See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/238473530

Invariance properties of Schoenberg's tone row system

Article in The Journal of the Australian Mathematical Society Series B Applied Mathematics · January 1980
DOI: 10.1017/S0334270000002368

CITATION READS

1 115

2 authors:

James Fill Alan Izenman

Johns Hopkins University Temple University
124 PUBLICATIONS 4,358 CITATIONS 72 PUBLICATIONS 4,393 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Handwriting View project

All content following this page was uploaded by Alan Izenman on 28 September 2014.

The user has requested enhancement of the downloaded file.

J. Austral. Math. Soc. 21 (Series B) (1980), 268-282

INVARIANCE PROPERTIES OF SCHOENBERG'S

TONE ROW SYSTEM

JAMES A. FILL and ALAN J. IZENMAN

(Received 13 July 1978)

(Revised 25 October 1978)

Abstract

This paper organizes in a systematic manner the major features of a general

theory of /w-tone rows. A special case of this development is the twelve-tone
row system of musical composition as introduced by Arnold Schoenberg and
his Viennese school. The theory as outlined here applies to tone rows of
arbitrary length, and can be applied to microtonal composition for electronic
media.

1. Introduction

Musical composition in the twentieth century has been enlivened by Arnold

Schoenberg's introduction of a structured system which emphasizes its serial and
atonal nature. Schoenberg called his system "A method of composing with twelve
tones which are related only with one another" [12] page 107. Although Schoen-
berg himself regarded his work as the logical outgrowth of tendencies inherent
in the development of Austro-German music during the previous one hundred
years, it has been criticized as purely "abstract and mathematical cerebration"
and a certain amount of controversy still surrounds the method.
The fundamental building-block in Schoenberg's system is the twelve-tone row,
a specific linear ordering of all twelve notes—C, C*, D, Eb, E, F, F s , G, Gs, A,
Bb, and B—of the equally tempered chromatic scale, each note appearing once
and only once within the row. A work, or section of a work, based on a particular
twelve-tone row (usually designated as the original or prime) is permitted to
include all possible transpositions of the prime, together with those of its retro-
grade, its inversion and its retrograde-inversion. (See the next section for explanations
of this terminology.) Furthermore, each note can be stated in any register.
268
[2] Schoenberg's tone row system 269

Theoretically, no note is more important than any other note, and none of the
four forms has primacy over the other three.
Although certain modifications have been made to this set of rules in recent
years, the basic principles have been followed fairly faithfully. Notable composers
who have used the twelve-tone row system in their musical compositions include
Alban Berg, Anton Webern, Karlheinz Stockhausen, Pierre Boulez, Igor
Stravinsky and Milton Babbitt. A more detailed description and a critical
examination of the method may be found in Perle [11].
The aim of this paper is to present the algebraic flavour and structure of
Schoenberg's system of musical composition. Numerous studies have appeared
in the musical literature dealing with various music-analytic properties of twelve-
tone rows, most notably those of Milton Babbitt ([1], [2], [3], [4] and [5]), who
has given the subject a great deal of its nomenclature. However, none of these
studies has attempted to determine the number of twelve-tone rows which possess
certain interesting properties, such as being invariant under any of the operations
of transposition, retrograde, inversion or retrograde-inversion. The inter-
relationships between the four forms of the prime twelve-tone row can be studied
through the set-complex, an ingenious tool devised by Babbitt in [1]. Each twelve-
tone row generates its own set-complex which contains all of its various forms;
it is of interest, therefore, to determine how many distinct set-complexes are
generated by the complete set of all 12! twelve-tone rows.
Such questions are of considerable importance to the musician who wishes to
compose in the twelve-tone row genre. In order to answer these questions, we
analyse the twelve-tone row system within the more general framework of m-tone
rows, where m is an arbitrary integer. The elements of such an m-tone row are
some permutation of the notes of an w-tone scale. The advantages of such a
generalization are not only mathematical; with the advent of electronic music
produced by a suitably programmed computer or synthesizer, the study and
composition of microtonal works in which intervals between notes are smaller
than a semi-tone is now feasible.
A list of all twelve-tone compositions by Schoenberg together with their
associated twelve-tone rows may be found in [10].

