(A) S(A)
300 0.930483 0.0341000 355 5.41070 45.7750
301 0.967643 0.360140 356 5.55213 45.9477
302 1.00597 0.686180 357 5.69622 46.1203
303 1.04549 1.01222 358 5.84298 46.2930
304 1.08623 1.33826 359 5.99244 46.4656
305 1.12821 1.66430 360 6.14462 46,6383
306 1.17147 1.99034 361 6.29955 47.1834
307 1.21602 2.31638 362 6.45724 47.7285
308 1.26188 2.64242 363 6.61774 48.2735
309 1.30910 2.96846 364 6.78105 48.8186
310 1.35769 3.29450 365 6.94720 49,3637
311 1.40768 4.98865 366 7.11621 49.9088
312 1.45910 6.68280 367 7.28811 50.4539
313 1.51198 8.37695 368 7.46292 50.9989
314 1.56633 10.0711 369 7.64066 51.5440
315 1.62219 11.7652 370 7.82135 52.0891
316 1.67959 13.4594 371 8.00501 51.8777
317 1.73855 15.1535 372 8.19167 51.6664
318 1.79910 16.8477 373 8.38134 51.4550
319 1.86127 18.5418 374 8.57404 51.2437
320 1.92508 20.2360 375 8.76980 51.0323
321 1.99057 21.9177 376 8.96864 50.8209
322 2.05776 23.5995 377 9.17056 50.6096
323 2.12667 25.2813 378 9.37561 50.3982
324 2.19734 26.9630 379 9.58378 50.1869
325 2.26980 28.6447 380 9.79510 49.9755
326 2.34406 30.3265 381 10.0096 50.4428
327 2.42017 32.0082 382 10.2273 50.9100
328 2.49814 33.6900 383 10.4481 51.3773
329 2.57801 35.3717 384 10.6722 51.8446
330 2.65981 37.0535 385 10.8996 52.3118
331 2.74355 37.3430 386 11.1302 52.7791
332 2.82928 37.6326 387 11.3640 53.2464
333 2.91701 37.9221 388 11.6012 53.7137
334 3.00678 38.2116 389 11.8416 54.1809
335 3.09861 38.5011 390 12.0853 54.6482
336 3.19253 38.7907 391 12.3324 57.4589
337 3.28857 39.0802 392 12,5828 60.2695
338 3.38676 39.3697 393 12.8366 63.0802
339 3.48712 39.6593 394 13.0938 65.8909
340 3.58968 39.9488 395 13.3543 68.7015
341 3.69447 40.4451 396 13.6182 71.5122
342 3.80152 40,9414 397 13.8855 74.3229
343 3.91085 41.4377 398 14.1563 77.1336
344 4.02250 41.9340 399 14.4304 79.9442
345 4.13648 42.4302 400 14.7080 82.7549
346 4.25282 42.9265 401 14.9891 83.6280
347 4.37156 43.4228 402 15.2736 84.5011
348 4.49272 43.9191 403 15,5616 85.3742 -
349 4.61631 44.4154 404 15.8530 86.2473
350 4.74238 44.9117 405 16.1480 87.1204
351 4.87095 45.0844 406 16.4464 87.9936
352 5.00204 45.2570 407 16.7484 88.8667
353 5.13568 45.4297 408 17.0538 89.7398
354 5.27189 45.6023 409 17.3628 90.6129

754
Table (3.3.4)— Continued

r A Ds A Ds
(nm) S(X) S(A) S(X) S(X)
410 17.6753 91.4860 40.3002 116.337
411 179913 91.6806 40.8076 116.041
412 18.3108 91.8752 41.3182 115.746
413 18.6339 92.0697 41.8320 115.451
414 18.9605 92.2643 42.349] 115.156

415 19.2907 92.4589 42.8693 114.861

416 19.6244 JUGS3) 43.3926 114.967
417 19.9617 92.8481 43.9192 115.073
418 20.3026 93.0426 44.4488 115.179
419 20.6470 93023512 44.9816 115.286

420 20.9950 93.4318 45.5174 115.392

421 21.3465 92.7568 46.0563 115.498
422 21.7016 92.0819 46.5983 115.604
423 22.0603 91.4069 47.1433 115.710
424 22.4225 90.7320 47.6913 115.817

425 22.7883 90.0570 48.2423 fe os:FA:

426 Phe) a i 89.3821 48.7963 11522
427 23.5307 88.7071 49.3533 114.500
428 23.9072 88.0322 49.9132 113.789
429 24.2873 8h3512 50.4760 113.078

430 24.6709 86.6823 51.0418 112.367

43] 25.0581 88.5006 51.6104 111.656
432 25.4489 90.3188 52.1818 110.944
433 25.8432 O2Ts 1 52.7561 110.233
434 26.2411 99.9953 533332 109.522

435 26.6425 95.7736 33.9 132 108.811

436 27.0475 979919 54.4958 108.865
437 27.4560 99.4101 55.0813 108.919
438 27.8681 101.228 55.6694 108.974
439 28.2836 103.047 56.2603 109.028

440 28.7027 104.865 56.8539 109.083

44] 29.4253 106.079 57.4501 109.137
442 29: 9579 107.293 58.0489 109.191
443 29.9811 108.508 58.6504 109.246
444 30.4142 109.722 59.2545 109.300

445 30.8508 110.936 59.8611 109.354

446 31.2909 [12461 60.4703 109.199
447 31.7345 113.365 61.0820 109.044
448 32.1815 114.579 61.6962 108.889
449 32.6320 115.793 62.3128 108.733

450 33.0859 117.008 62.9320 108.578

45] 33.5432 117.088 63.5535 108.423
452 34.0040 117.169 64.1775 108.268
453 34.4682 117.249 64.8038 108.112
454 34.9358 1197329 65.4325 107.957

455 35.4068 117.410 66.0635 107.802

456 35.8811 117.490 66.6968 107.501
457 36.3588 Lhy71 67.3324 107.199
458 36.8399 117.651 67.9702 106.898
459 37.3243 LiFsis2 68.6102 106.597

460 37.8121 117.812 69.2525 106.296

461 38.3031 117.517 69.8969 105.995
462 38.7975 117.222 70.5435 105.693
463 39.2951 116.927 TR19Z2 105.392
464 39.7960 116.632 71.8430 105.091

pe
 Table 1(3.3.4)— Continued

r A Ds A D¢5
(nm) S(A) S(A) S(A) S(A)

520 72.4959 104.790 110.803 96.0611

521 73.1508 105.080 TL1529 96.0065
$22 73.8077 105.370 112255 95.9519
523 74.4666 105.660 112.982 95.8972
524 Toae7 Ss 105.950 113.709 95.8426
ae) 75.7903 106.240 114.436 95.7880
526 76.4551 106.530 115.164 95.0778
527 TIA2T 106.820 115.893 94.3675
528 77.7902 107.110 116.622 93.6573
329 78.4605 107.400 1h 92.9470
530 79.1326 107.689 118.080 92.2368
531 79.8065 107.361 118.810 91.5266
532 80.4821 107.033 119.540 90.8163
533 81.1595 106.704 120.270 90.1061
534 81.8386 106.376 121.001 89.3958
535 82.5193 106.047 1203731 88.6856
536 83.2017 105.719 122.462 88.8177
537 83.8856 105.391 123.193 88.9497
538 84.5712 105.062 123.924 89.0818
539 85.2584 104.734 124.655 89.2138
540 85.9470 104.405 125.386 89.3459
541 86.6372 104.370 126.118 89.4780
542 87.3288 104.334 126.849 89.6100
543 88.0219 104,298 127.580 89.7421
544 88.7165 104.262 128.312 89.874]
545 89.4124 104.226 129.043 90.0062
546 90.1097 104.190 129.774 89.9655
547 90.8083 104.154 130.505 89.9248
548 91.5082 104.118 131.236 89.884]
549 92.2095 104.082 131.966 89.8434
550 92.9120 104.046 132.697 89.8026
551 93.6157 103.642 133.427 89.7619
552 94.3206 103237 134.157 89.72)2
S53 95.0267 102.832 134.887 89.6805
554 95-1339 102.428 135.617 89.6398
555 96.4423 102.023 136.346 89.599]
556 97.1518 101.618 137.075 89.4091
557 97.8623 101.214 137.804 89.2190
558 98.5739 100.809 138.532 89.0290
559 99.2864 100.405 139.260 88.8389
560 100.000 100.000 139.988 88.6489
561 100.715 99.6334 140.715 88.4589
562 101.430 99.2668 141.441 88.2688
563 102.146 98.9003 142.167 88.0788
564 102.864 98.5337 142.893 87.8887
565 103.582 98.1671 143.618 87.6987
566 104.301 97.8005 144.343 87.2577
567 105.020 97.4339 145.067 86.8167
568 105.741 97.0674 145.790 86.3757
569 106.462 96.7008 146.513 85.9347
570 107.184 96.3342 147.235 85.4936
571 107.906 96.2796 147.957 85.0526
572 108.630 96.2250 148.678 84.6116
a3 109.354 96.1703 149.398 84.1706
574 110.078 96.1157 150.117 83.7296
756
Table I(3.3.4)— Continued

rv A D5 A D¢5
(nm) S(X) S(A) S(A) S(A)
630 150.836 83.2886 188.701 74.0027
631 151.554 8353297 189.350 73.1465
632 S227) 83.3707 189.998 72.2902
633 152.988 83.4118 190.644 71.4339
634 153.704 83.4528 191.288 70.5776
635 154.418 83.4939 19R931 69.7213
636 155.132 83.5350 192.572 69.9101
637 155.845 83.5760 193,211 70.0989
638 156.558 83.6171 193.849 70.2876
639 157.269 83.6581 194.484 70.4764
640 1STIT9 83.6992 195.118 70.6652
641 158.689 83.3320 195.750 70.8540
642 159.39), 82.9647 196.381 71.0428
643 160.104 S200 197.009 MZ S15
644 160.811 82.2302 197.636 71.4203
645 161.516 81.8630 198.261 71.6091
646 162.221 81.4958 198.884 71.8831
647 162.924 81.1285 199.506 I2A571
648 163.626 80.7613 200.125 72.4311
649 164.327 80.3940 200.743 T2N0S |
650 165.028 80.0268 ZON 359 72.9790
651 165.726 80.0456 201.972 73.2530
652 166.424 80.0644 202.584 210
653 16%121 80.0831 203.195 73.8010
654 167.816 80.1019 203.803 74.0750
655 168.510 80.1207 204.409 74.3490
656 169.203 80.1395 205.013 73.0745
657 169.895 80.1583 205.616 71.8000
658 170.586 80.1770 206.216 70.5255
659 Led) 80.1958 206.815 69.2510
660 171.963 80.2146 207.411 67.9765
661 172.650 80.4209 208.006 66.7020
662 173,335 80.6272 208.599 65.4275
663 174.019 80.8336 209.189 64.1530
664 174.702 81.0399 209.778 62.8785
665 175.383 81.2462 210.365 61.6040
666 176.063 81.4525 210.949 62.4322
667 176.741 81.6588 21) 532 63.2603
668 177.419 81.8652 21D ee 64.0885
669 178.094 82.0715 212.691 64.9166

670 178.769 82.2778 213.268 65.7448

671 179.44] 81.8784 213.842 66.5730
672 180.113 81.479] 214.415 67.4011
673 180.783 81.0797 214.985 68.2293
674 181.451 80.6804 Oneee 69.0574

675 182.118 80.2810 216.120 69.8856

676 182.783 79.8816 216.684 70.4057
677 183.447 79.4823 217.246 FV P|
678 184.109 79.0829 217.806 71.4460
679 184.770 78.6836 218.364 71.9662

680 185.429 78.2842 218.920 72.4863

681 186.087 77.4279 219.473 73.0064
682 186.743 76.5716 220.025 73.5266
683 187.397 72.9153 220.574 74.0467
684 188.050 74.8590 eat i122 74.5669

757
 Table I(3.3.4)— Continued

r A D5 r A D5
(nm) S(X) S(A) (nm) S(A) S(A)
740 221.667 75.0870 785 243.924 63.8434
741 222.210 BATS 786 244.367 63.9355
742 222.751 72.7881 787 244.808 64.0276
743 223.290 71.6387 788 245.246 64.1198
744 223.826 70.4893 789 245.682 64.2119
745 224.361 69.3398 790 246.116 64.3040
746 224.893 68.1904 79] 246.548 63.8188
747 225.423 67.0410 t92 246.977 63.3336
748 220951 65.8916 793 247.404 62.8484
749 226.477 64.7421 794 247.829 62.3632
750 227.000 63.5927 795 248.251 61.8779
751 POM a9! 61.8752 796 248.671 61.3927
752 228.041 60.1578 797 249.089 60.9075
753 228.558 58.4404 798 249.505 60.4223
754 229.073 56.7229 799 249.918 59.9371
755 229.585 55.0054 800 250.329 59.4519
756 230.096 52.2880 801 250.738 58.7026
Le 230.604 51.5705 802 251.144 57.9533
758 2atallO 49.8531 803 251.548 57.2040
759 231.614 48.1356 804 251.950 56.4547
760 Pe 9S Me 46.4182 805 252.350 55.7054
761 232.615 48.4569 806 252.747 54.9562
762 piase V2 50.4956 807 253.142 54.2069
763 233.606 52.5344 808 253535 53.4576
764 234.099 54.5731 809 2ga0925 52.7083
765 234.589 56.6118 810 254.314 51.9590
766 235.078 58.6505 811 254.700 52.5072
767 235.564 60.6892 812 255.083 53.0553
768 236.047 62.7280 813 255.465 53.6035
769 236.5929 64.7667 814 255.844 54.1516
770 237.008 66.8054 815 256.221 54.6998
771 237.485 66.4631 816 256.595 55.2480
772 2at9S9 66.1209 817 256.968 55.7961
773 238.432 65.7786 818 257.338 56.3443
774 238.902 65.4364 819 257.706 56.8924
775 239.370 65.0941 820 258.071 57.4406
776 239.836 64.7518 82] 258.434 DF1248
777 240.299 64.4096 822 258.795 58.0150
778 240.760 64.0673 823 259.154 58.3022
419 241.219 63.7251 824 259.511 58.5894
780 241.675 63.3828 825 259.865 58.8765
781 242.130 63.4749 826 260.217 59.1637
782 242.582 63.5670 827 260.567 59.4509
783 243.031 63.6592 828 260.914 59.7381
784 243.479 63.7513 829 261.259 60.0253
830 261.602 60.3125

758
Table II(3.3.4) Abridged Values of Relative Spectral Radiant Power Distributions of CIE
Standard Illuminants A, B, C, and D,. from A = 300 to 830 nm at 5-nm Intervals (CIE, 1971)

A A B Cc Dgs r A B ¢ Ds
(nm) S(A) S(A) S(A) S(A) (nm) S(A) S(A) S(A) S(A)
300 0.93 0.03 ma 110.80 101.90 100.15 96.1
305 1.13 580 114.44 101.00 97.80 95.8
310 1.36 3.3 585 118.08 100.07 95.43 y
FP
315 1.62 11.8 590 121.73 99.20 93.20 88.7
320 Bee 0.02 0.01 20.2 595 123.59 98.44 O22 89.3
325 2und 0.26 0.20 28.6 600 129.04 98.00 89.70 90.0
330 2.66 0.50 0.40 3ia 605 132.70 98.08 88.83 89.8
aos 3.10 1.45 153 38.5 610 136.35 98.50 88.40 89.6
340 B49 2.40 aa 39.9 615 139.99 99.06 88.19 88.6
345 4.14 4.00 4.85 42.4 620 143.62 99.70 88.10 87.7
350 4.74 5.60 7.00 44.9 625 147.24 100.36 88.06 85.5
35> 5.41 7.60 9:95 45.8 630 150.84 101.00 88.00 83.3
360 6.14 9.60 12.90 46.6 635 154.42 101.56 87.86 83.5
365 6.95 12.40 17.20 49.4 640 157.98 102.20 87.80 83.7
370 T82 15.20 21.40 S21 645 161.52 103.05 87.99 81.9

FS 8.77 18.80 27.50 51.0 650 165.03 103.90 88.20 80.0

380 9.80 22.40 33.00 50.0 655 168.51 104.59 88.20 80.1
385 10.90 26.85 39.92 323 660 171.96 105.00 87.90 80.2
390 12.09 31.30 47.40 54.6 665 175.38 105.08 87.22 81.2
395 13.35 36.18 es 68.7 670 178.77 104.90 86.30 82.3

400 14.71 41.30 63.30 82.8 675 182.12 104.55 85.30 80.3
405 16.15 46.62 71.81 87.1 680 185.43 103.90 84.00 Ohare
410 17.68 52.10 80.60 gies 685 188.70 102.84 82.21 74.0
415 19.29 57.70 89.53 os 690 191.93 101.60 80.20 eh
420 20.99 63.20 98.10 93.4 695 1952 100.38 78.24 70.7

425 22.19 68.37 105.80 90.1 700 198.26 99.10 76.30 71.6
430 24.67 73.10 112.40 86.7 705 201.36 97.70 74.36 73.0
435 26.64 77.31 tas 95.8 710 204.41 96.20 72.40 74.3
440 28.70 80.80 121.50 104.9 715 207.41 94.60 70.40 68.0
445 30.85 83.44 123.45 110.9 720 210.36 92.90 68.30 61.6

450 23:09 85.40 124.00 117.0 725 213 29 91.10 66.30 65.7
455 35.41 86.88 123.60 117.4 730 216.12 89.40 64.40 69.9
460 37.81 88.30 123.10 117.8 735 218.92 88.00 62.80 25
465 40.30 90.08 123.30 116.3 740 221.69 86.90 61.50 nk
470 42.87 92.00 123.80 114.9 745 224.36 85.90 60.20 69.3

475 45.52 RES 124.09 115.4 750 227.00 85.20 59.20 63.6
480 48.24 95.20 123.90 H59 DS papa Bess 84.80 58.50 55.0
485 51.04 96.23 122,92 112.4 760 232.V2 84.70 58.10 46.4
490 53.91 96.50 120.70 108.8 765 234.59 84.90 58.00 56.6
495 56.85 95.71 116.90 109.1 770 237.01 85.40 58.20 66.8

500 59.86 94.20 112.10 109.4 775 PeteHs 65.1

505 62.93 P2554 106.98 108.6 780 241.68 63.4
510 66.06 90.70 102.30 107.8 785 243.92 63.8
515 69.25 89.65 98.81 106.3 790 246.12 64.3
520 72.50 89.50 96.90 104.8 files 248.25 61.9

525 Ep be 90.43 96.78 106.2 800 250.33 is

530 79.13 92.20 98.00 107.7 805 2o20a3 eee
535 82.52 94.46 99.94 106.0 810 254.31 52.0
540 85.95 96.90 102.10 104.4 815 256.22 54.7
545 89.41 99.16 103.95 104.2 820 258.07 57.4

550 92.91 101.00 105.20 104.0 825 259.86 58.9

555 96.44 102.20 105.67 102.0 830 261.60 60.3
560 100.00 102.80 105.30 100.0
565 103.58 102.92 104.11 98.2
570 107.18 102.60 102.30 96.3

7159
Table III(3.3.4) Relative Spectral Radiant Power Distributions of Daylight
Illuminants D5, (Correlated Color Temperature 5503 K) and D-,, (Correlated
Color Temperature 7504 K) Calculated in Accordance with CIE Method Given
in Text (CIE, 1971)

A Ds; D;;
(nm) S(A) S(A)
300 0.02 0.04
310 2u 5.1
320 11.2 29.8
330 20.6 54.9
340 23.9 2 Be:
350 27.8 62.7
360 30.6 63.0
370 34.3 70.3
380 32.6 66.7
390 38.1 70.0
400 61.0 101.9
410 68.6 111.9
420 71.6 112.8
430 67.9 103.1
440 85.6 121.2
450 98.0 133.0
460 100.5 132.4
470 o79 127.3
480 102.7 126.8
490 98.1 117.8
500 100.7 116.6
510 100.7 113.7
520 100.0 108.7
530 104.2 110.4
540 102.1 106.3
550 103.0 104.9
560 100.0 100.0
570 97.2 95.6
580 97.7 94.2
590 91.4 87.0
600 94.4 87.2
610 95.1 86.1
620 94.2 83.6
630 90.4 78.7
640 vfht 78.4
650 88.9 74.8
660 90.3 74.3
670 93.9 75.4
680 90.0 71.6
690 ti 63.9
700 82.8 65.1
710 84.8 68.1
720 70.2 56.4
730 79.3 64.2
740 85.0 69.2
750 71.9 58.6
760 52.8 42.6
770 75.9 61.4
780 71.8 58.3
790 72.9 59.1
800 67.3 54.7
810 58.7 47.9
820 65.0 52.9
830 68.3 55.5
Table IV(3.3.4) CIE 1931 (xp, yp)-Chromaticity Coordinates and Scalar Multipliers M,,M,
of Daylight of Different Correlated Color Temperatures TJ. (Planck’s Constant c,= 1.4388x
10 * m-K); (CIE, 1971)
T. Xp YD M, M,

4,000 0.3823 0.3838 ==1.005 2.827

4,100 0.3779 0.3812 — 1.464 2.460
4,200 0.3737 0.3786 seh ee! D127
4,300 0.3697 0.3760 —" 318 1.825
4,400 0.3658 0.3734 = 7333 1.550
4,500 0.3621 0.3709 — 1.286 1.302
4,600 0.3585 0.3684 — 1.238 1.076
4,700 0.3551 0.3659 = hl 9G 0.871
4,800 0.3519 0.3634 — 1.140 0.686
4,900 0.3487 0.3610 — 1.090 0.518

5,000 0.3457 0.3587 — 1.040 0.367

5,100 0.3429 0.3564 — 0.989 0.230
5,200 0.3401 0.3541 — (0.939 0.106
5,300 0.3375 0.3519 — 0.888 — 0.005
5,400 0.3349 0.3497 = 0:537 —0.105
5,500 0.3325 0.3476 — 0.786 ==0.195
5,600 0.3302 0.3455 — 0.736 — 0.276
5,700 0.3279 0.3435 — 0.685 — 0.348
5,800 0.3258 0.3416 — 0.635 —0.412
5,900 0.3237 0.3397 — 0.586 — 0.469

6,000 0.3217 0.3378 — 0.536 =U 19

6,100 0.3198 0.3360 — 0.487 — 0.563
6,200 0.3179 0.3342 — 0.439 — 0.602
6,300 0.3161 0.3325 —0.391 — 0.635
6,400 0.3144 0.3308 — 0.343 — 0.664
6,500 0.3128 0.3292 — 0.296 — 0.688
6,600 aki? 0.3276 =). 250 — 0.709
6,700 0.3097 0.3260 —0.204 — 0.726
6,800 0.3082 0.3245 =();159 — 0.739
6,900 0.3067 0.3231 —0.114 — 0.749

7,000 0.3054 0.3216 — 0.070 =O IS)

7,100 0.3040 0.3202 — 0.026 — 0.762
7,200 0.3027 0.3189 0.017 0.765
7,300 0.3015 0.3176 0.060 — 0.765
7,400 0.3003 0.3163 0.102 — 0.763
7,500 0.2991 0.3150 0.144 — 0.760
7,600 0.2980 0.3138 0.184 =O J50
7,700 0.2969 0.3126 0.225 — 0.748
7,800 0.2958 0.3115 0.264 — 0.740
7,900 0.2948 0.3103 0.303 =().730

8,000 0.2938 0.3092 0.342 —=() 720

8,100 0.2928 0.3081 0.380 — 0.708
8,200 0.2919 0.3071 0.417 — 0.695
8,300 0.2910 0.3061 0.454 — 0.682
8,400 0.2901 0.3051 0.490 — 0.667
8,500 0.2892 0.3041 0.526 — 0.652

9,000 0.2853 0.2996 0.697 — 0.566

9,500 0.2818 0.2956 0.856 — 0.471

10,000 0.2788 0.2920 1.003 — 0.369

11,000 0.2737 0.2858 1.266 — 0.160

12,000 0.2697 0.2808 1.495 0.045

13,000 0.2664 0.2767 1.693 0.239

14,000 0.2637 0.2732 1.868 0.419

15,000 0.2614 0.2702 2.021 0.586

20,000 0.2539 0.2603 2.571 1.231

 Table V(3.3.4) Mean Relative Spectral Radiant Power Distribution S,(A)
and First Two Eigenvectors $,(A) and S,(A) Used in CIE Method of
Calculating Daylight Muminants (CIE, 1971)

r
(nm) So(A) S\(X) S,(X)
300 0.04 0.02 0.0
310 6.0 4.5 2.0
320 29.6 22.4 4.0
330 33 42.0 8.5
340 ay BS) 40.6 7.8
350 61.8 41.6 6.7
360 61.5 38.0 he:
370 68.8 42.4 6.1
380 63.4 38.5 3.0
390 65.8 35.0 |e
400 94.8 43.4 11
410 104.8 46.3 a0)
420 105.9 43.9 wal
430 96.8 otil Set
440 113.9 36.7 SS
450 125.6 ce aw
460 1255 32.6 ere:
470 121.3 204 210
480 213 24.3 == nO
490 Ge lee 20.1 cous
500 i331 16.2 elMe:
510 110.8 1322 met ee
520 106.5 8.6 ene:
530 108.8 6.1 EP
540 105.3 4.2 ate
550 104.4 14 =0.3
560 100.0 0.0 0.0
570 96.0 1G 0.2
580 95.4 ot ioe 0.5
590 89.1 ape Be ne
600 90.5 fe ea 3.2
610 90.3 a er 4.1
620 88.4 mae 4.7
630 84.0 eke et
640 85.1 7109 6.7
650 81.9 107 a
660 82.6 —12.0 8.6
670 84.9 — 14.0 9.8
680 81.3 =) ES 10.2
690 TS 1 PAW 8.3
700 74.3 13.3 9.6
710 76.4 129 8.5
720 63.3 = 10.6 7.0
730 Ahad mee: 7.6
740 77.0 meee 8.0
750 G2 = 102 6.7
760 47.7 = ope
770 68.6 wae 7.4
780 65.0 — 10.4 6.8
790 66.0 — 10.6 7.0
800 61.0 ae dsl 6.4
810 33:3 bed 3:2
820 58.9 ee 6.1
830 We, 95 6.5
762 eS
MERRIE
DE ES wees
 =
JO

JO
AA
Sq

ALD

PAL
yor

FIR

Aq
Saleg

24)
AID
J9puyy

p96]
YoY
AID)
(S"¢"¢)]
S10JIVY

suIV
[eIedG
Wed
| ted
c ed
¢

Soydureg
ted
py 1d
¢

(0861
Paepurys

DdULIPLY

pomoarA
PPIAOIg
juBUTUINT]T

prepueyg
AA YISUITIAL ————

49A198qO
IAOWILJIIA]
(wu)y Og Sig 0) —- Qed.
— 9 Cy)
— Cy) ug
-- ig x Fg $00 Adhd

yusdsasonJUON

Arejuouatddng
OOP 80'°0 6700 8800 vv0'0 vv0'0 6c0'0 0 867 0 L6E 901°0 EL1°0
SOV S700 8700 7800 9¢60°0 cv0'0 8700 0 86C 0 L6E 901°0 SLI0
OLY €v0°0 L700 L800 €90°0 0v0'0 8700 00€'0 0 L6t 6010 LLI0
SIv 1700 9700 6800 €L0°0 8£0°0 L700 S0e'0 0 L6E oO vil 0810
OCP 0v0'0 9c0'0 6800 080°0 LE00 L700 I1€°0 96¢°0 Oc10 c81°0
S¢cV 8£00 9c0'°0 8800 880°0 9€0°0 9700 0 81E S6£°0 L710 €81°0
OtP LE00 9700 $800 6800 S£0°0 9700 0 9CE 0 COE 0 GEL €81°0
Sty 9€0°0 9700 7800 8800 ve00 9700 0 see 0 L8¢€ 9710 c81°0
Ovy $00 9700 8400 €80°0 €€0°0 L700 0 9bf 0 6Lt 9S10 0 6L1
SrP 0 veo L700 0 vL0 1800 c£0'0 8700 0 LSE CLEO 9910 0 SL1
OSP ve0'0 6c0'0 0L0'°0 9L0°0 c£0'0 0 8c0 69€'0 Eat SLs) 0 ILI
SSv S£0°0 0£0°0 990°0 1L0°0 c£0'0 6200 18¢°0 195.0 v810 L910
097 9£0°0 €£0°0 €90°0 990°0 0 c£0 100 16€0 goe0 1610 C910
4 Lt00 LE0'0 090°0 6500 €£0'0 ve00 0 86£ 0 8re S610 LS10
OLV 0 6£0 vv00 LS0°0 cS0'0 ve0'0 6£0°0 10v'0 0 Ore L610 ESO
SLY 170°0 vS0'0 vS0'0 8700 9€0°0 Lv0'0 00r'0 Ageal S610 6v1°0
O8P Sv0'0 890°0 cS0'0 S700 8€0°0 LS0'0 0 96£ 0 ocr 1610 9710
S8P 1S0°0 $800 0500 €r0'0 1700 L900 L8€0 6l€0 E810 0 Eri
06? 8S0°0 volo 8700 0v0'0 Sv0'0 LLO'0 9LE°0 O cle bL10 Irl‘O
S6v L90°0 1710 9700 LE00 6700 980°0 £9€°0 SOe'0 S910 6€1°0
00S LLO'0 9€1°0 vv0'0 ve0'0 $S0'0 7600 0 8re 0 667¢ 0 Sol 8e1°0
SOS 6800 9r1°0 cv0'0 ¢£0°0 790°0 $600 0 [ee 0 C67 °0 971 erg
OIS COLO 0 esl 1700 €£0°0 0L0°0 5600 0 Ele 0 987 0 LEl 9€1°0
SI¢ S110 9510 6£0°0 £t00 8L0°0 £600 0 L6c 180 0 6c1 pelo
OcS LEO LS10 8£0°0 ve0'0 980°0 0600 E870 9L70 (Gat [€l0
GS 6£1°0 6S10 LE00 ve0'0 7600 880°0 0 CLT OLT'0 S110 Lé10
OES ISt‘0 0 rot 9€0°0 ve0'0 L60°0 L800 C9C'0 $970 0 O11 0 vel
Ses COLO p90 0 $£0 po0'0 1010 980°0 ISc0 0970 LOI0 E710
Ovs 7 L910 ve0'0 ve0'0 volo $80°0 Ivc0 LSTO SOLO e710

763
Svs S810 CLIO €£00 0 €t0 901°0 £800 O0€TO 0 €Sc SOI0 0 €cl
 F142
4d
| Hed
c ed
¢ ed
p Ae
¢
AA, YISUITIAL a

—(S°¢€)I
panunuod)
(wu)y Crallg Cees Wile. Coto (we (wid -(yilts (mes (Yee eo

764
OSS 0 861 Cry c£00 €£00 LOI‘O £800 O OCT IScO SOl0 ver
CoS E170 ps0 1€00 0 c£0 LOI‘0 €80°0 C170 0 67 0 SOL 0 OCI
09S O O€cT £610 1€00 100 901°0 0 80 0 807 0 87c SOLO LILO
g9¢ IScO 0 907 1€0°0 0 c£0 0 pol 8800 0 LOC 0 977 SOLO 0 ell
OLS 0 9L7 0 €CC 1€0°0 9€0°0 0 LOL 1600 0 807 0 €vc LOL‘0 LILO

CLS 0 SOE 0 977 0 c£0 1v0'0 0 660 7600 0 807 0 8€c7 LITO e110
O8S¢ 0 9€£ CLTO €£00 0 ¢r0 960°0 960°0 0 807 O vee 0 OCI 6110
C8s 0 69 0 60€ ¢£00 0 60 0 £60 8600 0 907 0 Lec O Sel 0 LEl
06S 0 LOv 0 Lve L£00 0 vS0 060°0 0 OO1 0 pOcT 0 6cC 9510 6v1°0
66S 0 Lev 0 68£ 1700 0 €90 0 680 COLO 0 COC O 6CC e810 0 G&I.

009 0 6S 0 voV 9700 0L00 0 680 'O vO 0 £07 0 O£T VITO 8610
S09 0 C8 0 O8r £500 vL0O0 0 680 SOLO 0 60¢ O CET 0 OST 0 977
O19 10S'0 0 €cS 190°0 0800 0600 0 vol 0 Occ VETO 0 8c 0 9SC
C19 9190 0 09° 1LO'0 0 c80 1600 0 SOI 0 9€7 Gee\) Q PLE 0 8c
079 0 8cS 0 £69 0 780 800 0 60 0 SOL 0 997 O 8CC 0 eee 0 vie

Sco Les‘0 6190 ¢60°0 S800 600 0 vor Si ig 0 VCC 0 Ore 0 6££
O¢9 0 prs 0 179 °0 601 0 80 0 60 0 vol 0 86C 0 OCT 0 che 0 €9¢
C9 0 ISS LS90 0 Icl €80°0 £600 i) 0) O LI€ 8170 0 Ive 0 vse
Ov9 0 LSS 0 699 0 €£1 800 0 960 901°0 Loe 0 1cc 0 Sve 0 COV
co 0 €95 1890 0 Srl 0 C80 0 LOL 0 801 19€'0 0 9CCT 0 CIE 0 617

0OS9 L9S‘0 £69°0 0 9S1 €80°0 601°0 LITO 0 16¢ O LET 16¢'0 0 perv
¢o9 0 €LS 90L'0 991°0 7800 0 Oc 9110 Ocr'0 vSTO 0 per 0 8rr
099 6LS°0 0 61L LLIO 9800 pelo 0 €cl 0 9Lv 8L70 0 L8v 0 cor
699 C890 0 €€L 8810 0 980 0 pS O CEL 0 1S 0 60€ 0 Lvs 0 6Ltv
OL9 0 c6S 0 Lvl 0 10c 980°0 LLIO 0 €vl 0 686 0 Lv¢e °0 609 0 L6v

CLO 0 86S COLO LITO 6800 0 COT O LST Lv9'0 0 06£ L990 0 LIS
089 $090 8LL‘0 0 9€T 7300 0 8c7 Tor COLO 9€V7'0 eu tcl 0 8eo
689 €190 0 €6L O LSC 0 v30 O CST 0 161 0 6vl 0 v8v 0 99L 0 656
069 1790 L080 0 6L7 7300 SL70 LLGO L8L0 0 ves €08°0 0 6LS
669 0 6c9 6180 0 COE 6800 0 967 O CET 9180 0 08S O€8'0 0 86S

OOL Le9'0 0 O£8 0 9CE 980°0 0 91¢€ ¢Sc0 Ses8'0 0 vc9 6780 919°0

|, asay eiepase pasn

UI oy) AYD poyJoUl
10} BuIssasse
ay} Aytyenb
Jo oy} a[qistayredJo oy) wunsyoads
Jo yYSITAep SsIOVeNUMIS
JOJ

ee
Re

“Arj_
ee ee wa
00 “TOL'O 1S00™~0 Ocl 0" LIS STUO” T8007 “cee "0000 |SC9U- OCy Sli UCP 0" Sar0

765
9200 80) Ue eUeeseU
Cr Ltro €c0O0 9t00 POL) c00°0 |IISO SIvp I[tLO SIP 8hL0
ore Uo COU et. te 8200 U ccOO0 £890 900°0 j;cOvVO Olp CLIO OIP 60L0
9670
6850
9190
8900
I100
OLCcO
600°0
OLS0
tIcO
1100
1990
SOvp
SOP
volo
cOSO
JIITO
8790
8£9°0
v6l0
0670
8000
LtcO
[100
0Ob
OOF
Svc 0 velo 9100 C6E
0870'0
p9l0
Svc
O6£
0961°0
C8E
vrveT
9¢50'0
SET 0 0770 880'0 O8E
8170 877 0 LII0 CLE
1610 677 0 Srl 0 OLE
CSI O LTT 0 LLI0 G9€
IL1‘0 8170 vOT 0 O09¢€
780'0 voT 0 9170 CSE
S00 vLI0 VTT 0 OSE
$70 0 LEO OFT 0 Cre
700'0 COLO CET O OVE
100°0 SL0'0 (Gxal Cee
100°0 0S0'°0 OFT 0 O€E
000°0 €70'0 977 0 STE
900°0 0770 OTE
100°0 vI7TO Cle
100°0 S07 0 OLE
000'0 v61°0 COE
cs10 OOE

JUDBISIION[JUO
mee) INO tN CY) (a )O-LV ge CY)'O"g WY) Gy9 CVrgoge ay) vito Cu)
¢
adures Z
adures |
oJdures Xa t ¢ I X
jusdsar0N]
sojdures J soydures

NT
pd juvUIUNT] pxepuejys AID Aq pajerpeuy voy
S1OJIVY DUIPLY [v.IIdG [eoQUapP] JO sofdurvs jo eg & SULIOY SajduIeG jUddSaI0N],J BY) Jo / aquIsyA, SuIpUodsa.LIOZ
ay) YIM pourquio,y saydureg juadsaionyUON ay) Jo f saquiay, You ‘sajdureg yuassasonyy aemy, Jo (Y)T “(Y)"
‘CYYO saunuend jeysedg pur soydureg juadsas0NJUON saNJ], JO (Y)‘s si0}svy aouLIpey jededg (C'¢'E)IT AGL
 F198
yuUsdSaIONTJUON,

766
so,dures sojdures
judosaIoNyy

panuijuod
—(S'ee)IT
o[dures
| ajdures
Z ajdures
¢
X I é Ss iNaeN
(ura) Cy)ig (y)*g= (yg?
I Cru) CAO O)"d (i) CFO OOM RaO FeO) Le
Scr e830 8780 |SSLO Sty vel OLO0 1100 8IZL0 OL00 <80:0 t090 0200
OEP Ss80 8980 18980 Oth crLod S800 L000 0&L0
- S800 9700 9290 S800
Str 0980 8280 {8680 Sty OSL0 cw600 po00 0 FPL #600 01718000 cOUOTECIZ
Obr SS80 9180 |6680 Ovy 9620 0600 1000 SSZL°0 0600 6100 ItL0 0600
Shb 980 9980 {8880 Shy 1920 1800 000°0 7920 1800 €100 tvrLO 1800
OST 980 CS80 |cL80 OSb 99L'°0 ILOO 9920 1200 6000 0 CESL 100"
SSP cE80 9P80 {6980 Sop OLL0 p900 OLL0 7900 S000 v9L0 p900
O09PF 6780 I”80 19980 09p PLl0 9900 vLL0 9800 1000 BcLL.0 PSO;07
SOP Sc80 980 |6S80 Sov 8LLl0 8700 8LL0 8r00 0000 8LL0 8V00
OLP cc80 I€80 }0S80 OLb c8L0 600 c8Ll0 600 0 HCSL OLQWE
Slr Ic80 8c80 |tv80 SLb 8820 €t00 8820 €t00 8820 €t00
O08PF C780 8780 |IP80 08h v6L'0 8700 v6L'0 8700 0 v6L
= 8700
S&P CC80 Le8O0 |8e80 8p 66L0 ccOO0 66L'0
8 ccOO0 66L0
8 ccO'0
067 E780 L780 |9€80 06p 080 8100 080 8100 080 8100
SéP EC380 9C80 |£e80 S6vp 8080 vI00 8080 vl00 8080 vI00
ODS Pve80 L780 |€€80 OOS cI80 1100 Goa8
a (ee 0 cls HOG
SOS Le80 6c80 |ve80 SOS LI80 6000 LI80 6000 LI80 6000
-OlS 0. OME CESOW WLES OLS 0 ccs 8000 0 cs 8000 C780 8000
SIS sces0Or SES WS8ESOU SIS L780 9000 L780 9000 L780 9000
OCS St80 9€80 |6€80 OcS 0€8°0 S000 O0€8°0 S000 O0€8°0 S000
S7s SEBO 9E8O |LE8O StS £80 pooo 180 poood 180 yoo'o0
OES SE8O0 980 |LE8O OfS £80 poood 180 yooo I1£80 yoood
SEs PvEsO vesO |St80 StS I1€80 C000 [£80 C000 I£80 C000
Obs St80 St80 |9S80 Ops ced £000 I1€80 €00';0 180 c000
SPs SE8O SEBO |9E8O SPS 0 cts €00';0 ces0 €00';0 ces'0 £000
OSS PvE8O SE8O0 |St80 OSS €£€80 1000 £€80 1000 €€80 1000
sss peso S80 1St80 goS £€€8°0 1000 €£€8°0 1000 €£80 1000
09S St80 9€80 |9€80 O9¢ 0 vets 1000 £80 1000 v£8'0 1000
os6 Ula sre ee OER GOS vt80 1000 0 ves 1000 0 vets 1000
OLS 0 ves0" “Eee 1LE8.0 OLS S€80 1000 S€80 1000 ‘£80 1000
 SLs St80 St80 |St80 SLS SE80 0000 0 SEs 0000 S£80 0000
08S 980 980 |9€80 O8$ 9€8°0 9€8°0 9€8°0
SBS LE8O LE8O |Lt80 8S Le80 LEO LE8°0
06S LE80 LE8O |LE80 06S LE8°0 LE8°0 Lt8°0
S6S LE8O LE8O |LE80 6S LE30 0 LE8 0 Les
009 8£€80 8£80 |8€80 009 8£8°0 8£8°0 8£8°0
$09 6t80 6€80 |6€80 S09 6£8°0 m< SRE 6£8°0
O19 OF80 O80 |OP80 O19 0v8°0 O80 O80
SI9 ch80 ch80 |cr80 S19 0 v8 cv8°0 0 cv8
079 vy80 vH80 |hb80 0c9 vv30 0 vV8 vVvs'0
S79 9F80 9P80 |9P8'0 St9 9¢8°0 9780 9780
0£9 8r80 880 18780 O£9 8780 8780 8730
St9 OS80 OS80 10680 $e9 0S8°0 0S8°0 0S8°0
O09 cs80 cS8O |cS80 OvV9 cS8°0 cS8°0 cS8°0
SPh9 PpS80 S80 {S80 Sp9 0 vS8 vS3°0 vS3°0
089 9680 9680 19S80 0S9 968°0 968°0 9680
cso Ge LoSOo Oss 1LS8 C59 LS8°0 LS8°0 LS8°0
099 LS80 LS80 |LS80 099 LS8°0 LS8°0 LS8°0
$99 LS80 LS80 |LS80 $99 LS8°0 LS8°0 0 LS8
OL9 8680 8680 18680 OL9 8580 8680 8530
SL9 6S80 6S80 |6S80 SLO 6S8°0 0 6S8 6S8°0
089 0980 0980 10980 089 098°0 098°0 098°0
$89 1980 1980 |1980 $89 198°0 198°0 198°0
069 C980 c980 |c980 069 C980 C980 7980
$69 €980 £980 |£980 $69 £980 £980 £980
OOL 980 980 |7980 OOL 7980 798°0 798°0

LT,dy [ejOdS DOURIpEIs1OJDey (y)4g

Jo oy} jusdsaroNTJ
sajdures
ase poyeNoyeo
WOIJ
09P

=N WM (y)"S/(Wa-N +(V)
59 = (VK)
79 K (KOWY)'S
00E =X
WIJ)

aoyM
10}OeJ
‘(A[JUO

—(y)S¢_
eIOods

767
VoULIpELI
UOTOATJoI

[eryoods
JURIpe

Jo
0}
9q
389}
AoUdTOIJJO
‘(feusaj}xo)

‘gousosoutuny

JATIOIJJO
sanepas=CY)

19MOd
d01nos
‘(paztp
‘umouy
NV— yuRIPeL
pormnbe
‘UOTIL}IOXS
ferjods—(Y

eIeds
 6£0°0 1786 088°S 8£0°0 bLl’6 Ors's 9£0°0 CCC'6 OIS'*s 09
980°0 7£8°6 787 730°0 €09°6 €8I'P $L0'0 .896°8 Telit Oss
$610 6r1°6 S8L°7 881°0 p£8'8 689°Z 7910 009°L ELEC Or
88£'0 peo'L €7S'I ZLE‘0 p6S°L 8Sr'I 60£°0 7ZE9 rIZI 0€
ZIL‘0 79r'9 9L5°0 6990 7L0°9 IPs'0 S7S'0 ILL’? Str'0 0z
07ST €€8'P 680°0 vasa | 09" 180°0 696'0 080°€ LS0°0 Ol
798°7 10r'€ 7$0°0 6br'7Z 806°C rr0'0 60S "I Z6L'I L700 00S
pls 8SE°Z €9€'0 067'P 816'1 6670 p7E7 6£0'1 091°0 06
19h'6 819°T Bag 96€°L c97'I 0L8°0 9€9°€ 7790 87r'0 08
7L6'b1 8s0'l hee 61E TI 008°0 SILI 9IT'S Z9€0 9LL°0 OL
667 61 769°0 Z9E'E $80'+1 90°5°0 rSv'7 1¢3°¢ 0170 610'1 09
8£9°0Z evr’ SI6'€ Z9P bl O1E'0 vyl 7 Ser's LII‘0 1€0'T OSP
86661 797'0 SL6°€ 68r' El 8L1°0 689°Z Lv9'P 190°0 9760 Or
879 FI 7Z1°0 L66°7 8L9°6 180°0 €86'1 LOL'€ L70'0 6r9'0 O€
6r6'S L€0°0 SETI 668'€ 7Z0'0 Z18°0 9S7'I 800°0 7970 0Z
OLS'I 600°0 67E'0 €£0'1 900°0 LIZ‘0 Ore'0 Z00°0 1L0°0 Ol
p0r'0 Z00°0 $380°0 897°0 Z00°0 950°0 €60'0 100°0 610°0 00+
680°0 000°0 610°0 090°0 000°0 €10°0 €20'0 000°0 500°0 06
020°0 000°0 700°0 r10°0 000°0 €00'0 900°0 000°0 100°0 O8€
(W2W)S (V)A)S (v)xW)S (v)2(X) (v)AW)S (v) (Vv) (v)2(K)S (WAX) S (v) (VY) (wu)y
SS ee a a ee ‘YISUDTIAC AA
2D jueUIWN]|] q JueUIUIN]|] ||]
V jJueUIUIN
WU OT = YV S[RAIOJUT YSUdPOAVAA JV WU O// 0} OSE = Y Wor ‘oO ‘gy SjUvUTUINT]] prepurjig WID Jo suonnqiysiqg
JoMOg juLIpey peyoedg aaynepy oy Aq powsiayy (Y)2°(Y)4‘(Y)X SuOMoUNY suTpyePy-19}OD Te6l AID (SEO AIL

