Society for Music Theory

The Perception of Rhythm in Non-Tonal Music: Rhythmic Contours in the Music of Edgard
Varèse
Author(s): Elizabeth West Marvin
Reviewed work(s):
Source: Music Theory Spectrum, Vol. 13, No. 1 (Spring, 1991), pp. 61-78
Published by: University of California Press on behalf of the Society for Music Theory
Stable URL: http://www.jstor.org/stable/745974 .
Accessed: 14/12/2011 14:09

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

University of California Press and Society for Music Theory are collaborating with JSTOR to digitize, preserve
and extend access to Music Theory Spectrum.

http://www.jstor.org
The Perception of Rhythm in Non-Tonal Music:
Rhythmic Contours in the Music of Edgard Varese

ElizabethWest Marvin

To this point in its relatively brief history, the systematic decade have publications by the music-theoretical commu-
study of structure in non-tonal music has undergone tre- nity begun to focus attention consistently upon rhythmic
mendous changes.1 Yet its emphasis throughout has been structure in non-tonal music: for example, in articles by
upon pitch and pitch-class structureover rhythmicstructure, Allen Forte, Christopher Hasty, Martha Hyde, and David
in spite of the striking rhythmic innovations that have oc- Lewin, and in the recent books of Lewin, Jonathan Kramer,
curred in Western music during this century. Early writings and Robert Morris.3
on rhythm in non-tonal music were primarily the work of
composers, such as Olivier Messiaen, Milton Babbitt, and RhythmicStructureand the Electronic Medium," Perspectivesof New Music
KarlheinzStockhausen, whose contributionsdetailed the au- 1/1 (1962), 49-79, reprintedin Perspectiveson ContemporaryMusic Theory,
thors' compositional systems and aesthetics.2 Only in the last ed. Benjamin Boretz and Edward T. Cone (New York: Norton, 1972),
148-179; Karlheinz Stockhausen, ".. . how time passes ....," trans. Cor-
'Publications regarding non-tonal pitch and pitch-class structure now nelius Cardew, Die Reihe 3 (1959), 10-40; Stockhausen, "Structure and
span more than half a century, but few sources survey developments in the Experiential Time," trans. Leo Black, Die Reihe 2 (1959), 64-74. See also
field chronologically. The opening section of Janet Schmalfeldt's Berg's Charles Wuorinen, Simple Composition (New York: Longman, 1979), par-
"Wozzeck": Harmonic Language and Dramatic Design (New Haven and ticularlyChapters 10 and 12, for a discussion of Babbitt's time-point system.
London: Yale University Press, 1983), entitled "Pitch-Class Set Theory: 3Representative articles include Allen Forte, "Aspects of Rhythm in
HistoricalPerspective," gives one account of early developments in the field. Webern's Atonal Music," Music Theory Spectrum2 (1980), 90-109: Forte,
It also provides a succinct overview of Allen Forte's theories, as well as "Foreground Rhythm in Early Twentieth-CenturyMusic," Music Analysis
definitions for technical terms commonly used in the literature. See also 2 (1983), 239-268; ChristopherHasty, "Rhythm in Post-Tonal Music: Pre-
Elizabeth West Marvin, "A Generalized Theory of Musical Contour: Its liminary Questions of Duration and Motion," Journal of Music Theory 25
Application to Melodic and Rhythmic Analysis of Non-Tonal Music and its (1981), 183-216: Martha Hyde, "A Theory of Twelve-Tone Meter." Music
Perceptual and Pedagogical Implications"(Ph.D. dissertation:University of Theory Spectrum6 (1984), 14-51; David Lewin, "Some Investigations into
Rochester, 1988), which also begins with a historical overview (pp. 2-31) Foreground Rhythmic and Metric Patterning," in Music Theory: Special
and which supplements Schmalfeldt's account by including more current Topics, ed. Richmond Browne (New York, Academics Press, 1981), 101-
contributions to the field. 137. Other published investigations into the rhythmicstructureof non-tonal
2OlivierMessiaen. The Techniqueof My Musical Language(1944), trans. music may be found in Kramer'svery thorough "Studies of Time and Music:
John Satterfield (Paris: A. Leduc, 1956): Milton Babbitt, "Twelve-Tone A Bibliography," Music Theory Spectrum7 (1985), 72-106. Recent books
62 Music Theory Spectrum

A fundamental distinction must be made in theories of OVERVIEW OF MUSIC-PSYCHOLOGICAL STUDIES

musical time between those that posit an underlying system
of equally spaced time points and those that do not.4 These Almost without exception, psychologists agree that lis-
time points may take the form of a perceived beat, or may teners familiar with Western tonal music perceive musical
represent a small beat subdivision used as an analytical tool rhythms in relation to equally spaced, internally generated
for measuring durations. Theories that assume equally beats whenever possible.6 Dirk-Jan Povel and Peter Essens,
spaced time points are more common by far, since the for example, describe their beat-based model by contrasting
rhythm of the music that these theories model-virtually all three hypothetical "perceptual clocks": first, an absolute
Western tonal music and a great deal of non-tonal music as clock, pulsing at a single fixed rate; second, a clock that
well-has some basic pulse as its foundation. Two of the pulses at a rate derived from the smallest time unit of a given
publications cited above have dealt in some detail with rhythmic sequence; and third, a hierarchical beat-based
nonbeat-based musical time: Kramer, in his examination of clock.7 The authors reject the first two of these clock hy-
nonlinear time and perception of durational proportions; and potheses on the basis of their experimental results. The ab-
Morris, in his model of sequential time, in which the "'du- solute clock, for example, is "unable to explain why a tem-
rations between s-time points are undefined," and thus are poral pattern presented at a different tempo will be
not assumed to be equally spaced.5 This study proposes a recognized as structurally identical." Further, i'such a model
theory for analysis of nonbeat-based rhythms, one which would imply that all sequences having the same number of
differs from most previous work in that it models relative, [temporal] intervals will be equally well perceived and re-
rather than absolute, measured durations. These patterns of produced regardless of the durations of the intervals.'" Re-
relative durations, here termed rhythmic contours, are based
upon perceptual strategies that listeners use in the absence "Thisis not to say that such a perceptual model is universal. Listeners
of a beat framework. native or acculturatedto Arabic or Indian music hear rhythms as additive
ratherthan divisive. Thus, one would expect that the patterneighth-quarter-
quarterwould be as easily structuredcognitively as the patterneighth-eighth-
dealing with this issue include Lewin, Generalized Musical Intervals and quarterto these listeners, despite the fact that the former does not conform
Transformations(New Haven and London: Yale University Press, 1987); to a beat-based model. The psychological studies cited here are generally
Kramer, The Time of Music, New Meanings, New' Temporalities,New Lis- biased toward Western traditionby virtue of the musical backgroundsof the
tening Strategies(New York: SchirmerBooks, 1988); Robert Morris, Com- listeners who participatein these experiments. However, it is interesting to
position with Pitch Classes:A Theoryof CompositionalDesign (New Haven note that the duration-spaceclasses to be posited below model certainaspects
and London: Yale University Press, 1987). of Indian rhythmic practice, since some rhythmic talas considered to be
4It is assumed here that the presence of equally spaced time points is variationsof each other belong to the same d-space segment class (see note
requiredfor music to be metric. Meter will be understood, following Maury 30 and its accompanyingdiscussion, below).
Yeston's definition (The Stratificationof Musical Rhythm[New Haven: Yale 7Dirk-JanPovel and Peter J. Essens, "The Perception of Temporal Pat-
University Press, 1976], 151-152), as a consonant relationshipbetween two terns," Music Perception 2 (1985), 411-440 (see 413-414).
hierarchicallevels of equally spaced pulses, requiring both a faster- and a 8Ibid,413. For example, "the sequence 200 200 400 [msecs.] and 200 400(
slower-paced pulse. Rhythm will be understood here simply as a succession 400 . . . should both be equally well reproduced. In fact, however, subjects
of durations that may or may not be metrical, and thus may or may not reproduce the first sequence perfectly, but the second poorly." In musical
contain a perceived beat. notation, this would compare the sequence eighth-eighth-quarter with
5Morris,299. eighth-quarter-quarter.
 The Perceptionof Rhythmin Non-TonalMusic 63

