Page  37 ï~~Live Interactive Intelligent Computer Music in HMSL: Notes on Pieces 1984-1991 Larry Polansky, Dartmouth College, Music Dept., Hanover, NH 03755 email: larry.polansky@mac.dartmouth.edu Nick Didkovsky, 171 E. 99th St #20, NYC NY 10029 Abstract: This paper describes several pieces written and performed by the authors, individually and together, in the computer music language HMSL between 1984-199 1. These pieces explore different aspects of performer interaction with computer software, aspects of human-machine improvisation, and aesthetic issues in live computer music. Introduction This article documents works for live interactive computers and performers written in the computer music language HMSL between the years 1984 and 1991. The works are a small sample of the experiments, compositions and performances done in HMSL. For more descriptions of HMSL see Polansky, Burk, and Rosenboom [1987, 1990]; Polansky and Rosenboom [1985]; Riddell [1989]; Scholz [1990, 1988, 1988a], Didkovsky [1990b]. Aspects of our work in HMSL Interaction: Each piece explores interactivity in a different way. The performer(s) have a specific, new and unusual task, or form of interaction with the making of music, in communicating with the machine and other performers. Intelligence: Each piece approaches composition and creativity in a slightly different way, exploring notions of determinacy and indeterminacy, the co-involvement of the performer and the machine with the compositional process, and in many cases, the deliberate abdication of certain compositional decisions and determinations of musical parameters to the machine. Development of HMSL and a community of composition: One of our interests was to develop code and ideas that would be useful to and interactive with the work of other composers. In most cases, the code for these pieces was made available to the large community of HMSL users through BBS's, networks, and through copying disks. Low overhead: All of the pieces involve a minimal, inexpensive and portable computer music system, in which the success and musical intent of the piece is not determined by the expense of the hardware. The pieces needed to travel (most of them have been performed internationally). We were interested in deemphasizing the relationship between consumerism and music, a natural result of MIDI: the interest of a piece shouldn't be related to the economic means of a composer (something Ron Kuivila has jokingly and aptly termed "virtuoso consumerism"). The challenge was to make interesting pieces with inexpensive, flexible, and sometimes rather humble-sounding equipment. Pieces: by Larry Polansky 11' TN'1 (B'rey'sheet) (In the beginning... ) (Cantillation Study #1) (1984; revised 1987,1989) 1'1Z was one of the first pieces written in HMSL (by me or anyone). It was written for Jody Diamond, with whom I had studied cantillation of the Torah, and is the first of a set called the Cantillation Studies, based on computer-aided morphological transformations of the 11th and 12th century Masoretic cantillation melodies (tropes). Each piece is based on successive 17-verse sections of the Torah, and named, in traditional manner, after the first few words. The Torah tropes are used as a basis for melodic transformation by the computer. In B'rey'sheet, the tropes are sung, unadorned. The pitch of Jody's voice is captured by HMSL by a Pitch-To-MIDI converter, using a significant amount of software debouncing of incoming pitches. Jody sings in a just intoned scale I selected for the highly modal trope. The text is the first section of the Torah - describing creation, in which order is brought to the various cosmological parameters. With each of the 17 verses of the Torah section, the statistics of the computer's musical response are constrained. In this way, the computer "listens" to the melodies and generates its own events based on what it hears, and on where it is in the piece. There is a predefined trajectory of "computer attention," acheived by constraining the degree of randomness, which specifies that ICMC 37

