Peter Todd

Abstract

Prehistoric musical behavior did not fossilize very well, and relatively few species

are alive today with which we can take a comparative approach to the origins

of human musical ability. Evolutionary computer simulations provide another

means of exploring this question: by constructing a population of artificial music

producers whose behavior can be selected by various fitness-determining critics,

we can study hypothetical scenarios for the evolution of musical behavior. To

date, most simulations of this type were constructed to create new forms of com-

puterized music composition systems, rather than to answer scientific questions.

But we can survey these systems and the simulation techniques they use to learn

about the effects of different types of knowledge representations in music cre-

ators and critics on the evolutionary process. In addition, a simulation was explic-

itly designed to explore the question of the evolutionary impact of various forms

of selection, focusing on sexual selection by coevolving male song producers and

female song critics. In this model, coevolving creators and critics can increase

the diversity of musical behaviors seen both across generations and within any

one population. This kind of simulation approach to the evolution of musical

behavior can answer other questions as well.

Our ancestors did not leave us much to go on for piecing together the

evolution of our musical abilities. A disputed bear-bone fragment that

may or may not have been a Neanderthal flute some 40,000 years ago

(see Kunej and Turk, this volume) is one of the few clues we have to our

musical heritage; most of this ephemeral behavior did not leave a fossil

trace. A few species are around today whose quasi-musical behavior we

can compare, as other chapters in this volume attest. But environmental

pressures acting on this aspect of the evolution of songbirds, whales,

gibbons, and humans may have been rather different, and we are still left

with little evidence of past behavioral changes. We would like to be able

to replay the evolutionary tape from the beginning to hear the whole

piece, but all we have available are a few scattered snippets along with

bits from the end of some individual species’ records.

If we cannot replay the original tape, perhaps we can make a new one.

To gain insights into the origins of music in another way, we can turn to

a method of exploration that was technologically unfeasible even a

couple of decades ago: evolutionary computer simulations. Using this

approach we can construct artificial “worlds” within the computer in

which populations of simulated creatures create and possibly perceive

musical signals (typically just represented as sequences of numbers,

rather than as actual physical vibrations). Loosely speaking, if we allow

these creatures to reproduce differentially in response to their musi-

cal behaviors (e.g., those that create a certain kind of song might have

more offspring), successive generations of creatures will evolve different

20

Simulating the Evolution of Musical Behavior