garding the second type of clock, Essens and Povel have listeners use a hierarchical clock, in which equally spaced
undertaken further experimentation to determine "whether pulses of medium duration are subdivided or concatenated
subjects can use the smallest interval in a temporal pattern by the listener to structure a duration succession as it is
as a basic unit in representing other (longer) intervals in the heard.
same pattern."9 According to this hypothesis, duration suc- In a beat-based hierarchicalsystem, each of the durational
cessions in which intervalsrelate as 3:1 or 4:1, exact multiples units may also be subdivided. It is unclear, however, exactly
of the basic duration unit, should be reproduced more ac- how the listener perceives this subdivision. Experimentation
curately than ratios of 2.5:1 or 3.5:1. Their results do not has shown that the cognitive structure for perception of
support a distinction between such patterns, however, and rhythmic subdivisions is far from precise. Eric F. Clarke
they conclude that the smallest interval is not used in spec- proposes a perceptual model with two components: equally
ifying the time structureof such patterns. The implication of spaced metrical markers on one level, and a system of un-
this work is that rhythmic theories based upon tallied mul- timed procedures organized around these markers at an-
tiples of a composition's smallest durational value do not other, "specifyingsubdivisionsin terms of equal and unequal
model aural perception. That is not to say that such theories time spans, the unequal subdivisions using a simple distinc-
cannot reveal important aspects of a work's rhythmic struc- tion between long and short."'2 This categorization of un-
ture,10particularlyin compositions where serialized rhythm equal subdivisions simply into longs and shorts is an impre-
is directly linked to pitch structure and in certain non- cise measurement of relative duration, and explains the
Western musics.11Povel and Essens conclude, however, that common misperception of the dotted eighth-sixteenth beat
subdivision for a triplet's quarter-eighth subdivision, since

9Peter J. Essens and Dirk-Jan Povel, "Metrical and Nonmetrical Rep-

resentations of Temporal Patterns," Perception and Psychophysics 37
(1985), 3. erence to more subjective concepts of beat. For example, a beat of MM60
1?See, for example, Forte, "Aspects of Rhythm in Webern's Atonal can also be perceived as two beats at 120. Density referent, being faster than
Music." His proportional graph is designed in precisely this way: "The in- beat, is not subject to such ambiguities" (p. 396). While density referent is
teger value 1 is assigned to the smallest durationalvalue in the work (move- a useful analytical tool, Kauffmansuspects that it is not commonly used by
ment). The largest value is the least common multiple of all the other African performersor listeners to structurerhythmicpatterns perceptually.
values. . . . The result is a depiction of a precise calibration of component After observing a teacher instructingAfrican drummingstudents, Kauffman
durations, so that any temporal span or pattern can be compared with any notes that "Ayitee did not ask his students to count out eight fast pulses.
other" (p. 91). Instead he wanted them to respond to the gestaltof the two drum parts. This
llIn "African Rhythm: A Reassessment," (Ethnomusicology 24 [1980], would seem to suggest that density referent is only one level of a larger
393-415), Robert Kauffmanexpands upon variouscurrenttheories of rhyth- metrical organization" (p. 396).
mic structurein African music. Among these is analyticaluse of the "density t2EricF. Clarke, "Structureand Expression in RhythmicPerformance,"
referent," a concept defined by Mantle Hood in The Ethnomusicologist(New in MusicalStructureand Cognition, ed. Peter Howell, Ian Cross, and Robert
York: McGraw-Hill, 1971), that refers to the fastest regularly recurring West (London and New York: Academic Press, 1984), 225-226. Clarkecites
event. Kauffman notes that the density referent "can be used to study and W. Jay Dowling's "Scale and Contour: Two Components of a Theory of
understand temporal elements that would be rendered ambiguous by ref- Memory for Melodies," Psychological Review 85 (1978), 342-354.
64 MusicTheorySpectrum

