Professional Documents
Culture Documents
Chordal distance is a topic of great theoretical debate, as it suggests space in both the
physical and perceptual realms. The current paper introduces new tools for understanding
distance, or motion from endpoint s to endpoint t.” Figure 1 reproduces an interval from s to t:
In our paper, the endpoints of Figure 1 are represented by triads. Rather than focusing on
physical space, we aim to build a theory that depicts the perceptual distance between two triads.
This is particularly relevant to chromatic music of the nineteenth century, in which composers
R. 2 If these transformations represent “paths” between triads, an open question is: how do we
hear these paths? Is it possible that for any path, there are multiple “distances”?
First, we summarize some of the major theoretical works on chordal distance; we then
introduce several mathematical definitions and define our GIS; finally, we provide musical
examples of Beethoven (in his “Hammerklavier” Sonata, Op. 106) and Schubert (in his Piano
Sonata in Bb major, D. 960) that apply our new theoretical framework, underscoring the value of
1
Prior Theoretical Approaches
In his landmark 2001 book Tonal Pitch Space, Fred Lerdahl defines algorithms for
measuring distance between tonal chords. As an example, we illustrate how the pair of triads Bb
major (Bb+) and F# minor (F#-) are calculated within the diatonic “basic space,” which for Bb+
is the following:
a: Bb F
b: Bb F
c: Bb D F
d: Bb C D Eb F G A
e: Bb B C Db D Eb E F Gb G Ab A
Employing five levels, ^1 and ^5 are given a “weight” of 5; ^3 is given a weight of 3; other
diatonic scale degrees a weight of 2; and other chromatic notes a weight of 1. 3 The distance to
a: C# F#
b: C# F#
c: C# F# A
d: B C# D E F# G# A
e: A# B C C# D D# E E# F# G G# A
For Lerdahl, there are twelve distinctive pcs (as underlined) in the basic space of F#- compared
to those in the basic space of Bb+. In effect, this is mapping the relationship between any I and
bvi; Lerdahl shows that this is equivalent to one parallel move (I i) + one diatonic move (i
bVI) + one parallel move (bVI bvi). He concludes that the distance between these two
chords—in the context of the diatonic basic space—is a particularly high number (23). 4
The distance between these two chords changes substantially when, instead of using the
diatonic basic space, Lerdahl uses the triadic / hexatonic space, tapping into the hexatonic
2
properties of triads. 5 The basic space weights the tonic triad most heavily (^1, ^3, and ^5), but
instead of prioritizing diatonic scale degrees, it highlights the remaining members of the
hexatonic collection. As such, there are only six additional distinctive pcs between Bb major /
a: F#
b: C# F#
c: C# F# A
d: A# C# D E# F# A
e: A# B C C# D D# E E# F# G G# A
In fact, the total distance between these two chords in hexatonic space is 7, a significantly
smaller number than 23. From this, we already observe the possibility of viewing multiple
In his 2011 book A Geometry of Music, Dmitri Tymoczko claims that voice-leading size
should depend “only on how far the individual voices move, with the larger motion leading to
larger voice leadings.” 6 This seems fairly intuitive, as one could assume that when one travels
more distance, it requires more “work” to move from one triad to the next. So, Tymoczko’s
conclusion is based purely on counterpoint and de-emphasizes scalar and functional context.
“In principle, voice leading provides just one of many possible notions of musical
distance. We might sometimes want to conceive of musical distance harmonically
(based on common tones or shared interval content) or in a way that privileges
membership in the same diatonic scale—so that the F major triad is closer to C
major than E major is.” 7
This, however, generates an interesting case study. Take the following two pairs of chords, both
3
Figure 3a: C+ triad with its neighbor 6/4 chord (F+)
Figure 3b: Bb+ triad with its hexatonic pole, F#-, an “uncanny” juxtaposition 8
Although the total semitonal displacement between these two pairs of triads is the same, their
perceptual distances are qualitatively different, with the former possessing a diatonic
juxtaposition and the latter possessing a hexatonic one. Considering the Euclidean distance for
these triads, the results are even more “uncanny,” in that the functionally unambiguous neighbor
6/4 (Neo-Riemannian operation of “D”) possesses a larger distance (√02 + 12 + 22 = √5) than
√12 + 12 + 12 = √3.
