Some Reflections on the Nature of the Landscape within Which Systems are Designed Author(s): Joel Chadabe Source: Computer Music Journal, Vol. 1, No. 3 (Jun., 1977), pp. 5-11 Published by: The MIT Press Stable URL: http://www.jstor.org/stable/3679605 . Accessed: 21/03/2011 21:55

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=mitpress. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

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

http://www.jstor.org Some Reflections on the Nature of the Landscape

within which Computer Music Systems are Designed

Joel Chadabe Departmentof Music State Universityof New Yorkat Albany Copyright@ 1977 by JoelChadabe

INTRODUCTION generalconsiderations in system designthat mightbe mentionedat this time. Butbefore designing a system, In the broadestway, a systems-orientedattitude which is essentiallythe work of creatingan organiza- is characterizedby particularsensitivities towards: tion of softwareand hardwaremodules, it is essential (1) dynamicfunctioning as a transformationprocess to understandhow the systemis intendedto function. of some type, wherebyan input is acted upon to pro- The choices for functioningand theirimplications in duce an output; and(2) organizationas a multivariable organizationare the subjectof this brief essay. complexity of subsystemswhich arelinked to form the whole system. The functioningof the system is directlyrelated to its organization.In fact, one might THEFUNCTIONING OF THE SYSTEM define a system as a whole which functionsas it does becauseof the way its partsare organized. Let us take, as a model for generatingsound, a By the term "computermusic," I mean music versionof the classicalfeedback system as drawnin that is producedby a hardwareand softwaresystem Fig. 1. Themodel contains: (1) an input, by means that importantlyincludes a computer,but may also of which the composercontrols the system; (2) a include other analogor digitalsynthesis equipment. processingsection, which uses the input to produce The numerouspossibilities for computermusic system the output; (3) an output, which is the music as design,which result from the decreasingcost and con- sound; and (4) a feedbackloop, by meansof which sequenteasy availabilityof computersystem compo- the actualoutput can be comparedby the composer nents, makesthe questionof what type of system to with what was expected. designan extremelycompelling one. Specificanswers I will first describethe model as it representsthe to this questioncan be resolvedonly aftermuch more performanceof instrumentalmusic, as a point of researchin psychoacousticsis completed and after referencefor the followingdiscussion. The input is composershave had a chanceto workwith equipment musicalnotation, a score. The processingsection presentlybeing built. There are, however,certain representswhatever is necessaryto "transform"the

Composer Processing Output Input Section

FeedbackLoop

Figure 1. A model for sound generationbased on classicalfeedback theory.