both are heard as long-short.'3 In Clarke's "two-component much the same way that melodic contours represent relative
system," the perceived pattern of relative longs and shorts pitch height, without a precise calibration of the intervals
is overlaid upon a metrical matrix that represents beat struc- spanned.
ture and subdivision. Clarkenotes that this theory has certain Povel hypothesizes that listeners have at least two possible
similarities to W. Jay Dowling's two-component model of ways of understanding temporal sequences, and that it is the
melodic contour overlaid upon a diatonic scale framework.14 natureof the rhythmitself that determines which method will
Dowling and Dane L. Harwood note the similaritybetween be used: those that do not fit a beat-based coding are in-
theories as well, citing Monahan's suggestion that ternally represented as rather unstructured groups of tones. '"
He notes that for nonbeat-based rhythms, an "alternative
rhythmicsubdivisionpatternsare laid on the beat frameworkin a
coding, called 'figural coding' by Bamberger (1978), capi-
way analogous to the way melodic pitch contours are laid on the
talizes on the perceptual grouping of events . . . [and] de-
scale framework .... Rhythmic subdivisionscan thus be said to be
encoded in rhythmic contours of relative, not absolute, temporal tailed information about the relative durations of intervals
would seem to be left uncoded. "7 This hypothesis has been
relationships.Rhythmiccontours are like melodic contours in being
able to stretch to fit different frameworks(as with change of tempo) substantiated more recently by Jeffrey Summers, Simon
and in being able to slide along a given framework (as in displace- Hawkins, and Helen Mayers, who also describe two per-
ment of rhythmic accent).15 ceptual models of temporal organization: (1) Gestalt-like
Thus rhythmic contours may be understood as analogous groupings; (2) beat-based hierarchies.'s They note that the
first is used to interpret non-metrical rhythms, while the sec-
to melodic contours: they represent relative durations in
ond is used for metrical ones. It is this type of non-metrical
rhythmic processing that the theory to be discussed here
13Thistype of error has also been discussed with reference to African
attempts to model-a musical conception without the "sign-
drummingperformance.In discussingapplicationsof the "density referent"
posts" that beats provide, a conception marked by percep-
concept, Kauffman notes that: "A performer of the pattern o. must be
awareof the density referent , in order to avoid the errors , 2 or tual grouping according to temporal proximity, and retention
but he will ultimately respond to the larger gestalt of each beat or of the of relative rather than absolute measured durations.
entire measure. Thus it would also seem that African musicians respond to Nonbeat-based rhythms such as these abound in non-
some type of metricalorganization,which may include variouscombinations Western musics and in Western music of this century-in
of the density referent" ("African Rhythm: A Reassessment," 396-397).
some electronic compositions, in serialized rhythmic designs,
The rhythmsthat Kauffman cites are all unequal beat divisions which may
be categorized as long-short, and which are therefore easily confused. '6Povel, "Internal Representation," 16.
'4Clarkementions this analogy in "Some Aspects of Rhythm and Ex- '7Povel and Essens, "The Perception of Temporal Patterns," 437. The
pression," 324-325, and in "Structureand Expression in Rhythmic Perfor- article the authors cite is Jeanne Bamberger, "Intuitive and Formal Musical
mance," 226. Knowing: Parables of Cognitive Dissonance," in The Arts, Cognition, and
15W.Jay Dowling and Dane L. Harwood, Music Cognition (New York: Basic Skills, ed. S. S. Madeja (New Brunswick, N.J.: Transaction Books,
Academic Press, 1986), 187-188. The authors cite, in particular,Chapter 5 1978).
of Monahan's dissertation, 'Parallels between Pitch and Time: The Deter- '8JeffreyJ. Summers, Simon R. Hawkins, and Helen Mayers. "Imitation
minantsof MusicalSpace" (Ph.D. dissertation,Universityof California,Los and Production of Interval Ratios," Perception & Psychophysics39 (1986),
Angeles, 1984). 437.
 The Perception of Rhythm in Non-Tonal Music 65

and in rhythmically dissonant passages of works by Carter in the way intervallic spans are measured: in the first case
and Stockhausen, for example. as quotients and in the second as differences. The remaining
two temporal spaces are modular reductions of these two
systems.19
RHYTHMIC CONTOURS IN DURATION SPACE An additional type of temporal space is proposed here:
a duration space, analogous to contour space, that models
Two recent music-theoretical publications have discussed relative duration in the same way that contour space models
structural parallels between pitch spaces and temporal relative pitch height. However, unlike melodic contours in
spaces; they form an important point of departure for the contour space, which can be recognized aurally with some
theory of rhythmic contours that follows. Morris's Compo- accuracy regardless of context (tonal or non-tonal), rhythmic
sition with Pitch Classes defines three types of temporal contours in duration space are very much altered in listeners'
spaces, the structures of which are isomorphic with three perception by their metric contexts. Thus the works to be
pitch spaces: "sequential time" with contour space, "mea- studied here are non-metrical; they are, in fact, works where
sured time" with pitch space, and "modular time" with pitch- a consistent and uniform beat is hard to discern. It is in this
class space. Segments in Morris's sequential time and in con- type of context, where no consistent beat unit can be per-
tour space are represented numerically by integers from 0 to ceived, that rhythmic contours of relative shorts and longs
n-1, where n equals the cardinality of the segment. In both best model the listener's perception.20 At least one composer
sequential time and contour space, the precise interval be- concurs with the premise that listeners' temporal understand-
tween the successive elements of a segment is not calibrated; ing of nonbeat-based music is based upon perception of rel-
the integers simply model the concepts of "earlier/later" or ative durations; Gerard Grisey states that
"lower/higher" without a precise measurement of how much
without a reference pulse we are no longer talking of rhythm but
earlier or how much higher. Morris's measured time is a
of durations. Each duration is perceived quantitatively by its re-
temporal pitch-space analogy: both are spaces divided into lationship to preceding and successive durations. This is the case in
equal measured units numbered with positive and negative the rhythmicwriting of Messiaen and of the serialist school. In fact,
integers on either side of some midpoint. Lower pitches or
earlier time points are modeled by increasingly positive in- t9Morris'stemporal spaces are discussed in Composition with Pitch
Classes, 299-301; Lewin's are defined in GeneralizedMusical Intervals,22-
tegers. Finally, modular time is the temporal analogue to 25.
pitch-class space, derived from measured time and from 2'As Povel noted in the experimentsdiscussedpreviously, a beat emerges
pitch space by reduction mod n; its elements are numbered in duration successions where successive equally-spaced shorter durations
from 0 to cardinality n-1. Lewin's formulation, in Gener- may be heard as subdivisions of a longer duration. By extension, in beat-
alized Musical Intervals and Transformations, contains six based rhythms, stringsof equally spaced durationsare common (two eighths
or four sixteenths, for example). However, in durationsuccessions where no
temporal spaces, two of which correspond to Morris's mea- two durations are of equal length, a beat is difficult to hear unless supplied
sured time and modular time. In addition, he defines four in an accompanyingline. Musically-trainedlisteners may try to "impose" a
types of temporal spaces based not on sequential time points, beat to structure their listening, but if this strategy fails they too rely upon
but upon durations. Two of these duration spaces differ only a perceived rhythmic contour of relative shorts and longs.
66 Music Theory Spectrum

a micro-pulse allows the performer or conductor to count and ex- Example 1. Multiple realizations of <0 1 2 3> in contour space
ecute these durations, but it only exists as a way of working and J. S. Bach, The Musical Offering, BWV 1079, X. Canon a 2. Quaerendo invenictis,
has no perceptual reality. The more complex the durations . . . the mm. 1-2
more our appreciation of them is only relative (longer or shorter
than . . ).21 (a) ( IJ U I IJ J I
I I I I
Duration space (or d-space) is defined here as a type of
temporal space consisting of elements arranged from short Beethoven, Piano Sonata in F Minor, Op. 2 No. 1, I, mm. 1-2
to long. Elements in d-space are termed durations (durs) t I , , / _ \
(b) -bV - ,,L I I I
and, as mentioned previously, are numbered in order from
short to long, beginning with 0 up to (n-1), where n equals
Mikrokosmos
Bart6k, No. 144, "Minor Seconds, Major Sevenths," m. 63
the number of elements in the segment and where the pre-
Bart6k, Mik-okosmos No. 144, "Minor Seconds, Major Sevenths." m. 63
cise, calibrated duration of each dur is ignored and left un-
defined.22 A d-segment (dseg) is defined as an ordered set
of durations in d-space. Just as a contour-space cseg can be (c)
realized in pitch space in an infinite number of ways, so can *1~