map any triad (major or minor) into any other triad—in essence measuring how “far apart” two
triads are perceptually using different combinations of D, P, R, and L. This generates two
important conclusions: 1) including L in a model with D, P, and R improved the “fit” between
therefore is not redundant). 9 Of course, certain operations, such as “Slide,” (S) that rely on
compositions of these “basic” transformations (that is, the generators of the Schritt/Wechsel-
group), require larger amounts of perceptual work, despite the fact that only two semitones are
being displaced. 10 Studies by Rogers and Callender (2006) delve more deeply into the issues of
4
perceptual distance between trichords. While they conclude that the total sum of voice-leading
motion correlates with perceptual distance, this is shaded by the number of common tones. In the
“especially close.” 11 They believe that an accurate model for perceived musical distance “cannot
Thus, the primary goal of this paper is to provide a rigorous alternative method for
measuring voice-leading distances that accounts not only for geometric distance, but also
perceptual distance. The framework for our discussion was developed by neo-Riemannian
theorists over the last several decades, but in particular Richard Cohn, whose exploration of the
triadic universe and of Weitzmann and Hexatonic spaces generated the impetus for our current
paper. 13
As a brief review, let us examine the famous “Cube Dance” graph, as established by
Douthett and Steinbach, shown in Figure 4. 14 The twenty-four consonant triads can be
partitioned into eight pitch-class (pc) sums (1, 2, 4, 5, 7, 8, 10, 11); the remaining four slots
5
Each triad has two pitch classes in common with one augmented triad, as well as with two other
consonant triads. For example, F-, with pc sum = 1, has two pc’s in common with Ab aug (pc
sum = 0); additionally, F- has two pcs in common with F+ and Db+ (pc sum = 2). In addition,
each region is a group of six triads generated by the neo-Riemannian operations P and L, which
partitions the triads into four groups of six; each is called a “hexatonic region.” Ergo, our
triad can be viewed as either an upshift or a downshift of an augmented triad, generating four
different Weitzmann regions; for example, a Bb+ triad, with pc sum = 5, can be viewed as a
downshift from a Bb-D-F# augmented triad, located at pc sum = 6. As such, these two triads—
previously contained in the same hexatonic region—are now located in disparate Weitzmann
6
Figure 6: Weitzmann Regions
context, in which dual functional identities exist for any consonant triad—as up- or down-shifts
of one of the four augmented triads, or, on the other hand, as “floating” in one of the four
both W- and H-spaces, thus granting any consonant triad a dual identity (an extension of the
Our methodology for developing analytical applications using both W- and H-spaces is
inspired by Cohn’s 2012 book Audacious Euphony, in which he analyzes the first movement of
Schubert’s D. 960 in a similar fashion, 19 following his earlier work (1999) 20 as well as Michael
Siciliano’s 2002 dissertation. 21 Our main argument is expressed in the following axiom: if every
triad can possess multiple meaning, then there exist multiple perceptual voice-leading distances
between each triad, any one of which can be chosen to model specific analytical intuitions.
