Schuijer - Analyzing Atonal Music - Pitch-Class Set Theory and Its Contexts-University of Rochester Press (2008)

Analyzing Atonal Music
Schuijer.indd i 10/2/2008 10:29:33 PM

Eastman Studies in Music
Ralph P. Locke, Senior Editor
Eastman School of Music
Additional Titles in Music Analysis and on Music since 1920
Analyzing Wagner’s Operas: Alfred Lorenz Opera and Ideology in Prague:

and German Nationalist Ideology Polemics and Practice at the National
Stephen McClatchie Theater, 1900–1938
Brian S. Locke
Aspects of Unity in J. S. Bach’s
Partitas and Suites: An Analytical Study Pentatonicism from the Eighteenth Century
David W. Beach to Debussy
Jeremy Day-O’Connell
CageTalk: Dialogues with and about John Cage
Edited by Peter Dickinson The Pleasure of Modernist Music:
Listening, Meaning, Intention, Ideology
Concert Music, Rock, and Jazz since 1945:
Edited by Arved Ashby
Essays and Analytical Studies
Edited by Elizabeth West Marvin Ruth Crawford Seeger’s Worlds:
and Richard Hermann Innovation and Tradition in
Twentieth-Century American Music
Elliott Carter: Collected Essays
Edited by Ray Allen and Ellie M. Hisama
and Lectures, 1937–1995
Edited by Jonathan W. Bernard The Sea on Fire: Jean Barraqué
Paul Griffiths
Explaining Tonality:
Schenkerian Theory and Beyond The Substance of Things Heard: Writings
Matthew Brown about Music
Paul Griffiths
The Music of Luigi Dallapiccola
Raymond Fearn Theories of Fugue from the Age of Josquin to
the Age of Bach
Music Theory and Mathematics Chords,
Paul Mark Walker
Collections, and Transformations
Edited by Jack Douthett, Martha M. Variations on the Canon: Essays on Music
Hyde, and Charles J. Smith from Bach to Boulez in Honor of Charles
Rosen on His Eightieth Birthday
Music Theory in Concept and Practice
Edited by Robert Curry, David Gable,
Edited by James M. Baker,
and Robert L. Marshall
David W. Beach, and
Jonathan W. Bernard
A complete list of titles in the Eastman Studies in Music Series,
in order of publication, may be found at the end of this book.
Schuijer.indd ii 10/2/2008 10:30:17 PM

Analyzing Atonal Music
Pitch-Class Set Theory and Its Contexts
michiel schuijer
University of Rochester Press
Schuijer.indd iii 10/2/2008 10:30:18 PM

Copyright © 2008 Michiel Schuijer
All rights reserved. Except as permitted under current legislation,

no part of this work may be photocopied, stored in a retrieval system,
published, performed in public, adapted, broadcast, transmitted,
recorded, or reproduced in any form or by any means, without
the prior permission of the copyright owner.
First published 2008
University of Rochester Press

668 Mt. Hope Avenue, Rochester, NY 14620, USA
www.urpress.com
and Boydell & Brewer Limited
PO Box 9, Woodbridge, Suffolk IP12 3DF, UK
www.boydellandbrewer.com
ISBN-13: 978–1–58046–270-9
ISBN-10: 1–58046–270-7
ISSN: 1071–9989
Library of Congress Cataloging-in-Publication Data
Schuijer, Michiel.
Analyzing atonal music : pitch-class set theory and its contexts / Michiel Schuijer.
p. cm.—(Eastman studies in music, ISSN 1071–9989 ; v. 60)
Includes bibliographical references and index.
ISBN-13: 978–1-58046–270–9 (hardcover : alk. paper)
ISBN-10: 1–58046–270–7 (alk. paper)
1. Musical analysis—Data processing. 2. Atonality. 3. Computer composition.
4. Musical pitch. 5. Set theory. I. Title.
MT6.S35294A63 2008
781.2'67—dc22
2008030245
Portions of chapters 2 and 3 first appeared as “Was ist ein pitch-class set?” in Wahrneh-
mung und Begriff: Dokumentation des Internationalen Symposions, Freiburg 2.–3. Juni 2000
(Kassel: Gustav Bosse Verlag, 2000). A large part of chapter 3 was published as
“T & I: A History of Abstraction” in Tijdschrift voor Muziektheorie 6/1 (February 2001).
A catalogue record for this title is available from the British Library.
This publication is printed on acid-free paper.

Printed in the United States of America.
Schuijer.indd iv 10/2/2008 10:30:18 PM

in liefdevolle herinnering
Korien de Boer
Schuijer.indd v 10/2/2008 10:30:18 PM

Schuijer.indd vi 10/2/2008 10:30:18 PM
Contents
List of Illustrations ix
Preface xv
Acknowledgments xvii
1 Pitch-Class Set Theory: An Overture 1
2 Objects and Entities 29
3 Operations 49
4 Equivalence 84
5 Similarity 130
6 Inclusion 179
7 “Blurring the Boundaries”: Analysis, Performance, and History 218
8 Mise-en-Scène 236
Reference List 279
Index 293
Schuijer.indd vii 10/2/2008 10:30:18 PM

Schuijer.indd viii 10/2/2008 10:30:19 PM
Illustrations
Examples
1.1 Obvious and non-obvious combinations of tones in the opening
measures of Arnold Schoenberg’s Piano Piece, Op. 23, no. 3 5
1.2 Igor Stravinsky, Pas de Deux from Agon. An analysis of the
opening sonic field, mm. 411–13 8
1.3 A device of progression. The Webern-group 10
1.4 The octatonic scale as the common ground between the materials
of Schoenberg’s piano piece and Stravinsky’s Pas de Deux 13
1.5 Heinrich Schenker’s analysis of a song by Robert Schumann:
“Aus meinen Tränen sprießen” from Dichterliebe 14
2.1 Pitch numbers 30
2.2 Conversion of pitch numbers to pitch-class (PC) numbers 30
2.3 Pitch intervals 37
2.4 Conversion of pitch intervals to pitch-interval classes (PICs) 37
2.5 PICs are order-dependent 38
2.6 PICs associated with PC pairs 39
2.7 PICs and interval classes according to Allen Forte 39
2.8 Schoenberg, Piano Piece, Op. 11, no. 1, mm. 49–50 42
2.9 Instances of a pitch-class set. An example from Joseph Straus’s
Introduction to Post-Tonal Theory 42
3.1 The transposition of a melody. A musical example from
Heinrich Christoph Koch’s Musikalisches Lexikon 53
3.2 Twelve-tone transposition. The beginning of the prelude from
Schoenberg’s Suite for Piano, Op. 25 54
3.3 PC-set transposition in Anton Webern’s Movement for String
Quartet, Op. 5, no. 5 56
3.4 Instances of T9 in Webern’s Opus 5, no. 5 58
Schuijer.indd ix 10/2/2008 10:30:19 PM

x illustrations
3.5 Excerpts from Webern’s Opus 5, no. 5, selected by Allen Forte in

The Atonal Music of Anton Webern 61
3.6 The inversion of a fugue subject. An example from Nicola
Vicentino’s L’Antica Musica ridotta alla moderna prattica 63
3.7 The inversion of a fugue subject around B. An example from
Gioseffo Zarlino’s Le Istitutioni harmoniche 63
3.8 The inversion of a fugue subject according to
Giovanni Maria Bononcini 65
3.9 Friedrich Wilhelm Marpurg’s derivation of inverted forms
of the major and minor scales 66
3.10 Tonally modified inverted forms of subjects in fugues by
Johann Sebastian Bach 66
3.11 Axial symmetry in Webern’s Variations for Piano, Op. 27 70
3.12 The symmetrical arrangement of PCs in the closing section of
Schoenberg’s Piano Piece, Op. 23, no. 3 71
3.13 Inversionally related PC sets and symmetrical voice leading in
Webern’s Piece for Orchestra, Op. 6, no. 3 72
3.14 Inversion applied to the division of an interval (major sixth) 72
3.15 Inversionally related pitch intervals with respect to a referential
major sixth 73
3.16 The major sixth as a referential interval in Luciano Berio’s O King 76
3.17 Multiplication in PITCH 77
3.18 Interval expansion in Bela Bartók’s Music for Strings,
Percussion and Celesta 79
3.19 Evolution of the cycle-of-fourths and cycle-of-fifths transforms 80
4.1 PC sets extracted by Allen Forte from Schoenberg’s
Little Piano Pieces, Op. 19, no. 2 90
4.2 A part of Forte’s 1964 table of set-classes 93
4.3 Three instances of twelve-tone combinatoriality 95
4.4 The source set of Schoenberg’s Piano Piece, Op. 33a 96
4.5 “Z-related” pitch-class sets 99
4.6 Successive-interval arrays 101
Schuijer.indd x 10/2/2008 10:30:19 PM

illustrations xi
4.7 The opening of the third movement of Schoenberg’s Fourth String

Quartet, Op. 37 103
4.8 A part of Forte’s 1973 table of set-classes 108
4.9 The first two bars of Schoenberg’s Piano Piece, Op. 33a 112
4.10 John Rahn’s analysis of the opening measures of
Wagner’s Tristan und Isolde 112
4.11 Class types of pitch sets 125
4.12 PC set equivalence and sonic difference (Berg and Stravinsky) 127
5.1 Similarity relations in the Seventh Piano Sonata, Op. 64. by
Alexander Scriabin. An analysis by Allen Forte 135
5.2 Ernst Krenek’s chord classification 138
5.3 Paul Hindemith’s chord classification 140
5.4 Harmonisches Gefälle (Hindemith) 143
5.5 Maximally similar PC sets according to Forte (Webern,
Movement for String Quartet, Opus 5, no. 4) 145
5.6 Richard Teitelbaum’s computation of the “similarity index” (s.i.) 152
5.7 Vector subtraction 154
5.8 Maximal similarity in terms of REL2. Six pairs of PC sets singled
out by David Rogers 171
6.1 A segmentation of the main melody of Gabriel Fauré’s
Élégie for Cello and Piano, Op. 24 182
6.2 Members of a contour class 183
6.3 Segmentation and inclusion in Arnold Schoenberg’s
Piano Piece, Op. 11, no.1, mm. 1–11 184
6.4 The representation of set-class “6–16” in Schoenberg’s
Opus 11, no. 1 186
6.5 The representation of set-class “6-Z10” in the same piece 188
6.6 A family of subsets 192
6.7 A family of supersets 192
6.8 The inclusion of families of subsets and supersets 192
6.9 An analysis of the march section of Claude Debussy’s Fêtes 196
Schuijer.indd xi 10/2/2008 10:30:19 PM

xii illustrations
6.10 A tonal progression in Messiaen’s Mode 2, the octatonic.

An example from Le Banquet Céleste (1928) 201
6.11 The “Petrushka chord” 202
6.12 Johann Sebastian Bach, Nun lob, mein Seel, den Herren, BWV 17/7 203
6.13 Gottfried Weber’s chart of key relationships 204
6.14 Set-complex relations 208
6.15 The set complex about “5–33”/“7–33” 210
6.16 Analysis of Webern’s Bagatelle for String Quartet, Op. 9, no. 5 211
6.17 Families of chords in C major and C minor, projected on
Weber’s chart of key relationships 214
6.18 Harmonic dualism: the symmetrical alignment of major and
minor keys 216
7.1 Allen Forte’s and Richard Taruskin’s interpretations of a
passage from Stravinsky’s Rite of Spring 226
7.2 Nicholas Cook’s analysis of Schoenberg’s Little Piano Piece,
Op. 19, no. 6 231
7.3 Rudolph Réti’s reading of the opening measures of Mozart’s
Symphony No. 40, K. 550 233
8.1 Anton Webern, Piece for Violin and Piano, Op. 7, no. 1,
represented in the Ford-Columbia language (DARMS) 243
8.2 Webern’s Short Piece for Cello and Piano, Op. 11, no. 1, mm.
2–3. A segmentation following the rules of Forte’s
score-reading program 245
8.3 Mm. 1–3 of Webern’s Opus 11, no. 1 as segmented by
Forte in 1965 247
8.4 Dyadic invariance in Schoenberg’s Third String Quartet, Op. 30 253
Tables
1.1 Pitch-class set theory in the curricula of twenty teaching
institutions in the United States and Canada (1996) 19
2.1 The sum total of PC sets of successive cardinal numbers 47
2.2 PIC content 48
2.3 Absolute-PIC content 48
Schuijer.indd xii 10/2/2008 10:30:19 PM

illustrations xiii
3.1 An operation as defined by PC set theory 51

3.2 Milton Babbitt’s “group multiplication table” 51
3.3 The multiplication (mod 12) of PC values by factors from
1 through 11 77
4.1 Determination of the prime form of a PC set 106
4.2 The numbers of set-classes generated by three canonical
transformation groups 115
4.3 Early comprehensive chord counts in the twelve-tone universe 117
4.4 Fritz Heinrich Klein’s algorithm applied to dyads and triads 118
5.1 Distributive patterns in the relation 5:1 (Forte: second-order
maximal similarity) 146
5.2 Distributive patterns in the relation 5:1 (Forte: first-order
5.3 Distributive patterns in the relation 4:2 (Forte: first-order
5.4 Distributive patterns in the relation 6:4 (Forte: minimal similarity) 149
5.5 Distributive patterns in the relation 4:6 (Forte: minimal similarity) 149
5.6 Distributive patterns of PC sets with different numbers of elements 151
5.7 Distributive patterns in the relation 3:3. The difference is spread
over six interval classes 156
5.8 Distributive patterns in the relation 3:3. The difference only
involves two interval classes 156
5.9 The partition function of two PC sets 159
5.10 The partition function of two PC sets 159
5.11 The constant ratios between the subset classes of all cardinal
numbers in David Rogers’s six maximally similar PC set pairs 170
Schuijer.indd xiii 10/2/2008 10:30:19 PM

Schuijer.indd xiv 10/2/2008 10:30:20 PM
Preface
This book contains a musicological reflection on a body of music theory cover-
ing the second half of the twentieth century, early beginnings and late develop-
ments included. Few musicologists have concerned themselves with pitch-class
set theory. This is not surprising. Pitch-class set theory is a product of music the-
ory’s remarkable rise (or renaissance) as an independent academic discipline
in the United States—independent, that is, from musicology. A theory of pitch
relations in post-tonal music, it has chimed in with a deep interest in matters
of musical structure shown by composers and music theorists. And as far as the
study of musical works from the past is concerned, it has viewed the score as its
primary source of information. This goes against the grain of the musicologist
who, instead of isolating music, seeks to place it in a larger context—social, cul-
tural, historical, or otherwise.
However, I see no reason why this musicologist should avoid dealing with
pitch-class set theory altogether. Although it has developed into a highly dis-
tinct area of competence, it is part of the cultural environment in which music
is made, heard, and studied. It has provided a vocabulary for the description of
western art music from beyond the common practice period. It has established
an image of this music as being tightly, though not always conspicuously, struc-
tured, opening new avenues for analytical practice. Today, it is taught at colleges
and universities, and it is amply represented in the music literature. Seen thus,
pitch-class theory definitely falls within the scope of a contextualist approach
to music. By addressing its history, taking into account the variety of factors
that contributed to its development, this study aims to establish a new rapport
between musicology and music theory. As a consequence, the reader will find
historical narrative side by side—and sometimes inextricably entwined—with
theoretical and analytical discourse. It is my hope that this combination of per-
spectives will be to the enrichment of them both.
Schuijer.indd xv 10/2/2008 10:30:20 PM

Schuijer.indd xvi 10/2/2008 10:30:20 PM
Acknowledgments
The research for this book was conducted at the University of Utrecht as part
of a project funded by The Netherlands Organization for Scientific Research
(NWO). Paul op de Coul was a kind and stimulating supervisor. I retain good
memories of our spirited discussions. Rudolf Rasch provided expert guidance.
Over the course of twelve years, he has shown unbroken confidence in the
project, leaving me the freedom to shape it as I saw fit, and offering me good
advice when this freedom became unsettling. I thank him wholeheartedly for his
enduring support.
Several other people were involved in the review and improvement of the
manuscript: Henk Borgdorff, as a specialist in art theory and a close-enough
friend not to spare me the least bit of his criticism; Tijn Borghuis, because of
his expertise in mathematics and formal logic; and Paul van Emmerik, because
of his scrupulous scholarship, his knowledge of twentieth-century music, and
his interest in American culture. I also would like to acknowledge the percep-
tive remarks of Patrick van Deurzen, Erkki Huovinen, and Willem Wander van
Nieuwkerk. For answers to specific questions I am indebted to Jaap van Benthem,
David Fowler, Clemens Kemme, and Paul Scheepers. A special note of thanks is
due to Barbara Bleij, who provided critical comments and editorial advice dur-
ing the preparation of the manuscript, and to Gerard Goossens and Ger Vaessen
of Ascolta Music Publishing, who set the music examples with great skill.
This book was written directly in English. In matters of language, Johan Her-
renberg was an invaluable and always available mentor. Furthermore, I was happy
to receive English corrections from Ian Borthwick, Andrew Meyer, and Daniele
Sahr, who read parts of the manuscript in their capacity as native speakers.
Outside the circle of people who offered direct help in the process of writing,
there are some whose names deserve mention as well. The person with whom I
took the first steps to become knowledgeable about pitch-class set theory, in the
early 1990s, was Diderik Wagenaar, my analysis teacher at the Royal Conservatory
of The Hague. I recall with much pleasure the hours we spent on mastering an
analytical vocabulary that was new to both of us. I also enjoyed my conversations
and discussions with Allen Forte and Joseph Straus, which helped me in devel-
oping a balanced perspective on the theory. Allen kindly granted me permission
for the use of the photo on the cover.
The road to book publication can be long and hard, but I was fortunate
enough to have such an outstanding colleague and friend in Jonathan Cross,
whose enthusiasm for the project brought it to the attention of the University of
Schuijer.indd xvii 10/2/2008 10:30:20 PM

xviii acknowledgments
Rochester Press. And it was a privilege to work with URP’s friendly and profes-
sional staff. Special thanks to Eastman series editor Ralph Locke for his interest
and advice, and to managing editor Katie Hurley for her patient guidance. I owe
another debt to David McCarthy for his meticulous copyediting, and to Tracey
Engel, for helping me through the last stage of the production phase.
This project could only be finished with the support of two great families,
who did everything in their capacity to mend the loss that hit us all very hard in
1998. In particular, I want to thank my parents, Jan Schuijer and Elly Schuijer-
Brandenburg, and my parents-in-law, Otto de Boer and Thoma de Boer-Campagne,
for all their help and encouragement. I also want to mention my daughters,
Jantien and Marleen: tiny tots when the project started, strong allies when it had
to be finished under considerable pressure. Finally, my thoughts are with the
person who, although she has died, has never ceased reading over my shoulder.
This work is dedicated to her memory.
Schuijer.indd xviii 10/2/2008 10:30:20 PM

Chapter One
Pitch-Class Set Theory:

An Overture
A Tale of Two Continents
In the late afternoon of October 24, 1999, about one hundred people were
gathered in a large rehearsal room of the Rotterdam Conservatory. They were
listening to a discussion between representatives of nine European countries
about the teaching of music theory and music analysis. It was the third day
of the Fourth European Music Analysis Conference.1 Most participants in
the conference (which included a number of music theorists from Canada
and the United States) had been looking forward to this session: meetings
about the various analytical traditions and pedagogical practices in Europe
were rare, and a broad survey of teaching methods was lacking. Most felt a
need for information from beyond their country’s borders. This need was
reinforced by the mobility of music students and the resulting hodgepodge
of nationalities at renowned conservatories and music schools. Moreover, the
European systems of higher education were on the threshold of a harmoni-
zation operation. Earlier that year, on June 19, the governments of 29 coun-
tries had ratified the “Bologna Declaration,” a document that envisaged a
unified European area for higher education. Its enforcement added to the
urgency of the meeting in Rotterdam.
However, this meeting would not be remembered for the unusually broad rep-
resentation of nationalities or for its political timeliness. What would be remem-
bered was an incident which took place shortly after the audience had been
invited to join in the discussion. Somebody had raised a question about classroom
analysis of twentieth-century music, a recurring topic among music theory teach-
ers: whereas the music of the eighteenth and nineteenth centuries lent itself to
general analytical methodologies, the extremely diverse repertoire of the twen-
tieth century seemed only to invite ad hoc approaches; how could the analysis of
1. I have checked my account of the events described here with Patrick van Deurzen,
then coordinator of the Fourth European Music Analysis Conference. I am grateful for his
willingness to share his memories with me. For another account, see Cross 2000, 35. For
general reports of the conference mentioned, see Maas 2000 and Bernnat 2000.
Schuijer.indd Sec1:1 10/2/2008 10:30:20 PM

2 pitch-class set theory: an overture
works from this repertoire be tailored to the purpose of systematical training

without placing too much emphasis on particular styles of composition?
A late visitor entered the room, and seated himself on a chair in the middle
of the front row. He listened for a while to the discussion, his face expressing
growing astonishment. Then he raised his hand and said:
“You guys are discussing methods of analyzing twentieth-century music. Why

don’t you talk about pitch-class sets?”
He was American. The chairman, a professor from the Sorbonne, was quick
to respond:
“We do not talk about pitch-class sets, because we do not hear them!”
This dismissive response was not effective. The visitor said it was not his inten-
tion to discuss whether pitch-class sets could be heard, or had been used by com-
posers. He wanted to stress their value as an analytical tool. Pitch-class set theory,
he argued, enabled students to come to grips with complex music of the post-
tonal era. It was successfully applied in the United States, and he could hardly
believe that it was not taught anywhere in Europe!
Indeed it wasn’t, as he could infer from the reactions of some of the Euro-
peans present who appeared to be knowledgeable about the theory. They were
willing to admit that it worked well in some cases—for example, it was helpful in
clarifying the pitch structure of an early atonal composition by Anton Webern—
but they made it clear they would never encourage its general use. In their view,
it would force most music onto a Procrustean bed of preconceived relations.
The meeting had turned into a confrontation between European and Ameri-
can approaches to music analysis. And in the absence of a unified European
methodology, the participants ended up debating American practice only.
Although this turn took many by surprise—most of all the unhappy chairman—
it was to be expected at some point. For decades, the professional discourse on
music theory and analysis2 had been dominated by Americans. Whereas in most
European countries music theory offered training in basic musical skills as part
2. A note is necessary on my use of the terms “music theory” and “music analysis.”
These do not signify two separate disciplines. “Music theory” often functions as an umbrella
term comprising music analysis and other subjects, such as harmony and counterpoint.
However, it can also refer to a framework of concepts and/or protocols underlying the anal-
ysis of musical works. It is in this sense that I will use it. The corollary that music analysis is
“applied music theory” is not commonly embraced, as will be clear from what follows below.
European music theorists, including the British, often distinguish themselves from their
American counterparts by stressing the priority of analysis over theory. Characteristically,
one British study is entitled Music Analysis in Theory and Practice (Dunsby and Whittall 1988),
whereas the title of an American study is Music Theory in Concept and Practice (ed. James M.
Baker, David W. Beach, and Jonathan W. Bernard, 1997). For a historic discussion of the
relation between theory and analysis, see Cone 1967 and Lewin 1969.

pitch-class set theory: an overture 3
of the conservatory tradition, in the United States it had grown into an academic
discipline as well. Traditional pedagogy had been supplemented by an intensive
research program, which had made the Americans pre-eminent in the production
of music-theoretical knowledge. This pre-eminence was strongly apparent in the
field of music analysis, mainly because of the American adherence to two distinc-
tive bodies of analytical theory: the theory of Heinrich Schenker (1869–1935) for
tonal music, and pitch-class set theory3—commonly identified with the name of Allen
Forte (b. 1926)—for atonal music. On these twin pillars Forte himself based the
first American graduate program in music theory (at Yale University) in 1965.4
A division of analytical practice along the lines of these two theoretical bod-
ies is crude, of course. And today, as other perspectives on music analysis have
gained prominence in the United States, it is also outdated. However, it has been
of crucial importance to the image of American music theory. The Canadian
composer and music theorist William Benjamin has—perhaps unwillingly—
helped spread this image by addressing, in an article, the “curious marriage
of convenience” between what he saw as two contradictory streams of thought
(Benjamin 1981, 171). This metaphor came to the notice of the musicologist
Joseph Kerman, who cited it in Contemplating Music, his widely read, critical
account of postwar musical studies.5 It confirmed Kerman’s own critical assess-
ment of American music theory: “a small field built around one or two intense,
dogmatic personalities and their partisans” (Kerman 1985, 63). Meanwhile, Ker-
man saw more sense in the “marriage” than Benjamin, as did Patrick McCreless,
a commentator of later years:
Schenkerian theory and the current theories of atonal and twelve-tone music,
however mutually exclusive in terms of the repertoires that constitute their
objects, both share a value system that explicitly privileges rigor, system, and
theory-based analysis and implicitly share an aesthetic ideology whereby analy-
sis validates masterworks that exhibit an unquestioned structural autonomy.
(McCreless 1997, 32)
In Rotterdam, the dual image of American analytical practice was prevalent in

the minds of most conference-goers, and—much more importantly—so was its
privileging of “rigor” and “system.” When the chairman said: “We do not hear
these pitch-class sets,” he appealed to a common concern about “theory-based
3. Some people call it “set theory” (without the prefix “pitch-class”), reducing its
name to that of its mathematical model: Georg Cantor’s Mannigfaltigkeitslehre or Mengen-
lehre. Others speak of “set-class theory.” Although these alternative names are sometimes
used for good reasons—which will become obvious as we proceed—I have decided to
stick to the designation “pitch-class set theory” throughout this study.
4. Significantly, this program was not founded within Yale University’s professional
music school, but within the Department of Music, where it was associated with graduate
studies in musicology. Forte 1998b, 10.
5. Published in England under the title Musicology.

analysis” as practiced in the United States—a concern shared not only by those
European music theorists involved in the teaching of musical performers, but
also by American scholars like Kerman. He made himself the spokesman for
those who suspected such analysis of seeking to demonstrate the workings of the
theory rather than to reveal what is special about the music.
There may well have been grounds for this suspicion, although in Europe it
cannot always be dissociated from an anxiety about American hegemony over the
discipline. After all, Europe has its own breeds of analytical systems, and Heinrich
Schenker lived in Vienna! The European Music Analysis Conferences, for all the
work toward an entente cordiale between the national schools of musical analysis, have
also been motivated by a desire to “remind the theoretical community of its Euro-
pean heritage” (Street 1990, 357). Pitch-class set theory—which is not part of that
heritage—already divided Europeans and Americans at the very first of these con-
ferences, held in Colmar in 1989. Then, Allen Forte crossed swords with the Bel-
gian music theorist Célestin Deliège, who, in a preparatory paper, had criticized the
theory for lacking explanatory power (Deliège 1989). More than a decade later, the
controversy was still alive. What was it that aroused such persistent antagonism?
Paradigmatic Pieces
Pitch-class set theory addresses the notion of musical coherence. There are many
ways in which music can be said to cohere. For example, it may be in one key, or
be in a familiar form, it may obey a model of rhetoric, or it may have been set to a
single text. Pitch-class set theory seeks coherence in the relations between various
combinations of tones. A work susceptible to this approach is Arnold Schoenberg’s
Piano Piece Op. 23, no. 3 (1923). Example 1.1a shows its beginning.6
The piece first presents a single melodic line, involving the pitches B♭4, D4,
E4, B3, and C♯4 respectively.7 The same melody, now starting on F3, is formed by
the successive bass tones in measures 2–3. And in measure 4 there is an inverted
form of this melody starting on D5 in the top voice. These three statements are
isolated and bracketed in the lower half of example 1.1a.
6. This piece is a favorite among analysts, especially those concerned with Schoen-
berg’s path toward serial composition. Analyses of these opening measures have been pro-
vided by Stein (1925), Perle (1962), De Leeuw (1977), and Simms (2000), among others.
I do not pretend to add anything new to the work of these authors. It is my sole intention
to give the reader a basic understanding of what pitch-class set theory is about, in anticipa-
tion of the more systematic presentation in subsequent chapters.
7. With regard to the letter names of pitches, I follow the rule proclaimed by the
Acoustical Society of America. According to that rule, a pitch is identified by a capital let-
ter, a sharp or a flat if so required, and by the number of its octave range. Octave ranges
cover the pitches from any C through the next higher B, and are numbered from low to
high. Middle C is C4.

1 Langsam ( = ca 54) 2 3
dolce
poco rit.
5
4
2 3
5
4
Example 1.1a. The beginning of Schoenberg’s Piano Piece, Op. 23, no. 3. Obvious
combinations of tones: three statements of a five-tone melody.
FÜNF KLAVIERSTÜCKE, OP. 23, by Arnold Schoenberg. Copyright © Edition Wilhelm Hansen
AS. International Copyright Secured. All Rights Reserved. Used by Permission.
So far, an analysis of this piece does not seem to require a language other
than that used for any work featuring imitative counterpoint. The melody actu-
ally functions like the subject of a fugue, particularly since it is first answered “at
the fifth.” The rhythm of the melody is slightly different at the second entry, and
then is transformed beyond recognition at the third. Still, imitation—another
device to achieve musical coherence—sanctions the combinations shown in
example 1.1a. It is obvious that these tones belong together.
On closer scrutiny, however, less obvious groups of tones appear to relate to
the opening melody as well. These are shown in example 1.1b. Each of them
can be transformed into another statement of this melody, in recto or in inverso

α
δ
2 3
γ ε
2 3
Example 1.1b. Non-obvious combinations related to the opening melody of

Schoenberg’s Piano Piece, Op. 23, no. 3.
β:
8va
γ:
δ:
8va
ε:
Example 1.1c. New statements of the melody derived from the combinations in
ex. 1.1b.
(ex. 1.1c). We only need to reorder the pitches and replace some of them by
a higher or lower octave. Although these combinations are not musically artic-
ulated, we can define them on the basis of their hidden “substance.” Thus, a
dense web of relations emerges, in which the opening melody imposes its struc-
ture on the harmonies and accompanying figures.
We can continue this exploration, and add to the first statement of the melody
the first tone of the second statement (F3), which sounds between the former’s B3

2 3
Example 1.1d. Non-obvious six-tone combinations.
and C♯4. Now we have six tones instead of five (see ex. 1.1d, first 4 beats). These
six tones form a shape that similarly recurs a couple of times in the first measures.
Let us look at the second five-tone combination in example 1.1b, which is marked
β. If we enlarge it to include E♭3—a tone originally assigned to the next combina-
tion (γ)—we have another instance of our six-tone combination (as shown in ex.
1.1d, in the center). It appears that β and γ are connected in the same way as the
first two statements of the melody: γ answers β “at the fifth.” In both cases, the six-
tone combinations result from the progression from one five-tone combination
to the next. In measure 3, the progression from δ to ε is not a “fifth progression”;
however, it yields a similar six-tone combination (see ex. 1.1d, to the right).
The brief analysis above reveals a tightly woven pattern of recurrence. This
tightness is not characteristic for each measure of this piece. For example, mea-
sure 4 is already less “close-knit” than the previous measures.8 But it is the coher-
ence of measures 1–3 that concerns us here. What is it that recurs so consistently
throughout these measures? We have noted several related entities, but what do
these have in common? They can all be traced to the opening melody, but some
of them are not in the slightest way a representation of that melody. What they
actually represent is a basic property of the melody’s pitch material. Pitch-class
set theory is concerned with such properties. A pitch-class set is an abstract con-
cept of a combination of musical tones. It does not include the durations and
octave ranges of these tones; nor does it include the order in which they appear.
It reduces the combination to an unordered set of collective pitch designations
(“pitch classes”). By using this concept, analysts can trace relations that are inde-
pendent of shape and actual pitch content.
8. I agree with Bryan Simms, who writes: “As the piece progresses, dividing the entire
texture into variants of the initial shape becomes ever more difficult and requires analytic
strategies that fully bypass the musical context and, in all likelihood, the composer’s inten-
tions” (Simms 2000, 198). It is by no means foreign to the idea of contrapuntal writing to
vary the pace of thematic development. In a fugue, thematic passages often alternate with
freely constructed episodes. This holds true for Opus 23, no. 3 as well.

Adagio = 112
3
Violino
Solo
marc.
espress.
411 412 413
3
I
marc.
Violini espress.
3 3 3
II
arco 3
Viola
ma marc. tutti arco

Violoncelli
pizz.
Contrabassi
poco
Example 1.2a. Stravinsky, Pas de deux from Agon. The opening sonic field, mm. 411–13.
© Copyright 1957 by Hawkes & Son (London) LTD. Reprinted by kind permission of Boosey &
Hawkes Music Publishers LTD.
1 2
1 3
Example 1.2b. Abstract forms of the motifs in the first-violin part, mm. 412–13.
Another composition in which such relations have been found is Igor Stravinsky’s
ballet Agon (completed 1957). Example 1.2a shows the first three measures of the “Pas
de deux” (mm. 411–13). These have a more open texture than the first three mea-
sures of Schoenberg’s Opus 23, no. 3. Rather than a contrapuntal fabric, they pres-
ent a sonic field. This field is built from several tone combinations. Most conspicuous
are the motifs played by the first violins in measures 412–13, which mark the end of
these introductory measures. We can project the tones of each motif in a single octave
range, and then place them in ascending or descending order (ex. 1.2b). It appears
that the motifs reduce to different forms: a succession of a minor and major second
(in chromatic steps: 1 and 2), and a succession of a minor second and a minor third
(in chromatic steps: 1 and 3). One motif cannot be transformed into a statement of
the other by the rearrangement and octave displacement of pitches.

411 3 412 3 3 413
immobile
segment 3
411 arco 412 413 pizz.

mobile
segment
Example 1.2c. The mobile and immobile segments of the sonic field.
However, each of these motifs reflects the structure of a segment of the sonic
field from which they emerge. Example 1.2c shows that this field consists of a
mobile and an immobile segment. The former is a progression involving the
pitches C3 (mm. 411–12, violas), D♭3 (mm. 412–13, violas and cellos), and E♭2
(m. 413, double basses); the latter consists of the repeated D4s, B♭3s and C♭5s
played by second violins and violas. The first motif reflects the structure of the
mobile segment, and the second that of the immobile segment. The motifs and
segments, it should be noted, partition the total chromatic; the sonic field grows
into a twelve-tone field.
Significantly, each of the two pieces presented above was written just before
or at the time of its composer’s turn to twelve-tone serialism. Neither of them
is strictly twelve-tone, but they both display a remarkable consistency in the use
of pitch intervals. Henri Pousseur (1972) has noted that Stravinsky’s “Pas de
deux” employs a group of intervals taken from the twelve-tone series of Anton
Webern’s Variations for Orchestra, Op. 30—a work Stravinsky is known to have
admired (ex. 1.3a). This group of intervals—scalewise in chromatic steps: 1 2
1—combines the two intervallic patterns that emerged from our analysis of the
first three measures (see ex. 1.3b). Indeed, from measure 414 onwards we can
see that it is repeatedly generated by adjacent or overlapping occurrences of
these smaller patterns (ex. 1.3c and 1.3d). It thus seems to function as a device
of progression, similarly to the six-tone combination in Schoenberg’s Piano
Piece, Op. 23, no. 3 (ex. 1.1b and 1.1d above). In any case, it is the product of an
intervallic consistency reminiscent of—if not inspired by—twelve-tone serialism.
It is not by accident that two musical works approximating dodecaphonic
practice reflect so well the focus of interest of pitch-class set theory, for this theory
is itself an outgrowth of the theory of twelve-tone serialism. Its pioneer was the
composer Milton Babbitt (b. 1916), who had dedicated himself to extending, and
strengthening the theoretical underpinning of, Schoenberg’s twelve-tone tech-
nique. Babbitt set the example for the confluence of music theory and contempo-
rary composition under the aegis of the university, America’s principal employer
of composers. He taught at the music department of Princeton University,

Example 1.3a. A source of inspiration: a recurring group of intervals in the twelve-

tone series of Webern’s Variations for Orchestra, Op. 30.
1 2
1 3
1 2
1 3
Example 1.3b. The derivation of this group from the abstract forms of the motifs in
ex. 1.2b.
which owes to him its reputation as an American center of musical innovation.

His theory proceeded from a mathematical description of the familiar twelve-
tone operations (transposition, inversion and retrogradation). A good deal
of its conceptual apparatus was adopted by Allen Forte, who tailored it to the
analysis of non-serial music. For example, whereas Babbitt used the word “set”
for “twelve-tone series,” Forte also applied it to combinations involving less than
twelve tones and showing no definite ordering—combinations such as found in
Schoenberg’s piano piece and Stravinsky’s “Pas de deux.”
The relations shown in example 1.1 through 1.3 are instances of what Schoen-
berg saw as a continuous process of variation, the “endless reshaping of a basic
shape” (Schoenberg 1984, 290). The term “basic shape” is not very specific. It
might signify a twelve-tone series. If, alternatively, we replace it by “pitch-class
set,” we interpret it more generally—in other words, we impose fewer constraints
on the process of its “reshaping.” However, pitch-class sets do not necessarily
function as “basic shapes” in the Schoenbergian sense. The recurrence of a basic
shape is not the only measure of musical coherence posited by pitch-class set
theory. There are other categories of relations between groups of tones: rela-
tions between their interval contents, relations based on common tones, and
relations delineating hierarchical levels of structure, to mention but a few. Pitch-
class set theory imposes no limit on the number and nature of relations between
tone combinations.

Vln. solo 414 415
3
Vla.
pizz. 3
Vc. arco
poco
DB arco
poco
1: F G A A
3
2: F G A A
3: A A B C
3
4: B C D E
Example 1.3c. Stravinsky, Pas de deux from Agon. Four occurrences of the
Webern-group in mm. 414–15. The circled pitches form the same intervallic
patterns as those extracted from the motifs in mm. 412–13.

pizz.
Vln. I/II arco
459 460 461 462
Va. /Vc.
arco
pizz.
Example 1.3d. Stravinsky, Pas de deux from Agon, mm. 458–62. The Webern-group
as a device of progression. Each three-tone chord (circled) is complemented by a
member of the next, or the previous chord.
For example, the theory also deals with the question of how to conceive a rela-
tion between the two motifs in measures 412–13 of Stravinsky’s Agon (ex. 1.2). It
has provided ways to measure the degree to which such motifs are similar or dif-
ferent. (As shown in example 1.2b, the abstract forms of these motifs both con-
tain the intervals of a minor second and a minor third. The remaining intervals
are different: in one we find a major second, in the other a major third.) Even if
there is not a single basic shape underlying the different melodies, motifs, and
simultaneities, a composition may be shown to make a consistent use of particu-
lar tone combinations. These combinations may feature the same intervals, or
follow one another according to a rationale of progression.
We can take yet another view of the two motifs in Agon and stress that they are
both abstractly contained in the four-tone combination derived from Webern’s

Schoenberg (‘melody’)
1 2
1 3
Stravinsky
Example 1.4. The octatonic scale as the common ground between the materials of
Schoenberg’s piano piece and Stavinsky’s Pas de Deux.
Opus 30 (as can be seen from ex. 1.3b). Each of them “descends” from the lat-
ter. In turn, this four-tone combination “descends” from combinations of yet
higher magnitudes. In this way, we can plot entire family trees of pitch-class sets.
It is interesting that the five- and six-tone combinations found in Schoenberg’s
Piano Piece, Op. 23, no. 3, and the three- and four-tone combinations found
in the “Pas de deux” from Agon can be assigned to one family, since they all
“descend” from the octatonic scale (ex. 1.4). The octatonic background is most
conspicuous in measures 413–15 of the “Pas de deux,” where the overlapping
three- and four-tone combinations actually complement this scale (ex. 1.3c).
These family-like relations have enabled pitch-class set theory to account for
higher levels of organization in a musical work, and to hypothesize principles of
structure governing an entire repertoire of music.
This aspect of the theory—its working toward a hierarchy of structural lev-
els—is reminiscent of its tonal-music counterpart in the United States: the ana-
lytical theory of Heinrich Schenker.9 This is not surprising when we consider
that Allen Forte was a devoted Schenkerian. In 1982, he and his former student
Steven Gilbert published a much-consulted introduction to Schenker’s method
(Introduction to Schenkerian Analysis); and throughout his career he has used it for
the analysis of a wide range of music, including works from the late nineteenth
and early twentieth centuries (Contemporary Tone-Structures, 1955; “Schenker’s
Conception of Musical Structure,” 1959), and American popular songs from the
1930s and 40s (The American Popular Ballad of the Golden Era, 1995). Therefore, a
brief digression on Schenker’s approach to tonal music is now appropriate.
A Short Detour to Schenker

Schenkerian theory claims that harmonic progressions and voice-leading pat-
terns governing single phrases of tonal music are also at the basis of entire
9. For a general reflection on the relationship between Schenkerian theory and

pitch-class set theory, stressing their common roots in nineteenth-century organicism, see
Hinton 1988. For an analysis combining pitch-class sets with a Schenkerian approach, see
Forte 1988b.

∧ ∧ ∧
3 2 1
I V I
Example 1.5a. A Schenkerian fundamental structure (Ursatz).
Schumann, Dichterliebe,
∧
Op. 48II
3 ∧ ∧ ∧ ∧
b) 2 3 2 1
(= 3
2 1)
(Nbn)
( )
(8 7)
12
T 4 8 (!)
(Nbn)
(Wdh)
Kons
Dg. (
Vdg I IV V I V ) I
(A1 B A 2)
Example 1.5b. Heinrich Schenker’s analysis of a song by Robert Schumann:

“Aus meinen Tränen sprießen” from Dichterliebe. This analysis was published in
Schenker’s treatise Der Freie Satz (1935).
© 1935 by Universal Edition A.G., Wien/UE 6869A.
compositions. Small-scale patterns thus mark stages in the unfolding of larger

ones; they “prolong” these stages. Through a series of reductions a Schenkerian
analysis retraces the process of the work’s unfolding to the “fundamental struc-
ture” (Ursatz), which is itself the prolongation of the tonic triad. This fundamen-
tal structure consists of two simultaneous parts: a fundamental line—a stepwise
descent from the tonic third (or fifth, or octave) to the tonic—and a bass arpeg-
giation I V I that supports it (ex. 1.5a). It is important to note that the purpose
of the analysis is the discovery, not of the fundamental structure, but of the way
that leads from this fundament to the actual work.
Example 1.5b has been taken from Schenker’s treatise Der Freie Satz (1935).10
It shows three successive reductions of Schumann’s song “Aus meinen Tränen
sprießen” from the cycle Dichterliebe, Op. 48. Schenker himself did not waste
many words on this analytical graph. However, it has provoked much comment
from others. It has often been used as an example showing both the pros and
cons of Schenker’s analytical method: the insights it can offer, and the problems
it may raise.
The lowest staff looks like a rough draft, with analytical markings indicating
important tones and connections. A particular emphasis appears to be placed on
10. Translated in English as Free Composition (1979). The example is numbered 22a.

the first tone of the opening phrase (C♯5; beam, open note-head) and the first
tone of the contrasting middle section (B4 in m. 8; beam). These tones form
a large-scale progression: a descent from the third (^3) to the second (^2)of the
A-major scale. The same progression starts in measure 12, which makes sense,
as an altered recapitulation of the opening phrase begins here. But now the
progression is shown to reach the tonic A4. This tone and the B4 that precedes
it are represented by open note-heads, which suggests that they are more impor-
tant than the corresponding tones in measure 4.
The middle staff gives much less detail. Its function is to highlight the progres-
sion from C♯5 to B4, and later that from C♯5 to A4. The brief middle section
(“Und wenn du mich lieb hast”) is shown here to prolong the B4 in measure 8,
and to connect it to the C♯5 of the recapitulation. The middle staff thus rein-
forces the emphasis placed on the outer phrases. And we can now see why these
phrases are emphasized: they contain the elements of the fundamental structure
shown in ex. 1.5a.
The upper staff represents a basic elaboration of this fundamental structure.
It shows us that in Schumann’s song the descending third progression is first
interrupted and then stated completely; the resolution of the treble and bass
into the octave is suspended until just before the end.11 In Schenkerian theory,
“interruption” is an over-arching concept. It covers a span of music that is com-
posed of different parts. This can be a small span (like an eight-bar period) or
a large one (like a movement in sonata form). In other words, the term “inter-
ruption” can be applied to different levels of structure. It epitomizes Schenker’s
hierarchical approach to music analysis.
There are two reasons to dwell on example 1.5b. First, Forte presented
Schenker’s analytical graph in an article for Journal of Music Theory in 1959, and
went into great detail to explain it. This article—entitled “Schenker’s Concep-
tion of Musical Structure”—was also important for other reasons, which will
be clarified in chapter 8. Second, it may provoke the objection that it imposes
an a priori structural model on the song. It thus reinforces the image of a “the-
ory-based” or “system-oriented” analytical practice, an image that Schenkerian
analysis shares with pitch-class set analysis. Joseph Kerman, in his commen-
tary on both Schenker’s interpretation and Forte’s rendering of it, raised this
objection (Kerman 1980).12
It is questionable whether the graph of example 1.5b provides a good example
of Schenkerian analysis. In any case, Schumann’s song is a questionable example
of a Schenkerian interruption. This concept—a premature halt, followed by a
11. It is unclear to me why Schenker chose not to beam the progression B–A in the
last two measures, but did beam the corresponding progression in measure 15. It is obvi-
ous that the latter does not end the final phrase.
12. Kerman involved yet another analysis, which I take the liberty to exclude from
consideration: an analysis published by Arthur Komar in 1971.

new start—is perfectly illustrated by classical themes in period form;13 in “Aus

meinen Tränen sprießen,” however, the descending third progression starting
on C♯5 is not interrupted at the end of the first phrase—as it would be in a
period—but is stretched to the beginning of the next phrase. Furthermore,
one remarkable thing about this song is the stealthy beginning of its recapitula-
tion. The tonic third C♯5 is now the leading tone of a dominant seventh chord
applied to D. As a consequence, it does not demand the same kind of attention
that it received at the opening of the song. Neither the break in the continu-
ity of the descending third progression in measure 8, nor the new start of that
progression in measure 12 is musically articulated as such. Therefore, the term
“interruption” does not seem to apply naturally to Schumann’s song.14
Schenker’s approach to Schumann’s song seems to be geared too much
toward the validation of a general concept of tonal structure, ignoring what is
inconsistent with that concept. But what is it that we value in an analysis? Forte
selected this one as typical for Schenker’s achievement, which he summarized
as follows:
Schenker opened the way for a deeper understanding of musical structure

with his discovery that the manifold of surface events in a given composition is
related in specific ways to a fundamental organization. (Forte 1959, 4)
Forte liked the systematic, generalizing tendency of Schenker’s late output, and
saw his own role in the light of this tendency:
Although Schenker came very close to constructing a complete system, further

refinement and amplification are required if it is to fulfill its promise. (ibid., 16)
In spite of the dangers of overgeneralization, we should consider the reason

for imposing such concepts as the Schenkerian interruption: it strengthens the
image of music theory as a domain of competence distinct from musicology,
musical performance, and composition. Within these disciplines music theory
had traditionally played a subservient role. It provides historical musicologists
with the knowledge and tools to access musical sources from the past; it enables
performers to maximize their awareness and control of musical processes; and
13. As defined by William Caplin: “Essential to the concept of the period is the idea
that a musical unit of partial cadential closure is repeated so as to produce a stronger
cadential closure” (Caplin 1998, 49). An interruption in the Schenkerian sense demands
that the first unit end with a half cadence.
14. Forte did not make this point; nor did Kerman. This, however, does not detract
from the value of their observations. Forte saw the G in measures 12–13—which turns the
initial A-major chord into a dominant seventh chord—as resulting from a chromatically
descending inner voice (Forte 1959, 24). Kerman stressed this tone’s expressive quality in
relation to the word “klingen” (Kerman 1980, 326).

it offers composers a foundation for critical self-reflection. However, it does not

stand on its own two feet. It was for analytical theories like those of Schenker
and Forte to achieve that self-consistency (Forte’s “promise”), sanctioning inter-
pretations that, from other disciplinary perspectives, may strike one as subver-
sive rather than subservient.
Institutionalization and Criticism
Judging from the reactions it elicited at the European Music Analysis Confer-
ences, pitch-class set theory has not found an easy entry into Europe. Indeed,
only in Britain has it become part of the standard repertoire of analytical
resources. In North America, however, it has enjoyed a wide dissemination. As
noted earlier, it has played an important role in the emancipation of music the-
ory as an academic discipline in its own right. This emancipation was marked,
first, by the appearance of specialized journals such as Journal of Music Theory
(Yale University, 1957), Perspectives of New Music (Princeton University, 1961),
and The Music Forum (Columbia University, 1967–87); second, by the increasing
number of degree programs in music theory at American universities, especially
the PhD programs that were established after the example of Yale’s (1965); and,
third, by the foundation of the Society for Music Theory (SMT) in 1977. The
SMT was instrumental in, among other things, the organization of a large num-
ber of conferences, and provided its members with another important journal
from 1979 onwards (Music Theory Spectrum). By 1980, then, an infrastructure was
available that encouraged research, facilitated an ongoing professional debate,
and raised the profile of music theory as a “body of knowledge and a set of
shared practices.”15
Pitch-class set theory has not itself spurred this development from the begin-
ning. However, it has contributed to it by putting on the agenda the theoretical
underpinnings of the twentieth-century post-tonal repertoire. Apart from provid-
ing a technical vocabulary, it has helped to formulate the premises and questions
from which to proceed in the analysis of works from this repertoire: what was to
15. This is another quote from Patrick McCreless (1997, 17). He has described
the birth and growth of academic music theory from a perspective developed by
the French philosopher Michel Foucault, stressing the dependence of knowledge
on power: “Music theory is in fact, like all academic disciplines, a ‘docile body’—an
object of control—with respect to the university, just as, in another sense, most music
theorists, as individuals and employees of the university, are ‘docile bodies.’” (ibid.,
35) What is characteristic for this approach is the emphasis on a self-regulating aca-
demic discourse dictating the contributions of its individual participants. Although
the present study will pay a good deal of attention to the intrinsic dynamics of the
evolving discourse on pitch-class set theory, it will not refrain from also addressing the
decisive role of personal involvement.

be searched for, and how this was to be done.16 The influence it thus exerted
on the study (and the composition) of post-tonal music has been profound and
long lasting. The rise of pitch-class set theory has earned American music theory
a reputation for engagement with musical modernism. However, the flurry of
scholarly activity has left the modernist tradition of composition with mixed for-
tunes. On the one hand, this activity is a genuine response to that tradition,
which has yielded textbooks and manuals of analysis giving students access to
the music of such “difficult” composers as Schoenberg, Webern, Boulez, Carter,
Babbitt, and (late) Stravinsky. On the other hand, this music’s exposure in the
literature—combined with its minor role in public concert life—has added to its
reputation of being “cerebral” and “academic.” Pitch-class set theory employs a
mathematical vocabulary and mathematical models of presentation; and to des-
ignate pitches, it has substituted integers for the traditional letter names. Analy-
ses proceeding along set-theoretical lines, then, convey an image of rationality
that may confirm people in their rejection of the music concerned.
The integration of pitch-class set theory into the curricula of American col-
leges and universities is probably the most significant measure of its institution-
alization. Table 1.1 may give us a tentative impression of the extent to which it
has influenced music theory teaching at the college and graduate levels. It shows
its place in the curricula of twenty institutions spread over the United States
and Canada. The table reflects the situation in 1996. At that time, all of these
institutions offered degree programs in music. Apart from their geographical
distribution, they differed by type. They included private research universities
with music schools (University of Rochester) or music departments (Harvard
University, Columbia University), a private university with an integrated college
(Bradley University), state universities with music schools (Florida State Univer-
sity, Universities of Iowa and Michigan) or music departments (Universities of
Virginia and New Mexico), liberal arts colleges (Simpson College and Davidson
College), a college conservatory (Oberlin), and a community college (Pima, the
music program of which may have been exceptional for this type of institution).
In view of this diversity, it is significant that only one institution from this
group did not teach pitch-class set theory in 1996. This was Davidson College.
All the others had included it in their programs, although it received little cover-
age at the University of California in San Diego, at Memorial University of
16. Seen thus, pitch-class set theory has played a role comparable to that which
Thomas Kuhn (1962) called a “paradigm.” There are different interpretations of this
term, but in one sense a paradigm is an intellectual achievement that sets the course for
subsequent research in a particular field. More specifically, it confronts researchers with
“puzzles,” while at the same time providing them with tools for solving these “puzzles.”
For another comparison, Larry Laudan’s concept of a “research tradition” is worth con-
sidering: “a set of general assumptions about the entities and processes in a domain of
study, and about the appropriate methods to be used for investigating the problems and
constructing the theories in that domain.” (Laudan 1977, 81)

Table 1.1. Pitch-class set theory in the curricula of twenty teaching institutions in the United States and Canada (1996)
Schuijer.indd Sec1:19
Institute Representation Academic level Textbooks/manuals used
Bradley University, Peoria, IL. Separate course Undergraduate Rahn 1980

(Forte 1973 and Straus 1990,
articles, dissertations)
University of California, San Diego Integrated (Music Theory Undergaduate (Rahn 1980 and Straus 1990)
and Practice II )
Columbia University, New York Separate Course Graduate Own materials
(Perle 1962, Forte 1973, Rahn
1980, Straus 1990, Morris 1991)
Davidson College, Davidson, NC None
Florida State University, Tallahassee Integrated (20th-Century Undergraduate Straus 1990
Styles, Atonal Analysis) and graduate (Forte 1973, Rahn 1980, et al.)
Harvard University Integrated (Mathematical
Models for Music Theory) Graduate Own materials
(Lewin 1987, Lewin 1993)
University of Iowa Integrated (20th-Century Undergraduate Rahn 1980
Techniques) and graduate (Morris 1991, Lewin 1993)
Separate courses
University of Michigan, Ann Arbor Integrated (Basic Musicianship: Undergraduate, Own materials
Writing Skills; Analysis graduate (Morris 1991)
20th-Century Music
Separate course
10/2/2008 10:30:25 PM
Table 1.1. Pitch-class set theory in the curricula of twenty teaching institutions in the United States and Canada (1996)—(cont’d)
University of New Mexico Integrated (Composition I), Undergraduate Straus 1990, Morris 1991
Separate courses and graduate
Oberlin Conservatory Integrated (Music Theory IV ) Undergraduate Lester 1989
Pima Community College, Tucson, AZ Integrated (Theory and Structure Undergraduate Own materials, Straus 1990
of Modern Music)
University of Rochester Integrated (Various Courses) Undergraduate Straus 1990, Morris 1991
Eastman School of Music Separate Courses and graduate (Forte 1973, Rahn 1980)
Simpson College, Indianola, IA Integrated (Basic Theory) Undergraduate Kostka and Payne 1984
University of Texas at Austin Integrated Undergradute Rahn 1980, Straus 1990
Separate course and graduate
University of Virginia, Charlottesville Integrated (Materials of Undergraduate Straus 1990
Contemporary Music)
University of Washington, Seattle Integrated (Post-Tonal Theory Undergraduate Forte 1973, Rahn 1980,
and Analysis, Analytical and graduate Straus 1990
Techniques 1900-1950)
University of Wisconsin, Madison Integrated (Music Theory IV, Undergraduate Forte 1973, Morris 1987
Literature Survey of and graduate
20th-Century Music)
Separate course
University of Alberta, Integrated (Music Analysis) Undergraduate Forte 1973, Rahn 1980,
Edmonton, Canada and graduate Lewin 1987, own materials
10/2/2008 10:30:25 PM
Table 1.1. Pitch-class set theory in the curricula of twenty teaching institutions in the United States and Canada (1996)—(cont’d)
Dalhousie University, Halifax, Integrated (20th-Century Form Undergraduate Straus 1990
Nova Scotia, Canada and Analysis)
Memorial University of Integrated (Materials and Undergraduate Kostka and Payne 1984
Newfoundland, St. John’s, Canada Techniques of Music IV, (Forte 1973, Rahn 1980,
20th-Century Harmony, Straus 1990)
Theory/Composition Seminar)
Note: In the second column, one can see whether pitch-class set theory is the topic of a separate course, or is addressed in a more general context. In
the latter case, I have included the title or subject of the course in brackets. The third column specifies the academic level(s) on which the theory is
taught. The fourth column shows which texts students read. Here, the brackets indicate that a text is used for reference only, or is kept in reserve.
10/2/2008 10:30:26 PM
Newfoundland, at Oberlin Conservatory, and at Simpson College.17 In some

cases, it was a recent addition to the program (Bradley University and Simpson
College, 1993; University of Virginia, 1992; University of New Mexico, 1991).
From table 1.1 it appears that several institutions offered separate courses on
pitch-class set theory. Apart from the course at Bradley University—an under-
graduate course entitled “Theories of Atonal Music” but entirely devoted to the
theory of pitch-class sets—it concerned specialized graduate courses. A more
important indicator of the theory’s acceptance, however, was its appearance
in general courses, especially those intended for students at the undergraduate
level. This signified its inclusion in what was considered “basic knowledge” of
music. In this connection, we should take special note of its place in the music
theory core curriculum at universities like Michigan and Wisconsin. And the two
weeks spent on pitch-class set theory at Oberlin Conservatory may be a better
illustration of its success than its strong presence in the curriculum of the East-
man School of Music, an institution with a sixty-year tradition in the field of
advanced music theory.
Although it has achieved a firm position as an analytical “paradigm” in the
literature on twentieth-century music, and has found its way to the classroom,
the theory of pitch-class sets has seldom been short of criticism. Some of it has
already been mentioned. The theory has been criticized for its systematization
of analytical practice. It has been criticized for the way in which it has con-
structed “autonomous” musical objects, for failing to consider the historical
context of musical works of art, and for placing undue emphasis on their tech-
nical properties. But pitch-class set theory was not the only target of this criti-
cism; music theory itself, as an independent academic discipline, was under
attack, especially from those concerned with musical expression and imme-
diacy, or from those who posited music as social discourse and sought to study
it from a plurality of perspectives. I am referring to the members of the “New
Musicology” movement of the 1980s and 1990s. The names most commonly
associated with this movement are Lawrence Kramer, Susan McClary, and Gary
Tomlinson. However, its progenitor was the earlier-mentioned Joseph Kerman,
who, in Contemplating Music, had taken a stance against the narrow scope of
mainstream music scholarship—against its unreflecting ideology, its uniform
methodology, and its exclusive commitment to Western music in the high-art
17. The replies from these institutions to my inquiry (November 1996) included
remarks to this effect, but it can also be inferred from the fourth column in table 1.1.
At Simpson College and Memorial University, Stefan Kostka’s and Dorothy Payne’s Tonal
Harmony was used as a course text; more specifically the last chapter of this book, “An
Introduction to Twentieth-Century Practices,” only a few pages of which deal with “set
theory.” Students at Oberlin Conservatory read Joel Lester’s Analytic Approaches to Twenti-
eth-Century Music (1989), an elementary text discussing other topics as well. In San Diego
they did not work with a prescribed course text.

tradition. In this regard, Kerman made no distinction between theory-based

analysis and historical musicology:
Both were well calculated to thrive in the intellectual atmosphere of neoposi-

tivism. The appeal of systematic analysis was that it provided for a positivistic
approach to art, for a criticism that could draw on precisely defined, seemingly
objective operations and shun subjective criteria. (Kerman 1985, 73)
Whatever was new about “New Musicology,” it wasn’t the opposition to music
theory and analysis as an area of specialization. Such opposition is the natural con-
sequence of people claiming for themselves what others consider to be an integral
part of their own discipline. The musicologist Richard Taruskin, someone not typ-
ically associated with the movement,18 was equally averse to an Alleingang of music
theory, as appears from his polemic with Allen Forte about the latter’s analysis of
The Rite of Spring (Forte 1978, 1985, 1986; Taruskin 1979, 1986). His remarks will
concern us later (see chapter 7). And no spokesman for the New Musicologists
has been more succinct than the composer-theorist George Perle, who character-
ized pitch-class set theory as “martian musicology” (Perle 1990).19
Pitch-class set theory also stands condemned for its failure to explain how
music makes sense aurally. We often think that analyses of music should some-
how reflect the way in which we hear it, or at least could learn to hear it. This is a
concern of the music theory teacher, who helps students develop their hearing
skills. But it is also a concern of those looking for a basis of scientific verification
of analytical theories. A theory that tells us how we hear music can, in principle,
be tested (that is, if we come to an agreement about who “we” are); a theory
that tells us how it has been conceived cannot. Now, for a listener-based theory of
music to be potentially testable, it should not merely produce interpretations of
scores, but should also address the process through which such interpretations
come into being.
The composer and researcher Fred Lerdahl has pointed out why pitch-class
set theory falls short in this regard. For one thing, “it provides no criteria for seg-
menting the musical surface into sets” (Lerdahl 1989, 66). Indeed, without such
criteria there is no way of knowing with certainty which tones form a meaningful
combination. We might only refer to incidental corroborating evidence, such as
the resemblance between the opening of Schoenberg’s Opus 23, no. 3 and the
exposition of a fugue. From this resemblance it follows that the combinations
of example 1.1a stand out as important segments of structure. But how have we
managed to identify the non-obvious combinations of example 1.1b and 1.1d as
18. Yet, Kerman included his work in an appraisal of some “radical” trends in Ameri-
can musicology around 1990, together with that of Susan McClary, Gary Tomlinson, and
Carolyn Abbate. (Kerman 1991)
19. This expression had come to him through a remark of Richard Taruskin.

equally important segments? As said before, this involves a transformation, a

reshuffling of the pitch material so as to match it with that of the fugal sub-
ject—the “melody”—and its answers. Why do we, as analysts, go to such lengths
to show the importance of these non-obvious tone combinations? And in what
sense are they important? The same questions apply to our analysis of Stravin-
sky’s “Pas de deux.” It is one thing to note and describe the motifs played by
the violins in measures 412–13; to divide the accompanying sonic field into two
segments that can be reduced to the same abstract forms—as in example 1.2—is
quite another. Lerdahl would argue that a listener does not do this; the “hidden
substance” of these segments reveals itself to the analyst alone. The latter’s work
is met with suspicion:
Practitioners [i.e., analysts] have in effect relied on two external criteria for
set segmentation: its “musicality,” and its capacity to provide what the theory
denotes as significant set relationships. The first criterion is unexplicated, and
the second is self-reinforcing. (ibid.)
For this reason, Lerdahl wanted to describe music in terms of a “grammar” that
the listener attributes to it. Such a grammar, he thought, was not likely to evolve
from the abstractions of pitch-class set theory.
The principal milestone in the history of pitch-class set theory was the publi-
cation, in 1973, of Allen Forte’s book The Structure of Atonal Music. The segmen-
tation problem was immediately noted by William Benjamin in his review of it
for Perspectives of New Music (Benjamin 1974), and it has remained a potentially
fatal issue ever since.20 It loomed all the larger in view of the scientific spirit with
which the theory was suffused. This incongruity must have been deeply worri-
some to Lerdahl. The lack of consistent rules for segmentation was not the only
thing that bothered him, however. He also questioned the analytical relevance
of the concept of a pitch-class set, and of the concepts for relations between pitch-
class sets. In his view, these concepts did not account for the way in which he
believed music (whether it be tonal or atonal) was heard. For example, Lerdahl
argued that a pitch-class set does not appear to the listener as such—that is, as a
whole to which each member contributes in equal measure. “In a real context,
20. Benjamin’s criticism of Forte’s approach was diametrically opposed to Lerdahl’s.

Whereas Lerdahl wanted clear criteria for segmentation in pitch-class set theory, Benja-
min found Forte’s directions too restrictive: “Forte seems to regard pitches as associated
only if they sound simultaneously, form non-overlapping, uninterrupted successions, or
are otherwise in close proximity to one another. Notably absent is any basis for the associa-
tion of [pitch-classes] over longer time-spans, by means of registral distribution, similarity
of articulation, or the simple fact that each is somehow emphasized in its own context.”
(Benjamin 1974, 178–79). Yet, both Benjamin and Lerdahl might be right, the former
criticizing Forte’s own segmentations, and the latter the informal way in which he has
presented this topic in his book.

some pitches are heard as more or less structural than other pitches. Adjacent
pitches and chords form relationships that tug and pull at one another” (ibid.).
And he could cite a number of publications reporting experiments from which
it had appeared that set-theoretical concepts do not inform our hearing, or only
to a limited extent.
Lerdahl’s proposal for a listener-based theory of atonal music—a theory along
the lines developed in A Generative Theory of Tonal Music (Lerdahl and Jackend-
off 1983)—will not concern us here. Rather, we should ask ourselves whether
it is fair to want an analytical theory to be based on the musical intuitions of a
listener. Should it pass a “reality check”? Should examples from manuals and
textbooks of music analysis—or the observations of music students, for that mat-
ter—be rejected or modified when they fail such a check? Quite apart from the
question of which form an inquiry into the empirical groundings of the the-
ory should take—a very complex issue21—it might simply not be its purpose to
match a verifiable reality.
What, then, is an analytical theory of music? How does such a theory come
into being, and how does it function? To which needs does it respond, and what
kind of hold does it have on our musical imagination? In this study, these ques-
tions will be addressed regarding pitch-class set theory. If there is one thing that
has raised these questions, it is that, in spite of serious criticism, pitch-class set
theory has left such a big imprint on music scholarship and music teaching in
the United States and beyond. Perhaps paradoxically, we can add the criticism as
another measure of its institutionalization. This criticism betrays a deep engage-
ment with the issues that pitch-class set theory has addressed, opening up new
avenues of investigation that would have been unthinkable without it.
Aim, Scope, and Structure of this Study
The questions raised above invite a historical and contextual account of pitch-
class set theory. Such an account will be provided in this study, which can
thus be seen as concerned with the history of music theory. Like the seminal
works of this orientation—the narratives of François-Joseph Fétis (Esquisse de
l’histoire de l’harmonie, 1840) and Hugo Riemann (Geschichte der Musiktheorie im
IX.–XIX. Jahrhundert, 1898)—it looks at the history of music theory from the
vantage point of a contemporary theory. However, it does not see the contem-
porary theory as the summit or logical end point of that history. Pitch-class
set theory does not play the role of Fétis’s tonalité or Riemann’s theory of har-
monic functions. It is true that a comparison of the pitch-class set with older
musical concepts, such as chords, scales, motifs, or twelve-tone series, reveals
21. See for example Nicholas Cook’s summary of the critical reception of Lerdahl’s
and Jackendoff’s A Generative Theory of Tonal Music ( Cook 1989b, 118–20).

certain similarities; music theory builds on its own legacy. For this reason,
the present study describes the evolution of pitch-class set theory with occa-
sional reference to sources from a past more remote than the beginnings of
twelve-tone serialism. However, new musical concepts—or modifications of
traditional ones—do not only result from a self-generating theoretical dis-
course on music. And similarities between past and present concepts of musi-
cal structure do not always signify lineal relationships. Such relationships can
only be ascertained by a careful study of the use of these concepts. Apart
from a conceptual history, then, this study offers a view of pitch-class set the-
ory as a construction of its own time; a domain of musical competence that
reflects contemporary concerns, interests, and perceptions.
Earlier, I called pitch-class set theory a theory “for atonal music.” This does
not mean the same as “atonal theory.” A more comprehensive body of theory,
atonal theory includes, for example, the early attempts to establish a composi-
tional method on the basis of properties attributed to the equal-tempered chro-
matic scale, by composers such as Josef Matthias Hauer (Vom Melos zur Pauke,
1925, and Zwölftontechnik, 1926) and Herbert Eimert (Atonale Musiklehre, 1924).
It also includes the compendia of the harmonic and melodic resources con-
tained in the twelve-tone universe (from the writings of the nineteenth-century
French music theorist Anatole Loquin to Howard Hanson’s Harmonic Materials
of Modern Music, 1960). And it includes the manuals and theories of twelve-tone
serialism (e.g., Ernst Krenek’s Studies in Counterpoint, 1940, Eimert’s Lehrbuch der
Zwölftontechnik, 1950, and Babbitt’s articles on combinatoriality).
What all these theoretical works have in common is the idea of a tonal
equilibrium that allows any combination of tones to be formed in both the
horizontal and vertical dimensions. Any rule imposed on the combination of
tones is contextual—that is, it pertains to a single work.22 (The most obvious
example is a twelve-tone series serving as the referential structure for only
one composition.) However, work-specific rules can be subsumed under gen-
eral principles of organization (such as the principle of serial organization).
The more elaborate atonal theories deal with such principles. This is a very
broad delineation of the scope of atonal theory; it even allows us to analyze
tonal music from an atonal perspective. In such a case, the tonal equilibrium
is not what the music achieves, but what the theory takes as its starting point.
David Lewin’s theory of transformations can be seen as a late outgrowth of
atonal theory that has taken a portion of the tonal repertoire under its wings,
especially music of the late nineteenth century (Generalized Musical Intervals
and Transformations, 1987).
22. This use of the word “contextual” was introduced by Milton Babbitt: “Contex-
tuality . . . has to do with the extent to which a piece defines its materials within itself”
(Babbitt 1987, 167). Its antonym is “communal,” a word referring to materials that many
musical works share, such as the triadic structures and progressions of tonal music.

How should we distinguish pitch-class set theory from other varieties of atonal
theory? This is not immediately obvious. First, as noted before, the concept of
a “set” was already in use for the analysis of serial music. Second, pitch-class set
theory was never introduced under this name. In 1964, in an article for Journal
of Music Theory, Allen Forte presented what he called a “theory of set-complexes
for music.” This theory was not designed for the mere description of the struc-
ture of atonal music in terms of pitch-class sets and their various associations,
even though a considerable part of the article dealt with just that topic. It was
the statement of an organizational principle connecting various, if not all, major
pitch-entities in an atonal work: the “[pitch-class] set complex.” This is a special
case of a “family” of pitch-class sets, something in which Forte took a particu-
lar interest. In The Structure of Atonal Music, he developed the idea further, con-
vinced of its significance as a model for the pitch organization of movements or
entire compositions. And in 1988 he advanced an alternative type of family: the
pitch-class set “genus.”
Almost one half of The Structure of Atonal Music deals with the pitch-class set
complex. However, this has not proven the most durable part of this otherwise
very influential book. In 1997, the composer and theorist Robert Morris, an
advocate of the idea, noted that it had fallen, “if not by the wayside, at least
in frequency of use” (Morris 1997, 275). The pitch-class set complex did not
turn up in two successful, pedagogically inspired textbooks of pitch-class set
theory: John Rahn’s Basic Atonal Theory (1980) and Joseph Straus’s Introduction
to Post-Tonal Theory (1990), both of which are prominent in the fourth column
of table 1.1.
For the present study it is important to note that The Structure of Atonal Music
only represents a phase in the development of pitch-class set theory. Not every
part of the theory is contained in Forte’s book; nor has every chapter of this
book been of lasting influence. Pitch-class set theory is represented by a litera-
ture stretching from 1945 to the present day, with the years between 1960 and
1990 forming a period of crystallization and consolidation.
The focus of chapters 2 through 6 will be on pitch-class set theory’s conceptu-
alization of musical structure. These chapters trace the path that led to the defi-
nition of musical elements, sets, operations, and relations. Notwithstanding its
mathematical vocabulary, pitch-class set theory should be treated as the product
of, and the basis for, a music-theoretical discourse. It is not a mathematician’s
theory of post-tonal music, but an invention of composers and music theorists.
Therefore, it is appropriate for someone undertaking an inquiry into the rise
and development of pitch-class set theory to determine how its conceptual appa-
ratus relates to music history and the history of music theory; to ask, first, what
it has adopted from older theory and what it has added to it, and, second, which
cues it has taken from the repertoire and which constraints it has imposed on
its interpretation. It goes without saying that such an inquiry should also include
influences from outside the realm of music, such as from mathematics; and

should critically address the interaction of mathematical and musical concepts.

These principles guide the discussion in chapters 2 through 6.
Having placed pitch-class set theory historically this way, we are left with the
question of purpose and pragmatics. Chapter 7 (“Blurring the Boundaries: Anal-
ysis, Performance, History”) broadens the horizons to discuss music analysis, not
as the cool, taxonomic study of musical artworks, but as an activity representing
music in the here and now, like performance; an activity driven by its own con-
temporary agenda, which it finds mirrored in the music around which it revolves.
As a theory for analytical practice, pitch-class set theory came about because it
provided answers to questions of its time. What were these questions, and where
did they come from? Chapter 8 (“Mise-en-Scène”) relates the theory to some
crucial aspects and changes of its environment: to the impact of the computer
on the study of music, and to the American university in its double role as pro-
tector of high culture and provider of mass education. And it acknowledges that,
ultimately, individuals infused it with the sense of urgency that made it thrive.

Chapter Two
Objects and Entities

Pitch-class set theory is a musical application of mathematical set theory. As with
the latter, it defines a set as a collection of things, and it refers to these things as
the elements of the set. Sets are usually denoted by capital letters, like A, B, C,
etc. If not specified, the elements are denoted by lower-case letters: a, b, c, etc. In
pitch-class set theory, too, a set is defined by its elements only, not by a particular
arrangement of these elements or by their quantities. When two sets A and B
consist of the same elements, they are considered equal (A = B). What is special
about pitch-class set theory is the type of element it deals with: the pitch class.
Therefore, I will start this discussion of the theory with definitions of the terms
“pitch” and “pitch class.”
Pitch and Pitch Class
Pitch is one of the four distinguishing attributes of a musical tone (the others
being loudness, duration, and timbre). Broadly speaking, pitch is the position
of a tone on a spectrum that runs from low to high. In pitch-class set theory,
the term “pitch” designates a particular value assigned to that position. Thus, it
refers to what is often expressed by a letter name (C, D, E, etc.) together with an
indication of the octave range. Pitch-class set theory does not usually apply these
letter names. Under the postulate of equal temperament—the division of the
octave into (twelve) equal parts—it associates pitches with integers. It arbitrarily
assigns the number 0 to C4 (middle C), so that, by the equal distances between
all successive pitches, C♯4 = 1, B3 = −1, D4 = 2, B♭3 = 2, E♭4 = 3, and so on (ex.
2.1). Thus, each pitch can be associated with a pitch number, which is a positive
or negative integer. Of course, it is possible to assign the number 0 to a pitch
other than C4. This may help to clarify specific musical contexts. However, unless
otherwise stated, the pitch number 0 will represent C4 in this study. Enharmonic
notes (e.g., B♭ and A♯, or E and F♭) are given the same pitch numbers.
Another postulate of pitch-class set theory is that of octave equivalence.
Under this postulate, each pitch has the same value as the twelfth pitch above
and the twelfth pitch below it. The relation between two such pitches can be
expressed as the congruence modulo 12 of their pitch numbers. In other words:
these pitch numbers yield the same remainder when divided by twelve.
For a relation in a set S to be an equivalence relation, it has to satisfy three
conditions: it has to be reflexive (that is, for every element a in S, a is related to

30 objects and entities
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
-14 -13 -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0
Example 2.1. Pitch numbers.
0 1 2 3 4 5 6 7 8 9 10 11
Example 2.2. Conversion of pitch numbers to pitch-class (PC) numbers. Pitches

belonging to the same PC are arranged in columns.
itself), symmetrical (if a is related to b, then b is related to a), and transitive (if a
is related to b and b is related to c, then a is related to c). This appears to be the
case when S is the set of all pitch numbers and the open sentence “is related to”
is replaced by “is congruent modulo 12 with.” The relation is reflexive, for each
pitch number is congruent modulo 12 with itself; it is symmetrical, for the asser-
tion that pitch number a is congruent modulo 12 with b can always be reversed;
and it is transitive, for if a is congruent modulo 12 with b, and b, in turn, is con-
gruent modulo 12 with a third pitch number, c, then the same holds true for a
and c. It follows from this that an octave relation is an equivalence relation (see
also chapter 4).
The relation “is congruent modulo 12 with” partitions the set of all pitches
into twelve equivalence classes determined by their common remainder mod-
ulo 12 of pitch number. Such an equivalence class is called a pitch class (PC;
abbreviation of the plural: PCs). A pitch with pitch number a represents a
pitch class consisting of all pitches equivalent to a, and is denoted “a mod 12.”
In the following, this expression will be taken to mean the assignment to a
of its remainder modulo 12. For example, E♭4 is denoted 3 mod 12 = 3, E♭5
is denoted 15 mod 12 = 3, and E♭2 is denoted −21 mod 12 = 3. Example 2.2
shows the same pitches as example 2.1, but now the pitch numbers are con-
verted to PC numbers.
The notion of octave equivalence has a long history. It already expressed
itself in the equal letter names that medieval music theorists like Notker
Labeo and Guido of Arezzo assigned to pitches separated by an octave. Much
later, it enabled Baroque theorists like Johannes Lippius, Henricus Barypho-
nus, and Jean-Philippe Rameau to establish the concept of the invertible

objects and entities 31
triad.1 Today, we often refer to the concept without even realizing it. Con-
sider, as an example, this statement: “B is a chord tone in the II chord of D
Major.” No particular B is indicated; any one could function as a member
of the II chord in D major on the basis of octave equivalence. As far as the
octave relation is concerned, the integer model reinforces a traditional con-
ceptualization of pitch space.
PC set theory has derived the postulate of octave equivalence from the practice
of serial composition. The earliest reports of this practice stated that the octave
displacement of one or more tones did not affect the identity of a twelve-tone
series (see Greissle 1925, 64, and Stein 1925, 66). In Milton Babbitt’s wording,
the elements of a twelve-tone series were “considered independent of register
and durational values, and thus treated as equivalent if n octaves apart, where n
is any integer.” (Babbitt 1992, 1). This meant that analysts could describe music
without regard to its melodic contour. Serial composition thus went hand in
hand with an abstract notion of the material of a composition: a common “sub-
stance” wholly or partly underlying the most diverse musical statements. This
notion would become an essential—though by no means distinguishing—ingre-
dient of PC set theory.
Numerical pitch notations are distasteful to some. In a polemical article on
PC set theory, the Dutch composer Peter Schat wrote:
For the time being the notes will keep the names that thirty generations of
composers and theorists have managed with. He who wants to deprive some-
body of his name and his past should give him a number.2
An emotional outburst like this is unwarranted. Notes are not deprived of their
past by numbering them instead of using the traditional letter names. The his-
tory of Western pitch notation and pitch nomenclature encompasses numerical
systems as well. Aside from the numerical tablature notations (which indicated
not pitches, but fingerings for lute or keyboard) and the figured-bass notation
(which indicated chords to be played over a given bass line), several propos-
als for numerical pitch notations seem to have circulated in the past, especially
during the seventeenth century. One of these, a notation devised by William
Braythwaite in 1619, is still extant in print. It employs the numbers 1, 2, 3, . . .7;
these represent the notes of the diatonic scale, with 1 = do, 2 = re, 3 = mi and
so on (Gerads 1997/1998). A much more widespread numerical pitch notation
was the one used for the Rousseau-Galin-Paris-Chevé method of teaching sight
1. A discussion of this topic requires more thoroughness than can be provided on

these pages. I entrust the reader to Dahlhaus 1968 and Rivera 1984.
2. “Voorlopig houden de noten nog de namen waar dertig generaties componisten
en theoretici het mee gedaan hebben. Wie iemand van zijn naam, van zijn geschiedenis
wil beroven, moet hem een nummer geven.” (Schat 1998, 42, my translation)

singing. Like Braythwaite’s notation it was based on the diatonic scale, its com-
pass ranging from 1 (do) to 7 (si).
It should be noted that all the numerical notation systems just mentioned,
including the tablatures and the figured bass, served first and foremost practical
purposes. They were intended for musical training and performance. The pitch
notations were relative, not absolute. This means that the numbers were associ-
ated with intervals, and could be applied to different pitches (see below, “Pitch
Interval and Pitch-Interval Class”).
In the field of compositional theory and music analysis, the use of numbers
is not without historical precedents either. Words like “fifth” (Quinte), “sixth”
(Sexte), “second” (Sekunde) and “third” (Terz), as we encounter them in Hein-
rich Christoph Koch’s Versuch einer Anleitung zur Komposition (1782/1793), are
numerals indicating scale degrees. In the fragment below, Koch describes how
these scale degrees might assume the function of a new tonic in the course of a
symphonic Allegro:
The first and most usual construction of the first period of the second section
begins in the key of the fifth with the theme, occasionally also with another
main melodic idea, either note for note, in inversion, or also with other more
or less considerable alterations. After that it . . . modulates back into the main
key by means of another melodic idea, and from this [to the minor key of the
sixth, or otherwise] to the minor key of the second or third.3
By numbering the keys according to their positions in a scale (Tonart der Quinte,
Tonart der Sexte), Koch drew a hierarchical distinction between them. In that
respect he was a forerunner of Heinrich Schenker, who carried the tonal hier-
archy so far as to deny the existence of other keys than the main key at all. In
Schenker’s analytical graphs scale degrees are indicated not verbally but by cara-
ted Arabic numerals (see ex. 1.5a). Probably the best-known instance of the use
of numbers in music analysis is the practice of assigning Roman numerals to
tonal harmonies, a practice dating from the early nineteenth century and still an
important element of analysis courses today (Beach 1974). Again, it is a number-
ing of scale degrees. This means that in music analysis and compositional theory,
too, as well as in the aforementioned music notation systems, numbers are used
to refer to intervals rather than pitches.
3. “Die erste und gewöhnlichste Bauart dieses ersten Perioden des zweyten Theils
bestehet darinne, das er mit dem Thema, zuweilen auch mit einem andern melodischen
Haupttheile, und zwar entweder von Note zu Note, oder in verkehrter Bewegung, oder
auch mit andern mehr oder minder beträchtlichen Abänderungen in der Tonart der
Quinte angefangen wird, nach welchem . . . vermittelst eines andern melodischen Theils
die Modulation zurück in den Hauptton geführt, und von diesem in die weiche Tonart
der Sexte, oder auch in die weiche Tonart der Secunde oder Terz geleitet wird.” (Koch
1782/1793, IV, 307–8, transl. by Nancy Kovaleff Baker)

The practice of measuring an interval in half steps rather than in diatonic

steps ensued from the rise of what Jonathan Bernard has called “encyclopedic
music theory”: the project of mapping out all the pitch or interval combina-
tions that are possible within the twelve-tone universe, a project on which vari-
ous composers and music theorists embarked during the second half of the
nineteenth century and the first half of the twentieth (Bernard 1997). This
project, in turn, was encouraged by the relaxing of tonal strictures and the
enlargement of the harmonic palette in the music of that period, and by the
acceptance of the equal-tempered chromatic scale as a basis for theoretical
speculation. For the identification and tabulation of the vast amount of pos-
sible chords and scales a numerical notation of pitches and intervals using the
semitone (the “1”) as a unit of measure was most appropriate. Early examples
of this notation include the works of two nineteenth-century music theorists,
the Austrian Heinrich Vincent and the Frenchman Anatole Loquin.4 Later it
was applied by authors such as Ernst Bacon (“Our Musical Idiom,” 1917) and
Joseph Schillinger (Kaleidophone: New Resources of Melody and Harmony, 1940; The
Schillinger System of Musical Composition, 1946).5 Ernst Krenek recommended its
use for intervals in his Studies in Counterpoint (1940)—the first real manual of
twelve-tone serialism—though strangely enough without acting upon that rec-
ommendation himself.6
Babbitt’s integer notation for twelve-tone series—which he introduced in his
1946 thesis The Function of Set Structure in the Twelve-Tone System (published in 1992),
and in the more widely read articles that he published in subsequent years (Bab-
bitt 1955, 1960 and 1961a)—employed numbers from 0 through 11. Each num-
ber assumed the role of a “pitch number,” first called “set number,” which actually
referred to a PC (Babbitt’s definitions invoke modulo 12 arithmetic), and that of
an “order number.” Of a given twelve-tone series, the first tone received the “order
number” and “pitch number” 0, and was referred to as “0,0”; “3,7” indicated the
element with “order number” 3 (i.e., the fourth element of the series!), and
“pitch number” 7. With respect to pitch, Babbitt’s notation was relative: “[the]
initial assignment of any one of the integers to any one particular scale element
is purely arbitrary, but once having decided upon, this allocation must remain
fixed throughout the particular discussion” (Babbitt 1992, 1–2). It was devised
for the sake of transparency. Each of the basic twelve-tone operations can be
4. Vincent’s writings were first discussed by Robert Wason in a study of progressive

harmonic theory in the nineteenth century (Wason 1988). Catherine Nolan places the
works of both authors in the context of the rise of a combinatorial paradigm for the teach-
ing of harmony (Nolan 2003).
5. For accounts of the work of Bacon, see Neff (1991), Bernard (1997), and Nolan
(2003). Bernard also discusses Schillinger’s treatises.
6. Krenek either maintained the traditional interval names (put between quotation
marks, like “major third”) or he just mentioned the letter names of the notes involved.

expressed as an arithmetical operation, involving either “pitch numbers” (trans-

position, inversion) or “order numbers” (retrogradation), or both (retrograde-
inversion; see chapter 3).
How should we view Babbitt’s use of numbers? We might put it on a par with
the numerical music notations mentioned above, or with the analytical nota-
tions of someone like Schenker. However, it exceeded the merely practical, as
will become clear from the chapters that follow. Babbitt took musical relations as
manifesting principles of number, which is reminiscent of the medieval specula-
tive theory of intervals. Numerical relations embodied an immutable logic that
was very appealing to him. They served him, and other composers and theorists
of the twentieth century, as a model of coherence. Babbitt’s twelve-tone theory
was a renewal of the contract between music theory and number theory.
Babbitt already prefigured the conversion from a relative to an absolute
numerical pitch notation, with “0” representing C:
It may be found to be helpful to always associate a given integer and a certain

pitch name (0 for c, for instance) and maintain this correspondence through-
out. This device must, however, be looked upon purely as an associational sim-
plification, and in no sense as either a notational necessity, or as having any
effect on the generality of the discussion. (Babbitt 1992, 2)
Allen Forte actually took this step “for convenient reference,” as he wrote in
“A Theory of Set-Complexes for Music” (1964, 139). In Forte’s primary field of
study—the non-serial atonal music of the second Viennese school—it was usu-
ally hard to determine a contextual “0.” In serial twelve-tone music, however,
one could take the first note of the initial series to serve as a referential element.
Practical or not, Forte’s decision meant a radical departure from the analytical
tradition, for it was not customary to always assign the same number to a par-
ticular pitch. It is this aspect that may have provoked Peter Schat’s grumbling
reaction. Fixed pitch numberings have been used most prominently in research
into tuning systems.7 Therefore, they may seem inartistic and unduly “scientific”
to many of those primarily interested in music as an art form.
However, for composers of a modernist orientation, “scientific” was not an
abusive term at all. Anton Webern said he could no longer see any difference
between science and inspired creation (Webern 1960, 10). Edgar Varèse saw sci-
ence and art collaborating “on the threshold of beauty” (Varèse 1967, 196). Karl-
heinz Stockhausen demanded a “consistency of the single part and the whole,”8
invoking not only an organicist poetics of music but also positivist criteria for
7. For example, the pitch numbering system known as “semitone count,” in which
the keys of equally-tempered keyboard instruments are numbered consecutively, e.g.,
from 1 to 88 (piano).
8. “Die Sinnhaftigkeit einer Ordnung gründet in der Widerspruchslosigkeit zwischen
Einzelnem und dem Ganzen.”

scientific theories (Stockhausen 1963, 18). And the self-proclaimed logical-posi-

tivist Milton Babbitt insisted on a “scientific language” and a “scientific method”
when music was discussed (Babbitt 1961b, 398). The boundary between art and
science was permeable in such music journals as Die Reihe or Perspectives of New
Music. The reason why will be addressed later (see chapter 8B).
To summarize, numerical systems for the identification of pitches have
arisen naturally in Western musical thought. The numbers can be movable
or fixed, but in either case pitches are designated in terms of their intervallic
relations. Though practical and appropriate in many cases, the use of inter-
val measures as pitch labels has been subject to serious criticism, especially
with respect to non-serial “atonal” music. This criticism was brought up by
somebody who was not, like Schat, averse to PC set theory, but on the con-
trary has made important contributions to its development: the music theo-
rist and composer David Lewin. Writing in 1977, Lewin judged the practice
of numerical PC labeling “theoretically suspect,” “analytically confusing,” and
“systematically unnecessary,” among other things (Lewin 1977a, 43). Princi-
pal among the problems he addressed was the underlying, if perhaps unin-
tended, assumption of centricity. In example 2.1, D♯4 is labeled “3” because
it is three steps in an upward direction from the referential C4 (“0”); B♭3 is
labeled “−2” because it is two steps in a downward direction from the same C.
But in “atonal” music there supposedly are no referential pitches or PCs. And
if for a span of time a pitch does assume a referential function, it need not be
a C (Forte), or represent the first PC of the most significant series-form (Bab-
bitt). As a consequence, the association of pitches with numbers may obscure
“the exact nature of the musical structure involved” (Lewin 1977a, 30). We
shall see a few instances in due course.
Pitch Interval and Pitch-Interval Class
In view of the preceding discussion, we can expect PC set theory to define inter-
vals in a way that accommodates pitch numbers. A pitch interval (PI) is the num-
ber of semitones upward or downward that separates one pitch from another
(ex. 2.3). Direction matters; the pitches involved form an “ordered pair.” One
says that the pitch interval from a to b is b − a, in formal notation: PI(a,b) = b − a.
The brackets indicate that a and b are considered in the given order. The pitch
interval from C4 (0) to D♯4 (3), for example, is 3:
PI(0,3) = 3 − 0 = 3
This is also the pitch interval from, say, F4 (5) to A♭4 (or G♯4) (8):
PI(5,8) = 8 − 5 = 3

The pitch interval from C4 to D♯5 (15) is 15, which is also the pitch interval
from G♯4 to B5 (23):
PI(8,23) = 23 − 8 = 15
The pitch interval from C4 to A3 is −3, and so is the pitch interval from F♯4 to
E♭4:
PI(6,3) = 3 − 6 = −3
Under the assumption of octave equivalence each pitch interval is equivalent to

its congruent modulo 12, as is the case with each pitch. The pitch interval from
C4 to D♯4 (three semitones upward) has the same value as the pitch interval
from C4 to D♯5 (fifteen semitones upward). And the pitch interval from C4 to
A3 (three semitones downward) has the same value as the pitch interval from C4
to A4 (nine semitones upward). From this it follows that an intervallic analogue
of the PC can be conceived: the pitch-interval class (PIC). Like the set of pitches,
the set of pitch intervals is partitioned into twelve such classes. Every pitch inter-
val PI(a,b) represents a class of all pitch intervals equivalent to PI(a,b), denoted
PI(a,b) mod 12. This expression signifies the assignment to PI(a,b) of its remain-
der mod 12. Thus, for the PIC represented by the pitch interval from G♯4 to B5,
we write 3:
PIC(8,23) = PI(8,23) mod 12 = (23 − 8) mod 12 = 15 mod 12 = 3
And for the PIC represented by the pitch interval from F♯4 to E♭4, we write 9:
PIC(6,3) = PI(6,3) mod 12 = (3 − 6) mod 12 = −3 mod 12 = 9
Example 2.4 demonstrates the conversion of pitch intervals to PICs. The termi-
nology concerning PICs varies in the literature. Milton Babbitt (1992, 6–7) and
David Lewin (1959, 298–99), who established the concept as I presented it, just
called them “intervals.” Others speak of “ordered PC intervals” (Rahn 1980a,
Straus 1990a), or “directed PC intervals” (Regener 1974).9 The term “interval
class,” however, usually functions as a concept reducing the number of interval-
lic entities even further than “PIC,” as we shall see further below.
The above discussion may seem needlessly elaborate. However, this elaborate-
ness is justified by a counter-intuitive aspect of PICs. From example 2.5 it appears
that a minor third and a major sixth may represent the same PIC, while on the
9. These are not just different terms for, but actually different conceptions of, the
same phenomenon. One is based on a classification of pitch intervals (PIC), the other on
a classification of pitches (pitch-class interval).