a dseg be realized in measured time by an infinite number

of rhythms. Example 1 shows several realizations of the Stravinsky, L'Histoire du soldat, "Marche du soldat," mm. 57-58, trumpet in A

contour-space segment <0 1 2 3> realized in pitch space. A *. ; t -,

(d)
Segments in contour space are comprised of c-pitches num- 'VY
C)
1 _
I II ' C i
bered in order from low to high, thus <0 1 2 3> represents I L ___J
<3210>
a continuously rising melody, <3 2 1 0 > a continuously de-
scending line, and <2 3 0 1> a more angular melody with
two changes of direction. Note that the theory is general
ment realized in measured time.23 Examples (a) through (c)
enough to be applicable to Bach as well as Bart6k and that show different possible metrical interpretations, while ex-
the intervals spanned between contour pitches may vary;
thus the stepwise motion of the Bach excerpt is equivalent amples (d) through (f) show some non-metrical realizations
drawn from Varese's Octandre.
in contour space to the arpeggiation of the Beethoven ex-
In numbering durs from short to long, the determination
ample. Example 2 shows < 0 1 2 3> as a duration-space seg- is made from the onset of one dur to the onset of the next,
regardless of whether the pitch in question extends through
the entire temporal interval spanned or is interrupted by a
21GerardGrisey, "Tempus ex Machina. A Composer's Reflections on rest. Thus, dseg (b) of Example 2 still represents < 0 1 2 3 >
Musical Time," Contemporary Music Review 2 (1987), 240.
22Formulationof these definitions and those that follow are indebted to 23Note in Example 2 that the fourth dur of dseg (d) is ornamented by
those for contour space in Morris, cited above. Note that this application Dp-E grace notes. The grace notes will not be consideredtwo separate (very
of the COM-matrixdiffers somewhat from Morris's temporal applications short) durs, but rather ornaments (like a trill), belonging to the DO and
of contour theory. lengthening its duration slightly.
 The Perceptionof Rhythmin Non-TonalMusic 67

Example 2. Multiplerealizationsof dseg <0 1 2 3> in measuredtime:

overlaid upon three possible "temporal grids" from Varese, Octandre, I

(a) (d) m. 2, oboe

6~~P

(b)
(e) mm. 13-15, bassoon

)
L I4I i IJ. I
s i
4t7f.- r, F
(c)
(f)
mm.17-18, oboe
it J I. (f) mm. 17-18, oboe

6~~l~ l
n >6 ~+~~~~~~/Z
1- -\~
>4 5 Vt i If 7 r Ib
I

L,
I F.
I

M-mw
I .-Jff
<0123>

even though it contains a rest. Music psychologists call this sustaining instruments, the theory might be formulated to
temporal span the inter-onsetinterval. As Eric Clarke notes, exclude the duration of the final note, restricting its role so
"This is the most significant measure as far as the rhythmic that it serves only to define the length of the penultimate
function of the note is concerned since the other possible duration. However, in musical contexts-particularly those
measures (onset to offset or offset to onset) refer mainly to involving performance by sustaining instruments, such as
the articulation properties of the note."24 This requirement voice or wind instruments-the cut-off of the final note has
is one of the features that distinguishes contour space from some perceptual validity. It is for this reason that the ex-
duration space, since c-space compares points in space while amples following include the final note's duration. Finally,
d-space compares pairs of points (onset to onset). This pre- while a rest that is internal to a duration succession generally
sents certain analytical difficulties in determining the dura- adds to the duration of the note preceding, it is also an
tion of the final note of a succession, because there is no important criterion for dividing the succession into
following onset by which to measure the length of that final d-subsegments.
duration. In the abstract and in the case of certain non- A d-subsegment (dsubseg) is defined as any ordered sub-
grouping of a given dseg. Example 3 illustrates dsubsegs
drawn from a prominent oboe melody in the first movement
24Eric F. Clarke, "Structure and Expression in Rhythmic Performance," of Octandre (mm. 8-12). Dsubsegs may be compared more
212. easily by renumbering the segments through "translation,"
68 MusicTheorySpectrum

Example 3. Dsegs and dsubsegs in duration space:

Varese, Octandre, I, oboe, mm. 8-12
'
-'""D-.~ '
"r ~ longue
-l
tzi.'
LjrT
...L
T-&
_~i.g ~f, '-
i^ ^-
r 7m
v_ _ , _
f -mp zff
<210354>

Dsubsegs:

(a) : b, I
V^ r r
<210>

^bA^A A ? EE

(b)

<035>= <012>*

(c)
J r L - 1 1r Irf
<0354> = <0132>*

XY I 1_ 'I I1 I I
(d) . -
T Lj I

<1035>= <1023>*
* =by translation

as Example 3 shows. Translation is an operation through contiguous high cps aurally by their close proximity in pitch
which a dsubseg of n distinct durations, not numbered from height, for example, whereas the temporal nature of d-space
0 to (n-1), is renumbered from 0 for the shortest dur to (n-1) prevents the listener from grouping all long durs together
for the highest. Dsubsegs in duration space are assumed to simply by virtue of their length. Only in the case where
be contiguous subgroupings, unlike csubsegs in contour melodic contour interacts with perception of rhythmic con-
space. In contour space, the listener may group non- tour might a case for noncontiguous dsubsegs be made. In
 The Perceptionof Rhythmin Non-TonalMusic 69

compound melody, factors such as pitch proximity might Example4. Non-contiguoussubsegs:

cause the listener to perceive the higher or lower voice (or Varese, Octandre,I, oboe, mm. 2-3
both) as an independent duration stream. Example 4 shows
such an instance. Here, the reiterated low F might be heard (J J ? J J J J )
as a separate stream, resulting in a noncontiguous dsubseg L boJ ,- '

