Professional Documents
Culture Documents
Transformations
Author(s): Dan Tudor Vuza
Source: Perspectives of New Music, Vol. 26, No. 1 (Winter, 1988), pp. 258-287
Published by: Perspectives of New Music
Stable URL: http://www.jstor.org/stable/833330
Accessed: 10/08/2010 14:56
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=pnm.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
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 support@jstor.org.
Perspectives of New Music is collaborating with JSTOR to digitize, preserve and extend access to Perspectives
of New Music.
http://www.jstor.org
SOMEMATHEMATICAL
ASPECTS
OFDAVIDLEWIN'SBOOK
GENERA MUSICALIN IER
LTZED VALS
AND TRANSFORMATIONS
DAN TUDORVUZA
IHAVE STUDIED with great interest Professor David Lewin's book as,
beside my basic activities in functional analysis, I am working myself in
the domain of applicationsof algebraicstructuresto musical theory. It is a
great pleasure for me to review this prestigious book, from which I have
extracted a lot of valuablesuggestions.
The present review article, primarily concerned with the mathematical
aspects involved in Lewin's book (1987), is organized in four parts. Part1 is
a presentation of the content of the book viewed from a mathematical
angle, with emphasis on the author's algebraicconstructions derived from
his musical theories and analyses. In Part 2 I make some comments upon
the above constructions; in particularI establish the relationship between
them and some standard mathematicalobjects. In Part 3 I examine some
connections between Lewin 1987 and my work on Anatol Vieru's modal
GeneralizedMusicalIntervalsand Transformations 259
Lewin uses the word "transformation" for a map of a set into itself and
refers to bijective transformationsas "operations." I shall make use of his
terminology.
1.2. Readers of Lewin 1987 must master elementary set theory and some
algebraicgroup-theoretic concepts, such as groups and homomorphisms;
all the needed material is included in chapter 1, making the book the-
oretically self-contained. Nevertheless, I consider that the book would be
better understood by readers already possessing a certain amount of
algebraictraining.
1.4. Chapter 3 presents the formal study of GIS's and includes three dis-
tinct topics.
The problem of labelling the members of S with the aid of the members
of IVLS is first undertaken. Namely, given a referentialmember refof S, the
bijection LABEL of S onto IVLS is defined by
LABEL(s) = int(ref,s).
1.6. The last part of chapter 3 is devoted to the study of three basictypes of
operations on a GIS: transpositions, interval-preserving operations, and
inversion operations.
Given i in IVLS, the transpositionTi generated by i is defined by the
equation
int(s,T,(s)) = i, s S.
262 of NewMusic
Perspectives
int(v,jZ(s)) = int(s,u) , s E S.
int(X(s),X(t)) = int(s,t)
for all s, t in S.
There are a number of interesting theorems about these operations.
Thus, the group of transpositionsis canonicallyanti-isomorphicwith IVLS.
The group of interval-preservingoperations is (not canonically) isomorphic
with IVLS: in fact, to every LABEL-function there corresponds the iso-
morphism k -> Pk between them, where
1.7. The largest part of chapter 4 introduces the first example of a non-
commutative GIS. It is a very interesting rhythmic model. Earlier, when
illustratingthe notion of a direct product of a GIS, the author considered a
commutative rhythmic GIS which assumes the existence of an absolute
time-unit. This model deals with time spans, that is, ordered pairs (a,x)
with a E R and x E R ; it models an event which is separated a times
GeneralizedMusicalIntervalsand Transformations 263
the time-unit from a referential zero time-point and lasts for x times the
unit. Here the int function is given by
(IVLS is the direct product between the additive group R and the multi-
plicative group R+ ).
One sees that though the interval is invariant under changing the zero
time-point, it is not under changing the time-unit. Examination of some
musical situations projects the idea that the notions of an absolute concep-
tual time-unit and of an absolute zero time-point are fraught with meth-
odological problems. A solution could be to replace these absolute con-
cepts by contextual ones, "restricting our attention to music in which we
can identify and assert a referential time-unit and a referential zero time-
point contextually" (Lewin 1987, 66). In such a situation one could use
the above commutative GIS. "On the other hand, there are also pieces and
passagesin which we cannot identify such referentialentities contextually."