2. Notation

For notational simplicity and technical ease, we shall substitute the integers 0
through m — 1 for the ordered succession of m ascending notes that constitute the
w-note scale in question. Such a representation is standard for twelve-tone row
theory (where m = 12; see Chapter 1 of Perle [11]) and is probably necessary for
the theory of microtonal composition.
270 J. A. Fill and A. J. Izenman [3]

Let m > 2 be an integer and let

0 1 2 ... m-\
n=
be a permutation of Zm = {0, l,...,m— 1}. Then FI is called an m-tone row (or
m-phonic sequence). The set of all m! w-tone rows is denoted by £Pm. For example,
the twelve-tone row

/0 1 2 3 4 5 6 7 8 9 10 11 \
11 = , (*)
\9 10 3 11 4 6 0 1 7 8 2 5/

which can also be written in the familiar cycle notation as

n = (0 9 8 7 l 10 2 3 1 1 5 6)(4),
appears in Schoenberg's Violin Concerto of 1936.
We now define three operations on II. Let a be the cyclic permutation
(0 1 ... m— 1), which corresponds to a raising each tone by one unit. Then the
transposition of II by the integer a is the m-tone row II <xa. Thus the transposition
of II in (*) above by 3 yields the twelve-tone row
0 1 2 3 4 5 6 7 8 9 10 11
0 1 6 2 7 9 3 4 10 11 5 8

= (0)(l)(2 6 3)(4 7)(5 9 11 8 10).

Next, let j3 be the permutation (0 m—1)(1 m—2)(2 m —3) ..., which corresponds
to reversing the order of the elements. Then the retrograde (or crab) of II is the
m-tone row jSFI. Thus the retrograde of II in (*) above is

0 1 2 3 4 5 6 7 89 10 11
5 2 8 7 1 0 6 4 1 1 3 10 9

= (0 5)(1 2 8 11 9 3 7 4)(6)(10).

Note that jS2 = 1. Finally, let y be the permutation (1 m - l ) ( 2 m - 2 ) ( 3 m - 3 ) ....

Then the negative of II is the /M-tone row Fly. The negative of II in (*) above is

0 1 2 3 4 5 6 7 8 9 10 11
3 2 9 1 8 6 0 1 1 5 4 10 7

= (0 3 1 2 9 4 8 5 6)(7 ll)(10).
[4] Schoenberg's tone row system 271

Note that ay = ya" 1 . The inversion of an w-tone row II is now the w-tone row
II cr"2"°y ( = II ya2"0). For the II in (*) above, the inversion is

0 1 2 3 4 5 6 7 8 9 10 11
9 8 3 7 2 0 6 5 1110 4 1

= (0 9 10 4 2 3 7 5)(1 8 11)(6).

It is worth noting that this last definition is not the usual way in which inversion
is defined. Perle [11] page 3, for example, calls the negative of a tone row II the
inversion of II. However, for certain reasons which will become apparent later*
we wish to set the first element of the inversion of II equal to the first element of
II itself. This requirement leads naturally to the above alternative definition. The
retrograde-inversion (or crab-inversion) of II is the retrograde of the inversion of
II and is the w-tone row f$Ylori7'ay. Thus the retrograde-inversion of II in (*)
above is

/ 0 1 2 3 4 5 6 7 8 9 10 11
BIIa 6 y =
\ 1 4 10 11 5 6 0 2 7 3 8 9

= ( 0 1 4 5 6)(2 10 8 7)(3 11 9).

