Aesthetics and Techniques in the Electroacoustic Music of Iannis Xenakis

by NIKOLAS VALSAMAKIS

The Thought of Iannis Xenakis and the Electronic Medium Xenakis's thought was developed in the context of the postwar musical avant-guard in the fifties. His roots date back to the theories of Edgard Varese and Italian futurism. Xenakis showed a strong interest in formal and mathematical procedures in music. He was aware of the scientific and technological advances of his era and introduced many related ideas in his works. His work in electroacoustic music was a logical step toward a thorough investigation of various possibilities in musical composition. Historical Context: Modernism and the Musical Avant-garde Xenakis's artistic activity should be examined in the context of the postwar musical avant-guarde and the associated movements of modernism. A reference to some key features of modernism would help in the understanding of Xenakis's musical thought. A first aspect of modernism was the reaction against the prior aesthetic and philosophical forms of romanticism and classicism.
NIKOLAS VALSAMAKIS studied computers and music technology in

Athens and London. He received a Msc in Music Information Technology from City University, London. He has worked with KSYME (Center for Contemporary Music Research) in Athens since 1989. He is a founding member of the non-profit music and theater company "Prospecta," with a range of performances in Athens. Works have been performed in Athens FestivalHerodium Theater, Goethe Institute, Ancient Theater of Delphi, among others. He currently teaches Music Technology at AKMI College in Athens.

This negation of the principles of the previous tradition is expressed in music with the destruction and rejection of the earlier harmonic, melodic, and sonata forms of tonality in favor of the extension of dissonance and ambiguity. This approach led to the necessity of revolutionizing the "language" of music itself and to experimentation with musical form (Burger 1984). The desire for experimentation is associated with the fascination with technology and science. Technology affected not only the means of production and reproduction but also comprised a new aesthetic stimulus in terms of the subject matter of music itself (Burger 1984, Poggioli 1982). In parallel with the interest in technology and science modernism expressed a belief in progress, constant innovation, and change. Modern artists defined their role as leading this process through a radical intervention in art and culture (Poggioli 1982). Another feature was the trend toward theoriticism. This theoriticism manifested itself in many theoretical texts which preceded the creative process and intended to legitimize it. A characteristic feature was the publishing of polemical manifestos by the different schools (Poggioli 1982). Finally, the different modernist movements positioned themselves between the extremes of rationalism and irrationalism, objectivism and subjectivism. 1.2 Music and Mathematics Xenakis explains that in 1954 he started considering the application of mathematics to music. The reason was twofold. First, it grew out of a musical necessity to go beyond the impasse of serial composition. Second, Xenakis was interested in the musical characteristics of various sound events encountered in the environment. Moreover, Xenakis was influenced by the new paradigm in modern physics and the scientific thought of his contemporaries. Those reasons led him to conclude that the application of mathematics was necessary in order to cope with the problems of musical composition (Xenakis 1990: 8-9). The use of mathematics and formal logic soon led Xenakis to employ computers in the compositional process and investigate the problems of composition more thoroughly as he was freed from the tremendous amount of work and time he had previously spent on complex calculations. The computer provided him the appro8
JOURNAL OF THE HELLENIC DIASPORA

priate framework to explore both algorithmic composition and sound synthesis. Algorithmic composition is based on the utilization of logical procedures in music. An algorithm could be defined as a sequence of instructions for accomplishing a task (Dodge & Jerse 1997). Algorithms can be translated into a programming language and executed on a computer. Furthermore, besides organizing the different aspects of a composition, the computer is capable of producing the sound itself. With the help of an analog-to-digital converter, computers allowed the composer to control the finest details of the sound synthesis process to a much higher degree than older analog methods. In Xenakis's own words: With the aid of electronic computers the composer becomes a sort of pilot: he presses the buttons, introduces coordinates, and supervises the controls of a cosmic vessel sailing in the space of sound, across sonic constellations and galaxies that he could formerly glimpse only as a distant dream. (Xenakis 1992: 144)

1.2.1 Breaking with the Musical Orthodoxies of the Postwar Era

In 1955 Xenakis published his first theoretical writing on music. It was an article called "The Crisis of Serial Music," which was included in the first issue of Die Gravesaner Bldtter. In that article he strongly criticized the theory and practice of serial music. Serialism was the dominant musical direction in Europe at that time. It was pioneered by Boulez and Stockhausen and evolved as an extension of the twelve-note method introduced by Schoenberg and developed by Webern. That article was acknowledged as a landmark in the evolution of the musical thought of the twentieth century. Xenakis observed the weakness of the serial system and concluded that this was due to linear thinking and the increasing density of polyphonic structures. He criticized the inner contradiction of the system where the organization of the musical elements was unrelated to the audible result. Linear polyphony destroys itself by its very complexity; what one hears is in reality nothing but a mass of notes in various registers . . . There is consequently a contradiction

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

between the polyphonic linear system and the heard result, which is surface or mass. . . . When linear combinations and their polyphonic super positioning no longer operate, what will count will be the statistical mean of isolated states and of transformations of sonic components at a given moment. . . . The result is the introduction of the notion of probability, which implies, in this particular case, combinatory calculus (reprinted partly in Xenakis. 1992: 8) Xenakis proposed that similar musical phenomena could be better described by statistical means that suggest a closer relation between the compositional method and the perceived sonic result. Moreover, he charged the serial system that the focus on geometrical symmetries lent a static feature to the music. In contrast, the use and manipulation of probability functions was capable of producing dynamic structures subjected to transformations. This introduced the concept of large-scale control of complex sound events rather than their analytic and mechanistic elaboration. This is exactly as in science where statistical methods were used as tools for the overall understanding of exceedingly complex phenomena. A second, more general, criticism was that serial manipulations were nothing but limited cases in the vast domain of combinatorial analysis. Why not a continuous spectrum of frequencies? . . . of timbres, intensities, durations? The serial principle was only a particular case of combinatory calculus. Why limit the combinations to only 12 terms, why not extend them? (Matossian 1986: 85)

1.2.2 The Statistical Character of Natural Events

Xenakis was attracted to many sound events in the environment, such as the cicadas in the summer, the raindrops falling on a rooftop, or the waves crashing on the cliffs. He observed that all of these sound events consisted of a large number of individual sounds that are perceived as one distinct sound. For example, the listener does not listen to each separate raindrop but perceives the phenomenon of rain as a unified sound. The dissociation of the above examples (the cicadas, the rain,
10 JOURNAL OF THE HELLENIC DIASPORA

or the waves) from their semantic context would reveal their physical properties as sounds. All of these sounds were subjected to the same statistical laws. These laws can describe the passage from complete order to total disorder in a continuous or abrupt manner (Xenakis 1992: 9). Xenakis was interested in analyzing similar sound events found in the environment (natural or urban) and utilizing their model in the compositional process. Therefore, a scientific analysis of the various phenomena was necessary in order to derive the associated laws.

1.2.3 The Paradigm of Modern Physics

Xenakis's musical thought had been influenced by the advances of modern physics. He delivered ideas of new physics into musical composition. The twentieth century was characterized by a radical change in the analysis and description of the natural phenomena (Kuhn 1962). A shift in the dominant paradigm occurred with the abandonment of the Newtonian model of a mechanistic universe subjected to deterministic laws, toward a relativistic model of a non-determinant universe. Xenakis introduced in his musical thought such concepts as the Theory of Relativity proposed by Lorenz-Fitzerald and Einstein, the theory of quantum mechanics by Max Planck, and the theory of indeterminacy by Heisenberg (Xenakis 1992: 255-67). As an example, according to the Kinetic Theory of Gases, the molecular density in a cloud of gas is such that it is impossible to calculate the momentum and the location of each individual particle. Therefore, what concerns the observer of the phenomenon is the properties of the whole mass of the particles, which is calculated statistically. Xenakis transferred this model into music and applied it in order to handle the large amount of musical information occurring in a composition.

1.2.4 Stochastic Music

The utilization of statistical methods in his compositions led Xenakis to name his music stochastic. He used this term in honor of probability theory. It comes from the Greek word "stochos" meaning "target" or "aim." The word stochastic was first used by the mathematician Jacques Bernoulli. Bernoulli defined the law of large numbers and invented the Calculus of Probabilities. The law of large numbers implies that the more numerous the phenomena

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

11

the more they tend toward a stable state, toward a goal, or "stochos" (Xenakis 1990: 4-5). I think this has been my main contribution to contemporary music: masses of sounds controlled like clouds by means of probabilities that shape the clouds statistically. (Varga 1997: 142) Xenakis was convinced of the need to introduce scientific thought and mathematics into composition. Pithoprakta (195556) for orchestra was the first composition in which he applied stochastic procedures in his music. Stochastic procedures are capable of handling large amounts of data as was the case of Xenakis's clouds of sounds. A stochastic procedure integrates random numbers in the compositional process. Next, the output of the random generator was weighted by the use of a probability distribution or distribution lookup table. Different probability distributions (uniform, linear, exponential, Gaussian, Poisson, Cauchy, Bernoulli, etc.) produce different results. A random walk is a stochastic procedure in which the successive results of a random generator are added together. The results of the above procedure can be limited between predefined boundaries. A random walk that makes use of a Gaussian propability distribution with a mean value of 0 is called Brownian motion (Dodge & Jerse 1997). Xenakis also made use of Markov chains. A Markov chain is a system whereby the probability of occurrence of a future event is determined by the occurrence or state of one or more events in the past. As such, Markov chains introduce a sort of memory into the stochastic process (Lorain 1980). Despite the application of mathematics and formal procedures in his music, Xenakis never felt that he would be restricted to a formal logic or that he should accept unquestionably the results of a stochastic process. In contrast to other composers, such as Hiller, he was prepared to alter, rearrange, and refine the outputs of the stochastic functions according to his own aesthetic criteria.

1.2.5 Toward Electroacoustic Music

In the postwar period many of the leading composers of the musical avant-guard were affected by the technological advances 12 JOURNAL OF THE HELLENIC DIASPORA

and turned to the electronic medium for the analysis and generation of sound. Xenakis's involvement with the electronic medium was influenced by two distinctive instances in the early history of twentieth-century music, the sound experiments and philosophy of Italian futurism, and the work of the French composer Edgard Varese. Futurist theorist Severini wrote that art should evolve hand in hand with science. The composer Luigi Russolo argued in his manifesto "The Art of Noises" (1913) that "musical evolution is paralleled by the multiplication of machines." In his music he tried to express the new urban and industrial soundscape. Rejecting the limited range of timbres of the conventional instruments, he proposed the construction of entirely new instruments capable in emulating the large variety of sounds encountered in the natural and industrial environment (Apollonio 1973). Edgard Varese, from 1920 on, called for new sound materials and, like the futurists, linked the progress of music to the development of new instruments. He imagined electronic instruments which would one day enable the composer to obtain "any vibration of sound, . . . any differentiation of timbre, of sound combinations, . . . a sense of sound-projection in space" (Varese 1971). He called his music "organized sound" and defined himself as "a worker in rhythms, frequencies, and intensities." His music was concerned more with timbre, texture, and rhythm and his compositions were based on abstract forms created by blocks and masses of sound. 1.3 Classification of the Electroacoustic Works of Xenakis Between 1957 and 1994, Xenakis officially composed sixteen electroacoustic works (Salabert 1997). His works can be classified into three categories according to the techniques he applied. These categories also correspond to three different periods in his compositional career (Solomos & Hoffman 1998). The first period can be defined between 1955 and 1977. During that period, Xenakis applied to his electroacoustic music mainly tape manipulation techniques and used both concrete and synthesized sounds. The works composed during this long period can be further subdivided into two separate categories: In the first sub-category there are the compositions realised in the studios of Groupes de Recherches Musicales

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

13

(GRM): Diamorphoses (1955), Concret PH (1978), Analogique B (1959), Orient-Occident (1960), and Bohor (1962). It is worth noting that during that period Xenakis proposed and applied the concept of granular synthesis as a new sound synthesis technique. In the second sub-category there are the compositions intended as the musical part of Xenakis's Sound and Light spectacles: Hibiki Hana ma (1969-1971), Persepolis (1971), Polytope de Cluny (1972), and La Ligende d'Eer (1977). Although Mycenae Alpha (1978) was also the music of a spectacle it is classified into the next category according to the applied techniques. Finally, the ballet music of Kraanerg (1968-69) can be added to this sub-category. The second period can be defined by the utilization of the UPIC system, developed in CEMAMu. Between 1978 and 1989 Xenakis composed four works with the UPIC system: Mycenae Alpha (1978), Nur la Paix (1981), Tauhiphanie (1987), and Voyage absolu des Unari vers Andromede (1989). In the third period, between 1991 and 1994, Xenakis used the GENDY program, with which he investigated stochastic waveform synthesis. With the GENDY program he composed two works: Gendy3 (1991) and S.709 (1994).