Page  38 ï~~at the beginning of the work the computer mostly ignores the voice, but gradually and continuously increases its attention until the end, when follows the voice closely.The piece starts out in accompanimental statistical chaos in all of its parameters, and ends in unison with the voice. One variable controls the degree of all change in the piece, beginning high (17) and ending at 0. A second performer (usually me) changes the value of the variable (essentially telling the computer where it is in the piece) through a simple HMSL graphic interface By delaying or anticipating this change in relation to the singer, the mood and macro-rhythm of the performance can be varied tremendously -Jody and I have found through many performances that even though both tasks are highly constrained (she just sings the trope, I change a few variables more or less at specified times), the piece sounds and feels very different depending on our respective timings. Our growing sensitivity to these details makes performing the piece enjoyable, exciting, and unpredictable. The computer's sonic material is limited to four sine waves- a limitation accepted from my decision to use only Amiga local sound. The Amiga has four 8-bit DMA sound channels, which although of relatively low fidelity, are highly flexible in terms of intonation and timbre. Since a separate processor is responsible for updating the DAC outputs from a specified memory location, high-level software can change the output waveform transparently and quickly without interrupting the sound. In B'rey'sheet, I was interested in dynamic, point-by-point time domain waveshape modifications that were "orthogonal" to the kinds of higher level morphological transforms that HMSL was designed for. The sine tables are treated as long melodies, and the wavetable modulations have the following parameters: amount of wavetable to be modulated, degree, possibility and type of modulation. The modulation techniques are derived from the methods of the shape class in HMSL: simple ways of editing lists of musical data. For example, in B'rey'sheet, particular points in the sine table can be replaced with other points (a kind of spectral deformation), or portions of the table can be retrograded, inverted, scrambled, randomized, and so on. The degree of modulation is determined by location in the piece. Typically, at the beginning of the piece, all four sine tables are modulated at audio rates, as is the type of modulation itself. This modulation calms down both in rate and degree as the piece progresses. One of the most interesting aspects of B'rey'sheet is that the computer tunes "on the fly," making use of a simple intonational trajectory which guides the tuning decisions. This trajectory begins in a complex 17-limit tuning space, and ends in a kind of simple 3-limit one (or Pythagorean tuning). The current verse number serves as the limit for the tuning space, and random tuning values whose prime factors less than that number are chosen for numerators and denominators of just intervals to the just intervals of the melody itself. For example, in the first verse (17-limit), the computer might use 51/50 (a small minor second to the 8/7 using prime factors 17, 5, 3, and 2) to harmonize the septimal major second in the melody (8/7), resulting in an absolute "minor-third" interval of 204/175 (app. 266.5 cents). This particular pitch might only last a fraction of a second, since in the beginning of the piece everything is changing quite rapidly. The tuning process happens simultaneously in all four voices. The computer is always in harmony with the voice, but the nature of the harmonic space is highly stochastic, and has no real functionality. These tuning algorithms fall into the category of what I have called paratactical tuning [Polansky 1987b,c] - all of the tuning is done in real-time, in response to the input. No concept of scale, or gamut, is ever invoked. The pitches used are not chosen from a set of pitches, but generated by the machine in realtime, similar to the way a chorus or string quartet would dynamically tune to itself over the course of a piece, except that in B'rey'sheet the rules are extremely primitive! A natural extension to the algorithms would impose some sort of voice leading or functionality rules upon this tuning system, but this was not part of my aesthetic intent in this piece. Two Portable Pieces 17 Simple Melodies of the Same Length 17 Simple Melodies... (Polansky [1990b, 1988b]) was written for composer/performer Daniel Goode in 1987. It was intended to be a truly portable computer music piece which was more than simply a sequence of note information - offering a live, interactive, and flexible piece for any melodic performer who could input MIDI pitch data. This was motivated to some extent by the fact that when I wrote it, more and more extraordinary and experimental performers (like Daniel Goode, John Oswald, Ann laBerge, George Brooks, and others who have performed this piece) were starting to work with MIDI equipment in sophisticated ways: programming their own sounds, using various input devices and so on. From a software standpoint, what tended to be available to these artists was, in general, commercially oriented, offering limited ICMC 38