Consequently, one would like to replace the above int function by a func-
tion invariantunder changing the referentialentities. This is preciselywhat
Lewin does when constructing his non-commutative rhythmic GIS. In this
GIS, the space S is the set TMSPS of all time spans, IVLS is the set { (i,p)
i E R, p E R+ } with the composition
(i,p)(j,q) =(i+pj,pq),
Thus, though time spans rely numericallyon the referentialtime span (0,1),
the invarianceproperties of int ensure that no matter what the choice of a
time span as referential,the interval between two given time spans would
not change; or, as Lewin put it, "it does not matter to what percept we
attach the numericaltime-span label (0,1)." The definition of the int func-
tion reflects the idea that "any time span has the potential for becoming a
local contextual time-unit, setting a local tempo" (Lewin 1987, 67).
Indeed, int(s,t) = (i,p) means that the time span t begins i times the dura-
tion of s afters begins and lasts forp times the duration ofs.
Two mathematical results about this GIS are of importance. The first
asserts that the group of interval-preservingoperations is formed here pre-
ciselybythe operationsdescribingthe changeof referentials.The second is an iso-
morphism theorem, showing that there is an essentiallyunique GIS having
the time spans as space and sharing the same invariance properties as the
one constructed above.
264 of NewMusic
Perspectives
1.8. Chapter 5 is dominated by two major ideas. The first is the introduc-
tion of the X/Y interval function, a far-reachinggeneralization of Forte's
intervalvector. Given a GIS (S, IVLS, int), two finite subsets X,Tin S and i
in IVLS, IFUNC(X,Y) (i) is defined as the number of pairs (s,t) in Xx Y
such that int(s,t) = i. After some musical analysis certifying IFUNC as a
useful tool, a list of theorems on its formal propertiesis displayed. The the-
orems are especially concerned with the way in which IFUNC(X, Y)
changes when one applies to X and/or Y transpositions and interval-pre-
serving operations. As concerns inversion operations, Lewin mentions the
formula
1.11. Chapter 6 is devoted to the theory and the applicationsof the injec-
tion function. Given a set S, a transformationf on S, and two finite subsets
X, Y of S, INJ(X,Y)( f) is defined as the number of elements s in X such
that f(s) E Y. It is shown that INJ is a generalization of IFUNC: indeed,
when S is the space of a GIS, then IFUNC(X,Y)(i) = INJ(X,Y)(Ti). The
injection function allows one to analyze the behavior of general musi-
cal transformations. Thus, transformationsf for which INJ(X,X)( f) is
high are called "X-internal": "Intuitively, an X-internal transformation
tends to extend/elaborate/develop/prolong X in music, while a pro-
gressive transformation tends to urge X onwards, to become some-
thing else (like Y)" (Lewin 1987, 142). A "progressive" transformation
is one for which INJ(X,Y)( f) is high. Similarly, a transformation f
for which INJ(X,X) (f) is small is called "X-external," and one for
which INJ(X,Y) ( f) is small is called "dispersive." In this way, Lewin's
266 of NewMusic
Perspectives
The discourse opens with the idea that from a strict mathematicalview-
point, the theory of GIS's could be entirely subordinated to transforma-
tional theory as being a "special branch" of it, "namely that branch in
which we study a space S and a simply transitivegroup STRANS of opera-
tions on S" (Lewin 1987, 158) (simply transitivemeans that for every s,t in
S there is a unique operation OP in STRANS such that OP(s) = t); we shall
comment upon this idea in section 2.14. Thus, continues the author, one
could adopt a more formal attitude, by studying first the general theory of
groups of operations and then presenting GIS theory as a part of the for-
mer. However, he considers that from a methodological viewpoint, the
order of presentationadopted in his book is more convenient: "By starting
with a study of GIS structure, we have built a link between the historically
central concept of 'interval' and our present transformationalmachinery.