768
 769
9ELt 0 c9te'0 101¢ 0 8667¢ 0 LIS¢e 0 S8re'0 6brl 0 SLOv'0 9Lbr'0 (2 ‘& ‘x)
cOl'stl 000°00I TV0°86 €c7'S8 000°00I CLO 66 LYS'Se 000°001 878601 (Z ‘X ‘x) swing
000°0 000°0 100°0 000°0 000°0 100°0 000°0 000°0 c00°0 OL
000°0 100°0 100°0 000°0 100°0 c00°0 000°0 c00°0 ¥00°0 09
000°0 100°0 c00°0 000°0 100°0 c00°0 000°0 c00°0 900°0 OSL
000°0 c00°0 ¢00°0 000°0 c00°0 900°0 000°0 900°0 v10°0 OV
000°0 €00°0 800°0 000°0 y00°0 C100 000°0 0100 8700 O€
000°0 900°0 6100 000°0 600°0 970°0 000°0 610°0 LS0'0 02
000°0 v10°0 6£0°0 000°0 6100 €S0°0 000°0 0r0°0 OIT'O ol
000°0 6c0°0 €80°0 000°0 6£0°0 8010 000°0 $L0°0 60¢ 0 00L
000°0 7900 ILT0 000°0 080°0 0c7'0 000°0 9rI°0 vOv'0 06
000°0 vel 0 69¢'0 000°0 691°0 s9r'0 000°0 c6c 0 v08°0 08
000°0 6S7c 0 80L'°0 000°0 170 9L8°0 000°0 oes "0 8br'l OL
000°0 v0S 0 Stel 000°0 e190 gso'l 000°0 cL6 0 879°C 09
000°0 988°0 6rt'C 000°0 790° 1 SI8°c 000°0 LEO 1 Se 0S9
000°0 ter £69'¢ 000°0 60L'1 vLe v 000°0 C9S*C 8S¢°9 OF
000°0 061°C 60€°S 000°0 8SS°C 00c°9 000°0 volte 086°8 0€
c00°0 este OLO'L c00°0 O0L9'€ OrI'8 €00°0 ILO'S técelt 02
c00°0 9LI'b Ste'8 €00°0 vel v 9tb'6 ¥00°0 9cs¢°9 699°Cl Ol
L00°0 91e's 676'8 L00°0 606°S 8466 0100 LVS'L vOL cl 009
010°0 Fb oo| 4868 0100 SEL 8¢L 6 c10°0 Ors'8 6LS II 06
910°0 C66 'L LIV'8 910°0 96¢°8 tv3'8 810°0 8CC'6 6126 08
020°0 Lvl'6 CCEL 1700 vet 6 CLY'L 120°0 LSv'6 ILS°L OL
770
[eysedg
sumppyeyl-10jfOD
(Y¥)2‘°(Y)4
juvIpey
EOI
JMO
AGL
[E61
AIO
suoyouny
payysioyy
sayepy
“(Y)X
oy)
Aq
WU OT = YY S[eAsoquy yysuspaavAA Je UU OL/ 0} OSE = YX WOU“ G‘P'G**>q sjueUTUMNTT-q AID Jo suoynqiysiq

|] “Sq
jueUIWwO] |] $9q
jUueUIWIN] sq
JueUIWwNy]]
_
SOAR
ISU
Se
(wiu)Y (V)xW)SS (VAX) (V)2(K) S XV) (Vv) S (VAX) (v)2(v)S (V)AW)S (V)A)S (v)2(v)

O8E ¥00'0 000°0 0c0°0 900°0 000°0 1£0°0 600°0 000°0 0r0'0
06 S100 000°0 £L0°0 7700 100°0 0 yor 8720'0 100°0 4am)

00r £80°0 700°0 0 vot 'O CIT £00°0 I¢s'0 Lel'0 700'0 6¢9'0
Ol V8c'0 800°0 yse'l LLt'0 010°0 I S6L LSv'0 £10°0 °C O81
07 S160 L700 PV 86t 8811 ¢t0'0 80L'S I vcrv cv0'0 0v8°9
el
6l?
O£
6vL
0
CIT
ves
Soe
¢L0'0
1S6'8
°C
6c£
¢60°0
TT
61
688
OV
7970
9€8°C
$96'¢
L810
Fl
9Sr'¢
0
8cc
LI
9ce

OST
C7
6£1°
SLv'0
SELES
00C'P
PSe'0
‘91
€cs
COLN
Icr'0
17961
©
09 I8Z°7 PLS'0 096°SI £ CVC 699'0 80981 LI9'€ 9vL'0 07 6SL
OL LSs'1 698°0 CI 6£C Ccl'C 686'°0 €I 666 °C OEE 880° SI L6¢E
08 ¢¢6'0 sse'l L tv6 6v0'I I ScS L168 I 6£cI 999'1 £896
06 0 667 I cv6 4 (4458 0 Ore CHIC 06L'P 0 psec COL’? Lvl's

00S Lv0'0 ¢60°¢ 909°C 1S0'0 (64583 CI8°¢ yS0'°0 '¢ Ses 6L6°C
Ol 680°0 6I8'V OTS'l ¢60°0 Tels vI9'l 660°0 CLES 069'1
0¢ c09°0 SSL9 0 vol Lc9'0 OVO'L 9LL'0 9790 L 6re 0 66L
Of Ip9'l 9vS'8 8Ir'0 989°1 V8L'8 0 Orr I OIL 6£6'8 Ltv'0
Ov I@8°C L976 L610 698°C SCr'6 10¢°0 006°C 6 97S £070

Oss PV 8PC OSL'6 980°0 LIT 96L'6 980°0 ILC v08°6 980°0
09 999°¢ L9P'6
. Lt0'0 Sc9'S SIPV'6 L¢0'0 y8s's 9VC'6 L¢0'0
OL 8v0°L v08°8 610°0 Lv6'9 8L9°8 610°0 tvs'°9 6PS'8 610°0
 771
098¢ 0 OST’ 0 0667'0 18s¢0 16ce 0 Lele 0 00cE 0 9Lbe'0 yee’ 0 (2 ‘€ ‘x)
sorponewolys)
8SS CCl 000°001 6£6 v6 €18°801 000001 LI0'S6 $80°C6 000°001 Cv9 $6 (Z ‘X ‘X) swing
000°0 000°0 000°0 000°0 000°0 100°0 000°0 000°0 100°0 OL
000°0 000°0 100°0 000°0 000°0 100°0 000°0 000°0 100°0 09
000°0 100°0 c00°0 000°0 100°0 c00°0 000°0 100°0 c00°0 OSL
000°0 c00°0 ¢00°0 000°0 c00'°0 $00°0 000°0 c00°0 900°0 OP
000°0 £00°0 600°0 000°0 £00°0 010°0 000°0 b00'°0 I10°0 Of
000°0 900°0 $10°0 000°0 900°0 L100 000°0 L00°0 6100 0c
000°0 €10°0 Lt0'0 000°0 $10°0 1v0'0 000°0 L100 Lv0'0 Ol
000°0 $c0'0 690°0 000°0 870'0 LLO'0 000°0 c£0'0 680°0 00L
000°0 6r0'0 9¢1'0 000°0 bs0'0 OST 0 000°0 790°0 cLI'0 06
000°0 vil 0 sle0 000°0 9c1 0 ort 0 000°0 orl 0 000 08
000°0 L770 6190 000°0 6v7'0 089°0 000°0 987 0 18Z°0 OL
000°0 97h '0 IST 000°0 cor 0 CST 000°0 vcs 0 LIv'l 09
000°0 cSL'0 C66 I 000°0 018°0 Lyi? 000°0 $06'0 86¢ °C 0s9
000°0 687 1 86C'¢ 000°0 98¢'I LYS't 000°0 Les’ Ct6'€ OV
000°0 6S6'I 6rL P 000°0 680°C €90°S 000°0 6L7 7 ScS'S 0€
100°0 C66 °C OIL'9 c00°0 Coe e 160°L c00°0 SIVe 8S9°L 02
£00°0 890'P 801°8 £00°0 S9C'P 00S°8 £00°0 ISs'y 1L0°6 Ol
L00°0 89I°S 00L'8 L00°0 ples Lv0'6 L00°0 L99°S Ors 6 009
600°0 981°9 L8t'8 600°0 ese'9 €19°8 010°0 €8S°9 $768 06
S10°0 869°L 801°8 $10°0 988°L S0e'8 $10°0 L80°8 LIS’8 08
000°0 C7E'8 esol 000°0 0£9°8 9C6'L 000°0 700°6 LL7’8 OL
000°0 176°8 77E'9 000°0 $70°6 78°9 000°0 998 861°9 09
9€0'°0 788°8 rely 5€0°0 8188 LIL + €£0'0 660°8 97E"P oss
611°0 Z9€'8 SLZE LITO LOZ’8 vIZE r01°0 LOT'L Lv8°7 Or
$S7'0 ZOE"L €L6'I Lvz'0 COIL 076'1 Z1Z'0 980°9 vr9'l O€
10$°0 $879 1L6°0 8Lr'0 Z00°9 L760 L8€0 rS8'P 0SL°0 0z
9L6°0 p87'S L7E0 7680 Sr8'P 6670 059°0 EtS'E 8170 Ol
$80°Z 86£' L€0'0 7181 778'€ 7£0'0 OST I r7r'7 0Z0°0 00S
897 p8r'¢ LOI'0 87S € 088°Z Lel‘0 896'I L09'I 9L0°0 06
prl'8 $L9°7 6r8'0 OLr'9 S717 rL9°0 pl7'€ 9L0'1 Ive'0 08
L88°€I 7S6'1 790°7 1L9°01 00S" S8s I $96'F 869°0 L€L‘0 OL
76781 ere'l 89I'¢ 896 °€l L66'0 OSEZ Z08°S 97r'0 $00'I 09
85017 S¥6'0 pl6'€ L66'v1 €L9'0 L8L°Z Z08°S 097°0 8L0'1 OS
6rE 07 €v9'0 696'€ 766 '€l trr'0 67L ‘7 796'P 9S1°0 896'0 Or
$98'F1 OLE'0 I10'€ L666 6r7'0 $Z0'Z 89¢°€ €80°0 789°0 O€
7Z1°8 6L1°0 80L'I IIp's 611°0 LEl'l S6L'I 0r0°0 LLE0 02
7L9'7 090°0 18S°0 98L'I 0r0'0 88£°0 609°0 r10°0 Z£1'0 Ol
€9r'0 110°0 £010 El€0 L00°0 0L0°0 1110 €£00°0 $Z0°0 00b
€r0'0 100°0 600°0 620'0 100°0 L00°0 110°0 000°0 €00°0 06
Z00'0 000°0 100°0 Z00°0 000°0 000°0 100°0 000°0 000°0 O8€
Cums OMMs OWS GZS M)"OOS COMxMs COMZO0S COMAODS COMODS (wu)Y
——_——————————___ ‘ysuajaaen
OD JueurWNy|] q JueulUIN]|] V jueuluny|]
WU OL = VV S[RAdo}Uy Ysuapoav Ay Je WU OL, 0} OYE = Y Wo “D‘g°y sjuvUIUN]]] pavpuBly YTD JO suoyNqiysiq, 19M0d
jueipey jesoeds aayepy ayy Aq payysiany (Y)9!2*(y)old “(Cy )olx suoHoUny SUTyeP-10}/O) p96] ALD (SEC eT AAPL

Ti2
 773
SOLE'0 16l€0 pore’ $L67'0 9cSE'0 66h 0 6crl 0 6S0r'0 CIS 0 (2 ‘€ ‘x)
soonewolyD
LL Soll 000°00I 867 L6 6ve v8 000°001 LOC 66 00C'SE 000°00I 6ST ITT (Z ‘X ‘X) suing
ee ee ee ae sce ea nde ie aon se ateinensonaie oa
000°0 000°0 100°0 000°0 000°0 100°0 000°0 000°0 cO0';O OL
000°0 000°0 100°0 000°0 000°0 100°0 000°0 000°0 7000 09
000°0 100°0 700°0 000°0 100°0 €00°0 000°0 700°0 900°0 OSZ
000°0 100°0 €00°0 000°0 c00°0 ¢00°0 000°0 ¢00'0 010°0 Ov
000°0 £00°0 $00°0 000°0 b00°0 800°0 000°0 800°0 610°0 O£
000°0 700°0 €10°0 000°0 900°0 810°0 000°0 s10°0 0r0'0 0¢
000°0 110°0 8720'0 000°0 s10°0 6£0°0 000°0 €£0'0 €80°0 Ol
000°0 ¢z0'0 c90'0 000°0 €£0'0 — -¥80°0 000°0 ¢90'0 L9OT'0 00L
000°0 €S0'0 9e1°0 000°0 690°0 8L1°0 000°0 OcT'O 9€£°0 06
000°0 bIl'O €67'0 000°0 SPl'0 ble'0 000°0 6S¢'0 999°0 08
000°0 €€7'0 8650 000°0 v6c 0 ISL'0 000°0 66r 0 hdocogh OL
000°0 Isp'0 cel I 000°0 LSs0 OIv'l 000°0 1160 90£°C 09
000°0 808°0 bl0c 000°0 ¥86 0 bSP'c 000°0 19ST c68'¢ 0s9
000°0 bre l 92e..£ 000°0 8191 e88'e 000°0 960°C €66'S Ov
000°0 vcl'? ISs'y 000°0 OcS°7 LSL’S 000°0 BSL't €86°8 Of
000°0 986°C ccb'9 000°0 bore SIS°L 000°0 9720'S 60801 0c
000°0 vL6't 9SL'L 000°0 6LS'V St6'8 000°0 LGe 9 6be Cl Ol
000°0 Lc0'S €86°8 000°0 6L9°S L69'6 000°0 99P°L 8PL Tl 009
000°0 891'9 bL8'8 000°0 68L°9 89L'6 000°0 LIt’8 L96'II 06
000°0 SEC L vrr's 000°0 9CL'L L106 000°0 OrL'8 10c Ol 08
 000°0 €SL'L 6cI'L 000°0 C76'L S8CL 000°0 cols IshL OL
000°0 s9r'8 S86'S 000°0 €89°8 690°9 000°0 10L°8 esI°9 09
9£0°0 0£8'8 LIL'P 9£0°0 188°8 bel 9£0°0 ~ C168 I9L'b USS
€cl'0 6L9°8 66£ €c1 0 vb9'8 S8t'£ ccl 0 69S'8 9St'e OF
$8c'0 1078 IrCC C870 cII’8 COI? LLC'0 LS6°L OSI? 0€
0950 8c0'L 980°1 8rs'0 0L8°9 c90'1 0€S 0 979°9 L70'l 0c
180°1 Sss's c9€'0 Or0'T 6c9'°S 8re0 ¥86'0 lees 6ct 0 Ol
COI? 09S 'P 8£0'0 980°C Lee v 9£0°0 026'I 8h0'°P ve0'0 00S
CSIP 16t'¢€ c9T "0 688°¢ 9LI't ISTO pss £06 °C 8e1°0 06
60¢'8 6CL'C 998°0 COLL OLS °c €08°0 8169 CLEC IcL'0 08
Sec vl 100°7 SEL ScO'el 1es'1 pe6'l psy il viol SOL’! OL
€19 61 Ipr'l LOC L69'LI 00¢'T S90'€ vot ST bell 0S9°? 09
LIS °C O10! psi'y 880°0¢ 1060 3 2 9S0°LI S9L'0 OLT't OSb
9€T OC 6£9°0 Lv6't SSL LI 09¢°0 core c69 PI y9r'0 998°C OV
€6S el set 0 vSLc 68S II 687 0 8he °C £076 6¢C'0 y93'l 0€
I1€°6 $070 8661 Oc8°L cLI'0 vr9'l $L0°9 vel0 LL7'I 02
869° £80°0 $08°0 990°¢ 690°0 L99°0 OLE? cS0'0 LOS'0 Ol
vrl 0 L100 S9l0 Bo 0 bvl00 9€1°0 8SP 0 I10°0 cOr'O OOP
€90°0 c00°0 v10'0 6r0 0 100°0 1100 s¢0'0 100°0 800°0 06
y00°0 000°0 100°0 €00°0 000°0 100°0 c00°0 000°0 000°0 O8¢
W)TZ0)S WMS WEDS CONN) MOMS WOMANS CONZWS (CK) S — (K)XW)S (wiu)y
een a ATS IOIO A CAN
*q Jueurunyyy] *°q jueurun||] “q yueurwinyy]
*q sjuBUTUINIT]-G AID JO suonngqiysiqg 10M0g
WU OT = VY S[BAloJUT Ysuseaw AA Je WU Q// 0} OSE = Y Wor ‘4 G‘ Pq‘
jueIpey [eyoedg aayepy oy) Aq payysiayy (Y)%2(Y)NE (YX) X suopouny SUT eP-10}0) p961 AID (BE S)AT AIL

774
 775
Tese'0 eLI€ 0 9667 '0 esce'0 60£€ 0 8ele0 cLIe 0 L8ve'0 Ipee 0 (2 ‘A ‘x)
sonlonewolyy
I7L°071 000°001 8ch v6 18¢€°Lol 000°00T S78 V6 08606 000°00I 008°S6 (Z ‘X ‘X) swing
000°0 000°0 000°0 000°0 000°0 000°0 000°0 000°0 000°0 OL
000°0 000°0 000°0 000°0 000°0 100°0 000°0 000°0 100°0 09
000°0 000°0 100°0 000°0 100°0 100°0 000°0 100°0 c00'0 OSL
000°0 100°0 €00°0 000°0 100°0 €00°0 000°0 100°0 ¢00°0 Ov
000°0 700°0 900°0 000°0 c00'0 900°0 000°0 €00°0 L00°0 O¢
000°0 700°0 010°0 000°0 ¢00°0 7100 000°0 $00°0 €10°0 07
000°0 010°0 970°0 000°0 1100 670'0 000°0 €10°0 r£0'0 ol
000°0 1Z0°0 €$0°0 000°0 €c0'0 6$0°0 000°0 Lc0'0 690°0 00
000°0 cv0'0 801°0 000°0 9r0°0 0cI 0 000°0 bS0'0 6£1°0 06
000°0 L60°0 8r7'0 000°0 LOI'0 ¢L70 000°0 Scl'0 Ice0 08
000°0 €0c'0 0cS 0 000°0 $770 SLs'0 000°0 1970 999°0 OL
000°0 O8e'0 €96'0 000°0 9Tr'0 esol 000°0 SLv'0 cO7'T 09
000°0 €89°0 COLI 000°0 Ivl'0 8r8'1 000°0 S¢8'0 180° 0s9
000°0 Lol CL8°C 000°0 $6C'l 601° 000°0 8rr'l SLv'e OV
000°0 v68'1 Sce'V 000°0 C£0°C Ip9'p 000°0 GLCc LOIS O¢
000°0 v78°C 9L0°9 000°0 y00'€ c9r'9 000°0 Ga 3 8t0'L 02
000°0 BSB OES*L 000°0 ILO'P 9r6°L 000°0 I8¢'p OSs°8 ol
000°0 CL8'P 8Ie'8 000°0 OOT's LOL’8 000°0 CCP’ LS7'6 009
000°0 Orl’s 6S7'8 000°0 ve6's LEs'8 000°0 661°9 6168 06
000°0 Lv6'9 801°8 000°0 e9T°L 19€°8 000°0 LOV'L S¢9'8 08
 Wecumeiy.y
Llewosr
891I€0
9S90S0

8L1€0
cOvVeO
IL8E0°*

9c57
O0Gt

19VEO.
COL
Sse
50S €£9S0 Sreo 06090
6€£90
Ayloneuois

PABPULIS AID 0} Joadsay ym YNUIG s0joD eUNdG Jo sayeuTps00D AyOHeMOID-(4 *x) TEST ALD puwe syysuapeavA uowisuvsy,
So}eUIPIOOD

Ome

0.
OR

StSuP

6€I19S0

Ve9CO
werculr
¢tc6r0

80890

96rv0
sodso
cStS
S090
5605
Ge9S9cG90870 L6LEO0
OOPEO
coltO0
009

cugese a
<@C9s”0

Cece,O
SS9eoS 0

“O19S

PLD«6S
LEIS
6609
=A

E196

O89
609

1065
rcss LBS S$ 98S
SUIBUITOAe
uOoT}ISUeI] 7ss. I16S

6OI8S”
“99750. 09bS
*OLOV0 OLMV

auiggse Lisp
S0IV0 80cr

OESEO Ss6p
10070 I179r
OD0tPO 96S

clive FccS
LOVED Tels
LI9b
“Scrr’-cespO Vi6r 6SIS
SOS

Glse ScS
fPlyO: 6SS

0
1
A

L6¥VV
B8S7SO0 91650, SE8SO
98850
Ayloryeutolyy
Sa}eUIpIOOD

0.
0”

0.

0”
OLLO0.e8

¥6LV'0
90LSO0

cOoeso
06950

POOSO

6IZEO
cI8S0

tOLyO
6Y8o

POO
[SeeQ
SOIEQ SsScteO 66660
O06€0
O18t
OOL=A
0

(SOC SIS’. (SS --“6

oii0

0995

“E879
OIS9

CESS
L6rS

86bS

ecrs
90S
ors

prLeO Saopy 0SsTr

SYIBUDTOAL Stes Tv6S 86S
276° 9665
uOT}ISURI]

LICVO-*LoVS0
s'tcry

Y9E9EO0 L86Fb
CLIEO OELD
Lops.

OOS.
8EtPO OCCh

6L9E0 C6CS.
876E0 ISth
SOS

“L9LE0 StS
LIP E9Sh SOV £90
1720S
AA 6SSh0

8SEV0
08tr
LSSP'O™OMPPSO ..86¢50PvS6F0
OFISO
Aylotyeurory7)
Sd}
eUIPIOO?

“ROOSO
6IPSO0

66L7P0
O6IS0

O9C70

VOLEO
“869F0
99690

OI6VO
0

0
tO9S

PISS OW STSEP
CO9EO OOIPO 86570 O00SO
€08P0 ES8tSHeeOLy
608
SLLLON

69€9™
B8SPS
CLES“
v6H9

OLSS™
=A
CLES

ALES

SSES

;,
Tels
SYIBUITOAL CLOO 0609 SI7T9
.9919 ELTI
uoTIsUeL]
IStrPO.escloO
CLese

vreroD Piss
OLty
6cES

G66b Cee.
ELIPO D6Sh

COlvO Sees
6S8€0 SO8h
SOSPO) AEH

OD6EO EIS

166s
BOTY 98h COIS B8CTS
CLIS
AA,
O66E0)
VISPO
LILVO

Lbs
OLTI
SO8EO0
687F0
G6E8S
8879
I0SEO
OlEVO
PSlLS
6rt9
peleO
8levO
G6IEPO”IESPO 6P0SO
-LIOSO 66670
Ayornewoiys
Sd]
PUIPIOO.D
UL7T8V0

C9ISPO
OU
PNCHY
L86V0

GLOrO
POESO

959F0
87750
etree

SlévO

£t0OV0 B8SIPO0 UcssvO PvS8PO0

859F0
006
OPESO
S$ 0

0609
S$itZS

OLSS

=A
6859

60CS

CtES
SITS

E709
Hes

“CSCO S79 OLCO SrEd9

£679
Lit

SYIBUITIAR
uonIsuel]
ISITOS.P6cCr
F60cS

vretrkO. Ctrh

666€0 LTZIS
IEIVO O6SP
SLOEO OSLP

O80K0 PSPS
UPLS
O8IPO. S165
ELIS

619
LOOPO OPP

SS
OCChmECtn
OSbh GL8b
S6lh ECOS
AA
Oe
E8CRO.(7COVO
Lec
PAV
roto
elite
S35
CCTISA*G
TREO
STIORO.
SFIVO
COOLS?)
S0LEO"VCc6ld
SILS=*VEIVO
IXLbVO 6r8r0
YOI8PO
tCSPONRLSLPO.
Ayroneulolys
Sd]eUIPIOOZ
Oi
pvOSsrpO

pStTrhO0
ObLVO
88sP0

6EVO
106rP0
ISISO

OorCyOLrbrO CEOSO
I76F0
0S6=A
O0C90.
OFL9-
OOIS

O19S
O86S
S305

S6IS
80S

68 F0P9 O6P9
Cryo L959
SYIBUITOAT
pV jueuunyyy

uonisuel 6r9
(L°¢)1 qe

JDILP
Ossr

cSOS
9¢¢¢

ELED
ESP

60
OSH OO08F 0
O16P
16S

ScP 66h
219 8
8

[, A
ly
ty
NG
axa
OSC
OO
as
=
OR
0HSIOV'O
88tFi0
c8IPO
O09
SLSC
COORUT
“OOSSeVIoV
cas
B8EOEO
toS
O8¥P
O09EN
Lest
CLOD;:009E
OSE
L£6E9
.O0StrPO
1Sc9 ~~
0GEEP
0-0
-4,
~
0-1
0-2
—6115
6VP
"e858
LVCSO-COLCU
C0CE
P65
&-—COKF
SOS
E-0—9-L09.-—-
E08
——-1905-0-—-
809
OVE
6STI
G69
0-158
057 S$
SOEY OIP9
OSSPO
LOOPO
O09b
6679
80EV0
YOSLHO
Y96F
E119
66770
LOPSO
89
SH6S
S6EEO0
O8SS0
669
678S
86870
O16S0
O09b
S$
099F0
B87LPO
OCLb
LITD
B8Srh0
St6PO
IOS
67I9
COPVO
YCPSO
€6Lb
1966S
LESeO
90850
O8Lb
E€8S
O00EO
OvI90
OCHO
69P OU
“LE
OO.
UR
0
Om
CHBPO'VUSOS
bo7>
CScV
L6S
“L9GP
SOOSO
C8cSa
COED"
~STCS:0...
OSLP0
SCeS”
GIZV
OCIS
<SPonp
LEDS
OSLE
68IS0
BID
LOCS

716
 =A 00S =A 00vb =A O00 =A 00¢c =A O0I

uonIsueI Ayroryeuroiy uoTIsuel Ayloneurorys uoT}IsUeI Ayroryeworys uoT}ISUeL Ayrornewoiys) uOoT}IsUeIT Ayronewoiryy
SUISUITIAe UIPIOOZ
Sd} SYIBUITIAC SdCUIPIOOS)
A SUISUDTIAe So}eUIpIOO,
A SYIBUITOAR So}eUIPIOOD
MA SYISUITIAR sayeuIpIOO|D

ly ty x a ly ty x a ly ty a a ly ty x a ly cy a a

O09E DILS ISIZO B88Sr0 OO9E 6095 69L10 OIEVO ODDIE L6PS CHHIO EEO OOSE TLES L6ITO S8IEO OO9E 60S OcIILO S8610
OCH YILS 7ZTO SEOSO OISh TI9S OO8ITO S77SO HErH 86S LOVIO LesyO I79p YLES LLE0O0 €66F0 OSH E1CS 65800 LS6cTO
109P G6ILS €0€70 86950 LLIb S19 I8810 POIDO B8TSh OO0SS €6E10 S66P0 66Lb 98S 62600 60990 VPILP CCCS 6rS00 c6Sv0
LE69D TCLS 6ETO G6II90 E€9Lb 8I9S 85610 79590 SI9F TOSS O6EIO EESSCO Leh COPS ELOIO vESLO L16Lb 6ccoS ttr00 8pSS0
S6Lr LTZLS LOSTO E8p90 LI8h 7729S 61070 16490 I189h €0SS OHIO 80090 166h EPS L8II1O PHLLO €68r vHCS 9800 pILIO
HBr
S OELS PLStO SI990 I188r LZ9S 90I7TO 16690 LSLb LOSS G6EFIO 9PS90 EOS OHS SEEIO EC98L0 SS6b LStTS Ivv00 89€L0
£€06r DELS 09970 OCLIO 66h EOS vIETO ELILO OS8b PISS SESTO S9OILO IT60S CHrPS SOSTO 96810 S810S SLcS s8LS00 vesLO
T00S OSLS Cr8TO I8L90 E€C0S S9S SOv~O 8LILO Tp6r S7ss LOOTO OlPLO EIS ssSrs e890 CO8L0 OLOS “L6cS 98200 cOIsO
£905 S9LS 6670 ~7HLIO TOIS OLIS LO9TO 8IILZO L86h TESS E9LIO0 E0SL0 TOTS v6PS 8c20C0 O069L'0 LTIS Tres Of010 88180
Cris FO6LS rrceO S6S90 90CS PILES ETOEO G6E89:0 LOS eoss cO0TO’
0. src OLS F69e IvetO 9ce90' lis Cres 97cro iiso
L8PrS 8609 $70S0 19670 ESSS 0009 170650 896h0 DLIS T6SS €0KTO 99EL0 SSS 69LS 90CFh'0 YLLSO 60CS ITLES OISTO pSO8'0
DILS SLLL CEEIO P9O9EO ECS GSLLL OLSIO LeEO €6S8 CLLL 00890 861E0 CTLOD CLLL 800L0 16600 I1Sc9 89LL s88I1L0 cl8cO
S6IP DILS 96790 SEXED T8Iv EBS LCES9O OOPMEO B89Ih 86S 6SL90 ELIEO 86Ch CLOD 92L90 PESTO LEIb Sco CHILO tLLC0
LO YXILS vS090 LorEO C6EH ECBS 6790 POTEO TSEP 86S 88h90 DSODEO ElhMh €L09 OSE90 60970 6CEb CSt9 90990 69770
6ESH B8ILS €08S0 LSZTEO TSShr 78S
S O6SO CLETO O9bh 6£6S 80790 LE8TO OSh PLO9 07090 HSPTO DIPh €S¢C9 §=SI09O0 8ccCTO
H9r
8 OLS 00950 IIIEO PvHr9r LTS OOLSO SP8TO S8Sh Th6S £9850 LEXTO CC9Fb LLO9 109950 9b7TO 68bhr S$ST9 §=PO9SO cEOTO
618P 87LS OSESO LS67O GI8h OXE8S 96E50 $8970 89h rr6S 90990 O00STO IP8h cé609 s0d0SO LCcCO ILSr LSc9 6L1ISO0 6€810
896P SPLS LOTSO EI67O TH6P 68S 6PTSO B8P9TO €6Lb IT1S6S C8ESO COTO LP SOI9 EC8PO ELOTO E€0OLb $909 06870 90910
91S _108S 96670 OCIS..09670..
p'68S.0... LOOS 76970. 86h $969 89ISO 8SE7TO GEOS EID cEesvO. ecsocd C8lp SLC9 COEPO. 9cS1O
LBES TS6S ECOSPO I77EO 6L7TS 9866 O09F0 824870 OCIS 6109 I16LP0 SEr7O0 SOIS 9919 I8¢th0 B8IITO €68b 9679 %9P6E0 88710
0Ss
Ss 1809 O00r0 IISeO0 SOrs OII9 10070 S8I€O Les ZOI9 Sé6ehO Y9E9TO YOCS 6779 IS9E0 OCETO 866h LEED vIseO 6IStO
OCSS SLI9) L8Se0' ISLEO cLYS S009 ..tOSEO SPrrEO 6O0ES S919 .SO6EO LO8TO C9CS OES .OLOEO IcscO es80s Oro 6roc0 OLISTO
G79C 8679. SOLED IEOPO P7Ss GOED. 88600 LILEO GLES £979 O67E0 E80e0 LOES OCHO O00Sc0 iclce0 Tels tLlyd cS¥c0 90ELO0
F895 68h9 YIPSTO BSErFO HBSS LESD L8ITO OPIVO 6SPS LLy9O cOCT SLSE0 PFPRES P9So ScsTO 09600 E9TS CSS9 6W0CU LGLVO
JO

9Y}
9Y}
9Y}
IO)

YIIM

‘7
SOUT]

Jo
JO
IGRI

JOYS

syL,
Jojoo
pays]
2aogn

9Y}
IO)
SSOIOV
19}Ud9

NUNS

YIIM
sjDalQO

dA]
SaUT]
DATOAUT

“BIy)
JOYS
ewido

AA],
asou}
MOjaq
[eIJOIds

|poaisTT
SoAINS
saamno
sjOalgo
JATOAUT
‘((L'¢)Z
[eIOIds
sDULOATJaI

dDUROOTJOI
J0UR}}WISURI]

Tad
(J0UR}}TWUSURI}
778
0}

Jo

‘x)

pue
ALD
AID

WIM

[EEL

PBL
10j0-

s9
INU

(L'¢)IT
Pq
a

JeudO
oodsay
006
CX
Ae
OKO
OS6—
Snk
=A
009
PAEPURIS

uoyrsuesy
soyeurpsooy

sypsusjaavAy
uOT}ISURI Ayroneuloiy7 uonIsueIy Ayronewoimys) uOoT}IsUeLT Ayroneuoiys)
TL uUOT}IsUeI Ayronewoiys
TL

jueurumyyy
uOTISUeI Ayronewomys
AA, SYIBUITIAL Sd}eUIPIOO.
AA SUIBSUITOAR eUIPIOOZ)
Sd} A SUIBUITOAC SJVUIPIOOD
A SY}BUdTIAe SCUIPIOOD
A SYISUITOAL Sd}eUIPIOOD

AyoneMomyD-(4
ly ty x a ly ty x a ly cy x a ly ty x a ly ty x a

sOM9E “S969 ““SL8C0 “OCEED O09E OTIS <GE9~0 sLECt0 O'O9E B:16S CECTO
< .06CED .O;09€ s8LLS 91610 VOIt0 OOS€ L998 -VELOLO BEbCO
867 $979 6670 EISEO OIlp LII9 10870 ZESEO LLbh 976S v0b7O SrIvhO T8th I'8LS 85610 YIS9EO E€6EP 99S LLIIO Oz~seO
IPH ELTO LIDEO OO8EO B8Shr OCI9 9870 80070 SPSr ITt6S 96P70 POSHO 89h E8LS C00TO 6hp0F0 80Sr 899 OOLIO OclIVO
TISh L879 OEFTEO OSIPO 99Sh TEID 6S0E0 O98hh'0 YX6Shb SE6S E8S7TO0 I08V0 THSH LBLS $9070 S8hPrVO Y6Sb E€L9S 6blL1O0 c8Lv0
PLSP €0E9 89€€0 SIPhO OC9F IvlI9 C8IEO SPLHO O89b XF6S O9LTO B80ESO TIOF E€6LS OSITO €96F0 CS9b 8249S 10810 LSctSO
OOP IcE9 80SE0 PS9PF0 CTLOb €SI9 LIEEO v66V0 LLP 896 E20EO0 60850 LSI L6LS IctcCO S6cS0 8O0Lb €89° eL810 O€4S0
LIP
6 EPED PH9ED 96870 Help ELI9 EISEO B8L7SO TI8h FL6S c60E0 C6850 669b CO08S 86770 LOSSO 6LLH €69 6610 LSc~90
TLL 69€9 SOLED LOOSO S6Lr 619 LSX9EO I7PS0 r68r 7009 B8IE€0 IP090 OSLb 608 cOV~O 81650 PC8r I0LS 88000 €cS90
S9OLP v0PK9 L88E0 97IS0 S68r 9679 9b6E0 LESSO S6b
8 TEOD SISEO B8P090 HI8h IC8S OSScTO LETIO P86b ELS 90870 L790
SO8P IS79 €00V0 90750 PS6b OCED YICIPO OISSO SO00S 1909 61960 $6650 S06b PP8S P8LTO0 P8P9O 60S LILS €0L70 00690
P8h
I 6199 801F0 IS7SO BIOS 6579 PSErFO 90P50 9605 E519 O80F0 OSLSO 86h IL8S O00E0 17590 COS 96LS Of670 86490
06P
1 T9SL I8%h0 89790 1906 OSSL OFSPO £67SO B87%S CHSL 8S8h0 18050 SHES 8SSL B8rISO Sé8PO ISPS B8SSL =StpSO cSSpO

66IP E£06b PvOtTVO 60ISO B8EIP 1906 M8rh0 OIZTSO OEIKV 877s II8VO SOOSO ECIP SHES LOOSO eESLvVO IIIb csps 6ltSO e8pv0
I19CP PO06r CEIVO 65670 Pb6Ib I190S vrrOyO IEISO LICH 67CS PE9PO G6ILHO G6O0EF YIPES YDLLVO PvOEVO CLEP ESbS SLLYO ISLe0
STEP S06b IE0V0 ISLHO S8Lt% 7790S SctevVO 906F0 IEP 6CCS PISPO YCSPO YX6th 8rES B80SV0 CL6EO 6thh SSS CccSVO OSPEO
O6rb PP L69EO0 YILOVPO PEEP 90S SItTVO OOLHO ODMH TEtTS 66¢70 B8SIFO0 Pvb6br ISES COIPO SOSEDO CTSSh O9PS 8EIVO S000
09P
0 86h 86780 C69E0 OPH 8890S 066E0 PH8TPO LOSH SETS L00P0 OCLEO O9Sb SSeS SOOPO 6SceO0 II8h 98S «11960 cLbcO
899b LIP I0PE0 EISEO SESp LLOS 6PLEO GP8E0 ECOW HHTS CELEO GIEEO Y969P 99ES YOLEO 06870 C6h OISS L6veO SOVCO
LL 00S
Se OU hceOU TESS 99r—
G6 GOS.
SS 0. PSE UEPEO. GOP.LL ScoOU C09EOO GEE ciUe 9ro
6 0 998 Cesc) 90S. PecoO SotOU Buec
TZ; P87TS LIIEO G68IEO DI8h G6TIS G6HEEO YI6IEDO G6LLH 97S O0SEO 600E0 OS8h B88ES LISEO 66970 HHS SL9IS ~S6IEO 6cvpTO
Seesi piv0. O8PiC 0. Colt 9CIC OGCS LIceEO Pre0EeO F660 £9Eo LOECO
0 998C £10 SPS
6 Pyte veICU Leseh HoySLOU EOCUae SUC
3099 7995. COICO PITEOD PEsS OF9C 66080 PCIeO. LILS £009 OFLoO OS0E0. O6CS LO9S roltO If9C0. 389rS Vos_S IULC0 SoscU.
Gass E8650. 900E 0“ -68ce CRBe. 6509 COSCO StcheO. SO8s £19
££ GIScO 69120 Siva~ C9LS0 £O0E OLLcO Los) Vile ULLOU Pro
ELIO r8£9 00670 80€€0 6709 GIE9 IILCO 66780 HS8S 9679 OOKTO 6ITEO HESS C86S P99TO CL8TO 8809S LICD LEOTO O€8c0
 =A 00s =A 00b =A O0f =A 00c =A OOI

uoTNIsUeI] Ayronewoiys) uoT}IsUeI Ayroneuorys uOoTISUeI Ayroryeuorys UOTHISURIT, Ayloryeutorys) UOTISUeI Ayroneworsrysy
AA SYIBUITOAL eUIPIOO.)
$d} A SYIBUITIAC so}eUuIpIoOZ
AA SY}BUITIAC sayeuIpI007
A SYIBUITIAC soyeuIpIoo07
A SYISUITIAC PUIPIOOZ
Sd}

ly ty KG a ly ty xX a ly ty xe a ly ty Ne a ly ty ne a

ODDIE PSSS I6rPl0 61970 OOIE ISPS OIELO PHZETO OO9E SHES C8710 68810 OODOIE 8777S 89710 S9EITO OO09E 1190S 9PELO LPLOO
bbr
6 8°59 IbrlO IISEO Ohh SSrS 99710 OEOEO EISh CSES LIOIO COOEO SSrr E€€7S I8010 Pp86l0 69Sh 1180S 06600 LO9ITO
67SP 1965S 67Fl'O0 SZOPO G6OSP SPS
8 G6IZTIO POSED 6LSP 9XSES 06600 SESEO O8Sh OPTS 6800 99L70 6996 $606 ISL00 cOvc70
8SPSS S9SS 67Pl10 6LHrPO 69Sr I19PS E8IITO S86E0 67C9F O9ES 67600 IP0P0 L69b 7SéTS 09900 PvLOVO C6Lb CcCIS 1600 vpLovO
€69P ELSS TCLHIO SSO EC9P Pp9rPS SSIITO 60SP0 GOLP LIES 9b800 870S0 O9Lb 1197S 6bS00 IL6F0 SL8h 9FIS I1700 O6rS0
9OLP
Ss Z@sss 8PSI10 10790 €L9F 69r6 IPllO0 SsOo¢O
= LE8LH
> LILES 61800 07090. OP8r LLZS.. .6LP00°. 'L7C90 ..LSér. VLIS LLIOO. £6990)
TI8P 6859 I7910 OLSI9O YAL8b L6PS TZIELO LPOLO Lr PrsEeo Y9€800 I16h90 Ev6r OOES S9S00 CIELO LEOS SIcs vredOd ceLLO
S68Pr S095 O6LIO0 SEDLO €96h B8ISS YDISITO PSPLO 86h 80PS YPOOIO EfPLO0 B8S0S PHES L000 SO0080 6LOS 6£€7S 91500 $s08'0
OS6P 6195 67610 I07Z0 O90 rsss ES8I0 LBSLO LLOS 6SrS I8Pl0 LS8LO0 OCIS E8ES 68710 B8LO080 GIIS v9TS L7LOO €cc8'0
GODS 6£€9S PlIzZO ZLECLO “OCIS S8SS 67ITO OISLO TPIS? -P6rS <66L10 -L8LL0 BCOIS POS “61710 ~9708'0 9SIS
— 0675 6S600—
OF [9es
YI6IS PLS
6 16670 19890 PHLIS O79S SIPTO PrELO L6IS TESS 6II7TO 60920 E€6IS StrS P99TO Ih6L0 O6IS SITES 88110 €lc8O0
PE>SS OSSL IELSO 797PO 199 OSSL IP090 PS6E'O0 BLLS 8ccoL 89690 879¢0 16S
8 SScL B80L90 68760 OIID S8ESL StOLO $96c0

OILIP rsSSss 899950 S6IPO POIr 1996 69650 88860 860b 8LLS I8790 I9S€0 C60KV 816 16690 eEI7EO S80b OII9 cEes90 €S8cO
Tt SSSS 6PS0 600P0 O8t%h 7T99S Pv~SSO PH8PEO0 P8t%h 6LLS 789950 86060 OStTH 6165 88650 OC8TO OLIb ITII9 OLP9O €$9c0
LOY LSSsS S6LPO I87EO BEEP €999 LSTIO PpHeTEO YBSEr O8LS IL7SO P8L7TO SIEP OC6S PISSO ESTO LL CII9 LISSO ceIcO
ELPY 656SS vISPO P6670 E€6EF 99S O86F0 LO667TO PObPh I8LS LL6VO CISTO P8Er I6S 8s8I0SO LOITO 86th CII9 60€50 610C0
8SP
SS S9SS ELIPO 00970 ILrr 9999 86570 19970 G68rr P8LS POSPO TCITO 8Shh PCOS COSPO PL8IO Lé6th DII9 9PtrPO posto
€99P ILSS LO8EO0 I10P7T0 ESLb 6899 9690 6P61'0 6rSr B88LS 6I7TVO 80070 OFS OF6S SPOrVO I09I0 Ohh 6119 666£€0 pCELO
Tbr TI9S 60SE€0 6EI7TO 08h
8 8695 f£09E0 86810 vO9P T6LS 66660 6S810 £09F YXE6S COILED ErhIO 60SPh ScI9 6rSe€O IOIlTO
€0S
6 1696 I6EE0O0 YIZITO E88h EILS I0S€0 69810 $99b 86LS I08E€0 CELIO L69FP 66S OPPEO P8TIO BSLSP VEID LOTEO Sv600
ELIS TELS ILIZEO SSIZTO P86r EPplLS SLECO IP810 OO08P° 8185 -I6hS0- vLSIO —66Lb PL6S
~ -S8I1E0 96ITO -7TE9p PrI9 686C0- LS800.
LST LO8S CrOEO O077O %S67S vI109 I8S70 10070 P88rp B8E8S OSEEO YESITO VI6h O109 St6c0 YPIITO CLIV €S19 csscO 80800
OHS E709 99670 PLETO OSES VEIN OCCTO S6070 Slob ITL8S LOIEO Y9CSITO cSsOsS T1019 8cScO 68110 SELp OLI9 09970 SSLO00
Gers Izzose 19000 “LOSt0 CIrs OIES IZZVO° FIeTO €2eSe C1lC9 I200Q. ceLlO LLIS S819. SOCCO 6ccCEO .S'68P -9SC9 solaGe On)

L, sy yeudoIojoo Y[NuITspays] Ga0gn9Y} JOYS SOUT] SSOINVJY} 19]Ud9

JO 94} IQR} DATOAUT sjOa[qo
YIM [eIIDAds sOURIDaTJoI
IO) J0UR}IUISURI}
saamnd
JO AGAT
| “BIY) ‘((L"¢)Zasou) past] MOjaq9} JIOYS SOUT] dATOAUT sjda[qoYIM [eIJOOds IOULOOTJoI
IO) (JOUL}]TUISURI]
SOAIND
JO od]
‘7

779
 Table 1(3.8.2) Twenty-Seven Linearly Independent Spectral Reflectance Functions of
(Hypothetical) Black Objects Which are Metameric under CIE Standard Illuminant C
for the CIE 1931 Standard Observer (Wyszecki, 1958)

No. 450 nm 520 nm 600 nm

l 1.0 at 400 nm —0.024942 0.0024834 —0.0013165

2 1.0 at 410 nm —0.076313 0.0069882 —0.0038283
3 1.0 at 420 nm —0.28912 0.025181 —0.013477
4 1.0 at 430 nm —0.71059 0.052377 —0.027402
5 1.0 at 440 nm —0.96764 0.045355 —0.023781
6 1.0 at 460 nm —0.93258 —0.073952 0.037060
| 1.0 at 470 nm —0.71949 —0.17368 0.072058
8 1.0 at 480 nm —0.44830 —0.29436 0.090809
9 1.0 at 490 nm —0.24094 —0.42419 0.092148
10 1.0 at 500 nm —0.11855 —0.58715 0.083842
11 1.0 at 510 nm —0.046234 —0.79531 0.061471
12 1.0 at 530 nm 0.020643 —1.1423 —0.10570
13 1.0 at S540 nm 0.032539 —1.2146 —0.24726
14 1.0 at 550 nm 0.036672 —1.1797 —0.41860
ts) 1.0 at 560 nm 0.033744 — 1.0269 —0.60572
16 1.0 at S570 nm 0.026064 —(0.77591 —0.77965
17 1.0 at 580 nm 0.016230 —0.48391 —0.91651
18 1.0 at 590 nm 0.0070509 —0.20865 —0.99357
19 1.0 at 610 nm —0.0040730 0.12446 —0.93650
20 1.0 at 620 nm —0.0056672 0.16931 —0.79845
21 1.0 at 630 nm —0.0051740 0.15588 —0.60102
22 1.0 at 640 nm —0.0040473 0.12143 —0.41872
23 1.0 at 650 nm —0.0027519 0.082388 —0.26659
24 1.0 at 660 nm —0.0016466 0.049246 —0.15453
25 1.0 at 670 nm —0.00874309 0.026133 —0.080415
26 1.0 at 680 nm —0.00046118 0.013780 —0.041919
27 1.0 at 690 nm —0.00041286 0.012332 —0.037154

“Each reflectance function has only four reflectance values different from zero; three
are located at the fixed wavelengths 450, 520, and 620 nm, the fourth, which is always
equal to 1.0, varies its position in the spectrum from 400 to 690 nm in steps of 10 nm.

