219

SCALE NETWORKS AND DEBUSSY

Dmitri Tymoczko
It is frequently observed that over the course of the nineteenth century
the chromatic scale gradually supplanted the diatonic.
1
In earlier periods,
non-diatonic tones were typically understood to derive from diatonic
tones: for example, in C major, the pitch class F might be conceptual-
ized variously as the fifth of B, the leading tone of G, or as an inflection
of the more fundamental diatonic pitch class F. By the start of the twen-
tieth century, however, the diatonic scale was increasingly viewed as a
selection of seven notes from the more fundamental chromatic collec-
tion. No longer dependent on diatonic scale for their function and justifi-
cation, the chromatic notes had become entities in their own right.
Broadly speaking, composers approached this new chromatic context
in one of two ways.
2
The first, associated with composers like Wagner,
Strauss, and the early Schoenberg, de-emphasized scales other than the
chromatic.
3
Chord progressions were no longer constrained to lie within
diatonic or other scalar regions. Instead, they occurred directly in chro-
matic spaceoften by way of semitonal or stepwise voice leading.
Melodic activity also became increasingly chromatic, and conformed
less frequently to recognizable scales. Chromaticism thus transformed
not only the allowable chord progressions, but also the melodies they
accompanied.
JMT 48.2 8/2/07 2:53 PM Page 219
The second approach, associated with composers like Rimsky-
Korsakov, Debussy, and Ravel, preserved a more conventional under-
standing of the relation between chord and scale, but within a significantly
expanded musical vocabulary. New scales provided access to new chords,
while new chords, in turn, suggested new scales. The scalar tradition, as
I will call it, thus proposed a less radical, but also more hierarchically
structured, conception of musical space: scales continued to exert their
traditional influence in determining chord succession and melodic con-
tent, mediating between the musical surface and the underlying world of
the total chromatic. By contrast, the chromatic tradition tended toward a
flattening of musical space, as the chromatic scale exerted an increasingly
direct influence on harmonic and melodic activity.
This paper proposes a general theoretical framework for understand-
ing the scalar tradition in post-common-practice music. The argument has
three parts. Section I discusses a number of intuitive constraints on scale
formation, showing that the sets satisfying these constraints possess a
number of interesting theoretical properties. This constitutes a statics
of scalar collections. Section II turns to dynamicstechniques of mov-
ing between scales based on shared subsets or efficient voice leading. It
presents graphs depicting common-tone and voice-leading relations
among familiar scales, one of which is a three-dimensional Tonnetz for
an important class of seven-note scales. Section III applies this theoreti-
cal framework to the analysis of four of Debussys piano pieces. The con-
clusion suggests that the ideas developed here have wider analytical appli-
cations as well.
Though my chief analytical concern will be with Debussy, my broader
theoretical aim is to develop a set of concepts and structures useful for
analyzing a wide range of post-common-practice scalar music. This paper
draws examples from Debussy, Chopin, Liszt, Shostakovich, and
Rzewski. In previous work I have used similar ideas to analyze the music
of Ravel, Stravinsky, and several jazz improvisers.
4
Other theorists, such
as Elliott Antokoletz, Carlo Caballero, Clifton Callender, Ian Quinn, and
Daniel Zimmerman, have used similar methods to analyze Bartk, Al-
brecht, Faure, Scriabin, Reich, and Prokofiev.
5
Jose Antonio Martins uses
closely-related structures to study Bartk, although his approach is impor-
tantly different from mine.
6
The fact that similar methods can be used to
describe such a broad range of music suggests that we can indeed speak
of a scalar traditionand even a common practicein twentieth-
century music: a set of concepts and techniques common to a variety of
musical styles, including impressionism, jazz, and minimalism.
220
JMT 48.2 8/2/07 2:53 PM Page 220
221
I. Scalar Statics: Three Scalar Collections
(a) Scales, Sets, and Modes: Some Terminological Preliminaries
A scale is a series of pitches ordered by register. This ordering under-
writes a measure of musical distance distinct from the more general met-
rics provided by chromatic semitones and frequency ratios.
7
The interac-
tion between these contrasting metrics gives scalar music much of its
complexity. Example 1 illustrates, presenting a series of three-note chords
from Liszt. Understood in terms of the chromatic scale, the chords be-
long to three different set classes. Individual voices move by one of two
intervals as they pass from chord to chord. Understood in terms of Ex-
ample 1s seven-note octave-repeating scale, however, each chord is an
instance of the same set class: the triadic set class [024].
8
From this per-
spective, individual voices always move by the same intervala single
scale step. Analytically, then, these chords belong either to three differ-
ent types or to one single type, depending on whether we measure dis-
tance along the chromatic scale, or along the seven-note scale of Exam-
ple 1. I will use the terms chromatic distance and scalar distance to
refer to these two ways of measuring musical distance.
9
Any pitch-class set can be associated with an infinite number of non-
octave-repeating scales. There is, however, a unique octave-repeating
scale associated with it: the ascending registral ordering of all the pitches
belonging to all the pitch classes in the set. This infinite pitch series in
turn defines a unique circular ordering of pitch classes, through which the
pitch series repeatedly cycles.
10
(As Maria puts it in The Sound of Music,
the note Ti brings us back to Do.) In this article, the term scale
refers to such circular orderings of pitch classes. Consequently, we can
often dispense with talk of pitches and register, and speak instead about
relations between pitch classes themselves.
11
The scalar interval from
pitch class a to pitch class b is one ascending scale step if a immediately
Example 1. Parallel triads
(from Liszts Angelus! Prire aux anges gardiens)
a) mm. 177182
b) the underlying scale
JMT 48.2 8/2/07 2:53 PM Page 221
222
precedes b in the circular ordering; it is two ascending scale steps if a
immediately precedes a note that immediately precedes b, and so on.
Similarly, two notes form a scalar second or are separated by one
scale step, if they are adjacent in the ordering; they form a scalar third,
or are separated by two scale steps, if they are adjacent but for one note,
and so on. Note that scales in this sense do not have tonic notes, pitch
priority, or first or last notes. They are distinguished from sets chiefly in
that their circular orderings give rise to the non-chromatic distance met-
ric described above.
Since there is a unique octave-repeating scale associated with every
pitch-class set, and vice versa, I will use the same names (or variables) to
refer to both: thus the set S (or the collection S) refers to an
unordered collection of pitch classes, while the scale S is the unique
circular ordering of those pitch classes described in the preceding para-
graph. Octave-repeating scales and their associated sets share a great
number of properties: every subset of set S can be uniquely associated
with a subscale of scale S, and every voice-leading between sets S and
T can be uniquely represented as a mapping between the elements of
scales S and T. For this reason, it will sometimes be convenient to mix the
terminology of scales and sets, as in: the supersets of set S are scales
with property P. I trust this bit of linguistic shorthand will not confuse
readers about the underlying distinction between sets and scales.
12
The
terms scale class and scale type refer to classes of scales whose asso-
ciated sets are related by transposition.
13
Thus the C diatonic and D dia-
tonic scalesconsidered as circular orderings of pitch classesbelong
to the same scale class. Where the context is clear I will follow tradition,
using the term scale to mean scale class, as in the diatonic scale
and the octatonic scale.
Scales will be named according to their most familiar orderings.
14
This is a mere terminological convenience, and is not meant to privilege
any pitch or mode. The term mode will be used to describe a scale in
which a pitch class has been singled out as having priority over the oth-
ers. Though modes will not be a central concern of the theoretical por-
tions of the paper, they play a role in the analyses presented in Section
III. Diatonic modes are labeled using their familiar names: G natural
minor, D dorian, and so forth. Non-diatonic modes are named with
respect to the orderings in footnote 14. Thus the first mode of the C
acoustic scale (CDEFGAB) has C as its tonic note; the second
mode has D, and so forth.
(b) Three Scalar Constraints
Imagine that you were an early-twentieth-century composer trying to
devise scales and chords that were somehow similar to the scales and
JMT 48.2 8/2/07 2:53 PM Page 222
223
chords of the tonal tradition. What kinds of similarity would you be
interested in? How would you go about expanding the vocabulary of tra-
ditional classical practice?
This section suggests an answer to these questions, identifying three
constraints on scale-formation thatI believemay have influenced the
post-common-practice exploration of non-diatonic materials.
15
I begin
by describing these constraints; later sections show that the scales satis-
fying them are exceedingly well-known. My hope is that the intrinsic
plausibility of the constraints, the ubiquity of the objects they generate,
and their analytical utility will jointly support the claim that the con-
straints played a role in the development of twentieth-century music.
16
1. The diatonic seconds constraint. Example 2 presents a single
series of scalar intervals in the context of three different scales. The three
scales contain steps that are just one or two semitones large. This sug-
gests the following constraint:
DS (Diatonic Seconds): If the scalar interval between two notes is one
ascending step, then the chromatic interval between them is either one or
two ascending semitones.
We can think of the DS constraint as generalizing a purely conventional
feature of the diatonic scale; or we can think of it as expressing a deeper
perceptual fact, perhaps related to the critical bandwidth of the auditory
system.
17
The difference between these perspectives is not crucial to the
following discussion. What is important is that the DS constraint captures
a clear and salient way in which a non-diatonic scale can resemble the
traditional diatonic scale. This resemblance ensures an important degree
of consistency between the scalar and chromatic distance metrics de-
scribed in Section I(a).
18
Composers can therefore transport a scalar
Example 2. Three scalar passages (from the opening of
Debussys Fetes)
JMT 48.2 8/2/07 2:53 PM Page 223
224
interval pattern from one scale to another, or within a single scale by
means of scalar transposition, without radically transforming the size of
its steps. This is why the three melodies in Example 2 above are recog-
nizably similar.
2. The no consecutive semitones constraint. None of the scales
shown in Example 2 contain successive one-semitone steps. This sug-
gests a second constraint on scale formation:
NCS (No Consecutive Semitones): If the chromatic interval between
any two notes is two ascending semitones, then the scalar interval
between them is one ascending step; thus, a scale cannot contain succes-
sive one-semitone steps.
Where the DS constraint sets an upper limit on the size of a scales steps,
NCS sets a lower limit on the size of its leaps: if a scale does not con-
tain successive one-semitone steps, then any scalar interval that is larger
than a step (a leap) must span at least three chromatic semitones. The
two together ensure that there is a clear distinction between the size of a
scales steps and leaps.
19
This is illustrated in Example 2, where the
brackets identify the scalar leaps.
It is also possible to understand the NCS constraint as operating
directly on chords, or unordered sets. A set satisfies the NCS constraint
only if it does not contain an [012] trichord as a subset. (Thus, an octave-
repeating scale satisfies the NCS constraint if and only if its associated
set satisfies the NCS constraint.) The motivation for applying the NCS
constraint directly to sets is that the [012] trichord is often considered to
be a particularly dissonant object.
20
Perhaps for this reason, there are a
number of harmonically-rich, post-common-practice stylesincluding
impressionism, jazz, and minimalismin which sonorities containing
[012] trichords are rare.
21
One can think of the NCS constraint, when
applied to sets, as a way of modeling these styles: the sets satisfying NCS
constitute the harmonic space available to a composer who is willing to
countenance virtually any chord as a potential sonority, except for the
[012] trichord and its supersets.
Interpreting the NCS constraint as a constraint on chord formation
provides another reason to think that it might have influenced the choice
of scales themselves. For if an octave-repeating scale satisfies the NCS
constraint, then all of the unordered chords that can be constructed from
its elements also satisfy that constraint. This means that composers can
treat such scales pandiatonically, freely choosing elements from the
scale without generating collections that run afoul of NCS. Thus, insofar
as avoidance of [012] and its supersets was indeed an aspect of early
twentieth-century harmonic practice, and insofar as some of these com-
posers were interested in treating every scalar subset as a potential har-
JMT 48.2 8/2/07 2:53 PM Page 224
225
mony, it is natural that NCS might come to have an influence on the
choice of scales themselves.
22
3. The diatonic thirds constraint. The scales shown in Example 2
possess a third interesting property: notes with one note between them
(scalar thirds) are separated by either three or four semitones. This sug-
gests a final constraint on scale formation:
DT (Diatonic Thirds): If the scalar interval between any two notes is
two ascending steps, then the chromatic interval between them is three or
four ascending semitones.
The DT constraint again ensures an important degree of consistency
between scalar and chromatic distances. It permits a composer to trans-
port a single pattern of scalar intervals from one DT scale to another, or
within a single DT scale by scalar transposition, without radically chang-
ing the size of its two-step intervals. Consequently, all of the bracketed
intervals shown in Example 2 span three or four chromatic semitones.
This fact is particularly important in light of the Western classical tra-
ditions emphasis on tertian harmonies. The DT constraint ensures that
stacks of scalar thirds sound tertianthat is, they will be stacks of three-
or four-chromatic-semitone intervals. This property allows composers to
apply familiar routines of diatonic-scale composition in the context of a
wider range of scales, obtaining results that are different frombut not
too different fromfrom the music of the classical tradition. We can also
imagine early twentieth-century composers happening onto the DT scales
by the reverse process: beginning with a stack of three- and four-semi-
tone intervals, a composer might explore the various ways it can be trans-
formed into a scale by interleaving it with another such sonority. Such
investigations could well lead to the DT scales themselves.
The three scalar constraints just described are often evokedalbeit
tacitlyin accounts of the origin of the ascending form of the melodic
minor scale. Commentators frequently refer to the melodic awkward-
ness of the harmonic minor scales augmented second.
23
This awkward-
ness presumably derives from the harmonic minor scales violation of
the DS constraint. However, as Example 3 shows, there are a number of
alternatives to the harmonic minor scale that do not violate the DS con-
straint. Standard accounts simply assume that the ascending melodic minor
is the obvious choice from among these alternatives. Pressed to explain
this, historians would no doubt attribute melodic awkwardnessor per-
haps harmonic awkwardnessto those scales that add an additional
note to the harmonic minor scale. Examples 3(b) and 3(c) illustrate. This
awkwardness presumably reflects the fact that these scales violate the
NCS and DT constraints. The ascending form of the melodic minor scale,
by contrast, satisfies all three constraints. Indeed, it is the only scale sat-
JMT 48.2 8/2/07 2:53 PM Page 225
226
isfying them that contains both a minor triad and its leading tone. In this
sense it is the unique non-awkward alternative to the harmonic minor.
(c) The Locally Diatonic Scales
There are twenty-five types of scale satisfying the Diatonic Seconds
constraint. To understand them, it is helpful to observe that DS is inher-
ited by supersets. That is, if a scale S has this property, then so do all
scales containing S. (This is because you cannot make a scales steps
larger by adding notes to it!) The DS scales can therefore be character-
ized as supersets of the minimal DS scales. And the minimal DS scales
are precisely those without consecutive semitones.
24
These scales belong to just four types: diatonic, octatonic, whole-tone,
and acoustic.
25
In addition to satisfying the DS and NCS constraints, these
scales are locally diatonic within a three-note span: any three consec-
utive notes of one of these scales will be enharmonically equivalent to
Example 3. Scalar constraints in the genesis of the melodic minor scale
a) A harmonic minor
b) a hypothetical alternative
c) another alternative
d) A melodic minor, ascending
JMT 48.2 8/2/07 2:53 PM Page 226
some three-note span of some diatonic scale. (They therefore satisfy the
Diatonic Thirds constraint as well.) For this reason I refer to them as the
locally diatonic scales. We can expect significant and audible similarities
between traditional diatonic music and music that is based on locally dia-
tonic scales: stepwise locally diatonic melodies will use one- and two-
semitone intervals, while stacks of locally-diatonic scalar thirds will be
stacks of three- and four-semitone intervals. The four locally diatonic
scales represent natural objects of exploration for those early-twentieth-
century composers who wanted to expand the resources of traditional
tonal music without discarding such concepts as triad and scale step.
One point of clarification is in order. Although I have noted that all of
the DS scales can be represented as supersets of locally diatonic scales, I
do not think that it is always analytically profitable to do so. A number of
important twentieth-century scales satisfy DS, but not NCS or DT: Mes-
siaens modes 3, 6, and 7, the whole-tone plus one scale used by Bartk
among others, and the so-called bebop scales.
26
Whether to treat these
as supersets of the locally diatonic scales is a matter that very much de-
pends on analytical context. In Messiaens music, for example, the third
mode 912 [01245689T] is typically used as fundamental scale in its
own right, rather than as a superset of the whole-tone and acoustic scales.
Nevertheless, I do think that there is a substantial body of music in
which the locally diatonic scales have a privileged status. In such styles,
supersets of the locally diatonic scales typically appear as embellish-
ments of the locally diatonic scales. This is certainly true of the music of
Debussys early and middle periods. A similar point applies to much jazz
and minimalist music. The preceding discussion suggests numerous rea-
sons why this might be so. The conjunction of the DS, DT, and NCS
properties allow composers to obtain a wealth of nontraditional sounds
while continuing to compose in fairly traditional ways. They therefore
constitute an attractive set of tools for composers seeking to expand,
rather than replace, the vocabulary of traditional tonality.
(d) the Pressing Scales
The NCS constraint is inherited by subsets: if a set S does not contain
an 012 trichord, then neither do any of Ss subsets. We can therefore char-
acterize all the sets satisfying the NCS constraint as subsets of the maxi-
mal NCS sets. The situation is precisely the inverse of that which we
encountered with the Diatonic Seconds property: the fact that DS is
inherited by supersets led us to look for minimal DS sets. By contrast,
because NCS is inherited by subsets, we look for maximal NCS sets.
It can be shown that a set S is a maximal NCS set if and only if S, when
considered as a scale, possesses DT.
27
There are seven types of DT scale:
the four locally diatonic scales, the familiar harmonic minor scale, its
inversion (sometimes called the harmonic major scale because it can
227
JMT 48.2 8/2/07 2:53 PM Page 227
228
be obtained by raising the third degree of the harmonic minor), and the
620 [014589] hexatonic scale consisting of alternating one- and
three-semitone steps.
28
Each of these last three scale-types contains a
consecutive [0145] tetrachord, as illustrated in Example 4. None of the
four locally diatonic scales can contain this tetrachord, since its three-
semitone interval cannot be filled in without creating consecutive semi-
tones. I will call these scales the Pressing scales since the late Jeff
Pressing was the first person to write about them.
29
The Pressing scales represent a further extension of the locally dia-
tonic scales. Each permits the construction of tertian harmonies, andif
we count the three semitone intervalmore-or-less stepwise melodies.
30
Furthermore, these sets contain as subsets all and only those pitch-class
sets that do not themselves contain an [012] trichord. Consequently, com-
posers who are committed to NCS as a harmonic constraint will invari-
ably find themselves using subsets of these scales. A particularly impor-
tant corollary of this fact is that the seven scale types contain all the set
classes that can be formed by superimposing two triads of any quality
major, minor, diminished, or augmented. They are therefore the natural
Example 4. The seven Pressing scales
JMT 48.2 8/2/07 2:53 PM Page 228
scalar concomitants to the kinds of polytriadic sonorities much favored
by twentieth-century composers.
Thus a number of independent lines of investigation jointly converge
on the Pressing scales. One could arrive at them through purely melodic
processes, for instance by searching for the largest octave-repeating scales
not containing consecutive one-semitone steps. Or one could arrive at
these scales through purely harmonic processes, for instance by superim-
posing triadic sonorities. The DT constraint represents yet a third route
to these scales, one that emphasizes the correspondence between scalar
thirds and chromatic thirds. With so many independent considerations
pointing in the same direction, we should not be shocked to find the Press-
ing scales reappearing in a variety of musical and theoretical contexts.
(e) The Maximal Anhemitonic Scales
We conclude by discussing the complements of the DS sets. These are
the anhemitonic sets, since they cannot contain any semitones. (A set
contains a semitone if and only if its complements octave-repeating
scale contains a step of at least three semitones.) Since the locally dia-
tonic scales are the minimal DS sets, they have as their complements the
maximal anhemitonic set classes. These four set-classes are the only ones
whose octave-repeating scales satisfy the following, enlarged version of
the DS constraint:
DS+: If the scalar interval between any two notes is one ascending step,
then the chromatic interval between them is two or three ascending semi-
tones.
31
DS+ is similar to DS, except that chromatic distances have been in-
creased by one semitone each. The four scale types possessing this prop-
erty, and containing as subsets all the anhemitonic set classes, are the
diminished seventh chord, the 535 [02479] pentatonic scale, the 534
[02469] dominant ninth pentachord, and the whole-tone scale.
32
The pentatonic scale is, in addition, one of just three scale types satis-
fying an enlarged version of the DT property.
DT+: If the scalar interval between any two notes is two ascending steps,
then the chromatic interval between them is four or five ascending semi-
tones.
(The whole-tone and hexatonic scales also satisfy DT+, but in a less
interesting way, since their two-step intervals are always four semitones
large.) Example 5 shows that we can understand the pentatonic scale, like
the diatonic scale, as a stack of five-semitone intervals that has been per-
turbed so that it returns to its starting point without cycling through all
twelve pitch classes. In each case the perturbation is a minimal one. The
diatonic scale can be arranged as a stack of six perfect fourths (five-
229
JMT 48.2 8/2/07 2:53 PM Page 229
230
semitone intervals) plus one near fourth (or augmented fourth) of six
semitones. The pentatonic scale can be presented as a stack of four per-
fect fourths plus one near fourth (here a diminished fourth) of four
semitones. But whereas the five-semitone intervals in the diatonic scale
are all scalar fourths (lying three scale steps away from one another), the
five-semitone intervals in the pentatonic scale are scalar thirds. Thus the
second chord shown in Example 5 is not a fourth chord at all. Instead,
it is a stack of pentatonic thirds.
Example 6 presents a series of musical examples where what look like
fourths are more properly described as pentatonic thirds. Example
6(a), from the end of Debussys La fille aux cheveux de lin, shows a
passage in which diatonic fourths (left hand, m. 34) give way to penta-
tonic thirds (right hand, m. 35). (Related Debussian passages can be
found in Examples 23 and 28 below.) Example 6(b), from the second
movement of Ligetis Piano Concerto, features diatonic fifths in the right
hand and pentatonic fourths (the registral inversions of pentatonic thirds)
in the left. Example 6(c) shows the beginning of Miles Daviss So
What. The voicing that Bill Evans plays, which has become known as
the So What chord, is a pentatonic ninth chord, or a stack of five pen-
tatonic thirds.
33
Example 6(d) shows an excerpt from a Herbie Hancock
solo that linearizes a series of parallel pentatonic triads.
34
Finally, Exam-
ple 6(e) shows the chord formed by the open strings of a guitar, remind-
ing us that many fourth-based instruments are in fact tuned as penta-
tonic thirds. All of this music has as much right to be called tertian as
quartal, for the chords in question, though they contain five-semitone
intervals, are built out of scalar thirds.
Example 7 summarizes the main points of Section I in graphic form.
The upper black box contains those set classes whose associated scales
possess Diatonic Seconds. At the bottom of this box are the minimal DS
set classes, the locally diatonic scales. (Their supersets are shown, in con-
densed Fortean notation, in the columns above.) The lower black box
contains those set classes that satisfy the No Consecutive Semitones
constraint. At the top of this box are the Pressing scales, which are both
the maximal NCS sets and the only scales possessing the Diatonic Thirds
property. (Note that Example 7 collapses the harmonic major and har-
monic minor scales into a single harmonic scale type.) Finally, one can
Example 5. The diatonic and pentatonic scales as near fourth chords
JMT 48.2 8/2/07 2:53 PM Page 230
231
Example 6. Pentatonic thirds and diatonic fourths
a) Debussys La fille aux cheveux de lin
b) Ligeti, Piano Concerto, II, mm. 6061
c) Miles Daviss So What
d) Herbie Hancocks Eye of the Hurricane
(beginning of the fifth chorus)
e) the open strings of the guitar (cf. Example 6[c], m. 2)
JMT 48.2 8/2/07 2:53 PM Page 231
232
E
x
a
m
p
l
e

