Vous êtes sur la page 1sur 53

Perspectives of New Music

Some Compositional Uses of Projective Geometry


Author(s): David Lewin
Source: Perspectives of New Music, Vol. 42, No. 2 (Summer, 2004), pp. 12-63
Published by: Perspectives of New Music
Stable URL: http://www.jstor.org/stable/25164553
Accessed: 21-10-2015 00:44 UTC
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/25164553?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact 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

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Some Compositional
Uses
Geometry
Projective

of

Le?7ta
David Lewin

Part

I: Introduction

to the Projective

Plane; Diatonic

Modal

Examples

a school that has seven students: Ann, Bill, Carol, Dan, Eve,
and
Frank,
Gladys. Suppose the school offers seven courses: Latin,
Imagine
Math, Neurology, Psychology, Quantum mechanics, Russian, and Span
ish. Suppose the enrollment of students in courses is given by Example 1.
On the example, each asterisk indicates that the student whose initial
appears to the left is taking the course whose initial appears above. Thus
Ann is taking Math, Neurology, and Quantum mechanics;
the students
enrolled in Spanish are Carol, Frank, and Gladys; and so forth.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

13

ProjectiveGeometry

A
~B~
~C
~D
~E
~F~
~G

EXAMPLE 1
1 has the following properties:

Example

1. Any two students have just one course in common.


2. Any two courses have just one student in common.
3. Every course has at least three students; every student is taking at
least
4.

three

It is not

the

courses.
case

that

every

student

is taking

every

course.

The properties manifest a structure called a "projective plane." To


appreciate the geometric metaphor, we can replace the word "student"
by the word "point," the word "course" by the word "line," and the
notion of a student's taking a course by the notions of a point's "lying
on" a line and (equivalently) of the line's "passing through" the point.
The properties above then translate as follows:
1. Any two points lie on just one line.
2. Any two lines pass through just one point.
3. Every line passes through at least three points; every point lies on at
least three lines.
4. It is not the case that every point lies on every line.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

14

Perspectives

of New Music

the properties suggest things we apprehend about


"plane geometry." The things have to do in particular with incidence
relations of points and lines. Only property 2 departs from Euclidean
intuition: in Euclidean plane geometxy^parallel lines do not pass through
any common point. Projective geometry, unlike Euclidean geometry (or
As

translated,

"affine geometry") does not recognize the notion of parallel lines.


To turn the Euclidean plane into a projective plane one can adjoin a
"line at infinity," comprising formal "points" where parallel lines meet.
In the Euclidean plane one fixes a reference line; in a standard Cartesian
coordinate system this could be the horizontal X axis. For each number a

between 0 inclusive and 180 exclusive, stipulate a corresponding "point


at infinity."Each line in the Euclidean plane will be tilted with respect to
the reference line at some angle a degrees; the given linemeets the line
at-infinity at point a. So a line perpendicular to the reference line meets
the line-at-infinity at point 90. A line parallel to the reference linemeets
the line-at-infinity at point 0. More generally, two parallel lines will both
be tilted the same amount with respect to the reference line; if the com

mon

tilt is a degrees, then the two lines meet at infinite-point a. The


extended structure is called the "Real Projective Plane." The reader can
verify that it satisfies properties 1 through 4 above.1
A projective plane is highly structured, and that structure can be musi
cally suggestive in various contexts. As a first illustration, let us return to

Example

1 and

now

read

the

seven

as

"students"

pitch classes. The seven "courses" then become


leged Chords, the Chords L through S displayed
properties now read:

the

seven

white-note

a family of seven privi


in Example 2. The four

1. Any two notes belong to just one Chord.


2. Any two Chords have just one common

tone.

3. Every Chord has at least three notes; every note belongs to at least
three Chords.
4. It is not the case that every note belongs to every Chord.

EXAMPLE 2
This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

15

ProjectiveGeometry

EXAMPLE 3

Indeed in this particular geometry every Chord has exactly three notes,
and every note belongs to exactly three Chords. Chords L, M, and N,
taken in conjunction, suggest Phrygian modality; Chord P, when suitably
linearized, is appropriate for executing a melodic Phrygian cadence.

I constructed the example with these features in mind/ear.


Obviously
Later on, I shall show how I did that.
I shall present a Lehrst?ck that demonstrates how various
Meanwhile
the
of
geometry can be projected compositionally. After setting
aspects
seven-note
the
seven-chord projective plane as in Examples 1 and 2,
up
to build a species of "Phrygian mode,"
I decided to write a three-part
I
be a Chord.
vocal piece where every three-note verticality would
to base this piece on a cantus firmus, and had the idea of com

decided

posing

the

cantus

as

tune

twenty-one-note

comprising

concatenated

linearizations of the seven three-note Chords. Example 3 shows the can


tus, shaped to have a "Phrygian" character (and somewhat adjusted as
the composition went into later stages). As shown by the brackets and
letters

at

the

bottom

of

the

example,

the

first

three

notes

of

the

cantus

linearize Chord M, the next three linearize Chord L, and so on; the
melodic cadence is achieved through a linearized form of Chord P.
Imagining a vocal piece, I looked for a suitable text. I tried to find a
stylistically appropriate twenty-one-syllable text that could be sung to the
tune of the cantus firmus. Example 4, setting three seven-syllable verses
of text, seemed apt. (I modified my original cantus firmus slightly, the
better to accommodate this text.)

Ben - e - die -tus qui ven - it

in no -mi -ne Do -mi -ni,

in no -mi -ne Do

-mi -ni.

EXAMPLE 4

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

PerspectivesofNew Music

16

I wanted to harmonize each note of the cantus firmuswith each of the


three Chords to which it belongs. Example 5a sketches the idea. Wsym
bolizes the note of the cantus firmus; the lines on Example 5 a indicate
the three Chords

containing W, that is {W,XI, X2],

{W, Yl, Y2\, and

[W,Z1,Z2\.

a) b) c)
/W\
XI
XI

III

,WV

/Wl

Yl

Zl

XI-Yl-Zl

XI-Yl-W3

Yl

72

X2?Y2-22

X2-Y2-W3

III

W2

II

EXAMPLE 5

It is not hard to see that the seven "points" of Example 5a must be the
seven distinct notes of the mode.2 One naturally hopes to be able to
arrange matters as in Example 5b, so that {XI, Yl, Zl] and {X2, T2, Z2\
will be Chords (i.e., "lines" of the geometry), projected as vocal lines in
the two accompanying voices. But this is not possible. The line deter
mined by XI and Yl will indeed contain one of the Z-points, but that
same Z-point (not the other one) will also be the third point on theX2
72 line.3
This feature of the system suggests the compositional
algorithm
sketched by Example 5c. Here Wl symbolizes the present note of the
cantus firmus, and W2 symbolizes the next note of the cantus. (The can
tuswas constructed so as not to have repeated notes.) IfW3 is the third

containing Wl and W2 (the third point on the line


Wl
and
W2), then the X-notes and T-notes can be arranged
containing
so thatX1-Y1-W3
is a geometric line (i.e., a Chord), and X2-Y2-W3
is
also a geometric line (a Chord). Indeed, this arrangement is unique, up
to contrapuntal inversion of "voice 1" and "voice 2," and/or exchange
of notes Xn with notes Tw.4
The basic algorithm of Example 5 c produced the compositional study

note of the Chord

of Example 6.
Other features of the projective geometry suggest other compositional
possibilities. Consider for instance the fact that every pair of distinct notes
belongs to exactly one Chord, to which exactly one other note also
belongs. This enables one to add a third part automatically, to any two-part

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

17

Projective Geometry

nit,qui ve - - -nit,in

mi -

Do -

-ni_

-mi - ni,.

in_

no -

Do-mi-ni,_Do

-mi - - - ne inno

-mi -ni,

- mi -ne_

EXAMPLE 6

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

18 PerspectivesofNew Music

example

diatonic counterpoint that does not contain a vertical octave or unison.


Example 7 illustrates, adding a third part to an example fromZarlino.5
The example illustrates his "third mode";
given our particular Chord
a
course
to
it
is
natural
select
of
two-part model from one or
vocabulary,
another E-final mode. The exercise could be made more elegant if it did
not cling so doggedly to the notion that every three-note verticality had
to be a Chord of the geometric system; protocols for non-Chordal verti
calities in three voices could of course be worked out.
Another resource for composition in our geometric mode is "modula
tion" to secondary modes. This can be accomplished in several ways. First,
the Chords of Example 2 can all be transposed by a fixed number of semi

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

a)

19

Geometry

b)

L'

N'

P'

(M)

?r

R'

S'

M'

N'

P'

R'

S'

(Q)

(R)

(Q)

(P)
EXAMPLE 8

tones, introducing ficta as appropriate. An appropriate interval of transpo


sition is five semitones (Example 8a): that is because the two harmonic
triads of themode are related by T5; also the transposition introduces only

one ficta tone. The a-minor triad serves as pivot: Chord V of the trans
8b
posed mode coincides with Chord M of the original mode.Example
shows the original mode transposed by T2. This introduces ficta tones F|
and C|. Chord 5'of the transposed mode coincides with Chord Q of the
original mode. Example 8c shows a particularly idiomatic pitch-class inver
sion of the original mode, inversion-about-D. No ficta tones are intro
duced, and there aremany pivot-Chords: P'and il'coincide with R and P
respectively;Q 'coincides with Q. The mode of Example 8c could be taken
in itself as a sort of "Ionian/C-major"

structure.

one can make, from the


suggests another sort of modulation
one
can
to other diatonic
of
mode
2.
modulate
original
Example
Namely,
to
a projective
modes
other
choices
of
realize
modes,
Chords,
given by
seven
as
notes
can
with
the
white
One
plane
"points."
easily construct
That

large numbers of such modes. For that purpose, the abstract configura
tion of Example 9 is useful.
First one selects three basic Chords, which one might think of as "pri
mary." These are the Chords corresponding to lines 123, 345, and 561
on the figure. The primary Chords must be selected so that any two of
them have exactly one common tone. C-, G-, and F-major triads could
not be selected as the three Chords, because the G and the F triads have
no common tone. One could, however, select {C, E, G}, {F,A, C}, and
{G, F, B}. Arranging the notes to suit the configuration of Example 9,
one could take C = 1, G = 3, F = 5, E = 2, B = 4, A = 6. Then the remain
ing note, D, would be number 7 in the example.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

20

PerspectivesofNew Music

If 123, 345, and 561 are all formal "lines" of the geometry, then so are
174, 376, 572, and 246.
example

The choice of the "basic triangle" on Example 9 exhausts six notes of


the mode. The seventh note appears at the center of the triangle. Itmust
be the case that 174, 376, and 572 are all formal "lines" of the geometry;
the corresponding notes will then be formal Chords of the system.6
Finally, itmust be the case that 246 is a formal "line"; the corresponding
notes then form the seventh Chord.7

(a) (b) (c)

"Ionian/Cmaj";
primary Chords CEG,
FAC, GBF; other
Chords BCD, DGA,
DEF, EAB.

"Dorian"; primary
Chords DFA, GBD,
EGA; other Chords
CDE,
BEF.

ABC,

FGC,

"Mixolydian";
primary Chords GBD,
DFA, GAC; other
Chords CDE, EFG,
EAB, BCF.