780
 67000'0 creloO'O
$v100°0
€$S00°0 $6S10°0
€7L10°0 ru
8P110°0
ClccOO—.
O16V00—
VSE800—
PVSLEAs
Cissimy
cearreu
97980
8I1P6TO0— 86LL70—
vl
PlBPULIS I9A19SqO
puv
— >
91100°0
crE00'0
L66c0°0
I8c10°0 616c0°0
€79€0'0 OSE88ISO0—
S8t10°0 629F10—
SCCtCLvpsc
19760°0— UVSETE0—
$9078°0 PcLO~TO—
€l O-
Or)
0

€6100°0
cO1CO'0
$9S00'0
6S$8P70°0
SPls0'0
$9£0'0 —PIGCOU—=
99S10°0 P8b77O0—
ES880'0—
660S1'°0—
—Lsciv
SCriuo
8Csst'o L8LHI'O—
cl
ALD [C6]

0
—
=~]
6¢S00'0
18100°0 01670°0
66£70'0
1$610°0 LOTOO
€clt0'0 0691'0—
890b20—
ES8010—
“ES6rO0—" SI9TyO— $72000'0
/8re'O
I]
OS
0} 4}
Joodsay

=
v6c10°0
¢c100°0
9S$€00°0
LLLZO'0
769700
£€1800°0
OPECO
1990 >FL888'0
980L10—
LELITO— CO86EO—" L80r0'0
Ol
0" 0-—"

Se “OP
SyoK[_YIM

01000°0 9IP00'O—
L£0000°0
tcQ00 PESTO
TEECOO— -9P0ZI'0—
€€h76'°0 LOSte0—” $¢990'°0
SsSOr0—*
6 800=—
JOUIKJaIA[

U—" Os?
d9uRDIYOY

4
"C8660
9L1000—
PpSSs0OO0—
Pri
SSSOO1ODCESUOLCricro—”
8Z8t6'0 POrSTO—” 909L0°0
8 co
FEO
O=- 0 0O—
=
~ *
sdUL}Da[JoySUOIUNY

8tr00'0—~ OCS61L0
SP96l'O-—
STSCr EP8TIO—
8O6L1-LILI6U SOIDTO= 9£L90°0
L E iSO'0=
LOW
JO

= =:
- SCOPOS 9
OL COECOO
OSIZTO—
99680 91
ECOCE C6160 €Pr9r0'0
L00'0—"-POTIOO— SSO
9
O—s*esGe
Ore
0 0='
E89
PLTPEO—
GIL00180! Srcrio0 SPLOO'0—
[e.Iedg

Si C00 OL:
POS O
==
UeAeg-AUaM],JeUOSOYLIOG

8
a Is9STO—
rSElO0—68578°0“EO8ES 668£0°0 FE0TO'0—™
(SS61

rl
“PIPEZSAM)

v POU
O83
=
eCICO
..969000—
Prs0960 16SLC cOVCO'O S8T10
3
Oz OFS
JUBUTUNTTT

0)

eLOLeo
Z6I000— r09L0 96900'0 13500
D

ra
U OS OF
(Z7'R"E)IT
prepueyg

696560 eerGrcl 8P7c00'°0 CE

I 100
0 0—
-DARAA
see
FIGVL

OOFOlPOCLOCPOrrSaesrO9POLYO8P06P00SO1SOCSOrsObsOSs09S
AID

wu 5009 0£9 099OL9089

OLSO8¢06S O190c9 Ov9OS) 069

781
 ‘solaz o}0uap sooeds Aidwiq -“aj0N
a ne
686660 069
rE000'0— #186660 089
99000°;0— ¥vL0000— 676660 OL9
Ler00 O-= ~~ EP1I0V0—~ PLEO O0— -LtL66'0 099
LIZ0OO'O— SPrz00'0— OLPOO'O0— +0600'0— 817660 0S9
6£€000— 7Z8E000— ££€L000— 6071I00— LtPz70';0— 0S086°0 0)42)
O8r000— ~7rsooo0— s8£0l00— L66100— OLPEDO— I19SS00— £68560 0¢9
97900 0— SOLOO';0— 7ZSEIOO0— 70970O0— I7StO'0— 6r7L00— Pvré60l'0— 67760 0c9
60L00'0— O00800';0— €€SI00— I166700— 67ISO'O0— O€780'0— 6EP710— O6E8I'O0— 068ZL8°0 O19
EILOO'0O— +08000— 7ZrSIOO0— OL6Z0'0— S9ISOO— 667800— ILS7TI'O0— CP98lO— Pv9E9TO— PLEEEO— IST8EO— Y9868E'0— SL8re0— 009
6£9000— IZL00;0— F8E100— 999700— £r9P00— 6LPLOO0— ZLEI1O— 096910— P8IPT7O0— CSPS8O — 06S
Z6r00 0— 9S5S5000— 890100— 79070 0— 66S£0':0— LZ8S0'0— 67680':0— 79PELO— 66761'0— O016970— 098¢8°0 O8¢
86Z000— LEEOO'O— 6P9000— LSZIOO— s807Z00— 8sI9E00— Pr9S0O0— S7L80'0— LLOEIO— £7107O0— Pp6r97TO— P8ES80 OLS
060000— €0100:0— 10Z00';0— 86€00';0— ZZL000— ZSZIO0— +rIIZ00— P09E0'0— PL090'0— ZIPZ1O— 99L81'0— L8ES~TO— 79980 09¢
v6000°0 ¢0100°0 96100°0 ¢9€00°0 70900'0 19800°0 PSOl0'0 $c0100 87£00'°0 OLOSOO— ES OOO aS o SoOe COD Gime oss
L7C00'0 ¢$c00'0 ¢8700'°0 07¢600°0 99¢S10°0 807700 68€£0 0 €Lpro'oO ILISO'0 79600'0 COCHO0Se CPL LVS COv0Go= Ors
voc00'0 1ve00'°0 0S900°0 6£710°0 €T170'0 O1e€c0'O CLLVO'O 9S$90°0 881800 0c1S0'0 L7800°0 LO TIO OSes CVO= oes
TLE00'0 60L00°0 csel00 6C£C0'0 €S9¢0'0 17eS0'0 8erl00 1LS60'0 9¢€SL0°0 Secv0'0 PSLIO0= > 189600= OS
IT€€00'0
cPE00'0 €$900°0 87cl00 671700 c8E£0'0 956700 76690°0 Ivl60°0 9S6L0°0- ¢v9S0'0 €9600'°0 crSso 0— OIs
vOtL00'0
vryc00 0 LL700'0 0£S00°0 vIO10'0 LVLio0 9SL70'0 vsSoro 0 CSLSO'0 Ov6L0°0 96690'°0 cOsS0'0 6£1c0'0 STLZO'O0— 00S
¢1700°0 01700'°0 98L00°0 cselo0 €v1coo0 £91600 T1ISPO'O 01090°0 118S0'0 c¢L6r0'0 97970 0 17800 0— 06
16100°0
ve 1000 0S 100°0 88700'0 cSS00'0 €$600°0 OISIO0 8£ceC0 0 11c¢£0'0 61¢€70'0 06¢£70'°0 966£0°0 Lv9cO'0 1S700°0 O8P
1Z000°0 9¢€100°0 797000 €Sp00'0 07L00'0 ¢L010°0 8SS100 O€170'0 9¢£70'0 O££70'0 66810°0 LvOl0'0 OLY
€9000°0
60000';0— 810000— £€£€000';0— 9S50000— 98000'0— 611000— 9S1000— IL100:0— 80000°0 ¢7C700'0 0£S00°0 9S800°0 09
80000 0—
IZ000;0— 9€100':0— 19Z000— OSPO00— ZILODO0— 6hP0100— 86r100— 066100— SI6100— FvI9IOO0— ch8000— 00€00°0 OST
€9000':0—
SOl0o0O';0— 10Z00';0— S8f000— 99000— ZS0100— 9SS10':0— O€7Z0';0— L8670'0—- 686700— ¥9970'O0— 79910'0— 00100°0— Orr
€6000':0—
7Z6000';0— 9ZLI00';0— s8f€f000— €8S000— +76000— 69€100— $9610'0— 8£9700— O0L9700— LIP~OO0— SLsl00— 9¢€700 0— OC
87000 0—
Ip000';0—_ 8L000';0— OSI000— 8Sz00';0— OI1P00'0— L0900'0— 7L8000— ZLIIOO0— €61100— 680100— SzLo0'0— O0F1000— WA 4
9£000°0—
I10000— 1IZ000;0— IP000';0— OL000'0— ZII000— S9I000— 8s£€Z7000— O7£E000— LZE00'0— 667200'°0— 10c00°0— €v000'0— Olv
01000 0—
ZLO000';0— +1000';0— +Z000':0— 8s£0000— 950000— 08000'°0— 8g0100'0— O1100'0— I101000— 89000':0— v1000°0— OO
€9000';0— +0000;0—
e
Re ee er
6 VC GC CC IZ OC 61 SI ja! oT Cl wu
LG 9¢
TY
SS. ys ud]
d0uPR}DIYoyY
-OARM
panunjuod —(7'S'E)11 AGEL

782
Table III(3.8.2) Spectral Reflectances (in Percent) of 12 (Hypothetical Gray) Objects
Yielding Metameric Object-Color Stimuli with Respect to the CIE 1931 Standard
Observer and CIE Standard Illuminant C (Wyszecki, 1962)

A(nm) 1 2 3 4 ‘ 6 7 8 9 10 11 12

380 5.75 4.37 16.13 11.48 2.43 25.50 60.61 21.90 27.06 24.11 37.49 10.42
390 20.17 10.81 17.70 11.70 4.14 29.19 60.17 22.08 29.45 31.58 37.81 11.01
A esS:1D ty Goan 21 73 123 elites 31750. 55.03 922.08), 30.96. 37-03 « 37°70. 11°68
AN” 44,1922 792579: 91257 SBIO~ 368° 46.11 22.25 31-72 40.26 37AS 1226
AN eel] bones S0mn 28-95 1G 10,14. 9314.9 37.48: 222.43. 132.02 ©4167 36:55 12.99
430 46.17 28.34 30.64 13.30 12.98 33.27 32.30 22.61 32.10 41.06 35.69 13.64
440 41.97 29.64 31.26 13.42 15.65 32.97 29.86 22.96 31.91 39.15 34.51 14.43
450 35.29 30.04 30.89 29.92 28.58 32.30 28.90 23.50 31.48 35.21 32.99 15.63
460 27.44 30.73 29.61 48.87 43.89 31.21 28.47 26.68 30.77 29.06 30.86 22.58
ATO 16.11 30.23 25,95. 47/0) 46370817 282513985" 29118 1933 27-11 “S199
ASOCT 8.510 930.7309 2350 13742 43;4le 24:78.) 28.16 452.83 26.85 Ue 9.60 121.84 81:21
490 634 32.85 34.26 26.07 39.83 20.12 28.12 49.80 24.08 4.94 16.14 75.39
500 6.37 35.35 43.21 28.74 37.96 16.34 28.12 40.23 21.71 6.44 11.97 54.57
510 6.53 37.08 44.78 44.96 36.49 12.10 28.11 35.28 18.54 28.78 7.37 37.87
520 6.64 37.69 42.69 48.52 35.55 9.54 28.30 34.54 15.93 58.60 4.33 27.99
5300) 6.0 63664 37-98. £41096 8.978 0.89 | 3029. 3071. 14.75 $7.52 5.53 22:96
5400.14.04" 433.5495,32:33 31993) “30.98 4011)" 33°35.4.28.778 23.157 84341 “I9N0 2K41
550 56.96 28.51 27.07 22.22 26.43 35.82 32.54 26.20 41.36 28.22 48.95 20.48
560 67.92 21.45 22.75 16.76 20.02 29.29 30.94 24.33 51.46 16.64 66.21 20.29
570 60.46 14.65 19.85 15.58 13.62 38.19 29.71 23.22 47.11 9.04 60.02 21.65
680 48.11) 228.90? 18112 $4893 V8.18- 59162". 28°95 422.539 97.231 492 “4536 24/85
$000035.13 904 sl 02,504. 2704 «S061. 28.494 2299 8 28:13. 94.67. 31.76) 28.39
600 23.60 68.54 16.47 29.79 70.78 34.29 28.22 22.06 21.43 67.08 21.61 32.08
BiOie 14,34: 958566 17.05 B5520 5050 205654 128210 =22.5)e 17:27456-61 «15.206 35:87
620 8.09 36.76 30.35 40.21 38.32 11.79 28.15 31.64 14.46 36.73 11.64 39.37
610s 4 2h IAS TS96 B44.01 2287 626 29:15 63.99 13.26 21°75 ~9'56 42.07
640 1. 8S2011.77 80.00 147.37, 4259 . 29h 3394 765.389°17.29- 412.00 ©9112" -44.77
65078 0:94 “5.89 464.92 50:29 6.29 1.28 35/68-? 55.13. 42.31-' 6:00 ©1871 47.07
660° 0.31 — 2.94 49.84 52:49 3.15 0.74 34.90" 44.88° 62.23) 3.00 \ 8:78 48.88
670° 0.16 ' 1.26 39.24 54.24 1.57 0.83 33.85 37.68 65.28 1.50 9.53 50.44
680°) 0.00 6 0.42a B17 A578) 2089) | dsl3, B3SAK 32. 15e 61.63, 0.75 41310 51,80
Co0me EOCoun 000m 26.944 0 5OMMaISh- 34.4300 78 S57 20. 0.37. 21:55" 52.98
Tie 0.59 Om 25.30 pesoul oe 2) 3539.38 96 33, 25265). 0,00 W323ide 53.81
710 1.88 0.61 24.97 58.85 0.00 2.31 45.56 24.67 48.26 0.00 38.86 54.55
Pee so 0.92 OAS 5558s 00 2a 50.62 soo, a2? 0 4 e 8Ode
73092409 Fool) Gk 923.94. 16024: e000 (2110 854281 A922:73i5 40.98; ~.0.00%-44.29, «56.02
740° 634 2.06. 23.29 60.89 0.00 . 2.02. 58.17 22-18 37.17%--40.00.. 44.745 56.69
750 7.28 2.37 22.97 61.48 0.00 1.89 59.99 21.90 34.36 0.00 44.95 57.30
760 8.21 2.68 23.05 61.99 0.00 1.81 58.59 21.90 31.37. 0.00 45.18 57.87

“Rach of the 12 objects defined in this table leads to the following tristimulus values
and chromaticity coordinates:
X= 29.41, Y = 30.00, Z = 35.43, x = 0.310, y=0.316

783
Table IV(3.8.2) Spectral Reflectances (in Percent) of 12 (Hypothetical Gray) Objects Yielding Metameric
Object-Color Stimuli with Respect to CIE Standard Illuminant D,. and the CIE 1931 Standard Observer“

r l 2 3 4 5 6 us 8 9 10 11 12

400 35.11 17.66. (21,742 12 Als = 6.062531, 502 =55 03-) 22 07).@ 350.965 337.045 371 oe
410 4422: 22.78 (258)e012 500 gc8 12083268e 46.bl= 22 255 eee Slee Sale 12ers
420 47574 25.87 42901 "= 192.807.529.990 31S" S7A82 2237-3202 Neal SSERS6o 12
430 46.44 28.29 30.74 12.87 12.50 33.30 32.31 22.48 32.10 41.37 35.73 13.27
440 42.41 29.57 31.39 12.74 14.93 33.04 29.87 22.76 31.92 39.60 34.60 13.88
450 35.84. 29.97 {31.00.6279 125 227 19-32 Al) ono oo) 20m OkWee 5.000 5) anes
460 28.04 30.68 29.64 48.07 43.18 31.36 2847 26.44 30.86 29.45 31.08 22.04
470 16.74 30.21 25.88 47.02 45.88 28.96 28.24 39.65 29.32 19.54 27.40 51.64
480 905 30:73 523°3 1936.85" *43 Tov 225: 0le 26-145 7. Gi Uae. 0m eee tone
490 687 32.86 (33.97) 925.646 239.2)" (2036. e286. 10s <49.08- 224 35a lot Goo ee
500 = 7.03 35.34 §=42.78 928.27 =38.01 16.70 28.09 40.11 22.12 5.98 12.61 54.74
5100 741 37.04 (44.18) (44.39 236.61" 12.59" 28.07- 35.157 19. 149 98.02" 825 =oe9
520) | 7:73. 39.59. 641,96) ©47.865 635.699 210,145 —28.25- 334.40> 016-7 1557.56 4 5:49 282
530 =8.01 3644 37.20 40.63 34.06 21.05 30.24 31.58 15.67 56.25 6.86 23.46
540 15.23) 33.25 31.64 30.55 30.96 40.72 33.32 28.68 24.08 42.10 20.40 21.89
590°: 5803 2B2 26537 21.625 P26 2718 36.388 —32 Sle B26.16 E47 Zoe 226.C1 Ssh 20
560 68.78. 20.96 22.44 ~16.315 —16.69" ~29.65" “30.95; "24.357 =92.29- (15:46 ~ 67:20" 20a
570 61.06 14.08 19.80 15.30 13.12 38.36 29.74 23.31 47.74 8.03 60.73 21.84
580 48.45 8.24 18.35 18.84 7.51 59.59 29.01 22.69 37.70 3.90 45.79 24.90
390 35:20 «233357 917.47) ©24.02> (26.578 50,42* ©28.357" 122.435 128.4 lee 324.10 9h3 1689 928222
600 23.48 67.825 (17.09 29.99 °69.92°33,97 728.3 "2231" 2158 66.06, 920)
610 T4109 37°90." 172725 35.477 58.61" 20.288 SEB Os B22 1G Fs 32° 256,35 "214s
620 7.84 36.22 30.95 40.47 37.62 11.44 28.23 31.86 1446 36.58 11.39 39.17
630 4.02 21.07 78.41 44.21 2232 600 29.21 6451 13.24 2167 9.35 41.91
640) heT3 ddodl 80.33 u¢ AT32 s9l2.24-)° 22 33.98 «6349 ng Li 27s 11.9, 8.96 ae as
650 0:85. da74 9f(65.1255550.38) 46.0855. dulOe 35.705 g99.20 842 290, 5.98.) fOlo ee
660 0.27 2.86 49.96 52.54 3.03 0.67 3491 4492 62.22 2.99 8.72 48.84
670 ‘O14 82 839.308 54.2705 Lee OF. 33.867 #370088 69.27 149 950 50.42
680. = 0.00 - 0740 95351 20F So BU 702 oe obs ean eoeHLOe Ons 0),feel seth eeeee
690: O01 ~9.00 (26.95, 25725.= O38. 7 1 BU) o34.432 28.625 57.20" 0.37. 2E a ee
700, “200° 020 025.310 8.12 E000 279 39oe BO pe 2.00. 00 ue sa) eee

“The data are derived from those given in Table III(3.8.2) by means of the method based on variation
calculus (Takahama and Nayatani, 1972).

7184
Table I(3.9) Formulae of the Kubelka—Munk Theory of Colorant Layers (Judd and Wyszecki, 1975)

Quantity to
be Computed Formula Remarks

R (R = reflectance of a layer with pea (R, — Rwo)/Rwo — Ro(R, — 1/Ro)e8%"/Ro-Ro) Exponential solution of
tsi of reflectance equal to ase a RIOR Ga Ro) REDE ED differential equations

Hyperbolic solution of
g

ee ele
pee~ ecto OSX)
a—R,+ bctgh bSX differential equations
ean X(1 --1R,) + R;z Special case:
Rix=0) = P
SX(1 — R,) + 1 =
pi ee SX Special case:
(K=R,=0) CU eEL 1 K = R, =.0

pa Roll = Ro) + Ri Rl = Ro)

LiiReRs
p a Rox Rol2aRy — 1)
iv 1— RR,
TR
= R aha
nes Se
s an a RR;

Ry (Ro = reflectance of a layer re |

with ideal black background, ° 7 a+b ctghbsX
R, = 0) R, =a — (T2 + 52)%

pa oss R—R
°" 1 — R,(2a — R)
ne S Special case: S small
°K + ctgh KX) compared with K

R, (R, = Reflectance of a layer wiih kg te Rah, (lee)

with ideal white background, Ss Rees)
R, = 1)

& Kite Ka KA Ro also called reflec-

Renee he =e pices Mar Ss ~(F ae =] tivity of layer ;
(Ro = Reflectance of a layer so
thick that further increase in thick- Ro = 2— (a?— 1)% (1—Roo)
ness fails to change the reflectance) Ro =a—b Inverse: SOR
co

(T= 1
Pca I tre itt
ransmittance o fo. Are
a PSNRie b rei
eR

Fah cen at
TA=(R— Ba) 7 Ro]
1

1 Special case: S small

Te eS compared with K
‘sinh KX + cosh KX

1 =— =
SX: SAC 5(47 ctgh a7 — Ar ctgh cae Sk

1 1— aR
SX =- Arctgh Rs
pee UE
= 5{4r sinh —= — Arsinh 6)

: :
es i S

_ 1/Ro
+ Ro
2

1/Ro
— Ro
b ae ener em
b= (a2
— 1)%

785
Table I(3.9)— Continued

Quantity to
be Computed Formula Remarks

S (S = coefficient of scatter, rate of

increase of reflectance with thickness oe dXx->0

of a very thin layer for diffusely

incident radiant flux)
K (K = coefficient of absorption, C= |
rate of decrease of transmittance ~ \dX) x-+0
with thickness ofa very thin layer
for diffusely incident radiant flux)

Note. sinh uw = abbreviation for (e“ — 1/e“)/2, hyperbolic sine of wu.

cosh uv = abbreviation for (e” + I/e“)/2, hyperbolic cosine of u
(sinh uw + cosh uv = e”%).
ctgh w = abbreviation for (e“ + I/e")/(e“ — 1/e"), hyperbolic cotangent ofw; ctgh w = (cosh w)/(sinh w).
Ar sinh, Ar cosh, Ar ctgh = inverse hyperbolic functions: if v= sinh wv, then « = Ar sinh v, and similarly for Ar cosh and
Ar ctgh.

786
Table 1(4.1) Basic Photometric Quantities”
Light
Luminous efficacy,
K, K(A), K
Photopic luminous efficiency
function V(A) (Standard Pho-
tometric Observer for Photopic
Vision)
Scotopic luminous efficiency
function V’(A) (Standard Pho-
tometric Observer for Scotopic
Vision)
Luminous flux, F,,, or luminous
power, P,,
Light is radiant energy evaluated with respect to its ability to
stimulate the sense of sight of a human observer.
The /uminous efficacy of radiant energy is the quotient of luminous
flux by the corresponding radiant flux. The symbol K represents the
luminous efficacy of any radiant flux, which may include contribu-
tions of any or all wavelengths. The symbol K(A) represents
luminous efficacy of monochromatic radiant flux of wavelength A.
The symbol K,, represents the maximum luminous efficacy of
monochromatic radiant flux that will be obtained at the wavelength
A=A,, at which V(A) =1; K,,, is equal to 683 lumen per watt.
The photopic luminous efficiency function gives the ratio of the radiant
flux at wavelength A,, to that at wavelength A, when the two fluxes
produce the same photopic luminous sensations under specified
photometric conditions, A,,, being chosen so that the maximum value
of this ratio is unity (A,, =555 nm).
Unless otherwise indicated, the values used for the relative
luminous efficiency function relate to photopic vision by the
standard photometric observer having the characteristics laid down
by the CIE.
The scotopic luminous efficiency function gives the ratio of the radiant
flux at wavelength A,,, to that at wavelength A, when the two fluxes
produce the same scotopic luminous sensations under specified
photometric conditions, A,,, being chosen so that the maximum
value of this ratio is unity ee = 507 nm).
Unless otherwise indicated, the values used for the relative
luminous efficiency function relate to scotopic vision by the
standard photometric observer having the characteristics laid down
by the CIE.
Luminous flux (luminous power) is the quantity derived from radiant
flux (radiant power) by evaluating the radiant energy according to
its action upon a selective receptor, the spectral sensitivity of which
is defined by a standard luminous efficiency function.
Unless otherwise indicated, the luminous flux relates to photopic
vision, and is connected with the radiant flux by the following
formula adopted by the CIE:
F, (ot P,)= KymfPe.V(A)dd
v
Here P,,dA is the radiant flux emitted in the wavelength interval
dX containing the wavelength A, and V(A) is the photopic luminous
efficiency function. The factor K,, is the maximum luminous
efficacy corresponding to the wavelength for which V(A) =I.
The unit of luminous flux is the /umen (1m).
787
Table I(4.1)— Continued

Luminous intensity, The /uminous intensity in a given direction is the quotient of the
Ivo luminous flux emitted by a point source in an infinitesimal cone
containing the given direction, by the solid angle of that cone.
The unit of luminous intensity is the candela (cd).
Luminance, L,, The /uminance at a point of a surface and in a given direction is the
quotient of the luminous intensity in the given direction of an
infinitesimal element of the surface containing the point under
consideration, by the orthogonally projected area of the surface
element on a plane perpendicular to the given direction.
The unit of luminance is cd-m7?.
Illuminance, E,, The illuminance at a point of a surface is the quotient of the
luminous flux incident on an infinitesimal element of the surface
containing the point under consideration, by the area of that surface
element.
The unit of illuminance is Im-m’.
Luminous emittance, The /uminous emittance from a point of a surface is the quotient of
MD the luminous flux emitted from an infinitesimal element of the
surface containing the point under consideration, by the area of that
surface element. The unit of luminance is lm-m”.

“Note: The above photometric quantities relate to photopic vision, that is, involve the photopic luminous
efficiency function V(A). Analogous scotopic photometric quantities are defined by making use of the
scotopic luminous efficiency function V’(A). The symbols for the scotopic photometric quantities are the
same as those for the photopic quantities, but a prime is added (e.g., scotopic luminous flux, F’). The
maximum scotopic luminous efficacy occurring at \’,, = 507 nm (approx.) is equal to 1700 Im-W_!.

788
Table I(4.3.2) Values of Standard Spectral Luminous Efficiency Function V’(A) for Scotopic Vision at
One-Nanometer Intervals
Wavelength Wavelength Wavelength Wavelength
d (nm) V(r) A(nm) —-V"(A) A(nm) =V"(A) A(nm) V(X)
380 5.09 4° 10," 430 0.1998 480 0.793 530 0.811
I 6.65 1 0.2119 1 0.805 1 0.796
2 v.52 2 0.2243 2 0.817 2 0.781
3 8.54 3 0.2369 3 0.828 0.765
4 9.72 a 0.2496 4 0.840 4 0.749

385 1.108% 10-5 435 0.2625 485 0.851 535 0.733

6 1.268 6 0.2755 6 0.862 6 0.717
7 1.453 ‘i 0.2886 7 0.873 1 0.700
8 1.668 8 0.3017 8 0.884 8 0.683
9 1.918 9 0.3149 9 0.894 9 0.667

390 2.209 10-* 440 0.3281 490 0.904 540 0.650

1 2.547 I 0.3412 1 0.914 1 0.633
2 2.939 2 0.3543 2 0.923 2 0.616
3 3.394 3 0.3673 3 0.932 3 0.599
4 3.921 4 0.3803 4 0.941 4 0.581

395 4.53 9° 10-* 445 0.3931 495 0.949 545 0.564

6 5.24 6 0.406 6 0.957 6 0.548
7 6.05 7 0.418 i) 0.964 f 0.531
8 6.98 8 0.431 8 0.970 8 0.514
9 8.06 7 0.443 9 0.976 9 0.497

400 9.291 - 105? 450 0.455 500 0.982 550 0.481

1 1.070 - 1072 ys 0.467 1 0.986 1 0.465
2 1.231 2 0.479 2 0.990 2 0.448
3 1.413 3 0.490 5 0.994 3 0.433
4 1.619 4 0.502 4 0.997 4 0.417

405 1.852-10- 455 0.513 505 0.998 555 0.402

6 rs IB 6 0.524 6 1.000 6 0.3864
7 2.405 3 0.535 7 1.000 7 0.3715
8 2.730 8 0.546 8 1.000 8 0.3569
9 3.089 9 0.557 2 0.998 9 0.3427

410 3.484- 10? 460 0.567 510 0.997 560 0.3288

| 3.916 I 0.578 | 0.994 1 0.3151
2 4.39 = 0.588 2 0.990 z 0.3018
3 4.90 3 0.599 3 0.986 3 0.2888
+ 5.45 4 0.610 = 0.981 4 0.2762

415 6.04 -10-? 465 0.620 515 0.975 565 0.2639

6 6.68 6 0.631 6 0.968 6 0.2519
‘i 7.36 7 0.642 7 0.961 7 0.2403
8 8.08 8 0.653 8 0.953 8 0.2291
9 8.85 9 0.664 9 0.944 3 0.2182

420 9.66 5 10>* 470 0.676 520 0.935 570 0.2076

1 1052-2107 * 1 0.687 1 0.925 1 0.1974
2 1.141 2 0.699 2 0.915 2 0.1876
3 1.235 3 0.710 3 0.904 3 0.1782
4 1.334 4 0.722 4 0.892 4 0.1690

425 1.436: 107? 475 0.734 525 0.880 ai5 0.1602

6 1.541 6 0.745 6 0.867 6 0.1517
‘A 1.651 bi 0.757 iy 0.854 7 0.1436
8 1.764 8 0.769 8 0.840 8 0.1358
4) 1.879 9 0.781 9 0.826 9 0.1284

789
Table 1(4.3.2)— Continued

Wavelength Wavelength Wavelength Wavelength

A (nm) V(r) A (nm) V(X) A (nm) v"(A) A (nm) V"(X)
580 1.212- 10-1 630 3.335 - 10-3
680 7.15 - 10-5
730 2.546 - 10-6

1 1.143 3.079 1 6.66 2.393

1.078 2.842 6.20 2.250
1.015 2.623 5.78 eB t
9.56 - 10-2
2.421 5.38 1.989

8.99 - 10-2 2235; 5.01 - 10-5

1.870- 1
8.45 2.062 4.67 1.759
Fi 1.903 4.36 1.655
7.45 1.757 4.06 1.557
6.99 1.621 3.789 1.466

6.55 - 10-2
1.497 - 3.5335 10-° isi 1o-5
6.13 1.382 3.292 1.299
5.74 1.276 3.075 1.223
337 1.178 2.870 1.151
5.02 1.088 2.679 1.084

4.69 - 10-2
1.005 - 10-8 2.501 -10->
1.022 - 10-6
4.38 9.28 a 2 2.336 9.62 - 10-7
4.09 8.57 2.182 9.07
3.816 7.92 2.038 8.55
3.558 7.32 1.905 8.06

S343 - 10-2
6.77 - 10-4
1.780 - 7.60 -
3.087 6.26 1.664 7.16
2.874 5.79 1.556 6.75
2.674 5.36 1.454 6.37
2.487 4.96 1.360 6.01

2512+ 10-2 4.59 - 10-4

i273 10-° 5.67
2.147 4.25 1.191 5.35
1.994 3.935 1.114 5.05
1.851 3.645 1.043 4.77
1.718 3.37% 5.76 §- 4.50

[593 ew PA ee 9.14>- 4.25

1.477 2.901 8.56 4.01
1.369 2.689 8.02 3.790
1.269 2.493 fe 3.580
1175 aD) 7.04 3.382-

1.088 - 10-2 2.146: 6.60 - 3.196- 10-7

1.007 1.99] 6.18 3.021
9.32 - 10-3 1.848 5.80 2.855
8.62 1.716 5.44 2.699
1391 1.593 5.10 2.552

teat 1.480 - ns 4.78 2.413 -

6.82 1.375 4.49 2.282
6.30 1:27 4.21 2.159
5.82 1.187 3-091 2.042
5.38 1.104 3.709 1.932

4.97 - 10-3
1.026 - 1G 3.482 - 10-° 1.829 - 1o~*
4.59 9.54 - 10-5 3.270 1.731
4.24 8.88 3.070 1.638
3.913 8.26 2.884 1.551
3.613 7.69 2.710 1.468

780 1.390 - 10-7

790
Table 1(4.4) Calculation of Illuminance Produced by Lambert
Sources of Different Shapes
SOURCE ILLUMINANCE (E) NOTES

Inverse square law.

F = Luminous flux
L= Luminance
1 = Luminous intensity

RECEIVER
ii)Special case:
p= = SPHERE

2
pate poet = 7L sin*a
R cee Ay. or R +d

d a i
c

SPHERE
Sphere is equivalent
to point source of
luminous intensity
I=7LR* situated
at its center

I= luminous intensity
of source per unit
length, i.e. cd/m,
I= F/L 12
@,,@, =angles measured
in radians.

OH
 Table 1(4.4.)— Continued

SOURCE ILLUMINANCE (E) NOTES

STRAIGHT FLAT STRIP

dA»

length: 2L; width: 2a

a<«<d

RECTANGULAR AREA
i) Center (P)of receiver
element (dAz) opposite Horizontal illuminance:
one corner of source: AE L (a-ycos B) dAz at P in horizontal
2 plane.

SL(ysinB+8 sin a) Vertical illuminance:

dA2 at P in vertical
plane, parallel to plane
of source.

ii) Projected point Pin- Divide rectangle [ABCD] into

side rectangle of the four rectangular areas
source: (P'A] ,[P'8], [P'c], [P'D], and
compute for each case E,,
and E,as under i) and add
results. In finding E,, neglect
areas [P'D] and [P'C] since
their contributions to E,on
the upper side of dA» atP
is equal to zero.

iii) Projected point P Divide plane of source into

outside rectangle four rectangular areas [P'‘Al,
of source: [P'B] ,[P'C],[P'D] and compute
for each case E, and Ey
as under i). The resultant
illuminations are found
from the relation:
[P'a]+[P'c]-[P'B]-[P'D] = [ABCD]

792
 723
oT OY OL Ol Lis LE NOd
60 aS 9¢ Cl 80I VC cS SVs 06 WI1d
8°0 8°T CL C vCl C= 9C (Qu Dri ' v60 0 c6C vcv 0 80£°0 arum 9
al OE c9 8Te Sc cS NOd
Ont Ove SI 0 09 Ol— OV Ra 8c WI1d
cel iW! Ce a4 v6 9—- OSs Sa CV SOl 0 CEES 97r 0 Cc6v 0 aruMm ¢
66 EXE os €9 OSs 08 NOd
OT 8°~ VC Va O£ Ol 06 pe— 8C WI1d
OT Exe ce 9 SL Vag OL Ota CV 8800 Ce 8170 OLE 0 arumMm v
eal re OL Ol bee cs NOd
81 eG 8S (6 9¢€ 9—- SI s= VC W1d
IT ac L9e CAs COT Vie 9C (Gar v9 9€10 O8'Y c0Os 0 p9E 0 arumMm €
il IT €v Cv 9T— 8E NOd
vl 6C SC 0 09 Ell CV no SI WId
Soil BiG OV Ol v8 cls CV GC DE CSO 08'S OfC£'O CCV 0 arumMm G
cl GAS VC (4!) eC OC NDOd
cal OE 6S I Ol= VE o> OV Cl 91 WId
9'T OE ILI C= v9 c= (G4 4 al 9710 OVS Ive'0 08S 0 alruM if
9e01 0] (Sacidap)g tS £89 £29 ag alg 1g 07 Nol 7°70 = 9 0x Of} Oy JOAIISGQ. =JoquUINN
10} /V = 0 SJUDIOYJIOD dd (sj19quuey-}00} BUIPIOOD
Sd} ) 9x ur
UOT}9ag ssoID
Jeondi[yA JoUuBUIWUN Aonewoiy)

19}udz
1O]OD
(punoLing yieqg ‘suOyRAIaSq<Q 99.1d39p-0M J)
0 =/V 10} suoydIeg sso [eondyY pue sovdg-(7‘4 ‘x) ur sprosdyy SurysjeyAl-10j}0D (GPE) Wepyoey—uUMOIg (Z'p'S)I FIGBL
Li be Lt 69 Soo 9¢ NOd
S'l Le Sl v cs ols oP = vl W110
rt O'€ 6C (Ss 8V ‘aes ras 14 Sse Of 900°0 LO'| L9TO § P8r'0 aruM SI
ign oo,e IC 89 5 oe LI NOd
60 O¢ (44 y= VY (i vil 8c VE WI1d
VT 61 y9l cl 09 OTSe c8 oi cE 9L0°0 LY LI¢O0 899°0 aruM vl
$0 ST 69 Ss UE t> OO£ NOd
Ak am | oP 8c vel 81 O0O0E stc— rats W1d
9 Oigae tal oP 0 vol vl vrl v6o— 8S] cS0'0—- Ss'0 8710 Icc 0 aruM tl
LS A ie re cs 001 Cus SSI NOd
£0. 6d ct OV 911 OsS— 8el oL = OL WI1d
| neae 6€ 0 96 Ol= 88 62 hes c9 010: 0—- 68°0 te OS evs.0) arum al
Lull ev 61 VL vo— vl NOd
or ae E PLI om ve (Bes cel Ol 9€ W110
$0." oT I 8 09 btSec VSI (Sage 8E 8S1°0 0c 9 c6c0 ~~ 0590 arumM II
60 81 6 (atk Re SI NOd
cS el OSI fio 0c 2 Maa 9LI v9 9S W1d
TOR tt 9SI 4 vv sc VIC OL 06 LS10 O19 COCO cso0 aruMm 6
.
1 £9 = eT ge N9Od
W1d
St ak 06 CC vel (a 0c 0 oF €c1-0 Ol'r I8s'0 = 61 0 aruM 8
el Xe 89 Ic Ei= 9¢ NOd
OT Ev € Ol 907 Ova OL Ca Of W1d
VT ol 6v 9 88 Vig 9S cc co cs1 0 08'S OsrO 9bPO aruM L
GOI M01 (Soo1dap) 9 eg eg ey wey oly Wy 07 TZol 770 = 4 ox of Ox JOAIISQQ. JoquINN
0 = /1V 10J SJUDIOYJOOD SIO (SjJaquiey-}Ooj ul 9X ) Sd] PUIPIOOD
uoI}9ag ssoID yeondiyTy soueuIWIN | Ayonewoiry)
J9]U9D IOJOD
panunuod —(7'P'S)I AGEL

794
 795
NOd
WI1d
aruM
vol cE

89c0

IC
NOd
WI1d
NAN

OL
eS

0Sct
|$6
8Sc0 afruMm

OC
96
NOd
mm Oo
A ci a

8L W1d
c60 voc0 arumMm

61
v9
NOd
98 W1d

09'P
Gor
es css'0 aruM 81
NOd

cox
WI1d

06£°0
cL c8
88°C gcs'0 aqruMm LI

N
NOd
vec I= WI1d
91

Oe
OIC 99¢°0 9Ev'0 aruM

8
796
AGEL —(TP'S)T panuiuo)

1OJOD
19}U9,)
Ayonewoiy) JOUBUIUIN] JeondiySsOID UOI}N9g
BUIPIOOD
So} ) 9X ur (S}iaquuey-J00] INOW SJUSIOLYIOD 10] iV 0=

JOQuUINN,
= JOAIISGQ
Oy 07 ox = 770 TZo]07 ug a1g wey 2g ety *tS (Seoidep)g
01 9,01

CC aruMm POs” S360. ssl vso0 c9 sc 8C Co= O91 (44 19 8°C ia

WI1d 96 ee 08 Co col Ol Is S020
NOd trl o3— $9 LS a3 EY)

EC aruM “S570 SSc0. cel vc00 vil cL Of (ae 801 9¢ 69 €Sc¢ SO

WI1d 86 ae ey VV MiesNU vL CC 09 v7 60
NOd srl Nase
oti) OV 99 alc > 20

VC aruM €8£0 68r0 Oe'L cLI0 OL 0 Of oo= ccl C 06 81 GI

WId vv ame Ge r= VII 8 8L OC St
NOd 8S i= ct vl ce el

Sc aruM GECoO S580 9 0c 8S10 8V C Ol r= c6 9 £6 Cf vil

W1d 8V oa| 81 9 OIC el €8 as vil
NOd Iv $= 4 8L °3 cl

9¢ aruM vsSte0 8S€0 I¢9 091°0 OV $67 (44 re 961 Cc ev Gt LA

W1d 08 so v8 ae 89C cs ev ins 80
NOd 6 Fs 8c 9¢ VE ct

Le arum i9tvi0 95G0. 6l'V 0 8c1 vv so— 8S <—s v9 9 OV 81 Bi

W1d VC vce 8t Ol= 9¢ 9 Se ce vi
NOd IS sta v9 Or gee Gis

8c aruMm g61910 LIGO 81°9 se) te 2) SI st 98 8 oil 481 60°

W1d vs OV 8L 9s- O€T bo— trl re Ol
NOd 81 oe 89 Ic a el

6C afrum =the LIO LOL 9070 OS t= 88 0c— vrl oc Lc oi Ol

WI1d ve ooo 89 91 vol a= 81 Ee td
NOd Of 9t= 9¢ ce ec a
 Lé

S’s
Oe ara OC

Se
0's

ie

£8

SL
t< 61 we Lk Big be yf 67 Vs L's

(as
Se 69 69 L6 06 €8
SL

(44 cs 8L 19 69 69 86 86

19
cOl
C6
IL

c9
8f
ve
a +
vY vl

VVC vrl 08
SIT oot VET8Pe (G43 991 9L1 8ET vit VICvsl

vo—
of
Sas ef
Ol—a
Et

rN
VC

CROs tA
|
8c0'0
ISlO

891°0

€Il0 810 £c0'0

101°0—
nie

Oote 1l6v ts 969 |ae 18e

cv9'0
L9E'0

L9EO

00€c 160 00r'0 0vcs

8St'0

S80

Ice'0 0v6c 661°0

W1d
aruMm
aru

aruM
WI1d

W1d

WId
NOd

NOd

NOd

NOd

afruMm
W1dNOd aqruMmNOd
WI1d arum NOd
aqruM
wid

St
It

Of et ve 9¢

797
 ‘gouURUIUUN] UT AJUO JaYIP PUB SOT}IOIVBWIOIYS [edI}USp!
DABY EF PUL ZE SIO[OD ‘OJ Joquinu ou si soy} SWepYyor pur uUMOIg Aq UdAId dsOU} 0} puodsa1109 9[qe} SIY} UT USAT SIOJOS dy} JO SIaquINU sy, “fF AON
-}X9} JY} UI UDAIS UOISSNOSIP 94} OS|e 9aS ‘sprosdi|]a a4} JO SazIs oy}
Suipresoy ‘(Z7S6]) UMOI YIM DdULPIODIR UI _P9}991I09 INQ (6p) WepYyoe| pur UMOIG Kq uaaiZ asou) ale 91q¥} SI} UI UdAIS G ‘v puRs.“SoyT “€ AION
((Up'S) TZ AIL OF UoHeNsNy][! Jasur sas ‘astmyYDoPONUB Psinseaiu) g g[3ue oy} Aq uaais SI sIxe-2 94} JO UOTDaIIP ATISOd 9Y} 0} yoodsor YIM
sxe sofewl ay} Jo UoNeUTpOUT ay) {AJaATdadsor gq pue v Aq UaAId a1e UON}DES SSOID [eOI}IT]a BY} JO SAXBIUIAS IOUT Pu rofeur ay} Jo syysus] ou, “7 «AION
Op Penry he ny
ye — ¢ = ay ary ‘| = py ey 87 + Aiv)es+ IV hy 87 + fhiv)*s+ iy xy "87 + (ay)"F hq uaaid st sprosdija ay} Jo uonenbs sy, “7 a0N
s
a gaa lr e
Be L6 OV NOd
0c Ol

Ol v
tc 00! vL W1d
ioe)

6¢

Q9-—
of 98 96 vro0 Hb 6c£ 0 1610 aramM
0

8
16 Sel NOd
v0 60

8'T
HS 68 OLY W1d
180°0 ISs°c 6810 IL1V0 afruMm 8e
9°0

ard C6 VC 81 Ee
VY 101 9C NOd
ie | 61

cv 901 ve WI1d
6€ LOI 98 0v0'0 yLs'l 6IvO 070 afruMm at

VC
vl
v8
£39 wey 07 TZo, 770 = oF Of 0, JaAJasqQ.JaquINN,
EL?
ug
9201

DQ] (S92189P) 4
SJUSIOYJIOD I3| (sjaquie]-}O0} ul °{ ) $a} PUIPIOOD
Or
493
1V
uoloag ssorD g0ueuIwIN ] Ayonewoiyy
JeondTy
19}U97) 10]OD
oy a a ee ec Ee, ee ee aeSaerceeen EN ee araeeaeai
panunuod —(TP'S)I AGL

798
 799
roy Ye 6t°€ 98 0 ae Preece. ES 9L'8 1208 Co VY CC x TT
vOEe Svs tcl 6S °C EXC LOE a €7'S P8'~ Page VY €8v'0 c7r0 IT
80° O£'C $9 68'9-— LIAS Loe 8CIE v0 97— POL xOT
esl 86E €8 68°7 09°01 YO Si OL'9 SSI oa OL'6¢ as 8L1'0 8 OrE'O Ol
vl'0 COC LG vO'8— [Cree (6, Boo CELE Sy Wal Sie tOPELT *6
68°0 Los 6L Roe I8'Tl 9S'I Loe] Olece=— VOCCl ve LtloO 6070 6
0L0 v8 691 STf 88°91 t6L Co eG! IL'1€ IC Bt +8
c8'0 81°C OLI SII SE8 6 II 8S Er C8 IC pe Sc es 80€0 ¥l9°0 8
cO'T C8 I cil Os'Il OLS we 97 60 EVUC Sc L8 *L
SLI 00'€ Oll 9, ig 6°91 a2 99°¢ 1 OIL 910€ VY 9SV0 8 Sc0 L
SOc Ses vol LOY 00°17 9190 v9L 99°P SLICE «9
6t°C 90°9 cOl 18°C CGcl Se0 Cte Sit C891 Os 0f90 8870 9
860 06'1 v SCO Vole A See 9°'POI Does RC 8C *S
6r'l v9T ILI Ga ioe OV Ol tg OILY ec S$ LE Gl Is 89¢0 8LSr'0 ¢
Saaa | 06°C 691 roeWi Ce ILI OTP C£9OV ps9 60'¢1 *V
06° Sse 19] f= v9°9 19°C SYST Sr29 168 Eg Tivo 9£€°9°0 4
Tel 8CC eV 9L1°0 S8°VC LL 9S OV ool OS LE xf
A Be | ore tv 8V'0 09°L | LOC Pac = 66 61 a 6170 9770 £
LSI tLe 88 6E°S y8'9l VITO EGe Lone SL‘OV *C
60°C Ors v8 AES ely 8L'0 I8€ cL 61° ET v9 c6r0 6££°0 ra
80'1 LEC Le LG Ge SIT eC rE LOCLE FeLC— 6S°L8 «I
6¢ 1 COE v9 c60— CLE S10 vy si €e9I— LI vy ov 08c0 80£€0 i
9,01 POI 0 ee og Pg a3 as, mS oi fi % = aqUINN
(soo139p) O
(,01 X) S}USIDYJJa0D 9.9 (1-15) S$9}8UIPIOOD
0= AV 10J souRUIWIN'] Aytonewoiy9
WO1}99g SSOID
19\UdD) IOJOD
(punoLing AA aj
‘SUOBAIBSGQ) 99189P-UF]) UMOIg (LO6T) Aq
‘sI9AIBSGO pa}daT][OD AAJOM], SUIYIVIA[-1OJOD10J
Sprosdy|q (Z'p'S)IT asv19aAy AGeL
 panuyuo) —(7P'S)IT A192.

