Causal network structure predicts memory organization and neural reinstatement across events
Endel Tulving proposed that episodic memory is organized by spatiotemporal relationships, and current theories center the role of time and space in organizing memory. However, these theories are supported by paradigms that eliminate other forms of relationships, like causality. Here, we used fMRI to reveal the organization of neural representations that support episodic memory with a paradigm that decoupled time and causal relationships. Participants were scanned while watching and recalling a TV show featuring five temporally interleaved storylines. Behaviorally, the order of recalled events was best predicted by the strongest cause-effect relationship relative to temporal, spatial, or semantic associations. Neural similarity across different events in regions of the default mode network (DMN) reflected causal structure, and DMN patterns at event boundaries specifically reactivated prior, causally related events, suggesting causality allows us to stitch together related events across temporal gaps. Additionally, causal network distance between successive events predicted the strength of DMN pattern changes across event boundaries, suggesting causality also predicts stronger representational switching between unrelated events. Together, these findings suggest that the DMN supports the construction of the complex causal networks that play a central role in organizing how we understand and remember events.