Magnetite Biomineralization and Magnetoreception in Organisms

Magnetite Biomineralization
and Magnetoreception
in Organisms
A New Biomagnetism
TOPICS IN GEOBIOLOGY
Series Editor: F. G. Stehli, University of Oklahoma
Volume 1 SKELETAL GROWTH OF AQUATIC ORGANISMS

Biological Records of Environmental Change
Edited by Donald C. Rhoads and Richard A. Lutz
Volume 2 ANIMAL-SEDIMENT RELATIONS
The Biogenic Alteration of Sediments
Edited by Peter L. McCall and Michael J. S. Tevesz
Volume 3 BIOTIC INTERACTIONS IN RECENT AND FOSSIL
BENTHIC COMMUNITIES
Edited by Michael j. S. Tevesz and Peter L. McCall
Volume 4 THE GREAT AMERICAN BIOTIC INTERCHANGE
Edited by Francis G. Stehli and S. David Webb
Volume 5 MAGNETITE BIOMINERALIZATION AND
MAGNETORECEPTION IN ORGANISMS
A New Biomagnetism
Edited by joseph L. Kirschvink, Douglas S. Jones,
and Bruce j. MacFadden
Magnetite Biomineralization
and Magnetoreception
in Organisms
A New Biomagnetism
Edited by
Joseph L. Kirschvink
California Institute of Technology
Pasadena, California
Douglas S. Jones
University of Florida
Gainesville, Florida
and
BruceJ. ú ~ Åc ~ ÇÇÉå =
Florida State Museum
University of Florida
Gainesville, Florida
Plenum Press • New York and London

Library of Congress Cataloging in Publication Data
Main entry under title:
Magnetite biomineralization and magneto reception in organisms.
(Topics in geobiology; v. 5)
Bibliography: p.
Includes index.
1. Biomagnetism. 2. Biomineralization. 3. Magnetite. I. Kirschvink, Joseph L. II.
Jones, Douglas S. III. MacFadden, Bruce J. IV. Series.
QH504.M34 1985 591.19'214 85-17037
ISBN-13:978-1-4613-7992-8 e-ISBN-13 :978-1-4613-0313-8
DOl: 10.1007/978-1-4613-0313-8
©1985 Plenum Press, New York

Softcover reprint of the hardcover 1st edition 1985
A Division of Plenum Publishing Corporation
233 Spring Street, New York, N.Y. 10013
All rights reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted,
in any form or by any means, electronic, mechanical, photocopying, microfilming,
recording, or otherwise, without written permission from the Publisher
Contributors
Kenneth P. Able Department of Biology, State University of New York, Albany, New York
12222
Kraig Adler Section of Neurobiology and Behavior, Cornell University, Ithaca, New York
14853
R. Robin Baker Department of Zoology, University of Manchester, Manchester M13 9PL,

United Kingdom
Subir K. Banerjee Department of Geology and Geophysics, University of Minnesota, Min-

neapolis, Minnesota 55455
Gordon B. Bauer Department of Psychology, University of Hawaii, Honolulu, Hawaii

96822
Richard P. Blakemore Department of Microbiology, University of New Hampshire, Dur-

ham, New Hampshire 03824
Edward R. Buchler ORI, Inc., Silver Spring, Maryland 20910
Ruth E. Buskirk Institute for Geophysics, University of Texas, Austin, Texas 78712
Shih·Bin R. Chang Division of Geological and Planetary Sciences, California Institute of

Technology, Pasadena, California 91125
Michael Chwe Division of Geological and Planetary Sciences, California Institute of Tech-
nology, Pasadena, California 91125
Tom Dayton Department of Psychology, University of Oklahoma, Norman, Oklahoma

73019
Anne Demitrack Department of Geology, Stanford University, Stanford, California 94305
Andrew E. Dizon Southwest Fisheries Center La Jolla Laboratory, National Marine Fish-
eries Service, National Oceanic and Atmospheric Administration, La Jolla, California
92038
J. Robert Dunn Department of Geological Sciences, University of California, Santa Bar-

bara, California 93106
Darci Motta S. Esquivel Centro Brasileiro de Pesquisas Fisicas, CBPF/CNPq, Rio de Ja-
neiro, Brazil
Paul Filmer Division of Geological and Planetary Sciences, California Institute of Tech-
v
vi Contributors
Richard B. Frankel Francis Bitter National Magnet Laboratory, Massachusetts Institute

of Technology, Cambridge, Massachusetts 02139
Cliff Frohlich Institute for Geophysics, University of Texas, Austin, Texas 78712
Michael Fuller Department of Geological Sciences, University of California, Santa Bar-
bara, California 93106
William F. Gergits Department of Biology, State University of New York, Albany, New
York 12222
W. 1. Goodman 2-G Enterprises, Mountain View, California 94043

W. S. Goree 2-G Enterprises, Mountain View, California 94043
James 1. Gould Department of Biology, Princeton University, Princeton, New Jersey
08544
Douglas S. Jones Department of Geology, University of Florida, Gainesville, Florida 32611
Timothy K. Judge Department of Biological Sciences, State University of New York, Al-
bany, New York 12222
Roger 1. Jungerman Department of Physics, University of California, Santa Cruz, Cali-
fornia 95064. Present address: Department of Applied Physics, Stanford University,
Stanford, California 94305
Joseph 1. Kirschvink Division of Geological and Planetary Sciences, California Institute
of Technology, Pasadena, California 91125
Henrique G. P. Lins de Barros Centro Brasileiro de Pesquisas Fisicas, CBPF/CNPq, Rio
de Janeiro, Brazil
Laurent Longfellow Department of Physics, University of California, Santa Cruz, Cali-
fornia 95064
Heinz A. Lowenstam Division of Geological and Planetary Sciences, California Institute
of Technology, Pasadena, California 91125
Bruce J. MacFadden Florida State Museum, University of Florida, Gainesville, Florida
32611
Stephen Mann Inorganic Chemistry Laboratory, Oxford University, Oxford OX1 3QR,
United Kingdom
Janice G. Mather Department of Zoology, University of Manchester, Manchester M13
9PL, United Kingdom; Current address: Zoological Laboratory, Institute of Zoology
and Zoophysiology, University of Aarhus, DK-8000, Aarhus C, Denmark
Bruce M. Moskowitz Department of Geological and Geophysical Sciences, Princeton Uni-
versity, Princeton, New Jersey 08544
Michael H. Nesson Department of Agricultural Chemistry, Oregon State University, Cor-
vallis, Oregon 97331
William P. O'Brien, Jr. Institute for Geophysics, University of Texas, Austin, Texas 78712
Georgia C. Papaefthymiou Francis Bitter National Magnet Laboratory, Massachusetts In-
stitute of Technology, Cambridge, Massachusetts 02139
Contributors vii
Chris R. Pelkie Section of Neurobiology and Behavior, Cornell University, Ithaca, New
York 14853
Anjanette Perry Department of Oceanography, University of Hawaii, Honolulu, Hawaii
96822
Karla A. Peterson Division of Geological and Planetary Sciences, California Institute of
Technology, Pasadena, California 91125
David E. Presti Department of Biology, University of Oregon, Eugene, Oregon 97403
Brenda Roder Division of Geological and Planetary Sciences, California Institute of Tech-
Bruce Rosenblum Department of Physics, University of California, Santa Cruz, California
95064
Gary R. Scott Lodestar Magnetics, Inc.,' Oakland. CaUfornia 94608

Durward D. Skiles Seismographic Station, University of California, Berkeley, California
94720
Kenneth M. Towe Department of Paleobiology, Smithsonian Institution, Washington,
D.C. 20560
William F. Towne Department of Biology, Princeton University, Princeton, New Jersey
08544
Benjamin Walcott Department of Anatomical Sciences, School of Medicine, State Uni-
versity of New York, Stony Brook, New York 11794
Michael M. Walker Southwest Fisheries Center La Jolla Laboratory, National Marine
Fisheries Service, National Oceanic and Atmospheric Administration, La Jolla, Cali-
fornia 92038
Peter J. Wasilewski NASA Goddard Space Flight Center, Greenbelt, Maryland 20771
Ellen D. Yorke Department of Physics, University of Maryland Baltimore County, Ca-
tonsville, Maryland 21228
John Zoeger Los Angeles County Museum of Natural History, Los Angeles, California
90007
Preface
The mystery of how migrating animals find their way over unfamiliar terrain has intrigued
people for centuries, and has been the focus of productive research in the biological sci-
ences for several decades. Whether or not the earth's magnetic field had anything to do
with their navigational abilities has sufaced and been dismissed several times, beginning
at least in the mid to late 1800s. This topic generally remained out of the mainstream of
scientific research for two reasons: (1) The apparent irreproducibility of many of the be-
havioral experiments which were supposed to demonstrate the existence of the magnetic
sense; and (2) Perceived theoretical difficulties which were encountered when biophysi-
cists tried to understand how such a sensory system might operate. However, during the
mid to late 1960s as the science of ethology (animal behavior) grew, it became clear from
studies on bees and birds that the geomagnetic field is used under a variety of conditions.
As more and more organisms were found to have similar abilities, the problem shifted
back to the question as to the basis of this perception. Of the various schemes for trans-
ducing the geomagnetic field to the nervous system which have been proposed, the hy-
pothesis of magnetite-based magnetoreception discussed at length in this volume has per-
haps the best potential for explaining a wide range of these effects, even though this link
is as yet clear only in the case of magnetotactic bacteria.
The question then arises as to what is the proper term for this new field of research.
According to the definition of Williamson and Kauffman (1980), biomagnetism is supposed
to include the study of the magnetic fields produced by an organism, whereas magneto-
biology should include the study of responses or the detection of such fields by organisms.
In a strict sense, the fallacy of this term splitting is clear in the case of the magnetotactic
bacteria, where the magnetite is responsible both for the strong local magnetic fields (up
to 0.4 tesla at the end of a magnetosome) and for their magnetotactic behavior. In general,
however, we prefer to use the term biomagnetism (or biogenic ferromagnetism) for studies
which focus on the presence of magnetite for the simple reason that it is far easier to detect
the presence of minute concentrations of ferromagnetic material in tissues than it is to
determine what, if anything, they are used for.
We organized a special symposium on magnetite biomineralization at the 1981 meeting
of the American Geophysical Union in San Francisco, principally because much of the
active research in this branch of biomagnetism was being conducted in laboratories nor-
mally devoted to the study of rock- and paleomagnetism. The goal of this session was to
bring together scientists working on various aspects of the magnetite biomineralization
problem, and it attracted a large number of participants from the physical, geological, and
biological sciences. During this meeting, the point was raised that literature of direct im-
portance to this branch of biomagnetism was scattered in journals ranging from bacteriology
to geophysics, and that there was no common source to which one could turn for in-depth
discussions or guidance. This work was organized in response to that need. It seems clear
from the diversity of papers in this volume that we have been reasonably successful in
covering a wide spectrum of the subject matter involved, and we hope that it will be of
use as a basic reference source in years to come. Included in this compilation are papers
ix
x Preface
which deal with most aspects of magnetite biomineralization, including in-depth discus-
sions focused on the biomineralization process (e.g., the Lowenstam and Kirschvink, Nes-
son and Lowenstam, Frankel et 01., and Mann chapters), the physical properties of mag-
netite which make it more than just another biogenic iron oxide mineral (Banerjee and
Moskowitz, Chapter 2), the information content of the geomagnetic field which might be
of use to magnetically sensitive organisms (Skiles, Chapter 3), and the expanding number
of organisms which are known to have both a geomagnetic sensitivity and the ability to
precipitate magnetite biochemically (Chapters 13 through 25). Although the ferromagnetic
hypothesis for magnetoreception appears to have been first suggested and experimentally
tested by Gustav Ising (1945), the theoretical development of this model is quite recent
and is extended further in this volume by Yorke (Chapter 10), Kirschvink and Walker
(Chapter 11), and discussed in relation to the possible use of an induction-based electrical
sensitivity by Rosenblum et 01. (Chapter 9). During the past few years, significant advances
have also been made in laboratory techniques which allow the detection and identification
of subnanogram quantities of ferromagnetic materials. These include the ultrasensitive
superconducting (SQUID) magnetometers discussed by Fuller et 01. (Chapter 4), the in-
expensive magnetic shielding techniques developed by Scott and Frohlich (Chapter 8), the
magnetite extraction procedure of Walker et 01. (Chapter 5), Mossbauer spectroscopy as
used by Frankel et 01. (Chapter 13), and the ever-improving techniques of electron mi-
croscopy (reviewed here in separate chapters by Towe, Mann, and Walcott). A final ques-
tion is whether or not magnetite crystals formed by the magnetotactic bacteria can be found
and traced in the fossil record; differing views are expressed here by Demitrack (Chapter
35) and by Chang and Kirschvink (Chapter 36). Such "magnetofossils," if they exist, would
provide a much needed explanation for the great magnetic stability of many deep-sea
sediments and might eventually place constraints on the oxygen concentration in the bot-
tom waters of ancient seas.
Also included in this volume is a chapter which deals with the question of whether
or not primates possess the ability to biochemically precipitate magnetite, and whether or
not humans have a magnetic sensitivity similar to that of other organisms. As explained
in the editorial introduction to that chapter, this latter topic has generated far more con-
troversy than any other section in this volume, and as such we have expanded it to include
discussions and replies from scientists on both sides of the question. We encourage the
reader to examine this material closely and formulate a balanced opinion based on as much
of the raw behavioral evidence as possible, but in particular we hope that this discussion
will attract and encourage additional work on this difficult problem.
Advances in fields other than ethology have contributed a great deal to the devel-
opment of the hypothesis of magnetite-based magnetoreception. One example involves the
experimental and theoretical determination of the ë á å Ö ä É J Ç ç ú ~ á å = size for magnetite (e.g.,
Neel, 1955; Evans and McElhinny, 1969; Butler and Banerjee, 1975). However, it is clear
that the single most important discovery was the identification of magnetite in chiton teeth
by Heinz Lowenstam (1962a). The continuing work on biomineralization by Lowenstam
and his students (Lowenstam, 1963, 1967, 1974, 1980, 1981; Towe and Lowenstam, 1967;
Kirschvink and Lowenstam, 1979; Lowenstam and Weiner, 1983) ushered in the modern
era of the systematic study of biomineralization processes. Although it was not recognized
at the time, the magnetic properties of the chiton radulae also constituted the first bio-
magnetic effect to be discovered. The commonly accepted "founding" of biomagnetism,
the measurement of the magnetic field produced by the heart, was not reported until the
next year (Baule and McFee, 1963).
The story of how Lowenstam was led to the discovery of magnetite in chiton teeth
has never been told in print before, yet is one of the most interesting examples of a superb
naturalist at work. After the Nazis twice prevented him from receiving his Ph.D. from the
University of Munich for being Jewish, he fled from Germany in 1937 and came to the
United States. Heinz worked for several years during World War II as a paleontologist with
Preface xi
the Illinois Geological Survey and afterwards at the University of Chicago. As travel funds
for fieldwork were scarce, he studied the complexes of Silurian reefs in the Chicago area
which were within easy reach of the public transportation system. As a result of these
efforts, Heinz deduced the pattern of ecological communities in the reef and surrounding
sediments, and based on the assemblage of fossils present could determine their distance
from the main wave-resistant structures (Lowenstam, 1950). Mundane as this may seem,
it was an essential step which made it possible to locate subsurface petroleum reservoirs
in fossil reefs based on exploratory drilling, a technique many oil companies were quick
to exploit.
After his move to the California Institute of Technology in 1952, Heinz expanded his
studies of fossil reef communities to include the Tertiary deposits found on tropical islands
in the Central Pacific, with a particular focus on Palau. Due to sea level changes during
the past few million years, many of these reefs are now exposed on land and offer easy
access for study. Heinz noticed that when these limestone masses were exposed close to
the intertidal zone, they often eroded into peculiar mushroom-shaped features. These "nip"
islands were capped with vegetation and often had waves splashing around the base (Low-
enstam, 1974), one of which is shown on the jacket of this book. Geomorphologists at that
time thought that the action of the surf was responsible for the erosion in the intertidal
zone, but this did not seem to fit the pattern which Lowenstam observed. He noticed that
the depth that the nips cut into the limestone increased as one moved away from the active
surf area toward the more quiet waters of the lagoons, an observation which clearly ruled
out wave action as their source. Upon closer examination, Heinz discovered that the lime-
stone within the actively eroding portion of the nips was heavily striated with subparallel
chisel marks produced by the grazing action of polyplacophoran molluscs (chitons). This
was strange, as limestone is much harder than the protein, chitin, which was thought to
be the main structural component in the radular teeth of all molluscs. Yet the teeth were
clearly eroding the rock, and there had to be a simple explanation. Visual examination of
the radula (the tongue plate) revealed that the pairs of major lateral teeth contained a hard,
somewhat shiny black substance which upon further analysis was identified as magnetite.
For many years, petrologists and biologists alike were skeptical that the material in the
teeth could have been a true biochemical precipitate, rather than inorganic sand grains
which were merely assimilated from the local environment. Geologically, magnetite was
only known to form at high temperatures and pressures in igneous and metamorphic rocks,
and it came as quite a surprise to find it as a biochemical precipitate in the teeth of a
mollusc.
Heinz Lowenstam was finally awarded his Ph.D. by the University of Munich in 1981,
in recognition of the excellence of his work and in partial reparation for past injustices.
References
Baule, G. M., and McFee, R., 1963, Detection of the magnetic field of the heart, Am. Heart J. 66:95-
96.
Butler, R. F., and Banerjee, S. K., 1975, Theoretical single-domain grain size range in magnetite and
titanomagnetite, J. Geophys. Res. 80:4049-4058.
Evans, M. E., and McElhinny, M. W. 1969, An investigation of the origin of stable remanence in
magnetite bearing igneous rocks, J. Geomagn. Geoelectr. 21:757-773.
Ising, G. 1945, Die physicalische Molichkeit eines tierischen Orientierungssines auf Basics der Er-
drotation, Ark. Mat. Astron. Fys. 32A(18):1-23.
Kirschvink, J. 1., and Lowenstam, H. A., 1979, Mineralization and magnetization of chiton teeth:
Paleomagnetic, sedimentologic, and biologic implications of organic magnetite, Earth Planet. Sci.
Lett. 44:193-204.
Lowenstam, H. A., 1950, Niagaran reefs in Illinois and their relation to oil accumulation, Oil Gas J.
48:49-77.
xii Preface
Lowenstam, H. A., 1962a, Magnetite in denticle capping in recent chitons (Polyplacophora), Bull.
Geol. Soc. Am. 73:435-438.
Lowenstam, H. A., 1962b, Goethite in radular teeth of recent marine gastropods, Science 137:279-
280.
Lowenstam, H. A., 1963, Biologic problems relating to the composition and diagenesis of sediments,
in: The Earth Sciences: Problems and Progress in Current Research (T. W. Donnelly, ed.), Uni-
versity of Chicago Press, Chicago, pp. 137-195.
Lowenstam, H. A., 1967, Lepidocrocite, an apatite mineral, and magnetite in teeth of chitons (Poly-
placophora), Science 156:1373-1375.
Lowenstam, H. A., 1968, Weddellite in a Marine Gastropod and in Antarctic Sediments, Science
162:1129-1130.
Lowenstam, H. A., 1972, Phosphate hard tissues of marine invertebrates: their nature and mechanical
function, and some fossil implications. Chem. Geol. 9:153-166.
Lowenstam, H. A., 1974, Impact of life on chemical and physical processes, in: The Sea, Volume 5
(E. D. Goldberg, ed.), Wiley, New York, pp. 715-796.
Lowenstam, H. A., 1980, Bioinorganic constituents of hard parts, in: Biogeochemistry of Amino Acids
(P. E. Hare, T. D. Hoering, and K. King, Jr., eds.), Wiley, New York, pp. 3-16.
Lowenstam, H. A., 1981, Minerals formed by organisms, Science 211:1126-1131.
Lowenstam, H. A., 1984, Biomineralization processes and products and the evolution of biominer-
alization, 27th lnt. Geol. Congr. Sect. Paleontol. Theme C.02.1.5.
Lowenstam, H. A., and Weiner, S., 1983, Mineralization by organisms, and the evolution of biomi-
neralization, in: Biomineralization and Biological Metal Accumulation (P. Westbroek and E. W.
de Jong, eds.), Reidel, Dordrecht, pp. 191-203.
Neel, L., 1955, Some theoretical aspects of rock magnetism, Philos. Mag. Suppl. Adv. Phys. 4:191-
243.
Towe, K. M., and Lowenstam, H. A., 1967, Ulstrastructure and development of iron mineralization
in the radular teeth of Cryptochiton stelleri (Mollusca), J. Ulstrastruct. Res. 17:1-13.
Williamson, S. J., and Kaufman, 1., 1980, Biomagnetism, J. Magn. & Magn. Mater. 22:129-201.
Contents
I. Introduction and Background

Chapter 1 • Iron Biomineralization: A Geobiological Perspective
Heinz A. Lowenstam and Joseph L. Kirschvink

1. Introduction.......................................... 3
2. Biological Aspects of Iron Mineralization ................... 6
3. Biological Functions of Iron Biomineralization ............... 10
4. Geological Aspects of Biogenic Fe Oxides and Sulfides . . . . . . . . . 10
References ........................................... 13
Chapter 2 • Ferrimagnetic Properties of Magnetite
Subir K. Banerjee and Bruce M. Moskowitz

1. Introduction.......................................... 17
2. Basic Data ........................................... 18
3. Bulk Properties ....................................... 21
4. Magnetic Domain States ................................ 23
5. Remanent Magnetizations ............................... 31
6. Magnetic Granulometry ................................. 36
References ........................................... 38
Chapter 3 • The Geomagnetic Field: Its Nature, History, and

Biological Relevance
Durward D. Skiles
1. Introduction.......................................... 43
2. The Main Geomagnetic Field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3. The Field of External Origin ............................. 87
References ........................................... 98
II. Experimental Techniques and Instrumentation

Chapter 4 • An Introduction to the Use of SQUID Magnetometers in
Biomagnetism
M. Fuller, W. S. Goree, and W. 1. Goodman

1. Introduction.......................................... 103
2. Operating Principles of SQUIDs. . . . . . . . . . . . . . . . . . . . . . . . . . . 104
xiii
xiv Contents
3. Cryogenics........................................... 121
4. Instrument Configurations ............................... 129
5. Applications of SQUID Magnetometers in Biomagnetism ....... 136
6. Conclusions.......................................... 148
References ........................................... 149
Chapter 5 • Detection, Extraction, and Characterization of Biogenic

Magnetite
Michael M. Walker, Joseph 1. Kirschvink, Anjanette Perry,

and Andrew E. Dizon
1. Introduction.......................................... 155
2. Magnetometry Studies ...... . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
3. Extraction and Characterization of Biogenic Magnetite ......... 160
4. Discussion........................................... 163
5. Summary............................................ 164
References ........................................... 165
Chapter 6 • Studying Mineral Particulates of Biogenic Origin by

Transmission Electron Microscopy and Electron
Diffraction: Some Guidelines and Suggestions
Kenneth M. Towe
1. Introduction.......................................... 167
2. Sample Preparation for Electron Microscopy . . . . . . . . . . . . . . . . . 168
3. Studying the Sample in the Microscope .................... 173
4. Analysis of Electron Diffraction Powder Patterns. . . . . . . . . . . . . . 178
5. Conclusions.......................................... 179
Selected References .................................... 180
Chapter 7 • The Cellular Localization of Particulate Iron
Benjamin Walcott
1. Introduction.......................................... 183
2. Anatomical Techniques ................................. 184
3. An Example: The Bumblebee. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
4. Conclusions.......................................... 193
References ........................................... 195
Chapter 8 • Large-Volume, Magnetically Shielded Room: A New

Design and Material
Gary R. Scott and Cliff Frohlich

1. Introduction.......................................... 197
2. General Principles of Electric and Magnetic Shielding ......... 199
Contents xv
3. Practical Techniques for Building Magnetically Shielded Rooms 208

4. Three Specific Examples ................................ 214
5. Summary............................................ 219
References ........................................... 220
III. Magnetoreception: Theoretical Considerations

Chapter 9 • Limits to Induction-Based Magnetoreception
Bruce Rosenblum, Roger 1. Jungerman, and Laurent Longfellow
1. Introduction.......................................... 223
2. Noise and General Considerations . . . . . . . . . . . . . . . . . . . . . . . . . 224
3. The Induction Magnetoreception Organ. . . . . . . . . . . . . . . . . . . . . 225
4. Conclusion........................................... 231
5. Addendum: A Comment on Navigation . . . . . . . . . . . . . . . . . . . . . 231
References ........................................... 231
Chapter 10 • Energetics and Sensitivity Considerations of

Ferromagnetic
Magnetoreceptors
Ellen D. Yorke
1. Introduction.......................................... 233
2. Energy Considerations .................................. 234
3. Response Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
4. Sensitivity to Field Changes ............................. 238
5. Other Types of Receptors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
6. Tests of the Hypothesis ................................. 239
References ........................................... 241
Chapter 11 • Particle-Size Considerations for Magnetite-Based

Magnetoreceptors
Joseph 1. Kirschvink and Michael M. Walker

1. Introduction.......................................... 243
2. The Thermally Driven Variance Model of Magnetic Intensity
Reception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
3. Discussion........................................... 251
4. Summary............................................ 253
References ........................................... 253
Chapter 12 • Are Animal Maps Magnetic?
James 1. Gould
1. Introduction.......................................... 257
2. The Compass Sense ....... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
xvi Contents
3. The Map Sense ....................................... 259

4. Problems with Magnetic Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
5. Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
References ........................................... 266
IV. Magnetoreception and Magnetic Minerals in Living

Organisms
Chapter 13 • Mossbauer Spectroscopy of Iron Biomineralization

Products in Magnetotactic Bacteria
Richard B. Frankel, Georgia C. Papaefthymiou, and

Richard P. Blakemore
1. Introduction to Mossbauer Spectroscopy . . . . . . . . . . . . . . . . . . . . 269

2. Application of Mossbauer Spectroscopy to Magnetotactic Bacteria 279
References ........................................... 285
Chapter 14 • Magnetotactic Microorganisms Found in Muds from Rio

de Janeiro: A General View
Henrique G. P. Lins de Barros and Darci Motta S. Esquivel
1. Introduction.......................................... 289
2. The Geomagnetic Field ................................. 290
3. Results.............................................. 291
4. Conclusions.......................................... 305
References ........................................... 308
Chapter 15 • Structure, Morphology, and Crystal Growth of Bacterial

Magnetite
Stephen Mann
1. Introduction.......................................... 311
2. Instrumentation: High-Resolution Transmission Electron
Microscopy .......................................... 312
3. Materials and Methods ................................. 312
4. Results.............................................. 315
5. Discussion: Bioprecipitation of Bacterial Magnetite . . . . . . . . . . . . 323
6. Conclusions.......................................... 330
References ........................................... 331
Contents xvii
Chapter 16 • Biomineralization Processes of the Radula Teeth of

Chitons
Michael H. Nesson and Heinz A. Lowenstam

1. Introduction.......................................... 333
3. Anatomy and Operation of the Radula Apparatus ....... . . . . . . 335
4. Anatomy of the Radula Sac .............................. 337
5. Blood Chemistry ...................................... 341
6. The Ultrastructure of the Mineralization Zone . . . . . . . . . . . . . . . . 342
7. Concluding Remarks ................................... 361
References ........................................... 361
Chapter 17 • Magnetic Remanence and Response to Magnetic Fields

in Crustacea
Ruth E. Buskirk and William P. O'Brien, Jr.

1. Introduction.......................................... 365
2. Experimental Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
3. Discussion........................................... 377
4. Summary............................................ 380
References ........................................... 381
Chapter 18 • Magnetic Field Sensitivity in Honeybees
William F. Towne and James 1. Gould

1. Introduction.......................................... 385
2. Magnetic Fields Cause Misdirection in the Waggle Dance. . . . . . . 386
3. Magnetically Oriented Horizontal Dances ................... 392
4. Magnetic Orientation of Comb-Building. . . . . . . . . . . . . . . . . . . . . 393
5. Magnetic Fields and Orientation in Time ................... 395
6. The Magnetic Receptor System ........................... 398
7. Summary and Conclusions .............................. 403
References ........................................... 404
Chapter 19 • Magnetic Butterflies: A Case Study of the Monarch

(Lepidoptera, Danaidae)
Bruce J. MacFadden and Douglas S. Jones

1. Introduction.......................................... 407
2. Natural History of the Monarch Butterfly ................... 408
4. Induced Magnetization ................................. 410
5. Ontogeny of Magnetic Mineralization ...................... 411
6. Intraspecific and Interspecific Variation .................... 412
7. Attempts to Characterize the Magnetic Mineralogy ............ 413
xviii Contents
8. Summary and Conclusions .............................. 414

References ........................................... 415
Chapter 20 • Magnetoreception and Biomineralization of Magnetite:

Fish
Michael M. Walker, Joseph 1. Kirschvink, and Andrew E. Dizon

1. Introduction.......................................... 417
2. Magnetic Sensitivity in Yellowfin Tuna .................... 419
3. Detection of Magnetic Material in Fish ..................... 422
4. Characterization of the Magnetic Material ................... 426
5. Identification and Analysis of the Magnetic Material. . . . . . . . . . . 429
6. Discussion........................................... 431
References ............................................ 434
Chapter 21 • Magnetoreception and Biomineralization of Magnetite in

Amphibians and Reptiles
Anjanette Perry, Gordon B. Bauer, and Andrew E. Dizon

1. Introduction.......................................... 439
2. Amphibians.......................................... 440
3. Reptiles ............................................. 443
4. Conclusion........................................... 452
References ........................................... 452
Chapter 22 • Avian Navigation, Geomagnetic Field Sensitivity, and

Biogenic Magnetite
David E. Presti
1. The Sensory Basis of Bird Navigation ...................... 455
2. Orientation Experiments with Homing Pigeons ............... 459
3. Orientation Experiments with Migratory Birds ............... 464
4. Effects of Small Magnetic Field Changes on Navigation: The
Possibility of a Geomagnetic Map ......................... 469
5. Laboratory Attempts to Measure Avian Magnetic Field Sensitivity
472
6. Magnetite in Birds and Possible Mechanisms of Magnetic Field
Sensitivity ........................................... 474
References ........................................... 477
Chapter 23 • Magnetic Remanence in Bats
Edward R. Buchler and Peter J. Wasilewski

1. Introduction.......................................... 483
2. Methods............................................. 483
Contents xix
3. Results 484
4. Discussion........................................... 485
References ........................................... 486
Chapter 24 • Magnetoreception and Biomineralization of Magnetite in

Cetaceans
Gordon B. Bauer, Michael Fuller, AnjaneUe Perry, J. Robert Dunn,

and John Zoeger
1. Introduction.......................................... 489
2. Behavioral Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 490
3. Anatomical Studies .................................... 495
4. Conclusion........................................... 503
References ........................................... 505
Chapter 25 • Magnetoreception and the Search for Magnetic Material

in Rodents
Janice G. Mather
1. Introduction.......................................... 509
2. Influence of Magnetic Fields on Physiology. . . . . . . . . . . . . . . . . . 510
3. Magnetoreception ..................................... 510
4. The Search for the Magnetoreceptor ....................... 522
5. Summary............................................ 531
References ........................................... 532
V. Human Magnetoreception: An Editorial Introduction

Chapter 26 • Magnetoreception by Man and Other Primates
R. Robin Baker
1. Introduction.......................................... 537
2. Physiological Responses to Changes in the Ambient Magnetic
Field ............................................... 538
3. Magnetoreception ..................................... 539
4. Magnetoreceptors?.................................... 550
5. Discussion........................................... 556
6. Summary............................................ 559
References ........................................... 559
Chapter 27 • Statistical and Methodological Critique of Baker's

Chapter
Tom Dayton
1. Statistics in General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563
2. "Chair" Experiments Results Section. . . . . . . . . . . . . . . . . . . . . . . 565
xx Contents
3. Princeton Data Do Not Support Baker ...................... 565

4. Magnets vs. Controls for Baker's Experiments ................ 567
5. Magnets vs. Controls for K-6 ............................. 567
6. Physiology of Magnetoreceptors . . . . . . . . . . . . . . . . . . . . . . . . . . . 568
7. Summary............................................ 568
References ........................................... 568
Chapter 28 • Human Navigation: Attempts to Replicate Baker's

Displacement Experiment ........................ 569
Kenneth P. Able and William F. Gergits
Chapter 29 • Human Homing Orientation: Critique and Alternative

Hypotheses
Kraig Adler and Chris R. Pelkie

1. Introduction.......................................... 573
2. Bus Tests Conducted at Ithaca, New York . . . . . . . . . . . . . . . . . . . 574
3. Oriented Distributions from "Random" Data . . . . . . . . . . . . . . . . . 584
References ........................................... 587
Notes ............................................... 587
Reply to Baker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 591
Chapter 30 • Absence of Human Homing Ability as Measured by

Displacement Experiments ....................... 595
James L. Gould
Chapter 31 • A Study of the Homeward Orientation of Visually

Handicapped Humans
Timothy K. Judge
1. Introduction.......................................... 601
2. Methods............................................. 601
3. Results.............................................. 601
4. Discussion........................................... 602
References ........................................... 603
Chapter 32 • An Attempt to Replicate the Spinning Chair Experiment 605
Joseph 1. Kirschvink, Karla A. Peterson, Michael Chwe, Paul

Filmer, and Brenda Roder
Contents xxi
Chapter 33 • A Cautionary Note on Magnetoreception in Dowsers. . . . 609
Joseph L. Kirschvink
Chapter 34 • Human Navigation: A Summary of American Data and

Interpretation
R. Robin Baker
1. The American Data 611
2. The American Criticisms ............................... . 614
3. Concluding Remarks .................................. . 621
References ...................... .................... . 621
VI. Biogenic Magnetite in the Fossil Record

Chapter 35 • A Search for Bacterial Magnetite in the Sediments of Eel
Marsh, Woods Hole, Massachusetts
Anne Demitrack
1. Introduction.......................................... 625
2. Bacterial Magnetite .................................... 626
3. Methods............................................. 627
4. Results.............................................. 630
5. Discussion........................................... 639
Appendix 1: Eel Marsh NRM and Saturation Magnetization Data
643
Appendix 2: Description of Computer Procedure Used to Make
Stability Field Diagram 8a ................... . 644
References ........................................ .. . 644
Chapter 36 • Possible Biogenic Magnetite Fossils from the Late

Miocene Potamida Clays of Crete
Shih-Bin R. Chang and Joseph Kirschvink

1. Introduction.......................................... 647
2. Samples............................................. 648
3. Laboratory Extraction of Magnetite ........................ 650
4. Magnetic Studies ...................................... 653
5. Size and Shape Distribution of Magnetite ................... 654
6. Origin of Magnetite .................................... 655
7. Conclusion and Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 666
References ........................................... 667
Index ................................................... 671

I
Introduction and Background
During the past 10 years, three disciplines within the earth sciences (biomineralization,
rock magnetism, and geomagnetism) have contributed enormously to our understanding
of how organisms respond to geomagnetic stimuli. Because a basic understanding of these
subjects is a prerequisite for fruitful research in the field of biogenic ferrimagnetism, it is
essential for this volume to begin with introductory chapters on each, written for both
biologists and geologists.
The field of biomineralization is the study of the biochemical processes through which
organisms produce mineral hard parts. Of prime importance are the biogenic iron minerals,
of which magnetite is the only one known to be ferrimagnetic. Now that magnetite bio-
mineralization is known to occur in several separate phyla, is suspected in many more (see
Part IV), and may well be responsible for geomagnetic sensitivity in all animals, it is im-
portant to understand the evolutionary history of this mineral and the biochemistry of its
formation. As discussed by Lowenstam and Kirschvink in Part I, the most interesting ques-
tions in this area have not yet been identified, much less answered.
The field of rock magnetism is concerned with how the magnetic properties of minerals
vary with composition, grain size and shape. It has developed in conjunction with the
study of the history of the earth's magnetic field (paleomagnetism) and provides the basis
for our understanding of how rocks preserve a record of ancient magnetic fields. Due to
its abundance in many rock types and its role in their magnetization, magnetite has been
the subject of more experimental and theoretical studies in this regard than has any other
mineral; these properties are reviewed here in the chapter by Banerjee and Moskowitz.
These ferrimagnetic properties are what distinguish magnetite from the other biogenic iron
oxides, and they are of utmost importance in any analysis of magnetite-based magneto-
reception. For example, it has been observed through extensive electron microscopy that
the magnetotactic bacteria only make magnetite crystals with sizes in the range from 0.05
to 0.1 11m; this is also the range found from rock magnetism studies to yield particles with
single magnetic domains. Similar studies on magnetite crystals in chiton teeth (which are
used for their hardness rather than their magnetism) reveal particles with sizes both above
and below these bounds. These observations imply that the narrow size range of the bac-
terial crystals is maintained by natural selection for magnetotaxis, and that the bacteria
solved the question of the single-domain size range for magnetite long before rock magne-
tists did.
The field of geomagnetism is the study of the geomagnetic field including its source,
history, and expression within the biosphere. We now know that the geomagnetic field
contains a wealth of useful information beyond the simple ability to provide a north-south
compass for magnetically sensitive organisms. Diurnal variations and other magnetic pul-
sations contain information on latitude, seasonality, and time, and local magnetic an-
omalies and features like the marine magnetic lineations are potential reference cues for
migration routes. The important questions, of course, concern which features of the mag-
netosphere organisms actually respond to, and in what fashion the sensitivity is achieved,
if at all.
1
2 Part I Introduction
To begin with, however, one must have a firm understanding of what the geomagnetic
field is, how it is measured and reported, and which of its many features are reasonable
cues for organisms to recognize. The chapter by Skiles is the first attempt to review this
material for a biological audience, and as such is long overdue and of great importance
for all workers in this field.
Chapter 1
Iron Biomineralization
A Geobiological Perspective
HEINZ A. LOWENSTAM and JOSEPH L. KIRSCHVINK
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Biological Aspects of Iron Mineralization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
3. Biological Functions of Iron Biomineralization. . . . . . . . . . . . . . . . . . . . . . . . . . . 10
4. Geological Aspects of Biogenic Fe Oxides and Sulfides . . . . . . . . . . . . . . . . . . . . . 10
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1. Introduction
During the last two decades, the number of minerals identified as biological precipitates
increased fourfold, from the 10 known in 1963 (Lowenstam, 1963) to over 40 (Lowenstam
and Weiner, 1983) recognized today. And yet, from the continuing rate of discovery, it is
quite clear that many more probably remain to be recognized. Among the minerals which
have been discovered and for which the crystal structure has been clarified with modern
techniques are 10 iron minerals, including oxides, sulfates, sulfides, and phosphates. Table
I'lists these minerals and their distribution among the phyla, biomineralization modes,
and known functions. Several of these minerals are known to have a wide distribution
among prokaryotes and eukaryotes, and two of them (ferrihydrite and magnetite) are likely
to be the fourth and fifth most widely utilized biomineralization products in present-day
organisms. Ferrihydrite (2.5Fe203 . 4.5H 20) forms the core component (micelle) of the iron
storage protein ferritin in numerous animal classes, and related ferric oxide minerals occur
in ferritin like structures in other eukaryotes as well as in some prokaryotes (e.g., phy-
toferritin and bacterioferritin). Less than 25 years ago, magnetite had been identified in
chitons (class Polyplacophora; Lowenstam, 1962a), but its biogenic origin was regarded
with great skepticism. As shown by the papers in this volume, however, the phyletic
distribution of magnetite may soon compete in abundance with ferrihydrite for fourth
place.
The wide phyletic distribution of iron mainerals shown in Table I is not surprising
in view of the fundamental role that iron plays in biological processes. In the predominantly
oxidizing hydrosphere, the transport of dissolved iron, in the ferrous state, is very slow.
Hence, organisms have evolved a variety of mechanisms to satisfy this requirement for
iron acquisition, and internal storage sites have been developed to provide iron when it
is routinely needed and to provide high concentrations for emergencies such as blood loss.
In particular, bacteria have developed elegant mechanisms for acquiring and storing ex-
HEINZ A. LOWENSTAM and JOSEPH 1. KIRSCHVINK • Division of Geological and Planetary Sci-
ences, California Institute of Technology, Pasadena, California 91125.
3
ú =
TABLE I. Biochemically Precipitated Iron Minerals, Phyletic Occurrences, Mineralization Types, and Locations
Mineralization Mineral location

Minerals Occurrence process (es) (typical examples) References
Oxides
Magnetite Bacteria Matrix mediated Intracellular magneto somes Blakemore (1975)
(Fe 30 4) Protozoa (?) Unknown Unknown Lins de Barros et 01. (1981)
Molluscs Matrix mediated Extracellular in chiton Lowenstam (1962a)
teeth
Arthropods Unknown Honeybee abdomen Gould et 01. (1978)
Chordates Matrix mediated (?) Tuna dermethmoid bone, Walker et 01. (Chapter 20,
pigeon head this volume), Walcott et
01. (1979)
Ferrihydrite Bacteria Matrix mediated Bacterioferritin micelle Stiefel and Watt (1979),
(5Fe203'9H20) Yariv et 01. (1981)
Fungi Matrix mediated Ferritin micelle David and Easterbrook
(1971), Peat and Banbury
(1968)
Plantae Matrix mediated Phytoferritin micelle Hyde et 01. (1963)
Animalia Matrix mediated Ferritin micelle Ford et 01. (1984)
Lepidocrocite Porifera Induced Sponge granules Towe and Rutzler (1968) n
::r
(B-FeOOH) Mollusca Matrix mediated Chiton teeth Lowenstam (1967) ú =
Goethite Mollusca Matrix mediated Limpet teeth Lowenstam (1962b) "S!.

CD
...,
(a-FeOOH) .....
Amorphous Bacteria Unknown Variable or unknown Ehrlich (1981) ...,
0
-
ferric oxides Protozoa Unknown Foraminiferal test cement Towe (1967) ::l
I:Il
Annelida Matrix mediated (?) Polychaete tube cement Lowenstam (1972) o·
Mollusca Unknown Gastropod gizzard plates Lowenstam (1968) S
Amorphous Bacteria (?) Induced On surface of holothurian Lowenstam (unpublished) 5'
CI)
...,
Ilmenite skin e..
Sulfides N'
ú =
Pyrite Bacteria Induced Forms extracellularly e.g., Hallberg (1978) o·
(FeS 2) ::l
Hyrotroilite Bacteria Induced Forms extracellularly e.g., Hallberg (1978)

(FeS'nH 2O)
Sulfates
Jarosite Bacteria Induced Forms extracellularly, Lazaroff et a1. (1982)
[M-Fe(S04Jz(OH)6J only known from lab expo
Phosphates
Amorphous Annelida Matrix mediated Sternaspis div. sp. Lowenstam (1972)
hydroferric sternal shield
phosphate Mollusca Matrix mediated Chiton teeth Lowenstam (1972)
Enchinodermata Matrix mediated Holothurian dermal Lowenstam and Rossman
granules (1975)
C,1I
6 Chapter 1
tracellular iron, and they are known to produce numerous small iron-specific chelating
compounds known as siderophores. Some species produce these compounds in large quan-
tities and excrete them into the surrounding environment. Scavenging iron by this method
from the external medium, the bacteria then incorporate the molecules along with the
chelated iron (e.g., Nealson, 1984).
2. Biological Aspects of Iron Mineralization

Two distinct mineralization processes have been recognized (Lowenstam, 1981). One
of these has been termed "biologically induced" mineralization. In this process, minerals
are formed by interaction of biologically produced metabolite end products with cations
in the external environment. The other process has been termed "organic matrix-mediated"
mineralization. In this process, nucleation and subsequent development of minerals takes
place in contact with a preformed organic matrix. The two processes constitute end mem-
bers of a spectrum in which the organism exercises increasing control over the kind of
minerals to be produced as well as mineral growth. In "biologically induced" minerali ..
zation processes, which denote the spectral end member of least biological control, the
mineral precipitates resemble those formed by inorganic processes in crystal habits. Crystal
aggregates are devoid of unique shapes and usually show random orientation of variously
sized crystallites. There is a tendency for less vigorous control of these processes on the
mineral phases of the precipitates formed. At the other end of the spectrum in "matrix-
mediated" mineralization processes, as a rule, the minerals have unique crystal habits and
a narrow size distribution. Polycrystalline bodies adopt genetically controlled morphol-
ogies and show well-ordered crystal axis orientation as well as crystal fabrics (Lowenstam,
1980, 1981; Weiner and Traub, 1980). Several of the minerals formed today by "matrix-
mediated" processes in the hydrosphere are not known to be precipitated there by inorganic
processes. Examples are the minerals magnetite, fluorite, gypsum, and celestite (Lowen-
stam and Weiner, 1983). Minerals formed by these processes tend to have an overprint of
disequilibrium chemical signatures with reference to the environment (Lowenstam, 1981).
At present, this is best documented by the Sr and Mg contents and the oxygen and carbon
isotopic composition of carbonate minerals (summarized in Milliman, 1974; Dodd and
Stanton, 1981).
Biochemical and X-ray diffraction studies of the organic matrix, based on mollusc
shells, have shown that the matrix consists of a core component composed of a silk-fibroin-
like protein and commonly also of chitin (Weiner and Traub, 1980, 1981). The core com-
ponent is veneered on both sides by acidic surface layers. A proposed biomineralization
model envisions that the functions of the acidic surface layers include mineral nucleation,
crystallographic axis orientation, as well as crystal growth inhibition. The core component
is considered to provide a substrate for the acidic proteins and performs mechanical func-
tions. A review by Weiner et a1. (1983) elaborates on the matrix structure and function of
molluscs and other eukaryotes.
"Biologically induced" mineralization processes occur widely among prokaryotes, and
are found both intra- and extracellularly. Among the eukaryotes, the mineralization prod-
ucts are known to occur in some algal representatives of the protoctista, in many plant
phyla, and apparently also in many animal phyla. The minerals form intercellularly in
some algae and intracellularly in plants and animals (Arnott, 1973; Borowitzka et a1., 1974;
Lowenstam, 1981). Assignment of biominerals to "biologically induced" mineralization
processes is in most instances based on the crystal habits, shapes, and fabrics of aggregates
as seen in published illustrations.
Iron Biomineralization 7
The "organic matrix"-mediated mineralization processes are widespread among the

eukaryotes. The mineralization products are as a rule located at extracellular sites, where
they form commonly mineralized hard parts such as, e.g., exo- and endoskeletons, but they
may consist solely of numerous isolated mineral grains or skeletal pinpoint mineralizations
of crystal aggregates (Lowenstam and Margulis, 1980).
Recent reports on bacterially formed minerals with unique crystal habits and in two
cases with well-ordered crystal fabrics have been interpreted to indicate that prokaryotes
are also involved in "matrix-mediated" mineralization processes (Lowenstam and Weiner,
1983). The observations in support of this conclusion were as follows. Magnetotactic bac-
teria discovered by Blakemore (1975) form intracellular, membrane-bound magnetite crys-
tals with hexagonal prism (Towe and Moench, 1981; Matsuda et aI., 1983) or "teardrop"
shapes (Blakemore et aI., 1981). The crystals have a narrow size distribution (To we and
Moench, 1981). In addition, an unnamed bacterial species of Leptotrix has been shown to
form a hexagonally shaped grillwork of an as yet unidentified ferric oxide mineral inside
the sheath which covers the cell filaments (Caldwell and Caldwell, 1980). The unique
crystal fabric calls for organic framework control, but it is not known how the enclosing
sheath serves as the organic matrix. Other examples are provided by two cyanobacterial
species of GeitIeria. The filament surfaces are thickly covered by an orderly stacked fabric
of calcite crystals which form a square-shaped grillwork (Friedmann, 1955, 1979). Treat-
ment of the calcite coat with dilute hydrochloric acid leaves a delicate "pectinous" sheath
as a residue (Friedmann, 1955; Lowenstam and Weiner, 1983). SEM micrographs of critical
point dried and EDTA decalcified samples of freshly collected GeitIeria specimens have
recently been obtained by one of us (H.A.L.). The micrographs show that the spaces bor-
dered by the square-shaped frames of the calcite crystals are the sites of organic matrices
which are interconnected to form a continuous framework within the mineralized coat of
the bacteria (Lowenstam, 1984).
This observation extends documentation of "organic matrix-mediated" mineralization
among prokaryotes from the magnetotactic bacteria to the cyanobacteria and further shows
that these processes may take place extracellularly as well as intracellularly. The miner-
alized structures are formed by Geitleria on the filament surfaces with extensive matrix
development and grossly resemble skeletal structures of eukaryotes but seem to differ from
them in their crystal arrangement.
"Biologically induced" and "matrix-mediated" processes have also been noted to
occur in one organism at different anatomical sites and even at the same mineralization
site (Lowenstam and Weiner, 1983). An example of this latter process is found in the upper
beak of Nautilus, where regular matrix-mediated layers of calcite crystals are separated by
crystals of biologically induced brushite and weddelite (Lowenstam and Weiner, 1983).
Minerals formed by organisms often assume unique crystal habits and hence can be
recognized as biological precipitates on purely morphological grounds. A clear example
of this can be seen in the fluorite (CaF 2) produced in the gizzard plates of the gastropod
Eoscaphander (Fig. 1A) and in the statoliths of many opossum shrimp species (mysids)
(Lowenstam and McConnell, 1968). These crystals are acicular (Fig. 1BJ, and are totally
unlike the cubic and octahedral crystals produced through inorganic processes in nature.
Mineralogical changes during ontogeny have been noted in a number of animal groups.
The most spectacular change has been found to occur in the holothurian species Molpadia
intermedia s.l. (Lowen starn and Rossman, 1975). The mesodermal hard parts in early post-
larval individuals consist of calcitic spicules, as they do in other species of this class. In
the course of subsequent growth, most spicules outside the oral and caudal regions undergo
resorption and thereafter become replaced by granules, mineralogically composed of amor-
phous hydrous ferric phosphate, and amorphous silica in the form of opal. Figure 2A shows
8 Chapter 1
Figure 1. (A) Fluorite crystals from gizzard plates of the gastropod Eoscaphander. (B) Scale bar = 1
fLm, Fluorite crystals in a statolith of the opposum shrimp Praunus flexuosus . Scale bar = 5 fLm.
Figure 2. (A) Calcitic spicule from a young individual of Molpadia intermedia s.l. Scale bar = 100
IJ.m. (B) Amorphous hydrous ferric phosphate plus opal granule from a mature individual of M. in-
termedia s.l. Scale bar = 10 u,m.
10 Chapter 1
racquet-shaped calcitic spicules from a young individual and 2B, a granule composed of
the mineralogical replacement products formed in an older individual.
3. Biological Functions of Iron Biomineralization

Iron plays a central role in a wide variety of metabolic processes ranging, for example,
from the transport of oxygen in hemoglobin of vertebrate blood to a central role in the
cytochrome system. Similarly, iron minerals serve an equally wide range of functions,
including transport and storage, waste disposal, hardening of teeth, and navigation. Iron
transport and storage is perhaps the most universal function of the mineral ferrihydrite,
which as noted earlier forms the micelle of the iron storage protein, ferritin. Even organisms
which do not have hemoglobin as the oxygen-carrier require iron storage. For example,
although nautiloids use the copper-based oxygen transport agent, hemocyanin, for O2 trans-
port, their buccal muscles require iron for their myoglobin. Iron minerals also act as hard-
ening agents on the surface of the radula teeth of chitons and limpets. In tropical and
subtropical waters, both of these groups use these mineralized teeth to facilitate the ex-
traction of endolithic bacteria and algae. Both groups produce cliff recession on carbonate
islands where inorganic processes alone could not produce such effects (Lowenstam, 1974).
During this process, the teeth along the radulae are continually worn away and replaced,
and this loss of iron requires a balancing flux of ferritin to the tooth sites. Indeed, chiton
blood has been shown to have high ferritin concentrations (Nesson, 1969; Webb and Macey,
1983; Nesson and Lowenstam, this volume).
In rare cases, iron minerals are known to form in response to metabolic overloads
during the treatment of some human diseases. In particular, the symptoms of iron defi-
ciency in hemachromatosis and thallassemia are both treated by injection of colloidal iron
solutions which are readily absorbed. Although this temporarily alleviates anemia, the
excess iron is eventually dumped in various tissues (including the liver and spleen) in
irregularly shaped bodies of ferrihydrite called hemosiderin. Massive deposits of this ma-
terial can now be detected quickly and noninvasively by measuring the induced magnetic
moments with SQUID magnetometers (Ferrill, 1983).
Finally, by far the most surprising role of any iron mineral in biology is that of nav-
igation, a topic which is explored at length in this volume. Magnetite has been clearly
implicated in the cellular alignment of magnetotactic bacteria (Blakemore, 1975; Frankel
et a1., 1979) and could have a role in the homing instinct of chitons (Lowenstam, 1962a;
Mook, 1983). The development of extraction techniques which allow the purification and
high-resolution characterization of ferromagnetic minerals from tissues (Walker et a1., this
volume) has also confirmed that many of the magnetic particles detected with the SQUID
magnetometry are indeed biochemical precipitates.
4. Geological Aspects of Biogenic Fe Oxides and Sulfides
At present, very little is known about the evolutionary history of iron biominerali-
zation. The presence of a diverse assemblage of biogenic iron minerals in extant bacteria
(Table I) suggests that these abilities probably evolved prior to the development and ra-
diation of the eukaryotes [before about 1.4 b.y.]. The use of 34Sj32S ratios as a
tracer of "biologically induced" mineralization by sulfate-reducing bacteria in Precambrian
sedimentary pyrites has demonstrated that geochemical approaches can contribute ma-
terially to the elucidation of the history of biomineralization. It appears that "biologically
induced" -type mineralization was the first biomineralization mode to evolve in the Pre-
cambrian. This supposition is based on the sulfur isotopes from sedimentary pyrites which
show the beginning of dissimilatory fractionation by sulfate-reducing bacteria in samples
aged 2-2.4 b.y. (Cameron, 1982; Skyring and Donnelly, 1982). The use of ferrous iron in
solution as an electron donor for photosynthesis (rather than H2S or H2 0) as suggested by
Nealson (1984) and Y. Cohen (personal communication) may well have been one ofthe earliest
biochemical processes to induce iron oxide mineralization, and some of the oldest Banded
Iron Formations (BIFs) could have been formed as a result of this process. This suggests
that iron proteins were already in existence prior to the evolutionary pressures generated
by increasing amounts of free O2 in the atmosphere, an observation which is consistent
with the extensive use of iron proteins in the cytochrome system for regulating O2 in all
aerobic organisms. Therefore, iron-based proteins certainly must have evolved prior to the
massive change in oxidation state of the oceans which was signaled about 2 b.y. by the
widespread deposition of BIFs. Recent work on extant magnetotactic bacteria suggests that
the magnetite-forming step in the synthesis pathway is dependent upon the presence of
free O2 (Bazylinski and Blakemore, 1983; Blakemore et 01., 1985). This result, if ap-
plicable to all magnetite-precipitating bacteria, suggests that their fossil record should not
extend appreciably back beyond the 2 b.y. mark. However, the geomagnetic field probably
formed early in earth history (Stevenson, 1983) and there would have been a selective
advantage for magnetotaxis even in the absence of large O2 gradients.
At present, only "bacterially induced" mineralization is in evidence up to near the
end of the Proterozoic. Beginning with the Vendian-early Cambrian transition, eukaryotes
began to exercise more rigorous control over mineral-forming processes and "organic ma-
trix-mediated" mineralization evolved. As shown by skeletal remains, animals were the
first eukaryotes to develop "matrix-mediated" mineralization. Once initiated, this miner-
alization process began to spread rapidly to a wide range of animals in early Cambrian
time and continued to proliferate among the eukaryotes in the course of the Phanerozoic.
In extant prokaryotes and eukaryotes, there are examples of mineralization processes
which cover the spectrum between "biologically induced" and "matrix-mediated" mi-
neralization (Lowenstam and Weiner, 1983). The fossil record appears to present a very
different perspective of the history of biomineralization. It appears that "matrix-mediated"
mineralization was from the beginning limited to the eukaryotes and evolved considerably
later in geological time than "biologically induced" mineralization which continued to be
utilized only by prokaryotes. A number of seemingly unrelated aspects of present-day
"matrix-mediated" mineralization products have raised serious doubt about the validity
of the inference from the fossil records that mineralized hard parts, which began to appear
first in the geological record at the Vendian-Cambrian transition, actually mark the in-
ception of "matrix-mediated" processes. It is also unclear whether or not records at the
Vendian-Cambrian transition actually mark the inception of "matrix-mediated" processes.
It is also unclear whether or not animal representatives of the eukaryotes were the initiators
of these processes (Lowenstam and Weiner, 1983; Lowenstam, 1984).
It has been difficult to conceive that from its inception, "matrix-mediated" mineral-
ization should have been perfected as a process to yield in animals uniformly and suffi-
ciently massive skeletal mineralization to preserve their products morphologically intact.
Based on the observations that a number of early Cambrian skeletal records have only weak
surficial mineralization, it had been postulated that pinpoint mineralization may have
preceded massive mineralized hard parts (Lowenstam and Margulis, 1980). The indications
among representatives of the late Vendian soft-bodied Ediacara fauna of partial skeletal
mineralization, and the development of spicules in pennatulid-like forms (Glaessner and
Daily, 1959; Glaessner, 1976) seem to support the notion of an initial phase of patchy
skeletal mineralization.
Similar trends are in evidence with respect to increase in complexity in organization
of mineral products that are formed at homologous sites in several extant animal groups,
12 Chapter 1
e.g., the molluscs and fish (Lowenstam, 1980; Lowenstam and Weiner, 1983). Reflecting
evolutionary trends, they seem not only to lend support to the postulate that pinpoint
mineralization preceded total skeletal mineralization in animals, but seem to provide,
further, a clearer concept of the consecutive stages that trace the pathways of these pro-
cesses as shown in the following. Stage 1 is characterized by single crystals or single
colloidal bodies in great profusion at the depositional site. At Stage 2, the constituents
consist of a profusion of minute separate crystal aggregates instead of single crystals. At
Stage 3, most of the crystals are incorporated into a larger single body, but a few minute
crystal aggregates remain isolated. Stage 4 marks the aggregation of all crystals into a single
structure without differentiation in crystal fabric and mineralogy. Stage 5 resembles the
previous stage by the aggregation of all crystals into a single body but differs from Stage
4 in the subdivision of the structures into discrete micro architectural units with different
crystal fabrics which do not have to be mineralogically distinct. In some groups of orga-
nisms, the trend terminates at Stage 4 and in others, one or another intermediate stage
may be missing. As is usual for "matrix-mediated" mineralization products, the crystals
formed at Stage 1 commonly have specific crystal habits, may be unique in mineralogy,
and commonly have an overprint of disequilibrium chemical signature with reference to
the environment. From the beginning of localized crystal aggregation into multiple bodies
at Stage 2 to the incorporation of all crystals into single structures, as formed at Stages 4
and 5, one finds well-ordered crystal fabrics and well-defined, genetically controlled mor-
phologies of the polycrystalline bodies. Hence, individual crystals produced at Stage 1
and minute crystal aggregates formed at Stages 2 and 3 should be distinguishable from the
crystals and crystal aggregates of "biologically induced" and inorganic origin.
The sum of the considerations outlined above presents a very strong argument that at
the time the first completely mineralized animal skeletons made their apperance in the
fossil record,"organic matrix-mediated" processes had not just come into existence, but
had already evolved in evolutionary terms to near completion, that is, definitely to Stage
4 and possibly Stage 5. Therefore, "matrix-mediated" mineralization should have come
into existence earlier in Proterozoic time, and the seemingly unrecorded evolutionary in-
itial stages in development by these processes should be traceable in the sedimentary rocks
by means of single crystals and minute mineral aggregates, characterized by the criteria
outlined above. A further implication (of the possibility that the early evolution of "matrix-
mediated" processes took place in Proterozoic time) is that animal representatives may
not have been the initiator of these processes, contrary to the current view which has been
based on body fossils.
Independently derived support for a Precambrian origin of "matrix-mediated" mi-
neralization processes may be seen in an interphylum survey of the biochemistry of organic
matrices from the mineralized hard parts of eukaryotes (Weiner et 01., 1983). The data
come from a wide range of animal phyla and one group of protoctista. This survey reveals
common interphylum biochemical properties of the eukaryotic organic matrices, an ob-
servation which is consistent with the notion that "matrix-mediated" mineralization
"know-how" was inherited from some common Precambrian ancestral stock. The wide-
spread utilization of matrix-mediated mineralization for constructing solid exoskeletons
that came into existence at the end of the Precambrian would then again imply the cul-
mination of a long evolutionary history and not the inception of this mode of biominer-
alization. The alternative that each phylum "invented" independently the matrix strategy
would have to be qualified by invoking an explanation for the observed common inter-
phylum biochemical properties of organic matrices.
As noted above, it has recently been shown that extant prokaryotes also produce min-
erals with the aid of organic matrices. One of them, the magnetotactic bacteria, have mag-
netite crystals with unique crystal habits and are characterized by a narrow size distri-
bution. Magnetite crystals of this type have recently been located in deposits of Miocene
age (Kirschvink, 1982; Kirschvink and Chang, 1984; Chang and Kirschvink, this volume),
but there are as yet no data on the geological range of these magnetite-forming bacteria
nor on the two other mineral-forming bacterial groups. it is significant, however, that the
extant magnetite- and ferric oxide-producing bacteria may still reflect the reducing con-
dition and changeover from reducing to oxidizing conditions in the biosphere during the
early Precambrian. It is of further interest that the third group of mineral-forming bacteria
belongs to the cyanobacteria which extend in the sedimentary record back to 3.5 b.y. B.P.
(Awramik et a1., 1983). Nothing is as yet known about the biochemistry of the organic
matrices from prokaryotic mineralization products and hence we do not know as yet
whether they are analogous to eukaryotic matrices. Clarification of this aspect and also a
search for evidence of the existence of Precambrian prokaryotes involved in "matrix-me-
diated" mineralization is needed. Given this information, it should be possible to ascertain
whether "matrix-mediated" mineralization processes evolved independently in prokar-
yotes and eukaryotes, or whether the eukaryotes inherited the "know-how" of these pro-
cesses from a prokaryote progenitor in the Precambrian.
It is of interest to note that an inheritance of the "know-how" of "matrix-mediated"
mineralization by eukaryotes from prokaryotes can readily be accommodated in the ev-
olutionary scheme visualized by Margulis (1981), for in this scheme the eukaryotes evolved
directly from the prokaryotes, whereas in the scheme of Woese (1981), the origin of the
"know-how" is as yet not as clearly discernible.
ACKNOWLEDGMENT. This is Contribution No. 4150 from the Division of Geological and Pla-
netary Sciences, California Institute of Technology.
References
Arnott, H. J., 1973, Plant calcification, in: Biological Mineralization (1. Zipkin, ed.), Wiley, New York,
pp. 609-627.
Awramik, S. M., Schopf, J. A., and Walter, M. R., 1983, Filamentous fossil bacteria from the Archean
of Western Australia, Precambrian Res. 20:357-374.
Bazylinski, D. A., and Blakemore, R. P., 1983, Denitrification and assimilatory nitrate reduction in
Aquaspirillum magnetotacticum, Appl. Environ. Microbiol. 46:1118-1124.
Blakemore, R. P., 1975, Magnetotactic bacteria, Science 190:377-379.
Blakemore, R. P., Frankel, R. B., and Kalmijn, A. ]., 1981, South-seeking magnetotactic bacteria in the
southern hemisphere, Nature 286:384-385.
Blakemore, R. P., Short, K. A., Bazylinski, D. A., Rosenblatt, C., and Frankel, R. B., 1985, Microaerobic
conditions are required for magnetite formation with Aquaspirillum magnetotacticum, Geomi-
crobiol. J. 4:53-71.
Borowitzka, M. A., Larkum, A. W. D., and Nockolds, C. E., 1974, A scanning electron microscope
study of the structure and organization of the calcium carbonate deposits of algae, Phycologia
13:195-203.
Caldwell, D. E., and Caldwell, S. J., 1980, Fine structure of in situ microbial iron deposits, Geomi-
crobiol. J. 2:39-53.
Cameron, E. M., 1982, Sulphate and sulfite reduction in early Precambrian oceans, Nature 296:145-
148.
David, C. N., and Easterbrook, K., 1971, Ferritin in the fungus Phycomyces, J. Cell BioI. 48:15-28.
Dodd, J. R., and Stanton, R. J., 1981, Paleoecology: Concepts and Applications, Wiley, New York, pp.
138-142.
Ehrlich, H. L., 1981, Geomicrobiology, Dekker, New York, pp. 164-201.
Ferrill, D. E., Assessment of iron in human tissue: The magnetic biopsy, in: Biomagnetism: An In-
terdisc1ipinary Approach, (S. J. Williamson, G. L. Romani, L. Kaufman, and Ivo Modena eds.),
Plenum Press, New York, pp. 483-500.
14 Chapter 1
Ford, G. C., Harrison, P. M., Rice, D. W., Smith, J. M. A., Treffry, A., White, J. L., and Yariv, J., 1984,
Ferritin: Design and formation of an iron storage molecule, Philos. Trans. R. Soc. London Ser. B
304:551-565.
Frankel, R B., Blakemore, R P., and Wolfe, R S., 1979, Magnetite in freshwater magnetotactic bacteria,
Science 203:1355-1357.
Friedmann, I., 1955, Geitleria calcarea n.gen. et n.sp.: A new atmophylic lime-encrusting blue-green
algae, Bot. Not. Lund 108:439-445.
Friedmann, 1.,1979, The genus Geitleria (Cyanophyceae or Cyanobacteria): Distribution of G. calcarea
and G. floridana n.sp., Pl. Syst. Evol. 131:169-178.
Glaessner, M. F., 1976, Early Phanerozoic annelid worms and their geological and biological signif-
icance, J. Geol. Soc. London 132:259-275.
Glaessner, M. F., and Daily, B., 1959, The geology and late Precambrian fauna of the Ediacara fossil
reserve, Rec. South Aust. Mus. (Adelaide) 13:369-401.
Gould, J. L., Kirschvink, J. 1., and Deffeyes, K. S., 1978, Bees have magnetic remanence, Science
202:1026-1028.
Hallberg, R 0., 1978, Metal-organic interaction at the redoxcline, in: Environmental Biogeochemistry
and Geomicrobiology, Volume 3 (W. E. Krumbein, ed.), Ann Arbor Science Publishers, Ann Arbor,
Mich., pp. 947-953.
Hyde, B. B., Hodge, A. J., Kahn, A., and Birnstiel, M. L., 1963, Studies on phytoferritin. 1. Identification
and localization, J. Ultrastruct. Res. 9:248-258.
Kirschvink, J. 1., 1982, Paleomagnetic evidence for fossil biogenic magnetite in western Crete, Earth
Planet. Sci. Lett. 59:388-392.
Kirschvink, J. L., and Chang, S.-B. R, 1984, Ultrafine-grained magnetite in deep-sea sediments: Possible
bacterial magnetofossils, Geology 12:559-562.
Lazaroff, N., Sigal, W., and Wasserman, A., 1982, Iron oxidation and precipitation of ferric hydrox-
ysulfates by resting Thiobacillus ferrooxidans cells, Appl. Environ. Microbiol. 43:924-938.
Lins de Barros, H. G. P., Esquivel, D. M. S., Danon, J., and Oliveira, J. P. H., 1981, Magnetotactic algae,
Acad. Bras. Cienc. Notas Fis. CBPF-NF-048/81.
Lowenstam, H. A., 1962a, Magnetite in denticle capping in recent chitons (Polyplacophora), Bull.
Geol. Soc. Am. 73:435-438.
Lowenstam, H. A., 1962b, Goethite in radular teeth of recent marine gastropods, Science 137:279-
280.
Lowenstam, H. A., 1963, Biologic problems relating to the composition and diagenesis of sediments,
in: The Earth Sciences: Problems and Progress in Current Research (T. W. Donnelly, ed.), Uni-
versity of Chicago Press, Chicago, pp. 137-195.
Lowenstam, H. A., 1968, Weddellite in a marine gastropod and in antarctic sediments, Science
162:1129-1130.
Lowenstam, H. A., 1972, Phosphate hard tissues of marine invertebrates: Their nature and mechanical
function, and some fossil implications, Chern. Geol. 9:153-166.
Lowenstam, H. A., 1980, Bioinorganic constituents of hard parts, in: Biogeochemistry of Amino Acids
(P. E. Hare, T. D. Hoering, and K. King, Jr., eds.), Wiley, New York, pp. 3-16.
Lowenstam, H. A., 1984, Biomineralization processes and products and the evolution of biominer-
alization, 27th lnt. Geol. Congr. Sect. Paleontol. Theme C.02.1.5.
Lowenstam, H. A., and McConnell, D., 1968, Biologic precipitation of fluorite, Science 162:1496-1498.
Lowenstam, H. A., and Margulis, L., 1980, Evolutionary prerequisites for early Phanerozoic calcareous
skeletons, BioSystems 12:27-41.
Lowenstam, H. A., and Rossman, G. R, 1975, Amorphous, hydrous, ferric phosphatic dermal granules
in Molpadia (Holothuroidea): Physical and chemical characterization, and ecologic implication
of the bioinorganic fraction, Chern. Geol. 15:15-51.
Lowenstam, H. A., and Weiner, S., 1983, Mineralization by organisms, and the evolution of biomi-
Margulis, 1., 1981, Symbiosis in Cell Evolution, Freeman, San Francisco.

Matsuda, T., Endo, J., Osakabe, N., Tonomura, A., and Arii, T., 1983, Morphology and structure of
biogenic magnetite particles, Nature 302:411-412.
Milliman, J. D., 1974, Marine Carbonates, Springer-Verlag, Berlin, pp. 52-147.
Mook, D., 1983, Homing the West Indian chiton Ancanthopleura granulata Gmelin, 1791, Veliger
26:101-105.
Nealson, K. H., 1984, The microbial iron cycle, in: Microbial Geochemistry. (W. E. Krumbein, ed.),
Blackwell, Oxford, pp. 159-190.
Nesson, M. H., 1969, Studies on radulae tooth mineralization in the Polyplacophora, Ph.D. thesis,
California Institute of Technology, Pasadena.
Peat, A., and Banbury, G. H., 1968, Occurrence of ferritin-like particles in a fungus, Planta 79:268-
270.
Skyring, G. W., and Donnelly, T. H., 1982, Precambrian sulfur isotopes and a possible role for sulfite
in the evolution of biological sulfate reduction, Precambrian Res. 17:41-61.
Stevenson, D. J., 1983, The nature of the earth prior to the oldest known rock record: The hadean
earth, in: Earth's Earliest Biosphere G. W. Schopf, ed.), Princeton University Press, Princeton, N.
J., pp. 32-40.
Stiefel, E. I., and Watt, G. D., 1979, Azotobacter cytochrome 6557.5 is a bacterioferritin, Nature 279:81-
83.
Thode, H. G., Kleerekoper, H., and McElcheran, 0., 1951, Isotopic fractionation in the bacterial re-
duction of sulphate, Research 4:581-582.
Towe, K. M., 1967, Wall structure and cementation in Haplaphragmoides canariensis, Contributions
from the Cushman Foundation for Foraminiferal Research, Volume XV III, Part 4.
Towe, K. M., and Moench, T. T., 1981, Electron-optical characterization of bacterial magnetite, Earth
Towe, K. M., and Rutzler, K., 1968, Lepidocrocite iron mineralization in keratose sponge granules,
Science 162:268-269.
Walcott, C., Gould, J. L., and Kirschvink, J. 1., 1979, Pigeons have magnets, Science 205:1027-1029.
Webb, J., and Macey, D. C., 1983, Plasma ferritin in Polyplacophora and its possible role in the biom-
ineralization of iron, in: Biomineralization and Biological Metal Accumulation, (P. Westbroek
and E. W. de Jong, eds.), Reidel Dordrecht, pp. 423-428.
Weiner, S., and Traub, W. 1980, X-ray diffraction study of the insoluble organic matrix of mollusk
shells, FEBS Lett. 111:312-316.
Weiner, S., and Traub, W., 1981, Organic-matrix-mineral relationships in mollusk shell nacreous
layers, in: Structural Aspects of Recognition and Assembly in Biological Macromolecules (M.
Balaban, J. 1. Sussman, W. Traub, and A. Yonath, eds.), Balaban ISS, Rehovot, pp. 467-482.
Weiner. S.. Traub. W., and Lowenstam, H. A., 1983, Organic matrix in calcified exoskeletons, in:
Biomineralization and Biological Metal Accumulation (P. Westbroek and E. W. de Jong, eds.j,
Reidel, Dordrecht, pp. 205-224.
Woese, C. R, 1981, Archaebacteria, Sci. Am. 244:98-122.
Yariv, J., Kalb, A. J., Sperling, R, Bauminger, E. R. Cohen, S. G., and Offer, S., 1981, The composition
and the structure of bacteria ferritin of Escherichia coli, Biochem. J. 197:171.
Chapter 2
Ferrimagnetic Properties of Magnetite
SUBIR K. BANERJEE and BRUCE M. MOSKOWITZ
1. Introduction. . . . . . . . 17
2. Basic Data. . . . . . . . . 18
2.1. Spinal Structure ... 18
2.2. Magnetite . . . . . . . . . . . . . . . . . . ..... 18
3. Bulk Properties. . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . 21
3.1. Saturation Magnetism . . . . . . . . . . . . . . . . . 21
3.2. Anisotropy and Magnetostriction . . . . . . . . . . . . . . . . . . . . . . . 22
3.3. Temperature Dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4. Pressure Dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4. Magnetic Domain States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.1. Domain Theory . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2. PSD Particles . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3. Superparamagnetism .. 26
4.4. Critical Sizes .... . 27
4.5. Coercivity . . . . . . . 28
4.6. Interactions . . . . . . 30
5. Remanent Magnetizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.1. Isothermal Remanent Magnetization . . . . . . . . . . . . . . . . 31
5.2. Anhysteretic Remanent Magnetization. . . . . . . . . . . . 32
5.3. Chemical Remanent Magnetization. . . . . . . 33
5.4. Thermoremanent Magnetization. . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6. Magnetic Granulometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1. Introduction
Magnetite (Fe304, ferrous-ferric oxide) is ubiquitous as the source of the magnetism of

most biological magnetic systems. Although a cation-deficient form of it, maghemite (-y-
Fe203), and impurity-substituted magnetite (titanomagnetite) have on occasion been iden-
tified in biomagnetic systems, magnetite continues to be the primary magnetic source in
biology. It is of interest, therefore, to inquire into the origin of its magnetism, or more
properly, the ferrimagnetism of magnetite. In this chapter we deal with the ferrimagnetism
of magnetite single crystals first. We then occupy ourselves with the application of mag-
netic domain theory to the particle-size-dependent properties of magnetite and the various
kinds of intrinsic remanent magnetization contributed to by magnetite. The only type of
SUBIR K. BANERJEE • Department of Geology and Geophysics, University of Minnesota, Minne-

apolis, Minnesota 55455. Contribution No. 1059 of the School of Earth Sciences, Department of Geology
and Geophysics. University of Minnesota, Minneapolis. BRUCE M. MOSKOWITZ • Department
of Geological and Geophysical Sciences, Princeton University, Princeton, New Jersey 08544.
17
18 Chapter 2
natural remanent magnetization which we have not discussed here is depositional re-
manent magnetization (DRM), as it is still rare to find examples where biogenic magnetite
has been convincingly shown to be responsible for the DRM in a sediment. Future research
may prove otherwise. Finally, we deal with some practical magnetic techniques for de-
termining the magnetic domain state, hence the effective particle size, of magnetite.
2. Basic Data
2.1. Spinel Structure
A host of oxide minerals, including magnetite, crystallize with the spinel structure.
This structure can accept as cations at least 30 different elements with valence states from
+ 1 to + 6. Consequently, impurity cations will often drastically affect the physical and
magnetic properties of spinel oxides.
The spinel unit cell is face-centered cubic with space group Fd3m (in oxide spinels
containing 32 oxygens). The cations occupy interstitial sites within the oxygen framework
and these sites are of two types: tetrahedral or A sites which are located at the center of
a tetrahedron whose corners are occupied by 0 ions, and octahedral or B sites which are
located at the center of an octahedron whose corners are also occupied by 0 ions. The 0
anions, in contrast, occupy sites given in terms of the u parameter. Variations in u cor-
respond to displacements of the 0 ions along the [111] axis of the spinel structure and for
cubic close-packing u = 0.25. Details of the spinel unit cell are illustrated in Fig. 1.
In a unit cell of a spinel oxide, there are 64 possible A sites and 32 possible B sites.
However, only 8 of the A sites and 16 of the B sites are actually occupied by cations. The
general formula for a spinel is AB 20 4, where the A and B cations have different valence
states. In a normal spinel, the 8 A cations occupy A sites and the 16 B cations occupy B
sites (e.g., A[B 2]04)' In other spinels, an inverse structure is present, where 8 of the B
cations occupy the A sites and 8 A and 8 B cations occupy the B sites (e.g., B[AB]04),
while still other spinels are intermediate in their cation distribution between the normal
and the inverse structures.
2.2. Magnetite
Magnetite is an inverse spinel of structural formula Fe3 + [Fe 2 +Fe 3 + ]04 as determined
by neutron diffraction (Shull et 01., 1951). The unit-cell parameter(s) based on X-ray dif-
fraction data on both natural and synthetic samples range from 8.393 A to 8.3963 A. This
variation probably reflects varying amounts of impurity cations, or cation vacancies, or
both incorporated within the magnetite structure (Lindsley, 1976). Neutron diffraction
studies by Hamilton (1958) give u = 0.2548 ± 0.0002 (at 23°C), indicating almost perfect
cubic close-packing of O. A summary of some of the more important physical properties
of magnetite is given in Table 1.
The Fe ions on both A and B sites in magnetite give rise to a magnetic substructure
in which the net magnetic moment of the A sublattice is directed anti parallel to the net
magnetic moment of the B sublattice, as viewed along the [111] axis (see Fig. 1). This
anti parallel ordering occurs because of exchange interactions due to the overlapping of
the 3d electron orbital wave functions of the Fe ions on the A and B sublattice with the
2p electron orbital wave function of the 0 ion. The magnitudes of these exchange inter-
actions are dependent on the distance between Fe and 0 ions and their bond angles (Fe-
O-Fe); and these, in turn, are determined by the positions of the Fe3 + and Fe 2 + ions in
Ferrimagnetic Properties of Magnetite 19
#,1"'" - - - - - - - - - .... T - - - - - - - - -ú=

6}
.... '" I A .... .... I #'#'
' I
........
ú J J J J
I
J J J J J ....... --------;;;-f I " I
A .... ,\ I j ........ I
.. ' I I .... .... I I
D f= D ú f D f= I
ú
I #'
, --- ---- J ú J J J J --- 8 I I
I , ú| f bl ç?=
I /,
t'----.I
I ú KKKK=
I I
I
I
I I
I - --J ....
.... ....
I
Figure 1. The spinel unit cell showing the location of the ions in two of the eight octants. The other
six octants contain one or the other of these configurations; octants that share a face have different
ion configurations while octants that share an edge have the same ion configuration. A and B ions
are metallic ions in fourfold (A site) and six-fold (B site) coordination with oxygen (large circles). The
o ions are situated about halfway on the body diagonals; however, the 0 ions can be slightly displaced
along the body diagonals, resulting in a distortion of the A and B sites. The measure of this displacement
is the f.L parameter. The arrows denote the orientations of the magnetic moments for each of the A-
and B-site ions and illustrate the anti parallel alignment of the magnetic moments between A and B
sites (see text for further details). After Gorter (1955).
the spinel structure. In magnetite, the relationships between magnitude and type of in-
teraction are; AB 126 ú =BB90 > AA79 ú =AB 154 > BB 125 where the subscripts denote Fe-O-
Fe bond angle. Both AB 126 and BB90 interactions are negative (anti parallel alignment); the
AA79 interaction is weakly positive (parallel alignment); and AB 154 and BB 125 interactions
are negligible. Magnetite is therefore ferrimagnetic . The phenomenon of exchange inter-
actions which occur in magnetite (as well as in all other ferrites) is termed superexchange,
is extensively reviewed by Goodenough (1963). Superexchange refers to the major role
that the 0 ions play in mediating the exchange interactions; direct exchange refers to
exchange interaction between Fe ions only. However, in magnetite, direct exchange in-
teractions are insignificant because of the large distances between Fe ions.
Magnetite undergoes a cubic to orthorhombic crystallographic transition at 118°K. The
c axis of the new orthorhombic phase is parallel to , and 0.03% smaller than the cube edge
of the old cubic phase (Bickford, 1953). This transition is also accompanied by an electronic
ordering of the Fe 3 + and Fe 2 + ions on the B sites. Above 118°K, Fe 2 + and Fe 3 + ions are
20 Chapter 2
TABLE I. Physical and Magnetic Properties of Magnetite
Property" Value
Crystal type
T> 118°K Inverse cubic spinel
T < 118°K Orthorhombic
Lattice constant 8.393-8.3963 A
u parameter 0.2548
X-ray density 5.238 g/cm3
Electrical resistivity (at 3000 K) 7 x 10- 3 ohm/cm
Curie temperature 574°Cb
Crystallographic transition 119°K
Saturation magnetization (at 3000 K) 471 emu/cm3 (90 emu/g)C
Magnetic moment/molecule (at OaK) 4.1 /LB
Magnetocrystalline anisotropy constants Kl = -1.35 X 105 ergs/cm 3
K2 = - 0.44 X 105 ergs/cm3
Magnetostrictive constants
Polycrystalline As = 35 X 10- 6
Single crystal A111 = 72.6 X 10- 6
A100 = -19.5 X 10- 6
A110 = 55.1 X 10- 6
a Values taken from references in text and from Landolt-Boernstein. Series III/4b. pp. 65-66.
b To value from Pauthenet (1950). although values ranging from 575 to 585°C have been re-
ported in the literature.
o Again. reported values range from - 82 to - 93 emulg [see Bate (1980) for a discussion of
the variations in saturation magnetization and To values].
randomly distributed over the B sites, making electron hopping between B-site Fe ions
energetically favorable. The low electrical resistivity of magnetite (7 x 10- 3 ohm-cm at
3000 K) is attributed to this electron hopping (Verwey and Haayman, 1941). Below 118°K,
the Fe 2 + and Fe3+ ions become ordered along the (110) axes in the (100) planes. This ionic
ordering produces a large increase in electrical resistivity (100 x) (e.g., Parker, 1975).
The crystallographic transition is associated with a magnetic isotropic point where
the magnetocrystalline anisotropy constant (K1 ) changes sign from - to + (Le., Kl = 0
at 118°K). Electron hopping above 118°K results in a time-averaged valence state of FeZ . 5 +
on the B site and effectively removes the large anisotropy contribution due to Fe2+ . Below
118°K, electron hopping is considerably reduced and the Fe z + ions produce a large ani-
sotropy contribution, which are responsible for a large positive Kl (Stacey and Banerjee,
1974), the onset of which occurs between 118 and 130°K. This phenomenon will be dis-
cussed further in later sections.
Magnetites found naturally inevitably contain impurity cations, the most frequent ones
being Ti, AI, Mg, and Mn. Magnetite forms a solid solution series with ulvospinel
(Fe2 + [Fe2+Ti4 + ]04 ) whose members are called titanomagnetites (Fe3_Six04)' Titanom-
agnetites are by far the most common magnetic minerals found in basalts (e.g., Lindsley,
1976). The substitution of Ti for Fe occurs by replacing 2Fe3+ for Ti4 + + Fe z + (to maintain
charge balance) and with Ti ions occupying B sites (Ishikawa et a1., 1964). The effect of
Ti substitution in magnetite produces systematic variations in magnetic and physical prop-
erties: saturation magnetization Us) and Curie temperature (Te) decrease; coercivity (He),
magnetocrystalline anisotropy constant (K 1 ), cell parameter (a), and electrical resistivity
increase. The magnetic isotropic temperature (Tm) decreases for x < 0.3, but increases for
x > 0.3 (e.g., Stacey and Banerjee, 1974). Biogenic magnetite in some bacteria may be
slightly titaniferous (Towe and Moench, 1981).
i.6
0.8
0.6
Js
J;- 0.4
0.2
Figure 2. Variation of saturation magnetization of mag-
netite (relative to that at OOK) with absolute temperature a 0.2 0.4 0.6 0.8
(normalized to the Curie point). fo = 98.5 emu/g, Te = T
847°K. From Pauthenet (1950). Tc
Another magnetic mineral closely associated with magnetite is maghemite ('y-Fe 203)'
Maghemite is a cation-deficient spinel formed usually by the low-temperature oxidation
of magnetite, by the dehydration of lepidocrocite ('y-FeOOH), or by the direct precipitation
from solution (Taylor and Schwertmann, 1974). On oxidation to 'Y-Fe203, the cell parameter
slightly decreases to 8.33 A and cation vacancies are formed by the oxidation of Fe2+ to
Fe3+ (e.g., Lindsley, 1976). Maghemite is thermodynamically metastable with respect to
hematite (a-Fe203) and inverts to hematite by heating to temperatures above 350°C. It is
difficult to distinguish magnetite from maghemite using X-ray analysis. M6ssbauer and
magnetic measurements, however, can distinguish between these two oxide phases. Magh-
emite has also been suggested as occurring in magnetotactic bacteria (Frankel et aJ., 1979;
Towe and Moench, 1981).
3. Bulk Properties
3.1. Saturation Magnetism
A basic property of a ferrimagnetic mineral such as magnetite is its saturation mag-

netization, also known as spontaneous magnetization. Spontaneous magnetization is the
net magnetization which would be observed if a sensor could be placed inside a uniformly
magnetized single-domain particle. In practice, for example. a sample of known mass or
volume is placed inside a vibrating sample magnetometer within the é ç ä ú = pieces of an
electromagnet and the induced magnetization is measured as a function of increasing mag-
netic field. Lack of a significant increase in magnetization with increasing magnetic field
denotes saturation, i.e., maximum possible alignment of the magnetic domains at a given
temperature is called saturation magnetization corresponding to the given temperature.
Saturation magnetization is an intrinsic property of a ferrimagnet and is independent of
particle size. Like other thermodynamic parameters, however, saturation magnetization
(as well as magneto crystalline and magnetostrictive anisotropy constants) is a function of
pressure and temperature and it is necessary to define the ambient pressure and temper-
ature when quoting a saturation magnetization value.
Saturation magnetization per unit mass of magnetite is ú V M =emu/g and its dependence
on temperature down to 4.2°K has been measured by Pauthenet (1950). It is seen (Fig. 2)
that as predicted by Neel for ferrimagnetic magnetite, the largest value of saturation mag-
netization (Is) occurs at O°K. As the sample is heated, the observed magnetization decreases
according to the Brillouin function (the quantum mechanical description of J- T behavior;
e.g., Cullity, 1972) and Js goes to zero at the Curie point (Te) when the thermal energy (kTe,
22 Chapter 2
where k is Boltzmann's constant) equals the ferrimagnetic coupling energy or the super-
exchange energy. Te for magnetic is another intrinsic parameter for the material and is
equal to 847°K (574°C), although Te values ranging from 575 to 585°C have been reported
in the literature. Curie points can be measured simply by observing the temperature de-
pendence of Js. To ensure saturation, however, it is necessary to employ a moderately large
magnetic field (-10,000 Oe) during the experiment and this interferes with a clear-cut
determination of Te due to the presence of paramagnetic contributions which do not, unlike
ferrimagnetism, vanish at Te. Moskowitz (1981) has compared the efficiency of three dif-
ferent approaches to Curie point determination.
It should be noted that although magnetite follows generally the Neel theory of fer-
rimagnetism, its Js value when expressed in Bohr magnetons per molecule at 4.2°K is 4.1
rather than Neel's predicted value of 4.0. The explanation may lie in a slightly higher
contribution of magnetization by the octahedrally located Fe2 + ions.
3.2. Anisotropy and Magnetostriction
If saturation magnetization is called a first-order magnetic property, then magneto-

crystalline anisotropy is a second-order, although intrinsic, magnetic property of a ferri-
magnet. As the name implies, this anisotropy constant refers to the energy necessary to
deflect the magnetic spin moments in a single crystal from crystallographically "easy" (or
low-energy) to "hard" (high-energy) directions. The easy and hard directions arise from
the nature of the interactions between the crystal lattice and the magnetic spins (and is
called spin-orbit coupling) and these directions are different in different magnetic min-
erals. In magnetite, the easy, hard, and intermediate directions are the cubic [111], [100],
and [110] directions, respectively. Thus, the energy of magnetization (E K ) in an arbitrary
direction for a single crystal is given by
where Kl and K2 are the first- and second-order anisotropy constants and <Xl> <X2, <X3 are the
direction cosines. At 300oK, Kl and K2 for magnetite are - 1.35 x 105 ergs/cm3 and - 0.44
x 105 ergs/cm3 , respectively. For a multi domain sphere of magnetite in a demagnetized
condition, there will be six easy directions for the domains to align, corresponding to the
three [111] axes. Application of a large magnetic field, say 3000 Oe, will result in saturation
when the [111] axis nearest to the positive field direction is the preferred direction for all
the domains. In a saturated multi domain sphere or a single-domain sphere, the magneti-
zation can occupy one of two antiparallel directions along the preferred [111] axis.
Before discussing the temperature dependence of Kl (the first-order magneto crystalline
anisotropy constant), it is necessary to mention the magnetostriction constants, A11l and
AlOO, which arise from strain dependence of the anisotropy constants. Upon magnetization,
a previously demagnetized magnetic crystal experiences a strain which can be measured
as a function of the applied field along the principal crystallographic axes. Thus, A111 and
AlOO of magnetite are the maximum strains experienced along the [111] and [100] axes. If
one assumes an arbitrary direction (<Xl' <X2, <X3) for the applied field and if ú ä D = ú O D = ú P = are
the direction cosines for a strain gage, the magnetostriction (A) measured by the strain gage
will be given by
Thus, if a single crystal of magnetite is held under nonuniform stress (0'), the magnetic
energy will have two intrinsic components: EK , due to magnetocrystalline anisotropy con-
stants K1 and Kz, and ECTA, due to magnetostriction contains All1 and AlOo, In addition to
externally applied nonuniform stress, a magnetite crystal can experience internal non-
uniform stress due to crystal defects, dislocations, etc. For magnetite, A111 and AlOo are
+ 72.6 x 10- 6 and -19.5 x 10- 6 , respectively. These constants are to be multiplied by
the ambient strain to obtain magnetostrictive energy. For polycrystalline materials, the
effective magneto stricti on constant As is given by
As = (2AlOo + 3A111)/5
3.3. Temperature Dependence
Both (Kz, Kz) and (A111, AlOo) are functions of temperature and especially in magnetite
this results in interesting variations of magnetization direction with temperature. Both
types of constants owe their origin to the strong spin-orbit coupling of Fez + ions and the
coupling becomes much stronger as the ambient temperature is lowered. At about 118°K,
the anisotropy constant K1 changes its sign from negative to positive as a result of which
the effective K1 becomes zero at temperatures close to 118°K, the isotropic point (Tm) for
magnetite. If magneto crystalline anisotropy energy vanishes and then reappears with a
different sign, domain walls in a multi domain grain become diffuse and then renucleate,
thus losing all or a large part of the original remanence carried by such a multi domain
particle. This forms the basis of a convenient method for distinguishing true multi domain
(MD) particles of magnetite from single-domain (SD) and pseudo-single-domain (PSD) par-
ticles (see, e.g., Levi and Merrill, 1976). A similar cooling experiment in which low-field
magnetic suscepitbility is measured leads to an identification of MD magnetite particles
(Senanayake and McElhinny, 1981; Murthy and Patzold, 1982) by virtue of a peak in sus-
ceptibility around the isotropic point.
The same microscopic mechanism that leads to the changeover in sign of K1 also leads
to a sharp increase in the magnetostriction constants (A111, AlOo) upon cooling through the
isotropic point of magnetite. Useful applications of this phenomenon are not yet known
in rock magnetism or biomagnetism. However, any presence of residual nonuniform stress
in magnetite should become highly visible if cooled through the isotropic point and if
magneto stricti on or magnetostrictively controlled magnetization parameters are measured
concurrently.
3.4. Pressure Dependence

Hydrostatic pressure dependence of anisotropy, magnetostriction, and Curie point
have been studied and only a weak dependence has been detected. Nagata and Kinoshita
(1967) found that Kl and Kz decrease with pressure at the rate of - 5%/kbar while A111 and
AlOo increase at the rate of + 15%lkbar. Schult (1970) has experimentally observed that the
pressure dependence of Curie point for both magnetite and various titanomagnetites is
low, about 2°Klkbar.
4. Magnetic Domain States

4.1. Domain Theory
Some magnetic properties, particularly coercivity and remanence, vary greatly with
particle size. This is best illustrated in Fig. 3, which depicts schematically the variation
24 Chapter 2
Figure 3. Schematic illustration

showing the variation of coerciv-
ity (He) with grain size (in mi-
crometer) for magnetite. The su-
perparamagnetic (SP), single-
domain (SD), pseudo-single-do-
main (PSD), and multi domain
(MD) regions are also shown with
100 approximate transition sizes
Particle Diameter, a suitable for magnetite.
of coercivity with particle size for magnetite. As an introduction to magnetic domain the-
ory, Fig. 3 also shows the subdivision into superparamagnetic (SP), SD, PSD, and MD
particle sizes. The maximum coercivity occurs within the stable SD range; as particle size
increases from this maximum, coercivity decreases as the grain subdivides into domains;
as the particle size decreases from this maximum, coercivity again decreases due to thermal
fluctuations and goes to zero at the SP-SD transition. The large range in particle size-
from a few 100 A to greater than 100 !Lm (a factor of 10,000)-should also be noted.
Domains constitute a fundamental concept in magnetism. Magnetic domains are re-
gions within a magnetic material where the direction of spontaneous magnetization is
uniform; but different domains have different directions. Consequently, an MD particle
can have zero remanence if the magnetization directions in the domains mutually cancel
each other (Fig. 4a). An SD particle, on the other hand, is always magnetized to saturation
(Fig. 4b). Domain walls are boundaries between domains were the direction of magneti-
zation changes continuously from the direction in one domain in the direction of the other
domain. Domain walls are usually classified as (i) 180° walls, in which the direction of
magnetization changes by 180°, and (2) 90° walls, in which the direction changes by -90°
(e.g., Chikazumi. 1964).
Domains are formed in response to a particle's attempt to minimize its magnetic energy.
A large particle, uniformly magnetized, will induce magnetic poles on its surface, resulting
in a large magnetostatic energy contribution (Fig. 4b). To reduce this energy, the particle
can subdivide into domains (Fig. 4a), but only at the expense of introducing domain walls
Figure 4. (aJ Idealized sketch of a

multidomain particle. (hJ A single-
a b domain particle.
and the energy to produce them. Minimizing the magneto static and wall energies for a
given grain size, temperature, and applied field results in the equilibrium domain config-
uration.
The total magnetic energy is given by
(1)
The exchange energy (Eex) is produced by the angular deviations from parallel alignment
of neighboring spins within a domain wall. In order to reduce Eex, the angular deviations
should be small and, hence, the exchange contribution acts to make a domain wall as wide
as possible. The magneto crystalline anisotropy energy (Ea) is produced by angular devia-
tions from the easy directions; in order to reduce Ea , the domain wall should be as narrow
as possible to avoid the hard directions. The magnetoelastic energy (Es) is produced by
magnetostrictive strains and stresses within the particle and this energy determines
whether 180 or 90 domain walls are formed. The magnetostatic energy (Ems) is the dom-
0 0
inant term which causes domain formation. Finally, the magnetic energy (Eh ) is produced
by an external field; if the field is large enough, it will align the domains in the direction
of the field and destroy domain walls. Domain theory is reviewed by Kneller (1969), euUity
(1972), and Stacey and Banerjee (1974).
Equation (1) is a nonlinear partial differential equation in which the magnetization is
a continuous function of position with constant magnitude but of variable direction. The
result of minimizing (1) should lead automatically to domains and domain walls; however,
for finite three-dimensional crystal structures, the calculations have so far proved intract-
able (Brown, 1978). In specialized geometries (infinite cylinders, spheres) and in high
fields, the equations can be linearized and solved. This has resulted in a better under-
standing of domain nucleation and coercivity in fine particles and comes under the heading
of micromagnetics (Frei et 01., 1957; Brown, 1963). To circumvent the complexities of (1),
what is actually done to calculate domain transitions is to assume a domain configuration,
to calculate its energy, and then to compare it to the energy of a single domain. The lowest
energy configuration is then the equilibrium state (e.g .. Butler and Banerjee, 1975; Sato et
01., 1982). It should be noted, however, that the equilibrium domain state may not be the
domain state that is attained inevitably, but rather some other metastable states may also
form. Experimental evidence indicates that some very large particles, which should be
MD, remain in the SD state after application and removal of a saturating magnetic field
(Halgedahl and Fuller, 1980); and some very small particles, which should be SD, nucleate
domain walls after thermal demagnetization (Khrabrov et 01., 1974).
4.2. PSD Particles

The distinction between SD and MD particles is a straightforward one. However, for
particle sizes slightly larger than SD, the magnetic properties (particularly remanence and
coercivity) are much more SD-like than MD-like and hence the term pseudo-single domain
was coined by Stacey (1963). In MD particles, remanence and coercivity are controlled by
domain wall motion (an energetically easy process), whereas in SD particles, magnetization
changes only occur by rotation of its magnetic moment (an energetically harder process).
Thus, SD particles have high coercivities and more stable remanences than MD particles.
In theory, one should observe a drastic change in remanence and coercivity at the SD-
MD transition; however, there is instead a gradual change in magnetic properties and this
constitutes the PSD region. In a majority of rocks containing magnetite and titanomagnetite,
particle sizes are usually well above the SD size (e.g., Stacey and Banerjee, 1974) but are
characterized by stable NRM and high coercivities. It is suggested that magnetic moments
26 Chapter 2
due to domain walls (Dunlop, 1977, 1981), or magnetic moments due to surface effects
(Stacey and Banerjee, 1974; Banerjee, 1977), or both may help to bridge the gap between
SD and MD particles. PSD particles are assumed to contain just a few domains «10); but
the domain walls may constitute a large volume percent of the total grain volume; this
may explain their SD-like properties (Butler and Banerjee, 1975; Moskowitz and Banerjee,
1979; Dunlop, 1981). The opposing view is that SD particles do in fact exist either as
regions formed from exsolution between magnetite and ilmenite, or as ultrafine grains
(Evans, 1977). In sediments, a biogenic SD contribution may be a very real possibility
(Kirschvink and Lowenstam, 1979; Kirschvink, 1982).
4.3. Superparamagnetism
An SD particle of volume v has a uniform magnetization that is directed along some

axis where the anisotropy energy (Ea) is a local minimum. If v is small enough or the
temperature high enough, thermal fluctuations will be sufficient to overcome the aniso-
tropy energy barrier E = Emax - Emin , causing a spontaneous reversal of magnetization.
An ensemble of such particles will approach thermal equilibrium with a characteristic
relaxation time. In this ensemble, the net magnetic moment (fL) in zero field and at T >
OaK will average to zero (Jr = 0); but in an applied field there will be a net statistical
alignment of magnetic moments. The fraction of the total magnetization that will be aligned
by the field and at a particular T is given by the Langevin function
M(T) = coth (fLh/kT) - kT/fLh (2)
and is analogous to paramagnetism except now fL is not due to a single atom but due to
an SD particle consisting of > 105 atoms; hence, the name superparamagnetism (Bean and
Livingston, 1959; Kneller, 1969).
The approach to thermal equilibrium is governed by a relaxation time constant, first
derived by Neel (1949), and given by
(3)
In Eq. (3), fa is the frequency factor (-109 sec- 1 ; McNab et al., 1968); Is is the saturation
magnetization; Hk is the intrinsic coercive force determined by the dominant anisotropy.
In deriving Eq. (3), Neel assumed a uniaxial anisotropy, which is valid for elongated SD
particles. Bean and Livingston (1959), in contrast, have suggested a more generalized form
of Eq. (3):
lIT = fa exp( J ú b L â q F = (4)
where ú b =is now the barrier energy opposing spontaneous reversals.

The exponential nature of the relaxation time on v and T makes it possible to define
a temperature, TB (called the blocking temperature and TB < Te), at which the magneti-
zation goes from an unstable condition (T/t exp ú =1) to a stable condition (T/t exp ú =1) (Neel,
1949; Stacey and Banerjee, 1974). For example, a change in ú b =from 18kT to 57.7kT (a
factor of -3) results in a change in T from 1 sec to 4 X 109 years (a factor of 10 17 ). In
spherically shaped magnetite, this change in energy corresponds to an increase in grain
size from only -500 A to -740 A.
The coercive force can now be defined as the value of h which makes T = t exp in Eq.
(3):
He = Hk - [2kTHk In(tfa)/vIs]' (5)

where He is now a function of v, T, and t (Bean and Livingston, 1959). Only at T = OOK
does the coercive force He become equal to H k. The effects of thermal fluctuations for T
> OOK are to reduce He from its maximum values Hk and when He = 0 defines the SP-
SD transition (d s ). Equation (5) can be reformulated in terms of d s (Kneller and Luborsky,
1963):
(6)
4.4. Critical Sizes
4.4.1. Experimental Results

Experimental determinations of the SP-SD (d s ), SP-PSD (do), and PSD-MD (d m ) tran-
sitions are difficult because samples with narrow grain size and grain shape distributions
are difficult to obtain. Dunlop (1973) experimentally determined both d s and do in equant-
sized magnetite utilizing remanence ratio UrI!s) values at 77 and 300 0 K. At room temper-
ature, d s = (290-360) ± 50 A and do < 480 ± 50 A. Dunlop and Bina (1977), using thermal
fluctuation analysis, extrapolated their results to obtain a value of d s = 250 A. These data
indicate that there may be no stable SD region in equant-size magnetite at room temper-
ature, but rather a direct SP-MD transition (Dunlop, 1973). If there is a stable SD region,
then it must be extremely narrow (300-500 A). Slight elongation of the magnetite particles,
however, should ensure a stable SD region at room temperature (Butler and Banerjee, 1975;
Dunlop, 1981). It is interesting to note that most of the grain sizes of biogenic magnetite
are all elongated to some extent (Kirschvink and Gould, 1981; Towe and Moench, 1981).
Parry (1965) using TRM and coercivity data for magnetite determined d m to be 20 j,Lm.
Bailey (1975) and Day et al. (1977) determined d m to be between 10 and 20 j,Lm. Although
the PSD-MD transition is important in rock magnetism, this transition size is well above
the magnetite particle sizes found in biological systems.
The lower limit to SD behavior for maghemite was determined by Berkowitz et a1.
(1968) and found to be ú Q M M = A at room temperature. The upper limit to SD behavior has
not been experimentally determined, although theoretical estimates indicate that the tran-
sition sizes in maghemite should be slightly larger than those in magnetite (Morrish and
Yu, 1955; Morrish and Watt, 1957). The effect of Ti substitution is also an increase of the
transition sizes (Butler and Banerjee, 1975; Day et aI., 1977).
4.4.2. Theoretical Results

The difficulties inherent in experimentally determining domain transition sizes for
magnetite have prompted a considerable amount of work in estimating theoretical values
for these transitions. Theoretical estimates for d s and do have been made by Morrish and
Yu (1955), Frei et al. (1957), Murthy et a1. (1971), Evans (1972), Butler and Banerjee (1975),
and Moskowitz and Banerjee (1979), to name but a few.
In determining do, it is essential to know what is the lowest energy state of nonuniform
magnetization that will occur at do. Morrish and Yu (1955) assumed a circular spin con-
figuration; Frei et al. (1957) used a curling mode; Butler and Banerjee (1975) assumed a
two-domain structure. Butler and Banerjee (1975) also review the evidence and suggest
that in parallelepiped-shaped grains, an SD to two-domain (TD) transition should be the
one with the lowest energy. However, estimates from the various theoretical approaches
all agree with the experimental results for do, to within a factor of 2. The SP-SD transition
can be calculated by setting He = 0 in Eq. (5) (Butler and Banerjee, 1975). Table II sum-
marizes the theoretical and experimental estimates for ds, do, and d m in magnetite.
28 Chapter 2
TABLE II. Experimental and Theoretical Critical Grain Sizes for Magnetite
Experimental results
Dunlop (1973) 0.029-0.036° <0.048°
Dunlop and Bina (1977) 0.025 b
Day et 01. (1977) <O.l e 10-20e
Parry (1965) 15-20d
Theoretical results
Morrish and Yu (1955) 0.047"
Frei et 01. (1957) 0.048-0.05at
Butler and Banerjee (1975) 0.076g
Moskowitz and Banerjee (1979) 0.0808
0.44h
a Equant-sized magnetite (alb = 1).
b Extrapolation based on thermal fluctuation analysis.
C Synthetic, wet ground, irregular shapes.
d Natural, ground, irregular shapes .

• Circular spin mode (alb = 1).
f Curling mode (infinite cylinder, sphere).
g Two-domain mode (alb = 1).
h Two-domain mode (alb'" 0.4).
Indirect confirmation of the SD-TD transition model has come from the study of bio-
genic magnetites. There is strong biological evidence to suggest that biogenic magnetite
should be SD (Kirschvink and Lowenstam, 1979), and it is reassuring that almost all the
particle sizes measured for biogenic magnetites fall within the SD region calculated by
Butler and Banerjee (1975).
4.5. Coercivity
Coercivity is an extrinsic magnetic property dependent on particle size, particle shape,
grain-grain interactions, stress, and temperature and, as shown in Fig. 3, varies greatly
with particle size. The coercive force (He) is the reverse field required to reduce J. to zero
and it is most easily depicted on a hysteresis loop (Fig. 5). A related parameter is the
coercivity of remanence (Hre) which is the reverse field required to reduce a saturation
remanence (SIRM or Jr.) to zero.
The coercive force within the SD range (d. < d < do) varies as a - b(d)-3/2 (where a
and b are constants) and is well explained by the effects of thermal fluctuations (e.g.,
Kneller, 1969). The coercive force for d > do varies inversely with d (He d- n , 0.5 < n <
1); however, its dependence is not yet adequately understood on theoretical grounds. The
coercive force is zero for d < d., due to the onset of superparamagnetism.
Theoretical upper limits for He (in the absence of thermal fluctuations) are calculated
assuming a model in which the spins of all the atoms in the particle remain parallel during
rotation (Fig. 6a). This model is referred to as rotation in unison, coherent rotation, or the
Stoner-Wohlfarth (SW) model. Observed values of He are, however, almost always much
lower than those given by the SW model. Lower values for He can be calculated by invoking
noncoherent reversal modes whereby the spins do not remain parallel during reversing.
These reversal modes are called fanning, curling, and buckling and are shown in Fig. 6
Uacobs and Bean, 1955; Frei et al., 1957; Brown, 1963).
In Stoner and Wohlfarth's classic 1948 paper, coercivity and hysteresis were calculated
for ellipsoidal-shaped SD particles in which the magnetization reverses coherently against
Figure 5. Magnetic hysteresis loop illustrating saturation

magnetization (fs). saturation isothermal remanent mag-
netization (SIRM). coercive force (He). coercivity of reman-
ence (H re ). and initial susceptibility (Xo). H = applied field.
T = induced magnetization. -J
a uniaxial anisotropy. The total energy for a single SD particle in a field H is given by the
sum of the anisotropy and magnetic potential energy:
E = K sin 2 <1> - hI cos(9 - <1» (7)
where the angular orientations are shown in Fig. 7 and K is the anisotropy constant. If the
field is applied parallel to the long axis (9 = 0). then minimizing Eq. (7) will yield the
coercivity (Stoner and Wohlfarth. 1948):
He = 2K1ls (8)
The constant K for shape anisotropy is given by iIiNI;. where liN is the difference in
demagnetizing factors along the long (Na ) and short (Nb) axes of the grain. The maximum
value for liN is 2'IT (which occurs for an infinitely long. thin needle). The maximum coer-
civity due to shape is, therefore, 2'ITls (e.g., Cullity, 1972) and not 'lTls as erroneously given
by Stacey and Banerjee (1974). This corresponds to a maximum coercive force for magnetite
of 3000 Oe. For an ensemble of noninteracting, randomly oriented SD grains. the calculated
coercivity is reduced to 0.48(2K1ls) (Stoner and Wohlfarth, 1948). In rocks where the mag-
netic particles are finely disseminated, this reduced value for coercivity is more applicable;
however, for biogenic magnetite, where interactions and alignment of particles are more
the norm, Eq. (8) is probably more pertinent.
[ill B
(( ú =
,
\ \ \
//
//
UNISON
./
t
It
,\
'
I I
BUCKLING
\\ I ú J ú ú FK=
CURLING
ú=
FANNING
(0 ) ( b) (c) (d)
Figure 6. Magnetization reversal modes for single-domain particles. (a) Rotation in unison, (b) buckling
mode. (c) curling mode. and (d) fanning mode. After Kneller (1969).
30 Chapter 2
/
/
/
/
/ Figure 7. Effect of applied field H on magnetization Ms of
/ a single-domain particle.
The SW model can also be applied to particles with cubic anisotropy, although the
calculations are more complex (Johnson and Brown, 1959). The SW coercivities for shape,
stress, and magneto crystalline anisotropies are
He = tlNJs shape
He = 3A sU/Js stress
He = 2Kl/Js uniaxial
He = 4Kl/3Js cubic (Kl < 0)
The noncoherent reversal modes were first discussed by Jacobs and Bean (1955), Frei
et 01. (1957), and Brown (1963). Excellent reviews of these mechanisms are given by Kneller
(1969) and Brown (1963). Of particular interest is the fanning or chain-of-spheres model
proposed by Jacobs and Bean (1955) as a possible reversal mode in elongated SD particles.
Consider a chain of spheres, with either point contact or slight separation (Fig. 6d). Mag-
netization reversal can occur by (1) parallel rotation of each sphere's magnetization or by
(2) a symmetric fanning mechanism where the moments fan out in a plane, alternating in
their sense of rotation from one sphere to the next. The fanning mode envisioned by Jacobs
and Bean for elongated SD particles (Fig. 8a) seems most ideal for the particle configurations
in some magnetotactic bacteria (Fig. 8b) (Frankel and Blakemore, 1980; Denham et 01.,
1980). However, coercivity values based on the fanning mode are at least a factor of 2
greater than experimental results (Denham et 01.,1980). However, it should be remembered
that the experimental results are not on a single bacterium, but rather on a randomly
oriented ensemble which would probably cause the coercivity to be reduced (Jacobs and
Bean, 1955).
4.6. Interactions
Interacting SP particles can collectively act as SD particles (Radhakrishnamurthy et

01., 1973) and interacting SD particles can collectively act as MD particles (e.g., Kneller,
Figure 8. Electron micrograph of (a) electro-deposited elongated iron particles (after Luborsky, 1961)
and (b) magnetotactic spirillium strain MS-1 (after Frankel and Blakemore, 1980). The chain-of-spheres
model is used to describe the reversal mechanism in (a). The scale in (a) is 0.4 f.Lm x 0.7 f.Lm and in
(b) each bar = 0.5 f.Lm.
1969). The effects of interactions will decrease He (e.g., Dunlop and West, 1969; Cullity,
1972), but expand the SD range (Morrish and Watt, 1957). Indirect evidence of this comes
from the work on biogenic magnetite in chiton teeth. The magnetic particles in chiton teeth
are closely packed and separated from each other by -200 A of organic material (Kirschvink
and Lowenstam, 1979). Assuming that this magnetite is SD, then the magnetic particle
sizes in both Cryptochiton stelleri (d > do) and Lepidopleurus sp. (d < d s ), which plot
outside the SD region (see Fig. 4; Kirschvink and Lowenstam, 1979), should behave as SD
particles due to interactions (Kirschvink and Lowenstam, 1979). Cisowski (1981) has also
presented experimental evidence for the interacting nature of the magnetite in chiton teeth.
5. Remanent Magnetizations
5.1. Isothermal Remanent Magnetization
By definition, remanent magnetization refers to the residual magnetization in a sample

measured in a zero ambient field after it has been subjected to an applied magnetic field
with or without a simultaneous variation in temperature (or pressure). Isothermal remanent
magnetization (IRM) refers to that remanence left in the sample after a steady field (say,
32 Chapter 2
10-10,000 De) has been applied for a short time (-1 sec) and then switched off. For mag-
netite, hysteresis studies show that for fields lower than 500 De, IRM is acquired by ir-
reversible domain wall translation, while at fields above 500 De, irreversible domain wall
rotations occur. Particle size dependence of saturation IRM (SIRM) is due to the transition
from small SD through PSD to MD states and has been studied extensively by Parry (1965)
and Rahman et al. (1973). For magnetite of non-SD grain size He and SIRM (Jrs) are ap-
proximately related by the experimental relationship:
where He is the coercive force and N the demagnetizing factor due to shape anistropy.
While it had been assumed that both He and N can be independently determined from
hysteresis measurements (Neel, 1955; Stacey and Banerjee, 1974), it has now been shown
that the experimental determination of N from a hysteresis loop presupposes a strict and
a priori theoretical validity of the approximate relationship given above (Smith and Merrill,
1982). There is some cause for hope, however, as Dunlop (1983) has stated that theoretical
calculations of N for 2-domain to 10-domain particles show that irrespective of their mag-
netization state, N is indeed controlled by shape anisotropy and is constant to within
acceptable limits (± 5-20%). A knowledge of Jrs and He therefore can lead to an estimate
of N not only for SD grains but also for non-SD particles.
An IRM imparted at low fields (e.g., 20 De) may be moderately large but easily de-
magnetized by low peak values (-20-50 De) of alternating field (AF) while SIRM is highly
resistant to AF demagnetization. For nondestructive laboratory characterization of mag-
netic minerals, it is highly recommended, therefore, to restrain from applying SIRM to a
sample until other tests have been carried out lest it result in an irreversible magnetic state
for the sample. For a given suite of samples, SIRM magnitude (Jrs) can provide a convenient
normalization parameter for their natural remanent magnetization, as any variation in the
total magnetic carrier content will be reflected in their Jrs values. Fuller (1974) has used
this technique to provide relative paleointensity values for lunar rocks of different ages.
5.2. Anhysteretic Remanent Magnetization
Anhysteretic remanent magnetization (ARM) is a popular method of imparting lab-

oratory remanence to rocks and it is used primarily to characterize their magnetic carriers.
Unlike the case of IRM, in ARM it is necessary to employ both an AF of a given peak value
as well as a small steady field, of the order of 1 De. In imparting an ARM, the peak AF is
slowly reduced to zero in the presence of a constant steady field. Then the steady field is
switched off too, leading to the acquisition of an ARM. For SD particles of volume v, Jaep
(1971) has shown that the ARM intensity (JARM) at room temperature is given by
where Js(TR) = saturation magnetization at room temperature (TR), k = Boltzmann's con-

stant, he = coercive force, he = applied external steady field, and hi = the interaction
field or Lorentz field acting on an average particle but arising from all the other grains in
the sample. Banerjee and Mellema (1974) showed that hi can be measured from the de-
pendence of /ARM on he. For MD particles, Gillingham and Stacey (1971) show that
he ( 1 )
/ARM = N 1 + NXD
where he = external field, N = average demagnetizing factor of the magnetic particles,

and XD = magnetic susceptibility in the self-demagnetizing field of the particle.
Anticipating our discussion in Section 5.4 of the theory of thermoremanent magnet-
ization (TRM) , we point out at this stage that historically speaking, one of the strongest
reasons for studying ARM in SD particles of magnetite has been the similarity in behavior
between ARM and TRM as first shown by Rimbert (1959), and later confirmed by Levi and
Merrill (1976). One can point to a qualitative similarity by appealing to the similar dis-
ordering role of AF and temperature on the magnetization of a sample. Thus, while TRM
is acquired by magnetite cooling through its blocking temperatures (TB ; see next section),
ARM is acquired when the AF is slowly decreased from its peak value to zero. This analogy
is not rigorous, however, and it is more valid for SD and PSD particles than for MD as
shown by a comparison of the AF demagnetization curves of ARM and TRM for some MD
particles.
Partial ARM is defined as the fractional ARM acquired when the applied AF is de-
creased from one peak value to a value intermediate between that and zero. Thus, the total
ARM acquired by a sample when the AF is decreased from its maximum peak value Hmax
to zero can be compared against the sum of partial ARMs acquired in a series of steps
when the ambient AF is decreased from Hmax to Hn, then from Hn to Hn- 1 , and so on until
zero field is reached. It is found that the sum of such partial ARMs is equal to the total
ARM. Thus,
L Partial ARM = total ARM

n
The model for SD ARM assumes that a given peak AF activates magnetite particles with
remanence coercivity Hrc corresponding closely to the peak AF value. Therefore, the AF
demagnetizing curve of a given ARM or an ARM acquisition curve (as a function of applied
steady field) can be construed as a direct analog of the remanence coercivity spectrum.
Schmidbauer and Veitch (1980) have shown that for a 0.2- to O.4-jJ.m magnetite sample
containing perhaps one or two domains, ARMs acquired at a given peak AF can be de-
stroyed by AF demagnetization carried out at that peak value. All of this supports the
common practice of using the AF demagnetization curve of ARM of magnetite (especially
if SD grained) to characterize the distribution of remanence coercivity Hrc and thus, by
inference, the distribution of effective particle sizes.
ARM will be compared with TRM in Section 5.4. So far as the comparison of ARM
with IRM goes, a given steady field of weak to moderate strength (1-100 De) always pro-
duces a much larger intensity of ARM than IRM. This is because in the case of ARM, the
steady field is assisted by the AF in the acquisition process. Similarly, the remanence
coercivity fractions of the sample activated by the two processes are also dissimilar. Thus,
if the normalized AF demagnetization curves of ARM and IRM given at the same low steady
field are compared, it will be found that the ARM is much harder, i.e., has higher Hrc values
and demagnetizes at higher peak values of AF than the IRM. In fact, AF demagnetization
curves of ARM and IRM can be used to provide good ideas about the separate distributions
of higher and lower Hrc fractions, respectively, in the sample.
5.3. Chemical Remanent Magnetization
The origin of chemical remanent magnetization (eRM) can be explained by reference

to the concept of superparamagnetism introduced earlier. An SD grain of magnetite will
34 Chapter 2
Q)
E
l-
-
e:
o
c
)(
c
Q)
0::
Figure 9. Idealized relaxation time versus temper-

Temperature T ature diagram for TRM and CRM.
approach thermal equilibrium of its magnetization (induced or remanent) according to a

relaxation time constant T given by the following most general expression:
where fo = a constant frequency factor (-109 sec- 1 ), E = magnetic energy barrier opposing
spontaneous rotation of Is, k = Boltzmann's constant, and T = ambient temperature. In
the case of remanent magnetization (Le., no applied field) of magnetite, E is due mainly
to shape anisotropy or the demagnetization factor N. For perfect spheres, E is due mainly
to the first-order magnetocrystalline anisotropy constant K1 • In either case, as E denotes
total magnetic energy, it is a product of two terms, energy per unit volume and the volume
v itself. The relaxation time constant T is thus related to volume v and ambient temperature
T.
We can now proceed with the description of the acquisition of CRM. In Fig. 9, the
exponential temperature dependence of T has been indicated for two values of v where V2
> V1' The point A denotes the location of a small magnetite particle in the T-T space and
the particle diameter d ú = ds (the superparamagnetic threshold). The net IRM at point A
due to the application of a steady field (e.g., the earth's field) is therefore zero. However,
if due to a physicochemical process such as oxidation, hydrolysis, or dehydration the
magnetite particle increases in volume from V1 to V2, it will travel in the T-T space from
A to B where, let us assume, T ú =t exp (the experimental time constant) and the new diameter
d ú =ds . Somewhere between A and B the particle will have gone through a blocking volume
VB (or a blocking diameter dB) where the remanence will become finite and if the penul-
timate value of T at point B is suitably large, this new remanence or CRM will also be stable
with regard to time, temperature, or AF. Here again we see the difference between IRM
and CRM in that the additional process of grain growth in CRM leads to a much higher
intensity of remanence than what would be achieved by a pure IRM. Kobayashi (1959) has
shown convincingly that when magnetite is formed by chemical reduction of hematite,
the AF demagnetization curve of the CRM is much more stable than the IRM. The intensity
of CRM in SD particles depends directly on he (the external field) and inversely on the
anisotropy energy at the temperature T of CRM acquisition. This will be dealt with further
when we discuss TRM.
From the description of the CRM process, it should be clear that much of the SD grains
of magnetite of biological origin will carry a CRM. In fact, because of the simplicity of the
biological process of magnetite growth, it may be wise for us to use magnetite from bacteria
as preferred samples for the study of the origin and stability of CRM.
When magnetotactic bacteria die, the magnetite crystals contained in them add to the
sediments and the sediment as a whole should acquire a depositional or postdepositional
remanent magnetization (DRM or PDRM). Confirmed cases of such behavior are still rare
(e.g., Kirschvink, 1982).
5.4. Thermoremanent Magnetization
The concept of superparamagnetism can be used to explain the origin of TRM, as was
done for CRM. In Fig. 9, we showed that the acquisition of CRM results when T becomes
much greater than t exp because of grain growth through VB, the blocking volume. But the
same result can also be obtained by traveling along the path AC when a critical (or blocking)
temperature TB is reached upon cooling a sample from an elevated temperature. TRM is
therefore that remanent magnetization which is "frozen in" when a magnetic sample is
cooled in the presence of a field from its highest blocking temperature to room temperature.
Both the cooling rate and the magnitude of the ambient field can affect the blocking tem-
perature perceptibly. While a fast cooling rate raises the blocking temperature (York, 1978;
Dodson and McClelland-Brown, 1980; Halgedahl et 01., 1980), the presence of a moderate
to large (10-100 De) magnetic field depresses it (Sugiura, 1980; Clauter and Schmidt, 1981).
The intensity (hRM) of TRM for SD particles was studied by Neel (1949) for nonin-
teracting particles and was shown to be linearly proportional to the ambient field (he), if
the field is small (0.1-1 De), i.e., hRM/Js(TR) = ehe where C = (vJsB/3kT B). v represents
the average particle volume and JsB the saturation magnetization of magnetite at T = TB.
Although this provided the necessary theoretical support for archeomagnetists who wanted
to determine the paleointensity of the earth's magnetic field from archeological material,
it was soon found that the theoretically predicted constant of proportionality C was too
high compared to experimental results. Three decades of research following Neel's pioneer
work has failed, however, to resolve the difficulty. One approach, favored at one time by
Dunlop (1968), is to estimate the reduction of hRM due to particle interaction. The difficulty
with this approach is that it is hard to quantify the interaction field by independent ex-
periments or theoretical derivation. The second approach has been to take into account
the fact that true SD particles of magnetite are rare in nature (expecting magnetotactic
bacteria) and that most particles carrying stable TRM are PSD and hence they contribute
less to the hRM than true SD particles. Dunlop et 01. (1974) tested the PSD model of Stacey
and Banerjee (1974) and found agreement between observed PSD magnetic moment and
that predicted by Stacey and Banerjee (1974). In his review of TRM theories, however, Day
(1977) has pointed out that there is no satisfactory theory of thermoremanence extant today
that would explain satisfactorily the observed TRM in SD, PSD, and MD particles of mag-
netite. A recent development in the area is the application of the Langevin function to the
TRM of SD magnetite incorporating the eight easy axes of a magnetite crystal over which
the remanence can be distributed (R. J. Luce, V. A. Schmidt, and F. Keffer, 1983, personal
communication). It will be interesting to see if this model can better describe the exper-
imental results. Thermoremanence of MD magnetite has been approached theoretically by
a large number of workers, most recently by Merrill (1981). A satisfactory model that takes
into account the complexities of the real situation is still to emerge, however.
It is important to compare TRM with CRM and ARM. Kobayashi (1959) showed, in a
study of CRM acquisition in magnetite, that the stability of AF demagnetization of CRM
was closer to that of a total TRM than a partial thermoremanence acquired by the sample
36 Chapter 2
upon cooling from the temperature of magnetite formation. This seems to be the consensus
opinion although we lack numerous and convincing examples of comparison of CRM and
TRM in the same sample. Relative AF stability of ARM and TRM was mentioned in an
earlier part of this section. In general, they are similar. However, this similarity in AF
stability does not extend to the ratio of relative intensities of the two types of remanence
(!ARM/hRM) in a given small field. Banerjee and Mellema (1974) showed that for true SD
grains, the ratio approaches unity as the ambient temperature is raised. Levi and Merrill
(1976), however, have pointed out that the ratio is a strong function of particle size and
it is, therefore, hard to predict hRM from !ARM and vice versa unless the particle size is
carefully monitored.
6. Magnetic Granulometry
Some magnetic parameters can be used to differentiate between SP, SD, and MD par-
ticles and indirectly infer grain size. Admixtures of SP, SD, and MD particles or titaniferous
or oxidized magnetite or combinations of any or all of these can, however, lead to ambig-
uous results (Senanayake and McElhinny, 1981; Clark and Schmidt, 1982).
The hysteresis parameter J./f. and Hre/He are good indicators of domain state. In SD
particles, J./f. > 0.5 and 1 < Hre/He < 2; in MD particles, 0.01 < fr/f. < 0.3 and 2 < Hrel
He < 5; and in SP particles, fr/f. ú = 0.1 and Hre/He > 10. Strictly speaking, SP particles
have both fr and He = 0; but small admixtures of SP and SD particles usually occur, giving
finite, but low values for fr and He (Wasilewski, 1973; Dunlop, 1981). Hysteresis meas-
urements from magnetotactic bacteria give fr/f. = 0.47 and Hre/He = 1.23 which indicate
SD-like behavior (Denham et 01., 1980).
The parameters fr/xo and XolJ. are also useful in distinguishing SP from non-SP be-
havior. In SD or MD particles of magnetite, J./Xo will vary from -20 to 700 De, whereas
for SP particles, J./Xo < 0.12 De (Thompson et 01., 1980). Values for xolfs in SD and MD
particles of magnetite will rarely exceed -0.7 x 10- 3 De-I; values significantly greater
than this will therefore indicate a possible SP contribution (Dunlop, 1981).
The intensity of ARM (fARM) is found to be much more influenced by the smaller SD
and PSD particles than by the coarser-grained MD material, the latter contributing much
more effectively to the low-field susceptibility XO. Thus, when !ARM or XARM (Le., specific
ARM intensity acquired in a steady field of 1 De) is plotted against XO of pure magnetite
of different particle sizes, it is found that the slope XARMlXo shows continuous variation
as the magnetic domain state changes from MD through PSD to SD (King et 01.,1982). This
has been the basis for a rapid method of magnetic granulometry (Banerjee et 01.,1981; King
et 01., 1982).
The other techniques of magnetic granulometry involve the measurement of magnetic
parameters at elevated temperatures. Dunlop (1976) has used Neel's concept of thermal
fluctuation field (Hq) to devise a method of thermal fluctuation analysis. Hysteresis loop
parameters, fs, He' and Hre , are measured as a function of temperature (T) and the variation
of He or Hre versus T leads to the determination of the average volume v of the magnetic
material. Bol'shakov and Shcherbakova (1979) have devised a method based on thermal
demagnetization of partial thermoremanence (PTRM) which purports to show that while
the TRM of SD particles vanish when thermally demagnetized to the respective blocking
temperature, the TRM of MD grains have to be heated as high as the Curie temperature for
complete demagnetization. This may form the basis of a qualitative magnetic granulometric
method, but at face value it would appear that contrary to Levi and Merrill (1976),
Bol'shakov and Shcherbakova (1979) have found that Neel's law of additivity of TRM does
not hold. It will be interesting to see if this observation is confirmed by others in the future.
x
MO
so
Figure 10. Schematic representation of the

variation of initial susceptibility (x) with
temperature for magnetite, which at room
temperature (300 K) is in a multidomain
0 100 350
(MD), or single-domain (SD), or superpara-
magnetic (SP) state. Temperature (K)
Low-temperature measurements (77-300 0 K) of coercivity, remanence, and low-field

susceptibility are also diagnostic of domain state. The variation of Xo with temperature for
magnetite particles that are SP, or SD, or MD at room temperature is shown schematically
in Fig. 10. Susceptibility in MD particles is practically independent of temperature between
the isotropic point and the Curie temperature (Stacey and Banerjee, 1974); however, near
the isotropic point, XO increases and produces the characteristic low-temperature peak. In
SD particles, susceptibility is controlled by a combination of magneto crystalline and shape
anisotropies and will show a temperature dependence. Elongated SD particles, where shape
anisotropy predominates, will show a slight decrease with temperature and the low-tem-
perature peak at 118°K will be suppressed. SD particles with a significant magnetocrys-
talline anisotropy will show a peak at Tm, but the increase in XO at Tm will be much less
than that observed in MD particles (Clark and Schmidt, 1982). SP particles will produce
the most drastic changes in XO with temperature. As the temperature is decreased and passes
through the blocking temperatures of the SP particles, XO will decrease by anywhere from
a factor of 20 to 200 (Stephenson, 1970; Clark and Schmidt, 1982).
Superparamagnetic particles will also produce a large increase in Jr and He upon cool-
ing through their blocking temperature (e.g., Fig. 6 in Kirschvink and Lowenstam, 1979).
In SD and MD grains, Jr and He will not show such a drastic variation with decreasing
TABLE III. Characteristic Values and Behavior for Some Magnetic Parameters for
Magnetite
Parameter SP SD MD
JrlJs qO.01 0.3-0.5 0.01-0.3

Hrc/Hc p10 1-2 3-5
Jrlxo <0.120e 20-7000e 20-7000e
XolJs >0.0070e- 1 0.0070e- 1 0.0070e- 1
Xo- T large decrease small decrease peak at ú N N U Œh =
Hc T large increase small increase decrease at ú N N U Œh =
Jr-T large increase small increase decrease at ú N N U Œh =
38 Chapter 2
temperature. For MD particles as they cool through Tm, both He and Ir will decrease due
to the vanishing of K1 . In SD particles controlled by shape anisotropy, both He and Ir will
slightly increase concomitant with the increase in Is. The temperature dependence of Ir or
Xo for SD and SP particles can also be used to determine magnetic grain size distributions
(e.g., Kneller, 1969; Stephenson, 1970). Table III summarizes the magnetic parameters and
their characteristic behavior and values for SP, SD, and MD particles of magnetite.
References
Bailey, M. E., 1975, The magnetic properties of pseudo-single domain grains, M.Sc. thesis, University
of Toronto.
Banerjee, S. K., 1977, On the origin of stable remanence in pseudo-single domain grains, J. Geomagn.
Geoelectr. 24:319-330.
Banerjee, S. K., and Mellema, J. P .. 1974, A new method for determination of paleointensity from the
A.RM. properties of rocks, Earth Planet. Sci. Lett. 23:177-184.
Banerjee, S. K., King, J., and Marvin, J., 1981, A rapid method for magnetic granulometry with ap-
plications to environmental studies, Geophys. Res. Lett. 8:333-336.
Bate, G., 1980, Recording materials, in: Ferromagnetic Materials, Volume 2 (E. P. Wohlfarth, ed.),
North-Holland, Amsterdam, pp. 381-507.
Bean, C. P., and Livingston, J. D., 1959, Superparamagnetism, J. Appl., Phys. 30:1205-1295.
Berkowitz, A. E., Schuele, W. J., and Flanders, P. J., 1968, Influence of crystallite size in the magnetic
properties of acicular 'Y-Fe203 particles, ,. Appl. Phys. 39:1261-1263.
Bickford, L. R, 1953, The low temperature transformation in ferrite, Rev. Mod. Phys. 25:75-79.
Bol'shakov, A. S., and Shcherbakova, V. V., 1979, Thermomagnetic criterion for determining the do-
main structure of ferrimagnetics, Phys. Solid Earth. 15:111-117.
Brown, W. F., Jr., 1963, Micromagnetics, Interscience, New York.
Brown, W. F., Jr., 1978, Domains, micromagnetics, and beyond: Reminiscences and assessments, J.
Appl., Phys. 49:1937-1942.
Butler, R F., and Banerjee, S. K., 1975, Theoretical single-domain grain size range in magnetite and
titanomagnetite, ,. Geophys. Res. 80:4049-4058.
Chikazumi, S., 1964, Physics of Magnetism, Wiley, New York.
Cisowski, S., 1981, Interacting vs. non-interacting single domain behavior in natural and synthetic
samples, Phys. Earth Planet. Inter. 26:52-56.
Clark, D. A., and Schmidt, P. W., 1982, Theoretical analysis of thermomagnetic properties, low-tem-
perature hysteresis and domain structure of titanomagnetites, Phys. Earth Planet. Inter. 30:300-
316.
Clauter, D. A., and Schmidt, V. A., 1981, Shifts in blocking temperature spectra for magnetite powders
as a function of grain size and applied magnetic field, Phys. Earth Planet. Inter. 26:81-92.
Cullity, B. D., 1972, Introduction to Magnetic Materials, Addison-Wesley, Reading, Mass.
Day, R., 1977, TRM and its variation with grain size, Adv. Earth Planet. Sci. 1:1-33.
Day, R, Fuller, M. D., and Schmidt, V. A., 1977, Hysteresis properties of titanomagnetites: Grain-size
and compositional dependence, Phys. Earth Planet. Inter. 13:1206-1216.
Denham, C. R, Blakemore, R P., and Frankel, R B., 1980, Bulk magnetic properties of magnetostatic
bacteria, IEEE Trans. Magn. Mag-16:1006-1007.
Dodson, M. H., and McClelland-Brown, E., 1980, Magnetic blocking temperatures of single-domain
grains during slow cooling, ]. Geophys. Res. 85:2625-2637.
Dunlop, D. J., 1968, Monodomain theory: Experimental verification, Science 162:256-258.
Dunlop, D. J., 1973, Superparamagnetic and single domain threshold sizes in magnetite, J. Geophys.
Res. 78:1780-1793.
Dunlop, D. J., 1976, Thermal fluctuation analysis: A new technique in rock magnetism, J. Geophys.
Res. 81:3511-3517.
Dunlop, D. J., 1977, The hunting of the psark, f. Geomagn. Geoelectr. 24:243-318.
Dunlop, D. J., 1981, The rock magnetism of fine particles, Phys. Earth Planet. Inter. 26:1-26.
Dunlop, D. J., 1983, On the demagnetizing energy and demagnetizing factor of a multi domain ferro-
magnetic cube, Geophys. Res. Lett. 10:79-82.
Dunlop, D. J., and Bina, M. M., 1977, The coercive force spectrum of magnetite at high temperatures:
Evidence for thermal activation below the blocking temperature, Geophys. J. R. Astron. Soc.
51:121-147.
Dunlop, D. J., and West, G. F., 1969, An experimental evaluation of single domain theories, Rev.
Geophys. Space Phys. 1:709-757.
Dunlop, D. J., Stacey, F. D., and Gillingham, D. E. W., 1974, The origin of thermoremanent magen-
tization: Contribution of pseudo-single-domain magnetic moments, Earth Planet. Sci. Lett. 21:288-
294.
Evans, M. E., 1972, Single domain particles and TRM in rocks, Commun. Earth Sci. Geophys. 2:139-
148.
Evans, M. E., 1977, Single domain oxide particles as a source of thermoremanent magnetization, J.
Geomagn. Geoelectr. 29:267-276.
Frankel, R B., and Blakemore, R P., 1980, Navigational compass in magnetic bacteria, J. Magn. Magn.
Mater. 15-18:1562-1564.
Frankel, R B., Blakemore, R P., and Wolte, R. S., 1979, Magnetite in freshwater magnetotactic bacteria,
Science 203:1355-1356.
Frei, E. H., Shtrikman, S., and Treves, D., 1957, Critical size and nucleation field of ideal ferromagnetic
particles, Phys. Rev. 106:446-455.
Fuller, M., 1974, Lunar magnetism, Rev. Geophys. Space Phys. 12:23-70.
Gillingham, D.E.W., and Stacey, F. D., 1971, Anhysteretic remanent magnetization (ARM) in magnetic
grains, Pure Appl. Geophys. 8:160-165.
Goodenough, J. B., 1963, Magnetism and the Chemical Bond, Wiley-Interscience, New York.
Gorter, E. W., 1955, Some properties of ferrites in connection with their chemistry, Proc. lnst. Radio
Eng. 43:1945-1973.
Halgedahl, S., and Fuller, M., 1980, Magnetic domain observations of nucleation processes in fine
particles of intermediate titanomagnetite, Nature 288:70-72.
Halgedahl, S., Day, R, and Fuller, M., 1980, The effect of cooling rate on the intensity of weak-field
TRM in single-domain magnetite, J. Geophys. Res. 85:3690-3698.
Hamilton, W. c., 1958, Neutron diffraction investigation of the 119 K transition in magnetite, Phys.
Rev. 110:1050-1057.
Ishikawa, Y., Syono, Y., and Akimoto, S., 1964, Neutron diffraction study of Fe30.-FezTi04 series,
Annu. Prog. Rep. Rock Magn. Res. Group Jpn. 14.
Jacobs, 1. S., and Bean, C. P., 1955, An approach to elongated fine-particle magnets, Phys. Rev.
100:1060-1067.
Jaep, W. F., 1971, Role of interactions in magnetic tapes, J. Appl. Phys. 42:2790-2794.
Johnson, C. E., and Brown, W. F., 1959, Stoner-Wohlfarth calculation on particles with both mag-
netocrystalline and shape anisotropy, J. Appl. Phys. 30:3205-3225.
Khrabrov, V. I., Onoprienko, 1. G., and Shur, S. Y., 1974, Zh. Eksp. Tear. Fiz. 67:344-350 (in Russian).
King, J., Banerjee, S. K., Marvin, J., and bzdemir, b., 1982, A comparison of different magnetic methods
for determining the relative grain size of magnetite in natural materials: Some results from lake
sediments, Earth Planet. Sci. Lett. 59:404-419.
Planet. Sci. Letter. 54:388-392.
Kirschvink, J. 1., and Gould, J. L., 1981, Biogenic magnetite as a basis for magnetic field detection in
animals, BioSystems 13:181-201.
Paleomagnetic, sedimentologic, and biologic implications of organic magnetic, Earth Planet. Sci.
Lett. 44:193-204.
Kneller, E., 1969, Fine particle theory, in: Magnetism and Metallurgy, Volume 1 (A. E. Berkowitz and
E. Kneller, eds.), Academic Press, New York, pp. 366-465.
Kneller, E., and Luborsky, F. E., 1963, Particle size dependence of coercivity and remanence of single-
domain particles, J. Appl. Phys. 34:656-658.
Kobayashi, K., 1959, Chemical remanent magnetization of ferromagnetic minerals and its application
to rock magnetism, J. Geomagn. Geaelectr. 10:99.
Levi, S., and Merrill, R T., 1976, A comparison of ARM and TRM in magnetite, Earth Planet. Sci.
Lett. 32:171-184.
40 Chapter 2
Lindsley, D. H., 1976, The crystal chemistry and structure of oxide minerals as exemplified by the
Fe-Ti oxides, in: Oxide Minerals Short Course Notes (D. Rumble, ed.), Southern Printing Co.,
Blacksburg, Va., pp. L1-L52.
Lubrosky, F. E., 1961, Development of elongated particle magnets, J. Appl. Phys. 32:1715-1835.
McNab, T. R, Fox, R A., and Boyle, A. J. F., 1968, Some magnetic properties of magnetic (Fe 304)
microcrystals, J. Appl. Phys. 39:5703-5711.
Merrill, R T., 1981, Toward a better theory of thermal remanent magnetization, J. Geophys. Res.
86:937-949.
Morrish, A. H., and Watt, A. R, 1957, Effect of the interaction between magnetic particles on the
critical single-domain size, Phys. Rev. 105:1476-1478.
Morrish, A. H., and Yu, S. P., 1955, Dependence of the coercive force on the density of some iron
oxide powders, J. Appl. Phys. 26:1049-1055.
Moskowitz, B. M., 1981, Methods for estimating Curie temperatures of titanomaghemites from ex-
perimental Js- T data, Earth Planet. Sci. Lett. 53:84-88.
Moskowtiz, R, and Banerjee, S. K., 1979, Grain size limits for pseudosingle domain behavior in mag-
netite: Implications for paleomagnetism, IEEE Trans. Magn. Mag-15:1241-1246.
Murthy, G. S., and Patzold, R, 1982, Magnetic granulometry results from intrusive rock samples,
Nature 295:688-690.
Murthy, G. S., Evans, M. E., and Gough, D. I., 1971, Evidence for single domain magnetite in the
Michikaman anorthosite, Can. J. Earth Sci. 8:361-370.
Nagata, T., and Kinoshita, H., 1967, Effect of hydrostatic pressure on magnetostriction and megne-
tocrystalline anisotropy of magnetite, Phys. Earth Planet. Inter. 1:44-48.
Neel, L., 1949, Theorie du trainage fmagnetique des ferromagnetiques en grains fins avel applications
aux terres cuites, Ann Geophys. 5:99-136.
Neel, L., 1955, Some theoretical aspects of rock magnetism, Adv. Phys. 4:191-242.
Parker, R, 1975, Electrical transport properties, in: Magnetic Oxides (D. J. Craik, ed.), Wiley, New
York, pp. 421-482.
Parry, L. G., 1965, Magnetic properties of dispersal magnetic powders, Philos. Mag. 11:303-312.
Pauthenet, R, 1950, Variation thermique de l'aimantation spontanee des ferrites de nickel, cobalt, fer
et manganese, C.R. Acad. Sci. 230:1842-1844.
Radhakrishnamurthy, C., Sastry, N. P., and Deutsch, E. R, 1973, Ferromagnetic behavior of interacting
superparamagnetic particle aggregates in basaltic rocks, Pramana 1:61-65.
Rahman, A. A., Duncan, A. D., and Parry, L. G., 1973, Magnetization of multi domain magnetite par-
ticles, Riv. Ital. Ceotis. 22:259-266.
Rimbert, J., 1959, Contribution at'etude de I'action de champs alternatifs sur les aimantations rema-
nents des roches: Applications geophysiques, Rev. Inst. Fr. Pet. 14:123-155.
Sato, M., Yoshihiro, I., and Nakae, H., 1982, Magnetic domain structures and domain walls in iron
fine particles, J. Appl. Phys. 53:6331-6334.
Schmidbauer, E., and Veitch, R J., 1980, Anhysteretic remanent magnetization of small multi domain
Fe304 particles dispersed in various concentrations in a non-magnetic matrix, J. Ceophys. 48:148-
152.
Schult"A., 1970, Effects of pressure on Curie point of titanomagnetite (1-x)Fe304'xFe2Ti04, Earth
Senanayake, W. E., and McElhinny, M. W., 1981, Hysteresis and susceptibility characteristics of mag-
netite and titanomagnetites: Interpretation of results from basaltic rocks. Phys. Earth Planet. Inter.
26:47-55.
Shull, C. E., Wallan, E. 0., and Kochler, W. C., 1951, Neutron scattering and polarization by ferro-
magnetic materials, Phys. Rev. 84:912-921.
Smith, G., and Merrill, R T., 1982, The determination of the internal magnetic field in magnetic grains,
J. Geophys. Res. 87:9419-9423.
Stacey, F. D., 1963, The physical theory of rock magnetism, Adv. Phys. 12:45-133.
Stacey, F. D., and Banerjee, S. K., 1974, The Physical Principles of Rock Magnetism, Elsevier, Am-
sterdam.
Stephenson, A., 1970, Single domain grain distributions method for the determination of single domain
grain distributions, Phys. Earth Planet. Inter. 4:353-360.
Stoner, E. L., and Wohlfarth, E. P., 1948, A mechanism of magnetic hysteresis in heterogeneous alloys,
Philos. Trans. R. Soc. London Ser. A 240:599-642.
Sugiura, N., 1980, Field dependence of blocking temperature of single-domain magnetite, Earth Planet.
Sci. Lett. 46:438-442.
Taylor, R M., and Schwertmann, U., 1974, Maghemite in soils and its origin. II. Maghemite synthesis
at ambient temperatures and pH 7, Clay Miner. 10:299-310.
Thompson, R, Bloemendal, J., Dearing, J. A., Oldfield, F., Rummery, T. A., Stober, J. C., and Turner,
G. M., 1980, Environmental applications of magnetic measurements, Science 207:481-486.
Verwey, E. J. W., and Haayman, P. W., 1941, Electronic conductivity and transition point in magnetite,
Physico The Hague) 8:979-987.
Wasilewski, P. J., 1973, Magnetic hysteresis in natural materials, Earth Planet. Sci. Lett. 20:67-72.
York, D., 1978, Magnetic blocking temperature. Earth Planet, Sci. Lett. 39:94-97.
Chapter 3
The Geomagnetic Field
Its Nature, History, and Biological
Relevance .
DURWARD D. SKILES
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
1.1. The Biological Relevance of the Geomagnetic Field. . . . . . . . . . . . . . . . . . . . . 43
1.2. Electric and Magnetic Fields: Some Important Concepts. . . . . . . . . . . . . . . . . . 45
2. The Main Geomagnetic Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.1. The Nature of the Present Field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2. The Dipolar Configuration of the Field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.3. The Origin of the Field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.4. The Nature of the Field through Geologic Time . . . . . . . . . . . . . . . . . . . . . . . 64
2.5. Paleo intensities and the Age of the Field . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.6. Reversals of the Dipole Field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.7. Polarity Transitions: What Happens during a Reversal? . . . . . . . . . . . . . . . . . . 80
2.8. What Causes Reversals? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
2.9. Field Reversals and Phylogenetic Change. . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3. The Field of External Origin. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.1. The Solar Wind and the Magnetosphere. . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.2. The Solar Quiet and Lunar Daily Variations . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.3. Magnetic Storms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
3.4. Geomagnetic Indices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
1. Introduction
1.1. The Biological Relevance of the Geomagnetic Field
In the past decade, it has become apparent that the importance of magnetic interactions
is not confined to the world of physics but extends into the realm of physiology and whole
organism biology. That this should be so follows from two elementary observations: (1)
by virtue of their magnetic moments and electrical charges, the atoms and ions which make
up an organism can magnetically interact with the organism's environment, and (2) the
environment of practically every organism includes a highly ordered and stable (relative
to the lifetime of the organism) geomagnetic field which contains both spatial and temporal
information of potential value to the organism.
DURWARD D. SKILES • Seismographic Station, University of California, Berkeley, California 94720.
43
44 Chapter 3
The significance of these observations is considerably enhanced by the fact that in all
probability the existence of the geomagnetic field preceded the origin of life on earth, so
that the major portion, if not the whole, of organic evolution has occurred in the presence
of the geomagnetic field. Furthermore, except during brief periods spanning field reversals,
the field configuration and intensity have seldom differed radically from those of the pres-
ent. Hence, it would indeed be surprising if the world were devoid of species capable of
detecting, and deriving selective advantage from, the geomagnetic field.
The recent convincing demonstrations that various organisms ranging from bacteria
to vertebrates exhibit behavioral responses to the geomagnetic field or to laboratory fields
of geomagnetic intensity have established that the geomagnetic field is both a perceptible
and a relevant component of the environment of those organisms. The discovery that some
of those same organisms, as well as others yet to be examined for magnetoreception, contain
microscopic particles of magnetite which may be intimately associated with neurons (e.g.,
Zoeger et at, 1981; Kuterbach et al., 1982) indicates that magnetoreceptive organs may
soon be identified. Such identification would establish the existence of direct organismic
magnetoreception, as opposed to indirect organismic magnetoreception via induced elec-
tric fields and electroreceptive organs. [Direct, nonsensory responses to magnetic fields
have been invoked to explain the behavior of magnetotactic bacteria. Frankel et al. (1979)
calculate that the torque on the magnetic moment of a magnetotactic bacterium is sufficient
to orient the bacterium in the geomagnetic field.]
Regardless of the mode of magnetoreception, it is now clear that the geomagnetic field
must be regarded as an environmental factor of potential importance to a taxonomically
diverse array of organisms. The following are some of the more obvious ways in which
an organism might exploit or be affected by the geomagnetic field. The relative stability
of the field over ecological (as opposed to evolutionary) time, coupled with its geographic
axial and equatorial symmetry, provides for both orientation and navigation. Moreover,
because the field is perturbed by local and regional variations in lithology and topography,
various topographic features such as mountains, islands, and coastlines, as well as certain
localities in otherwise featureless terrains, might be recognized by their magnetic signa-
tures.
Because ionospheric electrical currents are influenced both by solar radiation and by
solar and lunar tidal forces, the magnetic field at the earth's surface undergoes low-am-
plitude variations with periods of the solar and lunar days, the synodic month, and the
tropical year. Any of these periodic variations might function as a Zeitgeber (synchronizer)
for a biological clock or enable a sufficiently sensitive organsim to mark the passage of
time.
The occasional observation of magnetic earthquake precursors has led to speculation
that detection of such precursors might explain some of the reports of unusual animal
behavior prior to earthquakes. Such speculation is readily extended to include other cat-
astrophic geophysical events such as severe storms or volcanic eruptions.
Finally, it has been suggested that during a geomagnetic field reversal, extinction and
speciation rates should increase if the reduced field intensity results in climatic changes
or in increased energetic particle and radiation flux at the earth's surface. Indeed, rough
correlations between field reversals during the last few million years and the appearance
and disappearance of several planktonic species have been observed in deep-sea cores.
Alternatively, it has been suggested that both the field reversal and the faunal changes are
caused by a third event such as a meteor impact. To these must now be added the possibility
that extinctions and speciation events could occur if populations dependent upon the
geomagnetic field for spatial or temporal information were deprived of that information
for many generations during a reversal.
The extent to which these speculations are valid must be determined from future
biomagnetic research. It is the purpose of this chapter to present a description of the geo-
The Geomagnetic Field 45
magnetic field. both past and present. The emphasis will be on phenomenology. with the
aim of providing the reader with a basis for intelligently assessing the relevance of a given
geomagnetic phenomenon to a particular biological situation. Sufficient theory and meth-
odology will. however. be indudd to give the reader some feeling for our depth of un-
derstanding. or lack thereof. of various geomagnetic phenomena. As much interesting and
valuable detail must of necessity be omitted here. the reader will find additional practical
information in several books on geomagnetism. particularly those by Chapman and Bartels
(1940). Matsushita and Campbell (1967). McElhinny (1973). and Merrill and McElhinny
(1983). However. when critically investigating a particular biomagnetic problem. the reader
should consult the original geomagnetic literature. A word of caution-many of the graphs.
diagrams. and maps contained therein. having been constructed with the primary intent
of aiding geophysicists in understanding the origins of geomagnetic phenomena. are ideal-
ized renditions of highly fitered data which often depart considerably from what a mag-
netosensitive organism would experience.
1.2. Electric and Magnetic Fields: Some Important Concepts
One of the primary objectives of biomagnetic research is to ascertain the nature of

biological magnetoreceptors. As the title of this volume indicates, attention is currently
focused on biogenic materials that can be either permanently or inductively magnetized.
From a theoretical point of view. however. consideration should not be restricted to such
materials. Because the motion of any electric charge is influenced by magnetic fields.
intraorganismic ionic and electronic currents must be included as potential mechanisms
for magnetoreception. Moreover. biomagnetism must also be concerned with electrore-
ception. because a magnetic field whose direction or intensity is changing with time is
always accompanied by an electric field (Faraday's law) and because an organism moving
in a magnetic field "sees" an electric field as well as a magnetic field. Therefore, should
an organism be observed to exhibit a behavioral or physiological response to a magnetic
field, it is a nontrivial matter to inquire whether that organism is responding directly to
the magnetic field via a magnetoreceptor or indirectly via detection of an associated electric
field. This is of particular concern in light of the proven electrosensory capabilities of
various organisms (see, e.g., Kalmijn, 1974). Strictly speaking, a similar concern exists
regarding organismic responses to electric fields. However. in the low-frequency. low-
velocity.low-electric-field-intensity approximation appropriate to the natural environment
of biological systems, the magnetic fields associated with temporal changes in electric
fields are generally exceedingly weak. A notable exception occurs in the laboratory when
an organism is subjected to pulsed electric fields. In such cases, the low-frequency ap-
proximation is violated (even if the number of pulses per unit time is low) by the sudden,
on-off and off-on changes in the electric field, and substantial magnetic fields may be
present.
Because an understanding of the fundamental relationship between electric and mag-
netic fields is essential to a proper interpretation of biological responses to the geomagnetic
field. a review of some basic concepts is in order. First and foremost among these is that
electric and magnetic fields are neither independent nor absolute entities. Rather. they-are
components of one and the same electromagnetic field. components whose spatial and
temporal characteristics will appear different to two or more observers who are moving
relative to one another. It is therefore critical that each observer measure all physical
quantities with instruments which are in his own frame of reference. i.e .. with instruments
which are not moving with respect to him.
46 Chapter 3
A B B E
Figure 1. The force on a negative charge q in the vicinity of a wire carrying current I producing a
magnetic field B. (a) The wire is at rest relative to the observer and q moves with velicity V q • (b) The
charge q is at rest relative to the observer and the wire moves with velocity Vw = - Vq •
The electromagnetic force Fq on a charge q moving with velocity v in an electric field

E and a magnetic field B is
Fq = q(E + kv x B) (1)
where x denotes the vector or cross product and k is a constant which depends on the
system of units used. In mks, electromagnetic (emu), and electrostatic (esu) units, k = 1,
and in cgs units, k = l/c. The electric force on q is thus collinear with E and the magnetic,
or Lorentz, force is perpendicular to both v and B. Equation (1) is valid for all observers,
regardless of their relative motions. However, the magnitudes and directions of the vectors
Fq, E, v, and B will, in general, be different for different observers. Because of charge
invariance, q is the same for all observers, and changing the sign of q reverses the direction
of Fqo For biologically relevant velocities, the differences in Fq and B between observers
are negligible.
To understand why an observer moving in a magnetic field sees an electric field, let
us consider what happens to an electric charge q moving parallel to a long, straight, current-
carrying wire. We shall look at the problem in two ways-first from the point of view of
an observer at rest with respect to the wire (Fig. la), and then from the point of view of
an observer at rest with respect to the charge (Fig. Ib). As all electric fields ultimately
originate from electric charges and all magnetic fields ultimately originate from electric
currents (the magnetic fields of magnetized bodies originate from subatomic currents), this
problem is equivalent to that of comparing the electric and magnetic fields experienced
by a pigeon at rest on its roost with the fields experienced by a pigeon flying by the roost.
Consider a roosting pigeon which sees a stationary wire carrying a current I from right
to left as shown in Fig. la. The wire is uncharged, so there is no electric field around the
wire (E = 0). The current I produces a magnetic field B whose lines circle about the wire
in a clockwise direction when viewed in the direction of the current I. (A magnetic field
line is a curve whose direction at each point gives the direction of the magnetic field vector
at that point.)
Let us inject a negative charge q into the vicinity of the wire with a velocity Vq directed
parallel to the wire (Fig. la). The roosting pigeon detects no electric force on the charge,
but measures a magnetic force qVq x B directed toward the wire. Consider now a second
pigeon which is flying along with the charge q. The flying pigeon also detects the magnetic
field B but measures no magnetic force on q because to him q is at rest (Fig. 1b). Clearly,
however, both pigeons will observe that q is accelerating toward the wire. Hence, to the
flying pigeon the wire must appear positively charged, producing an electric field E = Vq
X B which attracts the negative charge q to the wire. Why is the wire charged? The answer
is quite simple. Electric currents must ultimately form closed loops, so the wire in Fig. 1
must be part of a larger closed circuit. To the flying pigeon, the wire near q is moving with
velocity Vw = - Vq in a magnetic field produced by the current on the far side of the circuit.
In fact, each side of the circuit is moving in a magnetic field produced by the current on
the opposite side. The resultant magnetic forces on the charges in the wire cause positive
charges to accumulate in the wire near q and negative charges to accumulate in the wire
on the far side of the circuit.
Some readers (who will not be the first to do so) might disagree with the flying pigeon's
point of view and reason as follows. Because the moving circuit carries its magnetic field
along with it, there is no relative motion of the magnetic field lines and the charges in the
wire. Hence, the pigeon's deduction that the charge q is acted upon by an electric field
produced by a positively charged wire is wrong. What actually occurs is that magnetic
field lines moving past q produce a force on q. However, experiment will not bear out this
line of reasoning. If each pigeon is given an electric field meter, the needle on the flying
pigeon's meter will deflect and that on the roosting pigeon's will not. The flying pigeon's
"imagined" electric field is quite real.
Our errant readers have misunderstood Eq. (1). The velocity Vq is the velocity of the
charge q with respect to a given observer, and Band E are simply the fields measured at
the charge q by instruments at rest with respect to that observer. Nothing is stated or implied
about the motion of the fields or their sources. In fact, electrodynamic theory does not
define a "moving field line" in an unambiguous way. Consider, for example, that although
the flying pigeon sees electric field lines, the roosting pigeon does not see those same field
lines moving past with velocity -Vq • Rather, the roosting pigeon sees no field lines at all
(E = 0). The beauty of the field-theoretic approach to electromagnetism is that a given
observer need know nothing about the motion or any other properties of the circuit pro-
ducing an electromagnetic field. The observer need only measure the fields and apply Eq.
(1). For a detailed discussion of the electric fields experienced by moving and stationary
organisms, the reader should consult Kalmijn (1974).
2. The Main Geomagnetic Field
2.1. The Nature of the Present Field

A crude magnetic compass was apparently known to the Chinese in the 2nd century
B.C. (Needham, 1962) and since that time man has made considerable use of the fact that
over much of the earth's surface a magnetic compass points roughly to geographic north.
Indeed, if one were to take a spherical compass whose needle is free to move both hori-
zontally and vertically to Addis Ababa, Ethiopia, the needle would align itself in the hor-
izontal plane with its north-seeking end pointing directly to geographic north. The region
around Addis Ababa is, however, one of only three or four in the world where a compass
would orient so perfectly. In general, the direction of the needle would vary from place
to place, deviating several degrees from true north and inclining several degrees to the
horizontal.
It is therefore useful to choose a geographically based coordinate system in which the
magnitude and direction of the geomagnetic field vector at any point can be conveniently
specified. (At a given point, the magnetic field fector points in the same direction as the
48 Chapter 3
ZENITH (-)
NORTH (+)
EAST (+)/
D: DECLINATION
H: HORIZONTAL INTENSITY
Z: VERTICAL INTENSITY
X: NORTH-SOUTH COMPONENT Figure 2. The total geomagnetic
Y: EAST - WEST COMPONENT intensity F and the geographic
F: TOTAL INTENSITY orientation of the magnetic ele-
I: INCLINATION ments D, 1, H, X, Y, and Z. From
Chernosky et al. (1965) with per-
mission.
north-seeking end of a spherical compass needle.) The one generally adopted is a rectan-
gular system whose positive axes are directed northward, eastward, and downward (Fig.
2). The magnetic field vector, generally represented by F or T, rather than B, in the de-
scriptive geomagnetic literature, can then be resolved into Cartesian components desig-
nated X (north), Y (east), and Z (vertical). The component of the field lying in the horizontal
plane is the horizontal intensity H which points in the direction of the needle of the familiar
mariner's or map compass. The angle between H and true north is the magnetic declination,
or variation, D, which is taken to be positive (negative) when H points to the east (west)
of north. The inclination I is the angle between F and H and is positive (negative) when
F is directed below (above) the horizontal. These magnetic elements are related by the
equations
(2)
X = H cos D, Y = H sin D, Z = F sin I (3)
H = F cos I (4)
where H is the horizontal, and F = T is the total, field intensity. The magnitude and
direction of F are thus uniquely determined by any three independent elements, for ex-
ample, (X, Y, Z), (F, I, D) , or (H, Z, D).
In the mks system, the intensity of the magnetic field B (generally called the magnetic
induction in the formal literature of physics) has traditionally been expressed in newtons
per ampere-meter or webers per square meter. In recent years, the weber per square meter
has been renamed the tesla. In the Gaussian, or cgs, system, the unit of magnetic induction
is the gauss (designated f). In the geomagnetic and biomagnetic literature, one frequently
encounters the oersted, the cgs unit of the magnetic intensity vector H (not to be confused
with the horizontal component H of the geomagnetic field), which is equivalent to an
ampere-turn per meter in mks units. However, for all practical geomagnetic purposes, the
oersted and the gauss are numerically equal. For very weak magnetic fields, the nanotesla
(nT) is used in the mks system and the gamma ('y) in the cgs system. These units of magnetic
induction are related by
1 gauss = 10- 4 tesla = 10- 4 weber/m 2

1 gamma = 10 - 5 gauss = 10 - 9 tesla = 1 nanotesla
The intensity of the geomagnetic field at the earth's surface ranges from about 0.24
gauss near Rio de Janeiro, Brazil, to between 0.61 and 0.68 gauss near the geographic poles.
By laboratory standards, the field is rather weak. Continuous fields with intensities up to
100,000 gauss are readily produced in the laboratory, and the field near a toy magnet is
on the order of 100 gauss. In contrast, Wikswo et a1. (1980) have detected a field of 0.12
'Y produced by the action potential of an isolated sciatic nerve of a bullfrog. The latter
value is of interest as it suggests a rough intraorganismic noise level which extraorganismic
fields must exceed if they are to be biologically detectable.
One of the primary objectives of applied geomagnetics is the construction of maps of
the field elements. Magnetic surveys have been made by mariners since at least the 1500s,
but worldwide surveys did not begin in earnest until the beginning of the present century.
The early measurements were restricted to declination and inclination, as meaningful
measurements of the intensity had to await the theoretical and technological developments
of the 19th century. Today, the field is continually surveyed from land, sea, air, and sat-
ellite, with the most precise measurements being made at some 100 permanent terrestrial
magnetic observatories scattered around the world. Detailed knowledge of the structure
and behavior of the field is thus confined to the last 80 years or so, and statements regarding
the global nature of the field prior to that time must be inferred from extremely fragmentary
data by applying a sort of geomagnetic uniformitarian principle known as the "axial geo-
centric dipole hypothesis."
Surveys of the field culminated in the World Magnetic Survey, an international effort
which assembled data acquired from 1961 to 1967 and produced the International Geo-
magnetic Reference Field (IGRF) for epoch 1965.0. Figures 3, 4, and 5 are maps of F, D,
and I for the 1965 IGRF.* [For a detailed discussion of the derivation of the 1965 IGRF,
see Zmuda (1971).]
With these maps in mind, let us make a few observations about the field. Over most
of the northern hemisphere, the magnetic field lines are directed downward (I> 0). Over
most of the southern hemisphere, the field lines are directed upward (I < 0). (See also Fig.
13.) The lines of equal inclination are called isoclinics or magnetic latitudes. The isoclinic
girdling the earth for which I = 0 is called the magnetic equator and the point at which
I = + 90° (I = - 90 0 ) is called the north (south) magnetic, or dip, pole. Note that the
magnetic poles neither coincide with the geographic poles nor are antipodal. For the 1965
field, the north magnetic pole was located well within the Arctic circle on Bathurst Island,
Canada, at approximately 75SN, 101°W, whereas the south magnetic pole was located
* Far more detailed maps are published by the U.S. Naval Oceanographic Office.
50 Chapter 3
TOTAL FIELD (<JOuss) F
..,. ...
IGRF 1965·0
Figure 3. Total geomagnetic intensity. IGRF 1965.0. From Zmuda (1971) with permission.
just outside the Antarctic circle on the Adelie Coast of Antarctica at approximately 65SS,
140.3°E.
Clearly, the geomagnetic field contains considerable orientational and navigational
information. Lines of equal field intensity, called isodynamics, trend east-west over much
of the earth and thus permit a rough determination of latitude within each hemisphere.
Isoclinics provide for an even better determination of geographic latitude, and the change
of the sign of I at the equator discriminates between the hemispheres. If one has a magnetic
compass, then knowledge of the declination at any point is tantamount to knowledge of
geographic north. Lines of equal declination, called isogonies, and the isoclinics form a
grid analogous to that of geographic latitude and longitude. Note that there are two agonic
(D = 0) lines (Fig. 4). In light of the numerous investigations of animal orientation and
migration that have been performed in Europe in the last few decades, it is interesting that
one agonic line runs through central Europe in an almost perfect north-south direction.
With the notable exceptions of central Eurasia and southern Africa, over much of the
world's land mass and northern oceans, the isogonics trend roughly north- south and might
therefore be used to determine geographic longitude without the aid of a clock.
Construction of magnetic field maps is a nontrivial task. In general, the data must be
culled and weighted according to their reliability, air and satellite measurements must be
converted to appropriate surface values, and local crustal anomalies, due primarily to
DECLINATION (dI!:9rus) D
IGRF 1965·0
Figure 4. Magnetic declination, IGRF 1965.0. From Zmuda (1971) with permission.
geological features in the upper few kilometers of the earth's crust, must be removed.
Furthermore, the values of the magnetic elements are not constant in time but drift over
the years, fluctuate with daily, monthly, and annual periodicities, and are occasionally
perturbed by magnetic storms. The perturbations must be removed, the periodic fluctua-
tions must be averaged over time, and data spanning several years much be reduced to
the same date.
The drift of the values of the magnetic elements, known as the secular variation, is
sufficiently rapid to necessitate the revision of geomagnetic maps every 5 to 10 years.
Classic examples of the secular variation are found in the geomagnetic records obtained
near London and Paris over the last few hundred years (Fig. 6). At those localities, the
declination has changed by as much as 18° and the inclination by as much as 6° in 100
years. Such rapid changes in the field direction have an obvious biological implication.
While the ability to orient with respect to the geomagnetic field may be genetically pro-
grammed, the absolute direction of the field is unlikely to be. Successive generations must
learn to associate the field direction with a particular geographic direction.
Figures 7, 8, and 9 are maps of the annual secular variations in F, D, and I for the
1965 IGRF. It is clear from the maps that the secular variation is a global characteristic of
the field and that the changes in D and I at London and Paris are fairly typical. From the
London-Paris curves, it is also evident that both the rate and the sign of the secular var-
iation change with time. The geomagnetic field is thus an extremely dynamic phenomenon.
52 Chapter 3
INCLINATION (dt<1us) I
IGRF 1965·0
Figure 5. Magnetic inclination, IGRF 1965.0. From Zmuda (1971) with permission.
Because the magnetic elements change with time, and because the number of meas-
urements used to generate magnetic maps is necessarily finite, geomagnetic maps cannot
be precisely accurate at each point on the globe. But if the effects of crustal anomalies are
neglected, the agreement of the maps with measured values is remarkably good. Cain and
Hendricks (1968) compared values of the magnetic elements D, H, and Z from the field
model GSFC (12/66) (which for practical purposes is identical to the 1965 IGRF) with the
1965 annual means measured at the permanent magnetic observatories and found that the
values of D generally agreed within a few minutes and that the values of Hand Z often
agreed within a few tens of gammas and seldom differed by more than 100 'Y or so. Examples
of worst cases are a 550 'Y (- 2.5%) discrepancy in H at Mauritius in the Indian Ocean,
and a map value of about 2° for D as opposed to the observatory value of 0° at Tromso,
Norway. Such discrepancies are presumed to be due to local magnetic anomalies located
in the earth's crust.
Of course, care is taken not to locate magnetic observatories near strong crustal an-
omalies. Hence, the worst cases cited above are far from the worst cases one might en-
counter elsewhere. Anomalies from certain are bodies are sufficiently strong to render a
magnetic compass useless. For example, Chapman and Bartels (1940) noted that a magnetite
deposit near the Finnish island of Jussaro increases the vertical intensity by 0.63 gauss
and the resuling disturbance in D has caused many shipwrecks. One of the world's most
1541 •.ú ú •=•...•• !.580

1603 ..•...•..••·)
PARIS (... g ú f l I N S N U =
1780 1755 1730 1710 1695 1680 1662 ;630 :
_·_·-· __ •_ _· - - ·..i67i·_ .... ·1642· ......1620···
19!3 ........ 1967
65
D?K=Wú WJ ? J ? K ? =I''''''
ú
(J)
u.J
u.J
a:
'"
u.J
0
HARTLAND (LONDON)
"': f ú K> K=
z
Figure 6. Secular variation ú = 70 1805\ 1786
>-
of D and I recorded near <t
Z NT DRú = \ . ',780 1754
::; 1787\... ••.• • .• /1576
London and Paris over the U
z 1775·.... '7i2"
last 400 years. The solid " ...DI WWKYW>ú>KSN=................... ’ ’ ÑäúXPF=
curves represent continu-
'-. -. "1671
ous observatory records, 75
and the dashed curves are LONDON 11540) 1576
based upon historical val- NUMRJ ä à ú .=.... ä >Kú ú =......> > ú ú =.. ú ú ú ú =_. _........ ___ Kf > \ ú | K=ú ú ú ú .= ....ú ú > =___ . _ú ú | ú =.... ú ú => > ú ú > =ú ú =ú =ú WX=
1787 1775 1748 1665 1622 1580
ues given by Gaibar-Puer-
tas (1953). From Skiles -20 -10 o 10
(1970) with permission. DECLINATION IN DEGREES
intense anomalies occurs in Lapland where Z is increased to 3.6 gauss. The great anomaly
at Kursk, some 450 km south of Moscow, consists of two narrow features about 250 km
long and 60 km apart. D ranges from + 60° to -110°, and at the peak of the anomaly, Z
rises to 1.9 gauss (Fig. 10).
Magnetic anomalies are the result of the permanent and induced natural magnetization
of rock units. The intensity of the anomaly depends on the size and shape of the unit, its
depth below the surface, and the proportion of magnetic minerals it contains. Owing to
the presence of magnetic minerals in practically every rock type, the spectrum of magnetic
anomalies ranges from the strong but rarely encountered ones just described to a pattern
of low-intensity (100-1000 'Y) anomalies which blankets the earth. Figure 11 is a repre-
sentative example of magnetic topography obtained from an aeromagnetic survey. Typi-
cally, such surveys are flown at altitudes of a few hundred to 1000 m or so. At the surface,
the pattern would appear more intense and complex. Figure 12 shows the intensity an-
omalies recorded from ship surveys in the Gulf of Alaska. Often exceeding 1000 'Y, the
anomalies are typical of the magnetic background noise a migrating fish or bird would
encounter. Walcott (1978) has observed that the orientation of homing pigeons is perturbed
in the vicinity of slightly more intense, but much steeper, anomalies (3000-4000 'Y in 3
km). However, such anomalies might prove useful to a wide-ranging species which could
learn to recognize a particular region by its magnetic signature.
The ubiquity of magnetic anomalies has obvious implications for biomagnetic re-
search. If precise knowledge of the geomagnetic field at a particular locality is important,
the investigator should not rely on maps of the global field. Preferably, the investigator
should measure the field at the point of interest. If that is impractical, then local magnetic
maps from which anomalies have not been removed should be consulted.
2.2. The Dipolar Configuration of the Field

Because isolated magnetic charges (monopoles) do not exist (or are so rare or elusive
as to have escaped detection), the simplest form of a magnetic field is that of a dipole, the
54 Chapter 3
SECULAR CHANGE OF TOTAL FIELD (9ammas / y(ar) F
IGRF 1965·0
Figure 7. Secular variation in F, IGRF 1965.0. From Zmuda (1971) with permission.
most familiar example being the field of a bar magnet. To a fairly good approximation, the
configuration of the present geomagnetic field is that of an axial geocentric dipole, a dipole
located at the center of the earth and aligned with the earth's axis of rotation. A better
approximation is obtained if the dipole axis is inclined about 11 0 to the geographic axis
toward the 70 0 west meridian (Fig. 13). The best approximation is obtained if the inclined
dipole is displaced from the earth's center by a few hundred kilometers. There is, however,
no physical reality to the source dipole. It is merely an easily visualized way of representing
the gross configuration of the geomagnetic field .
For analytical purposes, it is most convenient to deal with a strictly geocentric dipole,
for such a dipole produces a field which is highly symmetric on the surface of the globe.
The best fitting central dipole is obtained as follows. In a region such as the earth's at-
mosphere near the ground, where there are no electric currents or magnetized matter, the
magnetic field satisfies the equation
vx B o (5)
and is derivable from a potential V,
B -VV (6)
SECULAR CHANGE OF DECLINATION (mlOut\!:s/yltor) D
IGRF 1965·0
Figure 8. Secular variation in D, IGRF 1965.0. From Zmuda (1971) with permission.
Because magnetic monopoles do not exist,
V· B = 0 (7)
and the magnetic potential satisfies Laplace's equation
V2 V =0 (8)
In a system of spherical coordinates where r is the distance from the origin, 6 is the col-
atitude, and <I> is the longitude, the solution to (8) can be written as an infinite series of
spherical harmonics:
n= l m =O
Here, P::'(6) are associated Legendre functions, HE is the radius of the earth. and
g::'. h::'. c::'. s::' are constants known as Gauss coefficients. The terms proportional to ú =
56 Chapter 3
SECULAR CHANGE OF INCLINATION (minutes/year)

.
I
IGRF 1965 ·0
Figure 9. Secular variation in T, IGRF 1965.0. From Zmuda (1971) with permission.
represent magnetic field sources outside the earth, and the terms proportional to (lIr)"+l
represent field sources inside the earth.
The potential function V is not directly measurable, but by applying Eq. (6) we can
generate expressions for the magnetic elements X, Y, and Z. If the spherical harmonic
series is truncated to finite values of nand m (generally n = m :s 12), then a regression
analysis can be performed to determine the values of the Gauss coefficients which provide
the best fit to the tens of thousands of values of the magnetic elements obtained from a
worldwide magnetic survey. When this is done, it is found that sources external to the
earth contribute less than 1% to the field at the earth's surface. (During a magnetic storm,
the contribution may be several percent, but as we have noted, such transient contributions
are eliminated from survey measurements.) Thus, as Gauss demonstrated over a century
ago, the geomagnetic field originates primarily from within the earth. The field of internal
origin is thus called the main geomagnetic field.
For the 1965 IGRF, the harmonic series was truncated to 80 terms, n = m = 8. The
Gauss coefficients of the main field are given in Table I. * It is from those coefficients that
* Coefficients for the 1970, 1975, and 1980 reference fields are given in Eos , Trans. Am. Geophys.
Union 62:1169 (1981).
/Jed/naHon flor/zonfallnfensity Vertic a/Intensify

SO' 55' .so'
oú KJ
= J ú ú J J ú ú J J
5 kl17.-
ú K =
Figure 10. Isomagnetic maps of the northern part of the Kursk anomaly. From Chapman and Bartels
(1940) with permission.
Figure 11. Magnetic anomaly map of the southern coast of England. From Hahn (1971) with permission.
58 Chapter 3
BRITISH
COLUMBIA
·00
Figure 12. Map of the northeast Pacific showing magnetic anomalies plotted along survey ship tracks.
The black portions represent positive anomalies. The magnitude of the large positive anomaly near
the extreme left of track HI is approximately 1500'Y. From Pitman and Hayes (1968) with permission.
the maps shown in Figs. 3-5 are generated. Also shown in Table I are the time rates of
change ú =and 11;:> of the Gauss coefficients. They are used to generate the maps of the
secular variation field shown in Figs. 7-9.
Ö ú K = which is by far the dominant coefficient. corresponds to an axial geocentric dipole
directed from north to south. Ö ú = and hl correspond to geocentric dipoles lying in the equa-
d b ú d ú b ä y ` = EQUAlOR
GEOGRAPHIC EQUATOR
Figure 13. Magnetic lines of force of an inclined geocentric dipole. N denotes the north geomagnetic
pole. S the south geomagnetic pole. and G the geographic axis.
TABLE I. IGRF 1965.0 Coefficientsa

Main field Secular change
(gammas) (gammas/year)
n m Öú = Üú = Öú = Üú =
1
1 °
1
-30,339
-2,123 5758
15.3
8.7 -2.3
2
2 °
1
-1,654
2,994 -2006
-24.4
0.3 -11.8
2 2 1,567 130 -1.6 -16.7
3
3 °
1
1,297
- 2,036 -403
0.2
-10.8 4.2
3 2 1,289 242 0.7 0.7
3 3 843 -176 -3.8 -7.7
4
4 °
1
958
805 149
-0.7
0.2 -0.1
4 2 492 -280 -3.0 1.6
4 3 -392 8 -0.1 2.9
4 4 256 -265 -2.1 -4.2
5
5 °
1
-223
357 16
1.9
1.1 2.3
5 2 246 125 2.9 1.7
5 3 -26 -123 0.6 -2.4
5 4 -161 -107 0.0 0.8
5 5 -51 77 1.3 -0.3
6
6 °
1
47
60 -14
-0.1
-0.3 -0.9
6 2 4 106 1.1 -0.4
6 3 -229 68 1.9 2.0
6 4 3 -32 -0.4 -1.1
6 5 -4 -10 -0.4 0.1
6 6 -112 -13 -0.2 0.9
7
7 °
1 -54
71
-57
-0.5
-0.3 -1.1
7
7
2
3 12° -27
-8
-0.7
-0.5
0.3
0.4
7 4 -25 9 0.3 0.2
7 5 -9 23 0.0 0.4
7 6 13 -19 -0.2 0.2
7 7 -2 -17 -0.6 0.3
8
8 °
1
10
9 3
0.1
0.4 0.1
8 2 -3 -13 0.6 -0.2
8 3 -12 5 0.0 -0.3
8 4 -4 -17 0.0 -0.2
8 5 7 4 -0.1 -0.3
8 6 -5 22 0.3 -0.4
8 7 12 -3 -0.3 -0.3
8 8 6 -16 -0.5 -0.3
a From J. Geophys. Res. 74:4407-4408 (1969).

60 Chapter 3
N
Geomagnetic
north pole
II ú ç =
Magnetic equator (!:..=.Q)

,/
--
Figure 14. Schematic illustration of

the distinctions between geo-
graphic, geomagnetic, and magnetic
s Geomagnetic poles and equators. From Mc-
(Geographic pale) sputh pole Elhinny (1973) with permission.
torial plane and are responsible for the inclination of the dipole from the earth's rotation
axis. The geocentric dipole of the 1965 IGRF has a magnetic moment M = o t Ö ú F O = +
(gU 2 + (hi)2p/2 = 8.01 X 10 25 gauss cm 3 • Because gg, g1, Ü ú = can be made to vanish by a
suitable choice of the origin of the coordinate system (r,e,<I», they can be viewed as pro-
ducing the eccentricity of the best-fitting inclined dipole. The remaining coefficients cor-
respond to weaker sources which become more eccentric as the value of n increases. How-
ever, it must be understood that individual terms or groups of terms in the harmonic series
do not correspond to individual, physically identifiable sources.
The justification of the earlier statement that the geomagnetic field is, to a good ap-
proximation, that of an inclined geocentric dipole is now clear. For the 1965 IGRF, the
axis of the best-fitting geocentric dipole (known as the geomagnetic axis) intersects the
earth's surface at the antipodal points 78.6°N, 69.8°W and 78.6°S, 110.ZoE, which are known
as the geomagnetic north and south poles, respectively. The geomagnetic poles should not
be confused with the north and south magnetic poles. The former have no physical reality,
whereas the latter can be located with a magnetic compass.
For theoretical purposes, the geomagnetic axis is used to define a system of geomag-
netic latitudes and longitudes. The great circle which is perpendicular to the geomagnetic
axis is the geomagnetic equator, the line of zero inclination of the main dipole field. Figure
14 illustrates the distinctions between the geomagnetic, magnetic, and geographic coor-
dinate systems.
If the main dipole field is subtracted from the observed global field, there remains a
residual field referred to as the nondipole field. Maps of the nondipole field are generated
from the Gauss coefficients for which n 2: Z. Figure 15, a map of the vertical component
of the nondipole field for epoch 1965, shows that the nondipole field is not completely
chaotic, but consists of a few anomalies of continental size. As might be expected, the
Figure 15. Vertical component of the nondipole field , epoch 1965.0 in units of 10 - 2 gauss. From
McDonald and Gunst (1967) with permission.
nondipole field changes with time, one of the more interesting recent changes being the
apparent tendency of the field to drift westward at about 0.2°/year (see e.g., Skiles, 1970).
2.3. The Origin of the Field
While a detailed knowledge of the magnetohydrodynamic processes which are pre-

sumed to be responsible for the geomagnetic field is not essential to the biomagnetic in-
vestigator, a general understanding of the origin of the field provides a valuable per-
spective, particularly for evaluating the evolutionary implications of geomagnetic
phenomena As we have seen, mathematical analysis establishes unequivocally that the
Ö É ç ã ~ Ö å É ú =Ie field originates almost entirely within the earth. It is now generally agreed
that the earth's fluid iron core is the location of the sources of the main field , so let us
first consider why that location is plausible. I use "plausible" because the earth's interior
beyond a depth of a few kilometers is not accessible to direct observation. Knowledge of
the deep interior is inferential, derived from mathematical models and consistency ar-
guments which are constrained by measurements made at the earth's surface.
Noting the similarity of the earth's field to that of a uniformly magnetized sphere,
Gilbert (1600) stated, "Magnus magnes ipse est globus terrestris." Despite the brilliance
of that bold assertion, we know the facts do not support it. To produce the main dipole
field, the entire earth would have to be uniformly magnetized with an intensity of about
0.075 gauss (Jacobs, 1963). However, below a depth of 25 km or so, the temperatures within
the earth exceed the laboratory temperature (the Curie temperature) at which iron can be
62 Chapter 3
permanently magnetized. Therefore, unless, contrary to the predictions of both theoretical

and experimental high-pressure physics, the pressures within the earth produce a form of
matter which can be magnetized at high temperatures, any magnetized material must be
confined to the earth's outer 25 km. That requires a uniform magnetization in excess of 6
gauss to produce the geomagnetic field. Both of these values of magnetization are far greater
than those ordinarily found in crustal rocks. The intensity of magnetization of iron ore
occasionally exceeds 1 gauss, but the intensity usually encountered is around 0.03 gauss.
In addition, the non-ore-bearing rocks which constitute the bulk of the earth's crust have
magnetizations which average less than 100 'Y (Hahn, 1971).
A second argument against a permanently magnetized source of the field is that such
a source could not account for the secular variation. For example, a westward drift rate
of 0.2°/year for the nondipole field implies crustal velocities of more than 20 km/year,
which is several orders of magnitude greater than the 2-5 cm/year inferred from sea-floor
spreading and continental drift.
A number of modern hypotheses of the origin of the field have been proposed. The
mechanisms include electromagnetic induction of the field by magnetic storms, thermoe-
lectric currents resulting from chemical and thermal inhomogeneities within the earth,
and an intrinsic magnetic moment resulting from the rotation of a massive body. The only
hypothesis to have survived both theoretical and experimental challenges is that of the
geodynamo, whose central assumption is that the earth's deep interior is an electrically
conducting liquid. The existence of a fluid core of some 3500-km radius is inferred from
the fact that below a depth of about 2900 km the velocity of seismic pressure, or P, waves
suddenly drops by 40% and seismic shear, or S, waves cease to propagate. The high density
that must obtain in the core in order to account for the mass of the earth, coupled with
geochemical arguments, indicates that the core is primarily iron and hence a good electrical
conductor.
Originally advanced by Elsasser (1946) and Bullard (1949), the dynamo theory of the
origin of the field is now generally accepted and its development is one of the major aims
of geophysics. In essence, the theory holds that motions in the core, probably driven by
thermal or chemical convection, amplify any electric currents present to the point that an
equilibrium is reached between the generation of those currents and their natural ohmic
decay. The rotation of the earth imposes powerful constraints on the fluid motions so that
the resulting currents produce a magnetic field which is roughly symmetric about the axis
of rotation. The approximate alignment of the dipole axis with the geographic axis is
therefore not accidental, but is a consequence of dynamo processes in a rotating system.
The dynamo theory of the origin of the geomagnetic field is discussed in detail and with
a mathematical clarity that will appeal to the nonspecialist in recent reviews by Busse
(1978, 1980, 1983). Here, a brief qualitative discussion is sufficient.
While the details of dynamo theory have yet to be worked out, the feasibility of geo-
dynamo has been demonstrated theoretically and experimentally. An intuitive grasp of the
dynamo process is provided by the Faraday disk dynamo. Consider an electrically con-
ducting disk rotating in the presence of a magnetic field as shown in Fig. 16. The Lorentz
force qV x B on the charges in the disk causes the periphery of the disk to become positively
charged and the axis negatively charged. If a stationary circuit with sliding contacts con-
nects the axis with the periphery, a current j will flow, producing a secondary magnetic
field. If the circuit is coiled so that the current flows through the coil in the same sense
as the sense of rotation of the disk, the secondary magnetic field will reinforce the original
field. If the radius of the disk is sufficiently large, or the rotation of the disk sufficiently
rapid, the source of the original field can be removed and the secondary field will increase
to an equilibrium point at which its rate of regeneration is equal to its rate of decay due
to ohmic dissipation. For a fluid iron sphere the size of the earth's core, the decay time
Figure 16. A disk dynamo rotating with angular

velocity n and producing a current j and magnetic
field B. From Busse (1980) with permission.
is several thousand years. Hence, it is at least plausible that the slow rotation of the earth
is sufficient to produce a self-sustaining dynamo.
At first glance, the disk dynamo might appear to contradict the laws of thermody-
namics. However, even if the disk were supported by frictionless bearings, it would still
experience a force proportional to j x B which opposed the rotation. Thus, energy must
continually be supplied to the disk to keep it rotating, with the result that the kinetic
energy of rotation is converted to electromagnetic energy and thence to heat via ohmic
dissipation. The energy source for the geodynamo may be thermal, either the heat from
the decay of radioactive elements in the core or as originally suggested by Verhoogen
(1961), the latent heat released as the liquid outer core slowly freezes to form a solid inner
core. The energy source currently favored by most geophysicists is the gravitational energy
released as light elements released by crystallization at the inner core-outer core boundary
are buoyed upward through the outer core (Stevenson, 1981).
The disk dynamo has two very interesting features. First, it produces a dipole field at
large distances. Second, if the sense of rotation of the disk and the sense of winding of
the coil are unchanged, the system will also work as a dynamo for a field whose direction
is the reverse of that shown in Fig. 16. This implies that the present direction of the earth's
main dipole field might just as well be reversed. As we shall soon see, there is abundant
evidence that the dipole field has indeed reversed direction many times over the course
of geological history.
Unfortunately, when applied to the earth's core, the disk dynamo encounters a fatal
problem-it is topologically distinct from the core. Whereas the core is a simply connected
body, the disk dynamo is like a donut-it has a hole in it. Any attempt to render the
Faraday dynamo simply connected abolishes its dynamo action. However, this difficulty
was resolved when Herzenberg (1958) proved theoretically, and Lowes and Wilkinson
(1963) demonstrated experimentally, that a simply connected dynamo is possible.
The Lowes-Wilkinson laboratory dynamo consists of two electrically conducting, ro-
tating cylinders oriented perpendicular to one another and immersed in a conducting me-
dium. The arrangement results in positive feedback in which the current in each cylinder
is induced by the magnetic field produced by the current in the other cylinder. Remarkably,
64 Chapter 3
the laboratory model even exhibits field "reversals"-at unpredictable intervals the dy-
namo fields become unstable and after a few seconds assume new directions.
While the two-cylinder dynamo satisfies the topological constraint of simple con-
nectedness, it does not produce a dipole field. Nor does its geometry correspond in any
obvious way to the geometry of the core. But without too great a stretch of the imagination,
one can visualize an analogous situation in which large-scale eddies in the core inductively
feed upon one another's fields to produce a self-exciting geodynamo. However, the analogy
is at best heuristic, and the exact nature of the origin of the geomagnetic field remains one
of the great unsolved problems of physics.
In a very qualitative way, the geodynamo can be viewed as an orderly process pro-
ducing a highly ordered dipole field, and the nondipole field, whose Gauss coefficients
are roughly 10% of the dipole coefficients, can be viewed as resulting from perturbations
of that orderly process, perhaps by large-scale eddies in the core. The secular variation
may thus be the result of the translational motion of those eddies (recall the westward
drift), changes in the intensity of motion within eddies, or the creation of new eddies and
the decay of existing eddies.
However, this idealized view may be misleading. Although the dipole term certainly
dominates Eq. (9), the magnetic field in the core is distinctly less ordered than the field
at the earth's surface. Whereas the dipole field drops off as N L ú I = the nondipole field drops
off at least as l/r4. Hence, at a radius of RE/2, corresponding to the outer portion of the
core, ratio of the nondipole terms to the dipole terms increases by a factor of 2 for the n
= 2 terms, by 4 for the n = 3 terms, etc. The dipolar configuration of the field at the
earth's surface is thus due in part to the geometric attenuation of the nondipole field. The
geodynamo may well result from a process which tends to be rather disorderly, perhaps
turbulent, but which has a certain global order imposed upon it by inertial forces due to
the earth's rotation. Whatever its origin, the dynamo is an unsteady process. From Table
I it is clear that the intensity of the geocentric dipole is presently decreasing at a rate which
could result in the demise of the dipole in about 2000 years.
A question which must inevitably be asked of any geomagnetic dynamo is, what was
the source of the original embryonic field which initiated dynamo action? The answer will
probably be forever unknown. Suffice it to say that practically any stray field would have
served the purpose, and of those there were undoubtedly many. Fields carried by the solar
wind and fields from thermoelectric currents produced by thermal and chemical inhom-
ogeneities within the earth are two prime candidates.
A more important question is the age of the geomagnetic field. If the field is indeed
of dynamo origin, then it is reasonable to suppose that it is as old as the earth's fluid core,
provided, of course, that a sufficient energy source was also present. Elsasser (1963) sug-
gested that core formation was not complete until about 3 b.y. B.P., but the present view is
that the core formed some 4.5 b.y. B.P. (Ringwood, 1979; Stevenson, 1981). More direct
evidence, in the form of a "fossil" magnetization of some of the oldest known rocks, in-
dicates that the field is at least 3.5 b.y. old (McElhinny and Senanayke, 1980). The potential
importance of these dates to organic evolution is clear. The geomagnetic field was probably
present at the time of biogenesis and certainly antedates the earliest time, 1.4 ± 0.1 b.y.
B.P. (Schopf and Oehler, 1976), suggested for the appearance of the eukaryotes.
2.4. The Nature of the Field through Geologic Time

Having observed that theoretical considerations indicate the age of the geomagnetic
field is comparable to the age of the earth, we shall now survey the experimental evidence
concerning the existence, the configuration, and the behavior of the field through geologic
time-the paleomagnetic record. Just as the major features and some of the details of
organic evolution can be inferred from the fossil record, so can the history of the geo-
magnetic field be inferred from fossil magnetism.
As they form, many rocks acquire a natural remanent magnetization (NRM), a weak
magnetization which is generally parallel to, and in the same direction as, the geomagnetic
field vector at the place and time of formation. Frequently, the NRM is quite "hard," i.e.,
stable and permanent, and provides a record of the geomagnetic field which persists in-
definitely. By measuring the NRM of rocks whose orientation at the time of formation can
be established, the direction and intensity of the paleomagnetic field can be determined.
The NRM resides in magnetic minerals, primarily oxides of iron and titanium, which
almost inevitably comprise at least a small percentage of any rock. Because events sub-
sequent to the 'formation of the rock, such as lightning strikes or chemical changes, may
alter or eliminate the original NRM, not all rocks possessing an NRM are useful for pa-
leomagnetic studies. The procedures for identifying, and in some cases removing, contam-
inating secondary remanences are well developed (secondary NRM is often quite "soft"
and easily removed by thermal or magnetic cleaning) and paleomagnetism now forms one
of the major branches of geophysics.
The physics of rock magnetism and the techniques used to measure very weak re-
manent magnetizations are discussed elsewhere in this volume. Here, it will suffice simply
to identify the kinds of NRM which provide the bulk of paleomagnetic data. As magma
cools and crystallizes to form igneous rocks, either as a lava flow on the earth's surface
or as a plutonic intrusion in a subsurface chamber, magnetic minerals pass through their
respective Curie temperatures (578°C for magnetite, Fe304) and acquire a thermoremanent
magnetization (TRM). Particularly in the case of plutonic bodies, a cooling front may sweep
slowly through the rock, leaving behind a continuous record of the field from which in-
ferences about the paleosecular variation may be drawn. In unconsolidated aquatic and
marine sediments, the magnetic moments of fine-grained magnetic materials tend to be-
come aligned with the ambient magnetic field so that the subsequent compacted sediments
possess a detrital, or depositional, remanent magnetization (DRM). Sedimentary cores from
lakes and the deep sea often provide continuous records of the paleomagnetic field span-
ning as much as a few million years. It is interesting to note that Kirschvink and Lowenstam
(1979) point out that a significant fraction of the DRM of oceanic sediments may be due
to biogenic magnetite, primarily from chiton teeth.
If a magnetic mineral is formed in the presence of a magnetic field at "room" tem-
peratures, it will acquire a chemical remanent magnetization (CRM) if the mineral grains
grow to exceed a critical size, the blocking diameter. For example, CRM can be produced
in sediments as magnetic grains grow by accretion, or in various rock types when iron
compounds are oxidized by weathering. Presumably, it is a CRM acquired during the pro-
cess of in vivo biomineralization that accounts for much of the magnetic remanence which
has been observed in various organisms ranging from Å Ü ú í ç å ë = (Kirschvink and Lowenstam,
1979) and bees (Gould et aI., 1978) to pigeons (Walcott et a1., 1979) and dolphins (Zoeger
et a1., 1981). To distinguish the NRM acquired by a living organism by whatever process
from the NRM of the abiotic environment, the term biogenic remanent magnetization or
bioremanent magnetization (BRM) is appropriate.
While the study of rocks which have acquired a remanent magnetization by natural
means provides the majority of paleomagnetic data, a relatively small but quite interesting
array of data has been obtained from archeological objects such as bricks, pottery, kilns,
and hearths which have been subjected to the heat of man-made fires. The latter, termed
archeomagnetic data, often have a precise date attached to them and thus provide a valuable
record of the geomagnetic field over the past several thousand years.
The impossibility of obtaining worldwide contemporaneous sets of paleomagnetic data
is obvious. Rock units such as lava flows are produced by events which are isolated in
both space and time. Sediments may provide continuous spans of data which overlap with
66 Chapter 3
similar data elsewhere. but the sedimentary data obtained thus far are hardly worldwide
in distribution. In addition. with but a few possible exceptions. such data cannot be dated
and correlated with sufficient accuracy to eliminate the dispersion due to the geomagnetic
secular variation. What is required. therefore. is a unifying concept which converts sparse
data into a synoptic model of the field appropriate to a particular geologic era. Such a
model would allow intelligent interpretation of noncontemporaneous paleomagnetic data
from different localities and. in particular. would provide a comparison of a paleomagnetic
field with the field of the present.
The appropriate model is inferred from both theoretical considerations and experi-
mental observations. First. although the geodynamo is a time-dependent phenomenon. it
is unlikely that the general nature of the dynamo process changes with time. Hence. the
general nature of the geomagnetic field should not change with time. Second. the hard
component of the NRM of rocks formed during the last few million years is almost in-
variably approximately parallel or anti parallel to the present geomagnetic field. As this is
a worldwide phenomenon. it implies the general configuration of the geomagnetic field
has not changed during the past few million years. These observations lead directly to the
dipole hypothesis. which holds that the geomagnetic field has always (except. possibly.
during reversals) closely approximated the field of a geocentric dipole. According to this
hypothesis. paleomagnetic data are to be treated as if they were produced by a perfect
dipolar field.
As the configuration and intensity of a dipole field are uniquely determined by meas-
urements of the magnetic elements at a single point. the dipole hypothesis provides for
the direct comparison of paleomagnetic data from different localities. At a distance HE from
a geocentric dipole directed from geomagnetic north to south. the magnetic elements Z.
H. and I are given by
Z = 2M cos V L e ú = (10)
H = M sin V N o ú = (11)
tan I = ZIH = 2 cot 9 (12)
where M is the dipole moment and 9 is the dipole or geomagnetic colatitude. i.e .• the angle
between the field point and the geomagnetic north pole. For a dipole field. H points directly
to the north geomagnetic pole. so that the magnetic inclination at any given point uniquely
determines the location of the geomagnetic north pole via Eq. (12). The intensity of the
field then determines the dipole moment M via Eq. (10) or (11).
Measurement of the direction of the NRM of a rock sample from a given locality gives
the direction. and hence the inclination I. of the paleomagnetic field at the time of formation
of the rock. Via Eq. (12). that value of I determines the location of a virtual geomagnetic
pole (VGP). the geomagnetic north pole appropriate to the hypothetical geocentric dipole
which could have produced the measured NRM. Similarly. a determination of the paleo-
magnetic field intensity establishes the magnitude of the virtual dipole moment of the
paleomagnetic field via Eq. (10) or (11). However. owing to the non dipole component of
a paleomagnetic field. a VGP only approximates the position of the north geomagnetic
pole that would have been obtained had we been able to perform a global survey of the
paleomagnetic field. In general. VGPs obtained from contemporaneous paleomagnetic data
from different localities will not coincide. but will be scattered by an amount proportional
to the nondipole component of the paleofield.
In the case of the present-day geomagnetic field. the direction of the magnetic field
vector at any point differs from the direction of the best-fitting inclined geocentric dipole
field by no more than about 25° (except near strong magnetic anomalies). so that a VGP
obtained from almost any point on the globe is a fairly reasonable approximation to the
Figure 17. Virtual geomagnetic poles calculated from the magnetic declination and inclination at
modern geomagnetic observatories of worldwide distribution. The mean VGP position ( + ) coincides
almost exactly with the geomagnetic north pole. From Doell and Cox (1961) with permission.
geomagnetic north pole. The effect of the present-day nondipole field is illustrated in Fig.
17, which shows the VGP positions determined from the field directions at several magnetic
observatories. All of the VGPs fall within about 20° of the north geomagnetic pole, and
the mean of the VGP positions coincides with the geomagnetic pole.
Because of the secular variation, Fig. 17 is similar to what one might expect to obtain
from measurements of the field direction at a single locality spanning, perhaps, several
centuries. However, owing to the regularity of the variation of D and I at a given locality
(see Fig. 6), the VGPs would not be scattered quite so irregularly, but would tend to lie
on smooth curves. The mean position of the VGPs would represent a mean geomagnetic
pole for the period of time covered. Based on surveys of the field made in this century, it
appears that on a time scale of centuries, the positions of the geomagnetic poles change
only slightly. Cain and Hendricks (1968) calculate that the position of the north magnetic
68 Chapter 3
pole varied by only 0.3° in latitude and lS in longitude between 1900 and 1965. During
that period the predominant motion of the pole was a westward drift averaging about 0.015°/
year. At that rate, the geomagnetic pole would circle the geographic pole in 24,000 years.
Hence, over the long term, the mean position of the geomagnetic pole should tend to
coincide with the geographic pole even though the instantaneous geomagnetic pole might
never coincide with the geographic pole. This observation has led to the axial geocentric
dipole hypothesis, which states that when averaged over sufficiently long periods of time,
perhaps 104 or 105 years, the dipole axis coincides with the geographic axis. From the
point of view of dynamo theory, this hypothesis is quite reasonable, for even though the
inclination of the dipole is probably a persistent feature of the field [possibly owing to the
different rates of precession of the core and the mantle, as suggested by Malkus (1968)],
there is no reason to suppose that the dipole should be tilted in any particular fixed di-
rection.
In a typical paleomagnetic study, many rock samples are obtained at a given locality.
These may traverse a succession of lava flows spanning thousands of years, or be taken
from scattered points in a pluton which took thousands of years to cool. Owing to the
secular variation, the corresponding array of VGPs will be scattered (see Fig. 17). But
according to the axial dipole hypothesis, the mean position of the VGPs, called the pa-
leomagnetic pole, will coincide with the paleogeographic pole.
The axial geocentric dipole hypothesis has been tested in two fundamentally different
ways. One compares the paleolatitude for a particular geological unit inferred from pa-
leoclimatic data, such as the latitudinal distribution of fossil coral reefs, with the paleo-
magnetic latitude for that unit. The agreement is reasonably good. (See, e.g., Briden and
Irving, 1964.) The other, and more precise, method checks the internal consistency of
paleomagnetic data from various localities.
Opdyke and Henry (1969) have confirmed the axial dipole hypothesis with paleo-
magnetic results from deep-sea cores covering the last 2.5 m.y. Because only the vertical
orientation of a deep-sea core is known, absolute determinations of paleomagnetic dec-
lination are not possible and virtual pole positions cannot be calculated. Opdyke and Henry
therefore compared a plot of Eq. (12) with a plot of the geographic latitude where a core
was collected versus the paleomagnetic inclincations determined from that core. Figure
18 shows results from over 50 cores collected at geographic latitudes between 55°N and
62°S. Each point in the figure was obtained from a single core and thus gives the mean
inclination of the field over the last 700,000 years, a time span more than adequate to
permit a test of the dipole hypothesis. Similar results were obtained from 15 cores spanning
the period 0.7-2.5 m.y. B.P.
If the paleomagnetic field was indeed approximately dipolar, then paleomagnetic pole
positions determined from different localities should be tightly grouped. Figure 19 shows
the Triassic (190-225 m.y. B.P.) pole positions determined from rock samples collected
from localities scattered across North America. As the data span some 35 m.y. and are
from formations about 200 m.y. old, the grouping is remarkable. According to Strangway
(1970), the most divergent poles (the ones scattered across the north coast of Siberia into
the Bering Sea) are from early paleomagnetic measurements which were not adequately
tested for contaminating NRM. The Triassic pole positions from Russian rocks are also
tightly grouped, but around a point in the Pacific Ocean off the tip of Kamchatka. (Such
separations of the mean paleomagnetic pole positions from different continents comprise
some of the most compelling evidence of continental drift.) The Triassic results confirm
the dipolar configuration of the field, but owing to their divergence from the present geo-
graphic pole caused by continental drift, say nothing of the field's geographically axial
nature. However, paleomagnetic poles obtained from rocks less than 25 m.y. old are only
slightly affected by continental drift and do cluster near the present geographic pole.
-9:).
BRUNHES EPOCH
-60·
'.
-30·
LA TI TLrr IN OCGREES
-9:). -60· -30· 30· 60·
tfI
W
ú= 30·
[oj
....
ú=
§
....
ú=
g 60· ".
ú=
ú=
9:).
Figure 18. Mean paleomagnetic inclination versus core latitude for several deep-sea cores (x) spanning
the last 700,000 years. The solid curve is the inclination versus latitude appropriate to an axial geo-
centric dipole. From Opdyke and Henry (1969) with permission.
The dipolar nature of the geomagnetic field over a considerable portion of geological
time seems well established. North American pole positions from the Quaternary, Triassic,
and Cretaceous are even more tightly grouped than the Triassic poles shown in Fig. 19,
while poles from the Jurassic, Permian, and Carboniferous show only slightly more scatter
than those of the Triassic. And though data from the lower Paleozoic are sparse, the pole
positions which have been calculated are grouped almost as well as those from the upper
Paleozoic (Strangway, 1970).
Figure 20 shows 71 paleomagnetic poles obtained prior to 1970 from Precambrian
rocks in North America. As the data span 2 b.y., the poles cannot be expected to be tightly
clustered. But they are far from randomly distributed, appearing instead to trace out a fairly
well-defined curve known as an apparent polar wander path (APWP). The smooth char-
acter of the curve and the low dispersion of most poles about the curve is clearly consistent
with a dipolar field. The APWP is presumably due largely to the drift of the North American
continent with respect to the geographic pole. However, it may contain a contribution from
true polar wander, a movement of the geographic axis with respect to the entire solid earth.
70 Chapter 3
ISO
Figure 19. Triassic paleomag-

netic pole positions from North
American rock units. From
o Strangway (1970) with permis-
sion.
In light of this array of impressive results, we must not lose sight of the fact that the
axial dipole hypothesis applies only to the configuration of the geomagnetic field when
averaged over 104 -105 years. In reviewing the theory of the geodynamo, Busse (1980) notes
that a dipole inclination of about 10° seems to be characteristic of planetary magnetic fields
and is probably an intrinsic property of planetary dynamos. McElhinny and Merrill (1975)
point out that about half of the published paleomagnetic poles for the last 5 m.y. obtained
Figure 20. Precambrian paleomag-

netic poles and apparent polar wan-
der path for North America. Age in
m.y. B.P. is indicated beside path.
From McElhinny (1973) with per-
mission.
from individual locality data do not coincide with the geographic pole because the data
do not span a sufficient period of time.
The scatter in paleomagnetic pole positions comes from several sources-experimental
error, continental drift, wobble of the geographic pole, and the geomagnetic secular var-
iation and nondipole field. Although the effects of these sources may tend to offset one
another, it is not unreasonable to view the pole scatter as a rough indication of the average
magnitudes of the secular variation and the nondipole field, and to infer from data such
as those shown in Figs. 17-20 that the secular variation and nondipole field have not
varied dramatically through geological time.
Careful analyses support this view. Brock (1971) showed that a useful comparative
measure of the paleo secular variation could be obtained from the dispersion in pole po-
sitions and found that the average secular variation prior to the Cenozoic was about 15%
lower than during the Cenozoic. From a more detailed analysis of paleomagnetic results
for the last 350 m.y., Irving and Pullaiah (1976) reached a similar conclusion-the secular
variation during the last 100 m.y. was somewhat higher than for the previous 250 m.y.
Brock (1971) also suggested that the frequency of geomagnetic field reversals is correlated
with the amplitude of the secular variation which, in turn, is correlated with the magnitude
of the nondipole field. If that is correct, then the variation in the reversal rate shown in
Fig. 30 provides a rough picture of the secular variation and the nondipole field during
the Phanerozoic.
McElhinny and Merrill (1975) have isolated the contribution of the secular variation
to the angular dispersion of the published VGPs for the last 5 m.y. Because that mean
angular dispersion is about 16° for the entire period but only about 9° for the Holocene,
they concluded that the secular variation during the Holocene has been somewhat lower
than the 5 m.y. norm.
2.5. Paleointensities and the Age of the Field
As noted previously, the Gauss coefficients of the dipole field are changing rapidly
(Table I), indicating that large changes in field intensity might occur in only a few centuries.
Indeed, during the time since Gauss was able to obtain the first reliable estimate of 8.55
X 1025 gauss cm 3 in 1835, the dipole moment has decreased at a rate of about 5%/century.
Is this rate typical? How long will it persist? The answer to the latter question is a matter
of conjecture. Some indication of the answer to the former can be obtained from paleo-
magnetic intensity determinations.
Unfortunately, reliable paleointensity determinations are far less numerous than VGP
determinations because the procedures for obtaining the former are far more complex and
time-consuming. Basically, the procedure is to first measure the natural TRM /p of a sample,
then heat the sample above the Curie temperature and allow it to cool and acquire a new
TRM /0 in a laboratory field of known intensity Fo. The paleomagnetic field intensity Fp
is then obtained from
(13)
Though simple in principle, the method often fails because heating the sample may alter
its magnetic properties so that the relation (13) does not apply. In addition, natural pro-
cesses may have altered the magnetic properties of the sample since its formation, so that
even though the paleomagnetic direction is preserved, the paleointensity is not. Fortu-
nately, unsuitable samples are readily detected and reliable (or at least internally consis-
tent) paleointensity estimates can be obtained. However, the process involves considerably
more than a straightforward application of Eq. (13). The preferred procedure, known as
72 Chapter 3
,
14
22
_12 24 + 21 1
•+
"':::I! 7 0
u
23
I
ú =10
T 5 I
8
+ 3 2 8 Figure 21. Variation of the
•
::)
oC[ I:J 7 8
l!)
geomagnetic dipole moment
:cc 8 during the last 9000 years in-
2 2 ferred from paleointensity de-
E! E!
ú =6
I
E!
terminations available prior
w to 1967. Each point is the av-
•
:::I!
c 5 erage of published virtual di-
:::I! 4
w pole moments for a 500-year
..J
cQ. interval. The number of in-
o 2 dividual determinations is
given above each point, and
the standard error is indi-
O
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 cated by a vertical line. From
AGE (YEARS BEFORE PRESENT) Cox (1968) with permission.
the Thellier method (Thellier and Thellier, 1959), involves heating and cooling the sample
through successively larger temperature intervals until the Curie temperature is reached,
and comparing the NRM destroyed by heating to the TRM acquired by cooling in a known
field. The process is then repeated to determine if the higher temperatures have altered the
magnetic behavior of the rock in the lower temperature intervals. If the magnetic behavior
is not altered, and if all temperature intervals produce the same paleointensity value via
Eq. (13), then that value is presumed to be valid. For a detailed review of paleointensity
determination, the reader is referred to Carmichael (1977).
Figure 21 summarizes the 127 Holocene paleointensity values determined prior to
1967 (Cox, 1968). The intensities were reduced to virtual dipole moments via Eqs. (10)-
(12) and averaged over 500-year intervals. Such time averaging serves to reduce the scatter
produced by the nondipole field and secular variation. During the last 9000 years, the
dipole intensity has apparently varied by 50% on either side of its present value. The
results cover too little time to provide any real indication of whether they are typical of
the behavior of the field over the long term, but they suggest a cyclic variation in amplitude
with a half-cycle of about 4000 years, a value which is remarkably consistent with the
theoretical free decay time of less than 104 years for electric currents in the earth's core.
Figure 22 shows the pre-Holocene paleointensity values assembled by Smith (1967).
They indicate that during the last 400 m.y., both the mean dipole moment and the am-
plitude of variation of the dipole moment have steadily increased. While the increases are
probably real, Smith points out that they could be artifacts of the aging of the rocks. Over
geological time, the natural TRM could have decayed or a substantial component of CRM
could have been acquired, so that the experimental method is invalid. Consistent with
these possibilities are the three unusually high dipole moments around 270 m.y. B.P.,
which were obtained from rocks with an exceptionally high magnetic stability. On the
other hand, McElhinny (1973) and Carmichael (1967) contend that the average field
strength passed through a real minimum about 500 m.y. ago, because the published Pre-
cambrian paleointensity values are comparable to, or greater than, the present field inten-
sity. For example, Bergh (1970) obtained a paleointensity of about two-thirds that of the
present from rocks 2.45 b.y. old, while McElhinny and Evans (1968) obtained an intensity
about 50% greater than that of the present from rocks 2.65 b.y. old.
18
a
16
14
o
12
10
mê ú ë É å =t
ú á é ç ä É = moment
8
•
11 6
.,".,
::l4
ú= • ••
• I
-
ú= 2
0
x
..., 0
•
• 0
••
•
ú=
§ 18 b
a
i. 16
"'-
:a
....
ú NQ= "'-
Figure 22. (a) Pre-Holocene Pha-
nerozoic virtual dipole moment ;; 12 "'- x
o
determinations available prior to "'-

