How Rare Is Symmetry in Musical 12-Tone Rows?
How Rare Is Symmetry in Musical 12-Tone Rows?
How Rare Is Symmetry in Musical 12-Tone Rows?
0
11 1
10 2
9 3
8 4
7 5
6
9 10 0 3 4 6 5 7 8 11 1 2
Figure 1. Clock diagram and musical notation for the row from Schoenberg’s Serenade, opus 24, movement 5.
the pitch occurs. A low C played on the cello and a high C played on the piccolo share the same chroma,
called C, D, or 1. In both music and psychology, chromas are represented in a circular format [15], [9]; this
justifies our use of clock diagrams.
124
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 110
0 0
11 1 11 1
10 2 10 2
9 T1 (p) 3 9 3
I(p)
8 4 8 4
0
11 1
7 5 7 5
6 6
10 2
9 p 3
8 4
0 0
11 1 11 1
7 5
10 2 6 10 2
9 C 1(p) 3 9 R(p) 3
8 4 8 4
7 5 7 5
6 6
Figure 2. The row p from Figure 1 (center) transformed by transposition, inversion, retrograde, and cyclic
shift (clockwise from upper left).
playing the row k semitones higher. On 12-tuples, Tk is given by the following map:
Tk
( p0 , p1 , . . . , p11 ) −→ ( p0 + k, p1 + k, . . . , p11 + k) mod 12.
Inversion I reflects the clock diagram across the diameter containing the first note p0
(Figure 2, upper right). Musically, this corresponds to playing the row upside-down.
On 12-tuples, it becomes
I
( p0 , p1 , . . . , p11 ) −→ ( p0 , 2 p0 − p1 , . . . , 2 p0 − p11 ) mod 12.
Retrograde R reverses the arrows of the clock diagram (Figure 2, lower right), playing
the row backwards:
R
( p0 , p1 , . . . , p11 ) −→ ( p11 , p10 , . . . , p0 ).
Cyclic shift Ck moves the last k notes from the end of the row to the beginning:
Ck
( p0 , p1 , . . . , p11 ) −→ ( p0+k , p1+k , . . . , p11+k ),
If we also allow cyclic shift, we can represent a row class by adding a final line
segment to the diagram, forming a closed polygon with twelve vertices.3 Because a
polygon has no starting point, it cannot be altered by cyclic shift (change of starting
point). Allowing for cyclic shift as well as the other three transformations, the row
class for Berg’s Violin Concerto is shown in Figure 4.
Composers sometimes choose a row that is symmetric. A symmetric row is one that
is invariant under some transformation; geometrically, its row class has a symmetric
diagram. The simplest type of symmetry occurs in the row for the Chamber Sym-
phony by Schoenberg’s student Anton Webern (Figure 5), which is its own transposed
retrograde, i.e., p = T6 (R( p)). Such rows are called palindromes because, up to trans-
position, they are the same forward and backward. A palindrome’s row-class diagram
has rotational symmetry; the diagram for Webern’s Chamber Symphony, for example,
looks identical if rotated 180 degrees.
A more complicated type of symmetry occurs in the row from movement 5 of
Schoenberg’s Serenade, opus 24 (Figure 1); this row is its own transposed retrograde
inversion, i.e., p = T5 (R(I ( p))). Rows like this, which are symmetric under three
transformations (T , R, and I ), have row-class diagrams with mirror symmetry. The
2 We have borrowed the term cyclic shift from computer science. Music theorists often call it rotation,
because it involves rotating the elements of the ordered list ( p0 , p1 , . . . , p11 ). But this use of the word rotation
can be confusing when clock diagrams are present. Rotating a clock diagram does not correspond to cyclic
shift; instead, it corresponds to transposition.
3 We allow nonsimple polygons, i.e., nonadjacent sides may intersect.
126
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 110
7 10 2 6 9 0 4 8 11 1 3 5
Figure 4. Polygon and musical notation for the row from Berg’s Violin Concerto.
5 8 7 6 10 9 3 4 0 1 2 11
diagram for Schoenberg’s Serenade, for example, looks like a brandy snifter, with mir-
ror symmetry about the stem (Figure 3).