2.0 Electroacoustic Music Realized in the Studios of Groupe de Recherches Musicales and the Granular Synthesis Concept
Xenakis began working seriously on music in the early 1950s. At the same time he was attracted by the new idiom of musique concrete. There is evidence that during that time he attended the first concerts of the Group de Recherche de Musique Concrete (Matossian 1986: 76). In that period Xenakis was highly interested in systematically examining the physical properties of sound. The studio of the French radio provided him with the opportunity to study and experiment directly with sound material, investigate a new composing medium, and participate in one of the most important places of that period's musical avant-garde.

14

JOURNAL OF THE HELLENIC DIASPORA

2.1 Early Electroacoustic Music Techniques

The studio of the Groupe de Recherches de Musique Concrete was the first dedicated electroacoustic music studio ever. Initially the studio was equipped with a disk-cutting lathe, four turntables, a four-channel mixer, filters, an echo chamber, and a mobile recording unit. Later, in 1951, the studio reconstructed and expanded with new equipment. Among other appliances there was a three-track tape recorder, which was used for tape manipulation techniques; the Motphophone, a machine with ten playback heads to replay tape loops producing multi-tap delay and echo effects; the Tolana Phonogene, a keyboard controlled machine to replay tape loops at twenty-four different speeds each adjusted by the composer; the Sareg Phonogene, a machine that played tape loops at a continuously variable range of speeds controlled by a slider; and the Potentiometre d'Espace, which was a device that distributed live an encoded soundtrack across four loudspeakers, including one hanging from the center of the ceiling (Palombini 1999). Xenakis used the above mentioned equipment and applied various techniques, mainly tape-machine manipulations. Tape speed changes altered the character of a sound, raising or lowering its pitch. Having at his disposal equipment such as the Tolana Phonogene or the Sareg Phonogene, he could experiment with controlling the tape speed by using a keyboard or a slider. The above equipment allowed the composer to control the tape playback in real-time by applying discrete, pre-adjusted speed changes as is the case of the keyboard or continuous speed changes as is the case of the slider. The character of a sound could be altered also by playing the tape backwards where the sound was heard in reverse. Tape splicing and editing produced interesting results. For example, radical transformation could be achieved by removing the initial attack of a sound or even by replacing the attack of one sound with that of another. One sound could be edited and joined after another. The edit from one sound to the next could be performed abruptly or gradually by applying various degrees of cross-fades. Moreover, two or more recordings could be mixed into a third recording. Reverberation was applied to a sound by projecting it to the echo chamber and then recording the room echo. Echoes could also be produced artificially by using the Morphophone. The studio also provided various filters that could be used to alter the timbre of a sound by attenuating or boosting different parts of its spectral content.

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

15

Although Xenakis was involved in the activities of the Paris studio for almost a decade, his objectives, interests, and compositional practice were never identical to the goals of the center as defined by Pierre Schaeffer. Schaeffer's objectives were to disassociate the recorded sounds from their original context. With the help of studio techniques he aimed in transforming recognizable sounds into sound objects where the composer could rely solely on their acoustic properties released from their original semantic context. Schaeffer suggested that the members of the group should follow a specific program, based on studying and classifying sound objects. For that purpose he proposed that for a period of time musical works should not be composed. Xenakis did not accepted this condition and preferred to conduct his own investigations as associate member without participating in the studio's research project. He wanted to study the physical properties of sound and investigate musically the possibilities of the electronic medium. His approach toward the materials found at his disposal (musical or architectural, physical or artificial) was to examine them in a scientific way. Furthermore, he explored and stretched these materials to their limits. His attitude to composition and the techniques he applied were in contrast to the objectives of musique concrete. For example, most musique concrete of that time was characterized by abrupt contrasts and juxtapositions of the sound objects without any transition. In contrast, Xenakis applied in his compositions gradual movements and transformations of the sound material. For example, in Orient Occident (1960) he examined various degrees of transition by applying cross-fades of different durations. Xenakis utilized also dense mixing of different sound materials or of different transpositions of the same sound material. This resulted in composite sound masses where the individual components lost their individual characteristics and became parts of a larger sound object. He was also practised in delicate editing of innumerable tiny tape splices that produced the granular textures of works such as Concret PH (1958), Analogique B (1959), or Bohor (1962). When Xenakis composed the electroacoustic work Analogique B (1959), Schaeffer spoke very roughly about it. The main reason was the application of statistical methods in the compositional process and especially the use of sine-waves as the only sound material. Finally, when Xenakis proposed in 1963 the use of mathematics and computers as a studio direction, Schaeffer did
-

16

JOURNAL OF THE HELLENIC DIASPORA

not accept it. That led Xenakis to withdraw himself from the Groupe de Recherches Musicales.
2.2 Granular Synthesis One of the most important of Xenakis's contributions in that period was the proposal of an alternative method for sound synthesis. Granular synthesis is a unique method of achieving complex sound events by the generation of high densities of small portions of sound, called grains. A sound grain is characterized by a very short duration, typically between 1 to 100 msec. Probability functions are often used to control the large amount of parameter data for the grains. By applying granular synthesis, Xenakis was able to control the formation of clouds of sound down to the microsound level.

2.2.1 Background
Although Xenakis was the first to propose a compositional theory applying granular synthesis, the origins of a granular representation of sound can be traced earlier in time. Xenakis confesses in his writings that when he proposed granular synthesis he was not aware of any previous related theory. He conceived the granular synthesis concept purely by intuition (Varga 1997: 197). It was many years later that he acknowledged the theories of quantum mechanics by M. Planck and Albert Einstein as the conceptual basis of granular synthesis (Xenakis 1992: preface). The British physicist Dennis Gabor was the first to propose a scientific theory based on a granular notion of sound. His theory proposes an alternative concept than the established Fourier analysis. According to Gabor: Fourier analysis is a timeless description in terms of exactly periodic waves of infinite duration. On the other hand it is our most elementary experience that sound has a time pattern as well as a frequency pattern. A mathematical description is wanted which ab ovo takes account of this duality. (Gabor 1947: 591) Therefore, Gabor's mathematical description took into account and correlated the time-domain (waveform) and the fre-

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

17

quency-domain (spectrum) representation of sound to obtain an acoustical quantum. These acoustical quanta are units of acoustic information. Each quantum can be represented as a harmonic oscillation within the range of audible frequencies, modulated by a short duration envelope. Furthermore, Gabor suggested that the human mechanism of hearing is not continuous and does not involve infinite resolution. Within a very short time limit (10-20 msec) the ear is capable of identifying only one sound event. In addition, the hearing mechanism is conditioned by discrete thresholds in the perception of frequency and amplitude (Roads 1985). In 1964 the cybernetitian Norbert Wiener argued that time, space, and matter are discontinuous: "In light as in matter, there is a granular rather than a continuous texture." He also doubted the representation capabilities of Fourier analysis of sound and proposed a granular approach (Wiener 1964). Abraham Moles, taking into account that the human hearing mechanism divides the frequency and amplitude dimensions of sound into quanta, proposed a segmentation of the audible frequency/amplitude plane in order to measure and represent the information of each sonic event. In advance, he estimated that the resolution of the ear is limited to 340,000 elements (Moles 1968: 108-110).

2.2.2 Granular Synthesis Proposal by Xenakis

Xenakis devoted two chapters of his book Formalised Music to the theory and applications of granular techniques. There he defined the granular concept: All sound is an integration of grains, of elementary sonic particles, of sonic quanta. Each of these elementary grains has a threefold nature: duration, frequency, and intensity. All sound, even all continuous sonic variation, is conceived as an assemblage of a large number of elementary grains adequately disposed in time. So every sonic complex can be analysed as a series of pure sinusoidal sounds even if the variations of these sinusoidal sounds are infinitely close, short, and complex. (Xenakis 1992: 43) According to the granular theory, any sound can be con18
JOURNAL OF THE HELLENIC DIASPORA

structed from a number of short duration acoustic elements. This elementary units are called grains. Xenakis realized the grains from sinusoidal sounds. The duration of each grain is constant, around 40 msec. Each grain should be shaped with an envelope. Xenakis used a rectangular envelope (Roads 1978). As mentioned above, granular synthesis is based on the discontinuous character of the hearing mechanism. Thus the frequencies and the intensities of the audible sounds can be quantized into discrete steps according to the thresholds of the hearing mechanism. Consequently, we can represent any possible frequency/loudness pair by forming the appropriate grain, Xenakis claims that the total number of elementary audible grains is about 340,000, a number identical to that suggested by Moles. The human hearing mechanism is not uniformly sensitive at all frequencies. This occurs due to the resonant frequencies of the outer ear which have a peak in the middle frequency region. Fletcher and Munson devised a set of curves to show the sensitivity of the ear at different frequencies across the audible range. Xenakis proposed to apply those curves to a one-to-one mapping of each of the 340,000 elementary units in order to represent the audible area in a homogenous manner (Xenakis 1992: 47) Xenakis used the concept of screens and books to represent a sound. Each screen represents a slice in time. The dimensions of the screen correspond to frequency and loudness, limited to the audible area. Each screen contains a grid, the cells of which can be filled with points representing the frequency and loudness of each grain. A variety of combinations of grain positions in a screen can be accomplished. For example, a screen can be limited to only one grain representing the occurrence of only a sinusoidal sound, or can be totally empty, representing silence. A completely filled screen with grains occupying the whole audible area represents a sound close to white noise. Therefore, Xenakis suggests that with the appropriate completion of the screens, a large variety of sounds can be produced. A book is a sequence of screens, describing the occurrence and attributes of grains in time, and therefore the evolution of a whole sound. As we move from one screen to another we notice that some grains remain at the same positions, some grains disappear, while other grains appear in positions that were not occupied before. This is representative of the transformation of a sound, where its