Page  39 ï~~experimental resources for composition and performance. I intended 17 Simple Melodies... as a kind of example of what could be done in designing unusual, interactive, and intelligent pieces for computer and performers of conventional instruments. By only supplying a musical form for the piece as a kind of "black box" software engine, leaving the timbres of the MIDI synthesizers as well as all the melodic material completely up to the performer, I was inviting what I hoped was an evolutionary and expansive collaboration. From its inception, 17 Simple Melodies... was distributed to the performers as a disk with instructions. The disk contained a fully executable version of HMSL with the code for the piece compiled as well. In this way, the score for the piece was the code itself. Source code is also supplied with the piece, and performers are encouraged to modify it, or use it as a model for their own pieces, which would be some sort of "collaboration" with me. The form of 17 Simple Melodies... is a kind of good-natured parody of "classic" artificial intelligence: data is gathered by a kind of perceptron (17 melodies of 17 notes each), the data is sorted (the intelligence), and finally, the data is re-played as a kind of simulation. In Section 1, the performer plays 17 melodies, signalling the computer through the ASCII keyboard when she is ready play the next melody (in one performance, John Oswald put the Macintosh keyboard on the floor and made these signals with his foot!). The computer informs the performer when it thinks it has heard 17 notes. Glissandi, rapid arpeggios, multiphonics, noises, and so on will confuse it, usually in an interesting way. Section 2 is silent, lasting a few moments while the computer sorts the 17 melodies into three independent lists, each in regard to some metric to the first melody played. All of the melodies are measured with respect to their similarity to the first melody by three different morphological metrics [Polansky, 1987a], or distance functions on melodies. The basic metric from which the three variations are derived is a simple version of what I have called the Ordered Combinatorial Direction metric: L-1 L-j _ d if f (sgn (Ni, Ni ), sgn (MiMij) ) d(N,M) = j=l i=l Lm where: N, M are two morphologies, or melodies (ordered lists of numbers); Ni, Mi, are the ith elements of morphologies N and M; sgn is a contour function: it returns a -1, 0, or 1 depending on whether or not first element is bigger, equal to, or smaller than the second; diff is a binary comparison: if the values are equal a zero is returned, if unequal, a 1; Lm is the binary coefficient of the length of the melodies (in this piece always 17), or the number of pairwise relationships. If the melody is L notes long: Lm =L2-L 2 (or in this case, 136). Lm describes a "half-matrix minus the diagonal." More simply put, this metric compares the combinatorial contours of two melodies by summing the difference of corresponding cells in the two Lm contour matrices generated by N and M. Each cells contain a 1, -1, or 0, depending on the combinatorial contour of the melody itself. (Polansky and Bassein [1990], Marvin and LaPrade [1987], Friedmann [1985, 1987]). If all cells of the matrix generated by N are equal to those generated by M, the metric would be 0, and these melodies would be considered to be the same in terms of their combinatorial contour. A value of 1 would indicate that the two melodies were as far apart as "they could be" in terms of their contour. Three lists of the 17 input melodies are made according to this metric in the pitch dimension, the duration dimension, and an equally-weighted average of pitch and duration dimension.There is no necessary correlation between the three lists: they could all be identical, or completely different depending on the correlation between the similarities in the duration and pitch dimensions to the first melody. Section 3 simply plays back the three lists, simultaneously, on three MIDI channels. The performer does nothing, creating a nice symmetry: in the first part only the performer plays, in the second nobody plays, in the third only the computer plays. Several variables are left to the performer, which are specified at performance time, including: the average number of repeats for each melody, the probability of changing a MID! preset at any given time in the piece, and the list of MID! presets available to each channel. By ICMC 39

