Investigating Whether Sleep Spindle Topography Reflects the Content of Prior Learning

Abstract

Sleep spindles, 0.5-3 second bursts of 11-15 Hz oscillations occurring during non-rapid eye movement (NREM) sleep, are thought to be an important index of memory reactivation. Prior work suggests that memory reactivation during NREM sleep happens during sleep spindle events, spindle events carry information about the content of learning prior to sleep, and spindle activity may be upregulated in cortical areas engaged during prior learning (Schreiner et al. 2021; Cairney et al. 2018; Petzka et al. 2022). We investigated if we could detect differences in spindle activity across the scalp based on the content of prior learning (learning involving either scenes or objects), and if the topography of the relationship between spindle activity and behavioral memory consolidation performance differed in our learning conditions. Each participant completed two study visits; during one visit, participants learned about animals paired with scenes, and during the other visit they learned about animals paired with objects. Each learning session was followed by an EEG-recorded nap and a post-nap memory test. We hypothesized that if spindles reflect processing of information in cortical areas engaged in learning, then our two learning conditions (animal-scene versus animal-object associations) would result in topographically distinct patterns of spindle activity and those spindles’ relationship to memory consolidation. We contrasted spindle activity (spindle density and sigma power) between conditions at each electrode site, and we tested for significance using cluster-based permutations. No significant clusters were found. We also correlated spindle activity with behavioral memory retention at each electrode and contrasted the spindle-memory relationship in the two learning conditions. Cluster-based permutations showed that power in the sigma band in the scene condition has a strong relationship to behavioral memory performance. This suggests that power in the sigma band may be predictive of memory reprocessing for scenes. We investigated if the observed effects for these two different analyses were occurring in the same electrode sites by correlating the contrast maps. We found a significant negative correlation between the contrast map of z-scored spindle activity and the contrast map of the spindle-memory relationship. This suggests that in the electrode sites that have a strong relationship to behavioral memory performance in one of the learning conditions, those same electrode sites will exhibit greater spindle activity for the opposite learning condition.