800
IOJOD
Ayionewoiys soueUullun’] io} 0=AZT

J9\Ud.)
SSOID
CUIPIOOD
$3} (1-4) I2W, S}UsI91JJ90)
X) (,01

UOI}99g
eee gene (saaidap)
JaquinnOx Of; LF 1g og seg ey 3 ee te 0 DOI 9-01
cl. c970 eveO0 LY C8LE ee Of S8 —5 96 96°0¢ C67 18 98°1 6ST
«Cl 69°8E OE SELE ee Oe COLE 880 LV ssl Ly'l

el 89£'0 cSt0 ee, cO'9C VIS 8S V6 is el lo 8V'C ILI 661 Oe

«tT CE 8C 08ST S188 08°SI— PV C6 890°0— sol 68'1 vol

eI ed) 0. TROL PSE EQ 6l Iv 0— S£0 SL’6 — og, 9L°8 09°T 0 Oc'e LEG
(.Z)e€1 LS‘0¢ OLS 33 5 Orol— I8°€¢ hoes
1 6S eee 89'T

vl £F9°0 voro Uv £1 66 lov Ol'801 61°6 SL'8 ed AN feat aye 96°0

«Vl sl se 90°01 VC! SO css 00'LI 00°c— SLI CVC 06°0

SI S8t'0 9PS0 LY 9S°6 861 6L'S 60°1— c8S 0= 9. ell 88t 6£€
«SI C96 coe p6'L l— sO SC VI beesMN rae 8cl 7 60 88°C

oh OLC0 6rS0 ES 961 CLL 69°6 veo LE ea 881 scl EL. toc
+91 COLI | cO'CI ee (8 €S°8C €99°0 Lcl IS'y voc

im! 80€°0 LO€O 09 98°9E 106— 6t SV OC? il ev Oe IV | 93 orl

«LI LO9€ 8— v6 v9'CY cel Ge LY reoU St vs evi

81 92950 cscQ Or 0¢ 6 — 31 ob 9126 Oly ol pr ke Be 1B el 6V'C 00°

«81 OE O€ sce tL 6cl Sv 69°6°0 v6LC ihepo) vl SOT 98°0

61 te.) ereUP 207 gs 6 VC 6c'S 9 eo $80 I 60°S oP 8cl S&.9 vlc

*61 11 8e 9.49 61°8 col 90°SS c$9'0 8cl $09 SVC

0c 96TO 0 foc 67 8 Iv 93 OSCE $60 LCC 81°C _ SEC em

OC* 81 6f C— 66° 8S°SP ea Go 1 LOSS Ori ? 96-¢ gr

LG 1910 0 OIv iS FO cel LEC S10 6l cy 817 vsl ve £61

«IC Or'SI 98°01 OLE Valse 9E°LP 6S°C 8ST vo’ gs'l

(a6 0 Ese GeO. ne 6L°91 61—19 OS €c Vou 08°0V oe ee Sc Cot Oe

*CC 8L VC or—Sr 91°¢9 98°81 vI9s OP'S OV 18°C COT

aon*/ ayy, uonenbs

Jo ayy sprosdija
st uaaldXq (ay)"S
+ 87 wy fy + (Ay)*S
+ 37 fy IV + S QV)"
+ 87 vy py “T= ary
/ = ZO SOL‘A
29 = 2 — x fy
= fh — RR 1V
= ‘7-1
AION
‘7 BY, SsO1D UONDES JO} JV = Q SI ue asdi[a asoym JofeLUpure Jou saxeluas
asev pue‘g “AjaAtoodsa
ayy, afdue
g SoAIdoY} UOTeUT|SUT
JO 94}
JofewsIxe YIM yodsaz
0} ay} dATpIsod UOTEIIP
JO dy} ‘sIxe-x Poinseoul OSIMYDOJINUe
99S) JaSUT UOIeIYSNIII
Ul AGLZ (Up's)
J/ON
“¢ JOY YORA JO[OD9} ISI MOL S2AIS SaN[eA PSUILIGO
WOOL] JYSICIIS SABRIOAV
JO dU} PIAIISGO
{IEP JY} PUODASMOI (YSLI9}SV)
SIAIZ SaN[eA POUIe}QO
WO] PaIYysioM SadeIDAR
JO dU} PAAIaSqO
“LIP 104 S[IeJoP
UO JY} ‘BUISLIDAR
998 "1X9}
 surpoep-10joD
sprosdyyy
puv NOY], SsOID SUOTIIG
JO SIdAIISGC
‘7H ‘YY pur AAD poulviqd
Aq rpezs{jy
pue Joppealy
(IL6T)
¢€)x .9 ewaedig
Ply WIM OF ayy, (punoung

P1421
(€y'S)I
38 UaINyJao_
X) (OL
JOON
‘P.SGO

IOjOD
SsOID
“ON "Sq... ox eh ane ug we ge 8 "8 (G@oap)g (,01x)> (.01x)¢g/m 'SoT
PY sjutog

U0T}Nag

»iajuay
I AD 98Lp0 LZEVO 6p SL?—- LIL rl 6Sp 6H rE 0's 67 ILI 1€€h— 6c
UV SPE 68€— 998 ZS 8h 6PI—
= 8 ‘Si L7 087 P6I'P— 8%
MD ¥L9 S69— 9LIL $6 €LET! OE SEy 59 72 66% MEr—- 09
Z AD 60660 SLEPO 069 8L9— 666 895—- €S8 OLT 6€ Ca Cx @ZE I6%r— O€
UV 969 PrE— 6Lh €S— 69% wa 19 67 O1Z IS%p— O€
MD ZI= — MOT ~—sdLSZs POI— CLES) TST)—C Sh 61 81 PLZ LSSr— O0€
€ AD te6rF0 ~9PEO 9rl €ZI LE9Z ISII— 8L91 SI@— LLI CL LI Ser p6fr— 6@
UV SOIl 8801— IS%Z Ih seo Lpl— I€ TP 91 soz 899b%— L@
MD SII OF I6bl 9P— I10Sl 9€1 6LI 78 ce LSE I€%r— O€
Pr AD 6LOE0 S6SEO 6661 7LS— vIOl blp— TZLEL 8%r— $9 TE 61 ELI 9@Lr-— O0€
UV HST OOI—-—Ss—s«éEL6Hs«BBL——Ss
899 = 6991S L9 9€ LI 917 I%Lp— O€
MD 8E6I 679— fPSOl 89%7— ISL 7@8E €9 €€ 61 9L'I 8ILr— 6S
AD 8SrrO0 6080 LOS SSE— TLE ob ZS AS aes,35 8°8 OF L8Z <SLOrP— 0€
UV 9€9 I16€— SIS L— 066 Ol 6b 9°9 87 EZ ECH-— O€
MD 769 €6€—- Ors I— 399 hl OS 09 BIZ LIZ O8%r— O0€
9 AD @ErO LIPEO 6LE 601 OS7Z I€S— I6€1 89€— LLL oP 61 97 @Sr— O0€
UV Z8OI 67E— ELSI SEE— 971 Sb L?@ 67 17 8€l IILb— O€
MOD @ES SL— SZEl Ell 689 9¢ ¢ 6€ v7 6S'1 €€Sb— LS
L AD @78E0 ZLEEO 869 60FV— ved ZI— Zl L8S— IZ 8°€ 17 I'l p09r— 67
UV ‘TLL 969-— Trrl €LZ 986 €87— 43 Sp O07 LEZ 60Sr— 0€
MOD S8II ~60I— pL07 = LO—-Sss*BEsdT'
0=— re vP LI OSZ 199h- 6t
=D ~S677O LIOL«L86Z0
«OSS—~—Ss_ Lol —~=SsECE 60L_—s—Ciéd'66H—C
86 re 5 LZ «OL! ehoP—
6 co
UV P91 ~7S9—- 856 861— 6ZIIl S€l 09 Le 61 0611 8r9r— 67
MOD 8291 80S— 6801 IpL— S69€ 19 6S TE 07 LSI OOLP— 6@
6 AD OIEEO 6PLZO Z6ZI Sh9- r8 97S— 601 Or cS cP lc E17 .StSr— 06
UV F181 VCE PRCOlS? 9 1S FOCOIE 39EPS. yf O'€ 91 I61 p7r— 68

801
MD ZISI 6S8— 9€6 68€— S08 OL pS (xs 61 657 9%br- 06
 PME FS) ponuijue)

‘ON
JO

802
Jojo) »1a}UaD 3 X)
UaDYJa0_5(,_O1 SsOID WONIG “P.SGO
ne ee Oa a Oe a ee ere ee
aeOn Dn ORcUO BE tg ug we ee 8 "8 Goep)g (,01x)> (01x)9/0 y"so
= symtog
ar
a A
uv 9Lb 96S1— 811 Frc ids Fel Bo. 8" fel — HOE [6¢sb— OE
0] S00L70) I) IP ,08Gm IST GISo— 296e OCE1 =. S761 Scr €'€ Ca ea lemme PaaS
MOD £619 IP8Z— SS8I €6S— 8681 16S1 co me HCI TOSE S8ier— Of

OS
85
BE
IZ
ee
UV 8pbl 6%— $99 601— 9011 € 99 0'P BT = EST — LOSS ES

PST
6981
CLEl
LPEIT
66P—
OLE—
—765—
psc
esc
MD SzPTI CbPe— FS8h ASL= O@KC--.6256
EL. vp LK 2601 §9LRP— 09

399—
atte 9¢
2ame
AZ
Poem
Ol
6'8
Olt
Olle
UV ooze ebl—" ORZCl 109% fbry cfL I 6 az = Tet Getty— ES

aiger
ame
avicrene
geco|—
SOs
(anUs¢
Ol
8L
Sc
Sb
ERIC
6L1
P8I
MOD
0°01
WESC
6S—
86szOl
eee!
pe eesSIC 2011 Ith ONE SLI 0C9, ESC SO OL t7 | yer. (meme

Yeo
SIQ0—
UV BLESc EHIS—y BRAS BEVI— 11TL 88h, OL as =A SITE B09Sip—B8
MOD KCREL BeLE— £u8Z% O0r— 86S O9EL= SL. 69 fez @167 H0Sy— BS
tal meee 0mmoc Us Ihcer O00 el GS Sxl l— occ eae 0G EO Lv tl tl: Uyak Dec
UV B89t7 GEIOI— IfOl SbGl— 19b6 H6IE C9 I'v EDI “6ET FRESP— 09
MOD 169p LZSI— 8901 LZI— 6% IZ OL 6'€ 281 RIZE Shisb— 09
asl ao 101E00 C91E _SIte sO16— 0821 ate mabh OG= Gra Ee atl 961) P= eco
UV 16zE 6POI— 86S1 SE 888 9€€ y9 L7 er 981 40t6—- 6S
MOD 90PE PSSI— IP9l 169— O€61 6€L 09 €"€ Or Orre — OS Oo
mena eee 17) CeCe PICe Tee SPLvh rop— cL hoe 1S (oad Sl . 8h1 166b~- O€-
UV BSdLZ ATES— pASOSI BOTT! ATOLL «OLY tT 97 OI LOT 988h— 09
MD RAILZ— C9S9T adhe ~EIvI— 2106 @BLS-7 QS aa AV OREC — $69 6c
eam] ooton) (LO ochcc — ale WUC .cis—= govol okcl- iJ. VT i Rae2 rs A a Sid
UV L8EZ 989— OIIl 7h SIS iG.” pom ee (Sty eee] obiey— Es
MOD S9E7 STI 197 =— 660 £891— BSE" LS 0'P eam
a isan) SANS
81 AD PrIEO SSEEO SBP O69I— IPhE. TEZ— SPL 98L1 96 $'1 ot] | 19 peeieks
UV cOboT Semele
ome purser oOl DOGlia Hae TE tie mscorlaqaue Vr eeu Ie
MO 9CHE 6SLI— SCSI 8l= 766 90€ 6S OV S| 2 8 ae Pe
ea |
 ‘-W- po 71 =X =°7
souvuluny & 0} 8uIpuodsas309 ‘9¢17'¢0 = °4 9' BOT ZC = 97 {sIOTOO 389} QZ [[e JOJ JULISUOD SI I9JU9D IO[OO ay} Jo 97 JJeUIPIOOS PITY} BY L,

803
gs STP «(OWT CS'E I's 0b 6 pL 661— POS LII— Ib MO
8% 8hkb— ~FlZ ONT cs 6€ 861 SIL €L— SIS Lrp— £19 UV
6 9b 861 Te vy 6 987— 976 LOP— 9ESI PLI- Sth 60670 8S0h0 AD 82
Scio) hs as Af Aa EL pl 69- SSL 8@Z £69 @I- LLI MO
6@ Seerp— LSI Te 8p 81 S6— LL9 LOI— 88L €hI— €8E UV
O0€ €6¢%r—- COZ 87 ie 691 I7Z— 965 SL7— 096 Tl S9% IED OS9O AD LZ
09 ~Sh—- 991 Sv I'v C0 OF SE nis BLT 1SZl —INS)
0€ SLor— 917% 07 ep tL 6€- «69h «=£9- SLL «OSH _—9*ET UV
67 809b—- P61 O07 6€ 9° WE— 86L IIE 166 pS9- O6bl I0LE0 O€L70 AD 9
cl trhOP—- «(OlZCOL'E g°L g gs— 08 ~«9- «= SS. BS— _—Ss*ODEL MO
O€ O1Zp- 6LI €F€ 6's EY weitif9 ni GalSSee SLM wilbOne Poa see LEE UV
0€ SNr—- 06% LZ 0'8 691 l— €1S he— 8001 ILI LSI OPlEO @7s0 AD Sz
O€ 6thr— OLE 07 19 jpn LE —<S9T= 61 96€1 PL8— LSB MOD
87 SESb— 9G1 = 7 ep If Lt S01 zI€— L416 96S— O9II UV
(eee Ps diel coe 0-0 y's OS vel— 9101 Z9I— LOO] 898— S8ZIl LO8E0 ESLEO AD v2
VTSa 4c en 0 A 8°9 78 «EOS «86 981 plI— $96 MO
0€¢ L9EP— SOE I7 a 7: 0 A ) A| UV
O0€ 9%Eh- SOT LT 9° tL «= @IZ— «ETL sPIZ~=—Sis«éOEE:;«SCSET— ~=—s«d6H—é«CMH'O-«OOLZO-s«AOD_—S—s EZ
6t vl0r— O€7Z LE rg g 80I— 896 I6I— €LS S9- 6lIl MD
SE —-R6Fib S81 -9'C 8p p Lt? Lev Ive O€Il 6r- SEE UV
O€ SOer— TZ LZ 6's Z SL— 1861 ISI— JLOIl 6%— €%% 9LLEO 87PSO AD
lL? 6Ilb— 9€€ LZ 06 IZ 6tZ— vEeOl IIZ 696 I€€—- £€@ MOD
8% LOPy— 891 = LZ 9p LZ p8-— sss oOl€ 116 6L7— 91S UV
67 S8rhr— O61 €E7 rp 9 p6T— Th6l MPT— CHI BOI- 80r P66EO EL8F0 AD IZ
6c wer— ILr 8il 9°8 BLL LSP Se LIRR TS, = SECC PHTOS SMURIOIZ MOD
O0€ 698h—- OTE 17 9°9 OG omer SrOlertiige ost ri |S esi UV
O0€ ler— B8ce 61 69 ch ok] ISL b— 96% 68r— 961 SZLPO 7%0E0 AD 2
LZ 998€- Wr TE Orl 9 Lhe SL8 LE—- 8 I8I— 69L MOD
LG——BBC—WBEE TT v'L 08 O9€ O€9 8S—- ZBI ~%he— 9951 UV
87 See OF &: eS £6 ae LOC 9cCV). 6L— LED Ch me SCS OS Co Ue CCPL Oe tL
804
SUOISSIG JUIIIJJIC] FOI], 10 OM], UI PIAIBSGE UOYIIG ssory Ady) pue sprosdyyY sumpoepA[-10jo = (p'p'S)I PAIGE
JO SUOISSAG JUSIZIIG] 9y) JO BIVG UBI, FIV JL SMOY UI polojUq Beg] MD pue Wy ‘AD Ssearasqg Aq
[((CLL6L “Aopp[ety pue ry9ezs{j\) UOeAIasq—)

paarasqg
IO[OD I9U9Z 3 s}uatpyyao_
x) OI (,_ SSOID WOT}DIG JOON

“ON UOISs9g x 04 11g clg cg tg t€3 £8 (Gap)g (,01X)2 (.01X)9

9/0 y"8o, s}UIog

Jaarlasqo
HD
6 I 0 OLE GLE’ SOC eS8TlgSiG—
yp Cll CLEse S17 4 is 61 LTC v— StS O€
G 9Z78I S68— v9Fl VILI— v86c Ov9 I¢ oe 81 LSI = Wy O€
€ Stet 6LE—
1 en$66 a L601 SLC TLE vS be Ex Srl 0f9v— Of
W Z6ZI SP9— bh 97S— 7601 cOv SS cv Ke E17? p— StS 06
I I S070 HHSTO PhZI 8ES— C9OI OvVE— 6SII O8— Ts oe 17 | "19 LISV Lé
c ECO LIP— eOlc:Se GPCL STO SLS—. 61 Oe 6'1 O09 8SLp— OG
W PEL
= o6r— Cocleli 69SheU6S
GLI 8E Ge Lx VS4 F— 399 €¢
el I -SPSE'D FSISKO LOCI
- AelOS— COlL—eRoloe
[9koe €L7 $9 L6 TT bry 6L1¢— Of
G ONG SCC 09 eOOl—9 OSE SI€ EL v9 87 C OF vras CSC Of
€ ePsl I8p— vey. LOZ 9871 077 69 vs ere 967 8tpy— 0€
W COLT eel 2908 PLI—
oe 989 €Sc 89 OL VT cle 6tohr, 06
a I OOMOIED LOvErtOlE
S9f== WISGl. —- [19099SSG Lo “ge vl “col == COOS O£
c ~ CRS? S9L—= OL6'9 SP LOS@S IL 09 se 81 607 9997— O€
W Sida OSclaDiE—
PE 676 66— 6S Ee | 96°1 Cove 09
 WV
6 I OIEEO 6PL70 8167 9807— 8962 €7l1— 888 OLE Sv O'€ Ci 627 S6P— Of
4 160% Scale C560 ~FUS— P96) She 8) 9°7 a 107 U6r=— Of

JdA1IsqO
€ 6901 pOS— ILEl S8l 6cEl 9¢P LE ve 17 6S'l wS9b— 67
W PLSI=. PLELI LOIC t= Ils 8S€ (44 O'€ 9'1 16] PS VER 68
Il I «=—9S07'0 sCHHSTO OPTI:
)=— 19T— 8p: th cecil Stal IL lp v7 OL'l vpOSr— O€
(4 Aes) ecie SOLS)| FeCl S085 PO c9 OP 6'1 6 tar0 apieyD LEAS iz
W 8bPrl 6¢%h— $99 601I— 9011 € 99 0 ae S8'l 7— 96S LS
| I 0 Shst SisrO tre US0L—= “Sey "SLI= 069 COP IL v9 81 O09'€ trpp— Of
c CSc e7OL—— “SOS = SOK VSP S8P cL 8'P Ly p87 S8Sb— 87
€ $76E le C90
= ELS = Chi GS0Cas b8s L9 ey cl £37 BsaPe O¢
W Chee Slee SRLS ERIS
ee leh 88P OL ia. Lil SIE COSy— 88
Sl I IOIEO ECOIEO L9EE 8IPI— O8IZ «LF 89 eLv 9S S7 v'l LB8'l vlL6y— 67
6 SECU OR—a)
me COLA OF v6czl br OL 67 Sl L6l 8L87— Of
W 16zE 6POl— 86S1 St 888 OEE v9 bie v'l 98'1 0%6'>— 6S

JOArasqoO
MD
6 I 0 OLE
Os 6rlc CLOls EcOcli= fol se Beso ci cs ae ct IP 66€7— O€
(4 cl6l let es0bl Lik
= BtR6 — 779s COLS S7 0'7 TCT 7= BOS Of
€ [Hic S606—2 “Iti cOll= MLOce™ ySc— 9¢ re Lil v8'l 68L'y— O€
W ZISI 6S$8-— 9€6 68€-— S08 6L vs 0's 6'1 6972 ¥— 9S 06
Il I S070 vbSz70 1891 StS—
9 tpL Ss9— vlL8 43 99 6'€ 0'7 6'l v09'>— O€
(4 166 = REEL 19> s6— veol 9L1 6L 8'P 87 So5Li1 & LEP O€
W CTI
= 6b7—
= 8b. GSo= Vs (444 eL vy v7 ©6651 ye SLE 09
el I SPrse0 8ISPO 90SI tfh— 08% 16 60S 961— cL v'L Ce 6t'¢
=~ V= SOC O€
(4 RT Bean FLO Sone eels CG €8 Ge 9'°7 6l'l p8Sr— 67
€ GED
= OLP—
| C8 36— “SIDI = ESE cL pige tir 89° L6~P— 0€
W Cee SPLEnee VOC. OP—
= e6cS JEL— as 69 EG ¢-L6:7 LOGirs 68
SI I 0 [OLE
Om ESTE COlRe L9Ce>
wl Cl all SRO LOSES 76S 8Y uc I'l 6c = S C90 O€
(4 -S20t= Foci G9TI- t6L—" O9cc. 656 €9 6'€ "| 69° cSLy— Of
W 90KE PSSI— IP9l 169— O€61 6£L 09 WF vl CPT
= Ve 195 09

805
 Table (5.5.2) Color-Matching Functions and Spectral Chromaticity Coordinates of CIE 1931
Standard Observer, Modified by Vos (1978)” °

Chromaticity coordinates Color-matching functions

A (nm) x"(X) yA) x(A) yA) Z(X)

380 0.17755 0.01320 0.26899(—2) 0.20000(— 3) 0.12260(— 1)
385 0.17744 0.01322 0.53105(—2) 0.39556(—3) 0.24222(— 1)
390 0.17723 0.01315 0.10781(—1) 0.80000(— 3) 0.49250(— 1)
395 0.17699 0.01316 0.20792(—1) 0.15457(—2) 0.95135(—1)
400 0.17676 0.01303 0.37981(— 1) 0.28000(—2) 0.17409
405 0.17644 0.01301 0.63157(—1) 0.46562(—2) 0.29013
410 0.17599 0.01303 0.99941(— 1) 0.74000(—2) 0.46053
415 0.17549 0.01306 0.15824 0.11779(— 1) 0.73166
420 0.17480 0.01333 0.22948 0.17500(— 1) 0.10658(1)
425 0.17368 0.01401 0.28108 0.22678(— 1) 0.13146(1)
430 0.17223 0.01512 0.31095 0.27300(— 1) 0.14672(1)
435 0.17022 0.01677 0.33072 0.32584(— 1) 0.15796(1)
440 0.16770 0.01907 0.33336 0.37900(— 1) 0.16166(1)
445 0.16433 0.02199 0.31672 0.42391(—1) 0.15682(1)
450 0.15980 0.02589 0.28882 0.46800(— 1) 0.14717(1)
455 0.15404 0.03092 0.25969 0.52122(—1) 0.13740(1)
460 0.14690 0.03787 0.23276 0.60000(— 1) 0.12917(1)
465 0.13828 0.04803 0.20999 0.72942(— 1) 0.12356(1)
470 0.12668 0.06595 0.17476 0.90980(— 1) 0.11138(1)
475 0.11185 0.09499 0.13287 0.11284 0.94220
480 0.09316 0.14086 0.91944(—1) 0.13902 0.75596
485 0.07007 0.20888 0.56985(—1) 0.16987 0.58640
490 0.04623 0.30304 0.31731(—1) 0.20802 0.44669
495 0.02380 0.42042 0.14613(—1) 0.25808 0.34116
500 0.00819 0.54540 0.48491(—2) 0.32300 0.26437
505 0.00378 0.66062 0.23215(—2) 0.40540 0.20594
510 0.01393 0.75440 0.92899(—2) 0.50300 0.15445
515 0.03922 0.81453 0.29278(—1) 0.60811 0.10918
520 0.07502 0.83492 0.63791(— 1) 0.71000 0.76585(— 1)
322) 0.11517 0.82638 0.11081 0.79510 0.56227(— 1)
530 0.15596 0.80538 0.16692 0.86200 0.41366(— 1)
535 0.19425 0.78070 0.22768 0.91505 0.29353(— 1)
540 0.23106 0.75311 0.29269 0.95400 0.20042(— 1)
545 0.26722 0.72295 0.36225 0.98004 0.13312(—1)
550 0.30300 0.69089 0.43635 0.99495 0.87823(—2)
2B 0.33865 0.65749 0.51513 0.10001(1) 0.58573(—2)
560 0.37423 0.62322 0.59748 0.99500 0.40493(—2)
565 0.40965 0.58858 0.68121 0.97875 0.29217(—2)
570 0.44471 0.55396 0.76425 0.95200 O72
SiS 0.47910 O31977 0.84394 0.91558 0.19706(—2)
580 0.51245 0.48653 0.91635 0.87000 0.18066(—2)
585 0.54436 0.45477 0.97703 0.81623 0.15449(—2)
590 0.57431 0.42499 0.10230(1) 0.75700 0.12348(—2)
595 0.60168 0.39768 0.10513(1) 0.69483 0.11177(—2)
600 0.62539 0.37406 0.10550(1) 0.63100 0.90564(— 3)
605 0.64622 0.35333 0.10362(1) 0.56654 0.69467(—3) -
610 0.66344 0.33627 0.99239 0.50300 0.42885(—3)
615 0.67749 0.32227 0.92861 0.44172 0.31817(—3)
620 0.68869 0.31109 0.84346 0.38100 0.25598(—3)

806
Table 1(5.5.2)— Continued

Chromaticity coordinates Color-matching functions

A (nm) x(A) yA) x"(X) y(X) z'(X)

625 0.69761 0.30223 0.73983 0.32052 0.15679(—3)
630 0.70478 0.29510 0.63289 0.26500 0.97694(—4)
635 0.71077 0.28913 0.53351 0.21702 0.68944(—4)
640 0.71566 0.28424 0.44062 0.17500 0.51165(—4)
645 0.71958 0.28034 0.35453 0.13812 0.36016(—4)
650 0.72248 0.27746 0.27862 0.10700 0.24238(—4)
655 0.72457 0.27537 0.21485 0.81652(—1) 0.16915(—4)
660 0.72593 0.27400 0.16161 0.61000(— 1) 0.11906(—4)
665 0.72721 0.27273 0.11820 0.44327(— 1) 0.81489(—5)
670 0.72820 0.27174 0.85753(—1) 0.32000(— 1) 0.56006(—5)
675 0.72891 0.27103 0.63077(— 1) 0.23454(— 1) 0.39544(—5)
680 0.72940 0.27054 0.45834(— 1) 0.17000(— 1) 0.27912(—5)
685 0.72970 027025 0.32057(—1) 0.11872(—1) 0.19176(—5)
690 0.72987 0.27008 0.22187(—1) 0.82100(—2) 0.13135(—5)
695 0.73003 0.26992 0.15612(—1) 0.57723(—2) 0.91519(—6)
700 0.73010 0.26985 0.11098(—1) 0.41020(—2) 0.64767(—6)
705 0.73006 0.26989 0.79233(—2) 0.29291(—2) 0.46352(—6)
710 O29) 0.27000 0.56531(—2) 0.20910(—2) 0.33304(—6)
715 U.72979 0.27016 0.40039( —2) 0.14822(—2) 0.23823(—6)
720 0.72958 0.27037 0.28253(—2) 0.10470(—2) 0.17026(—6)
725 0.72932 0.27062 0.19947(—2) 0.74015(—3) 0.12207(—6)
730 0.72904 0.27091 0.13994(—2) 0.52000(—3) 0.87107(—7)
735 0.72873 0.27121 0.96980(—3) 0.36093(—3) 0.61455(—7)
740 0.72840 0.27154 0.66847(—3) 0.24920(—3) 0.43162(—7)
745 0.72806 0.27189 0.46141(—3) 0.17231(—3) 0.30379(—7)
750 0.72769 0.27226 U.32073C— 3) 0.12000(—3) 0.21554(—7)
Lee) 0.72730 0.27264 0.22573(—3) 0.84620(—4) 0.15493(—7)
760 0.72689 0.27305 ONS97I35E—3) 0.60000(—4) 0.11204(—7)
765 0.72646 0.27348 ON 1275-3) 0.42446(—4) 0.80873(—8)
770 0.72601 0.27392 0.79513(— 4) 0.30000(— 4) 0.58340(—8)
775 0.72555 0.27438 0.56087(—4) 0.21210(—4) 0.42110(— 8)
780 0.72508 0.27486 0.39541(—4) 0.14989( —4) 0.30383(— 8)
785 0.72459 0.27535 0.27852(—4) 0.10584(—4) 0.21907(— 8)
790 0.72409 0.27587 0.19597(—4) 0.74656(—5) 0.15778(— 8)
795 0.72358 0.27636 0.13770(—4) 0.52592(—5) 0.11348(— 8)

800 0.72306 0.27687 0.96700(— 5) 0.37028(—5) 0.81565(—9)

805 0.72253 0.27740 0.67918(—5) 0.26076(— 5) 0.58626(—9)
810 0.72199 0.27794 0.47706(—5) 0.18365(—5) 0.42138(—9)
815 0.72144 0.27848 0.33550(—5) 0.12950(— 5) 0.30319(—9)
820 0.72089 0.27903 0.23534(—5) 0.91092(— 6) 0.21753(-—9)

825 0.72034 0.27958 0.16377(—5) 0.63564(—6) 0.15476(—9)

Totals 0.10702(3) 0.10748(3) 0.10505(3)
a

“Equal-energy stimulus (E): X;: = 0.33499, Y_ = 0.33618, Zp = 0.32883.

’Values given to five significant figures. Numbers in parentheses following the values x’(A), y (A), 2'(A)
denote powers of 10. Thus 0.85753(— 1) reads 0.85753 X 10 '.

807
 Table 1(5.5.3) Mean Color-Matching Functions and Spectral Chromaticity Coordinates of
Stiles 2° Pilot Group of Ten Observers, Interpolated from Table 1(5.3.3) at Intervals of
One Nanometer“

Wavelength Color-Matching Functions Chromaticity Coordinates

A
(nm) r(X) g(A) b(X) r(X) g(r)
390 0.18397(—2) — 0.45393(—3) 0.12152(—1) 0.1359 — 0.0335
391 0.22070(—2) —0.53474(—3) 0.14604(— 1) 0.1356 — 0.0329
392 0.2645 1(—2) —0.62941(—3) 0.17534(— 1) 0.1353 — 0.0322
393 0.31672(—2) —0.74019(—2) 0.21033(—1) 0.1350 —0.0316
394 0.37890(— 2) — 0.86974(—2) 0.25208(— 1) 0.1347 — 0.0309
395 0.46153(—2) —0.10464(—2) 0.31110(—1) 0.1331 — 0.0302
396 0.54135(—2) —0.11995(—2) 0.36165(— 1) 0.1341 — 0.0302
397 0.62593(—2) — 0.13528(—2) 0.41145(— 1) 0.1359 — 0.0294
398 0.72015(—2) — 0.15386(—2) 0.46652(— 1) 0.1377 —0.0294
399 0.83035(—2) —0.17975(—2) 0.53472(— 1) 0.1384 — 0.0300
400 0.96264( — 2) —0.21689(—2) 0.62371(— 1) 0.1379 —0.0311
401 0.11208(—1) —0.26695(—2) 0.73827(— 1) 0.1361 —0.0324
402 0.13000(— 1) — (1322992) 0.87262(— 1) 0.1340 — 0.0333
403 0.14934(— 1) — 0.37616(—2) 0.10186 0.1321 — 0.0333
404 0.16943(— 1) —0.41784(—2) 0.11682 0.1308 — 0.0322
405 0.18979( — 1) — 0.44304(— 2) 0.13161 0.1298 — 0.0303
406 0.21071(—1) — 0.46185(—2) 0.14684 0.1290 — 0.0283
407 0.23269( — 1) — 0.48824(— 2) 0.16337 0.1280 — 0.0269
408 0.25621(—1) — 0.53580(—2) 0.18206 0.1266 —0.0265
409 0.28156(— 1) —0.61520(—2) 0.20358 0.1248 — 0.0273
410 0.30803(— 1) — 0.72048(—2) 0.22750 0.1227 — 0.0287
41] 0.33450(— 1) — 0.83929(— 2) 0.25297 0.1203 — 0.0302
412 0.35991(—1) —0:959636=2) 0.27918 0.1178 —0.0314
413 0.38326(—1) —0.10704(—1) 0.30538 0.1151 — 0.0321
414 0.40459( — 1) —0.11689(— 1) 0.33172 0.1122 — 0.0324
415 0.42459( — 1) —0.12579(— 1) 0.35897 0.1092 — 0.0324
416 0.44398(— 1) —0.13405(— 1) 0.38787 0.1060 — 0.0320
417 0.46342(— 1) —0.14194(—1) 0.41914 0.1027 — 0.0315
418 0.48282(— 1) —0.14977(—1) 0.45284 0.0993 — 0.0308
419 0.50098( — 1) — 0.15786(— 1) 0.48810 0.0959 — 0.0302
420 0.51662(— 1) —0.16651(—1) 0.52396 0.0924 —0.0298
421 0.52851(—1) —0.17603(—1) 0.55953 0.0889 — 0.0296
422 0.53564(— 1) —0.18649(—1) 0.59396 0.0852 — 0.0297
423 0.53789(— 1) —0.19688(— 1) 0.62676 0.0814 — 0.0298
424 0.53542(— 1) — 0.20593(— 1) 0.65752 0.0775 — 0.0298
425 0.52837(— 1) —0.21240(— 1) 0.68586 0.0736 — 0.0296
426 0.51690(— 1) —0.21512(—1) 0.71140 0.0697 — 0.0290
427 0.50160(— 1) —0.21404(— 1) 0.73438 0.0657 — 0.0280
428 0.48348(— 1) —0.21029(—1) 0.75560 0.0618 — 0.0269
429 0.46357(— 1) — 0.20502(— 1) 0.77588 0.0578 — 0.0256
430 0.44287(— 1) —0.19936(—1) 0.79604 0.0540 —0.0243
43] 0.42209( — 1) —0.19414(—1) 0.81667 0.0503 —0.0231
432 0.40065(— 1) —0.18875(— 1) 0.83746 0.0467 — 0.0220
433 0.37751(—1) —0.18212(—1) 0.85779 0.0430 —0.0208
434 0.35168(—1) —0.17320(—1) 0.87703 0.0393 —0.0194
435 0.32220(— 1) —0.16097(—1) 0.89459 0.0354 —0.0177 -
436 0.28888(— 1) —0.14518(—1) 0.91025 0.0312 —0.0157
437 0.25318(—1) ~O. 127116 0.92446 0.0270 — 0.0136
438 0.21671(—1) —0.10821(—1) 0.93780 0.0228 —0.0114
439 0.18104(—1) — 0.89882(— 2) 0.95080 0.0189 — 0.0094

808
Table 1(5.5.3)— Continued

Color-Matching Functions Chromaticity Coordinates

Wavelength
r
(nm) r(A) g(r) b(A) r(X) g(A)
440 0.14763(—1) —0.73457(—2) 0.96395 0.0152 — 0.0076
44] 0.11643(—1) —0.58873(—2) 0.97687 0.0118 — 0.0060
442 0.85576(—2) —0.44508(—2) 0.98815 0.0086 — 0.0045
443 0.53193(—2) —0.28691(—2) 0.99635 0.0053 — 0.0029
444 0.17441(—2) —0.97985(—3) 0.10001(1) 0.0017 — 0.0010
445 —0.23392(—2) 0.13690(—2) 0.99814 — 0.0023 0.0014
446 —0.69620(—2) 0.42191(—2) 0.99033 — 0.0070 0.0043
447 —0.12038(— 1) 0.75200(—2) 0.97759 — 0.0124 0.0077
488 —0.17481(—1) 0.11219(—1) 0.96083 — 0.0183 0.0118
449 —0.23206(—1) 0.15266(—1) 0.94092 — 0.0249 0.0164
450 —0.29130(—1) 0.19610(—1) 0.91875 — 0.0320 0.0216
451 —0.35208(—1) 0.24189(— 1) 0.89561 — 0.0398 0.0273
452 —0.41419(—1) 0.28932(— 1) 0.87310 — 0.0481 0.0336
453 —0.47746(—1) 0.33770(— 1) 0.85281 — 0.0569 0.0403
454 —0.54170(— 1) 0.38634(— 1) 0.83625 — 0.0660 0.0471
455 —0.60677(— 1) 0.43464(— 1) 0.82487 — 0.0751 0.0538
456 —0.67291(—1) 0.48303(— 1) 0.81815 — 0.0842 0.0604
457 —0.74083(—1) 0.53311(—1) 0.81348 — 0.0935 0.0673
458 —0.81124(—1) 0.58651(—1) 0.80820 — 0.1032 0.0746
459 —0.88482(— 1) 0.64481(—1) 0.79972 —0.1141 0.0831
460 —0.96224(—1) 0.70954(— 1) 0.78554 — 0.1266 -0.0933
461 —0.10434 0.78117(—1) 0.76479 — 0.1413 0.1058
462 —0.11270 0.85807(—1) 0.73987 —0.1581 0.1203
463 —0.12113 0.93840(— 1) 0.71354 — 0.1765 0.1367
464 —0.12947 0.10203 0.68846 =1959 0.1544

465 —0,13759 0.11022 0.66723 —0.2150 0.1722

466 — 0.14535 0.11825 0.65182 — 0.2327 0.1893
467 — 0.15284 0.12620 0.64089 — 0.2488 0.2054
468 ~( 16019 0.13420 0.63201 — 0.2643 0.2214
469 — 0.16747 0.14239 0.62281 — 0.2802 0.2382

470 — 0.17486 0.15088 0.61098 get |ok 0.2570

471 — 0.18247 0.15981 0.59430 — 0.3197, 0.2796
472 = (1,19025 0.16916 0.57222 — 0.3452 0.3069
473 — 0.19801 0.17875 0.54617 — 0.3758 0.3392
474 — (20553 0.18840 0.51769 — 0.4106 0.3764

475 — 0.21260 0.19794 0.48829 — 0.4489 0.4179

476 — (71903 0.20720 0.45941 — 0.4894 0.4629
477 — 0.22465 0.21603 0.43234 — 0.5302 0.5098
478 — O27 0.22444 0.40726 — 0.5707 0.5581
479 — 0.23386 0.23255 0.38389 — 0.6113 0.6078

480 — 0.23780 0.24042 0.36195 — 0.6523 0.6595

481 — 0.24153 0.24816 0.34118 — 0.6944 0.7135
482 — 0.24521 0.25585 0.32131 — 0.7387 0.7708
483 — 0.24896 0.26359 0.30216 — 0.7859 0.8321
484 — 0.25281 0.27155 0.28379 — 0.8356 0.8976

485 — 0.25674 0.27993 0.26634 — 0.8868 0.9669

486 — 0.26076 0.28889 0.24990 0.9379 1.0391
487 — 0.26485 0.29862 0.23458 — 0.9869 1.1128
488 — 0.26901 0.30928 0.22050 — 1.0316 1.1860
489 ~O.27319 0.32095 0.20767 — 1.0695 1.2565