To this end, we return to our GIS framework. Every GIS contains a space 𝑆 of musical
elements; a mathematical group 𝐼𝑉𝐿𝑆 of intervals; and an interval function 𝑖𝑛𝑡 which is a map
from the Cartesian product 𝑆 × 𝑆 into the group of intervals. In our framework, the space S
7
contains each of the twenty-four major and minor triads, with each triad adopting coordinates in
both Weitzmann and Hexatonic spaces, for a total of forty-eight elements. Additionally, every
pair of elements in 𝑆 (a major or minor triad with either a “W” or “H” identity) is assigned a
Hexatonic element, a triad is contextualized as derived from one of the four hexatonic
collections. Each coordinate is represented as an ordered triple; the first entry is the pitch class
sum, where 3n represents the pitch-class sum of the nearest augmented triad (0, 3, 6, 9); the
second entry represents the semitone displacement (k = +/-1) from the nearest augmented triad;
and the third entry represents the pitch-class root r, allowing for the coordination of root-motion
Of course, when the set of elements is specifically major and minor triads (set class (037)), W
coordinates will always have a non-zero increment “k,” whereas H coordinates will contain a
middle entry of 0.
Figure 7 presents a complete chart that contains all major and minor (037) triads,
including their pitch-class sums and Weitzmann / Hexatonic coordinates. Using this chart as a
reference, let us return to the original pair of triads, Bb+ and F#-. Their pc sums are 5 and 4,
respectively, resulting in Hexatonic coordinates of (5, 0, 10) and (4, 0, 6). Their Weitzmann
coordinates, however, factor in the upshift and downshift from augmented triads; 5 is a
downshifted (k = -1) augmented triad of value 6, and 4 is an upshifted (k = +1) augmented triad
of value 3. This results in Weitzmann coordinates of (6, -1, 10) and (3, 1, 6), respectively.
8
Figure 7. Table of W and H coordinates for all major and minor triads (set class (037))
We should also note that the first two values in our triples are independent, despite their
appearance in the chart; e.g., when the first coordinate is 3n, the second coordinate is always
nonzero. This is the consequence of considering only (037) set classes. If we were to expand our
9
harmonic vocabulary beyond (037) set classes, however, then it would open up the possibility of
triples that are not being articulated in the current paper: for example, one can think of a C
augmented triad as an upshifted major triad, carrying the ordered triple (11, 1, 0) —just one
possibility that would open up the domain for our three values.
and H1 to W2, we can consider the four pairs as a compound interval, which can be written as a
four-dimensional vector:
Triad 1 Triad 2
< 𝑖𝑛𝑡 (W1, W2); 𝑖𝑛𝑡 (H1, H2); 𝑖𝑛𝑡 (H1, W2); 𝑖𝑛𝑡 (H1, W2) >
In order to fulfill the requirements of a GIS, we assert that each of the four elements of the
compound interval represents an interval in the Lewinian sense. We define each of these as a
“component” interval (𝑖𝑛𝑡 (A, B)). Consider one element from {W1, H1} of the form (xA, yA, zA);
and another from {H1, H2} of the form (xB, yB, zB). Then:
We choose the minimum value method (as opposed to absolute value) in order to preserve GIS
properties, resulting in the following interpretation: “i” is the change in perceptual pitch-class
sum (values from 0 to 6); “j” is the single semitone displacement of H- or W- spaces (values
from 0 to 2), and “k” is root motion (values from 0 to 11). The result forms our 𝐼𝑉𝐿𝑆: a group
structure that preserves modular arithmetic for each of our three values.
10
Consider the four possible paths from Bb+ to F#-, applying the four coordinate triples
from 𝑆: W1 = (6, -1, 10); W2 = (3, 1, 6), H1 = (5, 0, 10) , H2 = (4, 0, 6). The compound interval
(< 𝑖𝑛𝑡 (W1, W2); 𝑖𝑛𝑡 (H1, H2); 𝑖𝑛𝑡 (H1, W2); 𝑖𝑛𝑡 (W1, H2) >) would be:
The (W1, W2) component interval (3, 2, 4) can be interpreted as 3 pitch-class sums “away”; 2
displacement shifts; and root motion of 4 semitones. When considering these two triads in terms
of Figures 5 and 6, one can see that staying within the same hexatonic region requires
significantly less perceptual “work,” confirmed by our (H1, H2) of (1, 0, 4). However, choosing
In sum, our GIS coordinates changes of pitch-class sum (that is, the chord’s hour location
on the clock); changes in semitonal displacement from the nearest augmented triad; and changes
in root identity. Since every consonant triad possesses both Weitzmann and Hexatonic identities,
musical context allows for both of these personalities to be expressed in the same framework.
