Human statistical learning dynamically shapes the hippocampal processing of temporal associations

Abstract

AbstractRecent research suggests that the human hippocampus becomes sensitive to temporal associative patterns during statistical learning (SL). And yet, how such sensitivity develops throughout the course of learning remains unclear. We investigated this question in human adults performing a visual SL task, in which the transition probability among stimuli varied systematically across learning sessions. Hippocampal activity acutely reflected this cross-session change and was predictive of behavioral performance. Using a hidden Markov model to decode stimulus patterns from hippocampal activity, we observed a “mere-exposure” effect, such that brief experience with the temporally structured stimuli was sufficient to alter the hippocampal processing. As learning unfolded, feedback from the repeated exposure shaped the hippocampal processing to better accommodate the environmental statistics. Our findings provide direct evidence of how experience with associative patterns sculpts hippocampal activity in a nonlinear fashion, bringing new insight to its active role in SL.