Page  40 ï~~working with these parameters, and customizing the sounds and melodies, the performer has a tremendous degree of control over the piece. The duration of the piece is greatly affected by the average number of repeats for each melody. The performer specification of MIDI preset usage adds a distinctive individual voice to the performances. One interesting performance by Daniel Goode used simple sinusoid-like presets for all the voices, having the strange effect of blurring the linear polyphony into a more harmonic texture. 17 Simple Melodies... was a philosophical and a technological experiment. There is a conceivable, if extreme description of the piece which says that hearing it is to some extent irrelevant, since I purposefully avoid the decision about what it should sound like. This is of course at odds with the prevailing notion that music is, at its most fundamental definition, in some way inextricably tied to sound (I don't believe this to be necessarily true), and also against the more fashionable idea that if a composer needs to explain a piece, it can't be very good. 17 Simple Melodies... is heard, but it is also in some sense, pure explanation, both from the composer to the machine and the performer, and from the performer and machine to the audience. I was not so interested in removing sound from music as in refocussing the performer's and audience's attention: the particular timbres might be irrelevant (there is no real skill or compositional intelligence required to choose a MIDI preset). The very straighforward structure of the work should emerge as the salient feature. This has proved to be difficult for some listeners more accustomed to reacting to electronic music almost exclusively on the basis of its timbre. In 17 Simple Melodies... this is like reacting to the color of the chair you're sitting in: it's an important part of your experience, but not one that I wish to have any part in determining. As a result, this work has confused, angered (and, I hope, fascinated) many who have heard it. Pieces: by Nick Didkovsky DrNerve.hmsl Doctor Nerve is a seven piece band based in New York City that plays hard-edged, experimental atonal music for which I have been composing for over eight years. In February of 1989 I began work on a program called DrNerve.hmsl, which generates musical compositions orchestrated specifically for Doctor Nerve's seven instruments: soprano sax, trumpet, bass clarinet, vibraphone, electric guitar, electric bass, and drums. I wrote DrNerve.hmsl to give myself a compositional shock. I suspected that the resultant music would have an innocent freshness, unpredictability, and curious mixture of clumsiness and precision, as the program would consist of only a limited framework of instructions, devoid almost entirely of world knowledge. I suspected that it would generate music that my own prejudices overlooked, and I was interested in seeing how much of the resultant would work its way into my own aesthetic. The realization of a composition in DrNerve.hmsl takes four stages. First, the program is run, taking about thirty seconds of computer time, producing a piece of music from 1 to 3 minutes in length. Second, if a particular piece contains material worth developing, it is stored to disk. Third, the file is loaded into Deluxe Music Construction Set (DMCS), a commercial notation program, where I can rearrange, edit, and develop the ideas originally generated by DrNerve.hmsl. (On one occasion, DrNerve.hmsl generated a piece which I felt required absolutely no rearrangement or editing whatsoever, and is currently in the band's repertoire.) Fourth, when I feel that a composition is completed, it is touched up in DMCS, and saved to disk. I wrote a program called Copyist Companion which translates a DMCS score file into one readable by Dr. T's The Copyist, which generates high quality scores for the musicians. Finally, I make cassette tapes of the piece for each of Doctor Nerve's musicians. Since DrNerve.hmsl typically generates music which is almost completely insensitive to technical difficulty, these tapes speed the musicians' learning process. Description of the program DrNerve.hmsl is neither an "expert system" nor "rule-based" software: it can generate music very quickly - it knows very little. To generate a measure of music, it first chooses a subset of a pitch