15
9
3
-3
-9
-15
Example 2.3. Pitch intervals.
9 3
Example 2.4. Conversion of pitch intervals to pitch-interval classes (PICs).
other hand two minor thirds (or two major sixths, for that matter) may repre-
sent two different PICs. This is the consequence of conceiving an interval as an
ordered pair of pitches, a conception suggested by twelve-tone serialism. If the
first two tones of a twelve-tone series are, say, an A and a C—and if octave equiva-
lence is assumed—it is possible that they form an ascending minor third or a
descending major sixth. However, in the same series they cannot possibly form
an ascending major sixth and a descending minor third. Therefore, the concept
of an “interval class” should involve order when it is used with reference to a
twelve-tone series.
The consequence of considering the order of pitches is that with each pair of
PCs two PICs are associated, as appears from example 2.6. The two PICs associ-
ated with the PCs 0 and 9, for example, are 3 and 9; the two PICs associated with
the PCs 7 and 5 are 2 and 10. In each case, both PICs are complementary with
respect to twelve (the octave).
Now, one might wish that there were only one intervallic category associated
with each pair of PCs; in other words, one category for minor thirds, minor
tenths, major sixths, and major thirteenths, irrespective of the order of the PCs
involved. This is particularly relevant for PC set theory, which deals with rela-
tions that are independent of order. Allen Forte’s term “interval class” satisfied
the need for such an intervallic category. This is not the “pitch-interval class”
defined above, but a closely related concept, assuming the equivalence of PICs
complementary with respect to twelve. Forte’s 1964 definition runs:
Let S be the set of PC integers [0, 1, 2, . . . 11], with ordinary addition and sub-
traction (mod 12), absolute value differences, symbolized |d|, and x, y, any two

3 3 9 9
Example 2.5. PICs are order-dependent.
elements of S. Further, let a = |x−y| and b = x + y’, where y’ = 12−y and is called
the inverse of y. Then,
a ≡ b if and only if10

a + b = 0 (mod 12)
This definition is numbered (2.0) in Forte’s article. Forte continues:
The letters a and b represent interval-classes, and the equivalence relation

defined by (2.0) partitions S into 7 such classes comprising the familiar inter-
vals of the chromatic scale. . . . By convention, the [smallest of a and b] rep-
resents the class in each case. Since the class containing 0 is trivial we usually
speak of 6 instead of seven interval-classes. (Forte 1964, 140)11
In The Structure of Atonal Music, written nine years later, Forte confines him-
self to announcing that “inverse-related (modulo 12) intervals are defined as
equivalent” (1973, 14). According to Forte’s system—a system that has been
widely followed—an interval class is a class of unordered pairs of pitches
denoted PI{a,b} mod 12, and is represented by the smallest of (a − b) mod 12 and
(b − a) mod 12. The curly braces indicate that a and b are unordered. Thus, the
pitch intervals from G♯4 to B5 and from F♯4 to E♭4—which represent the PICs
3 and 9 respectively—both represent the Fortean interval class 3. Example 2.7
shows the relation between PICs and the interval classes as conceived by Forte.
A single twelve-tone series is not adequately represented by a succession of
interval classes à la Forte. But a succession like that would express what the 48
forms of a series have in common. In an attempt to define that essence, Carl
Dahlhaus called a twelve-tone series “a succession of interval classes comprising
the twelve PCs of the chromatic scale, which constitutes the relational system of
10. In sentential calculus, a division of mathematical logic, one speaks of the equiva-
lence of two sentences when these are connected by the expression “if and only if,” abbre-
viated “iff.” When a and b are sentences, the logical phrase “a iff b” means “a if b, and b if
a.” It is imperative that both sentences be either true or false, and that the conversion of
the two sentences does not change this condition. Mathematicians use sentential equiva-
lence when they define a new concept. The definition takes the form “a iff b,” where a sig-
nifies an expression involving the new concept, and b an equivalent expression in already
familiar terms (Tarski 1995).
11. The typography of the original text has been maintained in this quotation. This
will be the policy for every quotation in this study.

PCs 0,0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 0,10 0,11
PICs 0 1 2 3 4 5 6 7 8 9 10 11
11 10 9 8 7 5 4 3 2 1
Example 2.6. PICs associated with PC pairs.
PICs
0 1 2 3 4 5
PICs 0 11 10 9 8 7
PICs 1 2 3 4 5 6
Forte: 0 1 2 3 4 5 6
Example 2.7. PICs and the interval classes according to Allen Forte.
a composition.”12 Dahlhaus’s use of the term “interval class” was in agreement

with Forte’s. It united, for example, major thirds and tenths and minor sixths in
both directions.13
Aside from all the advantages it offers, Forte’s definition of “interval class” is
problematic. It is questionable whether an inverse relation of PICs is really an
equivalence relation; in other words, whether it satisfies the three conditions of
reflexivity, symmetry, and transitivity. Forte himself had summed them up in his
article, in a note attached to the above definition (Forte 1964, 179, note 5). In
the set of PICs the relation “is the inverse modulo 12 of” is not reflexive (a PIC
is not, in general, its own inverse; only 0 and 6 are); nor is it transitive (when b
is the inverse of a, and c is the inverse of b, c equals a, and is not, in general, the
inverse of a, except when b = 0 or b = 6). Therefore, when inverse-related PICs
are not equal, they are not equivalent either. Their relation is different from that
between octave-related pitches or pitch intervals. Consequently, Forte’s interval
class is not a “class” in the same sense as PCs and PICs are.
Treatises and manuals of more recent years, like John Rahn’s Basic Atonal Theory,
Robert Morris’s Composition with Pitch-Classes, or Michael Friedmann’s Ear-Training for
Twentieth-Century Music do not speak of the equivalence of inverse-related intervals.
However, Rahn and Morris still use the term “interval class” in the Fortean sense, as
12. “Eine die zwölf Tonqualitäten der chromatischen Skala umfassende Anordnung
von Intervallklassen, die das Bezugssystem einer Komposition darstellt.” (Dahlhaus 1970,
507, my translation)
13. “Eine Intervallklasse . . . umfaßt z.B. außer der großen Terz (c-e und c-As) die
große Dezime (c-e’ und c’-As) und die kleine Sexte (c-E und c-as).” (Dahlhaus 1970, 506)

does Joseph Straus (Introduction to Post-Tonal Theory). Friedmann, however, prefers

the term “unordered pitch-class interval.” This is a problem of vocabulary, a prob-
lem PC set theory has not been able to solve completely. “PIC” has correctly been
defined above as signifying a class of intervals associated with ordered pairs of pitches.
The almost identical term “interval class” (IC) refers to a different concept, a class of
intervals associated with unordered pairs of pitches.
In their contribution for the entry “interval class” in The New Grove Diction-
ary of 1980, Paul Lansky and George Perle called the Fortean “interval class”
an “absolute interval class.” This term refers to the mathematical concept of
the absolute value, by which positive and negative numbers of the same size
are equated. Likewise, it is possible to equate inverse-related pitch-interval
numbers—i.e., numbers assigned to ascending and descending intervals of the
same size—and inverse-related PIC numbers. The absolute value of a minus b—
notated |a − b|—is:
a − b, when b is smaller than or equal to a,
or:
b − a, when b is larger than a.
This means that the absolute value of PIC(a,b) equals that of PIC(b,a). For
example, if a and b are the pitches G4 (7) and E4 (4) respectively, the absolute
value is 3 in both cases:
|(4 − 7) mod 12| = (7 − 4) mod 12 = 3
and:
|(7 − 4) mod 12| = (7 − 4) mod 12 = 3
It can thus be said that 3 is the absolute pitch-interval class (APIC) represented
by the pitch interval from G4 to E4 and the pitch interval from E4 to G4.
Although the idea has met with no response, the absolute value can substitute
for Forte’s somewhat problematic “interval class.”
Pitch-Class Set
Pitch-class sets (PC sets) are specified according to the conventions of mathe-
matical set theory. When a PC set consists of, for example, the PCs 1 (C♯), 2 (D),
5 (F), 6 (F♯), 7 (G), and 9 (A), its notation is:
{1,2,5,6,7,9}

This notation does not take the quantity of an individual element into account.
Moreover, the curly braces indicate that the elements are displayed in arbitrary
order. Therefore, {1,2,5,6,7,9} does not mean anything different from {7,2,9,5,1,6}.
The number of elements in this PC set is 6. This number is called its cardinal
number, and is often represented by the sign “#” (“#A” means: “the cardinal num-
ber of set A”). The musical idea displayed in example 2.8 (a figure from Arnold
Schoenberg’s Piano Piece, Op. 11, no. 1) is a “realization” of our PC set. From the
accompanying table it appears that the PCs involved are the same.
PC sets can represent any combination of tones—scales, motifs, chords, and
harmonic-melodic progressions, for example. They do so regardless of the regis-
tral and temporal distribution of the notes, and of their harmonic and melodic
functions. Tone combinations are represented, in other words, as mere collec-
tions. We say that they are equal if they realize the same PCs. Example 2.9 is an
example from Joseph Straus’s Introduction to Post-Tonal Theory, which shows the
reader “a single pitch-class set expressed in five different ways” (Straus 1990a,
27). The PC set is {1,4,5,7}; the excerpts have been taken from a serial work: the
“Gavotte” from Schoenberg’s Suite for Piano, Op. 25.
As a musical concept, the PC set originates from two traditions. The first (and
oldest) of these is chordal theory. The PC set owes much to the concept of a
“chord,” the members of which may appear in different octaves and exchange
vertical order positions. A chord may also unfold linearly. In his Treatise on Har-
mony (1722), Jean-Philippe Rameau explains in which ways a melody can be
ornamented above a bass: one way is by using “consonant tones”:
Notes between beats which proceed by consonant intervals should be notes

which are in the chord of the first beat. When the next beat is reached, notes of
its chord should be used, and so on from one beat to the next until the end.14
The bass itself can be ornamented in the same way. Thus, a succession of “bro-
ken” or “arpeggiated” chords appears. Compositional instructions like these,
involving the “prolongation” or “linear unfolding” of triads, can be traced until
the 1630s (Rivera 1984, 74). So, what chords and PC sets have in common is
a flexibility of arrangement. Some have actually referred to PC set theory as a
“theory of chords” (Regener 1974). However, traditional chordal theory does
make a qualitative distinction between the different members of a chord (the
root, third, and fifth of a triad) and between the various possible bass positions
in which it may appear (the root position and the first and second “inversions”
of a triad). It also makes a distinction between chord tones and non-chord tones
14. “Lorsque l’on veut faire passer des Nottes entre les Temps par des Intervales
consonans, on ne peut en faire entendre d’autres que celles qui sont comprises dans l’Ac-
cord du premier Temps, pour tomber ensuite sur une Notte de l’Accord du Temps qui vient
immediament aprés, & ainsi d’un Temps à l’autre, jusqu’a la fin.” (Rameau 1722, 308,
transl. by Philip Gossett)

accel.
Pitch: F4 D4 C 5 G3 A4 F 4 F5 D5 C 6
Pitch Nr. 5 2 13 -5 9 6 17 14 25
PC 5 2 1 7 9 6 5 2 1
Example 2.8. Arnold Schoenberg, Piano Piece, Op. 11, no. 1, mm. 49–50.
Used by permission of Belmont Music Publishers, Pacific Palisades, CA 90272.
a. b. 7
1
p
p ^ ^ ^
sfp
sf
d. 24
c. 16
p5
1 5
4 5 2
^
dolce 3 1
e. 26
rit
ff
f
ff
Example 2.9. An example from Joseph Straus’s Introduction to Post-Tonal Theory (1990).

(like passing tones, neighbor tones, and suspensions). Furthermore, an impor-

tant aspect of a traditional chord is that it operates within the system of a partic-
ular key; in other words, its function is always defined in relation to that of other
chords. Finally, chords have always been more or less subject to the constraints
of meter and rhythm. All this does not necessarily apply to the PC set.
By the mid-nineteenth century the harmonic idiom of Western art music had
grown more complex; chord tones and non-chord tones could no longer be
so easily distinguished and the system of chordal relations gradually dissolved.
Chordal theory could cope for a considerable time, but as the process continued
some music theorists began to stop defining chords by their function, or impos-
ing restrictions upon their structure, and classified them on the basis of their
objective pitch or interval content.
Many such classifications soon faded into oblivion, but in recent years inter-
est in them has increased. Scholars now see them as having contributed to the
“conceptual preparation of the pitch-class set.” The quoted phrase comes from
the pen of Catherine Nolan (2003, 206). She and Jonathan Bernard (1997)
have investigated several of these chord classifications.15 Bernard discusses the
systems of Ernst Bacon, Joseph Schillinger, and Howard Hanson, which I men-
tioned earlier; and he evaluates similar efforts by composers like Fritz Heinrich
Klein (“Die Grenze der Halbtonwelt,” 1925) and Alois Hába (Neue Harmonielehre
des diatonischen-, chromatischen-, Viertel-, Drittel-, Sechstel-, und Zwölftel-tonsystems,
1927). One of his observations is that, in the process of enumeration, some of
these authors let the traditional distinction between harmony and melody fade;
in other words, they mixed the identities of chord, scale, and melodic pattern.
Bernard concludes that they were drifting away from chordal theory in a strict
sense, and entering the domain of PC set theory.
Nolan traces the “prehistory” of the pitch-class set back even further, to the
mid-nineteenth century. However, she relates the earliest comprehensive chord
classifications—such as Loquin’s Tableau de tous les effets harmoniques de une à cinq
notes inclusivement (1874)—to a general interest in the application of combinato-
rial techniques that goes back to the seventeenth century (Nolan 2003, 238).
This, too, indicates that the pursuit of comprehensiveness does not go hand in
hand with a concern for harmonic identity.
“PC set” is neither a basically harmonic nor a basically melodic concept. It
simply eliminates the distinction between the horizontal and the vertical. In that
respect, the idea of the PC set was prepared not so much by the above-mentioned
chord classifications as by twelve-tone serialism. We already have noted the prox-
imity of PC set theory and the practice of serial composition (see chapter 1, p.
9). And we have mentioned the role of Babbitt as the serialist who pioneered
the set-theoretical approach. Twelve-tone serialism, then, is the second tradition
15. I will again refer to their work in the section of chapter 4 that deals with some of
these classifications (“A Little More History: Musical Statistics”).

from which the PC set has sprung. Arnold Schoenberg stated that “the two-or-
more-dimensional space in which musical ideas are presented is a unit.” This
statement was made as a prefatory remark to the composer’s exposition of his
“method of composing with twelve tones.” It referred to the twelve-tone series,
which could generate melodic lines as well as sonorities.16
As we have seen, the elements of a twelve-tone series are considered indepen-
dent from register and duration. In other words, a twelve-tone series consists of
PCs. It is therefore a progenitor of the PC set. (Schoenberg himself called it a
“twelve-tone set.”17) Needless to say, “twelve-tone series” is a far more restricted
concept than “PC set,” since its cardinal number is fixed and its elements are
ordered. But the fact that adjacent elements can also be projected harmonically
implies that its ordering need not be musically manifest. Quite paradoxically,
serial techniques have enabled the unordered PC set of less than twelve ele-
ments to stand out as an important analytical category. This is already evident
from example 2.9, where the recurring association of the PCs 1 (D♭), 4 (E), 5
(F) and 7 (G) results from their order positions in a twelve-tone series.
PC set theory claims that such sets have also been living a life of their own in
music, independently from serial practice. Twelve-tone serialism is even claimed
to be only a special case of a much more general structural principle, involving
unordered PC sets of different magnitudes. As John Rahn has put it in the intro-
duction to his Basic Atonal Theory, the first real manual of PC set theory:
The relations taught here are basic to all atonal music, whether that music is
serial or non-serial. Indeed, serialism arose partly as a means of organizing
more coherently the relations used in the preserial “free-atonal” music. There
are relations—those deriving from syntactical order—that are peculiar to serial
atonal music. The theory of these particularly serial relations is an extension of
the “basic atonal theory” given here; serial theory builds on basic atonal theory
as its foundation. (Rahn 1980a, 2)
Any doubt cast on the generality of that hypothetical structural principle has
not prevented PC set theory from considerably widening its scope in the past
decades. Initially, in the 1960s and early 1970s, its focus was mainly on the non-
serial atonal repertoire of the second Viennese school. Later, Stravinsky’s music
16. Schoenberg supplied this statement with the following explanation: “The mutual
relation of tones regulates the succession of intervals as well as their association into har-
monies; the rhythm regulates the succession of tones as well as the succession of harmo-
nies and organizes phrasing. And this explains why . . . a basic set [sic] of twelve tones . . .
can be used in either dimension, as a whole or in parts” (Schoenberg 1984, 220). Schoen-
berg’s method reflected concerns that were shared by people from outside the serial tradi-
tion as well. Joseph Matthias Hauer’s twelve-tone technique did not differentiate between
the horizontal and the vertical either (Hauer 1925, 1926).
17. Milton Babbitt claims to have suggested this term to Schoenberg (Duckworth
1999, 63–64).

came within its reach (Forte 1973 and 1978, Van den Toorn 1983), followed by
the works of Debussy (Parks 1981 and 1989) and Scriabin (Baker 1986). By now,
it covers a variegated repertoire, ranging from late-nineteenth-century music to
contemporary jazz. This raises questions about its relation to other theories of
musical structure. For example, how does it relate to tonal theory? John Rahn,
continuing on the above-quoted line of thought, suggested that tonal theory
could, in a sense, be regarded as a special case of atonal theory (Rahn 1980a,
19). Seen thus, PC set theory functions like an “umbrella” theory of music.
Indeed, the quest for such a theory—which connects a multitude of musical
idioms, including contemporary ones—has given much impetus to American
music theory, and to PC set theory in particular, from the mid-1950s onwards.
Now, to return to the claim that a PC set corresponds to a verifiable musical
entity in non-serial music, how can such entities be found? Generally speaking,
they should be recognizable as subjects of a musical argument: they should be
repeated or transformed, or otherwise articulated. PC set theory has provided
models for a considerable number of transformations and relations. This is terri-
tory to be explored in the following chapters, after we have finished this prelimi-
nary examination of the “set” concept.
It is possible to cast the entire system of twelve PCs in terms of finite set the-
ory, and to calculate the total amount of PC sets. This calculation calls in the
terms “subset,” “superset,” “universal set,” “empty set,” and “power set.” Here is
an overview:
• Set A is a subset of set B (A ⊆ B) when each element of A is also an element

of B. Reversely, B is a superset of A (B ⊇ A) when it contains the elements of
A. A and B might be equal (namely when, simultaneously, A ⊆ B and B ⊆
A). In PC set theory, it would seem more natural to consider proper subsets
and supersets only. Indeed, this is what some prefer. Usually, A is called a
proper subset of B (A ⊂ B) when it is not equal to B (A ≠ B). The reverse
statement is that B is a proper superset of A (B ⊃ A).
• The universal set is the fixed set of which all sets considered are subsets. Its
counterpart is a set without elements, the empty or null set. This is a subset
of each of the sets considered. In PC set theory, the universal set consists
of integers, which, as we have seen, represent pitches under twelve-tone
equal temperament. From now, I will call it PITCH, using capitals for
quick identification. However, I will more often refer to PITCHCLASS,
the partition of PITCH that was described in the first section of this chap-
ter. This set consists of the integer values from 0 to 11 inclusive, which
represent the twelve PCs.
• The entire collection of subsets of a set A—including the empty set as well
as the set equaling A—is the power set of A, or ℘(A). The number of ele-
ments in ℘(A) is 2n, where n is the cardinal number of A. ℘(PITCHCLASS)

is the set of PC sets, henceforth designated PCSET. It contains 212 = 4,096

PC sets. Table 2.1 shows a partition of this amount into the quantities per
cardinal number. The general expression of the sum total of distinct PC
sets of a given cardinal number n is:
12!
n!(12 – n)!
Usually, PC set theory excludes certain sets from consideration when it comes
to music analysis. Of course the null set meets this fate, but others do as well,
for reasons to be discussed later. Partitioning PCSET into equivalence classes
further reduces the number of distinct PC sets (see chapter 4).
The Interval Content of a PC Set
Although a PC set is defined by its elements, one can say that the intervals between
these elements are what really matters. Indeed, many of the more significant rela-
tions between PC sets boil down to similarities between their interval contents.
Since “PC set” is an abstract concept, encompassing a large number of possible
musical realizations, the interval content of a PC set should be defined so that it
represents any of these realizations: first, it should be defined in terms of PICs—
absolute or not—and second, it should involve all possible pairs of PCs.
In 1959, in a contribution to the young Journal of Music Theory, David Lewin coined
the term interval function, which indicated the frequency of a given PIC (which Lewin
called an “interval”) among the ordered pairs of members of two PC sets:
Let P and Q be any two collections of notes. We impose no a priori restrictions

on P and Q. P may contain 2 notes and Q, 7; or P may contain 4 notes and Q,
12, etc.; possibly even Q may be the same collection as P.
We can define what I shall call the interval function between P and Q as follows:
For every integer i between 0 and 11 inclusive, let m(i) be the number of pairs
of notes [PCs] (x,y) such that x is a member of the collection P, y is a member
of the collection Q, and the interval between x and y is i. The function m will
be called the interval function between P and Q. (Lewin 1959, 299)18
It should be noted that the pairs (x,y) are ordered, which means that the interval
function is always directed from one PC set to the other. This may seem incon-
venient when there is no reason to assume a particular order of the collections
18. At the time it was not yet common usage to speak of “sets” in music theory, and
the word “collection” was equally apt to refer to unordered combinations of tones.

Table 2.1. The sum total of PC sets of successive cardinal numbers
Cardinal nr. Number of PC sets

0 1
1 12
2 66
3 220
4 495
5 792
6 924
7 792
8 495
9 220
10 66
11 12
12 1
4,096
P and Q. But the inconvenience is automatically removed in case P and Q are

the same. The interval function is very useful in that case; it is an adequate
measure of the interval content of a PC set. Lewin thus defined “interval con-
tent” as the interval function m(i) for every i between a PC set and its dupli-
cate (Lewin 1960, 99). In the following, we will use the term PIC content when
we refer to this property.
Given a PC set of n elements, the number of possible ordered PC pairs is n2
(including the unisons) or n2–n (excluding the unisons). The PIC content is
a partition of this number into the frequency rates of the successive PICs from
0 (or 1) to 11 inclusive. It can be determined by creating a matrix of the PICs
represented by all PC pairs and adding up the occurrences of each PIC. Table 2.2
demonstrates this, using the PC set {1,2,5,6,7,9} (from ex. 2.8). The resulting num-
bers form an ordered array, which is called a vector—a PIC vector in this instance.
Inverse-related PICs are represented equally. It was the composer Donald Mar-
tino, a former student of Babbitt, who started using ordered arrays of numbers to
model the interval content of a PC set (Martino 1961, 27). In this he was followed
by Forte, who assigned the term “vector” to such arrays (Forte 1964, 141). In math-
ematics and physics, row vectors are used to model spatial quantities. A row vector
represents the co-ordinates marking a certain point in an x-dimensional space.
Contrary to Lewin, Martino and Forte considered unordered pairs of PCs, of
which there are ½(n2 + n) in each set of n elements, including the unisons. Their
vector—commonly known as the “interval vector” or “interval-class vector,” but

Table 2.2. PIC content of PC set {1,2,5,6,7,9}
PIC matrix: 1 2 5 6 7 9
1 0 1 4 5 6 8
2 11 0 3 4 5 7
5 8 9 0 1 2 4
6 7 8 11 0 1 3
7 6 7 10 11 0 2
9 4 5 8 9 10 0
Multiplicity per PIC PIC: (0) 1 2 3 4 5 6 7 8 9 10 11

(PIC vector): (6) 3 2 2 4 3 2 3 4 2 2 3
Table 2.3. Absolute-PIC content of PC set {1,2,5,6,7,9}
APIC matrix: 1 2 5 6 7 9
1 0 1 4 5 6 4
2 0 3 4 5 5
5 0 1 2 4
6 0 1 3
7 0 2
9 0
Multiplicity per APIC APIC: (0) 1 2 3 4 5 6

(APIC vector): (6) 3 2 2 4 3 1
more properly designated absolute PIC vector (APIC vector)—consisted of

six “co-ordinates” only: the frequency rates of APICs 1 through 6. The fre-
quency rates of the “unison” pairs (the 0s) were left out as being redundant.
This changed the number of pairs to ½(n2 − n). The APIC vector of the PC set
{1,2,5,6,7,9} and its derivation are shown in table 2.3.
PIC and APIC vectors have both been used as representations of the interval-
lic content of a PC set. Although the choice between the two vectors is of limited
importance, it should be noted that the entry of PIC 6 is twice as large as the
entry of APIC 6. (Each PC pair represents two inverse-related PICs, and PIC 6
is its own inverse.) This forbids their exchange in at least one proposed applica-
tion, which awaits discussion in the last section of chapter 5.
In the preceding pages, various aspects of the PC set—both technical and
historical aspects—have been touched upon. We have introduced the PC set
as a representation of musical entities. The ensuing discussion will address the
nature of these entities more thoroughly. Among other things, this involves the
musical relations in which they are found, and the history of these relations.

Chapter Three
Operations
Each combination of tones can be identified on the basis of its PC set. However,
for such a combination to be considered of structural interest—that is, worth
identifying at all—it is a necessary (though not a sufficient) condition that it bear
a relation to other combinations. A relation between two combinations of tones
can sometimes be conceived of as a transformation. PC set theory has defined
several operations that transform one PC set into another, the most important
of which are transposition (T), inversion (I), and multiplication (M). No doubt,
transposition and inversion are backed by the longest history. The discussion of
them in this chapter will reveal the strong bonds that tie PC set theory to the
history of music theory. Multiplication is considerably younger as a concept of
a musical transformation, but it, too, antedates PC set theory and is rooted in
compositional practice.
Notes on the Term “Operation”
The term “operation” has been borrowed from mathematics. Some confusion
may arise over its correct use. A mathematical operation is defined with respect
to a collection of elements. Often, it is a protocol (a “function” or “mapping” in
mathematical language) assigning to each pair of elements of a collection one
element of the same collection. This is called a “binary operation.” If we think of
the collection of integers, ordinary addition, subtraction, and multiplication are
examples of binary operations. Under addition, for example, 7 is assigned to the
pair consisting of 3 and 4.
There are similar protocols for collections consisting of sets. For example, the
union of two sets A and B (denoted A ∪ B) is a set consisting of all elements
belonging to A and all elements belonging to B. The intersection of A and B
(denoted A ∩ B) is a set containing the elements that A and B have in common.
The difference of A and B (denoted A − B) is the set consisting of those elements
of A that do not belong to B.
A special case of the difference of two sets is the complement. This has been
an important concept in PC set theory, for reasons to be discussed in chapters 4
and 6. Given a set of things that I shall call ELEMENT, ℘(ELEMENT) is the set
of subsets of ELEMENT. One member of ℘(ELEMENT) is the set that equals
ELEMENT. Now, when A is another member of ℘(ELEMENT), the complement
A' of A is the difference of ELEMENT and A:

50 operations
A' = ELEMENT − A
In the case of PC set complementation, PITCHCLASS substitutes for ELEMENT,

and PCSET for ℘(ELEMENT). The complement of the PC set {1,2,5,6,7,9}—the
PC set taken from Schoenberg’s Opus 11, no. 1 (see ex. 2.8)—is {0,3,4,8,10,11}.
The latter results when we “subtract” {1,2,5,6,7,9} from PITCHCLASS.
PC set theory conceives of an operation in a slightly different way, though. We
shall see that it refers to a type of protocol assigning to each single element of a
collection a unique element of the same collection. If this collection is called S,
the operation is said to be an “operation on S” (Lewin 1987, 3). Suppose S con-
sists of five elements: a, b, c, d, and e. Then table 3.1a shows such an operation.
Each element must appear once in the left-hand column and once in the right-
hand column. In other words, the operation is a bijective function. (The term
“bijective” implies that the arrows in table 3.1a can also be reversed.) This defini-
tion of an operation does not exclude complementation. For each set that is a
member of PCSET there is a unique set in the same collection that is its comple-
ment. However, it has particularly served as a model for transposition, inversion,
and multiplication. Some historical background may help explain why.
PC set theory has adopted its definition of an operation from twelve-tone
theory. The rise of twelve-tone serialism invited a description of compositional
techniques in terms of formal operations. Milton Babbitt defined the familiar
four forms of a twelve-tone series—prime, retrograde, inversion, and retrograde-
inversion—as resulting from such operations. Each of these series-forms can be
seen as a one-to-one mapping of PITCHCLASS onto itself, corresponding to
table 3.1a. As we saw in chapter 2, Babbitt identified the elements of PITCH-
CLASS by their pitch and order numbers relative to the actual twelve-tone series
he was using (see chapter 2, p. 33). He let both categories of numbers run from
0 through 11. The retrograde resulted from the mapping of each order number
i onto 11 − i, whereas the inversion resulted from the mapping of each pitch
number j onto 12 − j. The retrograde-inversion was the result of a composite
operation: the mapping of each pitch number j onto 12 − j, and the subsequent
mapping of each order number i onto 11 − i.
Rather than combining two elements in order to obtain a third one, these
operations affect each element individually. They thus seem to fall into the cate-
gory of the “unary operations,” such as—if again we think of integers—squaring.
It would match with the concept of a binary operation if the protocol itself were
viewed as the element of another collection, that is, a collection G of operators.
In that case, an element of collection S is associated with each pair of elements
from G and S (table 3.1b). But the moot point here is that, in mathematics,
operations, irrespective of whether they are unary or binary operations, need
not be bijections. This requirement applies to a category of functions called
“group actions.” When every element of G combines with each element of S to
yield another (or the same) element of S—and the latter is the one and only

operations 51
Table 3.1. Operations as defined by PC set theory
(a) (b)
a → b (g,a) → b
b → d (g,b) → d
c → e (g,c) → e
d → a (g,d) → a
e → c (g,e) → c
Table 3.2. Milton Babbitt’s “group multiplication table”
S I R RI
I S RI R
R RI S I
RI R I S
element associated with that combination, as in table 3.1b—G as a whole can

be said to “act” on S. Now, for this to be a “group action,” G should qualify as
a “transformation group” with respect to S. This means two things. First, if a is
an element of S, G should contain an element that combines with a to yield a
(as when we add 0 to an integer). Second, if one element of G combines with a
to yield b (g(a) = b), and another combines with b to yield c (h(b) = c), then the
composition of these two elements of G should combine with a to yield c (hg(a) =
h(g(a)) = c, where a, b, and c are all elements of S).
The reader may be wondering what sense to make of this. Basically, the above
is a way to express the symmetry of S under the application of G. As a consequence
of this symmetry, the distinction between “element” and “transformation” fades.
The essence of a transformation group’s “acting” on a set is that every element
of the latter becomes itself a transformation. Babbitt understood this very well
when he entered the four forms of a twelve-tone series in a “group multiplica-
tion table” (table 3.2; Babbitt 1960, 252). Here, the word “multiplication” refers
to the combination of two or more serial operations (or transformations). In
the table, each entry is the result of combining the left-most entry of the same
row and the uppermost entry of the same column. The table enables us to verify
that the prime (“S”), retrograde (“R”), inversion (“I”), and retrograde-inversion
(“RI”) satisfy the criteria of a transformation group.
We will return to this topic later (in chapter 4). For now, it suffices to say that
the transposition, inversion, and multiplication operations of PC set theory, too,
combine in a number of ways to form group actions with respect to sets of pitch
values, such as PITCHCLASS. Apart from their historical backgrounds, this is
another reason why PC set theory presents them as operations.

52 operations
Transposition
Transposition can be defined on PITCH or PITCHCLASS. To start with the for-

mer, the integer value of a pitch corresponds to that of a pitch interval. The
value a of a pitch corresponds to PI(0,a). Therefore, a pitch interval value can
be added to a pitch value. For example, when the pitch interval 4 is added to
pitch 9 (A4 in example 2.1), the latter transforms into 13 (C♯5). However, PC
set theory commonly conceives of transposition as the addition of PICs. It thus
defines the operation on PITCHCLASS. This means that a PIC value is added to
a PC value (mod 12). When PIC 4 is added to PC 9, the latter transforms into 1.
This leaves the octave ranges unspecified:
(9 + 4) mod 12 = 13 mod 12 = 1
The transposition of a PC set A is the addition (mod 12) of an integer n to each

of the PC integers of that set. This operation is designated Tn(A), with the index
n representing a PIC.1 When the PC set {0,6,7,9} is transposed with n = 4, the
result is {4,10,11,1}:
T4({0,6,7,9}) = {(0 + 4) mod 12, (6 + 4) mod 12,

(7 + 4) mod 12, (9 + 4) mod 12} = {4,10,11,1}
This new PC set has the same interval content as the original set. The APIC
vector (see chapter 2, p. 48) is 112011 in both cases. The operation is defined
for unordered PC sets, but can also be applied to ordered ones (for example, a
twelve-tone series).
A definition from Heinrich Christoph Koch’s Musikalisches Lexicon (1865 edi-
tion) captures what transposition is still commonly understood to mean today:
The shifting of a melody, a harmonic progression or an entire musical piece to

another key, while maintaining the same tone structure, i.e. the same succes-
sion of whole tones and semitones and remaining melodic intervals.2
Example 3.1 is the example that illustrates the entry in Koch’s Lexicon. Tra-
ditionally, the term “transposition” belongs to the spheres of music theory
1. For labels of operations I use a separate font to avoid confusion with labels of
objects and measurements (like PC, PI and PIC).
2. “Das versetzen einer Melodie, Harmoniefolge, resp. eines ganzen Tonstückes in
eine andere Tonart unter Beibehaltung derselben Tonordnung, das heisst, derselben
Aufeinanderfolge der ganzen und halben Töne und übrigen Tonschritte.” (Koch 1865,
879, my translation) This concise definition is lacking in Koch’s original article in the
1802 edition of the lexicon. However, it accurately summarizes Koch’s lengthy explana-
tion. The original musical example consisted of just the first two measures.

operations 53
Example 3.1. The transposition of a melody. A musical example from Koch’s

Musikalisches Lexikon (second edition, 1865).
(e.g., the transposition of scales and modes), music notation (e.g., transpos-
ing instruments), and performance practice (e.g., the transposition of a song
to accommodate different voice ranges). A compositional device to which it
might refer is the “real” answer to the subject of an imitative piece—say, a
fugue, motet, or ricercare. Here is a description of this device, matching the
above arithmetical definition of transposition in all its precision, from Rameau’s
Treatise on Harmony:
If one part begins or ends on the tonic note, the other should begin or end
on the dominant, and so on for each note thus related within the octave of the
key in use. The notes between the tonic and its dominant should also corre-
spond in each part; i.e., the second note which is immediately above the tonic
should correspond to the sixth which is immediately above the dominant; the
same holds for each note a third, a fourth, or a fifth above or below the tonic
and that note which is the same degree above or below the dominant, follow-
ing the direction of the melody, which may ascend or descend. The conformity
which we say should be observed by the notes which begin and end the fugue
should also be observed by the entire phrase making up this fugue.3
It is important to note that real answers were not straightforwardly called

“transpositions” in fugal treatises. Characteristically, when the term “transposi-
tion” turns up in the context of fugal theory in Christoph Bernhard’s Tractatus
Compositionis Augmentatus (1650), it refers to the shift from one key to the other
(transpositio modorum) rather than to the shift of the fugal subject itself (Bern-
hard 1926, 97). In English translations of fugal treatises, however, the term is
3. “Si une partie commence ou finit par la Notte tonique, l’autre doit commencer ou
finir par la Dominante; & ainsi de chaque Notte qui se répond dans l’étendue de l’Octave
du Ton que l’on traite, faisant en sorte, que les Nottes qui se trouvent entre la Tonique &
sa Dominante se répondent également dans chaque partie, c’est-à-dire, que la Seconde Notte
qui est immediatement au-dessus de la Tonique doit répondre à la Sixième, qui est imme-
diatement au-dessus de la Dominante, selon le progrès du chant, qui peut monter ou des-
cendre; car la conformité que nous pretendons devoir être observée dans ces Nottes qui
commencent & terminent, la Fugue doit être également observée dans toute la suite du
Chant dont cette Fugue est composée.” (Rameau 1722, 333. Transl. by Philip Gossett)

54 operations
5 6 7 8 9
Rasch ( = 80) 10 11 12
2 3
5 6 7 8
9 10 11 12
Example 3.2. Twelve-tone transposition. This is the beginning of the prelude from
Schoenberg’s Suite for Piano, Op. 25. The left hand realizes an imitative texture by
playing a transposition (T6) of the twelve-tone series in the right hand. The melodic
contour of the series is largely retained, with rhythmic alterations. The last four
notes of the transposition (9–12), however, have been used to create a counterpoint
against the four preceding notes (5–8). Furthermore, in this section of the trans-
posed series the “original” downward leap of a major seventh (B4–C4; m. 3, right
hand) is transformed into a rising minor second (F2–F♯2; m. 2, left hand).
used quite often, not only with reference to real answers, but also with reference
to tonal answers and sequences.4
With the introduction of twelve-tone serialism in the early 1920s, the concept
of transposition started to play a more prominent role in compositional theory,
and hence in music analysis too. It was listed among the canonical transforma-
tions of a twelve-tone series. However, a twelve-tone series is a much less tangible
unit than a melody, a harmonic progression, or a musical composition. Its ele-
ments are not pitches but PCs, and these may generate both melodies and har-
monic progressions. A twelve-tone series may not even be presented as a single
musical unit. This means that in the context of twelve-tone serialism the term
4. In English, the terms “to transpose” and “transposition” may signify other proce-
dures than the one described by Koch. For example, “transposition” oddly signifies a dia-
tonic sequence in Alfred Mann’s translation of Friedrich Wilhelm Marpurg’s Abhandlung
von der Fuge: “The restatement of a subject by use of different tones in the same part is
called transposition.” It is clear from the accompanying musical example that Marpurg is
talking about a diatonic sequence (Mann 1987, 142). Consultation of the original German
edition reveals that Marpurg does not use the word transponi(e)ren but versetzen (Marpurg
1753, 6). The term Versetzung is found as a distinct entry in Koch’s Musikalisches Lexicon,
and it is reported there to differ from Transposition in that it allows the position of the
semitones to vary according to the degree of the scale on which the subject appears (Koch
1865, 919–20). Since transponi(e)ren is defined by Koch as a special case of versetzen, the lat-
ter term must have been considered a more general one, referring to any kind of moving
of a subject in pitch space. Therefore, it is recommended to use an English equivalent for
versetzen other than “to transpose.” “To shift” would be an appropriate alternative.

operations 55
“transposition” had a much more flexible meaning than in Koch’s Musikalisches

Lexicon. According to Koch, it was essential to preserve the melodic contour of
the transposed unit (“the same succession of whole tones and semitones and
remaining intervals”); by contrast, the transposition of a twelve-tone series might
involve a considerable change of its intervallic profile (ex. 3.2).
Milton Babbitt defined transposition in arithmetic terms. In his 1946 thesis
The Function of Set Structure in the Twelve-Tone System he wrote:
By applying the transposition operator (T) to a [twelve-tone] set we will mean

that every p of the set P is mapped homomorphically (with regard to order)
into a T(p) of the set T(P) according to the following operation:
To(pi,j) = pi,j + to
where to is any integer 0–11 inclusive, where, of course, the to remains fixed
for a given transposition. The + sign indicates ordinary transposition. (Babbitt
1992, 10)
Recall that in Babbitt’s notation every element (p) of the twelve-tone series (P) is
indicated by its position in that series (here the “order number” i) and its pitch
value modulo 12 (which is not yet called “pitch-class” but “set number,” in this
case j). The last sentence (“The + sign. . .”) designates addition as the standard
protocol of PC(-set) transposition. In other words, the PIC integer o is added to,
not subtracted from, the pitch value j.
The expression “is mapped . . . into” can be read, “is transformed into” or
“transforms into.” In 1946, however, Babbitt called transposition not a “trans-
formation” but a “translation,” a term borrowed from geometry and mechanics
referring to a motion without rotation. He used the term “transformation” with
reference to the inversion and retrogradation of a twelve-tone series only. Under
transposition, the succession of “interval numbers” (that is, of PICs) remained
invariant; for this reason, Babbitt considered transposition a “secondary” opera-
tion, an operation with “very little independent significance” (Babbitt 1992, 15–
16). This was in keeping with earlier treatises on twelve-tone serialism, in which
transposition used to be mentioned only, and with little emphasis, after the mir-
ror operations (inversion, retrogradation, and retrograde-inversion).5 Babbitt’s
later writings mark a shift in the valuation of twelve-tone transposition. In his
article “Twelve-Tone Invariants as Compositional Determinants,” published in
the Musical Quarterly in 1960, he put it on a par with inversion, arguing that it
likewise resulted in a permutation of all twelve “pitch numbers” (Babbitt 1960,
249) or “pitch class numbers” (ibid., 253). Transposition thus rose considerably
in rank as a musical operation.
5. See, for example, Stein 1925, Krenek 1940 or Schoenberg 1984.

56 operations
12 25
A: [2, 3, 7, 8, 9] B: [0, 1, 5, 6, 7]
Example 3.3a. PC-set transposition in the fifth of Anton Webern’s Five Movements
for String Quartet, Op. 5. An example from Allen Forte’s The Structure of Atonal
Music. In the first excerpt, the first beat should be divided into three. The triplet
sign is missing. Furthermore, it should be noted that Forte notates unordered PC
sets with square brackets instead of curly ones.
Copyright © 1973 by Yale University.
Sehr langsam ( = 40)

3 rit. Sehr langsam ( = ± 40) rit. verlöschend
12 arco 25
verlöschend
verlöschend
verlöschend
Example 3.3b. A less contracted version of Forte’s example. The boxes mark the
elements singled out by Forte.
© 1922 by Universal Edition A.G., Wien/UE 5888, UE 5889.
Babbitt’s definition of transposition showed a correspondence between

certain musical operations and operations on numbers. Babbitt used this
correspondence to explain and predict the effects, in terms of PC relation-
ships, of serial manipulations. This was the beginning of a new, distinctly
American branch of music theory. In European treatises on twelve-tone seri-
alism numbers were used merely as labels. Besides, no standard had emerged
or would crystallize in the years to follow, as a few examples will illustrate.
The numerals of Arnold Schoenberg (“Composition with Twelve Tones,” the
written version of a lecture dating from the 1930s, first published in Style and
Idea, 1950) and Josef Rufer (Die Komposition mit Zwölf Tönen, 1952) strangely

operations 57
enough indicated diatonic intervals.6 Ernst Krenek (Studies in Counterpoint,

1940),7 René Leibowitz (Schönberg et son école, 1947), and Hanns Jelinek
(Anleitung zur Zwölftonkomposition, 1952) numbered the twelve transpositions
of a series from 1 to 12 (instead of from 0 to 11). Herbert Eimert (Lehr-
buch der Zwölftontechnik, 1950) called the transpositions of a twelve-tone series
by the traditional letter names of their first PCs. Thus, “U(cis)” meant: the
inversion (“Umkehrung”) beginning on C♯. A similar rule was introduced
many years later by George Perle (Twelve-Tone Tonality, 1977). Instead of
Eimert’s bracketed letter names, Perle added absolute subscript numerals to
the series-form labels “P” (“prime form” or “prograde”), “I” (“inversion”),
“R” (“retrograde”) and “RI” (“retrograde-inversion”)—with “0” representing
C, as in Forte’s notation. He also used this protocol in his and Paul Lansky’s
contributions to the New Grove Dictionary of 1980.8
Allen Forte extended Babbitt’s definition of transposition to unordered PC
sets of cardinal numbers other than 12. In his 1964 essay “A Theory of Set-
Complexes for Music” he spoke of the mapping of a PC set “P” into the universal
set “S” (designated PITCHCLASS in chapter 2) under transposition, which he
defined as: “the addition mod 12 of any integer k in S to every integer p of P.”
Thus there are “12 transposed forms of P” (Forte 1964, 149).9 Any reference to
order is omitted in this definition, which basically equals the one given at the
beginning of this chapter. After the abandonment of contour preservation this
meant a significant further generalization of the concept of transposition.
George Perle had already enforced this generalization in his book Serial Com-
position and Atonality (1962), which was the standard American text on the analy-
sis of twentieth-century music in the 1960s and early 1970s (that is, before the
publication of Forte’s The Structure of Atonal Music). It is noteworthy that “serial
composition” preceded “atonality” in the title. This was indicative of Perle’s
approach, which laid stress on those aspects of “free” atonal music that antici-
pated the serial twelve-tone technique:
Microscopic elements are transposed, internally reordered, temporally or spa-

tially expanded or contracted, and otherwise revised, in a fluctuating context
that constantly transforms the unifying motif itself. (Perle 1962, 19)
6. For an explanation of Schoenberg’s protocol, see Schmidt 1988, 19. Schoenberg used
the signs “+” and “−” to differentiate between major (in case of a fourth: augmented) or minor
(in case of a fifth: diminished) intervals. Furthermore, he counted the transpositions of the
original series in upward direction and those of the inversion in downward direction.
7. I regard Krenek’s manual as a European contribution to twelve-tone theory,
although it was originally published in the United States.
8. See the entries “Serialism” and “Twelve-Note Composition.”
9. It should be noted that certain PC sets—for example the PC sets of augmented
triads—may transform into duplicates under transposition. The number of distinct trans-
posed forms is smaller in these cases.

58 operations
(a) 13 Tempo I 14
= ± 60 rit. = 48
pizz. arco
espr.
pizz.
= ± 60 rit.
2 3
mit Dämpfer pizz. arco
(b)
pizz.
[pizz.] pizz. arco 15
arco 11
espr.
[am Steg] pizz.
3 pizz.
( )
pizz.
pizz.
am Steg
pizz.
Example 3.4. Instances of T9 in Webern’s Opus 5, no. 5.

In Perle’s view, transposition was an operation applied to unordered collections

of any number of tones. For him, the meaning of the word “set”—which he had
borrowed from Babbitt—included such collections. Oddly enough, he regarded
them as attributes of “serial composition,” even when the compositions in which
he had found them did not feature any series, twelve-tone or otherwise:
In a strict sense the term “series” denotes an ordered succession of elements,

such as the Schoenbergian twelve-tone set . . . [In] the work of Scriabin and
some other composers the set is a collection of pitches the specific ordering
of which is purely compositional. The term “serial composition” is used in the
present study as a general designation for music based on any of these types of
set. (Perle 1962, 37n)

operations 59
From the last part of this quote it appears that Perle regarded the twelve-tone
series as the reference point for a PC set. Forte, however, considered the PC set
to be independent from the twelve-tone series. And whereas Perle, the composer,
spoke of transposition as a variation technique, Forte, the analyst, used transpo-
sition to identify relations between groups of tones, regardless of whether these
relations were consciously used by the composer.
In The Structure of Atonal Music, Forte provided three examples of the transpo-
sition operation. One of these is reproduced here as example 3.3a (Forte 1973,
5). It concerns two excerpts from the last of Anton Webern’s Five Movements
for String Quartet, Op. 5. In Forte’s book, musical examples are usually very
contracted, so as to enable maximal focusing on the properties discussed. As a
consequence, they veil a serious problem: when is a transpositional relationship
important? An examination of the context from which example 3.3a has been
extracted reveals this problem. Although the PC set of the second excerpt—the
“chord” (m. 25)—is irrefutably a transposition (T10) of that of the first—the
“melody” (m. 12)—the excerpts do not seem to interact very strongly. They are
separated by almost thirteen eventful measures, and the “chord” is not as clearly
articulated as the “melody.” As example 3.3b shows, the F♯4 in measure 25 is
part of a short line played by the first violin. This line refers back to the open-
ing of the piece played by the cello. Since tones cannot simply be classified as
“chord tones” or “non-chord tones” in the atonal idiom of this piece, it is hard to
prove that the first violin’s F♯4 also forms a meaningful entity with the sustained
tones of the other instruments in measure 25. It is surprising, by the way, that
Forte did not refer to the earlier and more marked appearance of the “chord”
on the second beat of measure 24.
The “melody” is noticeable, to be sure. It is played over a single harmony,
as we can see in example 3.3b, and it provides a strong timbral and textural
contrast with the measures that precede and follow it. What is more, an almost
literal transposition of it (T8) appears in measures 17–18. Only the rhythm has
changed a bit. Although this transposition is not particularly interesting from
the perspective that Forte’s definition offers,10 it is precisely its being “almost
literal” that qualifies the “melody” as contextually meaningful. PC sets cannot
convincingly be shown to function as transpositions of other PC sets without
invoking criteria other than the transpositional relationship alone—criteria such
as textural salience, the preservation of a melodic contour, or some significant
number of such relationships.
One wonders why Forte did not select two other excerpts from Opus 5, no.
5 to show PC set transposition. The operation T9 links the PC set of the cello’s
opening line to that of the pizzicato gesture in measure 13. Furthermore, it links
the latter to the next five notes in the piece, four of which are also played by the
cello alone (mm. 13–14; ex. 3.4a). These tone combinations stand out texturally
10. Forte calls it an “ordered transposition.” (Forte 1973, 62)

60 operations
and timbrally. Also, measures 13–14 refer back to the opening by featuring the
sound of the cello’s lower strings, solo. Besides, it marks the return of the initial
tempo (eighth-note: M.M. 60), which is slowed immediately afterwards, just as it
was in the two first measures (to eighth-note: M.M. 48). The melodic interval of
the perfect fourth, the opening line’s first interval (F♯2–B2), is also prominent
in the tone combinations of measures 13–14 (as A♭2–E♭2 and C2–F2 respec-
tively). Finally, the operation T9 defines the relationship between two other
important excerpts as well, and it does so much more obviously: measures 10–11
(sehr ruhig) and measures 14–15 (ex. 3.4b). This is an “almost literal” transposi-
tion, like the transposition of the “melody” in measures 17–18.
Forte’s book The Atonal Music of Anton Webern (1998) contains an analysis of
this movement that does not mention the transpositional relationship of example
3.3 anymore. Nor does it mention the transpositional relationship presented in
example 3.4a—that is, the relationship between the cello’s opening line and the
events in measures 13–14. Forte provides a reading that rules out the introductory
cello solo as a point of comparison. He detaches the perfect fourth F♯2–B2 from
the rest of that solo, reasoning that this interval is not “a conspicuous contributor
to the motivic fabric of the movement” (Forte 1998a, 85)—a challengeable posi-
tion. On account of the PC sets he has found articulated further on in the piece,
Forte singles out the successions B2–G3–G♯2–C3–E2–C♯2 (mm. 1–3), B2–G3–
G♯2–C3–E2 (mm. 1–2), and C3–E2–C♯2 (mm. 2–3) (ex. 3.5; ibid.).
Forte’s selection of PC sets is, of course, open to dispute. However, there is a
much bigger problem than its being right or wrong. The above discussion has
shown how hard it is to prove the importance of a selection of PC sets—or to
disprove it, for that matter. Unordered PC sets are quite malleable: there are
no criteria for their verification, how they are projected musically, and how they
connect to form larger structures. For example, from the standpoint of musical
phrasing one might object to Forte’s pasting the cello’s ostinato motif E2–C♯2
in measures 3ff. on to its introductory solo. The ostinato motif obviously defines
a new section. However, PC sets are not necessarily defined by phrases or sec-
tions. They are not necessarily defined by any obvious musical shape or Gestalt.
What, then, does it mean to say that the PC set represented by one combination
of tones is the transposition of the PC set represented by another? It suggests
that the musical piece in question is coherent in an inconspicuous way; that con-
figurations of tones, disparate though they may seem, are based on similar inter-
vallic structures. And this, in turn, may serve in a debate on the assessment of
that piece. From the very beginning of his involvement with atonal music, it has
been Forte’s concern to defend the early atonal repertoire against the criticism
that it lacked a structural principle of its own, that it was either distorted tonal
music, or half-baked serial music (Forte 1963, 72).
The claim that musical coherence resides in the consistent use of abstract
intervallic patterns obviously takes its cue from the paradigm of twelve-tone seri-
alism. However, unlike the concept of a twelve-tone series, that of an unordered

operations 61
(a)
(b)
21
Vl.I am Steg
Vc. 5
Example 3.5. Excerpts from Webern’s Opus 5, no. 5, selected by Allen Forte in The
Atonal Music of Anton Webern.
(a) Forte’s segmentation of the opening line of Webern’s Opus 5, no. 5.
(b) A motif transpositionally related to the succession B2–G3–G♯2–C3–E2 in m. 1–2 (m. 21).
PC set lacks a clear technical or procedural basis, due to the process of general-
ization from which it results.
This process of generalization has continued after the publication of The
Structure of Atonal Music. Robert Morris (Composition with Pitch-Classes, 1987)
and David Lewin (Generalized Musical Intervals, 1987) specified multiple
“spaces” to which this and other operations could be applied. (The word
“space” denotes a set, usually a set delimiting a world of specific musical
objects.) The twelve PCs under equal temperament provide but one such
space. For example, the transposition operation can be applied to objects
forming a (modular or non-modular) diatonic space,11 or to the space out-
lined by a chord. Spaces can also consist of objects other than pitches or PCs.
For example, objects can be chords, durations, or time-points. In a diatonic
space, or in a space consisting of triads (like the spaces pertaining to Richard
Cohn’s “hexatonic systems”; Cohn 1996), a minor triad can be a transposi-
tion of a major triad, and vice versa, which is impossible in spaces like PITCH
or PITCHCLASS.
Joseph Straus (1997) introduced the term “near-transposition” (along with
“near-inversion”), which he later replaced by “fuzzy transposition” (and “fuzzy
inversion”).12 He conceived of transposition as a voice-leading event, the “sending”
11. The objects of a modular diatonic space are PCs; the objects of a non-modular
diatonic space are pitches.
12. Joseph N. Straus, “Voice Leading in Atonal Music,” unpublished lecture for the
Dutch Society for Music Theory, delivered on April 11, 2003. Royal Flemish Conservatory
of Music, Ghent, Belgium.

62 operations
of each element of a given PC set to its Tn-correspondent. This view enabled him
to relate PC sets of two adjacent chords in terms of a transposition, even when
not all the “voices” participated fully in the transpositional move. The “forces”
determining successions of chords in atonal music were originally beyond the
scope of PC set theory. Straus’s invocation of voice leading can be seen as an
attempt to bring up the question of what these “forces” are—without question-
ing the theory.
Inversion
In the context of PC set theory, inversion is most appropriately defined as an

operation applied to pitch intervals or PICs. The inversion of a pitch interval (x,
x + a) yields (x, x − a). The latter is called the inverse of (x, x + a). The position
x on the pitch scale is the axis of inversion. According to this definition each
interval inverts around one of its own elements. For example, the inversion of
(G4,F5) or PI(7, 7 + 10), is (G4,A3), or PI(7, 7 − 10). Inversion of PICs is carried
out modulo 12. It relates a larger number of pitch intervals and it involves no
“real” axis of inversion. The pitch interval (G4,F5) may transform into (G3,A5)
or into (G4,A4), to mention just two possibilities.
It is customary to speak of the inversion of single PCs as well. This use matches
the above definition of inversion if a PC a mod 12 is conceived as a PIC of the
type PI(0,a) mod 12. Then, this PC inverts to −a mod 12. The inversion of a PC
set A means the application of this protocol to each of the members of that set,
and is designated I(A). The entire PC set A is thus, in an abstract sense, inverted
around 0. When A = {0,6,7,9}:
I({0,6,7,9}) = {0 mod 12, −6 mod 12, −7 mod 12, −9 mod 12} = {0,6,5,3}
Inversion does not affect the APIC vector of the original PC set (112011). Like
transposition, it can also be applied to ordered PC sets.
“Inversion” was an important subject in treatises dealing with fugal tech-
nique. Nicola Vicentino’s L’antica musica ridotta alla moderna prattica (1555) and
the third part of Gioseffo Zarlino’s Le Istitutioni harmoniche (1558) constitute
early milestones in this field. The authors pointed out that subjects of fugues
could be imitated alla riversa (Vicentino) or per mouimenti contrarij (Zarlino),
and that it was of particular interest when they were imitated strictly, so that
each falling semitone in a subject would correspond to a rising one in the
answer and vice versa.
Vicentino and Zarlino supplied their examples with notes that are descriptive
rather than explanatory. At first sight, the examples of both authors are without
sharps or flats. From Vicentino’s example it can be deduced that he exploited
the symmetrical division of the Dorian octave (in diatonic steps: 1 ½ 1 1 1 ½ 1),

Example 3.6. The inversion of a fugue subject. An example from Nicola Vicentino’s
L’Antica Musica ridotta alla moderna prattica. The interval measures are retained strictly.
Documenta Musicologica. Erste Reihe: Druckschriften-Faksimiles XVII. © Bärenreiter-Verlag,
Kassel.
Example 3.7. The inversion of a fugue subject around B. An example from Gioseffo
Zarlino’s Le Istitutioni harmoniche. A strict inversion of the subject requires that one
of two corresponding Bs be flattened in either guida or consequente.
Facsimile of the 1558 edition, Monuments of Music and Music Literature in Facsimile, II/1 (New
York: Broude Brothers Limited). Reproduced by arrangement with Broude Brothers Limited.

64 operations
in accordance with the directions of later theorists (ex. 3.6; Vicentino 1555, 89).
In the first of Zarlino’s examples, however, the subject appears to invert around
the double-faced B of the Renaissance tonal system. The desired inversional sym-
metry of guida (i.e., the subject) and consequente (the answer) arises only when
the b quadrato in one voice is answered by a b molle in the other and vice versa
(ex. 3.7; Zarlino 1558, 215). The rules of musica ficta do not always give a decisive
answer to the question which B should be taken as b quadrato and which as b
molle. For example, when the progression A–B–C is answered by C–B–A, it is only
reasonable to think that the latter B was flattened. This is not certain, though. In
any case, the Renaissance tonal system allowed D (and, by transposition, G) and
B to serve as centers of strict inversion.
It was for theorists of later generations, like Giovanni Maria Bononcini (Musico
Prattico, 1673), Johann Joseph Fux (Gradus ad Parnassum, 1725), and Friedrich
Wilhelm Marpurg (Abhandlung von der Fuge, 1753) to explain inversional rela-
tionships in more general terms, and to establish a systematic procedure that
would be of service to teachers of counterpoint for many years.13 The first step
of this procedure was to write down a diatonic scale spanning an octave and
showing the same succession of intervals in ascending and descending direction.
Of the seven diatonic octave species, only the Dorian had this property. Next,
one paired off the notes that were equally far removed from the opposite sides
of the scale: D and D, E and C, F and B, and G and A (ex. 3.8a). Finally, to pro-
duce the answer to a subject in strict contrary motion, one had to replace each
note of the subject by the one it formed a pair with (ex. 3.8b; Bononcini 1673,
84–85). According to Aloysius—the teacher in Fux’s dialogized treatise—it was
“inverted” to that other note:
Strict inversion arises if the progression of notes is so inverted that the relation
of mi to fa is always retained . . . If you compare the notes of the ascending
scale [from D4 to D5] on the left with those of the descending scale [from
D5 to D4] on the right, you will find that the d remains d by inversion, e is
inverted to c, f to b, and g to a.14
Theorists using the viewpoint of major-minor tonality, to whom speaking of the

Dorian scale must have been something of an anachronism, put this differently.
Marpurg’s instruction in his Abhandlung von der Fuge was to place the ascend-
ing octave of the first degree of a major key against the descending octave of
13. Reference to this procedure, and to its application by Bononcini, is made in Mann
1987, 144. That book also includes the relevant passages from the works of Fux and Marpurg.
14. “Contrarium autem reversum efficitur ita invertendis notis, ut ubique mi contra
fa eveniat . . . Aequiparentur notae à sinistris ascendendo, cum illis à dextris descen-
dendo, & reperies, D. per inversionem nihilominùs D. manere: E. in C., F. in B.mi, G. in
A. inverti.” (Fux 1725, 204–5; English translation in Mann 1987, 130)

operations 65
(a)
(b)
Example 3.8. The inversion of a fugue subject according to Giovanni Maria

Bononcini.
(a) Symmetrical relationships between the elements of the Dorian scale.
(b) The employment of these relationships in the answer.
Facsimile of the 1673 edition, Monuments of Music and Music Literature in Facsimile, II/78 (New
York: Broude Brothers Limited). Reproduced by arrangement with Broude Brothers Limited.
the third (that is, C–C/ascending versus E–E/descending), and the ascending
octave of the first degree of a minor key against the descending octave of the
minor seventh (that is, A–A/ascending versus G–G/descending; Marpurg 1753,
6). As example 3.9 shows, Marpurg thus arranged the notes in the same pairs as
Bononcini and Fux (C/E, D/D, E/C, F/B, and G/A), but he described the deri-
vation of these pairs through modern concepts of tonality.
In modern tonal practice, however, strict inversion was hardly, if at all, pur-
sued. Inversions of fugue subjects, such as in The Well-Tempered Clavier, or in The
Art of Fugue, left the framework of the key (consisting of tonic, mediant, and
dominant) intact. Intervallic strictness was subordinate to tonal firmness.15 In
15. For the same reason, “tonal” answers were preferred to “real” ones when the lat-
ter affected the main key’s framework right from the beginning. This would happen if in
the exposition of a fugue the subject started on the fifth. A “real” answer would then give
the main key’s supertonic as its first note. In a “tonal” answer that supertonic is replaced
by the tonic.

Example 3.9. Marpurg’s derivation of inverted forms of the major and minor scales.
Nederlands Muziekinstituut, Den Haag. NMI 6 E 22–23.
(a) ∧
1
[]
∧
5
∧
5
(b) ∧
1
(c) ∧ ∧
1 5
∧
5 ∧
1
Example 3.10. Tonally modified inverted forms of subjects in fugues by Johann

Sebastian Bach.
(a) The Well-Tempered Clavier I: Fugue in D minor, mm. 21–23.
(b) The Well-Tempered Clavier II: Fugue in B♭ minor, mm. 1–2 and 42–43.
(c) The Art of Fugue: Contrapunctus V, mm. 1–6.

operations 67
an inverted subject, the tonic was more likely to correspond to the dominant
than to the third (in major) or minor seventh (in minor). Example 3.10 pres-
ents three rectus/inversus pairs from The Well-Tempered Clavier and The Art of Fugue.
None of them can be classified as an instance of strict imitation. Not until the
twelve-tone system established itself as an alternative to the major-minor system
did the principle of strict inversion reappear.16
Unlike the diatonic collection, the fully circular twelve-tone collection allows
each tone to function as an axis of strict inversion. When the operation is
applied to a twelve-tone series, that series as a rule inverts around its first tone.
Twelve-tone serialism has affected the concept of inversion in the same way as it
has affected the concept of transposition. Since a twelve-tone series is basically
an abstract configuration, an inversional relationship between two such series
may not be obvious. The registral distribution of PCs is not essential to either of
them, although a composer could choose to observe the exact intervallic rela-
tionships in the series.
Like twelve-tone transposition, twelve-tone inversion was defined arithmeti-
cally by Milton Babbitt in his 1946 thesis:
Given a certain prime set, with general element pi,j; under the inversion oper-
ation, pi,j → I(pi, 12 − j); that is, each element of the prime set is mapped into
an element with identical order number but with set number the complement
(mod.12) [sic] of the original set number. (Babbitt 1992, 16)
Apart from the reference to an “order number,” this equals the definition of PC
set inversion with which this section opened. The “set number” is the PC integer,
and for “complement (mod 12)” one may read “inverse modulo 12.”
As with most texts on twelve-tone technique, Babbitt’s definition presumes
the axis of inversion to be the first PC of the prime form. For each twelve-tone
series there is one inversion that can be transposed eleven times. This means
that it requires a composite operation, designated TnI (or In), to get from the
prime form to an inverted form at a different pitch level (different, that is, with
respect to PC). Thus a hierarchy of series-forms suggests itself. Each series-form
is at some “distance” from the prime form, a “distance” measured in operations.
This hierarchy may or may not be exploited in musical works. It has been put
to use in an early serial work like Schoenberg’s Wind Quintet, Op. 26, which
plays with the four basic series-forms—the prime, the inversion, the retrograde,
16. For a discussion of symmetrical inversion in the context of late-nineteenth- and

early-twentieth-century harmonic dualism, see Bernstein 1993. Bernstein also discusses stud-
ies of symmetrical inversion by Georg Capellen (Fortschrittliche Harmonie- und Melodielehre,
1908) and Bernhard Ziehn (Harmonie- und Modulationslehre, 1887; “Über die symmetrische
Umkehrung,” in Gesammelte Aufsätze zur Geschichte und Theorie der Musik, 1927), both of whom
used it as a means to stretch the confines of harmonic tonality, and, in that respect, prefig-
ured the practice of twelve-tone composition. See also chapter 6, pp. 215–16.

68 operations
and the retrograde-inversion (first movement, mm. 1–24)—before it introduces

a series much further “removed” from the prime form: the T5 of the retrograde-
inversion (m. 24). In his analysis of this piece Schoenberg wrote:
While a piece usually begins with the basic set itself, the mirror forms and
other derivatives, such as the eleven transpositions of all the four basic forms,
are applied only later; the transpositions especially, like the modulations in for-
mer styles, serve to build subordinate ideas. (Schoenberg 1984, 227)
Schoenberg saw the twelve-tone system as hierarchical. The above statement

betrays a parallel between his hierarchy of series-forms and the hierarchy of
tonal keys. However, the basic serial operations often served just to generate
potential material for a composition. The composition itself could impose new
hierarchies on that material. In Schoenberg’s Variations for Orchestra, Op. 31,
the musical argument is dominated by series-forms containing dyads of B♭ and
E and of C♯ and G.17 Still, serial nomenclature conveys the notion of a hierarchy
of twelve-tone series in a musical work, relating them all to one “prime form.”
And this notion involves considering one axis of inversion only.
Now, even if there is a PC functioning that way, it need not be the first PC
of the prime form. In the second movement of Anton Webern’s Variations for
Piano, Op. 27, another axis clearly announces itself. The movement is based on
a succession of pairs of inversionally related series-forms set almost note-against-
note. Moreover, the series-forms of each pair are transposed with inversionally
related values (for example, 5 and 7), so that they keep forming the same pairs
of PCs.18 Example 3.11 shows the entire pitch material of this “classic” of axial
symmetry. The piece projects most PCs into a single octave range. Therefore,
its structure is best described in terms of pitches, not PCs. And the pitches, in
turn, are most appropriately defined in terms of pitch intervals measured from
A4, since each of them is audibly associated with its inverse about A4. Thus,
G♯3—that is, PI(A4,G♯3) or −13—is associated with B♭5—that is, PI(A4,B♭5)
or 13—on the upbeat of measure 1 and further in measures 5, 8–9, 11, 15, 18
and 22. Likewise, C♯4—that is, PI(A4,C♯4) or −8—is associated with F5—that is,
PI(A4,F5) or 8—in measures 2, 3–4, 8–9, 10, 16–17, 17 and 19–20.19 The inverse-
related PCs D and E are projected in various octave ranges, but always in such a
way that they invert to one another around A4 (cf. mm. 2–3, 6, 8, 15, 17 and 22).
The same is true for the PCs B and G (cf. mm. 2, 5–6, 8, 8–9, 12, 12–13, 19–20
17. Since the pitch levels of the two intervals of PIC 6 in the prime series-form of
Opus 31 are separated by an interval of PIC 3, a succession of B♭and E (or of E and B♭)
will necessarily occur with a succession of C♯ and G (or of G and C♯) in the same series.
In sum, there are 16 series-forms containing dyads of these PCs. This obviously has to do
with the symmetrical properties of the diminished seventh chord (C♯–E–G–B♭).
18. Perle (1977, 6) called such pairs P/I dyads.
19. The pitches G♯3 and C♯4 are sometimes spelled A♭ and D♭ in the score.

operations 69
and 20). The occurrences of E♭ (or D♯) are arranged symmetrically around A4
as well (mm. 6, 15 and 21). Finally, A4 inverts to itself, which accounts for the
striking tone repetitions in this piece (mm. 2, 9, 13 and 19).
When Babbitt analyzed this movement in his 1960 article “Twelve-tone Invari-
ants as Compositional Determinants”—so as to show that “the IT [sic] operation . . .
effects a categorization into [even and odd] interval classes”20—he labeled the
series-form starting on G♯3 in the right hand “prime form.” The inversion, which
starts on B♭5 in the left hand, was labeled “T2I,” in accordance with notational
conventions (Babbitt 1960, 254). This means that Babbitt considered his prime
form as inverted around G♯ and subsequently transposed with n = 2.21
Babbitt’s labeling is not incorrect, but in view of the pitch structure’s sym-
metry around A4 it is not too accurate either. The problem is the use of a label-
ing system that describes relationships between entities in only one, sometimes
circuitous way, whereas the music itself may simply take a “shortcut” from one
entity to the other. Here the reader may be reminded of David Lewin’s objec-
tions against the assignment of pitch values in terms of distance from an arbi-
trary reference point (like G♯3 in ex. 3.11). As Lewin pointed out, this involves
the risk of arbitrarily attributing a centric, or “tonic,” status to the latter. And it
may lead to expressions that obscure rather than clarify the actual musical struc-
ture (like “T2I” in Babbitt’s analysis of the same example).22
The problem looms larger when we consider Allen Forte’s definition of inver-
sion. According to Forte “each element p of [a given set] P is associated with one
and only one inverse element s = p’ in [the universal set] S” (Forte 1964, 144).
Order and magnitude of PC sets are no longer relevant in this definition. A PC
set is the inversion of another if it consists of all the inverse-related PCs, no mat-
ter what kind of structure they form. In this respect, the concept of inversion
has developed in the same way as the concept of transposition. What should be
noted, however, is Forte’s preliminary decision to assign the integer 0 to C as a
rule—so as to let this PC become a general axis of inversion—and for the rest to
retain the serial nomenclature, including the underlying notion of transposed
20. A sample of Babbitt’s characteristically dense prose. This sentence communicates

to the reader that the classes of intervals between corresponding notes of inversionally
related series are either all odd—e.g., minor seconds (1), minor thirds (3), major sixths
(9) and major sevenths (11)—or all even—e.g., major seconds (2), major thirds (4),
minor sixths (8) and minor sevenths (10).
21. The index n of the operation TnI is the sum of the integer values of a PC a and its
correspondent under that operation, i.e., (−a + n) mod 12:
n = a + (−a + n) mod 12
Thus, the PC values of each pair of inverse-related PCs in Webern’s movement add up to
2 (mod 12)—that is, if the value 0 applies to G♯. The index n of transposition (Tn), on
the other hand, represents the difference between the integer values of a PC (a) and its cor-
respondent (a + n).
22. Lewin 1977a; see also chapter 2, p. 35.

70 operations
(a)
= ± 160
(b)
Example 3.11. The symmetrical arrangement around A4 of the pitch material in

the second movement of Anton Webern’s Variations for Piano, Op. 27.
© 1937 by Universal Edition A.G., Wien/UE 10881.
inverted forms of PC sets. As a consequence of this decision, discrepancies such

as those arising in Babbitt’s Webern analysis are sanctioned by PC set theory.
We can find such a discrepancy in Forte’s analysis of measure 30 of Arnold
Schoenberg’s Piano Piece, Op. 23, number 3, which also discusses structural
aspects other than inversional relationships. Example 3.12 shows the fragment
(which has been enlarged to include measure 31) with a segmentation that Forte
may have borrowed from George Perle (1962, 45). This segmentation isolates
the dyad C4–G4 that the two sonorities of measure 30 have in common. Forte
obtains the PC set of the sonority labeled β by the operation T7I on the PC set of
the sonority labeled α (Forte 1973, 74–75). The same operation transforms the
PC set of γ into that of δ. This formulation does not make clear that the pitch
structure of this fragment is as symmetrical as that of the second movement of
Webern’s Opus 27. The “axis” lies between E♭4 and E4 this time. The dyad C4–
G4 is part of both sonorities, and is retained under the operation because C4
and G4 invert to one another around this axis. However, since the axis is not a
pitch, it is no use trying to show axial symmetry by an assignment of pitch values;
there is no pitch qualifying for the label “0” more obviously than any other.
On the positive side, it is significant that the interval retained under inver-
sion is a perfect fifth. This provides a link with the practice of tonal inversion,
where the tonic inverts to the dominant and vice versa (see ex. 3.10). Such links
mattered to Schoenberg. The analysis, in chapter 1, of the opening measures of
Opus 23, no. 3, showed a similar preponderance of the fifth relation. (In these
measures, the PC sets of α and δ underlie the fugal subject and its answer; see
ex. 1.1a.) Seen thus, the label T7I—“invert and transpose with seven semitones”
(i.e., a fifth)—may not be that bad after all. But its use is merely coincidental. In

operations 71
=
30 α γ
31
[ ]
β δ
Example 3.12. A symmetrical arrangement of PCs. Two measures from the closing
section of Schoenberg’s Piano Piece, Op. 23, no. 3. The segmentation was
proposed by George Perle and adopted from him by Allen Forte. An imaginary
axis lies between E♭4 and E4
FÜNF KLAVIERSTÜCKE, OP. 23, by Arnold Schoenberg. Copyright © Edition Wilhelm Han-
sen AS. International Copyright Secured. All Rights Reserved. Used by Permission.
Forte’s system, a fifth other than that between C and G would have required a
different label; and Schoenberg did not think of C as a kind of “fixed do.”
Although it is not a pitch, there still is an imaginary axis of inversion in exam-
ple 3.12. Sometimes, however, it is impossible even to imagine an axis, like in
example 3.13. This is a fragment of the third of Webern’s Six Pieces for Orches-
tra, Op. 6. The first flute, the horn, and the glockenspiel present two inversion-
ally related PC sets in measures 5–6 (ex. 3.13a). There are no pitches inverting
around the same axis (ex. 3.13b). In a case like this, we can view the inversional
relationship in terms of an operation TnI about the general axis C (0), and just
accept that this may not reflect the actual musical situation in every detail. The
expression “T4I” serves well to describe the progression in example 3.13a.
However, we can produce a more precise description of this fragment.
Another segmentation emphasizes the voice leading rather than the harmonic
entities. Example 3.13c shows that the progression outlined by the highest tones
of the flute (E♭5 and F5) is inversionally related to the progression outlined
by the lowest tones of the horn and the glockenspiel (F♯3 and E3). The axis of
inversion lies between E4 and F4. The highest tones of horn and glockenspiel
(B3 and B♭3) and the flute’s lowest (C5 and D♭5) form “inner parts” that invert
to one another about an axis between F4 and F♯4. Therefore, they seem to be
displaced upward by a semitone with respect to the “outer parts.” This reading of
the passage reveals a complex symmetry that is not apparent in example 3.13b.
Some models of inversional relationships can do more justice to the musical
context than TnI, as in David Lewin’s “Generalized Interval System” (GIS), an
algebraic structure representing various spaces of musical objects.23 The follow-
ing definition appeared as part of his presentation of the GIS:
23. For a formal definition of a GIS, see Lewin 1987, 26.

72 operations
(a) fl. (b)
hrn.
glsp.
(c)
Example 3.13
(a) Two measures from Anton Webern’s Piece for Orchestra, Op. 6, no. 3.
(b) The passage is based on a succession of two inversionally related PC sets. An axis
must bedefined for each pitch and its “inverse.”
(c) An alternative reading of these measures, stressing the symmetrical voice leading.
PI (5, 14) PI (5, 14)
PI (5, 9) PI (9, 14) PI (10, 5) PI (14, 10)
Example 3.14. Inversion applied to the division of a major sixth.
For each u and v in S (v may possibly equal u), we shall define an operation
Iv/u, which we shall call “u/v inversion.” . . .
. . . [W]e conceive any sample s and its inversion I(s) [. . .] as balanced about the
given u and v in a certain intervallic proportion. I(s) bears to v an intervallic rela-
tion which is the inverse of the relation that s bears to u. (Lewin 1987, 50)
S refers to “space,” the collection of objects of which u and v are members. It

should be noted that Lewin defined inversion with respect to a referential pair of
objects. This meant a generalization of the axis of inversion, smoothing out the
problem of its identification.
Imagining a referential pair of pitches or PCs improves our understanding of
inversion, and is more sensitive to the musical context than TnI. To show this,
I will modify my initial definition of inversion as an operation applied to pitch
intervals or PICs. My starting point is the division of an interval. For example,
it is possible to divide the major sixth (F4,D5)—i.e., PI(5,14)—into a major

operations 73
PI (5, 14) PI (5, 14)
PI (14, 16)
PI (5, 3)
Example 3.15a. Inversionally related pitch intervals stretching a referential interval

(major sixth).
PI (5, 14) PI (5, 14)
PI (5, 16) PI (14, 3)
Example 3.15b. Inversionally related pitch intervals breaking out of the referential
interval.
third and a perfect fourth ascending. Inversion divides the same major sixth
into a major third and a perfect fourth descending (ex. 3.14). To identify the
initial division it suffices to specify the dividing pitch by measuring the interval
separating it from one side of the major-sixth compass: PI(5, 5 + 4), or PI(5,9).
Then, for the inverted division we write: PI(14, 14 − 4), or PI(14,10). This means
that the inversion of the division of an interval can easily be stated in terms of
pitches. The pitches 9 and 10 are inversionally related with respect to the major
sixth (5,14). No axis is involved.
This concept of inversion is reminiscent of Bononcini and Fux, who applied
it to the Dorian octave (see ex. 3.8 above). Generally speaking, inversion is
defined with respect to a referential pitch interval—that is, an ordered pair of
pitches (x,y). To determine the inverse of a pitch interval (x, x + a), the latter
is measured from y in the opposite direction. As a consequence of this general
statement, inversionally related pitch intervals may stretch as well as divide the ref-
erential interval (ex. 3.15a). Furthermore, they can break out of it (ex. 3.15b).
The pitches x and y themselves are inversionally related of necessity. My initial
definition of inversion—that is, I(PI(x, x + a)) = PI(x, x − a)—is thus modified
to read:
IPI (x,y) (PI(x, x + a)) = PI(y, y − a)
where x may or may not be equal to y; and where a may be 0. When the opera-
tion is applied to single pitches the definition reads:
IPI (x,y) (x + a) = (y − a)
This definition adjusts to different situations. For example, when there is one
pitch obviously serving as an axis, y = x. This means that the referential pitch

74 operations
interval is a perfect unison (0). The second movement of Webern’s Opus 27

(see ex. 3.11) is a case in point. The referential pitch interval is (A4,A4), or
(9,9). For example, the pitch F5 (17, or 9 + 8) can be shown to invert to C♯4 (1)
as follows:
IPI (9,9) (9 + 8) = 9 − 8 = 1
Similarly, G♯4 (8, or 9 −1) can be shown to invert to B♭4 (10):
IPI (9,9) (9 − 1) = 9 + 1 = 10
When the axis lies between two adjacent pitches, x and y may represent these
pitches. Depending on the musical situation, they may represent another sym-
metrical pair as well.24 In the closing section of Schoenberg’s Opus 23, no. 3 (see
ex. 3.12), the best referential pitch interval is (C4,G4) or (0,7). It is contained in
all the sonorities. With respect to this pair, D♭5 (13) inverts to F♯3 (−6):
IPI (0,7) (0 + 13) = 7 − 13 = −6
Similarly, A♭5 (20) inverts to B2 (−13):
IPI (0,7) (0 + 20) = 7 − 20 = −13
When the operation applies to PCs instead of pitches, it is carried out modulo
12. Each represented pair of inversionally related PCs can be referential. Thus,
in the passage from Webern’s Opus 6, no. 3 (see ex. 3.13), the inversional rela-
tionships can be defined with respect to the PICs represented by the PC pairs
(E♭,D♭)/(3,1), (C,E)/(0,4), (B,F)/(11,5), and (F♯,B♭)/(6,10). For example,
if we take (E♭,D♭)—that is, (3,1)—as referential, the PC F (5) can be shown to
invert to B (11) as follows:
IPIC(3,1) (3 + 2) = (1 − 2) mod 12 = −1 mod 12 = 11
However, with respect to PIC(F♯,B♭) the same relationship is notated:
IPIC (6,10) ((6 + 11) mod 12) = (10 − 11) mod 12 = −1 mod 12 = 11
The alternative reading of the passage, shown in example 3.13c, considers

pitches as such, not as PC representatives. There are two sets of referential pairs:
24. David Lewin has pointed out that two objects of a system may invert to one
another with respect to various pairs, and that it is possible to consider each of these pairs
as referential (Lewin 1987, 52).

operations 75
{(E♭5,F♯3), (F♯3,E♭5), (F5,E3), . . . (E4,F4), . . .}, about which the “outer voices”
invert, and {(C5,B3), (B3,C5), (D♭5,B♭3), . . . (F4,F♯4), . . .}, about which the
“inner voices” invert.
In Luciano Berio’s 1967 tribute to Martin Luther King: O King, for voice and
five instrumentalists, there is an inversional relationship around the referential
pitch interval (F4,D5), which I used in example 3.14 and 3.15.25 The key feature
of this piece is a recurring melody of three phrases, which is sung on a set of
vowels gradually developing into the syllables of the subject’s name. The instru-
ments play along with bits and pieces of this melody, or they echo them softly.
Some notes receive a fortissimo accent. These accents, given by the piano and
one or more other instruments, and following one another ever more closely,
outline a large-scale projection of the melody.
The melody of O King employs seven PCs, all of which occur within the range
F4–D5 (ex. 3.16a). The remaining five PCs are represented by pitches spanning
the registers above and below the melody’s range (see ex. 3.16b). Initially, their
occurrences provide a faint accompaniment to the melody. The higher pitches
are G5 and E♭6. In the first half of the piece, they are played by the piano. In
the second half, the G5 is taken up by other instruments, and gradually becomes
more manifest—a process that continues until the climax (measure 4 after
rehearsal letter “F” in the score) and parallels the transformation of the vocal
sounds into the name “Martin Luther King.” The lower pitches are C4, E3 and
F♯2. After rehearsal letter “C” in the score, E, E♭ and F♯ begin to appear in
other registers as well. G and C, however, remain in theirs.
Now, the pitches G5 and C4—and, in the beginning, E♭6 and E3 as well—
are inversionally related with respect to the interval (F4,D5), which defines the
range of the melody of O King. This is a significant relationship, for the PCs G
and C mark the climax of the piece, the moment the dramatically sustained G5
of singer, flute and clarinet is suddenly contrasted with the C2 of the piano (a
pitch not heard before). The inversional relationship of G5 and C4 can only be
defined with respect to the interval (F4,D5), not only because it is a structurally
important pitch interval, but also because the melody’s pitches do not yield a
symmetrical subdivision of that interval. This means that the inversional rela-
tionship between G5 and C4 is not part of an overall symmetry.
There is not always a pair of inversionally related PCs that is so obviously
referential as the pair (F4,D5) in O King. A context-sensitive formula like
IPI (x,y) is of little use when the context does not give any clue to what x and y
might be. The TnI-protocol still has the advantage of providing a single refer-
ent for any inversional relation between PCs, but it may be at odds with the
musical context.
25. An orchestral arrangement of this piece serves as the second movement of Berio’s
Sinfonia (1968/1969).

76 operations
Example 3.16a. The central melody of Berio’s O King.

