Computational musicology is broadly defined as the study of Music by means of computer modeling and simulation. Evolutionary Computation-based modeling is particularly suitable to study the origins and evolution of music. This is an innovative approach to a puzzling old problem: if in Biology the fossils can be studied to understand the past and evolution of species, these ‘fossils’ do not exist in Music; musical notation is a relatively recent phenomenon and is most prominent only in the Western world.

If there is no possibility to trace back the rhythms that were created before musical representation was developed, we can try to simulate this behavior by creating artificial ‘worlds’ where agents interact by producing, listening and evaluating the sequences. The purpose of this research is not only to present hypothesis on why certain rhythms in real societies have evolved and others didn’t, but also to simulate hypothetical societies of players that can be used in a creative perspective.

There are two main questions to address here: one is the definition of the biological substrate of the agent, and the other is the interaction procedures that enable learning and consequent evolution of rhythms. Regarding the first question each agent needs an architecture that can solve the problems of synchronization, perception of beat hierarchy, quantization, categorization and memorization of rhythms. The second question, concerning the interaction, looks for very simple rules of communication and learning, which are responsible for changes to the internal state through an adaptive procedure.

This research presents a framework to study the evolution of rhythms in a society of virtual agents. This framework consists of a model of an agent containing a synchronization algorithm and connectionist approach to the categorization and memorization mechanism. The agents have different presets regarding their initial sense of pulse, emulating the human characteristic known as ‘Natural Timing’. Then, the experiments regarding the synchronization of the agents’ inter-onset intervals will be explained, as well as other experiments regarding the choice of a neural network that can code temporal sequences. Finally we will show details of the simulation algorithm, and present all the results achieved so far.