Episodic events are flexibly encoded in both integrated and separated neural representations

This study investigates how the brain encodes episodic events to support diverse memory functions. Thirty-six participants view movies simulating real-life interactions while EEG was recorded. They first watch movies featuring two characters (AB), followed by scenes where one original character interacts with a new one (BC). Memory is assessed for direct (AB and BC), indirect (AC) associations, and source memory about whether two characters have appeared together. Source memory is more accurate when participants correctly infer indirect associations. Time-resolved representational similarity analysis (RSA) of EEG data reveals both neural pattern similarities and dissimilarities during AB and BC encoding. Pattern similarities, indicating memory integration across episodes, predict successful indirect associations. Conversely, dissimilarities predict accurate source memory, indicating the preservation of distinct memory traces. Here we show that episodic events are flexibly encoded through both integrated and separated neural representations, enabling the brain to support multiple memory functions depending on task demands.