The Illiac Suite is the first musical composition for traditional instruments that was made through computer-assisted composition by Lejaren Hiller and Leonard Isaacson.

PremisesLejaren Hiller’s interest towards music, combined with his training as a chemist who had led to the use of computer, push the American researcher to groped a computer application in music, which was part of the wider set of disciplines outside the mathematical sciences. However it must be said that just the music, compared to the other arts, shares, with mathematics, a relation of broad perspectives, as is widely documented on a historical level (think in this regard to the mathematical relationships underlying to the monochord’s division of Pythagoras). The idea of musical application of computers is also born from a careful analysis that Hiller does on creative musical processes. According to Hiller, music could be defined as a sensible form governed by laws of organization that could be encoded in ways quite accurate.

Music is a sensible form. It is governed by laws of organization which permit fairly exact codification.

In this sense, therefore, the process of composition is essentially based on a set of organizational choices that the composer made on a, hypothetical, infinite variety of raw musical material (eg notes of a scale, ways of attack or links between a section and the other).

 the process of musical composition can be characterized as involving a series of choices of musical elements from an essentially limitless variety of musical raw materials.

These aspects make the computer particularly suited for composing musical works.

The mathematical model – Beyond the detailed analysis of the creative process, to use the computer for musical purposes was necessary identify a mathematical model, defined in computer programming terms, in order to generate data which, when decoded, would have returned a score in traditional notation. In reference to the analysis of the creative processes outlined above, it was to apply a model that allow the computer to make organizational decisions respect to musical composition features. The model adopted was the Monte Carlo method, an algorithm which uses the generation of random numbers.

Information Theory – There was one last important element in the Hiller’s experiment: that of Information Theory (also known as Theory of Communication) mentioned since the twenties. Richard Pinkerton, in the fifties, illustrated, in advance of all, its possible musical application, although not coming to any practical result. Ian Bent, in its Music Analysis book, provides a clear explanation of this theory: “The theory evaluates the ability of a system to receive, process, store or transmit information. And since for information is meant the choice of a message in a series of messages, the probability of arrival of any of these are affected by the increased frequency of some messages than others” and more “When a message occurs within a highly likely choice, we will say that it contains little information; vice versa when a message occurs with a choice unlikely, this will be highly informative”.[2]

The experiment – From these premises we can say that is implemented an information system with high information content of which has reduced the density of information through a selection process: in this system the information content is generated through the Monte Carlo algorithm, while the selection is carried out in accordance with the theoretical assumptions of the Information Theory. The research was carried out by Hiller in collaboration with Leonard Isaacson. The first goal was to program the algorithm by which generate data correlated with musical parameters. The operation of the algorithm consisted of three distinct phases: initiation, generation and verification.

Initialization – At this stage, was defined a table of rules, or conditions, that tell to the computer what combination of data could be considered legal and what not, so that all the choices could be made automatically. The rules could be derived in different ways. In the case of music was possible to assume rules by: a) traditional theory (eg prohibition of fifths and octaves parallel), b) dictated by the imagination of the programmer, even without relationship with music, c) derived from statistical analysis of other compositions and, finally, d) rules auto-generated by computer.

Generation – The second stage consisted with the generation of a single data at a time, put in relation with different musical parameters, such as a single pitch, rhythm, or expressions such as pizzicato and arco. The data were generated at a rate of a thousand numbers per second.

Verification – Once generated, the experiment continues with the comparison between data and set of rules, to validate it or not. If validated, then passed to the generation of a new parameter, otherwise it returns again to the generation phase to generate a new value for the same parameter, until you will get a validated parameter, or until becomes clear that no parameter is able to satisfy the requirements. Once a parameter is validated, before becoming part of the final composition, is allocated in memory until the end of the whole process of generation and selection. At the end of the work, the result is a score in alpha-numeric code which is decoded in traditional music notation, to be performed by an instrumental group.

Structure of experimentation – After defining the procedure and programmed the computer, it was necessary to plan experiments itself. Four different targets were defined, to be achieved with the same number of experiments. In the early stage there was still to define the type of instruments with which play the final score. Technical and structural limits pushed to exclude the use of analogical electronic instruments. Similarly, the piano was excluded and was chosen, instead, a string quartet that would allow a four-voice polyphonic. Reason for this choice was mainly the timbrical homogeneity of instrumental group. Having set a goal of four initial experiments also led to define a broad structure in four movements.