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

19

spectral characteristics evolve continuously. Finally, we can observe that the concept of screens and books is equivalent to a movie film where each screen corresponds to a film frame, while a book corresponds to a film sequence. Each sound can be represented by the formation of a cloud of grains. Even a sound with short duration can require the occurrence of hundreds or even thousands of grains. Therefore, the granular representation of sound requires the handling of a large amount of data. Since each grain can be defined by a number of parameters (frequency and amplitude in the simplest model) the amount of required data is a factor of the number of grains. The fine tuning of all of the parameters for each grain led to the necessity of the development of a higher level organization and control. This led Xenakis to shift his interest from the attributes of individual grains to the attributes of whole clouds of grains. He dealt with the problem of controlling large amount of data by using statistical methods. This attitude is equivalent to the shift from the methods of classical mechanics to the Maxwell-Boltzman Kinetic Theory concerning the description of a gaseous mass, where the scientist does not concern himself with details but tries to calculate the trajectory of the mass of particles statistically (Xenakis 1992: 49). According to Xenakis, to work like architects on the sonic material in order to construct complex sounds and evolutions of these entities means that we must use macroscopic methods of analysis and construction. Microsounds and elementary grains have no importance on the scale which we have chosen. Only groups of grains and the characteristics of these groups have any meaning. (Xenakis 1992: 50) Xenakis defined three factors that are important in the higherlevel control of the grains: density, topographic situation, and entropy, or the degree of order or disorder. Density refers to the number of grains that occur at the same interval of time. The higher the density of the events the more complex and thick is the produced sound. Topographic situation refers to the placement of the grains on the screen. Since the placement on the screen determines the frequency and amplitude of each grain, this factor determines the spectral characteristics of the sound. Entropy, the degree of order or disor20
JOURNAL OF THE HELLENIC DIASPORA

der in a group of elements, is a more complicated factor. Entropy in relation to time refers to the regularity of time intervals between the occurrence of each grain. Zero entropy means that a grain will be emitted in identical intervals in time. Entropy in relation to frequency refers to the degree of variation of the frequency of the grain. Zero entropy means that the grain can have only one stable frequency. Entropy in relation to loudness refers to the degree of the variations of the loudness of the grain. Finally, entropy in relation to density refers to the degree of the variations of the density of the grains (Xenakis 1992: 58-68). For the initial formation of screens, Xenakis proposed the application of statistical distribution of the grains. Therefore, the grains are placed randomly on the initial screens. Then, new screens can be constructed using set-theory operations like intersection, union, complement, and difference. For example, intersection of two screens produces a third screen containing grains that are common to both first two screens, while union produces a third screen containing all the grains of both first screens (Xenakis 1992: 58). Linking the screens is the final procedure for defining the evolution of a sound. Xenakis used probability functions, and Markov chains in particular, to define the succession of screens. Therefore, the occurrence of a screen determines the possibilities of the occurrence of the next new screen. Xenakis envisioned granular synthesis as an effective method for the simulation of the timbres of classical instruments and environmental sounds, as well as in the creation of rich new sounds. In fact within human limits, using all sorts of manipulations with these grain clusters, we can hope to produce not only the sounds of classical instruments and elastic bodies, and those sounds generally preferred in concrete music, but also sonic perturbations with evolutions, unparalleled and unimaginable until now. The basis of the timbre structures and transformations will have nothing in common with what has been known until now. (Xenakis 1992: 47)

2.2.3 Observations on Xenakis' Granular Synthesis Approach

The conception of granular synthesis by Xenakis is of great importance to the field of sound synthesis and electroacoustic composition for the following reasons:

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

21

First, he proposed a new synthesis method. Granular synthesis is a technique capable of producing a vast range of rich sounds. The complexity and dynamic quality of the sounds produced by granular synthesis makes the technique an attractive competitor for synthesis techniques based on fixed waveforms and envelopes (Truax 1988: 24). Second, Xenakis defined a graphical representation of the synthesis components, the grains. A graphical representation allows the composer to visualize various parameters of the components of a sound, thus better understanding the characteristics of sound. Third, Xenakis became aware of the necessity for a higherlevel organization of the grains and the macroscopic control of the sound. As granular synthesis of sound uses thousands of grains, the composer's focus must shift from the attributes of individual grains to the attributes of whole clouds of grains. Fourth, granular synthesis is a first attempt to control both the microstructure and the macrostructure of a whole composition by applying common compositional techniques. By applying stochastic methods and by using the electronic medium, Xenakis attempted to control all the levels of the composition from the individual waveforms to the structure of the piece. Fifth, the electroacoustic techniques of that period and especially the granular synthesis concept was Xenakis's first attempt to fill the gap between the two antagonistic directions of musique concrete and electronische musik. Granular synthesis could use electronic sounds as one of the properties of its building blocks, the grain, and was capable of producing timbres whose qualities could be compared to those of environmental sounds (Truax 1988: 25).

2.3 Compositions Realized in the Studios of GRM Diamorphoses (1957) is the first of Xenakis's electroacoustic
works. He began working on it in January 1957 and completed it six months later. Diamorphoses was first performed one year later, on October 5, 1958, at the Journees Internationales de Musique Experimentale in Brussels. The work is realized for four-channel tape and its duration is 7 minutes. According to Xenakis, the objectives of this work were: to mix timbres in order to arrive at a body of sound like white noise; to study the evolution of timbres, dynamics 22
JOURNAL OF THE HELLENIC DIASPORA

and register; to make unisons with attacks only with or without transposition: to make chromosomes of attacks. (Matossian 1986: 125) The source material of the composition consists of recordings of crashing cars, jet engines, earthquake shocks, as well as of bells, various precessions, and wind musical instruments. Concrete P.H. (1958) was designed to accompany the entrance and the departure of the audience into the audio-visual spectacle of the Philips Pavilion for the Brussels World Fair of 1958. The Pavilion was designed by Xenakis himself while the main music of the spectacle was Edgar Varese's electronic composition, Poem Electronique. The title of the work has a double reference. The first part of the title (concrete) refers to musique concrete and the activities of the GRM. The second part of the title (P.H.) corresponds to the initials of the words hyperbolic paraboloids, which refers to the S-curve architectural design of the surfaces of the Philips Pavilion. The duration of Concrete P.H. was 2 minutes and 45 seconds. The work was realized for two-channel tape, and in its original performance in the Philips Pavilion the sound was distributed over 350 loudspeakers. The sound material of the work is based on the recorded sounds of burning charcoal. The compositional techniques were restricted to simple tape manipulations like splicing, layering, and transposition, which, however, were highly extended. Thus, the composer involved in the splicing of innumerable pieces of tape, and the juxtaposition of many layers of sound material to obtain various degrees of sound density. Analogique B (1959) was the tape part of the instrumental piece Analogique A (1959) for nine string ensemble. The composition was realized for four-channel tape and lasts 7 minutes. The premiere of Analogique A+B took place at the Gravesano festival in 1959 under the direction of Hermann Scherchen. Analogique B was Xenakis's most systematic investigation into the creation of clouds of sounds by technical means on that period. Xenakis implemented granular synthesis by utilizing a sine wave generator and recording sine waves of different frequencies and amplitudes on tapes. Next, he spliced the tapes into innumerable tiny pieces, each representing a grain of sound. Finally, he edited these tiny fragments into sequences and mixed them together. For the organiza-

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

23

tion of the piece, Xenakis used stochastic methods (Poisson Law and Markov Chains). Orient-Occident (1960), was realized as the music for a film of the same name by Enrico Fulchignoni, commissioned by UNESCO. The film demonstrates and compares products and works of art from various civilizations, from prehistoric times to Alexander the Great. The composition is for two-channel tape and its duration is 12 minutes. The premiere of the work took place on May 1960, in Cannes. Xenakis derived his sound material from different sources. For example, he drew a cello bow across various objects such as cardboard boxes, sheets of metal, and various sizes of gongs and tam-tams. Another source was signals from the ionosphere which transformed into sound. Xenakis also used a recorded section of his earlier orchestral work Pithoprakta (1955), which transposed and slowed down in the studio. Bohor (1962), for eight-channel tape, is the last composition realized in the studio of GRM. The duration of the work is 21 minutes and 30 seconds. In composing Bohor, Xenakis used the sounds of various bracelets and jewelry from the Middle East as well as a Laotian mouth organ. In the studio, Xenakis was able to expand the tiny sounds of the bracelets and investigate the different sonorities in them. The main techniques employed in the realization of the work were tape speed manipulations and the layering of sound material through mixing. The work was conceived as a single massive crescendo with gradual inner evolution. According to the composer's own words, it can be characterized as: monistic with internal plurality, converging and contacting finally into the piercing angle at the end (sleeve notes: Bohor I Orient-Occident I concrete PH / Diamorphoses. Nonesuch H-71245).

3.0 Polytopes: The Light and Sound Spectacles

Almost ten years after he participated in his first large-scale project, the design and construction of the Philips Pavilion and the composition of Concrete PH performed therein, Xenakis started creating a series of light and sound spectacles, this time organized and produced independently. This corresponds to the period 1967 24
JOURNAL OF THE HELLENIC DIASPORA

to 1977, which is considered Xenakis's second period in electroacoustic music. One of Xenakis's biographers recounts that an early influence in the conception of these spectacles was the night-time bombing and machine gunning during the war in Greece (Matossian 1986: 212). In the Second World War, Xenakis witnessed many night bombings in Athens. He saw before his eyes an incredible spectacle of explosions and light beams produced by the tracer-bullets from the anti-aircraft guns. Those images, accompanied by the frightening roar of the bombers and the sounds of the machineguns, were etched forever in his memory (Matossian 1986: 213). Xenakis created six light and sound spectacles in total. This category of works are best known under the name Polytopes. The word Polytope is conceived by Xenakis and is a coinage from the Greek words poly, which means "many," and topos," which means "place." Only two of the works originally had the name Polytope: the Polytope de Montreal (1967) and the Polytope de Cluny (1972). However, all the works share common attributes and therefore could be classified in the same group. 3.1 Composing Space With his Light and Sound spectacles Xenakis wanted to create a new kind of abstract art in which both the musical and the visual part would be organized with the application of common notions and techniques (Varga 1996: 112). Xenakis would deliver his experience in musical composition and mathematics in the organization and control of the whole spectacle. Composing with sounds for the ear leads us to compose with light for the eyes. The laser beam and the electronic flash are the equivalents of beautiful sounds. To make them gleam in space is to create music for the eyes . . . The music for the eyes is created with concepts and procedures stemming from musical composition. The result is a new art of vision and hearing that it is neither ballet nor opera, but really an abstract spectacle in the sense of music, of the astral or terrestrial type. (Xenakis 1985b: 182) Xenakis, as early as his arrival in Paris and his association with Le Corbusier in 1947, began to interrelate the problems he

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

25

encountered in architecture with that of music composition. He followed a different path instead of the traditional approach to composition. He preferred to start working with the form and the general features of the composition and then to examine the individual elements and their relations. I found that problems in architecture were the same as in music. . . . Instead of starting from a detail, like a theme, and building up the whole thing with rules, you have the whole in mind and think about the details and the elements and, of course, the details . . . I thought that the best way to attack the problem was from both ends, detail and general. (Matossian 1985: 69) Xenakis's common approach to both music and architecture can be observed in the design of the Philips Pavilion in 1958 (pict Matossian: 114 and 115). There Xenakis used ideas and solutions that he had applied earlier in his orchestral composition Metastasis (1953). The form of the Pavilion and the use of hyperbolic paraboloids originated from the ruled surfaces that formed the string glissandi of Metastasis (Xenakis 1990: 10). Despite the analogies observed between the architectural and the musical space, Xenakis had recognized the multidimensional character of the latter instead of the three-dimensional character of the former. In addition, The common attitude against his materials led Xenakis to organize the audio and visual part of the Polytopes, where he handled light and sound as architectural elements. Finally, one should note Xenakis's idea of art as a cosmic landscaper. Xenakis imagined a future art form that would work as an extension of the light and sound spectacles, where the artist works on the cosmic level. Some first proposals of this kind of art included the utilization of powerful laser beams projected onto clouds, or satellites used as reflecting mirrors that would create a spectacle governing the earth and the moon on an enormous level (Xenakis 1985a: 5). 3.2 Synthesis Techniques in the Polytopes The music for the Polytopes was composed after Xenakis's departure from the studios of Groupe de Recherches Musicales (GRM) in 1963. Xenakis's sound synthesis approach on these 26
JOURNAL OF THE HELLENIC DIASPORA

works was similar to that of the period with the GRM. He primarily used tape manipulation and sound-processing techniques. The music had the same statistical aspects and Xenakis continued to use the granular concept to govern some of his sounds. Although he never investigated granular synthesis more than he did in Analogique B (1959), the music of the second period was also characterized by the gradual transformation of large clouds formed by tiny sounds. This was achieved by dense mixing of many layers of sound. In addition, he continued to process both instrumental and concrete sounds. He used sounds of instruments from all over the world, sounds of many non-musical objects and sounds from recordings of his earlier orchestral pieces. Another characteristic of the Polytopes was the application of multitrack tapes and the prominent spatialization of sound. The tape for the electronic works was prepared for seven to twelve audio tracks. The tracks would be distributed over a large number of loudspeakers (up to 800, as is the case of Hibiki Hana Ma). His only previous use of a large number of loudspeakers was in Concrete P.H. (1958). 3.3 Electroacoustic Compositions for the Light and Sound Spectacles

