Resolving the Deep Sleep Dual Indeterminacy Problem: Context-Dependent Slow-Wave Activity Modeling Predicts Neurobehavioral Fatigue Where Clinical Sleep Modeling Fails
Deep sleep is widely considered to be the most recuperative component of sleep restoration. Accordingly, a positive relationship between naturally occurring deep sleep and function (e.g., cognitive performance) is often assumed. However, this assumption warrants closer examination--particularly given the rise of sleep tracking that emphasizes traditional sleep metrics and their implied predictive value. We present evidence that while clinical deep sleep scoring provides no predictive value, slow-wave activity (SWA) exhibits a paradoxical association with both improved and worsened neurobehavioral fatigue following sleep deprivation. Specifically, we found that SWA-based models account for approximately 50-60% of the inter-individual variance in recovery from sleep deprivation. Remarkably, when regressed against recovery from sleep deprivation, SWA during the baseline sleep night showed a negative association (normalized {beta} = (-)0.5, p = 0.001) while in the same model SWA during the subsequent wakefulness period showed an opposite positive association (normalized {beta} = 0.5, p = 0.001). Furthermore, although the group-averaged SWA while behaviorally awake increased with impairment across the sleep deprivation period, individual-level data revealed an inverse relationship: individuals more resilient to sleep deprivation exhibited greater SWA in-between mental test sessions and less corresponding impairment during wakefulness suggestive of a protective effect. These findings identify a Deep Sleep Dual Indeterminacy Problem -- simultaneous measurement and causal indeterminacy -- that explains why clinical sleep staging fails as a functional biomarker across a wide range of outcomes, and provide a principled framework for next-generation sleep metrics grounded in continuous electrophysiology and temporal modeling.