Analysis
1. Schubert, D. 960
The first example excerpts the famous exposition from Schubert’s Piano Sonata in Bb
major, D. 960, in which Bb+ is abruptly confronted by F#-; this is a PLP neo-Riemannian path.
The perception of these two triads has not only harmonic, but also formal ramifications. When
the music asserts F#-, there is clearly a paradox regarding how “far we have gone.” Using a
strictly Hexatonic lens, we have not gone very far, remaining in the same region; this is
confirmed by the (H1, H2) component value of (1, 0, 4), interpreted as only moving one pitch-
class sum. Using a strictly Weitzmann lens, however, we have crossed over from one region to
11
Example 1. Schubert D. 960, Exposition
another; this is confirmed by the (W1, W2) value of (3, 2, 4). This directly ties into how one hears
this new thematic section; the (W1, W2) interpretation is more likely to suggest that the music has
moved into a new tonal area—potentially suggesting the onset of a secondary theme group.
This raises the question of how to interpret the components (H1, W2) and (H1, W2), which
appear to “split the difference” between the dually-hexatonic and dually-Weitzmann distances.
First, one notices that component intervals are an identical (2, 1, 4)—2 pitch-class sums away, 1
displacement shifts, and root motion of 4 semitones. This congruence between (H1, W2) and (H1,
W2) occurs for every triadic pair that does not employ a neo-Riemannian operation endowing a
shift of multiple semitones (e.g., R, which displaces two semitones). Additionally, this
underscores the paradox that this (and every) pair of triads contain attributes of both W and H
regions.
12
2. Beethoven, “Hammerklavier”
We now turn our attention to Beethoven’s “Hammerklavier” Sonata, Op. 106, which
offers similarly elusive harmonic juxtapositions as does the Schubert, including the same tonic
key of Bb major. Example 2 excerpts the end of the development and onset of the recapitulation.
It shows the interpolation of the confounding B+ triad (m. 201) interjected after an arrival on D+
(m. 197). The neo-Riemannian path from B+ to D+ is RP, resulting in distinct (W1, H2) and (H1,
W2) components; in fact, it is (W1, W2) and (H1, H2) that are identical! The full compound
interval (< (W1, W2); (H1, H2); (H1, W2); (W1, H2) >) is as follows:
13
In this case, it is the (H1, W2) path that is actually the “farthest.” This paradoxical path to B+
eventually takes us to its dominant, F#+ (also boxed), sharing the same pitch class sum as tonic
Bb+. So, what is the function of this B+? Ultimately, it serves to postpone Bb+, creating
mathematical (and thus perceptual) ramifications. By moving to B+, Bb+ is reached via F#, or
major would have gone directly back to Bb major via PL, without any interpolated B+ (and thus
no F#!). The compound interval for these two disparate neo-Riemannian paths LP and PL is
identical: (< (0, 0, 4); (0, 0, 4); (1, 1, 4); (1, 1, 4) >).
Bb major with its harmonic foil of the first movement, B minor, as well as considering long
range juxtaposition in later movements, including but not limited to the end of the second
movement and the beginning of the third movement, bearing the hexatonic relationship of Bb
major and F# minor, and, of course, endless possibilities in the fourth movement.
