Encoding of complex sounds in the auditory midbrain
by Dominika Lyzwa
Date of Examination:2014-06-17
Date of issue:2015-06-16
Advisor:Prof. Dr. Theo Geisel
Referee:Prof. Dr. Florentin Wörgötter
Referee:Prof. Dr. Theo Geisel
Files in this item
Name:Diss_Lyzwa.pdf
Size:11.8Mb
Format:PDF
Abstract
English
How complex natural sounds such as speech and vocalizations are encoded in the main converging center of the auditory midbrain is not yet fully understood. For multi-units, which are composed of several single neurons, the representation of individual vocalizations is investigated across gradients of neural preferences in the inferior colliculus. The multi-units allow making inferences about collective neural behavior. The cross-correlation of averaged spiking responses is used to analyze if the vocalization representation depends systematically on the spatial distance between multi-units. The measure of response correlations and correlated variability indicates if interactions between the multi-units exist. Neural discrimination is employed to examine whether the whole set of spectrotemporally rich vocalizations can be reliably distinguished. The question is addressed whether along the frequency gradient of the inferior colliculus, individual vocalizations are encoded following their specific spectral content. This leads to the question whether vocalizations are accurately represented by individual multi-units, or, if groups of multi-units provide a more detailed representation, and if the multi-units interact to facilitate encoding. Some of these vocalizations display envelope correlations, and neurons might have neural preferences to these correlations, allowing them to encode the complex sounds in more detail. Reverse correlation analysis can capture nonlinear neural response properties. This analysis is used to probe preferences of single neurons to stimulus correlations, separately for different frequency carriers of the envelope, and to characterize the neural spectrotemporal tuning. It is found that similarity of multi-unit spiking responses to vocalizations decreases linearly with spatial distance across the inferior colliculus, and extends to a few hundred micrometers. Neural correlations which are due to interactions between the multi-units are substantial but do not improve separability. Since interactions with other multi-units do not improve the neural separability, the multi-units act as independent encoders of vocalizations. However, neural discrimination improves substantially when the responses from more than one multi-unit are considered. In many cases, combining responses from a few multi-units results in a perfect discrimination of the whole set of vocalizations. The individual vocalizations are reliably encoded across wide frequency ranges, thus yielding a robust representation, which is relevant in a behavioral context. Optimal encoding does not strictly follow the spectral content of the vocalizations. The analysis of receptive fields reveals that the majority of the tested single neurons displayed preferences to envelope correlations. These are mostly found at the preferred frequency of the neuron and exhibit temporal tuning properties that enhance or modulate the neuron’s linear response. In conclusion, vocalizations are reliably encoded by independent multi-units across a wide frequency range of the inferior colliculus. Neural preferences to envelope correlations are present in the inferior colliculus and are likely to facilitate encoding of the vocalizations.
Keywords: information processing; complex sounds; auditory; inferior colliculus