Abstract
Humans can implicitly learn about multistep sequential relationships between events in the environment from their statistical co-occurrence. Theoretical work has suggested that neural replay is a candidate mechanism that aids such learning. Here, we used functional MRI (fMRI) to test whether replay is related to implicit learning of higher-order sequential relationships. Human participants viewed sequences of images that followed probabilistic transitions determined by ring-like graph structures. Behavioral modeling of response times revealed that participants acquired multistep transition knowledge in a manner consistent with gradual updating of an internal successor representation (SR) model. Yet, half of participants did not report being aware of any sequential task structure, and most participants failed to provide meaningful transition probability ratings in a posttask test. Analyses of temporal dynamics of multivariate fMRI patterns during brief 10 s pauses from the ongoing statistical learning task indicated backward sequential replay of multistep sequences in visual cortical areas. Variations in model parameters between participants that captured response time patterns related to strength of neural replay. No corresponding relations between replay and measures of explicit awareness were found. These findings indicate that implicit learning of higher-order relationships establishes an internal SR-based map of the task and is accompanied by cortical on-task replay.