You are on page 1of 8

Paradigms and Computer Music

Author(s): Andrew Gerzso


Source: Leonardo Music Journal, Vol. 2, No. 1 (1992), pp. 73-79
Published by: The MIT Press
Stable URL: http://www.jstor.org/stable/1513212
Accessed: 04-09-2016 14:27 UTC

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.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to Leonardo Music
Journal

This content downloaded from 5.51.164.112 on Sun, 04 Sep 2016 14:27:52 UTC
All use subject to http://about.jstor.org/terms
TH EORETICAL PERS PE CTIVE

Paradigms and Computer Music

Andrew Gerzso

omputer music is about 25 years old. A lations that we want to make on


great number of ideas have come out of the music-research those objects. What do I mean
community, some of which have found their way into the by objects and manipulations?
practical applications of commercial products as well. I For example, with a word
would like to examine some of the ideas underlying the processor, the objects are char-
machines and software (mostly the latter) used in music, by acters and words and the ma-
tracing where these ideas came from, pointing out some nipulations are copying, delet-
of the conceptual problems that experience has revealed ing, correcting the spelling,
A B S T R A C T
and, finally, making some suggestions for the future. In changing the font, etc. In mu-
particular I will discuss a family of systems that have been sic, the objects can be notes or
Any system used for making
used in the research community at large (the Music N chords that can be transposed, music embodies a particular way
languages [1]), two software systems that were developed edited, played, etc. The objects of looking at music. These ways
at the Institut de Recherche et Coordination Acoustique can also be sound tracks thatwe are called paradigms. The author
defines the general notions of
/Musique (IRCAM) (Max and Patchwork) [2] and the can manipulate, by changing
paradigm and representation and
sequencer, which is one of the main software tools used in the order, for example. They
explains how these concepts apply
the commercial music business [3]. can also be sound files that can to the design of computer-music
The systems I discuss below were all produced by people be manipulated by mixing, fil- systems. Several real systems are
of exceptional talent and imagination. Their accomplish- tering and so on. examined from the paradigmatic
point of view.
ments are milestones in the field of computer music. How- All the objects that we are
ever, pointing out problems (which is always easy to do with manipulating are represented
hindsight) is one of the ways to develop better systems for in some form or another by the
music in the future. software we are using. The question of representation, then,
is central to the design of any software system. What is it that
we want to represent? How do we want to manipulate what
is represented? These are the basic questions involved in the
PARADIGMS OF REPRESENTATION design of any software system.

Any time we use a system for making music, be it software, Computer music is not computer science, but many of
hardware or a combination of the two, we are dealing with the concepts and tools that are currently used in computer
an object that implicitly or explicitly embodies a paradigm music depend on the ideas that have been developed in
for making music. What do I mean by a paradigm? computer science. This is especially true of the concept of
In its most intuitive meaning, a paradigm is a guiding representation. If we are going to represent something in a
conceptual model for how something is thought to exist or computer, the way we do this depends on our concept of
behave. Therefore, the paradigm underlying a musical sys- representation. So, in fact, we are talking about the para-
tem is a model of what music is thought to be, the way it digms of representation. It would seem appropriate, there-
behaves and how it is created. The paradigm can be quite fore, to quickly summarize how the paradigm of repre-
(even excessively) loose or quite constraining. According to sentation has evolved over time.

Webster's Dictionary, a paradigm is "an outstandingly clear Recent years have seen the rise of a new discipline called
or typical example or archetype" [4] . But it is Thomas Kuhn cognitive science [6]. In fact, it is more a collection of
who has given a definition that is closer to our concerns disciplines with common goals than a discipline in itself.
here. He defines paradigms as "universally recognized sci- According to an interesting recent account by Varela,
entific achievements that for a time provide model problems Thompson and Rosch [7], thought about cognitivism has
and solutions to a community of practitioners" [5]. What is already gone through two stages and we are now on the
interesting here is the notion of 'model . . . solutions' that threshold of the third. What is not clear in this account is
are 'valid for a time'. Some of the systems described below, whether each stage replaces or builds upon the previous
in particular Music V, have been model solutions for many one. In any case, central to this evolution is the debate
years. concerning the paradigm for representation itself.
A mature musical system reflects what it is we want to
accomplish, so in a very direct way the software we use Andrew Gerzso (computer music researcher), IRCAM, 31, rue Saint-Merri, F-75004
Paris, France.
reflects what it is we want to do. How does this reflection
Manuscript solicited by Marc Battier.
take place? The software presents in some form or another
Received 28July 1992.
the objects we want to manipulate and the kinds of manipu-