7
.

A

g
r
a
p
h
i
c

p
o
r
t
r
a
y
a
l

o
f

t
h
e

r
e
l
a
t
i
o
n
s

b
e
t
w
e
e
n

D
S
,
D
T
,
N
C
S
,
a
n
d

t
h
e

a
n
h
e
m
i
t
o
n
i
c

s
e
t
s
JMT 48.2 8/2/07 2:53 PM Page 232
take the complement of the DS sets (reading 111 as 11, 102 as 22,
and so forth) to obtain the anhemitonic set classes, with the maximal
anhemitonic set classes contained at the bottom of the box surrounded by
the heavy black line.
The forgoing discussion will remind some readers of Richard Parkss
theory of pitch-class-set genera. Very much in the spirit of the present
study, Parks identifies a series of scalar collections that he considers to be
central to Debussys harmonic vocabularyincluding the diatonic, octa-
tonic, and whole-tone scales. Appendix I compares the two theories in
some detail.
II. Scalar Dynamics: Common-Tone and Voice-Leading
Relations Between Scales
One notable feature of traditional tonal practice is that modulation pro-
ceeds by way of efficient voice leading between scales sharing a large
number of common tones. For example, the C diatonic scale shares six
common tones with the G diatonic scale; moreover, these two scales can
be linked by what Richard Cohn calls maximally smooth voice lead-
ingvoice leading in which only a single pitch class moves, and it moves
by only a single semitone.
35
In this section we investigate common-tone
and voice leading relationships among the scales discussed in Section I.
Section II(a) considers common-tone relationships. Sections II(b)(d)
consider voice-leading in somewhat greater detail.
Note that in talking about common tones and voice leading we will be
talking about properties that are shared by sets and the scales associated
with them. Since the vocabulary of set theory is admirably well-suited for
this, much of the following discussion will be couched in terms of col-
lections and sets. In many musical styles, however, these relationships
are manifested by objects that are clearly scalarcircular orderings of
pitch classes that define their own distance metrics, as discussed in Sec-
tion I(a). It is the conjunction of these two distinct sorts of attributes
the common-tone and voice-leading properties discussed here, and the
scalar properties identified in Section Ithat endows the following
investigation with much of its interest.
Finally, although a detailed discussion of this matter is beyond the
scope of this paper, it is worth saying a word about why the scales dis-
cussed in Section I are linked by the voice-leading and common-tone
relationships discussed here. The reason is that the DS, NCS, and DT
constraints identify scales that resemble the diatonic collectionwhich
divides the twelve-note chromatic scale into seven nearly equal parts.
36
Since the diatonic scale is maximally even, its scalar intervals come in
just two consecutive-integer sizes: its seconds span one or two semi-
tones; its thirds span three or four semitones; its fourths span five or
233
JMT 48.2 8/2/07 2:53 PM Page 233
six semitones, and so forth.
37
In Section I we saw that this feature of the
diatonic scale ensures an important degree of consistency between the
scalar and chromatic measures of musical distance, and that this consis-
tency is to some extent inherited by locally diatonic and Pressing scales.
Elsewhere, I have shown that this same feature ensures that the various
transpositions of the diatonic scale will be linked by unusually small
voice leadings.
38
The scales described in Section I, by virtue of their
resemblance to the diatonic scale, inherit these special voice-leading
properties.
39
Thus the evenness of the diatonic collection plays two
essential but independent roles in our inquiry: it ensures the consistency
between the scalar and chromatic distance metrics, and it accounts for the
unusually complex network of voice-leadings we investigate below.
Appendix II(a) explores this matter further.
(a) Common-Tone Relationships Between Scales
If S and T are sets, neither of which entirely contains the other as a
subset, then they can share at most one note less than the cardinality of
the smaller set.
40
Two sets, not related by inclusion, maximally intersect
when they share this maximum number of notes. Thus if a whole-tone
collection shares five notes with some other set, then the two maximally
intersect, no matter how many notes that other set has. Two set classes
maximally intersect if there exist maximally-intersecting members of
those two set classes. Finally, a set class maximally intersects itself when
two distinct forms of that set share all but one of their notes. Note that
maximal intersection is a term of art that simply records the fact that
two sets, not related by inclusion, share a maximum number of common
tones. The term itself carries no implications about voice leading or
registral realization.
Example 8 shows which of the four locally diatonic set classes maxi-
mally intersect. Dark lines connect maximally intersecting set-classes,
while the numbers next to the arrows identify how many distinct transpo-
sitions a given set maximally intersects. Thus the number 4 on the line
connecting the octatonic to the acoustic set class indicates that every
octatonic collection maximally intersects four distinct acoustic collec-
tions. The number 1 on the same line indicates that every acoustic col-
lection maximally intersects just one octatonic collection. We see that the
acoustic set class is central to the graph, maximally intersecting the other
three set classes. Every acoustic collection can share six of its seven notes
with some diatonic collection, six of its seven notes with some octatonic,
and five of some whole-tone collections six notes.
41
None of the other
locally diatonic set classes maximally intersect. Instead, they can share at
most one less than the maximum number of notes with the other two:
every whole-tone collection shares four notes with some diatonic and
some octatonic collection, and every diatonic collection shares five notes
234
JMT 48.2 8/2/07 2:53 PM Page 234
235
with some octatonic collection. Finally, notice that the diatonic set class
maximally intersects itself: every diatonic collection shares six notes
with two of its transpositions. This is not true of any of the other locally
diatonic set classes.
42
An acoustic collection shares five notes with its
nearest transposition (by major second); an octatonic collection shares
four notes with both of its transpositions; and the two whole-tone collec-
tions share no notes.
One can draw a similar graph for the seven Pressing set classes, as
Example 9 demonstrates. The right side of Example 9 resembles Exam-
ple 8, without the circular arcs. It shows that the acoustic set class maxi-
mally intersects the octatonic, whole-tone, and diatonic set classes. The
left side of the graph shows that the octatonic, acoustic, and diatonic set
classes all maximally intersect the harmonic set class. (Note that there are
two arrows leading to and from the harmonic set class, indicating distinct
octatonic
(828)
acoustic
(734)
whole-tone
(635)
diatonic
(735)
5 notes
6

n
o
t
e
s

6

n
o
t
e
s

1 6


2

2

2


4

1

4

n
o
t
e
s
4

n
o
t
e
s
5

n
o
t
e
s
6 notes: T5 & T7
Example 8. Maximal connections among the four locally
diatonic scales
JMT 48.2 8/2/07 2:53 PM Page 235
236

1

6


2

3

1

6

n
o
t
e
s


4

1

6

n
o
t
e
s

6

n
o
t
e
s


2

2





1

1





1

1

6

n
o
t
e
s





1

4





1

4

6

n
o
t
e
s




1

1







1

1

6

n
o
t
e
s
:

T
5

a
n
d

T
7
o
c
t
a
t
o
n
i
c
(
8

2
8
)
h
e
x
a
t
o
n
i
c
(
6

2
0
)
a
c
o
u
s
t
i
c
(
7

3
4
)
d
i
a
t
o
n
i
c
(
7

3
5
)
w
h
o
l
e
-
t
o
n
e
(
6

3
5
)
5

n
o
t
e
s

5

n
o
t
e
s


h
a
r
m
-
m
i
n
(
7

3
2
A
)
h
a
r
m
-
m
a
j
(
7

3
2
B
)

1

2

2

E
x
a
m
p
l
e

9
.

M
a
x
i
m
a
l

i
n
t
e
r
s
e
c
t
i
o
n
s

a
m
o
n
g

t
h
e

s
e
v
e
n

P
r
e
s
s
i
n
g

s
c
a
l
e
s
JMT 48.2 8/2/07 2:53 PM Page 236
maximal intersections with its two inversional forms.) The harmonic set
class in turn maximally intersects itself, as well as the hexatonic set class.
All of the set classes not directly connected by lines can share one fewer
than the maximal number of notes, with the exception of the whole-tone/
hexatonic pair, which can share only three of six notes.
Finally, note that the transpositionally-symmetrical (or T-symmetrical)
43
Pressing set classes are at the top of Example 9, while the fifth-based dia-
tonic set class is at the bottom. In between are the harmonic and acoustic
set classes. These two seven-note set classes each maximally intersect the
diatonic set class as well as two T-symmetrical set classes. For this rea-
son, acoustic and harmonic scales can serve to mediate between diatonic
and T-symmetrical scales, providing a smooth way of modulating
between them. In much the same way, acoustic and harmonic scales can
mediate between distinct T-symmetrical scales.
44
Thus the acoustic scale
might be described as equally diatonic, octatonic, and whole-tone; just as
the harmonic scales are equally diatonic, octatonic, and hexatonic. As we
will see, these descriptions provide an important key to the scales func-
tion in twentieth-century music.
(b) Voice Leading Between Scales
Examples 8 and 9 show the maximal intersections among different set
classes. We will now look in some detail at voice-leading and common-
tone relationships between individual collections. We begin by consider-
ing the three seven-note Pressing set classes: the diatonic, acoustic, and
harmonic. The six- and eight-note Pressing set classes, all of which are
transpositionally symmetrical, will be considered in Section II(d) below.
Example 10(a) identifies all the seven-note Pressing collections max-
imally intersecting the C diatonic collection. Example 10(b) does the
same for C acoustic, while Example 10(c) shows the seven-note Pressing
collections maximally intersecting A harmonic minor and C harmonic
major. Two features of these graphs immediately attract attention. First,
almost all of the maximally intersecting collections can be linked by
maximally-smooth voice leading. (The one exception is the dotted line
in Example 10(c), linking A harmonic minor to C harmonic major. It is
discussed in Appendix II(b).) Equally striking is the fact that these
maximally-smooth voice leadings come in inversionally-related pairs.
For example, the C diatonic collection can be transformed into the G dia-
tonic by shifting F to F; similarly, C diatonic can be transformed into the
F diatonic by shifting B to B. The two voice-leading motions FF and
BB are symmetrical around the C diatonic collections D/A axis of
inversional symmetry: I
4
takes F into B, F into B, and the entire C dia-
tonic collection into itself. Globally, the entire graph of Example 10(a) is
inversionally symmetrical: I
4
simply rotates the graph 180, exchanging
I-related collections. (A glance at Examples 10(b) and 10(c) shows that
237
JMT 48.2 8/2/07 2:53 PM Page 237
238
E
x
a
m
p
l
e

1
0
.

