Extended Data Fig. 3: Slow wave, sharp wave, ripple and dentate spike properties by experiment type.

a, Same data as Fig. 2d and Supplementary Table 1, separated by experiment. In this and the following panels, bars above plots indicate general linear hypothesis tests on mixed effects models (n = 29 sessions of 3 kinds from 16 animals). Black bars reflect post hoc comparisons that were significant. Black bars with triangular endcaps indicate the presence of a significant interaction between experiment and condition, assessed using an asymptotic likelihood ratio test. Black bars with circular endcaps reflect the absence of a significant interaction between experiment and condition, but a significant main effect of condition. All tests were two-sided and corrected for multiple comparisons (see Supplementary Table 1 and Methods for details). Note that in the dual experiment, the time elapsed between early and late extended wake is longer than for novelty and locomotion, and the time elapsed between early and late recovery sleep is shorter. b, Same data as Fig. 2e, separated by experiment. c, Same data as Fig. 2f, separated by experiment. d, Same data as Fig. 3e, separated by experiment. e, Cortical power in the 2–6 Hz range during extended wake, replicating²⁰. Power was computed identically to SWA, but for the frequencies of interest (2–6 Hz). As for SWA, a linear mixed effects model was created using all six conditions of interest (early and late baseline NREM sleep, early and late extended wake, early and late recovery NREM sleep) and tested for an interaction of condition with the 2–6 Hz power. After finding a significant interaction (p < 1 × 10⁻⁷; f² = 0.447), we performed post hoc tests for a difference between early and late extended wake, which were significant for novelty (p = 0.017; d = 0.873; n = 12 subjects) and locomotion (p = 0.017; d = 1.23; n = 6 subjects), but not dual (p = 0.182; d = 0.755; n = 5 subjects). Boxplots show medians and quartiles. Whiskers are drawn to the farthest datapoint within 1.5 IQRs from the nearest hinge. f, Neocortical single unit mean firing rates in early and late baseline NREM sleep (ON periods; blue), early and late extended wake (green), and early and late recovery NREM sleep (red), for putative pyramidal cells (n = 721), narrow interneurons (n = 83), and wide interneurons (n = 55). All data are from novelty experiments, and only cells that could be tracked continuously for the full 48 h were analyzed. Bars above plots indicate post hoc tests performed on mixed effects models. Examination of residual plots revealed that model assumptions of constant variance and normality were better satisfied by square-root transformation of firing rates than log transformation, so the former was used for statistical analyses. The full model was formulated with condition as primary covariate and a nested random effects structure, with unique cell ID nested inside subject. A main effect of condition was found in every case, followed by post hoc tests for comparisons between conditions. The somewhat paradoxical increase in pyramidal cell firing rates from early to late recovery sleep may be explained by the relatively short elapsed duration between the two conditions (relative to early and late baseline sleep), since late recovery sleep was constrained to fall within the light cycle, to account for the known effects of immediate light exposure on cortical SWA (Methods). During this period, NREM sleep is prioritized over REM sleep, the latter of which may be associated with the largest decreases in firing rate²⁴. g, As in f, but for hippocampal single units (putative pyramidal cells, n = 27; narrow interneurons, n = 57; wide interneurons, n = 16). See Methods (spike sorting) for details. In f,g, boxen plots show the median and quartiles, and a number of additional quantiles proportional to the size of the data: In both f and g, additional boxes show the 12.5% and 87.5% quantiles. In f only, outer boxes show the 6.25% and 93.75% quantiles.