© 1970 by Universal Edition Ltd. London/UE 13781.
Example 3.16b. The initial projection of the remaining PCs. G5 and C4, and E♭6
and E3 are inversionally related with respect to the range (F4–D5) of the melody.
Multiplication
It is most convenient to define multiplication on PITCH first. This shows us what
the basic musical meaning of this protocol is. As said before, each member a
of PITCH corresponds to a pitch interval PI(0,a). Multiplication is conceived
as the enlargement of this interval by a factor n. PI(0,a) thus transforms into
PI(0,(n · a)). This protocol can be applied to pitches and sets of pitches. For
example, the multiplication of the pitch value 5 (or F4) by a factor of 7 yields
the pitch value 35 (B6). The multiplication of the pitch set {1,−3} (or {D♭4,A3})
by a factor of 2 yields {2,−6} ({D4,F♯3}). The multiplication of {−3,4,−5} (or
{A3,E4,G3}) by a factor of 3 yields {−9,12,−15} ({E♭3,C5,A2}). And the multi-
plication of {3,−2,−7} (or {E♭4,B♭3,F3}) by a factor of −4 yields {−12,8,28} (or
{C3,A♭4,E3}; ex. 3.17).
Applying multiplication with respect to PITCHCLASS means carrying it out
modulo 12. As a consequence, the relationships are no longer obvious. Multipli-
cation is thus abstracted from a basic musical intuition in much the same way as
transposition and inversion. It is important to note that not all factors n (mod
12) result in a one-to-one mapping, which is a prerequisite for an operation as
defined by PC set theory (table 3.3). For example, when the factor is 6, all twelve
PCs are mapped onto two PCs: 0 and 6 (representing the tritone). When the fac-
tor is 3 or 9, the twelve PCs are mapped onto four PCs: 0, 3, 6, and 9 (represent-
ing the diminished seventh chord). And when the factor is 2 or 10, the twelve
PCs are mapped onto six PCs: 0, 2, 4, 6, 8, and 10 (representing the whole-
tone scale). The only factors that do effect a one-to-one mapping on PITCH-
CLASS are 1, 5, 7, and 11. Of these, 1 and 11 are not commonly used, since
multiplication by the former maps each element onto itself (like T0), and multi-
plication by the latter maps each element onto its inverse mod 12 (like I). What

operations 77
5 1 4 3
-3 -3 -5 -2 -7
x7 x2 x3 x-4
35
2 12 8 28
-9 -15 -12
-6
Example 3.17. Multiplication in PITCH.
Table 3.3. The multiplication (mod 12) of PC values by factors from 1 through 11
Times
1 2 3 4 5 6 7 8 9 10 11
PC 0 0 0 0 0 0 0 0 0 0 0 0
1 1 2 3 4 5 6 7 8 9 10 11
2 2 4 6 8 10 0 2 4 6 8 10
3 3 6 9 0 3 6 9 0 3 6 9
4 4 8 0 4 8 0 4 8 0 4 8
5 5 10 3 8 1 6 11 4 9 2 7
6 6 0 6 0 6 0 6 0 6 0 6
7 7 2 9 4 11 6 1 8 3 10 5
8 8 4 0 8 4 0 8 4 0 8 4
9 9 6 3 0 9 6 3 0 9 6 3
10 10 8 6 4 2 0 10 8 6 4 2
11 11 10 9 8 7 6 5 4 3 2 1
remains, then, are the factors 5 and 7. For these, the same is true as for any pair
of factors n and 12 − n in table 3.3: M5(a) is the inverse mod 12 of M7(a).
The multiplication of a PC set A by a factor 5 is designated M5(A). When this
operation is applied to the PC set {0,2,3,7}, the latter transforms into {0,10,3,11}:
M5({0,2,3,7}) = {(0 · 5) mod 12, (2 · 5) mod 12,

(3 · 5) mod 12, (7 · 5) mod 12} = {0,10,3,11}

78 operations
When the same PC set is multiplied by a factor 7, the result is {0,2,9,1}:
M7({0,2,3,7}) = {(0 · 7) mod 12, (2 · 7) mod 12,

(3 · 7) mod 12, (7 · 7) mod 12} = {0,2,9,1}
Since M7(A) is the inverse of M5(A), most authors use the labels M (denoting
M5) and IM (denoting M7). Although I will follow this practice, I will keep the
indexed labels in reserve for convenient reference to the history of the PC set
multiplication.
Interval expansion, the intuitive basis of PC multiplication,26 plays a construc-
tive role in works of Béla Bartók and Alban Berg. In his Harvard Lectures of
1943, Bartók described how he used it:
The working with these chromatic degrees gave me another idea, which led to
the use of a new device. This consists of the change of the chromatic degrees
into diatonic degrees. In other words, the succession of chromatic degrees is
extended by leveling them over a diatonic terrain. (Bartók 1976, 381)
Thus, the successive intervals of the chromatic fugue theme of Bartók’s Music
for Strings, Percussion, and Celesta (PI 1,3,−1,−1,−2,1, etc.) are transformed into
diatonic ones, as example 3.18 shows.27 In the Third String Quartet, such an
“extension in range” (as Bartók called it) connects the opening held chromatic
tetrachord {C♯,D,D♯,E} and the closing fifths chord {C♯,G♯,D♯,A♯} (Antokoletz
1993, 260). When we assign the value 0 to C♯, which is the bass common to both
sonorities, (C♯,G♯,D♯,A♯) is the M7-transform of (C♯,D,D♯,E).
In the orchestral introduction of Berg’s orchestral song Seele wie bist du
schöner—the first of a song cycle on picture postcard texts by Peter Altenberg, Op.
4—each instrumental part, or group of instrumental parts, provides sequences
of a different motif. In some instrumental parts the sequential intervals increase.
The first violins, piccolo, glockenspiel, and xylophone repeat their motif at the
PIs 1, 3, 6, and 9, and the second violins and flutes repeat theirs at the PIs 1, 3,
6, 10, 15, 21, and 28.
26. Pierre Boulez provided a different musical interpretation of the concept of multi-
plication, defining it as an increase of number rather than size. Very basically, Boulez “mul-
tiplies” a PC set A—say, a segment of a series—by a PC set B by transposing A to the level
of each pitch element of B (Boulez 1963, 35). The product AB is thus a complex consist-
ing of as many transpositions of A as there are elements in B. Boulez’s idea started to
reverberate in the American music-theoretical discourse in the 1990s through the work
of, among others, Heinemann (1998).
27. Michael Friedmann (1990, 118) used the somewhat confusing term “modal
transposition” to denote this technique of musical derivation. In Bartók’s Music for
Strings, Percussion, and Celesta, diatonic themes are similarly transformed into chro-
matic ones.

operations 79
Andante tranquillo ( = 116)

1 2
etc.
Molto moderato ( = 144)
etc.
molto espress.
Example 3.18. An instance of interval expansion: Bela Bartók, Music for

Strings, Percussion and Celesta, first movement, mm. 1–5, and fourth movement,
mm. 204–9.
However, the actual antecedent of PC multiplication was a twelve-tone tech-

nique that Berg employed in his unfinished opera Lulu. The music critic Willi
Reich—Berg’s friend and, later, biographer—described this technique in an
article which appeared in the Musical Quarterly shortly after the composer’s
death. Berg had derived a new series from the prime series-form by taking every
seventh tone (starting with the first). Each tone with order number n—that is,
the nth note after the first28—was thus replaced by the tone with order number
(n times 7) mod 12. Taking every fifth tone of the prime series created another
new series. Each tone with order number n was thus replaced by the tone with
order number (n times 5) mod 12 (ex. 3.19a; Reich 1936, 394–95). Both opera-
tions yield the total chromatic, since the lowest common multiples of 5 and 12,
or 7 and 12, are their products.
Berg multiplied 5 or 7 by the order numbers of the tones in his series. Whose
idea was it to use their PC numbers instead? In 1936, Ernst Krenek, who was
deeply involved in a study of the relationship between music and mathematics,
described the same operations in one of six lectures he delivered in Vienna.
These lectures were published shortly thereafter under the title Über neue Musik.
When introducing the “five-series” (Fünferreihe) and “seven-series” (Siebenerreihe)
of a given twelve-tone series (Grundreihe), Krenek did not refer to Alban Berg or
Willi Reich, but to one “Ingenieur [engineer] Konrád in Brünn [i.e., Brno],”
who apparently had done some research in the field of twelve-tone serialism that
provided a slightly different perspective on the subject. Krenek pointed out that
the “n-times-five” and “n-times-seven” operations, when applied to the chromatic
scale, resulted in a cycle of fourths and a cycle of fifths, respectively (ex. 3.19b).
28. As a consequence, the first tone should have the number 0. Babbitt (1992, 1960),
too, let order numbers run from 0 through 11.

80 operations
etc.
etc.
Example 3.19a. Two order operations on the prime series-form of Alban Berg’s
Lulu, as demonstrated by Willi Reich.
Willi Reich, “Alban Berg’s Lulu,” Musical Quarterly 22/4 (1936): 394–95. Used by
permission of Oxford University Press.
8va
9 etc.
5
7
14 etc.
Example 3.19b. The transformation of a chromatic scale into cycles of fourths and
fifths by the same operations (Krenek 1937, 77). The “9” should be replaced with a
“10” (A♯).
f
g
Example 3.19c. The same transformations as in (b), now presented as the result of
mirror operations (Eimert 1950).
© Breitkopf & Härtel, Wiesbaden. Used with kind permission of the publisher.
Conversely, these one-interval cycles transformed into a chromatic scale—

another one-interval cycle—under the same operations (Krenek 1937, 77).
The next step was to replace each tone, as numbered according to its posi-
tion in the chromatic scale, by its cycle-of-fourths or cycle-of-fifths correspon-
dent irrespectively of its order position in the actual twelve-tone series. This step
was taken by the German composer and music theorist Herbert Eimert, who
sanctioned these operations by including them in his manual on twelve-tone
composition of 1950 (Krenek had not mentioned them in his Studies in Counter-
point). With a keen sense for historical connections, Eimert spoke of the “eight
modes” of a twelve-tone series. He presented all these “modes” as mirror forms
of one another. The inversion was obtained from the prime form through a
horizontal mirror. The retrograde and retrograde-inversion reflected the prime

operations 81
form and the inversion through a vertical mirror. Pursuing the analogy further,
Eimert described his “cycle-of-fourths-transform” (Quartverwandlung) and “cycle-
of-fifths-transform” (Quintverwandlung) as resulting from a slanting mirror:
Furthermore, one can sort of move the mirror at an angle, that is, the “angle”
of a fourth or fifth, so that the chromatic row is reflected in both cycles . . . In
this way, one obtains the cycle-of-fourths transform and the cycle-of-fifths trans-
form of the row.29
Eimert’s illustration is shown in example 3.19c. In turn, one could project the
cycle-of-fourths and cycle-of-fifths transforms through a vertical mirror, obtain-
ing their retrogrades. The mirror analogy served as a common denominator of
all the twelve-tone operations, but it was hardly enlightening in the case of these
new operations. (What is “the angle of a fourth”?) Eimert added instructions for
realizing the cycle transforms, which involved the concept of multiplication:
In general, the procedure is like this: to transform G [Grundreihe, i.e., the prime
series-form] into IV [“cycle-of-fourths-transform”] or V [“cycle-of-fifths-trans-
form”], one multiplies the interval numbers with 5 or 7, respectively. When the
products are larger than 12 . . ., their difference with the next-smaller number
in the multiplication table of 12 should be determined.30
Thus defined, the multiplication operations came to the notice of Princeton-

based composers and music theorists. James K. Randall, Godfrey Winham, and
Hubert S. Howe were the first to discuss and adopt them, not only with regard
to twelve-tone series. At the time, the unordered PC set had just established itself
in the literature, although different names were in use to denote it. In his 1965
paper “Some Combinational Properties of Pitch Structures,” published in Per-
spectives of New Music, Howe defined the multiplication operation with respect to
a “PC collection.” (Others, including Forte, used to call this a “pitch set”). Apart
from this broader field of application, Howe’s definition was very close in mean-
ing to Eimert’s. He wrote:
By a multiplicative operation Mn [the correct notation would be Mn] performed

upon a PC coll x, we mean the substitution, for each PC m in x, of the PC
obtained by multiplying m by n. (Howe 1965, 50)
29. “Ferner kann man den Spiegel gewissermaßen in Winkelstellung bringen, und
zwar in den ‘Winkel’ der Quarte und der Quinte, so daß sich die chromatische Reihe in
den beiden Zirkeln spiegelt . . . Auf diese Weise erhält man die Quartverwandlung und
die Quintverwandlung der Reihe.” (Eimert 1950, 29, my translation)
30. “Generell sieht das Verfahren so aus: Um G in IV bzw. V zu verwandeln, multipli-
ziert man die Intervallziffern mit 5 bzw. 7. Bei den Zahlen, die größer sind als 12 . . . muß
man die Differenz zur nächstniedrigen Zahl der Multiplikationsreihe von 12 feststellen.”
(Eimert 1950, 31, my translation)

82 operations
adding in a footnote that
“n” is a modulo 12 integer. (ibid., note 19)
Howe defined inversion as a multiplicative operation. He mapped a PC set

onto its inverse by M11. Thus, it was possible to present a coherent system of
four multiplicative operations similar to Schoenberg’s canon of twelve-tone
operations: M1 (“identity”), M11 (“inversion”), M5 (“cycle-of-fourths equiva-
lence”), and M7 (“cycle-of-fifths equivalence”).31 Each composition of these
operations equaled a single operation: M11M5 = M7, M7M5 = M11, M5M5 = M1,
M7M11M5 = M1, etc.32
Judging from his references, Howe had borrowed much from an unpub-
lished monograph by Randall. This monograph, entitled Pitch-Time Correlation,
was completed in 1962.33 In it, Randall had listed several properties of PC multi-
plication. For example, he pointed to the identical behavior of PCs and interval
classes under this kind of operation. When two PCs a and b are multiplied by n,
so is the value of the interval class associated with a and b:
PIC(M(a,b)) = M(PIC(a,b))
A transposition of the same pair does not affect the PIC, while under inversion it is
the APIC that remains invariant. Multiplication, however, affects PCs and APICs
alike. More specifically, it causes the APICs 1 and 5 to swap with one another:
PIC(M(0,1)) = PIC(0,5) = 5; PIC(M(0,5)) = PIC(0,25) = 1

PIC(M(6,7)) = PIC(30,35) = 5; PIC(M(−2,3)) = PIC(−10,15) = 1
This was a reason for Winham (1964, 110–11; 1970, 63–67) to question the
inclusion of multiplicative operations in the set of canonical twelve-tone opera-
tions. These operations did not preserve the “succession of interval classes” that
Carl Dahlhaus considered fundamental to a twelve-tone series, the property
remaining invariant under inversion, retrogradation, and any combination of
these operations (see chapter 2, p. 38). In view of the unordered PC set, it was
argued that multiplication could cause a change of interval content. In other
words, it could affect the “sound” that all realizations of the PC set, its transposi-
tions, and the transpositions of its inversion supposedly had in common.
31. Howe’s terminology.

32. The system exhibits the properties of a transformation group. More on this in
chapter 4.
33. According to a reference in one of the author’s published papers (Randall 1965,
92).

operations 83
On the other hand, certain PC sets map onto themselves under M, while this
operation equals a transposition or transposed inversion when applied to other
PC sets. These varying results are dependent on the original set’s interval con-
tent. If the APICs 1 and 5 are equally represented, M yields a PC set with the
same interval content. In a relatively small number of cases, this new PC set is not
a transposition or a transposed inversion; it just has the same interval content. It
has been a special attraction of PC multiplication that it provides a link between
some of these intriguing “Z-related” pairs of PC sets. John Rahn’s manual Basic
Atonal Theory made explicit mention of this property.34
The inclusion of a section on PC multiplication in a college textbook like
Rahn’s—Forte had ignored the topic in The Structure of Atonal Music seven years
earlier—was a response to the interest certain American composers had been
taking in the operation, rather than a token of its success as an analytical tool.
The composer Robert Morris had adopted multiplication as a basic twelve-tone
operation in a study he had written with the computer programmer Daniel
Starr. This study (“A General Theory of Combinatoriality and the Aggregate”)
was published in two installments in Perspectives of New Music, in 1977 and 1978.
Starr and Morris wanted to enlarge “the combinatorial possibilities of the
[twelve-tone] row.” In other words, they wanted to gain a larger number of
derivatives of the row that could be combined with it without doubling PCs.
(This was an issue of importance in twelve-tone theory; see also chapter 4.)
They did not worry about the possibly less audible relationship between a PC
set and its M-transform. As they argued, the “classical” twelve-tone operations
were equally abstract, since they were applied to PCs; why should a melodic
contour only expand or contract under multiplication, if it may dissolve
entirely under transposition or inversion (Starr and Morris 1977/1978, 5)?
Although this was a valid argument in itself, it wrongly suggested that audibil-
ity was a criterion of relatedness in “classical” twelve-tone serialism. From a
Schoenbergian viewpoint, the only possible objection one could raise against
multiplication concerned the changing order of interval classes. This objec-
tion was of a conceptual rather than an empirical nature.
34. See Rahn 1980, 104; “Z-related” pairs will be discussed in chapter 4.

Chapter Four
Equivalence
The operations discussed in chapter 3 are generalized representations of com-
positional techniques whereby PC sets are derived from each other. However,
not all relations between PC sets are based on derivation. This chapter and the
following ones will deal with other relations, seen within a historical framework.
In this chapter I will discuss the evolution of the concept of PC set equivalence.
Like the term “operation,” the term “relation” evokes the world of math-
ematics. In mathematics, more specifically in algebra, a relation is commonly
conceived as an open sentence, designated P, connecting the elements of two
collections, S and T. This open sentence is true or not true for each ordered pair
of these elements. If s is an element of S, and t is an element of T, a relation can
be defined as the collection of ordered pairs (s,t) for which the open sentence
P is true. This is a subcollection of the entire collection of these pairs. A relation
like this is called a relation “from S to T.”
We will deal here with the special case in which S = T. Then, the relation
is called “a relation in S.” In chapter 2, we defined the relation “is congruent
modulo 12 with” for the collection of pitch values under twelve-tone equal tem-
perament (PITCH). This relation singles out specific ordered pairs of pitch val-
ues from this collection, such as (14, −10), (5,29), or (10, −2). For these ordered
pairs the relation obtains: 14 is congruent modulo 12 with −10, 5 is congruent
modulo 12 with 29, and 10 is congruent modulo 12 with −2.
Definitions of Equivalence
The relation “is congruent modulo 12 with,” when it is defined on PITCH, is

what algebra refers to as an “equivalence relation.” In chapter 2, we have seen
that this relation satisfies all three conditions of equivalence: reflexivity, symme-
try, and transitivity. The concept of equivalence helps us bring large collections
of musical elements down to reasonable proportions. PITCH can be mapped
onto PITCHCLASS; and the collection of PC sets (PCSET) can be compressed
by defining equivalences on the basis of certain transformations or common
properties, as this chapter will show.
This may sound rather trivial. It is common practice to assign equal names
or labels to pitch values that are separated by one or more octaves, or to a
chord and its transpositions, or to the transpositions and inversions of the

equivalence 85
subject of a fugue. Indeed, an informal notion of equivalence has always

been part of music theory and analysis. PC set theory, however, has adhered
to formal definitions of equivalence.
Many such definitions invoke the three conditions mentioned above. How-
ever, the same kind of relation can be stated in another way, as follows. Equiva-
lent elements in a collection form subcollections that are called “equivalence
classes.” In turn, these equivalence classes form a “partition” of the original col-
lection. This means two things: first, the sum of all equivalence classes exhausts
the original collection; and, second, two equivalence classes do not share any
elements. Thus, a relation in S that partitions S is an equivalence relation in the
algebraic sense.1
Which relations between PC sets satisfy this concept of equivalence? Theorists
have answered this question in different ways. It has always been obvious that
they did not want to apply the concept whenever it was possible, but consid-
ered it appropriate to specific musical relations. In the 1960s it became custom-
ary to call pitch-class sets “equivalent” when they were related as transpositions
or transposed inversions, which may reflect musical intuitions molded by tra-
ditional training. However, does this use of “equivalence” satisfy the algebraic
definition of equivalence? It does, when the relation is defined in PCSET, and is
cast in the sentence: “PC set A is transpositionally related to PC set B.” This rela-
tion is reflexive, for any PC set is transpositionally related to itself by T0. It is also
symmetrical: when A is transpositionally related to B by Tn, B is transpositionally
related to A by T−n mod 12. And it is transitive: when A is transpositionally related
to B by Tn, and B is transpositionally related to C by Tm, then A is transposition-
ally related to C by T(n + m) mod 12.
Once we specify an index of transposition—that is, a PIC, represented by n—
these three properties vanish. The sentence “PC set A is related by the operation
Tn to PC set B”—in formal notation: “B = Tn(A)”—does not imply that A and B
might be equal. This means that the relation is not reflexive. Furthermore, the
sentence cannot be reversed; in other words, the expression “B = Tn(A)” does
not imply that A = Tn(B). This is the case only when n = −n mod 12 (that is, when
n = 0 or 6). Therefore, the relation is not symmetrical. Nor is it transitive: when
B = Tn(A), and C = Tn(B), it cannot be stated that C = Tn(A). This is the case
only when (n + n) mod 12 = n (that is, when n = 0).
In chapter 2, we already have seen that a relation between inverse-related
PICs does not satisfy the three conditions of algebraic equivalence. Since PC
values correspond to PIC values, the same holds true for PC sets and their inver-
sions. The relation “is the inversion of,” as defined in PCSET, is not reflexive;
most PC sets are not their own inversions. It is symmetrical: B = I(A) implies that
1. This partitional view of equivalence can be traced back to Richard Dedekind’s edi-
tion of Dirichlet’s papers on number theory (cf. Lejeune Dirichlet 1871, 136), and even
further to Carl Friedrich Gauss’s Disquisitiones arithmeticae (cf. Gauss 1863, 222–23.).

86 equivalence
A = I(B). However, it is not transitive. When B = I(A), and C = I(B), it follows

from the symmetrical property of the relation that C = A; and since the relation
is not reflexive, C ≠ I(A).
Transpositional and inversional relations between pitch-class sets do, how-
ever, satisfy a concept of equivalence defined with respect to relations between
sets. According to this set-theoretic definition, two sets A and B are equivalent
if there is a protocol defining a one-to-one correspondence between them—in
other words, if sets A and B have the same number of elements, and each ele-
ment of A is assigned to a different element of B. The relation “A is equivalent to
B” is a possible relation in a family of pitch-class sets. What is more, it defines an
equivalence in that family—that is, an equivalence in the algebraic sense—for it
is reflexive (A is equivalent to itself), it is symmetrical (the sentence “A is equiva-
lent to B” implies that B is equivalent to A), and it is transitive (the sentences “A
is equivalent to B” and “B is equivalent to C” imply that A is equivalent to C).
It may seem odd to say that the relation “A is equivalent to B” is itself an
equivalence. However, this statement involves two different meanings of the
term “equivalence,” one being a specific relation between sets (set-theoretic
equivalence), and the other a property of that relation (algebraic equivalence).
To summarize, Tn and I define relations between PC sets that are equivalence
relations in the set-theoretical sense only. The set-theoretic equivalence relation,
in turn, defines an algebraic equivalence relation in PCSET.
Despite its aura of precision, the term “equivalence” can cause confusion. It
has different mathematical meanings, and these meanings have been subject to
change. The partitional view of algebraic equivalence gained general currency
relatively recently, that is, in the 1930s and 1940s, through the algebra textbooks
of Van der Waerden (Moderne Algebra 1930/1931), Albert (Modern Higher Algebra,
1937), and Birkhoff and Mac Lane (A Survey of Modern Algebra, 1941).2 And com-
mon-sense notions of musical equivalence must be taken into account as well.
In view of this complex etymology, it is not surprising that music theorists have
different concepts of equivalence, as I will show later. It is easy to wave this prob-
lem aside, reasoning that we are dealing with music, and that musical notions
of equivalence differ from mathematical ones. Mathematical definitions can be
adjusted to musical conditions, provided that this is stated explicitly. Therefore
it is perhaps unfair to judge definitions of musical relations by purely mathe-
matical standards. However, if this is unfair, then why have these standards been
2. The concept of an equivalence class is described but not named in these books.
In an unfinished paper on the history of this concept, David Fowler (University of War-
wick, UK) has reported that the English term “equivalence class” appeared in print for
the first time in November 1941, in an article by Samuel Eilenberg and Saunders Mac
Lane (“Infinite Cycles and Homologies,” Proceedings of the National Academy of Sciences of the
United States of America, 27/11, 535–39). I am grateful to the late Dr. Fowler for sharing his
information with me.

equivalence 87
invoked by PC set theorists? Mathematics served them as a model of clarity, as a

badge of academic respectability, or as an authority encompassing music as well as
many other fields of human knowledge and activity. This is why, in a study of the
evolution of PC set theory, we have to deal with these mathematical definitions.
Equivalence and Equality
Milton Babbitt’s definition of equivalence (1946) at first sight seems peculiar

and arbitrary:
Two sets, P and P', will be considered equivalent if and only if, for any pi,j of the
first set and any p'i',j' of the second set, for all i’s and j’s, if i = i', then j = j'. (=
denotes numerical equality in the ordinary sense). (Babbitt 1992, 8–9)3
In Babbitt’s vocabulary, a “set” was a twelve-tone series. As he saw it, the equiva-
lence of two such series stipulated that they showed the same succession of PCs
(or “js”).4 This means that he considered the term “equivalent” as synonymous
with “equal.” Why did he speak of equivalence here? Perhaps he was guided by
musical considerations. A twelve-tone series can take many different shapes.
Therefore, a music theorist or analyst may prefer a word other than “equal” to
account for the relation between two realizations of it, a word referring to a
shared substance while providing scope for variation. “Equivalent” may have
been that word for Milton Babbitt. It is noteworthy that he reserved it for such
an exclusive relation. Apparently, he did not yet think of transpositional or inver-
sional relations as “equivalence” relations.5
Babbitt’s use of the term “equivalent” is informal, in spite of the formal air his
prose assumes (NB his use of sentential equivalence: “if, and only if”). However,
it does not require much effort to “improve” his definition, to fill the gap that
frustrates its full validation in terms of mathematical accuracy. From an algebraic
3. In this passage “P'” does not mean “the complement of P.”

4. When two series showed the same succession of “interval numbers,” or PICs, Babbitt
called them “dependent.” “Dependent” series are necessarily related as transpositions.
5. This betrays a much more Schoenbergian view of twelve-tone serialism than is usu-
ally ascribed to Babbitt. He is known as a composer and theorist less inclined than Schoen-
berg to make hierarchical distinctions between the various forms of a twelve-tone series.
He articulated this position in a review of René Leibowitz’s Schönberg et son école: “The func-
tionality of a twelve-tone composition is defined by the specific twelve-tone set. A functional
norm is stated, and deviations from this norm appear; but there is no degree of deviation,
no hierarchy of deviations such as presented in tonal music, to make possible progress and
growth—stated in terms of the functional context—through various stages of compositional
expansion.” (Babbitt 1950, 59) Characteristically, in Babbitt’s compositions the “functional
norm”—i.e., the prime form—is not necessarily stated at the beginning.

88 equivalence
perspective, equality is an equivalence relation. This is what Babbitt may have

had in mind when writing the passage quoted above. It renders clear what is now
so easily overlooked: that it is possible to conceive other equivalencies than the
equality relation. For example, “giving F the same order position” is a relation
that is reflexive, symmetrical, and transitive in the collection of all twelve-tone
series, and can thus be called an equivalence. On that basis, (B♭,E,F♯,D♯,F,A,D,
C♯,G,A♭,B,C)—the prime series-form of Schoenberg’s Variations for Orchestra,
Op. 31—is equivalent to (C,B,D,C♯,F,F♯,E♭,E,A♭,G,B♭,A)—the prime series-
form of Anton Webern’s String Quartet, Op. 28.
Equivalence, in the algebraic sense, is only a value assigned to a relation; it
is not the relation itself. Moreover, the equivalence of two things is contingent
upon a specific sentence expressing their relationship. This sentence does not
actually express the relationship, but is an integral part of it. For example, in the
collection of all people the relation “a is the sister of b” is not an equivalence,
since it is not reflexive (nobody is her own sister) or symmetrical (one may be
the sister of a brother). However, the phrase “is the sister of” can be replaced by
“has the same parents as.” Then, the relation between a and b fully complies with
the algebraic definition of equivalence.
Babbitt’s later writings confirm that he did not care too much about the exact
meaning of the term “equivalent.” In “Twelve-tone Invariants as Compositional
Determinants” (1960) he rightly equates “octave-equivalence” with “congruence
mod.12” (248), but in footnote 3, which discusses the fundamentals of group
structure, the term “equivalence relation” is applied to equations (249). In “Set
Structure as a Compositional Determinant,” an article from 1961, Babbitt calls
complementary intervals “equivalent,” in a sentence that must have inspired
Donald Martino to develop the APIC vector (see chapter 2, p. 47).6 Speaking of
the relations between complementary hexachords, he defines the “total interval-
lic content” of a hexachord as: “the 15 intervals . . ., specified most conveniently
in the form of the six possible intervals—since complementary intervals are
equivalent—with their associated multiplicities.” (Babbitt 1961a, 80). The use
of the word “equivalent” in relation to complementary (that is, inverse-related)
intervals is deserving of criticism, which was expressed in chapter 2.
For the next step in the evolution of the concept of PC set equivalence we
turn to an article from 1963 by Allen Forte, entitled “Context and Continuity
in an Atonal Work: A Set-Theoretic Approach.” This article—which presents an
analysis of Schoenberg’s Six Little Piano Pieces, Op. 19—was the first to make
explicit reference to set theory as an analytic tool, specifically for the non-serial
atonal music written by Schoenberg between approximately 1908 and 1921.
Forte began his argument by defining a universal set of twelve PCs—numbered
from 0 (C) to 11 (B)—and specifying several subsets that he considered sig-
6. Martino paid tribute to Babbitt’s articles, including this one, as main sources of
inspiration for his contribution to PC set theory. (Martino 1961, 224)

equivalence 89
nificant in Opus 19 (ex. 4.1). The abstract relations between these subsets were
his first object of study. He described them in terms of naive set theory, identify-
ing sets as the subsets or complements of other sets, or as the result of set union
or set intersection (see chapter 3, p. 49). Set theory meant to him a “system,” a
common tool for understanding all atonal music. His deductive approach was
clearly inspired by Milton Babbitt’s formulation of the (serial) twelve-tone sys-
tem. And it has become typical for the generations of American theorists after
Forte’s to speak of musical structure as the realization of a coherent set of gen-
eral principles. The system that Forte described in 1963 seems exclusively based
on his analysis of Schoenberg’s Opus 19. Forte did not refer to a totality of PC
sets extending beyond the ones he had selected from the score of this work. The
deductiveness of his argument was largely a matter of rhetoric; in fact, the argu-
ment was circular.
The term “equivalent” appears in the first section of Forte’s article, and is
taken by him to mean “identical” (Forte 1963, 76). Indeed, Forte considered two
subsets as equivalent when they consisted of the same elements. In such a case,
mathematical set theory speaks of the “equality,” not the “equivalence,” of sets.
The critique that can be leveled against Babbitt’s definition of equivalence holds
true for Forte’s as well. In any collection of sets, equality is, of course, an equiva-
lence relation; but it is pointless to say, without further qualification, that two
sets are “equivalent.” In example 4.1, the PC sets A4 and B5 are equivalent on the
basis of their identity relation (“A4 is equal to B5”). However, other equivalence
relations can be defined as well. For example, the PC sets A1 and B4 are equiva-
lent on the basis of their equal cardinal numbers (“A1 has the same number of
elements as B4”), and so are A2, B1, and B2.
The Study of Interval Content
Less than two years separate “Context and Continuity” from “A Theory of Set-
Complexes for Music” (1964), a period of mushrooming growth in PC set theory.
The system described in the latter article is much more elaborate than the one
presented in the former, and—above all—it is much more general in scope.7
Forte maps out a large number of possible musical relations. It is significant that
the role of music analysis, the study of one or more compositions, is considerably
reduced. The analytical section covers only five pages out of forty-two. In 1963, it
was five pages out of eleven. Meanwhile, the proposed field of analytical inquiry
exceeds the early atonal repertoire of the Second Viennese School, Forte’s main
focus in the earlier article. Pointing out the relevance of the PC set, Forte men-
tions the music of Liszt, Debussy, and Scriabin before that of Schoenberg and
7. Forte acknowledged that “it was the deficiencies of [the] analysis [in “Context and
Continuity”] which led to the present article” (Forte 1964, 182, note 18).

Example 4.1. PC sets extracted from Schoenberg’s Six Little Piano Pieces, Op. 19. An example from Allen Forte’s article “Context
and Continuity in an Atonal Work: A Set-Theoretic Approach.”
Perspectives of New Music, 1/2 (Spring 1963): 74–75. Used by permission of Perspectives of New Music.
10/2/2008 10:30:57 PM
equivalence 91
his pupils.8 He also refers to precedents in theoretical writings by François-

Joseph Fétis, Carl Friedrich Weitzmann, Arnold Schoenberg, Bernhard Ziehn,
and Charles Ives (Forte 1964, 136–37). Thus, he places the use of set theory in
music analysis historically, with reference to a variety of musical styles and theo-
retical concepts from the past.
Strangely enough, the musical examples do not reflect the broadening of
Forte’s perspective. Forte took them from the fourth of Anton Webern’s Five
Movements for String Quartet, Op. 5, one of the most accessible pieces from
the Viennese early atonal repertoire.9 During this early phase of its develop-
ment, PC set theory was still marked by a major commitment to the works of
Schoenberg and his pupils, notwithstanding Forte’s lip service to other compos-
ers. In 1965, shortly after the publication of “A Theory of Set-Complexes,” Forte
read a paper at a New York conference on computer applications in musicology,
discussing the implementation of his theory in computer-aided analysis. In this
paper—with the same title as his later book, The Structure of Atonal Music—he
was concerned with “a structural description of the revolutionary and enigmatic
non-tonal music of Schoenberg, Berg, and Webern composed during the period
1908–23, so-called atonal music” (Forte 1970, 10). The example he used was
Webern’s Three Pieces for Cello and Piano, Op. 11.10 The actual mission was
still the one he had stated in 1963, in “Context and Continuity”: to show that the
non-serial atonal music of the Second Viennese School satisfied criteria of musi-
cal logic and coherence as much as tonal or serial music.
Forte’s claim that the concept of a PC set could be applied to other reper-
toire may have been gratuitous at the time, but he had in any case compiled an
ingenious list of PC sets representing all possible combinations of tones under
twelve-tone equal temperament without regard to functional context. He had
reduced the sum total of 4,096 PC sets (see chapter 2, p. 47) by partitioning it into
8. The idea that these composers applied “non-tonal sets” (structurally significant
groupings of tones that do not constitute tonal chords) in their music, which can thus
be related to the atonal and serial repertoire of the Second Viennese School, had prob-
ably been suggested to Forte by George Perle’s book Serial Composition and Atonality, to
which he casually refers. Perle, in his discussion of the work of composers like Debussy,
Scriabin, and Roslavets, speaks of “nondodecaphonic” sets (structurally significant group-
ings of tones that are not part of a twelve-tone series). I believe that both authors refer to
the same class of things (structurally significant groupings of tones that are neither tonal
chords nor twelve-tone series), which they view from different historical perspectives.
9. Forte was not the first to deal with this piece analytically. For example, Perle—
again—had included an analysis of it in his Serial Composition and Atonality.
10. None of the important papers following on Forte’s—papers by Howe (1965),
Teitelbaum (1965), Clough (1965) and Chrisman (1971)—dealt with music from another
repertoire either, if they dealt with other works than Webern’s Opus 5 at all. Howe and
Teitelbaum examined this work once again; the papers of Clough and Chrisman discussed
conceptual issues without reference to written music.

92 equivalence
200 equivalence classes, or set-classes (example 4.2 shows the classes of cardinal
numbers 5 and 7). This time, Forte’s definition of equivalence fully complied
with the algebraic standard:
We now define an equivalence relation for arbitrary pitch-sets A and B. Let

v(A) represent the interval-vector of set A, and v(B) the interval vector of set
B. Then,
A ≡ B if and only if
v(A) = v(B)11
By “interval vector” Forte meant the APIC vector. He claimed that “the defined
equivalence [had] the reflexive, symmetric, and transitive properties” (Forte
1964, 143). This is true. The equal-vector relation obtains when the PC sets A
and B are equal; it also obtains when they are reversed; and if B has its APIC
vector in common with a third PC set C, then so does A. Ironically, it was this
new, sound basis of algebraic equivalence that would soon come under attack,
and would be withdrawn by Forte afterwards. This is not difficult to understand.
First, however, it has to be explained why Forte had decided to consider interval
content the prime distinguishing property of a PC set.
As we have seen, Lewin (1960), Babbitt (1961a), and Martino (1961) had
been pondering a way to encode the total interval content of a PC set. The issue
cropped up in relation to the serial technique of Arnold Schoenberg, especially
his use of hexachords. Schoenberg’s twelve-tone series are usually so constructed
that the two halves—the hexachords—have no PCs in common with their
inverted forms at some transpositional level. Each hexachord and its transposed
inversion thus fill out the total chromatic. In Babbitt’s words, they constitute the
twelve-tone “aggregate.”12 This is only possible if the first six PCs of a twelve-tone
series form a set that has a TnI-related complement. Then, reordering this com-
plement, i.e., the last six PCs, yields the first half of a transposed inversion of the
series (ex. 4.3a, inside the boxes). Of course, a similar relation exists between
the first half of the prime form and the second of the transposed inversion.
Babbitt called this property “hexachordal inversional combinatoriality.” The
term “combinatoriality” refers to combinatorics, a branch of mathematics deal-
ing with the permutations and combinations of elements of finite sets. The
property has long been regarded as peculiar to Schoenberg’s series, probably
because Schoenberg himself has written something about it. Here is that famous
statement from his paper “Composition with Twelve Tones”:
11. Among the 200 equivalence classes resulting from this definition there were
three that Forte considered trivial: the unique classes of sets with one, eleven and twelve
elements.
12. Babbitt used the word “aggregate” when he considered the total chromatic with-
out regard to order.

Example 4.2. A part of Forte’s 1964 table of set-classes, representing the PC sets of
cardinal numbers 5 and 7. Each class has an ordinal number, which is called a “set
number,” and can be found in the first column of each table. (The “Z”s will be
dealt with in a later section of this chapter.) This order number is based on the
magnitudes of the successive entries in the APIC vector. The first set-classes are
those with the highest magnitudes of APIC 1: 4 in the case of set-classes of cardinal
5, and 6 in the case of set-classes of cardinal 7. When the magnitudes of APIC 1 are
equal, those of APIC 2 are considered, and so on. The vectors are specified in the
third column of each table. The second column contains one representative of each
set-class, the PCs of which are given in “normal order” (another term awaiting
discussion). Finally, the fourth column gives the number of PC sets per class.
Equivalence classes of cardinal 5 and 7 with corresponding order numbers contain
PC sets related as complements with respect to PITCHCLASS.
Allen Forte, “A Theory of Set-Complexes for Music,” Journal of Music Theory, 8/2 (1964): 147.
Used by permission of Duke University Press and Yale University.

94 equivalence
Later, especially in larger works, I changed my original idea, if necessary, to

fit the following conditions: the inversion a fifth below of the first six tones,
the antecedent, should not produce a repetition of one of these six tones, but
should bring forth the hitherto unused six tones of the chromatic scale. Thus,
the consequent of the basic set, the tones 7 to 12, comprises the tones of this
inversion, but, of course, in a different order. (Schoenberg 1984, 225)
It is not so much the combinatorial property itself that is typical of Schoen-

berg’s series—for example, Webern’s series have this property as well—as rather
the use he made of it. Schoenberg’s idea was to simultaneously unfold two
series-forms while not allowing octave-related tones to coincide. Babbitt’s mis-
sion was to attain a more general level of understanding of this idea. He had
noted that the interval of a perfect fifth below (PIC 5) had no special meaning
in this regard, apart from the symbolic meaning Schoenberg may have attrib-
uted to it.13 Even in Schoenberg’s own works there are instances of hexachordal
inversional combinatoriality at other transpositional levels, such as the series of
the Variations for Orchestra, Op. 31 (T9I), or that of Moses und Aron (T3I).
Furthermore, Babbitt wanted to prove that the combinatorial property was
not just typical for inversionally related twelve-tone series (or for the simple case
of a series and its retrograde), but could also involve transpositionally related
series, or series related as a prime form and its transposed retrograde-inversion.
In these cases, hexachordal combinatoriality stipulated conditions other than
TnI-related hexachords: for the two halves of a series to complement the corre-
sponding halves of a transposition of that series, they need to be transpositionally
related to one another, and for the two halves of a series to complement the cor-
responding halves of a transposed retrograde-inversion of that series, they need
to be inversionally symmetrical. When all these conditions were met—that is,
when the two halves of a series were TnI-related, inversionally symmetrical, and
hence transpositionally related as well—Babbitt called the series “all-combinato-
rial,” indicating that its two halves complemented the corresponding halves of
all other series-forms at one or more transpositional levels. Example 4.3b shows
how the second half of such a series can be reordered to yield the first half, not
only of its retrograde but also of a transposition (lower staff, T6) and an inversion
13. It was a token of the past. It served to link Schoenberg’s music to that of the tonal
tradition, of which the composer could not stop seeing himself as a successor. (See the refer-
ences in chapters 1 and 3 to the role of the fifth in his Piano Piece, Op. 23, no. 3, pp. 7 and
70). Schoenberg’s remarks on another piece, the Wind Quintet, Op. 26, prove that, at the
time, hexachordal inversional combinatoriality mattered to him especially when it involved
two series related as prime form and T5I. What he claimed only is that the corresponding
hexachords of the prime and T5I forms do not fill out the total chromatic (Schoenberg
1984, 225). He failed to mention that the relation of the prime and T11I forms do satisfy this
condition. Significantly, in the piece itself this relation is not exploited.

equivalence 95
(0)
{C , D , D , E , G , G }
Example 4.3a. Hexachordal inversional combinatoriality; Arnold Schoenberg,

Piano Piece, Op. 33a. The first hexachord of the prime series-form is related to the
second by T5I. This means that the second hexachord can be reordered to yield the
first hexachord of an inversion.
(0)
{F, F , A , B , D , D}
Example 4.3b. Hexachordal all-combinatoriality; Schoenberg, Modern Psalm, Op.

50c. The first hexachord of the prime series-form is related to the second by T2, T6,
T10, T1I, T5I, and T9I. The second hexachord can thus be reordered to yield first
hexachords of all four series-forms.
(0)
{C , E , E , F} {F , G , A , B}
Example 4.3c. Tetrachordal combinatoriality; Schoenberg, Fourth String Quartet,

Op. 37. The second tetrachord of the prime series-form is related to the third by
T7I. The three tetrachords of this series-form can thus be reordered to yield those
of a retrograde-inversion.
(middle staff, T5I). Finally, it was Babbitt’s aim to show that tetrachords,
trichords, dyads, and even single PCs could be sources of combinatoriality as
much as hexachords (ex. 4.3c).
Combinatoriality, as it is generally understood, involves unordered PC sets
underlying non-corresponding sections of different series-forms. When Bab-
bitt considered a twelve-tone series in terms of such PC sets, he called it a

96 equivalence
Example 4.4. The source set of Schoenberg’s Piano Piece, Op. 33a.
“source set.”14 Example 4.4 shows what this means for the twelve-tone series of
Schoenberg’s Piano Piece, Op. 33a (cf. ex. 4.3a): it is reduced to a succession of
two hexachords defined with respect to PC content only, ignoring order. The
“source set” can be seen as providing the historical link between twelve-tone
theory and PC set theory: observing that two sections of different series-forms
contain exactly the same PCs comes close to stating that these sections represent
different orderings imposed on one, unordered PC set. This shows how closely
related the concepts of a PC set and a twelve-tone series actually are.
The actual step toward emancipation of the PC set as a musical concept was
taken when theorists started to define groupings of PCs in their own right, with-
out reference to a twelve-tone series. Since no particular intervallic relationships
among the PCs prevailed, and since the familiar twelve-tone operations did not
preserve the PCs themselves, these groupings were defined in terms of their
total interval content. This property, for which David Lewin and Donald Martino
had given formulas in 1960 and 1961 (see chapter 2, p. 47), is independent from
permutation, transposition, or inversion. Being less susceptible to change, it was
considered to be more fundamental to a PC set than the PCs themselves. In
music theory, it is customary to define a musical entity, not by its constituents
(such as pitches), but by the relations between its constituents (such as pitch or
time intervals). Entities thus defined include modes, hexachords, chords, fugue
subjects, and sonata themes.
The study of interval content had opened up new musical relations that
intrigued composers and theorists. And this study was all the more appealing
now that the tedious labor it involved could be left to computers.15 One particu-
larly exciting result was the discovery that any two hexachords complementing
one another with respect to PITCHCLASS are equal in terms of interval con-
tent, even if they are not related by transposition and/or inversion. The state-
ment of this relationship has become known as the “hexachord theorem,” and is
commonly attributed to Babbitt. Lewin (1960) presented it as a special case of a
14. The term “source set” was introduced in Babbitt 1955, 57n. Babbitt had described
the concept already in his 1946 doctoral thesis (Babbitt 1992, 104–16), occasionally refer-
ring to it as a “basic set.” Schoenberg, too, used the term “basic set,” though with refer-
ence to the non-transposed prime series-form.
15. The influence of the computer on the formulation of research questions and
the valuation of research results in the early 1960s can hardly be overestimated. Chapter
8A will give an account of the euphoria induced by the possibilities of computer-aided
research among American composers and musicologists.

equivalence 97
more general statement to the effect that the difference between the frequency
rates of a PIC in two complementary PC sets equals the difference between the
cardinal numbers of these sets. (This is called the “complement theorem.” In
the case of two complementary hexachords the difference is, of course, 0.)16
Lewin related the hexachord theorem to the typical way in which Schoenberg
partitioned his twelve-tone series. In the following quotation, the term “interval
function” refers to Lewin’s concept of the inclusive intervallic relation between
two PC sets:
If a piece is based on a hexachord P and is so constructed that, in general, some

form of P is always being played against the complementary form—P being dis-
tinguished from its complement at any given point, e.g. by instrumentation,
melody-accompaniment, etc.—and if the order of notes within the hexachords
is not too strictly constrained, then the interval function m(i) between P and
[its complement] P' will probably exert an aural effect in the long run. Since
m(i) represents the “total ‘contrapuntal’ relation” between P and P', of course
it also represents the “total ‘contrapuntal’ relation” between any transposed or
inverted form of P or P' and the corresponding form of P or P'. In this sense
the “total potential counterpoint” in any section of the piece is always the same.
And . . . this “total potential counterpoint” uniquely characterizes P, P', and their
transposed and inverted forms. This may explain in what sense a hexachord has
an aural “identity” in such a piece.” (Lewin 1960, 100)
In brief, the combinatorial properties of the twelve-tone series on which a work is

based place certain constraints on the harmonic idiom of that work. (Since “total
contrapuntal relation” implies linear statements of the two sets, which do not nec-
essarily occur, I prefer speaking of “harmony” instead of “counterpoint” here.)
Whereas Lewin thought of “harmonic idiom” only in terms of aural effect, Mil-
ton Babbitt invoked the composer’s intentions. In his words, “this characteristic of
singular harmonic definition [suggested] one important reason for Schoenberg’s
compositional concern with hexachords as combinational units” (Babbitt 1961a,
80). The claim that Schoenberg’s technique was informed by an awareness of
the equal distribution of intervals in complementary hexachords is probably too
bold; Schoenberg’s writings give no evidence of such awareness. However, what is
important here is the rise of interval content to the status of an analytical category,
a determinant of musical structure. Forte’s decision to choose interval content as a
basis of equivalence was a logical result of this development.17
16. Apart from Lewin, other theorists set out to prove the hexachord theorem mathemat-
ically as well. Proofs were published by Kassler (1961), Regener (1974), and Wilcox (1983).
17. He was not even the first to classify PC sets accordingly. Donald Martino, whose
research in the field of combinatoriality built on Babbitt’s, had already provided tables of
hexachords, tetrachords, trichords, and pentachords in his 1961 article “The Source Set
and its Aggregate Formations,” defining them in terms of interval content.

98 equivalence
“Operational Equivalence”
As mentioned, Forte’s interval-oriented definition of PC set equivalence was

soon challenged. In 1965, the equivalence problem was raised in an article that
appeared in the same Journal of Music Theory that had published Forte’s “A The-
ory of Set-Complexes” the year previously. The author was John Clough, a bud-
ding specialist in the field of music theory and mathematics. His main objection
to Forte’s definition of PC set equivalence was that Forte had invoked a rela-
tively unobvious relation—equality of interval content—whereas most PC sets
involved in analyses resembled each other in a more obvious way as well: as a set
and its transposition, inversion, or transposed inversion (a fact that Forte had
not failed to observe either). And the word “most” was the symptom of another,
even more acute problem: some PC sets with equal interval contents were not in
such an obvious relation. The set-classes that contain these PC sets are marked
with a “Z” in Forte’s table (for “zygotic,” a term originally referring to a pair of
animals under a yoke, but currently in use in the field of biology to denote the
fusion of two reproductive cells; see ex. 4.2); this peculiar relation between PC
sets has become known as the “Z-relation” (ex. 4.5).18
Clough raised the question of whether the “equality-of-intervallic-content” rela-
tion partitioned the collection of PC sets in a way that was musically meaningful and
unambiguous. He proposed to define the equivalence of PC sets A and B as follows:
A ≡ B if and only if
A and B are related by IT (Clough 1965, 169)
Clough had taken the label “IT” from Babbitt 1960 and Forte 1964. This label
denoted the same operations as TnI. (“Invert and then transpose A” and “trans-
pose the inversion of A” have the same results.) For Clough, however, the exact
protocol was not an issue at all. He used the label “IT” as an abbreviation of
“transposition and/or inversion” signifying a relation between PC sets that was
based on one of the two operations, or on both (165n).
Clough’s article was the first in the field of PC set theory to reveal the math-
ematical background of the term “equivalence,” and to come up with other, less
apparent, maybe less useful, but equally valid equivalencies among PC sets.19
18. The discovery of the relation—reported by Lewin (1960)—was another result of

the study of interval content. Some “Z-related” pairs are connected by M (or IM), a relation
depending on an APIC vector with identical entries for the interval classes 1 and 5 (see chap-
ter 3, p. 82). Forte revealed the meaning of the “Z” to the 25 participants of a seminar on the
history of serialism at the Orpheus Institute in Ghent, Belgium, on November 17, 2004.
19. For example, Clough defined a property which he called the odd-even characteristic
of a PC set: “the absolute value difference between the number of notes in one whole-tone
scale and number of tones in the other whole-tone scale” (Clough 1965, 164). The relation
between two PC sets sharing this characteristic is an equivalence in the algebraic sense.

equivalence 99
4
4
APICs: 5
3
1 1
{ 0, 1, 3, 4, 8 }
2 4
3
5
4
5
APICs: 4 4
3 1
{ 0, 3, 4, 5, 8 }
1 3
2
5
Example 4.5. Two PC sets not related as a set and its transposition, inversion or
transposed inversion, but nevertheless having the same interval content (Forte’s
“Z-relation”). With their common APIC vector 212320 they both belong to the
set-class of cardinal 5 listed as number 17 by Forte in his 1964 table of set-classes
(see ex. 4.2).
The relation that he proposed as an alternative to Forte’s equal-vector relation,

based on T and/or I, would ultimately become the paradigm of PC set equiva-
lence. Having first rejected it in a rather haughty reply to Clough’s article in the
same issue of Journal of Music Theory, Forte adopted it later, without reference to
Clough, in The Structure of Atonal Music (Forte 1965a, 1973).20
To define an algebraic equivalence on the basis of transposition and inver-
sion, as Clough did, seems problematical. Being not reflexive or transitive (in
general), an inversional relation is not a proper equivalence of this kind, and a
transpositional relation satisfies the criteria of algebraic equivalence only when
no index of transposition is specified. The term “equivalence” applies to these
relations in the way of Hubert S. Howe’s “operational equivalence”:
The PC coll[ection] y obtained by performing an operation o upon a PC

coll[ection] x will be called the operational equivalent of x under o. (Howe
1965, 50)
Yet it was the standard algebraic definition that Clough invoked. Transpositional
and inversional relations are consistent with this definition as long as it is imma-
terial which one of them obtains. Although Clough’s statement is valid in itself,
it may strike one as odd that each of the operations Tn, I, and TnI instantiates
20. It was necessary for him to do so, since the equal-vector relation contradicted the
relational concept that was crucial to Forte’s theory: the PC set complex (Clough 1965,
169; see also chapter 6.).

100 equivalence
PC set equivalence, for none of them can serve as a formal basis for equivalence
without the other. Is the algebraic concept of equivalence actually suited to rep-
resent these musical relations?
In any case, assigning objects to equivalence classes accurately models the
way in which we speak and write about music. When we call something a “major
triad,” a “pentatonic scale,” a “triplet,” or an “F♯,” we identify it as a member of
a class. These names refer to properties the objects concerned have in common
with other objects. And “having property x in common with” is an equivalence
in any collection of objects. However, when we define a class of transpositionally
and/or inversionally related PC sets, we are not, at first sight, dealing with a prop-
erty they share, but with a set of operations transforming one into another. Since
operations are not an obvious basis of algebraic equivalence, can we state their
relation in terms of a common property, a property invariant under transposi-
tion and inversion and not shared by PC sets of other classes?
There is at least one property invariant under transposition. The composer
and theorist Richard Chrisman drew attention to it in 1971, in an article entitled
“Identification and Correlation of Pitch-Sets.”21 Chrisman stated that two PC sets
with the same interval content would not yield two equal PIC successions if they
did not relate as transpositions. In other words, what transpositionally related
PC sets have in common is a set of PIC successions. It is possible to reduce the
number of these successions by arranging the PCs in ascending order and sub-
jecting them to cyclic permutation. A common application of this procedure is
the derivation, in manuals of tonal harmony, of the 6/3 and 6/4 chords from
a root position triad by successively transferring the lowest notes to a higher
octave. Chrisman would define these chords, not in terms of pitch intervals (or
PICs) measured from the bass—as in figured-bass practice—but in terms of adja-
cent pitch intervals bottom up, including the interval completing the range of
an octave. Thus, he would identify the three bass positions of a major triad as
“4–3–5,” “3–5–4,” and “5–4–3” respectively (ex. 4.6a). Chrisman called these suc-
cessions of pitch intervals “successive-interval arrays.” Each PC set generates a
number of these arrays equal to its cardinality.
Chrisman called two successive-interval arrays “equivalent” when one was a
cyclic reordering of the other. Apart from PC sets that were equal, this equiva-
lence involved transpositionally related PC sets. It did not involve PC sets related
as (transposed) inversions. Ultimately, however, Chrisman assigned PC sets and
their inversions to the same set-classes, which roughly equal the set-classes iden-
tified by Forte in 1964 (see ex. 4.2. Chrisman even used Forte’s “set numbers” to
label these classes. See his table 1 on p. 79). He did not include the inversions
on the basis of a defined equivalence. The inversional relation was described by
him as follows:
21. The article was derived from Chrisman’s dissertation: A Theory of Axis-Tonality for
Twentieth-Century Music (Yale University, 1969).

equivalence 101
(a)
4 3 5 3 5 4 5 4 3
(b)
3 4 5 4 5 3 5 3 4
Example 4.6. Richard Chrisman’s successive-interval array. An application to the

major and minor triads.
“The interval array of one [PC set is] equivalent by cyclic reordering to the retro-
version of the other.” (Chrisman 1971, 70; my italicization)
For example, the retroversion of the successive-interval arrays pertaining to the

major triad—i.e., “4–3–5,” “3–5–4,” and “5–4–3”—yields those pertaining to the
minor triad: “5–3–4,” “4–5–3,” and “3–4–5” (ex. 4.6b). However, the successive-
interval array of a PC set is not equivalent by cyclic reordering to that of its inver-
sion. This means that Chrisman’s set-classes do not satisfy the definition of an
equivalence class.
Chrisman’s article can be seen as an attempt to define PC sets (or “pitch
sets,” as they were called at the time) in terms of a more exclusive concept of
their interval contents than the APIC vector. An algebraic equivalence in PCSET,
involving only a PC set and its transpositions, can be defined on the basis of a
common set of successive-interval arrays. This equivalence did not involve inver-
sionally related PC sets. Notwithstanding the obvious relation between their suc-
cessive-interval arrays, it proved difficult to pin down a property held in common
between PC sets and their inversions.
Problems of Representation
Useful though it may have been, Chrisman’s concept of a PC set has not
established itself in the literature.22 A common set of successive-interval
arrays is not a property lending itself to concise modeling. For the sake of
conciseness, we can single out one of them as a referential array, but this
requires a fixed protocol.
22. Eric Regener (1974) and Robert Morris (1987, 40, 82; 1995) have proposed
concepts that are similar to Chrisman’s successive-interval array. Morris coined the term
“cyclic interval succession,” or “CINT.” His “pitch-class” interval succession” (PCINT) will
receive brief coverage later in this chapter.

102 equivalence
There already was such a protocol when Chrisman published his article. It
did not relate to successive-interval arrays, but to PC sets. One used to speak of
the “normal form” or “normal order” of a PC set, a concept that can be traced
back to the teaching and writings of Milton Babbitt, like so many other concepts
of PC set theory. The following definition appeared in “Set Structure as a Com-
positional Determinant”:
It is suggested that a given H[exachord] be ordered in “normal form” wherein

the interval determined by the first p.c.−last p.c. is greater than any interval
determined by p.cs in H. (Babbitt 1961a, 77)
Babbitt’s language is obscure here. In order to clarify his definition, let us con-
sider the example given by him: the first hexachord of the series-form opening
the third movement of Schoenberg’s Fourth String Quartet, Op. 37 (ex. 4.7).
Babbitt identifies (0,1,4,5,6,8) as its normal form. Two things are important: first,
we can obtain these numbers from the first six notes of Schoenberg’s movement
if we assign the value 0 to the PC G; second, in this order they represent the
most compact arrangement of the PCs involved. Apparently, what Babbitt means
to say is this: in any cyclic ordering of these six PCs—like (G,A♭,B,C,C♯,E♭),
(B,C,C♯,E♭,G,A♭), or (E♭,G,A♭,B,C,C♯)—the adjacent PCs E♭ and G yield the
largest difference modulo 12 of all adjacencies, namely 4. When this adjacency is
not included—in other words, when G is the first PC and E♭ the last—the PC set
is represented in normal, that is, most compact form.
From Babbitt’s rather casual definition, we can infer that the concept of
normal form had already been in use for some time. Otherwise, Babbitt would
have introduced it formally. Applying rules of scientific discourse in statements
about music was very important to him (as we shall see in chapter 8B). Another,
even more casual reference to normal form seems to confirm that the reader’s
familiarity with the concept was taken for granted. This reference comes from
Donald Martino’s article “The Source Set and its Aggregate Formations,” which
was published in the same year and in the same journal as Babbitt’s article (i.e.,
Journal of Music Theory):
Given a [hexachord] whose combinatorial properties are unknown, first

reduce it to normal form, then count up the fifteen intervals and consult the
table. (Martino 1961, 227)
Someone who had devised an algorithm for obtaining the normal form of a
PC set was James K. Randall, another composer from Babbitt’s circle at Princeton.
Hubert Howe provided the following summary of this description, with reference
to Randall’s unpublished study Pitch-Time Correlation from 1962. The term “pitch-
class collection” (“PC coll”) refers to a PC set. The term “pitch-structure” desig-
nates a collection of transpositionally related PC sets:

equivalence 103
614 [unis.] poco accel. a tempo

3 3
Example 4.7. The opening of the third movement of Schoenberg’s Fourth String
Quartet, Op. 37. The first hexachord of the twelve-tone series is marked.
STRING QUARTET, NO. 4, OP. 37, by Arnold Schoenberg. Copyright © 1939 (Renewed) by
G. Schirmer, Inc. (ASCAP). International Copyright Secured. All Rights Reserved. Used by
Permission.
(a) Select from among the pitch-class collections xi contained in a pitch-struc-

ture z the PC collections which contain the PC 0 (zero) and arrange the PCs
in numerical order. (For example, the pitch structure containing the PC coll
0127 and its transpositions yields the following four sets: 0127, 016 11, 05 10
11, and 0567.)
(b) Select from these PC colls the set or sets which have the smallest final num-
ber. (In the example above, this step yields the sets 0127 and 0567.)
(c) If step (b) yields more than one set, select from among these sets the set
with the smallest second, third, etc., number until only one set remains. (Con-
tinuing the above example, this step yields 0127.)
The unique ordered number-set thus obtained is defined as the normal form of
the pitch-structure z. (Howe 1965, 49)
The normal form corresponds to one of the successive-interval arrays of a PC

set. It represents all possible realizations of this PC set and its transpositions. Allen
Forte adopted the concept in “A Theory of Set-Complexes,” with a reference to
Martino’s article. It appeared in his table of set-classes, where it was designated
“normal order” (see ex. 4.2, the second column). However, Forte did not adopt it
without changes. The normally ordered PC sets in his table—all of which shared
the same first element (0, or C)—represented set-classes, but these set-classes did
not consist only of transpositionally related PC sets. Since the APIC vector served
as the basis of equivalence, they also contained the inversions of these PC sets.
And in a few cases they contained PC sets that were neither transpositionally nor
inversionally related to the normally ordered reference set (the “Z-related” sets).
This means that the normally ordered reference set represented many more
PC sets than just its transpositions. Forte had thus extended the range of PC sets
covered by the concept of a normal form. I believe he did this inadvertently; it
was the consequence of using a common representation (the normally ordered
reference set) based on a more exclusive relation (transposition, or a common
set of successive-interval arrays) than the relation that actually determined the
set-classes (equal interval content). This discrepancy between representation

104 equivalence
and content of Forte’s set-classes required clarification, which Forte eventually

provided in The Structure of Atonal Music.
In “A Theory of Set-Complexes,” Forte had not indicated by which protocol
he obtained the normally ordered reference set of a set-class. This was an omis-
sion, for if “normal” means “most compact,” there may be several normal forms
from which to choose. First, among the cyclic permutations of a PC set there
may be more than one with the smallest difference modulo 12 between the first
and last PC values. Second, in most cases a PC set and its inversion have differ-
ent normal forms. And third, if a set-class contains two members not related as
a set and its transposition, inversion, or transposed inversion—that is, a Z-pair—
one needs to take into account the normal forms of each. Forte had only casu-
ally referred to an “optimal” normal order (Forte 1964, 144). In The Structure of
Atonal Music, however, he described its derivation. Here is a summary:
Step 1 Consider a PC set. Put the PC values in ascending numerical order.
Step 2 Consider all cyclic permutations of this ordered set. Select the permu-
tations with the smallest PIC value determined by the ordered pair of
the first and last PCs. Proceed, if more than one permutation remains.
Step 3 For each selected permutation, consider the ordered pairs of the
first, and each of the remaining PCs. Select the permutation with the
smallest non-common PIC value determined by these ordered pairs.
If there is no such permutation among the ones that remain, each of
them may serve as normal order.
An example will be helpful here. The PC set (1,4,6,7,8,11) is already ordered by

step 1. Step 2 yields two other permutations: (4,6,7,8,11,1) and (11,1,4,6,7,8).
The ordered pairs of the first and last PCs of these permutations yield the small-
est possible PIC value, namely 9.
Since step 2 singles out two permutations, it is necessary to proceed with step 3.
In both permutations, the smallest PIC value determined by an ordered pair involv-
ing the first PC and one of the other PCs is 2: PIC(4,6) and PIC(11,1). The small-
est-but-one such value is 3 in the first permutation (PIC(4,7)), and 5 in the second
(PIC(11,4)). Therefore, 3 is the smallest non-common PIC value, and (4,6,7,8,11,1)
is the optimal, or “best,” normal order of this PC set (cf. Forte 1973, 4).
One additional step should be taken:
Step 4 Consider the remaining permutation of our PC set. Rename the PCs
in terms of the PIC values pertaining to the ordered pairs they form
with the first PC.
The result is an ordered PC set the first element of which is 0, like the sets in
the second column of the table in example 4.2. The PC set (4,6,7,8,11,1) thus

equivalence 105
transforms into (0,2,3,4,7,9). In The Structure of Atonal Music this ordered PC set is
called “prime form” (Forte 1973, 5). The term “prime form” originally belonged
to the vocabulary of twelve-tone serialism. “Prime set”—or simply “prime” (Bab-
bitt 1955, 1992; Perle 1962)—was an alternative designation of what Schoenberg
had called a Grundgestalt, or “basic set.” In “A Theory of Set-Complexes” Forte
had used the term “prime form” as the antonym of “inversion,” calling to mind
the traditional pair of rectus and inversus. (“Normally each PC set has two basic
and distinct forms: a prime form and an inversion”; p. 144.) But it ended up
referring to the final outcome of the above reductive protocol.
How did Forte deal with the conflict between the representation and the con-
tent of his set-classes? Only PC sets that are related as transpositions will produce
the same prime form. How can one prime form represent a PC set and its inver-
sion at all transpositional levels, let alone two “Z-related” PC sets? In The Structure
of Atonal Music, as will be recalled, set-classes consist only of PC sets that are
related transpositionally and/or inversionally. Here, Forte had finally met John
Clough’s criticism of his earlier definition, which proceeded from the interval
vector; he had abandoned equality of interval content as a basis of PC set equiva-
lence. However, there was still a need for a representational format applying
equally to the transpositions of a PC set and those of its inversion.
Forte’s proposal was to consider the PC set and its inversion in normal order,
and to repeat steps 2, 3, and 4 above (cf. Forte 1973, 12–13). One of these normally
ordered PC sets should prevail over the other (unless they were inversionally sym-
metrical, of course), and this set was to be reduced to its prime form. To return to
our example: the inversion of (4,6,7,8,11,1) is (8,6,5,4,1,11). The normal order of
this new PC set is (4,5,6,8,11,1). The PIC value of its initial pair—that is, (4,5)—is
smaller than that of the normally ordered original set. It is, in other words, the
smallest non-common PIC value of step 3. Therefore, (4,5,6,8,11,1) prevails over
(4,6,7,8,11,1). Step 4 reduces it to (0,1,2,4,7,9), the Fortean prime form of the set-
class to which these inversionally related PC sets belong.
We can also obtain the prime form without first determining the normal
order. We should then apply Randall’s algorithm to the transpositions of a PC
set and its inversion. Given an arbitrary set, which I call A:
Step 1 Consider all transpositions of A and I(A). Put them in ascending

numerical order.
Step 2 Select the PC sets beginning with 0, and having the smallest last integer.
Step 3 Select the PC set with the smallest non-common integer.
The PC set that remains is the ultimate prime form. Table 4.1a and 4.1b show
how, according to this protocol, (0,1,2,4,7,9) is identified as the prime form of
the PC set {1,4,6,7,8,11}.

Table 4.1a. The transpositions of the PC set {1,4,6,7,8,11} and its inversion
1 4 6 7 8 11 11 8 6 5 4 1
2 5 7 8 9 0 0 9 7 6 5 2
3 6 8 9 10 1 1 10 8 7 6 3
4 7 9 10 11 2 2 11 9 8 7 4
5 8 10 11 0 3 3 0 10 9 8 5
6 9 11 0 1 4 4 1 11 10 9 6
7 10 0 1 2 5 5 2 0 11 10 7
8 11 1 2 3 6 6 3 1 0 11 8
9 0 2 3 4 7 7 4 2 1 0 9
10 1 3 4 5 8 8 5 3 2 1 10
11 2 4 5 6 9 9 6 4 3 2 11
0 3 5 6 7 10 10 7 5 4 3 0
Table 4.1b. The same PC sets, put in ascending order . . .
1 4 6 7 8 11 1 4 5 6 8 11
0 2 5 7 8 9 0 2 5 6 7 9
1 3 6 8 9 10 1 3 6 7 8 10
2 4 7 9 10 11 2 4 7 8 9 11
0 3 5 8 10 11 0 3 5 8 9 10
0 1 4 6 9 11 1 4 6 9 10 11
0 1 2 5 7 10 0 2 5 7 10 11
1 2 3 6 8 11 0 1 3 6 8 11
0 2 3 4 7 9 0 1 2 4 7 9
1 3 4 5 8 10 1 2 3 5 8 10
2 4 5 6 9 11 2 3 4 6 9 11
0 3 5 6 7 10 0 3 4 5 7 10
Note: Twelve sets begin with 0, four of them having the smallest last integer. The boldly printed
set is the prime form.

equivalence 107
It is unfortunate that, in most cases, the PIC succession of the prime form
is not obtainable from all members of a set-class. Forte did not actually resolve
the conflict between representation and content.23 However, the prime form
figured prominently in Forte’s new definition of PC set equivalence:
Two pc sets will be said to be equivalent if and only if they are reducible to the
same prime form by transposition or by inversion followed by transposition.
(Forte 1973, 5)
There is a subtle difference between this statement and the one John Clough had
proposed eight years before. Clough spoke of operations transforming PC sets into
other PC sets (“related by IT”). Forte spoke of a property that PC sets may have
in common (“being reducible to the same prime form. . . .”). It is the same subtle,
but not unimportant, distinction as that between “being a sister of” and “having the
same parents as.” The elements are the same (PC sets, or people), but their relation
is conceived of in two different ways. One way tells us what the first element is with
respect to the second, and what, in turn, the second element is with respect to a
third (for example: “is related by Tn to,” “is a sister of”). This may not be what the
second element is with respect to the first (“is related by T−n mod 12 to,” “is a brother
of”), and what the first element is with respect to itself, or to the third element.
(Clough could solve this problem for the PC sets by not being too specific about
their relationship. As he saw it, they were mutually related just by T.) The other way
tells us what these three elements are with respect to something else (prime form,
parents), thus putting a lesser strain on the conditions of algebraic equivalence.
The substitution of the equal-vector relation by a set of relations based on
transposition and/or inversion resulted in 23 more equivalence classes than
Forte had listed in 1964. The 23 equivalence classes containing PC sets not
related by transposition and/or inversion (the “Z-”pairs) had to be split up.
On the other hand, the new list that Forte had published as an appendix to
The Structure of Atonal Music did not contain the twelve classes of PC sets with
two or ten elements. Forte had abandoned them, probably because he con-
sidered them too common to generate meaningful relationships. Example
4.8 shows a part of the new list: the part with the set-class representatives of
cardinal number 5 and 7, which can be compared to the corresponding part
of the 1964 list shown in example 4.2.
23. It would be an improvement if the normal order of a PC set would always clearly
reflect the prime form. For example, it is possible to define a normal order that is an arrange-
ment of the elements corresponding to that of their images in the prime form. (The term
“image” refers to the element to which a given element is assigned by some function; in this
case by transposition and/or inversion.) But this would contravene a by now common practice
in two ways: first, the normal order would be derived from the prime form, and not, as usual,
the other way around; and second, this alternative protocol would sanction descending normal
orders. For example, the normal order of the PC set {1,4,6,7,8,11} would be (8,7,6,4,1,11).

Example 4.8. A part of Forte’s 1973 table of set-classes, representing the PC sets of
cardinal numbers 5 and 7. Of each set-class, the name, prime form, and APIC vector
are given. The name is composed of the cardinal and ordinal numbers (the latter
of which is the “set number” from the table of 1964; see example 4.2.) In this table
there is no separate column for the number of PC sets per class. When it is not the
“normal” 24, this number appears in brackets in the “Name” column. On each side
of the table—that is, for each of the cardinal numbers 5 and 7—there are three
more PC sets than in the table of 1964. The new classes with ordinal numbers 36,
37, and 38 were originally part of those with ordinal numbers 12, 17, and 18
respectively. The “Z”s now refer to their common APIC vectors.
Allen Forte, The Structure of Atonal Music. Copyright © 1973 by Yale University.

equivalence 109
The number of items in the new list is greater, because the “Z-related” PC sets
now belong to different classes. Apart from this, the set-classes have “names.”
A set-class is not only identified by its prime form, but also by its hyphenated
cardinal and ordinal numbers.24 In spite of the criticism that the latter are
arbitrary identifiers (Regener 1974), they have been in general use since the
publication of The Structure of Atonal Music. Single PC sets are commonly identi-
fied by the names of the Fortean set-classes to which they belong. This has the
advantage of quickly revealing transpositional and inversional relations. Some
of Forte’s names have become very familiar among music theorists, because of
the peculiar sets to which they refer, or because of the role they play in the rep-
ertoire: “4-Z15” ({0,1,4,6}; one of the so-called “all-interval” tetrachords), “6-20”
({0,1,4,5,8,9}; the “Ode-to-Napoleon” hexachord), “5-Z17” ({0,1,3,4,8}; the “Far-
ben” chord), and “8-28” ({0,1,3,4,6,7,9,10}; the octatonic scale). The establish-
ment and rapid spread of this new vocabulary in the English-speaking world has
been one of Forte’s major successes.
Class Types
In the preceding sections I have more than once referred to the concept of a
set-class: an equivalence class of PC sets. An equivalence in a collection of things
(like PC sets) stipulates that this collection divides into equivalence classes. This
aspect of PC set equivalence was hardly mentioned until after the publication of
The Structure of Atonal Music. Forte did not use the term “set-class” in his treatise.
He did use the term “equivalence-class,” but not with reference to collections of
PC sets. In earlier publications, the concept of a set-class did not play an impor-
tant role either. To be sure, it was sometimes mentioned. For example, Howe’s
term “pitch-structure” referred to it (Howe 1965, 48). Usually, however, equiva-
lence was defined as a relation between individual PC sets satisfying the condi-
tions of reflexivity, symmetry, and transitivity.
As we have seen, it is difficult to state transpositional and inversional rela-
tions in a way that meets these three conditions. On the other hand, this set
of relations divides PCSET into non-overlapping subcollections, thus comply-
ing with the criteria of an equivalence class. It is possible to discuss PC set
equivalence from this angle, and avoid mentioning the relation itself alto-
gether. What is crucial for an equivalence relation between PC sets is not its
being reflexive, symmetrical, and transitive, but its partitioning the collec-
tion of all PC sets.
John Rahn took this approach in his Basic Atonal Theory. Rahn presented sev-
eral partitions of PCSET, each of which resulted in distinct “types” of PC sets.
24. Forte already used these names in his 1964 article, but the corresponding list
only provided the ordinal numbers of the set-classes.

110 equivalence
The first of these partitions was based on cardinal number, and resulted in 13
“cardinality types” (from cardinal number 0 to cardinal number 12). Rahn pro-
ceeded by partitioning these cardinality types on the basis of transposition and
of transposition and/or inversion. These partitions yielded “Tn-types” and
“Tn/TnI-types,” respectively. This is how Rahn defined the concept of a “type”:
“Types” are “equivalence classes” . . . Equivalence classes . . . must meet two main
criteria:
1. “exhaustivity”: the set of equivalence classes (the “partition”) must exhaust
the domain (i.e., the sum of the equivalence classes must equal the
domain) and
2. “exclusivity”: no two equivalence classes may have a member in common
(i.e., the intersection of every pair of the equivalence classes must equal
the null set). (Rahn 1980a, 74–75)
Any collection consisting of PC sets related as transpositions and (transposed)

inversions fulfills these criteria. Therefore, Rahn considered any two PC sets
equivalent that belonged to one such collection. PC sets could be equivalent
“under Tn,” or “under Tn or TnI, or both.”25 Furthermore, Rahn contemplated
equivalence classes of ordered sets (“lines”) as well as of unordered sets, and of
pitch sets as well as of PC sets. Of course, pitch sets and ordered sets are mem-
bers of much larger collections than PCSET.
Now that the partitioning of a collection was presented as the prime distinguish-
ing property of an equivalence relation, it was easier to accept other sets of relations
as equivalencies, even though they were less common and often less obvious than
transpositional and inversional relations. One such set was determined by transposi-
tion, inversion, and/or multiplication. As we have seen in chapter 3, multiplication
(by 5 or 7) transforms one PC set into another that may not be a transposition or
transposed inversion. Therefore, equivalence classes resulting from this set of rela-
tions tend to be larger than those resulting from transposition and inversion alone.
More specifically, a number of these equivalence classes comprise two classes of the
“Tn/TnI-type.” Rahn did not pay much attention to the “Tn/TnI/M5/M7-type” in his
manual, which nevertheless included a section on multiplicative operations; he only
mentioned it in an appendix (Rahn 1980a, 139).
In Rahn’s treatise, equivalence is not a specific relation, like “having the same
interval content” or “being reducible to the same prime form,” but a property of
various possible relations in various possible collections. As well as being more
25. It should be noted that in Rahn’s notation the index “n” does not represent a
specific integer value, but one from a given set of such values. Seen thus, B = “Tn”(A) is
an equivalence in the collection of all PC sets, for there is always a value “n” such that A =
“Tn”(A) (reflexivity), A = “Tn”(B) (symmetry), C = “Tn”(B), and C = “Tn”(A) (transitivity).

equivalence 111
mathematically precise, this is a more flexible use of the concept, enabling it to

describe musical relations in more or less detail, as the context requires.
Here is an example that requires consideration of the musical context. Arnold
Schoenberg’s Piano Piece, Op. 33a begins with a series of six four-note chords. A
classification of their PC sets according to “Tn/TnI”-type reveals a retrograde-sym-
metrical structure: α–β–γ–γ–β–α (ex. 4.9). This structure is reinforced by the over-
all pattern of rising and falling. Now, these observations are nice, but not entirely
accurate. Equivalence relations between PC sets do not provide scope for a more
important characteristic of these opening measures, one that entails the retrograde-
symmetrical arrangement of PC set classes. The two αs not only represent PC sets
of the same “Tn/TnI”-type; they are also real mirror inversions of one another. The
same holds true for the two βs and the two γs. Therefore, the relations between
these chords are more accurately conceived in a collection of pitch sets, a collection
that we might call PSET. The pitch sets are related pairwise as follows:26
The pitch sets of the two αs: IPI (−9, −2) ({−13, −12, −7, −2}) = {2,1, −4, −9}
The pitch sets of the two βs: IPI (−2,3) ({−3,1,3,6}) = {4,0, −2, −5}
The pitch sets of the two γs: IPI (3,10) ({−4,2,4,7}) = {17,11,9,6}
This representation adds a nuance that lies beyond the scope of PC sets. The
pattern of rising and falling is definitely not symmetrical, since the second γ is
projected higher than the first, so that the rising progression (α–β–γ–γ) is actu-
ally longer than the falling one (α–β–γ). Schoenberg’s dynamic markings stress
this asymmetrical pattern of rising and falling rather than the retrograde-sym-
metrical structure formed by the PC set classes of these measures.
The analytical exercises in Rahn’s chapter on “set types” refer to this musical
fragment. However, Rahn treats the chords as realizations of PC sets, not pitch
sets. He was inclined to generalize musical relationships, to strip them of all the
features referring to their actual musical context. Example 4.10 shows his analy-
sis of the beginning of the Prelude to Tristan und Isolde by Wagner. According to
Rahn, the first three phrases of this prelude consist of four or five simultaneities
each. If “Tn-types” are assumed, and one Greek letter is assigned to simultanei-
ties of the same type, the resulting pattern is α–β–β–γ. However, the assumption
of “Tn/TnI-types” results in a “retrograde-symmetrical structure,” namely α–β–β–α
(Rahn 1980a, 78).
26. The choice of (−9, −2), (−2,3), and (3,10) as referential pitch intervals is not
an arbitrary one. Each of these pitch intervals involves the PCs E♭ and B♭, which spell
out Schoenberg’s favorite perfect fifths and fourths. B♭ is the first element of the prime
series-form, and E♭ is its “image” (see note 23) in the T5I of the retrograde. The opening
measures involve just these two series-forms.

