Mesoscopic dynamics of pitch processing in human auditory cortex.

Abstract

Pitch is a perceptual correlate of sound periodicity elicited by vibrating bodies; it plays a crucial role in music and speech. Although perceptual phenomenology of pitch has been studied for centuries, a detailed understanding of its underlying neural mechanisms is still lacking. Early theories suggesting that pitch is decoded in the peripheral auditory system fail to explain the perception of complex stimuli. More recent mechanistic models, focused on how subcortical structures process periodic discharges of the auditory nerve activity, are unable to explain fully key aspects of the processing dynamics observed during electrophysiological recordings. In this thesis, we propose a novel theory describing how subcortical representations of pitch-related information are integrated in cortex and how this integratory process gives rise to the dynamics observed in magnetoencephalographic (MEG) experimental recordings. Auditory evoked fields recorded with MEG reveal a systematic deflection around 100 ms after stimulus’ onset known as the N100m. This deflection consists of several components reflecting the onset of different perceptual dimensions of auditory stimuli such as pitch, timbre and loudness. The exact latency of the component elicited by pitch onset, known as the pitch onset response (POR), shows a strong linear relationship with the pitch of the stimulus. Our theory links the POR latency with processing time and explains, in a quantitative manner, the substrate of the relationship between processing time and pitch. Cortical integration is described using a model of neural ensembles located in two adjacent areas, putatively located along the lateral portion of Heschl’s Gyrus in human auditory cortex. Cortical areas are hierarchically structured and communicate with each other in a top-down fashion. Pitch processing is modelled as a multi-attractor system whose dynamics are driven by subcortical input. After tone onset, the system evolves from an initial equilibrium position to a new equilibrium state that represents the pitch elicited by the tone. A computational implementation of the model shows that: 1) the transient dynamics between equilibrium points explains the POR; 2) the latency of the transient is directly linked with the time required to reach the new equilibrium state; and 3) that such processing time depends linearly on the pitch of the stimuli. Our theory also addresses the problem of how tones with several simultaneous pitch values are processed in cortex. In Western music, dyads comprising tones with different pitch values are judged as more consonant or more dissonant depending on the ratio of the periods of the involved sounds. The latency of the POR evoked by such dyads also presents a strong correlation with the perceived consonance: dissonant dyads generate later PORs than consonant dyads. Our theory of pitch processing describes consonance (dissonance) as a direct effect of harmonic collaboration (competition) during the cortical integration process: the cortical model shows that harmonic collaboration facilitates convergence, explaining why dissonant dyads require longer processing times and evoke later PORs than consonant dyads.