Neural coding of time-varying interaural time differences and its relation to perception

Abstract

In natural environments, sounds are often not static. Usually, moving objects require the most attention, e.g. for identifying the presence and direction of a moving vehicle, or detecting and tracking the trajectory of a predator or prey. Faster time-varying location cues can occur in acoustic environments containing many spatially distributed sound sources, like at a cocktail party. In this case, we can identify the locations of the sources by "glimpsing" at short-duration localization cues when the sound energy from one source dominates the mixture. Even faster time-varying spatial cues result from reverberation in an echoic environment and we perceive them as spatially diffuse. We qualitatively perceive motion, a cocktail party, and reverberation differently, and these three percepts are determined by how quickly the spatial cues are moving. How these percepts come about in the auditory system is unknown. Here, we studied how neurons encode time-varying location cues and how the neural code relates to perception. Our focus was on time-varying interaural time differences (ITD), one of the main cues for localizing sounds in the horizontal plane. We recorded from single neurons in the inferior colliculus (IC) in the auditory midbrain of unanesthetized rabbits. The IC is the site of an obligatory synapse in the auditory pathway and one of the first stages of processing following the initial extraction of spatial cues in the brainstem. We hypothesized that the IC exhibits limitations in its ability to encode time-varying ITD that give rise to these different percepts. First, we show that IC neurons are more "sluggish" on average at synchronizing to the time-varying ITD than to amplitude modulations presented at a static ITD. Binaural sluggishness has been proposed based on human psychophysics but never validated neuro physiologically in the IC. Second, we show that most neurons are unable to synchronize to the time-varying ITD at speeds where humans no longer perceive fluctuations. Instead, neurons exhibit a change in average firing rate that corresponds to binaural decorrelation of the noise for very fast time-varying ITD, and this may explain the percept of a spatially diffuse sound at these speeds. We further recorded neural responses to slow-moving ITDs in opposite directions within the range of perceived motion. Using a generalized linear model to parse the neuron's response into ITD-following and direction selectivity components, we show that the responses of IC neurons are dominated by their ability to follow the ITD more than direction selectivity. In parallel experiments, we asked human participants to either identify the motion direction or detect the slow-moving ITD in the same stimuli and determined the threshold durations for direction identification and for detection for each participant. Direction identification threshold durations were larger than detection threshold durations. We then implemented neural classifiers that either identified the motion direction or detected the slow-moving ITD based on single-neuron responses to the stimuli, and we found that the classifier exhibited duration thresholds that matched human thresholds on both tasks. Together, these results suggest that temporal limitations of neural responses in the IC may give rise to the limiting speeds of time-varying localization cues where we perceive motion, "glimpse" the position of a source amidst a mixture, and perceive a spatially diffuse background in a reverberant environment.