K) 1992 ISAST
Pergamon Press Ltd. Printed in Great Britain.
0961-1 215/92 $5.00+0.00 LEONARDO MUSICJOURNAL, Vol. 2, No. 1, pp. 7>79, 1992 73

This content downloaded from 5.51.164.112 on Sun, 04 Sep 2016 14:27:52 UTC
All use subject to http://about.jstor.org/terms
[the processes] enact a world as a
. , . . . . . .

domam ot ( .lstlnctlons that 1S lnSepara-


Fig. 1. (bottom) A
ble from the structure embodied by
small Music V
P6 P7 / the cognitive system" [15] (emphasis
\ program, with
\ (upper left) a mine). This third stage is still in a very

1 0/l 20
\ * graphicpatchrep-

491 51 1 resentation, which


speculative state.
I have strayed a little here, but my
SC I will produce a two-
purpose was to clarify the relationship
note musical
between paradigms for music and para-
phrase at tempo
digms for representation. In computer
1 60 to the quarter
B2 J= 60 note. F1 isafunc- music we attempt to create paradigms
Q Z tionthatdescribes that will guide the design of systems in
T jr .t the evolution of which we try to represent the objects
the amplitude, or
that we are interested in manipulating.
F2
loudness, of the
OSC + The techniques used for representa-
sound over time.
tion are grounded in ideas about repre-
F2 is the function
that specifies the sentation itself. At the same time, in the
timbre of the background so to speak, the paradigm
51 1 sound. for representation is evolving through
the efforts of computer science and the

l
OUT ) cognitive sciences. Representation in
the cognitive sciences has, then, gone
through the three phases discussed
above: the symbolic, the emergent and
the enactive.

1 INSO 1;
2 OSC P5 P6 B2 F1 F1 P30;
WHAT PARADIGMS
3 OSC B2 P7 B2 F2 P29; DO WE USE?
4 OUTB2B1;
5 END;
6 GEN O 1 1 0 0 .99 20 .99 491 0 511 The Music N Family
7 GEN O 1 2 0 0.99 50.99 205 -.99 ' 506 -.99 461 0 51 1; The field of computer music was first
8 NOT O 1 2 1000 .0128 6.70;
concerned with making sound. Max
9 NOT 2 1 1 1000 .0256 8.44;
10TER 3;
Mathews's Music V (circa 1960) was the
first language for programming sounds.
Later,John Chowning invented the fre-
In the first stage of cognitivism, the of highly interconnected simple com-
quency modulation (FM) synthesis tech-
key concept is the symbol. "The central ponents are the objects themselves.
nique (circa 1970) for the efficient cal-
intuition behind cognitivism is that in- The objects that are represented come
culation of rich musical timbres. In
telligence . . . so resembles computa- into being by themselves (or emerge)
order to meet the demands that music
tion in its essential characteristics that through interaction with the environ-
makes on processing, Peter Samson of
cognition can actually be defined as ment of the perceiver, rather than be-
Systems Concepts created the Samson
computations of symbolic representa- ing objects 'out there' that are then
Box for the Center for Computer Re-
tions" [8]. Cognition, therefore, is "in- represented (or simply transcribed) 'in
search in Music and Acoustics (CCRMA)
formation processing as symbolic com- here'. The simple input/output scheme
at Stanford University, California, and
putation, [or] rule-based manipulation of cybernetics is abandoned. In com-
IRCAM, in Paris, developed several gene-
of symbols" [9]. This stage is exempli- puter music this tendency is reflected,
rations of real-time machines [ 16] . But
fied in many of the systems developed for example, in real-time performance
without a doubt it was Mathews's work
systems, which are highly interactive.
,^ .

tor computer muslc.


