Fig. 1: Experimental paradigm and analytic scheme of memory representational transformation across sleep.

A The experimental procedure includes pre-sleep learning (i.e., encoding and maintenance), pre-sleep cued-recall and cued-recognition tests on half of the learned items, overnight sleep with TMR cueing during slow-wave sleep, and post-sleep tests, including cued-recall, mental retrieval with a written report and cued-recognition. EEG was recorded throughout the entire experiment. B Within-session RSA examined memory representations within each of the pre-sleep learning and post-sleep mental retrieval sessions; Cross-session RSA examined the memory representational similarity between these two sessions. Both the within-session RSA and cross-session RSA were performed at two different levels: item-level and category-level. Item-level representations were obtained by contrasting the within-item (WI) similarity versus the within-category (WC) similarity, while the category-level representations were obtained by contrasting the WC similarity versus between-category (BC) similarity. Memory representational transformation would be manifested by the following indices: decreased item-level representations, while persistently prominent category-level representations from pre- to post-sleep; and the low cross-session pre-post similarities. C Schematic illustration of the interactive roles of sleep stages and their associations with memory representational strength change. SWS*REM reflects the complementary roles, while REM/SWS reflects differential roles between these two sleep stages. SWS slow-wave sleep, REM rapid eye movement sleep.