ComputerMusic EnactionSkip other details (including permanent urls, DOI, citation information)
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. Please contact firstname.lastname@example.org to use this work in a way not covered by the license. :
For more information, read Michigan Publishing's access and usage policy.
Page 00000531 Computer Music Enaction Kevin Dahan CICM - Universite de Paris VIII email@example.com Abstract Most computer composition systems use a classical representation-based mechanism in which compositional artifacts are considered as objects and then processed through algorithms. While composers see the computer as a "tool ", allowing complex manipulations over objects and allowing greater control on the sound morphology itself very little work has been done to consider the computer as part of the composition process, fully integrating it. The key notion in computer music composition systems is to make available a host ofservices designed to permit "ready-made"" data manipulation and combination. We argue that another approach is desirable, by getting rid of assumptions based on a traditional, constructivist approach, and taking into account recent advances in cognitive science. seen as environments creating the conditions for a composition to take place rather than toolkits permitting to arrange and shape sonic materials. We will first present the different modelling strategies offered by current software environments for composition, and discuss their limitations. We will then introduce the notion of enaction" in a general context, and finally discuss the application of this concept in computer music. Composition Introduction The approach to computer music composition is largely based on rules inherited by the classical (i.e. occidental) approach to composition in poietic terms. There also exists an aural (as in perceptive-centered) tradition amongst electroacoustic composers, in which the sound "sensation" is central to the composition itself, and to composition mechanisms. In this perspective, computer music composition systems are generally considered either as environment based upon predefined "models" (Assayag 1993), which imply a reduction on the possibilities provided, in that there is a restrictive nature to the use of models, or as "sound generators" modules, which can be interconnected, but without any pre-defined "linguistic" constraints. As such, composition systems are largely considered by composers as tools, which permit direct manipulation and experimentation over sound, and allow to compose micro- and macro-structures in the same way. However, it is possible to have a different approach to composition systems by considering the computer as a medium providing information, filtering it and also allowing the user to shape its use more precisely, by removing biases and predefined views on the materials and avoiding to propose formal composition strategies. Composition systems could then be One of the main occupation in musicology has been to find "rules" that specify how to resolve composition problems, and therefore to propose analytical models, generally limited to a specific musical language, or even to a particular composer's style. Such "models" can then be used in "recomposition" environments (Cope 2001). 2.1 Models But what is a musical model? Since Popper rejected the inductive paradigm (Popper 1934), the reality of a scientific model has been questioned. In fact, Popper's definition implies that a model is indeed a representamen'. Such a paradigm shift results in invaliditing the notion of formal model, outside a well-defined situation - a model can therefore only be illustrative and holds no truth outside a restricted and particular case. Musical models In the case of music, there is a strong correlation between mathematical models and linguistic structures. Of course, some particular elements of linguistic structures found in music can be formalized and are well known. For example, medieval harmonization d la Guido d'Arezzo, 1 As Peirce defines it: A Sign, or Representamen, is a First which stands in such a genuine triadic relation to a Second, called its Object, as to be capable of determining a Third, called its Interpretant, to assume the same triadic relation to its Object in which it stands itself to the same Object. (Peirce Edition Project 1998) 531
Page 00000532 certain forms of tonal music (canons, aria, fugue), can easily be formalized in that they provide a global, archetypal structure to follow. On the other hand, classical composition treatises, such as (Falk 1958), give recipes for developing a theme (i.e. melodic augmentation, rythmic tmaformations, etc) which can be considered as formalization tools for local, singular structures. Lerdahl and Jackendoff(1983) somehow demonstrated this structural incoherence: We take the goal of music theory to be a formal description of the musical intuitions of a listener who is experienced in a musical idiom... implying that a generative grammar could be applied to any musical "style", provided there exists an "experienced listener". However, differential psychology and most of psychoacoustics showed the inability to define satisfactory "generic" models of musical perception (McAdams, Winsberg, de Soete, and Krimphoff 1995). Computer Music and Models Interesting results have been produced, mostly in the field on "reproduction" of ancient styles (from Lejaren Hiller's inital experiment to Cope's Virtual Music). It may seem difficult however to use the term "model" to characterize something which is more appropriately described as a set of rules. Experiments of this nature have been conducted using either the symbolic computation paradigm, by considering musical information (mostly notes) as symbols, and using analysis as finding rules in the temporal or structural associations of these symbols, or the connectionnism view, considering the composition as a lattice of interconnected elements. 