809
 Table 1(5.5.3)— Continued

Wayeleneth Color-Matching Functions Chromaticity Coordinates

r
(nm) r(X) (A) b(A) r(X) (A)
490 =O TZ? 0.33353 0.19593 — 1.0995 1.3225
491 — 0.28109 0.34686 0.18509 — 1.1205 1.3827
492 — 0.28453 0.36081 0.17498 — 1.1324 1.4360
493 — 0.28744 0.37526 0.16541 <s1-1550 1.4818
494 — 0.28969 0.39008 0.15622 — 1.1289 1.5201
495 — 0.29125 0.40521 0.14730 — 1.1148 1.5510
496 — 0.29228 0.42078 0.13865 — 1.0940 1.5750
497 — 0.29298 0.43696 0.13031 — 1.0681 1.5930
498 — 0.29354 0.45390 0.12232 — 1.0384 1.6057
499 — 0.29415 0.47173 0.11470 — 1.0064 1.6140
500 — 0.29500 0.49060 0.10749 — 0.9733 1.6186
501 — 0.29616 0.51059 0.10071 — 0.9398 1.6202
502 — 0.29734 0.53148 0.94295(— 1) —0.9053 1.6182
503 — 0.29813 0.55298 0.88199(— 1) — 0.8691 1.6120
504 — 0.29816 0.57482 0.82358(— 1) — 0.8305 1.6011
505 — 0.29706 0.59673 0.76714(— 1) — 0.7892 1.5854
506 — 0.29446 0.61845 0.71209(— 1) —0.7451 1.5649
507 — 0.29004 0.63975 0.65792(— 1) — 0.6980 1.5397
508 — 0.28387 0.66063 0.60468(— 1) — 0.6492 1.5109
509 — 0.27628 0.68127 0.55274(— 1) — 0.6003 1.4802
510 — 0.26759 0.70184 0.50248(— 1) 4+()/5523 1.4486
511 — 0.25812 0.72251 0.45426(— 1) — 0.5063 1.4172
S12 — 0.24816 0.74344 0.40844( — 1) — 0.4629 1.3867
513 — 0.23802 0.76478 0.36537(— 1) — 0.4225 1.3577
514 — 0.22781 0.78656 03252562 )) — 0.3853 1.3303
515 = 21725 0.80852 0.28781(—1) — 0.3504 1.3040
516 — 0.20600 0.83037 0.25287(— 1) —0.3171 1.2782
517 — 0.19374 0.85179 0.22016(— 1) — 0.2849 1.2525
518 —0.18014 0.87252 0.18945(— 1) = 2532 1.2266
519 — 0.16488 0.89226 0.16050(— 1) = (0.2298 1.2002
520 — 0.14768 0.91076 0.13309(— 1) — 0.1902 1.1731
521 — 0.12848 0.92793 0.10717(— 1) — 0.1586 1.1454
ape: — 0.10745 0.94384 0.82893(— 2) —(). 142 1.1174
9 fs) —0.84777(—1))) 0.95857 0.60380(— 2) — 0.0964 1.0895
524 —0.60631(—1) 0.97221 0.39763(— 2) — 0.0662 1.0619
20 —0.35184(—1) 0.98482 0.21170(— 2) — 0.0370 1.0347
526 —0.86031(—2) 0.99650 0.47268(— 3) — 0.0087 1.0082
527 0.18950(—1) —0.10073(1) += —0.94755(—3) 0.0185 0.9824
528 0.47338(—1)!) 0.10872) ™ — 02165762) 0.0446 0.9575
529 0.76441(—1) 0.10261(1) —0.32223(—2) 0.0695 0.9334
530 0.10614 0.10339(1) = —0.41574(—2) 0.0934 0.9102
531 0.13633 0.10405(1) = —0.50101(—2) 0.1163 0.8879
532 0.16688 0.10459(1) = —0.58186(—2) 0.1383 0.8665
533 0.19770 0.10499(1) = —0.66202(—2) 0.1593 0.8460
534 0.22868 0.10525(1) = —0.74487(— 2) 0.1795 0.8263
535 0.25981 0.10538¢1)-9 * —0383032( +2) 0.1990 0.8073
536 0.29116 0.10542(1) —0.91612(—2) 0.2179 0.7890
537 0.32277 0.105391) = —0.10001(—1) 0.2362 0.7711
538 0.35471 0.10531(1) = —0.10799(—1) 0.2539 0.7538
539 0.38702 0.105210)" ") — 01 1536G47) 0.2711 0.7370

810
Table 1(5.5.3)— Continued

Color-Matching Functions Chromaticity Coordinates

Wavelength
r
(nm) r(X) (A) b(A) r(A) g(A)
540 0.41976 0.10512(1) —0.12191(—1) 0.2877 0.7206
541 0.45298 0.10507(1) —0.12742(—1) 0.3038 0.7047
542 0.48676 0.10505(1) —0.13188(—1) 0.3194 0.6893
543 0.52121 0.10504(1) —0.13543(—1) 0.3345 0.6742
544 0.55645 0.10502(1) —0.13822(—1) 0.3493 0.6593
545 0.59259 0.10498(1) —0.14039(—1) 0.3639 0.6447
546 0.62973 0.10489(1) —0.14210(—1) 0.3783 0.6302
547 0.66800 0.10473(1) —0.14348(—1) 0.3927 0.6157
548 0.70748 0.10448(1) —0.14468(—1) 0.4071 0.6012
549 0.74822 0.10413(1) —0.14579(—1) 0.4215 0.5867
550 0.79004 0.10368(1) —0.14681(—1) 0.4360 0.5721
551 0.83271 0.10312(1) —0.14770(—1) 0.4503 0.5577
552 0.87603 0.10245(1) —0.14845(—1) 0.4646 0.5433
ely 0.91978 0.10168(1) —0.14901(—1) 0.4786 0.5291
554 0.96376 0.10080(1) —0.14936(—1) 0.4925 0.5151

550 0.10078(1) 0.99826 —0.14947(—1) 0.5061 0.5014

556 0.10516(1) 0.98745 ~0.14931(—1) 0.5195 0.4878
557 0.10953(1) 0.97574 —0.14888(—1) 0.5327 0.4745
558 0.11391(1) 0.96338 —0.14819(—1) 0.5456 0.4615
559 0.11834(1) 0.95066 —0.14727(—1) 0.5584 0.4486

560 0.12283(1) 0.93783 —0.14613(—1) 0.5709 0.4359

561 0.12743(1) 0.92515 —0.14480(—1) 0.5832 0.4234
562 0.13215(1) 0.91287 —0.14328(—1) 0.5953 0.4112
563 0.13702(1) 0.90123 —0.14160(—1) 0.6070 0.3993
564 0.14206(1) 0.89044 —0.13978(—1) 0.6184 0.3876

565 0.14727(1) 0.88039 —0.13782(—1) 0.6295 0.3763

566 0.15262(1) 0.87070 —0.13575(—1) 0.6404 0.3653
567 0.15807(1) 0.86100 —0.13357(—1) 0.6509 0.3546
568 0.16360(1) 0.85095 —0.13129(—1) 0.6613 0.3440
569 0.16917(1) 0.84018 —0.12893(—1) 0.6716 0.3335

570 0.17476(1) 0.82835 —0.12650(—1) 0.6818 0.3232

571 0.18034(1) 0.81513 —0.12401(—1) 0.6920 0.3128
572 0.18589(1) 0.80020 —0.12147(—1) 0.7023 0.3023
573 0.19138(1) 0.78362 —0.11888(—1) 0.7126 0.2918
574 0.19680(1) 0.76573 —0.11625(—1) 0.7230 0.2813

575 0.20214(1) 0.74686 —0.11356(—1) 0.7332 0.2709

576 0.20739(1) 0.72735 —0.11082(—1) 0.7433 0.2607
577 0.21254(1) 0.70752 —0.10803(—1) 0.7531 0.2507
578 0.21757(1) 0.68770 —0.10518(—1) 0.7626 0.2411
579 0.22247(1) 0.66819 —0.10228(—1) 0.7718 0.2318

580 0.22724(1) 0.64930 —0.99317(—2) 0.7804 0.2230

581 0.23186(1) 0.63117 —0.96303(—2) 0.7886 0.2147
582 0.23635(1) 0.61365 —0.93258(—2) 0.7964 0.2068
583 0.24069(1) 0.59659 —0.90203(—2) 0.8038 0.1992
584 0.24489(1) 0.57981 —0.87159(—2) 0.8109 0.1920

585 0.24896(1) 0.56317 —0.84148(—2) 0.8178 0.1850

586 0.25289(1) 0.54649 —0.81188(—2) 0.8245 0.1782
587 0.25668(1) 0.52963 —0.78300(—2) 0.8311 0.1715
588 0.26034(1) 0.51245 —0.75502(—2) 0.8376 0.1649
589 0.26387(1) 0.49480 —0.72812(—2) 0.8441 0.1583

811
 Table 1(5.5.3)— Continued

Color-Matching Functions Chromaticity Coordinates

Wavelength
r
(nm) r(A) 8(A) b(A) r(A) g(A)
590 0.26725(1) 0.47675 —0.70210(—2) 0.8505 0.1517
591 0.27044(1) 0.45842 —0.67665(—2) 0.8569 0.1452
692 0.27344(1) 0.43993 —0.65146(—2) 0.8632 0.1389
A935 0.27620(1) 0.42143 —0.62622(—2) 0.8693 0.1326
594 0.27871(1) 0.40302 —0.60062(—2) 0.8753 0.1266
395 0.28093(1) 0.38484 —0.57437(—2) 0.8811 0.1207
596 0.28286(1) 0.36699 —0.54718(—2) 0.8867 0.1150
597 0.28445(1) 0.34961 ~0,51875¢20) 0.8920 0.1096
598 0.28570(1) 0.33277 —0.48896(—2) 0.8971 0.1045
ay 0.28660(1) 0.31647 —0.45831(—2) 0.9019 0.0996
600 0.28717(1) 0.30069 —0.42743(—2) 0.9064 0.0949
601 0.28742(1) 0.28539 —0.39696(—2) 0.9108 0.0904
602 0.28734(1) 0.27056 —0.36751(—2) 0.9150 0.0862
603 0.28695(1) 0.25615 —0.33970(—2) 0.9190 0.0820
604 0.28625(1) 0.24215 —0.31410(—2) 0.9229 0.0781
605 0.28525(1) 0.22853 —0:29132(22) 0.9267 0.0742
606 0.28397(1) 0.21526 —0.27191(—2) 0.9304 0.0705
607 0.28240(1) 0.20233 —0.25626(—2) 0.9339 0.0669
608 0.28055(1) 0.18975 —0.24398(—2) 0.9374 0.0634
609 0.27842(1) 0.17755 —0.23441(—2) 0.9408 0.0600
610 0.27601(1) 0.16575 —0.22693(—2) 0.9441 0.0567
611 0.27333(1) 0.15438 —0.22092(—2) 0.9473 0.0535
612 0.27037(1) 0.14347 —0.21577(—2) 0.9503 0.0504
613 0.26714(1) 0.13305 —0.21090(—2) 0.9533 0.0475
614 0.26365(1) 0.12313 —0.20572(—2) 0.9561 0.0447
615 0.25989(1) 0.11373 —0.19966(—2) 0.9588 0.0420
616 0.25586(1) 0.10489 —0.19220(—2) 0.9613 0.0394
617 0.25158(1) 0.96593(—1) —0.18328(—2) 0.9637 0.0370
618 0.24707(1) 0.88810(—1) —0.17319(—2) 0.9660 0.0347
619 0.24235(1) 0.81516(—1) —0.16223(—2) 0.9681 0.0326
620 0.23743(1) 0.74682(— 1) —0.15069(—2) 0.9701 0.0305
621 0.23234(1) 0.68282(— 1) —0.13883(—2) 0.9720 0.0286
622 0.22708(1) 0.62289(— 1) —0.12695(—2) 0.9738 0.0267
623 0.22169(1) 0.56679(— 1) —0.11532(—2) 0.9756 0.0249
624 0.21617(1) 0.51426(—1) —0.10418(—2) 0.9772 0.0232
625 0.21054(1) 0.46504(— 1) —0.93822(—3) 0.9788 0.0216
626 0.20483(1) 0.41892(—1) —0.84426(—3) 0.9804 0.0201
627 0.19904(1) 0.37578(—1) —0.75965(—3) 0.9818 0.0185
628 0.19320(1) 0.33553(—1) —0.68351(—3) 0.9833 0.0171
629 0.18733(1) 0.29807(— 1) —0.61496(—3) 0.9847 0.0157
630 0.18145(1) 0.26333(—1) —0,55316(—3) 0.9860 0.0143
631 0.17557(1) 0.23120(—1) —0.49727( —3) 0.9873 0.0130
632 0.16972(1) 0.20162(—1) —0,44648(—3) 0.9885 0.0117
633 0.16390(1) 0.17448(— 1) —0.39998(—3) 0.9897 0.0105
634 0.15815(1) 0.14972(—1) —0,35697(—3) 0.9908 0.0094
635 0.15247(1) 0.12724(—1) —0.31668(—3) O99T9 0.0083
636 0.14687(1) 0.10695(— 1) —0.27849(—3) 0.9930 0.0072
637 0.14137(1) 0.88722(—2) —0.24217(—3) 0.9939 0.0062
638 0.13596(1) 0.72418(—2) —0.20760(—3) 0.9949 0.0053
639 0.13065(1) 0.57899(—2) —0.17465(—3) 0.9957 0.0044

812
Table 1(5.5.3)— Continued

Wavelength Color-Matching Functions Chromaticity Coordinates:

A
(nm) r(A) g(A) b(X) r(A) (A)
640 0.12543(1) 0.45033(— 2) —0.14319(—3) 0.9965 0.0036
641 0.12030(1) 0.33688(— 2) —0.11310(—3) 0.9973 0.0028
642 0.11527(1) 0.23734(— 2) — 0.84268(—4) 0.9980 0.0021
643 0.11033(1) 0.15045(—2) — 0.56564(— 4) 0.9987 0.0014
644 0.10550(1) 0.74966(— 3) — 0.29873(—4) 0.9993 0.0007
645 0.10076(1) 0.96611(—4) — 0.4083 1(—5) 0.9999 0.0001
646 0.96114 — 0.46618(—3) 0.20896(— 4) 1.0005 — 0.0005
647 0.91577 — 0.94692(— 3) 0.45000(— 4) 1.0010 — 0.0010
648 0.87149 —0.13524(— 2) 0.68102(—4) 1.0015 — 0.0016
649 0.82835 —0.16894(—2) 0.90080(— 4) 1.0019 — 0.0020
650 0.78642 —0.19645(—2) 0.11081(—3) 1.0024 — 0.0025
651 0.74574 —0.21840(— 2) 0.13018(—3) 1.0028 — 0.0029
652 0.70637 — 0.23544(— 2) 0.14807(— 3) 1.0031 — 0.0033
653 0.66835 —0.24819(—2) 0.16437(— 3) 1.0035 — 0.0037
654 0.63174 0, 25727042) 0.17897(— 3) 1.0038 — 0.0041
655 0.59659 —+026327(+ 2) 0.19175(—3) 1.0041 — 0.0044
656 0.56294 —0.26679( — 2) 02026 1( 3) 1.0044 — 0.0048
657 0.53081 — 0.26828(—2) 0.21149(—3) 1.0047 —0.0051
658 0.50018 70.267 96(-2) 0.21842(— 3) 1.0049 — 0.0054
659 0.47099 -40.26601(=2) 0.22343(— 3) 1.0052 — 0.0057
660 0.44320 — 0.26262(— 2) 0.22656(— 3) 1.0054 —0.0060
661 0.41679 0.25797 G2) 0.22784( — 3) 1.0057 — 0.0062
662 0.39172 —0.25224( —2) 0.2273 1(— 3) 1.0059 — 0.0065
663 0.36793 —0.24560(— 2) 0:22501(>-3) 1.0061 — 0.0067
664 0.34541 20038220 >2) 0.22096(— 3) 1.0063 — 0.0069
665 0.32410 “#23027(=22) 0:21520¢—3) 1.0065 — 0.0072
666 0.30398 + 022192(=2) 0.20776( — 3) 1.0067 — 0.0073
667 0.28502 +0;21332(—2) 0.19868(— 3) 1.0068 — 0.0075
668 0.26716 —0.20460(— 2) 0.18809( — 3) 1.0070 — 0.0077
669 0.25035 +0,19581( =2) 0.17630(—3) 1.0072 — 0.0079
670 0.23455 —0.18700(—2) 0.16361(—3) 1.0073 — 0.0080
671 0.21971 08 18226-32) 0.15033(— 3) 1.0075 — 0.0082
672 0.20576 —0.16950(— 2) 0.13675(— 3) 1.0076 — 0.0083
673 0.19267 —0.16090(— 2) 0.12317(—3) 1.0078 — 0.0084
674 0.18037 #0152472) 0.10988( — 3) 1.0079 — 0.0085
675 0.16884 —0.14424(—2) 0.97164(—4) 1.0080 — 0.0086
676 0.15801 0:1 3625(F22) 0.85290( — 4) 1.0081 — 0.0087
677 0.14784 + 0:12856(=2) 0.74535(— 4) 1.0083 — 0.0088
678 0.13829 HO Z20(=-2) 0.65168(—4) 1.0084 — 0.0088
679 0.12930 —0.11420(—2) 0.57370( — 4) 1.0085 — 0.0089
680 0.12086 40. 10755(F=2) 0.51033(—4) 1.0085 — 0.0090
681 0.11294 —0.10124(—2) 0.4598 1(—4) 1.0086 — 0.0090
682 0.10549 —0.95247(— 3) 0.42045(— 4) 1.0087 — 0.0091
683 0.985 10(— 1) —0.89554(— 3) 0.39058(— 4) 1.0088 — 0.0092
684 0.91958(— 1) —0.84145(— 3) 0.36853(— 4) 1.0088 — 0.0092
685 0.85811(— 1) — 0.79004( — 3) 0.35271(—4) 1.0089 — 0.0093
686 0.80044(— 1) —0.74116(—3) 0.34151(—4) 1.0089 — 0.0093
687 0.74630(— 1) — 0.69463(— 3) 0.33338(— 4) 1.0089 — 0.0094
688 0.69544(— 1) — 0.65030(— 3) 0.32676(— 4) 1.0090 — 0.0094
689 0.64762(— 1) — 0.60803(— 3) 0.32017(— 4) 1.0090 — 0.0095
 Table 1(5.5.3)— Continued
ee ————————e—e=SO

Color-Matching Functions Chromaticity Coordinates

Wavelength
A
(nm) r(X) g(r) B(A) r(X) g(r)
690 0.60260(— 1) — 0.56765(— 3) 0.31211(—4) 1.0090 — 0.0095
691 0.56019(— 1) 0.52909 —3) 0.30183(— 4) 1.0090 — 0.0095
692 0.52029(— 1) — 0.49231(—3) 0.28963(— 4) 1.0090 — 0.0095
693 0.48282(— 1) — 0.45733(—3) 0.27587(—4) 1.0090 — 0.0096
694 0.44769(— 1) — 0.42414(—3) 0.26090(— 4) 1.0090 — 0.0096
695 0.41480(— 1) — 0.39274(—3) 0.24508(— 4) 1.0090 — 0.0096
696 0.38408(— 1) — 0.36311(—3) 0.22876(— 4) 1.0089 — 0.0095
697 0.35543(— 1) 7 0'33526(= 3) 0.21228(—4) 1.0089 — 0.0095
698 0.32878(— 1) —0.30917(—3) 0.19597(— 4) 1.0089 — 0.0095
699 0.30405(— 1) — 0.28486(— 3) 0.18018(—4) 1.0089 — 0.0095
700 0.28114(—1) ==9.2623 (== 3) 0.16521(—4) 1.0088 — 0.0094
701 0.25999(— 1) —0.24151(—3) 0.15140(—4) 1.0088 — 0.0094
702 0.24051(— 1) — 0.22247(— 3) 0.13905(—4) 1.0087 — 0.0093
703 0.22261(—1) —0.20514(— 3) 0.12833(—4) 1.0087 — 0.0093
704 0.20620(— 1) —0.18940(—3) 0.11911(—4) 1.0087 — 0.0093
705 0.19117(—1) S(UPST2(>3) 0.11124(—4) 1.0087 — 0.0092
706 0.17740(— 1) = ON6218(=-3) 0.10457(— 4) 1.0086 — 0.0092
707 0.16481(—1) —0.15046(—3) 0.98953(—5) 1.0086 — 0.0092
708 0.15329(— 1) = O3983(--3) 0.94242(—5) 1.0086 — 0.0092
709 0.14273(— 1) —0.13018(—3) 0.90293(—5) 1.0086 — 0.0092
710 0.13305(— 1) —0.12140(—3) 0.86965(—5) 1.0085 — 0.0092
711 0.12415(—1) —0/11336(= 3) 0.84120(—5) 1.0085 — 0.0092
be 0.11593(—1) —0.10595(—3) 0.81621(—5) 1.0085 = 0092
773 0.10829(— 1) — 0.99062(— 4) 0.79333(—5) 1.0085 — 0.0092
714 0.10116(— 1) —0.92590(—4) OTRAS) 1.0085 — 0.0092
WS 0.94092(— 2) —0.85760(—4) 0.74351(—5) 1.0084 — 0.0092
716 0.87394(—2) —0.79184(—4) 0.71463(—5) 1.0083 —0.0091
717 0.81190(—2) — 0.73128(—4) 0.68695(— 5) 1.0082 —0.009i
718 0.75441(—2) —0.67550(—4) 0.66041(—5) 1.0081 — 0.0090
719 0.70114(—2) — 0.62412(—4) 0.63497(— 5) 1.0081 — 0.0090
720 O.651TI=2) —0.57677(—4) 0.61057(—5) 1.0080 —0.0089
721 0.60599(—2) —0.53313(—4) 0.58718(—5) 1.0079 — 0.0089
722 0.56354(— 2) —0.49290(— 4) 0.56474( — 5) 1.0078 — 0.0088
723 0.52417(—2) —0.45580(—4) 0.54322(—5) 1.0077 — 0.0088
724 0.48765(—2) — 0.42159(—4) 0.52258(—5) 1.0076 — 0.0087
725 0.45377(—2) —0.39003(—4) 0.50277(—5) 1.0075 — 0.0087
726 0.42232(—2) —0.36091(—4) 0.48377(— 5) 1.0075 — 0.0086
ip4) 039313G-2) — 0.33403(—4) 0.46553(—5) 1.0074 — 0.0086
728 0.36600(— 2) —0.30921(—4) 0.44728(—5) 1.0073 — 0.0085
729 0.2408 1(—2) — 0.28628(—4) 0.42952(—5) 1.0072 — 0.0085
730 0.31742(—2) —0.26511(—4) 0.41251(—5) 1.0071 — 0.0084

Some of the entries in the table have been interpolated numerically. The primary stimuli
are located at mp=15,500, mg=19,000, mz, = 22,500 cm~'. No correction for rod
intrusion has been applied to the data. Rod-intrusion corrected data are given in Table
3(5.5.4). Numbers in parentheses denote the power of ten. Thus, 0.18397(—2) reads
0.18397X 107.

814
Table 1(5.5.4) Mean 10° Color-Matching Functions, Spectral Chromaticity Coordinates, and W .D. W.
Coordinates of Stiles—Burch (1959) 49 Observers at Wavelengths from 380 to 760 nm and 5-nm intervals.
The Data are Interpolated from Table 1(5.5.4)

Color-Matching Functions Chromaticity Coordinates W. D. W. Coordinates

s(t a
Wavelen
A (nm) r(X) g(A) b(A) r(X) g(A) b(A) r*(X) g*(A)_ —b*(A)
380 0.0002 —0.0001 0.0008 0.23239%7-0.0551 0.8228 0.0245 —0.0259 1.0014
385 0.0006 —0.0001 0.0022 0.2222 —0.0549 0.8327 0.023237-=0.0255 1.0023
390 0.0015 —0.0004 0.0062 0.2102 —0.0546 0.8444 0.0216 —0.0250 1.0034
395 0.0038 —0.0010 0.0161 0.1999 —0.0542 0.8543 0.0203 —0.0246 1.0042
400 0.0089 —0.0025 0.0400 0: 19.1894 0.0539 0.8621 0.0194 —0.0242 1.0049
405 0.0188 —0.0059 0.0906 0.1820 —0.0566 0.8746 0.0181 —0.0251 1.0070
410 0.0350 —0.0119 0.1802 0.1723 —0.0584 0.8862 017025 0.0257 1.0087
415 0.0531 —0.0201 0.3088 0.1554 —0.0587 0.9034 0.0150 —0.0253 1.0103
420 0.0702 —0.0289 0.4670 0.1382 —0.0569 0.9187 0.0132 —0.0241 1.0110
425 0.0763 —0.0338 0.6152 0.1161 —0.0514 0.9353 0.0109 —0.0214 1.0106
430 0.0745 —0.0349 0.7638 0.0927 —0.0435 0.9507 0.0085 —0.0178 1.0093
435 0.0561 —0.0276 0.8778 0.0619 —0.0305 0.9686 0.0056 —0.0122 1.0066
440 0.0323 —0.0169 0.9755 0.0326 —0.0170 0.9844 0.0029 —0.0067 1.0038
445 — 0.0044 0.0024 1.0019 —0.0044 0.0024 1.0020 —0.0004 0.0009 0.9994
450 — 0.0478 0.0283 0.9996 —0.0488 0.0289 1.0198 —0.0041 0.0109 0.9932
455 — 0.0970 0.0636 0.9139 501102 0.0722 1.0380 —0.0090 0.0264 0.9827
460 —0.1586 0.1082 0.8297 —0.2035 0.1388 1.0647 —0.0160 0.0487 0.9674
465 — 0.2235 0.1617 0.7417 —0.3288 0.2378 1.0910 —0.0247 0.0795 0.9452
470 — 0.2848 0.2201 0.6134 —0.5190 0.4011 11 18ee 0.0366 01261 0.9106
475 — 0.3346 0.2796 0.4720 —0.8024 0.6705 1.1318 —0.0526 0.1958 0.8568
480 03776 0.3428 0.3495 —1.1998 1.0893 1.1105 —0.0728 0.2945 0.7783
485 — 0.4136 0.4086 0.2564 —1.6458 1.6256 1.0202 —0.0947 0.4167 0.6780
490 — 0.4317 0.4716 0.1819 —1.9468 2.1265 0.8203 —0.1145 0.5573 0.5573
495 — 0.4452 0.5491 D500 a— 89 1) 2.3400 USStie 04265 0.6968 0.4300
500 — 0.4350 0.6260 0.0910 —1.5426 2.2199 0.322708 0212718 0.8191 0.3087
505 — 0.4140 0.7097 0.0580 —1.1707 2.0067 0.163954 0: 1272 0.9252 0.1959
510 — 0.3673 0.7935 0.035.7270-7953 1.7180 0.0772 —0.1026 0.9875 0.1151
515 — 0.2845 0.8715 0.0200 —0.4687 1.4358 0.0329 —0.0743 1.0140 0.0603
520 —0.1855 0.9477 0.0095 —0.2404 1.2281 0.0123 —0.0447 1.0184 0.0263
525 — 0.0435 0.9945 0.0007 —0.0457 1.0450 0.0007 —0.0099 1.0081 0.0018
530 0.1270 1.0203 —0.0043 0.1111 0.8926 —0.0037 0.0275 0.9832 —0.0106
555 0.3129 1.0375 —0.0064 0.2328 0.7719 —0.0048 0.0644 0.9508 —0.0152
540 0.5362 1.0517 —0.0082 0.3394 0.6658 —0.0052 0.1046 0.9139 —0.0185
545 0.7722 1.0390 —0.0094 0.4286 0.5766 —0.0052 0.1459 0.8745 —0.0205
550 1.0059 1.0029 —0.0097 0.5032 0.5017 —0.0049 0.1876 0.8333 —0.0209
555 1.2710 0.9698 —0.0097 0.5697 0.4347 —0.0043 0.2320 0.7884 —0.0204
560 1.5574 0.9162 —0.0093 0.6320 0.3718 —0.0038 0.2816 0.7378 —0.0194
565 1.8465 0.8571 —0.0087 0.6852 0.3180 —0.0032 0:33:19 0.6862 —0.0181
570 2A511 0.7823 —0.0080 0.7353 0.2674 —0.0027 0.3881 0.6286 —0.0167

575 2.4250 0.6953 —0.0073 0.7790 0.2233 —0.0023 0.4459 0.5695 —0.0154
580 2.6574 0.5966 —0.0065 0.8183 0.1837 —0.0020 0.5071 0.5071 —0.0142
585 2.9151 0.5063 —0.0057 0.8535 0.1482 —0.0017 0.5712 0.4418 —0.0130
590 3.0779 0.4203 —0.0050 0.8811 0.1203 —0.0014 0.6291 0.3827 —0.0118
595 3.1613 0.3360 —0.0042 0.9050 0.0962 —0.0012 0.6858 0.3247 —0.0105

815
Table 1(5.5.4)— Continued

Color-Matching Functions Chromaticity Coordinates W. D. W. Coordinates

Wavelength
A (nm) r(X) g(A) b(X) r(X) g(a) b(X) r*(r) g™(A) bits
600 3.1673 0.2591 —0.0034 0.9253 0.0757 —0.0010 0.7397 0.2695, .—0.0093
605 3.1048 0;1917—-=0,0027 0.9426 0.0582 —0.0008 0.7906 0.2174 —0.0080
610 2.9462 0.1367 —0.0021 0.9563 0.0444 —0.0007 0.8344 0.1725 —0.0069
615 2.7194 0.0938 —0.0016 0.9672 0.0334 —0.0006 0.8718 0.1340 —0.0058
620 2.4526 0.0611 —0.0012 0.9761 0.0243 —0.0005 0.9045 0.1004 —0.0050
625 2.1700 0.0371 —0.0009 0.9836 0.0168 —0.0004 0.9333 0.0711 —0.0044
630 1.8358 0.0215 —0.0006 0.9887 0.0116 —0.0003 0.9536 0.0498 —0.0034
635 5179 0.0112 —0.0003 0.9928 0.0074 —0.0002 0.9702 0.0320 —0.0022
640 1.2428 0.0044 —0.0001 0.9966 0.0035 —0.0001 0.9856 0.0156 —0.0011
645 1.0070 0.0001 —0.0000 09999 0.0001 —0.0000 0.9996 0.0005 —0.0000
650 O782 Tana 0015 0.0000 1.0019 —0.0020 0.0001 1.0083 —0.0089 0.0006
655 0.5934 —0.0020 0.0001 1.0033 —0.0034 0.0001 1.0144 —0.0155 0.0011
660 0.4442 —0.0021 0.0001 1.0046 —0.0047 0.0001 T0198" =0.0273 0.0015
665 0.3283 —0.0019 0.0001 1.0057 —0.0058 0.0002 1.0246 —0.0265 0.0019
670 0.2394 —0.0016 0.0000 1.0065 —0.0067 0.0002 1.0281 —0.0305 0.0023
675 OVIZ2 a U00lZ 0.0000 L.OOTO: =— 0000.5 0.0002 1.0305" —-0.0332 0.0026
680 0.1220 0009. 0.0000 1.0074 —0.0076 0.0002 1.0320 —0.0348 0.0028
685 0.0853 —0.0007 0.0000 1.0076 —0.0078 0.0003 1.0327 —0.0358 0.0031
690 0.0586 —0.0005 0.0000 1.0076 —0.0078 0.0003 1032500 1-20-0356 0.0033
695 0.0408 —0.0003 0.0000 1.0077 —0.0080 0.0003 1.0328: 5540-0363 0.0035
700 0.0284 —0.0002 0.0000 1.0077 —0.0080 0.0003 1.0330 —0.0367 0.0038
705 0.0197 —0.0002 0.0000 1.0077 —0.0080 0.0003 1.0330 -—0.0367 ~ 0.0036
710 0.0135 —0.0001 0.0000 10076" 70,0079 0.0003 1.0325 —0.0360 0.0035
wAS 0.0092 —0.0001 0.0000 1.0073 —0.0076 0.0003 1.0309 —0.0348 0.0039
720 0.0064 —0.0000 0.0000 1.0059 —0.0062 0.0003 1.0247 —0.0283 0.0037
J25 0.0045 —0.0000 0.0000 1.0071 —0.0073 0.0002 1.0307 —0.0334 0.0026
730 0.0032 —0.0000 0.0000 ).O072he= 00015 0.0003 1.0304 —0.0341 0.0037
735 0.0023 —0.0000 0.0000 1.0074 —0.0074 0.0000 1.0340 —0.0340 0.0000
740 0.0017 —0.0000 0.0000 1.0073 —0.0073 0.0000 1.0334 —0.0334 0.0000
745 0.0012 —0.0000 0.0000 1.0068 —0.0068 0.0000 1.0308 —0.0308 0.0000
750 0.0009 —0.0000 0.0000 1.0070 —0.0070 0.0000 10320 0.0370 0.0000
755 0.0006 —0.0000 0.0000 1.0064 —0.0064 0.0000 10292 Mia O0292 0.0000
760 0.0005 —0.0000 0.0000 1.0066 —0.0066 0.0000 1.0300 —0.0300 0.0000

816
 FIqeI. (9°s‘S)] 9ar3aq-uay, SuTyDVYA[-10]0>
SUOIUNY
JO QZ [VNPIAIpUT ‘S19AIBSGC
OUT, ISI ST IV pajdIJ9G
Woyay)| [eyUaUtLIedx
ved pouresqO
Aq
saTnSpuL YONG *(6S61)
ayy SurureMay
“OMT, (SS)MD
PUL FAC (6S) a49A4 Osty paureyqggC
UO OU) “TN (S2[VS) “AOyeUTOAYDIAT,
ng Ja}Fy uoHadwo>
Jo
oy} UIE SuTYyd}VPA[-1OJOD
uoHesySaauT
JO sayypuv ‘IPing
(SSM (AW NWA (S) (€LINTAAD
v r oe
2 Pee
(wu) (M4 alXx Msx al X(¥)Qaol (YEX AOL (YZX awl X(Y)9aol (Y)4X al XYZal x(qdol (YeX «al (ys alx xqaol
06€ ES'T — 087—S— 001 €— 961 €— 9I'S—y— LOL €— Sle t— Bm 0S pe p91 Z— 69'S b— 9I—v— See €—
00 P77 €— 8Si—v— £1 t= I0'1 7Z— SvpE—€— €¢'S Z— TOL SEP—t— 8I'8 Z— €@'P €— I— OS €— 812 T—
Ol Sr'l ~— trs—e— L=sera 89° T— (I—7— I—8II'Z 7—"20'r brI—-,C— I—v19°7 98'I 7%— 169—€—- 1—ZOv'l
0Z 19 %— ssI—~— 3 08 3 eek S69 ~— 6hE—7— I—Spe's SS'9 C—" a) 0k Ur I—S8es¢ TELE OL wa. I—18S°¢
O€ tl trht—T— I—prl'L 069 7Z— Il'r?—7— I-78 Z—"91'9 99E—)ZZ I—b0E'8 £0'P -T— 9SE—.t— I—110°L
Or cog aroma CU I—8b9°6 WZ .7— SIZ—Z— I—86L'6 SEZ I—5.c—
98 T— I—?PEL'6 ZE'I 7— COI—.T— —087'6
I
OS 06°¢—t— OC Ge €10° 10'7—~— £97 T— r00'I b— OV 7— €6C 7— I—L6L°6 %— 6S ~— <6] re €90'I
09 —I I—-L9E G—1tt’6 I—OLb'8 I—€z7eI— Z—0£6'8 I1—996'8— I I—919 1=0I'] [—SeL'e 7—90£'6— 7—L76'L 000°I
OL I—L7S'7— I—S€0'7 I—6S€°9 I—966'7— [—?zl'¢ I—8L¢£'9 I—%00'€— I—7@peZ I—spe's I—L60'7— 1—676'1 I—I7Z0'L
08 1—LO0r'e— L-size I—€80'p I—7Zis'€— I—E8r's [—966'€ I—896'€— I—£€0L'€ l=LS7'¢ I—SLO'e— L—-0S7'¢ I—988'€
06 I—s8I3'¢— I—S6S'b [=Co1Z I—IL€7— I—786'p I—psL'l I—S9p'p— I—ISs'p 1=1797] I—79'E— I—LIt'b I—888'l
00S I—¢rs'e— I—9€1°9 1—990'I I—0S0'r— [—Z87'9 7—$96'L ¢— [—zIs I—8ZL'9 7—S8h'8 I—SOL'E— I—890°9 7—OIb'6
Ol I—oss'¢— I—Sb6'L 7—9'b I—ILVe€— I=TLLL Z—97Z'E ¢— [619 1997'S
5 T—SS8°S I—Z0L'7— [—98S'L 7—690'F
0z [—ZL61— I—1'9'6 2 ear I—TISI— I—I€L'6 €—9£6'L— Il I—p9r' 1—095°6 €—196°8 I—l671— I—8S7'6 Z—OSI'I
O€ I—p91'l 910° €—6LP'>— I—ILI 810° €—LZ0'E— [—LZ'I 610°I E—L88'E— I—Ob0'I €Z0'| €—06'>
Or I—StUP 8h0'I €—LLS'6— 1—66S'b 87Z0'| €—8ISL— I—$96'p rS0'l €—9E1'8— I—S79'b L901 €—160°6
Oss I—088°8 £00'1 7—9ETI— I—8L1°6 L10'I 7—8S11—6 L=Lt7 Z00'I 6— €—b0S I—SE€E'8 9>0'I €—61S'8
09 €8C'l 1—Z00'6 Z—Itvl'I— 88el I—p00'6 €—-L688— CSE 1—L07°6 t—vIOI—-l ele I—LES'6 €—O08S'8
OL | 768" I1—P08'L tr6—€— S16'l I—Sso'L pl8—€— I 188° I-6IL'L L76—€— 88L'I I—-060'°8 LLL—€—
08 oct IETI166
— 66.E— BEET I—E€IL'S 909—€— ZUb'7 1—200'9 199—5.€— €07'Z I-0SL'9 p¢9—€—
06 pe8'Z I—Z6l' ¥65—.€— LOL'Z I-P9I'h soS—€— OL8°Z I—P0e'b 79'S—.E— 88S°7 1—96S'b $6>—€—
009 866°C 1—S6s'Z 6EE—€— 8S8°7 I=—OrS'7Z €6E—€E— 0z0'€ L79'2
MI ¢— 0S t— ZSL‘Z I—OL8°7 I9€—€—
Ol 008°7 I—8LE'I 9F7—€— L99°7 I—OSE'l 4S7—€—- LE8'7 Sl'E—rl—£00'E
£— Lv9'7 78S"
—1 107—)€—
07 88E'7 7%—991'°9 681—E— LeGe ~T—OEI'9I—.TS €— O%r'Z ELS
ec P0'C=E— 60£°7 £ ite (Olt=
O€ $08'I Z—I8I'7 8€8—p— 8SL'I ~—107'7 SITL—p— 678" e ele Veer10 c— Z8L'I T—8rS'7 189—\y—
Or 9€7'I €—PLb'y 007—.y— I p77 €—9bS"p I9I— y— IpZ'l YP COL T—pnt
SC P— 1€Z'I €—-El€'S
sp €Sl=—iv—
0s9 I—IS8°L
= I— €—9PS 6 S— 1—806'L €—p99I—I's S— 'L 1781— E=—1S0-0:¢> c— 1—788"L
. T— €=—776LP S—
09 I—8LP'P
- €—8S0'7—99°] p— 1—O8S'P €—OPET—69 Sr I—ZsS'¢ 6661—*
of €h> Sc— I—-9ES'r €—1897—79 Ss-
OL v7’ 1— €—-19SI—8p'l p— I—PIS'Z €—6€8I—ss S— c Robi a8) Cie I—6Lh'7 €—pOl'Z—6 S—
08 I—SETI 01'6—y— 6'%f c— 11-0871 €E—IS0l—
£'§ C— P97! ClS 8 -bo98 Obes Sc— PSSLTI
A Isr EESC7 ae eda
06 7—ZI6S €Sb—y— ve S— 7—9L0°9 LOS—v— gl S- s <c—lteh— 0c '= 00) c— 7—PLI'9
9 O19—p— 61 S—
OOL —1S82
Ur tCC—¥— 97] c=" ~@—TIt67 IS%— v-— 60 S— eI9S LOC.P= CO f— Z—LOO'E LOE-p— 60 c=
Ol 7~—ZSE'l 10I—v— 30 S— Z—sseil SVI—y— v0 Ss— Cisco 6— c— 10 Ss— T—ZEP'l SpI—y— +0 c—
0Z €—H0r'9 vr— S—- £0 S— AST")aed8 Goerah eA 10 S— €—6rb'9 Tr— S— 70 S— €—$78'9 L9— c— T1°0 cs—
818
AIqeL
(PIEOW (SDLIG (9DINGA (6DdHW

—(9°S'S)J
v

panunjuo
(wu) (4alx xyesal X(y¥)9
aol (Y)4X aOl X(Y~)Z
a0 (YG
X aol (YsX «Ol (Ysx al X(y)9aol (Y)4X aOl (YZX aOl (v)9
xX aol
06€ tyc sat POST= I£8 t— She Cos toeSh Vow ee LC Saeko
Qt c= Cou be LOO ey ComteVeet Tee. Se

OOF Ceca Ca ton 6€°S c— 099 be taOL tre F 0t Cm, £8 ete 62 Le Vo 16-1 C= Of:¢ st os Sil teh 353 (Se
Ol tor Ge te ¢—*9C I—OITC e CI See90 t— b—C8at Geeae9 p96 I—v6et'l Cc C68OeCL - tr DLtb=
0c 30'S C= tag
EC €— I—0v0's 89S CT ser CCCoO I—SPr6P 7 OS: Cseere.
Lt > I—9LP'P 805 C— et WV.cG I—bP9'r
O£ 10'6 CA ae ChEGro? I—061°8 F 1 6 Cpe8 © 8 I—Cle 905 tare
O00 t= aL Lak OS— anc COs 1—Lcls
Ov Goer Lega
oP G7 6666I— SUE Cm ive,
30 C= I—€8S'°6 SY lagi
6S C= I—£€86°6 957 Cettecc0'T

OSP Sid Smet 6L7 C— I—189°6 ee Cave C0 100°1 SeOh sgl(A 6 Ores 1—0£6'6 t= Causey4 ie eel(4 I—t€r6'6
09 — I 1—06h L—¢90'T I—OI8°L le yrc b= G_9ILS I—6S7'8 be LSet b— C—8LL6 I—6t¢'8 tes tol Pe C—8E£9'8 I—LI8°8
OL os L=-Sci' C LCL G:9 LIS ce LOL) I—6S6'l RS LIP ge Lats eS I—£60°C [—890°9 ce [61 I—6l6'T LD CI b=
08 7 LCS. CCHS I—I6l'¢€ ogee
pe By cat S=98I'f flees es- Ott. } lPcebe 6SerSL ard LGC
S bes b—LOCE LOSEr=
06 I—Sbs'p— I—cpr0's I—LCL'I tLe I—LIS'¥ I=tSGl 1—688°€— I—8ss'p I—LL6'l 17096.fe— I—819'p 0c L==6£

00S v- t—C6C Fe ond

| C—Lb8'8 ye L=t67 1—060°9 C—S6L'6 7 FSO I—6L8°S cC—7C09'6SameWB OA ach I—601°96 -StCCG
Ol ta C198 I—eI1'8 C—9I0'E te ceL Da L=1EL8PF G90 Lasts I—L6L°L C—8C0'P I—86b'€—
f8 7s Vat CATS
07 ee ho eal I—$90°6 €—L90°6 S— Lth0 I—L6?'6 ¢—tL0'| te ts3 t= I—OPs'6 c—O8I'I te LOI T1616
I— c—E€C0'T
O€ I—ver'l 6c0'1 —7 t—c0s Tollb= €c0'l t=rcesS= bagcie vcO'l = ESESEAS I—v¢sll 820'l €—668'h—
Ov I—sp9's Iv0'T &£—S$sli6— t—tPrLs oe I ¢—9CL I—t€60°¢ $901 颗t00L— I—LSTS v60'1 620i c—-

Oss SLO'l Sr0'l 6 L=7Gr [—819'6 090° l= C-=3Pa L=9t2'6 6c0'1 oe co| Cort, 9rot= 810'T 2 Nee 16% (Somme
09 $39'l ELLG
LS Te G30 I vos )6 918eee CSV9C 19p'l 8166
L—eo Lae 6=C80) Ipp'l ECG Sie POONG—
OL OLe'C LOCO8 CS bo 901°C SES SSEP
PD bees
Ol o= 100°C OCSL saree LheOC= L66'1 VIG LObaaece—
08 668°C toe Ce ue aan9 US & 809°C LEED6s FL $= 899°C t—FL0:9 LuG-rst— CVC 976s.
hi See Le t—
06 E LCE oCle ee 0975to 890°€ PF Los teas 868°C -LOne [Serbt— $68°C |>-8rur
c0 68857UU

