Decoding MEG responses to musical pitch reveals the dynamic emergence of tonal structure in human cortex

Abstract

Tonal music is characterized by a hierarchical structuring of pitch, whereby certain tones appear stable and others unstable within their musical context. Despite its prevalence, the cortical mechanisms supporting such a percept remain poorly understood. We examined the neural processing dynamics underlying pitch-structure in Western Tonal Music. Listeners were presented with tones embedded within a musical context whilst their magnetoencephalographic (MEG) activity was recorded. Using multivariate pattern analysis, decoders attempted to classify the identity of tones from their corresponding MEG activity at each peristimulus time-sample, providing a dynamic measure of their cortical dissimilarity. Time-evolving neural distinctions were then compared with the predictions of several acoustic and perceptual models. Following the onset, a temporal evolution was witnessed in the representational structure in cortex. While MEG dissimilarities between tones initially corresponded to their fundamental frequency separation, distinctions beyond 200 ms reflected their status within the hierarchy of perceived stability. Transposing dissimilarities corresponding to this latter period into different keys, neural relations between keys correlated with the well-known circle of fifths. Convergent with fundamental principles of music-theory and perception, results detail the dynamics with which the complex perceptual structure of Western tonal music emerges in human cortex within the timescale of an individual tone.Tonal music is characterized by a hierarchical structuring of pitch, whereby certain tones appear stable and others unstable within their musical context. Despite its prevalence, the cortical mechanisms supporting such a percept remain poorly understood. We examined the neural processing dynamics underlying pitch-structure in Western Tonal Music. Listeners were presented with tones embedded within a musical context whilst their magnetoencephalographic (MEG) activity was recorded. Using multivariate pattern analysis, decoders attempted to classify the identity of tones from their corresponding MEG activity at each peristimulus time-sample, providing a dynamic measure of their cortical dissimilarity. Time-evolving neural distinctions were then compared with the predictions of several acoustic and perceptual models. Following the onset, a temporal evolution was witnessed in the representational structure in cortex. While MEG dissimilarities between tones initially corresponded to their fundamental frequency ...