Neural sensitivity to interaural time differences: beyond the Jeffress model.

Abstract

Interaural time differences (ITDs) are a major cue for localizing the azimuthal position of sounds. The dominant models for processing ITDs are based on the Jeffress model and predict neurons that fire maximally at a common ITD across their responsive frequency range. Such neurons are indeed found in the binaural pathways and are referred to as "peak-type." However, other neurons discharge minimally at a common ITD (trough-type), and others do not display a common ITD at the maxima or minima (intermediate-type). From recordings of neurons in the auditory cortex of the unanesthetized rabbit to low-frequency tones and envelopes of high-frequency sounds, we show that the different response types combine to form a continuous axis of best ITD. This axis extends to ITDs well beyond that allowed by the head width. In Jeffress-type models, sensitivity to large ITDs would require neural delay lines with large differences in path lengths between the two ears. Our results suggest instead that sensitivity to large ITDs is created with short delay lines, using neurons that display intermediate- and trough-type responses. We demonstrate that a neuron's best ITD can be predicted from (1) its characteristic delay, a rough measure of the delay line, (2) its characteristic phase, which defines the response type, and (3) its best frequency for ITD sensitivity. The intermediate- and trough-type neurons that have large best ITDs are predicted to be most active when sounds at the two ears are decorrelated and may transmit information about auditory space other than sound localization.