that has shaped at least three decades
In the second stage of cognitivism, In the third stage, the key concept is
of work in computer music.
the key concept is emergence. Here, that of enaction. "Brains make memo-
Central to Mathews's approach was
"the dominant theme shifts [away from ries, which change the ways we'll sub-
the paradigm of music as mathematical
the symbol] to the notion of emergent sequently think. The principal activities
function. Generally speaking, the sim-
properties" [ 10] . This approach to rep- of brains are making changes in them- plest function one can imagine is y = ( x),
resentation has given rise to the con- selves" [ 13] . This stage is characterized where the value of y is the result of the
nectionist strategies that are currently by the notion of processes that change action done by the function on the
being explored, of which the most spec- themselves. Fundamental to this ap- parameter x. In Mathews's text on Mu-
tacular example is the neural network proach is the attitude "that we move sic V we find: "Since the essence of
[11]. "The emergence of global states away from the idea of a world as inde- sound depends on the nature of the
in a network of simple components" pendent and extrinsic to the idea of a variations in pressure [in the air], we
becomes the primary behavioral char- world as inseparable from the structure will describe a sound wave by a pressure
acteristic of this approach [ 12] . Instead of these processes of self-modification" function p(t)" [17]. The unit gener-
of having objects that are explicitly rep- [ 14] . Thus, this approach inherits from ators of Music V can be viewed, then,
resented by symbols that then are used the concept of emergence. "Instead of as parameterized functions that, when
in calculation, the emergent global states representing an independent world, linked together, make a sound (see

74 Ge7zso, Paradigms and Computer Music

This content downloaded from 5.51.164.112 on Sun, 04 Sep 2016 14:27:52 UTC
All use subject to http://about.jstor.org/terms
MusFiforg..ithec2.from.Schema
.tractk_ic| a-yL
perform-
Fig. 1). The linkage itself is based on mathematical function itself but also in equal footing from a musical point of
the paradigm of the telephone switch- the way a traditional musical score is view. There is a parameter for pitch and
boardS which uses patch cords to link understood. For example, consider a another for timbre. Moreover, both
together people who want to talk. The quarter note played at a tempo of 60 to categories (pitch and timbre) are im-
patch paradigm has been applied in the quarter note. According to current plicitly assumed to correspond to a
innumerable hardware and software pr(} professional standards we need 44,100 'scale' or continuum of some sort. This
jects [18], and Music V led to the family numbers (called samples) per channel is true for pitch, of course, but not for
of Music N languages created during to represent this second of sound. The timbre. In fact, a paradigm for timbre
the 1970s and 1980s. All these systems pitch of this note can be represented that would allow us to establish some
were extensively used at IRCAM, with one number, say 440, for example. sort of classification method for musi-
CCRMA, Center for Music Experiments, So, the generator takes one parameter, cal purposes still does not exist. Increas-
University of California, San Diego, the pitch number, and turns the result ingly, timbre is being described intui-
Massachusetts Institute of Technology into 1 sec of sound that is 44)100 num- tively by musicians in terms of types of

(MIT) and hundreds of other places, in bers long. But does this produce a sat- processes or types of behavior [21 ] . For
a multitude of artistic projects. isfying result? We now know that a example, we can analyse sounds and
Intuitively, Mathews's functional para- sound that is musically rich is made up look for periodic (as opposed to cha-

digm views music as a constant flow of of many frequencies involving tiny vari- otic) behavior [22]. When a more spe-
data (due to the emphasis put on sig- ations over time. These variations must cific and complete description of these
nal-processing aspects), instead of as a be specified in some way. Therefore, types emerges we will be closer to a
series of events in time. The unit gener- the number of parameters needed to better way to represent timbre.

ators create this flow as a function of the represent the sound then starts to ap- What paradigm of representation does

parameters given to them. Note that proach the number of samples in the Music V correspond to, assumingVarelaSs

events are created by starting and stop- signal, and the dividing line between analysis is correct? Without a doubt, the

ping the flow. In Music 10, for example, generation and control becomes more one corresponding to the first phase,

we find two kinds of variables: I_TIME and more fuzzy. since Music V processes a continuous

and R_TIME. The first corresponds to Still another problem is that all the flow of symbols the samples of sounds

variables to which we assign values only parameters are implicitly put on an themselves.

