Sleep Preserves, Wake Differentiates: Strength-Dependent Forgetting of Declarative Memories Across the Retention Interval

Abstract

Sleep benefits memory, but whether this benefit depends on initial memory strength remains heavily debated. We asked whether sleep preferentially stabilizes weakly versus strongly encoded declarative memories, and whether classical sleep oscillations contribute to this effect. Ninety healthy adults (n=30/group) learned 80 unrelated word pairs; half presented twice (weak, S-), and half three times (strong, S+). Participants were tested either immediately ([~]40 min post-learning), after 9 hours containing nocturnal sleep monitored with polysomnography, or after a comparable daytime waking period. Memory performance was assessed using cued recall accuracy and reaction times (RT). Strongly encoded items (S+) were recalled better than weakly encoded items (S-) across all groups. Overall memory performance was lowest after wakefulness, whereas performance after sleep was slightly higher than immediate recall. Crucially, wakefulness selectively impaired the consolidation of weak items, with the WAKE group forgetting disproportionately more S- than S+ items. In contrast, the SLEEP group retained weak and strong items similarly, yielding performance comparable to immediate recall. RT analyses showed generally slower responses for weak items, particularly after wakefulness, while response times in the sleep and immediate recall groups did not differ. Analyses of classical sleep features, such as slow oscillations density, sleep spindle intensity, and their coupling, revealed no significant associations with overnight retention. Correlations between these oscillations and memory performance were present both during the adaptation and the learning nights, suggesting trait-like, rather than state-dependent relationships. Overall, our findings indicate that sleep protects declarative memories irrespective of initial encoding strength.