It will be convenient to use the following abbreviated version of an m-tone row Ft

in the discussion below. Instead of writing II in the form (1) above, we shall
henceforth write either

I I = (TTQ, 77 1( 7T2, . . . , 7 7 m _ 1 )

or simply

the permutational ordering of the elements being understood.

The following results are now straightforward consequences of the above
definitions.
(i) The retrograde of a transposition is the same transposition of the
retrograde; that is, jS(IIaa) = (j8II)a a .
(ii) The inversion of a transposition is the same transposition of the inversion;
that is, (na a )a- 2 ( l 7 » + a ) y = (na- 2 "°y)a a .
(iii) The retrograde-inversion of a transposition is the same transposition of the
retrograde-inversion; that is, j3(na II )«- 2( "» + ' ll y = (fiHor27">y)a.a.
272 J. A. Fill and A. J. Izenman 15]

Furthermore, because of the way in which we have defined the inversion of II,
the retrograde of the inversion of II is not in general the same as the inversion of
the retrograde of n ; that is,
(iv)
unless m is even and-|77m_1 —7ro| = w/2. On the other hand, however, the usual
definition of the inversion of II, namely IIy, always results in their equality.

3. The set-complex .

The four forms of an w-tone row and the m transpositions of each form can be
related to each other in the following useful way. The set-complex, •(!!)» corre-
sponding to a particular m-tone row II is the set of all w-tone rows of the four
forms Hoca, pU<xa, na~27r°ycto and /?na~2"°ycta, and may be represented as an
mxm square (hence the symbol -D(II)) whose entries are arranged in a "checker-
board" fashion as follows. Along the top line write the original m-tone row IT.
Next, write the m-tone row corresponding to the inversion of II, namely
1(11) = Hoc'2"oy, down the first column; from the way we have defined the
inversion of II, the starting element of both II and its inversion will be the same.
The square is completed by writing in each line the transposition of II which
starts with the element of the inversion in the first column. As an example we
give in Table 1 the set-complex of Schoenberg's Fourth String Quartet for which
m = 12. The rows reading from left to right give all 12 transpositions of II; the
columns reading from top to bottom give all 12 transpositions of /(FI) = Ua~27">y;
the rows reading from right to left give all 12 transpositions of -R(II) = jSII; and
the columns reading from bottom to top give all 12 transpositions of

TABLE 1

0 1 2 3 4 5 6 7 8 9 10 11

n ' 2 1 9 10 5 3 4 0 8 7 6 11
3 2 10 •11- 6 4 5 1 9 8 7 0
n«
no 5 7 6 2 3 10 8 9 5 1 0 11 4
n<*4 • 6
11
5 1 "2 9 7 8 4 0 11 10 3
10 6 7 2 0 1 9 5 4 3 8
n«»u 0 8 9 4 2 3 11 7 6 5 10
n«
Ila10
• 1
0 11 7 8 3 1 2 10 6 5 4 9
na 2 4 3 11 0 7 5 6 2 10 9 8 1
n«6
iia'
8
9
7 3 4 11 9 10 6 2 1 0 5
8 4 5 0 10 11 7 3 2 1 6
iia 8 10 9 5 6 1 11 0 8 4 3 2 7
net3 5' 4 0 1 8 6 7 3 11 10 9 2
 [6] Schoenberg's tone row system 273

Notice that the elements of such a "checkerboard" representation form a

Latin square (see [6]). However, this square array is just a convenient method of
writing down all of the 4m possible m-tone rows that can be generated from a
given m-tone row, and is not of any special interest by itself. The fact that •(!!)
is a Latin square is not used in the remaining sections of this paper.
Duplication of m-tone rows within a given set-complex will now occur if any
of the following relationships hold for some a:
n = ria a (i)
n = j8n« a (2)
n=Ila- a f f »ya o (3)
II = pna-2*<>y*a (4)
jSn =n.a- 2 "°ya a (5)
j8n=j3na- 2 "»ya° (6)
I! a"2"" y = j8n a"2"" ya°. (7)
These relationships express the possible invariance properties of an m-tone row
under the operations of transposition, retrograde, inversion, retrograde-inversion,
and of all combinations of such operations. The question is now which of these
invariances can occur and which cannot, and of those that can, to characterize
them and determine their exact number.
Clearly, an m-tone row can never be a nonzero transposition of itself; that is,
n ^ F I a 0 if a mod m ^ 0 . Moreover, an m-tone row can never be a transposition of
its own inversion; that is, for all a, n^na"~ 2 "°ya a . For, if