The most complicated type of symmetry occurs in the row from Berg’s Violin Con-
certo (Figure 4). This row is symmetric under four transformations; namely, it is a
transposed cyclic shift of its retrograde inversion, i.e., p = T4 (C3 (R(I ( p)))). Again,
the symmetry is evident from the row-class diagram, which has mirror symmetry
around the horizontal. Note that this row is not symmetric without cyclic shift; in
fact, removing any line from the diagram destroys the mirror symmetry.
Symmetric rows offer attractive musical possibilities. In the second movement of
his Chamber Symphony, Webern combines a palindromic row (Figure 5) with palin-
dromic rhythms and dynamics—creating a palindromic theme. In addition, symmetric
rows can be broken into shorter segments that are similar in content. (Some nonsym-
metric rows, called derived rows, share this property.) In Figure 1, for example, the
musical notation shows how the row breaks into two six-note halves, each of which is
a transposed retrograde inversion of the other. (Further inspection reveals that each six-
note segment can be broken into two three-note halves, each of which is a transposed
retrograde of the other. Thus, in addition to its ordinary symmetry, this row displays a
kind of nested symmetry—a fine point that we will not dwell on here.)
It is widely believed that the Viennese 12-tone composers, especially Webern, had a
penchant for symmetric rows [13], [8], [3]. Yet the number of symmetric rows used by
these composers is not overwhelming. Under transposition, retrograde, and inversion,
Let D = τ, ρ (or, equivalently, D = τ, σ ) and F = ρ. Note that D is the di-
hedral group of order 24. For any tone row α in S12 , the set of all tone rows that can
be obtained by applying a combination of retrograde, inversion, and transposition to
α (i.e., the row class of α) is the double coset Fα D. If we also allow cyclic shift, the
equivalence class becomes Dα D. We can detect a symmetric row by computing the
size of these double cosets. For a row α with no symmetry, every nontrivial transfor-
mation yields a new row, so |Fα D| = 48 and |Dα D| = 576. However, if α exhibits
symmetry under retrograde, inversion, and transposition, then |Fα D| < 48, and if a
nontrivial composite of (some of) all four transformations fixes α, then |Dα D| < 576.
Denote the set of classes of tone rows equivalent under transposition, retrograde,
and inversion (i.e., the set of row classes) by
T = {Fα D | α ∈ S12 }
and the set of equivalence classes under all four transformations (including cyclic
shift) by
R = {Dα D | α ∈ S12 }.
4 We multiply permutations left-to-right, so αβ means do α first, then do β.
128
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 110
We now investigate the structure of T and R. The number of row classes is provided
by two propositions:
Proposition 1. |T | = 9,985,920.
This result provides insight into the rareness of symmetric tone rows. The 26 · 6! · 7
αs for which |Fα D| = 24 are exactly the tone rows that can be transformed onto them-
selves by some composite of transposition, retrograde, and/or inversion. An argument
similar to the foregoing proof reveals that 26 · 6! of these tone rows exhibit symmetry
under a composite of transposition and retrograde only. (These rows are palindromic,
as defined earlier.)
Transposition and inversion are both right actions on S12 , and D is the group gener-
ated by these two actions, so the set of all rows that can be formed from a row α using
transpositions and inversions is just the left coset α D. Since |α D| = |D| for all α, there
are no tone rows symmetric via these two transformations alone. The corresponding
geometric observation is that, although a mirror reflection may fix the lines of a clock
diagram, it will reverse the arrows (see, for example, Figure 1).
Likewise, inversion and retrograde will not, by themselves, transform a tone row
onto itself, because inversion fixes the first note and retrograde does not. Thus the
remaining 26 · 6! · 6 symmetric tone rows are mapped onto themselves by a composite
of transposition, retrograde, and inversion. (The row in Figure 1 is an example.)
Table 1 enumerates the symmetric row classes under transposition, retrograde, and
inversion.
Table 2 also shows that very small coset sizes are especially rare. For example, only
two cosets have size 24; these represent highly symmetric structures that are important
in music theory. One is the chromatic scale, e.g., (0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11),
whose diagram is the dodecagon. The other is the circle of fifths, e.g.,
(0, 7, 2, 9, 4, 11, 6, 1, 8, 3, 10, 5), whose diagram is the dodecagram. (In music a
“perfect fifth” is seven semitones, so the circle of fifths is defined with pi = 7 + pi−1
mod 12.)
