UCSF

The primary auditory cortex (AI) is made up of highly interconnected populations of neurons that are responsible for integrating bottom-up auditory information from the lemniscal auditory pathway and top-down inputs from higher-order cortical areas. Despite that, most studies of information processing in AI focus on either single-unit spectro-temporal receptive field (STRF) estimation, or paired neuronal correlation analyses, and assume that AI neurons filter auditory information either as individual entities or pairs. Meanwhile, some recent studies have also shown how populations of AI neurons can also encode auditory behavior. Determining how AI encodes information will hence require an integrated approach that combines receptive field and multi-neuronal ensemble analyses.

In this dissertation, I show that I can accurately detect coordinated neuronal ensembles (cNEs), which we define as groups of neurons that have reliable synchronous activity, in AI. These cNEs are meaningful constructs that are active in both spontaneous and evoked activity, and their synchronous evoked activity cannot be trivially explained by receptive field overlap. cNEs also come in two flavors – one of them enhances stimulus representation over single neurons or simultaneously recorded random groups of neurons of the same size, while the other does not represent spectro-temporal features at all, and might reflect internally generated neuronal activity. Since single neurons can participate in multiple cNEs over the course of a recording, I also show that neurons can multiplex information, and encode slightly different spectro-temporal information, if they encode spectro-temporal information at all, when associated with different cNEs. Meanwhile, the enhancement of information processing and the reliability of representation of the stimulus by cNEs suggest that cNEs should be considered the principal unit of information processing in AI.