heard in the upper voice (indicated by stems up). Non-

contiguous dsubsegs are clearly a special case; thus the term 3-
3
dsubseg will generally refer only to contiguous dsubsegs un-
stems down: <0123> <11002>
less otherwise specified.
stems up:* <012> <102>
A precise profile of the structureof a duration succession * noncontiguous
dsubsegs
in d-space is provided by Morris's comparison-matrix (or
COM-matrix), shown in Example 5. This matrix is a two-
dimensional arraythat displays the results of the comparison resentatives of < 1 0 2 >. Dseg equivalence may also explain
function, COM(a,b). In this case, a and b represent any two why rhythmssuch as those of Example 5b are often confused
durs in d-space. If b is longer than a, the function returns by students in early stages of aural skills training, since the
' + "; if b is the same length as a, it returns "0"; if b is shorter two are equivalent in duration space. Finally, Example 5c
than a, COM(a,b) returns "-." Each of these matrices has illustrates an additional instance of dseg equivalence; here
symmetrical properties in which the diagonal of zeros from the durations of the two successions in measured time are
the upper left-hand to lower right-handcorner forms an axis not related by any precise mathematical relationship. Yet
of symmetry. Each value in the upper right-hand triangle is succession (4) is a free augmentation of (3); in d-space the
mirroredon the other side of this diagonal by its inverse. This two are equivalent representations of <0 1 3 2> and pro-
symmetrical structure is a natural consequence of the fact duce identical matrices as shown. Succession (4) is one of the
that the COM-matrix, as used here, compares a dseg with dsubsegs from Octandrecited previously in Example 3, and
itself. is numbered as in that example to show the translation op-
Two types of equivalence relations are posited for dsegs eration; succession (3) is comprised of the first four notes of
in duration space, based in part upon the COM-matrix.First, the movement. The two successions have clear aural
equivalent dsegs are those that generate identical matrices. associations-both are prominent solo oboe lines, and the
This definition asserts equivalence for any two duration suc- melody of succession (4) represents a rhythmic expansion,
cessions related as those in Example 5a. Measured in terms or development, of the melody with which the movement
of the smallest durational unit (the sixteenth note), succes- opened.
sion (1) may be represented as < 3 1 4 > and (2) as < 6 2 8 >. The second equivalence relation, the duration-space seg-
Succession (2) is an augmentation of (1) in measured time, ment class (dsegclass), is defined as an equivalence class
a relationship that may be shown numericallyby multiplying made up of all dsegs related by identity, translation, retro-
the durational values of (1) by 2. The two successions gen- grade, inversion, and retrograde inversion. The inversion of
erate identical matrices, and in d-space are equivalent rep- a dseg S of n distinct durs is written IS, and may be found
70 Music Theory Spectrum

Example 5. Dseg equivalence

(a) (c)
Varese, Octandre, I, oboe, mm. 9-12
(1) j =J 1

<3 14> in sixteenth-note durations = < 102 > in d-space

(3) ) r -
(2) j.I )I l mp -
<0132>
<628> in sixteenth-note durations = < 102 > in d-space

Varese, Octandre, I. oboe, m. 1

matrices:
L . ongue
(1) 3 1 4 (2) 6 2 8 -- -.-- r1 l r f -
3
1
0- +
+ 0 +
6
2
0- +
+ 0 + (4) Al ff L_
^ i^'i^ rTI7
4 - - 0 - - 0
8
-cff .fr
<0354>=<0132>

(b) matrices:

(3) 0 1 3 2 (4) 0 3 5 4
<102> in d-space 0 0+++ 0 0+++
1 - 0 + + 3 - 0 + +
3 -- 0- 5 -- 0-
2 - - +0 4 -- +0
<102> in d-space
'i, J., I 11

by subtracting each dur from (n-1), where n represents the Milton Babbitt in his Three Compositions for Piano.25 The
cardinality of the segment. In effect, this results in durations retrograde (RS) or retrograde inversion (RIS) may be found
"swapping" positions within the segment. Given an odd by listing the elements of S or IS in reverse order. Thus, in
value of (n-1), the longest and shortest durations swap po- Example 3, dsegs (a) and (b) are R-related, while (c) and
sitions, the next-to-longest and next-to-shortest swap posi- (d) are RI-related. Example 6 summarizes these relation-
tions, and so on. If (n-1) is even, the same holds true except ships for one representative of dsegclass <0 1 3 2>, rep-
that _5retains its position (see Fig. 1.). A comparison of the resenting S and its R, I, and RI transformations and the
P and I forms in Example 6, for example, shows that 0 and
3 (or the sixteenth and half) swap positions, as do 1 and 2 2^CharlesBurkhart describes this algorithm in some detail in his intro-
(or dotted eighth and quarter). This algorithm for finding the duction to Babbitt's composition in Anthologyfor MusicalAnalysis, 3rd ed.
inversion of a rhythmic segment is precisely the one used by (New York: Holt, Rinehart, and Winston, 1979), 578-584.
 The Perceptionof Rhythmin Non-TonalMusic 71

Figure 1. Example6. Dsegclass equivalence

n--r-
P:
n
<0213> P:
m n
<04132> P:
Im
<3124650> J fli
I: <3120> 1: <40312> I:
I <X
<3542016>
P: <0132>

0132
I: <3201> RI: <1023> R: <2310>

LJ LJ Li Li 3201 1023 2310

L1
0 0+++ 3 0 - - - 1 0 - ++ 2 0+-
1 - 0 + + 2 + 0- - 0 +0++ 3 0 - -
3 - - 0 - 0 ++ 0+ 2 - -0 + 1 + + 0 -
correspondingCOM-matrixfor each. Because d-space is iso- 2 - - +0 1 + + - 0 3 -0 0 + + + 0

morphic with c-space, the prime form algorithm and cseg-

class labels developed for contours in c-space, included as an rhythmic contours has not, up to this point, accounted for
appendix to a previously published article, may be adopted the instance of repeated equal durations. For the most part,
for use in analyzing rhythmic contours.26These labels are however, the theory can be extended to include such
hyphenated numbers, with the first number representing the rhythms. As shown in Example 7, a duration succession such
cardinality of the segclass, and the second representing its as eighth-quarter-half-quarter would be modeled as a
ordinal position on the list. As an alternative to reprinting "repeated-note contour," <0 1 2 1> in duration space.
the segclass table, duration segments are represented here Such a contour would generate a matrix containing zeros in
by their prime forms instead of their dsegclass labels.27 positions other than along the main diagonal, as shown. Its
To return briefly to the data on temporal perception dis- dsegclass label is a composite, combining the labels of the
cussed at the outset, listeners discern a beat most easily in two dsegs that are most similarto the repeated-note contour.
rhythmic patterns where successive equally spaced shorter These dsegs for <0 1 2 1 > are illustrated in Example 7; the
durations may be heard as subdivisions of a longer duration. first is generated by replacing all zeros that appear in the
It is only in nonbeat-based rhythms-by extension, rhythms upper right-handtriangle of the repeated-note contour with
with fewer instances of repeated equal durations-that lis- pluses, and the second with minuses.
teners perceive a rhythmic contour of relative shorts and In order to generalize contour theory from c-space to
longs without a precise notion of their proportional rela- d-space, a number of important differences between the
tionships. For this reason, the theory that models these spaces must be acknowledged, some of which have been
touched upon above. While these differences do not affect
26ElizabethW. Marvin and Paul A. Laprade, "Relating Musical Con- the general applicabilityof the theory to durationspace, they
tours: Extensions of a Theory for Contour." Journal of Music Theory 31 do have important implications for the perception of various
(1987), 225-267 (see 257-262). transformationsthat occur in the music to be studied below.
27A dseg's prime form is a representative form derived by the following
First, a fundamental perceptual difference between contour
algorithm: (1) if necessary, translate the segment so its content consists of
integers from 0 to (n-l): (2) if (n-1) minus the last element is less than the space and duration space is that while melodic contours are
first, invert the segment: (3) if the last element is less than the first, retrograde easily remembered, and same/different comparisons accu-
the segment. rately made by most listeners regardless of context (tonal or
72 MusicTheorySpectrum