Hibiki Hana Ma (1969-1970) could be translated as "Reverberation-Flower-Interval." It was commissioned for the Japanese Iron and Steel Federation's Pavilion at the Osaka Exhibition in 1970. The tape of the composition was created at the Radio Nippon Horo Kyokai (NHK) studio in Tokyo. The duration of the spectacle was 18 minutes. The twelve tracks of the composition (two synchronized six-track tape machines) were distributed over 800 loudspeakers. The music was played synchronized with a laser show designed by the Japanese sculptor Keiji Usami. Both the visual and the musical part of the spectacle were closely related as they followed in parallel the output of a specially designed stochastic program. The sound material of Hibiki Hana Ma is based entirely on instrumental sounds. Xenakis processed the recorded sound of an orchestra, as well as recording and processing a snare drum and a Japanese instrument called a biwa. Despite the sound processing, all the sounds on the piece maintained some of their original features and therefore remained recognizable. Aesthetics and Techniques in the Electroacoustic Music of Xenakis
27

Persepolis (1971), was premiered on August 26, 1971, in the ruins of the ancient city of Persepolis in Iran. It was commissioned for the Shiraz Festival for Arts and Music. The eight-track tape was composed at Studio Acousti in Paris. The duration of the work was 56 minutes. The spectacle was staged in the ruined palace of Darius, in Apadana, while the light show took place on the nearby hills. The near-desert conditions of the environment did not allow Xenakis to utilize a spectacle dependant on high technology. He used a multitrack tape playback system for the music reproduction and organized various performance events for the visual part. The sound was projected into space by a number of loudspeakers that was placed around the audience in three concentric circles. The composition was based on sound material, such as various Japanese wind instruments, sounds derived from bowed or shaken sheets of metal, and sounds from various jewelry, similar to those used in Bohor. The composition featured a grainy texture, subjected to gradual transformations. There was a profound contrast between the slowly evolving monolithic macrostructure of the work and the intense inner activity of the individual sounds. Hundreds of tiny sounds produced rough, metallic sound surfaces with distinctive spatial distribution. The Polytope de Cluny (1972) was staged at the Roman Baths of Cluny, in Paris. The installation premiered on October 13, 1972. A second version was played on the 17th of the same month. After that the spectacle was repeated several times during a day for a period over three months. Xenakis for the first time had at his disposal the appropriate computer technology to control and organize the complex configurations of the audio and visual part of the spectacle. The soundtrack was realised at the Studio Acousti in Paris. The duration of the eight-track tape was 24 minutes. For the visual part of the installation, 600 white flashes and 400 mirrors were staged over the walls and ceiling of the T-shaped hall. There were also three powerful lasers producing green, red, and blue light beams. The laser beams were pointed to the mirrors and were reflected many times by them. Xenakis had precisely calculated the angles of the mirrors in order to accurately reflect the laser beams and form various configurations of light in space. In addition, the mirrors were motorized so they could change orientation as a function of the program, thus forming different light configurations. The 600 lights were also programed to flash and form dif28
JOURNAL OF THE HELLENIC DIASPORA

ferent patterns that could be divided into three categories. First, various geometric shapes such as circles, spirals, or crossing lines. Second, the emulation of various natural phenomena such as galaxies, northern lights, rivers, ponds, etc. Third, patterns produced by stochastic functions. In contrast to the complexity of the visual part of the Polytope, the musical part was much simpler. The sound was distributed around the audience and was characterized mainly by modulating timbres and rhythmic pulses (Matossian 1986:221). La Legende d'Eer (1977) was the music of the Diatope, staged on the Place Beaubourg, outside the Pompidou Center in Paris to celebrate its opening. The work was created partly at the Centre d'Etudes Equipe de Mathematique et d'Automatique Musicales (CEMAMu) and partly at the Westdeucher Rundfunk (WDR) in Cologne, where the work was commissioned. It was first performed on February 11, 1978. The title of the work came from Plato's Republic, excerpts of which were included in the program. Xenakis also included three other texts in the program: Poimandres from Hermes Trigemestre, a discussion on the Infinite in Pascal's Pensees and a text of Robert Kirchener on supernovae. The duration of the seven-track tape is 46 minutes. The sound sources could be divided into three categories: recorded instruments from all over the world, such as the African mouth harp and the Japanese tzouzoumi, various sounds created by different objects, for example clapping wooden blocks or rubbing materials against one another, and electronic sounds created in the studio, either using the UPIC system or by applying mathematical functions in a first attempt at stochastic synthesis (Toop 1995). All these sounds were later treated by filtering, reverberation, transpositions, etc. The spatial position of the sounds also played an important role in the composition. Each of the seven audio tracks were distributed over eleven loudspeakers. In this work Xenakis not only created the audio and visual part of the spectacle but also designed the architecture of the Diatope. Xenakis decided on an architectural form that made use of hyperbolic paraboloids, the smooth curved surfaces that he used previously in the Philips Pavilion. The Diatope was 16 meters high and it was covered by a 1,000-square-meter red vinyl. The floor was made of translucent glass tiles. For the visual part of the spectacle, Xenakis used 1,680 flash lights, four lasers, and 400 rotating mirrors. The lights changed their state

Aesthetics and Techniques in the Electroacoustic Music of Xenakis 29

every 40 msec ( 1/25 sec), producing the illusion of continuous movement. The laser beams were reflected by the mirrors in a relevant manner as in the Polytope de Cluny. The commands that controlled the flashes, the laser beams, and the mirror positions were stored on a magnetic tape. The commands for the volume changes and the distribution in space of the seven audio tracks were also stored on the same tape. Mycenae Alpha (1978), was the music for a light and sound spectacle that was first performed on August 2, 1978, in Mycenae, Greece. For the realization of Mycenae Alpha, Xenakis used the UPIC system at CEMAMu. Therefore, this work is classified together with the compositions realized with the UPIC system (paragraph 4.4). Finally, Kraanerg (1968-1969) can be added to this period. It was not a spectacle but a ballet music for orchestra and four-channel tape. It was first performed on the June 2, 1969, in Ottawa for the inauguration of the National Arts Centere. The duration of the work is 75 minutes. Together with Analogique A+B they are the only compositions in which Xenakis combines physical instruments and electroacoustic sounds. The tape part was composed by recorded and processed orchestral sounds. According to Xenakis, the concept that underlies Kraanerg is "the biological struggle between generations . .. destroying existing political, social, urban, scientific, artistic, and ideological frameworks on a scale never before attempted by humanity, and unforseeable" (Toop 1989). 4.0 The UPIC System UPIC is a music composition and sound production system. The name is an acronym of Unite Polyagogique Informatique du CEMAMu. The word Polyagogique is a coinage from the Greek words poly, which means "many," and agogie, which means "training." Xenakis used this coinage to express the ability of the composer to control and interact with the various levels of the composition (symbols, syntax, structure, sounds, etc.) simultaneously. UPIC is a computer-based system where the user can graphically describe all the levels of the composition with the aid of an electronic tablet and listen to the result in real-time. It was devel-

30

JOURNAL OF THE HELLENIC DIASPORA

oped in CEMAMu (Centre d'Etudes Equipe de Mathematique et d'Automatique Musicales) and has been used by many composers. Xenakis composed four electroacoustic works using this system either entirely or in part: Le Legende d'Eer (1977), Mycenae Alpha (1978), Taurhiphanie (1987), and Voyage Absolu des U nari vers

Andromede (1989). 4.1.0 Background

Xenakis first started using graphical design to give visual representation of his works and to describe various aspects of his compositions in his first instrumental pieces. This technique derived from his architectural background as well as from the needs of the compositions themselves. The first piece in which he used graphics was Metastasis (1953-54) for sixty-one instruments. In this work he makes extensive use of glissandi, where the notes played by the strings glide continuously from one pitch to another covering a broad band of the spectrum. The evolution of the sound creates sound objects, which can be compared to the architectural objects he was working on. To visualize the form of the whole composition he used graphics in which he mapped the pitch dimension versus the time dimension. In his later works he used arborescences, where he mapped various dimensions of the composition (most commonly pitch versus time) into a Cartesian system of coordinates. This design gives a sense of sound continuity and clearly shows the evolution of the sound-masses, an integral aspect of his compositional processes. In spite of this, Xenakis still had to write the piece in traditional notation to prepare the parts for the musicians. This transcription of the graphical design took a great amount of work and time. Xenakis also believed that music creation should be open to everyone. For him, composition frees the imagination and other creative powers within man. Therefore, everyone should be able to get involved in the creative process regardless of his musical background. These thoughts led him to the conception of a system where the graphical description of the music could be transferred into sound. He believed graphical notation to be more general than traditional notation and therefore more easily understood by

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

31

everyone, even by little children. This approach could free the user from the restrictions of long-term studies on solfege, notation, and music theory. The system could also support people without proper training or dexterity on an instrument. The composition could be refined continuously, as the audible result would be available instantly. Finally, according to Xenakis, such a system could lead to a new approach to musical thinking, which would be positive for both the individual and society.

4.1.1 CEMAMu / Les Ateliers UPIC

Xenakis aspired to establish an institute devoted to research, composition, and pedagogy. The foundation of such a center would allow him to concentrate on his own specific needs, visions, and beliefs. He always had difficulties finding a studio to carry out his research and composition. In addition, he wanted to promote his ideas and he was very conscious of his social and pedagogical role as an artist and researcher. He didn't want to work in isolation and he had strong beliefs about the necessity of a new way of thinking, learning, and creating music. First, he founded EMAMu (Equipe de Mathematique et d'Automatique Musicales) in 1966. The institute's aim was to design equipment suitable for computer-aided sound synthesis and music production based on scientific thought and mathematics. It was located in CNET, the French National Center for Telecommunications Research. Later, in 1972, EMAMu became CEMAMu (Centre d'Etudes Equipe de Mathematique et d'Automatique Musicales). CEMAMu is a nonprofit organization whose activities are supported by the French Ministry of Culture. The initial objectives of the institute can be summarized in five points: 1) Fundamental research in musical and visual composition aided by various scientific fields such as mathematics, acoustics, and computer science; 2) Development of user-friendly tools such as computer systems, specialized programs, and methodologies; 3) Production of contemporary music using both the 32
JOURNAL OF THE HELLENIC DIASPORA

results of the research and the developed tools of the institute; 4) Establishment of new pedagogical methods for music and associated scientific fields with the cooperation of universities, conservatories, and other cultural organizations. 5) Dissemination of the knowledge gained by the research and the availability of the developed tools to a broad number of people in order to change the methods of creation and musical thinking (Lohner 1986a; Raczinski et al. 1992). Since 1986, the compositional and the pedagogical activities have been carried out by Les Ateliers UPIC, which consists of three electronic music studios. UPIC is the central system for each one of the studios, accompanied by a variety of computers with specialized music software for composition and sound synthesis, sound modules, effect units, mixing desks, recorders, etc. The studios are devoted to the creation of electroacoustic music. Many works are commissioned and numerous composers are encouraged to reside and work there. The pedagogical role of the institute is complemented by organizing workshops or classes on UPIC, there and throughout the world.

4.2.0 Description of the UPIC System

The UPIC system consists of a real-time unit, which handles all the sound synthesis and playback, a host computer running the human interface, and an optional digital graphic board. The concept is that the composer uses graphical notation to describe the various aspects of the composition. With the help of a graphical tablet or a computer mouse one can describe the music from its tiniest microlevel detail to its whole macrostructure.

4.2.1 The Drawing Board

The part of the system closest to the composer is the graphic digitizer. Although in UPIC's third version it was possible to substitute it for the computer's mouse, this graphic drawing board is still considered an integral part of the functionality and the expressiveness of the system. This board is a conductive pad consisting

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

33

of a network of numerous electronic fibres. The composer uses an electromagnetic pen to draw on the digitizer. When the pen tip touches the surface, the tablet senses the coordinates and sends the data to the computer. UPIC supports various sizes of digitizers, from A4 to A0, but the standard drawing area is around 60 cm high and 75 cm wide, giving an accuracy of 0.25 mm (Lohner 1986a). The tablet plays the role of the score while the composer uses arborescences divided into millimeter lines instead of the traditional staves.

