MUMT 303 New Media Production II Charalampos Saitis

Winter 2010

Computer-Assisted Composition
A short historical review
Computer-assisted composition is considered amongst the major musical developments that characterized
the twentieth century. The quest for ‘new music’ started with Erik Satie and the early electronic instruments
(Telharmonium, Theremin), explored the use of electricity, moved into the magnetic tape recording
(Stockhausen, Varese, Cage), and soon arrived to the computer era. Computers, science, and technology
promised new perspectives into sound, music, and composition. In this context computer-assisted
composition soon became a creative challenge – if not necessity. After all, composers were the first artists
to make substantive use of computers.

The first traces of computer-assisted composition are found in the Bells Labs, in the U.S.A, at the late 50s.
It was Max Matthews, an engineer there, who saw the possibilities of computer music while experimenting
on digital transmission of telephone calls. In 1957, the first ever computer programme to create sounds was
built. It was named Music I. Of course, this first attempt had many problems, e.g. it was monophonic and
had no attack or decay. Max Matthews went on improving the programme, introducing a series of
programmes named Music II, Music III, and so on until Music V. The idea of unit generators that could be
put together to from bigger blocks was introduced in Music III. Meanwhile, Lejaren Hiller was creating the
first ever computer-composed musical work: The Illiac Suit for String Quartet. This marked also a first
attempt towards algorithmic composition. A binary code was processed in the Illiac Computer at the
University of Illinois, producing the very first computer algorithmic composition. Later Hiller collaborated
with John Cage to create HPSCHD, a programme able to define equal-temperament scales, then choose
pitches and durations in them, and finally produce sounds. Back at the Bell Labs, James Tenney worked on
a composing programme, the PLF 2, which he used to compose his Four Stochastic Studies in 1962. The
software he produced was capable of compositional decisions, a fundamental concept in computer-assisted
and algorithmic composition. At the same time Music IV was completed. It was written for an IBM 7094,
one of the first computers to use transistors. Hubert Howe and Godfrey Winham at the Princeton University
improved Music IV, calling their new version Music IVB. When IBM introduced its new computers in
1965, new challenges for computer music appeared. Music I-IV had been written in low-level, machine-
specific assembly language, hence they would not run in other computers. In 1967, the Princeton group
presented a version of Music IVB written in FORTRAN, the Music 4BF. FORTRAN was major high-level
language at the time. Meanwhile, Max Matthews, Jean-Claude Risset, Richard Moore and Joan Miller were
developing the FORTRAN-based Music V at the Bell Labs. Music V was the culmination of the previous
programming environments. It included advanced software-defined unit generators that played notes,
discrete sounds containing transient information for the unit generators. The notion of the score, previously
presented in Music III, was further developed here, including note lists and function tables. Music V
marked the end of the first breakthrough of computer music. It was obvious that new ideas had to be
followed.

Hiller was not the first to work on algorithmic composition. Karlheinz Stockhausen, a pioneer of electronic
music, formulated new statistical criteria for composition, focussing on the aleatoric directional tendencies
of sound movement. Xenakis’ Metastasis for orchestra was premiered in 1955. Xenakis used stochastic
formulas that he worked out by hand to compose the piece. What Hiller showed, was the ability of the
computer to model algorithms. In the late 60s Xenakis went on developing his automated composition
programme, the Stochastic Music Programme (SMP). SMP eventually generates a score for conventional
instruments using complex stochastic formulas. Meanwhile, Gottfried Michael Koenig was studying
computer music programming at the WDR studio in Cologne. In 1971 he went to the Institute of Sonology
in Utrecht, where he completed PR1 (Project 1), a programme for algorithmic composition. PR1 generates
a score following both deterministic and aleatoric composition techniques. Koenig used the programme for
both his electronic and instrumental compositions. Unlike SMP and PR1, Barry Truax made a series of
programmes exclusively for direct digital sound synthesis. Those were named POD after POisson
Distribution, because the distribution of events in time and frequency in the programme follows the Poisson
distribution (Statistics and Probability Theory). In the next years, however, algorithmic composition more
likely became a (standard) part of more general programming environments.

1
MUMT 303 New Media Production II Charalampos Saitis
Winter 2010

One could argue that ‘new music’ has been a three-step adventure. These steps have always been taken
simultaneously, and most importantly interconnecting and interacting with each other. First it was the
challenge of new instrumentation beyond the orchestra. Second came the computers, which could either
compose or provide the parameters for the composer. The third step was the need for new sounds that are
not realizable with acoustical or analog electronic instruments. Sound synthesis provided the ground for
this ultimate step.

