Dissociating the neural codes for multiple pitch perception in humans

Abstract

Pitch is crucial for speech and music perception, yet its neural code remains contested. The classic rate-place theory suggests that pitch is computed from spatial patterns of average neural firing rates along the cochlear tonotopic axis. These cues are thought to be robust for isolated harmonic complex tones (HCTs) but degraded for spectrally dense mixtures of simultaneous HCTs (e.g., musical chords) due to auditory filtering. In many cases, pitch perception remains feasible for HCT mixtures, but it is unclear whether listeners utilize residual rate-place cues or alternative neural codes. To adjudicate between these possibilities, we generated rate-place metamers (META), which are synthetic stimuli with simulated auditory-nerve average-rate responses that are nearly identical to those of original pitch-evoking stimuli (ORIG). Using these stimuli, listeners completed four psychophysical experiments that spanned a wide range of acoustic and cognitive complexity, from single-HCT pitch discrimination to harmonic-expectation judgments. Combining the behavioral data with predictive models of listener performance, we found that listener performance on single pitch discrimination tasks could be adequately explained by a rate-place model, with or without simultaneous pitch maskers (Experiments 1 and 2). More complex tasks involving attention to all pitches in a three-pitch mixture (Experiments 3 and 4) showed a pattern of results more difficult to account for with a rate-place model, perhaps suggesting a role for integration of temporal cues. Significance StatementEveryday listening typically involves parsing or integrating several simultaneous pitches, for example, while listening to music or conversations in noisy environments, rather than hearing a single, isolated pitch. Despite their ubiquity, the neural mechanisms for encoding multiple pitches are not well understood, which partly explains the limitations of modern assistive listening devices in noisy settings. To address this, we developed model-based synthetic stimuli to measure how distinct pitch cues contribute to pitch perception. This approach advances our understanding of the neural basis of music perception, with a goal of informing the design of more effective hearing prostheses.