To some extent for some cultural-historicalreasons, it is easierfor us to hear
'intervals' between individual objects than to hear transpositionalrelations
between them; we are more used to conceiving transformationsas affecting
Gestalts built up from individual objects." The "intervallic" attitude cor-
responds to our position of observers outside music, while a "transforma-
tional" attitude would correspond to someone inside music (Lewin 1987,
158-59).
1.16. The book closes with two appendices. I shall remark only upon
Appendix B, the one concerned with group-theoretic ideas. Here one can
find a second example of a non-commutative GIS. It is constructed by
observing that a non-commutative group made up from some transposi-
tions and inversions acts simply transitivelyon the octatonic scale {C, CO,
DO, E, F#, G, A, At}. This example is chosen in order to illustrate an
abstract algebraic principle: given a simply transitive group STRANS of
operations on a set S, the group STRANS' of all operations commuting
with all members of STRANS is also simply transitive, and STRANS is pre-
cisely the group of all operations commuting with all members of
STRANS'.
Generalized
MusicalIntervals
and Transformations 269
1.17. Before ending this presentation, let me remark that Lewin suggests
ways in which mathematicsat a higher level than the ones used throughout
the book could be used for further study. Thus, on page 103 one sees how
problems about the intervallicfunction can generate problems about con-
volution on locally compact groups; on pages 152-56 one sees how meas-
ure theory, in particularHaar measure, could be used in order to extend
the constructions involving the injection function INJ(X,Y) to the case
when X and Y are infinite subsets.
2.1. I shall start with a list of definitions, which are by now classicaland well
known to every mathematician.
When the map act is clear from the context, one writesgs (respectively
sg) instead ofact(g,s) whenever act is a left (respectivelyright) action. One
also says that G operatesat left/right on S or that S is a left/right G-set.
One can pass from left to right actions and conversely by the following
procedure: if G is any group, let G? be the opposite group of G, that is, the
set G endowed with the composition a given by gah = hg (in the right side
one has the initial composition on G). Then any left action of G on S is a
right action of G? on S and conversely. By virtue of this principle, I shall
continue the list of definitions considering one type of action only.
Now consider two groups G1, G2 and suppose that Gi operates at right
on Si (i = 1,2). A map f from S] into S2 is said to be compatiblewith a group
homomorphism ( from G1 into G2 if =
f(sg) f(s)(p(g) for every s in S] andg
in G1.In particular,a G-isomorphism between two G-sets is a bijection com-
patible with the identity map on G.
A special situation occurs when G1and G2 operate on opposite sides.
Consider for instance the case when G1operates at right and G2 at left (for
the opposite case the definition is similar). The map f is said to be compati-
blewith the homomorphism p if f(sg) = (p(g-l) f(s) for every s in S, andg
in G1.The link between this situation and the preceding one is furnished by
the fact that, when regardingS2 as a right G2-set, f is compatible (in the
sense of the preceding definition) with the homomorphismg - (<(g-')
from G1into G2.
There are two more group-theoretic notions which will be used below.
An automorphism of the group G is an isomorphism of G onto itself; the set
of all automorphisms of G is a group under compositions of maps and will
be denoted by Aut(G). For everyg in G, the map ag given by og(h) = ghg-'
is an element of Aut(G), called the internalautomorphismdefined byg; the
set of all internal automorphisms of G is a subgroup of Aut(G) and will be
denoted by Int(G). The map takingg into a. is a homomorphism of G onto
Int(G), denoted by intaut.
All definitions above can be found in Bourbaki1971, chapter 1, ?5 with
the only exception of the locution "f is compatiblewith (p" which is taken
from Bourbaki1965 chapter 3, ?2; the same thing is expressed in Bourbaki
1971as "f is a (p-morphism," but for my present purposes, I preferthe for-
mer locution as more convenient.