M
a
x
i
m
a
l

i
n
t
e
r
s
e
c
t
i
o
n
s

a
m
o
n
g

t
h
e

s
e
v
e
n
-
n
o
t
e

P
r
e
s
s
i
n
g

s
c
a
l
e
s
a
)

d
i
a
t
o
n
i
c
;

b
)

a
c
o
u
s
t
i
c
;

c
)

h
a
r
m
o
n
i
c
JMT 48.2 8/2/07 2:53 PM Page 238
239
they are similarly symmetrical around D/A.) The inversional symmetry
of the voice leadings is a byproduct of the inversional symmetry of the
sets themselves.
45
The question now arises: is it possible to subsume the graphs of Ex-
amples 10(a)(c) within one graph? Example 11 shows that it is.
46
The
three-dimensional scale lattice of Example 11 is a Tonnetz for the
seven-note Pressing collections. Like the familiar Oettingen/Riemann
Tonnetz, it is a graphic representation that elegantly captures the
maximally-smooth voice-leading possibilities between the relevant
sets.
47
Unlike the triadic Tonnetz, however, it has a complex geometrical
structure that cannot be perspicuously represented in two dimensions.
Furthermore, Example 11 depicts objects belonging to three different set
classes, whereas the Oettingen/Riemann Tonnetz contains triads belong-
ing to just a single set class. The differences between these three set
classes produces asymmetries that give Example 11 its distinctive geo-
C dia
A HM
A hm
D ac
B ac
Fs dia
Fs HM
Fs hm
E HM
E dia
B dia
A dia
E ac
G dia
E hm
D dia
G ac
A ac
B HM
B hm
dia= diatonic
ac = acoustic
HM = harmonic major
hm = harmonic minor

G

G
s

E

E
s
C

Cs
AAs

F

F
s

D

D
s

Example 11. A portion of the scale lattice for the seven-note
Pressing scales
JMT 48.2 8/2/07 2:53 PM Page 239
240
metrical structure: every diatonic collection is connected by maximally-
smooth voice leading to six different seven-note Pressing collections;
every acoustic collection is connected by maximally-smooth voice lead-
ing to four different seven-note Pressing collections; and every harmonic
collection is connected by maximally-smooth voice leading to only three
Pressing collections. For this reason, harmonic collections lie on the
corners of the lattice, acoustic collections lie on edges, while dia-
tonic collections lie within it.
48
The three set classes thus have different
degrees of connectedness, giving rise to the complex structure shown
in Example 11.
There are two ways to understand this lattice: as a series of stacked
cubes, and as a series of intertwining strands. We will consider each in
turn.
(c) The Stacked Cubes Interpretation
Example 12 shows one of Example 11s cubic units. The graph con-
tains eight collections: four diatonic, two acoustic, and one each of the
harmonic major and minor. (This graph was originally devised by
Richard Cohn, who brought it to my attention in a personal communica-
tion.) All of the collections in Example 12 share the four pitch classes
DAEB. The graphs Cartesian coordinates represent pairs of
semitone-related pitch classes: the x axis corresponds to the pair {0, 1},
here labeled {C, C}; the y axis corresponds to the pair {5, 6} (labeled
{F, F}); while the z axis corresponds to {7, 8} (G and G).
49
A collec-
tions coordinates determine which element of each pair it contains: thus,
all collections with an x coordinate of 0 contain pitch class C, while all
with an x coordinate of 1 contain pitch class C, and so forth. Since Ex-
ample 12 is a cube, it contains a collection for every choice of one ele-
ment from each pair. Finally, notice that collections on a given face
jointly share five common tones. On the two faces with three diatonic
collections, the common five-note collection is the familiar pentatonic
scale. The four collections on the corners of these faces are the only
Pressing scales containing that scale.
The lattice in Example 11 is composed of twelve cubes, each related
by transposition to the one shown in Example 12. The first cube, in the
lower left of the graph, is identical to that in Example 12. The second
cube is stacked on top of it and shares a face with it. Every collection on
the second cube is T
7
of some collection in the first.
50
Note, however, that
the first and second cubes are oriented differently in three-dimensional
space: the second is a (transposed) 120 rotation of the first around the C
diatonic/A diatonic diagonal.
51
Similarly, the third cube lies to the right
of the second; it again shares a single face with it, and is its (transposed)
120 rotation around the same diagonal. The fourth cube lies behind the
JMT 48.2 8/2/07 2:53 PM Page 240
241
third, and is oriented in the same way as the first cube: every collection
on the fourth cube is T
3
of the collection in the corresponding position on
the first cube. This pattern repeats to produce twelve transpositionally-
related cubes, after which it returns to its starting point. Since it is diffi-
cult to depict the entire structure in two dimensions, Example 11 shows
only an excerpt of the whole.
52
Example 11 also records shared stack of fifths subsets. Every dia-
tonic hexachord (set class 632 [024579], a stack of five chromatic fifths)
is contained by exactly two Pressing collections: both are diatonic and are
connected by lines in Example 11. Every pentatonic collection (a stack
of four chromatic fifths, set class 535 [02479]) is contained by exactly
four Pressing scales, three diatonic and one acoustic. These share a cubic
face on Example 11. Every stack of three chromatic fifths (set class 423
[0257]) is contained by eight collections, whichas discussed above
form one of Example 11s cubes. Every stack of two fifths (set class 39
[027]) is contained by twelve collections, which appear on two cubes that
share a face. Finally, there are twenty Pressing collections sharing a sin-
C dia
A HM
A hm
D ac
A dia
G dia
D dia
G ac

G

G
s

C

C
s


F

F
s

Example 12. Richard Cohns scale cube
JMT 48.2 8/2/07 2:53 PM Page 241
gle fifth. These are the sixteen collections contained by three consecutive
cubes on Example 11, plus four others: one octatonic, one hexatonic, one
harmonic major and one harmonic minor.
53
Careful investigation of Example 11 reveals many other interesting
relationships which we cannot discuss here. Appendix II considers some
of these in more detail.
(d) The Intertwining Strands Interpretation
So far, we have been considering Example 11 as a series of stacked
cubes. But it is equally instructive to view the graph as the product of two
intertwined strands, one diatonic and one non-diatonic. The diatonic
strand consists of the familiar circle of fifths: it takes a jagged path
through Example 11, beginning in the lower left and moving up, right,
and into the paper successively. The non-diatonic strand is a less familiar
cyclic structure that also involves fifth transposition. Beginning with G
acoustic, we find the following sequence of non-diatonic collections:
G acousticA harmonic majorA harmonic minor
D acousticE harmonic majorE harmonic minor. . .
This is again a circle of fifths, but the unit of sequential repetition is
three collections long. The two strandsone diatonic, one non-diatonic
are intertwined somewhat as the strands of a double helix. Their interac-
tions are complicated by the fact that they do not have the same shape:
the diatonic strand takes a right-angled turn after every step, whereas
the non-diatonic strand moves in a straight line through each acoustic
collection.
Example 13 unfolds the lattice of Example 11 along its non-diatonic
strand. The central core of this graph features the full cycle of thirty-six
non-diatonic seven-note Pressing collections. On the outside of this cen-
tral circle runs the diatonic circle of fifths. Another diatonic circle of
fifths can be found within the central circle. As can be seen from this
graph, the circle of non-diatonic collections touches the diatonic circle of
fifths at two different points. The A harmonic major collection, for exam-
ple, is connected by maximally-smooth voice leading to the A diatonic
collection. The next collection in the non-diatonic circle, A harmonic
minor, is related by maximally-smooth voice leading to the C diatonic
collection. Thus moving one step forward on the non-diatonic circle
(from A harmonic major to A harmonic minor) brings us to a collection
related by maximally-smooth voice leading to a diatonic collection three
steps backward on the circle of fifths (C diatonic, which has three fewer
sharps than A diatonic). Moving another step forward on the non-
diatonic circle brings us to D acoustic, which is related by maximally-
smooth voice leading to both G diatonic (a fifth above C) and A diatonic.
In this way the non-diatonic circle of fifths continues to touch upon the
242
JMT 48.2 8/2/07 2:53 PM Page 242
243
diatonic circle of fifths at two distinct points. This is yet another reason
why the lattice is difficult to portray in two dimensions.
As Example 9 demonstrated, only the collections on the non-diatonic
strand maximally intersect the T-symmetrical Pressing collections.
Example 13 provides more information on these relationships. As can be
seen from this graph, the non-diatonic cycle divides into triplets, each
maximally intersecting the same octatonic collection. For example, A
harmonic minor, D acoustic, and E harmonic major all maximally inter-
sect Octatonic Collection II. (These triplets are collinear on Example 11.)
Adjacent harmonic-major and harmonic-minor collections maximally
intersect the same hexatonic collection. Acoustic collections maximally
intersect alternate whole-tone collections. The most efficient voice lead-
ings between these collections involve what Callender (1998) calls
C
H
M

C
h
m

F
a
c

G
H
M

G
h
m

C acD HM
D
hm

G
a
c

A
H
M

A
h
m

D
a
c

E
H
M

E
h
m

A
a
c

B
H
M

B
h
m

E
a
c

F
s
H
M

F
s
h
m

B
a
c

D
f
H
M

D
f
h
m

F s
a c A f H M A f h m

D
f
a
c

E
f
H
M

E
f
h
m

A
f
a
c

B
f
H
M

B
f
h
m

E
f
a
c

F
H
M

F
h
m

B
f
a
c

o
ct 3
oct 1
o
c
t
2
o
c
t
3
o
c
t
1
o
c
t
2
o
c
t
3
o c t 1
o
c
t
2
o
c
t
3
o
c
t
1
o
c
t
2
wt 2 hex 1
w
t
1
h
e
x
4
w
t
2
h
e
x
3
w
t
1
h
e
x
2
w
t
2
h
e
x
1
w
t
1
h
e
x
4
w
t 2
h e x 3
w
t
1
h
e
x
2
w
t
2
h
e
x
1
w
t
1
h
e
x
4
w
t
2
h
e
x
3
w
t
1
h
e
x
2
C

G
(=
)
G
D
(=)
D

A
(
=
)
A

E
(
=
)
E

B
(
=
)
B

F
s
(
=
)
F
s

C
s
( =
)
C s A f
( = )
A
f

E
f
(
=
)
E
f

B
f
(
=
)
B
f

F
(
=
)
F

C
(
=
)

C
(
=
)
C

G
(
=
)
G

D
(
=
)
D

A
(
=
)
A

E
(
=
)
E

B
( = )
B

G
f
(
=
)
G
f

D
f
(
=
)
D
f

A
f
(
=
)
A
f

E
f
(
=
)
E
f

B
f
(
=
)
B
f

F
(=)
Example 13. A circle of fifth-related, nondiatonic scales
JMT 48.2 8/2/07 2:53 PM Page 243
244
split and merge transformations. In moving from the whole tone col-
lection to the acoustic, or from the acoustic to the octatonic, a single pitch
class splits into its two chromatic neighbors. Conversely, in moving
from the octatonic to the acoustic, or from the acoustic to the whole tone,
the two notes of an [02] dyad merge into the pitch class between
them.
54
Voice leadings between T-symmetrical and harmonic collections
involve a similar, though more complex, process of splitting and
merging: here, either an [013] trichord becomes the semitone spanned
by its outer notes, or the reverse process occurs.
55
One problem with Example 13 is that a single T-symmetrical collec-
tion appears in multiple places on the graph. The example does not,
therefore, provide a particularly useful picture of how one might use
seven-note scales to modulate between T-symmetrical scales. Example
14 rectifies this by displaying the non-diatonic Pressing collections in
matrix form. The rows display all the collections maximally intersecting
a given octatonic collection. The columns display all the collections max-
imally intersecting a given hexatonic or whole-tone collection. The table
provides two modulatory routes between any octatonic/whole-tone pair,
or between any octatonic/hexatonic pair. G acoustic and D acoustic, for
example, maximally intersect both the Octatonic I and Whole-Tone I col-
lections, while E harmonic minor and G harmonic major maximally
intersect both the Octatonic II and Hexatonic II collections. There are no
Pressing collections maximally intersecting both a whole-tone and a
hexatonic collection. Instead, one must travel between them by way of
two-collection path involving both a harmonic and an acoustic scale.
III. Scales in Debussy
We have now developed a rather formidable technical apparatus for
understanding scales and their relationships. What remains to be shown
is that this apparatus is analytically useful. This section considers four of
Debussys piano compositions, all composed within a relatively short
Example 14. Maximal intersections involving T-symmetrical scales
WT I WT II HEX I HEX II HEX III HEX IV
OCT I D ac E ac D hm B hm A hm F hm
G ac B ac F HM E HM C HM A HM
OCT II F ac D ac F hm E hm C hm A hm
B ac A ac B HM G HM E HM C HM
OCT III E ac C ac B hm G hm E hm D hm
A ac F ac D HM B HM A HM F HM
JMT 48.2 8/2/07 2:53 PM Page 244
time: two preludes, Les Collines dAnacapri and Le vent dans la
plaine, both written in 1910; LIsle Joyeuse, written in 1904; and
Cloches travers les feuilles, the first in the second series of piano
Images, written in 1907. These pieces show progressively more complex
methods of scalar organization. Les Collines dAnacapri is relatively
simple, featuring a small network of seven-note scales surrounding a cen-
tral diatonic collection. Le vent dans la plaine resembles Collines,
but uses an expanded network that exploits the whole-tone scales trans-
positional symmetry. LIsle Joyeuse is a much more sophisticated piece,
involving a central acoustic scale and a greater variety of methods of
scalar organization. Finally, Cloches travers les feuillesthe most
difficult of the works we will examineuses a much wider range of
scale-to-scale transformations, and is rather more elusive in its large-
scale organization.
The following analyses will not aspire to completeness. Instead, they
will attempt to demonstrate that Debussy makes systematic use of the
voice-leading and common-tone relations between Pressing scales. My
goals are twofold: first, to demonstrate that the scales discussed in Sec-
tion I of this paper do indeed serve, to a good first approximation, as the
basic scalar vocabulary of at least some of Debussys pieces; and second,
that Debussy utilized the rich network of common-tone and voice-
leading relationships between these scales. I will also be concerned, on
occasion, to point out that the objects under discussion are indeed scales
rather than unordered collections. (See, in particular, the discussion of
Example 23, below.) But this will not be a major focus of the discussion.
For the most part, the scalar nature of Debussys music is so obvious as
to require little explicit commentary: his melodies tend to proceed in
stepwise fashion, his chords tend to be tertian, and his harmonies fre-
quently move in parallel within a scale. I trust that this is apparent even
to readers only modestly familiar with Debussys music.
Before proceeding, I should say a few words to allay the worry that
scales are, in general, very difficult or impossible to perceive. Although
it can sometimes be hard to identify scalesas for instance when they
are played as chords whose notes all lie within the span of a single
octavemy own experience is that Debussys scalar vocabulary is quite
perceptible. There are a number of reasons why this is so. First, his vo-
cabulary is relatively limited. He makes most frequent use of just four
scalesdiatonic, pentatonic, acoustic, and whole-tone. Other scales,
such as the octatonic, harmonic major, and harmonic minor, appear much
less frequently. (Debussy almost never used the hexatonic scale.) Second,
many of these scales are familiar, or otherwise distinctive, musical ob-
jects. Obviously, we are well-acquainted with the diatonic scale and its
modes. The whole-tone scale is intrinsically a highly distinctive musical
object, due to its limited interval content and its extreme transpositional
245
JMT 48.2 8/2/07 2:53 PM Page 245
246
E
x
a
m
p
l
e

1
5
.

T
h
e
m
a
t
i
c

i
d
e
a
s

i
n

t
h
e

o
p
e
n
i
n
g

o
f

L
e
s

C
o
l
l
i
n
e
s

d

A
n
a
c
a
p
r
i

JMT 48.2 8/2/07 2:53 PM Page 246

symmetry. And the pentatonic scale is not only aurally distinctive but also
familiar from Western and non-Western folk sources. Third, Debussy
tends to use some of his scales in just a few characteristic modes. In par-
ticular, the acoustic scale typically appears in the mode equivalent to the
mixolydian mode with raised fourth degree.
56
(Other modes of the scale
are occasionally found, however.) Fourth, Debussy tends to change
scales reasonably slowly. One often hears several measures of music con-
forming to a single scalar collection, in which all of the scales pitches
appear. In some casessuch as Voilesa single scale provides the
pitch material for an entire musical section. Finally, Debussy often uses
scales as melodic entities. Indeed, almost all of the pieces discussed
below involve fairly extensive passages in which scales are explicitly
stated melodically.
It should also be emphasized that perceptibility is not the only crite-
rion for evaluating musical analyses. Analysis does not simply describe
the way we hear music; it can also show us how the music was made. It
is possible to interpret the graphs presented in Section II as depicting an
abstract, precompositional space within which Debussys music moves.
57
The music may on occasion move quickly within this space, jumping
between scales without clearly demarcating them. It may sometimes be
hard to follow these transformations by ear. But this does not show that
scales are, in these circumstances, analytically useless. A rapid series of
closely-related scales can produce striking aural effects, even though the
listener cannot explicitly identify the scales used in the progression.
Scale-based analysis can, in turn, show us how to produce such effects.
(Analysis in this sense shows composers how to steal from each other.)
One might well compare Debussys procedures here to the Impressionist
painters subtle use of closely related hues. In both cases, we perceive a
single blended whole. Yet as theorists, analysts, or artists we may legiti-
mately want to look beneath the dazzling, multicolored surface to the
technical procedures animating it.
58
(a) Les Collines dAnacapri
Les Collines dAnacapri begins, as Example 15 shows, by introduc-
ing three thematic ideas: a rising pentatonic figure; a descending scale-
fragment touching on six notes of the diatonic scale; and a melody,
joyeux and lger, that starts with a six-note diatonic scale-fragment and
ends by emphasizing the black-note pentatonic scale. (These three ideas
are identified in the example as Themes , , and , respectively.) The
first portion of the piece is purely and placidly diatonic. Its sixty-one
measures involve just a few departures from the tonic B major
59
: m. 9
introduces two accidentals, suggesting familiar chromatic passing
motion between ii
7
and I
6
; mm. 2429 present the six pitch classes
ABCDEF, common to both B harmonic minor and Octatonic
247
JMT 48.2 8/2/07 2:53 PM Page 247
248
E
x
a
m
p
l
e

1
6
.