Example7. Repeated-notedseg <0 1 2 1> andrelated psychologists have not yet explored the question of whether
matrices listeners can perceive R, I, and RI transformations upon
duration successions. Retrograde rhythms have been used
for centuries in conjunction with pitch retrogradesin musical
<0121> composition, but it is unclearhow well listeners recognize the
0121 0132 0231 pitch transformation,much less the rhythmicretrograde. In
0 o ++ o o\+++ 0 0. + the case of metric music, rhythmic retrogrades violate such
1 -0+0 1 - 2 - \+ E
2 -- O- 3 - - 3 - - expectations as long notes coinciding with 'strong" beats or
1 - 0 + 2 - - + 1 - + + O initiating measures; thus the new metric context of the ret-
rogradesuccession makes this transformationdifficultto rec-
ognize aurally. Yet musical experience and intuition suggest
non-tonal), rhythmiccontours may not be recognized by lis- that rhythmicretrogradesmay be perceivable if their length
teners as identical if their underlyingmetrical structuresdif- is not excessive. Certainly the palindromic rhythms of We-
fer. It is for this reason that compositions chosen for analysis bern's Variations, Op. 27 and Symphony, Op. 21 can be
and discussion here are works that do not strongly invoke heard for a short while, if not for their full length.28The issue
a perceived beat or meter. A second consideration is that of rhythmicinversion is a more complex one. Few composers
since the elements of d-space are durationsand are therefore before this century attempted to "invert" duration succes-
measured from the onset of one event to the onset of the sions, since there was no established procedure as to how
next, each duration is dependent upon two points for its inversion might operate in a temporal space. Yet in both
identity ratherthan one. Thus, while a point in contour space contour space and durationspace, the operations of identity,
is immediately perceivable, a duration in d-space is not per- inversion, retrograde,and retrogradeinversioncan be shown
ceived until the second point of the pair defines its length. to model certain transformationsoccurringin non-tonal com-
Third, segments in c-space or d-space may be divided into positions. In contour space, experimentation has shown that
subsegments for comparison and analysis. Unlike c-space these transformationscan indeed be perceived;29in duration
segments, non-contiguous d-segments are probably per- space, however, some questions remain as to the percepti-
ceived by the listener only if their elements are associated bility of the I and RI operations.
by some other musical feature, such as extreme low or high
register (as in Ex. 4), by a pattern of accentuation, or by -'Of course, the pitch palindrome in these cases assists the listener in
some other mode of articulation.Thus, segmentation of non- perceiving the rhythmic palindrome.
contiguous d-segments should be considered analytically '9See W. J. Dowling and D. S. Fugitani, "Contour. Interval, and Pitch

only if one of these conditions holds. Recognition in Memory for Melodies," Journal of the Acoustical Society of
America 49 (1971), 524-531; Dowling. "Mental Structures Through Which
Finally, while the operations of inversion and retrograde Music is Perceived," Documentary Report of the Ann Arbor Symposium:
inversion have a clear perceptual basis in pitch-space and National Symposium on the Application of Psychology to the Teaching and
contour-space, applicationof these operations to successions Learning of Music (Reston, Va.: Music Educators' National Conference,
in duration-space may be more difficult to perceive. Music 1981), 141-149.
 The Perception of Rhythm in Non-Tonal Music 73

Qualifications aside, the theory proposed here does in- Table 1. Five forms of the tala dhruva
deed model aspects of rhythmic structure in non-metrical
differentjatis count total
music, not only in the Western non-tonal repertory that is of dhruva distribution counts dseg
considered here, but in certain non-Western musics as well.
1. tisra 3+2+3+3 11 <1011>
For example, in their discussion of South Indian rhythmic 2. chaturasra 4+2+4+4 14 <1011>
talas, Kanthimathi Kumar and Jean Stackhouse describe the 3. khanda 5+2+5+5 17 <1011>
divisions of tala in Karnatic music.30 They list the seven main 4. misra 7+2+7+7 23 <1011>
talas in terms of their number and grouping of counts, then 5. sankeerna 9+2+9+9 29 <1011>
note that each of these seven talas has five forms. An ad-
aptation of their table showing the five forms (Jatis) of the have been aimed at nearly complete and constant disruption
tala dhruva is given in Table 1.31 Note that each of the five of pulse, of any semblance of regularity in beat pattern ....
variations of dhruva has an equivalent rhythmic contour, It is difficult to find passages in Varese where the beat, or
< 1 0 1 1 >. Each of the other six main talas' Jatis shares the even some simple subdivision or compound of it, is literally
same dseg. Thus the seven talas are equivalence classes ac- stressed for more than a couple of measures."32 Bernard
cording to the definition of dseg proposed here. describes here precisely the type of nonbeat-based context
that listeners are most likely to perceive in terms of a rhyth-
AN ANALYTICAL APPLICATION mic contour.
Density 21.5 is the most often analyzed of Varese's com-
Generalization of contour theory to the temporal domain positions.33 Although the work's structure is as much
enables analysts to address two aspects of musical founded upon recurring rhythmic contours as it is upon re-
structure-melodic contour relations and non-metric rhyth-
mic structure-that are too often slighted in analyses of non- 32JonathanW. Bernard, The Music of Edgard Varese (New Haven and
tonal compositions. The analysis of Varese's Density 21.5 London: Yale University Press, 1987), 133.
33See, for example, Marc Wilkinson, "An Introduction to the Music of
that follows focuses upon these two types of analysis, noting
Edgar Varese," The Score and I.M.A. Magazine 19 (1957), 5-18; Martin
recurring melodic and rhythmic contours that work in con- Guimbel, "Versuch an Varese Density 21.5," Zeitschriftfiir Musiktheorie1
junction with pitch- and set-class structure to shape the (1970), 31-38; James Tenney with LarryPolansky, "Temporal Gestalt Per-
work's formal design. Varese's music provides an ideal con- ception in Music," Journal of Music Theory 24 (1980), 205-241; Jean-
text in which to illustrate analytical applications of the rhyth- Jacques Nattiez, "Varese's Density 21.5: A Study in Semiological Analysis,"
trans. Anna Barry, Music Analysis 1 (1982), 243-340: Carol K. Baron.
mic theory proposed here since, as Jonathan Bernard notes,
"Varese's Explication of Debussy's Syrinx in Density 21.5 and an Analysis
the composer's "penchant for rhythmic complexity seems to of Varese's Composition: A Secret Model Revealed," The Music Review 43
(1982), 121-134. In addition, Perspectivesof New Music 23 (1984) contains
3(KanthimathiKumar and Jean Stackhouse, Classical Music of South three articles in a Varese Forum:James Siddons, "On the Nature of Melody
India: KarnaticTraditionin WesternNotation (Stuyvesant, N.Y.: Pendragon in Varese's Density 21.5," 298-316; Jeffrey Kresky, "A Path through Den-
Press, 1988), 21-23. sity," 318-333; and Marion Guck, "A Flow of Energy: Density 21.5," 334-
31Ibid., 23. 347. Bernard's analysis concludes The Music of Edgard Varese, 217-232.
74 MusicTheorySpectrum

