Post-encoding slow-wave amplitude during a daytime nap predicts pattern completion from sparse visual cues

Abstract

Pattern completion refers to the reinstatement of a stored memory representation from partial or degraded cues. In this sense, it enables a form of cue-based generalization: the same memory representation can be retrieved across different, incomplete versions of the original input. Sleep supports hippocampus-dependent memory consolidation and may facilitate such flexible retrieval, but it remains unclear whether post-encoding sleep improves visual pattern completion from degraded cues. Previous sleep studies have mainly examined mnemonic discrimination or relational memory, leaving open whether sleep directly enhances the recovery of learned visual scenes from sparse perceptual information. We tested this question using the Memory Image Completion (MIC) task in a polysomnographic within-subject sleep-wake design. Twenty-eight healthy young adults (14 female; mean age 23.4 {+/-} 3.1 years) encoded scene-label associations and were tested immediately and after either a 90-min daytime nap or a matched wake interval. During retrieval, learned and new scenes were presented at five levels of visual completeness. A separate pre-encoding baseline nap assessed individual sleep physiology without prior learning. Sleep did not generally improve performance across all retrieval conditions. Instead, it selectively enhanced consolidation of learned scenes when visual cues were maximally degraded (p = .001) indicating increased pattern completion. No corresponding sleep effect was found for new scenes (all p > .31), suggesting that the benefit was specific to the recovery of previously encoded scene representations. Slow-wave amplitude during the post-encoding nap predicted consolidation of learned scenes under the most degraded condition (p = .011; FDR-corrected p = .042), whereas baseline slow-wave amplitude did not (p > .11). These findings suggest that post-encoding sleep facilitates cue-based recovery of learned visual representations from strongly degraded input, and that this benefit is linked to slow-wave amplitude during post-encoding sleep. Together, these results link sleep-dependent consolidation to visual pattern-completion-like retrieval and extend previous work on sleep-related memory transformation from verbal and relational paradigms to the recovery of learned scene representations from degraded cues.