Joel Chadabe: Some Reflections on the Nature of the Landscape within which Computer Music Systems are Designed Page 5 input into sound, and in this context it might include a keyboard; (2) other characteristicsof the notes, social activities - such as getting to know performers - such as dynamics,attacks and phrasing,are entered as well as technicalconsiderations. The output is the and adjustedline by line; then (3) all the voices are music as sound resultingfrom the performance.The playedback together,and a performer,using various feedbackloop is the meansby which the composer input devices, "conducts"the performanceby con- hears the music and comparesit with what was trollingoverall dynamics, tempos and balance. The expected. Since composers,especially unknown primarycharacteristic of this type of system is that it composers,often wait a long timebefore hearingtheir has been designedto rememberand play back specific music, feedback,and consequentlylearning, is often data but to vary its playbackin certainrespects delayed. accordingto realtimeinputs from a performer. The model representssound generationwith Memoryautomation, whereby a composerspeci- computersystems in two different ways: (1) the fies data that is rememberedand playedback, is dis- computeris used as a performer,or (2) the computer tinct from"process automation," whereby a composer is used as a composer-performer.If the computeris specifiesa processwhich automatically generates data. used as a performer,nothing essential is changedfrom At least in the sense that it generatesdata from pre- the traditionalprocess of composing-then-performing viouslyinput rules, a processautomation system could instrumentalmusic, although with computersthe field be said to "compose."And if the data it produced of possibilitiesfor sound constructionand transfor- were unpredictable,the system would producesur- mation is greatlyexpanded. The composersubmits prisesand take on a "personality"of its own. One to the computera previouslycomposed score which early example of this type of procedureis Lejaren the computeraccepts as input data. The data is Hiller'sexperimentation which resultedin the Illiac enteredmore slowly than it will be playedback, but Suite (1957). Hillerused the "MonteCarlo" method, the computer"remembers" it and eventuallyplays it a generate-and-testprocedure wherein many random back at the correct tempo. Becausethe computer numberswere generated and then judged, and accepted playsback automaticallywhat it remembers,I shall or rejected,according to specific criteria. The num- call this procedure"memory automation." Thereare bers were used to representpitches, among other myriadexamples of this way of workingin the litera- musicalattributes, and the criteriafor theiracceptance ture: CharlesDodge's Earth's Magnetic Field is one. includedthe followingrules: no tritones; the melody It is not typicalof Dodge'srecent work, but it provides had to startand end on middleC; and the rangeof the a dramaticexample of this use of a computer. Not melody from its lowest to its highestnote could not only werethe pitchesnot composedby the computer, exceed one octave. Once a note was accepted,it was they werenot composedby the composer. They were placedin a note list which was eventuallyprinted out derivedfrom a chartof the Kp indices*of the earth's and transcribedby hand into musicalnotation to be magneticfield during1961, and subsequentlyentered played with traditionalinstruments. into the computerto be realizedas sound. The com- In a non-realtimesystem, feedbackhas a cogni- positionis the resultof a translationfrom one medium tive value. The composerlearns from the feedback, to another,reminiscent of Villa Lobos'piano piece but to correctthe output, the input must be changed New YorkSkyline for which the pitcheswere deter- and processedagain from the beginning. Whatthe mined accordingto the levels of the buildingtops in a composerlearns from one performanceis appliedto pictureof the New York skyline. Dodge'stranslation the next one. Thisis true of Hiller'sprocedures in the was much more demanding,since it involvedthe 50's, and also true of instrumentalmusic composition manipulationand programmingof the machinesthat and performanceand the use of the computeras a performedhis score and the composingof all the performer.In fact, it is true of any situationwhere qualitiesin the musicthat were not indicatedby the composition and performance do not occur simul- chart,such as the organizationof the pitch scheme taneously. But when the music is output as sound and the timbres. while the composing is taking place, the composer is A realtimeinteractive version of a memoryauto- able to regulatethe musicas it is being composed. In mation system is the ConductorProgram, developed realtime interactive systems, feedback has a regulatory by Max Mathewsat Bell TelephoneLaboratories. as well as a cognitive function, and, depending upon A realtimeinteractive system has an output that is the natureof the system - whetherit is a memoryor virtuallyimmediate in responseto its input, making process automation system -- either the performance it possible to regulate the system during operation. or the composition is regulated. The ConductorProgram functions in three stages: One example of a realtime interactive process (1) pitches and durationsare enterednote by note via automation system is my composition Echoes (1972). In a performance of Echoes, an instrumentalist plays *TheKp index, measured 8 times per day, represents the into a microphone from a score which consists of changesin Earth'smagnetic field. - Ed. precise pitches but which allows for flexibility in