2.2 Limitations Consequently, the main drawback of the "rules/model" approach is to narrow the composition's field, hence hindering the composer's view by construing the chosen materials, as exemplified by (Nuhn, Eaglestone, Ford, Moore, and Brown 2002). Intrinsic and Extrinsic Constraints Switching between analysis (deconstructive) and composition (constructive) phases has always been a constant task for composers ( (Messiaen 1945), (Smalley 1986), (Mesnage and Riotte 1993)). Whereas analysis is necessary when dealing with a specific "parameter" in the material (as noted by Jean Barraque in (Feneyrou 2001)), formalism was always induced by intrinsic characteristics of the materials chosen, rather than imposed by an extrinsic environment - there is a difference between chosing a specific form outlining the development of musical discourse and having a bounded semantic and grammatical space implied by the use of specific tools. We call these constraints intrinsic and extrinsic. Such constraints are clearly defined by Horacio Vaggione (Budon 2000): A transformational system works not only on the basis of generative formulae but also on that of operations dealing directly with the formal properties of the data to be processed. To achieve independence between processes and data structures, composers are usually forced to move away from integrated composition environment, and to split tasks between a variety of software. Several attempts have been made to provide the composer with a non-intrusive composition environment ( (Courtot 1992), (Drummond 1997), (Dahan, Brown, and Eaglestone 2003)). Typing Computer-assisted composition systems are generally limited in the number of operations they provide the composer. These limits are imposed by the general choice of a strongly-typed environment, in which there are manipulation differences between the artifacts (sound and MIDI files, heterogeneous data) structuring the composition in the poietic stage (Whalley 2005). Depending on the environment, definition of complex data structures is more or less difficult: if the system permits it, the "patch" paradigm is used, acting as a black box for a specific manipulation on basic data objects. Typing also occurs in the definition of manipulation processes, which are usually strongly tied to a specific parameter of sound (i.e. amplitude, spectrum, etc). To this respect, typing is also present, in that it prevents linking objects by "transformational" relationship. The concept of an associative approach has been proposed to limit the use of strongtyping both in data structures and process definition (Dahan 2005). Finally, there exists a physical "typing", which ties the object to a particular perceptive scale, time. However, if we are considering compositional artifacts as compounds of physical data structures and non-morphological manipulation processes, the sense of time is more or less evacuated, in that any object can be adapted to any perceptive scale, from micro- to macro-time, effectively allowing a transversal and transformational approach to composition. Enaction The concept of enaction was originally introduced by Varela (1991) as 532
Page 00000533 A history of structural coupling that brings forth a world. The key problem underlying the questions of model-based and typed environment is that of representation, which enaction can help to overcome. 3.1 Definition The concept of enaction in cognitive science is a reactualization of the phenomenology as described by Merleau-Ponty (1945), and the ecological approach to perception (Gibson 1979). An Approach to Cognition Most of the concept is actually a criticism of the current paradigms in cognitive science, subtituting cognition as a mirror of the outside world to cognition as a part of the world, bringing forth structures and meaning out of it, instead of recreating it. One of the consequences of such an approach is the invalidation of "representation"2'2,in a system where there is no need of intermediary object, the body acting as an interface3. Enaction involves considering cognitive processes as a part of an environment, both resulting from and structuring it, rather than a recreation of an outside world. Enacting Music? To a certain extent, considering music from the enactive point of view is very similar to what "emerges" in a typical free improvisation situation, where players "Ireact" to the sonic environment with their own "structural coupling" (i.e. musical tastes, background, vocabulary and technique, feelings). In the case of (electroacoustic) composition, an enactive process may involve the system taking into account the composer behaviour and propose structural coupling between objects. Objects themselves shouldn't be considered as "sound artifacts", but more in an ontological perspective (Vaggione 2001), as "objects", i.e. complex of hetereogeneous data types and processes. This is quite different from considering a composition as a temporally-fixed set of symbols, musically meaningful, but lacking a functional perspective in the poietic (i.e. compositional) sense. Another feature of such compounds is the strong emphasis put on interaction instead of "outside" algorithmic rules. This puts the composer in a more "active" state, in which he not only devises how the object should sound, but also how a 2As connectionnism, to a certain extent, invalidated the use of symbols. 3 Varela defines the cognition as a network consisting of multiple levels of interconnected, sensorimotor subnetworks. clearly outlining the role of the body as an interface given object should sound when put under certain conditions. Using objects, the composer not only determine morphological features, but also behavioral aspects of sound structures to be, hence defing the composition as a dynamical system, instead of an exemplification of a rules/model system (Giunti 1997). 