REM sleep is considered a ‘hyperassociative’ state enhancing the connections between weakly related concepts, thereby facilitating memory restructuring — a critical process proposed in creative problem-solving. Here, we empirically explore this hypothesis by testing participants on a riddle before and after a 90-minute incubation period including only NREM sleep, REM sleep, or wakefulness. We quantified memory restructuring by computing differences in problem-related semantic memory network (SemNets) properties between pre- and post-incubation. For each sleep stage, we extracted fine-grained EEG measures linked to cognitive richness. REM sleep, compared to wake, broadly facilitated problem-related word associations but also reshaped the dominant problem representation by combining semantically remote concepts in memory or reducing strong but solution-irrelevant associations of ideas. Such REM sleep-related restructuring was associated with EEG measures indexing a richer cognitive state. While REM drove memory restructuring, it alone did not boost problem-solving success, suggesting that additional processes are necessary for creative breakthroughs.
Bieth et al. (2025) studied this question.