The GAP code for generating Table 2 is straightforward. To calculate the double
cosets Dα D for α in S12 , execute the following:
G:=SymmetricGroup(12);
D:=Group((1,2,3,4,5,6,7,8,9,10,11,12),(1,12)(2,11)(3,10)
(4,9)(5,8)(6,7));
x:=DoubleCosetRepsAndSizes(G,D,D);;
(Note that GAP uses the symbols 1, . . . 12 instead of 0, . . . , 11.) The list x is a collec-
tion of pairs: the first item of each pair is the coset representative, and the second item
is the size of the coset. The following functions return the total number of cosets, the
number of cosets of size 192, and a list of the coset representatives for all the size 192
cosets.
Size(x);
c192:=Filtered(x, L -> L[2]=192);;
Size(c192);
List(c192, L -> L[1]);
Given a representative of a double coset, the elements of the double coset are easy to
compute. For example,
AsList(DoubleCoset(D,(2,6)(3,11)(4,8)(5,9),D));
lists all of the elements in the coset Dα D for α = (1 5)(2 10)(3 7)(4 8). Thus it is
straightforward to implement our mathematical recipe for generating symmetric rows.
130
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 110
3. SYMMETRIC ROW CLASSES IN VIENNESE 12-TONE MUSIC. Under
transposition, retrograde, and inversion, symmetric row classes constitute just 0.13%
of the universe of possibilities. Yet they constitute 5% of the row classes in Schoenberg
(2 of 42) and 20% of the row classes used by Webern (4 of 21). It seems clear that
these composers liked symmetry. If they had chosen row classes at random, without
regard to symmetry, it is improbable that either composer would have used such a
large number of symmetric row classes. (For Schoenberg, the probability would be
.0015; for Webern, 1.25 × 10−8 .)5
Even when cyclic shift is allowed, just 1.16% of all row classes are symmetric. Yet
9% of the rows used by Berg display this property (2 of 23). Again, the probability
is low (.024) that Berg would have used so many symmetric row classes by chance
alone.6
The types of symmetry used by the Viennese composers are also of interest. Nei-
ther of the symmetric rows used by Schoenberg, and only one of the four symmetric
rows used by Webern, is a palindrome (symmetric under transposition and retrograde
alone). The rareness of palindromes in Webern has led one scholar to infer that Webern
preferred non-palindromic symmetry [3]. We cannot endorse this view, however, since
even in the universe of symmetric rows palindromes are outnumbered 6 to 1 (Table 1).
One of Berg’s symmetric rows, by the way, is also a palindrome; the other, as remarked
earlier, is symmetric under all four transformations (see Figure 4).
Although Viennese tone rows display above-chance levels of symmetry, more than
90% of them are not symmetric. Evidently most rows were chosen on the basis of other
criteria. This conclusion agrees with the judgment of music scholars, who have sug-
gested several other row types that may have interested these composers—including
“combinatorial,” “all-interval,” and “tonally colored” rows [2], [13].
Because Viennese composers had a variety of preferences, it can be hard to know
whether a row is symmetric because the composer wanted symmetry, or because the
composer wanted another property that happens to be related to symmetry. For ex-
ample, both Webern (once) and Schoenberg (twice) used rows that are symmetric only
when cyclic shift is allowed [10]. At first, this seems surprising: Webern never used
cyclic shift, and Schoenberg had abandoned it by the time of his cyclically symmetric
rows. However, both Schoenberg and Webern favored “hexachordally derived” rows
that could be broken into similar six-note halves [2]; here, that preference led to rows
that are symmetric using 6-fold cyclic shift (C6 ).
Because other preferences may be confounded with symmetry, it seems reason-
able to ask whether Viennese rows are remarkably symmetric given the constraints
imposed by the composers’ other preferences. This question is hard to answer alge-
braically; however, an approximate answer can be obtained using statistical methods.
First, one generates rows from (say) 1000 randomly selected row classes.7 Then, using
a statistical classification technique, one attempts to discriminate these random row
classes from those actually used by Viennese composers. Results using this approach
suggest that, even when other preferences are accounted for, Viennese tone rows still
display a significant penchant for symmetry [16].
5 These probabilities are based on a hypergeometric distribution with n = 9,985,920 row classes of which
r = 13,440 are symmetric (as in Table 1). For example, if Schoenberg had drawn 42 row classes from this
distribution, the probability that 2 or more would be symmetric is .0015.