Page 6 Computer Music Journal, Box E, Menlo Park, CA 94025 June, 1977 articulation,rhythm and speed. The sounds are Composingwith a realtimeinteractive process delayedfor a few seconds,then transformedin pitch automationsystem is an experience significantly and timbre and sent to one of severalloudspeakers differentfrom traditionalcomposing. To the degree placedaround the performancespace, from which they that a process automationsystem is active and un- are heard as distant echoes of what had originally predictablein its output, it producessomething in been played. The specific characterof each transfor- addition to what was input. Thus, the composer mation and the specificloudspeaker from which an "learns"by receivingnew informationin the form echo will be heardcannot be predicted becausethe of automaticallygenerated data, and the composer's informationwhich determinesthose processesis in regulationof the system's output occurs in an en- effect random. The sounds are an ambience,an vironmentof interactionand learning. This is dif- environment,within which the instrumentalistcon- ferent from typical instrumentalimprovisation tinues to play and to make performancedecisions. in that an improviserspecifies and performsdata The performer'sinput is a primarycontrol for the in reaction to another improviser,but here the system, but the system gives back somethingextra, composerspecifies and performscontrol strategies i.e. the randomtransformations of the input, with ratherthan data. In the same way that control which the performerinteracts. of the machinethat manufacturesthe product is An earlierexample of this type of systemis the more powerfulthan direct control over the details SalMarConstruction, first designedand built in 1970 of the product,control strategiesare more power- by SalvatoreMartirano in collaborationwith several ful than data specifications. The differencebetween engineersat the Universityof Illinois. It consistsof instrumentalimprovisation and control strategy a consolewhich contains digital and analog electronics interactionis, then, primarilythat of one's position and a control panel, and twenty-fourloudspeakers in a control heirarchy,but it is true that in both arrangedthroughout a performancespace. Four cases the performermust act and react in realtime. separateprograms of independentlyor interactively This is a fundamentallynew way of functioningfor changingsounds are generated,and each program a composer. It demandsperformance skill and travelsseparately in automatedroutings through the performance-orientedthinking, and it raisesquestions arrangementof loudspeakers.Martirano performs at as to how well a composercan deal with realtime the control panel which containsmore than 200 problems. Traditionalmethods of composingoffer touch-sensitiveswitches. The system continuously moreleisurely and thoughtful access to detail,but they and automaticallygenerates new data, which it do not offer realtimecontrol over composition. In- assemblesin varyingpatterns. The data is then used deed, it might be appropriateto ask if realtime to give each sound a differenttimbre, pitch, envelope, control over compositionand detailedspecification and loudness, so that every aspect of the sound, of data are in fact mutuallyexclusive procedures. includingits spatialrouting, is determinedby the One perspectiveof the problemwill be furtherelabo- automaticallygenerated data. Martirano'sperfor- ratedin a laterdiscussion of heirarchicalorganization, manceis a manipulationof the processin interaction but the point is that regardlessof the particular with what it produces,adjusting such parametersas attributesof any hardwaresystem, composerscannot rangeof variation,speed, octave placement,tuning think fast enoughto specify controlsin realtimefor system, and routing,as well as manualoverride func- quickly changingvariables. Thus, aspects of every tions such as stop, cycle, hold, and continue. realtimesystem will have to be automated. But must Processautomation systems pose a problemin the automationbe in the natureof a process? I have conceptualizingcomposition as a separateactivity presentedmemory automationand processautoma- becausesuch systems are generatingnew composi- tion as antipodalconcepts, and as an overalldescrip- tional data simultaneously with its performance. tion of different approachesto composition, it Further, every performance will be different in its probablyis a justifiablepresentation, but it is also true specifics. The music produced by these systems is that in any one system both memory and process identified more by the nature of its process than by automationsubsystems may be equally participant. its details. Since the details of process automation Memoryautomation subsystems could be part of a music will vary from performance to performance, realtimecomposing system in the form of stored one's attention shifts from the details for the process functions or other specified-in-detail"building itself and to the system which produces the process. blocks" which are assembledin realtimeby a com- Designing the system thus comes into focus as the poser, a proceduresomewhat akin to EarleBrown's first act of composition. It brings Yeats' words to practiceof "assembling"a compositionduring per- mind: "How can we know the dancer from the formance. That type of procedure- subsystems dance?" The system and the musical process it specified in detail and later assembled- can easily produces are related as organization is related to approach a substantialmultivariable complexity functioning. and unpredictability.

