We’ve updated our Terms of Use to reflect our new entity name and address. You can review the changes here.
We’ve updated our Terms of Use. You can review the changes here.

Rainbow Family

by George Lewis

  • Digital Album
    Streaming + Download

    Includes unlimited streaming via the free Bandcamp app, plus high-quality download in MP3, FLAC and more.
    Purchasable with gift card

      $7 USD  or more

    You own this


  • Compact disc
    Compact Disc (CD) + Digital Album

    Comes in an 8-panel gatefold case, with graphics from the 1984 IRCAM concert brochure, arranged by Jonathan Gardner. Includes a 6-page booklet with a note on Rainbow Family by the composer.

    Includes unlimited streaming of Rainbow Family via the free Bandcamp app, plus high-quality download in MP3, FLAC and more.

    Sold Out



A note about Rainbow Family
George Lewis

This audio documentation of my interactive computer music work Rainbow Family is taken from three days of live performances that took place in May of 1984 at the Salle de Projection of the Institut de Recherche et de Coordination Acoustique/Musique (IRCAM). The performances are the culmination of two years of sustained research and creative work in fulfillment of an IRCAM commission. Thanks to Tod Machover and the late and greatly missed David Wessel (1942-2014), I spent three years there, meeting Pierre Boulez (1925-2016), Tristan Murail, Philippe Manoury, and many others, and making some lifelong friends and associates, like Kaija Saariaho and Jean-Baptiste Barriere, as well as György Kurtág (the younger) and the late Oliver Johnson (1944-2002), a lifelong friend of Wessel’s (and a fellow drummer) with whom I performed in Steve Lacy’s ensembles.

For a number of years, the only account of Rainbow Family, the first of my “interactive virtual orchestra” works and a precursor to the better-known Voyager virtual orchestra and “interactive virtual improvisor” systems from 1987 and onward, was in Georgina Born’s still-controversial (in some circles) book, Rationalizing Culture: IRCAM, Boulez, and the Institutionalization of the Musical Avant-Garde (University of California Press, 1995). Despite the fact that most personages in the book are delineated with coded initials, those who venture into that complex ethnography will have no trouble recognizing the Rainbow projects and its composer. In all probability, this was the first commission from IRCAM for what were then called “microcomputers” (small systems) as well as the first that involved so-called improvising computer programs. Thus, for those interested in tracing the development of this aspect of my work, especially in the context of the history of interactive computer music-making more generally, this release could be useful.

Voyager and its offshoots, which are still being irregularly updated, have been taken by many as emblematic of my work in this area. However, Rainbow Family, which drew upon early ‘80s AI, 1950s cybernetics, and sociomusical networks of free improvisation in creating a social aesthetic that included “creative machines” as central actors, brought together the key methodologies and grounding aesthetics of interactivity that marked my later work:

(1) Human and/or non-human musicians engage in live, completely improvised dialogue with a computer-driven, interactive "virtual improvisor."

(2) One or more computer programs analyze aspects of the musical environment in real time, using that analysis to guide an automatic “improvisation” program that generates both complex responses to input and independent generative and analytic behavior that arises from its own internal processes.

(3) The system is an independent improvisor with its own set of musical approaches. When it detects outside input, it takes what it finds into account as part of its decision-making process. However, it should not need outside input to create music.

(4) As an independent improvisor, difference is asserted by the system in the form of program responses that are not necessarily related (or even deliberately unrelated to) outside input. These responses can be heard as the consequences of different modes of a construction of "listening," where the software attempts to create a detailed representation of what it receives from the improvisors.

(5) In playing with different people or in different musical situations, the system should play differently, while retaining something of its own personality.

(6) Any conceivable behavior choices should be designed into the system as part of its base of competencies for it to choose from freely, rather than incorporating hard-wired Pavlovian responses or transforming the system into an instrument to be manipulated by the human performer. Otherwise, I would not be learning much about the nature of communicative strategies in improvisation.

Thus, as new media theorist Simon Penny has maintained about interactive works more generally, “the machine system is constituted as a quasi-organism which responds to changes or perturbations in its environment.” Performances and computer programs such as this deal with the nature of music and, in particular, the processes by which improvising musicians produce it.
In this kind of work, the improvised musical encounter is experienced as a sonic negotiation between musicians, some of whom are people and others not. Any entity operating in this conceptual space would deal with issues of behavior, communication and intersubjectivity. As the anthropologist of technology Lucy Suchman has maintained, such interactions succeed through the communitarian construction of intelligibility. Suchman's logical conclusion here, one that has guided my work with interactive machines since the late 1970s, is that "machine-human differences are discursively enacted and available for refiguring."