Mäßig = 120
1 cantabile 2
γ β α
α β γ
Example 4.9. The first two bars of Schoenberg’s Piano Piece, Op. 33a.
Example 4.10. John Rahn’s analysis of the beginning of the Prelude to the opera
Tristan und Isolde by Richard Wagner. Each type of PC set is named after its prime
form, which Rahn called “representative form.” In Rahn’s notation, the representa-
tive form of a “Tn”-type is identified by parentheses. The representative form of a
“Tn/TnI”-type is identified by square brackets.
From Basic Atonal Theory 1st. edition by Rahn. 1980. Reprinted with permission by Wadsworth,
a division of Thomson Learning: www.thomson.com. Fax 600 730–2215.

equivalence 113
This sort of result appealed to Rahn. Why? First of all, one of its attractions
was its lack of obviousness. It showed him that analyzing music could yield pro-
found insights. He expected his generalizing approach to reveal musical rela-
tionships that would otherwise remain concealed. Few could write about the
gains of such an approach more compellingly than John Rahn:
Each step up the ladder of abstraction loses particular distinctions but gains
generality . . . Relations that may lie obscured in the thicket of the full particu-
larity of things can be perceived clearly when a process of generalization has
pruned away the underbrush of reality. A coherent structure of such abstract
relations does exert its own perceptional pull, forming associations strongly
in conjunction with and even in [spite] of the [criterion] of “same perceptual
neighborhood.” (Rahn 1980a, 77)
Rahn’s analysis of the first three phrases of Wagner’s prelude shows how “a
step up the ladder of abstraction”—i.e., from considering equivalence “under
Tn” to considering equivalence “under Tn or TnI or both”—reveals a retrograde-
symmetrical structure like that in the first measures of Schoenberg’s Opus 33a.
The “underbrush” that Rahn has “pruned away” in order to arrive at this result
is, first, the distinction between actual chords (like the simultaneities 4, 8, and
13 in example 4.10) and mere voice-leading events (like the simultaneities 3, 7
and 12), and, second, the difference in sound and tonal function of chords that
are based on inversionally related PC sets (like the “Tristan chord” 1 and domi-
nant-seventh chord 4).
Ignoring these distinctions is certainly questionable, but Rahn did reveal a
well-formed structure, one that called to mind twelve-tone serialism. Using Wag-
ner’s prelude along with the opening measures of Schoenberg’s Opus 33a can
be seen as a statement that Schoenberg’s compositional practice was built on
Wagner’s. Rahn thus provided Schoenberg’s music with a historical background
that added to its authority, while simultaneously substantiating his view of tonal
theory as a subbranch of atonal theory (see chapter 2, p. 45).
Canonical Transformation Groups
By stressing the class membership of PC sets rather than the operations con-
necting them, Rahn showed the equivalence of the transpositions of a PC set
and its inversion in quite a satisfactory way. Moreover, he pointed out other rela-
tions among PC sets similarly yielding (smaller or larger) equivalence classes.
In other words, equivalence came to be regarded as a category of different rela-
tions rather than as a relation in itself, and each of the relations of this category
counted as potentially significant. The significance of a relation was actually cer-
tified by its falling in the category of equivalencies; but it was indeed potential,
depending as it did on its musical function. The concept of musical equivalence

114 equivalence
thus became a truly “general” concept. Meanwhile, its traditional connotations

were bumped down into the realm of the “particular.”
One problem remains. Regarding equivalence, Rahn’s focus was not so much
on the quality of a relation between two PC sets—whether it was reflexive, sym-
metrical, and transitive—but rather on the partitioning of PCSET. However, if it
is not a single relation but a set of relations that partitions PCSET—such as the
Tns and TnIs—then how do these relations cohere? Is it possible to reduce them
to a single concept?
David Lewin, in the introduction to an article on different ways of model-
ing the intervallic properties of PC sets, invoked the mathematical concept of a
“transformation group” to account for the relations between operations jointly
effecting a partition of PCSET. For this reason, he referred to these operations
as “transformations” (Lewin 1977b, 195). The properties of a transformation
group have already been touched upon at the beginning of chapter 3, where we
examined the relation between the terms “operation” and “group action.” Now
follows a systematic recapitulation that complies with the general definition of a
mathematical group.
First of all, for a set G of transformations to be called a “group,” each combi-
nation of two or more of these transformations must equal a single other trans-
formation from the same set. In other words, if g and h are members of G, and
gh = k, then k must also be a member of G. (This property is called “closure”).
Secondly, combining one transformation with the result of combining a second
and a third must be the same as taking the third and combining it with the result
of combining the first and the second: g(hk) = (gh)k (“associativity”). The third
property is the presence of an “identity”-transformation e. The combination of
this and another transformation is equal in value to the latter alone: ge = eg = g.
The fourth and final property—not explicitly stated but merely implied by the
presentation in chapter 3—is that for each transformation g there is an inverse
g−1, a transformation with which it combines to yield the value of the identity
transformation: gg−1 = g−1g = e.
The sets of transformations underlying Rahn’s “Tn”-, “Tn/TnI”-, and “Tn/TnI/
M5/M7” class-types all satisfy these criteria. Each member of a set-class—that
is, each PC set—can be viewed as a transformation of every other member of
the same set-class, and as its own T0 (the identity transformation). Therefore,
all set-classes can be subsumed under the general concept of a “transformation
group.” Transformations defining a set-class type in conformity with established
practices are “canonical,” in Lewin’s words. The relation “is a canonical trans-
formation of” is an equivalence among PC sets. When two of them were in that
relation, Lewin called them “canonically equivalent” (Lewin 1987, 104).
There are groups of 12 (T), 24 (T and TI), and 48 (T, TI, TM, and TIM)
canonical transformations. However, set-classes are not always that large, since
some PC sets are self-reproductive (i.e., invariant) under one or more opera-
tions (apart from T0). For example, the PC set {2,3,5,6,10} is self-reproductive

equivalence 115
Table 4.2. The numbers of set-classes generated by three canonical transformation

groups
T, TI,
Cardinal nr. T T, TI TM, TIM
0 1 1 1
1 1 1 1
2 6 6 5
3 19 12 9
4 43 29 21
5 66 38 25
6 80 50 34
7 66 38 25
8 43 29 21
9 19 12 9
10 6 6 5
11 1 1 1
12 1 1 1
Note: See, by way of comparison, Morris 1987, 80.
under T8I; and the PC set {11,0,2,5} is self-reproductive under T0IM. The mem-
bers of any set-class are self-reproductive under an equal number of operations.
Table 4.2 shows the numbers of set-classes defined by each canonical transfor-
mation group. Each group can be said to “act” on the set-classes which it defines,
in accordance with the definition of a group action given in chapter 3. This
includes the class of PC sets of cardinal number 1, i.e., PITCHCLASS.
A Little More History: “Musical Statistics”
The PC set classifications shown in table 4.2 stand in a line of development

with comprehensive chord classifications from earlier periods. In chapter 2,
there was already brief mention of these older classifications, but now we can
study their relation to PC set theory in more depth. It should be recalled that
such classifications were a reaction to the apparent harmonic freedom in con-
temporary music, and that they stemmed from a spirit of objectivism typical
of modernist thought: as the chordal vocabulary seemed no longer subject

116 equivalence
to aesthetic and syntactical constraints, all a theory of harmony could do was

to inventory possible chords—given some set of available pitch elements—,
and perhaps to provide suggestions for their use. Or so was the opinion of the
compilers of these classifications. According to Fritz Heinrich Klein (1925), a
composer who had studied with Alban Berg, harmonic theory was to turn into
“musical statistics” (Musikstatistik).
The practitioners of this approach to harmony came up with lists of dis-
tinct chord types of each cardinality in (most often) the twelve-tone universe.
Some of them only provided the numbers of chord types they had found.
These numbers were significant insofar as they revealed the criteria by which
one assigned chords to the same classes. It is interesting to note that, appar-
ently, and in spite of all objectivity, these criteria could differ greatly. Table
4.3 shows the numbers pertaining to three classifications, and lists the authors
who calculated or cited them.27 The following discussion will cover the notions
of equivalence suggested by these classifications, while bearing in mind the
contexts in which they were presented.
Klein himself is one of the authors cited in table 4.3. He had presented the
results of his enumerations in an article entitled “Die Grenze der Halbtonwelt,”
which appeared in the German periodical Die Musik in January 1925. These
results are not the main focus of the article. They merely serve as an introductory
note to a discussion of compositional techniques proceeding from a twelve-tone
“mother chord.” Klein’s suggestions for the derivation of harmonic combina-
tions from this chord are important, for they reflect a mode of thought that
profoundly influenced his friend and teacher Alban Berg.28 Furthermore, the
“mother chord” (read from the bottom up: A0, A1, G♯2, E3, C♯4, B4, F♯5, C6,
F6, G6, B♭6, D7, E♭7) is the earliest known instance of an “all-interval series,” in
which each pair of adjacent PCs represents another PIC. Thus, it stands at the
beginning of a long search for other such series.29
The number of chord classes that Klein distinguished corresponds to the sum
total of non-equal PC sets (cf. table 2.1).30 This means that, for Klein, PC con-
tent was the distinguishing property of a chord class. From this viewpoint, all
27. This table summarizes a good deal of the information presented and discussed by
Bernard (1997) and Nolan (2003). It adds the reference to Bruno Weigl, whose work is
given consideration elsewhere by Menke (2005).
28. As Arved Ashby (1995) has shown, Berg’s practice of using more than one series
in a composition was not so much a liberty taken with the twelve-tone technique taught
by Schoenberg as the consequence of another systematic approach, for which Berg was
indebted to Klein.
29. See, for example, Krenek (1937), Babbitt (1992), and Eimert (1950, 1964). This
search ended with the publications of Jelinek (1961) and Bauer-Mengelberg and Ferentz
(1965). With help of a computer, these authors found the total number of 46,272 all-
interval series, transpositions included.
30. With the exception of the null set, which Klein did not take into account.

equivalence 117
Table 4.3. Early comprehensive chord-class counts in the twelve-tone universe
Cardinal nr. Number of “chords”

A. Loquin (1895)
A. Loquin (1874) E. Bacon (1917)
E. Stein (1925) J. M. Hauer (1925)
F. H. Klein (1925) B. Weigl (1925) W. Howard (1932)
1 12 1 1
2 66 11 6
3 220 55 19
4 495 165 43
5 792 330 66
6 924 462 80
7 792 462 66
8 495 330 43
9 220 165 19
10 66 55 6
11 12 11 1
12 1 1 1
Note: Not every author provided the numbers exactly as given in the respective column.
For example, Bruno Weigl counted 454 chord classes of cardinal 6, instead of 462. Joseph
Matthias Hauer, whose classification was not strictly one of chords, did not consider other
cardinalities than 6.
forms of a D-major triad belong to one class, whereas a D-major triad and an A-
major triad fall into different classes. It is difficult to determine whether this was
a choice of musical principle or the consequence of the chosen algorithm. In
view of Klein’s pursuit of objectivity, the answer is, perhaps, both. Nevertheless,
it is worth taking a closer look at his method. For reasons that will become obvi-
ous, Klein could only demonstrate its application to “dyads” (Zweiklänge) and
“triads” (Dreiklänge; Klein 1925, 282). Here it is rendered so as to facilitate com-
parison with the methods of others.
Table 4.4a shows how Klein derived his 66 dyad classes from the total chromatic.
In a first cycle, he combined the lowest tone of an octave range (C) with each of
the remaining eleven tones in that range (from B down to D♭). He then pro-
ceeded to the next cycle, combining C♯ with all others tones, except, of course, C.
After that, there followed combinations with D (excluding those of C and C♯), D♯

118 equivalence
Table 4.4a. Klein’s algorithm applied to dyads
B B B B B B B B B B B
B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ A♯
A A A A A A A A A A 11th cycle (1)
A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ G♯ 10th cycle (2)
G G G G G G G G 9th cycle (3)
G♭ G♭ G♭ G♭ G♭ G♭ F♯ 8th cycle (4)
F F F F F F 7th cycle (5)
E E E E E 6th cycle (6)
E♭ E♭ E♭ D♯ 5th cycle (7)
D D D 4th cycle (8)
D♭ C♯ 3rd cycle (9)
C 2nd cycle (10)
1st cycle (11)
(excluding those of C, C♯, and D), E (excluding those of C, C♯, D, and D♯),
etc. Thus, Klein’s algorithm consisted of eleven successively shorter cycles.
When the same method is applied to triads, it becomes more complex. As
appears from table 4.4b, each cycle is then built from of a number of subcycles. All
combinations from table 4.4a, except those with B, combine with the remaining
tones in the same octave range to yield Klein’s 220 classes of “triads,” redundan-
cies omitted.
Klein confessed his surprise at the total number of 4,095 chord classes of car-
dinality 1 through 12, a number much smaller than he had expected. He prob-
ably would have been astonished to hear that Erwin Stein had found only half
as many. A few months earlier, Stein, a student of Schoenberg, had contributed
an essay to a special issue of the periodical Musikblätter des Anbruch, published
on the occasion of his teacher’s fiftieth birthday. This essay was the first printed
exposition of Schoenberg’s twelve-tone method, laying much stress on the his-
torical conditions by which it was spurred. In the introduction, Stein welcomed
the maximization of harmonic resources ensuing from the “emancipation of the
dissonance” (one of Schoenberg’s tenets). With an air of offhanded authority
he advised how many chords were now at the free disposal of the contemporary
composer writing in a twelve-tone idiom:
Thus we get 55 constitutionally different three-note chords, 165 four-

note chords, 330 five-note chords, 462 six-note chords, the same number
of seven-note chords, and again 330 five-note chords, 165 nine-note, 55

Table 4.4b. Klein’s algorithm applied to triads
B B B B B B B B B B B B B B B B B B B B B B B B B B B
B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭
A A A A A A A A A A A A A A A A A A A A A A A A
A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭
G G G G G G G G G G G G G G G G G G
G♭ G♭ G♭ G♭ G♭ G♭ G♭ G♭ G♭ G♭ G♭ G♭ G♭ G♭ G♭
F F F F F F F F F F F F
E E E E E E E E E
E♭ E♭ E♭ E♭ E♭ E♭
D D D D
D♭ C♯ 3rd cycle (36)
C 2nd cycle (45)
1st cycle (55)
B B B B B B B B B B B B B B B B B B B B B B B B B B B B
B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭ B♭
A A A A A A A A A A A A A A A A A A A A A A
A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ A♭ G♯ 10th (1)
G G G G G G G G G G G 9th (3)
G♭ G♭ G♭ G♭ G♭ G♭ F♯ 8th cycle (6)
F F F F 7th cycle (10)
E E 6th cycle (15)
D♯ 5th cycle (21)
4th cycle (28)
10/2/2008 10:31:10 PM
120 equivalence
ten-note, 11 eleven-note chords, and one twelve-note chord, i.e. alto-

gether over 2,000 chords. 31
Stein did not reveal his algorithm, if he had any. He might just have heard about
these numbers. Literature on the subject was available at the time—as appears
from table 4.3—but it was probably not widespread. Most classifications seem to
have come about without prior knowledge of similar efforts, as Klein’s amaze-
ment at the results of his own count illustrates.
Someone who certainly knew how to obtain Stein’s numbers was his contem-
porary Bruno Weigl. This composer and music theorist from Brünn (Brno), in
the Moravian region of what is now the Czech Republic, had spelled out all the
different harmonic combinations within the total chromatic, arriving at the
same result reported by Stein.32 His list, which shows chords of three to five
tones in staff notation while the rest are rendered in letter notation, covers
almost fourteen pages of his Harmonielehre (1925).33 Although this book was
written with the aim of establishing a continuity between the music of the past
and that of the present—an aim which it shared with the harmonic treatises of,
among others, Bernhard Ziehn (Harmonie- und Modulationslehre, 1887), Rudolf
Louis and Ludwig Thuille (Harmonielehre, 1907), and Hermann Erpf (Studien
zur Harmonie- und Klangtechnik der neueren Musik, 1927)—it could hardly repre-
sent the most advanced achievements of contemporary practice beyond this
huge catalogue. However, it is not least because of this catalogue that it claims
the interest of the historian.
How did Weigl establish his chord classes? Which tone combinations did he
consider equivalent? If one looks at his “table of triads” (Dreiklangstabelle; Weigl
1925, 379), he seems, at first, to proceed in the same way as Klein (see table
4.4b). But he stops after the first cycle, seeing all possible combinations of three
tones represented therein. Now, since the first cycle involves all chromatic tones,
and since all combinations in it are built on the same “bass” (C), we can make
the following observations:
31. “So gibt es 55 ihrer Konstitution nach verschiedene dreistimmige Akkorde, 165
Vierklänge, 330 Fünfklänge, je 462 Sechs- und Siebenklänge, und wieder 330 Acht-, 165
Neun-, 55 Zehn-, 11 Elfklänge und einen Zwölfklang, ingesamt über 2000 . . . Akkorde.”
(Stein 1925, 60; transl. by Hans Keller)
32. Weigl’s number of six-tone combinations (454) differs from Stein’s (462).
The explanation is simple. Weigl forgot to list the following eight combinations:
(C,D♭,E,A♭,B♭,B), (C,D♭,G♭,G,A,B♭), (C,D♭,G♭,G,A,B), (C,D♭,G♭,G,B♭,B),
(C,D♭,G♭,A♭,B♭,B), (C,D♭,G,A♭,B♭,B}, (C,D,G,A♭,B♭,B), and (C,E,F,A♭,B♭,B)
33. A note by the author (p. IX) states that the publication of the Harmonielehre, by
the Schott firm in Mainz, was originally scheduled for 1922. Due to the financial crisis in
Germany, the book had to wait until 1925 to appear and was abridged to about two-thirds
of its original size.

equivalence 121
(1) Each combination is transpositionally related to two other combinations

in this cycle, except one combination that is self-reproductive under trans-
position (i.e., {C,E,A♭}).
(2) No two transpositionally related combinations appear in the same bass

position.
From this it follows that each combination in the first cycle is uniquely identified
with a set of pitch intervals measured from the “bass.” It represents all chords
that yield the same values when these PIs are assigned to PICs. Thus, in Weigl’s
view, two chords are equivalent when they are in the same bass position (or har-
monic inversion). This may or may not be true for two forms of a D-major triad;
and it may or may not be true for a D-major triad and an A-major triad.
Weigl’s view of a chord as a set of intervals relative to a bass is reminiscent of fig-
ured-bass practice. We may ask whether this was not a mere coincidence, the out-
come of a combinatorial protocol that, as Catherine Nolan points out, adds eleven
elements to “a disjunct, fixed element . . . in all possible combinations” (Nolan
2003, 214). It may have been so for Anatole Loquin, whom Nolan credits as the
first to apply this protocol to the elements of the chromatic scale. Loquin used the
resulting partition of the collection of all possible chords (our PSET) as a spring-
board to a partition of this collection into T-classes. And it may also have been so
for Erwin Stein, who was more than familiar with Schoenberg’s manipulation of
intervals and therefore conceivably receptive to the use of T/TI-classes.
For Weigl, however, the correspondence between his chord classification and
one along the lines of the figured-bass tradition was not entirely coincidental.
In the last section of his Harmonielehre, he retreated explicitly from the Rame-
auian, “root-oriented” concept of a chord, which he thought to be of no use with
respect to the chromatic idiom (Weigl 1925, 394–95). Since in this idiom each
tone could be combined with any other, he argued, the interpretation of a chord
as “root position” or “inversion” had lost its rational basis. These forms had now
become types. This explains why Weigl’s classification reflects a “bass-oriented”
concept of a chord. Such a concept facilitated a comprehensive listing of chords
while remaining relatively faithful to their sonic identities. Meanwhile, Weigl
realized that sonic identities were also shaped by their contexts. He cautioned
that his list did not account for the relations into which chords might enter. This,
he wrote, would have required him to include each chord in all its enharmonic
spellings (Weigl 1925, 378). Weigl was obviously closer to nineteenth-century
harmonic theory than to twentieth-century PC set theory, the visionary aspects
of his chord classification notwithstanding; but as to the limitations of a taxon-
omy of chords, he raised a point that would become a focus of criticism of PC set
theory in later years.
Weigl’s decision not to classify each chord of the chromatic scale with all its
bass positions was certainly rational. However, it was still a choice made from

122 equivalence
two possible options. For why would the absence of a root—or the impossi-
bility of assigning this role to a tone—preclude the association of chords on
the basis of their type? When the different bass positions of a chord have no
harmonic function to share, or no locus in a key, they still represent the same
“chord,” in this sense: a PC set with such-and-such intervallic properties, which
represents a harmonic entity. This definition may not capture the traditional
notion of a chord in all its complexity; but at least it points to an important
aspect of it—sufficiently important for writers other than Weigl to use it alone
as a basis of equivalence.
These others—like the composers Ernst Bacon and Josef Matthias Hauer,
and the musicologist Walther Howard—classified tone combinations accord-
ing to what John Rahn has called “Tn-type,” obtaining the numbers in the
right-most column of table 4.3 (cf. the left-most column of table 4.2). They did
not all speak of actual “chords.” Bacon’s numbers do refer to simultaneities,
but the essay in which he presented them, “Our Musical Idiom,” was a specula-
tive exercise rather than a record of compositional practice. Published in 1917
in the Monist, an American journal of philosophy, it had little impact on the
musical world around him. Today, its importance rests mainly in its adumbra-
tion of PC set theory. Bacon represented simultaneities as successive-interval
arrays partitioning the octave, as Richard Chrisman did with pitch sets many
years later (see above, p. 100). Like Chrisman, Bacon formed classes of such
arrays related by cyclic permutation.
Hauer’s classification is actually one of twelve-tone series, which he reduced
to complementary pairs of unordered hexachords, so-called “tropes” (Tropen).
There are 44 such pairs, which Hauer tabulated in his small composition treatise
Vom Melos zur Pauke (1925, 12). As eight tropes consist of transpositionally related
hexachords,34 there are eighty (88 – 8) distinct hexachord types. Hauer did not
enumerate tone combinations of cardinalities other than 6. The use of tropes
ensured the equal distribution of PCs in a composition, in accordance with the
“twelve-tone law” (Hauer 1925, 17). Within each hexachord, tones could sound
simultaneously or in succession; Hauer’s hexachords functioned equally in the
harmonic and melodic dimensions of music.
Walther Howard saw his tone combinations in a similar way—as the raw mate-
rial for harmonies as well as melodies (although he referred to them as “chords”
or “simultaneities”). Howard was the author of a Wissenschaftliche Harmonielehre des
Künstlers, which was published in Berlin in 1932. His aim was to present a theory
not so much concerned with musical uses and styles as with the relations embod-
ied by the pitch material itself (i.e., the twelve-tone chromatic). He expected this
approach to provide a basis for any possible or desirable application, and thus
34. In six of these tropes both hexachords are inversionally symmetrical. Therefore,
they relate not only as transpositions but also as transposed inversions. This relation, how-
ever, also exists between some hexachords from different tropes.

equivalence 123
to be most fit for pedagogical purposes (Howard 1932, 1). Counting and mea-
suring were methods that, in his view, ensured a fresh and unbiased attitude
towards music. In his book, Howard spelled out all T-classes of three-tone and
four-tone combinations and provided the numbers of classes of other cardinali-
ties (Howard 1932, 215).
The concept of a T-class lacks a clear sense of verticality, such as provided
by Weigl’s referential bass tone. Indeed, the members of such a class are sets
rather than chords. However, we must be careful in establishing a relationship
between the above-mentioned chord classifications and the concept of a PC set.
The compilers of these classifications were not involved in the development of
PC set theory, even though they used similar techniques of investigation. They
developed their systems in relative isolation, unaware of others who embarked
on similar projects and unable to create a stir. And their systems served differ-
ent goals. Seen thus, it is hard to maintain that they laid “important conceptual
groundwork for pc-set theory” (Bernard 1997, 12). Rather, the rise of PC set
theory has given us a focus on these unconnected attempts to come to terms
with the vastness of harmonic resources in the twelve-tone universe.
Sound
In the 1980s, the criteria of PC set equivalence, as a topic of scholarly debate,

had practically been exhausted. A general consensus of opinion had emerged
regarding the relations that qualified as “equivalencies.” Moreover, people
began to realize that the actual use of this designation required a good deal of
common sense. This made the discussion about its mathematical underpinning
more or less obsolete. In his book Composition with Pitch-Classes from 1987, Rob-
ert Morris wrote:
The question of which operations to choose is guided by the attempt to provide

a way of collecting “similar”-sounding sets into an equivalence class. Therefore,
not every (mathematically) competent definition will do. (Morris 1987, 79)
Morris himself may have been guided by his own perception in his efforts to
define useful equivalencies among pitch sets. Pitch sets are obviously not as
abstract as PC sets. Pitches and pitch intervals are more specific descriptors of
tone combinations than PCs and PICs. The latter help us discover certain con-
cealed relations between tone combinations, but relations of that kind are not
essential to every piece of music. As we have seen in our discussion of example
4.9 (the opening measures of Schoenberg’s Opus 33a), the use of PCs may some-
times obscure essential musical relations. John Rahn (1980a) already mentioned
the possibility of defining equivalencies in PSET, but he did not go beyond that.
Morris (1995) proposed three new class types. The first of these was the “pitch-

124 equivalence
set class” (PSC). A PSC contains pitch sets with the same intervallic spacing (i.e.,
the same succession of pitch intervals from low to high). Such pitch sets are, for
example, {D4,F4,A♭4,C♯5}, {F3,A♭3,C♭4,E4}, and {B2,D3,F3,A♯3} (ex. 4.11a).
To reduce the huge amount of possible PSCs, Morris suggested setting a limit to
their vertical spans. Any two members of a PSC are related by an operation on
PSET (Tn, where n is a positive or negative integer). For example:
T−9({D4,F4,A♭4,C♯5}) = {F3,A♭3,C♭4,E4}
The second class type in PSET was the “pitch-class interval succession” (PCINT).
A PCINT-class comprises PSCs with octave-related successive spacing intervals.
The PCINT-class of which {D4,F4,A♭4,C♯5} is a member also contains the pitch
sets {D3,F4,A♭5,C♯6} and {G3,B♭3,D♭5,F♯5} (ex. 4.11b). The last two pitch sets
are not related to {D4,F4,A♭4,C♯5}, or to each other, by Tn.
The third, and most inclusive, of Morris’s class-types was the “figured-bass”
(FB-) class. This is the class-type also invoked by Weigl (1925) and, before him,
Loquin (1874): an unordered set of PICs measured from a given bass (see table
4.3). In the figured-bass tradition, octave-related pitch-interval values—like a uni-
son and an octave, a second and a ninth, or a fifth and a twelfth—were congru-
ent mod 7. The chord symbol “6/3” indicated that each chord tone was related
to the bass by an interval congruent 6 (mod 7) or an interval congruent 3 (mod
7) if it was not a “doubling” of the bass. Morris proposed a similar labeling of FB-
classes partitioning PSET, which required the use of modulo-12 arithmetic.
Let us continue with the earlier examples. We have seen that {D4,F4,A♭4,C♯5},
{D3,F4,A♭5,C♯6}, and {G3,B♭3,D♭5,F♯5} belong to the same PCINT-class. They
also belong to the same FB-class, a class that includes pitch sets from different
PCINT-classes, for example {D4,C♯5,A♭5,F6} and {G2,B♭3,D♭4,F♯4,B♭4} (ex.
4.11c). In all these sets each pitch is related to the lowest pitch (the “bass”) by an
interval congruent 3 (mod 12), an interval congruent 6 (mod 12), or an interval
congruent 11 (mod 12), unless they represent the same PC as the “bass.” The
common designation for pitch sets with this property—that is, the “name” of
their FB-class—is “3 6 11.”35
The pitch sets of one FB-class are all realizations of transpositionally related
PC sets. These realizations are more likely to be heard as similar than other real-
izations of the same PC sets, especially when they are also members of the same
PCINT-class, or the same PSC. They satisfy certain criteria of presentation that
help the listener to relate them, like the spacing intervals or the pitch intervals
formed with the “bass.” These are not just aural criteria; Morris’s classification of
pitch sets bears a resemblance to the classification of chords in the music of the
common-practice period. It thus reflects a tradition that is still represented in
35. In Morris’s notation the PIC values 10 and 11 are represented by A and B, respec-
tively. His designation for this FB-class is “36B.”

equivalence 125
(a) (b) (c)
PSC
PCINT
FB
Example 4.11. Class types of pitch sets, after Robert Morris (1995).
concert life, in some branches of music production, and in methods of theoreti-

cal instruction. Therefore, one may feel a familiarity with Morris’s classification
as a conceptual device, while still being daunted by the sonorities themselves.
In the various definitions of PC set equivalence that we have discussed, “sound”
has only casually been mentioned, if at all. Theorists may have taken it for granted
that there is a correlation between PC set equivalence and aural similarity. However,
this correlation is questionable, as the case of dyadic sets clearly shows. Various har-
monic intervals are associated with a single class of such sets. Does this mean that
they are similar? Does a minor seventh sound similar to a major second, or does a
major tenth sound similar to a minor sixth? The relationships between these inter-
vals may be perceived on first hearing, but this depends on the musical context,
which could instead obscure such relationships. In tonal music, a major second and
minor seventh both tend to resolve, and the resulting interval progressions (diatoni-
cally: 2–3 and 7–6) are related by inversion at the octave. Thus, in a tonal context
we hear these intervals as if resulting from voice exchange. However, if the tonal
context is abandoned, we can no longer take their similarity as obvious.
In PC set theory, the concept of “interval content” has always been supposed
to account for a particular sound that is intrinsic to all musical realizations of a
PC set. Every other PC set with the same interval content would hence produce
the same “basic” sound. This involves the transpositions of the original PC set, its
inversion, and the transpositions of its inversion. It may also involve “Z-related”
sets. Straus (1990a, 67) wrote, “[sets] in the Z-relation will sound similar because
they have the same interval content.” Although it is hard to prove that there is
always such a common “basic” sound, the concept of interval content may help
one imagine it. It is true, there are some familiar cases in which a peculiar and
relatively invariant sound is reflected by a highly uneven distribution of intervals.
These cases include, for example, the PC sets of the “augmented triad”- type (APIC
vector 000300), the “diminished seventh chord”-type (APIC vector 004002), and
the “whole-tone scale”-type (APIC vector 060603). Furthermore, it is possible that
conventions of tonal hearing make us particularly insensitive to the similarity of
interval content between, say, the “Tristan chord” and a dominant-seventh chord

126 equivalence
(APIC vector 012111 in both cases). However, the sonic profile pertaining to a
specific interval content is, by and large, strongly variable. This variability should
not always be disregarded in the search of relations between pitch-class sets. The
following will illustrate this point.
In The Structure of Atonal Music Allen Forte posits a relation between the two
chords in example 4.12a. The first chord, α, has been taken from Alban Berg’s
song Sahst du nach dem Gewitterregen, one of the songs on picture postcard texts
by Peter Altenberg, Op. 4. The second chord, β, is from Stravinsky’s Symphonies
of Wind Instruments. Without suggesting a relationship between the two pieces,
Forte points out that the PC sets of these two chords reduce to the same prime
form (they are related as transpositions; Forte 1973, 12).
Forte’s two examples strike the reader as being very randomly chosen, but on
further investigation they do make an interesting comparison. First of all, both
constitute strongly articulated musical events, and, what is more, both are unify-
ing elements in their respective musical environments. Being sustained by the
strings for at least an entire measure, while the voice and the other instruments
remain silent, chord α effects a caesura of great power in Berg’s eleven-measure
song. Its PC set is inversionally related to that of one of the song cycle’s principal
melodic ideas: the line G–A♭–B♭–C♯–E (cf. the opening measures of the first and
last songs). Chord β is one of the recurring elements in Stravinsky’s Symphonies of
Wind Instruments, dominating the final section of this composition.36 It stands out
as a musical gesture, yet as a configuration of intervals it is tightly interwoven with
other elements of the Symphonies. For example, the upper section of chord β con-
tains the PCs G, B, D and F, which outline the melody of the opening refrain.
Both chords can be related to an octatonic background. If we view the octa-
tonic scale (see ex. 1.4) as a PC set, it is self-reproductive, not only under inver-
sion, but also under T3, T6, and T9. This means that there are only three such
sets. Pieter van den Toorn (1983, 31–98) called them “Collection I”: {C♯,D,E,F, . . .
A♯,B}, “II”: {D,E♭,F,F♯, . . . B.C}, and “III”: {D♯,E,F♯,G, . . . C,C♯}. The constitu-
ent five PCs of Stravinsky’s chord (β) form a part of Collection I. The chord itself
results from a transposition within this collection. It is the T9 of the chord that is
first presented just after the opening refrain of the Symphonies (ex. 4.12b).37
In mm. 3–5 of Berg’s song, all three octatonic collections are exposed in an
intricate web of short motifs. Although this passage exploits the total chromatic,
it is still possible to delineate these collections. As example 4.12c shows, they
are accentuated by the instrumentation. None of the instrumental combinations
(voice/harp/clarinet, horn, bassoon/viola, horns/viola) produces notes other
than those belonging to one collection (either I, II, or III).
36. This final section was published separately, in a piano reduction, as a contribu-
tion to a volume of the Parisian journal La Revue Musicale dedicated to the memory of
Claude Debussy (vol. 1/2, 1920, 22–23).
37. Here I follow the analysis of Taruskin (1996: 1494).

α 56 Tempo I = 72
β
Example 4.12a. Chords from Alban Berg’s Sahst du nach dem Gewitterregen, Op. 4, no.
3, and Igor Stravinsky’s Symphonies of Wind Instruments.
1 56 Tempo I = 72
T9
Octatonic collection I
Example 4.12b. Octatonic derivation of Stravinsky’s chord (after Taruskin). On the
lowest staff, black note-heads represent actual chord tones.
Coll. III Al les 4 ra stet . . . 5
hrp + clar.
3 Coll. I
3
Coll. II hrns. + va.

m.3
hrn. bn. + va.

Example 4.12c. Octatonic elements in Berg’s song.

128 equivalence
However, in spite of their common conceptual background, we should con-

sider what distinguishes these two chords. It is noteworthy that Berg has so
clearly projected the PC set {E♭,F,G♭,A,C} as a dominant-flat-ninth chord; with
the “root” (F) in the bass parts, the “fifth” (C) coming next in ascending order,
and the “ninth” (G♭) in the treble parts. One may object to the designation
“dominant-flat-ninth chord” because it has been adopted from the tonal idiom.
Indeed, this is not the idiom of this song; but it is not far away, and Altenberg’s
words provide an occasion to bring it to mind:
After the summer rain did you see the forest?!?!

All is glitter, quiet, and more beautiful than before.38
....................................
We can hear Berg’s chord as illustrating these words, as a sonority that has been
relieved from its traditional function of being dissonant with respect to other
sonorities, and that now exists only for its own sake.
In contrast to Berg, Stravinsky has laid out the elements of the transposi-
tionally related PC set {F,G,A♭,B,D} in such a way that the listener is not so eas-
ily reminded of a dominant-flat-ninth chord. Chord β would be in an unusual
position, with the fifth (D) in both bass and treble. Although Stravinsky’s
chord is not entirely devoid of tonal elements either (NB the major triad in
the upper voices), it is not as suggestive of functionally directed harmony as
Berg’s. It is by the relative autonomy of sound vis-à-vis PC set that these “equiv-
alent” chords are yet so different—as different as, for example, statues carved
from the same stone.
The concept of PC set equivalence embodies a view of musical coherence as
independent from the limits of aural perception. However, we have seen that it
has its own limits: it is rather indifferent to distinctions of sound. This indiffer-
ence is inherited from music theory rather than from mathematics. The search
described in this chapter was not a search for musical relations that satisfy a
mathematical definition. It was primarily a search for mathematical formula-
tions. These should underpin already well-established notions of musical relat-
edness, which could also help extend their range of application. Most of these
notions involve operations that preserve what are, in specific contexts, essential
characteristics of a tone combination—operations like the registral redistribu-
tion of PCs (context: figured-bass practice, theory of harmony), permutation/
harmonic inversion (context: double counterpoint, theory of harmony), trans-
position, mirror inversion (context: imitative counterpoint), and interval expan-
sion (context: melodic variation).
38. “Sahst du nach dem Gewitterregen den Wald?!?! / Alles rastet, blinkt, und ist
schöner als zuvor.” (transl. by A. Kitchin)

equivalence 129
PC set equivalence assumes the existence of a musical identity that remains

the same under these operations. Such a persistent identity can only be very
abstract; it has proven difficult to locate even in the PC sets themselves. One pos-
sible locus of identity, the total interval content, was abandoned as a criterion
for the classification of PC sets. What united such sets was to be found between
rather than in them. However, there remained an interest in relations between
PC sets on the basis of their intrinsic properties. “Similarity” was a concept by
which one hoped to approximate these relations.

Chapter Five
Similarity
The discussion of “similarity” has been one of the major threads running
through PC set theory. It revolves around the question of whether PC sets, when
they are not connected by a more or less obvious operation, or do not share a
more or less obvious property, can still be “related.” The aim has been to model
such relatedness, and thus render it tangible. There is no apparent link with the
familiar geometrical concept of similarity, which refers to the equal proportions
of different-sized objects. Nor is there a link with the set-theoretic relation of the
same name, which involves the order relations in two sets. The most likely source
of the concept is information science, especially insofar as it deals with uncer-
tainty. Similarity relations are defined there as “fuzzy” relations. They involve a
degree of imprecision, and thus better approximate everyday human thought
and speech than “crisp” equivalence relations (Klir and Folger 1988, 83–85.).
In PC set theory, “similarity” is an umbrella term covering concepts with dif-
ferent origins, ranging from geometry to statistics. In this chapter, these con-
cepts will be analyzed with respect to their underlying questions, intuitions, and
assumptions. None of them has become the “most common” concept of PC set
similarity; nor do they all refer to the same properties. PC sets can be similar in
different ways; generally speaking, however, similarity is a matter of degree. In
that respect, similarity relations are different from equivalence relations. PC sets
are either in an equivalence relation or not, but they can be more or less similar.
The study of PC similarity has been motivated by the prospect of a subtle and
flexible analytical framework.
The terms “equivalent” and “similar” may both be appropriate to a single pair
of PC sets. In other words, they are not mutually exclusive; they reflect different
ways of thinking about relations. However, similarity relations involve a much
greater number of PC sets: any pair of PC sets will be more or less similar on
some scale. This inclusiveness has been another reason to search for reliable
measures of PC set similarity.
Concept and Referent
What does it mean when two PC sets are called “similar”? How do these PC sets
resemble one another? And in what regard can PC set similarity be meaningful?
A PC set can be realized in a multitude of ways, causing different musical sensations.

similarity 131
Although in ordinary language the word “similarity” denotes a resemblance, the

similarity of two PC sets may not be apparent to a listener at all. It is, in other
words, an abstract relation, regardless of how one defines it. This is one of the
puzzling aspects of the debate on PC set similarity. A tool to describe relations
beyond techniques of direct musical derivation, the similarity concept might
account for the intuitive perception of a unified harmonic-melodic vocabulary.
But how can it do that if such perception is dependent upon the variables of
musical reality; upon the registral and temporal distribution of PCs; upon articu-
lation and syntax? And if PC set similarity does not tell us something significant
about our perception of a work, what can we learn from it?
The earliest texts dealing with PC set similarity did not raise this question.
The concept was brought into play without a proper introduction. Forte started
his discussion of it as follows:
To define an appropriate, musically meaningful similarity relation for any two

sets of different cardinality is a task beyond the scope of this paper . . . It is
possible, however, to define a simple yet effective criterion for similarity in the
case of two sets of the same cardinality. (Forte 1964, 149–50)
Forte did not state what he meant by “musically meaningful” and “effective.”
From his application of the concept—in an analysis of No. 4 of the Five Move-
ments for String Quartet, Op. 5 by Anton Webern—one can infer that he wanted
to come to grips with the relations between different equivalence classes of PC
sets. It is not difficult to see the reason for this. If equivalence (as defined on
the basis of transposition, transposition and/or inversion, or the equal-vector
relation) were the sole criterion of PC set-relatedness, it would be hard to prove
that PC sets are significant as units of structure in post-tonal music. Often, there
are relatively few equivalent pairs among the PC sets that can be extracted from
a musical composition (considering the most straightforward grouping criteria).
Therefore, if PC sets are important, it is logical to look for more possible con-
nections between them than pure equivalence.
The above suggests that measures of similarity supported the concept of
PC sets as real musical objects, while providing tools for their study. It sug-
gests, in other words, that analytical concepts may not only clarify musical
structures, but also enable, or uphold, an analytical practice, more or less for
its own sake. Indeed, this is a possibility we cannot ignore, particularly since
other authors working in the field of PC set theory have been just as reticent
as Allen Forte concerning the musical insights gained by the use of a similar-
ity measure.
In his search for “generally applicable structural criteria in the atonal works of
Schoenberg, Webern, and Berg,” Richard Teitelbaum investigated several relations
between the “harmonic collections” (i.e., the PC sets) in these works, including
“the degree of similarity (or dissimilarity) of intervallic distribution” (Teitelbaum

132 similarity
1965, 74–75, 88). However, he was equally concerned with the use of computers in
music analysis. As Teitelbaum wrote, the application of the computer was appro-
priate in the laborious process of determining and comparing the distribution of
intervals in a large number of PC sets (Teitelbaum 1965, 76). It is possible that
such an analytical endeavor was not only enabled, but also inspired by the avail-
ability of the computer. A computer can only deal with questions that lend them-
selves to algorithmic formulation. If analysts want to use it, they must raise such
questions. The question of which interval classes are associated with a particular
type of PC set belongs to this category. It translates into the following algorithm:
• List all unordered pairs of the PC set’s elements.
• Determine the interval class represented by each of these pairs.
• Compute the number of occurrences of each class.
• Order these numbers by class, from APIC 1 to APIC 6.1
The result of these consecutive operations is the familiar APIC vector (see chap-
ter 2, p. 48). Teitelbaum also used a protocol returning a value for the relation
between two different APIC vectors, a value called “similarity index.” This will con-
cern us later. The main point here is that he made no attempt to argue the need
for such a protocol; he just claimed his similarity index to be “a significant indica-
tor of the similarity or dissimilarity of interval content between two harmonic col-
lections” (Teitelbaum 1965, 88). Its significance, however, may have had a lot to do
with the state—and the status—of computer research (see further chapter 8A).
We must realize that at the time of Forte’s and Teitelbaum’s early explorations
in the field of PC set similarity little was known about the early atonal repertoire
that was their main research object. They were searching for adequate means of
description. Their similarity measures served as heuristic devices, scaffolding for
yet unknown determinants of musical coherence. Actually, this holds true for
PC set theory at large. A PC set analysis may reveal the consistent use of certain
intervallic configurations in a musical composition, but it may also reveal the
opposite. In both cases, the concept of a PC set has proven its worth, not as the
signifier of actual units of structure, but as an analytical tool.
This would mean that similarity measures should be seen—and judged—as
refinements of what is basically a heuristic apparatus. However, the heuristic
tools soon attained the status of descriptive models. Writing in 1964, Allen Forte
could not restrain himself from blurring the fundamental distinction between
these two categories:
The theory of set-complexes has special significance for the study of syntactic
structure in the case of atonal music, and work is now in progress to refine and
1. Cf. Teitelbaum’s description of this process (1965, 78).

similarity 133
test related research tools, with the expectation that useful general analytic
methods and descriptive techniques will result. (Forte 1964, 178)
The relationship between “analytic methods” and “descriptive techniques” is

unclear. It is possible that the former term refers to the heuristic dimension of
PC set theory; but Forte may have just wanted to draw a distinction between, on
the one hand, analyzing music and, on the other, communicating the results of
an analysis. In the preface to The Structure of Atonal Music, the heuristic function
of PC set theory is not even mentioned:
[I]t is the intention of the present work to provide a general theoretical frame-
work, with reference to which the processes underlying atonal music may be
systematically described. (Forte 1973, ix)
Forte assumes a correspondence between the theoretical framework and the

musical processes or structures. In other words, the theoretical framework is sup-
posed to be fully descriptive. But to which kind of process or structure do simi-
larity relations correspond? The PC set is itself a problematic concept, as we have
seen in previous chapters, but at least it is tied to music history, including the
history of music theory. It can be shown to derive from serial composition, and
although it is not actually an outgrowth of the chord or the scale, it has much in
common with them. Furthermore, when we identify PC sets and trace transpo-
sitional, inversional, complementary, and inclusional relations between them,
we are in the analytical tradition that has developed at professional schools of
music, and that has had a formative influence on composers, performers, schol-
ars, teachers, and editors alike. Whatever the geographical variety, this tradition
has always focused strongly on relations involving themes, motifs, harmonies,
and keys (correlative to its focusing on the repertoire of the common-practice
period). Notwithstanding its initial alliance with contemporary composition, PC
set analysis has established itself as an extension of this analytical tradition, and
it carries a good deal of weight derived from it.
This is not the case with similarity relations. PC set similarity was originally
a concept without a clearly defined function, and without a history that would
render it authoritative. This situation has never really changed. New similarity
measures have been introduced because they were more discriminating than
previous ones, or involved a wider range of PC set pairs.2 The relevance of PC
set similarity itself was taken for granted.
Robert Morris created a similarity index (labeled SIM) with the explicit
aim of approximating the judgment of the ear. According to Morris—who
2. See, for example, Isaacson (1990, 2), who had formulated three criteria for the
measurement of intervallic similarity: it should “1) provide a distinct value for every pair
of sets; 2) be useful . . . for sets of any size; 3) provide a wide range of discrete values.” See
further below.

134 similarity
had taken an interest in PC set theory primarily as a composer—SIM would

“provide a rationale for the selection of sets that insure predictable degrees of
aural similitude” (Morris 1980, 446).3 Here he is on thin ice. Aural similitude
always goes hand in hand with aural difference. It is the analyst’s decision to
stress the similitude. This decision is necessarily based on a particular concep-
tion of what is being heard.
For example, it is possible to stress the sameness of the chords C–E–G and E–
G–C by reference to their constituent “pitch-classes” (as Jean-Philippe Rameau
did); but it is just as possible to stress their distinctness, by reference to their
intervallic structures (as Gioseffo Zarlino did). Neither view is the negation of
the other. Rather, each rests on a different valuation of a single set of proper-
ties.4 In this case, these valuations correspond to certain structural or syntactic
functions. In other words, they concern the “nuts and bolts” of past composi-
tional and improvisational practices. PC set similarity, however, has always lacked
such a clear musical referent. Therefore, equating PC set similarity with aural
likeness is bound to remain a matter of faith.
Similarity measures have attained a high degree of sophistication. In combi-
nation with their lack of an obvious source of validation, this renders them unfit
for pedagogical purposes. Similarity is thus also an area of PC set theory falling
outside the traditional scope of music analysis (which has always had a generally
pedagogical orientation). In many texts dealing with this topic, the questions of
analytical procedure have been taken for granted as much as the questions of
musical relevance.
In The Structure of Atonal Music, Allen Forte—in many ways still a traditional
music theory teacher—offered a set of musical examples showing instances of
his R0-, R1-, R2-, and Rp-measures of PC set similarity (ex. 5.1; Forte 1973, 59; see
further below). These examples were probably not meant as evidence of the fre-
quency of these relations. At the time, Forte used to apply not empirical but logi-
cal criteria of significance. He considered a similarity relation worthier of note
the less likely it was to occur between any two PC sets. The question of what it
meant to encounter it in a work seems to have been of minor importance to him.
As to example 5.1, he made no effort to clarify the method of his analysis, or to
state its purpose. This musical example, like many others in his book, served
merely as illustration.
3. In a footnote added to his article—which responded to the criticism passed on an

earlier, spoken version of it—Morris confided: “I have been and still am concerned with
attempting to provide a similarity relation which corresponds in some way to what I—at
least—hear when two sets are presented in some similar musical setting.” (Morris 1980,
458) His “similarity index” was tested empirically by Cheryl Bruner (1984), who judged it
too narrowly defined to reflect a listener’s aural experience.
4. See Carl Dahlhaus’s discussion of this topic with reference to the theories of
Zarlino and Rameau. (Dahlhaus 1968, 211–14.)

similarity 135
61. Scriabin, Seventh Piano Sonata Op. 64 5
(a)
6-Z49 5-16
[4,6,9,10,0,1] 6-Z19
[1,2,5,6,8,9]
333
(b)
6-Z13 6-Z24 6-34

[9,10,0,1,3,4] [9,10,0,1,3,5] [3,4,6,8,10,0]
62. Similarity relations between parts a and b in example 61

R0, Rp
6-Z49 6-Z19
R1, Rp
6-Z13 6-Z24
6-34
R0, Rp
R0, Rp
Example 5.1. Similarity relations among PC sets in the Seventh Piano Sonata, Op. 64
by Alexander Scriabin, as pointed out by Allen Forte in The Structure of Atonal Music.
Contrary to Forte, Richard Teitelbaum took a truly heuristic approach, using

his similarity index (which he labeled “s.i.”), and other concepts, to characterize a
musical work’s harmonic idiom. The similarity indices resulting from Teitelbaum’s
analyses were not simply documented as analytical results, but as raw observations
from which certain conclusions could be drawn. For example, Teitelbaum noticed
that PC sets in Schoenberg’s early atonal works tended to be similar to a much
lesser degree than the PC sets he had extracted from Webern’s Five Movements
for String Quartet, Op. 5. He presented this as an indication of a difference in
“harmonic practice” between the two composers (Teitelbaum 1965, 108).
Another author who provided an analytical context for his discussion of PC set
similarity was Charles Lord. Lord’s “similarity function” (sf) was almost identical to
Morris’s similarity index, but he emphatically presented it as a corrective to Forte’s

136 similarity
R0-, R1-, R2-, and Rp-measures, claiming that it would give “a more complete and
specific picture of intervallic similarity” (Lord 1981, 109). This claim was justified,
for the sf denoted relative PC set similarity, and thus involved any possible pair
of PC sets. However, with the single musical example he presented—an excerpt
from the Ninth Piano Sonata, Op. 68 by Alexander Scriabin—Lord did not show
the musical functionality of PC set similarity any more than Forte did, particularly
since he had copied this example from The Structure of Atonal Music, complete with
Forte’s segmentation (cf. Forte 1973, 117).
Still other authors, including Regener (1974), Rahn (1980b), Morris (1980),
and Isaacson (1990), presented similarity measures without referring to musical
examples at all. Nevertheless, they were confident about their practical value:
All of the various similarity relations . . . can be combined into one monstrous
graph . . . through which wander all the paths of all the similarity relations.
And since these paths cross—that is, share element-nodes—the adventurous
composer can, with her or his moving finger, trace new similarity paths created
by the concatenations of segments from the crossing paths of the similarity
relations that created the graph. (Rahn 1980b, 497)
IcVSIM, as an intuitively satisfying measure of intervallic similarity in the

abstract, should prove beneficial for concrete applications in the analysis of
atonal music. (Isaacson 1990, 25)
It is not necessary to downplay the speculative origin of the similarity concept;

nor is speculation objectionable. It is perfectly possible for a composer to con-
ceive a musical process along John Rahn’s “similarity paths.”5 In that case, a lis-
tener’s ignorance of the process does not detract from the functionality of the
similarity concept. In music analysis, it is true, this functionality is hard to verify.
However, this in itself does not disqualify PC set similarity as an analytical con-
cept. Even though it is not grounded in pre-existing compositional or analytical
practices, it can still be viewed as a mental image of musical relatedness. And it is
impossible to ban such images from the history of music, since all musical rela-
tions—including the ones that do play a verifiable role in musical practice—rest
on conceptualization.6
5. Rahn’s article inspired a composition by James M. Bennighof. See Bennighof 1984.

6. The relationship between conceptual and perceptual structure is a recurring
topic in discussions about music analysis. Nobody wants to deny the important role of
the human ear when it comes to verifying the results of an analysis. On the other hand, it
cannot play this role without conceptual guidance; and sometimes it cannot play this role
at all. Research conducted by Nicholas Cook in the 1980s revealed that a fundamental
analytical concept like “tonal closure” exerts a relatively weak influence on music percep-
tion, and only within short spans of time (Cook 1987b; for a critical comment on Cook’s
experiment, see Gjerdingen 1999). Of course, it can be useful as an analytical concept,
but it may not tell us anything about perception.

similarity 137
What we need to know is what brings about these “images of musical related-
ness.” Besides a proximity between elements, they involve a purpose, and the
competence to make this proximity serve that purpose. A proximity of elements
may not be discerned if it does not reflect the light that purpose and compe-
tence cast over the music. The origination of Richard Teitelbaum’s similarity
index from an investigation into analytical applications of the computer is a case
in point. Going further back in time, we may think of Schoenberg’s preoccupa-
tion with the principle of “developing variation” in the music of, among others,
Beethoven and Brahms. This principle reflected his own compositional con-
cerns. Joseph Straus (1990b) has depicted Schoenberg as a composer who tried
to bring an omnipresent musical past into line with his own artistic mission. Carl
Dahlhaus did not make such a strong historical claim, but he, too, pointed to
the idiosyncrasy of Schoenberg’s analyses:
The abstraction with which Schoenberg the analyst operated consisted as a

rule in the fact that intervals or complexes of intervals appeared as the true
substance of music, whereas the other features of the composition . . . were
treated as the mere “surface,” more a matter of “presentation” than of “idea.”7
Schoenberg’s strong inclination to close reading reflects his devotion to the musi-
cal work of art as embodied in notation. This attitude was not idiosyncratic. It cor-
related with the work-concept originating in the Romantic aesthetic theory of fine
art. Lydia Goehr (1992) has pointed out the developments in the spheres of music
notation, performance practice, concert behavior, music criticism, and copyright
law by which this work-concept gained footholds in the nineteenth century.
For another example of a pragmatically conditioned analytical perspective,
we may think of Rameau’s basse fondamentale, whose complex and dynamic his-
torical background has been unraveled by Thomas Christensen (1993a). Else-
where, Christensen writes:
To understand Rameau’s conception of the fundamental bass in the Traité

de l’harmonie means reconstructing the musical conditions of his time that
impelled his discoveries, reconstituting the varied pedagogical traditions to
which Rameau was heir . . ., and identifying the intellectual ideas with which
he came into contact. (Christensen 1993b, 17)
Bearing in mind the richness and complexity of conceptualization, we will

now examine a number of similarity measures from the viewpoint of their under-
lying intuitions, and the origins of these intuitions.
7. “Die Abstraktion, mit der Schönberg als Analytiker operierte, bestand in der Regel
darin, daß Intervalle oder Intervall-Komplexe als die eigentliche Substanz der Musik
erscheinen, während die übrigen Tonsatzmerkmale . . . als bloße ‘außenschicht’—mehr
‘Darstellung’ als ‘Gedanke’—behandelt werden.” (Dahlhaus 1986, 282; transl. Derrick
Puffett and Alfred Clayton).

138 similarity
1. Triads consisting of three consonances:
2. Triads consisting of two consonances and one mild dissonance:
3. Triads consisting of one consonance and two mild dissonances:
4. Triads consisting of two consonances and a sharp dissonance:
5. Triads consisting of one consonance, one mild dissonance, and a sharp one:
6. Triads consisting of one mild dissonance, and two sharp ones:
Example 5.2. Krenek’s chord classification (Studies in Counterpoint, 1940).
Gradations of Harmonic Tension
When Forte started discussing PC set similarity in 1964, the topic had already
been raised twice, though in a different connection. Paul Hindemith, in his
Unterweisung im Tonsatz (1937), had proposed a comprehensive classification of
chords employing the twelve notes in the chromatic octave. And Ernst Krenek,
in his Studies in Counterpoint (1940), had classified all possible triads. Strictly
speaking, when chords are assigned to classes, we are dealing with equivalence
rather than similarity. The classifications of Hindemith and Krenek invite a com-
parison with PC set similarity because both composers plotted their classes on
scales of increasing harmonic tension. This means that they brought all chords
into a relationship of greater or lesser similarity.

similarity 139
Hindemith and Krenek gave different explanations for the experience of har-
monic tension. They both pointed to the qualities of the intervals contained in
a chord. Krenek distinguished consonances (PICs 3, 4, 7, 8, and 9), mild dis-
sonances (PICs 2 and 10), sharp dissonances (PICs 1 and 11), an interval that
could be perceived as consonant or dissonant, depending on the musical con-
text (PIC 5), and a “neutral” interval (PIC 6). He must have struggled with these
categories, for his argumentation was loose and contradictory on this point. On
the one hand he treated them as objective gradations of harmonic tension; on
the other, he pointed to the aesthetic, and hence arbitrary, basis of attributes
like “consonant” and “dissonant.”
The degree of the tension may be explained by vibration-ratios, combination-

tones, or other acoustical phenomena; yet, the decision of what shall be con-
sidered a dissonance, and how it should be handled is an arbitrary assumption
inherent in a particular musical style, for it depends exclusively on aesthetic
concepts. (Krenek 1940, 7)
In any case, Krenek correlated the tension of a triad with its constituent inter-
vals. For example, he classified a triad consisting of two consonances and a sharp
dissonance (for example, C4–D♭4–A4) as tenser than a triad consisting of a con-
sonance and two mild dissonances (for example, C4–D4–B♭4; ex. 5.2; Krenek
1940, 19–20).
Hindemith, who had grounded his theory of harmony physically—although
he sacrificed physics for tradition when the two were in conflict (see note 9
below)—arranged the intervals by the extent to which they were “burdened”
(belastet) by their combination tones, specifically their difference tones. When
the best-perceivable difference tones (designated D11 and D21)8 just amplified
the members of the interval, the latter sounded clear and stable. The most
strongly amplified member was designated the “root” of the interval. Thus, C4
was the root of the perfect fifth C4–G4 (D11 = C3 and D21 = C3), a very stable
interval. The perfect fourth G4–C5 (D11 = C3; D21 = C4)—which had its root
on top—was somewhat less stable. However, Hindemith attached a greater har-
monic value to this interval than to those producing non-amplifying difference
tones, like the minor sixth E4–C5 (D11 = G3). Hindemith’s arrangement of inter-
vals—his “series II”—was the following: (1) octave; (2) perfect fifth and (3) per-
fect fourth; (4) major third and (5) minor sixth; (6) minor third and (7) major
8. This is a common German notation, originally proposed by Heinrich Husmann.

D11 and D21 are called difference tones of the first and second order, respectively. When f1
and f2 are the frequencies of two pitches forming an interval, and f2 is higher than f1, then
the frequency of D11 equals f2 − f1, and the frequency of D21 equals 2f1 − f2 (Husmann
1953, 12–13). Husmann’s proposal to let the order indications “first,” “second” etc. begin
with the two primary pitches has not found general acceptance.

Example 5.3. Hindemith’s chord classification (Unterweisung im Tonsatz, 1937).
© By kind permission of the Schott Music GmbH & Co. KG, Mainz, Germany.
10/2/2008 10:31:16 PM
142 similarity
sixth; (8) major second and (9) minor seventh; (10) minor second and (11)
major seventh; (12) tritone.9
Like Krenek, Hindemith determined the value of a chord by considering the
values of the intervals of which it was formed. However, they valued certain inter-
vals differently, like the perfect fourth (a fairly stable interval according to Hin-
demith, but one that could also sound dissonant, according to Krenek) and the
tritone (a highly unstable interval according to Hindemith, but a “neutral” one
according to Krenek). Moreover, Hindemith also considered the mutual posi-
tions of the intervals in a chord. In his view, a chord was maximally stable when
its bass note was also the root, that is, the root of its “best” interval (in terms
of Series II). Hindemith distinguished six principal classes of chords, which he
identified by Roman numerals. A lower numeral indicated a greater harmonic
value (ex. 5.3; Hindemith 1937, 116–21).
The classifications of Hindemith and Krenek are, in fact, classifications of inter-
vallic properties, or combinations of such properties. Hindemith’s label “IIb2”
signifies the combined properties “containing a tritone, a major second and/or
minor seventh (i.e., not a minor second or a major seventh), and a bass note that is
not the root.” Having one such set of properties in common is an equivalence rela-
tion in the collection of chords encompassed by the twelve tones of the chromatic
octave. Hindemith’s classification of chords differs from the PC set classifications
of, say, Forte and Rahn in that there is no rule of transformation—or a coherent
set of such rules—connecting all members of one class, like transposition, trans-
position and/or inversion, etc. (The same goes for Krenek’s classification.) His
classes even consist of chords of different cardinalities. Moreover, since the root
concept serves as a criterion for the stability of a chord, two realizations of one PC
set may end up in different subclasses of his system. (Krenek—who did not use the
root concept—would not see them as distinct.) For example, Hindemith assigned
the chords C4–G4–B♭4 and B♭3–C4–G4 to the classes III1 and III2, respectively.
There is something else worth mentioning. Unlike the common PC set classi-
fications, Hindemith’s system does not seem to constitute a “phenomenology of
all chords” (Hindemith 1937, 120). It has obviously been devised for chords of
six tones or less, since it is unable to distinguish between larger ones. Each such
chord, of seven or more tones, contains representatives of all PICs. As a conse-
quence, all these chords would belong to just one of Hindemith’s classes, namely
IV. (Of course, Krenek’s classification of triads did not have this problem.) Rather
than simply viewing this imbalance as a shortcoming of Hindemith’s system, we
should consider its underlying pragmatic considerations: chords of more than
9. Series II is not entirely consistent with Hindemith’s basic principles. For example,
the members of a minor third are not amplified by either of this interval’s two prominent
difference tones. This means that the minor third would have to come after the major
second, the lower member of which is amplified by D11. In an apparent attempt to satisfy
the musical intuition of his readers, Hindemith presented the minor third as a “shading”
(Trübung) of the major third (Hindemith 1937, 89).

similarity 143
Example 5.4. Hindemith’s way of showing the increase and decrease of harmonic
tension (Harmonisches Gefälle) in his Unterweisung im Tonsatz.
© By kind permission of the Schott Music GmbH & Co. KG, Mainz, Germany.
six tones contradicted the principles of textural clarity and harmonic functional-
ity that guided the change in his style in the 1930s.
Which problems did Hindemith and Krenek actually want to solve? Their classi-
fications were obviously a response to the expansion of harmonic vocabulary in the
early twentieth century. In that respect, they fall in line with the chord-type inven-
tories discussed in chapter 4. However, Hindemith and Krenek did not content
themselves with the mere identification and tabulation of chords. They wanted to
provide new general criteria for their use, though for very different reasons. Hin-
demith aimed to keep harmonic theory as the intermediary between art and nature
at a time when it obviously had been shaken to its foundations.10 His “updating”
of it would not be complete without new instructions for the succession of chords.
Krenek—who, in the 1930s, acted as a spokesman for the musical modernism which
Hindemith opposed11—addressed his instructions for the succession of triads to stu-
dents of serial composition. Harmony was, in a way, a blind spot of early twelve-tone
serialism. Apart from the combinatoriality principle—which helped Schoenberg to
avoid octave doublings (see chapter 4, p. 94)—there were no general rules for the
treatment of chords. It was not Krenek’s aim to establish such a rule. Rather, he
wanted to bring the issue to the notice of his readers, for he felt that no composer
could manage without a keen sense of harmonic progression (Krenek 1940, 19).
In the opinion of these two men, one of the criteria for a good harmonic pro-
gression was the ebb and flow of harmonic tension—Hindemith’s harmonisches
Gefälle—which should be planned in order to make sense. Krenek showed how
the more dissonant simultaneities might prepare and underline the focal point
of a melody, and how the use of milder ones might induce a relaxation (Krenek
1940, 21–23). Similarly, Hindemith showed how a conscious choice of chords from
the various classes of his system might outline different tension curves (for example,
10. This was also the time that brought forth some electro-acoustical facilities with
which one could gain firm control of tone production. Hindemith’s decision to base
his new theory of harmony on the phenomenon of the combination tone—a phenom-
enon commonly known for almost two centuries—was inspired by his experiments with
the trautonium, an electronic musical instrument invented in the late 1920s (Hindemith
1937, 78). It is ironic that an electronic instrument helped Hindemith to take a conserva-
tive stance in view of musical modernism.
11. Especially in his six lectures on “new music” in 1936 (Krenek 1937).

144 similarity
curves rising and falling very rapidly, or curves rising and falling more slowly). Or he
did the reverse, drawing such tension curves from harmonic progressions (ex. 5.4;
Hindemith 1937, 138). Hindemith also invoked other criteria for harmonic pro-
gressions—criteria involving voice leading and root progression—but it is this one
that is of special interest in relation to PC set similarity. Harmonic tension, based
on intervallic structure, provided an index of relatedness which might be useful for
composers. In PC set theory, however, it was the intervallic structure itself that mat-
tered—not a slippery attribute like “tension.” Such an attribute implied causal rela-
tionships—tensions and resolutions—contradicting the sense of musical logic that
PC set theory had inherited from twelve-tone serialism.
Distributive Patterns
In “A Theory of Set Complexes” Allen Forte attempted to define meaningful
similarity relations given two PC sets with equal cardinal numbers. Since these
relations were based on the APIC vectors, they also involved the transpositions,
inversions, and Z-correspondents of these PC sets.
According to Forte, two PC sets were maximally similar if four interval
classes (i.e., APICs) were represented by the same number of unordered PC
pairs in each. In other words, four entries in the APIC vectors of these two
PC sets should be identical (Forte 1964, 150). (Four is the maximum number
of entries, for APIC 1–6, that two non-equal APIC vectors can have in com-
mon.) Eric Regener stated this relationship a little more formally. According
to Regener, two PC sets A and B, with a and b occurrences of APIC i respec-
tively, were maximally similar if:
ai = bi for four of the i [i ≠ 0]. (Regener 1974, 205)
The following example has been drawn from Forte’s analysis of Webern’s Five
Movements for String Quartet, Op. 5, no. 4, the analysis concluding his 1964
article. The violin tremolos in mm. 1–2 (ex. 5.5) realize the PC sets α = {B,C,E,F}
(that is, {11,0,4,5}; Forte name “4–8”) and β = {B,C,F,F♯} (that is, {11,0,5,6}; Forte
name “4–9”). A comparison of their APIC vectors (minus the entries for APIC 0)
reveals why Forte called them “maximally similar”:
α: 200121
β: 200022
Four entries correspond; two are different. (To facilitate comparison, different entries
between two APIC vectors appear in bold typeface in this and the following exam-
ples.) In PC set α there is one instance of APIC 4. This class is not represented in PC
set β. On the other hand, the single instance of APIC 6 in PC set α is outweighed by the

similarity 145
Sehr langsam = ca. 58
mit Dämpfer
am Steg . . . . . . . . . . . . . . . . . .
mit Dämpfer
am Steg . . . . . . . . . . . . . . . . . .
α β
Example 5.5. Maximal similarity between the PC sets {11,0,4,5} (α) and {11,0,5,6}
(β) at the beginning of Webern’s Opus 5, no. 4. An example mentioned by Allen
Forte in “A Theory of Set-Complexes for Music” (1964).
two instances of this interval class in PC set β. If we take these numbers into
account—and if we assume a similarity relation to involve two PC sets with non-
equal APIC vectors—the predicate “maximally similar” is fully appropriate.
Yet a further differentiation is possible. Some PC sets may be more “maximally
similar” than others. For example, Forte distinguished between “first-order” and
“second-order” maximal similarity, using common mathematical terminology to
denote levels of quantity.12 APIC vectors in a first-order maximal similarity rela-
tion consist of the same digits, whereas APIC vectors in a second-order maxi-
mal similarity relation—like those of the PC sets α and β above—do not. In
Regener’s formulation, a first-order maximal similarity relation meant that:
ai = bi for four of the i and in addition aj = bk (which implies ak = bj) for the
other two interval classes j and k (Regener 1974, 205)
In his analysis of Webern’s Opus 5, no. 4, Forte also mentioned an instance of

this relation, involving the PC sets {0,4,5} and {11,0,4} on the one hand (γ; Forte
name “3-4”), and {11,0,5} and {11,4,5} on the other (δ; Forte name “3-5”). These
PC sets have the following APIC vectors:
γ: 100110
δ: 100011
APIC 4 is represented in γ by the same number of unordered PC pairs as APIC 6

in δ. The same is true, of course, of APIC 6 in γ and of APIC 4 in δ.
12. Babbitt had used this expression with reference to combinatorial hexachords.
According to Babbitt, “first-order” hexachords created combinatorial relationships at only
one transpositional level, “second-order” hexachords at two, etc. (Babbitt 1955, 58).

146 similarity
Table 5.1. Distributive patterns in the relation 5:1 (Forte: second-order maxi-
mal similarity)
PC set α = {B,C,E,F}, or {11,0,4,5} (Forte name “4-8”)
{4,5} {0,5}
{11,0} {0,4} {11,4} {11,5}
APIC 1 2 3 4 5 6
PC set β = {B,C,F,F♯}, or {11,0,5,6}(Forte name “4-9”)
{5,6} {0,5} {0,6}

{11,0} {11,6} {11,5}
APIC 1 2 3 4 5 6
Note: The columns represent the successive APICs from 1 through 6. Five out of six PC pairs
from each set are distributed to the same classes. The black bars mark the common number of
pairs per class.
The maximal similarity relation appears to be even more flexible when the
sizes of the vector entries are taken into account. Each APIC vector represents a
distributive pattern: the unordered pairs formed by the elements of a PC set are
distributed to a maximum of six interval classes. When we compare the APIC vec-
tors of two PC sets of the same cardinality, we should know how many pairs from
either set belong to the same classes, and how many belong to different classes. The
ratio of these numbers can serve as an expression of the relation between these
PC sets. In the case of the PC sets α and β above, this ratio is 5:1, as table 5.1 shows.
Both sets have two PC pairs in APIC 1, two in APIC 5, and one in APIC 6. The
single remaining pairs in α and β are assigned to APIC 4 and 6, respectively.13
Now let us compare the APIC vectors of α and a new PC set: ε = {4,5,6,11}
(Forte name “4-6”):
α: 200121
ε: 210021
The APIC vectors of the PC sets α and ε represent distributive patterns in the same
relation as those of α and β, that is, 5:1 (table 5.2). Unlike β, however, ε is in a first-
order maximal similarity relation to α. Compared to the single APIC-4 represen-
tative {0,4} in PC set α, it accommodates an APIC-2 representative—{4,6}—which
does not have a “class-mate” either. In PC set β, however, the counterpart of {0,4} is
an APIC-6 representative, of which there was already one in PC set α.
13. The relation between the vectors of two PC sets α and β is symmetrical only if α
and β have the same cardinal number. In that case, there is one proportion representing
α’s relation to β as well as β’s relation to α.

similarity 147
Table 5.2. Distributive patterns in the relation 5:1 (Forte: first-order maximal
similarity)
PC set α = {11,0,4,5} (Forte name “4-8”)
{11,0} {0,5}
{4,5} {0,4} {11,4} {11,5}
APIC 1 2 3 4 5 6
PC set ε = {4,5,6,11} (Forte name “4-6”)
{5,6} {11,6}
{4,5} {4,6} {11,4} {11,5}
APIC 1 2 3 4 5 6
Table 5.3. Distributive patterns in the relation 4:2 (Forte: first-order maximal
similarity)
PC set ζ = {11,0,4,6} (Forte name “4-16”)
{11,4}
{11,0} {4,6} {0,4} {11,6} {0,6}
APIC 1 2 3 4 5 6
PC set η = {0,3,4,6} (Forte name “4-12”)
{0,3}
{3,4} {4,6} {3,6} {0,4} {0,6}
APIC 1 2 3 4 5 6
Two new PC sets—ζ = {11,0,4,6} (Forte name “4-16”) and η = {0,3,4,6} (Forte
name “4-12”)—are in a first-order maximal similarity relation as well:
ζ: 110121
η: 112101
However, from table 5.3 it appears that their distributive patterns are in the rela-
tion 4:2. Seen thus, they are more different than α and ε are, and even more
different than α and β, which are in a second-order maximal similarity relation
according to Forte. Apparently, the notion of “maximal similarity” may include
smaller or larger differences between the interval contents of two PC sets.
The same is true for another similarity measure that Forte had defined in 1964:
“minimal similarity.” In two PC sets that are minimally similar, each interval class

148 similarity
is represented by a different number of PC pairs (Forte 1964, 150). The APIC vec-
tors of the PC sets θ = {10,0,2,4,6} (Forte name “5-33”) and ι = {11,0,2,4,6} (Forte
name “5-24”) provide an instance of this relation:
θ: 040402
ι: 131221
Table 5.4 shows that these vectors represent distributive patterns in the relation
6:4. Now, θ is minimally similar to other sets as well, for example κ = {11,0,1,2,4}
(Forte name “5-2”):
θ: 040402
κ: 332110
In this case, however, the relation is 4:6 (table 5.5). Apparently, there are various
degrees of “minimal similarity.” This may seem contradictory, but it only shows
that PC set similarity can be defined in more than one way.
In The Structure of Atonal Music, Forte defined his similarity measures again,
now labeling them “R0” (minimal similarity), “R1” (first-order maximal similar-
ity), and “R2” (second-order maximal similarity).14 He also added a new mea-
sure: “maximal similarity with respect to pitch class,” designated “Rp.” Two PC
sets of cardinal number n are in this relation if they contain a common subset
of cardinal number n −1 under transposition and/or inversion (Forte 1973, 47,
50). Forte combined this measure of inclusion with each of the intervallic simi-
larity measures so as to reduce the number of PC sets satisfying either measure.
As said before, Forte regarded a relation as more significant when it involves a
smaller number of PC sets:
[T]he pc similarity relation Rp is not especially significant taken alone, since by

that measure a given set may be similar to many others. When Rp is combined
with the [interval-class] similarity relations, however, a considerable reduction
is effected. (Forte 1973, 49)
Forte thus spoke of musical relations as independent from musical context. He

measured their significance against the criteria of a self-contained system of
which the musical work was only an example. Forte’s approach to PC set simi-
larity was later criticized for failing to provide comprehensive coverage of that
system (Lord 1981; Isaacson 1990). Indeed, Forte considered a few special cases
14. Regener rightfully criticized these labels as being meaningless. The indices 0, 1,
and 2 follow the order: least similarity—great similarity—not as great similarity (Regener
1974, 205).

similarity 149
Table 5.4. Distributive patterns in the relation 6:4 (Forte: minimal similarity)
PC set θ = {10,0,2,4,6} (Forte name “5-33”)
{10,0} {6,10}
{4,6} {10,2}
{2,4} {2,6} {10,4}
{0,2} {0,4} {0,6}
APIC 1 2 3 4 5 6
PC set ι = {11,0,2,4,6} (Forte name “5-24”)
{4,6}
{2,4} {2,6} {11,6}
{11,0} {0,2} {11 {11,2},2} {0,4} {11,4} {0,6}
APIC 1 2 3 4 5 6
Table 5.5. Distributive patterns in the relation 4:6 (Forte: minimal similarity)
PC set θ = {10,0,2,4,6} (Forte name “5-33”)
{4,6} {2,6}
{10,0} {10,2}
{2,4} {6,10} {10,4}
{0,2} {0,4} {0,6}
APIC 1 2 3 4 5 6
PC set κ = {11,0,1,2,4} (Forte name “5-2”)
{1,2} {11,1}
{0,1} {2,4} {1,4}
{11,0} {0,2} {11,2} {0,4} {11,4}
APIC 1 2 3 4 5 6
of PC set similarity only, identifying them by distinct labels. He did not provide a
full scale of relative values. For this, he may have had his reasons: any two PC sets
are “similar” to some extent, but this does not mean that any similarity relation
is worth using. Forte’s measures are thus a selection from all the possible simi-
larity relations. On the basis of the vector entries, he could, for example, have
specified other degrees of similarity besides the maximal and minimal—APIC
vectors with one, two, or three corresponding entries—but he did not.
Which musical intuition underlies Forte’s intervallic similarity measures? It is
important to recall that they only apply to PC sets with the same number of ele-
ments. What does this restriction tell us? We already know that Forte, when he

150 similarity
presented the intervallic measures in 1964, referred to the possibility of compar-

ing other sets as well, saying that such an exploration would exceed the limits of
his paper. In The Structure of Atonal Music, however, he seems to have dropped
the matter. There is no mention of other similarities than those between PC sets
of equal size:
[I]t might be useful to define relations for sets of the same cardinal number so
that given two such sets known to be non-equivalent one could determine the
degree of similarity between them. (Forte 1973, 45)
Do Forte’s similarity relations themselves exclude comparisons between PC

sets of different sizes? Indeed, there can be no first-order maximal similar-
ity between sets of different cardinalities. Second-order maximal similarity is
possible if one set contains one more element than the other. The relation
between the PC sets ζ (see above) and λ = {0,1,4,6,9} (Forte name “5-32”) is a
case in point (table 5.6):
ζ: 110121
λ: 113221
Minimal intervallic similarity as defined by Forte becomes a more trivial

relation between PC sets as the difference between their cardinal numbers
increases. Therefore, it is best to use this relation with pairs of equal-sized
PC sets. But then there are conflicting reasons for defining this as the
domain of both minimal and maximal similarity. The significance of mini-
mal similarity is based on its relative infrequency among equal-sized PC
sets. Maximal similarity, however, can hardly be found in any other domain.
It would be logical if maximal similarity were defined for pairs of PC sets
with different numbers of elements, like ζ and λ above. For Forte, however,
comparing such pairs fell within the scope of PC set inclusion (see chapter
6), rather than similarity relations.
Forte’s intuition of intervallic similarity can be summarized in a number of
statements:
Statement 1:
PC pairs (intervals, subsets of cardinal number 2) are the key attributes of a
PC set.
Each pair of PCs receives a value (normally the APIC value). Similarity is
founded on the correspondence of these values. Statement 1 assumes interval
content as a distinctive property of a PC set.

similarity 151
Table 5.6. Distributive patterns of PC sets with different numbers of elements
PC set ζ = {11,0,4,6} (Forte name “4-16”)
{11,4}
{11,0} {4,6} {0,4} {11,6} {0,6}
APIC 1 2 3 4 5 6
PC set λ = {0,1,4,6,9} (Forte name “5-32”)
{1,4}
{6,9} {9,1} {1,6}
{0,1} {4,6} {9,0} {0,4} {4,9} {0,6}
APIC 1 2 3 4 5 6
Note: The PC sets satisfy Forte’s definition of second-order maximal similarity.
The next two statements are numbered 2a and 2b, since they both deal with
the same question: is a similarity relation exclusive to certain pairs of PC sets?
Statement 2a:
Similarity is a relation between equal-sized PC sets.
The APIC vectors of these PC sets constitute partitions of the same number (3 in
the case of PC sets with three elements, 6 in the case of PC sets with four elements,
10 in the case of PC sets with five elements, etc.). This is considered the basic
condition of intervallic similarity between two PC sets. It is actually an equivalence
relation, but PC sets in this relation may resemble one another to a greater or
lesser degree. The equal number of elements provides a common basis enabling
a relatively simple comparison between them. As an aside, it should be noted that
measuring the “size” of a PC set in distinct elements is not only a mathematical
convention. It also reflects a musical convention: the way in which tonal chords are
identified as triads or seventh chords without regard to any doublings.
Statement 2b:
Similar PC sets do not have the same interval content.
This statement is in keeping with the common use of the word “similar,” a
word that signifies a resemblance, but not an identity. When two APIC vectors
form partitions of the same number, they are supposed to be similar—unless
they are identical.
Statement 3:
The degree of similarity depends on the number of interval classes (APICs)
that are equally represented in both PC sets.

152 similarity
IVEC (I) = [3 2 1 0 0 0] (4-1)

IVEC (L) = [2 2 1 1 0 0] (4-2)
differences = [1 0 0 -1 0 0]
1 2 +0 2 +0 2 +1 2 +0 2 +0 2 2 = s.i.
Example 5.6. Richard Teitelbaum’s computation of the “similarity index” (s.i.) of

two PC sets with Forte names “4-1” (for example {0,1,2,3}) and “4-2” (for example
{0,1,2,4}), respectively. “IVEC” is Teitelbaum’s abbreviation of “interval vector”—that
is, our APIC vector.
Richard Teitelbaum, “Intervallic Relations in Atonal Music,” Journal of Music Theory 9/1 (1965):
88. Used by permission of Duke University Press and Yale University.
The fact that this number ranges between 0 (minimal similarity) and 4 (max-
imal similarity) reveals how rough Forte’s measures actually are. Five possible
degrees of similarity seem inadequate in view of the twelve set-classes of cardinal
number 3 that Forte distinguished, not to mention his thirty-eight set-classes of
cardinal 5, or his fifty set-classes of cardinal number 6.15 Forte’s measures can
tell us whether an interval class is equally represented in two PC sets. However,
they do not register the extent to which it is represented in either set, that is, its
share in the total number of PC pairs. As a consequence, we do not know in
what proportion similarity stands to difference. When the difference depends
on two interval classes only, as in the case of maximal similarity, it can still be
smaller or greater. Table 5.2 and table 5.3 illustrate this. In brief, any measure of
similarity should include a measure of difference.
Squaring the Difference

In 1965, shortly after Forte had presented his similarity measures, Richard Teitel-
baum came up with an alternative approach. Teitelbaum had just graduated
from Yale University, and would make a name for himself as a composer and
live-electronics musician in the following years. His article “Intervallic Relations
in Atonal Music” appeared in Journal of Music Theory, like Forte’s “A Theory of
Set-Complexes.” They seem to have been prepared simultaneously. Like Forte,
Teitelbaum claimed to fill the need of a systematic approach to the atonal works
of Schoenberg, Webern, and Berg; and he made no secret of his indebtedness
to Forte, much of whose terminology he had adopted. But he did not refer to
Forte’s similarity measures. He defined a new measure—the similarity index (s.i.
for short)—which responded to the problem that Forte had left unsolved: the
proportion of similarity to difference of two PC sets.
15. If the four statements above are taken for granted, the total number of similarity
relations is 66 for set-classes of cardinal number 3, and 595 for the set-classes of cardinal
numbers 5 and 6.

similarity 153
To compute the s.i. for two PC sets, Teitelbaum, like Forte, compared their
APIC vectors. Given two vectors, Teitelbaum proposed the following protocol:
[T]ake the difference between the corresponding entries, square each difference,
sum the squares and take the square root of the sum.(Teitelbaum 1965, 88)
He also provided an example, shown here as example 5.6. Teitelbaum’s proto-

col was actually based on a geometrical interpretation of the APIC vector. In
such an interpretation, the entries of a row vector represent the coordinates
marking a certain point in an x-dimensional space. The vector is an arrow con-
necting the origin of the coordinate system with this point (ex. 5.7a). Its value
is equal to its length, which can be computed by extracting the root of the
sum of the squared coordinates. (A simple case of this is Pythagoras’s theorem,
which applies to a two-dimensional space: a = a x 2 + a y 2, where a is the vector, and
where ax and ay are positions on the two axes of the coordinate system.) Seen
thus, the APIC vector is a vector in a six-dimensional space. It can be notated
(a1, a2, a3, a4, a5, a6), and its value is:
a1 + a 2 + a 3 + a 4 + a 5 + a6
2 2 2 2 2 2
Teitelbaum’s s.i. is the difference of two such vectors. To put it in a slightly dif-
ferent way, it is the value of the difference vector. The latter results from the sub-
traction of the entries of one vector from those of the other. Given two vectors
a = (a1, a2, a3, a4, a5, a6) and b = (b1, b2, b3, b4, b5, b6), the value of the difference
vector a − b is:
(a1 - b1 ) 2 + (a 2 - b2 ) 2 + (a 3 - b3 ) 2 + (a 4 - b4 ) 2 + (a 5 - b5 ) 2 + (a6 - b6 ) 2
Geometrically, the subtraction of one vector from another can be seen as a spe-
cific way of joining them together. Example 5.7b illustrates this, employing the
two-dimensional space of example 5.7a.
Teitelbaum’s s.i. offered more different values than the hidden scale against
which Forte had plotted his degrees of intervallic similarity. Also, it sometimes
contradicted Forte’s notion of PC set similarity. A comparison of the values that
Forte and Teitelbaum would assign to the same pairs of PC sets reveals their
varying degree of correspondence. Earlier on, the PC sets α = {11,0,4,5} and β =
{11,0,5,6} were shown to have APIC vectors representing distributive patterns in
the relation 5:1 (see table 5.1). Another pair of PC sets, α and ε = {4,5,6,11}
–
yielded the same result (see table 5.2). The s.i. would be 1.41 (√2)16 in both
16. The values of the s.i. are usually irrational. One can write them as the square root
of an integer, as Teitelbaum himself did (provoking the criticism of Charles Lord; see
Lord 1981, 111, note 7), or one can provide decimal numbers as approximations of these
values. Both notations are used in the present discussion.

154 similarity
ay
ax
Example 5.7a. The geometrical representation of a vector called a.
ay
-b
a
(a y - b y )
(a - b)
b
ax
(a x - b x )
Example 5.7b. The subtraction of a vector called b from vector a. An arrow as long
as b, but pointing in the reverse direction (−b), departs from the far end of a, mark-
ing a new point with the coordinates (ax− bx) and (ay − by). The difference of a and b
is a vector connecting the origin with this point.
cases, indicating a small difference of interval content. When seen from Forte’s
perspective, these pairs are maximally similar, but they fall into the “second-” and
“first-order” categories, respectively. The s.i. of the PC sets ζ = {11,0,4,6} and η =
{0,3,4,6}—with APIC vectors representing distributive patterns in the relation 4:2
–
(see table 5.3)—would be 2.82 (√8). In other words, they would be more different
than α and β or α and ε. According to Forte’s classification, however, ζ and η are in
a closer relation than α and β: a relation of first-order maximal similarity.
What is the cause of these discrepancies? We have seen that Forte considered
it important when two PC sets contained the same numbers of one interval class

similarity 155
(Statement 3). We also have seen that his measures of intervallic similarity are
insensitive to the actual sizes of these numbers. Now, the above observations sug-
gest that the s.i. of two PC sets increases with the difference between their dis-
tributive patterns. The explanation for this is simple. In two sets, the distribution
of PC pairs to the six interval classes is the same up to a specific number of these
pairs (cf. the black bars in Table 5.1 to 5.8). That leaves a number of pairs of
which the distribution is not the same. The s.i. registers this difference, the blind
spot of Forte’s similarity measures.
It does even more than that, as the following examples will show. First, let us
compare the APIC vectors of the PC sets µ = {2,3,4,5} (Forte name “4-1”) and
v = {2,3,4,8} (Forte name “4-5”):
μ: 321000
v: 210111
The distributive patterns of µ and v are in the relation 3:3 (table 5.7), which
suggests that these PC sets are less similar than, say, ζ and η. The distributive
patterns of the last mentioned pair were shown to be in the relation 4:2 (see table
–
5.3). However, the s.i. of µ and v—2.44 (√6)—is smaller than the s.i. of ζ and η—
–
2.82 (√8)—thus indicating a greater similarity of interval content. This discrepancy
can be explained as follows. The difference between ζ and η involves two interval
classes only (APIC 3 and APIC 5). In the case of µ and v, however, the difference is
spread over all six interval classes. By squaring the differences between the corre-
sponding vector entries, Teitelbaum’s s.i. suggests that two PC sets provide a larger
contrast as their difference is more concentrated in one interval class. When the
relation of the distributive patterns does not change, the s.i. may still indicate a
greater or lesser difference between PC sets. The comparison of µ’s vector with
that of a new PC set, ξ = {2,4,7,9} (Forte name “4-23”), illustrates this:
μ: 321000
ξ: 021030
These vectors represent distributive patterns that are in the same relation as
–
those of µ and v—that is, 3:3 (table 5.8). The s.i., however, is not 2.44 (√6), but
––
4.24 (√18). It strongly emphasizes the difference, as this involves the minimal
number of interval classes (two: APIC 1 and APIC 5). Interestingly, Forte would
assign a high degree of similarity to µ and ξ for the same reason. In his view, these
PC sets would be in a first-order maximal similarity relation.
Teitelbaum’s conception of intervallic similarity can be summarized as follows:
Statement 1:
PC pairs (intervals, subsets of cardinal number 2) are the key attributes of a
PC set.

156 similarity
Table 5.7. Distributive patterns in the relation 3:3
PC set μ = {2,3,4,5} (Forte name “4-1”)
{4,5}
{3,4} {3,5}
{2,3} {2,4} {2,5}
APIC 1 2 3 4 5 6
PC set v = {2,3,4,8} (Forte name “4-5”)
{3,4}
{2,3} {2,4} {4,8} {3,8} {2,8}
APIC 1 2 3 4 5 6
Note: The difference is spread over six interval classes.
Table 5.8. Distributive patterns in the relation 3:3
PC set μ = {2,3,4,5} (Forte name “4-1”)
{4,5}
{3,4} {3,5}
{2,3} {2,4} {2,5}
APIC 1 2 3 4 5 6
PC set ξ = {2,4,7,9} (Forte name “4-23”)
{2,9}
{7,9} {4,9}
{2,4} {4,7} {2,7}
APIC 1 2 3 4 5 6
Note: The difference only involves two interval classes.
Statement 1 was also made in connection with the similarity measures of

Forte.
Statement 2:
No pair of PC sets is excluded from the similarity relation.
The s.i. can be computed for each pair of PC sets, regardless of their car-
dinal numbers, and regardless of other relations they may have (such as an
identical interval vector). In other words, all PC sets are considered similar to
a greater or lesser degree. The fact that Teitelbaum did not apply his s.i. to PC

similarity 157
sets with different cardinal numbers does not change this condition. Nor is it
essential that the s.i. lose its effectiveness as this difference increases, as Eric
Isaacson (1990, 19) observed. This may only be a reason to limit the domain
of investigation.
Statement 3:
The similarity of two PC sets depends on the amounts by which the frequen-
cies of the interval classes in one set exceed those in the other.
The smaller these amounts, the greater the similarity. Teitelbaum’s s.i. is actually
a measure of dissimilarity.17
Statement 4:
The difference between two PC sets is greater as it involves a smaller number
of interval classes.
This becomes obvious when we consider two PC sets characterized by an uneven

distribution of intervals, one stressing an interval class that is entirely lacking from
the other. A PC set of the “diminished seventh-chord” type—stressing APIC 3 (four
PC pairs)—and a PC set of the “whole-tone tetrachord” type—stressing APICs 2
and 4 (three and two PC pairs, respectively) will together yield a high s.i.
Common Subsets
Most of the similarity measures discussed so far involve abstract intervals—that
is, intervals dissociated from their constituent elements. (The dissociation
was expressed in Statement 1 above.) Alternatively, two PC sets may be similar
because they share one or more subsets. We can say that set-classes of cardinali-
ties 3 and higher are more likely to be heard as common to a pair of PC sets
than interval classes. In other words, subset content may provide a more reliable
measure of similarity between PC sets than interval content does.
However, the definitions of PC set similarity that take subset content as their
starting point are no less abstract than those based on interval content. The mean-
ing of the term “subset” has been broadened by theorists: it may involve common
subsets under transposition and/or inversion, operations preserving the subset’s
interval content, though some theorists have excluded inversion from consider-
ation. Others, however, did not even specify operations, thus leaving open possibili-
ties such as a similarity based on common subsets under multiplication.
Forte had already brought up subset content as a measure of PC set similar-
ity in The Structure of Atonal Music: his Rp-relation was based on it. But the issue
received a more general treatment from Eric Regener (1974), who was a critical
17. This is true for other measures as well. See Rahn 1980b, 489, and Isaacson 1990,
27, note 7.

158 similarity
reviewer of Forte’s book. The general tenor of Regener’s comments—which were

published in Perspectives of New Music—was Forte’s lack of mathematical rigor, but
he also cast doubt upon the musical value of some of Forte’s relational concepts.
Regener proposed the partition function as a tool that would measure the similarity
of two PC sets more accurately than Forte’s Rp, R0, R1, and R2.
The partition function is the number of “configurations” (that is, transpo-
sitions) of one PC set having n elements in common with another. The parti-
tion functions for the successive ns—from 0 up to the cardinal number of the
smallest set—form the partition vector,18 which can detect structural similarities
between the two sets.
To illustrate how Regener’s concept works, let us return to the PC sets ζ =
{11,0,4,6} and λ = {0,1,4,6,9}. From table 5.6 we can see that the interval content
of the larger set λ includes that of ζ. Does this result bear a relation to the subset
content of ζ and λ? Table 5.9a shows the number of elements (“notes”) each
transposition of ζ has in common with λ. When ordered by interval of transposi-
tion, these numbers form the common-note vector of ζ and λ.19 Table 5.9b shows
how the partition vector can be derived from the common-note vector by count-
ing the numbers of the equal entries in the latter. There are no 0s and 4s in
the common-note vector. However, it contains four 1s, seven 2s and one 3. This
results in the partition vector 13710, ||13710|| in Regener’s notation.
One thing that can be read from the partition vector is the degree of inclusion.
This is the cardinal number of the largest common subset(s) under transposi-
tion. It is represented by the right-most non-zero entry of the partition vector. In
the case of ζ and λ, the degree of inclusion is 3, and is represented by the entry
1 (which means: “ζ has three elements in common with λ at only one transposi-
tional level”). This degree of inclusion tells us that no transposition of ζ—a set of
four discrete elements—is entirely included by λ, despite the total inclusion of ζ’s
interval content. We may conclude that the number of instances of an interval
class held in common between two PC sets is not indicative of the degree of sub-
set inclusion. However, Regener’s measure fails to provide sufficient evidence of
this, since it does not take into account every subset with the same interval con-
tent as ζ. The partition function should be extended to include the inversion of ζ
and its transpositions.20 Table 5.10 does this, and shows that the degree of inclu-
sion of I(ζ) and λ is the same, namely 3. Therefore, no straightforward relation
exists between common interval content and common subset content.
The degree of inclusion should, of course, be considered relative to the car-
dinal numbers of these sets. But even then it lacks discriminating power, as the
following examples illustrate. It would seem that the number 3 represents a
18. This row vector is a partition of the number of transpositions, i.e., 12.
19. This row vector is a partition of the product of the cardinal numbers (20 in this
case; each element of a set of cardinal number 5, like λ, must appear four times in a trans-
position of a set of cardinal number 4, like ζ (see table 5.9a).
20. David Lewin (1977b) proposed such an extension. See further below.

Table 5.9. Partition function of the PC sets ζ = {11,0,4,6} and λ = {0,1,4,6,9}
(a) Tn(ζ): Number of elements

in common with λ:
0 = { 11, 0, 4, 6 } 3
1 = { 0, 1, 5, 7 } 2
2 = { 1, 2, 6, 8 } 2
3 = { 2, 3, 7, 9 } 1
4 = { 3, 4, 8, 10 } 1
5 = { 4, 5, 9, 11 } 2
6 = { 5, 6, 10, 0 } 2
7 = { 6, 7, 11, 1 } 2
8 = { 7, 8, 0, 2 } 1
9 = { 8, 9, 1, 3 } 2
10 = { 9, 10, 2, 4 } 2 “Common-note vector”:
11 = { 10, 11, 3, 5 } 0 322112221220
(b) Number of elements in common with λ: 0 1 2 3 4

Instances of Tn(ζ) (“partition vector”) 1 3 7 1 0
Table 5.10. Partition function of the PC sets I(ζ) = {1,0,8,6} and λ = {0,1,4,6,9}
(a) Tn I(ζ): Number of elements

in common with λ:
0 = { 1, 0, 8, 6 } 3
1 = { 2, 1, 9, 7 } 2
2 = { 3, 2, 10, 8 } 0
3 = { 4, 3, 11, 9 } 2
4 = { 5, 4, 0, 10 } 2
5 = { 6, 5, 1, 11 } 2
6 = { 7, 6, 2, 0 } 2
7 = { 8, 7, 3, 1 } 1
8 = { 9, 8, 4, 2 } 2
9 = { 10, 9, 5, 3 } 1
10 = { 11, 10, 6, 4 } 2 “Common-note vector”:
11 = { 0, 11, 7, 5 } 1 320222212121
(b) Number of elements in common with λ: 0 1 2 3 4

Instances of Tn I(ζ) (“partition vector”) 1 3 7 1 0

160 similarity
significant degree of inclusion for two PC sets containing four elements each.
Yet, the PC sets µ = {2,3,4,5} and ξ = {2,4,7,9} are both included by this degree in
no less than eighteen T-classes of cardinal number 4. The same degree results
from their comparison to fifty-seven T-classes of cardinal number 6. And they
are included by a degree of 4—a degree as large as their cardinal number—in
twenty-one of these classes.
Now, let us compare PC sets of cardinal numbers 7 and 8. Three T-classes
of cardinal number 8 include the set {0,1,2,4,7,8,9} (Forte name “7–20”) with
a degree of 5; but 36 include it with a degree of 6, and four include it with a
degree of 7. Deriving a degree of inclusion from the partition vector thus leads
to a very uneven classification of similarity relations between PC sets.
The partition vector could also be used for an alternative measurement of
PC set similarity. Regener proposed comparing the partition functions of two
pairs of PC sets (Regener 1974, 208). What kind of result did he expect? Hav-
ing determined the partition vectors of these pairs, he could perhaps tell how
similar their relations were. This is not the same as telling how similar just two
PC sets are. However, both pairs may consist of two identical sets. For example,
one pair may consist of two µs, and the other of two ξs. In that case, µ and ξ are
compared with respect to their “self-partition functions.” We count the number
of elements each PC set has in common with each transposition of itself. This is
practically equivalent to applying Forte’s intervallic similarity measures. Regener
rightly noted that two PC sets in Forte’s R1-relation often have the same self-par-
tition functions (Regener 1974, 207). More specifically, they have the same self-
partition function unless they have different vector entries for APIC 6.21
Since self-partition functions are very similar to intervallic similarity as con-
ceived by Allen Forte, they involve the same problems. When it appears from the
self-partition vector that one PC set has, say, three elements in common with two
of its transpositions, and the same is true for the other set, the transpositional
levels concerned may or may not be the same for both sets. If they are not the
same, this is an indication of different interval contents, which the self-parti-
tion vectors fail to recognize as much as Forte’s R1-relation. For example, the
21. The APIC vector of a PC set reveals the numbers of PCs shared by the set and its
transpositions. Given a PC set A, a vector entry an (“a instances of APIC n”) most often
tells us that the number of elements in common between A and Tn(A) is a. If so, this is
also the number of elements in common between A and T−n mod 12 (A). Only when n = 6
will A and Tn(A) share 2a elements. In a previous section (on Forte’s similarity measures)
we have seen that the vectors of R1 -related sets consist of the same digits, and that four of
their entries correspond. This means that for most T(A)s sharing a elements with A, there
is a T(B) sharing a elements with the R1 -related set B. Consequently, A and B are likely to
have the same self-partition vector. This is not the case, however, when the entry for APIC
6 in the vector of A corresponds to the entry of another APIC in the vector of B—that is,
when a6 in the vector of A, and an in the vector of B (n ≠ 6). Then, B shares a elements
with Tn(B), whereas A shares 2a elements with T6 (A).

similarity 161
self-partition vectors of the PC sets µ and ξ (see table 5.8) are both ||52221||.
However, the transpositional levels at which µ retains three of its elements are 1
and 11, whereas ξ retains three elements when it is transposed by 5 or 7. This is
caused by the difference—so evident from table 5.8—between the interval con-
tents of µ and ξ. Compared to the three pairs of PCs in µ that belong to APIC 1,
ξ contains three pairs belonging to APIC 5.
Regener also considered the partition functions of two heterogeneous pairs
of PC sets. As said before, he thus applied the concept of similarity to relations
between PC sets. He did this because he had discovered that “[the] number of
different partition functions [was] relatively restricted.” Speaking of hexachords,
he pointed out that:
[T]o the 3240 (that is, 80·81/2) different pairs of hexachords correspond only
70 different partition functions, of which 14 cover the combinations of the 80
hexachords with themselves.
(Regener considered the collection of hexachords as partitioned by Tn alone.

