POSTER DETAILS
Neural manifolds in human superior temporal gyrus scaffold speech phonological processing
Pavo Orepic, Wilson Truccolo, Anne-Lise M. Giraud, TimothΓ©e Proix
Date / Location: Sunday, 10 July 2022 / S01-183
Add this poster event to your π
Add this poster event to your π
Recent work in early sensory and motor cortices has shown that functionally-relevant population activity can be represented as a set of trajectories confined to a low-dimensional neural manifold, a subspace spanned by the patterns of correlated neuronal activity. However, it remains unknown whether the same encoding principles apply to speech comprehension, a multi-level hierarchical neural processing that integrates several cognitive functions. We analyzed population activity recorded with a microelectrode Utah array implanted in the anterior superior temporal gyrus (aSTG) of a patient with pharmacoresistant epilepsy, and sought after a speech-associated neural manifold. We applied dimensionality reduction methods to the single and multiunit activity collected while the patient performed several auditory tasks encompassing a rich repertoire of stimuli. Our preliminary results show that distinct linguistic units (e.g., words, syllables, and phonemes) could be represented as distinct trajectories on a low-dimensional neural manifold. Moreover, our analysis suggests that differences in the dynamics of the trajectories spanned on the aSTG manifold preferably account for the variability present in the phonetic features of the stimuli. These findings extend previous work on single-neuron to neuronal population response to auditory stimuli in the aSTG, and shed new light on neuronal dynamics during speech processing by uncovering phoneme-specific trajectories confined to a speech-associated neural manifold.