Predictive Processing and Schema Effects in Episodic Memory

Schemas are associative knowledge structures that represent statistical regularities across multiple episodes. Schemas are known to enhance memory retention, with better memory observed for events that are either highly congruent or highly incongruent with our existing schemas. These schema effects are thought to arise due to predictive processing, whereby prior knowledge is used to make predictions about how an event will be structured. The subsequent affirmation or violation of these predictions is believed to potentiate memory encoding via prefrontal-hippocampal interactions. This thesis investigates whether schema effects extend to situations in which a core feature of the schema is the chronology of events, allowing for temporal predictions of the future. Participants were pre-trained on categorical sequences where schemas defined the order of semantic categories. Subsequently, they studied new sequences consisting of novel exemplars. Crucially, these sequences could either be congruent, incongruent, or unrelated to the pre-learned schemas. At test, participants were cued with an item from a sequence and required to recall the following two items in a four-alternative forced-choice task. Schema congruency effects were observed in all experiments. However, incongruency benefits were only observed for categorical violations, and not for violations of sequential order. MEG analyses revealed that schema violations were associated with modulations of theta-band activity, and that category, sequence position, and uncertainty signals could all be decoded from pre-stimulus neural activity. Overall, these findings are in line with predictive accounts of schema effects and refine our understanding of when memory advantages arise for schematically incongruent sequences.