This explains why he identified a larger number of different hexachords than
Forte—that is, 80 instead of 50.)
The number of pairs sharing the same function . . . ranges from 1 to 490
(for the class ||0044400||); fifty-one of these classes comprise 20 pairs or less.
(Regener 1974, 208)
These numbers, though interesting in themselves, are not particularly informa-

tive with respect to PC set similarity. What does it mean when two pairs of PC
sets have the same partition function? It means that the PC sets in question are
of the same sizes, and that the pairs have the same number of common subsets
under transposition. Moreover, the sizes of these common subsets correspond.
Thus, it can be said that the two pairs are similar to the same degree; but that
degree can still be high or low.
Although Regener’s alternatives to Forte’s similarity measures were by no
means a great success, David Lewin (1977b) took up this idea of comparing PC
set relations in terms of partition functions. Unlike Regener, Lewin considered
common elements under any group of canonical transformations (see chapter
4). The partition function of two PC sets A and B might not only involve the
twelve transpositions of A, but also the twelve transpositions of A’s inversion.
When the partition functions were the same, Lewin argued, two pairs of PC sets
could be similar to various degrees. They were “weakly similar” if the similar-
ity ended there. They were “superstrongly similar” if they also yielded the same
common-note vector. And they were only “strongly similar” if one pair yielded
the same common-note vector as did a canonical transformation of the other.
In Lewin’s view, then, the common-note function (the number of elements each
canonical transformation of one PC set has in common with the other) was

162 similarity
more significant than the partition function (the number of canonical transfor-
mations of one PC set sharing a specific number of elements with the other).
Lewin was right that the common-note function is a more selective measure. This
can be ascertained from the reflexive forms of the partition and common-note func-
tions. We have seen that the identical self-partition functions of the PC sets µ and ξ
conceal the difference between their interval contents. The common-note vectors
of µ and ξ and their duplicates, however, are most revealing in this regard:22
(µ,µ): 4 3 210 0 0 0 012 3
(ξ, ξ): 4 0 210 3 0 3 012 0
The PC sets µ and ξ are therefore only “weakly similar.” The conditions under
which any two PC sets A and B are strongly similar depend on the canonical
transformation group. Given the group of twelve transpositions, (A,A) and (B,B)
are superstrongly similar if A and B have the same interval content. Given the
group of transpositions and inversions, however, that condition would not even
ensure a strong similarity between the two. Lewin discovered that some of the As
and Bs with equal interval contents “fell by the wayside” when compared with
respect to their common-note functions. In other words, they were “weakly simi-
lar.” He found these As and Bs among the “Z-related” pairs (Lewin 1977b, 227).
Which statements could embody the ideas underlying the measures of
Regener and Lewin? First of all, what was their conception of a PC set?
Statement 1:
A PC set consists of discrete elements.
In other words, the PCs in a PC set are related only insofar as they belong to
the same set. Forte and Teitelbaum, when measuring the similarity of PC sets,
considered the intervals formed by the PCs. In other words, they saw each ele-
ment as a member of several pairs.
Statement 2 (“No pair of PC sets is excluded from the similarity relation.”) is not
different from the corresponding statement applying to Teitelbaum’s s.i. In the next
statement, however, we see how Regener and Lewin conceived of PC set similarity:
Statement 3:
Two PC sets are similar when they are partly the same.
Basically, Regener and Lewin measured the similarity of PC sets by enumerating

the elements they shared. It follows from Statements 2 and 3 that the “part that
is the same” may be one of the PC sets itself, or both of them (in which case the
PC sets are identical).
22. Apart from its first entry, the common-note vector of a PC set and its duplicate is
identical to the PIC vector of that PC set.

similarity 163
The next and final statement is “a step up the ladder of abstraction” (to quote
John Rahn):
Statement 4:
What is considered “the same thing” (PC set) may involve a transformation.
The elements may be replaced by others, as long as this is the consequence of

a canonical transformation. Regener was much more restrictive than Lewin in
regarding transformations as “canonical.” In the context presented here he
worked with transpositions only.
“Absolute” Measures
The “Set Theory” session of the second National Conference of the Society for Music
Theory, which took place in New York in 1979, contained two papers that addressed
the similarity issue. One was read by Robert Morris, the other was a response to it by
John Rahn. Both papers were published, along with the other papers of that session,
and with an additional contribution by David Lewin, in the journal Perspectives of New
Music. In its published form, this set of papers received considerable attention. It set
the agenda for future work in the field of PC set theory, as one can see from later pub-
lications like Morris (1987), Isaacson (1990 and 1996), and Rogers (1999). In these
years, PC set similarity was a very promising area of research.
Morris started off by proposing a new similarity measure: the similarity index
(SIM). At first sight, it looked like a rudimentary version of Teitelbaum’s mea-
sure of the same name (which Morris did not mention). Like Teitelbaum, Morris
compared the APIC vectors of two sets (“V,” in his notation), taking the differ-
ences of the corresponding entries. However, instead of squaring these differ-
ences, and taking the root of the sum of the squares, Morris simply added up the
differences. Here is his definition:23
If S and R are two sets:

V(S) = [a1,a2,a3,a4,a5,a6]
V(R) = [b1,b2,b3,b4,b5,b6]
. . . the similarity relation is written . . .
SIM(S,R)
. . . and is called the similarity index between (or of) R and S.
It is evaluated . . .
6
SIM(S,R)= Σ |an-bn|
n=1
(Morris 1980, 446)
23. In this definition, the numbers of instances per interval class (n, if not specified)
should actually have been given in subscript.

164 similarity
What is the consequence of a procedure like this? R and S both have n elements.
These elements, in turn, form ½(n2 − n) pairs in each set,24 which are distrib-
uted over six interval classes. As we have already observed, there is a number
of pairs in S and in R for which the distribution corresponds. Morris called this
number “k” (Morris 1980, 448). A number of pairs may then remain. Since R
and S are of the same size, this number is equally divided between them. The
pairs that remain in S are assigned to classes other than the remaining pairs in
R. Like Teitelbaum’s s.i., Morris’s similarity index (SIM) deals with these remain-
ing pairs. Both indices are actually measures of the difference between two PC
sets. However, Morris’s SIM does not measure the “spread” of the difference,
that is, the number of interval classes involved. Therefore, it is a less sensitive
tool than s.i. For example, Morris would not see a distinction between the rela-
tion of µ and v (s.i. = 2.44) and that of µ and ξ (s.i. = 4.24; see table 5.7 and 5.8).
His SIM would point at 6 in both cases.
As PC set similarity rose higher up the agenda of music theory at the begin-
ning of the 1980s, a special problem related to it became more obvious. Post-For-
tean similarity measures were supposed to return values for relations between
any two PC sets; as noted at the beginning of this chapter, this was one of their
main attractions. But these values could not always be compared, as they were
relative to the cardinal numbers of the PC sets involved. Morris was aware of this
problem, which also affected his own SIM:
Obviously, a similarity index of 2 between two hexachords is not directly com-

parable to the same index between two trichords. As the total number of [inter-
val classes] available becomes larger, a given index becomes a correspondingly
finer indication of similarity. (Morris 1980, 450)
Therefore, he provided an additional measure: an “absolute similarity index”

(ASIM). Morris’s idea was to divide the similarity index of two PC sets R and S by
the number of PC pairs in R plus the number of pairs in S.25 In his notation:
ASIM(R,S) = (SIM(R,S)) / (#V(R) + #V(S))
When the similarity index is equal to the total number of PC pairs in R and
S—indicating the maximum difference between any two PC sets (no PC pair
in S is distributed to the same interval class as any pair in R)—Morris’s ASIM
returns the value 1. As a consequence of Morris’s definition of PC set similarity,
24. The unisons are excluded; see chapter 2, p. 48.

25. Morris’s SIM is the sum of (a) the number of PC pairs by which S differs from
R, and (b) the number of PC pairs by which R differs from S. (These numbers are equal
when R and S have the same size.) It is thus a “two-way” relational concept, as is Teitel-
baum’s s.i. For this reason, the similarity index of R and S must be divided by the total
number of pairs in both sets.

similarity 165
only PC sets below cardinal number 5 can be maximally different.26 At the other
extreme, when two PC sets have the same APIC vectors, his ASIM returns the
value 0.
What does this mean for the PC set relations discussed so far? The PC sets
γ = {0,4,5) and δ = {11,0,5} yield the same similarity index as α = {11,0,4,5} and
ε = {4,5,6,11}—that is, 2. We have seen that these PC sets differ by one pair only.
Their distributive patterns are in the relations 2:1 and 5:1, respectively. Now,
Morris’s ASIM suggests that this difference loses significance when the total
number of PC pairs is larger. In the case of γ and δ (two times three PC pairs),
its value is 0.33. Therefore, γ and δ are less similar than α and ε (two times six PC
pairs), the absolute similarity index of which is 0.16.
This line of reasoning is clear. However, it proceeds from the APIC vector as an
abstract statistical format, that is, as a display of the frequency distribution of inter-
val classes in a PC set. Morris’s interpretation of it does not seem to consider the
nature and behavior of these “data.” Teitelbaum’s use of the APIC vector showed a
stronger sense of history in this regard. As we have seen, his s.i. responds strongly
and specifically to an unequal distribution of intervals, which reflects a musical
intuition buttressed by compositional practices (e.g., the use of equal divisions of
the octave). Morris’s contribution to PC set similarity expanded the area of the
thinkable. Others followed his lead, allowing the APIC vector to influence their
musical imagination and theoretical writings more and more.
John Rahn’s paper “Relating Sets” was more than just a response to Morris.
Apart from discussing and refining the tools the latter had offered—for exam-
ple, Rahn preferred Morris’s k to his SIM, and he proposed an “absolute k” as
the counterpart of Morris’s ASIM27—it also evaluated the similarity measures
provided by Regener (1974) and Lewin (1977b). And it contained a new set
of proposals, starting with a concept measuring the “mutual embedding in sets
A and B of sets X of size n,” designated “MEMBn(X,A,B)” (Rahn 1980b, 492).
Rahn’s concept was based on Lewin’s embedding number (EMB): “the number of
distinct forms of [a PC set] X . . . which are subcollections of [another PC set]
Y” (Lewin 1977b, 197; the word “form” refers to a canonical transformation).
Stressing subset content as a basis of PC set similarity, MEMBn was also reminis-
cent of Regener’s partition function.
It is one thing to invent a concept like the partition function or MEMBn; to
specify a value for the similarity of two PC sets is quite another. We have seen that
26. For two PC sets to be maximally different, the interval classes represented in one
should not be represented in the other, and is thus dependent on a sufficient number
of zero entries in both APIC vectors. The smaller the set, the larger this number. In this
regard, Morris’s ASIM reflects the intuition that chords consisting of just a few tones pro-
vide larger contrasts than chords containing many different tones.
27. To obtain this measure, 2k was divided by the number of PC pairs in R plus the
number of pairs in S. In Rahn’s notation:
ak(A,B) = 2·k(A,B)/#V(A) + #V(B) (Rahn 1980b, 489)

166 similarity
Regener had difficulties in deriving a scale of values from the partition function.
Rahn decided to count the number of all mutually embedded (common) sub-
sets of more than one element. Thus, he obtained a value for the “total mutual
embedding in A and B of subsets of all sizes greater than one” (Rahn 1980b,
493). The following formula tells us that TMEMB results from a summation of
the values of MEMBn for all n between 2 and 12 inclusive:
12
TMEMB(A,B) = Σ MEMBn(X,A,B) (ibid., 492)
n=1
It should be noted that Rahn did not just count the common subsets under trans-
position, like Regener. MEMBn was generalized to any group of canonical transfor-
mations. And Rahn focused on the number of common subsets, whereas Regener
had been more interested in their sizes. For example, Regener would have found
that the PC set ζ = {11,0,4,6} had a subset of three elements in common with λ =
{0,1,4,6,9}. According to Rahn, they would share four distinct subsets: {0,4,6}, {0,4},
{0,6}, and {4,6}. In his view, each common subset consisting of three elements
would entail three common subsets of two elements. Similarly, each common sub-
set of four elements would entail six common subsets of two, and four common
subsets of three elements, etc. Thus, the value of TMEMB grew exponentially with
the number of elements in common between two PC sets.
This was an acceptable idea in itself—much more acceptable than a concept
of PC set similarity fluctuating in exact proportion to the number of common
elements. What was problematical, however, was Rahn’s counting every instance
of a common subset:
X must be embedded at least once in both sets A and B to be counted; then all
instances of X in either set are counted. (Rahn 1980b, 492)
As we have seen, the PC set ζ = {11,0,4,6} has {0,4,6} in common with λ = {0,1,4,6,9}.
This common subset would thus count for two on Rahn’s scale. Of course, the
same goes for the smaller subset {0,6}. The common subset {0,4}, however, also
matches another subset of λ, namely {1,9}. Therefore, it would count for three.
The general idea underlying Rahn’s way of measuring PC set similarity can be
rendered as follows: imagine a thing (A) consisting of several components; next,
imagine a second, similar, thing (B). Suppose there is one component (X) of A
that is also a component of B; this is a similarity between A and B. Under Rahn’s
protocol, A and B are considered more similar as the number of X in B is greater.
(Of course, the similarity increases further when more X is also added to A.) For
the moment, it may be helpful to interpret A and B as two different dishes both
of which contain, say, ginger. What would happen if more ginger were added
to B? Although we can expect the flavor of ginger to become stronger in the
latter, the similarity between A and B need not increase as a result. Dish A may
not even taste of ginger much at all. This example may seem out of place, but it

similarity 167
is not clear why a similarity concept like Rahn’s would match the world of musi-
cal tones any better than the world of dishes and their ingredients. In any case,
when a common subset was represented by different numbers in two PC sets,
Rahn did not see this as a dissimilarity.
Like Morris, Rahn devised an “absolute” similarity measure for PC sets, using
TMEMB as a basis. To normalize the values of TMEMB between 0 and 1—repre-
senting minimal and maximal similarity, respectively—he divided TMEMB(A,B)
by the total number of relevant subsets in A and B. The result was “ATMEMB.”
PC sets that were related by a canonical transformation always yielded the value
1, which meant that each subset of A matched at least one in B. PC sets that were
not so related, but still had the same interval content, did not come out as maxi-
mally similar (as they do under Morris’s SIM).
The statistical approach to the similarity issue made itself felt in Rahn’s paper
as much as in Morris’s. Both authors defined PC sets in terms of loose com-
ponents: PC pairs (Morris) and subsets of any size between 2 and 12 inclusive
(Rahn). A component of one PC set might or might not match with one of
another. The number of matching components—or, in Morris’s case, the num-
ber of non-matching components—was set against the sum of all components in
both PC sets. Thus, similarities between PC sets were represented in proportion
to their cardinal number.
David Lewin’s “A Response to a Response” was the shortest, but by far the
most complex of the three papers on PC set similarity that were sent in for
publication in the 1980 Spring-Summer issue of Perspectives of New Music. It pro-
vided a measure, “REL,” that was meant as an alternative to Rahn’s ATMEMB.
REL now exists in various formulations. Lewin’s own was not incorrect, as he
himself suggested to Eric Isaacson (who reported this without further com-
ment; Isaacson 1990, 27, note 9), but the readers of his paper may have found
it a hard nut to crack:
REL(A,B) = (1/SQR(TOTAL(A)TOTAL(B))) times

ΣSQR(EMB(/X/,A)EMB(/X/,B)) (Lewin 1980, 500)
In his book Composition with Pitch-Classes, Robert Morris provided a more com-
pact version of the same definition, replacing some of Lewin’s labels (“SQR” for
“square root,” for example) by mathematical symbols:
1
REL(A,B) = ·V
W
V=
W=
and the sums are taken for all SCs [set-classes] X.

168 similarity
In his definition of W, however, Morris had wrongly substituted the plus sign for
the multiplication sign, so that in his version the square root was taken of the sum
of the numbers of X in A and B instead of their product (cf. Morris 1987, 107).
Although REL was a measure for the total common subset content of two PC
sets, Isaacson (1990) and Rogers (1999) applied it to subsets of a single cardinal
number: 2. Hence the term “REL2” in the following formula, which Lewin handed
down to them as an alternative for the one he had presented in 1980. This new
formula involved the cardinal numbers (#) of the two PC sets (X and Y):
REL2(X,Y) =
(Isaacson 1990, 11; Rogers 1999, 79)
All these formulas amounted to the same mathematical notion. When a is the
sum of n integers ai (ai = a1, a2, a3 . . . an), and b is the sum of n integers bi (bi =
b1, b2, b3 . . . bn)—in brief, when a = Σai and b = Σbi—the following equation
holds:
bi
ai
if and only if is a constant
The final clause stipulates a very important condition: for the equation to be valid,
ai and bi should be directly proportional. In Lewin’s interpretation, ai and bi rep-
resented the numbers of instances of a set-class i in the PC sets A and B. A valid
equation meant that A and B were maximally similar. Maximal similarity was thus
dependent upon a constant ratio between the numbers by which equivalent subsets
were contained in A and B. REL converted the above equation to a rational form:
or, alternatively,
(cf. Lewin 1980 and Morris 1987)
so that a value between 0 (minimal similarity) and 1 (maximal similarity) could

be obtained for any pair of PC sets. In the formula of REL2, the numerator was
multiplied by two so as to bring it into correspondence with the new denomina-
tor. The values of the terms #A(#A −1) and #B(#B −1) in this denominator were
twice the values of a and b, for which they had been substituted.28
28. REL2 can be thought of as measuring the similarity of interval-class vectors. In the
above definitions, a is the sum of the APIC vector entries of a PC set A; #A(#A −1) is the
sum of the PIC vector entries of that set (see chapter 2). It follows that #A(#A − 1) = 2a.

similarity 169
or, alternatively,
(cf. Isaacson 1990 and Rogers 1999)
Although Lewin’s measure was no doubt mechanically complex, the underly-

ing idea was straightforward; much more so, in any case, than the idea basic to
Rahn’s ATMEMB. Lewin’s measure also provided greater discrimination. Rahn
added the numbers of subsets of one equivalence class in A and B together. As
a consequence, he made no difference between a set-class represented by one
against four instances in A and B and a set-class represented by two against three.
The set-class in question counted for 5 in both cases. By taking the products of
these numbers, Lewin arrived at two different values, i.e., 4 and 6. As it was more
sensitive to differences such as these, REL could be seen as a more reliable mea-
sure of the delicacy of human perception than ATMEMB.
Lewin’s 1980 definition of REL, the one he would later reject, did not allow
PC sets to be maximally similar unless they were related by a canonical trans-
formation (transposition and/or inversion). In that case, the constant ratio
between the numbers of subsets X in A and B was 1:1. It would, however, have
been interesting to find another constant ratio as well. But such a thing could
only obtain for all X of a single cardinal number. This has possibly been the
reason for Lewin’s proposal of a specialized form of REL, namely REL2. David
Rogers, who provided a geometrical interpretation of the new measure (along
lines similar to our interpretation of Teitelbaum’s s.i.; cf. example 5.7), specified
six pairs of different-sized PC sets for which it returned the maximum value of
1 (Rogers 1999, 79). In both PC sets of each pair the same set-classes of cardi-
nal number 2—in other words, the same interval classes—were represented in
exactly the same proportion. This can be seen from their APIC vectors:
{0,1,4,5,8} (Forte name “5-21”): 202420
{0,1,4,5,8,9} (Forte name “6-20”): 303630
In this example the constant ratio is 2:3. In every PC set pair singled out as “max-
imally similar” by Rogers, there is a constant ratio for the classes of subsets of
each cardinal number. These ratios are all different, as table 5.11 shows. The
similarity of the PC sets presented in this table can easily be verified. One mem-
ber of each pair includes the other, which is not smaller in size by more than
one element. The larger member is transpositionally and inversionally symmet-
rical: it “closes the circuit” that is left open by the smaller one (ex. 5.8). The PC
sets {0,1,6}, {0,3,6}, and {0,2,4,6,8} relate to their counterparts—{0,1,6,7}, {0,3,6,9},
and {0,2,4,6,8,10}—as nearly complete cycles (modulo 12). All PC sets involve a
relatively small number of interval classes. The larger member of a pair does not
add new interval classes to those contained in the smaller.

Table 5.11. The constant ratios between the subset classes of all cardinal
numbers (#) in Roger’s six maximally similar pairs
3–5 4–9 5–21 6–20

#1 3 4 #1 5 6
#2 1 2 #2 2 3
#3 1 4 #3 1 2
#4 1 3
#5 1 6
{0,1,6} (Forte name “3-5”) {0,1,4,5,8} (Forte name “5-21”)
{0,1,6,7} (Forte name “4-9”) {0,1,4,5,8,9} (Forte name “6-20”)
3–8 4–25 5–33 6–35

#1 3 4 #1 5 6
#2 1 2 #2 2 3
#3 1 4 #3 1 2
#4 1 3
#5 1 6
{0,2,6} (Forte name “3-8”) {0,2,4,6,8} (Forte name “5-33”)
{0,2,6,8} (Forte name “4-25”) {0,2,4,6,8,10} (Forte name “6-35”)
3–5 4–9 7–31 8–28

#1 3 4 #1 7 8
#2 1 2 #2 3 4
#3 1 4 #3 5 8
#4 1 2
#5 3 8
#6 1 4
#7 1 8
{0,3,6} (Forte name “3-10”) {0.1.3,4,6,7,9} (Forte name “7-31”)
{0,3,6,9} (Forte name “4-28”) {0.1.3,4,6,7,9,10} (Forte name “8-28”)

1 5 6 2 4 6 3 3 6
1 5 1 5 2 4 2 4 3 3 3 3
{ 0, 1, 6 } , { 0, 2, 6 } , { 0, 3, 6 } ,
{ 0, 1, 6, 7 } { 0, 2, 6, 8 } { 0, 3, 6, 9 }
1 3 1 3 4 2 2 2 2 4 1 2 1 2 1 2 3
1 3 1 3 1 3 2 2 2 2 2 2 1 2 1 2 1 2 1 2
{ 0, 1, 4, 5, 8 } , { 0, 2, 4, 6, 8 } , { 0, 1, 3, 4, 6, 7, 9 } ,
{ 0, 1, 4, 5, 8, 9 } { 0, 2, 4, 6, 8, 10 } { 0, 1, 3, 4, 6, 7, 9, 10 }
Example 5.8. Maximal similarity in terms of REL2. Six pairs of PC sets singled out by David Rogers (1999).
10/2/2008 10:31:24 PM
172 similarity
When measuring the similarity of two PC sets A and B by means of REL, Lewin
did not count any set-class that was absent from either A or B.29 In that case, the
product of the share this set-class had in A (that is, ai) and its share in B (that is,
bi) is 0. Consequently, when the subset contents of A and B did not intersect at
all—as in the case of a “diminished-seventh chord” ({0,3,6,9}; Forte name “4-28”)
and a “whole-tone trichord” ({0,2,4}; Forte name “3-6”)—the value returned by
REL is 0. A and B are thus “minimally similar.”
David Lewin intended his REL to determine the degree of proportionality
between the subset contents of two PC sets. (To apply the “ginger” analogy
again: the taste of ginger in a dish does not change if one adapts the amount
of ginger to the amounts of the other ingredients.) Measuring PC set similar-
ity against size, he continued the line of investigation that had also yielded
Morris’s ASIM and Rahn’s ATMEMB. All these measures embody the following
basic ideas:
Statement 1:
A PC set is equal to the sum of its subsets.
But not all subsets were counted; not even all the proper ones. A count either
included all proper subsets consisting of two or more elements (such a count
was fundamental to Rahn’s ATMEMB or Lewin’s REL), or it included only those
consisting of two elements (as with Morris’s ASIM or Lewin’s REL2; from the per-
spective of Statement 1, the latter are consistent with Forte’s R0, R1, and R2, and
with Teitelbaum’s s.i.).
Statement 2 (“No pair of PC sets is excluded from the similarity relation”)
was already fundamental to the measures of Teitelbaum and Regener. The fol-
lowing two statements, however, concern the specifics of the similarity measures
discussed in the present section:
Statement 3:
Two PC sets are more similar as they contain more subsets that correspond
with respect to class.
Morris used the concept of an interval class (APIC). In this regard he followed
Teitelbaum. Morris’s values, too, increase as the similarity decreases. Rahn and
Lewin considered classes of subsets as being defined by a group of canonical
transformations, but they did not specify this group in their formulae.
Statement 4:
The correspondence takes on a higher value when the total number of sub-
sets in both sets is smaller.
29. Nor did Rahn. In the latter’s MEMBn and ATMEMB, however, the exclusion was
not the automatic result of the calculation. Rahn needed to add a clause to this effect to
his formula, which Lewin rejected as an “arithmetic awkwardness” (Lewin 1980, 501).

similarity 173
Every single subset is supposed to contribute to the identity of a PC set, just as

every instance of an interval class is supposed to contribute to the identity of a
chord. But when a PC set is compared to other PC sets, the absence of particular
classes of subsets is equally important. (For example, when a major triad is com-
pared to a chord with some dissonant intervals, the absence of such intervals
in the former acquires incidental significance.) “Being similar to A” not only
means “having a considerable number of things in common with A,” but also
“not having too many things that are absent from A,” or otherwise “not lacking
too many of the things that are found in A.”
An Almost Unreasonable Wish
In 1990, Eric Isaacson judged most similarity measures that had been developed
so far “inadequate.” Isaacson, who had chosen PC set similarity as a topic for his
dissertation at the University of Indiana, and who would reanimate the discus-
sion about it in the 1990s, found that no existing measure succeeded in ruling
out completely the influence of the cardinal number.
Few things are more natural and predictable than a range of similarity degrees
varying with the sizes of the PC sets in question. REL2 covers the whole range of
values from 0.0000 to 1.0000 when a trichord is compared to other trichords,
whereas it covers only a small range (from 0.5690 to 0.6395) when this trichord
is compared to nonachords. This makes perfect sense. Even though there are
as many set-classes of cardinal number 9 as set-classes of cardinal number 3, the
former simply cannot differ as much (cf. note 26).
However, Isaacson was concerned by this “inconsistency” (Isaacson 1990, 13).
He presented a measure ensuring the widest possible range of values for any
pair of cardinalities, thus fulfilling the almost unreasonable wish that the similar-
ity debate had been pursuing from the beginning; as Allen Forte had written in
1964: “[to] define an appropriate, musically meaningful similarity relation for
any two sets of different cardinality. . .”
Isaacson’s IcVSIM (“Interval-class Vector Similarity”) was an application of
a statistical tool: the “standard deviation function.” Given a series of numerical
values representing the frequencies of items in a sample, the standard devia-
tion function measures the spread of these values about their mean. In the
case of IcVSIM these numerical values were taken from the APIC difference
vector (the “Interval-difference Vector” in Isaacson’s words,”IdV” for short) of
two PC sets.
The difference vector had played a role in Teitelbaum’s concept of PC set
similarity, too. We have seen that his s.i. assigned a value to the difference vector.
This value increased with the differences between the corresponding terms of
the two APIC vectors. Since these differences tended to be greater in the case
of PC sets with different cardinal numbers, Teitelbaum’s s.i. failed as a measure

174 similarity
of the similarity between such sets. It was unable to return the lower values that
indicated greater similarity, except with pairs of equal-sized PC sets. As said
before, this was not a disadvantage in itself. However, by relating the differences
to their mean, one could undo the dramatic effect of a difference in size on the
s.i. of two PC sets. Here is how Isaacson defined his IcVSIM, which he called “a
scaled version of Teitelbaum’s s.i.”:
IcVSIM(X,Y) = σ(IdV), where

IdV = [(y1−x1) (y2 −x2 ). . . (y6 −x6 )]
Here, X and Y are IcVs [“interval-class vectors,” that is, APIC vectors] and σ
represents the standard deviation function. The standard deviation function,
defined in terms of the Interval-difference Vector, is:
σ= ∑(IdVi − IdV) 2
6
where:
IdVi = the ith term of the Interval-difference Vector, and
IdV = the average (mean) of the terms in the IdV
(Isaacson 1990, 18)
What does IcVSIM actually measure, if it is not just the difference in representa-
tion of the interval classes? It measures how evenly (or unevenly) this difference
is distributed over the various interval classes. A constant difference between the
entries of two APIC vectors, however big it might be, yields the value 0, indicating
maximal similarity. When the PC sets X and Y are of the same size, this constant
difference can only be 0. This means that X and Y are maximally similar if they
have equal APIC vectors—in other words, if they are related by transposition or
by inversion-plus-transposition, or if they are Z-related. When they do not have the
same number of elements, it is still possible that they are maximally similar if they
meet the following condition: the surplus of unordered PC pairs in the largest
PC set should amount to an integral multiple of six, so that it can be distributed
evenly over six interval classes.30 Isaacson provided the following example, involv-
ing the PC sets {0,3,6} (Forte name “3-10”) and {0,1,3,6,7,9} (Forte name “6-30”):
{0,3,6}: 002001
{0,1,3,6,7,9}: 224223
30. Isaacson stated this in more general terms: if the above-mentioned condition were
met, X might share an IcVSIM value relative to Y with a set of another cardinal number
(Isaacson 1990, 22). Thus, IcVSIM(X,Y) might equal IcVSIM(Y,Y) (maximal similarity).

similarity 175
Since a PC set of cardinal number 6 contains twelve more unordered PC pairs

than a PC set of cardinal number 3 (fifteen versus three), the only possible
constant difference between the APIC vector entries of these sets is 2. Isaacson
found eleven instances of maximal similarity between PC sets of different sizes
(Isaacson 1990, 19–22 and 27–28, note 16). The cardinal numbers of these PC
sets were 3 and 6 (three versus fifteen unordered PC pairs: a surplus of twelve
in the largest set; constant difference: 2), 5 and 8 (ten versus twenty-eight PC
pairs: a surplus of eighteen in the largest set; constant difference: 3), or 6 and
7 (fifteen versus twenty-one PC pairs: a surplus of six in the largest set; con-
stant difference: 1).
There are two more categories of PC set relations satisfying the above-
mentioned condition for maximal similarity in terms of IcVSIM. In PC sets
of cardinal number 7 there are eighteen more unordered PC pairs than in
PC sets of cardinal number 3 (twenty-one versus three). This surplus can be
divided by six, which is also true of the surplus in PC sets of cardinal number
9 with respect to PC sets of cardinal number 4 (thirty-six versus six, a surplus
of thirty PC pairs). However, these combinations of cardinal numbers do not
produce a single instance of maximal similarity.
One would expect members of complement-related set-classes to be
maximally similar in terms of IcVSIM. The complement theorem asserts
that the difference between the frequencies of an interval class in two com-
plementary PC sets is always equal to the difference between their cardinal
numbers (see chapter 4, p. 147). Unfortunately, this is not true of APICs.
The complement theorem was conceived with reference to PICs (Lewin
1960). And the PIC vector of a PC set differs from the APIC vector in that
the frequency of PIC 6 is twice the frequency of APIC 6 (see chapter 2,
tables 2.2 and 2.3). When we apply IcVSIM to PIC vectors instead of APIC
vectors, all complementary PC sets will turn out to be maximally similar;
but then {0,3,6} and {0,1,3,6,7,9} no longer deserve the designation they
received from Isaacson:
{0,3,6}: 00200200200
{0,1,3,6,7,9}: 22422622422
There is no constant difference between the corresponding terms of

these PIC vectors. When IcVSIM were modified to deal with PIC vectors
(a modification that would entail the addition of five more terms to the
“Interval-difference vector,” and the substitution of 11 for 6 as a divisor
in Isaacson’s formula), this would not be a case of maximal similarity. Of
all the similarity measures discussed in this chapter, only Teitelbaum’s s.i.

176 similarity
and Isaacson’s IcVSIM are profoundly affected by the use of PIC vectors
instead of APIC vectors. 31
We can start a summary of Isaacson’s ideas on PC set similarity with the first
two statements related to Teitelbaum’s s.i: “PC pairs (intervals, subsets of cardi-
nal number 2) are the key attributes of a PC set” (Statement 1); and “No pair of
PC sets is excluded from the similarity relation” (Statement 2). The third state-
ment, however, cannot be maintained. In Teitelbaum’s view, PC set similarity was
dependent on the differences between the frequencies per interval class. For
Isaacson’s IcVSIM, the following statement holds:
Statement 3:
The similarity (or dissimilarity) of two PC sets depends on the variation in dif-
ference between the frequencies of the successive interval classes.
The less marked the variation, the greater the similarity. The differences
themselves are of lesser importance. This means that PC sets of different car-
dinal numbers can be as similar as PC sets of the same cardinal number.
IcVSIM was not the last similarity measure to be developed. The debate
on PC set similarity has been ongoing, and has yielded several new mea-
sures, of which Marcus Castrén’s recursive relation (RECREL; Castrén 1994)
is the most prominent. Castrén not only compares two PC sets, but also all
subsets of these PC sets in terms of their subset contents, using a vector that
represents the percentual shares of all subset classes of a given cardinal-
ity in a PC set. The result is a conceptually complex, but highly nuanced,
measure of PC set similarity. Another noteworthy contribution is Robert
Morris’s pitch measure (PM), a similarity measure for pitch sets (instead of
PC sets; Morris 1995, 226). It registers the number of pitches and the num-
ber of (absolute) pitch intervals shared by two such sets. Its value lies not so
much in its subtlety or universal applicability as in its relative simplicity. It
is easy to integrate in everyday analytical practice, and capable of yielding
sensible results, especially when applied to open textures.
The increasing number of similarity measures called for surveys to evalu-
ate them, which were provided by authors such as Richard Hermann (1993),
Marcus Castrén (1994), Michael Buchler (1997), and Ian Quinn (2001). It
also elicited critical comments. Certain assumptions that had been taken
for granted began to meet with serious doubt. Isaacson (1996), for exam-
ple, questioned the commonly accepted idea that each interval class is
equally different from every other one. While acknowledging the merits of
the previous, mostly PC-based, similarity measures (including his own), he
advocated taking into account more dimensions than pitch alone.
31. This is a consequence of the fact that difference values—one of which may
change when the PIC vector replaces the APIC vector—are squared.

similarity 177
As stated before, there seems to be no similarity measure whose utility has

gained a general consensus. The second edition of Joseph Straus’s Introduc-
tion to Post-Tonal Theory (2000)—a book intended to represent PC set theory
to an audience of undergraduate music majors (p. vii)—does not mention
the topic at all. Yet, it is this topic that has been stimulating years of intensive
labor, reflected in papers that have won acclaim for their subtlety and pen-
etration. Apparently, it is so engaging intellectually as to survive the lack of
lasting success. How are we to assess all the work that has been done in the
field of PC set similarity?
Similarity relations provide an open concept of musical relatedness,
a concept yielding what John Rahn has called “one monstrous graph,”
in which each PC set is connected to all the other ones. Such a concept
enables the composer to trace—and the analyst to retrace—“similarity
paths” (Rahn) much in the same way that Hindemith created his “tension
paths” (his harmonisches Gefälle). Like the latter, similarity paths can be more
or less curved, depending on the similarity values of the PC sets underly-
ing successive sonorities. The similarity concept enables PC set theory to
address music as a temporal phenomenon. This is much more difficult to
accomplish with the common forms of PC set equivalence.
Similarity relations supplement equivalence relations in another way as
well. When we check a sample of sets for equivalent pairs, we show them
to satisfy (or not) an extrinsic criterion of relatedness. The term “extrin-
sic” means that this criterion is not necessarily a property of the two sets.
For example, when we say, with Allen Forte (1973), that equivalent PC sets
should reduce to the same prime form, we do not require them to be in
prime form. When we say, with John Rahn (1980a), that they should be
members of the same equivalence class, we do not demand that they appear
with all other members of that class. Actually, we do not even compare these
PC sets themselves, but just the results they yield after going through a spe-
cial kind of identification test. For the investigation of similarity relations
we define a property that each PC set exhibits in a specific way, or in some
measure. Our comparison now concerns these ways and measures, and is
thus more direct.
It is for this reason that similarity relations are supposed to accord with an
intuitive understanding of music better than equivalence relations do. They
bespeak a concern with musical experience, and with musical freedom, which
many found lacking in PC set theory originally. However, the purpose of mea-
suring these relations has been strangely ambiguous. For all their openness,
similarity measures have also been created to enhance our conceptual con-
trol over music. They require intellectual skill; the freedom to follow which-
ever similarity path can only be enjoyed at the highest level of abstraction.
What we see here is the interdependence of purpose and competence, the
self-perpetuating power of a technical vocabulary. Although it may seem that

178 similarity
similarity measures helped PC set theory deal with issues beyond its original
purview, these issues actually came up in response to the challenges implicit
in its own apparatus. The set and the row vector facilitated a comparative
study of relations among tone combinations, while perhaps seriously limiting
its scope in some ways.

Chapter Six
Inclusion
Much of what we call “pitch-class set theory” today originally served as the intro-
duction to a theory. An inventory of classes of musical objects, and of relations
between these classes, its function was to prepare the ground for a model of
large-scale pitch organization in music of the twentieth century: the pitch-class
set complex. This model was the main focus of Allen Forte’s seminal article “A
Theory of Set-Complexes for Music” from 1964 and of his book The Structure of
Atonal Music from 1973.
Set-complex theory deals with the analysis of entire compositions, or move-
ments and sections of compositions. More specifically, it deals with the question
of how these can be unified in terms of PC sets. This requires a PC set relation
that links several PC sets across boundaries of cardinality (unlike Tn/TnI equiva-
lence), but also sets a clear limit to their number (unlike the various “absolute”
similarity measures, like ASIM, ATMEMB, and REL). In other words, it should
delineate a cross-cardinality “family” of PC sets. Such a relation is the inclusion
relation. “Inclusion” means that one set is contained in another; the two sets
are related as subset and superset. It is on this relation that the PC set complex is
primarily based.
Over the years, the set complex faded into the background somewhat. Ironi-
cally, it never had the appeal of the ideas that once built up to it; even a topic so
notoriously abstruse as PC set similarity aroused more interest and discussion.
Although set-complex theory failed to make a lasting impact on the study of
post-tonal music, Forte remained dedicated to the aim of fitting PC set relations
into a larger pattern. In the late 1980s, he embraced the PC set genus as a source
of overarching coherence in music (Forte 1988a). This, too, is a “family”-like
concept based on PC-set inclusion. However, it is both more restrictive and more
universal in scope than the set complex. It is more restrictive, because one usu-
ally distinguishes a relatively small number of distinct PC-set genera;1 and it is
more universal in scope, because genera “exist” independently from individual
musical works. The set complex is defined contextually (like the subject of a
fugue, the theme of a variation cycle, or a twelve-tone series), while the genus is
defined communally (like “sonata form,” the Dorian mode, or C major).
The present chapter relates such concepts to tonal theory. It concentrates
mostly on the set complex, because of its role in the emergence of PC set theory.
1. Forte (1988a) lists twelve PC-set genera, Parks (1989) only five, compared to 114
set complexes.

180 inclusion
Considering that PC set theory was predicated on the structural autonomy of

“atonal” music, it is an interesting idea that it originally paralleled older sche-
mata of music explanation. The elaboration of this idea in subsequent pages
will necessarily involve some speculation. Forte was reluctant to interpret twen-
tieth-century works with reference to historical models. For Forte, the music of
Schoenberg or Stravinsky was interesting not for what it retained from the past,
but for the new knowledge he thought it embodied. And he believed that this
knowledge was only accessible through systematic observation. Hence Forte’s
unceremonious introduction of the set complex in The Structure of Atonal Music.
Where it came from mattered less than what it could tell about pitch relations in
the classical works of the twentieth century:
[The set complex] provides a comprehensive model of relations among pc sets

in general and establishes a framework for the description, interpretation, and
explanation of any atonal composition. (Forte 1973, 93)
Forte’s cursory introduction of the set complex recalls the taken-for-grantedness

of his ideas on PC-set similarity. However, the previous chapters have shown that
such ideas did not always evolve uniquely in response to new musical develop-
ments; they may already have their locus in the world. This chapter proposes
an ontology of PC set inclusion that shows the tonal descent of the set complex;
and, as a corollary, it offers a view of the tonal system through the lens of set-
complex theory.
An Analytical Exploration
Inclusion is the correlate of segmentation. Since music is organized in segments

of various kinds and magnitudes,2 there are many musical relations to which the
term “inclusion” is applicable. It may refer to relations between motifs and phrases,
intervals and chords, chords and polychords, or linear progressions and themes.
2. This seems a point of consensus between creative, analytical, and cognitive per-
spectives on music; witness the following three quotes: “It will be useful to start by building
musical blocks and connecting them intelligently. These musical blocks (phrases, motives,
etc.) will provide the material for building larger units of various kinds, according to the
requirements of structure” (Schoenberg 1967, 2); “While the phrase may be thought of as
a segment of some independence, it is usually associated with one or more other phrases
as an integral part of a larger structure” (Leon Stein 1979, 24); “If confronted with a series
of elements or a sequence of events, a person spontaneously segments or “chunks” the
elements or events into groups of some kind” (Lerdahl and Jackendoff 1983, 13). In some
publications the term “segment” functions at a higher level of abstraction than it does
here. Morris (1987) uses it as the designation of an ordered set, reserving the term “set”
itself for unordered sets.

inclusion 181
A closer examination of such relations reveals that the superset/subset definition

of that term is not always adequate or appropriate. Firstly, segmentations can be
ambiguous; and, secondly, the structural identity of each individual segment may
not reside in the sum of its elements and their associated intervals.
Consider, for example, Gabriel Fauré’s Élégie for Cello and Piano, Op. 24
(1880). On the face of it, the main melody in C minor (mm. 2–5) is divided
into four segments of one measure each by sequential repetition (ex. 6.1a).
Harmony, however, decides otherwise. Note the crucial difference between the
E♭4 in measure 2 and its melodic correspondent B♭3 in measure 3. The for-
mer is a relatively stable tone: the third of the tonic chord; the latter is the very
first accented dissonance. As such, it is the peak of an initial tension curve. This
means that B♭3 is still part of the first segment, which ends more convincingly
with its resolution to A♭3 than with the metrically weak G3 in measure 2. The
actual process of repetition, then, starts on the third beat of measure 3, and
another segmentation of the melody suggests itself (ex. 6.1b).
Although this segmentation is the most preferable from the viewpoint of
musical phrasing, it does not discard the previous one. The melodic parallelism
between the successive measures is too obvious to be missed; and such alterna-
tive, if subordinate, segmentations can gain in meaning as the piece progresses.
Other considerations may then come into play. For example, it may at first not
seem right to detach the eighth notes in measure 2 from the subsequent appog-
giatura. Yet, it is this eighth-note figure that launches the middle section in A♭
major (mm. 23ff.) and serves as an accompaniment to its ornate theme. This is
a reason to rate it as an important segment of the main melody (ex. 6.1c). Seg-
mentations, then, apart from possibly being ambiguous in themselves, may rest
on criteria of totally different orders.
Now, if the eighth-note figure represents an important segment of the Élé-
gie’s main melody, how should we characterize it? The first two forms shown in
example 6.1c (mm. 2 and 23) have the same succession of intervals, and each of
them occupies the same position relative to the local tonic (C and A♭, respec-
tively)—an interesting combination of properties in view of the modal change
from minor to major. However, the figure retains neither these intervals nor this
harmonic position over time. Apart from its rhythm, its most permanent charac-
teristic is the succession of an accented higher chord tone and an unaccented
lower one, the former being embellished by its upper neighbor.
It is hard to express such a characteristic set-theoretically. As will be recalled
from chapter 2, a PC (or pitch) set offers no differentiation between chord
tones and embellishments, or between accented and unaccented tones. Nor
does it differentiate between chord tones that are part of a melodic progres-
sion (such as C4 in ex. 6.1a, m. 2) and chord tones that are not (such as G3 in
the same measure). Seen thus, a PC set is a highly indeterminate concept. And
the general tendency of PC set theory has been toward even greater indetermi-
nacy—perhaps to expand its range of application, or to reflect the uncertainty of

182 inclusion
2
5
sempre
Example 6.1a. The main melody of Gabriel Fauré’s Élégie for Cello and Piano, Op. 24.
2
5
sempre
Example 6.1b. A segmentation of Fauré’s melody, based on considerations of

harmony and phrasing.
23 sempre molto adagio
26
sempre
Example 6.1c. A small segment and its transformations.
musical perception. The shift of focus from equivalence to similarity is a case in

point. Another typical example has been the development of a “theory of con-
tour” within the purview of set theory by Michael Friedmann (1985) and Robert
Morris (1987, 1993). A contour is an ordered set of pitch elements that are not
specified beyond their relative heights. This concept can help us cope with the
flexible identities of themes and motifs in tonal music. In example 6.1c, most
forms belong to the same contour class3—that is, they are contour-equivalent—
because their first tone is lower than the second, higher than the fourth, and
3. As defined by Friedmann (1985, 227). Morris adopted the term “contour class” in
1987, but replaced it with “cseg” (“contour segment”) in 1993.

inclusion 183
(a) (b) (c) (d)
A E Mm7 A B Mm7
(e) (f) (g)
E A E
Example 6.2. Members of a contour class (cf. ex. 6.1c).
equal to the third. However, other members of this class may not satisfy our ear-
lier description of these forms (ex. 6.2, the members labeled f and g). There-
fore, contour alone is not a distinguishing feature of Fauré’s eighth-note figure.
An adequate description of it ought to take into account the functional relations
between its elements.
At this point, the reader may be reminded of Fred Lerdahl’s doubts about
the musical relevance of the set concept (see chapter 1, pp. 23–25). His observa-
tion that “some pitches are heard as more or less structural than other pitches”
certainly holds true for all forms of the eighth-note figure in Fauré’s Élégie. And
it may also apply to tone combinations in the post-tonal repertoire that have
been treated in the literature as reified PC sets. Is the opening melodic line
of Schoenberg’s Piano Piece, Op. 11, no. 1 (ex. 6.3a, mm. 1–3) an occurrence
of “Set [-Class] 6-Z10,” as Forte had it in his analysis of this piece (“The Mag-
ical Kaleidoscope,” 1981, 136)? Such a statement implies that each pitch has
the same importance. It thus sanctions a comparison of this line with any other
six-tone combination. For example, Forte compared it to the left-hand part of
measures 4–5, which represents “6-Z39.” The PC sets in this class are the comple-
ments of those in “6-Z10,” which means that they have the same interval vectors
(333321). And he regarded the two accompanying chords in measures 2–3 as
another set of six PCs, as did Gary Wittlich (1974) before him. These chords
jointly represent the Fortean set-class “6-16,” which is related to “6-Z10” by the
common inclusion of, among others, “3-3”—the set-class of both the head motif
(B4,G♯4,G4) and the chord {B♭2,A3,D♭4}(ex. 6.3b).4
4. The inclusion of members of the same set-classes in two PC sets can be taken as
a measure of the similarity between these PC sets. Rahn’s MEMB-number, discussed in
chapter 5, is such a measure. A PC set of class “6-Z10” includes four members of “3-3,”
whereas a PC set of class “6-16” includes two. Hence,
MEMB3(“3-3,” “6-Z10,” “6-6”) = 6 (cf. Rahn 1980b, 492)
Of course, this number is only meaningful in comparison to similar numbers obtained

from other pairs of PC sets, or as part of a summation of all set-classes represented in the
two mentioned sets (Rahn’s TMEMB).

Mäßige 5
10
langsamer
rit.
Example 6.3a. Arnold Schoenberg, Piano Piece, Op. 11, no.1, mm. 1–11.
''6-Z10''
2 3
''3-3''
''6-16''
2 3 ''3-3''
Example 6.3b. A PC-set analysis of the first three measures of Schoenberg’s Piano Piece,
reflecting Forte’s emphasis on the two six-tone combinations in the right and left hands.
2 3
Example 6.3c. George Perle’s analysis of the same three measures.

George Perle, Serial Composition and Atonality: An Introduction to the Music of Schoenberg, Berg, and
Webern. Copyright © 1991, the University of California Press.

inclusion 185
The head motif is obviously part of a larger whole, and the melodic line that
includes it is itself a recurring element. But how do we know that the second
chord forms one structural unit with the first? Don’t we hear these chords as
two syntactically related events rather than as a single event taking two measures
(including a rest) to unfold? It is true, PC sets need not coincide with units of
musical syntax such as motifs or chords. But segmentation of a musical surface
into PC sets should nevertheless depend on a judgment of plausibility.
In this case, the segmentation is plausible. Whereas in this piece “6-Z10” is
primarily associated with restatements of the opening melody (mm. 34–36, 46–
47, and 53–55), “6-16” presents itself in different ways, as shown in example 6.4.
The six segments (a–f) in this example are all based on Tn-related forms of it.
We can view the succession of the two chords in measures 2–3 (a) as a partial
ordering on the PC set {5,6,9,10,11,1}, in the sense that each element of the subset
{5,6,11} precedes any element of {9,10,1}.5 The Tn-forms in the same column of
example 6.4 (b–d) preserve this ordering. In other words, the order relations
between the elements of {5,6,9,10,11,1} also hold between their correspondents
under Tn. Consider the relation between example 6.4a and 6.4b:
T3 ({5,6,11},{9,10,1}) = ({8,9,2},{0,1,4})
This partial ordering, together with the pointed beginning of the piece, empha-
sizes the role of “3-3” as a class of subsets held in common by the PC sets of “6-
Z10” and those of “6-16.” In the first column of example 6.4, “3-3” is represented
by the last tones of each segment. In two segments (b and d) these are also the
highest tones. The second column shows two occurrences of {1,2,5,6,7,9}—a set
we already know (see ex. 2.8), and another member of “6-16.” Here, “3-3” is even
more prominently set off as a subset. Each segment incorporates two statements
of the head motif, transposed and with octave displacements: (F4,D4,C♯5) and
(A4,F♯4,F5).6
The relations shown in example 6.4 assume the arbitrariness of the musical
surface. They exist independently of melodic shape and harmonic function.
Under the same assumption, George Perle (1962) interpreted the first measures
of Opus 11, no. 1 as a dense sequence of operations on the head motif (ex.
6.3c). Clearly, Perle’s segmentation is committed to the consistency of the inter-
vallic structure rather than to the integrity of the melodic line, which appears
5. The discussion of partially ordered sets was prompted by the “license of simul-
taneous statement of pitch-classes” in serial music (Babbitt 1962, 60). The concept of a
partial ordering enabled the study of relations between twelve-tone series beyond the 48
canonical transformations. Lewin (1976) explored this territory. Morris (1987, 198–202)
discussed the concept as part of a wider range of pitch-resources.
6. Recall that any PC set of class “6-16” includes two members of set-class “3-3” (see
note 4 above). In the last two segments of example 6.4 (e and f), this property is exploited
to the maximum.

(a) mm. 2−3
T8
(b) m. 12 (e) m. 13
[]
T3
(c) m. 29
T3
(d) m. 39 (f) mm. 49−50

T8
T11
T0
Example 6.4. The representation of Fortean set-class “6-16” in Schoenberg’s Opus

11, no. 1.

inclusion 187
as a composite of the higher tones of successive segments. But in view of the

year of composition (1909), and in the absence of any detailed information on
the actual conception of this piece,7 melodic and harmonic patterning cannot
simply be set aside as a function of what Forte has called “external form” (Forte
1981, 130). Now that we have examined the relation of the opening melody to
its accompaniment on PC-set basis, let us return to the question concerning this
melody: is it the realization of a PC set, or does it just represent itself?
As mentioned, “6-Z10’s” association with the opening melody of Schoenberg’s
piano piece is quite exclusive. Other occurrences of this set-class form a somewhat
heterogeneous group. Forte found eight of them, but not all of these are as con-
spicuous as the occurrences of “6-16” (ex. 6.5, cf. ex. 6.4).8
This suggests that mere content is not the essence of this melody. Rather, its
arrangement is what matters. The essence is a linear mold that can also accom-
modate PC sets of other classes, like “6-21” (mm. 9–11) or “5-Z17” (mm. 50–
52). This mold imposes a hierarchy on the elements of these sets. It gives us a
sense of tonal rhythm, which invites us to hear A4 in measure 2 as an incomplete
neighbor of G4, and F4 in measure 3 as a retardation of E4.9 Such hierarchical
distinctions challenge a set-theoretical interpretation of Schoenberg’s piece as
much as Schoenberg’s choice of harmony challenges a tonal interpretation.
It would be wrong to assign structural significance to the PC set of the open-
ing melody, if for no other reason than that the accompaniment represents a
set of such a significant class (i.e., “6-16”). We simply cannot offer a judgment
of plausibility as to the claim that this melody represents a PC set. It definitely is
a source of smaller sets, such as those of class “3-3,” but should these be consid-
ered subsets? Let us draw a few lines:
• A first condition for the musical pertinence of the set-theoretic inclusion

relation is that musical segments can be plausibly defined on the basis of
their abstract content.
7. There are no extant sketches for the Three Piano Pieces, Op. 11, and the only
theoretical text by Schoenberg that might shed a contemporary light on these pieces is his
Theory of Harmony of 1911. In this text, one searches in vain for references to procedures
so obviously foreshadowing the serial techniques of later years.
8. When hearing example 6.5e (mm. 45–46), the listener may recognize the suc-
cession of a minor third, i.e., {E4,C♯4}, and a major third one step lower, i.e., {D4,B♭4}.
This succession is also a characteristic of the opening melody, in which {A4,F4} follows
{B4,G♯4}. In measures 45–46, however, A3 and C3 should be added to form a set of class
“6-Z10.” A3 connects to B♭4 as E4 did to F4 in the original melody, but the role of C3—a
mere grace note to the following C♯4—is in no way comparable to that of its correspond-
ent G4 in measure 2. And if it is counted as a member of this segment, why exclude D♯3
in measure 44? Although in this excerpt the reference to the opening melody is obvious,
it does not obviously involve the melody’s PC set.
9. Lerdahl, in his analyses of this Piano Piece (1989, 82–84 and 2001, 353–66), also sees the E
as more important, though on the basis of a psychologically founded system of preference rules.

188 inclusion
(a) mm. 1−3
(b) mm. 9−11 (e) mm. 45−46

T0
T7
(c) m. 34 (f) m. 51
T3 I
T11 I
(d) m. 39 (g) m. 61
T9
T2
Example 6.5. The representation of Fortean set-class “6-Z10” in the same piece.
This condition is met by the accompaniment of the opening melody of Schoen-

berg’s Opus 11, no.1, but not by the melody itself, though it is met by the three-
tone head motif of this melody. The main melody of Fauré’s Élégie, however, also
fails to meet the condition, and so does the eighth-note figure that becomes so
important in the middle section of that piece.
• A second condition for the musical pertinence of the set-theoretic inclu-

sion relation is a salient representation of the set-classes of both the sub-
set and its superset.

inclusion 189
This is especially important in view of the role of this relation in the delinea-
tion of set-families (see further below). But what is “a salient representation”?
“Salience” is the generic term that Fred Lerdahl (1989) has used to denote a
number of properties by which a “pitch event”—that is, for Lerdahl, a tone or a
chord—can stand out as important in a sequence of such events. These proper-
ties range from the more sensory (for example, duration, loudness, or registral
position) to the more abstract (for example, structural prominence). In Ler-
dahl’s wake—but not quite in his spirit!—John Doerksen (1994, 1998) has devel-
oped salience criteria for set-classes, to determine which classes represented in
a composition have “the highest degree of involvement at the musical surface”
(Doerksen 1998, 196). Doerksen checks a composition for relevant segments,
thereby exclusively following boundaries created at the musical surface (voices,
simultaneities, rests). He then identifies these segments by their set-class names.
The degree of involvement of a set-class depends on the musical functions of the
segments that represent it (for example, the articulation of a formal division),
and on the number and nature of its relations to other set-classes in the compo-
sition (equivalence, complementarity, inclusion).
We need not fix the number of salience categories, nor restrict ourselves to
set-classes found on the musical surface. However, for the set-theoretic inclusion
relation to reflect a particular type of pitch organization, members of both set-
classes should indeed be woven into the musical fabric. To put this more point-
edly, for a relation to exist, its terms must exist. And the terms of an inclusion
relation typically do not exist through this relation, as one might claim for the
terms of an equivalence relation. This idea merits some elaboration:
With the Help of the Family

It is true, as we have said, that the single things are nothing as deter-
mined by their relations which are the negation of their singleness, but
they do not, therefore, cease to be single things . . . On the contrary,
if they did not survive in their singleness, there could be no relation
between them—nothing but a blank featureless identity.
—Thomas Hill Green, Prolegomena to Ethics
In a world of objects, one object can gain identity from a property it has in com-
mon with other objects. It is then no longer an anonymous individual object—
one out of many—but represents, or instantiates, this particular property. In
turn, the property concerned takes on general value. It is no longer restricted to
only one object. Consider the now familiar example of a tone that instantiates a
PC, for example E♭: this tone can be accentuated by other tones with which it
shares the E♭ property. But the property does not exist without multiple tones
sharing it. It is actually an equivalence relation that calls “E♭” to life.

190 inclusion
In so speaking, we involve ourselves in the metaphysical debate on “univer-

sals”: things that can be said of multiple concrete objects; things like “red,” “two
meters tall,” “parent,” “E♭.” Metaphysicians often distinguish between two kinds
of such things, namely, properties and relations. However, the example of a PC
shows that there can be much reciprocity between them. The term “property” is
used here as including relational as well as non-relational attributes.
What is it about universals that has triggered a debate? Philosophers struggled
with the question “how there can be oneness in the multiplicity” (Rodriguez-
Pereyra 2000, 256), or “how numerically different particulars can nevertheless be
identical in nature” (Armstrong 1978, 41). Universals have served as an answer
to these questions, but their ontological status has always been a matter of con-
troversy. Do they exist independently from the individual objects—the “particu-
lars”—to which they apply? Do they exist in these objects? Or do we only use
them as convenient labels for these particulars, thereby imposing distinctions of
language on a world that is otherwise unfathomable?10
PC set theory seems founded on the conviction that it makes sense to speak
of universals as real things, that what matters about an object is what can be
abstracted from that object. We have seen that it defines pitch events in terms
of certain properties, like PCs, interval classes, PC sets, interval vectors, etc. It
examines the relations between these properties, and it tacitly assumes that these
relations also hold between their bearers.
This assumption has drawn criticism. In the words of James A. Davis:
Once pitch-classes become the elements of an analysis, one is no longer dealing

with the pitches found within a composition. Instead, the analyst is manipulat-
ing classes which are by no means the same entities as the pitches which consti-
tute their membership. The problem is intensified when these pitch-classes are
gathered to form pitch-class sets, and these sets are then subjected to various
manipulations. At this point we have set-theoretical functions applied to sets
whose members are classes. Unfortunately, the original pitches found within
the composition are no longer a part of the ontological domain in which the
theorist is working. (Davis 1995, 513–14)
Is Davis right to have concerns about the level of abstraction on which PC set the-
ory operates? Not if his concern is that such a level is unprecedented in the history
of Western music theory, or that it is a priori musically irrelevant. When PC set
theory began, the notion of octave equivalence was already firmly established in
musical thought, and the concept of a PC set had its referents in compositional
practice. Were this not the case, the theory might not have achieved its institu-
10. Two text collections provide an overview of the debate: Properties, ed. by D. H. Mel-
lor and Alex Oliver (Oxford 1997), begins with contributions of Gottlob Frege and Bertrand
Russell; The Problem of Universals, ed. by Andrew B. Schoedinger (Atlantic Highlands 1992),
also covers ancient Greek and Medieval philosophy (Plato, Aristotle, Abelard, Aquinas).

inclusion 191
tional status. What Davis deplores in PC set theory (its abstraction) is actually con-
tinuous with the Western classical music-theoretical tradition. His observation of
an ontological ambiguity is well taken, but such ambiguity does not disqualify the
theory as a member of this tradition.
A PC set—or a PC set-class, for that matter—exists in the same way that “E♭”
exists: in a give-and-take with multiple concrete objects. It exists insofar as it is
instantiated by different tone combinations.11 In turn, it supplies these tone
combinations with a common structural identity. The important point here is,
however, that it contributes nothing to the reinforcement of an included smaller
combination or an including larger one. An inclusion relation is not based on
what its terms have in common. With respect to a set of sets, SET for short, the
proposition “A ⊆ X” (“A is a subset of X”) rather gives us a property (“⊆ X”) that
A might share with other sets ( A ⊆ X, B ⊆ X, . . ., X ⊆ X). Together, the sets with
this property form a subcollection of SET that we shall call the family of subsets of
X.12 In example 6.6a, its graphic image is a triangle with one upper vertex (asso-
ciated with X, the family’s “progenitor”) and two lower ones.13
Set A may not only have one such property. It may also be a subset of sets
other than X (A ⊆ Y, A ⊆ Z), and this holds true for any set with which it shares
the property “⊆ X.” Each of these inclusion relations (. . ., “⊆ X,” “⊆ Y,” “⊆ Z”)
generates a different, but not entirely distinct family (ex. 6.6b). The term “fam-
ily” is appropriate, because families, in the ordinary sense, have members in
common, members that are related by marriage: brothers-, sisters-, fathers-, and
mothers-in-law. And the term is applicable to various musical domains, not only
that of PC sets. Medieval hexachords, tonal keys, and twelve-tone series can be
seen as families (of diatonic pitches, of pitches and harmonies, and of series-
forms) loosely or tightly connected by the bonds of common membership.
The property “⊇ A” (“is a superset of A”) is the basis of yet another family. If
A is a member of the family of subsets of X, the family of supersets of A includes,
of course, X. In example 6.7a, the image of this family is a triangle with one
lower vertex (associated with the progenitor A) and two upper ones. The family
of subsets of X is still visible as a dotted triangle. Obviously, a family of n subsets
involves a further n families of supersets (ex. 6.7b).
11. For an illustration, see example 2.9.

12. The reader may wonder whether “the power set of X” is not a better term. However,
I intend the term “family” to have a broader meaning. In the following, it will be used for
sets of sets that are related by inclusion. This not only involves the subsets, but also the
supersets of a given set.
13. In his treatise on PC-set genera, Allen Forte (1988a) also uses the term “progeni-
tor.” In his view, the only PC sets that qualify as progenitors of a family (more specifically,
a genus) are those of cardinal number 3—first, because 3 is the smallest cardinal number
yielding more than one representative of an interval class; and second, because trichords
represent (uniquely or pairwise) all possible combinations of two APICs from 1 through
6. Each such combination forms the “DNA” of another genus.

(a) (b)
X Y X Z
B A C A
Example 6.6.
(a) A graphic representation of the family of subsets of X.
(b) Set A is a subset of a number of families.
(a) (b)
X X
A B A C
Example 6.7.
(a) A graphic representation of the family of supersets of A.
(b) Set X is a superset of a number of families.
(a) (b)
X X
A A
Example 6.8.
(a) Inclusion of the family of subsets of A.
(b) Inclusion of the family of supersets of A.

inclusion 193
All these families are unique.14 However, they are not easily delimited. They
cut across each other, and larger families contain smaller ones. Each family of
subsets that includes A also includes the family of subsets of A (ex. 6.8a). And
the family of supersets of A is included by the family of supersets of each subset
of A (ex. 6.8b).15
As a consequence of their criss-crossing boundaries, families of subsets and
supersets can be united to form larger entities; or they can be divided into
smaller ones. This suggests another reason why the term “family” serves us
well here. The size of a human family is equally indeterminate. The number
of its members varies with the “width of the frame.” We may think of a family
as a household of two parents and their children, or a group that additionally
includes grandparents, uncles, aunts, and cousins. We also use the term “family”
to refer to an entire genealogical tree. Furthermore, each member can be cho-
sen as a point of perspective on the family. From this perspective, certain mem-
bers are near while others are more remote. “Nearness” and “remoteness,” too,
are concepts that have a history in music discourse, like the concept of “having a
common family member”; but this history will concern us later.
To recapitulate, we are investigating the question of whether and when PC
set inclusion can rightfully be considered important. The answer may not lie in
the inclusion relation itself, but in the nature, power, and/or frequency of its
14. Theoretically speaking, the only two exceptions are the families of the universal set
and the null set. By their definitions (see chapter 2, p. 45), these sets have equal families.
15. For an accurate sense of the numerical constraints on the dimensions of such a
family network, we can take our perspective from one set, and calculate the number of
families of which it is a member. Let SET be the power set of ELEMENT. Further, let the
cardinal number of ELEMENT be n. Then, each set A in SET of cardinal number m is
a member of 2m families of supersets and 2n−m families of subsets (including such oddi-
ties as families with progenitors consisting of zero, single, or all elements). If we set n to
12—equating ELEMENT with PITCHCLASS—and m to 3, A is a member of eight families
of supersets and 512 families of subsets. If subsequently we set m to 7, A is a member of
128 families of supersets and 32 families of subsets.
These numbers are considerably smaller when we investigate the family affiliations of
not one, but two or more distinct sets. Let us take from SET an arbitrary number of sets,
A, B, C, . . ., regardless of cardinality. The number of families of supersets of which all of
A, B, C, . . . are members is dependent on the size of their intersection. If i is the cardinal
number of A ∩ (B ∩ (C ∩ . . .)), there are 2i families of supersets that contain A, B, C, . . .
The same sets can be found together in a number of families of subsets. This number,
however, depends on the size of their union. If u is the cardinal number of A ∪ (B ∪
(C ∪ . . .)), there are 2n−u families of subsets that contain A, B, C, . . . Thus, any two or
more sets belong to fewer families as their intersection is smaller, or, equivalently, as their
union is larger. In other words, the number of families involved decreases with the num-
ber of common elements between these sets. For example, if we keep n = 12 as the cardi-
nal number of ELEMENT, two sets of three elements with zero intersection are contained
in 64 families of subsets, 448 less than the number containing each of them.

194 inclusion
instantiation. “Including” a PC set, or “being included” in it, is a relation that

can become a property (like “E♭,” or like each PC set itself). It should then
be had in common with other PC sets, so as to transcend the multiplicity of
relations—not least inclusion relations—in the twelve-tone universe. Hence the
importance of families of subsets and supersets. But the abundance of such fam-
ilies poses a serious problem. How can PC set A instantiate the property “⊆ X” if
it is also included in Y, Z, and/or other sets? What makes X prevail as a superset
of A? And, conversely, how can X instantiate the property “⊇ A” if it also includes
B, C, and/or other sets? What makes A prevail as a subset of X? In short, how,
and to what extent, do these two PC sets exist?
Let us return once more to the opening melody of Arnold Schoenberg’s
Opus 11, no. 1 and its accompaniment. PC sets of highly involved set-classes are
those of the melody’s head motif, the chord succession in measures 2–3, and
the second chord of this succession (m. 3). We have not taken special note of
the first chord, {G♭2,F3,B3} (m. 2), which is a representative of class “3-5.” This
set-class is not as frequently and saliently represented in the piano piece as “3-3,”
the set-class of the head motif and the second chord. It does not coincide with
a foreground motif of the opening melody, nor does it notably occur indepen-
dently from “6-16.” And since, as we have seen, the opening melody cannot be
put on a par with its PC set, the only PC sets related as subset and superset are
those of the second chord and the chord succession as a whole.
This inclusion relation deserves to be noted, but it is not actually the focus of
instantiation. The subset and its superset exist by the properties that they rep-
resent for themselves, that is, by their recurrent set-classes. In other words, the
musical pertinence of PC set inclusion here actually rests on other relations (spe-
cifically Tn/TnI equivalence). Is it conceivable, though, that musically pertinent
relations rest on PC set inclusion?
The answer is yes, with the help of the family. Consider the family of subsets
of X. Recall that it is unique, in the sense that it is the only family containing
all these and no other sets. Although each member may also belong to other
families, X can be outlined by a number of them collectively. We have already
seen how in Stravinsky’s Agon, measures 414–15, the octatonic scale emerges
from a chain of interlocking three-tone combinations (see chapter 1, ex. 1.3c).
Each combination can be said to instantiate a portion of that scale by virtue of
appearing with the others. Thus, if A and X are PC sets represented in a musi-
cal composition, and A ⊆ X, A is a stronger instance of the property “⊆ X ” as
the number of PC sets with which it shares this property (B, C, . . . and X itself)
is larger. This number should perhaps not be the only measure of the “subset-
ness” of A, but for the moment we shall assume that under its increase “⊆ X” can
become “a way of being A.”
Similarly, the family of supersets of A can come to the aid of X, and pro-
vide it with “a way of being X.” All chords in the “march” section of Claude
Debussy’s Fêtes, the second of the orchestral Nocturnes, are members of the

inclusion 195
family of supersets of {8}, a one-element set. This is obvious, as the section is

built on a pedal A♭/G♯. But there is more. Debussy’s march theme has been
characterized as a “sui generis example” of a theme based on “non-tonally-func-
tional chromatic progressions” (Pomeroy 2003, 161). When viewed from our
“family” perspective, these progressions interact in an interesting way with the
structure of the theme, and with its instrumentation.
The march section begins in measure 116 (“Modéré mais toujours très rythmé”).
It is based on a repeated pair of phrases, α and β, each eight measures long. Muted
trumpets introduce this pair in measures 124–39 over an accompaniment of harps
and pizzicato lower strings. (The trumpet parts are shown in ex. 6.9a.) Woodwinds
take over in measure 140 with T3(α), and then play T6(β) in alternation with the
horns. Finally, from measure 156, trumpets and trombones play both phrases full
force at their original pitch amid a whirl of orchestral sound.
Example 6.9b shows the harmonic basis of the march section of Fêtes. All
chords are represented on the upper system. Below that we see a harmonic
reduction that gives visual prominence to the common tones between the struc-
tural chords. The parsimonious voice leading16 conceals the relation of model
and sequence between the progressions supporting phrase α and the whole of β
and T3(α). This relation is rendered distinct on the lowest staff in example 6.9b.
The next phrase, T6(β), could mark the beginning of yet another sequence.
However, the harmony loops back into the restatement of α.
As said before, the single unifying element of the entire chord succession
is the bass pedal A♭/G♯. But some other pitches are held for a considerable
time, too. Together with A♭/G♯, they define shorter spans of chordal accom-
paniment. We hear Fs throughout the first statement of phrase α. All chords of
phrase β, and a number of those supporting the subsequent T3(α), include B3
and D4. The remainder of T3(α) sees the return of F (m. 144), which does not
give way until the last statement of β (m. 164).
From these observations we can infer two things. First, apart from the family
of supersets of {8}, two smaller families can be distinguished: the family of super-
sets of {5,8} and the family of supersets of {2,8,11}. Second, the alternation of
these two smaller families stands in a contrapuntal relationship with the phrase
structure of the march. Example 6.9c shows where the families take over from
each other: at the beginning of the first statement of β, halfway through T3(α),
and at the beginning of the final statement of β. There is no such change at the
beginning of T3(α), nor at that of T6(β) or the final statement of α; but these
moments are marked by a major change of instrumentation, as if for want of
harmonic momentum.
16. Studies by Cohn (1996, 1997), Kopp (2002), and others provide support for the
idea that parsimonious voice leading—the transformation of one chord into another by a
minimum of melodic motion—was a principle of harmonic progression through a large
part of the nineteenth century.

124 3 3
α
129 3 3
β
134 3 3
Example 6.9a. Claude Debussy, the march theme from Fêtes, mm. 124–39.
α
124 128
Model:
β
132 136
Sequence:

T 3 (α)
140 144
T 6 (β)
148 152
Sequence: (interrupted)
Example 6.9b. The harmonic progressions underlying Debussy’s march.
Trp Wdw Hrn/Wdw Trp/Trb
α β T3 (α) T6 (β) α β
D D
F B F B
A A A A
{5,8} {2,8,11} {5,8} {2,8,11}
Example 6.9c. Changes of “family” and instrumentation plotted against the phrase
structure of the march.

198 inclusion
What do these families tell us about Debussy’s march? Perhaps this: certain
chord changes are “stronger,”17 more disruptive, than others, even when they
affect the same number of pitches or PCs, or when they are equivalent in some
way. For example, measure 144 may give us a stronger sense of harmonic change
than measure 128, in spite of the correspondence between these measures.
The progression from measure 143 to measure 144 eliminates a PC set that
has remained invariant under a number of chord changes: the set consisting of
G♯, B, and D. The progression from measure 127 to 128, on the other hand, is
the very first in the march section. Until that moment, we have heard only one
chord, albeit for eight measures.
Another reason to classify the chord change in measure 144 as “relatively
strong” is that it introduces—or actually reintroduces—a PC set that is held
through a considerable number of subsequent changes: the set consisting of A♭
and F. This assumes that the impact of a chord change increases with the per-
sistence of not only the eliminated PCs but also the new ones. (The degree of
that impact is thus established retrospectively.18) In this respect, too, the change
in measure 128 is weaker: a PC set of relatively great persistence is introduced
four measures later.19 By the same token, the chord change in measure 148 is of
lesser impact than the corresponding one in measure 132.
Thus, with the help of the family, we have discerned a misalignment between
phrase rhythm and harmonic accent in the march section of Fêtes. This can be
explained as a way to achieve gradual motion, as a remedy against sudden breaks
and abrupt shifts in the musical process. But it is also in keeping with the formal
function of this section. The pedal is traditionally a signal of anticipation, most
typically in the middle section of a small ternary form, or to the end of a sonata
17. I use the term “strong” in roughly the same sense as Arnold Schoenberg in his
Structural Functions of Harmony. According to Schoenberg, root progressions a fourth up
(e.g., V–I, I–IV, etc.) are strong, “because great changes in the constitution of the chord
are produced.” This is even more true of progressions like IV–V or V–VI. Hence, these
progressions are “superstrong” (Schoenberg 1969, 6ff).
18. From the phenomenological perspective presented by David Lewin (1986), the
event at the first beat of measure 144 is the object of various interrelated perceptions.
Each of these perceptions involves a specific context of which the event is part, for exam-
ple, measures 124–44 (perception: “sequential repetition”) or measures 132–44 (percep-
tion: “{G♯,B,D} eliminated”). And each perception provides itself a context for new events
to come. In these contexts, the event in measure 144 may be perceived in yet another
way. The chord change at the beginning of measure 148 denies the sequential repetition
noted there, insofar as this were to involve all of measures 124–39, but it places greater
value on the subset of PCs sustained from measure 144, i.e., {A♭,F}.
19. I mean {G♯,B,D}, but I realize that this is a questionable point. The chord change
in measure 128 does bring G♯ (alias A♭) and D together. Why should we not consider
this smaller PC set to be the progenitor of a family, since it is extended over more chord
changes than its superset {G♯,B,D}? My intuition is to give priority to the superset, unless
the number of chords that contain only the subset (i.e., {G♯,D}) is larger.

inclusion 199
development section. By slackening the pace of harmonic change, it delays the

moment of arrival, which consequently receives greater emphasis. A misalign-
ment of structural layers, such as described above, can reinforce this effect.
Metaphorically speaking, it can divide the music into different “streams,” which
create a strong sense of arrival when they finally converge. The hemiolas that
typically precede important cadences in eighteenth-century triple-meter dances
serve the same goal.
All of the above proceeds from the intuition that, if one, two, or more PCs are
shared by a number of chords in succession, each chord relates itself to them.
These PCs are not so much the center of gravity as the focus of contemplation,
more so as the number of chords increases. Such a relation can serve as an
instance of PC set inclusion.
It is important to note that our approach involves a reinterpretation of the
concept of a subset. It now refers, not to a distinct entity A within X, but to a
component of X. This is not a mere shift of emphasis. For example, we can hear
a tritone as included in a dominant seventh chord. However, this is not a sub-
set in the same sense as some of the three-tone configurations in Schoenberg’s
Piano Piece, Op. 11, no. 1. It receives no reinforcement whatsoever from
every, or any, other tritone. For convenience, we shall call this type of subset
“settled.” Subsets as they occur in Opus 11, no. 1, are then “nomadic,” because
they are defined in terms of properties that transcend the boundaries of any
particular superset.
Now the question is, with respect to subsets of the “settled” type: how can they
meet the requirement of a salient representation when their role is to make up
particular supersets, as in Stravinsky’s Agon? Since they can play this role only
collectively, as a family of subsets, what enables individual members of this fam-
ily to stand out from the rest? An important condition, as we shall see shortly,
is their membership in different families of subsets, or, equivalently, their being
progenitors of a family of supersets.
Plural Identities
It is not peculiar to PC set theory to define musical entities on the basis of their
constituent pitch elements and (or) abstract intervals; there are scalar or modal
theories that do the same. A scale may not represent a relational system, but
serve as a stock of tones (assuming octave equivalence) from which various com-
binations can be formed. Olivier Messiaen expressed this view when he wrote
about his “modes of limited transposition”:
[They] can be used melodically—and above all harmonically, melodies and

harmonies never going beyond the notes of the mode . . . They exist within
the spheres of influence of several keys at once, without polytonality—the

200 inclusion
composer being free to let one of the keys prevail, or to leave the tonal
impression undetermined.20
Example 6.10 is an excerpt from Le Banquet Céleste (1928), which Messi-

aen has analyzed himself in Technique de mon langage musical (Messiaen 1944,
96). The first chord {C♯,E♯,A♯,B} is a subset of the second transposition of
his octatonic Mode 2 (Pieter van den Toorn’s “Collection I”). At the same
time, this chord is highly suggestive of the key of F♯ major, in which it would
serve as a dominant thirteenth.21 Messiaen corroborates this suggestion by the
appearance of an actual F♯-major chord (with a non-modal E♯) on the down-
beat of the next measure, a chord marking the modulation from the second
to the first transposition of Mode 2 (“Collection III”). Inclusion is a crucial
concept here. Messiaen’s chords are subsets of his modal scales, but they can
form tonal patterns with other chords from the same scale, or with chords
from other scales. By their involvement in such patterns they achieve a certain
prominence. Their tonal affiliations cause them to stand out from their pre-
compositionally defined supersets.
The octatonic scale is noted for its capacity to accommodate various ele-
ments of tonal harmony: triads (major, minor, and diminished), seventh chords
(minor, dominant, and half-diminished), and French augmented sixth chords.
There are also more complex formations, such as the dominant thirteenth. Most
relations between these harmonic elements are not reminiscent of traditional
tonality. For example, the octatonic scale does not allow a dominant seventh
chord to resolve to its tonic. On the other hand, this chord can move to a domi-
nant seventh at another pitch-level. It is a consequence of the overall symmetry
of the octatonic scale that it contains a number of each available chord type. We
saw that symmetry at work earlier, in Stravinsky’s Symphonies of Wind Instruments
(chapter 4). The origin of the scale lies in sequential procedures of nineteenth-
century composers, particularly those procedures that employ one sequential
interval (PI 3, or −3) and preserve the interval content of the sequence model.
Taruskin’s account of the emergence of the octatonic scale stands as a classic of
musical scholarship (Taruskin 1985, 1996, 255–306).
Stravinsky’s “Petrushka chord” (ex. 6.11) is a subset of Collection III. It con-
tains two major triads with no common tones, or, looked at another way, two
dominant seventh chords with two common tones, their roots a tritone apart.
Before Arthur Berger’s discovery of the octatonic scale as a regulative force in
20. “[Ils] peuvent être utilisés mélodiquement—et surtout harmoniquement, mélo-

dies et harmonies ne sortant jamais des notes du mode . . . Ils sont dans l’atmosphère de
plusieurs tonalités à la fois, sans polytonalité—le compositeur étant libre de donner la pré-
dominance à l’une des tonalités, ou de laisser l’impression tonale flottante.” (Messiaen
1944, 85; my translation)
21. A chord owing its status of a “type” to the music of, among others, Schumann and
Chopin.

inclusion 201
(R:+ bourdon 16)

(chantant)
(clair)
(staccato bref, à la goutte d’eau)
Example 6.10. A tonal progression in Messiaen’s Mode 2, the octatonic. An exam-

ple from Le Banquet Céleste (1928) used by Messiaen in his Technique de mon langage
musical (1944).
By kind permission of Alphonse Leduc, owner and publisher for the worldwide.
the music of Stravinsky (Berger 1963), many took a chord like this as “polytonal.”
Today, this is a controversial designation, although the voicing in example 6.11a
does indeed articulate the two major triads, and although Stravinsky himself
confided that he “had conceived of the music [i.e., the music of the second tab-
leau of Petrushka] in two keys” (Stravinsky and Craft 1962, 156). Berger was right
to observe that an incomplete octatonic scale does not allow either key to pres-
ent itself in more than a fragmentary way. As he wrote:
[A] “polytonal” interpretation, insofar as it may have any validity at all, is even
more problematic than the determination of a single priority. For the “gapped”
scale affords far too little information for the delineation of “keys” of any kind.
(Berger 1963, 26)
However, such a “polytonal” interpretation can be meaningful in another

respect. The C-major and F♯-major triads in the Petrushka chord owe their
salience, not only to a particular musical articulation, but also, and maybe more
so, to their referentiality. “These triads—whose identity reflects past conventions
and traditions—are at one and the same time octatonic subsets and residual
emblematic reflections of common-practice tonality,” writes Kenneth Gloag (2003,
83). The latter role is reinforced by their relationship: C major and F♯ major are
maximally apart in the circle of fifths. It seems to be a statement on the condi-
tion of tonal music to blend them together as Stravinsky does. Notwithstanding
the inclusion of the Petrushka chord in an octatonic scale, its interpretation as
a synthesis of two opposing poles is entirely rational, given the weight and the
influence of the tonal tradition at the time of Petrushka’s composition. In turn,
this interpretation has added hugely to its salience as an octatonic subset. Berger’s

202 inclusion
a) 3
3
b)
3
3
Example 6.11. (a) The “Petrushka chord.” Igor Stravinsky, Petrushka (1911), from
the second tableau. (b) Octatonic derivation of the “Petrushka chord.”
© Copyright 1912 by Hawkes & Son (London) LTD. Revised version © Copyright 1948 by
Hawkes & Son (London) LTD. US Copyright renewed.
proposal to hear it as a “unitary sonic event” (Berger 1963, 22) would have been
pointless if it had not previously been categorized as a polychord.
What, we may ask, is the Petrushka chord? Or where is the Petrushka chord?
Can it be separated from the commentary it has elicited? This question invites a
general reflection on music theory and its relation to music practice, which will be
given space in chapter 7. For the moment, it suffices to say that an inclusion rela-
tion involves musical objects with plural identities. A subset of the octatonic scale
may also function as a tonal chord (Messiaen, Stravinsky); a segment of a melody
may reappear in the accompanying figuration of another melody (Fauré). A plu-
ral identity comes about by an object’s occurrence in different environments, most
typically when this object can be assimilated as an integral part of these environ-
ments—that is, when the environment comprises similar objects.
Two classic cases of a plural identity are pivot chords and temporary tonics.
Consider a family of chords that are diatonic to a given key. It is of no immedi-
ate concern here how many chords are in this family, or whether these chords
are frequent or rare; nor do we, at this point, distinguish between structural
and embellishing chords. What matters, for the present discussion, is that a
chord can gain salience in one key from its invoking another. However, it cannot
invoke a second key without the support of chords that are foreign to the first.
In example 6.12, the opening phrases of Bach’s chorale Nun lob, mein Seel, den
Herren, BWV 17/7, the first F♯-minor chord is a fully integrated member of the
family of chords in A major. The beginning of the second phrase brings to light
its affiliation to the F♯-minor family. This involves another chord, with which it
shares the property of being a member of this family, but which does not belong
to the family of A major (C♯ major, m. 5). It helps the F♯-minor chord stand out
from A major, almost like a “family member by marriage.”22
22. When it comes to instantiation of the property “Member of the F♯-minor Family
of Chords,” the help is mutual of course. However, the C♯-major chord does not have a
plural identity in this example.

inclusion 203


           
       
 
    
 
                    
           
F#m  C#M F#m
Example 6.12. Johann Sebastian Bach, Nun lob, mein Seel, den Herren, BWV 17/7.
The F♯-minor chord as a member of two families.
Bach’s chorale provides an example of what is generally known as “tonici-

zation.” Here, we describe it as an instance of family intersection, not because
this accounts better for tonal practice, but because an aspect of this practice
can be viewed through a set-theoretic lens—more specifically, through the
lens of PC set inclusion. Tonal chords are “subsets” of a key, that is, if we
define a key in terms of its constituent PCs. This definition is one-sided to be
sure, but it has been invoked in harmonic theories from early on, especially
in discussions of key relations. According to Abbé Georg Johann Vogler (Ton-
wissenschaft und Tonsetzkunst, 1776), closely related keys were those whose key
signatures differed by no more than one sharp or flat. Vogler thus looked at
the size of their intersection. Anton Reicha (Cours de Composition Musicale,
1816) observed that the tonic of a key serves as a diatonic triad in its closely
related keys. This statement reduces a chord to a set of pitches (or PCs) that
is included in various supersets. It is a set, not a chord, that is invariant under
a change of key.
Gottfried Weber (Versuch einer geordneten Theorie der Tonsetzkunst, Vol. 1, 1817)
actually used the word “family” to denote the sum total of diatonic triads and
seventh chords uniquely pertaining to a key.23 In his view, this family existed
prior to the scale, which he presented as the linear succession of the tones from
which the key’s triads and seventh chords were built (Weber 1817, 223). Like
Vogler and Reicha, Weber considered two keys more closely related as their
scales had more tones in common. However, relations between keys a fifth apart
were more important than others. Furthermore, Weber made an exception for
relations between parallel keys (e.g., C major and C minor), in that he added
them to the “first-degree” relations in spite of their relatively small intersection.
On his famous chart of key relationships (example 6.13; Weber 1917, 301) one
could measure the distance between any pair of keys. We will return to that chart
in the following section.
23. “Und so sind jeder Tonart nur gewissen Harmonieen angehörig, und diese, die
Familie der einer Tonart eigenthümlich angehörenden Harmonieen, nennen wir die
eigenthümlichen Harmonieen der Tonart.” (Weber 1817, 218; my italics)

Example 6.13. Gottfried Weber’s chart of key relationships (1817).
Nederlands Muziekinstituut, Den Haag. NMI 2 O 12.

inclusion 205
A later theorist for whom proximity was, among other things, a matter of shared
content was Arnold Schoenberg. In his Structural Functions of Harmony (posthumously
published 1954), Schoenberg presented a theory of “monotonality.” Instead of a sys-
tem of keys, this theory posited one tonality with different “regions.” These regions
were associated with consonant harmonies at a greater or lesser distance from the
tonic. One of the measures of this distance was the number of tones (PCs) a region
had in common with the tonic region. For example, regions in a “direct and close”
relationship had five or six tones in common, while an “indirect but close” relation-
ship was indicated by three or four tones being in common.
Schoenberg’s “classification of relationship” (Schoenberg 1969, 68) is not
consistent, due to his using a handful of criteria for the relative proximity
between regions (apart from a minimal intersection of three tones, he included
a fifth relation, a common dominant, the close proximity to a closely proximate
region, or a structural analogy); and this handful does not even include the
reason why the “Dorian” region (D minor with respect to C major) is ranked
as more distant from the tonic than the “minor mediant major” (E♭ major).24
However, common pitch content was important enough for Schoenberg to men-
tion as one basis of tonal relationships.
We have seen how family intersection can create the conditions for inclusion
relations to become musically manifest. To summarize: a member A of Family X
may appear to be also a member of Family Y when it is joined by a non-member
of X with which it shares the property “member of Y.” The inclusion relation
between A and Y may lead to a heightened salience of A in Family X. Such A are
the chord {C♯,E♯,A♯,B} in Messiaen’s Le Banquet Céleste (see ex. 6.10) and the
second F♯-minor chord in Bach’s setting of Nun lob, mein Seel, den Herren (see ex.
6.12). Now, the following is important to note. If we conceive of these chords as
plural identities, we abstract certain contextual properties away from them. In
that moment of abstract reflection, they become PC sets: blank, featureless iden-
tities, uninterpreted entities, members of no family. In reality, however, their sur-
roundings “tug and pull” at them, in Lerdahl’s words. It is this tug and pull that
can bring them into prominence.
In serial twelve-tone music, unordered PC sets emerge in a similar way: as a
neutral body of objects that are made prominent as they are mapped onto dif-
ferent series. In chapter 4, such relations were explained under the principle of
combinatoriality, a principle that ensured the conditions for a twelve-tone series
to retain particular PC sets under particular transformations.
24. Carl Schachter (1987, 316, note 22) already made this observation. Schoenberg’s
reason may have been that the fundamental-bass progression from C to D was “super-
strong” and hence disruptive (see note 17). The defectiveness of such progressions—or,
equivalently, the need to normalize them by an intermediate fundamental—was a tenet
of Viennese fundamental-bass theory, of which Schoenberg the harmony teacher was a
descendant. For an authoritative account of this tradition, see Wason 1985.

206 inclusion
The way in which Allen Forte introduces the inclusion relation in The Struc-
ture of Atonal Music reminds one of the studies of combinatoriality by Milton Bab-
bitt, David Lewin, and others. When confronting the question of how to assess
the relative importance of all the subsets of a PC set, he points to “the notion of
invariance.” Forte defines this as the intersection of transformationally related
PC sets:
Given equivalent pc sets A and B, a third set C is determined by the elements

that are both in A and B. This set is called the intersection of A and B. . . . The
elements of such a set will be called invariant pitch-classes. (Forte 1973, 30)
Recall that, in Forte’s view, equivalent PC sets were related as transpositions

or as transposed inversions. “Under these transformations some pc sets may
hold certain pitch-classes fixed, while other pcs will change,” he writes else-
where (Forte 1978, 9). Invariance provided him with a rationale for harmonic
progression, just as it provided Babbitt with a rationale for the succession of
twelve-tone series:
The concept of invariance is intimately bound up with the intuitive musical

notions of development, change, continuity, and discontinuity. (Forte 1973, 31)
Interestingly, these “intuitive musical notions” do not require PC sets with invari-
ants to be related transpositionally and/or inversionally. This arbitrary restric-
tion betrays the influence of twelve-tone serialism. It enables the “prediction”
of the number of PCs held over from one set to the other.25 As a corollary, it
helped Forte determine the conditions for “minimum,” “maximum,” and even
“complete invariance.” Thus, in the first chapters of The Structure of Atonal Music
his focus is on settled subsets. He considered these of structural importance
when he could find a larger representation of their supersets in the music; the
conjunction of these supersets was their primary means of instantiation.26 One
could say that Forte conceived of the property “⊇ A” as defining a family of
PC sets, and of A itself as the progenitor of this family. However, the notion of
25. See chapter 3, note 21. For any element a in a PC set A to be also included in
Tn(A), A should include b = (a − n) mod 12. Generally speaking, then, the intersection
between A and Tn(A) is as large as the frequency of the PIC n in A—which can be read
from the PIC vector. For any a in a PC set A to be also included in TnI(A), it must sum
to n with itself, or with another element of A. Therefore, the intersection between A and
TnI(A) is as large as the number of distinct PC values in either set that sum to n with
another PC value in the same set. These are but two of a number of “common-tone theo-
rems.” For a listing of these, see, for example, Rahn 1980a, 97–123.
26. Forte had not addressed the problem of a subset’s instantiation in “A Theory
of Set-Complexes for Music.” This paper covered inclusion relations only insofar as they
were relevant for an understanding of the PC set complex.

inclusion 207
“family” is latent in this introduction to PC set inclusion. It enters the discussion

later, in the second part of Forte’s book, where it effects a major change in the
notion of “subset.”
Study in K(h)
As noted before, about one half of The Structure of Atonal Music is devoted to an
elaboration of the theory of set complexes, Forte’s grand design of nine years
earlier. We shall now discuss this theory with reference to both the book and
the 1964 article. The changes that Forte made to it did not affect its essential
features, which we will be discussing here.
A set complex consists of a number of families of subsets and supersets. Con-
sider two PC sets that complement the twelve-tone chromatic. We shall label
them X and X '. In example 6.14, their elements are unspecified. Black note-
heads indicate PCs in X; white note-heads indicate PCs in X '. Basically, to the set
complex about X and X ' belongs any PC set A that satisfies one of four statements:
1.1 A ⊇ X
1.2 A ⊆ X
2.1 A ⊇ X '
2.2 A ⊆ X '
In any of these cases, A is in a set-complex relation to X and X '. From this it fol-
lows that a set complex is the union of four families: the family of subsets and
the family of supersets of a given set; and the family of subsets and the family of
supersets of the given set’s complement. It should be noted that the above four
statements reflect Forte’s 1964 presentation of the PC set complex. In 1973, he
assumed subsets and supersets to be “proper” (see chapter 2, p. 45), excluding
an A that is equal to X or X '. He even stipulated the exclusion of all A of the car-
dinal numbers of X and X '.27 Example 6.14 is in agreement with both definitions
of the set complex. (Hence the substitution of “⊂” and “⊃” for “⊆” and “⊇.”)
In set-complex theory, labels like X, X ', and A not only denote single PC sets,
but also all transpositions of these sets, and all transpositions of their inversions
as well. In other words, the set-complex relation is a relation between classes of PC
27. See Forte 1973: 93–95. This does not follow inevitably from the four relations
that define the membership of a PC set complex. When A is a proper subset or a proper
superset of X, it may have the same cardinal number as X '. (It may even be equal to X '.)
This is logically impossible only when the cardinal number of X is 6.