1967. Individual points repre- 10
"'-
"-
mê ú R É å í =
simt data from separate rock ú á é ç ä É = moment
units in cases where less than 10 8
""-
units were used. Vertical bars
represent the range in cases 6
""- ,,4
where data from 10 or more units "'-
were available. (b) Curves 1-3 4 "'-
"""-
are trends based on data in (a). -\-..:-.:.- ------:::-'-
Curve 4 is the best-fitting upper 2 3 ---- -
bound for all data except the +
0
three exceptionally high dipole
moments at 270 m.y. B.P. From o )00 200 300 400
Smith (1967) with permission. Age (x 10' yr)
The latter intensity determinations, together with the paleomagnetic pole distribution
shown in Fig. 20, clearly establish the presence of a geomagnetic field in the Precambrian.
Recently, McElhinny and Senanayke (1980) have obtained paleomagnetic evidence that
the field is at least 3.5 b.y. old, comparable to the age of the oldest known rocks.
From the preceding discussion, it is clear that the available evidence strongly supports
the view that although the exact configuration of the geomagnetic field at any particular
instant is unique, the field of the past has in general been substantially similar to that of
the present-a field whose average intensity at the earth's surface is on the order of 0.1-
1 gauss and whose direction at most points on the globe (except near large crustal an-
omalies!) does not vary by more than a small angle (25 0 maximum in the case of the present
field) from that of a geocentric dipole inclined about 10 to the geographic axis. Of course,
0
given the age of the field, we should not be surprised to find some exceptions to the general
rule. Indeed, it appears that during a reversal or near reversal of the direction of the main
74 Chapter 3
dipole, the field drops in intensity, ceases to be dipolar, and becomes highly variable and
irregular for perhaps a few thousand years.
2.6. Reversals of the Dipole Field
It was noted earlier that rocks are commonly found whose NRM is antiparallel to the
present geomagnetic field direction. There are two possible explanations for this-either
the geomagnetic field reversed and the NRM of the rocks faithfully recorded the event, or
the NRM of the rocks underwent a "self-reversal" by assuming a direction opposed to the
ambient magnetic field at the time of formation of the rocks or by spontaneously reversing
at a later date. That reversals are consistent with dynamo theory while self-reversing rocks
have rarely been observed in the laboratory (Cox et 01., 1964) argues strongly in favor of
the reality of field reversals. But the most compelling evidence is provided by several other
observations. In a stratigraphic sequence, the polarity of the NRM is correlated with the
age of the rocks but not with their mineralogy. Contemporaneous rocks of worldwide dis-
tribution contain NRMs of the same polarity. Polarity reversals are recorded simultaneously
in rocks throughout the world. A baked contact zone in a rock unit almost invariably
exhibits a polarity identical to that of the igneous rock that heated it, rather than to that
of the unbaked parent rock.
NRM which is roughly parallel to the present field is termed normal and that which
is roughly antiparallel is termed reversed. Similarly, a paleomagnetic field whose north
geomagnetic pole lies in the northern (southern) hemisphere is termed a normal (reversed)
field. Because the latter definition is made with reference to the appropriate paleogeo-
graphic pole, rather than with reference to the present geographic pole, care must be taken
when the APWP crosses the present equator as in Fig. 20. However, for rocks dating from
the Devonian onward, this problem does not arise (McElhinny, 1973) and need not be
considered in the following discussion of reversal sequences.
The most detailed and accurate picture of the sequence of geomagnetic reversals has
been obtained from rocks less than 5 m.y. old dated by the potassium-argon (K-Ar) isotopic
method. The polarity time scale for that period (Fig. 23) is divided into four polarity
epochs, each lasting 700,000 years or more, and includes several polarity events, lasting
from 20,000 to 200,000 years. The distinction between epochs and events is purely his-
torical and does not reflect a bimodal distribution of the times between successive reversals.
When the time scale was originally being worked out with sparse paleomagnetic data,
events escaped detection and it appeared that the field had reversed only once every million
years or so. Epochs thus denote longer periods of time during which the polarity of the
field was predominantly normal (e.g., the Brunhes normal epoch) or predominantly re-
versed (e.g., the Matuyama reversed epoch), and events denote shorter periods of time
within an epoch during which the polarity was opposite to that of the epoch. The time
between two successive reversals is generally termed a polarity interval, though the term
is used rather loosely in the literature and is occasionally applied to an epoch as well.
Apart from those events listed as questionable in Fig. 23 (e.g., the Laschamp event),
the field has reversed at least 23 times in the last 5 m.y. The length of polarity intervals
has ranged from about 20,000 years to 730,000 years and averaged about 217,400 years.
Particularly in the older data, there are gaps of 100,000 years or so, so that additional short-
duration reversals may yet be discovered. Without doubt, however, the field of the Brunhes
epoch has been almost exclusively normal, as the data are sufficiently dense to preclude
discovery of any but a few very short reversed events. As none of the questionable events
in Fig. 23 has been documented worldwide, it appears that the most recent dipole reversal
occurred about 730,000 years ago.
.-.... .......
!
...
-:".,
"., .
- -. "
ú= Éú =
çú=
".,
"'''''
e·-
to_
".,
- c c
.- '"
-0
".,
.-.. c
".,
.- s:.
K-Ar e"
.. 0 .0 [ ú= .111 2.
Agt úç =
00 -0 .0 01_ 0> Oa.
Im.y.l ZQ. Ell. czQ. «0 Q.w Q.w
--
Laschamp'
;e ,
II
úàyyúà= Bloke
11\
w
x
.--.
z
::::l
!iii cz
III
Illillll 073
-
,a = ..
-
:::!
090
097 Jaramillo
1i -= «
::I:
--
«
Itt,
.
1.67 ú =
::::l
I- Olduval ú =
I «
t 87
- ::I:
2 a
!!
,
-- ... -
'X'
55
;;
= 11\
11\
Koena ::::l
3 0- «
!!
C)
Mammoth
-
=
-
•
!OS
380
'" . toc:hiti
390
4 0-
ú =
4.05 cz
Nunivak w
-
III
Figure 23. Geomagnetic reversal sequence and 4.20 oJ
polarity time scale for 0-5 m.y. B.P. Each short 4.32 C)
SldufJall
horizontal line at the left indicates a K-Ar age
and magnetic polarity obtained from a single
--- 447
volcanic cooling unit. Normal (reversed) po-

larity intervals are stippled (white). Arrows in-
dicate polarity events which have yet to be 4.85
confirmed world wide. From Mankinen and . .... 5.00
Thvtro
5.0
Dalrymple (1979) with permission.
76 Chapter 3
The Laschamp event was named on the basis of two lava flows in France, dated between
8730 and 20,000 years ago, whose NRM is inclined 158° from the direction of a normal
axial dipole field (Bonhommet and Ziihringer, 1969). However, reversely magnetized rocks
of comparable age have yet to be found elsewhere. For example, Denham and Cox (1971)
studied the magnetic stratigraphy of a section of sediments from Mono Lake, California,
dating from about 30,400 to 13,300 years B.P., and found no evidence of the Laschamp
event. They did, however, locate an excursion of the field beginning around 24,600 years
B.P. and lasting some 600 years, during which the field apparently departed an average
of 25° in inclination and 78S in declination from its normal direction. The maximum
excursion of the declination was 130°, which implies an average rate of change of almost
OS/year. Such a rapid change in field direction might have drastic consequences for a
population of organisms which depends upon a relatively stable field direction for accurate
orientation or navigation.
Because they apparently do not represent field reversals, the Laschamp event and
Mono Lake excursion are generally believed to have been caused by unusually large-am-
plitude fluctuations in the non dipole field or by lowered dipole field intensities (Fig. 21)
which rendered the effects of the nondipole field more apparent. The existence of such
events and excursions therefore suggests that on occasion the dynamo process becomes
temporarily disordered.
For rocks older than 5 m.y., the precision of K-Ar dating is insufficient to establish
a reliable polarity time scale based on worldwide data. At present, the typical uncertainty
in a K-Ar date is about ± 2% (McDougall, 1979). A reversal dated at 5 m.y. ago might
therefore have occurred anywhere within a 200,000-year period (a period equal to or greater
than 16 of the 23 polarity intervals shown in Fig. 23), making it virtually impossible to
use K-Ar dates to correlate reversal sequences from different localities. Extensions of the
polarity time scale beyond 5 m.y. B.P. therefore rely heavily on individual stratigraphic
sequences deposited continuously over long periods of time. Correlation of data from dif-
ferent localities is possible via a pattern-matching procedure much like that used in tree
ring dating. Various geological and paleontological markers can then be used to fix the
date of certain points on the scale. (Conversely, owing to the worldwide nature of a geo-
magnetic field reversal, magnetic polarity stratigraphy shows considerable promise for
dating and correlating geological and paleontological events.)
With data obtained from marine magnetic anomalies similar to those shown in Fig.
12, Heirtzler et al. (1968) extended the polarity time scale to cover the last 80 m.y. (Fig.
24). They relied on the generally accepted model of sea floor spreading, which assumes
that hot material from the earth's mantle is continuously upwelled and extruded on the
surface at midoceanic ridges and subsequently spreads bilaterally away from the ridge at
rates of a few centimeters per year. As the material cools through the Curie temperature,
it becomes magnetized parallel to the ambient field and appears today as a positive (neg-
ative) magnetic anomaly if its NRM is normal (reversed). The process functions as a geo-
logical magnetic tape recorder which preserves a continuous record of geomagnetic po-
larity. To aid in the correlation of anomaly patterns from different regions, major positive
anomalies are identified by number. Anomalies 1-32 are shown in Fig. 24.
The reversal sequence in Fig. 24 was constructed from magnetic anomaly patterns
recorded in the North Pacific, South Pacific, South Atlantic, and South Indian Oceans. To
obtain the time scale, Heirtzler et al. (1968) first determined an average spreading rate for
the most recent part of the sequence by matching the first several marine anomalies to the
known terrestrial polarity sequence. Using a date of 3.35 m.y. for the end of the Gilbert
reversed epoch (see 3.40 m.y. in Fig. 23), they assumed a constant spreading rate and
obtained a time scale for the entire sequence via linear extrapolation. Because the 3.35
m.y. baseline was extrapolated to almost 25 times its length, several revisions of the time
scale have been proposed. (For a review, see Butler and Opdyke, 1979.) The major effect
MY
PLEIS
PLIO
MIO
,
0 LI
I ú =
EOC
PALAE
Figure 24. Polarity time scale for the last 80 m.y. determined from CRET
marine magnetic anomalies, Normal (reversed) polarity intervals are
black (white). The numbers assigned to prominent anomalies are given
to the left of the reversal sequence, and the time scale in millions of
years to the right. Geologic periods are indicated on the left. From
Heirtzler (1968) with permission. - 80
of these revisions is to compress the extrapolated portion of the Heirtzler et 01. (1968) time
scale by about 7% by placing the Cretaceous-Tertiary boundary (65 m.y. B.P.) just below
the base of anomaly 29 rather than between anomalies 26 and 27 (LaBrecque et 01.,1977).
The adjustment is based on the position with respect to the Cretaceous-Tertiary boundary
of a terrestrial normally magnetized zone correlated with marine anomaly 29. Of course,
any deviations from the assumed constant spreading rate as may have occurred have pro-
duced corresponding distortions in the polarity time scale. Because of such uncertainties,
78 Chapter 3
40
en
...J
«
>
II: 30
w
I-
Z
u.. 20
o
II:
W 10
III
::E
::J
Z
0.5 1.0 1.5 2.0 2.5

LENGTH OF INTERVAL (M.Y.)
Figure 25. Histogram of lengths of geomagnetic polarity intervals for the last 72 m.y., based on the
polarity time scale of LaBrecque et 01. (1977).
one should exercise considerable caution when applying polarity time scales and reversal
sequences to problems in stratigraphy, paleontology, evolutionary biology, and the like.
Although the polarity time scale derived from marine anomalies is inherently less
accurate than the late Cenozoic time scale of Fig. 23, it is remarkably detailed and provides
a very interesting picture of the behavior of the geomagnetic field over the last 80 m.y.
The analysis of LaBrecque et a1. (1977) reveals 188 reversals in the last 71.62 m.y. and
includes polarity events as short as 10,000 years. The mean length of a normal (reversed)
interval was 349,000 years (412,500 years), indicating the field was reversed about 54.2%
of the time. While that slight asymmetry in polarity may simply be an artifact of the im-
precision of the polarity time scale, it may well be the result of physical effects such as
thermoelectric currents or the NRM of crustal rocks which bias the dynamo process (Merrill
et a1., 1979). Figure 25, a histogram of the frequency distribution of the lengths of the
polarity reversals, shows a roughly exponential decrease of the frequency of occurrence
of longer intervals.
From Fig. 24, it appears that beginning with anomaly 18 the reversal frequency ab-
ruptly increased. Based on the LaBrecque et a1. (1977) polarity time scale, the mean length
of a polarity interval was about 777,000 years between 71.62 and 42.88 m.y. B.P. and
284,000 years between 42.88 m.y. B.P. and the beginning of the Gilbert reversed epoch at
5.12 m.y. B.P. The latter interval length is about 30% greater than the 217,400 year average
for the last 5 m.y. obtained from Fig. 23. The indicated increase in reversal frequency over
the last 72 m.y. is probably real, but its magnitude has almost certainly been exaggerated
by a failure to detect some very short events in the marine anomaly record and by the
obliteration of some older events by geological processes.
A mid-Mesozoic polarity time scale developed by Larson and Hilde (1975) from Ha-
waiian marine magnetic profiles is shown in Fig. 26. From those and other results, it
appears that the Cretaceous and Jurassic periods were dominated by normal geomagnetic
polarity. The existence of the Cretaceous Long Normal Zone, a period of practically un-
broken normal polarity extending from about 110 to 80 m.y. B.P. (Figs. 24 and 26), is fairly
well established. However, conflicting paleomagnetic results leave unresolved the question
of whether the Jurassic polarity was dominantly normal or reversed (Chan and Alvarez,
100
ALBIAN
APTIAN
<f)
MI
o=>
UJ BARREMIAN
U
<I:
t-
UJ ------ M5
n:
u
>- HAUTERIVIAN
:..J
------
n:
ú=
VALANGINIAN
130
BERRIASIAN MI5"
TITHONIAN
140
M20··
u
ú = KIMMERIDGIAN
<I:
n:
..., ------
=>
150
ú = OXFORDIAN
<I:
..J
Figure 26. Mid-Mesozoic polarity time scale obtained from marine

magnetic anomalies. Normal (reversed) polarity intervals are black CALLOVIAN
(white). Magnetostratigraphic nomenclature is at left. From Larson 160
and Hilde (1975) with permission. ----
1983). Generally speaking, the polarity time scale prior to the Cretaceous is poorly known
and will remain so for years to come. For recent reviews of the magnetostratigraphic time
scale, the reader is referred to Cox (1983) and Chan and Alvarez (1983).
McElhinny (1971) suggested that a measure of the relative frequency of reversals could
be obtained without knowing the reversal sequence simply by noting the proportion of
rock units of a given age which contain mixed, i.e., both normal and reversed, NRMs. In
Fig. 30 the percentages of mixed results obtained from worldwide paleomagnetic data are
plotted as a function of geological age. Of particular interest are the very low reversal
frequencies indicated for the Silurian and upper Carboniferous. The latter minimum cor-
responds to the Kiamann, or Permo-Carboniferous, quiet interval, a period spanning the
entire Permian and the upper Carboniferous in which few field reversals occurred. That
the same interval of reversed polarity has been observed in paleomagnetic data from the
Russian, North American, and Australian continents attests to its reality (McElhinny, 1969).
From a review of the paleomagnetic record, Irving and Pullaiah (1976) estimate the Kia-
mann interval lasted from 313 m.y. B.P. to 227 m.y. B.P.
Owing to the paucity of Precambrian paleomagnetic data and the problem of defining
polarity with respect to a poorly known APWP, a reliable Precambrian reversal sequence
80 Chapter 3
has yet to be obtained. However, McElhinny (1973) noted that 25% of the then available
126 Precambrian polarity measurements were mixed, as compared to 29% of the Phane-
rozoic measurements, and concluded that the reversal frequency during the Precambrian
was probably similar to that of the Phanerozoic. McElhinny also found that of 1231 pa-
leomagnetic results spanning the entire Phanerozoic, 54.1 % were reversed, a figure which
is remarkably consistent with the 54.2% reversed polarity time obtained from the data of
LaBrecque et 01. (1977) for the last 72 m.y.
2.7. Polarity Transitions: What Happens during a Reversal?
In principle, a dipolar geomagnetic field could reverse either by maintaining its align-
ment, decreasing to zero intensity, and then building up in the opposite direction, or by
maintaining its intensity and swinging through 180°. Alternatively, the dynamo process
could become disordered, resulting in the breakup of the dipole field into an irregular,
nondipole field, and eventually produce a new dipole field of reversed polarity when order
was restored. Clearly, any of these processes might have serious, worldwide consequences
for species which depend in any way upon a stable geomagnetic field. At the very least,
each process entails a considerable loss of magnetic compass information during a reversal.
Unfortunately, dynamo theory is not sufficiently developed to provide a choice among
these possibilities. We must therefore turn to the paleomagnetic record where we are still
not provided with a clear choice, owing to a very fragmentary record. As the reader has
no doubt inferred from the previous discussions, the reversal process is relatively rapid.
The theoretical decay time for the magnetic field in the earth's core is only a few thousand
years. Consistent with this, polarity events (which entail two reversals) as short as 10,000
years or so have occurred. Such a transient process is unlikely to be captured by intermittent
igneous events or recorded in any detail by relatively slow sedimentary processes. Never-
theless, persistent efforts by paleomagnetic investigators have produced several records of
polarity transitions. Though of variable quality, those records are beginning to provide a
rough picture of what happens during a reversal. It will be several years before a complete
picture of any particular reversal is derived, as a record of the reversal must be obtained
from several localities, and it will be far longer before we can determine how typical that
picture is of reversals in general.
The paleomagnetic record of polarity transitions has been reviewed by Fuller et 01.
(1979). Here, it will suffice simply to outline the general phenomenology of the reversal
process as it is presently interpreted from paleomagnetic data. Figures 27 and 28 are pa-
leomagnetic records of a late Tertiary reversal obtained from an igneous intrusion which
will aid the reader in visualizing the nature of a reversal. As each point represents the
mean value from a l-m rock section, Fig. 27 presents a smoothed version of the field
behavior prior to, during, and after the polarity transition. Figure 28 shows the unsmoothed
details of a 2-m section of the transition.
During a reversal, the field neither remains dipolar* (Hillhouse and Cox, 1976) nor
decreases to zero. Apparently, the intensity of the dipole component decreases below that
of the nondipole component, while the nondipole intensity remains roughly unchanged
and determines the transitional field geometry. While there is considerable evidence that
* The frequent use of VGP paths in describing field reversals should not be taken to imply that the
field is dipolar during a polarity transition. It is merely a useful way of describing the directional
behavior of the field at a particular site and comparing records from different sites.
90
z
52 0
I-
«
ú=
...J 270
u
w
0
180
90
-20 10 0 10 20 30 40
90
60
z
0 30
i=
« 0
z
d -30
ú=
-60
- 90
-20 -10 0 10 20 30 40
xIO- 4 G cm 3 gm- 1
5
>- 2
!:::
en
z 0·5
w
I-
Z 0·2
0'1
- 20 30 40
... N
DISTANCE (melers)
R
Figure 27. Paleomagnetic record of a reverse to normal geomagnetic reversal obtained from a late
Tertiary igneous intrusion. The abscissas show the distance of the paleomagnetic rock sample from
the midpoint of the reversal (I = 0). Declination and inclination are given in degrees. The intensity
of the NRM (after cleaning) is given in the lower frame. Each point represents the mean value of several
measurements from a 1-m section. From Fuller et a1. (1979) with permission.
the overall field intensity decreases during a reversal, possibly to about 10% of its normal
intensity (Opdyke et aI., 1973; Dodson et aI., 1978), there is some evidence that locally,
or for short periods, the field intensity may remain fairly high (Prevot, 1977; Shaw, 1977).
The transition in polarity requires around 4000-5000 years (Harrison and Somayajulu,
1966; Opdyke et 01.,1973) and the intensity decrease may either coincide with the polarity
transition (Opdyke et 01., 1973) or last two to four times longer than the polarity transition
(Hillhouse and Cox, 1976; Dodson et 01., 1978; Fig. 27).
As the dipole intensity decreases and the global field becomes less ordered, local field
directions show greater dispersion and change more rapidly. During the actual reversal of
polarity ( -1 to 10 m in Fig. 27), the magnitudes of these fluctuations increase dramatically
(Fig. 28), producing a secular variation whose general appearance is similar to, but whose
82 Chapter 3
Figure 28. VGP path obtained from a 2-m section of the actual polarity transition (I passing through
0) of Fig. 27. Geographic poles are denoted by X. From Fuller et a1. (1979) with permission.
amplitude is much larger than, the secular variation observed in historical times (ef. Figs.
6 and 17). Although highly irregular, the transitional field is apparently not completely
chaotic and may well be dominantly quadrupolar or octupolar (Hoffman and Fuller, 1978).
2.8. What Causes Reversals?
For a dramatic event such as a reversal of the earth's magnetic field, it is perhaps
natural to look for an exotic cause. Indeed, it has been proposed that cometary or meteoritic
impacts may have triggered the last two field reversals (Glass and Heezen, 1967; Durrani
and Kahn, 1971). However, although it may well be possible for a catastrophic event to
perturb the geodynamo sufficiently to cause a reversal, it is becoming increasingly clear
that it is generally unnecessary, and perhaps even inappropriate, to look beyond the dy-
namo process itself.
We have already noted that the Lowes-Wilkinson laboratory dynamo exhibits spon-
taneous changes in field directions after periods of stability. The solar magnetic field, as
evidenced by sunspots, reverses every 11 years. In addition, it has been demonstrated
analytically that the currents in a system of two coupled disk dynamos, in which the
stationary coiled circuit of each disk provides the magnetic field for the other disk, undergo
aperiodic reversals (Rikitake, 1958; Mathews and Gardner, 1963). The stochastic reversal
model of Cox (1968, 1969) assumed only that reversals result from the interaction of an
anomalously intense nondipole field with a low-intensity dipole field. It is now apparent
that various nonlinear physical systems, of which the geodynamo is but one example,
typically exhibit aperiodic chaotic behavior (Busse, 1983). Hence, reversals are probably
an intrinsic property of the geodynamo which need not be precipitated by an extrinsic
event.
On the other hand, changes in the long-term behavior of the geomagnetic field are
now viewed as the result of changes in the conditions under which the dynamo operates
(McElhinny, 1979; Busse, 1983). In a very qualitative way, an analogy can be drawn be-
tween fluid motions in the earth's core and meteorological phenomena (see, e.g., Cox,
1975). The motions producing the geomagnetic secular variation correspond to large-scale
weather patterns such as storms and high- and low-pressure regions moving across the
globe, whereas the extremely long-term changes in the dynamo process correspond to
climatic changes. Relatively sudden changes in the long-term (10-100 m.y.) average fre-
quency of reversals are thus presumed to reflect changes in the boundary conditions im-
posed on the dynamo, such as changes in the energy input or distortions in the geometry
or physical nature of the core-mantle interface. In particular, there is the intriguing pos-
sibility that the long-term behavior of the geodynamo is related to mantle convection and
hence to continental drift. Jacobs (1981) has associated changes in geothermal heat flux
with changes in reversal frequency, and Vogt (1975) has observed that changes in reversal
frequency apparently coincide with the major changes in plate tectonics around 110-120,
70-80, and 42-45 m.y. ago. (see Figs. 24 and 26). Very recently, Force (1984) observed
that the geological record contains a large body of evidence which suggests a connection
between field reversals, seafloor spreading rate, and climate.
2.9. Field Reversals and Phylogenetic Change
In 1963, shortly after the discovery of belts of geomagnetically trapped, high-energy

particles circling the earth (Van Allen, 1959), and at a time when the reality of geomagnetic
field reversals was firmly established Uacobs, 1963), Uffen (1963) published a note in
Nature with the provocative title "Influence of the Earth's Core on the Origin and Evolution
of Life." Uffen suggested that prior to the information of the earth's core and the devel-
opment of a geomagnetic shield from energetic charged particles from space, the surface
of the earth was continually bathed by a "sterilizing cosmic radiation" which inhibited
biogenesis.
After core formation and the ensuing evolution of life, a field reversal would cause
geomagnetically trapped particles to spill onto the earth, allow the solar wind (see next
section) and cosmic radiation to reach the surface, and expose various organisms to greatly
increased ionizing radiation. Mutation rates would rise, and evolutionary rates would
therefore be determined in part by the frequency of geomagnetic field reversals and by
conditions in the earth's core.
Uffen's hypothesis was quickly rejected (Sagan, 1965; Black, 1967; Waddington, 1967;
Harrison, 1968), primarily because quantitative estimates showed that during a reversal,
84 Chapter 3
GAUSS
EPOCH MATUYAMA EPOCH BRUNHES EPOCH
acquilanius
univ.rsus
Eucyrtidium matuyamai ::0

l>
blcornis 0
0
PtllrDCDnium
01
r
l>
:0 Z ::0
'"< 0
:0
l>
H. Vilma
'"en
:0
ú =
,...
l>
'"c
Figure 29. Geomagnetic polarity sequence for the last 2.5 m.y. and the fossil record of eight species
of Radiolaria that became extinct during that period. From Hays (1971) with permission.
cosmic radiation at the earth's surface would increase by only 10-15%, which amounted
to little more than the 10% increase from the equator to the poles that exists today. Such
a modest increase could hardly be expected to significantly alter mutation rates.
At the same time, however, a body of evidence was accumulating from the study of
deep-sea cores which indicated that the times of extinction or genesis of various species
of marine microfauna coincided with field reversals (Harrison and Funnel, 1964; Opdyke
et 01., 1966; Watkins and Goodell, 1967; Hays and Opdyke, 1967; Hays et 01., 1969; Kennett
and Watkins, 1970; Hays, 1971). For example, of eight species of Radiolaria from various
oceans that became extinct in the last 2.5 m.y., six disappeared from the fossil record near
the time of reversals (Fig. 29), and of eight species of Foraminifera which became extinct,
seven disappeared near reversals. It is important to note that the fossil records and the
magnetic records were obtained from the same cores, so that the correlations were not
artifacts of imprecise dating methods.
Simpson (1966) examined Uffen's contention that evolutionary rates should depend
on reversal frequency by correlating speciation rates with the occurrence of reversely mag-
netized rocks. However, as pointed out by McElhinny (1973), such a correlation does not
address the question, for, as the Kiamann interval clearly demonstrates, the frequent pres-
ence of reversely magnetized rocks need not imply that frequent reversals occurred.
McElhinny compared the frequency of reversals throughout the Phanerozoic with the ap-
pearance of new families and the extinction of existing families (Fig. 30) and concluded
there was no significant correlation, as the maxima of phylogenetic change coincided with
both maxima and minima of reversal rates. However, a perfect correlation can hardly be
expected as geomagnetic reversals are not the sole factor influencing evolutionary rates.
Moreover, evolutionary rates and reversal frequency do not appear to be entirely unrelated
in Fig. 30. Coincidence of extreme minima of phylogenetic change with the Silurian and
upper Carboniferous magnetically quiet intervals is particularly intriguing, as is the burst
in extinctions at the end of the Permian. Citing the large increases in extinctions at the
100
New families
75
25
0
£
SO
Extinctions
-
c
II
U
"-
II
Cl. 25
• Mixed polarities
50
25
• •
200 0
Figure 30. Comparison of the rates of genesis and extinction of faunal families. expressed as the
percentage of existing families (based on Newell. 1963). with the relative reversal frequencies during
the Phanerozoic. expressed as the percentage of rock units of a particular geological era containing
mixed (both reversed and normal) paleomagnetic polarities. From McElhinny (1973) with permission.
86 Chapter 3
end of the Permian and the end of the Cretaceous, Hays (1971) suggested that prolonged
periods of low reversal frequency might have allowed species adversely affected by geo-
magnetic reversals to accumulate.
Given these correlations, the question of the nature of the relation between field rev-
ersals and phylogenetic change remains a valid one. Assuming they are not fortuitous,
there are three possible explanations for the correlations: (1) A nongeomagnetic event
caused both the faunal changes and the field reversals. (2) The reversals critically altered
the nonmagnetic environment of the organisms. (3) The changes in the character of the
geomagnetic field during the reversals directly affected the organisms. Of course, anyone
of these possibilities could operate in conjunction with one or both of the others. In ad-
dition, survival of a particular species through several reversals indicates that the reversal
coincident with extinction or speciation may not have been the sole cause of the phylo-
genetic change, but may have merely tipped the scales for a species already under intense
pressure from other quite unrelated selective factors.
From a study of Antarctic deep-sea cores, Kennett and Watkins (1970) noted that
maxima in volcanic activity as well as geomagnetic reversals often coincided with micro-
faunal changes. If sufficient volcanic ash was injected into the atmosphere to reduce the
insolation at the earth's surface or to cause a climatic change, then the volcanism, rather
than the field reversal, might have been responsible for the faunal changes. Likewise, the
possible connection between core processes, tectonic events, and climate noted earlier
(Vogt, 1975; Force, 1984) suggests that field reversals and faunal changes may have co-
incided simply by virtue of their having had the same ultimate cause. Durrani and Kahn
(1971) offered another highly speculative third-party scenario. If a field reversal were trig-
gered by a cometary impact, then large quantities of gases such as methane and ammonia
released by the comet might produce faunal changes.
As an alternative to Uffen's hypothesis, Harrison (1968) suggested that a field reversal
caused climatic changes, possibly through increased ionization in the high atmosphere.
Rather tenuous evidence of climatic changes at the times of reversals has been found in
deep-sea cores, but not all reversals are accompanied by evidence of climatic change (Hays
and Opdyke, 1967; Hays et 01., 1969). Although Harrison's suggestion is currently of con-
siderable interest (Wollin et 01., 1973; King, 1974), no influence of the geomagnetic field
on climate has been proven.
More recently, Siscoe et 01. (1976) and Reid et 01. (1976) have constructed a hypothesis
which is essentially a modification of Uffen's original hypothesis. The increased charged
particle flux in the upper atmosphere during a reversal might produce catastrophic de-
pletions of stratospheric ozone and lead to increased UV radiation at the earth's surface.
Reid et 01. (1976) calculated that during a reversal, the solar proton event (an intense shower
of charged particles, primarily protons, on the upper atmosphere produced by a solar flare)
of August 1972, the most intense event then known, would have increased the biologically
effective UV dose by 15%. Proton events of far greater intensity should occur during the
thousands of years required to complete a polarity transition. Events 10 and 100 times as
intense would increase the effective UV dose by 55 and 160%, respectively. Reid et 01.
(1976) also point out that during a reversal, proton events might also affect the climate or
significantly reduce the ground-level intensity of visible radiation in the range 400-500
nm.
Hays (1971) noted that in addition to the effects of a field reversal on radiation levels
and climate, the possibility existed that reversals affected organisms directly, owing to
their sensitivity to magnetic fields. At the time, however, the experimental evidence of
biological influences of weak magnetic fields was rather tenuous, often contradictory, and
largely confined to invertebrates (e.g., Dull and Dull, 1935; Barnwell and Brown, 1961;
Palmer, 1963; Barnothy, 1969), and Hays's suggestion was ignored.
But as the balance of this volume clearly demonstrates, Hays's suggestion should be
reconsidered. In the past decade a compelling body of evidence has accumulated which
not only attests to organismic sensitivity to fields of geomagnetic intensities, but also in-
dicates that many organisms can derive selective advantage from the spatial and temporal
information content of the field. On the one hand, there are organisms whose electric sense
is sufficiently acute to detect the weak electric fields induced by their motion in the pres-
ence of the geomagnetic field (Kalmijn, 1974). Other organisms apparently respond directly
to magnetic fields, because motionally induced electric fields cannot be invoked to explain
the behavior of pigeons with magnets or current-carrying coils attached to their backs or
heads (Keeton, 1971; Walcott and Green, 1974) or the orientational responses of slow-
moving bacteria (Blakemore, 1975) and salamanders (Phillips, 1977) to magnetic fields. In
addition, the ability of stingrays to distinguish weak experimentally imposed electric fields
from motionally induced fields (Kalmijn, 1982) suggests that stingrays possess a magnetic
sense as well as an electric sense.
Certain features of the accumulating evidence of biological effects of weak magnetic
fields are particularly relevant to the problem of field reversals and phylogenetic changes.
First, the firm evidence encompasses a variety of taxa: bacteria (Blakemore, 1975), insects
(Lindauer and Martin, 1972), fish (Kalmijn, 1978; Quinn et al., 1981), amphibians (Phillips,
1977), and birds (Keeton, 1974). If the tenuous and circumstantial evidence is included,
then the list also includes flagellates (Palmer, 1963), protozoans and flatworms (Brown,
1962), snails (Barnwell and Brown, 1961), and mammals (Zoeger et al., 1981). Second,
much of the evidence pertains to orientation and migration, behaviors which are often
critical to survival and reproduction. Even small natural disturbances of the geomagnetic
field are reported to influence orientation and migration (Keeton et aI., 1974; Moore, 1977).
Third, there is some experimental evidence (Bliss and Heppner, 1976; Martin and Lindauer,
1977) and considerable speculation (Brown, 1976) that the geomagnetic daily variation
aids in the timing of circadian rhythms. Finally, Frankel et a1. (1981) have found that the
relative proportion of north-seeking and south-seeking magnetotactic bacteria in a popu-
lation depends on the inclination of the ambient magnetic field. Although probably the
result of a nongenetic modification of the bacterial phenotype in response to an environ-
mental condition, it is clear that the nature of a population can be influenced by the ambient
magnetic field. Frankel et a1. speculate that such phenotypic plasticity could enable mag-
netotactic bacteria to survive field reversals.
There is little need to elaborate upon the evolutionary implications of this evidence.
The deterioration of both the quality and the stability of geomagnetic compass information
during a reversal could be catastrophic to populations requiring reliable information for
orientation and navigation. The probable increase in the severity of geomagnetic disturb-
ances as a result of decreased geomagnetic intensity (see Siscoe et al., 1976) would also
affect species requiring magnetic compass information and might be deleterious to any
number of species by disturbing their biological clocks. There is no reason to suppose that
these are the only means by which geomagnetic reversals might directly influence evo-
lution-they are simply the most obvious in terms of the evidence now at hand. Biom-
agnetic investigators have often been confounded by the subtlety of the effects involved.
But if the last decade is any indication of things to come, refined experimental techniques
and improved equipment, together with the growing numbers of investigators, may well
reveal numerous less obvious but more effective means by which geomagnetic phenomena
might directly influence organic evolution.
3. The Field of External Origin
3.1. The Solar Wind and the Magnetosphere
The solar wind is a steady stream of charged particles, primarily protons and electrons,
flowing outward from the sun in all directions. Because the earth is immersed in the solar
88 Chapter 3
/
f k q Ñ Wú m i ^ k b =TAR
"'EOIUM MAGNE TOS><EA TM
úf =
r:
OAVStOE
BOu CARY
LAVER
I E TAv
SOl.AR
WINO
mä ^ p ú ^ =
MAGNlEr OSHE ATH
Figure 31. Interaction of the magnetosphere with the solar wind. From Friedman (1983) with per-
mission. (Figure originally due to J. G. Roederer.)
wind, the geomagnetic field does not extend indefinitely into space but is confined to a
well-defined region termed the magnetosphere, which is shaped like a comet with its tail
pointing away from the sun (Fig. 31). The sharp boundary of the magnetosphere is called
the magnetopause. Traveling at about 300 to 700 km/sec, the solar wind is "supersonic"
with respect to the hydromagnetic waves it can transmit, and upon encountering the mag-
netosphere it forms a stationary bow shock wave which stands sunward of the magneto-
pause by about three or four earth radii (one earth radius = 1RE = 3670 km). After passing
through the bow shock, the smooth flow of the solar wind becomes irregular and is deflected
around the magnetosphere. Some solar wind particles penetrate the magneto pause and
become trapped by the geomagnetic field.
The impact of the solar wind compresses the magnetosphere on the sunward side so
that the geocentric distance to the subsolar point on the magnetopause is normally about
11RE. Away from the sun, the magnetospheric tail extends beyond the moon's orbit to
perhaps 1000Re . Occasionally, solar flares perturb the solar wind by emitting clouds of
high-energy protons which rush through the solar wind and locally alter its speed and
density. The more intense disturbances can push the subsolar magnetopause to within 5R E
(Roederer, 1974). Siscoe et a1. (1976) calculated that during a geomagnetic field reversal,
the magneto pause may drop to about 2R E •
We have already noted that a Gaussian analysis [Eq. (9)) shows that a small part of
the geomagnetic field originates from sources external to the earth. Those sources are elec-
tric currents in the ionosphere, a region extending from an altitude of about 50 km to the
outer reaches of the atmosphere and consisting of charged particles either trapped from
the solar wind and cosmic radiation or produced by the ionization of upper atmospheric
atoms and molecules by high-energy solar radiation. The currents are large-scale flows of
ionospheric particles which are produced by electrical and mechanical forces and whose
amperages depend upon the densities and mean flow velocities of those particles. The
principle currents of interest here are: (1) the atmospheric dynamo, currents flowing at a
height of about 100 lan, driven by tidal motions of the ionosphere produced by solar and
lunar gravitational forces and by solar heating; (2) the ring current, an east-to-west flow
of geomagnetic ally trapped protons centered about the geomagnetic equator at a mean
geocentric distance of about 4RE ; and (3) currents in the magnetopause which are respon-
sible for confining and compressing the magnetosphere.
Changes in the configuration or intensity of any of these currents will produce cor-
responding changes in the geomagnetic field at the earth's surface. Such changes have time
scales ranging from seconds to decades and beyond. The very-short-period phenomena
(periods of from less than a second to a few minutes) are known as geomagnetic micro-
pulsations. They have various causes such as lightning storms, particle bombardment of
the ionosphere, and magnetospheric hydromagnetic waves and cavity resonances. In gen-
eral, their amplitudes are less than 1 'Y, but occasionally the longer-period pulsations have
amplitudes up to 1000 'Y, particularly in polar regions. Their amplitudes and frequencies
of occurrence undergo both daily and seasonal modulations. Micropulsations are reviewed
in detail by Campbell (1967) and will not be treated further here. Kalmijn (1974) discusses
the potential significance of micropulsations to electrosensitive organisms.
In general, one might expect to find geomagnetic phenomena with periodicities cor-
responding to any of the prominent geophysical and solar cycles such as the mean solar
(24.0 hr) and lunar (24 hr 50 min) days, the synodic month (29.5 days), the solar rotation
period (-27 days), the tropical year (365.24 days), and the sunspot cycle (11 years). Many
geomagnetic variations with these periods have been observed. In addition, prominent
aperiodic disturbances lasting from minutes to many hours are triggered by solar flares
(magnetic storms) and other solar disturbances (e.g., solar proton events), or by the sudden
release of magnetic energy from the magnetospheric tail (magnetic substorms). The lit-
erature on magnetospheric disturbances is vast, and the reader is referred to Chapman and
Bartels (1940), Matsushita and Campbell (1967), Ratcliffe (1972), and McPherron (1979)
for reviews. Here, we shall only touch upon some of the more salient features of those
phenomena which have obvious biomagnetic implications.
Finally, it should be noted that geomagnetic variations of ionospheric origin, prin-
cipally those with time scales on the order of hours, induce electric currents in the crust
and mantle, called telluric currents, which contribute to the internal sources of the geo-
magnetic field (Price, 1967). In most cases, the induced fields are only a fraction of the
ionospheric fields, but they can be quite significant near coastlines and solid-earth con-
ductivity anomalies. The amplitudes of temporal variations in the vertical element Z are
most affected, and in extreme cases may increase or decrease by an order of magnitude
(see, e.g., Rikitake, 1964).
3.2. The Solar Quiet and Lunar Daily Variations

On a few scattered days during each month, the magnetic elements at a given locality
exhibit particularly smooth variations with periods corresponding to the length of a solar
day. Such days are termed geomagnetically quiet and the variations are collectively called
the solar quiet daily variation, Sq. Sq is the result of the passage overhead of the solar-
driven atmospheric dynamo currents which are most intense on the sunward side of the
ionosphere owing to solar heating and ionization. Because the duration and intensity of
sunlight vary with latitude and with the season, the amplitude of Sq varies accordingly
(Figs. 32 and 33). Owing to the uneven distribution of the continents and oceans, there
are also longitudinal variations and hemispherical asymmetries in Sq. Sq averages around
50 'Y in intensity (compared to 50,000 'Y for the main field) and 0.10 in declination, but the
horizontal component of Sq reaches about 200 'Y at the magnetic equator. It is important
90 Chapter 3
HORIZONTAL DECLINATION VERTICAL
WJ WX WWú =
'_.'"
_-:7...ú WJ > ú =
ú WWT p =
C\ ......ú
.
=
20
<II
I£j
10 ú ..............
= j.: ;:
I£j
II:
CI
I£j
c
OS
ú ú KJ KJ ú = K ú K =
z
I£j 00 - .... -
C
::;)
to-
-20
-30
-40
Figure 32. The solar quiet daily variations
-50 ú = Sq for J months (see Fig. 33) for H, D, and
--- Z at various magnetic (dip) lattitudes and

J J ú =
:r.::=: three longitude zones: Europe and Africa

-60 ú =
Wú WK ú WK WW= J KZú J J J J -"---.-.,. (solid curves), Asia and the Pacific (bro-
120 y I 2.0' 120 y ken curves), and the Americas (chain
00 06 12 18 00 06 12 18 00 06 12 18 curves). From Matsushita and Maeda
LOCAL TIME (1965) with permission.
to note that the Sq variations shown in Figs. 32 and 33 are due to both ionospheric and
telluric currents (Matsushita, 1967).
The ionospheric tides driven by the gravitational attraction of the moon also affect
the geomagnetic daily variation. But because the magnitudes of the lunar gravitational
tides are generally less than l/loth those of the solar thermal tides, the lunar magnetic
effects are correspondingly smaller. The principle lunar effects are a semimonthly mod-
ulation of Sq and a lunar daily variation, L. Figure 34 shows an extreme example of the
lunar effects constructed from records from the magnetic observatory at Huancayo, Peru,
located at an altitude of 3350 m on the magnetic equator. L is evident as small humps in
the Sq minima between the full moon (new moon) and the last quarter (first quarter). L
may exceed 10-20 '" at the magnetic equator, but generally averages less than 5 ",. Like
the ocean tides, L actually has a semidiurnal period (Matsushita, 1967). The semimonthly
modulation of Sq in Fig. 34 is around 40 ",.
Brown (1972, 1976) has suggested that "subtle geophysical Zeitgeber" or time cues,
primarily geoelectromagnetic, are important, if not critical, to the timing of biological
o MONTHS E MONTHS J MONTHS
00
50 -
40 o. •• 0
ú ? D ? ê =•••,...
30 ú f ~ ç = eo eo
20
10
y
00 c
10
20
30 .... ú =
-""'000111000
40
... 000.0 00000°",,-
ú
-- ú
=
=
50
LOCAL TIME
Figure 33. Horizontal component of Sq for D months (January, February, November, December), E
months (March, April, September, October), and J months (May, June, July, August). From Matsushita
(1967) with permission.
clocks. Kalmijn (1974) also suggests that the 24-hr periodicity of micropulsations and the
geomagnetic daily variations may function as Zeitgeber in synchronizing circadian (circa
about, diem a day) rhythms, for the electric fields associated with those magnetic phe-
nomena are often within the range of detection of electrosensitive fish. Lindauer and Martin
(1972) reported that the small directional error, or Missweisung, of the waggle dance of
honeybees was correlated with the local geomagnetic daily variation and further (Martin
92 Chapter 3
'"
ú ú ú ú =
à O V S MMú ú ú =
DJ J J DJ J J DJ J J DX Ñf ú î i ê q i ? DZ ã T ç ç WJ WWå DJ J J J J DJ J J J J DJ J ? DJ J J J DJ J J J J DWi ~ X WJ ë WWX WJ Dí Qvarter
=
New moon FtrJt Quarter
Figure 34. Calculated lunar modulation of the solar quiet daily variation in H at Huancayo, Peru. From
Chapman and Bartels (1940) with permission.
and Lindauer, 1977) that honeybees could, in the absence of more important cues such as
photoperiod, synchronize their circadian rhythms to the geomagnetic daily variation.
However, in the absence of more compelling evidence to the contrary, the notion that
natural geomagnetic variations can function as reliable Zeitgeber for biological clocks must
be viewed 'Yith skepticism. The Sq fields shown in Figs. 32 and 33 were obtained through
extensive numerical processing of worldwide data from magnetically quiet days. Chapman
and Bartels (1940) note that Fig. 34 is not a real record but "is freed from the non-cyclic
and fortuitous changes which affect the actual records." Indeed, it is more often the case
that the pure solar and lunar variations are either phase-shifted or largely obscured by the
frequent occurrence of various magnetic disturbances (Figs. 35 and 36). Given the com-
putational efforts which are required to extract the lunar and Sq variations from magnetic
observatory records, coupled with the reports that birds (Keeton et al., 1974; Moore, 1977)
and bees (Lindauer and Martin, 1972) are apparently unable to orientationally compensate
for small magnetic disturbances, it is doubtful that an organism would benefit by relying
on such noise-ridden signals to entrain its biological rhythms. As the signals and the noise
are generally of comparable amplitudes and time scales, an appeal to the often remarkable
signal-discriminating abilities of various organisms is unlikely to remove that doubt. The
day-to-day variability in the timing of the maxima and minima of geomagnetic variations
is hardly comparable to the day-to-day precision in the timing of sunrise and sunset, the
oceanic tides, or even of the more variable thermoperiod and atmospheric pressure vari-
ations. However, it was only a short time ago that a similar skepticism existed regarding
organismic magnetoreception in general (e.g., Davis, 1948; Varian, 1948; Griffin, 1952).
3.3. Magnetic Storms
Of the various disturbances of the magnetosphere, the most dramatic in terms of their
effect on the magnetic field at the earth's surface are magnetic storms. Typically, the most
severe portion of a storm lasts from 2 to 12 hr, during which time the magnetic field
intensity at the earth's surface varies by a few hundred gammas and the magnetic decli-
nation by a degree or so. Intensity perturbations of tens of gammas may persist for over a
day. A magnetic storm is a worldwide phenomenon, but its intensity varies with locality,
generally being greater at higher latitudes. Figure 36 shows the record of the great magnetic
storm of May 13-16, 1921, obtained at Potsdam, Germany. The following ranges in the
magnetic elements were observed: ú a = = 3°19', ú e == 1060 'Y, ú w = = 1100 'Y. For an even
greater storm on April 16, 1938, Potsdam recorded: ú a = = 5°28', ú e == 1900 'Y, ú w = = 600
'Y (Chapman and Bartels, 1940). However, such great storms are rare, occurring only once
every 10-20 years. As for the general frequency of occurrence of magnetic storms, Chapman
and Bartels (1940) report that during the 53 years, 1875-1927, there were 343 magnetic
storms with ranges between 30' and 60' in D, and between 150 'Y and 300 'Y in H or Z, and
z z
n o
:lIIlIlIlIImtiffllHilR!lH
. .IIIIIIIIIN_O
ò K n ú K j ú ú WWX X W=
a ú ú =
Figure 35. Comparison of magneto grams from Huancayo, Peru, and Watheroo, Australia, for magnet-
ically quiet days (upper panels) and magnetically disturbed days (lower panels) . From Chapman and
Bartels (1940) with permission.
60 storms with ranges greater than 60' in D and 300 -y in H or Z, for an average of between
7 and 8 storms per year. However, the occurrence of magnetic storms is correlated with
the ll-year sunspot cycle, and during years of sunspot minima the average dropped to
about 1 per year. (Geomagnetic storms often show 27-day recurrence sequences, presum-
ably as a result of the 27-day rotation period of the sun.)
A magnetic storm is a direct result of a solar flare. Solar flares occur at random, but
their frequency is correlated with the solar cycle. During years of sunspot maxima, small
flares may occur hourly and very large flares monthly. From the numbers given above, it
94 Chapter 3
v r+115
,
K Ñ ú =K ú =
ID n
14m
l.hli
r:riJ ú lL "
= ..I ="-
:.an: IMI N' II" X
I,
ú=
VII ú K=
IlK
I
J
" 1-11 'fJv,

11,1
I
tlift
!1M If\,
II. H ú =
r--
In)
"1\ "'--,
/, IN
., t. 'r 1"-'
v
----" , ,-,,-".-
ú =
"." úf =
ú K J
ú =
'"
"- IV
\.I
r e-:rL 1----
v ú ? =
---- r---
--
a 6 12 18
Figure 36. The great magnetic storm of May 1921 as recorded at Potsdam, Germany. From Chapman
and Bartels (1940) with permission.
is clear that not all solar flares cause perceptible magnetic storms. However, the lower
limit of 150 'Y used in tabulating storms was arbitrary, so the reader should not conclude
that stormlike magnetic disturbances of intensity below 150 'Y do not occur. Randomly
occurring solar flares of random intensities perturb the magnetosphere accordingly.
The generalized phenomenology of a magnetic storm is nicely illustrated by the record
for the magnetic north component X in Fig. 36. When a flare occurs, a cloud of energetic
protons and electrons is ejected from the sun and generally reaches the earth within 10-
36 hr. When the cloud strikes the magnetopause, the magnetosphere is abruptly compressed
and the geomagnetic field intensity increases, signaling the beginning of the magnetic
storm. The abrupt increase, called the sudden commencement (SC), occurs within a few
minutes at a given locality, and may reach 100 'Y or more at the earth's surface. The SC is
propagated through the magnetosphere as a hydromagnetic wave and so is detected at
different localities at times which may differ by several minutes. The SC in Fig. 36 occurred
shortly after 2200 hr on May 14.
As the cloud of solar particles continues to strike the magnetopause, the storm enters
the initial phase, during which the field intensity remains high or may even increase
further. During this time the number of geomagnetic ally trapped particles increases, which
eventually leads to an increase in the ring current and a corresponding decrease in the
field intensity at the earth's surface. This decrease begins the storm's main phase. During
the main phase the intensity may decrease several hundred gammas. Eventually, however,
the excess of trapped particles declines and the field slowly returns to its normal state.
This latter portion of the main phase is known as the recovery phase. During a polarity
transition, or at other times of reduced geomagnetic field intensity (Fig. 21) when the
magnetopause moves closer to the earth and the auroral zone expands to lower latitudes
(Siscoe et 01., 1976), the frequency and severity of magnetic storms detected at the earth's
surface should increase.
The morphology of a magnetic storm does not always follow such a well-defined course
(e.g., some storms begin gradually rather than suddenly), but the general result is the
same-the geomagnetic field is significantly perturbed. Walcott (1978) observed that pi-
geons released near magnetic anomalies which perturbed the local field by 3000-5000 'Y
were significantly less well oriented toward home than pigeons released at nonanomalous
sites. Such intensities are only an order of magnitude greater than those of moderate mag-
netic storms and only slightly larger than those of great magnetic storms. Even moderate
magnetic storm disturbances are considerably in excess of the "normal" geomagnetic dis-
turbances that have been reported to influence bird orientation (Keeton et 01.,1974; Moore,
1977). Additional, though unsubstantiated, reports of biological correlates with geomag-
netic activity exist (Tshernyshev, 1968), including a few involving humans (Dull and Dull,
1935; Friedman et 01., 1963, 1965). However, the reported effects on humans have been
discounted by Pokorny and Mefferd (1966).
3.4. Geomagnetic Indices
In order to depict the character of geomagnetic activity throughout the day, the month,
the year, or even longer periods, and to facilitate the comparison of magnetic activity data
with other geophysical or solar data, several numerical indices of geomagnetic activity
have been defined. As correlations between bird orientation and several geomagnetic in-
dices have been reported (Moore, 1977), it is important to understand how such indices
are derived and what physical meaning may be attached to them. Geomagnetic indices
have been clearly and concisely reviewed by Lincoln (1967). A far more detailed and
penetrating (but probably less useful to the nonspecialist) analysis is presented by Mayaud
(1980). The following discussion will be confined to the K index, as it is that index for
which biological correlates have been published (Southern, 1971; Keeton et 01., 1974;
Moore, 1977). It will become clear from the discussion that any inferences regarding the
biological influences of geomagnetic activity based upon the K index must be viewed with
caution.
The geomagnetic K (from the German, Kennziffers) index is a 3-hr range index which
is intended to be a measure of the solar particle flux incident upon the magnetosphere as
reflected in the geomagnetic activity caused by changes in ionospheric currents resulting
from that particle flux (Lincoln, 1967). Hence, the effects of the solar daily variation Sq
and the lunar daily variation L as well as effects resulting from the incidence of electro-
magnetic radiation from solar flares upon the magnetosphere (variously called the sfe, the
solar-flare effect, or the crochet) must be removed from the observed magnetic variations.
Beginning with midnight, for each 3-hr interval during the day, the difference in the max-
imum and minimum values of each magnetic element D, H, and Z or X, Y, and Z recorded
at a given magnetic observatory is determined. That portion of each difference which is
due to Sq, L, and sfe (sfe can be 60 'Y or more) is then subtracted to obtain the range a for
each element (see Fig. 37). The latter process is rather subjective as it requires an estimate
on the part of the observatory worker of the undisturbed daily variation appropriate to the
observatory, taking into account the time of year and the phase of the moon. From 1939
to 1963 the largest of the three ranges was taken as the basis of K. But owing to the
sensitivity of Z to solid-earth induction effects, the variation in Z has been excluded from
the determination of K since January 1, 1964.
co
C')
3 6 9 12 15 18 Zl wú = 3 6 9 12
ú Z Dä W` i D\ a í = I;:::;T'= ãú= ["'"
Figure 37. Procedure used to determine the K index for each 3-hr interval from a typical magnetic record. The dashed line represents the estimated
undisturbed geomagnetic daily variation produced by Sq and L. a represents the amplitude of variation for a given 3-hr interval after Sq and L have
been removed. From Mayaud (1980) with permission.
9
ú =
'C
Si
w
ú MMê J ê J J J J J K J J J J J ê J J J J J ê J J J J J ê J J J J J ê J J J J J ú =
ú =2000
ú =
:::iE
...J
II:
III
ú =
g 1000
_.-
-
. .---. .
. ....
90"5 30°5
GEOMAGNETIC LATITUDE
Figure 38. Variation of the lower limit in gammas for K = 9 for several geomagnetic observatories
located at various geomagnetic latitudes. From Chernosky et a1. (1965) with permission.
For each 3-hr interval an integer from 0 to 9, the K value is assigned to the largest
value of a according to a permanent conversion scale appropriate to the observatory. For
the standard, midlatitude observatory at Niemegk, Germany (52°04'N, 12°40'E), the scale
is
K01 2 3 4 5 6 7 8 9
lower limit of a (-y) 0 5 10 20 40 70 120 200 330 500
Thus, if a is less than 5 'Y, K = O. If 120 'Y :::; a < 200 'Y, K = 6, etc. As the amplitude of
geomagnetic activity depends upon latitude and to a lesser degree upon longitude, the K
scales for different observatories are chosen accordingly. The scales are chosen by mul-
tiplying the 'Y range for each K value by a constant in order to make the frequency distri-
butions of K values as nearly the same as possible at all observatories. Figure 38 shows
the lower limit for K = 9 as a function of latitude for many observatories.
Several important properties of K should be noted. (1) It contains a subjective element
(the elimination of Sq, L, and sfe) which can be important for low K values. (2) K is a
measure only of the geomagnetic variation within 3 hr and except for very large K values,
often fails to adequately reflect how much the field is perturbed from its normal value.
For example, for the minimum value of H in Fig. 37 (the fifth box), K is based on a 3-hr
variation of about 80 'Y and does not reflect the fact that H is about 200 'Y from its undisturbed
value (the dashed line). (3) By excluding Z from consideration, K may considerably un-
derestimate the actual maximum magnetic variation. (4) A given K value is strictly ap-
propriate only to the locality where it was obtained. By using K values from an observatory
located 100 miles or more from the experimental site, one faces the very real possibility
that local solid-earth conductivity anomalies (which are distinct from magnetic anomalies)
or other effects render the observatory K index inappropriate. The dangers involved are
no less than the dangers encountered in using an IGRF map to determine the values of
98 Chapter 3
local magnetic field elements. The biomagnetic investigator would thus be wise to obtain
a few magnetic records at the experimental site for comparison with observatory records.
(5) Because K is designed primarily as a measure of solar particle flux rather than geo-
magnetic activity, inferences regarding the biological influences of geomagnetic activity
based on values of K are fraught with more than the usual perils of inferring causality from
statistical correlation. Some other correlate of solar activity such as subtle weather phe-
nomena (see, e.g., Wilcox et al., 1974) may well be responsible for the biological effect.
References
Barnothy, M. F., (ed.), 1969, Biological Effects of Magnetic Fields, Volume 2, Plenum Press, New York.
Barnwell, F. H., and Brown, F. A., Jr., 1961, Magnetic and photic responses in snails, Experientia
4:513.
Bergh, H. W., 1970, Paleomagnetism of the Stillwater Complex, Montana, in: Paleogeophysics (S. K.
Runcorn, ed.), Academic Press, New York, pp. 143-158.
Black, D. I., 1967, Cosmic ray effects and faunal extinctions at geomagnetic field reversals, Earth Planet.
Sci. Lett. 3:225-236.
Bliss, V. L., and Heppner, F. H., 1976, Circadian activity rhythm influenced by near zero magnetic
field, Nature 261:411-412.
Bonhommet, N., and Zahringer, J., 1969, Paleomagnetism and potassium argon age determinations of
the Laschamp geomagnetic polarity event, Earth Planet.. Sci. Lett. 6:43-46.
Briden, J. C., and Irving, K, 1964, Paleoclimatic spectra of sedimentary paleoclimatic indicators, in:
Problems in Paleoclimatology (A. K M. Nairn, ed.), Interscience, New York, pp. 199-250.
Brock, A., 1971, An experimental study of paleosecular variation, Geophys. J. R. Astron. Soc. 24:303-
317.
Brown, F. A., Jr., 1962, Responses of the planarian Dugesia, and the protozoan, Paramecium, to very
weak horizontal magnetic fields, Biol. Bull. 123:264-281.
Brown, F. A., Jr., 1972, The "clocks" timing biological rhythms, Am. Sci. 60:756-766.
Brown, F. A., Jr., 1976, Biological clocks: Endogenous cycles synchronized by subtle geophysical
rhythms, BioSystems 8:67-81.
Bullard, E. C., 1949, The magnetic field within the earth, Proc. R. Soc. London Ser. A 197:433-453.
Busse, F. H., 1978, Magnetohydrodynamics of the earth's dynamo, Annu. Rev. Fluid Mech. 10:435-
462.
Busse, F. H., 1980, Motions in the earth's core and the origin of geomagnetism, in: Physics of the
Earth's Interior, LXXVIII Corso, Soc. Italiana di Fisica, Bologna, Italy, pp. 493-507.
Busse, F. H., 1983, Recent developments in the dynamo theory of planetary magnetism, Ann. Rev.
Earth Planet. Sci. 11:241-268.
Butler, R. F., and Opdyke, N. D., 1979, Magnetic polarity stratigraphy, Rev. Geophys. Space phys.
17:235-244.
Cain, J. c., and Hendricks, S. J., 1968, The Geomagnetic Secular Variation, NASA Tech. Note D-4527.
Campbell, W. H., 1967, Geomagnetic pulsations, in: Physics of Geomagnetic Phenomena, Volume 2
(S. Matsushita and W. H. Campbell, eds.), Academic Press, New York, pp. 821-909.
Carmichael, C. M., 1967, An outline of the intensity of the paleomagnetic field of the earth, Earth
Carmichael, C. M. (ed.), 1977, Paleomagnetic Field Intensity: Its Measurement in Theory and Practice,
Phys. Earth Planet. Inter. 13.
Chan,1. S., and Alvarez, W., 1983, Magnetic polarity stratigraphy, Rev. Geophys. Space Phys. 21:620-
626
Chapman, S. and Bartels, J., 1940, Geomagnetism, Volumes 1 and 2, Oxford University Press (Clar-
endon), London.
Chernosky, E. J., Fougere, P. F., and Hutchinson, R. 0., 1965, The geomagnetic field, in: USAF Hand-
book of Geophysics and Space Environments (S. 1. Valley, ed.), Air Force Cambridge Research
Laboratories, Office of Aerospace Research, USAF, pp. 11-1-11-61.
Cox, A., 1968, Lengths of geomagnetic polarity intervals, J. Geophys. Res. 73:3247-3260.
Cox, A., 1969, Geomagnetic reversals, Science 163:237-245.

Cox, A., 1975, The frequency of geomagnetic reversals and the symmetry of the non dipole field, Rev.
Cox, A., 1983, Magnetostratigraphic time scale, in: A Geologic Time Scale (W. B. Harland, A. V. Cox,
P. G. Llewellyn, A. G. Smith, and R Walters, eds.), Cambridge University Press, London.
Cox, A., Doell, R R, and Dalrymple, G. S., 1964, Reversals of the earth's magnetic field, Science
144:1537-1543.
Davis, 1., 1948, Remarks on: "The physical basis of bird navigation," J. Appl. Phys. 19:157-160.
Denham, C. R, and Cox, A., 1971, Evidence that the Laschamp polarity event did not occur 13,300-
30,400 years ago, Earth Planet. Sci. Lett. 13:181-190.
Dodson, R, Dunn, J. R, Fuller, M., Williams, I., Ito, H., Schmidt, V. A., and Wu, Y.-M., 1978, Paleo-
magnetic record of a late Tertiary field reversal, Geophys. J. R. Astron. Soc. 53:373-412.
Doell, R R, and Cox, A., 1961, Paleomagnetism, Adv. Geophys. 8:221-313.
Dull, T., and Dull, B., 1935, Zuasammenhange zwichen Storungen des Erdmagnetismus and Haufungen
von Todesfallen, Dtsch. Med. Wochenschr. 61:95.
Durrani, S. A., and Kahn, H. A., 1971, Ivory Coast microtektites, fission track age and geomagnetic
reversals, Nature 232:320-323.
Elsasser, W. M., 1946, Induction effects in terrestrial magnetism. Part I. Theory, Phys. Rev. 69:106-
116.
Elsasser, W. M., 1963, Early history of the earth, in: Earth Sciences and Meteroritics U. Geiss and E.
D. Goldbert, eds.), North-Holland, Amsterdam, pp. 1-30.
Force, E. R, 1984, A relation among geomagnetic reversals, seafloor spreading rate, paleoclimate, and
black shales, ES, Trans. Am. Geophys. Union 65:18-19.
Science 203:1355-1356.
Frankel, R B., Blakemore, R P., Torres de Aranjo, F. F., Esquivel, D. M. S., and Danon, J., 1981,
Magnetotactic bacteria at the geomagnetic equator, Science 212:1269-1270.
Friedman, H., 1983, The legacy of the IGY, Eos, Trans. Am. Geophys. Union 32:497-499.
Friedman, H., Becker, R E., and Bachman, C. H., 1963, Geomagnetic parameters and psychiatric hos-
pital admissions, Nature 200:626-628.
Friedman, H., Becker, R E., and Bachman, C. H., 1965, Psychiatric ward behavior and geophysical
parameters, Nature 213:949-956.
Fuller, M., Williams, I., and Hoffman, K., 1979, Paleomagnetic records of geomagnetic field reversals
and the morphology of the transitional fields, Rev. Geophys. Space Phys. 17:179-205.
Gaibar-Puertas, c., 1953, Varacion secular del campo geomagnetico, Tortosa, Spain: Observ. del Ebro
Memo. No. 11.
Gilbert, Wm. of Colchester, 1600, De Magnete, magneticisque Corporibus, et de magno Magnete Tel-
lure; Physiologia nova, plurimus &' argumentis &' experimentis demonstrata, Peter Short, London.
Glass, B., and Heezen, B. C., 1967, Tektites and geomagnetic reversals, Nature 214:372.
Gould, J. L., Kirschvink, J. L., and Deffeyes, K. S., 1978, Bees have magnetic remanence, Science
201:1026-1028.
Griffin, D. R, 1952, Bird navigation, BioI. Rev. 27:359-400.
Hahn, A. G., 1971, Types of magnetic anomalies measured on land and general aspects of their geo-
logical meaning, in: World Magnetic Survey 1957-1969 (A. J. Zmuda, ed.), International Asso-
ciation of Geomagnetism and Aeronomy, pp. 134-143.
Harrison, G. C. A., 1968, Evolutionary processes and reversals of the earth's magnetic field, Nature
217:46-47.
Harrison, G. C. A., and Funnell, B. M., 1964, relationship of paleomagnetic reversals and micropa-
leontology in two late Cenozoic cores from the Pacific Ocean, Nature 204:566.
Harrison, G. C. A., and Somayajulu, B. L. K., 1966, Behavior of the earth's magnetic field during a
reversal, Nature 212:1193-1195.
Hays, J. D., 1971, Faunal extinctions and reversals of the earth's magnetic field, Geol. Soc. Am. Bull.
82:2433-2447.
Hays, J. D., and Opdyke, N. D., 1967, Antarctic Radiolaria, magnetic reversals and climatic change,
Science 158:1001-1011.
Hays, J. D., Saito, T., Opdyke, N. D., and Burckle, L., 1969, Pliocene/Pleistocene sediments of the
equatorial Pacific: Their paleomagnetic, biostratigraphic, and climatic record, Geol. Soc. Am. Bull.
80:1481-1514.
100 Chapter 3
Heirtzler, J. R., Dickson, G. 0., Herron, E. M., Pitman, W. C., III, and LePichon, X., 1968, Marine
magnetic anomalies, geomagnetic field reversals and motions of the ocean floor and continents,
J. Geophys. Res. 73:2119-2136.
Herzenberg, A., 1958, Geomagnetic dynamos, Philos. Trans. R. Soc. London Ser. A 250:543-585.
Hillhouse, J., and Cox, A., 1976, Brunhes-Matuyama polarity transition, Earth Planet. Sci. Lett. 29:51-
64.
Hoffman, K. A., an,d Fuller, M., 1978, Transitional field configurations and geomagnetic reversal,
Nature 273:715-718.
Irving, E., and Pullaiah, G., 1976, Reversals of the geomagnetic field, magnetostratigraphy, and relative
magnitude of paleosecular variation in the Phanerozoic, Earth-Sci. Rev. 12:35-64.
Jacobs, J. A., 1963, The Earth's Core and Geomagnetism, Pergamon Press, Elmsford, N.Y.
Jacobs, J. A., 1981, Heat flow and reversals of the earth's magnetic field, J. Geomagn. Geoelectr. 33:527-
529.
Kalmijn, A. J., 1974, The detection of electric fields from inanimate and animate sources other than
electric organs, in: Handbook of Sensory Physiology, Volume III/3 (A. Fessard, ed.l, Springer-
Verlag, Berlin, pp. 147-200.
Kalmijn, A. J., 1978, Experimental evidence of geomagnetic orientation in elasmobranch fishes, in:
Animal Migration, Navigation, and Homing (K. Schmidt-Koenig and W. T. Keeton, eds.l, Springer-
Kalmijn, A. J., 1982, Electric and magnetic field detection in elasmobranch fishes, Science 218:916-
918.
Keeton W. T., 1971, Magnets interfere with pigeon homing, Proc. Natl. Acad. Sci. USA 68:102-106.
Keeton, W. T., 1974, The orientational and navigational basis of homing in birds, Adv. Study Behav.
5:47-132.
Keeton, W. T., Larkin, T. S., and Windsor, D. M., 1974, Normal fluctuations in the earth's magnetic
field influence pigeon orientation, J. Compo Physiol. 95:95-103.
Kennett, J. P., and Watkins, N. D., 1970, Geomagnetic polarity change, volcanic maxima and faunal
extinction in the South Pacific, Nature 227:930-934.
King, J. W., 1974, Weather and the earth's magnetic field, Nature 247:131-134.
Lett. 44:193-204.
Kuterbach, D. A., Walcott, B., Reeder, R. J., and Frankel, R., B., 1982, Iron-containing cells in the honey
bee (Apis mellifera), Science 218:695-697.
LaBrecque, J. 1., Kent, D. V., and Cande, S. c., 1977, Revised magnetic polarity time scale for late
Cretaceous and Cenozoic time, Geology 5:330-335.
Larson, R. 1., and Hilde, T. W. c., 1975, A revised time scale of magnetic reversals for the early
Cretaceous and late Jurassic, J. Geophys. Res. 80:2586-2594.
Lincoln, J. V., 1967, Geomagnetic indices, in: Physics of Geomagnetic Phenomena, Volume 1 (S.
Matsushita and W. H. Campbell, eds.), Academic Press, New York, pp. 67-100.
Lindauer, M., and Martin, H., 1972, Magnetic effect on dancing bees, in: Animal Orientation and
Navigation (S. R. Galler, K. Schmidt-Koenig, G. J. Jacobs, and R. E. Belleville, eds.l, NASA SP-
262, U.S. Government Printing office, Washington, D. C.
Lowes, F. J., and Wilkinson, 1., 1963, Geomagnetic dynamo: A laboratory model, Nature 198:1158-
1160.
McDonald, K. L., and Gunst, R. H., 1967, An analysis of the earth's magnetic field from 1835 to 1965,
ESSA Tech. Rep. IER 46-IES 1, U.S. Deparment of Commerce, Boulder, Colo.
McDougall, 1., 1979, The present status of the geomagnetic polarity time scale, in: The Earth: Its Origin,
Structure, and Evolution (M. W. McElhinny, ed.l, Academic press, New York.
McElhinny, M. W., 1969, The paleomagnetism of the Permian of southeast Australia and its significance
regarding the problem of intercontinental correlation, Spec. Publ. Geol. Soc. Aust. 2:61-67.
McElhinny, M. W., 1971, Geomagnetic reversals during the Phanerozoic, Science 172:157-159.
McElhinny, M. W., 1973, Paleomagnetism and Plate Tectonics, Cambridge University Press, London.
McElhinny, M. W., 1979, Paleogeomagnetism and the core-mantle interface, in: The Earth: Its Origin,
Structure, and Evolution (M. W. McElhinny, ed.l, Academic Press, New York.
McElhinny, M. W., and Evans, M. E., 1968, An investigation of the strength of the geomagnetic field
in the early Precambrian, Phys. Earth Planet. Inter. 1:485-497.
McElhinny, M. W., and Merrill, R. T., 1975, Geomagnetic secular variation over the past 5 m.y., Rev.
McElhinny, M. W., and Senanayke, W. E., 1980, Paleomagnetic evidence for the existence of the
geomagnetic field 3.5 GA ago, J. Geophys. Res. 85:3523-3528.
McPherron, R. L., 1979, Magnetospheric substorms, Rev. Geophys. Space Phys. 17:657-681.
Malkus, W. V. R., 1968, Precession of the earth as the cause of geomagnetism, Science 160:259-264.
Mankinen, E. A., and Dalrymple, G. B., 1979, Revised geomagnetic polarity time scale for the interval
0-5 m.y. B.P., J. Geophys. Res. 84:615-626.
Martin, H., and Lindauer, M., 1977, Der Einfluss der Erdmagnetfelds and die Schwereorientierung der
Honigbiene, J. Compo Physiol. 122:145-187.
Mathews, J. H., and Gardner, W. K., 1963, Field reversals of 'paleomagnetic' type in coupled disk
dynamos, U.S. Naval Res. Lab. Rep. 5886.
Matsushita, S., 1967, Solar quiet and lunar daily variation fields, in: Physics of Geomagnetic phe-
nomena (S. Matsushita and W. H. Campbell, eds.l, Volume 1, Academic Press, New York, pp.
301-424.
Matsushita, S., and Campbell, W. H. (eds.l, 1967, Physics of Geomagnetic Phenomena, Volumes 1 and
2, Academic Press, New York.
Matsushita, S., and Maeda, H., 1965, On the geomagnetic solar quiet daily variation field during the
IGY, J. Geophys. Res. 70:2535-2558.
Mayaud, P. N., 1980, Derivation, Meaning, and Use of Geomagnetic Indices, Geophysical Mono-
graph 22, American Geophysical Union, Washington, D.C.
Merrill, R. T., and McElhinny, M. W., 1983, .The Earth's Magnetic Field: Its History, Origin, and
Planetary Perspective, Academic Press, New York.
Merrill, R. T., McElhinny, M. W., and Stevenson, D. J., 1979, Evidence for long-term asymmetries in
the earth's magnetic field and possible implications for dynamo theories, Phys. Earth Planet. Inter.
20:75-82.
Moore, F. R., 1977, Geomagnetic disturbance and the orientation of nocturnally migrating birds, Sci-
ence 196:682-6684.
Needham, J., 1962, Science and Civilization in China, Volume 4, Cambridge University Press, London.
Newell, N. D., 1963, Crises in the history of life, Sci. Am. 208:76-92.
Opdyke, N. D., and Henry, K. W., 1969, A test of the dipole hypothesis, Earth Planet. Sci. Lett. 6:139-
151.
Opdyke, N. D., Glass, G., Hays, J. D., and Foster, J., 1966, Paleomagnetic study of Antarctic deep-sea
cores, Science 154:349-357.
Opdyke, N. D., Kent, D. V., and Lowrie, W., 1973, Details of magnetic polarity transitions recorded
in a high deposition rate deep-sea core, Earth Planet. Sci. Lett 20:315-324.
Palmer, J. D., 1963, Organismic spatial orientation in very weak magnetic fields, Nature 198:1061-
1062.
Phillips, J. B., 1977, Use of the earth's magnetic field by orienting cave salamanders (Eurycea lucifugal,
J. Compo Physiol. A 273:273-288.
Pitman, W. C., III, and Hayes, D. E., 1968, Sea-floor spreading in the Gulf of Alaska, J. Geophys. Res.
73:6571-6580.
Pokorny, A. D., and Mefferd, R. B., Jr., 1966, Geomagnetic fluctuations and disturbed behavior, J. Nerv.
Ment. Dis. 143:140-151.
Prevot, M., 1977, Large intensity changes of the nondipole field during a polarity transition, Phys.
Earth Planet. Inter. 13:342-345.
Price, A. T., 1967, Electromagnetic induction within the earth, in: Physics of Geomagnetic Phenomena,
Volume 1 (S. Matsushita and W. H. Campbell, eds.l, Academic Press, New York, pp. 235-298.
Quinn, T. P., Merrill, R. T., and Brannon, E. L., 1981, Magnetic field detection in sockeye salmon, J.
Exp. Zool. 217:137-142.
Ratcliffe, J. A., 1972, An Introduction to the Ionosphere and Magnetosphere, Cambridge University
Press, London.
Reid, G. C., Isaksen, I. S. A., Holzer, T. E., and Crutzen, P. J., 1976, Influence of ancient solar-proton
events on the evolution of life, Nature 259:177-179.
Rikitake, T., 1958, Oscillations of a system of disk dynamos, Proc. Cambridge Philos. Soc. 54:89-105.
Rikitake, T., 1964, Outline of the anomaly of geomagnetic variations in Japan, J. Geomagn. Geoelectr.
15:181-184.
102 Chapter 3
Ringwood, A. E., 1979, Composition and origin of the earth, in: The Earth: Its Origin, Structure, and
Evolution (M. W. McElhinny, ed.). Academic Press, New York, pp. 1-58.
Roederer, J. G., 1974, The earth's magnetosphere, Science 183:37-46.
Sagan, C., 1965, Is the early evolution of life related to the development of the earth's core?, Nature
206:448.
Schopf, J. W., and Oehler, D. Z., 1976, How old are the eukaryotes?, Science 193:47-49.
Shaw, J., 1977, J:urther evidence for a strong intermediate state of the paleomagnetic field, Geophys.
J. R. Astron. Soc. 48:263-269.
Simpson, J. F., 1966, Evolutionary pulsations and geomagnetic polarity, Geol. Soc. Am. Bull. 77:197-
204.
Siscoe, G. L., Chen, C.-K., and Harel, M., 1976, On the altitude ofthe magnetopause during geomagnetic
reversals, J. Atmos. Terr. Phys. 38:1327-1331.
Skiles, D. D., 1970, A method of inferring the direction of drift of the geomagnetic field from paleo-
magnetic data, J. Geomagn. Geoelectr. 22:441-462.
Smith, P. J., 1967, Intensity of the earth's magnetic field in the geological past, Nature 216:989-990.
Southern, W. E., 1971, Gull orientation by magnetic cues: A hypothesis revisted, Ann. N.Y. Acad. Sci.
188:295-311.
Stevenson, D. J., 1981, Models of the earth's core, Science 214:611-619.
Strangway, D. W., 1970, History of the Earth's Magnetic Field, McGraw-Hill, New York.
Thellier, E., and Thellier, 0., 1959, Sur I'intensite du champ magnetique terrestre dans Ie passe his-
torique et geologique, Ann Geophys. 15:285-376.
Tshernyshev, W. B., 1968, Disturbances of the magnetic field of the earth and biological rhythms of
the beetle Trogoderma, Zh. Obshch. BioI. 29:719-722 (in Russian with English summary).
Uffen, R. J., 1963, Influence of the earth's core on the origin and evolution of life, Nature 198:143-
144.
Van Allen, J. A., 1959, The geomagnetic ally trapped corpuscular radiation, J. Geophys. Res. 64:1683-
1689.
Varian, R. H., 1948, Remarks on: "A preliminiary study of a physical basis of bird navigation," J. Appl.
Phys.19:306-307.
Verhoogen, J., 1961, Heat balance of the earth's core, Geophys. J. R. Astron. Soc. 4:276-281.
Vogt, P. R., 1975, Changes in geomagnetic reversal frequency at times of tectonic changes: Evidence
for coupling between core and upper mantle processes, Earth Planet. Sci. Lett. 25:313-321.
Waddington, C. J., 1967, Paleomagnetic field reversals and cosmic radiation, Science 158:913-915.
Walcott, C., 1978, Anomalies in the earth's magnetic field increase the scatter of pigeon vanishing
bearings, in: Animal Migration, Navigation, and Homing (K. Schmidt-Koenig and W. T. Keeton,
eds.), Springer-Verlag, Berlin, pp. 143-151.
Walcott, C., and Green, R. P., 1974, Orientation of homing pigeons altered by a change in the direction
of the applied magnetic field, Science 184:180-182.
Watkins, N. D., and Goodell, H. G., 1967, Geomagnetic polarity change and faunal extinction in the
southern ocean, Science 156:1083-1089.
Wikswo, J. P., Jr., Barach, J. P., and Freeman, J. A., 1980, Magnetic field of a nerve impulse: First
measurements, Science 208:53-55.
Wilcox, J.M., Scherrer, P. H., Svalgoard, 1., Roberts, W.O., Olsen, R. H., and Jenne, R. L., 1974, Influence
of solar magnetic sector structure on terrestrial atmospheric vorticity, J. Atmos. Sci. 31:581-588.
Wollin, G., Kukla, G. J., Ericson, D. B., Ryan, W. B. F., and Wollin, J., 1973, Magnetic intensity and
climatic changes 1925-1970, Nature 242:34-37.
Zmuda, A. J. (ed), 1971, World Magnetic Survey 1957-1969, International Association of Geomagne-
tism and Aeronomy, Bulletin No. 28.
Zoeger, J., Dunn, J. R., and Fuller, M., 1981, Magnetic material in the head of the Pacific dolphin,
Science 213:892-894.
Chapter 4
An Introduction to the Use of SQUID
Magnetometers in Biomagnetism
M. FULLER, W. S. GOREE, and W. L. GOODMAN
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
2. Operating Principles of SQUIDs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
2.1. The Superconducting State and Magnetic Field Detection by SQUIDs. . . . . . . . . . 104
2.2. The London and Josephson Equations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
2.3. SQUID Geometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
2.4. SQUID Electronics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
2.5. Flux Transformer Coupling of SQUIDs to B Fields . . . . . . . . . . . . . . . . . . . . . 116
2.6. Superconducting Magnetic Shields. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3. Cryogenics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
3.1. Conduction Heat Leak. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
3.2. Radiation Heat Leak. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 123
3.3. Vapor-Cooled Shields. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.4. Dewar Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 124
3.5. Gas Diffusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
3.6. Liquid Helium Storage and Transfer. . . . . . . . . . . . . . . . . . . . . . . . . . . 126
3.7. Safety and Neck Plug Hazards . . . . . . . . . . . . . . . . . . . . . . , . . . . . . . . . . 128
3.8. Cryocoolers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4. Instrument Configurations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
4.1. Magnetometers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 129
4.2. Gradiometers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.3. Rock Magnetometers and Susceptometers. . . . . . . . . . . . . . . . . . . . . . . . . .. 133
5. Applications of SQUID Magnetometers in Biomagnetism . . . . . . . . . . . . . . . . . . . . 136
5.1. Isolation of Magnetic Material in Biological Samples. . . . . . . . . . . . . . . . . . . . 136
5.2. Identification of Magnetic Material in Biological Samples. . . . . . . . . . . . . . . . . 137
5.3. Characterization of Magnetic Phases in Biological Material . . . . . . . . . . . . . . . . 142
6. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
1. Introduction
The advent of magnetometers with superconducting quantum interference device (SQUID)

sensors has brought about a major breakthrough in magnetic field measurement. SQUID
magnetometers have permitted field determinations which are orders of magnitude more
sensitive than previous measurements. They have also been used in magnetometers de-
signed to measure the magnetic moments of small specimens, such as the standard rock
M. FULLER • Department of Geological Sciences, University of California, Santa Barbara, California

93106. W. S. GOREE and W. 1. GOODMAN • 2-G Enterprises, Mountain View, California 94043.
103
104 Chapter 4
cylinders used in paleomagnetism and biological samples which may contain minute quan-
tities of magnetite, of interest to biomagnetists. In the absence of an inducing field, the
remanent moment of the specimen is measured. In the presence of a field, the induced
moment can be measured giving the susceptibility. SQUIDs have also been used in gra-
diometers to obtain magnetocardiograms and magnetoencephalograms. In the latter ap-
plication, fields of the order of femtoteslas (microgammas) are measured!
There are by now numerous descriptions of SQUIDs and their operation (e.g., Jaklevic
et 01., 1965; Zimmerman et 01., 1970; Mercereau, 1970; Nisenoff, 1970; Josephson, 1974;
Clarke, 1974). In addition, there are excellent descriptions of both SQUIDs and SQUID
magnetometers in recent review articles on biomagnetic instrumentation (e.g., Romani et
01., 1982; Erne', 1983). In the present paper, we attempt to provide a practical guide to the
use of SQUID magnetometers.
Section 2 begins with a historical development of SQUID magnetometers, including
a qualitative description of the operation of SQUIDs, designed to provide an elementary
introduction to the magnetometers. A more detailed discussion of the SQUID principles
follows, with a derivation of the London and Josephson equations and coverage of more
practical matters such as the geometry of the various SQUIDs used and their associated
electronics. Section 3 deals with cryogenics. Section 4 covers the actual magnetometer
configurations and includes discussions of field coupling to the SQUIDs, superconducting
shields, and cryogenics. In Section 5, applications of the magnetometers in biomagnetism
are described.
The discussion of instruments in Section 4 is centered around those used in the Rock
Magnetism and Paleomagnetism Laboratory at UCSB, which have been for the most part
designed and built by the 2-G Enterprises company and its various antecedents. There are
now other commercial suppliers of SQUID magnetometers and many groups have con-
structed their own. References to those newer instruments are given.
2. Operating Principles of SQUIDs

2.1. The Superconducting State and Magnetic Field Detection by
SQUIDs
The development of the SQUIDs followed theoretical and experimental work on su-
perconductivity, which culminated in the late 1950s and early 1960s in the BCS theory
of superconductivity (Bardeen, Cooper, and Schrieffer, 1957) and in the prediction (Jo-
sephson, 1962) and observation (Anderson and Rowell, 1963) of electron pair tunneling.
The phenomenon of superconductivity was discovered in 1911 by Kamerlingh Dnnes
shortly after he succeeded in liquifying helium. He found that the resistance of a sample
of mercury dropped to an immeasurably low value just above 4°K. Subsequent to this
discovery, many other superconducting metals have been discovered; two of the most
useful for instrument applications are lead (transition temperature 7.2°K) and niobium
(transition temperature 9.2°K).
In the superconducting state, metals are able to transport a dc current with zero re-
sistance. To understand this phenomenologically, assume that when a metal is cooled
below its transition temperature, a fraction of the electrons condense into the supercon-
ducting state. The colder the metal, the larger the fraction of condensed electrons becomes.
This model is commonly known as the two-fluid model.
When a dc field is impressed on a superconductor, the current is carried exclusively
by the superelectrons which are able to move through the lattice without scattering. When
an ac field is impressed, some energy is imparted to the normal electrons and hence the
resistance is not zero in this case.
Use of SQUID Magnetometers in Biomagnetism 105
Another important characteristic of suprconductors is that they spontaneously exclude

magnetic flux-the Meissner effect. When a superconducting body is cooled below its
.transition temperature in the presence of a magnetic field, currents are spontaneously
generated in the sphere so as to exclude any magnetic field from the interior of the body.
We know today that the Meissner effect is generally not perfect because much of the flux
penetrating a metallic body when it becomes superconducting is trapped in it. The flux
becomes pinned to local impurity sites in the metal and more energy is required to drive
the flux from the pinning site than is gained by removing it from the superconductor
volume. Superconducting shields therefore tend to trap the ambient field upon cooling
instead of excluding it (Meissner effect). Note, however, that because of the Meissner effect,
substantial field modification occurs upon cooling through the transition temperature.
The response to applied magnetic fields defines two types of superconductors, type
1 and 2. Type 1 superconductors are the mechanically soft elements such as lead, tin, and
niobium, which have transition temperatures below 9°K. They exhibit perfect flux shield-
ing in fields below the critical field He, which is generally less than 1000 gauss for most
type 1 superconductors. Type 2 superconductors are mechanically harder materials and
are for the most part alloys and compounds. They have higher critical fields of up to 105
gauss and critical temperatures of up to 21SK. Perfect flux shielding is only exhibited by
these materials in fields of up to about 102 gauss; in higher fields, the flux starts to enter
the material. This field value is the lower critical field Hel . The material is still super-
conducting and the flux continues to penetrate to the higher critical field He2 at which
point superconductivity is destroyed. In low-field applications such as the rock magne-
tomer, to be described later, one uses type 1 superconductors to provide magnetic shielding.
In contrast, in high-field applications such as superconducting magnets, one uses type 2
superconductors. Most SQUID sensors are constructed with type 2 materials, particularly
work hardened niobium and niobium-titanium.
From a phenomenological viewpoint, superconductivity can be understood by assum-
ing that below the transition temperature a fraction of the electrons in a metal condense
into a single-quantum macrostate which can be described by a single wavefunction. The
idea of such a macrostate was initially proposed by Fritz London, and was based upon his
concept of long-range order in momentum space. Subsequently, Ginzburg and London
further developed the concept by introducing an order parameter into the theory to char-
acterize the degree of superconductivity.
The microscopic theory of superconductivity was developed by Bardeen et 01. (1957)
and was based upon an electron pairing concept. In explaining superconductivity, BCS
theory postulates the existence of a weak attractive force between electrons in the super-
conductor. This is the long-range order, whose importance had earlier been appreciated
by London. The idea of correlated motion can be used to explain superconductivity, or
the absence of normal electrical resistance, in a simple qualitative way. Normal electrical
resistance is the result of scattering of the conduction electrons by the lattice. If the motion
of all of the electrons is correlated, then scattering of one electron would involve scattering
of the whole assembly of electrons taking part in the correlated motion. Such scattering
has very low probability. Therefore, when a supercurrent flows, it is not significantly
affected by the normal scattering that constitutes electrical resistance. London had not,
however, appreciated that it is the electron pair (rather than the single electron) that enters
into the correlated motion. This pair coupling is central to BCS theory. The coupling gives
rise to a lower-energy state, separated from the excited states by an energy gap. The physical
basis of the coupling is through the intermediary of the lattice; loosely speaking, one elec-
tron is attracted to the wake of the other. Such coupling is very weak as is evident from
the low temperature at which it is broken by thermal energy. The electron pairs are known
as Cooper pairs.
106 Chapter 4
TRANSPORT CURRENT
I
SUPERCONDUCTOR 2 - - . , DIELECTRIC
BARRIER
ú f l ^ = THICK
SUPERCONDUCTOR I J J WXXXç ú J J f =J WX D ú WWWWWX X O> > ú I K K E =
II II
a
Figure 1. Josephson junction.
Flux quantization in a superconducting ring was proposed by London (1950) and

observed by Deaver and Fairbank (1961) and Doll and Nabauer (1961). It plays an important
role in the SQUID magnetometers. When a superconducting ring is cooled through its
transition temperature in the presence of a magnetic field, flux will be excluded from the
material of the ring due to the Meissner effect and flux will be trapped in the center of
the ring. This trapped flux can only take on certain values which are ã ì ä í á é ä É ú = of the flux
quantum <1>0 = h/2e, where h is Planck's constant, 6.624 X 10- 34 J sec, and e the charge
of the electron, 1.602 x 10- 19 C. Persistent currents will flow in the circuit as it cools
through its transition temperature and becomes superconducting, if the flux does not have
the correct value. The flux quantum has a numerical value of 2.07 x 10- 15 Wb, or 2.07
x 10- 7 Mx. Thus, the persistent currents in superconducting rings are induced by ex-
tremely low levels of magnetic flux.
The phenomenon of flux quantization did not achieve much practical importance,
until the ideas developed by Josephson resulted in the development of practical devices
for detecting magnetic fields. Josephson (1962) pointed out that a supercurrent should be
able to pass across a thin dielectric barrier separating two superconductors. The essential
requirement is that the quantum mechanical wavefunctions for the Cooper pairs in the
two superconductors extend beyond the surface of the superconductors and coherently
link up across the barrier.
A typical Josephson junction consisting of two superconductors separated by a thin
dielectric barrier is shown in Fig. 1. Above the critical temperature of the superconductors,
the device has a resistance of a few ohms. Below the transition temperature, this resistance
vanishes and the tunneling of the Cooper pairs gives a zero voltage current across the
device.
To understand the current, we return to the concept of the correlated motion of Cooper
pairs with the same quantum mechanical phase angle in a superconductor upon which
BCS theory rests. In the Josephson junction, two superconductors are brought so close to
each other that the phases of the pairs in the two superconductors are coupled by tunneling
of pairs across the junction. Josephson predicted that for certain phase differences between
the two superconductors, the pairs would be preferentially transferred in one direction
across the junction. This is because motion in the one direction brings pairs from the first
superconductor to the second with the correct phase to join the correlated motion. Motion
in the opposite direction would not bring this about and so a pair moving in this direction
would therefore not be able to join the correlated motion in the first superconductor. Thus,
the current across the junction is dependent upon the phase difference of the pairs in the
two superconductors.
H'C
H' Ho
Figure 2. Current field relationship for Jo- o 2 3 4 5

sephson junction. H/Ho
The phase angle of the Cooper pairs is dependent upon the magnetic vector potential
A. Thus, given that a certain current flows when the B field is zero, the effect of the field
is to vary the phase angle across the junction, as shown in the upper part of Fig. 2. This
modulates the current across the junction and gives the characteristic periodic B-field
dependence of the current, which makes possible the measurement of magnetic fields with
these devices. It is therefore in principle a magnetometer in which the supercurrent is the
measure of the field. However, the intrinsic sensitivity of the device is small because the
field required to produce one flux quantum is relatively large due to the small area of the
junction. Much higher sensitivity can come from using a double junction.
Direct use of Josephson junctions was made in the dc SQUID developed by Clarke
(1974). It consists of a superconducting ring with two Josephson junctions arranged as
illustrated schematically in Fig. 3. The device is biased by a dc current T. In the absence
of a magnetic field, the maximum current through the device is simply twice the current
through each of the junctions. However, when a field is applied to the device perpendicular
to the plane of the ring, the accompanying magnetic vector potential is parallel to the
current in one arm of the ring and opposite in the other. This means that the phase shifts
through the individual junctions are opposite and the two currents interfere where they
meet at P. The phases of the two currents are equal when the magnetic field gives a flux
between the ring which is a multiple of <1>0' It is this interference of the currents in the
two arms of the ring which gave rise to the term superconducting quantum interference
device or SQUID. The expression for the current as a function of field has the same form
as double-slit Fraunhofer diffraction (Fig. 4). The Josephson junctions are in this case
coupled to a superconducting ring whose area is large compared with that of the Josephson
junction. This gives the enhanced sensitivity. Although the dc-driven SQUID may even-
tually prove of higher sensitivity than the rf (radio frequency)-driven SQUIDs, it is the
latter which have been principally used so far in magnetometers for biological and geo-
logical applications.
108 Chapter 4
TRANSPORT CURRENT
f I
Figure 3. Superconducting ring interrupted by two Jo-

sephson junctions.
The rf-driven SQUIDs use rf electronics to determine the changes in the state of a
superconducting ring brought about by the flux to be detected. Before this can be done
effectively, some modification of the ring is required. A macroscopic superconducting ring
does not switch to the normal state in a predictable or repeatable manner. The ring is
therefore weakened, so that the supercurrent must flow through a constricted region. This
then becomes the point at which the ring reverts to the normal state.
As we noted above, the superconducting ring responds to a change in B field by the
generation of a circulating current, in accordance with the Faraday-Lenz law. When the
supercurrent exceeds the critical current for the ring, a flux quantum enters or leaves the
ring according to the sense of the change in the B field. This is commonly refered to as a
flux quantum transit. To make a magnetometer to measure the B field seen by the sensor,
a means of measuring the changes in the supercurrent and the flux transits is required.
Both effects act as voltage sources, so that if a coil is coupled to the ring, emf's are induced
in that coil both by the continuous change in the supercurrent and by the discontinuous
changes associated with the change to and from the normal state when flux transits occur.
Hence, these voltages induced in a coil around the superconducting ring are a measure of
the field changes. The actual way in which these small signals are detected with rf elec-
tronics is discussed in the section on SQUID electronics.
In conclusion, it should be noted that the SQUID gives a measure of the change in
field from the time at which it initially became superconducting. The high sensitivity with
which these field changes can be measured arises from the nature of the periodic depen-
dence of the output on field. This is ultimately related to the flux quantum admitted to
the device when it is driven normal. As this is of the order of 10- 15 Wb, if the state of the
SQUID can be determined with a resolution of say 10- 3 of the period, then one has in
principle a sensitivity of 10- 18 Wb. In the following section we will develop more of the
details of the operation of the magnetometers.
Figure 4. Current field relationship for interferometer.

2.2. The London and Josephson Equations

It is not the purpose of this paper to give a rigorous development of the quantum
mechanical phenomena upon which the operation of the SQUIDs ultimately depends.
Nevertheless, some introduction to the London and Josephson equations seems appropri-
ate. The section can be skipped by those more interested in the practical operation of the
magnetometer. It is written in c.g.s. units.
The basic idea of the quantum macrostate is that all of the superconducting electrons
in a metal can be represented by a single wavefunction
1\1 = pl/2 exp(iB) (1)
where the amplitude factor p represents the superelectron density and B the phase angle.
In this form, the product 1\11\1* gives the superelectron particle density.
By analogy with the behavior of other particle wavefunctions, the superconducting
macrostate is assumed to satisfy the wave equation
-
1 (Ii _
e _) 2
7'il - - A 1\1 + VI\I =
al\l
ili- (2)
2m 1 e at
where e and m are the charge and mass, respectively, of the superelectron (or electron
pair).
The supercurrent within the superconductor has the form
Is = eli. (I\I*VI\I - I\IVI\I*) - e2 A 1\11\1* (3)

2ml me
which reduces to
Is = elip (VB _ eA) (4)

m lie
if the form of the wavefunction given in (1) is used.

The London equation, which describes the electromagnetic behavior of the supercon-
ducting macrostate, can be obtained from (4) by asserting that the superconductor be a
system of ì å á ú ç ê ã = phase. Then
VB = 0 (5)
Is = _ EÉú XF=
The uniformity of B in an isolated, simply connected superconductor suggests that the
superelectron momentum is related to 8. In fact, from (4), it follows by using
j)s = mVs + e (Ale) (6)
where j). is the superelectron momentum and Vs its velocity

Is = peVs (7)
that
ú _ = = -1
Ii J- 1>.
s
ds (8)
Many of the properties of superconductors are a reflection of the reality of the quantum
macrostate. For instance, the quantization of the flux follows from (8) by requiring that
the wavefunction phase change be equal to 21TTI over any closed path within a supercon-
ductor:
110 Chapter 4
il& = f- 21Tm c
V&·ds = 21Tn = --2-
<l>oe P
f-I"ds + - f -21T
<1>0
A· ds (9)
where <1>0 = hc!e, the flux quantum
c JAI,ds + JA·ds = n<l>o (10)
where
A = m/pe 2
For thick-walled cylinders, Is = 0 and this reduces to <I> = n<l>o which gives the flux quan-
tization.
The Josephson junction can also be analyzed directly using the macroscopic quantum
state formalism (Silver and Zimmerman, 1967; Feynman et aI., 1965; Mercereau, 1967).
In this case, the two superconductors are separated, but are close enough to leak super-
conducting particles from the one to the other, i.e., the macroscopic wavefunctions for the
superconducting particles overlap. In such a situation, the wavefunction can change in
time not only in proportion to the system energy but also by leaking particles from one
system to the other. The common method of indicating weak coupling is
(11)
where E is a small coupling coefficient between the two systems. Now substituting the
expressions:
1jJ1 = P11/2 exp(i&1) (12)

t/1z = PZ1/Z exp(Hlz) (13)
into (11) and collecting the real and imaginary terms, we obtain for systems of equal den-
sities
P1 (2/h) E P sin (&2 - Btl (14)
and
(15)
Equation (14) can be written in terms of a current density between the systems by using
Is = I./a = epT/2a (16)
where a is the weak link contact area and T is the superconductor volume. From (14),
Is = [ ePTE]
ah [ sin (&2 - Btl - 21T
<1>0 - ]
JA·ds (17)
where the magnetic vector potential term has been added to achieve gauge invariance.
Equation (15) can also be rewritten as
2eV
=- (18)
Ii
The magnetic potential is again needed to satisfy gauge invariance. The last expression
(where V represents the voltage between the two superconductors) follows from the fact
that the superelectrons are electron pairs at the Fermi surface.
Equations (17) and (18) are commonly called the Josephson relations. They indicate
that when two superconductors are brought into proximity, an initial transient current
flows to equalize the two wavefunction phases; the wavefunctions coherently connect
through the weak link. A static condition is reached when E1 and Ez have been equalized
[Eq. (18)] and the current will then be zero. If a current source is attached to the system,
the phase change across the link il8 = ú Ü = - 1>1 adjusts itself so that (17) is satisfied.
Whenever the current is changed, a voltage, given by (18), appears across the link. As soon
as the new phase separation has been created, this voltage vanishes and the current con-
tinues to flow in the zero-voltage condition. The largest possible current density which
can be supported by the junction, while still maintaining coherence, is given by (17) when
ill> = -rr/2, A = 0; currents exceeding this magnitude destroy the coherence.
2.3. SQUID Geometry
The rf SQUID sensor consists of a superconducting ring or loop of inductance L in

series with a weak link. The early designs of the weak link took many different forms
ranging from a piece of wire with a blob of solder on it to a cylinder with saw cuts. The
factors which govern the topology of the loops are their inductance, shielding, mechanical
stability, and flux coupling efficiency. We will describe two devices, the Dayhem bridge
and the point contact, which have been the most commonly used SQUIDs in the mag-
netometers to be discussed. Neither of these two devices are true Josephson junctions.
However, they do have the essential characteristic of giving an output which is a periodic
function of applied B fields.
Figure 5 illustrates the Dayhem bridge. A superconducting thin film is laid down on
a quartz substrate and cuts are made as shown. The cuts give a constriction through which
all of the current must flow. This then is the point at which the device goes normal. The
rf drive field is applied to the device via an rf coil. The output of the device is detected
inductively with a pickup coil. This type of SQUID was used in many of the earliest
magnetometers built for paleomagnetism laboratories. It was favored at that time over the
point contact SQUID because of difficulties encountered with mechanical instability of
the point contacts.
In its simplest form, the point contact has the geometry illustrated in Fig. 6a. The
point contact serves as the weak link in the superconducting ring. The contact consists of
a niobium point driven into a niobium surface. The simple point contact gave rise to the
"two-hole" point contact, from which in turn came the toroidal point contact. The various
geometries are illustrated in Fig. 6b and c. The configuration of the toroidal SQUID can
be visualized by regarding it as a two-hole SQUID in which the holes have been brought
back on themselves to form the toroidal hole. As we shall see, the toroidal contact is now
the most popular of the rf-driven SQUIDs.
'Fhe signal power developed at the sensor is inversely proportional to the SQUID
inductance, so that sensors with small inductances are preferred. For this reason, a sensor
made with a wire loop is not as desirable as one formed of a long cylinder. The lowest
inductance is usually achieved with a toroidal design. In this case, the toroidal cavity is
112 Chapter 4
DIELECTRIC CYLI NDER
SUPERCONDUCTING FILM
BRIDGE
SENSE EMF (300 A THICK BY I ú ã = WIDE)
AND
RF DRIVE
Figure 5. Dayhem bridge SQUID.
large enough that the inductance of the SQUID is limited by the inductance of the point
contact or the junction itself.
Shielding considerations relate to whether the field changes to be measured are cou-
pled directly to the SQUID sensor. If the sensor is used to measure the field changes directly,
this coupling must be a design feature. However, in most instances this is not the case;
the SQUID is coupled to the external field via a flux transformer coil. The SQUID should
therefore not couple to ambient external field changes and so the shielding afforded by
the toroidal design is another favorable feature.
The mechanical design of the SQUID is important because the sensor shrinks, typically
by 0.1-0.2%, as it is cooled to operating temperature. Symmetrical designs are generally
superior because of this. The principal problem is to eliminate thermal effects on the weak
link. This is particularly important in SQUIDs in which point contacts are used. The su-
perelectron tunneling is across a barrier of less than 100 A. Thus, the junction must be
mechanically stable so that thermal stresses do not give rise to changes of operation char-
acteristics.
Flux coupling to the SQUID is generally achieved via a flux transfer coil. The details
of the technique are discussed below. At this point we are only concerned with the nature
of the field coil in the sensor. This coil carries the supercurrent induced by the change in
flux in the pickup coil. The principal requirement is that the coil be coupled to the SQUID
sensor as efficiently as possible.
Because of the superior characteristics of the toroidal geometry in virtually all of the
above aspects, the toroidal SQUIDs are generally preferred. In such designs, the weak link
is generally placed at the center of the toroid. The weak link can be formed in a number
of ways, but the two most common are the bridge or the point contact. When point contacts
are used, they may either be clean or oxidized. Clean points tend to be mechanically
unstable. Oxidized points contain a thick niobium oxide layer over the point, which makes
them more mechanically stable on cooling. Other support barriers such as semiconductors
have been used to improve mechanical stability. Probably the most reliable weak links are
made with thin films. The approach utilizes superconducting thin films separated by a
small oxide, or semiconductor, to form the barrier Josephson junction. The films are often
deposited upon a silicon chip with contact pads that are connected to the center section
of a bulk toroid loop structure. The problem of making contact to such a chip within a
toroidal structure is formidable. Research to develop improved low-inductance means to
accomplish this are still under way. Thin films can also be deposited directly on insulated
o
I
\
I@--'"
I
\
I -- \
I
I
\
I
I
ú = - /
f J úL=
o Spring Washer
a
rf coil
j POint contact
screw
ú ï É~â ä á å â =
rf ` l á ä ú ä å é ì í =coil
b G
Figure 6. Point contact SQUIDs. (a) Toroidal point contact; (b) two-hole point contact; (c) simplest
point contact.
bulk niobium structures, thus eliminating the contact problems involved when chips are
used. Techniques which make use of this approach have also been developed and are
utilized in commercial SQUID sensors.
2.4. SQUID Electronics
A typical electronics system used with a SQUID sensor is illustrated in Fig. 7 (see also
Forgacs and Warnick, 1967). The sensor is excited by a sinusoidally varying rf magnetic
field of such magnitude that the critical current of the sensor is exceeded in each cycle.
The frequency of the exciting field is usually in the range 20-30 MHz. It is applied via a
small coil embedded in the cavity of the toroid (Fig. 6a). To increase the input impedance
of the device, the rf coil is frequently resonated with a small capacitor across it. The
114 Chapter 4
Low-frequency
,------- - - - - - - - - - - - 1 amplifier and
I I filter
I
I Drive and
sense coil
0 :
ê J J J J J ú f J J J J J J J J | Q=
Feedback circuit
I I
I I
I - Magnetometer - I Output
I sensor I signal
I I
- - - - CryogeniC J b ú î á ê ç å ã É å D J J - -
Figure 7. SQUID electronics.
impedance matching is necessary because the small voltage produced across the rf coil
must be amplified by a room-temperature FET amplifier whose optimum performance is
achieved with an input impedance in the range 1000-5000 n. The real part of the SQUID
impedance looking into the rf coil is of the order of 1 n, when the SQUID is operating.
The required impedance matching is achieved with a Q of about 30 in the resonant circuit.
When the SQUID is excited at a proper flux level, its critical current is exceeded twice
during each rf cycle. This allows flux to enter into the SQUID loop which in turn produces
a back emf in the rf coil. The exact way in which the flux enters into the loop depends
upon the detailed characteristics of the weak link, specifically upon the exact relationship
between the circulating weak link current and the superconducting wavefunction phase.
A simple qualitative model of the flux entry will, however, account for most of the observed
effects (Silver and Zimmerman, 1967). This model assumes that the SQUID is completely
self-shielding until the weak link critical current is exceeded. When this happens, a flux
quantum enters the SQUID loop. If the exciting flux changes and the weak link current
are reversed and the critical current is again exceeded, the flux quantum pops out of the
loop. Each time a flux quantum enters or exists the SQUID loop, it excites a voltage spike
across the rf coil. The frequency spectrum of the spike produces voltage components at
the resonant frequency of the tank circuit which depending upon their phase are either
positive or negative.
The phase of the voltage spikes induced in the rf coil with reference to the drive phase
is determined by the dc magnetic flux threading the SQUID loop. Hence, if this flux
changes, the phase of the spikes changes. To geophysicists and biomagnetists familiar with
the operation of fluxgate magnetometers, it may be helpful to recognize that the SQUID
operation is closely analogous to the operation of fluxgates. The effect is illustrated in Fig.
8. Note that in the absence of flux threading the loop, the phase of the voltage spikes is
given by the solid line in Fig. 8a. However, the presence of the dc flux makes the point
in the rf cycle at which the critical current is exceeded different from that in the flux-free
situation. This leads to the desired result, Le., the magnitude of the rf voltage across the
tuned circuit driving the SQUID is a periodic function of the dc flux threading the SQUID
loop. Specifically,
I V( t) I ex cos (21T<\>xI<Po) (19)
where <Px represents the dc flux through the SQUID loop and <Po the flux quantum.
@) 4
/1
/ I
/
/
/
...
( ( - - - r - t - - - Sias current induced by
I ambient field change
J J ú ú J J ú ú ú ú J H J J ú J H ú ú =
Sext
®2
b
CIRCULATING
EMF + RF DRIVE CURRENT +
--<D
---@ 4
4
wt wI wt
Figure 8. Field detection by rf-driven weak link SQUID. (a) Current field relationship in the presence
and absence of ambient field; (b) voltage spikes induced by flux transits.
The simplest way of obtaining an indication of the flux threading the SQUID loop is
to monitor the magnitude of the rf voltage across the SQUID drive coil. Such a scheme
has two basic problems. First, it is essentially a dc measurement and hence more liable
to drift than are ac techniques available for detection of comparable voltage levels. Second,
the output voltage is periodic with respect to the sensor flux, which greatly hampers meas-
urements of high accuracy and wide dynamic range because the flux sensitivity varies
periodically and is only rarely near the maximum value at
n = 0, ± 1. .. (20)
The first of the two problems can be overcome by applying a small low-frequency (50
kHz) magnetic flux to the SQUID loop and measuring the amplitude of the resulting tank
116 Chapter 4
voltage modulation envelope Em. For values of flux less than <1>0, Em is roughly proportional
to the slope lV(t)1 vs. <l>x relationship expressed in (19). Hence,
(21)
Therefore, Em is also periodic in <l>x and can be used as a measure of this quantity. Em,
however, is the amplitude of an ac signal and is therefore immune to thermal and other
drift effects that affect a dc signal. It can also be detected with a phase-sensitive detector
greatly increasing the signal-to-noise ratio.
To eliminate the problem of the nonlinearity of the SQUID output voltage, the SQUID
is used in a null detector mode by employing negative feedback to lock the magnetometer
at its maximum flux sensitivity position. The electronics used to achieve this are shown
in Fig. 7. In such a system, flux changes are sensed and essentially cancelled by application
of reversed flux. This flux is generated by applying current to the rf coil.
To increase the dynamic range of the magnetometer, one modifies the feedback loop
circuit. Many magnetometers have a number of sensitivity scales for which the full-scale
value will be 1, 10, and 100 <1>0' On the 1 <1>0 scale, the sensitivity of the instrument depends
upon the resolution with which the feedback current can be determined. This gives sen-
sitivities better than 1 part in 103 of a <1>0' The instrument resets when the <1>0 is reached.
On lower-sensitivity scales, the instrument will reset at the 10 or 100 <1>0 flux levels. The
resetting process consists of reducing the feedback current on the SQUID to zero. The
SQUID then sees the total flux from the pickup coils. This is sufficient to exceed the critical
current of the SQUID, which will then go normal and admit the flux quanta. In the earlier
systems, stability of this process was considered of prime importance and so the reset
circuits had relatively slow time constants of the order of milliseconds. Recently, consid-
erable improvements in dynamic range have been achieved, by making the time constants
of the resetting circuit of the order of microseconds. At the limit of the dynamic range,
one counts resets at the maximum possible frequency. This is then the measure of the
changes in flux threading the SQUID loop.
2.5. Flux Transformer Coupling of SQUIDs to B Fields
The physical size of the superconducting ring in the SQUID is constrained by its
inductance, Ls , the critical current Imm and the operating temperature as follows. The
output signal from a SQUID is a function of the amount that the critical current, ie, changes
with the application of magnetic flux to the SQUID ring. This variation, the modulation
depth, l1ie, is a maximum when
(22)
where <1>0 is the flux quantum. In this case, l1ie is of the same order as Imax. Further, the
energy associated with a quantum of magnetic flux, <1>0, being switched into and out of the
ring must be larger than the thermal energy, i.e.,
(23)
where k is Boltzmann's constant (1.38 x 10- 23 JI"K) and T is the absolute temperature.
These constraints set the critical current in the range of several microamperes and the ring
(SQUID loop) inductance in the 10- 9 H range or smaller.
Loops with inductance of 10- 9 H are of diameter 1-2 mm. One of the most significant
design problems with SQUID instruments is how to optimally couple a magnetic signal
(a) (b)
/'-'\ e ú =
\
\
\
\
KJ"OCtiOOi' )
SQUID loop ',_ / SQUID loop
ú p ~ã é ä É= ú ë ~ã é ä É=
(c)
Pick up coil - .--

Figure 9. Coupling of SQUID
sensor to fields to be meas-
ured. (a) Direct coupling; (b)
sample placed near SQUID;
/ \ /
(c) coupling via flux trans-
/ ',- -"
/
former.
to this very small sensing element. Three general solutions are used: (1) insert the sample
directly into the SQUID. This provides optimum flux coupling but restricts the measure-
ment to small samples at helium temperatures (see Fig. gal. (2) Place the sample as near
the SQUID as possible and measure the flux coupled to the SQUID (see Fig. gb). This
technique provides poorer coupling and the signal is a strong function of sample position.
(3) Couple the flux from a sample to the SQUID with a superconducting flux transformer
(Fig. gc).
The flux transformer is most often used with SQUID systems because of the flexibility
it provides. Pickup coils can be built with a configuration to match the particular meas-
urement requirement. Figure 10 shows a typical magnetometer pickup coil circuit with its
transformer coupling to the SQUID. The entire pickup coil-transformer circuit is usually
superconducting so its response is to net flux change <l> and not to time rate of change
(4)) as with a resistive coil. Also, the superconducting circuit is a noiseless field amplifier.
The field and flux transfer of this circuit can be simply derived remembering that the total
magnetic flux linkages of the circuit must remain constant, i.e.,
(24)
with Nl turns in the pickup coil 1, ú Y ä [ N = is the magnetic flux change applied to coil 1, Ll
and Lz are the pickup and field transfer inductances, and Ii is the current induced in the
superconducting circuit by ú Y ä [ N D = The flux applied to the SQUID ú Y ä [ ë = by the change ú Y ä [ N =
at the pickup coil is then
(25)
where Ms is the mutual inductance between the SQUID and coil 2, the field transfer coil, or
(26)
118 Chapter 4
Sensor Output
Wi ú W= t:(-
----J-
,lillJ
I
, I
SuperconductinO
, ú =Sensor
,, a:::::::3 V.
I I
,, 0=::::3
, I
I I
I , I
Id Tronsfer Coil
, I
, I :I _____ _
z y
Helmholtz
Geometry
Pickup Coils
Figure 10. Magnetometer pickup coils and transformer coupling to SQUIDs.
The value of Lz is its in situ value which differs from the free space value because of its
coupling to the SQUID. If the coupling coefficient between Lz and Ls is k where
(27)
then Lz in Eq. (24) is replaced by LZ(Fs/(l - kZ)l/Z; as k is usually 0.6 or less, the change
is about 25%, i.e., the effect of the coupling of Lz to the SQUID circuit is to increase the
effective inductance of Lz.
The magnetic flux applied to the SQUID can be maximized by varying Lz, assuming
the coupling coefficient k remains constant as Lz is varied, i.e., we differentiate Eq. (26)
with respect to Lz and set this equal to zero. The result is that
(28)
and the flux applied to the SQUID is
(29)
The magnetic induction field, I1B s, applied to the SQUID, in terms of the applied induction
field I1B 1 , is (since <I> = BA)
I1B = ú =Al (Ls)I/Z N I1B (30)

• 2 Az Ll 1 1
In typical magnetometer circuits, Ls == 1O- 9 h, Ll == 10- 7 h (one-turn coil with a diameter

of about 2, cm, h = henry), Nl = 1, and k = 0.6, then 11<1>. = 0.0311<1>1' That is, if we
apply one quantum of magnetic flux at coil 1, we couple 0.3 flux quantum to the SQUID.
For this same circuit, the induction field
Thus, the flux transfer with a pickup coil circuit is poor, but the induction field can be
amplified. The analysis presented above is applicable to gradiometer circuits where the
pickup coil circuit is now the sum of the inductances of the gradient coils.
In principle, arbitrarily high induction field amplification can be obtained with this
superconducting circuit, except that as the pickup coil inductance becomes larger it is
more difficult to match its inductance to Lz while maintaining a large coupling coefficient.
It is common to insert a superconducting transformer between the pickup coil and the
SQUID coil both for inductance matching as well as radio frequency interference (rfi)
shielding. The resulting field amplification equation becomes
_..& (Ls)I/Z
I1Bs - 2.
A L1
[1 + (1k k- kt)l/Z]-l dB l (31)
• t
where kT is the coupling coefficient of the intermediate transformer. Note that the in-
ductance and turns of the transformer coil do not appear in (31), but only the coupling
coefficient kT • The transformer can provide a good match between a large pickup coil (with
large inductance) and the very small SQUID coil if care is taken to keep kT large. It is
difficult to achieve a coupling larger than about 0.8 and in this case the factor kT/[l + (1
- kf )I/Z] is 0.5, i.e., a 2 to 1 loss of induction field amplification. The combination of
inductance matching and rfi shielding often makes this loss acceptable.
2.6. Superconducting Magnetic Shields
A magnetic shield is an essential component of the magnetometers discussed in this

paper if their full sensitivity is to be realized. Ferromagnetic shields can attenuate external
fields by factors of up to 104 , but superconducting shields can easily provide attenuation
factors of 108 and larger. The shielding is independent of frequency from dc to gigahertz.
A completely closed superconducting shell will trap whatever field is present when the
shell goes superconducting. External fields will be shielded unless the initial field of the
superconductor is exceeded. If the shell has an opening, the field can leak inside; however,
by proper design these fringing fields can be made very small.
The principle of operation of the shield is now illustrated by considering a super-
conducting ring which is cooled below its critical temperature in zero field. If a uniform
axial field is subsequently applied, a current will flow. As the resistance is zero, this current
will be whatever value is required to cancel the applied field. The field produced by the
current will be nonuniform, but when its value is integrated over the area of the ring, the
flux will exactly cancel the total applied flux.
120 Chapter 4
Figure 11. Superconducting magnetic shield. The applied field is in the negative z direction. The
induced field is shown on the right and the net field-the difference between applied and induced-
on the left.
Figure 11 shows the field around a superconducting cylinder treated in the same man-
ner as the ring. The field induced by the current is much more uniform, but the same
qualitative effects are present. The difference between applied and induced fields is shown
on the left side of the figure, and we see the remaining axial field which has leaked or
fringed into the shield. On the axis, the residual field is in the same direction as the applied
field, but close to the wall of the cylinder, it is opposed to the applied field. Near the end
of the cylinder, the induced field is much less radially uniform than it is near the center.
The field change inside the shield and on axis that will occur when an external field
is applied can be computed from
and
where BAZ = the axial field change, at a distance Z in from the open end of the shield;
BTZ = the transverse field change on axis at a distance Z from the open end of the shield;
ro = shield inside radius; BA = applied magnetic induction field in the axial direction;
BT = applied magnetic induction field in the transverse directions. For example, at a
distance of 4ro inside a cylindrical shield, the axial field will be attenuated by a factor of
(31)4 = 9.2 X 105 , and the transverse field attenuation will be (36)2 = 1.3 X 103 • Note
that these large attenuation factors apply to field changes. A superconducting shield is
often used to trap a given magnetic field environment by cooling the shield to its super-
conducting state in the desired field environment. The Meissner effect, the expulsion of
magnetic flux from a superconductor, is not very effective in the commercial purity lead,
niobium, etc. normally used for shield construction (99.9-99.99% pure) and is generally
in the few percent to 20% range. The field penetration, combined with the very small
superconductor volume as compared with the internal volume of the shield, makes it
straightforward to trap a field environment with less than a 1% change. Once the shield
is superconducting, the means of controlling the field, such as the ferromagnetic shield,
the Helmholtz coils, or the solenoid, may be removed and the trapped field will remain
stable as long as the shield is superconducting. This feature makes it possible to conduct
very sensitive measurements in a normal laboratory environment.
3. Cryogenics
The low temperatures at which presently available superconductors operate require
that the SQUIDs, and associated flux transformers, be maintained at temperatures in the
vicinity of lOOK. Eventually, such temperatures may be attainable without the use of a
cryogen, but for the moment the operation of SQUID magnetometers necessitates the use
of liquid helium.
The very low heat of vaporization of liquid helium (a heat load of only 28 mW will
evaporate 1 liter of liquid helium a day) requires that liquid helium dewars have high
thermal efficiency. The instrument system inserted into the liquid helium container must
have low internal heat production with a minimal amount of heat conducted down elec-
tricalleads to the instrument system from the outside. Figure 12 is a schematic of a vertical
dewar of the type typically used with SQUID systems. The efficiency of this dewar is
determined by the heat leaking down the access into the helium reservoir, or the neck tube
losses, and the heat conducted and radiated through the vacuum space. The neck tube
losses are comprised of radiation, solid conduction down the tube itself, conduction and
convection in the helium gas column, and conduction down shield tubes and electrical
leads. The heat load through the vacuum space is comprised of radiation between the
various thermal layers and conduction which can exist if reflecting layers placed within
the vacuum space have contact with each other. There can also be a heat leak due to residual
gas, especially helium gas in the vacuum space. Cryopumping by the liquid helium res-
ervoir will maintain an extremely high vacuum, so long as no helium gas enters the vacuum
space. More will be said later about helium diffusion.
While liquid helium has a low heat of vaporization, gaseous helium has a high specific
heat, and because the gaseous helium leaves the liquid surface at about 4.2°K, there is a
large potential cooling source available in warming this cold gas to room temperature. In
fact, this sensible heat is about 75 times larger than the heat of vaporization of the liquid
(Long and Loveday, 1968, p. 284). This large heat capacity can be used to intercept various
sources of heat leaking into the reservoir as will be described in the following sections.
There are many different ways to optimize dewar performance and most of these have
been implemented with commercially available SQUID systems. In the following section,
122 Chapter 4
Probe 5 ID
ú ` â =J q ú ú É = . .
I I ú ç é = Pia te
I I Vac. Seal
1 High ú=
Vac. ú f a=c. Jacket
1
20
37 30
V apor Cooled
Shields
lr--1 50 K
-'--
V ú > 50K
=
.--- -
r-- - AI uminized
72cm
I--- f-- Mylar a Spacer
L.He v--'40 layers
· v-- I o6 I ayers
1
20
S IY V- layers
a
u úI ó=
I
0
ú =
Figure 12. Schematic of

vertical dewar showing
superinsulation and vac-
uum jacket.
we will review the design requirements and techniques used in constructing dewars used
with SQUIDs.
3.1. Conduction Heat Leak
The largest source of heat leak, gas conduction between the inner and outer walls of
the dewar, is greatly reduced by evacuating this volume. The pressure in this space should
be at least 10 f.Lm or lower at room temperature. As the dewar is cooled, very effective
cryopumping can be achieved, if an adsorber material such as charcoal is thermally bonded
to the helium-temperature surfaces. This adsorber can reduce the pressure by several orders
of magnitude and maintain a low pressure, even in the presence of very small vacuum
leaks. The conduction heat load Qis given by k(AIl).:l T, where k is the thermal conduc-
tivity, A the cross-section area, 1 the length of the thermal path, and .:l T the temperature
difference. The thermal conductivity, k, is a function of temperature, and thermally sinking
electrical leads etc. to the helium boil-off gas can substantially change the temperature
profile and reduce the net heat leak to the liquid helium by factors of 2 to 100 (Scott,
1959). In general, alloys have much lower thermal conductivity than pure metals, and
wires and tubes made of silicon-copper, beryllium-copper, copper-nickel, and stainless
steel are used in efficient dewars. The thermal conductivity of many different metals,
alloys, and nonmetals is given by Gibbons (1971, p. 678) and Powell and Blanpied (1954).
Dewars for SQUID systems usually require nonmagnetic and even nonmetallic construc-
tion, and various epoxy composites with fiberglass, quartz cloth, and Kevlar cloth are often
used.
3.2. Radiation Heat Leak
Radiation often contributes about one-half of the heat leak reaching the liquid helium.
It occurs through the vacuum space, down all neck tubes, fill-vent lines, and any other
penetrations to the helium. The heat flow, Qr, per unit area due to radiation is given (Scott
etal., 1968, p. 186) by
. a(Tj - 11)
Q - (32)
2(n + 1)
E
r - av
where Eav is the emittance of the layers, a is the Stefan-Boltzmann constant (5.67 x 10- 12
W/cm2-0K), T1 is the high temperature (-290 0K), T2 is the low temperature (-4.2°K), and
n is the number of reflecting layers. In general, E may vary from layer to layer, but for
simplicity it is assumed equal for all layers. The equations are exact if all layers are free-
floating, Le., have no physical contact with adjacent layers.
For an unsilvered dewar with no radiation shields Qr is -0.02 W/cm2 of dewar surface
area. Adding one layer of aluminized mylar with Eav. -0.02 will reduce Qr by a factor of
about 200: 1 to 10- 4 W/cm2. Adding additional reflecting layers reduces the heat leak by
1/[2(n + 1)]. In practice, the space available for reflecting layers is small, to keep dewar size
and weight down, so the layers usually are in thermal contact, separated by a thin sheet
of fiberglass cloth, etc. This contact adds a conduction heat leak that increases with layer
density, whereas the radiation heat leak decreases with number of layers. The optimum
layer density is a function of the spacer material, technique of applying the layers, tem-
perature at each layer, and pressure between layers. Reviews of multilayer insulation are
given in Scott et al. (1968, p. 170) and Molnar (1971). Typically, about 50 layers per
centimeter of radial space are used near room-temperature surfaces and about 10 layers
per centimeter near helium-temperature surfaces.
3.3. Vapor-Cooled Shields
Further improvements in performance can- be achieved by adding thermally con-

ducting shields within the multiple reflecting layers and providing good heat exchange
between these shields and the helium boil-off gas. The combination of high vacuum, mul-
tiple reflecting layers, and vapor-cooled shields is termed superinsulation. Figure 12 gives
approximate dimensions for a "typical" SQUID dewar showing vapor-cooled shield at-
tachment, etc. The reflecting layers are usually 0.006-mm-thick aluminized mylar sheets
separated by a spacer such as 0.05-mm fiberglass cloth. The vapor-cooled shields are copper
or aluminum sheets 0.1 mm thick that completely surround the helium reservoir within
the reflecting layers. Heat exchange is provided by soldering or epoxy bonding these shields
directly to the dewar neck tube. A dewar of this general design with two vapor-cooled
124 Chapter 4
shields can exhibit overall thermal conductance of 1-5J.lW/cm2 of dewar room-temperature

surface area. For a 5-liter dewar as shown in Fig. 12, the total heat load reaching the liquid
helium can be about 20 mW giving a loss rate of 0.8 liter/day. About 50% of this loss is
due to conduction, convection, and radiation down the dewar neck tube.
3.4. Dewar Construction
Dewars must be rugged and preferably lightweight, and must meet stringent magnetic
and metallic requirements. In all SQUID applications, the dewar should be constructed of
nonmagnetic materials, with the specifications on the nonmagnetic components becoming
more critical nearer the pickup coils.
The most common nonmetallic materials are epoxy laminates using fiberglass, quartz,
or Kevlar cloth. Fiberglass is paramagnetic and should be avoided as coil forms and helium
reservoirs. Reservoirs and vacuum jackets have been built by filament winding and hand
layup to minimize weight but the most economical and most common construction is to
use flat plates and tubes, bonded together with epoxy. Metallic components are typically
made from aluminum alloys (6061), stainless steel (321), and copper alloys (alloyed with
nickel, beryllium, or silicon). Of these, stainless steel has the lowest thermal conductivity,
but is the most magnetic. It is especially susceptible to becoming strongly magnetized if
welded or silver soldered, and should be kept as far from the SQUID region as possible.
Stainless steel is often used for dewar neck tubes as it has low heat leak and eliminates
the helium diffusion problem. Silicon-copper alloys can be used to construct rfi shields
and helium reservoirs where the electrical resistivity of the structure can be varied with
the alloy composition.
Dewars for use with SQUIDs must have special features to minimize degradation of
SQUID performance. The following summary of requirements indicates the complexity of
these dewars.
The first priority is that magnetic noise be minimized. Eddy currents, thermal emf's,
and Johnson noise all can produce magnetic signals at the SQUID pickup coil. The dewar
vacuum jacket, superinsulation, and reservoir assembly must all be designed to eliminate
these sources.
Rfi shielding is a critical requirement for reliable operation of SQUID systems. It can
be accomplished with an rfi transformer added between the pickup coil and SQUID and/
or a metallic rf shield completely enclosing the pickup coil assembly. The rfi transformer
is the most effective shielding technique but it reduces the field coupling (ultimate sen-
sitivity) by factors of 2-3. Metallic shields can be very effective; for example, the helium
reservoir and dewar neck could be constructed of metal to provide excellent shielding,
but this introduces large sources of eddy current and Johnson noise. With careful design,
this technique is satisfactory with magnetometers for stationary applications, but is not
suitable for gradiometers. In an rfi shielded dewar, all electrical leads entering the dewar
must be well filtered or they will conduct rf inside the shield, eliminating its effec-
tiveness.
Temperature stability is critical because of the temperature sensitivity of SQUID per-
formance and the temperature dependence of the magnetization of parts of the system. The
temperature of a SQUID system is determined primarily by the pressure maintained on
the helium bath (White, 1979, p. 307). Field dewars need some type of pressure control
valves to help stabilize the temperature, as well as prevent back drifting (neck plugs) and
pump down loss as will be described later. The most common absolute pressure valve
(Tavco, Inc. and Scott et 01., 1968. p. 203) regulates the bath pressure to within about ±
0.03 atmospheres (± 40mK near 4.2K). A change in temperature can produce magnetic
signals by changes in the SQUID performance itself, and from paramagnetism of low tem-
perature components such as the helium reservoir, pickup coil forms, etc. Clarke et 01.
(1975) have measured the intrinsic change in magnetic flux as a function of temperature
for their dc SQUIDs to be from about 0.3<1>o/oK to about 5 x 1O- 3 <1>o/oK. This change is
reported to be due to the combination of critical current variations, paramagnetism, and
flux motion. For "typical" SQUID magnetometers, the flux quantum, <1>0, is roughly equal
to a magnetic field change of 10- 6 gauss. Therefore, the ±40 mOK temperature changes
associated with standard absolute pressure valves can produce signal changes from the
SQUID as large as ± 10- 8 gauss. Point contact SQUIDs have similar temperature sensi-
tivity, depending on the specific nature of the point stabilization and adjustment, and
SQUID package.
Temperature fluctuations can be greatly reduced by as much as a factor of 1000, by
using active control. One controller that has been used in field applications is described
by Gamble et 01. (1978).
All materials near the SQUID and pickup coils can be a large source of temperature-
dependent error signal due to their paramagnetism and diamagnetism. The magnetic sus-
ceptibility, X, is given by XO + CIT, where Xo is the diamagnetic (temperature independent)
component and C is the Curie constant. Therefore,
dxdt = idt (x 0
+ f)T = C
T2
The magnetic moment change, dm, produced by a change in temperature is
dm = - (CMIT2) H dt
where H is the applied magnetic field strength, M the sample mass, dT the temperature
change, and T the absolute temperature. Salinger and Wheatley (1961) published a sum-
mary of the magnetic susceptibility of numerous construction materials. Additional stud-
ies, all of which provide very useful data, have been published by Ginsberg (1970) and
Prober (1975). SQUID system temperature variations can result in large magnetic error
signals. Fortunately, the temperature variation due to the pressure control is at low fre-
quencies, order of 0.1 Hz, and may be out of the signal band. Further, because helium is
a poor thermal conductor, it tends to stratify with about half the temperature change oc-
curring near the top surface. This is especially true if the bath is held very still, and for
increasing bath pressure. Active bath temperature control can greatly reduce these para-
magnetic signals.
Superconducting structures such as SQUID packages and shield tubes are perfectly
diamagnetic and exclude any change in magnetic flux. Liquid helium is also diamagnetic,
its volume susceptibility being about 8 x 10- 7 compared with -1 for a superconductor.
As diamagnetism is not temperature dependent, its effect on SQUID performance will be
to distort field changes and to produce noise caused by motion of a diamagnetic object
near a SQUID pickup coil. The field distortion will occur, for example, if the SQUIDs are
placed too near a pickup coil and can change the mutual orthogonality of the coils. The
B field due to helium sloshing can be 10- 12 or 10- 13 tesla. It can be reduced by placing
baffles in the helium reservoir and separating the pickup coils from the liquid.
3.5. Gas Diffusion
Helium gas diffuses through most nonmetallic materials, such as glass and fiberglass.
Fortunately, this diffusion rate decreases with temperature to zero at liquid helium tem-
126 Chapter 4
perature. Because the dewar neck tube begins at room temperature, the upper section of
this tube will permit some helium gas to diffuse into the high-vacuum region. The diffusion
rate has been measured for various glasses (White, 1979, p. 3B), but is not accurately known
for fiberglass tubes. The diffusion problem can be reduced by bonding a helium getter such
as activated charcoal to the liquid helium reservoir (Molnar, 1971, p. 222). This getter
material will adsorb large quantities of helium gas and maintain the required high vacuum.
A getter volume of 5-10 cm3 can keep up with diffusion for 6 months to a year. Getter
material is often magnetic so it should be placed as far away from the pickup coils as
possible. If the dewar is warmed up, the getter will release the adsorbed helium gas, and
if helium gas is left in the reservoir during warm-up, the diffusion rate will increase rapidly
as more of the fiberglass reaches room temperature. This increase in diffusion can be min-
imized by keeping the dewar at liquid nitrogen temperature, or by flushing the reservoir
during warm-up with nitrogen gas to remove the helium.
Upon warming, the vacuum region should be evacuated to below a few millitorr and
then it can be pumped on at this pressure for a minimum of 6 hr to remove most of the
helium gas. Backfilling the high-vacuum region with nitrogen gas several times during this
pump-down can help flush out the helium gas. Care must be taken to backfill and pump
slowly to prevent collapse of the superinsulation.
3.6. Liquid Helium Storage and Transfer
Storage and transport of liquid helium is best accomplished in commercial superin-

sulated storage dewars such as those manufactured by Linde, Airco, M.V.E., and Cryofab.
These are available in lightweight and highly efficient models constructed from aluminum
andlor stainless steel, with volumes from 30 liters to over 100 liters. These dewars are
often available with built-in absolute pressure valves and safety vents for air transport.
The transfer and handling of liquid helium in the field is somewhat more complex
than in a laboratory environment. For example, trade-offs must be made between transfer
efficiency and system downtime. Also, the storage dewar and SQUID dewar may be at very
different pressures from each other and from ambient. This section addresses the various
losses that occur in preparation for and during a dewar cool-down, helium transfer, and
top-up.
An excellent discussion of cool-down losses is given by Scott et a1. (1968, p. 220).
Standard cryogenic practice is to precool the dewar with liquid nitrogen, as liquid nitrogen
is relatively inexpensive, has high heat capacity, and is easy to transfer. For field use,
nitrogen precool adds the logistics and handling of an extra cryogen plus the difficulty in
completely removing the liquid nitrogen from the precooled dewar before beginning the
liquid helium fill. It is critical that the liquid nitrogen be removed, as liquid (or solid)
nitrogen is a poor thermal conductor (_10- 4 W/cm-OK) and has high heat capacity between
77°K and 4.2°K. It can require tens of liters of liquid helium to cool 1 liter of nitrogen to
4.2°K. It is practical to fill a dewar with liquid helium without liquid nitrogen precool if
the transfer utilizes the very high sensible heat of the gaseous helium in removing most
of the heat from the helium reservoir and the vapor-cooled shields. A well-insulated helium
transfer line is crucial, as relatively long transfers (2-10 hr) are often required. The authors'
experience indicates that a transfer rate of about 2 liters of liquid helium per hour should
be maintained until the inner vapor-cooled shield reaches about BOOK, and the reservoir
assembly is near 4.2°K. The dewar is then filled at a rate of about 50 liters per hour as
with a normal refill. After the dewar is initially filled, the boil-off rate will be substantially
higher than its equilibrium value, by as much as a factor of 3, for about 12-24 hr.
When transferring helium from a storage dewar to the SQUID dewar, precautions
should be taken to minimize helium loss. The major sources of transfer loss are: (1) cool-
down of transfer lines, (2) heat input through the transfer line, (3) depressurization, (4)
pressurization, and (5) flashoff.
Transfer lines should be relatively light and compact, and flexible to permit easy
alignment of storage and receiver dewar. The stainless steel lines most frequently used
have a tube of 3- to 6-mm inside diameter, vacuum insulated over its full length. In some
lines, a few layers of aluminized mylar are wrapped around the center tube for radiation
shielding. These transfer lines typically contribute a steady-state heat input to the liquid
of from 1 to 3 W/m length. Lines with bayonet fittings or helium control valves have higher
heat loads by as much as a factor of 2. The liquid helium required to cool the transfer line
to 4.2°K is usually less than 1 liter because the sensible heat of the helium gas is optimally
used.
Depressurization and pressurization losses are complex processes because of thermal
stratification in the liquid helium. An excellent discussion of this is given by Scott et 01.
(1968, p. 223). The depressurization loss occurs when the storage dewar or SQUID dewar
has to be vented from a high pressure to a lower pressure, usually ambient, to begin a
transfer. This is often necessary with field systems where the absolute pressure valve is
set to about 0.05 atm above ambient at sea level. If a transfer is to be made at higher altitudes,
it may be necessary to vent both storage and receiver dewars to the new ambient. Scott et
01. (1968, pp. 209, 235) show that the depressurization loss is about 5% of the liquid
contents for a 2-psi pressure drop and about 10% for a 4-psi pressure drop.
Pressurization losses occur where warm gas or heat is added to a storage dewar to
force liquid helium through the transfer line. These losses, and the amount of pressurization
needed are discussed by Scott et 01. (1968, p. 227). The losses are a few percent if quick
pressurization is achieved with very rapid transfer (a few minutes). If longer pressurization
and transfer times are used, then more complete heat exchange occurs between the pres-
surization gas and the liquid, resulting in increased boil-off. The pressurization gas may
be obtained from a helium gas storage cylinder, a resistive heater in the storage dewar, or
a bladder connected to the storage dewar (White, 1979, p. 48). The resistive heater and
bladder are simpler for field use. Pressure is also built up at the storage dewar when the
warm transfer line is inserted and this can be adequate to transfer 10-20 liters of liquid.
Pressurization losses are similar to depressurization losses, often equaling 5-10% of the
helium transferred.
Flashoff or throttling losses are due to the isenthalpic expansion from storage pressure
to receiving dewar pressure, at the exit of the transfer line. Scott et 01. (1968, p. 223)
described this process in detail and show that it is similar in magnitude to the depres-
surization losses, about 6% for a 2-psi pressure drop and 10% for a 4-psi pressure drop.
Of the transfer losses described above, the pressurization loss and transfer line loss
can be minimized by very rapid transfer. Throttling losses are lower for slow transfers
(small transfer pressure). Careful planning can often minimize depressurization losses by
conducting the transfer when both dewars are at an elevation equivalent to their absolute
pressure valve settings. The authors' experience indicates that transfer made with a pres-
sure difference of 2-3 psig can be about 85% efficient [see also Long and Loveday (1968,
p. 276) where similiar efficiencies are reported).
Another potentially very large source of heat input can occur from a heat-pumping
effect of oscillations in tubes between the liquid helium- and the room-temperature regions.
Excellent discussions of these oscillations are given by Bannister (1966, p. 127) and Scott
et 01. (1968, p. 193). The oscillations are the same as those used in the helium dipstick
for measuring liquid level and can increase the heat leak down a tube by more than a factor
of 100. All tubes in a helium system that are open near the reservoir and closed near room
temperature are likely sources of these oscillations. Venting with a small hole inside the
dewar neck will prevent the oscillation. Outside the dewar, various damper volumes are
used to stop the oscillation.
128 Chapter 4
3.7. Safety and Neck Plug Hazards
Liquid helium is inert and does not present any chemical explosion hazards. Care
must be taken when transferring large quantities of helium to ensure that adequate ven-
tilation exists, as helium gas can displace the air and may result in asphyxiation (Scott et
al., p. 211). The major safety hazard with liquid helium dewars is plugging of ports to the
helium reservoir with solid air. This problem is more serious if the dewar is moved between
different altitudes. If the vent ports in a superinsulated dewar become plugged, the pressure
will build up rapidly because the vapor cooling stops (no gas flow) and the insulation.
efficiency is decreased at least 10 to 1. Scott et al. (1968, p. 166) present a discussion on
nonvented containers which are often used where large volumes of liquid are to be shipped
and the ultimate use requires gas and/or gas and liquid. They show that the pressure rise
in a 25-liter nonvented nitrogen-shrouded dewar initially 90% full, is about 0.5 psi/hr for
the first 16 hr until only a single phase exists, then increases to about 2.8 psi/hr. In a small-
neck dewar, the neck plug will often support more pressure than the reservoir walls and
the reservoir will rupture as the pressure rises; otherwise, the walls will hold until the
temperature gets high enough to soften and melt the air plug, allowing venting to occur.
This melting temperature is around 55°K which corresponds to a maximum internal pres-
sure buildup of about 18 atm.
Neck plug hazards are minimized by using a safety line into the reservoir with a high-
pressure or absolute-pressure vent valve at its room-temperature end. This line must be
properly designed so that pressure oscillations do not occur.
3.8. Cryocoolers
The optimum SQUID system would incorporate active cooling that could maintain
the system at its operating temperature (liquid helium range) without the use of liquid
cryogens. Helium liquefiers are commercially available that will easily provide enough
cooling but they often cost more than the SQUID system ($50,000 or more), are large and
very magnetic. The mechanical motion produces vibration and fluctuating magnetic signals
that can cause magnetic noise many orders of magnitude above the SQUID noise level.
SQUID systems, such as rock magnetometers, that have the pickup coils inside a su-
perconducting magnetic shield have been operated in a hybrid fashion where a small
cryocooler is used to cool a thermal shield surrounding the liquid helium reservoir to the
15°K range, thereby greatly reducing the liquid loss rate. A rock magnetometer system with
a cryocooler recently introduced by 2-G Enterprises achieves more than 500 days' contin-
uous operation on a single fill of the 100-liter helium reservoir.
Magnetometers and gradiometers that are used to measure signals from external
sources place much more severe restrictions on the cryocooler. These cryocoolers must be
small and lightweight and provide cooling with minimum magnetic, mechanical, and ther-
mal fluctuation.
Vibration isolation can largely eliminate direct mechanical coupling between a cry-
ocooler and a gradiometer-magnetometer assembly, but better mechanical isolation gen-
erally leads to more relative motion between the two. This places more stringent require-
ments on the magnetic signature of the cryocooler. Another design requirement for
cryocoolers is cool-down time. This requirement will be a compromise between cryocooler
size and capacity, and logistics of the SQUID system. Ideally, a very rapid cool-down (order
of minutes) would be desired. This would permit room-temperature standby of the SQUID
system. Conversely, ultrareliable cryocooler operation could permit continuous system
cooling with standby near the operating temperature. A review of small cryocooler de-
velopment is given by Brandt et a1. (1979).
4. Instrument Configurations
The instruments we will consider are

1. Magnetometers and gradiometers for measuring magnetic fields from external
sources, i.e., sample external to the dewar
2. Rock magnetometer systems for measuring samples that are inserted into a super-
conductive shielded region, usually at fields less than 10- 6 T and near room tem-
perature. Sample access diameters are typically 5-10 cm.
3. Susceptibility systems for measuring the field and temperature dependence of small
samples «10-mm diameter) at fields up to 5 T and over temperatures from 4 to
400°K.
4.1. Magnetometers
SQUID magnetometers have been developed for geophysical exploration (magneto-

tellurics and magnetic sounding), medical experiments (magnetocardiogram and ence-
phalogram) as well as general physics experiments. The pickup coils are typically one-
turn loops of diameter 1 to 5 cm. The ultimate sensitivity of the magnetometer depends
on the coil size, the SQUID noise level, and the coupling coefficient to the SQUID as
discussed previously. We can rewrite Eq. (31) to give the applied signal that can be detected,
aB 1 , in terms of the noise limit of the SQUID, aBs:
(33)
The SQUID noise is best characterized in terms of the minimum magnetic flux that can
be resolved when the flux is applied with a current in coil L2 (see Fig. 10) coupled to the
SQUID with coefficient k2. If the power spectral density of this flux is
(34)
then
For commercially available SQUIDs, .:l<l> varies from 10- 4 <1>0 rmsIHz 1/Z to 10- 5<1>0 rmsl
Hz 1/Z , and is white down to frequencies from 1 to 0.1 Hz. Below this frequency, the noise
of the SQUID typically increases as the reciprocal of the frequency, i.e., llf noise. If, as
in our example in Section 2.5, we assume Ls = 1O- 9 H, Ll = 1O- 7 H, Al = 3 X 10- 4 mZ
(a 0.02-m-diameter coil), K2 = 0.6, Nl = 1, a<l> = 5 x 10- 5 <1>0 rms/Hz 1/z , then
The magnetometer detects the vector component of the change in the induction field
normal to the plane of the pickup coil. Multiaxis coils (e.g., three pickup coils oriented
in orthogonal directions with attendant SQUIDs) will measure the change in the total field.
Multiaxis magnetometers are available from CTF Systems, 2-G Enterprises, S.H.E. Cor-
poration, and Cryogenic Consultants. Typically, the magnetometers are fitted into liquid
helium dewars with volumes of between 10 and 30 liters and can be operated continuously
130 Chapter 4
( a) ( b)
TO
TO ^ X WWZ Z Z Z Z Z ú [ =S QUID
ê J ú ? I ê = SQUID
Figure 13. Vertical gradiometer configurations. (a) First-derivative axial gradiometer; (b) first-derivative
off -axis gradiometer.
for up to 2 weeks between liquid helium refills. The magnetometer noise levels are usually
limited by the noise from the dewar and probe assembly to between 10- 13 and 10- 14 T.
4.2. Gradiometers
Gradiometer circuits are a straightforward extension of the magnetometer using a mul-

ticoil continuous superconducting circuit to detect the difference in field (or field gradient)
between two regions of space. Figure 13a shows a gradiometer for measuring the first
derivative of the field in an axial direction. Off-axis gradients can be measured with coils
configured as shown in Fig. 13b. The complete gradient tensor consisting of nine com-
ponents can be measured with multiple pickup coil-SQUID assemblies. Only five assem-
blies are needed because
and the gradient tensor is symmetric so
aBy = aB z
az ay
aB z = aB x
ax az
Higher-order gradients can also be measured with superconducting circuits. The cir-
cuit shown in Fig. 14 will measure the second derivative of the axial field ~ O_ ò L ~ ú K = This
circuit connects the output (induced current) of two first-derivative gradiometers in series
TO TO
SQUID SQUID
Figure 14. Second vertical derivative axial gradiometer.
opposition. The resulting net current is proportional to the difference in the gradient signal
between the two circuits, i.e., the derivative of the gradient. Higher-order axial and off-
axis gradients can be measured with similarly connected superconducting pickup coil
structures.
These gradiometer circuits measure the gradient of the induction field averaged over
the baseline (pickup coil separation). Often, gradiometers are actually used to measure the
magnetic field from a sample placed close to one of the pickup coils-and the remaining
coil (coils) is (are) used to cancel magnetic noise from sources more distant than the sample.
The very large noise immunity achieved with gradiometer circuits results from the rapid
falloff of the magnetic field from localized magnetic soures. This is referred to as near-
field mode of operation.
The dc magnetic field from a dipole source, M, as shown in Fig. 15 is
HR = 2M cos 8/41T r3
He = M sin 8/41T r3
132 Chapter 4
y
Figure 15. Dipole field resolved in ra-
dial (HR ) and tangential (He) compo-
x nents.
where r is the distance from the dipole source to the observation point and 9 is the polar
angle. The gradient of these field components is the derivative with respect to r and varies
as l/r4. The second-order gradient varies as N L ú I = etc. This very rapid decay with distance
means that remote field sources will produce very little signal at first- or second-derivative
gradiometer. A signal source located very near one of the pickup coils will produce flux
that mostly links this one coil and the circuit essentially measures the magnetic field of
the sample.
The degree to which the signal from distant sources is rejected by a gradiometer circuit
depends on the accuracy with which the coils are constructed and aligned. This noise
rejection is termed the common mode rejection, or balance of the gradiometer pickup coils,
and is measured by determining the signal that a first-order gradiometer would detect if
placed in a time-varying but perfectly uniform magnetic field. For example, if, in the
gradiometer circuit shown in Fig. 13a, the upper pickup coil had an area that was larger
than the lower coil, then a signal would be detected even if the gradient were zero. Another
source of balance is tilt of one coil relative to the other. This couples off-axis field changes
as net signals.
Because the gradient subtraction is done within the closed superconducting circuit,
the process is passive and is only dependent upon the mechanical accuracy and stability
of the pickup coil structure. Typically, the pickup coils are made of niobium-titanium
wire bonded to quartz or silicon substrates. The substrate is machined and ground to
tolerances of the order of 0.01 mm. The coil planes can be aligned to better that 0.010 giving
an intrinsic balance of the order of 1Q- 3 /m where the balance is defined as the error signal
in units of gradient (field/distance) divided by the uniform field change. A simple way to
measure the balance of a gradiometer circuit is to rotate the entire circuit (dewar included)
through a 3600 revolution in the earth's field in a site remote from large magnetic objects.
The measured peak-to-peak signal divided by twice the earth's component along the gra-
diometer axis is the balance.
The balance can be adjusted mechanically and electrically to the order of 10- 8 /m.
Mechanical adjustment consists of placing small pieces of superconducting material (bal-
ance disks) near one of the pickup coils to distort the field and couple more or less flux
to one coil relative to the other. These superconducting balance disks may be bonded to
the gradiometer substrate, requiring repeated warming and cooling of the gradiometer to
achieve a high balance, or they may be attached to rods that extend from the top of the
gradiometer probe to the pickup coils. The disks are then moved with micrometer ad-
justments at the probe top.
Electronic balancing is also very effective in improving the common mode rejection.
In this case, the error signal profile is measured with respect to the ambient magnetic field
by simultaneously sensing the field and the gradient. This balance scale factor can then
be used to provide a compensation signal that is fed back into the gradiometer or its output.
Mechanical adjustment is normally used to achieve a balance of 1O- 5 /m, and electronic
feedback is then used for further improvement (e.g., Vrba et 01., 1982).
Gradiometers are available from CTF Systems, Cryogenic Consultants, 2-G Enterprises,
and S.H.E. Corporation in a wide variety of configurations. Specific biological applications
are for magnetocardiograms and magnetoencephalograms, as well as testing samples for
the presence of magnetic inclusions. Noise levels are determined by the nature of the
ambient magnetic noise and subtleties in dewar design. In a relatively quiet unshielded
environment, noise levels of 2 x 10- 13 Tim can be achieved. These instruments have
achieved some of the most spectacular results in detecting the extremely weak fields ob-
served in magnetoencephalography. This is, however, an ac signal. To detect magnetic
inclusions in biological samples requires essentially a dc observation, as the animal is
moved past the sensor. For such measurements, it is likely that the best approach is to use
a shielded room, although as will be discussed below (Section 5.1) such observations have
been made in the magnetically noisy environment of a hospital.
4.3. Rock Magnetometers and Susceptometers
Two basic configurations have been adopted in the rock magnetometers: a straight-
through access system operated with the access either horizontal or vertical and a vertical
single-ended access system. The susceptometer is similar to the vertical magnetometer,
but includes a superconducting magnet. All of these instruments utilize the same rf-driven
SQUID or dc SQUID and electronics; they differ only in that their configurations are mod-
ified to suit the particular applications for which they are developed. The output is in
analog and digital format for recording and processing with standard computer interfacing.
Three types of cryogenic assemblies have been used at UCSB which are representative
of presently available systems. The simplest system has a 3-mm-inside-diameter access as
shown in Fig. 16a. The probe assembly is immersed directly in liquid helium, which is
contained in a standard superinsulated dewar. An advantage of this system is that the
sensors and other parts of the magnetometer are immediately accessible. A disadvantage
is that because the shield is immersed in liquid helium, it cannot be thermally switched
to trap desired fields without taking the probe out of the dewar or boiling off the helium.
This special vertical system was made to measure very small samples and has a sample
access diameter of 3 mm with a magnetic moment sensitivity of 2 x 10- 13 A m2 • This
brings within the range of measurement individual pseudo-single-domain particles. The
sample access region is evacuated so that the samples can be measured at room temperature.
Sample cooling occurs by radiation to the pickup coil support, which is at helium tem-
perature. This cooling rate extends up into the insulated region of the access tube. The
approximate sample temperature has been monitored by using the Morin transition in
hematite, and cooling rates of 0.5°K/min appear normal. The 3-mm system has an axial
and transverse measurement axis, so that the three orthogonal components of magneti-
zation can be measured by inserting the sample and then rotating it 90°.
The early rock magnetometers developed at Superconducting Technology Inc. used a
modification of the immersion type. In these systems, the shield and sensors are placed
in a partially evacuated region, as shown in Fig. 16b. A pressure of a few micrometers of
helium exchange gas provides a weak, but controllable, thermal link to the reservoir. This
design overcomes the difficulty of switching the shield, for the shield can be heated above
its critical temperature while the probe is still immersed in liquid helium. The probe
134 Chapter 4
/"
/V_.
SUPf"'!IISUl. • .,..,..
II
Figure 16. Instrument configurations. (a) 3-mm-access vertical system with direct immersion; (b) ver-
tical system with shield switching; (c) horizontal system with open access.
assembly is housed in a standard superinsulated dewar. Rock magnetometers of this design

are available from CTF Systems and Cryogenic Consultants.
The third cryogenic design is that used in the horizontal access magnetometer and
illustrated in Fig. 16c. The system also permits shield switching and makes particularly
efficient use of the vapor-cooling principle. It is well suited to the refrigerator approach
to minimize helium loss. The major advantages of this design are: (1) the proximity of the
shield to the sample entry, (2) low neck tube loss, (3) simplicity and versatility of straight-
through sample access, and (4) greatly improved thermal stability of the superconducting
components resulting in reduced noise.
The horizontal open-ended system has a sample access hole of 6.4 cm. The system
was designed in 1972 and has been in almost continuous use since then. Standard samples
are handled with a horizontal plastic holder. The open-ended access (Fig. 16c) permits
certain measurements not possible with the vertical single-ended systems. The most im-
portant of these is probably the measurement of long cores for geological purposes or
elongate samples in biomagnetic applications. It also makes other applications, such as
thermomagnetic analysis, more convenient, because it is easier to place the necessary de-
vices in the sample access region.
A new rock magnetometer suitable for biomagnetic work has been introduced by 2-G
Enterprises that incorporates the same basic design features of the first horizontal instru-
ment, but which has a reduced helium boil-off rate of 0.17 liters/day and a reduced noise
level of 10- 11 A m 2 (Fig. 17). With its 100-liter helium reservoir, it operates continuously
for more than 500 days between liquid refills.
The susceptometer system in use at UCSB is basically of the same design as the vertical
standard sample systems. It has a 7-mm access and a superconducting magnet capable of
giving a 2.5-T field. A superconducting shield is mounted inside the magnet and is used
c
'"CD
:a,
en
.g
8
ú=
Ol
ú=
;a.
o
:3CD
Cii
ú=
5·
ttl
o·
:3
Ol
ú=
;a.
00·
:3
Figure 17. 2-G Enterprises Model 760 R superconducting rock magnetometer.

ú =
ú =
ú =
136 Chapter 4
to trap the high field and provide the stability required to make sensitive field measure-
ments in the background of the strong dc field. It should be noted that if one wishes to
detect a B field of, say, 10- 10 T in the O.l-T field, then the dc field must, of course, be
stable to one part in 10 9 • The system temperature control unit makes possible continuous
magnetization measurements from 4°K to room temperature. It consists of a gas flow heat
exchanger in which helium gas entering at approximately 4.2°K is heated and temperature-
regulated with a resistance heater and diode thermometer and then passes through the
sample region. The sample is immersed in this gas environment at a pressure of 1 atm.
Sample temperatures may be changed quickly, and reproducibility and sample handling
are greatly simplified because the sample region is at 1 atm.
Susceptibility systems are available from S.H.E. Corporation. The newer systems have
applied fields of 5 T, temperature cycling up to 400°K, and the measurement is greatly
simplified with computer-controlled sample movement and field and temperature change.
5. Applications of SQUID Magnetometers in Biomagnetism

All of the magnetometers described in the previous sections of this paper have found
application in biomagnetism. In this section, the various applications are reviewed to help
potential users in the choice of instrument for the particular task they may have. They are
discussed in terms of isolation, identification, and characterization of magnetic materials
found in biological samples.
5.1. Isolation of Magnetic Material in Biological Samples
The isolation of magnetic material in biological samples is an aspect of biomagnetism

which has attracted considerable attention recently. Much of this work has been in con-
junction with behavioral experiments to see if the animal has the ability to detect and
make use of the geomagnetic field for navigational purposes. Of the instruments described
in this paper, two are well suited for such work-the rock magnetometer and the gradi-
ometer.
Given a small animal, which can be placed inside the room-temperature access hole
of the rock magnetometer, or the possibility of dissecting the animal to give samples com-
parable in size to the standard samples of paleomagnetism, the rock magnetometers afford
a convenient method of measurement. The standard instruments are usually vertically
nested dewars, having a vertical access hole, closed at the bottom, with a diameter of
approximately 2 inches. From a practical point of view, horizontal systems offer a little
more flexibility. They have open-ended access holes so that samples can be passed through
the instrument. Elongate samples can therefore be used. With such instruments, equivalent
moment sensitivities of the order of 10- 11 A m2 can be achieved. One can enhance the
detection level by giving the sample a saturation remanent magnetization (see Walker et
a1., Chapter 5, this volume).
The use of gradiometers affords an advantage over the rock magnetometers in that the
sample does not have to be placed in the instrument. The gradiometer can be used to scan
the sample by moving the one with respect to the other. Thus, an entire animal can be
studied without any dissection and the problem of contamination during dissection is
eliminated. As this problem of contamination has proved severe, the scanning method
using gradiometers may prove important. The method also permits noninvasive study of
live animals.
There are two basic strategies for the use of gradiometers to localize magnetic material
in biological samples. One can measure the fields due to the remanent moment, or one
ú J J ì É ï É ê =
Figure 18. Schematic of arrangement

at Case Western Reserve University
for biosusceptibility measurements
on the human liver. After Roman et al.
(1982).
can induce a magnetization in the sample and measure the field due to the induced mo-
ment. Both methods make use of measurements analogous to the classical use of an astatic
magnetometer. In that measurement, the sample is introduced close to one magnet of an
astatic pair, so that the field of the sample significantly affects only this magnet. The astatic
pair is insensitive to homogeneous fields in the laboratory and so responds only to the
field of the sample. Similarly, the SQUID gradiometer used in the near-field mode permits
determination of the field of the sample in the background of fields generated by distant
sources.
The use of the gradiometers to detect induced fields due to magnetic material in bi-
ological tissue has already resulted in a number of medical applications. The simplest
approach is to use a dc field. A homogeneous field is applied with a large Helmholtz coil
and the gradiometer used to detect the fields due to the induced moments of the sources.
Farrell et al. (1980) at Case Western Reserve University have used this method to measure
the susceptibility of the liver, giving a measure of iron in that organ, which can provide
an indication of malfunction in blood metabolism. They used a second-order gradiometer
in their initial approach, which was successful in a magnetically quiet environment. Later,
with a coil configuration which gives a drive field with a prescribed gradient over the
sample volume and a second-order gradiometer for detection, they were able to carry out
the experiment in the magnetically noisy environment of a hospital. This group also found
that it was necessary to place a water bag between the torso and the sense coils so that a
relatively uniform diamagnetic response was achieved (Fig. 18). In this way the weak
diamagnetism of the body is matched by that of the water.
Signal to noise can, in principle, be improved by using an ac drive field with phase-
sensitive detection. A 10-Hz frequency is adequate to permit convenient use of such de-
tectors. This method was investigated by the Case Western group, but they were not able
to improve upon their earlier results obtained with the dc technique. Romani et al. (1982)
point out that the twin difficulties of baseline drift of small field coils and the lack of
resolution with large coils have defeated attempts to improve these methods.
5.2. Identification of Magnetic Material in Biological Samples

The magnetic material localized using the techniques described above may contain a
variety of magnetic phases. These phases may exhibit any of the known forms of magnetic
138 Chapter 4
order. An essential part of studies of such magnetic material in biological samples is there-
fore to identify the magnetic phase. A common feature of the biological work is that the
signals are weak and so measurements have to be made at high sensitivity. The various
diagnostic characteristics of magnetic material are now reviewed and suggestions made
for the use of SQUIDs in identification of the magnetic phases.
Diamagnetic and paramagnetic material are readily identified because they do not
exhibit magnetization in the absence of a field. They do, however, become magnetized in
a field. Magnetic susceptibility (k) is the measure of how easily a material acquires mag-
netization:
M = kH
where M is the magnetization per unit volume and H the magnetizing field. In SI units,
susceptibility is dimensionless; M and H are both measured in amperes per meter.
Diamagnets have negative susceptibility which is invariant with temperature. As we
noted in the discussion of the measurement of the magnetic susceptibility of the liver, the
human body, like water, is diamagnetic, having a weak negative susceptibility. The ubiq-
uitous occurrence of diamagnetism makes it unlikely that its localization and identification
is likely to be of much interest.
Paramagnets have positive susceptibility due to the alignment of the moments of in-
dividual atoms or molecules by an applied magnetizing field, H. This alignment is opposed
by the randomizing effect of thermal energy. The classical theory of paramagnetism de-
scribes the magnetization acquired in the presence of a field in terms of the Langevin
function (L) operating on the ratio of the magneto static energy aligning the moments to
the randomizing thermal energy, e.g.,
M = Nj.LL(a) (35)
where
(36)
or
j.LH
M = Nj.L ( coth kBT - j.LH/kbT
1) (37)
For a ú = 1,
L(a) = i. (38)
where j.L is the moment of the atom or molecule, kB is Boltzmann's constant, and T is the
absolute temperature in degress Kelvin. It is evident that the susceptibility will have an
inverse dependence upon T. The quantum mechanical treatment yields analogous results
with the Brillouin function taking the place of the Langevin function. This inverse tem-
perature dependence of the susceptibility is the diagnostic characteristic of paramagnetism.
The presence of a paramagnetic phase can be readily determined using a high-field sus-
ceptometer. The temperature is varied from liquid helium to room temperature and the
high-field susceptibility measured to establish its temperature dependence.
If diamagnetic and paramagnetic material are present together in a sample, the amount
of paramagnetic material can bp. determined from the temperature-dependent susceptihil-
5 ...
....
E
u ....
......
...... TRM
' .
c
E
o
3 .....................
E
u
Kú =
c:
0'
o
ú =
KH ............:•••••••
.........
............. .....\.
Figure 19. Thermomagnetic analysis for Curie point de- 0' 145' 265' 373' 477'
termination. Temp DC
ity. The temperature-independent signal is due to the negative susceptibility of the dia-
magnet.
Diamagnets and paramagnets do not involve coupling between the magnetic moments
of individual atoms or molecules. In contrast, ferro magnets , antiferromagnets, and ferri-
magnets all have an internal magnetic order, the mutual alignment of the magnetic mo-
ments of individual atoms. In ferromagnets, this takes the form of the parallel alignment
of neighboring spins by the exchange interaction. In antiferromagnets, the alignment is
anti parallel , so that in a perfect antiferromagnet, no magnetization is seen in the absence
of a magnetic field. The origin of the coupling is exchange in metals and superexchange,
across an intermediate atom, in compounds. In ferrimagnets, the fundamental order is
antiparallel, as in antiferromagnets, but the magnetic moments in the two opposed di-
rections are unequal, giving a net moment in the absence of a field. In all cases, the align-
ment is opposed by the randomizing effect of thermal energy. At some temperature the
thermal energy overcomes the coupling that gives the magnetic order, and the material
enters a paramagnetic state. This temperature is called a Curie point in the case of fer-
romagnets and a Neel point in antiferromagnets and ferrimagnets. These temperatures are
a measure of the strength of the exchange and superexchange intensities in the various
materials and provide the primary means of identification of these magnetic phases.
The classical method of determining Curie points or Neel points is to heat the sample
in a thermomagnetic balance (e.g., Collinson et al., 1967). The cryogenic rock magneto-
meter, or a high-field susceptometer could be used to make the measurement. The principal
difficulty lies in the requirement to heat the sample and to observe its susceptibility at high
temperature. A number of methods of heating have been tried, e.g., a laser (Day et a1.,
1976b) and an rf heater (R. K. Walton, personal communication). Recently, we have found
that satisfactory high-temperature measurements can be achieved by heating the sample
outside of the instrument. It is placed in a nonmagnetic quartz holder, so that on insertion
of the sample into the instrument for measurement the temperature of the sample does
not change significantly (J. R. Dunn, unpublished work). Here, advantage is taken of the
fast response time of the SQUID magnetometers and the fact that the sample does not have
to be spun or vibrated to make a measurement. Figure 19 shows the result of such an
analysis. The field used in the experiment was less than 10- 4 T. The use of such low fields
has an advantage in determination of Curie points due to the phenomenon of the Hopkinson
140 Chapter 4
- ___u· 0
100 o o
M
Mo o
75 o
50
\
25
'0 0 0
ú =000
0-0
50 100 150 200 oK 250 300
Figure 20. Low-temperature analysis of IRMs to identify multi domain magnetite in stingray sample.
rise. This rise refers to the increase of susceptibility shown by magnetic materials below
the Curie point or Neel point. At such temperatures, the material has low anisotropy and
so is readily magnetized. Complications in the analysis can, however, arise due to the
difficulty of separating remanent and induced moments in such low fields and it may
prove better to use intermediate fields of a few tens of oersted to overcome this problem.
In antiferromagnets, no such difficulty arises and the use of the Neel point makes iden-
tification straightforward. Another possible approach is to use the gradiometer, which
permits the sample to be heated externally from the instrument. The Neel point of the
ferrimagnet magnetite, which is by far the most important magnetic phase in biological
material, is 586°C (859°K).
In addition to their distinctive Curie points and Neel points, ferromagnets, ferrimag-
nets, and antiferromagnets exhibit distinctive magnetic anisotropy transitions. Magnetic
anisotropy refers to preferred directions of magnetization within a grain. In such a direc-
tion, the energy associated with the magnetized state is minimized. One form of anistropy
is magnetocrystalline anistropy, which is defined with respect to the crystal lattice. Thus,
magnetite is most easily magnetized along the body diagonal (111) directions at room
temperature. However, at low temperature the easy directions become the (100) cube edge
directions, in association with a crystalline transition from cubic to orthorhombic. A simple
identification technique for magnetite was developed making use of the behavior of mag-
netic remanence across this transition (Nagata et 01., 1964; Fuller and Kobayashi, 1964).
The procedure is to magnetize the sample to saturation at liquid nitrogen temperature and
to observe the warming curve of this saturation remanent magnetization. This can be carried
PHASE IDENTIFICATION
ANTIFERROIoIAGNET FERROMAGNET
FERRIMAGNET
Figure 21. Identification scheme for magnetic material using SQUID magnetometers.
out either in the standard rock magnetometer, with suitable thermal lagging of the sample,
or in the high-field susceptometer with zero field. As the anisotropy transition is ap-
proached, the remanent magnetization decreases sharply, to give a minimum at the tran-
sition temperature. Above the transition, a memory of the low-temperature magnetization
is acquired (Kobayashi and Fuller, 1968). An example of identification of magnetite from
a stingray (Urolophus halleri) is illustrated in Fig. 20. This technique is suitable only for
magnetite which is relatively coarse-grained; the transition in remanence depends upon
the rearrangement of the domain pattern at low temperature and is not seen in single-
domain magnetite. More sophisticated techniques are required for the determination of its
presence in single-domain material, but the transition is reflected in changes in other
magnetic properties, such as coercive force which can be measured as a function of tem-
142 Chapter 4
(+) SATURATION
MAGNETIZ ATI ON
MAGNETIC
o b j ^ k b k ` b ú =
(Jr) ú =
(-) (+) MAGNETIC

K K K K WK K K K J J J ú J J J WX K J Ñ J J J Ñ J ú ú J J J J J FIELD
COERCIVE
FORCE
(He)
Figure 22. Hysteresis loop for en-

H semble of fine particles.
perature. An important advantage of using these anisotropy transitions for identification

is that the sample need not be heated.
The various techniques discussed in this section are summarized in a simplified flow
diagram (Fig. 21), which illustrates the use of the various cryogenic magnetometers to
identify the state of magnetic order and in the case of ferro magnets and ferrimagnets to
identify the phase.
5.2. Characterization of Magnetic Phases in Biological Material
Not only is the chemical composition of a magnetic phase found in biological samples
important, but so also are its grain size and whether interactions are present. Magnetic
sensors will be critically dependent upon whether the magnetic material is fine grain or
coarse. The magnetic behavior of materials is strongly dependent upon grain size so that
almost any magnetic property apart from the Curie point and saturation magnetization will
give an indication of grain size. A magnetic material may be characterized by its hysteresis
loop and so we consider first the effect of grain size on the loop. The hysteresis loop is
not conveniently measured with a cryogenic magnetometer because the necessary change
in the magnetizing field drives the SQUID sensor out of lock. The standard method is to
use a vibrating sample magnetometer as described in Collinson (1975). If the extreme sen-
sitivity of the SQUIDs were critical, one could use the susceptometer for a stepwise meas-
urement of hysteresis loops but it would be extremely time-consuming because one would
have to trap each required field in the superconducting shield before the measurement
could be made.
The behavior of fine ferromagnetic or ferrimagnetic particles, when they are exposed
to magnetic field cycling, is shown in Fig. 22. As the magnetic field is increased, the
magnetization increases to a saturation value (Ms). On reduction of the field, the mag-
netization decreases to give remanent magnetization (Jr or IRMsl when the field is removed.
The explanation of this phenomenon rests on two facts. First, the moments of the individual
atoms in the grains are aligned parallel to each other by exchange energy. Second, within
each grain there is a preferred axis of magnetization-the easy direction. In this example,
we assume that the grains are elongate and that the easy direction is the long axis of the
grain due to shape anisotropy. Initially, the magnetization lies in the easy directions, giving
(+1
SATURATION
MAGNETIZATION
Figure 23. Hysteresis loop for mul- H

tidomain material.
zero net moment. As the field is increased, the magneto static energy ( - M· H) requires that
the spins are pulled out of the easy directions to become parallel to the applied field. In
saturation, the applied field is sufficiently strong that all of the spins are parallel to the
field. When the field is reduced, the spins are free to fall back into the nearest easy direction.
This does not give a zero net moment because the easy directions chosen are systematically
biased toward the field direction. Application of a reversed or back field remagnetizes the
assemblage of particles in a negative sense, which eventually gives saturation magneti-
zation in that sense. The resulting trace of the magnetization is called the hysteresis loop.
Magnetic materials exhibit a great variety of hysteresis loops. An important variable
which determines magnetic behavior is the grain size. Thus, whereas an assemblage of
very fine grains of magnetite might exhibit a loop like seen in Fig. 22, large grains have
magnetically softer characteristics, i.e., they are magnetized and demagnetized by much
smaller fields (e.g., Fig. 23). This is because although small grains are magnetized ho-
mogeneously, large grains are subdivided by walls into domains of mutually opposed
magnetization. The introduction of domains reduces energy expended in the external field
of the grain, by reducing the net magnetic moment of the grain. In single-domain grains,
magnetization is by rotation (Fig. 24a). In multidomain grains, changes in magnetization
can be brought about by wall movement (Fig. 24b). This is a much lower energy process
than magnetization rotation, so multi domain grains are magnetized and demagnetized by
weaker fields than are single-domain grains.
In the hysteresis loops illustrated, three important states were noted. There is the state
of saturation, at which all spins are aligned parallel with the applied field. There is the
state of remanent magnetization, at which a memory of the field is retained, although the
field has been reduced to zero. Finally, there is the coercive force field, the back field
which is required to reduce the magnetization to zero. For single-domain grains, the re-
manent magnetization is a large fraction of saturation magnetization and the coercive force
tends to be high, say of the order of 104 /m. In multi domain materials, which can change
their state of magnetization by the movement of walls, coercive force is lower because wall
motion is a low-energy process and remanent magnetization tends to be a small fraction
of saturation magnetization.
The determination of grain size can be readily carried out using any of the properties
that reflect the different behavior in the single- and multi domain states (e.g., Day et a1.,
1976a). One of the simplest such indicators is the ratio of saturation remanent magneti-
zation to saturation magnetization. The magnetic hardness is also an indication of domain
144 Chapter 4
ROTATION
WALL MOTION
Figure 24. Magnetization processes. (a) Magnetization rotation in single-domain particles; (b) mag-
netization by wall movement in multidomain material.
state and is reflected in dc magnetization and demagnetization curves. Figure 25 illustrates

the behavior of a sample of whale dura. The material is evidently magnetically hard; a
field of nearly 500 Oe was required to reach a half-saturation value and 2kOe for saturation
remanence. The remanent coercivity, the back field to reduce the remanent magnetization
to zero, is roughly 300 Oe. The determination of these curves is experimentally simple; it
only requires the measurement of remanent magnetization, which is readily carried out
with the cryogenic rock magnetometers. The necessary magnetizing fields can be generated
Magnetic Moment (Ml
X 165 gauss cm3
-I Koe IKoe Koe
Field (H)
Figure 25. dc Magnetization and demagnetization.

with coils or electromagnets. AF demagnetization provides yet another means of estab-

lishing the magnetic hardness of the material (e.g., Collinson et al., 1967). AF demagne-
tization equipment is universally available in paleomagnetic laboratories.
The combined use of dc magnetization curves and AF demagnetization permits the
detection of interactions between magnetic particles (Cisowski, 1981). The test involves
the comparison of the dc magnetization curve and the AF demagnetization curve as in Fig.
26. If there are no interactions, the curves are symmetrical (as in Fig. 26a), and they intersect
at the remanent coercivity field value. In contrast, if interactions are important, the curves
are no longer symmetrical (Fig. 26b). This curve is for chiton teeth in which the fine
particles of magnetite are known to be so close to each other that interactions will be strong
(Lowenstam, 1962). This test also distinguishes between multidomain and interacting sin-
gle domain. Figure 26c gives the curves for a rock with multidomain magnetite. Note that
in comparison with the single-domain interacting material, the multi domain sample ac-
quires magnetization more readily in the low field and saturates at lower fields.
The problem of magnetic granulometry of lake sediments has attracted a good deal of
attention both because it has applications to limnology and because it is essential to an
understanding of the magnetization of these sediments. The techniques can be applied
immediately to biomagnetic analysis; indeed, it is possible that some of the magnetite in
lake sediments has a biological origin. A promising method of granulometry has recently
been suggested by King et al. (1982). They have chosen two parameters for their scheme.
The acquisition of anhysteretic remanence is one. Anhysteretic remanent magnetization
(ARM) is acquired when a magnetic material is exposed to an AF which is allowed to
decrease in amplitude in the presence of a biasing dc field. ARM is particularly sensitive
to the presence of single- domain particles. The other parameter is weak-field suscepti-
bility, which is sensitive to multi domain grains. Figure 27a shows the results of such an
analysis for certain assemblages and in the accompanying diagram a simple interpretational
model is presented. Such techniques could be particularly helpful in granulometry of bi-
ologically precipitated magnetite.
The use of these granulometry techniques permits the determination of grain size via
the effect of grain size on domain state and hence upon magnetic behavior. All of these
methods assume that the material is in a grain size range which is at least large enough
to be stable single domain. There remains the extremely fine-grain size range, which is
too small to exhibit stable single-domain behavior.
When the grain size of a ferromagnetic or ferrimagnetic phase decreases below a critical
size, in becomes superparamagnetic. Its magnetization is in thermal equilibrium with its
surroundings and follows the ambient field. Writing magnetization as a typical relaxation
phenomenon, we get
which gives after integration
Mt = Mo exp (-tIT)
where Mo is the integration constant from the initial conditions. The relaxation time is
dependent upon the anisotropy energy Ku , the volume V, and the thermal energy kT.
This dependence is exponential (e.g., Neel, 1957):
lIT = f 0 exp ( - KuVlkB T)
As a function of either T or V, particles exhibit sharp transitions between being stable

single domain with very long relaxation time, to a state in which relaxations (T) become
146 Chapter 4
LAMBERTVILLE PLAGIOCLASE
10 " - - - 10.000 OE- _ _ _ _ _ _ _ _ _ _ _ ú ñ J ñ | =
.______. /x
.------.
... .----------.ú = I
/
...
U
Z
Z
CI
...
:I
a:: .4 -
------R--- -- ------ J ú úx:\f =
/: .
____ x /X ú=,
2
___ HRC :
x_ _ x I
a 100 laO 500 1000 2000 5000 10.000

MAGNETIC FiElD (OERSTEDS)
CHITON MAGNETITE
:
1.0 --- -- ----10.000 OE- - - - - - - - - - - - - - - - ú x ---x
=
ú D y =- - - - - - 2.000 DE - - - - - - - - - -7 x
...
Q
ú =
J y y ú ú ú DG? J ú X L =
ú=
:I
a::
Q
Z y Wú K = x
ú ú ú J > J J ú ú Kú =
x'--x Hilc :::::::s;.
b 100 300 2000 5000 10.000
COARSE GRAINED BASALT
1.0 ,---- l.OOO DE - - - -- - - - x:=.=x-_x---x
/
\ e/
.
i
a ...
ú =ú = .6
Ii .4
ú=
\,/'
-!!-¥.
.2 X
..
1\
:H RC ........... - - -
C i J J I J l l ú ú J J ä ú l l J J ú ú Rú l ú l J J ú J J I l l ’ l J J J J J J J J J J ä ú l l l =
Figure 26. Acquisition of magnetization and AF demagnetization curves for (a) weakly interacting
single-domain particles; (b) strongly interacting single-domain particles; and (c) multi domain particles;
X, Acquisition of IRM; " AF Demagnetization of IRM. From Cisowski (1981).
540 Equ,d,menslonol
Fe304" 1% by volume
to
--
KEY
Ozdemlr B ooneqee (1981 )
to 0.025 f-Lm
520 200
• Donkers (978) LM 1
... 500 o Donkers (1978) LM3

0'
0.12 f-Lm QJ
o
to
<I.l
ú= 180 ú = 150
ú =
E
E <I.l
;)
",3 160 ,
I
toO.19f-Lm Q
Q x
x 140 ú N MM=
:E toO.99f-Lm <l:
a:: X
-"t 60
40
Wú ã =
'0 i
20
25-250 f-Lm -GO • KaK ú 0U =
0
o 10 20 30 40 50 60
X x 10-3 (emu/Oe'gr)
(a)
Figure 27. Granulometry by anhysteretic susceptibility (XARM) versus low-field susceptibility plots.
(a) Variation of XARM and X with grain size; (b) model of XARM versus X method. From King et 01.
(1982).
vanishingly small. This state is called superparamagnetism. The term superparamagnetic

alludes to the similarity to "paramagentism," whereby the magnetization is field dependent
with no remanence. The modifier "super" takes note of the fact that it is the magnetization
of the whole superparamagnetic particle which is behaving in this way, not the magnetic
moment of a single atom or molecule.
The superparamagnetic state could be of relevance to the biomagnetic studies dis-
cussed in this book (e.g., Kirschvink and Gould, 1981). Standard methods of detecting the
presence of superparamagnetic material have been developed, which make use of the in-
ability of superparamagnets to exhibit remanent magnetization although they do have
strong susceptibility. We noted that the phenomenon of superparamagnetism was neces-
sarily a function of temperature as well as grain size. The most direct way of demonstrating
that presence of a superparamagnet is therefore to study the relaxation time (T) of mag-
netization as a function of temperature. If the material is in the superparamagnetic state,
there must be some lower temperature at which the particles will achieve stable single-
domain behavior. T will sharply increase at this temperature. The determination of relax-
ation times is particularly easily carried out in the SQUID rock magnetometers: the sample
is magnetized in a field, placed in the magnetometer, and the decay of remanence observed.
The process is then repeated at progressively lower temperatures so that the relaxation
time can be observed as a function of temperature. A quick and dirty, but still effective
way of doing the experiment, is to measure the saturation remanent magnetization at dif-
ferent temperatures. When the grain size is stable single domain, the saturation remanence
will be large, but when it becomes superparamagnetic, the remanence decreases to zero.
Another possibility is to measure susceptibility. As the material changes from single do-
main to superparamagnetic, there will be a large associated increase in susceptibility. These
148 Chapter 4
Hard axis
aligned
S.D.
SUPERPARAMAGNETIC
AT ROOM
TEMPERATURE
SYMMETRICAL
NEGATIVE
Figure 28. Scheme for recognition of domain state by observations with SQUID magnetometers.
observations can also be made with the SQUID rock magnetometers, by trapping a weak
field in the superconducting shield.
The magnetic behavior of magnetite in superparamagnetic, single-domain, pseudo-
single-domain, and multidomain states is summarized in Fig. 28. As has been discussed
in this section, these parameters can be used to establish granulometry schemes. The choice
of the parameters used will be controlled by the availability of equipment to measure them
and the degree to which they uniquely determine domain state.
6. Conclusions
In the past 10 years the SQUID magnetometers have come to dominate magnetic meas-
urements in biomagnetism. Although it is notoriusly difficult to predict the future, which
has a habit of being controlled by factors unknown to those who attempt prediction, it
does seem likely that the SQUID magnetometers will continue to play an important role
in the near future. As experimental techniques evolve, most, if not all, of the observations
required to identify and characterize magnetic material in biological samples will probably
be carried out at high sensitivity with fast response times using the SQUID magnetometers.
Some of the most exciting developments in biomagnetism instrumentation are in areas not
specifically covered by this paper, namely the detection of the electromagnetic fields of
living tissues by the magnetometers. As we noted in the Introduction, the capability to
make magnetocardiograms, magnetoencephalograms and to study nerve impulses promises
a major field of endeavor. Moreover, if there is a magnetic sense developed in some animals,
it should be demonstrable in terms of associated nervous activity. Perhaps one day this
will be done with a SQUID magnetometer. About 10 years ago, in a paper on applications
of SQUIDs in rock magnetism and paleomagnetism (Goree and Fuller, 1976), we com-
mented on how the genius of Josephson and others had permitted a revolution in magnetic
measurements in geophysics within about 15 years of the original research. The technology
which derived from that insight has also made fundamental progress possible in the bi-
ological sciences.
References
Anderson, P. W., and Rowell, J. M., 1963, Probable observation of the Josephson superconducting
tunnelling effect, Phys. Rev. Lett. 10:230-232.
Bannister, J. D., 1966, Spontaneous pressure oscillations in tubes connecting liquid helium reservoirs
to 300 K environments, in: Pure and Applied Cryogenics, Volume 6, Pergamon Press, Elmsford,
0
New York.
Bardeen, J., Cooper, L. N., and Schrieffer, J. R., 1957, Theory of superconductivity, Phys. Rev. 108:1175.
Brandt, R., Nisenoff, M., and Edelsack, E., 1979, Superconductor Materials Science (S. Foner and B.
Schwartz, eds.), Plenum Press, New York.
Cisowski, S., 1981, Interacting vs. non-interacting single-domain behavior in natural and synthetic
Clarke, J., 1974, Josephson junction detection, Science 184:1235-1242.
Clarke, J., Gouban, W. M., and Ketchen, M. B., 1985, A reliable dc Squid made with tunnel junctions,
IEEE Trans. Magn. Mag-ll:724.
Collinson, D. W., 1975, Instruments and techniques in paleomagnetism and rock magnetism, Rev.
Collinson, D. W., Creer, K. M., and Runcorn, S. K., 1967, Methods in Paleomagnetism, Elsevier, Am-
sterdam.
Day, R., Fuller, M., and Schmidt, V. A., 1976a, Magnetic hysteresis properties of synthetic titano-
magnetites, J. Geophys. Res. 81:873.
Day, R., Dunn, J. R., and Fuller, M., 1976b, Intensity determination by continuous thermal cycling,
Phys. Earth Planet. Inter. 13:301-304.
Deaver, B., and Fairbank, W. M., 1961, Experimental evidence for quantized flux in superconducting
cylinders, Phys. Rev. Lett. 7:43.
Doll, R., and Nabauer, N., 1961, Experimental proof of magnetic flux quantization in a superconducting
ring, Phys. Rev. Lett. 7:51.
Erne, S., 1983, Squid sensors, in: Biomagnetism: An Interdisciplinary Approach (S. J. Williamson,
G.-L. Romani, L. Kaufman, and I. Modena, eds.), Plenum Press, New York, pp. 706.
Farrell, D. E., Tripp, J. H., Zanzucchi, P. E., Harris, J. W., Brittenham, G. M., and Muir, W. A., 1980,
Magnetic measurement of human iron stores, IEEE Trans. Magn. Mag-16:818-823.
Feynman, R., Leighton, R. B., and Sands, M., 1965, Lectures on Physics, Volume 3, Addison-Wesley,
Reading, Massachusetts.
Forgacs, R. L., and Warnick, A., 1967, Digital-analog magnetometer utilizing superconducting sensor,
Rev. Sci. Instrum. 38:214-220.
150 Chapter 4
Fuller, M., and Kobayashi, K., 1964, Identification of magnetite and hematite in rocks by magnetic
observations at low temperature, J. Geophys. Res. 69:2111-2120.
Gamble, T. D., Gouban, W. M., Ketchen, M. B., and Clarke, J., 1978, Regulator for controlling liquid
helium bath near 4.2K, Rev. Sci. Instrum. 49:119.
Gibbons, R. M., 1971, Thermophysical data for cryogenic materials, in: Cryogenic Fundamentals (G.
G. Haselden, ed.), Academic Press, New York.
Ginsberg, D. M., 1970, Magnetic susceptibilities of some materials which may be used in cryogenic
apparatus, Rev. Sci. Instrum. 41:1661.
Goree, W. S., and Fuller, M., 1976, Magnetometers using RF driven SQUIDs and their applications in
rock magnetism and paleomagnetism, Rev. Geophys. Space Phys. 14:591-608.
Jaklevic, R. C., Lamke, J., Mercereau, J. E., and Silver, A. H., 1965, Macroscopic quantum interference
in superconduction, Phys. Rev. A 140:1628-1637.
Josephson, B. D., 1962, Possible new effects in superconductive tunnelling, Phys. Lett. 1:251-253.
Josephson, B. D., 1974, The discovery of tunneling supercurrents, Science 184:527-530.
King, J., Banerjee, S. K., Marvin, J., and bzdemir, b., 1982, A comparison of different magnetic methods
for determining the relative grain size of magnetite in natural materials, Earth Planet. Sci. Lett.
59:404-419.
Kirschvink, J. 1., and Gould, J. L., 1981, Biogenic magnetite as a basis for magnetic field sensitivity
in animals, BioSystems 13:181-201.
Kobayashi, K., and Fuller, M., 1968, Stable remanence and memory phenomenon in multi domain
magnetite, Philos. Mag. 18:601-611.
London, F., 1950, Superfluids, Volume 1, Wiley, New York, p. 152.
Long, H. M., and Loveday, P. E., 1968, Safe and efficient use of liquid helium, in: Technology of Liquid
Helium (R. H. Kropschot, B. W. Birmingham, and D. B. Mann, eds.), U.S. National Bureau of
Standards Monograph III.
Lowenstam, H. A., 1962, Magnetite in denticle capping in recent chitons (Polyplacophora), Geol. Soc.
Am. Bull. 73:435-438.
Mercereau, J. E., 1967, Modulated flux flow in superconducting films, in: Proceedings of the Sym-
posium on the Physics of Superconducting Devices, Rep. No. 7023, Naval Research Laboratory,
Washington, D.C.
Mercereau, J. E., 1970, Superconducting magnetometers, Rev. Phys. Appl. 5:13-20.
Molnar, W., 1971, Insulation, in: Cryogenic Fundamentals (G. G. Haselden, ed.), Academic Press, New
York.
Nagata, T., Kobayashi, K., and Fuller, M., 1964, Identification of magnetite and hematite in rocks by
magnetic observations at low temperatures, J. Geophys. Res. 69:2111-2120.
Neel, L., 1955, Some theoretical aspects of rock magnetism, Adv. Phys. 4:191.
Nisenoff, M., 1970, Superconductive magnetometers with sensitivities approaching 10- 10 gauss, Rev.
Phys. Appl. 5:21-24.
Powell, R. L., and Blanpied, W. A., 1954, Thermal conductivity of metals and alloys at low temper-
atures, N.B.S. Circular 556.
Prober, D. E., 1975, Magnetic susceptibilities of construction materials useful for SQUID suscepto-
meters, Ph.D. thesis, Harvard University. Available as Technical Report No. 10, Tinkham Series,
Harvard University.
Romani, G. 1., Williamson, S. J., and Kaufman, 1., 1982, Biomagnetic instrumentation, Rev. Sci. In-
strum. 53:1815-1845.
Salinger, G. 1., and Wheatley, J. c., 1961, Magnetic susceptibility of materials commonly used in the
construction of cryogenics apparatus, Rev. Sci. Instrum. 32:872.
Scott, L. E., VanMeerbake, R. C., Kirk, B. S., and Nubel, G. C., 1968, Storage, distribution and handling,
in: Technology of Liquid Helium (R. H. Kropschot, B. W. Birmingham, and D. B. Mann, eds.),
U.S. National Bureau of Standards Monograph III.
Scott, R. B., 1959, Cryogenic Engineering, Van Nostrand, Princeton, N.J.
Silver, A. H., and Zimmerman, J. E., 1967, Quantum states and transitions in weakly connected su-
perconducting rings, Phys. Rev. 157:317-341.
Vrba, J., Fife, A. A., Burbank, M. B., Weinberg, H., and Brickett, F. A., 1982, Spatial discrimination
in SQUID gradiometer and 3rd order gradiometer performance, Can. J. Phys. 60:1-12.
White, G. K., 1979, Experimental Techniques in Low- Temperature Physics, 3rd ed., Oxford University
Press (Clarendon). London.
Zimmerman, J. E., Thiene, P., and Harding, J. T., 1970, Design and operation of stable RF biased
superconducting point contact quantum devices and a note on the properties of perfectly clean
metal contacts, J. Appl. Phys. 41:1572.
Experimental Techniques and
II
Instrumentation
Most of the advances in the field of magnetite biomineralization during the past 10 years
were made possible by using standard laboratory techniques from a variety of quite separate
fields and adapting them to the problems encountered when searching for nanogram quan-
tities of ferromagnetic minerals in animal tissue. Largely due t6 the rapid pace of these
continuing technical improvements, there is a clear gap in the published literature con-
cerning the methods which have emerged from the synthesis of these various techniques.
Consequently, new workers entering this field have been able to take advantage of the most
recent developments.
Although there are numerous other examples, three papers published in 1984 illustrate
how experiments would have been conducted and interpreted very differently had the
material in this volume been available. One group extracted magnetic material from within
the head of a small vertebrate by boiling the tissue for extended periods. A second group
found large, multi domain particles of volcanic magnetite in the sacculus (part of the inner
ear) of a primitive vertebrate and concluded that they were probably used as a magnetic
compass receptor. The third group embedded heads of a small vertebrate in a plastic resin,
detected strong magnetic r.emanence in them using conventional paleomagnetic tech-
niques, and then concluded that histochemical iron detected in sections of the embedded
tissues was probably part of the magnetoreceptor.
The chemical and rock magnetic properties of magnetite and the magneto physical
properties required for its use in magnetoreception suggest other interpretations of the
results of these experiments. Extremely fine-grained magnetite is oxidized very easily to
maghemite, hematite, and other iron oxides, particularly at elevated temperatures. The
above extraction procedure, therefore, is unlikely to have extracted magnetite intact from
the tissue. Large multidomain grains of magnetite are only weakly magnetic in nature,
although they will acquire a temporary isothermal remanent magnetization (IRM) after
exposure to strong magnetic fields. As a result of the high mass and weak moment of
multidomain particles, gravitational forces would dominate their motion in a viscous me-
dium and make their use in magnetoreception unlikely. Finally, many laboratory materials
that are presumed to be nonmagnetic (such as water, stainless steel dissection tools, Teflon,
and embedding resins) are often highly magnetic because they contain ferromagnetic con-
taminants. Careful control measurements, therefore, are necessary in any study employing
rock magnetic techniques to investigate biogenic ferrimagnetism. Because histochemical
iron is abundant in many tissues and stains specific for magnetite have yet to be developed,
caution also is necessary in studies of magnetite in the tissues where it occurs.
Examples like these illustrate the difficulty of independently synthesizing the ap-
proaches of several disciplines and the need for a detailed discussion of the basic laboratory
techniques now employed in the study of biogenic ferrimagnetism. Although many ad-
ditional techniques such as Mossbauer spectroscopy and various TEM procedures are also
mentioned by some authors in Part IV, the basic operation and use of the SQUID mag-
netometers, tissue extraction techniques for magnetite, the use and limitations of trans-
153
154 Part II Introduction
mission electron microscopy for biogenic minerals, and the construction of shielded lab-
oratory environments have never been adequately described in the literature before for
biomagnetic application. Many of these new and improved techniques and instrments are,
in turn, also of immediate interest to geophysicists as well, and we (the editors) hope this
part in particular will be of use to both fields.
Chapter 5
Detection, Extraction, and
Characterization of Biogenic Magnetite
MICHAEL M. WALKER, JOSEPH L. KIRSCHVINK,
ANJANETTE PERRY, and ANDREW E. DIZON
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 155
2. Magnetometry Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 156
3. Extraction and Characterization of Biogenic Magnetite. . . . . . . . . . . . . . . . . . . . .. 160
4. Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 163
5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 164
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 165
1. Introduction
Several difficulties arise when attempts are made to characterize the deposits of magnetite
found in metazoans. We are usually forced to deal with very small amounts of material,
dispersed in tissues, using indirect methods that are subject to contamination. Magnetite
crystals in the abdomens of bees (Gould et aI., 1978), and in the heads of pigeons (Walcott
et aI., 1979), and other vertebrates (Bauer et aI., this volume; Perry et aI., this volume;
Walker et al., this volume) are submicroscopic «100 nm), occupy a combined volume of
10- 10 to 10- 8 cm 3 , and have a mass of 1-100 ng. In organisms of up to 100 kg or more,
detecting such quantities of magnetite from its magnetic properties depends on the crystals
being highly concentrated in small, recognizable structures, and not uniformly dispersed
throughout all the tissues. Extraction and recovery of the crystals likewise depend on their
being sufficiently concentrated to be magnetically detectable.
The failure to recognize contaminants and the influence they can have on results of
biomagnetic studies has greatly hindered our progress in understanding the origin and
functions of biogenic magnetite. Magnetite is a common industrial pollutant and can often
find its way onto the external body surface or into the gut of higher animals (Kirschvink,
1983). A typical 100-nm crystal of the type found in bees and pigeons (Gould et aI., 1978;
Walcott et aI., 1979) has a moment of about 0.5 fAm 2 whereas a 10-j.Lm dust-sized particle
MICHAEL M. WALKER and ANDREW E. DIZON • Southwest Fisheries Center La Jolla Laboratory,
National Marine Fisheries Service, National Oceanic and Atmospheric Administration, La Jolla, Cal-
ifornia 92038. JOSEPH L. KIRSCHVINK • Division of Geological and Planetary Sciences, Cal-
ifornia Institute of Technology, Pasadena, California 91125. ANJANETTE PERRY • Department
of Oceanography, University of Hawaii, Honolulu, Hawaii 96822.
155
156 Chapter 5
may have moments of up to 500 pAm2. The moment of the multi domain particle is well
within the 1-10 pAm2 sensitivity limits of the superconducting magnetometers currently
in use whereas the moments of 10 3 to 104 of the single-domain particles must be aligned
to be detectable. Other ferro- or ferrimagnetic contaminants are frequently present within
the laboratory environment, particularly in paleomagnetic laboratories where rock samples
tend to leave a fine dust that is often rich in magnetic contaminants. The ease with which
contaminants can enter at all stages of biomagnetic studies dictates that we not only adopt
procedures to minimize the risk of contamination but that we specifically distinguish
between contaminants and true biochemical precipitates. It is therefore necessary to iden-
tify those properties that are likely to be unique to biogenic magnetite.
Interest in biogenic magnetite has focused on its potential use in the transduction of
the geomagnetic field to the nervous system. Kirschvink and Walker (this volume) argue
that the physical properties of the crystals are of primary importance and that single-
domain crystals are the most likely form of magnetite for use in magnetoreception. This
constraint should result in a restricted size-frequency distribution of the magnetite par-
ticles. Magnetite particles suitable for magnetoreception should therefore have coercivities
greater than that of multi domain magnetite «20 mT; Zoeger et 01., 1981) and less than
the theoretical maximum for single-domain magnetite (300 mT; McElhinny, 1973). Su-
perparamagnetic particles of biogenic magnetite (Gould et 01., 1978) are difficult to detect
without special facilities. Thus, in searching for magnetite suitable for use in magneto-
reception, we are primarily attempting to distinguish between single-domain and multi-
domain particles.
There is as yet no evidence that magnetite in the gut or the environment can naturally
enter the bloodstream and be transported to the places it is detected. Any such particles
used for magnetoreception are most likely produced within the organisms themselves,
presumably by enzyme catalysis. The specificity of enzyme pathways would be expected
to result in biogenic magnetite containing few of the impurities associated with geologic
magnetite or metals used to harden iron alloys (Lowen starn and Weiner, 1983). Magnetite
particles suitable for magnetoreception can therefore be reasonably expected to possess
physical and chemical properties distinguishing them from geologic and synthetic mag-
netites.
Magnetite or magnetic material without apparent magnetoreceptive function has been
detected in a variety of tissues in different species (Lowenstam, 1962; Presti and Pettigrew,
1980; Kirschvink, 1981; Kirschvink et 01., 1982). Except in the chitons, where magnetite
is used to harden radular teeth (Lowenstam, 1962), the function of these deposits is un-
known. Hypotheses are that the deposits store excess iron or that they may have a path-
ologic origin (Lowenstam and Weiner, 1983). It is therefore more difficult to predict char-
acteristics that will distinguish them from other magnetites. However, it is important that
attention be given to these deposits because they may predate the use of magnetite for
magnetoreception (Kirschvink and Gould, 1981; Walker et 01., this volume).
This chapter reviews the techniques which are in common use for detecting and char-
acterizing biogenic magnetite. Procedures for avoidance of contamination and examples
of specific tests for contaminants that are easy to conduct are included. Finally, we attempt
to identify further techniques that may prove useful in detecting, extracting, and analyzing
biogenic magnetite.
2. Magnetometry Studies
As noted above and by other authors in this volume, accidental contamination of

samples is a major problem in the search for biogenic magnetite. When working in paleo-
Biogenic Magnetite 157
magnetic laboratories, we have sought to minimize the risks of contamination by thor-

oughly scrubbing and lining the walls, roof, and floor with thin polyethylene sheets. The
recent development of a clean laboratory specifically designed for biomagnetic studies
eliminates many of the contamination problems previously experienced (Kirschvink,
1983). We still find, however, that contamination can enter in lint on the clothing or in
the hair of people using the laboratory. In the future, we hope to eliminate even this source
of contamination by wearing lint-free garments and by using a deionized water shower.
Nonmagnetic tools are a must in carrying out dissections in preparation for measure-
ments in the magnetometer. Typical metal dissection tools such as bone saws, scalpels,
and forceps can leave trails of highly magnetic particles behind them. Even tools made
from nonferrous metals such as aluminum or copper often contain small ferromagnetic
inclusions sufficient to prevent their use in dissections. Magnetic particles left in tissues
by these tools can easily be detected but can only be recognized from extensive tests of
their magnetic properties (see below).
Wood, plastic, and glass seem to be the materials most suitable for tools used in dis-
sections. We found glass microtome knives convenient for dissection and easily obtained
from electron microscopy laboratories. Disposable wood chopsticks are ideal for handling
tissue samples of the sizes measured in the magnetometer. Although they will acquire
magnetic moments, frequent washing and replacement of the chopsticks minimizes the
risk that they will pick up and transfer contaminants to samples. Frequent washing in
glass-distilled water and extended ultrasonic cleaning in either glass-distilled water or 6
N Hel of nondisposable equipment that comes in contact with the samples also serve to
minimize contamination.
The risk of contamination is further reduced if dissections are made from whole car-
casses rather than sections that have been reduced in size using metal saws or knives
(Walker et al., this volume). Saws appear to inject magnetic particles well into tissues.
These particles may be dispersed further during dissection and their' presence and con-
tribution to tissue moments cannot be determined other than by extensive tests. We have
also found from our work with fish that juvenile or subadult animals with incompletely
ossified bones are easier to dissect than adults. In large, thick-boned species such as the
blue marlin, Makaira nigricans, the easiest way to gain access to structures within the
skull was to score along bone sutures with a glass knife and break the bones apart. In turtles
and cetaceans, a rubber mallet or wooden-jawed vice was necessary to crack the skull
bones to provide access to the dura mater membrane (Bauer et 01., this volume; Perry et
01., this volume). Such techniques, while effective, make accurate localization and iden-
tification of magnetite-containing structures difficult.
After tissue samples have been dissected and washed in glass-distilled water, several
of their magnetic properties are of interest. These include the natural remanent magneti-
zation (NRM) , saturation isothermal remanent magnetization (sIRM) , and rate at which
magnetization is acquired or lost in progressively increased aligning or randomizing fields.
Before a tissue sample can be measured, it must be frozen so that any small magnetic
particles present will be immobilized. Otherwise, in the null field environment of the
magnetometer, the orientation of any magnetic particles suspended in a viscous medium
will be randomized by Brownian motion and any moments due to alignment of the particles
will be lost.
Simply freezing a sample is sufficient only for measuring its NRM. Determination of
other magnetic properties requires that the sample be magnetized prior to measurement.
When inducing the sIRM in a sample, it is desirable to expose it to a strong uniform field
in excess of 300 mT. In our early work we attempted to do this using a cobalt-samarium
magnet. Unfortunately, it is difficult to obtain homogeneous magnetization of large samples
using this method because of the rapid decay of field strength with distance from the
magnet. Inhomogeneous magnetization can lead to underestimates of the amount of mag-
158 Chapter 5
netic material present, making it possible to miss potentially important magnetic structures.
An air core solenoid delivering homogeneous fields of up to several Tesla has proven very
reliable in uniformly magnetizing samples for measurement in the magnetometer. This
solenoid also makes possible the progressive magnetization of samples in coercivity studies
(Kirschvink, 1983).
Great care is necessary in the choice of sample holders for magnetometry experiments.
The mylar, glass, or polyethylene plastic tubes commonly used by paleomagnetists are
adequate for use with biomagnetic samples, although they will acquire spurious moments
and contribute to background noise in the magnetometer. They will also acquire magnetic
moments if exposed to strong magnetic fields, so that samples must be magnetized sep-
arately from these holders and then loaded for measurement. Consequently, it is difficult
to maintain the same orientation of the sample to components of the process (solenoids
and magnetometer detection coils) in repeated measurements.
We have found two simple methods of attaching samples to a holder that maintain
constant orientation of the sample relative to the axes of the solenoids and magnetometer
pickup coils. A magnetized, frozen sample can be attached to the moistened end of a white
cotton thread and lowered vertically into the magnetometer. If thoroughly cleaned, the
thread does not show any NRM and can be used for repeated measurements such as al-
ternating field (AF) demagnetization. However, the "clean" threads may sometimes gain
moments if exposed to strong fields. Another method of loading the samples for rapid IRM
acquisition studies was therefore required. We found that a more effective sample holder
was a hook made of quartz glass fiber. The hook was inserted into the unfrozen tissue and
left within it throughout all the measurements. With this apparatus and the air core impulse
or AF solenoid mounted in line on the magnetometer, we were able to automate our ex-
periments and minimize time spent handling samples during repeated measurements. Con-
trol experiments conducted with the quartz fiber hook attached to an ice cube showed that
the fiber possessed no natural moment and did not acquire a moment even in high inducing
fields. However, such fibers have the disadvantage that they are fragile and break easily.
There are several paleomagnetic techniques that can be adapted for identifying and
characterizing the properties of biogenic magnetite. The most important of these is to de-
termine the coercivity spectrum of the particles present in a tissue sample. The coercive
field of a magnetite particle is the minimum intensity of an external field required to flip
the moment of the particle from one of its stable orientations in the long axis of the particle
to the other. The range of applied fields over which a sample acquires or loses magneti-
zation is dependent on the coercivities of the magnetic particles present in the sample.
Thus, the coercivity spectrum can eliminate a variety of minerals like hematite and goethite
as possible sources of remanence, and can give information concerning the size and shape
of any magnetite fraction present.
Two methods are available for determining the coercivity spectrum of particles present
in a sample: progressive IRM acquisition and AF demagnetization. An external field, B,
will shift the moments of magnetite particles with coercive fields less than B cos 6, where
6 is the angle between the particle and field vector directions. In IRM acquisition exper-
iments, the moments of particles in a sample are aligned by progressively increased fields.
In AF demagnetization experiments, the orientation of aligned moments of particles with
coercivities less than the peak applied field is randomized by sinusoidally oscillating fields
which slowly decrease in magnitude. Although the use of the cotton thread or quartz fiber
techniques described above makes these iterative measurements relatively easy, they make
it impossible to correct for the effect of the angle between the direction of the crystal axes
and the applied fields and cause slight overestimation of the coercivities of the particles.
However, for distinguishing single-domain magnetite from multi domain particles, and also
from high-coercivity minerals such as hematite, this error can be ignored. Thus, acquisition
100
--
0ú =
80
•
Af demap
I»
tit
0
IU
ú = 60
U
IU
C
ú =
0 40
Figure 1. Comparison of IRM ac-
quisition and AF demagnetiza- I» tit
tion experiments on an ethmoid 0IU
region sample from a human. u ú =
20
100% represents saturation mag- .5
netization and 0% represents the
natural remanent magnetization 0
of the sample. Sample supplied 3 ú= 10 30 úç = 100 300500 1000
by R. R. Baker (Department of Zo-
ology, University of Manchester). Peak Field, mT
of IRM in fields less than 20 mT or greater than 300 mT will indicate the presence of
magnetic contaminants in samples.
Plotting the results of IRM acquisition and AF demagnetization experiments together
can reveal much about the nature of the magnetic particles. For single-domain particles
of the type predicted for use in magnetoreception, magnetization will be acquired or lost
over a relatively small range of fields. Where all the particles are of single-domain size
and uniformly dispersed throughout a sample, the two curves should be mirror images of
each other over the same ranges of intensity of the applied fields. Where the crystals are
sufficiently close to interact with each other, IRM acquisition is inhibited and AF de-
magnetization is aided (Cisowski, 1981). Thus asymmetry of the curves about the 50%
magnetization point indicates the degree of interaction of the particles.
An example is given in Fig. 1 of the use of IRM acquisition to detect high-coercivity
contaminants in a sample of bones and tissue from the ethmoid region of a human head.
When we first tested this sample, we believed it had been dissected using nonmagnetic
instruments. However, the sample turned out to be far more magnetic than any other bi-
ological sample we have examined (moments >104 pA m2 ). The IRM acquisition curve
was incompatible with the presence of magnetite alone, as the sample did not become
saturated by 300 mT but continued to acquire magnetization up to fields of 800 mT. The
AF demagnetization curve also showed the extraordinary stability of the magnetic material
present in the sample. The tissue retained the majority of its magnetization up to fields
of 100 mT, the upper limit of our demagnetization unit. From these data we could only
conclude that there must have been high-coercivity contaminants in the tissue. On check-
ing, R. R. Baker (Department of Zoology, University of Manchester, personal communi-
cation, 1982) discovered that an assistant had inadvertently trimmed the sample with a
band saw. It would thus be impossible to conclude that biogenic magnetic particles con-
tained within the sample had been detected and identified as magnetite suitable for use
in magnetoreception.
Where suitable facilities are available, several other techniques have proven or should
prove useful in identifying and characterizing magnetite present in biologic samples. The
identity of the mineral can be determined from its Curie temperature, a procedure used
to identify magnetite in pigeons and honeybees (Gould et al., 1978; Walcott et al., 1979).
160 Chapter 5
The loss of IRM on warming through the isotropic point of magnetite has been used to
identify multi domain magnetite in the Pacific dolphin (Zoeger et al., 1981). As it distin-
guishes single-domain from multidomain material, this test is potentially very useful in
demonstrating the presence of contaminants in biologic samples. Finally, the presence of
superparamagnetic crystals can be demonstrated by continuously monitoring the remanent
moment of a sample versus temperature as it warms from liquid nitrogen temperature
(77°K) to room temperature (293-298°K). Freezing the samples to liquid nitrogen temper-
ature shifts the boundary between single-domain and superparamagnetic behavior toward
smaller particle sizes. Thus, as the crystals warm through their single-domain/superpar-
amagnetic transition, they will lose their stable remanence. A drop in total moment will
accompany this transition and the temperature at which it occurs will indicate the ap-
proximate sizes of the crystals. This experiment demonstrated the presence of over 108
such particles with sizes of 30-35 nm in the honeybee (Kirschvink and Gould, 1981) as
well as in one species of chiton (Kirschvink and Lowenstam, 1979).
3. Extraction and Characterization of Biogenic Magnetite
Much can be learned about the nature and organization of the magnetic material dis-
covered in biologic samples using its bulk magnetic properties as discussed above. How-
ever, it is eventually necessary to extract the material from the sample and apply a range
of techniques to its identification and characterization. Areas of highest magnetite con-
centration must be accurately identified if sufficient quantities of material are to be ob-
tained for analysis. In the fish this was done by magnetometric studies that exhaustively
sampled the tissues of one species, the yellowfin tuna, Thunnus albacares, until tissues
containing high magnetic remanence could be reliably located (Walker and Dizon, 1981;
Walker et a1., this volume). We were able to identify one specific and relatively small
structure, the dermethmoid bone, that was always magnetic in the tuna and in all the fish
species subsequently examined. Our experiments showed that the magnetite was concen-
trated in tissue contained within a sinus formed within the dermethmoid bone.
In other vertebrates, magnetite appeared to be concentrated in the dura mater (Bauer
et a1., this volume; Perry et a1., this volume) or in a region of the skull similar to the fish
(Mather and Baker, 1981; Baker et a1., 1983; Baker, Chapter 26, this volume; Mather, this
volume). In the turtles and cetaceans, subdivision of the dura revealed localization of
magnetic material within its anterior portions. Concentration on these regions made for
efficient extraction of magnetite in the green turtle, Chelonia mydas (Perry et al., this
volume).
For invertebrates the situation is less clear. Deposits of magnetite or magnetic material
have been located from magneto me try and coercivity studies in the abdomen of the ho-
neybee (Gould et a1., 1978) and in the head-thorax of the monarch butterfly (Jones and
MacFadden, 1982). There appears to be no reason why magnetite should not be detected
in the bodies of other invertebrate groups. If easily accessible magnetite-containing struc-
tures can be identified, the generally smaller size of invertebrates compared to vertebrates
could make them more suitable for magnetite extraction studies.
Once the magnetic structures have been identified, it is a simple matter to dissect and
combine a number of them for magnetite extraction. In this way we were able to treat the
dermethmoid tissues of up to five yellowfin tuna at once. The tissues were ground with
a little distilled water in a glass tissue grinder or in a test tube using a nonmagnetic pestle.
In the tuna this released fat and oil droplets into suspension. These were removed by
adding anhydrous ether to the suspension and shaking vigorously. After the aqueous and
ether phases separated, the ether was decanted and replaced. This procedure was repeated
until the aqueous phase became clear.
After ether extraction, the suspension was centrifuged, the aqueous supernatant re-
moved, and 5% Millipore* filtered hypochlorite solution (commercial bleach) added. The
mixture was centrifuged and the hypochlorite replaced periodically either until the tissue
disappeared completely or until no further digestion occurred. When digestion was com-
plete, the suspension was centrifuged and the supernatant replaced with distilled water.
This washing procedure was repeated at least five times. In the tuna a white residue as-
sociated with the magnetic material remained after tissue digestion. Treatment with buff-
ered EDTA (pH 7.1) carried out in similar fashion to the hypochlorite digestion freed the
crystals from the residue. They could then be separated magnetically from the residue
under a dissecting microscope.
To remove the magnetic particles for analysis, we held them in suspension using a
cobalt-samarium magnet held to the side of the test tube, and then pipetted them onto a
slide coated with xylene-based cement. The water in which the crystals were pipetted
onto the slide was allowed to evaporate and a second layer of cement covering the first
applied. The crystals were thus sealed in a cement sandwich which could be cut out,
removed from the slide, and placed in the beam of a mini-Debye Scherrer X-ray camera.
For electron microprobe analysis we pipetted the crystals onto clean microscope slides,
allowed them to dry completely, and transferred them to slides coated with epoxy resin.
It was then a simple matter to cure the resin, polish the crystals, and coat them with carbon
for electron microprobe analyses. Similarly, aggregates can be pipetted onto plugs and
prepared for examination in the scanning electron microscope (SEM) (see Perry ef a1., this
volume).
We exposed the crystals to Mo K.. X-irradiation (48-72 hr) in the mini-Debye Scherrer
camera. Development of the film and measurement of the band patterns was then routine
(see Perry ef a1., this volume; Walker ef a1., this volume). The procedure for electron mi-
croprobe analyses was more complex. These analyses determine the elemental composition
of minerals and permit tests of the purity and origin of the magnetite crystals. A survey
using EDAX probe was an appropriate beginning point as it indicated all the elements
present and their relative proportions in any sample (see Perry ef aI., this volume). Key
elements were then selected for quantitative analyses. In the tuna and turtle, we analyzed
for oxides of iron against a magnetite standard and for oxides of rare earth metals such as
titanium and manganese, which are commonly found as impurities in geologic magnetite
(Perry et a1., this volume; Walker et aI., this volume). We also chose to analyze for calcium
in an attempt to determine how closely the residue remaining after hypochlorite digestion
was associated with particle aggregates in the tuna.
Our analyses of the magnetite from the green turtle and the tuna showed that there
were very few oxides other than oxides of iron in the material. Although the standard
(NMNH 11487) used in the quantitative analyses was an unusually pure geologic magnetite
(M. O. Garcia, Hawaii Institute of Geophysics, University of Hawaii, personal communi-
cation), it still contained measurable oxides of the rare earth metals titanium and chro-
mium, whereas the biologic magnetites did not (see Perry et a1., this volume; Walker et
al., this volume). These data, taken with the absence of nickel from the turtle samples
(Perry et a1., this volume) and very small amounts of manganese in magnetite from both
the tuna and the turtles, strongly suggested to us that the magnetite was neither synthetic
nor geologic in origin.
The very small amounts of material obtained by digestion made it impossible to use
the approach developed by Towe (this volume) for preparing the crystals for transmission
electron microscopy (TEM). Chang (Division of Geology and Planetary Science, California
Institute of Technology, personal communication) developed a method for obtaining dis-
* Reference to trade names does not imply endorsement by the National Marine Fisheries Service,
NOAA.
162 Chapter 5
12
0
10
.!'
"-
vi 8 -
LU
t=
iii
zLU
6
I-
ú=
LU
> 4 -
ú=
..J
::-
LU
0::
l\i
2
00 0.05 0.20 0.25 0.35

d - SPACINGS (nm)
Figure 2. X-ray diffraction data for magnetic particles extracted from mouse tumors supplied by F. L.
Tabrah and S. Batkin (John A. Burns School of Medicine, University of Hawaii). Vertical lines indicate
relative intensities of the lines in the diffraction pattern. Numbers in parentheses indicate lines as-
sociated with metallic iron and the crystal plane giving rise to each line. Lines at d spacings of 0.30
and 0.25 nm could come from either magnetite or maghemite. The sources of the remaining lines in
the pattern are unknown.
persed crystals that compensated for this limitation. The aggregated crystals were pipetted
onto carbon-coated copper mesh grids and dispersed in an alternating magnetic field (100
mT). The grids were then air-dried and prepared for examination in TEM (see Walker et
aJ., this volume).
Extraction of the magnetic material immediately produced a substantial amount of
information that assisted in its identification. For example, the color of the particles ex-
tracted from the tuna and the turtle was sufficient to exclude maghemite. Maghemite has
magnetic properties similar to magnetite and therefore cannot be excluded as a source of
magnetic remanence by coercivity studies. However, tests were still necessary to identify
the crystals uniquely, demonstrate their biologic origin, and to exclude the possibility that
contaminants may have entered during dissection or extraction. For example, the particles
extracted from the yellowfin tuna and green turtle were uniquely identified as magnetite
by X-ray diffraction (Perry et aI., this volume; Walker et aI., this volume). In contrast, X-
ray diffraction of magnetic material extracted from tumors that had been shown to contain
probable single-domain magnetite crystals revealed native iron, and a mineral that could
have been magnetite or maghemite (Fig. 2). On the basis of color, maghemite could be
excluded. However, the iron was presumably a contaminant arising from an as yet unre-
cognized source.
Diffraction patterns are not conclusive proof of the origin of magnetite particles ex-
tracted from tissues, and care is necessary in their interpretation. Pure, fine-grained mag-
netite powders such as those predicted for use in magnetoreception will ideally yield sharp,
unambiguous diffraction patterns. Streaking of the spots or lines in X-ray and electron
diffraction patterns could arise from more than one source. Towe and Moench (1981)
suggest that vacancy defects in the crystal structure could have caused streaking of an
electron diffraction pattern taken from single-domain magnetite crystals purified from mag-
netotactic bacteria. However, multi domain particles could also give streaked diffraction
patterns. Therefore, procedures that combine identification of the particles with deter-
mination of their domain state are necessary. Electron diffraction and measurement of the
size and shape of isolated crystals can be conducted on the same samples mounted on
copper mesh grids. Thus, although electron diffraction is a more cumbersome technique
than X-ray diffraction, it does provide a conclusive test of the origin of the particles when
carried out in conjunction with determination of their size and morphology (Towe, this
volume; Walker et aI., this volume).
4. Discussion
The magnetometric methods developed so far permit detection and characterization
of the bulk properties of concentrations of magnetite in organisms. It is also now possible
to conduct analyses of polycrystalline aggregates and isolated crystals of magnetite ex-
tracted from magnetic tissues. An important conclusion arising from these studies is that
although it is relatively easy to detect the presence of magnetic material in organisms, it
is far more difficult to determine its origin and what, if anything, it does. We have attempted
to define a theoretical basis for determining what the form of biogenic magnetite used in
magnetoreception should be, and the experimental evidence to date has been consistent
with the theory. This permitted refinement of the hypothesis on the organization of mag-
netite-based magnetoreceptor organelles (Kirschvink and Gould, 1981; Kirschvink et aI.,
in press; Kirschvink and Walker, this volume; Walker et al., this volume), providing
opportunities for further tests of the ferromagnetic magnetoreception hypothesis.
An important result of this work is the demonstration of the need to test for contam-
inants at all stages of the research, and to check for consistency of results obtained from
different techniques. The differences between the bulk magnetic properties of the dura
mater of the Pacific dolphin (Zoeger et aI., 1981), the sample shown in Fig. 1, and the
ethmoid tissues of pelagic fish (Walker et al., this volume) illustrate quite clearly the effects
contaminants can have on magnetometry results. The mouse tumors from which we ex-
tracted magnetic particles had previously been shown to be magnetic and probably to
contain single-domain magnetite (Kirschvink et aI., 1982). X-ray diffraction of the magnetic
particles extracted from one of the tumor strains demonstrated the presence of native iron,
with the possible presence of magnetite. The presence of iron conflicts with the magne-
tometry data and suggests that contaminants may have entered during the extraction. Thus,
magnetic material detected in tissues and obtained by extractions should be shown to be
biogenic by independent means wherever possible.
The property of biogenic magnetite crystals that truly sets them apart from their syn-
thetic and geologic counterparts is their nonoctahedral crystal morphology in TEM (Towe
and Moench, 1981; Matsuda et aI., 1983; Walker et aI., this volume), a property not essential
to the magnetite-based magnetoreception hypothesis. This observation suggests that the
distinctive properties of biogenic magnetite arise from the biomineralization process. In
line with this, it seems reasonable that the chemical composition of biogenic magnetite
would be different from geologic magnetite. A potentially much more important charac-
teristic is the ratio of the oxygen isotopes present in biologic compared with nonbiologic
magnetite (Lowenstam and Kirschvink, this volume). At present, technology and availa-
bility of material limit the utility of oxygen isotopes for the recognition of biogenic mag-
netites. Currently available mass spectrometers require a minimum of about 20 fLg of sam-
ple, which is the equivalent of quantitative extraction of the magnetite from the
dermethmoid tissue of at least 100 yellowfin tuna. The potential for entry of contaminants
in such a large-scale extraction is enormous. An ion microprobe may require less sample
than a mass spectrometer and we are investigating the adaptation of this apparatus for
analysis of biogenic magnetites.
Thus, a substantial case can be made for our ability to determine what is and is not
biogenic magnetite suitable for use in magnetoreception. This does not imply that the case
164 Chapter 5
for magnetite-based magnetoreception is proven. However, the approach that has helped
in recognition of magnetite suitable for magnetoreception may prove useful in predicting
the important properties of other deposits of magnetite found in metazoans. Characteri-
zation of these deposits is important because they may predate the use of magnetite in
magnetoreception (Kirschvink and Gould, 1981; Walker et a1., this volume), and because
it is necessary to be able to distinguish the two from each other and from contaminants.
Simple hypotheses for the origin of biogenic magnetite are that it constitutes some
form of iron storage or is deposited pathologically (Lowenstam and Weiner, 1983; Walker
et al., this volume). In support of these hypotheses is the observation that anomalous
deposits of magnetic material are found more frequently in larger and older organisms (J.
G. Mather, Department of Zoology, University of Manchester, personal communication,
1981; Perry, unpublished data; Walker, unpublished data). Crystal size and form may be
less important in these cases than in magnetite used in magnetoreception. Presumably,
enzyme catalysis will be necessary to precipitate the mineral, so the deposits are still likely
to be chemically distinctive. Study of the biomineralization of other storage or patholog-
ically formed minerals may be instructive in helping predict the likely properties of bio-
genic magnetite not used in magnetoreception. This may also assist in predicting where
such deposits are likely to form and in their subsequent detection.
By carrying out studies with magnetite specifically in mind, we risk failing to recognize
other interesting biologically formed magnetic minerals. In recent years the number of
known biogenic minerals has increased substantially (Lowenstam, 1981; Lowenstam and
Kirschvink, this volume). In carrying out biomagnetic studies, we should therefore be
prepared to adopt appropriate approaches for correct identification and characterization
of other magnetic minerals. The basic magneto metric techniques suitable for detecting
these minerals are already well established through paleomagnetic studies. Buffered en-
zyme extractions aimed at digesting specific components of magnetic tissues may be a
suitable approach for extracting magnetic minerals unaltered. The specific methods for
detecting, extracting, and characterizing biogenic magnetite have undergone considerable
development in the last year or two. We expect this development to accelerate as interest
in biomineralization grows and hope that this chapter serves as a stimulus to development
of new, more efficient approaches.
5. Summary
Magnetite detected in cells or tissues is part of a recently described class of biologically

formed minerals and provides a possible physical basis for magnetoreception in living
organisms. Detection of magnetic material alone adds little to understanding of either the
biomineralization process or the possible use of the material in magnetoreception. Further
progress requires development of procedures that will distinguish between magnetic con-
taminants and biologic precipitates in tissue. Hypotheses concerning the nature and role
of the magnetic material can then be tested.
This chapter considers the special case of magnetite that is likely to be used in mag-
netoreception and attempts to define properties that will distinguish it from geologic and
synthetic magnetites. Methods developed to detect and characterize the magnetic prop-
erties of such material and to distinguish it from contaminants are then described. A similar
approach is used to develop techniques for extraction, identification, and characterization
of the properties of individual crystals and crystal aggregates. Examples of tests that have
identified magnetic material other than magnetite that could be used in magnetoreception
are given for magnetometry, extraction, and histologic studies.
The studies demonstrate that the biomineralization process is responsible for pro-
ducing the key distinguishing features of biologically formed magnetite. Although invasive
and noninvasive techniques can distinguish between biologic and nonbiologic magnetite,
use of both approaches at once can provide considerably more information on the nature
of the magnetic material as well as checks on results obtained by either set of techniques.
Extension of the approaches developed in this chapter to other magnetic minerals with
other possible roles in the physiology of the living organism could be equally fruitful.
ACKNOWLEDGMENTS. We are greatly indebted to Charles E. Helsley, Barbara H. Keating, and

Li Chung Ming, Hawaii Institute of Geophysics, University of Hawaii, for the use of pa-
leomagnetic laboratory facilities and assistance with X-ray diffraction analyses. Karla A.
Peterson assisted in magnetometry experiments at the California Institute of Technology.
Research funds and facilities were most generously provided by the Southwest Fisheries
Center Honolulu Laboratory, National Marine Fisheries Service, NOAA. This research was
supported in part by a graduate study award from the East-West Center, Honolulu, and a
grant from Sigma Xi to M.M.W. This is Contribution No. 3939 from the Division of Geo-
logical and Planetary Sciences, California Institute of Technology.
References
Baker, R. R., Mather, J. G., and Kennaugh, J. H., 1983, Magnetic bones in human sinuses, Nature 301:78-
80.
Gould, J. 1., Kirschvink, J. 1., and Deffeyes, K. S,. 1978, Bees have magnetic remanence, Science,
201:1026-1028.
Jones, D. S., and MacFadden, B. J., 1982, Induced magnetization in the monarch butterfly, Danaus
plexippus (Insecta, Lepidoptera), J. Exp. BioI. 96:1-9.
Kirschvink, J. L., 1981, Ferromagnetic crystals (magnetite?) in human tissue, J. Exp. BioI. 92:333-335.
Kirschvink, J. L., 1983, Biogenic ferrimagnetism: A new biomagnetism, in: Biomagnetism: An Inter-
disciplinary Approach (S. J. Williamson, G.-L. Romani, 1. Kaufman, and I. Modena, eds.), Plenum
Press, New York, pp. 501-532.
Kirschvink, J., and Gould, J. L., 1981, Biogenic magnetite as a basis for magnetic field detection in
Kirschvink, J. 1., and Lowenstam, H., A., 1979, Mineralization and magnetization of chiton teeth:
Lett. 44:193-204.
Kirschvink, J. L., Tabrah, F. 1., and Batkin, S., 1982, Ferromagnetism in two mouse tumours, J. Exp.
BioI. 101:321-326.
Kirschvink, J. 1., Walker, M. M., Chang, S.-B. R., Dizon, A. E., and Peterson, K. A., Chains of single-
domain magnetite particles in chinook salmon, Oncorhynchus tshawytscha, J. Compo Physiol., in
press.
Kuterbach, D. A., Walcott, B., Reeder, R. J., and Frankel, R. B., 1982, Iron-containing cells in the honey
Am. Bull. 73:435-438.
Lowenstam, H. A., and Weiner, S., 1983, Mineralization by organisms and the evolution of biomi-
McElhinny, M. W., 1973, Paleomagnetism and Plate Tectonics, Cambridge University Press, London.
Mather, J. G., and Baker, R. R., 1981, Magnetic sense of direction in woodmice for route-based nav-
igation, Nature 291:152-155.
Matsuda, T., Endo, J., Osakabe, N., Tonomura, A., and Arii, T., 1983, Morphology and structure of
biogenic magnetite particles, Nature 302:411-412.
166 Chapter 5
Presti, D., and Pettigrew, J. D., 1980, Ferromagnetic coupling to muscle receptors as a basis for geo-
magnetic field sensitivity in animals, Nature 285:99-101.
Planet. Sci Lett. 52:213-220.
Walcott, c., Gould, J. L., and Kirschvink, J. L., 1979, Pigeons have magnets, Science 205:1027-1029.
Walker, M. M., and Dizon, A. E., 1981, Identification of magnetite in tuna, Eos 62:850.
Zoeger, J. Dunn, J. R., and Fuller, M., 1981, Magnetic material in the head of the common Pacific
dolphin, Science 213:892-894.
Chapter 6
Studying Mineral Particulates of
Biogenic Origin by Transmission
Electron Microscopy and Electron
Diffraction
Some Guidelines and Suggestions
KENNETH M. TOWE
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
2. Sample Preparation for Electron Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
2.1. Preliminary Examination of the Specimen . . . . . . . . . . . . . . . . . . . . . . . . . . 168
2.2. Isolation and Separation of Mineral Phases . . . . . . . . . . . . . . . . . . . . . . . . . 169
2.3. Preparing Specimen Support Mounts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
2.4. Applying the Specimen to the Substrate . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
2.5. Shadowcasting and Surface Replication. . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
2.6. Tissue Preparation Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
3. Studying the Sample in the Microscope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
3.1. Preparing the Microscope for Use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
3.2. Working with the Microscope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
3.3. Identification of Particulates by Electron Diffraction . . . . . . . . . . . . . . . . . . . . 176
4. Analysis of Electron Diffraction Powder Patterns. . . . . . . . . . . . . . . . . . . . . . . . . 178
5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Selected References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
1. Introduction
The accurate characterization of very finely divided particles is often best accomplished
by means of electron microscopy and, where feasible, electron diffraction. The purpose of
this chapter is to describe or comment on some of the methods that I have found useful
for the isolation, identification, and characterization of mineral particles of biogenic origin
using techniques designed for the transmission electron microscope (TEM). These tech-
niques are useful for determining the size, shape, and mineralogy of both crystalline and
amorphous materials. The methods employed are generally applicable with minor mod-
ifications to both magnetic and nonmagnetic phases. As most textbooks on electron mi-
croscopy well illustrate, the variations on this theme are many. What follows are ap-
KENNETH M. TOWE • Department of Paleobiology, Smithsonian Institution, Washington, D.C.