R
e
h
a
r
m
o
n
i
z
a
t
i
o
n

i
n

t
h
e

c
o
n
c
l
u
d
i
n
g

p
o
r
t
i
o
n

o
f

L
e
s

C
o
l
l
i
n
e
s

d

A
n
a
c
a
p
r
i

JMT 48.2 8/2/07 2:53 PM Page 248

249
Collection III (with measure 28 interposing a D
7
chord that belongs nei-
ther to these collections nor to B major); while measures 46 and 51
involve straightforward chromatic passing notes.
The thematic ideas of Example 15 are largely absent from mm. 2461,
which has the character of an extended episode. When the themes return
in the last third of the piece, they are detached from their original B major
environment and set in the context of new scales. Example 16 shows how
the end of the piece cycles through, and reharmonizes, the themes of
Example 15. The second thematic cycle begins in m. 66 in G natural
minor. Halfway into Theme , however, the music replaces E with E,
producing G dorian. (Theme does not appear in this cycle.) The third
cycle begins in m. 73. The B major scale has again been perturbed by the
shift of a single pitch by a single semitone: D has replaced D, yielding
E acoustic. Theme again leads directly to Theme , which, touching on
D, neutralizes the E acoustic scale. (There is an ascending chromatic
line in the left hand of m. 77.) Measures 7880 present an oscillation
involving yet another single-semitone displacement: G alternates with
G, perhaps implying B major and B harmonic major respectively. (F
acoustic and Octatonic Collection III are also possible interpretations
here.
60
) Finally, m. 86 presents Themes and in the context of B
mixolydian, shifting B majors A to A.
Example 17 graphs the scales that I have identified in the latter por-
tion of the piece. The graph is a subgraph of Example 10(a): it contains
four of the six scales maximally intersecting the tonic B major. Several
features of this example deserve comment. First, it is centered, in the
sense that all of the scales on the graph maximally intersect a single col-
lection. Second, this central collection also serves as a stable tonic scale,
B diatonic
DDs

G

G
s


A

s


E

E
s

E acoustic
Fs diatonic
E diatonic
B harmonic
major
Example 17. Scales used in Les Collines dAnacapri
JMT 48.2 8/2/07 2:53 PM Page 249
250
to which the music repeatedly returns. These are common, though not
universal, features of Debussys use of scales.
61
(Cloches travers les
feuilles, discussed below, provides an example of a non-centered scale
network.) Third, the scales in Example 17 are presented with different
degrees of explicitness. The diatonic and acoustic scales are all complete,
and are given relatively extensive musical treatment. The harmonic major
(like the octatonic) is incomplete, fleeting, and ambiguous. This is again
typical of Debussys style, which makes frequent use of the diatonic,
whole-tone, and acoustic scales, but touches only occasionallyand
often ambiguouslyon octatonic and harmonic scales.
There is one further feature of Collines that distinguishes it from
Debussys other music. This is the way the themes themselves articulate
the subsets shared by its maximally intersecting scales: Theme , which
contains the six pitch classes common to the B diatonic and F diatonic
collections, appears as part of both scales. Likewise, Theme contains
the six pitch classes common to B diatonic and E diatonic, and again
appears within both scales. Finally, Theme contains the five pitch
classes common to the three scales in which it appears: B diatonic, E dia-
tonic, and E acoustic.
62
We might therefore say that Collines thema-
tizes the maximal intersections between its collections: not only does it
present each of its themes, with a fair degree of systematic rigor, in each
of its possible scalar contexts, but it also constructs its themes out of
exactly the notes common to the relevant scales.
63
(b) Le vent dans la plaine
The opening of Le vent dans la plaine, outlined in Example 18, is
similar in structure to the end of Collines, though its scales are less
clearly articulated. The piece begins with an ostinato pattern involving
the pitches BC. Inside this ostinato there is a simple four-note melody
that may suggest a pentatonic scale. If we take B as the primary pitch of
this section, and if we trust the key signature, then the mode is B phry-
gian. (It is also possiblein light of what eventually happensto hear
E as primary, in which case the mode is E natural minor.) Measure 5
shifts E to E, which I take to imply the third mode of G harmonic
major. As in mm. 7880 of Collines, the two forms of the note alternate
before settling, in measure 7, on the ostinato and its implied return to B
phrygian.
Measure 9 introduces a new theme: a series of descending seventh
chords over parallel fifths in the bass. C has replaced C, yielding E
dorian. (Retrospectively, it may be possible to conceive the preceding
music as an extended dominant chord, with its B-centered music lead-
ing to Measure 9s tonic E.) Four measures later, the ostinato reinstates
the C. In measure 15, B becomes B; if we hear this new B as cen-
tral, then the mode is the seventh mode of the C acoustic scale. The pen-
JMT 48.2 8/2/07 2:53 PM Page 250
251
tatonic tune returns in the new acoustic context. Measure 18 briefly
restores the original ostinato. In measure 19, the acoustic scale reappears;
the pentatonic tune has now been transposed up by scale step.
Measures 2022, shown in Example 19, present a more complex set
of scalar interactions than we have thus far considered. The acoustic scale
in mm. 1520 maximally intersects the whole-tone collection in measure
22. This suggests a continuation of the compositional logic that has thus
far animated the piece. However, this progression of maximally intersect-
ing scales is interrupted by the surprising and more distantly-related
Example 18. Thematic and harmonic summary of the opening of Le
vent dans la plaine
JMT 48.2 8/2/07 2:53 PM Page 251
252
E
x
a
m
p
l
e

1
9
.

L
e

v
e
n
t

d
a
n
s

l
a

p
l
a
i
n
e
,

m
m
.

1
9

2
6
JMT 48.2 8/2/07 2:53 PM Page 252
white-note scale in measure 21. What is the reason for this dramatic shift,
so different in character from the extremely smooth harmonic motion
that has characterized the piece up to now?
Example 19 shows that this measure involves a complex set of scalar
affiliations. The first three beats of m. 21 involve only white notes, sug-
gesting D dorian. However, the fourth beat of the measure adds a D
which, taken together with the earlier music, implies G acoustic.
64
This
fleeting acoustic sound gives way, in the next measure, to the maximally
intersecting Whole-Tone I collection.
65
Measures 21 and 22 thus suggest
a sequence of maximally intersecting scales: D dorianG acoustic
Whole-Tone I. This sequence is the inversion of the large-scale progres-
sion we have already discussed, B phrygianC acousticWhole-
Tone I.
66
The two distinct progressions meet in the whole-tone scale of
measure 22. Thus, what sounds like an interruption, the sudden appear-
ance of the white-note diatonic collection in measure 21, can be analyzed
as a continuation of the musics underlying logic.
The subsequent music reinforces this reading. The whole-tone scale
of measure 22 gives way to D dorian white note music in m. 23. (The
parallel fifths in the left hand recall the E dorian of mm. 912.) In m. 24,
G acoustic returns, suggesting two dominant ninth chords (one with
lowered fifth) a whole step apart.
67
In m. 25, the music slides chromati-
cally, transposing mm. 2223 up by half step. This produces the Whole-
Tone Collection II, and brings back the E dorian collection heard in mm.
912. (Again, this connection is reinforced by the parallel fifths in the
bass.) The transposition associates the E dorian mode with Whole-Tone
Collection II, just as the D dorian mode of mm. 21 and 23 was associated
with Whole-Tone Collection I. If the progression from E dorian to
Whole-Tone Collection II were to proceed by way of maximally-smooth
voice leading, it would involve an A acoustic collection not heard in the
piece.
68
Example 20, which graphs all but one of the scales in the piece,
attempts to capture this analysis.
69
The upper-right portion of the graph
resembles Examples 10(a) and 17: it features a central diatonic scale
(here, B phrygian) connected by maximally-smooth voice leading to
three seven-note scales. Here, however, the graph is augmented by a
series of additional scales: Example 20s C acoustic collection maxi-
mally intersects a scale (Whole-Tone Collection I) that does not maxi-
mally intersect the original diatonic collection. Furthermore, this whole-
tone scale participates in a second progression involving additional
diatonic and acoustic scales, neither of which maximally intersects the
original B phrygian. Finally, E dorian is itself associated with the other
whole-tone scale. While these scales do not maximally intersect, they can
be interpreted an elided transposition of the D dorianG acoustic
Whole-Tone I progression heard earlier in the piece.
253
JMT 48.2 8/2/07 2:53 PM Page 253
254
Viewed in this light, Le vent dans la plaine can be seen to possess a
fairly traditional scalar organization. The opening of the piece shows how
the central B phrygian (or if one prefers, E natural minor) maximally
intersects a series of closely related scalesmuch as the opening of a
classical major-key sonata emphasizes the maximally-intersecting tonic
and dominant scales. As we move into the middle part of the piece, these
scales themselves give rise to additional (maximally intersecting) scales.
Harmonic motion becomes somewhat freer, and mm. 2022 feature more
dramatic scalar shifts than those of the pieces opening. What follows, in
mm. 2843, is a passage of largely nonscalar music (not discussed
above) that briefly touches on, and rejects, the distantly-related G phry-
gian scale (mm. 3437). The final portion of the piece (mm. 44ff.) returns
to the thematic and harmonic material of the opening. In short, the piece
demonstrates a traditional musical logic, though in the context of an
expanded scalar vocabulary.
(c) LIsle Joyeuse
The previous two analyses featured networks largely centered on a
single diatonic scale. LIsle Joyeuse also features a central collection, but
it is different from the previous pieces in three important ways. First,
LIsle Joyeuse uses only the four locally diatonic scales, rather than the
seven Pressing scales: the hexatonic and harmonic scales do not appear
in the piece. Second, the central collection in LIsle Joyeuse is an acoustic
rather than a diatonic scale. Third, whereas the previous two pieces fea-
tured maximally-smooth voice leading, LIsle Joyeuse makes greater use
of cardinality-changing shifts of scale. As we will see, these last two dif-
Bf phryg.
C

C
s
Bfff G{Gf,Af}
{
D
,
E
}

E
f

E
f
f

f

C
f

C
n
Cf acoustic
Ef dorian
Gf harmonic
major
G acoustic
D dorian
Whole-Tone
Collection I
Whole-Tone
Collection II
(elided Af acoustic?)
Example 20. Scales in Le vent dans la plaine
JMT 48.2 8/2/07 2:53 PM Page 254
ferences permit a greater structural integration of the whole-tone and
octatonic collections.
I hear the piece as composed of three large sections, each of which
begins by exploring the same network of maximally intersecting scales.
70
Example 21 lists the main thematic ideas of the opening. The music of
measures 16 is based on the whole-tone scale and serves to articulate the
structural divisions of the piece: it appears at the beginning and end of the
work, as well as at the beginning of the second section (mm. 5263).
Measures 79 are acoustic, and contain the main theme of the work.
Measure 10 is diatonic; if we consider A to be its main pitch, it is in the
lydian mode. Measures 1112 are more complex. Here I have bracketed
three spans of music, each suggesting different scales: the first contains
the six notes common to both the acoustic and octatonic scales; the sec-
ond adds an A which belongs only to the octatonic scale; while the third
is in A major, with a passing chromatic F. Example 21 shows that meas-
ure 66 presents the relation between A acoustic and Octatonic Collection
III more explicitly.
Example 22 graphs the scales used in the first section of the piece.
Square boxes identify scales, which are linked by straight lines represent-
ing maximal intersection. The ovals contain measure numbers; they are
linked with curved lines in places where the music progresses so as to
reveal the maximal intersections between their underlying scales. Exam-
ple 22, which we have derived purely analytically, is essentially identical
to Example 8, which we derived from theoretical considerations. In the
center is the A acoustic scale, from which the main theme is drawn. It
maximally intersects one whole-tone, one octatonic, and one diatonic
scale; this last diatonic scale in turn maximally intersects two further dia-
tonic scales. The beginning of the piece emphasizes the maximal inter-
sections between the scales, tracing out a path on Example 22 that relates
the collections to the central A acoustic scale. I find the correspondence
between Examples 8 and 22 to be remarkable. Theory and analysis are
unusually close here, indicating that the ideas we have been exploring do
indeed capture something important about this music.
The scales in Example 22 constitute the main harmonic region of
LIsle Joyeuse. They appear, in various combinations, at the beginning of
each of the three main sections, as well as at the end of the piece. Over-
all they account for about three fifths of the music (roughly 150 of 255
measures).
71
However, this harmonic region engenders further develop-
ments in each of the works three sections. An in-depth exploration of
these developments is beyond the scope of this paper, but it will be worth
briefly summarizing each section in turn.
Section 1: mm. 151. Measures 135 have already been discussed in
connection with Example 21. Measures 3651, outlined in Example
23(a), contain a series of parallel triads in B major, twice interrupted by
255
JMT 48.2 8/2/07 2:53 PM Page 255
256
mysterious and distantly related chords. Example 23(b) shows that that
these mysterious chords are in fact parallel first inversion pentatonic
triads. (To see why, note that DGB a root position pentatonic triad
since it is a stack of two pentatonic thirds. Transposing the bass of the
chord up by octave therefore generates a first inversion pentatonic triad,
GBD.
72
) The shift between the two scales occurs between the G minor
triad (a root position triad belonging to the B major scale) and the appar-
Example 21. The tonic region of LIsle Joyeuse
JMT 48.2 8/2/07 2:53 PM Page 256
257
ent root position G major triad, which is in fact a first inversion triad
belonging to the G pentatonic scale. Example 23(c) rewrites the passage
on two staves, so that the two G triads are notationally equivalent.
To understand Example 23(a) it is necessary to think in terms of scales
and scalar distances. The first part of the phrase features scalar trans-
position, here generating parallel diatonic triads in the key of B major.
The end of the passage has exactly the same structure, although the
chords now move along the pentatonic rather than diatonic scale. One
can therefore depict Example 23(a) as deriving from a single, continuous
processparallel motion within a scalethat is expressed, at a more
surface level, using two distinct scales. A key aspect of this reading is that
the progression from the diatonic G minor triad to the pentatonic G
major triad results not simply from the movement of chord voices, but
from the shift of the underlying scale. This is clear from the spatial lay-
out of Example 23(c), which portrays the shift between the chords not as
motion within a scale, but as a jump from one staff to another.

Whole Tone I
1-6 21-24
A acoustic
7-9 15-20
A lydian
(E diatonic)
10
Octatonic III
11-13
12-13

A diatonic
25-35

36-51

B diatonic
(w/ pentatonic
interruptions)
11 (6 notes)
(chromaticized)

F

{
E
,

F
#
}
B{Bb, C}

G

G
#
A#A D#D

Example 22. Scales in LIsle Joyeuse, mm. 151
JMT 48.2 8/2/07 2:53 PM Page 257
258
Section 2: mm. 52159. Example 24 graphs the collections used in
this portion of the piece. The section begins with a return to the whole-
tone music of the opening, this time colored by the appearance of C
major triads. Measures 6466 briefly touch on the music of Example 21
before launching, in m. 67, into a new theme. Some analysts, such as Roy
Howat, hear this new theme as articulating a major structural division in
the piece; I do not, since it continues to inhabit the same scalar region
graphed in Example 22.
Measures 99159 comprise what I have called the harmonic devel-
opment section. The music begins with parallel fifths articulating a pair
of alternating thirds (EC, EC). The music then falls into a long
whole-tone passage. Example 25 portrays these whole-tone scales as
possessing attenuated dominant-chord functionality.
73
The G acoustic
scale in m. 105 can be understood as an altered dominant of the
immediately-preceding fifths on C. When the music repeats, the G
Example 23. Parallel diatonic and pentatonic triads in Debussys LIsle
Joyeuse, mm. 3651
a) a reduction of measures 3651
b) diatonic and pentatonic triads
c) mm. 3643 as composed of parallel chords belonging to two scales
JMT 48.2 8/2/07 2:54 PM Page 258
259

Whole Tone I
52-63
A acoustic
64-6
A lydian
(E diatonic)
83-6
C TRIAD
[= C diatonic?]
141-4
C acoustic
145-7 156-7


G# acoustic
(V of C#)
105-8

Whole Tone II
125-33

148-55 158-9

Whole Tone I
117-24

133-40
G TRIAD [= G acoustic?]
(V of C)
115-6
Octatonic III
71-82
67-70 95-8

A diatonic
87-94


F

{
E
,

F
#
}
B{Bb, C}
{
B

B
b
,



F

F
#
}
{D, E}Eb {D#, E#}E
D#D
A
b

{
G
,

A
}


G

G
#
99-104
FIFTHS ON E, C#, C
(Cs minor/C lydian)
109-14
HARMONIC
DEVELOPMENT
SECTION
(V of A)
(V of C)
66
all scales maximally intersect
harmonic motion tends
to suggest dominant-
tonic relationships
(V of C#)
{I V}
V I
I V
Example 24. Scales in LIsle Joyeuse, mm. 52159
JMT 48.2 8/2/07 2:54 PM Page 259
260
E
x
a
m
p
l
e

2
5
.

D
o
m
i
n
a
n
t
-
t
o
n
i
c

r
e
l
a
t
i
o
n
s

i
n

m
m
.