4.2.2 The Human Interface

Everything digitized on the drawing board is sent to the computer. The UPIC software runs on the Windows operating system and therefore it shares all the characteristics of the latter. The program environment of the system consists of various overlapped, resizable and zoomable "windows." Each "window" has its own functionality in the system and Xenakis calls them graphic objects. All of the objects are somehow interconnected. Every drawing made on the board refers to a specific object and it is pictured on it. Therefore the composer has simultaneous access to every parameter of the composition.

4.2.3 The Graphical Objects

The main graphical objects are the wavetable, the envelope, the page, the frequency table, the amplitude table, the sequence, and the audio channel. Each object according to its functionality can be composed from one or more draws. Xenakis calls each of the draws an arc. The wavetable object can be considered as the timbral source of the system. The wavetable holds a memory of the system in which digital data of one cycle of the sound waveform is stored. The table can be read repeatedly and the waveform is heard in various pitches according to the reading rate. The waveform can be drawn freely by the composer or it can be any extracted part of a digital recorded sound. The UPIC can hold up to sixty-four different waveforms. The envelope object describes the change of one of the attributes of the sound (usually amplitude) over time. In early sound synthesis the envelope consists of four segments: attack, decay, sustain, and release (ADSR). In UPIC, any shape can be drawn and the system can hold up to 128 different envelopes. The function of the envelope is applied not only to individual sounds but also to
34 JOURNAL OF THE HELLENIC DIASPORA

the whole composition. The program allows the extraction of the amplitude envelope from any recorded sound. The page object can be compared to the traditional staves score. Both describe the evolution of the composition over time. The horizontal axis represents time while the vertical axis represents pitch. The analogy cannot be taken further, though, as the UPIC system is designed to liberate the composer from the limitations of traditional notation. The time axis is read from left to right and the total duration of the page can be programed to last from 0.2 sec to any duration, limited only by the storage capacity of the hard diskc. The user is provided with various drawing modes (free hand, segmented line, etc.) to design each arc. The composer can draw up to 4,000 arcs on a page and there can be up to four pages active any time. The vertical axis is read from bottom to top. The number of the arcs that can be played simultaneously corresponds to the polyphony of the system, which at present limits to sixty-four oscillators. It should be obvious that a horizontal line would be heard as a note in stable pitch, while a diagonal line as a glissando. A dot has no actual meaning to the system since it would be read as a note with minimal duration or no duration at all and therefore it would not be heard. The arcs in each page can be grouped into four groups according to specific criteria defined by the composer. Each group can share its own attributes as waveform, envelope, etc. As its name suggests, the page object corresponds to a page of music. A whole composition can be completed by linking different pages one after the other. The frequency table describes the way that the page object translates the values of its vertical axis. First, it gives information on the frequency range of the page, limiting the boundaries between an upper and a lower value in Hertz. Second, it gives a description of how all the values inbetween are interpreted. If the interpretation is continuous, all the frequencies could be produced. If the interpretation is in discrete steps, then all the frequencies that fall outside these steps would be quantized to the closest valid value. This corresponds to the concept of scales. UPIC is capable of the design of an unlimited number of musical scales. Any scale, tempered or not tempered, could be constructed. For instance, UPIC can construct any tempered scale where the octave is divided into one to ninety-nine equal steps. There can be up to four frequency tables available per page.

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

35

The amplitude table functions in the same way as the frequency table object. Instead of frequencies it describes the interpretation of amplitudes. It works in conjunction with the envelope object. The sequencer object interprets the way the UPIC plays a page. UPIC plays back a page by reading it from left to right with a constant tempo predefined by the composer. In addition, the user can conduct the playback of the page by moving the pen on the board. In this performance mode the user can accelerate or slow down the tempo, jump from one location on the page to another, and read any portion of the page forward or backward. All these movements can be recorded and displayed on the sequencer object. The drawing on the sequencer can later be edited as any other arc in the system. UPIC can record up to 12 minutes of a performance into a sequence. Any part of the recorded sequence can be edited either by recording a new performance again for the specific part or by redrawing the sequencer arc. The accuracy of the system is limited to 6 msc, a resolution capable of capturing very fine details in the performance. There can be two sequencer objects available any time. The sequencer object can be assigned to control any page, which means that a performance on one page can be freely applied to any other. 4.2.4 The Host Computer The host computer can be any PC-based system with hard disk, a mouse, and a MIDI interface. The graphical digitizer and the real-time unit are both connected to the host computer. The requirements of the computer are not demanding since all the number-crunching computations are handled in the real-time unit. 4.2,5 The Real-time Synthesis Unit The real-time synthesis unit is dedicated to the sound synthesis of the system. The polyphony of the system is sixty-four oscillators. The unit performs in real-time, giving the composer immediate audible results. On board there are also two high quality analog-to-digital converters, allowing the composer to record from any sound source. There are also four digital-to-analog converters that play back the synthesized sounds. Since the implementation of the first version in 1977, there have been three more versions of the system. The second version added mixing capabil36
JOURNAL OF THE HELLENIC DIASPORA

ities in 1983, while the third version in 1987 was the first realtime system. In 1991, CEMAMu developed the first commercial UPIC system to use the Windows environment as its human interface. Nowadays, there are several UPIC systems installed throughout the world. 4.3 Observations on the UPIC System The graphical approach of the UPIC system is clearly an implementation of Xenakis's concept of outside-time structures. The graphic representations deal with parameters such as pitch and amplitude without any fixed reference to time. The time element is added later by the sequencer object. As the graphics can be played or randomly accessed forward and backward at variable speeds, the graphical parameters are clearly outside-time. Thanks to the metrical nature of time, one can furnish it within outside-time structure . . . In this way, time could be considered as a blank blackboard, on which symbols and relationships, architectures and abstract organisms are inscribed. (Xenakis 1992: 192) The system was designed in the context of the visionary optimism of the modernist movement. The goal was to give more people access to the creative process by designing a simple and intuitive interface. This idea led Nelson and many other researchers to argue that Xenakis attempted to implement a universal language with the UPIC system (Nelson 1997). However, on closer examination, such an argument turns out to be untenable. In spite of the accessibility inherent in its graphic environment, the system of symbolic representation must be learned in order to have successful results. For example, tests with children on UPIC showed that the sounds they imagined when drawing were not anything like what the system produceddrawing an airplane did not produce its sound. The user must in fact understand some basic concepts about physics and acoustics, as well as be introduced to the system's user interface, before being able to work effectively. One of the most difficult aspects of the UPIC system is the near impossibility for the user to predict the sound of a given handdrawn waveform, as visual appearance in the time domain is not

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

37

perceptually related to spectral content (Roads 1996). Further, different waveforms can have nearly identical spectrums if there are different phase relationships among their spectral components. Finally, complexity in the drawn waveform does not guarantee complexity in the resulting sound. In short, any aesthetic judgement on the drawing is not valid for sound itself. Still, one of the system's strong points is the exploitation of the intuitive expressivity of drawing by hand. Hands are the parts of the body best controlled by the brain, as they are required to make very precise movements (Greenfield 1997). This close relation between the hand and the brain gives UPIC a strong potential for expressivity. Drawing by hand also introduces a certain amount of randomness as it is almost impossible to draw perfect and accurate lines. The hand's instability introduces interesting variations in different parameters, which gives richness and inner movement to the sound. The hand itself stands between randomness and calculation. It is both an instrument of the mindso close to the handand an imperfect tool. The products of the intelligence are so complex that it is impossible to purify them in order to submit them totally to mathematical laws .. . The hand adds inner richness and charm. (Xenakis 19876: 23) Furthermore, with the introduction in 1987 of the first realtime UPIC system, new levels of the interaction between the composer and the computer were created. The feedback loop of the interaction involves the human brain, ear, and hand. In this creative process, the composer adjusts the movement of his hand to the sound he hears, continuously refining the audible result. A second advantage of the real-time system was the possible use of UPIC as a performance instrument. For example, the realtime control of the sequencer object can produce interesting results. This option was partly exploited by Xenakis in the performance of Taurhiphanie (1987). 4.4 Compositions Realized with the UPIC System Mycenae Alpha (1978) was the first work ever made entirely on UPIC. It was composed at CEMAMu. The duration of the two38
JOURNAL OF THE HELLENIC DIASPORA

channel tape is 10 minutes. It was first performed on August 2, 1978, in Mycenae, Greece. Xenakis admitted he made Mycenae Alpha as an exercise. He was experienced both with the technical and the musical capabilities of the first UPIC system. He was aware of the limitations of the system. That system had no realtime synthesis capabilities, so he had to wait for a long time, sometimes hours, before he could listen to the results. There were no mixing functions either. He prepared the pages of the score in such a way that they could be linked to each other by timewise juxtaposition (Lohner 1986b). Nevertheless, he was satisfied with the results and recognized the work as an official composition. Pour la Paix (1981) is a mixed-media work for chorus, narrators, and tape. The stereo tape was realized on the UPIC system at CEMAMu. It was first performed on April 23, 1982, at Radio France. The duration of the composition is 26 minutes and 45 seconds. The text was derived from Ecoute and Les Mort Pleureront by Francoise Xenakis. The phonemic text was written by Iannis Xenakis himself. The text is about the horrors of war. It tells the story of two friends that fight in two opposite camps and finally get killed in an explosion. There are four different versions of Pour la Paix in total. Three of them include a tape part. Taurhiphanie (1987) was the first work that exploited the capabilities of the new real-time unit. Part of the work was created on UPIC and stored on tape and part of the work was improvised live. The stereo tape lasts 10 minutes and 45 seconds. The first performance took place in the Arena at Arles on July 13, 1988. The title of the work is a coinage of the words tauros, which means "bull," hippos, which means "horse," and epiphany, which means "apparition." The sound material for both the tape part and the improvisation comes from digitized recordings of bulls. In the performance, Xenakis himself improvised live on UPIC and had control of the loudspeakers. The system was placed in the middle of the arena, which was full of real bulls and horses (Varga 1997). Voyage Absolu des Unari very Andromede (1989) was the last piece Xenakis made on UPIC. The duration of the stereo tape is 15 minutes and 30 seconds. It was first performed on April 1, 1989, at the Temple Kamejama Hontokuji in Osaka, Japan. The work is commissioned for the International Exposition of Paper Kites and organized by the local Goethe Institute. The title suggests the imaginary space trip of a kite toward the galaxy of Andromeda.

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

39

The piece consists of various episodes while the kite crosses the spaces between the stars.

