The Asymmetrical Architecture of Ultradian REM Sleep: Direction, NREM Currency, and a Finite Post-REM Predictive Horizon in Mice
Year: 2026
Type: preprint
Abstract: Two accounts published in 1994 placed the controlling variable of the ultradian NREM-REM cycle on opposite sides of the REM episode: propensity accumulating during NREM sleep (Benington and Heller, 1994), or a constraint imposed by the preceding episode (Vivaldi et al., 1994). The Asymmetrical Hypothesis (Le Bon, 2021) proposed a third position, in which the preceding episode determines only a bounded post-REM refractory period rather than the whole interval. I tested these accounts in 1,955 valid inter-REM intervals reconstructed from continuous 24-hour recordings in 25 mice of three strains. Accumulated NREM did not predict the duration of the following REM episode (Spearman rho = 0.015, p = 0.50), whereas preceding REM duration predicted subsequent NREM accumulation (rho = 0.506) in every animal considered individually; a mixed-effects model fitted on the log-log scale returned an elasticity of 0.56 ± 0.03, steeper than previously reported log-log slopes near 0.33. A lower envelope near 1 min of accumulated NREM per minute of preceding REM survived leave-one-mouse-out validation and within-mouse permutation (6.1% of held-out intervals crossed it, against 10.2% under permutation), but was convex rather than proportional to the preceding episode. The influence of the preceding episode became undetectable beyond approximately 10 min of accumulated NREM, which outperformed elapsed chronological time as the relevant clock in held-out mice. Episode-termination hazard rose with elapsed episode duration (Weibull shape 1.48, versus 0.84 for inter-REM intervals), and inter-animal variability in episode number exceeded that in mean episode duration approximately fourfold. Part of the non-monotonic, rise-and-fall shape reported in recent propensity analyses arises from survivor selection rather than from any additional mechanism: intervals still at risk late are enriched in long preceding episodes, and a mixture of exponential hazards with no individual decline reproduces the same pooled pattern. Together, these findings indicate that the early post-REM constraint is real but neither strictly bounded nor proportional, expiring within about ten minutes rather than governing the whole interval; because episode duration is itself internally constrained, 24-hour REM output is regulated through episode frequency rather than duration.
Author: Olivier Le Bon
Institution: Université Libre de Bruxelles
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7213105443