Retrieval-mediated and sleep-based memory consolidation provoke different neural and behavioural markers of memory generalisation.

Across sleep-based and retrieval-mediated consolidation, memories typically become generalised and less dependent on their episodic components for their recollection. However, memory transformations across sleep and retrieval training have not been directly compared. The current study aims to compare how sleep and retrieval training impact the endorsement of semantically similar and different lures, as well as their episodic recollection using the parietal old/new effect on the late positive component (LPC) in subjects' EEG. Thirty subjects (27F, 18–34, M=22.17) attended four sessions where they learnt different sets of 104 object-word pairs and completed one of four 120-minute memory interventions: retrieval training (i.e., cued recall practice), restudy (i.e., pair re-exposure), a nap opportunity, or a wakeful rest. EEG was recorded while subjects were tested on their recognition accuracy in an old/new paradigm with similar- and different-object lures. Our results revealed that retrieval training, but not sleep, lead to greater accuracy for identifying old pairs, but worse similar-lure discrimination. Whilst the parietal old/new effect did not differ between conditions, retrieval had lower LPC amplitudes for similar- than different-object false alarms, whilst restudy demonstrated the opposite. Sleep and wake demonstrated no differences in LPC amplitudes between hits and different false alarm types. Together, our study demonstrates evidence for gist-abstraction across retrieval training, and a task-relevant selective maintenance of episodic details across sleep. These results challenge theories that retrieval training replicates sleep-based consolidation mechanisms, instead acting as a fast route to semanticization regardless of the context.