9
9

1
5
9

o
f

L

I
s
l
e

J
o
y
e
u
s
e
JMT 48.2 8/2/07 2:54 PM Page 260
acoustic scale is replaced by a G triad (perhaps suggesting G acoustic)
that immediately gives way to the Whole-Tone I collection. Example 25
associates this music with the fifths on C. The reduction suggests that
the whole passage can be interpreted as an indecisive oscillation between
I and (altered) V, in both C and C. Of course, as this whole-tone music
continues (in mm. 117140), it begins to lose its dominant quality,
sounding more and more like a stable collection in its own right. But in
measure 141, the whole-tone music resolves triumphantly to a C major
triad, again suggestingif only retroactivelythat it had dominant
function. (Example 25 shows that this resolution articulates a IVI
progression spanning almost fifty measures.) This C major triad eventu-
ally gives way to the C acoustic scale (m. 145), which in turn dissolves
into Whole-Tone Collection II (m. 148).
74
A descending whole-tone scale
in mm. 158159 prepares for the return of the tonic A.
Section 3: mm. 160255. Example 26 graphs the final section of the
piece. The first 26 measures return again to the tonic network of scales
described in Example 22. Measure 186 marks a break in the musical tex-
ture, and the subsequent music demonstrates Debussys sophisticated
handling of scalar relationships. Example 27 summarizes. Measures
186195 (not shown on Example 27) are in C natural minor. Measure
196 turns to the whole-tone scale, and initiates a long melodic ascent by
whole-tones lasting until m. 220. The harmony of mm. 200207 implies
the E acoustic scale; measures 208215 repeat 200207 a major second
higher, now in F acoustic. Three features of mm. 196220 are notewor-
thy. First, these acoustic scales share five of their seven notes, which is
the most two acoustic scales can share. Second, the two acoustic scales
maximally intersect Whole-Tone Collection I, which appears in mm.
196199 and controls the melodic ascent in the passage. Third, the five
pitch classes common to both acoustic collections are precisely the
pitches that appear in the fanfare-like theme in mm. 208211. These
measures therefore belong to both the E and F acoustic collections, as
indicated by the overlapping brackets on Example 27. Measure 220
brings a return to the second theme and A lydian. The rest of the piece
explores the tonic scalar network depicted in Example 22, with brief
digressions to B mixolydian and F lydian.
The intricate scalar relationships in LIsle Joyeuse pose a critical and
interpretive challenge: how do we reconcile the improvisatory surface of
the piece with what seems to be its sophisticated scalar structure? More
generally, how is it that a joyous and exuberant work celebrating the
flowering of Debussys relationship with Emma Bardac, later to become
his second wife, could be among the most systematic of his musical con-
structions? Even more: how is it that the composer who despised rules
and declared pleasure is the law could have written a piece that so con-
sistently hews to its own compositional laws?
75
There seems to be almost
261
JMT 48.2 8/2/07 2:54 PM Page 261
262
too much structure in LIsle Joyeusetoo much to be consistent with
what we think we know about Debussy and his attitude toward
composition.
One way to approach these important and difficult questions is sug-
gested by what Robert Morris calls a compositional space. Composi-
tional spaces, as Morris defines them, are out-of-time networks of pitch
classes that can underlie compositional or improvisational action.
76
Thinking of scale networks in this way may ameliorate the cognitive dis-
sonance produced by the preceding analysis. For if Debussy had managed
to internalize some of the scalar relationships we have been discussing,
if he could move fluidly and easily between maximally-related scalar col-
lections, then it is clear how his music could be at once highly sponta-
neous and highly structured. A good part of the rigorous structure would
reside in the precompositional network of related scales; the music itself
would remain free to move spontaneously within this network, gambol-
ing lightheartedly in a field whose boundaries are drawn according to
strict principles. I find this image to be appealing, as it helps to reconcile
the results of the current investigation with our senseboth historical and
auralthat Debussys music is locally very free. Debussys music is in-
Whole Tone I
166-85
244-55 196-9
A acoustic
160-3
F acoustic
208-19
Eb acoustic
200-13
A lydian
(E diatonic)
220-35

A major


F

{
E
,

F
#
}
{
B
b
,
C
}

B
{
C
,
D
}

C
#

G

G
#
D#D
164-5 186-95
F lydian
(C diatonic)
236-7, 240-1
Bb mixolydian
(Eb diatonic)
238-9, 242-3
5 common tones
(see 208-13)
START
>
.
Example 26. Scales in LIsle Joyeuse, mm. 160255
JMT 48.2 8/2/07 2:54 PM Page 262
263
E
x
a
m
p
l
e

2
7
.

L

I
s
l
e

J
o
y
e
u
s
e
,
m
m
.

1
9
6

2
2
3
JMT 48.2 8/2/07 2:54 PM Page 263
264
deed spontaneous, but it operates within a compositional space that is
intricate and sophisticated.
(d) Cloches travers les feuilles
The three pieces we have analyzed so far are quite similar: each pres-
ents a central scale surrounded by a number of maximally related collec-
tions. Cloches travers les feuilles, the first in the second series of
piano Images, is a very different sort of piece. Not only are its scales less
fully articulated, but the relations between them are somewhat more
opaque. Cloches therefore provides an opportunity to test the limits of
the theoretical perspective I have presented. Do these concepts help us to
interpret even Debussys more mysterious and elusive compositions?
Example 28 presents a summary of the piece. We can see that maxi-
mally intersecting scales play a role at four places, two of which are
local, and two of which connect distant passages of music. The first
instance is found in the opening measures. The piece begins, as Example
28 shows, with a series of whole-tone fragments moving in different
directions at different speeds. (Arthur Wenk suggests that this music was
inspired by the gamelan.) The first non-whole-tone music, in measure 6,
uses a six-note subset of the E acoustic scale.
77
We then hear a return of
the opening whole-tone material. When the piece next leaves the whole-
tone realm it reaches something much closer to E mixolydian: the first
three beats of mm. 9 and 11 contain six of the seven notes of that mode,
while the next beat adds the missing A, albeit clouded by nonharmonic
passing tones. Thus the shift between whole-tone and diatonic is accom-
plished in stages, with the acoustic scale serving as a midpoint between
the two extremes. This modulatory schema here is, as we have seen,
extremely common throughout Debussys music.
78
Measures 1319 comprise the second thematic region of Cloches
(labeled Theme on the graph); they also contain the second instance
of maximally intersecting scales. Measures 1316 are octatonic, while
measures 1719 feature a series of extended sonorities that imply fifth-
related diatonic collections: B mixolydianE mixolydianE dorian.
By themselves, of course, these maximally-intersecting diatonic scales
are unremarkable. However, their appearance here underscores the simi-
larity between these traditional fifth-based progressions and the more
unusual scale-to-scale transformations we have been investigating.
Measures 2023 present the first return to the whole-tone music of the
opening. As can be seen from Example 28, the music has been trans-
formed so that it now conforms to the D acoustic scale, maximally inter-
secting the original Whole-Tone Collection I. (The reduction omits the
dominant ninth chords, on G and B, which interrupt the music.) Notice,
however, that the D acoustic scale in mm. 2023s is colored by the
presence of a non-scalar D that accompanies the melodic B throughout
JMT 48.2 8/2/07 2:54 PM Page 264
265
the passage. To my ear, these nonscalar tones complicate, but do not dis-
lodge, the acoustic-scale quality of the passage as a whole; this is largely
because the left hand very clearly states an acoustic-scale melody.
79
(I
hear the right hand in mm. 2023 as suggesting flamenco-style parallel
triads on B, C, and D.) Although m. 1 and mm. 2023 are separated by
a significant quantity of music, their thematic similarity makes the close
harmonic connection palpable.
Measures 4446 involve a second long-range relationship. This pas-
sage, a reprise of Theme , suggests the B harmonic minor scale, maxi-
mally intersecting Theme s original Octatonic Collection I. (Again, the
scalar qualities of the reprise are somewhat obscured, this time by a
descending chromatic line in the left hand of mm. 4546.) As before, the
thematic similarities dramatize the harmonic relationship, helping our
ears to associate two passages separated by a considerable quantity of
intervening music.
Unfortunately, the rest of the piece proves rather more difficult to
unravel. Looking again at Example 28, we can see that the opening pres-
ents a series of closely-related, but not always maximally-intersecting
scales. Measures 2439 constitute a contrasting middle section, featuring
distantly-related scales (E lydian, B pentatonic, and C major), and con-
trasting figuration in pentatonic thirds with figuration in diatonic fourths.
The conclusion (mm. 4049) brings back earlier material, but features
dramatic harmonic shifts whose underlying logic is not immediately
clear. Overall, one gets the impression of a meandering tour through a
series of basically unrelated harmonic regions.
Let us see what happens when we plot these scales on the scale lattice
of Example 11. Example 29 graphs almost all of the scales used in the
piece, laid out three-dimensionally as in Example 11. (The three scales
in brackets are placeholders, not appearing in the piece but necessary
to complete the graph.) For reasons that will become clear shortly, I have
omitted the T-symmetric scales, as well as B harmonic minor. What
remains are a series of adjacent faces from the cubic lattice of Example
11. Each contains four scales, three diatonic and one acoustic. Scales on
a single face have a common pentatonic subset. The pentatonic scale
associated with the topmost face is B pentatonic, which appears in
measures 31, 32, and 47 of the piece.
Example 29 is a portion of an interesting structure: the lattice display-
ing maximal intersections among the four locally diatonic scales. Exam-
ple 30 flattens Example 29 into two dimensions. The diatonic circle of
fifths runs in a zigzag fashion through the center, like the stripe on Char-
lie Browns sweater. Above and below the diatonic zigzag of fifths are a
sequence of T
2
-related acoustic scales, each sharing five notes with its
neighbors, and maximally intersecting the same whole-tone collection.
Neighboring acoustic scales maximally intersect distinct octatonic col-
JMT 48.2 8/2/07 2:54 PM Page 265
266
E
x
a
m
p
l
e

2
8
.

T
h
e
m
a
t
i
c

a
n
d

h
a
r
m
o
n
i
c

s
u
m
m
a
r
y

o
f

C
l
o
c
h
e
s

t
r
a
v
e
r
s

l
e
s

f
e
u
i
l
l
e
s

JMT 48.2 8/2/07 2:54 PM Page 266

267
E
x
a
m
p
l
e

2
8

(
c
o
n
t
i
n
u
e
d
)
JMT 48.2 8/2/07 2:54 PM Page 267
268
lections. (These octatonic collections must therefore appear at multiple
points on the graph.) Finally, each square of the graph contains all the
Pressing scales sharing a single pentatonic collection.
80
Example 31 displays all the scales used in Cloches, laid out as in
Example 30. This graph somewhat clarifies the mysterious structure of
the piece. It shows that the opening explores a small collection of rela-
tively closely-related scales. We begin at the top with Whole-Tone Col-
lection I. We then move to the maximally intersecting E acoustic scale,
back to the Whole-Tone I collection, and then on to A diatonic (E
mixolydian).
81
A diatonic jumps briefly to the non-scalar and distantly
related C major triad, but then returns to a point vertically below it, A
acoustic. This acoustic scale again leaps to a distant regionhere Octa-
tonic Collection Ibefore returning to the nearby E diatonic (m. 17).
(Octatonic Collection I maximally intersects D acoustic, as shown by
the dotted line on Example 31. I am unsure how much weight to place on
E lyd
Db ac
[Bb ac]
F dia
Bb mix
Bb pent
[Bb dia]
Eb mix
Eb ac
[Gb dia]
Eb dor. (= C# dia.)
[Gb ac]
Ab ac
Example 29. Scales used in Cloches travers les feuilles
JMT 48.2 8/2/07 2:54 PM Page 268
269
this connection.
82
) We now progress along the circle of fifths through A
diatonic (m. 18) to D diatonic (m. 19), then vertically upward to D
acoustic. Though it is perhaps an overstatement to describe these har-
monic moves as systematic, it would be just as wrong to dismiss them as
entirely capricious. For Cloches, no less than Le vent dans la plaine
or LIsle Joyeuse, begins by exploring a small group of closely-related
scales.
Furthermore, as in those other pieces, Cloches explores more dis-
tant regions as it progresses. The middle section of the piece is chiefly in
B diatonic (E lydian), lying on the far left side of Example 31. B dia-
tonic/E lydian alternates with the very distantly-related B pentatonic
(shared by all the scales on the rightmost square of Example 31) before
moving toward the more centrally located C (D) diatonic, heard previ-
ously at m. 19. The conclusion, after returning briefly to the whole-tone
ideas of the opening, shifts dramatically to F diatonic. (The music
emphasizes the three tones common to both collections, AGF, thereby
smoothing the transition somewhat.) This F diatonic scale might be
understood to balance or compensate for the B diatonic of the central
Bf dia
F dia Ef dia
Ef ac F ac
C ac
G ac
Af dia F pent
C pent B pent
E pent
D pent
G pent
G dia
C dia D dia
A dia
D ac Bf ac
Whole Tone I
Whole Tone II
oct 3 oct 2 oct 1
oct 1 oct 3 oct 2
f E E
f
Example 30. A portion of the locally diatonic scale lattice
JMT 48.2 8/2/07 2:54 PM Page 269
270
section: F diatonic and B diatonic are both a minor third away from the
central A diatonic (E mixolydian ) scale, the first diatonic collection in
the piece.
83
Example 31 thus suggests some new ways of understanding
the pieces distantly-related scalar regions.
Finally, Example 31 suggests a way of thinking about the different
roles of the various scales used in Cloches. The graph represents the
whole-tone, acoustic, and diatonic scales as structurally central: these
collections form a systematic network along which motion tends to occur
in fairly comprehensible ways. By contrast, the octatonic and harmonic
minor scales are represented as outliers, with the shift to octatonic Theme
being somewhat abrupt. Moreover, the octatonic and harmonic minor
collections, although maximally intersecting each other, do not clearly
relate to the other collections used in the piece. (Recall in this regard the
music in mm. 2429 of Collines.) These facts comport with my belief
that Debussy made the most systematic use of the whole-tone, acoustic,
and diatonic scales. Although octatonic and harmonic scales do appear in
his music, they play a fleeting and more adventitious role.
So should we then conclude that Cloches a travers les feuilles is not
improvisatory, free, or loosely-structured? No. For in comparison to the
three pieces studied earlier, Cloches exhibits considerably greater per-
missiveness, a wandering quality that we as analysts should be pre-
B dia
[Fs dia]
Bf pent Df dia
Af dia
Ef dia
[Bf dia]
[Bf ac] Af ac [Gf ac]
Df ac Ef ac
F dia
Whole Tone I
Octatonic
Collection I
B harmonic
minor
?
Example 31. Scales used in Cloches travers les feuilles
JMT 48.2 8/2/07 2:54 PM Page 270
pared to acknowledge. The point I want to insist upon, however, is that
this freedom is not arbitrariness. The procedures exemplified by a more
structured piece like LIsle Joyeuse are also present here, though in more
veiled form. We find striking instances of maximally intersecting scales,
expressed in both local and long-range ways. We see the acoustic scale
mediating between the fifth-based world of the diatonic collection and
the T-symmetrical world of the whole-tone scale. The harmonic minor
scale is clearly associated with the octatonic. And the scales in Cloches
participate in a venerable formal schema: the piece begins by exploring
a series of closely-related scales, broadens in its middle sections to
encompass more distantly-related scales, and concludes by recapitulat-
ing important collections presented earlier in the piece. In short, there is
structure to be found here, structure that has been clarified by the theo-
retical investigation undertaken in Sections I and II of the paper.
IV. Debussy in Context
The analytical challenge in approaching Debussys music has always
been to been to say something about the way it balances between
common-practice tonality and radical atonality. What makes this partic-
ularly difficult is that Debussys music is situated in the no-mans land
between our two most popular analytic methodologies: Schenkerian
analysis, which has its most natural application to music that is more
squarely conventional than Debussys; and set theory, which is most
comfortably at home with music that embraces the total chromatic. I do
not want to suggest that these methods cannot produce interesting
insights into this music. But I do think that one often has to engage in
analytical contortions to apply them, and I think there are cases where
neither technique (nor any combination of the two) tells the full story
about what the composer was up to. I hope to have begun to outline a
more indigenous approach, developing concepts and techniques that
have a natural home in this most alluring, and elusive, of musical styles.
More generally, the scales derived in Sections I and II themselves rep-
resent a natural way-station between the poles of traditional tonality and
complete atonality. Structurally similar to the diatonic scale, they are
capable of participating in the regime of tonally functional harmony. This
is precisely their role in jazz.
84
Yet at the same time, they possess an
extraordinarily rich variety of subsets, and are linked by a complex net-
work of common-tone and voice-leading relationships. One might go so
far as to say that the scales themselves are perched on the boundary
between tonality and atonality. As such, they are potentially of interest to
composers who seek rapprochement with tonal methods of musical
organization. I would hope that the preceding discussion, in addition to
enriching our understanding of existing music, may also interest com-
271
JMT 48.2 8/2/07 2:54 PM Page 271
272
E
x
a
m
p
l
e

3
2
.

M
a
x
i
m
a
l
l
y
-
i
n
t
e
r
s
e
c
t
i
n
g

s
c
a
l
e
s

i
n

C
h
o
p
i
n

s

M
a
z
u
r
k
a

O
p

4
1
,
n
o
.