II = na- 2 "oya° = nya 2 *° +a ,

then y = a ~ 2 *°- a , which is a contradiction. Relation (6) can also never occur
since (3) can never occur. This eliminates the possibilities of (1), (3) and (6) above.
On the other hand, if an m^tone row FI is a nonzero transposition of its own
retrograde (that is (2) holds), then m is even and II is the transposition of that
retrograde by m/2. For, suppose that II = /2Ilaa. Then £11 = IIa o , whence
II = IIa 2a and 2amodm = 0. This last implies that m is even and that
amodm = m/2, as asserted. With regard to (4) above, an m-tone row will be a
transposition of its own retrograde-inversion if and only if II = {SYlya? for some
ceZm; furthermore, if m is even, then c will be odd. To see this, note that

*2 vt = m(m-l)/2=mil (mod m).

i=0
274 J. A. Fill and A. J. Izenman [7]

If II = pilya? holds, then •n-i=c--rrm_1_i (modm), 0 < / < (wi/2) — 1, and so

m-X
2 w(=mc/2 (mod m).

Thus c is odd as asserted. Relation (5) follows directly from (4), while (7) (with
a replaced by —a) follows from (2). Finally, statements (2) (that II is a trans-
position of its own retrograde) and (4) (that II is a transposition of its own
retrograde-inversion) are mutually-exclusive by virtue of the fact that (3) can never
occur.
From the above results, we make the following observations. The set-complex
• ( f l ) corresponding to a particular m-tone row II by definition contains Am
m-tone rows. The above results suggest, however, that for certain rows II, the
number of distinct /w-tone rows contained in D(II) may be smaller than Am.
Indeed, we have seen that duplications will occur in • ( ! ! ) if and only if the /w-tone
row n satisfies either of the following two conditions:

m is even and II = j8IIam/2, (1)

Ft = /3FIyac, for some ceZm; and if m is even, then c must be odd. (2)

Sections 4 and 5 below are concerned with the determination of the exact number
of w-tone rows for which conditions (1) or (2) hold. Moreover, the above remarks
allow us to examine cases (1) and (2) separately.

4. Results for m even

Let m > 2 be an even integer throughout this section. We will sometimes write
2k for m.

THEOREM 4.1. There are exactly 2kk\ m-tone rows FI for which U = j3IIafe.

PROOF. N O W II = j9ITafc if and only if II is of the form

,..., (TTX + k)modm,(no + k)modm).

All such sequences may be generated without repetition according to the following
two-step process. For each i = 0,1, ...,k-1: (i) successively, for7 = 0,1 k— 1,
choose the pair of positions 7rf and 7r2t_1_f to be occupied by the pair of values j
and j+k; then (ii) for each j = 0,1, ...,k— 1, decide whether 7^=7 and
TT2ic-i-i=J+k> o r whether •ni=j+k and 7r 2 f t _ 1 _ i =/ There are k\ possible
results of step (i), and for each of those there are 2k possible results of step (ii).
[8] Schoenberg's tone row system 275

THEOREM 4.2. There are exactly k2k k! m-tone rows XI for which there exists an
odd constant ceZm such that XI = /?IIya c .

PROOF. We first prove that there are 2kk\ m-tone rows Ft for which II = f$Xlya.
N o w this latter equality holds if and only if II is of the form

II = ( ^ o ^ u •••>77'fc-i»(l — " f c - O n i o d m , ...,(l—n^)modm,(I—TT0)modm).