curring pitches or set classes, no analysis published to date Example8. Formaldesign of Varese, Density21.5
adequately addresses the issue of rhythmic structure. The A(mm. 1-23) B (mm. 24-40) A' (mm. 41-61)
work may be divided into three large sections, as illustrated
B-I "x" B-II A' B-II' "x" B-I'
in Example 8. Two of these sections are further subdivided;
24-28 29-32 32-40 41-45 46-51 51-53 53-61
B contains two contrastingsubsections marked B-I and B-II
in the example, while A' contains a return of material from
primary thematic material:
both the A and B sections. The boundary between subsec-
(mm. 1-3)
tions in each case is marked by a recurringfanfare-like mo-
tive labeled "x" in the diagram.
A recurringrhythmicfigure, dseg < 0 0 1 >, initiates most
A j I
phrases of the A section, as shown in melodies (a) through f - ===/ - tf

(f) of Example 9. Segment (d) consists entirely of this du-

ration succession. All of the remaining segments begin with (mm. 24-26)
either <00 3 2 1> or <0 0 2 3 1>. Although these two +
-+ +
successions do not share many embedded subsegments, com- - --3-- +
B-I
7 J. i1 I
parison of their matrices reveals a high degree of similarity; X I t I 3 77
P
F:P --:- -:::
their content is identical in nine out of ten positions. Three mp
of the six segments cited here-labeled (b), (c), and (f)-
begin with the rhythmiccontour < 0 0 2 3 1 >, yet their me- (mm. 32-33)

lodic contours and set-class structuresdiffer. Melody (c) may

8--------- ---- - - - -- - -- -
be heard as a variation of (b), since it immediately follows
(b) musicallyand since its rhythmiccontour differs only with
B-11
zXLz
f3 J
CI 3 --
I
respect to the final duration. Further, the pitches of melody
(c) are a literal subset of those in melody (b) (which, inci-
(mm. 29-30)
dentally, are the same pitches as melody (a)-thus (a), (b),
and (c) form a kind of "continuousvariation"). The melodic
contours of the first four notes of (b) and (c) differ a great
deal, however: in terms of adjacencies, <- + + > in (b) as ifr if'1::
JBVC ===- ff
opposed to < + - -> in (c). Although melody (f) begins
with the same rhythmic contour as segments (b) and (c), it
contains no common pitch classes with either segment. Fur-
ther, segment (f)'s melodic contour differs markedly from (e), are very similarrhythmically,since the rhythmiccontour
the others. This melody contains only three distinct pitches, of the latter, <0 0 3 2 1>, can be embedded contiguously
forming the repeated-note contour <2 1 2 0 1 2 0 1> and in the former. The first three notes of each forms the 3-1
set class 3-1 [0,1,2]. The remaining two segments, (a) and trichord, as did segment (f). Neither the melodic contour
 The Perceptionof Rhythmin Non-TonalMusic 75

Example 9. Primary rhythmic contours in Density 21.5, A and A' sections

A: A':- <00 1>, <00 2 1>, and < 0 1 2> embedded as dsubsegs
A 3--
(a) mm. 1-2 (g) mmn.4142 ;. |
ff~--
.
tL J!- "'u 1
S cf^f ff -=!==- p -

dseg = <0042135> cseg= <2130304> sc 54 dseg = <001> cseg = <102> sc 3-1

dsubsegs= <0021>,<00321> pcs = {1,4,5,6,7} pcs= {5,6,7}
by translation

-
A 3--- 3 L
(b) mm. 3-4 .-.I?n '-kJ (h) mm. 42-43 v i FJaMM- -i- n .. J - I -j
nf IP I I I
dseg= <002312> cseg= <213404> sc 54
dsubsegs= <001 2>,<0023 1> pcs = {1,4,5,6,7} dseg = <00333112224> cseg = <2131 2321304> sc 5-4
dsubseg= <001> pcs = {2,5,6,7,8}
bracketeddsubsegs: both = <00111>

. r 3
X - _ .
I
j_J_ I
(c) mm. 4-5
-O?c, t
"x" (based on rhythmic contour of (a)):
dseg= <002311> cseg = <232103> sc 4-13
dsubsegs= <0012>,<00231> pcs = {1,4,6,7}
O . _
nO\
(i) mm. 29-30
' -- ==
(d) m. 9 ,-^: b'r r r ff ff,
dseg = <0021> cseg= <3210> sc3-1
pcs = {5,6,7}
dseg = <001> cseg = <101> sc 2-1
pcs = {0, 1}

j) mm. 51-52 if;

I^- r-. -r I
(e)m. 15 'Ir F r
p dseg = <021> cseg = <210> sc 3-5
fz=-p pcs= {0,5,6}
dseg = <00321> cseg =<10202> sc 3-1
dsubseg= <0021> pcs= {3,4,5} 1 3 ?-1

(k) m. 52
~~-o
B 4P r ~
(f) mm. 21-22
p subito f ==-
dseg= <02111> cseg = <21021> sc 3-5
7.A dsubseg = <021> pcs= {0,5,6}
dseg = <00241345> cseg = <21201201> sc 3-1
dsubsegs= <0012>,<00231> pcs = {9,10,11}
76 MusicTheory Spectrum

