Yawning reveals energy-dynamics mismatch and neural inertia across state transitions

Yawning is a highly conserved behavior, yet its neural dynamics across arousal state transitions remain poorly understood. Classical reflex models fail to account for the precise timing of yawns during large-scale changes in brain network organization. Here, we investigated the neural signatures of yawning using simultaneous electroencephalography (EEG), electromyography (EMG), and kinematic recordings in a beagle model during wakefulness and propofol anesthesia. Propofol-associated yawning exhibited marked temporal asymmetry, with 89.2% of events occurring during emergence from anesthesia. To resolve the fast neural dynamics surrounding these events, we applied empirical mode decomposition coupled with the Hilbert–Huang transform (EMD-HHT) analyses. Across induction, recovery, and wakefulness, yawning was consistently associated with a transient decoupling between local and global network dynamics: a brief increase in local γ-band power coincided with a drop in network flexibility, quantified by instantaneous frequency volatility; IFV). We termed this reproducible motif as “Energy-Dynamics Mismatch” (EDM). A machine learning classifier leveraging EDM features reliably decoded the brain state immediately preceding yawning. We further validated the presence of this signature in an independent human cohort during anesthesia induction. These findings indicate that yawning is not a passive reflex, but is tightly linked to moments of network instability during state transitions. From a computational perspective, EDG may reflect a transient “control effort” that preserves information integration in the presence of “neural inertia.” This dynamic signature may provide a quantitative biomarker for monitoring arousal transitions brain-computer interface (BCI) interface state-alerting.