All such sequences may be generated without repetition according to the following
two-step process. For each i = 0,l,...,k—l, (i) successively, for j= 1,2,...,k,
choose the pair of positions •ni and iru-i-ito be occupied by the pair of values j
and (1—j)modm; then (ii) for each j=\,2,...,k, decide whether •ni=j and
•7T2k-i-i = (1 —j) mod m, or whether nt = (1 —j) mod m and ^k-i-i = J- There are k!
possible results of step (i), and for each of those there are 2k possible results of
step (ii). To complete the proof, observe that II = pXlya? holds if and only if
Yla-d = fiXlordya., where c = 2d+l. The k choices for c (that is, for d) gives the
factor of k'mklkk\.

An important example of a composition in which the prime twelve-tone row n

is a transposition of its own retrograde-inversion is that of Schoenberg's Suite,
Opus 29, written in 1926. The twelve-tone row in question, transposed by 9 so
that its first note is C (or 0), is

n = (0,1,4,5,8,9,2,3,6,7,10,11);

in this case, II = jSIIya11. (Note also that for this example the sequence of 11
intervals between the 12 notes, namely (1,3,1,3,1,5,1,3,1,3,1), is symmetric.
Similar relationships between intervals and notes are explored in detail in [7].)

Contained in the above theorems is the following:

COROLLARY 4.3. There are exactly 2k(k+1)! m-tone rows Ufor which O{Xl) has
fewer than 4m distinct members and, in such cases, the cardinality of D(II) is 2m.

THEOREM 4.4. If D(^ m ) = {D(II): I l e ^ J , then D(^ m ) has cardinality

(m\+2k(k+l)l)/(4m).

PROOF. Let us say that two wi-tone rows II and II' are equivalent if D(II) = •(!!').
The number of equivalence classes is then the cardinality of D(^n)- Now II and
IT are equivalent if and only if (at least) one of the following four statements
holds for some aeZm: (i) IT = IIa o ; (ii) II' = 0IIa o ; (hi) XI'= Uor2l"ya.a;
(iv) IT = j8nor2*°yaa. Let G be the group of 4/w different permutations <ri>a
276 J. A. Fill and A. J. Izenman [9]

with 1 ^ /< 4 and aeZm given by

a2>a:

a3>a:

The group operation here is composition of functions. It is easily verified that G

is indeed a permutation group, and that the equivalence relation induced by G
on 0>m results in the equivalence of II and II' if and only if D(n) = D(n').
According to Burnside's Lemma (see [9] page 136), the number of equivalence
classes is
2 <A
where | G\ is the order of G (here | G\ = Am) and ip{a) is the number of elements
of 0>m which are invariant under a. Clearly, «A(cr10) = \0irr\ = m\ and ^(CT1JO) = 0
if a^O. Furthermore, i/>(cr2ja) = 0 unless a = m/2, in which case

by Theorem 4.1. Similarly, ip(<j3a) = 0 for all a. Finally, 2 ^ < A K , J = k2kk\ by

Theorem 4.2. Thus, the number of equivalence classes is
= (m! + 2k k!+k2k k !)/(4w) = (m!+2\k +1) !)/(4m).

THEOREM 4.5. 77*e cardinality o/{D(II): I I e ^ m and | D(n)|<4w} = {D(n):

II e^»m and \ 0(11)1 = 2m} is 2k-\k+ l)!/w.

PROOF. Similar to that of Theorem 4.4.

5. Results for m odd

Let m > 2 be an odd integer throughout this section. We shall sometimes write
2/c+lfor™.

