Laminar organization of vocalization processing and attentional modulation in auditory cortex

This thesis investigates the neural mechanisms underlying vocalization categorization in the auditory system, focusing on feature-based representations and attentional modulation. Using electrophysiology, behavior, and computational modeling in guinea pigs, we examined how vocalizations are encoded along the auditory pathway and how task engagement shapes these representations. Our recordings revealed that while neurons in the thalamus and primary auditory cortex (A1) layer 4 responded broadly to spectral content, a significant proportion of A1 layer 2/3 neurons exhibited highly selective responses to specific vocalization features. This emergence of feature selectivity supports theoretical models of auditory categorization based on intermediate complexity features. We then trained guinea pigs on vocalization categorization tasks and found that our feature-based computational model, trained only on natural calls, successfully predicted animal behavior across various spectro-temporal manipulations. Finally, we investigated how task engagement modulates vocalization processing in A1 using chronic recordings during active and passive listening conditions. We found that attention altered the input-output relationships, particularly in A1 layer 2/3 neurons which can potentially lead to improved separability between target and non-target vocalizations at the population level. Together, these findings demonstrate that feature-based vocalization representations emerge in superficial layers of auditory cortex and are further enhanced by attention, providing new insights into the neural computations underlying robust vocalization categorization in complex acoustic environments.