5.0 Dynamic Stochastic Synthesis and the GENDY Program

In the chapter "New Proposals in Microsound StruCture," in

Formalized Music, Xenakis proposed an alternative method of

sound synthesis (Xenakis 1992). Instead of a method based on the Fourier analysis concept using periodic functions, he proposed a method that used stochastic variations directly on the pressure curve of the sound. An early attempt in stochastic synthesis is applied to some of the sounds in the composition La Legende d'Eer (1989). The completion and implementation of the concept of dynamic stochastic synthesis took place in the GENDY computer program. The program derived its name from the initials of the words generation (GEN) and dynamic (DY). Since 1991, Xenakis has been developing the program at CEMAMu, with the assistance of MH. Serra. There are two underlying goals of this project: First, to thoroughly exploit the stochastic waveform synthesis method, and second, to homogenize and unify both the microlevel (the sound events) and the macrolevel (the form) of musical composition. GENDY has been used in the composition of a series of computer music works. To date there are two official compositions (Salabert 1997) made by Xenakis: Gendy3 (1991) and S.709 (1994). 5.1 Background The idea behind the GENDY program derived directly from the concept of the early stochastic works, especially Achorripsis (1956-57) and the ST series of compositions (ST/4, ST/10, ST/48 ). In that period, Xenakis had the idea of a system that would produce music automatically using minimal rules. He imagined this system as a "black box," where he could enter data referring to musical parameters and the system would output music as a result (Varga 1996, p.80). In Achorripsis, Xenakis used a number of stochastic rules to calculate both the parameters of the individual notes (e.g. duration and pitch) as well as the structure of the whole composition. During the 1960s, Xenakis designed the ST computer pro40
JOURNAL OF THE HELLENIC DIASPORA

gram (hence the name of the series of the works) and used the computer in the composition of automated music for the first time. The program had the same structure as the one he had used in Achorripsis (Varga 1996:80). In addition, he achieved enormous acceleration of the computational process, which, in turn, gave him the ability to try a variety of different combinations of input parameters. When he was satisfied with the results, he transcribed the numerical outputs of the program into traditional notation. Referring to the ST program, Xenakis indicates the possibility of a stochastic synthesis program working directly at the sound level: Although this programme gives a satisfactory solution to the minimal structure, it is, however, necessary to jump to the stage of pure composition by coupling a digital to analogue converter (DAC) to the computer. The numerical calculations would then be changed into sound, whose internal organisation had been conceived beforehand. (Xenakis 1992: 144) Finally, it is worth noting the similarity between the organization of the macrostructure on these early instrumental pieces and that on the GENDY program. In Achorripsis, the macrostructure is organized using a vector matrix in which seven instrument families are distributed into rows, while the horizontal time axis is divided into twenty-eight sections. A similar approach was taken when implementing the organization of the voices in GENDY. 5.2 Stochastic Waveform Synthesis Xenakis conceived of stochastic waveform synthesis during the early 1970s (Xenakis 1992). The starting point in his thinking was the way the sound propagates in the air and is received by the ear. Considering sound as a psycho-physical phenomenon, we can say that the sound is variations in the atmospheric pressure produced by a vibrating source. The vibrations disturb the air molecules by pushing them together (condensation) and subsequently pulling them apart (rarefaction). When the sound reaches the ear it causes the eardrum to vibrate in sympathy with the wave (Roeder 1979). Thus, according to Xenakis:

Aesthetics and Techniques in the Electroacoustic Music of Xenakis 41

A sound may be completely represented by its curve of atmospheric pressure variation in time. It is this curve that strikes our ears and nothing else. Consequently, to judiciously construct pressure-time curves (linear forms) goes back, in theory, to fabricating any desired sound through D/A conversion. This curve and its corresponding sound (music) will be considered as an entity. (Xenakis 1985b:179) This statement is a departure from the widely spread method of Fourier analysis and synthesis of sound. This method is based on a theorem by Fourier, who proved that, under certain restrictions, any complex waveform can be analyzed into a series of sine waves with specific frequencies, amplitudes, and phases. Consequently, in the reverse procedure, a complex sound can be synthesized by adding elementary waveforms together with specific frequency, amplitude, and phase relationships. This method is called additive synthesis, or Fourier synthesis. In contrast, Xenakis has shifted his main interest in sound representation from the frequency domain to the time domain. In the 1970s, Xenakis highly criticized both the electronic music of the past and the electronic music of his time. He regarded the electronic instruments as incapable of reconstructing even the simple sounds of some orchestral instruments. He described all the pre-Second World War instruments such as the Trautonium, the Theremin, or the Martenot as failures. To him, the same was valid for the postwar electronic music that used oscillating electronic circuits to create waveforms based on trigonometric functions (sine wave, square wave, etc.). He found that approach pragmatic and the resulting sound dull and uninteresting. The only exception was the utilization of recorded sound material and its further treatment under the techniques of musique concrete. It is worth noting his underlying criticism of the serial composers who explored the electronic medium (e.g. Stockhausen). Any electronic music based on such sounds only, is marked by their simplistic sonority. Only when the "pure" electronic sounds were framed by other "concrete" sounds, which were much richer and much more interesting (thanks to E.Varese, P.Schaeffer, and P.Henry), could electronic music become really powerful. (Xenakis 1992: 243)

42

JOURNAL OF THE HELLENIC DIASPORA

Xenakis often wondered what would make a sound alive and interesting. He noticed that when analyzing the spectrum of a simple, sustained, instrumental sound there were minute variations both in frequency and amplitude of the different partials. The same phenomenon analyzed in the time domain would reveal alterations in the intensity and the rate of the pressure variation. These tiny fluctuations and irregularities in the sustained part of the sound are the determinant factors that distinguish an instrumental sound from its lifeless electronic simulation. For Xenakis, if someone wanted to synthesize a new sound while taking into account the characteristics that make an instrumental or natural sound rich and interesting, he should construct new theories of approach. These theories should use another functional basis and a harmonic analysis on a higher level. Xenakis was to construct his new theory on sound synthesis based on stochastic processes. In addition to the above, Xenakis noted the inefficiency of the Fourier method to represent sounds with noise components or that are very close to noise itself. The approximation of the Fourier method worked best with harmonically comprised sounds because that method had strong harmonic bias. That conclusion was important especially when taking into account the transient part of the sound. The transient part, which is determinant in timbre recognition, is characterized by non-linear behavior. That behavior is almost impossible to represent with the Fourier method. Furthermore, analysis of instrumental sounds showed that the spectrum of most instruments contained "detuned" harmonics, or sometimes both harmonic and inharmonic partials. In the 1970s, it was almost impossible to approximate such characteristics using additive synthesis because that would require a tremendous amount of calculation time and computer power. Xenakis approached this problem in the same way kinetic gas theory solved the problems of nineteenth-century classical mechanicsby approximating real variations with stochastic functions (Xenakis 1992: 244). Despite his criticism, Xenakis acknowledged the thorough research on the timbre of various instruments made by Jean-Claude Risset as well as the possibilities of the Music N series computer music programming language, developed by Max Matthews at the Bell Telephone Laboratories (Xenakis 1992: 246). For Xenakis, Music N language achieved the final step in the development of a Fourier-based analysis/synthesis method and

Aesthetics and Techniques in the Electroacoustic Music of Xenakis 43

represented the realization of the dream of electronic music composers of the fifties (Xenakis 1992: 246). Xenakis reached the conclusions mentioned above in his effort to investigate the inner characteristics that make sounds rich and interesting for music. His primary interest was not to simulate real instruments nor to approximate their timbres. On the contrary, his thought focused on how one can obtain rich, alive, previously unheard-of sounds (Xenakis 1985b: 181). Xenakis realized that there were two possible ways of looking at the problem. The first was to take the additive synthesis concept and extend it by introducing disorder in the parameters of the sound. That could be realized by applying stochastic functions to vary the frequency and the amplitude of each of the partials of the sound. Due to the inherent weaknesses of Fourier analysis as described earlier, Xenakis considered this solution inadequate. The second approach followed the opposite direction. Xenakis proposed a method by which the composer starts with an aperiodic waveform and then introduces different degrees of regularities. Instead of starting from the unit element concept and its tireless iteration and from the increasing irregular superposition of such iterated unit elements, we can start from a disorder concept and then introduce means that would increase or reduce it. This method would use stochastic variations of the sound pressure directly. (Xenakis 1992: 246)

5.2.1 Stochastic Waveform Synthesis Theory

The idea behind the method called dynamic stochastic synthesis works directly on the tiny details of the sound pressure curve. This very detailed operation requires an equally efficient sound analysis framework, which had emerged with the development of computers and digital technology. Working on the sound pressure curve means working with a high time resolution factor and handling a large amount of numbers. Xenakis's idea was to start with an aperiodic waveform and then inject regularities. The aperiodic waveform can be constructed with the help of a stochastic function. Suppose there is an initial waveform period that is reproduced numerous times but after each reproduction the points comprising the waveform curve are more 44
JOURNAL OF THE HELLENIC DIASPORA

or less displaced. This displacement is caused by a stochastic function. If this procedure is repeated enough times then the points of the waveform will eventually lose their continuity and finally come apart forming an amorphous, statistical cloud. Listening to this sound we will perceive it as white noise. For this reason it was necessary to introduce certain regularities by limiting the activity of the stochastic function. This can be done with the use of weights, constraints, and barriers. This will allow some periodicity to emerge and take control over the resulting timbre. Therefore, according to the rate of influence of the stochastic function, the sound can vary between two poles. On the one hand, if the influence of the function is reduced to zero, we have pure repetition or strict periodicity, which leads to a monotonous sound. On the other hand, if the influence is total, we have randomness or aperiodicity, which leads to noise. Consequently, we can say that the results of the stochastic synthesis method lie between determinancy and indeterminancy. It is obvious that with this synthesis method the composer can achieve a vast range of rich and continuously evolving sounds. Therefore, the role of the composer is to restrict and control the behavior of the stochastic function in order to achieve the appropriate musicality and quality in the resulting sound. Xenakis proposed eight strategies in conjunction with the dynamic stochastic synthesis (Xenakis 1992: 246-49; and Roads 1996: 341). In short they are the following: 1) The composer can choose among a variety of stochastic functions to create the waveforms. Each function has its own characteristics. The functions can be either continuous or discontinuous (e.g. Poisson, exponential, normal, uniform, Cauchy, arcsine, and logistic); 2) Stochastic functions can be combined in various ways (e.g. addition, multiplication, etc.); 3) The random variables governing the positions of the sound particles (samples) in the amplitude and time axes can be functions of elastic forces or other random variables; 4) The random variables can move between two elastic barriers or mirrors;

Aesthetics and Techniques in the Electroacoustic Music of Xenakis 45

5) The output of one stochastic function can be used as the variable of another stochastic function; 6) Various combinations of stochastic functions (linear, polynomial, etc.) can be considered as composite functions (e.g. mixtures through addition possibly over time); 7) The stochastic functions can be classified according to various criteria. Thereafter, these classes can be considered as elements of higher order sets; 8) Extending the previous strategy, the composer can assign the control of the higher order classes into a stochastic composition program. The latter controls the macrostructure of the composition. The eighth strategy comprises a central idea in Xenakis's thinking. With the development of the appropriate computer software and hardware, the composer is able to control both the sound material and the structure of a whole composition by applying common techniques to the various levels of the composition. Any theory or solution given on one level can be assigned to the solution of problems of another level. Thus the solutions in macrocomposition (programmed stochastic mechanisms) can engender simpler and more powerful new perspectives in the shaping of microsounds than the usual trigonometric functions can. Therefore, in considering clouds of points and their distribution over a pressure-time plane, we can bypass the heavy harmonic analyses and syntheses and create sounds that have never before existed. Only then will sound synthesis by computers and digitalto-analogue converters find its true position, free of the rooted but ineffectual tradition of electronic, concrete, and instrumental music that make use of Fourier synthesis despite the failure of this theory. All music is thus homogenised and unified. (Xenakis 1992: preface) 5.3 Description of the GENDY Program GENDY is a computer program running in the Windows 46
JOURNAL OF THE HELLENIC DIASPORA

environment. The user defines the initial waveform of up to sixteen voices and then the computer varies each waveform following specific probability functions. The computer also handles the combination of these voices using similar statistical laws. Therefore, the program, under the instructions of the composer, functions both on the sound-level and the global architecture of the piece. The final composition is stored as a soundfile in the computer's hard disk. The program operates in real-time, so the composer is able to alter the input parameters during the execution of the calculations.