009 OLE’ saat tee toeUE To brie & Bisse5— 65 t= blo’ a ea SePES ee Leo'e Suc, LES re Ce ee
Ol 99LL evra SC be Lv6c L157.) 09-5,oe 098°C Pel 8S ee EelSee 098°C P6tel
= of C= OS e-
0c S6S'C SS Le CLL te lpr'~ 1.0 G5 ee Relto ESE'C 9CI8
5 ShhBo 6 14 EroOF Or tee Peet
Of C8at
» Op SL mace lL te 1@8°1 1k G88
9 9S:EB 0781 Stee’
aC 88-0“.V— 618° CC OE8$
WT ele Pe
Ov STI Oe, etaLS Vo I 6c 2 —9LP- 10-6.BP l Lec Cole —yet— “G8 I 6c SSOP poor SCiLSBo

0s9 es bol CVS 8 0 ay Seer L OLe te Le PbO eae cas

on aal Vor PapOGL er 088.
21 le hoo te 207 o—'
09 our, CoeCLO netic Oe bre P Gtr ale 2 t—3LC.0—
2: ed 7 995 b—-py 66tsC—
S— O9 c= PF S95Sh IAC tens ts S—
b eee ee OLS Ig81 aedalt SieteOR | seg ede C Oy ee Sore te 544 a BrA RelNeen Hel5. arkSF A foie
08 it tere om Sl Wrap orc £ LECOCC. bere CO thet SCF Say T eeS9C
el Le C0 tome BE Se oS6GT
ie or LOsttae © 19 Sie
06 SPSeer VO ot T—866'S 81S—b— vc c= -c-706'S 08'?—"v= 0¢ s— OLE See useVE UR Same

00L Ooo "000ask mee

+ toe tc -tt9Ve cca
+ re “kg CatScemnin
TY dora
| Ml snag t G-L90
) cu Yammer 18.0 +
Ol | bee Se Sanit
A ae | 1 ag Foe t CHL PVT o_o COU
eC be Poe ged me LPs
so scaPe 0 4" [=
0c ee ee ee 00 s— tae daa
LU Se 11 Sete 9 SES
oe Ve,he ye oa CLOL HES Sle oe S100
 AIGeL —(9'°S'S)I panuyuod)
eee eee

(PO
MNU

14
(€7)va
(wu) x4al x(y)Zal (YaX aol

(LOW
(YsX aOl (YZxX al X(Y)9aol (YdX al xsal X(Y)9aol (YsX «Ol XZal X(Y)9aol

(67)MSd
06€ eset ¢ "Tet"Tes e— Seayee pe nile Eo
€L'9 p— —IV%— 06% €— e—IeTzZs—= =r O82 = soz ¢— T—"3L¢—301 t=
00r IS°6 €— 967—€— 6b'b 7— 9L'S €— p6I—€— 69C 2 Ics Z— OLYo- 107 T— EI t— Oc.f= 7's 7—
Ol PE ST —" ct I—€L8'1 L67 7Z— 801I—Z7— I—18¢'I 71ST T—181— I—1€6'l =T—VIES c—2084— I—LLb'Z
02 869 Z— €87—T— I—pL8'b T1189 Z="1LUT— I—OLS'P LOT I— 19'E—7— I—L78'p 7—tps's C—LOlge— I—b6S'S
O€ oen “Le e— Sty 1—908°L -c—ficw t—=075:¢— I—vI1'8 ETI 1 T—T8T9— I—901'8 7-066 C~-t80:'— I—9€7'8
Or TLL
= 9S Zr I—£€19°6 —FE0'e 7 Z—=08I— S101 20K T——- t—t60'2— 6Z0'I t—Pi9s T=—eEI— I—Z08°6
OSb Ol'v—Z— 4 tef Aceitl Z10'1 %6'F—Z— 60'E C= —789°6
I Ss‘s—7— di (Gracy I1—$99°6 16S— 7— eles C= I—S8L'6
09 I—9Se1— Z7—0£6'6 I—$69'8 I—€L9:1— I—96I'I fe Seed I—SL9'1— Z7—€L9°6 [—8SL'L I—976:1— r=0rG I—88L'L
OL I—Seyt— I—061'Z I—6S7'9 l—kk6c— I—S6€'7 L—WSs 1—687'¢— L=$8zz I—+r10°9— [=e Z [PLT 1—888'S
08 I—97S'€— I—Sbb'e I—9€b'€ I—SE0'r— I—60L'€ E—¢So4 p— I—z0t I—6Ib'€ I—60b'€ I—810'¢— I—O077'€ 1—96b'€
06 I—€90°r>— I—860'S I—676'1 I—OLS'>— I—L80°¢ I—S6s'I I—ISL'b— I—Z09'b I—9€L'I I—€>9'p— I—Ltt:¥ I—v0L'I
00S I—8ss'e— I—L8b'9 7-156 I—98b'b— I—69L'9 7—6L8°L I—se0's— I—96¢€°9 7—bb's 2 R165 I—016'S Z—SSL'S
Ol I—ZLE¢— I—O16'L 7—9L6°7 I—Sv9o'e— 1—660'8 7—169°7 t— I-77 [8671 5 epee — I—906'€ I—pbl'L Z—ETL'E
0Z 1—sor't— I—€£9°6 Z—LO0'! I—908°I— I—€£r'6 €—I8E'L 1—907°7— [£506 €—PI6'L [—683-1— I—617'6 €—6b0'6
O€ I—L60'I L10'1 €—6b8'b— 1—607'I L10'1 €—ILS'€—1 1-617" O€0'l E—osc:7— I—91€'1 ZZ0'I €—Sss'b—
Or I—I18'h 8S0'1 €—908'6— I—OL6'b pS0'l €—STEL— I—L78's
. prO'l €—900°8— I—86S'S Or0'l €—879'6—
Oss I—Sze's 090°! T—L9I1— I—8L'6 p10'l €—7ZS0'8— 160°1 I—9€L'6 €—1166— Lv0'l I—106°6 Z—900°I—
09 v7" I1—6L7'6 7—9801— 68P'I I1—091'6 €—OLL'L— €89'I I1—666'8 Z—000'I—I 6S" I—SZL'8 €—0866—
OL 9¢8"I I—6S8°L €16—€— 1Z0°7 1—89L'L OVL—€— L677 [sor g—lees— 7977 I—IS9'L €—€9h'8—
08 007 I—SL6'S 68L—€— OLr'Z [oies—- g—io Z18°7 I—Fis'¢ €--tOi— $787 I—SpE's 959—€—
06 789°7 I—W0Z'P 7'9—€— 006°7 I—9SI'P¢— 08 €— Lee'€ ZE0'v
=F G—teyo— THEE I[—I%p 88r—€—
009 968°7 I—L09'7% 6LE—€— pb0'€ I—8LS°7 IPf€—€— ELEE E—66P'eE— SG c= Z9E'E I1—T8¢7 Ole—f<—
Ol CEES ere c—SLt— L18‘% 1—Prel e—Et— OZI€ 1—097'1 €—""¥6T— €LO'E I~LPEl €E—981I—
0@ IZE7 7%—8PrI'9 67I—€— LSE ~%—£98's
I— Lv €— 966°7 2—-TES-S e—Vy61—°Z CES “Z—0209 y—-s0E8—
O€ 008"I t%—8L07 €L8—¥— 908°I 86" T—€
- $—OS—I 898" T—LLT
— E—MZI— 8r8"I Z—LOIZ PIE—y—
Or I bET €—IS7'p 98I—y— SEZ'T €—S00'P
I— PS y— OSz'l E—oIS'e 7—(OLZ— 9b7'I €—ZIer f—iro9—
0s9 I—L8s‘L €—v8hI—+9 c— I1—€98°L £—-LEVI—
ee c=" [82771I=" E=S8T ola (err e—eepi—£7 f—
09 1—79S'F €—I861—‘06 G—= 1—S0S"P €—99h'7—
‘9| c— I—LPE'b €—99L'I— 5&9 I—6I€'p €—£L61—Pp S—
OL I—80S'% €—80S'I—
3 teen p'@ 7S -1— L661 S—--€'0 f——-1=S0E-Z 1S= ss —~1-9L7'7 €—06F'1—6'f eo
08 [—88Z'l 668—y— 9'¢ Cap I—LS71 OVI—€- av S= [—691'1 I8L— p— ac Swe U-Z9I/1 698—t— .t7 S-
06 78079 b9b—v— ¢7T c— 7—890'9 80S5—y— 7 c— 7—S6S'S €8E—y— | Ge 7—9P9'S %b—y— | 7 =
OOL 7—966'7 6%7—¢t¢— 80 S—- T—Sé6'Cie Veetsr 60 c— ~%—p669'7 88I—y— OO Gens T—EPL' PHSLE—9°0 c=
Ol Z—p0Pr'l £0I—¢— 90 S— | 7c [— bf) ¢'0 f— 1 697 8— 9 C— ascii | c—s0E 66—" c—" $0 c="
02 €—696'9 pry— c— ¥0 Ss— €—I105°9 es— c— 70 Ss— £€—PH86'S S'€— c— 0 c—Iez £€—661'9 St— S— 10 c="

819
820
panunuod —(9'S'S)I A198.

(te)OLfs (OE)UVE A (OP) (Ip)dSf

4
(wu) (y)sxX al X(Y)SaL (Y)9X aL YdX al (2 aOlX X(Y)9aol ()e4x «al alxyze alxMq alxWYs xsal xy)aol
06€ €S7 €— +v—6IS—€L£8 €— “Cee cr PL c— I0°E — 90I—t+— 71 €— LL'I p— $8— c— ££°6 b—
00r A NS one SeerOR esa 655 T=" CET ~%— 709—€— 61°8 7— BS
a ea Ga4 a IST T— 107 £=— 0C6—y— | (S&S
Ol ty6 CT =O i~Cice~ 9L'9 7Z— €0'7—T— | I—p9S7 7—L0T €—"6L8— 1—cO14 991 7— 8LL—€— E8701
0Z TT Ve I—Pecse ZO! I— 7%@—OSE— I—666% OP'S T7— S97—Z— I—98°€ OPP @— 957—T— 1=S79°¢
O€ TES Z— 7%—O06E— I-—SI7S 001 I-— @Wry—tT— 1—866'L 769 T— etr—%— I—0102- 179 T— 96€—tT— 1—9L6°9
Or ERS Z—«= — IBZ I-ZLN6" SEZ ~=T— OTI—T— I-S€76. 780% 7—OL0:T— L—699:56— —20E 2 cee. 1—06L'6
OSr oC «t= 060° g—" I~£9L6 19'9—~— - PSE 7— 910°1 — C109 9I'E 7— I-8h66 6S'>—@— 68% 7— L00°I
09 I—606°I— 1—900°1 1-669 I—9LE°7— I—SIt'l I—p6E'8 1—e8z71'7— 1—O8I'I I—vI8'L I-ILbI— 1—?L1'1 1—897'8
OL I—SsIzve— cue 1—88s'¢ 1—9€8't— 1—179'% I—€€6'S 1—19€¢— I—LErz 1—969'¢ 1—7ze7— 1—8S1°Z I—SzE'9
08 I—900°r>— [cece [—ZECt I—110°'S— 1—066'€ I—Zot€ 1—6tr?— I—€8L'€ I—O€Z'E 1—76l'¢— I—€9€'€ I—L98'€
06 I—S8h'p— I—09@'p I—OLS'I— I—99°5 I—IL7'S 1—769'! 1—6F1:5— I—Z0€'S I—Lp9'l 1—PSL'E— 1—819'f I—-%07
00¢ I—TIS‘p— I—9L8°S Z7—87Z0'8 I—609°$— 1—868°9 7—SIE'8 1—19T's— I—b76'9 7—850°8 I—£€98'€— I—pes'9 7—L96'6
Ol I—p06'€— 1—809°L T—-LITE— [oe I—ZOE'8 Z7—bZ0'€ I—981'p— 1—667'8 7—€6'7 1—971'€— 1—898°L 7—6720'F
0Z I—vl6o1— I—07b'6 €—979'8 7— 1-8 1—619°6 €—SIP'L I—006°I— 1—965°6 €—066'°L I— 1-696 I—66S°6 7—810'1
O€ I—Ove'l 87Z0'I €—€€bb— I—p6e'l Z10'1 €—ZEpe— 1—7PE1 610°T €—99b'€— I—€60'I 1Z0'I €—6I1L'b—
Or I—76S'S 9¢0'I — e—1P73 I—8s0°9 1P0'I €—SI€L— I—196°S 8so'l €—779'9— I—8@b'p $So'l €—1IL'8—
oss LOI 1—1066 €—1106— ETI I1—£€89'6 €—LP9L— 1011 1-—S76'6 €=TLZL— I—III'8 L00°I 7—WILI—
09 1991 1-—FoL6 e—Piv6— OSL'I 1—091'8 €—9SL'L— 089'1 1—996'8 €—p80'L— S67'l T05°6 «I= 7=9F0'I—
OL ZIP'Z ee Coot LOv'Z I—IIP'L 489—€—- pOE'Z 1—9PL'L
. €—"60°9— rest I1-—977'8 LE8—€—
08 9€6°7 1—FPd'9 E—PESS— 796°C I—€h9'°S 08'>—€— 4 By 46 I—Set c—it6y— 6b7'7 9 Ire i— c=
06 p@b'€ ce— ce IZp'€ I—€E0'p O£E€—€— OIE ICEL E—NELE— SbL'z I1-OIS'Y €—<67—
009 Z6b'€ J—p1o¢ t—tri~— 10b'€ I—FSh'Z g0%7—€— a I—8er'e F1e—c= 7167 1—Tre'z ice—c=
Ol ZLI'€ 1e= t—teei— 601°€ I1—l6z'l €lI—€— 9667 T—tTer c—erri— p7L'Z 1=905'1 1LT=—c=
0Z OSs‘z eaeG oO €SS°7 7%—OZL'S 6SS—p— esp'Z T—LI8'S F—M8s— STEZ 7—L9L'9 Iv'L—v—
0€ 6S8'1 2-986 $—F5p1— 6981 ~—ELET @T—v— 16L'I 7—6S6'1 ILb—-y— 89L'T T=TrEe'z y—Ee—
Or Lee'l P—teot c—Gp— 0sz'l €—I0'> ¢—-j'1:s— €€7'I £€—166'¢ ¥—-'10I— Leet €—808'r SL— S—
0s9 Py i Ce tek I—P8L'L €—LeyI—6] c— I—OE8'L €—6fhI—3 Secs ‘I1—9L8'L €—L6L'I—87 c=
09 1—8zéy €—S68I—(9% c— I1—P9E'r €—S86I—67 c— I—PIP'? €—rL6I—
Ob S— I1-60S"p €E—SsS~T—OF c=
OL ERE GE— GE CS I—petEzZ €—@SI—€% S—- E'7 10S 101€— — 3'T C—— —8rr'Z
| €—100'T—-—-
ie eae
08 I—L9IT vS8—y— CT I—I8I'l 068—v— TT c= I-69I'I 988—y— ST Ce 1s7'1t= Zil=€=— 267 c=
06 Z—Is9's ifh—’— $0 cs— 7—899'S Ibb—— $0 S—- Z—PFst's ~—OPr— 10 c—- 7—000°9 €7S—v— 7% c—
00L 7%—PLZ 81%—-~~ v0 S—- ~Z—E£SL7Z b—8I%—£0 S- 7—PHrES'7 b—8I'%—
00 S—- 7—606'7 09%—y— 01 c—
Or 2 me Reece
bt4 4 aon enemy
| ier LIE 00I— .— 10 S—- %—£v7I OOI—tv— 00 S— 7T=ZET SCI—P— £0 s—
0Z Tir
— epee— 10S c— €—s7E9
: S— S— 10 S—- €—670°9 pp) S— 00 c= 6/¢'9f= 9—
1 c= 00 c—
 L F198 —(9°S'°S)I panunuo)

MUM (8h) (pSaWe (89)MD (6s)fad

v
(wu) «alx(y)s alx(@s
(y)qx aol (YAX al (YZxX «al (YGX dol (4xX al (ysx «al x(yaaol (Y)4X «Ol X(Y)SaOl (Y)9X a0l
06€ 6 £8 b basteOTe €— Oe Pe 6 t6 Vas reel ey Coe
6. V— 70°9 €— as hl dommeMN Fra 57S open
1
00r eope 6 (Sos)cs SEE cm la0 °C7 69'f—ta 167 ~~ eG eet Ge G9 C= 10L Cc GOTT ¥=-"0lE—Fil oes
Ol 60'e 80'S Ce I—v06'1 eae Fae> tL Cee, I—00'7 ="7—"Sr'e 601t— I—vrs'l ee fe ee)300 C= 6 THe
07 2a— Ce $e I—€89'p Cees€— 99 Ca I—006'F Ff SC" tee CSS I—9Sb'p G02 Cae Tree
O€ -GE1ES'9
TS THOS I—SpL'L ce WRC pe 89 ZEX ZL I46cL pas cst WP 2 059 LES Ge WeeCit IZLi9be
Ov € tIS Ieee C-WEr I—LS8°6 | t09 Wee LV Cet 6 [907 one rie ac Ver Ee o— I—€7P'6 0'1 Sex wear C—ELE 1—069°6
OSP P= Cor Tc COs —908°6
I = C905 SE LY CaP r00'I — C867 00°€ Z— I—£€68'6 i— St e— i c—tO1 610'1
09 I—Ss6eI—
| a 4 ear I—LOL'L [—LIgl=
| Feiv7 I—€88°L Olt P—peln I—1P9°L— Orneo- TILES 19£38i=
OL l—treee— 1—OS1'Z 1—091°9 L608. t60¢Es [OPES — 7 ol Le C L640 I—v7zs"9 l— E609 1 [—t87 1—6S8°9
08 I—s$60'€— —007'€
1 1—Ob9'€ t= 19S b= c5S'6lar I—L99°€ Ice0-bBe 0 ad < i=eltl L==896'7— [ft I—vse'p
06 tz Less I—90€'P 1—108'1 t= LOS [= LLi'¢
be | Legs L=0rik= I—sso's 1—96L'I Wee Seva: [SIL I—LOv'Z
00S = 089°Cbar ImLEL'S 7—990°6 Clit le I—S€79 7—820'6 I—096'€— I—0€9°9 7—660°6— 60GEi I—6L6'P I—O0€'I
Ol Om SeZ Loe I—p9L'L THLLYE TS RAYS ey I—7S0'8 7—SEP'E l—60ri6 [i—8k's B=Ocr'e Ste ke [—[s0'g 7—bLs'9
0z F=00Li— I—6££°6 €—96L'6 letcal= I—$89°6 €—000°6 lero I—S0€'6 €—86S°9 Ly I—6S8'8 7—966'1
O€ Learnt 6Z0'I ce St. C= ece,|b= Z10°1 fe £91 I L=Z81 1Z0'1 ¢— Loaser 1c01cE LEo'l f—Sto.6—
Or I—LI9"p rso'l le 7970 I—889'r rlorl €—Lr9'9— wis era €£0'l ae Wp Po) I—ZO0v'r 190°1 LSoL=
to
Oss I—O€S'8 Z10°1 ~80ce L8G Z00'I €—696'6— 160.6
=| one 6 C69
ls 6aS Saat I—€8L'8 15166
p COCs
Ee
09 9SE"I 6 Fiabe COLMA "1 66£ 868'8
1 628 C-I9S a rea 6 950 lauli—
900 B= €0r'| s1—ZOL'S L802
OL €68°1 S808 TISG= Ces L60°7 ols
I 62 EVE L00°7 Ceeae C908 £r6'l lets a-Sinl—
08 66£°7 98C.9TS 8S ee C= 69b°7 tO Seo ct: S0S*Z 9 C00 Somava LOGEC —tLe'9
| e—08tl—-
06 0S8°7 F 86r SF "GEC 66L°7 e I-69
Se
6 te C= 6S8°7 Py, Sor Vt Sk co €68°7 © 18S— S90
009 pS6'7 sc ver weet ta EV SL8°Z PASS hed ae Re S67 © Oe ee co) b06'7 e 00 Oc a
Ol SLLZ 2) 3 a NE eiceyaN Fon €18°Z ICr mL le 76 60L°7 9 he aad Nees earth 98L°7 littl— Gr
0@ 97E7 S059
7 -f0ICS prez 7H9 GC 6 PL CLEC 8865
eG 1 §C C= Slee 9 —0Lr
Pc theeoe
O€ €LL'I 7 bbeSC v= 68L'I C B6E Or Ce PALS 6LL'| LEE SC 6Y' Dot S08'I t7=—S06C Clint
Or 877'I —P9er
St Olet= eZ b S196 Se CET 6771 fb —9SE DAVEELo POX 9€7'I CLy
I — ObesePr
0S9 998°11 (e689 70r, C= BFL MLSEl= G19 aK at See, 80BE [i borerP19: €— Saal 6essT= = S-—-SPotlet
09 PF 98 E— ESTES Sr r alt 0CC—= ig geapo Ig8: en Bi ape Sryet 0c yet
OL ere CMCSA e cmc & 06E Le. lS €-2019' bee Car L9PiZ
= el te S60 +P Cal 1760S
=H E—O89T 5a gg
08 1 lye "Ter i— C490 CE Sted
— | GOL PR66 Cc CLCts Seo C= Nir Shoes I E91 Se te 9 c—
06 G9G'S aE
— PAIS Sai 09 | 189 Se eae ro: 98 nr va Pro yi] a 18"S tT ECR oc
00L 06:2 =—E= pL BO Sm PLOTT= Ce 65 b= —3-0 S— E076 PGSPB a ee ESTOSPHOS oa canes
Ol 9a £Cil——to 15KS Siar t cn l= V=26l aS) en cL) eC la Gl v=" St)ws Sa (0-600! OleP= 2500) a.
02 Goa SeNG Go 4 Ba dla Oe et eS eo caw (Nag Sood aili,
A ell U rl CLT
ee UL UU
“aon [[Y sanyea
aie paidiynus
Aq “AOI auL samod
std udAts Aayetpowuw!
SuIMoyjos
yore payeynge)
‘anjea ‘snY],
10j ‘O]dWexd
166°9,,
— ,.€ SUROW 165°9
X c-OI= *16$900'0Ajdwy
sadeds
UI ay) UUIN]OD-d3}0UdP “SO1az ‘snyJOJ ‘a]dWexa ,.¢/8°Z,, SUPDW SL8'7
X OI = ‘SL8°T

821
 822
A
AdUdID
BYYOSOYOY
SAPAIISGG)
snourum’y]
suoysuny
pourvjqd
sidosay,
SULYV]A
(Z°L'S)T
(ZLGL)
uBIPA,
AO}
IPVUIOAYIOIIJ
PING
AIGVL
IOI,
Aq
DUIIDJOY (UU KES) SY) JO S[OAT] DIULUTUUIN'T JUDIDIJFIC, ALY JW puv S9ZIG JUDIOFJIC, VOI], JO SPpoLy aQaedigD paMatA ATBUID UT SoYydRIA, SSOUYSUG
pShNuUys

PIP of 0$'6 PIP PITH of9

c—-Ol z-Ol ;—Ol I Ol ¢—-Ol z—Ol ;—Ol I Ol ¢—Ol z-Ol ;—0l I Ol

(uu)Y (, Ww. po) (, po) (,_W- po) (,_wW-po) (,_W-po) (,_W-po) (,_W-po) (,_w-po) (,_m-.po) (,_wW-po) (,_ur- po) z-W- po) - po) (,_W - po) (,_W - po) (,_W

OOP (t~—)Oov'l (7—)07'I —)00'6

(€ (€—)O0'€ (€—)Ol'l —)Ov'l (7 (7—)O7'I —)00°6 (€ —)00'€ (¢ (¢—)OL'l (7—)OP'l (7—)OTI (€—)006 (€—)OO0'€ (€—)OL'l
Olp —)O0'E
(7 OVE (7— (7—)OS'I (¢—)006 —)O0'€
(€ —)Ov'€
(7 (7—)O7'E —)O8'I (7 (7—)OL'l —)Op'€
(€ (Z7—)OE'E (7—)OZ'E (7—)06'1 (7—)OL'I (€—)OS'€
OtPr (T—)OL'9 (Z—)o€'9 )08'Z
— (Z (7—)06'I (€—)0S'6 (7—)06'L (7—)O7'L (7—)OL'E (7—)OE'Z (7—)O€'l (7—)00'6 (7—)OB'L (T—)00'b (7—)0S'7 (7—)0S'I
OfP (—)ol'l —)00'l (1 (Z—)00'S (7—)09'€ (T—)OS‘Z —)or'l (1 (1—)o€'l (7—)OS‘L (7—)Ov'b (7—)O0'E (1—)09'I (—)Orl (7—)006 (7—)00'S (7—)09'€
Orr —)08'1
(1 —)09'1
(1 (7—)OS'L —)06'S (7 (7—)00'P —)o¢e'z(1 —)00'7
(1 (1—)07'1 (7—)OS‘L (7—)07'S (1—)os'z U—)ov7z (—dost (7—)0S'8 (7—)os'9
OSP (1
—)OL'7 (1
—)Or'Z (1
—)07'1 — (7
(7—)OL'8 06'S (1
—)O7'€ (1
—)00'€ (1
—)08'I (1—)07'I (7—)O€'8 (I1—)O9'E (—)OE'E (I—)oe'z? (1—)os'I (1—)OL'l
09P —)os"€
(1 —)ol'e
(1 —)09'I
(1 (1—)0Z'I (%7—)00'8 (i—)08'r (I—)ozr —)0s‘Z(1 (1—)OL'I (1—)o€'l (I—DOE'S (i—)OL’y (U—)ove (i—)ov72 (1—)09'1
OLY (1—)OI'S (I—)00'r —)OE7Z
(1 —)09'1
(1 —)0€'l
(1 »=§ (1—)0779 —)os'¢
(1 —)06'€
(1 —)O€'Z
(1 —)06'1
(1 (1—)08'9 (—)079 (—)06'r (1—)os'? —)Or'z
(1
O8P —)os'9
(1 —)00°S
(1 —)00'€
(1 (1—)O1'Z —)06'I
(1 —)08'L
(1 —)Or'9
(1 —)00'¢
(1 —)08'Z
(1 —)OL'Z
(1 (I—)os's (U—)09L (@—)ol9 (1—)os'€ —)0Z'€
(1
06P (—d)os‘Z (d-)009 (i—oery —)08’°Z
(1 —)OL‘Z
(1 —)O€'8 (1 —)os‘*L
(1 —)o€'9
(1 —)09°€
(1 —)os‘€
(1 (I—d0E'6 (i—)oss U—)o7L (i—)oerp —)08’€
(1
00S —)0$'8
(1 (1—)07L (1-—)00°9 —)os’€
(1 —)09'€
(1 (1—)O1'6 —)07'8
(1 —)09°L
(1 (1—)OL'p —)Or'r
(1 (1—)096 (—)006 (i—)oss —)00'S
(1 (1—)OL'b
O1S (l—)OE6 U—)oL'8s (M—)olL (—)oos (i—wEes U—-)096 —)076 (i—)0s's (—)009 (i—)09S (1-066 (i—)096 (i—)ovr6 (i—)os9 (i—)06's
O7zS —)OL'6
(1 —)09°6
(1 §=(1—)06'8 (i1—)oL'9 —)00°L
(1 00'I —)08'6
(1 —)08°6
(1 (1—)OL'L (1—)O€E'L (1—)0S'6 (1—)066 (1—)08'6 —)00°8
(1 —)Ov'L
(1
OES —)08°6
(1 00'1 —)0S'6
(1 —)00°6
(1 —)06'8
(1 —)076
(1 —)0S°6
(1 00'I —)0S’6
(1 —)06°8
(1 (1—)006 (—)0E6 (1—)0L6 —)08°6
(1 —)06'8
(1
Ors —)07'6
(1 —)0S°6
(1 —)0S'6
(1 —)06'6
(1 —)OL’6
(1 (1—)OL‘L —)06'L
(1 —)00°6
(1 —)06'6
(1 —)09°6
(1 (1—)OS‘L (—OEL (i—)os‘8 —)08°6
(1 —)09'6
(1
oss (1—)os'9 (1—)0S‘L —)00°6
(1 (1—)0S°6 00'| (1—)OL’s —)07'9
(1 —)00'8
(1 —)07'6
(1 00'I (i—)or’s (—)oe's (i—)osL —)06'8
(1 00'1 0+
09¢ (—DdO€E'r (i—doss (U—)osz (I—)oss U—)096 Ci-—dole Ci—doery -—doco (i-—)ols (—)096 (I—)O9'E (i—)ovE (i—)009 (—)0eL (—)096
OLS (—)087 M—)ole CU—)009 (i—dose (U-)006 (—oez —)00°€
(1 (I—)oz's CU—o7L (—)006 (U—)o7z (i-dosz (—)osr (i—)oo. —)06'8(1
O8s (M—)O8'I CU—)OLZ (—)06r (d—)ovro (Q—dol's (I—)os'l (i—)ooz (—)O6'E (i—)ogeo9 —)07’8
(1 U—)or'l (i—)O8l (—)ove (i—)oo9 —)O1'8(1
06S (1—)07'I (1—)O'I (U—o6E U—)oss (—)ooL —)00'I
(1 (1—)Or'l (1—)oo'e (I—)OE'S (—)06'sS (7—)0S6 (—)o7TI U—)ol'z? (i—)oo's (i—)06s
009 (7—)09°8 (I—)OE'l U—)067 U—)oL'r U—)os's (7—)ooL (I—)00'! (i—oez (i—)ose (i—)o9s (%7—)0o79 (Z—-)oL'8 (W—)007 (I—)oerp (1—)os’s
O19 (Z@—OZ'S
= (Z@—)O8'8 07%)
— (I §=—s —)ov'e
(i )—S 009
§=606C §=(@—)ol'b (t—)oL9 (1—)0L'I (i-—dose (i—)oLy (%@—)ose (Z—)oss (—)orl (i—)oee (i—)ose
079 (7—)OT'E (T—)00'9 (1—)09'l U—)067 (i—)ore (%—-)orze? (T—)os'p (M—)OT'I C—)O9ZT CU—)OSE (7—)OOZT (%@—)ovE CU—)OOI (i—)orz (i—)ooe
0€9 (7—)O8'I (T—)oL'€ (—dOL'! CU—)007 (i—oez (Z—)Or'l (%—)or~% (T—)00'L (—dOL'I (i—)oe'z? (Z—)OI'l (7-061 (7—)00'9 (1—)09'I (@—oez
Ov9 (Z—)O0'l (%—)oezt (t—)os9 —)07'I
(1 (%—)Ov'l (c—)osL (Z—)OPr'l (T—)00'r (1—)00'1 (I—)Or'l (€—)009 (€-—)os6 (%—)ose (7-086 —)or'l
(1
0s9 (€—)07'9 (Z—)OPI (Z—OHE (%7—-)O7TL (Z—)os'8 (€—)00'r (€—-OL (—)Orze (—)079 (—)OL8 (€—-lOz7E (€—)0s'r (€—)061 (c—)os's (C—)osg
099 (€—)os'€ (Z—)OO'l (%—)O7TZ (%—)09r (%7—-)ozs (€—)orz (c—oe'r (Z—)OE'l (Z—)os'e (7—)oe's (€—)06'l (€—)osz (Z—)OOI (%—)ove (T—)ows
0L9 (E—OL'I
= CE—)OZ'S
= (Z—)OTI
= (@—)ow'@
366 (t—)097 (€—)OTZ'I (€—)OSZT (€—)OOL (7—)0O7Z (z7—-)o97 (—)os6 (€—)osI (¢—)o6'r (7—)O8'I (T—)09'7Z
089 (bh—)OB'L (€—-)097 (¢—)ol9 (7—)07'I (%—)Ov'l (p—)os's (€—)Or'l (€—)00'r (¢—)or6 (Z—)Or'l (h—)OL'> (—)Ors (€—)o97 (c—)006 (7—)or'l
069 (p—)OL'E (E—)OE'l (€—-)o6eE (€—-)009 (€—)0OOL W—)osz (—)oOLL (€—)OlI'Z (€—)Joos (€—lotTL (—)oez W—)oLy (€—)osl (€-)osr (c-—lowze
0OL (r—)007 (—)oos (€-)07Z (€—)ol'e (c—)ope (p—)or'l (p—)or'r (€—)O7I (€—)oLz~ (c—)oe (h—)OI'l (—)OLZ (h—)O78 (€—)OEZT (c—)ooe

oy], sio9quinu
ur sasayjuared
d}0uUap
ay} amod
Jo "U>) ‘snYyL (Z7—)Op'l
Speor OblX ‘2_Ol
Table 1(6.3) Selected Formulae for Calculating Lightness-scale Values V from Luminance Factors
Y of “Gray” Paint Chips
2
e Se e ee ee,
1. Priest, Gibson, and McNicholas (1920):

V=10yu<
Used in connection with the original Munsell system. Applies best to observations with a white
background (Munsell, Sloan, and Godlove, 1933; Godlove, 1933).
2. Munsell, Sloan, and Godlove (1933); Godlove (1933):

V = (1.474Y—0.00474Y7)'/7
A modified version of (1) above; applies best to observations with a middle-gray background of
luminance factor Y, =19.1.
3. Newhall, Nickerson, and Judd (1943):

LOY
1.2219V —0.23111V7 + 0.23951V? —0.021009V4 +0.0008404V>
Yugo

Used in connection with the Munsell renotation system. Applies best to observations with a
middle-gray background (Y, = 20). Defines Munsell value (_V =V) in the Munsell renotation system.
The luminance factor Y is relative to magnesium oxide taken as 97.5%; this gives a value of
Y=102.568 for V=10. The Munsell value function is tabulated in Table II(6.5.1) (see also,
Bridgeman, 1963; Billmeyer, 1963).
4. CIE (1976):

Paha
V= 25( Y, 16

CIE (1976) lightness function L*(=V). The luminance factor Y> refers to the nominally white
object-color stimulus; usually Y) = 100. (See also Wyszecki, 1963, Ladd and Pinney, 1955).
5. Foss, Nickerson, and Granville (1944):

V— 0-251 Slog..¥.
Defines the gray scale of color Harmony Manual and is based on Weber’s law. V = 0 corresponds to
a good pigment black (Y = 0.009); V =10 corresponds to a good pigment white (Y= 0.891). Applies
best to observations with a gray background whose luminance factor is close to that of the gray chips
being compared for their lightness difference.
6. Richter (1953):

V = 6.1723 log,)(40.7Y
+1)
Used in connection with the gray scale of the DIN Color Chart and based on a modified Delboeuff
formula. Applies best to observations with a gray background of luminance factor Y, ~ 50%.

823
Table I(6.4) Coefficients of Projective Transformation Yielding Approximately Uniform Chromaticity
Spacing.“

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Judd _Breckenridge— Farnsworth Hunter MacAdam Neugebauer OSA Sugiyama-— CiE
Coefficient (1935) Schaub (1939) (1944) (1941) (1937) (1949) (1963) Fukuda(1959) (1976)

ayy 2.7760 —0.74803 0.6600 2.4266 4.0 4.4210 0.4661 0.230 4.0
ay ZAP, . 7 1.58203 0.0000 — E3634 0.0 — 2.5263 0.1593 0.000 0.0
a}3 —0.1192 0.70001 0.0000 — 0.3214 0.0 — 0.6316 0.0000 0.000 0.0
a, — 2.9446 3.19700 0.0000 0.5710 0.0 0.6316 0.0000 0.000 0.0
ay 5.0322 —1.55045 1.2666 1.2447 6.0 2.2105 0.6581 0.600 9.0
a3 0.8283 —0.54884 0.0000 — 0.5708 0.0 0.9473 0.0000 0.000 0.0
a3, — 1.0000 1.00000 0.0000 1.0000 = 2a) —2.0000 —0.1574 —0.038 Al
a3) 6.3553. = 1.05336 1.0000 2.2633 12.0 12.0000 1.0000 1.000 12.0
a33 1.5405 —1.64023 0.2666 1.1054 3.0 3.0000 0.2424 0.163 3.0

Notes referring to the UCS systems (1) to (9).

(1) Coordinate system is represented by an equilateral triangle.
(2) Developed directly from Judd’s system. Coefficients so chosen that the diagram can be plotted in rectangular
coordinates. This UCS Diagram is usually referred to as the RUCS diagram (Rectangular-Uniform-Chromaticity-Scale
diagram) and places the chromaticity of the equal-energy source at the origin of the coordinate system. The
transformation coefficients given here differ from those onginally given by Breckenridge and Schaub (1939) but lead to
an identical diagram. The original transformation is carried out in two steps, a projective transformation followed up by
a simple translation of the origin. These two steps have been combined into a single step.
Developed for minimum deviation from circularity of loci of Munsell hues of chroma/10 and value 5/, with CIE
standard illuminant C as the center.
Developed to adjust the spacing of Judd’s diagram to conform more closely to the Munsell spacing. The chromaticity of
CIE standard illuminant C is represented by the origin of this system. Sometimes referred to as the “alpha—beta”
diagram.
and (7) Developed directly from Judd’s system. Coefficients so chosen that the diagram can be plotted in rectangular
coordinates. System (5) is deliberately designed to have simple coefficients. This system was adopted by the CIE in 1960
as a provisional standard UCS diagram and is called the CIE 1960 UCS diagram.
Identical to MacAdam’s diagram (5) but with the origin shifted into the point representing the chromaticity of the
equal-energy source.
Developed to conform to both the chromaticity spacing provided by the Munsell system and that provided by
MacAdam’s discrimination ellipses (MacAdam, 1942).
Modification of MacAdam’s diagram (5), which is also the CIE 1960 UCS diagram, by making the new coefficient a,,
equal to 1.5 times its original value.

“The transformation has the following form:

x (a,x +a,y+ 413) /(43,x + a3) y + 433)

= (a9,X + ayy + 43) /(a3,x + a3yy + 33)

(x, y = CIE chromaticity coordinates)

824
Table I(6.5.1) Three-Dimensional Color Spaces and Color-Difference Formulae
lS ee ee ee ee ee ee eee
1. Adams (1942)

(a) Chromatic-valence space:

i eK:
We= VFI
W, = 0.5V,
KES MDT |
eae

V,, = Munsell function defined by:

Y, =1.2219V, —0.23111V* +0.23951V,> —0.021009V,* + 0.0008404V,>

X= X/Xvgo = X/0.9804
Z.= Z/Zyy0 = Z/1.1810
(For CIE standard source C, X.=Y.= Z.=100.00). The coordinates W. and W, define, for every
W, =const., a UCS diagram. In this case, the abscissa (W,) is proportional to (X,—Y,), and the
ordinate (W,) is proportional to (Z,—Y,). The achromatic color lies in the origin of the diagram.
The total difference between two colors given in the chromatic-valence space is defined as:

aE=|(aw,) +(aw,y + (any)

(b) Chromatic-value diagram:

Wt=V.—-V,
We = VaoF:

We OD ACYy,)
V,=Munsell value function as defined above.
V.=Munsell value found from Munsell value function by substituting
X,.for Y, and V, for ¥,..
V,=Munsell value found from Munsell value function by substituting
Z..for Y, and V, for V,.
Table II(6.5.1) shows the interdependences of X and V,, Y and V,, Z and V,, respectively. The
tristimulus values X,Y, Z refer to the CIE 1931 standard observer and the CIE standard source C,
and are computed in accordance with the procedure outlined in Section 3.3.8.
The coordinates W* and W- define, for every W.*=const, a USC diagram. The total difference
between two colors given in the chromatic-value diagram is defined as:

)]}
az={(0.23av,) +[A(V,-¥,)] +[o.4a(v.-v,
The successful introduction of this formula to industrial applications is to a large extent due to the
work of Nickerson (1947, 1950), and thus the formula is often referred to as the Adams-Nickerson
color-difference formula.

825
Table 1(6.5.1)— Continued

2 Saunderson and Milner (1946)

“Zeta” space:
w= (y= V, )(9.37+0.79 cos 8)

= kV,

= (V,—V,)(3.33+0.87
sin8)
V..,V,,V,
Sans = Munsell values as defined in the chromatic-value diagram above and given in Table
11(6.5.1) as functions X, Y, Z, respectively.

0.4(V. —V,)
6 = angle calculated from tan @
V,—V,
=———_——

k =constant whose value may range from | to 5, depending on the observing conditions such as
fineness of the dividing line between the light patches being compared.
The coordinates §, and §, define, for every {, = const, a UCS diagram in zeta space.
The total difference between two colors given in zeta space is defined as

AE=[(As,)?
+(AE) +(4&)7] 7
Glasser, McKinney, Reilley, and Schnelle (1958)
Cube-root color coordinate system:

L* =25.29G'/? — 18.38
a* = K,(R'73—G'/?)

b* = K,(G'/3— B'/3)

with R=1.02X, G=Y, B=0.847Z

K,,= 106.0, K, = 42.34

For an improved approximation to the Munsell system, the constants are to be taken as follows:

K,=105.0 for R<G; K,=30.5 forB<G

KK, = 125,0" ior Re, K,=53.6 for B>G

The coordinates a*, b* define, for every L* = const, a UCS diagram in the cube-root color system.
The total difference between two colors given in the cube-root color coordinate system is defined
as:
AE=|(AL*)’ +(Aa*)’ +(Ab*)']'”°
~ Hunter (1942)
Modified Judd System (Judd, 1939):
Ue= 221 0

Us = 221Y'/4B

Uhre
826
Table 1(6.5.1)— Continued

with
ga 2.426% ~1.3631y ~0.3214
1.0000x + 2.2633 y+ 1.1054

_ 0.5710x +1.2447y —0.5708

P= +
0000x +2.2633y + 1.1054
x,y, Y=CIE 1931 chromaticity coordinates and luminance factor, respectively. (Y taken in
percent).
k=constant expressing the relative importance of the lightness (Uy) and chromatic compo-
nents (U,,Ug,) in the joint system; k =12 for comparisons made across a very narrow
dividing line. Increasing separation between the colored patches to be compared decreases
the value of k.

The total difference between two colors given in the modified Judd system is defined as:
2) V2
AE={,\ |221¥,1/4((da)? +(ag))'| +[kar?p |
where

Aa=a;— a, AB=B,—b,
when the first color is given by a,,8,, Y, and the second color by aj, B,, Y;.
__ constant factor which takes account of the masking influence of a
I, = glossy surface on the detection of color differences:

= ties

I= Yee. 5
The unit in which values of AE are obtained from the above formula is defined as the NBS unit of
color difference.

5. Scofield (1943)
L102

a=7La

be
a, B as defined in the modified Judd system; Y= luminance factor.
The total difference between two colors given in the Scofield system is defined as:

AE=|(AL)+(da) +(asy|
6. Hunter (1948, 1958)
Hunter L, a,, b;, system:
eG yas
Wie Gl Uh Coos
iL BS Gpeee™
yes
7.0(Y —0.847Z)
b, = —____—_—
yl/2

827
Table 1(6.5.1)— Continued

X,Y, Z= CIE 1931 tristimulus values (taken in percent). The Hunter photoelectric color-difference
meter measures the quantities L, a,,b, directly.
The total difference between two colors given in the Hunter L,a,,b, system is defined as:

AE=[(AL) +(Aa,) +(4b,)']

Wyszecki (1963)
CIE 1964 (U*V*W*)-system, (CIE, 1964):

dteorse Ei al GPa ag

. = 13W*(o as Uo)

Wo 2hy 7 1k te
where
u=4X/(X+15Y+3Z)

v=6Y/(X+15Y+3Z)

or u=4x/(—2x+12y
+3)

v=6y/(—2x+12y +3)

Ug, Up = Values of the variables u,v, respectively, for the achromatic color placed at the origin of
the (U*,V*)-system. For object colors the choice of up,v_9 to correspond to the illuminant is
generally satisfactory.
The total difference between color (U*,V*,W;*) and color (Ux, V3‘, W3*) is defined as:

AE =[(Us—US)’ +(Vt—VE) + (We WEY|

The CIE 1964 (U*V*W*) system is a development from the CIE 1960 UCS diagram.
In 1976 it has been superceded by an amended version known as the CIE 1976 (L*u*v*)-system
defined below (see also Section 3.3.9).

CIE (1976).
CIE 1976 (L* u*v*)-system:

L*=116(Y/Y,)'”°-16 with Y/Y>0.01

u*=13L*(u'—u')

p= Dia)

- where
u’=4X/(X+15Y+3Z)
v’=9Y/(X+15Y+3Z)
and
Ya AX CA Ia oe)

vo, = 9Y, /(X, + 15Y, +3Z,)

828
Table 1(6.5.1)— Continued
ee
rp Ee
with X,,, Y,, Z,, being the tristimulus values of the nominally white object-color stimulus. Usually,
the white object-color stimulus is given by the spectral radiant power distribution of one of the CIE
standard illuminants, typically Dg; or A, together with the perfect reflecting diffuser as the
reference surface viewed by the observer. Under these conditions, X,,Y,, Z, are the tristimulus
values of the standard illuminant with Y, =100.
The total difference between color (L*, u*, v*) and color (L3, uz, v3) is defined as :

AE*,=[(AL*)’ +(Au*)’ +(Av*)’] 7

The CIE 1976 (L*u*v*)-system is a combination of the CIE lightness index L* and the CIE 1976
(u’,v’)-UCS diagram. It is an improved version of the CIE 1964 (U*V*W*)-system, which it
supercedes (see No. 7 above). For more details, see Section (3.3.9).