EXAMPLE 10

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

ProjectiveGeometry

21

{C,
Example 10a shows themode generated by the "primary Chords"
E, G}, {F,A, C}, and {G, F, B}, as begim above. This turns out to be the
mode encountered earlier in Example 8c; itwas the
"Ionian/C-maj"
inversion of our original "Phrygian" mode about D. Example 10b shows
a "Dorian" mode generated by taking {D, F, A}, {G, B, D}, and {E, G,

A} as primary Chords. Example 10b inverts about D to yield Example


10c, a "Mixolydian" mode with primary Chords
{G, B, D},
{D, F, A},
serve as pivots between
and {G, A, C}. Chords DFA, GBD,
and CDE
our Dorian and Mixolydian modes; FGC pivots as a Chord for both our
Phrygian and our Dorian mode.
There is of course little reason, other than historicist sentimentality, to
assign special priority to harmonic triads as Chords, "primary" or other
wise, in constructing any formal diatonic mode (i.e., geometry).
are available. Some mathemati
Further techniques of "modulation"
ones
will
idiomatic
be
discussed
later, "collineations" in particular.
cally
* *

In any projective plane, there are many ways of choosing a structure


almost like that of Example 9, a structure consisting of seven points {1,2,
3, 4, 5, 6, 7} inwhich 123, 345, 561, 174, 376, and 572 are all formal
"lines" of the geometry. In general, 246 need not be a formal line.We
shall call any such structure a "reference quadrangle."
A generic reference quadrangle may be constructed as follows. Given
any projective plane (whether a seven-point plane or some other), select
"1" and "3" as any two distinct points. (Any line contains at least three
distinct points.) Since not all points of the plane lie on one line, select
"5" as any point not on the line 13.
The lines 13, 35, and 15 are all distinct. (If any two coincided, then
points 1,3, and 5 would all be collinear, and point 5 would lie on line
13, contrary to construction.) So, in particular, lines 13 and 15 are dis
tinct. Let L be a third line passing through point 1. (Every point lies on
at least three distinct lines.) Take "4" to be the point where line L meets
line 35. Point 4 is not the same as point 3, since Lis distinct from line 13.
Point 4 is not the same as point 5, since L is distinct from line 15.
Line L contains at least three distinct points, including points 1 and 4.
Take "7" to be a third point on line L. Then 7 cannot lie on any of the

other lines so far constructed (13, 15, or 35). If 7 were on line 13, then
line 17, which is L, would be the same as line 13?but L was taken to be
distinct from line 13. If 7 were on line 15, then line 17, which is L,
would be the same as line 15?but L was taken to be distinct from line

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

22

Perspectives

of New Music

Kl

[p2, p3, ql, ri}


J2
[pi, p3, q2, r2\
=
J3
[plyp2, q3, r3]
Jl

Ll

K2
K3

Ml
M2

[pi, ql, si, t)


L2 = [p2, q2, s2, t)
L3 = [p3, q3, s3, t]

M3

[pi, ri, s2, s3)


=
[p2, r2, si, s3}
=
[p3, r3, si, s2]
=
[q2, q3, ri, si)
=
[ql, q3, r2, s2\
=
[ql, q2, r3, s3)

N=[rl,r2,r3,t\
EXAMPLE 11

15. If 7 were on line 35, 7 would be the point where L meets line 35?
but that point is point 4, and 7 was chosen to be distinct from 4.
Take "2" to be the point where line 57 meets line 13. Point 2 is dis
tinct from point 3: if2 = 3, then line 25 = line 35, and point 7 would lie
on line 35?which
it doesn't. Point 2 is distinct from point 1: if 2 = 1,
=
it
then line 25
line 15, and point 7 would
lie on line 15?which
doesn't. So 1, 2, and 3 are different points of line 13.
In likemanner, taking "6" to be the point where line 37 meets line 15,
we see that 1, 6, and 5 are distinct points of line 15. In sum, points 1
through 7 are all distinct; 123, 345, 561, 275, 471, and 673 are all dis
tinct lines.
In the general projective plane, itneed not be the case that points 2, 4,
and 6 are collinear.

Part

II: Some Other

Projective

Planes, with Musical

Applications

as fol
A generic thirteen-point projective plane can be conceptualized
lows. Let the "points" of the system be symbolized as pi, p2, p3, ql, q2,
q3, rl, r2, r3, si, s2, s3, and t.Then take as the thirteen "lines" the sets of
points J1,J2, etc. given by Example 11.
The given thirteenpoints, togetherwith the stipulated thirteen lines, form
a projective plane. The readermay check the four required properties:
1. Any two points lie on just one line.
2. Any two lines pass through just one point.
3. Every line passes through at least three points; every point lies on at
least three lines.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

23

Geometry

4. It is not the case that every point lies on every line.


In the thirteen-point plane, every point lies on exactly four lines and
every line passes through exactly four points.
The thirteen-point plane can be used to structure "chromatic" modes,
if one adjoins a special symbol $ to the twelve chromatic pitch classes.
The special symbol could be realized musically as silence, or as a non
pitched event, or as a wild card. Example
such mode.

=
G,p3=?,
ql=D,
pl=B,p2
= $.
=A,?
?i=Et,?2=Ff,.tf
/i={G,F,D,C}

=
J2 {B,F, Bt, E}
=
J3 {B,G, Gf, Cf}

#2=Bt,

12 shows the structure of one

q3

K2

G$, ri=C,

r2=E,

r3-Cf,

tfi?{B,C,Ff,A}

=
{G,E, Et, A}
K3 - {F,Cf, Et, Ff}
Ml = {Bt,Gt, C, Et}
M2 - {D,Gf, E, Ff}
M3 = {D, Bt, C|, A}

LI - {B,D, Et, $}
L2 = {G,Bt, F|, $}
L3 = {F,Gf,A, $}
tf={C,E,Cf,$}
EXAMPLE 12

as follows. I
12 I proceeded
To construct the mode of Example
decided to make the special symbol $ the point t of Example 11. I then
considered the four lines of Example 11 that would contain the special
=
symbol t $. These four formal lines would be realized in the musical
mode by formal Chords thatwould each comprise three "genuine notes"
plus the special effect. I decided to structure themode so that those four
Chords would be transposed or inverted forms, each of any other. I spe
=
=
cifically decided that LI
[pi, ql, si, t)would be {B, D, Et, $}, that L2
=
[p3, q3, s3, t) would be
[p2, q2, s2, t) would be {Ff, G, Bt, $}, that L3
[rl, r2, r3, t) would be {C, Cf, E, $}. Thus
{F, Gf, A, $}, and thatN
the three notes B, D, and Et would be assigned in some order as the
points pi, ql, and si of Example 11; likewise the three notes Ff, G, and
Bt would be assigned in some order as the points p2, q2, and s2 of

Example 11, and so forth. Playing around with various possibilities for
those assignments, considering the other Chords that Example 11 would

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

24

Perspectives

of New Music

12. Evidently an

generate in each case, I arrived at the mode of Example


enormous number of such modes can be constructed.

rl

(t-1-1)-1-p3-s3v

r
/-s/-qJv
rl-p

pi-p

I-r3-(q3-q3-q3-q3)

I-P3-(ql?ql-ql-ql)-J--ql
s2-rl-qV

r3-ql-$3'

si

EXAMPLE 13

Example 13 demonstrates an algorithm for harmonizing certain types


of cantus firmus.The example supposes the succession t, q2, q3 as a seg
ment of such a cantus firmus.All vertically aligned sets of four events in
the example are Chords of the generic system. At the left, the cantus fir
mus event tis harmonized by the four vertical Chords that contain it: {t,

rl, r2, r3}, {t, si, pi, ql], {t, q3, p3, s3], and (moving onwards) {t,p2, q2,
s2). The voice leading can be arranged so that the three accompanying
voices of the first three chords all form formal lines with q2, the next
event of the cantus firmus. That is?as indicated by the horizontal and
si, q3, q2) is a formal line; so is \r2,
diagonal lines in Example 13?[rl,
so
and
is
[r3, ql, s3, q2\. q2 now takes over as cantus firmus
pi, p3, q2}y
and
it
is
first
harmonized
event,
by the (unique) chord that contains both
t.
The
and
then
process
repeats, aiming at q3, the next event of the
q2
cantus

firmus.

Evidently, in this arrangement, it is necessary for q3 not to be p2 or


s2?i.e., q3 must not lie on the line defined by t and q2. And of course t
and q2 must be distinct. The cantus is thus constrained as follows: itmust
contain no immediately repeated events, and no three consecutive events

may be collinear.
In Example 13, the three initial lines of "alto," "tenor," and "bass"
voices, under t in the cantus firmus, become verticalities when q2 takes
over the cantus role: the alto's initial line rl-sl-q3-q2
becomes the last
four-point verticality shown in the example; the tenor's initial line r2-pl
p3-q2 becomes the fifth four-note verticality, and so forth. This is a fea
ture of the algorithm.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

25

Projective Geometry

* * *

The thirteen-point plane can also be applied to other musical phenom


ena. For instance, there are exactly thirteen of what Larry Polansky calls
"ternary contours."8 These are illustrated in Example 14. The thirteen
point plane will organize these contours into mathematical "lines," each
of which is a group of four contours. Example 11 will help the reader
experiment with various modes of such grouping, assigning the symbols
ply p2, etc. in various ways to the thirteen symbols of Example 14.

EXAMPLE 14

* * *

is a twenty-one-point projective plane; every line therein has


exactly five points, and every point lies on exactly five lines. A generic
structure for the plane can be obtained by algebraic techniques to be dis
cussed later. The geometric structure could be applied, in various
"modes," to the family of twenty-one diatonic unordered dyads.
There is also a thirty-one-point projective plane; every line therein has
exactly six points, and every point lies on exactly six lines. The structure
might be of interest to those who are interested in the thirty-one-tone
scale; the geometry would single out a privileged family of thirty-one
hexachords as formal "lines" of the system. The thirty-one privileged
There

hexachords

Part

could be chosen in a vast variety of different "modes."

III: Algebraic

Structure

of the Projective

Plane

There is a common algebraic structure underlying all projective planes, a


structurewhich is very helpful for constructing a "generic" specimen of a

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

26

Perspectives

of New Music

given plane, along the lines of Example 11 earlier. To grasp such matters,
one must firstunderstand the notion of an algebraic "skew-field."
By that term, a mathematician understands a set F of objects x, y, etc.,
upon which two binary combinations are defined. The combinations are
usually denoted using conventional symbols for addition and multiplica
tion: thus objects x and ymay be combined so as to yield a formal "sum"
x + y, and a formal "product" xy. In order to constitute a "skew-field,"
the systemmust satisfy requirements 1, 2, and 3 following.
1. Pis

a commutative group under addition. The identity element


denoted by 0; the additive inverse of element a; is denoted "-#".

is

2. The non-zero

elements of F form a group under multiplication. The


element
of the group is denoted by 1; the multiplicative
identity
inverse of a non-zero #is denoted by "aT1" or by "1/V\

3. Multiplication distributes over addition, according to the laws x(y + z)


=
= xz
xy + xz\ (x +y)z
+yz.
Ifmultiplication
is commutative, the skew-field is called a "field." The
real numbers form a field. So do the rational numbers?those
which can
be expressed as quotients of integers. So do the numbers of form a +
is the (irrational)
b(Jl), where a and b are rational numbers and Jl
a
non-zero
a + b(Jl ) is (a
root
2.
of
The
inverse
of
square
multiplicative
cannot
denominator
vanish
because
a2 cannot
2bb)?the
-b(j2))/(aaa
not
and
b
rational
and
both
zero.9
possibly equal 2e2,
being
The integers mod 2 form a field. Those are the integers 0 and 1 with
all arithmetic reduced modulo 2?e.g.,
1 + 1 = 0. The integersmod 3 are
also a field, that is the symbols 0, 1, and 2 with arithmetic reduced mod
ulo 3: 2 + 1 = 0, 2 2 = 1 (4 is onemore than some multiple of 3), and so
forth. The integersmod 5 are a field, that is the symbols 0,1, 2, 3, and 4,
with all arithmetic reduced modulo 5. (4 3, for instance, = 2, since 12 is
twomore than some multiple of 5.) In general, for any prime number p,
the integersmod p form a field.
There are other finitefields as well. For instance, there is a field of four
elements which one can denote as 0,1, a, and a + 1. In this field, anything
added to itselfproduces 0, and a2 equals a + 1. These properties enable
one to carry out the necessary arithmetic, to verify that the structure is a
field. For instance, a(a + 1) = a2 + al = (a + I) + a = I + a + a = 1. Then (a
= 1 + 1 + a = a.
=
+ l)(a + 1) = a(a+
1) + l(a+
1)
(I) + (a + 1)
The cardinality of the field just studied, four, is not a prime number,
but it is a power of a prime. In general, for any prime number p and any
positive integer ?, there will be a field of cardinality p^. Furthermore, any

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

Geometry

27

finite field must be one of those. Indeed, any finite skew-field must be
one of those, since a remarkable theorem shows that any finite skew-field
must be a field.
* * *

Now we are ready to construct a generic projective plane. Let F be some


skew-field (any skew-field). Consider the family of all ordered triples <bl,
b2, b3> formed from members bl, b2, and b3 of F. We stipulate that two
such triples, say <bl, b2, b3> and <bl\ b2\ b3>, are "left-equivalent" if
there is some non-zero number k in F such that bl' = k bl, b2' = k b2,
=
and b3'
k b3. This in indeed an equivalence relation among the triples.
The relation is reflexive (bj- 1 bj,forj = 1, 2, 3); it is symmetric (if bj'=
=h
k bj, then bj = (l/k)
bjy, the relation is also transitive (if bj"
bj',

thenbj" = (hk) bj).