20560.
167
168 Chapter 6
proaches that have been used successfully in a variety of instances by the author. The
experienced transmission electron microscopist may wish to ignore much of what follows
and will even occasionally remark, "I know a better way"; this chapter is not written for
them. The content is not intended to be complete and rigorous. It is intended more as a
gathering of ideas, hints, and suggestions that may prove useful to researchers, who are
unfamiliar with TEM and electron diffraction, in general, or as it is applied to mineral
samples specifically.
Because each type of specimen has its own idiosyncrasies, there is no clear-cut meth-
odology that can be universally applied. The reader is warned against using this, or any
other guide as a "cookbook" manual. Successful TEM requires adaptation to the individual
needs of the particular specimens being studied. A flow-sheet approach is not easily re-
alized and the techniques are therefore as much an art as a science.
2. Sample Preparation for Electron Microscopy

Although the operation of the electron optical equipment is important, it is easily
fair to say at the outset that good specimen preparation is the key to successful electron
microscopy. The best equipment in the world cannot overcome poor sample preparation,
and if the specimen has been well prepared for the electron microscope, more than half
of the battle has been won.
2.1. Preliminary Examination ofthe Specimen
It may seem too obvious to suggest that a preliminary examination of the specimen
be undertaken before attempting to prepare it for electron microscopy, but commonly some
aspect of the sample is observed that may be important to the techniques to be applied,
or something is overlooked that later is deemed of significance. Therefore, it behooves the
investigator to take the time to study the specimen at a lower magnification before being
faced with the "needle-in-the-haystack" specter often raised by the high magnifications
of the TEM. Thus, the physical and chemical nature of each sample needs to be addressed
and a number of questions answered before moving forward. The following is a list of
some of the questions that are worthwhile considering at the beginning, although it is
recognized that many of them may not be answerable even under ideal conditions.
1. Can the mineral phases be seen individually with the light microscope? If so, then
they are almost certain to be too thick to transmit electrons without modification, and the
use of the TEM may even be unnecessary.
2. Is there more than one inorganic phase present within the sample? If so, difficulties
may arise in distinguishing one from another in the black-and-white world of electron
microscopy. Confusion in the evaluation of electron diffraction patterns can also arise.
3. Is there likely to be sufficient material to use for more than one analysis? In other
words, will the TEM study of the sample be a one-shot affair? If so, it may not be worthwhile
to use the TEM until, or unless more information on the nature of the sample has been
obtained through other methods of analysis.
4. Does other analytical data tell anything about either the mineralogy or the particle
size that could be useful in sample preparation? For example, does X-ray diffraction data
show that the phases are crystalline, but of very fine crystallite size, i.e., does the Debye-
Scherrer pattern show evidence of line broadening?
5. Have adventitious contaminant grains been looked for and removed from the spec-
imen to avoid the potential for misinterpretation?
TEM and Electron Diffraction 169
The answers to these and other related questions will lay the foundation for a plan
of specimen preparation for TEM. The more that is known about the sample beforehand,
the easier, and more productive the electron microscopy is likely to be.
2.2. Isolation and Separation of Mineral Phases
If a preliminary examination of the specimen indicates that an isolation of the inor-

ganic phases is feasible, the first step in sample preparation of particulate materials is the
removal of the mineral grains from the organism, and the separation of as much adhering
organic matter as possible. In general, if the mineral phases can be upgraded somehow
from the unwanted tissue by dissection, it is useful to do so. If the mineral matter occurs
in macroscopic, concentrated form, as is the case, for example, with the magnetite teeth
of some marine chitons, the minerals can be physically and mechanically separated first.
Careful dissection with nonmagnetic forceps and needles and concentration with a fine
camel-hair brush can be useful.
On the other hand, scattered mineral grains, such as those found in the magnetotactic
bacteria, cannot be handled in this fashion and other techniques need to be devised. Two
approaches, working together, have proven useful: the use of chemicals to solubilize the
organic matter combined with ultrasonic bath vibration to reduce aggregation. This process
begins with clean, 5-ml glass clinical centrifuge tubes. Small, approximately equal amounts
of the isolated tissues, or microorganisms, are placed in even numbers of tubes. Each tube
is then filled about two-thirds full of commercial household bleach (5.25% sodium hy-
pochlorite solution). Unequal amounts of tissue plus liquid in any of the centrifuge tubes
will unbalance the centrifuge and should be avoided. If only one tube contains the sample,
another containing a balancing amount of water can be used to offset it. The centrifuge
tubes are then placed in the water bath of an ultrasonic vibrator and allowed to undergo
ultrasonic treatment for several minutes. The bleach will react with proteinaceous material
rather quickly, but with carbohydrate material more slowly, if at all. One approach has
been to leave the samples in the bleach for several hours after the initial ultrasonic treat-
ment. The tubes are then placed into the clinical centrifuge and the undissolved material
spun down as a pellet. After the inorganic matter has been pelleted, new bleach is added
and the pellet resuspended with ultrasonic vibration. This process is repeated until it
appears that most of the organic matter has been dissolved. The bleach-treated grains
should then be washed in distilled water using the same routine of washing-pelleting-
resuspending until the bleach has been substantially removed.
As with any procedure, there are a few potential dangers that can arise: (1) Heat should
not be applied to assist the kinetics of organic matter dissolution because of the danger of
concomitant alteration of the mineralogy and/or the crystallinity of the inorganic material
being isolated. (2) The centrifuge pelleting steps should be of sufficient duration so as to
permit most of even the smallest mineral grains to reach the bottom of the tube. This can
be crudely evaluated by placing the centrifuge tubes in front of a light source. If the liquid
is clear, without a Tyndall effect, most of the material has been pelleted. (3) The danger
of alteration of the mineral phases by the bleach must be entertained. The potential for
oxidation of some of the reduced iron in magnetite exists, but experience has shown that
this is seldom realized. X-ray diffraction monitoring of such samples both before and after
extensive bleach treatment has failed to show the formation of new mineral phases. Never-
theless, the reader may wish to establish that no changes of this type have taken place.
(4) One further caveat: the oxidation of organic matter by the chlorine bleach could, in
some instances, release iron from chelated, or other unmineralized positions within an
organism. This iron could hydrolyze in the distilled water to form colloidal hydrous ferric
oxides not otherwise present. Although no instance of such an artifact has been docu-
170 Chapter 6
mented, it remains as a potential pitfall in any isolation procedure involving oxidation of

organic matter to remove mineral grains.
Once the mineral materials have been isolated and separated from the organic moieties,
they are generally ready for study. Because the TEM has a special limitation on specimen
thickness, a further treatment of the sample may be necessary to eliminate coarser grains.
Depending on the mean atomic number of the elements involved and their crystallographic
arrangement, specimens thicker than about 0.1 flm are seldom anything but silhouettes in
the TEM. Therefore, only those mineral grains thinner than this will be usefully studied
from a particulate mount. Coarser-grained preparations are better studied by scanning elec-
tron microscopy, reflected light, or replica methods.
2.3. Preparing Specimen Support Mounts
All particulate specimens to be studied by TEM need to be placed onto suitable spec-
imen mounts. The usual textbook procedure is to apply a drop of the dispersed particulate
onto a sample specimen holder-a grid or other small opening spanned with a very thin,
structureless, and electron-transparent substrate.
Specimen grids come in a variety of metal compositions and with a multitude of mesh
openings. For particulate dispersions, the Athene-type, 3-mm-diameter, copper grid having
small openings with thin, well-defined bars is adequate for most purposes. A 200-300
mesh grid will suffice. Grids having larger openings, slots, or single holes are better suited
for ultrathin-section or replica supports, but are less desirable for particulate dispersions
because of the greater potential for specimen drift and substrate damage in the electron
microscope.
Each grid must be covered with a support substrate. There are several choices, but
one that is commonly used is a thin film of collodion (nitrocellulose) covered with a thin
layer of evaporated carbon.
Most textbooks describe several techniques for preparing these thin films. In my opin-
ion, one of the easiest in producing uniformly thin films is the spreading drop-gravity
method. In this method a clean stock solution of about 1% (by weight) Parlodion in amyl
acetate is first prepared. A large glass crystallizing dish is filled with distilled water. In
the center of this dish a piece of copper (or other nonrusting) screening is placed on a
mounting block so that it rests just below the surface of the water. The copper mesh grids
are then carefully placed, one at a time, on the screening below the surface of the water.
When all the grids are so placed, more water is slowly added to the dish until a meniscus
is formed. The meniscus is "scraped" with a clean, glass stirring rod to remove dust par-
ticles. A single, small drop of the stock Parlodion solution is then applied to the clean
surface of the water and the amyl acetate allowed to evaporate, leaving the thin film of
plastic shimmering on the water surface. This should take only a few seconds. The film
is then allowed to settle onto the grids as the water is siphoned out of the crystallizing
dish. (In some laboratories, special glass dishes have been constructed with a nipple open-
ing at their base to allow the water to be removed without siphoning.) After the water has
been siphoned out, the screen containing the coated grids is then carefully removed and
allowed to air-dry in a dust-free place.
The air-dried grids may be used as is for specimen support, but it is highly desirable
to add a thin coating of evaporated carbon to their surface to strengthen them and to help
avoid charging effects in the microscope. A thin carbon film can be applied in a high-
vacuum evaporator which is standard equipment in most TEM laboratories. The desired
amount of carbon can be judged visually during its application using a simple technique.
Before the bell jar of the vacuum evaporator is pumped out, a small square of white, glazed
ceramic tile is placed next to the grids. In the center of this tile is placed a drop of high-
vacuum pump oil. When the carbon is evaporated, it will stick to the tile except beneath
the oil drop and the color difference will allow the investigator to control the relative
thickness of the carbon being deposited. A light gray color is adequate.
Carbon films, and to a lesser extent plastic films, are hydrophobic. This unfortunate
characteristic makes the uniform application of dispersions from aqueous suspensions
difficult. Therefore, some method of preparing the carbon and/or plastic support to receive
the specimen is required. The importance of this step cannot be overemphasized in the
study of particulate materials by TEM. The time spent in rendering grid surfaces hydro-
philic will pay handsome dividends!
Although the textbooks describe numerous methods for avoiding the beading effect
that a hydrophobic surface causes, one seldom-described method stands out as the cleanest
and most reliable: ion-cleaning through glow discharge. In this simple technique, the sur-
face of the grids, regardless of their composition, is bathed in a glow discharge atmosphere.
The process is best accomplished immediately after the carbon-coating step. If the vacuum
evaporator is equipped with an "ion-cleaning" transformer, it is turned on after closing
off the diffusion pump, after a small amount of air has been admitted to the bell jar, and
the pumping cycle "roughed" in a pre pump position. In the absence of a built-in trans-
former, a hand-operated Tesla high-frequency induction coil can be applied to an external
lead into the bell jar and will produce similar results. The bluish glow discharge will vary
with the degree of vacuum. Like an ionization gauge, it will be poor at either too low or
too high a vacuum. Trial and error will provide the investigator with the appropriate levels
to produce a good bathing effect for the grids. The grids should be given this treatment
for 30-60 sec. They may now be removed from the apparatus and are ready for use. A
retreatment may be necessary if the grids are not used within about a week's time.
2.4. Applying the Specimen to the Substrate
Having prepared the specimen as an aqueous suspension, the next step is to place an
appropriate amount of the sample onto the grid surface for study in the electron microscope.
Some method of holding the grids must be devised. The use of forceps can be troublesome,
especially with magnetic materials. A better procedure consists of first applying a small
strip of double-sided cellulose tape to a clean glass microscope slide. Individual, coated,
and ionized grids are then touched just at the edge of the tape strip, sufficiently to hold
them in position and far enough apart so as not to interfere with one another. The grids
are then ready to accept the sample.
The specimen should be well dispersed at an appropriate concentration in distilled
water. An "appropriate" concentration is learned through trial and error, and is judged by
visually estimating the optical density of the solution in a 5-ml centrifuge tube with back-
lighting. If the suspension appears to be too light, the solids can be repelleted in the clinical
centrifuge. If the suspension appears too dark, it can be diluted.
Before applying the sample to the grid, the centrifuge tube containing the sample
should be placed briefly into the ultrasonic bath to be sure that the mineral grains are
disaggregated. This can be especially important with magnetic materials that tend to reag-
gregate and form linear chains.
With a clean, disposable glass pipette, a small drop of the dispersion is applied to the
ionized surface of a grid. If the grid surface is properly hydrophilic, a small drop will have
a low, shallow meniscus profile. If the surface is hydrophobic, the drop will stand high.
If the latter occurs, it is generally inadvisable to proceed further because the sample will
not dry down uniformly on the grid surface. The particulates will tend to move to the
edges of the grid or occur in dense clumps. Rebathe the coated grids in the glow discharge
atmosphere (Section 2.3). If the drop assumes the low profile of the hydrophilic surface,
172 Chapter 6
the excess water can be allowed to air-dry, or with slightly concentrated suspensions it
can be carefully removed by placing a pointed piece of clean filter paper at the edge of
the grid. The water should seem to flatten out and uniformly disappear from a hydrophilic
surface, leaving behind the well-dispersed particulates. When thoroughly dry, the grid is
ready for electron microscopy. If the water does not dry down uniformly but appears to
be quickly and completely drawn into the filter paper, it is generally an indication that
the grid surface is too hydrophobic. Do not proceed further because few particles will be
retained on the grid surface and those that do will be clumped and poorly distributed. It
is best to rebathe the grids in the glow discharge atmosphere and begin again.
An alternate method of applying the specimen to the grids is sometimes useful, es-
pecially where aggregation of magnetic particles is to be avoided. (It is worth noting that
if magnetic grains reaggregate very rapidly, they are either too concentrated or too large
to be of immediate use in the TEM; further sample preparation is indicated.) This technique
involves placing the specimen suspension into a glass nebulizer and immediately following
ultrasonic dispersion, spraying a fine mist of the suspension onto the grids where the
microdroplets rapidly dry. Glass nebulizers for this purpose can be obtained from com-
mercial electron microscopy supply houses.
2.5. Shadowcasting and Surface Replication
Although the TEM allows for very high resolution, it suffers in that it cannot normally
provide an easily interpreted three-dimensional view of the specimen. To a very great
extent, the modern scanning electron microscope (SEM) has filled in this gap. It has allowed
for the determination of three-dimensional views of most objects, and it is especially useful
at low and intermediate magnifications. From time to time, however, very small particles
of colloidal dimensions can stretch the capability of the SEM to its limits. In such cases,
where a higher resolution and definition is required, the TEM can be useful. Here, a three-
dimensional impression is created through a technique known as shadowing or shad-
owcasting.
The method is well described in most general texts on electron microscopy and, simply
put, it involves the vacuum evaporation of a heavy metal (gold, platinum, chromium) from
a point source at an angle to the specimen already dried on the grid surface. The metal
piles ups on one surface creating a shadow behind the other, thus giving the impression
of depth when observed in the electron microscope. If the angle of shadowing is known,
the height of the shadowed material can be trigonometrically calculated. Where more ac-
curate dimensions are required, an internal calibration may be used to confirm the shad-
owing angle. Adding a few calibrating polystyrene latex spheres to the particulate sus-
pension will provide particles of known height from which the shadowing angle can be
back-calculated. Such spheres can also be obtained commercially.
For crisp shadows and a fine-grained metal coating, it is imperative that a high vacuum
be obtained before metal evaporation takes place.
The shadowcasting technique can be especially useful in helping to define the shapes
and dimensions of very thin flakes and fibers, both of which can be shadowed at very low
angles to enhance their visibility. Extremely small particles, below about 100 A, can present
difficulties associated with artifacts created by metal graininess, shadow softness, and
piling up of metal. The results of such experiments should be carefully evaluated.
With the advent of new higher-resolution SEMs, the time-consuming and difficult
techniques of single-stage surface replication to reveal surface topography have been ren-
dered all but obsolete. Although the method can still provide sharper, crisper photomi-
crographs at higher magnifications, its usefulness is substantially compromised and di-
minished by the increased work involved. In addition, the sample is destroyed. Suffice it
to say that where surface topographical information is the primary quest, the SEM is pre-
ferred. One exception is worth mentioning: freeze-fracture and freeze-etching of biological
tissue can perhaps be better dealt with using replica methods. The interested reader should
refer to papers on this specialized technique for details.
2.6. Tissue Preparation Methods
The investigator concerned with the TEM examination of in situ, tissue-bound ma-
terials of biogenic origin will be faced with the prospect of having an extensive laboratory
protocol of tissue preparation to deal with. It is beyond the scope of this chapter to provide
even a cursory discussion of the myriad of techniques involved in tissue handling, fixation,
embedding, sectioning, staining, and mounting. The reader intent on this approach is best
served by reading some of the more recent textbook accounts of these various procedures
and obtaining counsel from those experienced in such work.
On the other hand, a few words of warning may be useful under this heading. The
"normal" techniques involved in preparing most biological tissue samples for TEM were
often carefully and specifically designed for biological-biochemical purposes alien to the
study of mineral materials. The investigator primarily concerned with the localization and
identification of biogenic magnetic (or nonmagnetic) mineral materials should, at least at
the early stages of a TEM investigation, place this goal at a higher priority than the fidelity
with which cell tissues are preserved. After the minerals have been found and identified,
the procedures can be modified to deal with "good" fixation of the tissues themselves. It
is well worthwhile, therefore, to examine each of the "cookbook" recipes to be sure that
the procedures and the chemicals involved will be compatible with the goals of the study.
Some of the questions worth asking in the beginning might include:
1. Will the fixation chemistry (fixatives, rinses, buffers, solvents) significantly alter
the size, shape, or even the bulk mineralogy of the minerals known or suspected
to be present? This is especially important with minerals such as calcite or ara-
gonite. It is less likely to affect magnetic mineralogies.
2. Might the minerals (inherently electron-dense) be more readily observed, initially,
if heavy metal stains or fixatives were minimized, or even avoided? There is no
point in adding a stain to locate a mineral that can be readily seen anyhow.
3. Are the time-temperature procedures of embedding likely to cause mineralogical
changes? This is a real possibility for metastable hydrous ferric oxides and hy-
droxides.
4. Is the pH of the liquid in the microtome sectioning trough critical? Could any of
the minerals be dissolved (or oxidized) as the sections float prior to being placed
on grids?
3. Studying the Sample in the Microscope

Once the sample has been suitably prepared, it is ready for study in the electron
microscope. The purpose of this section is to review some of the general operating pro-
cedures that may be of value in the study of mineral materials. Many of these procedures
will be very well known to some people, and new to others, but all will require (and assume)
some basic familiarization with a TEM.
174 Chapter 6
3.1. Preparing the Microscope for Use
Unlike much of the early equipment, the modern TEM is generally reliable and is easy
to operate with minimal effort required to obtain good-quality micrographs. The improve-
ments in both electronic and mechanical stability have brought very good resolution within
the ready grasp of even a novice operator. Nevertheless, there are a few factors that require
caution in the study of most specimens, and some for mineral specimens in particular.
Before studying a specimen grid in any detail, it is worthwhile to establish that the
equipment is reasonably clean and well aligned. For all but the most exacting, high-res-
olution work, this can be accomplished with most microscopes rather readily, and the few
minutes spent in the beginning will usually be worthwhile.
After turning on the electron beam, both the filament alignment and the condenser
lens aperture alignment can be checked by bringing the condenser lens to a focus "cross-
over" position and adjusting the filament current slightly below its saturation. An aligned
filament will provide a symmetrical image of the filament tip and an aligned condenser
aperture will maintain the beam in a central position without significant sweep when
passing back and forth through cross-over. These two checks will save time and effort in
subsequent microscopy, especially when the illumination needs "fine tuning" for pho-
tographic purpOSeS. A misaligned filament or condenser aperture can easily result in poor
or uneven illumination.
The general alignment of the remaining lenses is adequate when an image holds its
central position on the viewing screen while the magnification and focus are varied. (In
many instruments, these alignments are fixed and require little or no adjustment.)
The objective lens aperture should be checked for its centering and for its influence
on image quality. Centering can be easily checked (with a specimen in place) by lowering
the magnification to zero, i.e., placing the microscope into its diffraction mode with only
a point at the center of the screen. The general background scattering of electrons produced
by the sample substrate will form an image of the objective aperture around the central,
focused diffraction point. The image of the aperture should be circular, free of "dirt," and
symmetrical with respect to the diffraction point.
The cleanliness of the objective aperture, and the general area around it, should be
established by checking for astigmatism and resolution. There are many ways to do this,
but one of the simplest is to examine a specimen at a magnification high enough to see
the very fine "salt-and-pepper" background generally observable on any carbon-coated
substrate. The very small grains of evaporated carbon should appear crisp and evenly
symmetrical with small changes in focus. They should appear to get smaller and lose
contrast as true focus is approached. if they appear to have any elongated or aligned ap-
pearance to them, the image is astigmatic and this should be corrected, especially if high-
magnification photomicrographs are anticipated. If the grains are not crisp and sharp as
focus is approached, the aperture and/or sample holder may be contaminated, or an elec-
tronic instability may be suspected. This so-called "soft" image will be emphasized on
photos and the condition should be eliminated. High resolution will be difficult or im-
possible to obtain otherwise.
These preliminary checks of the equipment can be made on the specimen to be studied,
or they can be done on specially prepared test grids. The latter should be used if the
microscope requires anything but a quick "trim" alignment. Continued use of the study
specimen may result in beam damage and will cause contamination.
3.2. Working with the Microscope

The study and photography of most average specimens in an electron microscope is
relatively straightforward unless very exacting high-resolution work is required. With par-
ticulate mineral samples that have been well prepared and reasonably well dispersed on
the grid, all that is required for routine work is the selection of several random fields for
photography at one or more magnifications. The focus should be carefully checked and
the specimen monitored for drift-a slow, directional movement due to substrate insta-
bility or charging effects. Except for high-resolution purposes, a slightly underfocused
image is best and, in practice, is usually the easiest to obtain. If drift is observed, no photos
should be attempted as they will invariably be poor.
The image formed by a TEM on the fluorescent screen is, as the name implies, from
electrons transmitted through the specimen. Therefore, variation in contrast in the spec-
imen can be the result of several causes. Differences in chemical composition may produce
differences simply as a result of mass absorption effects; heavier atoms absorbing more
electrons than lighter ones. Differences in sample thickness or crystallinity are also im-
portant. A grain may thus appear opaque because it is too thick or because it is made up
of electron-dense material, or because it is specially oriented crystallographic ally.
Crystalline materials commonly present varied images with either change in orien-
tation or focus. These marked changes in contrast are the result of diffraction by the crystal
structure, compounded by any differences in the thickness that the grain may have. A
small change in the Bragg angle can alter the image considerably and the biologist should
be warned against this when adjusting focus. A through-focus series may be helpful in
interpreting such images. Removal of the objective aperture will intensify these effects as
more diffracted beams contribute to the overall image. With noncrystalline materials, these
phenomena are lacking and, indeed, the presence or absence of diffraction contours and
the like may be useful for establishing crystallinity in unknown solids.
Most minerals, especially iron minerals, absorb electrons strongly as compared with
unstained biological tissues. This property may be used to advantage by the microscopist
in searching through a series of tissue ultrathin sections for biogenic mineral deposits.
Once the location of the mineralized regions has been established, new sections can be
poststained to bring out the biological structures. New sections are required because sec-
tions that have been studied in the TEM seldom accept stains well thereafter.
Photography of dispersed particulate grains can present problems with exposure. Au-
tomatic exposure meters can give false readings resulting in grains that are poorly exposed
with respect to the background. The investigator should decide which is more important,
outline shape or internal structure. A series of different exposures will ensure that most
grains are properly exposed, at least on one of the photos. With varied sizes and thicknesses
of particulates in a given field of view, a good exposure for all grains is not possible.
In working with the electron microscope, there are some structures that may be ob-
served from time to time that bear mention because of their possible artifactual nature.
Two that are especially important when studying particulates are fringes and inclusions.
Light or dark fringes at the edge of an object can be produced in several ways. When
they are more or less symmetrical around a grain, they can be focus artifacts due to strong
under- or overfocus. They can also be the result of a general contamination due to long
exposure to the electron beam. Or they can result from contamination due to drying down
of soluble organic matter, salts, buffers, or stains. When asymmetrical, they can be due to
a strong objective lens astigmatism. If they appear in photos, they can be due to specimen
drift or an accelerating voltage change during photography. And if the sample has been
shadowcast too heavily, a coating and heavy metal on the "sunny side" of a grain will
appear as a fringe.
Therefore, before interpreting any edge or fringe as a real structure, the possibility of
artifact should first be removed. If all, or most of the grains in the field of view have similar
fringes, or the fringes are all aligned in the same direction, the odds are good that the
structure is artifactual. Check for instrumental problems by changing focus and/or astig-
matism settings. Check for beam contamination by moving to a fresh area of the grid and
176 Chapter 6
making a rapid search for the fringe structures. Another check for beam contamination
consists of lowering the magnification and looking for a general darkening in the area
studied.
Chemical contamination resulting from inadequate washing may occur over the entire
grid, or it may be local. In either case, the observer will get the impression, especially at
lower magnifications, of looking through a foggy window. Sharp focus will also be difficult
to obtain. At higher magnifications, soluble salts that have crystallized out on the grid will
tend to sublime in the beam and seem to disappear completely, or lose contrast except for
the frailest outline.
"Inclusions" present another potential source of confusion. Because the TEM image
is like a shadowgraph, small particulates appearing to be within larger ones can be: (1)
surface pits, (2) surface coatings, or (3) true inclusions. True inclusions may be either solids
or voids. If the inclusions are solid, they will appear darker on the fluorescent screen than
the surrounding material; if voids, lighter than the surroundings. Surface pits will be
lighter; adherent coatings, darker. In order to distinguish surface from internal structures,
a side view may help or a surface replica may be necessary.
With metal-shadowed materials, it is useful to remember that on normal photographic
prints, the "shadows" will appear to be light. Only on a negative, or a negative print, will
they be dark.
3.3. Identification of Particulates by Electron Diffraction
In interpretation, application, and implementation, no aspect of TEM is as fraught

with difficulty as is electron diffraction. The identification of completely unknown par-
ticulates is especially hazardous and the reader is warned against high expectations for
the use of this technique. These difficulties are likely to be greater where small biogenic
magnetic particles of limited quantity are being encountered.
This section is not intended to be even an introduction to the theory and practice of
electron diffraction. It is intended to provide the reader with suggestions of a practical
nature for use with the types of samples likely to be encountered in the study of biomin-
eralization.
In a general way, the study of particulates by electron diffraction is similar to that of
X ray diffraction using the powder method. Instead of using a collimated beam of X rays
of fixed wavelength, a focused beam of electrons of fixed, but shorter wavelength is passed
through the specimen and is diffracted by the crystalline structure according to Bragg's
law. Unlike X rays, electrons can be diverted by magnetic fields and thus the image of the
resultant diffraction pattern can be focused, viewed on a fluorescent screen, and quickly
recorded on photographic film. The "trick" with the identification of particulate specimens
is to obtain the sharpest, most complete diffraction pattern possible, with the best cali-
bration possible.
Some samples, i.e., colloids, are ideally suited for electron diffraction. The ideal spec-
imen is one that is present in small «0.1 Ilm), thin crystals that are numerous, randomly
oriented, and closely spaced, but not clumped. Few natural samples meet all of these
criteria. If a sample must be crushed or ground to a very fine particle size for electron
diffraction, it probably occurs in sufficient quantity to be identified more easily by Debye-
Scherrer X ray methods. On the other hand, if it occurs in very limited quantity and the
crystals are large, or thick, it will not be suitable for electron diffraction because the electron
beam cannot pass through the sample.
Biogenic crystals present in ultrathin sections may be thin enough to transmit dif-
fracted beams, but not numerous enough to provide a pattern complete enough for iden-
tification. And it is seldom that such patterns can be accurately calibrated.
Although electron diffraction has its drawbacks, the reader should not be disheartened.
Begin with the concept that a diffraction pattern from a single crystal is made up of spots
that are related to the atomic structure of the material. The pattern formed by the spots
can be modified by the orientation of the crystal. Additional crystals will add to the number
of spots. The more crystals contributing to the pattern, the more spots, and the more the
spots will appear to form concentric rings with the undiffracted beam at the center. If the
number of crystals becomes very large, the spots will begin to lose their identity and blend
together with one another as uniform rings. The distance of the rings from the central beam
(and from one another) is related to the interplanar spacing of the crystal structure. Each
material has its own distinct pattern with its own "fingerprint" of spacings. These char-
acteristic d-spacings, and their relative intensities can be found in tables and books de-
signed for the purposes of identification. Identification, therefore, consists of matching the
d-spacings and the intensities of the pattern(s) with those in the tables.
To obtain an accurate measurement of the d-spacings, the pattern must be calibrated.
There are several methods that can be used. A rough instrument calibration can be obtained
by preparing a known sample grid and taking a photo of its diffraction pattern. From the
film (or plate), the diameter of each diffraction ring is measured, center-to-center through
the central beam. The known d-spacings in angstroms for each of the stronger reflections
are then multiplied by the diameters of the corresponding rings. The larger d-spacings will
be closer in to the central beam, the smaller further out. The number thus obtained for
each ring should be the same and will serve as a so-called camera constant for use with
unknown samples. In practice, the number obtained will vary slightly with each ring, and
an average value of the constant should be calculated.
Although the word "constant" is used, there are a number of things that can alter its
value. A change in accelerating voltage will cause a corresponding change in the wave-
length of the beam of electrons, and hence in the diameter of the rings. A change in the
height of the specimen grid will also cause a change in the ring diameter. And, not un-
commonly, the rings may be slightly oval rather than absolutely circular, so that care should
be used in this respect as well. Clearly, the use of such a predetermined camera constant
can be hazardous and may be very misleading for the identification of unknowns.
The most accurate work demands an internal standard, where the known and the
unknown can be compared with one another quite independently of the variables asso-
ciated with specimen height or accelerating voltage. One of the easiest and most reliable
methods involves the use of a vacuum-evaporated metal such as gold or aluminum. Here,
the specimen support film is given a moderate coating of the metal upon which the un-
known sample is then placed. Diffraction from this metal coating provides rings simul-
taneously with the rings (or spots) of the unknown.
A few hints involved with this technique deserve mention. As applied to particulate
dispersions, coating only part of each grid will permit diffraction patterns to be made both
with and without the internal standard. This avoids the necessity of changing or preparing
additional grids. This can be accomplished rather simply by lining up several Parlodion-
covered grids at the edge of double-sided cellulose tape (as earlier described) and then
covering half of them with a strip of any thin material so as to prevent that half from
receiving the evaporated metal, leaving the other half exposed.
For purposes of electron diffraction, the crystallite size of the evaporated metal should
not be small. Although a high vacuum inhibits grain growth of the metal and produces a
very fine background image, the diffraction rings will be broad rather than sharp. Therefore,
in contrast to shadowcasting (Section 2.5), the vacuum in the evaporator should not be
allowed to reach a high value (10- 3 mm mercury is adequate). And, of course, the surface
should be rendered hydrophilic (Section 2.3) before adding the unknown particulate.
Obtaining a good diffraction pattern suitable for identification purposes can be dif-
ficult. The microscopist should first try to locate an area on the grid where the particulate
178 Chapter 6
grains are numerous, but not badly clumped. This can be especially difficult with magnetic
materials, and if clumping is severe, one of the nebulizer techniques for sample preparation
(Section 2.4) should be tried. If the nebulizer technique does not work well, the suspension
is probably too concentrated, or too many large grains are present, or both.
After locating an area that looks promising, and with the objective lens aperture re-
moved, the instrument should be placed into the diffraction mode (Le., the magnification
reduced to zero) and the diffraction point sharply focused. If the sample is diffracting well,
a pattern should be clearly observed on the fluorescent screen. If the instrument is equipped
with a central beam stop, it may be inserted at this time to cut down on the glare from
the undiffracted beam.
Before making a photograph of the pattern, two additional adjustments can be valuable
in reducing background scattering and in producing a sharper pattern. A clean, selected-
area limiting aperture should be inserted and adjusted to surround the smallest area of
particles consistent with a complete diffraction pattern. This adjustment will vary con-
siderably with the specimen, and may not be feasible in all instances.
A second adjustment that will be useful is the defocusing of the condenser lens. This
will have the effect of darkening the screen, even to the point where the diffraction pattern
may not be visible to the observer on the screen. But the increase in pattern sharpness is
dramatic, and worth the effort. In fact, even a good diffraction pattern that is clearly seen
on the fluorescent screen will invariably be improved by this procedure; the results be-
coming apparent on a properly exposed and developed film.
The obtaining and photographing of electron diffraction patterns is a matter of varying
the several adjustments to suit the individual specimen. Apart from a focused central beam,
there are no "fixed" settings, and the "best" patterns are seldom the brightest on the screen.
The microscopist must be prepared to make constant adjustments-defocus the condenser;
refocus the diffraction spot.
The pattern will be in focus when the central beam is sharply focused to a small spot.
Several photographic exposures will be necessary to ensure that the pattern has been com-
pletely captured on film. The correct exposure is usually a trial-and-error situation and
only experience can be a guide. The diffraction pattern made with an intense beam near
cross-over can be photographed very quickly, but it suffers from high background scattering
and broadened rings or spots. In addition, an intense beam can be detrimental to the sample
because of charging effects and, indeed, some minerals can be altered significantly in this
manner through beam-induced heating.
4. Analysis of Electron Diffraction Powder Patterns
The identification of particulate materials by electron diffraction is similar in many

ways to that of X-ray diffraction using the Debye-Scherrer powder method. Basically, the
diffraction ring diameters are measured on the film, converted to d-spacings in angstrom
units, and the relative intensities of the lines noted. These data are then compared with
existing tables of d-spacings/intensities for various minerals until a match has been found.
This ideal procedure is seldom easily realized with electron diffraction because of either
a sample preferred orientation or a lack of sufficient grains to produce a suitably complete
pattern.
Preferred orientation may completely eliminate some of the stronger reflections that
are diagnostic for identification from the tables. Flake-shaped clay minerals are a classic
example of this problem because the most diagnostic d-spacings are the basal reflections
which will not appear when the electron beam is parallel to the crystal c-axis. In the case
of an isometric compound, such as magnetite, crystallographic preferred orientation is not
a problem for purposes of identification.
rEM and Electron Diffraction 179
More commonly, an incomplete pattern of spots is encountered, especially where only

a few grains give rise to the pattern. A confident identification can be difficult, if not
impossible. One approach may yield limited results. The film or plate is placed into a
photographic enlarger and an enlarged image of the pattern projected onto a white piece
of paper. The central beam location is carefully marked on the paper along with the po-
sitions of as many of the spots as possible. The radius of each of the spots is measured
and those with similar radii grouped together. If an internal standard has been used, the
d-spacings can be calculated using one of the standard rings as a point of reference. If no
internal standard was used, which is the usual situation, the previously determined camera
constant of the instrument can be used. Before calculating the ratios, this constant is first
recalculated for radii rather than diameters, and for the amount of projector enlargement.
The results of this procedure should be used with great caution, especially with totally
unknown materials. Stray reflections from other grains, so-called "forbidden" reflections,
and the presence of another phase can all complicate the situation considerably. Never-
theless, it can be used successfully to help confirm an educated guess about what the
mineral might be, or to rule out certain minerals.
A photographic enlargement of the pattern can be used if the rings are reasonably
complete and if all measurements are made in the same direction. It should not be used,
however, where incomplete groups of spots are all that exist, because the photographic
paper when dried will shrink or stretch differentially, making the locations of spots from
the central beam subject to distortion errors.
From time to time, good electron diffraction patterns will be obtained that resist iden-
tification, or produce an apparently spurious list of calculated d-spacings. This can be
especially annoying when the investigator is reasonably confident that one or another
mineral must be present in the sample. The situation is usually encountered when no
internal standard was used and the conditions under which the sample was prepared
differed in some way from that under which the camera constant "standard" was prepared.
These puzzling patterns can sometimes be reconciled in a trial-and-error fashion by making
two assumptions: (1) the camera constant did change and (2) the mineral producing the
pattern did have the suspected identity. At this point, the ring diameters of the most intense
lines on the pattern are multiplied by the d-spacing of the most intense line known for
the suspected mineral. If one of the resulting numbers is close to that of the original camera
constant, it is chosen as the new camera constant and the remaining d-spacings are re-
calculated. The recalculated spacings must all fit the suspected mineral if the two as-
sumptions were correct.
This procedure can even be used to identify electron diffraction patterns where no
camera constant at all is known, i.e., from photos in the literature. In this case, the ring
diameters are measured, a given mineral is assumed, a camera constant located, and the
resulting d-spacings used to confirm or deny the suspected mineral's identity.
5. Conclusions
The preceding paragraphs have presented some of the techniques that have worked
well for the author. As with most techniques, a modification here and there to suit the
particular problem is often desirable. Any laboratory equipped with a TEM will have avail-
able one or more of the standard reference works or textbooks on electron microscopy.
These works and the laboratory staff should be consulted for alternate approaches or more
details. The characterization of particulate materials by TEM involves careful sample prep-
aration and use of the equipment, both as a microscope and as a diffraction camera. There-
fore, before following this or any other guide for TEM, the reader should examine the overall
180 Chapter 6
scientific problem and decide at the very beginning what it is specifically that he or she
needs to learn about the sample. It is usually inadvisable to rush into the examination of
a specimen with the TEM simply to see what it looks like; a plan should be made and
priorities established. In this way, the experimental protocol for specimen preparation and
study can be designed to suit both the sample and the problem.
Selected References
Agar, A. W., Alderson, R H., and Chescoe, D., 1974, Principles and practice of electron microscope
operation, in: Practical Methods in Electron Microscopy, Volume 2 (A. M. Glauert, ed.), American
Elsevier, New York.
Beeston, B. E. P., Horne, R W., and Markham, R, 1972, Electron diffraction and optical diffraction
techniques, in: Practical Methods in Electron Microscopy, Volume 1 (A. M. Glauert, ed.), American
Elsevier, New York.
Breese, S. S. (ed.), 1962, Electron Microscopy, Academic Press, New York.
Chandler, J. A., 1977, X-ray microanalysis in the electron microscope, in: Practical Methods in Electron
Microscopy, Volume 5 (A. M. Glauert, ed.), American Elsevier, New York.
Dawes, C. J., 1971, Biological Techniques in Electron Microscopy, Barnes & Noble, New York.
Fischer, R B., 1954, Applied Electron Microscopy, Indiana University Press, Bloomington.
Gabriel, B. 1., 1982, Biological Electron Microscopy, Van Nostrand-Reinhold, Princeton, N. J.
Glauert, A. M., 1975, Fixation, dehydration and embedding of biological specimens, in: Practical
Methods in Electron Microscopy, Volume 3 (A. M. Glauert, ed.), American Elsevier, New York.
Gray, P., 1973, The Encyclopedia of Microscopy and Microtechnique, Van Nostrand-Reinhold, Prince-
ton, N. J.
Grimstone, A. V., 1977, The Electron Microscope in Biology, 2nd. ed., Arnold, London.
Haggis, G. H., 1966, The Electron Microscope in Molecular Biology, Longmans, London.
Hall, C. E., Introduction to Electron Microscopy, 2nd ed., McGraw-Hill, New York.
Hawkes, P. W., 1972, Electron Optics and Electron Microscopy, Taylor & Francis, London.
Hayat, M. A., 1972, Basic Electron Microscopy Techniques, Van Nostrand-Reinhold, Princeton, N.J.
Hayat, M. A., 1975, Positive Staining for Electron Microscopy, Van Nostrand-Reinhold, Princeton,
N.J.
Hayat, M. A., 1981, Principles and Techniques of Electron Microscopy: Biological Applications, 2nd
ed., University Park Press, Baltimore.
Hirsch, P. B., Howie, A., Nicholson, R B., Pashley, D. W., and Whelan, M. J., 1965, Electron Microscopy
of Thin Crystals, Butterworths, London.
Hren, J. J., Goldstein, J. 1., and Joy, D. C. (eds.), 1979, Introduction to Analytical Electron Microscopy,
Plenum Press, New York.
Kay, D. H., Techniques for Electron Microscopy, 2nd ed., Blackwell, Oxford.
Meek, G. A., 1976, Practical Electron Microscopy for Biologists, 2nd ed., Wiley-Interscience, New
York.
Mercer, E. H., and Birbeck, M. S. C., 1972, Electron Microscopy: A Handbook for Biologists, 3rd ed.,
Blackwell, Oxford.
Mokotoff, G. F., 1978, Electron Microscopy Laboratory Techniques: A Workbook, Library Research
Associates, Monroe, N.Y.
Nunn, R E., 1970, Electron Microscopy: Preparation of Biological Specimens, Butterworths, London.
Pease, D. C., 1964, Histological Techniques for Electron Microscopy, 2nd ed., Academic Press, New
York.
Pinsker, Z. G., 1953, Electron Diffraction, Butterworths, London.
Racker, D. 1., 1983, Transmission Electron Microscopy: Methods of Application, Thomas, Springfield,
Ill.
Reid, N., 1975, Ultramicrotomy, in: Practical Methods in Electron Microscopy, Volume 3 (A. M.
Glauert, ed.J, American Elsevier, New York.
Siegel, B. M., 1964, Modern Developments in Electron Microscopy, Academic Press, New York.
Siegel, B. M., and Beaman, D. R (eds.), 1975, Physical Aspects of Electron Microscopy and Microbeam
Analysis, Wiley, New York.
Sjostrand, F., 1968, Electron Microscopy of Cells and Tissues, Academic Press, New York.
Thomas, G., 1964, Transmission Electron Microscopy of Metals, Wiley, New York.
Venables, J. A. (ed.), 1976, Developments in Electron Microscopy and Analysis, Academic Press, New
York.
Weakley, B. S., 1981, A Beginner's Handbook in Biological Tranmission Electron Microscopy, 2nd
ed., Churchill/Livingstone, London.
Wenk, H.-R. (ed.), 1976, Electron Microscopy in Mineralogy, Springer-Verlag, Berlin.
Wi schnitzer, S., 1981, Introduction to Electron Microscopy, 3rd ed., Pergamon Press, Elmsford, N.Y.
Chapter 7
The Cellular Localization of Particulate
Iron
BENJAMIN WALCOTT
1. Introduction . . . . . . . . . . . . . . . . . . . . . . ......... 183

2. Anatomical Techniques .......... . ......... 184
2.1. Light Microscopy .... . 184
2.2. Electron Microscopy ...... . ....... . 186
3. An Example: The Bumblebee . . . . . ....... . 188
3.1. Introduction ..... . ............ . 188
3.2. Light Microscopy ........ . ................. . 188
3.3. Electron Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
3.4. Analytical Electron Microscopy 191
4. Conclusions 193
References . . . . . . . . . . . . . . . . 195
1. Introduction
There is considerable behavioral evidence which suggests that a number of diverse species
of animals (homing pigeons, warblers, salmon, salamanders, honeybees) are able to detect
weak earth-strength magnetic fields and that they may be able to use this information for
orientation. While other navigational clues have been proposed, e.g., olfaction (see Papi,
1982), the data suggest that some animals can sense magnetic fields. The identity and
character of the sensory system that is involved in this "new" sense, however, is unknown.
The discovery of magnetotactic bacteria by Blakemore (1975) and his colleagues suggested
that an iron oxide, specifically magnetite, could be involved in the sensory transduction
process. The observation that magnetite is also found in the abdomen of the honeybee
(Gould et aI., 1978) and the head of the homing pigeon (Walcott et aI., 1979) gave impetus
to the notion that magnetite could be of general importance in magnetic field sensitivity.
Thus, in principle, if one could localize magnetite in a cellular system that satisfied certain
requirements for a sensory system, one might be able to localize the sensory system re-
sponsible for magnetic field detection.
All sensory systems consist of receptor cells, often but not always neurons, that connect
to the central nervous system. In many systems, specialized molecules in the membrane
of the receptor cells are involved in the transduction process. In the visual systems of
animals, for example, a visual pigment absorbs the light quanta and that process alters the
permeability of the visual cell membrane to ions, usually sodium, which in turn alters the
membrane potential of the receptor, thus converting light energy to an electrical response
BENJAMIN WALCOTT • Department of Anatomical Sciences, School of Medicine, State University

of New York, Stony Brook, New York 11794.
183
184 Chapter 7
in the cell. This electrical response is conveyed to the central nervous system where it is
processed. In other sensory systems, however, the stimulus appears to act directly on the
membrane to alter ion permeability, for example in many mechanoreceptors. In all cases,
however, the stimulus induces a change in membrane permeability and that process causes
an electrical response. ,
The working hypothesis in the search for a magnetic field receptor, therefore, is that
an iron oxide, possibly magnetite, may be a common element in the transduction process
in many animals. In order to identify and characterize the sensory system that is involved
in magnetic field reception, it is important to localize intracellular iron and to characterize
it. Unfortunately, techniques such as SQUID magnetometry provide little information about
the cellular localization of the magnetic or inducibly magnetic material in tissues. There-
fore, we have been examining tissues of bees and homing pigeons histologically to deter-
mine the presence of iron at the light and electron microscopic levels. It is hoped that this
approach will enable us to find out which tissue(s) can be identified as a likely candidate
for a sensory system and further will give us some idea as to the mechanism of sensory
transduction in this system.
The purpose of this chapter, therefore, is to describe some of the methods that we
have used and to point out some of the limitations of the various techniques. In Section
3, original data on the iron-containing cells of the bumblebee are presented to serve as an
illustrative example.
2. Anatomical Techniques
2.1. Light Microscopy
The localization of free iron and its oxides can be achieved by the use of the Prussian
blue reaction. In this procedure, acidic potassium ferrocyanide is allowed to react with
the tissue; when iron is present, a bright blue precipitate, ferric ferrocyanide, is formed.
We use this method for whole mounts or large pieces of tissue, for thick sections of tissue
embedded in paraffin or soft plastic, and for l-j.l.m sections of tissue embedded for electron
microscopy.
For the first examination, it is advantageous to look at as large a piece of tissue as
possible. Because the stain does not penetrate very far into the tissue and because the
density of the tissue prevents one from seeing deep structures, the technique of whole
mount staining is useful mainly for sheets of tissue or the surface of smaller pieces. In the
honeybee, for example, one can dissect open the entire abdomen and stain the contents.
The procedure is to expose the tissue by dissection and to pin it with stainless-steel "min-
uten" pins onto a wax sheet. The preparation can then be dissected as much as necessary
so that tissues of interest are exposed. The tissue is then lightly fixed with buffered par-
aformaldehyde (4% in 0.1 M phosphate buffer at pH 7.2) for 30 min at room temperature.
It is washed in buffer, freed from the pins attaching it to the wax, and submerged in a fresh
mixture of equal parts of 0.5% hydrochloric acid and 0.5% potassium ferrocyanide. These
stock solutions may be made in large quantities well in advance and can be kept for long
periods of time. Aliquots of the separate stock solutions should be heated to about 60 a C
before mixing and the staining should be done in an oven at that temperature. After 15
min, the tissue is removed from the stain and washed in buffer. If the staining is not
sufficiently intense, the staining procedure can be repeated with freshly mixed solution.
It is not advisable to use a mixed batch of stain for more than 15 min as it loses its ability
to stain and can induce precipitation artifacts. These artifacts appear as dense irregular
blue deposits on the surface of the sections and often obscure the underlying structures.
Cellular Localization of Particulate Iron 185
The stained piece of tissue can then be examined directly and photographed. Cells
that contain iron or iron oxides will appear blue and the other tissues will be unstained.
The tissue can then be dehydrated with a series of alcohol solutions, cleared in xylene,
and embedded for sectioning. This embedding procedure is conventional and is described
in detail in many technique books, such as Humason (1967).
Whole mount preparations provide information about what tissues contain iron but
very little about the state of the iron or the characteristics of the cells associated with the
iron. More particularly, this technique does not tell one if the iron is intracellular or ex-
tracellular. Therefore, the tissue must be sectioned and examined with the light micro-
scope. Stained whole mount preparations can be embedded and then sectioned. The
embedding medium can be paraffin or glycol-methacrylate OB-4 plastic, DuPontiSorval,
or Polysciences). As most of our tissues have a large range of hardness, from bone to nervous
tissue, we use the harder plastic rather than paraffin. In this way the tissue integrity is
maintained and bone is cut as well as the softer adjacent tissue. The embedding procedure
follows the manufacturer's recommendation except that with some batches of plastic, less
accelerator should be used than recommended, for otherwise the plastic sets too soon and
embedding is imperfect. Relatively large pieces of tissue can be sectioned, up to 1.0 x 1.5
cm, on glass knives that are freshly broken. These knives are made on either an LKB knife
maker or a DuPont/Sorval knife maker and are made from either i- or l-inch-thick glass.
Sections 3 fLm thick are cut on a JB-4 microtome and floated on a water bath to flatten
them. They are picked up on glass slides, dried, and then processed. If the tissue has already
been stained as a whole mount, the sections can be coverslipped and examined. If the
tissue is unstained, the following procedure is used. The sections are heated to 60°C, then
rehydrated by submersion in hot distilled water. This procedure is necessary because one
must use glass staining racks and these must be warmed before going into the hot stain or
else they crack. The slides are treated for 15 min with the freshly mixed hot acidic po-
tassium ferrocyanide, then washed with distilled water and counterstained with Ker-
nechtrot's stain. The latter stain consists of 5% nuclear fast red in distilled water and the
sections are treated with the stain at 60°C for 30 min. The slides are then washed with
distilled water, dehydrated with an alcohol series and then with xylene. The slides are
then coverslipped conventionally.
When examined with the light microscope, tissue in the stained sections appears red,
due to the counter stain, and iron deposits appear a very intense blue. If care is taken to
use double-distilled water and to clean all slides, staining dishes, etc. very carefully, little
surface staining artifact will be observed. As iron is such an abundant element and, in
vertebrates, is intimately involved in respiration, its presence in a particular tissue does
not necessarily mean that it is always present in that tissue. Because it is our belief that
iron particles-if involved in the sensory process-should always be present in the re-
ceptor system, we have set up criteria for what we consider to be positive staining. The
staining must be found in the same cells in 10 or more consecutive serial sections and in
the same cells in the same tissue from at least 5 different animals. A positive result using
this staining technique does not prove that one has found magnetite, but it does demon-
strate iron and its location in the tissue. Clearly, some protein-bound iron, such as in
hemoglobin or the cytochrome enzymes, never stains, for red blood cells in pigeons and
mitochondria in all tissues are not observed to stain. Magnetite particles do stain as well
as any forms of metallic iron alloys so that no iron-containing tools can be used in tissue
sectioning or staining. Further, we observe staining in pigeon spleen which contains iron
from the breakdown of hemoglobin. In the honeybee, we have shown that most of the iron
granules are a hydrous ferric oxide, such as ferritin, and they stain intensely (Kuterbach
et al., 1982). Therefore, the potassium ferrocyanide staining technique is a first step in the
characterization of iron-containing tissues. The technique show those tissues and cells that
186 Chapter 7
contain iron and one can see if the iron is diffusely distributed within cells or tissues or
if it is in some particulate form.
2.2. Electron Microscopy
2.2.1. General Methods

Further characterization of the iron-containing tissues can be accomplished by ultra-
structural examination. Thus, one can further classify the iron-containing cells, and de-
termine the intracellular location of the iron. However, there must always be the "bridge"
between the light microscope and the electron microscope. Therefore, we stain l-f-Lm sec-
tions of our tissue that has been fixed and embedded for electron microscopy with the
potassium ferrocyanide technique. Then, subsequent thin (BOOA) sections are examined
in the electron microscope.
The general method for preparation of tissue for electron microscopy is to fix with
2.5% glutaraldehyde in 0.1 M phosphate buffer at pH 7.4 for 1 hr at room temperature. In
some cases, for example the pigeon, primary fixation is accomplished by perfusion of the
whole animal with the fixative. In other cases, such as the bee, the tissue is exposed by
dissection and fixed by immersion. After the primary fixation, the tissue is dissected free
and washed in buffer for 30 min and then postfixed in 1% osmium tetroxide in the phos-
phate buffer for 1 hr. For analytical electron microscopy, the osmium treatment is elim-
inated. The tissue is then dehydrated with an acetone series and embedded in plastic,
either "Polybed" (Polysciences) or Epon B12. The latter plastic is no longer available but
is preferable for analytical microscopy. For routine microscopy, any of a number of com-
mercially available plastics are adequate.
Thick or thin sections are cut on an ultramicrotome using either glass or diamond
knives. If the tissue is particularly rich in iron granules, a diamond knife is preferable as
the glass knife will dull rapidly. The sections are mounted on plain copper grids, stained
for 20-30 min with saturated aqueous uranyl acetate, washed, and then stained for 1 min
with Reynold's lead citrate. The use of the heavy metal stains (osmium tetroxide, uranyl
acetate, and lead citrate) is necessary to observe cellular detail. For analytical microscopy,
thicker sections (1500A) are cut from the non-osmium-treated Epon-embedded material.
Some sections are then stained with the uranium and lead, again to reveal cellular detail,
but the rest, which are to be analyzed, are not stained at all.
In the conventional thin sections examined with the electron microscope, tissue detail
is easly resolved. The iron, when present, appears as electron-dense particles or deposits,
at least in those tissues we have examined. Often, these deposits are hard to differentiate
from other heavily staining material. However, the iron granules often tend to shatter and
to rip the section as it is cut. Also, the iron is usually the most electron-dense object within
the section. Definitive identification cannot be made on the basis of ultrastructural ob-
servation alone. Comparison of the ultrastructural images with the staining pattern seen
in l-f-Lm sections of the same material at the light microscopic level is helpful.
The embedding plastics necessary for electron microscopy are much harder than gly-
col-methacrylate plastics used for light microscopy and therefore staining the l-f-Lm EM
sections is difficult, as stain penetration is poor. The sections are cut on the ultramicrotome
and floated on glass slides. The slides are dried and the sections baked onto the slides by
heating the slide over an alcohol flame for 30 sec or so. The slide is then immersed in the
hot staining solution for 15 min, washed in distilled water, and then stained again. This
process can be repeated until sufficient staining intensity is achieved. Usually, two stain-
ings are sufficient. The sections can then be counterstained with Kernechtrot's stain if
necessary. The end result will be blue precipitate within the section. Again, it is very
important to examine a number of sections to make sure the staining pattern is consistent
and that it is not due to comtamination on the section. The staining pattern will always
be less intense than with other embedding materials due to the thinness of the sections
and the relative hardness of the plastic. However, one can clearly localize iron deposits
and can identify cells that do have intracellular iron. The block can then be trimmed to
those cells and thin sections to be cut for electron microscopy. At low magnification in
the electron microscope, one can compare those cells that have electron-dense particles
with those that show positive iron staining in the light microscope. This is important
because not all dense particles seen in the electron microscope are in fact iron containing.
2.2.2. Analytical Electron Microscopy

X-ray microanalysis of sections is necessary to unequivocally determine that electron-
dense particles contain iron. As described earlier, the sections should be thicker than
conventional ultrathin sections and should not be osmicated or stained with lead or ura-
nium. The sections can be mounted on standard copper grids, although carbon grids are
preferable, and the sections should be coated with spectroscopic-grade carbon before ex-
amination. The carbon will prevent charging and provide some thermal stability. The anal-
ysis is performed in a transmission electron microscope equipped with a nondispersive
detector and analytical computer analysis system. We use a JEOL 200e with a side entry
goniometer stage and a horizontal detector. As the instrument itself is largely made of iron,
it is imperative that it be set up to minimize electron and X-ray scatter. Thus, one must
use a hard X-ray aperture and a graphite specimen holder. The procedure is to localize a
cell with granules in the unstained section. This can be facilitated by inserting a diffraction
aperture which will increase the contrast. Once the appropriate cell is located and granules
identified, the magnification is increased and a particle localized in the center of the field.
The condenser lens is focused to include only the particle and the X-ray analysis is begun.
We normally count for 200 sec and aim to have a low dead time of 4-5% so that the
detector is not saturated. The resulting X-ray spectrum is stored on computer disks with
others made from other particles. Additionally, we always make analyses from areas of
cytoplasm adjacent to the granules. However, as there is relatively little material there, the
number of X-ray counts is relatively low and thus it is dangerous to compare this back-
ground with the particles. The X rays produced by the particle could produce further X
rays from the iron of the microscope column or other material in the sections. Therefore.
we also make analyses from contaminating pieces of dirt that are found on the sections.
This control will give a comparable X-ray count to the granules and unless there is iron
in the contamination, should not show any iron. Such analyses will demonstrate whether
or not the electron-dense particle does contain iron and other similar elements and will
give some measure of the amount, although truly quantitative measurements are very dif-
ficult to obtain. X-ray microanalysis, however, cannot identify the species of iron present.
Therefore, other analytical techniques have to be employed.
2.2.3. Other Analytical Techniques

The most obvious technique is that of electron diffraction. A beam of electrons is
focused on the particle in the section with the instrument conventionally set up in the
diffraction mode. If the iron is organized in a crystal, then a diffraction pattern should be
seen. One can examine fields of particles or individual particles, using a large spot or a
microbeam. From an analysis of the pattern, one can determine the structure of the crystal
that produced it. It is possible to identify magnetite in this manner. However, if the crystals
are very small, are in random order and dispersed among other iron species, the pattern
188 Chapter 7
may not be visible. We have been consistently unable to detect any pattern for any of the
iron-containing granules from the honeybee.
Another technique that is still being developed is electron energy loss spectroscopy
(EELS). In this technique, thin sections are cut from unosmicated, unstained material and
carbon coated. The electron beam is passed through the area of interest, namely the granule,
and the electrons that get through are analyzed for their energy loss. This technique allows,
in principle, one to determine the presence of light elements such as oxygen as well as
the heavier elements that one can determine from conventional X-ray microanalysis. There-
fore, it should be possible to determine the iron/oxygen ratio in the granule and obtain
some measure of the oxidation state of the iron. However, because there are other elements
in the tissue, such as phosphorus and calcium, as well as protein and these elements bind
oxygen as well as iron, it will be difficult to determine accurate ratios.
Definitive tests for magnetite are Curie temperature determination as used by Gould
et 01. (1978) and Walcott et 01. (1979) and M6ssbauer spectroscopy as used by Ofer et 01.
(1981). These techniques have two major disadvantages, namely they require large amounts
of tissue and they provide no anatomic information. However, if one can isolate the cells
or tissue that have been determined by X-ray analysis and light microscopy to contain iron
granules, then these two analytical techniques should be able to characterize the iron.
Thus by a combination of structural, histologic, and analytical techniques, it is possible
to localize iron particles in cells and to characterize them.
3. An Example: The Bumblebee
3.1. Introduction
Gould et 01. (1978) were the first to show that honeybee (Apis mellifera) abdomens
contained particulate iron. Subsequently, Kuterbach et 01. (1982) showed that the parti-
culate iron was localized intracellularly in a group of cells, the oenocytes, that surround
each abdominal segment under the cuticle. Lately, we have been examining cells in the
abdomens of bumblebees (Bombus sp.) which have similar habits and diet to Apis but do
not dance to indicate the direction of food sources to others in the nest. We have found
that, like Apis, Bombus abdomens have cells that contain iron. This section will describe
the data that we have obtained from the Bombus preparation using the methods described
earlier.
A Bombus abdomen when cut along the dorsomedial axis, pinned open, fixed, and
then stained for iron with potassium ferrocyanide shows intense staining of bands of cells
that surround each abdominal segment, as in Apis (see Kuterbach et 01., 1982). Unlike
Apis, however, there is an extensive development of the fat cells so that the blue staining
oenocytes are less obvious. However, when one removes a sheet of the cells and looks at
a whole mount in the light microscope, the blue staining is clear. Abdominal segments
can be fixed and embedded in glycol-methacrylate plastic and sectioned. When these sec-
tions are then stained for iron, the positively stained cells are obvious (Fig. 1). In this
section there is a group of cells under the cuticle containing dark granular material which
is blue when viewed in transmitted light optics. At a higher magnification (Fig. 2), the
granular nature of the staining is more obvious and one can observe that not all the cells
stain. It should be pointed out that the field was selected to show positive staining of cells
c
• •
o
I
Figure 1. Low-magnification light micrograph of a glycol-methacrylate section stained with potassium

ferrocyanide and Kernechtrot's stain. When viewed in the microscope, the cells containing the dark
granular precipitate are blue. C, cuticle; 0, oenocyte; F, fat cell; M, muscle. Bar = 50 J!.m.
and that in other areas of the sections fewer of the cells stain. The granules appear to be
relatively uniformly spaced within the cells and at this magnification are not seen to be
associated with any specific structure. The nonstaining fat cells occur among the oenocytes
and have a homogeneous cytoplasm. The iron-containing oenocytes, however, have a vac-
uolated cytoplasm. From these sections, it is possible to map the distribution of the iron-
containing cells around each abdominal segment and observe that more cells are found
Figure 2. Higher-magnification light micrograph of a thick section. Note the granular staining and
density of particles in the vacuolated oenocytes, whereas the fat cells have a uniform cytoplasm. Bar
= 20 fJ.m.
190 Chapter 7
Figure 3. Low-magnification transmission electron micrograph of a fat cell and an adjacent oenocyte.
In this osmicated and stained material, the density of the cytoplasm of both cells is high. The fat cells
(F) have amorphous dense granules, whereas the oenocytes (0) have large vacuoles (V) and small
dense granules (arrowheads). Bar = 2.5 !-Lm.
on the ventral surface of the abdomen than on the dorsal surface. Large nerve bundles can
be seen to run in very close association with the sheets of cells, particularly near the
ventrally located segmental ganglia.
Segments of the abdomen were fixed and embedded as described for conventional
and analytical electron microscopy. Thick (1 J.Lm) sections of this material were stained
by the potassium ferrocyanide method and the results confirmed the staining pattern pre-
viously described. The blocks were then trimmed down to those regions that had the
highest density of iron-positive cells and thin sections were cut and stained for electron
microscopy. Two cell types are apparent in such sections (Fig. 3), one with small vesicles
and large irregular electron-dense inclusions and the other with large vesicles and smaller
very-electron-dense particles. From the light microscopy of the thick sections of the same
material, only these latter cells with the small dense particles and the large vesicles were
ever seen to have a positive reaction to the iron stain. The cytoplasm of both cell types
appears to be dense in conventional electron microscopy and the small granules of iron-
positive cells are often hard to see within the cytoplasm at low magnification. At higher
magnification, the granules become more obvious (Fig. 4) and are seen to be often associated
with the membranes of the endoplasmic reticulum and in some cases appear to be enclosed
by membrane. The particles, on the order of 0.7 !-Lm in diameter, often occur in clumps,
creating a larger electron-dense mass.
In order to demonstrate the particles more clearly, and to perform elemental analysis,
thick (1500 A) sections of unosmicated, Epon-embedded portions of the Bombus abdomen
Figure 4. Higher-magnification electron micrograph showing the cytoplasm of an oenocyte. Note the
small dense granules (arrowheads) that are often associated with membrane. Bar = 0.5 f.l.m.
were cut. When these sections were stained with uranyl acetate and lead citrate, cellular
detail could be seen (Fig. 5) . In this particular section, the edges of two adjacent cells can
be seen, one (a fat cell) with the larger less-electron-dense granules, and the other (an
oenocyte) with the smaller but more-electron-dense granules. The smaller granules of the
oenocyte tend to shatter and rip the matrix during sectioning, something that is never seen
with the large granules of the fat cells. This implies that the smaller granules are of a higher
density and hardness. Again, the smaller granules are often seen to occur in clusters and
range in size from about 0.2 to 1.0 f1m in diameter. Their shape is more often spherical
but can be oblong.
When such sections are examined unstained, the cellular detail is not apparent (Fig.
6) although the smaller granules of the oenocytes are still as electron dense as in the stained
sections. This is further evidence that the density of the small granules is high and due
to a heavy element rather than an organic molecule. The larger granules of the fat cells,
on the other hand, are very difficult to see in these unstained sections, which implies that
they contain organic molecules which are dense only when stained. Because these uns-
tained sections are free of extraneously applied heavy metals, they can be used for ele-
mental analysis. The inherent density of the small granules makes them sufficiently visible
so that they can be individually analyzed.
3.4. Analytical Electron Microscopy
For elemental analysis of the granules, we used the thick unstained sections of the
unosmicated tissue. To reduce charging and drifting of the specimen, the sections were
carbon coated with spectroscopic-grade carbon. Figure 6 is a transmission image of such
a section. The electron beam was adjusted to about a 0.5-f1m spot and was centered on
one of the small dense granules in an oenocyte. After a ZOO-run, the X-ray spectrum shown
192 Chapter 7
Figure 5. Low-magnification electron micrograph of a thick (1500 A) section of unosmicated tissue

stained with uranyl acetate and lead citrate. A fat cell (F) with large amorphous granules is adjacent
to an oenocyte (0) with many small dense granules. Note that many of these small granules have
shattered and torn the matrix of the section. Bar = 2 fLm.
Figure 6. Electron micrograph of an unosmicated. unstained thick section used for X-ray elemental
analysis. Note'the dense small granules in the oenocyte (0) and the much less dense granules of the
fat cell (F). Bar = 1.7 fLm.
Figure 7. (aJ X-ray spectrum of a small gran-

ule of an oenocyte. Horizontal axis is the en-
ergy of the X rays; the vertical axis is the
number of counts per channel. Vertical full
scale = 256 counts. Note the peaks charac-
teristic of iron. copper. calcium. and phos-
phorus. (bJ X-ray spectrum of an amorphous
granule of a fat cell. The copper peak is from
the support grid. the sulfur is probably from
the granule and the chlorine from the embed-
ding plastic.
in Fig. 7a was obtained. The count rate was about 60 per sec and the deadtime averaged
about 4%. Thus, the detector was not operated near saturation. This analysis clearly shows
the presence of iron, calcium, phosphorus, and a large copper peak. The copper. however.
can be seen to come from the support grid. This is shown when one moves the beam to
one of the larger less-dense granules of the fat cell. Here (Fig. 7b). there is a large copper
peak due to the grid and much smaller amounts of sulfur and chlorine. The latter is probably
from the Epon matrix and the sulfur could be from the particle. In no case. however. is
there any iron. calcium. or phosphorus. Thus. this analysis shows that the small electron-
dense particles of the oenocytes contain iron. Table I gives the results of the same analysis
as Fig. 7 in numerical form. The background X-ray counts have been subtracted by a con-
ventional computer program and thus the counts shown for each peak are the integrated
values. There are several important points to be made from the data in Table 1. First is that
the control analysis contains a copper peak of comparable magnitude (number of X-ray
counts) to that of the experimental sample. yet there was no iron peak in the control.
Second, in the spectrum for the small dense granule. the distribution of counts between
the iron K" and Ki3 peaks is as predicted from the atomic structure of iron. namely the Ki3
is 15% of the K". Under some conditions, copper can produce an "escape" peak that is
very close to that of the iron K",. However. as the iron ratios are as expected in this case,
the iron peaks are accurate. A number of analyses have been done. including analyses of
contaminant material on sections, and in all cases the only material to show iron. calcium.
and phosphorus has been the small electron-dense particles of the oenocytes.
4. Conclusions
194 Chapter 7
TABLE I. Qualitative Elemental AnalysisO

Energy Area Element Line
Fig. 7a 2.021 1856 P Ka
2.626 202 Cl K"
3.338 166 K Ka
3.703 943 Ca K"
6.411 1803 Fe K"
7.059 275 Fe hú=
8.054 2003 Cll K"
8.917 295 Cll hú=
Iron (Fe) h ú L h ? = (100) = 15.3
Fig. 7b 2.313 253 S K"

2.644 224 Cl K"
8.050 1832 Cll K"
8.912 276 Cll hú=
a The energy is given in keY, and the area in number of

counts that accumulated in 200 sec. Note that the iron h ú =
peak is 15% of the Ka as expected, which indicates no in-
terference from other elements.
the hypothesis that animals use iron oxides in the sensory transduction process for mag-
netic field detection. So far, magnetite has been reported in the honeybee (Gould et a1.,
1978), homing pigeon (Walcott et al., 1979; Presti and Pettigrew, 1980), and dolphin (Zoeger
et al., 1981). However, the techniques used in these studies have been varied and the data
are not consistent.
In the homing pigeon, Walcott et a1. (1979) reported magnetite in connective tissue
associated with the interior of the dorsal skull as determined by SQUID magnetometry and
the Curie temperature of the isolated material. Presti and Pettigrew (19aO), OJ;l the other
hand, using similar techniques, found that the magnetic material was associated with the
neck musculature and not the skull. Later, Walcott and Walcott (1982) reported that they
were unable to replicate either observation using magnetometry, even though they ex-
amined over 80 birds, some of which were experienced proficient homing pigeons. We
have examined histologically a number of pigeon heads and necks using the potassium
ferrocyanide technique described in this chapter. The results of this analysis show that
many cells stain positively for iron. In particular, cells that surround the olfactory nerve
in the beak and that are associated with the ventral surface of the olfactory lobes often
stain positively. However, the results are not consistent from bird to bird and only about
70% of the birds examined show that staining pattern. Some cells in various glands, par-
ticularly the lacrimal and salivary glands, show consistent staining. However, we believe
that these cells are macrophages involved in the degradation of red blood cells and their
hemoglobin. Examination of the spleen, an organ where massive red blood cell turnover
oC,curs, also shows many cells with large positively staining inclusions. This observation
tends to support our conclusions about the glandular staining.
Gould et a1. (1978) reported that honeybees were magnetic by magnetometry and that
the isolated magnetic particles obtained from dried bee abdomens were magnetite as shown
by their Curie temperature. Later, Kuterbach et a1. (1982), using a histologic approach,
showed that the only cells that appeared to contain iron granules were the oenocytes that
surround each abdominal segment. However, when these cells were removed and examined
by Mossbauer spectroscopy, at least 95% of the iron present was found to be in the form
of a hydrous iron oxide and not magnetite. Thus, there is confusion as to the location and
identity of the iron in the bee. Perhaps the different drying methods used (long-term room-
temperature desiccation vs. rapid freeze-drying in a lyophilizer) could account for the
different results.
In the dolphin, magnetometry of isolated tissues localized magnetic material that by
a low-temperature test for multi domain magnetite was determined to be magnetite (Zoeger
et al., 1981). Unfortunately, the histology, done with a scanning electron microscope, did
not reveal the nature of the association of this magnetic material with tissues and in par-
ticular the nervous system.
In conclusion, the evidence for the involvement of an iron oxide, specifically mag-
netite, in magnetic field detection in any multicelled animal is at present very weak. Much
more work has to be done. A11 the available techniques, magnetometry, histology, elemental
analysis, Mossbauer spectroscopy, and electrophysiology to name a few, must be brought
to bear on a few suitable preparations. No one technique can provide the answer to the
question whether magnetite or some other iron oxide is involved in the behavioral re-
sponses of animals to weak magnetic fields.
ACKNOWLEDGMENTS. I wish to thank R. J. Reeder for his help with the X-ray microanalysis
and C. McKeon for her technical assistance. I thank S. Schuh for preparation of the plastic
sections of the bumblebees. Supported by NIH Grant GM-28804.
References
201:1026-1028.
Humason, G. 1., 1967, Animal Tissue Techniques, 3rd ed., Freeman, San Francisco.
Kuterbach, D., Walcott, B., Reeder, R. J., and Frankel, R. B., 1982, Iron-containing cells in the honey
bee (Apis melliferaj, Science 218:695-697.
Ofer, S., Papaefthymiou, C., Frankel, R. B., and Lowenstam, H. A., 1981, Mossbauer spectroscopy of
iron-containing dermal granules from Molpadia intermedia, Biochim. Biophys. Acta 676:199-
204.
Papi, F., 1982, Olfaction and homing in pigeons: Ten years of experiments, in: Avian Navigation (F.
Papi and H. C. Wallraff, eds.), Springer-Verlag, Berlin.
Walcott, B., and Walcott, C., 1982, A search for magnetic field receptors in animals, in: Avian Nav-
igation (F. Papi and H. G. Wallraff, eds.). Springer-Verlag, Berlin.
Zoeger, J., Dunn, J. R., and Fuller, M., 1981, Magnetic material in the head of the common Pacific
Chapter 8
Large-Volume, Magnetically Shielded
Room
A New Design and Material
GARY R. SCOTT and CLIFF FROHLICH
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 197
2. General Principles of Electric and Magnetic Shielding. . . . . . . . . . . . . . . . . . . . . . 199
2.1. Types of Shielding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
2.2. Static Magnetic Shielding by Materials with Magnetic Remanence. . . . . . . . . . . . 202
3. Practical Techniques for Building Magnetically Shielded Rooms. . . . . . . . . . . . . . . . 208
3.1. Site Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
3.2. Openings, Entrances, and Access. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
3.3. Joints, Corners, and Ends of Sheets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . " 211
3.4. Processing and Handling of Steel Sheets. . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
3.5. Erection of the Shield and Placement of Sheets . . . . . . . . . . . . . . . . . . . . . . . 213
4. Three Specific Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
4.1. Woodward-Clyde, Oakland . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
4.2. Sierra Geophysics, Redmond. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
4.3. Caltech, Pasadena . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
1. Introduction
Technical developments of the last decade have opened up new fields of research in pa-
leomagnetism and biomagnetism. In particular, the main impetus has been the commercial
development of superconducting quantum interference device (SQUID) magnetometers,
which utilize the Josephson effect to measure extremely small magnetic fields. SQUID
magnetometers measure magnetic moments about 100 times smaller than had been possible
previously. This improved resolution has allowed paleomagnetists to study remanent mag-
netism in most sedimentary strata, and has extended considerably both the geographical
and the chronological data base available for paleomagnetic research. Similarly, super-
conducting magnetometers are opening new vistas in physiology and medicine.
For several reasons, the advent of the SQUID caused researchers to become more con-
cerned with magnetic shielding. First, a sample measured with a SQUID magnetometer is
not spun but is held in a fixed position, and hence it is possible for the earth's field to
induce a magnetization during the measurement process. Similarly, the earth's field can
GARY R. SCOTT • Lodestar Magnetics, Inc., Oakland, California 94608.

CLIFF FROHLICH • Institute for Geophysics, University of Texas, Austin, Texas 78712.
197
198 Chapter 8
Figure 1. Generally, the shielding provided by two

thin shells of material is superior to a single very
thick layer. For example, for a solid spherical shell
of material of constant permeability fL of 2000, inner
radius R and outer radius 2R (left illustration), the
interior field at A will be reduced a factor of 390
below the exterior field B. However, where the
shielding is provided by two thin shells of thickness
0.05R at radius Rand 2R (right illustration), the field in the interior will be reduced by a factor of
1732, according to the formulas reported by Schwiezer (1962).
cause samples to acquire a significant moment during preparation, storage, and laboratory
analysis which may be retained for days or months. As the SQUID magnetometer must
measure fields of 10- 8 of the earth's field or less for biological or paleomagnetic studies,
the measuring procedure is simplified if one can reduce the ambient or background field
around the sample and magnetometer. Finally, by reducing the ambient field, one can
realize the advantages of using a first-order SQUID gradiometer rather than using a higher-
order gradiometer which is necessary when the background magnetic field is larger.
Long before paleomagnetism and biomagnetism were recognized scientific disciplines,
scientists knew that it was possible to build containers or rooms to provide shielding from
low-frequency and de magnetic fields, such as the earth's field (Wills, 1899). For example,
if a thick-walled sphere of inner radius Hl and outer radius Hz is built of a ferromagnetic
material of high permeability fL, the field inside the sphere is reduced by a factor S
S = 2fL
9
(1 _m
m)
from the ambient magnetic field outside the sphere (e.g., Jackson, 1962). Wills (1899) and
subsequent workers (Wadley, 1956; Cravath, 1957; Schwiezer, 1962; Cohen, 1967; Thomas,
1968; Cohen et 01., 1970; Wikswo, 1975) showed conclusively that several layers or stages
of shielding material provided a more effective shield than a single thick layer of material
(Fig. 1). Using these principles, shields have been built with as many as six layers of mu
metal, which is reported to reduce the interior field by a factor of about 10,000 (Erne et
01., 1981; Mager, 1981; Kelha, 1981).
Until recently, there has been very little literature published which gives practical
advice for those wishing to construct magnetically shielded rooms. A notable exception
is the work of Patton (1967), who gives a detailed summary of pertinent equations of
estimating minimum wall thickness, expected shielding effectiveness, etc., as well as ex-
amples of shielded rooms that he himself constructed. Our experience and that of others
(Scott and Frohlich, 1980; Symons and Stupavsky, 1983) corroborates many of Patton's
conclusions and speculations. For example, he correctly notes that permanent or remanent
magnetization is responsible for providing much of the effective shielding in most existing
magnetic rooms. He also speculates that for constructing magnetostatic shields, inexpen-
sive materials such as transformer steel may be as effective as expansive high-fL materials
such as mu metal (Table I). Indeed, within many laboratory shields the field caused by
scientific equipment provides a significant proportion of the effective field in the shield's
interior. In these situations, incurring great expense to reduce the interior field is not cost
effective.
However, our subsequent research has not confirmed the usefulness of all the design
practices advocated by Patton and Fitch (1962), Patton (1967), and Thomas (1968). For
example, Patton (1967) suggests that a magnetic shield must be completely enclosed, and
Large-Volume, Magnetically Shielded Room 199
TABLE I. Comparison of Transformer Steel and Mu Metal
Transformer steel Mu Metal
Moderate permeability High permeability

Moderate remanence Low remanence
High saturation induction Moderate saturation induction
1982 cost $1.30/kg 1982 cost $9.00/kg, plus annealing
Annealed during manufacture High cost of annealing, 1982 cost $125/sheet
Fabrication and machining during construction of Fabrication and machining before annealing
shield
Flexibility in size, shape, and design changes Limits to size, shape, and design changes
Brittle-requires hardened cutting tools Ductile-easy to machine
Electrical resistivity 50 x 1O- 6 0hm/cm Electrical resistivity 62 x 10- 6 0hm/cm
Composition: 96% Fe, 4% Si; silicon used as an Composition: 75% Ni, 18% Fe, 5% Cu, 2% Cr;
deoxidizer in steel melts hydrogen annealed for deoxidation
Shielding performance by remanent alignment Shielding performance by spontaneous magnetic
alignment
thus he takes special care in describing how to construct tight metallic doors and over-
lapping joints linking adjacent sheets of shielding material. Our experience indicates that
the presence of open doors or close-tolerance joints is not critical for static magnetic shield-
ing; instead, it is important to orient these openings properly with respect to the direction
of the magnetic field, and with respect to variations or gradients in the field. Indeed, it is
possible to construct effective shields with only two parallel walls and with no floor or
ceiling material.
In the present paper, Section 2 discusses the different types of shields that can be
constructed to reduce electric and magnetic fields in enclosed areas. We shall outline
briefly the properties and materials which are desirable for obtaining effective shielding
with each type of material.
In Section 3 we discuss the practical aspects of building static room-size magnetic
shields out of sheets of electrical transformer steel. The main purpose of this is to explain
in some detail what we know about constructing magnetically shielded rooms. These tech-
niques have been developed principally from empirical studies and not from theoretical
considerations. One of us (GRS) designed and constructed seven shielded rooms between
1981 and 1983, and the construction methods discussed are distilled primarily from his
experience. In Section 4 we illustrate the principles covered in Sections 2 and 3 by de-
scribing particular features of several of these rooms.
2. General Principles of Electric and Magnetic Shielding
2.1. Types of Shielding
In spite of magnetic shielding'S long history, the principles that govern the design and
construction of magnetic shields are sometimes poorly understood even by scientists who
build magnetic shields. One reason for this confusion is that there are several different
reasons for building a "shielded room," each requiring a slightly different type of material
and construction. In addition, in a qualitative way, different types of shields look much
alike. For example, most electrostatically and magnetically shielded rooms are just large
boxes enclosed by metallic sheets. Because ferromagnetic metals are relatively good con-
200 Chapter 8
TABLE II. Physical Characteristics of Various Common Materialso

Absolute
initial Resistivity Relative
Material permeability (m!llcm) Remanence cost
Mu metal 20,000 62 Low High

(sheets)
Purified iron 5,000 10 High Low
(sheets)
Permalloy 2,500 45 High High
(sheets)
Nickel 600 7.8 ?? High
Aluminum 1 2.8 Low
Copper 1 1.7 Low
Silver 1 1.6 Very high
a From Weast (1973).
ductors, magnetically shielded rooms generally provide fairly good electrostatic and elec-
tromagnetic shielding as well, even though these features may not be a part of their design
criteria.
The three physical properties of a material which influence shield design are its elec-
trical resistivity p, its relative permeability fL, and its magnetic remanence. The electrical
resistivity of a material is a measure of its resistance to the flow of electrical currents. The
resistance R provided by a material of length L and cross section A is
R = pUA
The relative permeability fL is a measure of the ability of the material to affect the magnetic
field in its immediate neighborhood. If one changes the ambient magnetic field by ú _ ç I =
the field within the material changes by an amount ú _ =given by
ú _ ç = = Ñi ú _ ç =
In general, the permeability depends on the ambient field, the frequency of the applied
field, the orientation of the material, and numerous other parameters of varying mathe-
matical intractability. For this reason, to calculate explicitly the fields expected for special
geometries, fL is usually considered to have a single, fixed constant value.
Finally, when an: applied field Bo is removed, many materials retain some magnetic
moment. This is known as remanent magnetization, and materials which exhibit nonzero
remanent magnetization are said to have remanence. Resistivity, permeability, and re-
manence for a material all depend on how a material is prepared and on its orientation.
However, for convenience, in Table II we present approximate values for these properties
for various materials which might be used to construct shielded rooms.
Shielding against dc (0 Hz) electric fields, or electrostatic shielding, can be obtained
by enclosing a room or container with a thin layer of any conducting material such as
copper foil or even metallic paint. Regardless of the static electric field outside the enclosed
space, the inside field will be identically zero. Indeed, more than a century ago Faraday
used this fact to demonstrate th&t the electric field due to charges falls off as the square
of the distance. The only design criterion necessary for complete shielding from outside
dc electric fields is for the space to be completely enclosed by the conducting material.
Therefore, construction details include sealable doors and special fixtures for cables and
power lines that enter the shielded space. For shielding against exceedingly high electric
fields, one might have to be concerned about the breakdown of air. One would want to
avoid sharp points on the outside of the room, or enclose portions of the exterior within
a dielectric material.
Shielding against high-frequency (AF) electric and magnetic fields, or electromagnetic
shielding, can be accomplished by enclosing a room or container with a thick layer of
conducting material, usually copper or aluminum. Faraday induction within the con-
ducting material causes currents that oppose the external fields. Within the shielding ma-
terial, the electric and magnetic fields of frequency f (Hz) drop off exponentially (by a
factor of 2.718) over a distance 1) called the skin depth, given by
where p is the resistivity, foL is the permeability, and foLo = 41T X 10- 7 kg m/C 2 • For a
conductor like copper, 1) = 0.85 cm at 60 Hz but 1) = 0.06 cm at 10,000 Hz. Thus, a room
surrounded by copper walls a centimeter thick will reduce fields with a frequency of 104
Hz by a factor of more than a million, but at 60 Hz the reduction is about a factor of three.
A much thicker copper shield or shield of ferromagnetic material with a high permeability
would be needed to more completely shield against 60-Hz fields. The design principle for
this type of shielding is to determine the amplitude and frequency spectrum expected for
the external field, and then to make sure that at each frequency the thickness of the shield
material is a high enough multiple of the skin depth to lower the field appropriately. As
with electrostatic shields, the best shielding is obtained if there are no holes or gaps in
the shield, thereby allowing the induced currents to flow uninterrupted in any path sur-
rounding the room's interior. Clearly, a room constructed in this way will provide effective
electrostatic shielding as well as electromagnetic shielding.
Shielding against lower-frequency varying magnetic fields, or dynamic magnetic
shields, can be accomplished by enclosing a region with layers of very-high-permeability
magnetic material, such as mu metal or permalloy. This is the type of shielding discussed
by Wills (1899) and Jackson (1962). The ideal material for this type of shielding is similar
to the materials used for constructing transformers. It should have a high relative perme-
ability foL, but as little remanence as possible. Unlike electrostatic or electromagnetic shield-
ing, the geometry of dynamic magnetic shielding is critical, and will be discussed further
below. Instead of using a highly permeable material, it has been possible during the last
decade to use Helmholtz coils, a sensor, and a feedback loop to cancel the magnetic field
within the working volume. However, this shielded region is only a small portion of the
volume enclosed by the coils. The necessity of sensors and the expense of maintaining
currents in large coils limit the feasibility of using Helmholtz coils for shielding room-
sized volumes.
The fourth category, shielding against constant (0 Hz) magnetic fields, or magnetostatic
shielding, is the primary topic of the remainder of this chapter. Previously, there has been
little distinction between dynamic magnetic shielding and magnetostatic shielding, be-
cause a completely effective dynamic shield also will shield against static fields. However,
as some of the design criteria and the type (and expense) of materials are different for the
two types of shielding, a clear distinction is important. In general, magneto static shielding
can be obtained by using Helmoltz-type coils or permanent magnetic material. For room-
sized volumes, permanent magnetic material would be the preferred choice. A material
for magnetostatic shielding must have high remanence, must maintain its magnetization
without decay, and must be readily available. Electrical transformer steel is such a material.
202 Chapter 8
_ú=
r
X
: ( uoM f
1 ú
s
=t---
X
r ---<
Figure 2, When a magnetic dipole with moment M in an external
field B is in its minimum energy configuratiion, there will always
be a locus of points at a distance r from its equator where the
0
4nB dipole field exactly cancels the external field.
2.2. Static Magnetic Shielding by Materials with Magnetic Remanence

It may seem paradoxical that "permanent" magnetic materials can provide magnetic
shielding. Indeed, the following results show that one can construct an effective shield
from the earth's magnetic field using only one sheet of magnetic material "oriented" north-
south along the earth's ambient field. Even a simple magnetic dipole, however, will di-
minish the field at some points in space if oriented in its minimum potential energy po-
sition in the external field.
The potential energy U associated with a dipole of magnetic moment M in a field Bo
is
U = -M'B o
and so the minimum energy configuration will occur when the dipole is oriented along
the direction of Bo. If r is the position of a point in space relative to the dipole, the magnetic
field at r is
(1)
Thus, for points in the equatorial plane of the dipole where M'r = 0, there is a zero magnetic
field (see Fig. 2) at a distance fo from the dipole given by
For groups of dipoles, the energetically preferred orientation depends on the strength
of the applied field. For example, for two dipoles of moment M1 and M2 separated by a
vector distance r12, the contribution to the potential energy U12 due to the interaction
between the dipoles is
U 12 = fLo [ri2 M1'M2 - P ú j N K ê N OF E j OK ê Ç g =

41T r12
If r12 is along the direction of B1 , the minimum energy will always occur when the moments
line up along the field direction. However, if the field is perpendicular to the vector joining
the dipoles, the minimum energy configuration will depend on the strength of the applied
field. For large fields, the moments will be aligned parallel along the direction of the field,
but for smaller fields both the anti parallel alignment and the alignment perpendicular to
the applied field will have a lower energy (Fig. 3).
Qualitatively similar results hold for thin sheets of permanently magnetized material.
If the sheet is magnetized so that the moments of all the domains lie parallel within the
plane of the sheet, this provides a nearly constant magnetic field outside the sheet opposite
in direction to the direction of magnetization (Fig. 4).
Figure 3, Four possible configu-
ú=t
rations of two magnetic dipoles
in the presence of an external
A
_ú= 2
lJ A = --ú ç -
j= - 2MB
tm
magnetic field (upward arrow in 211 R3
each illustration). Configuration
A always has the least potential
energy; however, in the presence
m em
of an external field, dipoles con-
mt
strained as in B, C, and D may
prefer to be in one of several ori-
entations. In large external fields,
configuration C has the least en-
ergy, whereas in smaller external
fields, D will have the least. If the
dipoles are to lie parallel to the 0
IWl [C]J
field direction as in Band C, con-
figuration B will have the least
energy for low external fields.
ú=
Although it is difficult to calculate the field explicitly for rectangular sheets, one can
integrate Eq. (1) explicitly for the case of a flat thin circular disk of radius R, thickness t,
and magnetic moment m per unit volume directed parallel to the plane of the sheet. At a
distance r from the disk along the axis of the disk, the field is
(2)
Note that when r <if R, the field is nearly constant. Perhaps more surprising is the fact that
N S N S N N
S N S N S S
N N
úN= ú = Figure 4. The magnetic field due

N
to a sheet of magnetic dipoles
--,
S S (left illustration) is opposite to
r-- -----1-
N N the dipole direction and is ap-
I
I proximately constant in the re-
ú= ú= gion near the center of the sheet
t
I
N N
(area within dashed lines). In an
S
ú= applied field with intensity B o ,
N N the domains tend to line up along
B
- ------
I
the field direction, creating a
__ J 0 S S
L __ nearly constant low-field region
N N
between the two plates (center il-
S S lustration). In a box with closed
N N ends (right illustration), the do-
mains at the ends have difficulty
S S responding to the field in a co-
N S N S herent fashion, and thus the ends
S N S N provide little additional shield-
N S N S
ing in the interior regions.
204 Chapter 8
the field vanishes as R becomes very large, and we approach the case of an infinite sheet
of magnetic moment.
For our purposes, we can use the above formula to calculate approximately the shield-
ing factor S produced by the disk if all the magnetization is induced by the permeability
1.1. of the material. Suppose the disk is aligned so that its surface is parallel to the earth's
field Bo. From Maxwell's third equation, one can show that the magnetic moment m per
unit volume induced in a material by an external field Boutside that parallels the material's
surface is
1.1. - 1
m = - - Boutside (3)
1.1.0
However, the field Boutside includes both the earth's field Bo and the field due to the induced
magnetic moment. From (2), since r is zero this is
(4)
By combining (3) and (4) we can solve to find the shielding factor S, or the reduction in
the earth's field caused by the presence of the disk, i.e.,
ú =1
= + (1.1. - l)t =S (5)
Boutside 4R
Of course, this result is only approximate as we have tacitly assumed that the induced
magnetization is everywhere constant and equal to its value in the center of the disk.
These results show that one can construct an effective shield from the earth's magnetic
field using only one or two flat sheets of magnetic material oriented along the ambient
field direction (Fig. 4). In the case of two sheets, the lowest energy configuration is for
sheets oriented so that the magnetization direction is along the ambient field direction,
which produces an opposing field in the central region between the sheets. Note that the
shielding does not depend on having a closed region. If end sheets do exist, they contribute
little to the shielding, for the domains in the end sheets cannot align easily along the
ambient field direction (Fig. 4). However, the end sheets do react to the fringing fields
generated by the parallel sheets.
Of course, a limitation of calculations like the one above is that at ordinary temper-
atures, all real magnetic materials have both remanence and permeability. For this reason,
it is not possible to separate the effects of permanent magnetization and of permeability,
as we do in the discussion above, and as do authors such as Wills (1899) and Jackson
(1962). In addition, remanent dipole moments in a magnetic material do not have a com-
pletely fixed magnetization m; rather, m depends on the prior history of the material.
Finally, the permeability 1.1. depends on the strength and frequency of the applied field.
We can demonstrate the shielding effect possible with plane sheets of transformer
steel by performing a simple experiment (Fig. 5). We aligned the surface of a 66 x 71 x
0.062-cm rectangle of transformer steel along the earth's field and "demagnetized" it by
passing it through a coil consisting of about 200 turns of copper wire carrying a 60-Hz AF
current, thereby creating an alternating field (Fig. 6). The field created by the current is
initially strong enough to reorient domains, but it decreases to zero as the coil passes away
from the sheet. Thus, this "demagnetization" causes the permanently magnetized domains
in the transformer steel to realign so that they have a low potential energy in the earth's
field.
Figure 5. If sheets of transformer

steel are demagnetized in the
presence of the earth's field Do by
passing them through a large coil
carrying a 60-Hz current, the re-
sulting magnetic field near the
center of the sheets (at xl de-
pends strongly on the orientation
of the sheets during demagneti-
zation. When demagnetization
occurs while the sheet is oriented
parallel to the earth's field as in
the illustration on the left, the
field at x is reduced to only a few
percent of Do, showing that
shielding is occurring. However,
when demagnetization occurs
with the sheet oriented perpen-
dicular to the earth's field, al-
most no shielding occurs at x.
When we performed this experiment recently at the Woodward-Clyde Laboratory in

Oakland, the resulting magnetic field adjacent to the center of the sheet was 1500 nT. As
the earth's magnetic field in Oakland is about 50,000 nT, this demonstrated that a shielding
factor of about 35 could be obtained with only a single sheet of magnetic material.
We then took the steel sheet into a shielded room, and remeasured the magnetic field
near its center to get an idea of the remanent magnetization that remained. This remanent
field was about 4500 nT with a direction opposed to the earth's field during AF treatment.
This suggests that about 10% of the shielding provided by the sheet was from remanent
magnetization, and shows that closed boxes are not essential for effective shielding.
Now suppose we repeat the entire experiment but take care to align the plane of the
sheet perpendicular to the earth's field. When the field at the center of the sheet was
measured outside of the shielded room, it was 50,000 nT. Effectively, in this orientation
the sheet is transparent to the magnetic field. When the sheet was "demagnetized" in the
same fashion inside the shielded room, the field at the center of the sheet was about 100
nT, or about the same as the field inside the shielded room. This demonstrates that the
magnetization depends on the history of the material.
In general, because any two orthogonal planes contain all components of a vector,
constructing only two sets of surfaces may provide sufficient shielding. However, for any
given site (or vector), the simplest room design would include only a pair of walls precisely
aligned with the magnetic declination. This design is not practical, but serves to illustrate
the general principle that it is the planes parallel to a magnetic vector component that
provide most of the shielding from that magnetic component. The east and west walls are
of primary importance for shielding a typical site at intermediate latitudes, with the north
and south walls of secondary importance.
206 Chapter 8
Figure 6. Photograph of the large sweeper coil used by the authors to demagnetize individual sheets
of transformer steel and to reorient domains within sheets after they are in place in the shielded room.
The coil consists of approximately 200 turns of l8-gauge copper wire which can be connected to a
l20-V AC source. Smaller sweeper coils are used to demagnetize parts of the room with restricted
access.
A serious practical limitation of theoretical treatments of shielding (e.g., Wills, 1899)

is that they ignore remanence and also assume that the material permeability IL is a constant
independent of the field B. In fact, the permeability becomes markedly smaller at low
fields (Fig. 7) because the proportion of magnetic domains aligned with the field is reduced.
In theoretical treatments of magnetics, the fraction of domains that is aligned with the
applied field is known as the Langevin function L(a) (see, e.g., Kittel, 1966). Here, a =
MB/kT is the ratio of magnetic potential energy MB to thermal energy kT for individual
domains, with k being Boltzmann's constant and T the absolute temperature. For regular
groupings of identical isotropic magnetic domains, it can be shown that
20 l10,OOO
24 gauge transformer steel M22 FP
18 data from U.S.STEEL 9,000
16 8,000
>-
.-
14 7,000
.c
l\l
en 12 6,000 Q)
en E
.....
::::l
l\l Q)
Cl 10 5,000 a..
.Q
ú=
lii
8 4,000 E
.....
C 0
0 Z
:;::;
0 6 3,000
::::l
"C
.E
:t.
4 2,000
III 1,000
0 0
.1 10 100
H DC Magnetizing Force, oersted

Figure 7. Induced field B and permeability j.L as a function of dc magnetizing force for the transformer
steel used by the first author in the construction of magnetically shielded rooms. The arrow at 0.5 De
shows the approximate strength of the earth's field. The graphed data are provided by the manufacturer
of the steel sheet.
L(a) = coth (a) - l/a
As B is reduced, the fraction of aligned domains L(a) approaches zero, and thus the effec-
tive permeability of the bulk material also becomes small. For this reason, if one wishes
to use Wills's (1899) or Patton's (1967) formulas to obtain an accurate estimate of shield-
ing from several layers, one must use a different effective j.L for the permeability of each
layer.
Considering these problems, it at first may seem fortuitous that shields built using
Wills's principles provide shielding so effectively. In addition, because domains tend to
line up along the local field direction, why don't they sometimes produce a field which
more than cancels the earth's field, i.e., a net field in the opposite direction from the earth's
field? Unfortunately, in this chapter we cannot provide quantitative answers to these prob-
lems.
However, in a qualitative way we can explain why it is reasonable that properly pre-
pared rooms will behave as Wills calculated. In a high-j.L, nonremanent material, domains
are largely free to realign themselves along the local direction as the field changes. In
contrast, in a real material the direction of some of the domains remains fixed, especially
at lower fields. However, almost any activity such as hammering, thermal agitation, or
demagnetizing with AF "sweeper" coils (Fig. 6) allows the fixed domains to realign along
the local field direction, i.e., the same direction as for a nonremanent high-j.L material.
Thus, the shielding occurs because both remanent and nonremanent domains which realign
are aligned in the same manner as calculated for a high-permeability material, and the
remaining domains are either aligned randomly or aligned along some other historically
208 Chapter 8
The general tendency for domain alignment also helps explain why properly prepared
real materials provide even higher shielding than calculated from laboratory measurements
and formulas such as Eq. (5). The measured permeability for a material reflects only the
activity of the nonremanent domains. Because the preparation of the shield causes the
domains to align in the same manner, the effective permeability will be higher than the
measured permeability. In the experiment performed in Oakland with the 66 x 71 x 0.062-
cm sheet, we measured a shielding factor of 35. However, according to the manufacturer's
specifications (Fig. 7), fL for this electrical transformer steel in the earth's field is about
3000, and so from Eq. (5) we would calculate a shielding factor of about 2.5. This difference
was also demonstrated during construction of the outer layer of the shielded room in
Oakland. Patton's (1967) approximate formula for the shielding factor S provided by a
single layer of material of permeability fL of thickness t enclosing a space of interior di-
mension L is
S = 1 + 1.34 fLt/L
For the Oakland room, Scaled = 2.7, since t = 0.124 cm, L = 3.0 m, and for transformer
steel in the earth's field of 50,000 nT, fL = 3000 (Fig. 7). When the outer layer of the
Oakland room was constructed, the measured shielding factor was Smeasd = 6 (internal
field = 8500 nT). After some vibrational alignment (hammering) of domains, the internal
field was reduced to 2900 nT. This corresponds to an effective permeability of 28,000, far
above any measured value. This further demonstrates that with transformer steel, reman-
ence plays a major role in shielding.
3. Practical Techniques for Building Magnetically Shielded

Rooms
Many of the features of magnetic shield design and fabrication discussed here have
been used by previous workers. For example, Patton (1967) speculated that effective shields
could be constructed from thin sheets of electrical steel. Our usual construction technique
involves building a two-stage shield with a spacing of 25-30 cm between the stages. Each
stage has two thicknesses of 0.062-cm steel, for a total thickness of 0.124 cm/stage. Gen-
erally, we use the shielding factors reported by Patton (1967) and others to predict ap-
proximately the performance of a projected design. Like ordinary nonshielded rooms, the
framing for shielded rooms is constructed of wood; however, we always take special care
that the entire room is constructed with nonmagnetic materials except for the steel sheets.
For example, we use aluminum nails to put together the wooden frames and brass screws
to attach adjacent sheets to one another and to the frame.
Some of the construction practices we follow routinely are not always necessary to
obtain good shielding performance. For example, we generally demagnetize tools such as
screwdrivers or saws before using them to construct the interior shield of a room. We also
never lay tools (especially power tools) on the floor of a shield; rather, we place them on
wooden tables or boxes. However, localized magnetization induced by careless construc-
tion practices usually can be remedied afterwards with sweeper coils.
In this section we discuss in special detail certain features of our shielded room design
such as door entrances and joints which differ from the design suggested by Patton (1967).
We also discuss certain practices not emphasized by previous authors, such as conducting
a site survey and preparing the shielding materials before construction begins.
3.1. Site Survey
To ensure optimum performance of a magnetic shield, a detailed magnetic site survey

should be undertaken before construction begins. Besides finding the general features of
the magnetic field (direction and magnitude). a survey will reveal potential problems that
are the result of spatial or temporal variations at the site. The magnetic field inside larger
buildings is often distorted by structural or reinforcing steel and by service pipes. If these
spatial gradients are too large or extensive, they can reduce the effectiveness of a shield.
In practice, magnetic shields attenuate uniform fields more effectively than magnetic gra-
dients. If remanently magnetized material is the source of the gradient, it is often possible
to demagnetize the material with an AF sweeper coil. Throughgoing straight pipes or beams
are of little concern if they are oriented in an E-W direction, as they can be given a nearly
zero field by AF demagnetization. However, large fields are associated with ferrous ma-
terials having a N-S orientation, especially at bends or terminators. The best plan is to
have at least a double thickness of shielding material in and around areas having large
gradients. In addition, openings in the shield should be avoided in areas of large fields or
gradients. Finally, keep in mind that some kinds of laboratory equipment may produce
substantial fields within the shielded region.
A site survey should also search for temporal variations in the field. Because much
of the shielding produced by real materials is provided by remanent domains, the shield
will be less effective for temporal variations. Moving or movable magnetic sources are a
common problem. Examples are vehicles (both moving and parked). furniture, filing cab-
inets, laboratory equipment, and elevators. Avoidance of these problems is the best policy;
we recommend choosing a site that minimizes these features. It is helpful to request that
you be informed whenever equipment or furniture is moved into spaces near the shield.
Whenever the local magnetic field does change, the magnetic shield can be given a
new remanence which responds to those changes by using a sweeper coil as discussed in
a later section. The need for periodic servicing of a magneto static shield requires that the
shielding surfaces be accessible. In general, it is sufficient to give the inner surfaces a new
remanence, but for larger variations in the ambient field, the outer shield must be accessed.
3.2. Openings, Entrances, and Access
The problem of access is associated with all shielded enclosures. The solution differs
markedly for different types of shielding. For electromagnetic shields, the solution is to
completely enclose the shield with sealable doors without allowing any "line of site"
passages. The same solution works for magnetic shielding as well; however, while this
approach is effective, it restricts the design of magnetically shielded enclosures. More
flexible solutions to the problem of accessways can be found by considering the working
portions of a magnetic shield, and by contrasting the operation of magneto static and elec-
tromagnetic enclosures.
Two features important for electromagnetic shields are unnecessary for static magnetic
shields: "line of site" restrictions, and electrically sealed doors (e.g., see Fig. BA,B). In
practice, the two basic classes of shielding materials are those with high electrical con-
ductivity and those that exhibit ferromagnetism (Table II). Not only do these classes operate
using different physical mechanisms, but in a conceptual sense they provide shielding
from "sources" at right angles to one another. Because electromagnetic waves are transverse
waves, a plane sheet of eddy current shielding material such as aluminum will be most
effective against electromagnetic waves directed perpendicular to the plane of the material.
Alternatively, if the conducting material is parallel to the direction of wave propagations,
210 Chapter 8
Figure 8. Vertical north-south

cross section through 7-foot mu-
.. .y
metal cubic room at University of
Texas facility in Galveston, (A)
showing magnetic vectors out-
ú y j = ú = ú= i •••t
side the shield. Scale indicates
intensity of field vectors, indi-
\r
Ceiling 1 .!t--" cated by arrows. Stars are used
:: f----
/
I I I / ....
r---.------'
"ú ú =
for measurements greater than
0.1 mT. Ambient field, Bearth is
shown to scale. Declination = 6°,
Be
I I I I úy= \ inclination = 65°, Bearth =
}\ t 46,500 nT. (B) magnetic field

I I I I t
t
generated by shield material.
Vectors showing generated field
/'
, I
/ I / / are identical to magnetic field
\.
úú=
ú I =
vectors in (A), except the earth's
field of 46,500 nT has been sub-
tracted. Note the large variations
• Floor UT-Austin
in field direction and intensity
o I J J f ú
l ú=
ú _ _-.JI
=
1m that occur near the external edges
B o O.lmT of the shield. Stars are sites
ú =
where field exceeded 50,000 nT.
charges within the material can move only a small distance, and little shielding is achieved.
Thus, in practice this sheet is opaque to waves impinging on its surface, but is nearly
transparent when aligned with the waves. The situation for materials with ferromagnetic
properties is exactly the opposite. A sheet of this material is largely transparent when
oriented perpendicular to a magnetic field. However, by aligning the sheet along the field
direction, magnetic shielding occurs symmetrically on both sides of the sheet. Clearly,
different access openings must be designed for the two classes of shielding material.
We have built several quite effective magnetostatic shields which have no doors at
the entrance (e.g., see Fig. 9A,B). From the preceding discussion, it would appear that
access should be made where the magnetic field lines intersect the ends of the shielded
regions, for here the planes of shielding material are perpendicular to the field and have
little impact on the shield's overall effectiveness. Surprisingly enough, this is not a realistic
choice. Two serious flaws are encountered; first, to approach such an entrance, one must
pass through a high-field region (e.g., see Fig. 8). At the Woodward-Clyde shielded room
in Oakland, there are fields greater than four times the ambient (e.g., 200,000 nT) just
outside the lower-north and upper-south ends of the room. This suggests that openings
should face perpendicular to the ambient field, for then the field encountered decreases
steadily as the shield is approached from the exterior. Our experience in constructing
shielded rooms with openings which face perpendicular to the ambient field fully supports
this conclusion.
A second serious problem with openings which face parallel to the ambient field
direction stems from the high fields and large gradients produced by the shield itself at
the edges of the enclosure (Fig. 8B). When there are openings in these regions, a portion
of the ambient field, and/or the field produced by the shielding material intrudes into the
enclosure. If shielding material is placed over these openings, it produces shielding by
responding to these locally produced fields and gradients. Of course, if a very large shield
were built elongate and parallel to the ambient field, these effects could be avoided by
working only in the middle areas (e.g., Wikswo, 1975). However, in most practicallabo-
ratory shields, completely enclosing the end regions provides shielding from high fields
as well as from shield-produced gradients.
In summary, there are two general rules for locating openings and entrances. First,
access passages should face in a direction normal to the magnetic vector, and second, they
should be far removed from the magnetically extreme ends of the shield.
3.3. Joints, Corners, and Ends of Sheets
Previous discussions of shield design and construction have emphasized the impor-
tance of the joints between sheets of material (Patton, 1967; Thomas, 1968). This is a valid
concern for electromagnetic shields because electrical conductivity is important; however,
simple overlapping joints of ferromagnetic material give excellent results for magnetostatic
shields. The joint system we use incorporates approximately 5 cm of overlap, with the
two plates screwed together using solid brass fasteners spaced about every 15-20 cm. When
a covering is used, such as plywood or gypsum board, the spacing between screws can be
25 cm or more. These values for overlap and screw spacing are conservative, as we have
constructed joints with 1-cm overlaps which are virtually unnoticeable inside the shield.
Generally, overlapping joints cannot be detected magnetically. Some exceptions occur, but
this is restricted to within 10 cm of the joint surface and is only for joints in regions with
high magnetic fields.
Our design for corners has been simple and successful. We use a special nonmagnetic
bending brake, constructed from hardwoods, plywood, stainless steel and aluminum to
fold the steel sheets into a right angle. As there are two steel layers on all surfaces, a set
of folded sheets can be arraJ?ged so that a corner is fully covered except for the apex. For
the interior, whenever a sheet ends at a large opening such as an entranceway, the sheets
are bent approximately io cm outward. This places the end of the sheet and its attendant
large field further from the shielded space.
If possible, we complete the construction and demagnetization of the outer stage before
beginning the inner stage. For the interior sheets, we take care to perform the bending in
the low-field region produced by the outer stage of the shield. After the outer shield is in
place, the bending brake is moved into the partially shielded room (B = 3000 nT) and the
interior sheets are folded there. This apparently reduces the alignment of high-coercivity
magnetocrystalline domains in the deformed portions of the sheets.
212 Chapter 8
q ~ á é ú á K =Taiwan
AI
o
ú J ú
ú= J J
1m
= 1=
w
...
/
úáú= =
"
• r.1 -.1
f ú=
L
til 00
>j)
úLl á= ú=
l • •
ú =
.Ii
>j)
.1 ú=
.... \ 1 1
"• •
ú =
\ 1 1
.- • ... '"
f ú=
0- >j) 00
.. "'til" • • • •
>j) ú = <!' til >j)
ú=
....
Ii •
ú=
i
<!' >j)
i
>j)
i : ;
• •
=
Ii
..•
f ú=
ú=
....
II')
i
....
i
....
ú=
>j)
i i I
úJ
/"i
A S
Figure 9. (A) Plan view of shielded room at Institute of Earth Sciences, Academia Sinica, Taipei,
Taiwan. Numbers are field intensity (nT) measured 5 days after construction at points indicated by
the filled circles. Dashed lines show approximate contours of measured fields at intensities of 100
nT, 500 nT, 1000 nT, and 5000 nT. Arrows adjacent to "E" and "Woo show position of vertical cross
section in (B). Before construction, the ambient field intensity was 35,800 nT, with a declination of
8° and an inclination of 38°. (B) Vertical cross section through doorway in Taipei room, with field
intensities and contours labeled as in (A). Note that strong gradients in field intensities are restricted
to doorway area.
3.4. Processing and Handling of Steel Sheets

The shielding material we use is manufactured by u.s. Steel and is 24-gauge (0.62
mm), M-22, FP (fully processed), CP-3 (core plate) electrical steel with a saturation mag-
netization (Hs) of 400 Oe. It comes in rolls of up to 107 cm (42 inches) in width which
can be slit to any convenient width and cut to any desirable length. Generally, we have
used sheets of 66 x 132 cm (26 x 52 inches) or 91 X 132 cm (36 X 52 inches). These
sizes are convenient to handle during construction and require only a moderate number
of joints. The weight is 11.84 kg/m 2 (0.5 Iblft 2 ).
Before installing the shielding material, in many of our shields we demagnetize each
sheet to remove any remanent magnetization remaining from the processes of manufac-
turing, cutting, or handling. For demagnetization, we pass a large electromagnetic coil (60
Hz, 15 mT; see Fig. 6) around each sheet to modify the magnetic direction for most of the
remanent domains in the material. Because we control the orientation of the sheets relative
to the ambient field, when the AF electromagnet is removed, the sheets acquire a mag-
netization with known magnetic direction and relative intensity. The acquired magneti-
zation is refered to as anhysteretic remanent magnetization (ARM). Any orientation can
be used; one can even orient the sheets perpendicular to the earth's field to reduce the
magnetization acquired. However, we prefer to orient the ARM direction along the length
of the sheet. For this purpose we use an inclined table, with a tilt equal to the ambient
field inclination and facing toward magnetic north. The direction of the ARM is then
o..............____............_ _---J'1m
Taipei. Taiwan
.. ..-; 8
- -
"'
".. ú= ..:
!
I
.. ... ú=
-
"'
-
"' ú=
\
\
-
l
i
floo r
B L..-_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _.....
"'"''''']''' ?úDúDXX?m?????="." .• ".""".,,.
Figure g. (continued)
marked on each sheet so that it can be oriented later when the room is erected. During
this demagnetization-ARM acquisition process, the sheet is passed through the center of
the coil to expose the steel to the maximum fields available so that even high-coercivity
domains will be realigned. Subsequent procedures with the same coil will have insufficient
coercive force to realign the high-coercivity domains, but will affect only those with lower
coercivity.
3.5. Erection of the Shield and Placement of Sheets
The erection sequence we use for building rooms of electrical steel is designed to
simplify construction and to allow access to all surfaces (walls, floor, and ceilings) so that
they can be magnetized in place. The most convenient situation is a free-standing enclosure
where only the outer floor has restricted access. However, when space is at a premium,
shielded enclosures can be and have been built with four or five surfaces against existing
walls and ceilings. In these situations, planning must allow for these surfaces to be mag-
netized before they are closed off by later construction. If possible, some limited access
(even enough room for an electromagnetic coil) should be provided to allow for magnetizing
the surfaces. This may be necessary both during the construction of the shield as well as
later in cases where the ambient field changes because equipment or furniture is moved.
Each sheet is placed such that any previously acquired ARM is aligned with the max-
imum component of the ambient field in the plane. For example, for a shielded enclosure
with N-S orientation, inclination of about 60 0 , the sheets on the floor have their northward-
directed ARM aligned toward the north. For the north and south walls, the northward ARM
214 Chapter 8
is aligned downward. Although in principle one should install the sheets on the east and
west walls with their ARM inclined at 60°, in practice it is sufficient to align the ARM
downward, along the largest component of the field in the plane of the sheets.
Construction usually starts with the outer-stage floor. First, we give it an overall ARM
with the same coil that processed the individual sheets. In this case, we simply pass or
"sweep" over the surface in much the same manner as when cleaning with a mop or push
broom. The field continuously realigns perhaps 80% of the domains in an up and down
direction, perpendicular to the plane of the shielding material. As the coil moves away,
the ambient dc field preferentIally realigns a portion of these coerced domains along the
component of the ambient field in the plane of the material. The amount of ARM acquired
is proportional to the strength of the ambient dc field. Thus, the newly acquired remanence
is in the same direction and has the same relative strength as the ambient field components
in the plane of the shielding surface.
In general, the entire outer stage is erected and given an overall ARM. If access is
restricted, this ARM acquisition procedure is repeated after the construction of each wall
and ceiling. However, in the next section we will discuss an example where the inner
shield walls were built first and the enclosure still provided excellent overall shielding
performance. Regardless of the erection sequence, this conditioning process can be re-
peated subsequently if the interior field increases, either for the entire room or for a small
part of the shield. In most situations, magnetically mopping the inner shield (Fig. 6) will
yield the lowest interior fields.
4. Three Specific Examples
4.1. Woodward-Clyde, Oakland
We constructed our first two-stage electrical steel, shielded room in Oakland, Cali-
fornia, for Woodward-Clyde Consultants in 1980 (Fig. 10). This room had a shielded vol-
ume of 23 m3 • At this time we had not yet developed our methods for remagnetizing the
sheets in place with AF coils, although before installation we gave individual sheets a
relatively weak ARM. When the outer layer of shielding was in place, the internal field
was about 8500 nT. To improve the shielding, we first attempted to use physical shock or
Oakland
transformer steel
--"
Figure 10. Plan view of the

ramp ú= Woodward-Clyde shielded room.
Although this room was designed
.. with a compound entryway (a par-
tially shielded passageway lead-
ing into the room's interior), the
MN first author's subsequent experi-
t
10' X 10' X 8' inside
ence has shown that this feature is
not necessary to obtain high
16 X 12 X 10 outside shielding factors.
vibration to induce domain realignment. After experimenting with various mechanical

vibrators and hammers, the most effective method was found to be a hand-held rubber
mallet. By pounding the entire outer shield, we achieved an average field of 2950 nT,
corresponding to a shielding factor of about 15.
After installing the inner shield, the field was 490 nT. However, thoroughly pounding
the inner shield only reduced the internal field to 325 nT. Apparently, we had reached
the effective limits of shock magnetization, so we had to devise another more coercive
method. We found that by "sweeping" over the walls with the same electromagnetic coil
that processed the sheets, we immediately reduced the internal field to 126 nT. Sweeping
both the inner shield and two of the outer walls produced an average field of 103 nT. This
corresponds to a shielding factor of about 420.
While measuring the residual fields, we noticed an episodic magnetic field. The source
of this field was vehicular traffic on a nearby six-lane boulevard centered only 15 m to
the north of the shielded area. Magnetic pulses of 150 nT lasting 4-6 sec accompanied
passing cars. Buses and trucks produced fields of 800 nT inside the shield. The inability
to respond effectively to changes in the external field is a limitation of highly remanent
materials such as electrical steel, and underscores the importance of selecting a magnet-
ically quiet site.
4.2. Sierra Geophysics, Redmond
The shielded laboratory built for Sierra Geophysics near Seattle, Washington in 1981,
provided a graphic illustration of the remanence characteristics of electrical steel. This
room enclosed 28 m3 in a steeply inclined field (54,000 nT, inclination 72°). The con-
struction sequence was:
1. Step 1, build outer floor and ceiling and treat with sweep coils.
2. Step 2, build all interior surfaces and treat with sweep coile;.
3. Step 3, build outer walls.
We selected this sequence so that the inner walls could be aligned magnetically with the
nearly unshielded vertical field. After the first two steps were complete, the interior field
was surprisingly low, ranging from 290 nT to 820 nT with an average of 480 nT. The next
day, the outer walls were erected and given an in situ ARM with sweep coils, in effect
partially shielding the inner stage. The interior field now ranged from -12,000 nT to
-14,900 nT, with an average of -13,400 nT, and with a direction which now opposed
the ambient field. We allowed the shield to set in this configuration for 5 days without
further treatment, remeasured and found that the field ranged from -11,700 nT to -14,100
nT, with an average of -12,900 nT. We then swept the exterior walls again with a coil,
and the average field changed only slightly to -13,500 nT. These experiments clearly
demonstrate that remanence domains with long relaxation times (greater than 4 x 105 sec)
provide a significant portion of the shielding produced by electrical steel. It also shows
that the magnetic history of the shielding material is important.
After treating both interior and exterior walls extensively with sweeper coils as de-
scribed below, the shield reached an interior field of - 95 nT (range from - 43 nT to -147
nT), for a shielding factor of 570. However, the interior field remained directed opposite
to the ambient field. The process by which the interior field reached this low value il-
lustrates the coercivity of the remanence and the interaction of the two shielding stages
as they were remagnetized repeatedly (Fig. 11).
What we intended to do was to gradually demagnetize (and therefore remagnetize)
the inner stage in small steps until the interior field approached zero or reversed direction.
N
...=
Q
TABLE III. Characteristics of Various Shielded Rooms
Interior Number of Approximate
dimensions Volume ferromagnetic interior field Year of
Institution and location (m) (m 3 ) layers Material (nT) construction
University of Windsor 3.0 x 2.7 x 2.5 20 1 Steel 600-1400 1973

(Ontario)
University of California 3.7 x 4.3 x 3.0 48 2 Steel 100-250 1983

(Santa Cruz)
Princeton University 5.6 x 3.8 x 2.1 45 2 Steel 400-1200 1981
(Princeton. N.J.)
Stanford University 9.1 x 1.2 x 1.2 42 2 Mu metal 50-100 1976
(Stanford. Calif.) (cylinder)
University of California 4.9 x 4.0 x 2.1 41 2 Steel 60-125 1983
(San Diego)
California Institute of 4.2 x 3.3 x 2.8 39 2 Steel 120-220 1982
Technology (Pasadena)
Woodward-Clyde 3.1 x 4.3 x 2.4 32 2 Steel 75-250 1983 b
(Pleasant Hill. Calif.)
Sierra Geophysics 3.6 x 3.3 x 2.4 28 2 Steel 90-150 1981 n
::r
(Redmond. Wash.) 't:I
.-+
'"
Woodward-Clyde 3.0 x 3.0 x 2.4 22 2 Steel 75-250 1980 ..,C!l
(Oakland. Calif.) 0>
Academia Sinica 2.5 x 3.6 x 2.5 23 2 Steel 50-125 1983 ..,IIIt""'
(JQ
(Taipei, Taiwan) ct>
Phillips Petroleum 2.1 x 3.7 x 2.6 20 2 Steel 50-125 1983 <:
0
(Bartlesville, Okla.) 2"
South Dakota School of 2.7 x 2.4 x 2.7 17 2 Steel 300-400 1981 8
!I'
Mines (Rapid City)
ú=
University of Hawaii 2.4 x 2.4 x 2.4 14 2 Mu metal 75-120 1977 c III
(JQ
(Honolulu) ::l
ct>
C;.
California Institute of 1.8 x 1.8 x 1.8 6 2 Mu metal 50-200 1981 d -
Technology (Pasadena) e:..
University of Texas 1.8 x 1.8 x 1.8 6 2 Mu metal 75 1973/75
-<
rn
::l"
(Galveston) CD·
s:
ct>
0..
University of Houston 2.7 x 2.4 x 3.0 19 3 Steel 100-250 1983
:::0
(Houston, Tex.) 0
0
University of Windsor 3.0 x 2.7 x 2.1 17 3 Steel 50-100 1982 8
(Ontario)
Otaniemi (Finland) 2.4 x 2.4 x 2.4 14 3 Mu metal 5 1980
Berlin (Germany) 2.2 x 2.2 x 2.2 11 6 Mu metal 10 1981
a Values listed are from literature or from individuals who have built or used these rooms.
b Rebuilt from 1980 Oakland.
C Rebuilt from 1965 UT-Dallas.
d Rebuilt from 1967 JPL.
N
ú =
'I
This appeared possible because the inner shield had been given a relatively strong ARM.
After erecting the outer stage, the field at the inner walls was much smaller than the original
ambient field. By sweeping the coil over the inner shield we could impose a new weaker
ARM, and therefore reduce the remanence contributed by the inner shield. This new ARM
could be weakened in steps with increasingly larger AFs. This would realign effectively
the lowest coercivity domains first, and gradually involve more and more of the shielding
material (Fig. 11).
In addition to remagnetizing the inner shield with progressively higher AFs (Fig. 11),
the outer shield was remagnetized at the full coil strength of 12.5 mT after each interior
step. At alternating fields of 2 and 3 mT, we repeatedly realigned the interior and then
the exterior walls. This showed that it was possible to converge upon a single interior field
as a general equilibrium was reached between the shielding stages. This also demonstrates
that remanent magnetization of the electrical steel has a wide coercivity spectrum. This
experiment shows that shielding cannot be improved indefinitely by using domain
realignment to increase the effective permeability of a material. As the ambient field be-
comes weaker around the innermost shield, the alignment process becomes less effective.
This apparently limits the shielding that is possible. For the case of a two-stage enclosure,
the empirical shielding limit seems to be about 50 nT (Table III) regardless of whether iron
or high-permeability alloys are used. This corresponds to a shielding factor of about 1000.
4.3. Caltech, Pasadena
The Caltech Biomagnetic Clean Laboratory was built in 1982, and enclosed an interior
volume of 38 m 3 • Electrical steel shielding and clean-lab features were combined to form
a facility for research concerning the magnetic properties of biogenic material. For the
interior, epoxy paint and fireproof gypsum board were applied over the electrical steel
sheets. The lab also possesses magnetically filtered positive-pressure air flow, and a pass-
through shower in the entryway to the inner shield.
In addition to the clean-room features, the design of this laboratory presented two
unusual problems. First was size; a N-S length of 6.2 m was required, significantly larger
than any previous steel room and approaching the maximum horizontal size of mu-metal
shields. The interior length was 4.2 m, with a 1.7-m-wide spacing used for the north wall.
This space incorporates the compound entryway, shower, air ducts, and service pipes. The
second problem was the service pipes, which could not be eliminated as they provided
water and sewer needs for the entire building. Allowing these pipes, which are mostly
steel or cast iron, to pass between the shielding stages, required some planning. The re-
moval of pipes that terminated near the shield eliminated the large magnetic fields as-
sociated with their ends. Only E-W-directed pipes were kept, as this orientation produced
the minimum induced magnetization. We rearranged all other pipes to run E-W or rerouted
them around and away from the shield. Wherever convenient, pipes and ducts were re-
placed by nonmagnetic materials; however, concern about expense and compatibility with
building codes kept us from changing most of the pipes. Because the remaining service
pipes and ducts were oriented in an E-W direction, they could be demagnetized effectively.
Using the same coil as for processing the shield, we produced an ARM of nearly zero along
the length of the pipes. Inside the inner shield, no significant magnetic field could be
attributed to these service pipes, lying as close as 0.5 m away.
The outer shield enclosed 80 m3 , with an internal field of about 4000 nT after initial
treatment with the sweeper coils. With the inner shield erected, but before ARM treatment,
the field averaged about 200 nT, for a shielding factor of about 215. After treating the inner
shield with sweeper coils, the field was about 80 nT. The long-term stable field in the
interior is about 120-200 nT, for a shielding factor of approximately 300.
5. Summary
By taking advantage of the properties of remanent magnetization, effective and rela-

tively inexpensive magnetic shielding can be obtained against static fields such as the
earth's magnetic field. We here described three room-sized shields constructed from two
stages of ordinary transformer steel. We specially processed the steel with large AF coils
so that its remanent magnetization was aligned favorably for shielding. Our research sug-
gests that any ferromagnetic material will provide some level of shielding. However, the
effectiveness of magnetic shielding can be enhanced by considering the magnetic history
of the material and by orienting the shield and shield access ways in certain perferred
geometries.
ACKNOWLEDGMENTS We thank Wulf Gose, D. T. A. Symon, and an anonymous reviewer for

their suggestions concerning an earlier version of the manuscript. The motivation for using
this new shielding material sprang from a seminar at the University of Texas at Dallas lead
by Bob Patton in 1974. In attendance were John Foster and one of us (G.R.S.) who later
proposed to build a two-layer shield in 1979 for Woodward-Clyde Consultants. Mike Stu-
pavsky and David Symons generously supplied information from their experience with a
single-layer shield. Duane Packer and Jeff Johnston supplied support and technical as-
sistance through Woodward-Clyde Consultants. Several craftsmen and other workers were
essential to the existence of these shielded laboratories, including Michael Rosenbaum,
George Clark, William Richter, Henry Salameh, Pamela Cross, Carol Van Alstine, and John
Sporich. This is Contribution No. 578 from the University of Texas Institute for Geophysics.
220 Chapter 8
References
Cohen, D., 1967, A shielded facility for low-level magnetic measurements, J. Appl. Phys. 38:1295-
1296.
Cohen, D., Edelsack, E. A., and Zimmermann, J. E., 1970, Magnetcardiograms taken inside a shielded
room with a superconducting magnetometer, Appl. Phys. Lett. 16:278-282.
Cravath, A. M., 1957, Magnetic shielding with multiple cylindrical shells, Rev. Sci. Instrum. 28:659.
Erne, S. N., Hahlbohm, H. D., Scheer, H., and Trontelj, Z., 1981, The Berlin magnetically shielded
room: Section B: Performances, in: Biomagnetism: Proceedings Third International Workshop on
Biomagnetism (S. N. Erne, H. D. Hahlbohm, and H. Lubbig, eds.l, de Gruyter, Berlin, pp. 78-88.
Jackson, J. D., 1962, Classical Electrodynamics, Wiley, New York.
Kelha, V. 0., 1981, Construction and performance of the Otanii magnetically shielded room, in: Biom-
agnetism: Proceedings Third International Workshop on Biomagnetism (S. N. Erne, H. D. Hahl-
bohm, and H. Lubbig, eds.l, de Gruyter, Berlin, pp. 33-50.
Kittel, C., 1966, Introduction to Solid State Physics, Wiley, New York.
Mager, A., 1981, The Berlin magnetically shielded room: Section A: Design and construction, in:
Biomagnetism: Proceedings Third International Workshop on Biomagnetism (S. N. Erne, H. D.
Hahlbohm, and H. Lubbig, eds.l, de Gruyter, Berlin, pp. 51-78.
Patton, B. J., 1967, Magnetic shielding, in: Methods in Paleomagnetism (K. M. Creer and S. K. Runcorn,
eds.l, Elsevier, Amsterdam, pp. 569-588.
Patton, B. J., and Fitch, J. L., 1962, Design of a room-size magnetic shield, J. Geophys. Res. 67:1117-
1121.
Schwiezer, F., 1962, Magnetic shielding factors of a system of concentric spherical shells, J. Appl.
Phys. 33:1001-1003.
Scott, G. R., and Frohlich, C., 1980, Constructing a magnetically shielded room with transformer steel.
EOS 61:942.
Symons, D. T. A., and Stupavsky, M., 1983, A low cost magnetically shielded room for paleomagnetic
research, EOS 64:220.
Thomas, A. K., 1968, Magnetic shielded enclosure design in the DC and VLF region, IEEE Trans.
Electromagn. Com pat. EMC-l0:142-152.
Wadley, W. G., 1956, Magnetic shielding with multiple cylindrical shells, Rev. Sci. Instrum. 27:910-
916.
Weast, R. C. (ed.), 1973, Handbook of Chemistry and Physics, 54th ed., CRC Press, Cleveland.
Wikswo, J. P., 1975, Noninvasive magnetic measurements of the electrical and mechanical activity of
the heart, Ph.D. thesis, Stanford University.
Wills, A. P., 1899, On the magnetic shielding effect of trilamellar spherical and cylindrical shells,
Phys. Rev. 9:193-213.
III
Magnetoreception: Theoretical
Considerations
The hypothesis of magnetite-based magnetoreception is but one of many which have been
proposed during the past century, yet it has only been in the past few years that it has
been considered a serious explanation for magnetoreception in animals. This new status
has been gained in part through the widespread discovery of biogenic magnetite in animal
tissues, by the increasing number of animals known to possess geomagnetic sensitivity,
and by the recognition that magnetite is indeed the basis for the magnetotactic response
of bacteria. This bacterial example shows clearly that organisms have a very simple solution
to the problem of detecting magnetic field direction; indeed, only a few magnetosome-like
objects coupled to hair cells could account for all of the directional sensitivity exhibited
by organisms. However, there are still large gaps in our understanding of how a magnetite-
based sensory system would work in animals, and this is made more complex in attempts
to explain apparent sensitivities to small geomagnetic fluctuations and local magnetic
anomalies (on the order of 0.1% changes in the background field strength).
The first three chapters in this section take different approaches to this sensitivity
problem. Rosenblum, Jungerman, and Longfellow critically examine the hypothesis of in-
duction-based magnetoreception with the goal of discovering whether or not any known
form of electroreception could achieve high enough sensitivity. Yorke considers the en-
ergetics required for a magnetite-based sensory organ to achieve high resolution, while
Kirschvink and Walker explore the implications of one possible mechanism for large num-
bers of magnetite-based organelles to achieve this sensitivity, and make predictions-based
on natural selection-concerning the particle sizes for these structures. Finally, Gould
explores much of the behavioral evidence which suggests that minute features in the geo-
magnetic field may indeed playa role in the map sense of organisms.
221
Chapter 9
Limits to Induction-Based
Magnetoreception
BRUCE ROSENBLUM, ROGER L. JUNGERMAN,
and LAURENT LONGFELLOW
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
2. Noise and General Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
3. The Induction Magnetoreception Organ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
4. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
5. Addendum: A Comment on Navigation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
1. Introduction
While behavioral experiments apparently have established that several species perceive
magnetic fields as small as, or perhaps even substantially smaller than, the geomagnetic
field of the earth, in few species is an organ responsible for magnetoreception clearly
identified. Magnetite crystallites in bacteria and the ampullae of Lorenzini of the elas-
mobranch fish seem the only clear cases. Magnetoreception may also be the only perceptual
sense found in animals for which a similar sense is not manifest in humans.
A behavioral response without an obvious, corresponding sense organ presents a
unique challenge. Such a large number of species are reported to exhibit magnetoreception
that one suspects magnetoreception to be a quite general phenomenon. Grounds for such
suspicions-and clues to the nature of the sense organ-include the magnetic crystallites
widely found in amounts theoretically permitting magnetoreception, and without other
obvious function. Magnetoreception without magnetic material, by the use of induction
as in the elasmobranchs, also turns out to be a possibility for air-ambient animals. The
magnetoreception problem seems important, and many hints for solution are available.
The present paper investigates the possible form of magnetoreception organs based
on magnetic induction by considering the limitations imposed by noise. In other cases,
where one knows characteristics such as the size and form of a sense organ, one can cal-
culate noise limitations to ultimate sensitivity thresholds. Experience with different senses
in animals indicates that nature is often highly efficient in design, and in many cases organs
frequently perform close to these theoretical limits. In the present case, one does not know
the nature of the organ, but one does have empirical information on the detection thresh-
olds. We therefore turn the previous procedure around and attempt to deduce possible
BRUCE ROSENBLUM, ROGER 1. JUNGERMAN, and LAURENT LONGFELLOW • Department of

Physics, University of California, Santa Cruz, California 95064. Present address of RL.J.: Department
of Applied Physics. Stanford University. Stanford. California 94305.
223
224 Chapter 9
forms of the organ from the observed sensitivity and general biological and species-specific
constraints.
We will first review some general aspects of the noise limits to detection sensitivity
and then justify a narrowing down to thermal noise in a primary transducer section. We
finally specialize for a magnetoreception organ using Faraday induction.
2. Noise and General Considerations

A sense organ perceives an aspect of the environment, e.g., the strength of the ambient
magnetic field, and transmits that information in the form of an electrical signal to higher
neural levels for processing, response, and storage. The "noise" we consider is a signal
simultaneously transmitted along the same pathways which is indistinguishable from a
possible signal representative of the environment. The higher levels, e.g., central nervous
system, may correlate information received with data from other receptors, or with pre-
viously stored information, and might average over long time periods. If it is sufficiently
sophisticated, there is probably no simple and significant limit to sensitivity.
We will initially assume a situation without complex correlations or long averaging.
This assumption, while not invariably true, is a good approximation for many simple
threshold effects, such as the perception of weak sound and light. The minimum detectable
signal would then be of the order of the noise, whose character we will discuss. As we
wish to consider the limitations on the nature of the sense organ imposed by noise, we
are interested in noise intrinsic to the receptor. We therefore ignore aspects of the envi-
ronment ("environmental noise") which may "fool" the receptor.
Where in a magnetoreception organ does the intrinsic noise we wish to treat arise?
For some sense organs, and conceivable sense organs, it is feasible to consider the organ
as composed of a primary transducer which converts the environmental signal into a form
appropriate for processing in a distinct subsequent section, which then transmits the in-
formation to, say, the central nervous system. The ear is an example of this, while the eye,
on the other hand, is more a single unit, indeed, almost an extension of the central nervous
system. The fundamental reason for difference here, and in other cases, can be considered
the size of the quantum of energy involved. The optical quantum is large enough to directly
trigger a nerve impulse by an electrochemical process. The ear, responding to quanta five
orders of magnitude smaller, must respond with a primary structure-a macroscopic, non-
quantum, i.e., classical, transducer-between the nerve and the environmental signal.
Magnetic fields might conceivably affect neurons electrically or electrochemically in
a number of exotic ways that have been suggested. However, the quanta involved in the
interaction of the geomagnetic field at the molecular level are extremely small, and direct
neural stimulation processes seem most unlikely. A macroscopic primary transducer would
be expected a priori. In fact, the two magnetoreceptors which have been found in nature
(bacteria and elasmobranch fish) both utilize such macroscopic transducers. These two
mechanisms, torques on magnetic material and Faraday induction, are also the simplest
and most obvious ways of physically detecting magnetic fields.
We wish to deduce the nature of this macroscopic primary transducer for inductive
magnetoreception from noise and sensitivity considerations. The mechanism involved in
the secondary processing of the information output of the macroscopic transducer is likely
similar to other neuronal processes. We will assume that the latter operates at a level close
to the theoretical noise limit, as is often true in other senses. We therefore neglect any
noise introduced at later stages and focus attention on the primary transducer.
As a biological system, the primary transducer can in principle produce systematic
and random noise in ways too complex to evaluate. However, ever-present thermal noise
Limits to Induction-Based Magnetoreception 225
will set constraints which cannot be avoided, and the limits of efficient signal detection
systems, biological, electronic, and mechanical, are often so determined. We assume this
to be the present case. Because the primary magnetic transducer is in thermal contact with
the environment, whatever property is affected by the ambient magnetic field will also
vary due to thermal buffeting. For a variation in that property to be recognized as due to
the ambient field, it must exceed any qualitatively indistinguishable random thermal var-
iations.
Before narrowing our focus to induction, we will first contrast that mechanism with
one based on magnetic materials. The latter has been treated more extensively, and is the
subject of the following chapter (Yorke, this volume).
Magnetic materials for magnetoreception must almost necessarily be spontaneous mag-
netic dipoles in the form of submicrometer crystallites with a high content of iron or other
transition element. These crystallites must also somehow be innervated. The morphology,
size, and composition of an induction-based receptor is less defined than one based on
magnetic material. Before attempting to elucidate just these aspects of an induction organ,
we contrast the two.
A magnetic field exerts a torque on the magnetic crystallite, the primary transducer,
and the consequent stress (pressure) or strain (rotational motion) somehow creates a neural
electrical signal. For induction, the output of the primary transducer organ is immediately
electrical. For the magnetic crystallite, the torque exists without any behavior required of
the animal, while for induction, an appropriate motion is required to produce a signal.
The sensitivity for detection based on magnetic dipoles is calculated by determining
the probable error due to thermal agitation at any given time in the orientation of a dipole.
In the case of induction, one compares the thermally generated voltage to that generated
by a movement of the animal in the ambient magnetic field. In both cases, repeated ob-
servation by the animal of its magnetoreception signal or the averaging of a number of
receptors could, in principle, increase the sensitivity.
While it has not been established that magnetite crystallites are, in fact, magnetore-
ception organs in animals other than bacteria, it has already been shown that such materials
could be used to obtain the reported sensitivity to the earth's field and even the speculated
higher sensitivities. In the following section we explore the extent to which a magneto-
receptor based on induction could account for such phenomena and deduce parameters
such an organ would require.
3. The Induction Magnetoreception Organ

Certain elasmobranch fish sense magnetic fields as small as the earth's (Kalmijn, 1982),
presumably by induction. These fish (sharks and rays) detect electric fields by means of
conductive canals (the ampullary canals of Lorenzini) connecting pores on the skin with
sensitive voltage-sensing cells (electroreceptors) deep in the fish. Because some fish re-
spond to fields less than 1 ""Vim (Kalmijn, 1982), they need move with velocities of only
a few centimeters per second to generate an electromotive force (emf) large enough to sense
the earth's magnetic field with their electrodetection system.
Birds, or other air-ambient animals, could not accomplish magnetoreception by in-
duction in the same way as fish. In fish, the electric circuit consists of the ampullary canal,
the electroreceptor, and the external medium (seawater). Because the external section of
the path has comparatively low resistance, a substantial part of the induced emf appears
usefully across the electroreceptor. Increasing the resistance of the external path has, in
fact, been shown to decrease the fish's sensitivity (Rommel and McCleave, 1973). A similar
detection system in air would have essentially all of the emf across the air section and
226 Chapter 9
virtually none across the electroreceptor. For magnetic detection by induction in air, the
circuit must be closed within the animal.
For a circuit completely within the animal, there will be no induced emf if the animal
merely translates through a spatially uniform magnetic field, thus producing no magnetic
flux change through the circuit. The required flux change exists only if the circuit is rotated
or distorted to change its area.
The simplest inductive structure to consider is a coil of insulated conductor which
can be rotated in the magnetic field and whose ends are connected to an electroreceptor.
Other geometries would produce similar or poorer results, at least from electromagnetic
considerations. We will therefore treat the inductive structure as a circular coil with, per-
haps, a number of turns and formed of biological materials.
The induction voltages produced by the rotation of the coil will be small and a sensitive
electroreceptor will be required. Probably the most sensitive biological electroreceptor
known is that of the elasmobranchs, and it could be adequate for magnetoreception by
induction. However, while a sensitive electroreceptor is necessary, it is not sufficient.
Another necessary condition is that the signal power supplied by the coil to the electro-
receptor must exceed the indistinguishable part of the thermally generated noise power
also supplied. This thermal noise condition may well set the actual detection sensitivity
threshold. It surely sets constraints on the nature and form of the magnetoreception coil.
The charge carriers in the conducting material of the induction coil, thermally vibrate
at essentially all frequencies. This vibration causes a random voltage across the coil, and
noise power is therefore transferred to the electroreceptor. (The like power transferred in
the reverse direction need not concern us.) By noting that any system has, in thermal
equilibrium, an energy kT in each possible mode of vibration, where k is Boltzmann's
constant and T the absolute temperature, one can show that the effective rms noise voltage
appearing across our coil (of negligible inductance) is (see, e.g., Bennett, 1960)
(1)
R is the resistance of the coil, and AI is the range of frequencies to which the electroreceptor
is "sensitive," a point to be discussed when a value for ú Ñ =is considered.
The signal voltage, which must be recognized as such in the presence of the above
noise voltage, is created, we can assume, by a rotation of the coil. This could be by a 90°
movement of the animal changing the component of the earth's field perpendicular to the
plane of the coil from maximum to zero. The voltage so induced is given by
dB
V = n-'TTr2 (2)
• dt
where n is the number of turns, dB/dt is the rate of change of the component of magnetic
field perpendicular to the plane of the coil, and r is the radius of the circular coil.
With the 90° rotation of the coil assumed to be a smooth angular acceleration and
deceleration in a time T, V. will be a rounded pulse of approximate width T and amplitude
V. = nB'TTrZT- 1. We define T only roughly, but the approximate factor of two uncertainty
is in the spirit of the calculation.
In order to fully respond to this signal voltage, the electroreceptor (and neural mech-
anisms further along the perceptual path) must be sensitive to the range of frequencies AI
= (2'TTT)-1. A larger range of frequency response would allow finer discrimination of the
detailed form of V. for possible correlation with the perception of angular acceleration.
However, a larger frequency response range would also mean a larger effective noise volt-
age. High-sensitivity systems must be responsive to only the frequency components car-
rying the most significant information. We therefore set AI of Eq. (1) equal to (2'TTT)-1. The
perceptual level at which the frequency range is limited is immaterial; it could be in the
primary receptor or even in the level of "consciousness."
When the rms noise is equal to the anticipated signal voltage, there is approximately
a 50% probability that in any interval T there will be a voltage excursion indistinguishable
from a true signal. It is then not possible to respond to a signal with any confidence. One
can therefore use Vs = Vn as a criterion for the minimum perceptible signal.
Two considerations make this criterion excessively strict. The first is that the animal
can, in principle, correlate a conscious rotation of its induction coil with the signal and
reject voltage components not corresponding in phase with the rotation. However, with
the already optimized bandwidth to admit only the major signal components, this phase
criterion could only be used to reduce the effective noise by a factor of the order of 2,
which we will ignore.
A potentially more significant improvement could be achieved by taking several ob-
servations and averaging the results. In principle, this could reduce the effective noise by
the square root of the number of observations. Such an enhancement is, of course, possible
for any sensory system. Most animals, however, do not appear to increase the sensitivity
of their perception organs significantly over what would correspond to a single observation.
This can be a complex issue, but in the present case many observations would require
many rotations, which could presumably be noticed. In any event, this factor is readily
included as a modification of the final result at the end of the calculation.
To determine the magnetic field to an accuracy ú _ L _ = = q requires VsIVn = q. We will
proceed to calculate the sensitivity for a single observation by comparing Vs to Vn and
seeing what constraints this imposes on an inductive magnetic perception organ.
However, before considering air-ambient animals, let us check our analysis for an
elasmobranch, the ray, where sufficient data is available. The minimum electric field ob-
served to influence the behavior in the ray is 10- 5 Vim (Murray, 1965). For a 1-cm am-
pullary canal, this corresponds to a Vs = 1 X 10- 7 V. For the electric field perception to
be behaviorally useful, the ray should respond to this field with a "listening" time of not
much more than to of a second, yielding ú Ñ =about 2 Hz. With a resistance of the ampullary
canal of the ray of 3 x 104 ohms (Waltman, 1966), Eq. (1) gives a noise voltage of 3 x
10- 8 V. The signal voltage is, therefore, three times the thermal noise at the level at which
a behavioral response is first observed. This is not only consistent with our formulation,
but also suggests that thermal noise ã ú ó =actually set the detection threshold. *
To proceed with our treatment of air-ambient animals, we first write the resistance of
the coil in terms of more interesting and/or tractable parameters.
(3)
Here, p is the resistivity of the material of the conductor of the coil, d is the diameter of
the conductor, r is the radius of the loop of the coil, and n is the number of turns.
We can therefore write the "signal-to-noise" ratio as
(4)
The quantities in front of the square brackets are unrestricted at this point. The quantities
within the brackets can be quite readily estimated.
To be specific, we consider a medium-sized bird and assume the magnetic detector
is in the bird's head. Other situations are straightforward modifications. We assume that
* The ray actually has perhaps 50 ampullae. For a given field direction, these could, in principle, be
averaged to yield a theoretical signal-to-noise of almost 10 when the ray's first behavioral response
is observed.
228 Chapter 9
N Mú J J J J ú J J J J J J ú ú J J J J J J ú J J J J J J ú ú J J J J J J J J J J J J J J J J ú =
E
E 1.0 V=300q2 n"l q 2
ú =
n=IOq2
V=3000q2
l Kf ú J J J J J J J J J J J J J J ú ú J J J J J J J J J J J J J J ú J J J J J J J J J J J J J J ú =
0.01 0.1 1.0 10
d (mm)
Figure 1. Lines of constant minimum volume Vrn (mm 3 ). resistance R (ohms), and number of turns n
are plotted with signal-to-noise ratio q as a parameter on coordinates of coil radius r and conductor
diameter d (mm). The acceptable values for rand d for an induction detector are limited to the shaded
trapezoid.
the flux of the earth's 5 x 10- 5 T field through the coil is changed from zero to maximum
by a rotation of the bird's head by 90° in 0.1 sec. For the detection system to respond
efficiently to this voltage pulse, ilf = 2 Hz. The smallest resistivity we are aware of for
an organic biological material is that of the ampullary canals of the elasmobranchs, which
is 25 n cm (Waltman, 1966). This is less than an order of magnitude lower than other
fluids in animals, and we therefore take 50 n cm as a reasonable estimate for a good
biological conductor. (We treat the case of magnetite as a conductor below.) For T = 300°K,
kT = 4.2 x 10- 21 JrK. Substituting these numbers, we get
(5)
where we use mks units in all equations.

Equation (5) contains three independent parameters, which would seem to allow con-
siderable latitude in the nature of the magnetic detector. However, physical constraints
and some reasonable assumptions allow us to substantially narrow the range of these pa-
rameters. The attempt is to define as closely as we can the form an induction detector in
an animal must take.
On the log-log plot of Fig. 1, wide ranges of rand d are displayed. The heavy horizontal
line at the top indicates an upper limit of r = 5 mm if the detector is fit in a 1-cm bird
skull. The heavy 45° line, r = d, is the boundary of the physical region as r must be greater
than d.
The volume actually occupied by the conducting material making up the loops is
approximately given by
(6)
With Eq. (5), we eliminate n from Eq. (6) to get an equation for the minimum volume
required by the conducting material for a given q:
(7)
where we continue to use mks units. It does seem strange to see the volume depending
only on r to the inverse square, but this is for variable n and fixed q. In Fig. 1, we have
drawn lines of constant volume.
As the volume of the bird skull is of the order of 1000 mm 3 , we see that even for unity
signal-to-noise ratio, a significant fraction of the brain volume would be occupied by the
coil for small r. Because detection of magnetic fields does not seem to be an extremely
important function for the animal, it would probably not warrant a sense organ much larger
in size than those for sight or hearing. The region low r = 1 mm (volume greater than 300
mm3 ) is therefore conservatively excludable.
We have so far restricted rand d to the region bounded by the lines r = 5 mm, r =
d, and r = 1 mm. This region is still open to the left toward smaller d. We will argue that
both the large number of turns required and the high coil resistance make a system with
very small d unlikely.
From Eq. (5), we draw in Fig. 1 the family of lines indexed with n. We also eliminate
n from Eq. (3) with Eq. (4) and draw the family of lines for constant resistance R for given
q. For a given signal-to-noise ratio q, both the number of turns required and the resistance
of the coil increase rapidly as we move to smaller d. Let us, for now, consider q = 1,
probably the minimum value for a useful sense organ.
For a decreasing d which is still not very small, we leave the attractively simple and
compact possibility of n = 1. For much smaller d, the structure must become very non-
compact (r ú =d) and the required complexity increases rapidly (n - d- Z ). A large number
of turns also requires a considerable volume of insulating material, a factor we did not
include when we calculated the volume. These considerations provide a qualitative ar-
gument for limiting d to perhaps well above 0.1 mm. More quantitative arguments arise
when we consider the coil resistance.
We note from the constant resistance lines in Fig. 1 that R increases extremely rapidly
with decreasing d (R - d- 4 ). Resistances of the order lOB n might be a reasonable upper
limit on the basis of available insulating material. Actually, there is a more restrictive
upper limit.
Both the signal power and the thermal noise power delivered by the coil to an elec-
troreceptor depend on the internal resistance of the electroreceptor in the same way and
thus leave this signal-to-noise ratio unaffected by the electroreceptor resistance. However,
there is always some detection threshold or noise source intrinsic to the electroreceptor.
The transferred signal power must be greater than this noise as well as thermal noise. If
the coil resistance were much larger than the electroreceptor input resistance, the signal
power transferred to the detector would decrease as R-z. Therefore, d must be bounded
on the left by a line corresponding to a resistance not vastly greater than the input resistance
of the electroreceptor.
The constraints on the input resistance of a highly sensitive electroreceptor are not
clear to us. Large synaptic areas and the need to insulate from the surrounding environment
may be important factors. The most sensitive electroreceptor we know of is in certain
230 Chapter 9
elasmobranch fish, and we might suppose that other species employ a similar design.
Unfortunately, we know of no good measurement of the input resistance of the elasmo-
branch electroreceptor. There is, however, one measurement (Murray, 1965) from which
we can infer (Jungerman and Rosenblum, 1980) a resistance of the order of 10 5 n. To be
somewhat conservative, we take the R = 106 n line as the lower bound for d.
We have thus restricted and acceptable rand d for a magnetic induction detection
system to the shaded trapezoid in Fig. 1. We conclude that an induction magnetoreceptor
requires a coil with radius and conductor diameter within a factor of about three of r =
3 mm, and d = 1 mm. The coil should have only one, or at most a few turns. The conductor
material must have low resistance, be within an insulating membrane, and be innervated.
If such a macroscopic structure existed, it could hardly have escaped observation.
Parts of the circulatory system cannot be ruled out, but it is presumably not properly
innervated. A conclusion of our analysis might seem that a magnetic detection system
using induction is unlikely in animals in air. However, before dismissing a detector based
on induction, we note that there is, in fact, an organ which possibly fits our requirements.
The semicircular canals in the labyrinth of the inner ear are membranes of just the proper
dimensions filled with conducting endolymph. The heavily innervated cristae are can-
didates for electroreceptors. The cristae and the associated gelatinous mass holding the
hairs effectively close off the conducting path in the semicircular canal by pressing against
the walls of the ampulla (Kornhuber, 1974). This should be sufficiently insulating, as the
resistance of the gelatinous mass is likely to be substantially greater than that of the en-
dolymph. It is only when the head is rotated about an axis perpendicular to the plane of
a particular semicircular canal that the hairs and the gelatinous mass are displaced by the
flowing endolymph. A rotation about an axis in the plane of the canals is what is necessary
to induce a voltage by magnetic induction, and such a rotation causes no fluid flow. Other
circuits involving the labyrinth are also conceivable. There is a certain appeal to the organ
for orientation in the magnetic field being the same as the organ for gravitational orien-
tation, especially so if the angle between the magnetic and gravitational fields is the rel-
evant parameter in navigation (Wiltschko and Wiltschko, 1972).
Magnetite as a conductor: In our discussion of parameters for an induction coil, we
chose the resistivity of the conducting material to be 50 n em, high for any "normal"
biogenic material. Because magnetite now appears to be anabolized in a wide range of
species, we recalculate our results for conducting material with resistivity of 5 x 10- 3 n
em, that of magnetite. The conclusion, in summary, is that if the coil were made of a
material with the conductivity of magnetite, the dimensions could be smaller than pre-
viously stated by about an order of magnitude. A loop of magnetite with dimensions of
the order of a few tenths of a millimeter is a less attractive possibility than previous spec-
ulations. It becomes even less plausible when we apply the volume condition, Eq. (7). The
required volume of magnetite would be orders of magnitude greater than the amount re-
ported to be found in animals.
Because magnetite appears to be dispersed as sub micrometer crystallites, is it con-
ceivable that each of N of these cells could somehow function as a tiny induction-based
detector? Each would provide a signal-to-noise ratio q much less than unity, but the av-
eraged set of detectors would have the q of the set of N increased by N 1I2 over that of a
single detector and might thus yield q greater than unity. This possibility, however, violates
the constraint set by the limited amount of magnetite found in animals even more strongly
than does the assumption of a single magnetite coil. (Similar considerations hold for any
system of induction-based magnetoreceptors, each with q <Ii 1.) Furthermore, in the av-
eraging of an increasingly large number of any type of receptors, each with q <Ii 1, increas-
ingly lower threshold sensitivities and lower limits on correlated noise must be demanded
of the neural sensors associated with the primary receptor organs. This must ultimately
set limits on the numbers of receptors which can be separately sensed and averaged.
Experiments to distinguish between ferromagnetic- and induction-based magnetore-

ception are not trivial. Some aspects have been briefly outlined (Jungerman and Rosenblum,
1980).
4. Conclusion
A magnetoreceptor based on Faraday induction and not using magnetite is a possibility

in air ambient animals as small as birds. To be sensitive enough to detect the earth's field,
such an organ would need dimensions of the order of millimeters. The semicircular canals
appear candidates for such organs. In much smaller animals where the organs described
cannot exist, induction-based magnetoreception is ruled out.
To explain the "map sense" required by homing in pigeons, it has been speculated
that they may be able to detect variations in the earth's field to perhaps one part in 103
or 104 (Gould, Chapter 12, this volume). While within the range of possibility for magnetite-
based magnetoreceptor systems, this sensitivity appears greater than an induction-based
system could accomplish.
5. Addendum: A Comment on Navigation
Animal navigation data in general and homing in pigeons in particular implies the
animal must "know" the direction of the displacement vector to its goal. An orientation,
with respect to the earth's magnetic field, for example, is not sufficient. The suggestion
that pigeons in homing make use of a "map sense" based on spatial variations in the earth's
magnetic field has been plausibly and critically developed (Gould, Chapter 12, this vol-
ume). As pointed out above, inductin-based magnetoreception could not provide the sen-
sitivity needed for this ability. Induction cannot, however, be excluded from a role in a
different mechanism, one based in part on inertial navigation (a double integration of
acceleration). While not all published experiments accord well with an explanation based
on inertial navigation, none can rule it out, and all other suggested explanations have
problems as well.
It has been pointed out that as the largest errors accrue in the angular aspect, a coupling
of a compass orientation knowledge with an ability to integrate acceleration could con-
siderably relax the requirements on an inertial navigation system. An extremely good ac-
celeration sense would still be required, but the requirement is easily within theoretical
limits. We note that a good acceleration sense is also required for resolving the components
of the magnetic field for a map sense based on spatial variation of the earth's field. As
Gould points out, the accurate sensing of the direction of gravity-to one part in 104-is
needed here. This would necessarily imply the same accuracy for acceleration determi-
nation by the principle equivalence of acceleration and gravitational field.
We will shortly publish a noise and error analysis and computer model of a combined
inertial/compass-based navigation system.
References
Bennett, W. R, 1960, Electrical Noise, McGraw-Hill, New York.
Jungerman, R L., and Rosenblum, B., 1980, Magnetic induction for the sensing of magnetic fields-
An analysis, J. Theor. BioI. 87:25.
Kalmijn, A. J., 1982, Electric and magnetic field detection in elasmobranch fishes, Science 218:916.
232 Chapter 9
Lowenstein, O. E., 1974, Comparative morphology and physiology, Chapter II, in: Handbook of Sensory
Physiology, volume VIII, (Kornhuber, H. H., ed.), Springer-Verlag, Berlin, p. 77.
Murray, R W., 1965, Electroreceptor mechanisms: Relations of impulse frequency to stimulus strength
and responses to pulsed stimuli in ampullae of Lorenzini of elasmobranchs, J. Physiol. (London)
180:592.
Rommel, S. A., Jr., and McCleave, J. D., 1973, Sensitivity of American eels (Anguilla rostrata) and
Atlantic salmon (Salmo-salar) to weak electric and magnetic fields, J. Fish. Res. Board Can. 30:657.
Waltman, B. 1966, Electrical properties and fine structure of ampullary canals of Lorenzini, Acta
Physiol. Scand. 66(Suppl):264.
Wiltschko, W., and Wiltschko, R, 1972, Magnetic compass of European robins, Science 176:62.
Chapter 10
Energetics and Sensitivity
Considerations of Ferromagnetic
Magnetoreceptors
ELLEN D. YORKE
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . 233
2. Energy Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . 234
3. Response Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . 235
4. Sensitivity to Field Changes. . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . 238
5. Other Types of Receptors. . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . 239
6. Tests of the Hypothesis. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . 239
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . 241
1. Introduction
The suggestion that permanent magnets of biological origin are responsible for the sen-
sitivity of migratory animals to earth-strength magnetic fields predates both the discovery
of biogenic magnetite in many species and the generation of convincing behavioral evi-
dence for the existence of a magnetic sense. In 1945. Ising unsuccessfully searched for
remanence in the head of a swallow, and Griffin (1944) noted the absence of any known
magnetic material of biological origin in a review of bird navigation. Lowenstam (1962)
suggested that magnetite might be responsible for homing behavior in chitons. and Keeton
(1972) suggested that the magnetic sensitivity of birds might be the result of the "actual
mechanical displacement of some structure," though he did not speculate on its compo-
sition. Blakemore's (1975) discovery of magnetite-containing magnetotactic bacteria fol-
lowed the emergence of strong behavioral evidence that birds (Keeton, 1971; Wiltschko
and Wiltschko, 1972; Walcott and Green, 1974) and bees (Lindauer and Martin, 1968. 1972)
are sensitive to the earth's magnetic field. This gave impetus to the serious, quantitative
study of the conjecture that permanently magnetic material is the transducer of the mag-
netic sense (Gould et aI., 1978; Walcott et aI., 1979; Yorke, 1979; Kirschvink and Gould,
1981; reviewed by Kirschvink, 1982).
We shall examine in detail some of the arguments which support this conjecture. At
present, magnetite is known to be widely distributed among living organisms but its role
in magnetic sensing is not yet experimentally proven and a great deal of work remains to
be done.
ELLEN D. YORKE • Department of Physics, University of Maryland Baltimore County, Catonsville,

Maryland 21228.
233
234 Chapter 10
2. Energy Considerations
One way that permanently magnetic material can respond to a magnetic field is to
rotate into alignment with the field in the manner of a compass needle or a dip needle.
For a structure of organelle size-a micrometer or less-in a field of 0.5 G, thermal (Brown-
ian) motion can be important in degrading the quality of the alignment. A thin rod in an
environment of absolute temperature T will rotate randomly with an average kinetic energy
of ikT (k = Boltzmann's constant = 1.38 x 10- 16 erg/oK) for each of its two degrees of
freedom; at room temperature the energy associated with this random thermal motion is
4.14 x 10- 14 erg. In the presence of a magnetic field B, a rod with a magnetic moment f..l.
oriented at an angle 0 with the magnetic field has a potential energy of - f..l.B cos O. The
rod will be well aligned if f..l.B exceeds kT. For a single-domain magnetite grain (saturation
magnetization M = 480 gauss, length L, axial ratio a) in a field of 0.5 G, the condition f..l.B
"" kT implies that a grain of length 0.1 f..l.m and axial ratio exceeding 0.42 will achieve
reasonable orientation. This is well within the single-domain regime for magnetite and is
representative of the sizes of magnetite crystals that have been found in pigeons (Walcott
et al., 1979), bacteria (Blakemore, 1975), and bees (Gould et al., 1978). Improvement in
orientation of a single rod can be achieved by increasing the magnetic moment either by
increasing the volume of the grain while remaining in the single-domain regime or by
using geometric arrangements of single domains (such as chains) in which the magnetic
moments of single domains add and orient together as a unit. The latter is the strategy
adopted by magnetotactic bacteria. Another strategy which may be used by higher animals
is to average over the response of many independent units. We will discuss this later as
a possible technique for the perception of very small changes in magnetic field.
The competition between the alignment of noninteracting magnetic dipoles by a mag-
netic field and the disruption of this alignment by thermal motion is rigorously described
by the Langevin theory of paramagnetism, which is thoroughly discussed in many physics
books (Reif, 1965; Reitz et al., 1979). The degree of orientation of the moments is measured
by the average value of the projection of a magnetic moment on the direction of the magnetic
field (which is proportional to the magnetization of an assembly of dipoles). This is given
by the Langevin function
(cos 0) = L (f..l.BlkT) = coth (f..l.BlkT) - kTlf..l.B (1)
which is plotted in Fig. 1. It has been pointed out (Frankel and Blakemore, 1980; Kirsch-
vink, 1981) that when f..l.BlkT exceeds 6, alignment cannot be greatly improved and eco-
nomical use of material would militate against an organism's use of very large clumps of
magnetite. It is also clear from Fig. 1 that the alignment is very poor if f..l.BlkT s; 1.
As the Langevin theory assumes that individual dipoles do not interact with each
other, it is of interest to estimate the average field at the location of one dipole due to its
neighbors and compare it with that of the earth. Eldridge (1961) has shown that the average
interaction field in an unmagnetized assembly of point dipoles having a density D of
randomly placed dipoles per unit volume is 2.9 Df..l..
Recent work (Kuterbach et aI., 1982) on bees allows an estimate of the unknown den-
sity for bees. These authors have found granules of nonmagnetic compounds of iron in
bees; these granules occupy a fraction f1 "" 0.07, of the volume of cells analyzed, so if Vo
is the volume of a single granule, there are f1VIVo granules. They estimate that if only
0.33% of these granules were reduced to Fe304, the remanence reported by Gould et al.
(1978) could be accounted for. Thus, the density, D, of magnetite grains of volume Vo
would be given by 3.3 x 10- 3 f1IVo. Because f..l. = 480 Vo for magnetite, the estimated
average field of interaction of a dipole with its neighbor is 0.32 G which is fully comparable
to the earth's field. Thus, despite the computational simplicity of models based on non-
Ferromagnetic Magnetoreceptors 235
UP
S---------+---------N
Figure 1. The Langevin function, the

average projection of a magnetic mo-
ment on a magnetic field in the pres- DOWN
ence of thermal motion.
interacting grains of magnetite, it might be well to view them with caution, while using
their easily obtainable results as guideposts until further experimental results are available.
3. Response Times
In one simple theoretical model of a magnetoreceptor, a piece of magnetite, well
aligned by the earth's field, is able to rotate quite freely; its motion is opposed only by
the viscous drag of the medium in which it is embedded. As the animal moves around,
the magnetite rotates and tracks the field, changing its orientation relative to some nearby
physiological structure and thereby causing signals which are processed by the animal's
nervous system. For example, Kirschvink and Gould (1980) propose a "membrane short"
model with a magnetite grain embedded in an insulating membrane. As magnetite is highly
conducting, the surrounding region can be polarized or depolarized depending on the
orientation of the magnetite relative to the surface of the membrane and whether it is
unexposed or partially exposed to the fluid outside the membrane. Another model is pat-
terned after the gravity-sensing otoliths existing in many species. The rotation of a mag-
netite grain distorts surrounding cilia and the pattern and amount of distortion reflects
the orientation of the grain and thus the line along which the magnetic field vector lies.
H a freely rotating model can make any sense, it is essential that the magnetite reorient
rapidly enough as the animal moves for the resulting information to be of use to the animal.
Providing that the torque of the surrounding viscous material is proportional to the angular
velocity of the grain, the equation describing the grain'S motion is identical to the equation
236 Chapter 10
of motion of a damped compound pendulum (Yorke, 1979). Let I be the moment of inertia
of the grain about an axis perpendicular to its long axis and through its center, 0 be the
angle between the grain's magnetic moment (along the grain's long axis) and the field B,
and - F dO/dt be the viscous torque opposing the motion of the grain. Here, F is a constant
which depends on the size and shape of the grain and the viscosity of the surrounding
medium. The equation of motion of the grain (rate of change of angular momentum = net
torque) is
d 20 dO
I - = - liB sin 0 - F- (2)
de'" dt
The moment of inertia of the grain is found from its density (5.1 g/cm3 for magnetite)
and its dimensions. For convenience, we list below the expressions for the coefficients I,
(.LB, and F for a grain of length L and axial ratio a in a medium of viscosity TJ and evaluate
them for a magnetite grain with L = 0.1 (.Lm, a = 0.5 in a medium with a viscosity 100
times that of water and in a field of 0.5 gauss:
I = pa 2 L5 /12 == 1.1 x 10- 26 g cm 2

f,LB = MsB a 2 L3 = 6 X 10- 14 erg
F = (2'lT/3) TJ L3/[2 In(2/a) -1] = 1.2 x 10- 15 g cm 2 /sec
For small displacements (0 < 20°), sin 0 "" 0 and Eq. (2) reduces to the equation of a damped
harmonic oscillator. For parameters typical of single-domain magnetite, the oscillator is
heavily overdamped and its motion consists of an exponential decay from its initial dis-
placed position to alignment with B. The time constant for the decay is approximately F/
f,LB "" 20 msec. The magnet will track the field direction with a similar usefully short time
scale even for larger angular displacement. If the magnetite is not free to rotate, the time
scale for response may well be usefully short but then it depends on the unknown vis-
coelastic properties of the structure to which the grain is connected and cannot be cal-
culated from the above discussion.
Magnetoreception models based upon freely rotating grains of magnetite easily account
for two observed characteristics of the magnetic sense. The magnetic sense is "axial"
(Wiltschko and Wiltschko, 1972; Emlen, 1975) except possibly in the case of salmon (Quinn
et al., 1981). That is, migratory birds, bees, and tuna (Walker et al., 1982) seem to be unable
to distinguish between magnetic fields which are directed 180° apart (B and - B). For
example, the two field directions shown in Fig. 2 are both interpreted as having horizontal
component pointing north. As the orientation of elongated magnetite grains in the two
cases will differ only in the locations of their north and south poles, but not in the ori-
entation of the grain's long axis, the effect on surrounding cellular structures (e.g., the
membrane in a "membrane short" model) should be identical.
Second, such models would predict that there should be no behavioral change noted
if attempts are made to "demagnetize" the animal by subjecting it to a strong pulsed field
or by ac demagnetization. Reversing the magnetization of a grain will not change the ori-
entation of its long axis for more than a second or so, after its retUril to the earth's field.
The observation that bees (Gould et al., 1980) cannot be "demagnetized" is consistent with
this.
Some observations, however, may not lend themselves to easy interpretation with a
freely rotating model. The freely rotating model naturally predicts a loss of response at
low fields when the average orientation of the magnetite grains is poor. However, the
orientation at high fields saturates and, as can be seen from Fig. 1, there is little difference
between f,LB/kT = 6 (cos 0) = 0.83) and f,LB/kT = 60 (cos 0) = 0.98), at least in regard
to the average orientation predicted by the Langevin equation. If anything, the animal
0.75
ú =
ú =
U)
0
0.5
ú =
0.25
Eú F=
Figure 2. Both fields Band - B produce the same orienting in behavioral experiments. Note that an
elongated magnetic grain oriented with either field would lie along exactly the same line.
should respond better at higher fields. Nonetheless, a loss of orientation has been reported
for fields which are slightly stronger than that of the earth. Bees are reported to lose their
magnetic sense for fields exceeding 5 G (Lindauer and Martin, 1972). For fields greater
than approximately 5 G, they cease to orient their horizontal dance to the points of the
compass as determined by the local magnetic field direction. An interesting "window"
effect is reported for caged European robins in a controlled magnetic environment
(Wiltschko, 1978). The birds orient in preferred directions determined by the magnetic
field inside the cage in narrow ranges of magnetic field strength, approximately ± 0.2 G
around the local earth field strength. Outside this range of intensity, orientation fails tem-
porarily but the birds can be trained to respond to fields outside this range (either higher
or lower) by allowing them to live under altered field conditions for a few days. No such
window effect has been noticed for pigeons although in such experiments as that of Walcott
and Green (1974) they were certainly exposed to net fields (inside a coil mounted on the
bird's head) in excess of twice the earth's field. Kirschvink (1981) has shown that the
accuracy of the honeybees' horizontal dance for fields below 4.5 G can be fit to a Langevin
function. His work makes the freely rotating hypothesis particularly attractive for bees,
despite its apparent failure at "high" fields.
There are several ways of reconciling a freely rotating model with high-field lapses
of orientation. For example, the presence of precursor, superparamagnetic grains of mag-
netite is suggested by the finding of quantities of hydrous iron oxides in honeybees (Ku-
terbach et 01., 1982). It may be that orientation of these grains by sufficiently strong fields
produces signals which confuse the bee, causing it to disregard all magnetic field infor-
mation. Or this high-field cutoff in bees and the "window" effect in European robins may
be related to the way in which the animal interprets the information it receives rather than
to the physical nature of the transducer of its magnetic sense. As we shall discuss in the
next section, there is behavioral evidence that some animals make use of information about
the intensity of the field as well as its direction. Perhaps in fields which are anomalously
high or low (but not low enough to badly degrade orientation), the European robin tem-
porarily disregards all magnetic information and turns back to it only after it seems clear
that the increased field strength is a permanent part of its environment. Much more work
is needed on the response of animals as a function of field strength. The recent discovery
238 Chapter 10
(Walker et aI., 1982) that yellowfin tuna can be conditioned to respond to field changes
and the evidence that these animals do contain magnetite and are not sensitive to small
electric fields (thus probably do not use induction) bode well for the emergence of new
behavioral data to shed light on the mechanisms of magnetic field sensing.
4. Sensitivity to Field Changes

As magnetite was discovered in bees and then pigeons, it became clear from the low-
temperature-induced remanence that a large number of single-domain magnetite crystals
(10 6 -108 ) were present in an individual small animal. This is a great deal more magnetite
than is needed for reliable alignment in the earth's field. Why might such a large excess
of magnetic material be present?
Behavioral evidence suggests that migratory birds and bees are sensitive not only to
fields on the order of 0.5 G but to field variations which can be as small as 0.01% of the
average field. For example, the homing of pigeons is upset by the presence of naturally
occurring magnetic anomalies where the field strength varies by between 0.01 % and 10%
of the average strength (Walcott, 1978, 1980). Magnetic storms have also been shown to
affect homing in pigeons (Moore, 1977; Schreiber and Rossi, 1978). Other migratory birds
show a disturbance in flight path in response to man-made changes in the local field (Larkin
and Sutherland, 1977). Martin and Lindauer (1977) have observed changes in the foraging
patterns of bees when magnetic storms occur. This evidence led to two simultan.eous pro-
posals (Kirschvink and Gould, 1981; Yorke, 1981) that the abundant magnetic material is
needed to permit the detection of small changes in field strength or direction or both.
These changes are interpreted to provide additional useful information about the envi-
ronment. The animal may use the magnetic field for more than compass information.
It is known that for pigeons the sensitivity to small field variations manifests itself
under different conditions than the sensitivity to average-strength field direction. Mature
pigeons wearing field-changing coils or permanent magnets show deviations in homing
patterns only under overcast conditions (Keeton, 1971); in sunny weather, they prefer to
use the sun as a compass. They are, however, confused by very small field changes at
magnetic anomalies in both sunny and overcast weather. It has, therefore, been suggested
that the pigeons use the magnetic field in two ways (and may have two separate sets of
transducers). Under overcast conditions, the magnetic field is used as a backup compass
(and perhaps as a calibration standard for young birds). Under all conditions, small var-
iations in the field are used as a map to tell the bird where it is. A recent review of the
evidence for and against the use of magnetic field variations to provide a map for pigeons
has been made by Gould (1972).
In order to perceive small changes in the local field, the animal could integrate the
responses of many individual detectors. Assuming that the detectors are freely rotating
magnetite grains, Kirschvink and I have estimated the number of grains needed to account
for observed sensitivities. We agree that the numbers inferred from remanence measure-
ments would be large enough. The arguments are sketched in the following paragraph.
Suppose that the response of a single detector is governed by a probability distribution
r
leading to a mean response with variance or2. The probability distribution will depend
on the magnetic field and on the nature of the detector. For example, for freely rotating
magnets, the probability p(a,<\»dad<\> that the magnet is oriented in a small solid angle
centered on an angle a with the field direction is given by the Boltzmann distribution
p(a <\> )dad<\> = exp(f.LB cos alkT) sin a dad<\>

, (21TkT/f.LB)[exp(f.LBlkT) - exp( - f.LBlkT))
This was used to compute the Langevin function, Eq. (1). If N detectors are used, the
summed response will be normally distributed for large N with mean Nf and variance N8r.
A Å Ü ~ å ú =in external field will lead to a change in the average response so r will change
to r + .ir. This change will be distinguishable from statistical fluctuations if N.ir exceeds
the rms variation in response; that is, if .ir > (.ir/N)1/2. If the response depends on the
angle of orientation of a magnetite grain with respect to the field and to some local structure
(e.g., the membrane of the "membrane short" model), then 8r is proportional to .i1l2 which
is, for ILBlkT > 1, approximately ZkTIILB. A change in response due to a change in field
direction can detect a field component .iB at right angles to the original field if
.ill ú = .iBIB ú = (ZkTIILBN)1/2
which requires N > (ZkTIILB) (B/.iB)2 ú = 106 if .iBIB = 10- 3 and IL BlkT ú = 1. Kirschvink
(1981) has considered receptors which would be sensitive to intensity changes alone and
shows that similar estimates apply. Kirschvink and Gould (1981) have also pointed out
that repeated independent measurements, each made over a time interval t which is long
compared to the time constant T for realignment with the field that we have previously
estimated, could increase the effective number of measurements from N to Nth.
We have used freely rotating magnetite grains to estimate the sensitivity to small field
changes but similar considerations (with .very different numbers and different probability
distributions) would apply to other models. Whatever sort of detector is used, by the "map
sense" in birds and the corresponding small field variation sensitivity in bees, it is safe
to predict that there should be many such detectors present in a single animal.
5. Other Types of Receptors

The concept of magnetoreceptors based upon permanently magnetic material has the
attractive feature of simplicity. It also provides an immediate biological justification for
the widespread presence of magnetite in the animal kingdom. Several features of the mag-
netic sense can be explained by models using magnetite-based receptors. However, these
explanations are not unique and other proposals for possible magnetoreceptors have been
made. Induction has already been thoroughly discussed by Rosenblum et al. in this volume
(see also Jungerman and Rosenblum, 1980) and is used for magnetic field detection by
elasmobranch fish (Kalmijn, 198Z). Leask (1977) has proposed a mechanism based upon
optical pumping in molecules located in the eye of a bird. Leask's mechanism requires
light levels compatible with the animal's normal visual environment and would not be
operative in darkness. Experiments conducted in darkness have shown magnetic sensitivity
in salmon fry (Quinn et al., 1981) and perhaps other experiments are needed on vision-
deprived animals to rule out Leask's suggestion. Schulten and Schulten (1977) have briefly
proposed a mechanism based upon magnetic field alterations in chemical reaction rates.
Their proposal has not been further developed to determine whether it is capable of op-
erating at fields as low as 0.5 G with reasonable assumptions about molecular parameters.
Finally, Kirschvink and Gould (1981) have proposed and developed in detail several
models based upon assemblies of interacting superparamagnetic magnetite grains. Their
models can agree as well with most of the behavioral observations as the permanent magnet
models. Indeed, one of them affords a physical explanation for the high-field cutoff in the
magnetic sensitivity of bees.
6. Tests of the Hypothesis

At present, it has not been proven that there are any magnetoreceptors in insects or
vertebrates which use magnetite. Indeed, there are still substantial numbers of critics of
240 Chapter 10
the idea that animals can sense the earth's field, as can be seen from a recent essay by
Griffin (1982). Perhaps work with the yellowfin tuna will definitively lay such doubts to
rest. But even then the question of cleanly experimentally defining the nature of the mag-
netoreceptor will remain.
The most direct way of showing that magnetite is implicated in the magnetic sense
would be to locate magnetite deposits in a living animal and surgically remove them or
anesthetize or sever the nervous connections to the magnetite-containing region. Behav-
ioral experiments would then determine whether the animal has become "blind" to the
magnetic field. This is not an easy task (Walcott and Walcott, 1982; Walcott, this volume)
and such direct and convincing evidence is still far in the future.
A second possibility (Kirschvink and Gould, 1981) is to attempt to destroy the mag-
netoreceptors by subjecting them to a high magnetic field gradient. Receptors based upon
superparamagnetic or permanently magnetic material would be affected while other (chem-
ical or induction) types of receptors would not. Although a uniform magnetic field exerts
only a torque (and not a force) on a magnetic dipole, a nonuniform magnetic field does
exert a force. This force pulls an oriented dipole in the direction of increasing field strength
and is proportional to the gradient of the magnetic field. [More precisely, F = V (lfLIIBI
cosS).) In a medium of viscosity 1], a sphere of radius r moving with velocity v experiences
a viscous resistance of 6 TI1]rv. A sphere subjected to both the magnetic gradient force and
the viscous resistance would very quickly reach a terminal velocity for which the two
forces are equal and opposite. Assuming the dipole moment is parallel to B,
V = fL(VB) = ú =(480)r2 (VB) cm/sec

term 6TI1] r 9 1]
for a magnetite sphere. For r = 0.1 fLm,1] = 1 poise (100 times the viscosity of water), and
VB = 104 g/cm (easily attainable), Vterm = 10- 4 crolsec which, on a cellular scale, is fast.
This could result in a large translational displacement of weakly bound magnetite; perhaps
removing it from a region where it could function as a receptor. It thus seems reasonable
to try to "blind" an animal's magnetic sense by exposure to large gradients and then look
for behavioral changes. It should be emphasized, however, that failure to observe such
changes would not necessarily mean that the receptors are not magnetite. Magnetite which
is bound to one location by a structure which does not fail mechanically at a force on the
order of fLVB (and which might still be free to rotate) would survive such an experiment
intact.
An experiment performed by Kirschvink (1981) provides indirect evidence for a freely
rotating permanent magnet type of magnetoreceptor in bees. He has shown that the ac-
curacy of alignment to the cardinal points of the compass of the horizontal dance of the
honeybee can be fit to the Langevin function [Eq. (1)). There are two adjustable parameters
in his fit, but one of these is fL/kT and permits an independent estimate of the magnetic
moment of an individual magnet. He finds a moment appropriate to a magnetite crystal
of volume 10- 15 cm3 which is well within the single-domain range and of reasonable size
to give good alignment. While this internal consistency does not prove the existence of
rotating magnets, it is certainly very suggestive.
An additional indirect test can be made if it is possible to observe the bees' dance in
an alternating magnetic field. While most models should predict that the magnetic sense
becomes "confused" if it is exposed to a rapidly varying magnetic field, the simplicity of
the freely rotating model permits an estimate of the time scale for rotation and predicts a
time scale (F/fLB) which is inversely proportional to field strength. Earlier in this chapter,
with reasonable parameters, we have predicted a time of 20 msec in a 0.5g field. Thus,
one would expect to see the dance precision deteriorate in a field with a peak strength of
0.5g alternating at 50 Hz (or more slowly for higher viscosities; the frequency at which
the precision deteriorated would provide an experimental measurement of the response

time of the magnetoreceptor). However, this deterioration could be "cancelled" by in-
creasing the peak field strength because of the inverse dependence of the time scale on
field strength even though in de fields of higher strength both the Langevin function and
Kirschvink's experiment predict little or no improvement in orientation.
An observation of such an effect would not "prove" the existence of rotating permanent
magnets in bees but would certainly strengthen the case for such a model, while failure
to observe such an effect would encourage further investigations of other alternatives.
References
Blakemore, R P., 1975, Magnetotactic bacteria, Science 190:377.
Eldridge, D. K, 1961, Quantitative determination ofthe interaction fields in aggregates of single domain
particles, J. Appl. Phys. 32:2475.
Emlen, S. T., 1975, Migration: Orientation and navigation, in: Avian Biology, Volume V (D. S. Farner
and J. R King, eds.), Academic Press, New York, p. 129.
Mater. 15-18:1562.
Gould, J. L., 1982, The map sense of pigeons, Nature 296:205.
Gould, J. L., Kirschvink, J. L., and Deffeyes, K. S., 1978, Bees have magnetic remanence, Science 102:1-
26.
Gould, J. L., Kirschvink, J. L., Deffeyes, K. S., and Brines, M. L., 1980, Orientation of demagnetized
bees, J. Exp. BioI. 86:1.
Griffin, D. R, 1944, The sensory basis of bird navigation, Q. Rev. BioI. 19:15.
Griffin, D. R, 1982, Ecology of migration: Is magnetic orientation a reality?, Q. Rev. BioI. 57:293.
Ising, G., 1945, Die Physikalische Moglichkeit eines tierischen Orientierungssinnes auf Basis der Er-
drotation, Ark. Mat. Astron. Fys. 32A:1.
Jungerman, R L., and Rosenblum, B., 1980, Magnetic induction for the sensing of magnetic fields by
animals-An analysis, J. Theor. BioI. 87:25.
Kalmijn, A. J., 1982, Electric and magnetic field detection in elasmobranch fishes, Science 218:916.
Keeton, W. T., 1971, Magnets interfere with pigeon homing, Proc. Natl. Acad. Sci. USA 68:102.
Keeton, W. T., 1972, Effects of magnets on pigeon homing, in: Animal Orientation and Navigation
(S. R Galler, K. Schmidt-Koenig, G. J. Jacobs, and R K Belleville, eds.), NASA SP-262, U.S.
Government Printing Office, Washington, D.C., pp. 579-594.
Kirschvink, J. L., 1981, The horizontal magnetic dance of the honeybee is compatible with a single-
domain ferromagnetic magnetoreceptor, BioSystems 14:193.
Kirschvink, J. L., 1982, Birds, bees and magnetism: A new look at the old problem of magnetoreception,
Trends Neurosci. 5:160.
Kirschvink, J. L., and Gould, J. L., 1981, Biogenic magnetite as a basis for magnetic field detection in
animals, BioSystems 13:181.
Kuterbach, D. A., Walcott, B., Reeder, R J., and Frankel, R B., 1982, Iron-containing cells in honeybee
(Apis mellifera), Science 218:695.
Larkin, R P., and Sutherland, P. J., 1977, Migrating birds respond to Project Seafarer's electromagnetic
field, Science 195:777.
Lindauer, M., and Martin, H., 1968, Die Schwereorientierung der Bienen unter dem Einfluss der Erd-
magnetfelds, Z. Vgl. Physiol. 60:219.
Lindauer, M., and Martin, H., 1972, Magnetic effects on dancing bees, in: Animal Orientation and
Navigation S. R Galler, K. Schmidt-Koenig, G. J. Jacobs, and R K Belleville, eds.), NASA SP-
262, U.S. Government Printing Office, Washington, D.C., pp. 559-567.
Leask, M. J. M., 1977, A physicochemical mechanism for magnetic field detection by migratory birds
and homing pigeons, Nature 267:144.
Am. Bull. 73:435.
Martin, H., and Lindauer, M., 1977, Der Einfluss der Erdsmagnetfelds auf die Schwereorientierung
der Honigbiene, J. Compo Physiol. 122:145.
242 Chapter 10
Moore, F. R., 1977, Geomagnetic disturbance and the orientation of nocturnally migrating birds, Sci-
ence 196:682.
Exp. Zoo1. 217:137.
Reif, F., 1965, Fundamentals of Statistical and Thermal Physics, McGraw-Hill, New York.
Reitz, J. R., Milford, F. J., and Christy, R W., 1979, Foundations of Electromagnetic Theory, 3rd ed.,
Addison-Wesley, Reading, Mass.
Schreiber, B., and Rossi, 0.,1978, Correlation between magnetic storms due to solar spots and pigeon
homing performances, IEEE Trans. Magn. Mag-14:261.
Schulten, Z., and Schulten, K., 1977, The generation, diffusion, spin motion and recombination of
radical pairs in solution in the nanosecond time domain, J. Chern. Phys. 66:4616.
igation (F. Papi and H. G. Wallraff, eds.), Springer-Verlag, Berlin, pp. 338-343.
Walcott, c., 1978, Anomalies in the earth's magnetic field increase the scatter of pigeon's vanishing
Walcott, c., 1980, Magnetic orientation in homing pigeons, IEEE Trans. Magn. Mag-16:1008.
Walcott, C., and Green, R P., 1974, Orientation of homing pigeons is altered by a change in the direction
of an applied magnetic field, Science 184:180.
Walcott, C., Gould, J. L., and Kirschvink, J. L., 1979, Pigeons have magnets, Science 205:1027.
Walker, M. W., Dizon, A. E., and Kirschvink, J. L., 1982, Geomagnetic field detection by yellowfin
tuna, Oceans Institute Conference Record, IEEE, New York, pp. 755-758.
Wiltschko, W., 1978, Further analysis of the magnetic compass of migratory birds, in: Animal Migra-
tion, Navigation, and Homing (K. Schmidt-Koenig and W. T. Keeton, eds.), Springer-Verlag, Berlin,
pp. 302-310.
Wiltschko, W., and Wiltschko, R, 1972, Magnetic compass of European robins, Science 176:62.
Yorke, E. D., 1979, A possible magnetic transducer in birds, J. Theor. BioI. 77:101.
Yorke, E. D., 1981, Sensitivity of pigeons to small!l1agnetic field variations, J. Theor. BioI. 89:533.
Chapter 11
Particle-Size Considerations for
Magnetite-Based Magnetoreceptors
JOSEPH L. KIRSCHVINK and MICHAEL M. WALKER
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
1.1. Compass Organelles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
1.2. The High-Resolution Magnetic Sense. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
1.3. Magnetic Cues for a Navigational "Map" Sense . . . . . . . . . . . . . . . . . . . . . . . 245
2. The Thermally Driven Variance Model of Magnetic Intensity Reception . . . . . . . . . . . 248
3. Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
4. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 253
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
1. Introduction
The presence of a magnetic influence upon behavior now appears to be a fairly common
trait among a wide variety of organisms, as outlined and discussed elsewhere in this vol-
ume. In a broad manner, these behavioral responses can be grouped into two categories,
the first of which involves the use of a relatively insensitive "compass" to obtain directional
(north/south) information, and a more sensitive system involved in the "map" sense of
vertebrates and the time cue of insects.
1.1. Compass Organelles
Examples of the "compass" are numerous, and include the passive orientation of mag-
netotactic bacteria and algae (Blakemore, 1975; Lins de Barros et al., 1981), the waggle-
dance alignment of honeybees on horizontal honeycomb (Martin and Lindauer, 1977), the
swimming preference of salmon fry and smolt (Quinn, 1980, 1982; Quinn et a1., 1981),
elasmobranchs (Kalmijn, 1982), the stationary orientation of eels (Tesch, 1975, 1980), the
orientation preferences of amphibians (salamanders and newts; Phillips, 1977; Phillips
and Adler, 1978), and the cloudy-day backup compass of homing pigeons and migratory
birds (Keeton, 1972; Wiltschko, 1972; Walcott and Green, 1974). Single-domain crystals
of magnetite are clearly linked to the magnetotaxis of the magnetic bacteria (Frankel and
JOSEPH L. KIRSCHVINK • Division of Geological and Planetary Sciences, California Institute of

Technology, Pasadena, California 91125. MICHAEL M. WALKER • Department of Zoology, Uni-
versity of Hawaii, Honolulu, Hawaii 96822; and Southwest Fisheries Center Honolulu Laboratory,
National Marine Fisheries Service, National Oceanic and Atmospheric Administration, Honolulu,
Hawaii 96812.
243
244 Chapter 11
1.0
0.9
J J J J J J J J J J J J ú ú ú J J...ú K ú ú..-=..
1--1
--
Magnetotactic
Typical Magnetotactic Algae,5-1O fLm
0.8
Bacteria, 1-2 fLm
ú = 0.7 Honeybee
b Compass
.!:: 0.6
>
CI>
g' 0.5
c
...J 0.4 Figure 1. Alignment accuracy of
bees, bacteria, and algae plotted
0.3 as the Langevin function, versus
0.2 the ratio of magnetic to thermal
0.1
energies. The average alignment
of these organisms (or the in-
Mú J J WJ DNMWJ J J WDOMWJ J J J W> P MWJ J J J WDWQMWJ Ng J ú P ú J WNú RMWJ J J J WNú S MWJ J J J J WJ W> ê T M= ferred organelles in bees) varies
r - fLl-' only slightly despite the large
-kT range of magnetic moments.
Blakemore, 1980) where individual cells possess stable ferromagnetic moments which are
large enough to passively rotate the cells into a preferential alignment with the relatively
weak (50 f.l. T) geomagnetic field, despite the randomizing influence of Brownian motion.
The critical physical requirement for passive compass systems of this sort to function is
for the magnetic energy of orientation (f.l.B or f.l.B cos e, where f.l. is the cellular moment,
B the background flux density, and e the angle between the f.l. and B vectors) to be sig-
nificantly larger than the background thermal energy (kT, where k is Boltzmann's constant
and T the absolute temperature). A freely wandering moment of this sort will have a dis-
persion about the applied field direction which is given by the Boltzmann distribution,
exp( - f.l.BlkT) , which leads to an average alignment along the field direction given by the
Langevin function, L(f.l.BlkT) or coth (f.l.BlkT - kT/f.l.B. This function is illustrated in Fig.
1, with 'Y representing the ratio f.l.BlkT as will be used throughout this chapter.
At room temperature and in the geomagnetic field, the values of the parameter 'Y which
have been measured or inferred for the compass of bacteria, algae, and honeybee are shown
in Fig. 1. Frankel et 01. (1979) and Frankel and Blakemore (1980) directly determined the
moment of a typical magneto tactic bacterium by measuring the magnetite volume of single-
domain crystals aligned along the magnetosomes, and Kalmijn (1981) obtained a similar
result by comparing their average swimming velocity in a variety of differing strength
background fields to the Langevin function.
A similar swimming technique involving the radius of curvature when the field is
reversed has also been applied to the magnetotactic algae by Lins de Barros et 01. (1981)
and implies cellular moments (and 'Y values) approximately 10 times larger than for the
bacteria. In addition, a comparison between the accuracy of the honeybee magnetic com-
pass response in various background fields and the Langevin function yields an estimate
for their ferromagnetic compass moment corresponding to 'Y = 6 (Kirschvink, 1981). De-
spite this wide range of magnetic moments, the average alignment (given by the Langevin
function) varies much less, from about 0.82 for the honeybee, 0.94 for the bacteria, and
0.99 for the algae. As can be seen in Fig. 1, a compass organelle smaller than that of the
bees would be rather poorly aligned. The large moments for the bacteria and algae are
apparently related to the need for greater torque (f.l. x B) to align their large cellular mass
in the geomagnetic field with reasonable speed (Frankel and Blakemore, 1980; Lins de
Barros et 01., 1981). A larger torque would be required for the magnetotactic algae, as they
have diameters of 5-10 !Lm and far more surface area than a typical (1-2 !Lm) bacterium.
Extending this argument to the smaller size of the honeybee compass suggests that there
is relatively little additional organic material associated with the particle or chain of mag-
Particle-Size Considerations 245
netic particles. In summary, the compass behavior of bacteria, algae, and honeybee is com-
patible with a ferromagnetic transduction mechanism.
1.2. The High-Resolution Magnetic Sense
Although the high-resolution magnetic sense has not been reported from as wide a
variety of organisms as has the directional response mentioned above, the behavioral ev-
idence for its existence is compelling and reproducible in many birds (discussed in this
volume by Presti) and the honeybee (discussed by Towne and Gould). These behavioral
results raise the difficult problem of understanding how organisms might perceive or be
influenced by small perturbations in the background geomagnetic field «1 part in 104 ).
This problem is clearly more difficult than simply resolving the direction of the field.
Yorke (1981) and Kirschvink and Gould (1981) have independently analyzed this prob-
lem within the framework of the ferromagnetic hypothesis of magnetoreception. Both ap-
proaches are similar in that they assume that the high sensitivities are obtained by having
the nervous system of the animals extract information from numerous magnetite-based
receptors. In general, the quantity of magnetite located in organisms with this highly sen-
sitive system is sufficient to allow either of these strategies to work. Yorke develops her
strategy further elsewhere in this volume; in this chapter we consider some further im-
plications of the variance hypothesis proposed by Kirschvink and Gould (1981) and suggest
some behavioral experiments to test them. First, however, it seems worthwhile to examine
the range of possible ways in which organisms might process information from the geo-
magnetic field into forms useful for bi- or multicoordinate navigation.
1.3. Magnetic Cues for a Navigational "Map" Sense
Navigational map systems proposed for animals range from the "familiar area map"
discussed by Baker (1978) to bicoordinate grid maps involving two independent compo-
nents (e.g., Quinn, 1982). A magnetic familiar area map could involve learning by the
animal of a magnetic topography based on total magnetic field intensity, local gradients
in total intensity, or functions of both. In these cases the problem of fixing a position in
an unfamiliar area is reduced to the extrapolation of knowledge from the total intensity
and gradients in the familiar area (Gould et aJ., 1980). In the following section we shall
focus our discussion on possible map systems.
Magnetic sensory components of the "map" could operate alone or in combination
with one or more other senses. A variety of possible arrangements exist, each of which
could form one component of a navigational "grid" system as outlined in Fig. 2. The
simplest system would be one which operates alone, extracting information directly from
the field as indicated in Fig. 2A. Without any additional directional cues, a system of this
sort would only depend on a function of the total field strength. Because the intensity is
a scalar quantity, it is the simplest geomagnetic parameter which is independent of the
orientation of the animal. An array of magnetosensory organelles which could respond to
a component of magnetic intensity could be exceedingly sensitive, as deduced by Yorke
(1981) and Kirschvink and Gould (1981).
The next level would be to combine the magnetic and gravitational senses in some
fashion, as appears to be the case with the magnetic compass in birds discussed above. A
high-resolution component for the map sense could be formed by monitoring the spatial
angle between "down" and the field direction, and integrating these small inclination
differences over time (Fig. 2B). A map component of this sort would presumably be less
N
ú =
ú =
"Primary" senses Two-Level Derived Compass Responses Combinations of Three or More Cues
Sun or Other Nonmagnetic compasses---r-

"o...J C. Magnet;o Deo';nal;on G,;d
Gravity (Horizontal and Vertical Refer-
ence Frames)
Axial Magnetic Compass (e.g., Wiltschko and
Wiltschko, 1972)
B. Magnetic Inclination Grid - - - - - - - - - ,
Axial Magnetic Direction D. Horizontal and Vertical Field Components
(e.g., Yeagley, 1947)
A. Total Intensity Grid - - - - - - - - - - - - - - - - - - - - - - - - - - - - 1
E. Intensity gradient vector Field
"Circling" and other local behavior upon re-
lease
F. Marine Magnetic lineations for latitudinal
reference, magnetic inclination for relative
latitude
Figure 2. A possible hierarchy of geomagnetic-based cues for a navigational "map," based on bicoordinate grid systems.
n
::r
po
"S!-
et>
.....
I-'
I-'
sensitive than a pure intensity sense, as it must average out noise from both the magnetic
and gravitational systems.
The next and probably least-sensitive level would combine three or more responses
to form a map component. Two examples of this include the proposed use of the geo-
magnetic declination, or variations in the vertical or horizontal magnetic components as
proposed by Yeagley (1947). The declination sense (Fig. 2C) would need the magnetic
compass (gravity and axial magnetic direction) as well as an independent reference system
like a sun or star compass. Although this composite system would be subject to noise from
all of its component subsystems and would not work under cloud cover or for fish
migrating under ice, the geomagnetic declination varies greatly with longitude, particularly
at high latitudes, and thus it might serve as an occasional aid to the "map" when available.
Another third-level possibility would be to resolve the intensity and inclination into
their horizontal and vertical components, respectively, as suggested by Yeagley (1947) and
shown in Fig. 2D. However, it is not clear what additional information an organism would
gain from this extra processing, as the Cartesian components of the field (horizontal and
vertical) contain exactly the same information as the polar coordinates (intensity and in-
clination). It is difficult to conceive of "primary" organelles of any sort which would
directly yield horizontal and vertical field estimates-this would require an organelle
equally sensitive to both gravity and magnetism. For example, the large magnetite crystals
which would be of use for gravity reception would of necessity be multi domain in size,
making their net magnetic moments low and unsuitable for use as a ferromagnetic compass.
For single-domain magnetite crystals, the magnetic energy of orientation typically exceeds
that of gravity by factors of 10 2 to 104 , suggesting that the two senses are probably separate
at the receptor level.
A final third-level possibility of this sort would be for the organisms to keep track of
their positions in the vicinity of the release site while monitoring small shifts in the total
intensity. In this fashion a local "map" of the intensity gradient field could be constructed
and used as another component of a navigational grid. Some migratory and homing or-
ganisms like pigeons, turtles, and pelagic fish have been found to move in 100- to 200-m-
diameter circular paths immediately after release in an unfamiliar area (Walcott, 1978;
Rudloe, 1979); this behavior might be involved in the construction of such a gradient map.
For oceanic animals there is perhaps an even simpler arrangement for magnetic nav-
igation which would involve using the marine magnetic lineations. Unlike the continents,
the geomagnetic field over the ocean basins has a regular and simple pattern composed
of north-south-trending "ridges and valleys" produced by geomagnetic reversals recorded
in the basaltic seafloor (Vine, 1966). These could be used to keep track of relative changes
in east-west position by counting local minima or maxima, while an inclination compass
could keep track of north-south position. An even simpler strategy for animals with pri-
marily north-south migratory routes (e.g., fin whales in the Atlantic) would be to stay on
the magnetic lows. Kirschvink et a1. (in press) have recently found that cetacean strand-
ings along the eastern U.S. continental margin tend to happen where such minima intersect
the coastline, suggesting that such navigational cues may actually be used (Fig. 2F).
In view of this discussion, it therefore seems most likely that any magnetic components
to the "map" would involve functions of the total intensity, inclination, or declination
from true north in this respective order of decreasing probability. This is consistent with
the observation of Walcott et a1. (1979) that between 10 7 and 108 single-domain magnetite
crystals are present in homing pigeons, whereas only a few hundred to a thousand would
be needed for the compass responses. Magnetite-based receptors involved in the meas-
urement of either inclination or declination would be subject to the same Langevin-type
constraints as discussed above for compass organelles. A similar theoretical approach for
organelles involved in an intensity array is developed next, extending the thermally driven
variance model proposed by Kirschvink and Gould (1981) and Kirschvink (1982).
248 Chapter 11
2. The Thermally Driven Variance Model of Magnetic Intensity

