Experience without recall does not give way steeply to no-experience reports as slow-wave activity rises within a person: results of a preregistered reanalysis of the DREAM database

After an awakening from NREM sleep, a sleeper sometimes reports having experienced something without being able to say what: experience without recall (EWR). A preregistered test asked whether EWR gives way to reports of no experience (NE) as sleep deepens within a person, with relative slow-wave activity (SWA) in the 30 s before each awakening as the depth index. A floor account, on which experience itself fades with depth, predicts a steep decline of EWR against NE; a consolidation account, on which depth impairs the recording of experience more than the experience, allows at most a gentle one. The registered decision rule placed a margin of −0.6 logit per standard deviation on that slope. The analysis code was deposited before any EEG signal was read. In 756 NREM awakenings from 107 participants in five open datasets of the DREAM database, the slope was +0.03 logit per SD (95 % highest-density interval −0.26 to +0.33), entirely above the margin, so the registered verdict is that EWR survives depth. All thirteen registered robustness analyses agreed, among them leaving out the dataset that supplies most N3 awakenings, and adding time of night left the slope unchanged. Within-person SWA moved none of the three report categories credibly, including the loss of content that both accounts predict, which limits the result to excluding a steep within-person decline. In exploratory deciles of SWA standardised within each dataset, which mix depth with differences between participants and stages, reports thinned at the highest SWA and EWR fell from 0.47 to 0.26 of EWR-or-NE awakenings, the direction a floor account expects. The derived tables, model outputs and code are open. Files: Gruber-AIW192-DREAM-results.pdf, the results report; Gruber-AIW192-DREAM-preregistration.pdf, the registered text as filed on OSF Registries (https://osf.io/m4vju), still under embargo at the time of this deposit; aiw192-results-2f30cc2b.zip, the derived per-awakening tables, the model inputs and outputs of all fifteen analyses, the run logs and the script that draws Figure 1 (README inside). Analysis code, deposited before any EEG signal was read: https://doi.org/10.5281/zenodo.24060863