when a note begins. All the variables of


this kind are grouped together in a block
of code called 'I_C)NLY code' [19].
The second corresponds to variables
that are modified when each sample of
diagram of setup t7 <-r
sound is calculated. The purpose of
I_TIME and R_TIME variables is to
ance Dialogue de | \ /
make the distinction between context I 'Ombre Dauble by I | #
and flow, respectively. The need for the Pierre Boulez. |
I_C)NLY code was one of the first signs
1 of the tapeis | * track 1 > 2 2 ' 2 2 S Del
of the shortcomings of the pure flow
sent to the VCA
approach.
control unit na * track 2 SMP1 E code
I have already pointed out that the -

tne aucuo con-


paradigm of music as mathematical sole. The music is 1-
function excludes from the beginning then spatialised
the concept of musical event. The evalu- among six speak- |;

ation of a mathematical function itself ers using both the


timing informa- | ,,
does not even imply time. The function
tion coming from
is evaluated that is all. But discrete
track 2 of the
events are important in music and, tape recorder and -T
more generally, in any context where the spatial con- | audio l l
| input l l
the question of creating a language of figuratlons memo- UCR CONTBOL U

expression is a concern. In Charles audlo


control unit. 1 output
Hockett's interesting analysis of the
prerequisites for having a language of
any kind, discreteness (in the mathe-
matical sense of a clearly delimited
thing) is extremely important [20]. This
discreteness allows for the creation of a
differentiated vocabulary and for the
possibility of a grammar through the
combination of discrete objects.
Another problem is that this ap-
proach explicitly makes a distinction
between the generation of a sound (via
the mathematical function) and its con-
trol (via the parameters). This distinc-
tion has its roots in the nature of the

Gerzso, Paradigms and Computer Music 75

This content downloaded from 5.51.164.112 on Sun, 04 Sep 2016 14:27:52 UTC
All use subject to http://about.jstor.org/terms
lWRoman
-I 1 : =0-.>= X:0X-: = -->-DOD :_3 ___
U2 -a?. : s=.sxb,m;
_ _!=|."=W_1_,I,/_X.P.J.
tently travel along the patch lines.
Instead of having a constant flow of

bang reset
CUT MiDi 1 2 3 4 5 6 7 ON OFF information through the network of
D1 - EDIED113:>11321C3ED31:3
patch lines, as1:1
in theEDl
Music N family,
patcher } patcher patcher Vv | - /
patcher LOCAL sourse
_
information flows only when a message
1:2 C1
..

patcher MiDi source RQ


;
is sent [25]. Another major difference
>0
_W*=.
is that in Max a large variety of messages
Fer J 1 4-5 ++
Egtcher ,Jsigle initlal
(numbers, lists, 'bangs' [trigger mes-
t1 /74 0
sages], etc. ) travel between operators,
-

r space-ss t /89 21
=
whereas in the Music Nfamily it is num-
Btther | T 1- *+
mcher | T 5-6 + + bers representing signals (air-pressure
t /73 :

on off
t3 /57 Cj
variations) that travel between the
operators.
T2=3 Xie Finai
Wtcher | Sl9
One way to describe Max is to say that
L L] 23/68 [: it is a patch language with delays. If
Fatcher envelope tontrol
q
p3 Space-cs !
patcher | space