Reception
A geomagnetic intensity sense of the type outlined in Fig. 2 must somehow minimize
the directional aspect of the earth's field and extract only its scalar magnitude as would,
for example, a proton-precession magnetometer. Two approaches seem possible for this.
The simplest would be to use organelles which are only sensitive to some function of the
intensity. An integrated signal from a large number of such receptors would similarly be
free of directional dependence. A second approach would be to start with a less-than-
perfect organelle, one that responds to components of both intensity and direction. By
using a large number of such structures with random (e.g., uniformly distributed) spatial
orientations, the directional effects should average out and leave only an intensity-de-
pendent response for the overall system.
The total intensity of the field is probably not responsible for most of the map and
time responses, however. The K-index correlations of Keeton et al. (1974) and Southern
(1978) are best interpreted in terms of a sensitivity to small fluctuations in the background
field, and this interpretation looks plausible for the honeybee circadian rhythm data of
Lindauer (1977) as well as the pigeon magnetic anomaly effects (Walcott, 1978; Papi et
al., 1978). The ability to monitor small shifts in the background field would be of equal
use to a migrating animal as would monitoring total intensity.
Magnetite-based sensory organelles are in principle capable of generating the intensity-
dependent signal required for this type of system. A small particle in a fluid medium will
move randomly in response to thermal bombardment from molecules in the surrounding
fluid. This "Brownian" motion causes the moment of the particle (j,L) to deviate from the
direction of the external magnetic field (B) with a directional probability density given by
the Boltzmann distribution, exp (I' cos 6). If this field is strong, the parameter I' (= j,LBI
kT) will be large and j,L will closely align along the direction of B. An organelle which
fires as some repetitive function of grain orientation would produce a uniform frequency
of action potentials. In weaker fields, I' will be small and the position of the grain will
wander greatly; signals resulting from the organelle would occur with irregular frequency.
Although the mean firing frequency of an organelle would yield information about direc-
tion, the variance around the mean frequency from such a receptor would be a function
of intensity. Several explicit organelle configurations which could produce this type of
response have been discussed elsewhere (Kirschvink and Gould, 1981).
The sensitivity of a variance-based intensity system which responds to background
fluctuations may vary strongly with the magnetic moment of each organelle. The best
moment is clearly that at which the variance of the Boltzmann distribution changes the
most with small fluctuations in 1'. Components of the variance parallel and perpendicular
to the external field are
2
CT.L = (2/1') L(')') (1)
2
CT II = 1 - (2/1') L(')') - L2(')') (2)
where L(1') is the Langevin function as before. Similarly, it is straightforward to differentiate CT 1

andCT II withrespecttol',yielding
aCT.L = _1_ [LI(I') _ L(')')] (3)

al' I'CT .L I'
aCT II =.::..! [L(I')Llb) + CT 1- aCT1-] (4)

ú = CTII ú =
to 1.0
A 0"
B
.8
.2
K l H J ú ú ú ú ú ú ú ú ú ú ú ú = K l HJ ú ú ú ú ú ú ú ú ú J ê J ê ú =
o 2 4 6 8 10 12 0 2 4 6 8 10 12
Y Y
-.16
c -.08
o
ú D=
-.12 -.06
-.04
-.02
J K l l H J ú ú ú ú ú ú ú ú ú ú J ê ú =J K l l HJ ú ú ú ú ú ú ú ú ú ú J ê ú =
o 2 4 6 8 10 12 0 2 4 6 8 10 12
Y Y
Figure 3. Variance of motion components of the Langevin function, and the optimum moment for an
intensity receptor. A and B show the parallel and perpendicular components from equilibrium cal-
culated from Eqs. (1) and (2). The optimum moment for these receptors will be where these functions
change the most with small fluctuations in 'Y. These derivatives, calculated from Eqs. (3) and (4), are
shown in C and D and both clearly peak around 'Y = 2. Because Eq. (6) depends on a/a', the optimum
may be broadened somewhat and shifted to slightly larger 'Y.
These quantities are shown as a function of 'Y in Fig. 3. For very small organelle mo-
ments where 'Y is near zero, both perpendicular and parallel components of the thermally
driven deviation (Fig. 3A,B) approach constant values of (2/3)1/2 and (1/3)112, respectively.
Small fluctuations in 'Y therefore will not yield an appreciable changtl in the variance, and
an array under this condition will not be very sensitive. A similar situation holds for big
magnetite crystals where 'Y is large and the thermal deviations asymptotically approach
zero. As can be seen in Fig. 3C and D, the best organelle moment for both cases is near
'Y = 2, which is several times smaller than that predicted above for a compass organelle.
Note that this analysis does not depend on how the transduction between the magnetite
and the nervous system is achieved; any system which is subject to thermally driven var-
iance should be similarly constrained.
The sensitivity of such an array can be estimated in a straightforward fashion. Let f('Y)
be a typical signal received from one of these organelles, and O'("() be the associated standard
deviation around it (0'2 is the variance, and 'Y is as before). By integrating over N of these
identical organelles, the detectable standard deviation O'("() can be reduced by a factor .of
N 1I2 • Each grain will also have a characteristic rotational time constant, T, and by inte-
grating over a time length I the resolution of an array similarly improves by a factor of
(I/T)1I2. The sensitivity for the entire array to small fluctuations in 'Y, given by t:l"(, can now
250 Chapter 11
be found by assuming, for example, that the organism will be able to detect 1 S.D. shift
in the mean response, or:
cr('y)
[N(IITjp/2 = f('y + ú Dv F = - f('y) (5)
Dividing both sides by ú ='Y and assuming it is small relative to 'Y leads to
cr('y) = f('y + ú D v F = - f('y) = f'( )

ú D v x k E f f q à é L O= ú Dv = 'Y
implying that
(6)
As expected, Eq. (6) shows that the resolution of a magnetosensory system can be im-
proved by increasing the number of organelles (N), the integration time (TIT), or by max-
imizing the magnitude of f'('y) by selecting a suitable value for 'Y. Of course, the functions
f('Y) and cr('Y) will be determined by the particular model under consideration. According
to the intensity model outlined here, the function f('y) would reflect some component of
thermally produced variance generated by motion of the magnetite grain such as those
shown in Fig. 3A and B. The function of f'('y) would then correspond to either Fig. 3C or
D. With a constant temperature, the sensitivity to the magnetic field, ú _ I = would be given
by ú D v â q L à ä = or _ ú D v L D v K = For example, an array sensitive to the perpendicular variance com-
ponent averaging for about 3.5 sec (-100 T) over the 108 particles found in pigeons (Walcott
et al., 1979) should be able to detect an intensity fluctuation of less than 1 nT. in good
agreement with Yorke (1981) and Kirschvink and Gould (1981).
Finally, it should be noted that an intensity arra:y of this sort would also be sensitive
to small fluctuations in temperature. If the magnetic field strength is held constant, the
resolvable temperature change ú =T would be - q ú D v L D v K = where 1:1"( is given by Eq. (6). For
the pigeon system. this translates into a 0.006°C change across about 3.5 sec (-100 T).
Small temperature fluctuations might therefore be confused with magnetic field changes.
unless an independent correction system with separate thermoreceptors is used. It seems,
however, that the thermal inertia in birds is adequate to substantially dampen external
temperature shocks. For example, Torre-Bueno (1976) monitored by radio telemetry the
core temperature of starlings flying in a wind tunnel. By abruptly changing the air tem-
perature from an abnormally high 35°C to a more normal 16°C, he was able to produce a
rate of core temperature change of about 1°C/min (0.05°CI100 T). This change is slightly
more than 8 times the theoretical sensitivity that the variance-based system would have
in pigeons, but is comparable to the 10-nT level implied by their behavioral experiments.
Birds rarely encounter such extreme conditions in the wild. however, and the typical
steady-state shifts reported were much lower than this. Thermal correction of this sort is
therefore not likely to pose much of a problem for larger vertebrates, particularly if the
receptors are in thermal equilibrium with large volumes of internal fluid (blood. cerebro-
spinal fluid, etc.).
On the other hand. Martin and Lindauer (1977) observed temperature effects in the
magnetic behavior of honeybees, although at the time they interpreted their results as
supporting an unspecified paramagnetic transduction mechanism. The smaller volume of
the bee implies that it would have less thermal inertia to ambient temperature fluctuations,
and thermal effects might therefore be expected for them.
3. Discussion
The above analysis suggests that the total magnetic moment of an organelle could be
an important factor in discriminating between the compass and intensity sensory functions
of a magnetoreceptor. We can test this prediction by comparing the ratio fJ.BlkT with the
size ranges of magnetic crystals detected in the pigeons, honeybees, fish, etc. on the Butler-
Banerjee diagram (Fig. 4). Each point in Fig. 4 specifies the size and shape of a magnetite
parallelepiped, and therefore its magnetic moment is fixed as well. Superimposed are three
dotted curves which represent constant grain moments corresponding respectively to 'Y =
0.1, 1, and 10, assuming the grain is in the earth's magnetic field (50 fJ.T or 0.5 gauss) at
body temperature (3100K). As shown in Fig. 4, the bulk of these crystals fall slightly below
the 'Y = 1 curve in the pigeons and fish, indicating that they are in the main individually
too small for compass organelles, and may be too small for the intensity organelles in the
pigeons.
There are at least two ways by which particles of sufficient moment for compass re-
ceptors could occur. The crystals could be individually large enough to possess moments
sufficient to act as compass receptors. Unless they are relatively few in number, below
10,000 or so, these crystals should be detected in the coercivity spectra. It is interesting
to note, however, that Yorke (1979) and Kirschvink and Gould (1981) predict that up to
1000 single-domain magnetite crystals are sufficient to provide organisms with a very
sensitive compass receptor, and that such a small number of crystals could not be detected
by currently available magnetometers. Extraction of the crystals and analysis of their par-
ticle-size distribution are needed to test this point. On the other hand, if systems B, C, and
D of Fig. 2 are used in conjunction with A or as part of the map, more of these compass
forms might be present, although still not necessarily in sufficient numbers to be detectable.
ú =
1.00-r-----,-""T"'<"--,---r--.,...-,----.,...-,------.----.
...
• Y=IO MULTI-OOMAIN
Figure 4. Magnetic stability diagram for 0.50 .
rectangular parallelepipeds of magnetite
(after Butler and Banerjee, 1975, and ú =
. ..
SINGLE OOMAIN
• • Y= I •••
Kirschvink and Gould, 1981). Solid lines
represent the theoretically and experi-
15
:E ..
HONEY BEE- ... ............
• ••• .. ...
...........
mentally determined boundaries be- Y=Q.I • r;l 7....,,....
tween multidomain, single-domain, and
ú =
ú =
0.10
•
•
!
I
1-·..
I •••
Lr_-_-::....J__::
I r--- I,
superparamagnetic crystal sizes, with the
approximate size and shape ranges for ob-
ú =
ú = 9 ••
.. t· ....
'. U PIGEONS ' ••i X J X ê Z | J | J WWK Dú =
I
• •• • • 1. I
served biogenic magnetites superim- u 0.05 4 x10 Y •• BACTERIA ú ê ä ä D =
••• I " /
posed. The three dotted lines crossing the
....
ú= N MMp I ú ? ? ? ú =
single-domain and superparamagnetic ú= ú = ú =
areas represent contours of equal grain
volume and hence constant magnetic mo-
SIJPEI?PAI?AMAGNET/C
ment fL. Each curve is scaled such that the
ratio fLB/kT (= 'Y) is respectively 0.1,1, 0.01 +---,--..,..----.-----,--,------r-,----,-----,,......---l
and 10 at body temperature in the geo- 0.0 0.2 0.4 0.6 O.B 1.0
magnetic field. AXIAL RATIO (width/length)
252 Chapter 11
The second means by which the moment of receptor organelles can be increased is
by linking the crystals into chains to sum their moments. Examination of the electron
micrographs referred to by Walcott et al. (1979) suggests that the crystals are grouped in
chains several particles in length, making their total magnetic moment compatible with
that required for an intensity receptor. As yet there is no suggestion of longer chains such
as those which have been reported from magnetotactic bacteria (Balkwill et al., 1980),
although the possibility that they may have been sheared to a smaller size during the
extraction process cannot be ruled out. Like the larger crystals, chains should be detectable
through measurements of intergrain interactions. Walker et al. (1984, and Chapter 20, this
volume) demonstrate the presence of such effects in fish. Whether the chains are of suf-
ficient moment to serve as compass receptors is unknown, but if they are, it would suggest
that the compass sense might have higher resolution than behavioral experiments have
demonstrated so far.
Some behavioral tests of the hypotheses are also possible. Kirschvink (1981) shows
that the horizontal honeybee dance data of Martin and Lindauer (1977) and Gould et al.
(1980) are compatible with a single-domain ferromagnetic organelle with -y near 6. With
this receptor moment the accuracy of the bees' dance plotted against external field strength
quantitatively resembles the Langevin function, being poor in weak fields and asymptot-
ically improving in higher fields. The shape of the curve indicates the size of the moment
involved; for the bee compass receptor this moment is about 9 x 10- 16 A m2 •
A similar test should be possible for the intensity sense if and when a behavioral assay
for measuring sensitivity to small-intensity fluctuations is developed [the circadian rhythm
experiment of bees (Lindauer, 1977) is a possibility). As discussed above, the best grain
size for use in an intensity receptor is that at which the variance of the Boltzmann dis-
tribution changes the most with small variations in -y (see Fig. 3). With fields in the normal
geomagnetic range, the moment of the receptor is optimal when -y = 2. Outside this range,
this does not hold and sensitivity should decline. Tests for threshold sensitivity to small-
intensity changes over a range of external field values should therefore resemble the op-
timized moment plot (a6Ia-y vs. -y) and so enable -y to be estimated.
As noted by Nesson and Lowenstam (this volume), biomineralization of magnetite
occurs within the confines of an organic matrix which appears to control the size and
shape of the crystals. Within a given organism like bacteria, chitons, or tuna, the magnetite
crystals are all very similar in size and shape. Individually, the crystals do not have suf-
ficient moment to orient the bacterial cells, but their alignment in chains gives them suf-
ficient net moment to enable them to cause the bacteria to take up appropriate orientation
(Frankel and Blakemore, 1980). On the basis of this observation, we could predict that the
moment required, and hence the length of the magnetite crystal chain, is dependent on
cell size. Larger cells should therefore have longer chains of magnetite crystals. In the case
of the unicellular magnetotactic organisms, the length of the chain of magnetite crystals
has been determined by natural selection for sufficient net magnetic moment to bring about
the orientation of the cells.
Like the bacteria, the crystals of magnetite in metazoans have been found or inferred
to be similar in size and shape within species. The responses to magnetic fields by animals
fall into compass and map categories that can be easily explained by the magnetite-based
magnetoreception hypothesis. A corollary of this hypothesis is that there are separate,
optimum receptor moments for the two receptor types predicted here. These moments can
be produced by different sized crystals or the arrangement of similar sized crystals into
chains of different lengths. By arguments similar to those for the bacteria and algae, we
predict that the magnetoreceptors of metazoans will have been subject to selection for
optimum receptor moments for compass and map receptors and that size and shape of the
crystals will be under close genetic control. We would therefore expect to see convergence
on receptors of similar magnetic moments even though the magnetite crystals within the
receptors will be of different sizes and chain lengths in different species. We anticipate
that this will be a productive area for future research.
4. Summary
A variety of behavioral experiments suggest that many organisms are able to sense the
geomagnetic field. Many of these same organisms are also known to precipitate biochem-
ically the mineral magnetite, which is a suitable material for transducing the geomagnetic
field to the nervous system.' These behavioral experiments suggest that there are at least
two separate forms of magnetoreception, one which resolves the axial field direction or
maximum dip angle and one which monitors extremely small fluctuations in geomagnetic
intensity. An extension of the thermally driven variance hypothesis of magnetic intensity
reception suggests that the permanent magnetic moments for organelles of each type may
differ in size by factors of 3 or more, and leads to a variety of testable predictions for
magnetite-based magnetoreceptors.
ACKNOWLEDGMENTS. We greatly thank Drs. John D. Morgan, III, James L. Gould, George Car-
man, and Mark Muldoon for helpful discussions during the preparation of the manuscript.
This work was supported by NSF Grants SPI79-1485, PCM82-03627, BNS83-00301, and
PYI-8351370. This is Contribution No. 4135 from the Division of Geological and Planetary
Sciences, California Institute of Technology.
References
Baker, R R, 1978, The Evolutionary Ecology of Animal Migration, Hodder & Stoughton, London.
Balkwill, D. L., Maratea, D., and Blakemore, R P., 1980, Ultrastructure of a magnetotactic spirillum,
J. Bacteriol. 141:1399-1408.
Blakemore, R P., 1975, Magnetotactic bacteria, Science 190:377-379.
Butler, R F., and Banerjee, S. K., 1975, Theoretical single-domain grain size in magnetite and titan-
omagnetite, J. Geophys. Res. 80:4049-4058.
Mater. 15-18:1562-1564.
Frankel, R. B., Blakemore, R. P., and Wolfe, R. S., 1979, Magnetite in freshwater magnetotactic bacteria,
Science 203:1355.
Gould, J. L., Kirschvink, J. L., Deffeyes, K. S., and Brines, M. L., 1980, Orientation of demagnetized
bees, J. Exp. BioI. 86:1-7.
Kalmijn, A. J., 1981, Biophysics of geomagneticfield detection, IEEE Trans. Magn. Mag-17:1113-1124.
918.
(S. R Galler, K. Schmidt-Koenig, G. J. Jacobs, and R E. Belleville, eds.), NASA SP-262, U.S.
domain ferromagnetic magnetoreceptor, BioSystems 14:193-203.
Kirschvink, J. L., 1982, Birds, bees and magnetism: A new look at the old problem of magnetoreception,
Trends Neurosci. 5:160-167.
Kirschvink, J. L., and Gould, J. L., 1981, Biogenic magnetite as a basis for magnetic field sensitivity
Kirschvink, J. L., Dizon, A. E., and Westphal, J. A., 1985, Evidence from strandings for geomagnetic
sensitivity in cetaceans. J. Exp. BioI., in press.
254 Chapter 11
Lindauer, M., 1977, Recent advances in the orientation and learning of honeybees, Proc. Int. Entomol.
Congr. pp. 450-460.
Lins de Barros, H. G. P., Esquivel, D. M. S., Danon, J., and Oliveira, 1. P. H., 1981, Magnetotactic algae,
Martin, H., and Lindauer, M., 1977, Der Einfluss des Erdsmagnetfeldes auf die Schwereorientierung
der Honigbiene (Apis mellifica), J. Camp. Physiol. 122:145-187.
Papi, F., laale, P., Fiaschi, V., Benvenuti, S., and Baldaccini, N. E., 1978, Pigeon homing: Cues detected
during the outward journey influence initial orientation, in: Animal migration, Navigation, and
Homing (K. Schmidt-Koenig and W. T. Keeton, eds.), Springer-Verlag, Berlin, pp. 65-77.
Phillips, J. B., 1977, Use ofthe earth's magnetic field by orienting cave salamanders (Eurycea lucifuga),
Phillips, J. B., and Adler, K., 1978, Directional and discriminatory responses of salamanders to weak
magnetic fields, in: Animal Migration, Navigation, and Homing (K. Schmidt-Koenig and W. T.
Keeton, eds.), Springer-Verlag, Berlin, pp. 325-333.
Quinn, T. P., 1980, Evidence for celestial and magnetic compass orientation in lake migrating sockeye
salmon fry, J. Camp. Physiol. 137:243-248.
Quinn, T. P., 1982, Use of celestial and magnetic cues by orienting sockeye salmon smolts, J. Compo
Physiol. A 147:547-552.
Quinn, T. P., Merrill, R T., and Brannon, E. L., 1981, Magnetic field detection in sockeye salmon, J.
Exp. Zool. 217:137-142.
Rudloe, J. J., 1979, Time of the Turtle, Random House, New York.
Southern, W. E., 1978, Orientation responses of ring-billed chicks: A re-evaluation, in: Animal Mi-
gration, Navigation, and Homing (K. Schmidt-Koenig and W. T. Keeton, eds.), Springer-Verlag,
Berlin, pp. 311-317.
Tesch, F. W., 1975, Fishes: orientation in space, in: Marine Ecology: A Comprehensive Integrated
Treatise on Life in Oceans and Coastal Waters, Volume 2 (0. Kinne, ed.), Wiley, New York, p.
657.
Tesch, F. W., 1980, Migratory performance and environmental evidence of orientation, in: Environ-
mental Physiology of Fishes (N. A. Ali, ed.), NATO Adv. Study Inst. Ser A, Life Science 35:589-
612.
Torre-Bueno, J. R, 1976, Temperature regulation and heat dissipation during flight in birds, J. Exp.
BioI. 65:471-482.
Vine, F. J., 1966, Spreading of the ocean floor: New evidence, Science 154:1405-1415.
Walcott, C., 1978, Anomalies in the earth's magnetic field increase the scatter of pigeon's vanishing
Walcott, C., and Green, R P., 1974, Orientation of homing pigeons altered by a change in the direction
of an applied magnetic field, Science 184:180.
Walcott, C., Gould, J. 1., and Kirschvink, J. L., 1979, Pigeons have magnets, Science 205:1027-1029.
Walker, M. M., Kirschvink, J. L., Chang, S.-B. R, and Dizon, A. E., 1984, A candidate magnetic sense
organ in the yellowfin tuna, Thunnus albacores, Science 244:751-753.
Wiltschko, W., 1972, The influence of magnetic total intensity and inclination on directions preferred
by migrating European robins, in: Animal Orientation and Navigation (S. R Galler, K. Schmidt-
Koenig, G. J. Jacobs, and R E. Belleville, eds.), NASA SP-262, U.S. Government Printing Office,
Washington, D.C., pp. 569-578.
Wiltschko, W., and Wiltschko, R, 1972, Magnetic compass of european robins, Science 176:62-64.
Yeagley, H. L., 1947, A preliminary study of a physical basis of bird navigation, J. Appl. Phys. 18:1035-
1063.
Yorke, E. D., 1979, A possible magnetic transducer in birds, J. Theor. BioI. 77:101-105.
Yorke, E. D., 1981, Sensitivity of pigeons to small magnetic field variations, J. Theor. BioI. 89:533-
537.
Magnetoreception and Magnetic
IV
Minerals in Living Organisms
Although the link between the biogenic mineral magnetite and magnetic sensitivity in
living organisms has been firmly established only in the magnetotactic bacteria, the hy-
pothesis of magnetite-based magnetoreception leads to many testable predictions con-
cerning the biochemistry and behavior of magnetically sensitive organisms. Obviously, it
predicts that these particles will be present somewhere within an organism and the ex-
istence of specialized receptors which somehow use the motion of attached or included
magnetite particles to transduce the geomagnetic field to an organism's nervous system.
The constrants from rock magnetism imply that these particles should be uniformly and
stably magnetized, and they should have dimensions which fall within the narrow bound-
aries required to yield single magnetic domains. A related prediction is that these crystals
of magnetite are biochemical precipitates rather than inorganic particles of exogenous or-
igin. These and other predictions of the magnetite-based magnetoreception hypothesis are
the focus of the chapters in this section, and have guided most experimental studies in
this new field.
This section covers such a wide spectrum of the animal kingdom that it is impractical
to discuss each study individually here. It seems worthwhile, however, to mention some
of those general results which have particular relevance to the hypothesis of magnetite-
based magnetoreception. First, it is clear that ferromagnetic minerals are commonly found
in animal tissues, and, in the vertebrates, structures in the vicinity of the ethmoid bones
appear to be the major ferromagnetic site. In every case where the identity of the material
has been established, it is magnetite, and in the bacteria and tuna where the particles have
been extracted and examined by electron microscopy, they turn out to be uniform single-
domains. (This is not strictly true for the magnetite in chiton teeth, where some species
are known to produce superparamagnetic and/or multi domain crystals. It may, however,
be true for the magnetite spheres found in cetaceans and turtles.)
Detailed investigations of bacterial magnetosomes and chiton teeth have revealed that
both organisms are able to control precisely magnetite biomineralization, and they probably
use similar pathways involving the transport of ferric iron, precipitation of the mineral
ferrihydrite, and eventual reduction to magnetite. In chitons this process occurs at the
cellular level. and there is no suggestion that symbiotic bacteria are involved. This implies
that they must have the biochemical machinery for magnetite biomineralization, and thus
it is probable that members of other metazoan phyla have this ability as well.
In concluding this introduction, we must mention that there are several critical pre-
dictions made by the hypothesis of magnetite-based magnetoreception which must be
tested for a full and complete validation of the hypothesis. First, and perhaps most im-
portant, the microanatomy of the receptors needs to be worked out. Although large numbers
of magnetite-bearing structures have been located in the vicinity of nervous tissue in the
vertebrates, the actual volume density of crystals is only a few parts per billion. This implies
that the histological search for them will be a tedious, "magnetosome-in-the-haystack,"
affair, even with magnetite-specific staining techniques and thick-section TEM. Next, the
255
256 Part IV Introduction
neurons which convey information from the receptors to the brain need to be identified
and their responses to magnetic field stimuli recorded. Finally, laboratory-based behavioral
responses need to be developed for specific tests of predictions of the magnetite hypothesis
relative to other transduction mechanisms. For example, if honey bees could be conditioned
to distinguish magnetic north from south, an impulse-remagnetization experiment might
be able to flip the polarity of the compass and hence the conditioned behavioral preference.
This is a unique test for the presence of a ferromagnetic compass, and works extremely
well on the magnetotactic bacteria. Unfortunately, most of the known magnetic compass
responses in animals fail to distinguish magnetic north from south, which makes this
particular experiment inconclusive. Other possibilities exist, however, and this should
prove to be a fruitful area for future research.
Chapter 12
Are Animal Maps Magnetic?
JAMES L. GOULD
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
2. The Compass Sense. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
3. The Map Sense. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
4. Problems with Magnetic Maps. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
5. Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
1. Introduction
A variety of animals travel long distances to relatively small targets (reviewed in Gould,
1982a). For example, honey bees may venture out as far as 15 km from their hive and return
safely; vast numbers of monarch butterflies in the eastern United States fly thousands of
kilometers to small, isolated mountain forests in Mexico; green sea turtles which hatch on
tiny Ascension Island return there as adults years later from feeding grounds thousands
of kilometers away; and many species of birds regularly migrate between restricted summer
and winter territories, while homing pigeons can successfully return home after being
displaced hundreds of kilometers. In each of these examples, the ability to navigate appears
to be relatively unaffected by overcasts which block celestial cues, and all raise the question
of whether animals have a "map" sense.
A map sense is the ability of an animal to know where it is with repsect to its goal.
This can be based on information gathered on the journey away from the goal, or familiarity
with useful landmarks, or an actual ability to sense location independent of either of these
kinds of information. Honey bees provide an excellent example of the simplest sort of map
systems. Bees have two strategies for determining where they are. The first is known as
route-based navigation: on her outward flight, a forager bee keeps track of the distance and
direction of each leg of what may be a very circuitous flight (reviewed in von Frisch, 1967;
Gould, 1982a). (As we shall see below, she normally uses the sun's azimuth as her direc-
tional reference, though she has several, hierarchically arranged "backup" systems for use
when the sun is unavailable.) At the same time, she factors out the effects of crosswinds
on her heading as well as the westward movement of the sun. When she prepares to return
home after finding food, she integrates over the various legs of the journey and departs
directly for the hive. All the information necessary for her return was gathered on the way
out. Consequently, a naive bee passively transported from the hive to a release site will
fail to return to her colony. On the other hand, an experienced bee carried to a familiar
area will return. Bees appear to construct a mental map of their flight range as they explore
the vicinity of the hive in search of food, and call upon that map as needed. Whether this
JAMES L. GOULD • Department of Biology, Princeton University, Princeton, New Jersey 08544
257
258 Chapter 12
internal, experience-based map comes to take precedence over the route-based strategy of
navigation is not known.
Given that bees can navigate on the basis of familiarity and route-based information,
is there any reason to suppose that other animals require anything more? Obviously, fa-
miliarity alone cannot explain how sea turtles navigate across the trackless ocean or how
pigeons transported in closed cages well out of their normal flight range can horne. Mon-
archs add the further complication that most individuals have never been to the target,
and so have no route-based information at all. The issue seems settled in that standard
species for investigating vertebrate navigation, the homing pigeon. To use route-based
information, a pigeon would need to adopt either (1) an inertial strategy, measuring the
accelerations, decelerations, and angular rotations on the way out to compute a net dis-
placement, or (2) a compass strategy, measuring the direction of each leg and estimating
distance in some other way.
The inertial strategy is probably ruled out by experiments in which pigeons which
were transported after surgical removal or bisection of their semicircular canals-the or-
gans which measure these inertial parameters-homed successfully (Hachet-Souplet, 1911;
Huizinger, 1935; Wallraff, 1965, 1972; Keeton, 1974a). Similarly, pigeons and other birds
transported under deep anesthesia (Kluijver, 1935; Griffin, 1943; Walcott and Schmidt-
Koenig, 1973) or while being constantly rotated (Griffin, 1943; Wallraff, 1980a,b) appear
to home normally. The compass alternative seems unlikely as pigeons regularly horne after
being transported without access to celestial cues (reviewed in Keeton, 197 4a) and (usually)
when displaced in artificial magnetic fields which would seem at first glance to make
information in the earth's magnetic field useless (Keeton, 1974b; Wallraff, 1980a,b). There
are, however, two minor holes in the case against route-based navigation in pigeons. First,
in most experiments true magnetic direction probably could have been extracted from the
strong, artificial background field if the birds wearing the magnets have the appropriate
processing circuitry: e.g., if the bird can sense field strength along its body axis, flying in
a circle and determining the direction of maximum and minimum field strength would
reveal the north-south axis. Second, the lesson from studies of compass orientation (to be
mentioned below) is that normally there are at least two alternative systems available so
that the elimination of one does not destroy the behavior, because the other automatically
takes over. As it happens, no one has excluded both inertial and compass information in
the same experiment. As a result, the possibility that pigeons use a pair of route-based
strategies cannot be ignored. Nevertheless, the present consensus is strongly in favor of a
map sense which is based primarily on information obtained at the release site, and it is
the possible sensory basis of this apparent ability which I will explore.
2. The Compass Sense

As mentioned above, one of the most important principles to emerge from the study
of navigation is that animals frequently-indeed, usually-have alternative orientation
strategies which are used in a preferred order according to what cues are available. For
bees, where this hierarchial organization was first discovered, the primary cue is the sun
(reviewed in von Frisch, 1967; Gould, 1982a). As a result, contradictory information from
other sources will normally be ignored. If the sun is not visible, however, bees automat-
ically switch to using the patterns of polarized UV light in blue sky and so are able to
continue to navigate and communicate with good accuracy (reviewed in von Frisch, 1967;
see also Brines and Gould, 1979). Under overcast, bees switch to using their memory of
the sun's position with respect to prominent landmarks (Dyer and Gould, 1981). Because
honey bees continue to perform well under overcast even in the absence of useful land-
Are Animal Maps Magnetic? 259
Figure 1. On sunny days, initial ori-

• •
• • • ••••• •
entation of pigeons at a release site
is roughly the same for birds wearing • •• •• •••
\j
magnets as for control pigeons •
equipped with equivalent brass
bars. (The dashed line is the home
direction while the arrow is the
"mean vector" of the departure bear-
á å Ö ë ú í Ü É = center of mass of the dots
V sunny day
at the edge of the circle, each of

which represents one pigeon. A per-
fectly oriented set of birds would
yield a mean vector aligned with the
homeward direction and just touch-
ing the edge of the circle. Poorly ori- •• • •••
ented birds produce a short vector • •• •••
I •
\f
of unpredictable direction.) Under
total overcast, however, magnet- \
\
equipped birds were poorly oriented
whereas the pigeons wearing brass
bars displayed almost normal home-
ú= overcast day
ward orientation. The implication is
that pigeons use the sun for orien-
tation when it is available, but •
switch over to a magnetic compass •• •
when the sun is not visible. Redrawn
from Keeton (1974a). with magnets without magnets
marks, there is probably yet another backup in their repertoire-perhaps based on their
ability to sense the earth's magnetic field (reviewed in Gould, 1980a).
The lesson from bees has aided greatly in elucidating the compass system of birds.
For example, early experiments with pigeons showed that homing was virtually normal
under overcast, thereby (it was argued) ruling out the use of celestial cues (reviewed in
Keeton, 1974a). Similarly, homing was essentially normal in tests in which the pigeons
were equipped with magnets. These experiments (run, of course, on sunny days) were said
to exlude magnetic cues. Keeton, however, demonstrated quite elegantly that pigeons use
celestial cues and the earth's magnetic field as alternative compasses, with the sun strongly
preferred (Keeton, 1971). Hence, magnets have no real effect on sunny days, but virtually
destroy homing ability on cloudy days (Fig. i).
Work on the magnetic compass sense of birds has turned up several interesting twists
which we will do well to keep in mind when we look at the map sense. Experiments with
pigeons wearing a small pair of coils on the head and around the neck (Walcott and Green,
1974), as well as laboratory studies of migrating birds (e.g., Wiltschko and Wiltschko, 1972),
indicate that "north" to birds is the direction in which the earth's field lines "dip" into
the ground, and is utterly independent of the polarity of the field. Moreover, it appears
that pigeons and other birds must first calibrate their celestial compasses against their
magnetic ones before they can use the sun. Hence, birds reared with a conflict between
magnetic and celestial north will adopt magnetic north as their standard for calibrating
solar azimuth (as opposed, for example, to taking the point of highest sun elevation as
south) even though they will then ignore their magnetic compasses in favor of the sun
(Wiltschko et al., 1976, 1981).
3. The Map Sense

The evidence pointing toward a role of magnetic cues in the map sense comes from
a variety of sources, and magnetic map hypotheses of several sorts have been advanced
260 Chapter 12
12:31
N
Figure 2. Pigeons normally reorient their flight
1 paths and fly directly to the loft when familiar land-

marks are encountered. In this experiment, how-
ever, the birds were equipped with frosted contact
lenses to eliminate form vision. As a result, return-
ing pigeons (such as the one shown here) wandered
about in the vicinity of the loft. The implication is
)( that the map sense of pigeons has a resolution of
10:03 3-5 km. Redrawn from Schmidt-Koenig and Wal-
release site cott (1978).
(Yeagley, 1947; Gould, 1980, 1982b; Walcott, 1980a, 1982; Moore, 1980; Lednor, 1982).
To understand the relevance of the supporting data, however, we must first consider briefly
what sort of map information is available from magnetic cues. As Skiles (this volume)
points out, the earth's field changes in a more-or-Iess regular way from the magnetic equator
(where the dip angle is horizontal and the total field strength is roughly 25,000 ')') to the
magnetic poles (where the dip angle is vertical and the field strength is approximately
60,000 ')'). To use magnetic cues for latitude, birds would need to know the approximate
rate of change of the field and be able to measure it with sufficient accuracy. In the north-
eastern United States for instance, the dip angle changes by roughly 0.01°/km while the
total intensity goes up by 3-5 ')'/km. (The horizontal and vertical intensity change about
twice as fast.) The question, then, is just how accurate are the pigeons? The best answer
comes from experiments in which the pigeons were deprived of form vision and so had
to home without being able to see the loft if and when they got near it. From radio tracking
studies in which the birds wandered vainly in the vicinity of the loft after homing from
tens or hundreds of kilometers (Schmidt-Koenig and Walcott, 1978), we can place the
accuracy of the map sense at roughly 5 km (Fig. 2). This translates into a sensitivity of
0.05°, or 15-25 ')' of total field, and/or 30-50 ')' of anyone component of the field. As the
data in Skiles (this volume) immediately suggest, animals using magnetic map cues ought
to have problems during magnetic storms, as these events can alter field strength by tens
to thousands of gammas. Indeed, this is precisely what is observed. Yeagley (1951) showed
that homing speed and/or success in a series of nine pigeon races run in 1948 in New
Jersey was inversely correlated with sunspot activity (the source of the ions which cause
the storms). The same pattern is apparent in an 18-year analysis of pigeon races in Italy
ú Üç ã É=
70
]
ú = 60
ú =
'"
bO
c:
·cra
QJ
.0
Figure 3. At most release sites, pigeons tend to de- bO
c:
part systematically left or right of the true home-
ward direction. This "release-site bias" (shown
ú=
'2
here for Campbell, N.Y.) varies from day to day in ra
>
a manner roughly correlated with magnetic storm c:
strength, here averaged over 12 hr. Taken together ra
QJ
with the results of pigeon races on "stormy" days E
(discussed in the text), the implication is that the
birds' sense of where they are is disturbed. Re-
3 10 30 100 300 1000
drawn from Keeton et a1. (1974). magnetic variability (gammas)
(Schreiber and Rossi, 1976, 1978, 1979). A detailed examination of American race results
during one such storm reveal that the effect is more pronounced for races run east/west
vs. those run north/south, as would be expected (Carr et 01., 1982). As these effects are
seen under both sun and overcast, and are evident even at relatively low storm levels, it
seems unlikely that the birds' magnetic compass sense is involved.
Magnetic storms also affect initial orientation in a curious way. Pigeons are deflected
from their normal departure bearing from a release site in a systematic way which depends
at least roughly on the strength of the magnetic storm (Keeton et 01., 1974). Again, this
phenomenon is observed on sunny days and the magnitude of the deflection is far larger
than would be expected on the basis of any effect on a magnetic compass; instead, it seems
more likely that the pigeons' estimate of the location of the release site is altered (Fig. 3).
Along the same lines, the headings of migrating birds are also altered during magnetic
storms (Moore, 1977). In both cases, the threshold for an observable effect is in the range
of 10-20 "i-approximately the minimum sensitivity expected for a magnetic map sense.
The one obvious problem animals must overcome if they are to use magnetic field
information is clear from Skiles (this volume): the general pattern of magnetic field change
frequently is locally distorted. To construct a magnetic map, a pigeon would have to ex-
trapolate the local gradients sensed around the loft in order to "place" itself at greater
distances. If the local gradient is not representative, a systematic bias in homing direction
should be inevitable. Moreover, any release site could be "placed" accurately only to the
extent that it fits the local pattern. Hence, each release site should impose its own "bias"
on birds. In fact, precisely this kind of pattern is evident from release data. Around Ithaca,
New York, for instance, birds at most sites display a consistent and unique "release site
bias." The biases fall into a fairly obvious general pattern: those to the west of a line running
NNW/SSE through Ithaca are rotated CCW, while those to the east are rotated CW (Windsor,
1972, 1975; see Fig. 4). As would be expected if magnetic map information were being
used, the magnetic field gradient in the northeastern U.S. runs NNW. It is as though Ithaca
pigeons place Ithaca systematically NNW of its actual location (and, of course, place in-
dividual release sites north or south of their true position as well). The same pattern of
biases arranged systematically around the field gradient is seen for lofts in Lincoln, Mas-
sachusetts (Walcott, personal communication), and in Frankfurt (Windsor, 1972; Griiter et
01., 1982) though no clear pattern seems to exist around a loft in North Carolina (Windsor,
1972).
262 Chapter 12
---
_ _ú Q J
N .--
'f
LAKE ONTARIO ú =
Figure 4. The release-site biases

(the difference between the mean
w vector and the homeward direc-
tion) around Ithaca show a clear
tendency for the deflection to be
CW to the east of a line running
roughly NNW-SSE through Ith-
aca. and CCW to the west. (CW
biases are shown in black. CCW
biases in white.) Taken together
with similar patterns at other
lofts (discussed in the text). the
implication is that the pigeons
o 64KM systematically place Ithaca mag-
o
í J f J J ú f =
40MI netically north of its true loca-

tion. (From Windsor (1975).
Another prediction of the magnetic map hypothesis is that orientation at magnetic

anomalies-locations at which the field strength and/or gradients are highly abnormal-
should be strongly affected. Again. this appears to be the case. Walcott (1978) examined
the initial orientation of pigeons at a series of anomalies and found the birds' performance
(on both sunny and overcast days) was inversely correlated with the strength of the anomaly
(Fig. 5). A similar effect is seen at European anomalies (Kiepenheuer. 1982). Again. the
threshold for the effect is comparable to that expected for a magentic map system. A com-
parison of flight tracks from normal and anomalous sites indicates that the effect is quite
strong and persists until the birds are well away from the release site (Gould. 1982a.b;
based on data from Walcott. 1978; see Fig. 6). Pigeons on steep magnetic gradients (as
opposed to the irregular anomalies sampled by Walcott) display a strong tendency to fly
....
o
ic:: 1.0
ú = 0.6
.....o Figure 5. Pigeons released at magnetic anomalies are
-'= disoriented to a degree which corresponds to the
ú MK P = strength of the anomaly. (A value of 1.0 corresponds
to a mean vector which just touches the edge of the
- 0.2 circles in Fig. 1 and so reflects perfect agreement be-
tween the birds. whereas shorter vector lengths in-
dicate greater scatter between individuals.) The im-
0.1 '--------L_----L_--'--_--'---_-'---_ plication is that pigeons at anomalies have difficulty
30 100 300 1000 3000 judging where they are. Redrawn from Walcott
magnetic variability (gammas) (1978).
..:1
-c:..•
II>
'"
E
E
ú=
-p, home
direction
B km
Figure 6. The effect of a magnetic anomaly on pigeon behavior can be clearly seen in this comparison
of tracks of birds released at Iron Mine Hill in Rhode Island (A; this is the site which gave rise to the
data point in the lower right of Fig. 5) vs. tracks from magnetically normal Worcester, Massachusetts
(B). From Gould (1982a), based on data from Walcott (1978) and the u.s. Geological Survey.
264 Chapter 12
down the gradient regardless of the home direction (Wagner, 1976; Frei and Wagner,
1976)-a sensible strategy as an anomaly almost always has an unusually high field
strength so that flying down the gradient is more likely to lead a bird out of the anomalous
region than flying up or across it. Similar but less-well-documented effects of magnetic
topography have been reported by others (e.g., Graue, 1965; Talkington, 1967).
Although it is difficult to imagine any nonmagnetic or nonmap explanation for the
clear inability of pigeons to orient normally at magnetic anomalies, Walcott (personal
communication) has pointed out a potential weakness in this interpretation: most magnetic
anomalies are caused by large subterranean deposits of magnetite-containing iron which
also makes them minor gravitational anomalies. As the force of gravity varies slightly (but
more-or-Iess regularly) from the equator to the poles, it too could in theory provide map
information. Moreover, a very weak and curious correlation between daily site bias and
lunar phase has been reported which, if real, could be explained on the basis of the moon's
effect on the exact force of gravity felt at a release site (Larkin and Keeton, 1978). Walcott's
(personal communication) upcoming experiments at pure gravitational anomalies should
resolve this minor but nagging uncertainty.
Finally, there are several reports that transport to a release site in an abnormal but
well-controlled magnetic field can alter initial orientation (Kiepenheuer, 1978; Papi et aI.,
1978; Wiltschko and Wiltschko, 1978; Wiltschko et al., 1978). This sort of result, however,
is difficult to interpret because (1) the birds had a normal field available upon release, and
(2) other experiments (cited earlier) using apparently similar techniques show relatively
little effect. As reviewed elsewhere (Gould, 1982b) the weight of evidence is strongly
against the sun-arc hypothesis of Matthews, while the only other major theory of the pigeon
map-the olfaction hypothesis of Papi-is more inherently implausible than the magnetic
hypothesis and its experimental evidence is surprisingly weak. In sum, it seems reasonable
to suppose that magnetic field information probably plays some role in the map sense of
pigeons.
4. Problems with Magnetic Maps

There are several obvious and potentially revealing problems with magnetic map hy-
pothesis. The most obvious is that pigeons wearing magnets or coils on sunny days orient
almost normally (e.g., Keeton, 1971; Walcott and Green, 1974). A closer look at the data,
however, reveals that pigeons wearing magnets or coils show the same sort of rotation of
departure directions observed in ordinary pigeons during magnetic storms (Larkin and
Keeton, 1976). As pointed out earlier, the storm data are more consistent with a map
hypothesis than any possible compass effect. For a magnetic map sense to work, then,
pigeons would have to sense the relatively small change of tens or hundreds of gammas
between the field strength at the home loft and the release site though a strong background
field, whether that background is the normal one (of roughly 50,000 1') or an abnormal
one. The best evidence that this sort of processing is actually going on in pigeons is the
report that pigeons equipped with magnets and released from anomalies on sunny days
are still strongly affected by the anomalies (Walcott, 1980b). Because a 1000-1' anomaly
disorients a pigeon whether it is in a normal 50,000-1' field or an abnormal field of hundreds
of thousands of gammas, the ability to sense (and be misled by) small, map-level field
changes seems certain.
A more serious difficulty is that the earth's magnetic field cannot easily provide more
than one component of a map, and a map sense must have at least two useful variables.
Now it is true that there are three variables in field strength: horizontal field strength,
vertical strength, and declination (the deviation from true north). Declination, however,
is of no use without an independent reference to true north, and the unavailability of the
most reasonable candidate, the sun, has no major effect on homing. Both vertical and
horizontal field strength change systematically between the magnetic equator and the poles,
but in the NE U.S. their gradients intersect each other at an angle of 15-30°. If animals
can measure these quantities independenty (which requires also measuring their body
orientation-which is to say, the direction of gravity-to a precision of one part in 104 ),
then a nonorthogonal map grid is possible. For my own part, I have been more comfortable
with the strategy of using total field strength (which does not require measuring the vertical
at all) and simply not worrying about the other map component-after all, even under-
standing one component would be an enormous advance, and would probably make iden-
tifying the other(s) much less difficult. Nevertheless, a careful examination of the pattern
of total field strength surrounding Ithaca in western New York (which is much "lumpier"
and less regular than that of the areas of Massachusetts, Connecticut, and Rhode Island
surrounding Walcott's loft in Lincoln) suggests that the two-component version of the map
hypothesis or a one-component mechanism using dip angle (either of which would prob-
ably help make bettter sense of the pattern) may be more plausible.
Another obvious problem with magnetic maps is that no organ of the necessary sen-
sitivity is known, though species such as honey bees clearly demonstrate map-level sen-
sitivity (reviewed in Gould, 1980a). Of course, behavioral observations usually precede
the physiology, but the evidence for at least a magnetic compass sense has been strong
enough to prompt serious searches for just over two decades and, with two exceptions,
nothing consistent has yet emerged. The exceptions are the elasmobranchs (sharks and
rays) whose passive electrosensitive organs could be used to determine compass direction
(a feat which the animals demonstrate under laboratory conditions) (Kalmijn, 1978, 1982),
and certain unicellular organisms (several species of bacteria and a motile alga) who use
a chain of permanent magnets composed of magnetite to align themselves with the earth's
field lines (Blakemore, 1975; Frankel et 01.,1979; Frankel and Blakemore, 1980; Kirschvink,
1980; Barros et 01.,1982). Elasmobranchs succeed by using seawater as a volume conductor
to provide a return path for electrical current (Kalmijn, 1978). This strategy appears to be
out of the question for other fish as well as terrestrial animals. Neither elasmobranchs nor
the magnetotactic bacteria and algae have sufficient sensitivity for a map sense.
Substantial concentrations of magnetite have been found in a variety of animals in-
cluding honey bees (Gould et 01., 1978), homing pigeons (Walcott et 01., 1979), dolphins
(Zoeger and Fuller, 1981), tuna and salmon (Walker et al., Chapter 20, this volume), mon-
arch butterflies (Jones and McFadden, 1982), sea turtles (Perry et a1., this volume), field
mice (Mather and Baker, 1981), bats (Buchler and Wasilewski, this volume), and so on.
Moreover, the number of magnetite crystals is sufficient to provide the sensitivity necessary
for a map sense (Kirschvink and Gould, 1981), assuming that the nervous system integrates
over many receptors and averages over short intervals-a strategy to obtain high sensitivity
which is evident in many sensory systems.
5. Future Research
Although a magnetite-based map system is a realistic possibility, there is no good
evidence as yet which bears on the issue one way or the other. Decisive experiments could
come in two forms: behavioral or physiological. Neuorphysiological recordings from mag-
netically sensitive animals indicating appropriate levels of sensitivity would be strongly
suggestive. Alternatively, experiments in which animals are transported in carefully con-
trolled fields which provide map information appropriate for displacement in the opposite
direction, followed by equipping the animal at release with a small set of coils to maintain
266 Chapter 12
the small field change appropriate for the pretended displacement, could be decisive. There
are, however, obvious technical complications which arise from the difficulty of main-
taining three-axis field control with 10-,,( precision outside of the laboratory; and these
difficulties are compounded by our present ignorance of which (if any) and how the field
components are used. Another possibility would be to manipulate the potential map in-
formation during Wiltschko's experiments in which caged birds act out their migratory
behavior by hopping in the migratory direction. In particular, we would expect birds with
a two-leg migration pattern such as European warblers (Gwinner and Wiltschko, 1978;
Wiltschko, 1982) to take an interest in where they are in order to turn and then stop at
the appropriate place. Finally, a careful analysis of the behavioral effects of magnetic storms
with exact, three-axis magnetic records made at the experimental site ought to provide
evidence on which components are used and what the time course of the effect is (e.g.,
when is the "baseline" measurement of home made?, when is.the release-site measurement
made?, over what interval is the field averaged to reduce noise?, etc.). Each of these ex-
periments is waiting to be done.
References
Brines, M. L., and Gould, J. L., 1979, Bees have rules, Science 206:571-573.
Carr, H. P., Switzer, W. P., and Hollander, W. F., 1982, Evidence for interference with navigation of
homing pigeons by a magnetic storm, Iowa State J. Res. 56:327-340.
Dyer, F. C., and Gould, J. L., 1981, Honey bee orientation: A backup system for cloudy days, Science
214:1041-1042.
Mater. 15-18:1562-1564.
Frankel, R B., Blakemore, R. P., and Wolfe, R S., 1979, Magnetite in freshwater magnetic bacteria,
Science 203:1355-1357.
Frei, V., and Wagner, G., 1976, Die Anfangsorientierung von Brieftauben in erdmagnetisch gestarten
Gebiet des Mont Jorat, Rev. Suisse Zoo1. 83:891-897.
Gould, J. L., 1980, The case for magnetic sensitivity in birds and bees (such as it is) Am. Sci. 68:256-
267.
Gould, J. L., 1982a, Ethology: The Mechanisms and Evolution of Behavior, Norton, New York.
Gould, J. L., 1982b, The map sense of pigeons, Nature 296:205-211.
202:1026-1028.
Graue, L. C., 1965, Initial orientation in pigeon homing related to magnetic contours, Am. Zoo1. 5:704.
Griffin, D. R, 1943, Homing experiments with herring gulls and common terns, Bird Banding 14:7-
33.
Griiter, M., Wiltschko, R., and Wiltschko, W., 1982, Distribution of release-site biases around Frankfurt,
in: Avian Navigation (F. Papi and H. G. Wallraff, eds.), Springer-Verlag, Berlin, pp. 222-231.
Gwinner, E., and Wiltschko, W., 1978, Endogenously controlled changes in migratory direction of the
garden warbler, J. Compo Physio1. 125:267-273.
Hachet-Souplet, P., 1911, L'instinct du retour chez Ie pigeon voyageur, Rev. Sci. 29:231-238.
Huizinger, E., 1935, Durchschneidung aller bogengange bei der Taub, Pflugegers Arch. Gesamte Phys-
iol. Menschen Tiere 236:52-58.
Jones, D. S., and McFadden, B. J., 1982, Induced magnetization in the monarch butterfly, Danaus
plexippus, J. Exp. BioI. 96:1-9.
Animal Migration, Navigation, and Homing (K. Schmidt-Koenig and W. T. Keeton, eds.), Springer-
918.
Keeton, W. T., 1971, Magnets interfere with pigeion homing, Proc. Natl. Acad. Sci. USA 68:102-106.
Keeton, W. T., 1974a, The orientational and navigational basis of homing in birds, Adv. Study Behav.
5:47-132.
Keeton, W. T., 1974b, The mystery of pigeon homing, Sci. Am. 231(6):96-107.
field influence pigeon orientation, J. Compo Physio1. 95:95-103.
Kiepenheuer, J., 1978, Inversion of the magnetic field during transport: Its influence on the homing
behavior of pigeons, in: Animal Migration, Navigation, and Homing (K. Schmidt-Koenig, W. T.
Kiepenheuer, J., 1982, The effect of magnetic anomalies on the homing behavior of pigeons, in: Avian
Navigation (F. Papi and H. G. Wallraff, eds.), Springer-Verlag, Berlin, pp. 120-128.
Kirschvink, J. L., 1980, South-seeking magnetic bacteria, J. Exp. BioI. 86:345-347.
Kirschvink, J. L., and Gould, J. 1., 1981, Biogenic magnetite as a basis for magnetic field detection in
Kluijver, H. W., 1935, Ergebnisse eines Versuches fiber das Heimfindevermogen von Staren, Ardea
24:227-239.
Larkin, T. S., and Keeton, W. T., 1976, Bar magnets mask the effect of normal magnetic disturbances
on pigeon orientation, J. Compo Physiol. 110:227-231.
Larkin, T. S., and Keeton, W. T., 1978, An apparent lunar rhythm in the day-to-day variations in the
initial bearings of homing pigeons, in: Animal Migration, Navigation, and Homing (K. Schmidt-
Koenig and W. T. Keeton, eds.), Springer-Verlag, Berlin, pp. 92-106.
Lednor, A. J., 1982, Magnetic navigation in pigeons: Possibilities and problems, in: Avian Navigation
(F. Papi and H. G. Wallraff, eds.), Springer-Verlag, Berlin, pp. 109-119.
Acad. Bras. Cienc. Notas Fis CBPF-NF-048/81.
Mather, J. G., and Baker, R R, 1981, Magnetic sense of direction in woodmice for route-based nav-
Moore, B., 1980, Is the homing pigeon's map geomagnetic?, Nature 285:69-70.
Moore, F., 1977, Geomagnetic disturbance and the orientation of nocturnally migrating birds, Science
196:682-684.
Papi, F., laale, P., Fiaschi, V., Benvenuti, S., and Baldaccini, N. E., 1978, Pigeon homing: Cues detected
during the outward journey influence initial orientation, in: Animal Migration, Navigation, and
Homing (K. Schmidt-Koenig and W. T. Keeton, eds.), Springer-Verlag, Berlin, pp. 65-77.
Schmidt-Koenig, K., and Walcott, C., 1978, Tracks of pigeons homing with frosted lenses, Anim. Behav.
26:480-486.
Schreiber, B., and Rossi, 0., 1976, Correlation between race arrivals of homing pigeons and solar
activity, Boll. Zoo1. 43:317-320.
Schreiber, B., and Rossi, 0., 1978, Correlation between magnetic storms due to solar spots and pigeon
homing performances, IEEE Trans Magn. Mag-14:961-963.
Schreiber, B., and Rossi, 0., 1979, Observations on the homing behavior of pigeons during geomagnetic
storms of solar origin, Ital. J. Zoo1. 13:215-216.
Talkington, L., 1967, Bird navigation and geomagnetism, Am. Zoo1. 7:199.
von Frisch, K., 1967, The Dance Language and Orientation of Bees, Harvard University Press, Cam-
bridge, Mass.
Wagner, G., 1976, Das orientierungsverhalten von Brieftauben im erdmagnetisch gestorten Gebiete des
Chasseral, Rev. Suisse Zoo1. 83:883-890.
Walcott, C., 1978, Anomalies in the earth's magnetic field increase the scatter of pigeons' vanishing
ects.), Springer-Verlag, Berlin, pp. 143-151.
Walcott, C., 1980a, Magnetic orientation in homing pigeons, IEEE Trans. Magn. Mag-16:1008-1013.
Walcott, C., 1980b, Homing-pigeon vanishing bearings at magnetic anomalies are not altered by bar
magnets, J. b ú é K = BioI. 70:105-123.
Walcott, C., 1982, Is there evidence for a magnetic map in homing pigeons?, in: Avian Navigation (F.
Papi and H. G. Wallraff, eds.), Springer-Verlag, Berlin, pp. 99-108.
Walcott, C., and Green, R P., 1974, Orientation of homing pigeons is altered by a change in the direction
of an applied magnetic field, Science 184:180-182.
268 Chapter 12
Walcott, C., and Schmidt-Koenig, K., 1973, The effect on pigeon homing of anesthesia during dis-
placement, Auk 99:281-286.
Wallraff, H. G., 1965, Uber das Heimfindevermogen von Brieftauben mit durchtrenuten Bogeniingen,
Z. Vgl. Physiol. 50:313-330.
Wallraff, H. G., 1972, Homing of pigeons after extirpation of their cochleae and lagenae, Nature
263:223-224.
Wallraff, H. G., 1980a, Does pigeon homing depend on stimuli perceived during displacement? I, J.
Compo Physiol. 139:193-201.
Wallraff, H. G., 1980b, Does pigeon homing depend on stimuli perceived during displacement? II, J.
Wiltschko, R, Wiltschko, W., and Keeton, W. T., 1978, Effect of outward journey in an altered magnetic
field on the orientation of young homing pigeons, in: Animal Migration, Navigation, and Homing
(K. Schmidt-Koenig and W. T. Keeton, eds.), Springer-Verlag, Berlin, pp. 152-161.
Wiltschko, R, and Wiltschko, W., 1978, Evidence for the use of magnetic outward-journey information
in homing pigeons, Naturwissenschaften 65:112-113.
Wiltschko, R., Nohr, D., and Wiltschko, W., 1981, Pigeons with a deficient sun compass use the
magnetic compass, Science 214:343-345.
Wiltschko, W., 1982, The migratory orientation of garden warblers, in: Avian Naviagation (F. Papi
and H. G. Wallraff, eds.), Springer-Verlag, Berlin, pp. 50-58.
Wiltschko, W., and Wiltschko, R, 1972, Magnetic compasses of European robins, Science 176:62-64.
Wiltschko, W., Wiltschko, R, and Keeton, W. T., 1976, Effects of a "permanent" clock-shift on the
orientation of young homing pigeons, Behav. Ecol. Sociobiol. 1:229-243.
Windsor, D. M., 1972, Directional preferences and their relation to navigation in homing pigeons,
Dissertation, Cornell University.
Windsor, D. M., 1975, Regional expression of directional preferences by experienced homing pigeons,
Anim. Behav. 23:335-343.
Yeagley, H. L., 1947, A preliminary study of a physical basis of bird navigation,J. Appl. Phys. 18:1035-
1063.
Yeagley, H. L., 1951, A preliminary study of a physical basis of bird navigation, II, J. Appl. Phys.
22:746-760.
Zoeger, J., and Fuller, M., 1981, Magnetic material in the head of the common Pacific dolphin, Science
213:892-894.
Chapter 13
Mossbauer Spectroscopy of Iron
Biomineralization Products in
Magnetotactic Bacteria
RICHARD B. FRANKEL, GEORGIA C. PAPAEFTHYMIOU, and
RICHARD P. BLAKEMORE
1. Introduction to Mtissbauer Spectroscopy . . . . . . . . . . ...... . . . . . . . . . . . . .. 269

1.1. Nuclear 'V-Ray Absorption . . . . . . . . . . . . . . . . ...... . . . . . . . . . . . . .. 269
1.2. Hyperfine Interactions. . . . . . . . . . . . . . . . . . . ...... . . . . . . . . . . . . .. 271
1.3. Spectroscopy of Iron Oxides and Hydroxides. . . . . ...... . . . . . . . . . . . . .. 274
1.4. Biomineralization Products. . . . . . . . . . . . . . . . ...... . . . . . . . . . . . . .. 277
2. Application of Mtissbauer Spectroscopy to Magnetotactic Bacteria. . . . . . . . . . . . . .. 279
2.1. Magnetotaxis in Bacteria. . . . . . . . . . . . . . . . . ...... . . . . . . . . . . . . .. 279
2.2. Mtissbauer Spectroscopy . . . . . . . . . . . . . . . . . ...... . . . . . . . . . . . . .. 280
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 285
1. Introduction to Mossbauer Spectroscopy
Mossbauer spectroscopy is a nuclear 'Y-ray resonance technique that has been extensively
applied to the study of the electronic properties of iron in biological systems. The power
of the technique lies in its sensitivity to the effects of the physical environment (nature
and disposition of ligands, magnetic structure, etc.) on 57Fe nuclei. Moreover, it is sensitive
only to 57Fe so it is not affected by other elements in the system, except as they affect the
environment of the iron atoms. The technique allows discrimination of nonequivalent iron
sites as opposed to magnetic susceptibility or magnetization measurements which lump
together contributions from all the sources in the sample. As extensive reviews of Moss-
bauer spectroscopy are available (Greenwood and Gibb, 1971; Bancroft, 1973; Cohen, 1976,
1981) we will only give a brief introduction here, focusing on 57Fe. This will be followed
by a review of spectroscopy of iron in biomineralization products, and of magnetic inclu-
sions in magnetotactic bacteria.
1.1. Nuclear 'Y.Ray Absorption
Nuclei have discrete energy levels that correspond to different configurations of the
constituent protons and neutrons. Each energy level is characterized by a spin angular
RICHARD B. FRANKEL and GEORGIA C. PAPAEFTHYMIOU • Francis Bitter National Magnet

Laboratory, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139. RICHARD
P. BLAKEMORE • Department of Microbiology, University of New Hampshire, Durham, New
Hampshire 03824.
269
270 Chapter 13
Co -57 (270 day)
5
137 keV
2
10% 90%
ú = - - - + -...........,..--14.4 keV (T #11$10-7 sec)
1 _ .......- -.......- - 0
2
Figure 1. Decay scheme of 57CO to 57Fe.
momentum and other physical parameters. The low-lying energy levels of 2.2% abundant
57Fe are shown in Fig. 1. The stable ground state has a spin 10 = i and there is an excited
state at I1E = 14.4 keV with spin I 14 .4 = i. An 57Fe nucleus in the excited state, for example,
following the decay of radioactive 57CO, can decay to the ground state by emitting a 14.4-
keV "I ray (or by the alternative process of electron conversion), with a half-life of -10- 7
sec. Conversely, a nucleus in the ground state can be excited to the 14.4-keV state by
absorption of a "I ray with energy E'Y = 14.4 keV. The absorption probability or cross section
is a function of "I-ray energy. It is a maximum at I1E and falls to one-half its maximum
value when E'Y = I1E ± r/2, where r is the intrinsic linewidth of the 14.4-keV excited
state. The width is related to the lifetime of the state by the uncertainty principle:
r = h/27rT (1)
where h is Planck's constant. Because T = 10- 7 sec, f = 5 X 10- 9 eV for the 14.4-keV
state. Thus, f/ilE = 10- 12 !
Because electromagnetic radiation has momentum, the nucleus must recoil when it
absorbs the "I ray in order to conserve linear momentum. Thus, some of the "I-ray energy
goes into recoil energy and is not available for excitation. The momentum of the "I ray P
= E'Y/c, hence the recoil energy
(2)
where Mn is the mass of the nucleus and c is the speed of light. For 57Fe, R = 0.002 eV,
i.e., R ú = r. Because of nuclear recoil, the cross section for absorption will be low even
when E'Y is precisely equal to 11E. Mossbauer's great discovery was that if the nucleus is
Mossbauer Spectroscopy 271
embedded in a solid, there is a certain probability that the recoil momentum will be taken
up by the solid as a whole. In those cases, R ú =0 because Mn in the denominator of Eq.
(2) is replaced by the mass of the entire solid, Le., of the order of 1023 times the mass of
the single nucleus. Hence, the Mossbauer effect is recoilless nuclear 'V-ray resonance.
The probability of recoilless absorption f is a function of Mn , E'Y' and the vibrational
characteristics of the solid. It is temperature dependent, decreasing with increasing tem-
perature. These and various other considerations limit the number of isotopes in addition
to 57Fe for which resonant absorption can be observed.
The most convenient source of 14.4-keV 'V rays for Mossbauer spectroscopy of iron in
a particular material is radioactive 57CO which decays to 57Fe with a half-life of 270 days.
As in the case of 'V-ray absorption, 'V-ray emission also involves recoil momentum, but if
the radioactive nuclei are embedded in a solid, recoilless emission can occur. In practice,
57CO, which is produced in a cyclotron, is usually electroplated and diffused into a metallic
host such as Cr, Cu, Rh, Pd, or Pt.
In order to observe the resonance, it is necessary to vary the 'V-ray energy by an amount
of the order of several times r. This is accomplished by changing the effective frequency
of the 'V ray through the Doppler effect by moving the source of 'V rays relative to the
absorber. The change in energy
BE = (v/c)E'Y (3)
where v is the relative velocity. For 57Fe, BE = 2r when v = 0.2 mm/sec so velocities of
the order of 1 mm/sec to 1 cm/sec are sufficient to sweep out the entire line.
The spectrum is most often observed by measuring the intensity of 'V rays transmitted
through the absorbber as a function of relative velocity which is equivalent to energy [Eq.
(3)]. At high positive or negative velocities, high transmission occurs because the 'V rays
have been Doppler shifted off resonance. At velocities close to zero, the absorption cross
section is high and the transmitted intensity is relatively low. The experimentallinewidth
is at least 2r because it includes the intrinsic widths of both the source and the absorber.
1.2. Hyperfine Interactions
The usefulness of Mossbauer spectroscopy for the study of iron in materials stems
from the fact that interactions of the nucleus with the environment perturb the nuclear
levels with consequent changes in the 'V-ray absorption spectrum (Freeman and Frankel,
1967). These interactions are collectively known as hyperfine interactions because they
result in energy shifts or splittings of the nuclear levels which are tiny compared to 14.4
keV but which are nevertheless comparable or larger than r.
The three most important interactions leading to spectral features include (1) the iso-
mer shift, (2) the quadrupole splitting, and (3) the magnetic hyperfine splitting (Fig. 2). In
general, these interactions result in nuclear-level energy changes or splittings and can be
written as a product of nuclear and electronic factors. The nuclear factors are measured
properties associated with each nuclear level. The electronic factors depend on electronic
environment. Determination of the electronic factors gives information about the electron
structure in the solid.
The isomer shift arises from the fact that the ground state and the 14.4-keV excited
state have different mean-squared charge radii. Hence, the electrostatic interaction between
the nuclear charge and the atomic electrons will be different for ground and excited states
and the energy difference between the two states will be affected by the total electronic
charge density at the nucleus. This charge density is mostly due to the atomic s electrons
and is affected by the chemical environment. Thus, the splitting between ground and
272 Chapter 13
SOURCE ABSORBER
14.4T
kev
ú = ú ú =ú ú =ú = ú = ú ú ú ú ú =
O--L--
ISOMER QUADRUPOLE MAGNETIC

SHIFT SPLITTING SPLITTING
Figure 2. Effect of hyperfine interactions on the 57Fe nucleus.
excited nuclear states will be different in different chemical environments, resulting in a

relative shift of the centroid of the absorption line. If the 'V-ray source is chemically different
than the absorber, the absorption line will no longer be centered at v = 0, but will be
shifted to higher or lower velocities. Large differences in the s-electron density and con-
sequently large isomer shifts occur for iron atoms in different oxidation states. The shift
between Fe 2 + and Fe 3 + can be more than 1 mmlsec. Isomer shifts are generally quoted
relative to iron metal.
A splitting of the 14.4-keV nuclear levels can result from the interaction of the nuclear
quadrupole moment with an electric field gradient due to the electrons if the local sym-
metry about the iron atom is less than cubic. As the 14.4-keV state has spin I = ú I = it is 2I
+ 1 = fourfold degenerate. The quadrupole interaction splits the states into two sublevels,
each with twofold degeneracy. This results in an absorption spectrum with two lines of
equal intensity (for a polycrystalline absorber). The magnitude of the splitting AEQ is a
direct measure of the electric field gradient which depends on the local electronic envi-
ronment and the orbital angular momentum. Fe3+ (high spin) has a half-filled 3d shell
and no orbital angular momentum and the quadrupole splittings are typically small, AEQ
< 1.0 mm/sec. Fe 2 + can have unquenched orbital angular momentum and consequently
large quadrupole splittings, AEQ > 2.0 mmlsec, because of the extra electron outside of
the half-filled shell.
A magnetic field acting on the nucleus completely raises the fourfold degeneracy of
the I = ú = excited state and the twofold degeneracy of the I = i ground state. The splitting
between the nuclear sublevels depends on the magnitude of the magnetic field at the
nucleus and the magnetic moments of the ground and excited nuclear states. Although
there are eight possible combinations of the four excited sublevels with the two ground
sublevels, only six transitions are allowed and hence only six lines are usually observed
in the Mossbauer spectrum (Fig. 3). An effective magnetic field at the nucleus, the magnetic
hyperfine field Hhf , results from the interaction of the nuclear magnetic moment with the
atomic magnetic moment. In high-spin Fe 3 + with five unpaired electrons, the magnitude
of Hhf is typically of the order of 500 kOe and negative in sign, that is, the effective field
due to the magnetic hyperfine interaction is oriented anti parallel to the atomic magnetic
1.00
til,
.. ú = "rt ú =f'\ r.. r"'· an
T, I
.. .
ú I WK =
..
l ••• '.'.
:! .: :::: :. ,.
>-
o 00
00
00
I- 00
o
(j) 00 o
Z
W 0
0
00
I-
z
1-1
0.90 0
0 .
o
o
-10.0 -5.0 0 5.0 10.0

VELOCITY (mm/sec)
Figure 3. Mossbauer spectrum of iron metal at room temperature.
moment. In Fe2 + the atomic moments are typically smaller and more variable from com-
pound to compound because of orbital contributions; this is reflected in smaller and more
variable magnetic hyperfine fields. In iron metal where the moment per iron atom is 2.2
!LB, the magnetic hyperfine field is -330 kOe (at room temperature).
The magnetic field causes the nucleus to precess with a frequency VL that is propor-
tional to the field strength; the time for a single complete precession is known as the Larmor
precession time. For an Fe3 + atom with I Hhf I = 500 kOe, this is about 10- 7 sec. If the
field fluctuates or changes sign due to relaxation of the atomic moment on a time scale
less than the Larmor precession time, the magnetic splitting will not be observed and the
M6ssbauer spectrum will consist of a single line (or quadrupole doublet). This is generally
the case for paramagnetic iron atoms. In ferro-, ferri-, or antiferromagnetic iron compounds
where each atomic moment is oriented in a fixed crystallographic direction, the full mag-
netic splitting is observed in the spectrum. As the temperature increases through the Neel
or Curie point (the magnetic ordering temperature) above which the material is paramag-
netic, the spectrum splitting decreases and collapses to a single line (or quadrupole
doublet), reflecting the transition from magnetic order to paramagnetism.
The magnetic splitting can also be observed in paramagnets when the electron spin
is polarized by an external magnetic field. The magnetic hyperfine splitting depends on
degree of polarization or magnetization of a paramagnet, which varies with HolT up to its
full or saturation value. However, in an external field, two additional factors come into
play. First, the field at the nucleus Hn is now the vector sum of the applied field Ho and
the magnetic hyperfine field Hhf : -'
(4)
In paramagnets where the atomic moment is polarized parallel to Ho, the equation reduces
to a scalar equation and Ho adds or subtracts from H hf . Because Hhf is usually negative,
the net field is the difference between Hhf and Ho. For example, Fe 3 + with I Hhf I = 500
kOe in a 60-kOe external field at low temperature would have a spectral splitting corre-
sonding to 440 kOe. Second, the relative intensities of the spectral lines depend on the
orientation of Hn with respect to the 'V-ray propagation direction. If the 'V rays are propagated
parallel to the field direction, the relative intensities of the six lines are 3: 0: 1 : 1 : 0: 3. If
the 'V rays are propagated perpendicular to the field direction, the relative intensities are
274 Chapter 13
3: 4: 1 : 1 : 4: 3. In a polycrystalline, magnetically ordered material, the relative intensities

average to 3:2:1:1:2:3.
If a magnetically ordered material is placed in an external magnetic field, similar effects
occur (Chappert et 01., 1979). In a ferromagnet, the magnetization is generally polarized
parallel to Ho, in which case the field at the nucleus is the difference between Hhf and Ho.
The spectral intensities follow the same rules as outlined above. In a ferrimagnet, where
there are two or more magnetic sublattices with anti parallel moment orientation, the net
moment of the material will orient parallel to the applied field and the field will subtract
from the hyperfine field of those atoms whose moments are oriented parallel to the net
moment, but will add to the hyperfine fields of those atoms whose moments are oriented
anti parallel to the net moment. In antiferromagnets, the moments of the two anti parallel
sublattices cancel each other and there is no net moment to orient in the applied field. The
applied field simply broadens the lines without changing their positions or relative in-
tensities compared with zero field.
Superparamagnetism in small particles of magnetically ordered materials can be ob-
served in the Mossbauer spectrum. In this phenomenon, thermal energy excites transitions
of the magnetization, or of the sublattice magnetizations in antiferromagnets, between en-
ergetically equivalent crystallographic axes. These transitions are opposed by energy bar-
riers proportional to KV, where K is the anisotropy energy per unit volume and V is the
volume of the particle. K is a constant, characteristic of each magnetic material. The fre-
quency of transition is given by
f = fo exp( - KV/2kB T} (5)
where fo is a constant and kB is Boltzmann's constant. Changes in the Mossbauer spectrum

occur when f = VL, the Larmor precession frequency, as discussed above for paramagnets.
For Fe3+ with a hyperfine field of 500 kOe, VL = 10 7 sec- 1 • According to Eq. (5), this
condition will occur .at a temperature TB for a given particle size. For T < TB, one will
observe the full magnetic splitting in the spectrum. For T > TB, one will observe a collapsed
or paramagnetic spectrum, i.e., a single line or quadrupole doublet. If there is a distribution
of particle sizes in the sample, the condition f = VL will be satisfied in different particles
at different temperatures. Thus, there will be a region of temperature over which the six-
line spectrum and the collapsed spectrum coexist, with the latter increasing in intensity
and the former decreasing in intensity with increasing temperature. This is illustrated in
the case of the iron storage protein ferritin discussed below. Note that the condition for
observation of superparamagnetism by the Mossbauer effect is different than that for mag-
netization measurements. If f = 1 sec-t, the remanent magnetization will decay rapidly
but the particle will be stable on the Mossbauer time scale.
1.3. Spectroscopy of Iron Oxides and Hydroxides
Iron in a large number of compounds and minerals has been studied by Mossbauer
spectroscopy. Iron oxides and hydroxides are particularly relevant to iron biominerali-
zation processes so we will briefly review their spectral features here.
a-Fe203, hematite, has a close-packed oxygen lattice with Fe3 + ions in octahedral
sites. The magnetic ordering temperature is 950oK. Up to about 250 oK, the magnetic or-
dering is antiferromagnetic. At 250 oK, the material undergoes a transition known as the
Morin transition to a weakly ferromagnetic state in which the two sublattices cant toward
each other producing a small net moment. The Mossbauer spectrum is a single six-line
pattern with a magnetic splitting of 515 kOe at room temperature (Fig. 4) and 540 kOe at
1.00
. å. .å å.ê. ? y. . ú .ê. =
I .å
... .... ..' '... .... .. ..
.·
>-
.. .. .
··
ú =
0.90 '
.
CJ)
z
IJJ .. I
"
.
ú =
....
Z
.. .' '.
0.80 '.
-10.0 -5.0 0 5.0 10.0
VELOCITY (mm/sec)
Figure 4. Mossbauer spectrum of a-Fe20a at room temperature.
4.2°K. The spectrum undergoes subtle changes at the Morin transition (van der Woude,
1966).
'V-Fe203, maghemite, has a spinel (XY204) structure in which Fe 3+ ions in the x sites
are tetrahedrally coordinated and Fe3+ ions in the y sites are octahedrally coordinated.
Antiparallel alignment of iron atoms in x and y sites results in a ferrimagnetic structure.
The Mossbauer spectra of iron in the two sites are almost identical but they can be dis-
tinguished in an external magnetic field. The hyperfine fields are about 490 and 500 kOe
for the x and y sites, respectively, at room temperature (Armstrong et al., 1966).
Fe304, magnetite, has an inverse spinel structure (Xy 204) in which x is Fe 3+ in tetra-
hedrally coordinated sites and the two y's are Fe2+ and Fe3+ in octahedrally coordinated
sites. The magnetic structure is ferrimagnetic with the Fe2+ and Fe 3+ in octahedral sites
aligned parallel to each other and antiparallel to Fe3+ in tetrahedral sites. At room tem-
perature, there are rapid electron transitions between iron atoms in the y sites and the
Mossbauer spectrum consists of two overlapping subspectra, one due to Fe3+ in x sites,
the other an averaged spectrum due to Fe 2 + and Fe 3 + in y sites (Fig. 5a). The latter spectrum
has a smaller hyperfine splitting (Hhf = 453 kOe) and a positive isomer shift compared
to the x-site spectrum (Hhf = 491 kOe), reflecting the ferrous character of the y sites. Below
about 120oK, the temperatures of the so-called Verwey transition, Fe304 has a complex
magnetic structure which results in two partially resolved subspectra, one with Hhf = 500
kOe, the other with Hhf = 480 kOe (Fig. 5b). The former is due to Fe3+ in x and in y sites,
which, like 'V-Fe203, have similar spectra. The latter sub spectrum is due to Fe 2+ in y sites
(Banerjee et al., 1967; Hargrove and Kundig, 1970).
Mossbauer studies of small particles of Fe304 have also been carried out, with obser-
vation of superparamagnetic effects (McNabb et al., 1968).
a-FeOOH, geothite, has Fe3+ atoms in distorted octahedral sites. The structure is an-
tiferromagnetic below about 400°K. The magnetic splitting of the Mossbauer spectrum at
low temperature is about 500 kOe, which decreases to 380 at room temperature (Forsyth
et al., 1968).j3-FeOOH also has Fe3+ in octahedral sites, but the structure is nonstoichiom-
etric. The magnetic ordering is antiferromagnetic below 295°K and the magnetic splitting
of the Mossbauer spectrum is 475 kOe at low temperature (Dezsi et aI., 1967). 'Y-FeOOH,
lepidocrocite, is similar to j3-FeOOH, but with a complex layer structure. It is paramagnetic
to below 77°K where it becomes antiferromae:neticallv ordered. However. the Neel tem-
276 Chapter 13
1.00
" ' II' ' " ú = r".. r-.. r--\ I"". ".,-- "
i I :: .1:: .1 \ I
.t, .. 'a: Wú =...
...... ..
Wú =
:. ::
..
. •
'.'
... ::.
,.
ú= \f (0 )
., ... '
., "
I .,
>-
ú =
ú = 0.90
r
"" "'. ,". r" ..
ú = 1.00
Z ú y ú==
ú =
:. .! : j I :i
.j
.1
:;
\
:
I
, ,
•
. .., t
., .• : (b)
..
f
..!, ,
,.
.
.. of
•
0.90 ; ' .
.
-10.0 -5.0 0 5.0 10.0
VELOCITY (mm/sec)
Figure 5. Mossbauer spectrum of Fe 304 at (a) room temperature and (b) BOOK.
lOOK. The low-temperature magnetic hyperfine field is 460 kOe Uohnson, 1969). B-FeOOH
has Fe 3 + atoms in a hexagonally close-packed oxygen lattice with unequal numbers of
Fe3 + ions in different layers. Unlike the other FeOOH structures, the magnetic structure
is ferrimagnetic with two overlapping Mossbauer sub spectra at 80 K with magnetic hy- 0
perfine fields of 505 and 525 kOe (Dezsi et a1., 1967).

5Fe203'9H20, ferrihydrite, is a naturally occurring hydrous iron oxide which is
thought to be similar to the iron core of the iron storage protein ferritin. Natural and
synthetic samples exhibit varying degrees of crystallinity and corresponding variations in
X-ray diffraction patterns. It contains Fe3+ ions in octahedral sites, coordinated to 0, OH,
and OH 2. The Mossbauer spectrum at room temperature consists of a quadrupole doublet
with broad lines, indicating several slightly inequivalent iron sites. At 4.2°K, a magnetically
split spectrum with Hhf ::::: 500 kOe is obtained, but again with broad lines indicating a
distribution of magnetic hyperfine fields (Murad and Schwertmann, 1980).
Spectra for a naturally occurring amorphous iron oxide gel have also been obtained.
This material has a composition corresponding to Fe(OHla·0.9H2 0 with octahedrally co-
ordinated iron atoms but no long-range structural order. For T > 20oK, the spectrum con-
sists of a quadrupole doublet. At 4.2°K, the spectrum is split with a hyperfine field Hhf :::::
460 kOe. External field measurements indicate that the material is paramagnetic for T >
100oK, superparamagnetic for lOOK < T < 100oK, and magnetically ordered below lOOK
(Coey and Readman, 1973).
Mi:issbauer Spectroscopy 277
TABLE I. Mossbauer Parameters at 80 0K

Material Il (mm/sec)Q dE Q (mm/sec)b
A. magnetotacticum
Spectrum B 0.47 ± 0.03 0.65 ± 0.05
Spectrum C 1.32 3.17
Nonmagnetic cells 0.47 0.68
Cloned, nonmagnetic cells 0.51 0.65
Ferritin 0.47 0.73
E. coli
Storage material 0.50 0.66
Molpadia intermedia
Dermal granules 0.48 0.84
Ferrihydritec 0.47 0.74
Amorphous ferric gel d 0.47 0.81
a Isomer shift relative to iron metal at room temperature.

b Quadrupole splitting.
C Murad and Schwertmann (1980).
d Coey and Readman (1973).
1.4. Biomineralization Products
A number of iron biomineralization products have been studied by Mossbauer spec-

troscopy (Table I). An important class of iron biominerals occur in the cores of the iron
storage proteins including ferritin, hemosiderin, and gastroferrin. These proteins are large
spherical molecules, 120 A in diameter with 70-A-diameter iron-containing cores. The iron
is sequestered as a ferric oxyhydroxide of approximate composition (FeOOH)8'FeO'P04H2
(Blaise et al., 1965).
The Mossbauer spectrum of horse spleen ferritin shows evidence of superparamagne-
tism for 200K < T < 60°K. Below 20oK, the spectrum is magnetically split with Hhf = 500
kOe. Above 60 oK, the spectrum is a quadrupole doublet. Between 20 and 60 kOe, the
magnetically split spectrum and the quadrupole doublet coexist with the intensities of the
former and latter respectively decreasing and increasing with increasing temperature. The
spectral effects in this temperature range are consistent with an average particle diameter
of 70 A and an anisotropy constant of - 104 ergs/cm3. Hemosiderin gives spectra that are
very similar to those of ferritin (Fig. 6). The spectra of ferritin from the fungus Phycomyces
and of bacterioferritin from Azotobacter are similar to mammalian ferritin except that the
superparamagnetic behavior is observed over lower temperature ranges. If the core com-
positions in all ferritins are similar, we can assume that the decrease in the blocking tem-
peratures reflect smaller particle sizes in the plant and bacterioferritins (Oosterhuis and
Spartalian, 1976).
An iron-rich storage material of as yet unknown composition has been found in E.
coli and other prokaryotes, P. mirabilis and M. capricolum. The Mi:issbauer spectrum for
T> lOOK of the iron storage materials from E. coli is a quadrupole doublet with parameters
characteristic of high-spin Fe3 +. A six-line magnetic hyperfine spectrum with an effective
magnetic field at the nucleus of 430 kOe is observed at T < 1 K. Above 1OK, the lines
0
broaden and the splitting decreases with increasing T and collapses into the quadrupole
doublet at about 3SK. Between 1.2 and 3SK, the doublet and sextet are superposed,
indicating a spread of magnetic transition temperatures. This indicates lower-energy mag-
netic interactions between iron atoms than in ferritin, perhaps reflecting less dense packing
of the iron atoms than in ferritin (Bauminger et al., 1980).
278 Chapter 13
I ) .)It
- 0 -I '. - 4 0:2 • 10 12
VELOCITY (mm/sec) Figure 6. Mossbauer spectra of hemosiderin

in heart tissue. From Kaufman et 01. (1980).
Molpadia intermedia is a species of marine invertebrate that synthesizes iron- and

phosphate-rich dermal granules ranging in size from 10 to 350 IJom. These serve as strength-
ening agents in the connective tissues of their dermis. The granules consist of layers com-
posed of two types of spherical to ellipsoidal subunits 0.03 to 0.24 IJom in diameter, sep-
arated and alternately encapsulated by organic material. One type of subunit contains
water, iron, and phosphate with lesser amounts of calcium and magnesium. These deposits
are X-ray amorphous and in turn consist of electron-dense subunits 90-140 A in diameter.
The iron is present in the form of hydrous ferric polymeric units similar to the iron-con-
taining micelles of ferritin (Lowenstam and Rossman, 1975).
The Mossbauer spectrum at temperatures between 10 and 300 K is a broadened quad-
0
rupole doublet indicating a distribution of electric field gradients at the iron sites. Below
Mtissbauer Spectroscopy 279
(0)
ú=
c:
-
::J
ú= - --- -------.........
ú= (b)
.Q
ú= ------- I
-
..
c:
\.:-\'"
2!
-= (e)
I I
-1 2 -1 0 -8 -6 - 4 2 o 2 4 b
Veioci y (mm/sec)
Figure 7. Mtissbauer spectra of iron phosphatic dermal granules from Molpodio interrnedio at (a) 20,
(b) 7.2, and (c) 1.60 K. From Ofer et 01. (1981).
lOOK, the spectrum broadens and magnetic hyperfine structure appears, with the effective
magnetic field at the nucleus increasing with decreasing temperature. The breadth of the
lines indicates a distribution of magnetic hyperfine fields . At 1.6°K, the mode of the dis-
tribution is at 420 kOe and moves to progressively lower fields with increasing T, collapsing
at about lOOK (Fig. 7). The quadrupole doublet and the magnetically split spectra coexist
from - 8.0 to 10.0°K. The collapse of the magnetic hyperfine spectrum is indicative of a
magnetic transition at about lOOK. A longitudinal magnetic field of 80 kOe at 4.2°K broadens
the lines without substantially changing the line positions or relative intensities. This
indicates antiferromagnetic ordering of the iron atoms in the granules (Ofer et 01., 1981).
2. Application of Mossbauer Spectroscopy to Magnetotactic

Bacteria
2.1. Magnetotaxis in Bacteria

Magnetotactic bacteria are various species of aquatic microorganisms that orient and
swim along magnetic field lines (Blakemore, 1975, 1982; Moench and Konetzka, 1978;
Blakemore and Frankel, 1981). All magneto tactic cells examined to date by electron mi-
croscopy contain iron-rich, electron-opaque particles (Balkwill et 01 ., 1980; Towe and
Moench, 1981). In several species of magneto tactic bacteria, and possibly all, the particles
consist of magnetite, Fe304 (Frankel et 01. , 1979). Cuboidal, rectangular, parallelepiped,
and arrowhead-shaped particles occur in different species with typical dimensions of 400
to 1200 A. This places the Fe304 particles in the single-magnetic-domain size range. In
most species, the particles are arranged in chains, which impart a magnetic moment to the
cell, parallel to the axis of motility. The moment is sufficiently large that the bacterium
is oriented in the geomagnetic field at room temperature as it swims, i.e., the chain of
Fe304 particles functions as a biomagnetic compass (Frankel and Blakemore, 1980). The
organism thus propels itself along the geomagnetic field lines. The direction of migration
280 Chapter 13
depends on the orientation of the biomagnetic compass. Those with north-seeking pole
forward migrate north along the field lines. Those with south-seeking pole forward migrate
south. It has been found that north-seeking bacteria predominate in the northern hemi-
sphere while south-seeking bacteria predominate in the southern hemisphere (Blakemore
et al., 1981; Kirschvink, 1980). The vertical component of the inclined geomagnetic field
selects the predominant polarity in each hemisphere by favoring those cells whose polarity
causes them to be directed downward toward the sediments and away from the toxic effects
of the oxygen-rich surface waters. At the geomagnetic equator where the vertical com-
ponent is zero, both polarities coexist; presumably, horizontally directed motion is equally
beneficial to both polarities in reducing harmful upward migration (Frankel et al., 1981;
Frankel, 1982).
In the freshwater magnetotactic spirillum, Aquaspirillum magnetotacticum, iron com-
prises 2% or more of the cellular dry weight. Electron microscopic studies of this organism
show that the Fe304 particles are cuboidal, 400-500 A in width, and are arranged in a
chain that longitudinally traverses the cell (Fig. 8). The particles are enveloped by electron-
transparent and electron-dense layers; a particle and its enveloping membrane has been
termed a magnetosome (Balkwill et al., 1980).
Because A. magnetotacticum is cultured in a chemically defined medium in which
iron is available as soluble ferric quinate (Blakemore et a1., 1979), the presence of intra-
cellular Fe304 implies a process of bacterial precipitation of this mineral, with control of
particle size, number, and location in the cell.
In order to elucidate the Fe304 biomineralization process, we have studied cells and
cell fractions, some isotopically enriched in 57Fe, by Mossbauer spectroscopy. Cells of a
nonmagnetotactic variant that accumulated iron but did not make Fe304 and of a cloned,
nonmagnetotactic strain that accumulated less iron, were also studied. The results suggest
that Fe304 is precipitated by reduction of a hydrous iron oxide precursor (Frankel et al.,
1983).
2.2. Mossbauer Spectroscopy
Mossbauer spectra of wet packed cells enriched in 57Fe at 2000K and at 800K are shown
in Figs. 9 and 10, respectively. The 2000K spectrum can be analyzed as a superposition of
spectra corresponding to Fe304 (spectrum A), a broadened quadrupole doublet with pa-
rameters characteristic of ferric iron (spectrum B), and a weak quadrupole doublet with
parameters corresponding to ferrous iron (spectrum C) (Table I). Spectrum A is itself a
superposition of sub spectra corresponding to the Fe2+ and Fe3+ in octahedral sites (A2l
and the Fe3+ in tetrahedral sites in Fe304 (All. .
Spectrum B is also observed in lyophilized cells and has isomer shift and quadrupole
splitting parameters similar to iron in ferritin and in the mineral ferrihydrite, indicative
of ferric iron with oxygen coordination. The relative intensity of B to A is somewhat var-
iable from sample to sample, depending on growth conditions. At BOOK, spectrum A cor-
responds to Fe304 below the Verwey transition (Fig. 5) and the parameters of spectrum B
and the relative intensity of B to A are relatively unchanged compared to the spectrum at
250°K. Between BO and 4.2°K, however, the intensity of B decreases with decreasing tem-
perature so that at 4.2°K, only a residual doublet remains. A similar temperature depen-
dence for spectrum B is also obtained in lyophilized cells.
The isomer shift and quadrupole splitting parameters of spectrum C correspond to
high-spin ferrous iron in coordination with oxygen or nitrogen. This spectrum was not
observed with lyophilized cells, possibly as a result of oxidation during sample prepa-
ration. Wet packed cells kept unfrozen under anaerobic conditions contain increased
amounts of material responsible for spectrum C and correspondingly less material with
Miissbauer Spectroscopy 281
Figure 8. Electron micrograph of magnetosomes in A. magnetotacticum. The bacterium is approxi-

mately 3 fLm long.
282 Chapter 13
All
A21
1.00
>-
.....
en 0.95
zw
.....
z
0.90
0.85
Figure 9. Mossbauer spectrum of A. magnetotacticum wet packed cells at ZOO°K.
spectral characteristics H. Thawing and aeration of these frozen cells result in increases
in H spectral lines and concomitant decreases in C spectral lines. This indicates that the
iron atoms responsible for spectrum C came from reduction of the iron atoms giving spec-
trum H. Unlike that of spectrum H, the intensity of spectrum C does not decrease between
80 and 4.2°K.
The decrease in the intensity of spectrum H between 80 and 4.2°K can be explained
as the onset of magnetic hyperfine interactions at low temperature resulting in a concom-
itant decrease in the intensity of the central absorption doublet. This phenomenon has
been observed with Mossbauer spectroscopy of ferritin and hemosiderin (Fig. 6). However,
1.00
>-
!:: 0.95
(f)
z
w
.....
ú =0.90
0.85
-10.0 -5.0 0.0 5.0 10.0

VELOCITY (mmlseCl
Figure 10. Mossbauer spectrum of A. magnetotacticum wet packed cells at BOOK.
1.0
>-
ú =
(f)
Z 0.98
w
ú =
Z
0.96
- 5.0 0 5.0 10.0

VELOCITY (mm/sec)
Figure 11. Mossbauer spectrum of nonmagnetotactic cells at BOOK. Some residual Fe 3 04 is present in
the sample.
in the present case, the magnetic hyperfine lines are obscured by the magnetite spectral
lines (Ai and A 2). To further resolve the nature of the material responsible for spectrum
B, we studied the temperature-dependent Mossbauer spectra of nonmagnetotactic cells
which lacked the interfering magnetite.
For T ú = BOoK, the spectrum of lyophilized nonmagnetotactic cells (Fig. 11) consists
primarily of the quadrupole doublet characteristic of ferric iron as denoted by spectrum
B in Figs. 9 and 10. In addition, a very-low-intensity spectrum due to Fea04 (spectral lines
Ai + A2 in Fig. 9) is observed. These latter spectral lines might be due to a small fraction
of magnetotactic cells in the sample or trace amounts of magnetite possibly present in the
nonmagnetotactic cells. Below BOOK, the intensity of the quadrupole doublet decreased
with decreasing temperature while the intensity of a six-line spectrum flanking the doublet
increased. At 4.2°K, the spectrum (Fig. 12) consists primarily of the six broadened magnetic
hyperfine lines, with a small residual doublet in the center. The Fea04 spectrum was then
obscured by the six-line spectrum. Application of a longitudinal magnetic field of 60 kOe
produced broadening of the six-line spectrum but with no appreciable shifts in the line
positions and no decreases in any line intensities.
These spectral characteristics are indicative of small particles of hydrous ferric oxide
with antiferromagnetic exchange interactions similar to those of the ferrihydrite within
ferritin micelles. If we use values of K and to appropriate to ferritin, the experimental results
indicate that hydrous ferric oxide particles in the nonmagnetotactic cells are of the order
of 100 A in diameter, or less. Unlike ferritin or ferrihydrite, however, there is a residual
quadrupole doublet in the 4.2°K spectra of magnetotactic and nonmagnetotactic cells. The
intensity of this residual doublet varies somewhat from sample to sample, but its presence
suggests another high-spin ferric iron material with high-temperature spectral character-
istics similar to those of ferrihydrite, but with iron atoms less densely packed so that
magnetic exchange interactions between them are weaker and the spectrum is not mag-
netically split at 4.2°K, This latter material is more easily studied in a cloned, nonmag-
netotactic strain of A. magnetotacticum that accumulates less iron.
The Mossbauer spectrum of wet packed cells of the cloned, nonmagnetotactic strain
consists of a quadrupole absorption doublet for T;;::: 4.2°K (Fig. 13). The spectral parameters
obtained at BOOK were similar to those of spectrum B in magnetotactic cells (Table IJ,
284 Chapter 13
10
>-
ú =
U)
z
ú M KVVR=
z
0.99
-5.0 0 5.0
VELOCITY (mm/sec)
Figure 12. Miissbauer spectrum of nonmagnetotactic cells at 4.2°K.
indicating the presence of a high-spin ferric iron material. Application of an external 60-
kOe magnetic field at 4.2°K results in spectra with a broad distribution of hyperfine fields.
These spectral characteristics indicate the presence of high-spin Fe 3 + in a hydrous oxide
with magnetic exchange interactions of the order of 2-3°K, that is, where the iron atoms
are less densely packed than in ferrihydrite. This material has similar spectral character-
istics to the iron storage material in E. coli (Bauminger et 01., 1980).
When the wet packed cells were held above 275°K in an anaerobic environment, a
ferrous spectrum similar to spectrum C appeared, in addition to the ferric iron doublet.
This indicates that the hydrous ferric oxide in cells of this strain can be reduced to ferrous
iron as with cells of the other strains.
Diffusive motions of the magnetosomes in A. magnetotacticum have been observed
in the Mbssbauer spectrum of whole cells above 275°K. The temperature dependence of
>-
ú =
(/)
Z 098
w '
ú =
Z
092
- 4.0 0 4.0
VELOCITY (mm/sec)
Figure 13. Miissbauer spectrum of cloned. nonmagnetotactic strain of A. magnetotacticum at 4.2°K.

the ferrihydrite spectrum is consistent with the association of the ferrihydrite with the
magnetosomes. Ferrous iron in the cells appears to be associated with the peptidoglycan
of the cell wall (Ofer et 01., 1984).
The foregoing results suggest that A. magnetotacticum precipitates Fe304 in the se-
quence (1) Fe 3+ quinate ú =(2) low-density hydrous ferric oxide ú =(3) high-density hydrous
ferric oxide (ferrihydrite) ú =(4) Fe304 with Fe 2 + appearing as a transient between (1) and
(2). Because ferrihydrite contains ferric iron only, (3) ú =(4) implies reduction of one-third
of the iron atoms. The deposition of ferrihydrite and subsequent reductions and precipi-
tation of Fe304 occur in the magnetosome envelope. High resolution transmission electron
microscopy gives evidence for an amorphous (ferrihydrite) phase as well as single crystal
Fe304 with well defined morphology and orientation in the magneto somes (Mann et 01.,
1984).
Reduction of a ferrihydrite precursor to Fe304 occurs in the marine chiton, a mollusc
of the order Polyplacophora. In this organism, the radular teeth undergo a sequential mi-
neralization process that results in a surface coating of Fe304' Iron is transported to the
superior epithelial cells of the radula in the storage protein ferritin. Then, iron is transferred
to a preformed organic matrix on the tooth surface as ferrihydrite. Finally, the ferrihydrite
is reduced to Fe304' The resulting Fe304 particles have dimensions of the order of 0.1 11m
(Towe and Lowenstam, 1967; Kirschvink and Lowenstam, 1979).
Cells containing hydrous iron oxide granules have been found in the bands around
each abdominal segment in honeybees (Kuterbach et 01., 1982). Honeybees are sensitive
to the geomagnetic field and are known to have Fe304 in their abdomens (Gould et 01.,
1978). Some of the hydrous iron oxide in the granules could serve as a precursor to Fe304
formation.
Thus, Fe304 precipitation might follow similar pathways in a wide variety of orga-
nisms. In the bacteria, however, we have the best opportunity to elucidate the biochemical
details of the process and its connection to overall cellular metabolism.
ACKNOWLEDGMENTS. RB.F. and G.C.P. were partially supported by the Office of Naval Re-
search. The Francis Bitter National Magnet Laboratory is supported by the National Science
Foundation. RP.B. was supported by the Office of Naval Research and the National Science
Foundation.
References
Armstrong, R J., Morrish, A. H., and Sawatzky, G. A., 1966, Mtissbauer study of ferric ions in the
tetrahedral and octahedral sites of a spinel, Phys. Lett. 23:414-416.
Balkwill, D. 1., Maratea, D., and Blakemore, R P., 1980, Ultrastructure of a magnetotactic spirillum,
J. Bacterial. 141:1399-1408.
Bancroft, G. M., 1973, Mtissbauer Spectroscopy: An Introduction for Inorganic Chemists and Geoche-
mists, McGraw-Hill, New York.
Banerjee, S. K., O'Reilly, W., and Johnson, C. E., 1967, Mossbauer effect measurements in FeTi spinels
with local disorder, J. Appl. Phys. 38:1289-1291.
Bauminger, E. R, Cohen, S. G., Dickson, D. P. E., Levy, A., Ofer, S., and Yariv, J., 1980, Mossbauer
spectroscopy of E. coli and its iron storage protein, Biochim. Biophys. Acta 623:237-242.
Blaise, A., Chappert, J., and Givadet, J. 1., 1965, Observation par mesures magnetiques et effet Moss-
bauer d'un antiferromagnetisme de grains fins dans la ferritine, C. R. Acad. Sci. 261:2310-2313.
Blakemore, R P., 1982, Magnetotactic bacteria, Annu. Rev. Microbial. 36:217-238.
Blakemore, R P., and Frankel, R B., 1981, Magnetic navigation in bacteria, Sci. Am. 245(6):58-65.
Blakemore, R P., Maratea, D., and Wolfe, R S., 1979, Isolation and pure culture of a freshwater magnetic
spirillum in chemically defined medium, J. Bacterial. 140:720-729.
286 Chapter 13
Blakemore, R. P., Frankel, R. B., and Kalmijn, A. J., 1981, South-seeking magnetotactic bacteria in the
Chappert, J., Teillet, J., and Varret, F., 1979, Recent developments in high field Mossbauer spectros-
copy, J. Magn. Magn. Mater. 11:200-207.
Coey, J. M. D., and Readman, P. W., 1973, Characterization and magnetic properties of a natural ferric
gel, Earth Planet. Sci. Lett. 21:45-51.
Cohen, R. L. (ed.), 1976, Applications of Mosshauer Spectroscopy, Volume I, Academic Press, New
York.
Cohen, R. L. (ed.), 1981, Applications of Mosshauer Spectroscopy, Volume II, Academic Press, New
York.
Dezsi. I., Keszthelyi, L., Kulgawczuk, D., Molnar, B., and Eissa, N. A., 1967, Mossbauer study of J3 and
8-FeOOH, Phys. Status Solidi 22:617-629.
Forsyth, J. B., Hedley, I. G., and Johnson, C. E., 1968, The magnetic structure and hyperfine field of
goethite (8-FeOOH), J. Phys. C (Ser. 2) 2:179-188.
Frankel, R. B., 1982, Magnetotactic bacteria, Comments Mol. Cell. Biophys. 1:293-310.
Frankel, R. B., and Blakemore, R. P., 1980, Navigational compass in magnetic bacteria, J. Magn. Magn.
Mater. 15-18:1562-1564.
Science 203:1355-1356.
Frankel, R. B., Blakemore, R. P., Torres de Araujo, F. F., Esquivel, D. M. S., and Danon, J., 1981,
Frankel, R. B., Papaefthymiou, G. C., BlakeJllore, R. P., and O'Brien, W. D., 1983, Fe a04 precipitation
in magnetotactic bacteria, Biochim. Biophys. Acta 763:147-159.
Freeman, A. J., and Frankel, R. B., 1967, Hyper/ine Interactions, Academic Press, New York.
Gould, J. L., Krischvink, J. L., and Deffeyes, K. S., 1978, Bees have magnetic remanence, Science
201:1026-1028.
Greenwood, N. N., and Gibb, T. C., 1971, Mosshauer Spectroscopy, Chapman & Hall, London.
Hargrove, R. S., and Kundig, W., 1970, Mossbauer measurements of magnetite below the Verwey
transition, Solid State Commun. 8:303-308.
Johnson, C. E., 1969, Antiferromagnetism of 'Y-FeOOH: A Mossbauer effect study, J. Phys. C (Ser. 2)
2:1996-2002.
Kaufman, K. S., Papaefthymiou, G. C., Frankel, R. B., and Rosenthal, A., 1980, Nature of iron deposits
on the cardiac walls in ú J í Ü ~ ä ~ ë ë É ã á ~ = by Mossbauer spectroscopy. Biochim. Biophys. Acta
629:522-529.
Kirschvink. J. L., 1980, South-seeking magnetic bacteria, J. Exp. BioI. 86:345-347.
Kirschvink, J. L.• and Lowenstam, H. A.. 1979. Mineralization and magnetization of chiton teeth:
Paleomagnetic. sedimentologic and biologic implications of organic magnetite. Earth Planet. Sci.
Lett. 44:193-204.
Kuterbach. D. A., Walcott, B., Reeder, R. J.• and Frankel. R. B.. 1982. Iron-containing cells in the
honeybee (Apis mellifera). Science 218:695-697.
Lowenstam, H. A., and Rossman. G. R., 1975, Amorphous, hydrous, ferric phosphatic dermal granules
in Molpadia (Holothuroidea): Physical and chemical characterization and ecological implications
of the bioinorganic fraction, Chern. Geol. 15:15-51.
Mann, S., Frankel, R. B., and Blakemore. R. P .• 1984, Structure, morphology and crystal growth of
bacterial magnetite, Nature 310:405-407.
McNabb, T. K., Fox, R. A.• and Boyle, A. J. F., 1968. Some magnetic properties of magnetite (Fea04)
microcrystals, J. Appl. Phys. 39:5703-5711.
Moench, T. T., and Konetzka, W. A., 1978. A novel method for the isolation and study of magnetotactic
bacterium, Arch. Microhiol. 119:203-212.
Murad, E., and Schwertmann, V .• 1980, The Mossbauer spectrum of ferrihydrite and its relations to
those of other iron oxides. Am. Mineral. 65:1044-1049.
Ofer. S.. Nowik. I.. Bauminger. E. R, Papaefthymiou, G. C.• Frankel, R B., and Blakemore. R P., 1984,
Magnetosome dynamics in magnetotactic bacteria. Biophys. J. 46:57-64.
Ofer, S., Papaefthymiou. G. C.• Frankel, R. B., and Lowenstam, H. A.• 1981, Mossbauer spectroscopy
of iron-containing dermal granules from Molpadia intermedia, Biochim. Biophys. Acta 676:199-
204.
Oosterhuis, W. T., and Spartalian, K., 1976, Biological iron transport and storage compounds, in:
Applications of Mossbauer Spectroscopy, Volume I (R. L. Cohen, ed.), Academic Press, New York,
pp. 141-170.
van der Woude, F., 1966, Mossbauer effect in a-FezOa, Phys. Status Solid. 17:417-432.
Chapter 14
Magnetotactic Microorganisms Found
in Muds from Rio de Janeiro
A General View
HENRI QUE G. P. LINS DE BARROS and
DARCI MOTTA S. ESQUIVEL
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . " 289

2. The Geomagnetic Field. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 290
3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . " 291
3.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
3.2. Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
3.3. Analysis of the Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . " 305
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 308
1. Introduction
Among the many mechanisms used by living beings to secure information from the en-
vironment in the pursuit of survival, the perception of the geomagnetic field plays a unique
role. Weak, as compared to fields produced in the laboratory, but ever present, the geo-
magnetic field leaves behind records of its history in sediments, volcanic rocks, ceramic,
etc., and alters important characteristics of the biosphere. Present throughout the entire
process of the formation of species, its influence may be an important factor for the un-
derstanding of the behavior of living organisms.
The influence of "magnetic forces" on human behavior was considered for many cen-
turies, but after the rationalization of the concept of field at the end of the 18th century, *
magnetism lost its place in the study of living beings.
Studies in the 1950s of homing pigeons brought to light new information concerning
their navigational ability. It was hypothesized that these birds might be using the geo-
magnetic field as a compass, and this hypothesis was later confirmed (Walcott et a1., 1979;
* The study of many works reveals how the concept of magnetism has evolved (in the sense exposed
by Bachelard, 1966). Before the 17th century, the concept of magnetism (Plato; Aristotle; Hippocrates;
Lucretius; Gilbert, 1600) can be described as a naive realism. At the beginning of the 17th century
(Gilbert, 1600), we can notice a new approach that may be viewed as a clear empiricism. At the
beginning of the 19th century when the concept of magnetic field appeared (Whittaker, 1951), we
can identify a rationalist view of magnetism.
HENRIQUE G. P. LINS DE BARROS and DARCI MOTTA S. ESQUIVEL • Centro Brasileiro de Pes-
quisas Fisicas, CBPF/CNPq, Rio de Janeiro, Brazil.
289
290 Chapter 14
Gould, 1982). The effects of weak magnetic fields «5 gauss) on many different organisms
were studied and positive results were obtained, although the mechanisms of perception
involved were then unknown (Brown et 01., 1960; Palmer, 1963; Barnothy, 1969; Kirsch-
vink, 1981a, 1982). In spite of the macroscopic evidence, the effects of magnetic fields on
man have been observed only very recently (Baker, 1980a,b; Gould and Able, 1981; Kirsch-
vink, 1981b).
In 1975, Blakemore discovered bacteria that responded directly to the geomagnetic
field, moving in the direction of the lines of the magnetic field. This was the first unas-
sailable evidence that this field could directly influence the behavior of living beings.
The above considerations and some others outside the scope of this work suggest that
in science, the progress of one theory may block the development of other related areas
(Levi-Strauss, 1978). Most of all, they show that organisms are in complete harmony with
the environment, a harmony which is more than just ecological balance; that organisms
are sensitive to stimuli much weaker than science has imagined (Palmer, 1976; Schmidt-
Koenig and Keeton, 1978). The perception of weak fields by living beings is probably
related to the development of biological rhythms, which are extremely precise and are in-
herited. This perception requires a thorough examination, free from premature conclusions
(Gould, 1980), so that we can reach a better understanding of life on earth.
2. The Geomagnetic Field

The magnetic field is a three-dimensional vector field. In geomagnetism, it is usual
to represent the geomagnetic field vector in its declination, inclination, and amplitude
components. These components are defined in relation to an orthogonal frame of reference
whose axes are defined as functions of the geographical characteristics of the point on the
surface of the earth being considered. One of these axes coincides with the meridian at
this point and points to the geographical north. The other is perpendicular to the horizontal
plane tangent to the earth's surface and points to the center of the earth. The third is
perpendicular to the other two and points west. The magnetic declination is by definition
the angle between the projection of the geomagnetic field vector on the tangent plane and
the meridian. The magnetic inclination is the angle between the field vector and the tangent
plane. The amplitude or intensity is the magnitude of the geomagnetic field vector.
The deviation of the compass needle from the geographical north-south direction
suggested the need of defining and mapping the magnetic declination (Gilbert, 1600). The
first measurement of the magnetic inclination was made possible by Gilbert's empiricist
procedure in his works on magnetism and geomagnetism (Gilbert, 1600). Gauss's first meas-
urement of the amplitude of-the geomagnetic field in 1832 (Whittaker, 1951) became fea-
sible only after the rationalistic formulation of the concept of magnetic field. These are
three different instances in the history of the concept of magnetism, each one contributing
to the knowledge of one of the components of the field.
Gilbert's detailed study of the field produced by magnetized spheres (terrelas) showed
that the earth behaves like a magnetic dipole:-However, the reasons for the existence of
this field are still unknown. The analysis of a series expansion of spherical harmonic
functions of the field has shown that the main source of the geomagnetic field on the
surface of the earth is inside the earth. The external contributions are very small; they
account for daily variations and for some other effects. At present, the most widely accepted
model for the source of the geomagnetic field is a self-exciting hydromagnetic dynamo
inside the earth that produces an electric current which induces a magnetic field. This
model explains some of the fundamental characteristics of the earth's magnetism, but it
neither explains satisfactorily some important phenomena, such as the various reversions
Magnetotactic Microorganisms in Rio de Janeiro Muds 291
0.25
---------
--- ... -- - -- --- -- ----

Figure 1. The total intensity (gauss) of the geomagnetic field in Brazil. From Godoy (1981).
that the geomagnetic field has undergone in the last billion years (Opdyke, 1972), nor
describes the magnetic anomalies (Chernosky and Maple, 1961), regions where the earth's
field is very different from the field produced by a dipole. One of these regions, the vastest,
located near Rio de Janeiro, Brazil, is known as South Atlantic Magnetic Anomaly. In all
of Brazil's southeast region, the field values are very small in comparison to its values in
other regions of the earth.
Due to its extension, Brazil has peculiar field characteristics of the greatest interest
for the study of magnetism in living beings. In the north region of the country, the field
has positive inclination. Near Fortaleza, Ceara, in the magnetic equator, the field has null
inclination. At the South Atlantic Magnetic Anomaly region, the field values are below
0.25 G and the inclination is 25°. Throughout Brazil, the field magnitude values vary be-
tween 0.24 and 0.32 G (Godoy, 1981) (Fig. 1).
3. Results
3.1. Introduction
In this section, we describe the methods and techniques of sample preparation for
optical and electron microscopy of magnetotactic microorganisms found in muds of Rio
de Janeiro. We report the main results of the analyses of the characteristics of these mi-
croorganisms, as well as the size, number, and shape of the crystals which are responsible
292 Chapter 14
Figure 2. Detail of a coccus magnetotactic bacterium recovered from sediments of Rodrigo de Freitas
Lagoon. The high-density regions seem to be shaped as hexagonal prisms arranged in chains. The
arrow indicates a small high-density region at the end of the chain. Bar = 0.5 fLm.
for the magnetic properties of these microorganisms. The crystals are usually aggregated
in chains inside the cytoplasm of the cell. The existence of smaller particles at the ends
of the chain (Fig. 2) indicates that the mineralization process of magnetite formation in
the bacteria is under biological control and is genetically encoded (Blakemore, 1982; Fran-
kel, 1982). Frankel et a1. (1983) suggest that the enzyme activities that result in the for-
mation of magnetite are present in the magnetosome envelope (Frankel and Blakemore,
1984). The magnetite found inside these microorganisms is present in crystals with di-
mensions typical of single domains (Butler and Banerjee, 1975; Frankel and Blakemore,
1980). Particles with dimensions smaller than a single domain possess superparamagnetic
properties while the larger ones have multidomains structure. The magnetic moment of
single domains (below the Curie temperature) is the maximum magnetic moment of par-
ticles and is permanent. The identification of magnetite in chitons (Lowenstam, 1962),
bees (Gould et al., 1978, 1980; Kuterbach et aI., 1982)' pigeons (Walcott et al., 1979),
dolphins (Zoeger et al., 1981)' etc., shows that the biomineralization process of magnetite
formation is widespread among living organisms (Lowenstam, 1981).
The frequent and systematic observation of the waters of this region has revealed the
existence of a great number of different morphological types of south-seeking organisms,
many of which remain unidentified. We have found, in 6-J.Lm-diameter magnetotactic mi-
croorganism-rich samples from Rodrigo de Freitas Lagoon, some cells that exhibit char-
acteristics and behavior that suggest they are green algae of the Chlamydomonas genus
(Lins de Barros et a1., 1981). This is the first evidence of magnetotactic response in a
eukaryotic organism.
3.2. Techniques
The first step in the preparation of samples is to concentrate magnetically the living
magnetotactic microorganisms, attaching a magnet or a Helmholtz coil to the base of the
microscope (Blakemore and Frankel, 1981; Blakemore, 1982; Esquivel et a1., 1983). Very
high concentrations were obtained by keeping the samples in test tubes 30 cm long and
with diameters of 0.5 cm. In this case, after a few minutes, using Pasteur pipettes, sediment-
free magnetotactic microorganism-rich samples can be collected.
The material to be analyzed by transmission electron microscopy (TEM) was first
placed on a grid covered with collodion films (0.4% amyl acetate) and then fixed with
osmium tetroxide vapor.
For scanning electron microscopy (SEM), the samples were magnetically concentrated
over a slit with polY-L-lysine and then fixed in glutaraldehyde (2.5%) in phosphate buffer
(0.1 M) for more than 1 hr. Then they were washed repeatedly with phosphate buffer and
osmium acid (1 %) and dried by the CO 2 critical point technique. After the sputtering with
Au (200 A) we examined the slits with an optical microscope. It was thus possible to
observe the same microorganisms both optically and by SEM.
The glutaraldehyde-fixed, magnetically concentrated microorganisms were centr-
ifugued and dehydrated with increasing concentrations of acetone. The pellet was then
conventionally processed, placed in equal concentrations of Epon-acetone for 12 hr and
included in Epon. After the polymerization, ultrathin sections of the order of 1000 A were
made.
In order to film and observe optically these organisms, we used a Leitz-Ortholux
microscope with clear ground lighting, objectives of 10 to 100 x and oculars with mag-
nifications of 10 to 40.
We have analyzed the movement, determined the velocity, time, and radius of the U-
turn coupling cinematographical or video systems to the optical microscope.
It was thus possible to measure the mean velocity of migration as a function of the
external magnetic field (Teague et al., 1979).
Due to the small number of microorganisms per sample, it is very difficult to determine
the magnetic moment of magnetotactic microorganisms collected in a natural habitat. Ro-
senblatt et a1. (1982a,b) have measured the magnetic moment of bacteria in Aquaspirillum
magnetotacticum cultures using birefringence and light-scattering techniques. These tech-
niques, however, can only be used in culture samples (Frankel, 1984).
Electron microscopy, when feasible, produces images of chains of high-density regions
that can be identified, by spectroscopic techniques, as magnetite crystals (Towe and Low-
enstam, 1967; Frankel et al., 1979). If the chain is linear enough, the magnetic moment m
of bacteria can be estimated through the number and volume of these high-density regions:
m = volume of all crystals x saturation magnetization per unit volume (480 ergs/G cm3
for magnetite). If the organisms are just a few micrometers large, the images furnished by
TEM do not exhibit a good internal resolution, and so it is difficult to estimate m.
Another method of measuring the total magnetic moment is the U-turn method de-
veloped by C. Bean (personal communication). It relies upon the response of the organism
to the reversion of the magnetic field (Blakemore et al., 1979a). Subjected to a constant
magnetic field, magnetotactic microorganisms follow approximately a cylindrical helix
trajectory along the field lines. The stronger the field, the greater is the helical turn. When
the field is suddenly reverted, the microorganisms are subjected to a torque and the di-
294 Chapter 14
rection of the movement is reverted along a V-turn trajectory. The time T it takes to revert
the direction of the movement and the diameter L of the V-turn depend upon the total
magnetic moment of the organism and are given by
T = 81TTJR 3 In 2mBo (1)

mBo kT
(2)
where TJ is the viscosity (TJH20 = 10- 2 poise), k is the Boltzmann constant, and T is the
temperature.
Equations (1) and (2) are derived assuming that the organisms are spheres of radius
R that move at a velocity Vo independent of Bo in a viscous homogeneous environment,
and that all deviations from its rectilinear trajectory are due to Brownian motion. In the
case of living organisms, this model does not consider many aspects, particularly shape
characteristics and trajectory deviations due to mechanisms of movement (flagella, cilia,
etc.) associated with internal energies of living beings.
Even so, the measurement of the total magnetic moment by the V-turn technique is
the most general method to acquire information about the magnetic characteristics of these
organisms. To avoid contributions of biological energy to the movement, previously fixed
organisms can be used to measure the time T. Equation (1) can then be used to give the
total magnetic moment.
3.3. Analysis of the Results
We collected sedimented water samples from many Rio de Janeiro sites. We report
here the results of the analyses made of brackish waters from Rodrigo de Freitas Lagoon,
of marine waters from the head of Guanabara Bay, and of fresh waters from a small river
in the south of the city.
In all these sites, samples were collected from places where the waters are between
20 and 60 cm deep and exhibit a very slow motion. The samples were kept in the laboratory
at ambient temperatures with no chemical enrichment. After 4 or 5 days, the number of
magnetotactic microorganisms increased up to 10,000 cells/cm 3 • We tried some cultures
for these organisms, in particular those for A. magnetotacticum (Blakemore et 01., 1979b;
Escalante-Semerena et 01., 1980), with no positive results.
The strong oscillating magnetic field produced by a tape-recorder demagnetizer was
applied to all magnetotactic microorganisms. The original populations of mainly south-
seeking organisms then became 50% north-, 50% south-seeking organisms.
Because the waters from the Rodrigo de Freitas Lagoon are very much polluted with
organic material from the metropolitan area, they are very rich in microorganisms. Ex-
amination of water samples in the optical microscope revealed the existence of many
coccus and rod-shaped magnetotactic bacteria with sizes between 1 and 2 /-Lm and 2 and
3 /-Lm, respectively, which behave as described in the literature (Blakemore, 1975; Kalmijn
and Blakemore, 1978; Moench and Konetzka, 1978; Blakemore and Frankel, 1981; Esquivel
et 01., 1983).
We also observed some rod-shaped bacteria that respond differently to the reversal of
the magnetic field. When the field is reverted, these bacteria, which are concentrated at
the border of the drop, spin and move toward the center of the drop. After swimming for
some 20 /-Lm, they reverse the direction of movement without spinning and return to the
Magnetotactic Microorganisms in Rin de Janeiro Muds 295
Figure 3. Transmission electron micrograph of a magnetotactic coccus bacterium recovered from sed-
iments of Rodrigo de Freitas Lagoon. illustrating a linear chain of high-density regions inside the
cytoplasm. Bar = 111m.
border. These bacteria exhibit the same passive response to the magnetic field (Blakemore.
1975; Esquivel et 01. . 1983).
Observed by TEM. the coccus bacteria exhibit chains of high-density geometrically
shaped regions. From the micrographs, one can see that these regions seem to be shaped
as hexagonal prisms. Based on previous work (Towe and Lowenstam, 1967; Moench. 1978;
Frankel et 01 .• 1979; Towe and Moench, 1981), we believe that these regions are composed
of magnetite. The statistical distribution of the volumes of these regions is a narrow dis-
tribution centered at 5 x 10 - 16 cm 3 . The number of regions per cell. 8-10 for coccus
bacteria. is highly homogeneous in all the samples analyzed. In some samples. a great
number of bacteria with nonlinear chains (in L, S) (Figs. 3 and 4) were found . In very rich
fixed samples. the formation of chains of magnetotactic bacteria was detected. By TEM,
these chains display what seems to be an ordering of the crystal chains (Fig. 11).
We also observed the presence of greenish rounded magnetotactic microorganisms
with diameters between 4 and 7 J.lm that respond as bacteria but move and behave in a
different way. For fields of the order of the geomagnetic field. we cannot perceive any
magnetic orientation. For fields over 3 G, the trajectory is a cylindrical helix along the
field lines. and the velocity varies between 30 and 100 J.lm/sec. With magnifications over
500 x. they were seen to concentrate at the border of the drop and spin. This spinning
stopped when they swam in the middle of the drop or when the external field was greater
B
than 500 G. This may be an indication that the magnetic interaction energy for fields below
100 G is small in comparison to the energies (of biological origin) used for their movement.
The movement of the microorganisms is smoother and less aligned to the magnetic field
lines than the movement of the magnetotactic bacteria.
In rich samples of these microorganisms, we observed cells that exhibit a rounded to
pyriform shape, but never a fusiform contour, with 10-12 smooth undulations and no
blunt processes. When an iodine solution was used, it was possible to distinguish a cup-
shaped structure that we have identified as a chloroplast, a central blepharoplast for one
flagellum, and four to five pyrenoids. These characteristics have led us to suggest that these
cells are magnetotactic green algae of the Chlamydomonas genus (Lins de Barros et 01.,
1981).
We have met with two kinds of difficulties in our attempts to continue this research.
We have not managed to grow cultures, and the natural sample enrichment techniques
have not yet been mastered.
Based on frequent and continued optical observation, on the statistical analysis of the
velocity v and time T for the U-turn measurements, we have been led to admit the existence
of more than one type of magnetotactic microorganisms exhibiting these characteristics.
SEM results of such a sample seem to confirm this supposition. Figures 5 and 6 show the
morphology of two of these microorganisms: one exhibits a great number of microvillosities
and the other exhibits something of a globular structure disposed in a helicoidal arrange-
ment over a sphere. Nevertheless, it is possible that these two micrographs are of the same
microorganism at two different stages of development (Esquivel et 01., 1983).
Figure 7 is a transmission electron micrograph of one of these rounded microorga-
nisms. There are abundant fibers on the external surface, with about 1500 high-density
regions disposed in small aggregates, in parallel chains or displaying regular plane dis-
tributions in their interior. These are very well-defined regions, with many different regular
shapes, with dimensions between 400 and 800 A[inside the magnetic single-domain limits
of magnetite (Butler and Banerjee, 1975)]. Preliminary results from microanalysis of these
dense regions indicate the presence of iron, similar to the results of analysis of the mag-
netosomes of magnetotactic bacteria (Towe and Lowenstam, 1967; Balkwill et aI., 1980;
Towe and Moench, 1981). These regions must contain a percentage of magnetite (Fe304)
forming specialized structures that account for the observed magnetic orientation.
Ultrathin-section microscopy (Fig. 8) of one of these microorganisms shows more than
10 cells in a spiral arrangement and enveloped by an external coat. This coat exhibits a
radial periodic arrangement.
We observed some invaginations on the double outer membranes of some cells, as
well as specialized structures inside each cell, such as high-density regions, internal struc-
ture, and structures similar to "polar membranes" and organellas (Hickman and Frenkel,
1963; Remsen et 01., 1968; Moench, 1978; Moench and Konetzka, 1978). An enlargement
of the high-density regions indicates the possibility of an enveloping membrane and a
substructure composed of small grains of dense material. These regions are associated with
the observed magnetic orientation of these microorganisms, and must be organelles that
correspond to the magnetosomes found in magnetotactic bacteria. These characteristics
indicate that this microorganism is an unusual aggregate or colony of cells that exhibit
specialized structures inside each cell (Farina et 01., 1983; Remsen, 1983).
The study of waters (fresh and marine) collected in other sites has shown that mag-
netotaxy in microorganisms is a very general phenomenon, although they exhibit envi-
ronment-dependent responses.
Figure 4. Electron micrographs illustrating other distributions of high-density regions found in coccus
bacteria from Rodrigo de Freitas Lagoon (No.4 of Table IJ. (AJ L chain; (BJ S chain. Bars = 1 IJ.m.
298 Chapter 14
Figure 5. SEM image of magnetotactic microorganism (No.5 of Table I) found in Rodrigo de Freitas
Lagoon, presenting a globular structure disposed in a helicoidal arrangement. Bar = 5 11m.
Freshwater samples have disclosed four types of magneto tactic microorganisms living
in the same microhabitat. The main results obtained from these observations are shown
in Table I. Figure 9 is a transmission electron micrograph of a bacterium found in such
waters. We can see the crystal chains in the interior of the cell, in addition to some other
characteristics. Figure 10 is an optical image of some magnetotactic microorganisms found
in these waters.
In marine waters, there are mainly three types of magnetotactic microorganisms: coc-
cus bacteria with diameters of about 2 flm, greenish rounded microorganisms with di-
ameters of about 5 flm, and some ovoid organisms about 15 flm in size. In this last organism
the magnetic response, although direct and passive, is much more elaborate than those
previously described (Table I).
Figure 6. SEM image of magnetotactic microorganism (No.5 of Table I) found in the same sample
shown in Fig. 5. Note the great number of filaments around the surface of the microorganism. Bar =
5 11m.
Characteristics some of the microorganisms we have studied are shown in Table I:

average radius as obtained by means of optical and/or electron microscopy; average number
of high-density regions in their interiors; average velocities vo, U-turn times T and diameters
L for B = 9.3 G; magnetic moments as obtained by means of electron microscopy and/
or U-turn method; mBo/kT ratio and reversion time calculated for the local geomagnetic
field.
To calculate the magnetic moment by electron microscopy, we have assumed that
magnetite accounts for 80% of the volume of the dense regions (Towe and Lowenstam,
1967; Blakemore, 1982).
300 Chapter 14
Figure 7. Osmium vapor-fixed transmission electron microscopy of a whole magnetotactic microorganism sur-
rounded by fibers (Rodrigo de Freitas Lagoon. No.5 of Table J). (A) Abundant fibers on the external surface. Bar
= 5 fLm. (B) High-density grains containing iron. Bar = 1 fLm.
ú=
ú =
00
;:l
Table I. Characterization of the Magnetotactic Microorganisms Described in This PaperQ '"ú =
ú =
/l 'I .I (; III ( 10 I "11111 J ú =
iJI LJ n'
Sl'dilll,'nl ( I/) (-, ) TI __lllll 7 1 ""11 I. EII-Llron l/-Iurn (."Iul) ú=
úáäI >= Oplical IllIC rlJl-:r.lph ( J.l.f1l) ,\' ( J.l.lT1 isl'c) h"I.) (SI'/,) (J.l.f1l ) IlllCrcISU1P\' prw t'tiUfI' mil" kT (,., ) .. n'
..,
0
..,0
00
. ú =
Fr.'sh wilII'r ú K=
:.'. .. O.S .'i I IlII (UIH 011;' 8.
. ...... :1 O.'i n..! :1 () .5
'"
3
'"
2 5'
'"'"' ,.. r" Kú = 7 511 0 ..1 Il.l II 117 liS -1 lit ?:l
o·
...... 0-
3 1.5 )( Dú K R = 12 0 .'1 1..\ II
I...- Ii :!O '"
'0;"
;:l
::;.
'"
Rlltiril-:II Ill! 4 0
II .'I 10
• ••••
• () 0'1 1.-1 /I I;;
Fn'il,ls • ú=
c
J..lgllon 0-
5 '"
ú KN= > III!)!) ·10 1.. 1 1.-1 :10 :!.-I 'J;
ú I ä=
@ 17
(;u,lIldh.lr,1 B,II 6 II .'J 10 II , I t.:I I;'

/I
-• ...•
7 :! fl 70 n.J (1.-1 II /I -III lUI
:; • 'I ,\0 (1.-1 2. 1 2() .'i -l Uri 50

8
" R.mean radius determined by optical and/or electron microscopy; N, mean number of crystals found in the cytoplasm; 1', mean velocity at B = 9.3 G; Tcakd , reversion
time calculated by Eq. (1) with B = 9.3 G; Texp, mean revers ion time for the U-turn measured at B = 9.3 G; L, mean diameter of the U-turn at B = 9.3 G; m , estimated
"•
magnetic moment; mBo /kT. ratio between magnetic and th ermal energy at the local geomagnetic field (Bo = 0.25 G); TRo(calcd), reversion time for the U-turn calculated
by Eq. (1) with B = Bo = 0.25 G. W
0
ú =
302 Chapter 14
Figure 8. (Al and (Bl

Figure 8. Ultrathin sections of magnetotactic microorganism (No.5 of Table I) with high-density regions
suggesting that this microorganism is an unusual colony or aggregate of cells. (A) hd, high-density
regions containing iron; c, external coat exhibiting a radial arrangement; i, invagination; dm, double
membrane. Bar = 5 f1m. (B) pm, polar membrane; m, possible membrane around high-density region.
Bar = 0.1 f1m. (C) High magnification of grains showing possible substructure. Bar = 0.1 f1m. (D) 1.S.
Detail of internal structure found in some cells. Bar = 1 f1m . (E) c, external coating exhibiting a radial
arrangement. Bar = 1 f1m.
, ,
Figure 9. Transmission electron micrograph of a freshwater bacterium with two parallel chains of
dense regions. Bar = 111m.
Figure 10. Optical image of freshwater sample with three different microorganisms. Arrows show (1)
No.1, (2) No.2, and (3) No.3 of Table I. Bar = 10 IJ-m.
4. Conclusions
The various types of magnetotactic microorganisms found (Moench and Konetzka,
1978; Blakemore et a1., 1980; Kirschvink, 1980; Frankel et 01., 1981; Esquivel et 01., 1983)
and the diversity of behaviors observed in different waters from sites of different magnetic
fields [Fortaleza (Frankel et 01., 1981), Manaus, Salvador, Rio de Janeiro] have led us to
look at the phenomenon of magnetotaxy in a wider context.
Spaced magnetotactic microorganisms can be treated as noninteracting magnetic di-
poles. The average alignment of a magnetic dipole subjected to a field Bo is given by
(cos e) = L(mBo/kT) (3)
where L(x) = coth x + 1/x is the Langevin function of classic paramagnetism and mBol
kT is the ratio between the magnetic interaction energy and the thermal energy. For x ú =
1, L ú =0, and so the dipoles are weakly aligned to the field, while for x > 0, L - 1 and
we have an almost complete alignment ((cos e) - 1). We can also consider that they move
in their environment with a velocity Vo. The average migration velocity in the direction
of the field line (Teague et 01., 1979) is given by
(v) = Vo (cos e) (4)
where (cos e) is given by Eq. (3).

When the ratio mBo/kT is of the order of 1, the average migration velocity is about
30% of the instantaneous velocity; this means that even in this case there is a biological
advantage. Thus, in Rio de Janeiro (Bo = 0.25 G, inclination I =0 25°), for a microorganism
with mBo/kT = 1 (Le., mM = 1.6 x 10- 13 emu), we have (v) = 0.3vo, and the vertical
component of the migration velocity which gives the velocity with which the microor-
ganisms swim to the bottom is
Mv = (v) sin I - O.lvo
In Table I, the estimates of m show that for all the observed microorganisms, mBo/kT
> 1. These data indicate that magnetotaxy can be a more efficient mechanism than chem-
otaxy to produce a displacement toward the bottom (Frankel, 1982, 1984).
A magnetic moment inside these microorganisms produces a magnetic field around
them that decreases with the distance. The value of such a field one radius away from the
surface of these microorganisms should be greater than the value of the local field, so that
when the separation between these organisms is small enough, this interaction can be more
important than the interaction with the geomagnetic field.
Figure 11 shows coccus bacteria that have been fixed in the absence of a strong external
magnetic field. It seems to indicate the existence of magnetic interaction among these
bacteria in a configuration such as that obtained for magnetic crystals from A. magneto-
tacticum (Fig. 12). This magnetic interaction is probably masked when we observe living
microorganisms, for the energy of the movement, of biological origin, is much greater than
that due to interaction.
The same kind of bacteria, when collected from places with different geomagnetic
field, exhibit different shapes and numbers of magnetic crystals. It seems reasonable to
suppose that different geomagnetic field intensities should imprint magnetic moments such
that the product of a magnetic moment and the local geomagnetic field intensity would
be approximately constant.
The microorganisms shown in Figs. 5-8, with about 1500 nonaligned high-density
regions, suggest a new kind of mechanism to explain the magnetotaxic phenomenon. AI-
306 Chapter 14
Figure 11. Thirteen bacteria grouped in a distribution like that shown in Fig. 12. These coccus bacteria
have been fixed in the absence of a strong external magnetic field. Bar = 111m.
though these organisms have about 100 times more high-density regions than bacteria.
their magnetic moment is only about 10 times greater. Detailed analysis shows some or-
dering in the different aggregates of high-density regions into which these dense regions
are divided. This distribution may indicate another kind of sensibility to the magnetic
field. produced by a spatial distribution of magnetic dipoles that can detect spatial vari-
ations of the field .
On the other hand. the microorganism that uses the geomagnetic field as an orientation
mechanism must have some means to respond efficiently to variations of this field. or to
variations of their position in this field.
Hence. if there are external variations of the field that make the microorganisms deviate
from their trajectory. the correction has to be performed in a period of time shorter than
that between two consecutive perturbations of the environment. This means that the rev-
Figure 12. Crystals obtained in a sample of A. magnetotacticum culture from R. P. Blakemore. The
arrangement of these crystals is due to the magnetic interaction among them. Bar = 0.1 11m.
150
N MMú J J J J J J ú J J J J J J ú ú J J J J J J ú J J J J J J ú =
u
w
ú= a
w
ú=
i=
z
o
ii5
a:
w
>
w
a:
R Mú J J J J J J ú J J J J J J ú ú J J J J ú ú ú ú ú =
Figure 13. Relation between re-

version time (7) in the local geo-
magnetic field (Bo = 0.25 G) and
R3 of some magnetotactic micro-
organisms. Curve a, assuming m
= 1.6 x 10- 13 emu for all the
microorganisms and 7 calculated
from Eq. (1) . • , values of mean
reversion times obtained from
the estimated m in Table 1. ú I =ex- 0.1 10 100
pected region for the reversion
time (7).
ersal time has to be efficient from a biological perspective, that is, from the perspective
of the microorganism's life span and size.
As shown by Frankel and Blakemore (1980), magnetic orientation requires that the
magnetic interaction energy be greater than the thermal disorder energy, mBo/kT > 1. This
is true for magnetic moments of the order of mM = 1.6 x 10- 13 emu in the local geo-
magnetic field (Bo = 0.25 G). The reversal time T increases with the cube of the radius of
the microorganism for a fixed value of m [Eq. (1)]. For larger organisms, the rapid increase
of T at constant mM would make the response to magnetic stimuli inefficient. Curve a of
Fig. 13 shows these values calculated for mM and Bo = 0.25 G. Analysis of the data given
in Table I shows that m increases with the volume of the microorganism, implying a
decrease in T. Figure 13 shows the values of T for these values of m with Bo = 0.25 G.
These results lead us to believe that an upper limit to the size of the organisms should
exist, beyond which magnetotaxy would no longer be an efficient orientation mechanism.
Magnetotaxy seems to be an orientation mechanism that is effective when (1) the
magnetic interaction energy is much greater than the thermal energy and (2) the time
interval necessary for the torque produced by the geomagnetic field to orient the microor-
ganism is much smaller than the cycles of perturbations that may occur in their habitat.
308 Chapter 14
ACKNOWLEDGMENTS. We wish to ,thank Professor Jacques Danon, for discussion and partic-
ipation on this subject; Dr. R. B. Frankel, for introducing us to this subject and for con-
tinuous information; and Dr. R. P. Blakemore, for gifts of materials, and communications.
All three colleagues were a source of encouragement, inspiration, and ideas. To the late
Dr. 1. P. H. Oliveira, for special attention and a beautiful example of life; Marcos Farina,
for electron microscopy work; Drs. Wanderley de Souza and R. Machado, for discussions;
Roberto Eizemberg, for much work in our laboratory; and Marlene B. Mello, for preparation
of the manuscript.
References
Bachelard, G., 1966, La philo sophie du non, 4th ed., Presses Universite Paris, France.
Baker, R. B., 1980a, A sense of magnetism, New Sci. 87:844-846.
Baker, R. B., 1980b, Goal orientation by blind-folded human after long-distance displacement: Possible
involvement of a magnetic sense, Science 210:555-557.
Balkwill, D. L., Maratea, D., and Blakemore, R. P., 1980, Ultrastructure of a magnetotactic spirillum,
J. Bacteriol. 141:1399-1408.
Barnothy, M. F. (ed.), 1969, Biological Effects of Magnetic Fields, Volume 2, Plenum Press, New York.
Blakemore, R. P., 1982, Magnetotactic bacteria, Annu. Rev. Microbial. 36:217-238.
Blakemore, R. P., and Frankel, R. B., 1981, Magnetic navigation in bacteria, Sci. Am. 245(6):42-49.
Blakemore, R. P., Frankel, R. B., and Wolfe, R. S., 1979a, Ferromagnetism in freshwater bacteria, in:
Biological Effects of Extremely Low Frequency Electromagnetic Fields, Technical Information
Center, U.S. Department of Energy, Washington, D.C., pp. 400-407.
Blakemore, R. P., Maratea, D., and Wolfe, R. S., 1979b, Isolation and pure culture of a freshwater
magnetic spirillum in chemically defined medium, J. Bacteriol. 140:720-729.
Blakemore, R. P., Frankel, R. B., and Kalmijn, A. J., 1980, South-seeking magnetotactic bacteria in the
Brown, F. A., Jr., Britt, W. J., Benett, M. F., and Barnvell, F. H., 1960, Magnetic response of an organism
and its solar relationships, BioI. Bull. 118:367-381.
Chernosky, E. J., and Maple, E., 1961, Geomagnetism, in: USAF Handbook of Geophysics, Macmillan
Co., New York.
Escalante-Semerena, J. c., Blakemore, R. P., and Wolfe, R. S., 1980, Nitrate dissimilation under mi-
croaerophilic conditions by a magnetic spirillum, Appl. Environ. Microbiol. 40:429-430.
Esquivel, D. M. S., Lins de Barros, H. G. P., Farina, M., Aragao, P. H. A., and Danon, J., 1983, Mi-
croorganismes magnetotactiques de la region de Rio de Janeiro, BioI. Cell. 47:227-234.
Farina, M., Lins de Barros, H. G. P., Esquivel, D. M. S., and Danon, J., 1983, Ultrastructure of a mag-
netotactic microorganism, BioI. Cell. 48:85-86.
Frankel, R. B., 1982, Magnetotactic bacteria, Comments Mol. Cell. Biophys. 1:293-310.
Frankel, R. B., 1984, Magnetic guidance of microorganisms, Annu. Rev. Biophys. Bioeng. 13:85-103.
Frankel, R. B., and Blakemore, R. P., 1980, Navigation compass in magnetic bacteria, J. Magn. Magn.
Mater. 15-18:1562-1564.
Frankel, R. B., and Blakemore, R. P., 1984, Precipitation of Fea04 in magnetotactic bacteria, Philos.
Trans. R. Soc. London Ser. B 304:567-574.
Science 203:1355-1356.
Frankel, R. B., Blakemore, R. P., Torres de Araujo, F. F., Esquivel, D. M. S., and Danon, J., 1981,
Frankel, R. B., Papaefthymiou, G. c., Blakemore, R. P., and O'Brien, W. D., 1983, Fe a04 precipitation
in magnetotactic bacteria, Biochim. Biophys. Acta 763:147-159.
Gilbert, W., 1600, De magnete, magneticisque corporibus et de magno magnete tellure [On the load-
stone and magnetic bodies}, in: Great Books of the Western World, Volume 28 (R. M. Hutchins,
ed.), Encyclopaedia Britannica, London.
Godoy, R C., 1981, Variaciio secular da intensidade de campo geomagnetico no Brasil, Unpublished
Ph.D. dissertation, Inst. Geocifmcias UFRJ.
Gould, J. 1., 1980, The case for magnetic sensitivity in birds and bees (such as it is), Am. Sci. 68:256-
267.
Gould, J. L., 1982, The map sense of pigeons, Nature 296:205-211.
Gould, J. 1., and Able, K. P., 1981, Human homing: An elusive phenomenon, Science 212:4498-4500.
Gould, J. 1., Kirschvink, J. L., and Deffeyes, K. S., 1978, Bees have magnetic remanence, Science
201:1026-1028.
Gould, J. L., Kirschvink, J. 1., Deffeyes, K. S., and Brines, M. L., 1980, Orientation of demagnetized
bees, J. Exp. Biol. 86:1-8.
Hickman, D. D., and Frenkel, A. W., 1963, Observations on the structure of Rhodospirillum rubrum,
J. Cell Biol. 25:271-291.
Kalmijn, A. J., and Blakemore, R p . , 1978, The magnetic behavior of mud bacteria, in: Animal Mi-
gration, Navigation, and Homing (K. Schmidt-Koenig and W. T. Keeton, eds.), Springer-Verlag,
Kirschvink, J. L., 1980, South-seeking magnetic bacteria, J. Exp. Biol. 86:345-347.
Kirschvink, J. 1., 1981a, The horizontal magnetic dance of the honeybee is compatible with a single-
domain ferromagnetic magnetoreceptor, Biosystems 14:193-203.
Kirschvink, J. L., 1981b, Ferromagnetic crystals (magnetite?) in human tissue, J. Exp. Biol. 92:333-
335.
Kirschvink, J. 1., 1982, Birds, bees and magnetism, Trends Neurosci. 5:160-167.
Kuterbach, D. A., Walcott, B., Reeder, R J., and Frankel, R B., 1982, Iron-containing cells in the
honeybee (Apis mellifera), Science 218:695-697.
Levi-Strauss, C., 1978, Myth and Meaning, University of Toronto Press, Toronto.
An. Acad. Bras. Cienc. 54:258.
Am. Bull. 73:435-438.
Moench, T. T., 1978, Distribution, isolation and characterization of a magnetotactic bacterium, Un-
published Ph.D. dissertation, Indiana University.
Moench, T. T., and Konetzka, W. A., 1978, A novel method for the isolation and study of magnetotactic
bacterium, Arch. Microbiol. 119:203-212.
Opdyke, N. D., 1972, Paleomagnetism of deep-sea cores, Rev. Geophys. Space Phys. 10:213-249.
Palmer, J. D., 1963, Organismic spatial orientation in very weak magnetic fields, Nature 198:1061-
1062.
Palmer, J. D. (ed.), 1976, An Introduction to Biological Rhythms, Academic Press, New York.
Remsen, C. C., 1983. Structural attributes of membrane organelles in bacteria. lnt. Rev. Cytol. 79:195-
223.
Remsen, C. c., Watson, S. W., Waterbury, J. B., and Triiper, H. G., 1968, Fine structure of Ectothior-
hodospira mobilis Pelsh, J. Bacteriol. 95:2374-2392.
Rosenblatt, C., Torres de Araujo, F. F., and Frankel, R B., 1982a, Light scattering determination of
magnetic moment of magnetotactic bacteria, J. Appl. Phys. 53:2727-2729.
Rosenblatt, C., Torres de Araujo, F. F., and Frankel, R. B., 1982b, Birefringence determination of
magnetic moments of magnetotactic bacteria, Biophys. J. 40:83-85.
Schmidt-Koenig, K., and Keeton, W. T. (eds.), 1978, Animal Migration, Navigation, and Homing.
Springer-Verlag, Berlin.
Teague, B. D., Gilson, M., and Kalmijn, A. J., 1979, Migration rate of mud bacteria as a function of
magnetic field strength. Biol. Bull. 157:399.
Towe, K. M., and Lowenstam, H. A., 1967, Ultrastructure and development of iron mineralisation in
the radular teeth of Cryptochiton stelleri (Mollusca), J. Ultrastruct. Res. 17:1-13.
Towe, K. M., and Moench, T. T .• 1981, Electron-optical characterization of bacterial magnetite, Earth
Walcott, c., Gould, J. 1., and Kirschvink, J. 1., 1979, Pigeons have magnets, Science 205:1027-1029.
Whittaker, E., 1951, A History of the Theories of Aether and Electricity, Nelson, London.
Zoeger, J., Dunn, J. R, and Fuller, M., 1981, Magnetic material in the head of the common Pacific
Chapter 15
Structure, Morphology, and Crystal
Growth of Bacterial Magnetite
STEPHEN MANN
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
2. Instrumentation: High-Resolution Transmission Electron Microscopy. . . . . . 312
3. Materials and Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
3.1. Aquaspirillum magnetotacticum Cells. . . . . . . . . . . . . . . . . . . . . . . . . . . 313
3.2. Magnetotactic Coccoid Cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 314
3.3. Electron Microscopy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
4. Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
4.1. Microstructure, Morphology, and Crystallographic Alignment of Mature
Crystals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
4.2. Immature Crystals . . . . . . . . . . . . . . . . . . . . . . . " . . . . . . . . . . . . . 319
5. Discussion: Bioprecipitation of Bacterial Magnetite. . . . . . . . . . . . . . . . . . . . . . .. 323
5.1. Nucleation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
5.2. Crystal Growth and Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
5.3. Spatial Control of Bacterial Magnetite Formation . . . . . . . . . . . . . . . . . 326
5.4. Chemical Control of Bacterial Magnetite Formation. . . . . . . . . . . . . . . . 327
5.5. Structural Control of Bacterial Magnetite Formation . . . . . . . . . . . . . . . . . . . . 329
6. Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 331
1. Introduction
The biological synthesis of the mixed-valence iron oxide magnetite (Fe 3 04l poses several
fascinating and intriguing questions for the solid state and materials scientist. Preparation
of magnetite in the laboratory requires an experimental regime which often includes high
temperature, high pressure, and high pH. A further aspect of inorganic synthesis is the
experimental difficulty involved in the preparation of particles of homogeneous shape and
size. Biological reactions, however, are characterized by their selectivity and precision in
functional design. In this way biosolid-state reactions can arise which establish the for-
mation of magnetite under conditions of ambient temperature and pressure and at pH
values close to neutral. Furthermore, precise biological control over the activation and
regulation of the solid-state processes can result in magnetite particles of well-defined size
and crystallographic morphology. These aspects of chemical, stereo, and crystallographic
specificity in biosolid-state reactions are important criteria which need to be investigated
STEPHEN MANN • School of Chemistry, University of Bath, Bath BA2 7AY, United Kingdom
311
312 Chapter 15
and elucidated, as knowledge concerning such novel solid-state processes may be impor-
tant in the development and design of new materials for technological use.
In a recent overview (Mann, 1983) concerning the solid-state factors involved in biom-
ineralization processes, the importance of biological control over the chemical, spatial,
and structural dimensions of bioprecipitation has been emphasized. These factors are also
applicable in the understanding of bacterial magnetite formation. The aim of this article
is to discuss such a rationalization in the light of recent structural and morphological
studies of bacterial magnetite using the technique of high-resolution transmission electron
microscopy (HRTEM).
2. Instrumentation: High-Resolution Transmission Electron

Microscopy
HRTEM has become an established method for investigating the local structure of
crystalline, pseudocrystalline, and amorphous materials at the nanometer level. It has been
applied to the study of a wide variety of inorganic minerals (Hutchison et al., 1977) and,
more recently, to the investigation of biogenic inorganic solids such as calcium carbonate
(Mann et al., 1983), silica (Mann and Williams, 1982), and iron oxide (Fe304) (Mann et
al., 1984a,b) as described in this report.
The general applications of HRTEM in solid-state chemistry are summarized in Fig.
1. The general aim is to investigate the structure and properties of materials through direct
imaging of the crystal lattice in the electron microscope. In this way information concerning
the nucleation and growth of materials can be studied and deviations from ideality in
structure and stoichiometry elucidated.
The potential of HRTEM lies in the principle that for crystals of sufficient thickness
(often less than 100 A), the phase contrast observed in an electron micrograph recorded
under specific conditions of defocus (Schwerzer defocus) is closely related to the projection
of the atomic potential distribution in the crystal. In such ideal circumstances there is a
one-to-one correspondence between the object under study and the recorded image. How-
ever, in practice, the interpretation of high-resolution electron micrographs is more difficult
due to the sensitivity of such images to microscope aberrations, sample thickness (multiple
scattering effects), and defocus conditions. These factors can be accommodated in theo-
retical calculations of image contrast such that the experimental and simulated images can
be directly compared allowing evaluation of the information transferred from the object
to the recording plate in the electron microscope.
The combination of HRTEM and selected-area electron diffraction (or other electron
microdiffraction techniques) provides a means of studying individual microcrystals such
as the magnetite particles synthesized in magnetotactic bacteria. In this way information
concerning the nucleation and growth of individual crystals can in principle be deduced.
No such information can be obtained from conventional structure-determination tech-
niques such as X-ray diffraction, for the data obtained are statistical in nature. However,
care must be taken in the inference of general mechanisms of crystal formation from lattice
images, for not only may there be problems in interpretation but only a small fraction of
the total sample population will be chosen for investigation and questions concerning the
selectivity of the experimental data must be addressed.
3. Materials and Methods
Two different sources of bacterial magnetite have been studied by HRTEM.

Bacterial Magnetite 313
I HRTEM
I
I
..v {t
Geometrical Structural
information information
I
.J, w ú= \It
Size Morphology Crystalline Noncrystalline
of of materials materials
microparticles microparticles
I J v "
Local disorder
(0, CS, OS, SF)
Local
order j
-!- J.,
Nucleation
Nonstoichiometry
and
crystal growth
I
-!t
I
Chemical
,\. information ./
on solids
Figure 1. Flow diagram showing the general application of HRTEM in the study of solids. D = dis-
locations, CS = crystallographic shear, DS = domain structures (twin and intergrowth boundaries),
SF = stacking faults.
3.1. Aquaspirillum magnetotacticum Cells
A. magnetotacticum is a microaerophilic bacterium which can be grown in pure cul-

ture (Blakemore et aI., 1979; Maratea and Blakemore, 1981). Low-resolution electron mi-
croscopy has shown the cells to be approximately 3 !-Lm in length and to contain, on average,
20 intracellular enveloped magnetite particles (magnetosomes) of diameter 40-50 nm
which are organized in a single chain that traverses the cell longitudinally (Balkwill et aI.,
1980). Cells of A. magnetotacticum synthesize magnetite only under microaerobic con-
ditions, accumulating Fe some 20,000- to 40,ooo-fold over the extracellular concentrations
(Blakemore et aI., 1979). The magnetite particles are in the single magnetic domain size
range and the chain of magnetosomes imparts a permanent magnetic dipole moment par-
allel to the cell's axis of mobility such that the bacteria orient in the geomagnetic field
(Frankel and Blakemore, 1980).
A recent Mossbauer spectroscopy study of this organism (Frankel et aI., 1983) indi-
cated, in addition to magnetite, the presence of hydrated ferric oxide phases together with
the magnetosomes. Because such data are statistical in nature and reveal no information
concerning the local structure and organization of the hydrated ferric oxide phase, HRTEM
has been used to locate and investigate the interrelationships of the two iron oxide phases
within individual magnetosomes.
314 Chapter 15
The experimental strategy involved the elucidation of crystallographic structure, de-

gree of order, morphology, and crystal growth by the direct lattice imaging of magnetite
crystals within intact bacterial cells. The study of crystals in situ would then eliminate
many of the potential artifacts associated with the isolation techniques and would permit
a wide range of particles of different stages of development to be examined. Also, infor-
mation concerning the crystallographic orientation and extent of crystallization in anyone
single chain could be evaluated.
3.2. Magnetotactic Coccoid Cells
These cells were not available in culture and were isolated from a complex simulated
natural environment described in detail elsewhere (Moench and Konetzka, 1978). To sep-
arate the magnetite particles, the procedure described by Towe and Moench (1981) was
employed. A portion of the aggregated magnetite sample was treated overnight with 5.25%
sodium hypochlorite solution at room temperature to remove any organic material which
may have adhered to the magnetite particles. The magnetite dispersion was then repeatedly
centrifuged and resuspended in distilled water.
Due to availability, no crystals within intact cells were studied. Thus, the information
obtained concerning crystal growth within this biological system was derived only from
studies on isolated crystals.
For studies on A. magnetotacticum, intact unfixed cells were dried down in air onto
carbon-coated, Formvar-covered, copper electron microscope grids (2.3 mm). In the in-
vestigation of crystals from magnetotactic coccoid cells, the isolated aggregated magnetite
particles were redispersed ultrasonically in distilled water and a drop of the suspension
air-dried onto grids as above. Both sets of experiments were undertaken using a JEOL 200CX
electron microscope fitted with a high-brightness LaB 6 cathode and which has a point-to-
point resolution of 2.5 A. All experiments were carried out at 200 keY with an objective
aperture diameter of 40 f.Lm.
Crystals were imaged at several different defocus conditions such that a "through-
focus" series of micrographs was recorded, with the optimum defocus position at - 650
A. In this way the change of image structure with focus position could be evaluated. How-
ever, both sets of biogenic crystals were generally too thick for two-dimensional "structure-
images" to be obtained except at very thin edges, and in most particles the resolution was
such that only one-dimensional lattice fringes could be imaged. These fringes, such as
those shown in Fig. 2, represent lattice planes oriented parallel to the electron beam and
are indexed in the normal crystallographic way by Miller indices in the form (hkl). As
magnetite has a cubic symmetry, many lattice planes such as (100), (010), and (001) are
equivalent; in such cases, the set are indexed in the form {hkl}. The spacings between
fringes of anyone set and the angles between different sets allow the identification of the
crystal projection within the electron microscope, indexed in the form [uvw). By studying
many different crystals in different orientations in the electron microscope, the three-
dimensional morphology of the particles can be determined as discussed in Section 4.1.
Lattice fringes will only be observed when they lie in the Bragg orientation for dif-
fraction and thus the absence of lattice images in a recorded image does not necessarily
imply an amorphous structure but often an unaligned sample. Regions of local disorder
within periodic arrays of atoms can be recognized by modulation of the lattice fringes
although variations in crystal thickness and local crystal bending can have similar effects.
Figure 2. HRTEM image of a magnetite crystal imaged in the [0111 zone showing a well-ordered single-
domain crystal and a characteristic morphology based on an octahedral prism of {111} faces truncated
by {100} faces (Fig. 4). The lattice fringes shown correspond to the (022) planes and run perpendicular
to the (100) face. Note that the crystal edges are not smooth and show outgrowths on the well-developed
{111} faces (arrows) . Bar = 10 nm.
4. Results
4.1. Microstructure, Morphology, and Crystallographic Alignment of
Mature Crystals
Magnetite crystals grown within A. magnetotacticum and magnetotactic coccoid cells

both show particles of characteristic shape and size. This section reports the data con-
cerning crystals which show mature crystallographic habits and it must be stressed that
in both sets of crystal populations there were many crystals which had not obtained this
characteristic morphology. Thus, the morphologies described below are idealized con-
structs and may only apply to biogenic crystals of substantial age. However, they are in-
dicative of the modes of crystal formation in these biological systems.
316 Chapter 15
Figure 3. Selected-area electron diffraction

pattern from a mature magnetite crystal of
characteristic morphology. The single-crystal
pattern corresponds to the [011J zone. The
(200) and (ZOO) reflections arise from double
diffraction.
4.1.1. A. magnetotacticum Cells

Crystals (ca. 50 nm in size) of distinct morphology showed lattice images consistent
with the cubic (Fd3m) inverse spinel structure of magnetite. Consider Fig. 2 which is
representative of the many crystals studied. The image shows lattice fringes corresponding
to the (022) planes which are well defined and run continuously throughout the particle.
There are no breaks or distortions in the fringes, indicating that the particle has a high
degree of perfection and is a single-crystal domain. In other crystals, lattice fringes cor-
responding to the {111}, {222}, {220}, {200}, and {400} planes were imaged and in all cases
the crystals were single domain and free of dislocations and stacking faults. By considering
the relative orientation of these fringes, the zone of projection in Fig. 2 can be identified
as [011] . This zone was observed in many mature crystals and was confirmed by selective-
area electron diffraction patterns (Fig. 3). Thus, the crystal edges seen in Fig. 2 are crystal
faces aligned perpendicular to the zone of projection and can be identified as shown. Note
that the (022) fringes run perpendicular to the (100) face as expected.
The morphology of these mature crystals can be inferred from the relative directions
of the fringes and crystal edges and is based on an octahedron of {111} faces truncated by
{100} faces (Fig. 4). Thus, crystals imaged in the [011] zone often showed edges meeting at
characteristic angles of 125 and 110° corresponding to the intersection of (100) and (111),
and (111) and (111) faces. In other mature crystals, these angles were distorted from their
theoretical value which may be a consequence of slight crystal misalignment with respect
to the electron beam or it may be a real growth effect.
On the basis of the above morphology, the crystal alignment in A . magnetotacticum
cells containing chains of well-developed crystals was determined. The crystals were found
to be preferentially aligned with the [111] direction parallel to the chain axis (Fig. 5). This
result is important with regard to magnetotaxis, as the easy axis of magnetization in Fe304
also lies along the [111] direction.
4.1.2. Magnetotactic Coccoid Cells

Magnetite crystals isolated from coccoid cells have been shown by low-resolution
electron microscopy to be parallelepiped crystals with a mean length of ca. 100 nm and
Figure 4. Idealized morphology for biogenic magnetite crystals from

A. magnetotacticum. An octahedral prism of {Ill} faces is truncated
by {I oo} faces . Such a crystal form indicates the stabilization of the
{Ill} planes over other low-index faces such as {100}.
mean width of ca. 60 nm (Towe and Moench, 1981). High-resolution images showed rec-
tangular crystals with truncated faces and lattice fringes which were consistent with the
cubic structure of magnetite. Lattice fringes corresponding to the {111} (4.8 A) , {220} (2.9
A), {200} (4.2 A) , and {311} (2.5 A) planes were observed from crystals before and after
hypochlorite treatment, the {111} fringes being observed in most crystals. Three different
zone directions were imaged. In the majority of crystals, lattice images corresponding to
the [011) zone were observed (Fig. 6). Note how different the crystal morphology is com-
pared with the [011) zone for magnetite crystals in A. magnetotacticum (Fig. 2). This, and
other zones, were confirmed by selected-area diffraction patterns and by measurement of
Figure 5. Alignment of mature magnetite crystals in A. magnetotacticum . The {111} faces are aligned
perpendicular to the chain axis (arrow) . Thus, the chain axis corresponds to the [111] direction of the
magnetite lattice which is also the easy axis for magnetization. Bar = 40 nm.
318 Chapter 15
Figures 6-9. Electron micrographs of bacterial magnetite from coccoid cells imaged at high resolution
in different crystal zones. For each zone the particles are well-ordered single crystals, rectangular in
shape with truncated faces.
Figure 6. [011] zone of magnetite. The (011) face is truncated by (011) and (100) faces as shown. Bar
= 20 nm.
the angles between different lattice planes. Figure 7 shows an enlarged image from the
edge of a magnetite crystal in the [011] zone showing (111), (111), and (200) planes in the
expected directions. The white dots are not Fe atoms but columns of holes aligned parallel
to the electron beam. In other crystals, the (211] (Fig. 8) and [100] (Fig. 9) directions were
imaged. In each of the three zones, lattice fringes were observed to run continuously across
the crystals, indicating that the particles were single-domain crystals as for those imaged
in A. magnetotacticum. The degree of crystal perfection in these images was exceedingly
high and stacking faults and dislocations were not observed.
The idealized morphology of the crystals was determined from the identification and
measurement of crystal faces in relation to the imaged zone axis. Simulated diffraction
patterns (Skarnulis, 1979) were used to aid crystal face indexing. The identified faces are
shown in Figs. 6, 8 , and 9. The three different shapes corresponding to the three different
zone directions can be rationalized for a crystal of hexagonal habit in the [111] direction,
truncated by a threefold centro symmetric arrangement of {100} and {011} planes as shown
in Fig. 10. Morphologically, the magnetite is prismatic and appears to have '3 21m symmetry.
However, the high-resolution electron microscope images are consistent with the cubic
structure of magnetite, and indices in Fig. 10 have been assigned in accordance with cubic
Figure 7. Enlarged micrograph of the [011] zone showing lattice fringes corresponding to the (111),
(111), and (200) planes. Bar = 5 nm.
symmetry. Thus, crystals imaged in the [011] zone showed centro symmetric truncated faces
of characteristic angles 145 and 125 0 to the (111) plane corresponding to the (011) and
(100) planes (Fig. 7), respectively. Images observed in the [211] zone showed rounded
corners (Fig. 8), as there are no low-index faces which can truncate the prism in this zone.
Crystals imaged in the [100] direction (Fig. 9) showed four truncated faces with vectors at
an angle of 45 to the [011] direction which correspond to the (010) and (001) faces.
0
In summary, HRTEM images have shown that in both biological systems studied,
magnetite crystals of mature crystal habit are synthesized and in A. magnetotacticum the
crystals are crystallographic ally oriented within a chain. These crystals, although highly
perfect and single domain in both systems, have very different but specific idealized mor-
phologies. These results are of upmost importance in the elucidation of mechanisms of
biological control of crystal development in these bacterial cells.
4.2. Immature Crystals
Crystals which did not show the well-defined crystallographic habits discussed in
Section 4.1 have also been studied by HRTEM. Such crystals should gi ve important insights
into the development of crystal formation within the bacterial cells, as they may be rep-
resentative of early or incomplete stages of crystal growth.
320 Chapter 15
Figure 8. [211) zone of magnetite. Rounded corners were generally observed in this zone. Lattice fringes
imaged are (022) planes. Bar = 20 nm.
4.2.1. A. magnetotacticum Cells

Figure. 11 shows a chain of magnetite crystals imaged within an intact cell. Unlike
many chains which showed crystals of mature habit in regular alignment, this chain com-
prises crystals with marked variations in size and morphology ranging from particles with
characteristic cubo-octahedron morphologies (crystal a in Fig. 11) through distorted forms
to very irregular particles of sizes in the range of 30 nm (particle B in Fig. 11). Many chains
of this type have been studied. No linear sequence of crystallographic development along
the chain has been observed although crystals at the ends of the chains often appeared to
be smaller in dimension.
Lattice imaging of the irregular particles has shown the presence of contiguous crys-
talline and noncrystalline regions within magnetosomes of this type (Fig. 12). The crys-
talline zone was always observed to be single domain with well-ordered lattice planes of
magnetite. No other crystalline phases, such as -y-FeOOH, have been observed. As can be
seen in Fig. 12, the lattice fringes often appeared to extend into the amorphous region in
a preferential direction which may indicate a preferred nucleation and growth direction.
These multi phase particles probably represent the early stages of magnetite formation with
the noncrystalline material corresponding to the hydrated ferric oxide phases identified
by Mossbauer spectroscopy (Frankel et 01., 1983).
Figure 9. [100] zone of magnetite. The crystal appears truncated symmetrically by four {DOl} faces.
Lattice fringes. imaged are (020) planes. Bar = 20 nm.
101 .0'" 110
,}", ,
" .........
/' ..
Figure 10. Idealized crystal morphology for

magnetite particles from magnetococcoid
bacterial cells. The crystals have a hexagonal
shape truncated by a threefold centrosym- I
metric arrangement of {Oll} and {100} {Ol l ]
planes.
322 Chapter 15
Figure 11. Chain of magnetite particles imaged within an intact unstained A. magnetotacticum cell
showing a combination of large and small particles. Particle A has a characteristic morphology seen
in many mature crystals (see Fig. 4 .) Particle B is irregular in form and appears to be at a different
stage in development than A. Lattice images of this particle showed crystalline and noncrystalline
zones (Fig. 12). The cluster of four particles near B were also imaged as single crystals with localized
amorphous regions. They are spatially separate and therefore not a crystallographic multidomain
aggregate. Bar = 100 nm.
4.2.2. Magnetotactic Coccoid Cells

Many crystals isolated from magneto tactic coccoid cells did not show the well-defined
crystallographic morphology shown in Fig. 10. Other crystals, both before and after hy-
pochlorite treatment, showed high-resolution images which had rounded edges and regions
of discontinuity in the lattice fringes traversing the crystal zone. However, these crystals
appeared to be structurally perfect, for through-focus imaging showed that the seemingly
atomic irregularities were in fact due to rapidly changing diffraction conditions due to
variations in crystal thickness and not due to structural disorder in the crystal. Thus, these
highly ordered crystals appear to have extensive surface disorder. Figure 13 shows a mag-
Figure 12. HRTEM image of particle B of Fig. 11 showing the coexistence of crystalline and non-
crystalline phases. The crystalline zone shows well-ordered (222) lattice fringes and is a single domain.
The fringes extend into the amorphous phase in a preferential direction. The superimposed black
dashed line indicates the extent of the low-contrast edge of the particle against the background carbon
noise of the grid. Bar = 5 nm.
netite particle of high crystal perfection as indicated by the continuous (022) lattice planes.
The crystal has well-developed (011) faces but rather poorly developed (100) faces trun-
cating the hexagonal prism. There is a localized "graininess" superimposed on the fringes.
The edges of the crystal are ill-defined but appear in places to have a periodic arrangement
of crystalline outgrowths.
Figure 14 shows a similar crystal imaged in the [011 zone. Careful examination of the
micrograph indicates that the (011) lattice planes are regularly ordered within the crystal
but are very low in contrast and disappear at the edges except in the [011] direction,
suggesting that there is extensive noncrystalline material overlying the crystal surface. It
is unlikely that these images result from electron beam damage to the crystal, as crystal
images and diffraction patterns did not change appreciably with time during the period
of recording the micrographs. It is thus inferred that such images are of crystals in different
stages of development to those which showed highly ordered crystal faces and edges.
5. Discussion: Bioprecipitation of Bacterial Magnetite

Figure 13. Well-ordered single crystal of magnetite imaged in the [OIl) zone. Lattice fringes correspond
to (022) planes. The surface of the crystal is irregular as shown by localized regions of "graininess"
superimposed on the continuous lattice fringes (see regions between white arrows). A periodicity of
crystalline outgrowths can be observed on several of the crystal faces (black arrows). Bar = 20 nm.
Figure 14. Magnetite crystal imaged in the [011J zone showing (011) lattice planes and extensive
noncrystalline material overlying the crystal surface. Bar = 20 nm.
can be proposed. The sequence of events in these organisms can be postulated as
Fe 3 + -chelateoUT hydrous ferric oxide
1
Fe304
magnetosome membrane
cytoplasmic membrane
In this sequence the unique stage appears to be the transformation of the hydrous ferric
oxide to magnetite, as the initial stages of the sequence are likely to involve processes
similar to the formation of ferrihydrite cores in the iron storage protein ferritin.
The critical events in bioprecipitation to be activated and regulated are the nucleation,
growth, and phase transformation processes. Precise replication of bacterial magnetite crys-
tals implies the coordination of control systems for the spatial, chemical, and structural
dimensions of solid formation.
5.1. Nucleation
Nucleation of bacterial magnetite can arise directly from aqueous solution or via an
intermediate (precursor) solid phase which is thermodynamically less stable. Nucleation
processes can be homogeneous or heterogeneous. Homogeneous nucleation occurs due to
the spontaneous formation of nuclei in the bulk of supersaturated solutions whereas het-
erogeneous nucleation involves the formation of nuclei on the surfaces of a substrate pres-
ent within the aqueous medium. Homogeneous nucleation is very unlikely in biological
systems, as there are many organic surfaces present which will energetically favor het-
erogeneous nucleation.
The rate at which nucleation occurs will be dependent on the activation energy re-
quired to create the new interface and thus nucleation will only occur when the energy
released in the formation of bonds in the bulk of the growing solid phase is sufficient to
overcome the energy required to form the new surface. Critical factors which will determine
the nucleation activation energy barrier are the level of supersaturation of ions (the measure
of thermodynamic instability) and the surface energy of the embryonic nucleus which may
be crystalline, pseudocrystalline, or amorphous and which may be highly hydrated. Both
these factors can be biologically influenced through the molecular design of organic sub-
strates and the regulation of ion concentration gradients.
HRTEM of the immature, irregular magnetosomes imaged in A. magnetotacticum cells
suggests that the nucleation of the magnetite phase occurs within or contiguous with an
amorphous ferric oxide precursor. The amorphous precurser will be kinetically (but not
thermodynamically) preferred over the magnetite phase, for the change in entropy of for-
mation will be much lower in the transformation of aqueous ions to sites in a disordered
lattice. As the crystals in both bacterial species are single domain, there must be one
primary site for magnetite nucleation which grows at the expense of other potential sites.
If crystal nuclei do initiate at other sites, then they must rapidly redissolve and reprecipitate
at the primary site. It is therefore tempting to suggest that the surrounding magnetosome
membrane may playa crucial role in directing nucleation. The possible influence exerted
by an organic surface is discussed in Section 5.5.
326 Chapter 15
5.2. Crystal Growth and Morphology
In both A. mognetotocticum and magnetotactic coccoid cells, magnetite crystals have

been observed in close association with amorphous phases. These results suggest that in
these systems the growth of magnetite occurs from structural modification of precursors
rather than from direct precipitation of ions in aqueous solution. The growth of crystals
from phases of different crystallographic structure involves a process of ion translocation
to new lattice coordinates. Phase transformations of this type can occur via surface dis-
solution of the precursor followed by reprecipitation of the second phase upon particles
in the medium. Alternatively, the second phase can be formed via an in situ solid-state
transformation particularly when there is a close structural match (topotaxis) and low
interfacial energies between the two phases. For the transformation of amorphous hydrated
ferric oxide to magnetite, the former mechanism is most probable because there must be
extensive structural rearrangement, dehydration, and partial reduction of the ferric iron.
Thus, in the early stages of magnetite growth in A. mognetotocticum (Fig. 12), the
solid-state rearrangement could occur through a solution front at the interface of the crys-
talline and amorphous phases. At present, there is no information concerning the further
growth of these crystals except that crystal growth is highly ordered. Questions concerning
whether the hydrated ferric oxide continues to be precipitated within the magnetosome
compartment or whether the precursor forms in a single event still remain unanswered.
For the crystals isolated from magnetotactic coccoid cells, it does seem that amorphous
material of as yet unidentified composition is associated with the surface of the growing
magnetite crystals, suggesting that the primary particles of the precursor are continually
transported to or precipitated within the magneto some compartment during crystal growth.
Aggregation of such particles at the magnetite crystal surface is probable, as the isoelectric
point for magnetite is ca. 6.5 at ambient temperature. The periodic edge structure observed
in Fig. 13 may then indicate the active sites of crystal growth at the surface. As the aged
crystals have a high degree of perfection, the surface reactions must be relatively slow.
Figures 4 and 10 show highly specific but different crystal habits for magnetite particles
synthesized in the two bacterial species described in this report. The hexagonal prism
shape is similar to that described for crystals from an unidentified bacterium extracted
from sediment (Matsuda et 01., 1983) but the truncated faces are very different. As the
morphology of a crystal arises from the interaction between its crystal structure and the
environment in which the crystal grows, it is possible that the ultimate forms of these
bacterial magnetite crystals are determined by biological control over the conditions of
crystal growth rather than passive crystallization.
Abiogenic magnetite is often found in octahedral, rhombododecahedral, and cubic
habits which can be rationalized from the spinel (cubic) crystal structure and the relatively
low surface energy of the low-index {lll}, {llO}, and {100} crystal planes. However, crystal
habits are very sensitive to changes in the environmental conditions of growth such that
the level of supersaturation, the direction of supply of ions, the concentration of extraneous
ions and molecules, pH, redox potential, and temperature can all modify the crystal shape.
Thus, the selective control of the chemistry of the precipitation environment could result
in species-specific crystal morphologies.
5.3. Spatial Control of Bacterial Magnetite Formation
Synthesis of magnetite in the bacteria studied above takes place in localized regions
of the cells. Biological compartments can be utilized in the spatial control of biominer-
alization as limiting volumes for mineralization or as surfaces for controlling the crystal
shape and orientation. Thus, the ultimate size of the bacterial magnetite crystals (ca. 50
nm in A. magnetotacticum and ca. 100 nm in coccoid cells) may be a reflection of the
spatial constraints placed on the reaction volume.
In a similar manner, spatial constraints may influence morphology. For example, an
important observation concerning the idealized crystal morphology as shown in Fig. 10
is that faces of the form {011} do not grow at the same rate. One explanation could be that
there is a small degree of anisotropy in the crystal lattice which modifies the crystal space
group from cubic to some other symmetry. However, it is also possible that six of the {On}
faces are limited in growth by the surrounding membrane, permitting anisotropic growth
of the crystal and the development of the unique crystal morphology.
5.4. Chemical Control of Bacterial Magnetite Formation
In principle, biomineral structure and morphology can be determined solely by the

physicochemical properties of the mineralizing environment without any dependence
being set on the stereotactic properties of organic surfaces. A critical factor in the control
of the chemistry of biological precipitation is the requirement of localized biological com-
partments (Mann, 1983). It is therefore possible that the membrane surrounding the mag-
netite crystals plays an important role in controlling the chemistry of the magnetosome
compartment.
The influence of chemical factors on the modulation of nucleation and growth of
bacterial magnetite can arise through the regulation of supersaturation levels within the
magnetosome compartment by active ion-transport processes. Three different processes of
microbial iron transport are known at present (Neilands, 1977; Raymond and Carrano,
1979) and thus different processes could lead to different supersaturation levels within
the magnetos orne envelope resulting in different crystal morphologies and growth rates.
Alternatively, control of the chemical composition of the mineralization zone could result
in precise modifications in the crystal growth processes. For example, by maintaining a
constant environment within the envelope, single crystals of perfect order could be grown
from solutions of low supersaturation. A simple change in anion (CI-, p l ú J I = H2P0 4 )
concentration is known to have a marked influence on magnetite precipitation in vitro
(Sidhu et a1., 1978; Tamaura et a1., 1981) which suggests that there must be biological
control over extraneous ions within the envelope surrounding the magnetite crystals.
The phase transformation of amorphous hydrated ferric oxide to magnetite can occur
at neutral pH provided the redox potential of the reaction environment is established in
the range of -100 mY. It is important to stress that the redox potential will be extremely
sensitive to pH such that small changes in pH could have a marked influence on the phase
transformation processes. For example, a slight lowering of pH will favor more positive
values of the redox potential which ultimately favor transformation to less hydrated, more
crystalline ferric oxides such as geothite (a-FeOOH). This mineral has been observed in
the radular teeth of limpets (Lowenstam, 1962) whereas the radular teeth of chitons are
known to contain magnetite (Towe and Lowenstam, 1967).
The rate of transformation of hydrated ferric oxide to magnetite is likely to be slow
as dehydration, dissolution, reprecipitation, and partial reduction of the ferric ion are
probably involved. Investigations of the transformation of ferric oxides to magnetite under
aqueous conditions in inorganic systems have shown that the critical step is the involve-
ment of aqueous Fe2 + ions at the ferric oxide surface (Tamaura et a1., 1981, 1983). The
adsorption capacity of iron oxides for aqueous metal ions is high (Swallow et a1., 1980),
and the absorption step generally involves the hydrolysis of the cation and releases protons
into solution (Benjamin and Leckie, 1981). Thus, at room temperature above pH 7.3, the
transformation of 'Y-FeOOH to magnetite is triggered by the adsorption of Fe 2 + (aq) ions
328 Chapter 15
[as FeOH+(aq)] onto the ferric oxide surface (Tamaura et a1., 1983) forming soluble ferric-
ferrous hydroxo complexes. Similarly, the transformation of ferrihydrite to magnetite has
been observed in the presence of Fe2 +(aq) ions at pH 7.8 and 25°C under a nitrogen at-
mosphere (Mann, unpublished data).
Tamaura et a1. (1983) have shown that one proton is released when FeOH+(aq) is
adsorbed onto the surface of amorphous ferric oxide:
The kinetics of the adsorption process is first order with respect to the concentration
of Fe2 + ions in solution provided the number of active surface sites remains constant during
the reaction. The next stage is the destabilization of the surface intermediate resulting in
the dissolution and release of mixed-valence hydroxo complexes into solution which sub-
sequently reprecipitate as magnetite:
The rate of magnetite formation appears to be essentially first order with respect to the
concentration of the surface intermediate (Tamaura et a1., 1983). Although the composition
of the intermediate released into solution is unknown, the formation of Fe304 in the second
step involves the release of one further proton.
Thus, it seems probable that the growing magnetite crystal shown in Fig. 12 occurs
through a solution interface between the crystalline and amorphous phases in which the
dissolution of the amorphous precursor takes place via soluble ferric-ferrous hydroxo
complexes due to an increase in Fe 2 +(aq) concentration within the magneto some com-
partment. A similar mechanism can be postulated for the growth of mature crystals isolated
from magnetotactic coccoid cells.
The presence of Fe 2 + ions in A. magnetotacticum has been shown by Mossbauer
spectroscopy (Frankel et a1., 1983). Although the origin of these ions in the crystal growth
process is unclear, the Mossbauer results have shown the Fe2 + ions to be in close asso-
ciation with the hydrated ferric oxide phase. It seems, therefore, that the Fe2 + ions within
the magnetosome compartment must arise from partial reduction of the hydrated precursor
phase due to changes in the local redox potential rather than from direct transport from
the cytoplasm or periplasmic space.
Because the adsorption of Fe2 + ions onto ferric oxide surfaces is accompanied by the
release of protons, there will be a resultant lowering in the reaction pH and an increase
DOD]
Figure 15. Possible growth of the (011) face of magnetite on an

organic substrate with subsequent growth in the [011) and [100)
directions. Square nuclei of this orientation have been observed
in epitaxial studies of inorganic magnetite crystallization (Shi-
gematsu et aI .• 1980).
(more positive) in the redox potential of the reaction system. Thus. in order for the phase
transformation to magnetite to proceed to any significant extent within the bacteria. there
must be precise regulation of the pH and hence redox potential within the localized mi-
neralization zone.
5.5. Structural Control of Bacterial Magnetite Formation
There has been much discussion about the involvement of organic surfaces in bio-
logical mineralization. It has been suggested (Lowenstam. 1981) that organic-matrix-
mediated mechanisms of biomineralization result in highly specific crystallographic types
and orientation. The role of the organic surface as a substrate for epitaxis has been suggested
although several other different functions of the organic surface in biomineralization have
been discussed (Mann. 1983). The crystallographic morphology and alignment of the {111}
faces in the magnetite particles in A. magnetotacticum suggest that there is a stereotactic
relationship at the interface of the crystal and surrounding organic membrane which en-
ergetically stabilizes the [111] vector in crystal growth. In crystals from magnetotactic
coccoid cells. it is the {011} faces which are well developed and stabilized. In these crystals
the final crystal habit could arise by preferential growth on the (011) face with subsequent
growth in the [011] and [100] directions (Fig. 15). Such a mechanism is feasible as it has
been observed that epitaxial growth of the (011) face of inorganic magnetite on the (001)
face of sodium chloride developed from crystalline nuclei ca. 10 Ain thickness and growing
in the [011] and [100] directions (Shigematsu et 01 .• 1980). These nuclei were almost square
in geometry. implying that the rates of growth in the [011] and [100] directions were ap-
proximately the same. The greater length of the biogenic magnetite crystals as compared
with their width or height would then be a consequence of the different rates of growth
of the (111) and (011) faces. respectively. The important point is that because the preferred
orientation of the magnetite nucleus will be dependent on the energetics of interaction
between the crystal faces and the organic substrate. a change in the composition and struc-
ture of the surrounding membrane could in principle result in different initial orientations
and hence different crystal morphologies.
It is clear from the HRTEM results that some mature magnetite crystals in both bacterial
species do not attain the characteristic idealized morphology and crystallographic align-
ment. It may be that nucleation. growth. and organization processes inherent in these
biological systems are very susceptible to physical and chemical fluctuations in their intra-
and extracellular environments. Crystallographic alignment may be significantly distorted
by drying processes on the electron microscope grid and by entry into the high-vacuum
conditions of the sample chamber.
330 Chapter 15
6. Conclusions
The direct lattice imaging of bacterial magnetite particles by HRTEM has led to some
important developments in the elucidation of the structure, morphology, and crystal
growth of this biogenic material. The particles are highly ordered single-domain crystals,
have characteristic morphologies, and can be crystallographically ordered into chains. The
mechanisms of their synthesis are intriguing and it has been shown that there is a close
Biological
organization
I Control
I
-l. 'lit I
Functional Process
design
,... design
Product
design
I ,"
Space
I Cellular " Crystal size
compartmentalization and shape
Chemistry ,'- I Ion ·fl ux/gradients I ,"-

I I I Crystal nucleation
growth and
composition
Structure I ," I Stereotaxis l ,,

I I I Crystal structure
and orientation
Figure 16. Flow diagram showing the control principles inherent in biomineralization.
association at the nanometer level between the growing crystal and an amorphous pre-
cursor.
The study of biomineralization products such as bacterial magnetite involves the in-
corporation of solid-state chemical data in the wider context of biological organization.
Thus, the subject is challenging for the physical and biological scientist. Figure 16 shows
in outline a possible rationalization of biomineralization as an organizational and control
system. There is a hierarchy of control levels leading from functional to process to product
design. Control over functional properties is directed at the spatial, chemical, and structural
dimensions of biological processes which involve the activation and regulation of localized
cellular compartments, ion fluxes across these compartment boundaries, and molecular
design of organic surfaces which selectively influence crystal formation (stereotaxis). The
resulting product (the biomineral) may then have characteristic size, morphology, com-
position, crystallographic structure and orientation through the dynamic interplay of these
control processes.
Many questions remain to be answered. In particular, the structure and composition
of the magnetosome membrane remains a complete mystery. The study of biogenic mag-
netites of such remarkable crystallographic specificity has far-reaching implications not
only in the heuristic value of studying biological organization and control in biosolid-state
reactions but also in the applications of such knowledge in the practical fields of catalysis
and materials science.
ACKNOWLEDGMENTS. The HRTEM work reported in this article could not have been under-
taken without the assistance of Drs. R. B. Frankel, R. P. Blakemore, and T. T. Moench who
provided the bacterial samples. Many thanks to Professor R. J. P. Williams for his en-
couragement and help throughout the course of this work.
References
Balkwill, D. L., Maratea, D., and Blakemore, R. P., 1980, Ultrastructure of a magnetotactic spirillum,
J. Bacteriol. 141:1399-1408.
Benjamin, M. M., and Leckie, J. 0., 1981, Multiple-site adsorption of Cd, Cu, Zn and Pb on amorphous
iron oxyhydroxide, J. Colloid Interface Sci. 79:209-221.
Blakemore, R. P., Maratea, D., and Wolfe, R. S., 1979, Isolation and pure culture of a freshwater magnetic
spirillum in chemically defined medium, J. Bacterial. 140:720-729.
Frankel, R. B., and Blakemore, R. P., 1980, Navigational compass in magnetic bacteria, J. Magn. Magn.
Mater. 15-18:1562-1564.
Frankel, R. B., Papaefthymiou, G. C., Blakemore, R. P., and O'Brien, W., 1983, Fe a04 precipitation in
magnetotactic bacteria, Biochim. Biophys. Acta 763:147-159.
Hutchison, J. L., Jefferson, D. A., and Thomas, J. M., 1977, The ultrastructure of minerals as revealed
by high resolution electron microscopy, in: Chemical Society Specialist Reports in Surface and
Defect Properties of Solids Volume 6, pp. 320-358.
Lowenstam, H. A., 1962, Goethite in radular teeth of recent marine gastropods, Science 137:279-280.
Mann, S., 1983, Mineralization in biological systems, Struct. Bonding (Berlin) 54:125-174.
Mann, S., and Williams, R. J. P., 1982, High resolution electron microscopy studies of the silica lorica
in the choanoflagellate Stephanoeca diplocostata Ellis, Proc. R. Soc. London Ser. B 216:137-146.
Mann, S., Parker, S. B., Ross, M. D., Skarnulis, A. J., and Williams, R. J. P., 1983, The ultrastructure
of the calcium carbonate balance organs of the inner ear: An ultra-high resolution electron
microscopy study, Proc. R. Soc. London Ser. B 218:415-424.
Mann, S., Moench, T. T., and Williams, R. J. P., 1984a, A high resolution electron microscopic in-
vestigation of bacterial magnetite: Implications for crystal growth, Proc. R. Soc. London Ser. B
221:385-393.
332 Chapter 15
Mann, S., Frankel, R. B., and Blakemore, R. P., 1984b, Structure, morphology, and crystal growth of
bacterial magnetite, Nature 310:405-407.
Maratea, D., and Blakemore, R. P., 1981, A. magnetotacticum sp. nov., a magnetic spirillum, Int. J.
Syst. Bacteriol. 31:452-455.
Matsuda, T., Endo, J., Osakabe, N., and Tonomura, A., 1983, Morphology and structure of biogenic
magnetite particles, Nature 302:411-412.
\1oench, T. T., and Konetzka, W. A., 1978, A novel method for the isolation and study of a magnetotactic
Neilands, J. B., 1977, Siderophores; Biochemical ecology and mechanism of iron transport in enter-
obacteria, Adv. Chern. Ser. 162:3-32.
Raymond, K. N., and Carrano, C. J., 1979, Coordination chemistry and microbial iron transport, Acc.
Chern. Res. 12:183-190.
Shigematsu, T., Ushigome, H., Bando, Y., and Takada, T., 1980, Epitaxial growth of magnetite on rock
salt by reactive evaporation, J. Cryst. Growth 50:801-806.
Sidhu, P. S., Gilkes, R. J., and Posner, A. M., 1978, The synthesis and some properties of Co, Ni, Zn,
Cu, Mn, and Cd substituted magnetites, J. Inorg. Nucl. Chern. 40:429-435.
Skarnulis, A.-J" 1979, A system for interactive electron image calculations, J. Appl. Crystallogr. 12:636-
635.
Swallow, K. C., Hume, D. N., and Morel, F. M., 1980, Sorption of copper and lead by hydrous ferric
oxide, Environ. Sci. Technol. 14:1326-1331.
Tamaura, Y., Buduan, P. V., and Katsura, T., 1981, Studies on the oxidation of iron(II) ion during the
formation of Fe a04 and !X-FeOOH by air oxidation of Fe(OH)z suspensions, J. Chern. Soc. Dalton
Trans. 1981:1807-1811.
Tamaura, Y., Ito, K., and Katsura, T., 1983, Transformation of 'Y-FeOOH to Fe a04 by adsorption of
iron(II) ion on 'Y-FeOOH, J. Chern. Soc. Dalton Trans. 1983:189-194.
Towe, K. M., and Lowenstam, H. A., 1967, Ultrastructure and development of iron mineralization in
Chapter 16
Biomineralization Processes of the
Radula Teeth of Chitons
MICHAEL H. NESSON and HEINZ A. LOWENST AM
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 333
2. Materials and Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
2.1. Animals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
2.2. Tissue Preparation for Electron Microscopy. . . . . . . . . . . . . . . . . . . . . . . . .. 334
2.3. Staining for Electron Microscopy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
2.4. Light Microscope Staining. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 335
3. Anatomy and Operation of the Radula Apparatus. . . . . . . . . . . . . . . . . . . . . . . .. 335
3.1. The Radula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
3.2. The Radula Apparatus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 335
3.3. The Feeding Process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
3.4. The Radula Replacement Process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 337
4. Anatomy of the Radula Sac . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
4.1. Radula Formation and Maturation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 337
4.2. The Mineralization Zone. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
4.3. Light Microscopy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 338
4.4. The Dorsal Sinus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 341
5. Blood Chemistry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 341
6. The Ultrastructure of the Mineralization Zone. . . . . . . . . . . . . . . . . . . . . . . . . . . 342
6.1. The Dorsal Sinus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 342
6.2. The Basement Material. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
6.3. The SE Cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
6.4. Ultrastructure of the Tooth Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
7. Concluding Remarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 361
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 361
1. Introduction
Magnetite was first identified as a biogenic mineral in the radula teeth of chitons (Mollusca,
Polyplacophora) (Lowenstam, 1962). The caps of mature major lateral teeth are composed
in part of crystalline magnetite in an ordered matrix of organic fibrils (Towe and Low-
enstam, 1967; Kirschvink and Lowenstam, 1979). The biomineralization proceeds in two
steps: (1) an initial deposition of ferrihydrite (5Fe z0 3'9H zO) (Towe and Bradley, 1967)
within the organic matrix of the tooth cap (Towe and Lowenstam, 1967), and (2) the trans-
formation of the ferrihydrite into magnetite (FeO'FeZ03) (Towe and Lowenstam, 1967).
MICHAEL H. NESSON • Department of Agricultural Chemistry, Oregon State University, Corvallis,

Oregon 97331. HEINZ A. LOWENSTAM • Division of Geological and Planetary Sciences, Cal-
ifornia Institute of Technology, Pasadena, California 91125.
333
334 Chapter 16
The mineralization process occurs within the radula sac; the teeth are associated with
the superior epithelial tissue of the sac. Within the zone of mineralization, cells of the
superior epithelium which contact the major lateral tooth caps contain a large accumulation
of ferric iron-containing granules (Prenant, 1928; Gabe and Prenant, 1948; Carefoot, 1965).
We have isolated and identified ferritin from these cells in Cryptochiton stelleri (Towe et
al., 1963) and have, with the electron microscope, observed in this tissue membrane-bound
vesicles, some filled with ferritin, and others with a less-ordered electron-dense material,
resembling vertebrate hemosiderin (Towe and Lowenstam, 1967).
In this report, we describe the results of our further investigation of some biological
aspects of the process of tooth mineralization. Our findings on the anatomical relationships
and detailed ultrastructure of the iron-depositing cells may be of value to understanding
the process of magnetite biomineralization in other biological organisms.
2.1. Animals
Specimens of Lepidochitona (Cyanoplax) hartwegii and Mopalia muscosa were col-

lected from the rocky intertidal zone at Corona del Mar, California, and were maintained
in the laboratory in recirculating seawater aquaria at 15°C. Alga-encrusted rocks from the
collection area served as substrate and food source.
2.2. Tissue Preparation for Electron Microscopy
The three anterior valves of an animal were removed to expose the buccal region. Ice-
cold 2.5% glutaraldehyde (Aldrich Chemical Co.) in filtered, unbuffered seawater was
poured into the body cavity. The posterior portion of the radula sac was dissected free,
removed from the animal, and placed in two changes of fresh cold fixative for 30 min and
12 hr. After washing (2 X 15 min) in cold, filtered seawater, most material was postfixed
for 2 hr in 2% OS04 in filtered seawater and washed in cold seawater (3 X 5 min). Specimens
were then dehydrated through a concentration series of ethyl alcohol, infiltrated with
propylene oxide, and embedded in Epon 812 (Shell Chemical Co.), following the procedure
of Luft (1961). The epoxy was polymerized and cured at 60°C. Both I-fLm sections for light
microscopy and thin sections for electron microscopy were cut with glass knives on an
LKB Ultratome.
2.3. Staining for Electron Microscopy
Electron microscope sections were stained with either uranyl acetate (Brody, 1959) or
lead citrate (Venable and Coggeshall, 1965), or most frequently both.
Fixation quality of the relatively large M. muscosa radula sac was quite poor. Hence,
all of the accompanying electron micrographs are of 1. hartwegii tissue. However, our
observations indicated no differences between the two species.
Thin sections were examined with a Phillips EM-200 electron microscope. Selected-
area electron diffraction studies of unstained thin sections were also performed following
the methods described in the microscope operating manual.
Radula Teeth of Chitons 335
2.4. Light Microscope Staining
One-micrometer Epon sections were stained for light microscope observation by the
procedure of Richardson et a1. (1960), except for the replacement of methylene blue with
1% toluidine blue. The Prussian blue method was modified to demonstrate ferric iron in
Epon-embedded material. Sections were placed in 1% potassium ferrocyanide in 1% HCI
for 1-2 hr at 60°C. The reagent was replaced every 15 min to prevent nonspecific deposition
of the Prussian blue. The sections were counterstained with 5% aqueous neutral red at
60°C for 15 min.
3. Anatomy and Operation of the Radula Apparatus

3.1. The Radula
Throughout the Polyplacophora, the radula is of quite uniform construction. The long,
ribbonlike radula membrane bears many transverse rows of teeth; each row contains a
central tooth and eight flanking pairs of teeth. The central tooth and the first laterals are
small and weakly scoop-shaped in profile. The third laterals have a narrow elongate shape;
the fourth and fifth laterals and the three marginal pairs are all small, polygonal, and of
low profile. The second laterals, referred to as the major lateral teeth, are the largest and
most distinctive teeth (Fig. 1B). They are composed of two parts (Runham, 1963a): a long
tooth base, attached at one end to the radula membrane, and a clearly demarcated, pos-
teriorly recurved tooth cap. The opaque black tooth caps, formed in part of magnetite
(Lowenstam, 1962), present a sharp contrast to the tooth bases and to the other teeth which
are all nearly transparent and only lightly pigmented. The shape of the tooth caps ranges
from a broad chisellike form (in the Chitonidae) to a one- to three-pronged cusp (in most
other families). The cap shape is useful as a diagnostic taxonomic characteristic (Thiele,
1909). The occurrence of additional mineral species in the tooth caps also appears to be
taxonomically correlated: the Chitonidae deposit lepidocrocite (a-Fe DOH) and francolite
(a carbonate fluorapatite) into distinct micro architectural units (Lowenstam, 1967); Cryp-
tochiton (Lowenstam, 1972) and Mopalia (unpublished observations) deposit an amor-
phous hydrous ferric phosphate.
The major lateral tooth caps in an individual radula display marked changes in ap-
pearance from one end to the other. Posteriorly, the first 8 to 12 rows of caps are colorless,
transparent, and soft. Runham's (1963a) histochemical studies on Acanthochitona com-
munis suggest that the organic framework of the tooth caps is composed of chitin and
protein which become cross-linked prior to mineralization. Over the next 2 to 4 rows, the
caps are impregnated with increasing amounts of opaque, golden- to reddish-brown fer-
rihydrite. Next come several rows with increasingly dense magnetite deposits, and finally
20 to 70 rows of shiny black, magnetite-bearing, tooth caps. The last few anterior rows of
caps usually show signs of mechanical wear.
3.2. The Radula Apparatus
The superb anatomical studies of Plate (1897, 1899, 1902) show that the feeding ap-
paratus of chitons (Fig. 1A) consists of (1) the radula itself, which extends posteriorly
within the radula sac, a long outpocketing of the wall of the buccal cavity, (2) a lateral
pair of elongate-ovoid cartilaginous supports, and (3) a complex array of over 40 myo-
globin-containing buccal muscles (Giese, 1952; Manwell, 1958). The supports, together with
336 Chapter 16
RAOULA RETRACTOR MUSCLES
BUCCAL CAVITY
RAOUL A
FOOT
Figure 1. Diagrammatic transverse section through the radula apparatus of a chiton (after Plate). (8)
A part of the radula of C. stelleri: eight rows of paired mature major lateral tooth caps. At the bottom,
the marginal teeth show increasing pigmentation toward the anterior of the radula (right).
several associated muscles, form a discrete tonguelike mass, the odontophore, which com-
prises much of the posterior wall of the buccal cavity. The anteriormost 10-15 tooth rows
of the radula extend over the dorsal and anterior surfaces of the odontophore.
3.3. The Feeding Process
We have studied the feeding process from 16-mm motion pictures of Cryptochiton
stelleri grazing on the algae growing on transparent aquarium walls. The first step of feeding
is the appearance, in the dilated oral opening, of the subradula organ which is presumed
to perform chemosensory testing of the substrate (Heath, 1903). If suitable food is present,
the subradula organ is withdrawn and the odontophore is rotated downward and forward
to the mouth. This motion places the anterior 8 to 10 rows of radula teeth in close proximity
to the substrate through the oral opening. Next, the entire radula is retracted over the
immobile odontophore. As the radula membrane is drawn over the bending-place formed
at the anterodorsal edge of the odontophore, each pair of major lateral teeth is sequentially
rotated toward the midline by the curling of the radula membrane. As they move toward
the midline, the major lateral tooth caps firmly scrape the substrate in the effective rasping
stroke, so as to produce the characteristic chevron-shaped feeding trail. The tooth caps,
the first lateral and the central teeth all rake the abraded material into the mouth. The
odontophore is then retracted and the chiton is ready to initiate another feeding cycle.
3.4. The Radula Replacement Process
The radula in all radula-bearing molluscs is continually replaced throughout the life
of the organism (reviewed by Raven, 1966). The replacement mechanism is comprised of
several distinct but closely integrated processes: (1) new radula material is secreted at the
posterior end of the radula sac; (2) the entire radula gradually moves anteriorly within the
sac; (3) the newly formed radula material undergoes various maturational changes as it
moves forward; and (4) the anteriormost, usually use-worn, radula is broken down and
eliminated. Thus, by the nature of the replacement mechanism, essentially all stages of
tooth development and maturation are present in linear array in an individual radula.
4. Anatomy of the Radula Sac
4.1. Radula Formation and Maturation
New radula material is produced by a hemispherical cushion of narrow epithelial cells,

the odontoblasts, which form the posterior end of the radula sac. The first appearance of
a tooth is a thin membrane secreted onto the surface of the odontoblasts. As additional
material is secreted, the membrane thickens and assumes the shape of the mature tooth
(Rottmann, 1901; Schnabel, 1903). In pulmonate gastropods, autoradiographic studies pro-
vide evidence that the odontoblasts are permanent (Runham, 1963b).
The cells at the lateral and ventral edges of the odontoblast cushion secrete the radula
membrane; the cells bordering these edges divide to form the inferior epithelium (Fig. 2)
which is responsible for maturational changes of the radula membrane (Runham, 1963b).
Cells adjacent to the dorsal edge of the odontoblast cushion continually divide to form
the superior epithelium (SE) (Figs. 2-4). These columnar cells extend to the surfaces of
all teeth and almost entirely fill the intervening spaces between the teeth (Gabe and Prenant,
1959). Autoradiographic studies on the pulmonate gastropod Lymnaea stagnalis (Runham,
1963b) have established that the cells of the SE move forward at exactly the same rate as
the radula itself. Thus, the complex interdigitation of cells and teeth is not disturbed as
the radula grows anteriorward. The same relationship between SE cells and radula teeth
is likely to hold true for all radula-bearing molluscs.
The cells of the SE secrete the substances necessary for maturation of the teeth. In the
gastropods Bulla and Haminea, the SE cells are rich in polysaccharides in the same region
of the radula sac where these substances appear in the teeth (Gabe and Prenant, 1952).
Similarly, in molluscs whose mature teeth are impregnated with calcium- or iron-con-
taining minerals, the SE cells in the mineralization zone are rich in calcium or iron, re-
338 Chapter 16
100Hm
Figure 2. Cross section of the radula sac of 1. hartwegii. The pair of major lateral tooth caps (tc) are
impregnated with brown ferrihydrite mineral. They are surrounded by the superior epithelial cap cells
(cc) which can be distinguished from the minor cells (mc) by their darker staining cytoplasm. The
prismatic inferior epithelial cell layer (ie) covers the curled-up radula membrane (rm). The cartila-
ginous plate (cp) forms the top of a dorsal sinus (ds). Except for the tooth caps , all of the teeth exhibit
parallel horizontal fractures. A small portion of the major lateral tooth base is attached to each tooth
cap and overlies the tooth base (tb) of the next anterior row. Note the cell-filled pores within the tooth
bases and the filling by minor cells of the spaces between all of the minor teeth. It, lateral tooth; mt,
marginal teeth. One micrometer Epon section; toluidine blue, azure II.
4.2. The Mineralization Zone

The iron-containing (Fe 3 +) granules first appear in the cytoplasm of the SE cells one
to three rows posterior to the first row in which ferrihydrite can be detected in the tooth
caps, in Acanthochites fascicularis (Gabe and Prenant, 1948) and in L. hartwegii and M.
muscosa. The concentration of granules increases greatly over a few rows, and then grad-
ually decreases. We have observed iron-containing granules over a segment of the SE at
least 10 rows in length. We are unsure of the anterior extent of the granule-containing
zone, because we have been unsuccessful in preparing tissue sections in the mature tooth
cap region. However, direct measurements of total iron content of individual tooth caps
in several individuals of M. muscosa indicate that iron deposition is complete 10 to 12
rows anterior to the first iron-containing cap (unpublished observation).

