Change-Point Structure in Longitudinal EEG: A Two-Channel Decomposition of Consciousness Regime Transitions

Transitions in consciousness occur across diverse conditions including anesthesia, seizure evolution, and sepsis-associated encephalopathy. Existing models typically treat these transitions as scalar reductions in global integration or network efficiency. This paper proposes a dynamical decomposition in which state transitions reflect changes in two separable latent channels: (1) local constraint-resilience dynamics, observable through perturbation-recovery and spectral structure, and (2) global realizability capacity, observable through EEG reactivity and effective propagation. The model predicts that these channels can degrade independently and with distinct temporal ordering. Specifically, some transitions (e.g., pharmacologic anesthesia) should show preserved local perturbation geometry with reduced global propagation, whereas others (e.g., pre-ictal or inflammatory ramp states) should show early degradation of local dynamical resilience preceding global collapse. The framework further predicts that the informational content of EEG spectral properties shifts across ramp progression—from reflecting local constraint geometry early to reflecting substrate capacity later. The paper formalizes this decomposition within a constraint-manifold model, defines measurable change-point criteria for each channel, and specifies falsifiable lead–lag predictions testable in existing longitudinal EEG datasets. The central claim is not metaphysical but dynamical: consciousness regime shifts correspond to boundary crossings in a two-channel viability space whose components can be empirically dissociated.