nor the pitch-class content of (e) can be embedded literally used in this work, the analysis and discussion of which are
in (a), however. greatly enhanced by the precise language of contour theory.
The three-note figure with which the A' section begins, The first of these involves the composer's development of
segment (g) of Example 9, marks a return to the duration melodic material by registral expansion-that is, by varying
succession, melodic contour, and 3-1 set class of the com- the pitch-space realization of a reiterated c-segment. The
position's opening, although the precise pitches differ by a second involves his use of contour equivalence spanningboth
semitone. The composer repeats and expands this motive in the pitch-registraland temporal domains. Example 10 shows
melody (h), which follows (g) immediately in the score. Seg- three instances of pitch-space expansion within recurring
ment (h) is a member of set class 5-4 [0,1,2,3,6], as were equivalent contours, a technique that plays an importantrole
the opening two melodies of the work. The rhythmiccontour in linking the A and A' sections, as well as the B and B'
of this final reference to the material of the A section differs sections. The first example of contour expansion occurs
most stronglyfrom the contours of the rhythmsthat preceded within the A section in mm. 3-4 and mm. 13-14. Here, the
it. This duration succession, <0 0 3 3 3 1 12 2 2 4>, con- cseg < 1 0 2 3 > is expanded from a total pitch compass of
tains four instances of repeated equal durations, and more a minor third to one of a perfect twelfth. Likewise, the
closely resembles the repeated-durationsuccessions that are rhythm of mm. 13-14 represents a contour expansion of the
featured in the B section than the rhythmsof the A section. first in sequential time; both are d-space statements of
Motive "x," dividingboth the B and A' sections, provides <0 0 1 2>. Second, mm. 9-10 of the A section contain a
contrast to the material that surrounds it by virtue of its long melody that for two measures oscillates up and down
sudden change of register and dynamic, but the contour seg- a semitone, creating the cseg < 1 0 1 10 1 0 0>. This
ments and duration successions used here are not new to the contour recurs in the A' section, mm. 46-48, this time as a
work. Segment (i) of Example 9 contains three short state- minor-third oscillation. In the third instance, the melody
ments of "x"-the first and last forming dseg <0 0 2 1> beginning in the second measure of the B section (m. 22)
(which was heard previously in segments (a) and (e)) and the forms cseg <0 1 3 0 3 0>. It returns in the B' section ex-
central statement forming dseg <0 0 1>, the segment that panded in register by a semitone, and with one additional
has been heard repeatedly in the work as a kind of rhythmic cp .34
"head motive" of every melody discussed thus far. In this A strikingfeature of this work is the occurrence of several
case, however, the motive is initiated with equal-duration segments that are equivalent in contour space and duration
thirty-second notes rather than the sixteenths of most pre- space. Melody (a) of Example 11 is one such example. This
vious statements. The melodies of segments (i), (j), and the melody is structuredsuch that each successive pitch is both
initial notes of (k) are the continuously descending melodic higher and longer than the one that preceded it; thus both
contours < 2 1 0> and < 3 2 1 0 >; these contours recurreg- facets of its structure can be represented by the succession
ularly throughout the piece as do their inversions, to be <0 1 2 3>. Although melody (b) begins with a repeated
discussed in Example 10 following. duration, the rhythmicand melodic contours of the last three
This analysisof melodic and rhythmiccontours in Density
21.5 closes with a discussionof two compositional techniques 34Bernardnotes this relationship as well (p. 230).
 The Perception of Rhythm in Non-Tonal Music 77

Example 10. Contour expansion in Density 21.5

A section
mm. 34
f\ 3 mm
1314-
I .9p"f I
v ::
mf
cseg <1023> expanded
dseg<001 2> expanded
A and A' sections
mm. 9-10
A r--3---

?S
. e ' 3
mf subito
----------------------------------- ----------
8 ---
mm.46-48 3 >

I2Oeffxd
cseg
cseg< 1 01 0 1 01 01> expanded

B and B' sections

mm. 25-27
A\ + +

- 7 t ;.1 .
.g

4c .
.b
~Li -
-- r
l~ i 1 ".
> p5=- p ==- >p
>mp
- mp
mm. 56-58
ri
f).P
C *r. r- i
-
f aJ 1u
cseg<0 1 3030>expanded
(second time adds one cp 1)

notes can be represented by the succession <0 1 2>. In the registral and temporal domains as <0 1 2 3>. A case
addition, if the melody is aurally segmented by register, the might be made for the function of the continuously ascending
non-contiguous dsubseg formed by the three highest pitches melody as a cadential gesture in this work, since melody (c)
is also <0 1 2>. Melody (c) also contains one repeated du- serves this function immediately before the B-II section be-
ration (this time at the end of the melody) but, disregarding gins, and the ascending melody of mm. 44-45 (not shown
this repetition, its initial contour may be represented in both in Ex. 11) concludes the first subsection of the A' section.
78 Music Theory Spectrum

Example 11. A recurring melodic/rhythmic contour in Further, the melody that concludes the entire composition,
Density 21.5 melody (e) of Example 11, is the longest continuously rising
contour of the work, composed of contour pitches
3 ---- <0 1 2 3 4 5 6>. Although this duration succession does
(a) m.6 not represent continuously longer durations, the subseg-
ments (as indicated by the composer's slurring) contain two
fdse <01
dseg = <0123> cseg = <0123> sc 4-12 embedded statements of dsegment < 0 1 2 >. Thus, the final
pcs = {7,9,10,1}
melody of the work has been prepared aurally by similar
gestures, with both its contour and rhythmic structure heard
"_a- . f _
k1- ~ previously in other cadential contexts. Finally, segments (b),
-
(b) mm. 13-14 A e _ I I I I V
(c), and (e) also feature a continuous crescendo. Each pitch
ff I I in the bracketed successions is higher, longer, and louder
than the previous. Thus <0 1 2 3> is also represented in a
dseg = <0012> cseg = <1023> sc 3-5

csubseg = < 01 2 > (twice) "dynamic space," measured from soft to loud.
pcs = {4,9,10} In summary, generalization of contour theory into other
domains enables the analyst to compare diverse facets of
I musical structure along a single sequential scale. The ana-
(c) mm. 31-32 ' C!
,-
W,
-1
I> - lytical examples drawn from Octandre and Density 21.5 have
shown ways in which analysis of duration successions as
dseg= <01233> cseg = <01234> sc 5-5
i I
pcs = {4,8,9,10,11} rhythmic contours may clarify some aspects of one compos-
er's musical language.

(d) m.55
( 7Y ABSTRACT
This paper develops a theory that models nonbeat-based rhythms
d seg = <012> cseg = <012> sc 3--3 as "rhythmiccontours" of relative longs and shorts, drawing upon
pcs= {4,7,9} discussions of temporal spaces appearing in recent work of Robert
Morris and David Lewin and upon various music-psychologicalin-
n
vestigations of rhythmicperception. A new type of temporal space
;'
i 1I
W
(e) mm. 58-61
=Wr r i iI is proposed: a duration space (d-space) analogous to Morris'scon-
tour space, in which elements are ordered sequentially from short
I I I to long. After developing equivalence relations for d-space seg-
dsegs = < 012 >,<012> segmented by slurs ments, illustrated by excerpts from Edgard Varese's Octandre, the
cseg (entire excerpt) = < 01 2345 6> sc 7-Z37
pcs = I10, 11,1,2,3,5,6} paper concludes with an analysis of Density 21.5 that focuses upon
relationships among rhythmic contours.