NB-CNS Seminar - Edward Chang

"The Cortical Machinery for Words"

Spoken language is one of the most distinctive capacities of the human brain, yet how neural populations transform fleeting patterns of sound into recognizable words has remained a central open question in neuroscience. In this keynote, I will present a synthesis of work from our laboratory using high-density direct cortical recordings in humans to reveal how the superior temporal gyrus (STG) and surrounding auditory cortex give rise to the perception of speech.

I will first describe how local populations in the STG encode the acoustic-phonetic building blocks of speech, with neurons tuned to specific phonetic features, speaker characteristics, and prosodic contours such as pitch and stress. These representations are not simple acoustic reflections but are organized around linguistically meaningful dimensions, and they emerge within tens of milliseconds of sound onset.

I will then turn to recent discoveries showing that the STG contains representations of whole words as distinct neural objects—encoded separately from, and in parallel with, their constituent phonetic features. Critically, these are representations of word form: the stored phonological patterns that allow a listener to recognize a familiar spoken word as a unit, independent of its meaning. These word-form representations are sensitive to lexical identity, frequency, and phonological neighborhood structure, and they reveal a functional specialization within the STG that had not previously been appreciated. Together, these findings suggest that the STG is not merely an acoustic-phonetic analyzer but a key site where continuous speech is parsed into the lexical sound patterns that serve as the entry points to the mental lexicon.

Host: Markus Meister