Auditory Streaming: Behavior, Physiology, and Modeling

Abstract

Auditory streaming is a fundamental aspect of auditory perception. It refers to the ability to parse mixed acoustic events into meaningful streams where each stream is assumed to originate from a separate source. Despite wide interest and increasing scientific investigations over the last decade, the neural mechanisms underlying streaming still remain largely unknown. A simple example of this mystery concerns the streaming of simple tone sequences, and the general assumption that separation along the tonotopic axis is sufficient for stream segregation. However, this dissertation research casts doubt on the validity of this assumption. First, behavioral measures of auditory streaming in ferrets prove that they can be used as an animal model to study auditory streaming. Second, responses from neurons in the primary auditory cortex (A1) of ferrets show that spectral components that are well-separated in frequency produce comparably segregated responses along the tonotopic axis, no matter whether presented synchronously or consecutively, despite the substantial differences in their streaming percepts when measured psychoacoustically in humans. These results argue against the notion that tonotopic separation <italic>per se</italic> is a sufficient neural correlate of stream segregation. Thirdly, comparing responses during behavior to those during the passive condition, the temporal correlations of spiking activity between neurons belonging to the same stream display an increased correlation, while responses among neurons belonging to different streams become less correlated. Rapid task-related plasticity of neural receptive fields shows a pattern that is consistent with the changes in correlation. Taken together these results indicate that temporal coherence is a plausible neural correlate of auditory streaming. Finally, inspired by the above biological findings, we propose a computational model of auditory scene analysis, which uses temporal coherence as the primary criterion for predicting stream formation. The promising results of this dissertation research significantly advance our understanding of auditory streaming and perception.