Cross sections (Fig. 2) reveal that the radula membrane is curled into aU-shaped
channel within the radula sac. The inferior epithelium consists of a thin, single layer of
100 jJm
Figure 3. Longitudinal section through the early mineralization zone of the radula sac of L. hartwegii.
From posterior to anterior (left to right), there are two colorless tooth caps, two brown ferrihydrite-
impregnated ones, and the first two magnetite-containing caps (tc). Magnetite is visible as the dark
material on the posterior surface of the anteriormost tooth cap. The plane of sectioning passes through
two prongs of the tricuspid tooth caps. The basement material (bm) is discernible mainly where it
has lifted off the superior epithelial cells during fixation. The cap cells (cc) form discrete tissue masses
surrounding each cap, separated from each other by strands of minor cells (mc). Within the tissue
mass, cap cells extend from the dorsal sinus (ds) to the tooth cap surface. tb, tooth bases; rm, radula
membrane; cp, cartilaginous plate. One-micrometer Epon section; toluidine blue, azure II.
prismatic cells which directly contacts the outer surface of the radula membrane. The
inferior epithelium is, in turn, covered by a layer of connective tissue which extends be-
yond the edges of the radula membrane over the dorsal quadrant of the sac to form a
continuous sheath around the sac. This outer sheath is the insertion site for a number of
muscles including the major radula retractors.
All of the radula teeth are embedded in the radula membrane (Figs. 2 and 3). The
tooth bases of the major lateral teeth project toward the center of the sac (Fig. 2) and extend
anteriorly (Fig. 3), so that the major lateral caps of one row lie over the bases of the next
anterior row. Serial sections show that a long cell-filled canal extends through each major
lateral tooth base and terminates close to the base-cap junction. The canal opens to the
exterior through a pore that is located on the anterior face of the stylus near its basal end.
The function of this canal is unknown.
In the mineralization zone, both the tooth bases and the minor teeth generally crack
or shatter during sectioning, presum. .bly because they are poorly infiltrated by the plastic
embedment medium. The major lateral tooth caps are of markedly different consistency
from the other teeth, as evidenced by the ease with which they can be cleanly sectioned.
340 Chapter 16
Figure 4. High-magnification light microscope detail of a cross section similar to Fig. 2. The basement
material, the endothelium (e), and portions of the basal part of some cap cells (cc) have pulled away
during specimen preparation. Each cap cell extends from the dorsal sinus (ds) to the surface of the
tooth cap. Scattered dark-staining Fe 3 + -containing granules are visible at the basal pole and central
region of the cap cells. A dense accumulation of granules defines the granule zone (gz) which is
separated from the tooth cap surface by the cell apical region which exhibits an alternating arrangement
of dark- and light-stained areas. n, nuclei of the cap cells; tb, tooth base; It, lateral teeth; mc, minor
cells. One-micrometer Epon section; toluidine blue, azure II.
The cells of the SE (Figs. 2 and 3) form a solid mass of tissue completely enveloping
the radula teeth throughout the zone of tooth maturation. Within the mineralization zone,
we see two distinct classes of SE cells. The first type, referred to here as the cap cells (cc,
Figs. 2-4) , are those which terminate on the major lateral tooth caps. The second type,
which we call minor cells (mc), terminate either on the minor teeth or on the surfaces of
the major lateral tooth bases. The cap cells of each transverse tooth row form a single
compact mass that envelops both tooth caps (Fig. 2). Each tissue mass is clearly separated
from the adjacent ones by strands of the minor cells (Fig. 3) . Every cap cell arises at the
dorsal sinus (vide infra), extends toward the tooth cap by a path that may include a com-
plete 180 0 turn, and terminates directly on, and usually nearly perpendicular to , the cap
surface (Figs. 3 and 4).
Most of the minor cells appear to extend from the lateral margins of the dorsal sinus
floor to the surfaces of the major lateral tooth bases and the minor teeth. Strands of minor
cells also enter the pore of each tooth base and fill the canal (Fig. 2).
The cap cells of L. hartwegii and M. muscosa are cytologically identical to the cells
described by Prenant (1928) and Gabe and Prenant (1948) for several European chiton
species. Fe3 + -containing granules occur at the basal pole, scattered through the central
Radula Teeth of Chi tons 341
cytoplasm, and in a dense accumulation 15 jJ.m wide near the apical pole (Figs. 2-4). The
granule zone is separated from the tooth surface by a 5- to 10-jJ.m-wide zone which contains
a complex pattern of light- and dark-staining material (with toluidine blue). This granule-
free apical zone contains no Fe3 +, detectable by the Prussian blue method. Attempts to
detect Fe2+ iron in l-jJ.m sections and in paraffin-embedded thick sections were negative
both for the granules and for the apical zone.
4.4. The Dorsal Sinus
The dorsal portion of the radula sac is occupied by a hitherto undescribed cavity that
we call the dorsal sinus (ds, Figs. 2-4). The sinus is not readily discernible in material
prepared for light microscopy by paraffin embedment methods. However, in l-jJ.m Epon
sections of the radula sac, it can be seen to occupy the entire volume between the outer
connective tissue sheath and the SE tissue mass. The lumen of the sinus is limited by a
dark-staining, very thin layer which is directly apposed to the dorsal connective tissue
sheath, but separated from the SE cells by a layer of light-staining extracellular matter, the
basement material (Fig. 3). The sinus extends nearly to the posterior end of the radula sac.
We have confirmed the existence of a thick cartilaginous plate (cp, Figs. 2 and 3) formed
as a thickening of the dorsal portion of the sheath lying over the posterior region of the
sac, first described by Carefoot (1965). While he suggests that this plate might function to
support the soft, newly formed radula until tanning is completed, we consider it more
likely that it serves to prevent the collapse of the dorsal sinus where the radula membrane
itself is not firm enough to do so.
The anterior extent and site of origin of the dorsal sinus have not been determined;
the highly mineralized mature tooth caps prevent the cutting of useful sections in this
region. However, certain anatomical findings of Plate (1897, 1899, 1902) are of likely rel-
evance with regard to the anterior origin of the dorsal sinus. In most chiton families, the
posterior portion of the radula sac lies entirely within a large blood vessel, the visceral
artery, which arises at the site where the radula sac penetrates the diaphragm which sep-
arates the cephalic blood sinus from the abdomen. The visceral artery transports a portion
of the blood from the head region to branches which ramify in several organs of the digestive
tract. A dorsal aorta supplies blood directly from the heart to the cephalic blood sinus,
where it bathes the complex musculature of the radula apparatus. Thus, the radula sac is
provided with an abundant supply of blood and with a potential source of blood to fill
the dorsal sinus.
5. Blood Chemistry
The blood (or hemolymph) of the Polyplacophora contains the copper protein, hem-
ocyanin, as its principal respiratory pigment (Manwell, 1958). A small number of nu-
cleated, basophilic cells (103 _104 cells/mm3 ) are also present in the hemal fluid (Arvy and
Gabe, 1949) (Fig. 9). The presence of iron in the blood was first reported by Carefoot (1965)
who found 85 jJ.g Fe/ml whole blood in a specimen of C. stelleri.
We h;lVe determined the total iron content of the blood of 10 individuals of M. muscosa,
each analyzed in duplicate by the colorimetric method of Fischer and Price (1964). The
measured iron concentrations range from 42 to 113 jJ.g/ml. In our initial attempts to char-
acterize the blood iron, we observed a single, homogeneous-sized iron-containing peak by
agarose-gel chromatographic fractionation; we found this material to possess a broad range
of sedimentation coefficients (from 66 S to 20 S) upon sucrose gradient centrifugation.
342 Chapter 16
This combination of properties is characteristic of ferritin, which has a unique size, but a
range of molecular weights dependent upon the iron content of individual molecules.
Electron microscope examination of samples of Mopalia blood, both unstained and neg-
ative-stained with phosphotungstic acid, conclusively demonstrated the presence of fer-
ritin in the blood. Recently, Webb and Macey (1983) have confirmed these findings for
several Australian chiton species.
Ferric iron-containing granules have been observed in the digestive gland and the
intestinal epithelium of M. muscosa (unpublished observation) and of several other chiton
species (Fretter, 1937; Gabe and Prenant, 1948). As both of these organs are involved in
the absorption and mobilization of ingested food constituents, both are potential loci of
the uptake of dietary iron and its incorporation into ferritin as well as the release of ferritin-
sequestered iron into the circulatory system.
6. The Ultrastructure of the Mineralization Zone
6.1. The Dorsal Sinus
The lumen of the dorsal sinus is lined with a single layer of flattened endothelial cells
(Figs. 5-7). Dorsally, the endothelium is directly apposed to the connective tissue sheath
(data not shown); ventrally, the endothelial cells are separated from the basal ends of the
SE cells by a layer of extracellular substances, the basement material (Figs. 5-8).
We have not been able unequivocally to demonstrate the continuity of the endothelial
lining. Often, gaps occur between the endothelial cells, the cells are frequently pulled
away from the basement material, and in many areas the cells cannot be detected at all
(Fig. 5). However, the endothelial layer is probably quite prone to artifactual distortion
and disruption in the course of specimen preparation.
The cells of the endothelium are nucleated (Fig. 5) and their cytoplasm contains scat-
tered cisternae of endoplasmic reticulum and a few mitochondria (Figs. 6 and 7). Mem-
brane-bound vesicles are also present; these vesicles contain amorphous-looking material
of low electron density in which 6-nm electron-dense micelles, presumably ferritin, are
commonly present (Fig. 6).
Similar ferritin-containing clumps of amorphous material are present within the lumen
of the dorsal sinus. They are especially abundant in the posterior portion of the sinus,
close to its blind end where they occur in association with cells (Fig. 9), which resemble
the blood cells described by Arvy and Gabe (1949). These cells contain a large nucleus
and a small amount of basophilic (ribosome-rich) cytoplasm. Studies on capillaries of
mammalian tissues indicate that ferritin transport across the endothelial cells occurs in
membrane-bound vesicles which pick up ferritin on the lumenal side and dump it on the
basement layer side (Bruns and Palade, 1968). We suggest that a similar process occurs in
the dorsal sinus endothelial cells. These lines of circumstantial evidence lead us to believe
that the dorsal sinus is a specialized branch of the circulatory system which transports
iron, in the form of plasma ferritin, to the cap cells in the zone of tooth mineralization.
Conclusive proof of this hypothesis will require a determination of the anatomy of the
anterior portion of the sinus.
6.2. The Basement Material
The basement material forms a continuous layer between the sinus endothelium and
the SE cells (Fig. 5). Its thickness varies from less than 100 nm, where it is contiguous to
Figure 5. The basal pole of the cap cells (CC). The thick basement material layer is closely apposed
to the cap cells. A nucleated endothelial cell (Endo) lines the dorsal sinus (DS). The cap cells are
penetrated by many light-colored extracellular spaces, and the cytoplasm contains electron-dense
vesicles of many sizes. Note, especially at the top of the field, how the basement material exhibits a
heterogeneous appearance of alternating light and dark areas .
344 Chapter 16
0,5 11m
Figure 6. The endothelial cell (Endo) contains a large vesicle (V) with ferritinlike material in its lumen
(small arrow) . The large arrows indicate dark regions of the basement material which are composed
of concentrations of 6-nm micelles in a matrix of moderate electron density. DS, dorsal sinus; MT,
mitochondria; ER, endoplasmic reticulum; CC, cap cell.
the basal ends of the minor cells (data not shown), to 1-1.5 11m, where it contacts the cap
cells (Figs. 5-7). The basement material appears to be attached to the SE cells, for it remains
intact even in regions where the endothelial cells are not present (Fig. 5).
Near the cap cells, the basement material is of rather uniform thickness (- 111m), but
is inhomogeneous in appearance (Fig. 5). In transverse sections, electron-dense bands, 40-
50 nm thick, alternate with wider, less-dense zones, with a repeat period of about 200 nm
(Fig. 6). In tangential section, the electron-dense regions form a regular array of polygonal
units which are about 200 nm on a side. From its appearance in several orientations, we
deduce that the dense material is arranged in a honeycomblike pattern with its partitions
Figure 7. Basal pole of superior epithelial cells. The cap cell (CC) displays many stages of endocytosis
of ferritin molecules. Ferritin-coated membranes and invaginations abound. At the lower right, two
minor cells (MC) are distinguished by the presence of mitochondria (Mt) in their cytoplasm and the
absence of ferritin vesicles. The arrows denote an intercellular space containing many ferritin micelles.
Note that the cap cell membrane is actively endocytotic. while the minor cell membrane is free of any
bound ferritin. Endo. endothelial cell; ER. endoplasmic reticulum; OS , dorsal sinus; BM, basement
material.
346 Chapter 16
Figure 8. Detailed views of ferritin endocytosis by the basal end of cap cells (CC). Note the ramifying
extracellular spaces. The heterogeneity of the basement material (BM) is especially clear in A.
perpendicular to the basal face of the cap cells. When viewed at higher magnification (Figs.
6-8), the electron-dense regions are seen to consist of concentrations of ferritin micelles
embedded in material of intermediate electron density (presumably the apoferritin pro-
tein). The least dense areas of the basement material are composed of very fine, possibly
fibrillar, material less than 3 nm in diameter. There is no apparent association of the dense
regions with any visible structures of either the endothelial cells or the cap cells. Our
electron micrographs do not allow us to distinguish between the presence of preformed
channels or the occurrence of preferential ferritin-binding sites within the basement ma-
terial. We have not seen any published reports describing similar ordered substructure
within a basement layer in other organisms. We can only conjecture whether this pattern
is indicative of a specialized mechanism for the concentration of ferritin or for the transport
of ferritin across the basement material.
Figure 9. Within the dorsal sinus, a small basophilic cell is surrounded with clumps of plasma proteins
in which 6-nm ferritin micelles are embedded . N, nucleus ; R, ribosomes; Fn, ferritin micelles.
348 Chapter 16
6.3. The SE Cells
6.3.1. The Basal Pole

Our findings indicate that ferritin enters the cap cells by the process of endocytosis.
Ferritin molecules are first bound to the plasma membrane, presumably at regions rich in
ferritin-specific binding sites. The ferritin-coated membrane then invaginates and pinches
off to form a small intracellular membrane-bound vesicle containing ferritin (Figs. 7 and
8). The basal region of the cap cells has a "spongy" appearance (Fig. 5) due to the prolif-
eration of ramifying extracellular spaces which provide a tremendous increase in the area
of cell membrane available for endocytosis (Figs. 7 and 8). The ferritin endocytosis zone
extends deep into the cap cell tissue mass. We have observed ferritin molecules within
the intercellular spaces even at the level of the cap cell nuclei and have occasionally seen
ferritin-filled endocytotic invaginations in this region (data not shown).
The larger, membrane-bound ferritin-filled granules (Figs. 5, 7, and 8) are presumably
formed by fusion of the endocytotic vesicles. Similar fusion has been reported to occur
during ferritin uptake by mammalian renal epithelial cells (Farquhar and Palade, 1960) and
by Ehrlich ascites tumor cells (Ryser et aI., 1962). In the latter case, the uptake and ag-
gregation of ferritin was found to occur even under anaerobic conditions. The absence of
any mitochondria in the basal portion of the cap cells suggests that ferritin endocytosis in
chitons may also occur without any need for the energy provided by mitochondrial oxi-
dative phosphorylation.
The basal pole of the minor cells can be readily distinguished from the cap cells by
several criteria. They are generally of smaller diameter, and always contain a large number
of densely staining mitochondria, which are uniformly about 0.2 f.Lm in diameter and 1.0
f.Lm long (Fig. 7). Although the well-developed intercellular spaces generally contain fer-
ritin molecules, we have not observed any ferritin endocytosis into the minor cells. In Fig.
7, the arrows denote an intercellular space between a mitochondrion-containing minor
cell and a cap cell. Although the cap cell plasma membrane is almost completely coated
with ferritin, no ferritin is visible on the cell membrane of the adjacent minor cell. We see
no ferritin-containing granules within the basal region of the minor cells.
6.3.2. The Central Region of the Cap Cells

From the basal pole to the level of the nuclei, the cap cells possess few distinctive
features, except for the ferritin-containing granules. With increasing distance away from
the basal pole, the v·olume of free space within a ferritin granule is reduced. The granules
in the nuclear region are round or oval membrane-bound vesicles completely filled with
ferritin molecules.
The large ovoid nucleus is situated in the apical half of the cap cell (Fig. 4). In the
perinuclear region, the rough endoplasmic reticulum is extensively developed. The interior
of the cisternae are generally filled with a moderately electron-dense material.
Beyond the apical end of the nucleus, the cytoplasm of each cell contains a well-
developed Golgi apparatus, usually arranged parallel to the long axis of the cell (Fig. 10).
The Golgi apparatus is comprised of two to six flattened Golgi cisternae followed by several
layers of swollen cisternae and Golgi vacuoles. Many smooth-surfaced vesicles (up to 150
nm in diameter), which are probably Golgi vesicles, are scattered throughout the apical
cytoplasm, often near the membrane-bound iron-containing granules (Fig. 10).
The paucity of ferritin granules in the central region of the cap cells suggests that their
transport from the basal pole to the granule zone is rapid.
Figure 10. The cap cells contain a well-developed Golgi apparatus between the nuclei and the granule
zone. The cap cell apical cytoplasm contains abundant Golgi vesicles. AmG. amorphous granules; Fn ,
a paracrystalline ferritin granule; ER, endoplasmic reticulum.
6.3.3. The Apical Pole of the Cap Cells

The apical ends of the cap cells terminate directly on the surface of the major lateral
tooth cusps (Figs. 4 and 11). Hence, the apical poles of the cap cells are intimately involved
in the processes which eventually result in the impregnation of the cusps with the mineral
magnetite. The apical pole of each cap cell is divisible into two distinct regions: the zone
of iron-containing granules, and the zone of microvilli.
6.3.3a. The Granule Zone. The cytoplasm of the cap cells contains a dense ac-
cumulation of iron-containing, electron-dense granules (Figs. 4 and 11). The granule zone
begins at approximately the level of the Golgi apparatus, extends for 5-15 J..lm, and ter-
minates at a distance 5-10 J..lm from the tooth surface. Sections through it contain granules
with a wide range of diameters. As the thickness of an ultrathin section (-80 nm) is small
compared to the diameter of an individual granule (-1000 nm or 1 J..lm). the apparent size
of a granule is a function of its intact diameter and of the distance of its center from the
plane of sectioning.
An estimate of the range of granule sizes can be achieved by measuring the diameters
of only those granules whose outer membrane appears to be sectioned nearly transversely
(Le., the membrane exhibits a clear trilayered structure). By this method, we estimate that
the iron-containing granules range from 0.3 to 3.0 J..lm in diameter.
From the basal pole to the level of the nucleus, only ferritin-filled granules can be
found in the cap cell cytoplasm. At the level of the Golgi apparatus, near the basal edge
350 Chapter 16
Figure 11. A ferrihydrite-impregnated tooth cap (TC) and the apical pole of the cap cells. The cells
terminate nearly perpendicular to the tooth surface. The iron-containing electron-dense granules adjoin
a mitochondrion-rich apical cytoplasm which forms into bundles of microvilli (MV) that extend di-
rectly to the tooth surface. The pattern of ferrihydrate deposits follows the underlying matrix fibers
(which are visible toward the tooth tip). Note the much greater ferrihydrite density in the posterior
half of the tooth. Note the amorphous layer over the tooth tip and continuing along the posterior
surface of the cap. GZ, granule zone.
of the granule zone, granules filled with other forms of electron-dense iron-containing
material first appear. It is probable that the ferritin granules are the source from which the
other types of granules arise.
The ferritin granules are filled with the 6-nm electron-dense micelles characteristic
of ferritin. As previously reported (Towe et al., 1963; Towe and Lowenstam, 1967), there
are two types of ferritin-containing granules; crystalline (Fig. 12A) and paracrystalline (Fig.
12B). The crystalline granules contain hexagonally packed ferritin molecules with a lattice
spacing of 9.5 ± 0.5 nm. Usually, several crystal domains are found in a single granule,
but single-domain granules do occur (Fig. 12A). Crystalline granules are rare; they make
up less than 2% of the ferritin granule population. Paracrystalline ferritin granules are, on
average, about 1 fLm in diameter, but a small proportion are as large as 3 fLm. The center-
to-center spacing between micelles is 11.4 ± 0.1 nm. The paracrystalline granules comprise
about one-third of the total population within the granule zone.
The intermediate granules, also approximately 1 fLm in diameter, contain an internal
core of paracrystalline ferritin and an outer amorphous layer of small electron-dense par-
ticles which form aggregates of irregular shape and variable size (Fig. 13). The range of
appearance of intermediate granules extends from those with a thin shell of amorphous
particles and a large core of ferritin (Fig. 13A), through those with roughly equal amorphous
and ferritin layers (Fig. 13C), to those with only a tiny central core of ferritin (not shown).
We observe that intermediate granules comprise 3-5% of the population. This is a low
estimate, as these granules will be properly identified only if the plane of sectioning in-
cludes the ferritin core.
Within the core region, the 6-nm ferritin micelles are embedded in a matrix of moderate
electron density (Fig. 13D) which we presume to be apoferritin protein. At the junction
with the amorphous zone, there are usually areas of much lower electron density (Fig.
13A,C,D) which contain scattered typical6-nm micelles. We interpret these areas as regions
where the apoferritin has disappeared, either by proteolysis or by solubilization. Upon
their release from the apoferritin shells, the ferrihydrite micelles either are actively con-
verted or spontaneously dissociated into the amorphous electron-dense material. The char-
acteristic "nibbled" outline of many of the intermediate granules (Fig. 13A,B) may arise
from fusion of the granules with small primary lysosomes (Golgi vesicles) containing pro-
teolytic enzymes.
The previously described amorphous granules (Towe and Lowenstam, 1967) (Fig.
12D,E) are the most abundant class, comprising well over one-half of the total. They are
entirely filled with amorphous electron-dense particles identical to the outer layer of the
intermediate granules. Even at high magnification (Fig. 12C), the size and shape of the
individual particles are difficult to discern within the amorphous aggregates; the particles
are at most 2 nm in diameter, but the majority are appreciably smaller. The amorphous
granules range in diameter from 0.2 to 0.8 fLm, about 20-30% smaller than the ferritin
granules. The amorphous granules within a single thin section may differ substantially in
electron density; this suggests that they contain a range of concentrations of electron-dense
particles.
There is no observable arrangement of the several granule classes within the granule
zone; granules of all types are seemingly randomly distributed throughout.
Our observations on the iron-containing granules suggest a close parallel with the
process of hemosiderin formation in vertebrates (Sturgeon and Shod en, 1964). Hemosiderin
is an insoluble form of iron present as yellow intracellular, membrane-bound granules
which give an intense positive Fe 3 + staining reaction. Hemosiderin is generally considered
to be a product of intracellular degradation and aggregation of ferritin. X-ray diffraction
studies of hemosiderin indicate that it is composed of ferrihydrite, identical to the micellar
mineral of ferritin (Fischbach et aI., 1971). Electron micrographs of hemosiderin vesicles
352 Chapter 16
Figure 12. (A) Crystalline ferritin granule. (B) Paracrystalline ferritin granule. (C) Detail comparing
ferritin granule and amorphous granule. (0, E) Amorphous granules. Bars = 100 nm.
Figure 13. (A-C) Intermediate granules. (D) Detail of (C). Bars = 100 nm.
Uacobs et a1., 1978). Although we have been unsuccessful in our efforts to obtain useful
data from selected-area electron diffraction of unstained sections through individual amor-
phous granules, we infer that they are equivalent to hemosiderin and are composed of
ferrih ydrite.
6.3.3b. The Microvillus Zone. The apical end of each cap cell is formed into a
bundle of 500-1000 microvilli which arise from the apical cytoplasm and terminate at the
surface of the tooth cap (Figs. 11, 14, and 16-18). The apical cytoplasm also extends nearly
354 Chapter 16
Figure 14. (AJ Cross section through the microvillus bundles of the apical end of the cap cells. Note
the mitochondrion-filled cytoplasmic shells around the microvillus bundles. Also note the central
mitochondrion-filled cytoplasmic protuberances. (BJ Detail of section similar to (AJ. CS. cytoplasmic
shell; CP, cytoplasmic protuberance; Mt, Mitochondria; MV, microvilli.
Figure 14. (continued)
to the tooth surface as a thin mitochondrion-filled shell which surrounds the entire bundle
of microvilli of each individual cell (Fig. 14) . In addition, one or more broad fingers of
mitochondrion-rich cytoplasm commonly protrude toward the tooth from the center of the
cell (Fig. 14). Consequently, the microvilli of a single cell vary widely in length, from 2
, , • ú ? = n ú = • _ _
356 Chapter 16
outer cytoplasmic shell, or the main body of apical cytoplasm (Fig. 11). Rough calculations
indicate that the microvilli provide a greater than 100-fold increase in surface area of the
cell terminus.
The microvilli, 80-120 nm in diameter, are usually arranged in a tightly packed, hex-
agonal array (Figs. 14, 15). Each microvillus is delimited by a single (trilayered) unit mem-
brane approximately 6 nm thick (Figs. 14B and 15B). The outer surface is covered with a
fuzzy coating of finely fibrillar material which may extend from 20 to 30 nm. The interior
of each microvillus contains a loose network of thin fibrils, 3-4 nm in diameter, which
lie roughly parallel to its long axis and are more circumferential than central (Figs. 14 and
15).
The fact that the microvilli of the cap cells terminate directly on the surface of the
tooth caps is unusual. Microvilli are usually found on the surfaces of cells which line the
walls of a lumen (e.g., the epithelial lining of the digestive tract), where they serve to
increase the surface area for the purpose of absorption of soluble materials. In the cap cells,
the likely function of the microvilli is to increase the surface area for the purpose of se-
cretion.
The mass of apical cytoplasm, from which the microvilli arise, is characterized by the
presence of numerous mitochondria (Figs. 4, 11, and 14); these account for the dense
staining by toluidine blue of this region in I-fJ.m sections (Fig. 4). The mitochondrion-rich
cytoplasm extends from the base of the granule zone to the microvilli, with the highest
concentration of mitochondria occurring in the shell of cytoplasm surrounding the mi-
crovillus bundles and within the cytoplasmic protuberances which approach the cusp
surface (Fig. 14A). These mitochondria are densely stained and highly variable in size and
shape; many, particularly within the surrounding shell, are long and sinuous. The high
concentration of mitochondria at the apical pole of the cap cells indicates that large
amounts of energy are available for any or all of the following steps: the final intercellular
processing of iron; iron transport across the microvillus membranes; the precipitation of
ferrihydrite in the matrix fibrils; the transformation of ferrihydrite to magnetite within the
matrix.
At the apical edge of the granule zone, the iron-containing granules are in close prox-
imity to the microvilli (Figs. 11 and 15). Only within this specific region do we observe
a double-membrane structure tightly apposed to the inner face of the plasma membrane
of the cytoplasm surrounding the microvillus bundles (Fig. 15). The structure occurs both
in the cytoplasmic shell and in the cytoplasmic protuberances described above. Closer
examination reveals the structure to be identical to a cisterna of endoplasmic reticulum
(Fig. 15B). Ribosomes are attached only to the cytoplasmic face of its inner membrane (Fig.
15B); the outer membrane is bare and lies about 6-7 nm from the cell membrane. We have
seen no published reports of similar specialized cisternae. From a consideration of its
appearance, our initial surmise as to the function of the double-membrane structure was
that it served as a barrier to diffusion of something through the plasma membrane, but
whether the diffusion was into or out of the cell was unclear. In this context, however,
the membrane structure's extremely limited areal extent is puzzling. It forms sheets sur-
rounding only the proximal 1-2 fJ.m of the microvillus bundles. We now consider it more
likely that these specialized cisternae serve to prevent exocytosis of the iron-containing
granules. Exocytosis is equivalent to the reverse of endocytosis (see Section 6.3.1): it occurs
by the fusion of an intracellular vesicle membrane with the cell membrane, resulting in
the secretion of the vesicle contents into the extracellular space. The specialized cisternae
prevent the iron-containing granules from reaching and fusing with the plasma membrane
at the apical pole of the cap cells. Thus, the granule iron cannot be shunted directly into
the extracellular space, but must undergo the appropriate terminal processing before being
secreted by the microvilli.
Figure 15. Cross section through the upper part of the microvillus bundles near the granule zone. The
specialized cisternae (SC) are directly apposed to the cell membranes of the cytoplasmic shell (CYTO)
surrounding the microvillus bundles. The arrows indicate ribosomes on the inner face of the cisternae.
Fn, ferritin granule; MV, microvilli; Mt, mitochondria.
358 Chapter 16
Figure 16. Ferrihydrite spherules in the tooth cap matrix. Note how the fine fibrils aggregate into
layers to define cavity walls. FH, ferrihydrite; Ma, matrix fibrils; MV, microvilli of cap cells.
We have no positive findings that elucidate the nature of the terminal processing of
iron within the cap cells. However, the following negative findings are relevant: (1) we
have not observed the lysis of any iron-containing granules, releasing their electron-dense
contents into the apical cytoplasm; (2) we see no membrane-bound vesicles within the
microvilli; (3) we see no electron-dense 'particulate material (larger than - 2 nm) either
in the apical cytoplasm or within the microvilli.
6.4. Ultrastructure of the Tooth Matrix
The protein-chitin fibrils which comprise the tooth cap matrix of L. hartwegii are
about 2 nm across (Figs. 16 and 18), whereas those of C. stelleri are 5 nm (Towe and
Lowenstam, 1967). The 2 nm fibrils are massed into sheets, approximately 20 nm thick,
which form the walls of an array of elongate polygonal cavities (Figs. 11 and 16). The
deposition of ferrihydrite in the tooth cap occurs predominately in the form of 100- to
150-nm, roughly spherical, aggregates (Fig. 16), composed of tiny crystallites of indeter-
minate size (Towe and Lowenstam, 1967). It had been suggested, but not fully demon-
strated, by Towe and Lowenstam (1967) that the deposition of ferrihydrite occurs directly
on the organic framework. Low-magnification micrographs of ferrihydrite-containing tooth
caps of L. hartwegii fully validate this suggestion (Fig. 11). The pattern of ferrihydrite
deposition matches exactly the arrangement of the underlying organic matrix. The tooth
matrix possesses a high degree of order. The long axes of the polygonal cavities extend
Figure 17. Ferrihydrite spherules (FH) in the exterior portion of the amorphous layer (AmL). These
spherules are larger than the ferrihydrite aggregates on the tooth matrix (TM) (at lower right) . MV,
microvilli; Mt, mitochondria.
360 Chapter 16
Figure 18. Earliest magnetite deposition. The magnetite crystals (Mag) appear as tabulate parallele-
pipeds. Note how they are frequently at a specific angle to the organic framework of the tooth cap.
FH, ferrihydrite deposits; MV, microvilli.
nearly parallel to the anterior face of the tooth cusp. Centrally, the cavity array curves
through a nearly 1800 turn so that the cavity axes abut the posterior face of the cap at an
acute angle. This pattern is maintained by the magnetite crystals of mature caps examined
by SEM (Kirschvink and Lowenstam, 1979). The ordered array of crystals may serve to
maximize the mechanical strength or abrasion resistance of the mature teeth.
During the initial mineralization stage, ferrihydrite deposition is much heavier in a
5- to 7-f.Lm layer along the posterior face of each tooth cap (Fig. 11). We are uncertain as to
whether this variation of deposition is a result of differential cellular activity or localized
differences in matrix composition.
The whole posterior surface of the ferrihydrite-impregnated tooth caps is covered with
a layer of amorphous material, 1.5-2.0 f.Lm thick (Figs. 11 and 17); the layer becomes much
thicker between the basal portions of the tricuspid prongs in the caps of L. hartwegii and
M. muscosa (not shown). The coating material contains 200-nm spherical aggregates of
ferrihydrite (identified by electron diffraction), concentrated toward its outer surface. Al-
though the time of formation of this layer is not known, nor is its function at all understood,
we call attention to its existence because its presence may confuse the interpretation of
scanning electron micrographs of cell-free tooth cap surfaces.
The initial occurrence of magnetite within the tooth caps has been identified by the
appearance of its characteristic diffraction pattern superimposed on the ferrihydrite pattern
in studies of unstained sections of the tooth cap. Electron microscope examination of
similar stained material shows a firm correlation of the magnetite diffraction pattern with
the presence of tabulate electron-dense crystals in the teeth (Fig. 18). The magnetite crystals
form among or in close proximity to, but without readily apparent direct association with,
the ferrihydrite deposits. However, most of the magnetite crystals appear to lie at a specific
angle to the long axis of the matrix cavities (Fig. 18), which suggests a role of the organic
matrix in the nucleation of the magnetite crystallites. We do not yet understand the roles
played by the organic matrix and by possible additional cell-secreted substances in the
biochemical processes which lead to the conversion of ferrihydrite to magnetite.
7. Concluding Remarks
We have established that in chitons, the ongoing delivery of the large amounts of iron
needed for radula tooth magnetite biomineralization occurs in the form of ferritin which
functions as an iron transport protein both in the circulatory system and, aggregated within
vesicules, through the cytoplasm of the epithelium. The involvement of membrane-bound
accumulations of ferritin in the formation of iron-containing tooth-capping minerals is not
limited to chitons: patellid gastropods (unpublished observations), urodelan amphibians
(Randall, 1966), and rats (Reith, 1961) all contain numerous ferritin-filled, membrane-
bound granules in the dental epithelium near the site of iron mineral deposition. The
utilization of ferritin for large-scale iron transport may well be universal.
In vertebrates, the conversion of ferritin iron to particulate ferrihydrite (Le., hemo-
siderin) has generally been considered to be a mechanism for dead-end storage of excess
iron in a nontoxic poorly mobilizable form. In the chitons, it appears that the vesicle-
bound free ferrihydrite may be a transitional stage in iron transport. We do not yet know
by what means the iron moves from the amorphous granules to the tooth matrix. How
many different transport agents are required to effect the transmembrane and trans cytosol
movement of iron? In what chemical form(s) is the iron carried? Answers to these questions
will require additional biochemical studies; the results may well be of applicability in
unraveling the mechanisms involved in iron mineral formation in other organisms.
ACKNOWLEDGMENTS. We express our appreciation to Dr. E. R. Berger for initiating us into

the art of thin sectioning. We thank Pat Koen for his valuable assistance with the electron
microscope and Joyce Nesson for her help with preparation of the manuscript. M.H.N. was
supported by USPHS Training Grants to the Division of Biology, California Institute of
Technology. The study was also supported by NSF Grant EAR-76-03725 to H.A.L. This is
Contribution No. 4038 from the California Institute of Technology.
References
Arvy, L., and Gabe, M., 1949, Contribution a l'etude morphologique du sang des Polyplacophores,
Bull. Soc. Zoo1. Fr. 74:172-179.
Brody, I., 1959, The keratinization of epidermal cells of normal guinea pig skin as revealed by electron
microscopy, J. Ultrastruct. Res. 2:482-511.
Bruns, R R, and Palade, G. E., 1968, Studies on blood capillaries. II. Transport of ferritin molecules
across the wall of muscle capillaries, J. Cell BioI. 37:277.
Carefoot, T. H., 1965, Magnetite in the radula of the Polyplacophora, Proc. Malacol. Soc. London
36:203-212.
Farquhar, M. G., and Palade, G. E., 1960, Segregation of ferritin in glomerular protein absorption
droplets, J. Biochem. Biophys. eytol. 7:297-304.
Fischbach, F. A., Gregory, D. W., Harrison, P. M., Hoy, T. G., and Williams, I. M., 1971, On the structure
of hemosiderin and its relationship to ferritin, J. Ultrastruct. Res. 37:495-503.
362 Chapter 16
Fischer, D. S., and Price, D. C., 1964, A simple serum iron method using the new sensitive chromogen
tripyridyl-s-triazine, Clin. Chern. 10:21-31.
Fretter, V., 1937, The structure and function ofthe alimentary canal of some species ofPolyplacophora,
Trans. R. Soc. Edinburgh 59:119-164.
Gabe, M., and Prenant, M., 1948, Quelques aspects cytologiques du metabolisme du fer chez Acan-
thochites fascicularis L., Arch. Anat. Microsc. Morphol. Exp. 37:136-154.
Gabe, M., and Prenant, M., 1952, Sur Ie role des odontoblastes dans l'elaboration des dents radulaires,
C. R. Acad. Sci. 235:1050-1052.
Gabe, M., and Prenant, M., 1959, Particularites histochimiques de l'appareil radulaire chez quelques
mollusques, Ann. Histochim. 3:95-112.
Giese, A. C., 1952, Myoglobin in radular muscles of chitons, Anat. Rec. 113:609a.
Heath, H., 1903, The function of the chiton subradular organ, Anat. Anz. 23:92.
Jacobs, A., Hoy, T. G., Humphrys, J., and Perera, P., 1978, Iron overload in Chang cell cultures, Br. J.
Exp. Pathol. 59:589-598.
Kirschvink, J. L., and Lowenstam, H. A., 1979, Mineralization and magnetization of chiton teeth:
Lett. 44:193-204.
Am. Bull. 73:435-438.
Lowenstam, H. A., 1972, Phosphatic hard tissues of marine invertebrates: Their nature and mechanical
function and some fossil implications, Chern. Geol. 9:153-166.
Luft, J. H., 1961, Improvements in epoxy resin embedding methods, J. Biophys. Biochem. Cytol. 9:409-
414.
Manwell, C., 1958, The oxygen-respiratory pigment equilibrium of the hemocyanin and myoglobin of
the amphineuran mollusc, Cryptochiton stelleri, J. Cell. Compo Physiol. 52:341-352.
Markel, K., 1958, Bau and Funktion der Pulmonaten-radula, Z. Wiss. Zool. 160:213-289.
Plate, 1. H., 1897, 1899, 1902, Die Anatomie und Phylogenie der Chitonen: 1897, Zool. Jahrb. Suppl.
IV (Fauna Chilensis I), 1-243; 1899, ibid. Part B 2:15-216; 1902, ibid. Part C 2:281-600.
Prenant, M., 1928, Quelques aspects histologiques du metabolisme du fer chez les chitons, Arch. Anat.
Microsc. Morphol. Exp. 24:1-7.
Randall, M., 1966, Electron microscopical demonstration of ferritin in the dental epithelial cells of
urodeles, Nature 210:1325-1326.
Raven, C. P., 1966, Morphogenesis: The Analysis of Molluscan Development, 2nd ed., Pergamon Press,
Elmsford, N.Y.
Reith, E. J., 1961, The ultrastructure of ameloblasts during matrix formation and the maturation of
enamel, J. Biophys. Biochem. Gytol. 9:825-839.
Richardson, K. C., Jarett, L., and Finke, E. H., 1960, Embedding in epoxy resins for ultrathin sectioning
in electron microscopy, Stain Technol. 35:313-323.
Rottman, G., 1901, Uber die Embryonalentwicklung der Radula bei den Mollusken. I. Die Entwicklung
der Radula bei den Gephalopoden, Z. Wiss. Zoo1. 70:236.
Runham, N. W., 1963a, The histochemistry of the radulas of Acanthochitona communis, Lymnaea
stagnalis, Helix pomatia, Scaphander lignarius and Archidoris pseudoargus, Ann. Histochim.
8:433-442.
Runham, N. W., 1963b, A study of the replacement mechanism of the pulmonate radula, Q. J. Microsc.
Sci. 104:271-277.
Ryser, H., Caulfield, J. B., and Aub, J. c., 1962, Studies on protein uptake by isolated tumor cells. I.
Electron microscopic evidence of ferritin uptake by Ehrlich ascites tumor cells. J. Cell BioI. 14:255-
268.
Schnabel, H., 1903, Uber die Embryonalentwicklung der Radula bei den Mollusken. II. Die Entwicklung
der Radula bei den Gastropoden, Z. Wiss. Zoo1. 74:616-655.
Sturgeon, P., and Shoden, A., 1964, Mechanism of iron storage, in: Iron Metabolism: An International
Symposium (F. Gross, ed.), Springer-Verlag, Berlin, pp. 121-146.
Thiele, J., 1909, ijevision des Systems der Ghitonen, Part I, Schwiezerbartsche, Stuttgart.
Towe, K. M., and Bradley, W. F., 1967, Mineralogical constitution of colloidal "hydrous ferric oxides,"
J. Colloid Interface Sci. 24:382-392.
Towe, K. M., and Lowenstam, R. A., 1967, Ultrastructure and development of iron mineralization in
Towe, K. M., Lowenstam, R. A., and Nesson, M. R., 1963, Invertebrate ferritin: Occurrence in Mollusca,
Science 142:63-64.
Venable, J. R., and Coggeshall, R, 1965, A simplified lead citrate stain for use in electron microscopy,
J. Cell BioI. 25:407-408.
Webb, J., and Macey, D. J., 1983, Plasma ferritin in Polyplacophora and its possible role in the biom-
ineralization of iron, in: Biomineralization and Biological Metal Accumulation (P. Westbroek and
E. W. de Jong, eds.), Reidel, Dordrecht, pp. 423-427.
Chapter 17
Magnetic Remanence and Response to
Magnetic Fields in Crustacea
RUTH E. BUSKIRK and WILLIAM P. O'BRIEN, JR.
1. Introduction, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , 365
1.1, Major Questions and Approaches in This Field , , , , , , , , , , , , , , , , , , , , , , , 365
1.2, Effects of Magnetic Fields on Crustacea, , , , , , , , , , , , , , , , , , , , , , , , , , " 366
2, Experimental Studies, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , " 368
2,1. Experimental Rationale-Migratory vs, Sedentary Crustacea , , , , , , , , , , , , , , , 368
2,2, General Methods and Techniques, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , 370
2,3, Measurement Results , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , 371
3, Discussion, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , 377
3,1. Magnetic Remanence in Shrimp and Barnacles, , , , , , , , , , , , , , , , , , , , , , " 377
3,2, Relation to Studies of Other Arthropods, " " " " " " " " " " " " " , 378
3,3, Significance of This Study and Suggestions for Future Work, , , , , , , , , , , , , " 379
4, Summary , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , " 380
References, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , " 381
1. Introduction
Crustacea comprise a major portion of the diversity and biomass of marine fauna, yet their
biology is little known in comparison to their terrestrial counterparts, the insects, These
marine arthropods display oriented movements in their habitat as well as timing of be-
havior with respect to environmental stimuli. There have been very few studies, however,
of either the effects of magnetic fields on Crustacea or the magnetic fields generated by
living Crustacea, In this chapter, we report the first measures of significant magnetic re-
manence in Crustacea, we relate this biological magnetism to that described in other an-
imals, and we discuss its possible significance in the life history of shrimp and barnacles.
1.1. Major Questions and Approaches in This Field
The recent discovery that ferrimagnetic material can be synthesized by living orga-
nisms (reviewed by Kirschvink and Lowenstam, 1979; Lowenstam, 1981; Kirschvink, 1983)
has led to renewed investigations in biomagnetism, One line of research pursues a de-
scription of the distribution and composition of the magnetic material in various species,
Another is beginning to describe the chemical pathways involved in biosynthesis of mag-
RUTH E. BUSKIRK and WILLIAM p, O'BRIEN, JR, • Institute for Geophysics, University of Texas,
Austin, Texas 78712.
365
366 Chapter 17
netite (Lowenstam, 1981). In other studies the level of biological magnetic remanence is
being compared to remanence in the earth's sediments (Kirschvink, 1982a; Demitrack, this
volume).
The evolutionary advantages of possessing magnetite or other ferrimagnetic materials
are also areas of current conjecture. In some cases, specific physical properties of magnetite
such as its density, hardness, or electrical resistivity (Kirschvink, 1982b) could convey a
selective advantage to the organisms. It is also an intriguing possibility that biological
magnetite could be part of a mechanism for organisms detecting and orienting to the earth's
magnetic field. Such a response could involve the simple motion of a dipole in the earth's
field, as in the magnetotactic bacteria (Blakemore and Frankel, 1981) or, in larger animals,
a more complex neural transducer (Kirschvink and Gould, 1981; Kirschvink, 1982b). A
magnetic transducer could be capable of detecting both temporal and spatial variations in
the earth's magnetic field. With appropriate sensitivity and filtering of background noise,
animals could synchronize their circadian rhythms with daily variations in the earth's
field (Brown et 01., 1970) or could use inclination and/or declination angles to devise or
position a magnetic map (Gould, 1980; Quinn et 01.,1981). It is not surprising that in many
of the recent searches for biomagnetic remanence, investigators have used animals that
appear capable of long-distance navigation and homing (Gould et 01., 1978; Walcott et 01.,
1979; Zoeger et 01., 1981; Perry et 01., 1981; Quinn et 01., 1981; Walker and Dizon, 1981;
Jones and MacFadden, 1982). In this study we consider both migratory and sedentary
representatives of the class Crustacea. Previous laboratory and field work indicates that
some members of this class are affected by changes in magnetic fields, as described below.
1.2. Effects of Magnetic Fields on Crustacea
Research on the effects of magnetic fields on arthropods has concentrated on three

types of phenomena. (1) Unusually high or low magnetic fields affect physiological pro-
cesses, including rates of development and reproduction. (2) Periodic changes in geo-
magnetic field intensity may be associated with activity rhythms. (3) Changes in magnetic
field intensity or direction can cause changes in orientation behavior. In comparison to
insects, in which all of these phenomena have been addressed, relatively little study has
been devoted to the effects of magnetic fields on Crustacea.
For a variety of crustaceans, it has been shown that increased magnetic fields affect
metabolic rates. However, the experiments have not incorporated sufficient controls to
pinpoint the mechanism or to remove the possibility of a stress reaction to the large applied
field. Vasil'yev et 01. (1974) found for the water flea Daphnia magna that growth in a field
of 136 mT increased its rate of development but decreased its fecundity. Respiration rates
of brine shrimp Artemia salina were increased when they were hatched in fields of 20 or
180 mT (200 or 1800 De) (Taneyeva, 1978). Increases were greatest for larvae about 24 hr
old and were more pronounced for larvae in containers near the south pole of the magnet.
When brine shrimp were reared in a constant field of 105 mT (1050 De), their life span
increased by 21 %, and they displayed greater resistance to bacterial infections (Taneyeva
and Dolgopol'skaya, 1974). The increased magnetic field could simply have had a negative
effect on bacterial growth, but this was not investigated directly.
Taneyeva (1978) hypothesized that the mechanism for the increase in growth in un-
usually high magnetic fields was the facilitation of water absorption. She used barnacles
(Balanus eburneus) to test the effect of magnetic fields on osmotic responses to water of
different salinities. A field of 105 mT significantly increased the water uptake in a high-
salinity (55%) medium and increased the water loss in a low-salinity (5%) medium in
comparison to controls in seawater.
Crustacea 367
In a study of the effects of magnetic fields on regeneration of limbs in fiddler crabs

(Uca pugilator and U. pugnax), Lee and Weis (1980) stressed the significance of the mag-
netic field gradient. Following loss of appendages, crabs placed at the south pole end of
a magnetic field of 10 mT regenerated and molted sooner than controls, while crabs at the
north pole were delayed in development. The investigators suggested that a differential
sodium ion influx would affect development rates in such a way.
Many species of Crustacea display temporal patterns of behavior that, as demonstrated
experimentally, may be correlated with diverse environmental stimuli (Palmer, 1974; En-
right, 1978). It is possible that variations in the earth's magnetic field serve as time cues
for these animals. Although few species have been investigated for response to magnetic
field changes, controversy over the nature of biological clocks (Brown et aI., 1970) stim-
ulated a great deal of experimental work on the response of crustaceans to environmental
stimuli. For example, fiddler crabs are known to display daily cycles of metabolic rate and
color changes related to the timing of the tides (Barnwell, 1966). Although the phase of
the circadian rhythm can be reset in an artificial light or temperature regime, the rhythm
persists 4-5 days in crabs isolated from any light, temperature, or tidal cues (Webb and
Brown, 1965). Variations in the metabolic cycle appear to be related to variations in the
geomagnetic field (Brown et al., 1970); however, direct tests of the influence of magnetic
fields on fiddler crabs have yet to be made.
Only a few Crustacea have been tested for behavioral responses to magnetic fields.
Work is currently under way to determine if spiny lobsters (Panulirus argus) are capable
of responding behaviorally to magnetic fields (Kenneth Lohmann, personal communica-
tion). Caribbean spiny lobsters display striking mass migrations from shallow areas to more
open water in the autumn (Herrnkind, 1969). Following periods of queue formation, es-
pecially on days after storms, group movement is strongly directional (Herrnkind and
Kanciruk, 1978). Several factors, including the direction of wave surge and water currents,
appear to be involved in orientation. In preliminary experiments by Walton and Herrnkind
(1977), bar magnets were attached to the walking legs of blinded spiny lobsters in order
to distort the magnetic field around their heads. Their orientation did not differ from that
of controls; instead, the experimental animals oriented into the surge prevalent at the time
of testing.
To our knowledge there are no published studies of orientation by shrimp, crayfish,
or crabs in a magnetic field. In the often-cited research involving upside-down orientation
of crayfish in magnetic fields (Ozeki and Osada, 1977), iron particles were experimentally
substituted for the natural rock grains serving as statoliths in the sensory capsule of the
exoskeleton. The response of neural receptors, disturbed normally when gravity pulls the
rock grains downwards, was in this case triggered by iron particles being deflected upwards
by the magnet. Thus, this example is not a case of magnetic orientation.
The only documented cases of magnetic orientation in Crustacea involve the beach
orientation movements of sand hoppers (Amphipoda). Sand beach organisms rely on a
variety of cues for directional homing including visual landmarks, slope, sun and moon
directions, and hydrostatic pressure (Enright, 1978). In addition, it appears that the type
of cue employed depends upon several conditions such as the amount of body water and
relative humidity, the level and anisotropy of light (for directional information), and
whether the animal is crawling (as in feeding) or jumping (M. C. Arendse, personal com-
munication). The sandhopper Talitrus saltator oriented to the earth's magnetic field in
complete darkness by jumping toward the direction parallel to the orientation of the coast-
line from which the specimens were taken (Arendse, 1978). When the local geomagnetic
field was cancelled with Helmholtz coils, the orientation of the sandhoppers became ran-
dom. When the natural field was artificially compensated and shifted, the animals oriented
to the artificial field (Arendse and Kruyswijk, 1981). The nocturnal, freshwater or estuarine
amphipod Orchestia cavimana displays the same jumping orientation to natural and al-
368 Chapter 17
tered magnetic fields, within acceptable humidity levels, in the absence of light and gravity
cues (Arendse and Barendregt, 1981).
2. Experimental Studies
2.1. Experimental Rationale-Migratory vs. Sedentary Crustacea
The Crustacea species chosen for this study were brown shrimp and acorn barnacles.
Both species have planktonic larvae and postlarval stages which live in bays and estuaries.
While barnacles become sedentary and fixed in a particular orientation as they grow,
shrimp can swim great distances and burrow into the sediments at a new site nightly.
Intertidal barnacles and brown shrimp both display circadian or tidal activity rhythms
(Palmer, 1974). The migratory brown shrimp and sessile barnacles are compared here to
each other as well as to Palaemonetes grass shrimp from the same habitat.
2.1.1. Shrimp Movements and Cycles

Several species of Penaeus shrimp (Crustacea, Penaidea) show long-distance migration
in the Gulf of Mexico and Atlantic Ocean associated with specific stages of their life cycles.
Adult prawns spawn in offshore waters where the larvae develop. Postlarvae move into
coastal estuaries, and young shrimp remain there for a few months, feeding and growing.
Juveniles and subadults emigrate from the estuaries to the open ocean during the summer
and fall. In the case of the brown shrimp (Penaeus aztecus), postlarvae of body length 8
to 14 mm move into shallow water in winter or spring, preferring shallow habitats over
deep bays (Fry, 1981), and leave the estuaries at a size of about 80 mm (Trent, 1966).
Mark-recapture experiments with brown (P. aztecus), white (P. setiferus), and pink
(P. duorarum) shrimp indicate movements of individuals up to 115 km in Texas popu-
lations (Klima, 1963) and up to 575 km in the Atlantic (North Carolina Department of
Conservation and Development, 1967). General movement patterns run parallel to the
coastline (in water depth ranges typical of the species). Nearly all Carolina recaptures
indicated a movement southward, while shrimp along the Gulf of Mexico coast moved
either to the east or to the west.
Laboratory experiments with P. duorarum indicate that movements into and out of
the estuaries are facilitated by selective use of tidal currents (Hughes, 1972). Postlarvae
swim upstream at times of flood tides and downstream during ebb tides, and this tidal
rhythm of responses persists for several days in the laboratory with no externally changing
cues. Juveniles, which are caught more frequently in the ebbing tide, demonstrated a similar
trend, although they did not swim downstream at times in the closed laboratory corre-
sponding to daytime low tides.
Shrimp show a rhythm of nocturnal swimming and diurnal burrowing (Aldrich et al.,
1968), and the time of emergence from the sand is quite synchronous. Tests with pink
shrimp (Hughes, 1968) indicate that responses to light cues account for the synchronous
emergence of shrimp from the sand to begin feeding, and larger individuals (greater than
40 mm in body length) are more responsive to external cues. In the laboratory a daily cycle
of shrimp activity can be recorded, and it persists at least 3 days in a regime of continuous
darkness. A second activity cycle is also present that resembles a tidal rhythm (Hughes,
1968). The external cues and internal mechanisms controlling such rhythms in shrimp
have not been determined.
Although movements out of the estuary appear to be related to tides and salinity
preference, factors directing shrimp movements along the coast are unknown. Long-dis-
Crustacea 369
Figure 1. Schematic representa- y í J J l K Rã ú H? ? DJ J Rã J J J J J K Ñ=
bJ
tion of the orientation of the two
magnetic fields (each 100 mT)
l,51
through which seawater flowed
in each of the two magnetic test
SEAWATER
----7
}TES?
ANEL )
lines. Fouling densities were de-
termined from measurements of
mass accumulation on remova- 'i' 151 fNT
ble PVC test panels downstream
from the magnetic fields. "'pvc PIPE
tance movements of tagged Penaeus shrimp are well documented, but the mechanisms
underlying orientation and navigation have not been established. Neither marine recapture
studies nor laboratory experiments designed to detect possible homing or return move-
ments of individuals in the spring have been completed. However, many areas along the
Atlantic and Gulf coasts retain high shrimp concentrations, so it is doubtful whether the
documented movements represent only a one-way dispersal phase. If some adults do return
to areas near where they were spawned, present data on shrimp response to water currents,
salinity, temperature, and sediments (Grady, 1971) could not explain such movement pat-
terns. A long-distance reference cue such as the earth's magnetic field would be needed.
2.1.2. Marine Fouling Experiments

One application of knowledge of biomagnetism in Crustacea may be in the design of
technology for controlling their settling and growth. Balanus eburneus and B. amphitrite
niveus are barnacles common in low to middle intertidal zones in the Gulf of Mexico. The
planktonic larvae settle and quickly colonize exposed intertidal surfaces.
In a recent experimental study in Galveston Bay, Gulf of Mexico, O'Brien (1985)
showed that barnacles and other marine macrofouling organisms grew more densely in
seawater that had been channeled through magnetic fields of 100 mT than in control chan-
nels. Growth of fouling organisms in PVC pipelines containing flowing seawater that had
passed through two transverse magnetic fields (Fig. 1) was compared to that in otherwise
identical control lines. The mass of fouling organisms growing on PVC plates (Table I) was
significantly greater in the two magnetic field lines than in the control lines, even though
the settling density of barnacles (the dominant fouling organisms) was about the same for
Table I. Observed Marine Fouling

Densities DO
Day 42 Day 117
Test line D (kglm 2 ) D (kglm 2 )
Magnetized #1 1.06 5.87

Magnetized #2 1.18 4.54
Control #1 0.63 3.93
Control #2 0.83 3.85
a After 42 days and 117 days of growth in PVC
pipes for two lines where the seawater had
passed through two magnetic fields (100 mT)
and for two identical control lines containing
no magnetic fields.
370 Chapter 17
all four plates. In the experiment, the seawater flow rates were 80-120 liters/min (fluid
velocities about 0.3-0.4 m/sec). The observed response of macrofouling organisms to mag-
netic fields in their environments suggests that more information is needed about the
magnetic properties of the organisms themselves. Future experiments could be designed
to determine the conditions and parameters that maximize growth (e.g., mariculture ap-
plications) or to determine if there exist combinations of magnetic field intensities and
fluid flow velocities that inhibit growth (e.g., marine biofouling control; Becka et al., 1981).
2.2. General Methods and Techniques
All measurements were made in the University of Texas Paleomagnetism Laboratory

in Galveston during the fall of 1980 and summer 1981. Living specimens were held in
natural or artificial seawater in glass or plastic containers. Any dissection or manipulation
of specimens was done with plastic picnic knives or pieces of broken glass. Excess water
was blotted off with cotton cloth, and the specimens were weighed or measured before
being placed in the magnetometer.
Initial tests showed that use of plastic or paper sample holders increased magnetic
noise above acceptable levels. Therefore, during measurements the specimens rested in a
short, grooved plug of Styrofoam inserted inside a Mylar tube as a sample holder. For the
shrimp samples, magnetic intensity of the tube and plug alone was measured following
one or two sample measurements and was subtracted from the sample reading. The tube,
plugs, and other materials were kept separate from rock paleomagnetism equipment and
were cleaned and demagnetized when their remanence exceeded background levels.
All magnetic measurements were made with a superconducting magnetometer (Goree
and Fuller, 1976) that is sensitive to extremely weak magnetic fields, on the order of 10- 11
A m2 (equal to 10- 8 emu). Specimens were held at room temperature inside a vertical,
3.8-cm-diameter access port. Specimens were consistently oriented with their axis of bi-
lateral symmetry oriented along the x axis of measurement. Magnetometer readings were
replicated (each reading was actually an average of 10 or 20 scans) in each of three axes
while the specimen was in place in the holder. When any of the pairs of average readings
differed by more than 20%, they were rejected and the measurement repeated. At the time
of each measurement, we recorded the data and calculated magnetic vectors and intensities
on a Tektronix 4051 computer. Final specimen preparation and magnetometer measures
were conducted in a magnetically shielded room (see Scott and Frohlich, this volume) in
which the average field intensity was about 100 nT (100 'Y).
Initially, the natural remanent magnetization (NRM) (see McElhinny, 1973, for dis-
cussion of magnetization) was measured for each freshly collected specimen. To determine
values of saturation isothermal remanent magnetization (sIRM) , specimens were magnet-
ized using a calibrated electromagnet with variable fields up to 800 mT. The specimen
rested on a clean wooden holder at the center of the gap between the poles of the elec-
tromagnet for 10 or 20 sec. For demagnetization studies we used a Schoenstedt single-axis
alternating field (AF) demagnetizer, repeating each demagnetization for the three orthog-
onal axes of each specimen. In most cases, AF demagnetization involved a sequence of
increasingly strong peak fields and increasing decay rates up to a peak field setting of 600
Oe with a decay rate of 25 mae/half-cycle.
Contamination is a major problem in studies of biological remanent magnetism. In
addition to avoiding potential contamination sources during the process of collecting,
holding, and preparing the specimen for measurement, it was also necessary to avoid
feeding the organisms contaminated food. For example, shrimp food prepared and used
by the National Marine Fisheries Laboratory (NOAA, Galveston) had a high magnetic re-
manence (NRM of 10- 7 A m2 /cm 3 , sIRM of 10- 6 A m2 /cm 3 ), probably because the material
Crustacea 371
....
N
E
4
ú =
o....I
;-9
Figure 2. Magnetization per in- o
dividual cephalothorax of 45 ú =
4
brown shrimp. Open circles rep- N
resent NRM of freshly killed ú=
specimens (except circles con- W
Z
taining an X, for which the ú J N=
shrimp had been dead at least 1 4 0
hr before measurement). Solid
:I
Do
circles represent IRM of speci-
mens which were magnetized in !a::
a ZOO-mT field prior to measure- ::J:
ment. Vertical lines connect tn
NRM and IRM measurements of 25 50 75
the same specimen. SHRIMP BODY LENGTH (mm)
was put through a metal grinder in the final stages of preparation. In addition, natural
sources of food can affect magnetization levels of specimens. In preliminary work, two
Palaemonetes grass shrimp held overnight in a plastic container fed on a dying brown
shrimp. Of the 22 grass shrimp we measured during the entire study, these two were the
only ones displaying any sIRM levels greater than measurement noise. Contamination can
be reduced further if freshly molted animals are measured, because Crustacea empty their
stomach and gut just prior to molting.
2.3. Measurement Results
2.3.1. Shrimp
Juvenile and subadult brown shrimp (Penaeus aztecus) collected by trawling in Gal-
veston Bay were placed in holding tanks for 1-28 hr before measuring. While initially we
used shrimp from bait camp holding tanks, most Penaeus prawns and all Palaemonetes
grass shrimp were caught in the Bay especially for this study and held in aerated plastic
containers. Some individuals had been reared from postlarvae in aquaria at the National
Marine Fisheries Laboratory at Galveston. Measures of magnetic intensity (both NRM and
sIRM) overlapped for individuals from different sources, so data are combined in this study.
For 50 Penaeus (45 freshly killed) the mean NRM level of the cephalothorax (Le., tail
removed) was 2 X 10- 10 A m 2 /cm 3 (range 0.08 to 9.5 X 10- 10 A m 2 /cm 3 ). sIRM levels
averaged 4 X 10- 9 A m 2 /cm 3 (range 0.018 to 3.6 X 10- 9 A m 2 /cm 3 ). As indicated in Fig.
2, there was no simple relationship of NRM or sIRM to body length of the Penaeus shrimp.
However, the freshly killed animals of body length greater than 80 mm (subadults) tended
to have NRM values closer to their sIRM values than did smaller individuals. Fourteen of
the fifty Penaeus cephalothorax specimens measured had NRM values at or below the level
of measurement noise (3 X 10- 11 A m 2 ); however, when those specimens were magnetized,
all but two showed sIRM values greater than 1 X 10- 10 A m2 • Freshly killed Penaeus differ
in NRM from shrimp that have died several hours earlier. NRM was greater than meas-
372 Chapter 17
HEAD THORAX ABDOMEN
MAGNETIZATION 180 90 6
(x 10-11Am 2 / cm 3 )
PERCENT OF
TOTAL SHRIMP 40%
REMANENCE
21% 39%
./
./
./
./
./ '"
I
Figure 3. Average distribution of magnetic remanence in brown shrimp. The magnetization is given
for each of the three major body sections.
urement noise in 67% of Penaeus specimens measured immediately after they were killed
by severing the cephalothorax from the tail. These values decreased to 21-56% of the
original intensity within 1 hr after death, and only 39% of the specimens that had been
dead at least 4 hr had significant NRM values.
The magnetic material was more concentrated in the shrimp cephalothorax but was
not localized within one particular structure. The cephalothorax accounted for over half
of the sIRM intensity in 17 of 22 individuals measured. Despite the larger volume of the
abdomen, it possessed an average of only 39% of the magnetic intensity. The small head
section (anterior to the cervical sulcus) on the average had intensities 30 times that of the
entire abdomen (Fig. 3). The diffuse nature of the magnetic material made it impractical
to make an extraction to determine positively by spectral analysis the chemical carrier of
the remanence.
The orientation of the NRM was not consistent from one cephalothorax specimen to
another, but some trends do appear in the data. Figure 4A (for NRM) and B (for magnetized
specimens) present equal-area projections of the magnetic declination (0° to 360° to clock-
wise, with the orientation of the head at 0°) and inclination (90° at center, 0° at periphery).
Only measurements in which specimens were oriented properly in the sample holder are
included in this analysis. There is a weak trend in the NRM orientation for the x-y vector
to point to the right (in 11 of 14 specimens). The orientation appears to be more nearly
vertical (all but one with inclination greater than 50°), and the z component tends to be
Crustacea 373
t::.
10
90
5• • 4
•
ú U =M ú N =
t::. .7 t::.06
t::.
Figure 4. Orientation of magnetization in brown shrimp (cir-

cles) and acorn barnacles (triangles). An equal-area projection
of the magnetic declination (0 to 360° clockwise, with the
0
orientation of the head at 0°) and inclination (90 at center,

0
0° at periphery, negative values presented as solid symbols)

indicates the shrimp and barnacle NRM (A) and shrimp IRM
(B). Specimens for which both NRM and IRM values were
measured and plotted are numbered individually for com-
parison of the two plots.
positive (upwards, open symbols, in 9 of 14 specimens). These specimens were magnetized

in fields of 200 to 600 mT with the body oriented so that the north pole corresponded to
the right side of the shrimp (90° in Fig. 4B). Most sIRM declination values (12 of 13) cluster
between 35° and 135°; however, the inclinations vary considerably. Ten specimens with
good measurement orientation in both NRM and sIRM determinations are numbered in-
dividually in Fig. 4A and B. Note that the magnetic orientation of freshly killed specimens
did not always move toward the direction of the applied field; e.g., #9 did not change
when magnetized. Specimens that had been dead 1 hr or longer (e.g., numbers 3, 4, and
7 in Fig. 4A and B) tended to track the direction of the applied magnetic field more closely
than did freshly killed shrimp.
When shrimp were exposed to stepped increases in magnetic field by using a calibrated
electromagnet, we found the pattern of acquired magnetization given in Fig. 5. Most spec-
imens reached sIRM values in fields of 150-200 mT, and magnetic intensity increased
only slightly in stronger fields. All magnetized specimens acquired at least half their sat-
uration magnetization intensity levels in fields of 100 mT.
Relatively small fields were required for demagnetization of shrimp. Four shrimp spec-
imens which displayed NRM values significantly above background noise levels were
demagnetized. Progressive AF demagnetization revealed that NRM values were greatly
reduced in 20-mT fields and appeared mostly demagnetized in 40-mT fields (Fig. 6, open
circles, and Fig. 9A, solid circles). On the other hand, shrimp magnetized in the lab required
greater field strengths for complete demagnetization. Demagnetization did not level off
until fields were greater than 60 mT (Fig. 6, solid circles, and Fig. 9A, solid circles).
374 Chapter 17
"'""
E
cr:
01
b...
>< 3.0
Figure 5. Acquisition of magnetization of brown shrimp

100 200 300 400 is given as a function of magnetic field to which the
MAGNETIC FIELD (mT) specimen was exposed.
ti"
E
<
ú=
b
Figure 6. Stepwise three-axis demagneti-

zation of brown shrimp. Open circles in-
dicate measurements that began with non-
magnetized specimens (NRM). and solid
circles indicate measurements of speci-
mens magnetized with an electromagnet
(200 mT) before stepwise demagnetization.
20 40 Horizontal axis is the peak field at each AF
MAGNETIC FIELD (mT) demagnetization step.
Crustacea 375
-
N
E
.
•
II
e(
"
0- 8
.J
0
.J
e(
:::l
C
• U Qú = Qú =
>
C -9
Z
"
Z
...
0
e(
00
...UJ
N
0
-10
Figure 7. NRM (open circles) and IRM Z ()
(solid circles. following magnetization in
a 500-mT field) of acorn barnacles. Ver-
tical lines connect NRM and IRM meas-
"2
e(
11 0.4 0.6 0.8 1.0 3.7

urements of the same specimen. BARNACLE MASS (g)
2.3.2. Barnacles
The barnacles (B. eburneus and B. amphitrite niveus) used in these experiments were
grown on glass microscope slides or strips of plastic held by an open plastic framework
suspended about 1 m below sea level on the seaward side of a breakwater in Galveston
Bay. The attached barnacles were brought to the lab immersed in seawater and were re-
moved from the glass or plastic bases using nonmagnetic tools. No consistent differences
in magnetization between the two barnacle species were detected. so measurement results
are combined in this report.
NRM measurements were taken on 12 individual intact live barnacles. Of these. 7 had
NRM values that exceeded the background noise determined at the time of each meas-
urement. The measured NRM values ranged from 5.15 to 18.0 x 10- 11 A m 2 /barnacle (Fig.
7. open circles) with an average of 11.7 x 10- 11 A m2 /barnacle. In one case the barnacle
could not be removed from the glass slide. and measurements were made with the barnacle
still attached to a piece of glass the size of its base (approximately l-cm diameter). NRM
measurements on another piece of glass from the same slide showed negligible magneti-
zation. The declination and inclination of the NRM magnetic moment of five barnacles
are plotted in Fig. 4A (open triangles indicate positive inclination values; solid triangles
indicate negative inclination values). In general. the magnetic moment vector is roughly
aligned with the barnacle body axis of bilateral symmetry.
Plots of the IRM values as a function of magnetizing field strength are shown in Fig.
8. Note that the IRM values are saturated for intensity values greater than 200 mT. In
addition to the acquired IRM measurements made on these three specimens. sIRM values
were determined for four other live barnacles using a magnetizing field of 500 mT. The
sIRM values measured for the seven barnacles (Fig. 7. solid circles) ranged from 3.75 to
17.7 x 10- 9 A m 2 /barnacle with a mean value of 8.58 x 10- 9 A m2 /barnacle. When
considered on a mass basis. the mean induced magnetization was 7.7 x 10- 9 A m2 /g. For
two of these seven barnacles. even though the NRM was indistinguishable from noise. the
sIRM values were 3.81 and 14.8 x 10- 9 A m2 /barnacle.
376 Chapter 17
....
N
E
ú NR=
b
...
><
z
o
j:
ú N M=
t-
W
Z
"
<C
ú =
W
..J 5
o
<C
Z
ex:
<C
!XI
Figure 8. Stepwise acquired magnetization of bar-
nacles is given for each level of magnetic intensity
200 400 of the electromagnet to which the specimen was
MAGNETIC FIELD (mT) exposed.
Results of a progressive AF demagnetization study of one barnacle are shown in Fig.

9B (solid circles). Prior to AF demagnetization, the intact barnacle had been magnetized
at increasing fields (see top curve in Fig. 8) up to a maximum field of 500 mT. As shown
in the demagnetization curve of Fig. 9B. the IRM value was reduced to 57% of the sIRM
value using a field of 20 mT and was still about 25% of the sIRM value in a demagnetization
field of 50 mT.
Measurements were taken of the soft internal tissue of two freshly killed barnacles,
and the NRM values were 2.5 x 10- 11 and 5.1 x 10- 11 A m 2 lbarnacle. For the second
barnacle, both the separated soft tissue and the shells were then magnetized in a field of
300 mT. The measured IRM values of the tissue and shells were 2.79 x 10- 9 and 17.3 x
10- 9 A m 2 , respectively. As the mass of the tissue was 0.035 g and that of the shells was
0.46 g, the magnetization was 7.9 x 10- 8 and 3.76 x 10- 8 A m 2 /g for the tissue and shells,
respectively, of one individual barnacle. Although no conclusions can be drawn from a
single specimen, these measurements appear to indicate that the magnetic material was
not highly localized in this individual barnacle. While there was six times as much mag-
netizable material in the shells than in the soft tissue, the magnetization was twice as high
in the tissue as in the shells.
2.3.3. Other Measurements

Palaemonetes pugio is a small grass shrimp (Crustacea: Caridea) that spends its entire
life cycle in estuaries or salt marshes and shows no long-distance movement. The species
feeds primarily on seagrass epiphytes or as a scavenger and tolerates a wide range of water
salinities (Morgan, 1980). Specimens were collected from the same habitat and at the same
time as the juvenile Penaeus aztecus. Fifteen washed specimens, ranging in body length
Crustacea 377
A. SHRIMP
50
ú =
ú = 25
en
LL
o
ú =
Z
11.1 B. BARNACLE
ú = 75
11.1
a..
50
Figure 9. A comparison of normalized 25

curves for stepwise acquired magnetiza-
tion (open circles) and AF demagnetiza-
tion (solid circles) for (A) brown shrimp 20 40 60
and (B) acorn barnacles. MAGNETIC FIELD (mT)
from 21 to 31 mm, were measured in the magnetometer when freshly killed but intact.
Magnetic remanence was not significantly greater than background noise for either the
NRM (range of 0.4 to 2.0 X 10- 11 A m2 l or the sIRM (range of 1.1 to 6.0 X 10- 11 A m 2 l
for these specimens.
Two seawater samples were tested for levels of induced magnetization. The first, ap-
proximately 1 cm 3 of artificial seawater, was frozen in a magnetic field of about 100 mT.
Its measured magnetization was 1.5 x 10- 10 A m 2 • The NRM of a second water sample,
taken from Galveston Bay, was 3.1 X 10- 11 A m2 • When magnetized at 400 mT, its IRM
was 4.0 X 10- 11 A m 2 . Thus, both samples of magnetized frozen seawater showed an
induced remanence slightly greater than measurement noise. However, the levels were
lower than NRM values for most specimens of barnacles and brown shrimp measured.
3. Discussion
3.1. Magnetic Remanence in Shrimp and Barnacles
Brown shrimp and acorn barnacles show significant levels of induced magnetization.
Many individual specimens of freshly killed shrimp and barnacles show NRM as well.
The level of magnetization cannot be attributed to the seawater in which these animals
live. Measurement results of washed and dissected specimens indicate that external or gut
contamination was also not significant. In addition, magnetized specimens of grass shrimp
from the same habitat show nonsignificant levels of magnetization. Therefore, we conclude
that these Penaeus shrimp and Balanus barnacles possess naturally occurring magnetic
material in their bodies.
In the shrimp, the magnetic material is clearly more concentrated in the cephalothorax,
but it is not localized in any specific structure. For example, although the magnetic in-
378 Chapter 17
tensity is highest in the head region, there is diffuse magnetic material in the thorax muscle
tissue as well. The orientation of the magnetic moment is not consistent from one ce-
phalothorax specimen to another, but the data suggest a vertical orientation. For the bar-
nacles measured, the orientation of the magnetization is roughly aligned with the axis of
bilateral symmetry and is generally in a plane parallel to its flat base. The magnetic material
was present in both the shells and the soft tissue of the barnacle specimen analyzed, but
the details of its distribution in the body remain unresolved.
The magnetic properties measured in these Crustacea strongly suggest that the re-
manence carrier is magnetite even though definitive chemical tests have not been made.
For both groups the curve for acquired magnetic remanence begins to level off near 200
mT, and IRM is near saturation before 300 mT. In AF demagnetization, field strengths of
20 mT are sufficient to reduce the NRM values for shrimp to one-quarter of original values,
while the levels for magnetized shrimp are reduced to one-half. Demagnetization of mag-
netized specimens of both shrimp and barnacles is generally compl-ete with fields of 60
mT. These general indicators of soft magnetization agree with those measured for magnetite
(McElhinny, 1973) and make it unlikely that the samples have been contaminated with
hematite or other iron compounds.
The domain state of the magnetic material in shrimp and barnacles has not been es-
tablished, but evidence suggests that the magnetic units are not simple isolated single
domains. A comparison of progressive acquisition and AF demagnetization of the IRM
indicates that the two curves cross at field strengths of about 35 mT for the shrimp (Fig.
9A) and at 33 mT for the one barnacle measured (Fig. 9B). At this intersection point, the
ratio of demagnetized remanence to saturation remanence is about 0.30 for shrimp and
about 0.38 for the barnacle. This point of intersection corresponds to the remanent coercive
field, and the degree of symmetry between the two curves is related to the strength of the
interaction between the magnetic domains of the material (Cisowski, 1981). The coercive
field is thus estimated to be about 35 mT for shrimp and 33 mT for the barnacle measured,
while the AF required to reduce the remanence to half is about 20 mT for shrimp and 33
mT for the barnacle. At the coercive field, the remanence ratio values (about 0.30 for
shrimp, 0.38 for the barnacle) are clearly less than the ratio of 0.5 expected for noninter-
acting single-domain particles of magnetite. However, they are near the value of 0.27 for
chiton teeth (strongly interacting single-domain particles) and 0.30 determined for small-
grain, single-domain synthetic magnetite (Cisowski, 1981). The low resistance to demag-
netization in the shrimp measured suggests that the magnetic grains of the remanence
carrier are influenced by interacting fields. On the other hand, material with multi domain
grains is comparatively easier to magnetize and would show a steeper acquisition curve
in fields below 50 mT than do the curves for shrimp in Fig. 5. The coercive field ratio is
slightly larger for barnacles than for shrimp, suggesting there may be somewhat less in-
terdomain interaction for the barnacle magnetic material than for that in the shrimp.
3.2. Relation to Studies of Other Arthropods
The level of induced magnetization in the Crustacea measured is of the same order
of magnitude as that reported for honeybees (Gould et al., 1978) and monarch butterflies
Ganes and MacFadden, 1982), i.e., 10- 9 A m 2 ( = 10- 6 emu). Ratios of IRM signal to
background noise were greater than 12 for all magnetized specimens and in most cases
were over 50. As in the honeybees, NRM values for shrimp were not significant in dead
specimens but were high in many of the freshly killed specimens. In the case of both shrimp
and honeybees, the orientation of magnetization in dead specimens tracked the applied
field rather closely, while this was less so in freshly killed specimens.
Crustacea 379
The location of most of the magnetic material is in the head-thorax region in both
shrimp and the monarch butterfly, while in honeybees it is concentrated at the anterior
abdomen. The magnetic material appears to be more diffuse in shrimp and is clearly not
restricted to a distinct strudure. In the one barnuclQ IDQ::lwn!d. both the ú Ü É ä ä = and the ú n Ñí =
tissue contributed significantly to the magnetic remanence.
The only other crustacean which is known to possess magnetic material is the spiny
lobster Panulirus argus (Decapoda, Palinuridae). NRM values in the range of 10- 10 to 10- 8
A m 2 (10- 7 to 10- 5 emu) are found in regions of the posterior cephalothorax, the area of
highest concentration of magnetic material (Lohmann, 1984). Thus, the crustacean species
in which significant magnetic remanence has been documented appear not to have mag-
netic material concentrated in a particular organ, in contrast to honeybees. The cephalo-
thorax of these Crustacea, however, possesses the greatest concentration of magnetic ma-
terial.
3.3. Significance of This Study and Suggestions for Further Work
This study has documented the presence of magnetic material in several marine Crus-
tacea (Penaeus aztecus, B. amphitrite niveus, B. eburneus) and its absence from others
(Pa1aemonetes pugio) from the same habitat. These results plus current work on spiny
lobsters (Lohmann, 1984) indicate that certain Crustacea have magnetic remanence of a
similar nature to that found in honeybees (Gould et a1., 1978) and butterflies (Jones and
MacFadden, 1982).
The remanence carrier of crustacean ferrimagnetism appears to be magnetite, as de-
termined by magnetization patterns. Chemical tests are needed, however, to confirm the
presence of magnetite and its domain states in these animals. Because of the somewhat
diffuse arrangement of the magnetic material in the shrimp and barnacles, initial work on
extraction and purification techniques is necessary.
The magnetic material in the Crustacea could contribute significantly to the magnetic
remanence of marine sediments. Lowenstam (1974) and Kirschvink and Lowenstam (1979)
have calculated that magnetite synthesized by chitons could account for a detrital remanent
magnetization intensity near 0.1 nT (10- 6 G). While the magnetic intensity measured in
tissue of these Crustacea appears to be somewhat lower than that for chiton teeth, the
shrimp and barnacles could still make a notable contribution. Penaeid shrimp are common
in shallow marine habitats. For instance, the estimated commercial shrimp catch for a
single year (for 1968: Diener, 1975) was over 5000 kg/km 2 for waters shallower than 100
m along the Texas coast. When converted to individuals of the size measured in this study,
the year's catch would average nearly one magnetic shrimp per two square meters of coastal
seas. Barnacles, which abound in mangrove and rocky intertidal zones, also have a sig-
nificant remanence and could contribute magnetite to coastal sediments. Biogenic material
is one major source of magnetite in oceanic sediments (Lowenstam, 1974; Henshaw and
Merrill, 1980); however, the depth and chemical composition of the burial environment
are important determinants of the amount of unmodified magnetite actually deposited.
The presence of ferrimagnetic material occurring naturally in crustaceans could pos-
sibly be significant in at least three aspects of the life history of these organisms: (1) settling
and growth rates of sedentary forms, (2) migration or movement orientation in nonmigra-
tory forms, and (3) biological rhythms. However, as detailed below, the effect of possessing
magnetic material has not been investigated for any of these aspects in shrimp and bar-
nacles. In addition, no research has suggested that the density, hardness, or electrical
properties of magnetite might convey a selective advantage to crustaceans that possess it,
as it has in chitons (Kirschvink, 1982b).
380 Chapter 17
Laboratory studies (e.g., Taneyeva, 1978; Lee and Weis, 1980) and preliminary field
experiments (O'Brien, 1985) indicate that in a variety of Crustacea, growth rates are en-
hanced in the presence of an artificial magnetic field. The mechanisms for this phenomenon
are still unknown, but strengthened ionic gradients might be involved. For example, sec-
ondary effects from solar flares influenced the productivity of offspring of fruit flies (Dro-
sophila melanogaster) grown in artificial magnetic fields of sufficient strength (35-100
mT) to alter air ionization (Levengood and Shinkle, 1962). The presence of magnetite, per
se, has not been shown to affect growth rates. It can be speculated, however, that the small
electrical currents induced in biological tissues bearing magnetite might have a favorable
effect on growth or metabolism, perhaps by affecting ionic exchange across membranes or
altering the ion specificity of receptor sites. Such an effect might be tested most precisely
in rapidly growing eggs or larvae.
Migration movements of Penaeid shrimp have been documented by marine tagging
studies, but there has been no controlled laboratory work on orientation response of shrimp
to magnetic fields. Although such an investigation might be profitable for understanding
movements of brown shrimp, it remains unclear how magnetite that is not well localized
in the shrimp body might serve as a sensory frame of reference for orientation behavior.
In many marine Crustacea, possession of a daily time reference may convey a selective
advantage. Sedentary animals would have benefits in terms of the timing of protective
color change rhythms (fiddler crabs: Palmer, 1974) or in terms of successful foraging while
avoiding desiccation (barnacles). In mobile animals, the benefits of coordinating loco-
motion with tidal cycles (estuarine crab larvae: Cronin and Forward, 1979; beach amphi-
pods: Enright, 1978) or of timely burial in sediments (shrimp: Hughes, 1968) also appear
important. The daily fluctuations of intensity in the earth's magnetic field (Kalmijn, 1974)
could possibly provide information on time of day or serve as one of several alternate cues
for periodic behavior. Such a behavioral response to daily changes in the earth's magnetic
field has been suggested from data on honeybee foraging behavior (Lindauer and Martin,
1972; Gould, 1980). Previous studies on circadian rhythms in Crustacea have not directly
addressed this possibility.
4. Summary
The magnetization of shrimp (Penaeus aztec us) and barnacles (Balanus species) was
measured with a cryogenic magnetometer. The mean level of NRM was 2.0 x 10- 10 A
m 2 for 45 living or freshly killed shrimp. For 7 living barnacles in their shells, the NRM
exceeded background noise and averaged 1.2 x 10- 10 A m 2 • Samples were magnetically
saturated in fields of about 300 mT, and sIRM levels averaged 4.2 x 10- 9 A m 2 for the
shrimp and 8.6 x 10- 9 A m2 for the barnacles. Acquisition of sIRM and response to AF
demagnetization suggest that the observed magnetic material is magnetite. The magnetic
material in shrimp is not localized in any particular structure but appears to be more
concentrated in the cephalothorax. Samples of seawater as well as water in which some
laboratory shrimp were reared showed little significant magnetization. In addition, meas-
ures of NRM as well as magnetized specimens of nonmigratory grass shrimp from the same
habitat as the migratory brown shrimp showed magnetization levels no greater than the
level of background noise (about 10- 11 A m 2 ).
ACKNOWLEDGMENTS. Measurements were made in the University of Texas Paleomagnetism

Laboratory, and we greatly appreciate the aid and technical advice of Wulf Gose and Mar-
garet Testarmata. Specimens were obtained through the facilities of the National Marine
Fisheries Laboratory, Marine Biomedical Institute, and the U.S. Coast Guard, all in Gal-
Crustacea 381
veston, Texas, and in particular we thank Ray Hixon, James Lyon, James McVey, Elaine
O'Brien, and Zula Zein-Eldin for their help. The perspective and encouragement provided
by Gary Scott, Joe Kirschvink, and Cliff Frohlich in the early stages of the study were also
critical. We appreciate helpful comments on the manuscript from M. C. Arendse and Ken-
neth Lohmann. This is Contribution No. 564 from the University of Texas Institute for
Geophysics.
References
Aldrich, D. V., Wood, C. K, and Baxter, K. N., 1968, An ecological interpretation of low temperature
responses in Penaeus aztec us and P. setiferus postlarvae, Bull. Mar. Sci. 18:61-71.
Arendse, M. c., 1978, Magnetic field detection is distinct from light detection in the invertebrates
Tenebrio and Talitrus, Nature 274:358-362.
Arendse, M. C., and Barendregt, A., 1981, Magnetic orientation in Orchestia, Physiol. Entomol. 6:333-
342.
Arendse, M. c., and Kruyswijk, C. J., 1981, Orientation of Talitrus saltator to magnetic fields, Neth.
J. Sea Res. 15:23-32.
Barnwell, F. H., 1966, Daily and tidal patterns of activity in individual fiddler crabs (genus Vca) from
the Woods Hole region, BioI. Bull. 130:1-17.
Becka, A. M., Castelli, V. J., and Fischer, K c., 1981, Bibliography on fouling, biodeterioration and
their control, David W. Taylor Naval Ship Research and Development Center, Report SME-81/43,
Bethesda.
Blakemore, R P., and Frankel, R B., 1981, Magnetic navigation in bacteria, Sci. Am. 245:58-65.
Brown, F. A., Jr., Hastings, J. W., and Palmer, J. D., 1970, The Biological Clock: Two Views, Academic
Press, New York.
Cronin, T. W., and Forward, R B., Jr., 1979, Tidal vertical migration: An endogenous rhythm in
estuarine crab larvae, Science 205:1020-1022.
Diener, R A., 1975, Cooperative Gulf of Mexico estuarine inventory and study-Texas: area descrip-
tion, NOAA Technical Report NMFS CIRC-393, Seattle.
Enright, J. T., 1978, Migration and homing of marine invertebrates: A potpourri of strategies, in: Animal
Migration, Navigation, and Homing (K. Schmidt-Koenig and W. T. Keeton, eds.), Springer-Verlag,
Fry, B., 1981, Natural stable carbon isotope tag trace Texas, USA, shrimp (Penaeus aztecus) migrations,
V S Natl. Mar. Fish. Servo Fish. Bull. 79:337-346.
Goree, W. S., and Fuller, M., 1976, Magnetometers using RF-driven Squids and their applications in
Gould, J. L., 1980, The case for magnetic sensitivity in birds and bees (such as it is), Am. Sci. 68:256-
267.
201:1026-1028.
Grady, J. R, 1971, The distribution of sediment properties and shrimp catch on two shrimping grounds
on the continental shelf of the Gulf of Mexico, Proc. Gulf Caribb. Fish. Inst., 23rd Annu. Sess.
pp. 139-148.
Henshaw, P. C., Jr., and Merrill, R T., 1980, Magnetic and chemical changes in marine sediments,
Rev. Geophys. Space Phys. 18:483-504.
Herrnkind, W., 1969, Queuing behavior of spiny lobsters, Science 164:1425-1427.
Herrnkind, W., and Kanciruk, P., 1978, Mass migration of spiny lobster Panulirus argus (Crustacea:
Panuliridae): Synopsis and orientation, in: Animal Migration, Navigation. and Homing (K.
Schmidt-Koenig and W. T. Keeton, eds.), Springer-Verlag, Berlin, pp. 430-439.
Hughes, D. A., 1968, Factors controlling emergence of pink shrimp (Penaeus duorarum) from the
substrate, Bioi. Bull. 134:48-59.
Hughes, D. A., 1972, On the endogenous control of tide-associated displacements of pink shrimp
Penaeus duorarum, Bioi. Bull. 142:271-280.
382 Chapter 17
electric organs, in: Handbook of Sensory Physiology, Volume III/3 (A. Fessard, ed.), Springer-
Kirschvink, J. L., 1982a, Paleomagnetic evidence for fossil biogenic magnetite in western Crete, Earth
Kirschvink, J. L., 1982b, Birds, bees and magnetism, Trends Neurosci. 5:160-167.
disciplinary Approach (S. Williamson, ed.), Plenum Press, New York, pp. 501-532.
Lett. 44:193-204.
Klima, E. F., 1963, Mark-recapture experiments with brown and white shrimp in the northern Gulf
of Mexico, Proc. Gulf Caribb. Fish. Inst., 16th Sess. pp. 52-64.
Lee, P. H., and Weis, J. S., 1980, Effects of magnetic fields on regeneration in fiddler crabs, BioI. Bull.
159:681-691.
Levengood, W. C., and Shinkle, M. P., 1962, Solar flare effects on living organisms confined in magnetic
fields, Nature 195:967-970.
Navigation (S. R. Galler, K. Schmidt-Koenig, G. J. Jacobs, and R. E. Belleville, eds.), NASA SP-
262, U.S. Government Printing Office, Washington, D.C., pp. 559-567.
Lohmann, K. J., 1984, Magnetic remanence in the western Atlantic spiny lobster Panulirus argus, J.
Exp. BioI. 113:29-41.
McElhinny, M. W., 1973, Palaeomagnetism and Plate Tectonics, Cambridge University Press, London.
Morgan, M. D., 1980, Grazing and predation of the grass shrimp Palaemonetes pugio, Limnol. Oceanogr.
25:896-902.
North Carolina Department of Conservation and Development, 1967, Migration and growth of com-
mercial Penaeid shrimps in North Carolina, Special Science Report No. 11, pp. 1-29.
O'Brien, W. P., Jr., 1985, Effects of magnetic fields and fluid velocity on the growth of marine bio-
fouling organisms, in: A Review of the Applications of Applied Fields for Energy Conservation,
Water Treatment and Industrial Applications (R. Reimers, L. White and S. Bock, eds.) United
States Department of Energy, Washington D.C., in press.
Ozeki, M., and Osada, S., 1977, Upside-down orientation in the crayfish with ferrite statoliths during
magnetic field stimulation, Annot. Zoo1. Jpn. 50:220-230.
Palmer, J. D., 1974, Biological Clocks in Marine Organisms: The Control of Physiological and Be-
havioral Tidal Rhythms, Wiley, New York.
Perry, A., Bauer, G. B., and Dizon, A. E., 1981, Magnetite in the green turtle, Eos, Trans. Am. Geophys.
Union 62:850-851.
Quinn, T. P., Merrill, R. T., and Brannon, E. T., 1981, Magnetic field detection in sockeye salmon, J.
Exp. Zoo1. 217:137-142.
Taneyeva, A.I., 1978, Changes in the physiological processes of aquatic organisms under the influence
of a constant magnetic field, Hydrobiol. J. 14(5):48-54.
Taneyeva, A. I., and Dolgopol'skaya, M. A., 1974, The biological effect of a constant magnetic field
on Artemia salina, Hydrobiol. J. 10(4):47-52.
Trent, L., 1966, Size of brown shrimp and time of emigration from the Galveston Bay system, Texas,
Proc. Gulf Caribb. Fish. Inst., 19th Sess. pp. 7-16.
Vasil'yev, A. S., Bednarskiy, A. D., Vasil'yeva, L. A., and Chubur, V. P., 1974, The reproduction and
development of Daphnia magna in a magnetic field, Hydrobiol. J. 10(2):54-57.
Walcott, C., Gould, J. L., and Kirschvink, J. L., 1979, Pigeons have magnets, Science 205:1027-1029.
Walker, M. M., and Dizon, A. E., 1981, Identification of magnetite in tuna, Eos, Trans. Am. Geophys.
Union 62:850.
Crustacea 383
Walton, A. S., and Herrnkind, W. F., 1977, Hydrodynamic orientation of spiny lobster Panulirus argus
(Crustacea: Palinuridae): Wave surge and unidirectional currents, in: Proceedings of the Annual
Northeastern Meeting of the Animal Behavior Society, Memorial University of Newfoundland,
Mar. Sci. Res. LHb" TflGhniGHI Rilport ú Má N ú N J ú N N K =
Webb, H. M., and Brown, F. A., Jr., 1965. Interactions of diurnal and tidal rhythms in the fiddler crab
Uca pugnax, BioI. Bull. 129:582-591.
Zoeger, J., Dunn, J. R., and Fuller, M., 1981, Magnetic material in the head of a common Pacific dolphin,
Science 213:892-894.
Chapter 18
Magnetic Field Sensitivity in
Honeybees
WILLIAM F. TOWNE and JAMES L. GOULD
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
2. Magnetic Fields Cause Misdirection in the Waggle Dance . . . . . . . . . . . . . . . . 386
2.1. The Waggle Dance and "Residual Misdirection" . . . . . . . . . . . . . 386
2.2. No "Residual Misdirection" in Dances in Null Magnetic Fields .. . 387
2.3. No Misdirection in Dances Oriented along Magnetic Field Lines .. . 388
2.4. A Mathematical Description of Misdirection . . . . . . . . 389
3. Magnetically Oriented Horizontal Dances . . . . . . . . . . . . . . . . . . . . 392
4. Magnetic Orientation of Comb-Building . . . . . . . . . . . . . . . . . . . . . 393
5. Magnetic Fields and Orientation in Time . . . . . . . . . . . . . . . . . . . . . . 395
5.1. The Time Sense of Bees. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . 395
5.2. The Magnetic Fields and the Time Sense . . . . . . . . . 397
6. The Magnetic Receptor System . . . . . . . . . . . . . . . . . . . . . . . . 398
6.1. The Implied Sensitivity of the Detector . . . . . . . . . . . . . . . . . . 398
6.2. The Nature of the Detector. . . . . ... . .............. . 399
6.3. The Bees' Magnets . . . . . . . . . . . . . . . 401
6.4. Anatomy . . . . . . . . . . . . . . . . . . . . . 403
7. Summary and Conclusions . . . . . . . . . . . . . 403
References. . . . . . . . . . . . . . . . . . . . . . . 404
1. Introduction
In the past 15 years an abundance of evidence has accumulated suggesting that the be-
havior-particularly the orientation-of a variety of organisms is affected by the earth's
magnetic field. The orientation and navigation of pigeons and honeybees have been un-
usually well studied (see reviews by von Frisch, 1967; Keeton, 1974; Gould, 1982), and
of all terrestrial animals, it is in these two that the effects of magnetic fields on orientation
have been most clearly and abundantly documented. Gould (Chapter 12, this volume)
reviews the pigeon findings. Here we review the evidence for an effect of the earth's mag-
netic field on honeybee orientation. Where we see the evidence as weak, we say so and
suggest experiments that should help to show whether or not the effects are real. We then
discuss the nature and sensitivity of the receptor mechanism as inferred from behavioral
observations and anatomical studies. Again, we suggest experiments which should help
us to locate and understand the magnetic field detector.
WILLIAM F. TOWNE and JAMES 1. GOULD • Department of Biology, Princeton University, Prince-
ton, New Jersey 08544.
385
386 Chapter 18
Time of day
'.
.....
D N MD ú J Q J J J H J J Q J J J H J J J ú J Q J J J H J J Q J J J H J J J ê J ú J J J H J J J ê J J ú ú ú g ú D í J ú ê J ú =
c:
0
....U .5' ." '
I' ,
'
...
Q)
0 ã J J í J J Hú Z HWJ., J •J í J J WJ HJ K••ú ’ =! .I,.;.,.'J í I KJ HJ J HJ J J J > J J HJ J HKKKKKKKI WXKú
rl'.I ú= =..ú WK';'
"'C
.!!! . . '. .' .'. . . IJ •
W=
:?!
-5'
.1 • I. ú K = ,, t •
-10'
. I" •
.' , •
. .. :,
I , . I, •
' . I .. ., • ,
-15' L - - - - 1 - - -4"'=0-=-·ú D J J J J Z Z S MWJ WJ D J J J D J J ú U MWWJ ’ J J J D J J J WWN MMZ J ’ -= -'----:12:::0-=-·J J J D D J J J N WT QMWWJ ’ J J J D J J J J ú N S Z J M’ =-- --,ao----
Dance angle
Figure 1. Typical misdirection curve. Abscissa: predicted dance angle (measured clockwise from up-
ward vertical) assuming the bees make errorless sun/gravity transpositions; the predicted angle changes
over the course of the day as the sun moves across the sky. Ordinate: misdirection, i.e., the deviations
of the measured dance angles from their predicted values; +, clockwise; -, counterclockwise. Each
dot represents the average of 10 waggle runs. The comb was aligned north/south, and the dance floor
faced east. The arrow indicates a predicted zero-crossing (see text). From Lindauer and Martin (1968).
2. Magnetic Fields Cause Misdirection in the Waggle Dance
2.1. The Waggle Dance and "Residual Misdirection"
When a forager honeybee returns from a successful foraging trip, she performs a "wag-
gle dance" -a ritualized pantomime of the outward flight to the food source-while other
bees, potential recruits, follow attentively. In her pantomime the dancer waggles her body
from side to side while walking forward along a line, the direction of which corresponds
to the direction of the food. The dance is normally performed on a vertical sheet of comb
in the darkness of the hive, and gravity is used as the reference: "up" during the dance
refers to the direction of the sun during the outward flight. Hence, a dance 90° to the right
of vertical corresponds to an outward flight bearing 90° to the right of the sun. The duration
of waggling is proportional to the distance to the food. Each miniaturized version of the
outward flight is called a "waggle run," and a bee usually performs at least several waggle
runs during a single bout of dancing.
Although the dance does indeed convey fairly accurate information to recruit bees
(von Frisch, 1967; Gould, 1975), there are almost always small, systematic errors-up to
20° at times-in the direction indications given by dancers. These errors are not just "noise"
in the system, however, in that all of the bees dancing at any given time make the same
error in both magnitude and direction. A typical example of the course of the misdirection
over time throughout a single day is shown in Fig. 1. The direction of the dance with
respect to gravity changes over the course of the day as the sun moves across the sky
altering the relative positions of the sun and food. Note how the misdirection changes
irregularly as the dance angle changes. von Frisch and colleagues (see von Frisch, 1967,
pp. 196 ff.) have abundantly documented this phenomenon and named it "residual mis-
direction."
Magnetic Field Sensitivity in Honeybees 387
Time of day
-
'10'
I J J ú ú J
J Nú =
........ .... .
... . -..... ..
c '5'
.\
...u
.2 '
... ., ú ú =r-r.-'.ILU./• ....K WJ ú-K =,."11
..
CI)
o· I .::'!
Wú =ú ? ê ç D =
... ú Kr f =
T .'
-:. ".
':-
"0
.!!!
::?i
-5 . - - - - - !------- - - - -
-10'
140' 160' 180' 200 . 220 . 240 . 260 .
Dance angle
Figure 2. No misdirection in a null magnetic field. Details as in Fig. 1 except that the earth's field
was compensated to within 5% (2500 'Y) just as the measurements began. From Lindauer and Martin
(1968).
2.2. No "Residual Misdirection" in Dances in Null Magnetic Fields
In an initial attempt to explain how the misdirection arises, von Frisch and Lindauer
(1961) focused on an apparent relationship between the misdirection and the direction of
the waggle run with respect to gravity and hypothesized that the misdirection was due to
peculiarity of the gravity detector system of bees. A similar relationship has been observed
in ants in experiments in which the ants were required to navigate to a goal on a vertical
surface (Markl, 1964), suggesting that any "nonlinearity" in the operation of the gravity
organ might not be peculiar to bees. The "gravity organ" hypothesis was further supported
by the observation that there is no misdirection in dances performed on horizontal surfaces
and referenced to celestial cues. (Such horizontal dances are used frequently in experi-
ments, but they also occur naturally on busy summer days when bees occasionally dance
just outside the nest entrance.) The hypothesis as it stood could not account for all of the
data, however, especially the data indicating that the courses of the daily misdirection
curves generally are not reproducible from day to day, even with the same bees foraging
from the same feeder (Lindauer and Martin, 1968; Martin and Lindauer, 1977).
Lindauer and Martin (1968), in an attempt to see if the misdirection is in any way
dependent upon the earth's magnetic field, performed a series of experiments in which
they compensated for the earth's field with a set of Helmholtz coils to produce an ap-
proximate null field. Although this same experiment has been performed earlier by von
Frisch (1967, p. 460) with negative results, Lindauer and Martin found that 45 min after
the coils were turned on, the misdirection vanished entirely (Fig. 2). Thus, misdirection
seems to arise in some way as a result of the earth's field. The 45-min time lag seems at
first curious, but in fact it poses no major problem; such time lags are not unusual in the
responses of bees to changing stimuli (e.g., Lindauer, 1963; Gould, 1980a, 1984; Hepworth
et al., 1980) and may be a manifestation of some higher-level processing strategy-a "run-
ning-average" strategy, for instance-which could serve as a built-in stimulus-noise filter
(Gould, 1984). The lag noticed by Hepworth et al. (1980) is particularly noteworthy here
as it involved a 40- to 60-min delay in the responses of their bee's overall activity levels
to (intermittent) changes in the magnetic field strength. In any case, Lindauer and Martin's
result is clear and repeatable and provides some of the best evidence to date that honeybees
can sense the earth's magnetic field. The magnitude of the effect in the field-compensatiOll
experiments depends upon the degree to which the earth's field is compensated. As Fig.
388 Chapter 18
Figure 3. The earth's field (in Germany) as expe-

a East-facing
comb
-N rienced by bees dancing on east- west-, and north-
facing dance floors. In (a), the comb is aligned
north/south and is viewed from the east. In this
case, the earth's field vector runs parallel to the
M comb's surface, and waggle runs oriented toward
either 155 or 335 0 (measured clockwise from the
upward vertical) coincide exactly with the field
b West-facing
comb -5
lines. In (b), a north/south comb is viewed from the
west. If bees dance on this west-facing surface, wag-
gle runs oriented toward either 25 or 205 0 will co-
incide with the field lines. In (c), the comb is
aligned east/west, and the projection of the field
vector onto the comb is a vertical line. In this case,
W waggle runs oriented toward either 0 or 1800 will
C North-facing LN coincide with the field vector's projection irres-

pective of the side of the comb on which the dances
comb take place. For all three comb alignments, Lindauer
and Martin (1972) report that dances oriented along
the field lines (or their projections onto the comb)
contain no misdirection.
2 shows, the misdirection is completely eliminated when the field is compensated to within
5% (Le., within 2500 "y of a null field). In artificially high fields-up to 13 times normal-
the magnitude of the misdirection does not change appreciably, but the angular dispersion
of the dances increases greatly (Lindauer and Martin, 1968). Given these results, it would
seem unreasonable to question the existence of some sort of influence of magnetic fields
on the dance orientation, but why and how these effects occur and why bees would wish
to use the earth's field to insert calibrated errors into their dances are as yet quite unclear.
2.3. No Misdirection in Dances Oriented along Magnetic Field Lines

Lindauer and Martin (1972) examined closely a large number of daily misdirection
curves in light of the information they had on the direction of the magnetic field lines at
the time of the dance recordings. They noticed that when the bees danced on the east-
facing side of a sheet of comb aligned north/south, there was almost always no misdi-
rection-the misdirection curves passed through zero-when the waggle runs were ori-
ented toward either 155 or 335° measured clockwise from the upward vertical (see, e.g.,
the arrow in Fig. 1). In Germany, where the dance recordings were made, the earth's mag-
netic field vector points north and is inclined downward at an angle of 65° from the hor-
izontal, or 155° from the upward vertical (see Fig. 3a). Because the magnetic field vector
runs parallel to the surface of a north/south-aligned sheet of comb, the waggle directions
of 155 and 335° coincide exactly with the field lines (Fig. 3a). In dances on the west-facing
side of a north/south comb, Lindauer and Martin never found misdirection for dance di-
rections of 25 and 205°, again the angles for which waggle runs were aligned exactly along
the field lines (Fig. 3b). When a comb is aligned east/west, the field vector does not run
parallel to its surface, and the projection of the field vector onto the comb is a vertical
vector pointed down (Fig. 3c). Lindauer and Martin (1972) found, as one might expect,
that the misdirection curves for dances recorded on such east/west-running combs always
passed through zero at dance angles of 0 and 180°, regardless of the side of the comb on
which the dances took place (see, e.g., the arrow in Fig. 4). (This latter effect was the source
of the apparent but weak relationship between the dance angle relative to gravity and
misdirection that originally led von Frisch and Lindauer to suspect that misdirection was
due entirely to a peculiarity of the gravity orientation system.) Lindauer and Martin report
that these rules hold for all but one of the 62 misdirection curves that were available to
them at the time, and they have published several examples such as those shown in Figs.
1 and 4 (Martin and Lindauer, 1977). The bulk of the data has yet to be published, however.
In summary, collapsing the above three rules into one, it seems that there is no mis-
direction in dances oriented along the projection of the magnetic field lines onto the dance
floor, irrespective of the polarity of the dance or, alternatively, of the magnetic field. One
other kind of evidence that supports the rule comes from dances that Martin and Lindauer
(1977) recorded in Morocco and in the Canary Islands where the magnetic field vectors
point down with less of an incline than in Germany; 49 and 42° from the horizontal,
respectively. The misdirection curves obtained at both locations always pass through zero
at the predicted dance angles for both north/south and east/west combs. In fact, at both
places Lindauer and Martin were initially given erroneous values for the local magnetic
inclinations, and the dance recordings later led them to investigate and discover that the
values they were working with were indeed wrong. Although it seems clear that Lindauer
and Martin's "zero-crossing rule" reflects a real phenomenon, we should point out that
there are usually zero-crossings in the misdirection curves that are not predicted by the
above rules (see Figs. 1 and 4), and it still seems possible, albeit remotely given the pre-
cision with which the curves generally seem to cross zero at the predicted places, that the
rule would not stand up to a statistical analysis of the data.
2.4. A Mathematical Description of Misdirection
Although the zero-crossings of the misdirection curves seem to be highly reproducible

from day to day for dance floors in given locations and orientations, the courses of the
rest of the curves, as mentioned earlier, generally are not. With their stockpile of misdi-
rection curves before them, Lindauer and Martin (1972), Martin and Lindauer (1973,1977)'
and Lindauer (1977) set out to construct a mathematical formula to describe the curves in
hopes of isolating the important variables other than the dance angle. After going through
four preliminary versions that seemd to fit some of the misdirection curves fairly well,
they arrived at a complicated formula that seems to fit a wide range of these bumpy curves
for various positions of the dance floor and various directions to the food. The formula is
where Mi = misdirection at time 2; eA l ,2 = ä ç Ö E ú c N I O = + 1) x I.L x sin am x aa m ; ú c ä I O =

= rate of change of field intensity at times 1 and 2, respectively; am = dance angle relative
to the projection of the magnetic field lines onto the dance floor at time 2; aa m = change
in am from time 1 to time 2; v = the Raleigh constant (= 0.2 or 0.4); I.L = magnetic
permeability of a bee = 1.0; A = scaling factor. [In Lindauer's 1977 version of the equation,
aa m is missing entirely, the quantity (eA2 - eA l ) is not squared, and the first term, eA 2,
is not multiplied by A.]
Figure 4a shows one of the most dramatic of several curves fitted to data by Martin
and Lindauer (1977). According to the formula, the two crucial parameters determining
the misdirection are (1) the angle, am, between the projection of the magnetic field lines
onto the dance floor and the line of waggling-as one might have expected-and (2) the
time course of variations in the total intensity of the magnetic field (see ú =F 1 and ú =F2 in
the formula). One of the constants, I.L, is of a fixed value, while the other, v, has two possible
390 Chapter 18
t.F
I50 g amma
20
c: 15
...0
t.l
10
ú= 5
"tl
III 0
ú= -5
-10
120' 140'
a
! !
25
-
0
20
c: 15
... 0
t.l 10
ú=
"tl
5
III
0
ú=
-5
am 80' 100' 120' 140' 160' 180' 160'
í ú á J ã J É ú f J N ú l ú Ü J J J J ú N J N ú N i Ü J J J J i f =ú N I Å O Ü ú ú f ú J J D N P ú Ü ú f J J J J J J N ú ú T Ü J J J J J J ú N D ú R Ü =
b
Figure 4. (a) Misdirection over the course of a day and the corresponding curve fitted by Martin and
Lindauer (1977) with their formula. In addition, the upper curve shows the variations in the intensity
of the earth's magnetic field recorded simultaneously with the dances. (An absolute scale for field
intensity is not given, presumably because only variations in intensity are important in the formula.
The static component of the earth's field is about 50,000 'I at this latitude.) am is the expected angle
of the dance (without misdirection) relative to the projection of the earth's field lines onto the dance
floor. The comb was aligned east/west. Note that the actual data seem to intercept the zero misdirection
line at am = 180 as predicted. The distance to the feeder was 400 m. (b) Same as in (a) except that
0
the distance to the feeder was 3000 m. From Martin and Lindauer (1977).
values depending upon temperature fluctuations during the dance recordings. The scaling
factor, A, apparently has a different value for each curve. There is no rule which specifies
the interval of time between time 1 and time 2, and Lindauer and Martin have apparently
changed the interval at will in fitting their curves. The undefined sign of the second term
is said to help to account for apparent time lags in the bees' responses to changes in IlF.
(The logic of this argument escapes us, but it is certainly the case that the curves could
not be fitted to the data without the option to change the sign of the second term at will.)
The formula works only for misdirection curves for dances indicating food sources about
400 m distant; for greater distances to the food, the formula tracks the curves only poorly,
and the misdirection is usually smaller than predicted (see Fig. 4b; Martin and Lindauer,
1977).
Given all of this, it would not seem unreasonable to disregard the formula altogether
were it not for its ability to track a variety of the highly irregular misdirection curves so
closely. From the curves presented by Martin and Lindauer, it seems possible that the
small natural variations in the magnetic field intensity and the dance angle relative to the
field lines do indeed together determine the misdirection-within, at least, certain limits
of several parameters. We feel, however, that the details of Martin and Lindauer's formula
must not be taken too literally at this point as the rationale behind its derivation is tenuous
at best, the units of the two sides do not match, and the evidence for its validity is based
entirely on empirical, albeit seemingly accurate, curve-fitting. The formula can and should
be used heuristically, however, as it makes specific predictions which can be tested through
artificial manipulations of the field.
Unless the field is artificially manipulated, with each relevant parameter of the field-
direction, polarity, total intensity, magnitude of variations in intensity, and time course
of variations in intensity-controlled spearately, we will continue to be forced to draw
inferences on the nature of the putative magnetic detector based on phenomena whose
very existence is supported by correlational evidence alone. A good first step would be to
attempt a repetition of the compensation experiments of Lindauer and Martin (1968) to
confirm their result. (We have tried and succeeded once.) There would then be three major
questions to address, each of which would best be handled separately: (1) Do the angles
of the dance relative to the field lines indeed affect the misdirection as Lindauer and
Martin's zero-crossing data suggest? This could be tested directly using an artificial field
of invariant intensity but the direction of which is varied at will. We would expect this
result to hold up. (2) Do variations in the total intensity of the field affect the misdirection
as Lindauer and Martin's curve-fitting suggests, and can we predictably "shape" the curve?
This could likewise be tested in artificial-field experiments in which the angle of the field
lines relative to the dance is held constant and field intensity is varied at will. (3) What
is the sensitivity of the detector, and what is its temporal resolution? These questions
could be addressed with a design similar to that just suggested where the amplitudes and
time courses of the imposed intensity variations are each varied in turn and the effects
observed.
In summary, a little bit of very good evidence suggests that magnetic fields are re-
sponsible for the misdirection in the gravity-oriented dances of bees, and quite a bit of
weaker evidence hints at how. The system seems to be one of great potential for the une-
quivocal demonstration of a magnetic effect on orientation and for precise experimental
studies of the detector system and the processing of magnetic information. One problem
is that we have no idea why magnetic fields should affect the orientation of dances, par-
ticularly only dances referenced to gravity and not to celestial cues. It seems unlikely that
the misdirection serves any useful biological function for the bees given that it appears to
have no effect on the communication (von Frisch and Lindauer, 1961). Our difficulty in
understanding the precise mechanism of the misdirection is compounded by the discovery
of Kilbert (1979) that modulations in the frequencies of the faint sounds produced by bees
392 Chapter 18
Figure 5. Horizontal magnetically oriented dances in

an earth-strength field. Each arrow is proportional to
the number of waggle runs oriented in that direction
(± 11.25°). A total of 84% of the waggle runs were ori-
ented toward the eight cardinal points (N = 24,601).
From Lindauer and Martin (1972).
as they dance run exactly parallel in every way to changes in the misdirection; the sounds'
frequencies always return to a "baseline" value in null magnetic fields, when the mis-
direction curves pass through zero, and when the dances take place on a horizontal comb.
Apparently the influence of magnetic fields on these systems is exerted in some very
general way and yet, at the same time, is restricted to behaviors linked to gravity orientation.
Perhaps misdirection and the associated sound frequency modulations are artifacts of the
gravity receptor system's construction (Lindauer and Martin, 1972) or of a magnetic field
detector that is useful in some other context (Gould, 1980b) or both. In any case, it seems
ironic that a phenomenon as confusing as misdirection should have such great experi-
mental potential.
3. Magnetically Oriented Horizontal Dances

A second effect of magnetic fields on dance orientation is apparent in the spontaneous
orientation of waggle runs toward the eight cardinal points of the compass (N, NE, E, ... )
when bees are forced to dance on horizontal surfaces in the absence of visual cues (Fig.
5). The bees must be forced to live and dance on a horizontal comb for at least 2 weeks-
an unnatural but seemingly tolerable situation for them-before they begin to show the
orientation (Lindauer and Martin, 1972), and it is certainly useless to them for the dance
directions are unrelated to the directions of the food sources visited by the dancers. The
existence of the phenomenon is not nearly as incomprehensible as that of misdirection,
however, in that such spontaneous orientation, serving no apparent purpose, is known to
exist in a variety of animals-although its physiological basis has yet to be worked out.
von Frisch (1967, pp. 407-409 and 442-443), for example, catalogs numerous species of
insects (including, of course, bees), crustaceans, and arachnids which show spontaneous
orientation to the direction of vibration of polarized light. In many of these cases, the
animals orient their bodies either parallel to, perpendicular to, or 45° from the direction
of vibration of the light-just as do bees to the direction of the magnetic field lines.
The spontaneous orientation of horizontal dances was first noticed by Lindauer and
Martin (1972), and their basic result has been reproduced several times since then (Martin
and Lindauer, 1977; Brines, 1978; Gould et al., 1980). As far as we know, no attempt to
reproduce the effect has ever failed. The best evidence that the orientation is based on
magnetic fields comes from field-compensation and intensification experiments by Martin
and Lindauer (1977): the orientation becomes very weak when the earth's field is com-
100
"0
...cQ)
80
Q) ...c
II)
0 &.
II)
c
2 ú = 60
Figure 6. Effect of field strength Q) "E ________________ J I ? ~ K WWWÇ ú å WK =
on the quality of orientation in - <1l

ú= (J
the horizontal dances. The
dashed line shows the results ex- ú ? b<1l= 40
pected if the orientation were .... 3:
random (50% of the waggle runs
o ...0
oriented toward the eight cardi-
nal points and 50% toward the 20
eight noncardinal points). N = eartn's field
64,375 waggle runs. Data from
Lindauer and Martin (1972) and
Martin and Lindauer (1977);
0 ú=
0 2 3 4 5
compiled in Kirschvink (1981,
Table 1). Field strength (Gauss)
pensated to within 0-5% (Le., 0-2500 'Y) and is improved slightly in stronger fields. Figure
6 summarizes the results comprising measurements of over 64,000 individual waggle runs.
As in the field-compensation experiments on misdirection, there was a time lag of between
30 and 90 min in the responses of the bees to rapid changes in field intensity.
Given the large amount of data in direct support of a magnetic basis for the orientation,
there can be little doubt that it is real. One interesting experiment that has not been at-
tempted seriously-although Brines (1978) made a very preliminary attempt which failed,
probably for technical reasons-would be to alter the main axes of orientation by placing
a hive in an artificial field whose horizontal component points toward a noncardinal point
of the compass (e.g., NNE). There is no reason to think that such an experiment would not
succeed, and it would exclude the remaining remote possibility that the orientation is
based on something other than magnetic fields.
The horizontal magnetic dance could provide a more straightforward experimental
system for studying the magnetic field detector than misdirection, and, as we will discuss
later, Gould et a1. (1980) and Kirschvink (1981) have already used it in initial attempts to
discover the nature of the detector. With the horizontal dance, one could easily measure
the absolute sensitivity of the detector and begin to look into the nature of the time lags
in the bees' responses to "jumps" in field strength. Unfortunately, it is probably not pos-
sible to study the sensitivity of the detector to small temporal variations in field strength
with the horizontal dance, for such variations probably have little or no effect on the
orientation. In any case, the horizontal magnetic dance is the most convincing case of
"real" orientation to the magnetic field by honeybees to date, and the phenomenon will
undoubtedly prove useful for future experimental studies of the detection and processing
of magnetic field information.
4. Magnetic Orientation of Comb-Building

Another situation in which bees seem to orient to magnetic fields is when a swarm-
a large group of workers and a single queen who have left their parent colony to found a
394 Chapter 18
new one-builds a comb in its new nest site. A honeybee nest usually consists of several
parallel sheets of comb suspended from the walls and ceiling of a cavity, and it would
make sense for the hundreds or, sometimes, thousands of workers starting to build new
sheets of comb in the darkness of their cavity to agree on the direction in which it should
be oriented (von Frisch, 1974, p. 94). Lindauer and Martin (1972), based on the results of
experiments with eight swarms, report that the bees in a swarm do in fact agree; the new
comb is oriented in the same direction (± 2°) as that of the comb in the parent colony. As
all other obvious cues (light, direction of nest box entrance, gravity) were eliminated by
the design of the boxes in which Lindauer and Martin's swarms built their combs, the
bees' maintenance of comb orientation suggests that they were using the earth's magnetic
field as their orientation cue. Lindauer and Martin also report that the comb orientation
can be predictably deflected by artificial magnetic fields and that very strong fields (10
times normal) greatly disrupt normal comb-building (Lindauer and Martin, 1972; Martin
and Lindauer, 1973). Few of the details of these experiments have been published, however.
Several attempts to repeat Lindauer and Martin's original observation that swarms
tend to build comb in the same direction as that of the parent colony have failed (Gould,
1980b; De Jong, 1982), but De Jong (1982) has been able both to repeat it and to demonstrate
a likely cause for the previous failures. De Jong found that when swarms were transferred
from one nest box to another, they would maintain their comb orientations only if they
had occupied the first nest box for fewer than about 9 days; "old" swarms do not maintain
their comb orientations, but "young" ones do. Having demonstrated significant mainte-
nance of comb orientation in young swarms, De Jong went on to ask whether or not the
earth's magnetic field provided the orientation cue. He collected five young swarms and
placed them in fresh nest boxes. After 2 days, by which time the bees had built enough
comb for its orientation to be measured, he transferred the bees to a new box and allowed
them to build again. In this way he got several comb orientation measurements from each
swarm. The critical thing is that upon each transfer, De Jong altered the horizontal com-
ponent of the magnetic field in the box by 90° with a set of Ruben's coils by switching the
coils either on or off. When the coils were off, the magnetic field vector pointed north;
when they were on, east. Altogether he measured 15 combs from trials in which the field
was rotated upon transfer of the swarm into its new box. The data are shown in Fig. 7.
The plot is arranged so that a comb oriented in the same absolute direction as that of the
previous comb built by that swarm-implying that the bees used some fixed external
reference for orientation-would plot at the top of the circle (0°), whereas a comb oriented
in the same direction relative to the (now-deflected) magnetic field-implying that the
bees used the horizontal component of the earth's field for orientation-would plot at the
bottom of the circle (180°). The distribution of comb orientations is significantly nonran-
dom (p < 0.05 by the Raleigh test), and the mean direction is not significantly different
from 180° (p = 0.02 by the V-test). Using De Jong's methods, our laboratory has now been
able to repeat this phenomenon. Thus, it seems that bees can indeed use the horizontal
component of the earth's field in orienting their comb.
The large amount of scatter in De Jong's data, which is much greater than the ± 2°
reported by Lindauer and Martin, raises the issue of the adaptive significance of the ori-
entation. It seems unlikely that orientation with so much scatter could be a very important
organizing force in comb construction. Two possibilities come to mind. First, it could be
that the precise age of the swarm is important in the orientation and that Lindauer and
Martin's results more closely reflect the natural situation, so that the orientation is, in fact,
normally useful in organizing comb construction. This seems entirely possible, for in Lin-
dauer and Martin's experiments each swarm apparently built a comb only once when it
was very young, whereas in De Jong's experiments most swarms were forced to build
several times and were of variable ages. The second possiblity is that the orientation is
simply an adaptively neutral aritifact of one sort or another. With regard to these possi-
Fixed Reference
Figure 7. Comb orientation by swarms
in De Jong's "coil on"/"coil off" ex-
periment. Each symbol represents the I
direction in which a swarm oriented I
its comb after being transferred to a
new box with the direction of the mag-
I
I
netic field deflected by 90° from that
in the previous box. Open circles:
o.• I •
"coil off"; horizontal component of
the field pointing north. Solid circles:
"coil on"; horizontal component
pointing east. The actual measure-
jI
ments could have ranged between 0
and 180°; they were doubled for anal- I
ysis and plotting. The arrow repre- I
sents the mean vector. <P is the mean
direction; r, the length of the mean
vector; u, the u-statistic for the V-test; 0: 198 0
•
0
and P, the probability value from the Magnetic Field
V-test with an expected direction of r : 0.39
180°. Statistics as in Batschelet (1965,
1972). Data from De Jong (1982, Table
U :2.06
3). p: 0.02
bilities, it would be interesting to repeat De Jong's protocol with swarms of precisely known
age (De Jong's were not) to see exactly what the effects of swarm age and multiple comb-
building are. Even more interesting would be to place a few swarms in null magnetic fields
to see, first, if their comb orientations would become random as one would expect given
De Jong's results and, second, if they do, to see if the organization with which the bees
build is substantially impaired by the absence of the presumed organizing force.
The horizontal magnetic dance and the dance misdirection together clearly have more
potential for being useful in studies of magnetic field reception in bees than comb-ori-
entation does, and both of the former provide much less cumbersome experimental sys-
tems. Thus, the magnetic orientation of comb-building may ultimately turn out to be most
interesting in the context of the coordination and adaptive significance of honeybee social
behavior.
5. Magnetic Fields and Orientation in Time
One way in which bees might use magnetic fields is to take advantage of the small
(50-100'Y or 0.1-0.2% of the total field) but regular daily cycles in the strength of the
earth's field for their orientation in time (Lindauer, 1977). The relevant characteristics of
these magnetic field cycles, generated by ion flow in the upper atmosphere, are discussed
by Skiles (Chapter 3) in some detail.
5.1. The Time Sense of Bees
It is well known that honeybees can keep track of the times of day when specific food
sources are offering food. If, for example, a group of bees is trained by being allowed to
396 Chapter 18
70
60 a
50
40
30
20
.. 10
G)
-...
ú =
G)
G)
o
-
.-...
-
I/)
I/)
60
b
> 50 r
40
- r
- .
30 r
- ...
r -
-
I-
20 I-
,
l
I-r-
IThn
10
2200 2400 :zoo 400 &00 8 00 1()OO 1:ZOO 1400 1&00 lSOO 20"" 2:ZOO
Time of day
Figure 8. (a) Results from a typical time-training experiment under conditions of constant light. tem-
perature. and humidity. The bees were trained to forage from a feeder that offered food only between
10 a.m. and noon each day (shaded region). On the test day. no food was offered. and the bees' visits
to the feeder were recorded. The bars show the number of visits recorded in each 1-hr interval. (b)
Same as in (a) except that the hive was placed in a strong (10 times greater), aperiodic (Le .. static)
magnetic field for the test. Data from Lindauer (1977) .
forage for several days from a feeder that offers sugar water at a particular time each day.
and then on a day when no food is offered the bees' visits to the feeder are recorded. one
finds that the bees begin to investigate the feeder well before the usual feeding time. show
a peak visitation rate just before feeding time. and give up investigating the (still empty)
feeder toward the end of the usual feeding period. This general result prevails even if the
experiment takes place indoors under conditions of constant light. temperature. and hu-
midity (e.g.. Fig. 8a) and has been reproduced numerous times in several laboratories
(Martin et 01 .• 1977; Gould. 1980b; reviewed by von Frisch. 1967). The obvious potential
explanation is that the bees rely on their endogenous circadian clocks for the timing in-
formation they seem to be using. Two kinds of evidence (reviewed in von Frisch, 1967,
pp. 352-359) suggest that this is, in fact, the case. The first is that if bees are chilled for
several hours after the feeding period on the day before testing, they show up late to the
feeder in the test (Renner, 1957). Chilling a bee might well slow down its circadian clock
(Sweeney and Hastings, 1960; reviewed in, e.g., Hastings, 1976) but would be less likely
to interfere with a bee's tracking of some external cue. The second kind of evidence comes
from longitudinal displacement experiments by Renner (1957): Renner found that if bees
were trained in a flight room under conditions of constant light, temperature, and humidity
in Paris and then transported overnight to New York for testing in an identical flight room,
the bees in New York visited the feeder almost exactly 24 hr after the last feeding in Paris,
not 5 hr later as would be expected if bees could measure local time according to some
external cue. Considering only these results, the"endogenous rhythm" hypothesis seems
a likely explanation for the bees' timing abilities under conditions of constant light, tem-
perature, and humidity.
5.2. Magnetic Fields and the Time Sense
A more recent experiment reported by Lindauer (1976, 1977) suggests that the real
explanation may be even more interesting. In the experiment, Lindauer and colleagues
followed the usual time-training protocol described above, but on the day of the test they
placed the hive in a magnetic field 10 times stronger than normal. The surprising result
was that the bees showed little sign of knowing the feeding time (Fig. 8b). (We have been
able to reproduce this result in a preliminary way.) This led Lindauer (1977) to conclude
that "bees use the diurnal periodicity of the magnetic field as a timer." It would certainly
make sense for animals that spend most of their time in dark, temperature-regulated cavities
to have some external timing reference other than the daily light and temperature cycles
used by most animals.
There are at least a couple of reasonable mechanisms by which bees could use the
temporal information in the geomagnetic field cycles. They could, for example, measure
the field continuously and simply memorize which food sources are producing food during
which part of the cycle. Alternatively, they could use the earth's field as a circadian "Zeit-
geber" or "time giver" (Aschoff, 1960)-Le., "set" their circadian clocks by the field cy-
cles-and then memorize feeding times relative to the phase of their internal clocks. This
second option would represent a less direct but more general role for the magnetic field
cycles in the bees' time sense. In either case, the bees would almost certainly use only the
variations in the field intensity; to use the direction variations, the bees would need to be
able to measure their body orientations to within about 0.01° (Gould, 1980b). Both pos-
sibilities could, in principle, be reconciled with Renner's (1957) observations discussed
above, but we will refrain for the moment from further speculation as the important issue
at hand is not how bees use the timing information in the field cycles, but rather whether
or not they use it in the first place.
Unfortunately, Lindauer's (1976, 1977) original experiment is impossible to interpret
critically because it confounds two potentially important parameters of the magnetic
field-its strength and its periodicity. Recall that the field used in the experiment was
both aperiodic (Le., static) and 10 times the strength of the earth's field. In all three cases
we know of in which bees were subjected to field strengths 10-times normal (approximately
5 G) or greater, the behavior of the bees was unusual or disrupted. The first is the field-
intensification experiments on misdirection by Lindauer and Martin (1968) discussed ear-
lier: strong magnetic fields (9-13 times normal) greatly increase the dispersion of the dance
angles. The second case is Martin and Lindauer's (1973) high-field experiment on comb-
398 Chapter 18
building in which the bees built an unusual, cone-shaped comb. The last is a recent report
by Tomlinsin et al. (1981) that placing a strong magnet next to dancing bees significantly
curtails the mean duration of their dance bouts, Le., they perform fewer waggle runs before
quitting. Thus, very strong magnetic fields seem to disorient bees-make them "sick"-
in some way, and it is possible that this general effect explains Lindauer's result. Two
other bits of evidence argue against the "sickness" hypothesis-at least as an explanation
of Lindauer's result-however. First, Lindauer (1976, 1977) points out that of 19 normal
(no artificial magnetic fields used) time-training experiments he performed, 4 failed, Le.,
the bees seemed not to know when the feeding time was. On the very same days on which
those four experiments were done, Lindauer's magnetometer had registered "magnetic
disturbances," suggesting that relatively low-intensity abnormalities in the field can di-
sorient the bees in time. The second bit of evidence comes from an experiment in which
trained bees were displaced in latitude from Germany to the Arctic Circle. Even though
solar time was roughly the same in both places, the daily patterns of variation in the earth's
field were quite different, and the bees performed poorly in the timing test (Lindauer,
1977). Given this indirect evidence, it is not possible to distinguish whether the "sickness"
hypothesis or Lindauer's "magnetic timer" hypothesis (if either) is correct. One quick test
of the "sickness" hypothesis would be to observe the effect of a field greater than 10 times
normal on the orientation of horizontal magnetic dances. Recall that in the horizontal
dances recorded by Martin and Lindauer (1977), a 4.5-G (approximately 9 times) field
increases the percentage of waggle runs oriented toward the cardinal points over that
obtained with weaker fields (Fig. 6). According to the "sickness" hypothesis, slightly
stronger fields should cause the orientation to break down.
The critical test of Lindauer's hypothesis, as suggested by Gould (1980b), would be
to see if we can predictably change the bees' subjective day length by canceling the natural
variations in the earth's field and inserting artificial ones with a different period in their
place. [Circadian rhythms can generally be stably entrained to periods at least 2 or 3 hr
shorter or longer than the optimal 24 hr (see, e.g., Hastings and Schweiger, 1976).] If this
works, one would be able to go on and ask exactly how the bees use the variations (Le.,
directly or indirectly via their circadian rhythms), what parameters of the field they meas-
ure (intensity or direction), and what their sensitivity is.
In summary, it is easy to imagine an adaptive value for the bees' use of the timing
information in the daily magnetic field cycles, but the only evidence we have so far that
the bees actually' do use it is equivocal. Fortunately, given the sophisticated techniques
now available for artificially manipulating magnetic fields, it should not be difficult to
perform the critical experiments. Then too, if bees do use the field cycles, we should be
able to use the tried-and-true time-training protocol to further understand how they do it
and how the magnetic detector works.
6. The Magnetic Receptor System

6.1. The Implied Sensitivity of the Detector
Unfortunately, our best guesses as to the sensitivity of the detector to magnetic field
intensity are based on the existence of the two phenomena for which we have the least
convincing evidence-the reported relationship between misdirection and variations in
field intensity, and the bees' apparent use of the daily intensity cycles as timing cues. If
misdirection does depend on variations in field intensity as Martin and Lindauer (1977)
suggest, then it implies a sensitivity on the order of 1-10 'Y if the detector measures absolute
field intensity, or 0.1-1.0 'Y/min if it measures rates of change in field intensity (Martin
and Lindauer, 1977; Gould, 1980b; Kirschvink and Gould, 1981). The misdirection system
appears to saturate at rates of change greater than about 1-3 'Y/min; at these rates the
dispersion of the dance angles becomes abnormally high (see Martin and Lindauer, 1977,
Fig. 18). These estimates, derived from the misdirection data, correspond reasonably well
to those obtained if we assume Lindauer's "magnetic timer" hypothesis is correct: in order
to achieve the ± 15 min accuracy displayed by Lindauer's bees (Fig. 8a), the bees would
have to be able to measure total field intensity to within about 2-5 'Y, or rates of change
in intensity of about 0.3 'Y/min, depending on which parameter they used (Gould, 1980b).
Although the correspondence between these independent estimates is encouraging, the
estimates are, unfortunately, not to be trusted, at least until the phenomena underlying
them are substantiated more rigorously. We suspect that the artificial-field experiments
we discussed earlier (Sections 2 and 5) will soon yield the critical measurements we need,
but for the moment the estimates are all we have.
The information available on the directional sensitivity of the detector is somewhat
more reliable. From the horizontal dance data it is clear that the accuracy of the detector
in earth-strength fields is on the order of ± 9° (SD; from the data in Gould et a1., 1980).
The real accuracy is probably even better, as the measurement errors of experimenters and
the inevitable battering a dancer experiences in attempting to maneuver among nestmates
on a crowded dance floor both probably contributed significantly to the variation in the
dance measurements we used for the estimate (Kirschvink, 1981). The precision with which
Lindauer and Martin's misdirection curves seem to pass through zero at the predicted
places implies a directional accuracy for the detector of about ± 2° in earth-strength fields
(e.g., Martin and Lindauer, 1977), and the comb-orientation data suggest that a swarm is
capable of measuring field direction with an accuracy of either ± 2 or ± 40°, depending
upon whose results one chooses to use. Finally, we know from the horizontal dance data
that a weak directional sensitivity still persists in 0-5% (0-2500 'Y) fields (Fig. 6). We still
do not know the absolute minimum intensity required for directional sensitivity, but the
appropriate experiments would be simple to do.
In any case, it is clear that honeybees can, at the very least, detect the direction (but
not necessarily the polarity) of magnetic fields somewhat weaker than that of the earth,
and it seems possible, according to some of the behavioral evidence, that they can sense
field intensity with an amazing 1-10 'Y accuracy. The obvious next step, aside from at-
tempting to confirm the reality of the magnetic effects for which the current evidence is
weak and to sharpen our estimates of the sensitivity, is to find out how they do it.
6.2. The Nature of the Detector
Because bees are terrestrial animals living in a nonconductive milieu and because they
seem to be able to sense magnetic fields when they are nearly immobile, it is not likely
that their magnetic field reception is based on induction-the principle that moving an
electrical conductor through a magnetic field will generate an electric potential (Gould,
1980b; Kirschvink and Gould, 1981; see Rosenblum et a1., this volume). Induction could
account for the magnetic field sensitivity of elasmobranch fish, however (Kalmijn, 1974,
1978). In addition, it seems unlikely that the receptor mechanism in bees is based on
paramagnetic effects. Paramagnetic fields are weak, local magnetic fields that arise from
the temporary net alignment of the spins of unpaired electrons in many kinds of molecules
under the influence of strong external fields. It is conceivable that these weak fields might
be detected in some way, but the interaction energy of a single electron in a typical organic
molecule with an earth-strength magnetic field is only one-millionth the energy with which
the electron is bombarded with thermal "noise" from the background (Yorke, 1979; Kirsch-
vink and Gould, 1981; see Yorke, this volume). It seems unlikely that the bees' nervous
400 Chapter 18
systems could make do with such a small signal-to-noise ratio. Although we should prob-
ably not dismiss induction (e.g., Cope, 1973) and paramagnetism (e.g., Leask, 1977) entirely
based on these considerations alone, the discovery of biologically synthesized magnetite
(FeO·Fe20a) in chitons (Lowenstam, 1962), then bees (Gould et 01., 1978), and now nu-
merous other organisms (reviewed in several chapters in this volume) suggests two other,
more plausible mechanisms based on two of magnetite's ferrimagnetic properties-per-
manent magnetism and superparamagnetism.
The magnetic behavior of a magnetite crystal depends upon the crystal's size and shape
(Butler and Banerjee, 1975; Kirschvink and Gould, 1981; Kirschvink and Walker, this vol-
ume). If the crystal is large enough or anisotropic enough, the spins of the unnpaired
electrons in the Fe2+ atoms in the crystal will, because of the physical arrangement of
the atoms, spontaneously align themselves with one another, generating a strong net mag-
netic field (See Banerjee and Moskowitz, this volume). The smallest magnetically self-
stable unit of magnetite is usually on the order of 0.1 j.Lm (but varies with crystal shape)
and is called a single domain. Thus, a single-domain crystal is permanently magnetized
and behaves like a compass needle; in the presence of an external magnetic field, it will
attempt to rotate until its own field is in alignment with the external one. It is not difficult
to imagine a number of ways in which single-domain crystals of magnetite could be in-
corporated into transducing organelles and linked to the nervous system, and Kirschvink
and Gould (1981) and Kirschvink (1981) propose several plausible models (see Kirschvink
and Walker, this volume). One model, for example, is a simple "compass needle" detector
that measures the torque on one or more single-domain grains to detect the direction of
an external field. The overriding consideration in all of the models is the interaction energy
between an earth-strength field and the magnetite grains as compared to the energy of
thermal "noise" (kT, where k is Boltzmann's constant and T is absolute temperature); if
the interaction energy is not on the order of kT or greater, the magnetic interaction will
tend to be swamped by thermal "noise." In short, Kirschvink and Gould found that the
interaction energies of the single-domain crystals in their models could easily equal or
exceed kT, so the "permanent magnet" mechanism seems entirely plausible.
The second property of magnetite which could be used in a magnetoreceptor, super-
paramagnetism, is a bit more subtle. Grains of magnetite which are just below the critical
self-stable size nonetheless possess magnetic moments which-unlike those of paramag-
netic substances-persist in the absence of external fields. The directions of these mo-
ments, however, unlike those of permanent magnets, are unstable and tend to track the
directions of external fields. And in the absence of external fields of sufficient strength,
they tend to be batted about at random by thermal energy. Thus, the direction of the
magnetic moment of a superparamagnetic grain can be changed without having to move
the grain. As with the single-domain-"compass needle" (Gould, 1980b)-mechanisms,
numerous "superparamagnetic" transduction geometries are imaginable (Kirschvink and
Gould, 1981), although they are less intuitively obvious. One example, in which we imagine
a series of superparamagnetic grains embedded in an elastic, stretch-detecting rod, is shown
in Fig. 9. As the figure indicates, in the presence of an external magnetic field perpendicular
to the rod, the magnetic moments of the grains would align themselves side-by-side parallel
to the field, causing the grains to repel each other and the rod to expand. In an external
field parallel to the rod, on the other hand, the moments of the grains would align them-
selves end-to-end, causing a mutual attraction among the grains, and the rod would con-
tract. With a few of these receptors arranged properly, an animal could measure both the
intensity and the direction (but not the polarity) of an ambient field. (With only one rod
there would be an ambiguity between intensity and direction.) As in the single-domain
models, the interaction energies are large enough to overcome thermal "noise," and the
forces generated could conceivably be detected by a nervous system if the right number
of grains were involved (Kirschvink and Gould, 1981).
A
_t_t_t-------I'lAB'"
Expansion
,,---t • •
B -----I-Bu. --
Contraction
Figure 9. Elastic-rod superparamagnetic transducer. In (AJ, the external field (Bextl aligns the super-
paramagnetic moments of all grains perpendicular to the rod. This causes the grains to repel each
other, and the rod expands. In (BJ, the external field aligns the moments parallel to the rod. This
causes the grains to attract each other, and the rod contracts. After Kirschvink and Gould (1981).
6.3. The Bees' Magnets
Bees seem to contain the magnetite that would be required to build both permanent
magnet and superparamagnetic detectors, and with the evidence available we cannot dis-
tinguish which mechanism they use, if either. Gould et 01. (1978) first discovered that bees
have permanent magnetic moments. These moments are aligned transverse to the bees'
long axis in the horizontal plane and most likely arise from the net alignment of the mo-
ments of numerous single-domain grains of magnetite in each bee. The models of Kirsch-
vink and Gould (1981) and Kirschvink (1981), as mentioned above, show that these single-
domain grains could conceivably account for the bees' abilities to detect magnetic fields.
Gould et 01. (1980) found, however, that some of the bees they tested possessed no de-
tectable permanent moments and that these "nonmagnetic" bees, in addition to other bees
that had been artificially "demagnetized" with a strong alternating field, could nonetheless
orient their horizontal dances to external fields. This aruges against the "permanent mag-
net" hypothesis, but does not allow us to rule it out for two reasons. First, as the single-
domain grains of bees are probably highly anisotropic (Gould et 01.,1978), and such grains
have a very strong tendency to align their magnetic moments along their long axes, the
alternating-field demagnetization procedure used by Gould et a1. (1980) probably only
scrambled the directions of the grains' moments insofar as it aligned the moment of each
grain in one of only two possible directions at random. If the magnetic moments of half
of the grains in a bee were oriented in one direction and the other half in the opposite
direction, there would be no net alignment of the moments, and the "demagnetized" bee
would possess no net field. However, only the polarities of the grains' fields would be
randomized, while their axes would not be changed. If bees were to use permanent magnet
detectors that ignored polarity-a simple "compass needle" torque detector, for example
(see Kirschvink and Gould, 1981, for others)-then the demagnetization procedure of
Gould et 01. would not have been expected to affect the bees' magnetic field orientation
in the first place. Further, the notion that the real detector does, in fact, ignore polarity is
supported by the behavioral data; bees-and birds for that matter (Wiltschko and
Wiltschko, 1972; Walcott and Green, 1974)-show no sign of being able to detect field
polarity. Our second reason for not being able to rule out the "permanent magnet" hy-
pothesis is simpler: it could be that Gould and colleagues' "nonmagnetic" bees did, in
fact, have net magnetic fields and that the fields were simply too weak for their instruments
to detect. As Kirschvink (1981) points out, however, this argument does not apply if bees
use their hypothesized permanent magnet detectors for the high-resolution magnetoin-
402 Chapter 18
6 Figure 10. Superparamagnetic

grains in a worker bee. The re-
laxation time for a magnetic do-
main increases exponentially as
5 the temperature is lowered.
Thus, some grains which are su-
perparamagnetic at room tem-
""'
::s 2 - - - -_ _ perature (300 K) will behave as
e 4
0
single domains capable of hold-

>DO) 3
I ing a magnetic remanence at liq-
o
uid nitrogen temperature (77°K).
Conversely, any net remanence
held by these domains at low
temperature will disappear as
they warm in field-free space
across their superparamagnetic/
stable domain transition (their
blocking temperature). The tem-
perature at which the remanence
is lost indicates the grain size re-
sponsible (see labels on ab-
scissa). To check for superpara-
Temperature CK) magnetic grains, the bees were
cooled to 77°K in the presence of
24 56 109 189 300 4417 a strong magnetic field (3000 G),
1
oI-I----+I-----i-----tr-----tlt----;1
forcing all the domain moments
150 200 250 300 350 400 into alignment. Curve 1 shows
Diameter of grain at SD/SPM the remanence of one bee as it
transition (A) warmed up inside the field-free
region of a cryogenic magnetom-
eter. Most of the remanence is lost as the stable/superparamagnetic size transition boundary goes from
300 to 325 A. Because a magnetite sphere 325 A in diameter has a moment of about 8.6 x 10- 15 emu,
about 2 x 108 of them are necessary to account for the lost remanence. Curves 2 and 3 show subsequent
warming cycles without exposure to the strong magnet. From Kirschvink and Gould (1981).
tensity reception implied by some of the behavioral data (see Section 6.1). This is because
calculations (Kirschvink and Gould, 1981; Yorke, 1981) have shown that the quantity of
single-domain magnetite that would be required for such a sensitivity would have been
easily detected by Gould and colleagues' instruments. Unfortunately, we do not know for
certain that bees are capable of such high-resolution intensity reception (Section 6.1), so
these issues must remain open.
Although the "axial" variant of the "permanent magnet" hypothesis cannot be ruled
out, the "superparamagnetic" hypothesis looks more attractive. For one thing, the single-
domain grains in bees seem to vary widely in both size and number (Gould et 01., 1978,
1980), whereas all adult bees so far tested seem to have substantial numbers (2 x 10 8 ) of
superparamagnetic grains that are almost all within the relatively narrow size range of
300-350 A (Fig. 10). This uniformity seems to suggest a biological function (Gould et 01.,
1980). In addition, just like an "axial" permanent magnet detector, a detector based on
superparamagnetic grains would be immune to alternating-field demagnetization, so there
is no conflict between the "superparamagnetic" hypothesis and Gould and colleagues'
(1980) results. Finally, most mechanisms based on transversely aligned single domains are
difficult to reconcile with Lindauer and Martin's (1972) observations on misdirection (Sec-
tion 2). Recall that there is no misdirection in dances oriented along magnetic field lines-
a situation in which the torque on transversely aligned single domains would be at a

maximum-and that there is also no misdirection in dances in null fields-a situation in
which the torque on the same domains would be at a minimum. As Gould et 01. (1980)
point out, it is difficult to understand how both a maximum and a minimum detector
response could produce the same behavioral output. Equally puzzling, if bees used per-
manent magnet detectors, would be the observation that bees dancing perpendicular to
the field lines-a situation similar to the null field circumstance in that the torque on
transversely aligned domains would be at a minimum-nonetheless display a maximum
misdirection (Gould et 01., 1980). Several of the superparamagnetic mechanisms proposed
by Kirschvink and Gould (1981) do not have these problems, so the "superparamagnetic"
hypothesis seems our best bet at the moment.
6.4. Anatomy
There has been some progress in locating the putative magnetite-based receptors. First,
Gould et 01. (1978) separated the heads, thoraxes, and abdomens of 100 dried bees, crushed
them, and passed them through a device which separated both the permanent magnets
and the superparamagnetic material from the rest. Only the abdomens contained a magnetic
fraction. Further, Gould et 01. dissected two dried bees with nonmagnetic tools and tested
various parts for (permanent) magnetism. They found permanent magnets in only the front
third of the bees' abdomens. (This says nothing as to the location of the superparamagnetic
grains.) Taking their cue from Gould and colleagues' findings, Kuterbach et 01. (1982)
examined the abdomens of bees for iron-containing cells and found bands of (seemingly
innervated) cells in each abdominal segment that contain numerous iron rich granules.
These granules are somewhat variable in size and shape, on the order of 0.5-1.0 /lm across,
and are composed of hydrous iron oxides in noncrystalline form. Kuterbach et 01. point
out that such hydrous iron oxides are typically paramagnetic at room temperature and
would not, therefore, contribute to the permanent magnetism measured by Gould et 01.
(1978). They point out further, however, that such materials are precursors of magnetite
in the magnetite biosynthesis of chitons (see Towe and Lowenstam, 1967; Kirschvink and
Lowenstam, 1979) and magnetotactic bacteria (Frankel et 01., 1979), and that only about
0.33%-an undetectably small amount-of the hydrous iron oxides in their bees would
need to be reduced to magnetite to account for the magnetic moments of Gould and col-
leagues' bees.
The findings of Kuterbach et 01. have numerous possible implications, and although
they mayor may not ultimately bring us closer to localizing and characterizing the bees'
magnetic detectors, they will certainly help us to focus our anatomical searches, at least
for the time being.
7. Summary and Conclusions

It is clear, largely from the behavioral observations and experiments of Lindauer and
Martin discussed above, that bees can at the very least detect and orient to the direction
(but not the polarity) of magnetic fields somewhat weaker than that of the earth. That bees
can detect field intensity, especially with the 1-10 'Y accuracy suggested by some of Lin-
dauer and Martin's results and later assumed (usually cautiously) by several authors, is
only weakly supported by the evidence, however, and badly needs to be studied in ex-
periments in which the magnetic field is artificially manipulated. Such experiments are
under way.
404 Chapter 18
It is likely that the magnetic field reception of bees is based on a mechanism involving
small grains of magnetite for two reasons: (1) the alternative mechanisms-based on in-
duction and paramagnetism-are unlikely, and (2) bees contain substantial amounts of
magnetite of both single-domain and superparamagnetic sizes which could, according
to theory, be used to build the necessary receptors. We do not know which of the two
possible magnetite-based mechanisms the bees use (if either), but the indirect evidence
available favors the "superparamagnetic" hypothesis over the "permanent magnet" hy-
pothesis, if only slightly. Behavioral experiments may never provide conclusive evidence
on this issue, but sharpening our knowledge of what the detector can actually accomplish
would help greatly.
Iron-containing cells in the abdomens of bees have recently been discovered, but the
visible iron-containing granules are not in the form of magnetite; instead, they are in the
form of hydrous iron oxides-known biosynthetic precursors of magnetite. We will need
both more information on the location of the bee's magnetite grains (both superparamag-
netic and single-domain) and more precise behavioral studies of the detector's capabilities
before we can justifiably begin to turn our questions over to the neurophysiologists-from
whose work the final answers will almost certainly have to come.
Note Added in Proof

Walker and Bitterman (1985) have shown that foraging honeybees can be conditioned to
respond to magnetic fields. Their procedure, which involves the pairing of sucrose rewards
with certain manipulations of the magnetic field, gives strong conditioning, and it seems
highly likely that the system will prove useful in experimental analyses of the bees' mag-
netic sense.
References
Aschoff, J., 1960, Exogenous and endogenous components in circadian rhythms, Cold Spring Harbor
Symp. Quant. Biol. 25:11-28.
Batschelet, K, 1965, Statistical methods for the analysis of problems in animal orientation and certain
biological rhythms, American Institute of Biological Sciences Monograph, Washington, D.C.
Batschelet, K, 1972, Recent statistical methods for orientation data, in: Animal Orientation and Nav-
igation (S. R. Galler, K. Schmidt-Koenig, G. J. Jacobs, and R. K Belleville, eds.), NASA SP-262,
U. S. Government Printing Office, Washington, D. c., pp. 61-91.
Brines, M. 1., 1978, Skylight polarization patterns as cues for honeybee orientation: Physical meas-
urements and behavioral experiments, Thesis, Rockfeller University, Appendix B.
titanomagnetite, ]. Geophys. Res. 80:4049-4058.
Cope, F. W., 1973, Biological sensitivity to weak magnetic fields due to biological superconductive
Josephson junctions?, Physiol. Chern. Phys. 5:173-176.
De Jong, D., 1982, The orientation of comb-building by honeybees, ]. Compo Physiol. 147:495-501.
Frankel, R. B., Blakemore, R. P., and Wolfe, R. S., 1979, Magnetite in freshwater magnetic bacteria,
Science 203:1355-1357.
Gould, J. 1., 1975, Honeybee recruitment: The dance language controversy, Science 189:685-693.
Gould, J. 1., 1980a, Sun compensation by bees, Science 207:545-547.
Gould, J. 1., 1980b, The case for magnetic sensitivity in birds and bees (such as it is), Am.Sci. 68:256-
267.
Gould, J. 1., 1982, The map sense of pigeons, Nature 296:205-211.
Gould, J. 1., 1984, Processing of sun-azimuth information by honeybees, Anim. Behav. 32:149-152.
202:1026-1028.
Gould, J. L., Kirschvink, J. L., Deffeyes, K. S., and Brines, M. L., 1980, Orientation by demagnetized
bees, J. Exp. Bio1. 86:1-8.
Hastings, J. W., 1976, Basic features, in: The Molecular Basis of Circadian Rhythhms (J. W. Hastings
and H. G. Schweiger, eds.), Dahlem, Berlin, pp. 49-62.
Hastings,J. W., and Schweiger, H. G. (eds.), 1976, The Molecular Basis of Circadian Rhythms, Dahlem,
Berlin.
Hepworth, D., Pickard, R. S., and Overshott, K. J., 1980, Effects of the periodically intermittent ap-
plication of a constant magnetic field on the mobility in darkness of worker honeybees, J. Apic.
Res. 19(3):179-186.
electric organs, in: Handbook of Sensory Physiology, Volume 3, (A. Fessard, ed.), Springer-Verlag,
Kalmijn, A. J., 1978, Experimental evidence of geomagnetic orientation in elasmobgranch fishes, in:
Animal Migration, Navigation, and Homing (K. Schmidt-Koenig and W. T. Keeton, eds.). Springer-
5:47-132.
Kilbert, K., 1979, Gerauschanalyze der Tanzlaute der Honigbiene (Apis mellifica) in unterschiedlichen
magnetischen Feldsituationen, J. Compo Physio1. 132:11-26.
Kirschvink, J. L., 1981, The horizontal magnetic dance of the honey bee is compatible with a single
Lett. 44:193-204.
Kuterbach, D. A., Walcott, B., Reeder, R. J., and Frankel, R. B., 1982, Iron-containing cells in the
honeybee (Apis mellifera), Science 218:695-697.
Leask, M. J. M., 1977, A physiological mechanism for magnetic field detection by migratory birds and
homing pigeons, Nature 267:144-145.
Lindauer, M., 1963, Kompassorientierung, Ergeb. BioI. 26:158-181.
Lindauer, M., 1976, Orientierung der Tiere, Verh. Dtsch. Zool. Ges. 1976:156-183.
Lindauer, M., 1977, Recent advances in the orientation and learning of honeybees, Proc. XV Int. Congr.
Entomo1. pp. 450-460.
Lindauer, M., and Martin, H., 1968, Die Schwereorientierung der Bienen unter dem Einfluss der Erd-
magnetfelds, Z. Vgl. Physio1. 60:219-243.
Lindauer, M., and Martin, H., 1972, Magnetic effects on dancing bees, in: Animal Orientation and
262, U. S. Government Printing Office, Washington, D. c., pp. 559-567.
Lowenstam, H. A., 1962, Magnetite in dental capping in recent chitons (Polyplacophora). Geol. Soc.
Am. Bull. 73:435-438.
Markl, H., 1964, Geomenotaktische Fehloroentierung bei Formica po1yctena Forster, Z. Vgl. Physio1.
48:552-586.
Martin, H., and Lindauer, M., 1973, Orientierung im Erdmagnetfeld, Fortschr. Zool. 21:211-228.
Martin, H., and Lindauer, M., 1977, Der Einfluss der Erdmagnetfelds und die Schwerorientierung der
Honigbiene, J. Compo Physio1. 122:145-187.
Martin, U., Martin, H., and Lindauer, M., 1977, Transplantation of a time-signal in honeybees, J. Compo
Physio1. 124:193-201.
Renner, M., 1957, Neue Versuche uden Zeitsinn der Honigbiene, Z. Vg1. Physio1. 40:85-118.
Sweeney, B. M., and Hastings, J. W., 1960, Effects of temperature upon diurnal rhythms, Cold Spring
Harbor Symp. Quant. BioI. 25:87-104.
Tomlinsin, J., McGinty, S., and Kish, J., 1981, Magnets curtail honeybee dancing, Anim. Behav. 29:307.
Towe, K. M., and Lowenstam, H. A., 1967, Ultrastructure in development of iron mineralization in
406 Chapter 18
bridge, Massachusetts.
von Frisch, K., 1974, Animal Architecture, Harcourt Brace Jovanovich, New York, pp. 93-94.
von Frisch, K., and Lindauer, M., 1961, Uber die "Missweisung" bei den rich tungsweisenden Tiinzen
der Bienen, Naturwissenschaften 48:585-594.
Walcott, C., and Green, R. P., 1974, Orientation of homing pigeons is altered by a change in the direction
Walker, H. M., and Bitterman, M. E., 1985, Conditioned responding to magnetic fields by honeybees,
,. Compo Physiol. in press.
Wiltschko, W., and Wiltschko, R., 1972, Magnetic compasses of European robins, Science 176:62-64.
Yorke, E. D., 1981, Two consequences of magnetic material found in pigeons, ,. Theor. BioI. 89:533-
537.
Chapter 19
Magnetic Butterflies
A Case Study of the Monarch (Lepidoptera,
Danaidae)
BRUCE J. MAcFADDEN and DOUGLAS S. JONES
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 407
2. Natural History of the Monarch Butterfly. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 408
3. Materials and Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 408
4. Induced Magnetization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 410
5. Ontogeny of Magnetic Mineralization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 411
6. Intraspecific and Interspecific Variation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 412
7. Attempts to Characterize the Magnetic Mineralogy. . . . . . . . . . . . . . . . . . . . . . . . 413
8. Summary and Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 414
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . " 415
1. Introduction
The search for magnetic material in organisms has encompassed a wide spectrum of the
biosphere from bacteria to humans. Until recently, the presence of the principal organically
synthesized magnetic mineral, magnetite, was known only from a few phyla (Lowenstam,
1981). As is evident from the other contributions in this volume, the last few years have
witnessed a large body of research dealing with the detection of magnetic material and
\subsequent behavioral experiments for all of the vertebrate classes. However, similar stud-
ies involving the detection of biogenic magnetite in many invertebrate phyla and of the
interactions of magnetic fields and behavioral responses are relatively scarce. This paper
discusses recent biomagnetic research concerning lepidopterans, focusing upon the mon-
arch butterfly, Danaus plexippus.
Relatively little attention has been paid to lepidopterans from a biomagnetic per-
spective. Baker and Mather (1982) tested for a compass sense in the large yellow underwing
moth, Noctua pronubo. The results of their experiments suggest that when placed in an
orientation cage (Helmholtz apparatus) with the ambient field reversed, these moths reo-
riented themselves apparently in response to this cue. They concluded that (as with ho-
neybees and pigeons; e.g., Dyer and Gould, 1981; Wiltschko et 01., 1981) the magnetic
sense can serve as a backup orientation device particularly on cloudy days and nights
when neither the sun compass nor celestial navigation is available. Jones and MacFadden
BRUCE J. MAcFADDEN • Florida State Museum, University of Florida, Gainesville, Florida

32611. DOUGLAS S. JONES • Department of Geology, University of Florida, Gainesville, Flor-
ida 32611.
407
408 Chapter 19
(1982) demonstrated the presence of magnetic material in the highly migratory monarch
butterfly, D. plexippus (discussed below). Jungreis (1982) investigated the presence of
biomagnetic material in nine species of migratory and nonmigratory butterflies, moths,
and crickets. She too found that only the monarch butterfly exhibited a significant mag-
netization.
This paper presents the results of experiments dealing with the detection of magnetic
biomineralization in the monarch butterfly. Such a study was initially undertaken because
this species is highly migratory and it seemed appropriate to inquire whether magnetism
played a part in this behavior. In addition to testing for the presence of biogenic magnetic
minerals in the monarch, this investigation provided an opportunity to address several
general biological questions that remained unanswered from previous biomagnetic studies
of other taxa, including: (1) Is the quantity of magnetic material significantly different in
males and females? (2) When during ontogeny is the magnetic material sequestered or
synthesized? (3) Are there significant differences in the amount of magnetic material pres-
ent in migratory versus nonmigratory species? We will first review the study reported by
Jones and MacFadden (1982) and then present (1) new experimental data concerning when
during ontogeny the magnetic material is synthesized and (2) a comparison of D. plexippus
with other species of the lepidopteran family Danaidae.
2. Natural History of the Monarch Butterfly

This section presents a brief synopsis of observations published elsewhere (Brower,
1977) concerning the natural history and life cycle of the monarch. The monarch butterfly,
D. plexippus, is a member of the milkweed butterflies (Danaidae). The Danaidae is a diverse
family with a worldwide tropical distribution. So far as is known, most of the danaids are
relatively sedentary. In a notable exception, the New World D. plexippus is highly mi-
gratory. The monarch is a very wide-ranging species, the limits of its range extending as
far north as Canada and as far south as Patagonia (Clark, 1941). During the late spring,
summer, and early autumn, populations of the monarch butterfly are found throughout
much of North America where they exploit the food resource Asclepias (which is a very
diverse genus of milkweed containing in excess of 100 species). In the autumn, these North
American populations undergo a spectacular southward migration (for distances up to 4000
km) and in so doing, avoid the extreme winters. Some western colonies overwinter in
California; however, until only recently the overwintering sites of the eastern populations
were unknown. During the 1970s a spectacular overwintering site was discovered in a very
localized area of the mountainous region in northern Michoacan, central Mexico. It has
been estimated that within one well-studied site (a) there was a population of 13.8 million
butterflies. In this mountainous region, the number of overwintering monarchs has been
estimated to be as high as 100 million individuals (Brower, 1977).
While spending the summer in North America, the monarchs reproduce. During the
reproductive phase, the monarch passes through several stages of the life cycle, from egg
to larva (with five instars) to pupa and then adult. This entire cycle takes about 1 month.
The young adults are reproductively mature within several days after they emerge from
the chrysalid. In terms of the migratory phase, Brower (1977, p. 41) noted that: "The mi-
gration of the monarch is very different from that of birds because the individual butterflies
that fly south in the fall are several generations removed from their ancestors that flew
north the previous spring. Their annual return to the overwintering groves is thus a com-
plex, inherited behavior pattern in which learning plays no role."

To ascertain whether the migratory monarchs possessed magnetic material which
could possibly by used in orientation and navigation, specimens were collected and ana-
Magnetic Butterflies 409
Till'
1)0 l\ I
(\)o° ---------------- ------------ I 0
ú T M=
SOBom
(I)
Figure 1. Orientation conventions of the monarch butterfly (Danaus pJexippus L.) specimens used in
this report. The four directions in the horizontal plane in which the body was brought close to the
magnet are as follows: 0°, anterior; 90°, right lateral ; 180°, posterior; 270°, left lateral. The two directions
in the vertical axis are top (dorsal) and bottom (ventral). The directions of orientation relative to the
SQUID sensor axes in the magnetometer are represented by (x), (y), (z). The position of wings and
legs, which were removed during this study, are indicated by dashed lines. From Jones and MacFadden
(1982).
Iyzed, beginning in 1980. A cryogenic magnetometer was used to test for the presence of
magnetic material in the tissue of the butterflies. This section will describe our procedures
for cleaning and measurement of butterfly specimens. Other techniques related to the de-
sign of growth, extraction, and other experiments will be discussed in the relevant sections
below.
The low level of magnetism detected in many organisms E ú N l J V | N M J N M= A m 2 ) requires
carefully planning and testing so that measurements are biologically meaningful and not
merely a result of ambient contamination. It is extremely important to perform careful
dissections with nonmagnetic tools. During our study we used plastic or Teflon instru-
ments. The tools and specimens were cleaned in double-distilled water before measure-
ments were taken. In order to determine that contamination was not introduced from the
water, frozen cubes of the double-distilled water were measured in the magnetometer; the
results revealed no detectable magnetic moment.
410 Chapter 19
To ensure that external contamination had been removed from the system, all spec-
imens were first measured uncleaned. After cleaning, each specimen was then remeasured.
This process of measurement, cleaning, and remeasurement was repeated until there was
negligible change in the magnetic moment (and presumably as much external contami-
nation as possible was removed from the system).
In the several experiments that involved adult specimens, we could not satisfactorily
clean the wings and appendages. These were probably contaminated by magnetic particles
trapped in the scales and setae. We therefore analyzed only the body segments (head,
thorax, and abdomen) of adults for the presence of magnetic material. These body segments
were cleaned by first removing as much of the external setae and scales as possible, washing
with double-distilled water, then blow-drying the specimens in air.
The magnetic moments of all specimens were measured using the two-axis supercon-
ducting cryogenic magnetometer at the University of Florida Paleomagnetics Laboratory.
Goree and Fuller (1976; also see Fuller et 01., this volume) describe the design and spec-
ifications of this instrument. All calculations of magnetic vectors were done on a PET 2001
microcomputer. Specimens were lowered into the magnetometer on a sample holder made
of "scotch" tape. The background magnetic moment of each piece of tape was measured
before it was used to hold a specimen. (It was found that the plastic "micromount" boxes
used in some other studies were subject to a wide variation in magnetic moment; they
were not used in this research.) Individual pieces of tape were rejected for use as specimen
holders if background noise exceeded -5 x 10- 11 A m2 • Five to ten measurements in each
of three directions were taken for the magnetization of the sample holders, uncleaned
specimens, and cleaned specimens. Each of these was then averaged and the background
noise of the tape was subtracted before the magnetic vector was calculated.
In order to determine whether the monarchs could be induced with a magnetization,
specimens were removed from the tape holders, moved to an opposite corner of the lab-
oratory, and either a 500- or 3000-G permanent magnet was held next to each specimen
for 10 sec. To ensure that the specimens never came in direct contact with the magnet,
they were placed in clean plastic vials and the magnet was kept in a plastic bag. Therefore,
any magnetic particles attracted to the magnet did not contaminat«;l the specimens. The
specimens (except for the eggs) were magnetized in one of six possible directions. The
conventions used were as follows (see Fig. 1): 0°, 90°, 180°, 270°, top, and bottom.
4. Induced Magnetization
This section summarizes the research presented in Jones and MacFadden (1982). Eight-
een (vacuum-)preserved monarch specimens were analyzed for their magnetic moments,
following the procedures outlined above.
The thoroughly cleaned bodies of these specimens were measured for any natural
remanent magnetization (NRM). None of these exhibited an NRM higher than the back-
ground noise of the magnetometer (-2 x 10- 11 A m2 ). Each specimen was then subjected
to a relatively strong magnetic field with a 5 x 10- 2 T magnet in a known orientation
(Fig. 1). In all specimens a significant induced isothermal remanent magnetization (IRM)
was produced which ranged from 0.37 to 1.67 x 10- 9 A m2 (X = 1.27 X 10- 9 A m2 ; s
= 0.77). Subsequent experiments with a stronger (3 x 10- 1 T) permanent magnet indicated
that our previous measurements using the 5 x 10- 2 T magnet were at or near saturation.
In order to localize the site or concentration of the magnetization, 10 specimens were
dissected (using nonmagnetic tools) into two parts: the head-thorax and the abdomen. The
head-thorax yielded about two-thirds of the magnetization with a mean value of 0.88 x
10- 9 A m 2 (O.R. = 0.43 to 1.66 x 10- 9 A m 2 ; s = 0.36). Nine abdominal portions
4
....
'"EI 2
t}S
II '
1x 10- 9
....« 8
c
.2
1j1 1
6
iiiN
I
;: 4
IIII!lij
CD
j, j
C
CI 2
as
:E 1x 10-
1
"C
CD 8
U
:::J
"C 6
.E
4
2
1x10- 11
LLx- Noise level of Magnetometer
rnmIVVit 1 2345 6 7891011 it 12345678910x J&M '82
Instar Days Following Days Following
Pupation Eclosion
'-' I I I I
Eggs larvae Pupae Adults
N=15 N=12 N=12 N=10
Figure 2. Acquisition of induced IRM for Florida Danaus plexippus during different ontogenetic stages.
Means and ranges of values are shown. Relative rates of synthesis for pupae and young adults are
indicated by slopes of "best-fit" lines: y = 0.69x + 0.09, r = 0.61 (pupae) and y = 2.13x + 0.61, r
= 0.86 (adults). The data plotted to the extreme right of the x axis ("J & M '82") are for D. plexippus
from site a, Michoacan, central Mexico. From Jones and MacFadden (1982).
yielded a significantly lower mean magnetization of 0.24 x 10- 9 A m 2 (s = 0.18; O.R. =

0.14 to 0.49 x 10- 9 A m 2 ). Further analysis indicated that most of the magnetic moment
is carried in the thorax.
For each specimen, the direction of induced IRM (Le., the position in which the magnet
was held) was determined for one of six possible directions (Fig. 1). It was found that the
direction of the induced field closely tracked that of the applied field for the 0°, 90°, 180°,
and 2700 orientations, a result that was previously demonstrated for honeybees (Gould et
al., 1978). For some unexplained reason, the "top" and "bottom" orientations did not
produce concordant results. These results could arise either from some internal property
of the magnetic material in the butterflies or anisotropy of the magnet. It was found that
no significant difference in either the intensity or the direction of induced magnetization
could be attributed to sexual dimorphism.
5. Ontogeny of Magnetic Mineralization

To discover when during ontogeny the monarch synthesizes magnetic material, in-
dividuals were raised in captivity and measured daily for their magnetic moment during
each developmental stage. On April 13 and 14, 1982, eggs and larvae (instars I-III) were
collected from stands of the milkweed, Asclepias humistrata, at Kanapaha Botanical Gar-
dens, Gainesville, Florida. The life cycle of these individuals was continuously monitored
throughout the next 3 to 4 weeks.
Initial NRM measurements of the eggs and larvae yielded values in the range of 1.73
x 10- 11 to 2.45 X 10- 11 A m 2 , virtually indistinguishable from the level of instrument
background noise (2 x 10- 11 A m 2 ; (Fig. 2). Larval instars I-V were measured daily for
412 Chapter 19
NRM, then placed in a strong (3 x 10- 1 T) magnetic field and remeasured for induced
IRM. The absence of NRM and extremely low IRM values for all larval stages (Fig. 2)
indicate that no magnetic material is present during this early phase of ontogeny.
Pupation occurred between April 22 and 27. Chrysalids were cleaned, exposed to the
3 x 10- 1 T field, and measured. This procedure was repeated daily throughout the 10-
to Il-day chrysalid stage. Induced IRM values for the pupae (Fig. 2) were initially low but
increased into the range of 10- tO A m2 by the end of pupation. The mean value for this
interval, 0.95 x 10- tO A m2 , suggests that a small but significant amount of magnetic
material is synthesized during this transition period leading to the adult (butterfly) stage.
Because only gas exchange occurs between the chrysalid and its surrounding environment,
the acquisition of magnetic material during pupation must represent in vivo synthesis
during histological reorganization (Le., it cannot represent ambient contamination).
Emergence of the adults from their chrysalids (eclosion) occurred between May 3 and
7. All adult specimens were again cleaned, exposed to a strong magnetic field, and meas-
ured daily for the following 8-10 days. As indicated in Fig. 2, induced IRM values of newly
emerged adults were comparable to those of advanced pupae. Over the first 4 days following
eclosion, mean IRM values increased to levels within the range of those previously meas-
ured in adult monarchs. As with the 18 adults during the earlier study (Jones and
MacFadden, 1982), the direction of the induced IRM closely tracked the orientation of the
applied field. Dissection and remeasurement revealed that the majority (about 65%) of the
magnetic moment is located in the thorax, consistent with previous results (Jones and
MacFadden, 1982). As indicated by the IRM values (Fig. 2), the rate of synthesis of magnetic
material in young adults is substantially greater than that in pupae.
The data thus indicate that monarch eggs and larvae possess no discernible magnetic
material. Significant quantities of magnetic material are synthesized during pupation and
this process continues in young adults. Within 4 days following eclosion, induced IRM
values are normally within the range of those seen in mature adults (Fig. 2; Jones and
MacFadden, 1982).
6. Intraspecific and Interspecific Variation

In order to assess the extent of biomagnetic variation among apparently allopatric
populations of the monarch, specimens were analyzed from four localities in Florida and
Central America (Table I). Both live and preserved specimens were measured for the Apa-
lachicola, Florida, population. In contrast to the relatively strong NRM (1.2 x 10- 9 A m 2 )
reported for bees (Gould et 01., 1978), the live monarchs from Apalachicola exhibited a
weaker but significant NRM (x = 1.91 X 10- tO A m2 , N = 3). The fresh specimens were
then preserved for several days by freezing, prepared, exposed to a strong field, and meas-
ured for IRM. These specimens had a mean IRM of 1.75 x 10- 9 A m2 that closely tracked
the direction of the applied field. The preserved monarch specimens from other localities
(Table I) also exhibited weak NRM. However, upon inducing a magnetic moment with a
strong field, these specimens showed mean values of 2.98 x 10- 9 A m2 (EI Salvador),
2.21 x 10- 9 A m2 (Oaxaca), and 1.27 x 10- 9 A m2 (Chiapas, Mexico). These results
demonstrate that different monarch populations have relatively high IRM, and by inference,
quantity of magnetic material, with respect to other danaid butterflies (see below).
In order to determine the extent of biomagnetic variation among different species of
danaid butterflies, carefully cleaned and prepared specimens were analyzed for D. gilippus,
Lycorea c1eobea, and D. philene (Table I). These specimens exhibited mean induced IRM
values of 0.78 x 10- 9 ,0.52 X 10- 9 , and 0.09 x 10- 9 A m2 , respectively. Model I ANOVA
indicates that mean induced IRM values are statistically significant both between danaid
Table I. Induced Magnetization (IRM) for Cleaned Specimens of the Monarch, Danaus
plexippus, and Other Milkweed Butterflies (Family Danaidae)Q
Species Locality N x(x 10- 9 A m2 )
D. plexippus Site ex. Michoacan. 18 1.27' 0.77

Mexico
St. Marks Lighthouse. 3 1.75 1.28
Apalachicola. Florida
EI Salvador 3 2.98 1.23
Candelaria. Loxicha 2 2.21 0.69
Oaxaca. Mexico
All D. plexippus The above 25 1.61 1.02
D. gilippus Jacarei. Brazil 2 1.23 0.54
Cali, Columbia 3 0.41 0.32
La Libertad. El Salvador 3 0.84 0.23
All D. gilippus The above 8 0.78t 0.45
Lycorea c1eobea Muste Chiapas. Mexico 3 0.30 0.17
Santa Tecla. El Salva.dor 3 0.74 0.74
All L. c1eobea The above 6 0.52t 0.54
D. philene Trobiand Islands. Papua. 3 0.09t 0.01
New Guinea
a See Jones and MacFadden (1982).
, Statistically significant. P < 0.05.
t Statistically significant. P < 0.01.
species (t in Table I, p < 0.01) and between populations of D. plexippus (* in Table 1. p

< 0.05). Further analysis using the Student-Newman-Keuls test suggests that the mean
induced IRM of D. plexippus is statistically different from the other three species, which
are not different from one another (p < 0.05); that mean IRMs among D. plexippus pop-
ulations from El Salvador and Michoacan are statistically different (p < 0.05); and that
mean IRMs among D. plexippus populations from El Salvador and Michoacan are statis-
tically different (p < 0.05). However, small sample sizes require caution in the interpre-
tation of these results and further research involving seasonal, geographic, and specific
variations is warranted. These data are interesting because, along with the results for D.
plexippus, they suggest a spectrum of biomagnetic variation for these danaid butterflies.
7. Attempts to Characterize the Magnetic Mineralogy
Several techniques were used to attempt a determination of the magnetic mineral and
domain structure that carries the induced IRM in the monarch. Unfortunately, our results
are not conclusive; all we can say is that the magnetic mineral is probably magnetite, but
none of our tests could unambiguously identify the mineral phase and domain structure.
We first attempted to directly isolate the magnetic mineral grains using the techniques
outlined by Walker et al. (Chapter 5, this volume). About two dozen thoroughly cleaned
monarch specimens (head, thorax, and abdomen; wings and legs removed) were placed in
a concentrated solution of hypochlorite (commercially available bleach). After several
days, most of the protein and fat had been dissolved, leaving the relatively inert residue,
most of which included chitinous material floating at the surface. We then took a 5 x
10- 2 T permanent magnet and placed it close to this solution to try to attract or aggregate
any magnetic mineral particles. This technique was unsuccessful in aggregating any par-
414 Chapter 19
tieIes visible to the unaided eye. The chitinous material floating at the surface was then
removed and the remaining liquid was pipetted onto microscope slides which were dried
in air. Subsequent examination using an oil immersion light microscope failed to reveal
any visible concentration of magnetic minerals. Furthermore, X-ray diffraction of the dried
residue did not reveal the presence of any magnetic mineral phases. The explanation for
the unsuccessful extraction is probably that there was not a sufficient number of monarch
specimens to produce enough magnetic material to be identified either under the micro-
scope or with the X-ray diffractometer.
In an indirect attempt to identify the magnetic material in the monarchs, we subjected
five specimens to low-temperature experiments. We first froze the cleaned butterfly spec-
imens to liquid nitrogen temperature (-196°C), placed them in the room-temperature ac-
cess tube of the magnetometer with a thermocouple attached to their bodies, and generated
a warming curve in that environment which described the rise in temperature as a function
of time. We then froze and quickly placed the specimens in the magnetometer without the
thermocouple and monitored the change in intensity of magnetization with respect to time.
Our experiments indicated a significant drop in magnetization as the specimens warmed
up past about -150°C. This phenomenon was also observed for honeybees, where Gould
et al. (1978, p. 1028) state that: "The presence of magnetite in three live bees was tested
by freezing them to -196°C in liquid nitrogen, and continually monitoring the remanence
as they warmed up through the isotropic point of magnetite (-143°C). In each case the
remanence decreased slightly near this temperature, indicating the presence of magnetite
... " (Ozima et al., 1964; Levi and Merrill, 1978). Therefore, the experimental results for
both monarch butterflies and honeybees are similar, and suggest the presence of magnetite
(although further work is required to determine the grain and domain size of this mineral).
8. Summary and Conclusions

Based on the new data presented here and those previously reported by Jones and
MacFadden (1982)' we conclude that the monarch butterfly possesses organically synthe-
sized magnetic material. The magnetic material, which is principally concentrated in the
thorax, is synthesized in vivo during pupation and young adulthood. Eggs and larvae do
not possess any magnetic material. The magnetic mineral in monarchs seems to be mag-
netite although this has not been demonstrated conclusively. Several allopatric populations
of D. plexippus show consistently high induced IRM in contrast to values obtained for
three other relatively sedentary species of danaids. Is there a correlation between the
amount of magnetic material and its use as a magnetoreceptor? Is it possible that the mag-
netic material in D. plexippus serves as a magnetoreceptor during migration? The tax-
onomic variation studies suggest that the presence of magnetic material in other danaid
species is not an all-or-nothing phenomenon. Nevertheless, a significant difference exists
between migratory and nonmigratory species.
These results do not prove that monarchs use this magnetic material to orient with
the geomagnetic field. Confirmation of magnetoreception in D. plexippus must await fur-
ther behavioral studies. The seasonal, biogeographic, and ontogenetic variations of mag-
netization within the monarch, other danaid species, and other lepidopterans in general
are all in need of further investigation as is the identity of the magnetic mineral involved.
In light of the discovery of biogenic magnetic minerals in several species of lepidopterans,
it is tempting to generalize to other species as yet not investigated. However, we know
that many lepidopterans possess no magnetic material (Jungreis, 1982). Much additional
work is needed before we can begin to account for the presence, distribution, and character
of biogenic magnetic material among the Lepidoptera.
ACKNOWLEDGMENTS. We thank L. P. Brower and J. Calvert of the University of Florida and

L. Miller and J. Miller of the Allyn Museum for the loan of relevant specimens and their
encouragement during this study. L. P. Brower, J. Cohen, J. L. Kirschvink, L. Miller, J.
Miller, N. Williams, and K. T. Wilkins critically read earlier versions of the manuscript.
This research was supported by NIH Biomedical Grants to the University of Florida. This
is Contribution No.5 from the University of Florida Paleomagnetic Laboratory.
References
Baker, R. R., and Mather, J. G., 1982, Magnetic compass sense in the large yellow underwing moth,
Noctua pronuba L., Anim. Behav. 30:543-548.
Brower, 1. P., 1977, Monarch migration, Nat. Hist. 87:40-53.
Clark, A. H., 1941, Notes on some North and Middle American danaid butterflies, Proc. U.S. Natl.
Mus. 90:531-542.
Dyer, F. C., and Gould, J. 1., 1981, Honey bee orientation: A backup system for cloudy days, Science
214:1041-1042.
Goree, W. S., and Fuller, M., 1976, Magnetometers using RF-driven SQUIDS and their applications
in rock magnetism and paleomagnetism, Rev. Geophys. Space Phys. 14:591-608.
Gould, J. L., Kirschvink, J. 1., and Deffeyes, K. S., 1978, Bees have magnetic remanence, Science
201:1026-1028.
Jungreis, S. A., 1982, Biomagnetism: A possible orientation mechanism in migrating and non-migrating
insects, Unpublished Master's thesis, University of Florida, Gainesville.
Levi, S., and Merrill, R. T., 1978, Properties of single-domain, pseudo-single-domain, and multi domain
magnetite, J. Geophys. Res. 83:309-323.
Ozima, M., Ozima, M., and ú â á ã ç í ç I = S., 1964, Low temperature characteristics of remanent magnet-
ization of magnetite, J. Geomagn. Geoelectr. 16:165-177.
Chapter 20
Magnetoreception and
Biomineralization of Magnetite
Fish
MICHAEL M. WALKER, JOSEPH L. KIRSCHVINK, and
ANDREW E. DIZON
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 417
2. Magnetic Sensitivity in Yellowfin Tuna. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 419
3. Detection of Magnetic Material in Fish . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 422
4. Characterization of the Magnetic Material. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 426
5. Identification and Analysis of the Magnetic Material. . . . . . . . . . . . . . . . . . . . . .. 429
6. Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 431
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 434
1. Introduction
Many species from different taxa respond to one or more features of the geomagnetic field
(Keeton, 1971, 1972; Lindauer and Martin, 1972; Wiltschko, 1972; Walcott and Green, 1974;
Martin and Lindauer, 1977; Quinn, 1980; Wiltschko et 01.,1981). These responses fall into
two general categories: responses to magnetic field direction and to magnetic field inten-
sity. Magnetic compass responses include the vanishing bearings of homing pigeons (Wal-
cott and Green, 1974) and directional preferences of migratory species in orientation arena
experiments (Wiltschko, 1972; Tesch, 1974; Quinn, 1980). The postulated magnetic in-
tensity, or "map", response (Gould, 1980, 1982; Moore, 1980; Walcott, 1980) refers to the
apparent ability of homing pigeons to determine their position to within a kilometer or
two using some feature related to geomagnetic field intensity. This response has been
inferred from the vanishing bearings and homing speeds of birds released at geomagnetic
field anomalies and during magnetic storms (Keeton, 1969, 1971, 1972; Gould, 1980, 1982;
Walcott, 1980). Gould (Chapter 12, this volume) provides a full discussion of this research.
Experimental evidence is accumulating that fish also possess a magnetic compass and
that they can learn to respond to magnetic fields in conditioning experiments. Quinn
MICHAEL M. WALKER • Southwest Fisheries Center Honolulu Laboratory, National Marine Fish-
eries Service, National Oceanic and Atmospheric Administration, Honolulu, Hawaii 96812, and De-
partment of Zoology, University of Hawaii, Honolulu, Hawaii 96822. JOSEPH L. KIRSCH-
VINK • Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena,
California 91125. ANDREW E. DIZON • Southwest Fisheries Center La Jolla Laboratory, Na-
tional Marine Fisheries Service, National Oceanic and Atmospheric Administration, La Jolla, California
92038.
417
418 Chapter 20
(1980), Quinn et 01. (1981), and Quinn and Brannon (1982) demonstrated unconditioned
compass responses in sockeye salmon fry and smolts. Fish held in the center of four-armed
wooden tanks (Quinn, 1980; Quinn et 01., 1981), or circular arenas with eight escape traps
(Quinn and Brannon, 1982) were released and allowed to enter the tank arm or escape trap
corresponding to their directional preference. In the absence of visual cues, the fish made
their preferences using magnetic field cues. These preferences were appropriate for bring-
ing the fish to their nursery lake or its downstream outlet.
Kalmijn (1978) conditioned a response to magnetic field direction in the ray, Urolo-
phus halleri, in which the fish were rewarded for entering an enclosure in the magnetic
east and punished for entering an identical enclosure placed in the magnetic west of the
experimental tank. When the magnetic field direction in the tank was reversed randomly,
the fish selected the enclosure in the magnetic east of the tank with accuracies of up to
90% in training sessions.
There are two central problems in the study of magnetic sensitivity in animals. The
first is that many of the behavioral results obtained so far are subject to methodological
criticisms, may be unreproducible, and tell little about the functioning of the sense (Emlen,
1975; Able, 1980; Griffin, 1982). The second is that as yet it is unknown how and where
the magnetic field is detected (Able, 1980). Thus, it is difficult to design and conduct
explicit experiments to obtain the necessary behavioral, anatomical, and neurophysiol-
ogical proofs of the existence of the sense, and to analyze its capacities.
Conditioning experiments can provide the necessary reproducibility and power for
unequivocal demonstration of the existence of a magnetic sense. However, attempts to
condition animals to magnetic fields have largely failed (Able, 1980). Where conditioning
has been obtained (Reille, 1968; Bookman, 1977; Phillips, 1977; Kalmijn, 1978), either the
experiments were unreproducible (Kreithen and Keeton, 1974; Beaugrand, 1976; Griffin,
1982) or subsequent psychophysical analyses of the sense were not carried out (Phillips,
1977; Kalmijn, 1978). These inconsistent results suggest that the experimental designs may
have been inappropriate for demonstrating responses to magnetic fields (Ossenkopp and
Barbeito, 1978). Thus, it is not yet possible to reject or accept without reservation the
hypothesis that animals can detect magnetic fields. Further attempts must be made to
obtain robust, reproducible responses to magnetic field stimuli and to identify their sensory
bases.
Any hypothesis seeking to explain geomagnetic field sensitivity in animals must pro-
vide a mechanism by which the action of the geomagnetic field can bring about orderly
displacement of the electrical potential of a receptor cell membrane. That is, the geomag-
netic field must act on the magnetoreceptor cell with a neural coupling energy greater than
the background thermal energy, kT (Jungerman and Rosenblum, 1980). The mechanism
must also explain both the general compass response and the very high sensitivities in-
ferred for detection of changes in magnetic field intensity (Martin and Lindauer, 1977;
Gould, 1980, 1982). Finally, the hypothesis should make testable predictions concerning
magnetoreceptor operation.
Among the hypotheses for magnetoreception that have been suggested are forms of
electrical induction (Kalmijn, 1974; Jungerman and Rosenblum, 1980); optical pumping
(Leask, 1977); liquid crystal effects (Russo and Caldwell, 1971); and biological supercon-
ductivity (Cope, 1971, 1973). These hypotheses demonstrate stimulus energies sufficient
to depolarize the membrane of a receptor cell and make geomagnetic field detection pos-
sible. However, few explain both the directional responses to magnetic fields made by
animals and the sensitivity to very small variations in magnetic field intensity exhibited
by homing pigeons and other birds (Southern, 1978, Gould, 1980, 1982; Walcott, 1980).
In addition, there may be a lack of evidence for receptor cells which behave in the required
fashion (e.g., Cope, 1973). Finally, magnetoreception is sometimes known to occur under
conditions where special requirements of the hypotheses are not met (Quinn et 01., 1981).
Fish 419
We do not intend to dismiss this theoretical work prematurely. However, the difficulties
with these hypothesized transduction mechanisms summarized above suggest that the
older and simpler ferromagnetic transduction hypothesis warrants further examination.
The possibility that the force exerted on magnetic particles could be transduced to
the nervous system has been independently suggested by Ising (1945). Lowenstam (1962),
and Keeton (1972). Support for the idea came with the discovery of magnetite in mag-
netotactic bacteria (Frankel et a1., 1979; Frankel and Blakemore, 1980), and in bees (Gould
et a1., 1978), birds (Walcott et a1., 1979; Presti and Pettigrew, 1980), and mammals (Zoeger
et a1., 1981). Theoretical analyses (Yorke, 1979, 1981; Kirschvink and Gould, 1981) show
that where the magnetite is present in a suitable form and in sufficient quantities, it could
provide the basis for a very sensitive magnetoreceptor system capable of deriving infor-
mation about direction and intensity of the geomagnetic field. Important predictions of
these analyses are (1) that the magnetoreceptors should not be dependent on special con-
ditions other than the presence of a magnetic field to function (Yorke, 1981), (2) that there
should be separate compass and intensity receptors (Kirschvink and Walker, this volume),
and (3) that compass accuracy and threshold sensitivity to intensity changes should be
constrained by the physical properties of the magnetite crystals (Kirschvink and Walker,
this volume).
The primary tests of the magnetite-based magnetoreception hypothesis must be be-
havioral and physiological. However, there are physical tests described in this chapter that
bear on the structure and function of a magnetite-based magnetoreceptor. Such tests include
magnetometry experiments designed to identify deposits of magnetite within the bodies
of different organisms, and to analyze the size, shape, and arrangement of the crystals.
Extraction and analysis of diffraction spectra of the magnetic material provide us with a
second means of identifying the mineral. Extraction also provides measurements of the
size and shape of individual crystals and information about the biomineralization process.
A feature of these studies is the very small amounts of material involved and the ease
with which contaminants can affect the results of experiments (Quinn et a1.'. 1981; Jones
and MacFadden, 1982; Walker et a1., this volume). Thus, it is important in testing for the
presence of magnetite in biologic samples to test for the presence of contaminants wherever
possible. Methods for doing this in whole tissues include testing for the effect of vigorous
washing and ultrasonic cleaning on the magnetic properties of samples used in magne-
tometry experiments (Jones and MacFadden, 1982) and testing for the magnetic properties
of ferromagnetic contaminants (Kirschvink et a1., 1985; Walker et a1., this volume). As-
saying for trace elements associated with geologic and artificial ferromagnetic materials
and examining the morphology of isolated crystals provide further means of determining
the origin of the crystals.
This paper reports results of our studies of magnetoreception and its possible trans-
duction mechanism in the yellowfin tuna, Thunnus a1bacares, and other pelagic fish. We
obtained reproducible responses to earth-strength magnetic fields in the yellowfin tuna
using an orthodox discrimination training procedure (Woodard and Bitterman, 1974).
Using a variety of magneto metric and mineralogic techniques, we have detected, extracted,
and characterized magnetite within the yellowfin tuna and other pelagic fishes. Although
these studies are not as complete as we would desire, they do permit us to compare results
between species from different orders, providing us with a more general insight into the
putative magnetoreceptor system of fish. We then consider the results in relation to mag-
netoreception and magnetite biomineralization.
2. Magnetic Sensitivity in Yellowfin Tuna

The first hypothesis to test in magnetic field conditioning experiments is that animals
can distinguish between different magnetic field stimuli. The procedures chosen should
420 Chapter 20
therefore be appropriate to measure simple distinguishability of different magnetic fields.

Decisions are required on experimental situation, testing procedure, response measure,
response produced, and stimuli to be distinguished. Once it is established that the stimuli
are distinguishable, different hypotheses will dictate different choices among these com-
ponents of learning experiments (Kling, 1971).
Learning is detected as a relatively permanent change in behavior resulting from con-
ditions of practice (Kling, 1971). Thus, in discrimination learning experiments, some meas-
urable bit of behavior is modified by experience gained by the subjects during testing. In
unitary, or go-no go, procedures, a single, generalized response is defined and then either
positively or negatively reinforced under different stimulus conditions. Discrimination
learning is then measured by comparing the readiness with which the response is expressed
between the stimulus conditions (Bitterman, 1966). In choice procedures, two discrete
responses that cannot be produced together are defined. In one stimulus condition, one
of the responses is rewarded and the other punished. In the alternate stimulus condition,
the consequences of the two responses are reversed. Discrimination is detected from the
choices the animal makes between the alternate responses under the different stimulus
conditions (Bitterman, 1966).
The approaches used to study magnetic field conditioning in animals can be sum-
marized as follows. Fields were uniform and movement was restricted, limiting the sub-
ject's ability to sample the magnetic field environment (Meyer and Lambe, 1966; Reille,
1968; Kreithen and Keeton, 1974). Magnetic field polarity was the most commonly used
discriminative stimulus (Phillips, 1977; Kalmijn, 1978; Griffin, 1982). The subject animal
was usually required to make a choice between two alternate responses and, with one
exception (Meyer and Lambe, 1966), multiple responding was not required (Bookman,
1977; Kalmijn, 1978; Griffin, 1982).
Magnetic fields are pervasive stimuli that can only be presented singly-first one, then
the other-in experimental situations. Choice conditioning experiments using stimuli,
such as magnetic fields, that can only be presented singly are among the most difficult
discrimination problems that can be presented to experimental animals (Mackintosh, 1974)
and will often fail with well-understood, salient stimuli (Bitterman, 1979). In choice pro-
cedures, trials most commonly end after the first response by the subject (e.g., Bookman,
1977; Kalmijn, 1978). However, discrimination will be sharpened if subjects are required
to produce multiple responses (Bitterman, 1979), as often occurs in unitary procedures.
Unitary discriminative training procedures are therefore more appropriate than choice
procedures for use with magnetic fields. They should permit the subjects freedom of move-
ment (Kreithen and Keeton, 1974), and sufficient time to sample the magnetic field en-
vironment and to produce multiple responses during trials.
For conditioning yellowfin tuna to magnetic fields, we defined a single response,
required the fish to produce that response more than once, rewarded that response under
one set of magnetic field conditions and not under another, and compared the readiness
with which the response was expressed under the two conditions. The fish were required
to swim through a 60 x 30-cm pipe frame lowered into their tank for 3D-sec trial periods
and retracted for intertrial periods. Within the trial periods the fish were able to swim
repeatedly through the frame. Thus, the measure of behavior compared between the stim-
ulus conditions was the rate of performance of the conditioned response. The primary
advantage of rate as a measure of discrimination is its sensitivity: it can vary widely and
rapidly in response to changes in experimental conditions and can accommodate short-
term variability in behavior (Kling, 1971).
These experiments were conducted at the Kewalo Research Facility of the Southwest
Fisheries Center Honolulu Laboratory, National Marine Fisheries Service. The fish used
were captive juvenile yellowfin tuna (40- to 50-cm fork length) tested individually in one
of two cylindrical test tanks (6-m diameter, 0.75-m depth). The experimental tanks con-
Fish 421
tained no metal and each had 100 turns of No. 18 AWG magnet wire wrapped around its
perimeter. A l-A direct current passed through these wires added a vertical magnetic field
to the background field. This field was nonuniform, adding from 10 j..lT in the center to
50 j..lT at the edge of each tank. The pipe frame, the magnetic field, and a semiautomatic
feeder mounted at the side of each tank (Jemison et aI., 1982) were operated by mechanical
and electrical linkages from an experimental control room. The control room was physi-
cally isolated from the experimental tank and was darkened during experiments. The fish
were observed through small viewing ports, and their responses were recorded manually.
The differences between the two magnetic fields used in the discrimination experi-
ments were as follows. The local Hawaiian field was uniform throughout the tanks. That
is, inclination, declination, and total field intensity were the same at any point in the area
occupied by the fish. The altered field introduced significant radially-oriented gradients
of both intensity and inclination within the tanks. These experiments therefore provided
the fish with two very different magnetic fields as discriminative stimuli. The fish could
conceivably monitor differences in magnetic field inclination, intensity, or the gradients
in these two parameters to make the discrimination.
In discrimination testing, a trial began with simultaneous presentation of the pipe
frame and either the reinforced (S+) or nonreinforced (S-) stimulus. All responses by the
fish within each 3D-sec trial period were counted. In S + trials the fish was rewarded with
a piece of food immediately following the first response after 30 sec. In S- trials, a 10-sec
penalty timer started at the end of the trial period. The timer was reset by each subsequent
response by the fish until either the penalty time elapsed or a total of 30 sec of penalty
had accumulated. Thus, response to S- was penalized by extending the trial without any
possibility of the fish obtaining food. The fish were given 3D-trial training sessions held
once daily. In any trial session the S + and S - were presented 15 times, with no more than
three S + or S - trials in succession. Testing was balanced by training two fish with the
normal Hawaiian field and two with the altered field as S+.
Discrimination between the two magnetic field conditions became evident after two
3D-trial sessions (Fig. 1). During the first day of testing, the response rates to the two stimuli
exhibited by the fish fluctuated randomly about each other. By the end of the second day,
the fish began to show a higher rate of response to the reinforced than to the nonreinforced
stimulus. Some individuals showed higher response rates during nonreinforced trials just
before beginning to make the discrimination. However, all individuals showed consistently
higher rates of response to S + than to S - on the third and fourth days of training, suggesting
that they were correctly anticipating the onset of positive or nonreinforcement.
All the fish completed at least 120 trials over four or more days of testing. An analysis
of variance comparing S + and S - response rates over 120 trials divided into blocks of
before and after 60 trials (the first two and the third and fourth training sessions) yielded
an F1,3 stimuli = 7.61, P = 0.07 and an F1,3 stimuli by blocks = 102.55, P = 0.002. All
other comparisons within the analysis did not approach significance. The main effect
(stimuli) failed to be statistically significant. This is attributed to the small differences in
response rates between S + and S - and to the variability of responding compared with the
mean rate of response. Dividing the data into blocks demonstrated that the behavior of the
fish on the third and fourth days of testing was different from their behavior over the first
2 days, and that the change in behavior that occurred was dependent on the discriminative
stimuli. Thus, the behavior of the fish changed as a result of their learning that production
of the response had different outcomes under different magnetic field conditions.
To test whether the fish were responding to possible equipment- or observer-related
cues, control trials were conducted with one fish. One of the wires connecting the power
supply to the coil around the tank was disconnected and all other procedures followed as
before. The response rates during reinforced and nonreinforced trials fluctuated randomly
about each other during this period (Fig. 2). When the circuit between the power supply
422 Chapter 20
I f', ,
,
7 I 'll
I
I ", ",IJ.
lIJ
/ 'tJ....." " , ..,.tJ. 5
Dú ? I I I I =
+
t-
:l
ú= 6
,
ú =
(/)
lIJ
(/)
(/)
ú= 5
5-
5 TRIAL BLOCKS
Figure 1. Magnetic field discrimination learning in yellowfin tuna. Each point is the average of five
S+ or S- trials for the four fish tested. Data from Walker (1984).
and the coil was reestablished, the fish was again able to anticipate positive and nonrein-
forcement and respond accordingly (Fig. 2). However, the separation between response
rates was less than before the control trials were conducted. In subsequent experiments
run using double-blind procedures, two fish were easily able to discriminate between the
two magnetic fields (Walker, 1984). From these results we conclude that the fish used only
the magnetic fields presented to anticipate the onset of positive or nonreinforcement.
3. Detection of Magnetic Material in Fish

As noted above, to act in magnetoreception, the energy of interaction between mag-
netite particles and the geomagnetic field must bring about the orderly displacement of
the potential of a receptor cell membrane. Kirschvink and Walker (this volume) assume
that, at the receptor level, energetic considerations are of primary importance and show
that single-domain particles are best suited for use in magnetite-based magnetoreception
by animals. To transduce the energy of their interaction with the geomagnetic field to the
nervous system, the particles must be at least partly free to move. They will then be aligned
by an external field and their position or movement will transform magnetic field stimuli
to mechanical stimuli (Kirschvink and Walker, this volume).
Freedom of the magnetite crystals in magnetoreceptor organelles to rotate and their
small mass mean that, in the absence of an external field, their orientation will be ran-
Fish 423
5
..J
ú=
a::
I-
"- 4
(/)
w
(/)
z
0
Q.
(/) 3
w
a::
u..
0
a:: 2
w
m
:IE
:::J
Z
W
(!)
ú =
a::
w
ú= 00
5 TRIAL BLOCKS
Figure 2. Magnetic field discrimination learning in individual yellowfin tuna. Shaded region in plot
indicates control tests. Data from Walker (1984).
domized by thermal agitation. The particle moments will therefore cancel and not be de-
tectable. Consequently, magnetite-based magnetoreceptor organelles will exhibit no natural
remanent magnetization (NRM) in the null field environment of a superconducting mag-
netometer. Freezing Ä á ç ä ç Öá ú =samples prevents any magnetic particles present from moving.
Their moments can be realigned by momentary exposure to a strong magnetic field (>0.3
T) and will then sum to produce a net moment, the saturation isothermal remanent mag-
netization (sIRM).
The sIRM has been the basic sample measurement in our biomagnetic studies. Presence
of magnetic material in a sample will be demonstrated by a ratio of the signal from the
magnetized sample to the background signal in the magnetometer, designated the signal-
to-noise (SIN) ratio, greater than unity. A scaling effect on the SIN ratio can result from
sample size: Large samples are more likely than small samples to have large SIN ratios.
To take account of this effect, we determined relative concentrations of magnetic material,
or intensity of magnetization, in different samples by dividing their moments by their mass.
In this case a reverse scaling effect could occur. Small samples with moments within or
close to the background noise in the magnetometer may appear more intensely magnetized
than larger samples with high SIN ratios. Determining whether or not a sample is magnetic
using either of these measures alone is arbitrary. Consequently, we chose to recognize as
magnetic only samples which gave both high SIN ratios and intensities of magnetization.
To distinguish magnetic material with a possible magnetoreceptive function from other
deposits, we sought to identify a tissue having the following characteristics: (1) It should
have a high remanent magnetic moment (or high SIN ratio) concentrated in a small volume
of sample (indicated by high intensity of magnetization) compared with other tissues from
the same fish; (2) the anatomic position of the magnetic tissue must be consistent from
fish to fish; and (3) the bulk magnetic properties, including particle coercivity, should be
similar in different individuals and in different species of fish.
Our first studies set out to determine whether magnetic material is consistently lo-
calized at any point in the body of the yellowfin tuna. Tissue and organ samples, including
424 Chapter 20
Table I. Saturation Moments. Signal-to-Noise Ratios. and Intensities of Magnetization

for Tissue and Organ Samples from Different Yellowfin TunaG
Mean moment
(pAm2) Intensity (pT) ±
Sample ± S.D. (N) Signal/noise S.D. (N)
Liver 105.0 ± 134.4 (2) 2.1 1.8
Pyloric cecum 49.6 ± 50.0 (3) 1.0 3.5
Intestine 14.5 ± 20.5 (2) 0.3 4.8
Red muscle 184.0 ± 274.7 (3) 3.7 3.5
White muscle 155.0 ± 211.2 (6) 3.1 5.7 ± 5.3 (3)
Brain 36.4 ± 50.3 (5) 0.7 7.5
Gill 95.0 ± 143.4 (6) 1.9 20.6

Skin 41.7 ± 80.1 (6) 0.8 35.7
Peduncle tendon 120.0 ± 169.7 (2) 2.4 41.4
Frontal bone 202.0 ± 129.0 (6) 4.0 103.6 ± 86.9 (2)
Pectoral fin 325.0 ± 427.0 (2) 6.5 62.5
Posterior brain case 375.0 ± 455.0 (6) 7.5 ND
Dorsal fin 400.0 ± 628.5 (5) 8.0 ND
Cardiac muscle 500.0 ± 707.1 (2) 10.0 4.5
Eye 1242.5 ± 1052.6 (4) 24.9 13.7 ± 14.1 (2)

Dermethmoid bone 1320.6 ± 867.5 (15) 26.4 127.0 ± 86.7 (7)
a Variability estimates are standard deviations and numbers in parentheses are the number of samples measured
for the saturation moments and intensities of magnetization. Intensity estimates for many samples came from one
fish only; "ND" indicates no data. Signal-to-noise ratios were calculated by dividing the mean saturation moment
by the maximum background noise (50 pAm 2 ) in the magnetometer. Samples are grouped into those which were
clearly nonmagnetic, those which were magnetic from their signal-to-noise ratio or their intensity of magnetization
only, and those which were magnetic from both measures.
bones of the body and skull. skin. sense organs. viscera. and swimming muscles. were
dissected from three fish (fork length 40-50 em) with glass microtome knives and handled
with nonmetallic tools in a magnetically shielded. dust-free. clean room. Although sub-
sequent dissections focused on the magnetic tissue. other samples were measured in all
fish. Samples were washed in glass-distilled water. frozen in liquid nitrogen. exposed to
strong fields from a cobalt-samarium magnet or an air-core impulse solenoid. and tested
for IRM in a superconducting magnetometer. Techniques for laboratory preparation and
dissection of tissue samples have been described elsewhere (Kirschvink. 1983; Walker et
01 .• this volume).
Six tissue or organ samples showed neither high SIN ratios nor high intensities of
magnetization (Table I). Seven other samples had moments less than 10 times the back-
ground noise in the magnetometer. However. because of the scaling effect caused by their
small mass. these samples showed high intensities of magnetization. Cardiac muscle and
eye samples acquired high moments but had relatively low intensities of magnetization.
Subdivision and remeasurement showed that the moments acquired by eye samples were
not associated with the lens. retina. or optic nerve. Because all these tissue samples were
either clearly nonmagnetic. appeared magnetic from one measure only. or were not re-
producible in different fish. it seemed unlikely that they would come from a sensory organ;
for this reason we focused our work on tissues that acquired large. reproducible moments.
Only the dermethmoid bone gave high values for all measures of magnetization used
in all fish examined (Table I). A scatter diagram. plotting intensity of magnetization against
SIN ratio for the data presented in Table 1, clearly identified the dermethmoid bone as the
most magnetic sample measured. Subdivision and remeasurement of the dermethmoids
Fish 425
Table II. Saturation Isothermal Remanent Magnetization in Selected Tissues of Different

Fish Q
Tissue IRM (pAm2)

Number
Species sampled Oermethmoid Muscle Eye
Thunnus albacares 16 260-3000 20-500 100-2600

T. alalunga 2 100-200 30 1700
T.obesus 1 480 400 NOb
Sarda orientalis 1500 700-1200 5000
Scomber sp. 1 1750 60 90
Makaira nigricans 4 110-170 NO NO
Oncorhynchus tshawytscha C 4 310-360 50-73 100-2300
Engraulis mordax 5 235-2850 40-170 NO
a Background signal in magnetometer 10-50 pAm'.
b ND. no data.
o Olfactory rosette and brain always nonmagnetic in O. tshawytscha; other tissues. especially viscera, variable in
all species.
in a number of fish suggested that the magnetic material was contained in tissue in a sinus
formed within the bone.
Taken alone. the SIN ratio data could lead to the conclusion that many tissues are
magnetic in the yellowfin tuna. However. most samples were only occasionally magnetic
and often vigorous washing or ultrasonic cleaning would reduce the IRM acquired by such
samples. This suggests the presence of external contaminants in samples where the clean-
ing procedure reduced the moment and anomalous deposits of magnetic material where
it did not.
We then tested for the presence of magnetic material in the bodies of other pelagic
fish species. We chose to ignore many of the corresponding tissues that were nonmagnetic
in the yellowfin tuna. and concentrated instead on those associated with the head. As in
the yellowfin tuna. these experiments did not identify or characterize the magnetic material
but they did allow us to (1) identify tissues that were magnetic in all individuals. (2)
identify tissues that were probably magnetic due to the presence of contaminants. and (3)
identify areas that may have been magnetic due to the presence of anomalous deposits of
magnetic material.
Fish from the orders Perciformes (families Scombridae. Istiophoridae. Coryphaenidae).
Clupeiformes (family Engraulidae). and Salmoniformes (family Salmonidae). all had mag-
netic material associated with the dermethmoid bone or the anterior region of the skull
(Table II). The sIRM values ranged between 100 and 3000 pAm2 for dermethmoid bones
from nine species of pelagic fish. Most of these sIRM values fell in the range from 100 to
1000 pAm2. being most consistent among individuals for the blue marlin. Makaira ni-
gricans. and chinook salmon. Oncorhynchus tshawytscha.
In our survey of different species. we worked mostly with the heads of fish. These
heads had often been cut with either a metal saw or a knife. which could easily have
contaminated tissues in the region of the cuts (see Bauer et al.. this volume). Where we
were able to work with whole fish. only tissues in contact with the environment frequently
acquired magnetic moments (Table II). For example. the northern anchovy. Engraulis mor-
dax. we studied were obtained whole from the Southwest Fisheries Center La Jolla Lab-
oratory. National Marine Fisheries Service. Their gills and guts frequently acquired mo-
ments. Vigorous rinsing with distilled water sometimes led to loss of the sIRM acquired
by these tissues. Muscle samples. which could not have been exposed to the environment.
were not magnetic in any of the anchovies we examined. Magnetic muscle samples in the
426 Chapter 20
heads of two chinook salmon could have been contaminated by knife cuts but did not lose
their moments after cleaning. The moments acquired by these samples may therefore be
due to biochemical deposits of magnetic material, and a number of tests will be necessary
to characterize them. Walker et 01. (this volume) describe these tests and discuss the anal-
ysis of a tissue contaminated by saw cuts. Here we focus on tests conducted on the der-
methmoid tissue in a number of species.
4. Characterization of the Magnetic Material