points of the F Projective Plane are labeled by the non-zero left


equivalence-classes of F-triples. That is, <0, 0, 0> is excluded as a "point"
of the geometry, but every other F-triple represents a point; among those
triples <bl, b2, b3> and <cl, c2, c3> represent the same point ifand only if
there is some non-zero ?in F such that cj = k- ?/for each y = 1,2,3.
We shall label the formal lines of the F Projective Plane by the non
zero right equivalence-classes
of F-triples, which we shall write using
The

to the triple
square brackets: the triple [XI, X2, X3] is r^i-equivalent
=
some
non-zero
if
there
is
k
in
_Fsuch
that
[XI ',XT, X_H
k, for
Xj'
Xj
each j = 1, 2, 3. Every non-zero triple represents a line; among those
triples [XI, X2, X3] and [Tl, T2, T3] represent the same line if and only
if there is some non-zero k in _Fsuch that Yj =Xj- k, for each j = 1,2,3.
The point b= <bl, b2, b3> lies on the lineX=
[XI, X2, X3] ifand only

ifthe sixnumberssatisfy
theequation (bl)(Xl) + (b2)(X2) + (b3)(X3) =

0 in the skew-field F. One sees that if b' is left-equivalent to b, and/or


right-equivalent toX, the numbers ?y'and Xj'wi?
satisfy the same equa
tion. Hence the relation "? lies on _?" iswell-defined for points and lines.
The known algebraic behavior of such equations in skew-fields guaran
tees that the "points" and "lines" of the "F Projective Plane," as defined
above, do indeed constitute a projective plane. That is:
1. Any two points lie on just one line.
2. Any two lines pass through just one point.
3. Every line passes through at least three points; every point lies on at
least three lines.
4. It is not the case that every point lies on every line.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

28

Perspectives

of New Music

The Property of Desargues


(discussed in note 1) also obtains. Remark
ably enough, every projective plane (i.e., every aggregation of formal
points and formal lines satisfying the four properties above and the Prop
can be labeled by numbers from some F Projective
erty of Desargues)
Plane so as to realize the algebraic structure described above. That is,

every (Desarguesian) Projective Plane "is" infact an F Projective Plane, for


some suitable F.
If the Pis finite (and hence a field) we can count the number of points
in the F Projective Plane. Say the cardinality of Pis q,where q is a power
of some prime p (as earlier). The number of F-triples is then q3, and the
number of non-zero F-triples is q3 - 1. Of those, each is equivalent to q
1 others, since each triple can be multiplied by any of the q - 1 non-zero
members of F to form an equivalent triple. The number of non-zero
- 1
divided by q - 1. That number is equal
equivalence classes is then q*
to 1 + q + q2, which will be the number of points in the corresponding
=
=
+ 4 = 7; themod-2 projective
projective plane. If # 2,l + #+ #2 l+2
= 1+ 3 + 9
our
is
"diatonic"
If
seven-point
plane
plane.
#=3, 1 + ^ + ^
=
our
or
the
mod-3
is
"chromatic"
13;
projective plane
thirteen-point
=
we
contour"
If
the
obtain
q 4,
"ternary
plane.
twenty-one-point plane;
if q = 5, we obtain the thirty-one-point plane. Six is not a prime power;
there is no six-element (skew-)field, so there is no forty-three-point pro
jective plane. But there is a seven-element field, so there is a fifty-seven
point projective plane. Eight and nine are each prime powers, so there is
a

seventy-three-point

plane,

and

Part IV: Applications?Mathematical


the Twenty-One-Point
Plane

ninety-one-point

plane.

and Musical?of

Part

III to

21 = 1+4+16;
the twenty-one-point projective plane "is" in fact the F4
where
F4 is the four-element field discussed in Part III.
Projective Plane,
=
F4 = (0,1,?,#+1};
anything added to itself is 0; a + 1 02. One verifies
=
=
=
=
= 1. 1 +
that 0(0 + 1)
03
0; (0 + l)3
1; (0 + l)2
(0 + 1)0
(0 + 1)
=
+
+
so
1
1
and
forth.
a,
[=
0]

Given a twenty-one-point projective plane, we shall now see how it


manifests the structure of the F4 Projective Plane. Using themethod dis
cussed at the end of Part I earlier, we select any "reference quadrangle"
in the plane; let its points and lines be denoted as in Example 15: points
pi, p2, p3, q, rl, r2, and r3\ lines Jl, J2, J3, Ml, M2, and M3, each line
containing inter alia the points specified on the figure.
=
=
Now, using numbers from F4, we label pi
<1, 0, 0>, p2
<0, 1, 0>,
=
=
<0, 0, 1>, and q <1, 1, 1>. The algebraic structure of the F4 Plane
p3

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

29

ProjectiveGeometry

Jl=[p2,p3,rl,
=
J2
[p3,pl,r2,
=
J3
[pl,p2,r3,
Ml = [pl, q, rl,
M2 = [p2, q, r2,
M3 = [p3, q, r3,

EXAMPLE 15
tells us what the line coordinates forJl will be. Both p2 and p3 have first
coordinate equal to zero. That is, both are points b = <bl, b2, b3> satisfy
- 0.
Thus, both points satisfy the equation (bl)(Xl)
ing the equation bl
+ (b2)(X2) + (b3)(X3) = 0, where XI = 1,X2 = 0, and X3 = 0.We know,
from the algebraic structure developed in Part III, that the coordinates
forJl will then be [1, 0, 0]: Jlis that line containing exactly such points
b as satisfy bl =0 (more explicitly, bl 1, plus b2 0, plus b3 0 equals 0).
=
=
[0,1, 0]; J2 contains just those points b satisfying b2 0.
Similarly,J2
?
=
0.
And J3
[0, 0, 1]; J3 contains just those points b satisfying b3
Example 16 updates Example 15 to display the new information.

pl

<1, 0, 0>, pl=<0,1,

0>,p3

<0, 0,1>

q-<lyl,l>

/Z-[1,0,0]
J2 =[0,1,0]
=
[0,0,1]
J3
Ml=[pl,q,rl,.
=
M2
[ply q, rl,

[pl,p3,rl,..
[p3,pl,rl,..

[pl p2yv3y...

l(W-O)
.K*2-0)
}(b3=0)

M3 = [p3yq} r3, .

EXAMPLE 16

=
in F4. The lineMl
shall find the line coordinates forMl
=
=
<1, 0, 0> and q <1,1, 1>.
[XI, X2, X3] passes through the points pl
we
must
1
XI
X2
have
0
plus
plus 0 X3 equal to 0 (since
Accordingly,
Next we

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

30

of New Music

Perspectives

and also 1 XI plus 1 X2 plus 1 X3 equal to 0 (since q


pi lies on Ml),
lies on Ml). That is,we must have XI = 0, XI +X2 +X3 = 0. Or: XI = 0,
X2 =X3. Accordingly Ml = [0, 1, l].10
In similar fashion, we see thatM2 = [1, 0, 1], and thatM3 = [1, 1, 0].
The new information is displayed in Example

pi

<1, 0, 0>, p2

/i- [1,0,0]
=
/2 [0,1,0]
=
J3 [0,0,1]

<0, 1, 0>, p3

{p2,p3,rl,.
[p3,pl,r2,.
\pl,p2,r3,.
?fi-[0,1,1]:
{?I, ft ri,.
?I? - [1, 0,1]:
{?2, ftr2,.

M3-[l,l,0]:f^,ftr3,.

17.

<0, 0, 1>, q

<1, 1,1>

(M-0)
(?2=0)
(W-0)
=
(b2 b3)
(bl=b3)
(bl=b2)
EXAMPLE 17

Now we shall find the point coordinates for point rl. Suppose those
coordinates are <bl, b2, b3>. Since rl lies on line/i = [1, 0, 0], we have bl
=
=
= 0. Since rl lies on lineMl =
[0, 1, 1], we have b2 + b3 0, whence b2
b3 (in F4). In sum: bl = 0; b2 = ?3. ri is then labeled by the triple <0,1,1>.

Similarly, r2 is labeled by the triple <1, 0,1>, and r3 is labeled by the triple
<1,1, 0>. Example 18 updates Example 17 with the new information.
In the particular field F4, 1 + 1 = 0. Therefore, in the particular sys
temwe are now considering, the sum of rl's three coordinates is zero. So
is the sum of r2's three coordinates, and so is the sum of r3,s three coor
dinates. That is, the points rl, r2, and r3 all satisfy the equation bl + b2 +
b3 = 0. (This would not be the case in a field where things added to
themselves were not zero.) Hence the three r-points all lie on the line L =
[1, 1, 1]. Example 19 updates Example 18 with the new information.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

ProjectiveGeometry

31

=
=
=
=
<1, 0, 0>, p2
<0, 1, 0>, p3
<0, 0, 1>, q <1, 1, 1>
pl
=
=
=
rl
<0,1, 1>, r2 <1, 0, 1>, r3 <1, 1, 0>
Jl =[1,0,0]:
=
[0,1,0]:
J2
=
[0,0,1]:
J3
Ml =[0,1,1]:
M2 = [1,0,1]:
M3 = [1,1,0]:

[p2,p3,rl,.
.
[p3,pl, r2,
[pl,p2,r3,.
[pl,q,rl,.
[P2,q,r2,.
[p3,q,r3,.

(bl=0)
(b2=0)
=
(b3 0)
=
(b2 b3)
(bl=b3)
(bl=b2)

EXAMPLE 18

19, we now see, realizes a seven-point plane embedded


Example
within the twenty-one-point plane. That is because the two-element field
F2 = {0, 1} is a subfield of F4. The seven points and seven lines of
Example 19 are exactly those points and lines whose coordinates can be
written using only numbers from F2. Since our original choice of refer
ence quadrangle was arbitrary,we can pause to note the interesting fact:
any choice of reference quadrangle
realizes

the seven-point

in the twenty-one-point projective plane

plane.

The visual image of Example 19, on the Euclidean plane of the page,
instance it does not show us that rl, r2,
begins to be deceptive now?for
and r3 are collinear.11 So we shall continue onwards from Example 19
purely algebraically, adjoining other points and lines of the twenty-one
point plane using number-triples of F4. The coordinates of our new points
and lines will all involve numbers of F4 that are not numbers of F2; thus
theywill all involve the number a of F4. Example 20 lists the twenty-one

points of the plane, identifying them with twenty-one non-equivalent


number-triples from F4. And Example 21 lists the twenty-one lines of the
plane, giving their line coordinates in square brackets; Example 21 also
shows which points of Example 20 lie on which of those lines.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

32

Perspectives

of New Music

=
=
=
<1, 0, 0>, p2
<0,1, 0>, p3
<0, 0,1>,
pi
?
=
=
ri
<0, 1, 1>, r2 <1, 0, 1>, r3 <1, 1, 0>
Jl=[l,0,0]:{p2,p3,rl,..
=
J2
[Q,l,0]:{p3,pl,r2,..
=
[0,0,1]:
J3
[pl,p2,r3,.
1,1]: [ri, r2,r3,...
L-[l,
Ml=[0,l,l]:[pl,q,rl,..
M2 = [l,Q,l]:{p2,q,r2,..
M3 = [l,l,0]:{p3,q,r3,..

<1,1,1>

.}(W-0)
.}(f?-0)
.}(f??0)
}(M + *2 + *3-0)
. }(t?-*3)
, }(M-*3)
. )(W-t?)

EXAMPLE 19

<0,

1,0>,

?i=<l,0,0>,

?2

ri=<0,l>l>)
si =<0,1,
#>,
ii' = <0,1, a + 1>,
tl =<a,
1,1>,
fi' =<? + !, 1, 1>,

r2-<1,0,1>,
s2 = <1, 0, #>,
? + 1>,
?2 =<1,0,
?2 = <1, #, 1>,
+ 1, 1>,
?2' -<1,*

/>3

<0,0,

1>,

r3?<l,l,0>
j3 = <1, a, 0>
??? -<1, ? + 1, 0>
t3 = <1,1, #>,
# + 1>,
r?'
=<1,1,

r-<l,

1, 1>

U =<1, #, 0 + 1>
f = <1, a + 1, #>

EXAMPLE 20

The

incidence relations of Example

21 can be checked, using the equa

tion (bl)(Xl) + (b2)(X2) + (b3)(X3) = 0 as thecriterionfordetermining


'
that point b = <bl, b2, b3> lies on lineX = [XI, X2, X3]. Thus si = <0,
=
1, a + 1> (on Example 20) and Nl'
[0, # + 1, 1] (on Example 21);
=
+
+
+
+
so
as indi
0,
(1)(?
1)
l)(l)
(a
(0)(0)
point si' lies on lineNT,
cated on Example 21. sT = <0, 1, a + 1> also lies on line G = [1, a, a +

=
1], since (0)(1) + (l)(a) + (a + 1)(* +1) 0.