Joel Chadabe: Some Reflections on the Nature of the Landscape within which Computer Music Systems are Designed Page 7 The unpredictability with which a system oper- THEORGANIZATION ates can be measured by the extent to which the input OF THEPROCESSING SECTION is reflected in the output. Since unpredictability may result either from complexity or randomness in the The organization of the processing section can functioning of subsystems, distinctions between be described as a hierarchy. A hierarchy is a multi- memory and process automation may be less impor- level system of subsystems, a series of boxes within tant than the idea associated earlier with process boxes, where each box contains several smaller boxes. automation systems - that if an output is unpredic- It is the nature of hierarchies that each subsystem table, i.e. if it reflects something "added" to the input, communicates with subsystems at other levels only the composer composes by interaction with an auto- through adjacent levels above or below. Whateverthe mated system. "Unpredictable," to be sure, does not character of the individual subsystems, however they mean that the output from the system has no connec- are organized and function, and regardlessof whether tion with the input, because all parts of the system they are synchronized or non-synchronized, their function interactively, and if the input becomes relationships with other subsystems are determined noninfluential in shaping the output, the input will by their position in the hierarchy. And beyond a lose its meaning. And the overall point is that the use generalized description of hierarchical organization, of computers in intelligent composing systems offers we might differentiate between control hierarchies a methodology for composing that is substantively and structural hierarchies. In control hierarchies, new and potentially beneficial. the vertical alignments of levels function as channels The means by which the composer interacts for communication and influence, so that one level with the system is also important. The nature of a "causes" events to occur at a lower level. Structural particular input device and its associated performance hierarchiesare simply additive rankingsfrom the many gesture is a determinant in the functioning of the and/or fast at the lowest level to the few and/or slow system. There is a great variety of devices and ges- at the highest level. An example of a structural hier- tures that might be considered. Playing keyboards or archy might be the organization of a book as many footpedals, waving one's hands in front of a television letters, fewer words, fewer sentences, fewer para- camera, communicating brainwave activity via attach- graphs, fewer chapters, etc. In a control hierarchy, ments to one's head, dancing in a costume with mer- we assume that changes occur because of control, and cury switches attached to it, moving within a capaci- we have a ranking of controls from the lowest level to tance field, for examples, are all possible inputs. The the highest, with each ascending level exerting more choice of input is important to system functioning in influence over a larger number of lower-level controls. two respects: (1) each particular device suggests Sounds change simultaneously at many speeds, certain performance skills; and (2) the devices and the and thus can be described as structuralhierarchies. The gestures used to perform them have psychological, fastest changes occur in the millisecond range, what I artistic and social connotations that can influence the call "audio-time," and give us information about pitch, nature of the information conveyed by the perfor- timbre, loudness, and location, among other things. mance. The potential flexibility in system design that Slower changes, in the range of large fractions of a sec- derives from choice of input and, also, choice of a ond to many seconds, which I call "conscious-time," meaning for the input, might be clarified by the give us information about melodies or successions of following comparison with a piano. Although pianos sounds. We perceive audio-time changes and draw con- can be performed by striking the casing, strumming clusions about them, but we are not consciously aware or plucking the strings directly, vibrating the strings from listening to the sounds what structural aspects in sympathy with another sound, and perhaps in other of the sounds give us our impresssions. But we are unorthodox ways, the normal input to the piano is aware of the nature of the changesin conscious-time - via the keys. A pianist normally plays keys. The key in fact, musical training usually includes a conscious triggers a mechanism that strikes a string. Thus, there recognition and writing down of those changes as is one input device and it always functions in the same melodies, harmonies and rhythms, in dictation exer- way. With computer systems, any appropriate input cises. There is yet a slower rate of change, occurring device may be used and the input can function as in what I would call "longterm-time," in which the a control for any aspect of the system within the changes occur so slowly that we do not have a sense of context of the design of the system. From a system- connection between events. The hierarchy as it organization point of view, it is the processing sec- applies to analog is usually resolved tion that determines how an input will affect the into two levels, audio-time and conscious-time, or system in producing an output, and so from that what Donald Buchla calls "sound" and "structure." point of view the processing section more than the The resolution could be more finely drawn, as it other components defines the way the system will often is in computer systems, from individual partials function. within a spectrum, for example, to single sounds, to