Further Considerations
We have demonstrated that any consonant triad can be conceptualized as possessing both
W- or H-identities, and that it is possible to measure four different types of perceptual voice-
leading intervals among these multiple meanings, thus reintroducing a specific notion of function
displacements of diminished seventh chords (termed “Boretz Spiders by Cohn 22) and the
intervening spaces (the three octatonic pools). While set-class (0258)—that is, half-diminished
14
voice-leading enjoyed by the privileged members of set-class (037), we can bridge the 2-
Thus, in the same manner as earlier with regard to triads, any seventh chord can be characterized
as being derived from a diminished seventh chord (that is, Boretz-flavored), or as floating in an
octatonic pool (that is, derived from the referential collection of one of the three octatonic
scales). Consequently, one can construct an equivalent GIS to model perceptual distances among
functionally contextualized seventh chords; the Boretz regions are akin to the Weitzmann
regions, and the octatonic pools are akin to the hexatonic spaces. 23
having dual functional identities: one related to its derivation from a scalar collection (Hexatonic
or Octatonic), and one related to minimal displacements of an equal division of the octave (an
perceptual intervallic distances, and enriches our hearing of the spaces between chords.
15
1
See David Lewin, Generalized Musical Intervals and Transformations (New Haven, CT: Yale University Press,
1987; reprint: Oxford University Press, 2007), see p. xxix. “A Generalized Interval System (GIS) is an ordered triple
(S, IVLS, int), where S, the space of the GIS, is a family of elements, IVLS, the group of intervals for the GIS, is a
mathematical group, and int is a function mapping S x S into IVLS, all subject to the two conditions (A) and (B)
following. (A): For all r, s, and t in S, int (r, s) int (s, t) = int (r, t). (B): For every s in S and every i in IVLS, there is
a unique t in S which lies the interval i from s, that is a unique t which satisfies the equation int (s, t) = I” (p. 26).
2
To name a few: Richard Cohn, “Neo-Riemannian Operations, Parsimonious Trichords, and Their ‘Tonnetz’
Representations,” Journal of Music Theory 41/1 (1997): 1-66; Julian Hook, “Uniform Triadic Transformations,”
Journal of Music Theory 46/1-2 (2002), 57-126; Brian Hyer, “Reimag(in)ing Riemann,” Journal of Music Theory
39/1 (1995): 101-138; David Kopp, Chromatic Transformations in Nineteenth-Century Music (Cambridge
University Press, 2002).
3
See Fred Lerdahl, Tonal Pitch Space (Oxford University Press, 2001). Specifically: “Level a is octave (or root)
space, level b is fifth space, level c is triadic space, level d is diatonic space, and level e is chromatic space. In
contrast to Deutsch and Feroe, a fifth space is included here, because the fifth is more stable than the third of a triad
and because the fifth becomes the basis for shifting the space. The seventh-chord level is excluded because in
Classical music seventh chords have little independent status, the interval of a seventh usually behaving as a local
dissonance governed by voice-leading principles. If however, a seventh is judged to be harmonic, it can be added at
the triadic level” (p. 47).
4
Ibid., p. 69.
5
Ibid., pp. 258-63.
6
See Dmitri Tymoczko, A Geometry of Music: Harmony and Counterpoint in the Extended Common Practice
(Oxford University Press), p. 50.
7
Ibid., p. 51, n. 30.
8
This juxtaposition is the central topic in Richard Cohn’s 2004 article “Uncanny Resemblances: Tonal Signification
in the Freudian Age,” Journal of the American Musicological Society, 57/2 (2004), pp. 285-323.
9
See Carol Krumhansl, “Perceived Triad Distance; Evidence Supporting the Psychological Reading of Neo-
Riemannian Transformations,” Journal of Music Theory 42/2 (1998): pp. 265-282.
10
Of course, including other Schritt and Wechsel operations of Riemann (as noted in Krumhansl 1998, p. 275)
would generate different psychological “paths.” As in her article, the “neo”-Riemannian operations are most central
to this paper.
11
See Nancy Rogers and Clifton Callender, “Judgments of Distance Between Trichords,” Proceeding of the 9th
Annual International Conference of Perception and Cognition (2006), University of Bologna, p. 1687.