2.4. Every group G operates on itself at right via the map act defined by
act(g,s) = sg for everyg,s in G. The G-set so obtained is PH and Bourbaki
denotes it by Gd.IfS is a PH right G-set, the G-isomorphismsfrom Gdonto
S are preciselythe orbital maps; if (S, IVLS, int) is a GIS and if we consider
(as in 2.2) S as a right /VLS-set, then the IVLS-isomorphismsfrom S onto
IVLSdare precisely the LABEL-functions, that is, the inverses of orbital
maps.
2.6. Let two GIS's (Si, IVLSi, inti) (i= 1,2) be given and let us consider Si
as a right IVLS/-set. Then a map f from S1onto S2 is compatible with a
homomorphism (pfrom IVLS1into IVLS2if and only if
int2(f(s),f(t)) = (c(intl(s,t))
(except for the case of affine spaces, where one uses the word "affine map"
to design a map compatible with a linear map between vector spaces); I
shall refer these maps asprincipalhomomorphisms.
The notion of a principal homomorphism provides a reasonableway to
work with transformationson a GIS which are not in the group PETINV
(Lewin 1987, 57). For instance if we work within S12and if we would like
to introduce circle-of-fifths transformationswithout resorting (as in Rahn
1980) to a particularLABEL-function, we should define them as principal
homomorphisms M for which M is multiplication of intervals by 7. The
readerof Lewin 1987 could note the use, on the staves of pages 166-67, of
several principal homomorphisms M such that M is multiplication of
intervalsby 2.
In sections 2.7-2.9 below we shall fix a GIS (S, IVLS, int).
kx(P) = int(x,P(x)).
The formula
kx = (1)
inyt(xy)Y
shows that Int(IVLS) operates at left on the set K = {kx I xES} and that
the map x I- kx (which is not one-to-one in general!) from the right
IVLS-set S into the left Int(IVLS)-set K is compatible with the homo-
morphism intaut. In fact, K is a PH Int(IVLS)-set. When IVLS is com-
mutative, Int(IVLS) reduces to the identity map and K reduces to a single-
element set, the element being the canonical isomorphism can between IP
and IVLS.
GeneralizedMusicalIntervals
and Transformations 273
T, t = I where x = T(u)
TX = Tn V= I= I.
3) The formula in theorem 3.5.8 can be written in a label-free manner:
4 w= kv- (int(x,v))Tintw,u,.
274 of NewMusic
Perspectives
for some k in K. When IVLS is commutative, all k equal can and the com-
position between int' and can equals int; hence, (2) reduces in this case to
formula (1) from 1.6.
2.10. Consider now two GIS's (Si, IVLSj, intj) (1=1,2) and let f be a one-
to-one principal homomorphism from SI into S2. Then denoting by
IFUNCj the intervallicfunction (Lewin 1987, 88) attached to the j-th GIS,
one has
(the prime stands for the intervallicfunction computed in (S, IPo, int')).
This relation is more symmetric than (2) in section 1.8, as an intervallic
function occurs in both sides, and more "self-contained," as it relies nei-
ther on any particularLABEL-function, nor on any particularrepresenta-
tion of Ias I. Of course, when one writes down an explicit form for I, one
finds again Lewin's formula (2) from 1.8.
2.11. Earlier, we have seen that Lewin's procedures for constructing new
GIS's from old are concerned only with quotients and direct products. I
think that the notion of a subsystem-the dual notion for a quotient sys-
tem-is of equal importance in the theory of GIS's. It could be defined as
follows:
For instance, if (S12, Z12, int) is the classical twelve-tone system, then
GeneralizedMusicalIntervals
and Transformations 275
2.12. There are other kinds of G-sets, beside PH G-sets, which occur in
Lewin's book. Some of them are connected with the canonicalequivalence
discussed in chapter 5. Given a group CANON of operations on a set S, a
left action of CANON on the collection of all finite subsets of S is defined
by act(A,X) = A(X). Canonical equivalence is precisely conjugation with
respect to the above action; the set IXI (definition 5.2.2 in Lewin 1987) is
precisely the orbit of X.