THEOREM 5.1. There are precisely m2kk\ m-tone rows U for which there exists a
constant ceZm such that II =

PROOF. We first prove that there are 2kk\ m-tone rows n for which IT = /3Ily.
Indeed, II = /Slly if and only if FI is of the form
I I = (T7 0 , 7 7 l 5 . . . , 7r f c _ 1 ,0, — 7Tfc_i m o d w , . . . , — TTX m o d m, —nq m o d w ) .
[10] Schoenberg's tone row system 277

All such sequences may be generated without repetition according to the following
two-step process. For each / = 0,1, ...,k-l: (i) successively, for j = 1,2, ...,k,
choose the pair of positions ^ and •nm_x_i to be occupied by the pair of values j
and —j raodm; then (ii) for each / = 1,2, ...,k, decide whether iti=j and
Trm_1_i = —j mod m, or whether •ni = —j mod m and Trm_1_i = j . There are k!
possible results of step (i), and for each of those there are 2k possible results of
step (ii). To complete the proof, notice that II = fillya? if and only if
Ylord = fHla.-dy, where c = 2d modm. The m choices for d (and hence for c)
give the factor of m in m2kk\.

Corresponding to Corollary 4.3 we have:

COROLLARY 5.2. There are exactly m2kk\ m-tone rows IT for which •(!!) has
fewer than Am distinct members, and in such cases the cardinality of •(!!) is 2m.

THEOREM 5.3. The set-complex n\(^m) has cardinality

(m\ + m2kk\)4m = (m-l)!/

PROOF. Use Burnside's Lemma as in the proof of Theorem 4.4.

THEOREM 5.4. The cardinality of {\J(n):Ue^>m and \ D(n)|<4w} = {D(n):

II e0>m and\ D(n)| = 2m} is 2k~1kl.

PROOF. Similar to that of Theorem 4.5.

6. Examples

In this section we illustrate the results of the theorems in Sections 4 and 5.

Example 1. There are exactly 6 three-tone rows II for which II = f$Hcx~27">ya.a
for some a. They are (using the method of proof of Theorem 5.1):

(1,0,2), (2,0,1), (2,1,0), (0,1,2), (0,2,1), (1,2,0).

These are all the three-tone rows. For all 6 three-tone rows it is true (compare
with Corollary 5.2) that • ( ! ! ) has 6, not 12, distinct members. According to
Theorem 5.3, there is (3-l)!/4+2 1 - 2 l! = 1 set-complex in Hi&Q, and Theorem
278 J. A. Fill and A. J. Izenman [11]

5.4 again assures us that it has only 6 members. Of course, the only set-complex
is the above set of 6 three-tone rows. It may be written in the following form:

0 l 2

Example 2. There are exactly 8 four-tone rows fl for which II = /SO a", and in
all these cases a = 2. They are (using the method of proof of Theorem 4.1):

ni = (0,1,3,2), II, = (0,3,1,2), n 3 = (2,l,3,0), n 4 = (2,3,l,0), j

n 5 = (1,0,2,3), n 6 = (3,0,2,1), n , = (1,2,0,3), n 8 = (3,2,0,1). j

There are exactly 16 four-tone rows II for which II = jSIIa-^ya" for some a.
They are (using the method of proof of Theorem 4.2):

= (0,2,3,1), n i o = (O,3,2,l), n u = (1,2,3,0), n i 2 = (1,3,2,0),

= (2,0,l,3), n i 4 = (3,0,l,2), n i 5 = (2,l,0,3), n i 6 = (3,l,0,2),
(2)
= (1,3,0,2), II18 = (1,0,3,2), n i 9 = (2,3,0,l), n 2 0 = (2,0,3,1),
= (3,l,2,0), n 2 2 = (0,1,2,3), I I . ^ (3,2,1,0), nu = (0,2,1,3).

Notice that the set (1) of four-tone rows is disjoint from the set (2). For all 24
four-tone rows II, it is true (compare with Corollary 4.3) that D(n) has 8, not 16,
distinct members. According to Theorem 4.4, there are (4! + 22(2+l)!)/(4-4) = 3
distinct set-complexes in D ^ ) , and Theorem 4.5 again assures us that each of
the three has 8, rather than 16, distinct members. The three set-complexes are those
sequences in (1) plus

{n 9 , n 17 , n 13 , n 21 , n ^ , n 12 , n ^ , n16} and {n10, n 18 , n 15 , n ^ , n 22 , n u , n 19) n 14 }.

