Low-dimensional neuronal population dynamics in anterior superior temporal gyrus reactivate phonetic representations during semantic processing.

Traditional models of speech perception posit that neural activity sequentially encodes speech through a hierarchy of cognitive processes, from early representations of acoustic and phonetic features to late semantic encoding. Yet the mechanisms by which neural representations are transformed across the speech hierarchy remain poorly specified. Here, we analyzed unique microelectrode array recordings of neuronal spiking activity from the human left anterior superior temporal gyrus, a brain region at the interface between phonetic and semantic auditory processing, during a semantic categorization task and natural speech perception. In both conditions, low-dimensional neuronal population dynamics revealed a distributed and parallel encoding of phonetic and semantic representations. During natural speech, the low-dimensional dynamics were simultaneous to a power increase in the beta and low-gamma local field potentials, reflecting concurrent instantiation of top-down predictive and bottom-up cumulative processes. Our results support a mechanism for phonetic-to-semantic transformations encoded at the neuronal population level.