CIE (1976).
CIE 1976 (L*a*b*)-system

= Y/yajr > = 16

a* = 500[(X/X;)'? -(¥7Y,)'”)]
b* = 200|(¥/Y,)'”° -(Z/z,)'”"]
with

LOR TTY ASAIN

The tristimulus values X.,, Y,,, Z,, of the nominally white object-color stimulus are defined as in No.
8, above.
The total difference between color (Li, af, b7) and color (L¥, a¥, b¥) is defined as:

AE*,=|(AL*) +(Aa*) +(Ab*))]'”

For more details, see Section (3.3.9).

MacAdam (1942, 1943), Brown—MacAdam (1949).

2 2 2 AY
(As) = g,,(Ax) +2g,,AxAy+ g,(Ay) T 833 >

Y=(Y,+Y,)/2, where (x,, y,;, ¥;) and (x, y2, Y,) are the 1931 CIE chromaticity coordinates and
luminances of neighboring colors | and 2.
211, B12» 27 — constants whose values depend on the chromaticity (x,, y,) of color 1 and are read
off special graphs.
33—constant whose value is approximately unity for samples viewed at an angular
subtense of about two degrees and separated by a sharp and narrow dividing line. For
a less well-defined line, g,,; may be considerably less (for example, for a 5° separation
between large samples, g3; is approximately 0.005). For extremely small samples,
contrasted in color with their background, g,;,;~0.5 and g,),22)5, 85. are about
one % of their nominal values (Am. Inst. Phys. Handbook, 2nd ed., pp. 6-149, 1963).

829
Table 1(6.5.1)— Continued

The unit of color difference (As =1) defined by the above formula is called the MacAdam unit of
color difference (See also Sections 5.4.1 and 5.4.2). A formula has been developed by Chickering
(1967) which allows the calculation of As without the use of graphs for the values of the g;,
coefficients.

K. Richter (1978/79, 1980)

LABNAU 1977 color space

L* =116(Y/100)'’° —16
A* = 500(A’—
A’) ¥'79

B* = 500( B’— B,)Y'”°

where

1/3

a= alF*5]

This space is intended to be an improvement over the CIE 1976 (L*a*b*)-space (see No. 9 above).
It includes a chromaticity diagram (A’, B’). The capital letters LABNHU describing the color space
denote the following perceptual attributes: L = L* “photometric” or p-lightness; 4 = A* =“red-
green” or RG-chroma; B = B* =“yellow-blue” or J/B-chroma; H = H* “colorimetric” or c-light-
ness; N = N* = blackness (N = noir); U =U* = unique hueness.

830
 £160]
laa) \O
al yy
ras) ™~
o

620
OMmONANNMOKDMWEr-ONTNA-
OAAWE™OMN
30 2
EBS EEN Be TE ERE a DUES eB BINH BIRS RM

S oO oO L0 9°0

(é

ie

iS
vel

0
0

vL'l

G
6L'I
8

g
66'1
8

¢
9

v9"
8
b

6S'1
8
L
€

8
L

€
I

I
(SLOI ‘PY99zZSAAA puv ppne Woy) X anfeA SKINMWIYSIT, ALD Jo uopoung v se “4 anjea [esunf,

Ors's
060°9

Ores
cOE'S

psls
696 '°¢

psps
976'S

9TL'S

600°S

C06'P

96L'Y
COL'Y
g9T'S
8C7'S
800°9

SO8'S

809°S

c6b's

Le6ov

L98'P
811s

SP0'S
BLES
€L1°9
€8£°9

€69'P
tL6P
180°S
Ive9

le8'p
0
0
v6TC

0
plc
6L°¢

9
c
66°C

E
9

v9°?
8
L
8

Si
8

6S °C

¢
iS

c
t

8
€

8
I

I
I670°9

Ly9's
I

iS
91P's
Si
S
£1e1°9

LS88°s Lv6 01
9SP8's
L6679

SiS
a)
SH

C989°S
(é
29ESCO

)
98CL'P
69°C
161°S 8L8'6
69°]
68'1
68°C
L96'S ILO

P81
p8'~
99L'S
OIs'Tl

poll

7S ‘Ol

pre Ol

0$9°6

OLE6

0976

OS1°6
9Pb II

600711

97801

LGG.O|

Lts’6
T8P'6
9Cth'6
tr9 Ol

89101

£666

Slt'6

S076

£606
1OL'T1

18°6
v9L'6
St6'6

LOL’6

t6S'6
Ol
€8°

Ol
p9r

60101
1S0'O1
|
RSH

Ol
S9L

ol
vor

Ol
S8c
TI
ese
ll
Ice
Il
8Sc
Sé6l
ll

98801
Leo
ll
plsil

SBS

Ore
NA—ORAOWr-UONnRNATMNA-—OADAAOErONnNTFTMNA-OAOrONnNTMNA—DAGOrON

St ge ee Eee

€ C I 0 8 (iB 9 iS v € (é I 0 6 8 je 9 Si blyt G l 0 6 0 L 9 iS v £ (é l 0 6 8 iE o iS
oP
LY

68°P oP 9°S SP
8's
6S

OMmUONAMNNANHOADAOMWrONHNTAMA-OCAMWFON
TMA -—-CAMr-UONnN TANNA -—-CAOr-ON
ZA aS ORS
66'S 8S EiS
yy)

OMmUONTAMAH—OAMWFUONAMA—CAPMWrFONFMNNA—-DAMOWFrUONTMNNAN-CAMOCrON
89

69 89 29
dys
SL

NAN—OKDOSKrUONTMNNAN-—ODAWSOFrUONHNTANMNA-—-OCAMOrOoHnTETnnN-CAAOr-UON
tA BL LL fe Wye
9°8

9°8

R00 BASES A FTA A 00 Be NOE SF Nyt ER COIND WARE ONT SG OGESAS AG

ww
x co
co 0 L’8 L’8 co
(WT'S’9)IT P48.

95001b0'001£S°66 0S 66'L6 6b'L6 87°96 86'S6 VL’ LC cO'€6

1S°€6 18°16 ££ $8°06 Lt €1°06 99°68
$0°C6
Xx 86 P6 26 16 06
6L'66 L766 9L°86 S786 bL
10°66 L6 VC €L°96 €7'96 €L°S6
L6 86°96 76 76
$6 €C'S6
87 86 60 St 9L't6
v6 00°76 8L°C6
€S°C6
C6
6C LS'16 60°16 19°06 68°68 cr
68
001
Of

NOMmONFMNAMOKDOEr-ONTMA
HK OCOAADAMr™-OMN
L6

9°6 9°6

831
 c0
OmONnTMNAHOAMrFONHNAMNA—DCAMrUONTMA—O

v0 So 00

0°]
cl
vl

vil cal Gal 01

8 iE 9 iS € c I 0
vc (6x6 Ge OC
pst

ve

WrONPTMARODWMONTMA=OS
foe)
vv Ss a

16:07 9'0C
SS'0C 9°02
L702
81°07 00°07
60°07 16°61 9S'6l
$961 Lv’
6l
C807
€L°0¢ SPr'07 C861
61
vl 8t'6l

v9
—(V1'S'9)IIpanunuoy
L A148

v6

832
 1€60
=
~~
Fe
i=)

|
6L°0

9°0
Dr-ONTMNAHRODOMOMONAMNAROAWMEON

L'0 9°0

RA CY se ONBOTS NOU RE EY tS G0 TS1D sn AO ch 08 tS

ont

Pt
ws Ea

MAABODOMONATMNABORDOOONEMNABODHOMON

97
ro LC IG

1L9°6
OvL TI
Se6o Tl

OSE TT

vor TT

08601
£8 Il

6cC TT

C19 01

vel ol

v19'6
TI9 TT

cST OL

SLO 01
LIO 0!
£61 Ol

€v8'6

Srb'6
68£°6

LL76
€£t6
16 01
Ter Ol

ERE Or
a
106°6 $8L°6
6566 8CL'6 LSS'6
°6
10S
OL8
IT
IT
$08

SLOT

IT
Lvs

IT
6Iv

TT
C67

EST
CcvO
TT

81601
01
9S8
01
c6L
Ol
bel
€L9°01

ol
Iss

Ol
IL€
€ C I 0 6 8 L 9 SI v € (6 I 0 6L€8 L 9 S € G |! 0 6 8 if 9 S v € C I 0 6 8 L 3) g

9°€

ce
pat
¢vl
66°F

oP
LY

DHm-ONTFTMNNANRMROKDOMON
AMA HK OKHOHr-ONnNTMNARODSERHONHTMNAROAMOr-ON
4A
87

6v oP cP
8s

OMmOMNTFTMNNAHKMODWMWErONTMNARKODOMOErFONAMNAHKODTME-OHMHTATMNARKOAOMron
44 66'S 6s Ls 9°¢
Ls 9°¢
89

Om~ONAMNNAHOARAOFONAMNAHKDAMCFONnNAMNARKODAMWErONHTAMNARKODAOrON
44
Pe)
LID

669 i) 959
BL

/ejE

Ee eer et ee YO a eNOS ee Be CNT, eae hae a eC Dee ae ee ee ar

FA ay =
Lae SE OEE ~

o Dio TON Fim Ci OD OOM Ei Ce A 0010

44 668 68 88 L’8 06 9°8
\o
ioe)
va)
co

com OM NA BRONDOMm~ONHNTANA
SO AoronwtrnnnOnworwon
68°6

9°6

v8'6 L’6 L6 9°6

wos)
(4OJDeJ
Jo
AID
sNNUIYSLLY,
anjeA
soULUTUM])
{ppne¢
uoyouny
[PesuN,
anjea
“PY99ZSKAA
44
(AT'S'9)IT
(SL6I
pu
Ade
vse

833
 c0
comruontrann Oo

v0 60°0 00

8 & 9 S £ Cc I 0
v0"

8 fh 9 S £ C I 0 6 8 L 9 iS v £ G I 0 6 8 L 2 S £ Cc I 0 6 8 L a S v y Cc I 0 6 8 L 9 S v € c I 0

07
vr? (Syiré ME CC ve? 0°

8 L 5 ¢ v £ C l 0 6t'€8 £ 9 iS £ c I 0 6 8 (3 9 ¢ Poey c I 0 6 8 L 9 § v Me (4 I 0 6 8 L 9 iS 4 £ Cc I 0

Ove
v

ised (68 Oe

96°07
98°07

6S 0¢

0°07

9861
6b 07

LL’6l
S661
LL‘0@
89°07

Ov'0C
1€°0c
OG
co
£107
NAH ODADMONTMNA- OCOAMOMrFONnNTMNA— ODACOCrUONHNTMNA=
_
8 jE 9 ¢ vO'S€ c I 0
loa) AQ st
vrs £'S yD cs va) a) va)

MNF ODOr-ONTMAN—CAorwonwtrnn-o
(O88)

v9 Gp) vco9 0°9

—(A1'S'9)IIpanuyuo)

vS'L

vO'L
PIL

0 vel
le

(é

jh
0
4

9
L

C
5

2
L

£
I

9
8

0
¢

€
I
(6
vr'L

I
S

6E°L
8
£

5
e

I
66°89
6L°89
65°89

LS'p9

68°C9
81°69

91°C9
80°€9

86°19

66°6¢
10°89
18°L9

O1'6S
619

C89

S719

L109

£9°6S
Sv'6S
87'6S
10°79

S09

18°6¢
[O19

0v'89 ces9
19 tv

‘09 cS
vO'L9

99°99
96°99
HE 99
£p'L9
tc Lo

$8°99

80°99
68°99

v6'v9
9L'¥9

|
vS'8|

v1'8

¥0'8

0 ve
Us

Cc
0
L
C

C
€

9
L

C
&

9
L

0
g
I

9
I
Cc

Cc
vv8

0
S

6£°8

I
£

8
L
I

8
0£°68
6¥'06

£S°68

CV'P8
£e°S8

0S°18
$9°V8
€L'06

88°78
$706

67°€8

6£°C8
L606

Or'08
€L'18
LIC8

8718

99°8L
81°08

£S°6L
SL’6L
L66L

OL6L
88°8Z
1¢°6L

10°06 $798
£8 cS
78°88
6S °88

C188

9S°S8
Se°88

88°L8
S9°L8

$698

6L°S8
81°48
1vL8
A981

vl'6
6L°19
S'6

088°09
S618

v0'6
8°08

0 ve'6€0°98
Cc

990°68
2COLO 92 8h°98
S¢Ig*s9 Ea€1°s9
tf

CL'98
OL'S9 g61°78
£9°€9
S
0

€8
€PL
£9GE9 90°18
LO'19
I
8
L

LL vEe'c9
0LO'€8
C9 c9°C8
8
9
8

9
3

iS
L

C
iS

0OL'09
£
I

)
O1'S8

L
8

0
I

e
8°79

c9'08
vbv'6

0
c
S

6£°6

ls

iS
€

I
I

6C°6
6C°L
6C°8

60°6
60°8
60°L

Aa
4A
A44 LL’68
6h
6r'8
0c'89
6r'6
L
||

L6¢8
v7'8
SPe9
VOL
vC'6 8°78
618
C9
cS
61L
61°6

834
 835
S So's S 95601
OCOAAGrwon
CAworwomn

COUet UN. OO) AOI

CAPCrONTMNNAN—-CAGrwon

OAOTrON
ILLOl
é20' TI

MARODOMONATMARSCRHMONMNIMA-CHRHMMmMON
9 $69°S 9

ANRODOHMONATMNAAMOHRHMrON
—-CAMrUONTMARK
10°80]
L 9EL'S fi 680°11 801
67
8 8LL'S 8 SSIIT
g's

801 8S
6ST LeeLt
cP

6S'I O78°S

SE
Sed

va)
nN
ct L8°801
0 £98'S | 0 L87 Il $1601
! $06'S I PSE ll

OCrUonTFTnne
60! tv
iCLv6'S
cer
Sr-onytTnn

68h II
cL’601
£ 066'S €
Diy.

eye
100
vac || OSS Ut
SE

9°8
91 VeOiO™)

ros
9°0 Dig DES eS) OT
OL
S8L0°9
LI
gp79
RAED CT OO ENNELY

OIL
6S
9 [elo a9 £69 II 68°01
LTESMS)
£CoE BIL
86079
TT
OL itt Ly
69'1 tSc9 | 697 || 668 LI

99
9°6

=
SS 9'P EH
[I0696
L6c9
| 90°T11
IChE'9
8t0'TI
OSr-ONTMNNANK

SeCll
CLSG9
801'CI
(6
>|

TMA
SOIT
gSLI
€ttv9

PEM EA
v6ClI
Cl
8re
VLC
8Lv'9
PL

LY
fits L9 peel
g6le
Sv7s'9
CI

N-GOLE AD OM

MM OADAMOSrFUONHN
Seal
9 OLS'9 9 06€ ZI

BO
ell es
L S19°9 fl 19¢Cl

OO
Plt
CI
8799°9
Tl
ees

lee)
vit
cy
6L'I 80L'9 6L°T |} S09 TI

Ls
rE
+
6L'¢

OI
cL
0SSE9—
TI
LLY
Vit
|

RKO NAAR CP Me,

co's
tt
I £08°9 I aa | cestt
c0S8°9
18°71
(4

AWOmONFANNAN
AO
II co's
€ 868°9 t v68°Cl cost

8°8
vs" 9669 v8°T || 896°C

89
se

eo
a)
2
+
NS OE
OL co
S£66°9
IvO'el 9L
cS
9 1vO'L g vil el e8°91T
L 060°L jl 88 el Ch2uy
86eIL
El
coc

OE

EN(ONGE

68°8
.
Lil vy
68'1 88I°L 68°C || SEE El

gt

89
co
a)
=
+
Lil PL
O0Leck
El
CIV
|
8 L 9 S 60£ ré I 0

yO'stt
I L87°L I 98P El

Ee a

ROSE DO BE AP ON
sestl
C LEel C COS ET 99°811
€ 98E°L € [oe 24! L681I
v6't 9Ev'L vec || CIL el

“4
+
L76ll
S L8V°L S 68L El 6ll
8S
y LES'L 9 S98'el

NR
06611
fi 88S°L L Cr6 EL 17071
8 6£9'L 8 610'FI
660

66'S
Is‘0cI
66'1 069°L 667| ||
960°FT C8071
vI'Izl
(SLOT
‘PpezsKAy wosy)7ppn¢ pue ane
snfnUIysLy Jo GID
uoyouny (OI1'S'9)IT PesuN,anjea74 se v AGEL
 c0
v0
6c'0 0°0 0°0

1 c6TC 686'1
696 Oc6'l
6r6 cL8'1 ves
eS3'l ¢18'l O9L'I
cr SOL’
€cL' SEDs
Is9'l p9s'l
18¢'l Lys’ TIS’96r' COP
6LV'1 Srl
AS
I 1 | I I I
LLY'T
PSPC
~ce 600°C
L8€°C
p9E'T
CECIZE°C
76678LTC
9STT
SETT Vey 671°C
era Osl'c 601°C
680°C
690°C
670°C
620°C
600°C 1161
168°1 I8LL I 1
96L'I 889°]
6991 SRST!
I66S |6cS I6cr

om~runtrnn- oO

Ol
(oa!
(Gu!
vl

60°1

60°C

OT
CC
DAOWOMmONTMUANAOKHROMONAMNAHOHOWEFOnNTMNA
RK OHOMOr-ONnNTMNA SO om~ruounrnn-oo
SE Le _

VT VT N N (oxé ro N

POLL
GULL
OS6'L

9P8°L
668

OIL'8

69¢'8
S178
C918

950°8

868°L
061 0!

09L°6
00L°6
0796

09-6

L868

99L°8

tvs
8Lt°8

£00°8
97801

000!

7886

Che6

€C8°8

$s9°8
c9L Ol

sos ‘Ol

Le ol
$9001

6196

£876
869 Ol

8Lt Ol
cle Ol
€Sc Ol

€v6'6
££9°0I

178°6

1$0°6
16801

lpr Ol

8
brs
8878

vCE'8

601°8
088°8

6668
L916
6016
6LS'6

10¢°6

6
SCC
69601

ve

vy ev Cv (G54 Ov OV

ANK ONDAMDE™-ONTMNA—-OCOAHAWOWOEr-ONAFTMAHK—OHDOWFONnHNTAMNA—DACrUOHTMNNA—O
(EY)

(EY)

0-9

09

3 c I 0 6r'L8 L 9 S bol€ (6 I 0 6E°L8 L 9 ¢ € Cc I 0 8 L 9 iS AE € Cc 0 é I 0 60°L8 fe vO'L I

PSL
ZA

TE
panuyuoD —(DVS'9)II A981

vols
8r'18 82°08
10°18 ¢S'08 60°08
ce'08 98°6L Ov 81°6L
£9°6L CL’8L
$6°8L 6 LO vO'8L
C8°LL
6S ES OL'9L
c6'9L 09°SL
OCiOL 18°SL
8Pr9OL LEsl 6 IL OS
Sasi 90°PL ve'cl 8 OC
69°IL
SL 1612 v9'OL 10°02
IL'18

6L 8L 8L LL Tee PL vl PL ae IL IL
Zz

9 4 I 0 6v'88 L 9 iS brs€ (6 I 0 6£°88 le 9 iS £ C l 0 VL 0

8 Ls 9 SI vos€ rd I8C 8CONGL
is
L 9OCISE
Iyel SI
iS
86°CL €3SS°CL ¢ v0°8€cy
Cc 1 0 60°88 Ih 990°1L
iS
OL
$8 OL(4 C I 008°69
OL
IC
ps8
ZA

Ralcke
8pel
ve 6c
678
L£0°9L 61°L
618
el
ps I's
9LCL
VIL
£6°€6
[2
prLOlL8°901
S1'LOI bL'SOl
9b°SOI06-01
Z9°P01
81°SOl re LO'P0IZS€01
p7'€01
96'701
69°701
Zp'z0198°101
65101
Ze'101
S0'101
8/001+7001
1$001 L1°66 9°86
L6'66 £b'66 1€°L6 66'S6 69'b6 61°76
P01
901Z0'901
65°90I
0€ 6L'€01 b1'Z0l 0L'66 06°86 LE'8611°86
p8'L6
85°L6 S0'L6
6L'96
£5°96
9796 £L'S6
Lv'S6
17'S6
566 br'r6 L9'€6 Zb'£6
91°€6
06°76
60°6

© C I 0 8 L a ¢
vS'6

onr~unrnn-o

66h vr'6 C6 0°6

836
Table 1(6.5.3) Whiteness Formulae (Whiteness Measure Denoted by W)

W=Y
Y= luminance factor of white surface relative to a given standard white and with respect to the
CIE standard observer [see, for example, ASTM Method E97-53T (1953); Federal Specification
TT-P-141b, Method 612.1, Method 412.1, Method 613.1].

W=B
B=“blue reflectance” = reflectance of white surface when illuminated with blue light of specified
spectral radiant power distribution. In some cases, the spectral distribution is that of CIE standard
illuminant A, weighted by the CIE 1931 z(A)-function [see, for example, TAPPI Method T425m-48
(1948); ASTM Method D985-50(1950)]. In this case, B is often called “paper brightness.”

w= R99 20 —R 450
R709
R409, R459 = reflectances of white surface at wavelengths 700 and 450 nm, respectively. This W is
usually referred to as an index of “yellowness” [see, for example, Hunter (1958)].

A—B
SCE
B,G, A= reflectances of white surface when illuminated with lights of specified spectral radiant
power distributions. Usually, the spectral distributions are those of CIE standard illuminant C
weighted respectively by the z(A), y(A) and red portion of the x(A) color-matching functions of the
CIE 1931 standard observer. This W is usually referred to as an index of “yellowness” (see, for
example, Federal Specification TT-P-141b, Method 613.1)

W = 2 R439 — Rezo
R 430, R670 = reflectances of white surface at wavelengths 430 and 670 nm, respectively. This
formula is called the Stephansen formula |[see, for example, Berger (1959)].

R 439, R679 aS in Formula No. 5. This formula is called the Harrison formula |see, for example,
Berger (1959)].

W=G-4G— B)=4B—3G
B,G=reflectances as defined in Formula No. 4. This formula is called the Taube formula [see
Hunter (1960)]. The quantity (G— B) (in percent) denotes an index of yellowness, whereas G (in
percent) is the luminance factor representing an index of lightness.

W = 0.333Y+ 1.060Z — 1.277X

X,Y, Z= CIE 1931 tristimulus values (Yy499 = 100) with respect to CIE standard illuminant C. This
is the Berger formula (Berger, 1959).

W=L-—3b
L, b= Hunter coordinates [see Table 1(6.5.1)] (Hunter, 1960).

W=(¥— kp?)'/?
Y = luminance factor of white surface
p. = colorimetric purity
k = constant
This formula is called the MacAdam formula (MacAdam, 1934; Judd, 1935).

837
Table 1(6.5.3)— Continued

11. W=(Y— KS*)'/?

Y = luminance factor of white surface;
S=distance in Judd’s UCS triangle [See Table 1(6.5.1)] between the point representing the
standard white and the point representing the white test surface;
K =constant, which for application to the grading of laundered white goods is 6700. This formula
is called the Judd formula (Judd, 1935, 1936).

“a7} 1/2
2: W=1 |[30(a? Hamat
Y= luminance factor (Yy_0 = 1.0)
a, B=coordinates of Hunter UCS diagram [see Table I(6.5.1)]. This formula is called the Hunter-—
Judd formula (Judd, 1941).
Variations of this formula have been proposed by Selling and Friele (1950) and
Hunter (1958; see Formula No. 11, as an example).

13:
,
AN ontaes 8 = ae
K,(G +0.242B) 200
G, B=reflectances as defined in Formula No. 4.
K,=constant which varies with the observing conditions. Typical values are K, = 20, or 70, or
(preferred) 38.6 (Hunter, 1960).

W=100—[(L, — L)* +(a* + b*)]'7*

L,a,b= Hunter coordinates [see Table I(6.5.1)]
Lp= lightness index of the preferred white of whatever material is being studied. If magnesium
oxide is preferred, L, is set equal to 100 (Hunter, 1958).

W =100—[(100— L)? + 10b7]'/2

L, b= Hunter coordinates [see Table I(6.5.1)] (Hunter, 1960).

Ww =100— OAT ells + k[As]?]!72

A( yi} == Lae Yee, sample
Y= jumdinanice factor
As =[(Au)? +(Av)?]'/? = distance on MacAdam’s UCS diagram [also the CIE 1960 UCS dia-
gram, see Table 1(6.5.1)] between the point representing the standard white (MgQ) and the point
representing the white test surface;
k = const. |
This formula is called the first Selling formula (Selling and Friele, 1950). A simplified formula is
given below as No. 17.

WE W=100-[(AY)* + k’(As)*]!7?
AY Yast, \
test gamples © — JUIDINANCe tacior,
As =as defined in Formula No. 16;
= const; a typical value is k’= 9.5 10°.
This is the second Selling formula (Selling and Friele, 1950; Friele, 1959).

W=Y-13.2Y[(Au)* + Av)?]!/2
Y = luminance factor of white test surface;
Au=u- uo,
Av =v — 0;
- u,v =coordinates of white test surface in MacAdam UCS diagram [also the CIE 1960 UCS
diagram, see Table I(6.5.1)];

838
Table 1(6.5.3)— Continued

Uy, Vy = coordinates of standard white in the same UCS diagram; for magnesium oxide surface
under CIE standard illuminant C, uy = 0.2009, vy = 0.3073 (CIE 1931 standard observer). This is
the Croes formula (Croes, 1959).

ve|av-(a (8) 2 2]1/2

L,M,S= length of long, medium, short axis of color discrimination ellipsoid centered on the
standard white.
A,b, c= constants
This is the Friele formula (Friele, 1959).

20. W=Y + a(x—x9)t+ b(y— yo)

Y = luminance factor of white test surface;
x, y= CIE chromaticity coordinates of white test surface under CIE standard illuminant D,; and
with respect to either the CIE 1931 or the CIE 1964 standard observer;
Xo; Yo = CIE chromaticity coordinates of perfect reflecting diffuser under CIE standard illuminant
D,; and with respect to either the CIE 1931 or the CIE 1964 standard observer;
a, b=coefficients whose values depend on the “hue preference” that prevails among observers
judging whiteness.
With no hue preference, referred to as “neutral hue preference”:

a= —800, b= —1700

With “green hue preference”:

a= —1700, b= —900

With “red hue preference”:

a= +800, b= —3000

This is the Ganz formula (Ganz, 1979).

839
 OS9P'0 Sttro €c9v'09IPh'0 LOLE'O

pereO LOPE'O
60L7'0

vipro

VELEVUALOLE0
Ofre'O
89960
69070
BuUoOI,)

88IS'0 L99b'0 O0CS'0P80 1980

SPv7S'0
C6EE'O

€SLv'0
=L29¢'0
66170
Ol
YAD

O8Pr'O

OOTr'O
cS6E'0
OLLE'0
096¢€°0
cOIP0 O£0v'0E8re'0
I6rE'0

6LEt
8Etr0
0pIIP'O
66th 0CELE
ISLLt
PSre'O
EPPEe'O

cech
9L7S°0I8sp'0
OOLb'0PLos'0
006F'0
O8br'0 9€S¢'06rcr'0
OS0S'0OLSP'O
OFTr'O
[fesuNnyA, Puosa¢g cei JW SonyeA

OL9P'0
6LIS'0

cCOIv'O0
O8¢h'0
P66E'0
968¢°0
OF6r'0
OfSv'0

L?Sr'0
89tP'0
6L0S'0
€19r'0
Iv6t'0
LL8€°0
O£6£'0
€9Le'0

Oc8e'0 LOIPY'O
966¢'°0 SIPb'0

L6ee'0 LEPEO
C8cv'0

CS9E'O O98E'0
c6ry'0

EPSE Oe clLct0

O0@7r'0 H06h'0
y90b'0 96SP'°0
9L8£°0 ChcrO

I@ve'0 L8Pe'o
I8tv'0

Ictro SPris'o
h

90EV'0
89Sh'°0
9IES'0

OI8P'0
Sc0s'0

S6EE
0000

0
0
OS6E'0
OC?tr'O

pLIs0
OL6P'0
LIPs'0
L
YAS

Sé6IS'0
8ISPh'O

98S€°0
669€°0
verry
16es'0
SPOY-MOT[IA

0
6S9€'0

SceEO
O8Et'0
LLEEO
S8S¢'0
6L9€'0

89bS'0
8Lrbv'0
6l1es'0
6bry'0

vOLbr'O
ISIP'0

OCES'O
CIvh'O
89IP'0 ch8t OL7h'0
6€crv'06lIv0 OSLE'O0
006£'0
ccOv'0

Ofte.0
Oc8e'0
0SOv'0
8£6e°0

£990
8P6E'0

ESteO

0
YAOS

cS7S'0 cObr L6SS'06610

SILS'0 6c7h'0
cé6Sr'0
=176v'0

OEP 0 c99t'0 SL?v'0

E2t&0
ILSh'0

886£'0
6P8r'0
60SE°0
899€°0

0
0LEPS'O
86cr'0
LS9S'0
L9CV'O
v9SS'0 TILv0I0L£°0
8¢ch180¢'0
cL6e'0
LO0S'0 1600
0O£Se'0
OSLE
06ree'
Icve'o Ecrs'0
88170 PLye'O
€Lete'0
v9Se’O
OV8E'0
880$'0
SrIP'O

069€°0
OISe'O0
686£0
ycOr'08€6t'089L¢'0OLSt'O0
198¢°06L9t'06ere
[fosuNnyA, SUINY puw SonF

9ECE'O “VEEEO
OSsSt'0 L760
Tere'0 IP9E'O
eLcet0 OCEE'O

066£°0
Lpse'0 096€'0
6cre'0 L99E'O
Lpse'0 CcS8b'0
I9L¢°0 cSsPv'0

908¢'0
IZOv'0

Icee'0
0
c7ss'0
$69S°0L6cS ILevo
IL9P'0
100S°0 SILE0O
€SOPr'O

6L8S°0
8699°0
YASS

eSPe'o
168h°0
0
9P0V'0
%78S'0 86ce'0
COLe'O

88rS°0
Lp6t'0

O8IPF'0
009£°0
SIcs'0
L88t°0

908E°0
L9vE'O
eesr'o
80Lt'0
(Cr6l

A
998L°0 0160 90¢F'0 S00¢°0
‘ppne pues “UOSIDYIIN] *TTVUMANT) UINUITXB JAI JUSUISIg [e9H9100, [| 94} 0} O17
[fesunyA, JOJ SUOTVIOUNY

6eve'0 97SE'0
Zhre'0
6rE£'O OcLE'0 eLpe
o9ee'0
6$9t'0

eSce'0
eese0

h
0
O88e O16E'0
ZIZP'017Z79€°0 6009°0 COlr'0
06h 0

O9EE'O
80tr0

IL9EO

0
H8TEO
€€ce'O
8PeeO
O09E'O O6bb'0
68S£°0 LEZE'O
IOEEO
CcIPE
ELCE ITce'0

€IZe'0
IPLs'0
L69¢°0
89bS'0
Ovec
OSIS'0
L99¢°0
0619€0
C18r LIPVE'O
89ce'0
87cE
6bbr'0I8ee'0
OSSt'089Lt°0
6cLt'0
61SS°0
oO
veces
OOLE'O

0090
96S£°0
0
P86t
CShEO

Serec8ee
[jesun]Ay jo sjuayeamby

OcrE'O
S8te
Seee
v9SEO 6Lc£

Ieve'0
96E£

16ct'0
8cre

e8et
88er
8IIv

0 oO 0
0Bret
0€8ce
0Olcte
0

CSPE-0
0OSre eve09fee
OC8CE
OccE'O
0

0
Pe
6I
'0
0
O£8t 0
0

6S06 119€°0 09S¢°0 Ss9v'0

81ev'0
cI8€0

€9cE0
Isse0

0
0
0

‘0
IvesLLLY'0
0OLbY08880seer
09617 06¢8S¢9cS'0
196¢'0 000r'0
I8ee'0
spoy

C690
0

9SceccCE
88I£

cic

061£°0
pree'0
Orce

09Lb

0 O
879E
€rec

OLTEO
€9ct'0
08Pc£
0vCce
098IT£ O8EcE
0CSTE
Q9cEO
0prceeO6IE
0
0

OcEv
9S7£
0
977EO

0
8ele
6L1¢°0

OSct
0
14540)

Isce
0

Icce
0
pSce0
100¢°0
S6re'0
vEele

L6cS'0

L8I1v'0
O8rr'0
csss
(a> x Fe) I¢cé6l HID

1Z6€°0
0cOS

0
0

oO
0
oO

8r8r'0
c6sv0 0SO8e
0L90PcSSE0
90£€°0
0

‘0
6bcr
0

eve
0
O1Se
0

ese
6LI£ O91£°0SLIt €867c80£°0vrie'o
61 1£°0

OLIE0
OIcE'O
$99£°0
SPre

S887

0 0
stro

v8ct

0
87Lt
196¢°0

cE8e

0
8Iee

891£0 IL1€'0
0LLI£
691£°0 087670poe 8I1£
0 8Sle0
e910
99T£°0
SclPr0IL9€°0
6£0£
0
c80£
0

brie
oO
O9TE'O
OLIE
0
ILT€
0
09970

06LV'0
es8Iv

$90¢'0

99S¢°0
89670
66r£

c9cS

OcEr
1p0S
0

0
90€r'00

0
0

0006t
0

0O9PE
O9ECcE
O16S'°0

soo0¢e'0
0

‘0
0

0
(1'°9'9)I P1921

x
WOT] ETS

OI/8
|

|
DIA

OT/L

81/9
|
998L°0

Zz
v
9/6

8
9

é
v

8
9
v
Z
rl
ZI
Ol
|

840
 LL61°0 v8Ls'0 61L7°0 88£9°0
0L672°0 crl9'O
0 Ovss
0870 8E€0e'0 816S°0
1919°0

02/¢
00ES
‘0 08870 cOrE’O Le€9s'0
106S°0 ££6S'0 686¢°0

9I7E'0
Lvos'0 Ipes'0
0S67'0 8SIe'0 06SS'0 I€Zs'0 £S6£°0 cr9S'0 10cr'0 90S¢°0 OSrr'0

LLCt'O
L679'0
ZP9£'0

0 0
Oc8r'0 ZO0E'0 ILOS‘0 v6lt'O‘0 087s

Ieee
C8PS'0 606¢'0 0 cChS Ivly'0 Sees'0 8Ltv0 009¢'0

LL61°0

L¢€09°0
LS9£'0
eesr'o0 8S0£'0 Lyly'0 L7éce'0 L7@6bv'0 ‘0
SLIS 0 pest I9TS°0 v90r'0‘0 801s 9L7b'0

OLEE
'0 68h

ILLS°0
P99€0
CSCb'0 101€0 elvr'o
0 Ovet£9Sr'0 S6Lv'0 8SLe°00 O£8b 096¢°0 O@8h'0 Ivlv'0 8eer'O

I8¢s'0
099€'0
I1zs'0

0
096£°0 O£IE0 8L0v'0
0 8€ce O8Iv'0 S9Ev'0 0 Ov9t Otrb'0 8080 Ovrr'0 pS6t'0 8cIPr'0

O£9€0

0
099¢°0 0
O¢CE
SrIe0 Ovlt'0 908¢°0 SC6E'0
0 v6re
896£°0 vI9E'O 166£°0 vILe'O 0 Ose

SLSE‘0
OLLb'0

0
O9Ee'0 gSI€0O T6EL0 c6Ie 44580) 90S€°0 LEE O£St'0 S6£E°0 Orse'0 SPre'o

elI¢‘066Cb
vIse’o

66bE'0

68££°0L8etv6ce
86££°0
6L8£°0
$c0s'0 $66t°0

9089°0

66££°08ree'06CCE'O
eILr0 S9ve'0
8LZ£'0
8crr'09bSE'0
88670
868S'0 C7970 18870 8690
O01 £0
07950 0 vCL7 0 8L67
09790 eese'0

00710
0
0
c61£
O187c
69£S'°0 0 LSOt0 66S 89S¢'0

6¢£9'0

0
vs
0
697£°0
0 cLOS0 L687 6c1€°0 £0950 88St°0 °0
6086 O16£°0
67LS 691¢'0

190

6£09°0
0 0
Ieee0
vLLv'0 69670 9LIE'0 0 Secs O8S¢°0 9S8£°0
SLvS'0 cers L60v'0 cher
CI/p
ISsce'o 9S€S'0 eLsv'0
cLeYv'0 Ig0e'0 607£'0
OS8r'0 LSS¢'0

0S8es
‘0 IL0S0 LLLtOLOS'0 v66t'0‘0 8£0S
Ol
vocr'0 virro

60¢9°0108S°0
00cI'0
6SE£°0
Iviv'O S800 977E'0 SIvr'o 0 OOS C190 vl9oe'0 IS9P'0 6S8€°0 SS9P'0 6c0bv'0 €lcr'o

vels€v0s'0
81bs'0
OScs'0

0
Ovee
908¢'0
Sclt'0 €cce'O 066£'0 c1PE'O 6€St°0
Ivly'O L8Iv0 6L9€'0 80c¢r'0
0 608¢ 8P6E'0

069¢'0
$66¢'0
$96r'0

OOfEE'O
19be'0 1£°0OS 00cE*O 8ese'0 0 voce vC9e'0 EVEEKO [S90 (4440 c99€°0
0 pose 06S¢°0

66Cr'0
sesvo
ououtrnNn
8190

BecerO

916£°0
8LO0v'0
681b'0

$S$90°0 91190 07S9'°0 09970 L189°0

0 CTL8C

80S£°0
C8S£°0

0
099¢°0

0
8728S vO79'0 68L7°0
0 c6r9 CT1OE'O
£0L9°0
9€S¢9°0 0 88S vO6C'0
85190 6c1¢'0 c7E9'0

0 O
161S°0 00SS°0 0 veOt 0
Oct
O£LS'0 IL8¢'0 I/€ $S$90°0

06LS7
Ip6s'0 818t°0

6rct
1780 '0 90S vITe'O 0
IScs'0 0
L6cL
t6tS

9S~716970
IYER
SLvS'0 ILLE'0 9SrS'0
Ov0r'0 0 06€S 90EF'0 6SSv'0
60rb'0 c6SV'0 8910 91££°0
8eLv0 vSs8r'0 C690
bs6rv'0 806¢°0

118720
996r'0 O0£6b'0 9TIv'O 97EPV'O
1cOv'0 0
8rIr 0
06I1£'0
Over’ cOEE'O 80Er'0 O9Er'0 €9S¢°0 9LEb'0 SILt'0

0LLbe
S98t°0
8Ltr0 810r'0

816c°0
SOes'0

16Se'0 SP9e'0 061£'0 069¢°0

0 8Pc£ 8cLt'0 LSLE'0 16ce'0 ILL¢°0 9LPE'O ILL€'0 6bS¢'0 0 O£9e

600£°0
SCMmUOtrTN

0
CL8V'0

Te
Of
9LOE'O
0Iver

Obre'0 CIvE'O
‘0 PELs cOE9'0
0 L8C7 0
16L9°0
OcSc
LPLe'0

0 8tvs 0 0£6S

971
S9b7'0 c6£9'0
vOLC'O

£0°0
0 cZIS LSSS'‘0
0 €£9T cS6S'0
0 PL8c

£8070
SALA
9LLV'0 evis'o 0
008720
eevs LeOL'0 E00 971 $66S'0
06S¢'0

CELIO
0
06th cv9v'0
0 ve6c SL8v'0
0
ecle

08cr~
O8CS'0 I8se'0 9tbS'0 S76E'0‘0 SLPS

0LOPEvIee'OvELeoOcIE'O

Lv79'0)

0
1ZOv'0 0
ctor
v8Ir'0 Stevo 691£°0 0
86Sv'0 80S¢°0 0 8eLt
bLor 069%'0 9L9b'0

06SC£
PI9E'O c69E°0 ITLe'0 L$i£.0 I8l¢'O

Lt6c'0 Ieee

vscec'0 O9LT
TS8E'0 S9EE'O 0
O88t'0 9LPE 688t°0 0

$600
CcL8E
ILcr'0

£6S7'000670
0 0Ofte
89Ir'0

b96£°0

OITIO'O 8S0S'0 vO9S'0 00170 I119°0 06770

C4LS0 I18¢°0
889¢°0

06S£°0

0 CI8b 0 C875 L6¢7'0

0 L8P¢c'0
CCLS OIc10'0 1ZL9'0

0
SISO S880 0
Secs
SISc0 86970 8P09'0

0plete6670
I8bv'0 1999°0
991P'0 Ocrb'0
0 87L7 099F'0
0 888C I1¢s'0

€1L70
8L19°0
8S0€°0

89L£°0 806¢°0
0 6767 OCOr'0
0
S6Lt°0
veoe

COI6ctC
8Scb'0

17670
BLLE'O 9brb'0

v8ss'0
OLTE'O

O8se'0 T86t0

0€c0£0061°00 69570
0tc6v
890£°0
ILte'0
099S°0

91870
OPree

8cIv'0
pSle'o
LLero0
OfPrr'O

841
842
panunjuoy —(1°9°9)I AGB
SMO][PA X-UdIIHHSMO][9

AD0°S ADSL AD0°01

|
07/6 998L'°0 O€8h'0 76050
SI 7Z8LP'0 6P0S'0 £99F'0}| 88IS‘O OPSP'0} OZES'O 81/6| 998270 PSEr'0} |
80SS'°0 |
80IP'0 6695°0 |
ZOIE'O OZ6S'0 ZEOE'O 8bLS'0
9I TILP'O LL6r'0 |
S6SP'0| POIS'O LLrr'O ¢7ZS'0 91 8870 |
E8ES'O |
8S0r'0 ISSO |
I8SE'0 PS9S°O 6LOE'O Orbs'0
bl 7ZO09F'0 |]
698b'0 |
£OSP'O €66h'0 €6€h'0 IOIS'0 vl ZIZPO |
LETS'O |
€66E'0 67ES'O |
ISSE'O 6EES'0 SIIEO 6ZIS'O
ra 69S¢'0 |
L7SP'O SSbhb'O 6ILb'O |
69€b0] 678h'0 ILZb'0 O%6b'0 Tl 80I¥'0 |
8Z0S'0 |
II6E'0 780S'0 |
8ISE'O ZPOS'O 6EIEO 678h'0
OI OLEV'O |
69EF'0 SLZV'0 |
67SP'O |
10ZP'O 7TZ9P'O OZIb'0 p69r'0 Ol EL6E°0 |
I9LPO |
OI8E'0 16Lb'0 |
ILPE'O SELP'O SSIEO scsr'0
8 PSIP'O |
98IP'O O80rF0 |
6IEV'O 61070 Z6EV'O| LS6E'0 OStbr'O 8 PESE'O |
O6Pr'O |
869€'0 L6rv'0 |
PIPE'O SIbh'O LSIEO 6STh'0
9 OI6E'0 |
TL6E0 8S8E'0 |}
ILOPO TI8€0O EZIP'O I9LE0] CSIv'O 9 OL9IE'O |
8LIP'O |
TLSE0 6LIP'O |
ISEEO IIIPO ESIE'O 800r'0
v SS9E0 |
8ELEO IZ79E0 |
66LE'0 I16SE0 TE8EO| 8SSE'O TS8E'O v 66rE'0 |
998E'0 |
LEPE'O 198E'0 |
PLZE'O €6LEO PrIEO IIZE'O
||]
PZIEO
PSPE'O
86IEO
OOSE'O
6ESEO
H8SZEO|
PESE'O
S9EEO
IZEEO
6PEe'O
LESeO
SLEEO
POSED
re
O6EE'0
TLHEO
ZL7SEO

2/8 O16S'0| I8L7°0 0r89'0

7 9b87'0 69590
02/8| O16S'°0 160S°0| 006r'0 07 |
LZIPO SS8S'0 Z6SE'O |
SE79'O 81670 ¢S79'0
SI E££0S'0 |
SS8r'0 LP8r'0 |
69050 6OLF'0 |
OZZS'O OLSP'0 99€S'0 81 O ILE |
LSSS°O |
POIP'O S8LS°0 S8SE'0 |
€909'°0 L867'0 616S'0
91 LS6b'0 |
O08r'0 I6LP'0 |
Z10S'°O 8S9b'0 |
8SIS'0 SZSP'0 ¢6cS'0 91 L7ZEPO |
SLPS‘O |
190h'0 IP9S'0 69S€'0 |
86L5°0 EPOE'O 8LSS'0
al TH8r'0 |
ZILP'O 66970 |
O%6P'0 PLSb'0 |
79060 OSbr'O I8IS‘0 vl I197P'0 |
PESO |
IIOF'0 89PS'0 9PSE'O |
O6PS'0 I60€'0 LvZS'0
rat 8L9b'0 |
68Sh'0 Z9Sb'O |
88Lh'0 SSPr'O |
LI6b'0 IPEh'O OZOS'0 a PSIP'O |
ELIS'O |
P76E'O 661S'0 TISEO |
PHIS'O PZIE'O 976P'0
|
E9PE'O
OFIEO
109¢'0
16Lh0
YI8E'O
6L8P'0
IZ0r0
6980
061770
16L¢°0
OLEP'O
|-109F'0
E870
ZILP'O
69Pr'0
EZPb'O
Ol
8 IE7HO IEZPO
| 8SIP0 8LEP'O
| 880F'0 99Ph'0
| 800b'0 O%St'0 8 8S8E'0 OSSP'O
| 969E'0 THSP'O
| 8ObE'O TShr'O
| 6bIE'O p8Tr'0
9 696£°0 600b'0
| EI16E'0 LIIPO
| Z98E'0 SLIP'O
| C€O8E'O 9I7P'O 9 069E0 OEZP'O
| ELSEO PIZP'O
| 6EEE'O 6ZIb'O
| OSTEO PlOr'O
|P89E'0
OPIE'O
LTLE'O
99ZE'O
6O8E'0
CEPEO
TL8EO
POSEO
I8S€0
€88lO
7ZZ9E'O
198E°O
OS9EO
9Z8E'O
ISLE'O
vL88EO
Zz 9OFE'O P8~EO
| PEELED 8ISEO
|] 6LEEO OVSEDO
| 6SEEO ZTSSE'O ve LZEEO SSSE'O
| P8ZE'O THSE'O
| PHEIE'O ZOSE'O
| IZIEO 6SPE'0

