are so far undiscovered. The paradigm seems better suited to unrehearsed and unprepared activities improvisation, for example, in which learning is not segregated from composition, or composition from performance.

New Paradigms: (6) Composition as a Genetic Process

There is no familiar ring to the phrase 'biological model of music', no well-worn series of thinkers who have subscribed to the view that musical procedures can be explained by concepts that derive from the study of living organisms. Yet recent experiments with algorithmic composition have revealed that to a striking degree many highly diverse repertories may be simulated, to degree that makes individual styles recognizable, from procedures that treat musical attributes as if they obeyed the rules of genetics. I refer in particular to the Experiments in Musical Intelligence by David Cope, a composer and professor of music at the University of California at Santa Cruz. At the outset, Cope's interest was in identifying more precisely than music theorists, historians, and analysts had done the procedures that typified the style of a given composer. He wanted to know for his own benefit as a composer what made Mozart Mozart and Bach Bach. He made the assumption, based on readings in computational linguistics and artificial intelligence, that there could be a universal procedure on which most musical composition is founded and in which only the details varied from case to case.

His own explanations of his work favor the language of linguistics and artificial intelligence. The core procedure is a 'grammar' and the points at which individual identity is acquired are likened to the parts of speech. I find it easiest to think of this top-down approach as an analogue to the American parlor game Mad-Libs. A Mad-Lib is a short story missing most of its nouns, verbs, and adjectives. A group of players is asked to construct a list of nouns, verbs, and adjectives in an appropriate order. Then a host reads the story with each entrant's list of words inserted in the appropriate places. Each telling is different. Some are hilarious, some non-sensical.

In place of the players and their lists, Cope assembles lexicons of material gathered from pieces of a like nature that have been closely inspected and carefully dissected. He segregates short strings of pitches, of durations, and of other attributes of the repertory under investigation. In the pattern-matching phase, he looks for concordances among the stored items. The number found will depend, of course, on the parameters for length and so forth, in short on the quality of the match sought. Only strings that are matched are preserved; those that are matched too abundantly are discarded. In the style-specification phase, he runs a query program that enables him to determine the external specifications for a new work its meter, key, and so forth. In the generative phase, a composing program is run. This program seeks to conjugate, within the prescribed style, five musical analogues of parts of speech statement, preparation, exten-