These three set-complexes can be written in the following form:

0 1 2 3 0 1 2 3 0 1 2 3

0 1 3 2 0 2 3 1 0 3 2 1
3 0 2 1 2 0 1 3 1 0 3 2
1 2 0 3 1 3 0 2 2 1 0 3
2 3 1 0 3 1 2 0 3 2 1 0
[12] Schoenberg's tone row system 279

Example 3. There are exactly 121 = 479,001,600 twelve-tone rows. Of these

(1) 26-6! = 46,080 are transpositions of their own retrogrades (in each case,
transposition by 6); (2) 6-26-6! = 276,480 are transpositions of their own
retrograde-inversions (for each c = 1, 3, 5, 7, 9, and 11, there are 46,080 twelve-
tone rows which are transpositions of their own retrograde-inversion by c).
According to Corollary 4.3 there are, therefore, 322,560 twelve-tone rows II for
which •(!!) has 24, rather than 48, distinct members. The number of distinct
•(II) is, by Theorem 4.4, equal to (12! + 2 6 (6+1)0/(4-12) = 9,985,920. By
Theorem 4.5, 2 6 - 1 (6+l)i/12 = 13,440 of these have 24, rather than 48, distinct
members. Note that

(13,440 x 24)+ (9,972,480 x 48) = 479,001,600.

Table 2 sets out, for each m between 3 and 15, the appropriate number of m!
m-tone rows which are transpositions either of their own retrogrades or of their
own retrograde-inversions. Also included in Table 2 is the number of distinct
set-complexes, divided according to whether the set-complex contains 2m or Am
distinct w-tone rows. It is worth noting the following identity:

ml = ((*{D(II): | 0(11)1 = 2m}) x (2m)) + (( f {•(!!): | 0(11)1 = 4m}) x (4m)),

a special case of which is illustrated at the end of Example 3.

7. Results for large m

For m>15, the values that would appear in Table 2 are too large to print
exactly or even to comprehend easily. However, using Stirling's formula for ml
we can make several simple approximations for the entries with large values of m.
The most interesting results are the following. The proofs are omitted.
(1) The square of the number of m-tone rows which are transpositions of their
own retrogrades behaves asymptotically like the total number of m-tone
rows times {-until)* for m even and is exactly 0 for m odd.
(2) The square of the number of m-tone rows which are transpositions of their
own retrograde-inversions behaves asymptotically like the total number of
m-tone rows times (7rm3/2)i/m, where fm is m/4 if m is even and is 1 if m
is odd.
(3) The number of distinct set-complexes behaves asymptotically like (m—1)!/4
for all m.
(4) The square of the number of set-complexes with 2m distinct m-tone rows
behaves asymptotically like the number of distinct set-complexes times
(7rm/2)*/m, where fm is defined in (2).
 to
8

TABLE 2
The number of invariant m-tone rows and distinct set-complexes for each m
*
S *n *n 277 *nai) •nai)
m m-tone rows n = jSna- II = j8na~ ° ytx" 1 aw l I D (II) | = 2m | D (II) | = 4m

3 6 0 6 I 1 0 31
4 24 8 16 3 3 0
5 120 0 40 8 4 4 3
D.
6 720 48 144 38 16 22
7 5,040 0 336 192 24 168
8 40,320 384 1,536 1,320 120 1,200
9 362,880 0 3,456 10,176 192 9,984 s
10 3,628,800 3,840 19,200 91,296 1,152 90,144
11 39,916,800 0 42,240 908,160 1,920 906,240 i
3
12 479,001,600 46,080 276,480 9,985,920 13,440 9,972,480
13 6,227,020,800 0 599,040 119,761,920 23,040 , 119,738,880
14 87,178,291,200 645,120 4,515,840 1,556,847,360 184,320 1,556,663,040
15 1,307,674,368,000 0 9,676,800 21,794,734,080 322,560 21,794,411,520
[14] Schoenberg's tone row system 281