Studies of the acquisition and loss of magnetization by samples permitted us to identify
tentatively the source of the sIRM in the dermethmoid bone of the yellowfin tuna and to
make inferences on the organization of the magnetic crystals. A prediction of the magnetite-
based magnetoreception hypothesis is that the particles are free to move and that their
orientation will be randomized by thermal agitation at room temperature in a null field
environment. If the dermethmoid bone is the site of magnetoreceptor organelles, it should
possess no NRM and its acquired moment should be lost if it is allowed to thaw. The frozen
dermethmoid bones of seven yellowfin tuna first were tested for NRM. We magnetized these
samples, allowed them to warm to room temperature, and measured their moments at 5
min-intervals as they thawed. Four of these samples were then washed, refrozen, and
subjected to progressively increasing magnetic fields in an impulse magnetizer (Kirschvink,
....
ú =
.,....0
.,
0
0
0
0 0
20
10
00 60 70
TIME (MINUTES)
Figure 3. Loss of remanent magnetization with time on warming from liquid nitrogen temperature
(77°K) to room temperature (293°K) in the dermethmoid tissues of seven yellowfin tuna. Only two
measurements were taken at 35 min.
Fish 427
100
^ | ú J S=
r/
ú =
ú I ñ l Dú =
ú ÑI I ú =
80
ILl
(J)
<l:
ú Ñ? I I ú = TT/
(f
vY
ILl
II::
0
ILl 60
ú Ñ=
0
II::
úí = I/f
úk=
0
i, ( T/
ILl
(J)
<l: 40
ILl
II::
f úDf ú='tt
0
ú=
0ú =
20
/
S ú = ,
01
2 4 6 8 10 20 40 60 200 400 600 1000
PEAK FIELD (mT)
Figure 4. Progressive acquisition and loss of remanent magnetization by the dermethmoid tissues of
four yellowfin tuna, Vertical bars indicate standard errors, The ordinate and abscissa of the intersection
point are approximately 30% (R = 0,3) and 40 mT, respectively (see text), From Walker et 01. (1984),
1983). This procedure was followed by progressive alternating field (AF) demagnetization
of the samples after they were saturated, After each step in these magnetization and de-
magnetization experiments, we remeasured the moments of the samples.
The frozen dermethmoid bones of the yellowfin tuna showed no NRM (moments 3-
30 pA m2 ). After saturation, all showed an exponential decay with time in the moments
they retained (Fig. 3). This suggests that the orientation of the crystals was randomized
by thermal agitation as the tissue thawed. From this observation we conclude that the
crystals are at least partly free to rotate.
The dermethmoid tissues of four yellowfin tuna acquired virtually all of their reman-
ence in fields less than 200 mT and lost it again in unblocking fields between 10 and 100
mT (Fig. 4). The relatively narrow range of fields over which the dermethmoid tissues
acquired and lost remanent magnetization is consistent with the acquired moments re-
sulting from single-domain crystals of magnetite. However, the estimates of median coer-
civity of the magnetic particles obtained from the two methods differed significantly. This
discrepancy results from interactions between the magnetic particles.
The AF demagnetization curve for a dispersion of noninteracting single-domain crys-
tals should be symmetrical about the 50% level with the IRM acquisition curve. Asymmetry
in the two curves implies that the neighboring particles are sufficiently close for their
magnetic moments to interact to aid AF demagnetization and inhibit IRM acquisition (Ci-
sowski, 1981). These interaction effects lead to under- and overestimates of the median
microscopic coercivities of the crystals, respectively. Cisowski (1981) found that the shift
toward higher coercive fields in the IRM curve, and the shift to lower coercive fields in
the AF demagnetization curve are almost exactly the same. As a result, the abscissa of the
intersection point is independent of the interaction effect and provides an estimate of the
remanent coercive field. This has a value of 40 mT for the crystals in the yellowfin tuna
dermethmoid tissue. The crystals therefore fit into the single-domain magnetite region with
particle lengths of approximately 50 nm and axial ratios of about 0.8 in the Butler-Banerjee
diagram (see Fig. 4 in Kirschvink and Walker, this volume). Depending on the size dis-
428 Chapter 20
100
801- (A)
W
rJ)
«w
a::
u
w
0 60
a::
0
w
rJ)
«W 40
a::
u
ú=
•,,!! 20
0
100
(B)
w 80
rJ)
«
w
a::
u
w 60
0
a::
0
w
rJ)
«
w
40
a::
u
ú=
•ú = 20
0,
2 4 6 8 10 20 40 60 80 100 200 400 600 1000
PEAK FIELD (m T )
Figure 5. Progressive acquisition and loss of remanent magnetization by the dermethmoid bones of
(A) four chinook salmon and (B) one chub mackerel (dashed lines) and one striped bonito (solid lines).
Vertical bars in (A) indicate standard errors. R values are 0.32 (chinook salmon), 0.37 (chub mackerel),
and 0.32 (striped bonito). Estimates of median coercivities of the magnetic particles taken from the
abscissae of the intersection points are 46 mT (chinook salmon), 58 mT (chub mackerel), and 60 mT
(striped bonito).
tribution of particles, between 1 million and 100 million crystals would be necessary to
produce the sIRMs observed. These numbers are comparable to the estimates of the number
of single-domain crystals possessed by honeybees (Gould et 01.,1978) and homing pigeons
(Walcott et 01., 1979).
The intersection of the AF demagnetization and IRM acquisition curves for the tuna
dermethmoids is at approximately 30% magnetization [or R = 0.3 as discussed by Cisowski
(1981)] (Fig. 4). This is well below the value of 0.5 expected for completely noninteracting
single-domains. The crystals of magnetite in the dermethmoid tissue of the yellowfin tuna
Fish 429
therefore interact significantly and are about as closely associated as the crystals in the
semidispersed powder (R = 0.3) studied by Cisowski (1981).
Comparative data are available for a number of species. The IRM acquisition and AF
demagnetization curves (Fig. 5) for dermethmoid samples from individual chub mackerel,
Scomber japonicus, and striped bonito, Sarda orientalis, and for four chinook salmon are
almost identical in form to those obtained for the yellowfin tuna. The AF demagnetization
curves for the dermethmoids of two blue marlin were also very similar to the other AF
demagnetization curves and gave a median coercivity estimate of 18 mT (Walker, unpub-
lished data). These results give particle length estimates of 50-60 nm and axial ratios of
0.5-0.8 for magnetite particles in all these fish. The consistency of the results in species
of very different sizes and from different orders argues against the possibility that their
dermethmoids contain contaminants. We have carried out a number of tests which con-
clusively exclude contaminants as the source of remanence in the dermethmoid tissue of
the yellowfin tuna and which further demonstrate the high degree of control exercised
over the deposition of the magnetic particles.
The IRM acquisition and AF demagnetization curves for yellowfin tuna, chinook
salmon, striped bonito, and chub mackerel are very flat at fields less than 10 mT and also
at fields greater than 200 mT. This excludes multi domain magnetites, commonly found
in igneous rocks and laboratory dust, as the source of remanence in the dermethmoid
tissues. Multidomain magnetites are magnetically soft (Kirschvink, 1983) and acquire or
lose magnetization in much lower fields than did the dermethmoid tissues of these pelagic
fishes. The flattening of the IRM acquisition curves above 200 mT excludes hematite and
many iron alloys, which will continue to acquire remanence in fields above 1 T. Therefore,
we can exclude almost all ferromagnetic minerals except maghemite and synthetic mag-
netite as the source of remanence in the dermethmoid tissues of pelagic fishes. The extraction
and analyses of the magnetic particles discussed next enable us to exclude even these
possible sources of magnetic remanence.
5. Identification and Analysis of the Magnetic Material
The extraction techniques discussed by Walker et a1. (this volume) permitted a number
of distinctive assays for magnetite in the tuna dermethmoid tissue. Magnetic particles
extracted from the tissue were black, both to the naked eye and when viewed under a
dissecting microscope. This excluded maghemite as a possible source of the remanence
in the dermethmoid tissue and strongly suggested that the only magnetic mineral present
was magnetite. In an attempt to determine whether normally nonmagnetic tissues also
contained finely dispersed magnetic material, a large sample (about 10 g) of the white
muscle of one fish was digested using the same techniques. No magnetic particles were
obtained, presumably because any particles present in the swimming muscle must have
been present in concentrations too small to be extracted using these techniques.
X-ray diffraction, the technique used to identify the crystals, depends on the inter-
action between the collimated X-ray beam, the ions in the mineral, and their orientation
in the crystal lattice. The beam enters the sample and is scattered at angles characteristic
of the position of each ion in the lattice. The scattered beam is detected by X-ray photo-
graphic film which, after development, shows a series of concentric arcs. The distance of
each arc from the center is thus characteristic of the structure and composition of the sample
crystals. From these distances are calculated the distances (known as d spacings) between
adjacent ions in the unit cell of the lattice.
X-ray diffraction of the magnetic material extracted from the dermethmoid tissue of
the yellowfin tuna uniquely identified magnetite as the source of remanence (Fig. 6). The
430 Chapter 20
12 I- I I
. 10 f- -
,--..,
(/) ú =
ILl
i= 8 - -
enz ú =
ú =
ILl 6 - 0 -
I- ú =
C\I
Z 0 C\I
'It
4- --
ú =
'It
ú =
ILl -
> ú =
It)
ú=
II
ú =
2 -
...J
ILl
a: II I
0.15 0.20 0.25 0.30 0.35 0.40 0.45
d - SPACINGS (nm)
Figure 6. X-ray diffraction data for magnetite extracted from tissue contained within the dermethmoid
bone sinus of the yellowfin tuna. Vertical lines indicate relative intensities of lines in the diffraction
pattern. Numbers in parentheses indicate lines associated with magnetite and the crystal axis giving
rise to each line.
lattice parameter estimated from X-ray diffraction is 0.8358 ± 0.004 nm (reference value
0.8396 nm). The origin of the lines not associated with magnetite in the pattern is unknown,
although they do not arise from any known ferromagnetic mineral. Two possible sources
of these lines are connective tissue associated with the crystal aggregate and insoluble
proteins involved in an organic matrix in which the crystals are deposited (Weiner ef a1.,
1983).
Electron microprobe analysis of aggregated crystals showed that the magnetite from
the yellowfin tuna is remarkably pure. The crystals contained no measurable titanium and
almost no manganese (Table III), which are common components of geologic magnetites.
The crystals also contained no measurable chromium, excluding many possible synthetic
ferromagnetic minerals. As in the magneto me try studies, we can thus rule out almost all
nonbiologic origins for the magnetite crystals found in the tuna.
We were able to use transmission electron microscope (TEM) studies to measure the
size and shape of the isolated magnetite particles. These crystals averaged 45 nm in length,
38 nm in diameter, and had a subcubic form (Fig. 7). The crystals are thus single-domains
and conform to the size and axial ratio ranges predicted from their coercivities (see above).
Table III. Electron Microprobe Analyses of Magnetite Particles isolated from Yellowfin
Tuna
Magnetite
standard Weight (%) of
Oxide (NMNH 11487) sample
FeD 90.9 86.3 + 7.7
Ti0 2 0.2 0.0 ± 0.0
Cr 2 03 <0.25 0.0 ± 0.0
MnO <0.0 0.2 ± 0.1
CaD --- 0.2 ± 0.0
Total 91.4 86.7
Fish 431
Figure 7. Free magnetite grains extracted from the dermethmoid tissue of the yellowfin tuna. Trans-
mission electron micrograph courtesy of R. S.-B. Chang. From Walker et a1. (1984).
Their morphology differs from the octahedral crystal form found in all nonbiogenic mag-
netites. On the basis of this crystal form, we conclude that these crystals could not have
come from nonbiologic sources, but must have a biogenic origin. This is consistent with
the results obtained from the different approaches used in this study. The evidence from
a variety of methods for biogenic origin of the crystals gives us greater confidence when
discussing our results and their implications for magnetoreception and magnetite biomi-
neralization in fish.
6. Discussion
There is a growing body of evidence that fish respond to magnetic fields. Elasmobranch
(Kalmijn, 1978) and teleost (this study) fish have been conditioned to respond to magnetic
field stimuli while salmoniform (Quinn, 1980; Quinn et al., 1981; Quinn and Brannon,
1982) and anguilliform (Tesch, 1974) fish have shown unconditioned directional responses
to magnetic fields. Directional responses imply that fish possess a magnetic compass. The
conditioning experiment reported here (Figs. 1 and 2) provided changes in magnetic field
432 Chapter 20
intensity, intensity gradients, and inclination to fish being conditioned to discriminate

between magnetic fields. Thus, it is possible that the fish responded to intensity or intensity
gradients in these experiments. There is an important need for further conditioning ex-
periments to test the responses of fish to these components of the geomagnetic field and
to test the predictions of the ferromagnetic magnetoreception hypothesis (Kirschvink and
Walker, this volume).
It has not yet been established that fish use the geomagnetic field for direction finding
or navigation in the ocean. Quinn (1982) presents arguments that migrating Pacific salmon
must be able to navigate to achieve the distances traveled in the time taken to return from
the open ocean to the North American coast. Quinn goes on to suggest that these fish may
use a compass and a magnetic map based on a bicoordinate grid of magnetic field incli-
nation and declination for navigation.
There is some field evidence that fish are able to determine their position and swim-
ming direction, although the sensory bases for these abilities are unknown. Short-term
tracking experiments show that a number of different pelagic fishes, including swordfish,
Xiphios glodius (Carey and Robison, 1981), and Atlantic salmon, Solmo solar (Smith et
01., 1981), maintain relatively constant compass courses for substantial periods (up to
several days in the swordfish tracked by Carey and Robison). In addition, skipjack tuna
(Yuen, 1970) and swordfish (Carey and Robison, 1981) make diurnal return movements
to discrete areas (shallow banks) without retracing their outward path. The sensory mech-
anism or mechanisms responsible for guiding these movements are as yet unknown. How-
ever, the ability of fish to detect magnetic fields suggests that the possibility that they use
the geomagnetic field to guide their movements is worth serious experimental investiga-
tion.
The lack of feasible transduction mechanisms to explain magnetoreception has hind-
ered the testing of hypotheses concerning magnetic sensitivity in animals. One mechanism
for transduction of magnetic field information to the nervous system has been suggested
for the elasmobranchs by Kalmijn (1974, 1978). On the basis of his theoretical analysis,
Kalmijn (1974) predicted that the elasmobranchs should be able to detect magnetic fields
using their ampullae of Lorenzini. This prediction appeared to be borne out by his later
successful conditioning of rays to respond to magnetic fields (Kalmijn, 1978). However, the
critical behavioral experiment to determine whether or not electrical field information is
necessary for magnetoreception in the elasmobranchs has yet to be carried out.
Discovery of single-domain magnetite in a variety of metazoan groups provides the
basis for a general magnetoreception mechanism suitable for use in both aquatic and ter-
restrial environments. In the vertebrates, single-domain magnetite has been found in the
anterior dura mater membrane or in association with the ethmoid areas of the skull (Baker,
Chapter 26; this volume; Bauer et 01., this volume; Mather, this volume; Perry et 01., this
volume). We have found single-domain magnetite with virtually identical magnetic prop-
erties in the dermethmoid tissue of representatives of different orders of teleost fish. Our
studies of the magnetite crystals in the yellowfin tuna showed that the crystals average 45
nm in length by 38 nm in diameter (Fig. 7) and are arranged in interacting arrays that
appear at least partly free to rotate. These results enable us to refine our predictions con-
cerning magnetoreceptor organization and sensitivity in fish.
The magnetic properties of the magnetite crystals will determine and constrain the
operation of magnetite-based magnetoreceptor organelles (Kirschvink and Walker, this vol-
ume). From the dimensions of the crystals extracted from the yellowfin tuna, we can cal-
culate that 8.5 x 10 7 particles are necessary to produce the mean remanence observed in
the dermethmoid tissue. The energy of interaction of the individual crystals with the geo-
magnetic field will be about 0.1 kT (see Fig. 4 in Kirschvink and Walker, this volume). To
achieve coupling energies with the geomagnetic field large enough for detection by the
nervous system, the crystals must therefore be organized into interacting arrays. Depending
Fish 433
on the numbers of crystals in the arrays, the maximum theoretical sensitivity of a magnetite-
based magnetoreception system in fish can easily be estimated (Yorke, 1981). The apparent
freedom of the crystal arrays to rotate suggests that a mechanoreceptor monitoring position
or movement of the arrays is a suitable means to link the magnetite crystals to the nervous
system.
The appeal of the ferromagnetic magnetoreception hypothesis is that it can theoreti-
cally account for the responses to magnetic fields exhibited in such diverse groups as the
bacteria (Frankel and Blakemore, 1980), algae (Lins de Barros et a1., 1981; Lins de Barros
and Esquivel, this volume), bees (Lindauer and Martin, 1972), fish (Quinn, 1980), and birds
(Keeton, 1969). In the unicellular organisms and honeybees, the hypothesis has been tested
experimentally (Kalmijn, 1981; Kirschvink, 1981; Frankel et a1., this volume). For fish,
some indirect evidence for the hypothesis is also available. Quinn et a1. (1981) state that
the magnetoreceptor of the sockeye salmon must be able to function in the dark, in salt
and fresh water, in the absence of water flow, and be evolutionarily adaptable to magnetic
field reversals. Although not a test of the hypothesis, these behaviorally determined con-
straints argue against the optical pumping and electrical induction hypotheses for mag-
netoreception (Leask, 1977; Kalmijn, 1978) and are all compatible with possession by the
salmon of a ferromagnetic magnetoreceptor (Kirschvink et a1., 1985).
The case for magnetite-based magnetoreception in metazoans will not be proven until
linkage of the crystals to functioning sensory nerves transmitting magnetic field infor-
mation to the central nervous system has been demonstrated. However, the discovery of
biogenic magnetite, suitable for use in magnetoreception in different vertebrate and in-
vertebrate organisms, cracks the conceptual nut concerning a general magnetic sensory
mechanism and provides a good working hypothesis for the testing of many ideas about
magnetoreception. Substantial support for the hypothesis will come from experiments that
test for ferromagnetic effects on behavioral responses to magnetic fields. Kirschvink and
Walker (this volume) suggest experiments to estimate the magnetic moments of magnetite-
based magnetoreceptors and to test the theoretical constraints on ferromagnetic compass
and intensity receptors.
The variety of techniques used in these studies provided us with a number of internal
checks on our work. Through magnetometry studies we were able not only to identify
those areas of the bodies of fish that were magnetic, but also to show that the sIRM acquired
by the dermethmoid tissue was due to single-domain particles and not to magnetically
very soft or very hard contaminants. The microprobe analyses showed that the crystals
from the yellowfin tuna contained almost none of the impurities characteristic of geologic
magnetites. Finally, the unique nonoctahedral morphology of the crystals compared with
the octahedral crystal form of all other magnetites demonstrates the biogenic origin of the
crystals seen in TEM (Fig. 7; Lowenstam and Kirschvink, this volume).
Our studies on the magnetite crystals extracted from yellowfin tuna bear on the process
of magnetite biomineralization. The diffraction spectra show that the crystals are very pure,
implying that they are formed under close chemical control. The crystals are also very
uniform in size and shape. These properties are characteristic of biominerals formed under
matrix-mediated control (Lowenstam, 1981; Lowenstam and Kirschvink, this volume).
The only well-studied examples of magnetite biomineralization are in the chitons and
bacteria. These organisms appear to lay down a template of organic matrix and biochem-
ically precipitate magnetite within it under enzymatic control (Kirschvink and Lowenstam,
1979; Balkwill et a1., 1980; Nesson and Lowenstam, this volume). Matrix-mediated biom-
ineralization may be the mechanism for magnetite precipitation in fish as it could provide
the means for control of the size, shape, and composition necessary for use of the crystals
in magnetoreception. Locations of the crystals and their site of deposition in intact tissues
could provide us with valuable understanding of both their biomineralization and their
role in magnetoreception.
434 Chapter 20
Biomineralization processes have a long evolutionary history. The widespread use of

very similar matrix-mediated biomineralization processes in metazoans for skeleton build-
ing from the late Precambrian period (Weiner et al., 1983) suggests a common origin for
matrix-mediated biomineralization dating to the early Precambrian (Lowenstam and Wei-
ner, 1983). Magnetite, presumably formed by matrix-mediated biomineralization, has now
been identified in many phylogenetically distant metazoan groups. From arguments similar
to those advanced by Lowenstam and Weiner (1983), two interpretations of the appearance
of magnetite in these groups can now be offered. Biogenic magnetite in the metazoans arose
from a long history in the Precambrian which predates the differentiation of the metazoan
phyla, or it arose from multiple, independent origins in the late Precambrian, after the
divergence of the metazoan phyla. Resolution of the problem of the origin of biogenic
magnetite in the metazoans may depend on identification of biogenic magnetites of clear
metazoan origin from before the late Precambrian.
Lowenstam and Weiner (1983) suggest that the ability to produce magnetite evolved
in the early Precambrian as a means of iron storage in the reducing environment of that
time. In bacteria possessing magnetite crystals, the ability to perform directed swimming
responses may have provided selective advantages for magnetotaxis, even before the advent
of an oxygen-rich environment. In the metazoans, it seems necessary to postulate selection
operating on an association between magnetite crystals and sensory nerves as the origin
of magnetoreception. This assumes that magnetite played some other, prior role in the
body of the organism. Other studies have reported the presence of magnetite or magnetic
material which does not have an apparent magnetoreceptive function (e.g., Presti and Pet-
tigrew, 1980) and which may be lost in anemic individuals (Baker et al., 1983). Study of
these deposits and the sites and conditions under which they form may elucidate the early
functions of magnetite in the metazoans.
ACKNOWLEDGMENTS. Heinz A. Lowenstam and AnjaneUe Perry provided helpful discussions

during the course of this research. We are greatly indebted to Charles E. Helsley, Barbara
H. Keating, Li Chung Ming, and Michaela. Garcia of the Hawaii Institute of Geophysics,
University of Hawaii, for the use of paleomagnetic laboratory facilities and assistance with
X-ray diffraction and electron microprobe analyses. Shih-Bin Robin Chang and Karla A.
Peterson carried out the transmission electron microscope study and assisted in magne-
tometry experiments at the California Institute of Technology. Research funds and facilities
were most generously provided by the Southwest Fisheries Center Honolulu Laboratory,
National Marine Fisheries Service. This research was supported in part by a graduate study
award from the East-West Center, Honolulu, and a grant from Sigma Xi to M.M.W. This
is Contribution No. 3938 from the Division of Geological and Planetary Sciences, California
Institute of Technology.
References
Able, K. P., 1980, Mechanisms of orientation, navigation, and homing, in: Animal Migration, Ori-
entation, and Navigation (S. A. Gauthreaux, Jr., ed.), Academic Press, New York, p. 283-373.
Baker, R. R., Mather, J. G., and Kennaugh, J. H., 1983, Magnetic bones in human sinuses, Nature 301: 78-
80.
Balkwill, D. 1., Maratea, D., and Blakemore, R. P., 1980, Ultrastructure of a magnetotactic spirillum,
J. Bacterial. 141:1399-1408.
Beaugrand, J. P., 1976, An attempt to confirm magnetic sensitivity in the pigeon, Columba livia, J.
Camp. Physiol. A 110:343-355.
Bitterman, M. E., 1966, Animal learning, in: Experimental Methods and Instrumentation in Psychology
0. B. Sidowski, ed.), McGraw-Hill, New York, pp. 451-484.
Fish 435
Bitterman, M. E., 1979, Discrimination, in: Animal Learning: Survey and Analysis (M. E. Bitterman,
V. M. LoLordo, J. B. Overmier, and M. F. Rashotte, eds.), Plenum Press, New York, pp. 413-443.
Bookman, M. A., 1977, Sensitivity of the homing pigeon to an earth-strength magnetic field, Nature
267:340-342.
Carey, F. G., and Robison, B. H., 1981, Daily patterns in the activities of swordfish, Xiphias gladius,
observed by acoustic telemetry, U. S. Fish Bull. 79:277-292.
Cope, F. W., 1971, Evidence from activation energies for superconductive tunneling in biological
systems at physiological temperatures, Physioi. Chern. Phys. 3:403-410.
Cope, F. W., 1973, Biological sensitivity to weak magnetic fields due to biological superconductive
Josephson junctions?, Physiol. Chern. Phys. 5:173-176.
Emlen, S. T., 1975, Migration: Orientation and navigation, in: Avian Biology, Volume V (D. S. Farner,
J. R. King, and K. C. Parkes, eds.), Academic Press, New York, pp. 129-219.
Mater. 15-18:1561-1564.
Frankel, R B., Blakemore, R P., and Wolfe, R S., 1979, Magnetite in freshwater magnetota.ctic bacteria,
Science 203:1355-1356.
267.
Gould, J. L., 1982, The map sense of pigeons, Nature 296:205-211.
Gould, J. 1., Kirschvink, J. L., and Deffeyes, K. S., 1978, Bees have magnetic remanence, Science
201:1026-1028.
Griffin, D. R, 1982, Ecology of migration: Is magnetic orientation a reality?, Q. Rev. BioI. 57t3):293-
295.
Ising, G., 1945, Die physikalische Moglichkeit eines tierischen Orientierungssinnes auf Basis der Er-
drotation, Ark. Mat. Astron. Fys. 32A(18):1-23.
Jemison, H. A., III, Dizon, A. E., and Walker, M. M., 1982, An automatic feeder for liquids and wet
or dry solids, Behav. Res. Methods Instrum. 24(1):54-55.
Jungerman, R 1., and Rosenblum, B., 1980, Magnetic induction for the sensing of magnetic fields by
animals-An analysis, J. Theor. BioI. 87:25-32.
electric organs, in: Handbook of Sensory Physiology, Volume IlI/3 (A. Fessard, ed.), Springer-
Animal Migration, Navigation, and Homing, (K. Schmidt-Koenig and W. T. Keeton, eds.), Sprin-
ger-Verlag, Berlin, pp. 347-353.
Kalmijn, A. J., 1981, Biophysics of geomagnetic field detection, IEEE Trans. Magn. Mag-17:113-1124.
Keeton, W. T., 1969, Orientation by pigeons: Is the sun necessary?, Science 165:922-928.
Keeton, W. T., 1971, Magnets interfere with pigeon homing, Proc. NaIl. Acad. Sci. USA, 68:102-106.
(S. R Galler, K. Schmidt-Koenig, G. J. Jacobs, and R. E. Belleville, eds.), NASA SP-262, U.S.
Kirschvink, J. 1., and Gould, J. 1., 1981, Biogenic magnetite as a basis for magnetic field detection in
animals, Biosystems 13:181-201.
Lett. 44:193-204.
Kirschvink, J. 1., Walker, M. M., Chang, S.-B. R, Dizon, A. E., and Peterson, K. A., 1985, Chains of
single-domain magnetite particles in chinook salmon, Oncorhynchus tshawytscha, J. Compo Phys-
iol. A., in press.
436 Chapter 20
Kling, J. W., 1971, Learning: Introductory survey, in: Woodworth &- Schlosberg's Experimental Psy-
chology U. W. Kling and 1. A. Riggs, eds.), Holt, Rinehart & Winston, New York, pp. 551-613.
Kreithen, M. L., and Keeton, W. T., 1974, Attempts to condition homing pigeons to magnetic stimuli,
Leask, M. J. M., 1977, A physicochemical mechanism for magnetic field detection by migratory birds
and homing pigeons, Nature 267:144-145.
262, U.S. Government Printing Office, Washington, D. C., pp. 559-567.
Am. Bull. 73:435-438.
Lowenstam, H. A., 1981, Minerals formed by organisms, Science, 211:1126-1131.
Mackintosh, N. J., 1974, The Psychology of Animal Learning, Academic Press, New York.
Martin, M., and Lindauer, M., 1977, The effect of the earth's magnetic field on gravity orientation in
the honey bee (Apis mellifica), J. Compo Physiol. A 122:145-187.
Meyer, M. E., and Lambe, D. R., 1966, Sensitivity of the pigeon to changes in the magnetic field,
Psychon. Sci. 5(9):349-350.
Moore, B. R., 1980, Is the homing pigeon's map geomagnetic?, Nature 285:69-70.
Ossenkopp, K.-P., and Barbeito, R., 1978, Bird orientation and the geomagnetic field: A review, Neu-
rosci. Biobehav. Rev. 2:255-270.
Phillips, J. B., 1977, Use of the earth's magnetic field by orienting cave salamanders (Eurycea lucifuga),
salmon fry, J. Compo Physiol. A 137:243-248.
Quinn, T. P., 1982, A model for salmon navigation on the high seas, in: Proceedings of the Salmon
and Trout Migratory Behavior Symposium, (E. 1. Brannon and E. o. Salo, eds.), pp. 229-237.
Quinn, T. P., and Brannon, E. 1., 1982, The use of celestial and magnetic cues by orienting sockeye
salmon smolts, J. Compo Physiol. A 147:547-552.
Exp. Zoo1. 217:137-142.
Reille, A., 1968, Essai de mise en evidence d'une sensibilite du pigeon au champ magnetique a l'aide
d'un conditionnement nociceptif, J. Physiol. (Paris) 60:85-92.
Russo, F., and Caldwell, W. E., 1971, Biomagnetic phenomena: Some implications for the behavioral
and neurophysiological sciences, Genet. Psychol. Monogr. 84:177-243.
Smith, G. W., Hawkins, A. D., Urquhart, G. G., and Shearer, W. M., 1981, Orientation and energetic
efficiency in the offshore movements of returning Atlantic salmon, Salmo salar 1., Scott. Fish.
Res. Rep. 21, ISSN 03088022.
Southern, W. E., 1978, Orientation of ring-billed gull chicks: A reevaluation, in: Animal Migration,
Navigation, and Homing, (K. Schmidt-Koenig and W. T. Keeton, eds.), Springer-Verlag, Berlin,
pp. 311-317.
Tesch, F.-W., 1974, Influence of geomagnetism and salinity on the directional choice of eels, Helgol.
Wiss. Meeresunters. 26:382-395.
Walcott, C., 1980, Magnetic orientation in homing pigeons, IEEE Trans. Magn. Mag-16:1008-1013.
Walcott, C., and Green, R. P., 1974, Orientation of homing pigeons altered by a change in the direction
Walker, M. M., 1984, Learned magnetic field discrimination in yellowfin tuna, Thunnus albacares, J.
Compo Physiol. A. 155:673-679.
Walker, M. M., Kirschvink, J. L., Chang, S.-B. R., and Dizon, A. E., 1984, A candidate magnetic sense
organ in the yellowfin tuna, Thunnus albacares, Science 224:751-753.
Fish 437
Weiner, S., Traub, W., and Lowenstam, H. A., 1983, Organic matrix in calcified exoskeletons, in:
Biomineralization and Biological Metal Accumulation (P. Westbroek and E. W. de Jong, eds.l,
Reidel, Dordrecht.
by migrating European robins (Erithacus rubeculal, in: Animal Orientation and Navigation (S.
R. Galler, K., Schmidt-Koenig, G. J. Jacobs, and R. E. Belleville, eds.l, NASA SP-262, U. S. Gov-
ernment Printing Office, Washington, D.C., pp. 569-577.
Woodward, W. T., and Bitterman, M. E., 1974, A discrete-trialslfixed-interval method of discrimination
training, Behav. Res. Methods Instrum. 6:389-392.
Yorke, E. D., 1979, A possible magnetic transducer in birds, J. Theor. Bioi. 77:101-105.
537.
Yuen, H. S. H., 1970, Behavior of skipjack tuna, Katsuwonus pelamis, as determined by tracking with
ultrasonic devices, J. Fish. Res. Board Can. 27:2071-2079.
Zoeger, J., Dunn, J. R., and Fuller, M., 1981, Magnetic material in the head of the common Pacific
Chapter 21
Biomineralization of Magnetite in
Amphibians and Reptiles
ANJANETTE PERRY, GORDON B. BAUER,
and ANDREW E. DIZON
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 439
2. Amphibians. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 440
3. Reptiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 443
3.1. Magnetoreception in Sea Turtles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 443
3.2. Isolation and Identification of Magnetite in Sea Turtles . . . . . . . . . . . . . . . . .. 447
4. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 452
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 452
1. Introduction
Many amphibians and reptiles are seasonally migratory, traveling to and from suitable
breeding, feeding, or hibernation grounds. These creatures also carry out small-scale di-
rected movements in local areas. As juveniles, they must locate appropriate areas for growth
and maturation, often in environments vastly different from those into which they hatched.
Thus, soon after hatching, sea turtles scrabble up through the sand and find their way
down the beach to the ocean, while tadpoles, after spending several weeks in an aquatic
environment, metamorphose into frogs and climb out of their ponds onto dry land. As
adults, the amphibians and reptiles slither, stalk, or swim about in search of food and
shelter and to escape predators.
The distances covered by amphibians and reptiles in all of these movements are, with
notable exceptions, not of great extent, and in most cases the use of sight, smell, hearing,
and touch are probably sufficient for direction finding. In order to orient in familiar areas
while engaged in foraging, escaping from predators, or other routine movements, these
animals probably depend upon reference to learned landmarks, routes, and features of
topography (Ferguson, 1971). The occasionallongerforay into unfamiliar territory, unusual
weather or other adverse environmental conditions, and ontogenetic and seasonal migra-
tions may, on the other hand, require the use of a host of sensory capabilities. Studies
have shown that amphibians and reptiles can call on multiple sensory bases for direction
ANJANETTE PERRY • Department of Oceanography, University of Hawaii, Honolulu, Hawaii

96822. GORDON B. BAUER • Department of Psychology, University of Hawaii, Honolulu, Ha-
waii 96822. ANDREW E. DIZON • Southwest Fisheries Center La Jolla Laboratory, National
Marine Fisheries Service, National Oceanic and Atmospheric Administration, California 92038.
439
440 Chapter 21
finding (reviewed in Adler, 1982), and are capable of employing alternative cues and sen-
sory mechanisms when circumstances dictate or when denied use of a normally functional
sense.
The use of magnetic information for orientation has been demonstrated in other ver-
tebrates, including homing pigeons (Keeton, 1971), salmon (Quinn, 1980), elasmobranchs
(Kalmijn, 1978), eels (Tesch, 1974), and woodmice (Mather and Baker, 1981). Only pre-
liminary studies, however, have looked specifically at magnetic sensitivity as a component
of amphibian and reptile orientation. Magnetic cues could be of particular importance to
amphibia!1s and reptiles in environments where visual cues are limited, as is the case for
two of the animals discussed below, the cave-dwelling salamander and the marine turtle.
2. Amphibians
Amphibian movements are generally restricted to distances less than a few kilometers.
Within their limited range, seasonal migrations and movements around local areas are well
oriented (Schmidt-Koenig, 1975), and, in displacement studies, amphibians have exhibited
remarkable homing abilities. In one of many such experiments designed to examine am-
phibian orientation, Twitty et 01. (1967) displaced red-bellied newts up to 8 km away from
their native stream segment, either to other segments of the same stream or across a ridge
to some foreign stream. Over 60% of the displaced newts were able to find their way back
to their home stream segment, usually in the next breeding season. To account for such
homing accuracy among amphibians, researchers have investigated the use of olfaction,
vision, extraoptic receptors, and sun and celestial compasses, but have rarely questioned
whether magnetoreception may also be involved.
Phillips (1977) provided evidence for a learned directional response to earth-strength
magnetic fields in the cave salamander, Eurycea lucifuga. His objective was to determine
if animals trained to move through passageways under specific magnetic field conditions
would, when tested, use the learned magnetic information as an orientation cue. For train-
ing, he confined two groups of 15 salamanders each in two separate training corridors.
Each corridor consisted of two compartments filled with pieces of limestone, connected
by a darkened central passageway. Moisture was supplied alternately to one of the two
end compartments, forcing the animals to move at intervals from one compartment to the
other through the passageway. The two corridors were aligned along the same topograph-
ical axis, parallel to the earth's N-S magnetic field, with one, the "a" corridor, enclosed
in a coil which rotated the magnetic field 90 0 clockwise, and the other, the "b" corridor,
in the normal field. Movement of the A group was therefore perpendicular to the magnetic
N-S axis, while that of the B group was parallel to the earth's field. For testing, the two
groups were released simultaneously into the center of a cross-shaped assembly made up
of the two training corridors connected perpendicularly to one another. Animals were
released into the center of the testing assembly, and after 40 min the location of each
salamander was recorded. Animals were tested in both the natural and the altered magnetic
field, and with the cross assembly positioned with either the "a" or the "b" corridor parallel
to the topographical axis along which both were aligned in training.
In 6 of 16 tests there was a significant difference between the groups in their choice
of corridors, with the difference resulting from the groups moving in the direction predicted
from training, that is, perpendicular to the N-S magnetic axis for the A group and parallel
to the N-S magnetic axis for the B group (Fig. 1). In these cases the animals' distribution
shifted in the manner predicted by the orientation of the magnetic field only. Although a
directional response to magnetic fields was clear in these six tests, in the majority of tests
the distribution of the two groups was not significantly different. Phillips believes this
Amphibians and Reptiles 441
76-1 76-4
UT ú =
A
..L" ..L.. tl
II A
...
J J ú ú J J ú J J J J ú ú J J ú J J
ú = II II
X2 7.00, p< .01, n 28 I X2 5.82, P( .02, n 29
--r--------------------
76-5
r \'
76-8
A 8 A
.1. II .1. II .1. 11 ... II
X2 4 .17,p .05, n29 X2 9 . 14, P(.Ol, n 28
Figure 1. Results of tests with salamander groups A and B, previously trained in corridors a and b,
respectively. These tests represent the four possible combinations (insets) of the expected magnetic
directions relative to the home corridors and the topographical axis. The histograms show the per-
centage of each group of animals which moved perpendicular and parallel to the natural or altered
field. For example, in test 76-1, animals in group A, which were originally trained to move perpen-
dicularly to the magnetic field, do so preferentially (85%) even though the corridor perpendicular to
the magnetic axis (b) is not their home corridor; this preferred movement also coincides with the
topographical axis (represented by the vertical axis in each inset) . For statistical purposes, the com-
bined number of A- and B-group animals which moved along the expected axis is tested (chi-square,
one-tailed, one degree of freedom) against the null hypothesis of a 50 : 50 distribution of both groups
along the two perpendicular axes. From Phillips (1977).
discrepancy may have been due to different factors for each group. The A group may have
been influenced by familiar vibrations, as the tests were carried out in the location where
this group had been trained. Natural disturbance in the earth's magnetic field appeared to
affect the B group's performance, with increased deterioration in performance correlated
with the magnitude of disturbance.
Related experiments were carried out by Phillips and Adler (1978) with eastern red-
spotted newts (Notophthalmus viridescens), to determine if newts exhibit a locomotor or
positional response when placed in a magnetic gradient. Newts were individually tested
by placing them in aquaria oriented along the earth's N-S magnetic axis. As in Phillips's
442 Chapter 21
Table I. Mean Positions of Newts under Three Magnetic Conditionsu

Test Test
sequence Date A 8 C D E F mean
Mag-O tests
1 July 14 N 18.47 N 22.07 N 22.10 N 12.12 N 15.85 N 0.84 N 15.21
4 July 18 N 9.02 S 4.75 N 1.17 N 18.87 S 2.54 1.27N 3.43
7 July 21 N 10.43 N 18.77 excluded excluded N 9.22 excluded N 12.78
10 July 27 N 23.06 S 24.40 N 13.46 S 22.23 S 20.19 N 1.35 S 4.83
Mag-N tests
2 July 15 S 0.43 N 8.46 N 17.04 N 9.07 N 0.97 N 5.08 N 6.68
5 July 19 N 1.45 N 15.75 S 13.72 S 2.06 S 13.54 N 1.14 S 4.45
8 July 22 N 6.02 S 11.18 N 0.18 S 13.64 N 13.49 N 5.97 N 0.15
11 July 29 N 2.79 S 22.10 S 6.02 S 23.70 S 3.18 N 17.70 S 5.74
Mag-S tests
3 July 16 N 23.55 N 4.75 excluded N 16.08 N 11.35 N 15.06 N 14.15
6 July 20 N 6.81 N 18.47 S 3.56 S 9.96 N 16.69 N 5.46 N 5.66
9 July 25 N 15.32 S 18.47 excluded S 12.19 N 23.29 S 0.28 N 1.52
12 August 1 X X N 13.84 excluded excluded S 8.89 N 2.49
a Six newts (animals A-F) were tested in the sequence noted. with the bar magnet absent (Mag-O) or present at the
north (Mag-N) or south (Mag-S) end of the tank corridor. Individual means are corrected to exclude positions when
newts climbed on the tank wall and the mean is excluded altogether if the number of data points is less than 10.
Climbing animals. Le .. those climbing on the tank walls 10 or more times during a given test. are indicated by
italics. Mean values are given in centimeters. north (N) or south (S) of the center of the tank where each animal
was released. Animals A and B died of unknown causes on August 1. From Phillips and Adler (1978).
salamander experiments, light was kept dim and diffuse to avoid influencing the response.
Newts were exposed to three magnetic conditions: (1) the natural magnetic field, (2) mag-
netic north tests (Mag-N) in which a bar magnet was placed horizontally 5 cm beyond the
north end of the tank, and (3) magnetic south tests (Mag-S) in which the magnet was placed
an equal distance from the tank's south end. Newts were observed for a gO-min period
under each condition, and their position, as well as whether they were on the floor or
climbing the walls, was noted. From comparison of the mean floor positions of six animals
(Table I), and of the positions on the wall of climbing animals relative to the placements
of the bar magnet, the authors concluded that newts are capable of perceiving magnetic
cues and prefer positions where the field strength and inclination approximate ambient
values.
The apparent existence of magnetic sensitivity in newts and cave salamanders suggests
that the capability may be present in salamanders generally (Phillips and Adler, 1978).
The two salamanders studied belong to markedly different habitats, thus their magnetic
sensitivity probably did not evolve simply in response to specific environmental con-
straints, such as cave dwelling. While the terrestrial cave salamanders probably perceive
the magnetic field directly, the newts, like other aquatic vertebrates, could indirectly per-
ceive the field through electric induction. The two species studied thus far, however, are
phylogenetic ally related rather closely, and it is likely that they share a common type of
magnetoreceptor, possibly involving magnetite.
At this time, the two papers discussed above comprise the published research on
magnetic sensitivity in amphibians. Many questions still remain to be answered. Are other
amphibians sensitive to magnetic fields? How is such sensitivity used in orientation? Is
magnetic perception in amphibians magnetite-mediated, or is some other sensory mech-
anism responsible? Studies under the direction of Dr. J. B. Phillips and Dr. K. Adler are
currently in progress to examine magnetoreception in amphibians in greater detail.
3. Reptiles
Most reptile movements are confined within small-scale natural boundaries, such as
the shores of a pond for aquatic turtles, or to a few kilometers of desert or forest terrain
for the terrestrial snakes, lizards, and turtles. Only the sea snakes, the saltwater crocodile,
and the marine turtles travel across long distances.
Little is known about the migratory patterns and mechanisms of the sea snakes (Dun-
son, 1975) or the crocodile (Minton and Minton, 1973). Sea snakes are without well-defined
musculature, and are not capable of vigorous or prolonged swimming. It appears that these
snakes are almost planktonic and drift passively with the currents, sometimes over
hundreds or thousands of kilometers when they are caught up in the major ocean gyres.
The crocodiles may be more adept at charting a course than the snakes, but their movements
are almost wholly unresearched. Of course, long-distance migration is not a prerequisite
for magnetic sensitivity in reptiles. Nonmigratory snakes, alligators, crocodiles, lizards,
and turtles may use a magnetic sense for short-range orientation, but studies have not yet
been conducted to test this hypothesis.
Among the reptiles, the best-researched long-distance migrant is the green turtle, Che-
lonia mydas. Green turtles can accurately navigate in open ocean waters over distances
up to a few thousands of kilometers. Tagged individuals from populations throughout the
world have been shown to migrate from coastal feeding pastures to far-removed breeding
and nesting grounds on remote island beaches, returning to the same nesting sites re-
peatedly (Carr and Carr, 1970; Pritchard, 1973; Balazs, 1976). Hirth (1971) has suggested
that green turtles employ a multiplicity of cues and several senses, possibly including
magnetoreception, for navigation. Magnetic cues available to the turtles include the geo-
magnetic field's intensity, polarity, and inclination. Sea turtles could also use magnetic
anomalies such as the striped pattern on the ocean bottom that is a result of continuous
seafloor spreading during epochs of normal and reversed magnetic field direction, and
volcanic islands and seamounts that are sites of magnetic anomaly due to the high iron
content of the basaltic lavas that formed them.
A small amount of pilot work with green turtles and another of the migratory sea
turtles, the loggerhead Caretta caretta, has produced some results which could be sup-
portive of a magnetic sensitivity in these species.
3.1. Magnetoreception in Sea Turtles
Baldwin (1972) compared the headings in the wild of four radio-equipped green turtles
that had magnets attached to their plastrons with their headings when equipped with
nonmagnetic brass bars. On clear days or in shallow water, no differences in headings
before and after magnet placement were apparent. However, the heading of one turtle tested
in deep water under overcast conditions suggested that where the bottom is not visible,
sensing of the magnetic field could be important in turtle orientation. Further experiments
with the remaining turtles were hampered by difficulties with weather and with the radio-
tracking equipment, and to our knowledge these experiments have not been repeated.
We have conducted preliminary learning experiments to test for magneto reception in
sea turtles. All our experiments were carried out at the National Marine Fisheries Service
Laboratory in Honolulu, Hawaii. We first tested the ability of two 1-year-old green turtles
to discriminate between normal and altered magnetic field conditions (Perry, 1982). The
turtles were tested for magnetic sensitivity using a discrete-trials/fixed-interval condition-
ing technique (Woodward and Bitterman, 1974). Experiments were conducted in a 6-m-
diameter, circular, nonmagnetic pool that was wrapped with a Helmholtz coil constructed
444 Chapter 21
of 100 turns of magnet wire. A I-A dc current through the coil added a vertical field of
from 0.30 G at tank center to 0.50 G along the periphery to the Hawaiian magnetic field
of about 0.35 G. The turtles were trained to repeatedly press a paddle presented at tank
edge. At the end of 30-sec trial periods, they were rewarded with a small piece of fish.
The paddle was manually lowered into the tank at the beginning of each trial by the
experimenter, who was stationed behind a screen. A bar attached perpendicularly to the
paddle handle enabled the experimenter to lower the paddle while remaining out of the
turtle's view. A microswitch attached to the back of the paddle and connected to an au-
tomatic counter recorded the number of presses per trial.
For discrimination testing, the turtles were rewarded for pressing the paddle only
during stimulus-present (S +) trials. One turtle (turtle A) was rewarded for paddle presses
in the altered magnetic field, the other (turtle B) for presses in the normal field. Each trial
began after a random interval of 20-60 sec with simultaneous presentation of the paddle
and the altered or normal magnetic field. For S+ trials, the first paddle press after 30 sec
had elapsed resulted in reward presentation and paddle removal. During stimulus-absent
(S -) trials, presses during the 30-sec period earned nothing and, as punishment, the paddle
was not removed for another 20 sec, thus delaying the start of a new trial. Paddle presses
during this punishment period resulted in subsequent 20-sec delays in paddle removal,
up to 60 sec total. S+ and S- trials were presented in quasi-random order, with no more
than three trials of one type in a row. Each day's session consisted of 20 S+ and 20 S-
trials.
After the first 3 days of discrimination testing, turtle A's response rate during the S +
condition was significantly greater than during S- trials for 5 days (200 trials), using a t-
test for paired comparisons (t4 = 6.4, P < 0.01). In later trials, however, correct responding
by turtle A did not exceed 50%, and the experiment was halted (Fig. 2a). Although there
were no changes in the experimental setup for trials with turtle B, except to reverse the
rewarded and unrewarded conditions, turtle B never showed the ability to discriminate
between the two field conditions (Fig. 2b).
We also conducted learning experiments with an adult male loggerhead turtle of ap-
proximately 20-25 years of age. The experiments were conducted in one arm of aU-shaped
concrete tank. A free operant conditioning format was used in which the subject was
reinforced for pressing a paddle in the presence of a magnetic stimulus and was not rein-
forced for pressing in the absence of the stimulus. A buzzer (secondary reinforcer) followed
by a fish (primary reinforcer) was delivered after every correct paddle press, i.e., a con-
tinuous reinforcement (CRF) schedule was employed. Presentation of magnetic stimuli
and recording of data were fully automated; food was delivered by an experimenter blind
to the stimulus condition. There were no adversive experimental contingencies for incor-
rect responding. S+ and S- trials were counterbalanced over groups of 30 trials for ex-
periment I and of 12 trials for experiment II. It should be noted that the S- condition was
not equivalent to the normal Hawaiian magnetic intensity of about 0.35 G. Steel reinforcing
bars in the walls of the tank distorted the inclination and reduced the intensity to 0.27 G
at the experimental paddle.
In experiment I a small Helmholtz coil was placed perpendicularly to the N-S axis
of the experimental tank. This generated a graded field which amplified the horizontal
component of the ambient field and produced an overall intensity of 0.77 G at the point
of response, the paddle. Dc current was used to reduce acoustic artifacts, as ac generates
a 60-cycle hum.
As in the green turtle trials, the results from this experiment were somewhat ambig-
uous. Approximately 540, 30-sec trials were run over 6 days before performance stabilized
at 50% correct response. Over the following 4 days of trials (90 trials a day), performance
exceeded 50% on 3 of 4 days (range 48-54%). There was some indication of a deterioration
in performance at the end of each session, perhaps due to reduced motivation. If the last
a
...J 3
<C
a:
I-
D y ú D b i a =CONDITION
a
w
en \ \
/
0
M 2 L í WWK ú =
..... \
en \ / "- I
w \ /" /
en
en
w
a:
0..
t::./ 't::.---t::.
/ ' ..... 't::./
I
u. S- FIELD CONDITION
0
0z
w
(!)
<C
a: 0
w
>
<C
0 6 7
DAYS
b
...J
<C 3
a:
I-
aw FIELD CONDITION
en
0
(¥)
..... 2 '/
t::.
\
en /
w / ú =
en "-
en I
't::.
w /
"-
a: t::.
0..
"'t::.
""
u.
0 S- FIELD CONDITION
0
z
w
(!)
<C
a: 0
w 3 4
> 0 2 5 6 7
«
DAYS
Figure 2. Daily averaged rate of paddle presses by turtle A (a) and turtle B (b) after the first 3 days of
discrimination testing.
30 trials for each session are dropped from consideration, the subject exceeded 50% on
all 4 days (range 51-60%). Data considered over counterbalanced groups of 30 trials (15
S+ and 15 S- trials) indicated a significant difference using a t-test for paired comparisons
(t7 = 2.67, P < 0.05).
In experiment II the same free operant procedure was used, but the coil was suspended
horizontally above the subject. A field was generated which added to the vertical com-
ponent of the ambient field and produced an intensity of 0.58 G at the point of response.
Because of the turtle's relative stability in the horizontal plane, we felt that a vertical field
alteration would present the subject with a more stable stimulus as it moved within the
experimental area. Performance over 266 discrimination trials, however, remained at ap-
446 Chapter 21
proximately 50%. In other words, the turtle was responding equally in S+ and S- con-
ditions, i.e., at chance levels.
Although there were differences in inclination and intensity between the two logger-
head experiments, and between the loggerhead and green turtle experiments, methodo-
logical deficiencies preclude any intelligent discussion of differences in our results based
on these factors. The results of the experiments were equivocal from several standpoints:
1. The slow consummatory response of the loggerhead deprived it of response time
during the S + period, possibly artificially depressing discriminative performance.
2. A free operant procedure, especially using a CRF, allows the subject to track re-
inforcement, i.e., the act of rewarding a response increases the probability of another
response no matter what the state of the discrimination stimulus. In other words,
a better than 50% performance might be attained solely by using reinforcement cues
rather than magnetic cues.
3. The effect, when present at all, was small.
Due to differences in experimental design, items (1) and (2) were not problems in the green
turtle study, but the effect was also small (54-58% correct response) during trials with
turtle A, and was absent in turtle B's experiment. Under such circumstances, one has to
be concerned with subtle alternative cues or chance effects.
If one keeps the results of these discrimination experiments in perspective, they can
be considered promising pilot work. There are various procedural changes that can be
employed in future experiments which will reduce the ambiguity of results.
1. There are a number of learning formats which can reduce or eliminate reinforcement
cues. These include lean partial reinforcement schedules, testing on extinction trials, and
the discrete-trials/fixed-interval method reported by Woodward and Bitterman (1974) and
used in the green turtle experiments (see also Walker et 01., Chapter 20, this volume).
2. Punishment, e.g., time-out, or delays before the beginning of another trial, has been
shown to be effective in reducing incorrect responses in discriminatory learning by green
turtles (Manton et 01., 1972) and in discriminatory learning of magnetic cues by tuna (M.
Walker, personal communication). Mild punishment procedures (delay of paddle removal)
were used in the green turtle experiment, and might prove effective in training loggerheads
and other turtles in magnetic discrimination formats.
3. The use of only one or two subjects and a single two-choice format leaves open to
question the influence of nonexperimental cues, i.e., discriminative cues other than the
magnetic field which may bias responding. Using single subject designs, the influence of
alternative cues can be lessened by reversals of reinforcement contingencies. For example,
initially reward responses during the presence of a magnetic field but not in its absence.
After a criterion percentage correct is attained, reverse the contingencies, reward only
responses in the absence of the magnetic field, but not in its presence. In this manner,
cues that might accompany one magnetic condition but not the other are equated over the
experiment. Another alternative is to increase the sample size and make reward contin-
gencies differ for different groups, i.e., one group is only rewarded in the presence of a
magnetic stimulus, the other is rewarded only in the absence of the magnetic stimulus.
4. Fixing the presentation of food at a given site and fully automating the feeding
procedure can reduce potential experimenter cues as well as encourage a more time ef-
ficient consummatory response by the subject.
5. The use of more magnetically clean procedures can enhance the likelihood of a
subject detecting a stimulus as well as giving the experimenter a clearer picture of exactly
what stimulus characteristics are salient. For example, in the loggerhead experiments the
steel reinforcing bars in the tank walls created a disorderly ambient field and distorted
the gradient of the experimental field.
Additional issues in discrimination learning experiments using magnetic stimuli are

discussed in Ossenkopp and Barbeito (1978) and in the chapter on cetaceans (Bauer et 01.)
in this volume.
We are presently conducting a study on the resting orientation of green turtles in
response to altered magnetic field conditions. In these experiments, individual turtles are
placed in a small, indoor tank that is enclosed in black plastic to eliminate light and visual
distractions. Helmholtz coils outside the tank are supplied with power to alter the normal
magnetic field. Turtles are kept in the tank overnight in complete darkness, except for a
brief flash every 20 min to provide light for an overhead camera that records their positions.
The turtles are being tested for orientation responses to shifts in both the horizontal and
the vertical components of the magnetic field. Similar experiments conducted with eels
(Tesch, 1974) demonstrated a clear orientation shift with changes in the horizontal field
components. We hope that the results from these tests will be less ambiguous than those
obtained in our learning experiments, and that they will aid us in designing future ex-
periments to determine what aspects of the field might be relevant to migrational navigation
in the sea turtles.
3.2. Isolation and Identification of Magnetite in Sea Turtles
During the course of our operant conditioning experiments, we conducted magneto-

metry studies to look for magnetic material in Chelonia mydas. Specimens of hatchling,
juvenile, and adult green turtles were examined for magnetic remanence using SQUID
magnetometers. To reduce the risk of contamination, the magnetometer enclosure was
thoroughly cleaned, and only nonmagnetic glass and wood instruments were used in all
handling and dissection. Whole hatchlings and samples of tissue from all age classes were
rinsed with glass-distilled water, frozen in liquid nitrogen, and saturated with a 3000-G
cobalt/samarium magnet. Saturation isothermal remanent magnetization (sIRM) was then
measured in the SQUID. Background signal measurements were taken periodically for
comparison with tissue signals, and to ensure that the magnetometer enclosure remained
magnetically clean.
Magnetic remanence was found in the head region of all turtles examined (four hatchl-
ings, three juveniles, and two adults; Table II). The greatest concentration of magnetic
material occurred in the anterior portion of the dura mater, although it was also present
diffusely throughout the facial muscle. sIRM in the adult turtle duras was 9 x 10- 6 emu,
about 50 times greater than background. Because of this high remanence, dura tissue was
the primary material examined in further tests.
Remanence was also found in whole hatchlings. When these were further dissected,
the highest readings were obtained from the stomach; they probably came from magnetic
particles present in sand and dirt ingested by the turtles. Remanence associated with sur-
face tissues, such as carapace scutes, could be the result of external contamination and
was therefore not measured separately or examined in subsequent tests.
The sIRM measured in the turtle is within the range of that found in other vertebrate
species (see Bauer et 01.; Presti; Walker et 01., Chapter 20, this volume). Adult turtles contain
more magnetic material than juveniles, a phenomenon also observed in yellowfin tuna
(Walker et 01., Chapter 20, this volume), honeybees (Gould et 01., 1978), and woodmice
(J. Mather, personal communication).
Remanence in the dura was found from alternating-field demagnetization to be mag-
netically hard. The median unblocking field for net magnetic alignment in the dura was
approximately 225 G, which indicates that single-domain magnetite crystals are present
448 Chapter 21
Table II. sIRM in Three Age Classes of Green Turtles

Sample emu
Whole hatchlings 3.8 x 10- 6

8.7 X 10- 6
1.3 X 10- 5
2.3 X 10- 5
Hatchling heads 1.4 X 10- 6
1.6 X 10- 6
Hatchling necks 7.0 X 10- 8
8.3 X 10- 8
Juvenile heads 1.1 X 10- 6
3.0 X 10- 6
9.5 X 10- 6
Juvenile flippers
Front left 3.7 X 10- 7
Front right 2.8 X 10- 7
Rear left 9.7 X 10- 7
Rear right 6.3 X 10- 7
Adult brain 3.2 X 10- 7
6.0 X 10- 7
Adult eyeball 6.3 X 10- 7
Adult facial muscle blocks
1 x 1 x 1 cm 4.4 X 10- 7
1x1x2cm 5.5 X 10- 7
1x2x3cm 4.4 X 10- 6
1 x 3 x 3 cm 6.1 X 10- 6
Adult dura mater
Posterior 1.0 X 10- 6
Posterior 1.5 X 10- 6
Anterior 9.0 X 10- 6
Anterior 9.2 X 10- 6
Empty sample holder 4.2 x 10 -8_2.3 X 10- 7
In order to identify the source of magnetic remanence in turtle tissue, we extracted

and analyzed magnetic particles from adults and juveniles (for methods used see Walker
et 01., Chapter 5, this volume). Particle samples were examined via X-ray diffraction, elec-
tron microprobe analysis, EDAX line spectra analysis, and scanning electron microscopy.
The X-ray diffraction pattern produced by aggregated particles from turtle duras is
that expected for magnetite (Fig. 3). The lattice spacing parameter estimate calculated from
this pattern was 0.8375 ± 0.004 nm, which compares well to the reference parameter for
magnetite of 0.8396 nm, indicating that we had extracted a very pure magnetite sample
from the turtles. The number and relative intensities of rings in the pattern, specific for
each crystal type, also closely match the reference values (Joint Committee on Powder
Diffraction Studies, 1974).
Electron microprobe analysis revealed that the extracted particles were very rich in
iron and contained no measurable titanium or chromium, as would be expected for mag-
netite isolated from rock. Unlike geologic magnetite, small amounts of manganese and
calcium oxides were persistently associated with the crystals (Table III). Tissue residue
often found associated with the magnetite particles after extraction (see Walker et 01.,
Chapter 5, this volume) may account for the presence of these oxides. Analysis by ED AX
probe corroborated the microprobe data, exhibiting strong peaks only for iron, and showing
Figure 3. X·ray diffraction pattern of magnetite extracted from green turtle dura maters.
Table III. Electron Microprobe Analysis of

Magnetite Particles Isolated from Green
Turtles
Weight % of sample
Magnetite stan-
dard (NMNH
Oxide 11487) Turtle
FeO 90.9 85.5 ± 1.7

Ti0 2 0.2 0.0 ± 0.0
Cr Z 0 3 0.2 0.0 ± 0.0
MnO 0.0 0.3 ± 0.0
CaO 0.3 ± 0.1
Total 91.3 86.1
450 Chapter 21
Figure 4. EDAX spectrum for magnetite

sample from green turtle dura maters .
Numbered peaks correspond to the follow-
ing elements: 1, silica (from glass coverslip
sample was placed on); 2, gold (from goldl
paladium coating sprayed on sample to in-
6 crease conductivity for scanning electron
microscopy) ; 3, chlorine (from bleach used
L_........___.._ ....L _ _ _ _... to digest dura membrane tissue); 4, palad-
1. ium (from goldlpaladium coating) ; 5 and 6,
iron.
Figure 5. Scanning electron micrograph of colloidal crystal structure from green turtle dura mater
sample.
Figure 6. Scanning electron micrograph of sphere from green turtle dura mater sample.
no presence of other metals that would be expected as geologic contaminants, including

nickel (Fig. 4).
Scanning electron microscopy revealed two types of crystal structure in dura samples:
colloids (Fig. 5), and spheres (Fig. 6). The colloidal structures may comprise clusters of
single-domain magnetite crystals. Although the apparent size of each nodule in the cluster
is larger than that which would produce single-domain behavior in magnetite, each crystal
may be surrounded by an organic coat. The spheres of magnetite, such as the one depicted
in Fig. 6, have not previously been observed in organisms, although magnetite-containing
cosmic spherules are common (Parkin et 01.,1977). It is possible that the turtle dura spheres
are not of biological origin, but are contaminating cosmic spherules. We believe, however,
that the spheres we observed are the result of biologic precipitation rather than cosmic
contamination for several reasons . Our sample preparations were carried out in two dif-
ferent laboratories at different times , and as cosmic spherules are not a common laboratory
contaminant it seems unlikely that they would appear in both sample sets. Cosmic spher-
ules usually contain high proportions of magnesium and nickel, neither of which were
present in our samples. Our strongest evidence for in situ information was that the spheres
were not seen in control samples or in tissues other than dura. At this time the function
of the large (10-50 11m) spheres is unknown. Dr. J. L. Kirschvink speculates that they may
be involved in a system that detects geoelectric fields (Cromie, 1982).
The green turtle is a threatened species, and we therefore use only specimens that
have died from natural causes in our magnetometry studies. As specimens become avail-
able, we hope to isolate more spheres from the duras and to study their properties in greater
detail. We can then attempt to elucidate the role the spheres play, if any, in turtle mag-
netoreception.
452 Chapter 21
4. Conclusion
The amphibians and reptiles form an evolutionary link between higher and lower
vertebrates. Because of this phylogenetic position, the examination of magnetic perception
and the sensory apparatus responsible for such perception is of particular interest in these
classes. Evidence is accumulating for magnetoreception and magnetite biomineralization
among fish, mammals, and birds. Preliminary work on magnetic susceptibility in sala-
manders and turtles, and the discovery of magnetite in turtles, is an important step toward
understanding the evolutionary processes involved in magnetic sensitivity in all verte-
brates. As demonstrated by this chapter, research on magnetoreception in the amphibians
and reptiles has just begun. Additional studies are needed if the role of magnetite and
magnetic sensitivity in vertebrates as a group is to be clearly understood.
ACKNOWLEDGMENTS. We wish to thank William G. Gilmartin and the National Marine Fish-
eries Service for providing turtles, equipment, and research facilities. John McIntosh, Kath-
leen Sobon, and Catherine Uyehara gave freely of their time to help in carrying out the
behavioral experiments. Personnel of the Hawaii Institute of Geophysics, University of
Hawaii, provided paleomagnetic laboratory facilities and assistance with electron mi-
croscopy and diffraction analyses. Joseph L. Kirschvink introduced us to the use of pa-
leomagnetic techniques to look for magnetite in organisms, and gave us much-needed
advice and instruction. Research funds were generously provided by the Southwest Fish-
eries Center Honolulu Laboratory, National Marine Fisheries Service. The loggerhead stud-
ies were supported in part by Sea Grant Project NA79AA-D-00085.
References
Adler, K., 1982, Sensory aspects of amphibian orientation and navigation, Vertebr. Hung. 21:7-18.
Balazs, G. H., 1976, Green turtle migrations in the Hawaiian archipelago, BioI. Conserv. 9:125-140.
Baldwin, H. R., 1972, Long-range radio tracking of sea turtles and polar bear: Instrumentation and
preliminary results, in: Animal Orientation and Navigation (S. R. Galler, K. Schmidt-Koenig, G.
J. Jacobs, and R. E. Belleville, eds.), NASA SP-262, U.S. Government Printing Office, Washington,
D.C., pp. 19-37.
Carr, A., and Carr, M. H., 1970, Modulated reproductive periodicity in Chelonia, Ecology 5:335-337.
Cromie, W. J., 1982, Born to navigate, Mosaic July/August:17-23.
Dunson, W. A. (ed.), 1975, The Biology of Sea Snakes, University Park Press, Baltimore.
Ferguson, D. E., 1971, The sensory basis of orientation in amphibians, Ann. N.Y. Acad. Sci. 188:30-
36.
201:1026-1028.
Hirth, H. F., 1971, Synopsis of Biological Data on the Green Sea Turtle (Chelonia mydas Linnaeus
1758), FAO Fish. Syn. No. 85.
Joint Committee on Powder Diffraction Studies, 1974, Selected Powder Diffraction Data for Minerals,
Swathmore, Pa.
Keeton, W. T., 1971, Magnets interfere with pigeon homing, Proc. Natl. Acad. Sci. USA 68:102-106.
Lett. 44:193-204.
Manton, M., Karr, A., and Ehrenfeld, D. W., 1972, An operant method for the study of chemoreception
in the green turtle, Chelonia mydas, Brain Behav. Evol. 5:188-201.
Minton, S. A., Jr., and Minton, M. R., 1973, Giant Reptiles, Scribner's, New York.
Ossenkopp, K.-P., and Barbeito, R., 1978, Bird orientation and the geomagnetic field: A review, Neu-
Parkin, D. W., Sullivan, R. A. L., and Andrews, J. N., 1977, Cosmic spherules as rounded bodies in
space, Nature 266:515-517.
Perry, A., 1982, Magnetite and magnetic sensitivity in the green turtle, Chelonia mydas, Master's thesis,
Department of Oceanography, University of Hawaii, Honolulu.
Phillips, J. B., 1977, Use ofthe earth's magneticfield by orienting cave salamanders (Eurycea lucifugal,
J. Compo Physisol. A 121:273-288.
Keeton, eds.l, Springer-Verlag, Berlin, pp. 325-333.
Pritchard, P. C. H. 1973, International migrations of South American sea turtles (Cheloniid and Der-
mochelididl, Anim. Behav. 21:18-27.
salmon fry, J. Compo Physiol. A 137:243-248.
Schmidt-Koenig, K., 1975, Migration and Horning in Animals, Springer-Verlag, Berlin.
Tesch, F. W., 1974, Influence of geomagnetism and salinity on the directional choice of eels, Helgol.
Wiss. Meeresunters. 26:382-395.
Twitty, V. C., Grant, D., and Anderson, 0., 1967, Initial homeward orientation after long-distance
displacements in the newt Taricha rivularis, Proc. Natl. Acad. Sci. USA 57:342-348.
Woodward, W. T., and Bitterman, M. E., 1974, A discrete-trialslfixed-interval method of discrimination
training, Behav. Res. Methods Instrum. 6:389-392.
Chapter 22
Avian Navigation, Geomagnetic Field
Sensitivity, and Biogenic Magnetite
DAVID E. PRESTI
1. The Sensory Basis of Bird Navigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 455

2. Orientation Experiments with Homing Pigeons . . . . . . . . . . . . . . . . . . . . . . . . .. 459
2.1. Imposed Magnetic Fields during Flight. . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
2.2. Imposed Magnetic Fields during Transport to Release Site . . . . . . . . . . . . . . .. 462
3. Orientation Experiments with Migratory Birds. . . . . . . . . . . . . . . . . . . . . . . . . .. 464
4. Effects of Small Magnetic Field Changes on Navigation: The Possibility of a Geomagnetic
Map. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 469
5. Laboratory Attempts to Measure Avian Magnetic Field Sensitivity. . . . . . . . . . . . . . . 472
6. Magnetite in Birds and Possible Mechanisms of Magnetic Field Sensitivity. . . . . . . .. 474
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 477
1. The Sensory Basis of Bird Navigation

How birds manage to find their way around in the world has long been a source of fas-
cination. These animals range widely when foraging for food, sometimes traveling 100 km
or more from their nest or roost. Moreover, many species of birds fly long distances between
breeding grounds and wintering areas, returning with extraordinary precision to specific
locales year after year. Distances traveled during such migrations typically average between
1000 and 6000 km each way between breeding and wintering grounds. Several species of
birds travel close to 30,000 km annually between arctic breeding grounds and wintering
areas in or near the antarctic. It would seem that birds, as well as other animals which
forage and migrate over large distances, must possess navigational abilities of a fair degree
of sophistication. Indeed, given the essential importance to such animals of an accurate
and fail-safe navigational system, one might not be surprised to find that evolution has
been opportunistic in exploiting available environmental sources of navigationally useful
information.
Experiments in which wild birds-ranging from sparrows to albatrosses-have re-
turned to their home territories after capture and artificial displacement over distances of
up to several thousand kilometers serve as further illustrations of the effectiveness of avian
navigation (Griffin, 1944; Matthews, 1953b; Mewaldt, 1964; Keeton, 1974a). This so-called
"homing" behavior of birds has been most extensively studied in the modern-day homing
pigeon l!':eeton, 1974b; Walcott, 1974). These birds derive from the Old World rock dove
(Columba livia) and have been selectively bred for an improved ability to return home
quickly after being taken many kilometers away and released. Such pigeons can, for ex-
DAVID E. PRESTI • Departments of Biology and Psychology, University of Oregon, Eugene, Oregon
97403.
455
456 Chapter 22
11:40
10:25
Figure 1. Homing behavior of a pigeon

wearing frosted contact lenses so that
no visual patterns were discernible.
Upon release the bird flew toward its
home loft and spent more than 1 hr cir-
cling that area, approaching several
10:15 times to within 1 Ian of the loft. Fol-
lowing this frustrated effort to locate
home, it turned and flew back in the
direction from which it had come.
5 km Points represent positions located by ra-
diotelemetry. Numbers indicate times
10:03 for some of the tracking fixes. From
x- RELEASE SITE Schmidt-Koenig and Walcott (1978).
ample, be taken hundreds of kilometers away from their home (under conditions in which
they cannot see where they are being taken) to a place where they have never before been;
when released, a well-bred and properly cared-for homing pigeon will select the proper
direction and begin flying toward home within several minutes.
For many years it was thought that bird navigation could be explained on the basis
of random (or calculated) searching for familiar visual landmarks-and, perhaps, the as-
sociation of geographical features (such as mountain ranges, coastlines, and rivers) with
the directions of certain celestial cues (such as sunrise and sunset). Much of the evidence
for navigational capabilities in birds seemed to be anecdotal at best and certainly not the
result of carefully designed experimentation, or even careful interpretation of otherwise
clear experimental results. However, over the past three decades or so, a variety of ex-
periments with both migratory birds and homing pigeons have demonstrated the presence
of remarkable navigational capabilities (Keeton, 1974a; Emlen, 1975a). Young pigeons, for
example, that have never previously been taken away from their home loft, have been
shown to orient toward home within a minute of their release at a site 100 km away (e.g.,
Wallraff, 1970). Such a result indicates that these birds cannot be flying in a visual search
pattern until they encounter familiar landmarks. Moreover, homing pigeons fitted with
frosted contact lenses, so that objects farther away than several meters cannot be discerned,
still orient toward home when released at distant sites (Schmidt-Koenig and Schlichte,
1972; Schmidt-Koenig and Walcott, 1978). Even more strikingly, many of these birds man-
age to fly to within 1 km of their home loft (Fig. 1), sometimes landing nearby. This clearly
indicates that pigeons possess a navigational system capable of phenomenal precision
which does not depend upon the visual perception of landmarks at all!
Birds 457
In the early 1950s, Gustav Kramer, a pioneer of modern experimentation in bird nav-
igation, hypothesized that navigation might best be considered as a two-step process, con-
sisting of map and compass components. According to Kramer's map-and-compass con-
cept, a bird, in deciding which way it wanted to fly, would first have to know where it
was relative to where it wanted to go. The determination of its current position relative
to the goal constitutes the map (positional) step of Kramer's model. Second, such a bird
would perform a compass (directional) step to determine how the goal direction related
to the geophysical environment. This map-and-compass hypothesis (to describe bird nav-
igation) derives from an analogy with human navigation. A human navigator uses a map
to place his location relative to that of a goal, determines the relative direction of the goal
from the map, then utilizes a compass (be it solar, stellar, geomagnetic, or inertial) to
associate the direction obtained from the map with a direction in the environment.
In a beautiful series of experiments conducted during 1950 and 1951, Kramer showed
that birds utilize the position of the sun as a source of directional information (Kramer,
1952). He monitored the directional preference of activity which individual songbirds
(European starlings, Sturn us vulgaris) exhibited when kept in circular cages during their
spring and autumn periods of migration. Kramer and his colleagues also devised a paradigm
in which birds were trained to go to a feeder in a particular compass direction, placed
among many identical feeders around the periphery of a circular cage. By manipulating
the apparent position of the sun with mirrors and by using artificial light sources, they
showed that birds were determining compass directions by observing the sun's position.
In order to do this, birds must accurately know the time of day, for only then can solar
position be converted into a compass direction. This discovery of the ability of birds to
use the sun as a compass was, in fact, one of the timely events which led to the development
of interest in the study of biological clocks during the 1950s. Experiments over the next
several years demonstrated that migratory birds and homing pigeons whose endogenous
biological clock had been phase-shifted by exposure to an altered light-dark cycle err in
choosing the proper compass direction when tested behaviorally (Hoffmann, 1954, 1960;
Schmidt-Koenig, 1960). The magnitude and direction of the error corresponds to the rate
of the sun's movement across the sky during the day (=15°/hr). Such clock-shift experi-
ments suggest that the information-processing strategy used by birds when navigating
contains a distinct compass component, which can be manipulated without affecting the
bird's position-finding ability (map component). This is in agreement with Kramer's map-
and-compass hypothesis.
Many birds migrate at night and thus cannot use the sun as a source of directional
information. Such birds do, however, utilize the stars as a compass. Some species determine
compass direction by assessing the center of rotation of the starry sky, an indicator of the
north direction in the northern hemisphere (Emlen, 1975a,b). These birds then learn star
patterns (presumably akin to what are called constellations by humans) associated with
the rotational axis and thereafter rely on the recognition of the§e patterns to determine
compass direction. This provides a compass reference that is independent of time. There
is also evidence that some birds may use the direction of the setting sun as a source of
compass information when they set out on migratory flights at dusk (Moore, 1978, 1980).
Although it is well established that celestial cues provide birds with information re-
garding compass direction, how birds actually know where they are-that is, what con-
stitutes their map-has remained more of a mystery. It was recognized long ago that the
positions of the sun (Matthews, 1953a) and the stars (Sauer, 1957), used in conjunction
with an accurate chronometer, are also capable of providing birds with positional infor-
mation. The altitude of the sun above the horizon at a given time of the day varies as a
function of latitude on the earth's surface. Thus, for example, in order to determine its
latitude, a bird could extrapolate the sun's arc of motion across the sky to its highest point
(which occurs at local noon time) and compare the measured altitude with a remembered
458 Chapter 22
CONTROL BIRDS CLOCK - SHIFTED

NON-SHIFTED 6 HOURS AHEAD
A Figure 2. Clock-shift experiments with homing pi-

geons. Open circles indicate visual vanishing bear-
ings of control birds with a history of exposure to
normal day-night cycle. Solid circles indicate van-
ishing bearings of birds who have had their endo-
E ú ú F = CJ
úúl = úú= genous biological clock phase-shifted 6 hr ahead by
exposure to an altered day-night cycle. Vanishing
bearings are determined by observing (with bin-
B oculars) the direction in which a bird disappears
(to the nearest 5°) following individual releases at
a distant site. The direction of home is indicated
by a dashed line. Each symbol on the periphery of
a circle graph indicates the bearing of one bird. The
0fj} ° ú = mean of these bearings is denoted by an arrow
c ú ` g í =s
If : °
00
(2)':
,
•
r .....
V
:.
whose length is inversely proportional to the extent
of scatter. (A) Birds released under sun; (B) birds
released under total overcast at sites they may have
seen before; (C) birds released under total overcast
° ° • • at unfamiliar sites. From Keeton (1974).
noon altitude at home or at another goal. Longitudinal information could be determined

by comparing local sun time with home time as indicated by the bird's internal clock.
Human navigators, in fact, use a sextant to accurately measure the altitude of the sun or
specific stars; this information is then used in conjunction with a chronometer to determine
position. As attractive as this possibility may sound for bird navigation, a variety of ex-
periments (such as clock-shift experiments like those described above) have nevertheless
demonstrated that birds do not use the sun and stars as sources of map (positional) in-
formation, but only as indicators of compass dirction (Keeton, 1974a; Emlen, 1975a).
Navigation by celestial cues of course requires that the sky be sufficiently free of clouds
for the sun or stars to be located. However, it has been known for some years that migratory
orientation occurs under overcast conditions as well. Radar observations also indicate that
migratory birds can be well oriented even when flying between thick layers of cloud (Grif-
fin, 1972, 1973). Finally, following decades of belief that homing pigeons simply cannot
orient under overcast conditions (a belief held within the academic community, not, how-
ever, among racing pigeon fanciers), a series of experiments reported in 1969 by Keeton
established definitively that properly trained and motivated homing pigeons are able to
orient toward home from an unfamiliar location on overcast days. Moreover, Keeton (1969)
obtained another result that generated great excitement among investigators of avian nav-
igation: phase-shifting of a bird's internal clock does not affect its orientation on overcast
days as it does when the sun is visible (Fig. 2). Thus, the method used by these birds to
orient on overcast days does not require time-compensation as does the sun compass.
The old concept that birds navigate exclusively via the recognition of familiar visual
landmarks has been replaced by the idea that a number of different environmental stimuli
and information-processing strategies may be important for navigation (Keeton, 1974a;
Emlen, 1975a; Able, 1980). The sun, the stars, and topographic landmarks are well known
to be used by birds when navigating. The discovery that at least some birds can perceive
Birds 459
ultraviolet light (Wright, 1972; Kreithen and Eisner, 1978; Goldsmith, 1980; Parrish et 01.,
1981) and the polarization of linearly polarized light (Kreithen and Keeton, 1974b; Deluis
et 01., 1976) raises the possibility that they might use skylight polarization patterns (which
are most intense in the ultraviolet region of the spectrum) as an aid to navigation, a process
known to occur in a number of invertebrate species (von Frisch, 1967; Waterman, 1981).
Recently, it has been found that pigeons are capable of perceiving sounds having fre-
quencies as small as 0.1 Hz (Yodlowski et 01.,1977; Kreithen and Quine, 1979). It is thus
possible that pigeons use constant environmental sources of very low-frequency sound
(waves breaking on a seacoast, winds over a mountain range) to construct a navigationally
useful map of their environment. Although such very low-frequency (and therefore very
long wavelength) sound cannot be localized by the usual method of measuring binaural
differences in phase and arrival time, its direction could, in principle, be ascertained by
measuring frequency Doppler shifts while flying in various directions (Quine and Kreithen,
1981). Pigeons are also capable of perceiving extremely small changes in ambient air pres-
sure, equivalent to pressure differentials experienced during altitude changes of as small
as 10 m (Kreithen and Keeton, 1974a). Windborne olfactory cues have also been considered
as possible indicators of location and direction, a proposal which has lately generated
much experimentation and discussion (papi, 1976, 1982; Wallraff, 1980b; Gould, 1982;
Wallraff et 01., 1982).
Human navigators have for nearly a thousand years used magnetic compasses to de-
termine direction. In principle, birds, too, as well as other animals, could obtain directional
(compass-type) information from the geomagnetic field. Moreover, the earth's magnetic
field may also be capable of providing positional (map-type) information. The intensity
and inclination of the geomagnetic field vary as a function of position on the earth's surface,
possibly in a regular enough fashion to be of use to navigating animals.
It is clear that the sensory worlds of other animals are vastly different from our own
such world and that any of a number of environmental stimuli, to which humans might
be insensitive, may play important roles in avian navigational behavior.
2. Orientation Experiments with Homing Pigeons

For at least a century the suggestion that the earth's magnetic field might provide
information to navigating birds has occurred in the scientific literature. Viguier, in 1882,
proposed that birds might construct a navigational grid based on the intensity and the
inclination of the geomagnetic field. The idea that birds might navigate using a bicoordinate
grid system composed of intersecting lines of magnetic latitude (as measured, for example,
by the intensity of the vertical component of the geomagnetic field) and lines of geographic
latitude (as measured by the magnitude of the Coriolis force generated by the earth's ro-
tation) was set forth by Yeagley over 35 years ago (Yeagley, 1947). [The idea that Coriolis
forces might be utilized by navigating birds was first proposed by Gustaf Ising (1945), a
Swedish physicist.]
Yeagley's proposal of geomagnetic-Coriolis bicoordinate navigation was based on pi-
geon homing experiments begun in the early 1940s. In one such experiment, Yeagley at-
tached small permanent magnets to the wings of one group of homing pigeons and non-
magnetic copper plates of approximately the same size and mass to the wings of another
group of pigeons. He reported that the homing performance of the birds carrying magnets
was significantly poorer (Yeagley, 1947). In another series of experiments, birds raised and/
or trained at a home loft in Pennsylvania were displaced 1800 km west to the vicinity of
a "conjugate" point in Nebraska where the combination of geographic and magnetic lat-
itudes was the same as at the home loft. When released in Nebraska, Yeagley reported that
460 Chapter 22
BRASS BARS MAGNETS
(!)
yúúú=
A
:C) "!
.
1 ...
.
Figure 3. Effect of magnets on pigeon homing. Open cir-

cles indicate visual vanishing bearings of control birds
wearing small brass bars. Solid circles indicate vanish-
ing bearings of birds wearing small magnets. (A) Expe-
rienced birds released under sun at 27-50 km; (B) ex-
c perienced birds released under total overcast at 27-50
km; (C) first-flight young birds released under sun at 27
km. From Keeton (1974).
these birds flew preferentially toward the nearby (but unfamiliar) conjugate point, rather
than in the direction of their home (Yeagley, 1947, 1951). The implication is that the birds
were utilizing a measure of the geographic and magnetic latitudes and were thus misled
(by the proximity of the conjugate point) into believing that home was nearby, rather than
1800 km to the east.
Yeagley further proposed that pigeons (and other birds) might detect the geomagnetic
field intensity by measuring the magnitude of an electric voltage induced as they fly
through the earth's field. His detection mechanism was promptly and severely attacked
on theoretical grounds by several people (see Griffin, 1952). Moreover, other experiments
(Gordon, 1948; Matthews, 1951; van Riper and Kalmbach, 1952), including one by Yeagley
(1951) himself, failed to reproduce the result that magnets attached to the wings of pigeons
disrupt their navigational ability (Griffin, 1952). Although Yeagley's ideas generated much
discussion, the overall negative reception which they received probably contributed to
discouraging further serious consideration of a role for the geomagnetic field in bird nav-
igation for a number of years. A role for the geomagnetic field in animal navigation was
apparently an idea whose time had not yet come.
2.1. Imposed Magnetic Fields during Flight
Two decades after Yeagley's original publication, the striking finding (Keeton, 1969)
that under conditions of heavy overcast pigeons were well oriented and, moreover, not
affected in their orientation by clock-shifts, led Keeton to reconsider the possibility that
the geomagnetic field might be providing directional information to navigating birds-at
least in the absence of such information from the sun. Keeton (1971) found that small
magnets attached to experienced homing pigeons (to the back, at the base of the neck;
producing a field of about 0.5 G at the bird's head) resulted in disorientation when the
birds were released from distances of 27-50 km under conditions of total overcast. No
such disorientation occurred during similar releases when the sun was visible (Fig. 3A,B).
Birds 461
Figure 4. Vanishing bearings (determined by ra-

diotelemetry) of homing pigeons wearing Helm-
holtz-type coils around the head and neck.
"North-down" indicates that the applied field NORTH-DOWN NORTH-UP
through the head was such that the north-seek- COILS COILS
ing end of a compass needle pointed down. KKI ú =
"North-up" indicates the opposite field config-
uration produced by reversing the direction of
current through the coils. Birds released on
sunny days vanished in the homeward direction
independent of the applied magnetic field con-
SUN ôúF=
figuration. Birds released on overcast days and
wearing the "north-down" field configuration
vanished in the homeward direction (now with
increased scatter), while birds wearing the OVERCAST
"north-up" field configuration vanished in the
opposite direction. From Walcott and Green
(1974).
Interestingly, although the initial orientation (as measured by vanishing bearings) of the
birds was dramatically affected by the presence of magnets on overcast days, many of these
birds nevertheless got home, sometimes just as rapidly as the control birds without mag-
nets. It thus appears that whatever disturbance the magnets produce, the birds are able to
compensate for it relatively quickly-correcting their orientation even though the magnets
remain attached to their bodies.
Walcott devised an elegant refinement of the paradigm in which magnets were attached
to birds. He fitted pigeons with a pair of Helmholtz-type coils around the head and neck,
such that a magnetic field of approximately 0.6 G was produced through the bird's head
and neck when the coils were supplied with current from an attached battery. The applied
field could be in either of two directions, depending on the direction of current flow in
the coils. When released under sunny skies at an unfamiliar site, birds with an applied
field in either of the two directions oriented homeward (Walcott and Green, 1974; Walcott,
1977). However, when released under total overcast, one group of birds (with the applied
field through the head such that the north-seeking end of a compass needle pointed down)
oriented homeward, while the other group (with the applied field through the head such
that the north-seeking end of a compass needle pointed up) vanished in a direction nearly
straight away from home (Walcott and Green, 1974; Visalberghi and Alleva, 1979 (Fig. 4).
These results are consistent with a model of geomagnetic field navigation, proposed by
Wiltschko and Wiltschko (1972) and further described in Section 3, in which the north
direction is taken to be that in which the magnetic field vector makes the smallest angle
with gravity. Birds wearing a coil with the north-down configuration would find the net
vector resulting from the combination of the geomagnetic field with the coil field to be
most steeply inclined when flying north, as occurs naturally in the absence of an applied
field from the coil. On the other hand, birds wearing a coil with the north-up configuration
would find the resultant magnetic field vector to be most steeply inclined when flying
south, the opposite of the natural situation.
Although the effects of attached magnets and Helmholtz-type coils on the initial ori-
entation of homing pigeons are most pronounced on overcast days, some disorientation
of experienced birds is produced even when the sun is visible. Keeton (1971), for example,
reported that attached magnets affected vanishing bearings and homing speeds of expe-
rienced pigeons which were released under sunny skies at a site more distant from home
(85 km) than the other release sites (27-50 km) in his study. Walcott, in an extensive series
of sunny-day releases conducted during the early 1970s, found a significant increase in
462 Chapter 22
the scattter of vanishing bearings for experienced pigeons wearing Helmholtz-type coils
which produced magnetic fields of between 0.1 and 0.6 G in intensity (Walcott, 1977).
Similar results have been obtained by Visalberghi and Alleva (1979). Such results might
indicate that the geomagnetic compass is still being used to some extent and is capable of
being affected even in experienced birds on sunny days, or they may indicate that unnatural
magnetic fields are interfering with some other aspect of the avian navigation system-
perhaps a geomagnetic map.
Whereas attached magnets had no effect on the orientation of experienced homing
pigeons (released less than 50 km from home) on sunny days in Keeton's (1971) experi-
ments, inexperienced birds (young birds being taken away from home for the first time)
released under clear skies and carrying magnets were disoriented (Fig. 3C). More recent
experiments have shown that the sun compass in pigeons is learned rather than innate
(Wiltschko et 01., 1976; R. Wiltschko and Wiltschko, 1981), and suggest that young birds
are able to utilize the geomagnetic field as a compass prior to their developing the ability
to use the sun as a compass (Wiltschko et 01., 1981). Thus, although the sun may be the
preferred source of compass information for experienced birds-as indicated by the results
of clock-shift experiments-the geomagnetic field appears to be the primary source of
compass information used by young, inexperienced pigeons.
Wiltschko et 01. (1983) have raised young homing pigeons in an altered magnetic
environment (horizontal field component rotated by approximately 65 or 120°) in order
to ascertain whether the sun compass is calibrated against the magnetic compass during
ontogeny. Pigeons raised under such conditions and released for the first time under sunny
skies exhibited vanishing bearings different from those of control birds raised in the normal
geomagnetic field. On subsequent releases the differences in orientation between exper-
imental and control birds changed in various ways. The results of Wiltschko et 01. (1983)
suggest an involvement of the magnetic compass in the established of the sun compass in
pigeons, but the precise relationship between the two systems is not yet clear.
The strategy used by a bird to process compass information might well be simpler for
geomagnetic field information, which remains constant in time, as opposed to solar in-
formation, which requires the monitoring of a constantly changing stimulus (the sun's
position in the sky) and compensation for the change using the endogenous biological
clock. The additional complication of time-compensation might account for the ontoge-
netic priority of the geomagnetic compass over the sun compass in pigeons.
2.2. Imposed Magnetic Fields during Transport to Release Site
It has often been wondered whether homing in pigeons depends upon information
collected by the birds during the displacement from their home to a distant release site.
Pigeons could, for example, monitor changes in linear acceleration in conjunction with
changes á ú =angular acceleration or changes in the direction of the geomagnetic field vector,
and by integration of these changes determine their position relative to the starting point
of their journey. Such birds might also monitor other environmental stimuli during the
displacement journey-local odors, for example. Several recent studies suggest that the
magnetic field experienced during the outward journey from home can affect the orien-
tation of homing pigeons.
Papi et 01. (1978) and Benvenuti et 01. (1982) transported pigeons in iron containers
which reduce the intensity of the geomagnetic field to about 0.5% of its normal value.
Control birds were transported in nonmagnetic aluminum containers which did not distort
the geomagnetic field but which affected other potential cues (visual, olfactory, auditory)
in the same way as the iron containers. While the control birds were well oriented toward
home upon release, the pigeons that had traveled in the iron containers were less well
.0;.°0
. ..
Birds 463
00
Figure 5. Effect of manipulation during transport to release site 0: .:..

' •..•
u
Q f
on the initial orientation of inexperienced homing pigeons. Open i
.:.
.
\
circles denote visual vanishing bearings of control birds. Solid ú I =
circles denote vanishing bearings of birds which experienced a J' . . . . ., ...:.
distorted magnetic field during transport to the release site. A
Home direction is indicated by a dashed line. (A) Experimentally
manipulated birds experienced an additional horizontal mag-
netic field. Each circle graph is the summary of four release ex-
periments from the same site, located 65 km from the home loft. çú= I
Birds were released shortly after arriving at the release site. From ,j • II
W. Wiltschko and Wiltschko (1981). (B) Experimentally manip- '. ú=0 0°·
ulated birds experienced an additional horizontal magnetic field,

W8re in closed containers, were rotated at 1.5 rev/sec, and were
supplied with bottled air. This treatment effectively prevented
the acquisition of any potential navigational information which
might be based on geomagnetic, visual, inertial, or olfactory stim-
uli. Release took place the day following the displacement jour-
ney. There were four separate release experiments from four dif-
B
ferent sites. Each site was located in a different cardinal direction
approximately 180 km from home. From Wallraff (1980a).
oriented and generally did not depart in the homeward direction. Homing performance
(Le., the percentage of birds reaching home and their average speeds of return), however,
did not differ significantly between the control birds and the birds that had experienced
an altered magnetic field during their journey to the release site.
Benvenuti et al. (1982) also exposed pigeons during displacement to a distorted mag-
netic field generated by three orthogonal pairs of Helmholtz-type coils. Polarity and in-
tensity (ranging between 0 and 0.6 G) oscillated sinusoidally with periods between 15 and
39 sec, different for the different pairs of coils. Control birds were homeward-oriented
upon release, while birds exposed to the oscillating magnetic field were poorly oriented
in nonhomeward directions. Again, homing performance was not affected.
Using a single pair of Helmholtz-type coils, Kiepenheuer (1978a,b) exposed homing
pigeons during displacement to an ambient magnetic field with a vertical component op-
posite to that of the normal geomagnetic field. He found a slight effect of this treatment
on the vanishing bearings of both experienced and inexperienced birds: a rotation of the
average vanishing direction about 30° clockwise relative to that of the control birds. Homing
performance was not significantly affected.
Wiltschko and Wiltschko (1982) report that the less homing experience a bird has, the
more likely it is to be disoriented by manipulation of the ambient magnetic field during
transport to the release site. Moreover, even among inexperienced birds, those younger
than 11-12 weeks were more disoriented than older pigeons. Working with young «12
weeks of age) inexperienced homing pigeons, the Wiltschkos have measured a dramatic
effect on the initial orientation of birds exposed to altered magnetic fields during trans-
portation to the release site (Wiltschko et a1., 1978; Wiltschko and Wiltschko, 1978; W.
Wiltschko and Wiltschko, 1981). They found that birds subjected to an additional hori-
zontal magnetic field (produce by a pair of Helmholtz-type coils) departed randomly when
released at a site 65 km from home (Fig. 5A). [W. Wiltschko and Wiltschko (1981) also
found that transportation to the release site in total darkness-while exposed to the normal
geomagnetic field-also results in random vanishing bearings for inexperienced homing
pigeons.) Homing performance was not evaluated because the distance proved to be too
large for the young inexperienced birds to negotiate. Birds which experienced the normal
geomagnetic field during their displacement journey from home and then were exposed
464 Chapter 22
to an altered magnetic field for 30 min at the release site prior to their release were not
disoriented.
Results such as these suggest that homing pigeons might use geomagnetic field in-
formation collected during the displacement journey to tell them where they have been
taken. However, the fact that homing performance (in those cases where it could be as-
sessed) was not impaired in birds exposed to various sorts of distorted magnetic fields
during displacement-even though initial orientation was-indicates that the birds very
soon realize where they are and navigate home appropriately. The results are somewhat
suggestive of a temporary effect on a geomagnetic compass, like that produced by an at-
tached magnet on an overcast day.
In a very extensive series of experiments, Wallraff (1980a) and Wallraff et 01. (1980)
have strengthened the contention that information collected during displacement is not
necessary for homing in pigeons. Inexperienced and experienced birds were transported
to release sites in an apparatus designed to prevent the acquisition of any potentially useful
navigational information. The apparatus consisted of a closed (but lighted) container which
was continuously rotated quite rapidly (1.5 rev/sec). Two pairs of Helmholtz-type coils
generated additional horizontal magnetic fields and the birds were supplied with bottled
air to prevent the perception of olfactory information during the outbound journey. The
sense of rotation of the container and the configuration of the magnetic field were changed
throughout the journey at random intervals between 0.5 and 9 min. Control birds were
transported at the same time on the outside of the vehicle with complete access to the
normal geomagnetic field, as well as visual, olfactory, and inertial cues. Upon arrival at
the release site, birds which had been enclosed in the rotating containers were under-
standably quite upset. Something akin to motion sickness was likely their major complaint.
In contrast to the experiments described above, Wallraff allowed his birds to rest-and
perhaps adjust to their new location-until the following day; only then were the birds
individually released. The results are definitive: there were no differences in navigational
behavior between the experimentally manipulated and the control birds-assessed by vis-
ual vanishing bearings (Fig. 5B), homing performance, and the distribution of recoveries
of birds which did not make it the entire distance home. Wallraff's results clearly indicate
that pigeons (experienced and inexperienced) can home perfectly well from distant sites
even when prevented from collecting navigationally useful information-geomagnetic, vis-
ual, inertial, and olfactory-while being transported from home to the release site.
3. Orientation Experiments with Migratory Birds

In one of the few field studies dealing with magnetic effects on avian orientation not
done with homing pigeons, Southern (1972b) measured the vanishing bearings of fledgling
ring-billed gulls (Larus deIawarensis) released 29 km west of their colony. Control birds
generally disappeared to the south or southeast, the normal autumn migratory direction,
whereas gulls with small magnets attached to the tops of their heads vanished randomly.
The disorienting effect of the magnets in these experiments seemed to be just as pronounced
on sunny days as on overcast days. This is analogous to the disorientation experienced
by young first-flight homing pigeons carrying magnets on sunny days.
The methods pioneered by Kramer for the study of orientation in caged birds have
also been applied to the investigation of the geomagnetic field as a source of orientational
information in migratory songbirds. Experiments by Merkel and Wiltschko (1965;
Wiltschko, 1968) during the 1960s showed that the European robin (Erithacus rubecula),
a nocturnally migrating songbird, exhibits oriented nocturnal activity (called Zugunruhe
or migratory restlessness)-with direction appropriate to the seasonal migratory direc-
Birds 465
tion-in a cage isolated from visual environmental cues (stars and landmarks) but exposed
to the normal geomagnetic field. When the horizontal component of the ambient magnetic
field was changed in direction using Helmholtz-type coils encompassing the cage, the
preferred direction of the birds' activity also changed in a corresponding manner. Wiltschko
(1972) also reported that migratory activity of European robins became disoriented if the
magnetic field at the test cage, which had a normal intensity of 0.46 G, was artificially
decreased to 0.34 G or lower or increased to 0.64 G or higher. However, if the birds were
exposed to the abnormal field intensity for 3 days or longer, they regained the ability to
orient in the appropriate migratory direction in the abnormal intensity, while also retaining
the ability to orient properly in the normal intensity (Wiltschko, 1978). Interestingly, birds
adapted to an increased intensity of 1.5 G were found to orient properly in both 1.5- and
0.46-G ambient fields, but were disoriented in a magnetic field of intermediate intensity
(0.81 G). It is as if the birds are learning to utilize ambient magnetic field information at
new intensities; they appear to be able to use only those field intensities with which they
have become familiar through several days of exposure.
In an elegant series of experiments in which both the polarity and the inclination of
the magnetic field around the test cage were manipulated, the Wiltschkos obtained results
which indicate that, unlike man-made magnetic compasses, the magnetic compass of the
European robin does not use the polarity of the magnetic field for detection of direction
(Wiltschko and Wiltschko, 1972). Rather, these birds appear to obtain directional infor-
mation by measuring the inclination of the magnetic field, taking the north direction (in
the northern hemisphere) to be that in which the magnetic field vector makes the smallest
angle with the direction of gravity (Fig. 6). Such a detection strategy will fail when the
vertical component of the magnetic field is zero, such as occurs at the geomagnetic equator.
Indeed, European robins tested in orientation cages with the ambient magnetic field totally
horizontal were randomly oriented (Fig. 6E). The Wiltschkos point out that the European
robin does not cross the geomagnetic equator during its migration and thus should not
encounter this confusing situation under natural conditions. Bird species which migrate
between northern and southern hemispheres presumably employ strategies for the pro-
cessing of compass information from the magnetic field and other sources which avoid
confusion at the geomagnetic equator.
The Wiltschkos went on to investigate the interaction between the stellar and geo-
magnetic compass systems in the European robin and in several other species of nocturnally
migrating songbirds, the European warblers Sylvia borin (garden warbler) and S. communis
(whitethroat) (Wiltschko and Wiltschko, 1975a,b, 1976). In all cases the geomagnetic field
was found to be the primary source of compass information. The star compass was found
to be a secondary orientation system established by transferring directional information
obtained from the magnetic field to the stars. A very interesting difference, however, ap-
peared between the robins-which are relatively short-distance migrants, confining their
journey to the northern hemisphere-and the two warbler species-which are relatively
long-distance migrants, with large parts of their wintering grounds in the southern hem-
isphere. After a change in the ambient field direction, warblers responded on the first test
night with a corresponding change in migratory activity orientation, in spite of contra-
dictory directional information from the stars. On the other hand, robins required two or
three test nights for an ambient magnetic field change to produce a corresponding alteration
in the orientation of their nocturnal migratory activity. During these two or three nights,
they continued to use the stars as an indicator of direction. The Wiltschkos suggest that
the long-distance migrant warblers frequently recalibrate their steller compass against their
magnetic compass, and thus respond rapidly to a change in the magnetic field, whereas
the shorter-distance migrant robins less frequently check their stellar compass against their
magnetic one. During migratory flight, many of the stars seen by a robin at the beginning
466 Chapter 22
.@> -t·
5
Figure 6. Orientation behavior of European robins

tested in Kramer-type cages during spring migration.
On the left, the average headings of individual birds
c on single nights are denoted by solid triangles on the
periphery of the circle graphs. Arrows indicate di-
.E£). -t·
rections of mean vectors. Inner circles indicate 5%
5
(dashed) and 1% (solid) levels of significance com-
puted by the Rayleigh test. On the right, ambient
magnetic field configurations produced by the com-
bination of the geomagnetic field and the field gen-
D
erated by Helmholtz-type coils encompassing the ori-
"EB .+U
entation cages. N, S, U, and D denote the north, south,
5 up, and down directions, respectively, The intensity
of the resultant field in all configurations is 0.46 G.
(A) Normal geomagnetic field; (B) normal geomag-
netic field direction with opposite polarity; (C) ver-
s tical component of normal geomagnetic field re-
E w " " E
versed; (D) opposite polarity of configuration C; (E)
horizontal field only, vertical component zero. From
D
5 Wiltschko and Wiltschko (1972).
of its journey are still visible at the end, while the night sky might change virtually com-
pletely between the beginning and the end of a warbler's migratory journey.
Using similar methods, the indigo bunting (Passerina cyanea) (Emlen et al., 1976),
blackcap (Sylvia atricapilla) (Viehmann, 1979), savannah sparrow (Passerculus sand-
wichensis) (Bingman, 1981), and pied flycatcher (Ficedula hypoleuca) (Beck and
Wiltschko, 1982)-all of which are nocturnally migrating songbirds-have also been
shown to orient their migrational activity according to the direction of the ambient mag-
netic field.
A series of experiments by Gwinner and Wiltschko has demonstrated that the magnetic
compass sense in the garden warbler, at least, is innate (Wiltschko and Gwinner, 1974;
Gwinner and Wiltschko, 1978). Birds were taken from their nests before their eyes had
opened and hand-raised in the absence of environmental celestial cues (sun and stars).
When tested in closed cages several months later during the period of their first autumn
migration, the birds exhibited activity oriented in the appropriate southwesterly direction.
This is a very powerful result. There were no celestial cues available to these birds. More-
over, possible sources of navigational information such as olfactory cues and environ-
mental infrasound could not be innately programmed; they would have to be learned and
thus could not have been used by the birds. Thus, the earth's magnetic field was the only
known source of directional information available to these birds. As their orientation is
Birds 467
N N
w w E
Figure 7. Orientation behavior of

garden warblers during autumn
migration. Circle graphs denote
the orientation of hand-raised war-
blers tested during their first mi-
grational season in Kramer-type
cages. The average headings
(rounded to the nearest 5°) of in-
dividual birds on single nights are
denoted by solid triangles on the
periphery of the circles. Arrows in-
dicate direction of mean vector.
Upper left graph shows data rec-
orded during August and Septem-
ber. Upper right graph shows data
recorded during October, Novem-
ber, and December. Map indicates
directional aspects of the autumn
migration of free-flying garden
warblers from central Europe to
their wintering grounds in Africa
(hatched area); determined from
the monitoring of banded birds.
From Gwinner and Wiltschko
(1978).
known from other experiments to be dependent upon ambient magnetic field (Wiltschko
and Wiltschko, 1975a), the most parsimonious interpretation of these results is that the
geomagnetic field is supplying the necessary orientational reference to the young warblers.
Gwinner and Wiltschko (1978) also found that during the latter part of the period of
autumn migration, the orientation of the caged birds' activity shifted from southwest to
southeast. This shift is consistent with a change in flight direction which migrating garden
warblers make along their route from Europe to Africa (Fig. 7). Furthermore, if birds are
kept isolated from celestial cues, and under a constant 12-hr photoperiod and constant
temperature conditions for several months beyond their period of autumn migratory ac-
tivity, they will again exhibit migratory restlessness, now oriented in a northernly direction
appropriate to spring migration (Gwinner and Wiltschko, 1980). Thus, not only is the pro-
468 Chapter 22
cessing of magnetic compass information in these birds innate, but so also is an endogenous
program to control the onset, duration, and direction of flight during migration.
Hand-raised birds which had never been exposed to celestial cues were also used in
the orientation experiments of Bingman (1981) with savannah sparrows and Beck and
Wiltschko (1982) with pied flycatchers. In these experiments, changing the ambient mag-
netic field with Helmholtz-type coils also changed orientated migratory activity of the
birds. As for the European robin (Wiltschko and Wiltschko, 1972), the orientation of the
pied flycatcher appears to depend on the direction of the ambient magnetic field, rather
than on the polarity. Random orientation occurred when the ambient field was reduced
in intensity from the normal 0.46 G to 0.32 G (Beck and Wiltschko, 1982). This work
provides further strong support for an innately programmed magnetic compass in at least
some species of birds.
Recently, Alerstam and Hogstedt (1983) rotated by 90° the horizontal component of
the magnetic field surrounding the nests of pied flycatchers. Nestling birds could see the
sky when they peered out from the nest boxes. Young birds raised in this rotated magnetic
field exhibited a corresponding 90° deviation in the orientation of Zugunruhe when tested
during autumn migration under conditions of normal geomagnetic field and evening sky.
These results suggest that young pied flycatchers learn a corrspondence between an in-
nately programmed magnetic direction and observed celestial cues. Later, during migratory
orientation, the recalibrated celestial compass system seemed to be used as the primary
navigational reference.
Bingman (1983) raised savannah sparrows in view of the sky in an ambient magnetic
field rotated by 90° and then tested them during autumn migration under the normal geo-
magnetic field but without a view of the sky. He reported that oriented migratory rest-
lessness of the experimentally manipulated birds deviated by 90° from the oriented activity
of control birds raised under the normal geomagnetic field. As these birds could not see
the sky when tested, they were presumably orienting with respect to the geomagnetic field.
The deviant orientation of the experimental birds thus suggests that these birds changed
their magnetic compass on the basis of a comparison between celestial and magnetic in-
formation during the period prior to autumn migration.
As Bingman (1983) points out, it may be very important for migratory birds to be able
to modify an innately programmed magnetic compass after checking it against an unam-
biguous reference defining geographic direction-such as the rotational axis of the night
sky. The geomagnetic field varies so considerably in both space and time that it would
not be possible for birds to rely on an innately programmed and nonmodifiable geomagnetic
compass for navigation, For example, within the breeding range of the savannah sparrow
(extending from Alaska to the northeastern United States) the geomagnetic declination
varies by 80° (Bingman, 1983). Another problem facing an innately programmed magnetic
compass direction is the reversal of polarity in the earth's magnetic field which occurs
every few hundred thousand years or so. Such polarity reversals probably require several
thousand years to take place, with considerable wandering of the direction of the geo-
magnetic field vector during this period. Thus, over a span of hundreds of generations of
birds, the geomagnetic field would not be a stable directional reference. By checking the
magnetic compass against a stable celestial reference and modifying the former accordingly,
birds could later use their magnetic compass to select and maintain a desired geographic
direction during the absence of celestial cues.
In Bingman's (1981, 1983) savannah sparrow experiments, a curious bimodal pattern
of orientation, with two clusters separated by 180°, was measured. Individual birds were
not dually oriented; the bimodal clustering appeared only when the results from many
birds were pooled. When the horizontal component of the ambient magnetic field was
shifted by 90°, both clusters also shifted by 90°. Such a bimodal distribution was always
observed with groups of hand-raised birds that had never had flight experience. In contrast,
Birds 469
savannah sparrows captured in the field after their first migratory season exhibited uni-
modal orientation in the seasonally appropriate migratory direction when tested in ori-
entation cages (Bingman, 1983). This suggests that naive savannah sparrows need to cal-
ibrate the absolute direction of their geomagnetic compass by some means during actual
(perhaps migratory) flight experience.
4. Effects of Small Magnetic Field Changes on Navigation: The

Possibility of a Geomagnetic Map
The earth's magnetic field is subject to relatively small «5%) fluctuations in intensity
and direction as a result of storms which take place on the sun. Such solar storms can
result in significant increases in the stream of charged particles which flows outward from
the sun and sweeps by the earth. These bursts of charged particles generate currents in
the earth's upper atmosphere which affect the intensity and direction of the magnetic field
at the earth's surface. An irregular intensity change in the horizontal or vertical component
of the geomagnetic field of more than about 40 nT (1 nT = 1 'Y = 10- 5 G) over a period
of 3 hr is generally considered to be indicative of a magnetic storm. Major storms, which
occur not infrequently, produce intensity changes of 120 nT or more. On rare occasions,
intensity changes as large as 2500 nT (5% of the average geomagnetic field intensity) have
been measured.
Working with ring-billed gulls in orientation cages, Southern (1971, 1972a, 1978) has
reported that the tendency of these gulls to orient in the appropriate migratory direction
is disrupted when magnetic storms produc!3 intensity changes of 50 nT or more. Keeton
and co-workers (Keeton et aI., 1974; Larkin and Keeton, 1976) have reported that homing
pigeon vanishing bearings tend to be deflected counterclockwise during magnetic dis-
turbances, the magnitude of the deflection increasing with increasing magnetic storm ac-
tivity. The magnitude of this deflection also depends upon the release site: a relatively
large effect being measured at one release site (Fig. 8), while a much smaller effect was
found at a second site. The geomagnetic fluctuations during these studies were very likely
less than 100 nT, and the indication is that magnetic field intensity changes of less than
30 nT are sufficient to affect the initial orientation of homing pigeons.
In other studies demonstrating an effect of magnetic disturbances on avian navigation,
Moore (1977) found a correlation between the scatter in flight direction of free-flying noc-
turnal passerine migrants and magnetic storm activity. Yeagley (1951) pointed to a negative
correlation between homing times-as measured in several 100- to 300-km New Jersey
pigeon races-and sunspot number, an observational parameter known to be associated
with increased solar magnetic activity. More recently, Schreiber and Rossi (1976) examined
records from 18 pigeon races held between 1932 and 1957 over a 736-km route in Italy;
they found a strong negative correlation between sunspot number and the percentage of
pigeons which returned home on the same day as their release 736 km away. Finally, Carr
et al. (1982) have reported that a major magnetic storm during the summer of 1972 appears
to have significantly affected the performance of homing pigeons in several east-west races
in the United States, but not in north-south races.
The orientational behavior of homing pigeons released at a number of magnetic an-
omalies in the northeastern United States has been examined by Walcott (1978). Such
magnetic anomalies are locations where the geomagnetic field is permanently distorted,
by the presence of large quantities of magnetic mineral in the local rock. Walcott found
that experienced pigeons released under sunny skies at these locations were far more
scattered in their homeward orientation than they were when released at magnetically
normal sites. Moreover, the extent of the scatter was correlated with the magnetic anomaly'S
470 Chapter 22
III
Q)
60
...
Q)
CI
Figure 8. Influence of geomagnetic storm activity
Q)
'tl on the initial orientation of homing pigeons. Ex-
50 perienced pigeons were released at a site in New
z
o • York approximately 70 km from their home loft.
i= Pigeons released at this site normally departed
u 40 •
w about 12° counterclockwise from the home direc-
a:
15 • tion on magnetically quiet days; this deviation from
30 the home direction increased with increasing geo-
magnetic disturbance. Points indicate the average
deviation of vanishing bearings from the home di-
ú = 20
o
a: • • rection for seven release experiments conducted
ú =
during 1973 (each experiment involved an unre-
z 10 ported number of birds). Geomagnetic disturbance
o is measured as the sum of four 3-hr K values for the
ú= o 12-hr period ending at the time of completion of
:>
w the release. The K index is a parameter used to de-
o 5 10
scribe geomagnetic storm activity over a 3-hr pe-
GEOMAGNETIC DISTURBANCE riod. For these data, summed K values of 4 and 12,
(12-hour summed K values) for example, reflect approximate geomagnetic dis-
turbances of 5-10 and 40-70 nT, respectively. Cur-
iously, this 12-hr sum apparently provided a better independent variable than did a 6-hr sum or the
single 3-hr K value at the time of the release. From Keeton et 01. (1974).
strength, as measured by the maximum variability of the total field intensity over the first
kilometer in the homeward direction from the release site. This variability ranged from
approximately 30 nT to nearly 3000 nT, depending on the release site. Again, the suggestion
from Walcott's data is that pigeons are sensitive to field distortions as small as 30 nT (Fig.
9). Kiepenheuer (1982) has found a similar correlation between disorientation and strength
of magnetic anomaly for homing pigeons released under sunny skies in southern Germany.
Walcott (1980) has further found that pigeons carrying magnets and released on sunny
days at magnetic anomalies are just as disoriented as birds carrying nonmagnetic brass
bars. Thus, the sensitivity of these birds to the field distortion produced by an anomaly
persists even on top of the additional relatively large (-1 G) field of the attached bar magnet.
[This result should be compared to that of Larkin and Keeton (1976), who found that
1.0
:I:
0.8 Figure 9. Correlation between the strengths of
•
I-
<.? geomagnetic anomalies and the orientation of
z 0.6
W homing pigeons released in the neighborhoods of
...J
the anomalies. Geomagnetic distortion at a mag-
a: 0.4 netic anomaly is measured by the maximum
o
I-
U
• change in the total intensity of the geomagnetic
W field measured along a l-km line in the direction
> 0.2 of home. Homing orientation is measured by the
length of the mean vector of vanishing bearings.
A mean vector length of 1.0 indicates that all pi-
geons vanished in the same direction (no scatter
30 100 300 1000 3000 in vanishing bearings). A mean vector length of
GEOMAGNETIC DISTORTION (nanotesla) 0.0 indicates a random distribution of vanishing
bearings (maximum scatter). Mean vectors are
based on vanishing bearings (measured by radiotelemetry) for between 23 and 44 pigeon releases per
site. From Walcott (1978).
Birds 471
attached bar magnets seem to disrupt the effects of magnetic storms on vanishing bearings,
suggesting that perhaps the stronger field of the attached magnet prohibits a bird from
sensing the relatively small field changes associated with a magnetic storm.]
When released at a location away from home, pigeons generally do not depart directly
toward home, but exhibit a consistent deviation from the home direction (Kramer, 1959;
Wallraff, 1970, 1978; Keeton, 1973; Windsor, 1975). This "release-site bias" or "preferred
compass direction" is relatively stable over time and somewhat characteristic of a given
site. Keeton (1973) showed that the release-site bias is independent of weather condition
(sun versus overcast), age and experience of the birds, location of the home loft, and even
bird species (pigeon versus swallow)-and concluded that some environmental factor basic
to the avian navigational map is characteristically rotated at certain release sites. Gould
(1980, 1982) has considered the possibility that such release-site biases are representative
of variations in geomagnetic topography. In order to use the geomagnetic field as a map,
pigeons, for example, would presumably learn patterns of variation in magnetic field in-
tensity (and perhaps direction) in the vicinity of their home loft. The direction and distance
of a displacement from their loft could then be ascertained by extrapolating their locally
learned magnetic map. Nonlinearities in the local magnetic gradients might thus result in
systematic errors as the birds consult their extrapolated magnetic map at a release site.
Wagner and Frei (Wagner, 1976; Frei and Wagner, 1976; Frei, 1982) have reported that
homing pigeons released in the neighborhood of a geomagnetic anomaly in Switzerland
tended to exhibit deviations from the homeward direction toward that direction which
maximized the magnetic field intensity change between release site and home. Such results
again suggest that small changes in magnetic field intensity may be used as navigational
cues by homing pigeons and support a possible relationship between magnetic topography
and release-site bias.
Young homing pigeons are generally allowed to practice flying in the vicinity of their
loft on a dialy basis. It is during such exercise flights that birds are presumed to learn
patterns in the geomagnetic field which they might later extrapolate to construct a navi-
gational map. However, experiments begun by Kramer (1959) and continued by Wallraff
(1970, 1979) indicate that pigeons can orient and begin flying toward home even when
they have been confined in aviaries since birth, never before let out to practice flying in
the vicinity of their loft. As it is virtually impossible to imagine how anything useful could
be learned about long-range geomagnetic patterns over the space of an aviary only several
meters wide, such results suggest that pigeons have access to more than just extrapolated
local geomagnetic field information for the construction of a navigational map. One pos-
sibility is that pigeons rely on information detected during the journey to the release site
if they have been prevented from constructing a map based on extrapolation. Experiments
of the type described in Section 2.2 in which pigeons are prevented from collecting useful
magnetic field information during transport to the release site have thus far not been carried
out with birds confined to aviaries since birth. One would predict that if the geomagnetic
field were the only source of map information, a bird so raised and so manipulated would
be completely confused upon release at a distant site. More definitive experiments could
be done by imposing controlled magnetic field changes during the outward journey mim-
icking the changes which would be experienced during transit in a completely different
direction.
In summary, our present knowledge of avian navigation is consistent with the hy-
pothesis that birds utilize the geomagnetic field as an information source for both a compass
(directional) and a map (positional) sense. The experimental evidence supporting a geo-
magnetic compass in birds is quite compelling. Young, inexperienced birds appear to use
a geomagnetic compass to determine direction even before the ability to use the sun as a
compass develops. Older, experienced birds still rely on a geomagnetic compass to de-
termine direction in some circumstances, especially in the absence of information from
472 Chapter 22
celestial sources. Such a compass sense appears to involve the determination of the geo-
magnetic field vector's direction, and not its polarity. It is perhaps worth noting that such
a strategy might be less affected by the reversals of polarity which occur in the earth's
magnetic field every few hundred thousand years.
Because bird orientation is affected by relatively small changes in the ambient mag-
netic field (produced by magnetic storms and geomagnetic anomalies), it appears that the
geomagnetic field is being utilized for more than compass information. A 500-nT magnetic
storm might, at most, rotate a compass needle a mere 10 or so, and thus little effect of such
a storm on orientation would be expected if birds were using the geomagnetic field only
as a source of compass information. However, if the geomagnetic field is being used as a
navigational map, the perception of small changes in intensity (or direction) is essential.
The intensity of the geomagnetic field changes by roughly 3-5 nT/km. Thus, to account
for the homing accuracy of pigeons wearing frosted contact lenses, a sensitivity to intensity
changes of approximately 10 nT must be postulated, the same order of sensitivity that is
derived from the results of geomagnetic storm and magnetic anomaly experiments (Gould,
1980, 1982; Walcott, 1982). The significance of the frosted contact lens experiments
(Schmidt-Koenig and Schlichte, 1972; Schmidt-Koenig and Walcott, 1978) can hardly be
overstated. Birds wearing these lenses cannot be using visual patterns as navigational aids.
Furthermore, it seems unlikely that navigational maps based on olfactory cues or envi-
ronmental infrasound sources could account for an accuracy of 1-2 km. Moreover, if ol-
factory cues playa critical role in avian navigation, it is surprising that birds should not
have been able to pinpoint the location of the loft-by way of its "characteristic odor"
(Schmidt-Koenig and Walcott, 1978)-after having arrived in its vicinity. While a navi-
gational map based on the geomagnetic field is, indeed, a most attractive explanation of
the existing data, there is as yet no direct evidence that such a geomagnetic map exists for
birds or for any other animal. It has even been argued that large-scale anomalies in the
geomagnetic field-having amplitudes of hundreds of nanotesla and extending over many
kilometers-should, in principle, preclude its use as a navigational map (Lednor, 1982).
5. Laboratory Attempts to Measure Avian Magnetic Field

Sensitivity
Navigational experiments with homing pigeons are conducted by transporting birds
to a distant location and determining the direction (through observation with binoculars
or by radiotelemetry) in which individual birds vanish from sight after being released. By
following radiotelemetrically monitored homing pigeons in an airplane, Michener and
Walcott (1967) were further able to determine a bird's entire flight path, or at least a large
part of it. Migratory birds are less amenable to navigational experimentation in the field,
although some homing-type experiments have been conducted with various species. Most
of the work concerning the possible role of the geomagnetic field in the navigation of
migratory birds has come from studies of orientation in Kramer-type cages. Individual birds
confined in circular or octagonal cages display oriented activity (Zugunruhe or migratory
restlessness) during those periods when they would normally be on migratory flights. The
direction of the oriented activity can be monitored by using microswitch-equipped perches
around the periphery of the cage or by using funnel-shaped cages designed to record the
direction in which a bird hops (Emlen, 1975a). As discussed in Section 3, this paradigm
has been used to obtain effects of ambient magnetic field configuration on direction of
orientation, thus suggesting the presence of a magnetic compass in many birds. The results
of these experiments have been subjected to some criticism because well-defined orien-
tation seems to depend strongly on certain aspects of cage geometry (apparently, radially
Birds 473
oriented perches produce oriented behavior, while tangentially oriented ones do not; the
funnel cage works for some species, whereas the radial-perch cage does not) and on the
methods of data analysis (Emlen, 1975a). Beck and Wiltschko (1983) have, however, dis-
cussed cage geometry in terms of the specific locomotor behavior of some bird species;
for example, pied flycatchers tend to hover in such a way that orientation is better recorded
in a funnel cage than it is in a radial-perch cage. It seems that when the canonical methods
of experiment and data analysis are utilized, clear and reproducible effects of the ambient
magnetic field are obtained on the orientation of migratory activity. As these effects have
been found by investigators in several different laboratories using several different pas-
serine species, it is likely that they indicate a real influence of ambient earth-strength
magnetic fields on avian orientation. To obtain clear results, nevertheless, it is necessary
to observe many birds over many days during periods of migrational activity. Even then,
the signal-to-noise ratio (a feeling for which can be obtained from Figs. 6 and 7) is still
too low for the results to win the approval of everyone. Unfortunately, it may be the case
that the signal-to-noise for experiments such as those done in orientation cages will never
improve much beyond the present level.
In order to further substantiate (as a prelude to better characterization) the perception
of magnetic fields by birds, a number of investigators have attempted to evoke a response
to various aspects of the ambient magnetic field in a laboratory situation. Griffin, for ex-
ample, has alluded on several occasions (Griffin, 1952, 1982) to various laboratory inves-
tigations (conducted on and off over the last four decades) of whether birds might perceive
earth-strength magnetic fields. One recent attempt utilized a paradigm in which European
starlings were trained to' fly in a wind tunnel for extended periods of time (Torre-Bueno,
1976; Griffin, 1982). Using negative reinforcement with electric shock, birds were con-
ditioned to move to one or the other side of the wind tunnel in response to the presence
or absence of a light stimulus paired with a magnetic field stimulus (being a change in the
direction of the ambient magnetic field by approximately 45°). The results were negative;
when the light intensity was dimmed to below visual threshold, the magnetic field alone
did not cue the birds to alter their flight direction.
Other efforts to condition birds to magnetic stimuli have employed a paradigm which
monitors heart rate increases elicited by an electric shock which is preceded by the stimulus
of interest. This paradigm has proven satisfactory for the study of the perception of a variety
of sensory stimuli, including light, sound, odor, and barometric pressure. Although Reille
(1968) had reported success in conditioning a response in homing pigeons to a magnetic
field change, attempts to reproduce these results have not been successful (Kreithen and
Keeton, 1974c; Beaugrand, 1976). In the studies of Kreithen and Keeton (1974c), the am-
bient magnetic field was changed during a 10-sec period to a configuration in which the
total intensity and inclination remained approximately the same as the normal earth's field,
but with the declination rotated between 90 and 120° from normal. A sensory receptor
sensitive to magnetic compass direction might be expected to respond to such a change.
A similar series of experiments by Beaugrand (1976) also failed to condition a cardiac
response in homing pigeons to magnetic field changes delivered over a period of approx-
imately 10 sec. In these experiments the change encompassed total intensity, inclination,
and declination, but the altered field remained within the natural range of variation of the
geomagnetic field.
The cardiac conditioning paradigm has employed a rather short stimulus duration-
approximately 10 sec-perhaps too short for the bird to respond to. Pigeons usually require
a few minutes to orient toward home upon being released and there are also other reasons
to believe that the processing of magnetic field information by birds may require several
minutes (Gould, 1982). For example, Walcott's studies of pigeon orientation in the neigh-
borhood of magnetic anomalies indicate that birds do not recover from the disorientation
they experience when released at anomalies until they have been out of the distorted field
474 Chapter 22
for some time; on the other hand, pigeons released elsewhere appear to fly through geo-
magnetic anomalies with little effect (Walcott, 1978).
Another possible shortcoming of some of the laboratory paradigms (although not the
wind-tunnel experiment with starlings) which attempt to measure magnetic field sensi-
tivity is that the bird is held immobile in the center of a set of Helmholtz-type coils with
which the ambient field can be altered. If motion or muscular activity of some sort is
necessary for the perception of magnetic field stimuli, then such methods would perforce
fail. Interestingly, Bookman (1977, 1978) has reported success in training pigeons to dis-
criminate between the presence and absence of an earth-strength (0.5 G) magnetic field.
In his paradigm, birds were trained to associate one or the other of two food containers at
the end of a 3.5-m tunnel with each of the two magnetic field configurations. His results
indicate that successful discrimination was accompanied by "fluttering activity" (sus-
tained hovering, short flights) of the birds as they negotiated the length of the tunnel before
making a choice. If the birds walked the length of the tunnel, their association of a specific
food container with a specific magnetic field configuration was random. These results are
consistent with a mechanism in which some sort of movement activity is required for the
reception and/or processing of magnetic field information. The experiments for which
Bookman reported success were, however, done with mated pairs of pigeons together in
the tunnel and the presence or absence of flutter activity may have reflected different
motivational states of these birds; this may well have been the key factor in determining
sensitivity to the ambient magnetic field.
The lack of a definitive laboratory assay for avian magnetic field sensitivity is certainly
an impediment to better characterization of this sense. However, negative results obtained
in an environment so grossly unnatural to the utilization of sensory processes normally
used only during navigation can hardly be considered as strong evidence against the very
existence of such processes.
6. Magnetite in Birds and Possible Mechanisms of Magnetic

Field Sensitivity
A biological detector capable of measuring the direction and/or intensity of the geo-
magnetic field would presumably be based on one of the known ways in which magnetic
fields may influence matter. Many attempts to hypothesize mechanisms to account for the
presumed sensitivity of birds to the earth's magnetic field have been frustrated by the
relative weakness of the interactions of earth-strength magnetic fields with matter. For
example, the interactions of diamagnetic and paramagnetic substances with the geomag-
netic field are too weak to overcome thermal randomizing effects. A rather novel proposed
mechanism-involving polarized light emission from a molecular triplet state split in en-
ergy by the ambient magnetic field-suffers primarily from the unlikely existence of a
molecule with sufficiently small zero-field splittings of the triplet state (Leask, 1977). Elec-
tromagnetic induction seems to be a plausible mechanism for geomagnetic field sensitivity
in marine elasmobranch fishes (sharks, skates, and rays) which possess extraordinarily
sensitive electroreceptive organs (Kalmijn, 1978, 1981). Elasmobranch electroreceptors
achieve their high sensitivity by utilizing the low resistivity of their seawater environment
to complete an electric circuit. An electroreceptor could not be of similar design in birds
because of the high resistivity of air. Thus, for birds to detect the geomagnetic field by
means of electromagnetic induction, the detector circuit must be contained completely
within the animal. Jungerman and Rosenblum (1980) have discussed the size requirements
of such a detector and conclude that it would need to be several millimeters in diameter.
They discuss the semicircular canals of the inner ear in this context. Other innervated
conducting loops might also be hypothesized.
Birds 475
Energetically, the most feasible way to construct a biological detector capable of meas-
uring earth-strength magnetic fields and fractional changes thereof is to utilize permanently
magnetic particles. Indeed, Donald Griffin perspicaciously pointed out in a 1944 article
on bird navigation that "sensitivity to as weak a field as the earth's is made extremely
unlikely by the fact that living tissues are not known to contain any of the ferrimagnetic
substances (such as metallic iron, magnetic iron oxide or magnetite, nickel and cobalt)
which alone are capable of exerting appreciable mechanical forces in the earth's magnetic
field." Shortly thereafter, Ising (1945) reported that magnetometer measurements of bird
specimens did not reveal the presence of any permanently magnetic material. (Ising then
proposed a theory of avian navigation based on the perception of Coriolis forces.) The first
detection of magnetite of biosynthetic origin was as the capping material in chiton teeth,
a discovery made by Lowenstam (1962) some 20 years ago. Lowenstam even suggested
that such a magnetic mineral might be involved in mediating the claimed sensitivity of
various organisms to weak magnetic fields. Despite this discovery and proposal, it was
Blakemore's more recent (1975) discovery of magnetotactic bacteria which catalyzed se-
rious thinking regarding the possibility of permanent magnets being employed by animals
for the detection of the geomagnetic field. In 1978, Gould et al. used SQUID magneto me try
to investigate honeybees, insects whose orientational behavior is affected by earth-strength
magnetic fields. They found fine-grained, permanently magnetic material within these bees
and identified it as magnetite on the basis of Curie temperature (Gould et a1., 1978). Using
the same methodology, Walcott et a1. (1979) discovered deposits of magnetite inside the
skull of homing pigeons. Subsequent studies by Presti and Pettigrew (1980) revealed the
presence of fine-grained, iron-rich, permanently magnetic material throughout the head
and neck region of homing pigeons, feral pigeons, white-crowned sparrows (Zonotrichia
leucophrys), and a variety of other avian species. In several white-crowned sparrows and
older (several years of age) homing pigeons, it was possible to visualize magnetic material
embedded in the neck musculature and associated fascia. This material had the form of
aggregates of very small particles. What the function, if any, of such amorphous structures
in avian muscle tissue might be is quite a mystery. It is unlikely to be sensory. Within the
heads of birds, induced remanent magnetization was found to be largely associated with
the skull. This remanence was diffuse, being spread rather uniformly throughout the skull
(Presti and Pettigrew, 1980). Induced remanent magnetization was also measured in other
avian tissue-heart muscle, for example. Walcott and Walcott (1982) are continuing the
investigation of iron-containing regions of the pigeon head through the use of iron-staining
techniques and microscopy.
The existence of large numbers of magnetic particles-probably magnetite and likely
biogenic-in animal tissue has important implications for the study of magnetic field sen-
sitivity in birds. The potential utilization of permanent magnets to construct a magnetic
field detector makes possible a variety of designs with sufficient sensitivity to account for
the known behavioral data (Yorke, 1979, 1981; Presti and Pettigrew, 1980; Kirschvink and
Gould, 1981). Perhaps the most obvious way to build a magnetic field detector from such
tiny magnets is to couple them with one of the sensitive mechanoreceptors found in an-
imals, such as a hair cell, pressure receptor, or stretch receptor. A variety of other clever
designs may also be imagined (Kirschvink and Gould, 1981). Certainly, with the discovery
of fine-grained magnetite in animal tissue, it is no longer tenable to exclude the perception
of weak magnetic fields simply on the grounds that an effective sensory mechanism cannot
even be envisioned.
Hypothesized mechanisms of magnetite-based magnetic field perception in animals
depend upon the alignment of arrays of single-domain magnetite grains in the ambient
magnetic field. Detection of the average direction of alignment by the nervous system would
provide the basis for a compass sense, while detection of the variance of the alignment
about the average direction would provide the basis for intensity perception and thereby
476 Chapter 22
a map sense. A small number (=10 2 ) of magnetite grains are probably sufficient to give
an accurate compass direction, whereas a relatively large number (=10 6 _10 7 ) of grains
seem necessary to construct a sensitive intensity detector (Yorke, 1981; Kirschvink and
Gould, 1981; Kirschvink, 1982). The measured remanence in avian tissue is consistent
with the presence of such large numbers of magnetite grains (Walcott et al., 1979; Presti
and Pettigrew, 1980).
The alignment of single-domain grains in a magnetic field is a function not only of
ambient magnetic field intensity, but also of ambient temperature. The greater the ambient
magnetic field intensity, the smaller is the variance in grain alignment-whereas the greater
the ambient temperature, the larger the variance in grain alignment, due to thermal buf-
feting of the grains. If the temperature is constant, then changes in the variance of grain
alignment will be produced only by changes in the ambient magnetic field intensity. If
the ambient temperature of the magnetic sensor system is not constant, then it may be
impossible to distinguish temperature changes from magnetic field intensity changes (un-
less the detector is equipped with some unknown mechanism to correct for temperature).
By summing responses from many individual grains or by averaging over an extended
period of time, the signal-to-noise ratio of a magnetic detection system can be increased,
thus improving the resolution of the system (Yorke, 1979, 1981; Kirschvink and Gould,
1981). Such integration and averaging will not, however, remove the relative dependence
of the detection system on changes in ambient temperature and field intensity, although
the effects of thermal fluctuations can be averaged out.
The average alignment of a single-domain particle of magnetic moment I.L in an ambient
magnetic field of intensity B can be represented as the Boltzmann average of cos 8 over
the surface of a sphere, where 8 is the angle between the directions of I.L and B:
(cos 8) = f(cos 8) exp[I.LB cos 8/kTldO

X K K K K WK K J ú J ú J J J J J Z J J
fexp[I.LB cos 8/kTldO
= coth[I.LBlkTl - (kT/I.LB)
= L[I.LBlkTj
Here, k is Boltzmann's constant and T is the absolute temperature. This defines the Lan-
gevin function, L, originally derived by the French physicist Paul Langevin in a description
of molecules possessing magnetic dipole moments (Langevin, 1905a,b). The variance, or
root mean square deviation, of this Boltzmann average of cos 8 is given by
a = ((cos 2 8) - (cos 8)2)1/2

= (1 - 2LkT/I.LB - L2)1/2
For particles in the single-domain size range, the relative dependence of the alignment
variance on ambient temperature and magnetic field is given by the ratio (iialiiB)/(iialiiT).
This ratio turns out to be simply equal to - TIB and thus is independent of the exact value
of the single-domain magnetic moment. One finds that at a temperature of 41°e (approx-
imate avian body temperature) and an ambient magnetic field intensity of 0.5 G, a O.l°e
change in the temperature will be indistinguishable from a magnetic field intensity change
of 16 nT.
In birds and mammals, the homeothermic animals, body temperature is carefully reg-
ulated and maintained relatively constant. While rapid increases (of 1-4°C) in core body
temperature and brain temperature occur after the onset of flight in birds, during extended
flight a relatively constant elevated body temperature is maintained (Torre-Bueno, 1976;
Bernstein, 1982). Torre-Bueno (1976) found, for example, that the core temperature of
Birds 477
European starlings flying in a wind tunnel was maintained constant to within 0.1-0.4°C
for extended periods of time. If temperature control in active homeotherms is limited to
± 0.1 °C or so over extended periods of time, then this may play an important role in limiting
the sensitivity of a hypothetical geomagnetic field intensity detector. Again, the sensitivity
limits estimated on this basis fall within the range of 10-30 nT predicted from the various
orientation experiments described in Section 4.
In poikilothermic animals, such as insects, amphibians, reptiles, and some fish, body
temperature can fluctuate over a wide range of many degrees. This may limit the ability
of such animals to extract detailed positional information from the geomagnetic field.
However, like birds and mammals, certain fish (tuna, for example) and insects (honeybees,
for example) have various means of regulating body temperature to keep it relatively con-
stant under certain conditions.
The apparent occurrence of fine-grained, permanently magnetic material in several
kinds of tissue presents the possibility that several roles may be played by these particles.
These roles would presumably utilize one or more of magnetite's unique properties of
ferrimagnetism, high density, hardness, and conductivity. With respect to the possible
utilization of biogenic magnetite by birds and other animals as a sensory basis for geo-
magnetic field detection, it is tempting to say now, as Dirac (1931) wrote a half-century
ago in another context involving electromagnetism, that "under these circumstances one
would be surprised if Nature had made no use of it."
ACKNOWLEDGMENTS. I thank Professor Donald Griffin and an anonymous reviewer for

thoughtful comments on the manuscript. The author is a Monsanto Fellow of the Life
Sciences Research Foundation.
References
Able, K. P., 1980, Mechanisms of orientation, navigation, and homing. in: Animal Migration. Ori-
entation. and Navigation (S. A. Gauthreaux. Jr .• ed.). Academic Press. New York. pp. 283-373.
Alerstam. T .• and Hiigstedt. G.• 1983. The role of the geomagnetic field in the development of birds'
compass sense. Nature 306:463-465.
Beaugrand. J. P .• 1976. An attempt to confirm magnetic sensitivity in the pigeon. Columba livia, J.
Beck, W., and Wiltschko, W.• 1982, The magnetic field as a reference system for genetically encoded
migratory direction in pied flycatchers (Ficedula hypoleuca Pallas), Z. Tierpsychol. 60:41-46.
Beck, W .• and Wiltschko, W., 1983. Orientation behaviour recorded in registration cages: A comparison
of funnel cages and radial perch cages. Behaviour 87:145-156.
Benvenuti. S., Baldaccini. N. E., and loale. P .. 1982, Pigeon homing: Effect of altered magnetic field
during displacement on initial orientation, in: Avian Navigation (F. Papi and H. G. Wallraff. eds.),
Springer-Verlag, Berlin. pp. 140-148.
Bernstein, M. H .• 1982, Temperature regulation in exercising birds, in: A Companion to Animal Phys-
iology, (C. R. Taylor, K. Johansen, and L. Bolis, eds.), Cambridge University Press, London. pp.
189-197.
Bingman, V. P., 1981, Savannah sparrows have a magnetic compass. Anim. Behav. 29:962-963.
Bingman, V. P., 1983, Magnetic field orientation of migratory savannah sparrows with different first
summer experience. Behaviour 87:43-53.
Blakemore, R. P .• 1975. Magnetotactic bacteria. Science 190:377-379.
Bookman. M. A.. 1977. Sensitivity of the homing pigeon to an earth-strength magnetic field. Nature
267:340-342.
Bookman, M. A., 1978, Sensitivity of the homing pigeon to an earth-strength magnetic field. in: Animal
Migration, Navigation. and Homing (K. Schmidt-Koenig and W. T. Keeton. eds.), Springer-Verlag,
Berlin. pp. 127-134.
Carr. P. H., Switzer, W. P., and Hollander, W. F., 1982. Evidence for interference with navigation of
homing pigeons by a magnetic storm. Iowa State J. Res. 56:327-340.
478 Chapter 22
Delius, J. D., Perchard, R. J., and Emmerton, J., 1976, Polarized light discrimination by pigeons and
an electroretinographic correlate, J. Compo Physiol. Psycho1. 90:560-571.
Dirac, P. A. M., 1931, Quantised singularities in the electromagnetic field, Proc. R. Soc. London Ser.
A 133:60-72.
Emlen, S. T., 1975a, Migration: Orientation and navigation, in: Avian Biology, Volume 5 (D. S. Farner
and J. R. King, eds.), Academic Press, New York, pp. 129-219.
Emlen, S. T., 1975b, The stellar-orientation system of a migratory bird, Sci. Am. 233(2):102-111.
Emlen, S. T., Wiltschko, W., Demong, N. J., Wiltschko, R., and Bergman, S., 1976, Magnetic direction
finding: Evidence for its use in migratory indigo buntings, Science 193:505-508.
Frei, V., 1982, Homing pigeons' behaviour in the irregular magnetic field of western Switzerland, in:
Avian Navigation (F. Papi and H. G. Wallraff, eds.), Springer-Verlag, Berlin, pp. 129-139.
Frei, V., and Wagner, G., 1976, Die Anfangsorientierung von Brieftauben im erdmagnetisch gestorten
Gebiet des Mont Jorat, Rev. Suisse Zoo1. 83:891-897.
Goldsmith, T. H., 1980, Hummingbirds see near ultraviolet light, Science 207:786-788.
Gordon, D. A., 1948, Sensitivity of the homing pigeon to the magnetic field of the earth, Science
108:710-711.
267.
Gould, J. 1., 1982, The map sense of pigeons, Nature 296:205-211.
201:1026-1028.
Griffin, D. R., 1944, The sensory basis of bird navigation, Q. Rev. BioI. 19:15-31.
Griffin, D. R., 1952, Bird navigation, BioI. Rev. Cambridge Philos. Soc. 27:359-393.
Griffin, D. R., 1972, Nocturnal bird migration in opaque clouds, in: Animal Orientation and Navigation
(S. R. Galler, K. Schmidt-Koenig, G. J. Jacobs, and R. E. Belleville, eds.), NASA SP-262, V.S.
Griffin, D. R., 1973, Oriented bird migration in or between opaque cloud layers, Proc. Am. Philos.
Soc. 117:117-141.
Griffin, D. R., 1982, Ecology of migration: Is magnetic orientation a reality?, Q. Rev. BioI. 57:293-295.
Gwinner, E., and Wiltschko, W., 1978, Endogeneously controlled changes in migratory direction of
the garden warbler, Sylvia borin, J. Compo Physiol. 125:267-273.
Gwinner, E., and Wiltschko, W., 1980, Circannual changes in migratory orientation of the garden
warbler, Sylvia borin, Behav. Ecol. Sociobiol. 7:73-78.
Hoffmann, K., 1954, Versuche zu der im Richtungsfinden der Vogel enthaltenen Zeitschiitzung, Z.
Tierpsychol. 11:453-475.
Hoffmann, K., 1960, Experimental manipulation of the orientational clock in birds, Cold Spring Harbor
Symp. Quant. BioI. 25:379-387.
Ising, G., 1945, Die physikalische Moglichkeit eines Tierischen Orientierungsinnes auf Basis der Er-
drotation, Ark. Mat. Astron. Fys. 32A:1-23.
Jungerman, R. L., and Rosenblum, B., 1980, Magnetic induction for the sensing of magnetic fields by
animals-An analysis, J. Theor. BioI. 87:25-32.
Kalmijn, A. J., 1978, Experimental evidence of geomagnetic field orientation in elasmobranch fishes,
in: Animal Migration, Navigation, and Homing (K. Schmidt-Koenig and W. T. Keeton, eds.),
Springer-Verlag, Berlin, pp. 347-353.
Kalmijn, A. J., 1981, Biophysics of geomagnetic field detection, IEEE Trans. Magn. Mag-17:1113-1124.
Keeton, W. T., 1969, Orientation by pigeons: Is the sun necessary?, Science 165:922-928.
Keeton, W. T., 1971, Magnets interfere with pigeon homing, Proc. Natl. Acad. Sci. USA 68:102-106.
Keeton, W. T., 1973, Release-site bias as a possible guide to the "map" component in pigeon homing,
J. Compo Physio1. 86:1-16.
Keeton, W. T., 1974a, The orientational and navigational basis of homing in birds, Adv. Stud. Behav.
5:47-132.
Keeton, W. T., 1974b, The mystery of pigeon homing, Sci. Am. 231(6):96-107.
Kiepenheuer, J., 19 78a, Pigeon navigation and magnetic field: Information collected during the outward
journey is used during the homing process, Naturwissenschaften 65:113-114.
Birds 479
Kiepenheuer, J., 1978b, lversion of the magnetic field during transport: Its influence on the homing
behavior of pigeons, in: Animal Migration, Navigation, and Homing (K. Schmidt-Koenig and W.
T. Keeton, eds.). Springer-Verlag, Berlin, pp. 135-142.
Kiepenheuer, J., 1982, The effect of magnetic anomalies on the homing behaviour of pigeons: An
attempt to analyse the possible factors involved, in: Avian Navigation (F. Papi and H. G. Wallraff,
eds.). Springer-Verlag, Berlin, pp. 120-128.
Kirschvink, J. 1., 1982, Birds, bees and biomagnetism: A new look at the old problem of magnetore-
ception, Trends Neurosci. 5:160-167.
Kirschvink, J. 1., and Gould, J. 1., 1981, Biogenic magnetite as a basis for magnetic field detection in
Kramer, G., 1952, Experiments on bird orientation, Ibis 94:265-285.
Kramer, G., 1959, Recent experiments on bird orientation, Ibis 101:399-416.
Kreithen, M. 1. and Eisner, T., 1978, Ultraviolet light detection by the homing pigeon, Nature 272:347-
348.
Kreithen, M. 1., and Keeton, W. T., 1974a, Detection of changes in atmospheric pressure by the homing
pigeon, J. Compo Physiol. 89:73-82.
Kreithen, M. 1., and Keeton, W. T., 1974b, Detection of polarized light by the homing pigeon, Columba
livia, ,. Compo Physiol. 89:83-92.
Kreithen, M. 1., and Keeton, W. T., 1974c, Attempts to condition homing pigeons to magnetic stimuli,
J. Compo Physiol. 91:355-362.
Kreithen, M. 1., and Quine, D. B., 1979, lnfrasound detection by the homing pigeon: A behavioral
audiogram, J. Compo Physiol. 129:1-4.
Langevin, P., 1905a, Magnetisme et theorie des electrons, Ann. Chim. Phys. 8(5):70-127.
Langevin, P., 1905b, Sur la tMorie du magnetisme, J. Phys. (Paris) 4(4):678-693.
Leask, M. J. M., 1977, A physiochemical mechanism for magnetic field detection by migratory birds
and homing pigeons, Nature 267:144-145.
Lednor, A. J., 1982, Magnetic navigation in pigeons: Possibilities and problems, in: Avian Navigation
(F. Papi and H. G. Wallraff, eds.). Springer-Verlag, Berlin, pp. 109-119.
Lowenstam, H. A., 1962, Magnetite in denticle capping in recent chitons (Polyplacophora). Geol. Soc.
Am. Bull. 73:435-438.
Matthews, G. V. T., 1951, The experimental investigation of navigation in homing pigeons, ,. Exp.
BioI. 28:508-536.
Matthews, G. V. T., 1953a, Sun navigation in homing pigeons, J. Exp. BioI. 30:243-267.
Matthews, G. V. T., 1953b, Navigation in the manx shearwater, ,. Exp. Biol. 30:370-396.
Merkel. F. W., and Wiltschko, W., 1965, Magnetismus and Richtungsfinden zugunruhiger Rotkehlchen,
Erithacus rubecula, Vogelwarte 23:71-77.
Mewaldt, 1. R, 1964, California sparrows returns from displacement to Maryland, Science 146:941-
942.
Michener, M. c., and Walcott, c., 1967, Homing of single pigeons-Analysis of tracks, J. Exp. Bioi.
47:99-131.
Moore, F. R, 1977, Geomagnetic disturbance and the orientation of nocturnally migrating birds, Sci-
ence 196:682-684.
Moore, F. R, 1978, Sunset and the orientation of a nocturnal migrant bird, Nature 274:154-156.
Moore, F. R, 1980, Solar cues in the migratory orientation of the savannah sparrow, Passerculus
sandwichensis, Anim. Behav. 28:684-704.
Papi, F., 1976, The olfactory navigation system of the homing pigeon, Verh. Dtsch. Zool. Ges. 69:184-
205.
Papi, F., 1982, Olfaction and homing in pigeons: Ten years of experiments, in: Avian Navigation (F.
Papi and H. G. Wallraff, eds.). Springer-Verlag, Berlin, pp. 149-159.
Papi, F., loale, P., Fiaschi, V., Benvenuti, S. and Baldaccini, N. E., 1978, Pigeon homing: Cues detected
during the outward journey influence initial orientation, in: Animal Migration, Navigation, and
Homing (K. Schmidt-Koenig and W. T. Keeton, eds.). Springer-Verlag, Berlin, pp. 65-77.
Parrish, J., Benjamin, R, and Smith, R, 1981, Near-ultraviolet reception in the mallard, Auk 98:627-
628.
480 Chapter 22
Quine, D. B., and Kreithen, M. 1., 1981, Frequency shift discrimination: Can homing pigeons locate
infrasounds by Doppler shifts?, J. Compo Physiol. 141:153-155.
d'un conditionnement nociceptif, J. Physiol. (Paris) 60:85-92.
Sauer, E. F. G., 1957, Die Sternenorientierung nachtlich ziehender Grasmiicken (Sylvia atricapilla,
borin und curruca), Z. Tierpsychol. 14:29-70.
Schmidt-Koenig, K., 1960, Internal clocks and homing, Cold Spring Harbor Symp. Quant. BioI. 25:389-
393.
Schmidt-Koenig, K., and Schlichte, H. J., 1972, Homing in pigeons with impaired vision, Proc. Natl.
Acad. Sci. USA 69:2246-2247.
Schmidt-Koenig, K., and Walcott, C., 1978, Tracks of pigeons homing with frosted lenses, Anim. Behav.
26:480-486.
Schreiber, B., and Rossi, 0., 1976, Correlation between race arrivals of homing pigeons and solar
activity, Boll. Zoo1. 43:317-320.
Southern, W. E., 1971, Gull orientation by magnetic cues: A hypothesis revisited, Ann. N.Y. Acad.
Sci. 188:295-311.
Southern, W. E., 1972a, Influence of disturbances in the earth's magnetic field on ring-billed gull
orientation, Condor 74:102-105.
Southern, W. E., 1972b, Magnets disrupt the orientation of juvenile ring-billed gulls, BioScience
22:476-479.
Southern, W. E., 1978, Orientational responses of ring-billed gull chicks: A re-evaluation, in: Animal
Torre-Bueno, J. R, 1976, Temperature regulation and heat dissipation during flight in birds, J. Exp.
BioI. 65:471-482.
van Riper, W., and Kalmbach, E. R, 1952, Homing not hindered by wing magnets, Science 115:577-
578.
Viehmann, W., 1979, The magnetic compass of blackcaps (Sylvia atricapilla), Behaviour 68:24-30.
Viguier, c., 1882, Le sens de l'orientation et ses organes chez les animaux et chez l'homme, Rev.
Phil. 14:1-36.
Visalberghi, E., and Alleva, E., 1979, Magnetic influences on pigeon homing, BioI. Bull. 156:246-256.
bridge, Mass.
Wagner, G., 1976, Das Orientierungsverhalten von Brieftauben im erdmagnetisch gestorten Gebiete
des Chasseral, Rev. Suisse Zoo1. 83:883-890.
igation (F. Papi and H. G. Wallraff, .eds.), Springer-Verlag, Berlin, pp. 338-343.
Walcott, C., 1974, The homing of pigeons, Am. Sci. 62:542-552.
Walcott, C., 1977, Magnetic fields and the orientation of homing pigeons under the sun, J. Exp. BioI.
70:105-123.
Walcott, C., 1978, Anomolies in the earth's magnetic field increase the scatter of pigeon's vanishing
Walcott, C., 1980, Homing-pigeon vanishing bearings at magnetic anomalies are not altered by bar
magnets, J. Exp. BioI. 86:349-352.
Walcott, C., 1982, Is there evidence for a magnetic map in homing pigeons?, in: Avian Navigation (F.
Walcott, C., and Green, R P., 1974, Orientation of homing pigeons altered by a change in the direction
Wallraff, H. G., 1970, Ober die Flugrichtungen verfrachteter Brieftauben in Abhangigkeit vom Hei-
matort und vom Ort der Freilassung, Z. Tierpsychol. 27:303-351.
Wallraff, H. G., 1978, Preferred compass directions in initial orientation of homing pigeons, in: Animal
Birds 481
Wallraff, H. G., 1979, Goal-oriented and compass-oriented movements of displaced homing pigeons
after confinement in differentially shielded aviaries, Behav. Ecol. Sociobiol. 5:201-225.
Wallraff, H. G., 1980a, Does pigeon homing depend on stimuli perceived during displacement? 1.
Experiments in Germany, J. Compo Physiol. 139:193-201.
Wallraff, H. G., 1980b, Olfaction and homing in pigeons: Nerve-section experiments, critique, hy-
potheses, J. Compo Physiol. 139:209-224.
Wallraff, H. D., Foa, A., and Ioale, P., 1980, Does pigeon homing depend on stimuli perceived during
displacement? II. Experiments in Italy, J. Compo Physiol. 139:203-208.
Wallraff, H. G., Benvenuti, S., Papi, F., and Gould, J. L., 1982, The homing mechanism of pigeons,
Nature 300:293-294.
Waterman, T. H., 1981, Polarization sensitivity (H. Autrum, ed.), in: Handbook of Sensory Physiology,
Springer-Verlag, Berlin pp. 281-469. Volume VII/6B.
Wiltschko, R, and Wiltschko, W., 1978, Evidence for the use of magnetic outward-journey information
in homing pigeons, Naturwissenschaften 65:112-113.
Wiltschko, R, and Wiltschko, W., 1981, The development of sun compass orientation in young homing
pigeons, Behav. Ecol. Sociobiol. 9:135-141.
Wiltschko, R, Wiltschko, W., and Keeton, W. T., 1978, Effect of outward journey in an altered magnetic
field on the orientation of young homing piegons, in: Animal Migration, Navigation, and Homing
(K. Schmidt-Koenig and W. T. Keeton, eds.), Springer-Verlag, Berlin, pp. 152-161.
Wiltschko, R, Nohr, D., and Wiltschko, W., 1981, Pigeons with a deficient sun compass use the
Wiltschko, W., 1968, Ober den Einfluss statischer Magnetfelder auf die Zugorientierung der Rot-
kehlchen (Erithacus rubecula), Z. Tierpsychol. 25:537-558.
by migrating European robins (Erithacus rubecula), in Animal Orientation and Navigation (S. R
Galler, K. Schmidt-Koenig, G. J. Jacobs, and R E. Belleville, eds.), NASA SP-262, U.S. Government
Printing Office, Washington, D.C., pp. 569-578.
pp. 302-310.
Wiltschko, W., and Gwinner, E., 1974, Evidence for an innate magnetic compass in garden warblers,
Naturwissenschaften 61:406.
Wiltschko, W., and Wiltschko, R, 1972, Magnetic compass of European robins, Science 176:62-64.
Wiltschko, W., and Wiltschko, R, 1975a, The interaction of stars and magnetic field in the orientation
system of night migrating birds. I. Autumn experiments with European warblers (genus Sylvia),
Z. Tierpsychol. 37:337-355.
Wiltschko, W., and Wiltschko, R, 1975b, The interaction of stars and magnetic field in the orientation
system of night migrating birds. II. Spring experiments with European robins (Erithacus rubecula),
Z. Tierpsychol. 39:265-282.
Wiltschko, W., and Wiltschko, R, 1976, Interrelation of magnetic compass and star orientation in
night-migrating birds, J. Compo Physiol. 109:91-99.
Wiltschko, W., and Wiltschko, R, 1981, Disorientation of inexperienced young pigeons after trans-
portation in total darkness, Nature 291:433-434.
Wiltschko, W., and Wiltschko, R, 1982, The role of outward journey information in the orientation
of homing pigeons, in: Avian Navigation (F. Papi and H. G. Wallraff, eds.), Springer-Verlag, Berlin,
Wiltschko, W., Wiltschko, R, and Keeton, W. T., 1976, Effects of a "permanent" clock-shift on the
orientation of young homing pigeons, Behav. Ecol. Sociobiol. 1:229-243.
Wiltschko, W., Wiltschko, R, Keeton, W. T., and Madden, R, 1983, Growing up in an altered magnetic
field affects the initial orientation of young homing pigeons, Behav. Ecol. Sociobiol. 12:135-142.
Windsor, D. M., 1975, Regional expression of directional preferences by experienced homing pigeons,
Anim. Behav. 23:335-343.
Wright, A. A., 1972, The infleunce of ultraviolet radiation on the pigeon's color discrimination, J. Exp.
Anal. Behav. 17:325-337.
Yeagley, H. L., 1947, A preliminary study of a physical basis of bird navigation,/. Appl. Phys. 18:1035-
1063.
482 Chapter 22
Yeagley, H. 1., 1951, A preliminary study of a physical basis of bird navigation. Part II, J. Appl. Phys.
22:746-760.
Yodlowski, M. 1., Kreithen, M. 1., and Keeton, W. T., 1977, Detection of atmospheric infrasound by
homing pigeons, Nature 265:725-727.
537.
Chapter 23
Magnetic Remanence in Bats
EDWARD R. BUCHLER and PETER J. WASILEWSKI
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 483
2. Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 483
3. Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 484
4. Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 485
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 486
1. Introduction
Evidence which has been accumulating over the past 40 years shows that many species
of temperate zone bats migrate and accurately locate their summer roost sites and winter
hibernacula. They also home reasonably well when artificially displaced (Griffin. 1970;
Baker. 1978). Despite abundant documentation of these behaviors, we still have no ac-
ceptable explanation for how they are accomplished. It is clear that echolocation alone is
not sufficient. The visual capabilities of bats are much more refined than is generally
believed (Childs and Buchler. 1981; Buchler and Childs, 1982). However. it seems likely
that still other orientation cues must be available to them because a disconcerting number
of displaced. blinded bats still have been able to home effectively (e.g .. Mueller. 1966,
1968; Davis and Barbour. 1970). Encouraged in particular by recent research with homing
pigeons (reviewed elsewhere in this volume). we were interested in determining ultimately
if bats might also use the earth's magnetic field as a cue during homing or migration.
Because of the expected complexity and duration of the necessary experiments to test for
the perception of a magnetic field by bats, we chose instead to first search for the presence
of magnetic material.
2. Methods
Measurements were made using a SQUID cryogenic magnetometer that is located at
Goddard Space Flight Center. Greenbelt. Maryland. It is capable of detecting approximately
10- 8 emu. When a live big brown bat (Eptesicus fuscus) was confined in a Mylar tube and
lowered into the magnetometer. it produced substantial magnetic moments. However. the
likelihood of contamination required a more rigorous approach. Before measuring the mag-
netic remanences of four species of bats and two species of other small mammals. each
was carefully prepared using only plastic and glass instruments. The carcasses were
skinned and the digestive tracts removed to minimize contamination from the exterior and
EDWARD R. BUCHLER • ORI, Inc., Rockville, Maryland 20850. PETER J. WASILEWSKI •

NASA, Goddard Space Flight Center. Greenbelt, Maryland 20771.
483
484 Chapter 23
from foodstuffs. The fore and hind limbs were severed at the knee and elbow to eliminate
contaminants on the digits. Each carcass then was rinsed in deionized, demineralized,
triple-distilled water, weighed, and individually frozen in a plastic bag. To measure natural
remanent magnetization (NRM) , a specimen was placed head downward in a previously
demagnetized Mylar cylinder. Any slight remanence remaining in the cylinder was sub-
tracted from the specimen-plus-cylinder values. The NRM of each specimen was monitored
using digital and analog readouts as it was lowered into the magnetometer. Maximum total
moments and moments in the x, y, and z axes were recorded. The dorsoventral plane of
the carcass was the same as the x-y plane of the magnetometer field. Saturated induced
remanence magnetization (sIRM) was measured for all specimens immediately after sat-
uration in a 10-kOe dc field. Each sample was oriented in the same position in the field.
Following saturation, the demagnetization curve was also determined for nearly all car-
casses by measuring the remaining moments after each step of a progressive demagneti-
zation in an ac field using the following sequence: 50, 100, 200, 300, 500, 700, 1000 Oe.
In some instances we determined the demagnetization curve starting with the NRM rather
than sIRM. In an attempt to localize the source of the magnetic material, we decapitated
two of the E. fuscus carcasses after saturation, resaturated the parts, and again measured
the magnetic moments.
Because red blood cells contain iron in the hemoglobin molecules, we also were in-
terested in the influence of blood on the measured remanences. Two whole blood samples
from mice of 0.5 cm 3 each, a volume appropriate for a 5-g carcass, were tested, one fresh
and one frozen.
Finally, we wished to quantify the contamination that would have occurred if the
digestive tracts had not been removed. Several species of insects in addition to bees are
known to have substantial NRMs (Gould et al., 1978; Gould, this volume). Thus, the diges-
tive tracts of insectivorous bats could be serious sources of error.
The following small mammals were used in the study:
1. Six adult big brown bats (Eptesicus fuscus)-not generally distant migrators but
have contributed many impressive homing records
2. Six adult little brown bats (Myotis lucifugus)-well known for their relatively long-
distance migrations and homing ability
3. Two adult Seminole bats (Lasiurus seminolus)-Althought little is known about
the seasonal movements of this bat of the southeastern U.S., it may migrate over
short distances
4. One adult red bat (Lasiurus borealis)-probably migrates farther, on the average,
than any other North American bat
5. Five juvenile white laboratory mice (Mus musculus, Jackson Laboratories strain P-
57)-This species was chosen as a possible "control" before evidence was pub-
lished supporting magnetic sensitivity by, and the presence of magnetic material
in, another rodent, the European woodmouse (Apodemus sylvaticus) (Mather and
Baker, 1981)
6. One adult short-tailed shrew (Blarina brevicauda)-nonmigratory, fossorial, and,
like the bats tested, insectivorous
3. Results
All the animals showed substantial natural remanences, the total moments indicating
a preferential (nonrandom) alignment of magnetic domains. After saturation, all had sIRMs
significantly greater than their NRMs. The sIRM can be used as an indicator of the total
amount of magnetic material present (Table I).
Magnetic Remanence in Bats 485
Table I. Summary of Measurements of Magnetic Remanences

Mean magnetic moments (x 10 - 6 emu)
Mean carcass RM after

weight (g) demagnetization
Species N (range) NRM (range) sIRM (range) (range)
Eptesicus fuscus 6 7.8 (6.5-9.0) 0.98 (0.21-2.7) 15 (8.6-21) 2.2 (1.4-2.8)

Myotis lucifugus 6 3.4 (2.9-3.7) 0.42 (0.15-0.91) 6.5 (1.8-19) 1.8 (0.77-4.4)
Lasiurus seminolus 2 4.9 (3.9-5.8) 2.8 (2.3-3.2) 18 (14-21) 2.7 (2.5-2.9)
Lasiurus borealis 1 4.8 0.66 20
Mus musculus 5 5.5 (5.3-5.7) 0.68 (0.45-0.88) 2.1 (0.96-3.2) 0.79 (0.41-1.2)
Blarina brevi cauda 1 8.0 1.2 21
Table II. Measurements of Head and Torso sIRM Values for Eptesicus fuscus
Mean magnetic moments (x 10- 6 emu)
Intact Head after Torso after Head Torso

No. carcass decapitation decapitation Sum resaturated resaturated Sum
1 9.5 0.7 + 4.3 5.0 0.9 + 10 10.9

2 21 1.3 + 5.6 6.9 6.5 + 11 17.5
The decapitation experiments with E. fuscus revealed that the torso contained sub-
stantially more magnetic material than the head. The act of decapitation reduced the total
remanence of the head plus torso to approximately half that of the intact carcass. The total
moments of the subsequently resaturated head and torso were about the same as the original
sIRM for each intact carcass (Table II).
A plotting of the coercive field of magnetic particles (course of force) for one E. fuscus
yielded a zero-crossing point of 350. This suggests that the magnetic material is magnetite
with a particle size of a few micrometers.
As one might predict, fresh blood had no magnetic remanence. However, the frozen
sample, while having no NRM, had an sIRM of 1.6 x 10- 6 emu. This is sufficient to account
for most of the relatively small increase in magnetic moments of the saturated mouse
carcasses but not for the larger increases in the other species.
The insect-filled digestive tract of an E. fuscus had an NRM of 3.8 X 10- 6 emu, greater
than that of a carcass of the same species, and an sIRM of 1.5 X 10- 5 emu, equal to the
mean sIRM value of a carcass for this species. Measurement of the digestive tract of a short-
tailed shrew (B. brevicauda). also insectivorous, produced similar results: an NRM of 2.3
X 10- 6 emu and an sIRM of 1.1 X 10- 5 emu.
4. Discussion
The magnetic remanences of the small mammals measured in this study are very
similar to those obtained for homing pigeons by Walcott et a1. (1979) who reported NRMs
of 10- 7 to 10- 6 emu and sIRMs of 10- 6 to 10- 5 emu. The values are also of the same order
of magnitude as those reported for honeybees by Gould et a1. (1978) and for the heads of
woodmice by Mather and Baker (1981).
486 Chapter 23
However, the decapitation experiments suggest potentially important differences be-

tween bats and pigeons or woodmice regarding the location of the magnetic material. In
the latter two species it is probably concentrated in only one or a few areas in the head.
In E. fuscus the measurements indicate that while some of the material is in the head, the
majority resides in the torso. At this time it is unknown whether the material, probably
magnetite, is present in two or more discrete masses or dispersed more diffusely. Even if
the torso contains more magnetic material, there is still more than enough present in the
head to provide the basis for a very sensitive magnetoreceptor system (see Yorke, 1981).
The measurement of a substantial sIRM, but no NRM, for frozen blood samples suggests
that part of the total sIRM measured for frozen organisms with hemoglobin-based blood
may be attributable to the iron in the red blood cells. This is a potentially significant source
of error if the blood content in the frozen sample is relatively high and the amount of other
magnetizable material is low. Because this is the first report of sIRM in frozen blood and
was determined using only one sample, the validity of the results is questionable and
inadvertent contamination may have been the cause. The high values for remanences of
the digestive tracts containing insect remains illustrate the importance of considering nat-
ural sources of contamination and the method of preparation when making such meas-
urements.
Demonstrating the presence of magnetizable material is only one step in the experi-
mental procedures necessary to establish that an organism uses magnets to detect the earth's
magnetic field and orients relative to it. Thus far we have not been able to demonstrate
that bats can sense a magnetic field. In one series of laboratory experiments we attempted
unsuccessfully to condition two stationary, restrained E. fuscus to associate the reversal
of a surrounding earth-strength magnetic field generated by Helmholtz coils with a sub-
sequent electric shock.
Because magnetic perception may require the bat to be flying, we next tested five E.
fuscus individually as they flew freely in a large, darkened, L-shaped room. They were
flown about 1 hr/day for several weeks and established a predictable flight path about the
room. When flying through the narrower leg of the room, they passed through a 2 x 2 x
2-m Rubens coil (Rubens, 1945) which ptoduced a fairly uniform field of about 0.5 Gin
the horizontal axis. When a bat was in flight at the far end of the room, the coil was activated
occasionally and, as the bat flew near and through the coil, we were alert for any behavioral
changes suggesting detection. The behavior of the bats was not noticeably different when
the coil was generating a field. Perhaps the most critical problem with these experiments
was one of motivation. In the first instance, the stress of restraint may have been too extreme
and, in the second case, the need to attend to a subtle cue was absent.
The experiments necessary to test for a magnetic sense in bats should be performed
on two fronts to gather electrophysiological evidence and field evidence of a predictable
response. The former implies localization of the magnetic material and the details of in-
nervation. The latter requires that the experimenter control for the many acoustical and
visual cues to which bats attend while they are in flight in a natural environment. With
time and careful investigation, we suspect that demonstrations of magnetic material in
animals and their use of it in orientation and migration will be common.
ACKNOWLEDGMENT. This research was funded by a National Science Foundation grant to the
first author while at the Department of Zoology, University of Maryland, College Park.
References
Baker, R R, 1978, The Evolutionary Ecology of Animal Migration. Holmes & Meier, New York, pp.
556-603.
Magnetic Remanence in Bats 487
Buchler, E. R, and Childs, S. B., 1982, Use of the post-sunset glow as an orientation cue by big brown
bats (Eptesicus fuscus), J. Mammal. 63:243-247.
Childs, S. B., and Buchler, E. R, 1981, Perception of simulated stars by Eptesicus fuscus (Vesperti-
lionidae): A potential navigational mechanism, Anim. Behav. 29:1028-1035.
Davis, W. H., and Barbour, R W., 1970, Homing in blinded bats (Myotis sodalis), J. Mammal. 51:182-
184.
Griffin, D. R, 1970, Migrations and homing of bats, in: Biology of Bats, Volume 1 (W. A. Wimsatt,
ed.), Academic Press, New York, pp. 233-264.
201:1026-1028.
Mueller, H. C., 1966, Homing and distance-orientation in bats, Z. Tierpsychol. 23:403-421.
Mueller, H. C., 1968, The role of vision in vespertilionid bats, Am. MidI. Nat. 79:524-525.
Rubens, S. M., 1945, Cube-surface coil for producing a uniform magnetic field, Rev. Sci. Instrum.
16:243-245.
Walcott, C., Gould, J. L., and Kirschvink, J. L., 1979, Pigeons have magnets, Science 205:1027-1029.
537.
Chapter 24
Biomineralization of Magnetite in
Cetaceans
GORDON B. BAUER, MICHAEL FULLER, ANJANETTE PERRY,
J. ROBERT DUNN, and JOHN ZOEGER
1. Introduction............................................................................................... 489
2. Behavioral Studies....................................................................................... 490
2.1. Correlational Studies.............................................................................. 490
2.2. Laboratory Experiments........................................................................... 492
3. Anatomical Studies of Cetaceans...................................................................... 495
3.1. Common Dolphin, D. delphis.................................................................... 495
3.2. Cuvier's Beaked Whale, Z. cavirostris.......................................................... 497
3.3. Humpback Whale, M. novaeangliae ............................................................ 498
3.4. Atlantic Bottlenose Dolphin, T. truncatus..................................................... 499
3.5. DaB's Porpoise, P. dalli ........................................................................... 500
3.6. Summary............................................................................................ 501
4. Conclusion................................................................................................ 503
References ............................ , ..................................................... " ., .. .. .. .. .. . 505
1. Introduction
Living cetaceans comprise two suborders, Mysticeti (the baleen whales) and Odontoceti
(the toothed whales). Within both groups are species which migrate over extended areas
(reviews in Kinne, 1975; Baker, 1978). In general, the mysticetes are more predictable in
their migration patterns and appear to range over a larger area. Although much remains
to be learned about specific migration routes, it is probably safe to say that most mysticetes
follow a similar pattern. They feed during the summer in polar and subpolar waters and
breed in temperate to tropical waters in the winter. Bowhead whales, which are predom-
inantly arctic in distribution, and Bryde's whales, which are primarily tropical, appear to
be the only exceptions. The migrations between feeding and breeding grounds can be
extensive, e.g., 5000 miles in the case of the gray whale (Pike, 1962). They can also be
remarkable in their precision, e.g., humpback whales are able to locate the Hawaiian Islands
after a journey over thousands of miles of open ocean on their southern journey from Alaska
GORDON B. BAUER • Department of Psychology, University of Hawaii, Honolulu, Hawaii

96822. MICHAEL FULLER and J. ROBERT DUNN • Department of Geological Sciences, Uni-
versity of California, Santa Barbara, California 93106. ANJANETTE PERRY • Department of
Oceanography, University of Hawaii, Honolulu, Hawaii 96822, JOHN ZOEGER • Los Angeles
County Museum of Natural History, Los Angeles, California 90007.
489
490 Chapter 24
(Baker et 01., 1982; Darling and Jurasz, 1983; Darling and McSweeny, 1983). The odon-
tocetes have less spectacular migrations, but they still travel over substantial stretches of
ocean.
Multiple sensory cues are no doubt utilized in navigation, but direct evidence is min-
imal (Kinne, 1975; Baker, 1978; Madsen and Herman, 1980). Potential cues include light,
sounds, odors (in mysticetes only), tastes, thermal gradients, wind, and currents. Common
dolphins, Delphinus delphis, in the Mediterranean Sea have been observed to swim east-
erly in the morning and westerly in the evening, presumably using the sun as a cue (Busnel
and Dziedzic, 1966; Pilleri and Knuckey, 1968). Spy-hopping, in which a whale lifts its
head vertically out of the water, suggests a means to orient toward surface features (Gilmore,
1960). Evans (1971) has observed a close correlation between herd movements of Delphinus
and sea-bottom topography (maximum depth 260 m).
Although various sensory mechanisms are present in cetaceans, special navigational
problems occur when animals are in a pelagic environment which can minimize available
cues. Dawbin (1966) has observed that the migration routes of humpback whales cannot
be consistently related to bottom topography, direction of ocean currents, or the nature of
water masses. There are few distinctive surface features and the bottom topography is
frequently too distant to provide precise cues. During overcast conditions, celestial features
are unavailable. In an ocean environment of limited navigational cues, the earth's magnetic
field could provide necessary information for orientation. There is evidence that a number
of marine vertebrates can detect magnetic fields (Walker et 01., Chapter 20, this volume).
Recently, Klinowska (in press) has reported evidence that cetacean strandings in the United
Kingdom all occur at areas of minimum magnetic field. This evidence has important im-
plications for the ability of cetaceans to use magnetic fields and use them to guide mi-
grations.
Our own research on magnetic detection by cetaceans followed two tracks. In our
behavioral studies, we took a preliminary look at the relationship of geomagnetic events
with cetacean strandings and migrations, but concentrated our efforts on a variety of con-
ditioning experiments to investigate magnetic perceptual abilities in two bottlenose dol-
phins, Tursiops truncatus. In our anatomical studies, we looked for a potential magnetic
receptor in a diverse group of subjects, four odontocetes and a mysticete: two species of
dolphins, the common dolphin (D. delphis), a pelagic animal, and the bottlenose dolphin
(T. truncatus), a coastal species; Cuvier's beaked whale, Ziphius cavirostris, a large, deep-
water species; Dall's porpoise, Phocoenoides dalli, a coastal animal; and a humpback
whale, Megaptera novaeangliae, the only mysticete studied.
2. Behavioral Studies
2.1. Correlational Studies
A possible window on the effect of geomagnetism on cetacean behavior is provided

by terrestrial strandings. Current evidence points to multifactor causes, including disease,
population density, acoustic interference, social cohesion, predator avoidance, and rever-
sion to primitive behavior (Geraci and st. Aubin, 1979). Another factor could be the in-
fluence of geomagnetic anomalies and storms.
The records of cetacean strandings in British waters are carefully documented by the
British Museum (Natural History) back to mid-1913. Klinowska (in press) used these rec-
ords to investigate the relationships between strandings and geomagnetic events. Animals
for which there were adequately accurate records were divided into two groups: those that
were dead at the time of stranding and those that were alive. The locations of dead strand-
Cetaceans 491
ings were positively correlated with tide and current patterns; that is, the animals were
carried passively to shore. On the other hand, live stranding sites did not share any common
oceanographic or geographic characteristics. Live strandings occurred exclusively at areas
of minimum magnetic field. Additionally, Klinowska cited a number of cases which suggest
that cetaceans might be reluctant to cross anomalous areas of high magnetic intensity.
Analysis of species differences in stranding rate indicates that pelagic species were the
most likely to run around.
Kirschvink et 01. (1984) analyzed 212 live stranding events on the East Coast of the
United States. Stranding information was derived from the Smithsonian Institute's com-
puterized data base, Smithsonian Scientific Event Network (SEAN). Kirschvink et 01. found
significant correlations between live strandings of noncoastal species and local magnetic
minima and/or subaverage geomagnetic intensity. They confirmed these findings with a
series of Monte-Carlo simulations. Cornwell-Huston (personal communication, 1984) in-
dependently analyzed 296 East Coast cases drawn from the same data base as Kirschvink
et 01. and came to the same conclusion, i.e., live cetacean strandings occur along magnetic
troughs. Dizon and Bauer (unpublished data) were unable to find any correlation between
magnetic storms, as measured by K indices, and strandings in the Hawaiian Islands. Sim-
ilarly, Cornwell-Huston (personal communication) was unable to find any association be-
tween magnetic storms and strandings along the New England Coast.
Kirschvink et 01. (1984) also reported that preliminary analysis of open-ocean obser-
vations of some dolphin species and finback whales showed those locations to be areas
of low magnetic intensity. They hypothesized that the relationship of cetacean locations
with magnetic minima might be attributable to the relative continuity of lows compared
to highs, which are more discretely situated. This continuity, in turn, could provide ce-
taceans with relatively continuous navigational cues.
Changes in cyclical magnetic characteristics, such as planetary intensity, may provide
cues for the seasonal migrations of some cetaceans. Preliminary analyses of correlations
of whale migrations with monthly magnetic range (ap) indices (Mayaud, 1980) provide
tentative support for this hypothesis. Dawbin (1956) reported on migrations of humpback
whales through Cook Strait, New Zealand, between 1914 and 1955 (excluding 1919 and
1921-1925). Pearson product-moment correlations of the number of whales sighted
weekly with ap indices indicate some significant associations (Bauer, unpublished data).
Specifically, the higher the ap indices for the months preceding arrival, the earlier was
the influx of whales. Migration through the Strait occurred primarily in June and July.
Dawbin (1966) proposed that humpbacks migrate at a rate of about 15° of latitude per
month. He used 60 0 S latitude as the boundary for the southern feeding terminus. Traveling
from this boundary to Cook Strait (40°C) at 15°/month would place the humpback whale
departure time at between 1 and 2 months before arrival at Cook Strait, approximately
April through June. The strongest correlations were for these months when the cues for
migrating departure should be present. The timing of humpback whale migrations shows
considerable year-to-year variation, shifting as much as 3 weeks over a 3-year period (Nish-
iwaki, 1960; Baker and Herman, 1981), and up to 8 weeks over 40 years (Dawbin, 1956).
Previous attempts to correlate these shifts with environmental cues such as day length,
water temperature, or lunar cycles have not been successful (Whitehead et 01., 1982;
Baker et 01., 1985). Cyclical magnetic phenomena may provide some of the sensory
cues utilized by migrating whales and at least partially explain variations in migration
timing.
The association of cetacean strandings and open-water locations with geomagnetic
minima has important implications for magnetic detection and orientation. It suggests that
noncoastal cetaceans are using magnetic cues for navigational purposes. This does not
preclude the possibility that inshore species also detect magnetic fields, but they may rely
more on other sensory mechanisms, such as echolocation and vision, to monitor shallow
492 Chapter 24
coastal areas. Although promising, these correlational studies should be interpreted cau-
tiously. Alternative explanatory variables need to be more thoroughly investigated. These
include topographical features and meteorological conditions, as well as the possibility
that cetacean movements are influenced by tracking fish that use magnetic orientation
cues. Correlations of magnetic range indices and whale migrations are still preliminary
and need to be supported by the same investigations of alternative vtlriables as those sug-
gested for strandings. Additionally, differential whaling of various age and sex classes can
have an effect on migration rates, which can in turn produce spurious correlations with
magnetic phenomena. Future studies should make use of data bases unaffected by whaling.
Analysis should also utilize a more detailed record of magnetic characteristics and whale
movements, e.g., daily as opposed to weekly and monthly measures.
2.2. Laboratory Experiments
Correlational analyses of magnetic anomalies and strandings do not by themselves

establish a causal relationship. Although suggestive, they do not constitute proof of a
magnetic detection capability. Experimental studies are necessary for this purpose. All the
experimental studies of behavior in response to magnetic fields were done with two female
Atlantic bottlenose dolphins, T. truncatus. Both dolphins, Akeakamai and Phoenix, were
approximately 6 years of age and had been in captivity for 4 years at the time of the study.
They had been extensively involved in experimental projects investigating delphinid cog-
nitive capacities. They were housed in a 50-m-diameter cement tank at the Kewalo Basin
Marine Mammal Laboratory, University of Hawaii. For purposes of these experiments, it
is important to note that the structure of the tank was reinforced with iron bars which
produced some distortion of the ambient magnetic field.
The first behavioral experiment was modeled after an experiment described by Kalmijn
(1978). Kalmijn introduced a local magnetic field into a tank by means of a small induction
coil. Leopard sharks, Treakis semifaciata, were observed to veer off toward the center of
the tank upon swimming into the region of the induced field. Our similar experiment
yielded no such dramatic results. An induction coil was placed at various locations around
the sides of the tank such that it altered the horizontal component of the ambient magnetic
field. In no location was there any obvious response indicating that a field had been de-
tected.
We next attempted a series of conditioning experiments. In all the experiments, we
passed a dc current through a I-m coil suspended above the tank which generated a mag-
netic field that added to the vertical component of an ambient field of 0.37 G (3.7 x 10- 5
T). With this coil arrangement, the dolphins encountered a graded field with a maximum
overall intensity of about 0.60 G.
We tried a variety of experimental formats without any indication of a magnetic dis-
crimination. Briefly, the experiments were as follows. (1) Two-choice discrimination: The
subject swam under the induction coil, then had to press one response paddle if a local
anomaly was present, i.e., current was passing through the coil, and another paddle, dif-
ferent in location and form, if the local anomaly was absent. (2) Two-choice discrimination
with an acoustic cue: The procedure was the same as for experiment 1, except that an
acoustic cue or "name" was paired with the magnetic stimulus. The acoustic cue was then
faded (attenuated) until only the magnetic cue remained. The fading technique used in
this and the following experiments was based on a cross-modal transfer technique (Herman,
1980). Under such a procedure, it has been demonstrated that a performance limitation
related to one sensory modality can be overcome by pairing the limiting modality with a
dominant one. For example, Tursiops, despite having good visual acuity, have great dif-
ficulty with cognitive tasks requiring a discrimination between static visual stimuli. They
Cetaceans 493
are quite proficient at the same tasks if acoustic stimuli are used. If acoustic stimuli are
paired uniquely with visual stimuli and subsequently faded, performance on the visual
tasks greatly improves. It was thought that this technique might be helpful in uncovering
magnetic detection abilities. (3) "Go/no go" discrimination with a visual (light) cue: In the
golno go task, only one paddle was used. The subject was expected to press the paddle in
the presence of a lotal anomaly and refrain from pressing in the absence of such a field.
A light was paired with the magnetic stimulus and faded. (4) "Go/no go" with a visuall
acoustic cue: This was the same procedure as for experiment 3, but a cue was used which
could provide the subject with both visual and acoustic (through echolocation) informa-
tion. One of two hoops constructed of PVC pipe was placed under the induction coil, a
hoop with pipes of large diameter in the presence of the local anomaly and a hoop with
narrow-diameter pipes in the absence of the field. The difference in pipe diameter was
faded until only one size pipe was used, leaving the magnetic stimulus as the only guide
for correct responding.
We made a variety of experimental changes within the above formats including varying
the strength of the local field, changing the direction of movement of the subject (Le., for
some experiments the subject moved along the north-south axis of the tank, for others
along the east-west axis), and increasing movement with respect to the coil. Although
movement was increased, both total distance traveled through the local field and speed,
the dolphins were always required to move in a straight line.
Although a null hypothesis cannot be proven, it is frequently tempting to accept it
after a series of experiments that offer no contradicting evidence. In this case, there are
some good reasons for avoiding a hasty judgment about cetacean magnetic perception. A
strong case can be made for the magnetic detection abilities of pigeons based on field
experiments (Keeton, 1974; Presti, this volume), yet laboratory conditioning studies have
rarely supported a magnetic detection hypothesis (Reille, 1968; Bookman, 1978). These
few supportive experiments have proven difficult to replicate. When numerous condi-
tioning experiments fail to confirm an ability demonstrated in other formats, the meth-
odology has to be questioned. One possibility is an inadequate stimulus. Wiltschko (1978)
has shown the importance of using fields which are within normal earth-strength bound-
aries. However, many experiments have used such fields and still have been unsuccessful.
Bookman (1978), after a successful experiment in conditioning magnetic discrimination,
suggested that freedom of movement might be the key. He pointed out that previous ex-
periments had indicated that subjects were constrained in their movement. Even within
his own experiment, only on those trials where the subjects "fluttered" was performance
above chance. Keeton (1974), noting that most unsuccessful studies made use of short
pulses, proposed that long pulses might be necessary.
It is difficult to evaluate experiments with negative results, for the significant reason
that they are usually not reported, or when reported are frequently quickly discussed with-
out attention to methodological detail that might be helpful in pointing research in po-
tentially more promising directions. Fortunately, two review articles, de Lorge and Marr
(1974) and Sheppard and Eisenbud (1977), describe negative magnetic conditioning studies
with a variety of species including pigeons, rats, and squirrel monkeys. Both reviews in-
dicate that the procedures did not allow much subject movement.
Two successful conditioning studies, one with salamanders (Phillips, 1977; Phillips
and Adler, 1978) and the other with yellowfin tuna (Walker et a1., Chapter 20, this volume),
provide some instructive methodology. Although rather different in format, they have three
characteristics in common: (1) subjects had considerable freedom of movement; (2) they
remained in the experimental field for relatively long periods; (3) they were exposed to
relatively homogeneous fields. (In the tuna studies the experimental tank was encircled
by a coil which introduced a gradient near the edges. However, most of the tank had a
uniform field.) Our ignorance of the mechanisms of magnetic detection includes ignorance
494 Chapter 24
of the appropriate orienting responses necessary for adequate perception. Experiments that
constrain the subject in time and place may be putting significant limits on appropriate
orientation. Additionally, there is some indication that the magnetic sensor may be some-
what slow in responding to new field conditions (Wiltschko, 1978). Another consideration
is that without contrary empirical evidence that an animal does not perceive a field in-
directly, we cannot rule out detection through electrical induction. That is, the subject
may perceive the electrical current it generated by moving through a magnetic field. The
magnetic perception mechanism may have evolved to be operative in a moving organism
and inoperative in a stationary organism, either because of receptor or information-pro-
cessing (cognitive) characteristics. This can apply to an organism using a direct detection
mechanism as well as an induction mechanism.
The importance of considering homogeneous fields, duration of exposure, and subject
movement lies in the ecological validity of the experimental situation, i.e., the approxi-
mation to the natural conditions associated with the evolution of a magnetic detector. None
of the models we are familiar with preclude a detector of graded fields. However, it is
possible that an animal's cognitive apparatus may preclude response to particular field
characteristics. A possible explanation for the difficulties encountered in training re-
sponses to magnetic fields can be found in the literature on constraints and specializations
in learning and memory (Shettleworth, 1972; Seligman and Hager, 1972; Hinde and Ste-
venson-Hind, 1973; Herman, 1980). For many years, behaviorists operated under the equi-
potentiality premise; it was assumed that general learning processes existed in all animals,
and that the specific organisms, reinforcers, responses, and stimuli were not especially
important. [See Ossenkopp and Barbeito (1978, p. 263) for review of learning constraints
and magnetic detection in experiments with birds.] This theoretical orientation is depicted
rather clearly by Teitelbaum (1966, p. 567): " ... in any operant situation, the stimulus,
the response, and the reinforcement are completely arbitrary and interchangeable. No one
of them bears any biologically built-in fixed connection to the others." In recent years, a
large number of studies have been reported that run contrary to the equipotentiality prem-
ise. These studies indicate that, just as morphological characteristics have evolved through
natural selection, so have cognitive characteristics. The consequence of this evolution
appears to be that an animal's information-processing characteristics are adapted to its
specific environmental niche. Bottlenose dolphins live in an environment which is fre-
quently visually restricted; a heavy premium has been placed on acoustic perception of
the environment through echolocation and passive listening. Nevertheless, under the equi-
potentiality premise, it would be expected that given their good visual and acoustic acuity,
dolphins should be able to process information equivalently in both modalities. This is
not the case (Herman, 1980). They process acoustic stimuli with relative ease, while having
a serious problem with static visual stimuli. They appear to have a constraint on cognitive
processing of static visual stimuli, stimuli that are probably of minimal importance in their
natural habitat.
There are numerous examples that suggest that an animal's interaction with its natural
environment has a strong influence on the adequacy of stimuli, reinforcers, and responses
incorporated into laboratory experiments. If we assume that the primary function of a
potential magnetic detector in cetaceans is probably the perception of the earth's magnetic
field for use in navigation, laboratory experiments should be designed which best match
geomagnetic characteristics. As this has already been demonstrated to apply to intensity
(Wiltschko, 1978), it might be prudent to assume that it applies to other properties, e.g,
homogeneity. It is probable that those animals that do detect local graded fields in labo-
ratory experiments, such as elasmobranchs (Kalmijn, 1978), use their magnetic perception
capabilities for detection of proximal objects, for example, prey. Under any circumstances,
graded fields and restricted subject movement have not yielded positive results in con-
ditioning dolphins. A "win/stay lose/shift" strategy suggests that we change procedures.
Cetaceans 495
Other areas of change might include using pelagic subjects rather than the coastal Tursiops
which has more access to a variety of terrestrial and shallow-bottom navigational cues. A
coastal environment may place few magnetic detection demands on a dolphin. The cor-
relational analyses of strandings indicated that predominantly pelagic species were in-
volved. It would be prudent to choose experimental subjects based on prior indications
of a magnetic sense. Another experimental consideration, noted by Kinne (1975), is that
animals may lose their navigational abilities in captivity. It may be important to work with
more recently captured subjects.
3. Anatomical Studies
Various authors have proposed that magnetite could function as part of a transduction
mechanism for detecting magnetic fields (Yorke, 1979, 1981; Kirschvink and Gould, 1981).
This has been demonstrated to be the case with magnetotactic bacteria (Frankel et al.,
1979). For more complex organisms, the function of magnetite remains hypothetical, albeit
an intriguing possibility.
The search for magnetic material in the tissues of cetaceans presents some unique
problems in obtaining magnetically clean preparations, problems not encountered in con-
siderably smaller animals. Because of the enormous size of mysticetes and many odon-
tocetes, dissections may have to be done in situ. Thick layers of blubber, muscle, and bone
make entry into the cranial cavity difficult to manage without introducing magnetic con-
taminants. Our methodology has progressively evolved and continues to evolve toward
cleaner procedures. The following discussion is organized according to the order in which
the species were studied so as to mark the changes in dissection technique and the probable
correlative lessening of artifactual results due to contamination. Anatomical work was done
with previously frozen heads. All magnetic measurements were made with SQUID mag-
netometers (Goree and Fuller, 1976).
The cetacean cranial and brain structure is unique; the organization differs dramati-
cally from other mammals (Morgane and Jacobs, 1972). As a consequence, the normal
orientation axes of the mammalian brain become somewhat confused. Figure 1 indicates
the orientation of a stylized cetacean brain with the dura mater intact. The olfactory bulbs,
which are present in mysticetes but absent in odontocetes, have been omitted. The dura
mater is a thick, inelastic brain envelope made up of two layers: an inner, meningeal and
an outer, periosteal layer (Millen and Wollam, 1962). In our discussion, the layers are
considered as a unit. We discuss two dural processes: falx cerebri and tentorium cerebelli.
Formally, they are tissues derived from the meningeal layer which separate the cerebral
hemispheres and the cerebrum and cerebellum, respectively. We use the terms in a less
precise manner to refer to an area approximately 0.5 cm wide located contiguously with
the dural processes (Fig. 1). The dura is richly innervated and vascularized.
3.1. Common Dolphin, D. delphis
The first reported magnetite found in cetaceans and apparently in mammals was dis-
covered in the common dolphin, D. delphis (Zoeger et a1., 1981). A Stryker autopsy saw
was used to cut the head into coronal sections which were then further divided parasa-
gittally. Four dolphins were satisfactorily dissected for analysis. All tissue samples showed
a detectable natural remanent magnetization (NRM), but an especially strong moment was
found in the dura mater near the juncture of the falx cerebri and-tentorium Å É ê ú Ä É ä ä á K = Tissues
were found in this area with a moment of approximately 2 x 10- 5 G cm 3 or 2 x 10- 8 A
496 Chapter 24
m 2 (1 G cm 3 = 10- 3 A m2 ). Acquisition of remanent magnetization and the demagnetization

of saturated remanence indicated that the material was magnetically soft, i.e., multi domain.
The material was substantially demagnetized by a field on the order of 10- 2 Oe (1 Oe -80
Aim). A particle removed from one of the dolphins was further analyzed and found to be
predominantly iron with no detectable nickel or chromium. A low-temperature test for
magnetite (Kobayashi and Fuller, 1968) indicated that the material was multi domain mag-
netite. Scanning electron microscopy revealed numerous fibers around the magnetic par-
ticle which the authors speculated were nerve fibers, although tissue deterioration pre-
cluded positive identification.
Cerebral multidomain magnetite has not been reported for other species, e.g., fish
(Walker et 01., Chapter 20, this volume), turtles (Perry et 01., this volume), and pigeons
(Walcott et 01.,1979). In fact, it is considered an indicant of contamination by some authors
(Walker et 01., Chapter 5, this volume). This particle was deeply embedded in tissue. More-
over, the absence of nickel and chromium in the particle argues against contamination
from the Stryker saw and stainless steel scalpel used in the dissection. Nevertheless, in
order to lessen problems of potential contamination, subsequent studies gradually deem-
phasized the use of metal tools.
3.2. Cuvier's Beaked Whale, Z. cavirostris
A Ziphius was dissected in the open air at the National Marine Fisheries Laboratory,
Honolulu, Hawaii. Blubber, muscle, and other extracranial tissues were removed with steel
knives and portions of occipital bones were removed with a Stryker saw in order to gain
access to the cranial cavity. Thereafter, plastic and glass knives were used to remove the
dura. No other tissues were removed for analysis. The dura was rinsed in tap water, then
stored in alcohol. Before analysis, the tissues were rinsed in distilled water. Large pieces
of dura were analyzed for saturation isothermal remanent magnetization (sIRM), and then
smaller pieces were checked for more exact loci. A strong magnetic moment was found
in a narrow strip of the anterior falx cerebri (Table I). The magnetic moment was found
to be progressively weaker in the dorsal direction. The principal magnetic area was a 2-
to 3-cm longitudinal strip at the juncture of the ventral and anterior falx cerebri with a
moment of 2.91 x 10- 5 G cm3 • Most of the lateral and dorsal dura covering the cerebral
cortex was magnetically quiet. The remaining dura covering the posterior and ventral por-
tions of the brain, including cerebellum, hindbrain, and midbrain structures, showed strong
saturation remanence similar to the anterior falx cerebri. One 0.2-g (dry weight) section
had a moment of 1.84 x 10- 5 G cm3 , or a magnetization per unit mass of 9.2 x 10- 5 G
cm 3 /g or A m2 /kg (1 G cm 3 /g = A m2 /kg). A small piece of midline occipital bone exhibited
a moment of 9.01 x 10- 6 G cm3 •
AF demagnetization of saturated tissues suggested some differences in coercivity
among different areas of the dura, as well as bone (Table V). Coercivity of the anterior strip
was much greater than for the posterior areas. The median AF demagnetizing field for the
anterior area was between 200 and 275 Oe vs. 100 and 125 Oe for the posterior dura and
occipital bone.
, Results were tempered by the finding of some obvious contaminants; one large mag-
netite particle found before rinsing in distilled water had a diatom attached to it. Although
later rinsing did seem to remove any large particles, the initial contamination is cause for
concern. There were a variety of means for entry of contaminants, including air currents
in the open dissection environment and off the external surface of the Ziphius. Addition-
ally, the use of a saw to cut through the occipital bones introduced another possibility of
contamination. A number of considerations, however, suggest that our findings, although
compromised, may be an accurate representation of the natural state. The areas cut by the
Cetaceans 497
Table I. Topological Representation of the Dura of

Ziphius cavirostris by sIRM (Units 10- 6 G cm 3 )G.b
Left Medial Right
Ventral anterior NM c 19.10 0.91

(orbital lobes)
Anterior NM 14.00 0.95
(orbital lobes) 3.12
Dorsal NM 0.57 3.72
(parietal, occipital,
temporal lobes)
Posterior 43.70 18.40 NM
(cerebellum)
Ventral posterior 4.29 12.6 1.25
(midbrain)
a Background noise: 0.48-1.0 x lO-7 Gem'.
b Occipital bone sample: 9.01 x lO-6 Gem'.
C NM, not measured.
saw were relatively small. Most of the areas of significant remanence, except for the dura
over the dorsal cerebellum, were shielded from any particles that the saw might have
introduced. The loci of magnetic areas, the magnitude of remanence, and the coercivity
have parallels with other cetaceans and vertebrates, e.g., fish (Walker et 01., Chapter 20,
this volume), turtles (Perry et 01., this volume), and pigeons (Walcott et 01., 1979). Further
tests such as X-ray diffraction should offer other clues as to the accuracy of the findings.
3.3. Humpback Whale, M. novaeangliae
A male calf, approximately 2 months of age, was dissected at the same open-air site
as the Ziphius. The procedures were exactly the same except that the tissues were preserved
in buffered formalin. There was a detectable remanence in all pieces of dura measured
(Table II). The remanence in a piece of tissue taken from the olfactory bulbs was negligible.
The NRM in the dura gave moments of between 1.14 x 10- 7 and 1.93 x 10- 6 G em 3 • The
largest moments were found in the anterior areas of the falx cerebri, where the sIRM values
for magnetization ranged from 2.5 x 10 5 to 1.0 X 10- 4 G cm 3 /g. Unlike the other adult
cetacean species studied, but similar to the juvenile Phocoenoides discussed below, the
posterior area did not show a strong remanence, sIRM = 1.0 x 10- 6 G cm 3 /g, although
more sampling needs to be done. Also, unlike the other species studied, the coercivity of
the anterior areas seemed to be weak; the median unblocking field occurred between 100
and 150 Oe. However, there were indications of substantial interaction effects. IRM ac-
quisition of one anterior sample showed a median inducing field of 700 Oe; sIRM AF
demagnetization yielded a median unblocking field of 150 Oe. The resulting remanent
coercivity was 325 Oe. Cisowski (1981) has demonstrated that the interaction of closely
spaced single-domain grains retards the IRM acquisition, yielding overestimates of coer-
civity, while sIRM demagnetization is accelerated, yielding underestimates. The results of
the coercivity measures indicated single-domain grains, probably magnetite, of a size,
shape, and proximity similar to those found in chiton teeth. A low-temperature test for
magnetite (Kobayashi and Fuller, 1968) showed a transition indicating single-domain mag-
netite. Given a 50 Oe thermoremanent magnetization (TRM), the tissue was essentially
demagnetized at 575 Oe. This also indicated that the material was probably single-domain
magnetite (Lowrie and Fuller, 1971).
498 Chapter 24
Table II. Topological Representation of the Dura of

Megaptera novaeangliae by NRM (Units 10- 6 G cm 3 )U
Left Medial Right
Ventral anterior 0.53

(orbital lobes) 0.76 b
Anterior 1.93 2.13 c
(orbital lobes) 1.79
Dorsal
temporal lobes) All other areas had a fairly evenly
Posterior distributed remanence, range
(cerebellum) 0.15-0.48
Ventral posterior
(midbrain)
a Background noise has been subtracted from the measurements.
b sIRM = 3.3 x 10- 5 G cm 3 •
C sIRM = 2.6 x 10- 5 G cm 3 .
Again, contamination was a problem. Various large granules with substantial magnetic
moments were found. These sat loosely on the surface of the dura and were most likely
airborne contaminants. A thorough rinsing in distilled water seemed to remove them. The
strong similarity of the magnetic properties of the material found in the dura to other
biological magnetite, in addition to the distinctive frontal location of highest remanence,
suggests that the material is not contamination. Ultimately, X-ray diffraction could resolve
the issue by giving us a measure of the purity of the samples; samples more pure than
geological particles would support a biological origin (Perry et al., this volume; Walker et
al., Chapter 20, this volume).
3.4. Atlantic Bottlenose Dolphin, T. truncatus
The large size of the Ziphius and Megaptera made it difficult to work in a clean en-
vironment. The serious problem of pieces of "beach" finding their way into our prepa-
rations encouraged us to use alternative procedures when we obtained smaller species.
The Tursiops was an adult of unknown sex. The dissection was done indoors. Steel knives
were used to cut off blubber and muscle from around the skull. Thereafter, no metal tools
were used. The cranial cavity was entered by breaking the bones around the occipital
condyle with a rubber mallet. All cutting and tissue removal were done with nonmetallic
tools. Only distilled water was used for rinsing.
Detectable saturation remanence was found throughout the dura, as well as in some
muscle tissues and bone (Table III). A sample of parietal-occipital cerebral tissue showed
a negligible remanence. In the dura, large sIRM values for magnetization were found over
the cerebellum, 3.82 x 10- 5 G cm 3 /g (wet weight), and ventral anterior falx cerebri, 6.99
x 10- 6 cm 3 /g and 1.31 x 10- 5 G cm 3 /g, just to the right of this area. The major portion
of the dorsal dura covering the cerebral cortex had a smaller moment, range 0.51 x 10- 6
to 2.34 X 10- 6 G cm 3 /g. Substantial saturation remanence was found in the cranial muscles,
1.28-6.22 x 10- 5 G cm 3 /g, and occipital condyle, 5.65 X 10- 6 G cm 3 jg. However, the
latter areas came into contact with steel carving knives and were exposed to a number of
possibly contaminated surfaces.
AF demagnetization indicated a greater coercivity of magnetic material from the ven-
tral anterior falx cerebri than from two lateral orbital areas analyzed, with a median AF
Cetaceans 499
Table III. Topological Representation of the Dura of Tursiops truncatus by sIRM (Units
10- 6 G cm 3/g, wet wt)Q,b
Left Medial Right
Ventral anterior 1.73 7.43 6.99 13.07 3.5

(orbital lobes)
Anterior 4.43 2.41 10.66
(orbital lobes)
Dorsal 4.61 2.20 1.43 2.34 3.92
temporal lobes)
Posterior Measurements not 8.34 38.23
(cerebellum) valid
Ventral posterior Measurements not valid
(midbrain)
Q Note that units in this table are expressed as magnetization per unit mass. Background noise has been subtracted
from the measurements.
b Cerebral cortex range: 0.01-0.06 x 10- 6 G cm 3 /g; muscle: 6.22 x 10- 5 G cm 3/g; occipital bone: 3.96 x 10- 6 G
cm 3/g.
demagnetizing field of 275 Oe in the former vs. 150-175 Oe in the latter. There were no
obvious indications of any contamination.
3.5. Dall's Porpoise, P. dalli
Five male Phocoenoides, two juveniles and three adults, were studied. The procedures
were exactly the same as those used for the Tursiops except that before the muscle and
blubber were cut away with metal knives, samples were taken of each tissue using glass
knives. A large magnetic particle, probably contamination, was found sitting loosely on
the dura of one of the adult specimens; that specimen was dropped from the overall anal;
ysis.
A detectable moment (NRM) was found in most tissues, but there was considerable
variability (Table IV). In only one subject, an adult, did the dura covering the cerebrum
have a detectable moment. Relative to other species, the overall NRM of the anterior dura
was not large (mean = 2.87 x 10- 7 G cm 3 ); however, the intensities were, e.g., NRM =
6.71 x 10- 7 G cm 3 /g, sIRM = 2.37 x 10- 5 G cm 3 /g (dry weight) for one adult subject.
There was a difference in remanence in the area of the cerebellum between adults and
juveniles, with a mean moment of 4.75 x 10 7 G cm3 and no detectable remanence, re-
spectively. The mean moment for the ventral dura (with the exception of orbital dura) was
3.26 x 10- 7 G cm 3 (N = 3); one juvenile subject did not have these tissues measured.
Differences among these areas were neither large nor consistent. Muscle had moments
ranging from 1.37 x 10- 7 to 9.51 X 10- 7 G cm 3 , and blubber 1.69 to 2.38 x 10- 7 G cm 3 •
The cerebrum had a very small moment, but the cerebellum had a considerable moment,
one sample yielding a magnetization of 1.14 x 10- 6 G cm 3 /g (dry weight). AF demag-
netization indicated the median unblocking field occurred consistently at between 150
and 200 Oe for tissues in the anterior falx cerebri but averaged less for other tissues (Table
V). Most measurements of saturation remanence indicated magnetic moments between one
and two orders of magnitude greater than NRM. There were two exceptions: sIRM of one
section of posterior dorsal dura and a piece of blubber measured only slightly more than
double the NRM. One implication of these exceptions is that there may be more than one
500 Chapter 24
Table IV. Topological Representation of the Dura of

Phocoenoides dalli by NRM (Units 10- 7 G cm 3 ; Means
± S.D., N = 4)Q
Left Medial Right
Ventral anterior 2.87 ± 0.69 b

(orbital lobes) and
anterior (orbital lobes)
Dorsal c
(orbital, parietal,
occipital, temporal)
Posterior 9.94 ± 4.75 (adults)d
(cerebellum) 0.00 (juveniles)
Ventral posterior 3.26 ± 2.18"
(midbrain)
a Background noise has been subtracted from the measurements.
b Individual sample (adult): NRM = 6.71 x 10- 7/g, sIRM = 2.37 x 10- 51
g (dry weight).
cThree subjects had no detectable NRM; one adult had an NRM of 8.08 x
lO-7. Measurement encompassed the dura from the left and right ventral
anterior and anterior regions in addition to the dorsal region.
d Individual sample (adult): NRM = 5.60 x 10- 7/g, sIRM = 1.74 x lO- 51
g (dry weight).
e N = 3: two adults, one juvenile.
Table V. sIRM AF Demagnetization, Jllo after 200 De (-16,000 Alm)Q

Ziphius Megaptera Tursiops Phocoenoides
Ventral anterior dura (orbital lobes) 0.51 0.34 b 0.60 0.47 (juvenile)
or anterior dura (orbital lobes) 0.43 0.29 0.44 (juvenile)
0.46 (adult)
0.49 (adult)
Dorsal dura (parietal, occipital, 0.46
temporal lobes) 0.26
Posterior dura (cerebellum) 0.25
0.45
Cerebellum 0.25 (adult)
0.47 (adult)
Occipital bones 0.35
Blubber 0.32
a / = magnetic moment after demagnetization; /0 = initial magnetic moment.
b Remanent coercivity = 325 De (-26.000 Aim).
type of magnetic substance present, e.g., hematite which requires greater energy than mag-
netite to attain full saturation.
3.6. Summary
Much of our effort was directed toward finding multi domain material similar to that
found in the analyses of Delphinus tissues, Le., we concentrated on obtaining measures
of NRM with considerably less frequent measurement of sIRM. This effort appears to have
Cetaceans 501
DORSAL
ú J J c ~ ä ñ =Cerebo
A
VENTRAL
ANTERIOR
Tentorium Cerebetli
POSTERIOR
Figure 1. Four views of a stylized cetacean brain with dura mater intact. (A) Rostrolateral; (B) rostro-
lateral. showing two dural processes; (C) dorsal; (D) ventral. The relative distribution of magnetic
material in the dura mater of adult cetaceans is indicated in solid black. The one mysticete studied,
a juvenile humpback whale, lacked a differential concentration in the posterior areas, as did the
juvenile DaU's porpoises.
been misdirected. Despite this problem, some interesting and robust trends seem to be
present across three odontocete species, Ziphius, Tursiops, and Phocoenoides. Megaptera
showed some similarities to the odontocetes but also showed some important differences.
Among the odontocetes, the largest moments within each species, NRM and/or sIRM,
seemed to be located in the dura frontally near the falx cerebri and posteriorly around the
cerebellum (Fig. 1). In Tursiops the cerebral dura was more magnetic ventrally; in Pho-
coenoides and Ziphius the ventral dura covering the midbrain had a significant moment.
In all species the dorsal and lateral cerebral dura were found to be considerably less mag-
netic than other dural tissues. In the two species tested, Tursiops and Phocoenoides, the
502 Chapter 24
cerebrum was found to have a very weak remanence. On the other hand, the cerebellum
had a strong remanence. The anterior falx cerebri tissues consistently had a greater coer-
civity than other tissues.
Megaptera also had its largest moment in the anterior falx cerebri area, no moment in
olfactory tissue adjacent to the anterior falx, but a large posterior remanence was not found.
It is not known whether the distribution reflects species or age differences, or considering
our small sample size, normal variation. As was noted earlier, a strong cerebellar remanence
in the dura was not present in juvenile Phocoenoides. Studies of various species-ho-
neybees (Gould et a1., 1978), tuna (Walker et a1., Chapter 20, this volume), turtles (Perry
et a1., this volume), and woodmice (J. Mather, personal communication)-have found that
juveniles have less magnetic material than adults. In Phocoenoides this proved to be the
case, although only in the cerebellar area of the dura. The relative magnetic softness of
anterior tissues in Megaptera indicated by AF demagnetization proved to be a function of
interaction effects and not representative of the real coercivity.
The nature of the magnetic material in cetaceans is still incompletely determined.
Analysis of Megaptera indicates that single-domain magnetite is present. The saturated
remanence and coercivity in odontocete tissues are comparable to other vertebrate species
(see Perry et a1.; Presti; Walker et a1., Chapter 20, this volume). The coercivity of the
magnetic tissues, although suggestive of at least some single-domain particles, showed less
resistance to AF demagnetization than single-domain grains found in geological forma-
tions, i.e., median unblocking fields of about 400 Oe (Dunlop and West, 1969; Levi and
Merrill, 1978). But, as was shown to be the case with Megaptera, this was probably due
to interaction effects (Kirschvink and Lowenstam, 1979; Cisowski, 1981). Tissues from the
Ziphius and the Megaptera were dissolved in 5% Millipore-filtered hypochlorite solution
(commercial bleach). After centrifuging, a residue of apparently black, magnetic particles
settled out. We say "apparently" because given the small quantity of aggregated material,
there was some color ambiguity. Walcott et a1. (1979) observe that magnetite is the only
magnetic material that is black. It is also probable that the magnetic contribution of iron
from hemoglobin is trivial. When the retia (large blood vessel complexes) are detached
from the dura, the magnetic moment remains with the dura. Although most areas of sig-
nificant remanence are heavily vascularized, there are also areas of little vascularity which
have a strong moment.
The similarities among different odontocetes of the loci, magnitude, and coercivity of
magnetic material in conjunction with the apparent color of the magnetic aggregate found
in Ziphius tissues, suggest that at least some of the magnetic material is magnetite. The
case for magnetite in Megaptera is considerably stronger. Of special interest is the location
of the magnetic material found in the area of the anterior falx cerebri in all the cetaceans
studied. This area sits against the ethmoid bones of the cranium. Walker et a1. (Chapter
20, this volume) have located the primary site of magnetite in yellowfin tuna in the ethmoid
complex. It is much too early in our investigations to suggest a homology, but the asso-
ciation of magnetic material with the ethmoid region is worth investigating in other species.
Also worth comparative study is the distribution of remanence in the actual brain tissues.
The report of magnetic material in the hind- and midbrain tissues of the rhesus monkey
(Kirschvink, 1981) and a corresponding nonmagnetic cerebral cortex is similar to our find-
ings for Tursiops and Phocoenoides.
The data on Delphinus need to be put into the context of the other studies. Zoeger et
a1. (1981) reported a low NRM in most of the tissues that they looked at, but chose to
concentrate on a "hot spot" of multi domain magnetite. The former probably corresponds
to the material we found in other species. We did not find any similar large particles of
soft magnetic material except as contamination in the other cetaceans. At this time, we
would have to consider the soft magnetic material specific to the sample studied, specific
to the species, or alternatively as contamination.
Cetaceans 503
Additional analyses have to be done with our present samples in terms of more pre-
cisely characterizing the magnetic material, e.g., what elements are present, what the shapes
of the particles are, and how they are distributed with respect to each other. Our studies
have concentrated on the cranium, especially the dura mater. This was based on previous
work on other species. It is important to sample from a whole animal. Thus far, availability
and size of subjects have proven to be obstacles. Findings of a noticeable remanence in
the cranial blubber and muscle of some specimens indicate that magnetic material may
be more pervasive than we had previously thought. However, using our present technique
of cutting with plastic and glass knives to sample an entire dolphin or whale would be
prohibitively cumbersome.
There is a potential alternative. Third-order SQUID gradiometers are currently avail-
able which offer some potential for measuring the magnetism of tissues in a nonintrusive
manner (Vrba et aI., 1982). A biogradiometer such as the one described by Vrba et al. offers
a number of advantages. (1) It overcomes the problem of introducing contaminants into
tissues. (2) It facilitates sampling over large areas by reducing time-consuming dissections.
(3) It offers a unique opportunity to study dynamic magnetic processes in live animals,
e.g., a third-order magnetometer has been used to produce a magnetoencephalogram in
humans (Vrba et al., 1982).
Our previous studies have concentrated on the rough localization of magnetic reman-
ence; future studies should concentrate on finer detail, especially more precise localization
and innervation. Because of the various difficulties of working with cetaceans, not the least
of which is expense, and in light of the parallels that seem to exist with other species, it
might be prudent to work with some smaller, more accessible animals. After some neural
models are developed with other species, cetaceans might be more efficiently studied.
4. Conclusion
Cetaceans present some unique problems for magnetic research. Notwithstanding dif-
ficulties with contamination and a changing methodology, we discovered some surprising
consistencies in the location, magnitude, and coercivity of magnetic tissues. From the
standpoint of looking at magnetic navigational abilities, the ultimate tests are behavioral.
Anatomical and physiological correlates may be suggestive, but they do not prove the
existence of a perceptual ability. The difficulties encountered in demonstrating magnetic
detection in cetaceans and other species, e.g., homing pigeons (Keeton, 1974), suggest that
a multiphase approach is necessary. Ideally, field homing experiments should be imple-
mented similar to those done with homing pigeons, in which Helmholtz coils were attached
to the heads of subjects (Walcott and Green, 1974). Unfortunately, the expense of doing
open-water work with cetaceans is a severe limitation. The correlation of pelagic cetacean
strandings with areas of minimum magnetic field is highly promising. Supportive evidence
from other geographic areas in conjunction with more quantitative analysis should elu-
cidate this relationship. Both Klinowska and our group have found that a number of prob-
lems exist for elucidating geomagnetic-stranding relationships: (1) the multiple causal
nature of strandings; (2) errors of reporting and record keeping; (3) biased or missing data
due to unobserved strandings; and (4) different effort at different locations and times. All
these tend to introduce a lot of noise into the data. In the future we expect to analyze larger
data bases and concentrate on species with extensive records of strandings, e.g., pilot
whales, Globicephala. The larger data base in addition to the elimination of interspecies
variance should provide somewhat cleaner data.
Experimental procedures are necessary to establish causal connections as well as to
provide precise analyses of magnetic detection capabilities. If magnetic detection is oc-
504 Chapter 24
curring, the pertinent field characteristics need to be determined. For example, are ceta-
ceans detecting differences in intensity, declination, or inclination? Also, what are the
relevant organism variables, e.g., species, movement, duration of exposure to magnetic
fields? Of course, we have yet to demonstrate a magnetic detection capability in the lab-
oratory. However, a variety of changes in experimental procedure were proposed earlier
in this chapter which should enhance our ability to demonstrate magnetic detection abil-
ities, if they exist. Despite the obvious necessity of continued behavioral studies, there are
still important reasons for continuing to emphasize anatomical studies as well. (1) The
conditioning work with dolphins has not yielded positive results. The difficulties en-
countered in conditioning other species believed to have a magnetic sense, e.g., pigeons
(Bookman, 1978), suggests that we should not count on a rapid breakthrough in this area.
(2) Other, nonconditioning, laboratory experiments with various species have been de-
pendent, at least in part, on rather stereotyped behaviors; for example, the migrating "in-
stinct" in birds has been used to show changes in orientation in flight cages in response
to changes in magnetic fields (Wiltschko, 1978). Such inflexible patterns of behavior have
not been identified in small cetaceans appropriate for laboratory study. (3) Correlational
studies of natural behavior and magnetic events will probably continue to suffer from a
substantial noise problem. (4) Open-water work with dolphins is extremely expensive, and
consequently difficult to get funded. (5) Any type of experimental behavioral work with
large whales and magnetic field detection is unlikely. (6) As experimentation with more
accessible animals, e.g., tuna, pigeons, and possibly rodents, yields data about anatomical
and physiological correlates of magnetic perception, we can look for parallel structures
and neural processes in cetaceans. An argument by analogy with other species could form
the basis for a hypothetical model of cetacean magnetic detection.
Another reason for continuing the anatomical studies is that they may lead to some
insight on alternative functions of biomagnetic material. Based on prior research with
noncetacean species (see Kirschvink and Gould, 1981) in addition to our knowledge of
the ecology of cetaceans, we originally hypothesized that magnetite might be implicated
in the detection of magnetic fields for purposes of orientation and navigation. There are
many other possible reasons for the presence of magnetite. The findings of magnetic ma-
terial in a broad range of tissues such as muscle, blubber, and bone in addition to dura
and neural tissue is perplexing. One possibility is that it is a natural, environmental con-
taminant (as opposed to a contaminant introduced during dissection and analysis). Gaskin
(1982) reports finding a wide variety of heavy metals in cetacean tissues. Some of these
are, of course, involved in metabolic processes, e.g., iron in hemoglobin. Other metals,
e.g., mercury, are exogenously derived toxins. One possibility to consider is that at least
some magnetic material may exist pathologically in cetacean tissues.
Magnetite may be involved in structural support. It increases the hardness of chiton
teeth (Kirschvink and Lowenstam, 1979). The dura of adult cetaceans is substantially os-
sified in various areas. If this reflects selective pressure to provide extra hardness or rigidity,
magnetite could contribute to this process. Conley (1969) lists a wide variety of weak
magnetic effects on biological systems, including accelerated growth in paramecia, changes
in motor activity in sparrows, and reduction in enzyme activity, infertility, shortened life
span, and hyperphagia in mice. Brown (1964; Brown et a1., 1970) discusses the influence
of magnetic fields on circadian rhythms. It is quite apparent that the potential functions
of magnetic material in biological systems are considerably broader than the scope of our
present research. A comprehensive understanding of the function of magnetic materials
in cetaceans calls for a greatly enhanced research effort.
ACKNOWLEDGMENTS. This research was supported in part by grants to Louis M. Herman from
the National Science Foundation (BNS-8109653) and Sea Grant (NA79AA-D-00085). Spec-
imens and facilities for dissections were provided by Linda Jones, National Marine Mam-
Cetaceans 505
mal Laboratory, Seattle, Washington, and Bill Gilmartin, National Marine Fisheries Service,
Honolulu, Hawaii. Sam Ridgway and Pete Schroeder of the Naval Ocean Systems Center
assisted in the dissection of the humpback whale. Facilities for behavioral research were
provided by the Kewalo Basin Marine Mammal Laboratory, University of Hawaii. Andrew
E. Dizon and Joseph L. Kirschvink gave freely of their time and expertise to assist in many
phases of the research. Louis M. Herman offered helpful suggestions on the laboratory
experiments. Behavioral experiments were conducted with the invaluable assistance of
students from the University of Hawaii: Terry Annon, Sayward Ayre, Carolyn Borden,
Paula Borden, Debbie Ching, Bob Cisneros, Elly Forrest, Barbara Hay, Jean Osumi, Kathy
Sobon, Amy Suzuki, and Jason Yoshida. Brooks G. Bays did the illustration. Jayn Roush,
Barbara Kuljis, and C. Scott Baker assisted with the literature review and manuscript prep-
aration. Michael Walker reviewed the manuscript. We wish to thank all of these people
for their generous support of this project.
References
Baker, C. S., and Herman, L. M., 1981, Migration and local movement of humpback whales (Megaptera
novaeangliae) through Hawaiian waters, Can J. Zoo1. 59:460-469.
Baker, C. S., Herman, L. M., Bays, B. G., and Stifel, W. F., 1982, The impact of vessel traffic on the
behavior of humpback whales in southeast Alaska, Unpublished report to the National Marine
Fisheries Services.
Baker, C. S., Herman, L. M., Perry, A., Lawton, W. S., Straley, J., and Straley, J., 1985, Population
characteristics of summer and late-season humpback whales (Megaptera novaeangliae) in south-
eastern Alaska, Marine Mammal Science, Vol. 4, No. 1.
Baker, R R, 1978, The Evolutionary Ecology of Animal Migration, Holmes & Meier, New York.
Bookman, M. A., 1978, Sensitivity of the homing pigeon to an earth-strength magnetic field, in: Animal
Brown, F. A., Jr., 1964, A unified theory for biological rhythms, in: Circadian Clocks U. Aschoff, ed.),
North-Holland, Amsterdam, pp. 231-261.
Brown, F. A., Hastings, J. W., and Palmer, J. D., 1970, The Biological Clock, Academic Press, New
York.
Busnel, R-G., and Dziedzic, A., 1966, Acoustic signals of the pilot whale Globicephala melaena and
the porpoises Delphinus delphis and Phocoena phocoena, in: Whales, Dolphins, and Porpoises
(K. S. Norris, ed.), University of California Press, Berkeley, pp. 607-646.
samples, Phys. Earth Planet Inter. 26:56-62.
Conley, C. c., 1969, Effects of near-zero magnetic fields on biological systems, in: Biological Effects
of Magnetic Fields (M. F. Barnothy, ed.), Plenum Press, New York, pp. 29-51.
Darling, J. D., and Jurasz, C. M., 1983, Migratory destinations of North Pacific humpback whales
(Megaptera novaeangliae), in: Communication and Behavior of Whales (R. Payne, ed.), Westview
Press, Boulder, pp. 359-368.
Darling, J. D., and McSweeny, D. J., 1983, Observations on the migrations of North Pacific humpback
whales (Megaptera novaeangliae), in: Proceedings of the 5th Biennial Conference on the Biology
of Marine Mammals Abstracts, p. 21.
Dawbin, W. H., 1956, The migrations of humpback whales which pass the New Zealand coast, Trans.
H. Soc. N.Z. 84:147-196.
Dawbin, W. H., 1966, The seasonal migratory cycle of humpback whales, in: Whales, Dolphins, and
Porpoises (K. S. Norris, ed.), University of California Press, Berkeley, pp. 145-170.
de Lorge, J., and Marr, M. J., 1974, Operant methods assessing the effects of elf electromagnetic fields,
in: ELF and VLF Electromagnetic Field Effects (M. A. Persinger, ed.), Plenum Press, New York,
pp. 145-175.
Geophys. 7:709-757.
506 Chapter 24
Evans, W.E., 1971, Orientation behavior of delphinids: Radio telemetric studies, Ann. N.Y. Acad. Sci.
188:142-160.
Science 203:1355-1356.
Gaskin, D. E., 1982, The Ecology of Whales and Dolphins, Heinemann, London.
Geraci, J. B., and st. Aubin, D. J., 1979, Biology of Marine Mammals: Insights Through Strandings,
Report No. MMC-77113, U.S. Marine Mammal Commission.
Gilmore, R M., 1960, Census and migration of the California gray whale, Nor. Hyalfangst.- Tid. 49:409-
431.
Goree, W. S., and Fuller, M., 1976, Magnetometers using RF-driven Squids and their applications in
Gould, J. 1., Kirschvink, J. 1., and Deffeyes, K. S., 1978, Bees have magnetic remanence, Science
201:1026-1028.
Herman, L. M., 1980, Cognitive characteristics of dolphins, in: Cetacean Behavior (1. M. Herman, ed.),
Wiley, New York, pp. 319-429.
Hinde, R A., and Stevenson-Hinde, J., 1973, Constraints on Learning, Academic Press, New York.
5:47-132.
Kinne, 0., 1975, Orientation in space: Animals, mammals, in: Marine Ecology, Volume II, Part 2 (0.
Kinne, ed.), Wiley, New York, pp. 709-916.
Kirschvink, J. 1., 1981, Biogenic magnetite (Fe 3 04): A ferrimagnetic mineral in bacteria, animals, and
man, in: Ferrites, Proceedings of the International Conference, Japan, 1980, pp. 135-137.
Kirschvink, J. 1., and Gould, J. 1., 1981, Biogenic magnetite as a basis for magnetic-field detection in
Lett. 44:193-204.
Kirschvink, J. 1., Westphal, J. A., and Dizon, A. E., 1984, Cetacean strandings along the North American
Atlantic coast: Evidence for a geomagnetic influence on pelagic navigation, EOS, Transactions of
the American Geophysical Union, San Francisco, Abstract, 65:865.
Klinowska, M., in press, Strandings and behaviour, in: Research on Dolphins (M. M. Bryden and R.
J. Harrison, eds.), Oxford University Press, London, in press.
Kobayashi, K., and Fuller, M., 1968, Stable remanence and memory of multi domain materials with
special reference to magnetite, Philos. Mag. 18:601-624.
Levi, S., and Merrill, R T., 1978, Properties of single-domain, pseudo-single domain, and multi-domain
magnetite, J. Geophys. Res. 83:309-323.
Lowrie, W., and Fuller, M., 1971, On the alternating field demagnetization characteristics of multi-
domain thermoremanent magnetization in magnetite, J. Geophys. Res. 76:6339-6349.
Madsen, C. J., and Herman, 1. M., 1980, Social and ecological correlates of cetacean vision and visual
appearance, in: Cetacean Behavior (1. M. Herman, ed.), Wiley, New York, pp. 101-147.
Mayaud, P. N., 1980, Derivation meaning and use of geomagnetic indices, Geophysical Monograph
22, American Geophysical Union, Washington, D. C.
Millen, J. W., and Wollam, D. H. M., 1962, The Anatomy of Cerebrospinal Fluid, Oxford University
Press, London.
Morgane, P. J., and Jacobs, M. S., 1972, Comparative anatomy of the cetacean nervous system, in:
Functional Anatomy of Marine Mammals (R J. Harrison, ed.), Academic Press, New York, pp.
117-244.
Nishiwaki, M., 1960, Ryukyuan humpback whaling in 1960, Sci. Rep. Whales Res. lnst. 15:1-15.
Ossenkopp, K.-P., and Barbeito, R, 1978, Bird orientation and the geomagnetic field: A review, Neu-
Phillips, J. B., 1977, Use of the earth's magnetic field by orienting cave salamanders (Eurycea lucifuga),
J. Compo Physiol. 121:273-288.
Cetaceans 507
Pike, G. c., 1962, Migration and feeding of the gray whale (Eschrichtius gibbosus), J. Fish. Res. Board.
Can. 19:815-838.
Pilleri, G., and Knuckey, J., 1968, The distribution, navigation, and orientation by the sun of Delphinus
delphis L. in the western Mediterranean, Experientia 24:394-396.
d'un conditionement nociceptif, J. Physiol. (Paris) 60:85-92.
Seligman, M. E. P., and Hager, J. L., 1972, Biological Boundaries of Learning, Appleton-Century-
Crofts, New York.
Sheppard, A. R., and Eisenbud, M., 1977, Biological Effects of Electric and Magnetic Fields of Ex-
tremely Low Frequency, New York University Press, New York.
Shettleworth, S. J., 1972, Constraints on learning, Study Behav. 4:1-68.
Teitelbaum, P., 1966, The use of operant methods in the assessment and control of motivational state,
in: Operant Behavior: Areas of Research and Application (W. K. Honig, ed.), Appleton-Century-
Crofts, New York, pp. 565-608.
Vrba, J., Fife, A. A., Burbank, M. B., Weinberg, H., and Brickett, P. A., 1982, Spatial discrimination
in Squid gradiometers and 3rd order gradiometer performance, Can. J. Phys. 60:1060-1073.
Walcott, c., and Green, R. P., 1974, Orientation of homing pigeons altered by a change in the direction
Walcott, c., Gould, J. L., and Kirschvink, J. L., 1979, Pigeons have magnets, Science 205:1027-1028.
Whitehead, H., Silver, R., and Harcourt, P., 1982, The migration of humpback whales along the north-
east coast of Newfoundland, Can. J. 2001. 60:2173-2179.
pp. 302-210.
Yorke, E. D., 1981, Sensitivity of pigeons to small magnetic-field variations, J. Theor. Bioi. 89:533-
537.
2oeger, J., Dunn, J. R., and Fuller, M., 1981, Magnetic material in the head of the common Pacific
Chapter 25
Magnetoreception and the Search for
Magnetic Material in Rodents
JANICE G. MATHER
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
2. Influence of Magnetic Fields on Physiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510
3. Magnetoreception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510
3.1. Detection of Magnetic Fields. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510
3.2. Use of the Geomagnetic Field for Compass Information . . . . . . . . . . . . . . . . . . 513
4. The Search for the Magnetoreceptor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
4.1. Magnetometric Investigations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
4.2. Histological Investigations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 524
4.3. Neurophysiological Investigations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 530
5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 531
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 532
1. Introduction
The influence of magnetic fields on small mammals, particularly rodents, has been in-
vestigated for the past 30 years. Early studies were aimed at determining the effect of
anomalous magnetic conditions on mammalian biology in order to assess possible hazards
associated with human occupations. As a consequence, many experiments have involved
exposing laboratory rodents to magnetic fields that are either far stronger or weaker than
the earth's magnetic field, so that any influence on physiology and behavior can be as-
certained.
More recently, research on rodents has branched into the field of magnetoreception,
that is, the ability to detect magnetic fields per se as opposed to merely being affected by
them. In particular, the latest research has been concerned with the ability of wild and
laboratory rodents to acquire compass information from the geomagnetic field, and so has
involved exposing animals to more natural magnetic environments during experimenta-
tion. These magnetic orientation investigations have stimulated a new line of research
concerned with the sensory mechanisms which enable rodents to detect ambient magnetic
fields. *
* Owing to the vast literature regarding the influence of magnetic fields on organisms, reference to
specific examples of animals other than rodents is restricted to the section covering the search for
the magnetoreceptor (Section 4).
JANICE G. MATHER • Department of Zoology, University of Manchester, Manchester M13 9PL,

United Kingdom. Current address: Zoological Laboratory, Institute of Zoology and Zoophysiology,
University of Aarhus, Denmark.
509
510 Chapter 25
2. Influence of Magnetic Fields on Physiology

