|dc.description.abstract||The neuron is a powerful computational unit. However, neurons do not exist in isolation, rather they are embedded within intricate networks where each neuron is wired to tens of thousands of other neurons. Because of this interconnectivity the co-variability and timing of responses across neurons has significant impact on signal propagation. On the other hand, response co-variability is inherently linked to the underlying circuitry. In this thesis I seek to improve our understanding of neural coding in the auditory system through the analysis of neuronal spiking correlations. This work integrates computational models and analysis of the effect of correlations on information with physiological recordings throughout the sound localization system of the barn owl.
The barn owl midbrain contains a topographic map of auditory space in the optic tectum (OT). Conversely, maps are not found in the forebrain. Neurons in the auditory arcopallium (AAr), a forebrain region involved in sound localization display stereotypic monotonic tuning to auditory space with preference to the contralateral side and a sharp increase across the midline. This particular tuning shape is predicted by models of sound localization in mammals, the two-channel rate-code, where the firing rate of hemispheric populations with opposing tuning signals the laterality of the sound source. We examined the correlation structure of this population and its implications regarding information.
The information encoded by a neural population depends on its correlation structure. AAr neurons displayed similar spatial tuning across recording sites, and nearby cells showed low response co- variability (known as noise correlation), a correlation structure that may increase sparseness.
Decoding analysis of the data matched the expectation that positive noise correlations impaired performance, suggesting that, given the tuning properties of AAr neurons, low noise correlations would be beneficial for sound source discrimination. On the other hand, nearby neurons in OT displayed high noise and temporal correlations, suggesting shared inputs. Taken together the auditory forebrain in the barn owl displays a correlation structure which is in stark contrast to the midbrain, supporting the hypothesis that these areas rely on different coding schemes.
The firing pattern of neurons throughout the auditory system is entrained by the amplitude modulations of the sound, known as the envelope. Sensitivity to the envelope is linked to the spectrotemporal tuning of neurons. In the brainstem of the barn owl, the tunings to binaural spatial cues and spectrotemporal sound features are interdependent. However, whether this interdependence has implications for coding sound location remains an open question. We investigated this question in the midbrain where the convergence of brainstem pathways underlies the first emergence of space specific neurons. Here, the across-trial spike-timing reproducibility, a metric of entrainment, was correlated to response strength as sound direction changed. We hypothesized that this effect would drive temporal correlations of downstream OT neurons, which we had previously shown share inputs. This was indeed the case as neurons in OT displayed both across-trial reproducibility and synchrony which were rate-dependent.
The findings presented here are significant for supporting the hypothesis of a rate-code for sound localization in forebrain regions, as postulated in mammalian systems. In addition, we provide a link between temporal entrainment by sound features and spatial sensitivity in the midbrain of the barn owl.||en_US