row and a subset of a duration row, which stay fixed for the entire composition. Using these subsets, a bass line is generated for one measure by splitting the measure into some number of phrases, and cycling through the subsets of pitch and duration. Once the bass line is composed, it generates a derivative drum part for kick drum, snare drum, hihat, and toms. Doctor Nerve's drummer, Leo Ciesa helped develop a drum part algorithm which creates an accompaniment that is beautifully tied to the bass. The results of this algorithm are shown in the score excerpt below. DrNerve.hmsl has contributed a tremendous vocabulary to my compositions. I was repeatedly presented with ideas that should have been obvious to me, but which I'd never realized. For examples of the music see Didkovsky [1991, 1990a]1. ICMC 40

Page  41 ï~~Toms/Hihat 1 r,' -3-3 ---i - 5--I- 31 Kick fSnare I I I,,rI-I [ F' F! I v,, F. Bass I 3- 5r-3--"~ II ~2r 5 j- L_ 5 i-- (excerpt from Nerveware #2) Lottery Lottery is based on a social model which promotes responsible resource sharing. Performers are temporary members of an evolving microsociety, the behavior of which is correlated to audible change. Lottery's goals are: 1) that its participants recognize their joint situation, and address this interdependence rationally 2) that its participants evolve their microsociety into responsible resource-sharing behavior, and 3) to produce a music which closely models the development, successes, and failures of the first two goals. The piece is derived from the work of evolutionary biologist Garrett Hardin and computer scientist/theorist Douglas R. Hofstadter [1983]. It was programmed by myself, with the assistance of Phil Burk, Larry Polansky, and Robert Marsanyi. The piece was first performed by Burk, Polansky, Marsanyi, and myself at the Mills College CCM on 4/1/90. Lottery models a society whose participants vie for control of a common resource. "The social arrangements that produce responsibility are arrangements that create coercion." (Hardin). A shared musical environment constitutes this resource, and Lottery employs a coercive device to oversee its allocation called a Luring Lottery. A Luring Lottery introduces a contradiction between the desire to win, and the desire to gain. The awarded prize is a quantity X/N where X is some maximum prize value (like a large sum of money), and N represents the total number of ballots submitted to the lottery. Participants may enter as many ballots as desired, but as the number of ballots N increases, the awarded prize X/N decreases! An overly zealous participant might enter millions of ballots, assuring a victory but the awarded prize would be negligible. This simple and beautiful inverse relationship has tremendous power, and is inherently resistant to the hoarding and over-exploitation of common resources. Lottery 's common resource is musical control over Sound States. A Luring Lottery is called periodically, which awards temporary control of the current Sound State to the winner. The magnitude of control given to the winner is divided by the total number of ballots submitted. Participants are in a simultaneously cooperative and competitive environment, where they must balance the desire to control the piece, and assuring that such control will not be diluted to a meaningless level. This behavior is explored in a highly quantified manner, where the evolution of the performers' microsociety is readily analyzed. Lottery is for any number of performers, each with an Amiga. All Amigas are linked together via MIDI, used purely as a networking protocol. The sound generation of Lottery is handled by the Amigas. The common resource is a Sound State of audio waveforms, built from the sum of sine waves in natural harmonic ratios. The number of waveforms is four times the number of Amigas in the piece. Each new Sound State, realized every thirty seconds as the result of a new Luring Lottery, is a weighted mix of the piece's previous Sound State and the winning Sound State. During the thirty second interval, each performer may construct a potential Sound State by editing the magnitudes of the partials in any or all waveforms. Each performer may also submit any number of ballots to the lottery during this interval. The host machine selects a winner at random, and implements the winner's Sound State. As required by the Luring Lottery, the extent of this implementation is scaled down by the total number of ballots. If participants in a performance submit a huge amount of ballots, the piece will not change noticeably from its current Sound State. ICMC 41