To make a musical application of this structure, we shall begin sketch


ingmusic which assigns the twenty-one unordered diatonic dyads to the

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

Jl =[1,0,0]
=
J2 [0,1,0]
=
/5
[0,0,1]
=
?
[1,1,1]:
Ml =[0,1,1]
Ml[1,0,1]
M3 = [1,1,0]
M
=[0,0,1]
#2 = [0,0,1]
N3 = [0,1,0]
ATi' -[0,* + l,l]
N2' = [0 +1,0,1]
AT3' = [0 + 1,1,0]
?-[^,1,1]
K2-[1,*>1]
K3-[l,l,*]
Kl'
=[0 + 1,1,1]
Kl'
=[1,0 + 1,1]
+ 1]
K3' =[1,1,0
G = [l,0,0 + 1]:
+ 1,0]:

#=[1,0

[p2yP3ytlySlySV\
[plyp3yrlyslysi]
[pl,pl,r3,s3,s3'}
[rly rly r3y u, v)

[plyqyTlytlyt?]
[pl, qyrly tly ti }
[p3yqyr3yt3yt3' )
[PlySlyt? yt3yU)
[flySlyt? yt3yV)
[p3yS3ytly t? yU)
[plySl' ytlyt3' yV)
[flyS? ytlyt3' yU)
[p3yS3' ytlyti yV)
[rlySlyS3yti' yt? }
[TlySlyS3' yti' yt3' }
[r3ySl' yS? yti yti
[rlyS? yS3' ytly t3\
[rlyS3ySly

tly t3\

[r3, Sly Sly tly tl\

[slysi ys3' yq} u)


[Sly S3y S?

33

Geometry

y0, V)

(arithmetic inF4(\))
(W-0)
(?2-0)
(?5=0)
=
(bl+bl + b3 0)

(bl = b3)

(bl=b3)
(bl-bl)
(b3-a-bl)
(b3?a-bl)
(bl-a-bl)
(b3-(a + l)-bl)
=
(b3 (a + l)- bl)
(bl=(a + l)-bl)
+ b3)
(a-bl-bl
(a-bl-bl+bJ)
=
(a-b3
bl+bl)
+ b3)
((a + l)bl=bl
=
((a + l)bl
bl+b3)
((a + l)b3-bl+b2>)
+ (a + l)b3=0)
(bl+a(bl)
=
(bl+(a + l)bl + a(b3)
0)

EXAMPLE 21

listed

points

in

Example

20.

One

strategy?not

the

only

one!?is

to

begin with a reference quadrangle. Example 22 assigns to Example 19


the values pl = {CE}, pl = {FA}, p3 = {GB}, r3 = {EG}, rl = {DA}, r2 -

{BFU

{AC}.

EXAMPLE 22

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

34

Perspectives

of New Music

These assignments give a "C-major" sort of reference to the quadrangle?a


feature evident inExample 23a, where the bracketed segments are collinear.

pi r3 pi

rl p3 rl pi

pi

rl

r3 q

p3

p2

r2 r3 rl

p3 rl pi

EXAMPLE 23

Example 23b projects the same formal "lines" as verticalities, indicat


ing another textural possibility. No rhythm is indicated. I have not exer
cised any special control over the voice leading here, simply trying to
project vaguely Stravinskyish spacings of the dyad-triples.
Example 23 illustrates how the reference quadrangle of Example 22
yields a seven-point subplane of the twenty-one-point plane. In the
present context, it is natural to arrive at the full plane by "diminuting"
Example 23a. To see what that means, consider the first three dyads of
Example 23a. These dyads, playing the roles of pi, r3, and p2, all lie on
line J3 of the full plane, but they do not constitute the entire line J3 of
that plane. The full line J3, as we see in Example 21, contains not only
the points pi, r3, and r2, but also the points s3 and s3', points not
appearing on the reference quadrangle. We can naturally imagine dyads
s3 and s3', in this context, appearing "in passing" to elaborate the pro

gression pi, r3, p2 of Example 23a. The firstphrase of Example 24a


materializes that notion.
In that phrase, the progression pi, r3, p2, which began Example 23a, is
diminuted by transitional dyads {CG} and {EB} that appear on the off
beat eighths. To subsume those dyads onto line J3, we assign them the
point-values s3 and s3', the other points of lineJ3. The firstfive dyads of
Example 24a will then project line J3 in its entirety. In like spirit, the
second five dyads of Example 24a take the progression p2, rl, p3 from

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

pi

35

Geometry

r3 p2 p2 rl p3 p3 r2 pi pi q
ri r3 q
p3
si si'
tl tl'
s3 s3'
s2 s2'
t3 t3'

p2

t2

r2 r2 r3 rl
v u
t2'

EXAMPLE 24

Example 23a, and diminute it by inserting transitional dyads {FG} and


{CD} on the off-beat eighths; the transitional dyads are then assigned as
=
points si and si', so the phrase as a whole projects line Jl
{p2, p3, rl,
same
in
its
24a
in
the
si, si'}
entirety. Example
proceeds
spirit through
out. It turns out (in thisplane!) that every remaining point of the plane
will be referenced exactly once, by diminuting Example 23a in such a
way. One can see that by inspecting Example 21; there one sees that the
linesJl, J2, J3, L, Ml, M2, and M3 collectively reference point u, point v,
each

?-point,

each

s' -point,

each

i-point,

and

each

t' -point

exactly

once.

And those are the lineswhich appear in part on Example 23a, as progres
sions to be "diminuted."
Example 24b realizes the five-dyad lines as verticalities, elaborating the
three-dyad verticalities of Example 23b. The new dyads are shown with
filled-in noteheads. Orchestration
and/or other compositional means
entrance
amount of sustained time, etc.) could
order
of
and
exit,
(dynamics,
be used to project the idea that the open-notehead parts of each sonority is
more in the nature of a Zentralklang, while the filled-in noteheads project
Akzidentien}2 On the other hand, a composer might not choose to project
explicitly the "referentiality" of Example 23b, treating that only as a work
ingmethod to arrive at the full twenty-one-point structure.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

36

of New Music

Perspectives

DEFGABEFGABFGAB
CCCCCCDDDDDEEEE
si' pi

t3' s3

tT

t2' s2'

t3

rl

t3

r3

s3'

G A B A B B
F F F G G A
si p2 r2 s2 p3 t2
EXAMPLE 25

Example 25 shows the assignment of dyads to points, that eventuates


from the diminutions of Example 24. In working out those diminutions,
it is helpful of course to keep a running log of Example 25 as itdevelops,
to be sure that each dyad is used, and used only once.
Now thatwe have assigned a dyad to each formal point of Example 20,

we automatically know the formal lines of Example 21?in particular the


fourteen lines of that example not yet involved in the constructions of
Example 24.

t3

si

tT

pi

t3'

tl

pi

si'

rl

si

s3

t3'

tT

EXAMPLE 26

and Kl of Example 21 into the


Example 26a projects lines Nl, Nlf,
musical texture of Examples 23a and 24a, loosening up the rhythm and
"bowing" a bit. Example 26b imagines a more flexible texture than any
so far presented; in that texture itprojects linesJ3, G, and if as indicated.
The

interested reader, taking hints from this example, will quickly dis

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

37

Geometry

cover many other textures inwhich to project formal lines of the plane,
and many other musical resources for linking the lines together.
Many, many other "modes" are of course available for assigning dia
tonic dyads to the formal points of Example 20. The choice of our "C
major" mode was only to facilitate hearing for the firstsuch mode to be
studied. Within a composition, one could "modulate" between different
such modes, particularly if they have formal lines in common, with which
to "pivot." Or one could transpose a mode via any number of semitones,
introducing new tones. Example 24a, using a ficta F| within our mode,
suggests one such possibility. Finally one can use mathematical
mations that are particularly characteristic for projective planes.

Part V: Collineations

in the Projective

transfor

Plane

It is time now to bring such transformations into the picture. A collinea


tion is a function f that permutes the points of a projective plane among
themselves in such fashion that,whenever points p, q, and r are collinear,
so are ?(p), ?(q), and f(r). Loosely speaking, a collineation is an
operation
on points that "preserves lines." If point p varies along line L, then f(L),

the locus of the points f(p), will itself be a line. f(Z) may or may not be
the same line as L.
Obviously, collineations are particularly idiomatic sorts of transforma
tions to consider, given the points of a projective plane. The collineations

form a group of operations.13


Here is the basic theorem concerning collineations: let pl, p2, p3, and
p4 be points, no three of which are collinear; let pi, p2', p3', and p4'
also be such a quartet of points; then there exists a collineation f such
=
=
=
that f(pl) =pl',
f(p2)
pi,
f(p3) p3', and ?(p4)
p4'. In this context
the p' -points may all be distinct from the ^-points, or various of the p' points may be the same as various of the ^-points; the theorem obtains in
any

case.14

As one

intuits from the theorem, there are a "large number" of col


lineations on a given plane. To sharpen that intuition, let us examine col
lineations on the seven-point plane.

For that purpose Example 27 essentially reproduces an earlier sche


matic for the plane. Points 1, 3, and 5 of the example are non-collinear;
we can take them as the generic "pl, p2, and p3" of the theorem. We are

also to consider a generic "?>4" not lying on line 13, or on line 35, or on
line 51. Point 7 is the only possible such p4, for this plane. Given points
pi, pi, p3', and p4' as stipulated, let us find a collineation f as in the
theorem. We know that any such fmust satisfyf(l) = pl ', f(3) = pi, f(5)

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

38

Perspectives

of New Music

A seven-point plane: 123, 345, and 561 are all formal "lines" of the
geometry; so are 174, 376, 572, and 246
example

27

=
p3', f(7)
p4'. Subject to those conditions, we find that all other
ues for fare forced, iffis to be a collineation. Specifically, since point
collinear with points 1 and 3, f(2) must be collinear with f(l) = pi'
=
f(3)
p2' ; f(2) must then be the unique third point on the unique

val
2 is
and

line
must
third
and
be
the
f(4)
p2'. Similarly,
passing through pi'
unique
=
=
f(3) and p3'
f(5). And
point on the unique line passing through p2'
on
the unique line passing through
f(6) must be the unique third point

=f(5).
=f(l)and?3'
Some tedious reasoning shows that this f, the unique fwhich could be
a collineation subject to the conditions given, is in fact a collineation.
etc. but also f(l)f(7)f(4)
etc. are all lines of
That is, not only f(l)f(2)f(3)
as
is
f(2)f(4)f(6).
Example 27,
In the above arguments we did not need to stipulate f(7) = p4' ; the
=
=
and f(5) = p3'. For if that be
equation is forced by f(l)
pT, f(3)
pi,
as
must
then
and
be
described
the case,
above; each of those
f(2), f(4),
f(6)
some
with
from
among {pi', p2', p3'}, so none of
points is collinear
pair
can
must
them
be p4' ; hence f(7)
be p4', no other f-value being p4'. So
we have a stronger theorem for the seven-point plane. In that plane, a
pi'

unique collineation is determined by any non-collinear {pi, p2, p3\, plus


any non-collinear {pi', p2', p3'}. This enables us to count how many col
lineations

there

are

on

the

seven-point

plane.

There

are

seven

ways

of

must be dis
selecting pi' ; then there are sixways of selecting p2' ?which
tinct from pT . The line pi'p2'
contains three points; p3' must be
selected so as not to be any of those points; hence there are fourways of
selecting p3'. The total number of choices for pi', p2', and p3' is then
seven times six times four, or 168; there are 168 possible collineations.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

39

ProjectiveGeometry

is a big group of operations, for a structure of only seven points.


