Neural Representation of Vocalizations in Noise in the Primary Auditory Cortex of Marmoset Monkeys
Date of Award
Doctor of Philosophy (PhD)
Robust auditory perception plays a pivotal function in processing behaviorally relevant sounds, particularly when there are auditory distractions from the environment. The neuronal coding enabling this ability, however, is still not well understood. In this study we recorded single-unit activity from the primary auditory cortex of alert common marmoset monkeys (Callithrix jacchus) while delivering conspecific vocalizations degraded by two different background noises: broadband white noise (WGN) and vocalization babble (Babble).
Noise effects on single-unit neural representation of target vocalizations were quantified by measuring the response similarity elicited by natural vocalizations as a function of signal-to-noise ratio (SNR). Four consistent response classes (robust, balanced, insensitive, and brittle) were found under both noise conditions, with an average of about two-thirds of the neurons changing their response class when encountering different noises. These results indicate that the distortion induced by one particular masking background in single-unit responses is not necessarily predictable from that induced by another, which further suggests the low likelihood of a unique group of noise-invariant neurons across different background conditions in the primary auditory cortex. In addition, for a relatively large fraction of neurons, strong synchronized responses can be elicited by white noise alone, countering the conventional wisdom that white noise elicits relatively few temporally aligned spikes in higher auditory regions.
The variable single-unit responses yet consistent population responses imply that the primate primary auditory cortex performs scene analysis predominately at the population level. Next, by pooling all single units together, pseudo-population analysis was implemented to gain more insight on how individual neurons work together to encode and discriminate vocalizations at various intensities and SNR levels. Population response variability with respect to time was found to synchronize well with the stimulus-driven firing rate of vocalizations at multiple intensities in a negative way. A much weaker trend was observed for vocalizations in noise. By applying dimensionality reduction techniques to the pooled single neuron responses, we were able to visualize the dynamics of neural ensemble responses to vocalizations in noise as trajectories in low-dimensional space. The resulting trajectories showed a clear separation between neural responses to vocalizations and WGN, while trajectories of neural responses to vocalization and Babble were much closer to each other together. Discrimination of neural populations evaluated by neural response classifiers revealed that a finer optimal temporal resolution and longer time scale of temporal dynamics were needed for vocalizations in noise than vocalizations at multiple different intensities. Last, among the whole population, a subpopulation of neurons yielded optimal discrimination performance.
Together, for different background noises, the results in this dissertation provide evidence for heterogeneous responses on the individual neuron level, and for consistent response properties on the population level.
Dennis L. Barbour
ShiNung Ching, Daniel W. Moran, Baranidharan Raman, Mitchell S. Sommers
Permanent URL: https://doi.org/10.7936/K7HQ3X9G