Numerous investigations over the past few decades report an influence of magnetic
fields on rodent physiology and behavior (for review see Barnothy, 1964, 1969; Ketchen
et 01.,1978). A wide variety of magnetic field conditions have been investigated, with field
strengths ranging from less than 0.5 mG (usually produced by Helmholtz coils and magnetic
shielding) to 9000 G (usually produced by an electromagnet or a permanent magnet), and
the duration of exposure ranging from minutes to years. Studies have involved several
species of rodents, mainly the laboratory mouse, Mus musculus, the laboratory rat, Rattus
norvegicus, and the golden hamster, Mesocricetus auratus.
For rodents, the physiological and associated behavioral effects resulting from ex-
posure to strong magnetic fields (400-9000 G) are manyfold. Briefly, reports include in-
fluences on the CNS, bone marrow, adrenal cortex, spleen, liver, kidney, blood, and urine,
as well as adverse effects on conception rate, embryo development, and ovary morphology.
Other influences include increases in longevity, vigor, and appetite, and decreases in aging
rate, weight, wound healing, and aggression. High field strengths also appear to increase
the rejection rate of cancer transplants in rodents. The physiological effects of near-zero-
strength magnetic fields (-0.5-1.0 mG) include tissue hyperphasia and a reduction in
enzyme activity, fertility, and lifespan. Weak magnetic fields also seem to induce behav-
ioral changes such as supine positioning and cannibalism.
In general, the various investigations of magnetic fields of very high and very low
strength suggest that the biological effect of the field is not instantaneous, but rather that
a continuous exposure over some length of time is required for the effect to be observed.
Possible mechanisms by which magnetic fields could exert an influence on general phys-
iology, for example, by affecting the shape, diffusion rate, or oxidation rate of certain
molecules, are outlined in Barnothy (1969).
3. Magnetoreception
Considerable evidence has accumulated which indicates that in addition to the many
effects of magnetic fields on rodent physiology, magnetic fields can actually be detected
by rodents. Such an ability for magnetoreception implies an influence, either direct or
indirect, of the field on the animal's central nervous system. Evidence for magnetic field
detection by rodents is available from a variety of investigations which tend to fall into
two groups: those reporting a general behavioral response to the magnetic field and those
reporting a more specific response related to accessing compass information from the field.
3.1. Detection of Magnetic Fields
Within the group of investigations dealing with magnetic field detection without nec-
essarily involving directional information, some findings remain ambiguous with regard
to the means by which the magnetic field influences the behavior of the animal. For ex-
ample, the tilt-cage activity of the Mongolian gerbil, Meriones unguiculatus, appears to
correlate with changes in the intensity of the horizontal vector of the geomagnetic field
(Stutz, 1971, 1972). Although this correlation could be due to the animal's detection of
the magnetic field, the animal modifying its level of activity accordingly, it could alter-
natively result from an influence of the ambient field on some aspect of the animal's
physiology, e.g., by affecting muscle performance. Other investigations, such as those in
which the animal is allowed to choose between a high- and a low-intensity magnetic field
Rodents 511
Figure 1. Orientation cage used for dis-

placement experiments on rodents. The ap-
paratus comprised a cross-shaped Perspex úê~åëé~êä>åí =
chamber which had transparent sides and pl!rspl!x door
lid, and a black opaque floor. An activity (OOl! per armJ
wheel (diameter 13 cm) was present in

each of the four arms of the cage, which
allowed the animal to wheel-run and also
provided some degree of shelter. Trans-
parent Perspex doors were inserted at the
ends of the four arms for experiments on
magnetic orientation whereas aluminum
mesh doors were used for other orientation
experiments. The arms of the cage were
aligned with geomagnetic north, south,
east, and west during testing. The animal
entered the cage through a centrally posi-
tioned hole in the lid. In overnight exper-
iments, a circular nest box containing bed- ú =
ding material and food was present in the
center of the cage, a water bottle was positioned directly above the nest, and sawdust was scattered
over the floor. All provisions were renewed for each experiment on a different individual. Redrawn
from Mather and Baker (1980).
environment, offer more support for magnetoreceptive ability. For example, in a study of
the effect of a permanent magnetic field of 1100 G on the activity and food/water con-
sumption of rodents, l-year-old Swiss female mice were retained for periods of 22 hr in
a lightproof T-shaped apparatus which had a permanent magnet at one end and a dummy
magnet at the other. The mice, which were permitted to move freely within the apparatus,
showed a tendency for increased activity and increased food and water intake in the high-
field environment, but tended to spend more time in the low-field environment (Russell
and Hedrick, 1969).
I have recently found a similar preference for low-intensity magnetic conditions in a
pilot investigation using the European woodmouse, Apodemus sylvaticus (Mather, un-
published). This investigation was conducted in the laboratory, the experimental animals
being from a laboratory colony. A relatively simple experimental technique was employed.
The apparatus consisted of a cross-shaped Perspex chamber (Fig. 1) which had an activity
wheel placed in each of the four arms. Attached to each wheel were two mechanical
counters, one to register the number of revolutions in a clockwise direction, and the other
to register those in a counterclockwise direction. Exterior to the chamber, a brass bar was
positioned below each of three of the wheels, and a bar magnet (pole strength 130 G) was
positioned below the remaining wheel. The long axes of the bars and arms coincided and
the north pole of the magnet always pointed away from the center of the chamber. All bars
were wrapped in polythene to remove possible differences in odor. Consequently, an an-
imal running in a wheel above a brass bar experienced the normal laboratory magnetic
field of 0.5 G, whereas an animal wheel-running above a bar magnet experienced a rela-
tively strong magnetic field of approximately 14.0 G (i.e., field strength 3 cm up from the
center of the magnet). The entire apparatus was housed within an opaque plastic screen
plus lid, so restricting visual and olfactory cues. During displacement to the experimental
room, attempts were made to disorient the animal in order to reduce the possibility that
any influence of the bar magnet may be overridden by a "homing" response. Each mouse
was transported in an aluminum container covered by black polythene so reducing di-
512 Chapter 25
(a) (b) direc tion of

wheel with arm 'with magnet Figure 2. Influence of a strong mag-
magnet
t netic field on the wheel activity of
woodmice. Each mouse was housed
a=+ 1590
in a cross-shaped chamber which of-
a=+161°
r=0.472 r=0.345 fered a choice of four activity
wheel wheels, one per chamber arm. A
\ \
-90 brass bar was positioned below each
of three of the wheels and a bar mag-
net was positioned below the fourth
•• • • ••
wheel. A total of nine woodmice,
• • four males and five females, were
wheel opposite each tested on one occasion, males
magnet Z=D.956
u= 1.314 and females being used alternately.
Z=1.787 One female did not wheel-run.
u=1.771· (a) Influence of the bar magnet
on the choice of activity wheel. Each
dot represents an individual's mean angle of orientation of the wheel-running vector, derived from
the total number of revolutions in each of the four arms. Angles are expressed relative to the arm with
the magnet.
(b) Influence of the bar magnet on the preferred orientation of wheel activity. Each dot represents
an individual's mean angle of orientation of wheel activity, derived from the total number of revo-
lutions in the four cardinal compass directions regardless of arm. Angles are expressed relative to the
direction of the arm with the magnet.
Arrows with circles show the second-order mean vectors for the group, this vector having a mean
angle, a, and a particular bias, r. Rayleigh's Z-test examines for nonuniformity, and the V-test, statistic
U, examines the significance of the component in a predicted direction. V-test relative to 180 from 0
the magnet. All circular statistics according to Batschelet (1981). *P < 0.05; **P < 0.01; ***P < 0.001.
rectional information during travel, and was taken along a fairly complex route (involving
a distance of 50 m and four flights of stairs). Also, at the beginning and end of the dis-
placement journey, the container plus animal was rotated through 360 on three occasions.
0
Once within the test site screen, the mouse was allowed to enter the cross-shaped chamber
where it remained for a period of 15 hr. The total number of wheel revolutions in both
clockwise and counterclockwise directions for each of the four wheels was read from the
counters the following morning. The chamber was maintained in the same physical ori-
entation for the duration of the investigation but the magnet was moved clockwise to the
next arm before each experiment on a different individual. (This procedure controlled for
such factors as: orientation to the home cage or 'some feature in the room; a possible pre-
ferred compass direction; any influence of smell of previous occupant; and slight differ-
ences in the wheels or chamber arms.)
For each mouse, the total number of revolutions per wheel was determined and the
wheels ranked according to amount of activity. This revealed that those wheels above brass
bars are preferred to the wheel above the bar magnet. No animal carried out its highest
level of activity in the wheel over the magnet, and moreover, six of the eight animals ran
least in the "magnetic" wheel [for the latter, p < 0.01, binomial test (Siegel, 1956)]. This
tendency for the mice to dislike the wheel associated with the strong magnetic field is
again evident when the total amount of wheel activity per arm is used to derive a mean
vector of running (Fig. 2a). Furthermore, if the mean vector is derived from the direction
of wheel revolution regardless of arm, the mice appear to adopt a heading away from the
source of the magnetic anomaly although this trend does not reach significance (Fig. 2b).
This tendency for avoidance of a high-intensity magnetic field in a four-choice situ-
ation seems to have been confirmed. Laboratory rats retained in a radical four-arm maze
Rodents 513
for a 3-min observation period are reported to orient away from an arm containing a magnet,
though data have not yet been published (Chafetz, 1982).
That the low level of activity associated with strong magnetic conditions in the wood-
mouse investigation is in contrast to the high level of activity found for strong magnetic
conditions in the laboratory mouse investigation by Russell and Hedrick (1969) almost
certainly results from the use of different activity recording techniques, the activity wheel
and the photocell, respectively. These techniques often yield data with dissimilar trends
in response to an assortment of factors, simply because they monitor different behavior
patterns. Whereas the photocell records the animal's general movement, the wheel records
only running activity which has recently been shown to reflect "migratory restlessness,"
similar to Zugunruhe exhibited by caged birds (for discussion see Mather, 1981a,b).
It is worth noting that migratory behavior often involves the use of compass infor-
mation, particularly when the animal is in unfamiliar terrain. Any animal that makes use
of a geomagnetic compass may be expected to avoid magnetic anomalies as these are likely
to interfere with magnetoreception ability. This may explain, in part, the avoidance re-
sponse to strong magnetic fields shown by rodents.
3.2. Use of the Geomagnetic Field for Compass Information
An ability to determine direction would be of particular advantage to any animal that

at some stage of its life has to travel either in a specific direction, i.e., orientation, or to a
specific location, i.e., navigation. Many species of animals are now known to have such
an ability, directional information being derived from a variety of sources including the
earth's magnetic field
A sense of direction in rodents has been demonstrated for a number of species in both
the field and the laboratory. Field investigations generally make use of the fact that small
mammals tend to spend most of their time in a comparatively limited area, the home range,
to which they usually return after exploratory forays. For example, "homing" experiments
in which animals are caught, displaced various directions and distances, released, and
finally recaptured at the original capture site, have shown that rodents are able to return
to a specific location after displacement from that area (see Bovet, 1978). That this homing
ability is not simply the result of random scatter away from the point of release, but involves
some kind of navigational mechanism, is shown by displacement experiments in which
rodents are able to orient toward home while retained in some form of orientation cage at
the release site (Lindenlaub, 1955, 1960; Bovet, 1960; Mather and Baker, 1980, 1981). This
apparatus facilitates the monitoring of the animal's directional preferences which are man-
ifest as either more frequent movements in, or more time spent in, some directions than
others.
A clear demonstration of navigational ability involves testing the animal's orientation
response in four different compass quadrants relative to the goal. An investigation em-
ploying this procedure has been conducted in the field by myself and R. Robin Baker
(University of Manchester) using three species of wild rodents, the European woodmouse,
A. sylvaticus, the yellow-necked mouse, A. flavicollis, and the bankvole, Clethrionomys
glareolus (Mather and Baker, 1980). After capture in a Longworth trap, each animal was
displaced with visual and olfactory cues reduced and then tested in a cross-shaped ori-
entation cage made from transparent Perspex (Fig. 1). Each test comprised a 4-min obser-
vation period during which the time spent by the animal in each of the four arms of the
cage was monitored. Testing on the same individual was usually repeated at four different
displacement sites, each varying in compass direction and straight-line distance (up to 200
m) from the point of capture. In displacement experiments on wild rodents, the trap of
514 Chapter 25
East South
• •
Upslope
.----
•• a=+23
r=O.6S9
Figure 3. A demonstration of
navigational ability by rodents.
Wild rodents (19 in all), mainly
woodmice, were displaced and
, then tested in an orientation
- •• l.
I
• •• cage. Data are lumped for dis-

Z=S.934** Z=3.47S*
placements in four different
u=3.384 *** u=2.427** compass sectors relative to the
trap of capture. Each dot repre-
sents an individual's mean angle
•••••• of orientation, derived from the
l •• , time spent in each of the four
,, •• arms of the cage during a 4-min
I a=+14 ,, •• \ / a=+43
test. Angles are expressed rela-
r=O.S12 " r=O.656 tive to trap direction. Arrows
j
show the mean vectors for the
• four sectors and the dashed lines
denote the 95% confidence in-
• •• / terval around each sample mean.
North West Statistics as in Fig. 2. V-test rel-
Z=4.192* Z=6.S87*** ative to trap direction. Redrawn
u=2.S10** u=2.714** from Mather and Baker (1980).
capture is usually presumed to be in the vicinity of the animal's home and is therefore
regarded as the goal during orientation.
Analysis of the animals' performance in the orientation cage indicated an ability for
homeward orientation for displacement distances between 6 and 80 m. Figure 3 shows
that the directional preferences for all four compass sectors relative to the trap are non-
uniform (Z-test) with a nonrandom component in the direction of the trap (V-test). Fur-
thermore, for three of the four sectors the sample mean is actually in the trap direction
(Le., the 95% confidence interval around the mean angle includes trap direction), uphill
orientation tending to prevail from the west sector. Such results provide a convincing
demonstration of navigation by rodents.
Having shown that rodents do have navigational ability, the next step is to examine
the mechanisms involved in this goal orientation response. Basically, directional infor-
mation can be gained from cues available en route and from cues available on site.
Investigations of the goal orientation of woodmice suggest that visual and perhaps
olfactory cues can provide valuable directional information (Mather and Baker, 1980,
1981). Prominent landmarks appear to be of value for orientation over longer distances,
as sugges,ted by observations of the Scandinavian lemming, Lemmus lemmus, (Marsden,
1964), and inertial mechanisms may playa role over shorter distances, as proposed for
the Mongolian gerbil, Meriones unguiculatus (Mittelstaedt and Mittelstaedt, 1980).
It is also known that rodents can make use of compass systems for directional cues.
The involvement of a sun compass in homeward orientation has been shown in the field
for the meadow vole, Microtus pennsylvanicus (Fluharty et aJ., 1976) and the thirteen-
lined ground squirrel, Spermophilus tridecemlineatus (Haigh, 1979), and sun compass
orientation has been demonstrated in the laboratory for the striped field mouse, Apodemus
agrarius (Lliters and Birukow, 1963).
The use of the geomagnetic field as an additional source of compass information for
small mammals such as rodents would seem to be advantageous for several reasons. Many
Rodents 515
species of rodents inhabit subterranean burrows and have runs beneath leaf litter and
vegetation. Moreover, many species are primarily nocturnal, exhibiting peak levels of ex-
ploratory, food foraging, and reproductive behavior during the hours of darkness (Corbet
and Southern, 1977; Mather, 1981a,b). Under such conditions, the animals are afforded
little or no view of celestial and topographic cues which are potentially useful for ori-
entation. Furthermore, celestial cues in particular are not constant but change with time
of day, month, and year, and their availability may be limited by weather conditions such
as thick cloud cover. Considering these points, some other compass system based on non-
visual features of the environment would, therefore, be of advantage.
The most convincing evidence for magnetoreception in rodents comes from the dem-
onstration of an ability to derive directional information from the ambient magnetic field.
Over the past three decades, a range of data implicating the earth's magnetic field for
compass information has arisen from a variety of investigations.
Circumstantial support for a magnetic compass is available from an investigation of
woodmice and bankvoles which involved subjecting the animals to a training procedure
designed to induce them to adopt a particular orientation with respect to the cardinal points
of the compass (Bovet, 1965). The rodents were trained in a radial eight-arm chamber in
the absence of optical, acoustic, and olfactory cues. Even under these restricted conditions,
the animals were capable of recognizing particular compass directions, although Bovet
remained undecided as to the nature of the cues used for orientation. Similar findings are
reported for a study of the ability of the golden hamster to carry food back to the nest along
a fairly direct path (Etienne, 1980). Each hamster was housed and tested in a circular arena
which contained a central food supply and which had a nest box attached to the outside,
access to the nest being gained through one of 12 doors positioned symmetrically around
the arena's periphery. Even after passive transport to the food source and the simultaneous
elimination of visual, olfactory, and acoustic cues (including rotation of the entire arena),
most of the hamsters were still able to orient in the usual compass direction of the nest.
Furthermore, when tested in an unfamiliar arena, the hamsters again adopted a heading
consistent with orientation toward the usual location of the nest box. Etienne suggested
that the animals were orienting in a given direction with respect to a compass, the nature
of which still had to be determined, although the geomagnetic field was considered a
possible contender.
Several field investigations also appear to lend circumstantial support for a magnetic
compass. Displacement experiments on a number of wild rodent species, mainly the wood-
mouse, the yellow-necked mouse, and the bankvole, and involving the use of circular
labyrinth orientation cages, have demonstrated homeward orientation at distances of up
to 2-3.7 km despite the restriction of visual and olfactory cues (Lindenlaub, 1955, 1960;
Bovet, 1960). Similarly, during displacement experiments on woodmice and bankvoles
(Mather and Baker, 1980), animals tested before being allowed to see their surroundings
still showed an indication of homeward orientation (Mather, 1981a).
The first clear evidence for an ability by rodents to access compass information from
the ambient magnetic field is provided by a more recent field investigation using the Eu-
ropean woodmouse (Mather and Baker, 1981). For this investigation, an attempt was made
to alter in a predictable manner the directional cues available to the animal both during
the displacement journey and at the release site. Hence, two main items of apparatus were
employed: a transportation cage (Fig. 4) which allowed manipulation of magnetic, visual,
and olfactory information en route, and an orientation cage (Fig. 1) plus opaque screen
which allowed manipulation of visual and olfactory cues at the test site. The design of
screen ensured that only the sky directly overhead could be viewed by the mouse from
inside the orientation cage. Two experimental series were conducted, involving three dif-
ferent test treatments, as follows. Series I comprised two test treatments, control and ex-
perimental, both of which employed the screen at the test site. The control treatment
516 Chaptllr 25
cotton
wool scroll of translucent
plastic tubing
Figure 4. Transportation cage
used to manipulate directional
information during the displace-
ment of rodents. The cage com-
prised a plastic container which
could be made lightproof and air-
tight. The inside of the container
outer plastic container supported two lateral coils, each
consisting of 100 turns of 20
s.w.g. insulated copper wire,
which were connected in series
and powered by a 9-V battery. An
82-ohm resistor in the circuit re-
sulted in a current of 65 rnA,
\
thereby producing a horizontal
groove to receive
scroll magnetic field of 0.37 G between
Scm the coils. When this field was op-
NORTH
posed to the horizontal compo-
ú =
nent of the earth's magnetic field

(0.17 G), this resulted in a net horizontal component of 0.20 G opposite in direction to that of the
geomagnetic field. Thus, when the transportation cage was oriented along a north-south axis, the
horizontal component of the magnetic field within the cage could be shifted by 180°, so causing a
north-seeking compass needle to rotate and point toward geographic south. When the coils were
activated, therefore, a near-mirror image of the local geomagnetic field was produced within the cage,
with a horizontal component (H) of 0.20 G, a vertical component (Z) of 0.49 G, a total field intensity
(F) of 0.52 G, and an angle of inclination (I) of 68°. When the coils were deactivated, the magnetic
field within the cage was that of the earth's normal field at 52° latitude: H = 0.17 G; Z = 0.45 G; F
= 0.48 G; I = 69°. Redrawn from Mather and Baker (1981).
involved the normal geomagnetic field plus restriction of visual and olfactory cues during
displacement, followed by similar conditions at the test site. The experimental treatment
involved reversal of the horizontal component of the magnetic field plus restriction of
visual and olfactory cues during displacement, followed by continued restriction of visual
and olfactory cues at the test site. Tests in this series were arranged in a sequence of two
of each condition alternately. Series II followed the first series and comprised a single test
treatment which did not employ the test site screen. The treatment involved reversal of
the horizontal component of the magnetic field plus restriction of visual and olfactory cues
during displacement, followed by the normal geomagnetic field and relatively unrestricted
access to visual and olfactory cues at the test site. The orientation cage was rotated through
90° between tests in this series.
After removal from the trap of capture, the mouse was allowed to enter a translucent
scroll, the diameter of which could be adjusted to suit each animal's size. This scroll loosely
restricted the mouse's movements, so ensuring that the animal both faced the direction of
travel and was retained in a central position between the two lateral coils during dis-
placement (see Fig. 4). The scroll plus mouse was placed inside the transportation cage
which was then made airtight by sealing the lid and lightproof by covering with black
polythene sheeting. At this juncture, depending on the test treatment, the coils could be
activated and onwards from this point the transportation cage was maintained along a
north-south axis. A 2-min adaptation period elapsed before the transportation cage was
carried in a fairly direct path 40 m due north to the test site where the orientation cage
had been positioned with its four arms aligned with the cardinal compass points. Once at
the test site, the coils could be deactivated, care being taken to ensure that the sun's position
Rodents 517
SERIES I
Figure 5. Influence of magnetic conditions during With screen
displacement on the goal orientation of woodmice. Normal Reversed
The woodmice were displaced and then tested in magnetic field magnetic field
an orientation cage. Two experimental series were
trap trap
conducted, involving three different test treatments
and a total of 53 separate displacement occasions. I I
Thirty-five individual mice (25 males and 10 fe-
males) were used, no animals being tested more
than once for any treatment. Series I comprised two
test treatments: control, involving the geomagnetic
field during displacement, and experimental, in-
volving reversal of the ambient magnetic field dur-
ing displacement. In these two treatments, visual
and olfactory cues were restricted both en route and Z=2.940 Z=4.0SS·
at the test site. Series II comprised a single test treat- u=2.353** u=1.S76*
ment involving reversal of the magnetic field dur-
ing displacement, and restriction of visual and ol-
factory cues en route but not on site. Each dot trap
represents an individual's mean angle of orienta- t
,•
tion, derived from the time spent in each of the four
arms of the cage during a 4-min test. Angles are
expressed relative to trap direction. Arrows show
the mean vectors for the three treatments and the SERIES II
dashed lines denote the 95% confidence interval Without screen
around each sample mean. Statistics as in Fig. 2.
V-test relative to trap direction for Series I controls
and Series II, and relative to 180 from the trap for
0 •
Series I experimentals. Redrawn from Mather and Z=4.662*·
u=3.031···
Baker (1981).
could not be detected by the mouse during the period from activation to deactivation of
the artificial magnetic field. The scroll was then removed from the transportation cage (the
animal spent approximately 4 min inside this apparatus) and the mouse was transferred
to the orientation cage. Once more, testing consisted of a 4-min observation period during
which the time spent in each of the four arms of the cage was recorded (small peepholes
were present in the screen) and again the trap of capture was considered as the goal.
The results of Series I (Fig. 5) show a clear dichotomy in the directional preferences
of the control and experimental mice, the mean angles differing by 117° and the distri-
butions of the two groups being significantly different (Watson's test, U217.16 = 0.402, P
< 0.001). After a displacement journey involving the earth's magnetic field and restricted
access to visual and olfactory cues, the mice were able to orient in the direction of the
trap despite the continuing restriction of visual and olfactory information by the screen
and cage at the test site. However, when the mice experienced similar displacement and
test site conditions except that the declination of the magnetic field was deflected by 180°
during displacement, mean orientation shifted away from the point of capture by 131°.
The orientation of these experimental animals has a nonrandom component in a direction
opposite to the trap (V-test), although the 95% confidence interval around the sample mean
does not quite include the predicted direction (see Aneshansley and Larkin, 1981). A
nonrandom component in a direction opposite to the controls' mean angle is not apparent
(V-test, u = 1.298, P > 0.05). The nonuniformity in orientation by the experimentals
implies that the artificial reversed magnetic field during displacement does not simply
disrupt the animals' ability to orientate, but rather causes a rotation of the animals' estimate
518 Chapter 25
of trap direction. This probably results from a misinterpretation of the true direction of
travel, although a less likely possibility is that short-term aftereffects of exposure to altered
magnetic fields may somehow affect subsequent navigation at the test site. The significant
difference in the directional preferences of the control and experimental mice does in-
dicate, however, that the animals can use directional information from the ambient mag-
netic field for goal orientation, and so provides convincing evidence for magnetoreception
ability in rodents.
The results of Series II (Fig. 5) show that when subjected to a reversed magnetic field
and restricted access to visual and olfactory cues during displacement but with relatively
unrestricted access to cues at the test site, the mice were again able to orient toward the
trap, unlike mice subjected to similar displacement conditions but with the screen present
at the test site (Le., Series I, experimentals). The mean angles of orientation for these two
groups experiencing a reversed magnetic field during displacement differ by 124°, and the
orientation distributions are significantly different (Watson's test U220.16 = 0.273, P <
0.01). This indicates that magnetic information gathered en route may be overridden by
geographical information available on site should there be any discrepancy among the
different cues. Orientation by mice experiencing the geomagnetic field plus the screen
(Series I, controls) does not appear to differ from that by mice experiencing a reversed
magnetic field but no screen (Series II) (Watson's U220.17 = 0.063, P > 0.05). No difference
in the orientation performance of male and female mice is evident for any of the test
treatments (Series I, control: U2 11 •6 = 0.094, P > 0.05); Series I, experimental U2 12 •4 =
0.092, P > 0.05; Series II: U 2 15 •5 = 0.085, P > 0.05}.
Having monitored their outward journey by reference to the ambient magnetic field,
thereby establishing displacement location relative to the trap, the woodmice then have
to translate trap direction into a heading for travel. In theory, this could be achieved by
reference to test site compass cues (such as the sun, polarized light, or the geomagnetic
field) and/or, when available, geographical information. For example, we might expect
better orientation under sunny skies than under overcast due to better cues for the trans-
lation of geomagnetic orientation into a travel heading. Orientation by the woodmice has
been analyzed with respect to conditions of sun (Le., if the sun shone at some stage during
testing, although the sun's disk was rarely viewed directly from within the orientation
cage when the screen was in position) and cloud (Fig. 6) to examine for the use of compass
cues at the test site.
In Series I (screen present), under sun, the directional preferences of the control and
experimental mice do have nonrandom components in their respective predicted direc-
tions, the mean angles separating by 118° and the two distributions being significantly
different (Watson's U2 1O •1O = 0.296, P < 0.001). Under overcast, however, only for the
controls does the sample mean have a nonrandom component as predicted, no difference
in the two distributions being found (Watson's U2 7 •6 = 0.066, P > 0.05) though sample
size is small. Such results are suggestive of the involvement of a celestial compass at the
test site, but comparison of the control group's orientation under sun to its orientation
under cloud does not show any significant difference (Watson's U2 10 •7 = 0.064, P > 0.05),
a similar result being found for the experimentals (Watson's U2 1O •6 = 0.055, P > 0.05).
In Series II (no screen), orientation toward the trap occurs for both conditions of sun
and overcast, and although the animals' performance does seem to improve when sunny,
the difference does not reach significance (Watson's U 2 13 •7 = 0.122, P > 0.05). The avail-
ability of geographical information during this series, however, may lessen any reliance
on compass cues.
It seems clear that rodents can make use of a geomagnetic compass, but it is not yet
known whether they use the dip angle (as found for birds; Wiltschko and Wiltschko, 1972)
or the polarity of the magnetic field, as investigations so far have involved only manip-
ulation of the horizontal component of the ambient field which alters both elements.
Rodents 519
Sun Overcast
trap trap
! 1
SERIES I
With screen
eNormal field
oReversed field
Figure 6. Influence of sun and
overcast on the goal orientation of
woodmice subjected to various •
Z=1.72S
Il
Z=4.466** •
Z=1.382
Il
Z=O.S50
displacement conditions. Data u=1.671* u=2.429*· u=1.662* u=O.099
from Series I and II, presented in
Fig. 5, are analyzed with regard to trap trap
conditions of sun and cloud at the ! t
time of testing. Each dot (Series I:
., control; 0, experimental) rep- SERIES II
resents an individual's mean angle o
o
of orientation during a 4-min test Without screen
in the orientation cage. Angles are Reversed field a=+15
expressed relative to trap direc- r=O.781
tion. Arrows show the mean vec-
tors for the various test treatments
under sun and cloud. Statistics as Z=4.26S·· Z=1.415
in Fig. 2. V-test as in Fig. 5. u=2.S2S*· u=1.6S1·
I have further examined the ability to access compass information from the magnetic
field in a pilot investigation of the ontogeny of the magnetic sense in rodents (Mather,
1981a; Mather et al., 1982). These laboratory experiments involved juvenile golden ham-
sters aged between 20 and 30 days at the time of testing, and the cross-shaped orientation
cage (Fig. 1) was again employed to determine the animals' directional preferences. In each
of the four arms of the cage, an activity wheel which could rotate in a single direction
only was positioned such that the direction of rotation was away from the cage center. A
mechanical counter attached to each wheel recorded the number of revolutions per arm.
Inside the experimental room, dim illumination was provided by a red light bulb. The
orientation cage was positioned within a pair of Helmholtz coils (Fig. 7) which, when
activated, induced reversal of the horizontal component of the magnetic field within the
cage. The entire apparatus was housed within a lightproof screen (a cuboid wooden frame
supporting black opaque polythene sheeting), access being gained via two overlapping
sheets directly above the cage. Each experiment on an individual hamster comprised eight
consecutive 10-min tests (80 min total) and the animal was exposed to either the normal
laboratory magnetic field (N) or the reversed magnetic field (R) during each 10-min period.
Experiments were conducted using alternately the test sequences R-N-R-R-N-R-N-N
and N-R-N-N-R-N-R-R. After removal from the natal cage, the hamster was transported
in a small cardboard box to the experimental room (the displacement journey involving
a distance of 7 m and two 90° turns) and introduced into the orientation cage. The cage
was then made lightproof and, depending on the particular test sequence, the magnetic
field either remained unaltered or was reversed. The investigator returned at the end of
each 10-min period to determine the number of wheel revolutions from the four counters,
and when necessary to change the status of the coils according to the test sequence. The
cage was rotated through 90° between experiments on each individual.
520 Chapter 25
Figure 7. Helmholtz coil system

used to manipulate magnetic field
conditions at the test site. The ori-
entation cage was positioned on a
wooden stool within a pair of
Helmholtz coils. The four arms of
the cage were aligned with the car-
dinal compass points, the coils
ORTH being sited at the ends of the north
and south arms. Each coil con-
sisted of 200 turns of 40 s.w.g. in-
sulated copper wire, the two coils
being connected in series and
powered by a 12-V car battery.
When connected (via a 5-m lead)
to the battery, the coils produced
a horizontal magnetic field of 0.35
O.S m G, which when opposed to the hor-
izontal component of the normal
magnetic field in the laboratory
(0.17 G), resulted in a net horizon-
2 v car _ tal component of 0.18 G opposite
ba tery in direction to that of the labora-
tory field . When the coils were ac-
tivated, therefore, there Was in the
position of the orientation cage a
near-mirror image of the normal laboratory magnetic field, with a horizontal component (H) of 0.18
G, a vertical component (Z) of 0.43 G, a total field intensity (F) of 0.46 G, and an angle of inclination
(I) of 67°. When the coils were deactivated, the magnetic field within the orientation cage was that
of the normal laboratory field : H = 0.17 G; Z = 0.43 G; F = 0.46 G; I = 68°.
Each hamster's mean orientation of wheel activity during both normal and reversed
magnetic field conditions is presented in Fig.8a. Orientation in the normal field is directed
toward the natal cage despite directional information being restricted both during dis-
placement and at the test site. Under similar experimental conditions but with the magnetic
field reversed at the test site, orientation seems to shift away from the natal cage, though
a nonrandom component is not evident either in a direction 180° from the natal cage (V-
test) or 180° from the mean orientation in the normal field (V-test, u = 1.132, P > 0.05).
Although orientation performance for the two conditions appears to differ, the difference
is not significant (Watson's U215 .15 = 0.142, P >; 0.05). However, if the angular difference
between an individual's mean orientation in the normal field and its mean orientation in
the reversed field is determined, the group data show a bimodal distribution (Rao's test,
U = 199.0°, P < 0.01), implying that some animals respond to changes in the ambient
magnetic field whereas others do not.
Further analysis considers the age of the animals at the time of testing. For hamsters
aged 20-25 days, orientation is uniform in both the normal (0 = + 27°; r = 0.375 ; Z =
0.845, P > 0.05; u = 1.158, P > 0.05) and the reversed (0 = + 52°; r = 0.278; Z = 0.464;
P > 0.05; u = - 0.593, P > 0.05) magnetic field, and the distributions do not differ (Wat-
son's U 2 6 •6 = 0.032, P > 0.05). For hamsters aged 26-30 days, however, orientation in the
normal field is directed toward the natal cage (0 = + 13°; r = 0.497; Z = 2.225, P > 0.05;
u = 2.055, P < 0.05), and orientation in the reversed field (0 = -127°; r = 0.606; Z =
3.303, P < 0.05), although not directed 180° away from the natal cage (u = 1.547, P >
Rodents 521
(a) Natal cage
Figure 8. Influence of magnetic field con- I

ditions at the test site on the orientation of
juvenile golden hamsters. Hamsters (15 an-
.WJ. •
0:••
0
imals from three litters) were displaced and

o a=+18
\
then tested in an orientation cage which • normal field ú= r=O·445 •
was positioned within Helmholtz coils and o reversed field o l!.
surrounded by a lightproof screen. Each ex- 00 a=-127
r=0252
• Z=O·955
u=O·832
periment on an individual comprised eight 0.0
tests, involving a sequence of normal and .0 0
00
reversed magnetic field conditions, during
which wheel activity was recorded.
(a) Orientation of wheel activity dur-
ing normal and reversed magnetic field (b) Mean orientation In
normal magnetic field
conditions. For each hamster, the total
number of wheel revolutions in each of the I
four arms of the cage during each of the
eight 10-min tests was used to calculate a
mean angle of orientation for each 10-min
period. An individual's eight mean angles • age 20 - 25 days
were then used to calculate two second- o age 26 - 30 days l!.
order mean angles, one derived from the o a=-148 Z=0·459
four mean angles for the normal magnetic r=O 226 o u=0·813
field and the other from the four mean an-
gles for the reversed magnetic field. Each
dot thus represents a second-order mean
angle for an individual's orientation of wheel activity during the particular magnetic condition. Angles
are expressed relative to the direction of the natal cage. Arrows show the third-order mean vectors
for the two magnetic conditions. Statistics as in Fig. 2 but at third-order level. V-test relative to natal
cage direction for normal magnetic field data, and relative to 180° from the natal cage for reversed
field data.
(b) Angular difference between the mean orientation of wheel activity in normal and in reversed
magnetic field conditions. Each dot represents the angular difference between an animal's second-
order mean orientation of wheel activity in the normal magnetic field and the same individual's mean
orientation in the reversed field. Angular difference is expressed as positive or negative deviation
from mean orientation in the normal field, which is set at 0°. The arrows show the third-order mean
vectors of the angular differences for the two age categories. Statistics as in Fig. 2 but at third-order
level. V-test relative to 0° for younger hamsters, and relative to 180° for older hamsters.
0.05), is directed 180° from the mean orientation in the normal field (u = 1.969, P < 0.05).
For these older animals, even though the sample means for the two magnetic field con-
ditions differ by 140°, no significant difference in distribution is found (Watson's U 2 g •g =
0.147, P > 0.05). However, if age is taken into consideration when examining the angular
difference data for which bimodality has been shown (Fig. 8b), a highly significant dif-
ference in the distributions of the two age categories becomes evident (Watson's U2 g •6 =
0.311, P < 0.001).
It is tempting to interpret these results as suggesting that orientation response to the
magnetic field changes with age, and that a particular stage in ontogeny has to be attained
before the hamster is capable of deriving directional information from the magnetic field.
Such a latency in magnetoreceptive ability in rodents could, perhaps, be due to physio-
logical constraints such as incomplete development of the sensory mechanism during early
life, or due to the animals' having to learn how to make use of a magnetic compass. In
view of the post hoc analysis of age category, however, suggestions must remain tentative
until further investigations have been completed.
522 Chapter 25
4. The Search for the Magnetoreceptor
The influence of magnetic fields on the orientation of a wide variety of organisms

from bacteria to vertebrates is well documented (see Kirschvink, 1983). However, excepting
bacteria, the physiological basis of magnetic field detection remains obscure, although
several theories have been advanced such as optical resonance, induction, and the in-
volvement of permanent magnets and superparamagnetic substances (for discussion see
Kirschvink and Gould, 1981). Biogenic magnetite is directly involved in the orientation
response of bacteria to magnetic fields (Frankel et 01.,1979), and the subsequent discovery
of magnetite deposits in animals known to be sensitive to earth-strength fields, such as
bees (Gould et 01., 1978) and homing pigeons (Walcott et 01., 1979), suggests a similar
mechanism in eukaryotes. The demonstration of magnetoreception ability in rodents has
led to a search for the magnetoreceptor in these animals. So far, this search has incorporated
several lines of investigation including magneto metric , histological, and neurophysiol-
ogical techniques.
4.1. Magnetometric Investigations
Following the discovery of magnetite in the pigeon head (Walcott et 01., 1979), I de-
cided to ascertain whether similar deposits are present in the rodent head (Mather and
Baker, 1981). Initially, efforts concentrated upon magnetometric techniques which in-
volved the use of a par-astatic magnetometer. Three species of wild rodents, the European
woodmouse, the bankvole, and the house mouse, Mus musculus, were examined, all an-
imals having been killed in the field by domestic cats and deep frozen for up to 3 months
before testing. Preparation of the heads ready for insertion into the magnetometer involved
skinning (using glass dissecting instruments to prevent possible contamination by metallic
tools), followed by embedding in deionized ice within a small plastic container (volume
10 cm 3 ).
Intact heads were first tested for natural remanent magnetization (NRM, a residual
field resulting from a preferential alignment among individual magnetic particles), and
then tested for isothermal remanent magnetization (IRM, a measure of the total magnetic
material present). Before testing for IRM, the specimens were exposed to a 2000-G elec-
tromagnetic field to align any magnetic material existing in the sample, exposure occurring
immediately prior to testing in the magnetometer as the IRM tended to decay fairly rapidly.
Readings obtained for intact heads are presented in Table 1. Only 1 of 11 specimens ex-
hibited NRM significantly above mean background noise level, whereas 8 of 12 specimens
exhibited IRM significantly above background. The mean IRM value, approximately 6 x
10- 6 emu/g, is nearly five times greater than the mean NRM value. Such results indicate
the presence of magnetic material in the rodent head.
In an attempt to localize the site of the magnetic deposits, the heads were subsequently
dissected (using glass tools), and the various tissue samples were again exposed to the
2000-G field and retested for IRM (Table 1). The highest levels of IRM, 5 x 10- 6 emu/g,
were found in the anterior/dorsal part of the head. Unfortunately, owing to cumbersome
dissecting tools, a precise location for the magnetic material is difficult to define, although
the area of high IRM seemed to be associated with the olfactory region exterior to the brain
(removal of brain tissue had little influence on the readings).
Similar levels of IRM for the woodmouse head have been found in a more recent
magnetometry investigation using a SQUID cryogenic magnetometer, the mean saturated
IRM for three adult specimens, one male and two females, being 2 x 10- 6 emu. (Kirschvink,
personal communication). A magnetometric examination of M. musculus (Jackson's strain)
Rodents 523
Table I. Summary of NRM and IRM in the Intact and Dissected

Rodent Head a
Mean magnetic
moment (x
N 10- 6 emu/g) Range
Container
Empty NRM 4 0.63 0.28- 1.05
Plus ice NRM 4 0.83 0.59- 1.23
IRM 4 0.96 0.51- 1.45
Whole head NRM 11 1.33 0.29- 7.81
IRM 12 6.31 1.14-18.31
Dissected head
Cranium IRM 3 3.08 1.20- 5.57
Lower jaw IRM 7 0.75 0.43- 1.64
Cranium
Anterior IRM 6 5.13 1.26-16.13
Posterior IRM 9 1.23 0.59- 2.55
Anterior cranium
Anterior IRM 5 1.93 0.47- 4.22
Posterior IRM 5 0.82 0.27- 2.01
Dorsal IRM 6 1.92 0.71- 4.01
Ventral IRM 6 1.31 0.71- 2.99
a Three species of wild rodents were tested, a total of 12 animals in all. Only readings
greater than twice mean background noise level, where background includes the
NRM or IRM of a control plastic container plus ice, were considered to be signif-
icant. From Mather and Baker (1981).
has also been conducted (Buchler and Wasilewski, 1982). Three male and two female mice,
aged 2-3 months, were prepared (i.e., skinned, digestive tracts removed, and limbs severed
at the median joints) using nonmetallic tools, and tested in a SQUID magnetometer. The
carcasses, mean weight 5.5 g, had a mean natural remanence of 6.8 x 10 -7 emu and a mean
saturated remanence of 2.1 x 10- 6 emu, the mean IRM being three times greater than the
mean NRM.
Levels of IRM obtained for the rodent head are comparable to those reported for the
heads of other vertebrates, such as pigeons (Walcott et 01., 1979) and dolphins (Zoeger et
01., 1981). The finding of significant levels of IRM for tissues of both wild and laboratory
rodents suggests that substantial amounts of magnetic material are of common occurrence
in all rodent species.
The character of the magnetic material in the rodent head has been examined by look-
ing at the coercivity spectrum of the magnetic particles, this research being in collaboration
with ]. L. Kirschvink (California Institute of Technology). The coercivity spectrum is de-
termined using the two techniques of progressive alternating field (AF) demagnetization
and IRM acquisition, and gives an indication of particle size, shape, and domain state
(Kirschvink, 1983). The AF demagnetization and IRM acquisition curves for the wood-
mouse head are shown in Fig. 9, the curves being an average for three specimens. The
intersection point of the two curves gives an estimate of the coercive field of 44 mT (440
G), suggesting that the magnetic material is in the form of single-domain magnetite. The
measure of intergrain interaction (R) of 37% and the asymmetry of the two curves indicate
a fair degree of grouping among the magnetic particles, perhaps as chains similar to the
magnetite crystals in magnetotactic bacteria (Frankel et a]" 1979), or as small isolated
524 Chapter 25
100
Af de mag IRM
ú=
!!.... 60
QI
en
'ú"=
u
QI
c:J
'-
o
40
QI
en
'"t;
QI
1: 20
ç HJ J J J J ú ú ú J J J J ú J J J J J K J J | K J J ú J J J J J J | K J J ú J J ú =
1 3 5 10 30 50 100 300 500 1000
Peak Field, mT
Figure 9. Progressive AF demagnetization and IRM acquisition curves for the woodmouse head. Plots
show the average results from tests on three adult woodmice, one male and two females. The inter-
section point of the two curves gives an estimate of the coercive field of 44 mT, and a measure of
intergrain interaction (R) of 37%. Such results suggest the presence of moderately interacting single-
domain magnetic crystals. Data from J. L. Kirschvink (personal communication).
clumps. A coercivity spectrum resembling that found for the woodmouse head is reported
for other vertebrate tissue, such as salmon ethmoid (Kirschvink, 1983).
4.2. Histological Investigations

As the magnetometric investigation of the rodent head yielded only a general area,
the anterior/dorsal cranium, for the location of the magnetic material, histological tech-
niques were employed to search for a more precise anatomical region (Mather ef 01.,1982).
So far, four species of rodents have been examined: the European woodmouse (subadult
and adult, age unknown as these wild animals were the quarry of domestic cats); the
laboratory mouse (juvenile, age 10 days); the laboratory rat, Brattleboro strain (juvenile,
age 5 days), and Long-Evans strain (sub adult, age 7 weeks); and the golden hamster (adult,
age 18 months).
Histological preparation of the specimens was carried out in collaboration with John
H. Kennaugh (University of Manchester). Whole heads were initially fixed in 5% neutral
formalin for at least 2 days before undergoing standard wax embedding, after which the
embedded specimens were sectioned using an MSE sledge microtome. The sections (12
11m thick) were picked up using pressure-sensitive tape, then mounted on slides coated
with Mayer's albumin and dried in an oven at 50°C for 2 days. Tape and adhesive were
removed by immersing the slides first in acetone and then chloroform, and the slides were
taken through xylene and down the alcohol series to distilled water. Sections were stained
for ferric iron using the Perl (Prussian or Berlin blue) reaction which involved placing the
slides in a warm solution of equal parts of 4% potassium ferrocyanide and 4% HCI either
Rodents 525
for a maximum of 10 min in an oven at 60°C (Hutchinson, 1953) or for 30 min at room
temperature. Following this, the slides were washed in distilled water, dehydrated in the
alcohol series, lightly counterstained in eosin, cleared in xylene, and finally mounted in
Depex. Conventional light microscopy was used to examine the sections, any deposits of
ferric iron appearing as a deep blue coloration. Control tests showed that magnetite par-
ticles of geological origin produced an intense blue precipitate after the above procedure.
Taking the magnetometry results into consideration, initial histological examination
focused on the anterior/dorsal region of the rodent head (see Fig. lOa). For all animals
except the laboratory mouse and the Brattleboro strain of rat, sections of this region stained
positively in the Perl reaction, indicating the presence of ferric iron material. It is perhaps
significant that those animals for which no staining was apparent were juveniles (10 and
5 days old, respectively), especially as younger rodents appear not to respond to ambient
magnetic fields (vide supra). For the woodmouse, the Long-Evans rat, and the golden ham-
ster, all of which were either subadult or adult, a positive result was observed in the scroll
bones of the olfactory region, in particular the ethmoturbinal bones (Fig. lOb). The ferric
material was in the form of a continuous band ranging from 5 to 10 f.Lm in thickness and
5 to 20 f.L beneath the bone surface. No staining was evident in any of the associated soft
tissue, such as brain, membrane, or muscle. As reported in other studies (Walcott and
Walcott, 1982), hemoglobin, an iron/protein complex, does not stain in the Perl reaction.
Nervous innervation of the iron-containing region was not apparent.
At this stage in the search for the rodent magnetoreceptor, it seemed possible that the
subsurface layer of ferric iron in the ethmoturbinal bones might be the source of the elevated
levels of IRM found for the anterior/dorsal region of the head. This idea was encouraged
by the subsequent finding for the human head of a visually identical layer of ferric-staining
material beneath the surface of the sphenoid/ethmoid sinus bones (Baker and Mather, 1982;
Baker et a1., 1982), an area later reported to be magnetic (Baker et a1., 1983). Further support
came from reports of the presence of magnetite in the ethmoid bone complex of tuna
(Walker and Dizon, 1981) and salmon (Kirschvink, 1983), and also the finding of iron-
staining sites in the vicinity of the olfactory region in pigeons (Walcott and Walcott, 1982).
However, this circumstantial association between the ferric layer and the high remanence
was soon to be questioned when it was discovered for the common marmoset, Callithrix
jacchus, that these subsurface deposits were not confined to bones in the olfactory region
but extended throughout all bones of the head, although layer width and density did vary
(Mather and Kennaugh, unpublished).
In view of this, we carried out an extensive search of the rodent head. Serial sections
of the entire woodmouse head (Fig. lOa) confirmed the widespread distribution of the
ferric layer, the band of blue-staining material being present beneath the surfaces, often
both inner and outer, of most bones of the skull. Furthermore, small isolated concentrations
of ferric iron were found scattered throughout some areas of bone marrow, particularly in
the ethmoturbinal region. Moreover, the root areas of the teeth in the upper and lower jaw
were filled with dense ferric deposits (Fig. 10c). The search for iron was then extended to
include the woodmouse femur, and examination revealed a ferric layer along the inner
surface of the bone next to the medullary cavity, as well as small ferric concentrations
within the bone marrow. Evidently, ferric iron material is abundant in rodent skeletal
tissue.
A variety of iron-staining sites are reported for other vertebrates, some being analogous
to those found in the woodmouse. For pigeons, for example, in addition to the three sites
close to the olfactory region, diffuse blue staining is found in the bone matrix of the dorsal
cranium, particularly at the surfaces and in areas of muscle attachment. Also, for cave
salamanders, a diffuse blue stain occurs in the bone matrix at the base of the teeth (Walcott
and Walcott, 1982).
526 Chapter 25
a.
Posterior
Sm
Figure 10. Sections through the head of a wood mouse, showing deposits of ferric iron. (aJ Horizontal
section through the entire head showing: A, the ethmoturbinal bones of the olfactory region; B, the
Rodents 527
At present, the significance of the various ferric iron deposits remains obscure, al-
though tentative suggestions may be offered. As the subsurface ferric layer appears to be
present in most bones of the body, it seems unlikely that its primary function is concerned
with magnetoreception. One possible explanation for its occurrence is that bones are used
as a storage or dumping site for iron. Another possibility is that the deposits may be in-
volved in the growth and repair of bones. Interestingly, the healing rate of fractures seems
to be enhanced by applied electromagnetic fields (Bassett et 01., 1974); perhaps the iron
exerts some influence on the orientation of new bone fibers, so affecting the organization
and strength of the repair process. Similarly, if the small isolated concentrations of ferric
iron within bone marrow are of common occurrence in most bones, it is again likely that
their major role is concerned with something other than magnetic field detection, perhaps
erythrocyte production. Certainly, the iron-storage compound ferritin is reported to be
present in bone marrow (Granick, 1946).
However, even if the function of these ferric iron deposits in skeletal tissue is largely
unrelated to magnetoreception, this does not preclude the possibility that the ferric material
in a specific region of the body may be involved in the detection of magnetic fields. Con-
sidering the magneto me try results for the rodent head, it is possible that ferric deposits
within the bones of the olfactory region may have different magnetic charcteristics to
deposits elsewhere, perhaps resulting from the conversion of some of the iron material
into a more magnetic form suitable for magnetoreception. In particular, magnetite in the
form of superparamagnetic or single-domain crystals would be suited to this purpose
(Kirschvink and Gould, 1981). If bones in the nasal region have become specialized for
magnetic field detection, it is likely that only a small fraction of the total iron present
would have to be converted into magnetite in order both to form a magnetoreceptor and
c.
SOO ).1m
Figure 10. (continued)

root of a tooth in the upper jaw; C, the nasal septum; D, the eye (C and D given for reference). (b)
Sagittal section through the ethmoturbinal bones of the olfactory region, showing a continuous layer
of ferric iron material in a band beneath the surface of the bone. (c) Horizontal section through the
region of a tooth in the upper jaw, showing both the dense deposits of ferric iron in the tooth root
and the continuous layer of ferric material beneath the bone surface.
Adult, female specimen. Sections stained using the Perl reaction. Sections prepared by J. H. Ken-
naugh.
528 Chapter 25
to account for the levels of remanence found for the rodent head. For example, it has been
calculated that only a few hundred single-domain crystals of magnetite would be sufficient
to provide an animal with an accurate magnetic compass (Yorke, 1979). Also, it is reported
that although a substantial quantity of blue-staining hydrous iron oxide (typically para-
magnetic at room temperature and so not contributing to remanence levels) is present in
the abdomen of the honeybee, only 0.33% would have to be reduced to magnetite in order
to account for the magnetic moment of bees (Kuterbach et al., 1982). Thus, it is possible
that the ferric deposits in rodent skeletal tissue are principally in the form of hydrous iron
oxides, such as ferritin, but that in the olfactory region some of the iron may be reduced
to form single-domain magnetite, so accounting for both the magnetometry results (Table
I) and the coercivity spectrum results (Fig. 9) obtained for the rodent head.
In a recent collaborative investigation with J0rgen M0rup J0rgensen (University of
Aarhus, Denmark), transmission electron microscopy has been used to examine the ultra-
structure of the ferric material within the rodent head. The heads of two adult male woodm-
ice were dissected using glass tools and bones were removed from the ethmoturbinal,
nasoturbinal, and posterior cranial regions. Excised tissue was fixed and embedded con-
ventionally and the sections (2 J.Lm thick) were stained with uranyl acetate and lead citrate.
The micrographs (Fig. 11) show that the ethmoturbinal bones contain groups of electron-
opaque crystal-like particles which appear to be located along the interface between the
bone and marrow. Individual particles within each group range in size from 4 x 2 J.Lm,
comparable to hydrous iron oxides, down to 100 nm, within the size range of single-domain
magnetite. The particles are situated close to nerve tissue, although the exact relationship
between particles and nerves is as yet unclear but should soon be elucidated from serial
sectioning. Sections of the same material as used for TEM have also been cut and stained
for iron using the Perl reaction. Positive staining is found in groups like the TEM findings,
indicating that these crystallike particles contain ferric iron. From preliminary examina-
tions, such ferric particles do not appear to be present in the nasoturbinal or posterior
cranial bones.
If the vetebrate magnetoreceptor mechanism is based on magnetite and requires a fixed
orientation, rather than rotation, of the magnetite particles, it may be of advantage to embed
the ferric deposits within a bony matrix. As bone. is permeated by blood and lymphatic
vessels, the site of magnetite deposition could remain accessible to transport fluids, in
which case a constant turnover of iron may occur such that it is present in a variety of
forms from magnetite precursor (ferritin?) through to stable magnetite. This would allow
the periodic renewal of those particles involved in the magnetic compass, which may be
necessary if the size, alignment, or clumping of crystals alters with time, thereby interfering
with the compass mechanism.
Other explanations are possible for the high remanence levels found for the anterior/
dorsal region of the rodent head. Assuming the entire subsurface layer to be weakly mag-
netic, it may be simply that the large surface area of the scroll bones in the nasal region
results in a greater concentration of iron material, so increasing remanence to a level de-
tectable by our magnetometer. Of course, it is also possible that all ferric deposits so far
located are paramagnetic and unrelated to either the IRM values or magnetoreception abil-
ity, in which case the source of the remanence and the location of the magnetic sense
organ have still to be determined.
A further point to note is that even though the rodent jawbones contain considerable
dense ferric deposits in the roots of teeth, remanence levels for the jaw were low, implying
that such deposits are composed of a relatively nonmagnetic materiaL This material also
may be ferritin, as this is known to be present in rodent teeth, e.g., in the cells associated
with the maturation of enamel (Reith, 1961).
As ferric iron deposits such as magnetite (Kirschvink, 1983) and ferritin (Granick,
1946) are widespread in animal tissue, these compounds may well play other roles in
Rodents 529
Figure 11. Transmission electron micrographs of ferric particles in the ethmoturbinal bones of the
woodmouse head. (a) Low magnification showing a group of electron-opaque crystal-like particles
which are located along the bone-marrow interface. The particles are situated close to nerves and
stain positively in the Perl reaction. (b) High magnification of a single crystal-like ferric particle.
Adult, male specimen. Sections stained with uranyl acetate and lead citrate. Micrographs provided
by J. M0rup J0rgensen.
530 Chapter 25
biological systems, which may account, in part, for the influence of magnetic fields on
many aspects of animal physiology.
4.3. Neurophysiological Investigations
The presence of magnetite in the rodent body does not confirm its use in magnetic
field detection; obviously, the magnetoreceptor must have the appropriate connections to
the central nervous system. Consequently, in addition to the many lines of research aimed
at locating deposits of magnetic material, investigations have also been directed toward
the identification of some central nervous structure which may be part of the magnetic
sensory system.
A neurophysiological investigation of the guinea pig, Cavia porcellus, has shown that
the electrophysiological activity of single cells in the posterior region of the pineal organ
is influenced by ambient earth-strength magnetic fields (Semm et a1., 1980). Helmholtz
coils below the jaw and above the head generated an artificial magnetic field, the polarity
of which could be altered such that it added to, compensated, or inverted the vertical
component of the local magnetic field. Inversion of the vertical component resulted in the
angle of inclination being changed from + 63° to - 63°, while total intensity and north
direction remained the same. When a magnetic field stimulus of + 0.5 Oe was applied for
17 min, out of 71 cells tested (from 16 male animals), 15 cells (from 11 animals) showed
a significant depression in activity, clear effects being apparent after a latency period of
approximately 2 min. After switching off this field, the cells retained a diminished activity
and a second stimulus of shorter duration had no effect even when the field strength was
varied. Inversion of the magnetic field ( - 0.5 Oe) then resulted in 3 of the 15 responsive
cells restoring their activity to initial output level, two cells reaching this level after 8-
12 min. When the inverted field was switched off, the cells retained their normal activity
which could be depressed again by a low-magnetic-field stimulus of + 0.1 Oe, suggesting
a relatively high sensitivity of the cells. In these experiments, the pineal cells were re-
sponding to a rapid change in the vertical component of the magnetic field, and a more
recent report suggests also a response to gradual changes in the ambient field (Semm et
a1., 1982).
The electrophysiological activity of pineal cells of pigeons is also influenced by earth-
strength fields (Semm et a1., 1982), but the latency period for pigeon cells is in the range
of milliseconds, far shorter than found so far for guinea pigs. It is not yet clear whether
the cells react primarily to changes in the dip angle or to changes in intensity of the
magnetic field. Nor is it known whether or not the pineal itself senses changes in the
magnetic field. According to the investigators, the effect may well be indirect, especially
as the organ is hevily innervated by sympathetic fibers from the superior cervical ganglia
and the sympathetic nervous system can be affected by magnetic stimuli. Furthermore,
the lack of appreciable quantities of iron in the pineal organ makes it unlikely that mag-
netite for the possible transduction of magnetic stimuli is present. As a result of their
research, Semm et a1. have proposed that the pineal organ may be part of a magnetic
compass; they consider that as the pineal is a light-sensitive timekeeping organ and ti-
mekeeping is required for sun-compass orientation, this part of the brain would be an ideal
site for integration of a combined compass-solar-clock system.
Whether other regions of the central nervous system are involved in the perception
and/or transmission of magnetic information remains to be discovered. Other brain struc-
tures, such as the inferior and superior colliculi, corpus callosum, and epithalamus, have
been tested but no reaction to magnetic stimuli could be measured (Semm et a1., 1980).
Biochemical investigations of the influence of magnetic fields on the secretory activity
of the pineal gland of rats have revealed that the activity of the enzyme N-acetyltransferase,
Rodents 531
and correspondingly the melatonin content, of the pineal during the night is strongly de-
pressed following exposure of freely moving animals to an altered magnetic field (inversion
of the horizontal component) for 15 min (Welker et aI., 1982). It has been proposed (Semm
et aI., 1982) that melatonin may influence the compass system in the pineal or transmit
information regarding altered magnetic fields to other regions of the brain.
5. Summary
Magnetic fields of various strengths, ranging from 0.5 mG to 9000 G, exert an influence
on numerous aspects of rodent physiology. In general, the biological effects are not in-
stantaneous but are observed some time after the application of the artificial field. A more
immediate response to the ambient magnetic field, indicating detection of the field per se
and therefore the involvement of the central nervous system, is demonstrated by behavioral
research on rodents. For example, the avoidance of high-intensity magnetic anomalies
suggests an ability to detect strong magnetic fields, while orientation in a given direction
without access to visual, olfactory, and acoustic cues suggests the detection of earth-
strength fields and their use for compass information. The first clear evidence for a magnetic
sense of direction in rodents is provided by displacement experiments in which manip-
ulation of the ambient magnetic field en route to the test site subsequently influences
"homeward" orientation, implicating a magnetic compass as a navigational aid. A prelim-
inary investigation of the orientation of juvenile rodents suggests that this ability to use
a magnetic compass may not prevail until a particular stage in ontogeny.
Attempts to locate the magnetoreceptor in rodents have included searching for mag-
netic deposits in tissue and also searching for structures in the CNS which may be involved
in the perception and/or transmission of magnetic stimuli. Magnetometric investigations
have found elevated levels of IRM in the anterior/dorsal region of the rodent head. Further
examination of the head indicates the presence of material with a coercivity spectrum
characteristic of single-domain magnetite, with an indication of a fair degree of interaction
between the magnetite crystals. Widespread deposits of ferric iron in the rodent head and
body have been located histologically, and so far three main categories of ferric material
can be identified: a subsurface layer present in most bones; small isolated concentrations
scattered within some areas of bone marrow; and dense material in the roots of teeth.
Transmission electron microscopy has revealed that the ethmoturbinal bones of the ol-
factory region contain groups of crystallike ferric particles which are in the vicinity of
nerves and which appear to be located along the bone-marrow interface. Neurophysiol-
ogical investigations show that the cells in the pineal organ are influenced by earth-strength
magnetic fields, and biochemical analysis suggests an effect of magnetic fields on the
secretory activity of this structure.
For rodents, as for other animals so far investigated, clear connections between the
various findings of magnetoreception ability, tissues with magnetic remanence, single-
domain magnetite, deposits of ferric iron, and structure in the CNS either associated with
ferric material or influenced by magnetic fields, remain to be established, so offering a
broad and exciting horizon for future research.
ACKNOWLEDGMENTS. My thanks are due to: Dr. J. L. Kirschvink of the Department of Geo-
logical and Planetary Sciences, California Institute of Technology, Dr. E. R. Buchler of the
Zoology Department, University of Maryland, and Dr. J. M0rup J0rgensen of the Institute
of Zoology and Zoophysiology, University of Aarhus, Denmark, for allowing me to present
unpublished data; Dr. P. Dagley and the Geophysics Department, University of Liverpool,
u.K., for the use of their magnetometer; Professor F. Oldfield of the Geography Department,
532 Chapter 25
University of Liverpool, U.K., and Dr. J. Comer of the Physics Department, University of
Manchester, U.K., for magnetometry advice; Dr. D. W. Yalden of the Zoology Department,
University of Manchester, U.K., for rodent specimens; Dr. R. R. Baker and Dr. J. H. Kennaugh
of the Zoology Department, University of Manchester, U.K., for assistance with some of
the research projects; and Dr. N. M. Stone for a valuable criticism of the manuscript. This
work was supported by SERC Grant GRlB74337
References
Aneshansley, D. ]., and Larkin, T. S., 1981, V-test is not a statistical test of 'homeward' direction,
Nature 293:239.
Baker, R R, and Mather, ]. G., 1982, A comparative approach to bird navigation: Implications of
parallel studies on mammals, in: Avian Navigation (F. Papi and H. G. Wallraff, eds.), Springer,
Baker, R R, Mather,]. G., and Kennaugh,]. H., 1982, The human compass?, EOS, Trans. Am. Geophys.
Union 63:156a.
Baker, R R, Mather,]. G., and Kennaugh,]. H., 1983, Magnetic bones in human sinuses, Nature 301:78-
80.
Barnothy, M. F. (ed.), 1964, Biological Effects of Magnetic Fields, Volume I, Plenum Press, New York.
Barnothy, M. F. (ed.), 1969, Biological Effects of Magnetic Fields, Volume II, Plenum Press, New York.
Bassett, C. A 1., Pawluk, R ]., and Pilla, A A, 1974, Augmentation of bone repair by inductively
coupled electromagnetic fields, Science 184:575-577.
Batschelet, K, 1981, Circular Statistics in Biology, Academic Press, New York.
Bovet, ]., 1960, Experimentelle Untersuchungen tiber das Heimfindevermogen von Miiusen, Z. Tierp-
sychol. 17:728-755.
Bovet, ]., 1965, Ein Versuch, wilde Miiuse unter Ausschluss optischer, akustischer und osmischer
Merkmale auf Himmelsrichtungen zu dressieren, Z. Tierpsychol. 22:839-859.
Bovet, ]., 1978, Homing in wild myomorph rodents: Current problems, in : Animal Migration, Nav-
igation, and Homing, (K. Schmidt-Koenig and W. T. Keeton, eds.), Springer, Berlin, pp. 405-412.
Buchler, K R, and Wasilewski, P.]. 1982, Bats have magnets, Eos, Trans. Am. Geophys. Union 63:156a.
Chafetz, M. D., 1982, Geomagnetic orienting in the radial eight-arm maze?, J. Gen. Psychol. 107:287-
295.
Corbet, C. B., and Southern, H. N. (eds.), 1977, The Handbook of British Mammals, Blackwell, Oxford.
Etienne, AS., 1980, The orientation of the golden hamster to its nest site after the elimination of
various sensory cues, Experientia 36:1048-1050.
Fluharty, S. 1., Taylor, D. H., and Barrett, G. W., 1976, Sun compass orientation in the meadow vole,
Microtus pennsylvanicus, J. Mammal. 57:1-9.
Science 203:1355-1356.
Gould, ]. 1., Kirschvink, ]. 1., and Deffeyes, K. S., 1978, Bees have magnetic remanence, Science
201:1026-1028.
Granick, S., 1946, Ferritin: Its properties and significance for iron metabolism, Chern. Rev. 38:379-
403.
Haigh, G. R, 1979, Sun-compass orientation in the thirteen-lined ground squirrel, Spermophilus tri-
decemlineatus, J. Mammal. 60:629-632.
Hutchinson, H. K, 1953, The significance of stainable iron in sternal marrow sections, Blood 8:236-
248.
Ketchen, K K, Porter, W. K, and Bolton, N. K, 1978, The biological effects of magnetic fields on man,
Am. Ind. Hyg. Assoc. J. 39:1-11.
Kirschvink, ]. 1., 1983, Biogenic ferrimagnetism: A new biomagnetism, in: Biomagnetism: An Inter-
Kirschvink, ]. L., and Gould, ]. L., 1981, Biogenic magnetite as a basis for magnetic field detection in
Kuterbach, D. A, Walcott, B., Reeder, R ]., and Frankel, R B., 1982, Iron-containing cells in the honey
Rodents 533
Lindenlaub, K, 1955, tIber das Heimfindevermogen von Siiugetieren. II. Versuche an Miiusen, Z.
Tierpsychol. 12:452-458.
Lindenlaub, K, 1960, Neue Befunde iiber die Anfangsorientierung von Miiusen, Z. Tierpsychol.
17:555-578.
Liiters, W., and Birukow, G., 1963, Sonnenkompassorientierung der Brandmaus (Apodemus agrarius
Pall.), Naturwissenschaften 50:737-738.
Marsden, W., 1964, The Lemming Year, Chatto & Windus, London.
Mather, J. G., 1981a, Wheel activity, exploration and goal orientation by small mammals, Ph.D. thesis,
University of Manchester, U.K.
Mather, J. G., 1981b, Wheel running activity: A new interpretation, Mammal Rev. 11:41-51.
Mather, J. G., and Baker, R R, 1980, A demonstration of navigation by small rodents using an ori-
entation cage, Nature 284:259-262.
Mather, J. G., Baker, R R, and Kennaugh, J. H., 1982, Magnetic field detection by small mammals,
Eos, Trans. Am. Geophys. Union 63:156a.
Mittelstaedt, M. 1., and Mittelstaedt, H., 1980, Homing by path integration in a mammal, Naturwis-
senschaften 67:566-567.
Reith, K J., 1961, The ultrastructure of ameloblasts during matrix formation and the maturation of
enamel, J. Biophys. Biochem. Cytol. 9:825-839.
Russell, D. R, and Hedrick, H. G., 1969, Preferences of mice to consume food and water in an envi-
ronment of high magnetic field, in: Biological Effects of Magnetic Fields, Volume II (M. F. Bar-
nothy, ed.), Plenum Press, New York, pp. 233-239.
Semm, P., Schneider, T., and Vollrath, 1.,1980, Effects of an earth-strength magnetic field on electrical
activity of pineal cells, Nature 288:607-608.
Semm, P., Schneider, T., Vollrath, L., and Wiltschko, W., 1982, Magnetic sensitive pineal cells in
pigeons, in: Avian Navigation (F. Papi and H. G. Wallraff, eds.), Springer, Berlin, pp. 329-337.
Siegel, S., 1956, Nonparametric Statistics for the Behavioral Sciences, McGraw-Hill, New York.
Stutz, A. M., 1971, Effects of weak magnetic fields on gerbil spontaneous activity, Ann. N.Y. Acad.
Sci. 188:312-324.
Stutz, A. M., 1972, Diurnal rhythms of spontaneous activity in the Mongolian gerbil, Physiol. Zool.
45:325-334.
igation (F. Papi and H. G. Wallraff, eds.), Springer, Berlin, pp. 338-343.
Walcott, C., Gould, J. 1., and Kirschvink, J. 1., 1979, Pigeons have magnets, Science 205:1027-1028.
Walker, M. M., and Dizon, A. K, 1981, Identification of magnetite in tuna, Eos, Trans. Am. Geophys.
Union 62:850a.
Welker, H. A., Semm, P., and Vollrath, 1., 1982, Effects of an artificial magnetic field on the secretory
activity of the rat pineal gland, Symp. Ger. Endocrinol. Soc. Saltzburk, p. 122.
Wiltschko, W., and Wiltschko, R, 1972, Magnetic compass of European robins, Science 176:62-64.
Yorke, K D., 1979, a possible magnetic transducer in birds, J. Theor. Biol. 77:101-105.
Zoeger, J., Dunn, J. R, and Fuller, M., 1981, Magnetic material in the head of a common Pacific dolphin,
Science 213:892-894.
Human Magnetoreception
v
Few areas of research have engendered as much controversy or heated debate as has the
question of whether or not humans possess a magnetic sensitivity comparable to that of
other organisms. For obvious reasons, this question has attracted a great deal of media
attention which at times seems to have interfered with the analysis of the data at hand.
In the ideal case, it is clear that tests of human magnetoreception should be performed
and interpreted with the same stringent standards as are used with any other species (in-
cluding, for example, double- or triple-blind procedures and an agreement on test pro-
cedures; see Dayton, this section). At the same time, we recognize that the study of mag-
netoreception in the context of the magnetite hypothesis is as yet a new field, and only
now are criteria developed elsewhere for assessing reliability of data or laboratory results
being applied.
During the review process for Dr. Robin Baker's chapter on magnetoreception in this
volume, we discovered a great deal of confusion concerning the protocol which had been
used in some of his experiments. It also became apparent that other investigators had
attempted to replicate his experiments or variants of them without success, but few of
them had been discussed in the literature (with the notable exception of Gould and Able,
1981). In the long run, reproducibility is by far the single most important factor which
must be judged in the evaluation of any experiment; this is what happened to Yeagley's
work during the 1940s on homing pigeons when the role of the sun compass was not
understood and the magnetic experiments were not run on cloudy days. There is an un-
fortunate tendency in many fields of science to report only those experiments which yield
"positive" results, and to refer only in passing to things which do not work as expected.
For these reasons, we invited the most vocal of Dr. Baker's critics to participate in an
expanded discussion and reply section which focused on the experimental attempts to
repeat Baker's human magnetic experiments elsewhere. In this manner, both sides of the
issue could be presented for the first time in the same publication, and the readers would
have the opportunity to decide for themselves the relative merits of both camps. The editors
feel it is particularly important to take this approach when the work is published in book
form, as there are rarely, if ever, subsequent issues in which a normal discussion and reply
forum could exist.
This section begins with the article on primate magnetoreception by Dr. Baker (which,
like every other chapter in this volume, was initially criticized by at least two anonymous
referees and revised by the author), followed by several somewhat shorter critiques and
discussions of attempts to replicate these or similar experiments in the United States. These
are in turn discussed in a reply by Dr. Baker. Only in the case of the Adler and Pelkie
paper did we feel that it was necessary to allow a further comment, and this was done
because it was subsequently pointed out that in the Cornell bus experiments the subjects
had removed their blindfolds after each stop on the trip, something which was not done
in other U.S. experiments in which Dr. Baker participated. Without such information, the
Monte-Carlo simulation described by Adler and Pelkie does not make much sense.
535
Chapter 26
Magnetoreception by Man and Other
Primates
R. ROBIN BAKER
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537
2. Physiological Responses to Changes in the Ambient Magnetic Field . . . . . . . . . . . . . . 538
2.1. Strong Fields. . . . . . . . . . .. .................... . .... . 538
2.2. Weak Fields. . . . . . . . . . . . . . . . . . . . . . . .... . 538
2.3. Magnetic Storms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
2.4. Electromagnetic Fields Applied to Bones . . . . . . . . . . . . . . . . . 539
3. Magnetoreception . . . . . . . . . . . . . . . 539
3.1. The Dowsing Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
3.2. Compass Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 540
3.3. Goal Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545
4. Magnetoreceptors? .. . 550
4.1. Location ..... . 550
4.2. Structure? .. . 552
4.3. Physiology ..... . 554
5. Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 556
6. Summary . 559
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 559
1. Introduction
Magnetoreception may be said to have occurred when a specific change in the ambient
magnetic field, or in the orientation of an organism relative to the ambient magnetic field,
is converted within the organism into a characteristic pattern of nerve impulses. In most
cases we should expect that these nerve impulses would reach the central nervous system.
We might also expect that, upon integration with other information, there may be some
form of behavioral or other physiological response that could then be measured. The site
at which the ambient magnetic field is converted into nerve impulses may be termed a
magnetoreceptor.
The physiology of humans and other primates has long been known to be sensitive
to changes in the ambient magnetic field. Most of the known responses are observed in
very strong magnetic fields and seem to be disruptive in nature. Further adverse physio-
logical responses have been attributed to small changes in the geomagnetic field due to
magnetic storms. However, any disruptive influence of artificial magnetic fields or mag-
netic storms need only indicate interference with normal physiological processes and need
not necessarily indicate magnetoreception.
R. ROBIN BAKER • Department of Zoology, University of Manchester, Manchester M13 9PL, United
Kingdom.
537
538 Chapter 26
The study of primate magnetoreception has so far been restricted to humans and has
always been contentious. In large part this is because until recently the only behavioral
response to be studied was the enigmatic "dowsing" reflex. Increasingly convincing ev-
idence that a wide range of organisms, including other mammals, could detect and use
the earth's magnetic field for orientation and navigation, suggested new ways to investigate
the possibility of human magnetoreception. Experimental techniques similar to those used
on a wide range of animals have been developed to study the possible involvement of
magnetoreception in compass and goal orientation by humans. The results remain con-
tentious but can at least now be viewed in the perspective of similar results for other
animals.
This review is concerned primarily with the evidence for magnetoreception by humans
and with the possibility that such an ability derives in some way from magnetoreceptors
based on biogenic deposits of magnetic material. The review begins, however, with a brief
summary of other physiological responses by primates to changes in the ambient magnetic
field. The picture that is emerging for other animals is that biogenic magnetic deposits are
perhaps more widespread than necessary solely for magnetoreception. It is possible, there-
fore, that many such deposits could be involved in some of these more general physiological
responses.
2. Physiological Responses to Changes in the Ambient Magnetic

Field
2.1. Strong Fields
In their everyday working lives, many humans are exposed to strong magnetic fields.
Evidence that such fields influence primate physiology has been reviewed by Barnothy
(1964) and Ketchen et 01. (1978).
Magnetic fields thousands of times stronger than the geomagnetic field alter the heart
action of squirrel monkeys, Saimiri sciureus (Beischer and Knepton, 1944; Beischer, 1969),
and influence the orientation of sickled, but not normal deoxygenated, erythrocytes in
humans (Murayama, 1965). When an alternating magnetic field of 10-90 Hz and 300 G or
greater is applied near the eyes of a human subject, a shimmering luminosity appears at
the borders of the visual field. This is known as the "magnetic phosphene effect" (Alex-
ander, 1962) and seems to be limited to alternating magnetic fields (Drinker and Thompson,
1921). However, stationary magnetic fields of 200 to 1000 G do alter the brain waves of
monkeys and humans (Kholodov et 01., 1969).
None of these influences of strong magnetic fields are necessarily adverse in the long
term and there is no direct evidence for primates that such adverse effects occur. Never-
theless, Soviet workers engaged in fabrication of permanent magnets have reported such
symptoms as irritability, fatigue, occasional dizziness, altered appetite, headache, changes
in heart action, decrease in arterial blood pressure, and a numbness of, and marbling pattern
on, their hands (Beischer and Reno, 1971).
2.2. Weak Fields
Beischer (1971) exposed pairs of men for up to 2 weeks in a reduced, virtually null,
magnetic field within a large Helmholtz-coil system. A variety of standard physiological
and psychological tests were carried out, including tests on visual fields and acuity, time
estimation, and equilibrium. No influence of the reduced field was observed.
Magnetoreception by Man and Other Primates 539
2.3. Magnetic Storms
Small but irregular fluctuations in the geomagnetic field occur as protons and electrons
brush past the earth's atmosphere. These charged particles originate in the outer layer of
the sun and are carried across space by the "solar wind." The more extreme fluctuations
in geomagnetic field generated in this way are known as "magnetic storms." Even the most
extreme magnetic storm, however, causes no more than about a 5% perturbation in the
geomagnetic field, which in midtemperate latitudes has an intensity of about 50,000 nT.
Despite the small changes in intensity that are involved, magnetic storms have been
claimed to have an adverse affect on human behavior and physiology. In a recent review,
Srivastava and Saxena (1980) supported claims that the incidence of (1) hospital admis-
sions for myocardial infarction, (2) road accidents, and (3) air accidents due to pilot error,
all correlate in some way with the level of magnetic storm activity. These and other possible
correlates of magnetic storms are discussed elsewhere in this volume.
2.4. Electromagnetic Fields Applied to Bones
Most studies of human physiology and behavior in relation to changes in the ambient
magnetic field have been concerned with the possibility of adverse effects. However, puls-
ing electromagnetic fields of low frequency and strength, inductively coupled across skin
to fractured bone in dogs, were found to enhance the organization and strength of the repair
process. The method has since been applied successfully in the treatment of humans (Bas-
sett et aI., 1974).
3. Magnetoreception
Of all the reactions described in the previous section, only the magnetic phosphene
effect clearly qualifies as magnetoreception, though changes in brain activity could indicate
that magnetoreception has occurred. In both cases, however, the reaction only occurred
at field intensities more than 100 times greater than the geomagnetic field.
The possibility that primates might be able to detect earth-strength magnetic fields
has been studied only for humans and only in the context of three types of behavior. These
are (1) the dowsing response, (2) compass orientation, and (3) goal orientation. In this
section, I review briefly evidence that magnetoreception is involved in these three elements
of human behavior.
3.1. The Dowsing Response
Traditionally, a dowser is a person who searches for hidden water or minerals with
a forked twig or other divining rod which dips suddenly when held over the correct spot.
The divining rod can be made of any material and is generally considered to serve merely
as a mechanical amplifier which provides a visual signal of minute muscular contractions
in the arms (Bird, 1979).
Rocard (1964) has shown that the traditional function of the dowsing response is
unreliable, there being no correlation between the response and the presence of under-
ground water. On the other hand, both Rocard (1964) and Harvalik (1978) describe ex-
periments which are claimed to show that the dowsing reflex is a ê ú ä á ~ Ä ä ú = indicator of
magnetic gradients.
540 Chapter 26
The majority of experiments on dowsing have been inadequate to provide unequivocal

evidence for magnetoreception, usually for one of three reasons. Either (1) there is no proper
control group; (2) if there are controls, the experiment is not performed "double-blind"
(i.e., neither subject nor experimenter knowing whether the "target" is present or activated)
or often even "blind" (i.e., the subject not knowing the location or state of the target); or
(3) the sample size is too small, many experiments being carried out on just one or two
subjects. However, some of the experiments described by Rocard (1964) and Harvalik
(1978), in which the dowsing reflex was used to test for the presence of magnetic gradients,
do seem to be free of these or other obvious faults in experimental protocol.
In one experimental series carried out by Harvalik (1978), 14 subjects took part in 694
trials in which each walked through an electromagnetic "beam" in the frequency ranges
1 Hz to 1 MHz as produced by a low-power, high-frequency generator. Double-blind pro-
tocol was achieved by means of a mechanical "randomizer" to switch the beam on and
off. The dowsing response was used to measure whether the subjects could detect the
presence or absence of the magnetic anomaly. Of the total 694 trials, 661 gave a correct
response as compared to 33 incorrect responses.
In another series, 300 randomly selected subjects walked through an artificial magnetic
field produced by connecting an electric power supply (either ac or dc) to two electrodes
implanted in the ground 20 m apart. Eighty percent showed a dowsing response when the
current was activated, none when the current was inactivated (Harvalik, 1978).
Rocard (1964) describes experiments in support of his claim that some subjects may
be able to detect a rate of change in magnetic field intensity as low as 0.3 mOe/sec as they
walk through a magnetic anomaly.
3.2. Compass Orientation
Over the past century, many people have been convinced that humans have a separate
sense devoted to geographical direction based on a direct awareness of the geomagnetic
field (Viguier, 1882; Hudson, 1922; Lucannas, 1924; reviewed by Jaccard, 1931; Howard
and Templeton, 1966). Serious attempts to obtain evidence, however, have been few (re-
viewed by Baker, 1981) and in recent years, students of human navigation (e.g., Gatty,
1958; Lewis, 1972) have inclined to the view that humans do not have such a sense.
3.2.1. "Chair" Experiments

The simplest test for a magnetic compass sense is to sit someone on a rotatable chair,
deprive them of all useful cues except the ambient magnetic field, turn the chair, and ask
them to estimate the compass direction in which they are facing each time the chair stops.
If an ability to judge direction under such conditions is shown, some means of changing
the magnetic field then checks whether the subject is using magnetoreception or some
other, nonmagnetic, mechanism.
Since December 1980, over 800 different subjects have taken part in such "chair"
experiments carried out by myself and my students from Manchester University. The equip-
ment used is illustrated in Fig. 1.
The chair is made of wood with brass and aluminum fittings. The subject is double-
blindfolded (an inner, heavy-cotton, blindfold and an outer pair of swimming goggles lined
with plasticine) and fitted with ear muffs (Fig. 1). Each estimate is preceded by the ex-
perimenter rotating the chair clockwise (270°-540°) and then anticlockwise (angle ú =360°).
Each test involves nine (or eight, pre-1983) estimates during which time the experimenter
does not speak.
Figure 1. Experimental setup for chair experiments. The TRS 80 computer which runs the experiment
is normally behind the subject but has been placed to the side for illustration.
542 Chapter 26
In all these, subjects knew that the magnetic field through their head might be artificial;
they wore either electromagnetic helmets or a metal bar or bars. Helmets could be acti-
vated or not. Bars could be magnets or brass. Double-blind protocol was achieved by having
the conditions determined by computer and implemented by a third party in a way un-
known to either experimenter or subject.
Sequences of directions are determined by a TRS 80 "pocket" computer and printer
(Fig. 1). A random number program produces directions that are printed by the computer,
one at a time, during the experiment. The computer accepts and prints each estimate as
it is made and has the facility to compute and print out the mean vector (eO, r) of the nine
angular errors associated with the nine estimates of compass direction (where eO = mean
error; r = length of the mean vector; see Batschelet, 1981).
3.2.2. Statistics
Chair experiments generate data that can be analyzed at three different levels. For
example, suppose 10 subjects were each tested on five occasions, with nine estimates in
each test. Levell would use all 450 estimates; level 2 would use 50 mean errors (one mean
error per test); and level 3 would use 10 mean errors (one mean error per subject).
The data have many similarities with those generated by orientation cage studies on
birds (e.g., Wiltschko and Wiltschko, 1978). On the grounds that successive hops by a bird
during a single test are not independent, it has become conventional in such experiments
not to use level 1 analysis, except for special purposes. Similarly, successive estimates of
direction by humans during the course of a single test in chair experiments cannot be
independent and level 1 analysis again cannot be justified. Instead, for birds, a single mean
direction of hop during the course of a test is calculated. The means for many different
tests are then subjected to higher-order analysis. Usually, however, analysis is at level 2,
even though individual birds are tested on more than one occasion. Such analysis, though
widely accepted, is still open to criticism for a possible lack of independence of data points.
Apart from statements of probability based on nonindependent data points (see Section
3.3.2), there are other statistical abuses which, although common in the orientation lit-
erature, ought to be avoided in the analysis of data from chair experiments. These are (1)
multiple hypothesis testing (particularly the use of both Rayleigh z- and V-tests on the
same data sets; see Section 3.3.2) and (2) post-hoc analysis.
A lack of independence among data points in chair experiments can be reduced, as
in orientation cage experiments on birds, by the use of level 2 analysis or preferably avoided
by the use of level 3 analysis. Multiple hypothesis testing and post-hoc analysis can be
avoided by adopting a single, universal hypothesis for all samples and another single,
universal hypothesis for all two-sample comparisons.
In chair experiments, the null hypothesis for all samples is that there is no significant
vector in the direction eO = 0 or, if this vector is significant, that the 95% confidence
interval of the mean error does not include 0 (see Aneshansley and Larkin, 1981). Only
0
the V-test, not the z-test (Batschelet, 1981), is appropriate. In the comparison of two sam-
ples, the null hypothesis is that there is no difference in the strength of compass orientation.
The most suitable test is Wallraff's modification of the Mann-Whitney test (see Batschelet,
1981, pp. 127-128). In most cases, there is a clear prediction as to which sample should
show the better orientation and a one-tailed test is appropriate. When this is not the case,
a two-tailed test is used.
3.2.3. Results
Since 1980, 875 individuals from 11 recognizable groups have been tested in chair
experiments. The results for all tests in an unaltered magnetic field are presented in Table 1.
Table I. Summary of All Chair Experiments in Which Subjects Were Tested in an

Unchanged Magnetic Field a
Group Year(s) N eO r h p
Manchester students 1980-81 44 -8 0.173 0.171 0.054

University staff 1981 26 2 0.115 0.115 0.202
University visitors 1981 15 -21 0.307 0.287 0.058
Field course students 1981 41 -27 0.140 0.125 0.129
Bramhall residents 1981-82 17 -2 0.656 0.656 <0.001
Manchester naturists 1982 39 -7 0.552 0.548 <0.001
British orienteer squad 1982 22 64 0.148 0.065 0.334
Field course students 1982 31 3 0.162 0.162 0.100
Manchester schoolchildren 1982-83 494 -13 0.155 0.151 <0.001
Watford dyslexics 1983 122 26 0.079 0.071 0.135
Field course students 1983 24 -4 0.416 0.415 0.002
Total 875 -7 0.175 0.174 <0.0001
a N. number of individuals; eO. mean error; r. length of mean vector; h. homeward component; p. probability (V-
test).
People from a modern, industrial environment show a weak [h = 0.174; where h, the
"homeward" component (Batschelet, 1981), is the component of the mean vector in the
direction e = 0°] but significant (e = -7 ± 12°, P < 0.0001, V-test) ability to estimate
compass direction in chair experiments. With such a weak response, significant results
are unlikely to be obtained with small sample sizes (e.g., Zusne and Allen, 1981).
One way that subjects could solve the orientation problems set by chair experiments
would be to note their initial orientation upon being blindfolded and then thereafter to
follow the turns of the chair using an inertial sense (Barlow, 1964). If this were the tech-
nique used, the data should show two characteristics. First, the accuracy of the estimates
should deteriorate during the course of a single test, the last estimate being less accurate
than the first. Second, a subject who is grossly in error with his or her first estimate should
be less accurate with the second and subsequent estimates than subjects who are accurate
with their first estimate. Neither of these characteristics was found upon separate analysis
of the first to last (eighth) estimates within tests of 76 subjects from mid-1982 (Table II).
Table II. Lack of Influence of (1) Time into Test; and (2) Accuracy of First Estimate; on
Accuracy of Compass Orientation in Chair Experiments
Mean vector
No. of
individuals r V-test p
Time into test
Estimate No.1 76 5 0.243 0.0015
8 76 1 0.290 0.0002
Accuracy of
First estimate
Second estimate
following
First eO 0 ± 45 44 -27 0.253 0.018
First eO 180 ± 45 20 o 0.270 0.043
544 Chapter 26
brass
Figure 2. Influence of alignment of bar magnet on

compass orientation in a chair experiment. Each
dot is the mean error for an individual calculated
over 40 estimates (5 tests) made in each condition.
Vertical dotted line indicates the position e = 0°.
Arrows and data show the mean vector for the 10
points. Dashed lines indicate 95% confidence in-
terval of the mean error. Subjects were all N-S
sleepers (see Section 4.3). Earmuffs are omitted
from the drawing for clarity. From Baker (1984a),
N-up magnet S-up magnet simplified from Baker (1984b).
The data are more consistent with the view that nonvisual orientation in chair experiments
involves judgment of direction rather than inertial judgment of turns.
Evidence that this sense of direction involves magnetoreception comes from an ex-
periment on the compass orientation of subjects wearing bar magnets of different polarities
(Baker, 1984a,b). Ten Bramhall residents (Table I) were selected for testing on 15 further
occasions (1 week between each test) with either a bar magnet or a brass bar aligned ver-
tically on the right temple as shown in Fig. 2. Bars, which were enclosed in an opaque
and sealed cotton envelope, measured 77 x 15 x 6 mm. Magnets had a pole strength of
about 200 G. Conditions were randomly allocated by computer and implemented by a third
party. Tests continued over 4 months until each subject had been tested 5 times wearing
a brass bar and 10 times wearing a bar magnet (5 times with N pole uppermost, 5 with S
pole uppermost).
Subjects were all N-S sleepers (see Section 4.3) and were always tested between 1000
and 1500 GMT. They, and the experimenter, wore only clothes made of 100% cotton and
had not worn stereo headphones, traveled by train, or used nose drops or nasal spray during
the preceding 48 hr. The 10 subjects produced highly significant orientation while wearing
the brass bar (N = 10, e = 17 ± 37°, r = 0.678, P = 0.003, V-test).
No means were available within the test room for either subject or experimenter ac-
cidentally to discover which pole of the magnet within the envelope was uppermost nor
even whether the bar was a magnet or brass. Any concern over the latter, however, thereby
negating the double-blind design of the experiment, can be overcome by restricting com-
parison to orientation performance while wearing magnets of different polarities.
Compass orientation was significant while wearing a S-up magnet (N = 10, e = 20
± 25°, r = 0.841, P = 0.00036, V-test) but not while wearing a N-up magnet (N = 10, e
= 166 ± 32°, r = 0.756, P = 0.99, V-test). The difference in orientation performance (Fig.
2) was significant [U = 0, P < 0.002 (2-tailed), Wallraff's test]. Level 2 data from this
experiment are given in Baker (1984b).
Further evidence that compass orientation in chair experiments is based on magne-
toreception is given in Section 4.3 (Fig. 4).
3.3. Goal Orientation
Goal orientation is the ability of an animal to determine the direction of a particular

distant point in space. It is more complex than, but usually involves, compass orientation
(see discussions by Baker, 1981, 1982, 1984b).
The usual way to study goal orientation is to carry out some form of "displacement-
release" or "homing" experiment. Among primates, such experiments have so far been
performed only on humans and have become known as "bus" experiments.
3.3.1. Bus Experiments

Bus experiments involve transporting groups of subjects along tortuous routes. At the
destination( s) they are asked to indicate the direction of home, either in terms of its compass
direction or by pointing. Each estimate is then expressed as an error, eO, relative to home
direction. Perfect homeward orientation is therefore described by e = 0°. Usually, subjects
are blindfolded both throughout the outward journey and while making their estimates.
Details of the protocol of bus experiments have been given in Baker (1980,1981) and Gould
and Able (1981).
The first bus experiment seems to have been carried out by Lord (1941) who displaced
schoolchildren along a 3-km course through Ann Arbor, Michigan. However, the results
were not analyzed with a view to answering the questions that are conventional in studies
of other animals: (1) does a significant ability for goal orientation exist; and (2) if so, under
what range of conditions is it manifest? The same comment can be made on experiments
in which human subjects have walked short distances and attempted to point back to the
starting point of their journey (e.g., Worchel, 1951; Juurmaa, 1966; Garling et a1., 1979).
The only published bus experiments aimed specifically at answering these questions
are those of Baker (1980, 1981), Gould (1980), and Gould and Able (1981).
3.3.2. Statistics
Gould and Able (1981)' in their Table 1, present data from three bus experiments.
These data are subdivided into 17 subsets, each relating to a single test site. To each of
these 17 subsets, the authors apply both z- and V-tests. The analysis is then summarized
(Gould and Able, 1981, p. 1063) by the sentence: "The one instance of orientation at the
5 per cent level out of the 34 statistical tests in Table 1 is about what ought to turn up by
chance."
This example illustrates two of the statistical abuses that are commonplace in the
orientation literature: (1) the use of nonindependent data and (2) multiple hypothesis test-
ing. First, the 17 subsets are not independent samples. Within any journey, a person's
estimate of home direction at one site must build upon, and be influenced by, estimates
of home direction at previous sites. The 17 subsets, therefore, derive from only three in-
dependent data sets. Second, z- and V-tests address different null hypotheses. The z-test
is appropriate when a unimodal departure from uniformity in any direction is predicted;
the V-test when the mean vector is expected to have a significant component in a particular
direction (Batschelet, 1981). Published probabilities for these tests assume that only one
546 Chapter 26
of the two tests has been applied. Probability levels change if both are used. On independent
data sets, application of both tests illegitimately improves the change of obtaining a "sig-
nificant" result; on nonindependent data sets the effect is problematic.
When bus experiments involve several test sites during the journey, they generate data
that can be analyzed at three levels. Suppose that 50 people are displaced, 10 on each of
five journeys, and estimate home direction at six sites on each journey. Levell analysis
would use all 300 estimates; level 2 would use 50 estimates (one mean error per person
per journey); level 3 would use 5 estimates (one mean error per journey, each calculated
from the 10 individual means). As we have already seen, however, level 1 analysis uses
nonindependent data and cannot be justified.
The danger of multiple hypothesis testing can be avoided for bus experiments in the
same way as for chair experiments (Section 3.2.2). In all tests, there is a clear prediction
that orientation should be toward home (Le., the mean error, e, of estimates of home di-
rection should be 0°) and that treatments should influence the strength of homeward ori-
entation. Only the V-test (against e = 0°) for single samples and Wallraff's nonparametric
two-sample test for the strength of homeward orientation would seem to be appropriate
unless there is a clear alternative prediction before the experiment is carried out.
3.3.3. Results
Between October 1976 and November 1982, 33 bus experiments have been carried out
by myself or my students from Manchester University. Details of the protocol for 1976-
1979 were given in Baker (1981). The electromagnetic helmets described therein continued
to be used to alter the magnetic field through the head in 1980, but in 1981 and 1982 we
reverted to the use of bar magnets and brass bars. In all 33 journeys, at least some of the
subjects experienced an unaltered magnetic field. All results for these subjects are pre-
sented in Table III. Whether analyzed at level 2 or level 3, the picture is one of a weak
but significant ability for goal orientation following blindfolded displacement.
Published accounts of attempts to replicate the Manchester bus experiments all refer
to journeys in the United States. Four experiments were carried out by Gould at Princeton
in 1980 (Gould, 1980; Gould and Able, 1981) and three more were carried out at Princeton
in 1981 by Gould in collaboration with myself and Janice Mather (Gould and Able, 1981).
Two experiments were performed by Able at Albany in 1980 (Gould and Able, 1981) and,
finally, one experiment was carried out by Adler at Cornell in 1981, again in collaboration
with myself and Mather. Although data from this last experiment have never formally been
published, they have entered the public domain through being used in discussion by media
such as The Los Angeles Times.
Unfortunately, evaluation of Gould's original four experiments at Princeton is hind-
ered by inconsistencies in the published data. The first experiment (May 2, 1980) is par-
ticularly important, being the only one of the series in which some subjects wore magnets,
and Gould has published the results on two occasions (Gould, 1980; site 1 in Fig. 1, Gould
and Able, 1981). However, for both controls and experimentals, the published results are
different in the two sources. For the controls, an estimate of "W" in Gould (1980) has
become "ESE" in Gould and Able (1981), thus reducing the vector length from 0.371 to
0.334. To confuse matters further, the mean vector that accompanies Fig. 1 in Gould and
Able (1981) is appropriate to the data in Gould (1980). For the experimentals, Gould (1980)
originally showed the mean vector to be to the left of home but later states (in Gould and
Able, 1981, p. 1061) that the mean vector was 12° to the right of the homeward bearing.
In the same sentence, Gould also states that the mean vector length of 0.27 is not statistically
significant. This is misleading, for with N = 20, a mean vector of e = 12° and r = 0.27
is significant (P = 0.047, V-test).
Table III. Summary of Estimates by Subjects with an Unaltered Magnetic Field through
Their Head in all Manchester Bus Experiments
Compass estimates Pointing estimates
Date Stops N r p (V-test) N r p (V-test)

10.11.76 3 8 60 0.366 0.228
10.18.76 3 8 0.551 0.015
10.20.76 3 6 29 0.586 0.038 - 29 0.367 0.128
11.01. 76 3 7 -8 0.867 0.001 - 28 0.909 0.002
11.17.76 1 5 21 0.571 0.045 128 0.639 0.898
11.22.76 1 11 44 0.609 0.020 55 0.188 0.304
12.06.76 1 8 3 0.919 0.0002 -60 0.376 0.222
10.24.77 1 6 40 0.739 0.026 78 0.483 0.360
10.31.77 1 6 -1 0.412 0.074 151 0.518 0.943
12.04.78 1 11 71 0.491 0.223 86 0.398 0.447
06.29.79 1 42 4 0.380 0.0003
06.29.79 2 16 -20 0.608 0.0008
10.15.79 3 16 30 0.553 0.004 - 51 0.514 0.034
10.22.79 5 12 8 0.248 0.113 - 38 0.433 0.047
10.29.79 6 15 5 0.557 0.002 -86 0.688 0.395
11.05.79 8 17 69 0.221 0.320 141 0.375 0.955
10.03.80 2 12 -79 0.196 0.426 39 0.296 0.128
10.13.80 5 15 -7 0.407 0.014 - 29 0.366 0.040
10.20.80 5 15 -98 0.299 0.591 o 0.396 0.016
10.27.80 5 16 136 0.235 0.833 -47 0.482 0.032
11.10.80 1 9 -93 0.345 0.531 38 0.553 0.031
11.10.80 9 9 -86 0.133 0.484 - 5 0.428 0.036
11.17.80 1 10 18 0.694 0.002 164 0.130 0.715
11.17.80 9 10 61 0.141 0.378 12 0.823 0.003
11.24.80 1 9 13 0.419 0.042 104 0.539 0.714
11.24.80 9 9 86 0.340 0.459 4 0.718 0.002
10.19.81 6 25 - 23 0.091 0.276 33 0.231 0.085
10.26.81 6 22 41 0.327 0.051 14 0.340 0.015
11.02.81 6 25 32 0.192 0.124 69 0.423 0.141
11.09.81 6 23 105 0.161 0.612 -38 0.459 0.007
10.25.82 6 9 81 0.426 0.386 -13 0.596 0.008
11.01.82 6 9 -125 0.267 0.746 23 0.302 0.116
11.08.82 6 9 109 0.187 0.604 - 36 0.463 0.055
Total
Level 2 414 18 0.256 <0.00001 372 2 0.245 <0.00001
Level 3 31 21 0.558 0.00003 31 13 0.509 0.00006
It is not only site 1 in Fig. 1 of Gould and Able (1981) that has a written mean vector
which is incompatible with the accompanying illustrated data. The same is true for sites
3, 4, 6, 12, and 13. Gould kindly sent me a copy of the raw data for these experiments so
that I could check his calculations. It appears that for all except sites 1, 3, and 13, the
incompatibility is due simply to incorrectly placed dots in the illustrations. The confusion
over site 1 persists, however, for neither of the two published accounts (Le., Gould, 1980;
Gould and Able, 1981) agrees with the raw data. These instead show a mean vector for
controls of e = - 86°, r = 0.425 and for experimentals wearing magnets of e = -10°, r
= 0.335. Sample size is 20 in both cases. This first experiment, therefore, gave a mean
548 Chapter 26
Table IV. Estimates of the Compass Direction of "Home" in Ten American Bus
Experiments by Subjects in Unaltered Magnetic Field
Location Date Stops N eO r p (V-test)

Princeton (1) 5.02.80 1 20 -86 0.425 0.428
(J.L.G.)a 10.30.80 6 15 -147 0.316 0.927
10.31.80 6 8 77 0.522 0.313
11.24.80 5 11 -68 0.168 0.381
Total 54 -101 0.165 0.631
Princeton (2) 2.16.81 4 20 -29 0.155 0.193
(J.L.G., RRB., 2.17.81 4-5 19 7 0.193 0.118
J.G.M.) 2.18.81 4 12 28 0.199 0.191
Total 51 0 0.166 0.047
Albany 1 10 77 0.140 0.443
(K.P.A.) 1 9 79 0.170 0.444
Total 19 78 0.154 0.421
Cornell 2.20.81 4 30 20 0.541 0.00006
(K.A., R.RB.,
J.G.M.)
Total: level 2 154 -1 ± 50 0.147 0.005
Total: level 3 10 13 0.404 0.039
a Experimenters: j. L. Gould; R. R. Baker; j. G. Mather; K. P. Able; K. Adler.
vector for controls that would be significant by the z-test used by Gould and for experi-
mentals that is significant by the V-test (P = 0.019).
In view of the confusion over the published data for Gould's 1980 Princeton experi-
ments, I have used the raw data for these four journeys in preparing Table IV which sum-
marizes the 10 journeys carried out on American soil. Whether analyzed at level 2 or level
3, the results show, as for the Manchester data, a weak but significant ability for goal
orientation following blindfolded displacement.
Although the mean error for a journey is usually less than 90 (Tables III and IV), the
0
homeward component varies markedly from one journey to the next. Goal orientation by
pigeons (Keeton et aI., 1974; Larkin and Keeton, 1976) and compass orientation by gulls
(Southern, 1971) and salamanders (Phillips and Adler, 1978) all vary in accuracy from day
to day in relation to magnetic storm activity. In pigeons, the best correlation is with respect
to K 12 , the sum of the K indices of magnetic storm activity for the four 3-hr periods before
the end of the experiment. We have reported (Baker and Mather, 1982; Baker et a1., 1982)
that the ability of humans to describe the compass direction of home also correlates with
magnetic storm activity, the best correlation being with K6 , the sum of the K indices for
the two 3-hr periods before the end of the experiment. The data for 1976-1982 have been
illustrated in Baker (1984b) and are summarized in Table V and Fig. 3. Although the rank-
order correlation between K6 and ability to estimate the compass direction of home is
significant (Table V), there is an indication (Fig. 3) that peak homeward orientation may
occur at a value for K6 of 2 rather than under the very quietest magnetic conditions. The
ability to point toward home shows no correlation with magnetic storm activity (Table V).
The first indication that goal orientation during bus experiments may be based on
magnetoreception came from an experiment at Barnard Castle, County Durham, in which
subjects wearing bar magnets on the back of the head failed to produce significant home-
ward orientation while controls wearing magnetically inert bars produced good homeward
orientation (Baker, 1980). At the first of the two sites, the difference in strength of home-
ward orientation was significant [N 1 •2 = 15,16, U = 69, P < 0.025 (I-tailed), Wallraff's
two-sample test for the strength of homeward orientation].
Table V. Spearman's Correlation Coefficients for the

Relationship between Magnetic Storm Activity (K6) and
Homeward Component in Goal Orientation Experiments
(See Fig. 3)
N (for K6 = 1-8) rs (2-tailed)
Compass estimates
Controls" 8 -0.905 <0.02
Experimentals 8 - 0.310 ns
Pointing estimates
Controls 8 0.452 ns
Experimentals 8 -0.310 ns
Q Controls, normal magnetic field through head; experimentals, artificial mag-
netic field through head.
The Barnard Castle experiment was followed by a series in 1979 (Baker, 1981) and
1980 in which bars were replaced by electromagnetic helmets as a means of inducing an
artificial magnetic field through the head. Details of these helmets are given in Baker (1981).
Controls wore deactivated helmets.
In pigeons, the homeward orientation of birds exposed to an artificial magnetic field
shows no correlation with magnetic storm activity (Larkin and Keeton, 1976). The same
is so far true for humans (Table V). The electromagnetic helmets failed, however, to rep-
licate the reduction in homeward orientation that had been indicated in the Barnard Castle
experiment. Indeed, by the end of 1980, it was clear that if the helmets were having any
effect, it was to improve homeward orientation (Table VI), a trend that had begun to emerge
even after the first four experiments (Baker, 1981, pp. 56-57).
As the 1979-80 bus experiments had been designed to test the hypothesis that an
artificial magnetic field would reduce homeward orientation, the data could not then be
analysed post hoc to test for an improvement in homeward orientation. The news that a
strong magnetic field may lead to an improvement in homeward orientation by homing
pigeons (Gould and Walcott, in Walcott, 1982) led in late 1981 to bus experiments designed
to test the possibility that applied magnetic fields may also improve homeward orientation
by humans. These experiments are continuing (Baker, 1984a; 1985).
p
0.8
• <0.00001
Figure 3. Relationship between magnetic
•
storm activity and ability to estimate the 0.6 38 • <0.02
compass direction of home during goal ori- • ns
• •ú Q P =
entation experiments on humans. Mag- h 0.4 84/ "'77
netic storm activity is given by K6 ; goal ori-
entation by the homeward component (h;
see Batschelet, 1981) of compass estimates 0.2 63
• 25
of home direction. Numbers show total in- ú Qó ’ J K =
dividuals that have taken part in experi-
ments at each K value. One mean error per
person per K value was calculated before
0.0
.
_0.2 L--_L------''-----'_---'_---'_---'_---'-_---'
2 3 4 5 6 7 B
.
J J J J J J J J J J J J J J J ú J J J J J J J y V=
calculating h. p values show significance

of each h (V-test). magnetic storm activity (K 6 I
550 Chapter 26
Table VI. Summary of Compass Estimates of Home Direction during Bus Experiments
in Which Some Subjects Were Exposed to an Artificial Magnetic Field through the
Head during Displacement
Normal field Artificial field
No. of
journeys N eO h p (V-test) N eO h p (V-test)
Manchester 8 118 10 0.198 0.0011 117 -4 0.215 0.0005

1979-80
Princeton 2 32 -70 0.083 0.2498 34 -5 0.257 0.0172
1980-81
Manchester 7 77 114 -0.053 0.7437 98 18 0.160 0.0126
1981-82
In 1981, four journeys were carried out along similar routes to those described in Baker
(1981). There were six test sites along each route. The blindfolded subjects wore nothing
on their heads until after site 3 when magnets (S pole uppermost; dimensions and pole
strength as described in Section 3.2.3) or inert bars were placed on the right temple (as in
Fig. 2) for the remainder of the journey. In 1982, three similar journeys were made but the
magnets (some N-up, some S-up) or brass bars were worn throughout the experiment. At
the completion of these two series, homeward orientation by subjects wearing magnets
was significantly beUer than that by controls (Table VI).
Despite the widespread interest in whether the American bus experiments did or did
not produce significant homeward orientation, it seems to have escaped general notice
that over the two Princeton experiments in which magnets were used, subjects wearing
magnets produced clearly significant results (N = 34, e = - 5 ± 70°, r = 0.258, P =
0.017, V-test) (see also Table VI). In part, this oversight may have been aided by Gould's
misleading statement, described earlier in this section, that the mean vector length for
experimentals was not significant, when in fact it was. As both of the Princeton experiments
involving magnets had been designed to test for a reduction in homeward orientation
relative to controls, we cannot now retrospectively claim support for the opposite hy-
pothesis. Nevertheless, it is gratifying to observe that had Gould's first experiment been
testing our current hypothesis, he would have been able to report a significant difference
between controls and experiments [N1 •2 = 20,20, U = 134, P < 0.05 (1-tailed), Wallraff's
two-sample test for the strength of homeward orientation).
4. ú ~ Öå É í ç ê É Å É é í ç ê ë \ =
4.1. Location
All three fields of study into the possibility of magnetoreception by humans have
sought an indication through experiment of the location of the magnetoreceptor(s) through
which the behavioral response is mediated.
Using the dowsing response and the high-frequency electromagnetic beam described
in Section 3.1, Harvalik (1978) screened progressively smaller parts of the body with metal
sheets that would cut off the beam. The experiments implicated magnetoreceptors in the
vicinity of the adrenal glands and another in the head. Progressive shielding of parts of
the head also implied an unpaired receptor on the midline between but slightly above the
eyes and about 13 mm in front of the ear orifice. Support for the importance of the adrenals
is claimed from the medical histories of dowsers subjected to surgery of the kidney region.
Only removal of the adrenal glands is claimed to block the dowsing response.
Q
) î ú=
Figure 4. Influence of the position of a bar magnet on

the ability to estimate compass direction when the
magnet is removed. Subjects were tested wearing two
brass bars in the positions shown (earmuffs not shown
for clarity). Previously. up to about 3 min before test-
ing. subjects had worn two other bars. one of which
(black in diagram) was a bar magnet (200 G; N pole
uppermost). the other (white in diagram) being brass.
All bars were in opaque cotton envelopes and protocol
was double-blind. Other conventions as in Fig. 2. Sim-
plified from Baker (1984b).
Chair experiments have suggested that nonvisual compass orientation might involve
a magnetoreceptor in the front of the head (Baker, 1984b). Figure 4 shows the results of
an experiment in which blindfolded subjects first wore two bars, one of which was brass,
the other a N-up bar magnet as used in other chair experiments (Section 3.2). Two zones
on the head were used (Fig. 4), broadly in front of the ear or behind. After 10 min, the
blindfold and bars were removed. Two minutes later the subject was tested, again blind-
folded and wearing two bars, but now both were brass. Experiments were double-blind.
The bars were placed in opaque, sealed, and labeled cotton envelopes by a third party and
their location on the head determined by the TRS 80 computer shown in Fig. 1. Tests took
place in a hut made entirely of wood, with brass fittings, on the edge of woodland in rural
Gloucestershire. Experimenters and subjects removed all metal objects from their persons
before entering the hut. There was no possibility of accidental identification of which bars
were magnets and which brass.
Figure 4 shows that after wearing a magnet on the front of the head. the subjects (Field
Course Students, 1982; Table I) were less able to identify compass direction than after
wearing a magnet on the back of the head [N 1 •2 = 21,20, U = 112, P = 0.011 (2-tailed),
Wallraff's two-sample test for the strength of homeward orientation].
It is known from other chair experiments (Section 3.2.3) that subjects wearing mag-
netically inert bars are able to judge compass direction. However, in the experiment in
Fig. 4, due primarily to limitations to sample size, there was no formal control group (Le.,
subjects exposed to exactly the same procedure but wearing two brass bars in both halves
of the experiment). Although it seems likely, we cannot, therefore. properly conclude from
Fig. 4 that a N-up magnet on the front of the head is more disruptive to compass orientation
than a similar magnet on the back.
Bus experiments have suggested that the ability to describe the compass direction of
displacement might also involve a magnetoreceptor in the front of the head. The electro-
magnetic helmets used in the Manchester bus experiments produce an uneven field through
the head. We have mapped this field and, following the first four journeys on which they
were used, sought a match between the direction of lines of force through different sites
in the head and observed patterns of influence of the helmets on orientation (Baker and
Bailey, in Baker, 1981). As a result, it was predicted (in Baker, 1981) that a magnetoreceptor
involved in orientation was located on the midline, between but slightly below the eyes,
552 Chapter 26
and about 3-4 cm in from the front of the head; roughly where the brain, olfactory, and
optic nerves rest on the sphenoid bone.
In summary, all three lines of investigation into human magnetoreception have im-
plicated a magnetoreceptor in the front of the head. In addition, dowsing studies suggest
paired magnetoreceptors in the adrenal glands.
Perhaps it is pure coincidence that pre-Columbian sculptures carved 3000-4000 years
ago have consistent and distinctive patterns of magnetism. Not only have pre-Columbian
sculptors in Mexico carved a turtle head with a magnetic pole located in the snout (Malms-
trom, 1976; see Perry et aI., this volume) but in Guatemala there are human figures in
which the sculptors seem consistently to have located magnetic poles in the temples of
the figures' heads and in the region of their midriff (P. A. Dunn and V. H. Malmstrom,
personal communication).
4.2. Structure?
Anatomical search for human magnetoreceptors has so far been based solely on the
hypothesis that such receptors will in some way involve biogenic deposits of magnetic
material, probably magnetite (Lowenstam, 1962; Blakemore, 1975; Kirschvink and Gould,
1981; Yorke, 1981).
The first discovery of magnetic material in primate tissue was by Kirschvink (1981a).
The cerebellum, midbrain, and corpus callosum, but not the cerebral cortex, of the brain
of rhesus monkeys, Macaca mulatta, were found to have a diffuse saturation isothermal
remanent magnetization (sIRM) of about 25 pT, suggesting the presence of between 1 and
5 million single-domain crystals per gram of tissue.
Kirschvink (1981b) has also examined human adrenal glands for magnetic remanence.
He found a measurable amount of high-coercivity ferromagnetic material which appeared
to be finely disseminated throughout the tissue. Between 1 and 10 million single-domain
magnetite crystals per gram of tissue would be necessary to account for the observed mag-
netic remanence.
Histological as well as magneto metric techniques have been used to search for mag-
netic deposits in the human head. Details of the techniques used are in Baker et a1. (1983).
Magnetometric measurements on "soft" tissues from the human head revealed no areas
of significant magnetic remanence (Le., no readings greater than twice background noise
level where background includes the sIRM of the plastic container holding the tissue).
Specimens of skull, rib, and sphenoid bone also gave readings less than twice background
levels. In the thin but very hard bones that form the walls of the sphenoid/ethmoid sinus
complex, however, magnetic remanence was obtained that was consistently greater than
twice background. The range of sIRM (emu/g) for all tissues tested is given in Baker et al.
(1983). The sIRM of the sinus bones ranged from 3.02 to 31.58 x 10- 6 emu/g.
Tissue from all of the regions tested magneto metrically was examined histologically
except for the bone of the skull and the sphenoid bone itself. Although a scattering of iron-
staining materials was found throughout the soft tissues and in bone marrow, extensive
positive staining for ferric iron was found only within the bone of the sphenoid/ethmoid
sinus complex (Fig. 5). There appears to be a continuous layer of iron-staining material
roughly 2 fLm thick and about 5 fLm beneath the surface of the bone. However, a collab-
orative study (with ]. Takacs, Nuclear Physics, Oxford University) using a proton probe
failed to detect iron in this layer, even though sections of the same specimens of sphenoid
sinus bone stained positively in histological studies (Kennaugh, personal communication).
None of the above examples of magnetic remanence in primate tissue (Le., monkey
brain, human adrenals, human sinus bones) has positively been attributed to the presence
of magnetite. In the case of the sinus bones, the source of the magnetic remanence is
Figure 5. Section through bone from the wall of the sphenoid sinus of a 57-year-old woman after
staining for ferric iron. The slide shows a layer (about 5 f.Lm beneath the surface of the bone) that
stains positively for ferric iron by the Perl reaction. The bone is about 200 f.Lm thick. From Baker et
oJ. (1983).
currently being investigated in collaboration with Joe Kirschvink. California Institute of

Technology. The only specimen so far available for this investigation was contaminated
through first having been specially prepared to be photographed by being trimmed with
a band saw. The specimen is awaiting being "cleaned" for further testing.
In collaboration with D. M. Guthrie and J. H. Kennaugh . a pilot histological study of
the secretory lining of the human sphenoid sinus has been carried out with particular
attention to nerve fibers. These appear to be few and in no obvious way unusual for such
tissue. Unfortunately. though hopefully temporarily. we no longer have access to suitable
human specimens. our supply disappearing before we could attempt to study the inner-
vation and other features of the sinus bones by electron microscopy. For the moment.
therefore. we are restricted to making inferences concerning the distribution of iron and
nerve fibers in human sinus bones from our parallel studies of the analogous ethmoturbinal
bones of rodents (Mather et 01 .. 1982) . This work is described in detail by Mather (this
volume) and only those features that may be relevant to our human research are sum-
marized here.
Like humans. wood mice have (1) a magnetic compass sense (Mather and Baker. 1981);
(2) in the head. strongest magnetic remanence in the vicinity of the olfactory region (Mather
and Baker. 1981); and (3) a layer that stains positively for ferric iron beneath the surface
of the bones of the ethmoturbinal region (Mather et 01.. 1982). The woodmouse head also
contains magnetite (Kirschvink. personal communication).
Whether the iron-staining layer beneath the surface of the sinus bones of humans and
rodents is. or contains. magnetite and is thus the source of remanence is as yet unknown.
Recent developments in our studies of woodmice. however. along with our proton probe
studies of human sinus bones at Oxford . suggest this to be less likely than at first it ap-
peared. We now know for woodmice (Mather et 01.. in Mather. this volume) , for example.
554 Chapter 26
that the iron-staining layer is found in bones in parts of the head that show much reduced
sIRM compared to the anterior dorsal region. The layer is also found in at least the femur
among bones in other parts of the body. In marmosets, Callithrix, Mather and Kennaugh
(personal communication) have again found an iron-staining layer beneath the surface of
bones in various regions throughout the skull. By analogy, it seems we should assume that
the ferric layer is also widespread through human bones, including those (e.g., sphenoid,
skull, rib) that did not show above-background remanence in our magnetometer.
Kuterbach et 01. (1982) have shown for bees that, in the region of the abdomen that
shows magnetic remanence due to magnetite (Gould et 01., 1978), there are deposits of a
ferric material that stains by the same histological technique that was applied to human
sinus bones. These deposits are of hydrous iron oxides that are themselves only para-
magnetic but which could be biochemical precursors of magnetite. Such hydrous iron
oxides are often biological storage materials for iron and are precursors to magnetite for-
mation in chitons and magnetotactic bacteria (Towe and Lowenstam, 1967; Frankel et 01.,
1979). Only 0.33% of the iron visible in the bee's abdomen need be magnetite to account
for the level of remanence in the region (Kuterbach et 01., 1982).
Electron micrographs of bones from the heads of woodmice taken by J. M0rup J0r-
gens en (University of Aarhus, Denmark) show crystallike structures ranging in size from
4 x 2 !-Lm, comparable to hydrous iron oxides, down to 100 nm, within the size range of
single-domain magnetite (J0rgensen, personal communication). These str\lctures have so
far (December 1983) been found only in the ethmoturbinal bones and are always clustered
around nerve fibers. As far as we can tell, however, they are further from the surface of
the bone than the layer that we observe histologically. Indeed, so far there is no indication
of dense particulate material in this layer. Investigation is continuing.
In summary, human bones from the sphenoid/ethmoid sinus complex appear to have
an elevated level of magnetic remanence. We cannot yet feel confident, however, that we
have found the source of this remanence.
4.3. Physiology
A number of factors have been found to influence the accuracy of compass orientation
in chair experiments. Some of these may eventually give us insight into the nature and
physiology of the magnetoreceptor(s) by which nonvisual compass orientation may be
derived; others may be irrelevant to magnetoreception.
Accuracy of compass orientation varies with time of day. The initial report of this
effect (Baker and Mather, 1982) suggested two peaks of orientation, around noon and mid-
night. With more data, however, the midnight peak has disappeared and the pattern now
(Baker, 1984b) is for a gradual decrease in orientation ability with time of day after a period
of better orientation between about 1000 and 1500 GMT.
It was reported (Baker et 01.,1982) that, even between 1000 and 1500 GMT, orientation
may be influenced by clothing materials. Since then, experiments (between 1000 and 1500
GMT) on naturists (Anonymous, in Baker, 1984b) have shown that subjects wearing no
clothes (N = 39, eO = 2 ± 32°, r = 0.400, P = 0.00025, V-test) and subjects wearing only
a full-length robe made of 100% cotton (N = 39, eO = 7 ± 34°, r = 0.372, P = 0.0006,
V-test) both show good compass orientation, with no difference in strength of orientation
[N 1 •2 = 39,39, U = 706, P = 0.726 (2-tailed), Wallraff's two-sample test for the strength
of homeward orientation]. Subjects wearing an otherwise identical robe made of 100%
polyester, however, were disoriented (N = 39, eO = -3°, r = 0.073, P = 0.257, V-test).
This reduction in strength of orientation was significant, both relative to when wearing
nothing [N 1 •2 = 39,39, U = 607, P = 0.045 (l-tailed), Wallraff's two-sample test for the
strength of homeward orientation] and when wearing cotton [N 1 •2 = 39,39, U = 576, P
= 0.033 (l-tailed), Wallraff's two-sample test for the strength of homeward orientation].
All subjects wore a brass bar (as in Fig. 2) which they were told could be a magnet. The
experimenter wore no clothes in all tests.
The basis of the influence of polyester material on orientation has not yet been in-
vestigated, but our initial hypothesis is that it is related in some way to the electrostatic
fields that are such a feature of clothes made from such material. Magnetoreception by a
beetle is reported to be influenced by electric fields (Schneider, 1961), so perhaps the
polyester effect is due to a specific influence of electrostatic fields on the mechanism of
magnetoreception. Alternatively, the effect may be due to a nonspecific influence on gen-
eral attentiveness and concentration.
The aftereffect of wearing a magnet on the front of the head (Fig. 4) does not disappear
quickly. Indeed, preliminary results suggest that the original level of compass orientation
does not reappear until 24-48 hr after a 10-min exposure to a 200-G N-up magnet on the
right temple (Karen Tricker, personal communication). Rate of recovery may be influenced
by the orientation of subjects during sleep.
Human brain rhythms during sleep are reported to be influenced by bed alignment
(Kotleba et 01., 1973) and we have reported previously that performance in chair experi-
ments is also influenced by the alignment of the subject's bed (Baker et 01., 1982; Baker
and Mather, 1982). Since that report, bed alignment has been included as a matter of routine
on the questionnaire that all subjects complete before being tested. These questionnaires
are not seen by the experimenter. Figure 6 summarizes all data on bed orientation for the
eight experimental series completed by the end of 1982, broken down by clothing material;
data from the same series, broken down by time of day, are summarized in Baker (1984b).
The different series were carried out on different subjects (except for the two tests on
naturists), under different conditions, often in different locations, and each was aimed
primarily at the study of other influences. Groups tested have been: University staff; public
visitors to a University Open Day; Students (1981); Bramhall residents; Students (1982);
British National Orienteering Squad; and Manchester naturists, the last group being tested
wearing cotton or polyester.
Several features emerge from the data in Fig. 6 and Baker (1984b) [n.b. level 3 (see
Section 3.2.2) is used for all single-sample analyses; comparisons between samples use
Wallraff's nonparametric two-sample test for strength of orientation in a predicted direction
(Batschelet, 1981, pp. 127-128)]. N-sleepers (Le., people who sleep in beds aligned such
that their feet point north ± 45°) show significant orientation between 1000 and 1500 GMT
(N = 68, eO = 16 ± 28°, r = 0.342, P = 0.00008, V-test) as do S-sleepers (N = 52, eO =
-4 ± 68°, r = 0.207, P = 0.017, V-test), the difference between the two being not sig-
nificant [N t •2 = 68,52, U = 1602, P = 0.189 (l-tailed)]. E-W sleepers, on the other hand,
are not significantly oriented (N = 87, eO = 5°, r = 0.090, P = 0.119, V-test) and their
performance is significantly weaker than that of N-S sleepers [N t •2 = 87,120, U = 4430,
P = 0.032 (l-tailed)].
Relative to their orientation around midday, S-sleepers show a significant deterioration
after midnight [Nt .2 = 12,52, U = 106, P = 0.00046 (2-tailed)] whereas N-sleepers [N1,2
= 10,68, U = 279, P = 0.362 (2-tailed)] and E-W sleepers [N t •2 = 11,87, U = 345, P =
0.134 (2-tailed)] do not. After midnight, therefore, the compass orientation of N-sleepers
is significantly better than E-, S-, and W-sleepers combined [N t •2 = 10,23, U = 58, P =
0.013 (l-tailed)].
Polyester material has little influence (Fig. 6) on the compass orientation of N-sleepers
[N t •2 = 28,48, U = 663, P = 0.46 (l-tailed)] but seems to interfere with compass orientation
by S-sleepers [N1,2 = 21,33, U = 163, P = 0.0007 (l-tailed)], E-sleepers [N1,2 = 18,26, U
= 117, P = 0.0026 (l-tailed)], and W-sleepers [N1,2 = 25,35, U = 376, P = 0.178 (1-
tailed)], though in the last case the difference is not significant.
556 Chapter 26
e=16° e =-23'
r =0.43 r =0.20
0'-
no 0
polyester o
000
....... o
ú ç =
e=-157
. . 0
e=-170
r =0.21 r =0.11
N-SLEEPERS S-SLEEPERS E-SLEEPERS W-SLEEPERS

Figure 6. Apparent synergistic influence of bed alignment and clothing materials on compass ori-
entation in chair experiments. Five groups of subjects were tested wearing clothes that were normal
or made from a deliberately high level of polyester material. Three groups were tested wearing nothing
or clothes made of 100% cotton. Most conventions as in Fig. 2, except the mean vector for the dots
shown is placed outside the circle. Significance of this mean vector (V-test) shown inside circle. Lines
inside circle show mean vectors for individual groups. N-sleeper, a person who sleeps on a bed the
foot of which points north ± 45° (see Fig. 7).
The general impression from these data is that the ability for compass orientation of
N-sleepers is more robust than that of other groups. Preliminary data suggest that such
robustness appears within 1 week of becoming aN-sleeper.
5. Discussion
A unique feature of the study of humans in the general search for animal magneto-
receptors is the combination of behavioral. magnetometric, and anatomical evidence. Stud-
ies of the dowsing response and of orientation in chair and bus experiments all implicate
a magnetoreceptor in the front of the head in a region that has elevated magnetic remanence
and appears to contain ferric material.
The hypothesis that the sinus region is the site of magnetoreception involved in com-
pass and goal orientation by humans awaits testing by micromanipulation of magnetic
fields in localized regions of the head during chair and bus experiments. Until then, en-
couragement comes primarily from the approximate coincidence of this region with that
implicated by experiments on compass and goal orientation and, for that matter, dowsing.
N4--
Figure 7. Hypothesis for the influence of bed alignment

on human magnetoreceptors in terms of the alignment
of magnetic particles in the head. Horizontal and ver-
tical components (dashed arrows) of the geomagnetic
field are shown for a location in midtemperate lati-
tudes in the northern hemisphere. As a N-sleeper rolls
from side to side during sleep, the vertical component
cancels out giving a net alignment as shown. In con-
sequence, when a N-sleeper stands and walks around
during the day, any dipole in the head will more or
less complement the geomagnetic field. When a S-
sleeper stands, any dipole will be opposed to the geo-
magnetic field. As E-W sleepers roll from side to side
during sleep, both vertical and horizontal components
cancel out. Net alignment results only if the person
spends asymmetric times during sleep on right or left
sides. From Baker (1984a).
Further encouragement could be derived from the development of orientation ability in

children. Preliminary results suggest that nonvisual orientation, as measured in chair ex-
periments, first appears at about the age of 5 years, and reaches adult accuracy (or better)
by the age of 12 years. Sphenoid and ethmoid sinuses are small at birth, are growing quickly
by the age of 5 years, and are nearly adult in size by the age of 12 years (Maresh, 1940;
Caffey, 1978).
Whether or not the iron-staining materials observed in the sinus bones contain mag-
netite, there are various features of human compass and goal orientation that could be
consistent with a magnetoreceptor based on deposits of magnetic particles. This view is
strengthened by recent work on homing pigeons.
Gould and Walcott (in Walcott, 1982) have placed the heads of live homing pigeons
in strong magnetic fields and then measured homing performance. Preliminary data suggest
that after a pigeon's head has been placed in a strong, alternating magnetic field, mag-
netoreception is less accurate, whereas after the head is placed in a strong, static field,
magnetoreception may actually be improved. Gould and Walcott suggest that the net align-
ment of magnetic particles in a magneto receptor in the pigeon's head may be reduced by
an alternating field but improved in a static field.
Adopting the thesis that more-strongly aligned particles allows more accurate, and
perhaps more robust, magnetoreception, Fig. 7 presents a hypothesis (from Baker, 1984a)
for the influence of bed alignment on human compass orientation. The hypothesis would
explain (1) the more robust and accurate performance of N-sleepers and (2) the finding
that, although both N- and S-sleepers begin the day with an accurate compass sense, the
performance of S-sleepers, as the day progresses, deteriorates more rapidly than the per-
formance of N-sleepers. In addition, the hypothesis makes a number of testable predictions.
First, in the tropics, N- and S-sleepers should perform equally well. Second, in the southern
hemisphere, S-sleepers should have the more-robust compass sense. Third, N-up magnets
placed vertically alongside the magnetoreceptor should interfere with the orientation of
S-sleepers more than with that of N-sleepers, the converse being true for S-up magnets.
Post-hoc analysis of old data and preliminary results from new experiments (Baker,
1984a, 1985) support this last prediction.
558 Chapter 26
Other features of the results from chair and bus experiments could also be interpreted
in terms of the alignment and realignment of magnetic particles. The improvement in goal
orientation while exposed to an artificial magnetic field (Table VI) and the "aftereffect,"
the influence of a magnet on orientation persisting even after the magnet is removed (Fig.
6 and in Baker, 1984a,b, 1985), are both compatible with the alignment and realignment
of magnetic particles. Even the apparent relationship between magnetic storm activity before
an experiment and the performance of humans (and pigeons) could conceivably be inter-
preted in terms of the influence of the magnetic storms on the net alignment of magnetic
particles.
If this latter suggestion is true, it would imply that even under natural conditions the
hypothetical particle-based magnetoreceptor is in a dynamic state. Perhaps, therefore, one
explanation for the high precursor/magnetite ratio that Kuterbach et 01. (1982) seem to have
found in bees is that magnetoreceptors are characterized by a high and continuous turnover
of magnetite. High turnover could be due to a clumping or other disruption of magnetite
particles with time and the need to replace them with further particles of optimum size
and spacing; in this case, orientation during sleep could be an important element in mag-
netoreceptor maintenance.
Suppose that each person has a characteristic alignment of magnetic particles in their
head that is a function of that person's particular sleep habits (Fig. 7). Suppose, also, that
each person learns to associate a particular signal from their magnetoreceptor with a par-
ticular compass direction. Any change in alignment of magnetic particles could cause a
change in the signals generated by the magnetoreceptor for each compass direction. The
result could be a misinterpretation of compass direction that would persist even after the
magnet is removed. This aftereffect would then persist either until the old particles or
newly deposited particles regain the original alignment as a result of bed orientation, or
until the person learns to associate the new signals from the magnetoreceptor with the
correct compass directions. Even if realignment of particles does cause a misinterpretation
of compass direction, however, there is no obvious reason for it also to cause a misinter-
pretation of angle of turn within the geomagnetic field.
Most, if not all, results from bus and chair experiments seem compatible with a mag-
netoreceptor based on particles subject to alignment and realignment, either by natural
events or by experimental manipulation. The exciting challenge we now face is to produce
an internally consistent model that allows the influence on orientation of specific manip-
ulations of particle alignment to be predicted in advance. When we are able to obtain
predictable and replicable effects from experiments based on such a model, we may feel
that we are at last nearing an understanding of one of the most contentious of biological
enigmas.
ACKNOWLEDGMENTS. Professor Yates, Dr. M. I. Wright, and the staff of the Pathology De-
partment of Manchester University provided human tissue for magnetometric and histo-
logical studies and Dr. P. Dagley and the Geophysics Department, University of Liverpool,
allowed the use of their magnetometer. The geomagnetism unit of the IGS, Edinburgh,
provided indices of magnetic storm activity. My own bus experiments have been carried
out in collaboration with Dr. S. E. R. Bailey 11979), Dr. J. G. Mather (1979-1982), and R.
G. Murphy (1982) at Manchester, Professor J. L. Gould (1981) at Princeton, and Professor
K. Adler (1981) at Cornell. R. G. Murphy carried out the exhausting series of chair ex-
periments on Manchester schoolchildren, and nearly 100 friends, family, and students
have been experimenters in other chair experiments. Dr. P. Semm, Dr. J. Ratcliffe, Professor
J. Kirschvink, and S. Hemmings provided valuable references. To all of these and the 1000
or so people who have been subjects, I offer my sincere thanks. This work was supported
by SERC Grant GRlB74337.
6. Summary
The physiological responses of primates to changes in the ambient magnetic field have
been studied using monkeys and humans. Magnetic storms and strong, artificial magnetic
fields have both been claimed to have an adverse influence on primate physiology. On the
other hand, when applied to fractured bones, artificial but weak electromagnetic fields
have been claimed to aid repair. Few of these physiological responses, however, obviously
involve magnetoreception. Studies of the latter have been restricted to humans and have
used three types of behavioral response: the dowsing reflex; compass orientation; and goal
orientation. All three series of experiments implicate a magnetoreceptor in the front half
of the head, perhaps behind but roughly level with the eyes, while studies of the dowsing
response also implicate magnetoreceptors in the region of the adrenal glands. Magneto-
metric and histological studies of human tissue have found that the adrenal glands and
bones from the sphenoid/ethmoid sinus complex are both regions of elevated magnetic
remanence and seem to contain ferric deposits. Studies of compass and goal orientation
suggest that ambient magnetic fields can physically alter the magnetoreceptor in a manner
that could be consistent with the alignment and realignment of magnetic particles. Such
realignment seems sometimes to improve rather than disrupt orientation.
References
Alexander, H. S., 1962, Biomagnetics-The biological effect of magnetic fields, Am. J. Med. Electron.
1:181-187.
Aneshansley, D. J., and Larkin, T. S., 1981, V-test is not a statistical test of "homeward" direction,
Nature 293:239.
Baker, R R, 1980, Goal orientation by blindfolded humans after long-distance displacement: Possible
Baker, R R, 1981, Human Navigation and the Sixth Sense, Hodder & Stoughton, London.
Baker, R R, 1982, Migration: Paths through Time and Space, Hodder & Stoughton, London.
Baker, R R, 1984a, Sinal magnetite and direction finding, Phys. Technol. 15:30-36.
Baker, R R, 1984b, Bird Navigation: The Solution of a Mystery?, Hodder & Stoughton, London.
Baker, R R, 1985, Exploration and navigation: The foundation of vertebrate migration, in: Migration:
Mechanisms and Adaptive Significance (M. A. Rankin, ed.j, Port Aransas Marine Laboratory, Port
Aransas.
Baker, R R, and Mather, J. G., 1982, A comparative approach to bird navigation: Implications of
parallel studies on mammals, in: Avian Navigation (F. Papi and H. G. Wallraff, eds.), Springer-
Baker, R. R, Mather, J. G., and Kennaugh, J. H., 1982, The human compass?, EOS 63:156.
Baker, R R, Mather, J. G., and Kennaugh, J. H., 1983, Magnetic bones in human sinuses, Nature 301:78-
80.
Barlow, J. S., 1964, Inertial navigation as a basis for animal navigation, J. Theor. BioI. 6:76-117.
Barnothy, M. F. (ed.), 1964, Biological Effects of Magnetic Fields, Volume 1, Plenum Press, New York.
Bassett, C. A. L., Pawluk, R J., and Pilla, A. A., 1974, Augmentation of bone repair by inductively
coupled electromagnetic fields, Science 184:575-577.
Batschelet, E., 1981, Circular Statistics in Biology, Academic Press, New York.
Beischer, D. E., 1969, Vectorcardiogram and aortic blood flow of squirrel monkeys (Saimiri sciureus)
in a strong superconductive electromagnet, in: Biological Effects of Magnetic Fields, Volume 2
(M. F. Barnothy, ed.), Plenum Press, New York, pp. 241-259.
Beischer, D. E., 1971, The null magnetic field as reference for the study of geomagnetic directional
effects in animals and man, Ann. N.Y. Acad. Sci. 188:324-330.
Beischer, D. E., and Knepton, J. C., 1944, Influence of strong magnetic fields on the electrocardiogram
of squirrel monkeys. Aerosp. Med. 35:939-945.
560 Chapter 26
Beischer, D. E., and Reno, V. R., 1971, Magnetic fields and man; where do we stand today?, in: AGARD
Conference Proceedings N. 95, Part III; Special Biophysical Problems in Aerospace Medicine (A.
M. Pfister, ed.), pp. C-12-1-C-12-7.
Bird, c., 1979, The Divining Hand: The 500-year-old Mystery of Dowsing, Dutton, New York.
Caffey, J., 1978, Pediatric X-ray Diagnosis, 7th ed., Year Book, Chicago.
Drinker, C. K., and Thompson, R. M., 1921, Does the magnetic field constitute an industrial hazard?,
J. Ind. Hyg. 3:117.
Science 203:1355.
Giirling, T., B6i:ik, A., and Lindberg, E., 1979, The acquisition and use of an internal representation
of the spatial layout of the environment during locomotion, Man-Environ. Syst. 9:200-208.
Gatty, H., 1958, Nature is your guide, Collins, London.
Gould, J. L., 1980, Homing in on the home front, Psychol. Today 14:62-71.
Gould, J. L., and Able, K. P., 1981, Human homing: An elusive phenomenon, Science 212:1061-1063.
201:1026-1028.
Harvalik, Z. V., 1978, Anatomical localization of human detection of weak electromagnetic radiation:
Experiments with dowsers, Physiol. Chern. Phys. 10:525-534.
Howard, 1. P., and Templeton, W. B., 1966, Human Spatial Orientation, Wiley, New York.
Hudson, W. H., 1922, On the sense of direction, Cent. Mag. 104:693-701.
Jaccard, P., 1931, Le sens de direction et l'orientation lointain chez l'homme, Payot, Paris.
Juurmaa, J., 1966, An analysis of the ability for orientation and operations with spatial relationships
in general, Work Environ. Health 2:45-52.
Ketchen, E. E., Porter, W. E., and Bolton, N. E., 1978, The biological effects of magnetic fields on man,
Am. Ind. Hyg. Assoc. J. 39:1-11.
Kholodov, J. A., Alexandrovskaya, S. N., Lukjanova, S. N., and Udarova, N. S., 1969, Investigations
of the reactions of mammalian brain to static magnetic fields, in: Biological Effects of Magnetic
Fields, Volume 2 (M. F. Barnothy, ed.), Plenum Press, New York, pp. 215-219.
Kirschvink, J. L., 1981a, Biogenic magnetite (Fea04): A ferrimagnetic mineral in bacteria, animals, and
man, in: Ferrites: Proceedings of the International Conference Japan, 1980, pp. 135-137.
Kirschvink, J. L., 1981b, Ferromagnetic crystals (magnetite?) in human tissue, J. Exp. BioI. 92:333-
335.
Kotleba, J., Bielek, J., Glos, J., and Barta, J., 1973, 0 moznom vplyve magnetickeho pol'a zeme na
spanok cloveka, Cs. Fysiol. 22:459-460.
Kuterbach, D. A., Walcott, B., Reeder, R. J., and Frankel, R. B., 1982, Iron-containing cells in the honey
Lewis, D., 1972, We, the Navigators, Australian National University Press, Canberra.
Lord, F. E., 1941, A study of spatial orientation of children, J. Educ. Res. 34:481-505.
Am. Bull. 73:435-438.
Malmstrom, V. H., 1976, Knowledge of magnetism in pre-Columbian Mesoamerica, Nature 259:390-
391.
Maresh, M. M., 1940, Paranasal sinuses from birth to late adolescence, Am. J. Dis. Child. 60:58.
Mather, J. G., Baker, R. R., and Kennaugh, J. H., 1982, Magnetic field detection by small mammals,
EOS 63:156.
Murayama, M., 1965, Orientation of sickled erythrocytes in a magnetic field, Nature 206:420-422.
Rocard, Y., 1964, Le Signal du Sourcier, Dunod, Paris.

Schneider, F., 1961, Beeinflussung der Activat des Maikafers durch Veranderung der gegenseitigen
Lage Magnetischer und elektrischer Felder, Mitt. Schweiz. Entomol. Ges. 33:232-237.
Southern, W. E., 1971, Gull orientation by magnetic cues: A hypothesis revisited, Ann. N.Y. Acad.
Sci. 188:295-311.
Srivastava, B. J., and Saxena, S., 1980, Geomagnetic-biological correlations: Some new results, Indian
J. Radio Space Phys. 9:121-126.
rowe, K. M., and Lowenstam, H. A., 1967, Ultrastructure and development of iron mineralization in
the radular teeth of Chryptochiton stelleri (Mollusca), J. Ultrastruct. Res. 17:1-13.
Viguier, c., 1882, Le sens de I'orientation et ses organs chez les animaux et chez I'homme, Rev.
Philomath. 14:1-36.
von Lucannas, F., 1924, On the sense of locality in men and animals, Rev. Revs. 70:218.
Walcott, c., 1982, Is there evidence for a magnetic map in homing pigeons?, in: Avian Navigation (F.
Wiltschko, R., and Wiltschko, W., 1978, Relative importance of stars and magnetic field for the accuracy
of orientation in night-migrating birds, Oikos 30:195-206.
Worchel, P., 1951, Space perception and orientation in the blind, PsychoI. Monogr. 65:3-32.
537.
Zusne, 1., and Allen, B., 1981, Magnetic sense in humans? Percept. Mot. Skills 52:910.
Chapter 27
Statistical and Methodological Critique
of Baker's Chapter
TOM DAYTON
1. Statistics in General. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ., 563

2. "Chair" Experiments Results Section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 565
3. Princeton Data Do Not Support Baker. . . . . . . . . . . . .. . .. . . . . . .. . . . . . . . . 565
4. Magnets vs. Controls for Baker's Experiments . . . . . . . .. . . .. . . . . . .. . . . . . . . 567
5. Magnets vs. Controls for K-6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 567
6. Physiology of Magnetoreceptors. . . . . . . . . . . . . . . . . . . .. . . .. . . . . . . . . . . . 568
7. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . .. . . . . . .. . . . . . . . . 568
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . .. . . . . . . . . . . .. . 568
Robin Baker's chapter contains methodological and factual errors in the "Statistics" and
"Results" subsections of the two sections "Compass Orientation" and "Goal Orientation";
there are also problems with the "Physiology" subsection of "Magnetoreceptors?" Cor-
rections of these errors substantially weaken Baker's conclusions.
1. Statistics in General
Comparison of two conditions' means without formal hypothesis construction and
testing is inappropriate for confirmatory data analysis. Comparisons of two groups' V-test
statistics, correlation coefficients, or any other statistics are just as much in need of formal
analyses as are comparisons of more common statistics (e.g., means). There are several
instances in which Baker tried to contrast such statistics of two groups inappropriately by
comparing the results of the two one-sample tests. For example, Baker should have applied
his chosen test-Wallraff's modification of the Wilcoxon-Mann-Whitney U test (Bat-
schelet, 1981, pp. 125-128)-to all comparisons rather than to only a select few.
One-tailed tests usually are not justified merely by "clear prediction" of the direction
of outcome. This point is a difference between statistics and research methodology, and
is attested to by the relatively rare usage of one-tailed tests in the behavioral literature
despite the prevalence of "clear predictions." Pure science infrequently uses truly direc-
tional hypotheses; researchers are usually interested in any difference between treatment
groups even though they prefer or expect one direction over another. Baker was obviously
interested in any effects of artificial magnetic fields and so did not treat a difference in
an unexpected direction the same as he treated no difference; the underlying hypothesis
was thus actually non directional.
TOM DAYTON • Department of Psychology, University of Oklahoma, Norman, Oklahoma, 73019

563
564 Chapter 27
The danger of using a one-tailed test when a two-tailed is appropriate is the doubled
risk of Type I error (Le., mistakenly rejecting the null hypothesis). An obtained statistic
that falls at a < 0.10 under a two-tailed hypothesis can be made to fall at a < 0.05 merely
by formal adoption of a directional hypothesis.
I believe nondirectional tests are appropriate for the work discussed in Baker's chapter
because we are so uncertain of possible mechanisms underlying any potential human mag-
netic sense that we are not at all sure of the potential effects of the experimental treatments.
In his chapter the possibility for a priori adoption of directional hypotheses for brand new
data did not exist because it was a review.
Sometimes the argument is made that one-tailed tests are appropriate for exploratory
research because the cost of mistakenly failing to reject the null hypothesis-and so failing
to detect a new phenomenon-is greater than the cost of mistakenly and temporarily be-
lieving in the existence of the effect. The proper procedure for this case is use of a two-
tailed test (especially in exploratory research) and explicit adoption of a higher a of 0.10
or whatever seems appropriate. Use of a one-tailed test hides the fact that a large Type I
probability is being accepted.
Baker correctly points out the necessity of analyzing data beyond "level one" -usually
referred to in the literature as "first-order"-when data points are not independent of each
other. The correct procedure in such a case is to collapse across individuals, occasions,
etc., until the resulting data points are independent. Second-order analysis is common in
the orientation literature because animals almost certainly do not make judgments of di-
rection independent of immediately preceding judgments. In the human bus experiments,
for example, if a subject decides that the first orientation site is considerably south of home
and is then taken for a brief ride to the second site, the subject will probably not decide
he or she is now north of home. Consequently, that subject's errors of estimate from all
the sites should be averaged and that average error taken anew as a single data point with
r = 1.
Third-order analysis, however, is rarely justified. Contrary to Baker's assertion, it is
untrue that if second-order is good, then third-order is better. Third order is appropriate
in these experiments only if there is reason to suspect lack of independence of subjects
from each other. I see no reason to be that suspicious, but if I did, I would suspect that
the experiments were so poorly designed and conducted that the data shouldn't be trusted
regardless of statistical manipulation. There are definite and important costs undertaken
with higher-order analysis so it should not be done on whim. The basis of higher-order
analysis is elimination of information from the data in order to obtain a solid estimate of
exactly how much independent data there are. If the dependence is very small, then a
relatively large amount of independent information is lost by collapsing into higher order,
and perhaps it is better to use lower order. If the dependence is large, then collapsing yields
a much truer picture of the world at the cost of only a very small amount of independent
information, and higher order is appropriate.
Examples of this principle: Subjects in bus experiments are probably independent of
each other so second-order analysis is appropriate but third-order is not. Subjects partic-
ipating in more than one journey in a single experiment probably do not make independent
judgments on each journey (e.g., "Yesterday I think they took me south and there is only
one long road leading south; they probably won't take me to the same place two days in
a row, so today I must not be south of home") and so third order is appropriate for them.
Finally, the large number of statistical tests in this chapter cause the experimentwise
probability of a Type I error (Le., mistakenly rejecting the null hypothesis) to be much
larger than the individual test probability of 0.05; this is a danger in all meta-analyses.
Experimentwise a hits 0.50 at about 13.5 independent tests with individual a = 0.05.
Statistical Critique 565
(Because Baker's tests are not all independent of each other, more than 13.5 tests are
necessary to reach experimentwise a = 0.50.) An occasional significant result is to be
expected merely by chance, so it is important to verify such a result by looking for con-
sistency of results among all experiments of the same type. Formal meta-analysis-which
Baker did not do-does just this.
2. "Chair" Experiments Results Section
Baker failed to adequately test for use of an inertial sense by subjects in chair exper-
iments when he looked for poorer mean accuracy (the mean across all subjects) for the
last than for the first of eight trials. Once again he failed to use a formal test for the difference
between two groups. I grant that the direction of the difference does not support an inertial
method of orientation-r is larger for estimate eight than for estimate one-but the 0.047
difference in r is so small that it may well be due to random fluctuation. Note that only
the vector length r is relevant to this question because the inertial error should be random
in direction and strength across subjects; the mean angle of estimate one should be equal
to that of estimate eight. The expectation based on use of inertial sensing is that the number
of subjects with extreme errors should be larger in estimate eight than in estimate one, so
r should be smaller in estimate eight than in estimate one. I have strong doubts about the
power of this two-estimate comparison, though. The cumulative error due to inertial sen-
sing should probably be quite small relative to other effects, so the adequate test would
be a within-subjects trend analysis across all eight trials.
Baker's second method for detecting use of an inertial sense is in fact merely a test of
good versus poor subjects in general. One would expect poor subjects to do poorly on their
second estimate as well as on their first, regardless of the method (magnetic or other) they
use. Consequently, this test is useless for isolating cumulative error of inertial sensing.
3. Princeton Data Do Not Support Baker

I was present at Princeton-2, scored some of the response cards, and computed second-
order statistics from the raw data. To Gould I pointed out the discrepancies between my
data and those appearing in Gould and Able (1981), so Gould sent me corrections obtained
from checking of the response cards. Gould also kindly sent me the Princeton-l raw data,
from which I computed second-order statistics. Of all the Princeton experiments, only the
magnets group in Princeton-l was significantly oriented toward home (see Table I). Note
that the direction of difference between magnets and controls is opposite in Princeton-l
and -2, which hints that the significant Princeton-l result is due to random fluctuation;
the probability of obtaining at least one significant result by chance in the 12 Princeton
tests performed in Table I is about 40%
Baker never reported the nonsignificant Princeton-2 magnets vs. controls test, but only
the significant (according to his calculations) difference in Princeton-1. (In fact, his test
of Princeton-1 is incorrectly one-tailed; the correct two-tailed test using his incorrect num-
bers is not significant.) My calculations reveal no significant difference between magnets
and controls in Princeton-lor -2. Two-tailed tests are especially appropriate here because
the hypothesis fluctuates between facilitative effects of magnets (e.g., Manchester 1981-
82 as reported in Baker's chapter) and detrimental effects (e.g., Barnard Castle as originally
reported in 1980 by Baker).
566 Chapter 27
Table I. Second-Order Analysis of Princeton Experiments,

from Raw DataO
Data n eb r V
Princeton-l
Controls
5-2-80
10-30-80 20 -94.3 0.3709 -0.551
10-31-80 and 15 -172.6 0.3074 -4.573
11-24-80 19 -95.7 0.2783 -0.527
TotalC
Magnets 50 -107.4 0.2329 -3.485
5-2-80
20 -6.4 0.3348 6.653**
Princeton-2
Controls
2-16-81 20 -8.8 0.1733 3.426
2-17-81 19 44.7 0.1928 2.605
2-18-81 d 10 -8.9 0.1235 1.220
Total C 45 19.0 0.1436 6.110*
Magnets
2-18-81 15 13.1 0.2131 3.113
Total American control data

Princeton-l C 50 -107.4 0.2329 - 3.485
Princeton-2 C 45 19.0 0.1436 6.110*
Albany· 10 77 0.14 0.315
9 79 0.17 0.292
Cornell" 30 20 0.541 15.251* **
Total 144 -1.8 0.1284 18.485**
Magnets vs. controls

nmagnets ncontrols Wallraff's U
Princeton-l (5-2-80) 20 20 188.5
Princeton-2 (2-18-81) 15 10 62
a All tests are two-tailed. Compare this table to Baker's Table IV in this volume.
b e is second-order angular error from home direction.
C Third-order was used for four Princeton-l subjects in the Total tests because they
participated in both 10-30-80 and 10-31-80; I judged other subject duplications

as non problematic because of the greater time elapsed between experiments. I
treated four subjects in Princeton-2 controls similarly.
d One subject had r = 0 (Le., no mean angle) and so was excluded.
e I did not compute Albany and Cornell from raw data or check them in any way.
* 0.05 < P < 0.10; ** 0.01 < P < 0.05; *** p ú =0.0001.
In summary of the Princeton data, contrary to Baker's assertion, neither magnets nor
controls are significantly oriented toward home in Princeton-2, and only magnets are sig-
nificant in Princeton-I. Though the test of the total American control data (see Table I) is
significant. this obviously derives solely from the highly significant and atypical Cornell
results which are questionable-see the comment by Adler and Pelkie in this volume. The
differences between magnets and controls are not significant in either Princeton-lor -2.
Thus, the only bus experiment results that directly address magnetic detection (i.e., mag-
nets vs. controls) have not been replicated by anyone but Baker.
Statistical Critique 567
Table II. Fisher's z-transform Comparisons of Magnets' vs. Controls'

Correlations of K-6 (Magnetic Storm Activity) with Homeward
Component in Goal Orientation ExperimentsQ
Controls Magnets F p
Compass estimates
Correlation -0.905 -0.310
z-score -1.499 -0.320 0.9686 0.063
Pointing estimates
Correlation 0.452 -0.310
z-scores 0.4872 -0.320 0.8997 0.200
a Correlations of subjects wearing magnets are not significantly different from correlations of
control subjects. Correlations are from Baker's Table V in this volume. n = 8 in each of the
four conditions. Both tests are two-tailed.
4. Magnets vs. Controls for Baker's Experiments

Baker reported Barnard Castle magnets vs. controls only for site 1! There were two
sites (Baker, 1980); following Baker's explicit logic in the rest of his chapter and the usual
practice, sites 1 and 2 should be averaged for each subject and a second-order Wallraff's
test of magnets vs. controls done. I have computed Wallraff's test for Barnard Castle Site
1 and in contrast to Baker's figures here (U = 69) found it not significant (U 15 •16 = 93, P
> 0.05, two-tailed); data were from the illustrations in Baker (1980). Inclusion of Site 2
would have made the numbers considerably 'worse as I found for Site 2, U15 •16 = 113, P
ú = 0.05. Though I could not perform a second-order analysis for lack of identification of
subjects, I did a first-order analysis out of curiosity and obtained U30 •32 = 405, P s; 0.29.
(Note that all of these analyses used the best tie-breaking technique of assigning the average
rank to ties.)
It appears, then, that in contrast to Baker's assertions, the Barnard Castle magnets vs.
controls difference is not significant at Site 1, Site 2, 1 and 2 combined first-order, and
probably not second-order either.
The remaining Manchester 1979-80 experiments (of which Barnard Castle was a mem-
ber) were not given a statistical test of significance of difference between magnets and
controls despite Baker's implication of a significant difference. Neither did Manchester
1981-82 have a test, despite Baker's assertion that the difference is significant.
If Baker is correct about Barnard Castle magnets reducing performance, why did later
experiments (Le., the rest of Manchester 1979-80 and Manchester 1981-82) show a re-
versed trend? This discrepancy at the least indicates the appropriateness of two-tailed tests,
and at the most suggests that the results are due to chance. Note that even use of the same
type of magnets (bar magnets) in Manchester 1981-82 as in Barnard Castle was not enough
to guarantee the same direction of outcome.
5. Magnets vs. Controls for K·6
Baker did not do formal tests of the differences between correlations of magnets with
K-6 (magnetic storm activity) and controls with K-6 (Baker'S Table V in this volume). I did
Fisher's z-transforms and discovered no significant differences (see Table II). Thus, wearing
of magnets seems to have no effect on the correlation between magnetic storm activity and
orientation. Further evidence for the appropriateness of two-tailed tests is the contradiction
of direction of effect of magnets in pointing vs. compass estimates.
568 Chapter 27
6. Physiology of Magnetoreceptors
Section 4 of Baker's chapter contains several inappropriate uses of one-tailed tests.
The clothing hypothesis test and the sleeping direction N-S vs. E-W test are not significant
two-tailed though Baker reported them as significant one-tailed. The interaction of sleeping
direction with time of day and with polyester clothing are reported incorrectly one-tailed
but their conclusions are not changed by adoption of two-tailed tests.
7. Summary
There are four basic methodological problems with Baker's chapter. In decreasing order
of importance:
1. The Princeton data do not support Baker's hypothesis, contrary to his assertions.
2. Baker frequently failed to use statistical tests of comparison between magnets and
controls; correct tests usually reveal no difference.
3. Use of one- instead of two-tailed tests inflated by a factor of two the probability of
mistakenly accepting existence of effects. Most of Baker's one-tailed tests do not
reach a = 0.05 two-tailed.
4, Inappropriate use of third-order analyses reduced the amount of information ex-
tracted from the raw data by the statistical tests.
ACKNOWLEDGMENTS. I thank Joseph Lee Rodgers for comments on the manuscript.

This material is based in part upon work supported under a National Science Foun-
dation Graduate Fellowship. Any opinions, findings, conclusions, or recommendations
expressed herein are those of the author and do not necessarily reflect the views of the
National Science Foundation.
References
Chapter 28
Human Navigation
Attempts to Replicate Baker's Displacement
Experiment
KENNETH P. ABLE and WILLIAM F. GERGITS
Baker (1980, 1981) reported a seemingly robust ability of blindfolded human subjects to
indicate the direction toward home (the starting point of the journey) by written and verbal
estimates or by pointing. The orientation of the subjects was sufficiently good that statis-
tically significant results were obtained with sample sizes of 10-20 individuals. Further-
more, the ability to indicate correctly the homeward direction seemed sometimes to de-
teriorate when the subjects wore magnets or Helmholtz coils (but see Dayton's commentary,
this volume). With the apparent exception of one collaborative test at Cornell (see Adler
and Pelkie, this volume). several attempts to repeat Baker's straightforward experiment
failed to yield comparable results (Gould and Able, 1981; Dayton, this volume).
Baker's chapter in this volume presents a considerably modified picture with regard
to human goal orientation. More recent displacement experiments, especially those with
several stops, have produced much poorer results (see Baker's Table III) so that overall
only a weak ability is evident upon higher-order statistical analysis. In addition, it has
been reported that applied magnets probably facilitate rather than disrupt the ability of
the subjects to perform goal orientation. At best, the situation is unclear.
Since the initial reports of Baker's results, we have made a number of attempts to
replicate the basic phenomenon, an ability by blindfolded subjects to indicate the home-
ward direction. Two such tests were described briefly in Gould and Able (1981). Here we
present a summary of displacement experiments that we have performed to date. For com-
parison, we have followed Baker's latest analysis procedures, performing second-order
analysis in cases where multiple stops were made during a displacement.
With respect to pointing at the final destination of the displacement journey, all in-
dividual runs and the pooled data were random (Table I, Fig. lA). The mean errors of the
five tests were all >85° and the nonsignificant mean of the pooled data was almost exactly
opposite homeward.
Written compass estimates of the homeward direction tended to be more accurate.
The mean errors for each of four tests were <79°, although only one test produced a sig-
nificant mean vector by the V-test. In the tests involving multiple stops, the data from
only a single stop (the first stop on April 16, 1981) yielded statistical significance. There
was, however, a tendency for the mean errors to be small at the first three stops on that
date as well as on May 7, 1981. The pooled distribution (Fig. lB) shows a weak orientation
(r = 0.289, P <0.05, V-test) with a mean + 38° relative to homeward.
KENNETH P. ABLE and WILLIAM F. GERGITS • Department of Biology, State University of New
York, Albany, New York, 12222 .
569
570 Chapter 28
Table I. Summary of Pooled Data for Blindfolded Pointing and Written Compass
Estimates of the Homeward Direction during Displacement Experiments U
Distance No. Mean vector
traveled of No. of
Date Destination (km) stops subjects eO r P (V-test)
Pointing while blindfolded

April 24, 1980 Smalley Road 16 1 10 +105 0.25 n.s.
April 24, 1980 Old Chatham 17 1 9 +140 0.19 n.s.
April 16, 1981 Mechanicville 77 5 17 -85 0.18 n.s.
May 7, 1981 Mechanicville 77 5 9 -116 0.33 n.s.
March 2, 1984 Old Chatham 17 1 22 +144 0.16 n.s.
Written compass directions

April 24, 1980 Smalley Road 16 1 10 +77 0.14 n.s.
April 24, 1980 Old Chatham 17 1 9 +79 0.17 n.s.
April 16, 1981 Mechanicville 77 5 17 -17 0.35 <0.05
May 7, 1981 Mechanicville 77 5 9 +50 0.37 n.s.
a Pointing was performed only at the destination of each journey. On trips with multiple stops, second-order analysis
of the written compass directions has been performed, data from each subject being averaged over the several stops
of the trip.
Several points are relevant to the interpretation of these (and some of Baker's) data:
1. We have performed no tests involving magnets, but have attempted first to obtain
a repeatable homeward orientation behavior. Having been unable to obtain a consistent
result, we have had no clear phenomenon with which to attempt magnetic manipulations.
2. If Baker's hypothesis of magnetic route-based navigation is correct, it is not obvious
to us why written compass bearings should be better predictors of the homeward direction
than pointing. Indeed, just the reverse seems intuitively more likely. Even in Baker's own
HOMEWARD
Figure 1. Plots of the estimates of
homeward direction by blind-
folded human subjects transported
in vans. Van windows were cov-
ered with aluminum foil. (A) Each
subject's attempt to point toward
the starting point of the displace-
ment journey (home) at the trip's
destination. Subjects remained
blindfolded, but were removed
o ú =• from the van one by one and asked
to point homeward while outside
A. N = 67 B. N = 45 the van. Each dot represents a sin-
r = .080 r = .289 gle estimate by each subject. (B)
Written compass estimates of the
homeward direction. Open dots represent a single estimate made at the destination of journeys in-
volving only one stop. Solid dots represent the means of each subject's five estimates during a single
trip involving multiple stops. r represents the length of the mean vector; significance levels refer to
the V-test.
Displacement Experiments 571
1.0
.8
.6
.4
.2
u 0
Cl
::c -.2
-.4
Figure 2. The homeward directional com-
....,6
ponent (HDC) [= costa - ú F K = where a is
the mean direction of the distribution and -.8
ú = is the homeward direction] of the -1.0
pooled data at each successive stop along 0 10 20 30 40 50 60 70 80
a displacement route with five stops. DISTANCE TRAVELED (Km.l
data (see his Table III), pointing estimates were overall much poorer than written compass
estimates.
If, on the other hand, subjects are attempting to construct a mental map based on turns,
accelerations, and other consciously perceived aspects of the route of travel, we should
expect the written estimates to be better. To be quite accurate, one need only be able to
figure out that he or she is in, say, the northeast quadrant with respect to home. To be
able, in addition, to point toward southwest is a much more difficult proposition based
on inertial information, whereas in theory it should not be if a functional magnetic compass
exists.
3. Based on our own experience as subjects in displacement experiments and the
responses of other subjects to questionnaires filled out after the tests, we believe that
inertial cues perceived during the outward journey may be importantly involved in the
weak orientation of the written home directions: In our tests employing multiple stops,
there was a striking pattern of better homeward orientation at the first three stops and a
rapid deterioration thereafter (Fig. 2). A similar trend seems to exist in the Princeton data.
This pattern would be predicted by the hypothesis that inertial cues are predominant (errors
should accumulate as the journey continues), but should not occur by Baker's mechanism.
4. If our suspicion concerning the importance of inertial cues is correct, selection of
appropriately complex displacement routes is critical. Routes should contain many slow,
gentle turns and should be set up so that stops occur in as many quadrants as possible
with respect to home. Even a route with many turns will be inadequate if the distances
are short: a subject who has determined correctly that he or she is being displaced west-
ward, cannot be east of home later in the journey unless a considerable distance has been
covered.
We think it necessary, therefore, to exercise caution in the interpretation of our
results and trust they will not be misrepresented by others. Most importantly, they cannot
be taken to represent a replication of Baker's experiments. Attempts by subjects to point
toward home were entirely random and the trend in the written compass directions was
not nearly so robust as the data in Baker's original tests. Statistical significance rests on
the results from a single test, which itself was heavily influenced by the data from the first
stop. There is a weak homeward trend which runs through the written directions, but we
can provide no evidence (positive or negative) regarding a role of magnetism in that ori-
entation. Given all the data with which we are familiar, we think more attention should
be paid to the admittedly less interesting possibility that the weak orientation occasionally
572 Chapter 28
exhibited by displaced human subjects is based on nonmagnetic cues such as inertial

information.
References
Chapter 29
Human Homing Orientation
Critique and Alternative Hypotheses
KRAIG ADLER and CHRIS R. PELKIE
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 573
2. Bus Tests Conducted at Ithaca, New York. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 574
3. Oriented Distributions from "Random" Data. . . . . . . . . . . . . . . . . . . . . . . . . . .. 584
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 587
Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 587
Reply to Baker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 591
1. Introduction
Baker (this volume) has provided a comprehensive review of tests on human orientation
which permits a detailed evaluation of this potentially interesting and highly controversial
phenomenon. The idea that humans, when blindfolded and displaced, exhibit an uncon-
scious compass sense based on detection of the earth's magnetic field was first proposed
by Baker (1980) and followed up by him with a series of additional experiments (references
in Baker's chapter, this volume). However, the only attempts at replication that have been
published (Gould and Able, 1981) failed to support Baker's hypothesis and there has been,
to date, no satisfactory resolution to the controversy.
We were intrigued by the original experiments (Baker, 1980) and at our invitation
Baker came to Cornell University in February 1981 to demonstrate his procedures, prep-
aratory to a series of tests that we planned to do ourselves. The tests Baker performed at
Cornell yielded the most statistically significant results ever reported (see Baker's Tables
III and IV, this volume) and it behooves us, therefore, to examine them and our subsequent
Cornell tests in more detail. In this paper we wish to summarize the Cornell tests and to
offer an alternative, equally supportable explanation for the results that does not invoke
a human compass sense at all. We also wish to report on a Monte-Carlo simulation, using
the sample sizes and number of test stops reported by Baker for all of his English bus
experiments (his Table III, this volume) and his procedures in data analysis. Our simu-
lation, however, rather than using Baker's actual data, employs "random" numbers but
produces statistically significant "orientation" comparable to Baker's results. Our purpose
here is to show that alternative explanations exist for the human bus test results. Unless
the reality of a compass sense in humans can be demonstrated conclusively, it is premature,
we believe, to move beyond this stage to draw any conclusions about its sensory basis.
KRAIG ADLER and CHRIS R. PELKIE • Section of Neurobiology and Behavior, Cornell University,
Ithaca, New York 14853.
573
574 Chapter 29
2. Bus Tests Conducted at Ithaca, New York
Baker (this volume, Table IV) reports on a test conducted at Ithaca in February 1981
which yielded a highly significant result (V-test, p = 0.00006). For the record, although
Baker claims that this "experiment was carried out by Adler ... in collaboration with
myself ... ," the procedures and route were decided by him and the route was unknown
to Adler who was, in fact, a subject in this test. We can confirm Baker's statistical analysis
of this test (see Note 1; although he improperly lumped nonindependent data, as discussed
in Section 3 below), but his demonstration left us critical of the test procedures used and
suspicious that the topography peculiar to the Ithaca area may have contributed to, or
indeed may have been the sole cause of the strongly oriented distributions that Baker
obtained.
Ithaca is nestled in a valley at the southern tip of a long (65 km) lake, and Cornell
University, from which the subjects were drawn, is situated on the east hill overlooking
the lake. Thus, from a starting place on the campus, a bus could not displace subjects west
or south without going down relatively steep hills into the town of Ithaca, an elevational
change of about 215 m. Even if it were possible to choose some circuitous route from
campus to the town which avoided steep grades and thus was not perceived by the subjects
as a change in elevation, it seems plausible to us that subjects may have summarily ex-
cluded the possibility of displacements to the south or west-if they did not detect an
altitudinal change-simply because such a displacement, given their knowledge of Ithaca's
topography, was thought "impossible." (The route Baker chose, in fact, was roughly NNE
of the starting point.) The problem with such a topography is that it severely narrows
the range of possible choices on the part of subjects and, therefore, ought to bias their
choices in the direction of statistical significance (see further discussion in Section 3 and
in Note 2).
We attempted to overcome those procedures that we believed were open to subjective
bias, by altering the protocol in our own experiments. External smells and, especially
sounds had been potential sources of orientation cues for subjects in Baker's demonstration
test. For example, in a small town like Ithaca there are few highways where traffic moves
at high speed; thus, subjects might have noticed traffic sounds during the Baker test and
used them, if not to orient per se, then to reduce the possible choices. In other words, use
of such cues may not allow subjects to precisely determine the compass direction but may
induce clumping in their choices in some other direction by excluding a sector of possible
directions. Therefore, in our tests we played tape-recorded background noise (natural surf
sounds) and sprayed the bus with room freshener during the entire displacement journey,
in an attempt to mask these cues. Even though subjects wore double blindfolds, windows
were covered with opaque paper (except in our first test) to eliminate differential heat cues
due to the sun's location.
To avoid biasing subjects' estimates or the transcription of data, a number of "double-
blind" procedures were instituted. First, all instructions were tape-recorded, from a script
written by us but read by an individual who knew nothing of the displacement journey;
the same tape was used in all of our tests. By tape-recording the requests that were to be
made at ea·ch stop (rather than giving them orally and extemporaneously as Baker had
done), we eliminated the possibility that subconscious changes either in intonation or in
information could occur at the different stops which might be utilized by the subjects.
Second, subjects were allowed to remove their blindfolds at each stop after making their
estimates. We then held up a four-armed sign showing the cardinal compass directions
which a tape-recorded message informed subjects was correctly oriented, unlike Baker
who pointed out and orally identified the compass directions. Third, all data sheets were
transcribed by a trained assistant who had no knowledge of the route or of the treatments.
Human Homing Critique 575
Because one of the pieces of data obtained from the subjects was the direction of an arrow
written by them on their tablet, such transcription might otherwise be open to subjective
bias. Finally, in that experiment in which the subjects had magnets or brasses attached to
the bridge of their noses (near the location of the putative magnetoreceptor, according to
Baker), the two kinds of metal pieces were of equal size and weight and in indistinguishable
casings and were attached by another assistant who had no knowledge of the route and
was not otherwise involved in the test. Thus, the test supervisors did not know which
subjects were the experimentals and could not preferentially bias their responses; moreover,
although the identity of subjects wearing magnets or brasses was confirmed at the end of
the test, their locations were coded so that the assistant transcribing the data did not know
which were the experimentals (see Fig. 1).
With these presumed improvements in procedure, we expected no clumped distri-
butions in our tests, if, in fact, Baker's Ithaca results were merely due to procedural
artifacts. To our surprise, there were more significantly clumped distributions among our
tests than one would expect by chance alone (Table I). We were perplexed, at first, until
an important clue was found in the comments written on the post test questionnaires that
all subjects were requested to fill out. This volunteered information disclosed an unex-
pected approach that was used by some subjects in determining direction and suggested
a hypothesis that is open to objective analysis. These subjects claimed to have treated the
front of the bus (and, therefore, the top of their writing tablets) as north, despite our tape-
recorded pretest instructions in which we attempted to demonstrate that true compass
directions were unrelated to the heading of the bus. The implication of this misinterpre-
tation of instructions is a warning to those who do these tests with humans: even when
presumably clear instructions are given, some subjects may, nevertheless, record their
choices according to a subjective frame of reference of their own choosing.
Using the input from the posttest questionnaires filled out by our subjects, we have
devised a series of alternative mechanisms that individual subjects could have used in
answering questions. Fortunately, it is possible to analyze the actual data with respect to
each of these mechanisms and to determine whether there is any evidence to support them.
Several of these mechanisms do not invoke a human compass sense at all but appear to
be as fully supported by the Cornell data as is Baker's original mechanism. Of course, these
are post hoc analyses and therefore one cannot draw final conclusions from them. We
regard our own tests as pilot studies only, but the hypotheses that are described below
and that are to varying degrees supported by our data, may be used as a priori predictions
on which to base future tests.
The first of the two requests made at each stop was the compass estimate: "Using one
of the eight compass directions, on the tablet write down the direction from here to the
last stop." The eight compass directions were illustrated on an information sheet that was
read by subjects and then collected before the test began. The eight directions were N, NE,
E, SE, S, SW, W, and NW; a compass diagram on the information sheet illustrated their
relationship to one another. The subjects were told to use letters (not degrees) when writing
these directions.
Two different ways that subjects could have interpreted this request are as follows:
Mechanism A. A subject may envision a "bird's-eye view" of his present position on
a map and determine the absolute direction to the last stop in the "real world" (= external
to the bus) frame of reference (see Fig. 2A). Consequently, although the subject was obliged
to face forward on the bus throughout the entire journey, he would not consider the present
heading of the bus when writing his estimate. This was the frame of reference that subjects
were instructed to use in determining compass direction. A short pretest demonstration
of the independence of "direction to the last stop" from "bus heading" was designed to
help subjects visualize this approach in responding to this request. Baker's proposed mag-
576 Chapter 29
3
Mag N
\
lkm
ú =
1 mile
\DISPLACEMENT
- 2
ROUTE
1 START
ú =
BRASSES MAGNETS
êJJJJJJJJJJJJJJJJJúJJJJJJJJJJJJJJúäêI JJJJJJJJJJJJJJJ^^JJJJJJJJJJJJJJJJJJú=
STOP 1 o· STOP 1 o·
Il.
a
I-
III
a
l-
N
Il.
a
Iii MY, 289.3' ; r, 0.24
u,0.4673 ;P.0.3183
MY,344.6'. r, 0.58
u,3.2785 ; P , 0.0007
MY ,306.2'; r, 0.75
u'2.6563 ; p, 0.0044
MY, 39.3'; r, 0.47
u,2.1946 ; p, 0.0146
STOP 2 o· STOP 2 o·
N
Il.
a
Iii
a
I-
M
Il.
a
Iii MY,37.0'; r, 0.11
u,0.5324 ; P, 0.2955
MY, 340.9' ; r, 0.41
u d.3302 ; p, 0.0104
MY'294.2"; r, 0.25
u,0.6193 ; P, 0.2660
MY, 327.8"; r, 0.36
u,1.8082 • P, 0.0354
Figure 1. Orientation of blindfolded human subjects after displacement (Cornell Test 4). The bus route
began on campus at 1, followed by 10 loops and circles in a large parking lot; two requests were made
at both stops 2 and 3. Data are divided into four quadrants, depending upon stop location and magnet!
brass treatment. For example, in the upper left quadrant the requests made were as follows: (1) "Write
down the compass direction from here to the last stop" (stop 1). (A, "write compass letters"; B, "draw
arrow on tablet"); see text for exact wording of these requests. For data shown in the other three
quadrants the identical requests were made; e.g., in D, subjects were asked to draw an arrow on their
tablet pointing to the last stop (now stop 2). For comparison between stops, the expected direction
in each case is rotated to 0°; thus, in A the expected direction (to stop 1) of 0° is actually 249° to
magnetic north. Note that in Table I, however. all group mean vector bearings are recorded relative
to magnetic north = 0° in order that the results from separate tests can be compared.
The distance by road between stops 1 and 2 is 12.7 km (excluding the parking lot) and the di-
rection from stop 2 to stop 1 is 249°; distance between stops 2 and 3 is 7.4 km, direction from stop 3
netically based compass sense is one possible mechanism for determining direction with
respect to an external frame of reference.
Hypothesis A. The angular values of the subjects' compass estimates will be uniformly
distributed with regard to the theoretical direction (= "direction to the last stop"). If the
V-test statistic exceeds a critical value at the p < 0.05 level of significance, the sample is
significantly clustered about the theoretical direction (Batschelet, 1981, p. 59).
Result A. In the Ithaca tests reported in Table I, all four of Baker's test stops, but only
two of the ten Cornell test stops, exhibit significant clustering. Note, however, that not all
of these test stops are independent (see Section 3 for a discussion of problems arising from
nonindependence) and one should not simply tally the number of significant distributions
for a given hypothesis and compare that number to others. Regardless of whether one treats
all 14 test stops as independent or only the six that are truly independent (Baker's test
stop 2 > 1 and the first stop in each of the five Cornell tests), it is clear that the proportion
of stops showing clustering in our Cornell tests is lower than for Baker's Ithaca test. Possibly
this difference may be due to the more rigorous and double-blind procedures used for the
Cornell tests.
Mechanism B. A subject may have a "feeling" (however acquired, possibly via an
ideothetic reference system), but no name for the correct direction to the last stop (Le., "I
know the direction is off to my right, but I don't know whether that direction is north or
east"). Baker (1980, p. 555) has reported that subjects sometimes confuse directions when
attempting to name them. To give a correct name to the perceived direction, the subject
must have some sort of external reference. Under normal circumstances, this frame of
reference may be provided by street signs, or a hand-held compass, etc. In the test design,
nearly all external frames of reference were believed to have been eliminated. However,
one possibly very important frame of reference was the subject's heading at the time he
made the determination. If the front of the bus (= top of the subject's writing tablet) was
taken to be a 0° or a north reference, then the answer would be rotated away from the
correct direction by an angle mathematically equal to the heading of the bus (see Fig. 2B).
The V-test predicted angle is "direction to the last stop minus the bus heading," and
if the V-test probability is < 0.05, then this hypothesis is a valid alternative to Hypo-
thesis A.
Hypothesis B. The angular values of the subjects' compass estimates will be uniformly
distributed with regard to the theoretical direction (= "direction to the last stop minus
bus heading"). If the V-test statistic exceeds a critical value at the p < 0.05 level of sig-
nificance, the sample is significantly clustered about the theoretical direction.
Result B. Data from three of the Ithaca test stops listed in Table I are significantly
clustered when analyzed in this way. It is interesting to note that these same test stops
also show significance according to Hypothesis A. We cannot conclude that either Hy-
pothesis A or B is better supported by comparing the results from only these three stops
to stop 2 is 143°. Double blindfolds were in place during displacement and during questioning. Sta-
tistics given are MV, mean vector (also depicted as arrow inside each circle); r (length of arrow in
circle, a measure of dispersion about the mean); u (V-test statistic); p (absolute probability). In all tests,
the distribution is tested against an expected (but rotated) direction of 0°.
Most of the individual tests exhibit significant orientation (5 of 8). None of the pairwise com-
parisons of magnet group to brass group are significantly different, using the Watson U 2 n, m test; the
most different distributions (B vs. F) have a U2 of 0.15, p > 0.1. Apparently, presence of a bar magnet
(field strength at 10 em was 0.44 G, with north pole up) on the bridge of the nose has no demonstrable
effect when compared to controls wearing a nonmagnetic brass piece of the same size and shape. The
location of magnets was decided in order for them to affect any cranial magnetic field receptor, in-
cluding that proposed by Baker (1981) to be located near the olfactory bulbs.
Table I. Summary and Test Statistics for Ithaca Bus Experiments Q
Baker test Cornell test 1
Test date 810220 810220 810220 810220 820416 820416

Test stop > last stop 2> 1 3> 2 4> 2 5>2 2> 1 3> 2
Direction to the last stop (DLS) 180 243 304 17 210 146
Bus heading (BH) 300 190 275 145 322 188
Number of subjects (N) 30 28 28 29 8 10
Group mean compass score (MCS) 208.9 271.8 306.1 34.8 218.8 45
Mean vector length (r) 0.3 0.39 0.23 0.31 0.34 0.06
• Hypothesis A: compass estimate made relative to magnetic north = 0° (Fig. 2A) (V-test tests MCS against DLS)
Predicted compass estimate (DLS) 180 243 304 17 210 146
V-test probability [Vp(DLS)] 0.0225 0.0052 0.0415 0.0128 0.0871 0.5203
• Hypothesis B: compass estimate made relative to bus front = 0° (Fig. 2B) (V-test tests MCS against DLS - BH)
Predicted compass estimate (DLS - BH) 240 53 29 232 248 318
V-test probability [Vp(DLS - BH)] 0.025 0.9889 0.3288 0.9875 0.1142 0.4944
Group mean arrow scores measured with tablet top 233.1 138.9 153.8 262.1 254.8 205.6
= 0° (MTA)
Group mean arrow scores measured with magnetic 173.1 328.9 68.8 47.1 216.8 33.6
north = 0° (MTA + BH) (=MNA)
• Hypothesis C: arrow estimate (MNA) made relative to magnetic north = 0° (Fig. 2C) (V-test tests MNA against
DLS)
Predicted arrow direction (DLS) 180 243 304 17 210 146
V-test probability [Vp(DLS)] 0.0177 0.4347 0.7817 0.0353 0.158 0.8223
• Hypothesis D: arrow estimate (MNA) made relative to bus front (tablet top) = 0° (Fig. 2D) (V-test tests MNA
against DLS + BH)
Predicted arrow direction (DLS + BH) 120 73 219 162 172 334
V-test probability [Vp(DLS + BH)] 0.1003 0.7117 0.8811 0.812 0.2359 0.1107
Group mean of individual difference angles [each individual difference angle = individual's arrow estimate (NA)
minus individual's compass score (CS) both measured in magnetic north = 0° frame of'referencel
Group mean difference angle (DAC) 302.7 139.4 316.3 109 338 164.9
• Hypothesis E: compass and arrow estimates are made by the subject in a magnetic north = 0° frame of
reference (Fig. 2E) (V-test tests DAC against 0°)
Predicted difference angle (0°) 0 0 0 0 0 0
V-test probability [Vp(OO)] 0.1245 0.B14 0.1462 0.743 0.0666 0.70B7
• Hypothesis F: compass estimate is made by subject in magnetic north = 0° frame of reference but arrow
estimate is drawn by the subject in a bus front = 0° frame of reference (Fig. 2F) (V-test tests DAC against
BH)
Predicted difference angle (BH) 300 190 275 145 322 188
V-test probability [Vp(BH)] 0.0171 0.2277 0.1371 0.0531 0.06 0.3002
a The "Baker Test" (data columns 1-4) corresponds to the "Cornell" row in Baker's Table IV (this volume); see Note
1 for differences between our data and Baker's. For Cornell Test 4, the data for subjects with brasses and magnets
(see Fig. 1) are given in separate columns for each stop. Column headings include test date and an abbreviation
representing the test stop and last stop (= target site); in addition, direction to the last stop, bus heading, and
sample size are given. Sample sizes within individual tests vary because data were useful only in those instances
where subjects gave a legible response to both the compass and arrow questions, since for Hypotheses E and F it
was necessary to have both answers for a given individual.
The table is subdivided into three sections representing separate categories of analyses. In the first section,
compass estimate statistics and two alternative hypotheses (A and B) relating to that question are presented. Next,
the arrow estimate statistics are given, with two alternative hypotheses (C and D) relating to it. In the last section,
the relationship between the subject's compass and arrow scores is analyzed with respect to alternative Hypotheses
E and F.
Consider the first data column (Baker test, stop 2 > 1) as an example. The test date was February 20, 1981; the
estimates were made at stop 2 relative to stop 1, which was 180° (= DLS, or "direction to the last stop") from
stop 2; at stop 2 the bus was parked heading 300° (= BH); and 30 subjects (= N) gave legible answers. The group
mean vector for compass scores (= MCS) and its vector length (= r), here 208.9° and 0.3, respectively, represent
the mean of individual compass scores (Le., if a subject wrote "NE," then his raw score was recorded as 45° to
magnetic north). Hypotheses A and B are described in the text and illustrated in Fig. 2A, B; briefly, this table
gives the compass estimate predicted by each hypothesis and the relevant V-test probability. For Hypothesis A,
the predicted direction was 180° (V-test, p = 0.0225); for Hypothesis B, on the other hand, the predicted direction
was 240° (p = 0.025). For Hypotheses C and D, this table first gives the group mean vector bearing for the arrow
scores, as measured relative to the top of each subject's writing tablet (= MTA, here 233.n, and the group mean
vector bearing for arrow scores as measured relative to magnetic north (= MNA, here 173.1°). To calculate the
Table I. (continued)
Cornell test 2 Cornell test 3 Cornell test 4
820427 820427 820511 820511 820629 820629 820629 820629

2> 1 3> 2 2> 1 3> 2 2 > 1 brass 2> 1 magnet 3 > 2 brass 3> 2 magnet
210 146 249 143 249 249 143 143
322 188 98 322 100 100 321 321
20 20 11 12 18 18 18 18
99.7 247.5 226.7 167.7 178.3 195.2 180 77.2

0.12 0.17 0.37 0.51 0.24 0.75 0.11 0.25
210 146 249 143 249 249 143 143

0.6067 0.5851 0.0515 0.0128 0.3183 0.0044 0.2955 0.266
248 318 151 181 149 149 182 182

0.7467 0.3595 0.3294 0.0086 0.1075 0.0012 0.2507 0.6514
144.6 155.6 153.4 182.9 133.6 188.3 162.9 149.8
106.6 343.6 251.4 144.9 233.6 288.3 123.9 110.8
0.5 0.15 0.5 0.76 0.58 0.47 0.41 0.36
210 146 249 143 249 249 143 143

0.7693 0.8211 0.0108 0.0002 0.0007 0.0146 0.0104 0.0354
172 334 347 105 349 349 104 104

0.0927 0.1708 0.5912 0.0027 0.9282 0.0819 0.0108 0.0174
1.9 218.8 96.4 326 89.5 104.9 294.3 351

0.42 0.22 0.38 0.73 0.56 0.56 0.42 0.12
o o o o o o o o
0.0042 0.8585 0.5798 0.002 0.4878 0.8089 0.1492 0.2345
322 188 98 322 100 100 321 321

0.0207 0.1185 0.0369 0.0003 0.0008 0.0005 0.0129 0.2626
mean angle of the subjects' arrow estimates made in a "bus front = 00" frame of reference (MTA), the top of the
page was assumed to be 00 and the angle of each subject's arrow was measured on the page with a protractor; Le.,
if a subject drew an arrow pointing to the right edge of the tablet, it was scored as 900. At each stop, every.subject's
tablet top was pointed to the bus front; at each stop, the bus heading (BH) was measured with a compass. To calculate
the mean angle of the subjects' arrow estimates in the "real world" frame of reference (MNA), the bus heading was
added to each page arrow angle. For example, if a subject drew an arrow pointed to the right side of his tablet, and
the bus heading at that stop was 90 0, then the subject's arrow was scored as 1800 in the "real world" frame of
reference. Mathematically, it does not matter whether the bus heading is added to each individual score before
calculating the mean, or simply added to the group tablet mean (MTA) afterward. In either case, the same mean
vector length (= r, here 0.27) will result. So that the results of analyses by Hypotheses C and D can be directly
compared, MNA was used as the basis for both analyses; however, the predicted angles for the respective V-tests
are related to the frame of reference in which the subjects were presumed to have drawn their arrows. Hypotheses
C and D are described in the text and illustrated in Fig. 2C, D; the arrow estimates predicted by each hypothesis
(and their V-test probabilities) are also given.
Continuing with Baker test column 1, for Hypotheses E and F the difference angle between each subject's compass
and arrow scores was calculated, where compass and arrow were both measured and rotated by us to the same
frame of reference (Le., relative to magnetic north = 00); the mean of the difference angles for all subjects (= DAC]
here is 302.70, with a mean vector length of 0.27. Hypothesis E predicts a difference angle of 00 (V-test, p = 0.1245).
In Hypothesis F, the difference angle is equal to the bus heading, here 3000 (V-test, p = 0.0171). Hypotheses E and
F are described in the text and illustrated in Fig. 2E, F.
The statistically significant V-tests (p < 0.05) are highlighted in boldface. See text for discussion of individual
test results.
580 Chapter 29
COMPASS ESTIMATES
NORTH NORTH
a· a·
of reference
L . -_ _---+j DLS (90·)
A B
ARROW ESTIMATES
NORTH NORTH
a· a·
1
1
",-I J J J ú a i p E V MK F = ú J J ú a i p =(90·)
ú =
C o
DIFFERENCE ANGLES
(Calculated as Arrow
Estimate minus
ER ROR Compass Estimate) ERROR
o· a·
'tl
CI>
U
'tl
CI>
....
a.
-90· +90· -90·
Calculated as
Circle C
minus Circle A
E F
Figure 2. Illustration of the six alternative hypotheses. In circles A-D, magnetic north (= 0°) is at the
top of each circle; BH is the bus heading and DLS is the direction to the last stop. In this example,
DLS is east (90°) and BH is northeast (45°). In A and B, subjects were asked to write letters for compass
direction on their writing tablets. As illustrated, tablets were aligned with the bus, with their tops
toward the front of the bus. For simplicity we have assumed that the subject has written or drawn
the correct estimate in whichever frame of reference he or she used.
In circle A, the subject used a "real world" frame of reference and wrote the correct compass
direction of DLS on the tablet. In B, the subject regarded the tablet top (= bus front) as north; thus,
that represent truly independent data sets (Baker stop 2 > 1; Cornell test 3, stop 3 > 2 and
test 4, stop 2 > 1 magnet).
The second request made at each stop was the arrow estimate: "Please hold your tablet
with its upper section or bound edge toward the front of the bus. An assistant will check
the alignment. The tablet must be correctly aligned for the data to be meaningful to us.
On the tablet draw an arrow pointing from here to the last stop."
There are at least two ways in which subjects apparently interpreted this seemingly
straightforward request:
Mechanism C. A subject could determine the correct direction from his or her present
position to the last stop and draw an arrow on the tablet that pointed directly to the last
stop. Thus, the subject would make reference not to the bus heading, but to the "real world"
frame of reference (see Fig. 2C); this was the response that we expected subjects to give.
(If the subject had been allowed to stand, turn freely, and point to the last site, presumably
his or her arm would point in the same direction as the arrow the subject had drawn on
the constrained tablet.) Our method for determining the "real world" direction of subjects'
arrow estimates is as follows: Subjects were instructed to hold their tablets with the top
toward the front of the bus when drawing the arrows. While subjects were drawing their
answers, an assistant left the bus and measured the bus's heading with a compass. Tablet
arrows were later measured with a protractor and thus by correcting for the bus heading
at each stop the true compass headings of the tablet arrows were determined.
Hypothesis C. The directions of the arrows drawn by subjects, as measured in a mag-
netic north = 0° frame of reference, are randomly distributed with respect to the theoretical
direction (= "direction to the last stop"). If the V-statistic exceeds a critical value at the
p < 0.05 level of significance, the sample is significantly clustered about the theoretical
direction.
when attempting to name the compass direction of the right front corner of the bus, true east became
northeast in the subject's bus frame of reference.
In circles C and D, subjects were asked to draw an arrow on their writing tablets pointing to DL5.
In C, the subject used a "real world" reference, and drew an arrow that pointed correctly to east (90°).
In D, where the subject regarded the tablet top as subjective north, the arrow was drawn to subjective
east, or true 135° as measured later by us.
In circles E and F, the hypotheses are illustrated in a fundamentally different way. Here, the
difference angle between each subject's arrow and compass estimates is calculated. Therefore, the top
of each circle does not represent magnetic north but zero difference between compass and arrow
estimates. E shows the predicted difference angle in the case where the subject used the same frame
of reference for both determinations. For example, in both circles A and C, the subject used the "real
world" frame of reference; because he or she scored both compass and arrow estimates as east or 90°,
the difference angle is zero (as shown in E). F shows the predicted difference angle in the case where
the subject used two separate frames of reference to make the two estimates. In this example, the
subject used the "real world" reference for the compass estimate (as in circle A, east or 90°) but used
the tablet top as 0° reference when drawing the arrow (as in circle D, the subject drew an arrow pointing
to true 135° as measured by us, although the subject thought the arrow was pointed east). To enable
direct comparison of E and F, all compass and arrow estimates were measured in the "real world"
frame of reference. Thus, in F we calculated the subject's difference angle as arrow (135°) minus
compass (90°) equals difference angle (45°). In this specific case in which the subject used true north
when making the compass estimate, and regarded the bus heading (or tablet top) as north when drawing
the arrow, the predicted difference angle between the two estimates is the angle equal to the bus
heading. Although subjects may use other reference systems, no predictions can be made until such
references are identified. Further details are given in the text.
582 Chapter 29
Result C. Table I shows that, of the 14 test stops, eight of them gave results that were
significantly clustered. It would be more intellectually satisfying to discover that humans
had an ability to point (or draw an arrow) to a desired goal rather than simply name a
compass direction to that goal. Under natural circumstances, the ability merely to name
the direction, even if it is the correct one, is insufficient for goal orientation without the
ability to move off in that direction. From these results, it would appear that subjects were
able to determine goal direction, by drawing an arrow (but see results of Hypothesis F,
below).
Mechanism D. A subject could have the "feeling" that he or she knew the name of
the correct direction (Le., "I know it is east"), but would need an external frame of reference
in order to determine the direction in which to draw the arrow. One possibility for this
external frame of reference would be the front of the bus being taken as 0° or north. Thus, a
subject who thinks the correct direction is east, and assumes the front of the bus is 0°
(north), would draw an arrow pointing to the right side of the bus (see Fig. 2D). It is as if
the subject treats the tablet like most standard maps where north is at the top. Only in the
case where the bus really pointed north would this arrow be correct in the "real world"
frame of reference. The subject's frame of reference and the subject's estimate is rotated
from the "real world" (where 0° = north) in a clockwise direction by an angle equal to
the bus heading. Thus, the V-test predicted direction is mathematically equal to the "di-
rection to the last stop plus bus heading."
Hypothesis D. Subjects' arrows, as measured in a magnetic north = 0° frame of ref-
erence, are distributed randomly with regard to the theoretical direction (= "direction to
the last stop plus bus heading"). If the V-statistic exceeds a critical value at the p < 0.05
level of significance, the sample is significantly clustered about the theoretical direction.
Result D. Only three of the 14 test stops gave results that were significantly clustered
(Table I). It is interesting to note that the same three stops also show clustering with respect
to Hypothesis C.
Recall that the arrow estimate immediately followed the compass estimate in the test-
ing sequence. Therefore, it seems likely that a subject who had just made what he presumed
to be a correct compass estimate would then utilize that information in his choice of an
arrow direction, and choose a direction that is consistent with the earlier compass estimate.
The problem confronting a subject is which frame of reference to use when drawing his
atrow: (1) the "real world" reference where north equals 0°, or (2) a subjective reference
not necessarily related to the "real world." Baker (this volume) presumes that subjects are
utilizing "real world" information, detected magnetically, to fix their frame of reference
for purposes of orientation. This mechanism appeared reasonable to us also, until com-
ments made by some subjects made us consider the alternative mechanism, namely, that
a purely subjective reference system might be used to draw the arrows.
Mechanism E. If the frame of reference that the subject used to draw the arrow is the
"real world," then the arrow estimate should equal the compass estimate that the subject
just made. When the angle of each subject's compass estimate (as measured in a north =
0° frame of reference) is subtracted from that subject's arrow (measured in a north = 0°
frame of reference), it yields a "difference angle" for that subject (see Fig. 2E). We would
expect that the two angles would not differ (Le., the difference angle shduld equal 0° or
zero error; this should not be confused with 0° in a compass sense) in the case where a
subject makes both of his or her estimates agree in the "real world" frame of reference.
Hypothesis E. All subjects' difference angles at any stop should be randomly distrib-
uted with respect to a predicted V-test direction of 0° error. The alternate hypothesis would
state that if subjects' arrow estimates are based on their compass estimates, then the group
mean error (computed from the "arrow minus compass" values for individuals) should
cluster at 0° error.
Result E. Of the 14 test stops reported in Table I, only two show significant clustering
of 0° error but apparently for totally different reasons. Recall that 0° error is achieved when
an individual's compass and arrow estimates are in agreement, but note that this does not
require that the actual direction chosen bears any relationship to the correct direction in
the "real world." For example, if an individual chose east and drew an arrow pointing to
east, then his difference angle is 0°, even if the expected direction in a "real world" frame
of reference was, say, southwest. Thus, if a number of subjects chose directions in this
manner even though there was no agreement among them, the mean of the difference angles
for the group could be significantly clustered around 0° error.
In Cornell test 3, stop 3 > 2, it appears that subjects were able as a group to write the
correct compass estimate (Hypotheses A and B) as well as draw the correct arrow estimate
(Hypotheses C and D); thus, it is not surprising that at this stop the difference errors between
individual compass and arrow estimates were small and clustered around 0° error by Hy-
pothesis E (V-test, p = 0.002). But note that significant clustering also occurs for Cornell
test 2, 2 > 1 (V-test, p = 0.0042) even though there were no significant clusterings under
Hypotheses A-D. In the latter instance, we apparently have a case where most subjects
were individually consistent in choosing compass and arrow directions but these bore no
relationship to the expected "real world" direction or to other subjects' estimates.
In three other cases (Baker test, stops 2 > 1 and 5 > 2; Cornell test 4, stop 2 > 1
magnet), the compass and arrow tests show significant clustering but, under Hypothesis
E, there is no clustering. This result at first seems incomprehensible, until we consider an
alternative mechanism (F).
Mechanism F. If the subject's arrow estimate depends on his or her compass estimate
(as for Mechanism E) but, instead, the subject uses the "bus heading = 0°" as a frame of
reference for drawing the arrow, then a different prediction for the difference angle must
be made. This might well arise if the subject has maintained a "real world" sense of di-
rection in nameable compass terms, but the twists and turns of the bus have confused him
or her so that the subject cannot point in the compass direction that he or she has just
named. Baker (1981a, pp. 68-69) reports an analogous situation in which subjects reported
that they were confused by last-minute turns just before choosing directions.
To analyze the data, we consistently transformed all compass and arrow estimates to
a magnetic north = 0° frame of reference. We added the bus heading angle to the angle
of the arrow as measured on the tablet, to yield a "real world" direction for the arrow. For
example, a subject may have given a compass estimate of east (as in Fig. 2A) but then drew
an arrow on the tablet pointing to the right side of the bus (as in Fig. 2D). The bus heading
is 45°. Consequently, the subject has drawn an arrow that actually points to 135° (southeast)
as viewed from the "real world" frame of reference. The subject has "correctly" drawn an
east arrow in his or her own frame of reference ("bus heading = 0°"), but the difference
angle as scored in our data analysis is equal to arrow (135°) minus compass (90°), or 45°.
One can see from Fig. 2D that this difference angle is simply due to the frame of reference
rotation caused by the bus turning to 45°. As Hypothesis F assumes that a subject would
attempt to maintain an internally consistent frame of reference, then the subject's difference
angle (from his or her point of view) is 0°, but from our point of view is equal mathematically
to the bus heading (see Fig. 2F).
Hypothesis F. The arrow estimate made in a bus front = 0° frame of reference minus
the compass estimate made in a magnetic north = 0° frame of reference (both measured
in the magnetic north = 0° "real world" frame of reference) yields a difference angle for
each individual. Hypothesis F predicts that the individual difference angles will be ran-
domly distributed with respect to "bus heading." The alternative hypothesis states that
the difference angles will be signficantly clustered around an error equal to "bus heading."
Result F. In half of the 14 test stops described in Table I, there is significant clustering
under Hypothesis F. The results from two test stops (Cornell test 2, stop 2 > 1 and test 3,
584 Chapter 29
stop 3 > 2) are equivocal as they also show significant clustering under Hypothesis E. (In
fact, the results from the latter stop apparently support all six hypotheses!) In the data from
the other five test stops, Hypothesis F (but not Hypothesis E) is supported. On close ex-
amination of the columns of data for these five test stops, no clear pattern emerges that
explains why Hypothesis F should be supported in these five instances.
Discussion. In Baker's test stop 2> 1, the compass estimates are significantly clustered
(under both Hypotheses A and B, so the basis for this clustering is equivocal) and the arrow
scores are also clustered with respect to magnetic north. One would expect, therefore,
according to Baker's reasoning, that the two estimates-compass and arrow-are depen-
dent upon the same (external) reference system and these data ought to be significantly
clustered under Hypothesis E. In fact, they are not (V-test, p = 0.1245) and, instead, there
is significant clustering under Hypothesis F (V-test, p = 0.0171). Cornell test 4, stop 2 >
1 magnet shows the identical pattern. These results taken together with those of most of
the Cornell tests, imply that Hypothesis F (which presumes that subjects are using a sub-
jective and not a "real world" frame of reference in making the arrow estimate) may be
correct. We realize that there are other possible permutations of the compass/arrow rela-
tionship that yield other predictions. Our purpose here is not to exhaustively describe
every possible permutation, but merely to point out that some alternative ways of exam-
ining the same data set support alternative hypotheses at least as well as those hypotheses
predicted by the naive point of view that the subjects were making estimates according to
the method we originally expected. There is no easy way to avoid this problem-asking
subjects what they did or how they did it is not a reliable method of collecting this very
important information. Baker himself has pointed out that his subjects were never able to
report reliably how they performed the tasks (Baker, 1981a, p. 37). It is possible, however,
to try to understand how subjects actually worked out the answers to questions by analyzing
the data from the point of view of alternative hypotheses, as we have begun to do.
On the basis of our retrospective analyses, we do not claim to have proven that Baker's
interpretation is wrong. What we hope to have demonstrated is that the Ithaca data (both
Baker's and our own) are open to several equally-supportable interpretations and that future
bUE tests must be designed to distinguish between the alternative hypotheses. Based on
our experience, future tests of this sort must be more carefully designed to take into account
such variables that interact in the V-test as the bus heading, direction to the last stop, and
sample size. While we agree with Gould and Able (1981) that the bus tests per se are
"technically undemanding," no one should misinterpret this to mean that these experi-
ments are either simple to design properly or easy to interpret.
3. Oriented Distributions from "Random" Data
Baker (this volume, Section 3.3.2) rightly points out the statistical abuses that are often
to be found in the orientation literature. Among other criticisms, he comments upon the
improper lumping of nonindependent data. Yet, after taking others to task for their abuses,
Baker himself proceeds to lump data (Baker's Table III, this volume) in ways that are
improper on two grounds. First, in deriving his column headed "Compass estimates" he
lumps data from several stops for those test days where there were two or more stops (23
of 31 days). Second, in calculating the grand totals at the end of his table, Baker lumped
all 31 tests in a manner that violates statistical norms in at least two ways: (1) some subjects
were used in more than one test (see Note 3), so that test days are not independent samples;
and (2) sample sizes vary from 5 to 42 (see Note 4). Statisticians, of course, warn us of
precautions we should take in using particular statistical tests, and with good reason: the
possibility of a Type I error is markedly increased if certain assumptions of the statistical
test are not met (Batschelet, 1981). In the bus experiments in question, lumping of data
from the several stops of a given test would be quite improper (see further discussion
below). Baker himself (this volume, Section 3.3.2) also gives us a perfectly plausible reason,
on biological grounds, for not lumping data from several stops of the same test: "a person's
estimate of home direction at one site must build upon, and be influenced by, estimates
of home direction at previous sites." In addition, it may not be appropriate in doing second-
order analyses to lump together data from single-stop tests with data from multiple-stop
tests in which correct compass information is supplied to subjects at each stop (see
Note 5).
Given the dangers, both statistical and biological, of lumping such data, it occurred
to us that Baker's summary of the bus tests ("a weak but significant ability for goal ori-
entation following blindfolded displacement," Baker, this volume, Section 3.3.3) might,
in fact, be the result of small biases whose influence is magnified when large data sets are
combined. Thus, using a procedure pioneered by Cole (1957), we decided to test this idea
by calculating the result using Baker's sample sizes and arithmetic procedures but begin-
ning with "random" numbers instead of actual data. Cole attributed his "data" to the
unicorn, a mythical animal which, at least according to Cole's analysis, using incorrect
but then-fashionanble procedures, has a perfectly good circadian rhythm!
In the absence of Baker's raw data and to convince ourselves that we were computing
our compass estimates in the same manner as he did, we recalculated his results for several
tests. For example, data from the second test conducted on June 29, 1979 (Baker, this
volume, Table III) were published at least three times previously (Baker, 1980, Fig. 4; 1981a,
Fig. 6.2; 1981b, Fig. 7), and in each instance Baker provided figures from which the raw
data can be reconstructed. Using data from these tests, we calculated the same mean vector
(his eO), vector length (r), and V-test probability given by Baker (this volume, Table III);
we have performed confirmatory recalculations on several other Baker tests. We were also
able to derive the same second- and third-order total values reported in Baker's Table III
and are convinced that in developing our computer simulation, that we have employed
exactly the same procedures used by Baker (see Note 6).
In performing this simulation, however, we chose to alter a single variable that we
have reason to believe might be an important one in actual tests with humans. It is possible
to point to several biasing factors, including some that even Baker himself has pointed out
(see Note 7). Knowledge of local geography (peculiar topography, local road systems, lo-
calizable sounds or odors) or of other external reference cues (such as the sun's heat) could
subtly influence the choices made by subjects and result in a statistically significant dis-
tribution.
In the first phase of the simulation, random "compass scores" in the range of 1-360°
were generated by an Apple Macintosh computer (see Note 8). The same numbers of "sub-
jects" and "stops" per test as in Baker's Table III were used and the "random" numbers
were handled in the same way that Baker treated his data. In each succeeding phase of
our simulation, the arc in which the "random" scores were generated was decremented
by 10°, and all of the same analyses were performed. This was designed to simulate a bias
on the part of subjects to make estimates in restricted arcs somewhat smaller than the full
360°. A total of 110 runs was performed, in a Monte-Carlo simulation. We realize that
restricting the artificially generated "compass scores" to an arc slightly less than a 360°
circle means that our data are not random in a strict mathematical sense, but this small
departure from true randomness was necessary in order to model the effect of a bias. As
mentioned in Section 2, once subjects had made a determination of displacement direction
from the starting point, their choice of directions at subsequent stops may have been de-
pendent on the choice made at stop 1. For example, if a subject thought he or she was
being moved south and thus chose the direction north at stop 1, and the bus were to move
in some direction that did not appear to retrace the original path to stop 1, then the subject
586 Chapter 29
0.05
!P 'j' . ·n l·
J J J J J ú J J J J J J J J J J J J J J J ú J ú J --------------
0.01 J J J
. .,'
ú J J J J J J J J J J J J J J J J
0.001 J J J J J J J ú J ú J J J J J J J J J J
0.0001
c..1
0.00001
>-
!::
...J
iii
«
III
ú= 0.000001
c..
z
o
;:::
u
UJ
-,
UJ
c::
t;;
UJ
l-
I
>
A B
E 0.0000001 - - - - - - - - - - - ;-.. :':...;; ::::: ::::: :::::
360· 3'0· 320· 300· 280· 360· 3'0· 320· 300· 280·
ARC OF BIAS ANGLE ARC OF BIAS ANGLE
Figure 3. Results of Monte-Carlo simulation of human orientation tests depicted in Baker's Table III
but using random numbers in place of real data points. The purpose of the simulation in A was to
examine the effect of lumping biased data on second-order analysis (= "level 2" in Baker's Table III).
The independent variable in this simulation is the arc within which an individual chose each compass
direction. Thus, within a 360° arc the choices of a subject were random, but within a 320° arc possible
estimates were reduced by 40° from random; as shown in the diagrams at top, the arc of possible
directions is centered at 0°. Thus, excluded directions (unshaded areas) are centered at 180°.
Ten runs were done within each of nine arcs ranging from 270 to 350 and 20 were performed at 0
360° (total, 110 runs). Note that one of the runs in the 360° arc is statistically significant (p < 0.05),
and 2 of 10 are significant within the 350° arc (p < 0.05). In an arc of 330°, a majority (7 of 10) of the
runs are significant at p < 0.00001; this rejection level is of interest because it is the one reported by
Baker for his data (this volume, his Table III).
In B, the same 110 sets of data were analyzed by third-order analysis (= "level 3" in Baker's
Table III). By comparison to the results of the second-order analysis (in A), one can see that the third-
order analysis appears to be even more influenced by the multiplicative effects of small biases.
The conclusion one draws from this simulation is that lumping of nonindependent data, as Baker
has done, magnifies any bias in the second- and third-order analyses and that this leads to statistical
significance, even though in our simulation the original "data" were random numbers. See text for
detailed discussions.
might preferentially choose a compass direction at stop 2 that was not completely random
with respect to the choice at stop 1. To put it another way, his or her choice at stop 2
might not be selected from all possible directions in a circle (Le., an arc of 360°) but,
instead, from a somewhat reduced arc of possibilities (Le., an arc of 270° or 300°). If this
were true, to what extent would a small directional bias influence the statistical result
when data from many stops were lumped?
It turns out to have an enormous influence (see Fig. 3). From these graphs it can be
seen that if the choice of direction at a stop is only slightly related to the choice at previous
stops, then the second-order summary results obtained are nearly always statistically sig-
nificant at the 0.05 level; as shown in Fig. 3, the arc of possible choices need be reduced
by only 20°, to a range of "random" possibilities within an arc as large as 340°. When the
arc is as large as 330°, then the level of significance is approximately as great as that
calculated by Baker (p < 0.00001; his Table III) for data from human tests.
The conclusion we draw from these analyses is that Baker's admittedly "weak but
significant" evidence for a human compass response may not result from a compass sense
at all. Instead, equally significant results can be obtained due to the cumulative effects in
large samples of very small biases that subjects can acquire from a variety of external
sources. Clearly, the proof for a human compass sense has not been made beyond any
reasonable level of doubt.
ACKNOWLEDGMENTS. We are indebted to the many Cornell students and staff who participated
as volunteer subjects in our tests; the vast majority of them at the time were freshman
biology students. Our tests of human orientation were supported by the National Science
Foundation (BNS-7924525) and the data analyzed with support from the National Institutes
of Health (NS-19089), both grants to K.A. We thank members of the Cornell Orientation
Group for their advice in designing our own experiments: Irene Brown, John B. Phillips,
Timothy Larkin, and Jerry Waldvogel. We also thank Irene Brown, John D. Crawford,
Thomas Dayton, John B. Phillips, and Timothy Larkin for critiques of the manuscript.
References
Baker, R R, 19aO, Goal orientation by blindfolded humans after long-distance displacement: Possible
involvement of a magnetic sense, Science 210:555-557
Baker, R R, 19a1a, Human Navigation and the Sixth Sense, Hodder & Stoughton, London.
Baker, R R, 19a1b, Man and other vertebrates: A common perspective to migration and navigation,
in Animal Migration (D. J. Aidley, ed.), Cambridge University Press, London, pp. 241-260.
Batschelet, E., 19a1, Circular Statistics in Biology, Academic Press, New York.
Cole, L. C., 1957, Biological clock in the unicorn, Science 125:874-876.
Emlen, S. T., Wiltschko, W., Demong, N. J., Wiltschko, R, and Bergman, S., 1976, Magnetic direction
finding: Evidence for its use in migratory indigo buntings, Science 193:505-508.
Notes
Note 1. Baker first converted each compass estimate from a raw score to an error from
expected, measured in degrees (see his Table IV, this volume). He then calculated a first-
order mean vector for each subject by treating each individual error score (at the four test
stops) as a unit vector, then calculated the mean of the four unit vectors. As all 30 subjects
in Baker's Cornell test gave four usable compass estimates, the number of first-order mean
vectors was 30 (= N). Each first-order mean vector was treated as a unit vector, and the
second-order mean vector was calculated (r, eO, in his Table IV). The V-test probability that
the second-order mean vector was not significantly clustered about 0° error was calculated
[p (V-test), in his Table IV].
In our Table I, we have calculated the mean vector of compass estimates at each stop
on the Baker test. For the purpose of our comparative analysis of compass and arrow
estimates, we could only use data from those subjects who gave both a legible arrow and
588 Chapter 29
a compass score at each stop. Consequently, we did not use data from one or two subjects
at three of the four stops. Some subjects wrote either no arrow, two arrows, or an arrow
without a definite "arrowhead" to indicate the orientation. There was also an occasional
curlicue from a subject who found the task apparently too difficult. (Baker's use of all 30
subjects in his analysis was perfectly proper, as in fact all 30 subjects did give legible
compass responses at all four stops.)
Note 2. The route chosen by Baker was such that the range of choices made by subjects
could have been severely limited. The original plan for Baker's Cornell test was to use two
starting points in an attempt to minimize the problem of subjects giving responses in a
restricted arc, and thus biasing the group mean vector. The subjects had been informed
by Baker at a meeting before the bus journey began that this "two-start" procedure would
be used; in order to have a record for later use in the design of our own tests, we tape-
recorded all of his instructions. However, the choice of test stops may still not have suc-
ceeded in overcoming the biasing effect. The first stop on the Baker trip (called Stop 2 >
1; see Table I for identification of stops) was made at the end of an essentially straight 10-
km stretch of a major highway. Subjects were mostly long-term residents of the Ithaca area
so it was likely that they had knowledge of the peculiar topography and road system of
the region, and this factor alone could have resulted in the highly clustered data obtained
at this site. The original intention was not to stop at that particular point, but the bus
driver missed a planned 90° right turn that would have occurred about 1 km before the
planned first stop. At that time, Baker requested that the bus be halted and compass and
arrow estimates made, rather than doing a turn of 270° and proceeding to the planned stop;
this request was based on Baker's experience that subjects were confused if a sharp turn
was made just before the questions were asked (Baker, 1981a, Fig. 7.4). After the questions
were completed at this provisional stop (Stop 2 > 1), the bus moved to the site originally
designated to have been Stop 2. At this point, everyone removed their blindfolds and the
direction of north was pointed out. This stop was then announced to be the beginning of
the second part of the journey to which each subsequent stop would refer as goal (3 > 2,
4 > 2, 5 > 2). Although these three stops were intended to be distributed within a wide
arc around the goal, in fact they fell within an arc of 134°, and indeed all four stops were
within a 197° arc (see Fig. 3 for the statistical effect of restricting choices to such a small
arc). If subjects had determined that they could not have gone as far west (toward the lake)
as they were taken east, then they might be biased to answer that the goal was generally
westward of the subsequent stops. Even answering northwest or southwest would result
in significantly clustered orientation by the V-test if the majority of the group gave that
response.
Note 3. In animal orientation research it is often necessary to combine results from
several test samples to do second-order analyses, but in doing so, certain conditions must
be met. One of these conditions is that the samples be independent (Batschelet, 1981, pp.
197-198); in the case of bus experiments with humans, this condition would be met if
individual subjects were tested only once. However, in computing his "level 2" and "level
3" summary statistics, Baker (this volume, Table III) has combined results from 31 tests
in which numerous subjects were tested more than once. This fact becomes clear when
one carefully compares Baker's descriptions of individual tests (Baker, 1981a, pp. 14-19)
that constitute the first 16 tests in Table III (Baker, this volume) to Baker's summary of
the 1976-1979 experiments (1981a, p. 21). For example, the four tests listed in Table III,
conducted from October 15 to November 5,1979, correspond to the "Series III Manchester"
series in Baker's summary; in the first, he shows a total sample size of 60, but his summary
clearly indicates that these data were drawn from only 34 individuals. Concerning the
1976-1978 series (10 tests), 76 data points are included in Baker's Table III, but his earlier
summary notes that only 64 subjects were involved; Baker himself (1980, p. 556) points
this out: "A few subjects went on two journeys (in Series I tests)." In addition, based upon
the reported sample sizes, the six tests conducted by Baker in November 1980 (his Table
III) look suspiciously as if the same subjects were used for each of the three pairs of tests.
Clearly, Baker's tests are not independent data sets, and thus he violates one of the fun-
damental assumptions of the statistical tests that he used.
Note 4. A further condition for second-order analyses that needs to be satisfied is that
each test must be "based on the same number of observations" (Batschelet, 1981, p. 198).
In the present instance, this means that in order for the result of tests to be combined to
do a second-order analysis, the sample size in each test must be the same. According to
Batschelet, "small deviations ... can be tolerated. However, large deviations complicate
the statistical analysis considerably." Batschelet presents graphic evidence (1981, Fig.
10.1.1) which shows that even a twofold difference in sample size has a major effect on
the distribution of r values. Baker's sample sizes (his Table III) range from 5 to 42, an
eightfold difference. It is hard to imagine that such a wide range in sample sizes would
qualify as "small" in Batschelet's sense, yet there is no indication in any of Baker's pub-
lications that he is cognizant of this problem or that he adjusts the result according to some
statistically appropriate correction factor. It would seem, therefore, that Baker's "level 2"
and "level 3" analyses violate another statistical assumption.
Note 5. An unanswered question regarding the bus experiment protocol is, what is
the effect on subjects' compass estimates at successive stops, considering the fact that
subjects receive more information about the displacement route by traveling longer and
farther, and also receive updated correct compass information at each stop? While it is
statistically proper to combine a subject's estimates to generate a first-order mean vector,
this statistical lumping presupposes that the subject is answering the same question with-
out receiving updated correct compass information during the course of the tests. A bird
hopping in an orientation cage may make repeated estimates of its migratory direction
based on a changing cue, such as a rotating star pattern; the bird's individual estimates
are lumped together to generate a first-order mean for the night (Emlen et aI., 1976). How-
ever, this lumping would not be statistically valid if the experimenter provided a fixed
orientation cue once an hour for the bird to calibrate against. Are human subjects influenced
when the test supervisor points out true north and its direction agrees exactly with the
subjects' estimate? What do those subjects whose estimates were completely wrong infer
when confronted with this divergent information? Does it cause a sudden adjustment to
a "correct" interpretation of where they must be? In some of Baker's tests, as many as nine
stops were made on a single journey. It seems likely that subjects on a test of this nature
were actually receiving a good deal more information about the displacement than subjects
who were taken to only one stop. Our tests at Cornell used a protocol in which each stop
became the "goal" for the compass estimate made at the next stop; we thought that this
might reduce the additive effects of the multiple stops referring to a single goal. In ret-
rospect, however, we feel that this was not completely successful. It appears that it is very
difficult to not provide additional information to subjects on multiple-stop tests. The most
conservative approach for future bus tests may be to use a single-stop protocol, which
would also make controlling for direction and distance traveled much easier.
Note 6. In the simulation, random "compass scores" were generated as described in
Note 8. In all cases, it was assumed that the "compass scores" had been adjusted to reflect
"error from predicted"; that is, predicted angles had been rotated to coincide. Thus, all
V-tests could conveniently be performed using 0° (no error) as the predicted direction. The
same numbers of "individuals" and "tests" as in Baker's Table III (this volume) were used.
For example, as in Baker's test of November 20, 1976, six individuals each gave three
estimates. The simulation would have generated six sets of three random scores, and com-
bined those bearings, three at a time, to yield six first-order mean vectors. The first-order
vectors were then combined to yield a second-order test mean vector that corresponds to
the data presented by Baker for each of the 31 tests. The V-test probability was calculated
590 Chapter 29
for this second-order mean vector. Following Baker, the 31 test mean vectors would then
be combined in two different ways. First, a "grand mean" vector, referred to as "level 2"
analysis by Baker, was calculated by combining all 414 first-order vectors as if they were
all independent (they were not; see Note 3). Second, the 31 second-order mean vectors
were each assigned a length of 1, and a "level three," or third-order analysis was performed.
V-tests were performed on both the "grand mean" and the third-order results. For instance,
in Fig. 3A, a data point plotted at 300° on the abscissa represents the "grand mean" vector
of a set of 414 first-order vectors. The first-order vectors were calculated from "scores"
that fell in the range of -150° to + 150°, with 0° at the center of the range.
First-order analysis, in which every raw score is treated as if it is independent, is
statistically improper for analyzing the bus data, as Baker points out (this volume, Section
3.3.2). Nevertheless, we also performed a first-order analysis to determine what the effect
of bias would be if the data were mishandled in this manner. The resulting distribution
was not noticeably different from the second-order analysis reported here (see Fig. 3A).
Note 7. The effects of subtle and not-so-subtle environmental cues have been men-
tioned, but Baker did not effectively control for these. As discussed in the text, the Ithaca
area has a distinct topography related to its location in the heart of the Finger Lakes region
of upstate New York, an area sculpted by the last glaciation with lakes and valleys running
generally along a north-south axis. In general, steep slopes are found on the east and west
sides of these prominent geographic features. Thus, people familiar with the area (as in
Baker's test; see Note 2) would be able to narrow their estimates of displacement direction
by utilizing altitudinal information (ears "popping" or kinesthetic feedback), if it was
provided; by contrast, if such information was not detected, their directional choices would
be equally narrowed but to a different area of displacement. Baker's Cornell test route was
chosen to avoid any steep descents. The only relatively flat area large enough for a bus
journey, reachable without a hill climb or descent, is NNE of Cornell campus. As subjects
could have inferred that this was the only possible displacement direction, Baker used a
second "start" after the bus had reached the flat region. Nevertheless, if at the second start
subjects knew where they were, they would also know that Cayuga Lake was only a short
distance to the west, and thus the distances and travel times used in later displacements
would necessarily position them eastward of the second start.
The same initial directional bias may also have played a role in Baker's Manchester
displacements. In Fig. 3.7(b) of Baker (1981a), he illustrates, and states clearly in the cap-
tion, that "all journeys began with a stage 8 km to SSE." One would guess that this route
leads from his university to a roundabout link with the major roadways in the Manchester
area. Were subjects familiar enough with the area so that they could have had a strong
conviction that they were taken southward? In Fig. 3.7(b), it appears that all journeys and
stops (except "E") would fall below a straight line running approximately NE-SW through
the campus. Thus, a subject who sensed that he was being taken south needed only to
answer "north" to score correctly.
Baker reports that some of his subjects claimed to have used the heat of the sun on
their faces (1981a, p. 37), but then states that if subjects can orient without the sun cue,
the sun cue is therefore nonessential. However, he concludes his short discussion on re-
jecting the sun as a useful cue, by admitting that even if the "perception of the sun's position
relative to direction of displacement is not the basis of route-based navigation, ... it may
be used when possible" (1981a, p. 38). The point is that the bus journeys are of such short
duration and extent that the sun, when available, could well be used as a virtually fixed
orientation cue. Only tests done during a short time span around local noon would serve
as an adequate control, as even the blindfolds and window curtains still allow some light
and heat to be detected when the sun is shining. Baker also reports that his subjects claim
to have made some use of industrial odors and airport sounds as directional cues, but
presents no evidence that these cues were either helpful or controlled for in his tests (1981a,
p. 37).
To quote Baker, "There is no doubt that all of the cues mentioned by the students
could provide information useful for route-based navigation" (1981a, p. 37). The fact that
subjects may be either unaware of or unable to describe which cues they may have used
does not foreclose the possibility that they were nevertheless subconsciously using those
(or other) cues. Without adequate controls for all of these cues, including the topographic
awareness of the region (which is almost impossible to eliminate), the bus tests as they
have been performed are not sufficient evidence for a magnetically based orientation sense.
Note 8. Because the "seed" is determined by a finite number of possible values in a
particular hardware location, microcomputers generate pseudorandom number sequences.
The Macintosh offers the capability to seed the random number generator with 65,536
different starting values under program control (in Microsoft BASIC, different positive
random number seeds produce different random number sequences). The program can
read the built-in clock and use the number of seconds since midnight as a constantly
changing seed for the random number generator. In fact, we used this method of "seeding
the timer value" before generating each "score" during every run (1760 random "scores"
were generated per run; that is, in Fig. 3, a total of 110 runs consisting of 193,600 random
"scores" were utilized). Because of the speed of program execution (approximately 7.25
min per run), every single "score" generation did not start with a different seed number;
on average, approximately 125 different seeds were generated per run.
Reply to Baker
Having seen, in manuscript, Robin Baker's response to our critique ("A Summary of
American Data and Interpretations," this volume), we wish to emphasize or clarify several
issues that apparently were not clear to him. We provide readers with these brief comments
so that they can better focus on the substantive points of disagreement which, in large part,
Baker has failed to address.
Bus Tests at Ithaca. Under "Experimental Protocol," Baker comes to an erroneous
conclusion when comparing results from his test performed at Ithaca and our later ones
in which we took more elaborate precautions (the so-called "stringent tests" in Baker's
terminology). Baker claims that "The number of stops showing significant homeward ori-
entation ... is greater in the stringent tests" (emphasis added). True, but the careful reader
will note from our Table I that there were only four stops in Baker's Cornell test while in
our later tests there was a total of ten stops, if one includes all conditions. Baker confuses
absolute numbers with proportions, the latter clearly being the proper method for com-
parison; we had specifically warned against making such an erroneous interpretation (see
our section "Result A"). Thus, using the procedure we recommended of comparing only
the truly independent first-stop results (which Baker, in his "Summary," now agrees is
best), 100% of Baker's test stops (1 of 1, under Hypothesis C in our Table I) but only 60%
(3 of 5 first stops) in our more stringent tests are significantly oriented, the opposite of
Baker's claim. Incidentally, Baker's references here to showing orientation by "pointing"
are apparently a lapsus on his part. In none of our tests were subjects ever asked to point;
doubtless Baker meant to say "by drawing an arrow" (see Hypothesis C). Drawing an arrow
represents an important difference as it forces the subject to think in terms of the tablet
(as in Hypotheses D and F), and not in kinesthetic terms as when allowed to stand, turn,
and point.
In this same section, Baker laments that the homeward orientation demonstrated in
our stringent tests "is lost in a maze of post-hoc analysis." He completely misses the point
592 Chapter 29
of this part of our critique. We freely admitted that our analysis was done post hoc, in an
effort to ask whether other hypotheses-in addition to that originally proposed by Baker-
conceivably could account for the Ithaca results and which could lead to new hypotheses
for later testing. The answer is a resounding "yes," but Baker fails to acknowledge or discuss
these alternative hypotheses. Specifically, Hypothesis F explains the Ithaca results fully
as well as Baker's original hypothesis. As we carefully pointed out in our critique, this
does not prove that Baker's interpretation is wrong. What we are proposing to Baker and
any others who may wish to conduct definitive bus tests in the future is that it is necessary
to deal with multiple hypotheses (rather than a single one) and to design critical experi-
ments to distinguish between them.
Oriented Distributions from "Random" Data. Under "Second-Order Analysis," Baker
states that we were incorrect in assuming that he had lumped data from several stops
whereas he had used second-order analyses instead. We were well aware that Baker had
used second-order analyses (which we had to know, in order to use his exact computational
techniques to model in our Monte-Carlo simulation) and a careful reading of the first
paragraph of this section, as elsewhere in our critique, should make it clear that we regard
second-order analysis as one kind of lumping. Now, second-order techniques are perfectly
proper to use so long as certain conditions are met. Unfortunately, as discussed in Notes
3 and 4 of our critique, Baker's second-order analyses violate statistical norms on at least
two grounds, but we were forced to use these flawed methods of lumping in order to do
our simulation properly. Thus, in his "Summary" rebuttal, Baker says that we-along with
his other critics-agree with him that second-order statistical techniques are proper to use
(and we do, if the conditions are met), but readers should not be misled to infer that we
agree with Baker's second-order methods, which are incorrect.
Turning to our Monte-Carlo simulation of Baker's Manchester data, we wish to make
two points. First, it is not at all essential to that simulation whether blindfolds were re-
moved at each stop or not. In his summary, Baker informs us that removal of blindfolds
at each stop " ... is not part of our procedure at Manchester ... ," but his published writ-
ings are not clear on this point. For example, he reports (1981a, p. 37) that in two tests in
series I, half of the subjects did not have blindfolds, yet in the caption to his Table III (this
volume) we are not told whether these subjects are included there or not. Later in his book
(1981a, p. 76, Fig. 7.11) he reports results from test series I and III at five different stops
(the figure depicts an unblindfolded person pointing). If one compares the total sample
size in this figure (N = 300) to the total for series I and III in his Table III (this volume;
series I equals rows 1-10, series III rows 13-16) where N = 468, one could infer that in
168 instances subjects replaced their blindfolds before going on to additional stops. Even
in his review (this volume, Section 3.3.1) he tells us "Usually, subjects are blindfolded
both throughout the outward journey and while making their estimates" (emphasis added).
It is little wonder that readers like ourselves were confused about Baker's blindfolding
procedures, so that the unequivocal statement in his "Summary" (quoted above) is most
important to have on record.
Second, Baker fails to comment on the central conclusion from the Monte-Carlo sim-
ulation, that small biases introduced into test subjects' choices become magnified when
multiple stops are lumped via second-order analyses and almost always lead to highly
significant results. Now, even if we exclude possible bias due to removal of blindfolds (as
Baker now tells us that blindfolds were not removed), there are still many other factors
such as topography or kinesthetic cues that could introduce such a bias; this was discussed
at length in our Note 7 but ignored by Baker. Of course, it is possible (and Baker would
have us believe) that magnetic cues provide the bias that allows for nonrandom choices,
but what Baker seems unwilling to consider is that some nonmagnetic explanation may,
in fact, be involved. As our simulation shows clearly, a small bias of 30° (that is, reduction
of possible choices from random, or 360°, to a range of random choices within a still very
broad arc of 330°) can produce statistically significant results 70% of the time (see our Fig.
3) when data are lumped from multiple stops by second-order analysis in the manner
employed by Baker.
Conclusion. Thus, Baker's insistence, in his "Concluding Remarks," that the American
data rival his own Manchester data in demonstrating the existence of a magnetically based
navigational ability is simply false. We have not explained away our results with" ... post-
hoc analysis, statistical and arithmetic quibbles ... " as Baker claims. To the contrary, we
have attempted to demonstrate not that Baker's interpretation is wrong, but that other
explanations are possible. Post-hoc analyses are an important way to develop new, testable
hypotheses (like our Hypothesis F) which hopefully will allow us to better understand
how humans navigate, but in doing these analyses we must employ statistically proper
techniques lest our efforts be plagued with artifact.
There is no question that the phenomenon of human navigation, as championed by
Baker, is intriguing and worthy of further study. However, we will do this study a disservice
if we constrain our thinking to only a single mechanism or employ improper experimental
or statistical techniques in pursuit of an answer.
Chapter 30
Absence of Human Homing Ability as
Measured by Displacement
Experiments
JAMES 1. GOULD
Baker asserts that an individual's direction estimates made at separate stops are not in-
dependent, and therefore must be collapsed. It is certainly true that such estimates might
not be entirely independent, but Dayton's chapter points out that there are major costs
incurred (in terms of potentially useful data lost) when this sort of second-order analysis
is invoked. We (Gould and Able, 1981) chose to present first-order analyses because (1)
Baker (1980a,b, 1981) had occasionally done so, and (2) an analysis of the actual degree
of independence between subsequent estimates (as opposed to speculation) convinced us
that the data were largely independent, and so too much information would be lost if we
blindly applied second-order analysis. (Our original data and first-order analyses are given
in Fig. 1 and Table 1.) The logic of our examination of the data for independence was as
follows: if estimates were wholly dependent on the previous estimate of direction, they
would differ by 0 on average; if, on the other hand, estimates were wholly independent,
0
the subsequent estimate would differ by ± 90 0 on average. The data on the difference
between subsequent estimates of control subjects are shown in Table II: the average dif-
ference for all judgments of direction was 83.6 0 , a value much closer to the prediction
based on total independence of the data. The reader, however, may accept the arguments
for second-order analysis; as Table III indicates, there is no significance at this level either.
Beyond the question of whether the Princeton data are best analyzed by first-order or
second-order techniques, Baker's analysis of the Princeton bus-displacement experiments
(Gould and Able, 1981) is seriously mistaken in two ways as Dayton's paper points out:
the statistical techniques he employed are frequently inappropriate, and where the correct
technique is used, his computations are incorrect. In addition, there are other potential
errors, apparent misinterpretations, and misleading implications which tend to weaken or
contradict his conclusions. Perhaps the most fundamental is the magnitude of the effect
he reports. Even before he began running experiments at "optimum" times of day, or
selecting subjects on the basis of their clothing, sleeping preferences, use of nosedrops,
and so on, Baker found an astonishing strong effect. The average homeward component
of the mean vectors reported in his book (Baker, 1981) are in the range of 0.49, while mean
angular errors average only 13.5". Even well-trained homing pigeons rarely do this well.
Even if Baker's various manipulations of the Princeton/Albany bus data were correct, the
size of the alleged effect would be negligible compared to his: the homeward component
of the mean vector is 0.01, while the mean angular error is 96.4 0 (90 0 is expected by chance).
JAMES L. GOULD • Department of Biology, Princeton University, Princeton, New Jersey 08544.
595
596 Chapter 30
Figure 1. The 209 individual estimates of homeward direction in the first four Princeton experiments
are indicated by dots on the periphery of the 13 circles, each circle representing one "release" site.
The routes to the sites are also shown. The dotted line in each circle is the homeward bearing, while
the arrow is the mean vector of the estimates; e is the angular difference (that is, error) between the
mean vector and the homeward direction; r is the length of the vector; and the asterisk indicates
statistical significance at the 5% level. All statistics are Rayleigh tests. By the modified Rayleigh V
test-a statistical procedure which takes into account the expected direction-no site was significant
at the 5% level.
Readers may wish to entertain the possibility that any magnetic orientation ability in hu-
mans-or at least Americans-is too weak to be of any adaptive significance.
Readers should also be aware that, although the Princeton and Albany experiments
were attempted replications of Baker's work, they were in some respects better controlled.
For instance, Baker refers to his routes as "extremely circuitous," but even a quick com-
parison of the various routes in Baker's book with those in Gould and Able (1981) reveals
that Baker's routes are comparatively simple. When conducting our joint experiments in
Princeton, in fact, Baker insisted on simplifying the routes, though we insisted on retaining
the early "circling" of the vans. The blindfolds used also deserve some comment. Baker
used so-called sleep blindfolds, which are certainly adequate for people trying to go to
sleep in a dark room, but which allow a relatively clear view of the floor because of the
Absence of Human Homing 597
Table I. Summary of Tests for Direction Finding

Statistical
Mean vector
significanced
Location Q
Treatmentb N Errore Length z-test V-test
A Millstone (20 km NNE) C 20 _78° 0.24 n.s. n.s.