On those seven points, one out of thirtypermutations will be a collinea
tion.15 One naturally searches for particularly interesting subgroups of
collineations. And there are some.
In particular, there exist permutations which are seven-cycles, which are
also collineations. Consider once again the numerical diagram of Example
27; let us apply to that diagram the permutation (1734652). This permu
tation corresponds to the function fwhich maps f(l) = 7, f(7) = 3, f(3) = 4,
= 1.
. . .,
f(2)
Example 28 shows the effect of the permutation on the dia
=
gram: where point 1 used to be we find point 7
f(l); where point 7 used
=
=
to be we find point 3
f(7); where point 3 used to be we find point 4
=
. . , and where
f(3),.
point 2 used to be we findpoint 1 f(2).
That

EXAMPLE 28

Where line 123 used to be on Example 27, we now find on Example


28 the set of points [7, 1, 4} - {f(l), f(2), f(3)}. Checking back on
Example 27, we see that 714 is in fact a line ofthat example. So {f(l),
f(2), f(3)} is a line of Example 27. Likewise, where line 345 used to be on
=
Example 27, we now find on Example 28 the set of points {4, 6, 2}
{f(3), f(4), f(5)}. Checking back on Example 27, we see that 462 is in fact
a line ofthat example. So {f(3), f(4), f(5)} is a line of
Example 27. And so
forth; in this fashion we check that the permutation fmaps all lines of
Example 27 onto lines, fis thus a collineation.
That is particularly nice because f, as a permutation, is also a seven
To see why that is nice, let us begin by calculating the
cycle (1734652).
iterates of the function f; let us write f1for f, f2 for ff, f3for fff,. . . , f6for
ffffff,and f? for the identitymap on the seven symbols. The

seven iter

ates are then f*= (1734652), f2= (1362745), f3= (1423576), f4=
=
=
=
(1675324), f5 (1547263), f6 (1256437), and f? ( ).16

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

40

Perspectives

of New Music

Since fffffff
(seven times) is the identity function, it follows that (fw)(P)
_=f(m + n)where the sum m + nis
computed modulo 7. (We can throw away
seven
"f's that appears in fff. . . taken m + n times.) Thus f
any string of
and its iterates form a group of transformations on the seven points, a
group isomorphic to the additive group of integersmod 7.
Looking back two paragraphs at the permutations which express f and
its iterates, we can see that the group is simply transitive on the seven

symbols. That means: given any numerical symbol i from one through
=
seven, and any symbol y (possibly
i), there exists a member of the group
{f?, f1, f2, f3, f4, f5, f6} that permutes i to j; further, there exists a unique
such member of the group. A little thought will persuade the reader that
this simple transitivity is a direct consequence of af"s being a cycle.Given i
and j, we find those symbols in the expression f = (1734652); we then
count how many steps to the "right" we must proceed, to get from i to j
in that expression. Ifwe proceed n steps to the "right," then fn (f iterated
n times) will map i to j; furthermore no other fmwill do the trick?it will
map i to the symbol that lies m steps to the "right."
The group of f-iterates, being simply transitive, induces on the family
of seven points a GIS structure in the sense of Lewin.17 We can, that is,
treatP as a formal "interval" between symbols i and y, ifP is that partic
ular iterate of fwhich maps i to j. Or we can consider fn to be the opera
tion of formal "transposition" by that formal interval.
To illustrate the phenomenon,
let us consider the firstseven-point dia
tonic mode we explored earlier, the one which can be symbolized as in
Example

29.

Lines

123, 345, 561, 174,


275, 376, 246

Chords EGB, BFA, ACE,


EDF, GDA, BDC, GFC

EXAMPLE 29

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

41

Geometry

=
=
=
=
=
=
3, F
4, A
5, C
6, and D
1, G
2, B
Example 29 identifies E
=
= 7. Under this
becomes f
identification, the permutation f (1734652)
=
f, that is replaces E by D, D by B, B by F, F by C, C by
(EDBFCAG).
A, A by G, and G by E. This mapping can be considered as a formal
in the diatonic GIS under consideration.
"collineation-transposition"
Example 7 earlier added a third part to a Zarlino example, so that
every three-part verticality would be a formal Chord (line) of Example
29. Example 30 applies the f transformation to every note from the
opening phrase of Example 7, producing a "collineation-transposition"
of the phrase in the pertinent GIS. A one-sharp signature has been
applied for purposes of euphony. That does not change the diatonic loca
tion of the note F, whether as F natural or F sharp.
i=
?

=w=

rJ9 J
??

wm

gEf

?=

e?*E

m
Opening

phrase of Example 7, collineation-transposed


=
permutation f (EDBFCAG)
Example

by the

30

of the same
Example 31 displays another collineation-transposition
=
the
phrase from Example 7, now by the permutation f6 (EGACFBD),
inverse of f (the "complementary interval" of fin the GIS).
hears that the
Examples 30 and 31 make the collineation audible?one
Chords remain the same, en masse, even though the individual notes are
transformed. That is, the three-note verticalities of Examples 30 and 31 are
the same enmasse as the three-note verticalities of Example 7, even though
the verticalities appear in different places within the different compositions.
Another choice of permutation g, provided that g is both a seven-cycle
and a collineation, would lead to another collineation-GIS on the struc
ture of Example 29. A considerable number of such g exist for the seven

point plane.18
Many

The

resources

become

available

from

interested reader is referred to Lewin's

the presence

of GIS

structure.

book, cited in note 18.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

42

Perspectives

of New Music

^s

Opening

^^

phrase of Example 7, collineation-transposed


=
permutation f6 (EGACFBD)
Example

by the inverse

31

are particularly interested in other transformations on a


projective plane called "correlations" and "polarities." A correlation f
maps the points p of the plane onto the lines f(p), in such fashion that
collinear points map to concurrent lines: ifp, q, and r all lie on one line,
then the lines f(p)y f(q)> and f(r) all meet at one point. A correlation f
determines a mapping f* of lines onto points: given a line L, as p varies
along the line L, the varying line f(p) must continually pass through a
certain point q, where all the lines f(p) "converge." f*(L) is defined to be
that point q. One can show: if the lines LI, L2, and L3 are concurrent,
then the points f*(Ll), f*(L2), and f*(Z3) must be collinear.
The situation gives rise to a mapping Pf of points to points, which
must be a collineation. The correlation fis called a "polarity" if f*fis the
= L if and
only if
identity collineation. In case f is a polarity, then f(p)
=
a
some
f
will
which
lie
For
there
in
be
p.
f*(L)
points p
polarity
general
on their own f-transforms; the locus of such points is called a "conic,"
and the lines L which are the f-transforms of those points are called "tan
=
gents" to the conic. If f(p) L and p lies on L, then L is "tangent to the

Mathematicians

conic at point p.n The terminology is of course suggested by the way in


which polarities behave on the Real Projective Plane.
To explore the musical applications of such matters would
lengthen
this article unduly; the interested reader is referred to pertinent math
ematical texts, and should be able?if interested?to find musical applica
tions

galore.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective Geometry

Part VI: Higher


Applications

Dimensional

Projective

43

Spaces and Musical

We can extrapolate from projective planes, to consider projective geo


metric spaces of three dimensions, which we shall call "(projective)
3spaces." Abstractly, we could give incidence rules applying to certain
formal objects called "points," "lines," and "planes," to define such
structures. For instance, every pair of distinct points lies on just one line;

every triple of distinct and non-collinear points lies in just one plane;
every pair of distinct planes has in common just one line; every triple of
distinct planes, not all including any one line, has in common exactly one
point; if two points both lie in a certain plane, then so does the line
determined by the points; if a point lies on a line and that line lies on a
plane, then the point lies in the plane; and so forth.
Refining such notions into a set of axioms is important and worth
study. But here itwill be much easier to define projective 3spaces algebra
ically.Given any skew-field F, we take the formal points of the _F3space to

be the left equivalence-classes


of non-zero number-quadruples
from F.
We take the formal planes ofthat space to be right equivalence-classes of
non-zero number-quadruples. We say that the point <bl, b2, b3, b4> "lies
in" the plane [XI, X2, X3, X4] if the equation

=
(bl)(Xl) + (b2)(X2) + (b3)(X3) + (b4)(X4) 0
is satisfied, (kbl, kb2, kb3, and kb4 satisfy this equation if and only if bl,
b2, b3, and b4 do; Xlk, X2k, X3k, and X4k satisfy this equation if and
only ifXI, X2, X3, and X4 do.)
A line is then defined as the set of points lying on any two distinct
planes. Those will be the points whose numerical coordinates satisfy two
non-equivalent

such

equations.

The desired incidence rules for points, lines, and planes all fall out
from the algebraic structure so defined. These are essentially the inci
dence rules of our Euclidean geometric intuition, except that there is no
such concept as "parallel." E.g., a line cannot be parallel to a plane; a line
not in a plane must meet that plane at just one point. And so forth. The
algebraic definition ensures that every formal plane of F projective 3space
has the structure of an F Projective Plane (Fprojective 2space).
If F is a finite (skew-)field, of cardinality q, then F projective 3space
contains exactly 1 + q + q1 + q^ points. Thus F2 3space contains 1+2 + 4
+ 8 = 15 points; F3 space contains 1 + 3 + 9 +27 = 40 points.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

44

Perspectives

of New Music

of the formal study of projective 3spaces carries over from


The
notion of collineation transfers intact; correlations now map
2spaces.
to
points
planes, and so forth.
Much

POINTS
pi =<1000>
= <0100>
?2
= <00\0>
p3
= <0001>
p4

= <1100>
ql2
=
#I.? <1010>
= <1001>
ql4
= <0110>
q23
= <0101>
q24
= <00l\>
q34

rl
r2
r3
r4

=<0111>
= <1011>
= <1101>
= <1110>

? = <1111>

PLANES
.O =[1000]:
K2 = [0100]:
.O = [0010]:
K4 = [0001]:
L12 = [1100]:
L13 = [1010]:
L14 - [1001]:
L23 = [0110]:

?24 = [0101]:
L34 = [0011 ]:
Ml = [0111]:
M2 = [1011]:
M3 = [1101]:
M4 = [0111]:
^=[1111]:

{p2,p3,p4, q23, q24, q34, rl]


{pl,p3,p4,ql3,ql4,q34,r2}
{pi,p2,p4, ql2, ql4, q24, r3\
[pi,p2, p3, ql2, ql3, q23, r4\
[p3,p4,ql2,q34,r3,r4,s]
[p2,p4, ql3, q24, r2, r4,s]
[p2,p3, q!4, q23, r2,r3, s]

(W-0)
=
(62 0)
(W-0)
=
(b4 0)
=
=
(bl + b2 0; bl b2)
=
=
+
(bl b3 0; bl b3)
=
(bl + b4 0;bl= b4)

[pl,p3,ql3,q24,rl,r3,s]
[pi,p2, ql2, q34, rl, r2, s\
[pi, q23, q24, q34, r2,r3,r4\
[p2,ql3,ql4,q34,rl,r3,r4\
[p3,ql2, ql4, q24, rl, r2,r4\
[p4,ql2,ql3,q23,rl,r2,r3\

=
=
(b2 + b4 0; b2 b4)
=
=
+
(b3 b4 0; b3 b4)
=
+
+
(b2 b3 b4 0)
=
(bl+b3 + b4 0)
=
+
+b2
b4
(bl
0)
=
(bl + b2 + b3 0)

[pi, p4, ql4, q23, rl, r4, s)

(all ^-points, and s\

=
=
(b2 + b3 0; b2 b3)

(bl + b2 + b3 +W

0)

EXAMPLE 32

Example 32 displays the points and planes of F2 projective 3space. To


relieve the page of visual clutter, pi = <1, 0, 0, 0> has been abbreviated as
= <1000> and so forth.
pi
We can find the lines of the 3space in several ways. Let us consider the
line common to planes K3 and K4, for a start.That linewill comprise the
points common to the two planes, and we can read those off Example 32
as [pi, p2, ql2\. We see that those are exactly the points whose coordi
nates satisfyboth b3 =0 and b4 = 0?the equations determining both K3

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

Geometry

45

and K4. There is an interesting feature about the line [pi, p2, ql2\: ifwe
"add" pi to ply coordinate by coordinate, we obtain the coordinates of
q!2y the third point on the plp2 line. Schematically:

<1000>
plus <0100>,
equals <1100>.