6 This probability is based on a hypergeometric distribution with n = 836,017 row classes of which r =
ever, this will tend to underweight the symmetric rows, which have smaller row classes (cosets). One must
compensate for this underweighting, or avoid it by sampling row classes directly.
REFERENCES
1. E. Amiot, Mathématiques et analyse musicale: une fécondation réciproque, Analyse Musicale 28 (1992)
37–41.
2. M. Babbitt, Some aspects of twelve-tone composition, The Score and I.M.A. Magazine 12 (1955) 53–61.
3. K. Bailey, The Twelve-Note Music of Anton Webern: Old Forms in a New Language, Cambridge Univer-
sity Press, Cambridge, 1991.
4. The GAP Group, GAP—Groups, Algorithms, and Programming, Version 4.2, Aachen, St Andrews, 2000,
(http://www-gap.dcs.st-and.ac.uk/∼gap).
5. S. W. Golomb and L. R. Welch, On the enumeration of polygons, this M ONTHLY 67 (1960) 349–353.
6. E. Haimo, Schoenberg’s Serial Odyssey: The Evolution of His Twelve-Tone Method, 1914–1928, Claren-
don Press, Oxford, 1990.
7. E. Haimo, The evolution of the twelve-tone method, in The Arnold Schoenberg Companion, W. B. Bailey,
ed., Greenwood Press, Westport, CT, 1998, pp. 101–128.
8. D. Headlam, The Music of Alban Berg, Yale University Press, New Haven, 1996.
9. E. Krenek, Musik und Mathematik, in Über neue Musik, Verlag der Ringbuchhandlung, Vienna, 1937,
pp. 71–89.
10. J. London, P. von Hippel, D. Huron, J. Cartano, K. Kingery, B. Olsen, and T. Santelli (P. von Hippel, ed.),
Row Forms in the Serial Works of Schoenberg, Berg, and Webern, Center for Computer-Assisted Research
in the Humanities, Stanford, CA, 2001; http://www.ccarh.org/publications/data/humdrum/tonerow/
11. R. C. Read, Combinatorial problems in the theory of music, Discrete Math. 167 (1997) 543–551.
12. D. L. Reiner, Enumeration in music theory, this M ONTHLY 92 (1985) 51–54.
13. J. Rufer, Composition with Twelve Notes Related Only to One Another (trans. H. Searle), Macmillan,
New York, 1954; original German book published in 1952.
14. A. Schoenberg, Composition with twelve tones (1), in Style and Idea (trans. L. Black), L. Stein, ed.,
Faber & Faber, London, 1975, pp. 214–244; original German essay published in 1941.
15. R. N. Shepard, Geometric approximations to the structure of musical pitch, Psychological Review 89
(1982) 305–333.
16. P. T. von Hippel and D. Huron, Tonal and contra-tonal structure in Viennese 12-tone rows, unpublished
manuscript (2001).
DAVID J. HUNTER teaches mathematics and computer science at Westmont College. He received his bach-
elor’s degree in 1990 from the University of Illinois, with a mathematical year abroad at the University of
Manchester. After a brief stint teaching high school, he left for the University of Virginia, where he received
his Ph.D. in algebraic topology under the supervision of Nick Kuhn. He enjoys listening to the atonal piano
improvisations of his son Andy (age five) and daughter Ruth (age three). He was the best man at Paul von
Hippel’s wedding.
Department of Mathematics and Computer Science, Westmont College, 955 La Paz Road, Santa Barbara, CA
93108
[email protected]
PAUL T. VON HIPPEL has published articles on music psychology and statistical models for musical struc-
ture. He holds a Ph.D in music from Stanford University and a master’s degree in statistics from Ohio State
University. Formerly a postdoctoral fellow in Ohio State’s School of Music, he now works as a statistician
in the same university’s Department of Sociology. His one-minute piano piece “Kodaly’s Music Box” can be
heard on the recent CD Don’t Panic! Sixty Seconds for Piano. He was the best man at David Hunter’s wedding.
300 Bricker Hall, 190 North Oval Mall, Ohio State University, Columbus OH 43210
[email protected]
132
c THE MATHEMATICAL ASSOCIATION OF AMERICA [Monthly 110