1
JMT 48.2 8/2/07 2:54 PM Page 272
273
posers looking to explore new musical techniques. Certainly, my own
compositional practice has been enriched by this perspective.
I will conclude with a few final musical examples that indicate our
investigation has identified features of a larger musical practice. The first
two examples come from the nineteenth century. Example 32 presents
two excerpts from Chopins C minor Mazurka, Op. 41 no. 1. The piece
opens in what seems to be C phrygian. When this music returns, in mm.
7380, the C phrygian scale is transformed, by way of maximally-
smooth voice leading EE, into a mode of the F harmonic minor scale.
The resulting combination of modality with maximally related scales
anticipates Debussys more systematic use of the same techniques.
Example 33 presents a strikingly Debussian passage from Liszts third
Mephisto Waltz. The passage implies three scales: D dorian, G acoustic
(D melodic minor ascending), and the Whole-Tone Collection I, with G
acoustic (D melodic minor) maximally intersecting the other two scales.
As occurs frequently in Debussys music, the acoustic scale mediates be-
tween diatonic and whole-tone collections. (This is just one of a number
of remarkable passages in Liszts late works that anticipate Debussys
compositional procedures.
85
) These examples from Chopin and Liszt
suggest a more general question: what was the scope of pre-twentieth-
century exploration of non-diatonic scales, and how does it relate to the
more systematic scalar thinking we find in Debussy? Both questions
deserve further investigation.
86
Example 34 is an excerpt from Stravinskys Petrouchka, in which a B
lydian scale progresses to a maximally intersecting B acoustic scale.
This is one of a large number of similar passages to be found in the works
Example 33. Maximally intersecting scales in Liszts third Mephisto
Waltz, mm. 1926
JMT 48.2 8/2/07 2:54 PM Page 273
274
of disparate early twentieth-century composers, including Scriabin,
Prokofiev, Ravel, and others. (Klokkeklang, one of Griegs more radical
works, could also be mentioned here.) Again, we might ask: How sys-
tematically did these twentieth-century composers explore non-diatonic
scales? Do the similarities between them result from historical influence
or convergent evolution? And what happened to this scalar tradition in
early twentieth-century music? Did it continue, unnoticed by theorists, to
play a significant role in later styles? Or was it displaced by more radical
Example 34. Maximally intersecting scales in in Petrouchka, R65:58
Example 35. Shostakovichs Fs-minor fugue, mm. 2129
JMT 48.2 8/2/07 2:54 PM Page 274
275
methods such as polytonality and atonality? Example 35 hints at an
answer to these questions: it presents an excerpt of the F minor fugue
from Shostakovichs Preludes and Fugues. This passage, which appears
multiple times throughout the fugue, involves a two-semitone shift
between G harmonic major and F phrygian. It gives us reason to suspect
that the scalar tradition plays a role in Shostakovichs music, interacting
in complex ways with post-Wagnerian chromaticism, polytonality, and
atonality.
Finally, it is clear that the late twentieth century witnessed the gradual
reemergence of diatonic thinking, largely under the influence of mini-
malism. But non-diatonic scales also play an important role in this music
as well. Example 36 presents an early minimalist piece, Fredrick Rzew-
skis Les Moutons de Panurge.
87
The entire composition centers on a
maximally-smooth progression from B acoustic to F diatonic. (The
scales involved are the same as those appearing in Example 34.) This
should lead us to ask: how does the scalar music in early twentieth-
century art music relate to the scalar music we find at the end of the cen-
tury? What are the similarities uniting these superficially quite distinct
musical styles?
I would not want, at this point, to make too much of the connections
uniting Examples 3236, nor of their relation to the Debussy pieces stud-
ied in this paper. But I do think it is likely that further investigation will
reveal something like a hidden common practice, a way of thinking
about scales that is the exclusive property of no one composer. This
common practice is, to be sure, just one of many threads that run
through the heterogeneous cloth that is twentieth-century music. But it is
an important and colorful thread, one thatalmost a hundred years after
the factwe are just beginning to understand.
Example 36. Rzewskis Les Moutons de Panurge
JMT 48.2 8/2/07 2:54 PM Page 275
Appendix I: The Genera of Richard Parks
In The Music of Claude Debussy, Richard Parks identifies five gen-
era that he takes to be important in Debussys music: the diatonic,
whole-tone, octatonic, 817/18/19 and chromatic.
88
The dia-
tonic, whole-tone, and octatonic genera consist of all the subsets and
supersets of the diatonic, whole-tone, and octatonic collections. These
genera center on three of my four locally diatonic scales. Though Parks
acknowledges that Debussy makes frequent use of the fourth locally dia-
tonic scale, the acoustic, he does not consider it to engender a genus of
its own. He prefers to interpret it, like the harmonic major and minor
scales, as a distortion of the diatonic. Parks considers the prevalence of
these scales in Debussys music to be a manifestation of Debussys post-
Romantic tendencies.
89
Parkss fourth genus, the 817/18/19 genus, consists of all the sub-
sets and supersets of the octochords 817, 818, and 819. His fifth, the
chromatic, is derived according to substantially different principles,
which bear little resemblance to the theory developed here.
90
I will there-
fore restrict my attention to Parkss treatment of the diatonic, whole-
tone, and octatonic genera, since it is here that his genus theory
comes closest to the concerns of this paper. The similarities between our
two approaches should be clear, even from this brief description: in the
grand scheme of things, Parkss views are quite similar to my own. How-
ever, more detailed consideration shows that there are also interesting
differences between us.
1. As mentioned above, Parks does not consider the acoustic or har-
monic scales to be objects in their own right, treating them as mere mod-
ifications of the diatonic scale. By contrast, I consider them to be inde-
pendent scales that interact with the diatonic and T-symmetrical scales in
complex and interesting ways. I can see no theoretical reason, and cer-
tainly no reason in Debussys music, for assimilating these non-diatonic
scales to the diatonic. Indeed, it is a central point of the present paper that
the octatonic, whole-tone, and hexatonic scales relate to the diatonic
scale in much the same way as the acoustic and harmonic scales do.
2. Unlike Parks, I attempt to provide an explanation of why Debussy
and other composers used the scales they did. Parkss genera are merely
descriptivethey are simply the collections he finds most frequently in
Debussys music. My own approach, though it is also useful for describ-
ing Debussys music, attempts to push deeper, into the realm of underly-
ing principles. Thus I emphasize, not just that Debussy used the scales he
did, but also that these scales are related in interesting and specifiable
ways. This in turn leads me to suggest reasons why Debussyas well as
so many other composersmight have used these scales.
3. Parkss approach is largely set-theoretic, typically treating the dia-
276
JMT 48.2 8/2/07 2:54 PM Page 276
tonic, whole-tone, and octatonic collections as unordered sets. By con-
trast, I emphasize that Debussy typically uses these objects as scales
whose constituents are ordered in pitch-class space (and typically, regis-
ter). This ordering underwrites a notion of distance that permits analysts
to talk about scalar intervals, transposition within a scale, parallel motion
within a scale, and so on. Examples 2 and 6, among many others, demon-
strate that these concepts are useful in analyzing Debussys music.
4. Parkss genera consist in the subsets and supersets of a central
cynosural set class. However, I am not at all convinced that it is neces-
sary to appeal to large scalar supersets in analyzing Debussys music. Not
only is it hard to hear such sets but there is scant precedent for their use
in any of the music that Debussy might have known. Furthermore, the
seeming appearance of these scalar supersets can easily be explained as
the result of modulation (change of scale) or nonharmonic tones.
5. Finally, I disagree with Parks on the importance of complementa-
tion. Unlike Allen Forte, Parks does not consider complementation to be
essential to the process of genus formation. Nevertheless, the comple-
ment relation does play a major role in his analyses. I am suspicious of
this, chiefly because scales satisfying the DS constraint necessarily con-
tain their own complements. (As discussed in Section I, the complements
of these sets are anhemitonic, sharing no common tones with their semi-
tone transpositions.) Thus, any analysis that identifies an abundance of
subsets and supersets of the locally diatonic scales will ipso facto iden-
tify an abundance of complement-related pairs. It is therefore incumbent
on the analyst to show that he or she is revealing something interesting
about the musics structure, rather than describing an inevitable byprod-
uct of the artists choice of materials. Parks, to my knowledge, never
addresses this fundamental issue; instead, he takes the mere presence of
complements as evidence of the analytical and musical relevance of
complementation.
277
JMT 48.2 8/2/07 2:54 PM Page 277
Appendix II: Advanced Scale Geometry
This appendix discusses further features of scale geometry. I begin by
providing a general explanation of the geometrical representations con-
tained in this paper. I then investigate voice leading between harmonic
major and minor scales.
(a) Where Does the Scale Lattice Come From?
The scale lattice of Example 11 belongs to a family of closely-related
structures. Indeed, virtually the same lattice depicts single-step voice
leading relations among trichords belonging to four of the five diatonic
trichord-types, as well single-step voice leading among trichords belong-
ing to four of the eight octatonic trichord types. Examples 37(a) and (b)
illustrate, providing analogues to Example 12s scale cube. These
cubes can be stacked in the manner described in Section II(c). The result-
ing lattices can be decomposed into intertwining strands as described
in Section II(d). The only difference between the respective lattices is the
number of cubes involved: there are twelve distinct transpositions of each
seven-note Pressing scale, producing a lattice of twelve cubes; since
there are only seven transpositions of each diatonic trichord. and only
eight transpositions of octatonic trichord, these lattices involve seven and
eight cubes, respectively.
How can this possibly be? What is the relation between Pressing scales,
diatonic trichords, and octatonic trichords? The answer is that any maxi-
mally even set that is not transpositionally symmetrical can be used to
build a generalized circle of fifthsa cycle of transpositionally related
sets linked by single-step voice leading.
91
It follows that we can always
build a cubic lattice as shown in Example 37(c). Here, our generalized
circle of fifths takes Set 1 to (transpositionally related) Set 2 by way of
the single-step voice leading
1

1
, Set 2 to Set 3 by way of the single-
step voice leading
2

2
, and Set 3 to Set 4 by way of the single-step
voice leading
3

3
. The cube in Example 37(c) associates each voice
leading motion with a spatial dimension: here, the y-axis is associated
with
1

1
, the x-axis is associated with
2

2
, and the z-axis is asso-
ciated with
3

3
. To build the rest of the cube, we simply apply these
single-step voice leadings to the appropriate sets. For example, we get
from Set 1 to W, the set in the lower front-right corner, by applying the
voice leading
2

12
. It follows that this set will be linked to Scale 3 by
the single-step voice leading
1

1
. It is fairly straightforward to
describe the structural properties of the sets that result: for example, set
W must be inversionally symmetrical and transpositionally related to Z,
the set in the rear upper-left corner; meanwhile, sets X and Y, on the lower
rear corners, are related by inversion.
92
The lattices described here are also related to another group of cubic
278
JMT 48.2 8/2/07 2:54 PM Page 278
279
lattices. When an n-note set divides the octave perfectly evenly, then one
can build a circular lattice of n-dimensional cubes, each sharing a single
vertex with its neighbors. For example, single-semitone voice leading
between major, minor, and augmented trichords can be represented using
four cubes, each of which shares a vertex with its neighbors. This struc-
ture was first investigated by Douthett and Steinbach, who call it Cube
Dance.
93
Similarly, single-semitone voice leading between dimin-
ished seventh chords, half-diminished seventh chords, minor seventh
chords, dominant seventh chords, and [0268] French Sixth chords can
CEG
DFG
DEG
DEA
DFA
CEA
CFA
CFG

C

D

E

F

G

A

Set 1
X
Y
Z
Set 4
Set 2
Set 3
W

3



2

1

CEG
CsFsG
CFsG
CFsA
CsFsA
CEA
CsEA
CsEG

E

F
s

C

C
s

G

A

(a) (b)
(c)
Example 37. Other scale lattices
(a) Diatonic trichords 024, 014, 013, and 023
(b) Octatonic trichords 025, 024, 014, 034
(c) A recipe for forming a generalized scale lattice
JMT 48.2 8/2/07 2:54 PM Page 279
be represented using a necklace of three four-dimensional hypercubes,
each of which again shares a vertex with its neighbors.
94
A deep understanding of all of these lattices, as well as the relations
between them, is well beyond the scope of this paper. I believe it can only
be obtained when we adopt a geometrical perspective on set classes. In
Tymoczko 2006 I show that there exist geometrical spaces in which
points represent unordered pitch-class sets and line segments represent
voice leadings between their endpoints, with the length of a line segment
equal to the size of the voice leading it represents. These spaces are com-
plete, representing all pitch class setsin every imaginable tuning sys-
temof a given cardinality. All of the lattices we have discussed appear
naturally in these geometrical spaces. Their structure is largely deter-
mined by the twists of the space containing them. Readers who wish to
explore the matter further will find additional resources on the internet:
on my website, there is a draft of an extensive collaborative project with
Ian Quinn and Clifton Callender, in which we explore the geometrical
perspective in great depth.
95
I have also written a free software program
that produces 3D models of these geometrical spaces.
96
This program can
be used to display the complete trichord lattices generated by Examples
37(a) and (b).
(b) The Circle of Harmonic Scales
Example 10(c) shows that the C harmonic major scale shares six of its
seven notes with both the C and A harmonic minor scales. The voice
leading between C harmonic major and C harmonic minor is maximally
smooth, since it involves the single-semitone displacement EE. But in
moving from C harmonic major to A harmonic minor, the pitch class G
is displaced by two semitones to A, leading to an enharmonic respelling
of A as G. Alternatively, one can represent this voice leading as involv-
ing two single-semitone shifts: G moves to G, producing the multiset
C-D-EFGAB, which combines an augmented triad with a dimin-
ished seventh chord.
97
Then A moves to A. In Douthett and Steinbachs
terminology, the C harmonic major and A harmonic minor scales are P
0,1
-
related. We now turn our attention to the maximal intersectionbut not
maximally-smooth voice leadingbetween them.
To understand these relationships, it is useful to note that every har-
monic major or minor scale can be partitioned into a triad and a nonin-
tersecting diminished seventh chord. The four harmonic scales in Exam-
ple 10(c)C harmonic minor, C harmonic major, A harmonic minor, and
A harmonic majorall share the same diminished seventh chord
(BDFA/G). The triads not intersecting this diminished seventh
chord are related by the neo-Riemannian P and R voice leadings.
98
Example 38 demonstrates, presenting the eight harmonic scales contain-
ing the diminished seventh chord BDFA/G. The scales are identi-
280
JMT 48.2 8/2/07 2:54 PM Page 280
281
fied on the outside of the octagon; inside it, one finds eight triads, linked
to their neighbors by the parallel and relative voice leadings and together
comprising an octatonic scale. Scales whose triads are linked by the P
voice leading are adjacent on Example 11. (For example, AhmAHM is
an edge of the lowest cube.) Scales linked by the R operation are not adja-
cent on Example 11. Instead, they correspond to short cuts that leap ten
places on the non-diatonic circle of fifths, cutting diagonally through
empty space. On Example 11, for instance, the A harmonic major scale
in the first cube is linked, via the R operation, to the F harmonic minor
scale in the fourth cube.
99
The effect of considering these additional voice leadings is to trans-
form the circle of non-diatonic scales, shown in Example 13, into a more
complex structure: a doubly periodic lattice lying on the surface of a 2-
torus. One can think of this structure as a series of three octagons, each
transpositionally equivalent to Example 38. One moves from octagon to
octagon by way of maximally smooth voice leading involving an inter-
vening acoustic scale. For example, one can move from C harmonic
minor, which lies on the octagon shown in Example 38, to G harmonic
major, which lies on a different octagon, by way of the F acoustic scale.
C maj C min
Fs min
Ef maj
Ef min
A min
A maj
Fs maj
common tones:
{B, D, F, Af}
C harmonic
major
C harmonic
minor
A harmonic
minor
A harmonic
major
Fs harmonic
minor
Ef harmonic
major
Ef harmonic
minor
Fs harmonic
major
P