coordinate by coordinate,

Ifwe "add" pi to ql2 in the same way, we obtain the coordinates of p2y
the third point on the plql2
line. Schematically:
<1000>
plus <1100>,
equals <0100>.

coordinate by coordinate,

feature generalizes to all lines in F2 3space. That is, suppose a =


=
=
<aly a2y a3y a4> and b <bl, b2, b3, b4> and c <cl, c2, c3, c4> are the
three distinct points of a line in that space. Then cl = al + bl, c2 = a2 +
=
=
b2, c3 a3 + b3, and c4 a4 + b4. This enables us to find the third point
of a line very quickly, knowing the other two.
We have seen that [pi, p2, qli\ is a line, specifically the line common
The

to planes K3 and K4. Adding together the plane-coordinates


for those
two planes, [0010] and [0001], we obtain [0011], the coordinates for
plane L34. Checking Example 32, we see that the line indeed lieswithin
L34y as well as K3 and K4. K3, K4, and L34 are the three planes which
contain the line.We could thus denote the line as well by \K3, K4, L34\.
To be symmetrical, we shall then write /pi, p2, ql2/ for "the line [pi,

p2, ql2)"

In general, we see that ifI and j are distinct numbers from one to four
inclusive, with i < j, then /pi, pj, qij/ will be a line. If m and n are the
other two numbers between one and four inclusive, with m < n, then the
line /pi, pj, qij/ is the same as the line \Km, Kn, Lmn\, the line where

those three planes all meet. There are sixways of choosing such an i and
j, so there are six such lines. Example 33 continues the tally of lines
within F2 3space in this fashion.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

PerspectivesofNew Music

46

/piy PJ>
W/

/pi, qmn, rj/


/pi, ri, s/

\Km,Kny Lmn\
=

\Kj,Mi, Lmn\

\Ljm,Ljn, Lmn\

/qjm, qjn, qmn/


/qij, qmn, s/
/qij, ri, rj/

\Ki,Mi, N\

\Lij,Lmn, N\

\Lmn,Mm, Mn\

(i< j; m< n the other numbers)

6 such lines
(anydistinct i and j; m< n the other numbers)
12 such lines
m<
n
the
other
numbers)
(any i;j<
4 such lines
(any i;j< m< n the other numbers)
4 such lines
m<
n
other
the
(i< j;
numbers)
3 such lines
(i< j; m< n the other numbers)
6 such lines
example

33

The example shows that there are thirty-fivelines in all. Let us check out
the coordinate addition feature for the line containing points p2 and r4.
p2 =<0100>;
=
adding r4 <1110>, coordinate by coordinate,
we obtain <1010>, which is ql3.
As we see from Example 33, /p2, ql3, r4/ is indeed a line. The same
addition feature will obtain for the coordinates of any three planes pass
ing through one line. The feature depends on the fact that F2 has only
the two elements 0 and 1,with 1 + 1 = 0.

F2 3space applies very naturally to a number of musical


sider for instance Example 34.

contexts. Con

smalltom-tom

jWijP]jIPiJ]J]J?]]iPJ7P
example

34

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

47

Geometry

The example is taken frommeasures 57-64 ofMilton Babbitt's All Set.


Let us write a "1" every time an eighth note in the stream is played by
the tom-tom, and a "0" every time the note is played by the snare drum.
Then the firstgroup of four notes from the example can by symbolized as
the second group of four notes can be symbolized as "1111,"
"1011,"
and so forth; the pertinent aspect of the sixteen four-note groups can be

as 1011 1111 1000 0010 0101 1110 0110 0011 1100 1001
transcribed
1010 0001 0000 0100 0111 1101.19

^
r'rr'rrrr'r*" '" r*?*r*r!rrr*!*rr*1"
rr- irurir*r*i- *ri
example

same

abstract

structure.

Here

\ir l*rrrlrr*r

35

35, taken from measures

Example
the

89-104

"1"

of the same piece,

represents

quarter

note

shows
in

the

bass drum, and "0" represents a quarter rest in that instrument. The six
teen groups of four quarters transcribe into exactly the same binary code:

1011 1111 1000 0010 0101 1110 0110 0011 1100 1001 1010 0001
0000 0100 0111 1101.
article on Babbitt's
Christopher Wintle's
Semi-Simple Variations
includes an analysis of other such rhythmic series, there involving groups

of four

sixteenths

within

Babbitt's The Head


the

singer,

there

are

the quarter-note

pulses.20

of theBed sets a poem of sixteen stanzas. Besides


four

accompanying

instruments;

for

present

pur

poses let us call them "instrument 1" through "instrument 4." The first
stanza is "set" as an epigraph: all instruments are silent. (It might be
appropriate to have the singer speak these verses.) Each of the other fif
teen stanzas is setwith its own unique combination of instruments. One
can attach a code number to each stanza by writing a "1" for each instru

ment that is participating, and a "0" for each instrument that is sitting
out. Thus "1011" encodes that stanza where instruments 1,3, and 4 are
playing, while instrument 2 is sitting out. The epigraph encodes as 0000.
One can easily imagine many other contexts for which such code is
appropriate. For instance, we could define four registers in a piece, within
one

of which

themusic,

any

note

must

sound.

Given

a certain

articulated

span

of

the symbol "1" could indicate "being loud" (constantly or on

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

48

PerspectivesofNew Music

the average) during that span; the symbol "0" could indicate "being
soft" (ditto) during that span. Supposing that events in each register
could be unequivocally determined as "loud" or "soft" during any partic
ular formal "span," then each span would generate a code label to indi
cate the loudness or softness of its various registers. 0101, for instance,
would mean that events in registers 1 and 3 were soft during the span,
while events in registers 2 and 4 were loud.
In the musical contexts discussed so far, each four-digit binary number
encodes what can properly be called a state of a musical system, that is, a

particular condition in which we may find the system over a pertinent


span of time. In each case, we may find the system in state 0000 just as
well as any of the other states. That is one reason why it seems strained to
states, with the points <1000>,
identify (only) the fifteen "non-zero"
etc. of F2 3space. Another reason is that the incidence relations
<0100>
of points, lines, and planes in the geometric space are not particularly well

reflected in the relations of corresponding states. For example, in the


are collinear. (Observe
and <1001>
geometry, points <0111>,
<1110>,
the addition feature.) But ifwe try to realize these points in the bass
drum texture of Example 35, we shall find ourselves trying to group in
some

"natural"

way

the

three

measures

<rest,

beat,

beat,

beat>,

<beat,

beat, beat, rest>, and <beat, rest, rest, beat> ofthat example; the three
measures do not seem to group musically in any particularly natural way.
Of course we have made plenty of "artificial" identifications, of formal
geometric points with various musical quiddities, in earlier parts of this
paper. But we can in fact find a more natural sort of identification here.
to indicate the corresponding state of a musical
Let us write "@1001"
the
is
system;
"@," that is, "at" or "in" the corresponding state.
system
We can then identify the fifteen non-zero number-quadruples with the
fifteen possible changes of state. The change of state <1100> means:
change what things 1 and 2 are doing; do not change what things 3 and
4 are doing. For example, suppose the system is in the state @1001;
apply

the change of state <1100>. Then the firstand second digits of @1001
will change, while the third and fourth digits remain what theywere. The
change of state thus produces the new state @0101.
Here we envisage the change of state as a dynamic entity, operating
upon the static states. It has somewhat the nature of a formal "transposi
tion," of one state into another, or of a formal "interval" between con
secutive states.We could develop the notion farther in that direction, if
we wished, but then we would have to allow the null change "<0000>"
as a formal "interval" or "transposition." That would lead to a different
mathematical model
It
(a "vector space" rather than a "geometry).
would also mar the idea of change, since in some ways it is problematic to

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

49

Geometry

hear a formal "change of state" of "<0000>," when none of the musical


parameters actually changes: would we not more intuitively hear a "repe
tition" or "prolongation"? Of course, ifa composition forces us to accept
the idea of "one state per measure," or something of the sort, then we
could force ourselves to hear a formal "change of state" with each new

measure.
teen

Still, there is something attractive about allowing only the fif

non-zero

The

number-quadruples

to represent

algebra of state changes is easy:

if the system is in state @ al


the change of state
produces the new state @(al

a2
<bl
+ bl)(a2

"changes"

a4,a3
b2 b4>
+ b2)(a3

b3
+ b3)(a4

of

state.

+ b4),

where all the algebra is in F2. That is because al "changes" if and only if
it becomes al + 1, mod 2; ifal does not change, it remains al + 0. So if
bl =1, al changes to al + bl; if bl = 0, al remains al + bl. And similarly
for a2,

a3,

a4.

=
=
=
<dl, d2,
Suppose that b <bl, b2, b3, b4>, c <cl, c2, c3, c4>, and d
are
on
some
us
d4>
the
three
line
of
the
Let
d3,
geometry.
points
imagine
and applying the changes of state b, c,
starting at state @(al)(a2)(a3)(a4)y

and d seriatim thereafter.At the end of this process we will arrive back in
the opening state, @(al)(a2)(a3)(a4).
That is because, starting in the
we
will
seriatim
progress
opening state,
through states
@(al+bl
@(al + bl + cl
@(al + bl + cl + dl

)(a2 + b2
){a3 + b3
)(a4 +),b4
)(a2 + b2 + c2
)(a3 + b3 + c3
)(a4 + b4 + c4 ), and
)(a2 + b2 + c2 + d2 )(a3 + b3 + c3 + d3) (a4 + b4 + c4 + d4).

= bl +
Now, as we observed earlier, dl
cl; adding dl mod 2 to both sides
ofthat equation, we derive 0 = bl + cl + dl. Likewise 0 = b2 + c2 + d2,
and so forth. Hence
(al + bl + cl + dl) is simply al, and so forth; the
as asserted.
final state above is simply @(al)(a2)(a3)(a4),
Thus the spatial "closure" of the line /b, c, d/ is reflected in a musical
closure here, the closure of the progression we can symbolically denote
=
by @a, @(a + ?), @(a + b + r), @(a + b + c + d) @a.
Since b + c = d, we can rewrite the above progression as @a, @(a + ?),
@(a + d), @0. State @(a + c) is not traversed during the progression, but
the state can easily be accessed from any of the states @a, @(a + b),or@(a

+ ?), by applying an appropriate change of state from among the changes


b, c, and d: @(a + c) can be attained by applying change c to state @a, or
by applying change d to state @(a + ?), or by applying change b to state
@(a + d).

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

50

Perspectives

of New Music

If one starts at a statewithin the orbit [@a, @(a + b), @(a + c), @(a + d)},
and ifone applies successively any sequence of changes from within the line
/by c, d/, one always remains within that orbit of four states. The sixteen
states partition up into four such orbits for the line /b, c, d/. A different
linewill partition the sixteen states into four different four-state orbits.
Similarly, if change b varies within any plane of the geometry, the six
teen states are partitioned into two eight-state orbits. For instance, the
points bwithin the plane Kl are those bwhose firstcoordinate bl is zero;
when such b are applied in any sequence starting from a certain state @a,
then the first state-coordinate will never change from its original value
the particular progression. If instrument 1 was silent to

al, whatever

begin with, instrument 1 will remain silent throughout the plane


progression; if instrument 1 was playing to begin with, instrument 1 will
remain playing throughout the plane-progression.
If instrument 1 was
silent to begin with, then any other state inwhich instrument 1 is silent
will be accessible via some ?-change within the plane Kl.

The mechanics of collineations and other algebraic/geometric notions


lend even more structuring to this abstract system. The interested reader
may explore to what extent the system may or may not be analytically

suggestive in connection with pertinent music by Babbitt (or others). I


have not myself as yet undertaken such analyses; the present article is
meant only to suggest and develop compositional possibilities. An appen
dix will

offer

some

observations

on

change

structure

in the

state-rows

of

All Set, and of the Semi-Simple Variations.