208 inclusion
Set X Set X'
1.1 A⊃X
1.2 A⊂X
2.1 A ⊃ X'
2.2 A ⊂ X'
(1.2 + 2.2)
T/ Tn I
Example 6.14. Set-complex relations. In all cases A is member of the set complex
about X and X '. Below: a subcomplex relation.
sets. As a set-class, A may satisfy two of our four statements. For example, one PC
set of class A may be included in a PC set of class X, while another includes, or is
included in, a PC set of class X '. A then belongs to a more exclusive complex, the
so-called subcomplex about X and X '. Every member of this subcomplex satisfies a
statement of number 1 (1.1 or 1.2) and a statement of number 2 (2.1 or 2.2).
The set-complex and subcomplex relations are symmetrical, that is, they
extend from A to X/X ' and from X to A/A'. This can be shown by replacing each
of the above four statements by an equivalent statement:
1.1 X ⊆ A
1.2 X ⊇ A
2.1 X ⊇ A '
2.2 X ⊆ A '
The statements of number 1 have simply been reversed. Those of number 2,

however, have traveled a longer way, under the license of what Forte called the
“natural association of inclusion and complementation” (Forte 1973, 94). This

inclusion 209
is what he meant: the complement of any subset of any set A is a superset of the
complement of A. Thus, from the reversal of the original statement 2.1 (“X ' ⊆
A”) we can deduce that X ⊇ A'. Similarly, from the reversal of the original state-
ment 2.2 (“X ' ⊇ A”) it follows that X ⊆ A'. It was this property of the inclusion
relation that led Forte to the development of the PC set complex:
By property 3 it is possible to construct about each of the distinct sets [that is,
A, A', X, or X'] and its complement a symmetrical array of sets to which they
are in the inclusion relation. Such an array will be called a set-complex, and
the complementary pair about which the complex is arranged will be called
the reference-pair. (Forte 1964, 162)
Example 6.15 shows the (“total”) set complex about the reference-pair con-
sisting of “5-33 and “7-33” (above and middle).28 It brings out very clearly the
symmetry between the family of supersets of “5-33” and the family of subsets of
“7-33” by virtue of the assignment of the same ordinal numbers to complemen-
tary set-classes. The smaller families of subsets of “5-3” and supersets of “7-33”
are shown to be equally symmetrical; Forte’s “property 3” ensures that for each
subset of “5-33” (e.g., “4-24”) there is a superset of “7-33” (in this case, “8-24”)
Furthermore, all set-classes that bear an inclusion relation to both members of
the reference-pair—and that thus appear in two of the four families instead of
in one—are entitled to the membership of the (“most significant”) subcomplex
about “5-33/7-33” (ex. 6.15, bottom).
Example 6.15 dates from 1964. What would have happened to it if Forte had
included it in The Structure of Atonal Music nine years later? He would have left
out all set-classes of cardinal numbers 2 and 10, which by then were no longer
represented on his list of prime forms. Furthermore, all set-classes of cardinal
numbers 5 and 7, including the duplicates of “5-33” and “7-33,” would have dis-
appeared from the chart. However, the PC set complex did not only develop
with respect to the number of restrictions on its membership. A striking feature
of its 1973 presentation was the emphasis on analytical procedure. This empha-
sis was key to resolving an issue of vital importance: in what practical way could
the PC set complex serve as the “framework for the description, interpretation,
and explanation of any atonal work,” as Forte hoped?
Once an analysis had yielded a number of important set-classes, the suggested
procedure could not be more straightforward. The first step was to check each
pair of these set-classes for a set-complex relation (labeled “K”) or subcomplex
relation (labeled “Kh”). Each member of such a pair should then belong to the
28. The set-class with Forte name “5-33” is that of the whole-tone pentachord. The
prime forms of both referential set-classes are contained in examples 4.2 and 4.8. Note
that Forte uses “K” and “Ks” as labels for “set complex” and “subcomplex.” (He would
replace “Ks” by “Kh” later.) “N” and “M” are variables, like our A and X.

Example 6.15. The set complex (above) and subcomplex (below) about “5-33/7-33.” An
illustration from Allen Forte’s article “A Theory of Set-Complexes for Music,”Journal
of Music Theory 8/2 (1964): 169.
Used by permission of Duke University Press and Yale University.

inclusion 211
(a)
Set-classes
Mm: PC sets: (Forte):
1: C C♯ D♯ E 4–3
2–4: D E♭ E F B 5–4
4–5: C F G♭ B♭ B 5–7
2–5: 8–5
6–7: F# G A♭ A B♭ 5–1
8: E♭ E F 3–1
9–10: C C♯ D G♭ B 5–5
11: G♯ A
12: G A♭
11–12: 3–1
13: C♯ D E♭ E B 5–2
(b)
3–1
4–3 K
8–5 Kh 4–3 8–5
5–1 Kh Kh K
5–2 Kh K K
5–4 Kh K Kh
5–5 Kh K Kh
5–7 Kh Kh
Example 6.16. (a) PC sets of the main segments of Anton Webern’s Bagatelle for
String Quartet, Op. 9, no. 5. (b) A set-complex analysis of the Bagatelle
set complex or subcomplex about the other. The second step was to group the
results in a matrix. A very small example is shown here as example 6.16b. It
depicts the relations between the set-classes represented by the principal seg-
ments of Webern’s Bagatelle, Op. 9, no. 5 for String Quartet. (Table 6.16a. shows
the PC sets of these segments.) Such a matrix was a basis for assertions about
different types and degrees of coherence. One would note the relative density
of set-complex relations in Webern’s Bagatelle, and especially the involvement

212 inclusion
of “3-1” in subcomplex relations with all but one of the other set-classes. This
involvement reflects the tendency toward a clustering of minor seconds in this
piece.29 In another work, the set-classes might all be members of the set com-
plex about one of them, but not have many mutual relations of this kind. Or a
work might not appear to be particularly coherent in this sense at all.
Let us dwell for a moment on this procedure. When carrying it out, what is
our purpose? Suppose we have selected a number of tone combinations in a
work for further investigation, on the basis of the recurrence of their set-classes,
or on account of their importance in other respects. So far, the only thing we
have is a set of guided observations, like “A and B have the same PC set,” or “A
and B have the same similarity index as C and D.” There is a fair chance that
these will not connect to reveal a larger pattern of relationships. Yet it is the
possibility of such a pattern that motivates our procedure. We want to ascertain
whether larger structural entities can be built from smaller ones, so as to widen
our analytical scope. The interaction of these larger bodies of musical material
may help us explain whole works or (most often) sections thereof.30
One aspect of set-complex theory is especially intriguing: a small entity can
represent a larger one. More specifically, a PC set can stand for a set complex or
subcomplex, as the “head of a family” (to quote Gottfried Weber). This means
that it is a potential source of coherence over a larger span of music. For a PC set
to realize this potential, it should be in a set-complex relation with all (or most)
other set-classes represented in that span. It is then called a nexus set. On its first
appearance, in The Structure of Atonal Music, this term seems to serve as a mere
substitute for the older “reference-pair,” but later on Forte used it to refer to a
PC set that exerts a dominance over concurrent sets on account of the number
of its set-complex relations (Forte 1973, 101, cf. 113–14).
The above will sound familiar to the reader who is well versed in the history
of music theory pedagogy and in traditional methods of music analysis. PC set
(complex) theory aims to proceed smoothly from elements to complex forms.
29. Set-class “3-1” is the class of chromatic trichords. Its prime form is (0,1,2). The
set-complex analysis does not record the important tendency towards equal distribution
of PCs in Webern’s Bagatelle. This is what the arrangement of PC sets in example 6.16a
reveals. The same example also shows the instantiation of {E♭/D♯,E} as a subset of four
PC sets.
30. Examples of the application of set-complex theory to larger works include Forte’s
own analysis of Stravinsky’s The Rite of Spring (Forte 1978) and Janet Schmalfeldt’s study of
Alban Berg’s Wozzeck (Schmalfeldt 1983). Forte traces the harmonic materials in The Rite
of Spring to the set complexes about three “prominent” set-classes of cardinal number 7
and four set-classes of cardinal number 8. Schmalfeldt, in a more hermeneutic approach,
assigns the “significant” components of the harmonic language of Berg’s opera to three
set complexes, which in her view support the main psychological threads in the drama.
The quotation marks in the foregoing sentences are not intended to cast doubt on the
judgment of these authors; they should remind the reader of the critical role of the deci-
sions preceding the set-complex analysis.

inclusion 213
This smoothness is important in theories of harmony and musical form, which

provide ways to assure a clear connection between the small detail and the
larger whole of a composition. Harmonic functions can be attributed to chords,
sequences of chords, phrases, and sections. Similarly, certain formal patterns
underlie themes as well as movements of compositions.
Can the nexus set be compared with the tonic triad, as another instance of
a concept that provides a link between different structural levels? And, conse-
quently, does the PC set complex owe something to the concept of a key? In the
previous section, we referred to Gottfried Weber’s definition of a key as a “family
of chords.” It would probably not be right to regard this expression as an anticipa-
tion of twentieth-century musical thought. More likely, PC-set (complex) theory
has expanded on elements of earlier music-theoretical vocabularies. Therefore,
it is less instructive to look for vestigial remnants of the tonal system in the PC
set complex than to study the tonal system—the system of keys and key relation-
ships—from the perspective of set-complex theory. However contrived such an
exercise may seem, it enables the concepts of key and complex to shed light on
each other without falsely suggesting an organic development between the two.
Let us take another look, then, at Weber’s chart of key relationships (ex. 6.13).
Each key is represented by a letter name. We can also take this letter as referring
to that key’s tonic chord. Thus, we read the chart as a topographical survey of
all major and minor chords, each of which serves as the tonic of a key, and is
surrounded—or at least one-half surrounded—by the consonant triads belong-
ing to that key. This reading matches with the eighteenth-century intuition that
keys are closely related when the tonic chord of one exists in the other.31 If we
were to define an “in-key relation” between two consonant triads—requiring one
triad to be a diatonic member of the key of the other—this relation would, of
course, be symmetrical, like the set-complex relation. A complication, however,
is the presence of the major dominant triad in a minor key family, which is more
natural (or frequent) than the presence of the minor key’s tonic in the family
of its major dominant. To maintain the symmetry, we would have to include the
minor subdominant as a member of a major key family (ex. 6.17).32
Another complication is the almost complete intersection of the families of
a major key and its relative minor. This may lead to a wrong assignment of tonic
function (that is, if we equate “tonic” and “nexus set”). For example, if an analy-
sis of Bach’s chorale setting in example 6.12 would follow the same lines as that
31. Earlier on, I quoted Reicha on this subject. For more, see Lester 1992, 214–17.
32. A rule of correspondence between these two harmonies is provided by harmonic
dualism. According to Oettingen (1913) one changed the mode of the Unterdominante
(i.e., subdominant) in major and its mirror image, the Oberregnante (i.e., dominant) in
minor so as to reinforce the relation with their tonic and phonic Klänge (“klangs”). These
transformations added a leading tone for the fundamental of the major triad (C) and
its structural equivalent, the “phonic overtone” of the minor triad (G), respectively. See
further below.

214 inclusion
(a)
f F d D b B g♯
b♭ B♭ g G e E c♯
e♭ E♭ c C a A f♯
g♯ A♭ f F d D b
c♯ D♭ b♭ B♭ g G e
(b)
f F d D b B g♯
b♭ B♭ g G e E c♯
e♭ E♭ c C a A f♯
g♯ A♭ f F d D b
c♯ D♭ b♭ B♭ g G e
Example 6.17.
(a) The family of chords in C major, projected on Weber’s chart of key relationships
(b) The family of chords in C minor
of Webern’s Bagatelle Op. 9, no. 5 in example 6.16b—using an “in-key relation”

instead of a “set-complex” relation—the F♯-minor chord would turn out to be
the “nexus set” instead of the tonic A major. The only chord in this excerpt
that does not belong to the A-major family (C♯ major) is still a member of the
F♯-minor family. Therefore, F♯ minor has the most in-key relations among the
chords in example 6.12.
To prepare the next step in this comparative study, we should note once
more that relations between keys exist in various degrees. Some keys are closely
related, others more remotely so. Furthermore, in the context of a musical com-
position, one key may be secondary to another. For this to be seen in the light
of set-complex theory, it seems proper to use the notion of transitivity. Consider
the family of subsets of X. As will be recalled, if such a family includes A, it also
includes the family of subsets of A (see ex. 6.8a). In turn, each subset of A has
a similar family, which is included by the families of subsets of A and X, respec-
tively. Now, in a sense, we “move away” from X as we traverse subsets of suc-
cessively lower cardinal numbers, and as the number of families that exercise a
claim on these subsets increases. This may give us a sense of “distance.”
It takes some mental effort to speak about key relations in terms of families
of subsets and supersets—let alone in terms of subsets or supersets that involve
their own families—but Arnold Schoenberg has given us more than a hint of
how to achieve this. Schoenberg’s theory of monotonality presents us with a

inclusion 215
great expansion of Weber’s “family of chords.” Each member of this family now
supports its own family, under the central governance of the tonic. For example,
F major is a member of the family of chords of C major; and B♭ major is a mem-
ber of the family of chords of F major. Seen thus, B♭ major is also a member of
the C major family, albeit a more remote one than F major.
However, distance is not the only measure of relationship within a family. The
inclusion relation is transitive, but a family relation—like a set-complex relation,
or our in-key relation—is not. Given that A ⊆ X, there are members of the fam-
ily of subsets of X that do not belong to the family of subsets of A. This means
that a family of subsets (or supersets, for that matter) has subdivisions, each of
which intersects with other families. Apart from a sense of distance, then, family
relations provide a sense of direction, depending on the clarity of these subdivi-
sions. Although Schoenberg’s theory employs geographical imagery (“regions”),
its tendency toward a unification of tonal space may somewhat undermine our
orientation. Obviously, most chords from the region of the “Mediant Major”
(in C major: the family of chords of E major) do not belong to the region of
the Submediant Minor (in C major: the family of chords of A minor); but since
these regions are both integral to the territory of one tonic, one may seriously
ask whether there still exists a clear line of demarcation between these regions.
In Schoenberg’s tonal space an in-key relation may yet be transitive.
Finally, the involvement of complement-related families in one set com-
plex seems to have no parallel in tonal theory. Yet the picture changes when
we express the relation between such families in a more general way—as two
non-equal families with progenitors X and X ', which are arranged symmetrically
about an axis (see ex. 6.15). For each member A of the family of X there is a
unique member A' of the family of X ', such that the relation from A to X extends
in reverse from A' to X '. The members of both families are thus related by inver-
sion (as defined in chapter 3).
This holds true, not only for the subsets of X and the supersets of X ' in a set
complex, but also for the members of the families of major and minor keys in
the context of nineteenth-century harmonic dualism. In that context, the minor
triad was the mirror image of the major triad—a chord composed of the same
intervals measured down from the fifth. Proceeding from this inverse relation,
theorists like Arthur Joachim von Oettingen (Harmoniesystem in dualer Entwick-
lung, 1866) and Hugo Riemann (e.g., Skizze einer neuen Methode der Harmonielehre,
1880) constructed the minor tonality as a system of harmonic progressions in
contrary motion to those of the major tonality. This placed the subdominant of
the minor family in exact opposition to the dominant of the major family, and
likewise the subdominant of major in opposition to the dominant of minor. In
example 6.18, the relations between the major and minor families are depicted
so as to convey the parallels with the set complex in example 6.15. By show-
ing two pairs of families, this example indicates that the choice of the “progeni-
tors”—i.e., the actual tonics—was arbitrary. Whereas Oettingen contrasted the

216 inclusion
(a) Oettingen:
E♭ A♭ D♭
c f b♭
F C G
d a e
(b) Riemann:
G C F
e a d
F C G
d a e
Examples 6.18. Harmonic dualism: the symmetrical alignment of major and minor
keys (cf. ex. 6.15).
major and minor tonalities of the same Hauptton (C, as the fundamental of
C major—or “c+”—and the “phonic overtone” of F minor—or “co”), Riemann
chose to stay within the collection of natural tones, setting off C major (“c+”)
against A minor (“oe”).
The structure of the set complex is certainly reminiscent of harmonic dual-
ism; however, the meaning of the resemblance is unclear. Although harmonic
dualism allowed modal mixture to occur (see note 32), it did not, of course,
envisage a merger of the major and minor tonalities into one family—like the
merger of four families of subsets and supersets into one set complex. In fact,
all parallels observed between set-complex theory and the key system—as inter-
preted by theorists of different generations—exist exclusively on a conceptual
plane. They tell us little about the musical nature of the set-complex itself. In
one essential respect, the set complex is even at odds with notions of tonality:
its constituents are set-classes, not pitch-class sets; set-complex theory deals with
nomadic subsets and supersets, not settled ones. One may conceive of the set-com-
plex as a family of pitch materials; but these pitch materials owe their structural
identities to relations other than their inclusion in that family.
However, there are two reasons to take note of the abstract similarities
between set-complex theory and the tonal system. One is the later development
of PC set genera as tightly knit families with more distinct harmonic profiles
than the set complex. This bespeaks the ambition to imbue PC set theory with
a sense of tonality-like hierarchy. Forte’s PC-set genus of 1988 consists only of a
family of supersets. And from this family he omitted some “black sheep”—most
significantly the members whose complements do not include the three-tone
progenitor (Forte 1988a, 192; see also note 13). In the following year, Richard

inclusion 217
Parks presented a genus that combines two families: the subsets and supersets
of what he called a “cynosural set.” This role was assigned to the diatonic (Forte
name “7-35”), whole-tone (“6-35”), and octatonic (“8-28”) scales among others.
Parks selected these scales not only for their historical significance, or for the
role they play in the music of Debussy (the object of his study), but also for their
symmetrical properties. Because of these, they accommodate all subsets under
transposition alone, making it possible, as Parks writes, “to associate the sound
of a given set to its genus” (Parks 1989, 58).
The other reason to pinpoint elements of tonal thought in set-complex the-
ory is that these may help achieve continuity with the music-theoretical past—
not some peripheral movement, but the grand tradition. Rather than relegating
such elements to the category of the “incidental,” or of the “outer form,” we can
talk about them in terms of their effect on the interpretation of music. We thus
regard them as integral to the theory from a performative perspective. The fol-
lowing chapter will develop such a perspective.

Chapter Seven
“Blurring the Boundaries”

Analysis, Performance, and History
Music theory, it was argued in chapter 1, can be a discipline at the service of
music historians, providing them with the conceptual apparatus to analyze and
assess music from the past. But music theory is also a part of music history, pre-
senting itself for analysis and assessment. This double identity can be viewed as
a special case of a problem fundamental to historiography: the problem of a his-
torian’s own historicity, which renders an objective truth outside history inher-
ently impossible. Historians trained in modern hermeneutics know about this.
They know that their reading of a source is at best a mediation, a give-and-take
between two historical viewpoints, and that as such it is temporally conditioned.1
This explains why history is, on a fundamental level, at odds with science. For
the result of a scientific experiment to be declared valid, it should not be tempo-
rally conditioned, but, on the contrary, be obtainable under all circumstances.
It should make no difference at which time, in which location, and by whom the
experiment is carried out. This requirement of repeatability goes against the
very nature of historical research.
What Music Analysts Do
What music analysis has in common with the historian’s account of past events
is its historicity. The analysis results from a mediation between an observer and
an object, and between the worlds they represent. It tells us something about the
music and its historical background, while at the same time reflecting the craft,
the experience, the interests, and the expectations of the analyst. Even a hun-
dred music students supplying the harmonies of a four-voice chorale setting by
1. This renders synoptically what Hans-Georg Gadamer, in Wahrheit und Methode

(1960), called a “fusion of horizons” (Horizont-verschmelzung). A “horizon” delineates the
world in which a person operates, and the means by which he understands that world. In
Gadamer’s view, to understand the other does not mean to empathize with him fully, but
rather to integrate one’s horizon with his. Thus, in historical writing, one integrates one’s
horizon with the historical horizon.

“blurring the boundaries” 219
Johann Sebastian Bach with exactly the same function labels do not alter the fact
that such an analysis represents a historically determined state of affairs: har-
monic functions having become a special focus of music theory; the professional
education of musicians having been assigned to institutions and coordinated in
training programs; composition, performance, and analysis having drifted apart
as musical disciplines; and works from different periods and of different styles
having been integrated into a “classical” canon. In any case, this state of affairs
was largely unknown to Bach. This example shows how complex analyzing music
can be from a historiographical point of view. It may not only involve the histo-
ricity of an observer and an object, but also the historicity of the methods used.
People who care about the scholarly status of music theory and analysis, who
are sensitive to the criticism that these disciplines lack historical perspective
and uphold the “fiction” of self-contained musical structures, may argue that
this “fiction” results from conditions also delimiting the writing of history, or
indeed any process of signification, like the reading of a poem, the performance
of a symphony, or the staging of a play. In other words, they may not be willing
to suppress their interest in an aspect of music only because it was presumably
unimportant to the composer and his contemporaries. Thomas Christensen
(1993b), for example, wanted to safeguard the music analyst from a perma-
nent obligation to represent musical works by means of contemporary modes
of understanding (without giving up the obligation to be aware of these modes
of understanding).2 He argued against Joseph Kerman, Gary Tomlinson, and
Leo Treitler, who had been critical of music analysis in its role of an academic
specialization. But the scholar whose position especially bothered him was Rich-
ard Taruskin. Like other theorists, Christensen could not reconcile Taruskin’s
notorious campaign against the orthodox historicism of “authentic” perfor-
mance practice with his repudiation of analytical tools not contemporary with
the object analyzed. Taruskin had played the card of historical correctness with
characteristic eloquence in his debates with music theorists such as Allen Forte
(Taruskin 1979, 1986; see further below), Matthew Brown, and Douglas Demp-
ster (Taruskin 1989, 1995):
What one hopes to achieve by means of analysis is not merely a taxonomy of

musical configurations, but insight into praxis: methods, routines, devices of
composition. Since one cannot claim to understand a praxis unless one can
state the theoretical basis on which it rests, one is going to be interested in
historical viewpoints as well as any more efficient ones we may now propose.
(Taruskin 1989, 157)
As for music analysis, if it is not to be merely another reinforcement of tran-

scendentalist myths—if, that is, it is going to tell us something we don’t already
“know”—it should provide us with information (yes, historical information)
2. Christensen referred to Gadamer’s model of interpretation (see note 1).

220 “blurring the boundaries”
about composing techniques . . . If we are not interested in learning that (and

many have by now so declared themselves) then analysis has nothing to tell us
but comforting bedtime stories, for which just as many retain a seemingly insa-
tiable appetite. (Taruskin 1995, 24)
Taruskin thus denied the music analyst the freedom of interpretation he granted
to the performer. The latter’s primary task was to please and move “an audience
in the here and now” (1995, 23). This view of the performer’s mission had already
triggered a polemic from the champions of “authenticity.” Performers who prided
themselves on their historical research were given to understand that:
Research alone has never given, and is never likely to give . . ., enough infor-
mation to achieve that wholeness of conception and that sureness of style—in
a word, that fearlessness—any authentic, which is to say authoritative, perfor-
mance must embody. (Taruskin 1982, 343)
Christensen pointed out the difference in treatment of performers and analysts

in Taruskin’s writings. And he tried to draw Taruskin’s analyst into the firing line
of his own criticism:
Why, . . . one wonders, is Taruskin not led to conclude that historical contextu-
ality in music analysis is but another form of “authenticity”? The search for his-
torically indigenous modes of analytical interpretation, after all, is based upon
an “authentistic” aesthetic just as historical performance practice is. (Chris-
tensen 1993b, 23)
The answer to Christensen’s question is quite simple: in Taruskin’s view, schol-

ars, more specifically analysts, had no choice. They did the right thing, following
the “authentic” course. It was their duty to search for “historically indigenous
modes of analytical interpretation.” They were not in the position to communi-
cate an artistic experience. Taruskin thus drew a line between musical practice
and the scholarly study of music, which he considered subject to different rules
of conduct vis-à-vis the musical work. Music analysis was consigned to the side of
scholarly study; its goal was to yield knowledge.
However, one cannot play down the role of signification in the analytical liter-
ature. In the first decades of the twentieth century, the analyses of Ernst Kurth,
Heinrich Schenker, and Arnold Schoenberg were intended to stress certain val-
ues that music—the best music, in the opinion of these authors—embodied.
Kurth regarded music as a direct expression of the inner forces of human life.
Schenker devised a concept of organic unity which only the works of a group
of great—and, coincidentally, for the most part German—masters could illus-
trate. And for Schoenberg, aesthetic value came from an immutable substance
pervading an ever-changing musical surface. These men did not analyze with a
view to stating knowable facts about a musical composition, but to highlight it as

a work of art, to celebrate its wholeness and inner logic; in a way, they acted as
performers. Analysis can be seen as a mode of representation alongside concert
performance. Like the latter, it reinforces the image of a self-contained musical
whole—of Opusmusik, as Hans-Heinrich Eggebrecht (1975) called it—in an envi-
ronment in which it stands out in bold relief.3
Can this view also be applied to current analytical practices? Both Schenker
and Schoenberg have left their marks, while Kurth, on the other hand, has not
had a lasting influence. Schenker’s method—gradually stripping down a musi-
cal text to a basic progression, the Ursatz—is still being used today. As noted
in chapter 1, his name has become attached to a dominant analytical tradition
in the United States, which has spread to Great Britain, Australia, and Canada.
Schenkerian analysis has changed in the academic environment that adopted
it. American music theorists have downplayed Schenker’s unedifying polemic,
and his exclusive commitment to an “aesthetically superior” repertoire. They
have decoupled his method from his “unfortunate excursions into the realm of
the political, social, and mystical” (Babbitt 1952, 264). And they have taken the
Ursatz-concept—with its Romantic connotations, and the uniformity it imposes
on the deep levels of musical structures—for granted, stressing the value of the
analytical process at the expense of its final outcome.4
William Rothstein (1990) has documented the method’s assimilation to the
norms of scholarly discourse maintained at American universities. And Robert
Snarrenberg (1994) has reported the transition from a “natural” to a “scientific”
imagery reflecting this assimilation in the literature, a transition that has caused
people to associate Schenkerian analysis with academic dryness and formalism.
This association may be justified in a number of cases, but it does not apply to
everyone, and not to Schenkerian analysis in itself. Notwithstanding its transfor-
mation, it is still essentially a practice meant to elevate musical works of art. A
Schenkerian analysis is the written, graphed-out, or spoken counterpart of the
concert performance, from which one should not expect historical information,
but an artistic interpretation.
Schoenberg’s influence on twentieth-century analytical discourse has been
no less strong, if a little more diffuse than Schenker’s. It has been felt in the field
of classical form—due to his posthumously published Fundamentals of Musical
Composition (1967)—in analyses focusing on large-scale motivic coherence—due
3. One may think of other environments that do not allow musical works to be per-
ceived as tightly organized statements. And, of course, there is music to which the concept
of a “work of art” is not relevant at all. See Cook 1989a (10–70) and Stockfelt 1993 for some
reflections on the interdependence of musical structure, space, and mode of listening.
4. This line of defense was taken by Pieter van den Toorn: “[W]hat counts are the
musical issues that are brought to the fore, not the product itself, its ‘solution’ of a musi-
cal puzzle, or the fact that alternative readings may suggest themselves as equally valid.
The latter are, indeed, encouraged as an essential part of the process.” (Van den Toorn
1995, 98)

to the motivic analyses of his own music, and of earlier music—and, more gener-
ally, in the very close alliance between theory and contemporary composition
in the years after the Second World War, an alliance to which PC set theory ulti-
mately owes its existence. Schoenberg had fewer sides to screen from the public
gaze and from scholarly attention than Schenker. His uneasy modernism—which
seems to have grown organically from a deeply felt commitment to the musical
tradition—was more agreeable than Schenker’s forthright conservatism. Schoen-
berg’s impact as a composer testified to the relevance of his analytical observa-
tions, although his analyses may actually have been meant as justifications of his
own music.5 In any case, the message communicated by these analyses was that
a full appreciation of the music required total immersion in the score. By point-
ing out the intricacy of a movement by Brahms or Beethoven, by shedding light
on the recurring intervallic configurations unifying it on a deep structural level,
Schoenberg argued the necessity of a textual interpretation.6
Taking the score as representing the work, and searching it for unifying
elements, has been common practice in academia for years. PC set theory, for
example, has extended Schoenberg’s viewpoint that the outer appearance of
things may conceal their sameness, and that it is their sameness that matters,
even when it is not obvious at a first hearing. PC set theory considers a PC set
as being the essence of a chord, a melody, or a motif, or any other musical entity
involving traditional pitches, so that when two such entities have their PC sets
in common—or when their PC sets are of the same equivalence class—they are
equated, by supplying them with identical labels. When the PC sets are not equiv-
alent, one may measure the degree by which their relationship is close or dis-
tant (the common ground among the various similarity measures). A difference
between PC sets, however, is valued as such only when it comes under the head-
ing of PC set complementation (a category still involving intervallic similarity).
PC set theory can thus be seen as dealing with a concept of unity based on
recurrence and variation. Like Schenker’s theory of tonal unity (a concept based
on prolongation), it has transformed an aesthetic creed into a set of analytical
tools. Not surprisingly, then, it has met with the same kind of criticism that was
meted out to the Schenkerians. It was considered to be insensitive to musical
5. See also chapter 5, p. 137. I am referring to the analytical sections in Schoenberg’s

papers “Brahms the Progressive” (1933; Schoenberg 1984, 98–441), and “Composing
with Twelve Tones” (1941; Schoenberg 1984, 213–45), and to the analysis of his own Four
Orchestral Songs, Op. 22 (1932; Schoenberg 1965).
6. Schoenberg may have gone much further than that. The American composer
Dika Newlin remembered this characteristic utterance of her teacher, Arnold Schoen-
berg: “Music need not be performed any more than books need to be read aloud, for its
logic is perfectly represented on the printed page; and the performer [i.e., the player],
for all his intolerable arrogance, is totally unnecessary except as his interpretations make
the music understandable to an audience unfortunate enough not to be able to read it in
print.” (Newlin 1980, 164)

experience, and ignorant of what may render a work unique. Roger Scruton’s
judgment was particularly severe:
The dry pseudo-science of the language draws our attention only to what is
most lifeless in the music, and seems to forbid us to hear its expressive power.
(Scruton 1997, 415–16)
The professionalization of music analysis, through the systematic application of

Schenkerian theory and PC set theory, was deplored as degenerate by those who
saw a natural affiliation of analysis with criticism, like Scruton, Kerman (1980,
1985), or Treitler (1980). Yet, this professionalization has never been—and will
never be—able to change the fact that music analysis is essentially a matter of
signification, that it gives meaning to an object, even if analytical techniques are
governed by fixed protocols enforcing conformity without regard to the individ-
uality of that object. Even the strictest codification of analytical practice cannot
diminish its expressive purport.
Of course, this does not preclude complete failure. Analysts are fallible
to the same extent as performers are. As Carolyn Abbate wrote, “[verbally]
couched interpretations of music are performances and can only be more
or less convincing” (Abbate 1991, xi). The question is: how does an analysis
convince us as a performance, quite apart from the empirical or historical evi-
dence that it may provide? For one thing, it should demonstrate knowledge and
skill, the latter comprising both the power of observation and the ability to
arrange the various observations into a structured statement. For another, it
should convey an experience, that is, the impact the musical work has made on
the analyst. The balance is rather delicate; the problems some people have
with Schenkerian theory and PC set theory can be attributed to the upsetting
of this balance by an extreme systematization of analytical techniques. But a
highly sophisticated analytical apparatus need not be an impediment to ana-
lytical understanding (no more, in any case, than a highly developed technical
proficiency impedes a performance). If it is handled with all due awareness of
the analyst’s aesthetic creed, and with all due concern for the compositional
routines and the conventions of hearing that the musical work follows or
breaks, it may be the key to profound insights.
However, an analysis should also be convincing as an act. That is, one should
be made to believe that the musical work reveals itself through the analysis. There
is a growing literature on this aspect of the discipline, an objective that is as
vital as it is difficult to achieve. With reference to writings of David Lewin, Fred
Maus and Jonathan Dunsby, among others, Nicholas Cook (1999, 252) speaks of
an emerging “performative epistemology” of music theory, according to which
analyses produce, rather than trace, musical meaning. This has led to several re-
readings of the analytical literature. As an example, Cook discusses the perfor-
mative content of Schenker’s analysis of Beethoven’s Ninth Symphony:

In its wealth of striking literary images, in its fervent advocacy of the “absolute
music” tradition, and in its attempt to erase Wagner from that tradition, as well
as in its documentation for all time of a now defunct style of performance,
Schenker’s book constitutes an irreversible event in the reception history of
the Ninth Symphony. (Cook 1999, 254)7
The word “event” takes us back to the beginning of this chapter. Cook does not
present Schenker’s analysis as a commentary from the sidelines of music history,
but as a moment of music history. Music analysis is treated by him as an activity
and as such as an essential part of musical life itself. Consequently, it is not only
judged by its accuracy, but also by its capacity to engage people and to interact
with other ranges of activity (like composition, performance, and history). Anal-
ysis may not only reveal how music was composed, and how it should be played
or heard; it is also a process that can be enjoyed in its own right. By the same
token, a musical work is as much a possible object of an analysis as an incentive
to analyze. Any representation of this work is induced by reciprocal interaction
with a particular craft. The musical work enables people to express themselves
and to display their talents. People do not just mediate, unselfishly passing on
musical works to their contemporaries. They also appropriate, using them to
serve personal, artistic, scholarly, pedagogical, or political goals.
When we analyze works from the past, we make them contemporaneous with our
own present. A history of music analysis, then, is concerned with degrees of media-
tion and appropriation in the products of that discipline. The writing of such a his-
tory is not just for practitioners. Nor is it meta-history, if such a thing exists at all. The
history of music analysis is a branch of music history—and, ultimately, of history—
focusing on a particular aspect of musical life: a practice that may either be closely
allied with composition and performance, or pursued independently. Music analysis
history can be put on a par with music performance history, as it relates a musical
practice to its cultural environment. It allows us to see the larger complex of which
this practice is part; to see how musical meaning arises from this complex, and is
acted out—”performed”—by the analyst. As such, it is of interest to non-specialists if
we for the moment disregard the expertise its subject requires.
Before we start to study aspects of PC set theory’s interaction with its envi-
ronment, it is appropriate to emphasize the egalitarian spirit of the present dis-
cussion. It may be disturbing to some readers that analysis is treated on a level
with performance, and that analysts and performers are supposed to construct,
rather than just to channel, musical meaning. The work of historians, as stated,
is likewise constructed and creative. The problem is a confusion concerning dis-
ciplines that we have learned to distinguish so carefully. Scholars are not auto-
matically artists; yet, earlier on, we have declared them capable of taking artistic
7. See Cook 1995 for a more extensive reading, along the same lines, of Schenker’s
analysis.

action (even worse, by exploiting the work of others). Performance is not the
same as composition. Yet, what we have taken as a model for music analysis is not
performance as the reproduction of a text—in this case, a score—but perfor-
mance as an event, in which the performer takes charge of the text.8 This might
be an invitation to think of a score as an “ideal” performance (like the score of
Cassandra’s Dream Song by Brian Ferneyhough), an arrangement as an analysis
(like Webern’s orchestration of the Ricercar from The Musical Offering), a record-
ing as a score (like Miles Davis’s 1964 version of My Funny Valentine), or a compo-
sition as a historical essay (like Berio’s Sinfonia). Stretching the point, we might
say that there is only one range of musical activity, involving various disciplines; a
continuum of actions, reactions and reflections, deploying sounds, tones, words,
and/or gestures. What is the reason for suggesting this? Why should we blur the
traditional boundaries between these disciplines?
In the previous chapters, we have traced the evolution of a conceptual frame-
work pertaining to a vigorous analytical practice. Up to this point, we have seen
that PC set theory was driven by a major development in the field of twentieth-
century composition, that is, twelve-tone serialism. However, it is unlikely that
serialism was the only driving force for PC set theory. First, PC set theory has
never found a firm footing in continental Europe, even though serialism was for
a time just as pervasive there. Second, although its foundations were laid by a
composer and theorist of serial music—Milton Babbitt—PC set theory has never
shown an exclusive commitment to such music.
On the other hand, twelve-tone serialism has certainly served as the principal
paradigm of PC set theory, sometimes with far-reaching consequences for the
music to which it was applied. For example, we have seen how serial concepts
were redefined through PC set theory so as to become applicable to non-serial
music. In view of the origin of these concepts this music was deemed “atonal” or
“non-tonal,” even if there was evidence to the contrary, for example, in Stravin-
sky’s The Rite of Spring. Allen Forte’s analysis of this piece (Forte 1978) might have
constituted an “irreversible event” in its reception history, to paraphrase Nicho-
las Cook, if Richard Taruskin (1979) had not seriously questioned Forte’s stance
of “phenomenological virginity” (Taruskin’s words). Forte took up the gauntlet
a few years later (1985). The ensuing exchange between the two authors, in the
British journal Music Analysis, about a short passage from the Ritual of the Rival
Tribes, certainly constitutes an irreversible event in the history of music analysis. It
made the papers, even if only because of its unusual fierceness.9 The discussion
8. For a critique of the still-persistent idea of performance as reproduction, and the

radical negation of this idea in the work of music philosophers like Stan Godlovitch, Rob-
ert Martin and Christopher Small—which involves the subordination of musical works to
performances—see Cook 2001.
9. It was reported about in the New York Times (“Should We Care Who Wrote It?” by
Donal Henahan; May 17, 1987, 102). For another commentary, see Hinton 1988.

8va
W.W. 3
Tpt.
4-18:[2,3,6,9] 4-7:[11,0,3,4] 4-17:[11,2,3,6] 4-7 4-18:[9,10,1,4] 4-18:[10,11,2,5] 4-8:[6,7,11,0] (4-18)
Example 7.1a. A passage from Stravinsky’s Rite of Spring (Rehearsal No. 60). An
analysis by Allen Forte, The Harmonic Organization of The Rite of Spring.
8va
p.t. p.t. p.t.
ant.
**
B: I V* I V (VII) I V G: I VII ***
* (or "modally", VII)
** Stravinsky spells this note B
*** "anomalous"
Example 7.1b. The same passage. An analysis by Richard Taruskin (1979).
Example 7.1c. Taruskin’s model: Esprit Philippe Chédeville, Musette, La Sincère.

Richard Taruskin, “Letter to the Editor,” Music Analysis 5/2–3 (1986): 314. Used by permission
of Blackwell Publishing.
centered on the question of what were the most appropriate grouping criteria
for the tones sounding at Rehearsal No. 60. Forte characteristically took his cue
from recurring set-classes (such as his “4-18”), whereas Taruskin, in an equally
characteristic manner, provided an example from common (tonal) practice,
which could have served as a model for the passage (ex. 7.1).
This “event” shows that to talk of music theory as depending on actual music
needs a critical attitude, and is indeed very problematic with respect to PC set
theory. This theory has not only developed through a dialogue with a musical
repertoire, with respect to which it has tried to attain an optimal level of accu-
racy, but also through solving problems inherent to its own methodology. In
other words, it has also been self-driven, as for example the interest in segmen-
tation as a topic of theoretical speculation illustrates. Segmentation—the pars-
ing of a composition into significant musical units—is an important analytical

issue, especially when the music does not seem to suggest obvious criteria for
the process on a first hearing or reading. (It was actually the heart of the mat-
ter in the debate between Forte and Taruskin.) It is one thing to find relations
between PC sets; quite another to claim that these relations assure coherence.
This required a supplementary theory of segmentation, including typologies of
possible segments and procedural guidelines (Forte 1973, 83–92, Hasty 1981,
Schaffer 1991). PC set similarity, as we have seen in chapter 5, is another topic
that emerged from the self-imposed limitations of the theory; it compensated
for the rigid relational concepts PC set theory had adopted from twelve-tone
theory (like transposition and inversion). However, the decision to represent
musical structures by PC sets—and PC sets, in turn, by frequencies of interval
classes (the interval vector)—has deeply informed the notions of similarity that
were applied to these structures.
If a theory of music is not entirely music-driven (which is not as odd as it
sounds), if it exploits musical works as much as it explains them, then what is it
that makes this theory and its concomitant analytical practice flourish? Why,
and to whom, is it gratifying? These are important questions, for, whatever its
motivation, an analytical practice causes music to resonate with certain mean-
ings and to persist through time. It can even provide a “safe haven” for written
music that would otherwise have little chance to survive. Thus, music theory and
music analysis, while using the riches of great music, reward music with reflec-
tive thought enhancing its depth.10 To return to the earlier question: we do not
blur the boundaries traditionally dividing musical and scholarly activity, com-
position and performance, etc. These boundaries are already blurred, allowing
various practices to interpenetrate.
Historical Models and Topical Meanings
There are various notational devices to help us clarify music, like chord symbols,
harmonic reductions, voice-leading graphs, and form charts using letters. Fur-
thermore, we can draw on an analytical vocabulary offering many angles from
which to view musical compositions. However, these devices and vocabulary
are also attributes of a skilled activity seeking to maintain and improve itself by
means of music. We can describe this activity with regard to its effects, to the way
in which it reacts with the music, thus stressing its performative dimension. Sub-
sequently, we can put this activity into historical perspective, with regard to the
goals it serves and the competence it requires.
10. Joseph Kerman (1983) has called attention to the role of musical discourse in the
complex mechanism by which canons of respected musical works arise. For a more recent
study of this topic, considering craft, repertory, criticism, and ideology as principles of
canon formation, see William Weber (1999).

Let us consider the notational devices and the vocabulary of PC set theory. What
effect do they have on music? Is there a property or quality they accentuate? PC set
theory can be seen as basically providing an inventory of relations between PC sets.
However, this is not what defines it as a theory. PC set relations are in fact properties
of the algebraic structure that has been chosen as a model for pitch structure. PC
set theory is a theory only in so far as it claims that some of these relations can be,
and have been, exploited in musical compositions.11 What it does is to group notes
together and to encode the resulting combinations in order to substantiate this
claim. Michael Friedmann, whether he was conscious of it or not, has described this
kind of analysis as the listener’s active interference with the musical structure:
To perceive . . . a musical structure, one must first segment it from its surround-
ings in the musical continuity, so that it becomes a unit with sufficient internal
integrity to be independently scrutinized, and then identify characteristics that
can later be used to liken or contrast the unit to others. (Friedmann 1990, xxii,
my italics)
This is worth a moment of contemplation. What are the grouping criteria? They
are very hard to pin down. Joseph Straus probably gave an accurate characteriza-
tion of the general practice when he advised his students “to poke around in the
piece, proposing and testing hypotheses as you go” (Straus 2000, 51). And what
does the encoding involve? As we have seen, the analyst notates the PCs repre-
sented in a musical segment, and assigns the segment to a class. Most often this
is a Tn/TnI-class, the segment’s membership of which is certified by the “prime
form” (Forte) of its PC set or by its interval vector. The predominance of this
class type in the notation and vocabulary of PC set theory is no doubt a token
of its common utility. On the other hand, using transposition and inversion to
identify classes of PC sets means the generalization and perpetuation of the time-
bound, since the way in which these musical operations are defined—that is,
without regard to order and contour—reflects a specific phase in the history of
composition (see chapter 3). Paradoxically, this is the phase of twelve-tone seri-
alism. The paradox is that the “unorderedness” of PC sets can only be observed
in a context in which order is central.
11. The question of whether composers like Arnold Schoenberg or Igor Stravinsky
actually composed with PC sets—decades before their theoretical formulation—has been
subsumed under the “intentional fallacy” by Forte in his exchange with Taruskin: “I sub-
mit that we can never know with any certainty what ‘the composer thought he was about’
and that to attempt to do so to validate an analysis is an empty pursuit” (Forte 1986, 335).
However, if PC set theory does not address the question of how, and by whom, PC set
relations were used, it cannot not claim anything except that the music of Schoenberg is
composed of pitches under twelve-tone equal temperament (a statement entailing all the
relations between PC sets). Ethan Haimo (1996) criticized Forte for invoking the inten-
tional fallacy, and provided analyses of Schoenberg’s music taking into account Schoen-
berg’s compositional thought, as evidenced by his manuscripts and writings.

PC set theory fundamentally conceives of music as composed of segments

that are different in some ways, but connected by recurring intervallic patterns
or qualities. In other words, it claims that music achieves coherence by a con-
tinuous process of variation. This process is sometimes concealed by the articula-
tion of the musical surface. The phrases and rests, the changes of register and
instrumentation—in short, the salient features of a musical composition—may
not always help to identify the “relevant” segmentation. Or one may find several
competing segmentations, each of which involves different PC sets. Such ambi-
guities are not valued negatively, though. On the contrary, the prevailing view is
that conflicting segmentations produce rich musical structures requiring a deep
understanding, as the following statements illustrate:
Schoenberg often conceals the main structural components of his music

behind primary configurations. (Forte 1973, 92)
It seems most reasonable to begin by identifying the primary segments, that

is, those segments clearly articulated as distinct musical units. But it would be
dangerous [sic] to stop here—to do so is to assume that meaningful relation-
ships can exist only at the surface level. We must look for other, less obvious,
relationships. (Beach 1979, 15)
The presence of ambiguity in segmentation . . . allows for many interrelated

lines of development to take place and this makes it possible for the music to
achieve great structural richness and depth. (Hasty 1981, 60)
Some of these groupings may span across rests or phrasing boundaries, and
that is okay. A rich interaction between phrase structure and set-class structure
is a familiar feature of post-tonal music. (Straus 2000, 51)
Not every conflict is to be valued as aesthetically enriching, though. In A Guide

to Musical Analysis (1987), Nicholas Cook demonstrated an analytical application
of PC set theory as expounded by Forte (1973), using Arnold Schoenberg’s Lit-
tle Piano Piece, Op. 19, no. 6. Cook—a scholar from the English tradition; not
a “pure” set-theorist, but somebody deeply interested in music analysis and its
pedagogy—had set himself no easy task. On the one hand, he wanted to explain
why the set-theoretical approach was so appealing to some. On the other, he
had to interest the uninitiated by rewarding their immediate perceptions about
the piece. Now, as we can infer from the above-quoted statements, analysts felt
attracted to PC set theory because it transcended those immediate perceptions.
Therefore, Cook’s analysis is a rather unusual one, dealing only with the PC sets
of discrete musical units.12
12. Cf. the analysis of Opus 19, no. 6 made by Forte (1973, 97).

Cook’s segmentation is reproduced here as example 7.2a. Cook claimed to

have based it on “surface features like texture, rhythm and dynamics” (Cook
1987a, 124). At first sight, this seems certainly acceptable. The first three seg-
ments can be seen as having distinct formal functions: segment “A” presents an
idea, segment “B” repeats and elaborates this idea, and segment “C” carries it
further. The segments “D,” “E,” and “F” build a less cogent succession, but the
framing silences and contrasting textures give support to Cook’s segmentation.
(Moreover, segment “F” seems to serve as the recapitulation in a ternary form, as
it brings back the opening chords.) The PC sets of these six segments have inter-
esting relations. For example, “E” includes “D” and the complement of “F.” “D”
is also included in the complement of “A” (under inversion and transposition),
and the complement of “F” is also included in “B” (under transposition). Even-
tually, Cook shows that all the PC sets belong to one set-complex about “E.”
One of Cook’s more interesting observations is that “C” stands apart from
the other segments, although it belongs to the same set-complex. It is not
related to the two surrounding segments (“B” and “D”) in the same way that
these segments are related to one another. In other words, it interrupts a chain
of logical connections (Cook 1987a, 130). This is a statement about the musi-
cal content of Opus 19, no. 6, and that statement is supported by Cook’s neatly
drawn Figure 62 (reproduced here as ex. 7.2b). What Cook wanted his readers
to see is that PC set theory is capable of producing such statements. But what
is their value? On closer investigation, “C” is not a complete segment based on
the surface features used by Cook. The two silent beats in measure 4, followed
by the third appearance of the chord {A4,F♯5,B5} (a chord marking the begin-
ning of the segments “A,” “B,” and “F”), suggest that “C” should actually start
three beats earlier, that is, on the last beat of measure 4. Mark Delaere (1993,
82) has interpreted the entire first half of Opus 19, no. 6 as a classical sen-
tence, with the enlarged “C”-segment functioning as the continuation phrase.
The accelerating activity in this segment validates Delaere’s comparison, but
this activity is obscured by Cook’s segmentation.
Is it revealing to speak of a conflict between “set-class structure” and “phrase
structure” inherent in this work, a conflict by which it achieves “structural rich-
ness and depth”? Cook’s analysis does not pose such a distinction. That is why, in
this case, the conflict is not enriching but just problematic. To speak of “C” as an
interruption is inconsistent with its function, which is that of a culmination.
Admittedly, Cook’s interpretation is not unattractive. The idea of breaking
up an otherwise logical sequence of events is conceptually simple and artistically
potent. It is therefore worth being considered as an explanatory model for a piece
like Opus 19, no. 6, which is enigmatic, leaving much scope for interpretation.
What is more, “interruption” is also an important concept in Schenkerian theory,
as Cook does not fail to point out. It is the cadence on the dominant at which the
fundamental line of a composition comes to a “premature” halt (see example 1.5).
This historical background adds some weight to Cook’s interpretation. However,

Example 7.2a Nicholas Cook’s segmentation of Schoenberg’s Little Piano Piece,
Op. 19, no. 6.
© 1913, 1940 by Universal Edition A.G., Wien/UE 5069.

Example 7.2b. PC set relations between the segments identified by Cook.

A Guide to Music Analysis, by Nicholas Cook (1994). By permission of Oxford
University Press.
Cook’s interpretation is actually dictated by his starting point. As already stated,

he sought to demonstrate the analytical theory of Allen Forte. According to this
theory, segment “C” cannot begin in measure 4, because this would raise the car-
dinality of its PC set to 10, a number that Forte had decided not to take into
account (as Cook points out on page 125; see also chapter 4, p. 107). Conse-
quently, segment “C” is brief, and its pitch content suggests a discontinuity.
This just shows that an analytical theory imposes certain a priori conditions
on the analysis of a work; the music is made to respond to the analyst’s perspec-
tive. In any case, now that we have discussed what music analysts do, we may ask
what they want to achieve. Cook’s purpose was primarily educational: to teach
an analytical skill, not as an “esoteric specialism,” but as a part of the “taken-
for-granted professional equipment of the historical musicologist and the eth-
nomusicologist” (Cook 1987a, 3). He wanted to show his readers how one goes
about analyzing with PC sets, using the surface of Schoenberg’s piano piece as
a sample “playing field.” As we have seen, the more ardent set-theorists have
pursued their aims further. They speak of discovering relations beneath the musi-
cal surface, relations that might even contradict the articulation of that surface.
They analyze musical pieces as complex and multivalent networks.
Such an activity is certainly not without precedents or historical models. The
crystallization and elaboration of classical musical forms from 1750 onwards,
the subtle interaction between the tonal and thematic strands in, for exam-
ple, sonata forms, tonal harmony’s capability of generating various structural
levels (by “prolonging” tones and harmonies, or by moving into subordinate
key areas), and the intricate motivic relationships in the music of Beethoven,
Brahms, and Schoenberg, have all been factors in the rise of music analysis as
a discipline in its own right. As such, they also account for the biases of that dis-
cipline. One cultivates those fields that have yielded well. Analytical methods

a b
Example 7.3. Rudolph Réti’s reading of the opening measures of Mozart’s Sym-
phony No. 40, K. 550: the first three notes of the bass (a) form “a literal transposition
of the symphony’s main motif” (b).
Rudolph Reti, The Thematic Process in Music. By permission of Faber and Faber Ltd.
are based on historical canons—sets of works that respond to those methods—

which their practitioners often seek to enlarge. Felix Salzer (Structural Hearing,
1952), for example, has tried to expand the horizon of Schenkerian theory by
adding the music of, among others, Monteverdi, Ravel, Hindemith, Bartók, and
Stravinsky to what was originally one of the most exclusive analytical canons.13
Rudolph Réti (The Thematic Process in Music, 1951) and David Epstein (Beyond
Orpheus: Studies in Musical Structure, 1979) followed in Schoenberg’s footsteps, ana-
lyzing Classical and Romantic compositions in terms of the recurrence and devel-
opment of small intervallic cells. Thus, Réti could represent Mozart’s Symphony
No. 40 as a musical statement almost as motivically tight as Schoenberg’s Four
Orchestral Songs, Op. 22 (ex. 7.3; Réti 1961, 116). And Erwin Ratz (Einführung in
die musikalische Formenlehre, 1951) reformulated the theory of musical form, with its
inherent orientation towards the Viennese classics (Beethoven in particular), in
such a way that it also encompassed the music of Johann Sebastian Bach.
Seen thus, it is not strange that the canon of PC set theory has been expand-
ing beyond the serial and proto-serial repertoires of the second Viennese
school, as noted in chapter 2. Its productiveness in one field entitled it to move
on to other fields, just like Schenkerian theory and the theory of musical form.
Forte’s article “Pitch-Class Set Analysis Today” from 1985, which reports the use
of PC set theory in a wide variety of musical repertoires, reminds the reader of
a showcase of trophies. One such “trophy” is this quotation from an article on
Debussy’s prelude Brouillards, by Richard Parks. It argues that PC sets can serve
as a resource for the analysis of the music of this composer:
To a rather large extent, Debussy’s music (of which Brouillards may be considered
typical) reveals, in its pitch resources, combinations which exhibit characteristics
lying beyond traditional notions of harmony, voice leading, and a referential
tone and sonority (tonic). (Parks 1981, 134, quoted in Forte 1985, 35)
13. Schenker’s own canon consisted of the works of the great German and Aus-
trian masters from Bach to Brahms, supplemented with those of Domenico Scarlatti
and Frédéric Chopin. In recent years, Schenkerian theory has been applied to Japa-
nese koto music (Loeb 1976), jazz (e.g., Larson 1998), and American popular song
(e.g., Forte 1995).

When music is analyzed, it is often implicitly measured against a historical model:

a work—or a set of works—that have been shown to respond to the method
applied. Whether the choice of model is appropriate is sometimes a moot point.
The works of Debussy, like those of Schoenberg’s middle period, have sometimes
been analyzed with tools specifically designed for the tonal music of the com-
mon-practice period.14 In other words, they have been represented as the end of
nineteenth-century practice rather than the start of twentieth-century practice.
Forte and others have criticized such analyses for neglecting what is novel—and,
in their view, essential—in these works. (Hence Forte’s aversion to Taruskin’s
analysis of that famous passage from The Rite of Spring, which he expressed in
the same article; see ex. 7.1b–c.) Parks’s words on Debussy must have been very
welcome indeed.
It may seem that music analysis is involved in a large, ongoing historical proj-
ect: grouping works by identifying threads running through the history of music
composition. However, in view of the fact that music analysis is interwoven with
its own cultural environment, we may ask why analysts engage with particular
historical models—for example, the sonatas and symphonies of Beethoven, or
Webern’s Movement for String Quartet, Op. 5, no. 4 (the textbook example of
“free atonal” music). What has attracted them to these works?
The previous discussion perhaps suggests a cynical answer to this question:
these works underpin the relevance of their enterprise. This answer is not only
cynical but also incomplete. There is no doubt reciprocity between the analyti-
cal armamentarium and particular musical repertoires. However, this reciprocity
need not inhibit individual creative thought. Music analysts have broken fresh
ground of their own accord, and they have received impulses from their cul-
tural environment. An analysis may reflect current orientations in composition
and performance, or the culture and concerns of an institution; it may borrow
insights from other research disciplines; or it may be affected by developments
in the fields of technology, teaching, and education policy. In short, the analyst’s
perspective is subject to change, and so is what appeals to analysts in music. This
means that what they value in a historical model is of topical interest.
The reception history of the music of Debussy and Schoenberg, to which I
just referred, is an example of how music can take on different meanings in the
light of different historical models (in those cases, tonal versus non-tonal). In
the 1980s and 1990s, approaches emphasizing the self-containment of the musi-
cal work of art, which had dominated the music-analytical scene in the 1960s
14. Hugo Leichtentritt’s attempts to analyze Schoenberg’s “atonal” music as an exten-

sion of nineteenth-century tonal practice (Leichtentritt 1927, 436–57, 1928) are docu-
ments of historical interest. Many have followed in his footsteps, writing analyses dealing
with harmonic constraints in works of Schoenberg. For a historical overview, and some
additional analytical observations, see Schuijer 1996. For a persuasive tonal analysis of
Debussy’s music, see Pomeroy 2003.

and 1970s—and which had been targeted by Joseph Kerman—had to compete

against a growing interest in music as a social and cultural phenomenon, an
interest seen in a preference for scholarly criticism15 above scientific inquiry.
This shift of focus has also resulted in new, or renewed, historical models. For
example, gender studies have brought to the attention of music analysts reper-
toires that were either neglected (music by women), or considered of limited
analytical interest (popular music). They have also helped vocal genres—like
opera, oratorio, and song—rise through the ranks of a discipline traditionally
concerned with issues of instrumental music. Studies of “intertextuality” in
music were inspired by Harold Bloom’s theory of literary influence, which sees
the history of Western poetry as driven by a struggle of younger poets against
their predecessors, and the individual poem as an attempt to subdue an older
poem. The music of Brahms, Schoenberg, and Stravinsky—already firmly estab-
lished in the canons of music analysis—epitomized the anxiety about a rich, but
at the same time endangering, past, and figured prominently in this new branch
of study (Straus 1990b, Korsyn 1991).
Each of these examples reflects a state of affairs comprising a lot more than
a score, a set of analytical techniques, and a perceptive mind—and this is true
for PC set analysis as well. We need to determine what has led to an analytical
practice outside of all the cues it has taken from actual music, or from the his-
tory of music analysis. The following chapter will focus on two conditions that
have encouraged the use of PC set theory in the United States. The first is the
availability, from the 1960s onwards, of the computer as a research tool in the
humanities. The computer was welcomed not only because of its practicality, but
also because of its apparent capability to transcend the human mind. The sec-
ond condition is the American university itself, which has not only offered train-
ing to and employment for music theorists, but has also been a forum where
they could meet and exchange ideas and views in an intellectual atmosphere.
15. The term “criticism” was used by scholars like Leo Treitler and Joseph Kerman to
denote a way of writing about music that is historically and technically informed and con-
veys an artistic experience. According to a definition from Kerman himself, dating from
1965, “criticism” is “the way of looking at art that tries to take into account the meaning it
conveys, the pleasure it initiates, and the value it assumes, for us today.” (Quoted in Ker-
man 1985, 123)

Chapter Eight
Mise-en-Scène
From a narrativistic point of view, history is a staging of the past.1 This book has
placed PC set theory, an evolving theoretical discourse on music, in the fore-
ground, without adding much scenery. It has only made casual reference to its
social, cultural, and technological setting. Now that we are about to bring two
crucial elements of that setting onstage—the computer and the university—we
should realize that their mere presence does not provide an explanation for
their influence on music theory. They have exerted that influence because music
theorists reacted to them, probably for a complex of reasons. This chapter seeks
to investigate these reasons, with particular emphasis on the roles of Allen Forte
and Milton Babbitt.
A: Formalization and Forgetfulness

The Computer as a Role Model
Analyzing with PC sets can be laborious. The process of coding tone combina-
tions and establishing their relations in terms of their intervallic structures, the
operations that connect them, or the complexes to which they belong, is time-
consuming. Moreover, it is extremely prone to inaccuracy. The slightest error
can result in the wrong classification of a PC set, and this, in turn, can have
distorting consequences for an analysis. PC set relations do not necessarily fol-
low syntactical rules, so that, in the absence of detailed information about the
composition process, we have little that can help us verify them. Accuracy is the
only foundation on which the analyst can build.
PC set theory has done a lot to regulate the extraction of information from
scores and to ensure its correct classification. It has provided very precise coding
protocols, and it has mapped out a considerable number of musical relations in
the abstract. But what is most important, its notation also allows the electronic
storage and manipulation of this information. This has not been a response to the
1. I owe this expression to a number of writers. For example, Susan Sontag (1980,
116) used it to describe the workings of memory. Here, it is intended to state that what
is told about the past cannot be separated from how it is told. The impact of this idea on
historiography has been described as a “narrativistic turn” (cf. Ankersmit 1990, 23–33).

mise-en-scène 237
cognitive demands PC set theory made upon analysts; composers and music analysts
did not start to use computers because the number of PC sets was so vast, because
their interval vectors were so hard to memorize, or because their relations were so
intricate. On the contrary, it was by means of computer technology that this infor-
mation had become available in a comprehensive form in the first place.2
Few people realize that PC set theory was actually devised for computer-aided
analysis—today, this fact is ignored in most analytical textbooks. John Rahn’s
Basic Atonal Theory, for example, contains a few chapters exploring a single piece
by way of instructions and assignments. In spite of the integer notation and the
mathematical vocabulary, these instructions and assignments require only a
score, a keyboard, and audio equipment. With regard to set type identification,
Rahn comforts his readers:
With a little practice, using the shortcuts, you will soon find yourself doing this
quickly in your head. For example, learn to identify {7,8,1} as “up 1, up 5” in nor-
mal form ({7,+18,+51}) which, starting on zero, produces {0,+11,+56}. The inver-
sion will be “down 1, down 5,” and the reverse of this upwards will be “up 5, up
1” for normal form, which, starting on zero, gives {0,+55,+16}. Obviously [0,1,6] is
more normal than [0,5,6] so the Tn/TnI-type is [0,1,6]. (Rahn 1980, 82)
The following quotation has been taken from the 2000 edition of Joseph Straus’s
Introduction to Post-Tonal Theory. The student is taught here to relate the numeri-
cal notation of a PC set found in the eleventh song from Schoenberg’s Book of the
Hanging Gardens, Op. 15 to a tonal musical shape:
The four-note gesture is a member of set class 4-17 (0347). [Straus uses both
Forte name and prime form to identify PC sets.] It is easy to visualize this mel-
ody as a triad with both major and minor thirds, although, as we will see, it
occurs later in the song in a variety of guises. (Straus 2000, 61)
Michael Friedmann, in his Ear Training for Twentieth-Century Music, proceeds from
formal definitions (“DEFINITION 5.1: Z-related set classes”; Friedmann 1990,
74) to ear-training and sight-singing exercises (“EXERCISE 5.2: Distinguishing
Z-related set classes”; ibid., 75). The following exercise should be used to train
the skills of segmentation and set-class identification:
2. In the academic year 1963–64 Hubert Howe had worked in Princeton on a project
“involving the application of digital computers to music theory and musicology.” One of the
results of this project was “a list of pitch structures of sizes 2 through 10 together with their
complements, interval-content, and another list which shows the number of transpositions
with n common tones, for each possible n with each size PS.” To achieve this result, Howe
had used a program written in FORTRAN for an IBM 7094 computer. Allen Forte published
his first list of PC sets in the same year, but he only mentioned the program by which he had
mapped out the set-complex relations that were the topic of his research. This program was
written for a similar computer installed at Yale (Howe 1965, 45, 56; Forte 1964, 182, note 21).

238 mise-en-scène
As with trichords, the reading off tetrachord types should be practiced with
any one-voiced example, first by reading off adjacent tetrachords (the first
four pitch classes of a melody, then the next four, and so on), then by over-
lapping tetrachord types (notes 1,2,3,4, then 2,3,4,5, then 3,4,5,6, and so on).
(ibid., 100)
Even when, in these textbooks, PC set theory is presented as a set of practical

tools rather than as a compositional theory, it seems to be dealing only with the
question of how pieces of music have been put together, how they “work,” not
with the questions that spurred its development in the early 1960s: how to trans-
fer conventionally notated music to the computer, and how to retrieve new and
interesting information about that music from the computer memory. Now that
PC set theory is found in pedagogical programs, now that music students learn
to analyze with PC sets, to sing them, and to identify them by ear, we are hardly
ever reminded of those seminal questions.
In the mid-1960s the computer was a hot issue in the humanities. This was the
time that scholarly communities became aware of its potential as a research tool.
In the United States, the years between 1964 and 1967 saw a burst of conferences
on the subject, where computer applications in a number of disciplines, includ-
ing archeology, history, and literature, were demonstrated. The momentum was
felt in music research as well: computers were used in the spheres of electronic
composition, style analysis, structural analysis, music education, bibliography,
and thematic indexing. Computer centers at universities functioned as meeting
places where unlikely alliances were forged. “One of the virtues of a university
computer center,” the composer James Randall wrote, “is that it seems at once to
create, because of its numerous functions, an interdependence among people
of widely divergent positions, interests and skills.” (Randall 1965, 85).
At Princeton University, Randall himself, Godfrey Winham, Hubert Howe,
Tobias Robison, and Eric Regener—all of whom were interested in the use of
the computer as a tool for sound generation and electronic composition—
helped the historical musicologists Lewis Lockwood and Arthur Mendel, who
were carrying out a computer-aided investigation of stylistic properties of the
masses of Josquin Desprez. Lockwood (1970) described how he and Men-
del had the 18 masses key-punched manually from the Smijers edition in a
computer language called IML (“Intermediate Music Language”). This ardu-
ous task was carried out by paid students.3 Lockwood and Mendel were also
provided with programs for proof-reading and playing the music thus stored.
The higher aim, however, was to obtain answers to questions about musical
3. At the time, the punched card devised by Herman Hollerith (1860–1929) was still
a common input medium. It was prepared by typing characters on a keypunch, which
appeared as holes on the card. When scanned by a card-reading device, the patterns
formed by these holes generated a series of electronic impulses, which could be processed
by a computer.

mise-en-scène 239
structure and style. Such questions were translated into a computer language
for music information retrieval (MIR) that had been developed by Michael
Kassler and Tobias Robison. According to Robison, this language enabled an
investigation “fundamentally concerned with the occurrence of discrete sur-
face phenomena in the music.” He explained that:
[these] events, which might be certain chords, patterns of notes, rhythms,

pitches and so on, all share the characteristic that they can easily be sought out
and found by short sequences of MIR instructions. It is thus a simple matter to
write a MIR program which will look . . . for these phenomena. The results of
such searches can then be used in several ways. (Robison 1967, 127)4
For example, a typical question that could be handled by MIR was “to search a
piece of music for all examples of a rising second followed by a rising third” (P. H.
Smith 1967, 20). This question illustrates how the computer could help people
establish contacts and discover similarities between their scholarly orientations. It
was the kind of question that theorists of a more serial bent raised in connection
with the music of Schoenberg, Berg, and Webern. However, it was equally relevant
to an exploration of Josquin’s musical style, which should match Knud Jeppesen’s
“handmade” study of Palestrina’s—as Lockwood said, “in a tiny fraction of the
time and in a more flexible, more exhaustive, and more reliable manner.5
So, apart from speed and accuracy, the computer provided a sense of com-
munity among a differentiated group of music scholars. It also helped musicology
to develop self-confidence. Musicology was still a very young academic discipline
in the United States,6 but with respect to the use of the computer it was as old
as other branches of the humanities. Music was an area of research considered
4. For another description of MIR, and its relation to other early input languages
for music, see Kassler 1966. Kassler, a student of Babbitt, later developed a computer pro-
gram carrying out certain tasks involved in a Schenkerian analysis.
5. Lockwood 1970, 24. Lockwood’s statement reflects the situation in 1965. He and
Mendel were dealing with many other questions as well, including questions of musica ficta,
modality, and text setting. Judging from later reports (Mendel 1969 and 1976), much of
the initial optimism about the rewards of using computers had ebbed away by around 1970.
6. The years between 1930 and 1960 saw a rapid increase of activity in American
musicology, but it took some time before it was recognized as an independent academic
discipline. The first American chair in musicology was established in 1930 at Cornell Uni-
versity. Four years later the American Musicological Society was founded. In 1951, the AMS
became a constituent member of the American Council of Learned Societies. According
to Charles Seeger, “departments of music in colleges and universities and conservatories
of music knew little or nothing of European musicology and were not friendly to attempts
to introduce it [in the 1930s].” (Seeger 1969, ix). In his book The Place of Musicology (post-
humously published in 1957), Manfred Bukofzer wrote that that “musicology . . . entered
the American university by the back door, by way of established nonmusic departments,”
and he characterized its recognition as an academic discipline as a “slow and somewhat
roundabout process” (Bukofzer 1957, 3).

240 mise-en-scène
both challenging and promising in this connection: it was considered challeng-

ing because it demanded a specific language to represent music in all its com-
plexity; and promising, because, from the structuralist point of view that was en
vogue at the time, it seemed to lend itself to formalization without losing infor-
mation. Moreover, it was expected to yield numerous data that would otherwise
be hard to retrieve in the same quantities. Music would thus allow computer
technology to show off its great power.
The involvement of data processing techniques contributed to the aca-
demic status of both musicology and music analysis; the ability to employ the
computer counted as a measure of a discipline’s maturity. When summing up
the advantages of the new technology, hardly anyone failed to mention the
mental exercise it demanded. It was obvious that the computer could not just
solve every problem right away. In order to be “understood” by it, a problem
had to be broken down into a number of smaller questions. Each question had
to be converted into an algorithm, a sequence of instructions. Of course, the
range of these instructions was dependent upon the categories in which prior
information—for example, a musical score, a literary text, the census data of
a city, or a list of artifacts in an archeological site—was stored. This meant that
data, if not already formulated for the computer, underwent a considerable
transformation before an answer could come into view.
Some denounced the strong influence that the computer—a mere tool, after
all—exerted on the scholar’s perspective, narrowing it down to the measurable.
Many others, however, welcomed the opportunity it offered to test their research
questions and methods, and to meet its demands of exactitude and consistency,
thus raising the standard of accuracy of scholarly research. Those who incorpo-
rated the new technology into their work stressed the methodological virtues
of working with computers, even if they were aware of the dangers of scientistic
reductionism. Lewis Lockwood was one of them:
[T]he logical organization of computer investigation not only invites us but

compels us to frame our small-scale questions in precisely defined terms, and
thus to formulate our large-scale problems in ways more exact than seems likely
to be the case if we were merely making tentative steps towards answers that in
large measure appeared to be hopelessly beyond reach. (Lockwood 1970, 23)
The musicologist Jan LaRue, whose specialization was the study of musical styles,
saw a new age dawning:
May I recommend the computer to you as an instrument without human preju-

dices. It has its own prejudices, numerical and procedural. But these often act
as stimulances and correctives, as healthy balances and supplements to human
attitudes. With this new aid, the coming generation of musicologists should
develop a style analysis that is comprehensive rather than selective, broad
rather than personal, and rich in musical insight. (LaRue 1970, 197)

mise-en-scène 241
According to these authors, the computer held a mirror up to the scholars, mak-
ing them aware of their limitations. However, it also let them overcome these
limitations. The computer served them as a role model. By letting it perform
certain complex tasks, one hoped to gain a better understanding of these tasks,
and, at the same time, of the processes in one’s own mind. For example, analysis
programs would probe the consistency of common analytical methods. Likewise,
composition programs, such as those developed by Lejaren A. Hiller and Leon-
ard M. Isaacson, shed light on fundamental processes in existing music. Thus, it
was thought, the computer would not only assist, but also educate its users.
When we answer the question of what stimulated scholars to apply the com-
puter around 1965, we should not overlook the fact that the computer industry
had taken an interest in the evolving “market” of the humanities. It provided
funds and equipment for scholars who worked with computers. The Interna-
tional Business Machines Corporation (IBM) was frequently mentioned on
sponsor lists of conferences and research projects, together with universities,
and institutions like the Rockefeller Foundation and the National Science Foun-
dation. Apart from supporting current research projects, it tried to increase the
awareness of data processing techniques at universities. IBM employed the musi-
cologist Edmund A. Bowles as a manager of professional services in its Depart-
ment of University Relations. In this capacity, Bowles—himself a specialist in
medieval music—developed contacts with scholars in the United States (whose
equal he was) and acquainted them with the computer as a research tool.
In 1964 and 1965, for example, Bowles organized a series of six regional con-
ferences in the United States. The universities that hosted and sponsored these
conferences received financial support from IBM, as Bowles wrote in his pref-
ace to a published selection of conference papers.7 He reported that some 1200
people had visited at least one of the conferences. This number, and the rapid
succession of the conferences, testify to the intensity of IBM’s marketing:
The Use of Computers in Humanistic Research (Rutgers, the State University of

New Jersey; December 4, 1964)
Computers for the Humanities? (Yale University; January 22–23, 1965)
The Digital Computer as a Servant of Research in the Humanities and Arts (Univer-
sity of California, Los Angeles; April 15–16, 1965)
Humanistic Studies and the Computer (Georgetown University, Washington DC;

October 21–22, 1965)
Computers and Research in the Humanities (Purdue University, October 29–30,

1965)
7. This collection was published by Prentice Hall in 1967, under the title Computers in
Humanistic Research: Readings and Perspectives.

242 mise-en-scène
The Impact of the Computer on the Humanities and the Arts (Boston University,
November 4–5,1965)
The company also supported conferences and symposia devoted to single

topics. For example, it hosted a Literary Data Processing Conference, which was
held on September 9–11, 1964, and a symposium on input languages for repre-
senting music, on May 10, 1965. And it did even more. It granted financial sup-
port to the journal Computers and the Humanities, founded in 1966,8 it enabled
the American Council of Learned Societies (ACLS) and New York University to
establish a National Center for Bibliographic Data Processing in the Humani-
ties,9 and it sponsored fellowships that the ACLS provided for computer-ori-
ented research in the humanities.10
Obviously, research projects involving computers were beginning to gen-
erate funds at that time, and many opportunities were offered to share the
results with others. Apart from the Josquin project in Princeton, several other
computer-aided ventures in the field of music attracted attention: stylistic
investigations of the 16th-century Italian frottola repertoire (Lincoln 1967),
the 16th-century chanson repertoire (Bernstein and Olive 1969), and Joseph
Haydn’s symphonies (LaRue 1967); the bibliographic RILM project conducted
by Barry Brook (Répertoire International de Littérature Musicale; Brook 1967), the
generation of a complete list of all-interval twelve-tone series (Eimert 1964,
Bauer-Mengelberg and Ferentz 1965), and last, but not least, a project con-
cerned with the structure of non-serial atonal music. The initiator of that proj-
ect was Allen Forte.
Computer Language and the Structure of Atonal Music
Forte’s project comprised three levels of activity: encoding (representing a musi-

cal score in machine-readable language), parsing (dividing the encoded music
into smaller components), and interpreting (determining relations between
those components). For encoding, the first level of activity, Forte used a lan-
guage that had been developed by the multi-talented Stefan Bauer-Mengelberg
for the purpose of automatic music printing. Bauer-Mengelberg, a mathemati-
cian, lawyer, and conductor, worked as a programmer for IBM, and was Leonard
8. In the first issue’s editorial (“Prospect”), the contribution of “the major computer
manufacturer” is gratefully acknowledged (p. 1).
9. “News and Notes,” in Computers and the Humanities 1 (1966–67), 11.
10. For a statement on the ACLS Program for Computer Studies in the Humanities,
see Lieb 1966. Allen Forte held a fellowship in the ACLS when working on a score-read-
ing program during the academic year 1965–66 (see further below). His report includes a
footnote to the effect that this fellowship was sponsored by IBM. (Forte 1966, 363)

Sehr langsam ( = ca 50) espress.
mit Dämpfer 3
Geige
Klavier
5 col legno
weich gezogen
sempre äußerst zart

3 3
rit.
pizz.
Example 8.1a. Anton Webern, Piece for Violin and Piano, Op. 7, no. 1.

244 mise-en-scène
Example 8.1b. Forte’s representation of Webern’s score in Ford-Columbia

(DARMS). Each graphic symbol in the score is represented by one or more items
from the character set of an IBM Type 29 keypunch or typewriter. Numerical values
indicate positions on the staff. (Odd numbers indicate lines, even numbers spaces;
the numbers “21” and “29” refer to the first and fifth lines, respectively.) The music
is arranged per staff from top to bottom of a page in the score.
Allen Forte, “A Program for the Analytic Reading of Scores,” Journal of Music Theory 10/2
(1966): 334. Used by permission of Duke University Press and Yale University.
Bernstein’s assistant at the New York Philharmonic before he became president

of Mannes College of Music in New York in 1966. His “Ford-Columbia” input
language—also known as DARMS (“Digital Alternative Representation of Music
Scores”)—was very successful. It was straightforward with respect to standard
music notation, using an ordinary typewriter character set; furthermore, it could
cope with complex music, and was adaptable to various purposes.11
Example 8.1a shows Webern’s first Piece for Violin and Piano, Op. 7. Ford-
Columbia made such a piece of music appear as a single character string (ex.
8.1b; Forte 1966, 334). In this form it was ready for a computer program writ-
ten in SNOBOL (an “acronym” for “String-Oriented Symbolic Language”).
This language executed operations on free character strings. For example, it
could divide a string, or paste strings together; it could look for and take out
substrings, transpose them, substitute them, erase them, etc. Allen Forte was
an expert user of this language. He published a manual for SNOBOL3 in 1967
(Forte 1967b). In his own project, it served as the basis for a score-reading pro-
gram, the purpose of which was to extract segments from music encoded by
keypunch in the Ford-Columbia language. This was the second level of activity
in this project.
11. See Bauer-Mengelberg 1970 for a brief introduction.

(a) ‘Primary segments’ :
PSGT
1 [] (Cello)
PSGT
2 [] (piano RH)
PSGT
3 [] (piano LH)
(b) ‘secondary segments’ 1-5:
PSGT
1
SSGT 3 SSGT 1
SSGT 2
PSGT
2
PSGT
1
SSGT 5
SSGT 4
PSGT
2
(c) ‘secondary segments’ 5-8:
PSGT
1
SSGT 8
SSGT 6
SSGT 7
PSGT
3
PSGT
1
SSGT 5
PSGT
3
(d) ‘secondary segments’ 9-12:
PSGT
2
SSGT 10 SSGT
11
PSGT
3
SSGT 9
SSGT 12
PSGT
2
PSGT
3
Example 8.2. Anton Webern, Three Short Pieces for Cello and Piano, Op. 11, no. 1,
mm. 2–3. A segmentation following the rules of Forte’s score-reading program. The
example shows secondary segments (SSGT) resulting from the interaction between
three different pairs of primary segments (PSGT).
© 1924, 1952 by Universal Edition A.G., Wien/UE 7577

246 mise-en-scène
The way in which the score-reading program operated—Forte has described

this in much detail (Forte 1966 and 1967a)—was constrained by the properties
of SNOBOL. First, the instrumental or vocal parts had to be reassembled from
the Ford-Columbia character string, which obscured them (as ex. 8.1b shows);
new strings had to be created from substrings of the original character string.
Second, to each code that implied a temporal position—i.e., a note or a rest—a
position value had to be assigned, so as to enable the location of musical seg-
ments. So-called primary segments were extracted from character strings that rep-
resented single parts. They were defined as one-dimensional substrings delimited
by rest codes. In other words, any succession of notes in one part beginning after
a rest and continuing until the next rest counted as a primary segment. Example
8.2a shows the three primary segments extracted from measures 2–3 of the first
of Anton Webern’s Three Short Pieces for Cello and Piano, Op. 11. Next, the
program took all pairs of character strings, each member of which represented
a single part, and scanned them for secondary segments, which resulted from the
interaction between two contiguous primary segments. Each of the coinciding
substrings of these primary segments, the union of these substrings, and the
remaining substrings of each segment counted as a secondary segment (ex. 8.2b
through 8.2d).
Further programmed operations on the strings of primary and secondary
segments resulted in configurations involving more than two parts. In the seg-
mentation process, the pitch codes of the input language—a space code, or the
combination of a space and an accidental code—were replaced by PCs, while
other characters were deleted. In an output string, a segment like SSGT 9 in
example 8.2d would take the following form:
(P72*1*4*5*8*9 P80*7),
where “P72” and “P80” are position values with which the subsequent PCs are
associated.
The segmentation shown in example 8.2 differs greatly from the segmenta-
tion of the same excerpt in Forte 1970, which actually dates from 1965 (example
8.3; this example also includes the first measure of the piece). For my segmenta-
tion I have followed the protocols of the SNOBOL-based score-reading program,
which was not available in 1965. Forte’s segmentation is probably hand-made
and, as a consequence, involves some intuitive judgments. This may explain why
it is the better one. For example, it makes a lot of sense to distinguish the cello’s
B–B♭ motif (11,10), the melody in the right hand of the piano part (3,0,1,7),
and the chord {4,5,8,9} as primary segments. What, by contrast, causes the seg-
mentation of example 8.2 to be so awkward? The blame has to be placed on the
Ford-Columbia input language. Flexible though it may have been in its applica-
tion, it did not really lend itself to analytical purposes, especially not when the
music had a piano part. Example 8.1b shows that Ford-Columbia “read” a score

mise-en-scène 247
Example 8.3. Mm. 1–3 of Webern’s Opus 11, no. 1 as segmented by Forte in 1965.
Barry S. Brook, ed. Musicology and the Computer. New York: City University of New York
Press, 1970. Used by permission.
staff by staff. This meant that the left and right hands of a piano part were fed
into the computer as separate entities. Example 8.2 is in strict compliance with
this procedure. As a result, the G♯3 (8) of PSGT 2 (right hand) is unfortunately
separated from the tones of PSGT 3 (left hand) with which it forms a chord. Of
course, one can always rearrange a score before encoding, moving notes from
one staff to another. As we shall see, however, Forte had reasons to refrain from
such interventions.
Having identified a large number of segments, Forte could proceed to the
third level of activity, and trace relations between these segments. To this end,
he had written a sequence of programs in another language, called MAD (for
“Michigan Algorithm Decoder”). After determining the PC set of each segment,
these programs would (in Forte’s own words):
(1) determine the class to which each set belongs; (2) list and count all occur-
rences of each set-class represented; (3) compute, for each pair of set-class
representatives, an index of order similarity; (4) determine the transposition-
inversion relation for each pair of set-class representatives; (5) list, for each
set-class represented, those classes which are in one of three defined similar-
ity relations to it [R0, R1 and R2; see chapter 5] and which occur in the work
being examined; (6) summarize in matrix format the set-complex structure
of classes represented in the work; (7) accumulate and retrieve historical and
other informal comments in natural language. (Forte 1966, 357)
This enumerative description maps out a large part of the terrain of PC set the-
ory; several lines of development have departed from here. Some of these lines

248 mise-en-scène
were short-lived, while others have continued, owing to the work of Forte him-
self and many of his students. “PC set,” “normal” or “prime form,” and “interval
vector” have become household concepts for many music theorists, but in the
mid-1960s they met a very specific need: the need for definitions of musical rela-
tions that a computer program could recognize. A comparison of two musical
segments, such as those obtained by Forte’s score-reading program, now had to
involve an algorithm, a series of reductive operations applied to both with a view
to reveal possible correspondences between them.12
Forte seems to have worked on the three levels of his project in reverse order.
In 1964 he presented his work on PC set relations, enabled by MAD (Forte 1964,
1970). The score-reading program was developed in 1965 and 1966 (see foot-
note 10). No doubt, the implementation of this program was the most prob-
lematic part of the project. First, it was apparently written with a very specific
repertoire in mind: ultra-brief atonal compositions, a music of small gestures
and evocative silences.13 It is difficult to imagine what the program would have
done with long chains of chords and melody. Second, the score reading pro-
gram applied parsing rules that were context-independent and utterly rigid.
Obviously, the formulation of such rules was constrained by the properties of the
computer language, but Forte regarded these constraints as salutary rather than
restrictive. As we have seen, he was not the only humanities scholar who greeted
data processing techniques in the 1960s, but he was one of those showing the
fewest reservations. The cause to which he had devoted himself—the recogni-
tion of the structural integrity of the non-serial atonal repertoire—demanded a
fresh outlook, an attitude unimpaired by traditional notions and personal biases.
Forte believed that the computer could enforce such an attitude, and thus con-
tribute to a proper understanding of the “revolutionary” music of Schoenberg,
Berg, Webern, and others:
[T]he choice of atonal music as the object of research . . . brings with it the
need for an extensive revision of outlook. For example, conventional descrip-
tive language (“melody,” “harmony,” “counterpoint”) is not very useful, and
may even be a significant hindrance to new formulations, since it is oriented
towards older music. When, in addition, problems of structural analysis are
stated in terms accessible to computer programming, many conventions are
set aside and many familiar concepts are rejected after serious scrutiny. (Forte
1966, 331)
12. For algorithms yielding normal and prime forms of PC sets, see chapter 4, pp.
102–6. For the algorithm on which the interval vector is based, see chapter 5, p. 132.
13. Forte’s published demonstrations involve music of Anton Webern: Piece for Vio-
lin and Piano, Op. 7, no. 1 (Forte 1966; see ex. 8.1), Short Piece for Cello and Piano, Op.
11, no. 1 (Forte 1970; an excerpt of this piece is shown in ex. 7.2), and one of the Baga-
telles for String Quartet, Op. 9, no. 5 (Forte 1966).

mise-en-scène 249
Added to this came a more general epistemological viewpoint. More than once,
Forte has expressed the opinion that a musical analysis should have a clearly
defined object, providing all the evidence that could possibly support the analy-
sis. A valid statement about a musical work—a statement beyond a conjecture
or an opinion—should restrict itself to what verifiably constitutes that work: the
score. When Forte spoke about the “completeness,” “consistency,” and “testabil-
ity” of an analysis, as he did as late as 1985, the score was still his sole reference
point (Forte 1985, 42). This view of the object of a musical analysis—which was
informed by positivist epistemology14—fit with his efforts to computerize the
analytical process:
The musical score . . . constitutes a complete system of graphic signs and prop-
erly presented for computer input, may be analyzed by a program as a logical
image of the unfolding musical events that make up the composition. (Forte
1966, 332)
Can we disentangle PC set theory from the history of computer applications

in the humanities? Can we use it without reference to that history? Of course,
PC set theory has evolved. It has left the rigidities of its initial “technological”
phase behind. In The Structure of Atonal Music, for example, a primary segment
is defined loosely as “a configuration that is isolated as a unit by conventional
means.” Forte here refers not only to units demarcated by rest codes—as he
did in 1966—but also to “rhythmically distinct melodic figures” and “beamed
groups.” Secondary segments are now called “composite segments,” and the way
in which they are formed is left more open as well (Forte 1973, 83). However,
Forte still suggests a systematic and easily programmable procedure to obtain
less obvious, but possibly significant segments from a score: sequential imbrica-
tion, or overlapping. (This is the procedure introduced as an aural exercise by
Michael Friedmann in 1990: “notes 1,2,3,4, then 2,3,4,5, then 3,4,5,6, and so
on.”) Joseph Straus’s advice “to poke around in a piece” to obtain meaningful
segments is a long way from such attempts to formalize the analytical process.
Notwithstanding its greater flexibility now, PC set theory bears the marks of its
origin as a computer project. To begin with, the similarity and set-complex rela-
tions (5 and 6 in Forte’s enumeration; see above, p. 247) had never been defined
before the start of this project. And relations that had already made their appear-
ance in the analytical literature—such as the (unordered) transposition-inversion
relation (4 in Forte’s enumeration)—became more important, since the computer
could detect them so well. Furthermore, PC set theory presupposes a practice of
close reading of the score. Although this practice already existed before computer
programs were applied to music analysis—we have seen how much this meant to
14. For the relationship between atonal music theory and positivistic philosophy, see
Davis 1995. More on this in the next section of this chapter.