The 1984 IRCAM Rainbow performances featured from one to four human improvisors (contrabassist Joëlle Léandre, soprano saxophonist Steve Lacy, flutist and bass clarinetist Douglas Ewart, and guitarist Derek Bailey) performing with three networked Apple II computers, which performed on three of the then-new Yamaha DX-7 synthesizers. Audio input from the instrumentalists went to analog pitch- and envelope-following hardware, which generated voltage values that were converted to digital form. This input information was collected by one computer and distributed over a MIDI-like network to the others; each machine analyzed the data locally and created its own responses to what it “heard,” as well as generating and developing original material with no necessary direct relationship to the input.

Both the computer programming and the hardware hacking for Rainbow were completely my own, and the documentary film about the work that IRCAM produced in 1984 shows me working in the computer language FORTH, which I learned about from my mentor David Behrman. The reason why FORTH became useful in these pre-Max/MSP days was that it was interactive in real human time with the programmer. New modules were quickly compiled and easily tested, and FORTH “words,” as subroutines were called, could be connected to each other and run interactively. Time-critical processes could be coded in assembly language and easily linked to the rest of the system.

As I recall, the form of Rainbow’s musical process resembled a kind of simple generative grammar. Unfortunately, due to my nomadic lifestyle in the 1980s, I no longer have access to Rainbow’s source code. However, I did program the display to show where the system was in its generative process, as well as which of the four musicians it was taking input data from. The documentary film actually shows the display at multiple points, which allowed me to regain a sense of the system’s process grammar. From this I can recall that Rainbow used a “phrase cycle,” a syntactical “theme and variations” idea with a set of optional behaviors that ran in strict order with options:

Create a new phrase, choosing melodies, harmonies, rhythms, dynamics, register, interval width and timbre (DX-7 settings, based on timbres created and generously provided by the now-legendary David Bristow, who was in residence at IRCAM, working with the already legendary John Chowning). I think this process could also choose material from a set of pitches and durations it detected in the input; certainly the later Voyager does this.

Create a variation on what was created in PLAY NEW 1.

PLAY 3 (optional)
Create another variation on what was created in PLAY NEW 1.

PLAY (optional)
Play “freely” without regard to previous themes.

PHRASE REST (optional)
A short pause, probably around 2-3 seconds, that could be asserted after PLAY NEW 1, PLAY 2, PLAY 3, or PLAY.

LONG REST (optional)
A pause, generally asserted after the end of a completed phrase cycle, but not that often.

So you would get combinations like:


The system’s pitch and volume information came from an analog pitch follower designed by Carl Fravel and sold through his Gentle Electric company (gentleelectric.com/documentation/Gentle_Electric_PEF_Pitch_and_Envelope_Follower-Service_Manual.pdf). Serge Tcherepnin graciously told me that this formed the basis for his Serge Modular pitch follower, but I would have to wire up this OEM version myself using the schematic that came with the device. We also had some help from the Yamaha Corporation, which provided the DX-7s. Apple Computer France provided the Apple IIe computers; in 1983, Wessel, who seemed to know everybody, brought John Sculley, then CEO of the company, to IRCAM. Sculley provided a tantalizing preview of a new kind of Apple, called a “Macintosh,” that came out a bit too late to use in the Rainbow project.

I did not feel the need to take volume outputs from the musicians individually; the assumption/hope was that the human players would mostly match their volumes to each other, so one volume number would be enough. The microphone outputs from each player were sent to an 8-in/1-out Analog Devices analog switch multiplexer, which could be digitally controlled with a single byte. Three bits was enough for eight channels, as you binary heads know; 000 was channel 1, 001 was channel 2, and so on. One channel at a time was opened to the pitch follower. The system would always try to match input volume, durations, and register, and those parameters would affect its choices of musical output. The time between two successive pitches was taken as duration, but to avoid glitches, durations that were too short weren’t registered. If the system did not detect any input for more than three seconds, it would pause its playing and you would see “READY” onscreen.