The geometric system as it stands can be applied to anymusical context
where "states" are determined by each of fourmusical quiddities necessar
ily doing either "this" or "that" over each well-defined "span" of the
music.21 The entire apparatus generalizes easily to fivemusical quiddities,
sixmusical quiddities, D musical quiddities each doing "this" or "that";
themathematics follows in theway the reader will expect, involving projec
tive 4space, projective 5space, projective (D- 1) space, all with respect to
the field F2. Many of the observations made above rely heavily on the
properties of F2; the systematics do not generalize very satisfactorily to F3
situations where each musical quiddity is necessarily
spaces, modeling
one
of
three
doing
things (rather than one of two things) over any span, or
F4 spaces, modeling situations where each musical quiddity is necessarily
doing one of four things over any span, etc. Nevertheless, musical situa
tions which can be articulated by distinguishing "this or that" among suffi
cientlymany quiddities are ubiquitous enough, so that the applicability of

the F2 spaces, modeling changes of state, seems significant.22


On another tack: there are thirty-one points in F2 4space. (31 = 1+2
+ 4 + 8 + 16.) Itmight be interesting to try applying that structure to the

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

ProjectiveGeometry

51

thirty-one-tone scale, in various modes. There would be "lines" of three


notes each, "planes" of seven notes each, and "hyperplanes" of fifteen
notes each. One might begin, for instance, by staking out a "basic" dia
tonic

seven-note

plane.

so that the scale could


We have already observed that 31 = 1 + 5+25,
also be structured by the F5 plane; there the "lines," within any particu
larmode, would be thirty-one hexachords.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

52

PerspectivesofNew Music

Appendix

Here we shall examine some aspects of change-structure in the rhythmic


rows of Babbitt's All Set, and of his Semi-Simple Variations.

STATES CHANGES
1. @1011
2.

HEXCODE

<0100>

<0111>

<1010>

<0111>

@1111

3. @1000
4.

6.

7.
8.
9.

@0010

5. @0101

D
C
<1011>

<1000>

<0101>

<1111>

<0101>

<0011>

<1011>

<0001>

<0100>

<0011>

<1010>

@1110

@0110

THIRD POINT

3
D

?0011

@1100

10. @1001

11. ?1010

12. ?0001

13. ?0000

14. ?0100

15. ?0111

16. ?1101
EXAMPLE 36

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

ProjectiveGeometry

53

column of Example 36 headed "STATES"


lists,from top to bot
row
states
of
the
from
All
the
sixteen
the
list of states may be
tom,
Set,
checked against Examples 34-5 and the earlier discussion thereof.
The next column, headed "CHANGES,"
lists the changes between
at the top
consecutive pairs of states. For example, the change <0100>,
is
#1
of the CHANGES
the
STATES
and #2 to
column,
change between
The

the left.STATE 1 is@1 Oil and STATE 2 is?111 1; in passingfrom

1 to STATE 2, the first digit of the STATE


code does not
nor
nor
does the third digit change,
the fourth digit; the second
change,
entry
digit does change. That behavior is encoded by the CHANGE
no change in the first, third, and fourth digits; change in the
<0100>:
STATE

second digit. As discussed earlier, the CHANGE


may also be computed
algebraically, by adding the two STATES digit by digit mod 2: 0, the first
is 1 + 1 (mod 2), the sum of the firstdigits from
digit of the CHANGE,
STATE 1 and STATE 2; 1, the second digit of the CHANGE,
is 0 + 1
(mod 2), the sum of the second digits from STATE 1 and STATE 2; and
so forth.

The next column of Example 36, headed "HEXCODE,"


provides a
convenient shorthand for labeling the various CHANGES.
Each HEX
CODE
symbol is the hexadecimal number corresponding to the four
to its left. Thus
the first
digit binary number of the CHANGE
CHANGE,
<0100>, when read as a binary number, corresponds to the
decimal number 4. (It has 0 ones, 0 twos, 1 four, and 0 eights.) The dec
imal number 4 corresponds to the hexadecimal number 4, the firstHEX
CODE
number. The third CHANGE,
<1010>, when read as a binary
number, corresponds to the decimal number 10. (It has 0 ones, 1 two, 0
fours, and 1 eight.) The decimal number 10 corresponds to the hexadec
imal number A, the thirdHEXCODE
number. (After reaching 9, hexa
decimal numbers continue with A = decimal 10, B = decimal 11, C =
decimal 12, D = decimal 13, E = decimal 14, and F = decimal 15; decimal
16 is then hexadecimal 10.) The eighth CHANGE,
<1111>, when read
as a binary number, corresponds to the decimal number 15. (It has 1
one, 1 two, 1 four, and 1 eight.) The decimal number 15 corresponds to

the hexadecimal number F, the eighth HEXCODE


number.
Each CHANGE
number represents a formal "point" in the projective
are the same; hence
geometry at hand. No two consecutive CHANGES
determines a formal "line" of the
every pair of consecutive CHANGES
geometry.23 Since that geometry is a geometry over the mod-2 field,

every line contains exactly three points. As discussed in the text earlier,
given two points, we can determine the third point on their line by add
ing theirmod-2 coordinates, digit by digit. Let us consider, for example,
in Example 36, the CHANGES
the first two CHANGES
<0100>
and

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

54

Perspectives

of New Music

<0111>. Adding those two CHANGES


digit by digit mod 2, we obtain
the CHANGE
<0011>.
So <0011>
is the third point on the line deter
mined by <0100> and <0111>. That third point, <0011>, has the hexa
decimal code number 3. (It has 1 one, 1 two, 0 fours, and 0 eights, so it
corresponds to decimal 3, which is hexadecimal 3.) The final column of
enters the hexadecimal code
POINT,"
Example 36, headed "THIRD
number "3" in the appropriate place, to show that the point with HEX
CODE
3 is the third point on the line given by the points with HEX
4 and 7. (NB: we do not perform any arithmetic with the hex
numbers, which are only convenient labels; all our arithmetic is per
formed mod 2 on the four-digit binary numbers, digit by digit.)

CODES

For another illustration we can consider the fifthand sixth CHANGES


on Example 36, the CHANGES
<1011> and <1000>. Adding those two

CHANGES digitby digitmod 2, we obtain theCHANGE <0011>. So

is the third point on the line determined by <0100>


and
<0111>. That third point, <0011>,
is again the point with HEXCODE
3. Accordingly the hex number "3" is entered in the final column of
Example 36, to the right of the fifthand sixth changes, to show that the
3 is the third point on the line given by the
point with HEXCODE
<0011>

B and 8.
points with HEXCODES
on Example 36,
In general, then, any two consecutive HEXCODES
to
with
POINT
number
the
THIRD
code
the
together
right, constitute
some

formal

"line"

of

our

geometry.

We

have

just

seen,

in connection

with the fifth and sixthCHANGES,


that the points with HEXCODES
B,
a
recurs
on the table, in connection with the
and
3
form
line.
That
line
8,
tenth and eleventh CHANGES.
We can thus observe that the line deter
mined by the tenth and eleventh CHANGES,
hex3 and hexB, is the same
as the line determined by the fifth and sixth CHANGES,
hexB and hex8.
the
line
the
thirteenth
and
fourteenth
determined
Similarly,
by
hex4 and hex3, is the same as the line determined by the
CHANGES,
firstand second CHANGES,
hex4 and hex7. Our geometry thus enables
us to find interesting sorts of "recurrences" in the pattern of consecutive
CHANGES.
Itmay be significant that our two recurrent "lines," hexB83
and hex473, are coplanar. (The points with HEXCODES
B, 8, 3, 4, and
=
+
7 all lie in one plane?the
"b3
b4
0.")
plane
The pattern of CHANGES
has a number of other interesting struc
tural features as well. The palindrome hex7-hexA-hex7,
formed by the
is one such feature; so is the palin
second through fourth CHANGES,
drome hex5-hexF-hex5,
formed by the seventh through ninth
CHANGES.
Itmay be significant that the points involved in the two pal
indromes all lie in one plane. (The points with HEXCODES
7, A, F, and
5 all lie in the plane "?2 + b4 = 0.")

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective Geometry

55

The hex5-hexF-hex5
palindrome (seventh through ninth CHANGES)
is preceded by points hexB and hex8, from the hexB8 line, and followed by
points hex3 and hexB, also from the hexB8 line. Thus a looser sort of pal
indromic structure extends from the hexB of the fifthCHANGE
through
That articulates the first four
the hexB of the eleventh CHANGE.
the loose palindrome,
and the last four
CHANGES,
preceding
the loose palindrome. The first four CHANGES
CHANGES, Mowing
to
the
line hex473; so do the last four CHANGES.
include reference
Of course, all thiswork is pre-analytic, not strictlyanalytic. Still, itdoes
suggest things to investigate during the course of an analysis. Further
things are suggested by themeter of Example 34, which groups STATES
in pairs. We could accordingly examine CHANGES
between STATES
between consecu
lying two order-numbers apart, as well as CHANGES
on
we
tive states. Following up
that idea,
could also examine CHANGES
between STATES yet more widely separated in order position, wherever

any aspects of themusic make that analytic strategy plausible.


Example 37 is the analog for Example 36, using the rhythmic state
row from the Semi-Simple Variations. (Wintle's ample analysis is cited in

note 21.) Here we find strong patterning in themiddle of the CHANGE


are hexl-hex6-hex2
succession: the fifth through eighth CHANGES
are hexl-hex6-hex4
hexE; the ninth through thirteenth CHANGES
can
hexE-hex2. Hex4-following-hex6
(tenth and eleventh CHANGES)
be regarded as a substitute for hex2-following-hex6
and
seventh
(sixth

because the points with HEXCODES


6, 2, and 4 form a
CHANGES),
line. (That is, the line hex64 is the same as the line hex62.) After the sub
stitution of hex4 for hex2, at the eleventh CHANGE,
hex2 appears after
that cements
all, following hexE (twelfth and thirteenth CHANGES);
the association to the seventh and eighth CHANGES,
where hexE fol
lowed hex2.
Babbitt changes by <1111> each STATE of the row, to obtain what he
treats as "the inversion" of the row.When a given STATE of the row is
transformed in this fashion, each of its attacks becomes a rest in the
"inverted" STATE,
and each of its rests becomes an attack in the
"inverted" STATE. The "inverted" STATE thus complements the rhyth
mic pattern of attacks and rests in the given STATE.
Given our
is rather a
algebraic/geometric
point of view, Babbitt's "inversion"
CHANGE
of the entire row seriatim, CHANGING
each STATE by
<1111>
seriatim. The transformation behaves more like a formal "trans
each
position," than a formal "inversion"; one could similarlyCHANGE
STATE of the row seriatim by any other <bl b2 b3 b4>. Given any two
s and t of the original row, let s' and t' be the transformed
STATES
STATES under any such "transposition"
(including Babbitt's). Let us

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

56

PerspectivesofNew Music

STATES CHANGES
1. @1111

HEXCODE

<0011>

<1101>

<1010>

<1001>

<0001>
7
<0110>
7. @0101 4
<0010>
8. @0111 C

2. @1100 E
3. @0001 7
4. @1011

THIRD POINT

5. @0010 8

6. @0011

<1110>

9. @1001 F
<0001>
10.@1000 7

6
2
E

ll.@1110

<0110>

<0100>
A
<1110>

12.@1010

<0010>

13.@0100 C
14.@0110 9

<1011>
15.@1101 6
<1101>
16.@0000

B
D
EXAMPLE 37

denote by ub" the transposing CHANGE


<bl b2 b3 b4>. We can write s'
= s + b and t' =t + where the
denotes
b,
plus sign
digit-by-digit addition
from s' to t' is given by the four-digit
mod 2. Then the CHANGE
binary number s' + t', and that number is (s + b) + (t + b), which is s + t,

from s to t. In sum, the CHANGE


which is the CHANGE
from the
transformed STATE s' to the transformed STATE t' is exactly the same

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

Projective

Geometry

as the CHANGE
from the original STATE s to the original STATE
The transformation preserves the CHANGE-structure
exactly.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

57

t.

58 PerspectivesofNew Music

Notes

1. All the projective planes we shall be studying satisfy another property


as well, the Property of Desargues. The property is cumbersome to
state, but it is highly important, and it should be put down here:
Let x, y, z, and u be four distinct points, no three of which are col
linear. (The four Properties guarantee the existence of such a quartet,
as we shall later see.) Let x' be any third point on the line xu; let
y' be
on
on
zu.
third
line
let
z'
the
be
the
third
line
any
yu;
any
point
point
can
:
not
show
that
x'
does
if
it
then
the
line
did,
(One
equal y'
x' u would equal the line y' u, i.e., the line xu would equal the line yu.
But then x, y, and u would be collinear, contradicting the original
conditions of choice. In similar fashion, one shows that y' is not z',
and that z' is not x'. Thus x' and y' determine a line x' y', etc.)
Let h be the point where line xymeets line x' y' ; let i be the point
where line yz meets line y' z' ; let j be the point where line zx meets
line z' x'. The Property of Desargues
is this: the three (distinct)

points h, iyand j will be collinear.