Example 38. The circle of harmonic scales

JMT 48.2 8/2/07 2:54 PM Page 281
282
(Four acoustic scales thus lie between any two octagons, serving to con-
nect a harmonic major scale on one octagon to a harmonic minor scale
on the other.) The 36-scale circle of non-diatonic fifths winds its way
along this structure, alternately moving along each octagon and moving
between octagons by way of acoustic scales.
We are now in a position to appreciate an attractively simple fact: all
of the Pressing scales, with the exception of the octatonic, maximally
intersect precisely six other Pressing scales; the octatonic maximally
intersects twelve other Pressing scales. Example 39 explains. Every
whole-tone scale maximally intersects six acoustic scales. Every hexa-
tonic scale maximally intersects six harmonic scales, three harmonic
major, and three harmonic minor. Every octatonic scale maximally inter-
sects four acoustic, four harmonic major, and four harmonic minor
scales. All of these intersections involve scales with different cardinali-
ties, and hence do not give rise to maximally-smooth voice leading.
Every diatonic scale maximally intersects six seven-note scales (two dia-
tonic, two acoustic, and two harmonic). It can be linked to all six of these
by maximally smooth voice leading. The acoustic scale maximally inter-
sects two T-symmetrical scales (one whole-tone, one octatonic) and four
seven-note scales (two diatonic, and two harmonic). It can be linked by
maximally-smooth voice leading to all four of the seven-note scales.
Finally, every harmonic scale maximally intersects two T-symmetrical
scales (one hexatonic, one octatonic), and four seven note scales (two
harmonic, one diatonic, and one acoustic). It can be linked to three of the
four seven-note scales by maximally-smooth voice-leading, and to the
remaining one by way of the non-maximally-smooth R relation dis-
cussed above. There is therefore an inverse relationship between a scales
transpositional symmetry and the number of different scale-types it
maximally intersects. The highly T-symmetrical whole-tone scale maxi-
mally intersects only one type of scale; while the highly variegated,
T-asymmetrical harmonic scales (which contain three different step sizes
and a very diverse collection of subsets) maximally intersect five of the
six scale types.
Example 39. Maximal intersections by scale type
Scale Type diatonic acoustic harmonic octatonic whole-tone hexatonic total #
whole-tone 6 6
hexatonic 6 6
octatonic 4 8 12
diatonic 2 2 2 6
acoustic 2 2 1 1 6
harmonic 1 1 2 1 1 6
JMT 48.2 8/2/07 2:54 PM Page 282
283
NOTES
1. See, for example, McCreless 1996, especially pp. 110111.
2. Samson 1977 proposes a similar dichotomy.
3. Schoenberg 1911/1978, ch. 20, explicitly asserts the primacy of the chromatic scale.
4. Tymoczko 1997, 2002, 2003.
5. See Antokoletz 1993, Caballero 2004, Callender 1998, Quinn 2002, and Zimmer-
man 2002.
6. Martins 2005.
7. We typically interpret the term one chromatic semitone to mean one scale step
in the standard chromatic scale. However, since the chromatic scale is equal-
tempered, we can also take the term semitone to refer to a scale-independent
unit of distance, namely 1/12 of an octave. Tymoczko 2005 develops this perspec-
tive at length.
8. Clough 1979 explores diatonic set theory in detail.
9. Clough and Douthett 1991 use the terms chromatic length and diatonic length,
abbreviated clen and dlen, to refer to these two ways of measuring distance.
10. For example, corresponding to the pitch-class set [024579E] there is the infinite
ascending ordering of pitches (. . . , 55, 57, 59, 60, 62, 64, 65, 67, . . .). Integers
here represent MIDI note numbers, which provide a spelling-free alternative to the
more familiar, spelling-specific system of designating pitches; see Tymoczko
2005. If one replaces each pitch by the pitch class containing it, the infinite,
ordered series of pitches becomes a unique circular ordering of pitch classes. We
can notate this as (0, 2, 4, 5, 7, 9, 11, [0]), with the bracket indicating a circular
sequence that returns to its starting point.
11. The language of pitch classes provides an efficient way to make general claims
about pitches. For instance, when one says pitch class B is one semitone, and one
scale step below C in the C diatonic scale, one means: for every pitch C, there
is a pitch B exactly one semitone and one scale step below it in the C major scale.
See Tymoczko 2005.
12. Neologisms such as subscale and superscale are in my judgment unnecessar-
ily confusing, as are such convoluted locutions as the set associated with scale S
has subsets that are themselves associated with scales that . . . The mixed termi-
nology is significantly more direct.
13. In general, I will not consider inversionally-related scales to belong to the same
class: thus, the harmonic major and hamonic minor scales are different.
14. The C diatonic scale is the unique octave-repeating scale containing only the pitch
classes 0, 2, 4, 5, 7, 9, and 11. The C acoustic scale is the unique octave-repeating
scale containing only the pitch classes 0, 2, 4, 6, 7, 9, and 10in letters
CDEFGAB. (The scale is familiar from the classical tradition as the
ascending form of G melodic minor.) The C harmonic minor scale contains the
pitch classes 0, 2, 3, 5, 7, 8, and 11, or CDEFGAB. The C harmonic
major scale contains pitch classes 0, 2, 4, 5, 7, 8, 11, or CDEFGAB. The
C pentatonic collection contains 0, 2, 4, 7, 9, or CDEGA. Whole-Tone Col-
lection I is the whole-tone collection containing C (=1, in integer notation).
Whole-Tone Collection II is the collection containing D (=2). Octatonic Col-
lection I is the octatonic collection containing the dyad CD (= 1, 2). Octatonic
JMT 48.2 8/2/07 2:54 PM Page 283
284
Collection II contains DD (= 2, 3), and Octatonic Collection III contains
DE (=3, 4). Hexatonic Collection I contains the dyad CD, Hexatonic Col-
lection II contains DD and so forth.
15. Lerdahl (2001, 268274) likewise proposes a set of desirable scalar properties.
Though these properties differ from those discussed in this paper, there is consid-
erable overlap between his scales and mine.
16. Music theorists sometimes shy away from historical explanations of this sort. At
the end of Section I(b), however, I argue that conventional accounts of the origin
of the melodic minor scale presuppose the three scalar constraints described here.
My account simply generalizes the widely-accepted explanation of the origin of
the melodic minor scale, adapting it to a twentieth-century, post-common-practice
context.
17. Lerdahl (2001, 316) has suggested that something like the DS property reflects a
natural, psychoacoustically-based boundary between the melodic and har-
monic realms. Pitches separated by one or two semitones are subject to forces of
melodic attraction that interfere with their ability to coexist in stable harmonies;
pitches separated by three or more semitones, since they are subject to much
smaller melodic attractions, are more likely to coexist harmonically.
18. Jay Rahn (1991) investigates a related kind of consistency between scalar and
chromatic distance metrics.
19. This is an instance of what Rahn (1991) calls the absence of contradictions in a
scales interval spectrum.
20. Zimmerman 2002 attributes the dissonance of the [012] trichord to the fact that
all three of its intervals would be judged dissonant in any tonal context. Huron
1994 presents a quantitative model of acoustical dissonance in which the 012 tri-
chord is the most dissonant of the trichords. Friedman (1990), though he does not
emphasize the term dissonance, asserts the distinctiveness of the [012] trichord,
advising students to rely on its presence or absence when identifying larger set
classes aurally. Of course, the notion of musical dissonance is both conventional
and contextual, and is notoriously hard to quantify. (Even the more restrictive
notion of acoustic or sensory dissonance is complicated, since it depends on tim-
bre, register, and loudness.) The assertion in the text should therefore be taken as
a reflection of what some composers and theorists have believed, rather than a
simple fact of nature.
21. A number of musical analysts, working on a wide range of music, have arrived at
the NCS constraint independently. Jeff Pressing (1978, 1982) initially identified
the constraint as a key feature of jazz harmony. Philip Russom (1985) applied the
constraint in analyzing Ravel. (It is unclear whether Russom, who has since left
music theory, developed the NCS constraint on his own, or took it from Pressing.)
The constraint also plays a role in Larson 1992. Daniel Zimmerman (2002) inde-
pendently formulated the NCS constraint in considering Prokofiev. My own use
of the constraint in analyzing jazz and impressionism (1997) was likewise carried
out without knowledge of this other work. The simplest explanation for this theo-
retical convergence is that it results from the music itselfthat there indeed are
styles of post-common-practice music that employ a wide range of harmonies, but
not those containing [012] subsets.
22. Explicit articulations of this pandiationic approach to harmony can be found in
JMT 48.2 8/2/07 2:54 PM Page 284
285
the works of jazz theorists who assert that chord voicings can be constructed out
of any of the notes of the appropriate scale. See, for instance Holdsworth 1993 and
Levine 1989.
23. See, for instance, The New Harvard Dictionary of Music.
24. The only way you can remove a pitch class from a DS scale, and still end up with
another DS scale, is if you remove the middle note of an [012] trichord.
25. See Tymoczko 1997.
26. See Messiaen 1956, Martins 2005, and Baker 1985, respectively.
27. Suppose S is a maximal NCS set. Consider any consecutive trichord of the scale
S. Since the trichord cannot be an instance of set class [012], it must span at least
three semitones. But since S is maximal, the trichord can span no more than four
semitones. (If the trichord spanned five semitones, then we could, contrary to
hypothesis, insert another note within it without violating NCS.) Therefore the
scale S possesses the DT property. Conversely, suppose a scale S possesses DT. Its
consecutive trichords span either three or four semitones, which means they
belong to one of three types: [013], [024], and [014]. (The argument that follows
mentions only the prime forms of the [013] and [014] trichords. However, it also
applies to their inversions [023] and [034].) Furthermore, if a DT scale contains a
consecutive [014] trichord, that trichord must be embedded within an [0145] tetra-
chord. (This is because the middle note of the [014] trichord must lie two scale
steps, and four semitones, below a scale tone.) Now let us see whether any note
can be inserted within the span of these trichords. [013] and [024] are internally
NCS maximal, since adding a note within the span of either trichord violates
NCS. [014] is not itself internally NCS maximal, since a note can be inserted
within the trichord to create an [0134] tetrachord. However, we have determined
that the [014] trichord must be embedded within an [0145] tetrachord. [0145] is
internally NCS maximal. We conclude that the set S must be a maximal NCS set.
This shows that every DT scale is a maximal NCS set, and vice versa.
28. See Pressing (1978), Tymoczko (1997). The computer scientist David Rappaport
(2006) has observed that the multiset CCDEFGA also possesses dia-
tonic thirds.
29. The theorists who have independently discovered the NCS constraint have all
gone on identify the Pressing scales: first Jeff Pressing (1978), then Larson (1992),
Tymoczko (1997), and Zimmerman (2002). Allan Holdsworth (1993) explores re-
lated ideas. I recently learned that Sam Davis, who teaches guitar privately in Cam-
bridge, Massachusetts, has developed a substantially similar scale theory. The fact
that so many musicians have independently gravitated toward the same scalar col-
lections helps support the idea that a similar convergence occurred among early
twentieth-century composers.
30. Note that the two harmonic scales minimally violate DS, since they contain only
a single three-semitone scale-step. The hexatonic scale involves a more radical
violation of DS, since it contains three three-semitone steps. For this reason, per-
haps, it is less frequently encountered as a melodic entity.
31. Since a DS+ scale contains no one-semitone steps, its complement contains no
steps larger than two semitones. Since a DS+ scale contains no steps larger than
three semitones, its complement contains no consecutive semitone steps.
32. Note that since an anhemitonic set class shares no notes with its T
1
and T
11
forms,
JMT 48.2 8/2/07 2:54 PM Page 285
286
it must be contained twice within its complement; in the case of the whole-tone
scale, the T
1
and T
11
forms are equivalent. This becomes relevant in Appendix I,
which discusses Richard Parkss work.
33. For an earlier jazz example featuring pentatonic thirds, see Art Tatums 1933
recording of Tea for Two, chorus 3, bars 56. A transcription can be found in
Edstrom and Schiff 1996.
34. This solo is transcribed in Waters 1996.
35. See Cohn 1996. The voice leading I have in mind is, of course, FF. For an ac-
count of scale-to-scale transformations as voice leadings between pitch-class sets,
see Tymoczko 2005.
36. Clough and Douthett 1991.
37. See Clough and Meyerson 1985, Clough and Douthett 1991.
38. This point, which is developed in Tymoczko 2005 and 2006, generalizes an obser-
vation first made by Richard Cohn (1996, 1997). It can, for example, be shown
that the minimal bijective voice leading between two transpositionally-related dia-
tonic sets is at least as small as the minimal bijective voice leading between any
other seven-note sets related by the same transposition operation.
39. For example, the smallest bijective voice leading between any DS set and its T
1
-
or T
2
-form will be at least as small as the smallest bijective voice leading between
any other T
1
- or T
2
-related sets of the same cardinality. Similar statements can be
made for DT scales and their T
3
- or T
4
-forms. These facts follow from the struc-
ture of the relevant scalar path matrices as described in Tymoczko 2005.
40. This sounds more complicated than it is: if the larger set were to share all the notes
of the smaller set, then it would contain that set!
41. This is one reason why the acoustic collection is so often mistaken for other scales
(Tymoczko 2002, 2003). Among theorists who have discussed it, Perle (1984) and
Callender (1998) have both pointed out that it maximally intersects the whole-tone
and octatonic scales. Taruskin (1996) makes a similar observation with regard to
the mystic chord, a six-note subset of the acoustic scale. Howat (1983) notes the
maximal intersection between the whole-tone and acoustic scales. And Antokoletz
(1993) notes the maximal intersection with the diatonic collection.
42. An inversionally symmetrical scale must be an incomplete interval cycle to have
this property, and the diatonic is the only such scale in the group.
43. I will use T-symmetrical and I-symmetrical as shorthand for transposition-
ally symmetrical and inversionally symmetrical.
44. Two distinct T-symmetrical sets can never maximally intersect.
45. If one can transform set A into set B by raising a single pitch class by a single semi-
tone, then one can transform the inversion of set A into some inversion of set B by
lowering a single pitch class by a single semitone. In the special case where A and
B are, like the diatonic and acoustic scales, inversionally symmetrical, inversion
yields a second voice leading between set A and a set of Bs type.
46. Example 11 continues to omit the non-maximally-smooth voice leading between
A harmonic minor and C harmonic major, discussed in Appendix II(b).
47. My graph most closely resembles that version of the triadic Tonnetz in which tri-
ads appear as nodes, rather than faces, of the graph. This is the geometric dual
of the Tonnetz featuring pitch classes as its nodes. See Douthett and Steinbach
1998. Note that not all the connections on the triadic Tonnetz involve maximally-
smooth voice leading whereas all the connections on my scale lattice do.
JMT 48.2 8/2/07 2:54 PM Page 286
287
48. On Example 11, only the A diatonic collection is surrounded by six scales. This is
because the example shows only a portion of the complete lattice. The subgraphs
shown in Example 10 may provide a clearer perspective.
49. I use spelling-specific pitch-class labels for pitch-classes here because they are
somewhat more convenient than numerical labels. However, I do not mean to priv-
ilege any one spelling over any other: C refers to the same pitch class as B,
just as Bob Dylan refers to the same individual as Robert Zimmerman.
50. Since each cube is itself inversionally symmetrical, the second cube is also the
inversion (around E) of the first.
51. If this is difficult to visualize, produce a physical model of Example 12 by draw-
ing the scale names on the corners of a cardboard box. Then rotate the box so that
the line connecting C major and G major occupies the position held by the line
connecting C major to G acoustic. Then transpose each scale by seven semitones.
52. The first cube in Example 11 should share a face with another cube that is in front
of it, while the last cube should share a face with another cube on top of it, and
so on.
53. I describe the relationship that gives rise to these additional harmonic scales in
Appendix II(b).
54. For example, Octatonic Collection I can be transformed into G acoustic by replac-
ing the dyad {G, A} with the pitch class A. G acoustic can be transformed into
Whole-Tone Collection I by replacing the dyad {D, E} with the pitch class E.
55. Octatonic Collection III can be transformed into C harmonic minor by replacing
the trichord {G, A, B} with the dyad {G, A}. It can be transformed into F har-
monic major by replacing the trichord {D, E, F} with the dyad {E, F}. Similarly,
C harmonic minor can be transformed into Hexatonic Collection IV by replacing
the trichord {D, E, F} with the dyad {E, F}. F harmonic major can be trans-
formed into Hexatonic Collection IV by replacing the trichord {G, A, B} with the
dyad {G, A}. Observe that the two harmonic scales define two pathways between
Octatonic Collection III and Hexatonic Collection IV; the two pathways involve
the same trichord-for-dyad substitutions, but in reverse order.
56. The term acoustic scale is often used to refer to this particular mode, though in
this paper I have appropriated it for the scale itself.
57. On the notion of a compositional space see Morris 1995 and 1998. I will return
to this topic at the end of Section III(c).
58. Analysis can also give us useful information about how a composer thought about
his or her music, even if those thoughts do not correspond to any perceptual real-
ity whatsoever. Analysis in this sense is part of the history of ideas, a kind of con-
ceptual archaeology.
59. While I have no qualms about describing Collines as being in B major, it is
worth mentioning that centricity in post-common-practice is a subtle matter, as it
may involve a tonic note, a tonic collection, or both at once. These notions corre-
spond only vaguely to the traditional conception of key. For instance, a piece
may have the white-note scale as its tonic scale, and pitch class C as its tonic note,
and yet it may follow few or none of the harmonic and voice-leading conventions
of traditional tonality. It is slightly misleading to describe such a piece as being
in C major; the more neutral description, in C ionian, seems more appropriate.
60. My analysis of mm. 7880 is somewhat speculative. The total pitch content of the
first halves of mm. 7880 is CDEFGA, common to both B major and
JMT 48.2 8/2/07 2:54 PM Page 287
288
F acoustic. The total pitch content of the second halves of mm. 7880 is
CDEFGA, common to both B harmonic major and Octatonic Collection
III. This music can therefore be analyzed using a variety of underlying scales. I
favor B major for the first halves because Debussy uses diatonic scales more fre-
quently than acoustic scales, and because I find nothing in the music that leads me
to postulate a missing B. My hypothesis of B harmonic major for the second
half may strike readers as more questionable, especially if theylike meare
inclined to analyze this same set class as octatonic when it is presented in mm.
2429. I can give two reasons for my decision: first, the local context of mm.
7880 is quite different from that of 2429, clearly suggesting a series of scales
connected to the tonic B major by maximally-smooth voice leading; second, the
music of 2429, with its distinctive half-step, whole-step melodic pattern, is more
stereotypically octatonic than the music of 7880. Other readers may prefer to
conceive of mm. 7880 in terms of a fleeting chromatic inflection rather than full-
blown change of scale. My response is that Debussy had a distinct preference for
those fleeting inflections that produce scales that obey NCS. Thus, to characterize
Debussys practice of fleeting inflection it helps to think in terms of the Press-
ing scales. Finally, I should emphasize that my analysis of these measures repre-
sents a hypothesis about the techniques embodied in the music, rather than the
way we perceive them, and that this hypothesis is supported not just by the piece
in question, but by a broader investigation of Debussys compositional practices.
61. The prelude Des pas sur la neige features a similar network of scales. Here the
tonic collection is D natural minor (mm. 14, 19), surrounded by the maximally-
intersecting D dorian (mm. 57, 2021), B acoustic (mm. 1618), and D har-
monic minor (mm. 3236). Other scales appear in the piece as well, including A
mixolydian, G major, and Whole-Tone Collection II.
62. Each of the themes is, furthermore, a stack of perfect fifths: Themes and are
diatonic hexachords, or stacks of five perfect fifths, while Theme is a pentatonic
scale, or stack of four fifths. David Kopps 1997 article makes a related point,
emphasizing the pentatonic quality of this pieces themes.
63. Theme could also appear in an A diatonic context. However, the A diatonic scale
does not maximally intersect the tonic B major. So one should really say:
Collines presents each of its themes in the context of every possible Pressing
scale that maximally intersects the tonic collection.
64. The entrance of the D on the fourth beat of the measurewhich can be concep-
tualized as a Csuggests a retrospective reinterpretation of the earlier part of the
measure in terms of the G acoustic scale. Since the notes in the first three beats of
the measure belong to both scales, this reinterpretation complements, rather than
supplanting, the D dorian reinterpretation.
65. Note that the fourth beat of m. 21, taken together with the first beat of m. 22, con-
tains the five pitches common to both the G acoustic and the Whole-Tone I
collection.
66. I
10
maps C diatonic to G diatonic, B acoustic to G acoustic, and Whole-Tone Col-
lection I to itself. As a result of this inversional equivalence, the progression D
dorianG acoustic raises one note by semitone, while the progression B phry-
gianB acoustic lowers one note by semitone. The voice leading between
acoustic and whole-tone scales involves an inversionally symmetrical merge.
JMT 48.2 8/2/07 2:54 PM Page 288
67. I use the term dominant ninth chord to refer to a collection rather than a tonally-
functional object.
68. One can hear hints of this missing A acoustic collection on the downbeats of
mm. 25 and 26.
69. Not appearing on the graph is the G phrygian scale implied by mm. 3537. This
collection does not maximally intersect the other collections in the piece. This
serves as a reminder that, although Debussys music often features networks of
maximally-intersecting scales, he has no qualms about leaping between distantly-
related scalar regions.
70. These are not the divisions proposed in Howat 1983, nor would I insist on their
unqualified correctness. LIsle Joyeuse, like Prelude to the Afternoon of a Faun,
has an ambiguous and somewhat elusive formal structure. It is not surprising that
different analysts parse it differently. Fortunately, my remarks do not depend on
the precise location of section boundaries.
71. Specifically, they account for the music of mm. 198, 160185, and most of 220
255. Note that this count includes the largely B major music of 3651. Although
the B major collection does not reappear in the piece, it is maximally connected
to the tonic network shown in Example 22.
72. Note that a first-inversion diatonic triad (in close position) contains, from bottom
to top, scalar intervals third-fourth-third. A close-position first-inversion penta-
tonic triad contains, from bottom to top, scalar intervals third-second-third.
73. In Tymoczko 1997, I describe how jazz and impressionism both use non-diatonic
scales in tonally functional ways.
74. The C acoustic collection briefly resurfaces in 156157, acting as a kind of
neighbor scale to the Whole-Tone II collection. Note that if one continues to
interpret the music functionally then the acoustic and whole-tone scales in mm.
145157 have tonic function. By contrast, in jazz the acoustic and whole-tone
scales almost invariably have dominant function.
75. Lockspeiser 1962: Volume 1, 206.
76. Morris 1998, 177.
77. The spelling here is CDFGAC, rather than the more usual DE
FGAC. In discussing this passage, Richard Parks (1989) considers the C in
measure 6 to be part of the whole-tone superset 733 [012468A], the complement
of the set heard in m. 1. This strikes me as odd: the C displaces the B of mm.
15; it does not commingle or coexist with it. Our disagreement here testifies to a
more fundamental difference in analytical approach: Parks tends to think in terms
of large scalar supersets, whereas I am more inclined to postulate changes of scale.
See Appendix I for further discussion.
78. We have already encountered this schema in both Le vent dans la plaine and
LIsle Joyeuse. Mouvement, the third in the first series of piano Images, pro-
vides another instance at mm. 97114. The opening of Fetes, shown in Exam-
ple 2, involves an interesting variant of the schema: the dorian music of mm. 18
moves to the D acoustic music of mm. 1114, which moves to the whole-tone
music of mm. 1518. F dorian maximally intersects D acoustic, but the Whole-
Tone Collection II of mm. 1518 does not. Whole-Tone Collection I appears in
m. 19, however.
79. Debussy and Ravel frequently harmonize a melodic scale in ways that imply other
289
JMT 48.2 8/2/07 2:54 PM Page 289
scales. For related passages, see Example 27, above, and Tymoczko 1997, Ex-
ample 26. Satyendra 1992 uses the term stacked space to refer to this situation.
80. It may be helpful to contemplate how the lattice of Example 11 results from fold-
ing up the lattice of Example 29, connecting fifth-related acoustic collections by
way of harmonic scales.
81. From here on, I will revert to speaking of scales rather than modes: E diatonic will
stand equally for B mixolydian, F dorian, and so forth.
82. In my initial thinking about the piece, I was not at all tempted to associate the D
acoustic collection with the octatonic scale in m. 13. However, I now find myself
intrigued by the thought that mm. 2023 represent a hybrid of the whole-tone
Theme with the octatonic Theme . On this interpretation, the D acoustic scale
is a whole-tone collection that has been infected with elementsspecifically an
[0134] tetrachordacquired from the octatonic second theme.
83. Note that the E lydian scale appears on the left of Example 31, suggesting motion
in the flat direction from the central collections of the beginning. In this sense
the sharp-based notation is misleading. Note also that Example 31 is globally
I-symmetrical. Reflecting the graph through a vertical line containing A major is
equivalent to inverting each collection around E/B. This operation preserves the
graphs acoustic and diatonic scales; in particular, it maps the middle sections
(distant) B major onto the closing sections (distant) F major.
84. See Tymoczko 1997.
85. Liszts late works use a variety of Debussian devices: fourth chords, non-
functional harmony, whole-tone scales, acoustic scales, and perhaps even the har-
monic major collection. Semitonal voice leading, as Ramon Satyendra (1992,
1997) emphasizes, is a pervasive feature of the style.
86. See Taruskin 1996, Satyendra 1992, 1997.
87. The piece is meant to be played by multiple players in unison at rapid tempo. Per-
formers begin by playing the first note, then the first and second, then the first, sec-
ond, and third, and so forth, until the whole sequence is played completely. At that
point musicians begin eliminating notes from the beginning of the sequence, play-
ing the second through the last notes, then the third through the last, and so forth.
It is expected that players will make mistakes and lose their place; Rzewski
instructs them not to try to return to the fold. The lost sheep are to proceed at
their own pace, creating a pandiatonic halo that surrounds the main line. (NB:
the existence of this halo means that the transition from B acoustic to F major
will occur by way of a superimposition of notes belonging to both scales.)
88. Allen Forte has also developed a theory of genera that is quite different from
Parkssand much farther from the spirit of the current investigation. Unlike
Parks, whose theory is principally motivated by analytical considerations relating
to Debussys music, Forte aims to provide a fully general taxonomy of pitch-class
sets, not bound to the procedures associated with any one composer. A fuller com-
parison of Parkss and Fortes theories can be found in Parks 1998.
89. Parks 1989, 59.
90. Parks (1998, 207) claims that the chromatic genus is a simple genus, in the sense
that it consists of all the subsets and supersets of a single cynosural set. Else-
where, Parks (1989, 74) explicitly contradicts this statement. I do not see how to
resolve the discrepancy.
91. See Lewin 1996. Moreover, any incomplete interval cycle will give rise to an anal-
290
JMT 48.2 8/2/07 2:54 PM Page 290
291
ogous scale lattice whose elements are connected by single-voice (but not neces-
sarily single-step) voice leading.
92. Any maximally even set that is not transpositionally symmetrical can be written
as an ordered series whose successive elements are separated by the interval c, and
the last element is separated from the first by the interval c + i, with i 1. For
example, the diatonic scale can be represented as the series (F, C, G, D, A, E, B),
with c = 7 and i = 1. The voice leading motion
1