250 mise-en-scène
Schoenberg (see chapter 7, p. 222)—it was sanctioned, and reinforced, by the spe-
cial conditions that the computer imposed on the study of music.
All in all, PC set theory seems to have received considerable impetus from the
computer. The reason why this is not mentioned in textbooks anymore is the role
of music analysis as it has been sketched in chapter 6, more specifically its kinship
to performance. A successful performance makes us forget about its own contin-
gent nature. An analysis can also successfully represent a musical work of art, but
by transcending a mere exposition of musical techniques. What is more, it has
to transcend its own history. Therefore, it is understandable why the canonical
textbooks of PC set theory and PC set analysis are silent on the naive reception of
computer technology in the 1960s. A discussion of this would run counter to the
purpose of these textbooks: to present an authoritative body of music theory.
B: Contemporary Music, Science, and Education

Babbitt’s Positivism
The existence of a “distinctly American version of music theory,” presenting

itself as a “legitimate academic discipline rather than as a service discipline for
conservatories and university music schools” (McCreless 1996, paragraph 5), is
generally acknowledged. And there is no doubt that PC set theory is a product
of music theory’s adaptation to an academic environment under the sway of pos-
itivistic philosophy, as James Davis (1995) has shown. Positivism dominated the
American universities of the Ivy League in the 1950s and 1960s. Davis sees the
following tendencies as characteristic of this philosophical orientation:
[A] rigid empiricism which leads to an emulation of science in both methodol-

ogy and terminology; a rejection or avoidance of metaphysics; the use of linguis-
tic and logical analysis; and, partly as result of these characteristics, the avoidance
of subjective interpretation in the process of analysis. (Davis 1995, 508)
Milton Babbitt—who, while he was working on twelve-tone theory, laid much

of the groundwork for PC set theory—was affiliated with Princeton University,
where one of the European proponents of logical positivism, Carl Hempel,
taught philosophy of science. Inspired by Hempel, Rudolf Carnap,15 and others,
15. Rudolf Carnap was one of the founding members of the famous Viennese Circle
in the 1920s, but Babbitt liked to see him as a member of a “Viennese triangle,” together
with Schoenberg and Schenker (Babbitt 1987, 17 and 1999). Carnap fled the Nazi ter-
ror and settled in the United States (like Hempel, who came from Berlin), where he was
appointed at the University of Chicago. Later, he spent two years in Princeton, where Bab-
bitt had the opportunity to meet him (Babbitt 1999, 47).

mise-en-scène 251
Babbitt wrote a few methodological papers (“Past and Present Concepts of the
Nature and Limits of Music,” 1961;16 “The Structure and Function of Musical
Theory,” 1965; “Contemporary Music Composition and Music Theory as Con-
temporary Intellectual History,” 1972), in which he argued that any music the-
ory, while bearing on sense-experience, should comply with the requirements of
a formal theory. He defined such a theory as “a deductively interrelated system
of laws, statable as a connected set of axioms, definitions, and theorems, the
proofs of which are derived by an appropriate logic” (Babbitt 1961b: 399).
Babbitt was an adherent of the doctrine of the unity of science, a doctrine
to the effect that all sciences are representatives of a single universal science.
Although this “universal science” was first and foremost physics, Babbitt wanted
to subject statements about music to its regime as well:
[T]here is but one kind of language, one kind of method for the verbal for-
mulation of “concepts” and the verbal analysis of such formulations: “scientific
language” and “scientific” method. Without even engaging oneself in dispos-
ing of that easily disposable, if persistent, dichotomy of “arts” and “sciences”
(or, relatedly, “humanities” and “sciences”)—that historical remnant of collo-
quial distinction—it only need be insisted here that our concern is not whether
music has been, is, can be, will be, or should be a “science,” whatever that may
be assumed to mean, but simply that statements about music must conform to
those verbal and methodological requirements which attend the possibility of
meaningful discourse in any domain.17 (Babbitt 1961b, 398)
According to Babbitt, theories of music had failed to meet these requirements in the
past. More specifically, they had failed to define properly their empirical domains
(the musical repertoires to which they applied), and to choose properly their primi-
tives (Babbitt 1961b, 399). For example, harmonic theory had never succeeded in
making a reasonable case for the overtone series as the ultimate cause of the “tri-
adic-tonal system.” It is obvious that the first five overtones form a structure cor-
responding to that of a major triad—a potential tonic—but how could one infer
from this that chords tend to appear in particular successions, or that minor triads
can assume the role of tonic? Babbitt’s criticism did not target the assumption of a
“natural” cause for music, but rather the inability to base a theory without gaps and
internal contradictions on this axiom (Babbitt 1965a, 57–58).
16. This paper was delivered at the Eighth Congress of the International Musicologi-
cal Society in New York. The title was not Babbitt’s own; it had been given to the session
in which he spoke. Another paper was delivered by the German musicologist Heinrich
Hüschen.
17. This quotation shows Carnap’s influence on Babbitt’s thinking. It was a life-long
enterprise of Carnap to construct a meta-language of science by way of an analysis of the
concepts, laws, and theories of the special sciences. (Der logische Aufbau der Welt, 1928; Test-
ability and Meaning, 1936; The Methodological Character of Theoretical Concepts, 1956.)

252 mise-en-scène
Babbitt made an exception for Schenkerian theory, which he helped to gain

academic credibility at a time when it was not yet commonly known among
music scholars in the United States. In one of his few published statements about
Schenkerian theory—a review of Felix Salzer’s Structural Hearing—he argued
that the Urlinie (actually the Ursatz) was an integral part of the empirical domain
of this theory, while at the same time “it was completely acceptable as an axiom-
atic statement (not necessarily the axiomatic statement) of the dynamic nature
of structural tonality” (Babbitt 1952, 260). And he referred to his own twelve-
tone theory as a music theory evolving logically from two very simple perceptual
assumptions: “the capacity to perceive pitch-class identity and non-identity, and
interval-class identity and non-identity” (Babbitt 1961b, 402). These assumptions
enabled the reduction of the chromatic universe (PITCH) to a set of twelve
PCs (PITCHCLASS). Furthermore, they enabled the representation of musical
transformations by operations on this finite set of numbers. The consequences
of these operations, in terms of the variance and invariance of PCs and PICs,
could be derived from finite group theory.
For example, one could deduce from a PC set how many of its elements
remained invariant under a transposition operation. The number of PCs held
in common by a PC set A and its transposition Tn(A) is equal to the frequency
of the PIC of transposition (n) in that set.18 Babbitt (1960) extended this rule
to a PC set A contained in a twelve-tone series S, showing that for A to recur in
the Tn-form of this series, S should also include B = T12−n(A). He cited the series
of Schoenberg’s Third String Quartet, Op. 30, in which each of the six disjunct
dyads maps onto another, or onto itself, under T6 (ex. 8.4a and 8.4b).
While enhancing the scientific image of music theory, this also proved reward-
ing material for textbooks. In Serial Composition and Atonality, George Perle iden-
tified two more operations causing the same dyads to interchange (T3I and T9I;
Perle 1962, 129). Joel Lester used the Schoenberg example in an introductory
text for undergraduate students on analytical approaches to twentieth-century
music—a text drawing on several analyses by Babbitt. With reference to the
beginning of the second movement of Schoenberg’s quartet (ex. 8.4c), Lester
wrote the following assignment:
[The] intervals between first and second violin are pairs of pitch-classes from
P0 . . . Write out series forms P0, P6 [T6], I3 [T3I], and I9 [T9I]. What happens
to the intervals in P0 in these series-forms? What is there in the construction of
the series that allows these relationships to arise? (Lester 1989, 204)
This provided students with a basic awareness of the relation between the
structure of a series and the effects of particular operations performed on that
series—an awareness, in other words, of a system expressing itself in a different
18. See chapter 3, note 21, and chapter 6, note 25.

mise-en-scène 253
PIC: 9 6 5 5 3 6
Example 8.4a. The series of Schoenberg’s Third String Quartet, Op. 30, divided in
disjunct dyads.
PIC: 9 6 5 5 3 6
Example 8.4b. The T6-form of the same series. “[T]he interval succession deter-
mined by disjunct dyads is 9,6,5,5,3,6 [in PICs]. The interval between the identical
5’s is 6, between the complementary intervals 9 and 3 is also 6, and interval 6 is its
own complement. So, under the application of t = 6, the pitch content of disjunct
dyads is preserved and ex. 2 [here (b)] can thus be regarded as a permutation of
the dyads of ex. 1 [here (a)].” (Milton Babbitt’s commentary; Babbitt 1960, 251)
Adagio ( = 60)
Vln. I
2 3 4
Vln. II
Va.
Example 8.4c. The beginning of the second movement of Schoenberg’s quartet.

way in each new twelve-tone composition. The excerpts from Schoenberg’s Opus
30 thus came to serve a traditional pedagogical purpose. Originally, however,
this material was supposed to provide ammunition for the professionalization,
in an academic environment, of music theory. It was a means for composers and
music theorists to be at the forefront of scientific research.
Anxieties about Musical Discourse
Babbitt’s deep involvement in the positivistic philosophy of science can be viewed

from a historical or a contemporary angle. One may point to the long history of
the relationship between music theory and science, or refer to ideas and experi-
ences that gave Babbitt the incentive to extend this philosophy to music theory.

254 mise-en-scène
What do we learn from the first approach? As one of the seven “Liberal Arts,”
music was an established part of the late-Medieval university curriculum. It was
concerned with numbers and their ratios representing harmonic intervals. These
were believed to contain the key to the order of the universe. Therefore, music
was a truly scientific discipline (in the Aristotelian sense), transcending what was
only practically useful, and providing a model for understanding the physical
world at large. As Penelope Gouk (1999, 2002) has shown, music continued to
play this role far into the seventeenth century—for example for Robert Fludd,
Johannes Kepler, and even, to a lesser extent, for Isaac Newton. Babbitt, one
might say, has renewed music theory’s old claim to the status of a science. He has
attempted to reposition music theory in the vanguard of intellectual life.
This is certainly a viable perspective from which to view Babbitt’s mission. Need-
less to say, Babbitt was a composer and a teacher of composition, not a twentieth-
century theoros in splendid isolation from musical practice. And he was scornful
of those who represented the relationship between music and science in the past,
like Marin Mersenne or Jean-Philippe Rameau.19 However, his own invocation of a
source of authority beyond the canon of Western art music—a source of authority
that not only constrains and validates musical thought, but also brings it into align-
ment with other fields of human endeavor—can be seen as an echo of this relation-
ship. It reflects the concern with ultimate foundational principles that continued to
exist in music theory when, from the eighteenth century onwards, it became more
of an aesthetic endeavor than a scientific one.20 Babbitt spoke out so strongly against
Mersenne and Rameau—his seniors by 328 and 233 years respectively—because, in
his experience, the authority still accorded their theoretical works frustrated the
advancement of contemporary music, and impeded the recognition of music the-
ory as a contemporary intellectual discipline. It annoyed him beyond measure that
“music in the form in which it would appear pertinently in the documentation of
the intellectual activity of our time, as discourse on music by competent profession-
als, has not been admitted to membership in that activity” (Babbitt 1972, 157).
19. Babbitt had no time for Mersenne and Rameau. He called Rameau’s Génération
harmonique “a collection of usually unsatisfactorily formulated definitions, unconnected
protocol sentences containing these defined terms, and superfluous insular constructs.”
And he blamed Mersenne for justifying the rule of the number six (the correspondence
between the interval content of the major triad and the first six divisions of the vibrating
string) by the number of planets that were known at the time: “Beyond the intimations
of the cosmic scope and affinities of music, there is the implication that certain classes of
objects hierarchically ‘justify’ others, and the pressure of the still persistent numerologi-
cal fallacy (the assumption that two different exemplifications of the same number class
therefore possess other properties in common).” (Babbitt 1961b, 399, 400)
20. For example, acoustics provided backing for the harmonic theories of Riemann
(1880) and Hindemith (1937), and psychology for Kurth’s dissection of Romantic har-
mony (1923). Krenek (1937) found support for a fundamental revision of music theory in
Hilbertian axiomatics.

mise-en-scène 255
The role of the “competent professional” is crucial, and a recurring topic in

Babbitt’s writings. In his view, musical discourse should be limited to those with
musical competence. Showing a lack of such competence made one’s opinion
irrelevant. For this excessively protective attitude, Babbitt took another cue from
the sciences: the ongoing process of specialization. In the paper that earned
him the widely held reputation of an “alienated modernist”21—“Who Cares if
You Listen?,” first published in High Fidelity in 195822—he wrote:
The time has passed when the normally well-educated man without special
preparation could understand the most advanced work in, for example, math-
ematics, philosophy, and physics. Advanced music, to the extent that it reflects
the knowledge and originality of the informed composer, scarcely can be
expected to appear more intelligible than these arts and sciences to the person
whose musical education has been even less extensive than his background in
other fields. (Babbitt 1958, 39)
Specialized knowledge was a measure of the dignity of music. We can see how this
fit with Babbitt’s concerns as a composer, a point to be discussed below. But what
kind of knowledge would regain the musical specialist the communal respect he
once had earned on account of the “universality” of his insights? How could music
theory be recognized as a modern intellectual discipline if its practitioners were
not answerable to a community larger than their own? On the one hand, it was
necessary for Babbitt to disentangle himself from the foundationalism inherent in
Western music theory, and clear the way for musical experimentation:
[T]he informed musician . . . has been obliged to recognize the possibility,

and actuality, of alternatives to what were once regarded as musical absolutes.
He lives no longer in a unitary musical universe of “common practice,” but in a
variety of universes of diverse practice. (Babbitt 1958, 38)
21. A characterization attributed to Joseph Kerman by Martin Brody (1993, 180).

Kerman had written about the “sense of modernist alienation” that Babbitt’s utterances,
like those of Schoenberg and Adorno, betrayed (Kerman 1985, 101).
22. Apart from its being first published in a periodical for music listeners, the article
was reprinted in The American Composer Speaks (1966), edited by Gilbert Chase, and in Con-
temporary Composers on Contemporary Music (1967), edited by Elliott Schwarz and Barney
Childs. Babbitt and his advocates later stressed that the provocative title was not his own, but
was imposed by the editor. However, Babbitt did not change it when the article re-appeared
in the above-mentioned anthologies. It was only in a much later reprint, in Richard Kostela-
netz’s Esthetics Contemporary (1978), that the original title was used (“The Composer as Spe-
cialist”). Whatever its origin, the catchphrase “Who Cares if You Listen?” has exerted too
great an influence over the reception of Babbitt’s text to remove it from history. And, in all
fairness, the gist of Babbitt’s argument was hardly less provocative. Babbitt urged the neces-
sity of “a total, resolute, and voluntary withdrawal from this public world to one of private
performance and electronic media, with its very real possibility of complete elimination of
the public and social aspects of musical composition.” (Babbitt 1958, 126)

256 mise-en-scène
On the other hand, he had to create a new strand of foundationalism to war-

rant a meaningful discourse on music. Martin Brody, in his article “‘Music for
the Masses’: Milton Babbitt’s Cold War Music Theory” (published in the Musi-
cal Quarterly in 1993), has traced Babbitt’s insistence on the use of “scientific
language” back to this dilemma: Babbitt did not want to invoke absolute criteria
for musical practices or compositional techniques as such. He regarded them
as constrained, not by any doctrine on the nature of music, but by rules of dis-
course. Therefore, Brody speaks of a “metatheory,” a theory not of music itself,
but of its conceptualization. No a priori limits were set to the range of musical
invention, but statements that characterized, explained, justified, or criticized
music were to meet absolute criteria of scientific rigor.
Thus, in the words of Brody, Babbitt managed “to refashion a conservative
orthodoxy out of a radical, if anxious, acknowledgement of cultural relativism”
(Brody 1993, 167). How far this acknowledgement really went, and whether it
could hold out in spite of the “conservative orthodoxy,” is a moot point. How
flexible was Babbitt’s metatheory? Could it embrace “a variety of musical uni-
verses,” or was it mostly pitched toward his own brand of music? We might ask
with Brody why Babbitt imposed such stringent rules on what should be “an
open conversation among the participants of a musical culture.” Additionally, we
might ask why the forbidding complexity of Babbitt’s own writing should be the
corollary of the application of these rules—a question that Brody does not raise
in his article. These questions shift the focus away from Babbitt’s place in intel-
lectual history—from his defining the terms for a new rapprochement between
music and science—toward his cultural context.
This, then, is the second approach to an interpretation of Babbitt’s scientism,
an approach that may help to iron out some of the remaining incongruities. For
one thing, it is quite amazing that language has been so central to the thought
of this composer. Its importance as the last basis of communal discourse in a
fragmented musical world does not explain the fierceness with which Babbitt
has called for its purification, a fierceness that must have made it difficult for
many readers to take his side. Joseph Kerman (1985, 101) perceived “an under-
tone of distress, even rage” in Babbitt’s writings, “erupting into repeated assaults
and innuendos directed against various predictable targets.” According to
Brody, these emotions reflected Babbitt’s deep-seated aversion to nationalism,
proletarianism, and mass culture as exemplified by the cultural politics of the
Soviet-Union—a cultural politics that had outspoken proponents in the United
States in Babbitt’s formative years, the 1930s.23 However, we must also take into
23. See Brody 1993. The 1930s were a period in the cultural history of the United
States in which notions like “high art” or “autonomous art,” etc., came under great pres-
sure. First, there was much dissatisfaction with America’s cultural dependence on for-
eign nations. Composers like Aaron Copland argued a turn to the country’s own musical

mise-en-scène 257
account Babbitt’s experiences in the public domain as a composer of demand-

ing music with no roots in common practice.24 For such a composer, language
was much more than just a medium of conversation. It was something that could
make him or break him:
Perhaps there have been eras in the musical past when discourse about music
was not a primary factor in determining what was performed, published, there-
fore disseminated, and—therefore—composed, . . . when—indeed—the com-
positional situation was such as not to require that knowing composers make
fundamental choices and decisions that require eventual verbal formulation,
clarification, and—to an important extent—resolution. But the problems of
our time certainly cannot be expressed in or discussed in what has passed gen-
erally for the language of musical discourse, that language in which the incor-
rigible personal statement is granted the grammatical form of an attributive
proposition, and in which negation—therefore—does not produce a contra-
diction; that wonderful language which permits anything to be said, and noth-
ing to be communicated. (Babbitt 1965a, 50)
Babbitt felt his own music, the music of his students, and of the composers
whom he admired—like Schoenberg, Webern, Stravinsky, Varèse, Sessions, or
Dallapiccola—to be extremely vulnerable to misconceptions. And the effect
of such misconceptions could be disastrous in a cultural society in which con-
temporary art music was weakly represented. Since this music obviously did not
“speak for itself,” language could either be a threat to its survival or rescue it
from extinction. This may have motivated the density and armored rhetoric of
Babbitt’s written language—a language that does not comfort, but overpowers;
that does not provide access to music, but rather surrounds it with a barricade
of abstractions.
vernaculars, which involved a rejection of the imported European “grand tradition” of art
music. Second, the economic depression following 1929 was a serious blow to the institu-
tions that had been fostering Western art music, such as orchestras and opera companies.
Artists became reinvolved with social issues (Alexander 1980, 152ff.). Third, mention
should be made of the Popular Front, the broad coalition of working-class and middle-
class parties against Fascism that was founded in 1935 under the auspices of the Moscow
Comintern. For some time, the American branch of this movement successfully pursued a
cultural politics that emphasized the national history and traditions of America. (Alexan-
der 1980, 192ff, and Guilbault 1983, 17ff.)
24. These experiences are recalled in papers (Babbitt 1987, 1988, 1989) and inter-
views (especially Duckworth 1999). Sources of frustration include the economics of per-
formance, the uncompromising rehearsal schedules of orchestras, the level of journalism,
the lack of publishing opportunities, many fruitless attempts to apply for a Guggenheim
Fellowship (eventually awarded to Babbitt in 1960), and the lack of interest on the part of
the general public.

258 mise-en-scène
The University as a Refuge?
Whereas Babbitt’s methodology served to delimit what counted as knowledge

about music, his language may have restricted the number of those involved in
the production and the assessment of such knowledge—thus creating a caste of
specialists. However, it could not do this without the backing of an institution
to which the provision of knowledge was already assigned—a much-respected
and essentially meritocratic institution with a highly developed “training and
certification machinery” (Jencks and Riesman 1968, 45). Babbitt’s view of the
university as a “home for the ‘complex,’ ‘difficult,’ and ‘problematical’ in music”
(Babbitt 1958, 126), an environment at some distance from normal society, in
which market and consumer pressures were not all-pervading, indicates that his
theoretical writings—with their overt appeal to a learned audience beyond the
discipline of music—were a call for protection. In any case, he has never made
a secret of his affection for this institution, which he regarded as the modern
source of artistic patronage:
Perhaps, then, I should begin by bringing you the latest tidings from our sec-
tor: serious musical composition is still alive and well, and living in our univer-
sities, including my own. Or, in the interest of historical subjectivity, I could
with equivalent candor report that serious musical composition is fighting for
its life, is well on its way to extinction, and is still breathing only in a number
of well-heeled nursing homes for the musically self-indulgent, of which my uni-
versity is fortunately one. (Babbitt 1972, 152)
[The] survival [of serious music] seems unlikely when the conditions neces-
sary for that survival are so seriously threatened. These conditions are the cor-
poral survival of the composer in his role as a composer, then the survival of
his creations in some kind of communicable, permanent, and readable form,
and finally, perhaps above all, the survival of the university in a role which
universities seem less and less able or willing to assume: that of the mightiest
of fortresses against the overwhelming, outnumbering forces . . . of anti-intel-
lectualism, cultural populism, and passing fashion. (Babbitt 1987, 163)
This raises the question of how society has taken care of composers—or not.
There have been many complaints about the unwillingness of concert agencies,
orchestras, publishing houses, record companies, and broadcasting corporations
in the United States to bring the work of living American composers of “seri-
ous” music into the public realm. Babbitt was not the only one to seek refuge in
the university. In 1965, a society was formed to serve the interests of composers
whose public musical activities—such as performance, publication, and profes-
sional discussion—were mostly confined to the university: the American Society
of University Composers. Its founding members were unanimous in their rejec-
tion of the working conditions of composers in the public domain:

mise-en-scène 259
[An] environment where music is regarded as entertainment, where profes-

sional standards are set by non-professionals, and where writing about music
is dominated by a belief in amateurism, is inadequate to our professional
requirements.25
Babbitt was among these founding members, as were several of his closest allies,
such as Benjamin Boretz and Donald Martino. But the list also included the
names of other composers, for example Norman Dello Joio, Otto Luening,
and Gunther Schuller; concern for composition as an art was shared by many
people. Even a music historian like Nicholas Tawa—a known critic of art for
art’s sake, and an avowed adversary of Babbitt’s “self-perpetuating circle of serial-
ists”—observed that “the rejection of contemporary American composers and
their works reached epidemic proportions in the three decades after World War
II,” and that the composer was not the only culprit of this development:
[T]he United States, in social, economical, political, and cultural ways, has
managed to defeat its native art composers by imposing conditions on them
that are impossible to meet. Among these conditions are the equating of artis-
tic value with a huge audience, large monetary profits, and only amusement
rather than also the enlightenment of listeners. (Tawa 1995, 179, ix, x)
Between 1959 and 1970, the programs of American symphony orchestras

were surveyed annually by Broadcast Music, Inc. (BMI) and the American Sym-
phony Orchestra League. Oliver Daniel, who was responsible for the publication
of these surveys, aired a scathing judgment on the symphonic repertoire in the
United States:
No other country so ignores the creative efforts of its own citizens. And
nowhere is the generation gap more noticeable. Because so little attention is
given to the work of our young composers, they are turning toward smaller
instrumental combinations and electronic instruments.26
Still, we must keep a sense of perspective about this. It would be going too
far to say that American contemporary art music has virtually been banned
from American cultural life—a point so persistently hammered home by Bab-
bitt. It has received support through intertwined channels of public and private
funding, and it has benefited from the work of devoted organizations such as
the Composers’ Forum (founded 1935), the American Music Center (founded
1939), and the American Composers Orchestra (founded 1977). One problem,
however, is that the efforts to provide footholds for American composers
25. Quoted from an advertisement in Perspectives of New Music 4/1 (fall–winter, 1965).
26. Quoted in Hart 1973, 417.

260 mise-en-scène
of contemporary art music have had no self-sustaining effect. In other words,

for these composers to reach an audience in the public arena, the combina-
tion of special funds and “missionary” activity has remained a sine qua non.
Another problem is that this combination usually does not bring advantage to
everyone. Funding, commissioning, and honoring can only be selective, leav-
ing some composers “in the cold.” Either you should contribute to the goal of
the funding agency, or fall within its scope.
Substantial contributions to the dissemination of new music have been made
by the Rockefeller and Ford Foundations (established in 1913 and 1936, respec-
tively). These philanthropic agencies have helped to launch large projects, such
as the “Composers in Public Schools” Program (Ford Foundation, 1959–69)
and the “Orchestra Residencies” Program (Rockefeller Foundation, 1982–92).
They have also channeled funds to performance and composer service organiza-
tions, so as to broaden the scope for contemporary art music (Wiprud 2000, 8).
In other words, they have invested in the building of a community receptive to
musical innovation—an approach in keeping with the humanitarian ideals upon
which they were founded. Individual fellowships for scholars and artists (includ-
ing composers) have been provided by the John Simon Guggenheim Memorial
Foundation (established in 1925), but these fellowships have tended to fall into
the hands of the already successful.27 Federal support comes from the National
Endowment for the Arts, an organ of the National Foundation on the Arts and
the Humanities, which was established by the United States Congress in 1965. Its
task was to spread funds over the full range of native activity in the arts. Regard-
ing music, it has focused on organizations and institutions, rather than individu-
als, especially in the sphere of performance.28
The flow of funds to new music has sometimes followed a tortuous course.
For example, Babbitt’s Philomel, for soprano and tape (1964), resulted from a
commission of the American Music Center, which in turn was supported by the
Ford Foundation; and his Concerto for Piano and Orchestra (No. 1, 1985) was
written for the American Composers Orchestra, which had received a grant from
Francis Goelet—a New York-based real estate developer, and one of America’s
private patrons of new music (Wiprud 2000, 3, 13).
27. They were intended for “scholars and artists of proven abilities” from the
beginning. Senator Simon Guggenheim’s second Letter of Gift (June 7, 1929) stated so
explicitly.
28. Such initiatives are listed in The National Endowment for the Arts 1965–2000: A Brief
Chronology of Federal Support for the Arts (2000). Examples include the Audience Develop-
ment Project of 1967 (“to fund presenters of local concert series for young or unknown
artists,” 14–15), the Opera-Musical Theater Program launched in 1979 (“to help broaden
the concept of music theater and to make this art available to a larger audience,” 29), and
the Meet the Composer Program of the same year (“to help young artists work with com-
posers in residence,” 30).

mise-en-scène 261
In view of the supply of funds from government agencies and non-profit orga-
nizations—which are eligible for tax-exempt status in the United States29—we
might ask to what extent Babbitt’s doom-laden perspective on the condition of
serious contemporary music outside the university was justified. Babbitt spoke
warmly of an organization that was established in 1952 by the German-born
wine importer Paul Fromm: the Fromm Music Foundation. He considered this a
rare example of a foundation “devoted to the contemporary, mainly American,
composer,” and gratefully endorsed its founder’s belief that music “was not a
‘performing art’ but a creative art.”30 This distinction is revealing. Apart from
the implication that performing arts are “non-creative,” and therefore probably
subservient to other arts, it shows that Babbitt regarded composition as the most
vital aspect of a musical culture, the source of all other musical activity. That is
why, in his view, the composer’s craft warranted special protection.
Babbitt did not so much deny that American society provided opportuni-
ties for contemporary composers, as regret that these opportunities were not
tailored to their own needs. Indeed, one can say that many funding programs
have required an effort on the part of composers to find their own audience,
or to establish their own contacts with performers, rather than letting them
write music and taking care of everything else. This is not surprising; funders,
understandably, want to ensure that commissioned works meet with a posi-
tive response. However, Babbitt expected a greater commitment from society
to support the composer. This expectation was most probably strengthened by
the successes of his European contemporaries—Pierre Boulez, Herbert Eimert,
Karlheinz Stockhausen, and others with whom he shared an interest in serial
control, mathematics, and science. These composers had contracts with major
publishers, most notably Schott in Mainz and Universal Edition in Vienna.31
Some of them occupied key positions in musical life. For example, Eimert was
on the staff of the West German State Radio (WDR, for Westdeutscher Rundfunk)
in Cologne, which he made into an important platform for new music.32 The
29. See Section 501(c)(3) of the US Internal Revenue Code of 1954.

30. Babbitt 1988, 3, 4. The Fromm Foundation ran a commissioning program and
sponsored the first, and also other, performances of the works that had resulted from this
program. It brought composers and performers together in the Tanglewood Festival of
Contemporary Music (from 1964), and it sponsored the Seminars in Advanced Musical
Studies at Princeton University (1959, 1960). These seminars were intended as the Ameri-
can counterpart of the famous Darmstadt Summer Courses for New Music (see further
below), and gave birth to the journal Perspectives of New Music in 1962.
31. Universal Edition had also issued the journal Die Reihe, which served as the prin-
cipal forum for European serialism between 1955 and 1962. An American edition of this
journal was published between 1958 and 1968. The lacuna it left in relation to American
developments was filled by Perspectives of New Music (see note 30). For the competitive rela-
tionship between both journals, see Kerman 1985, 102, and Grant 2001, 2–3.
32. This forum was the famous “Musical Night Program,” which Eimert had launched
in 1948. At the time, the radio station’s name was Nordwestdeutscher Rundfunk.

262 mise-en-scène
WDR acted as a patron of Stockhausen, who worked at the electronic studio that
Eimert and Werner Meyer-Eppler had founded there in 1951. Boulez’s contri-
butions to the debate on cultural politics in France, together with his brilliant
international career as a conductor, caused the French government to appoint
him director of the musical department of the Beaubourg Center, a museum of
modern art and a center of creativity bequeathed to the French people by Presi-
dent Georges Pompidou.33 This was in 1974. Under the leadership of Boulez,
the musical department became known, independently, as the Institut de Recher-
che et de Coordination Acoustique/Musique (IRCAM), where musical research and
experimentation took place on a permanent basis. Babbitt could only dream of
being so prominently involved in a rich nation’s struggle for cultural prestige.
As an American composer, he considered himself to be at a disadvantage, and at
times he could barely contain his bitterness about this:
The Continental composer may be obliged to court program directors of radio

stations, but he also can expect to encounter even conductors—for example—
who not only share his nationality and cultural background but have the
authority and professional ambition to perform works other than the riff-raff-
ish displays of evanescently flashy timbral patinas which make life easy for the
American conductor, his performers, his audience, and his employers, and yet
are to count as a generous gesture to contemporary music.
The European composer can even expect to find a devoted publisher to
relieve him of many of all those onerous and demeaning tasks and obligations
which attend the preparation of a composition for performance and publica-
tion. Any American composer who anticipates the time when, if not just for
eminence at least for advanced age, he may receive such treatment, is very
likely to be disappointed. . .
The Continental composer may guide his professional behavior with a view
to enjoying the material rewards of genuine celebrity; the American composer
of highly “cultivated” music, with no illusions as to what are the cultural heroes
in a people’s cultural democracy, may attain bush-league celebrity, with many
of the disadvantages of materially genuine celebrity and but few of the advan-
tages. (Babbitt 1989, 110)
Babbitt was right insofar as the composers in Europe (whom he was hardly
willing to consider his equals) seemed to have received unconditional support
for their work. In the words of Paul Griffiths, “their sole responsibility was to
33. Boulez’s outspoken articles and interviews, and his successful organizational
initiatives—such as the Domaine Musical series of concerts in Paris (1954–73)—had
increased his celebrity. After the French government had refused to act upon his pro-
posals for a reform of musical life, in 1966, Boulez turned his back on officialdom in his
home country. The subsequent breach ended with his appointment at the Beaubourg
Center in 1974.

mise-en-scène 263
create” (Griffiths 1995, 36). However, Babbitt may have been unaware of the
sometimes very special conditions that warranted this “unconditional” sup-
port. The history of the Darmstadt Summer Courses for New Music is most
illustrative in this respect. This annual contemporary music festival, which fea-
tured concerts, lectures, and workshops, provided another major platform for
the European avant-garde, but it had come about, in 1946, under auspices of
the American occupying military government in Germany. The initiative of a
local administrator, it had been adopted as part of a program to rebuild the
German cultural infrastructure, and at the same time infuse it with American
cultural and political values. American music played a much more prominent
role in the Summer Courses than was suggested by their reputation as a vehi-
cle for the group around Stockhausen and Boulez34—the “Darmstadt group”
as it is called in Donald Grout’s widely read A History of Western Music (1980,
740). This throws an interesting sidelight both on the complaints about the
marginal status this music held in the United States itself, and on the denigrat-
ing remarks about American composers with which Boulez shocked readers of
the New York Times in 1969:
Americans are jealous—I’m not sure if that’s the right word—thinking the
Europeans are taking attention away from them. The Americans do operate
under a severe handicap, of course; they have no strong personality in the
field. If they were strong enough to establish their personality on the world,
they would see that no national favoritism exists.
Boulez directed the bulk of his criticism at the university affiliations of American
composers:
European music is not connected with the university. There is no ivory castle
for us . . . The university situation is incestuous. It is one big marriage in which
the progeny deteriorates, like the progeny of old and noble families. The uni-
versity musician is in a self-made ghetto, and what is worse, he likes it there.35
34. Amy Beal (2000) has shown that the director of the Summer Courses, the former
music critic Wolfgang Steinecke, received financial and material support on request from the
Music and Theater Branch of the American military government. And American officials,
like the composer Everett Helm, had a hand in the organization of the Summer Courses:
“In addition to supplying money . . ., Helm helped Steinecke contact American compos-
ers and obtained scores of American music from the United States. During the fifties he
remained actively involved with the [Summer Courses] by lecturing frequently on American
music.” (Beal 2000, 114) Milton Babbitt taught in Darmstadt no earlier than 1964.
35. Boulez interviewed by Joan Peyser for the New York Times (“A Fighter From Way
Back”; March 9, 1969, 19). Boulez followed in the footsteps of Stravinsky, who had warned
young composers, “Americans especially,” against university teaching in his conversations
with Robert Craft (Stravinsky and Craft 1959, 132). Boulez later became a vigorous cham-
pion of the music of Elliott Carter.

264 mise-en-scène
Were American composers a more appropriate target of the “ivory tower” accu-
sation than their European counterparts? One cannot reproach somebody for
seizing the best opportunities that are available in his profession. Boulez et al.
no doubt have achieved a stronger presence on the international music scene—
they have kept in touch with the public world to which Babbitt had bidden fare-
well in “Who Cares if You Listen”—but America’s cultural politics in Germany in
the aftermath of the Second World War had been a factor in their rise to fame.
We know that Babbitt “liked it there.” He felt at home at the university. He
had positioned himself as someone who spoke the language of contempo-
rary science. As such, he claimed to have a more thorough understanding of
Schoenberg’s twelve-tone music than Schoenberg himself. Babbitt thought that
Schoenberg “like many other great innovators, was not . . . entirely aware of the
implications of his own discovery.”36 And he scorned the European serialists of
his own generation particularly for what he saw as their methodological shortcom-
ings, while leaving aesthetic issues unaddressed:
Mathematics—or, more correctly, arithmetic—is used, not as a means of char-

acterizing or discovering general systematic, pre-compositional relationships,
but as compositional device, resulting in the most literal sort of “programme
music” . . . The alleged “total organization” is achieved by applying dissimi-
lar, essentially unrelated criteria of organization to each of the components,
criteria often derived from outside the system, so that—for example—the
rhythm is independent of and thus separable from the pitch structure. (Bab-
bitt 1955, 55)
Babbitt and Boulez both saw the university as separate from the larger world.
But they fundamentally disagreed in their assessments of this relation. Whereas
Boulez offhandedly depicted the university as insulated—an “ivory castle,” a
“ghetto”—Babbitt rather saw it as a “fortress.” For him, the university was not a
greenhouse of impractical knowledge but an outpost of a developing society.
Paradoxically, the idea that the university offered a relatively safe haven for
innovative composers arose from the interaction of American universities with
society at large. These universities have a long history of societal involvement.37
They are regarded as directly contributing to the progress and prosperity of the
nation, which naturally involves a responsiveness to problems and questions
36. Babbitt 1950, 266. Babbitt expressed this view again many years later, in the lec-
tures he delivered at the University of Wisconsin, Madison, in 1986: “Schoenberg never
understood the generality of the principles involved.” (Babbitt 1987, 14)
37. Clark Kerr, president of the University of California from 1958 till 1967, and an
acknowledged chronicler of American higher education, has described this history (Kerr
1995). It began with the Morrill Act of 1862, which authorized grants of public land for the
establishment of universities contributing to the industrial and agricultural development
of the United States. These “land grant” universities (e.g., Rutgers, Cornell University,

mise-en-scène 265
raised from outside the groves of academe. From early on, they have established
or accommodated professional schools, so as to ensure the transmission of new
knowledge to the applied professions. Of course, this utilitarian image of the
university has not gone unchallenged. There has also been the image of an insti-
tution devoted to the disinterested pursuit of truth. For example, Allan Bloom
thought of the university as a vital counterbalance to the regime of public opin-
ion in a modern democracy.38 But the disappointment of this classic man of
letters with the way in which the university had developed in the United States—
especially with its growing neglect of the canon of Western literature and intel-
lectual thought—is a natural result of its relative openness to influences from
outside the campus. This openness has long distinguished American universities
from European ones.39 In the words of Francisco Ramirez, a specialist in com-
parative higher education, “American universities are more likely to historically
develop as universities embedded in civil society while European ones are more
likely to be buffered from society and linked to the state” (Ramirez 2002, 257).
The employment of composers by universities is a testimony to this “embed-
dedness” rather than to the recognition of music as a scientific discipline in its
own right. American composers have been affiliated with universities since long
before Milton Babbitt gave us the archetype of the “university composer.” The
first Chairs of Music were occupied by John Knowles Paine (Harvard, 1875–
1906), Horatio Parker (Yale, 1894–1919), and Edward MacDowell (Columbia,
1896–1904). The American university, due to its spectacular growth in the twen-
tieth century (both in the size of enrollments and in the range of programs
offered), has become a home not only for the sciences but also for the creative
arts. It is “the place where post-secondary education ought to occur, regardless
of its activity or goal.” (Risenhoover and Blackburn 1976, 7).
the University of California at Berkeley, and the University of Wisconsin), offered a mix-
ture of academic and vocational instruction, and much of their research was intended to
meet practical needs. Other instances of societal involvement include the participation
of universities in programs of military research during and after the Second World War,
and the G.I. Bill (1944) enabling war veterans to enroll at colleges and universities (which
necessitated a diversification of programs). The final stage of the history described by
Kerr is that of the “multiversity,” a large but disunited knowledge factory accommodating
various communities and serving a disparate clientele.
38. “It is to prevent or cure this peculiar democratic blindness that the university
may be said to exist in a democracy, not for the sake of establishing an aristocracy, but
for the sake of democracy and for the sake of preserving the freedom of the mind.”
(Bloom 1987, 252)
39. Dutch scientists and scholars who traveled in the United States between 1900 and
1940, such as the historian Johan Huizinga, the pharmacologist W. Storm van Leeuwen,
and the chemist Hugo Rudolph Kruyt, were very surprised to see the role of vocational
instruction at American universities, the pragmatic orientation of scientific research, and
the tendencies towards the popularization of science (Van Berkel 1990).

266 mise-en-scène
Composers were not recruited as researchers, but as educators who could trans-
fer hands-on experience to students of music and musicology. The musicologist
Edward Lowinsky stressed the value of their presence on the faculty: composers,
he wrote in 1963, “impart to musical training an acuteness, a liveliness, and stan-
dards of musical quality that spring from their constant translation of musical craft
into musical art in their own work” (Lowinsky 1963, 47). In the same year, Rus-
sell F. W. Smith, then associate dean of general education at New York University,
expressed a similar view with reference to the recruitment of artists in general:
[W]e need scholars in a university because students are given their best chance
if they learn philosophy from philosophers, sociology from sociologists, and
biology from biologists, not from historians and appreciators of philosophy,
sociology, and biology, so they have their most real introduction to the arts
from artists, not from historians or appreciators of the arts. (Russell F. W.
Smith 1963, 69)
A high value was placed on the contribution of artistic practices to academic

programs. It was considered important to nourish these practices and to keep
them within reach. This was the main reason to offer doctoral degrees—such
as the DMA (for “Doctor of Musical Arts”) or PhD (for Philosophiae Doctor)—in
composition from the 1950s onwards. These degrees were not so much intended
to reward “scientific accomplishments” as to create job opportunities for com-
posers. As certificates of eligibility, they were especially required by less distin-
guished colleges, which could not afford to hire composers who had already
proven themselves by commercial success (Beglarian 1971, 465).
The first PhDs in composition were awarded to students of Babbitt at Princeton
University, after the degree program had been established there in 1961.40 Whether
this brought academic recognition to Babbitt’s program is a matter of doubt. The
musicologist Arthur Mendel, who chaired Princeton’s music department from 1952
till 1967, gave remarkably shallow reasons for the existence of a PhD in composition:
(1) it was more difficult for composers than for musicologists to achieve publication,
so they actually had more need for the upper-grade license; (2) there was already
a Master’s degree for composers, and it would be illogical not to offer a Doctor’s
degree as well; (3) no other degree than that of PhD was offered in Princeton in
any other field, and it was not desirable to create an entirely new one for compos-
ers (Mendel 1963). This suggests that the degree was actually a sort of compromise
resulting from the multiple functions and goals of the American university.
40. These students were required to hand in a composition and an analytical or theo-
retical essay, but over the years the emphasis shifted toward the essay. Examples of such
essays include: Godfrey Winham, “Composition with Arrays” (1964), Michael Kassler, “A
Trinity of Essays” (1967), and Benjamin Boretz, “Meta-Variations: Studies in the Founda-
tion of Musical Thought” (1970).

mise-en-scène 267
Another indication that composers did not owe their university positions
to their image as “scientific researchers” is the relatively small number of seri-
alist composers working at universities. After the Second World War serialism
became associated with scientism. However, the commonly held view that serial
music thus naturally secured a home within the American university—at the
expense of other musical orientations—has come under scrutiny. An article by
Joseph Straus in the Musical Quarterly (1999) provides statistical information that
suggests otherwise. Straus examined composers who taught in universities from
1950 through 1969. He discovered who earned recognition by performances,
recordings, and publications, and who received important grants or prizes in
this period. From these data he inferred that the American contemporary music
scene was not dominated by composers of serial music. Even at the four alleged
hotbeds of musical modernism (Harvard, Yale, Columbia, and Princeton), these
composers represented a small, if high-profile, minority.
Certain aspects of Straus’s research invite criticism. For example, it is a bit
odd to draw one’s information about the stylistic orientations among university
composers in the period 1950–69 from the 1970–71 edition of the Directory of
Music Faculties in Colleges and Universities.41 Furthermore, even if one wants
to denounce the “myth of serial tyranny,” its existence cannot be brushed aside
as resulting from gossip, a lack of knowledge on the part of journalists and his-
torians, or the “provocative rhetoric” of a few serial composers (Straus 1999,
333–34). Musical discourse does not cover music like a layer of dust. It takes part
in the chain of actions and reactions constituting the history of music—which is
exactly the reason Milton Babbitt was so anxious about it.42
Notwithstanding these objections, Straus deserves credit for pouring the cold
water of statistics over the heated claims of favoritism and straitjacketing. One
of the things he has convincingly shown is that the American academy at large
has had no special interest in furthering serial composition per se. It has been
indifferent to the emulation of a scientific attitude by composers of serial ori-
entation. It is unlikely that it has ever required composers to serve the progress
of music under the same methodological regime that prescribes how scientists
should serve the growth of knowledge. If there has been a trend toward equating
composition with scientific research, this concerned the academic freedom—
specifically, the freedom from extensive supervision—allowed to composers and
researchers alike. The composer’s task has been to maintain and teach an artis-
tic practice, as part of the university’s involvement in cultural life. In this regard,
41. See the caption of Figure 2 in Straus’s article. Straus refers to this figure as
follows: “By 1970, at the end of the period under study and after twenty years of presumed serial
dominance, one might expect to find at least a majority of these . . . composers to be
working in the serial idiom. The facts however are somewhat different.” (Straus 1999,
307; my italics)
42. For a substantial criticism of Strauss’s article, see Shreffler 2000.

268 mise-en-scène
the American university is better depicted as a complement of society rather than

as a refuge from it.
It is not surprising that Boulez, coming from Europe, failed to take the his-
tory and societal context of the American university into account when criticiz-
ing its role as employer of composers (and performers). What is surprising is
that Babbitt, as an advocate and beneficiary of this relationship, seems to have
misunderstood its role as well.43 His scientific rigor was largely self-imposed, and
his view of the university as a “fortress” ignored this organization’s permeability
to the surrounding society as much as Boulez’s image of an “ivory castle.” This
raises a number of questions, the very last this study will deal with. Babbitt is
credited by many with pioneering the American theoretical discourse on music,
and with bringing academic respectability to it. How could he have been so suc-
cessful if, as the foregoing discussion suggests, this was largely his own preoc-
cupation, stemming from his concerns as a composer, and not so much that of
the academic environment in which he was working? Furthermore, what is the
extent of his influence? Did this theoretical discourse develop in alignment with
his own writings, or has it been shaped by other factors? As for PC set theory, we
have identified one such factor earlier on: the rise of computer technology. But
we shall consider one more: the practice of teaching music theory.
A Democracy of Learning
No doubt, Babbitt’s work has been instrumental in the birth of what is called
“contemporary” music theory by some (like McCreless 1996, 1997), and “profes-
sional” music theory by others (like Kraehenbuehl 1960, and Babbitt himself;
1987, 121)—a discipline that, judging from these characterizations, was intended
to obliterate old-fashioned and dilettantish music theory. However, Babbitt’s
view of music theory as an empirical-theoretical system should not be seen as the
Open Sesame! to the research departments of American universities.44 Nor did
his view become a paradigm for all research endeavors in the field.
Babbitt’s scientistic conception of the discipline certainly elicited responses.
It was developed further, or modified, in methodological papers by younger
colleagues like Benjamin Boretz (1969/1973), John Rahn (1979a, 1979b), and
43. The gap between Babbitt’s ideal university and the realities of academic life
appears from his remark that “the university’s deadly deficiencies . . . are the extent to
which it reflects the ‘real world’ outside the university.” (Babbitt 1970, 65)
44. Accepted wisdom has it that the establishment of the doctoral program at Yale
meant American music theory’s breakthrough as an academic discipline. However, it was
not a breakthrough on all fronts. As McCreless (1997, 38) has noted, the discipline has
gained entrance more easily to university-based music schools and liberal arts colleges
than to research universities.

mise-en-scène 269
Matthew Brown and Douglas Dempster (1989). And there were theories that
obviously, if perhaps tacitly, took it as their starting point. Fred Lerdahl’s and
Ray Jackendoff’s application of generative linguistics to the structure of classi-
cal Western tonal music is a case in point. Starting from the notion of “group-
ing”—that is, from an experienced listener’s capacity to organize sound signals
into larger entities (motifs, phrases, themes, etc.)—they developed a theory of
musical understanding (A Generative Theory of Tonal Music, 1983). This theory
consisted of logically connected rules of musical grammar, rules according to
which Lerdahl’s and Jackendoff’s “expert listener” would decide how to under-
stand a musical composition. Whatever its shortcomings, it was a serious attempt
to combine a formal organization with a focus on the empirical world.45
But not all the work that was published in the wake of music theory’s advent
as a research discipline was influenced by Babbitt’s attitude, not even when that
work seemed to show signs of obedience to the rules of scientific explanation. It
certainly did not hold true for PC set theory, which could be formalized, but was
hard to operationalize. In other words, it was hard to provide empirical indica-
tors for the importance of PC sets, which the theory sought to underpin.
This problem has been raised in previous chapters. In chapter 3, it was
shown that the assumed independence of PC sets from musical articulation
seriously hampers their verification. Chapter 4 dealt with the concept of PC
set equivalence, which served to unite what could be heard as different, and
thus to represent a musical composition as coherent. But a concept like that
does not help to explain why we hear a musical composition that way, at least
not according to the rules of scientific explanation invoked by Babbitt. This
would entail the condition that PC sets correspond to perceptual units—that
is, to groups of notes that are heard as belonging together—whereas they
often serve as a kind of heuristic tool to find relations that the ear might have
missed. “Coherence” is not a perceptual category if the elements by which a
piece of music coheres are not directly perceived—still less if their relations
are as abstract as the PC set similarity relations discussed in chapter 5. In short,
PC set theory is not a theory of music cognition, like Lerdahl and Jackendoff’s
“generative grammar” of tonal music. As a consequence, it should not be
judged from this viewpoint. (This answers Fred Lerdahl’s criticism of the the-
ory, discussed in chapter 1.)
Nor did PC set theory provide what Richard Taruskin demanded: insight
into past compositional practices (once again, if we take scientific criteria into
45. It would be going too far to see Lerdahl’s and Jackendoff’s work as a response
to Babbitt’s methodological papers. Their theory was intended as a musical counter-
part of Chomskian linguistics, and was spurred by Leonard Bernstein’s enthusiasm for a
similar approach to music. I refer the reader to the 1973 Charles Eliot Norton Lectures
of Bernstein at Harvard University, which were published in 1976 under the title The
Unanswered Question.

270 mise-en-scène
consideration). What could positively indicate the role of the PC set as a device
of composition if we limit our inquiry to the “music itself” (which is usual, if
not a matter of principle, in music analysis)? One might think of the frequent
occurrence of a particular set-type in a work, or of harmonic progressions that
involve particular PC set relations (like the inversion relation in ex. 3.13). But
these criteria are utterly ambiguous. How frequent is “frequent”? And what are
the conditions in which these set-types occur? The PC set {1,4,5,7} was shown to
appear a number of times in the Gavotte from Schoenberg’s Suite, Op. 25 (see
ex. 2.9); but then it consists of the first four elements of a recurring twelve-tone
series: a fairly trivial case of a PC set leaving a strong mark on a piece of music.
If PC set theory claims that PC set relations form the common theoretical basis
of a body of twentieth-century art music larger than serial twelve-tone music, it
should enable us to validate that claim. This requires the consultation of sources
outside the domain of the theory, which is restricted to scores.
Seen thus, PC set theory is not an explanatory theory, but the codification
of an analytical practice. As such, it bears no substantial relation to modern sci-
ence, even though it appeals to a scientific sense of objectivity and consistency.
It is true that it incorporates what one might call a hypothesis about pitch orga-
nization (that atonal pieces are structured around groups of notes that can be
related to each other), but this “hypothesis” is not supposed to be tested—at least
not by analysts. On the contrary, the hypothesis actually serves to guide the analy-
sis. As Straus writes in one of the first chapters of his Introduction to Post-Tonal
Theory: “When we listen to or analyze music, we search for coherence. In a great
deal of post-tonal music, that coherence is assured through the use of pitch-class
sets.” (Straus 1990a, 26) Such a statement hardly invites an empirical test.
But the perhaps obvious verdict of circularity should not be made without tak-
ing into account the pedagogical orientation inherent in PC set theory. It allows
one to analyze musical works which are otherwise closed off. It has generalized
and extended structural concepts like “motif,” “chord,” and “series”—with their
legacy—so as to render them applicable to a large repertoire of music from the
twentieth century and beyond. Music analysis at large belongs more naturally to
the world of education than to the world of science; music analysis in the United
States is no different, in spite of the great attraction science has exerted upon
American music theorists. It is the principal aim of an analytical theory that it be
taught, that the insights it offers trickle down to the classrooms where new gen-
erations of musicians, scholars, and teachers are bred—much as Babbitt’s highly
involved study of twelve-tone invariants in Schoenberg’s Third String Quartet
reached a student audience through the manuals of Perle and Lester.
Book-length theories of musical structure—like the study of Lerdahl and Jack-
endoff mentioned earlier, or like William Caplin’s Classical Form (1997)—often
suggest a didactic purpose. They are set up progressively—starting with analyses
of small fragments and then proceeding to larger and more complex structures;
they have abundant musical examples; they show a deep concern with the

mise-en-scène 271
consistency and clarity of the applied terminology; and the pace of the argu-
ment is relaxed, encouraging the interested novice to persevere. If such a publi-
cation is still too much a theoretical treatise, it may be followed by another one
better suited to the classroom. Heinrich Schenker’s theory has been presented
to a student readership in books like Introduction to Schenkerian Analysis (Allen
Forte and Steven Gilbert, 1982) and Analysis of Tonal Music (Allen Cadwallader
and David Gagné, 1998). And the books of John Rahn (Basic Atonal Theory) and
Joseph Straus (Introduction to Post-Tonal Theory) provided easier access to PC set
theory than Forte’s own The Structure of Atonal Music.46
The professionalization of music theory has been described as a process of
emancipation. According to Patrick McCreless, it was in the late 1950s and early
1960s that American music theory could “stake a claim for admission to the
modern academy,” after it had thrown off “the shackles of its old pedagogical
self.” McCreless speaks about the birth of the music theorist “out of the music
theory teacher” (McCreless 1996). In reality, however, the discipline has never
truly disentangled itself from the practice of teaching. PC set theory, for exam-
ple, does not embody music theory’s emancipation from a teaching tradition;
rather, its existence is a measure of the importance, endurance, and flexibil-
ity of this tradition. The new formulations that PC set theory offered—whether
they resulted from the application of computer technology to music analysis or
from the adoption of scientific language—found their way into pedagogical con-
texts, where they were used without specific reference to the conditions that had
engendered them. They were absorbed, and transformed, by a practice with a
much longer history.
This does not mean that PC set theory’s formulations failed to affect the teach-
ing practice in the United States; they affected it greatly. They helped revital-
ize music theory instruction—a need that was very urgent in 1959, when David
Kraehenbuehl—a former student of Hindemith at Yale University, and the first
editor of Journal of Music Theory—wrote:
Theory has fallen into a state of almost universal academic disgrace. In many
schools, it is the program in which those unfortunates who have failed in every
practical musical activity are placed; in others, the hours devoted to it, obvi-
ously a waste, are quite reasonably pared down; in still others, it is maintained
as a kind of traditional house-keeping service which no one has had the cour-
age to improve or eliminate. (Kraehenbuehl 1959, 31)
46. This does not mean that Rahn and Straus merely reproduced the contents of The
Structure of Atonal Music. Rahn extended Forte’s theory (operations on pitch sets, multipli-
cation), but also deleted a considerable portion of it (set-complex theory, the intended
crown on Forte’s achievement). Moreover, his nomenclature differed from Forte’s. Straus
included the most successful elements of Forte’s theory in a general survey of analytical
concepts and techniques applied to post-tonal music.

272 mise-en-scène
And it was still urgent in 1965, to judge from Howard Boatwright. In the view
of this composer, most of the theory texts in use at teaching institutions had
“neither kept up with the present nor incorporated much of our improved
understanding of the past.” He called on those involved in the teaching of music
theory in colleges and elsewhere to “eliminate anachronistic practices . . ., to
make [theory] serve more meaningfully in developing general understanding
of our musical heritage from the distant and recent past, and to connect it more
realistically with the events of the present” (Boatwright 1965, 48).
In this period, the topic of music theory instruction was discussed in a num-
ber of forums. Periodicals such as Journal of Music Theory and College Music Sympo-
sium devoted space to these discussions.47 A thread running through them was
music theory’s failure in dealing with music of the present. Music education at
large was felt to be serving the preservation of past heritages rather than their
evolution. This was the main reason why, in 1963, the Music Educators National
Conference launched the so-called “Contemporary Music Project for Creativity
in Music Education.” Enabled by a grant of $ 1,380,000 from the Ford Founda-
tion, it investigated music programs in public schools and colleges, it organized
workshops and seminars for music educators, and it conducted experiments
with new approaches to music teaching with a view to “modernize and broaden
the quality and scope of music education on all levels.”48 Apart from a greater
focus on contemporary composition in music education—for example via the
“Composers in Public Schools Program” mentioned earlier—the Contemporary
Music Project aimed at an integration of the various fields of knowledge and
skill with which students of music had to be acquainted.
Thus, a general training in the practice of composition for all student musi-
cians was proposed. Aural training programs would draw on a greater variety of
musical materials and benefit from advances in the field of learning psychology.
Courses in music history would support the study and performance of music,
while music analysis, apart from being taught systematically in separate classes,
would inform music instruction in any class at any level. The Contemporary
Music Project lasted until 1969, but continued to inspire curricular experiments
afterwards (Wason 2002, 71). It went a long a way toward the kind of reform that
David Kraehenbuehl had proposed in 1959. Kraehenbuehl saw music theory as
an area of creative thought, interacting with the music that was currently com-
posed, played, and heard, as well as with other domains of knowledge. “Such a
47. “On the Nature and Value of Theoretical Training: A Forum,” Journal of Music The-
ory 3/1 (1959); “The Professional Music Theorist: A Forum,” Journal of Music Theory 4/1;
“Symposium: The Crisis in Music Theory Teaching,” College Music Symposium 5 (Fall 1965).
48. Quoted from the introduction to Comprehensive Musicianship: The Foundation for
College Education in Music: A Report of the Seminar sponsored by the Contemporary Music Project
at Northwestern University, April 1965. Washington: Contemporary Music Project/National
Music Educators Conference, 1965, 3.

mise-en-scène 273
dynamic theory would always be welcome in the training of any practicing musi-
cian” (Kraehenbuehl 1959, 32).
The issue of Journal of Music Theory in which these words appeared also con-
tained Allen Forte’s introduction to Heinrich Schenker and his theory, which
was cited in chapter 1 (“Schenker’s Conception of Musical Structure”). Although
Schenker’s work was already known through the teaching and writings of Adele
Katz, Felix Salzer, Roger Sessions, and Milton Babbitt, it is this article that marks
the beginning of its march through the American academy.49 Forte was 32 years
old, and about to be appointed at Yale University. His article was actually a pro-
grammatic statement, as the reader could infer from the subtitle (“A Review and
an Appraisal with Reference to Current Problems in Music Theory”). As such, it
is highly relevant to the present discussion.
In Forte’s view, one of the main attractions of Schenkerian theory was “the
discovery of a fundamental principle, which . . . [opened] the way for the disclo-
sure of further new relationships, new meanings.” (Forte 1959, 3). This funda-
mental principle was not the Ursatz—which Babbitt had given axiomatic status
seven years earlier—but the concept of structural levels. Apart from being “far
more inclusive” than the Ursatz (9), this concept lent itself to pedagogical appli-
cations. It involved a repeatable analytical procedure: the successive reductions
leading from the surface to the fundamental structure. Unlike Babbitt in 1952,
Forte stressed the fact that Schenker was a music teacher, and that “almost all his
writings are intended to instruct—in the most practical sense of that term.” He
brought the educational significance of Schenkerian theory to the notice of his
readers, claiming that it led to “an understanding of the total [musical] work in
all its complexity” (5). And he particularly addressed its relevance to the con-
cerns of practicing musicians. It would bring music theory teaching into a natu-
ral relationship to performance, a relationship of which Schenker had said:
If I had my way, every instrumentalist would have theory as his major study. It
is not enough for him to play mechanically, as though he has Czerny exercises
before him. They say they have practiced (especially the ladies). But what is
the use of that? What Geist makes their practicing vital? In painting and poetry,
Czerny exercises do not exist. (Schenker quoted by Forte 1959, 4–5)
Forte also touched on the relevance of Schenkerian theory for musicologists,

with reference to Schenker’s autograph studies and his editorship of the sonatas
of Carl Philipp Emanuel Bach and Beethoven. For Forte, Schenkerian analysis
meant the reconciliation of diverse kinds of musical activity; it meant compre-
hensive musicianship. Since this was one of the main objectives of the later Con-
temporary Music Project, it is worth exploring a bit further the relation between
49. This is acknowledged by several writers, such as Kerman (1985), Rothstein

(1990), and Snarrenberg (1994).

274 mise-en-scène
this project and the rise of Schenkerian theory in the 1960s, to the point where
the issue involves the relation between music theory as teaching subject and
music theory as an academic discipline.
Forte’s article on Schenker prefigured the Contemporary Music Proj-
ect in two ways. It not only stated that the highest degree of aesthetic
response was achieved by a combination of practical skill and theoretical
insight—following the example of Schenker’s teaching—but it also pre-
sented Schenkerian theory as based on a general model for understand-
ing all music. Thus, Forte tried to establish a line of continuity between
the music of the past and that of the present, another objective of the
Contemporary Music Project. Forte was well aware that the avowed con-
servative and doomsayer Heinrich Schenker would not have danced to
this tune. So, he picked his words carefully:
It seems contradictory that Schenker’s work should contribute significantly

to the solution of certain problems in advanced contemporary music. Yet,
his general concept of structure, apart from his specific formulations of
triadic tonal events, lends itself to modern thought regarding music. (Forte
1959, 26)
This general concept of structure was the “totally organized work” (26), in which
every detail contributed to the experience of the whole. Although at first sight
this concept does not seem to be in conflict with Schenker’s own ideas, Forte’s
rendering of an analysis by Schenker—his analysis of Robert Schumann’s song
“Aus meinen Tränen sprießen” (ex. 1.5)—has struck some readers as very dif-
ferent in tone and style from its model. William Rothstein calls Schenker’s voice
“that of a prophet, pronouncing sacred mysteries from on high,” whereas he
sees Forte as “the cool taxonomist, concerned above all with rationalism and
clarity” (Rothstein 1990, 198–99). Robert Snarrenberg (1994, 51) speaks of a
“scientistic transformation of Schenker,” which turns works of genius into pas-
sive subjects of dissection.
While all of this is to some extent true, the most radical departure from
Schenker consisted in Forte’s tendency to generalize. He suggested that Schen-
kerian theory could be subsumed under a more inclusive concept of musical
structure, a concept independent of tonality, triadic or otherwise.50 It is when he
speaks of “essential relationships” (Forte 1959, 26), or “similarities at other than
50. The more inclusive, the more insightful: this seems to have been the motto guid-
ing the attempts to conceive of musical structure in the most general terms. On the one
hand, this generalization enabled comparative studies across boundaries of style and
genre; on the other, it took the focus away from the composer and historical context (see
also chapter 7, note 11). In these two respects, Forte’s agenda can be characterized as a
genuinely structuralistic one.

mise-en-scène 275
the surface level” (27) in “advanced contemporary,” “non-triadic,” or “problem-

atic modern music” that PC set theory comes into view.51
Forte’s “general concept of structure” also appeared on the agenda of the
Contemporary Music Project—perhaps by his own agency. The documentation
of the project includes the proceedings of a Seminar on Comprehensive Musi-
cianship, which was held at Northwestern University in Evanston, Illinois, from
April 22 through 25, 1965. The purpose of this seminar was to lay down the
aims of the project with regard to college education in music. Forte was one of
the participants. At the time, he was already deeply involved in the computer-
aided investigation of the atonal music of Schoenberg, Berg, and Webern. Less
than two weeks earlier, on April 10, he had presented the first results of this
investigation at a meeting at the Rockefeller Center in New York (cf. Forte
1970). Now he was invited to deliver a position paper on “The Role of Music
Theory in the Development of Musical Understanding.” Furthermore, he was
a member of a group of participants assigned to make recommendations with
respect to training programs in music analysis and aural skills.52 Most of these
recommendations dealt with the necessity of expanding the musician’s frame
of reference (“Materials may be drawn from any era or culture”), achieving
an integration of musical skills (“the repertoire used in aural training should
be actively related to the repertoire heard in performance”), and of develop-
ing effective means of communication (“the musician should . . . have the wid-
est possible acquaintance with the various relevant formulations of systematic
music theory”). For the present discussion, however, the following brief pas-
sage is important:
A basic prerequisite for the analytical training process is a generalized

framework for the study of musical materials, which might be in the form
of a systematic description of musical structure cutting across stylistic bar-
riers and approaching music from a less narrow point of view than the now
generally current.53
51. In his article, Forte used all these labels to designate a repertoire of twentieth-
century music that he regarded as one of the five great challenges for music theorists of
his day. His brief sample analysis of Debussy’s Prelude La Cathédrale Engloutie (1910)—a
piece he apparently considered as belonging to that repertoire—still made use of “Schen-
ker-derived” techniques. Besides the analysis of this and other “problematic modern
music,” the challenges he mentioned were: “constructing a theory of rhythm for tonal
music,” “determining the sources and development of triadic tonality,” “gaining informa-
tion about compositional technique,” and “improving theory instruction.”
52. Other participants in this group included the composers Leslie Bassett (Univer-
sity of Michigan, Ann Arbor) and Robert Cogan (New England Conservatory), and the
music educator Charles Leonhard (University of Illinois, Urbana).
53. All quotes have been taken from the proceedings of the seminar (Comprehensive
Musicianship, 15–17; see note 48).

276 mise-en-scène
These words seem too abstract to have ever elicited action; however, when they
were put to paper, action was already under way. The passage can be read as an
introduction to the projects Forte had already embarked on. It stressed the need
of a generalized analytical vocabulary as a remedy to the narrow-mindedness of
teaching programs offering an inventory of past practice. We know that Forte
put little trust in the knowledge and skills handed down by tradition. In his posi-
tion paper, he had said that music theory owed regained respectability to com-
puter technology, the use of which had helped widen the scope of the discipline.
Now, his aim was to bridge the gap between advanced research in music theory
and music theory as taught in the classroom:
The implications and responsibilities of advanced research with respect to

education often are totally disregarded by persons deeply involved in such
work, just as the implications of these recent developments appear to have
been overlooked in the teaching of music theory at the classroom level. (Forte
1965b, 39)
The study of skills and techniques . . . must be informed by the highest level
of scholarship, for the task of leading to an understanding of complicated
art music requires a knowledge of the role of systematic generalization and a
comprehension of the significant characteristics of musical abstractions and of
symbolic processes in general. (Forte 1965b, 40)
Forte thus aligned his program for the development of music theory at the
college level—including Schenkerian theory and PC set theory—with a broad
education policy aiming at maximum participation in what was called “the
information revolution.” It is interesting to compare his words with those of
Ole Sand, who headed the Center for the Study of Instruction of the National
Education Association, and was invited to speak about general trends in cur-
riculum and instruction:
Probably the most immediate factor forcing change upon education is the
explosion of knowledge—the “information revolution.” Furthermore, because
scientific and scholarly work is now quite extensive and many are engaged in
it, the rate of revision is swift. Teaching the disciplines in this situation clearly
requires teaching something more permanent and pervasive than a catalog of
factual knowledge, although some facts are essential, and it is clear that there
is a continuing need for drill and repetition for learning basic information.
Educators are not only concerned with the amount of knowledge students
possess but also with students’ lack of understanding about what they presum-
ably know. Since about 1955, a vivid awareness of this latter problem has led
some scholars and researchers to explore ways of selecting, organizing, and
teaching available information to make it intelligible and more usable. In gen-
eral, the recent studies shift the balance in learning from inventory to trans-
action. The structure of a discipline, its methods of inquiry, and the styles of

mise-en-scène 277
thinking of its scholars and specialists offer important keys to this educational
task. (Sand 1965, 73)
A “generalized framework for the study of musical materials,” as recommended

by the Seminar on Comprehensive Musicianship, can be seen as a response to
the changing conditions of learning pointed out by Sand. Apart from the cul-
tural and artistic values it may have involved (Forte’s “totally organized work”),
apart from its role in retaining a sense of continuity between the past and the
present, such a framework was expected to assure the distribution of new knowl-
edge through a system of mass education. Thus, it would serve the “democracy
of learning” so firmly anchored in the American self-image. For American soci-
ety mass education is more than the product of social forces and demographic
change. Mass education entails the idea that everybody should have access to
knowledge, wield decision-making power, and be able to change their status.
Mass eduation is “the United States’ assurance that democracy may thrive,” to
borrow a word from Samuel B. Gould, once a university chancellor and propaga-
tor of educational television (Gould 1971, 355–56).
This idea points to the rise of PC set theory in the United States. PC set theory
is not only the product of the scientification of music theory, but also of a com-
mitment to education. Whoever rejects its formalism and reductionalism, and
finds its language inappropriate, should bear in mind that these aspects serve
the purpose of equal opportunity. They enable anybody who is interested, and
perseveres, to develop an expertise in some of the art music of the post-1900 era
and to hand it down to others. It may be hard to believe that the mathematical
borrowings and the more arcane relational concepts are interesting to those out-
side a highbrow elite; but whoever thinks they are, therefore, incompatible with
an idealistic notion of mass education, should be reminded that this notion does
not exclude the emergence of such elites. Rather, it allows various levels of prior
ability and acknowledges various levels of achievement. This is what European
composers (like Arnold Schoenberg, Ernst Toch, and Paul Hindemith), who
had settled in the United States during the Nazi period and assumed teaching
positions at colleges and universities, found so difficult to cope with. As Alan P.
Lessem has shown, they objected to the numbers of students they were required
to teach, and to the quantification of their achievements:
[The American democratic ideal] was . . . irreconcilable with the belief, so much
a part of the meaning of ‘culture’ in its nineteenth-century middle-class Euro-
pean milieu, that significant artistic achievement was given to only a chosen few,
from whose hands the multitude gratefully received secondary benefits. And it
was a belief to which many of the émigrés still clung. (Lessem 1988, 11)
PC set theory has never become—and may never become—a part of the stan-
dard equipment of the “musically literate”; but it is itself equipped to meet the

278 mise-en-scène
demands of mass education in music, in the sense that it provides a conceptual

framework the intrinsic logic of which could be taught to many. An answer to
the question of why PC set theory has become so vital to the image of American
music theory should not ignore this aspect; it may even have been the key to the
theory’s success. On the other hand, it may have precluded its triumphal prog-
ress in Europe, where the aristocratic notion of culture that Lessem described is
more deeply embedded than in the United States.
Finally, then, we can answer the question left over from the previous section:
how did Milton Babbitt bring music theory so successfully into the American
academy? Babbitt has not been successful in every respect. Music theory has
not, on the whole, developed into something scientifically viable—that is, from
his original empirical-theoretical viewpoint. In other words, it has not quite
secured its place among the academic disciplines on his conditions. While being
informed by an academic discourse, it has kept in touch with its “old pedagogical
self”: a discipline at the service of music education. PC set theory is one manifes-
tation of this double mindset. It testifies to Babbitt’s influence, but at the same
time it reflects the pedagogical tradition from which he had tried to cut music
theory loose; it may well be that Babbitt’s own intellectual legacy owes much to
the vitality of this tradition.

Reference List
Abbate, Carolyn. 1991. Unsung Voices: Opera and Musical Narrative in the Nineteenth Cen-
tury. Princeton, NJ: Princeton University Press.
Alexander, Charles C. 1980. Here the Country Lies: Nationalism and the Arts in 20th-Cen-
tury America. Bloomington: Indiana University Press.
Ankersmit, F. R. 1990. De navel van de geschiedenis: over interpretatie, representatie, en histo-
rische realiteit. Groningen: Historische Uitgeverij.
Antokoletz, Elliott. 1993. “Middle Period String Quartets.” In The Bartók Companion,
ed. Malcolm Gillies, 257–77. London: Faber and Faber.
Armstrong, David M. 1978. Universals and Scientific Realism. Vol. 1, Nominalism and
Realism. Cambridge: Cambridge University Press.
Ashby, Arved. 1995. “Of Modell-Typen and Reihenformen: Berg, Schoenberg, F. H. Klein,
and the Concept of Row Derivation.” Journal of the American Musicological Society
48: 67–105.
Babbitt, Milton. 1950. Review of Schoenberg et son école and Qu’est-ce que de la musique de
douze sons? by René Leibowitz. Journal of the American Musicological Society 3: 57–60.
———. 1952. Review of Structural Hearing by Felix Salzer. Journal of the American Musi-
cological Society 5: 260–65.
———. 1955. “Some Aspects of Twelve-Tone Composition.” The Score and IMA Maga-
zine 12: 53–61.
———. 1958. “Who Cares if You Listen?” High Fidelity 8/2: 38–40, 126–27.
———. 1960. “Twelve-Tone Invariants as Compositional Determinants.” Musical
Quarterly 46/2: 246–59.
———. 1961a. “Set Structure as a Compositional Determinant.” Journal of Music The-
ory 5/1: 72–94.
———. 1961b. “Past and Present Concepts of the Nature and Limits of Music.” In
International Musicological Society: Report of the Eighth Congress, New York 1961,
398–403. Kassel: Bärenreiter.
———. 1962. “Twelve-Tone Rhythmic Structure and The Electronic Medium.” Per-
spectives of New Music 1/1: 49–79.
———. 1965a “The Structure and Function of Musical Theory.” College Music Sympo-
sium 5: 49–60.
———. 1965b. “The Use of Computers in Musicological Research.” Perspectives of New
Music 3/2: 74–83.
———. 1970. “The Composer in Academia: Reflections on a Theme of Stravinsky.”
College Music Symposium 10: 63–65.
———. 1972. “Contemporary Music Composition and Music Theory as Contemporary
Intellectual History.” In Perspectives in Musicology: The Inaugural Lectures of the Ph.D.
Program in Music at the City University of New York, ed. Barry S. Brook, Edward O.
Downs, and Sherman van Solkema, 151–84. New York: W. W. Norton.

280 reference list
———. 1987. Words about Music, ed. Stephen Dembski and Joseph N. Straus. Madi-
son: University of Wisconsin Press.
———. 1988. “Paul Fromm: A Memoir.” In A Life for New Music: Selected Papers of Paul
Fromm, ed. David Gable and Christoph Wolff, 3–5. Cambridge, MA: Harvard
University Press.
———. 1989. “On Having Been and Still Being an American Composer.” Perspectives
of New Music 27/1:106–12.
———. 1992. The Function of Set Structure in the Twelve-Tone System [1946]. PhD disser-
tation, Princeton University.
———. 1999. “My Vienna Triangle at Washington Square.” In Driven into Paradise:
The Musical Migration from Nazi Germany to the United States, ed. Reinhold Brink-
mann and Christoph Wolff, 33–53. Berkeley: University of California Press.
Bacon, Ernst. 1917. “Our Musical Idiom,” with an introduction by G. D. Dunn. The
Monist 27:560–607.
Baker, James M. 1986. The Music of Alexander Scriabin. New Haven, CT: Yale University
Press.
Bartók, Béla. 1976. “Harvard Lectures” [1943]. In Béla Bartók Essays, ed. Benjamin
Suchoff. London: Faber and Faber.
Bauer-Mengelberg, Stefan, and Melvin Ferentz. 1965. “On Eleven-Interval Twelve-
Tone Rows.” Perspectives of New Music 3/2:93–103.
Bauer-Mengelberg, Stefan. 1970. “The Ford-Columbia Input Language.” In Musicol-
ogy and the Computer/ Musicology 1966–2000: A Practical Program, ed. Barry S.
Brook, 48–52. New York: City University of New York Press.
Beach, David W. 1974. “The Origins of Harmonic Analysis.” Journal of Music Theory
18/2:274–306.
———. 1979. “Pitch Structure and the Analytic Process in Atonal Music: An Interpre-
tation of the Theory of Sets.” Music Theory Spectrum 1:7–22.
Beal, Amy C. 2000. “Negotiating Cultural Allies: American Music in Darmstadt,
1946–1956.” Journal of the American Musicological Society 53/1 (2000): 105–39.
Beglarian, Grant. 1971. “Education of American Composers.” In The Encyclopedia of
Education, vol. 6, ed. Lee C. Deighton, 462–67. New York: The MacMillan Com-
pany & The Free Press.
Benjamin, William E. 1974. Review of The Structure of Atonal Music by Allen Forte.
Perspectives of New Music 13/1:170–90.
———. 1981. “Schenker’s Theory and the Future of Music.” Journal of Music Theory
25/1:155–73.
Bennighof, James M. 1984. A Theory of Harmonic Areas Defined by Pitch-Class Sets. PhD
dissertation, University of Iowa.
Berger,. Arthur. 1963. “Problems of Pitch Organization in Stravinsky.” Perspectives of
New Music 2/1:11–42.
Berkel, K. van. 1990. “De amerikanisering van de Europese universiteit: het begin
van de discussie in Nederland (1900–1940).” In Amerika in Europese ogen: facet-
ten van de Europese beeldvorming van het moderne Europa, ed. K. van Berkel, 135–
60. ‘s-Gravenhage: SDU uitgeverij.
Bernard, Jonathan W. 1997. “Chord, Collection, and Set in Twentieth-Century The-
ory.” In Music Theory in Concept and Practice, ed. James M. Baker, David W. Beach,
and Jonathan W. Bernard, 11–51. Rochester, NY: University of Rochester Press.

reference list 281
Bernhard, Christoph. 1926. Die Kompositionslehre Heinrich Schützens in der Fassung sei-
nes Schülers Christoph Bernhard, ed. Josef Maria Müller-Blattau. Leipzig: Breit-
kopf und Härtel.
Bernnat, Andreas. 2000. “Einheit und Mannigfaltigkeit: Die ‘4th European Music
Analysis Conference’ in Rotterdam.” Musik & Ästhetik 14:87–92.
Bernstein, David W. 1993. “Symmetry and Symmetrical Inversion in Turn-of-the-Century
Theory and Practice.” In Music Theory and the Exploration of the Past, ed. Christopher
Hatch and David W. Bernstein, 377–408. Chicago: University of Chicago Press.
Bernstein, Lawrence, and Joseph P. Olive. 1969. “Computers and the 16-Century
Chanson: A Pilot Project at the University of Chicago.” Computers and the
Humanities 3: 153–63.
Bloom, Allan. 1987. The Closing of the American Mind. New York: Simon & Schuster.
Boatwright, Howard. 1965. “The Crisis in Theory Teaching.” College Music Symposium
5:42–48.
Bononcini, Giovanni Maria. 1673. Musico Prattico (facsimile of the edition of 1673).
New York: Broude Brothers, 1969.
Boretz, Benjamin. 1969/1973. “Metavariations: Studies in the Foundations of Musical
Thought.” Perspectives of New Music 8/1:1–74; 8/2:49–111; 9/1:23–42; 9/2:232–
70; 11/1:146–223; 11/2:156–203.
Boulez, Pierre. 1963. Musikdenken Heute. Mainz: Schott.
Brody, Martin. 1993. “‘Music for the Masses’: Milton Babbitt’s Cold War Music The-
ory.” Musical Quarterly 77/2:.161–92.
Brook, Barry S. 1967. “RILM: Répertoire International de la Littérature Musicale,”
Computer and the Humanities 1:103–8.
Brown, Matthew, and Douglas Dempster. 1989. “The Scientific Image of Music The-
ory.” Journal of Music Theory 33/1:65–106.
Bruner, Cheryl L.1984. “The Perception of Contemporary Pitch Structures.” Music
Perception 2:25–39.
Buchler, Michael H. 1997. Relative Saturation of Subsets and Interval Cycles as a Means for
Determining Set-Class Similarity. PhD dissertation, University of Rochester.
Bukofzer, Manfred. 1957. The Place of Musicology in American Institutions of Higher
Learning. New York: The Liberal Arts Press.
Cadwallader, Allen, and David Gagné. 1998. Analysis of Tonal Music: A Schenkerian
Approach. New York: Oxford University Press.
Caplin, William E. 1998. Classical Form: A Theory of Formal Functions for the Instrumental
Music of Haydn, Mozart, and Beethoven. New York: Oxford University Press.
Castrén, Marcus. 1994. RECREL: A Similarity Measure for Set-Classes. Helsinki: Sibelius
Academy.
Chrisman, Richard. 1971. “Identification and Correlation of Pitch-Sets.” Journal of
Music Theory 15:58–83.
Christensen, Thomas. 1993a. Rameau and Musical Thought in the Enlightenment. Cam-
bridge: Cambridge University Press.
———. 1993b. “Music Theory and its Histories.” In Music Theory and the Exploration
of the Past, ed. Christopher Hatch and David W. Bernstein, 9–39. Chicago: Uni-
versity of Chicago Press.
Clough, John. 1965. “Pitch-Set Equivalence and Inclusion (A Comment on Forte’s
Theory of Set-Complexes).” Journal of Music Theory 9/1:163–71.

282 reference list
Cohn, Richard. 1996. “Maximally Smooth Cycles, Hexatonic Systems, and the Analy-
sis of Late-Romantic Triadic Progressions.” Music Analysis 15/1:9–40.
———. 1997. “Neo-Riemannian Operations, Parsimonious Trichords, and their ‘Ton-
netz’ Representations.” Journal of Music Theory 41/1:1–66.
Cone, Edward T. 1967. “Beyond Analysis.” Perspectives of New Music 6/1:33–51.
Cook, Nicholas. 1987a. A Guide to Musical Analysis. Oxford: Oxford University Press.
———. 1987b. “The Perception of Large-Scale Tonal Closure.” Music Perception
5:197–205.
———. 1989a. Music, Imagination and Culture. Oxford: Clarendon Press.
———. 1989b. “Music Theory and ‘Good Comparison’: A Viennese Perspective.”
Journal of Music Theory 33/1:117–41.
———. 1995. “Heinrich Schenker, Polemicist: A Reading of the Ninth Symphony
Monograph.” Music Analysis 14/1:89–105.
———. 1999. “Analysing Performance and Performing Analysis.” In Rethinking Music,
ed. Nicholas Cook and Marc Everist, 239–61. Oxford: Oxford University Press.
———. 2001. “Between Process and Product: Music and/as Performance.” Music The-
ory Online 7/2, http://www.societymusictheory.org/mto/issues/mto.01.7.2/
toc.7.2.html.
Cross, Jonathan. 2000. “Modernism and the Challenge to Music Theory.” In Wahrneh-
mung und Begriff: Dokumentation des Internationalen Symposiums Freiburg, 2.–3. Juni
2000, ed. Wilfried Gruhn and Hartmut Möller. Kassel: Gustav Bosse Verlag.
Dahlhaus, Carl. 1968. Untersuchungen über die Entstehung der harmonischen Tonalität.
Kassel: Bärenreiter.
———. 1970. “Was ist eine Zwölftonreihe?” Neue Zeitschrift für Musik 131:506–7.
———. 1985. “Was heisst ‘Geschichte der Musiktheorie’?” In Geschichte der Musiktheo-
rie, vol. 1, ed. Frieder Zaminer, 8–39. Darmstadt: Staatliches Institut für Musik-
forschung Preussischer Kulturbesitz, Wissenschaftliche Buchgesellschaft.
———. 1986. “Was heisst ‘Entwickelnde Variation’?” In Bericht über den 2. Kongres der
internationalen Schönberg-Gesellschaft [1984], 280–84. Vienna: Elisabeth Lafite.
Transl. as “What is ‘Developing Variation’?” by Derrick Puffett and Alfred
Clayton, in Schoenberg and the New Music: Essays by Carl Dahlhaus, 128–33. Cam-
bridge: Cambridge University Press, 1987.
Davis, James A. 1995. “Philosophical Positivism and American Atonal Music Theory.”
Journal of the History of Ideas 56/3:501–22.
Delaere, Mark. 1993. Funktionelle Atonalität: Analytische Strategien für die frei-atonale
Musik der Wiener Schule. Wilhelmshaven: Florian Noetzel Verlag.
Deliège, Célestin. 1989. “La Set-Theory ou les enjeux du pléonasme.” Analyse Musicale
17:64–79.
Doerksen, John F. 1994. A Theory of Set-Class Salience for Post-Tonal Music, with Analyses of
Selected Lieder by Anton Webern. PhD dissertation, University of Western Ontario.
———. 1998. “Set-Class Salience and Forte’s Theory of Genera.” Music Analysis
17/2:195–205.
Duckworth, William. 1999. Talking Music: Conversations with John Cage, Philip Glass,
Laurie Anderson, and Five Generations of American Experimental Composers. New
York: Da Capo Press.
Dunsby, Jonathan, and Arnold Whittall. 1988. Music Analysis in Theory and Practice.
London: Faber Music.

reference list 283
Eggebrecht, Hans Heinrich. 1975. “Opusmusik.” Schweizerische Musikzeitung 115:2–11.

Eimert, Herbert. 1924. Atonale Musiklehre. Leipzig: Breitkopf & Härtel.
———. 1950. Lehrbuch der Zwölftontechnik. Wiesbaden: Breitkopf & Härtel.
———. 1964. Grundlagen der musikalischen Reihentechnik. Vienna: Universal Edition.
Epstein, David. 1979. Beyond Orpheus: Studies in Musical Structure. Cambridge, MA:
The MIT Press.
Forte, Allen. 1959. “Schenker’s Conception of Musical Structure: A Review and an
Appraisal with Reference to Current Problems in Music Theory.” Journal of
Music Theory 3/1:1–30.
———. 1963. “Context and Continuity in an Atonal Work: A Set-Theoretic Approach.”
———. 1964. “A Theory of Set-Complexes for Music.” Journal of Music Theory 8/2:136–
83.
———. 1965a. “The Domain and Relations of Set-Complex Theory.” Journal of Music
Theory 9/1:173–80.
———. 1965b. “The Role of Music Theory in the Development of Musical Under-
standing.” In Comprehensive Musicianship: The Foundation for College Education
in Music: A Report of the Seminar sponsored by the Contemporary Music Project at
Northwestern University, April 1965. Washington: Contemporary Music Project/
National Music Educators Conference.
———. 1966. “A Program for the Analytic Reading of Scores.” Journal of Music Theory
10/2:331–64.
———. 1967a. “Computer-Implemented Analysis of Musical Structure.” In Papers from
the West Virginia University Conference on Computer Applications in Music, ed. Ger-
ald Lefkoff, 29–42. Morgantown: West Virginia University Foundation, 1967.
———. 1967b. SNOBOL3 Primer: An Introduction to the Computer Language. Cambridge,
Massachusetts: The MIT Press.
———. 1970. “The Structure of Atonal Music: Practical Aspects of a Computer-Ori-
ented Research Project.” In Musicology and the Computer/Musicology 1966–2000:
A Practical Program, ed. Barry S. Brook, 10–18. New York: City University of New
York Press.
———. 1973. The Structure of Atonal Music. New Haven, CT: Yale University Press.
———. 1978. The Harmonic Organization of The Rite of Spring. New Haven, CT: Yale
University Press.
———. 1981. “The Magical Kaleidoscope: Schoenberg’s First Atonal Masterwork,
Opus 11, Number 1.” Journal of the Arnold Schoenberg Institute 5:127–68.
———. 1985. “Pitch-Class Set Analysis Today.” Music Analysis 4/1–2:29–58.
———. 1986. “Letter to the Editor in Reply to Richard Taruskin from Allen Forte,”
Music Analysis 5/2–3, pp. 321–37.
———. 1988a. “Pitch-Class Set Genera and the Origin of Modern Harmonic Spe-
cies.” Journal of Music Theory 32/2:187–270.
———. 1988b. “New Approaches to the Linear Analysis of Music.” Journal of the Ameri-
can Musicological Society 41/2:315–48.
———. 1995. The American Popular Ballad of the Golden Era 1924–1950. Princeton, NJ:
Princeton University Press.
———. 1998a. The Atonal Music of Anton Webern. New Haven, CT: Yale University
Press.

