Mechanisms and response properties of duration-tuned neurons in the vertebrate auditory midbrain

Abstract

<p>This thesis aims to elucidate the mechanisms and response characteristics of neural circuits in the vertebrate brain capable of responding selectively to stimulus duration. The research within focused on, but is not limited to, auditory neurons; however, most of the results extend to other sensory modalities. These neurons are known, appropriately, as duration-tuned neurons (DTNs). Duration-tuned neurons tend to prefer stimulus durations similar to the duration of species-specific vocalizations and have preferred durations ranging from 1 ms up to over 100 ms across species.</p> <p>To study the mechanisms underlying DTNs, biologically inspired computational models were produced to explore previously hypothesized mechanisms of duration tuning. These models support the mechanisms by reproducing the responses of <em>in vivo</em> DTNs and predicting additional <em>in vivo</em> response characteristics. The models demonstrate an inherent flexibility in the mechanisms to extend across a wide range of durations and also reveal subtleties in response profiles that arise from particular model parameters.</p> <p>To quantify the encoding efficiency of <em>in vivo</em> DTNs, information theoretic measures were applied to the responses of 97 DTNs recorded from the auditory midbrain (inferior colliculus) of the big brown bat. Stimulus duration encoding robustness, as measured by stimulus-specific information, tended to align with the stimulus durations that produce the largest responses. In contrast, stimulus durations with the most sensitivity to changes in stimulus duration, as measured with an approximation of the observed Fisher information, tended to be stimulus durations shorter or longer than the duration evoking the largest response. Remarkably, both optimal and non-optional Bayesian decoding methods were able to accurately recover stimulus duration from population responses, including durations that lacked neurons dedicated to best representing that duration. These results suggest that DTNs are excellent at encoding stimulus duration, a feature that has been generally assumed but not previously quantified.</p>