Page  42 ï~~The first rehearsal of Lottery ought to have been public - the performers' behavior was the most radically evolutionary and self-aware. The "microsociety" evolved most painfully during this rehearsal, as we witnessed group behavior ranging from "arms escalation" (all participants begin to submit higher and higher numbers of ballots), "de-escalation" (participants see the futility of their joint behavior and begin dropping out of the balloting process, or submitting very low counts), "peace" (a group behavior of low balloting), and "de-stabilization" (one participant gets greedy and starts to submit significantly more ballots that the rest of the group; starts to win disproportionately often; others follow suit). By the time we performed the piece, the group as a cooperative organism had already been established, and the performance was rather tame. But what it lacked in social storminess, it gained in a cooperative music making. The group never established a definition of responsible behavior, that is, as defined by Hardin, which is duplicated by all and which benefits each member maximally. Our behavior developed intuitively into low balloting. Recently, I have derived more formal approaches which maximize each participant's personal gain. These models require access to a random number generator which determines whether or not a participant enters a given lottery, and how many ballots she should submit. The next version of Lottery will have this tool available to its participants, who can pursue a more quantified notion of cooperation. Collaboration: Slippers of Steel There Is More Headroom, But One's Feet Are Forced Into Slippers Of Steel is a live performance for two electric guitarists, two computers linked via MIDI, and electronic sound-generating hardware. The title is an excerpt from Melody Sumner's booklength poem, The Time is Now. Slippers... was composed and programmed by Nick Didkovsky and Larry Polansky in the spring of 1991, and premiered on 5/22/91 at Dartmouth College. It addresses human and computer reactions, pushing the limits of the capabilities of live and inorganic performers to perceive and react to a rapidly changing performance environment. The piece alternates six times between two performance behaviors: Layers and Hits. The decision to change from one to the other can be made by either computer, which informs the other machine (via MIDI) and the live performers (via text on the screen) of the change. The machines continuously sample each others' behaviors via MIDI, keeping all aspects of the performance in sync. Layers consist of parametrically defined streams of sound, each lasting from 4 to 20 seconds. During a Layers section, any number of sound streams can occur, separated by 2 to 13 seconds of silence. As both computers and performers play during layers, four sets of such sound streams are audible, overlapping in various combinations. The computers use the following parameters to generate their sound streams: Pitch Mean and Range, Loudness Mean and Range, Horizontal Density Mean and Range, Vertical Density Mean and Range, and Legato Mean and Range. Each computer-performed sound stream is played on MIDI devices. Each stream may be performed in one of three modes: 1) independently of the other computer, 2) imitating the other computer, or 3) opposing the other computer. In independent mode, random sets of starting and ending parameters are created.The starting values evolve to the ending values either linearly or by following a half cosine curve [Tenney, 1987] over the course of the 4-20 second stream. A stream might go from low high pitches, while its dynamics fall from ffl to mf, while evolving from monophony to polyphony, while changing from a very fast, dense cloud of sound to a very sparse one, and from a great degree of overlap to a very short staccato treatment of pitch events. Each of these parameters may have their own trajectories, resulting in unusual and quite beautiful phrasing. We found that the common cliches such as fast=loud and sparse=quiet were shattered by this approach to phrase generation, which deliberately avoids such artificial correlations. A sound stream in imitation mode continuously samples the parameters of the other computer, and plays them. Each composer is responsible for programming his own "parametric player" which interprets these values musically. The skeleton of the overall program runs identically on both computers, but provides for