The work was originally conceived not as a network of computers, but as three separate computer improvisors, each running a copy of the Rainbow Family software and making its own performance decisions. The differences in decision-making would be interpreted as a form of sonic individuation--perhaps along lines suggested by Gilbert Simondon, but also based on the terminology developed by African American improvisors, in which players strive to develop their own unique “sound.” Developing one’s own sound amounted to a process of sonic subjectivation.

This was the theory. However, in rehearsals the musicians began referring not to computer A, B, or C, but to “the machine.” To enhance the impression of individuation, I tried strategies of spatialization by having separate speakers for each machine’s output, as well as limiting the set of sounds each machine had available for output. Neither of these strategies could shake the impression of a unitary “machine” subject. I eventually realized that what I had not done was to create three separate programs, each with its own behavior. Three improvisors, even performing on the same instrument, would be heard as three individuals, because the behavior of the individual performers would be different. In the case of Rainbow Family, the three machines were clones that pursued the same kinds of behavior over time, even if individual decisions taken at certain times differed. The response to this by the human performers was what made Rainbow Family an interactive virtual orchestra.


Over the years I’ve had time to think about Georgina Born’s account of my activities at IRCAM, while being reminded by her book that a few people there really were as dismissive of my work as I thought they were, even after three days of interactive computer music concerts in which the system never failed once. Moreover, it was embarrassing, if true, to read that “while subverting many aspects of IRCAM and its technologies, PL [me] did not contest the dominant forms of rhetoric around music and the aesthetic.” Indeed, phrases from the 1984 program notes, such as “The structure of Rainbow Family includes both improvised and precomposed elements,” show me trying to read the dominant aesthetic and adopt its tenets.

At the same time, I was in a DuBoisian condition of double consciousness, what Cornel West has identified as characteristic of “a new breed of cultural workers who simultaneously position themselves within (or alongside) the mainstream while clearly aligned with groups who vow to keep alive potent traditions of critique and resistance.” Thus, what actually happened in Rainbow Family performances might have been a radical position in the musique contemporaine of the time—free improvisations among computers and people in which, as the program notes suggested, “computers and performers have almost the same degree of initiative and responsibility over the course of the performance.”

In my further defense I can only say that I was an immigrant, living in Europe for the first time, struggling with the languages, and trying to assimilate into an international institutional culture from an outsider position, not only in terms of aesthetics and methodology, but in terms of race, a term that was controversially removed from the French constitution in 2018. Indeed, the title of the work referred both to the multi-timbral sound of the piece, and to its dedicatees, the biracial family of my longtime friend and colleague, clarinetist J.D. Parran.

One claim made by European policymakers is that the refusal to collect and analyze statistics on race avoids identity politics and promotes national unity. Subaltern populations, however, are not deceived by all the universalist chatter; the resulting lack of documentation of racialized inequity, or "the choice of ignorance," as social scientist Patrick Simon put it, can simply be a tactic for consolidating the position of dominant groups. Hopping into a Parisian taxi just after the Rainbow Family documentary was shown on the French television channel TF1, the driver, who was black, asked if it was really me in the film, remarking that he hadn’t known that there were black people at IRCAM. Indeed, in all probability I was the first to have received such a commission, although I would be pleased now for that impression to be proven incorrect. Then I remember him saying, “But you are American. It’s different for you.”

We can conclude from the foregoing discussion that interactive computer music can provoke questions that encompass not only technological or music-theoretical interests but philosophical, political, cultural and social concerns as well. In fact, I can say that much of my later scholarly work in critical improvisation studies emerges from my practices of interactive computing. Rainbow Family was an important point of departure for my computer-aided inquiry, over more than forty years, into agency, communication, subjectivity, listening, intentionality, social responsibility, and freedom.


released October 30, 2020

Douglas Ewart - bass clarinet, flute, and Ewart bamboo flute
Joëlle Leandre - contrabass
Derek Bailey - electric guitar
Steve Lacy - soprano saxophone


all rights reserved



Carrier Records New York, New York

Carrier is dedicated to bringing the best, most unique, and most adventurous music from the New York City region, the US, and abroad, and to release a wide range of styles and sound worlds, while maintaining the highest quality releases possible.

contact / help

Contact Carrier Records

Streaming and
Download help

Redeem code

Report this album or account