First experiment – Divided into three phases, was designed to obtain simple diatonic melodies. To do this, it was necessary to refer to a method of writing music enough known but not too complex; to be used as a model of construction of the musical material. Hiller and Isaacson chose a strict counterpoint, as expressed in the treatise by Joseph Fux, Gradus ad Parnassum. Compared to this method, you made reference to the counterpoint of the first kind (prima specie) from which were obtained sixteen rules, including prohibitions permits and concessions, which would form the content of the initialization tables. There are three categories of rules laid down: melodic (notes had to be contained within an octave, repetitions of notes were not allowed, etc.) harmonic (were allowed only intervals such as unison, octave, fifth, sixth and third major, or minor; tritone was considered dissonant, etc.) and mixed rules (prohibition of fifths and octaves parallel, in the link from one chord to one of the four voices had to move to stepwise or remain stationary, etc.). The first phase of the experiment was finished with the creation of a diatonic melody, subsequently used as cantus firmus for the construction of other simple polyphony: two voices, in the second phase of experiment, and four voices in the third. All the material formed the three sections of the first movement: presto, andante and allegro.

Second experiment – The musical approach is not very different from the first: you make a four-voice polyphony, but overall structure is not divided into three, but in one section, adagio non troppo lento, which closes with a coda. A new feature is detected at the technical-computing level. Hiller and Isaacson improve the encoding of musical rules transduced in mathematical terms. The goal was to demonstrate that, even if not all, most of the traditional musical notation could be properly encoded in computer terms. Hiller shows as, in the second experiment, computer programming had reached a high level of complexity with a final result obtained at the end of 1900 arithmetic operations.

Third experiment – The third experiment, which gives an Allegro con brio, leads researchers to a musical goals. In this case, they wanted to demonstrate the ability to define, through the computer, parameters such as rhythm and dynamics. A general goal of the research was to be able to computerize all the distinctive elements of music, a fact that remained unresolved at the end of the research. Decisions about metre and tempo, in fact, were taken outside computer programming, referring to other research and clinical study how to computerize musical parameters considered to be more complex. In the new experiment rules tables were revised with respect to rhythm and dynamics problems: were entered all the information needed to perform validation tests on these parameters. Beyond this, the third experiment differs from previous for a chromatic musical writing rather than diatonic. The overall structure was divided into three long sections, followed by a coda. Initially the computer was programmed to produce music through the application of random processes. The first results obtained, a highly dissonant music, forced the researchers to modify the approach by introducing a series of traditional rules which would have improved the work of selection. The result obtained after the imposition of the new methodology was characterized to be very similar to the musical works of contemporary composers.

Fourth experiment – If the first three experiments, although different each other, shared a selection of musical parameters through the imposition of traditional rules, obtained by Fux or by trends of contemporary music, in the fourth and final experiment, the first new element is just replace those rules with other derived from non-musical disciplines. Hiller and Isaacson were determined to prove that a properly programmed computer could compose music, good music, even by using rules that had nothing to do with the art of sound. The new rules system was prepared by drawing on probabilistic method of Markov chains, a stochastic process based on the principle that within a sequence of events, the choice of a new event is closely related to that immediately preceding, without any consideration of events in a time past. Expressed in musical terms the principle tells us that the choice of a note, or interval, is closely related to the note, or interval, immediately preceding. At the end of the experiment, the researchers obtained the fourth movement of the suite, tanto presto che possibile, with the addition of a conclusive coda. From a stylistic point of view, there is a music notation very close to that of contemporary trends, even if obtained through non-musical procedures.

Performance – The research, which we have described the most relevant aspects, was launched in September 1955. In July of 1956 had been completed the first three movements, not including the coda of the third. In the following months would also completed the fourth and last movement. Meanwhile, on August 9, 1956, was organized at the University of Illinois, who had financed the entire research, the first public performance of the Illiac Suite for Strings Quartet, although in its three movements version. The Quartet called to perform the work consisted of students of the University’s Department of Music. The official presentation to the scientific community that was some weeks later, on August 28, 1956 on the occasion of the 11th National Meeting of the Association for Computing Machinery. In the weeks following the coda was also completed in the third movement and the Illiac Suite was complete, so that in November of 1956 was formalized the finished work. Subsequently Wladimir Ussachevsky also became interested in the Illiac Suite and, in general, to all research conducted by Hiller and Isaacson. He became Chairman of the Editorial Board of the New Music Edition, New York, a publication directed primarily to the publication of scores of contemporary music. On whose pages was, in fact, published the final score of the Illiac Suite.

[1] Lejaren Hiller, Leonard Isaaccson, Experimental Music: Composition with an Electronic Computer, McGraw-Hill, New York, 1959.
[2] Lejaren Hiller, Music Composed with Computers: a Historical Survey in The Computer and Music, edited by Harry B. Lincoln, Cornell University Press, 1970, pp. 42 – 97.
[3] Gagne, Cole; Caras, Tracy. 1982. Soundpieces: Interviews With American Composers, Scarecrow.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.