=-- - : 1 - ro
tont I
there were no delays, then everything
t REF IX
Eginpace-tabl j would be executed as quickly as possi-
Dialosue de liombre double (roman) >>>>>>>>> a <<<(<<<<e ble. The delays are either explicitly pro-
4 .X*89
Ltcher 71 l 3-4
patcher | grammed or induced by the system when
iZ5o b
_ - _
1 ** set reArence vatues
2 ** choose section it is waiting for an external signal to
T 1-2 T 4=S T 5-6
C 1 , w proceed. The temporal structure of the
1Qm
musical emerges as a result of these
Fig. 3. The Max program for D?alogue de l 'Ombre Double by Pierre Boulez. Each one of the delays.
little boxes labeled 'patcher' contains subprograms that control the sound spatialisation.
I have had two occasions for using
The square items with the circles inside are called 'bangs', which are trigger buttons for
Max in the context of a musical produc-
setting off a subprogram. The long vertical rectangles are potentiometers, which are used
interactively to set loudness or speed. tion [26]. The first was in the realiza-
tion of Dialogue de L Vmbre Double
(1985) by the French composer Pierre
Max in Max is done graphically. Max has
Boulez, in which a live clarinet dia-
The Max programming language, been extremely successful in both the
logues with a clarinet on tape that is
named in honor of Max Mathews, was research and commercial music com-
spatialised using an audio matrix [27]
initially conceived as a programming munities [24].
controlled by a Max program (Figs 2
environment for creating real-time mu- Max is based on the idea of objects
that send each other messages. Max and 3). The tape has two tracks. The
sical applications. The most common
interacts with the outside world by first contains the recorded music that is
applications output MIDI data for con-
accepting and sending MIDI informa- to be spatialised, and the second con-
trolling synthesizers. In its most recent
tion. The objects, which are themselves tains SMPTE code that is sent to the
version, which runs on the IRCAM Mu-
operators or functions made up of Max program.
sical Workstation [23], it operates not
Max operators, are interconnected in a The second composition, also by
only at the control level but at the sam-
ple level as well. All programming patch-like fashion. Messages intermit- Boulez, was Explosante-Fixe (1991) for
chamber ensemble, solo MIDI flute [28],
digital signal processor [29], audio ma-
trix and samplers [30] (Fig. 4). Here
the solo flute sends MIDI information
Fig. 4. Schematic
diagram of equip- to a Max program containing a score
ment setup tor a follower (Fig. 5). The score follower
Sampler A a
_ / performance of determines where the flutist is in the
S a m p 1 e r B # r,xXblosante-rtxe Dy score and triggers events (sound trans-
_ # Pierre Boulez. The
formations, spatialisations or sampling)
# # notes from the solo
/ # flutist are sent to a
associated with particular notes in the

# # Maxprogram con- score. The main advantage of score-


# # taining the score fol- following resides in the accurate musi-
> t lower. Depending cal synchrony that is made possible. An
Arrssosz>sovfossif>> on the actions to be
example of the burden put on this sys-
taken at each cue in
tem by Explosante-Fixe is that, in the
the score, the score
space of 6 min, the flutist triggers some
follower then issues
' commands for the 250 cues.
# control of the sam- Two related problems were encoun-
# plers, spatialisation tered during these two production pro-
0 (Matrix32) or
jects, particularly during Explosante-Fixe.
# sound transforma-
The first concerns what I call the
- tion (second Max
program control- 'state problem'. At any given point in
# ling the 4X real- time, one cannot determine what state
Control # time digital signal a particular Max program is in. Since
processor). the overall paradigm stresses message
sound: >>zAw

flow and continuity (as opposed to

76 Gerzso, Paradigms and Computer Music

This content downloaded from 5.51.164.112 on Sun, 04 Sep 2016 14:27:52 UTC
All use subject to http://about.jstor.org/terms
states), it is not surprising that we should
Time
encounter this problem. No provision
. , . . . Fig. 5. Schematic
exlsts tor
diagram of the
ticular
score follower. st
notes
troublesome whenever the initial state Notes coming in
played by the in- from f 1 ute
of a program is important. Practically
strumentalist are
speaking, I solved the problem by writ-
. . . . . . . . compared by the
ng extenslve lnltla lzatlon code. In
score follower
the commercial version of Max, an ad program with the
hoc solution was found in the form of score stored in
'presets' [31]. the computer