8. Additional remarks

It is important to note that the results of this paper apply not only to m-tone
rows constructed from a single w-tonal scale (that is, from a single octave), but
also to compositions in which each note of an m-tone row can be stated in any
octave that the composer desires; indeed, this is part of the Schoenberg system.
The extension is straightforward: the composer states the number of octaves that
he is willing to consider in his choice of notes, say five; since there are m\ distinct
w-tone rows that can be obtained from a single octave, there must be 5mm\
w-tone rows that can be obtained from five octaves. Similar modifications apply
to each of the results of Sections 4 and 5.
In a further paper (see [7]) the authors present an efficient method for generating
all twelve-tone rows which are transpositions of their own retrograde-inversions.
This is an alternative procedure to that outlined in the proof of Theorem 4.2.
The method also applies to deriving all tone rows of even length which are trans-
positions of their own retrograde-inversions.
A related subject to that presented here concerns the medieval art of change-
ringing on church bells. We refer the interested reader to reference [8].

9. Acknowledgements

We should like to thank Professor Easley Blackwood of the Department of

Music at the University of Chicago for introducing us to this subject and for his
constant encouragement. Comments by Professor Irving Kaplansky on a previous
draft of this paper proved invaluable, and discussions with Allen Moy, Ronald
Thisted and Sandy Zabell were very helpful. We should also like to thank the
Arnold Schoenberg Institute for providing us with reference [10].
Work on this paper was carried out while Alan J. Izenman was a Visiting
Assistant Professor in the Department of Statistics, University of Chicago,
Chicago, Illinois. Support for this research was provided in part by National
Science Foundation Grants No. SOC72-05228 A04 and No. MCS72-04364 A04.
James Fill is a National Science Foundation Fellow and Robert R. McCormick
Fellow.

References

[1] M. Babbitt, "The function of set structure in the twelve-tone system", Unpublished
manuscript, Department of Music, Princeton University (1946).
[2] M. Babbitt, "Some aspects of twelve-tone composition", The Score and I.M.A. Magazine
12 (1955), 53-61.
282 J. A. Fill and A. J. Izenman [15]

[3] M. Babbitt, "Twelve-tone invariants as compositional determinants", The Musical

Quarterly 46 (1960), 246-259.
[4] M. Babbitt, "Set structure as a compositional determinant", Journal of Music Theory
5 (1961), 72-92.
[5] M. Babbitt, "Since Schoenberg", -Perspectives of New Music (Fall-Winter 1973, Spring-
Summer 1974, Special Double Issue), 3-28.
[6] J. Denes and A. D. Keedwell, Latin squares and their application (Academic Press, New
York, 1974).
[7] J. A. Fill and A. J. Izenman, "The structure of Rl-invariant twelve-tone rows",/. Austral.
Math. Soc. Ser. B (to appear).
[8] T. L. Fletcher, "Campanological groups", American Mathematical Monthly (1956),
619-626.
[9] C. L. Liu, Introduction to combinatorial mathematics (McGraw-Hill, New York, 1968).
[10] J. Maegaard, "Schonbergs Zwolftonreihen", Die Musikforschung 29 (1976), 385^*25.
[11] G. Perle, Serial composition and atonality (University of California Press, 2nd edition,
1968).
[12] A. Schoenberg, Style and idea (Philosophical Library, New York, 1950).

Department of Statistics
University of Chicago
5734 South University Avenue
Chicago
Illinois 60637
U.S.A.
and
Department of Applied Statistics
School of Statistics
University of Minnesota
Saint Paul
Minnesota 55108
U.S.A.

View publication stats