The notion of a stabilitysubgroup, though not explicitly stated in Lewin
1987, is implicitly assumed by the formula
EMB(X Y) = RGNPF(X,Y)(card X)
RGNPF (X,X) (cardX)
given (without proof) on p. 152 and also, by the organization of the set of
tritones as the space of a GIS (p. 203), the latter representing in fact the
expression, in this particularcase, of the generalalgebraicprincipleasserting
276 of NewMusic
Perspectives
2.13. Let a GIS (S, IVLS, int) be given. Consider a subset P of IVLS satisfy-
ing the following requirements:
Such a set always can be extracted from a finite group. For instance, in
Z12one usually works with the set {1,2,3,4,5,6}. One can do that also for
infinite groups; in general one must invoke the axiom of choice, but in
some concrete cases one can write down P without resortingto that axiom.
By a dyadI mean a subset of S consisting of two distinct elements; by a
directeddyadI mean an ordered pair (s,t) such that int(s,t) EP.
The group of interval-preservingoperations operates at left on the col-
lection C2 of all dyads, breakingit into orbits. The extension to an arbitrary
GIS of Lewin's theorem on dyad classification in his non-commutative
rhythmic GIS sounds: Let (sl,tl) and (s2,t2) be two directed dyads. Then
the (undirected) dyads {s1,tj} and {s2,t2} are conjugated if and only if
int(sl,t) = int(s2,t2). Consequently, there is a one-to-one correspondence
between P and the orbits of C2.
Now suppose that IVLS satisfiesthe condition
and that we can choose P so that, beside (P1) and (P2), it also verifies
This cannot be done for every group satisfying(*). When one can do it,
the relation < on S defined by
s< t if int(s,t) EP
uniquely written as an ordered n-tuple (s,... ,s,) so that s<s2 < ... <s,. It
turns out that two such n-tuples (sl,... ,s) and (t1,... ,tn) are conjugated if
and only if int (s,sj+ 1) = int(tj,t.+ 1) forj = 1,... ,n -1. Consequently, there
is a one-to-one correspondence between P"-1 (the Cartesian product of
n - 1 copies ofP) and the orbits of C,.
For instance, if (TMSPS, IVLS, int) is Lewin's non-commutative rhyth-
mic GIS, one can make at least two choices Pa, Pr for P:
int2(s,t) = p(intl(s,t))
for every s,t in S. The reader of Lewin 1987 has already encountered this
kind of equivalencein a particularcontext, namely in the statement of the-
orem 4.1.5 about the essential uniqueness of Lewin's non-commutative
rhythmic GIS. One can prove that two GIS's are equivalent if and only if
their groups of transpositions are equal as sets of operations on S; hence,
278 of NewMusic
Perspectives
3.2. Let me examine some formal connections between Lewin's work and
mine.
As in Lewin 1987, the basis for my study is provided by a PH G-set A; I
make the additional assumption that G is commutative.
In order to establish convenient bijections between G and A, I use
orbital maps; these are the inverses of Lewin's LABEL-functions.
Lewin attaches to a GIS transpositions, interval-preservingoperations,
and inversion operations as basic types of transformationson that GIS; as a
rule, he does not consider maps acting between differentGIS's. In my the-
ory, I consider principal homomorphisms as a basic class of maps. Such
maps may act between different PH sets; transpositions and inversions
belong to this class.