!
Related to the state problem is the memory. The sys-
tem must 'look'
problem regarding context. If one sim-
to the left and t; + fl , t t t t
ply plays a composition from start to
right in the score
finish, there is no problem. But making to catch mistakes
music involves rehearsals that, in turn, on the part of the notes stored
require starting at different places in instrumentalist. ln memory
the score of a piece. Most music is made
up of a number of processes that over-
lap in time, so starting at a particular
piler outside of Max. One has to be a a LISP function always returns one
point in a piece requires knowing what
programming expert to handle these value). The outputs offunctions can be
state one is in at that point, as well as
code objects easily. connected to the inputs of other func-
what context one is starting in. We need
One could argue that Max embodies tions via patch lines, hence the name
to have a simple mechanism for deter-
the symbolic paradigm of representa- Patchwork. We can already see that there
mining what processes were already ac-
tion. Like Music V and its children, Max is also a strong conceptual tie between
tive and what processes should be in-
processes a flow of symbols, although Patchwork and LISP.
itialized and started.
the vocabulary of symbols has become Patchwork, especially the Esquisse
The context problem is found again much larger since it can include both Musical Toolbox, is useful as long as
in the 'follow' module that is responsi- samples at the level of sound produc- one uses a functional approach to the
ble for score-following in Max. As I have tion and more complex messages at the generation of music. A simple example
pointed out, this module was absolutely control level. But in Max one can also of the functional approach is chord
crucial to the overall strategy of having very easily create musical automatons, transposition. The three parameters
the flutist be master of all the different for example, whose interesting behav-
given to the function are the initial
types of events triggered in real timeS ior is a priori quite unpredictable. In chord, the interval and direction of
and it worked extremelywell. However, this respect Max embodies, in a weak transposition (up or down). The result
the score follower looks for notes in a sense, the emergent paradigm as well. is the transposed chord. A more sophis-
list without taking into consideration
ticated example is chord interpolation.
tempo or metric context [32]. The main
Patchwork The parameters here are the two chords
advantage of having such a context sen-
Patchwork is a programming environ- between which one will interpolate and
sitivity would be that prediction would
ment that resides in a Common LISP the number of steps in the interpola-
become possible. A system's knowledge
interpreter. It was designed as a toolbox tion. At each step the function calcu-
of the tempo and the current rhythmic
for computer-aided musical composi- lates a chord whose notes are obtained
value would be sufficient for predicting
tion. Patchwork can be viewed at several by interpolating between correspond-
when the next event is most likely to
levels. At the most elementary level ing notes in the two initial chords. But
happen. Preparation (for sending large
Patchwork can be used as an environ- what happens when the problem to be
amounts of data or for calculating re-
ment for programming graphically in solved cannot be handled by the func-
sults just before they are needed) is
Common LISP. (As opposed to the situ- tional paradigm? A typical request on
then possible, instead of the mad dash
ation in Max, going back and forth be- the part of composers involves the use
that results from the trigger strategy.
tween LISP and Patchwork is very easy.) of constraints. For example, perhaps
Max also maintains the sound
At a higher level, which makes use of the one would like to obtain a family or
production/sound control dichotomy
Esquisse Musical Toolbox [33], the user a collection of chords that obey con-
found in the Music N languages This
can create compositional algorithms straints such as interval content, regis-
is true in whatever version of Max one
that produce musical material destined ter or density. Programming experts
uses, be it the Macintosh version or the
for score sketches or that can be used will point out that constraint program-
IRCAM/NeXT version.
for creating data files that will be input ming (which is derived from the para-
One final point has to do with the
into a sound synthesis program, for ex- digm of logic programming [34] ) can
unavailability of an ordinary program-
ample (see Fig. 6). From these algo-
be done in LISP very easily. This is true,
ming language (such as C, Basic, LISP, rithms one can easily create entire per-of course (the experts are always right),
etc.) easily accessible from Max. The sonalized libraries of musical functions.
but it avoids the real issue. The impor-
present strategy (in the version for theThe functions are represented on the tant point is that when programming
Macintosh) forces the user to create screen as little boxes that contain a is needed, the environment should
'code resources' (a concept special to number of inputs (corresponding to provide a toolbox of directly accessible
Apple Computer), which are chunks of the parameters of the function) and programming paradigms (functional
executable code generated by a com- one output (a result of the fact that paradigms, constraint paradigms, etc.) .

Gerzso, Paradigms and Computer Music 77

This content downloaded from 5.51.164.112 on Sun, 04 Sep 2016 14:27:52 UTC
All use subject to http://about.jstor.org/terms
MN6 o ER o standard music
Because of the strong tie between
cies cropping up it is probably because

of a shift of interest towards other areas,


value to

tereste
the composer who is more in-

Patchwork and I,ISP, Patchwork corre- extrapo


such as the act of musical composition
sponds to the first of Varela's three proach
itself.
P zases. This is not surprising, since LISP
it is to
Despite the fact that composition is
is a language that stresses symbolic cal- assume
a highly personal affair, composers are
culation. LISP, by the way, stands discove
realizing that common concerns do ex-
for LISt Processing and, by implication, scientis
ist. The challenge is to formalize this
the processing of symbols in a list. tromag
common ground. Early efforts at
My survey of systems thus far reveals once an
formalizing musical composition have
a heavy dependence on the patch, or time. T
taken traditional music or, rather, mu-
functional, paradigm. What is remark- not get
sical history, as a sort of progress to be
able is that this paradigm has been use- useful
formalized. The basic idea is that one
ful for so long in so many different incorp
first formalizes early counterpoint,
contexts. But this dependence under- musica
then harmony, then fugues, etc., until
lines again the fact that the first few musica
the present day is reached. While at a
decades of computer music were char- graphi
certain level this can be an interesting
acterized by the search for systems for ments
exercise for musicologists, it is of little
making sound. If we now feel inadequa- possibi
turing
:

and, fi
Fig. 6. (left) An the com
example of a
sound-
Patchwork pro-
gram of an inter-
polation betwe n
the chord on the The S
I l (upper) left and Sequen
00 the one note cal pro
chord on the
(upper) right. The The pa
interpolation is on the
calculated by the prising
. module labeled cordin
'interchord'. (bot- music
c h o r d " tom) The result in approa
notation is shown.
that c
which
with i
drives
sizers
etc.), t
corded
tape re
The s
ventio
orche
score a
The om
sical f
are sy
real-t
sent. A
chordllde
X I s |
I
sical f
lFurs llvels I one pa

r Z-
ones i
From
it wou
local m

k "t 111t' 1S+' ti° i S; b; 1


78 Gso Paradigms

This content downloaded from 5.51.164.112 on Sun, 04 Sep 2016 14:27:52 UTC
All use subject to http://about.jstor.org/terms
by calling on a personalized toolbox of Lastly, general systems that try to do 18. In Donald Buchla's analog systems, modules
r * , ^ . . . , , . are patched together. The patch languages 4CED
IUnCtlOnS. . llS lmp les detlnlng a strat- eveIything should be forgotten and
and CPAT were designed for IRCAM's 4C and 4X
egy for importing functions from out- replaced by communities of programs processors by Curtis Abbott and myself, respec-
side of the sequencer. A good model to that can live independently and at the tively. It is interesting to note that Miller Puckette's
Max was initially called Patcher.
follow is HyperCard. If the user finds same time cooperate and interact with
intuitive programming with Hypertalk each other. Aboare all, computer music 19. Marc Battier, "Le langage de synthese MUSIC
10", IRCAM internal document, 1980, pp. 2-6. See
inadequate, then he or she can create should quit living on borrowed para- also Tovar, "The Music Manual", CCRMA internal
functions using languages such as C or digms. document.

PASCAL and import them as external 20. Charles Hockett, "The Problem of Universals
commands (XCMDs). There is a stand- References and Notes in l anguage" inJoseph Greenberg, ed., Universals
of Language (Cambridge, MA: MIT Press, 1963) pp.
ard interface designed so that the 1. Music V was invented circa 1960 by Max
1-29.
XCMDs can communicate easily with Mathews. The generations of languages that fol-
lowed include Music 4BF by Hubert S. Howe, Jr., 21. The notion of process is quite present in the
the HyperCard internal data structures,
Music 10 by Tovar, Csound by BarryVercoe, 4CED minds of French composers. In a recent lecture at
which are invisible to the user. by Curtis Abbott, CPAT by Andrew Gerzso, and the College de France, Pierre Boulez suggested that
others. composition should be viewed as the organization
Now, of courseS tape recorders are
of musical processes. The notion of process on the
disappearing and once they are re- 2. IRCAM is part of the Centre Georges Pompidou
level of sound production is a major preoccupation
art center in Paris, France. Max was invented
placed by hard-disk systems or some- of composers such as Tristan Murail and Marc-An-
by Miller Puckette at IRCAM. Sold since 1991 as
dre Dalbavie (both of France).
thing else it will be interesting to see a commercial product by Opcode Systems (Palo
how the sequencer paradigm changes. Alto, California), Max runs on Apple computers.
22. This approach to sound analysis and re-synthe-
Another version of Max runs on the IRCAM Music sis was developed by Xavier Serra at CCRMA.
No doubt the track concept will disap- Workstation, which uses the NeXT computer.
pear and be replaced by something less Patchwork is the result of an international project 23. The IRCDAM Musical Workstation is built on the
principally involving Mikael Laurson (the inven- NeXT computer, which can contain up to three
rigid and more fragment-like in nature.
tor) of Finland and Jacques Duthen and Camilo IRCAM digital signal-processing cards, each of
Interaction (on the control or signal Rueda at IRCAM. I managed this project in its last which makes use of two Intel i860 chips.
level) between tracks might be interest- phase of development, which led to an exportable
version of Patchwork in Common LISP. 24. Since it appeared on the market in the early
ing. A toolkit for building sequencers, part of 1991, dozens of musical applications have
3. The most widely used sequencers are Studio been written.
which would include tools for writing
Vision by Opcode Systems; Digital Performer by
and generating music (using a local Mark of the Unicorn; and Cubase by Steinberg. All 25. This is true only of the version of Max that does
programming language), might be three run on the Macintosh family of computers not generate sound signals (which are continu-
from Apple Computer. ous) .
more useful. Dynamic 'orchestras' are
4. Websters Ninth New Collegiate Dictionaty 26. I used the pre-commercial version of Max
another possibility.
(Springfield, MA: Merriam-Webster, 1984). called Patcher.

