Documentos de Académico
Documentos de Profesional
Documentos de Cultura
ISBN-13: 978–1–58046–270-9
ISBN-10: 1–58046–270-7
ISSN: 1071–9989
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.
Korien de Boer
Preface xv
Acknowledgments xvii
3 Operations 49
4 Equivalence 84
5 Similarity 130
6 Inclusion 179
8 Mise-en-Scène 236
Index 293
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
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.
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.
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.
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)
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
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
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
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.
immobile
segment 3
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,
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.
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.
© Copyright 1957 by Hawkes & Son (London) LTD. Reprinted by kind permission of Boosey &
Hawkes Music Publishers LTD.
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.
© Copyright 1957 by Hawkes & Son (London) LTD. Reprinted by kind permission of Boosey &
Hawkes Music Publishers LTD.
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.
∧ ∧ ∧
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)
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.
Forte liked the systematic, generalizing tendency of Schenker’s late output, and
saw his own role in the light of this tendency:
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).
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)
Schuijer.indd Sec1:19
Institute Representation Academic level Textbooks/manuals used
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)
Schuijer.indd Sec1:20
Institute Representation Academic level Textbooks/manuals used
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)
Schuijer.indd Sec1:21
Institute Representation Academic level Textbooks/manuals used
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
22 pitch-class set theory: an overture
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.
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.
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,
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.
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
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
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
0 1 2 3 4 5 6 7 8 9 10 11
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
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
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)
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.”
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
And for the PIC represented by the pitch interval from F♯4 to E♭4, we write 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
9 3
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,
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
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.
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)
or:
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:
and:
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”:
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)
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).
Used by permission of Belmont Music Publishers, Pacific Palisades, CA 90272.
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:
• 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)
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).
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:
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.
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
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
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.
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:
A' = ELEMENT − A
(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
S I R RI
I S RI R
R RI S I
RI R I S
Transposition
(9 + 4) mod 12 = 13 mod 12 = 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:
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.
(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
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)
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 by permission of Belmont Music Publishers, Pacific Palisades, CA 90272.
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.
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.
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.
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.
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.
(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.
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
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
(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).
Copyright © 1998 by Yale University.
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.
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
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 3.7. 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.
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
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)
(a)
(b)
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.
(a) ∧
1
[]
∧
5
∧
5
(b) ∧
1
(c) ∧ ∧
1 5
∧
5 ∧
1
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,
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)
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.
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
(a)
= ± 160
(b)
=
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:
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.
© 1956 by Universal Edition A.G., Wien/UE 12012.
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)
PI (14, 16)
PI (5, 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:
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
IPI (9,9) (9 + 8) = 9 − 8 = 1
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):
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 (6,10) ((6 + 11) mod 12) = (10 − 11) mod 12 = −1 mod 12 = 11
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).
{(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).
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
5 1 4 3
-3 -3 -5 -2 -7
x7 x2 x3 x-4
35
2 12 8 28
-9 -15 -12
-6
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}:
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.
etc.
molto espress.
28. As a consequence, the first tone should have the number 0. Babbitt (1992, 1960),
too, let order numbers run from 0 through 11.
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.
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
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)
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:
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.
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.
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
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.).
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.
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
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)
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.
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).
10/2/2008 10:30:57 PM
equivalence 91
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.
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:
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.
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.
(0)
{C , D , D , E , G , G }
(0)
{F, F , A , B , D , D}
(0)
{C , E , E , F} {F , G , A , B}
(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
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.
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:
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.
“Operational Equivalence”
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
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).
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.).
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).
(a)
4 3 5 3 5 4 5 4 3
(b)
3 4 5 4 5 3 5 3 4
“The interval array of one [PC set is] equivalent by cyclic reordering to the retro-
version of the other.” (Chrisman 1971, 70; my italicization)
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.
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”:
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):
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:
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.
(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)
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.
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
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 2 Select the PC sets beginning with 0, and having the smallest last 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}.
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
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.
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).
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.
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)
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).
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.
γ β α
α β γ
Example 4.9. The first two bars of Schoenberg’s Piano Piece, Op. 33a.
Used by permission of Belmont Music Publishers, Pacific Palisades, CA 90272.
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.
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).
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
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
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.
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.
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♯
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:
Schuijer.indd Sec1:119
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
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.
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
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.
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
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-
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.”
PSC
PCINT
FB
Example 4.11. Class types of pitch sets, after Robert Morris (1995).
(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).
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.
© 1953 by Universal Edition A.G., Wien/UE 14325.
© Copyright 1926 by Hawkes & Son (London) LTD. Reprinted by kind permission of Boosey &
Hawkes Music Publishers LTD.
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.
hrp + clar.
3 Coll. I
3
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)
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.
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.
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
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:
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
test related research tools, with the expectation that useful general analytic
methods and descriptive techniques will result. (Forte 1964, 178)
[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)
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.
(a)
6-Z49 5-16
[4,6,9,10,0,1] 6-Z19
[1,2,5,6,8,9]
333
(b)
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.
Copyright © 1973 by Yale University.
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)
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:
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:
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).
5. Triads consisting of one consonance, one mild dissonance, and a sharp one:
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.
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.”
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
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).
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).
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:
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
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).
© 1922 by Universal Edition A.G., Wien/UE 5888, UE 5889.
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)
γ: 100110
δ: 100011
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).
