Page  00000001 THE PSYCHO-ADAPTIVE LISTENING MACHINE: AN APPLICATION OF PERCEPTUAL CONTROL THEORY TO COMPUTER MUSIC Erik Gottesman 3607 Jane Dr. Midland, MI 48642-9751 USA emgottes @umich.edu Abstract The Psycho-Adaptive Listening Machine (PALM) is a conceptual framework for composition and the study of musical cognition. PALM is based on perceptual control theory, a theory of animal behavior. 1 Introduction Living organisms do not produce repeatable actions, they produce repeatable results. When we make aesthetic evaluations of the results of creative behavior, we do so in a dynamic fashion, influenced by preconditioning, intrinsic and extrinsic disturbances. To understand this process, requires insight into how we accept and interpret environmental stimuli. The computer is an appropriate tool for the study of this phenomenon. 2 Background The use of electrical signals emanating from living organisms (bioelectric signals) to create music is not new. While a complete history of bioelectric music experiments is beyond the scope of this paper, some background is appropriate. The history of bioelectric music experiments can be traced back to 1934 with the translation of the human electroencephalogram (EEG) into audio signals (Adrian and Matthews 1934); these experiments went largely unnoticed until the late 1960s. In the decades that followed, the efforts of many composers and artists (Lucier 1976; Arcadiou 1986; Rosenboom 1976, 1989; Knapp and Lusted 1990) established work in this arena as meritorious. These systems may produce some interesting musical results, but they will tend to be stylistically limited, and due to their reliance on ill-structured methods of acquisition and analysis they tell us very little about the underlying processes of musical cognition. We must seek a new paradigm that closely resembles our own context-sensitive behavioral practices, both innate and experiential. 3 The Nature of Control In the field of psychology, theories of behavior are implemented with the intent of explaining phenomena related to conscious and unconscious processes via explicit mechanisms assumed to be simpler than the phenomena explained (Laske 1974). Perceptual control theory (PCT) is a theory of human and animal behavior, an extension of control theory principles first explored in the early 1950's by the cyberneticists Weiner and Ashby; however, William T. Powers' seminal 1973 book, Behavior: The Control of Perception, contributed the first substantial insight into the requirements and consequences of applying physical science to the field of psychology. In accordance with other theories of behavior, PCT strives to explain a phenomenon, namely the phenomenon of control. Let us define control as follows: A is said to control B if for every disturbing influence acting on B, A generates an action that tends strongly to counteract the effect of the disturbing influence on B. Using this definition, we can begin to construct a useful system for the explication of behavior via bioelectric signals. Such a system is not a biocontroller (Knapp and Lusted 1990), as it does not control anything. Rather, it provides something far more useful: schemata for the process of perceptual control.

Page  00000002 As we lead our daily lives, we encounter a massive amount of data, a veritable plethora of meaningless syntactical mechanisms, potential energy. As living symbol-manipulating systems, we must convert this staggering amount of data into active knowledge, a sort of kinetic energy that carries with it assigned semantic meanings. Only when information is (internally) represented in this manner does it become useful. We define peiformance as the general set of creative processes that contribute to musical behavior. Consider the study of performance via a signal-processing system (SPS). An SPS must treat the results of musical behavior as a dependent variable, a response to stimuli. Such a model cannot provide an explanation for the achievement of consistent ends through varying means. Under this model, extensive statistical analysis is required to reconstruct the intermediary processes that terminate in some musical result. Such an analysis provides modest (and often meaningless) correlations. It is unfortunate that nearly all bioelectric music experiments have adopted this obviously inadequate model, relegating their function to one of artificial and arbitrary generation and control. An information processing approach is more appropriate. As opposed to an SPS that operates on signals, an information processing system (IPS) performs analyses and transformations on symbolic representations of data. Making use of this paradigm, PCT becomes readily applicable to the study of performance. PCT contends that we are not responsible for (musical) behavior, but (musical) perceptions; musical behavior is controlled through our perceptions. Thus, performance may be described as a collection of programs or perceptual control schemata. It appears that we have conceived a truly useful tool, a vehicle for the real-time study of musical cognition and perception. 4 Conceptual Framework The Psycho-Adaptive Listening Machine (PALM) is a conceptual framework for the study of natural and artificial creative processes. At the most abstract level, PALM contains two distinct entities, the listener and his surrounding environment. The listener is an interacting entity which may at any time assume the role of composer, audience, or both simultaneously. The environment is separated into two subenvironments, the first being a spaLtio-temporal musical environment containing all dependent sonic effects readily influenced by the listener, and the second being the sum of all independent effects as a single equival~ent disturbalnce. The organization of the listener model is significantly complex, as it must account for both innate and experiential processing at multiple syntactic and semantic levels. The adopted model draws upon many sources, but pays its greatest tribute to the brilliant insight delineated in Otto Laske' s many papers on cognitive musicology (Laske 1972, 1974, 1975, 1988) and Robert Newcomb's Music In Thle Air. A Thleoretical Model and Software System for 1Music Analysis and Composition (Newcomb 1998). The listener receives all environmental stimuli as a continuous signal. Before perceptual grouping may occur, lexical processing must be performed to break down the continuous input signal into discrete, atomic units called genotypes, representing the collection of all potentially perceptible events. In turn, syntactic and sonological parsing filters and groups these genotypes into a new collection, called phenotypes, representing all readily perceptible events. The chosen nomenclature is one adapted for our purposes from Dora Hanninen's A General Thleory for Context-Sensitive Music Analysis (Hanninen 1996). As stated above, phenotypes are selected based upon sonic and contextual criteria specific to the listener' s personal behavioral profile. As suggested by Hanninen, sonic criteria define boundaries and imply segments while contextual criteria define segments and imply boundaries through the recognition of associations between two or more atomic units (HaLnninen 1996). This is important to realize since previous experience dictates what sonic and contextual criteria are available, thus directly affecting the listener's ability to distill information from an ongoing stream of stimuli. It should also be noted that this ability to differentiate between sonic and contextual criteria is one of the many deficiencies of the SPS approach discussed earlier. The modeling phase, implemented at multiple syntactic and semantic levels, assigns meaning to phenotypes found during the parsing phase. The listener' s ability to extract meaning is again based on an experiential database psychologists refer to as long-term memory (LTM). Functionally speaking, the modeling phase identifies, in the form of a problem, any musical effects (non-equilibria). Finally, the