5. Thomas S. Kuhn, The Structure of Scientific RerJF 27. The IRCAIW MATRIX32 was designed by Michel
lutions (Chicago IL: Univ. of Chicago Press, 1962) Starkier and Didier Roncin.
SOME MUSTS p. Vii.

28. The MIDI flute is an IRCAM invention by


6. According to Webster's International Diction-
FOR THE FUTURE ary, 'cognition' is "the act or process of knowing
Michel Starkier. Information from the key posi-
tions and a microphone placed near the em-
includiIlg both awareness and judgment".
Computer music needs new paradigms. bouchere is turned into MIDI note-on/note-off
7. Francisco Varela, Evan Thompson and Eleanor messages.
Time, context, state, musical writing
Rosch, The Embodied Mind (Cambridge, MA: MIT
(as opposed to sound generation), dy- 29. The IRCAM 4X processor.
Press, 1991 ) .

namic musical languages, toolboxes of 8. Varela, Thompson and Rosch [7] p. 40.
30. The AKAI S1000.

computational paradigms, and the con- 31. This and many other imaginative and practical
9. Varela, Thompson and Rosch [7] p. 42.
struction and reconstruction of repre- features were invented by David Zicarelli.
10. Varela, Thompson and Rosch [7] p. 87.
sentations are but a few of the impor- 32. This was achieved in earlier work done at IR-
11. Neural networks are being applied to many
tant concerns for the future. areas such as character recognition, bird-song
CAM on score-following, principally by the flutist
Lawrence Beauregard of the Ensemble InterCon-
Important to these concerns are what
. . . .

earnlng, pltC z detectlon, etc.


temporain (with the help of Xavier Chabot) and
Marvin Minsky calls "Level-Bands", 12. Varela, Thompson and Rosch [7] p. 99. Barry Vercoe (inventor of Csound) of MIT.
which involve making "strong connec- 13. Marvin Minsky, Society of Mind (London: Heine- 33. This is the result of the collaboration of a
tions at a certain level of detail . . . [and mann, 1985) p. 288. number of people, principally Pierre-Francois Bais-
making] weaker connections at higher nee, Jean-Baptiste Barriere, and the composers
14. Varela, Thompson and Rosch [7] p. 139.
Marc-Andre Dalbavie, Magnus Lindberg and Kaija
and lower levels" [36] (emphases mine). 15. Varela, Thompson and Rosch [7] p. 140. Saariaho.
Too often in computer music, systems 16. The 4A (1976), 4B, 4C and 4X (1980) real- 34. PROLOG is an example of a language using
are provided from which-in princi- time digital signal processors were designed by this paradigm.
ple-more complex systems can be cre- Giuseppe DiGiugno at IRCAM. The IRCAM Musi-
cal Workstation ( 1991 ) was created by a team headed 35. The basic ideas of this approach are found in
ated. In realityS the complex level is sim- by Eric Lindemann. Umberto Eco's L'Oeavre Ouverte.
ply never achieved7 because the level of
17. Max V. Mathews, The Technology of Computer 36. Marvin Minsky, Society of Mind (London: Heine-
detail one starts with is simply too low. Music (Cambridge, MA: MIT Press, 1969) p. 2. mann, 1985) p. 86.

Cerzso, Paradigms and Computer Music 79

This content downloaded from 5.51.164.112 on Sun, 04 Sep 2016 14:27:52 UTC
All use subject to http://about.jstor.org/terms

You might also like