Page 8 Computer Music Journal, Box E, Menlo Park, CA 94025 June, 1977 groupings of sounds and to larger groupings. The par- The changes that best define and illuminate a musical ticular definition of a hierarchy will depend upon the composition are found neither at the lowest nor the exigencies of a particular musical idea and sound- highest levels of the hierarchy, but somewhere in generating technique and therefore will vary from between. As the music evolves, we consciously per- composer to composer and from composition to com- ceive neither individual impulses, which move too fast, position. Further, the choice of which sound- nor the overall form of the composition, which moves generating technique to use will affect the entire sys- too slowly. The music is communicated in conscious- tem, since every method contains a built-in set of vari- time, where we are aware of our awareness of change. ables in terms of which the sound can be controlled. Changeat this one level is a meaningful representation The implications of the choice can be stated in the of the whole hierarchy, and powerful variables are context of a comparison between two techniques: thought of as offering direct access to this level. One additive synthesis and 's technique for of the attractions of analog synthesizers is that sound sine wave frequency modulation. generation, which occurs at the lowest level of the Additive synthesis defines a sound as a discrete hierarchy, is automatic, as it is with acoustic instru- specturm, a spectrum composed of separate sine wave ments, allowing the composer to compose "music" components which are usually controlled in terms of rather than "sound." their frequencies, amplitudes and phases. This calls Yet the situation is not so simple. An easy accep- for a vast store of data. If a spectrum consists of, say, tance of variablesat one level, to the relative exclusion the modest number of fourteen sine waves, and each of concern for other levels, might lead a composer, sine wave needs separate amplitude and frequency data without being aware of it, to compose sounds which specified eight times per second, we will have a total do not contain the microchanges that convey qualities of 224 separately controllable variables per second. such as "warmth." And since computer techniques Since the variables are numerous, they are weak. A offer composers the unique potential of composing variable is "weak" or "powerful" to the extent that sound as well as music, it would be ignoring one of the it causes significant change. Weak variables offer the principal advantages of the medium not to deal effec- potential for a detailed specification of sound, but tively with making good sounds. There is, then, an they do not permit realtime specification by a com- argument for weak variables that offer detailed access poser because no composer can specify great amounts to sound and consequently which make possible a of data quickly enough. greater variety of timbre. But even frequency modu- The significant advantage of sine wave frequency lation techniques can become complex with additional modulation is that timbral aspects of sound can be variablessuch as frequency deviation, vibrato, random controlled through two powerful variables. In fre- elements in the waveform, and, in some cases, the quency modulation, the frequency of one waveform number of variables in frequency modulation might is modulated according to the instantaneous amplitude approach that of additive synthesis. From the com- changes of the other. If the frequency of both signals poser's point of view, the problem is one of dealing is in the audio range, audible partials will be added to with the maximum complexity that realtime will the spectrum of the modulated signal. The frequency allow - or not composing in realtime. If realtime is relationships between the component partials in the considered desirable, the answer to realtime com- modulated spectrum will depend upon the ratio of plexity is automation, and automation is achieved by modulated to modulating frequencies. The number of the linking of subsystems throughout the hierarchy. partials present in the spectrum will depend upon the Subsystems may be combined "additively" or ratio of amplitude to frequency of the modulating "interactively." An additive combination consists waveform. That ratio is called the Modulation Index. of component processes which add together in consti- Thus, by varying the ratio between the two frequen- tuting the whole. The components may be seen as cies we vary the nature of the modulated spectrum, independently variable, as are the sine waves in addi- and by varying the Modulation Index we vary the tive synthesis, but at the same time they "cooperate" number of partialsin the spectrum. Significant change in the shaping of a particular totality. An interactive is achieved with extremely efficient means. However, relationship between subsystems has one subsystem because all the frequencies of the partials are in fixed changing proportionally with some aspect of another. relationships with the ratio of the modulated to One sine wave might modulate another, for example. modulating frequencies, and all the amplitudes are in In an interactive relationship, we refer to the modu- fixed relationships with the Modulation Index, all the lating signal as a control and the modulated signal partials will change as a group. This technique offers as a variable. Controls and variables, on the one hand, powerful variables, but it does not offer detailed and processes which are additive, on the other, can access to sound. occur at any level of the structural hierarchy. The There is much to be said for choosing a sound- fabric of the system is made up of all the additively generating method which offers powerful variables. and interactively related subsystems throughout the

Joel Chadabe: Some Reflections on the Nature of the Landscape within which Computer Music Systems are Designed Page 9 hierarchy, and the dynamics of the system are deter- initialized, i.e. turned on or preset; (2) the cybernetic mined by their functioning in relation to one another. system, which pursues the accomplishment of a goal via feedback to the composer and the continual THE REORGANIZATIONAND FUNCTIONING regulation of variables; and (3) self-organizing OF THE SYSTEM systems, where feedback functions not only as a basis for the regulation of variables but also as a basis for A hierarchy is specialized because connections changing the connections between subsystems. It are limited to occur only between adjacent sub- should be noted, however parenthetically, that when systems. In contrast, a reticulated system, in which reorganization occurs and controls are routed to dif- any subsystem can connect with any other subsystem, ferent variables, the meanings of the controls change. is general-purpose. In the functioning of systems, Therefore, taking optimization into account, the specialized systems are defined, i.e. extracted, from controls themselves will have to change accordingly. general-purposesystems. One example of this process An example of a realtime interactive general- in computer music is the composing of a "score" purpose system with reorganization capabilities is with Music V. Donald Buchla's Series 500 Electric Music Box. It Music V, a computer program for direct digital consists of a console which contains performance and synthesis of sound developed by at Bell- other input devices, a computer, a gating matrix and a Telephone Laboratories, is organized as a modular system of analog modules. Control voltage waveforms system of building blocks which are conceptually are defined by the composer and routed from the similar to the modules of analog systems. The basic computer to any other module via the matrix. Their modules are called "unit generators," and include values and their triggers may be controlled by any of functions such as OSC, an "oscillator" which generates the performance devices. Thus, a composer may a waveform by calling a stored function, AD2, a two- specify a system of differentiated variables, connect input adder, and RAN, a random function generator. them in any way, and decide which among them will There are many other unit generators, and any number be performed. But the Series 500 also contains a of them may be connected together to make an program called "Patch," by means of which the con- "instrument." One simple instrument might be two nections between modules may be defined at first, oscillators whose outputs are combined in a two- stored on cassette, and redefined in realtime. This is input adder. A different instrument might be based a distinguishing feature of the system, in that it pro- on frequency modulation. Since each instrument can vides for a complexity of system functioning that consist of any number of unit generators which can would not otherwise be possible. be connected in a variety of ways, and since each unit The process of changing organization might be generator requires different frequency, amplitude or graphically conceptualized as a "reorganization other data, the instruments can vary widely in their spiral," shown in Fig. 2, wherein a hierarchy is complexity and operation. The instruments play extracted from a reticulated system only to dissolve discrete sounds called "notes." A note is specified by again and be reconstructed again. an instrument definition, a set of instructions for the unit generators in the instrument, a starting time, and a duration. The complete listing of all the instru- ments, their data, their starting times, and their dura- El tions, is called a "score." Creating a score consists of choosing certain connections from all the possible connections and deciding upon data for the unit gener- ators involved, which will function at different speeds. Thus, a specialized hierarchy is formed from a general-purposereticulated system. Complexity derives from a multivariable system specialized of subsystems which are affecting and changing each other through their various relationships. But there is general- another dimension to complexity. It derives from a purpose system's ability to reorganize. By "reorganization," I mean a changing of the connections between sub- systems. We might differentiate between three types of systems: (1) mechanistic, fixed logic systems, such as oscillators or random number generators, which produce an output that does not vary according to continual input, but which might or might not be Figure2. A "ReorganizationSpiral."