5.3.1 Micro-level Synthesis 5.3.1.1 Waveform Definition

In the first step, the composer defines the initial waveform. At the present time, the program can only represent the waveform as a polygon. This restriction serves a certain purpose, since a polygonized waveform is a much simpler model and therefore the program algorithm is more efficient in the calculations (Serra 1993). The waveform consists of a number of segments. Each segment is defined by the coordinates of its two endpoints, which correspond to vertices in the time and the amplitude axes. The program then computes all the intermediate samples with linear interpolation, a mathematical technique for generating a line between two endpoints. The total duration of each waveform is equal to the sum of the duration of its segments. Therefore, the total duration of the segments is inversely proportional to the frequency of the sound. After defining the initial waveform, the composer has to choose the functions that will stochastically displace the points of the polygon waveform after each period. The program provides a selection of stochastic distributions. To date, there are six stochastic distributions available: Cauchy, Poisson, Gaussian, arcsine, uniform, and logistic. The computation of the polygonal waveforms makes use of one stochastic law that governs the amplitudes and another one that governs the durations of the time segments. The program applies these distributions and synthesizes a large sequence of successive waveforms. In this sequence, each waveform is a "distorted" variation of the previous one. The probability function is applied onto each different endpoint in the waveform. The smooth continuation between two successive waveforms is

Aesthetics and Techniques in the Electroacoustic Music of Xenakis 47

achieved by ensuring that the first endpoint in each waveform shares the same vertices with the last endpoint of the previous waveform, thus avoiding abrupt jumps. These stochastic variations of the polygonized segments result in continuous variations on the produced sound. For example, variations of the segments on the amplitude axis lead to variations of the amplitude and the timbre of the sound, while variations of the segments on the time axis, alter the periodicity of the waveform, which results in variations of the frequency. Thus, if the composer applies variations on only the amplitude of the segments while the time positions of the endpoints remain unaltered, then the resulting sound has a constant stable pitch. If the composer applies variations on the positions of the endpoints on the time axis, then the frequency can be increased or decreased.

5.3.1.2 Elastic Barriers

The logic of the program algorithm imposes the use of boundaries to limit the stochastic variations. These boundaries are called mirrors, or elastic barriers. There are two reasons for needing to apply boundaries. First, there are some physical limits imposed by the computer itself. The amplitude of the sound samples must not exceed the resolution of the analog-to-digital converters. Second, if the stochastic variations are not limited, the sound signal tends quickly toward noise. In order to give a sense of timbre or pitch continuity to the produced sound, the successive waveforms should share similarities. These similarities could be either periodicities in the time axis or symmetries in the amplitude axis (Xenakis 1992: 289). The lack of similarities caused by the absence of any limit in the waveform variations produces sounds that are perceived as noise. It is therefore necessary to put limits on the stochastic processes in order to both avoid stable repetition (resulting in monotone sounds) and total instability (leading to irregularity and noise). There are both time and amplitude elastic barriers. When the stochastic function defines new coordinates for the segment points of the polygonized waveform, the program examines whether each endpoint falls within the barriers. All the points that fall outside the barriers are reflected back into a permissible position into the barrier range. These barriers have been named mirrors, due to their reflective property. There can be as many reflections as needed so
48 JOURNAL OF THE HELLENIC DIASPORA

that the final position of the endpoint falls within the barriers. The mirrors are like filters on the stochastic variations. The composer can control the quantity of reflected points by determining the position of each barrier. Since reflections introduce discontinuities in the waveform, we can say that the mirrors impose timbral control. In addition, the time barriers determine the time interval between the segment points, and therefore they are related to the frequency of the sound.

5.3.2 Macro-level Composition

On the macro-synthesis level, the program uses stochastic functions to handle the organization of the sounds. These stochastic functions are similar to the ones used on the micro-level. Therefore, the program controls both the macro- and the micro-level using a common approach. GENDY handles the macro-level structure of the composition as a two-dimensional space where the horizontal axis represents time and the vertical axis is used for the layout of different voices. The composer assigns a number of initial parameters and the program structures the whole composition.

5.3.2.1 The Voice Configuration

The vertical axis represents the voice configuration of the composition. The composer assigns up to sixteen different voices. Each voice corresponds to a synthesised sound produced at the microlevel. Each voice is defined by a different set of parameters. These parameters refer to the waveform and its evolution (see paragraph 5.3.1 above ).

5.3.2.2 The Sound/Silence Decision

The time axis is divided into a series of successive time-fields. Each time-field corresponds to a portion of time where a voice is either silent or active. The duration and the succession of each time-field are different for each voice. The computer decides whether each time-field is to be silent or not. The decision is made by using the Bernoulli trial. The Bernoulli trial is a stochastic function that outputs only two possible results (success/failure, yes/no, etc.). The composer initially sets a probability of success p for each different voice. The computer generates a random number z between 0 and 1 using uniform distribution. Then the program compares the two numbers (p and z). The decision is

Aesthetics and Techniques in the Electroacoustic Music of Xenakis 49

successful if the probability of success p is above or equal to the random number z (p>=z). If the decision corresponds to success then the voice is active, while if it corresponds to failure the voice is silent. Furthermore, the probability of success controls the ratio of the active and the silent sound-fields of each entire voice. For example, if p is 0.5 then the active sound-fields tend to be equal in number (though not in duration) to the silent sound-fields.

5.3.2.3 The Durations The duration of each time-field is also decided by the program. For this it uses an exponential stochastic function, the formula being: d = (-1/D) log(1-z)
The composer sets the mean duration D while the computer generates a random number z with uniform distribution between 0 and 1. The mean duration is fixed for each voice (Serra 1993). Therefore, the program decides on the duration of each time-field applying a value ranging around the mean duration specified by the composer.

5.4 Observations on Dynamic Stochastic Synthesis

For Xenakis, dynamic stochastic synthesis is a method emerging out of the necessity to vitalize the sound products of the electronic medium. The sound space in question is one which will produce a likeness of live sounds or music, unpredictable in the short or long run, but, for example, being able to vary their timbre from pure "sine-wave" sound to noise. (Xenakis 1992: 295) After investigating its possibilities, both in theory and in practice, one can observe the method's ability to yield a wide range of sounds, with great inner activity and complexity. As mentioned above, the synthesis algorithm operates directly on the sound pressure curve. However, despite the advantages, there seem to be some weak points. First, in the stage where the composer defines the initial period 50
JOURNAL OF THE HELLENIC DIASPORA

of the waveform, it is difficult to predict the sound result. It is almost impossible to estimate the timbre of a sound from its visual appearance in the time domain. This is similar to the problem encountered with the UPIC system discussed earlier. Second, it is difficult to predict the evolution of the sound. This would require a deep knowledge of the influence of the stochastic function over the sound waveform. This problem of control becomes harder if the composer moves toward various combinations of the stochastic functions by following a more complicated strategy. This is also obvious when moving toward the control of multiple voices. Serra, Xenakis's assistant on the GENDY program, explains: At this time we do not know exactly how to formalise and quintise the effect of the mirrors on the sound parameters. The mathematical aspect of the stochastic synthesis algorithm is at this time under study. (Serra 1993: 247) Dynamic stochastic synthesis is still in the beginning. Only after a thorough investigation of the behavior of the various stochastic functions and the associated strategies can the composer obtain a more stable compositional framework. Xenakis is aware of both points stated above (Robindore 1996). His long experience with the applications of stochastic functions in music have gained him great familiarity with them. He proposes an intuitive approach toward the decision of the specific parameters in the program. He has an experimental attitude toward the program, which enables him to have the final decision on the results. A common scenario of working with the program is that the composer initially sets the parameters, listens to the results, revises his decisions, and continues this procedure until he is satisfied with the result. Xenakis is not restricted to the deterministic character of the mathematical functions but he is more concerned about the acoustical results. Finally, in the current state of the program, and for computational efficiency, the waveforms that are used are polygonized and linear interpolation is applied for the completion of the waveforms. The use of linear interpolation for waveform synthesis yields specific side-effects in the sound. This is because the sharp angles in the waveform produce certain harshness in the sound

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

51

with uncontrollable high-frequency partials. Xenakis was always attracted to some kind of roughness in the sounds he used in his compositions, both instrumental and electronic. However, a sound synthesis algorithm should not impose specific qualities on the sound material, especially if this is induced by some restrictions in the code as is the case of linear interpolation. In the future one hopes that this aspect of the method will be revised to include another interpolation algorithm (exponential, logarithmic, etc.) that generates a different family of curves between the breakpoints. This would reduce that emphasis in the high frequencies of the sound spectrum.

5.5 Compositions Realized with the GENDY Program GENDY3 (1991) is the first work made with the GENDY
program. The stereo tape lasts 20 minutes. It was premiered in Metz, France, in November 1991. S.709 (1994) is the last official electroacoustic work made by Xenakis. The duration of the stereo tape is 7 minutes. It was first performed on December 1994 in Paris at Radio France.

6.0 Final Conclusions

An overall look into the electroacoustic work of lannis Xenakis may lead to a few main conclusions. First, Xenakis was one of the first composers to deviate from the two orthodoxies of the postwar era, Elektronische Musik and musique concrete. By introducing mathematical and scientific concepts into composition and by using both concrete and electronic sounds in his works, he contributed to the overcoming of the polarization of the above two dominant forces during the 1950s. Consequently, he helped in the historical evolution of the genre. Second, Xenakis also proposed new models for the micro-time sound synthesis. The theory behind all those models focused on the atmospheric pressure curve (waveform) that represents sound. This approach was in opposition to the Fourier concept of sound. The three different sound synthesis models proposed by Xenakis were: granular, graphical, and stochastic waveform synthesis. All three models were capable of producing rich and lively timbres. In granular and stochastic waveform synthesis, Xenakis introduced con52
JOURNAL OF THE HELLENIC DIASPORA

cepts deriving from the fields of new sciences and mathematics. In graphical synthesis he exploited the intuitive expressivity of drawing by hand. Those models opened new paths in timbral composition but their musical possibilities have yet to be explored. In particular, concerning the stochastic waveform synthesis model, a higher degree of control by the composer has to be established. However, the granular synthesis model has been extensively explored and has paved the way for a whole new direction in research and composition. Third, Xenakis contributed to the technology of electroacoustic music by developing two different compositional environments for electroacoustic music at CEMAMu. First, he developed the UPIC system, a computer-based workstation, in which he implemented graphical synthesis. Second, he developed the GENDY computer program, in which he implemented stochastic waveform synthesis. Both systems are capable of creating a complete composition. Finally, Xenakis attempted to overcome the cognitive division whereby material and structure are considered to be two different aspects of a composition. He adopted a holistic attitude according to which all the levels of the composition could be unified through the application of common compositional models. In classical aesthetical theory, the sound is neutralized by the compositional process and becomes material with which to build. For example, the sound of a violin becomes a neutral material, which is used in the structuring of compositions. Only in this way can the relationship between the materials and the inner logic of the work of art be revealed (Boulez 1987). With the advent of music technology and the ability to synthesize sound, the composer makes the listener hear the material as something composed, something subjected to aesthetic criteria. Thus, a sound that is the result of a creative process implies some kind of form. This process of composing-the-sound reflects a totally different approach from the established approach of composing-with-sounds and reveals a new cognitive relation concerning the form and the materials (Di Scipio 1994). All three sound synthesis models proposed by Xenakis (granular, graphical, and stochastic waveform synthesis) also supply ideas on musical form. Thus, it could be said that theoretical models for composing-the-sound propose ideas on models for composing-with-sounds. Furthermore, on all the three sound synthe-

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

53

sis models, the micro-level synthesis theory could be identical to the macro-structure composition theory. This relation between the theory for timbral design and the theory for macro-structural organization provides a strong coherence between the material and the form in the composition. Consequently, homogenization is achieved throughout the composition.