customized interpretation. Thus, even in imitation mode, the performance of each machine differs enough to be musically vital, yet follows the parameters closely enough to make the notion of imitation obvious to the listener. For example, one computer's imitative response might be to play dyads in perfect fifths with the pitch values it sampled. A sound stream in opposition mode also continuously samples the other computer, fetches its current parameters, then "flips" these values with a composer-written custom opposition function. Again, each composer provides his own opposition functions, some of which are simple, some more elaborate. While performing their layers electronically, the computers also give performance instructions to the two guitarists, which are identical to the modes of the machines: 1) "Play independently" 2) "Imitate the ICMC 42

Page  43 ï~~other player" 3) "Oppose the other player." An additional "Stop Playing" command provides for the silence between live sound streams. The guitarists apply their own intuitive notions of opposition and imitation to the other's performance. When independent, each guitarist may improvise freely, bearing in mind that his playing must evolve from a starting state to a goal state.This scheme creates fascinating interactions. Performance modes can combine in the six following ways ("unit" refers to either guitarist or computer): Unit I Unit II (1) independent independent (2) independent imitate (3) independent oppose (4) imitate imitate (5) imitate oppose (6) oppose oppose Behaviors created by (1), (2), and (3) are self-evident. (4) - (6) are more interesting. (We will use "fast" and "slow" as generic descriptors of high and low parametric values respectively). (4) causes both units to converge immediately on the same unchanging parameter values. If I plays "fast," so does II (imitation). As each unit polls the other, confirmation of "fast" is fetched, and imitated, thus never changing. This behavior is static and identical. (5) generates rapidly oscillating behavior. Following the "fast/slow" terminology, if I fetches "fast" from II, it will imitate, and play "fast." An instant later, II fetches "fast" from 1, and opposes with "slow." An instant later, I fetches "slow" from II and imitates, causing 2 to swing back to "fast" by opposition. On the average, these oscillations rebound at the rate at which each machine reads the other's data. For human performers, this speed of oscillation is directly proportional to the performer's ability to perceive the other's activity and react decisively. One of our goals is to reduce the live "sampling rate" as much as possible. (6) results in static, opposite behavior. If I fetches "fast" from II, it will oppose it with "slow." When II fetches this "slow" value from I, it will oppose with "fast." This exchange settles almost immediately into static opposites. At any time during Layers, one of the computers may decide stochastically to switch the whole piece to Hits, a sequence of reaction exercises, separated by unpredictable silences. Performers wait in silence until one of the computers initiates a "hit" by emitting a single sound event, at which time the other computer and the two guitarists must react as quickly as possible by responding with their own single sound event. After this flurry, all wait in silence again for the next "hit." The idea is to minimize the duration between the initiating sound event and the response events. The computers' limitations are easily quantified (the MIDI data rate and the efficiency of the software response). It is more difficult to quantify the response time for live performers, whose ability to react quickly is a complex summation of personal concentration, speed at which sound reaches the performer from the PA system, and the rate at which nerve signals travel through the human body. We aim to minimize this human interval as much as possible. During Hits, any computer may give the signal to switch back to Layers, which the ensemble follows. Performances of Slippers of Steel vary tremendously. The use of stochastic elements in structuring the changes from Layers to Hits makes it difficult to predict the overall length of the piece. We have rehearsed 5 minute versions and 20 minute versions! The use of chance elements in the selection of performance parameters and performance modes also ensures that the middle level elements of the piece are unlikely to be repeated. As the piece addresses the very elemental musical disciplines of reaction time and fast decision making, it is a raw and exposed performance experience, exhilarating and exhausting to play. A Selected List of Other HMSL Works by the authors Larry Polansky Bedhaya Sadra/Bedhaya Guthrie (for the Astra Choir) [1990c]. Collaboration with I Wayan Sadra. L.S~' T1 L71 (V'leem'shol). 