284 reference list
———. 1998b. “Paul Hindemith’s Contribution to Music Theory in the United

States.” Journal of Music Theory 42/1:1–14.
Forte, Allen, and Steven Gilbert. 1982. Introduction to Schenkerian Analysis. New York:
W. W. Norton.
Friedmann, Michael L. 1985. “A Methodology for the Discussion of Contour: Its
Application to Schoenberg’s Music.” Journal of Music Theory 29/2:223–48.
———. 1990. Ear Training for Twentieth-Century Music. New Haven, CT: Yale University Press.
Fux, Johann Joseph. 1725. Gradus ad Parnassum (facsimile of the edition of 1725),
ed. A. Mann. Kassel: Bärenreiter, 1967.
Gadamer, Hans-Georg. 1960. Wahrheit und Methode. Tübingen: J. C. B. Mohr.
Gauss, Carl Friedrich. 1863. Disquisitiones Arithmeticae [1801]. Vol. 1 of Werke. Göttin-
gen: Königlichen Gesellschaft der Wissenschaften.
Gerads, Alida. 1997/1998. “Verborgen in een kist: Motetten in zeventiende-eeuwse
Engelse cijfernotatie.” Muziek & Wetenschap 6:239–52.
Gjerdingen, Robert. 1999. “An Experimental Music Theory?” In Rethinking Music, ed.
Nicholas Cook and Mark Everist, 161–70. Oxford: Oxford University Press.
Gloag, Kenneth. 2003. “Russian Rites: Petrushka, The Rite of Spring, and Les Noces.” In
The Cambridge Companion to Stravinsky, ed. Jonathan Cross, 79–97. Cambridge:
Cambridge University Press.
Goehr, Lydia. 1992. The Imaginary Museum of Musical Works: An Essay in the Philosophy
of Music. Oxford: Clarendon Press.
Gouk, Penelope M. 1999. Music, Science, and Natural Magic in Seventeenth-Century Eng-
land. New Haven, CT: Yale University Press.
———. 2002. “The Role of Harmonics in the Scientific Revolution.” In The Cambridge
History of Western Music Theory, ed. Thomas Christensen, 223–45. Cambridge:
Gould, Samuel B. 1971. “Universities, Intellectual Role of.” In The Encyclopedia of Edu-
cation, vol. 9, ed. Lee C. Deighton, 354–58. New York: The MacMillan Com-
pany & The Free Press.
Grant, M. J. 2001. Serial Music, Serial Aesthetics: Compositional Theory in Post-War Europe.
Cambridge: Cambridge University Press.
Greissle, Felix. 1925. “Die formalen Grundlagen des Bläserquintetts von Arnold
Schönberg.” Musikblätter des Anbruch 7:63–68.
Griffiths, Paul. 1995. Modern Music and After: Directions since 1945. Oxford: Oxford
University Press.
Grout, Donald Jay. 1980. A History of Western Music, 3rd ed. with Claude V. Palisca.
London: J. M. Dent.
Guilbault, Serge. 1983. How New York Stole the Idea of Modern Art: Abstract Expressionism,
Freedom, and the Cold War, trans. Arthur Goldhammer. Chicago: University of
Chicago Press.
Hába, Alois. 1927. Neue Harmonielehre des diatonischen-, chromatischen-, Viertel-, Drittel-
, Sechstel-, und Zwölftel-tonsystems (facsimile of the edition of 1927). München:
Universal Edition, 1978.
Haimo, Ethan. 1996. “Atonality, Analysis, and the Intentional Fallacy.” Music Theory
Spectrum 18/2:167–99.
Hanson, Howard. 1960. The Harmonic Materials in Twentieth-Century Music. New York:
Appleton Century Crofts.

reference list 285
Hart, Philip. 1973. Orpheus in the New World: The Symphony Orchestra as an American
Cultural Institution. New York: W. W. Norton.
Hasty, Christopher. 1981. “Segmentation and Process in Post-Tonal Music.” Music
Theory Spectrum 3/1:54–73.
Hauer, Josef Matthias. 1925. Vom Melos zur Pauke. Vienna: Universal Edition.
———. 1926. Zwölftontechnik. Vienna: Universal Edition.
Heinemann, Stephen. 1998. “Pitch-Class Set Multiplication in Theory and Practice.”
Music Theory Spectrum 20/1:72–96.
Hermann, Richard. 1993. A General Measurement for Similarity Relations: A Heuristic for
Constructing or Evaluating Aspects of Possible Musical Grammars. PhD dissertation,
University of Rochester.
Hindemith, Paul. 1937. Unterweisung im Tonsatz: Theoretischer Teil. Mainz: Schott.
Hinton, Stephen. 1988. “Musikwissenschaft und Musiktheorie oder die Frage nach
der Phänomenologischen Jungfräulichkeit.” Musiktheorie 3/3:195–204.
Hohenberg, John. 1974. The Pulitzer Prizes: A History of the Awards in Books, Drama,
Music, and Journalism Based on the Private Files Over Six Decades. New York: Colum-
bia University Press.
Howe, Hubert S. 1965. “Some Combinational Properties of Pitch Structures.” Perspec-
tives of New Music 4/1:45–61.
Howard, Walther. 1932. Wissenschäftliche Harmonielehre des Künstlers. Berlin: Verlag für
Kultur und Kunst.
Husmann, Heinrich. 1953. Vom Wesen der Konsonanz. Heidelberg: Müller-Thiergarten
Verlag.
Isaacson, Eric J. 1990. “Similarity of Interval-Class Content between Pitch-Class Sets:
The IcVSIM Relation.” Journal of Music Theory 34/1:1–28.
———. 1996. “Issues in the Study of Similarity in Atonal Music.” Music Theory Online
2/7, http://www.societymusictheory.org/mto/issues/mto.96.2.7/toc.2.7.html.
Jelinek, Hanns.1952. Anleitung zur Zwölftonkomposition. Vienna: Universal Edition.
———. 1961. “Die krebsgleichen Allintervalreihen.” Archiv für Musikwissenschaft 18
(1961): 115–25.
Jencks, Christopher, and David Riesman. 1968. The Academic Revolution. Garden City,
NY: Doubleday.
Kassler, Michael. 1961. The Decision of Arnold Schoenberg’s Twelve-Note-Class System and
Related Systems. Princeton, NJ: Princeton University Press.
———. 1966. “Toward Musical Information Retrieval.” Perspectives of New Music
4/2:59–67.
Kerman, Joseph. 1980. “How We Got into Analysis, and How to Get Out.” Critical
Inquiry 7/2:311–31.
———. 1983. “A Few Canonic Variations.” Critical Inquiry 10/1:107–26.
———. 1985. Contemplating Music: Challenges to Musicology. Cambridge, MA: Harvard
University Press.
———. 1991. “American Musicology in the 1990s.” Journal of Musicology 9/2:131–44.
Kerr, Clark. 1995. The Uses of the University, 4th. ed. Cambridge, MA: Harvard Univer-
sity Press.
Klein, Fritz Heinrich. 1925. “Die Grenze der Halbtonwelt.” Die Musik 17/4:281–86.
Klir, George J., and Tina A. Folger. 1988. Fuzzy Sets, Uncertainty, and Information. Eng-
lewood Cliffs, NJ: Prentice Hall.

286 reference list
Koch, Heinrich Christoph. 1782/1793. Versuch einer Anleitung zur Komposition. Leipzig:
A. F. Boehme. Sections 3 and 4 trans. as Introductory Essay on Composition: The
Mechanical Rules of Melody by Nancy Kovaleff Baker. New Haven, CT: Yale Uni-
versity Press, 1983.
———. 1865. Heinrich Christoph Kochs Musikalisches Lexikon: zweite durchaus umgearbei-
tete und vermehrte Auflage von Arrey von Dommer. Heidelberg: Mohr.
Komar, Arthur. 1971. “The Music of Dichterliebe: The Whole and its Parts.” In Schu-
mann: Dichterliebe, ed. Arthur Komar. New York: W. W. Norton.
Kopp, David. 2002. Chromatic Transformations in Nineteenth-Century Music. Cambridge:
Korsyn, Kevin. 1991. “Towards A New Poetics of Musical Influence.” Music Analysis
10/1:3–72.
Kostka, Stefan, and Dorothy Payne. 1984. Tonal Harmony: With an Introduction to Twen-
tieth-Century Music. New York: McGraw-Hill.
Kraehenbuehl, David. 1959. [Introduction to] “On the Nature and Value of Theo-
retical Training: A Forum.” Journal of Music Theory 3/1:31–32.
———. 1960. “The Professional Music Theorist—His Habits and Training.” Journal of
Krenek, Ernst. 1937. Über neue Musik: sechs Vorlesungen zur Einführung in die theoretische
Grundlagen. Vienna: Ringbuchhandlung.
———. 1940. Studies in Counterpoint. New York: Schirmer.
Kuhn, Thomas S. 1962. The Structure of Scientific Revolutions. Chicago: University of
Chicago Press.
Kurth, Ernst. 1923. Romantische Harmonik und ihre Krise in Wagners “Tristan” (facsimile
of the edition of 1923). Hildesheim: Georg Olms, 1975.
Lansky, Paul, and George Perle. 1980. “Interval Class.” In The New Grove Dictionary of
Music and Musicians, vol 9, ed. Stanley Sadie. London: Macmillan.
Larson, Steve. 1998. “Schenkerian Analysis of Modern Jazz: Questions about Method.”
Music Theory Spectrum 20/2:209–41.
LaRue, Jan. 1967. “Two Problems in Music Analysis: The Computer Lends a Hand.”
In Computers in Humanistic Research: Readings and Perspectives, ed. Edmund A.
Bowles, 194–203. Englewood Cliffs, NJ: Prentice Hall.
———. 1970. “New Directions for Style Analysis.” In Musicology and the Computer/
Musicology 1966–2000: A Practical Program, ed. Barry S, Brook, 194–97. New
York: City University of New York Press.
Laudan, Larry. 1977. Progress and its Problems: Towards a Theory of Scientific Growth. Ber-
keley: University of California Press.
Leeuw, Ton de. 1977. Muziek van de twintigste eeuw: een onderzoek naar haar elementen en
structuur, 3rd ed. Utrecht: Bohn, Scheltema & Holkema.
Leibowitz, René. 1947. Schönberg et son école: L’étape contemporaine du langage musical.
Paris: J. B. Janin.
Leichtentritt, Hugo. 1927. Musikalische Formenlehre, 3rd ed. Leipzig: Breitkopf & Härtel.
———. 1928. “Schönberg and Tonality.” Modern Music 5/4:3–10.
Lejeune Dirichlet, P. G. 1871. Vorlesungen über Zahlentheorie, ed. by Richard Dedekind.
Braunschweig: Friedrich Vieweg und Sohn.
Lerdahl, Fred. 1989. “Atonal Prolongational Structure.” Contemporary Music Review
4:65–87.

reference list 287
———. 2001. Tonal Pitch Space. New York: Oxford University Press.
Lerdahl, Fred, and Ray Jackendoff. 1983. A Generative Theory of Tonal Music. Cam-
bridge, MA: The MIT Press.
Lessem, Alan P. 1988. “Teaching Americans Music: Some Émigré Composer View-
points, ca. 1930–1955.” Journal of the Arnold Schoenberg Institute 11/1:4–22.
Lester, Joel. 1989. Analytic Approaches to Twentieth-Century Music. New York: W. W. Norton.
———. 1992. Compositional Theory in the Eighteenth Century. Cambridge, MA: Harvard
University Press.
Lewin, David. 1959. “Intervallic Relations between Two Collections of Notes.” Journal
of Music Theory 3/2: 298–301.
———. 1960. “The Intervallic Content of a Collection of Notes, Intervallic Relations
between a Collection of Notes and its Complement: An Application to Schoen-
berg’s Hexachordal Pieces.” Journal of Music Theory 4/1:98–101.
———. 1969. “Behind the Beyond: A Response to Edward T. Cone.” Perspectives of
New Music 7/2:59–69.
———. 1976. “On Partial Ordering.” Perspectives of New Music 14/2:252–57.
———. 1977a. “A Label-Free Development for 12-Pitch-Class Systems.” Journal of
———. 1977b. “Forte’s Interval Vector, My Interval Function, and Regener’s Com-
mon-Note Function.” Journal of Music Theory 21/2:194–237.
———. 1980. “A Response to a Response: On PCSet Relatedness.” Perspectives of New
Music 18/2: 498–502.
———. 1986. “Music Theory, Phenomenology, and Modes of Perception.” Music Per-
ception 3/4:327–92.
———. 1987. Generalized Musical Intervals and Transformations. New Haven, CT: Yale
University Press.
———. 1993. Musical Form and Transformation: Four Analytic Essays. New Haven, CT:
Yale University Press.
Lieb, Irwin C. 1966. “The ACLS Program for Computer Studies in the Humani-
ties: Notes on Computers and the Humanities.” Computers and the Humanities
1:7–11.
Lincoln, Harry B. 1967. “Musicology and the Computer: The Thematic Index.”
In Computers in Humanistic Research: Readings and Perspectives, ed. Edmund A.
Bowles, 184–93. Englewood Cliffs, NJ: Prentice Hall.
Lockwood, Lewis. 1970. “A Stylistic Investigation of the Masses of Josquin Desprez
with the Aid of the Computer: A Progress Report.” In Musicology and the Com-
puter/Musicology 1966–2000: A Practical Program, ed. Barry S, Brook, 19–27. New
York: City University of New York Press.
Loeb, David. 1976. “An Analytic Study of Japanese Koto Music.” Music Forum IV:335–95.
Loquin, Anatole. 1874. Tableau de tous les effets harmoniques de une à cinq notes inclusive-
ment. Bordeaux: Féret et fils.
———. 1895. L’harmonic rendue claire et misc à la la portée de tous les musiciens. Paris:
Richault.
Lord, Charles H. 1981. “Intervallic Similarity Relations in Atonal Set Analysis.” Jour-
nal of Music Theory 25/1:91–111.
Lowinsky, Edward E. 1963. “On the Matter of a Doctor’s Degree for Composers.” Col-
lege Music Symposium 3:43–52.

288 reference list
Maas, Hans. 2000. “Fourth European Music Analysis Conference.” Tijdschrift voor
Muziektheorie 5/1:40–42.
McCreless, Patrick P. 1996. “Contemporary Music Theory and the New Musicology:
An Introduction.” Music Theory Online 2/2, http://www.societymusictheory.
org/mto/issues/mto.96.2.2/toc.2.2.html.
———. 1997. “Rethinking Contemporary Music Theory.” In Keeping Score: Music, Discipli-
narity, Culture, ed. David Schwarz, 13–53. Charlottesville: University of Virginia Press.
Mann, Alfred. 1987. The Study of Fugue [1958]. New York: Dover.
Marpurg, Friedrich Wilhelm. 1753. Abhandlung von der Fuge nach den Grundsätzen und
Exempeln der besten Deutschen und ausländischen Meister. Berlin: Haude und Spener.
Martino, Donald. 1961. “The Source Set and its Aggregate Formations.” Journal of
Mendel, Arthur. 1963. “The Doctorate in Composition.” College Music Symposium 3:53–59.
———. 1969. “Some Preliminary Attempts at Computer-Assisted Style-Analysis in
Music.” Computers and the Humanities 4:41–52.
———. 1976. “Towards Objective Criteria for Establishing Chronology and Authen-
ticity: What Help Can the Computer Give?” In Josquin des Prez: Proceedings of the
International Josquin Festival-Conference held at the Juilliard School at Lincoln Center
in New York City, 21–25 June 1971, ed. Edward E. Lowinsky, in collaboration with
Bonnie J. Blackburn, 297–308. London: Oxford University Press.
Menke, Johannes. 2005. “Harmonielehren ‘nach’ Hugo Riemann.” In Musiktheorie,
ed. Helga de la Motte-Haber and Oliver Schwab-Felish (Handbuch der syste-
matischen Musikwissenschaft 2), 263–80. Laaber: Laber-Verlag.
Messiaen, Olivier. 1944. Technique de mon langage musical. Paris: Alphonse Leduc.
Morris, Robert D. 1980. “A Similarity Index for Pitch-Class Sets.” Perspectives of New
Music 18/2:445–60.
———. 1987. Composition with Pitch-Classes: A Theory of Compositional Design. New
Haven, CT: Yale University Press.
———. 1991. Class Notes for Atonal Music Theory. Lebanon, NH: Frog Peak Music.
———. 1993. “New Directions in the Theory and Analysis of Musical Contour.” Music
Theory Spectrum 15/2:205–28.
———. 1995. “Equivalence and Similarity in Pitch and their Interaction with Pcset
Theory.” Journal of Music Theory 39/2:207–43.
———. 1997. “K, Kh, and Beyond.” In Music Theory in Concept and Practice, ed. James
M. Baker, David W. Beach, and Jonathan W. Bernard, 275–306. Rochester, NY:
University of Rochester Press.
Neff, Severine. 1991. “An American Precursor of Non-Tonal Theory: Ernst Bacon
(1898–1990).” Current Musicology 48:5–26.
Newlin, Dika. 1980. Schoenberg Remembered: Diaries and Recollections (1938–76). New
York: Pendragon Press.
Nolan, Catherine. 2003. “Combinatorial Space in Nineteenth- and Early Twentieth-
Century Music Theory.” Music Theory Spectrum 25/2:205–41.
Oettingen, Athur von. 1913. Das duale Harmoniesystem [1866]. Leipzig: C. F. W. Siegel.
Parks, Richard S. 1981. “Pitch Organization in Debussy: Unordered Sets in Brouil-
lards.” Music Theory Spectrum 2:119–34.
———. 1989. The Music of Claude Debussy. New Haven, CT: Yale University Press.
Perle, George. 1962. Serial Composition and Atonality. London: Faber and Faber.

reference list 289
———. 1977. Twelve-Tone Tonality. Berkeley: University of California Press.

———. 1990. “Pitch-Class Set Analysis: An Evaluation.” Journal of Musicology 8/2:151–72.
Pomeroy, Boyd. 2003. “Debussy’s Tonality: A Formal Perspective.” In The Cambridge
Companion to Debussy, ed. Simon Trezise, 155–78. Cambridge: Cambridge Uni-
versity Press.
Pousseur, Henri. 1972. “Stravinsky by Way of Webern: The Consistency of a Syntax.”
Perspectives of New Music 10/2:13–51; 11/1:112–45.
Quinn, Ian. 2001. “Listening to Similarity Relations,” Perspectives of New Music 39/2: 108–58.
Rahn, John. 1979a. “Aspects of Musical Explanation.” Perspectives of New Music
17/2:204–24.
———. 1979b. “Logic, Set Theory, Music Theory.” College Music Symposium 19:114–27.
———. 1980a. Basic Atonal Theory. New York: Longman.
———. 1980b. “Relating Sets.” Perspectives of New Music 18/2:483–98.
Rameau, Jean-Philippe. 1722. Traité de l’Harmonie Reduite à ses Principes Naturels. Paris:
Ballard. Trans. as Treatise on Harmony by Philip Gosset. New York: Dover, 1971.
Ramirez, Francisco O. 2002. “Eyes Wide Shut: University, State and Society.” European
Educational Research Journal 1/2: 256–73.
Randall, J. K. 1965. “A Report from Princeton.” Perspectives of New Music 3/2:84–92.
Ratz, Erwin. 1951. Einführung in die musikalische Formenlehre: über Formprinzipien in den
Inventionen und Fugen J.S. Bachs und ihre Bedeutung für die Kompositionstechnik
Beethovens. Vienna: Österreichischer Bundesverlag.
Regener, Eric. 1974. “On Allen Forte’s Theory of Chords.” Perspectives of New Music
13/1:191–212.
Reich, Willi. 1936. “Alban Berg’s Lulu.” Musical Quarterly 22/4:383–401.
Reicha, Antoine. 1816. Cours de Composition Musicale, ou Traité Complet et Raisonné
d’Harmonie Pratique. Paris: Gambaro.
Réti, Rudolph. 1961. The Thematic Process in Music [1951]. London: Faber and Faber.
Riemann, Hugo. 1880. Skizze einer neuen Methode der Harmonielehre. Leipzig: Breitkopf
& Härtel.
———. 1914/1915. “Ideen zu einer ‘Lehre von den Tonvorstellungen,’” Jahrbuch der
Musikbibliothek Peters 21/22:1–26.
Risenhoover, Morris, and Robert T. Blackburn, 1976. Artists as Professors: Conversations
with Musicians, Painters, Sculptors. Urbana: University of Illinois Press.
Rivera, Benito V. 1984. “The Seventeenth-Century Theory of Triadic Generation and
Invertibility and its Application in Contemporaneous Rules of Composition.”
Music Theory Spectrum 6:63–78.
Robison, Tobias D. 1967. “IML-MIR: A Data-Processing System for the Analysis of
Music.” In Elektronische Datenverarbeitung in der Musikwissenschaft, ed, 103–35.
Harald Heckmannn. Regensburg: Gustav Bosse.
Rodriguez-Pereyra, Gonzalo. 2000. “What is the Problem of Universals?” Mind
109/434: 255–73.
Rogers, David W. 1999. “A Geometric Approach to Pcset Similarity.” Perspectives of
New Music 37/1:77–90.
Rothstein, William. 1990. “The Americanization of Heinrich Schenker.” In Schenker
Studies, ed. Hedi Siegel, 193–203. Cambridge: Cambridge University Press.
Rufer, Josef. 1952. Die Komposition mit zwölf Tönen. Berlin: Max Hesse.
Salzer, Felix. 1952. Structural Hearing: Tonal Coherence in Music. New York: Charles Boni.

290 reference list
Sand, Ole. 1965. “Current Trends in Curriculum and Instruction.” In Comprehensive Musi-
cianship: The Foundation for College Education in Music: A Report of the Seminar Sponsored
by the Contemporary Music Project at Northwestern University, April 1965. Washington:
Contemporary Music Project/National Music Educators Conference.
Schachter, Carl. 1983. “Analysis by Key: Another Look at Modulation.” Music Analysis
6/3:289–318.
Schaffer, Sarah. 1991. Analytical Issues in the Segmentation of Atonal Music: An Investiga-
tion Based on Selected Pre-Serial Works of Schoenberg, Berg, and Webern. PhD disserta-
tion, Indiana University.
Schat, Peter. 1998. “Splijt de leeuwerik! Muziek in het tijdperk van de herverlich-
ting.” Tijdschrift voor Muziektheorie 3/1:40–44.
Schillinger, Joseph. 1940. Kaleidophone: New Resources of Melody and Harmony. New
York: M. Witmark & Sons.
———. 1946. The Schillinger System of Musical Composition. New York: Carl Fischer.
Schmalfeldt, Janet. 1983. Berg’s Wozzeck: Harmonic Language and Dramatic Design. New
Haven, CT: Yale University Press.
Schmidt, Christian Martin. 1988. Schönberg’s Oper “Moses und Aron”: Analyse der diaste-
matischen, formalen und musikdramatischen Komposition. Mainz: Schott.
Schoenberg, Arnold. 1965. “Analysis of the Four Orchestral Songs Opus 22” [1932].
———. 1967. Fundamentals of Musical Composition, ed. Gerald Strang. London: Faber
and Faber.
———. 1969. Structural Functions of Harmony [1954], ed. L. Stein. New York: W. W.
Norton.
———. 1984. Style and Idea: Selected Writings of Arnold Schoenberg [1950], ed. L. Stein.
London: Faber and Faber.
Schuijer, Michiel. 1996. “Leonhard Deutsch and Schoenberg’s Harmony.” Tijdschrift
voor Muziektheorie 3/2:81–91; 3/3:190–207.
Scruton, Roger. 1997. The Aesthetics of Music. Oxford: Clarendon Press.
Seeger, Charles. 1969. [Foreword to] Studies in Musicology: Essays in the History, Style
and Bibliography of Music, ed. James W. Pruett. Chapel Hill: University of North
Carolina Press.
Shreffler, Anne C. 2000. “The Myth of Empirical Historiography: A Response to
Joseph N. Straus.” Musical Quarterly 84/1: 5–29.
Simms, Bryan R. 2000. The Atonal Music of Arnold Schoenberg 1908–1923. New York:
Oxford University Press.
Smith, P. H. 1967. “The Computer and the Humanist.” In Computers in Humanistic
Research: Readings and Perspectives, ed. Edmund A. Bowles, 16–28. Englewood
Cliffs, NJ: Prentice Hall.
Smith, Russell F. W. 1963. “A Community of Artists and Scholars.” Arts in Society
2/3:66–74.
Snarrenberg, Robert. 1994. “Competing Myths: The American Abandonment of
Schenker’s Organicism.” In Theory, Analysis and Meaning in Music, ed. Anthony
Pople, 29–56. Cambridge: Cambridge University Press.
Sontag, Susan. 1980. Under the Sign of Saturn. New York: Farrar, Straus & Giroux.
Starr, Daniel, and Robert Morris. 1977/1978. “A General Theory of Combinatoriality
and the Aggregate.” Perspectives of New Music 16/1:3–35, and 16/2:50–84.

reference list 291
Stein, Erwin. 1925. “Neue Formprinzipien” [1924]. In Von neuer Musik: Beiträge zur
Erkenntnis der neuzeitlichen Tonkunst, ed. H. Grues, E. Kruttge and E. Thalmeier,
59–77. Cologne: Marcan. Trans. as “New Formal Principles” by Hans Keller, in
Orpheus in New Guises, 57–85. London: Rockliff, 1953.
Stein, Leon. 1979. Structure and Style: The Study and Analysis of Musical Forms. Miami:
Summy-Birchard Inc.
Stockfelt, Ola. 1993. “Adequate Modes of Listening.” Trans. by Anahid Kassabian and
Leo G. Svendsen, Stanford Humanities Review 3/2:153–69.
Stockhausen, Karlheinz. 1963. Texte zur elektronischen und instrumentalen Musik.
Cologne: M. Dumont Schauberg.
Straus, Joseph N. 1990a. Introduction to Post-Tonal Theory. Englewood Cliffs, NJ: Pren-
tice Hall.
———. 1990b. Remaking the Past: Musical Modernism and the Influence of the Tonal Tradi-
tion Cambridge, MA: Harvard University Press.
———. 1997. “Voice Leading in Atonal Music.” In Music Theory in Concept and Practice,
ed. James M. Baker, David W. Beach, and Jonathan W. Bernard, 237–74. Roch-
ester, NY: University of Rochester Press.
———. 1999. “The Myth of Serial “Tyranny” in the 1950s and 1960s.” Musical Quar-
terly 83/3:301–43.
———. 2000. Introduction to Post-Tonal Theory, 2nd. ed. Englewood Cliffs, NJ: Prentice
Hall.
Stravinsky, Igor, and Robert Craft. 1959. Conversations with Igor Stravinsky. London:
Faber and Faber.
———. 1962. Expositions and Developments. New York: Doubleday.
Street, Alan. 1990. “Conference Report: First European Congress of Music Analysis:
Colmar, 26–8 October 1989.” Music Analysis 9/3:357–60.
Tarski, Alfred. 1995. Introduction to Logic and to the Methodology of Deductive Sciences.
Trans. Olaf Helmer. New York, Dover, 1995.
Taruskin, Richard. 1979. Review of The Harmonic Organization of The Rite of Spring by
Allen Forte. Current Musicology 28:114–29.
———. 1985. “Chernomor to Kashchei: Harmonic Sorcery; Or, Stravinsky’s Angle.”
Journal of the American Musicological Society 38:72–142.
———. 1982. “On Letting the Music Speak for Itself.” Journal of Musicology 1/3:338–
49. Repr. in Taruskin 1995.
———. 1986. “Letter to the Editor from Richard Taruskin.” Music Analysis 5/2–
3:313–20.
———. 1989. “Reply to Brown and Dempster.” Journal of Music Theory 33/1:155–64.
———. 1995. Text & Act: Essays on Music and Performance. New York: Oxford Univer-
sity Press.
———. 1996. Stravinsky and the Russian Traditions: A Biography of the Works through
Mavra. 2 vols. Oxford: Oxford University Press.
Tawa, Nicholas. 1995. American Composers and Their Public: A Critical Look. Metuchen,
Metuchen, NJ: Scarecrow Press.
Teitelbaum, Richard. 1965. “Intervallic Relations in Atonal Music.” Journal of Music
Theory 9/1:72–127.
Treitler, Leo. 1980. “History, Criticism, and Beethoven’s Ninth Symphony.” 19th-Century
Music 3/3:193–210.

292 reference list
———. 1982. “‘To Worship that Celestial Sound’: Motives for Analysis.” Journal of
Musicology 1/2:153–70.
Van den Toorn, Pieter C. 1983. The Music of Igor Stravinsky. New Haven, CT: Yale Uni-
versity Press.
———. 1995. Music, Politics, and the Academy. Berkeley: University of California Press.
Varèse, Edgar. 1967. “The Liberation of Sound.” In Contemporary Composers on Contem-
porary Music, ed. Elliott Schwarz and Barney Childs, 196–208. New York: Holt,
Rinehart, and Winston.
Vicentino, Nicola. 1555. L’Antica Musica ridotta alla moderna prattica (facsimile of the
edition of 1555), ed. E. Lowinsky. Kassel: Bärenreiter, 1959.
Vogler, Georg Joseph. 1776. Tonwissenschaft und Tonsetzkunst. Mannheim: Kuhrfürst-
liche Hofdruckerei.
Wason, Robert W. 1985. Viennese Harmonic Theory from Albrechtsberger to Schenker and
Schoenberg. Ann Arbor: UMI Research Press.
———. 1988. “Progressive Harmonic Theory in the Mid-Nineteenth Century.” Jour-
nal of Musicological Research 8:55–90.
———. 2002. “Musica Practica: Music Theory as Pedagogy.” In The Cambridge History of
Western Music Theory, ed. Thomas Christensen, 46–77. Cambridge: Cambridge
University Press.
Weber, Gottfried. 1817. Versuch einer geordneten Theorie der Tonsetzkunst zum Selbstunter-
richt mit Anmerkungen für Gelehrten, vol. 1. Mainz: Schott.
Weber, William. 1999. “The History of Musical Canon.” In Rethinking Music, ed. Nich-
olas Cook and Marc Everist, 336–55. Oxford: Oxford University Press.
Webern, Anton. 1960. Wege zur neuen Musik, ed. Willi Reich. Vienna: Universal Edition.
Weigl, Bruno. 1925. Harmonielehre. Mainz: Schott.
Wilcox, Howard J. 1983. “Group Tables and the Generalized Hexachord Theorem.”
Winham, Godfrey. 1964. “Henry Weinberg, Three Songs (1959).” Perspectives of New
Music 2/2:106–11.
———. 1970. “Composition with Arrays.” Perspectives of New Music 9/1:43–67.
Wiprud, Theodore. 2000. “On the Money: New Music Funding in the United States.”
NewMusicBox 2/4 (2000), http://www.newmusicbox.org/page.nmbx?id=804.
Wittlich, Gary. 1974. “Interval Set Structure in Schoenberg’s Opus 11, No. 1.” Perspec-
tives of New Music 13/1:41–55.
Zarlino, Gioseffo. 1558. Le Istitutioni harmoniche (facsimile of the edition of 1558).
New York: Broude Brothers, 1965.

Index
An italicized page number indicates a figure or table.
Abbate, Carolyn, 23n18, 223 Armstrong, David M., 190

absolute pitch-interval class (APIC), 48, Ashby, Arved, 116n28
82–83, 99, 151, 175, 191n13; definition “Aus meinen Tränen sprießen” (Schu-
of, 40. See also interval class, Forte’s mann, Dichterliebe, Op. 48, no. 2), 14,
definition of 16, 274
absolute value, 40
Adorno, Theodor W., 255n21 Babbitt, Milton, 9–10, 18, 26, 31, 36,
aggregate, twelve-tone, 92 43, 44n17, 47, 58, 92, 94–97, 98,
Albert, A. A., 86 105, 116, 145n12, 206, 225, 236,
algebra, 84, 86 239n4, 278; Carnap’s influence on,
algorithm, 117–18, 102, 105, 119, 120, 132, 251n17; on contemporary music cul-
240, 248. See also APIC vector; normal ture in the USA, 257–59, 261; group
form; normal order; prime form multiplication table, 51; and the lan-
all-interval series, 116, 242 guage of science, 35, 251–58, 264,
American Composers Orchestra, 259 268–69, 278; on Mersenne, 254;
American Council of Learned Societies metatheory of, 256; on Rameau,
(ACLS), 242 254; relation to his European con-
American Music Center, 259 temporaries, 261–64; on Schenker-
American Musicological Society (AMS), ian theory, 221, 252, 273. See also
239n6 combinatoriality; communality and
American Society of University Compos- contextuality; equivalence; integers;
ers, 258–59 inversion; normal form; partial
American Symphony Orchestra League, ordering; positivism; Princeton
259 University; retrograde; retrograde-
Analytic Approaches to Twentieth-Century inversion; transposition (T); twelve-
Music (Lester 1989), 20, 22n17, 252, tone serialism; university
270 Babbitt, Milton, works by: The Func-
Ankersmit, Frank, 236n1 tion of Set Structure in the Twelve-Tone
APIC. See Absolute pitch-interval class System (published 1992), 33, 55, 67,
APIC vector, 48, 52, 62, 88, 98n18, 87, 96n14; “Set Structure as Compo-
99, 108, 125–26; algorithm, 132; as sitional Determinant” (1961), 88, 102;
a basis of equivalence, 92–93, 103; “Twelve-Tone Invariants as Composi-
definition of, 48; similarity measures tional Determinants” (1960), 55, 69,
based on the, 132 144–46, 148–49, 88, 252–53, 270; “Who Cares if You
151–155, 160n21, 163, 165, 168–69, Listen?”(1958), 255n22, 264.
173–175. See also Z-relation Bach, Carl Philipp Emanuel, 273

294 index
Bach, Johann Sebastian, 66, 202–3, 205, Boulez, Pierre, 18, 78n26, 261–64, 268
213, 219, 233; Art of Fugue, The, 65, 66; Bowles, Edmund A., 241
Nun lob, mein Seel, den Herren, BWV Brahms, Johannes, 222, 233n13
17/7, 202, 203, 205, 213; Well-Tempered Braythwaite, William, 31–32
Clavier, The, 65, 66 Broadcast Music, Inc., 259
Bacon, Ernst, 33, 43, 117, 122 Brody, Martin, 255n2, 256
Baker, James M., 2n2, 45 Brook, Barry S., 242
Bartók, Bela, 78, 233; Music for Strings, Brown, Matthew, 219, 269
Percussion, and Celesta, 78, 79; String Buchler, Michael H., 176
Quartet, No. 3, 78. See also interval Bukofzer, Manfred, 239n6
expansion; multiplication (M).
Baryphonus, Henricus, 30 Cadwallader, Allen, 271
“basic set,” 44n16, 68, 94, 96n14, 105 canonical transformation, 54, 113–14,
“basic shape,” 10, 12 161–62, 163, 165–67, 169, 172, 185n5.
bass position, 41, 100, 121. See also har- canonical transformation group. See
monic inversion transformation group
basse fondamentale (Rameau). See funda- Cantor, Georg, 3n3
mental bass Capellen, Georg, 67n16
Bauer-Mengelberg, Stefan, 116n29, 242– Caplin, William E., 16n13, 270
43, 244n11. See also all-interval series; cardinal number, 41, 45–46, 89, 97, 100,
Ford-Columbia (input language) 158, 167, 173, 179, 232
Beach, David W., 2n2, 32, 229 cardinality. See cardinal number
Beal, Amy C., 263 Carnap, Rudolf, 250, 251n17
Beethoven, Ludwig van, 222, 223–24, Carter, Elliott, 18, 263n35
233, 234, 273 Castrén, Marcus, 176
Benjamin, William E., 3, 24. See also seg- Chopin, Frédéric, 233n13
mentation chord class, 116–18, 120
Berg, Alban, 78–79, 91, 116, 126–28, 152, chord classification, 43, 115, 121, 123;
212n30, 239; Five Orchestral Songs after Hindemith’s, 138, 140–41, 142;
Picture Postcard Texts by Peter Altenberg, Klein’s, 116–18; Krenek’s, 138, 142;
Op. 4, 78 (no. 1), 126, 127, 128 (no. Weigl’s, 120–21
2); Lulu, 79, 80 Chrisman, Richard, 91n10, 100–102, 122.
Berger, Arthur, 200–2 See also equivalence; successive inter-
Berio, Luciano, 75–76, 225 val array
Bernard, Jonathan W., 2n2, 33, 43, Christensen, Thomas, 137, 219–20
116n27, 123 CINT. See cyclic interval succession
Bernhard, Christoph, 53 class. See equivalence class
Bernstein, David W., 67n16 Clough, John, 91n10, 98–99, 105, 107. See
Bernstein, Lawrence, 242 also equivalence
Bernstein, Leonard, 244, 269n45 Cohn, Richard, 61, 195n16
Birkhoff, George D., 86 Columbia University, 17, 18, 19, 265, 267
Bloom, Allan, 265 combinatoriality, 26, 83, 92–94, 95, 97,
Bloom, Harold, 235 143, 145n12, 205–6; hexachordal
BMI. See Broadcast Music, Inc. all-, 94, 95; hexachordal inversional,
Boatwright, Howard, 272 92, 94, 95; tetrachordal, 95. See also
Bononcini, Giovanni Maria, 64–65, 73 Schoenberg, Arnold: use of hexa-
Boretz, Benjamin, 259, 266n40, 268 chords

index 295
combinatorics, 43, 92 cyclic interval succession (CINT), 101n22.

common-note function, 161–62 See also successive interval array
common-note vector, 158, 159, 161–62 cynosural set, 217
communality and contextuality (Babbitt),
26, 179 Dahlhaus, Carl, 31n1, 38–39, 82, 134n4,
complement (set operation), 49–50, 137
87n3, 89, 92, 93, 94, 96–97, 122, Daniel, Oliver, 259
175, 183, 207, 209, 215–16, 222, 230, DARMS. See Digital Alternative Repre-
237n2; definition of, 49. See also set sentation of Music Scores
complex Darmstadt Summer Courses for New
complement theorem, 97, 175 Music, 261n30, 263
Composers’ Forum, 259 Davis, James A., 190–91, 249n14, 250
Composers in Public Schools Program, Davis, Miles, 225
260, 272 Debussy, Claude, 45, 89, 91n8, 126n36,
comprehensive musicianship, 273 234; Fêtes (Nocturnes, No. 2), 194–98,
computer, 28, 91, 96, 116n29, 132, 137, 217; Prelude Brouillards, 233; Prelude
268, 271, 275–76; Forte’s view on the, La Cathédrale Engloutie, 275n51
248–49, 276; influence of the, on PC Dedekind, Richard, 85n1
set theory, 96n15, 237–38, 247–50; Delaere, Mark, 230
influence of the, on the scholar’s per- Deliège, Célestin, 4
spective, 240–41; technology, positive Dello Joio, Norman, 259
reception of, by scholars, 238–41, Dempster, Douglas, 219, 269
248. See also segmentation “dependent” series. See twelve-tone series
Computer and the Humanities, 242 developing variation, 137
Cone, Edward T., 2n2 difference (set operation), 49–50
Contemplating Music (Kerman 1985), 3, difference tone, 139, 142n9
22–23 difference vector. See interval-difference
contemporary music, 254, 274; funding vector
of American, 259–61; rejected by Digital Alternative Representation of
American orchestras, 259 Music Scores (DARMS), 244. See also
Contemporary Music Project, 272–75 Ford-Columbia (input language)
contextuality. See communality and con- directed PC interval. See pitch-interval
textuality (Babbitt) class (PIC)
contour, theory of, 182 Dirichlet. See Lejeune Dirichlet
contour class, 182–83 dualism. See harmonic dualism
contour segment (cseg), 182n3 Doerksen, John F., 189
Cook, Nicholas, 25n21, 136n6, 221n3, dominant, 53, 65, 67, 70, 205, 213; -flat-
223–24, 225. See also segmentation ninth chord, 128; -seventh chord, 113,
Copland, Aaron, 256n23 125, 199, 200; -thirteenth chord, 200
Cornell University, 239n6, 264n37 Dunsby, Jonathan, 2n2, 223
Craft, Robert, 201, 163n35
criticism, 23, 223, 227n10, 235n15 Eastman School of Music, 20, 22
cseg. See contour segment Eggebrecht, Hans-Heinrich, 221
“cycle-of-fourths-equivalence” (Howe). Eilenberg, Samuel, 86
See equivalence Eimert, Herbert, 26, 57, 80–81, 116, 242,
“cycle-of-fifths-equivalence” (Howe). See 261–62. See also all-interval series;
equivalence multiplication (M)

296 index
Élégie (Fauré), 181–83, 188 figured bass, 31, 32, 100, 121, 124, 128
Epstein, David, 233 Fludd, Robert, 254
equal temperament, 26, 29, 33, 34n7, 45, Ford Foundation, 260, 272
61, 84, 91, 228n11 Ford-Columbia (input language), 244,
equality of sets, 29, 41, 87–88, 89 246
equal-vector relation, 92, 98–99, 107, Forte, Allen, 3, 4, 10, 47, 103–9, 179–80,
131, 174. See also Z-relation 229, 232, 236, 237n2, 271n46;
equivalence, 29–30, 142, 151, 168, 177, analyses of Webern’s Op. 5, no.
179, 182, 189, 269; algebraic, 84–86, 5, 56, 59–60, 61; computer-aided
88, 92, 98n19, 99, 100, 101, 107; analysis project, 91, 242, 246–48, 275;
canonical, of PC sets (Lewin’s defini- educational philosophy of, 275–76;
tion), 114; of chords, 116, 120–22; influence of positivist epistemology
contour-, 182; “cycle-of-fourths-” on, 248–49; involvement with atonal
(Howe), 82; “cycle-of-fifths-” (Howe), music, 60, 89, 91, 132–33, 180, 248;
82; informal notions of, 84–85, 86, polemical exchange with Taruskin
87, 100; of intervals, 36–39, 85, 88; on The Rite of Spring, 219, 225–26,
of “lines” (Rahn), 110; octave-, 29–31, 228n11, 234; programming expertise
36–37, 190, 199; operational, of PC of, 244; on Schenkerian theory, 13,
sets (Howe), 99; partitional view of, 16, 273–74; score-reading program,
85–86, 109, 114; of PC sets (Clough’s 246–47. See also computer; equiva-
definition), 98–99; of PC sets (Forte’s lence; genus; integers; interval class;
definition), 89, 92, 97–98, 103, 105, interval content; invariance; inversion
107, 177, 206; of PC sets (Rahn’s (I); Michigan Algorithm Decoder
definition), 110, 113–14, 177; of pitch (MAD); normal form; normal order;
sets (Morris, Rahn), 110, 123–25; prime form; progenitor; segmenta-
sentential, 38n10, 87; set-theoretic, tion; table of set-classes; set complex;
86; sound and, 123–28; of successive similarity; similarity measures; trans-
interval-arrays (Chrisman), 100; of position (T); Yale University; Z-rela-
twelve-tone series (Babbitt), 87–88, tion
89. See also APIC vector; inversion (I); Forte, Allen, works by: The Atonal Music
multiplication (M); similarity; transpo- of Anton Webern (1998), 60, 61; “Con-
sition (T) text and Continuity in an Atonal
equivalence class, 30, 46, 85, 86n2, 92, Work” (1963), 88–89, 90, 91; The Har-
100–101, 107, 109–10, 113, 123, 131, monic Organization of the Rite of Spring
169, 177, 222. See also “type”; chord (1978), 23, 225–26, 227; “The Magical
class; contour class; contour segment Kaleidoscope” (1981), 183–84, 187,
(cseg); FB-class; PCINT-class; pitch- 191n13; “Pitch-Class Set Analysis
class (PC); pitch-interval class (PIC); Today” (1985), 233–34; “Schenker’s
pitch-set class (PSC); set-class Conception of Musical Structure”
Erpf, Hermann, 120 (1959), 15–16, 273; The Structure of
Atonal Music (1973), 19–20, 24, 27, 38,
FB. See figured bass 56, 57, 83, 91, 99, 104–5, 107, 108,
FB-class, 124, 125 109, 126, 134–35, 136, 148, 179–80,
Fauré, Gabriel, 181–83, 202 206–7, 209, 212, 271; “Theory of Set-
Ferentz, Melvin, 116n29, 242 Complexes in Music, A” (1964), 27,
Ferneyhough, Brian, 225 89, 91–92, 98, 104–5, 134, 144–48,
Fétis, François-Joseph, 25, 91 179, 206n26, 210

index 297
FORTRAN, 237n2 Hába, Alois, 43

Foucault, Michel, 17n15 Haimo, Ethan, 228n11
Fowler, David, 86n2 Hanson, Howard, 26, 43
Friedmann, Michael L., 39–40, 78n27, harmonic dualism, 67n16, 213n32,
182, 228, 237–38, 249. See also seg- 215–16
mentation harmonic function, 25, 122, 185, 213, 219
Fromm, Paul, 261 harmonic inversion, 121, 128. See also
Fromm Music Foundation, 261 bass position
fugue, 5, 7n8, 23, 53, 62, 63, 65, 66. See Harmonielehre (Weigl), 120–21
also rectus / inversus harmonisches Gefälle, 143, 177
fundamental bass, 137, 205n24 harmony, tonal, 2n2, 32, 100, 200, 232,
fundamental structure. See Ursatz 233; theory of, 25, 33n4, 116, 120–21,
Fux, Johann Joseph, 64–65, 73 128, 143, 203–5, 213, 251, 254n20.
“fuzzy inversion.” See inversion: “near-” See also bass position; dominant; har-
(Straus) monic dualism; harmonic function;
“fuzzy” relations, 130 harmonic inversion; subdominant;
“fuzzy transposition.” See transposition tonic; tonicization
(T): “near-” (Straus) Harvard University, 18, 19, 265, 267,
269n45
Gadamer, Hans-Georg, 218n1, 219n2 Hasty, Christopher, 227, 229
Gagné, David, 271 Hauer, Joseph Matthias, 26, 44n16, 117,
Gauss, Carl Friedrich, 85n1 122
gender studies, 235 Haydn, Joseph, 242
generalized interval system (GIS), 71 Heinemann, Stephen, 78n26
Generative Theory of Tonal Music, A (Ler- Helm, Everett, 263
dahl and Jackendoff 1983), 25, 269 hemiola, 199
genus, 27, 179, 191n13, 216–17 Hempel, Carl, 250
geometry, 55, 130, 153–54, 169 Hermann, Richard, 176
Gilbert, Steven, 13, 271 hexachord, 88, 92, 94–97, 102–3, 109,
GIS. See generalized interval system 122, 145n12, 161, 164; medieval, 191.
Gjerdingen, Robert, 136n6 See also combinatoriality; Schoenberg,
Gloag, Kenneth, 201 Arnold
Godlovitch, Stan, 225 hexachord theorem, 96. See also comple-
Goehr, Lydia, 137 ment theorem
Goelet, Francis, 260 High Fidelity, 255
Gouk, Penelope, 254 Hilbert, David, 254n20
Gould, Samuel B., 277 Hiller, Lejaren A., 241
Greissle, Felix, 31 Hindemith, Paul, 233, 254n20, 271, 277;
“Grenze der Halbtonwelt, Die” (Klein discussion of interval qualities, 139,
1925), 43, 116 142; on the use of chord relations,
Griffiths, Paul, 262 143–44, 177. See also chord classifica-
group. See transformation group tion; harmonisches Gefälle
group action, 50–51, 114–15 Hollerith, Herman, 238n3
Grout, Donald J., 263 Howard, Walther, 117, 122–23
Grundgestalt. See “basic set” Howe, Hubert S., 81–82, 91n10, 99,
Guggenheim Fellowship, 257n24 102–3, 109. See also equivalence; mul-
Guido of Arezzo, 30 tiplication (M)

298 index
Huizinga, Johan, 265n39 and sound, 97, 125–26. See also APIC
Hüschen, Heinrich, 251n16 vector; multiplication (M); PIC vec-
Husmann, Heinrich, 139n8 tor; Z-relation
interval expansion, 78–79, 128. See also
IBM. See International Business multiplication (M)
Machines Corporation interval function, 46–47, 49
IML. See Intermediate Music Language interval vector, 47, 92, 105, 156, 183,
inclusion, 133, 148, 150; application of 227, 228, 237, 248. See also APIC vec-
the concept of, 180–81, 187–89, 193– tor; PIC vector
94, 199; and complementation, 208– interval-difference vector, 153, 173–75
9; degree of, 158, 160; and invariance, “Intervallic Relations in Atonal Music”
206; involving musical objects with (Teitelbaum 1965), 152
plural identities, 202, 205; relation, Introduction to Post-Tonal Theory (Straus
the, as a property, 191, 194; relation, 1990, 2000), 19–21, 27, 40, 41, 42, 177,
transitivity of the, 214–15;. See also set 237, 270, 271
complex; subset; superset invariance, 206, 252
information science, 130 inverse, 38–40, 47–48, 62, 67, 68, 69,
Institut de Recherche et de Coordination 72–73, 76–77, 78, 82, 85, 88, 114, 215
Acoustique/Musique (IRCAM), 262 inversion (I), 10, 49, 55, 57, 62–76,
integers, 18, 29, 31, 33, 37, 45, 46, 52, 82–83, 96, 122n34, 125, 126, 133, 142,
69, 82, 105–6, 110, 124, 237; Babbitt’s 144, 148, 157, 158, 161, 162, 169, 174,
use of, 33–34, 50, 55, 67, 69, 79n28; 207, 215, 227, 228, 230, 237, 247, 249,
Forte’s use of, 34, 35, 57, 69–71, 88 270; axis of, 62, 67–74; defined as a
intentional fallacy, 228n11 multiplicative operation (Howe), 82;
International Business Machines Corpo- definition of, 62, 72–73; and equiva-
ration (IBM), 237n2, 241–42, 244 lence, 84–87, 98–101, 103, 107–9, 110,
Intermediate Music Language (IML), 113, 131, 206; Forte’s definition of, 69;
238 in fugal theory, 62–67; harmonic, see
International Musicological Society harmonic inversion; and harmonic
(IMS), 251n16 dualism, 67n16, 215–16; index of,
interruption, Schenkerian, 15–16, 230 69n21; Lewin’s definition of, 71–72,
intersection (set operation), 89, 110, 74n24; “near-” (Straus), 61; normal
193n15, 203, 205–6, 213; definition form of, 104, 105; of pitch intervals
of, 49 (pitch-sets), 62, 68, 72, 73–74, 112;
intertextuality, 235 and prime form, 105–7; in tonal
interval class, 36–40, 69, 190, 191n13, music, 65–66, 70; of twelve-tone
227, 252; Forte’s definition of, 37–40. series (Babbitt), 34, 50, 51, 67, 69; of
See also absolute pitch-interval class twelve-tone series (Eimert), 80–81;
(APIC); pitch-interval class (PIC) by voice exchange, see harmonic
interval content, 10, 43, 46–47, 52, inversion, bass position. See also com-
82–83, 132, 200, 237n2; considered binatoriality; rectus / inversus; “type”
as the prime distinguishing property (Rahn): “Tn/TnI-”
of a PC set, 92, 96; complementary IRCAM. See Institut de Recherche et de
hexachords equal in terms of, 96–97; Coordination Acoustique/Musique
equality of, abandoned as a basis of Isaacson, Eric J., 133n2, 136, 148, 157,
PC set equivalence, 98, 100–101, 103, 163, 167–69, 173–76. See also similarity
105, 129; Lewin’s definition of, 47; measures

index 299
Isaacson, Leonard M., 241 LaRue, Jan, 240, 242

Ives, Charles, 91 Laudan, Larry, 18n16
Leeuw, Ton de, 4n6
Jackendoff, Ray, 25, 180n2, 269, 270 Lehrbuch der Zwölftontechnik (Eimert
Jelinek, Hanns, 57, 116n29. See also all- 1950), 26, 57, 80–81
interval series Leibowitz, René, 57, 87n5
Jeppesen, Knud, 239 Leichtentritt, Hugo, 234n14
Josquin Desprez, 238–39, 242 Lejeune Dirichlet, Johann Peter Gustav,
Journal of Music Theory, 15, 17, 27, 93, 98– 85n1
99, 102, 152, 210, 244, 271, 272, 273 Lerdahl, Fred, 23–25, 180n2, 183, 187n9,
189, 205, 269, 270. See also music cog-
K. See set complex nition; segmentation
Kassler, Michael, 97n16, 239, 266n40 Lessem, Alan P., 277
Katz, Adele, 273 Lester, Joel, 22n17, 213n31, 252, 270
Kepler, Johannes, 254 Lewin, David, 2n2, 26, 36, 46–47, 50,
Kerman, Joseph, 3–4, 15, 16n14, 22–23, 71–72, 74n24, 92, 96–97, 98n18, 114,
219, 227n10, 235, 255n21, 256, 158n20, 161, 163–65, 175, 185n5, 206,
261n31, 273n49 223; critique of numerical PC labels,
Kerr, Clark, 264n37 35, 69; Generalized Musical Intervals
key, 32, 43, 53, 64–65, 199–200, 201, and Transformations (1987), 19–20, 61;
202–4; chords as subsets of a, 203, phenomenological perspective of,
213. See also regions, tonal 198n18; “A Response to a Response”
key relationships, 68, 203, 205, 213–16; (1980), 167–69, 172. See also comple-
Weber’s chart of, 204 ment theorem; equivalence; general-
Kh. See subcomplex ized interval system (GIS); interval
King, Martin Luther, 75 content; interval function; inversion
Klein, Fritz Heinrich, 43, 116–20. See also (I); partial ordering; similarity mea-
chord classification sures; transformation group; transfor-
Koch, Heinrich Christoph, 32, 52, 53, mations, theory of
54n4, 55 “Liberal Arts,” 254
Komar, Arthur, 15n12 Lincoln, Harry B., 242
Kopp, David, 195n16 Lippius, Johannes, 30
Korsyn, Kevin, 235 Liszt, Franz, 89
Kostka, Stefan, 19–21, 22n17 Lockwood, Lewis, 238–40
Kraehenbuehl, David, 268, 271–73 logic, mathematical, 38n10
Kramer, Lawrence, 22 logical positivism. See positivism
Krenek, Ernst, 55n5, 116n29, 254n20; Loquin, Anatole, 26, 33, 43, 117, 121, 124
discussion of interval qualities, 139, Lord, Charles H., 135–36, 148, 153n16.
142; Studies in Counterpoint (1940), See also similarity measures
26, 33, 57, 80; Über neue Musik (1937), Louis, Rudolf, 120
79–80; on the use of chord relations, Lowinsky, Edward E., 266
143. See also chord classification Luening, Otto, 259
Kruyt, Hugo Rudolph, 265n39
Kuhn, Thomas S., 18n16 Mac Lane, Saunders, 86
Kurth, Ernst, 220–21, 254n20 MacDowell, Edward, 265
MAD. See Michigan Algorithm Decoder
Lansky, Paul, 40, 57 Mann, Alfred, 54n4, 64n13, 64n14

300 index
Mannes College of Music, 244 interval succession (PCINT); pitch

Marpurg, Friedrich Wilhelm, 54n4, set; pitch-set class (PSC) ; segment;
64–65, 66 similarity measures; Starr, Daniel
Martin, Robert, 225n8 “mother chord,” 116. See also all-interval
Martino, Donald, 47, 88, 92, 96, 97n17, series
102–3, 259 Mozart, Wolfgang Amadeus, 233
mass education, 28, 277–78 multiplication (M), 49–50, 51. 76–83,
mathematics, 10, 18, 27–28, 79, 86–87, 157, 271n46; applied to twelve-tone
123, 128, 145, 151, 167–68, 237, 255, series, 79–81; Boulez’s definition of,
261, 264, 277. See also absolute value, 78n26; defined by Eimert, 81; defined
algebra, combinatorics, equivalence, by Howe, 81; definition of, 76, 77–78;
geometry; group action; logic, math- and equivalence, 110, 114–15; and
ematical; operation; relation; set the- interval content, 82–83; of order
ory, mathematical; standard deviation numbers, 79; of pitch intervals (pitch
function; statistics; transformation sets), 76; transforming a chromatic
group; vector scale into cycles of fourths or fifths,
Maus, Fred E., 223 79–80. See also equivalence: cycle-of-
McClary, Susan, 22, 23n18 fifths-; equivalence: cycle-of-fourths-;
McCreless, Patrick, 3, 17n15, 250, 268, interval expansion; transformation
271 group; “type” (Rahn): “Tn/TnI/M5/
Mellor, D. H., 190n10 M7-”
Mendel, Arthur, 238, 239n5, 266 Music Analysis, 225
Menke, Johannes, 116 music cognition, 269
Mersenne, Marin, 254 Music Forum, 17
Messiaen, Olivier, 199–201, 202 Music Information Retrieval (MIR), 239
Meyer-Eppler, Werner, 262 “Music for the Masses” (Brody 1993),
Michigan Algorithm Decoder (MAD), 256
247–48 Music Theory Spectrum, 17
MIR. See Music Information Retrieval musica ficta, 239n5, 264
“modes of limited transposition” (Mes- Musical Quarterly, 55, 79, 256, 267
siaen), 199 musicology, 239; new, 22–23
mod(ulo) 7, 124
mod(ulo) 12, 29–30 National Center for Bibliographic Data
Monteverdi, Claudio, 233 Processing in the Humanities, 242
Morris, Robert D., 19–20, 101n22, 115, National Endowment for the Arts, 260
123–25, 135, 136, 163–65, 167–68, National Science Foundation, 241
172, 176, 180n2, 185n5; on the audi- Neff, Severine, 33n5
bility of PC set relations, 123, 133–34; new musicology. See musicology
Composition with pitch-classes (1987), New York Philharmonic, 244
39, 61, 123; “A General Theory of New York Times, 225n9, 263
Combinatoriality and the Aggregate” New York University, 242
(1977/1978), 83. See also combinato- Newlin, Dika, 222n6
riality; contour, theory of; contour Newton, Isaac, 254
class; contour segment (cseg); cyclic nexus set, 212–13, 214. See also set com-
interval succession (CINT); FB-class; plex
interval class; multiplication (M); par- Nolan, Catherine, 33n4, 33n5, 43,
tial ordering; PCINT-class; pitch-class 116n27, 121

index 301
normal form, 102–4, 237, 248; algo- PC. See pitch-class

rithm (Randall), 102–3, 105; Babbitt’s PCINT. See pitch-class interval succession
description of, 102. See also normal PCINT-class, 124
order PCSET (collection), 50, 84, 85, 86, 101,
normal order, 93, 102–4, 105, 107n23; 109, 110, 114; definition of, 45–46
algorithm (Forte), 104; “best” (or pedal, 195, 198
“optimal”), 104. See also normal form performance, analysis and, 220–21,
Notker Labeo, 30 223–24
null set, 45, 46, 110, 116n30, 193n14 period (musical form), 15, 16
numerical notation. See integers Perle, George, 4n6, 23, 57, 68n18, 70–71,
91n8, 91n9, 105, 184–85, 252, 270.
O King (Berio), 75–76 See also absolute pitch-interval class;
octatonic scale, 13, 109, 126, 194, 200– segmentation
202, 217; “collection I,” 126–27, 200; Perspectives of New Music, 17, 24, 35,
“collection II,” 126–27; “collection 81, 83, 90, 158, 163, 167, 259n25,
III,” 126–27, 200 261n30, 261n31
octave equivalence. See equivalence “Petrushka chord,” 200–202
Oettingen, Arthur Joachim von, 215–16 PhD, 17, 266
Olive, Joseph P., 242 PI. See pitch interval
Oliver, Alex, 190n10 PIC. See pitch-interval class
operation, 27, 84, 85, 96, 99–100, 107, PIC vector, 47–48, 162n22, 168n28,
110, 113–14, 115, 123, 128–29, 130, 175–76, 206n25
157, 252, 271n46; binary, 49; as PITCH (collection), 52, 61, 76–77, 84,
defined in PC set theory, 50–51, 76; 252; definition of, 45
twelve-tone, 10, 33–34, 50, 51, 67–68, pitch interval (PI), 9, 35–36, 37, 38, 52,
81, 82–83, 84, 96; unary, 50. See 62, 68, 72–74, 75, 76, 123–24, 176;
also complement; difference; group definition of, 35. See also inversion;
action; intersection; inversion (I); multiplication (M); transposition (T)
multiplication (M); retrograde; ret- pitch set, 76, 110, 123–25, 176, 181,
rograde-inversion; transposition (T); 271n46
union pitch-class, definition of, 30
ordered PC interval. See pitch-interval pitch-class interval succession (PCINT),
class (PIC) 101n22, 124
“Our Musical Idiom” (Bacon 1917), 33, pitch-class set (PC set), definition of,
122 40–41
overtone series, 251 pitch-class set complex. See set complex
pitch-class set genus. See genus
Paine, John Knowles, 265 PITCHCLASS (collection), 50–51, 52,
Palestrina, Giovanni Pierluigi da, 239 57, 61, 76, 84, 93, 96, 115, 193n15,
Parker, Horatio, 265 252; definition of, 45
Parks, Richard S., 45, 179n1, 216–17, pitch-interval class (PIC), 36–37, 39–40,
233–34 46–48, 52, 55, 62, 72, 74, 82–83,
parsimonious voice-leading, 195 97, 100, 121, 124, 206n25. defini-
partial ordering, 185 tion of, 36. See also absolute pitch-
partition function, 158–62, 165–66 interval class (APIC); inversion;
partition vector, 158–61 multiplication(M); transposition (T)
Payne, Dorothy, 19–21, 22n17 pitch-set class (PSC), 123–25

302 index
Pitch-Time Correlation (Randall), 82, 102 relation, definition of, 84

Pomeroy, Boyd, 195 Répertoire International de Littérature Musi-
Popular Front, The, 256n23 cale (RILM), 242
positivism, 23, 34–35, 249, 250, 253 representative form (of a PC set). See
Pousseur, Henri, 9 prime form
power set, 45, 191n12, 193 Réti, Rudolph, 233
prime form, 105–7, 108, 109, 110, 112, retrograde, 34, 51, 55, 67, 80–81, 94, 111,
126, 177, 209, 212n29, 228, 237, 248; 113; defined by Babbitt, 50
algorithm, 104, 105–6; of a twelve- retrograde-inversion, 34, 51, 68, 80,
tone series, 57, 67–69, 80, 87n5, 92, 94–95; defined by Babbitt, 50
94n13 Riemann, Hugo, 25, 215–16, 254n20
Princeton University, 9, 17, 237n2, 238, RILM. See Répertoire International de Lit-
250, 261n30, 266. See also Perspectives térature Musicale.
of New Music; Seminar in Advanced Rivera, Benito V., 31n1, 41
Musical Studies Robison, Tobias D., 238–39
progenitor, 191, 193n15, 198n19, 199, Rockefeller Center, 275
206, 215, 216; Forte’s definition of, Rockefeller Foundation, 241, 260
191n13 Rodriguez-Pereyra, Gonzalo, 190
PSC. See pitch-set class Rogers, David W., 163, 168–69, 171
PSET (collection), 111, 121, 123–24 Roslavets, Nikolai Andreyevich, 91n8
Pythagoras, theorem of, 153 Rothstein, William, 221, 273n49, 274
Roussau-Galin-Paris-Chevé method for
Rahn, John, 36, 114, 122, 123, 136, 142, sight-singing, 31
157n17, 163, 169, 172, 177, 183n4, Rufer, Josef, 56
206n25, 268; Basic Atonal Theory
(1980), 19–21, 27, 39, 44–45, 83, 109– salience, 59, 188–89, 194, 199, 201–2,
13, 237, 271; “Relating Sets,” (1980), 205. See also set-class
165–67; on tonal theory, 45, 113. See Salzer, Felix, 233, 252, 273
also equivalence; multiplication (M); Sand, Ole, 276–77
similarity measures, “type” Scarlatti, Domenico, 233n13
Rameau, Jean-Philippe, 30–31, 41, 53, Schachter, Carl, 205n24
121, 134, 137, 254 Schaffer, Sarah, 227
Ramirez, Francisco O., 265 Schat, Peter, 31, 34, 35
Randall, James K., 81–82, 102, 105, 238 Schenker, Heinrich, 3–4, 13–17, 32, 34,
Ratz, Erwin, 233 220–22, 223–24, 230, 233, 239n4,
Ravel, Maurice, 233 250n15, 252, 271, 273–74, 275n51,
rectus / inversus, 5, 67, 105 276; Der Freie Satz, 14. See also inter-
Regener, Eric, 36, 41, 97n16, 101n22, ruption; Ursatz
109, 136, 144–45, 148n14, 157–61, Schillinger, Joseph M., 33, 43
162–63, 165–66, 172, 238. See also Schmalfeldt, Janet, 212n30
common-note function; common-note Schoedinger, Andrew B., 190n10
vector; inclusion: degree of; partition Schoenberg, Arnold, 9, 10, 18, 44, 55n5,
function; partition vector 82, 83, 87n5, 89, 91, 96n14, 105,
regions, tonal, 205, 215 116n28, 118, 121, 131, 135, 137, 143,
Reich, Willi, 79–80 152, 180, 229, 234, 235, 239, 248,
Reicha, Anton, 203, 213n31 250n15, 255n21, 264, 277; as an analyst,
Reihe, Die, 35, 261n31 137, 220–22, 233; historical awareness

index 303
of, 70, 94n13, 137; interval notation 244–47; Cook’s, of Schoenberg’s

of, 56–57; PC-set awareness of, 97, Piano Piece, Op. 19, no. 6, 229–32;
228n11; use of hexachords, 92, 94, Lerdahl’s view on, 23–24, 180n2, 183,
97. See also “basic set”; “basic shape”; 269; Perle’s, of the first measures of
developing variation; set; regions, tonal; Schoenberg’s Piano Piece, Opus 11,
“strong” and “superstrong” progres- no. 1, 184, 185; by sequential imbrica-
sions; twelve-tone series tion (Friedmann, Forte), 237–38, 249;
Schoenberg, Arnold, works by: Book of the Straus’s view on, 228, 229, 249
Hanging Gardens, Op. 15, no. 11, 237; Seminar in Advanced Musical Studies,
“Composition with Twelve Tones” 261n30
(1950), 44, 56–57, 92, 222n5; Four Seminar on Comprehensive Musician-
Orchestral Songs, Op. 22, 222n5, 233; ship, 275, 277
Fundamentals of Musical Composition sentential calculus. See logic, mathematical
(published 1967), 180n2; Modern Psalm, Serial Composition and Atonality (Perle
Op. 50c, 95; Moses und Aron, 94; Piano 1962), 19, 57–59, 91n8, 91n9, 184, 252
Piece, Op. 11, no. 1, 41, 50, 183–87, Sessions, Roger, 257, 273
188, 194, 199; Piano Piece, Op. 23, set, definition of, 29
no. 3, 4–7, 8, 9, 10, 13, 23, 70–71, 74; set complex, 27, 99n20, 179–80, 207–17,
Piano Piece, Op. 33a, 95, 96, 111, 112, 230, 237n2, 247, 249, 271; analysis,
113, 123, see also segmentation; Six 209, 211–12; evolution of the, 209;
Little Piano Pieces, Op. 19, 88–89, 90 reference pair of a, 209, 212; tonal
(no. 2), see also segmentation (no. 6); theory through the lens of the, 213–
String Quartet No. 3, Op. 30, 252–53, 16. See also subcomplex
270; String Quartet No. 4, Op. 37, 95, set theory, mathematical, 3n3, 29, 40, 89
102–3; Structural Functions of Harmony set-class, 92, 100–101, 109, 207–8, 216,
(published 1954), 198n17, 205, 214–15; 226, 229–30, 237, 247, 216, 226,
Suite for Piano, Op. 25, 54 (Prelude), 229–30; identification, 237, 247; rep-
41, 42, 270 (Gavotte); Theory of Har- resenting a, 103–5, 107, 109; salience,
mony (1911), 187n7; Variations for 188–89, 191, 194; and transformation
Orchestra, Op. 31, 68, 88, 94; Wind group, 114–15. See also equivalence
Quintet, Op. 26, 67–68, 94n13 class; table of set-classes
Schott Music, 120n33, 261 Shreffler, Anne C., 267n42
Schuller, Gunther, 259 similarity, 179, 180, 227, 247, 249, 269;
Schumann, Robert, 14–16, 200n21, 274 aural experience of, 131, 133–34, 136n6,
science, history and, 218 177; equivalence and, 130, 177, 182;
science, music theory and, 23–24, 34–35, from the composer’s perspective, 136,
102, 250–56, 264, 265–71, 274, 277– 177; as a focus of the study of musical
78. See also Babbitt, Milton; university structure and style, 132, 135; Forte’s
Scriabin, Alexander N. 45, 58, 89, 91n8, introduction of PC set, 131, 173; maxi-
135, 136 mal (first-order), see similarity measures:
Scruton, Roger, 223 R1 (Forte); maximal (second-order), see
Seeger, Charles, 239n6 similarity measures: R2 (Forte); mini-
segment, Morris’s definition of, 180n2 mal, see similarity measures: R0 (Forte);
segmentation, 23–25, 61, 70–71, 185, origins of the concept of, 130; and the
189, 226–27, 228; ambiguity in, 181, pedagogical tradition of music analysis,
229; Benjamin’s criticism of Forte’s 133, 134; “strong, “superstrong,” and
approach to, 24n20; computerized, “weak” (Lewin), 161–62

304 index
similarity measures, 132, 133, 134, 136, Steinecke, Wolfgang, 263n34

179, 222 ; ASIM (Morris), 164–65, Stockhausen, Karlheinz, 34–35, 261–62,
172, 179; ATMEMB (Rahn), 167, 263
169, 172, 179; EMB (Lewin), 165, Storm van Leeuwen, W., 265n39
167; IcVSIM (Isaacson), 136, 173– Straus, Joseph N., 27, 40, 41–42, 61–62,
76; PM (Morris), 176; R0 (Forte), 125, 137, 177, 228–29, 235, 237, 249,
134, 135, 136, 147–49, 150, 152, 158, 267, 270, 271; and the “myth of serial
172, 247; R1 (Forte), 134, 135, 136, tyranny,” 267. See also inversion; seg-
145–47, 148, 150, 154, 155, 158, mentation; transposition (T)
160, 172, 247; R2 (Forte), 134, 136, Stravinsky, Igor, 18, 44–45, 180, 233,
145–47, 148, 150–51, 154, 158, 172, 235, 257, 263n35; Agon (“Pas de
247; Rp (Forte), 134, 135, 136, 148, deux”), 8–13, 24, 194, 199; PC-set
157–58; RECREL (Castrén), 176; awareness of, 228n11; Petrushka, 201;
REL2 (Lewin), 167–69, 171, 172, 173, Symphonies of Wind Instruments, 126–
179; S.i., (Teitelbaum), 132, 152–57, 28, 200; The Rite of Spring, 212n30,
162, 164, 165, 169, 172, 173–74, 225–26, 228. See also octatonic scale;
175; sf (Lord), 135–36; SIM (Mor- “Petrushka chord”
ris), 133–34, 163–65, 167; TMEMB Street, Alan, 4
(Rahn), 166–67, 183n4. See also “strong” and “superstrong” progressions
APIC vector; common-note func- (Schoenberg), 198n17, 205n24
tion; common-note vector; inclu- “strong,” superstrong,” and “weak” simi-
sion: degree of; partition function; larity (Lewin). See similarity
partition vector subcomplex, 208–12
Simms, Bryan R., 4n6, 7n8 subdominant, 213, 215
Skizze einer neuen Methode der Harmonie- subset, 49, 88–89, 165, 176, 179, 181;
lehre (Riemann 1880), 215 common, 148, 157–58, 161, 166–67,
Small, Christopher, 225n8 168; content, similarity measures
Smith, P. H., 239 based on, 148, 157–163, 165, 168, 172,
Smith, Russel F. W., 266 176; definition of, 45; invariant, 206;
Snarrenberg, Robert, 221, 273n49, 274 nomadic, 199, 216; proper, 45, 207;
SNOBOL, 244, 246 settled, 199, 206, 216. See also inclu-
Society for Music Theory (SMT), 17 sion; key; set complex
“Some Combinational Properties of Pitch subsets, family of, 191–94, 199, 207, 209,
Structures” (Howe 1965), 81, 99, 102–3 214–16
Sontag, Susan, 236n1 successive interval array, 100–102, 103,
“sound,” 82, 113, 125–26, 128, 217. See 122. See also cyclic interval succession
also equivalence; interval content (CINT)
source set, 96 superset, 179, 181; definition of, 45; proper,
“Source Set and Its Aggregate Forma- 45, 207. See also inclusion; set complex
tions, The” (Martino 1961), 47, 97n17, supersets, family of, 191, 194, 199, 200,
102–3 209, 214, 216
standard deviation function, 173
Starr, Daniel, 83. See also Morris, Robert D. tablature, 31, 32
statistics, 130, 165, 167, 173 table of set-classes, 93, 99, 103, 108
Stein, Erwin, 4n6, 31, 55n5, 117, 118, Tanglewood Festival of Contemporary
120–21 Music, 261n30
Stein, Leon, 180n2 Tarski, Alfred, 38n10

index 305
Taruskin, Richard, 23, 126n37, 127, 200, trope, 122

219–20, 227, 234, 269; polemical twelve-tone operations. See operation
exchange with Forte on The Rite of twelve-tone serialism, 9, 26, 33, 37, 83,
Spring, 219, 225–26, 228n11, 234 113, 143, 228; European versus Amer-
Tawa, Nicholas, 259 ican, 56, 225, 261n31. Milton; PC set
Teitelbaum, Richard, 91n10, 131–32, 135, theory influenced by, 31, 43–44, 60,
137, 152, 165, 172, 173, 176. See also 105, 206, 225; and scientism, 267. See
similarity measures also Babbitt, Milton
Thuille, Ludwig, 120 twelve-tone series, 26, 37, 54–55;
“Tn-type.” See “type” Dahlhaus’s definition of, 38–39,
“Tn/TnI-type.” See “type” 82; “dependent,” 87n4; forms, hier-
“Tn/TnI/M5/M7-type.” See “type” archy of, 67–68, 87n5; and octave
Toch, Ernst, 277 equivalence, 31; and PC set, 10, 44,
Tomlinson, Gary, 22, 23n18, 219 59, 60–61; of Schoenberg’s Piano
Tonal Harmony (Kostka and Payne 1984), Piece, Op. 33a, 95, 96; of Schoen-
19–21, 22n17 berg’s Third String Quartet, op. 30,
tonic, 53, 65, 67, 69, 70, 200, 205, 213– 252–53; of Schoenberg’s Variations
14, 215, 233 for Orchestra, Op. 31, 68, 88, 94;
tonicization, 203 of Webern’s String Quartet, Op.
Traîté de l’harmonie (Rameau). See Treatise 28, 88; of Webern’s Variations for
on Harmony Orchestra, Op. 30, 9, 10, 12–13. See
transformation group, 51, 82n32, 114; also all-interval series; combinatorial-
canonical, 113, 115, 162; the concept ity; equivalence; integers; inversion;
of a, invoked by Lewin, 114 multiplication; operation; prime form;
transformations, theory of, 26 retrograde; retrograde-inversion;
translation, 55 transposition (T)
transposition (T), 10, 49, 59–61, 78n26, “type” (Rahn), 109–10, 111, 112, 114, 123;
82–83, 87n4; definition of, 52; dif- “Tn-,” 122; “Tn/TnI-,” 237. See also
ferent labels for, 56–57; and equiva- Equivalence class
lence, 85–86, 98–101, 110, 126, 128;
Forte’s definition of, 57; in fugal union (set operation), 89, 193n15, 207;
theory, 53–54; as an inappropriate definition of, 49
equivalent of Versetzung, 54n4; index Universal Edition, 261
of, 52, 69n21, 85, 99; Koch’s defini- universal set, 45, 57, 69, 88, 193n14
tion of, 52–53; “modal” (Friedmann), universals, 190
78n27; of twelve-tone series (Babbitt), university, 9, 17n15, 239n6, 268n44;
34, 50, 51, 55–56; “near-” (Straus), Babbitt’s view of the, 258–59, 264,
61–62; of pitch intervals (pitch sets), 268; Boulez’s view of the, 263–64, 268;
52. See also invariance; “modes of lim- employment of composers by the Amer-
ited transposition” (Messiaen); parti- ican, 265–67; history of the American,
tion function; transformation group. 264–65; as a home for the creative arts,
transposition number. See 265–66; teaching, Stravinsky’s aversion
transposition(T): index of to, 263n35; types, 18. See also Columbia
Treatise on Harmony (Rameau 1722), 41, University; Cornell University; Harvard
53, 137 University; New York University; PhD;
Treitler, Leo, 219, 223, 235n15 Princeton University; University of Wis-
“Tristan” chord, 113, 125 consin; Yale University

306 index
University of Wisconsin, 20, 22, 264n36, Quartet, Op. 9, no. 5, 211–12, 214;
264n37 Five Movements for String Quartet,
Unterweisung im Tonsatz (Hindemith), 138 Op. 5, 135 (general), 91, 131, 144–45,
Ursatz, 14, 221, 252, 273 234 (no. 4), 56, 58, 59–60 (no. 5); Six
Pieces for Orchestra, Op. 6, no. 3,
Van den Toorn, Pieter C., 221n4, 45, 71–72, 74; Three Pieces for Cello and
126, 200 Piano, Op. 11, 91 (general), 245, 246
Varèse, Edgar, 34, 257 (no. 1); Three Pieces for Violin and
vector, 47, 153. See also APIC vector; Piano, Op. 7, no. 1, 243, 244; Varia-
common-note vector; interval-dif- tions for Piano, Op. 27, 68–69, 70. See
ference vector; partition vector, PIC also twelve-tone series
vector Weigl, Bruno, 116n27, 117, 120–22, 123,
Vicentino, Nicola, 62, 63 124. See also chord classification
Vincent, Heinrich, 33 Whittall, Arnold, 2n2
Vogler, Georg Johann, 203 Wilcox, Howard J., 97n16
Winham, Godfrey, 81–82, 238, 266n40
Waerden, B. L. van der, 86 Wittlich, Gary, 183
Wagner, Richard, 124; Tristan und Isolde
(prelude), 111–13. See also “Tristan Yale University, 3, 17, 237n2, 265, 267,
chord” 268n44. See also Journal of Music
Wason, Robert W., 33n4, 205n24, 272 Theory
Weber, Gottfried, 203, 212. See also key
relationships Zarlino, Gioseffo, 62, 63, 64
Weber, William, 227n10, 134 Ziehn, Bernard, 67n16, 91, 120
Webern, Anton, 2, 18, 34, 91, 94, 131, Z-relation, 83, 98, 99, 103, 104, 107, 109,
152, 239, 248; Bagatelles for String 125, 144, 162, 174, 237

Eastman Studies in Music
The Poetic Debussy: A Collection of The Gardano Music Printing Firms,
His Song Texts and Selected Letters 1569–1611
(Revised Second Edition) Richard J. Agee
Edited by Margaret G. Cobb
“The Broadway Sound”:
Concert Music, Rock, and Jazz since 1945: The Autobiography and Selected Essays of
Essays and Analytical Studies Robert Russell Bennett
Edited by Elizabeth West Marvin Edited by George J. Ferencz
and Richard Hermann
Theories of Fugue from the Age of
Music and the Occult: Josquin to the Age of Bach
French Musical Philosophies, 1750–1950 Paul Mark Walker
Joscelyn Godwin
The Chansons of Orlando di Lasso and
“Wanderjahre of a Revolutionist” and Other Their Protestant Listeners: Music, Piety, and
Essays on American Music Print in Sixteenth-Century France
Arthur Farwell, Richard Freedman
edited by Thomas Stoner
Berlioz’s Semi-Operas: Roméo et
French Organ Music from the Revolution to Juliette and La damnation de
Franck and Widor Faust Daniel Albright
Edited by Lawrence Archbold
The Gamelan Digul and the Prison-Camp
and William J. Peterson
Musician Who Built It: An Australian Link
Musical Creativity in Twentieth-Century with the Indonesian Revolution
China: Abing, His Music, (includes CD)
and Its Changing Meanings Margaret J. Kartomi
(includes CD)
“The Music of American Folk Song”
Jonathan P. J. Stock
and Selected Other Writings on
Elliott Carter: Collected Essays and American Folk Music
Lectures, 1937–1995 Ruth Crawford Seeger, edited by
Edited by Jonathan W. Bernard Larry Polansky and Judith Tick
Music Theory in Concept and Practice Portrait of Percy Grainger
Edited by James M. Baker, David Edited by Malcolm Gillies
W. Beach, and Jonathan W. Bernard and David Pear
Music and Musicians in the Escorial Liturgy Berlioz: Past, Present, Future
under the Habsburgs, 1563–1700 Edited by Peter Bloom
Michael J. Noone
The Musical Madhouse
Analyzing Wagner’s Operas: Alfred Lorenz (Les Grotesques de la musique)
and German Nationalist Ideology Hector Berlioz
Stephen McClatchie Translated and edited by Alastair Bruce
Introduction by Hugh Macdonald

The Music of Luigi Dallapiccola Letters I Never Mailed: Clues to a Life
Raymond Fearn Alec Wilder
Annotated by David Demsey
Music’s Modern Muse: A Life of Winnaretta
Foreword by Marian McPartland
Singer, Princesse de Polignac
Sylvia Kahan Wagner and Wagnerism in Nineteenth-
Century Sweden, Finland, and the Baltic
The Sea on Fire: Jean Barraqué
Provinces: Reception, Enthusiasm, Cult
Paul Griffiths
Hannu Salmi
“Claude Debussy As I Knew Him” and
Bach’s Changing World:
Other Writings of Arthur Hartmann
Voices in the Community
Edited by Samuel Hsu,
Edited by Carol K. Baron
Sidney Grolnic, and Mark Peters
Foreword by David Grayson CageTalk: Dialogues with and about
John Cage
Schumann’s Piano Cycles and the
Edited by Peter Dickinson
Novels of Jean Paul
Erika Reiman European Music and Musicians
in New York City, 1840–1900
Bach and the Pedal Clavichord:
Edited by John Graziano
An Organist’s Guide
Joel Speerstra Schubert in the European Imagination,
Volume 1: The Romantic and Victorian Eras
Historical Musicology: Sources,
Scott Messing
Methods, Interpretations
Edited by Stephen A. Crist and Opera and Ideology in Prague:
Roberta Montemorra Marvin Polemics and Practice at the National
Theater, 1900–1938
The Pleasure of Modernist Music:
Brian S. Locke
Listening, Meaning, Intention, Ideology
Edited by Arved Ashby Ruth Crawford Seeger’s Worlds:
Innovation and Tradition in Twentieth-
Debussy’s Letters to Inghelbrecht:
Century American Music
The Story of a Musical Friendship
Edited by Ray Allen and Ellie M. Hisama
Annotated by Margaret G. Cobb
Schubert in the European Imagination,
Explaining Tonality:
Volume 2: Fin-de-Siècle Vienna
Schenkerian Theory and Beyond
Scott Messing
Matthew Brown
Mendelssohn, Goethe, and the Walpurgis
The Substance of Things Heard:
Night: The Heathen Muse in European
Writings about Music
Culture, 1700–1850
Paul Griffiths
John Michael Cooper
Musical Encounters at the
Dieterich Buxtehude: Organist in Lübeck
1889 Paris World’s Fair
(includes CD)
Annegret Fauser
Kerala J. Snyder
Aspects of Unity in J. S. Bach’s
Musicking Shakespeare: A Conflict of Theatres
Partitas and Suites: An Analytical Study
Daniel Albright
David W. Beach

Pentatonicism from the French Music, Culture, and
Eighteenth Century to Debussy National Identity, 1870–1939
Jeremy Day-O’Connell Edited by Barbara L. Kelly
Maurice Duruflé: The Art of Musical Phrasing in the
The Man and His Music Eighteenth Century: Punctuating the
James E. Frazier Classical “Period”
Stephanie D. Vial
Representing Non-Western Music in
Nineteenth-Century Britain Beethoven’s Century:
Bennett Zon Essays on Composers and Themes
Hugh Macdonald
The Music of the Moravian
Church in America Composing for Japanese Instruments
Edited by Nola Reed Knouse (includes 2 CDs)
Minoru Miki
Music Theory and Mathematics:
Translated by Marty Regan
Chords, Collections, and Transformations
Edited by Philip Flavin
Edited by Jack Douthett, Martha M.
Hyde, and Charles J. Smith Variations on the Canon:
Essays on Music from Bach to Boulez in
The Rosary Cantoral: Ritual and
Honor of Charles Rosen on
Social Design in a Chantbook from
His Eightieth Birthday
Early Renaissance Toledo
Edited by Robert Curry, David Gable,
Lorenzo Candelaria
and Robert L. Marshall
Berlioz: Scenes from the Life and Work
Wagner and Venice
Edited by Peter Bloom
John W. Barker
Beyond The Art of Finger Dexterity:
Analyzing Atonal Music:
Reassessing Carl Czerny
Pitch-Class Set Theory and Its Contexts
Edited by David Gramit
Michiel Schuijer

For the past forty years, pitch-class set theory has served as a frame of refer-
ence for the study of atonal music, through the efforts of Allen Forte, Milton Bab-
bitt, and others. It has also been the subject of sometimes furious debates between
music theorists and historically oriented musicologists, debates that only helped
heighten its profile. Today, as oppositions have become less clear-cut, and other
analytical approaches to music are gaining prominence, the time has come for a
history of pitch-class set theory, its dissemination, and its role in the reception of
the music of Schoenberg, Stravinsky, and other modernist composers.
Analyzing Atonal Music: Pitch-Class Set Theory and Its Contexts combines thor-
ough discussions of musical concepts with an engaging historical narrative.
Pitch-class theory is treated here as part of the musical and cultural landscape
of the United States. The theory’s remarkable rise to authority is related to the
impact of the computer on the study of music in the 1960s, and to the Ameri-
can university in its double role as protector of high culture and provider of
mass education.
Michiel Schuijer teaches at the Conservatory of Amsterdam and the Univer-
sity of Amsterdam. His research focuses on topics at the interface between music
theory and historical musicology.
“A penetrating study of pitch-class set theory, its mathematical foundations, and

the context in which it was developed.”
—Mark Delaere, University of Leuven (Belgium)
“Schuijer’s book brilliantly situates pitch-class set theory—our dominant mode

for analyzing atonal music—in its historical, social, and cultural contexts: this is
a tour de force of intellectual history. What is more, it also provides a magiste-
rial if, at times, pointedly critical summary of the theory itself. A deeply thought
evaluation of a theoretical approach that has largely escaped critical scrutiny,
despite its dominant position in contemporary North American music theory.”
—Joseph N. Straus, Distinguished Professor,
Graduate Center, City University of New York

Schuijer - Analyzing Atonal Music - Pitch-Class Set Theory and Its Contexts-University of Rochester Press (2008)

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

Schuijer - Analyzing Atonal Music - Pitch-Class Set Theory and Its Contexts-University of Rochester Press (2008)

Uploaded by

Copyright:

Available Formats

Analyzing Atonal Music

Schuijer.indd i 10/2/2008 10:29:33 PM

Additional Titles in Music Analysis and on Music since 1920

Analyzing Wagner’s Operas: Alfred Lorenz Opera and Ideology in Prague:

Schuijer.indd ii 10/2/2008 10:30:17 PM

University of Rochester Press

Schuijer.indd iii 10/2/2008 10:30:18 PM

All rights reserved. Except as permitted under current legislation,

First published 2008

University of Rochester Press

Library of Congress Cataloging-in-Publication Data

This publication is printed on acid-free paper.

Schuijer.indd iv 10/2/2008 10:30:18 PM

Schuijer.indd v 10/2/2008 10:30:18 PM

1 Pitch-Class Set Theory: An Overture 1

2 Objects and Entities 29

7 “Blurring the Boundaries”: Analysis, Performance, and History 218

Reference List 279

Schuijer.indd vii 10/2/2008 10:30:18 PM

Schuijer.indd ix 10/2/2008 10:30:19 PM

3.5 Excerpts from Webern’s Opus 5, no. 5, selected by Allen Forte in

Schuijer.indd x 10/2/2008 10:30:19 PM

4.7 The opening of the third movement of Schoenberg’s Fourth String

Schuijer.indd xi 10/2/2008 10:30:19 PM

6.10 A tonal progression in Messiaen’s Mode 2, the octatonic.

Schuijer.indd xii 10/2/2008 10:30:19 PM

3.1 An operation as defined by PC set theory 51

Schuijer.indd xiii 10/2/2008 10:30:19 PM

Schuijer.indd xv 10/2/2008 10:30:20 PM

Schuijer.indd xvii 10/2/2008 10:30:20 PM

Schuijer.indd xviii 10/2/2008 10:30:20 PM

Pitch-Class Set Theory:

Schuijer.indd Sec1:1 10/2/2008 10:30:20 PM

works from this repertoire be tailored to the purpose of systematical training

“You guys are discussing methods of analyzing twentieth-century music. Why

Schuijer.indd Sec1:2 10/2/2008 10:30:21 PM

In Rotterdam, the dual image of American analytical practice was prevalent in

Schuijer.indd Sec1:3 10/2/2008 10:30:21 PM

Schuijer.indd Sec1:4 10/2/2008 10:30:21 PM

Schuijer.indd Sec1:5 10/2/2008 10:30:21 PM

Example 1.1b. Non-obvious combinations related to the opening melody of

Schuijer.indd Sec1:6 10/2/2008 10:30:22 PM

Example 1.1d. Non-obvious six-tone combinations.

Schuijer.indd Sec1:7 10/2/2008 10:30:22 PM

ma marc. tutti arco

Schuijer.indd Sec1:8 10/2/2008 10:30:22 PM

411 3 412 3 3 413

411 arco 412 413 pizz.

Schuijer.indd Sec1:9 10/2/2008 10:30:23 PM

Example 1.3a. A source of inspiration: a recurring group of intervals in the twelve-

which owes to him its reputation as an American center of musical innovation.

Schuijer.indd Sec1:10 10/2/2008 10:30:23 PM

Schuijer.indd Sec1:11 10/2/2008 10:30:23 PM

Schuijer.indd Sec1:12 10/2/2008 10:30:23 PM

A Short Detour to Schenker

9. For a general reflection on the relationship between Schenkerian theory and

Schuijer.indd Sec1:13 10/2/2008 10:30:24 PM

Example 1.5b. Heinrich Schenker’s analysis of a song by Robert Schumann:

compositions. Small-scale patterns thus mark stages in the unfolding of larger

Schuijer.indd Sec1:14 10/2/2008 10:30:24 PM

Schuijer.indd Sec1:15 10/2/2008 10:30:25 PM

new start—is perfectly illustrated by classical themes in period form;13 in “Aus

Schenker opened the way for a deeper understanding of musical structure