A main topic in my theory is the study of mathematicalstructureson the
set M(S). The latter set is constructed in Vuza 1982a as follows: given a
group G and a left action of G on a set S, define a left action of G on the set
P(S) of all subsets of S by act(g,X) = { gx Ix EX} for everyg in G and X in
P(S). By definition, M(S) is the collection of all orbits determined by the
action of G on P(S). In order to compare this construction with Lewin's
classesof canonical equivalence, let me remarkthat the latter are the orbits
determined by the restriction of the action of G on P(S) to the subset of
P(S) consisting of all finite subsets of S (in Lewin's theory, G is a group
CANON of operations on S). It is true that my theory is especially con-
cerned with the sets M(S) in the case when S is a PH G-set and G is com-
mutative, while in Lewin's theory, CANON is in general neither com-
mutative, nor simply transitive on S. On the other side, while Lewin
considersonly the canonicalequivalenceof finite sets, my model of periodic
rhythm makes extensive use of a collection of elements in M(S) such that
each of them is an orbit consisting of infinite subsets of S.
When A is a PH G-set, I shall refer the elements of P(A) as "modes"
and the elements of M(A) as "modal classes."
GeneralizedMusicalIntervalsand Transformations 281
A periodicrhythmis a (possibly
empty)subsetR of Q satisfying(RI) and (R2)
below:
4.3. Looking first at the differences, one sees that while Lewin's model
applies to rhythm in general, my model applies to a special class of periodic
rhythms. Time axis is R in Lewin's model, Q in mine; in other words, I
consider a temporal space in which events are separatedby commensurable
durationsonly. This has at least two formal advantages:First, the collection
of all periodic rhythms is closed under taking unions, intersections, and dif-
ferences (that is, it is what in mathematicsis called a ring of sets). Second,
one can compose rhythmic classes in the same way as it is done for modal
classes. Composition of rhythmic classes has a natural musical interpreta-
tion and is very useful when studying the construction of canons.
does not depend upon the particulartime-unit used when we have written
down R and R'. Hence, Div(R) plays here the role of a contextual time-
unit, the admissiblemoments we are looking for being separatedby integral
multiples of that unit from the startingmoment. A more formalway to put
this is to say that the collection of all periodic rhythms is stable under the
action of the group formed by all operations t-->at+ b (a EQ+, b EQ)-
the Q-analog of the group of interval-preserving maps used in Lewin's
model-and the set (2) is an invariantfor that group.
The readershould note how the numbers Div(R) are used in Vuza 1985
to model polyrhythm.
I would like to conclude this review with the idea that the publication of
Lewin's book representsan event markingin a decisive way the process of
the introduction of mathematical structures and reasoning into music-
theoretic activities. Consequently, I consider that musicians as well as
mathematicianswill be interested in studying it.
Musicianswill discover through it the role played by the modelling func-
tion of mathematicsin better understanding the structuralunity of a phe-
nomenon which is manifest through a variety of forms. Indeed, it is not a
fortuitous but a rather deep fact that the same formal tools can describe
aspects from different domains (such as melodic, rhythmic, timbral) of the
musical phenomenon. And, last but not least, the book responds to a
necessity which will become (at least I think so) quite pressing in the near
future: to accustom musicians to the rigors of mathematicalreasoning, by
offering them real mathematicaldefinitions, theorems, and proofs.
It is quite possible that the professional mathematician would find the
mathematicaltheory in the book ratherelementary. It is my conviction that
the author consciously confined his exposition to a medium mathematical
level, taking into account the audience he addresses. Nevertheless, the
mathematician's attention will be captivated by the way in which the
author gives new and, sometimes, unexpected interpretations to classical
mathematicalideas when applied to musical contexts. These new interpre-
tations lead in turn to new mathematicalproblems; and this is the aspect
which should especially concern the mathematician, most of these prob-
lems requiring, as the author himself points out, mathematicalmethods of
a higher level. Let me recall to the reader that Lewin-whose academic
training is both musical and mathematical-employed as early as 1959
sophisticated devices such as convolution and Fourier transformin order to
solve problems inspired by musical considerations (see Lewin 1959).
In 1982 I was closing my conference on mathematicalmethods in music
delivered at a scientificsession of the Romanian Academy in the spirit of the
idea that the future will decide if music theory will integrate mathematical
reasoning as a response to an internal necessity. Professor David Lewin's
book pleads for a strongly affirmativeanswer.
286 of NewMusic
Perspectives
REFERENCES