To see how the property ismanifested in Example 1, let us take for
instance x = Ann, y = Bill, z = Carol, and u = Gladys. One checks that
no three of these students are in any one course
together, so no three
of

them

are

"collinear."

Take

x'

to be

any

third

student

sharing

course with x = Ann and u = Gladys; here x' can only be Dan. Take
= Bill and u =
y' to be any third student sharing a course with y
can
to
here
Eve.
be
z'
Take
be
third
student shar
any
Gladys;
y'
only
=
=
ing a course with z Carol and u Gladys; here z' can only be Frank.
Now construct the point hwhere lines xy and x' y' meet: find the stu
dent hwho is taking both a course along with Ann and Bill, and also
a course along with Dan and Eve. That student, as we see from
=
=
Example 1, is Frank. So h Frank, i the point where line yz meets
line y' z' = the student taking both a course along with Bill and
=
= the
Carol, and a course along with Eve and Frank. So i Dan. j
= the student
zx
meets
where
line
line
z'
x'
point
taking both a
course along with Carol and Ann, and a course along with Frank and
Dan. So j = Eve. According to the Property of Desargues, h, i, and j
are collinear; that is, therewill be a course whose students include h =
=
=
Frank, i Dan, and j Eve. And in fact there is such a course.
which
Projective planes
satisfy the Property of Desargues are called
Since
all
the
Desarguesian.
planes we shall be studying are Desargues
we
assume
shall
this
ian,
property as well, during the present article,
"
when we speak ofa ccprojective
plane.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

ProjectiveGeometry 59

are distinct by construction; so are Tl and 72; so are Zl


and Tl are distinct because distinct lines have only one
in
common, and Wis already the one point in common between
point
the WXlinc and the WT line. By the same argument, either of theX
and X2
and Z2. XI

2. XI

points on Example 5 a is distinct from either of the T-points; similarly


theXs and theZs are distinct; similarly the Ts and the Zs are distinct.

3.

Proof: Let v symbolize the line containing XI and Tl, in Example 5a.
Let F symbolize the third point of v. F cannot be W, for the WXline
meets line v in only one point, and that point must be XI (not V). V
cannot be X2, for the same reason. F cannot be T2, for the WT line
meets line v in only one point, and that point must be Tl (not V).

Hence F must be either Zl or Z2.


Without
loss of generality, we can label Zl and Z2 so that V = Zl.
Tl -Zl
is a line.
Thus we may suppose that v =XI

Let v' be the line containing XI and 72. Let V be the third point
of v'. By the same sort of reasoning just traversed for v and V, we can
of Example 5a. F
in factmust
conclude that F must be a Zpoint
be Z2. For ifF were Zl, then the distinct lines v and v' would have
two distinct common points?both XI and Zl.
72 -Z2 is a line.Now let v" be the line containing
Thus vf =XI
X2 and 72; let F'
be the third point of v". As before, V" must be
one of theZ-points. As before, too, V" cannot be Z2. For in that case
72 -Z2, and v" would have two distinct points in
v" would be X2
common with v' =X1 -T2 -Z2. So V" must be Zl, and v" must be
X2-T2-Z1.

This demonstrates what was to be proved: the third point F of the


line v containing XI and Tl is the same as the third point V" of the
line v" containing X2 and 72.
4.

The interested reader may verify this as an exercise, using arguments


like those in note 3 above.

5. Gioseffo Zarlino, On theModes: Part Four ofLe istitutioni harmon


iche, 1558, trans.Vered Cohen, ed. Claude Palisca (New Haven: Yale
University Press, 1983), p. 62.
6.

Points 1 and 7 determine a line, line 17. Line 17 intersects line 345
in a unique point. 3 cannot be that point; if itwere, then lines 173
and 123 would share two distinct points, a contradiction. Similarly, 5
cannot be the point where line 17 intersects line 345. Therefore, 4
must be the point where line 17 intersects line 345. So 174 is a line.
Similar reasoning shows that 376 is a line, and that 572 is a line.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

60 PerspectivesofNew Music

7.

Points 2 and 4 determine a unique line,which contains three points.


Line 24 is not any of the three "primary" lines 123, 345, and 561.
Nor is it any of the three lines observed in note 5, that is, lines 174,
line 24 is the "seventh line" of the system. A
376, and 572. Hence
similar argument shows that line 46 is also "the seventh line," as is
une 26. That is, points 2, 4, and 6 all lie on "the seventh line," which
is therefore 246.

8.

Larry Polansky and Richard Bassein, "Possible and Impossible Mel


Music Theory 36.2
ody: Some Formal Aspects of Contour," Journal of
259-84.
(Fall 1992),

9.

Proof: Suppose a and b are rational numbers, not both zero, such
that 02 = 2b1.We will demonstrate a contradiction.
Since not both a and b are zero, neither a nor b is zero (consider
ing the equation they are assumed to satisfy). In particular, b is not
zero. So we may divide by b; let r be the absolute value of a/b. r is a
=
positive rational number, and by the equation assumed, r2 2. Thus r
=
Jl, and the square root of 2 is a rational number. That is not true,
so our contradiction is established.

10. Any other non-zero number-triple from F4 whose firstcoordinate is


zero and whose second and third coordinates are
equal will be of
form [0, k, k] for some non-zero k. That triple is (right-equivalent
to [0,1,1].
11.

[Lewin's original typescript figures depicted


points in his reference quadrangles.]

no

lines connecting

terminology ismeant in the sense of Stockhausen's


(1963, Universal Edition 13993).

12. The

Plus Minus

13. It is clear that ifboth fand g are collineations, then so is fg. (It must
"preserve lines.") It is not immediately clear that the inverse function
of fwill be a collineation, but such is the fact. It is easily demon
strated by using the algebra of the appropriate skew-field.
14. The proof ismost easily provided by algebraic techniques; readers
with some background in linear algebra should be able to provide a
proof, or to follow one in a mathematical text.
15. There are 7! possible permutations on the seven points (7 6 4) of
these are collineations. 7! divided by that number equals (5-3-2),
which is 30.

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

ProjectiveGeometry 61

16. f1maps 1 to 7 and f*maps 7 to 3, so f2 (f iterated) maps 1 to 7 to 3,


=
i.e., 1 to 3. f2 (13 something).
to
4 and f1maps 4 to 6, so f2 (f iterated) maps 3 to 4 to
3
f1maps
=
to
3
6.
f2
6, i.e.,
(136 something).
And so on, up to f2= (1362745). Then:
f1maps 1 to 7 and f2maps 7 to 4, so f3 (f twice iterated) maps 1 to 7
to 4, i.e., 1 to 4. Or one can just count three steps to the right of the
=
symbol 1, in the expression (1734652) for f. f3 (14 something).
to
the right of the symbol 4, in f = (1734652),
Counting three steps
we find the symbol 2. f3maps 4 to 2. f3= (142 something).
three steps to the "right" of the symbol 2, in f =
Counting
we
find the symbol 3. f3maps 2 to 3. f* = (1423 some
(1734652),
so
forth.
thing). And
17. David Lewin, Generalized Musical Intervals and Transformations (New
Haven: Yale University Press, 1987). A discussion of simply transitive
groups and GIS structure appears on pages 157 and following.
18. This situation generalizes to finiteprojective planes in general. That is,
given a finiteprojective plane with _Vpoints, there always exists a permu
tation f of thoseNpoints which is both an _V-cycleand a collineation.
The situation generalizes even farther: given a finiteprojective space
of any dimension, with N points, there always exists a permutation f of
those N points which is both an _V-cycleand a collineation. We shall
discuss projective spaces of three and more dimensions later on.
I am grateful to Prof. Noam Elkies of the Mathematics Depart
ment, Harvard University, for supplying me with a proof of the
above fact (in a private conversation).

19. I thank Anton Vishio


my

for calling this passage,

and the next one, to

attention.

20. Christopher Wintie, "Milton Babbitt's Semi-Simple Variations," Per


spectives ofNew Music 14.2 and 15.1 (1976), 111-54. The rhythmic
analysis

appears

on

pp.

141-5.

21. The "spans" are perhaps most naturally defined as temporal seg
ments. They could easily be defined formally by other musical criteria
as well, but I do not see how to get enough of them thatway to gen
erate interesting patterns, let alone an interesting composition.

22.

If each quiddity is doing one of three things, one can formally break
the performance down analytically into two binary oppositions, first
doing "either thing 1 or one of the other things," then, ifdoing one
of the other things, doing "either thing 2 or thing 3." And so forth

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

62 PerspectivesofNew Music

for higher numbers of things. The logical possibility, however, will


probably evoke a suspicious or hostile musical reaction from most
readers. (It does from me.)
Indeed, the idea of analyzing virtually everything by sufficiently
many binary oppositions is under some fire these days, from a num
ber of sides in a number of ways. So, in fact, is the notion of analyz
ing anything at all by any binary opposition. The issues touch very
closely on the potentialities of computers, and on their interrelations
with various aspects of human activity, individual or social. I do not
presume to address pertinent matters of computer science, logic, psy
chology, metaphor, gender, race, etc. in the present context, but I
cannot resist quoting

John Cage, who died yesterday.

ismaintained here is the concept of pairs of opposites-.hav


black
and white, as itwere, and then composing with the
ing
of
these
play
opposites. One can then engage in all the games
that academic composition has led us to know how to play. One
can balance this with that, produce climaxes, and so on. I'm
afraid all I can say is that it doesn't interestme.
It doesn't seem to me to radically change the situation from
the familiar convention. It simply takes these new ways of work

What

ing and consolidates

remains

at home

with

them with the old knowledges,


one's

familiar

ideas

of

the

so that one

drama?of

the

play of the opposites. So, one wouldn't have to change one's


mind. Whereas, I thinkwe are in a more urgent situation, where
it is absolutely essential for us to change our minds funda
mentally. And in this sense, I could be likened to a fundamental
ist Protestant

preacher.

. . What
.
we need is a use of our Art which alters our lives?is
our
useful in
lives.We are familiarwith those plays of balance, so
they couldn't possibly do anything more to us, no matter how

novel theywere, than they have already done. "New wine in old
bottles." (Interview with Roger Reynolds, inContemporary Com
posers on Contemporary Musky ed. Elliott Schwartz and Barney
Childs (New York: Holt, Rinehart andWinston, 1967), 345-6.)

the
The reader may gauge my own attitudes?or
ambivalences?by
material I have written in the present Part of this paper, and by the
material in the present note. And by my having written such a note.
It seems manifest to me that some binary oppositions are meaning
ful, and that others can be made so in certain contexts. It also seems
silly and destructive to me to worship that notion. A certain ambiva

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions

ProjectiveGeometry 63

lence feels comfortable to me in this context, or perhaps a Zen


"mu." If one refuses to value binary choices on principle, then one
cannot

value

one's

own

binary

choice

not

to value

choices.

One

can

deny being interested, as Cage does; that is different from making a


he comes very close to courting such a danger in the
choice?though
quoted passage. Or one can simply refuse to engage in dispute,
which (at present) seems advisable to me.
can be the same. If the CHANGE
two consecutive CHANGES
+
from STATE N to STATE N
1) were the same as the CHANGE
(

23. No

fromSTATE (N+ 1) to STATE (2f+ 2), thenSTATE (N + 2) would

have to be the same as STATE N. (We shall see why presently.) But
that cannot happen here, since Babbitt's STATE-row does not con
tain any duplicate STATES.
To see why STATE (N + 2) would have to be the same as STATE
AT in the case above, we can recall that the CHANGE
from STATE
+
N to STATE
is
the
(N
1)
given by summing
digits of the two

STATES, digitby digit.Likewise theCHANGE fromSTATE (N +


(N + 2) is given by summing the digits of the two
1) to STATE
were the same,
STATES, digit by digit. So if the two CHANGES

thenwe couldwrite "(STATE N) + (STATE (N+l))

(STATE (N

+ 1)) + (STATE (N + 2))," where the plus sign symbolizes addition


mod 2, digit by digit. In this algebraic situation, we could cancel the
summand "(STATE
(N + 1))" from both sides of the equation, and
infer that STATE N = STATE (N + 2).

This content downloaded from 169.229.11.216 on Wed, 21 Oct 2015 00:44:16 UTC
All use subject to JSTOR Terms and Conditions