Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Investigating hippocampal predictive processing in speech-based statistical learning

Poster Session D - Monday, March 9, 2026, 8:00 – 10:00 am PDT, Fairview/Kitsilano Ballrooms

Daniela Herrera Chaves¹, Emily Cordeiro¹, Arun Thurairajah¹, Greydon Gilmore¹, Lyle Muller¹, Ana Suller-Marti¹, Seyed M. Mirsattari², Laura Batterink*¹, Stefan Köhler*¹; ¹Western University, London, ON, ²Mayo Clinic, Jacksonville, FL

Statistical learning (SL) is a fundamental learning mechanism that supports the extraction of regularities from the environment to support many different aspects of perception and cognition. The hippocampus has been proposed to play a central role in SL, but current evidence regarding its involvement is mixed. While neuroimaging evidence suggests that the hippocampus is involved in SL, prior intracranial EEG data do not directly support a role for the hippocampus. To reconcile these conflicting findings, we hypothesized that the hippocampus does not directly support learning but rather retrieves already learned regularities to generate predictions of upcoming stimuli. In the current study, 21 patients with drug-resistant epilepsy implanted with intracranial depth electrodes listened to a continuous speech stream containing repeating “hidden” trisyllabic words. In the following test phase, participants made speeded responses to specific target syllables embedded within shorter snippets of the speech stream while their brain activity was recorded. At the behavioral level, we found faster reaction times to syllables in later, more predictable positions within a word, compared to word-initial syllables. At the neural level, preliminary event-related potential (ERP) analyses revealed a subset of electrode contacts that showed an ERP effect prior to the onset of a target syllable, relative to non-target syllables, suggesting task-relevant predictive processing. These electrode contacts were mainly located within the hippocampus, insula, and middle temporal cortex. These results suggest that the hippocampus is part of a network of brain regions that supports the prediction of upcoming syllables as a consequence of statistical learning.

Topic Area: LANGUAGE: Other