Cognitive Science seminar: Learning to listen, listening to learn

When: Wednesday, June 18, 2025, 11:00 AM - 12:00 PM
Where:

Title: Learning to listen, listening to learn: Music perception as a process of probabilistic prediction based on statistical learning

Abstract: Like natural language, music is an ubiquitous auditory component of human culture with most individuals achieving maturity in music perception through their experience of developing within a musical culture, just as they achieve maturity in speech perception, through developing within a linguistic culture. How does this maturity arise? Traditionally, the debate has been dominated by nativism but this has been challenged in recent years by two developments: first, increasing research on non-Western cultures has highlighted cross-cultural differences in music perception; second, more powerful computational models of learning have shown that it is possible to simulate music perception without positing pre-existing innate representations. My thesis is that music perception reflects an ongoing process of statistical learning of regularities in auditory experience limited by architectural constraints on learning but not innate representational constraints. I will introduce a computational model that generates probabilistic predictions based on statistical learning (both long-term and short-term) and show that it simulates a wide variety of psychological processes involved in music perception: expectation, complexity perception, auditory memory, perceptual segmentation, affective experience and pleasure. The model also predicts cross-cultural differences in music perception and the developmental trajectories that produce them. I will conclude by considering the limitations of the approach and the most pressing questions remaining to be answered.

Location: Peter Landin Room 4.24, 4th floor

Key contact name: Haim Dubossarsky