Multiple memories can be simultaneously reactivated during sleep as effectively as a single memory

Memory consolidation, the process by which memories become stable and long-lasting, is significantly enhanced by sleep. A prominent hypothesis, known as active systems consolidation, proposes that memories initially stored in the hippocampus are reactivated during sleep, strengthening neocortical representations. While reactivation of individual memories during sleep has been demonstrated in both humans and animals, the question of how the brain manages the simultaneous reactivation of multiple memories remains largely unexplored. This study addresses this gap by examining whether reactivating multiple memories concurrently impacts the sleep-related benefits of memory consolidation. Previous studies have primarily focused on reactivating single memories, using techniques such as targeted memory reactivation (TMR), where auditory or olfactory cues presented during sleep selectively reactivate specific memories. However, the question of whether the brain can efficiently manage multiple, simultaneous reactivation processes without incurring a cost to memory consolidation remains open. This research directly compares the effects of TMR on single versus multiple memory reactivation, testing whether parallel processing depletes resources critical for effective memory consolidation. The authors hypothesize that the benefit of TMR might be dependent on the number of memories reactivated simultaneously, suggesting a limited capacity for parallel reactivation. Alternatively, they propose that multiple memories can be reactivated in parallel without compromising the benefits for each individual memory.

The literature review summarizes existing research on memory consolidation and reactivation during sleep. It highlights the active systems consolidation hypothesis and cites previous studies demonstrating memory reactivation in rodents and humans using techniques like fMRI and MEG. The review emphasizes the lack of research directly comparing single versus multiple memory reactivation during sleep and the need to understand potential resource limitations and interference effects. The researchers point to studies using TMR with odor or auditory cues, noting that while these studies showed benefits for cued memories, they did not directly compare single versus multiple item reactivation. The absence of this direct comparison motivates the current study's design.

Participants learned the locations of 90 images on a circular grid. Images were grouped into sets of one, two, or six items (e.g., six different cat images). During an afternoon nap, sounds associated with half the sets were unobtrusively presented during NREM sleep stages (N2 and N3). Spatial memory was assessed before and after sleep. The experiment used a within-subjects design, allowing researchers to compare the effects of single versus multiple item reactivation within the same participants. EEG data were collected during sleep to assess sleep spindles and delta-theta power modulations following cue presentation. Statistical analyses included repeated-measures ANOVAs, paired t-tests, correlations, and linear mixed models to examine the effects of set size and cueing on memory performance and neural activity. To address potential confounds related to pre-sleep accuracy differences between set sizes, two complementary analyses were conducted: one using a regression-based approach to adjust for pre-sleep differences, and another using bootstrapping to create subsamples with balanced pre-sleep accuracy. Swap errors (mistaking one item's location for another within the same set) were also carefully considered and removed from the primary accuracy analyses to prevent bias in memory performance assessments. EEG data were analyzed to examine sleep spindles and delta-theta power modulations following cue presentation, specifically looking at the effects of set size on these neural measures. Advanced signal processing techniques were utilized for data cleaning and analysis of EEG data.

The study's key findings demonstrate that the benefits of TMR on memory are not influenced by the number of items within a cued set. Participants showed significantly better recall for cued items compared to non-cued items, regardless of whether the cued set contained one, two, or six items. This suggests that the brain can reactivate multiple memories simultaneously without diminishing the benefits for each individual memory. There was no significant interaction between set size and cueing status on memory performance. Analyses examining cumulative benefits for entire sets showed a marginal trend toward larger benefits for larger sets, but this effect was weaker than the effect of cuing. This is noteworthy as it suggests that while memory improvement is not limited by the number of items, larger sets still show larger improvements in total, suggesting a cumulative effect. Further analyses testing alternative hypotheses that assume only a subset of items within a set receive the reactivation benefit were not supported by the data. Regarding EEG analyses, power in both the delta-theta and sigma frequency bands, associated with sleep spindles, increased following cue presentation, and this increase was directly related to set size. Specifically, larger sets showed larger increases in power in these bands. This increase was primarily driven by changes in spindle probability, with the number of spindles being positively modulated by set size, rather than by changes in spindle amplitude.

The findings challenge models of sleep-dependent memory consolidation that posit a limited capacity for reactivation and suggest a capacity for parallel processing of multiple memories during sleep. The results support a parallel reactivation hypothesis (PRH), where multiple memories are reactivated independently and simultaneously. An alternative, but less probable interpretation, is that reactivation occurs in rapid succession. While the authors acknowledge the possibility of rapid sequential reactivation, the data better support the PRH because it doesn't require unrealistically rapid reactivation cycles to account for the observed lack of set size effects on memory benefits. The study also suggests that TMR may either directly activate multiple specific memory traces or it may reactivate generalized contextual representations that subsequently benefit the embedded individual memories. The relationship between set size-dependent increases in delta-theta and sigma band activity and the set size-independent memory benefits is discussed. The findings suggest that while the reactivation benefit to each item is not limited by set size, the overall benefit to the set may show some size dependency, and the neural correlates (delta-theta and sigma power) clearly show set-size dependence. These neural findings are interpreted in the context of existing research on sleep spindles, slow waves, and K-complexes. The results highlight the complexity of sleep-dependent memory consolidation and the need for further research.

This study provides strong evidence that the brain can simultaneously reactivate multiple memories during sleep without compromising the benefits of reactivation for each individual memory. This challenges the notion of a limited capacity for reactivation and highlights the potential for parallel processing during memory consolidation. Further research should investigate the upper limits of reactivation capacity, the role of contextual representations in reactivation, and individual differences in the capacity for parallel processing during sleep.

The study's limitations include the use of a relatively small sample size of university students. The generalizability of findings to other populations may be limited. The researchers acknowledge that the choice of stimulus categories (highly associated items with the same sound) might influence the results and that differences in relatedness between sets could introduce noise. The relatively small set sizes used might not reveal potential interference effects at higher memory loads. Further research is needed to investigate potential confounds related to stimulus category, inter-set relatedness, and the influence of larger set sizes on simultaneous reactivation and consolidation.