1985 [1988a, 1988c] Simple Actions 1985-90 [ 1990a, 1988b, 1987d] Four Voice Canon #6 [1990a, 1988a, 1988c] Horn (for Chris Bobrowsld) 1990 3 Sttudies. I. Rhythm II. Melody III. Harmony 1990. For the Dowtown Ensemble. [1990d] Cocks crow, dogs bark....1988. Collaboration with Melody Sumner and John Bischoff. [1988b] Distance Musics 1987. [1987d] Mod.mania; The World's Longest Melody; Buy Some for Spare Parts. 1986-88. Ambient pieces. ICMC 43

Page  44 ï~~Nick Didkovsky Fast Fourier Fugue. 1990. Solo four voice computer performance. IMPv2.0. 1990. Software instrument for live improvisation. Fourier Music I. 1990. For flute and Amiga Interference (Cranked and Crammed). 1990. For live performer and computer. Metamusic/Metatext. 1989. Commissioned by The Downtown Ensemble. Phoneme Music. 1989. Commissioned by Gamelan Son of Lion. References Burk, Phil. 1987. "HMSL - An Object Oriented Music Language." In RoboCity News. Volume III. May. Didkovsky, Nick. 1991 (with Dr. Nerve). Beta 14 ok. CD using DrNerve.HMSL. Cunieform Records 26. - 199 la. (with Dr. Nerve) Beta 14 ok. CD using DrNerve.HMSL. Cuneiform Records. RUNE26. - 1990a. (with Dr. Nerve). Did Sprinting Die? CD using DrNerve.HMSL. Wayside Music Archives Series. WMAS2 - 1990b. "HMSL: In all Languages." Ear Magazine. Vol. 14. #10. February. - 1989. "Computer Generated Composition for Doctor Nerve." Proc. of the 9th Symp. on Small Computers in the Arts. November. Friedmann, Michael. 1987. "A Response: My Contour, Their Contour." J. Mus. Theory. Fall. 268-274. - 1985. "A Methodology for the Discussion of Contour..."J. Mus. Theory. 29:223-248. Hardin, Garrett. 1968. "The Tragedy of the Commons."Science. V.162, No 3859. December. Hofstadter, Douglas. 1983. In "Metamagical Themas."Sci. American. Vol 248, #6. June. Marvin, Elizabeth West, and Laprade, Paul A. 1987. "Relating Music Contours: Extensions of a Theory for Contour." Journal of Music Theory. 31:225-267. Polansky, L., Burk, P., with Rosenboom, D. 1990. "HMSL: A Theorerical Overview." PNM. Summer. Polansky, L. 1990a. The Theory of Impossible Melody. CD. Artifact Recordings. ART04. - 1990b. 17 Simple Melodies of the Same Length. Software. Frog Peak Music. - 1990c. Bedhaya Sadra/Bedhaya Guthrie. With I Wayan Sadra. Score. For voices, kemanak, and gamelan. Frog Peak Music. - 1990d. 3 Studies. Software. Frog Peak Music. - 1988a. Four Voice Canons #3-6/Cantillation Studies #1-2. Cassette. Frog Peak Music. - 1988b. Works for Performers and Live Interactive Computer. Frog Peak Music. - 1988c. Selected Compositions. Frog Peak Music. - 1987a. "Morphological Metrics: An Introduction to a theory of formal distances." In Proceedings of the ICMC compiled by James Beauchamp. San Francisco, CA: Computer Music Association. - 1987b. "Paratactical Tuning: An Agenda for the Future Use of Computers in Experimental Intonation." Computer Music Journal. Volume 11. Number 1. Spring. - 1987c. "Item: Lou Harrison's Role as a Speculative Theorist...." In A Lou Harrison Reader. Soundings Press. Santa Fe. - 1987d. "Distance Music I-VI, For Any Number of Programmer/Performers and Live, Programmable Computer Music Systems." In PNM. 537-544. Vol. 25. Nos. 1 and 2. Winter/Summer. Polansky, L. and Bassein, Richard. 1990. "Possible and Impossible Melody: Some Formal Aspects of Contour." Paper delivered to the Society for Music Theory, Oakland, 1990. Submitted for publication. Polansky, L, and Rosenboom, D. 1985. "HMSL: A Real-Time Environment for Formal, Perceptual and Compositional Experimentation." In Proc. of the ICMC. Barry Truax, editor. 243-50. San Francisco, CA: Computer Music Association. Polansky, L., Rosenboom, D. and Burk, P. 1987. "HMSL: Overview (Version 3.1) and Notes on Intelligent Instrument Design." In Proc. of the ICMC. Urbana, Illinois. Riddell, Alistair. 1989. "Interview with Larry Polansky." ARRAY, Newsl. of the Australian Computer Music Association. #2. Scholz, Carter. 1990. "Computer Partnerships." Keyboard. October. -- 1988. "MIDI Resources." In Keyboard magazine. 74-88. November. -- 1988a. "HMSL Software Language."Music Technology. September. Sumner, Melody. 1983. The Time is Now. Burning Books. Oakland. Tenney, James. "About Changes: Sixty-Four Studies for Six Harps." In PNM. 537-544. V.25. N. 1/2. ICMC 44