Prediction in Human Auditory Cortex

Spoken language is thought to be facilitated by an ensemble of predictive mechanisms, yet the neurobiology of prediction for both speech perception and production remains unknown. We used intracranial recordings (31 patients, 6580 electrodes) from depth probes implanted along the anteroposterior extent of the supratemporal plane during rhythm listening, speech perception, and speech production. This revealed a frequency-multiplexed encoding of sublexical features during entrainment and a traveling wave of high-frequency activity across Heschl’s gyrus. Critically, we isolated two predictive mechanisms in early auditory cortex with distinct anatomical and functional characteristics. The first mechanism, localized to bilateral Heschl’s gyrus and indexed by low-frequency phase, predicts the timing of acoustic events (“when”). The second mechanism, localized to planum temporale in the language-dominant hemisphere and indexed by gamma power, predicts the acoustic consequence of speech motor plans (“what”). This work grounds cognitive models of speech perception and production in human neurobiology, illuminating the fundamental acoustic infrastructure – both architecture and function – for spoken language.