Table 5.1. Distributive patterns in the relation 5:1 (Forte: second-order maxi-
mal similarity)
{4,5} {0,5}
{11,0} {0,4} {11,4} {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.
Table 5.2. Distributive patterns in the relation 5:1 (Forte: first-order maximal
similarity)
{11,0} {0,5}
{4,5} {0,4} {11,4} {11,5}
APIC 1 2 3 4 5 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)
{11,4}
{11,0} {4,6} {0,4} {11,6} {0,6}
APIC 1 2 3 4 5 6
{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
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:
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).
Table 5.4. Distributive patterns in the relation 6:4 (Forte: minimal similarity)
{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
{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)
{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
{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
[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)
ζ: 110121
λ: 113221
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.
{11,4}
{11,0} {4,6} {0,4} {11,6} {0,6}
APIC 1 2 3 4 5 6
{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
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.
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.
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.
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)
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.
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
(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.
{4,5}
{3,4} {3,5}
{2,3} {2,4} {2,5}
APIC 1 2 3 4 5 6
{3,4}
{2,3} {2,4} {4,8} {3,8} {2,8}
APIC 1 2 3 4 5 6
{4,5}
{3,4} {3,5}
{2,3} {2,4} {2,5}
APIC 1 2 3 4 5 6
{2,9}
{7,9} {4,9}
{2,4} {4,7} {2,7}
APIC 1 2 3 4 5 6
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
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.
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.
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 5.10. Partition function of the PC sets I(ζ) = {1,0,8,6} and λ = {0,1,4,6,9}
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).
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.
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)
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
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.
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.
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.
“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
23. In this definition, the numbers of instances per interval class (n, if not specified)
should actually have been given in subscript.
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:
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,
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)
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
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:
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=
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) =
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,
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.
or, alternatively,
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.
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).
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
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.”:
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).
{0,3,6}: 00200200200
{0,1,3,6,7,9}: 22422622422
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 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.
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.
An Analytical Exploration
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.
2
5
sempre
Example 6.1a. The main melody of Gabriel Fauré’s Élégie for Cello and Piano, Op. 24.
2
5
sempre
26
sempre
3. As defined by Friedmann (1985, 227). Morris adopted the term “contour class” in
1987, but replaced it with “cseg” (“contour segment”) in 1993.
A E Mm7 A B Mm7
(e) (f) (g)
E A E
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,
10
langsamer
rit.
Example 6.3a. Arnold Schoenberg, Piano Piece, Op. 11, no.1, mm. 1–11.
Used by permission of Belmont Music Publishers, Pacific Palisades, CA 90272.
''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
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.
T8
(b) m. 12 (e) m. 13
[]
T3
(c) m. 29
T3
T11
T0
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.
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.
Used by permission of Belmont Music Publishers, Pacific Palisades, CA 90272.
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:
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.
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).
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).
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.
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.
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.
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 6 (β)
148 152
Sequence: (interrupted)
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.
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.
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”:
composer being free to let one of the keys prevail, or to leave the tonal
impression undetermined.20
(clair)
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)
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.
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.
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)
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.
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:
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.
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.
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 '
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.
(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
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.
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.
(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
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
(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
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.
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
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):
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)
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)
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
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)
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
8va
W.W. 3
Tpt.
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.
Copyright © 1978 by Yale University.
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"
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.
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.
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)
12. Cf. the analysis of Opus 19, no. 6 made by Forte (1973, 97).
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.
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).
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)
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.
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).
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).
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)
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.
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:
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).
The musicologist Jan LaRue, whose specialization was the study of musical styles,
saw a new age dawning:
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 Digital Computer as a Servant of Research in the Humanities and Arts (Univer-
sity of California, Los Angeles; April 15–16, 1965)
7. This collection was published by Prentice Hall in 1967, under the title Computers in
Humanistic Research: Readings and Perspectives.
The Impact of the Computer on the Humanities and the Arts (Boston University,
November 4–5,1965)
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)
Geige
Klavier
5 col legno
weich gezogen
rit.
pizz.
Example 8.1a. Anton Webern, Piece for Violin and Piano, Op. 7, no. 1.
© 1922 by Universal Edition A.G., Wien/UE 6642.
PSGT
1 [] (Cello)
PSGT
2 [] (piano RH)
PSGT
3 [] (piano LH)
PSGT
1
SSGT 3 SSGT 1
SSGT 2
PSGT
2
PSGT
1
SSGT 5
SSGT 4
PSGT
2
PSGT
1
SSGT 8
SSGT 6
SSGT 7
PSGT
3
PSGT
1
SSGT 5
PSGT
3
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
(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
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
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).
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)
14. For the relationship between atonal music theory and positivistic philosophy, see
Davis 1995. More on this in the next section of this chapter.
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.
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).
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.)
[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
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.
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.
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.
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:
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
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.
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:
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)
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.
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).
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
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:
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.
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.
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:
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,
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).
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)
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).
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.
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.
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.
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
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.
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.
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)
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:
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.
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).
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 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
thinking of its scholars and specialists offer important keys to this educational
task. (Sand 1965, 73)
[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
———. 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.
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.
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.
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.
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:
Cambridge University Press.
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
Music Theory 4/1:62–71.
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.
———. 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
Music Theory 21/1:29–48.
———. 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.
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
Music Theory 5/2:224–73.
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.
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].
Perspectives of New Music 3/2:1–21.
———. 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.
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.
———. 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.”
Perspectives of New Music 21/2:535–39.
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.
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
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
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
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