12
Ibid., p. 1691.
13
See Richard Cohn, “Square Dances with Cubes” Journal of Music Theory 42/2 (1998): 283-296; “Weitzmann’s
Regions, My Cycles, and Douthett’s Dancing Cubes,” Music Theory Spectrum 22/1 (2000): 89-103; Audacious
Euphony: Chromatic Harmony and the Triad’s Second Nature (Oxford University Press, 2012).
14
See Jack Douthett and Peter Steinbach. “Parsimonious Graphs: A Study in Parsimony, Contextual
Transformations, and Modes of Limited Transposition,” Journal of Music Theory 42/2 (1998): pp. 241-263.
15
Ibid., p. 254.
16
Figures 5 and 6 are alternative illustrations of Cohn’s original figures (2000, p. 97).
16
17 Cohn 2012, p. 105: “We define three such transformation classes: an H class, consisting of the three members of
the hexatonic group (L, P, and H); a W class, consisting of the three members of the Weitzmann group (R, N, and
S); and an E class, consisting of the transformations that map triads within their own zone (LP, PL, and the identity
operation E).”
18
See Janna K. Saslaw and James P. Walsh, “Musical Invariance as a Cognitive Structure: ‘Multiple Meaning’ in
the Early Nineteenth Century,” in Music Theory in the Age of Romanticism, ed. Ian Bent, pp. 211-232 (Cambridge
University Press, 1996). Saslaw and Walsh state that for Vogler, “Multiple Meaning is an important compositional
tool to be used to lead the listener into unexpected territory,” (p. 217); the augmented triad (as well as the
diminished seventh) is among those that possess this quality (p. 218). Weber also treats these two chords, with the
“modern-day augmented-sixth chords as transformations of the half-diminished-seventh chord on scale degree 2 in
minor” (p. 220). In summary: “For Weber, Multiple Meaning is used by composers to create variety and richness,
and to smooth the transition between distantly related keys. For Vogler, Multiple Meaning creates a sense of surprise
or deception—allowing the composer to “remodel” the listener’s heart at will (p. 222).
19
See Cohn 2012, p. 128.
20
See Richard Cohn, “As Wonderful as Star Clusters: Instruments for Gazing at Tonality in Schubert,” Nineteenth-
Century Music 22/3 (1999): pp. 213-232.
21
See Michael Siciliano, “Neo-Riemannian Transformation and the Harmony of Franz Schubert,” Ph.D. diss.,
University of Chicago, 2002. Siciliano comments on Cohn’s early (1999) conception of traversing W- and H- spaces
at once, highlight the paradoxical nature of voice-leading “function”: “It is not self-contradictory because the aspects
occur in two distinct ‘spaces’: a harmonic/triadic space and a motivic/melodic space. However, we are asked to
interpret the same phenomenon, shared pitch classes, in both spaces. In the triadic space F#- shares no pitch classes
with Bb+, and thus although both still in the tonic region, is as far removed from Bb+ as possible. At the same time
in the melodic space, F#- shares two pitch classes with the augmented triad that includes F, and thus has a
(melodic?) affinity for the dominant. These spaces are not well defined or differentiated. The augmented triad
appealed to by the melodic space is defined by the roots of the triads in the dominant LP region. Conversely, the
harmonic space of the LP regions is defined by melodic relations (the minimal displacements). Further, the melodic
sharing of pitch classes with an augmented triad associated with a region creates the paradoxical result that ALL the
minor triads in the tonic region have a greater affinity to the dominant region than ANY of the minor triads in the
dominant region” (p. 96).
22
See Cohn 2012, p. 157.
23 A similar approach could be adopted for every possible space generated by equal division of the octave. For
instance, we could generate a whole-tone division in which the minimal displacements are Mystic chords (but, of
course, again encountering the reality of (037)’s privileged status, in that there are no efficient voice-leadings among
Mystic chords derived from different whole-tone scales).
17