Over and Over: Understanding Non-Music to Music Transformations With A Predictive Coding Model

Abstract

Music is something of a universal language, transcending culture, nationality, age, and gender. But what makes music music — what makes a sequence of tones or sounds come together and tickle the ear in just the right way — is still for the most part not completely known to us. The speech-to-sound and sound-to-music phenomenon presents us with insight into the auditory processes that underlie the formation of musical perception, revealing repetition to be a necessary factor in this formation. Centering this musicalization process within the perspective of the predictive coding models of perception in the auditory system allows us to reason with the influences of repetition on this process. The mismatch-negativity (MMN) reflects the prediction error in auditory processing and can therefore be used as a probe to test if this predictive model is what drives the influence of repetition on perceptions of musicality. The speech-to-song illusion acts as a compelling analytical tool to supplement out knowledge the sound-to-music phenomenon. In analyzing both within a predictive coding framework, I piece together a potential model for their musical transformation that rely on prediction error and perception modes. A rudimentary experiment on myself resulted in replication of the sound-to-music illusion, as well as evidence supporting a moderate negative relationship between a sound's (median) fundamental frequency and the degree of its musical transformation. Further research on acoustic features such as fundamental frequency and pitch are necessary to not only broaden our understand of the STM illusion but probe for a predictive coding mechanism within it.