€689'°0
87LZ7'0
T/L
90€b'0|
91870
€9S9'0
072
SI |
SSSE'0 7hZ9'0 $0670 9819°0
OI/L 90Er0]|
| 6P0S0 EP8b'0
| SL8P'0 LPOS'O
| 8ZLP'0 SIZS'O
| 78Sbh'0 CLES'O co) 99€b'0 8LSS°0
| 9LOb'0 E8LS°0
| 6rSE'0 0009'0
| 186770 C£8S'0
+l | OS6r'0 ELLY'O
| I6Lb'0 S96b'0
| 7S9P'O 8ZIS‘O
| 9ISb'0 LL7S‘0 vl 60EF'0 6S¢S°0
| LZOF'O SI19S'0
| ZESE'O OOLS'O
| LPOe'O 8srs'0
a + 908h'0 999%'0
| LL9OP‘O LS8r'0
| LrSrO SOOS'O
| O@bP'O 1€1S°0 TI EIZH'O OLZS'O
| 6r6E'O L9ES'O
| ZOSE'O 8ZES'O
| Z60E°0 ¢60S°0
OL | 909F'0 9ISb'0
| 60SP'0 969b'0
| OObP'O O€8h'O
| 687r'0 LE6v'0 Ol I160r'0 O£OS'O
| TZS8E'0 IS0S'0
| I9PE'O OS6b'0
| EZIEO TELV'O
8 ESEV'O ZIEVO
| ILZP'O T9Ph'O
| P8IP'O 89Sr'0
| 060F'0 1p9r'0 8 616E'0 P89P'0
| TZLE'0 6990
| 9OKE'O 8SSP'0
| OFIE'O L8€r'0
9 ELOb'O ELOF'O
| 600F'0 86IP'O
| £P6E'O P9ITP'O
| F98EO SOLh'O 9 8ZLE'0 YIEP'O
| I8SE'0 167P'0
| IPEEO 16Ib'0
| ZrIEO 8S0r'0
v I9LE'O OO8E'O
| BILEO S88E'0
| LLOE'O S76EO
|] b79EO 1S6€°0 v PESE'O ES6E'O
| LEPE'O 676E'0
| LOZEO 8P8E'O
| ELIEO POLEO
o XEPE'O LOSE'O
| 6IPEO OPSE'O
| 96EE0 8SSE'O 69£€E°0| 69S£°0 c 8ZEEO 69SE'O
| P8ZEO 6SSEO
| OGIEO YISEO
| LIIEO 69r£'0
 6r6r'0

SLIv'0

96bE'0
vOOL'O

£9Sr'0
6619°0

9709°0

880¢°0 8LS¢°0
L887 19€S'0
TO8S
00L9°0
6LI1L ISLs'06SL¢'0
80¢9'0 188¢'0 09€L°0 cL9S'0
1l0er'0 $60$'0 Orcr
¥189°0cL6b'0

OcLE'0
c6£9'0
£06r'0
8Ses'0
9199°0

C78E0

00L9°0

C6600 LILv'0
eclr'o
€87C'0 €cbl'0
080£°0
6rS7'0

900€°0
690¢'0

CCLTO
98670
0

ITT€'O
8870
0
"0

1€sz'0
LEcs'0 80S¢°0 8S89°0
c8C9'0 Ossr'0
810r'0
OSsse'0 086LLLEss'0 O8Se0

OTE
80670690£°0
800¢'0
960
Leoe'0
980€°0

CLIE:0
9TIE'O
8970

8clt 0
€9L7°0

0v67°0
cOLC'O

L061°0
LOET'O
87970
8010

veLe

CECT
8c0e

S870
SO
O8Te0 br9e'0

OS9E'0
0

60$9°0
997¢'0 8Phb9'0

Orse
0CCPT
06$7'0
8SL7°0 OOTE'O

O
601£°0

0
0

O
pele
0
CL87'0

C8790 961S Icer'0 See9'0 9L6b'0b66£

€8bPr'0 8Pes

LS¢rv'0
ELEEO L96P'0

O8Es'0
6Les'0
00LS

OLTEO 88cr
0 0 0

0
$860
96SS°089Lh'0CC6E'O
OSst'0 066S
006rS 09St'0 0€16$
8990 06tLv
LSIv0
0voor

0
0
cle
6crtISpe'0
86b£0

€9Pe'0
8Ive'0

OOrE0
SSte0
c6le 0

Ssl¢e0
Isce'0

I8ce0

Ipce
0 0
OSre'O 0psec
0vLce
88Ie0 O9TE'O 0S9Tt 0erie
09TE'O
08PcL psle'o

0
88Pre'0

S6ee'0
86be'0

SLée'0

8ree'o
088r'0 vI9v'O
£60S'0 ps8cr

6cLEt0 6I€€0

VLE 0

C86£°0
8Prls'0

cr6S'0
88LS°0 0
v9ss'0
v9TS'08trr'0
80070
c6SE'0 S8Ps'0 L60v'0
cO8S
‘0 0cI9E 0S8s°0
0pses
0CS8h08L9£
crOr'0 cC9E c8S¢'0
6e8e

6cbr'0
19ve'0

cess
0
6010

0s9r'0
pSse0
OSLE0

6SEC°0
S9LE0.
€86t0
68C£0

cIEe 0 90LE°0
T10r'0

0 0

OLIS'0

89L€°0
168¢'0
086£°0CLLE'0 0887ce 060E£

vc6e
0
826
0

€99€°0
S180

C8PE
0

898e°0
8ILt
0 LS8v'0
8ese'0
vIPs'0 890S'°0erly

cSLV0
£08t'0

88070
00S 0

v8rr
900¢'0

ChEe'O
66Lt

o
v6Ss'00610
S880
OLPt'0
8e0r'0
LO9€'0 69€S'09P9b'09€9¢°0
0c09S

0
bLiv
6c£Pr
0

CcIPE'O
69070
6L8¢0

cSee0
[c9e-0
pservo
69Ch'0
6SIv'0

CLSE'0

0
0

0896£
80L¢°0
0c8EL 0Icve
188¢°0 0Orse

cLLE'O
vecr
0
8800
ecer
0

$S$90°0
00c1 0
$00¢'0

LL61°0 97I£0'0 017100

02/9

ré

SI
v
9

c
Zz
vl

Ol
ZI

ol

9
8
9

C
cl

8
9I
Z6£S'0

GELeE0
LEZS'0
890S'0
ZI8b'0
7Sbr'0

06€S0

L96¢'0

S6Lr'0
60¢S0
€€0v'0

8P9t'0

SISO

Icer'0

8Pr8e0 9SS¢°0
119€'0

1c9v'0

cSPr0

68Lt0 88170

cIcr0 cO8t0
68Lt0 cise 0
970S0 Srev0

cLEv'0096£°0 LOtr'OCOLE
89bb'0 06170
v8SIv0

€6Sr'0
88rr'0l0cbv'06L9¢'0
086tt 0 0cere
06SP'0 cLOV'0 Otrr'0IL8¢'0
0OLE
196¢
0

poor
tr6r'00Orr'0 OcIv
80¢S Issy'0 066170CLEP
8S8t0 099€°0

L8cVv'0
€cLyv0 10vr'0

0 0 0 O
00c7S
0L8066ILv'0 109€°0 ‘0 OS8r'0 0Ov9e
LS0S 08897EcLEO 06Ltv
08LLt
06887
vSLv'0
8£9Pr'0 OS8t'0
ceor66170
C86t0
cHSt0

68S¢'0
9tSV'0
Iger0

0
Icer'0
cISb'0 leve'0
090¢'0
0SPLe LOLV'0Ocbr'0 0OLPE
7rOr'0
|
S6sr
0

98070

80s0c6r'0 sotvo LSOv'0

610$'0 strr' O18r88170 €8cr'0
I1ZP ELSv'0

'0 0 0 0
06LP'0 O8Se'0
88Sr'0SS6t'0 c69v'0 0079
9L8P'0 OSsPr'010L€°0 08PLt 06€8£ c9trv'0
ce6v coer
6LSv
LLLV'0 ¢16¢'0
06S 0

4
LSve'0

LSLE0
S06r'0

69070

9P9e'0
6£9F'0

SrLv0

OL9V'0

ersy'0

0 0 0
a0)

000st
Ispr'0

l6lv'0
08LrV'0

9tbP'0
OrivO
v6Let'0

cI8vycsvps6e 008170 9LOV'0 99170

Iesp0
Ovse0

cé6th0 L79v'0
S8Lt'0 Sc8e0
678h'0

9LIv'0
LO9F'0

L98t'0
I@pr'o

0 0 0
0OLS¢
€897'0c6cr'0 16tr'0ps9e'0
Sc9r'0 0O0OLE
876r'0 c0cr
190S°0 LLIVO
»
OeLv'0

c9EPh'O | OIZIO'O | Z/T

C8080

chSv0
O8tr0

pese0

v8Lv'0
LL¢v'0
$98r'0

ec9e'0

cOLt0
S89P'0

896£°0

0
O8re'0
LISv'0Ovse'O
09LV'0
8elr
0
$06r'0

0 00cT'O
OcIS

SCIEOD
$S$$90°0
S000

LL61'0

9/€
I/p

843
844
AGL
Sud0ID suddID-on|gq

—({'9°9)I
panurjuod
DS'T Dds 90'S Ods'L 900!

DSL
h

Zi x fi a fi x h x

OIA
91/6
| 99820
| 0£97'0 99670
vl 0 TILT 9¢Lb'0
cl 98270 l6br'0 87S70 O9IPO S7ETO 96LE°0
Ol IS8c°0 SLcev0 6£97°0 100¢'0 LSbZ'0 COLE'O 998L°0 10€c'0
8 cl6c°0 vsov'0 SEL70 vS8£0 blsz70 819€0

1/6

61¢7
0S$86£°0
0
Ltvco
9967°0
99P8£°0 CE87O L69E'0 €0L70 €IseO
97CE

SS8£0
$1770

C8EcT
Sore'0

089S¢€°0
0

0
0 0 0

0
66SC
8100
v909€°0 €£67°0 6I1S€0 OF87'0 CcCOPE'O
Scce

08LEt

cS97
1L9€70

8ELE°0
LOSt€'0
0
89L7'0
C 8S0£°0 OOPE'O LIOEO LSeeo $9670 €6ce'0

Oootrn
10sec

OccE

€9L70
ecpve
8Ile€0

cS9TO

LO9E'O
0

0€6c0
L8ct
88I1¢

SPs70 1I9be'0
C8870
Orle'0

SO8T0
80C£'0

8cLc'0

6ree'O
OOLCT'O

#2/8
| O16S'0
| 16070 €£09'0

€cee
0L86c
08EeeOcece
6STE'0

60S¢'0 L9CE0
evsc0T16c'0
L06c'0

Lv6c'0
CT 17770 66LS'0 I78I'0 OvEer0
0c 6£€70 19¢S'0 PELIO v9IIbO
0
9S61°0 | 908¢°0
81 cS9b'0
IS¢7'0 60€S'0
EOI 9981°0 980¢'0 6SL1°0
C8LE0
91 £9S7°0 Sv0S'0 OFTC'O O0St'0 T1070 c66£°0

cé6rh
0SP8I°0
O 0
0
SI6TO
CELE 891€'0 IcL1°0 OSPre
vl 19970 O8Lr'0
8rEbO
89ET 8b170 €06£°0

cLEvVO
016S°0

O0861°0
LS0c'0
189€°0 9€6C7'0
88LI1°0 6LIt'0
8981°0 0
cere
(4! €Prl70 pssvo0 6870 161v'0 c87¢c'0 IT8€0
visto

0cscv'0
96170
O£9E0 Lt6l0 881e
0 010c'0 cIvt'0 8L67'0
Ol 67870 l0€P'0
E1970
920b'0 Ofbe'0 OILEO

ScIvO

vScc0
8S61°0

O
99S¢€°0
CSET ScOE'0
OCICO 961€°0 v8Ic
oO €8ee0
8 96870 $90¢'0 Ycle'0 S98bOQ v9Se0 L19€°0
10170

cOOr'0

O8E7'0
00st 0 O00SC'0 CcOETO
€90€'0
861€'0 0 CSEEO
cSEC
9 cS6c0 IS8e€0 C870 COLE'O £6970
CISEO
0v9CT

L98E0
OCI7 SIS7c0
0 0
0
Lv9C'O
6cre 660£°0
68¢7°0 861£
0 ScSc0 8Iee
v 600£°0 vI9EO
vc6c
€cst0 COve'O
87870

€£Lt
061Vc

06£970
0

0
0
0

16270
IsteO 98970 00cE
0 8ILc0 8Lc£ O£€TEO
G £000 pore 600€°0 6Stt'0 LS6c'0 €6c£'0
088Sc

809€°0
O6C'0 89cL'0 6870 este0 00670 8cce0 CSTE'O
CSL

v9re

PL87
9Z/L
| 90€7'0 6891'G| 6bS9°0 0
LOE
8S8h0

18670
06162

PSL7'09CEEO
£4 SZL8T0 §9¢9.:0 SIPFI0 8LLb:0
CC 62070 L109°0 6tS10
€89b0 90£r'0

L6ET
0cIes'0
O

0¢ 18170 pels'0 88910 OLSP'O

IcST'O
00cS'0
0LLtevoO
0
O8e1°0
81 8770 LOVS'0 8rrr
IP8l'0

vLOS'0
6S9I1°0
0
SIStO 9LOE0 Levl'o
oT 8Prbce'0 €OcS'0 Ofer
C8610

S080
€t6r'0
68S1°0
Lé8¢'0

OL€lOcCh'0
O6F1'0

0
0
SL9OI'O 68r1°0
89170 1OT€'0 Psst
0
vl 89570 Ie6r'0 0 6£17 66170

ILLv'0
selvo
97910

L961°0
88Lt'0

velclo
8810 OSTEOO 6cI€'0
ISLI0 CESC
cl CL9TO L99¢'0 S6¢7'0
85070
90€v0 6r0r

EVID 188I'0
0cIPe10re'0

L661°0 C6870
[P81'0
8rIe'O PI6l
Oo
Ol | SLL70) S6tvO0 O SPT0 pl6e

996£

CITT
0€£0C0

OOSby'0
0CE6I'O

0
e917 St0C'0 S9Te'0 v60C'0 96670
8 ° 19870} 6710 S6S70
p9LEO
S6l7

91970L9ChO
0pS8t0

ITI7YIb7'0
0
psec vlOeEO
SETCO 8LI€'0
C6770
9 €£6c°0 EL8t0 8CLZO
CC9E'O
8PLE
0CSECO

L80¢:0
006tt19ct£'0

yo9te9LSE°0

PSs7z'0
0 O

ersc'0 8rbc'0
6900 981£°0
06r7C0
v 76670 vr9e'O OS87'0 C8PE0
OIPe'0 6LEe'0sete

L890
106£°0
EIS7O
se9e'0
OL8t'0 6€Lt'09€9€°080St'0

PEEl 88L1°0COICO6tPe'0

Oo

tHCO cHr9T'O
601€°0
68I1£°0 L970
¢ Loco elveO0 CL6c'0 efeeO

10870
cOEL

IZL€°0
C€99C
9CT7'080970

09cS¢'0
Ofre'O

86870 69870 c8I¢'0 8L870 €vIe'O

0697E

pste

C0670
SIve0

8PSt'0
£08C°0
v9Le7O

S670

T00€'0
L670
0S¢tce

L6ct'0

99€£°0
697t'0
 9ILTO 09L7°0
S8r1'0 €997'0
68¢7°0

0£L70
86770

£9870
¥867
160€°0
9LEC
ITT€
1$97'0cO87'0
18$c0 18870SI0€°0
8L0€°0

C8S7°0

c$6C°0
OvOE'O
0L61°0 09870

00970
v8Prc
vecc0

96L7
cISc
0

0
6CL7'0 0$6c'0 cEle

0
0

0
6061°0

Setc0
8691°0

8LS$7°0
L870
I8Tl°O

OvlLc0
O8r1°0
€cOl'0
8Pc10
80£1°0
80110

0
0
LEErO
8ISt'O

9TIcC
0

8880°0

O9LT°0

P8Ec
0
$907'°0
c£6c
9L6cc80€OsTe
csoe'0

cLIE'0
180€°0
186c°0

0 0 0 0 I9T€O
08787
088720 0ceOEScle'0 0L997
08IL7Ov8c'0
0vLL7 016¢c'0
Ss8st019610vstc €860°0
L9ITO 9LLI0c6CC 89L0°0 86c1
OTE
0

SSTE0
selec
'o

69T£°0
L10€'0

LIZE'0
oO 0 0
8010COLI'OILIZ°00970
8PrT1°0 6870 y9ETO
0Lest0£€0C'0
0OSsc
C180 c601°0Orst'O
€660°0
14450)
Spee 067£'0
ITee'0c9CE bree O8ceO8ct9PCE'0 8010 OLIE
IPle'0

0 0
O 0 0
0Sree
oO
evee 06C£E 0 617t'0
0Lette I€ce019ct'0 0OIcE
SLTE'OO8ce'0OLCE'O 0SLOe 0 6170
861£ 0Ovce80cE'0
0cece
0bece
PI9TO 16Sc 87800OLITO 0681°0
Lt0c'0

CL8C0
89110
ScEl‘O

0
18L0°09¢01°0
0evel
¥060°0 8rrl'0OSssI'0
00I1C'0 0Ip87
09€C'0 ¢£90°07660°06L£1°0
8190C8170 66L7°0
0O08P7c
prslo
l6v1'0
c99L'O

9ECTO
Ipvce'0
87970

098¢°0 SP9t'0 69€£°0 SLLt 889€°0 6P9£Ovse'0

€9ve'0

OLE 0
S8Lt'0 S9ITO

909€°0
80Lt°0

678E°0
06L£°0
08PLe 0p8Se
669€'0 8Pre'0897TE'0
0cCSE 0 pSLe'0 0
008¢°0
088Lt 0OLE
0OcLE 009¢€°0 SLEEO
6$97 cOrTTOc6rl
OSLt'0 8relO

899¢'0 SeLl0
I¢se0 8¢L0'0
cE8E
0 1980°0

Leste
0 $0cc'0
CSE
0
69€£°0
vI8e SOOT"O
0
OcIT'O

0091°0

COLT0
€067'0
8rI7'0

0 0
08870 01S0'0
9£90°0¢160°0
89L0°0 08L77
0€8c18EL1°0
09007 0Ov8c
cSSc0
08PbTe
-
6SSI'0

0861°0
8crl'0

6LL1°0
S610

ceETO
9¢S7'0
69710

LL6I'0

S000 ‘0
00cI
c6lv0

c66£ 0 CS8l0

OI€e 0 O16c0
o9er'0
OcSv'0

80crh'0

S10r'0

8Sst'0

Ovrr'0
€OEE'O

LZEEO 08870
OLOr'O cO9T'0
OLrtr 0 90010

86rvt'0 8c970
66LE'0 PcIlc0
SPr7v0 CICI 0:
OCIIO

69P1°0
672670
$681°0

cOLO0
ISSO

16Sc0
1¥60°0

cO9E'O
8ScS
O€L0°0 Te1S'0

cC8E
Ists‘0

6lvb

ILEEO
0
88cr
SLCI

IL9I0
68070

$6070
€S8t'0
0
L6cc'0
OLLt

IIL70
Sspe0
l6rr'0
11800
0
8trb
85600

61S70
L8S¢'0
96Lt'0
Lb~cO

06S~7'0
CLS0°0
09070
vIl6£°0

OChT'O
6L9t°0
0
epee
6hL7'0

Srsv0
00v0'0
0
€SS00
cory
O€clO
8Ltro0
C8El'0
66cr'0

CcCLtO
€llpO

0690°0
cbsv'0
8S¢vr'0
OLOT'O

0S800
88tr0

0
PLEc
SS9f'0
9L81°0
€£6£'0
0
891P
86t1

Ol6r'0
810S°0 vipb CLIE'O
8Lcr'0 6c8t'0
SOLb'0

SPSe'0
SStt0
LI8¢'0

19Sh'0

ILtv'0

c6t00

L9bT'O
ZOLT'O
9OLI'O

61670
18S0°0

oO
0

OLOI'O
6SII0 10070 OIS7
ceslo C9970
9610
CLEI'O

Sv6C°0
CITIO

SLL7°0
ILSI0

OvOS'0
OLLO'0
cv6v'0

ccOv'0
9801°0
Cv8r'0

6861°0
c9SP'O
OOSI'0
87600

COLP'0
£6710

61ch
CECT

0
S8PI'0
LLOV'O
8$80°0 S7El'0
L7IS‘0 S6Lh'0Issp'0 OSEt7'0 0cCSE'O
Pvs9l'0 ILI7O
8eIvO
6L6t°0 LO87 vccS
$8S0°0
0pree'O
8S67‘0
0
9LL10
SIppO

0
S6E7
$16t'0
86S70
VCLE'O
6€0S'0
82800
OSOT'O
Lé6v'0

0070
c80F'0

868S°08c9S'0
I9LS°0 Ices'0186r'08LS¢h'0 098£°0
€9pS'0 OOPS
vIcS
S10S°0 997h'0
066£°0
06Ev'0

C8EE'O
S6St'0
S6LE°0
09SS°0

cScS'0
$69S°0

IsLp'0
160$°0

0 0
$8L0°0
6090°0 8Iel0$6910OIC
PrIIO
€$60°0 06970 COVI'O
88IT'OLc9I'0 6SE7'0
0961°0

cI970 6tcr0
8PL70
89870
88670
£6770
6091°0
8060°0
6LOI'0

CEPI'O

0
¥L99°0

$609°0

Spree
SSIL‘0

ILIS‘0€LLp0
68¢1°0CI610 0LOIVO
0I€trp0
6cec 81670LOvO'O
c6EEO
IISc8790
1870 0109°0
¥VI90°0
LS8S°0
17800
y89S'0
81010
€pss'0 cespo0c66£°0
€r8I0
LO8r'0
SIIc0 18S7'068670
LIvt'0
VOLE'O
18L7°0
‘0 ILOS
99¢7°0
I8Iv'0
S8LT°0
y76v'0

O£I70
ILSP'0
CSTI'O
80¢S'0

OScL'090L9°0IT19°0
6LLS'0
Ofrs'0

SS At)

Leve0
€96£0
eels '0
81£9°0
$£09°0
cCIL'O

O8tr'0
SvOr'0
SOLv'0
6SLS'0
8169°0

C6£9°0
S8tL°0

I1vs'0

Sele
0

CcOSL'O$L69°0I¢v9'0 0Scrs
0900S
L6SV'0 0OLE
SIcr'0
I@8€'0
0

SPIO9est'0 8L77PLSc
6£L1°0cOIT'O L967'0 Lice. OIL7
881106LS1°0$00¢'0 Iv870 6001°0
Oecl
9brl'O
C8910
60610SSET
S69€°0
vI8Pr'0
LELs'0
$099°0
1L89°0

0 0 0 0
Orel C7610 9TPC’006970
oO 66L7°0
0C687 €660°0LEX
06£0£¥6L0°0 0 c8L1°00 S8tc0
0$9S7 tv6c'0 0920°0Oo
OfO0e'08¢cS0'0 rata)
S€Lc0
16870
8217019Sc'0
$00¢'0

LL61°0 00cI'0
97
87/9 91 cl or 8 9
£4 (46 07 81 a!

845
 846
panurjuod —(1'9'9)1 A198L
Sud21DH (panu1juo2) suse1H-on|g (panul1juod)

DST 50°01 DdS'T 20's Dds'L 22001

OIA A x

Ore0'0 1T09°0
0
06£0'0 | $$S90°0 | 7Z/E 89PL'0 e£t0'°0 brrro
0790°0 CcO8S C800 $69t°0

0c
LZIL‘0
O?LO'O 8cs00 €6tr0
6P0I'0 99L9°0 €880°0 $0950 8IL00 Orer Oo C890
8790'0 08S0°0
Ov6c'0
0
OcIT'O vIPs €v80°0 L99C'0 S€L0'0
0 6L67 1690°0 6SSc'0

c80S'0
$S$90°0
~IPETO 0cr9'0 $760°0 SLCvV'0
979T'O c$09°0 C8El'O L6Is'‘0 c6lvO0
I9TT'O ISOI'O
8P9E°0 OF60'0 LcOE'0
L797'0
¥L80°0 8620°0 ISIZ0
8r6r'0
099T°0 88c1'0
0790 8SIT'O ILO€'O 9801°0 90L7'0 81010 [8c7'0

89S0°0€cOl'0
8I18h0
¢60¢v'0 TIvl'0
Se6l'0
C8970 cSStO O8S¢0 OIPlO 8IT€O0
9CET0
v8Lc0 OScr'o IIvz'0

L99¢'0
0 0
pL6t 0 889I°0
0 8¢C70 O8tr SPsl'0
Tese €OLT'O 6SIE'O
cL87
OC9T'O ISst'0 IL$7'0

cEE008620°0 9IST'O

TO6T'O
ch9S'0
90¢S'0vS6r'0 SOsrv'O
OL6I'0
Ip8e'0
v0
OO
ILvc'0 cE1TO
89be'0 O00 88I1€'0 876I'0
0 8S6C I198T°0 CCLEO

T1zs'0
OO8T'O
79TI'OOl€rO0

OLI7'0
0
0
0
669£°0
OCC
O8Lt
IILc°0 0
Leve'0 98E£ EPET'O OOcTE'0
CL77'0
Ip0E'O [7770 988c°0

88070
IOIv'0

SEbCO
cSLV'0

0 0
LEse0 Scsz70
0 S€6c 6tre'0 66L7'0
0 ILce CHLTO
0 c6lE 6697°0 OcIE0 0997°0 Osoe'0

chev

cr9T
106€°0
9PET'O

C
Lette0 Pr8c'0

S160

9€87
81970
L99¢'0
T6£€°0

6667°0
81 9T a! cl ol 8 a v OOSE'0
|
9I/Z
8S€L°0
9ZIEO'O
6700

vl
07800 0989'0
692L0°0
08820 v7L0'0 8L¢2'0
971
£00

Letvo0

986¢'0
$87c0'0

0
cl LOETO 80£9°0
86990
eLLI'0
Ol OSOI'0
088S

96620 1660°0 c8S7'0 67260°0 €€I70

0687'0 €L6v'0
6790°0
66S0°0
$ss0'0

OL@b'0
0

LL7O'0OcIT
8
C6170
cvOS'0 0
Leoe
SOVT'O Szeto OIL70 8sSclo T€ec'O

€8Ir0
1$80°0
9LSE0

6SLv'0
0
9ccSv
€6bc'0 ersl'o OIT€O LPLI'0 €$8c'0 6991°0 OLS7'0
Iss¢0

ICEl'O

169S°0186v°0
€SIS'0 SOSP'0
6S0P'0

9L70°0 CchrlO
v £9Lc°0 866£°0 O
PEt
OSTEO 79170 18670 9602°0 06L2'0

vrcv0

€8or'0

8890°0 6L61°0
crsl'0
O61LSS\'0

oO

0
0

0
4 8L62'0 LOSE'0 L697°0 SLI€0 1S97'0 860€°0 9097'0 O10E'0

8170
Tecpv
vE60'0 06€L£0
IL61°0

ccoOl0 00770
€86£'0
CSPE'O

Sse
SOLI'OCre
evec
LISt’08LEe

O09ST'O Ov9c
SOLE0
OVSTO
|
072900
89689°0
OITIO0| O£S0°0 tp6r'0
c60cIpet'0

69870
07870

OOFE'O
9 TILI0 619S°0 brEl'O SOSr'O 6S01'0 0
S8¢c'0
PLOr'O
6cI7

0
v PS¢e'0 68br'0 6S170 L96£°0 COLT'O 89L7°0 8S9T'0 96¢7'0
0ILc£08Sre

I1S0'0
T16€°06SSE0 8SIr'0
9L¢0°0

0
OIc710'0

c 0 O167
0 vest 8SLl7°0 v8re0 OEVT'O €cOE'O C9IETO C88C'0

8910
966¢
6rcl0
S9LT0 69110

610b'0
£601'0

CSPE'0
098¢c'0

PELE
090rt
T0ce'0

0goes 0 0OSPeOILS'08Icr'0
€881'0

0816c6$S0°006¢¢'0
OFOT'O
€SZLI'O

0€£8C
Ivl€0

68970
LOVE'0
68¢£°0

009¢'0
00Sc'0

P9SE'0
 0$87'0
vSOE'O

ObSTO 9LIZ0

£9870 LPL70O
$9270 IL9I'O

OPLT0O €£STO

8867'0 19620
8r87'0 85070S$tb7'0

O€07'O

TI’Z'O

$0670
01670
8£0E'0
ITI0£'0 S970OL9T'O
01
qd0

8LP7'0

98970

6$67°0
CSET'O
OFb7'0
6€81°0

LEIT'O
8661°0

cSETO
LL970
tvpc7'O

SOSZ'0 Lrtez~O
6P97'0
76L7'0Se0£'0
LZOE'0

IP77'0 SL61°0
OPt7'0
£9870T1I9T'O
~OLLT°0
SO00E'0
O000E"'0

8Lt70 89120
9887'0
1087'0

SS$67'0 £9670
8t970 I€Sc'0

78070
S$cCTI'0
7170
6c£1°O
PLIZ'O
prrl'O
%777'0
SSSI°0
66770
8691°0
b8t70
LS81'0

CLS70
I1@Z@'0

17870
6$97'0
8P97'0
9P87'0 PCCT'0
8IP7'O60870
add'l

S9ETO
19910 vocc'
IISlO

0
COLT0
9S87'°0 OIPT'O
9PS7'0€€87'0 T98IO}chOTO LIPTO|
Sv6l'0}
sonjq-oding

6661°0

S9E7'0

8L67°0
8Ss70
88170
| | | | | P6LI'0 |
6tcl'0

|
|

|
|
OL9T'0
1987°0 L970
8SP7'087870 6891°0 6770
S9ETO
q@d0's

IcStO

61170)
66L1°0
C7810

86L7'0
8LL7'0
CSLTO
vL6c'0
vi9z'0

vScc'0
86L7°0

L7éve'0
96S7'0

tS67'0

L617'0

velLeo
€L81°0

eesz'0
16670

€767
097
|LSOE'0
SIOEO
csoe'0 |6£0E'0
£00£'0
bveoe’oO |Eb9T'O
~L897T0
T1970|TIOEO
78670
€00£'0 LSITO}
OrslO
8S8l'0
$8770}
02070€987'0|
IPr07'0
LIvc'0
6bb7c'°0
L897'0
6€L7'0|
99970
|£7670
81670
80670

|
8Ist0
8e9l'0

€LL7'0

97070
LS67'0 Lvot'0

166700
LL97'0

c£6C Sc0e'0

90700

vO8T'0

6SS1°0

v681°0
S¢tcc

8L07'0

Ch67'0
€LL1°0
81610SSt7'0
080700Lbrc'0
0c€997

88rc
C8870
q@dS

£90€°0

60L7°0
6L87'0

c9IT'O 60£7'0
86b7'°0

8981°0

66S7'0

8cL1°0

8L77'0
8t0c
6CL7'0 8870

07L7
$L67'0

8SL7°0

CSET

OPT
0

0
0

0
0
PVSL10 78970 S6rl'0 89610
e910
7

c9S7'0

8ESc
O

tl6l'O
0vL7C
$6070S9vT'0
0L687 cH9l'O
€6L1°0LSIc@0
S9t70
Lvsc'0
00970
016¢°0

LL61°0
S00t'0

998L°0 90EP'0
COorTnNn

ZI/L
Or8c'0
cl0t'0
89b7'0

vZ67'0 09L7'0 ceotO 6b17

vrecOLSL7O
Lt9T0
68170

IL87°0

0
TILT TIS70 Octo O981°0
L907LvScO
000C'0
PSPrl'O

66¢7°0
0$97'0
Lv6l'0

86cC'0
PIITO
8LL1'0
62910
£0810

O 0
6b67'°0
9LOE'0 S£67'0
c90E'0
6727011620
L8S7'0
8IL7'0 €0CC'0
E8810LL7C0
6SS70
08Lv7'0
0c8ETO
8L07 87L7 80670
6£0£'0
$8970
98870 SOST'OL6LI'0
€07I'O76vl'0
999T°0 8PS7'0
v961'0 667709967'0
17870
19670 17870 L8L7
1£97°0 tv0c'0VLtecO
6L81°0 Ltsco18870 crv07'0 cl9c
8L81°0
ITZ1°0 80870
porto
v8sto

8PC7'0

£0870
L000

0
LISz0

L80¢°0 0
LLOE'O £0770 80L7'0Leot’0
O£cl

899709670 £0£7'0 8S0¢°0

9970 Lt6c'0 O£C7'0
LIv7'0 000£°0
81810 9tPC 9LE1'0 ZEIT‘
ZSET'0
CS7C'0
CLYT0

C7670

c6LI'0

0
9SS1°0P6610 Z097'0
rELI'0 S870
cello £981°00
8t6c'0

€6vc'0 6L87°0
£90£°0

Lvec0

88970
L€67'0
6tt7

88870
ceOe

$77701$97°0
$970

91070
S$1c7'0 10L7'0
6170

Ocel 17070
88970

19L7°0
S00e°0
cOre’O

cLIZ'0
68L7°0

8S61°0 61Sc
L8vc
8070

SOST'O

v6L7'0
6167'0

0
0

960€°0 LOEc'0
88870
86670 O£PTe'0
06CL7 08L0t
vS87'0
6LS7°0 TL67'0
0
0

Lttz0 98610 ££97'0 Ol€el €881°0 6LSC°0

$L97'0

6CL1°0

0
C6770

6S7'0

$891088070
OcETO
0CH87
£7870
vS6c

80670 $1910
IL97'0 OlPce'0SL87'0 96r1'0
8LLI°0vO?C'0
16L7°:0 ILOt'0

LSv7'0
ZSL7'0£7670
6£87'0 CLYT
18S7'0 6S¢7'°0
19S7'0
S7IE'0

Lv6l'0 L897'0
99170

90

c6v7'0
€8c1°0
1I9v1'0

000€'0pZ1E°00
L90€°0
0

CcL9T'0
IL87°0
SLL70 09670
08e£0t 08Ptc C8970
OLTEe'O 068L7800¢°0
66870L60t°0
0

08970 99070 ve9l'0

Sevl0L6L10 81¢7'°0 0991°008070
cIETO
6L81°0
€LOt
0

0601°0

L691°0

0170

60670LL81°0v9CT'O L687°0
C9PT'O
89970 8077067970
v661'0 0v6cl
L987°0 O8r1'0 ILS7'0
S870
LL61°0
016S°0
998L°0

090h S00t'0
Or
71/8

847
848
panuruod —(19'°9)T F198L

g soni (panuijuo2)

soni
7 q@dS q@d0's GdS'L Ol q#d0

g-ading
(panuijuod)

A
h

DIA
cS6l'0
8LL0°0
0r80'0
IZL6I°0
6S91°0 $780°0 ¥661'0 v060'°0

0070
¥89I°0 6680°0 0700 $8600
¢€IZI°0 0860°0 8070 y90I'0
Leol'o CrLI'O 8SO01'0 SLO7'O OvIIO
8Icl0 LOTITO 86Z1°0 S8IIl'O OZITO 9ScI'0
9I/r
| 00710 0060°0 €L61°0 LI€l‘0
19810
9TETO 0 OLIC tte
88710

0
|
tO
SSIIO 9€E1°0
LZOl'0 LS0c0 vOcI’O
S8ZI°0 SS91°0 6Lv1°0
IP6I'0
89rl°0 OC7@C'0 €0S1'O
vos

Olel0 eLvlo
LPTI'O 60070 £961°0
| €6£1°0
0 Lest 6$91°0
0 LEO 67910

SIP
L8rl0 veol'O 8677'0 6S91°0
L9PIO vSETO 8PrI7O
| 0
10910 8207 80C1'0eIsto evslo 8170 IT8t°O 88€70 LtslO

O8sI°0
1891°0 SO81°0
LELI0 vesz7O SPETO
| I@81'0 c£770 0S0C'0
vOETO
€c07'0 L670 8£0C'0

66C1°0
86910
€LL1°0

09Z1°0
€681°0 ¢661°0
80L7'0
8070 CLS7'0
| cOIZ'O OLYT'0 00ET'O ILb7'0 9970819770
€977'0

cIST'O
6re1°0 8881°0

vS61°0

0
0

LSI70 S£C7'0
0970
CL87'0 78L7°0
| 8870 vOLT'0 09SC¢°0 LS97'0 82S7'0 6SL7'0 CCSTO

09170

8601°0 6SL1°0
¥60C'0

6cV7'0 L8¢70
LZL7'0 8€0€°0 76670
| €€Lc7'0 Lv67'0 CH8C'0 19870 61870 11670 vO8c'0
€vez

09070
Love
$c61°0ScETO

€SL7'0 78L7°0

€9€7'0
87970
L6S70
0c6eTL791'0 £0170c9ST7'0

EcLc'0
01670
91870

9L87°0
8091°0
08r70'0

$$$90°0
ITSO'O

ZI9T'O
cvS0'0

I79I°0
9$S0°0

ct9I°0
6090'°0
97610OS6I1'0
8I61°08£€61°0£961°0

$$90°0
€0$0°066$0°080L0°0

¢CPr9l'0
IIZ0'0

0
8cCl $680°0
0$90°0 0070

LL9I'O
L’80°0

0)
£40)
8Itlo
£980°0
CLLO'00£60°0

C8610O£07°0

vI/¢
| $$S90°0 S901°0 S8cl'0 IScl0 0 Itvl silo
87600
c8L0°0OFLI'O
0907°0
OcOl'O

6860°0 €961°0
crOlO 189L0
| Mello 0
8ccl 0 O09F1 86t1°0 LSSI‘0 9SE1°0
8S91°0COLI0S9LT'O
8POl'O
8III'O

7607°0

OCCI'O cEITO6SC1°0 6L81°0

| trelO certo SL91°0 9LS1°0 8IZ1°0 c9ST'0
CHIT

8IclO
88110

Presto
OScl’O

TIStO
0 leew LeSl'0 61170
| esslto 8S91°0 $0610 O8Z1°0 8061°0 66L1°0
£0610
eStl0
907c'0
LOFVI'O

97810 9€S7T0 Sesl0 SLE70| SZ81°0 tlic

e€c6l0 0
cc0C 0
cCITO cS0Cc
$0070
9€S1°0
S9SI'O

8L77'0

oO

E8170 8PL70 9LITO TE9TO| 00ZC'O 0 L9¢e'0

9PCC cltc0 0 £6tc 19¢7'0
I9L1°0
L8tc0

6FI7'0
98Z1°0

9£97'°0 €862'0 |
L197@'‘0 17670 919¢'0 I£97°0 10870 £9970 80L7°0
0 61L7
.

9SL1°08S77'0
IIlEec
[e0c'0

IIs7c'0

O10c'0

9€ST0
6170

09970

S6t1°0ecs810SOve'0
61£7'0

O¢S7'0

0091°0101709SL70

crSl'0L861°0 LS8C°0
LLLCO
£8970
L¥87'0
0L9C'0
 preoo0cP0'0 81Z0°0
8080°0
€v90'°0 Ofel'O
ISsl'O0 0vc0°0
18700 9£P0'0v9S0'0S$PL0'0
8£90°0 OLFI'O
8781°0

97£0°0
08t0'0£6¢0°0 6980°0
6c01'0
0
8ccl 08770
II6I'0Scé6I'0 8661°0
8Z6I'017070 O0TT'O
v6c7'0 82610 9L61°0
$9610 0cIc'006770

8I61'0Le6l’O
ILp0'O
6L£0'°0 796I°0
8LS0°0
661°0
07S0'°0 cS07'0
0160°0
L807°0
9ZOT'0
6€170 Ovsl'O
OLIT'O £0870
L9S7'0
00970
c9ICO
OFbe'O 9¢€61°0
cv6l'0
cS61°0 1661°0
8007'0
04070
8070 0061706S77
LL9CTO
0820°0
O1€0°0€L¢£0°0
60%0'0 v6S0°0 6£80°0
ISvo'08£S0°0 $$60°0
v60I"0S6rl'O
8ScI°0 8rI7'0
06L1°0 Ov10°0
c8S7'00910°0
0810°0
c0CO'00L70°0
vec0'060£0°0cOv0'0
bSP0'0
81S0°0
€8S0°0 cCrl'O
IvL1O17810

0891°0

9891°0
O£9T°O

9691°0
6891°0

OcLI'0

000C'0
OL9T'O

vost'o
60ZL1°0
0991°0

8cL1°0
Se9l'0

€I8l'o

Lpsc'0
8eZ1°0
€c91°0

$0070
68170
87910

IOLT'O

1691°0
18910

890

IOLI‘O

COLO
Ove0'0 chL0'0
c6r0'0 9990°0 cS¢0'0 8890°0
¥080°0
0$60°0 0O1Ec
c9OLI'O

Ocre'0

c891°0
c891°0
S891°0

cL81°0
Lv9l'O

CcILC
c881°0
Ov9lO

€s9l0

cECT
OOST'O OPcI'O
€S7I0 £€L80°0

8£97'0 v79T'0
£9770 C6170
£9ET'O 8POI'O

S89I1'0 T6bl'0
¢@r6l‘0 II8l°0
0

O
veloL981°0
0L80°0

L?ve'0
8L91°0
T1070
S810
Lbrl'o89€7'0

Lbvl'o
Svc7'0
LS81°0

$1970

IT61°0
9LOL'0

Oest'O
8Lc1'0

LSII‘0 0Och
16rl°0
OvslOSLIZ 6€S1°0
$681°0
eL¢cl.0
9TIEO'O

0
99T1°O
ceelSc81°0c6SC0
Oo 09€7'0

OIZ10'0
9€ ve O€ 8Z 81 vl ral Ol
9I 8
ge 97 v7 ref6 4
8/7

c091°0
LSII0

8SS70

9PET°0Le6l'0
SCLT0
0£Tec 8Icl
0 16¢c'0
0pl6l
t9Sl°0
c6tl
96€1°0

cOTC'O
9ILI0

0 060€C
c691°0

66L7°0

LLOIOe8Z1°0
O8cI'087070
6£91'O
00rC'0
972070
CCH
O

[SOLO £9070 YILIO

8960°0
COLO
SPLI'O

elelo 0SbsSc
8S91°0 06LSC'0
16cc
8I1ScO
CITI

8070
O

$960°0
8SS1°0
SPIO
L781°0
LIGEOPL870
COTGO6SS70 LL9TO
16770
9L07'067970
80610
87810

OFCl'O090¢'°0
8III0O
11600 | 9ZI£0°0

8570OP67'0
8LS7'0
Ic9I'0 C?e70
I8L70
|6910
LIO9T'O
89170
HCETO
OIZ10°0

0 8 9 v c 8 9 v C
iT

1/Z

849
850
F198 —(1-9'9)T panuluoy

9/6 ; : 8IZE0 Sr87'0 9/6 998L'0|

| ZZEEO O167'0
| IEPEO 88670
|] ZISEO TSOEO
| O6SEO SIIEO
v £9670 $9870
| E£00E'0 OL8Z'O
| LITEO 876770
| OLIEO 99670 v PEZE'O OLOE'O
| IOEE'O O90E'O
| OSEE'O 6600
| OOPE'O OrIe'O
z OSOE'O ISOL'O
| L9DE'O 090E'0
| LOIE'O I80€'0
| 8ZIE0 v60£'0 Z 6rIEO SOIEO
| ZLIEO 9ZIEO O6IEO| IPIEO SOTEO| SSIEO

r1/8 olen THEEO 6bE7TO FI/8 O16S°0|

| 1Z79€°0 96¢7'0
ZI LIIEO OLEZO
| TZIEEO OLPZ7'0 ZI ZSSE'O P6S7'0
| 8I8E'O THLZ'0
| ZOOr'O 6S$87°0
Ol OL8Z‘0 O8EZ'O
| YIIE'O L6rZ'0
| TZ8ZEO Z8S7'0 Ol 6LHE'O 66970
| S89C'0 87870
| OF8E'O O€67'0
| €86E'0 6POE'O
8 00870 88rZ'0
| PI6Z°O PES7TO
| OIIEO 97970
| OSZEO 0OLZ'0 8 90FE'O £6L7'0
| OLSE'O 00670
| Z89E'O £8670