Memory strength at reactivation, not memory age, governs prediction error-driven updating of naturalistic event...

Prediction error during memory reactivation is posited to be a key driver of memory updating. However, prediction error leads to varied memory outcomes, and it is unclear what specific factors influence memory malleability. Two factors have long been theorized to shape a memory's susceptibility to modification-the age and strength of the memory at reactivation-but their interaction with prediction error has not been directly tested in complex, naturalistic events. In a three-session paradigm using short videos with salient action-outcome narratives, we elicited prediction error by interrupting expected outcome endings during reactivation. Following reactivation, participants viewed semantically similar videos introducing new information that could be incorporated into an updated memory. Memory age was manipulated by varying the delay between encoding and reactivation (1 h, 24 h or 2 weeks) and memory strength was measured with participants' ratings of recollection quality after each video reactivation. We found that prediction error at reactivation selectively increased the intrusion of newly experienced information for weak, but not strong, memories at reactivation. By contrast, the age of a memory at reactivation neither interacted with prediction error nor influenced memory outcomes. These findings identify memory strength, not age, as the critical boundary condition for prediction error-driven memory updating in human episodic memory. This article is part of the theme issue 'The role of hippocampal predictions in cognition: bridging perception and memory'.