3.2 Enactive Computer Music Environments Several criterions must be met if we are to achieve the definition of a computer music composition system allowing enaction: * Objects and processes must be initially untyped, * Interaction must take place at any level of the environment4, * There must be no assumptions on the associative rules, and on the manipulative rules, * Meaning must be enacted through constant interaction between the different parts of the environment (i.e. the composer, the software, the materials)5. Structuration and articulation of events may then be planned at any level of the time scale, from micro- to macrolevel, without needing special processes tied to morphologically induced constraints. This involves providing the system with basic sets of (untyped) operators, and let the user create associations between operators and data structures, therefore defining complex objects which can be used as basis for more complex operators. The process of object typing and definition is transferred back to the composer, allowing a truly dynamic approach to sound and music composition. This, however, does not prevent predefined constraints and rules/models to be used, but prevents the system itself to be the basis for such assumptions, which can impair the composer's perspective and hence the composing strategy (Moore 2006). There is a question left, that of visually representing the information contained in a enactive composition environment. The main possibility is to allow any type of data representation: statistical, physical, symbolic, to be displayed, leaving the user with a choice depending on the situation, and allowing symbolic representation without specific meaning for the system, as in figure 1. 4 That includes interaction between objects, between the environment and the objects, between the user and the objects, between the environment itself and the composer 5This implies no predefined model or set of rules. 533
Page 00000534 --- - @0.0 e --p Figure 1: symbolic representations in Sketcher 4 Conclusion Creating such an environment may seem a difficult task at first, on the engineering side. However it seems a natural evolution of the computer music composition systems: from algorithmic, systems evolved to object-orientation, and from there could benefit from more complex computer engineering techniques such as aspect-oriented programming and interaction-based techniques. Using the enaction paradigm, computers could change from their current "number-crunching" device role to something more challenging for a composer, taking a more central place in the composition process, evolving from tool to medium, and from a static environment to a dynamic system. These issues raise important questions, since there is a need to define new strategies for internal representation of sound objects and structures. Common denominators are needed to define "untyped" objects and their associated representational strategies. 4.1 Acknowledgements The author wishes to thank Horacio Vaggione and Myriam Desainte-Catherine for numerous inputs and support for this ongoing research. 4.2 References References Assayag, G. (1993). La Composition Assistee par Ordinateur, Chapter CAO: Vers la Partition Potentielle. Number 3. Paris: IRCAM. Budon, 0. (2000). Composing with objects, networks, and time scales: an interview with horacio vaggione. Computer Music Journal 24(3), 9-22. Cope, D. (2001). Virtual Music. Cambridge, Mass.: MIT Press. Courtot, F. (1992). CARLA: Acquisition et Induction sur le Materiau Compositionnel. Ph. D. thesis, IRISA. Dahan, K. (2005). An associative approach to computer-assisted music composition. In Proceedings of the ICMC 2005 Conference, Barcelona, pp. 223-226. ICMA. Dahan, K., G. Brown, and B. Eaglestone (2003). New strategies for computer-assisted composition system: a perspective. In Proceedings of the International Computer Music Conference, Singapore. ICMA. Drummond, J. (1997). Integrated development environment for computer music composition. In Proceedings of the ICMC 97 Conference, Thessaloniki. ICMA: ICMA. Falk, J. (1958). Precis Technique de Composition Musicale, theorique etpratique. Alphonse Leduc. Feneyrou, L. (Ed.) (2001). Jean Barraque: Ecrits, Volume 3 of Esthetique. Paris: Publications de la Sorbonne. Gibson, J. J. (1979). The Ecological Approach to Visual Perception. Boston: Houghton Mifflin. Giunti, M. (1997). Computation, Dynamics and Cognition. Oxford: Oxford University Press. Lerdahl, F. and R. Jackendoff (1983). A Generative Theory of Tonal Music. Cambridge, Mass.: MIT Press. McAdams, S., S. Winsberg, G. de Soete, and J. Krimphoff (1995). Perceptual scaling of synthetized musical timbres: Common dimensions, specificities and latent subject classes. Psychological Research (58), 177-192. Merleau-Ponty, M. (1945). Phenomenologie de la Perception. Paris: Gallimard. Mesnage, M. and A. Riotte (1993). La composition assistee par ordinateur, Chapter Modelisation informatique de partitions, analyse et composition assistee. Les cahiers de l'IRCAM. Paris: IRCAM. Messiaen, 0. (1945). Technique de mon language musical. Paris: Alphonse Leduc. Moore, A. J. (2002-2006). Personal Communications. Nuhn, R., B. Eaglestone, N. Ford, A. J. Moore, and G. Brown (2002). A qualitative survey of composers at work. In Proceedings of the International Computer Music Conference, Cuba, pp. 572-598. ICMA. Peirce Edition Project (Ed.) (1998). The Essential Peirce. Selected Philosophical Writings (1893-1913), Volume 2. Indiana University Press. 534
Page 00000535 Popper, K. R. (1934). Logik der Forschnung: Zur Erkenntnistheorie der Modernen Naturwissenschaft. Vienna. Smalley, D. (1986). The Language of Electroacoustic Music, Chapter Spectromorphology and Structuring Processes. London: MacMillan. Vaggione, H. (2001). Some ontological remarks about music composition processes. Computer Music Journal 25(1), 54 -64. Varela, F. (1991). The Embodied Mind: Cognitive Science and Human Experience. Cambridge, Mass.: MIT Press. Whalley, I. (2005). Software agents and creating music/sound art: Frames, directions, and where to from here In Proceedings of the International Computer Music Conference, Barcelona, pp. 691-694. ICMA. 535