Page 10 Computer Music Journal, Box E, Menlo Park, CA 94025 June, 1977 Why is the potential for reorganizationimpor- REFERENCES tant? Since the cyberneticmodel for a systtm is single-goaledand does not have a capabilityfor reor- 1. DonaldBuchla, personal communication. ganization,its output is limited to changewithin a 2. John Chowning. "The Synthesisof Complex relativelyfixed range. An "instrument"designation, Audio Spectraby Meansof FrequencyModulation." for example, cannot exceed the changepermitted Journalof the AudioEngineering Society, 21: 526-34, by its variables. Another manifestationof non- 1973; reprintedin ComputerMusic Journal, 1 (2): reorganizationis the "overallness"that characterizes 46-54, April, 1977. "process"art and music, such as paintingsby Jackson 3. Heinzvon Foerster,ed. Cyberneticsof Cyber- Pollockand many compositionsby John Cagewhich netics. Urbana: BiologicalComputer Laboratory, do not havesectional differentiation. Assumingthat 1974. a process does not spin itself out so slowly that 4. LejarenHiller and LeonardIsaacson. Experimen- changeswould normallyseem to be sectional,signifi- tal Music. New York: McGraw-Hill,1959. cant long-termchange is the resultof reorganization. 5. Otto Laske. "The Information-Processing To put that more stronglyin the converse,if a system Approachto MusicalCognition." Interface,3:109- is to have the potentialfor a rich varietyof significant 136, 1974. change,it must have reorganizationcapability. 6. BurtLevy, personal communication. 7. SalvatoreMartirano, personal communication. SUMMARY 8. MaxV. Mathews. The Technologyof Computer Music. Cambridge,Mass.: M.I.T.Press, 1969. We now have a picture of the whole system 9. Max V. Mathews. "The ConductorProgram." organizedas a hierarchyof subsystemsand variables Unpublishedpaper. which function at different speeds, which can be 10. JamesA. Moorer. "SignalProcessing Aspects prespecifiedin detail or automated by means of of ComputerMusic - A Survey." ComputerMusic some process,and which can be combinedadditively Journal, I (1): 4-37, February,1977. or interactively.The system'sfunctioning derives from 11. DexterMorrill, personal communication. its (1) automation,which could be thought of as 12. HerbertSimon. "TheOrganization of Complex internally generatedcontrols and additions, and Systems,"in HowardPattee, ed., HierarchyTheory. (2) controlsfrom the outside, either of which can be New York: Braziller,1973, pp. 1-28. directedtowards changing variables anywhere in the 13. BarryTruax. "A CommunicationalApproach hierarchyor towardschanging connections between to ComputerSound Programs." Journal of Music subsystems. Perhapsthe most importantinnovations Theory, 20 (2): 1-64, Fall 1976; reprintedin part in this approachare the idea of composingwith real- elsewherein this issue of ComputerMusic Journal. time interactivecontrols in an interestingenviron- ment, and the idea that the composerfunctions in two ways - first as the designerof an external system, then as a participantin that system's functioning.

Joel Chadabe: Some Reflections on the Nature of the Landscape within which Computer Music Systems are Designed Page 11