Page  00000003 solving phase poses possible actions to counteract these effects. The output function is the solution to the posed problem, a set of actions that control our environment based upon our perceptions. Listener Environment Output Quantity Fig. 1 PALM Block Diagram 5 Divisive Issues and Applications Compositional activity may be either purely exploratory or oriented to the achievement of some goal-state (i.e. binary form). PALM is well-suited to activity as it assumes either of these forms; however, the greatest strength of PALM is in it' s ability to explore specific problems related to musical cognition. For instance, consider a perceptual control schema, as described earlier and pictured in figure 2a. This schema is a progression of knowledge states (Sn) via transformations (Tn) over time, until some terminal state is reached. Now consider the case in which certain intermediary knowledge states are reordered or omitted from the same progression as in figure 2b. Such a difference might be noted when two students solve a counterpoint problem, doing so via different means. If such a schema containing omitted states converges to the same terminal state as a similar schema, then we might refer to these schema as ergodic, belonging to the same grammnar. This phenomenon appears to be the case for extrinsic grammnars, which are taught, as opposed to intrinsic grammrars, which are learnt, may vary radically from one listener to the next, and imply stylistic similarity. Si1 12 tS3 14_ 55 i 1 5 Ti 5f2 13 5~f4 15 5 6-f rTi 52 13 54I 15 56S (a) (b) Fig. 2 (a) Complete Perceptual Process (b) Perceptual Process with Reord-ering/Omission

Page  00000004 Another issue of great intrigue is how sensations of pleasure manifest themselves in a sonic environment and what factors might promote or impede upon such sensations. Recent research has presented significant proof of the existence of both conscious and unconscious defense mechanisms in the brain (Shevrin et al. 1996). PALM's sensitivity to semantic encoding makes it adequate for further study into this phenomena and its application to computer music composition. Additionally, PALM offers great promise in the areas of rehabilitation and music therapy. 6 Summary We are now endowed with a system by which we can easily study concepts of musical cognition such as the roles of environmental disturbance and deterministic and synthetic parameterizations. More importantly, we may reap unparalleled benefits by utilizing the results of our rigorous analysis in a compositional capacity (Newcomb 1998). Thus, PALM provides a radically new approach to the study of performance. Through modularity, ergonomic design and artistic orientation we may indulge ourselves in the exploration of art's psychological roots, without infringing on art's power over the body and imagination. 7 References [Arcadiou 1986], Stelios Arcadiou (STELARC), Beyond the Body: Amplified Body, Laser Eyes, & Third Hand.NMA, Vol. 6, pp. 27-30, NMA Publications, 1986. [Hanninen 1996], Dora A. Hanninen, A General Theory for Context-Sensitive Music Analysis, Society for Music Theory National Conference, Baton Rouge, Louisiana, October 31, 1996. [Knapp and Lusted 1990], R. Benjamin Knapp and Hugh S. Lusted, A Bioelectric Controller for Computer Music Applications. Computer Music Journal, Vol. 14, No. 1, pp. 42-47, The MIT Press, 1990. [Laske 1972], Otto E. Laske, On Musical Strategies with a View to a Generative Theory of Music. Interface, Vol. 1, pp. 111-125, Swets & Zeitlinger, 1972. [Laske, 1974], Otto E. Laske, The Information Processing Approach to Musical Cognition. Interface, Vol. 3, pp. 109-136, Swets & Zeitlinger, 1974. [Laske 1975], Otto E. Laske, Toward a Theory of Musical Cognition. Interface, Vol. 4, pp. 147-208, Swets & Zeitlinger, 1975. [Laske 1988], Otto E. Laske, Introduction to Cognitive Musicology. Computer Music Journal, Vol. 12, No. 1, pp. 43-57, The MIT Press, 1988. [Lucier 1976], Alvin Lucier, Statement On: Music for Solo Performer, 1971, in (Rosenboom 1976). [Newcomb 1998], Robert Newcomb, Music In the Air: A Theoretical Model and Software System for Music Analysis and Composition, Organised Sound, Vol. 3, No. 1, Cambridge University Press, 1998. [Rosenboom 1976], David Rosenboom, ed., Biofeedback and the Art, Results of Early Experiments, 2nd ed. Aesthetic Research Center of Canada Publications, 1976. [Rosenboom 1990], David Rosenboom, The Performing Brain. Computer Music Journal, Vol. 14, No. 1, pp. 48-66, The MIT Press, 1990. [Shevrin et al. 1996], Howard Shevrin et al., Conscious and Unconscious Processes, The Guilford Press, 1996.