References
A. BIBLIOGRAPHY Attali, J. Noise: The Political Economy of Music. Manchester: Manchester University Press, 1985. Ames, Ch. "Automated Composition in Retrospect: 1956-1986." Leonardo 20, no. 2 (Cambridge, Mass.: MIT Press, 1987):169-86. "The Markov Process as a Compositional Model: A Survey and Tutor ial." Leonardo 22, no. 2 (Cambridge, Mass.: MIT Press, 1989):175-88. Apollonio, U., ed. Futurist Manifestos. London: Thames & Hudson, 1973. Bastiaans M. "Gabor's Expansion of a Signal into Gaussian Elementary Signals." Proceedings of the IEEE 68 (1980): 538-39. Bois, M. Xenakis, The Man and His Music. London: Boosey & Hawkes, 1967. Boulez, P. (1977). "Technology and the Composer," in S.Emmerson, ed., The Language of Electroacoustic Music. London: Macmillan, 1986. "Timbre and Composition." Contemporary Music Review 12, no. 1 (Har wood: G+B Arts International, 1987). Bregman, A.S. Auditory Scene Analysis. Cambridge, Mass.: MIT Press, 1990. Burger, P. Theory of the Avant-Guarde. Manchester: Manchester University Press, 1984. Busoni, F. "Sketch of a New Aesthetic of Music," in F.Busoni, C. Debussy, and Ch. Ives, Three Classics in the Aesthetic of Music. Dover, 1907. Chadabe, J. Electric Sound: The Past and Promise of Electronic Music. Englewood Cliffs, N.J.: Prentice-Hall, 1996. Cope, D.H. New Directions in Music. Iowa: W.C.Brown, 1989. D'Allonnes, O.R. Xenakis/Les Polytopes. Paris: Balland, 1975. De Poli, G. "A Tutorial on Digital Sound Synthesis Techniques." Computer Music Journal 7, no. 4 (Cambridge, Mass.: MIT Press, 1983): 8-26. De Poli, G., A. Piccialli, and C. Roads, eds. Representations of Musical Signals. Cambridge, Mass.: MIT Press, 1991. Di Scipio, A., "Formal Processes of Timbre Composition Challenging the Dualistic Paradigm of Computer Music." Helsinki: Proc. of ISEA, 1994. Die Reihe. "vol.1 Elektronische Musik." Vienna: Universal Edition, 1955. Dodge, Ch., and Th.A. Jerse. Computer Music, revised edition. New York: Schrimer Books, 1997.

54

JOURNAL OF THE HELLENIC DIASPORA

Emmerson, S., ed. The Language of Electroacoustic Music. London: Macmillan, 1986. Gabor, D. "Acoustical Quanta and the Theory of Hearing." Nature 159 (1947): 591-94. Greenfield, S. The Human Brain. London: Weidenfeld & Nicolson, 1997. Griffiths, P. Modern Music: A Concise History from Debussy to Boulez. London: Thames & Hudson, 1978. Hoffmann, P. "Implementing the Dynamic Stochastic Synthesis," in G. Assayag, M. Chemillier, Ch. Eloy, eds., Proccedings of the T hird Journies d'Informatique Musicale JIM 96. Caen: Group de Recherche en Informatique, Image, Instrumentation de Caen, 1996, pp. 341-47. Jones, K. "Compositional Applications of Stochastic Processes." Computer Music Journal 5, no. 2 (Cambridge, Mass.: MIT Press, 1981): 45-61. Jones, D., and T. Parks. "Generation and Combination of Grains for Music Synthesis." Computer Music Journal 12 no. 2 (Cambridge, Mass.: MIT Press, 1988): 27-34. Julesz, B. "The Elements of Texture Perception and their Interactions", Nature 290 (1981): 91-97. Koenig, G.M. "Aesthetic Integration of Computer-Composed Scores." Computer Music Journal 7, no. 4 (Cambridge, Mass.: MIT Press, 1983). Kuhn, Th. The Structure of Scientific Revolutions. Chicago: University of Chicago Press, 1962. Kundera, M. "Xenakis, Prophet of Insensibility," sleeve notes in the CD Kraanerg, KTC 1075, ETCETERA, 1989. Lerdahl, Fr. "Timbral Hierarchies." Contemporary Music Review 2, no. 1 (1987). Lohner, H. "The UPIC System: A User's Report." Computer MusicJournal 10, no. 4 (Cambridge, Mass.: MIT Press, 1986a): 42-49. "Interview with Iannis Xenakis." Computer Music Journal 10, no. 4 (Cambridge, Mass.: MIT Press, 1986b): 50-55. Lorrain, D. "A Panoply of Stochastic Canons." Computer Music Journal 4, no. 1 (Cambridge, Mass.: MIT Press, 1980): 53-81. Loy, D.G. "Composing with ComputersA Survey of Some Compositional Formalisms and Music Programming Languages," in M. Mathews and J.R. Pierce, eds., Current Directions in Computer Music Research. Cambridge, Mass.: MIT Press, 1989. Manning, P. Electronic and Computer Music. Oxford: Oxford University Press, 1987. Marino, G., J.-M. Raczinski and M.-H. Serra. "The New UPIC System," in Proceedings of the 1990 International Computer Music Conference. San Francisco: ICMA, 1990, pp. 249-52. "The UPIC System, Origins and Innovations." Perspectives of New Music 31, no. 1 (1992): 258-69. Mathews, M., J. Miller, F.R. Moore, J.R. Pierce and J.C. Risset. The Technology of Computer Music. Cambridge, Mass.: MIT Press, 1969. Mathews, M., and J.R. Pierce, eds. Current Directions in Computer Music Research. Cambridge, Mass.: MIT Press, 1989. Matossian, N. Iannis Xenakis. London: Kahn & Averill, 1986.

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

55

Moles, A. Information Theory and Esthetic Perception. Urbana: University of Illinois Press, 1968. Moore, F.R. Elements of Computer Music. Englewood Cliffs, N.J.: Prentice-Hall, 1990. Myhill, J. "Some Simplifications and Improvements in the Stochastic Music Program," in Proceedings of the 1978 International Computer Music Conference. Evanston: ICMA, 1979, pp. 272-317. Nelson, P. "The UPIC System as an Instrument of Learning." Organised Sound 2, no. 1 (Cambridge: Cambridge University Press, 1997): 35-42. New Grove Dictionary of Music. London: Macmillan, 1980. Palombini, C. "Technology & Pierre Schaeffer." MikroPolyphonie 4, no. 1 (1997). "Musique Concrete Revisited." UFPr Arts Department Electronic Musicological Review 4 (June 1999). Pape, G. "Some Musical Possibilities of the New UPIC System," in Computer Music Conference and Festival at Delphi. Athens: CCMR, 1992a. "Studio ReportLes ateliers UPIC," in Computer Music Conference and Festival at Delphi. Athens: CCMR, 1992b. Piaget, J. Child's Conception of the World. Littlefield: Adams, 1980. Poggioli, R. Theory of the Avant-Guarde. Cambridge, Mass.: Belknap/Harvard, 1982. Raczinski, J.-M., G. Marino, and M.-H.Serra. "New UPIC System Demonstration," in Proceedings of the 1991 International Computer Music Conference. San Francisco: ICMA, 1991, pp. 567-70. "CEMAMu-Studio Report," in Computer Music Conference and Festival at Delphi. Athens: CCMR, 1992. Reynolds, K., transcription and editing. "Xenakis, Reynolds, Lansky and Mache Discuss Computer Music," public discussion held on April 7, 1992 at the Computer Music Conference and Festival in Delphi, Greece. Roads, C. "Granular Synthesis of Sound," in C. Roads and J. Strawn, eds., Foundations of Computer Music. Cambridge, Mass.: MIT Press, 1985. , ed. The Music Machine. Cambridge, Mass.: MIT Press, 1989. The Computer Music Tutorial. Cambridge, Mass.: MIT Press, 1996. Robindore, Br. "Eskhate Ereuna: Extending the Limits of Musical Thought Comments On and By Iannis Xenakis." Computer Music Journal 20, no. 4 (Cambridge, Mass.: MIT Press, 1996): 11-16. Roeder, J.C. Introduction to the Physics and Psychophysics of Music. New York: Springer-Verlag, 1979. Salabett. Iannis Xenakis: Catalogue of Works. Paris: Edition Salabert, 1997. Serra, M.-H. "Stochastic Composition and Stochastic Timbre: GENDY3 by Iannis Xenakis." Perspectives of New Music 31, no. 1 (1993): 236-57. Smalley, D. "Spectro-morphology and Structuring Processes," in S. Emmerson, ed., The Language of Electroacoustic Music. London: Macmillan, 1986. Solomos, M. Sleeve notes in the CD Xenakis-Electronic Music, CD003, EMF, 1997. Solomos, M., and P. Hoffmann. "The Electroacoustic Music of Iannis Xenakis," in First Symposium in Music and Computers. Corfu: Ionian University, 1998. Squibbs, R.J. "Images of Sound in Xenakis' Mycenae-Alpha," in G. Assayag, M. Chemillier, and Ch. Eloy, eds., Proceedings of the Third Journe'es d'Infor-

56

JOURNAL OF THE HELLENIC DIASPORA

matique Musicale JIM 96. Caen: Group de Recherche en Informatique, Image, Instrumentation de Caen, 1996, pp. 208-219 Sutherland, R. "Xenakis: The Instrumental Works," and "Xenakis: The Electronic Works," in New Perspectives in Music. London: Sun Tavern Fields, 1994. Schwartz, E., and B. Childs, eds. Contemporary Composers on Contemporary Music. New York: Da Capo, 1983. Toop, R. "The Music of lannis Xenakis," sleeve notes in the CD Kraanerg, KTC 1075, ETCETERA, 1989. "Beyond the History of Music . . . ," sleeve notes in the CD lannis Xenakis 2-La Legende d'Eer, M0782058, Montaigne, 1995. Truax, B. "Real-time Granular Synthesis with a Digital Signal Processing Computer." Computer Music Journal 12, no. 2 (Cambridge, Mass.: MIT Press, 1988): 14-26. "Musical Creativity and Complexity at the Threshold of the 21st Cen tury." Interface 21, no. 1 (1992): 29-92. Varese, E. "The Liberation of Sound," in B.Boredz and E.Cone, eds., Perspectives on American Composers. New York: Norton, 1971. Varga, B.A. Conversations with lannis Xenakis. London: Faber and Faber, 1997. Wiener, N. (1964). "Spatial-temporal Continuity, Quantum Theory and Music," in M.Capek, ed., The Concepts of Space and Time. Boston: Norton, 1975. Woodward, R. "Kraanerg," sleeve notes in the CD Kraanerg, KTC 1075, ETCETERA, 1989. Xenakis, I. Arts/Sciences-Alloys. Stuyvesant, N.Y.: Pendragon, 1985a. "Music Composition Treks," in C. Roads, ed., Composers and the Com puter. Madison: A-R Editions, 1985b. "Mycenae Alpha." Perspectives of New Music 25 (1987a): 12-15. "Xenakis on Xenakis." Perspectives of New Music 25 (1987b): 16-49. . Formalized Music, revised edition. Stuyvesant, N.Y.: Pentagon Press, 1992. B. DISCOGRAPHY Diamorphoses (1957). "Xenakis: Electronic Music." Electronic Music Foundation, EMF CD 003, 1997. Concret PH (1958). "Xenakis: Electronic Music." Electronic Music Foundation, EMF CD 003, 1997. Analogique A+ B (1958-59). "Iannissimo!" Vandenburg, VAN 0003, 1997. Orient-Occident (1960). "Xenakis: Electronic Music." Electronic Music Foundation, EMF CD 003, 1997. Bohor (1962). "Xenakis: Electronic Music." Electronic Music Foundation, EMF CD 003, 1997. Kraanerg (1968-69). "Iannis Xenakis: Kraanerg." Asphodel, Sombient 0975, 1997. Hibiki-Hana-Ma (1969-70). "Xenakis: Electronic Music." Electronic Music Foundation, EMF CD 003, 1997.

Aesthetics and Techniques in the Electroacoustic Music of Xenakis

57

La Legende d'Eer (1977). "Iannis Xenakis 2." Auvidis Montaigne, MO 782058,

1995.
Mycenae Alpha (1978). "Electro Acoutic Music: Classics." Neuma 45074, 1993. Tauthiphanie (1987) "Iannis Xenakis." Neuma 45086, 1994. Voyage absolu des Unari vers Andrornide (1989). Perspectives of New Music, PNM

28, 1990.
Gendy3 (1991). "Iannis Xenakis." Neuma 45086, 1994.
5.709 (1994). "Xenakis: Electronic Music." Electronic Music Foundation, EMF

CD 003, 1997.

58

JOURNAL OF THE HELLENIC DIASPORA