John Chowning of Stanford University got involved with the work in Bell Labs in the mid 60s. He brought
Music IV in Stanford and had it run in a PDP-1 computer at the Artificial Intelligence Lab of the
University. When the PDP-1 was replaced by the PDP-6, Chowning together with David Poole wrote
MUS10. Meanwhile, he was working on new techniques for sound synthesis. After several experiments and
mathematical confirmation he introduced frequency modulation (FM) as a synthesis technique with a great
advantage: “extreme economy”, i.e. producing sound with rich spectra from just two oscillators. It was a
milestone in computer music. FM was soon licensed by Yamaha and later patented. During the same
period, digital technology as well as psychoacoustics were making rapid jumps. This created ground for the
cultivation of digital sound synthesis, the new computer music adventure. The objective was the design of
digital synthesizers that would take advantage of FM as well as other computer music developments. On
the grounds of more systematic research, two poles were created in the late 70s and became the leading
centres of research and experimentation in computer music: CCRMA in Stanford, and IRCAM in Paris.

Another significant development of that time was speech synthesis, introduced by Charles Dodge. Gerald
Bennett and Xavier Rodet at IRCAM worked towards generating a singing voice with a computer. In 1978,
they presented CHANT. CHANT was a programme using the human vocal tract as a model of synthesis.
Soon it proved capable of non-vocal sounds synthesis. CHANT also introduced the new area of physics-
based synthesis, also referred to as physical modelling, where a natural sound is synthesized according to
the physics of the vibration structure that produces it. In 1981, the same people developed FORMES, an
improved programme based on object-oriented programming. It was used by Jean-Baptiste Barriere to
compose Chreode in 1983. During that period, computer music was becoming highly popular. More
research centres were formed, and composers-researchers – or researchers-composers – were exploring all
possibilities of using the computer as a musical instrument and/or a compositional tool. However, access to
the established centres was limited and moreover expensive if there was no funding. As a result, the need
for a compositional workstation for the individual composer was growing greater and greater.

In the early 80s a group of composers living in York (UK) was formed. The group was called Interface and
its first members were Trevor Wishart, Richard Orton and Tom Endrich. By 1986, more composers joined
the group. The vision of the group was to build a compositional environment that would be affordable and
accessible by the individual composer. They started working on the Atari ST computer. What attracted
them to it were its affordability, its 16-bit technology, and a built-in MIDI interface. With the software
contribution of Martin Atkins, and the hardware contribution of Dave Malham, the Composer’s Desktop
Project (CDP) was created. Almost a year ago, Barry Vencoe made Csound, a unit-generator-based
software synthesis language. Csound was an attempt to ‘adjust’ the ideas of Music I-V to personal
computers. Csound was a non-realtime environment until 1989, when it was turned into a real-time control
language. Before David Zicarelli became involved with Max/MSP, he contributed in the design of
Intelligent Music’s M in 1987. M was a graphical algorithmic environment allowing the composer to
manipulate sounds recorded through a MIDI keyboard. However, as Digital Signal Processing was
becoming a standard in computer music, more sophisticated programmes were needed to include DSP
techniques.

Several years earlier, in 1981, Pierre Boulez presented Repons, a composition for a twenty-four-piece
orchestra and six soloists. The novelty was that each soloist was independently connected to 4X, where the
sound was processed and routed to different loudspeakers around the concert hall. 4X was built by
Giuseppe Di Giugno at IRCAM and is regarded the first ever made digital signal processor. It was quite
successful and composers such as Luciano Berio, Pierre Henry, and Jean-Baptiste Barriere. In 1985 Miller
Puckette went to IRCAM and started programming software for 4X. While working together with the
composer Philippe Manoury on the later’s Jupiter, they faced timing issues with the existed software. They
were interested in having the performer triggering electronic sounds. The main challenge was to time

2
MUMT 303 New Media Production II Charalampos Saitis
Winter 2010

musical events independently of one another. The solution came from Puckette in 1987. He overcame the
timing problem by programming a realtime scheduler, which he named Max, after Max Matthews. To make
things easier, he moved 4X and his scheduler on a Macintosh with MIDI support. He went on designing a
graphical interface, which he named Patcher. Soon the interface became the main part of the project, the
project went further than the initial concept, and in 1988 the first version of Max was presented. The next
step was taken by Zicarelli, who developed MSP, an environment for real-time audio synthesis, DSP, and
algorithmic composition, and extended Max to Max/MSP. Max/MSP was an unprecedented breakthrough,
still regarded as the leading real-time audio synthesis platform. Puckette went on re-designing an open-
source version, which he called Pd (PureData). Pd has some fundamental differences from Max/MSP, but
is quite as popular, and in many cases preferable because of its free distribution. In 1996 James McCartney
wrote SuperCollider, a programming environment with object-oriented language for real-time audio
synthesis and algorithmic composition. Max/MSP, Pd, and SuperCollider are now used by many
composers-programmers, as well as by sound artists, and sound engineers. In the dawn of the twenty-first
century computers are more powerful and thus capable of things yet to be done. Scientists, programmers,
composers, and musicians work interactively towards new challenges for music and composition.

References
Chadabe, J. (1996). Electric sound: The Past and Promise of Electronic Music. Upper Saddle River, NJ:
Prentice Hall.

Roads, C. (1996). Computer Music Tutorial. Cambridge, MA: The MIT Press.