Timbre Encoding in the Inferior Colliculus

Timbre, or the quality of a sound, is a critical component in speech and music. One percept of timbre, brightness, is correlated with the spectral centroid and the spectral envelope of harmonic sounds. Little is known about how this aspect of timbre is encoded in the subcortical auditory system. We used physiological and computational modeling methods to investigate the representation of spectral peaks in a harmonic complex tone with a broad, triangular-shaped spectrum. Extracellular single-neuron recordings were made in the central nucleus of the inferior colliculus in awake, female Dutch-belted rabbits. A population response to the timbre stimulus was inferred by shifting the stimulus spectrum above and below the characteristic frequency of each neuron. Spectral peaks in the stimulus were encoded in peaks in the average-rate profiles of most neurons, and this representation was robust over a range of suprathreshold levels. Neural discrimination thresholds were also sufficient to describe human behavioral thresholds. Temporal responses were complex and often exhibited phase locking to the fundamental frequency and integer multiples of the fundamental frequency. Computational models that included neural-fluctuation sensitivity and amplitude-modulation-sensitive broadband inhibition captured the major trends in physiological results. These findings demonstrate that multiple mechanisms may influence robust spectral-peak encoding in inferior colliculus neurons. Significance Statement Spectral features, such as the spectral centroid and spectral peaks, are critical features of sounds with timbre, including speech and music. This study addresses a critical gap in knowledge about timbre representation in the auditory system by investigating neural encoding of spectral peaks in the inferior colliculus. Results demonstrate that spectral peaks are robustly represented in average rate responses across sound levels, and that neural discrimination thresholds are on a par with human thresholds. We found that a computational model with broad inhibition that is sensitive to neural fluctuations in peripheral responses may underly spectral-peak encoding. These findings extend previous vowel encoding studies to provide broader insights into timbre processing, with potential applications for improving hearing-aid and cochlear-implant technologies.