B Hillsborough (18 km N) C 20 12° 0.30 n.s. 0.5
C Reaville (20 km NW) C 20 -145° 0.18 n.s. n.s.
D Linvale (23 km WNW) C 20 -176° 0.05 n.s. n.s.
E Kingston (4 km NE) C 19 3° 0.26 n.s. n.s.
F Pennington (18 km W) C 19 _34° 0.16 n.s. n.s.
G Ringoes (25 km WNW) C 19 54° 0.29 n.s. n.s.
H Flemington (26 km NW) C 19 149° 0.14 n.s. n.s.
Croton (33 km WNW) C 9 -147° 0.16 n.s. n.s.
Rosedale (5 km N) C 12 -166° 0.12 n.s. n.s.
M 14 -140° 0.08 n.s. n.s.
K Rocky Hill (16 km NNE) C 12 40° 0.31 n.s. n.s.
M 14 -144° 0.30 n.s. n.s.
L Belle Mead (22 km NNE) C 12 22° 0.23 n.s. n.s.
M 14 128° 0.35 n.s. n.s.
M Woods Tavern (27km NNE) C 12 _49° 0.27 n.s. n.s.
M 14 56° 0.19 n.s. n.s.
° Distances are those from the departure site.
b C, control; M, magnet.
C Relative to expected direction: home in the case of controls, and 180· from home for subjects wearing magnets.
d z is the Rayleigh test with a 0.05 criterion; V is the modified Rayleigh test with a 5% criterion.
difficult-to-follow and varied shapes of noses; and, after dark adaptation, subjects can even
see through the blindfold itself. The Princeton blindfolds consisted of two layers of black
velour and a layer of black felt, both extending well below the nose. We did not feel that
even this was adequate, and so we also asked our subjects to keep their eyes closed, an
instruction Baker does not report giving (and did not give when he ran two experiments
in Princeton). We always closed and covered our bus windows, whereas Baker reports
taking this precaution only occasionally. The Albany experiments used equally effective
blindfolds and hoods, and in any case were run at night, thereby minimizing visual cues.
Judging from the way Baker ran two Princeton experiments, there are at least three
other differences: we always used a written protocol, refused to answer questions except
Table II. Examination of Princeton Bus Data for Independence

No. of No. of Average difference between
Test data subjects sites subsequent estimates
Oct. 30, 1980 15 5 74.6°
Oct. 31, 1980 8 6 92.8°
Nov. 24, 1980 11 5 91.3°
Feb. 16, 1981 20 4 80.3°
Feb. 17, 1981 19 4/5 82.0°
Feb. 18, 1981 11 4 85.9°
Weighted average 83.6°
598 Chapter 30
Table III. Second-Order Analysis of Princeton Bus

Data
Experiment Sites Second-order statistics
May 2, 1980 1 N.A.
Oct. 30, 1980 2-7 >0.1
Oct. 31, 1980 8-13 >0.1
Nov. 24, 1980 9-13 >0.1
Combined 2-13 >0.1
Feb. 16, 1981 A-D >0.1
Feb. 17, 1981 E-H/I >0.1
Feb. 18, 1981 J-M
Controls >0.1
Magnets >0.1
Combined A-M >0.1
Combined 2-13; A-M >0.1
by rereading the relevant part of the protocol, and did not issue instructions while either
pointing in the direction chosen for the journey or at the equivalent part of a large compass
card used to explain how to write directions down. From the photographs in Baker's book
and a film he showed at his seminar in Princeton, we can add further to this list of dif-
ferences-for example, we did not take our subjects to the top of a campus building for a
look at the surrounding topography (since, obviously, we wished to minimize the use of
topographic cues).
Baker's discussion of other bus experiments contains several errors of which the reader
may wish to be aware. First, he claims that only the V- or z-test is correct, though he does
not say which. Now the z-test simply looks for any nonrandom orientation at a site, and
so is to be preferred when there is no particular expected direction, as when subjects wear
magnets. Baker, for example, applies this statistic to claim, post hoc, that Sherpa vans
substantially rotate compass estimates (Baker, 1981). Then too there are strong release-site
biases in some animals. The V-test, by comparison, is more forgiving of short mean vectors
but requires orientation to be in a predicted direction. We presented both statistics, as we
did not wish to create the impression that we might have selected our statistical techniques
post hoc to support our conclusions. The z-test results are summarized in Fig. 1 and Table
I; the V-test results are [as stated in the text and caption in Gould and Able (1981)] in-
significant in all cases for Fig. 1 and are listed in Table 1.
Baker chooses to cite data reported in newspapers and lay magazines, but the reader
may wish to refer to reviewed articles. For example, Psychology Today, which published
an article by Baker (1980a) and a preliminary report of one Princeton experiment (Gould,
1980), misplaced one data point in my figure, leading to the ominous "inconsistency"
cited by Baker. Readers comparing that diagram with the one in Gould and Able (1981)
will see that the critical measure-the mean vector, which they did not redraw-is the
same in both. Baker's problems with calculating mean vectors have already been discussed
by Dayton, and so nothing further need be said about the other contradictions between
Gould and Able (1981) and the allegedly "correct" figure computed in Baker's article. The
calculations are not particularly difficult, and the reader is invited to try them and compare
his or her answer with the published values. Alternatively, the reader may consult Dayton's
chapter where the computations appear to have been carefully made from the original data
cards marked by subjects. I find Dayton's conclusions convincing, and congruent with
Absence of Human Homing 599
those of Gould and Able (1981). Xeroxes of the original data cards, provided to both Baker
and Dayton in 1981, are available to any interested scientist.
The reader should also not be misled by Baker's statement that only Series I compared
magnets with controls; indeed, Baker was present at another such experiment, reported in
Table I of Gould and Able (1981). Baker's claim that there was significant orientation among
Princeton subjects wearing magnets is incorrect. The data for the four sites used in the
joint test, for example, were r = 0.08, a = -140°; r = 0.30, a = -144°; r = 0.35, a =
128°; r = 0.19, a = -49°. The reader can readily imagine that there is no consistent pattern
here. The correct, second-order, statistically nonsignificant values are presented in Table
III and by Dayton.
Finally, it is probably worth emphasizing the danger of post hoc analyses once again.
When statistical techniques and hypotheses are chosen post hoc, there are significant risks.
Baker shows a fondness for this practice by citing an article by Srivastava and Saxena
(1980) reporting three post hoc correlations. It should be obvious to most readers that if
we accept the 0.05 level as a statistical criterion, about 1 in every 20 tests will yield a
spurious correlation. Hence, if we try correlating, say, three different magnetic parameters
against 20 different kinds of sickness or accident, we will get three spurious correlations.
Similarly, if we try the same data with different statistical techniques for variety, the odds
for a spurious correlation are improved. Indeed, if subjects are subdivided by time of
testing, direction of sleeping, and a variety of personal preferences and habits, spurious
correlations are inevitable. The only proper course is to treat such data as defining a hy-
pothesis which is then to be rigorously tested using the same methods.
To summarize, Baker's calculations appear to be incorrect and potentially misleading;
his techniques are sometimes questionable; and his results cannot be replicated. I conclude,
therefore, that humans probably lack any significant ability to sense direction magnetically.
Note Added in Proof

An extreme series of attempted replications by G. W. M. Westby and K. J. Partridge
(personal communication; paper in review) based on 450 tests and incorporating many of
Baker's initial criticisms has also failed to turn up any evidence for homing ability in
humans.
References
Baker, R R, 1980a, Homing instincts of humans, Psychol. Today 14:61-72.
Baker, R R, 1980b, Goal orientation by blindfolded humans after long-distance displacements: Pos-
sible involvement of a magnetic sense, Science 210:555-557.
Baker, R. R, 1981, Human Navigation and the Sixth Sense, Hodder & Stoughton, London.
Gould, J. L., 1980, Homing in on the home front, Psychol. Today 14:62-71.
Gould, J. L., Able, K. P., 1981, Human homing: An elusive phenomenon, Science 212:1061-1063.
Srivastava, B. J., and Saxena, S., 1980, Geomagnetic-biological correlations: Some new results, Indian
J. Radio Space Phys. 9:121-126.
Chapter 31
A Study of the Homeward Orientation
of Visually Handicapped Humans
TIMOTHY K. JUDGE
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ., 601
2. Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 601
3. Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 601
4. Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 602
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 603
1. Introduction
Homing ability after displacement to a release site has been reported for many species.
Recent experiments have also shown this ability in humans who have been displaced
(Baker, 1980). However, the results of these experiments have not always been reproducible
(Gould and Able, 1981). This paper describes a displacement experiment using visually
handicapped humans.
2. Methods
Nine volunteers from the Northeastern Association of the Blind, Albany, New York,
were blindfolded and hooded. All of the test subjects were taken by van, in which the
windows had been covered with aluminum foil, over a 43.5-km route composed of city
streets and county highways that had many curves and sharp turns. Six of the nine vol-
unteers were partially sighted, two were congenitally blind, and one had totally normal
vision. At the first four stops along the route, written estimates of the homeward direction
were asked of the test subjects. At the fifth and final stop, written estimates were again
taken and the volunteers were removed from the van one by one and asked to point in the
direction of the home site, the Northeastern Association of the Blind at Albany. All of the
data were subjected to statistical tests as described by Batschelet (1981).
3. Results
Written estimates of the homeward direction were analyzed using second-order sta-
tistics. This involved taking the deviation of the written estimate from the homeward
TIMOTHY K. JUDGE • Department of Biological Sciences, State University of New York, Albany,
New York 12222.
601
602 Chapter 31
•r=.168 o
r=.159
z=.228 z=.228
0= 106 0 0= 1210
Figure 1. Solid circles are the mean errors from homeward of the
written estimates for each individual across all stops. Open circles
are pointing estimates of the homeward direction. The difference
in the sample sizes is due to the difficulty in determining the
written estimates of one of the test subjects. r is the vector length,
z is the score from Rayleigh's z-test of uniformity, a is the mean
direction.
direction of each stop for each individual using the procedure of Baker (this volume).
These values were then pooled and a mean direction and vector length (r) were computed
from the five values. Estimates of the homeward direction by pointing were analyzed using
first-order statistics to determine the mean direction and vector length.
Neither the pooled written nor pointing estimates of the homeward direction showed
significant orientation (Fig. 1). Written estimates of the homeward direction yielded a mean
of 1060 with a vector length of r = 0.168. Estimates of the homeward direction from pointing
yielded a mean of 121 0 with a vector length of r = 0.159. Neither was significant using
the Rayleigh test or V-test.
4. Discussion
Recent experiments have alluded to the ability of humans to use the earth's magnetic
field in orienting when displaced to a release site (Baker, this volume). In Baker's exper-
iments, test subjects showed statistically significant homeward orientation. However, when
these experiments were repeated in the United States, homeward orientation was not al-
ways obtained (Gould and Able, 1981).
Baker argues that humans possess the ability to detect changes in the geomagnetic
field and that they can use these changes in navigating. This experiment followed Baker's
(1980) protocol, but used visually handicapped humans. If there is a magnetic compass
sense in humans, it might be more pronounced in individuals who are deprived of visual
cues. As with many of the previous experiments conducted in the United States, significant
homeward orientation was not obtained. This may be due, in part, to the small sample
size; however, this was unavoidable due to the number of visually handicapped humans
in the general population.
ACKNOWLEDGMENTS. I wish to thank the clients and staff of the Northeastern Association of
the Blind at Albany for their support and cooperation in conducting this experiment, and
Drs. K. P. Able and H. M. Judge for their support and guidance.
Sightless Human Orientation 603
References
Chapter 32
An Attempt to Replicate the Spinning
Chair Experiment
JOSEPH L. KIRSCHVINK, KARLA A. PETERSON,
MICHAEL CHWE, PAUL FILMER, and BRENDA RODER
Baker's (this volume) spinning chair experiment is an attractive approach for testing the
hypothesis of human magnetoreception. It requires only a few students at a time rather
than busloads, and can be run at a low level over a suitably long period of time. It is further
amenable to a variety of double-blind controls over the magnetic field and subject positions
which are difficult to achieve in the bus experiments.
We ran a modified version of this experiment during the winter and spring of 1982
using 10 undergraduate students at the California Institute of Technology. On the advice
of Dr. Baker, subjects were initially screened through a questionnaire, and only N-sleepers
who did not regularly use stereo headphones or public transportation facilities were in-
cluded. Our spinning chair was built of wood like that of Baker, but differed in that it
spun smoothly on a nonmagnetic, stainless-steel bearing system rather than on four wheels.
Following the recommendation of magician James Randi (personal communication), we
used opaqued swimming goggles as a blindfold and built nonmagnetic earmuffs using
acoustic fiberboard insulation.
Experiments were conducted in a large basement chamber underneath a student dor-
mitory. The location was chosen primarily for its close proximity to the subjects and for
the relatively low local magnetic gradients. Large field values as monitored with a fluxgate
magnetometer were only observed near water pipes in the 3-m-high ceiling but no meas-
urable change was observed within the area of the spinning chair.
Rather than changing the field through a subject's head with strapped-on bar magnets
as Dr. Baker did, we used two large pairs of square (- 2-m diameter) coils to deflect the
horizontal component of the geomagnetic field. The coils formed a cube with one pair
aligned along magnetic north-south which could completely null or reverse the horizontal
component around the subject's head, and another which could produce an east or west
component of equal intensity. Power was fed to each pair of coils through a long cable
which ran up through a narrow conduit hole in the cement ceiling, ran 15 m down a hallway
in the overlying student dormitory and to the power supply which was located in a small
student library. The circuits were controlled by two switches in the library wired so as to
yield either N, S, E, or W resultant fields in the experimental chamber, and pairs of even/
odd random numbers read from a table were used to select these directions with equal
probability. Although the basement experimenter could signal the library with a small
buzzer to indicate the start of a spinning trial and the need for a new random field setting,
JOSEPH L. KIRSCHVINK, KARLA A. PETERSON, MICHAEL CHWE, PAUL FILMER, and BRENDA
RODER • Division of Geological and Planetary Sciences, California Institute of Technology, Pasa-
dena, California, 91125.
605
606 Chapter 32
w t 27121 9121
cD> • 191
cD> • 195
R •• 11196
R •• 1476
A 5 B 18121
Pointing Estlmotes Relotlve Pointing Estlmotes Relotlve
to True North to Mognetlc North
Figure 1. Results of the Caltch spinning chair experiment. Each dot is the average direction from one
trail of eight spins. In (a) the directions are measured with respect to geomagnetic (true) north within
the basement room, and those in (b) are with respect to the direction of the field during each spin.
the experiment was double-blind, as the "librarian" had no means to signal the basement.
New field settings were changed slowly over a period of about 4-5 sec to avoid spark
transients or motions in the coils.
Experimental sessions consisted of eight consecutive trials for each subject. At the
start of each session, the subjects were shown magnetic north with the power to the coils
turned off, and then the blindfolds and earmuffs were positioned. As described by Baker,
the experimenter always remained behind the subject and slowly rotated the chair in ar-
bitrary directions, stopping randomly in one of eight directions (N, NE, E, ... ). Stopping
directions were chosen with even/odd groups of three digits from random number tables,
and a light tap on the shoulder by the experimenter was used to signal the subject for a
response. Subjects were instructed to first point in the direction they thought was north,
and then state the direction in which they thought they were facing. Although this double
response gives the same estimate of where the subjects thought north was, we chose to do
it this way as a check on the consistency of their spatial orientation.
All subjects quickly learned the spatial orientations and there was usually little, if
any, difference between the direction estimates from the two responses. After each session
of eight spins had been completed, the responses were compared with the record of mag-
netic field directions from the library. Results were analyzed both with respect to true
geographic north (ignoring any magnetic changes) and by using experimentally altered
magnetic north.
It is clear that each successive directional response within a trial of eight spins depends
at least somewhat on the direction in which the subject thought he or she was facing before
the spin began. For this reason, we have used the vector average of all eight responses as
independent estimates of the orientation accuracy rather than those from each spinning
response (see Dayton, this volume). Figure 1 shows these second-order results for the
...,
::r
CD
r:n
'0
S·
::J
S·
ao
(")
::r
e:.... .
Q
Table I. Second-Order Results Grouped According to Subjects from the Caltech Spinning Chair Experiments
Estimates relative to room Estimates relative to magnetic field
No. of 8-
spin trial Pointing Verbal Pointing Verbal
Subject sessions azimuth r p azimuth r p azimuth r p azimuth r p
A 13 219/0.2545 >0.10 224/0.2672 >0.10 176/0.1519 >0.10 169/0.1468 >0.10

B 11 237/0.2165 >0.10 225/0.2508 >0.10 52/0.2669 >0.10 121/0.2654 >0.10
C 10 30/0.2089 >0.10 359/0.1923 >0.10 202/0.2950 >0.10 200/0.3662 >0.10
D 9 125/0.5526 >0.05 122/0.5787 >0.025 91/0.4633 >0.10 96/0.4178 >0.10
E 8 34/0.3565 >0.10 32/0.3765 >0.10 242/0.5559 >0.05 251/0.4307 >0.10
F 7 250/0.3668 >0.10 243/0.4250 >0.10 271/0.4310 >0.10 265/0.4813 >0.10
G 6 236/0.2711 >0.10 208/0.1358 >0.10 202/0.3581 >0.10 207/0.4861 >0.10
a Three subjects who had too few sessions (2. 1, and 1, respectively) have not been listed although they are included in the 68 trial sessions plotted in Fig. 1. Significance
levels for the Rayleigh test of randomness are given under the columns labeled "p" for the numbers of trials and resultant values shown. Only subject D had marginally
significant responses relative to the room, and only subject E gave one marginally significant result relative to the magnetic field.
=
o
-...]
608 Chapter 32
pointing responses grouped relative to geographic north and experimental magnetic north.
Neither grouping shows a significant orientation toward any direction (Rayleigh test of
randomness p < 0.10), although there is a weak tendency in both cases for the residual
vector to point south. Second-order responses grouped according to individuals (Table I)
also show only one subject out of seven with a significant orientation, and that direction
is southeast!
Although our results contradict those of Baker, there are several procedural differences
between our experiments and his which might conceivably have influenced our results.
These include: (1) our subjects were spun for a longer time and through wider arcs than
were Baker's. (2) Most of our experiments were conducted in the early afternoon and
evenings, rather than from 10 a.m. to 3 p.m. (3) Our subjects were permitted to wear their
clothing, and we did not check their undergarments for the synthetic fiber or silk content.
(4) Finally, we ran the experiment in the basement of a dormitory rather than an isolated
wooden shed. We believe these differences to be of iittle or no importance, and tentatively
conclude that Caltech students and perhaps humans in general lack any strong or useful
ability to sense magnetic direction. We encourage others to try this experiment.
Chapter 33
A Cautionary Note on
Magnetoreception in Dowsers
JOSEPH L. KIRSCHVINK
In his discussion concerning human magnetoreception, Baker (this volume) cites the work
of Harvalik (1978) on human dowsers as an example of prior work suggesting a magnetic
sense in humans. This is a poor example. Harvalik claims to have measured the sensitivity
of his subjects by passing direct electric current through the ground between the copper
electrodes separated by about 20 m. With the current on, the dowsers walked between the
electrodes and gave signals and claimed to detect the small field present. In one publication
(Harvalik, 1975), he built and used a "randomizer" to make things double-blind. In several
other papers published in The American Dowser (Harvalik, 1973, 1974,1976), he describes
localizing the site of the receptors in the pineal and adrenal glands and measuring field
sensitivities down to an astounding 0.00001 "I (10- 14 Tesla).
In June 1979, I ran a duplicate of this experiment at Princeton University with a local
dowser who occasionally "helped" the local water company find lost pipes. Prior to the
experiment, the dowser located an area which gave him no "response" and we used it for
the experiment. In this area, we used an 80-mA ground current between two copper stakes
separated by 20 m which was controlled by a student in a distant building and switched
on and off with equal probability using a random number table (polarity remained con-
stant). We asked the dowser to tell us whether the current was on or off, and allowed him
to "examine" the area with the current on before the series of double-blind trials began.
We ran two series of 10 trials each, with 5 min spacing between each dowsing attempt as
suggested by Harvalik (1973). Although all responses given were clear-cut (either "yes"
or "no"), only 13 of the 20 were later found to be correct (p ;> 0.05 two-tailed on the
binomial distribution). Despite the small number of trails, our failure rate (35%) was far
higher than the 5% (33/694) reported by Harvalik (1978) for the same experimental setup.
The most serious problem with the work of Harvalik, however, is the claim of a
0.00001-"1 (10 IT) sensitivity for one dowser mentioned above. This is 6 orders of magnitude
smaller than that inferred for the homing pigeon or honeybee. The analyses of Kirschvink
and Gould (1981), Yorke (1981), and Kirschvink and Walker (this volume) imply that they
would need roughly 1012 times more magnetoreceptors than would a bird. If based on
magnetite, the Fea04 in them would weigh a total of about 10 kg, more than the pineal
and adrenal glands combined. Although the SQUID magnetometers described by Fuller et
01. (this volume) have sensitivities in this range, it is physically impractical to use them
to detect fields in the femtotesla range without some form of magnetic shielding to elim-
inate geomagnetic noise; the dowsers clearly did not have this.
Randi (1980) describes a much more thorough experiment on Italian dowsers looking
for water under controlled, double-blind conditions. None were successful. One is com-
JOSEPH L. KIRSCHVINK • Division of Geological and Planetary Sciences, California Institute of

Technology, Pasadena, California 91125 .
609
610 Chapter 33
pelled to view the claims of dowsers with skepticism until a series of well-controlled and
reproducible experiments shows otherwise.
References
Harvalik, Z. v., 1973, Sensitivity tests on a dowser exposed to artifical dc magnetic fields, The Amer-
ican Dowser 13:85-87.
Harvalik, Z. V., 1974, Locating the dowsing sensors by the high-frequency beam method, The American
Dowser 14:4-9.
Harvalik, Z. V., 1975, The randomizer, The American Dowser 15:19-21.
Harvalik, Z. V., 1976, Locating the dowsing sensor-processor in the Brain, The American Dowser
16:106-108.
Harvalik, Z. V., 1978, Anatomical localization of human detection of weak electromagnetic radiation:
Experiments with dowsers, Physiol. Chern. Phys. 10:525-534.
in animals, BioSysterns 13:181-201.
Randi, J., 1980, Flirn-Flarn! The Truth about Unicorns, Parapsychology and Other Delusions, Lip-
pincott & Crowell, New York.
Yorke, E. D., 1981, Sensitivity of pigeons to small magnetic field variations, J. Theor. Biol. 89:533-
537.
Chapter 34
Human Navigation
A Summary of American Data and
Interpretations
R. ROBIN BAKER
1. The American Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 611

2. The American Criticisms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 614
2.1. Second-Order Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 614
2.2. Post-Hoc Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 615
2.3. Experimental Protocol. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 616
2.4. Influence of Magnets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 619
2.5. Accuracy of Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 620
2.6. Anonymous Comments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 620
3. Concluding Remarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 621
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 621
This paper is divided into two parts. The first summarizes American data on human nav-
igation; the second evaluates criticisms of comparable data from Britain.
When collected together, American data overwhelmingly replicate their British fore-
runners. In stark contrast, the interpretation of their own data by American authors is
universally negative. The reader is invited to draw his own conclusion from this contra-
diction.
1. The American Data

The collected papers by Gould (1980), Gould and Able (1981), Able and Gergits (this
volume), Adler and Pelkie (this volume), Dayton (this volume), and Judge (this volume)
present and analyze data from 18 bus experiments performed on American soil. Attempts
to estimate the compass direction of home after blindfolded displacement have been pub-
lished for 17 such experiments; attempts to point, or draw an arrow, toward home for only
12. To this latter, I can add an analysis for one of the collaborative Princeton experiments
(February 16, 1981), making 13 in all. The data for all of these experiments are collected
together in Table I.
Differences in analysis of the collaborative Princeton experiments presented by Dayton
(this volume) and myself (Baker, this volume) are discussed in the next section. To be
R. ROBIN BAKER • Department of Zoology, University of Manchester, Manchester M13 9PL, United
Kingdom.
611
612 Cha] )ter 34
Table I. Summary of American Bus Experiments
Mean vector
N eO ± CI r p (V-tesW Author(s)
Compass estimates
Princeton-l
Controls 50 -107 ± 55 0.233 0.758 Dayton (this volume)
Magnets 20 -6 ± 66 0.335 1.8 x 10- 2 Dayton (this volume)
Princeton-2
Controls 45 19 0.144 0.098 Dayton (this volume)
Magnets 15 13 0.213 0.126 Dayton (this volume)
Albany
Controls 45 38 ± 42 0.289 1.6 x 10- 2 Able and Gergits
(this volume)
Controls-"blind" 8 106 0.168 0.575 Judge (this volume)
Cornell-l
Controls 30 20 ± 26 0.541 6.0 x 10- 5 Baker (this volume)
Cornell-2
Controls 56 -46 0.179 0.094 Adler and Pelkie
(this volume)
Magnets 18 -54 ± 21 0.750 4.0 x 10- 3 Adler and Pelkie
(this volume)
Total
Controls 226 -4 ± 38 0.156 4.9 x 10- 4
Magnets 53 - 32 ± 29 0.393 3.3 x 10- 4
Pointing Estimates
Princeton-l
Controls 20 -78 0.140 0.426 Gould and Able
(1981]
Magnets 20 -42 0.100 0.318 Gould and Able
(1981)
Princeton-2
Controls 20 -45 0.281 0.102 Baker (unpublished)
Albany
Controls 67 -178 0.080 0.822 Able and Gergits
(this volume)
Controls-"blind" 9 121 0.159 0.638 Judge (this volume)
Cornell-l
Controls 30 56 ± 55 0.310 0.087 Pelkie (personal
communication)
Cornell-2
Controls 56 -40 ± 31 0.342 3.0 x 10- 3 Adler and Pelkie
(this volume)
Magnets 18 39 ± 43 0.470 1.5 x 10- 2 Adler and Pelkie
(this volume)
Total
Controls 202 -29 ± 71 0.104 3.4 x 10- 2
Magnets 38 26 ± 69 0.237 3.2 x 10- 2
a Significant p values (V-test) are shown in exponential form.

Reply 613
Figure 1. Significant influence of (a) brass/lead (b) magnets

magnets on estimates of the com-
pass direction of home in Amer-
ican bus experiments. Estimates
of the compass direction of home
at the first (or only) stop in all
American bus experiments in
which subjects wore: (a) magnet-
ically inert bars or weights; (b)
magnets. In the Princeton-1 (e)
and Cornell-2 (et) experiments,
subjects wore magnets on the
front of the head; in Princeton-2
(0) on the back. Each dot is
the estimate of home direction
(vertical dotted line) by one
person. Arrow shows the mean
vector (eO, r: where eO = mean error, r = length of mean vector) of the estimates. When r = 1.0, arrow
length = radius of the circle. Dashed lines show 95% confidence limits of mean error.
The strength of homeward orientation by subjects wearing magnets is significantly different from
that of controls wearing inert bars (N 1 •2 = 48,53, U = 939, Z = - 2.276, P (2-tailed) = 2.3 x 10 -2,
Wallraff's test).
Drawn from data in Gould (1980), Dayton (this volume), Adler and Pelkie (this volume).
conservative, however, I have used Dayton's figures in Table I. As Adler and Pelkie (this
volume) do not present second-order analyses for their Cornell experiments, I have fol-
lowed the recommendation in their paper and used data from only the first site on each
of their four journeys.
In my various publications (e.g., Baker, 1981), I have claimed that bus experiments at
Manchester have shown two things: (1) humans have a nonvisual element to their navi-
gational armory and (2) at least part of this nonvisual ability is based on magnetoreception.
We now have to ask whether the collected evidence from American experiments shown
in Table I supports either or both of these claims.
The evidence for a nonvisual element to human navigational ability seems over-
whelming. Blindfolded subjects tested in an unaltered magnetic field show a significant
abilitybothtopointtowardhome(N = 202,e = -29 ± 71°,r = 0.104,p = 3.4 x 10- 2 ,
V-test) and to estimate its compass direction (N = 226, e = -4 ± 38°, r = 0.156, P =
4.9 x 10-4, V-test) (in this paper I follow the convention of giving exact probabilities
where possible and distinguishing significant values by expressing them in exponential
form).
Three of the American experiments have used magnets on the head to alter the ambient
magnetic field during displacement. Twice, magnets have been placed on the front of the
head (Princeton-1; Cornell-2); once on the back (Princeton-2). Table I shows that, in general,
subjects wearing magnets produce stronger homeward orientation than subjects who ex-
perience an unaltered magnetic field. Figure 1 compares estimates of home direction on
the same three journeys by "experimentals" (wearing magnets) and "controls" (wearing
magnetically inert brass or lead weights). The difference in homeward orientation is sig-
nificant (N 1 ,2 = 48,53, U = 939, Z = - 2.276, P (2-tailed) = 2.3 x 10- 2 , Wallraff's test).
Figure 1 meets the most stringent of conditions laid down by the various authors in
this volume. Adler and Pelkie (this volume) are critical of the more liberal procedure
adopted by both Dayton (this volume) and myself (Baker, this volume) of lumping second-
614 Chapter 34
order mean vectors (from journeys with several stops) with first-order mean vectors (from
journeys with just one stop). As far as the experiments with magnets are concerned, the
Princeton-l journey (May 2, 1980) had only one stop, the Princeton-2 journey (February
18,1981) had four, and the Cornell-2 journey (June 29, 1982) had two. Moreover, at the
second site on the Cornell journey, subjects estimated the direction of the previous stop,
not "home" as in Princeton-2. Adler and Pelkie (this volume) consider that the only truly
independent data from journeys with two or more stops are those from the first stop; these
are the data presented in Fig. 1.
Kirschvink et aI. (this volume) report on a chair experiment of their own design in
which, between each estimate, the ambient magnetic field changes in relation to the room.
If subjects use a magnetic sense, estimates of direction should relate more to the artificial
magnetic field than to the room. Relevant data are shown in Kirschvink et aI. (this volume,
Table I). It can be calculated from the mean errors for the seven subjects that the lengths
of the mean vectors relative to the artificial field (pointing: 0.402; verbal: 0.558) are indeed
greater than for the mean vectors relative to the room (pointing: 0.250; verbal: 0.288),
though with a sample size of only seven the differences are not significant. The results
are at least encouraging and hardly the basis for even a "tentative" conclusion that "Caltech
students and perhaps humans in general lack any strong or useful ability to sense magnetic
direction" (Kirschvink et aI., this volume).
In summary, data from American bus experiments show nothing but support for the
earlier British data. Both British and American series show overwhelming evidence of an
ability for nonvisual navigation by humans and both indicate that at least part of this ability
involves magnetoreception. American chair experiments also tend to support, rather than
contradict, the suggestion that humans have and use an ability for magnetoreception. De-
spite all of this evidence from their own experiments, my American colleagues remain
unanimous in their opposition to such a suggestion and attempt to bolster their stand with
a veritable barrage of criticisms of the British data.
In my view, the results presented in Table I and Fig. 1 answer my critics far more
eloquently than pages of argument. Nevertheless, I feel obliged to address those criticisms
that involve points of importance to the design, analysis, and interpretation of future ex-
periments.
2. The American Criticisms
2.1. Second-Order Analysis
I originally introduced multiple stops in bus experiments as a means of (1) studying

the way navigational accuracy changed with time and distance and (2) testing whether
orientation used primarily a sense of rotation or direction (Baker, 1981). Able and Gergits
(this volume) make similar use of such data.
Most authors (Able and Gergits, this volume; Adler and Pelkie, this volume; Baker,
this volume; Dayton, this volume) agree that, when multiple stops are made, the correct
procedure is to use second-order statistics. Such analysis is adopted by Able and Gergits
(this volume), Dayton (this volume), Judge (this volume), and myself (Baker, this volume)
[n.b. Adler and Pelkie (this volume, Section 3) are incorrect in assuming I lump data from
several stops on those 23 of 31 days that two or more stops were made-second-order
analysis was used]. However, Gould (this volume) argues that successive estimates on a
single journey may be relatively independent. If so, second-order analysis would be less
necessary and the lumping of data from several stops may sometimes be justified. In support
of his suggestion, he shows that at Princeton the mean difference between successive es-
timates is 83° rather than 0°. The analysis, however, is inappropriate.
Reply 615
On journeys with multiple stops, subjects undoubtedly make a direct attempt to es-
timate home direction at the first stop. At subsequent stops, however, an individual often
compromises between making navigational judgements and "playing percentages" in the
hope that at least one estimate during the journey will be absolutely correct. To this end,
many deliberately err to one side or other of their previous estimate. A mean difference
of 83° as calculated by Gould is consistent with this practice and is in no way evidence
for a lack of dependence of one estimate on another. Indeed, the possibility that subjects
may use such a "percentage" strategy emphasizes the importance of second-order analysis
as a means of isolating the element of navigational ability from other factors.
There is less agreement on how to handle data from subjects who make more than
one journey. Adler and Pelkie (this volume) feel that such subjects should contribute only
one data point to higher-order analysis. Dayton (this volume) sometimes uses only one
data point per subject and sometimes more than one, depending on some subjective as-
sessment of the influence of the time interval between successive journeys. Researchers
on pigeon navigation regularly allow individual birds to contribute more than one data
point to specific analyses.
I have no strong feelings either way on this matter, lay down no guidelines in my
chapter (Baker, this volume, Section 3.3.2), and, as pointed out by Adler and Pelkie (this
volume, Note 3), combine results from experiments in which subjects were tested on more
than one journey (Baker, this volume, Table III). Reanalysis of the data, using only subjects
on their first trip, gives figures of: N = 173, e = 18 ± 20°, r = 0.305, P = 9 X 10- 9 , V-
test, for compass estimates; and N = 147, e = -6 ± 23°, r = 0.285, P = 7 X 10- 7 , V-
test, for pointing estimates. No conclusions are affected. Indeed, homeward components
are improved.
In contrast, I feel more concerned that chair experiments should generate for higher-
order analysis only one data point per person (Baker, this volume, Section 3.2.2). Other-
wise, there is a temptation to make more frequent use of "good" or "bad" subjects, de-
pending on prejudice. Thus, I should accept for higher-order analysis only the seven in-
dividual mean errors shown under each category in Kirschvink et aI. (this volume, Table
I), not the 68 mean errors shown in their Fig. 1.
Adler and Pelkie (this volume) accept the correctness of second-order analysis in bus
experiments. At the same time, they see a danger that such an analytical technique may
exaggerate the actual level of homeward orientation. They illustrate their point by means
of a Monte-Carlo simulation. The reader should beware, however, of their use of the word
"random" in describing this simulation. Just once in their paper, Adler and Pelkie warn
that their data are not random but designed to model a bias in the correct direction. In
their Introduction, however, they give no such warning. Second-order analysis cannot
spuriously produce homeward orientation unless something causes a bias in the correct
direction; a bias that may be due to a variety of factors, including magnetoreception.
Despite general support for the use and validity of second-order analysis, I feel we
should consider seriously the suggestion by Adler and Pelkie (this volume) that future bus
experiments should make just one stop per journey. This would not only remove one area
of dissent but also have the benefit that measured levels of navigational performance by
humans and other animals would be more comparable.
2.2. Post-Hoc Analysis
Gould (this volume) correctly warns of the dangers of post-hoc analysis but incorrectly
implies that conclusions in my review of the Manchester experiments (Baker, this volume)
derive from such procedure. For example, hypotheses concerning the influence of bed
616 Chapter 34
September (Baker and Mather, 1982) and San Francisco in December (Baker et al., 1982).
Subsequent experiments have been designed to test specific predictions (e.g., N-sleepers
orient better than So, E-, and W-sleepers; N-S sleepers orient better than E-W sleepers;
subjects not wearing polyester material orient better than subjects wearing such material).
Thus, the procedure used is precisely that advocated by Gould (this volume).
When, as in the two examples above, research hypotheses (sensu Siegel, 1956) are
clearly directional, I see no violation in applying one-tailed tests of probability. Dayton
(this volume) suggests that one-tailed tests are inappropriate, not only for the data on bed
orientation and clothing, but also for the interactive influence of magnets and magnetic
storms.
Several years ago, Larkin and Keeton (1976) showed that the influence of magnetic
storms on the orientation of homing pigeons was masked by magnets on or near the birds'
heads. As a consequence, the research hypothesis adopted in our study of the interactive
effects of magnetic storms and magnets on human navigation was that an artificial magnetic
field through the head would similarly mask any influence of magnetic storms. Statisti-
cally, such a masking effect should emerge as a weaker correlation between K indices and
homeward orientation for subjects wearing magnets than for controls. A one-tailed test
would seem to be totally justified. Using Dayton's own figures (Dayton, this volume, Table
II; Fisher's z-transform), the probability that the correlation coefficient for controls is ab-
solutely greater than for subjects wearing magnets is 3.15 x 10- 2 for compass estimates
and 0.100 for pointing estimates (l-tailed tests).
Policy over the use of one-tailed and two-tailed tests in the face of directional research
hypotheses remains a matter for discussion rather than dogma. It should be noted, however,
that no major conclusion in my review (Baker, this volume), other recent publications
(Baker, 1984, 1985), or this paper is negated by adoption of two-tailed probabilities.
Returning to the question of post-hoc analysis, the only gross example of such pro-
cedure in this volume, acknowledged by the authors themselves, is that by Adler and
Pelkie. However, there are other, more subtle, examples that are not acknowledged. When
Dayton (this volume) states that the American compass data show significant homeward
orientation only because of the collaborative experiment at Cornell, he is performing post-
hoc analysis. In any case, his implicit conclusion is negated by the significant homeward
orientation since reported at Albany by Able and Gergits (this volume) and in noncolla-
borative tests at Cornell by Adler and Pelkie (this volume). Similarly, when Able and
Gergits (this volume) suggest that significant homeward orientation at Albany is heavily
influenced by good orientation at one site on one journey, they too are carrying out post-
hoc analysis. They are also showing bias. For example, their results are also heavily in-
fluenced in the opposite direction by poor homeward orientation at other stops on other
journeys (e.g., sites 4 and 5, April 16, 1981, Able and Gergits, Fig. 2), but the authors choose
not to mention this in the same context.
2.3. Experimental Protocol
Gould (1980, this volume) and Adler and Pelkie (this volume) express concern over
various features of the Manchester protocol for bus experiments. This concern derives in
part from their interpretation of my published accounts but primarily from first hand ob-
servation of the collaborative experiments my colleague, Dr. Janice G. Mather, and I ran
with Gould at Princeton and with Adler's colleagues Chris Pelkie and Irene Brown at
Cornell. The points of concern are all important and deserve scientific evaluation.
Gould (this volume) questions the efficacy of the blindfolds used at Manchester and
both Gould (this volume) and Adler and Pelkie (this volume) suggest the Manchester routes
may be too simple. The same authors suspect that verbal instructions from an experimenter
Reply 617
who knows the route about to be taken may give unconscious information to the subjects
in accordance with some "Clever Hans" effect. Adler and Pelkie (this volume) are further
concerned that olfactory, acoustic, and topographic cues may be used and Able and Gergits
(this volume) are "suspicious" that inertial navigation is being employed.
All of the American experiments took elaborate precautions to prevent the use of visual
information, often at the cost of considerable discomfort to their subjects (Gould, 1980;
plus my own personal experience as a subject at Princeton). Various combinations of dou-
ble-blindfolds and hoods plus blacked-out windows were used. Despite these precautions,
homeward orientation still occurred (Table I). It is a safe conclusion that some nonvisual
ability is being used.
Able and Gergits (this volume) "suspect" that this nonvisual ability may be based on
inertial navigation, presumably using some form of internal gyroscope as suggested by
Barlow (1964). Gould (this volume) also implies such a mechanism when he suggests I
carry out bus experiments over routes that are too simple. Able and Gergits see support
for their suspicion in a tendency for homeward orientation to deteriorate as the journey
progresses. They suggest this would not happen if magnetoreception were involved.
Analogy with what is known for pigeons makes such an argument surprising. First,
there is no a priori reason for magnetoreception to prevent deterioration in homeward
orientation with increasing distance. Second, pigeons are known to use magnetoreception
during displacement and yet from many, if not the majority of, lofts show just such a
deterioration with distance (Schmidt-Koenig, 1979). Third, despite this observed deteri-
oration, few recent reviewers argue that pigeons use inertial information in navigation.
Not only is such an argument for human navigation surprising, it has no support from
the available evidence. It is true that homeward orientation deteriorated with distance in
two of the Albany experiments (Able and Gergits, this volume, Fig. 2), and on February
16 and 17, 1981 during the Princeton-2 (collaborative) series. However, on February 18,
1981 during the Princeton-2 series, homeward orientation by controls improved with dis-
tance (Gould and Able, 1981) and on both of the journeys with multiple stops during the
Princeton-l series, homeward orientation was better at the last stop than at the first two
stops (Gould and Able, 1981).
At both Princeton and Cornell a mixture of complex and less complex routes have
been used. These allow some assessment of whether navigational accuracy is dependent
on route complexity as might be expected if inertial cues are involved. In the Princeton-
2 experiments, the first route, organized by Gould, was relatively complex; the second,
organized by myself and Dr. Mather, perhaps less so. The third was intermediate. At Cor-
nell, the collaborative experiment apparently used a less complex route than subsequent
experiments. The results (see Dayton, this volume, Table I; Adler and Pelkie, this volume,
Table I) show no clear relationship between route complexity and navigational accuracy.
So far, the most elaborate precautions during bus experiments were those taken by
Adler and Pelkie (this volume) at Cornell. Routes were complex and there was stringent
care to prevent the use of visual cues. In addition, verbal instructions were given by tape
recorder, compass directions were displayed by signpost, and all aspects of the execution
of experiments and transcription of results were double-blind. Olfactory cues were masked
by sprays; acoustic cues by playing the sound of surf. Experimental protocol was admirable
and contrasts with the much less stringent design of the earlier collaborative experiment
at Cornell.
Having gone to such lengths to control all of these factors, Adler and Pelkie (this
volume) might have been expected to present a direct comparison between results from
the collaborative and subsequent experiments. Instead, the results are presented in such
a way that comparison is difficult. No second-order calculations are made and data are
presented (Adler and Pelkie, Table I) separately for each site. The important question of
whether stringent experimental protocol reduces homeward orientation is lost in a maze
618 Chapter 34
of post-hoc analysis. We are told only that the number of stops showing significant compass
estimates of home is lower for the stringent tests than for the collaborative test. The fact
that the number of stops showing significant homeward orientation by pointing is greater
in the stringent tests is not mentioned.
Adler and Pelkie (this volume) warn that the only truly independent estimates of home
direction at Cornell are those at the first stop on each of the single collaborative and four
subsequent journeys. Thus, one acceptable measure of the effect of stringent protocol on
homeward orientation can be obtained by summating estimates at all first stops on the
Cornell-2 experiments and comparing the mean vector with that at the first stop on the
collaborative journey. Compass estimates give: N = 30, e = 29 ± 58°, r = 0.296, P = 2.2
X 10- 2 , V-test, for the collaborative experiment; and N = 74, e = -50 ± 29°, r = 0.317,
P = 7 x 10 - 3, V-test, for the stringently controlled experiments. Pointing estimates give:
N = 30, e = -7 ± 70°, r = 0.274, P = 1.7 X 10- 2 , V-test, and N = 74, e = -17 ± 30°,
r = 0.300, P = 2.7 X 10- 4 , V-test. The stringent experimental protocol, therefore, has no
clear effect, reducing the homeward component for compass estimates from 0.259 to 0.202
but actually increasing that for pointing estimates from 0.272 to 0.287. Moreover, despite
the larger sample size for the more-stringent tests, the length of the mean vector increases
in both cases (compass: 0.296 to 0.317; pointing: 0.274 to 0.300).
Much of the discussion in the paper by Adler and Pelkie (this volume), including
virtually all of their Note 8, is based on the erroneous assumption that the Manchester
protocol involved subjects' removing their blindfolds at each stop. This is not part of our
procedure at Manchester nor was it part of the experiments at Princeton, Albany, or even
of the collaborative experiment at Cornell. In this last case, subjects removed their blind-
folds at the first stop, for the reasons correctly given by Adler and Pelkie. Thereafter,
blindfolds and hoods remained in place until the experiment was over. There was no
further update of compass information at the remaining three stops. Comments on the
possible effect of such procedure in generating the Manchester results are therefore in-
applicable.
The chair experiment described by Kirschvink et 01. (this volume) did not attempt to
follow the protocol used at Manchester. The authors themselves point out, somewhat
tongue-in-cheek, a number of differences between their protocol and my own. Even though
I have presented data that their items (2) and (3) have a significant influence (Baker, 1984a,
this volume) and despite the fact that they can offer no data to the contrary, the authors
express a confident "belief" that the differences are "of little or no importance." The most
important difference in our respective protocols, however, is not mentioned: their exper-
iment requires humans to orient to a field that changes from one estimate to the next. There
is no evidence that any animal can show accurate magnetoreception under such conditions.
Indeed, Wallraff and Gelderloos (1978) were unable to demonstrate magnetoreception in
birds when they exposed passerines to a slowly rotating magnetic field.
Although Caltech students were not accurately oriented to the changing magnetic field
in chair experiments, they were, in accordance with the use of a magnetic sense, more
oriented to the magnetic field than to the room (see above). With larger sample sizes, the
Caltech protocol may yet be suitable for the study of magnetoreception.
In my account of the Manchester chair experiments (Baker, this volume), I describe
how double-blind protocol was achieved in experiments involving magnets, but do not
indicate which of the control groups listed in Table I of that review were exposed to such
protocol. With the exception of the pilot experiment on "Manchester students, 1980-81"
carried out by Meharg (personal communication), all experiments concerned with magnetic
effects ("Field course students, 1981, 1982, 1983" and "Bramhall residents") were double-
blind for magnetic treatment. The remainder were single-blind, subjects wearing a helmet
or brass bar that they knew may be magnetic but which the experimenter knew to be inert.
Even so, all experiments, except that on naturists, were double-blind with respect to the
Reply 619
main point of the experiment (e.g., bed orientation, dyslexic status). The combined single-
blind (Le., magnetically) experiments in Baker (this volume, Table I) give: N = 762, e =
-7 ± 12°, r = 0.160, P = 3.1 X 10- 11 , V-test. Double-blind experiments give: N = 113,
e = -6 ± 28°, r = 0.278, P = 2 X 10- 5 , V-test. The respective homeward components
are 0.159 and 0.276. There is thus no indication that orientation is improved by uncon-
scious communication between experimenter and subject.
Dayton (this volume) suggests that the data I present in Baker (this volume, Table II)
do not rule out the possibility that Manchester chair experiments test for an inertial sense
of rotation rather than a sense of direction. His grounds are that (1) my comparison of first
and last estimates gives no formal statistical support and (2) comparison of the second
estimate of subjects who orient either well or badly on their first estimate is merely a
comparison of good and bad subjects. Neither of his points have weight. The inertial hy-
pothesis predicts that the last estimate in a test should be less accurate than the first. As
the strength of compass orientation at the last estimate is marginally better than at the first,
p (l-tailed) is automatically greater than 0.5 and no formal statistics seem necessary. Sim-
ilarly, the strength of compass orientation at the second estimate by subjects incorrect at
the first is again marginally greater than that by subjects correct at the first. It follows we
can reject both the suggestion that inertial cues are being used and Dayton's suggestion
that the two groups are merely "good" and "bad" subjects.
2.4. Influence of Magnets
One benefit of the American search to find a weakness in the Manchester bus exper-
iments is that tests have now been carried out under the most stringent of conditions. There
is one respect, however, in which the Manchester protocol is equally as stringent: through-
out the journey and while arrows and compass estimates are being measured and tran-
scribed, experimenters do not know which subjects have been exposed to magnets and
which to brass.
I have always stressed that magnetoreception may be only one of a multiplicity of
nonvisual elements in the navigational armory of humans, as of other animals (Baker, 1981,
1984a). As yet, however, it is the only element for which there is direct experimental
support. The Cornell group found that the masking of smells and sounds during displace-
ment had no obvious influence on homeward orientation. The Manchester data are more
consistent with a sense of direction than a sense of rotation (Baker, 1981). Finally, the use
of topographical cues has not yet been tested and, indeed, suitable experiments may be
difficult to design. In contrast, data both from Manchester (Baker, 1984, 1985 this volume)
and now from America (Fig. 1, this chapter) show that magnets on the head have a sig-
nificant influence on homeward orientation. Thus, the only element of nonvisual navi-
gation by humans for which there is any experimental support so far is magnetoreception.
The significant influence of magnets on orientation in American bus experiments only
emerges when the data from Princeton and Cornell are combined. The authors concerned,
therefore, may be excused for overlooking the effect, though the reader may be surprised
that Adler and Pelkie (this volume) did not comment on the contrast between the results
in their Fig. 1, A and E [the difference in dispersion is significant: U = 78, Z = - 2.525,
P (2-tailed) = 0.012, Wallraff's test].
Dayton (this volume) seeks to minimize the extent of the influence of magnets in the
two Princeton bus experiments by suggesting that the magnets have opposite effects on
the two journeys. In fact, as his own figures show clearly, on both journeys magnets improve
homeward orientation. The comparable compass estimates of home at Cornell replicate
this effect (Adler and Pelkie, this volume).
620 Chapter 34
Able and Gergits (this volume) express confusion over whether magnets are expected
to improve or disrupt orientation and do not see nor seek a pattern in the results. I have
pointed out (Baker, 1985) that the combined Manchester, Barnard Castle, and Princeton
bus experiments do show a trend. Thus, magnets on the front of the head (Princeton-l) or
between the ear and the eye (Manchester) both produce significant homeward orientation
that for the combined data represents a significant improvement over orientation by con-
trols. Magnets on the back of the head (Barnard Castle, Princeton-2) have less, if any, effect.
Magnets behind the ear (Manchester) have an intermediate effect. The new data for compass
estimates from Cornell (Adler and Pelkie, this volume) reinforce this trend.
There is still a long way to go before we have a complete picture of how the position
of magnets on the head, as well as their strength, polarity, and axis of alignment each
influences homeward orientation and whether this influence is different for compass and
pointing estimates. Perhaps, with the publication of these American data, we can now stop
arguing over whether magnets have an influence and begin instead to collaborate in un-
raveling the nature of the influence.
2.5. Accuracy of Analysis
Dayton (this volume) and myself (Baker, this volume) report errors and inconsistencies
in the original analysis of the Princeton bus experiments (Gould, 1980; Gould and Able,
1981). Gould (this volume) acknowledges at least some of these and states that the analysis
presented by Dayton (this volume) is now correct. Able and Gergits (this volume, Table
I) also now give mean vectors of N = 10, e = +43°, r = 0.26 and N = 9, e = +96°, r =
0.52 which differ from those of N = 10, e = + 77°, r = 0.14 and N = 9, e = + 79, r =
0.17 originally published (Gould and Able, 1981, p. 1062) for the first two journeys at
Albany.
Gould (this volume) and Dayton (this volume) both state that my analysis of the Prince-
ton experiments (Baker, this volume, Table IV) is "incorrect." Dayton's analysis clearly
differs from my own (compare his Table I with my Table IV). My analysis is based on a
data set sent to me by Gould in 1981 and gives mean vectors for individual sites during
Princeton-2 (Le., collaborative) experiments that, with three minor but conservative ex-
ceptions, are the same as those given by Gould and Able (1981). I feel confident, therefore,
that my data set is to all intents and purposes the same as that used by those authors.
According to Dayton (this volume], his analysis had the benefit of "corrections" sub-
sequently sent to him by Gould. Comparison of Table I (in Dayton, this volume) with Table
1 (in Gould and Able, 1981) and Table IV (in Baker, this volume) shows that these "cor-
rections" have the effect of reducing second-order sample size for controls from 51 to 49,
increasing the sample size for subjects wearing magnets from 14 to 15, and changing prob-
ability values (V-test) for total controls from 0.047 to 0.071. Dayton (this volume) further
reduces effective sample size for controls to 45 (for which P = 0.098) by using third-order
mean errors for subjects taking part in more than one journey. Our analyses differ, therefore,
partly due to the "corrections" sent to him, but not to me, and partly due to differences
in what we have treated as independent data points. In the wider context of other American
bus experiments (Table I, this chapter), it seems relatively trivial whether the Princeton-
2 experiments give p = 0.047 or 0.098. The problem of which, if either, is actually "correct"
appears insoluble.
2.6. Anonymous Comments

The editors have received comments from two American scientists stating that some
features of my review (Baker, this volume) are inconsistent with claims made in earlier
Reply 621
manuscripts that were submitted to scientific journals, reviewed by the scientists con-
cerned, and never published. Specifically, the editors were told that there are differences
in the claimed influence of magnetic storms and that details of chair experiments presented
in my current review make the protocol appear more rigorous than originally described.
According to these scientists, earmuffs were not worn and experiments were not double-
blind. The editors have invited me to reply publicly.
The influence of magnetic storms demonstrated in Baker (this volume, Fig. 3 and Table
V) is the same as that claimed in unpublished manuscripts.
No chair experiment at Manchester has ever been performed without earmuffs.
Since June 1981, all chair experiments in which magnets were involved have used
double-blind protocol.
A variety of manuscripts, documents, and films exist to support these statements and
some have been sent to, or seen by, the editors.
3. Concluding Remarks
The American data now rival the Manchester data in the clarity with which they
demonstrate the existence of nonvisual navigational ability in humans and the involvement
in this ability of magnetoreception. Despite this, my American colleagues remain united
in their opposition to accepting the É ñ á ú í É å Å É = of human magnetoreception. Had the Amer-
ican data as a whole shown:
1. Absence of homeward orientation in bus experiments; or
2. Abolition of homeward orientation upon stringent control for visual, olfactory,
acoustic, inertial, and "Clever Hans" effects;
3. No significant influence of magnets on homeward orientation;
4. r values in Caltech chair experiments that related more to the testing room than to
the magnetic field;
there would have been some scientific justification for continuing opposition to the Man-
chester data. Yet not one of these effects is found; the results are totally consistent with
the British claims.
In spite of the clarity and strength of their combined results, not once does an American
researcher acknowledge support for the experiments they have each attempted to replicate.
Instead, the various authors voice confident disclaimers based on a mixture of post-hoc
analysis, statistical and arithmetic quibbles, unsupported "belief" and "suspicion," and
finally anonymous and unfounded aspersions.
ACKNOWLEDGMENTS. The Manchester project on human magnetoreception is supported by

SERC Grant GRlB74337.
References
Baker, R R, 1984, Bird Navigation: The Solution of a Mystery?, Hodder & Stoughton, London.
Baker, R R, 1985, Exploration and navigation: The foundation of vertebrate migration, in: Migration:
Mechanisms and Adaptive Significance (M. A. Rankin, ed.), Port Aransas Marine Laboratory, Port
Aransas.
Baker, R R, and Mather, J. G., 1982, A comparative approach to bird navigation: Implications of
parallel studies on mammals, in: Avian Navigation (F. Papi and H. G. Wallraff, eds.), Springer-
622 Chapter 34
Baker, R. R., Mather, J. G., and Kennaugh, J. H., 1982, The human compass?, EOS 63:156.
Barlow, J. 5., 1964, Inertial navigation as a basis for animal navigation, J. Theor. BioI. 6:76-117.
Gould, J. 1., 1980, Homing in on the home front, Psychol. Today 14:62-71.
on pigeon orientation, J. Compo PhysioI. 110:227-231.
Schmidt-Koenig, K., 1979, Avian Orientation and Navigation, Academic Press, New York.
Siegel, S., 1956, Nonparametric Statistics for the Behavioral Sciences, McGraw-Hill, New York.
Wallraff, H. G., and Gelderloos, O. G., 1978, Experiments on migratory orientation of birds with sim-
ulated stellar sky and geomagnetic field: Method and preliminary results, Oikos 30:207-215.
Biogenic Magnetite in the Fossil Record
VI
It is reasonable to conclude from the wide spectrum of organisms with the ability to pre-
cipitate magnetite that it probably evolved prior to the major radiation of animal phyla in
the late Precambrian. From a variety of paleomagnetic studies, it is also known that fine-
grained magnetite is one of the most common carriers of stable natural remanent mag-
netization in marine sediments, and this appears to have been true for most of Phanerozoic
time. Although the source of this fine-grained magnetite has long been a mystery, partic-
ularly in deep-sea sediments, two analyses predicted that the contribution of bacterial
magnetite to the sediments could be responsible (Kirschvink and Lowenstam, 1979; Towe
and Moench, 1980). Small particles of magnetite, however, are easily destroyed in an ox-
idizing environment and it is not known a priori which rock types would be best to examine
for bacterial magnetofossils. The two chapters in this section document the first steps in
the direct search for these objects in the fossil record, and the results show a great deal of
promise. If the magnetite-formation step in all magnetotactic bacteria is indeed oxygen-
dependent, as appears to be the case from laboratory cultures, these "magnetofossils" may
prove to be useful indicators for microaerophilic conditions at the time of deposition.
623
Chapter 35
A Search for Bacterial Magnetite in the
Sediments of Eel Marsh, Woods Hole,
Massachusetts
ANNE DEMITRACK
1. Introduction. . . . . . . 625
2. Bacterial Magnetite .. 626
3. Methods . . . . . . . . . 627
3.1. Sampling Site Description, Core Sampling Procedure. . .. . . . . . . . . . . . . . . 627
3.2. Magnetic and Thermomagnetic Procedures. . . . . . . . . . . . . . . . . . . . . . . . . 627
3.3. Magnetic Separation Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 628
3.4. Transmission Electron Microscopy and Electron Diffraction. . . . . . . . 629
3.5. Scanning Electron Microscopy and Energy-Dispersive X-Ray Analysis . . . . . . . . 629
3.6 X-Ray Diffraction. . . . . . . .. .......................... 630
4. Results. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . 630
4.1. Whole Mud. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . 630
4.2. Separate . . . . . . . . . . . . . . 633
4.3. Summary of Results. . . . . . . . .. . . . . . . . . . . . . . . 639
5. Discussion. . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . 639
Appendix 1: Eel Marsh NRM and Saturation Magnetization Data. . . 643
Appendix 2: Description of Computer Procedure Used to Make Stability Field Diagram
8a. . . . . . . . . . . . . .. .............. 644
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644
1. Introduction
Magnetite is the magnetic mineral most prominent as a remanence carrier in sediments.

Not surprisingly, the discovery of a magnetic bacterium (Blakemore, 1975) caught the at-
tention of researchers like Kirschvink and Lowenstam (1979) who recognized its impor-
tance to the field of paleomagnetics. If these microscopic magnetite makers were present
in large numbers in their environment, they might supply a significant quantity of fine-
grained magnetite to the sediments. Furthermore, as they are known in both marine and
nonmarine environments (Moench and Konetzka, 1978), and have been found in both the
northern and southern hemispheres and at the equator (Kirschvink, 1980; Frankel et a1.,
1981), they are potentially a widely distributed source.
Like the bacteria, magnetite appears in a wide range of environments. It is known to
be the major carrier of remanence in modern lake sediments from Scotland (Thompson
and Morton, 1979), Northern Ireland (Thompson et a1., 1979), England (Mackereth, 1971;
ANNE DEMITRACK • Department of Geology, Stanford University, Stanford, California 94305 .
625
626 Chapter 35
Thompson, 1973), and Finland (Stober and Thompson, 1979). It is also known in marine
sediments from the Atlantic, Pacific, and Indian oceans (Lovlie et aI., 1971), and from the
Black Sea (Creer, 1974).
Many researchers have recognized that magnetite can be supplied inorganically to
sediments by (1) erosion of local bedrock and soils (Thompson and Morton, 1979), (2)
fallout of cosmic spherules or industrially derived particles (Oldfield et aI., 1977; Doyle
et al., 1976), and (3) authigenic precipitation from solution (Mackereth, 1971). On the other
hand, organic sources remain poorly explored. Although Lowenstam first described bio-
genic magnetite in the denticle cap pings of chitons in 1962, the contribution of biogenic
magnetite to remanence in sediments was considered insignificant until a large-scale and
cosmopolitan producer, like the bacteria, was found.
Previously published data, reexamined in light of a bacterial source for magnetite,
suggest a possible biogenic role in the magnetization of certain sediments. For example,
Mackereth (1971) noted a correlation between higher remanence intensity and higher or-
ganic matter content in cores from Lake Windermere, England. He concluded that episodes
of increased bacterial activity may have caused an increase in the precipitation of authi-
genic magnetite from solution. Alternatively, this correlation might indicate the presence
of magnetic bacteria within the bacterial bloom, which would naturally be reflected as an
increased production of biogenic magnetite. Stober and Thompson (1979) noted that mag-
netite was present in greater concentration in Finnish lake sediments than in its source
terrain, and that it was concentrated to a greater degree than were other magnetic phases.
They interpreted this as evidence of preferential segregation of magnetite, due to density
differences, during transport. On the other hand, this concentration may reflect the addition
of a biogenic magnetite component to the sediments in the lake. Kirschvink (1982) has
reevaluated published data on cores of Miocene marine clays from western Crete which
showed a correlation between intervals of magnetic reversal and decreased quantity of
magnetic material. The authors of the study could not explain the correlation, but Kirsch-
vink has suggested that it reflects a decrease in the bacterial magnetite contribution as a
result of intensity decreases in the geomagnetic field during a reversal. The bacteria would
have to manufacture more magneto somes in order to continue to orient in the dwindling
geomagnetic field, and eventually, this would become a competitive disadvantage if not
a physical impossibility. Hence, selection against the bacteria would manifest itself as a
decrease in the amount of magnetic material contributed to the sediment.
This study pursues this theme by exploring the possibility of a biogenic origin for the
remanence carrier in the sediments of Eel Marsh in Woods Hole, Massachusetts. Eel Marsh
was one of the first environments found by Blakemore to contain magnetic bacteria (Blake-
more, personal communication) and was thought to be suitable for protecting fine-grained
magnetite from oxidation. It provides an appropriate setting in which to consider a bacterial
source for the remanence carrier.
2. Bacterial Magnetite
Magnetotactic bacteria were first described by Blakemore (1975) who had observed
their migration to the north end of a microscope slide. These organisms have evolved the
specialized ability to navigate in the ambient geomagnetic field by means of a chain of
electron-dense magnetic particles or magneto somes contained within their cytoplasm
(Frankel et aI., 1979; Balkwill et aI., 1980). Although it was originally thought that the
bacteria used magnetotaxis to find the less oxygenated parts of their environment, their
discovery at the equator has instead implied that they use it for more efficient and directed
movement (Frankel et aI., 1981).
Eel Marsh Greigite 627
Towe and Moench (1981) recognized that all bacterial magnetite particles they had
observed were single domain in size, measuring approximately 0.1 f.Lm on a side, and that
if they were incorporated into sediments, they would impart a detrital remanent magnet-
ization that would be stable over geologic time. As Kirschvink (1982) has observed, ori-
entation within a magnetic field would be impossible for an organism with superpara-
magnetic or multi domain crystals, and that if indeed the bacteria use the magnetosomes
for orientation, then natural selection virtually assures that they will be of single domain
size. The question is whether or not they are incorporated into sediments and preserved
through lithification.
In a simple model for the incorporation of bacterial magnetite into sediments, (1)
bacterial cells would settle to the bottom, (2) the cells would lyse and release their magnetic
particles, (3) the particles would align in the ambient magnetic field, and (4) this net mag-
netization would be preserved during burial, compaction, and lithification of the sedi-
ments. If this model is broadly correct, then it is likely to be repeated in the range of marine
and nonmarine environments in which magnetic bacteria have been found (Moench and
Konetzka, 1978), which implies that bacterial magnetite may account for part of the mag-
netic signal in shallow marine sediments, river sediments, and lake sediments. In addition,
Kirschvink (1982) has noted that as bacteria are among the most ancient life forms, bacterial
magnetite may be present in sediments as old as Precambrian.
Calculations show that bacterial magnetite alone can yield measurable NRM intensities
(Kirschvink and Lowenstam, 1979; Towe and Moench, 1981). A magnetotactic cell has a
magnetic moment of 1.3E-12 emu (Frankel, et a1., 1979). Given natural population densities
(Blakemore, 1975; Moench and Konetzka, 1978), and generation time (Blakemore et aJ.,
1979), a population of constant size could yield an NRM intensity of 4.75E-07 to 1.2E-02
emu/cm 2 per year. Actual intensities would vary with population size fluctuation and
changes in sedimentation rate, but these figures stress the potential significance of the
biogenic magnetic component. However, the presence of bacterial magnetite as a reman-
ence carrier in sediments remains undemonstrated.
3. Methods
3.1. Sampling Site Description, Core Sampling Procedure
Samples were taken from two coastal marine marshes located near Woods Hole, Mas-
sachusetts, and identified by Blakemore (personal communication) as having magnetic
bacterial populations. Both contained live bacteria at the time of sampling.
Eel Marsh is brackish with a water depth of about 30 cm and a strong odor of hydrogen
sulfide gase. Marsh deposits consist mainly of organic-rich muck, gray mud, and reddish
brown peat. Cedar Marsh is an embayment of Oyster pond, which is described in Emery's
(1969) monograph A Coastal Pond. As the Cedar Marsh samples had a lower remanence
than those of Eel Marsh, they did not become an important part of this study.
Oriented cores of diameter 2-% inches and length of about 1 m were collected manually
in Lucite tubes. They were dried in their tubes under a heat lamp (36°C) to 84-74% of
their original wet height, sliced in half, and boxed into oriented plastic sample boxes.
3.2. Magnetic and Thermomagnetic Procedures
NRM measurements were made at the California Institute of Technology on box sam-
ples from both marshes using an SCT cryogenic magnetometer set in a mu metal room.
628 Chapter 35
10-3 l'
ú=
a,l
!
j
1
:
1
I
CI
"
ú =
E 10- 4 d
CD
:E
a:
(/)
c
Figure 1. SRM intensities for unseparated and
b
separated samples, compared for different sepa-
ration techniques: (a) unseparated: initial; (b) sep-
arated: untreated; (c) separated: ultrasonicated;
10- 5 (d) separated: Calgon treated.
SRM and AF demagnetization measurements were made at Princeton University on a

Schonstedt Model DSM-l spinner magnetometer.
For SRM measurements, the samples were saturated between the poles of an electro-
magnet for 10 sec at 8000 G, and measured'immediately after saturation. For the AF de-
magnetization measurements, the box samples were demagnetized in a solenoid that gen-
erated a peak field of 1000 G and was contained within a field free space provided by a
Helmholtz system to an accuracy of ± 100 -y. Samples were kept in motion during de-
magnetization runs with a two-axis tumbler.
Saturation magnetization measurements were made at Princeton University on a ver-
tical balance in which the intensity of magnetization (J) of the sample could be measured
as a function of the applied field (H),
Curie temperature measurements were made on a similar vertical balance at the U.S.
Geological Survey Paleomagnetics Laboratory in Menlo Park, California.
3.3. Magnetic Separation Procedure
Mud from Eel Marsh was prepared for separation by wet sieving through a Standard
Testing Sieve No. 14 (1.40 mm) and further diluted with distilled water to make a thin
slurry. It was otherwise untreated. Initially, attempts were made to increase the effective-
ness of the separation by treating the slurry with a 0,5% solution of sodium hexameta-
phosphate (Calgon), ultrasonication, and household bleach, but the untreated mud gave
the best yield (Fig. 1),
Magnetic separations were made by pumping the fine-grained mud slurry upward
through an empty S.G. Frantz Model L-1 cannister separator (Fig. 2), which is magnetized
by the enclosing electromagnet (a) but loses its magnetization instantaneously when the
magnet is turned off. The slurry was circulated through the system for 3-6 hr at progres-
sively slower speeds; at the end, the cannister was rinsed with the magnet still on and
ÄJJZZZí ZZZí ZZú ä i =
Figure 2. Magnetic separator: a,

electromagnet; b, pole pieces; c,
stainless steel cannister. The
slurry is circulated through the
cannister from the bottom.
then flushed with the magnet off to recover the separate. During the separation, the elec-
tromagnet was at maximum field (13,500 G).
The separated mud was reconstituted into plastic sample boxes for magnetic meas-
urements that were used to monitor the effectiveness of the separation. A quantity of slurry
was pipetted into the boxes and allowed to air dry, a procedure which was repeated until
the sample box contained approximately 3-4 g of moist but solid mud, enough to saturate
and measure in the magnetometer.
In one early separation (AJ, the slurry was not pumped through the system, but rather
poured into the cannister and allowed to settle for approximately 30 sec before being
drained and replaced by more. The results of this separation compare with those of the
other method (B) in Fig. 5.
3.4. Transmission Electron Microscopy (TEM) and Electron Diffraction
TEM was performed at the Smithsonian Institution on a Phillips EM-200 transmission

electron microscope specially equipped with a rotating-tilting stage and operated at 80 kV.
Photomicrographs were taken on 35-mm Kodak fine-grained positive film.
The separate was prepared for the sample grids by further concentration with a strong
hand magnet held to the outside of a test tube. A drop of the concentrated suspension was
placed on a carbon-coated, Formvar-covered, copper-mesh grid, and dried in an oven at
low heat for several minutes.
3.5. Scanning Electron Microscopy (SEM) and Energy-Dispersive X-

Ray Analysis (ED AX)
The SEM and EDAX analysis were performed on an AMR Model 1000 scanning elec-
tron microscope, with attached Kevex Ray System 5000A.
630 Chapter 35
The samples used in the SEM were the same samples used in the TEM, and had been
mounted directly onto SEM aluminum stubs with graphite paint. The stubs were left un-
coated.
3.6. X-Ray Diffraction
X-ray diffraction was performed at Princeton University on a diffractometer equipped

with an LiF focusing monochronometer, using a Cu filament with normal filter.
The sample was prepared for the diffractometer by concentrating it at the bottom of
a test tube by centrifugation, resuspending it by ultrasonication, and passing it by vacuum
through a 0.2 f.Lm Millipore filter so that the particles collected on the surface of the filter.
The filter was rolled with its coated side down onto a glass slide which was then mounted
directly into the diffractometer.
4. Results
The purpose of this study was to examine the magnetic properties of the muds in
brackish Eel Marsh in order to identify the .magnetic remanence carrier. If fine-grained
magnetite were an important remanence carrier and if the grains were similar in form to
bacterial magnetite particles, then this would support the biogenic supply model for mag-
netite in these sediments. Conversely, the absence of magnetite in these sediments would
be negative evidence with regard to the model, as Eel Marsh is positively known to support
magnetic bacteria.
The mud was analyzed before and after magnetic separation. For the whole mud, this
analysis included (1) NRM, (2) SRM, (3) AF demagnetization behavior, and (4) saturation
magnetization. For the magnetic separate, it included (1) saturation magnetization, (2)
thermomagnetic behavior, (3) X-ray diffraction, (4) TEM and electron diffraction, and (5)
SEM with EDAX analysis.
The results of the investigation follow in individual sections below.
4.1. Whole Mud
4.1.1. NRM and SRM Measurements

The two oriented long cores from Eel Marsh both had a detectable NRM (Appendix
1). Samples from core EM 1 had an average NRM intensity of 0.647E-07 emu/cm 3 • Those
from core EM 2 had an average of 0.436E-05 emu/cm3 • These intensities are typical of
Recent sediments.
The distribution of NRM directional vectors from EM 1 is shown in Fig. 3. Although
dispersion is large (K = 4.6), the present field direction at Woods Hole lies within the circle
of 95% confidence for the sample mean, indicating the presence of a nonrandom component
parallel to the present field direction. The scatter in directions may be due in part to the
method used to box the samples which introduces error by disturbing the sediment near
the box edges, or in part to magnetization acquired during dewatering.
Similar measurements on a core from Cedar Marsh, located about 3.5 km from Eel
Marsh, showed no detectable remanence. The magnetic intensity of the Cedar Marsh sam-
ples, approximately 0.522E-08 emu/cm3 , was at or below the noise level of the cryogenic
magnetometer on which they were measured. Cedar Marsh is also known to support mag-
•
• •
•
• •
•
•
•
•
• box samp les

Figure 3. NRM directions for Eel • average vector
Marsh core EM 1. • present field ú á ê ÉÅí á ç å =
netic bacteria (Blakemore, personal communication), but differs from Eel Marsh in being
somewhat more isolated from the ocean and less brackish.
Upon saturation in an 8000-G field, the Eel Marsh samples increased in remanence
intensity by three orders of magnitude (Appendix 1). The magnitude of this increase sug-
gests a strongly ferrimagnetic material like magnetite, which would acquire a strong SRM
even at low concentration. Weakly ferromagnetic minerals like hematite would have to be
present in quantities at least an order of magnitude greater than that of magnetite in order
to give the same saturation remanence intensity.
4.1.2. AF Demagnetization of SRM

The AF demagnetization curve of the Eel Marsh sample shows a magnetically soft
material which retains less than 10% of its original magnetization at 400 G, and less than
2% .at 800 G (Fig. 4). The original saturation remanence was produced in a field of 8000
G, but its coercive force spectrum suggests that it saturated at a lower field strength. The
behavior of the sample upon demagnetization suggests that of magnetite or a magnetite-
like mineral, in contrast to hematite or a hematite-like mineral.
Figure 4a-c compares the AF demagnetization curve of the Eel Marsh sample with
published AF demagnetization curves for samples containing magnetite, maghemite, and
greigite, all strongly magnetic minerals.
Figure 4a is adapted from Lowrie and Fuller's (1971) Fig. 1, which illustrates the
experimental AF demagnetization results of Rimbert (1959). Rimbert's samples were mag-
. - -
632 Chapter 35
1.0 ....•.......
\ .....
.,o ·111•••
.,"- ...
100 300 500 oe

a Peak AC Field
1.0
.,o .,o
.,"- .,"-
úK=
'Ott.,
?Dú?= 2
................... .... ..................
300 600 oe 300 600 oe
b Peak AC Field c Peak AC Field
Figure 4. Eel Marsh demagnetization curve compared to published curves. Initial remanence is an
IRM (····Eel Marsh data, --Published data). (a) Adapted from Lowrie and Fuller (1971) after Rimbert.
Samples are magnetite powder dispersions, mean grain size 0.1 !Lm, with IRMs produced in 700- and
1200-0e fields as shown. (b) Adapted from Dunlop and West (1969). 1: Powder dispersion from ox-
idized basalt; dominantly magnetite; 60% 0.5- to 5-!Lm unaltered grains. 2: Powder dispersion of
maghemite; equant single-domain grains. (c) Adapted from Suthill et a1. (1982). Sample is magnetic
concentrate from tidal flat sediment; predominantly multi domain magnetite, with greigite >8%.
magnetized a series of these powders whose IRMs had been induced in successively higher
fields. Note the similarity in shape between the published curves and the Eel Marsh curve,
particularly below 50 Oe and above 300 Oe. The shape of a demagnetization curve is a
function of sample grain size, coercivity, and the nature of the remanence (Stacey and
Banerjee, 1974), so similarity between curves suggests common properties between sam-
ples.
Figure 4b compares Dunlop and West's (1969) experimental results, their Fig. 15b,
with the Eel Marsh result. Their samples are also dispersed powders, and the magnetization
is an IRM. The Eel Marsh curve plots lower than the published curve for the oxidized
basalt whose dominant magnetic grains are pseudo-single-domain and multi domain mag-
netite with strong shape anisotropy (7: 1). It coincides strikingly with the published curve
for the synthetic maghemite powder comprised of equant, single-domain grains, which
probably implies some similarity in grain size and coercivity between the Eel Marsh re-
manence carrier and the fine maghemite powder.
Figure 4c is adapted from Fig. 10 of Suthill et al. (1982). The published curve shows
the AF demagnetization behavior of an IRM; the sample is a magnetic concentrate from a
tidal flat silt. The remanence carrier is predominantly multi domain magnetite, but the
magnetic iron sulfide greigite is present and contains >8% of the total Fe in the sediment.
The Suthill et al. curve is the only published AF demagnetization curve for a sample known
to contain greigite. We hesitate to draw any particular conclusion from this comparison
except to note that the inflection points of the two curves are similar. Nevertheless, it
seemed helpful to compare our results with those of a sediment known to contain greigite.
4.1.3. Saturation Magnetization

Saturation magnetization intensity is weak but measurable for a sample of raw mud
(Fig. 5). The shape of the saturation curve is complex, but suggests the presence of a softer
component that saturates below 1000 G and a harder component that has not yet saturated
at 8000 G. The saturating component causes the change in slope of the curve between 50
and 200 G, while the nonsaturating component is responsible for the continued rise of the
curve at high fields.
Projecting back along the curve to the y axis, we can obtain a y-intercept value which
represents the magnitude of the total saturation magnetization for the sample's saturating
component (y = 0.036 emu/g). This figure allows us to approximate the amount of magnetic
material in the sample. If the saturating component is magnetite, with a saturation mag-
netization of 90 emu/g at 3000 K (Stacey and Banerjee, 1974), then the sample must contain
0.04% magnetite. If the saturating component is greigite, with a saturation magnetization
of 30 emu/g at 300 0 K (Spender et al., 1972), then the sample must contain 0.12% greigite.
These two minerals, along with maghemite and pyrrhotite, exhibit the high susceptibility,
low coercivity, and strong magnetization that characterize the remanence carrier in Eel
Marsh.
From these calculations, it appears that the magnetic material in the Eel Marsh mud
may be present in very small quantity. Much more hematite would be necessary to cause
the intensity of magnetization found in the sample. With a saturation magnetization of
0.04 emu/g at 3000 K (Stacey and Banerjee, 1974), hematite would have to comprise 9% of
the sample, and we might expect to detect it easily.
4.2. Separate
4.2.1. Saturation Magnetization

The saturation magnetization curves for two different separates are shown in Fig. 5.
The two samples differ in their separation procedures (see Methods). At seen from Fig. 5,
sample B contains a large paramagnetic component which causes the sample to resist
saturation. Its presence is probably a direct result of the separation procedure in which
sample B collected on the magnet over a longer period of time than sample A. In contrast,
sample A contains a well-defined, saturating component.
For both of these curves, we can approximate the amount of saturating magnetic ma-
terial as we did for the whole mud, by projecting back along the curve to the y intercept.
The y intercept for sample A is about 0.19 emu/g and for sample B, 0.83 emu/g. Hence,
634 Chapter 35
...•
... '.'" ,,'
B
..•.... ....
o ...•.
ú =.. ' .. ú =..
o
,.: .$ ....
....
.,
o
CQ
o
o A
C'I
..•. ..•..............•..............•....
.....•.. .•.
...•. " .•.... ill' " .....
ci
"
M
KWW?úW=..•...•...
1000 3000 5000 7000 gauss
H
Figure 5. Saturation magnetization curves for Eel Marsh. M, whole mud; A, separate A; B, separate
B.
sample A contains 0.2% magnetite or 0.6% greigite, while sample B contains 0.9% mag-
netite or 2.8% greigite. Comparing these values to the values for the whole mud obtained
in the previous section, we see that the magnetic fraction has been concentrated by ap-
proximately an order of magnitude (Table I).
Using another approximation, we have independent evidence for the effectiveness of
separation, from saturation remanence measurements of the mud before and after sepa-
ration. That is, rather than considering the concentration of magnetite in the separate, we
can consider the removal of magnetic material from the mud. As seen in Table I, this
comparison also suggests that the separation has decreased the amount of magnetic material
in the mud by an order of magnitude.
4.2.2. Thermomagnetic Behavior

The thermomagnetic behavior of the separate is complex. The magnetization-tem-
perature curve in Fig. 6 shows at least four magnetization intensity increases superimposed
on a generally declining intensity curve. These bumps probably record phase changes in
the sample with increasing temperature, despite a vacuum in the sample chamber. The
increase at 195°C may be due to the formation of pyrrhotite which has a Curie point of
326°C. The increase at 356°C is of unknown origin. The strong increase that peaks at 505°C
is probably due to the formation of magnetite, which has a Curie temperature near 575°C
where the curve again reaches a minimum. The rising tail on the curve above 586°C is
curious and may be due to the formation of elemental iron. Upon cooling, the sample has
Table I. Effectiveness of Separation Q
A. Based on Saturation Magnetization Data
Concentration of magnetic
Separate fraction
Mineral Whole mud A B A B
Magnetite 0.04% 0.2% 0.9% 5 23

Greigite 0.12% 0.6% 2.8% 18 24
B. Based on Saturation Remanence Data
SRM average S.D. Fraction of magnetic

Sample treatment (emu/g) (emu/g) N material in mud
Whole mud 0.578E-03 ±0.406E-03 10

Separated mud: 0.214E-04 ±0.141E-04 3 0.04
untreated
Separated mud: 0.310E-04 ±0.062E-04 2 0.05
ultrasonicated
Separated mud: 0.988E-04 ±0.445E-04 5 0.17
Calgon
a In part A, the separate contains roughly an order of magnitude more magnetic material than the whole mud. In
part B, the separated mud contains roughly an order of magnitude less magnetic material than the whole mud.
acquired an intensity of magnetization about 2.7 times larger than its original value. The
increase suggests that the sample has altered irreversibly to a new, more magnetic phase.
When the sample is rerun, it shows a Curie temperature of approximately 586°C and
increases in intensity above 645°C. None of the bumps in the initial curve are reproduced.
The sample appears to contain only magnetite in this run. Upon cooling a second time,
the sample has acquired a further magnetization which brings it to 4.3 times its initial
magnetization intensity, probably due to the formation of additional magnetite.
The complexity of the thermomagnetic behavior of the sample may be summarized
as follows:
1. The original magnetic material is unstable above 184°C where it undergoes a phase
change to pyrrhotite. This is probably not a Curie point, but rather a temperature
of instability under the vacuum conditions of the experiment.
2. The dominant magnetic phase in the original sample is probably not pyrrhotite or
magnetite, which form during the run.
3. The initial magnetic phase is probably a sulfide, as deduced from the formation of
pyrrhotite. If it were an oxide, pyrrhotite probably would not have appeared during
the transformations.
4.2.3. TEM
TEM gave important information as to the morphology of the grains in the separate.
Although it is strictly impossible to determine whether the grains in the photographs are
magnetic or not, it is a reasonable assumption based on (1) the procedure used to obtain
the sample, (2) the predominance of the depicted grains among other particles in the sam-
ple, and (3) the tendency of the particles to clump. Among the variety of shapes within
the sample, the illustrated particles predominate, and among the rest, a sizable number
are clay particles which are identified by their typical form and diffraction pattern.
636 Chapter 35
,
,
..
1.0
,,
.,
'.. . ,,
"
" .. -.,
....
"
.'........
c '" , . II.'
.., heat i ng---=--" ••• 1;., ••••
..,
'-.. 0.5
,,
cooling-' ,
"
,, ,
,
100 300 500 'e

Temperature
Figure 6. Thermomagnetic behavior of magnetic separate.
The grains have a cubic morphology with sides approximately 0.1 J.Lm (Fig. 7). They
appear to be uniform in both size and shape from grain to grain.
4.2.4. SEM and EDAX

The SEM was most useful for its EDAX information on the composition of the observed
grains. For the SEM, the copper TEM grid was mounted directly onto an aluminum SEM
stub.
During SEM scan, particles were selected for their visual properties and then analyzed
by EDAX. The selective criteria were: (1) size of less than 1 ILm, (2) equant form, and (3)
high conductivity (brightness on the screen). Of particles with the desired characteristics,
all had composition Fe + S. None lacked sulfur. As elements of low atomic number are
invisible to the electron beam, the EDAX would only detect Fe if the solid were magnetite.
A total of 65 particles were analyzed for composition. With experience, it became
possible to accurately predict the Fe + S particles because of their easily recognizable size
and brightness. Among particles that were randomly selected, most compositions suggested
a variety of iron silicate, including some of high Al content which were probably clays.
Several particles contained only Si among detectable elements, suggesting quartz. One
composition (Fe + Ti) suggested a detrital magnetite grain, too large to be of bacterial
origin.
The SEM work suggests that particles of iron sulfide composition are numerous and
distinctive within the separate. Submicron magnetite is apparently absent. Among the
Figure 7. Transmission electron micrographs of magnetic separate. Cubic mineral is greigite, with
dimensions of approximately 0.1 11m on a side. Magnification varies between photos.
rest of the particles, most appear to be iron silicates. Based on the absence of other can-
didates, the predominant magnetic material in the separate appears to be an iron sulfide.
4.2.5. Electron Diffraction

Electron diffraction on the particles in Fig. 7 yielded an uncalibrated diffraction pattern
which can be used comparatively. The procedure followed in evaluating the pattern (Towe,
this volume) is as follows:
1. The measured diameter (cm) of the brightest unknown ring is arbitrarily designated
d(lOO) .
638 Chapter 35
Table II. Predicted and Observed Ring Diameters for the

Electron Diffraction PatternU
Actual d Predicted Observed
(in A) Intensity (in cm) (in cm)
Greigite 2.980 100 assumed 7.24

1.746 75 4.24 4.22
2.470 55 6.00 6.10
3.500 30 8.51 8.38
1.901 30 4.61
1.001 30 2.43
Magnetite 2.532 100 assumed 7.24
1.485 40 4.25 4.22
1.616 30 4.62
2.967 30 8.48 8.38
2.099 20 6.00 6.10
1.093 12 3.12
a Measured ring diameters (em): 4.22, 6.10, 7.24, 8.38, 11.12, 12.58, 14.54,
15.20.
2. A known mineral is selected for comparison and a scaling constant is calculated

from the ratio of the unknown d(IOO) value (cm) to the reciprocal of the known
d(IOO) value (A).
3. A predicted diameter (cm) is calculated for each of the known d spacings (A) by
dividing the knowns by the scaling constant.
4. The predicted diffraction line diameters for the known mineral are compared to
the measured diameters of the unknown mineral.
The pattern was evaluated with respect to the minerals magnetite, greigite, pyrrhotite,
smithite, and pyrite. It matched poorly with all but magnetite and greigite, for which the
predicted diffraction line diameters match well with the observed (Table II). Because mag-
netite and greigite are isostructural, the relative spacing of their diffraction lines should
be the same and a match with both minerals assures that the unknown· mineral has an
inverse spinel structure. As the pattern is uncalibrated, it should not be possible to dis-
tinguish between them further. In practice, however, the greigite pattern appears to be a
closer match because its four maximum intensity lines are present, whereas for magnetite,
one of its more intense lines (30) is missing and a weaker line (20) is present.
Although no one piece of evidence has unequivocally identified greigite as the mag-
netic remanence carrier, the electron diffraction evidence for an inverse spinel structure
strongly supports this identification. Combined with EDAX evidence for an iron sulfide
composition, TEM evidence for cubic crystal form, and thermomagnetic evidence, the
cumulative evidence is strongly suggestive of greigite.
4.2.6. X-Ray Diffraction

X-ray diffraction on the magnetic separate fails to demonstrate the presence of greigite
or of any other magnetic phase. The X-ray pattern suggests the presence of chlorite-ver-
miculite, hornblende, epidote, possible illite, and quartz, all nonmagnetic phases. Hematite
may be present, but its peaks are masked by those of more abundant phases so that it
cannot be satisfactorily identified. Absent from the pattern are greigite, magnetite, goethite,
pyrite, and pyrrhotite.
It seems probable that the magnetic phase in the separate is present in quantities too
small for resolution by X-ray diffraction. As the magnetic phase apparently comprises less
than 2% by weight from saturation magnetization data, this result is entirely consistent.
4.3. Summary of Results
1. Eel Marsh mud is magnetic with directions of magnetization highly scattered

around present field direction.
2. It acquires a saturation remanence three orders of magnitude greater than its natural
remanent magnetization.
3. It may contain as little as 0.5% by weight of magnetic material.
4. The AF demagnetization behavior of the mud resembles published data for mag-
netite- and maghemite-bearing samples.
5. Magnetic separation removes approximately 90% of the magnetic material.
6. The magnetic separate is unstable above 184°C. At higher temperatures, it undergoes
irreversible phase transformations to pyrrhotite and magnetite.
7. The dominant magnetic mineral is present as regular cubic crystals of dimensions
approximately 0.1 IJ.m, with an iron and sulfur composition. Based on its mor-
phology, composition, and electron diffraction pattern, it is identified as greigite.
8. There is no bacterial magnetite present in the magnetic separate.
5. Discussion
The results of this study show that mud from a magnetotactic bacterial environment
can be magnetic without preserving bacterial magnetite. In Eel Marsh, at least part of the
remanence is carried by the iron sulfide greigite. In this section, we discuss the implications
of the presence of greigite.
Greigite and magnetite share the same cubic inverse spinel structure (Skinner et 01.,
1964), but differ in their anions. As a consequence, the two minerals appear to have similar
saturation remanence behavior, AF demagnetization behavior, and saturation magnetiza-
tion behavior. Although the magnetic moment of greigite is only one-third that of magnetite
due to enhanced electrical conduction in the sulfide lattice (Spender et a1., 1972), this may
not be recognized immediately in room-temperature magnetic tests, and the two minerals
may appear indistinguishable without compositional evidence. However, at higher tem-
peratures, their magnetic properties no longer coincide, for above 180°C, far below the
Curie point of magnetite (580°C), greigite becomes unstable and undergoes a phase trans-
formation to pyrrhotite.
If the depositional model for incorporating bacterial magnetite into sediments is ever
correct, then we must explain why magnetite was absent from the mud in Eel Marsh. The
most straightforward explanation follows from the difference in the stability conditions
of magnetite and greigite. As a sulfide, greigite should be stable under reducing conditions
with sufficient available S2-. On the contrary, the oxide, magnetite, should be stable under
low-sulfur, more-oxygen-rich conditions. In the laboratory, Berner (1964) found that greig-
ite precipitated in solutions saturated with H2S where iron was in the form Fe2+ and oxygen
was at low partial pressure. If the conditions varied, other iron sulfides like amorphous
FeS, mackinawite, pyrrhotite, or pyrite formed. Naturally occurring greigite has been found
in reduced varve layers from a Tertiary lacustrine sequence in California (Skinner et 01.,
1964), Pleistocene clay varves from Lake Superior (Dell, 1972), organic-rich mud from
Belgium (Jedwab, 1967), anoxic mud from Chesapeake Bay (Biggs, 1963), and anoxic mud
640 Chapter 35
Table III. Concentrations' Used in Making the

Stability Field Diagrams for Eel Marsh a .b
Ca 88 ppm Cl 3940 ppm
Mg 279 S04 529
Na 2170 Total C0 3 57
K 92 H2 S 9.8
Total Fe >0.247
pH = 6.65 Eh< -0.135

a Except for the Fe concentration and the pH. all values were
taken from A Coastal Pond (Emery. 1969). The two exceptions
were measured in Eel Marsh for this study.
b The procedure used to construct the stability field diagrams
is described in Appendix 2.
from the Black Sea (Berner, 1974); all from environments which would not appear to be
favorable to magnetite preservation or precipitation.
The stability field diagrams in Fig. 8 emphasize the bleak preservation potential of
magnetite in reducing environments like Eel Marsh. In these diagrams, the stability fields
for greigite, magnetite, pyrite, and goethite have been plotted for reducing conditions (Eh
< 0), through a range of pH. Ideally, the chemical information used in making these dia-
grams should have been obtained from a single water sample taken at the site in question.
Unfortunately, only pH and total iron measurements were obtained in Eel Marsh. Rather
than use arbitrary chemical information to complete our data, we chose to use Emery's
analysis published in A Coastal Pond (Table III). Emery's sampling site is located about
3.5 km from Eel Marsh and it resembles our site in that both contain a mixture of fresh
and ocean water, both are obstructed to some degree in their communication with the
ocean, and both are underlain by the same glacial deposits. Fortunately, only the total iron
and the total sulfur concentrations are critical to the placement of the stability field bound-
aries in the diagrams, and even so, variations in these concentrations up to a factor of five
do not cause significant changes in the diagram boundaries (Garrels and Christ, 1965).
Because of the apparent insensitivity of the stability field boundaries to moderate changes
in concentration, we feel that these diagrams should be essentially correct for Eel Marsh.
With regard to the Eh, the odor of H2 S, which was present at the sampling site, would
suggest that Emery's Eh measurement is appropriate as an approximation for that in Eel
Marsh, and in any case, the Eh should decrease with depth in the sediment (Emery, 1969).
Figure 8a illustrates the stability fields for greigite and magnetite. If we assume that
for Eel Marsh, pH = 6.65 (this study) and Eh decreases with depth from approximately
- 0.135 at the sediment-water interface, then greigite is more stable than magnetite. Be-
cause Eh decreases with depth, conditions are unlikely to become more favorable to mag-
netite preservation upon burial. Rather, any remaining magnetite should disappear and
greigite should be the sulfide that forms.
Figure 8a ignores the stability fields of pyrite and goethite. When these minerals are
included in the calculations, the stability fields of greigite and magnetite disappear (Fig.
8b). This follows directly from the magnitude and direction of the free energy change for
the reaction of greigite to pyrite and magnetite to goethite. Thermodynamically, pyrite and
goethite are the stable phases under Eel Marsh conditions. In Berner's (1964) experiments,
greigite converted to pyrite or pyrrhotite upon aging, and he concluded that it would not
persist in ancient sediments. If this is correct, then it may only exist metastably in Eel
Marsh, and it should disappear with time. For magnetite, the possibility of preservation
a.
ú=
Magnetite
-:
s:;
w
ú =
7 8
b.
>Q)
Goethite
Figure 8. Stability field diagrams for

Eel Marsh. Pyrite and goethite are
Pyrite
the thermodynamically stable
phases. but they are omitted from (a) Mte
so that metastable greigite and mag-
netite will appear (Berner. 1967. 5 6
1971; Garrels and Christ. 1965; pH
Langmuir. 1969).
seems equally grim. Under reducing conditions. it should dissolve and its iron reprecipitate
as a sulfide. Under less reducing conditions. it should convert to goethite. This is graph-
ically expressed in Fig. 8b where the magnetite stability field under Eel Marsh conditions
is limited to pH = 8 and Eh < - 0.210.
In an environment like Eel Marsh. where magnetite yields to greigite. both the original
DRM held in bacterial magnetite and the subsequent CRM held in greigite would be lost
with burial. In the rock record . lithified equivalents of Eel Marsh-type deposits. that may
have originally contained biogenic magnetite. would appear as nonmagnetic. sulfide-rich
peats. Only if the transformation of greigite were arrested. as appears to be the case in the
Black Sea (Berner. 1974). or if pyrrhotite were formed with the pyrite. would any mag-
netization be retained . In neither case would it be a bacterial remanence.
According to our analysis then. the normal order of events in Eel Marsh following
deposition of bacterial magnetite should be the dissolution of magnetite and the inorganic
precipitation of greigite. This greigite is likely to be in the form of very fine crystals (Berner.
1964), like those of near-micrometer size which caused a broadening of X-ray peaks for
Skinner et 01. (1964). Hence. the identification of fine crystals of greigite in Eel Marsh is
entirely consistent with an inorganic model for the formation of greigite.
It is intriguing to consider the possibility of bacterial mediation in the formation of
greigite. a possibility which is suggested by the work of Freke and Tate (1961). These
642 Chapter 35
authors report the precipitation of a strongly magnetic, fine-grained iron sulfide in an

anaerobic enrichment culture of the bacterium Desulfovibrio. Berner (1974) describes the
bacterially controlled sulfur cycle which appears to determine the formation and persist-
ence of greigite in the Pleistocene of the Black Sea. The rate and continuity of the sulfur
cycle is governed by bacterially mediated reactions (1) and (3):
(1)
HzS + detrital iron minerals ú =greigite, mackinawite (2)
(3)
SO + iron sulfides ú =pyrite, pyrrhotite (4)
In reaction (1), the bacterium Desulfovibrio reduces the Sm- ion for energy with
which to oxidize organic matter. The resulting production of HzS may be limited by a
short supply of p l ú =- , due to slowed diffusion from a high sedimentation rate, or an absence
of organic matter for oxidation by the bacteria. Once formed, HzS reacts rapidly (2) with
detrital iron minerals to form intermediate sulfides like greigite. Contingent upon the avail-
ability of excess HzS from reaction (2), the bacterium Thiobacillus oxidizes HzS to form
elemental sulfur (3). Like reaction (1), reaction (3) is limited only by insufficient Desul-
fovibrio activity. Elemental sulfur reacts (4) with the intermediate sulfides to make pyrite
or pyrrhotite.
The results of Freke and Tate (1961) apparently demonstrate an incomplete cycle
where the enrichment iron in the culture substitutes for detrital iron minerals. The presence
of greigite in Eel Marsh may suggest a similar sulfur cycle, particularly in light of its high
organic content and rich bacterial populations (Emery, 1969).
Whether bacterially mediated or inorganically controlled, the dissolution of bacterial
magnetite and the precipitation of greigite appear to be the normal processes acting in Eel
Marsh. Bacterial magnetite is not suited for preservation outside the cell under marsh
conditions, and it is unlikely to be found in lithified equivalents of Eel Marsh muds.
Apparently, the depositional model for incorporating bacterial magnetite into sediments
is not universal. If there is an environment for which it is correct, then that environment
remains to be found.
ACKNOWLEDGMENTS. I wish to thank Kenneth M. Towe of the Smithsonian Institution, Robert

F. Butler of the University of Arizona, Sherman Gromme of the U.S. Geological Survey,
Menlo Park, California, and Lisa Pratt and Robert Stoddard of Princeton University for
their help with the TEM work and electron diffraction analysis, the thermomagnetic runs
and thermomagnetic interpretation, and the X-ray diffraction work. I also thank George A.
Parks of Stanford University for help with geochemical considerations and Robert B. Har-
graves of Princeton University, under whose direction this study was begun.
This research was supported in part by National Science Foundation Grant EAR
8008207.
Appendix 1: Eel Marsh NRM and Saturation Magnetization Data
Depth in
Sample No." core Dec!. Inc!. Total moment Saturation moment
013 0.0 em 233.2° 53.7° 0.202E-06 emu 0.890E-03 emu

012 3.0 275.1 15.4 O.310E-06 O.740E-03
011 5.5 12.6 43.8 0.374E-06 0.903E-03
010 8.5 135.7 40.8 0.938E-06 0.636E-03
009 11.0 57.4 76.9 0.273E-06 0.791E-03
008 13.75 326.1 54.9 0.286E-06 0.106E-02
007 16.5 15.2 30.0 O.692E-06 0.119E-02
006 19.5 32.4 80.5 0.162E-06 0.704E-03
005 22.0 171.2 79.9 0.369E-06 0.929E-03
004 24.75 8.3 76.9 0.498E-06 0.874E-03
003 27.5 57.9 82.2 0.510E-06 0.725E-03
002 30.5 294.8 14.8 0.142E-05 0.802E-03
001 32.5 37.4 71.6 0.212E-06 0.577E-03
109 0.0 72.0 52.3 0.165E-06 0.625E-03

108 4.0 23.9 59.9 0.339E-06 0.102E-02
107 8.5 321.2 70.2 0.533E-06 0.629E-03
106 13.0 164.7 7.1 0.638E-06 0.126E-02
105 16.0 255.8 86.4 0.110E-06 0.554E-03
104 20.5 84.5 58.2 0.262E-06 0.107E-02
103 22.5 104.8 71.5 0.334E-06 0.949E-03
102 25.0 35.2 34.5 0.476E-06 0.505E-03
101 27.5 33.7 64.3 0.298E-06 0.702E-03
Averages 13.8 72.4 0.427E-06 emu 0.824E-03 emu
sample box (6.6 em 3 ): 0.647E-07 emu/em 3 0.436E-05 emu/em 3
Fischer statistics: K = 4.6 0.95 = 15.9
a Samples 001-013 are from the western half of the split core and samples 101-109 are from the eastern half.
644 Chapter 35
Appendix 2: Description of Computer Procedure Used to Make

Stability Field Diagram 8a
Given:
1. Total S, total Fe
2. Constituents that determine salinity
3. Constituents that may alter pH through hydrolysis
4. Constituents that may complex with S or Fe
Constraints:
1. Mass balance (for each constituent)
2. Reaction and equilibrium constants
Procedure:
1. Estimates ionic strengths
2. Corrects equilibrium constants for ionic strengths
3. Solves reactions, equilibria, mass balance (for S)
4. Recalculates ionic strengths
5. Iterates until changes in ionic strengths are negligible
6. Considers list of given solids (mackinawite, greigite, magnetite, sulfur, amorphous
FeS, Na,K,H-jarosite) and "precipiates" the minimum compatible set of solids from
the given volume of water
With the computer data from repeated runs through a range of Eh and pH, we can
complete a diagram like Fig. 8a.
References
Balkwill, D. 1., Maratea, D., and Blakemore, R P., 1980, Ultrastructure of a magnetotactic spirillum,
J. Bacteriol. 141:1399-1408.
Berner, R A., 1964, Iron sulphides formed from aqueous solution at low temperatures and atmospheric
pressure, J. Geol. 72:293-306.
Berner, R A., 1967, Thermodynamic stability of sedimentary iron sulfides, Am. J. Sci. 265:773-785.
Berner, R A., 1971, Principles of Chemical Sedimentology, McGraw-Hill, New York, pp. 115-137,
193-209.
Berner, R A., 1974, Iron sulphides in Pleistocene deep Black Sea sediments and their paleooceano-
graphic significance, in: The Black Sea-Geology, Chemistry, Biology, AAPG Memoir 20, pp.
524-531.
Biggs, R B., 1963, Deposition and early diagenesis of modern Chesapeake Bay muds, Ph.D. dissertation,
Lehigh University, unpublished.
Blakemore, R P., Maratea, D., and Wolfe, R S., 1979, Freshwater magnetic spirillum, J. Bacterial.
140:720ff.
Creer, K. M., 1974, Geomagnetic variations for the interval 7000-25,000 yr. B. P. as recorded in a core
of sediment from station 1474 of the Black Sea cruise of "Atlantis II," Earth Planet. Sci. Lett.
23:34-42.
Dell, C. I., 1972, An occurrence of greigite in Lake Superior sediments, Am. Mineral. 57:1303-1304.
Doyle, 1. J., Hopkins, T. 1., and Betzer, P. R, 1976, Black magnetic spherule fallout in the eastern
Gulf of Mexico, Science 194:1157-1159.
Geophys. 7:709-757.
Emery. K. 0., 1969, A Coastal Pond, Elsevier, Amsterdam.
Science 203:1355-1356.
Frankel, R B., Blakemore, R P., Torres de Araujo, F. F;, Esquivel, D. M. S., and Danon, J., 1981,
Freke, M., and Tate, D., 1961, The formation of a magnetic iron sulphide by bacterial reduction of
iron solutions, J. Biochem. Microbiol. Technol. Eng. 3:29-39.
Garrels, R M., and Christ, C. L., 1965, Solutions, Minerals, and Equilibria, Harper & Row, New York,
pp. 178-223.
Jedwab, J., 1967, Mineralisation en greigite de debris vegetaux d'une vase recente (Grot Geul) , Soc.
Belg. Geol. Bull. 76:1-19.
Kirschvink, J. L., 1980, South seeking magnetic bacteria, ]. Exp. Biol. 86:345-347.
r.::irschvink, J. L., 1982, Paleomagnetic evidence for fossil biogenic magnetite in western Crete, Earth
Paleomagnetic, sedimentologic, and biologic implications, Earth Planet. Sci. Lett. 44:193-204.
Langmuir, D., 1969, The Gibbs free energies of substances in the system Fe-Oz-HzO-CO z at 25°C,
U.S. Geol. Surv. Prof. Pap. 650-B:B180-B184.
Lovlie, R, Lowrie, W., and Jacobs, M., 1971, Magnetic properties and mineralogy of four deep sea
cores, Earth Planet. Sci. Lett. 15:157-168.
Am. Bull. 73:435-438.
Lowrie, W., and Fuller, M., 1971, On the alternating field demagnetization characteristics of multi-
domain thermoremanent magnetization in magnetite, ]. Geophys. Res., 76:6339-6349.
Mackereth, F. J. H., 1971, On the variation in direction of the horizontal component of remanent
magnetization in lake sediments, Earth Planet. Sci. Lett. 12:332-338.
Moench, T. T., and Konetzka, W. A., 1978, A novel method for the isolation and study of a magnetotactic
Oldfield, F., Thompson, R, and Barber, K. E., 1978, Changing atmospheric fallout of magnetic particles
recorded in recent ombrotrophic peat sections, Science 199:679-680.
Rimbert, F., 1959, Contribution a l'etude de l'action de champs alternatifs sur les aimantations re-
manentes de roches, Rev. lnst. Fr. Petrole. Ann. Combust. Liquides, 14:17ff.
Skinner, B. J., Erd, R c., and Grimaldi, F. S., 1964, Greigite, the thiospinel of iron; a new mineral,
Am. Mineral., 49:543-555.
Spender, M. R, Coey, J. M. D., and Morrish, A. H., 1972, The magnetic properties and Mossbauer
spectra of synthetic samples of Fe 3 S4, Can. J. Phys. 50:2313-2326.
Stacey, F. D., and Banerjee, S. K., 1974, The Physical Principles of Rock Magnetism, Elsevier, Am-
sterdam.
Stober, J. C., and Thompson, R, 1-979, An investigation into the source of magnetic materials in some
Finnish lake sediments, Earth Planet. Sci. Lett. 45:464-474.
Suthill, R ]., Turner, P., and Vaughn, D. ].,1982, The geochemistry of iron in recent tidal flat sediments
of the Wash area, England, Geochim. Cosmochim. Acta 46:205-217.
Thompson, R, 1973, Paleolimnology and paleomagnetism, Nature 242:182-184.
Thompson, R, and Morton, D. ]., 1979, Magnetic susceptibility and particle size distribution in recent
sediments of the Loch Lomond Drainage Basin, Scotland, J. Sediment. Petrol. 49:801-812.
Thompson, R, Batterbee, R W., O'Sullivan, P. E., and Oldfield, F., 1979, Magnetic susceptibility of
lake sediments, Limnol. Oceanogr. 20:687-698.
Towe, K. M., and Moench, T. T., 1981, Electron optical characterization of bacterial magnetite, Earth
Chapter 36
Possible Biogenic Magnetite Fossils
from the Late Miocene Potamida Clays
of Crete
SHIH-BIN R. CHANG and JOSEPH L. KIRSCHVINK
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 647
2. Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 648
3. Laboratory Extraction of Magnetite. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 650
4. Magnetic Studies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 653
5. Size and Shape Distribution of Magnetite. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 654
6. Origin of Magnetite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655
7. Conclusion and Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 666
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 667
1. Introduction
In the 23 years since Lowenstam (1962) first discovered the mineral magnetite in chiton
teeth, many other organisms have been reported to be able to form this mineral as well
(Blakemore, 1975; Gould et al., 1978; Frankel et a1., 1979; Walcott et al., 1979; Kirschvink,
1981a; Walker and Dizon, 1981; Lins de Barros et a1., 1981). Magnetite is now the fourth
most common biogenic mineral after carbonate, opal, and ferrihydrite and related ferric
oxide in terms of its production by different groups of organisms (Lowenstam and Weiner,
1982). A variety of magnetite-forming organisms live in aquatic environments and hence
there is the question whether magnetite formed by organisms can be preserved in sedi-
ments. In particular, one group of magnetite-synthesizing organisms, the magnetotactic
bacteria, are cosmopolitan in their aquatic distribution. Based on calculations considering
their natural population density, sedimentation rates, and volume of magnetite per cell,
the biologic contribution of magnetic remanence in sediments has been estimated by
Kirschvink and Lowenstam (1979) to reach the 10- 4 Aim level. Towe and Moench (1981)
revised this estimate upwards to 10- 3 Aim which is more compatible with the remanence
generally observed in sediments.
Recently, a variety of paleomagnetic studies have attempted to unravel the detailed
behavior of the geomagnetic field during geomagnetic polarity transitions as recorded in
marine and lacustrine sediments and volcanics (e.g., Valet and Laj, 1981; Bogue and Coe,
1982). The reliability of sediments to record accurately the geomagnetic field depends on
the constant influx of suitable magnetic particles; consequently, many authors have care-
SHIH-BIN R. CHANG and JOSEPH L. KIRSCHVINK • Division of Geological and Planetary Sciences,
California Institute of Technology, Pasadena, California 91125.
647
648 Chapter 36
fully studied the magnetic mineralogy of sediments deposited during transitions. Of par-
ticular interest in the search for fossil biologically formed magnetite is the work of Langereis
(1979) and Valet and Laj (1981), on the marine clays of the Potamida section in north-
western Crete. During the two geomagnetic transitions studied, they found that both the
anhysteretic remanent magnetization (ARM) and the saturation isothermal remanent mag-
netization (sIRM) decreased, indicating that the sedimentary magnetic fraction was tem-
porarily reduced. Closer examination of their data also reveals that the ARM/sIRM ratio
decreased during the transitions. Because sIRM is proportional to the total amount of mag-
netic mineral and the ARM level is biased toward the extremely fine-grained (SD or PSD)
fraction, a drop in the ARMlsIRM ratio suggests that the concentration of fine-grained
material disappeared during the reversal. In addition, the IRM acquisition curve argues
that magnetite might be the major carrier of the stable magnetic remanence in this clay
section.
The suggested decrease of fine-grained magnetic material associated with the geo-
magnetic transition led Kirschvink (1982) to propose that some of the magnetic material
normally contributing remanence to the sediments was bacterially produced, and that the
period of reduced geomagnetic intensity surrounding the transition led to a decrease in
the bacterial contribution to the sediments. Several lines of evidence suggest that the se-
lective advantage of magnetotaxis would be diminished during periods of low geomagnetic
intensity (Kirschvink and Gould, 1981). This hypothesis would predict the decrease of
fine-grained magnetite particles observed during the transitions preserved in the Potamida
clays of Crete. This hypothesis leads to two testable predictions. First, SD magnetite crystals
with characteristics similar to those of extant magneto tactic bacteria ought to be present
in the nontransitional portions of the Potamida clays and extensive purification followed
by electron microscopy should reveal their presence. Second, if such particles are indeed
found, their absolute concentration should decrease during the transition. A test of this
first prediction is the focus of this study; an answer to the question of fluctuations in the
bacterially formed magnetite concentration requires the development of a quantitative
method for assessing the absolute abundance of bacterially formed magnetite in sediments.
Previous results with scanning electron microscopy on magnetic extracts from modern
sediments and ancient sedimentary rocks (Lovlie et aI., 1971; Lowrie and Heller, 1982;
Demitrack, this volume) indicate that this technique does not have the ability to resolve
the morphology and internal characteristics of magnetite grains in the SD size region (D.l
fLm). The higher resolution of transmission electron microscopy (Towe, this volume) is
required. Furthermore, it is necessary to eliminate all minerals except magnetite prior to
electron microscopic examination; if this is not done, the identity of ferromagnetic phases
will remain in question.
The purpose of this study is to search for biogenic magnetite in the Potamida clays.
We developed a new extraction technique which efficiently separates fine-grained mag-
netite from all the other ferromagnetic minerals in the sediments. The recovery and iden-
tification of bacterial fossils is an important test of the hypothesis that biochemically
formed magnetite contributes a substantial fraction of the magnetic material in marine
sediments. In addition, the ability to recognize these fossils in old sediments would help
unravel when the magnetotactic bacteria evolved and may ultimately shed light on the
origin of iron-mediating enzymes.
2. Samples
Eighteen samples from the Potamida clays were supplied by Drs. Zijderveld and C.
G. Langereis at the State University of Utrecht. These cores span 4D m of strata which
Magnetofossils in Crete 649
P010mido Sec1ion Polari1y Skouloudhina Sec1ion
Langereis' Sample Sample

Measurements Number 40 Number
90 FeV
N
, NRM Intensity 85
FeIV 35
• NRM (50 mT) 80
75
Feill R
70 30
Epoch 5
65 ]
:c 25
'"W
I N
u
.2 O2A
i
C. IIA
ú= 20 16A
Fell 21A
'"
ú=
if,
15
ú= 268
31A
6CC
758
1018
R
10
FeI
o 5 10 15 20 25 o 10 20 0 10 20 30 o 10 20 30 o 10 20
Magnetic Intensity First Extract Bulk Sample Intensity Bulk p ~ ú ä É = Intensity First Extract
(xI0-3 Aim) IntenSJty/Bulk (XI0-8 Am2/g) (xIO- Am2/g) Intenslty/Bulk
Sample Intensltv Sample Intensity
Figure 1. Variations of magnetic intensity of bulk sample and the ratio of first extract intensity and
bulk sample intensity for the Potamida and the Skouloudhina sections. Langereis's measurements
(personal communication) have been listed for comparison.
correspond to a continuous sedimentary record of approximately 1 My and have a relatively

high sedimentation rate of 4 cm/lOOO years. Drooger et al. (1979) provide a detailed geologic
description of the sampling locality, and Langereis (personal communication) has meas-
ured the magnetic intensities of these 18 samples before and after 50-mT alternating field
(AF) demagnetization (shown in Fig. 1). The NRM-intensity decrease occurs in all three
transitions, which is consistent with the result of Valet and Laj (1981). The sedimentary
environment of the Potamida clays has been determined by facies analyses to be an open
basin (Meulenkamp et 01., 1979). Detailed biostratigraphic and magnetostratigraphic cor-
relations indicate that this clay was deposited between 6.3 My and 5.3 My ago (Tortoni an/
Messinian; Langereis and Zachariasse, 1981). Small evaporite deposits have been observed
near the Potamida section (Meulenkamp et 01., 1979), and studies of the large-scale tectonic
history of this area show that this sedimentary basin was connected to the Mediterranean
during the late Miocene Messinian salinity crisis (Meulenkamp et 01.,1979; Drooger et 01.,
1979). The whole sequence of this section appears to be continuous. Five iron-rich layers
have been found separately around sampling localities 15,45, 70, 80, and 85. One of them
(sampling locality 45) corresponds to the distant part of one turbiditic layer but represents
no sedimentary hiatus (Drooger et 01., 1979). An abrupt appearance of Globorotalia con-
nomiozea (TortonianlMessinian boundary) has been reported just above 70 (Zachariasse,
1979).
In addition to the Potamida section, another section at Skouloudhina in northwestern
Crete recorded the same R-N transition as Potamida, but did not show the same ARM drop
(Laj, personal communication; Valet et 01.,1983). Nine cores of this section have been sent
by Dr. Laj in France, and the grain-size distribution and major mineral occurrences of the
650 Chapter 36
two sections show no obvious contrast. The upper five levels of the Potamida section
samples available for this study have relatively larger grains. The average grains of the
other cores are mainly below 10 11m in size. Kaolinite, illite, calcite, and quartz are the
dominant mineral phases observed in all the samples.
3. Laboratory Extraction of Magnetite

The extraction technique used in this study is based on the physical and chemical
characteristics of magnetite: high magnetization, high density, and resistance to dithionite-
citric acid dissolution. Although this technique may still need revision and improvement
before it can be more widely applied, it is at present the most efficient way to extract pure
magnetite from poorly consolidated sedimentary samples and works well on the material
from Crete.
The extraction begins by preparing a fine-grained slurry from the dry mud by adding
distilled water, and stirring the resulting mixture evenly. An initial magnetic separation
is made using a 60-ml manual speed-controlled separatory funnel attached to a Franz
isodynamic magnetic separator with a rubber tube fitted with a clamp to allow a slow flow
of water through the system. The current of the separator should be set at about 0.5 A. The
slurry passes through this funnel slowly, and all of the magnetic grains will be held at the
side of the funnel. After this step, the average magnetic intensity of the sample will increase
from about 1-2 x 10- 4 A m 2 /kg for bulk material to 2-4 x 10- 3 A m 2 /kg.
A second magnetic separation is done by transferring the first magnetic separate into
a 10-ml conical centrifuge tube with distilled water. In order to disperse the quartz and
clay mineral grains which clump together with magnetite, the tube is placed in an ultra-
sonic shaker for about 10 min. The highly magnetic grains can then be pulled to the edge
of tube by attaching a small magnet to the side and gently sloshing the fluid/sediment
mixture back and forth. After discarding the supernatant liquid and washing the residue
twice more, the magnetic intensity of the resulting magnetic separate is in the range of 5-
8 X 10- 2 A m 2 /kg.
Next, the residual material is cleaned off all other ferric iron oxide minerals by treating
it for 2 days in a buffered sodium dithionite-citrate solution (Mehra and Jackson, 1960).
This solution under these conditions is known to dissolve hematite, maghemite, and the
other possible iron oxide minerals in the samples, but will not attack magnetite at room
temperature for several months (Kirschvink, 1981b). Previous work has shown that he-
matite, maghemite, and goethite dissolve quickly in this solution and we have experi-
mentally established that it works for pyrrhotite as well. A TEM control experiment on a
fine-grained magnetite standard powder confirmed that there was no observable change
after 2 days of dithionite treatment. This is the only chemical treatment used in the whole
extraction procedure and it should not alter the morphology of magnetite, unless the surface
has been oxidized to form a thin layer of maghemite.
After 2 days of chemical dissolution, the solution is centrifuged, the liquid discarded,
and the sample ultrasonically resuspended and washed several times with distilled water.
A final magnetic extraction is done by pipetting the solution onto a glass microscope slide
and letting it form a thin film with most of particles in the center. A small hand magnet
can then be used to work the magnetic particles away from the center leaving any non-
magnetic grains behind.
Samples were prepared for TEM using the procedures described by Towe (this vol-
ume), with one minor modification. Clumps of magnetic particles were partially dispersed
using a 100-mT, 400-Hz peak-to-peak oscillating magnetic field produced by our AF de-
magnetizing unit. In theory, this procedure should first break the clumps into mutually
Figure 2. X-ray diffraction patterns of (a) magnetite standard and
(b) final extract of t4e Potamida clays.
652 Chapter 36
Figure 3. Electron diffraction patterns of grains in the magnetic extract prior to dithionite treatment
from the Potamida clays in Crete. The wavelength of electrons is 6 pm, and the camera length is 180
mm. Positive was enlarged 7 times.
repelling particle chains aligned parallel to the oscillating field. As these chains settle
through the liquid onto the grids, the oscillating field hits them perpendicularly to their
length, and will free some of the individual crystals. This works fairly well in practice.
A check of the purity of the resulting powder was made using the standard Debye-
Scherrer method. Almost all diffraction lines from our sample matched those from a pure
100
Figure 4. AF demagnetization curve for
90
final extract from the Potamida clays.
Due to low demagnetization resistance 80
of MD magnetite and strong interacting 70
field of dense-packed SD magnetite, ú =
0:: 60
sIRM demagnetization normalized re-
manence for both MD and interacting <fl 50
SD magnetite grains will increase ab- 0ú = 40
ruptly (Cisowski, 1981). The soft part of 30
this final extract, which corresponds to
20
the abrupt decrease of % sIRM during
< 100-G demagnetization, can be char- 10
acterized as MD or interacting SD mag- 0
netite. The smooth decrease of % sIRM 0 100 200 300 400 500 600 700
at demagnetization level > 100 G is Demagnetization Field (Gauss)
characteristic of SD magnetite.
magnetite standard (Fig. 2). Cell dimensions and mole fraction of Fe zTi0 4 of these mag-
netite extracts can be determined by comparing their diffraction patterns with those of the
silicon standard as 8.39 A and 0.02. Before surveying the final magnetic extracts of all the
samples, we also examined the magnetic extract, prior to dithionite treatment under the
TEM. The electron diffraction patterns (Fig. 3) of the dominating dark mineral grains in
the extract confirm that most, if not all, of the material in the samples is magnetite. These
two direct observations are consistent with the conclusion derived indirectly from the
magnetic measurements (Langereis, 1979; Valet and Laj, 1981).
4. Magnetic Studies
Another measure of the concentration of fine-grained magnetite in each sample is the

ratio between the IRM intensity of the first extract and that of the bulk sample. Figure 1
shows variations in this ratio as well as those for the magnetic intensity of bulk material
for each sample of both Potamida and Skouloudhina sections. Both of these parameters
decrease during the lower two transitions at the Potamida section, but no such decrease
is observed in the corresponding R-N transition at the Skouloudhina section. The intensity
also decreases during the upper transition of the Potamida section (Langereis, personal
communication). These data are consistent with the disappearance of large amounts of SD
magnetite in some parts of the formation.
We also ran a progressive AF demagnetization of the sIRM on the final extract from
the lower portion of the Potamida section (Fig. 4). Two magnetic fractions are clearly
present, a relatively soft one with coercivity less than 10 mT which probably is a result
of MD grains, and a magnetically harder portion corresponding to smaller, presumably SD
grains.
A summary of all the data obtained from Valet and Laj's (1981) work and this study
gives us an overview of the magnetic mineral phase variations in these sections:
1. Magnetite is the major remanence carrier in both sections.
2. A large amount of the fine-grained (probably SD) magnetite disappears during the
lower two transitions of the Potamida section.
3. The remanence of the upper third of the Potamida section and most of the Skou-
loudhina section is probably mainly due to large-grained (PSD or MD) magnetite
or other magnetic minerals.
654 Chapter 36
ú = [j)
1.00 r--r---.,-,.--r-,----,--r---r---,--.---. 103
Multi -Domain úQ)=
0.50 Qj
Two Domain E
if) oc
c a
0
U Z
ú= .r:::.
-£
0.10 10 2 gQ)
0'>
c
Q)
ú =
Figure 5. Plotted on this domain sta-
0.05 bility field diagram for magnetite (But-
-
ú = ú =
Kú =
Q)
'-
1er and Banerjee, 1975), the size and
U a
'+"
..... 0... shape distribution of fine-grained mag-
a netite in the final extract of samples
0...
;: others Superparamagnetic from the lower half of the Potamida sec-
0.0 1ú J D J J J f J J J J D | K ä K K K K J J I J J J i | I J J K K K K f J J J I J J J J f = 10 I tion shows overlapping with the re-
0.0 0.2 0.4 0.6 0.8 1.0 ported size and shape distribution of
Axial Ratio (width/length) bacterially formed magnetite.
These phenomena do not unambiguously constrain the origin of the magnetite in these
clays. However, the other possible sources (e.g., volcanic eruptions, weathering, falling of
cosmic dust, etc.) would not lead to a drop in concentration of magnetite during magnetic
reversal as would the biogenic hypothesis.
5. Size and Shape Distribution of Magnetite

The detailed size and shape distribution of the final magnetite extract from samples
5, 15, 25, 35, 45, and 60 of the Potamida section and sample 21 B of the Skouloudhina
section have been studied by TEM. The majority of the extract grains in the first five samples
are of SD size. The latter two samples are primarily composed of TD or MD magnetite.
These direct magnetic phase variation observations reinforce arguments based on magnetic
measurements in these formations. A plot of the general size and shape distribution of
grains from the final extract of the first five samples on the magnetite domain stability
diagram of Butler and Banerjee (1975) shows that the range of grain size from these extracts
overlaps with the size and shape distribution of bacterial magnetite (Fig. 5).
Another characteristic feature used to distinguish biogenic magnetite from inorganic
magnetite is its morphology. Both the hexagonal shape reported by Towe and Moench
(1981) and the teardrop shape found by Blakemore et a1. (1980) are diagnostic.
Surveying the grains in the final magnetic extract of the sample, we found that they
can be grouped into four modes of occurrence: The first is composed of rounded or elongate
grains with frothy surfaces (A; Figs. 6, 7). The modal size distribution is from 0.1 to 0.2
f,Lm, but occasionally, large grains of up to 1 f,Lm have been found. Euhedral crystals with
typical octahedral (B) or hexagonal (C) shapes in their microscopic image are also fre-
quently observed (Fig. 8). The smooth surfaces of these two types of grains suggest amaz-
ingly good preservation of magnetite in the Potamida clay. The size of the octahedral grains
varies from 0.05 to 0.2 f,Lm. In addition to their shape, a narrow range of size distributions
(0.1 ± 0.02 f,Lm) is another intriguing feature of the hexagonal grains. A very small number
of the fourth type of grains consisting of crystals with prismatic or oval forms (D; Figs. 9,
10), are compatible with anomalous shapes of magnetite found in bacteria (cuboidal and
teardrop; Blakemore et a1., 1981) and turtles (spherical; Perry et a1., this volume).
Figures 6-17. Transmission electron micrographs of the final extract from samples of the Potamida
section in Crete (Figs. 6-10, 12-17) and magnetite standard (Fig. 11). Scales are variable. Arrow in
Fig. 13 indicates a possible characteristic of pseudohexagonal grains-an uneven crystal face.
Figure 6. 392,000 x.
6. Origin of Magnetite
Grain-size distribution histograms have long been used by sedimentologists as a prove-
nance indicator. Applying this to delineate the source of ultrafine magnetite in sediments
is very important from a paleomagnetic point of view. Due to its high susceptibility and
resistance to demagnetization, the presence of SO magnetite makes a significant contri-
bution to the remanent magnetism of the rocks. In addition, the small size of SO magnetite
particles ensures that it will have a slow settling velocity in aquatic environments. De-
positional effects should not affect its alignment along the magnetic field.
SO magnetite particles have been observed directly by TEM in a variety of igneous
rocks. Evans et 01. (1968) and Evans and Wayman (1970) found such particles as inclusions
656 Chapter 36
Figure 7. 172,000 x.
in pyroxene and plagioclase grains in gabbroic intrusions Similarly, Smith (1979) found
them as an auxiliary phase in glass of a midocean ridge basalt and Geissman et al. (1983)
have reported it from an ash flow-tuff. However, the high Ti contents of these grains dis-
tinguish them from SD magnetite grains observed in this study. Various authors have used
the revised Lowrie-Fuller test (Johnson et al., 1975) as a tool to determine the existence
of SD magnetite particles in various types of sediments (e.g., limestone, Lowrie and Heller,
1982; lake sediments, Stober and Thompson, 1979), but no TEM observations have been
made in these studies and the size and shape distribution of the grains inferred to be present
cannot be determined precisely. Of all the common magnetic mineral phases in soils tab-
Figure 8. 586,000 x .
658 Chapter 36
oI 0.1 o, 2
Figure 10. 248,000 x .
ulated by Schwertmann and Taylor (1977), magnetite was not listed. However, Ozdemir
and Banerjee (1982) found ultrafine-grained magnetite particles with a size distribution in
the SD to PSD stability field as a major magnetic component in the soil samples from west-
central Minnesota. The other two occurrences of SD magnetite are biogenic and synthetic.
Because the magnetic extraction techniques for sediments are still under development.
reports on other occurrence of SD magnetite in nature are to be expected in the future. For
the purpose of this study, one obvious question to be raised is, what are the sources of SD
magnetite in marine sediments?
Magnetite in marine sedimentary environments might have one or more of the fol-
lowing origins: (1) cosmic, (2) volcanic, (3) terrigenous, (4) submarine weathering, (5) hy-
drothermal, (6) diagenetic, (7) hydrogenous (authigenic), and (8) biogenic. Industrial fallout
(Doyle et a1 ., 1976) should not have made any contributions to the magnetic mineralogy
in ancient sedimentary rocks .
Magnetic spherules (10-200 f.Lm in diameter) consisting of metallic iron nuclei sur-
rounded by a shell usually of magnetite, have been observed in recent sediments (Chester
and Aston, 1976). Although their origin is not clear, cosmic production has been proposed
(Castaing and Frederickson, 1958) . Their scarcity implies that they are not an important
source of magnetite in sediments.
Terrigenous magnetic phases playa major role as the remanence carrier in some marine
sediments (Verosub, 1977). Rounded and eroded shapes would presumably be a result of
water transport while sharp and angular forms would suggest eolian transport of volcanic
dust, for example. Magnetic studies have confirmed the presence of MD magnetite in a
variety of marine sedimentary rocks (Lowrie and Heller, 1982; Ensley and Verosub, 1982)
and deep-sea cores (Lovlie et 01., 1971). Rounded grains with pitted surfaces observed by
SEM imply these magnetite particles are of detrital origin. Again, their size (0.2-30 j.Lm)
is larger than the size of magnetite grains extracted from the Potamida clays. Nevertheless,
the low resolution of SEM prevents delineation of detailed grain size distribution histo-
grams for grains less than 1 j.Lm.
Magnetite might be formed in hydrothermal systems by the reaction (Shanks et 01.,
1981)
(46 - 0.5x)Fe zSi04 + U p M ú H = + 8Mgz+ + 4H zO

= 4FeSz + 4FeMgzSi4 0 lO (OHlz + (28 - X)Fe304 + xFeZ03 + (30 - 0.5x)SiO z(aq)
or it might be formed by submarine weathering of oceanic basalt (Fyfe and Lonsdale, 1981).
The occurrences of magnetite in Red Sea hydrothermal system sediments (Hackett and
Bischoff 1973) and manganese nodules (Carpenter et 01., 1972) have been confirmed by
magnetic measurements, but no discrete magnetite grains have been observed. Geologic
evidence also does not support the presence of a hydrothermal system around Crete during
the late Miocene.
Diagenetic magnetite is thermodynamically permissible in anoxic marine sediments
(Berner, 1964). Recently, botryoidal and spheroid magnetite crystal aggregates (3-20 j.Lm
;:'1 diameter) found in the upper Silurian and lower Devonian Helderberg limestones and
the Cambrian Bonneterre dolomite have been inferred to be of diagenetic origin (McCabe
et 01., 1983). However, geochemical models of diagenetic magnetite formation are not yet
well developed. In addition, the size (0.5-2 j.Lm) of their individual crystallites does not
fall in the SD field.
As mentioned before, a systematic survey of the size and shape distribution of ul-
trafine-grained magnetite of different origins has not yet been made. Therefore, the above
comparison does not exclude any of the six occurrence modes (cosmic, volcanic, terri-
genous, hydrothermal, submarine weathering, and diagenetic) as a possible source of mag-
netite in the Potamida clays. However, the comparisons do suggest that the majority of
these magnetite grains are probably are of hydrogenous or biogenic origin.
Henshaw and Merrill (1980) recently developed an Eh-pH stability diagram for iron
phases in marine depositional environments, using the actual average concentration of S
and Fe in the ocean and revised energy data for iron phase reactions. Their work indicates
that magnetite may be an authigenic phase under suitably reducing conditions. Earlier,
Harrison and Peterson (1965) suggested that the major remanence carrier in one of the
Indian Ocean cores might be of authigenic origin, based on an unusual vertical elongation
of the susceptibility anisotropy ellipsoid. However, no mechanism has been suggested to
relate this abnormal susceptibility anisotropy with authigenesis. In any case, these analyses
contradict the traditional view that magnetite is not a stable phase in most aqueous en-
vironments (Garrels and Christ, 1965).
On the other hand, SD magnetite has been successfully precipitated in the laboratory
(Sugimoto and Matijevic, 1980) under conditions of slow reaction rate and relatively high
temperature. These conditions rarely occur in natural depositional environments, but the
similarity in shape of the type B grains to the magnetite standard (Fig. 11) strongly argues
for its inorganic origin. Two possible biologically mediated mechanisms might promote
this reaction directly or indirectly.
660 Chapter 36
o 0 1 0.2 J4IA
IL-_-"-_--J.
Figure 11. 116,000 x.
Iron bacteria, a type of chemoautotrophic microorganism with a typical body size from
1 to 10 11m, will oxidize ferrous ion in the water and obtain energy from this reaction. The
oxidized product of this biologic mechanism is generally believed to be an iron hydroxide
(Ehrlich, 1981). Because a conclusive phase identification has not been made, other pos-
sibilities cannot be excluded.
The other possible way to precipitate iron oxide from seawater is through iron co-
agulation with organic material to form colloids of amorphous iron hydroxide or hematite.
In estuarine environments, apparently large amounts of iron are removed through an iron
coagulation mechanism of this sort (Hunter, 1983). Due to the small size (0.45 11m) of
colloids, no phase identification has been successfully made (J. Edmond, personal com-
munication). The application of this mechanism to the marine environment has not been
carefully studied. Biologic removal apparently affects the Al and Ca concentration of sur-
face seawater (Deuser et a1.. 1983). The short residence time of iron in seawater suggests
that the biologic removal mechanism of iron might work more efficiently. After these
colloids settle down to the water-sediment interface, the Eh change in their surroundings
might reduce them to magnetite as follows (Zen, 1963):
0...._ _ ._
0.....I | K àú = l r-
Figure 12. 116,000 x .
Mackereth (1971) and Thompson et a1. (1980) both report positive correlations between
the magnetite and organic material contents in lake sediments. Either biologically mediated
or organically catalyzed production of magnetite could be responsible for these observa-
tions.
Type C grains are the most probable candidate for biogenic magnetite. The size of these
grains overlaps with the typical bacterial magnetite size. We noticed that at certain viewing
angles the octahedral grains will have a hexagonal image by TEM (Fig. 12). But if the grains
were pseudohexagonal. the surface might not be straight or the interior of the grains might
show asymmetric light intensity contrast. Portions of the grains show features which sug-
gest an inorganic origin (Fig. 13). Other hexagonal grains, observed under very high mag-
662 Chapter 36
Figure 13. 348.000 x .
nification (>200,000 x), have perfect crystal faces and symmetric interior light intensity
contrasts (Figs. 8, 14-17). These grains do not appear to be pseudohexagonal.
Although the biologic origin of these grains cannot be ascertained just from morpho-
logic observation, there appear to be hexagonal SD magnetite particles in these late Miocene
sediments from northwest Crete. If these grains are really biogenic, they should be regarded
as fossils as are other remains of organisms. However, the term "microfossil" is inappro-
priate for these particles as they are much smaller than any other microfossil reported to
date, their size being less than the wavelength of visible light. The term "nanofossil"
already has been applied to other groups, so we propose using the term "picofossil" for
any biogenic objects in this size range.
o os oI I u.
Figure 14. 542,000 x .
Some of the type C grains are not distinguishable from cubes (Fig. 14), another rare
crystal form of magnetite which is also (apparently) present in some bacteria (Balkwill et
01., 1980). This occurrence of type C grains could be either biogenic or authigenic. Type
A grains probably represent micro scale oxidation or dissolution products of original bio-
genic or authigenic magnetite. After being put in dithionite-citrate solution for 2 days, the
surficial oxidation products (maghemite?) of these grains would be leached away to pro-
duce frothy surfaces.
Even though we cannot determine which source (biogenic or authigenic) contributes
the larger portion of magnetite in the Potamida clays, the significant point is that both
authigenesis and biogenesis depend on the activity of organisms. The observed decrease
664 Chapter 36
o, 0.05 a t
Figure 15. 560,000 x .
of fine-grained magnetite in this section during a reversal might be caused by the decrease
of biologic activity due to environmental changes (small increase of the sun's radiation
reaching the earth's surface, variations of atmospheric circulation, etc.) during a transition
(Black, 1967; Tarling, 1971) or by the decrease in magnetite precipitation by bacteria as
Kirschvink (1982) suggested.
As the results show, the grain size of sample 60 in the Potamida section, which rec-
orded normal polarity, is larger than that of sample 45 in the same section, which was
deposited during a transition. This contradicts the predictions of the biogenic hypothesis,
but as sample 45 is near a ferruginous layer (Fell), it might not record the exact variation
oI
0.05 q
I
Figure 16. 422,000 x .
of grain size distribution during a reversal. The samples from the Skouloudhina section
do not show magnetic intensity variations during the transition (Fig. 1). The grain size of
extracted magnetite from sample in the Skouloudhina section also fall in the MD field.
Such differences in magnetite grain size distribution between the lower Potamida section
and the Skouloudhina section might correspond to changes in the depositional environ-
ments.
The Skouloudhina section is located in the same basin as Potamida (Laj et 01., 1982),
but is laterally closer to the neritic Fotokadhon Formation (Meulenkamp et 01., 1979; Lan-
gereis, personal communication). Another source of magnetite, such as terrigenous detritus ,
666 Chapter 36
°'--____°.....-_____5. 2 M-
Figure 17. 268,000 x.
might be more significant for contributing magnetite in this relatively shallower environ-
ment. In addition, the depositional age of both sections is close to the Mediterranean
salinity crisis in the Messinian. The increase in magnetite grain size in the upper Potamida
section probably also reflects a transition from open-marine environments to an early Mes-
sinian overall sea level submergence, resulting in semiclosed shallow embayments (Meu-
lenkamp et aI., 1979; Valet et al., 1983; Langereis et a!., 1984) and terrigenous detritus
might become the dominant phase. The fact that no euhedral magnetite grains have been
found in samples 60 and 02A supports this deduction. The real relation between envi-
ronmental changes and such magnetite grain size variations needs to be determined by
detailed local depositional phase analysis.
7. Conclusion and Applications

The results shown above do not exclude the possibility of a nonbiogenic origin for
the magnetite in these sediments. But the evidence suggests a biologic origin for certain
grains. Whether the bacterial magnetite is the dominant magnetite phase as predicted by
Kirschvink (1982) cannot as yet be determined from this study. We plan to investigate
magnetite extracts from a variety of sediments and sedimentary rocks to see the size and
shape distribution of detrital magnetite in different environments and different geologic
periods. In particular, comparison of the magnetite extracts from sediments with abundant
fossils and organic material (like marl) and sediments which are apparently of inorganic
origin (like loess) might provide a solution to this problem.
We can also use this method to evaluate the existence of SD magnetite as a primary
component in varieties of rocks. This is extremely important for paleomagnetic studies on
shale and other sedimentary rocks. Because the major mineral phases in shales are usually
diagenetic clays, the remanent magnetism of shale might be secondary in origin. Apparently
this is not the case for the Potamida clays. It is highly possible that other shales also have
primary remanent magnetism obtained by biologic precipitation of iron oxide. More than
half of all sedimentary rocks are shale, and an understanding of the processes through
which they become magnetized is important to the science of paleomagnetism.
ACKNOWLEDGMENTS. Drs. J. D. A. Zijderveld, C. G. Langereis, and C. Laj generously contrib-

uted samples to this study. Dr. J. P. Revel made available the transmission electron mi-
croscope in his laboratory. P. F. Koen provided technical assistance on TEM handling. H.
A. Lowenstam, S. M. Awramik, M. A. S. McMenamin, K. M. Towe, and R. L. Edwards
gave helpful comments and suggestions during preparation of this paper. This study was
partially supported by National Science Foundation Grant EAR 81-21377. This is Contri-
bution No. 3913 from the Division of Geological and Planetary Sciences, California Institute
of Technology.
References
Balkwill, D. L., Maratea, D., and Blakemore, R P., 1980, Ultrastructure of a magnetotactic spirillum,
J. Bacteriol. 141:1399-1408.
Berner, R A., 1964, Stability fields of the iron mineral in anaerobic marine sediments, J. Geol. 72:826-
834.
Blake, D. J., 1967, Cosmic ray effects and foraminifera extinctions at geomagnetic field reversals, Earth
Blakemore, R P., Frankel, R B., and Kalmijn, Ad. J., 1980, South-seeking magnetotactic bacteria in
the Southern Hemisphere, Nature 286:384-385.
Bogue, S. W., and Coe, R S., 1982, Successive paleomagnetic reversal records from Kawai, Nature
295:399-401.
Butler, R R, and Banerjee, S. K., 1975, Theoretical single-domain grain size range in magnetite and
Carpenter, R, Johnson, H. P., Johnson, and Twiss, E. S., 1972, Thermomagnetic behavior of manganese
nodules, J. Geophys. Res. 77:7163-7177.
Castaing, R, and Frederickson, K., 1958, Analyses of cosmic spherules with an X-ray microanalyse,
Geochim. Cosmochim. Acta 14:114-117.
Chester, R., and Aston, S. R, 1976, The geochemistry of deep sea sediments, in: Chemical Ocean-
ography, Volume 6 (J. P. Riley and R Chester, eds.), Academic Press, New York, pp. 281-390.
samples, Phys. Earth Planet Inter. 26:56-62.
Deuser, W. G., Brewer, P. G., Jichells, T. D., and Commean, R F., 1983, Biological control of the removal
of abiogenic particles from the surface sea, Science 219:388-391.
Doyle, L. J., Thomas, L. H., and Betzer, P. R, 1976, Black magnetic spherule fallout in the eastern Gulf
of Mexico, Science 197:1157-1159.
Drooger, C. W., Meulenkamp, J. E., Langereis, C. G., Wonders, A. A. H., Van der Zwaan, G. J., Drooger,
M. M., Raju, D. S. N., Doeven, P. H., Zachariasse, W. J., Schmidt, R R, and Zijderveld, J. D. A.,
1979, Problems of detailed biostratigraphic and magnetostratigraphic correlation in the Potamida
and Apostoli sections of the Creten Neogene, Utrecht Micropaleontol. Bull. 21.
Ehrlich, H. L., 1982, Geomicrobiology. Dekker. New York.
668 Chapter 36
Ensley, R. A., and Verosub, K. 1., 1982, A magnetostratigraphic study of the sediments of the Ridge
Basin, Southern California and its tectonic and sedimentologic implications, Earth Planet. Sci.
Lett. 59:192-207.
Evans, M. E., and Wayman, M. 1., 1970, An investigation of small magnetic particles by means of
electron microscopy, Earth Planet. Sci. Lett. 9:365-370.
Evans, M. E., McElhinny, M. W., and Gifford, A. C., 1968, Single domain magnetite and high coer-
civities in a gabbroic intrusion, Earth Planet. Sci. Lett. 4:142-146.
Frankel, R. B., Blakemore, R. P., and Wolfe, R. S., 1979, Magnetite in fresh-water magnetotactic bacteria,
Science 203:1355-1356.
Fyfe, W. S., and Lonsdale, P., 1981, Ocean floor hydrothermal activity, in: The Sea, Volume 7 (c.
Emiliani, ed.), John Wiley & Sons, New York, pp. 589-638.
Garrels, R. M., and Christ, C. 1., 1965, Solutions, Minerals and Equilibria, Harper & Row, New York.
Geissman, J. W., Newberry, N. G., and Peacor, D. R., 1983, Discrete single-domain and pseudo-single-
domain titanomagnetite particles in silicic glass of an ash-flow tuff, Can. J. Earth Sci. 20:334-
338.
Gould, J. 1., Kirschvink, J. 1., and Deffeyes, K. S., 1978, Bees have magnetic remanence, Science
201:1026-1028.
Hackett, J. P., and Bischoff, J. 1., 1973, New data on the stratigraphy, extent and geologic history of
the Red Sea geothermal deposits, Econ. Geol. 68:563-584.
Harrison, C. G. A., and Peterson, M. N. A., 1965, A magnetic mineral from the Indian Ocean, Am.
Mineral. 50:704-713.
Henshaw, P. c., and Merrill, R. T., 1980, Magnetic and chemical changes in marine sediments, Rev.
Hunter, K. A., 1983, On the estuarine mixing of dissolved substances in relation to colloid stability
and surface properties, Geochim. Cosmochim. Acta 47:467-474.
Johnson, H. P., Lowrie, P., and Kent, D. V., 1975, Stability of anhysteretic remanent magnetization in
fine and coarse magnetite and maghemite particles, Geophys. J. R. Astron. Soc. 41:1-10.
Kirschvink, J. L., 1981a, Biogenic magnetite (Fe 3 04j: A ferrimagnetic mineral in bacteria, animals, and
man, in: Ferrites: Proceedings of the International Conference Japan, 1980, pp. 135-137.
Kirschvink, J. 1., 1981b, A rapid, non-acidic chemical demagnetization technique for dissolving ferric
minerals, Eos 62:848.
Kirschvink, J. L., and Gould, J. 1., 1981, Biogenic magnetite as a basis for magnetic field sensitivity
Paleomagnetic, sedimentologic and biologic implications of organic magnetite, Earth Planet. Sci.
Lett. 44:193-204.
Laj, c., Jamet, M., Sorel, D., and Valente, J. P., 1982, First paleomagnetic results from Mio-Pliocene
series of the Hellenic sedimentary arc, Tectonophysics 86:45-67.
Langereis, C. G., 1979, An attempt to correlate two adjacent Tortonian marine clay in western Crete
using magnetostratigraphic methods, Utrecht Micropaleontol. Bull. 21:193-214.
Langereis, C. G., and Zachariasse, W. J., and Zijderveld, J. D. A., 1984, Late Miocene magnetobiostra-
tigraphy of Crete, Mar. Micropaleontol. 8:261-281.
Acad. Bras. Cienc. Notas Fis. CBPF-NP-048/81.
Lovlie, R. W., Lowrie, W., and Jacobs, M., 1971, Magnetic properties and mineralogy of four deep-sea
cores, Earth Planet. Sci. Lett. 15:157-162.
Lowenstam, H. A., 1962, Magnetite in denticle capping in recent chitons (Polyplacophoraj, Geol. Soc.
Am. Bull. 73:435-438.
neralization, in: Biomineralization and Biological Metal Accumulation (P. Westbroek and E. D.
de Jong, eds.j, Reidel, Dordrecht, pp. 191-203.
Lowrie, W., and Heller, F., 1982, Magnetic properties of marine limestones, Rev. Geophys. Space Phys.
20:171-192.
McCabe, C., Van der Voo, R., Peacor, D. R., Scotese, C. R., and Freeman, R., 1983, Diagenetic magnetite
carries ancient yet secondary remanence in some Paleozoic sedimentary carbonates, Geology
11:221-223.
Mackereth, F. H. J., 1971, On the variation in direction of the horizontal component of remanent
magnetism in lake sediments, Earth Planet. Sci. Lett. 24:414-418.
Mehra, O. P., and Jackson, M. 1., 1960, Iron oxide removal from soils and clays by a dithionite-citrate
system buffered with sodium bicarbonate, Clays Clay Miner. 7:317-327.
Meulenkamp, J. E., Jonkers, A., and Spaak, P., 1979, Late Miocene to early Pliocene development of
Crete, in: VI Colloquium on the Geology of the Aegean Region pp. 138-149.
Dzdemir, D., and Banerjee, S. K., 1982, A preliminary magnetic study of soil samples from west-central
Minnesota, Earth Planet. Sci. Lett. 59:393-403.
Schwertmann, U., and Taylor, R M., 1977, Iron oxides, in: Minerals in Soil Environments 0. B. Dixon
and S. B. Weed, eds.), American Society of Agronomy, Madison, Wise., pp. 145-180.
Shanks, W. c., III, Bischoff, J. L., and Rosenbauer, R J., 1981, Seawater sulfate reduction and sulfur
isotope fractionation in basaltic systems: Interaction of seawater with fayalite and magnetite at
200-350°C, Geochim. Cosmochim. Acta 45:1977-1995.
Smith, P. P. K., 1979, the identification of single-domain titanomagnetite particles by means of trans-
mission electron microscopy, Can. J. Earth Sci. 16:375-379.
Stober, J. c., and Thompson, R, 1979, Magnetic remanence acquisition in Finnish lake sediments,
Geophys. J. R. Astron. Soc. 57:727-739.
Sugimoto, T., and Matijevic, E., 1980, Formation of uniform spherical magnetite particles by crystal-
lization from ferrous hydroxide gels, J. Colloid Interface Sci. 74:227-243.
Tariing, D. H., 1971, Principles and Applications of Paleomagnetism, Chapman & Hall, London.
Thompson, R, Bloemendal, J., Dearing, J. A., Oldfield, F., Rummery, T. A., Stober, J. c., and Turner,
G. M., 1980, Environmental applications of magnetic measurements, Science 207:481-486.
Valet, J. P., and Laj, c., 1981, Paleomagnetic record of two successive Miocene geomagnetic reversal
in western Crete, Earth Planet. Sci. lett. 54:53-63.
Valet, J. P., Laj, c., and Langereis, C. G., 1983, A study of the two different R-N geomagnetic reversals
recorded at the same site, Nature 304:330-332.
Verosub, K. 1., 1977, Depositional and postdepositional processes in the magnetization of sediments,
Rev. Geophys. Space Phys. 15:129-142.
Walcott, c., Gould, J. 1., and Kirschvink, J. 1., 1979, Pigeons have magnets, Science 205:1027-1029.
Walker, M. M., and Dizon, A. E., 1981, Identification of magnetite in Tuna, Eos 62:850.
Zachariasse, W. J., 1979, Planktonic foraminifera from section Potamida. 1. Taxonomic and phyletic
aspects of keeled globorotaliids and some paleoenvironmental estimates, Utrecht Micropaleontol.
Bull. 21:129-166.
Zen, E., 1963, Components, phases, and criteria of chemical equilibrium in rocks, Am. J. Sci. 261:919-
942.
Index
Acanthopleura granulate, 15 Bacterium(-a) (continued)

Adrenal glands, 550, 552 magnetic crystals from, 252
"Aftereffect", 551, 555, 557-558 population density, 647
humans, 551, 555, 557-558 rod-shaped magnetotactic, 294
pigeons, 557 sulfate-reducing, 10
Aggregate, 297, 303, 306 Bacterioferritin, 3, 4, 15, 277
Algae Balanus, see Barnacles
compass of, 244 Barnacles
green, 293, 297 growth of in magnetic field, 366, 369, 370
magnetotactic, 297 magnetic properties, 375-377
magnetotactic, 14, 243, 252, 254, 433, 668 Bardeen, Cooper, and Schrieffer (BCS) theory,
Alternating field (AF) demagnetization, 145, 105-108
146,158,159,401,427-429,649,650, Bats, 483-487
653 Bed orientation, 544, 555-558, 616
Aluminum, 177 hypothesis, 557, 558
Amphibians, 243, 439, 440-442, 452 Bees, 292
Ampullae of Lorenzini, 223, 225, 227 Behavioral investigations, 510-521, 530, 531
Amyl acetate, 170 Biochemical investigations, 530, 531
Animalia, 4 Biogenic ferromagnetism, see Ferromagnetism
Annelida, 5, 14 Biogenic minerals, see Minerals, biogenic
Antiferromagnetic, 139-141 Biogenic origin, 626
Apatite, 14 Biological clock, 44, 87, 90-92, 457, 458, 462
Apis mellifica, see Honeybee Biological energy, 294
Aquaspirillum magnetotacticum, 13, 280, 293, Biological rhythms, 290
294, 305, 306; see also Magnetite, Biomineralization 1, 163, 164
bacterial induced, 4-6, 11
Archaebacteria, 15 matrix-mediated, 4-7,11,12,433,434
Arthropods, 4 pinpoint, 7, 12
Asclepias, 408, 411 process(es), 419, 433, 434
Ash flow-tuff, 656 process of magnetite, 292
Atmospheric circulation, 664 products, 277, 280, 285
Azotobacter, 15 skeletal, 11
Bacterium(-a), 290, 292-295, 297, 298, 304-306, Bioremanent magnetization, 65
495 Biosphere redox potential, 13
coccus, 294, 295, 297, 298, 305, 306 Biostratigraphy, 649
compass, of 244 Birds, 225, 245, 250, 493-494, 497, 498,
cyano-, 7, 13 503-504
fossil, 13 migration, 257-265
magnetic, 625, 630 migratory, 243
magnetotactic, 1, 4, 7, 10-14, 162, 243, 244, Blackcap, 466
252,279,294,295,297,419,433,434, Bleach, 169
626, 647, 648, 668 Blepharoplast, 297
coccus, 292, 294, 295 Blind subjects, 601, 602, 612
spirillum, 667 congenitally, 601
671
672 Index
Blindfolds, 540, 576, 616-618 Chiton (continued)

Blocking temperature, 35, 402 tooth mineralization, ultrastructure of,
Blocking volume, 35 342-361
Blood Chlamydomonas, 293, 297
chiton, 10 Chloroplast, 297
vertebrate, 10, 250 Chordates, 4
Blue marlin, 157, 425, 429 Chromium, 172,430
Boltzmann distribution, 244, 248, 252 Chub mackeral, 425, 429
Boltzmann's constant, 244, 400 Circadian rhythms, 87, 91, 395-398
Bombus, 188 Clay, 178, 650, 667
Bone repair, 539 "Clever Hans" effect, 617-619
Bonneterre dolomite, 659 Clothing, effects on orientation by, 544,
Brackish waters, 294 554-556, 616, 618
Bragg angle, 175 Coercive force, 26, 28, 142, 143
Bragg's law, 176 Coercivity, 20, 25, 28, 423, 427, 429, 496-513
Brain of magnetite, 158,419, 431, 433
cetacean, 495-502 spectrum, 158, 523, 524
waves, 538, 555 studies, 158, 160, 162
Brillouin function, 21 Cognition, information processing, 294
Brownian motion, 244, 248, 294 Coils, magnetic
Bumblebee, 188 alternating field (AF), 207, 213, 214
Brushite, 7 electromagnetic (sweeper), 204, 205, 207-209,
Bus experiments, 545-550, 556, 574-584, 589, 213-215, 218, 219
591,592,611-621 Helmholtz, 201
alternative hypotheses, 577-584, 592 Colloids, 176
American, 545-550, 574-584, 611-621 Columba livia, see Rock dove
British, 546-550 Compass, 231, 289, 290
post hoc analyses, 575, 584, 591, 615, 616, axial magnetic, 246
618 bacterial, 244
Butler-Banerjee diagram, see Magnetite, single- ferromagnetic, 247, 254
domain geomagnetic, 509, 513-521
magnetic, 417, 431, 459, 462, 465-469, 471,
Calcification (of plant), 13 472, 571
Calcite, 7, 9, 650 organelles, 243, 249, 250
Cambrian, 11,659 orientation
Carbon-evaporated film, 170 by birds, 548, 618
Carbonate, 647 by humans, 538, 540-545, 554, 556
Cardinal points, 392 by rodents, 553
Caretta caretta, 444-446 by salamanders, 548
Cave Salamander, 254, 440-441 and time of day, 554
Celestite, 6 sense, 258, 259
Centrifuge, 169 star, 247, 254, 457, 465
Cerebrospinal fluid, 250 sun, 246,457,462, 514, 518, 519
Cetacean, 157, 160 "Compass-needle" detector, 400
brain, 495-502 Conditioning experiments, 417, 418, 431, 432
geomagnetic sensitivity, of 253 Contamination
Chair experiments, 540-545, 554, 556, 615, 618 adventitious grains, 168
American, 614, 615, 618, 621 chemical, 176
British, 540-545, 554-556, 619 electron beam, 175-176
Chelonia mydas, 443, 444, 447-451 ferromagnetic, 153
Chemotaxy, 305 magnetic, 156, 158, 159, 164, 419, 425, 429,
Chiton 3, 6, 156, 160, 433, 668 495, 497-500, 502-504
blood chemistry, 341, 342 sources of, 370-371
homing by, 10, 14 Continent
magnetite in, 252, 292 magnetic anomalies, 247
radula teeth of, 169, 333-363 U.S. Atlantic margin, 247
teeth, 1, 668 Cooper pairs, 106, 107
Index 673
Cosmic dust, 654 Domain states, 23

Cosmic spherules, 667 Domain theory, 23
Course of Force, 485 Double-blind protocol, 535, 540, 542, 544, 551,
Crete, 14 574, 576, 618, 619, 621
Fotokadhon Formation of, 665 Dowsing, 538-540, 550, 556
Potamida Clays of, 647-655, 665 Dura mater, 160, 163, 432, 495-504
Skonlandhina section of, 649, 653, 654, 665 green turtle, 447-451
Crickets, 408 Dynamic magnetic shield, 201
Crocodile, Saltwater, 443 Dyslexics, 543, 619
Crustacea
growth in magnetic field, 366, 367, 380
magnetic properties, 371-379 Earmuffs, 540, 544, 621
significance of ferrimagnetism, 378-380 Echinodermata, 5
Cryocoolers, 128 EDAX probe, 448, 450
Cryogenics, 121-128 Ediacara fossil reserve, 14
Cryptochiton, 334-336, 358, 360 Eel Marsh, 627
Crystal chains, 295, 298 Eels, orientation of, 243
Crystallographic transition, 19, 20 Eh-pH stability diagram, 659
Crystals, 291, 292, 295 Elasmobranch fish, 225, 227, 253, 399
Culture, 297 Elastic-rod superparamagnetic transducer, 400,
Curie temperature, 20-22, 61, 139-141 401
of magnetite, 159 Electric fields, 555
Cytochromes, 10, 11 perception of, 227
Cytoplasm, 292 shielding of, 199-201
high frequency fields, 201
skin depth, 201
D spacing, 177-179 static fields, 200, 201
Danaus gilippus, 412, 413 Electrical properties of materials
Danaus philene, 412, 413 conductivity, 199-201, 209, 211
Danaus plexippus, 407-415; see also Monarch resistivity, 199-201
butterfly Electromagnetic field, 45-47
Deep-sea cores, 659, 668 and dogs, 539
Delphinus delphis, 490, 495-497, 501-502 Electromagnetic shielding, 200, 209
Depositional effect, 655 Electron, 175
Depositional environment, relation to grain size, Electron diffraction, 167, 176-179, 187, 652
665 calibration, 177
Dermal granules, 278 identification of particulates by, 176-179
Dermethmoid bone, 4, 160, 424-427, 430 pattern, 177, 653
Dermethmoid tissue, 425-429, 431-433 Electron micrograph transmission, 655; see also
Devonian, 659 Electron microscope
Diamagnetism, 138, 139 Electron microprobe, 430, 433
Dipole, 291-305 analysis, 448, 449
Direct current (DC) magnetization, 144-146 Electron microscope, microscopy, 167-179, 186,
Discrimination learning, 492, 493 190, 291, 293, 299, 301, 430, 433, 448,
in salamanders, 440, 441 450, 451, 554, 648
in sea turtles, 444-447 analytical, 191
Discrimination testing, training, 419, 421 higher resolution SEM, 172
Diseases, human, 10 scanning (SEM), 293, 297, 298, 299, 648, 659
Displacement, 601, 602 transmission (rEM), 293, 295, 298, 300, 304,
Displacement experiments, 513-521, 545, 569 528, 529, 648, 653-656, 667, 668
Dithionite-citric acid dissolution, 650, 652 artifacts, 169, 172, 175
Dithionite-citrate solution, 663, 669 contamination, 175-176
Diurnal variations 1; see also Circadian rhythms and crystalline materials, 175
Divining rod, 539 diffraction mode, 174, 176-179
Dolphins 292; see also Cetaceans fixation, 173
bottlenose, 490, 492-495, 499-502 freeze-fracture, 173
Pacific, 163 photography, 175
Domain walls, 24 preparation for use, 174
674 Index
Electron microscope, microscopy (continued) Fish(es). 12, 251, 492-494, 497, 498, 502, 504
transmission (TEM), (continued) elasmobranch, 225, 227, 253, 399, 431, 432
replica methods, 170 migration, 247
sample preparation, 168-173 pelagic, 247, 419, 425, 429, 432
shadowcasting, 172 teleost, 431, 432
"soft" image, 174 Flagellum, 297
specimen drift, 170 Fluorite, 6, 14
specimen support, 170, 171 Flux quantization, 110
and hydrophobicity, 171, 177 Flux quantum, 106, 107, 108, 110
specimen thickness, 170, 175 Foraminifera, 5
"through focus" series, 175 Fossils, see Magnetite, fossils
ultrathin sections, 170, 175, 176 Fouling organisms, 369
variation in contrast, 175; Franz isodynamic magnetic separator, 650
see also Electron diffraction Frequency range, 226
Electroreception, 474 Fresh water, 294, 298, 304
Electroreceptor, 229 Fungi,4
Electrostatic shielding, 199, 200
Energy Gabbroic intrusion, 656
magnetic, 244 Gastropod, 5
quantum of, 224 Geitleria, 7, 14
thermal, 244 Geomagnetic field, 43-102, 156, 244, 251, 289,
Engraulis mordax, 425 290,291,295,299,301,305,306,307,
Eolian transport, 659 422,432,647
Eoscaphander, 7, 8 age of, 44, 64
Erithacus rubecu1a, 464, 465, 466 amplitude of, 290
Errors in analysis, 546-548, 620 animal orientation and migration, influence
Escherichia coli, 15 on,43,95
Estuarine environment, 660 anomalies, 1, 417
Ethmoid sinus, 552 apparent polar wander path, 69
Ethmoid/sphenoid bone complex, 525-529 atmospheric dynamo, 89
Ethmoturbinal bones, 553-554 biologic clock, influence on, 44, 87, 90-92
Eukaryotes, 3, 6, 7, 11-13, 293 biologic relevance, 43-45
Eurycea 1ucifuga, 254, 440, 441 configuration
Exchange energy, 25 past, 44, 64-71
Experimental protocol, magnetoreception, 535, present, 47-61
616-619 Cretaceous Long Normal Zone, 78
daily variation, 89-92, 395-398
declination of, 48, 50, 247, 290, 421, 432
Falx cerebri, 495-498, 500-502 dipolar configuration, 53-61
Faraday induction, 201, 224 dipole hypothesis, 49, 54, 66, 68
Fayalite, 669 dipole moment, 60
Ferric iron deposits, biologic, 525-529 direction, 397, 417-419
Ferrihydrite, 3, 4, 10, 276, 647 excursions, 76
deposition, 333, 350, 358, 359, 360 features, 417
Ferrimagnet, 140-142 gauss coefficients, 55-61
Ferrimagnetism, 17, 19, 22, 140-142, 145 geodynamo, 62, 63
Ferritin, 3,4, 10, 13-15, 277 geomagnetic axis, 60
chiton, 334, 342, 344, 345-347, 349-353, 361 geomagnetic equator, 58, 60
endocytosis, 344-346, 348 geomagnetic indices, 95-98
Ferromagnetic materials, 198, 199, 209, 211, geomagnetic latitudes, 60
219 geomagnetic poles, 60, 67
metals, 198, 200, 201, 213 gradient, 421, 432
properties of, 199-201, 210, 218 Hawaiian, 421
Ferromagnetism, biogenic, 1 inclination, 48, 290, 421, 432
Ficedu1a hypoleuca, 466, 468, 473 intensity of, 49, 248, 417, 419
Field-compensation experiments, 387, 391 International Geomagnetic Reference Field, 49
Fin whales, 247 isodynamics, 50
Index 675
Geomagnetic field (continued) Holothurians, 5

isogonies, 50 Homeward component of direction, 543, 548-
lines, 46-47, 388 549, 601, 602, 618-619
magnetic anomalies, 52, 53, 57, 58 Homeward orientation, 602
magnetic elements, 48 Homing, 231, 601
magnetic equator, 49, 60 Homing experiments, see Displacement
magnetic latitude, 49 experiments
magnetic pole, 49, 60 Honeybees, 155, 159, 160, 245, 250, 251, 254,
magnetic storms, 92-95 419,428,433, 502, 554, 558
magnitude, 291 abdomen, 4
magnetosphere, 87-89 circadian rhythm, 248, 252
main field, 56 comb-building, 393
origin of, 61-64 compass of, 244
paleointensity, 44, 71-74 conditioned responses to magnetic fields, 404
paleomagnetic pole, 68 demagnetized, 401
paleomagnetic record, 64, 68-74 horizontal dances of, 392-393, 398
polarity 403, 420 magnetic-field receptor of, 398-400
polarity intervals, 74 magnetic orientation of, 385, 393-397, 399
polarity time scale, 74 magnetic remanence of, 401, 402
polarity transitions, 80-82, 648, 653, 664 magnetite in, 403
reversals, 74-83,433,467,472,654,664,667, misdirection in dance, 386-392
669 navigation, 258-259
cause of, 82-83 and null magnetic fields, 387, 391
and evolution, 83-87 permanent magnetism of, 399
and foraminifera extinction, 667 residual misdirection in dances of, 386-392,
Kawai record, 667 398
secular variation, 51, 58 field compensation experiments on, 387,
transition, see Polarity transition 391
virtual dipole moment, 66 zero-crossing of, 389, 391
virtual geomagnetic pole (VGP), 66 sounds of, 391-392
Geomagnetism, 1, 290 and strong magnetic fields, 388, 393, 394,
Geomicrobiology, 13 396, 397
Gizzard Plates, 5 and superparamagnetism, 401-403
Glow Discharge, 171 swarm, 393
Goal orientation 539, 545-550; see also time sense of, 395-398
Navigation waggle dance of, 243, 252, 386-392
Goethite, 4, 14, 158, 274, 650 Honeycomb, 243
Gold, 172, 177 Human, 159, 601, 602
Gradiometer, 130-133, 136, 137, 140 orientation
Grass shrimp, 376-377 blindfolded, 576
Gravitational orientation, 230 effect of magnets, 576-577, 579
Gravity, 246, 247 homing, 573-593
Greigite, 633, 638, 639, 640, 641, 642 tests, procedures, 574;
Gull, ring-billed, 464 see also "Bus" experiments; "Chair"
Gypsum, 6 experiments
Humpback whale, 490, 498-499, 501-502
Hydrotroilite, 5
Haplophragmoides canariensis, 15 Hydrous ferric oxides, 276, 283
Helderberg limestones, 659 Hyperfine interactions, 271
Helium, liquid, storage and transfer, 126-128
Hellenic sedimentary are, 668
Helmholtz coil, 293, 486 Igneous rocks, xi
Hematite, 21, 153, 158, 274, 429, 650 Illite, 650
Hemocyanin, 10 Ilmenite, 5
Hemoglobin, 10, 484-486 Inclination, 291
Higher-order analysis, 564 Indigo bunting, 466
Histological investigations, 524-530 Induction, 399
Histology, 184, 188, 552-554 Induction-based magnetoreception, 223, 225
676 Index
Inertial cues, 571 Living beings, magnetism in, 291

Inertial navigation, 231 Loess, 667
Inertial sense, 543, 617, 619 London equations, 109
Inner ear, 230 Lowrie-Fuller test, 656
Innervation, 496, 503 Lyeorea c1eobea, 412-413
Insects, 484-486
time cue, 243 Maghemite, 17, 21, 153, 162, 275, 429, 650, 663
Interaction energy, 399 Magnet, 293
Internal standard, 177 cobalt-samarium, 157, 161,424
Invaginations, 297, 303 in pigeon, 669
Ion-cleaning, 171 Magnetic anisotropy, 142
Iron, 297 transitions, 140
coagulation, 660 Magnetic anomaly, 52, 53, 57, 58, 261-264, 291,
hydrous ferric oxides, 169, 173, 276, 283 469-474
localization, 184, 188 continental, 247
native, 162, 163 and pigeons, 248
oxides, 153, 161 South Atlantic, 291
oxides and hydroxides, Mossbauer Magnetic biopsy, 13
spectroscopy of, 274 Magnetic characterization, 142-148
particulate, 183 Magnetic coils, see Coils, magnetic
proteins, 11 Magnetic contamination, 156, 158, 159, 164,
release by oxidation, 169 419, 425, 429, 495, 497-500, 502-504
storage, 164, 434 sources of, 370-371
sulfide, 637, 639 Magnetic crystals, 305
Iron-mediating enzymes, 648 Magnetic cues, 592
Isothermal remanent magnetism (IRM), 153, Magnetic declination, 290
522-525, 528 Magnetic deposits, 552-554
acquisition, 499 humans, 552-554
Isotropic point, 20, 37 monkeys, 552
woodmice,553-554
Jarosite, 5 Magnetic detection, 490-495, 503-504
Josephson junction, 106, 107, 110, 111 Magnetic dipole, 290, 305, 306
Josephson relations, 111 remanent dipole moment, 202, 203
Magnetic disturbances, 398
Kaolinite, 650 Magnetic domain, 24, 206, 207; see also
Katsuwonus pe1amis, 423 Magnetite, multidomain, single-domain
alignment, 207, 208, 214, 215, 218
Laboratory high coercivity, 204, 213, 215, 218
clean, 157 non-remanent domains, 207
biomagnetic, 218, 219 permanent, 202, 207, 215
electron microscopy, 157 thermal agitation of, 206, 207
paleomagnetic, 156 vibration alignment of, 208
Langevin function, 26, 44, 206, 207, 252, 305 Magnetic domain theory, 17, 24; see also
as constraint on magnetoreception, 247 Magnetite, multi domain, single-domain
domain alignment, 207 Magnetic field, 176, 290, 293-295, 305, 306,
effective permeability, 208 602; see also Geomagnetic field
thermal agitation, 207 altered, 440, 441, 444, 445
Langevin theory, 234, 236, 237 compensation experiments, 387, 391
Larus de1awarensis, 464 Crustacea growth in, 366-367, 380
Learning (conditioning), 420, 421, 490, and induced electric field, 44, 45-47
492-495, 504 intensity, 393, 397, 417, 418, 421, 432
constraints and specializations, 494 measurement, 103, 108, 129, 131, 132, 133
Lepidochitona, 334, 338, 340, 358, 360 accuracy of, 227
Lepidocrocite, 4, 14, 15, 275 pathological effects of, 509-510, 527, 530
Lepidoptera, 407-415 physiological response to
Leptotrix, 7 strong fields, 538
Limestone, 656, 668 weak fields, 538-539
Limpets, 4 receptor (honeybee), 398-400
Index 677
Magnetic field (continued) Magnetite (continued)

static, 198, 201, 219, 396 authigenic, 659
stimuli, 418, 419 bacterial, 252, 313, 315, 316, 320, 325-329,
strong, 396 627, 630, 654, 661, 669
avoidance of, 510-513 chemical control of, 327
transducer, 399,401 crystal alignment, 315-319
units of measurement, 49 crystal growth, 326
varying, 198, 201 high resolution transmission electron
vector, 388 microscopy (TEM) studies, 311-332
Magnetic gradient, 441, 444 magnetotactic coccoid cells, 314-316, 322,
Magnetic granulometry, 36, 147 326, 329
Magnetic history of materials, 215, 219 microstructure, 315, 319
Magnetic hysteresis, 142-144 morphology, 315-319, 326
Magnetic inclination, 290 nucleation, 325
Magnetic interactions, 145, 297, 305, 306, 307 spatial control, 326
Magnetic map, 417, 432, 459, 471, 472 structural control, 329
Magnetic materials, 225 biogenic, 155, 156, 158, 159, 163, 164, 251,
Magnetic moment, 157, 158, 292-294, 299, 301, 252,433,434,641,647,648,658,659,
305-307,423,433 661
induced, 204 biomineralization, 419, 431, 433
remanent, 203 biosynthesis, 11, 323-329
Magnetic orientation, 307 botryoidal, 659
bacteria, 295, 297 in chitons, 252
human, 563-568 teeth, 169, 360, 361, 647;
Magnetic phosphene effect, 538-539 see also Chiton
Magnetic pulsations, 1 coercivity of, 158,427,430
Magnetic remanence, 153, 160, 162,429, 647 as a conductor, 230
of bees, 668 cosmic, 658, 659
of green turtles, 447, 448 in Crustacea, 378
of honeybees, 402 crystals, 293
Magnetic sense, sensitivity, 418, 432, 540 chains, 252
Magnetic separations, 628 morphology, 431, 433
Magnetic shielding, 197-220 shape, 400
dynamic, 201 size, 400
electromagnetic, 200, 201 Curie temperature, 159
magnetostatic, 198, 201-210, 219 density, 650
Magnetic spherules, 658 diagenetic, 658, 659
Magnetic stability, in Crustacea, 378 euhedral, 666
Magnetic storms, 260-261, 417, 469, 470, 471, extraction, 160, 658
472,539 ferrimagnetic, 21
and accidents, 539 fossils, 658, 659
and heart attacks, 539 bacterial, 648
orientation and navigation, 548-549, 558, biogenic, 647
616, 621 geologic, 156, 161, 163, 164, 430, 433
Magnetic susceptibility (low field), 147 hexagonal, 654
Magnetic tissues, 163 hydrogenous, 658, 659
Magnetically shielded room, 197-220 hydrothermal, 658, 659
construction of, 208-219 in hydrothermal system, 659
design and fabrication of, 198, 199, 205, 208- inorganic, 654
219 interparticle interactions, 252
examples, 210-212, 214-219 isotropic point, 160
site survey for, 208, 209, 215 and magnetic deposits, 522-528
Magnetism, 289, 290, 291 magnetization, 650
Magnetite, 1, 3,4,6,12, 17, 178, 275, 292, 293, in manganese nodule, 659
295,297,299,475-477,485,495-496, morphology, 654
504,552-554,648,650,653,669 multidomain, 156, 158, 160, 251, 429,
alignment energy, 234, 235 496-497, 501-502, 653, 654, 659
678 Index
Magnetite (continued) Magnetoreception (continued)

and nervous system, 249 by electrical induction, 418, 433
precipitation in magnetotactic bacteria, 285 ferromagnetic, 432, 433
pseudosingle domain (PSD), 648, 658 hypothesis, 245, 252
reduced iron in, 169 influence of magnets, 544-545, 548-551, 555,
in sea turtles, 447-451 557-558, 613-614, 619-620
single-domain (SD), 156, 158, 162, 163, 247, intensity, 251
251,292,297,400,422,427,430,432, Langevin function constraint, 244, 247
499, 502, 653-656, 662, 668 by liquid crystal effects, 418
resistance to demagnetization, 655 magnetite-based, 419, 422, 426, 433
spheroid, 659, 669 in newts, 442
superparamagnetic, 654 ontogeny of, 519-521, 525
synthetic, 156, 161, 164, 429, 658 by optical pumping, 418, 433
teardrop, 654 paramagnetic hypothesis, 250
terrigenous, 658, 659 in salamanders, 440, 441, 442
tissue extraction, 252 in sea turtles, 443-446
tooth-capping mineral in chitons, 333, 335 Magnetoreceptor 153, 522-530
in tuna, 252, 669 alternating current (AC) response, 240, 241
in turtle, 654 cells, 418, 419
volcanic, 153, 658, 659 magnetite based, 419, 423, 432, 433
x-ray diffraction pattern of, 651 definition, 537
Magnetization location, 550-552
change by wall movement, 144 adrenal glands, 552
direct current (DC), 144-146 front of head, 552
homogenous, 157 sinuses, 552-554, 556-557
inhomogenous, 157 organelles, 422, 423, 426, 432
intensity of, 423, 424 physiology, 554-558
rotation, 144 response time, 235, 236, 240
spontaneous, 21; sensitivity to small field changes, 238, 239
see also Remanent magnetization, experimental, 238
Saturation magnetization structure, 552-554, 557-558
Magnetocrystalline anisotropy, 22, 23 system, 419
constant, 20 Magnetosome, 4, 244, 280, 292, 297
energy, 25 Magnetostatic energy, 25
Magnetofossils, 14 Magnetostatic shield, 198, 201, 202, 209, 210,
Magnetometer, 103, 129-136, 156-158 211
cryogenic, 402 Magnetostratigraphy, 649
proton-precession, 247, 248 Ridge Basin, California, 668
superconducting, 423, 424 Magnetostriction constants, 22, 23
vibration, 142 Magnetotactic response, 293, 298
Magnetometric investigations, 447, 451, Magnetotactic spirillum, 253; see also Bacteria,
522-525, 528, 552-553 magnetotactic
Magnetoreception, 156, 159, 162, 163, 164, Magnetotaxis, 279, 297, 305, 307, 648
509-521, 530-531 Makaira nigricans, 157,425,429
by biological superconductivity, 418 Manganese, 430
compasses, 251-252 nodule, 667
definition, 537 Map sense, 231, 259-265
by electrical induction, 418, 433 of vertebrates, 243
ferromagnetic, 432, 433 Marine magnetic lineations, 1, 246, 247
hypothesis, 245, 252 Marine waters, 294, 298
influence of magnets, 544-545, 548-551, 555, Marl, 667
557-558,613-614,619-620 Megaptera novaeangliae, 490, 498-499, 501-
intensity, 251 502
Langevin function constraint, 244, 247 Meissner effect, 105
by liquid crystal effects, 418 Membranes, 297, 303
magnetite-based, 419, 422, 426, 433 Messinian, 649
in newts, 442 salinity crisis, 649, 666
Index 679
Metazoans, 252 Navigation (continued)

Methodology, 563-568 magnetic, 247
Microfossil, 662 map sense, 259-265
Microorganism, 291-294, 297, 299, 303-307 pigeons, 535, 548-549, 617
chemoautotrophic, 660 route-based, 570
magnetotactic, 291-295, 297-301, 303, 305, Navigational armory, 613, 619
307 Near field mode, 131-132
Microscopy, see Electron microscope, Neurophysiological investigations, 530
microscopy; Optical microscopy Newt, eastern red-spotted, 441, 442
Microvillosities, 297 Nitrocellulose, 170
Midocean ridge basalt, 656 Nitrogen, liquid, 402
Migration, 1, 257-265, 489-492, 589 Noctua pronuba, 407
bird, 455, 464-469 Noise, 224
mean velocity of, 293, 305 thermal, 226, 399
Milkweed, 408-411 Non-magnetic tools, 157, 159, 169
Minimum perceptible signal, 227 North-seeking organisms, 294
Minnesota, 658 Northern anchovy, 425
Miocene (late), 647, 662 Nose drops, 544
Molluscs, 4, 5, 12 Notophthalmus viridescens, 441-442
Molpadia intermedia, 7, 9, 14 Nuclear 'V-ray absorption, 269
Monarch butterfly, 160,407-415 Null magnetic fields, 538
IRM,410-414 experiments, 387, 391
magnetic mineralogy, 413-414
migration, 408 Odontocetes, 489, 491, 501-504
natural history, 408 Oenocyte, 188
NRM,410-413 Oncorhynchus nerka, 418, 433
ontogeny of magnetic mineralization, 411- Oncorhyncus tshawytscha, 425, 426, 429
412 Ontogeny of magnetoreception, 519-521
Monkeys, 493, 502, 538, 573 Opal, 7, 9, 525, 647
marmosets, 554 Open marine environment, 666
rhesus, 552 Optical microscope, 293, 294, 299
squirrel, 538 Optical microscopy, 184, 188, 291, 299, 301
Monte-Carlo simulation, 535, 585, 586, 592, 615 reflected light, 170
Mopalia, 334, 335, 338, 340, 342 Organelles, 297
Mossbauer spectroscopy, 153, 269, 280 Organic material, solubilization by chemicals,
Moths, 407-408 169
Mouse (laboratory), 484-486 Organic matrix, 12, 13, 569, 602
Movement, 293, 294, 297, 305 Orientation, 513-521, 569, 602
Mu-metal, 198, 199, 200, 201, 210, 219 in amphibians, 439, 440
Multidomain (MD). 22, 23, 24, 25, 36, 143-148, effects on by clothing, 544, 554-556,616, 618
247, 665
human homing, 573-593
particles, 143-148
in reptiles, 439, 443
structure of, 292
Orientation cage, 511, 513-521, 542, 589
Myoglobin, 10
birds, 542, 618
Mysids,7
Orientation mechanism, 307
Mysticetes, 489-491
Orienteers, 543, 555
Oxidation (by bleach). 169
Nanofossil, 662
Natural remanent magnetization (NRM), see Paleomonetes, 376-377
Remanent magnetization, natural Paleomagnetism, 1, 153, 647, 655, 667
Natural selection, 1, 252 Panuliris (lobsters)
Naturists, 543, 554 magnetic material, 379
Nautilus, 7, 10 orientation, 367
Navigation, 245, 490-491, 494-495, 503-504, Paramagnetism, 138, 139, 141, 147, 305, 399
513-519,602 Paramecium, 504
compass sense, 51, 258-259 Parlodion, 170
humans 545-550, 569, 573-593, 611-621 Partially sighted, 601
680 Index
Particulate dispersions, 170-173 Remanence carrier, 626, 630, 653

Passerculus sandwichensis, 466, 468, 469 Remanence, 25
Passerina cyanea, 466 Remanent coercivity, 144, 145
Penaeus (shrimp) Remanent magnetism, 655, 667
behavior, 368-369 Remanent magnetization, 142-146, 198, 200,
magnetic properties, 371-374, 377 201,206-209,213,215,218,219
Permalloy, 200, 201 anhysteretic (ARM), 32, 36, 145, 213, 214,
Permeability, 198, 200, 201, 204, 207 218,219,647,649,668
effective, 200, 218 bio-,65
relative, 200, 208 chemical. 33, 65
Phocoenoides dalli, 490, 498, 500-502 depositional, 35
Phosphates, 5, 9 isothermal (lRM), 31, 153, 522-525, 528
Photosynthesis, 11 natural (NRM) , 18, 65, 157, 158,423,426,
Phytoferritin, 3, 4, 14 427, 484-486, 495-496, 498, 500-501,
Picofossil, 662 522-523
Pied flycatcher, 446, 468, 473 permanent magnetic domains, 202, 204, 215
Pigeon, 155, 159, 250, 251, 289, 292, 417, 418, relaxation time, 215
428 saturation isothermal (sIRM), 157, 158, 159,
homing, 243, 247, 254,455-464,469-475 160, 423-429, 433, 496-501, 648, 653
homing behavior in, 259-265 shock magnetization, 215
and magnetic anomalies, 248 thermo- (TRM). 33, 35
Pineal organ, 530-531 Reproducibility of results, 535, 546, 611-614,
Plagioclase, 656 621
Plantae, 4 Reptiles, 439, 443-452
Plastic resin, 153 Resins, 153
Platinum, 172 Reversion time, 299, 301, 306, 307
Polar membranes, 297, 303 Rib-bone, 552
Polychaeta, 5 Ring-billed gull, 464
Polyplacophora, 3, 14, 15; see also Chitons Robin, european, 464, 465, 466
Polystyrene latex spheres, 172 Rock dove, 455
Porifera, 4 Rock magnetism, 1
Potential energy (magnetic). 202-204, 206
Rodents, 493, 502, 504, 553-554
Praunus flexuosus, 8 woodmice, 553-554
Preservation, 640 Ruben's coils, 394, 486
Prokaryotes, 3, 7, 11, 12, 13
Protoctista, 12
Proton probe analysis, 552-553 Sacculus, 153
Protozoa, 4, 5 Salamander, 243, 254, 440-441, 493
Prussian blue reaction, 184, 189 Salmo salar, 432
Pseudosingle domain (PSD), 24, 25 Salmon, 254
model,35 Atlantic, 432
Pseudosingle domain-multidomain (d m ). 27 chinook, 425,426, 429
Pyrenoids, 297
fry, 243,418
Pyrite, 5, 10
smolt,418
Pyroxene, 656
sockeye, 418,433
Pyrrhotite, 635, 650
Sarda orientalis, 425, 429, 432
Saturation isothermal remanent magnetization
Quartz, 650 (sIRM) , see Remanent magnetism,
saturation isothermal
Radula Saturation magnetization, 20, 21, 142, 143, 293
apparatus operation (Polyplacophora), Scomber japonicus, 425, 429
335-337 Sculptures, pre-Columbian, 552
of chitons, 333-363 Seawater 660, 669; see also Marine water
sac anatomy, polyplacophora, 337-342 Sedimentary rock, 648
Red Sea hydrothermal system, 659, 668 shale, 667
Relaxation time, 402 marine, 659
Release-site bias, 471 Sedimentation rate, 647
Index 681
Sediments, 647 SQUID (cryogenic magnetometer), 103-105,

anoxic, 659 107, 108, 111-119, 121-130, 133, 136-
deep sea, 14, 667 139,142,147-149,197,198,483-484
grain-size distribution, 649, 655 direct current (DC), 107-111
lacustrine, 647, 656, 661, 669 electronics, 113-116
magnetic mineralogy of, 648 first-order gradiometer, 198
marine, 647, 650, 659, 667 flux coupling of SQUID sensors, 112, 116-
modern, 648 119
water transport of, 659 higher-order gradiometer, 198
Semicircular canals, 230 radio frequency (RF), 106, 107, 115
Sensitivity, 223 sensor, 111-113
Sensory mechanisms, 490-491 Stability field diagrams, 640
Shale, 667 Starlings, 250,457
Shape anisotropy, 29 Statistics, 542, 545-546, 563-568
Sharks, 225, 227 abuses, 584
first-order analysis, 590
Shielding factor(s). 198, 204, 205, 208, 215, 218,
independence, 542, 545-546, 613-615, 618
219
lumping, 584
Shielding layers, 198, 207, 213-215, 218, 219
multiple hypothesis testing, 542, 545-546
Shielding materials, 209
one-tailed tests, 542, 616
ferromagnetic metals, 198
one- vs. two- tailed tests, 563-565
high mu-materials, 198, 201, 218
post-hoc analysis, 542, 575, 584, 591, 615-
mu-metal, 198-201, 210, 219 616, 618
permalloy, 200, 201 second-order analysis, 542, 545-546, 586,
permanent magnetic materials, 208 588-590, 592, 593, 613-615
transformer steel, 198, 199, 201, 204-219 third-order analysis, 586, 588, 590
Shrew, short-tailed, 484-486 V4est 577-579, 581-586, 588
Sickle-cell erythrocytes, 538 Statoliths, 7
Siderophores, 6 Steel tools, 153
Signal/noise ratio (SIN), 227, 423, 427, 429 Stereo headphones, 544
Silica, 7 Sternaspis, 5
Silicon standard, 653 Stimulus
Silurian, 659 conditions, 420
Single-domain, 1, 21, 22, 24, 25, 36, 292, 648, discriminative, 419, 421
658, 667; see also Magnetite, single magnetic field, 418, 419, 420, 422
domain nonreinforced, 421
particle, 144-148 reinforced, 421
Single domain-pseudosingle domain (do), 27 Stingray, round, 418, 432
Skull bone, 552 Strandings, 490-492, 503
Snake, sea, 443 Striped bonito, 425, 429, 432
Sodium hypochlorite, 169 Sturnus vulgaris, 457
Soil, 669 Sulfate reduction, 669
Solar radiation, 664 Sulfur isotope fractionation, 669
Solar wind, 87 Superconducting magnetic shields, 119-121
South Atlantic magnetic anomaly, 291 Superconducting quantum interference device,
South-seeking organisms, 292, 294 see SQUID
Sparrow, savannah, 466, 468, 469 Superconducting ring, 107, 108, 111, 112
white-crowned, 475 Superconducting state, 104-106
Specimen grids, 170-172, 174 Superparamagnetic-single domain (d.), 27
Sphenoid bone, 552 Superparamagnetic particles, 292
Sphenoid sinus, 552-553 Superparamagnetism, 24, 26, 33, 35, 36, 37,
Spicules, 7, 8 147, 274
Spinel structure, 18, 19 of honeybees, 401
Spirillum, magnetotactic, 667 Susceptibility, 655
Sponge granules, 7, 15 anhysteretic, 145, 147
Spontaneous magnetization, 21 temperature dependence, 138, 139, 144
Spreading-drop gravity method, 170 Swallow, 471
682 Index
Swordfish, 432 Tursiops truncatus, 490, 492-495, 499-502

Sylvia atricapilla, 466 Turtle, 157, 159-162,247, 254,497-498, 502
Sylvia borin, 465, 466, 467 green, 443-444, 447-451
Sylvia communis, 465 loggerhead, 444-446
Symbiosis, 14 sea, 443
Tyndall effect, 169
Talitrus (sandhopper) orientation, 367
Teflon, 153 V-turn, 293, 294, 297, 299, 301
Temperature, 476, 477 Ultrasonic bath, 169
fluctuations, 250, 251 Ultrasonic shaker, 650
regulation, 254 Urolophus halleri, 418, 432
Thermal noise, 399
Thermomagnetic balance, 139 Vacuum evaporation, evaporator, 170-172, 177
Thermoreceptor, 250 Variance hypothesis, 245
Thin film, 170 Vendian, 11
Thiobacillus ferrooxidans, 14 Vertebrates, 250
Thunnus alalunga, 425 map sense of, 243
Thunnus albacares, see Tuna, yellowfin Volcanic dust, 659
Thunnus obesus, 425 Volcanic eruption, 654
Tissue Volcanic rock, 647
dermethmoid, 160, 161, 163
digestion, 160 Waggle dance, 386-393
Titanium, 430 Warbler
Titanomagnetites, 17, 20 garden, 465, 466,467
Tortonian, 649, 668 whitethroat, 465
Trains, 544 Water transport (sediment), 659
Transformer (or electrical) steel, 198, 199, 201, Weathering, 654
204-220 Weddelite, 7, 14
composition, 199 Whale, toothed, 489, 491, 501-504; see also
conductivity/resistivity of, 199 Cetaceans
examples for shielding, 214-219 Woods Hole, Massachusetts, 627
magnetic properties of, 207, 215, 218
permeability of, 207, 208 Xiphias gladius, 432
remanence of, 204, 208, 213, 214, 215, 218, X-ray camera, 161
219 X-ray diffraction, 162, 163, 168, 176-179,429,
Transportation cage, 515-518 430, 448, 449
Tumors, 162, 163 Debye-Sherrer method, 168, 176, 652
Tuna, 159-163 powder method, 176, 178
albacore, 425 X-ray microanalysis, 187, 191
bigeye, 425
magnetite of, 252 Zeitgeber, 397
skipjack, 423 Ziphius cavirostris, 490, 497-502
yellowfin, 254, 419, 420, 423-430, 432, 433 Zonotrichia leucophrys, 475

Magnetite Biomineralization and Magnetoreception in Organisms

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Magnetite Biomineralization and Magnetoreception in Organisms

Uploaded by

Copyright:

Available Formats

Magnetite Biomineralization

Volume 1 SKELETAL GROWTH OF AQUATIC ORGANISMS

Plenum Press • New York and London

©1985 Plenum Press, New York

R. Robin Baker Department of Zoology, University of Manchester, Manchester M13 9PL,

Subir K. Banerjee Department of Geology and Geophysics, University of Minnesota, Min-

Gordon B. Bauer Department of Psychology, University of Hawaii, Honolulu, Hawaii

Richard P. Blakemore Department of Microbiology, University of New Hampshire, Dur-

Edward R. Buchler ORI, Inc., Silver Spring, Maryland 20910

Shih·Bin R. Chang Division of Geological and Planetary Sciences, California Institute of

Tom Dayton Department of Psychology, University of Oklahoma, Norman, Oklahoma

Anne Demitrack Department of Geology, Stanford University, Stanford, California 94305

J. Robert Dunn Department of Geological Sciences, University of California, Santa Bar-

Richard B. Frankel Francis Bitter National Magnet Laboratory, Massachusetts Institute

W. 1. Goodman 2-G Enterprises, Mountain View, California 94043

Gary R. Scott Lodestar Magnetics, Inc.,' Oakland. CaUfornia 94608

I. Introduction and Background

Heinz A. Lowenstam and Joseph L. Kirschvink

Chapter 2 • Ferrimagnetic Properties of Magnetite

Subir K. Banerjee and Bruce M. Moskowitz

Chapter 3 • The Geomagnetic Field: Its Nature, History, and

II. Experimental Techniques and Instrumentation

M. Fuller, W. S. Goree, and W. 1. Goodman

Chapter 5 • Detection, Extraction, and Characterization of Biogenic

Michael M. Walker, Joseph 1. Kirschvink, Anjanette Perry,

Chapter 6 • Studying Mineral Particulates of Biogenic Origin by

Chapter 7 • The Cellular Localization of Particulate Iron

Chapter 8 • Large-Volume, Magnetically Shielded Room: A New

Gary R. Scott and Cliff Frohlich

3. Practical Techniques for Building Magnetically Shielded Rooms 208

III. Magnetoreception: Theoretical Considerations

Chapter 10 • Energetics and Sensitivity Considerations of

Chapter 11 • Particle-Size Considerations for Magnetite-Based

Joseph 1. Kirschvink and Michael M. Walker

Chapter 12 • Are Animal Maps Magnetic?

3. The Map Sense ....................................... 259

IV. Magnetoreception and Magnetic Minerals in Living

Chapter 13 • Mossbauer Spectroscopy of Iron Biomineralization

Richard B. Frankel, Georgia C. Papaefthymiou, and

1. Introduction to Mossbauer Spectroscopy . . . . . . . . . . . . . . . . . . . . 269

Chapter 14 • Magnetotactic Microorganisms Found in Muds from Rio

Henrique G. P. Lins de Barros and Darci Motta S. Esquivel

Chapter 15 • Structure, Morphology, and Crystal Growth of Bacterial

Chapter 16 • Biomineralization Processes of the Radula Teeth of

Michael H. Nesson and Heinz A. Lowenstam

Chapter 17 • Magnetic Remanence and Response to Magnetic Fields

Ruth E. Buskirk and William P. O'Brien, Jr.

Chapter 18 • Magnetic Field Sensitivity in Honeybees

William F. Towne and James 1. Gould

Chapter 19 • Magnetic Butterflies: A Case Study of the Monarch

Bruce J. MacFadden and Douglas S. Jones

8. Summary and Conclusions .............................. 414

Chapter 20 • Magnetoreception and Biomineralization of Magnetite:

Michael M. Walker, Joseph 1. Kirschvink, and Andrew E. Dizon

Chapter 21 • Magnetoreception and Biomineralization of Magnetite in

Anjanette Perry, Gordon B. Bauer, and Andrew E. Dizon

Chapter 22 • Avian Navigation, Geomagnetic Field Sensitivity, and

Chapter 23 • Magnetic Remanence in Bats

Edward R. Buchler and Peter J. Wasilewski

Chapter 24 • Magnetoreception and Biomineralization of Magnetite in

Gordon B. Bauer, Michael Fuller, AnjaneUe Perry, J. Robert Dunn,

Chapter 25 • Magnetoreception and the Search for Magnetic Material

V. Human Magnetoreception: An Editorial Introduction