1
displaces the first element
in the pitch class series by i, producing a series whose successive elements are
separated by the intervals (c + i, c, c, ..., c), with the first element now separated
from the last by interval c. Similarly,
2

2
displaces the second element by i,
producing a set with interval series (c, c + i, c, ..., c), and
3

3
displaces the third
element by i, producing the interval series (c, c, c + i, ..., c). It follows that apply-
ing
2

2
to the initial element in the cycle produces the series (ci, c + i, c, ...,
c). Since the interval from the last element to the first is c + i, this element is inver-
sionally symmetrical. It is the analogue of the acoustic scale on Exx. 1112. Sim-
ilar reasoning can be used to derive analogues of the harmonic minor and har-
monic major scales, and to show that they will be inversionally related. One needs
to make slight adjustments to the argument when the underlying chords have only
three notes, but the essential idea is the same.
93. See Douthett and Steinbach 1998. Cube Dance contains within it the familiar
triadic Tonnetz. In some ways, however, Cube Dance is a far better model of
voice-leading relationships than the Tonnetz: on Cube Dance, the F minor triad
is closer to the C major triad than the F major triad, whereas this is not true on the
Tonnetz. Thus Cube Dance can be used to represent the fact that the F minor
triad can appear as a passing chord between F major and C major.
94. This graph subsumes Douthett and Steinbachs Power Towers, but is not equiv-
alent to it. It can be obtained from Power Towers by incorporating the six
French Sixth chords omitted from that graph. Jack Douthett informs me that he
and Peter Steinbach omitted the French Sixth from Power Towers merely for
orthographic reasons.
95. See http://music.princeton.edu/~dmitri.
96. See http://music.princeton.edu/~dmitri/ChordGeometries.html.
97. As David Rappaport points out, this multiset possesses the DT property. If one
includes it on the scale lattice of Example 11, the resulting structure is locally iso-
morphic to the graph of single-step voice-leading relationships between diatonic
trichords.
98. The neo-Riemannian P (parallel) voice leading holds constant the root and fifth of
a (major or minor) triad, while changing the quality of its third. The R (relative)
voice leading holds constant the major third of a triad, while moving the remain-
ing note by two semitones to form another major or minor triad. The L (leading-
tone exchange) voice leading is similar, though it holds fixed the minor third.
99. Note that although these scales are far apart on the circle of nondiatonic scales,
they are not far apart on Example 11 itself. For instance, A harmonic major can
move to F harmonic minor by way of an intervening A diatonic collection. This
two-step path is shown on Example 11.
JMT 48.2 8/2/07 2:54 PM Page 291
292
WORKS CITED
Antokoletz, Elliott. 1993. Transformations of a Special Non-Diatonic Mode in
Twentieth-Century Music: Bartok, Stravinsky, Scriabin, and Albrecht. Music
Analysis 12/1: 2545.
Baker, David. 1985. How to Play Bebop, Vol. 1. Bloomington: Frangipani.
Caballero, Carlo. 2004. Multimodality in Faure. Seattle, WA: Paper Read at the
Joint Meeting of the Society for Music Theory and American Musicological
Society.
Callender, Clifton. 1998. Voice-Leading Parsimony in the Music of Alexander Scri-
abin. Journal of Music Theory 42: 219233.
Callender, Clifton, Ian Quinn, and Dmitri Tymoczko. 2007. Generalized Voice Lead-
ing Spaces. Preprint. http://music.princeton.edu/~dmitri.
Childs, Adrian. 1998. Moving beyond Neo-Riemannian Triads: Exploring a Trans-
formational Model for Seventh Chords. Journal of Music Theory 42: 181193.
Clough, John. 1979. Aspects of Diatonic Sets. Journal of Music Theory 23: 4561.
Clough, John and Jack Douthett. 1991. Maximally Even Sets. Journal of Music The-
ory 35: 93173.
Clough, John, and Gerald Myerson. 1985. Variety and Multiplicity in Diatonic Sys-
tems. Journal of Music Theory 29: 249270.
Cohn, Richard. 1996. Maximally Smooth Cycles, Hexatonic Systems, and the
Analysis of Late-Romantic Triadic Progressions. Music Analysis 15: 940.
. 1997. Neo-Riemannian Operations, Parsimonious Trichords, and their
Tonnetz representations. Journal of Music Theory 41: 166.
Douthett, Jack and Peter Steinbach. 1998. Parsimonious Graphs: a Study in Parsi-
mony, Contextual Transformations, and Modes of Limited Transposition. Journal
of Music Theory 42: 241263.
Edstrom, Brent and Ronny S. Schiff. 1996. The Art Tatum Collection: Artist Tran-
scriptions. Milwaukee: Hal Leonard.
Forte, Allen. 1988. Pitch-Class Set Genera and the Origin of Modern Harmonic
Species. Journal of Music Theory 32: 187270.
Friedmann, Michael. 1990. Ear Training for Twentieth-Century Music. New Haven:
Yale University Press.
Holdsworth, Allan. 1993. Just for the Curious. Miami: Warner.
Howat, Roy. 1983. Debussy in Proportion. New York: Cambridge.
Huron, David. 1994. Interval-class content in equally tempered pitch-class sets:
Common scales exhibit optimum tonal consonance. Music Perception 11: 289
305.
Kopp, David. 1997. Pentatonic organization in two piano pieces of Debussy. Jour-
nal of Music Theory 41: 261287.
Larson, Steve. 1992. Scale-Degree Function: Cognition Research and Its Application
to Aural-Skills Pedagogy. CRCC Technical Report No. 67. Indiana University:
Center for Research on Concepts and Cognition.
Lerdahl, Fred. 2001. Tonal Pitch Space. New York: Oxford University Press.
Levine, Mark. 1989. The Jazz Piano Book. Petaluma: Sher Music.
Lewin, David. 1996. Cohn Functions. Journal of Music Theory 40: 181216.
JMT 48.2 8/2/07 2:54 PM Page 292
Lockspeiser, Edward. 1962. Debussy: His Life and Mind. 2 vols. New York:
Macmillan.
Martins, Jose Antonio. 2005. The Dasian, Guidonian, and Cyclic Spaces in Twentieth-
Century Music. Ph.D. dissertation, The University of Chicago.
McCreless, Patrick. 1996. An Evolutionary Perspective on Nineteenth-Century
Semitonal Relations. In The Second Practice of Nineteenth-Century Tonality, ed.
William Kinderman and Harald Krebs. Lincoln: University of Nebraska.
Messiaen, Oliver. 1956. The Technique of My Musical Language. Trans. John Satter-
field. Paris: A. Leduc.
Morris, Robert. 1995. Compositional Spaces and Other Territories. Perspectives of
New Music 33/1, 2: 328358.
. 1998. Voice-Leading Spaces. Music Theory Spectrum 20: 175208.
Parks, Richard S. 1989. The Music of Claude Debussy. New Haven: Yale University
Press.
. 1998. Pitch-Class Set Genera: My Theory, Fortes Theory. Music Analysis
17: 206226.
Perle, George. 1984. Scriabins Self-Analyses. Music Analysis 3: 10124.
Pressing, Jeff. 1978. Towards an Understanding of Scales in Jazz. Jazz Research 9:
2535.
. 1982. Pitch Class Set Structures in Contemporary Jazz. Jazz Research 14:
13372.
Quinn, Ian. 2002. Foundations of Steve Reichs Later Harmonic Practice: The Vari-
ations and Tehilim. Columbus, OH: Paper Read at the Joint Meeting of the Soci-
ety for Music Theory and American Musicological Society.
Rahn, Jay. 1991. Coordination of Interval Sizes in Seven-Tone Collections. Journal
of Music Theory 35: 3360.
Rappaport, David. 2006. Musical Scales, Integer Partitions, Necklaces, and Poly-
gons. Proceedings of the 2006 Bridges Conference on Mathematical Connections
in Art, Music, and Science.
Russell, George. 1959. The Lydian Chromatic Concept of Tonal Organization for
Improvisation. New York: Concept.
Russom, Philip Wade. 1985. A Theory of Pitch of Pitch Organization for the Early
Music of Maurice Ravel. Ph. D. dissertation, Yale University.
Santa, Matthew. Defining Modular Transformations. Music Theory Spectrum 21:
200229.
Samson, Jim. 1977. Music in Transition : A Study of Tonal Expansion and Atonality,
19001920. New York: Norton.
Satyendra, Ramon. 1992. Chromatic Tonality and Semitonal Relationships in Liszts
Late Style. Ph.D. dissertation, The University of Chicago.
. 1997. Conceptualizing Expressive Chromaticism in Liszts Music. Music
Analysis 16: 219252.
Schoenberg, Arnold. [1911] 1978. Theory of Harmony. Translated by Roy E. Carter.
Berkeley: University of California Press, 1978.
Taruskin, Richard. 1996. Stravinsky and the Russian Traditions. Berkeley: University
of California Press.
Tymoczko, Dmitri. 1997. The Consecutive Semitone Constraint: A Link Between
Impressionism and Jazz. Intgral 11: 13579.
293
JMT 48.2 8/2/07 2:54 PM Page 293
. 2002. Stravinsky and the Octatonic: A Reconsideration. Music Theory
Spectrum 24: 68102.
. 2003. Octatonicism Reconsidered Again. Music Theory Spectrum 25:
185202.
. 2005. Voice Leadings as Generalized Key Signatures: A rational reconstruc-
tion of conventional notation. Music Theory Online 11.4.
. 2006. The Geometry of Musical Chords. Science 313: 7274.
Waters, Keith. 1996. Blurring the Barline: Metric Displacement in the Piano Solos
of Herbie Hancock. Annual Review of Jazz Studies 8: 1937.
Wenk, Arthur. 1983. Claude Debussy and Twentieth-Century Music. Boston: Twayne.
Zimmerman, Daniel. 2002. Families Without Clusters in the Early Works of Sergei
Prokofiev. Ph.D. dissertation, The University of Chicago.
294
JMT 48.2 8/2/07 2:54 PM Page 294