Replay Of Abstract Temporal Structure In Episodic Memory

Temporal information is a core component of episodic memories and plays a critical role in both encoding and retrieval of memories.However,what we know about how it is involved in those processes remains little.Therefore,in this study,we aimed to investigate how temporal information evolves during the encoding and retrieval of episodic memories,using neural decoding technique based on intracranial EEG(i EEG)signals from epileptic patients.In particular,we described our research question as: 1)whether there is a replay of temporal structure during online memory retrieval and offline post-retrieval rest,2)whether online and offline replay differ in their underlying mechanism,and 3)whether replay happens on multiple time scales.To obtain the neural representation of temporal information,we firstly trained a machine learning model to decode the temporal states from encoding of single-exposure video materials.By analyzing the performance of temporal state decoders,we further showed that this neural representation of abstract temporal structure presented a continuous trend,gradually changed,during encoding and generalized to new,untrained videos.Then we found replay of the temporal structure during both memory retrieval and post-retrieval resting,with replay in the retrieval phase predicted the performance of temporal order judgement task and replay in the rest phase reflected the confidence ratings,but not vice versa.Finally,we found replay with larger compression factors within sharp-wave ripples(SWRs)in hippocampus,indicating the existence of replay on multiple time scales.In all,our study evidenced both online and offline replay of abstract temporal structure in human episodic memory with potential different underlying mechanisms and further indicated the existence of multiple time-scaled replays happened withinand outside-SWRs.