This manuscript introduces a computational and falsifiable framework for evaluating whether time-varying geomagnetic activity can influence neural state dynamics during sleep. The approach centers on a latent controllability metric, Γ(t), defined from EEG/EOG/EMG-derived state transitions as a stability-weighted measure of directional sensitivity. Rather than asserting an effect, the work formalizes the conditions under which such a coupling could be meaningfully detected. A simulation-based “RealStorm” perturbation model is used to establish expected signal structure, followed by stress testing and adversarial validation against a hierarchy of null models, including temporally shifted, phase-randomized, shuffled, and synthetic storm-like predictors. Results demonstrate that autocorrelation in geomagnetic signals can produce spurious alignment with neural dynamics, allowing shifted predictors to remain competitive. To address this, the manuscript defines strict acceptance criteria requiring that real geomagnetic predictors outperform all adversarial controls, exhibit REM-specific modulation, and replicate across independent observations. This study does not provide empirical evidence for geomagnetic–neural coupling. Instead, it establishes a rigorous, reproducible testing framework designed to distinguish genuine weak external perturbation effects from autocorrelation-driven false positives. The framework is broadly applicable to neural time-series analysis and to the study of weakly coupled dynamical systems.
Joan Kaliff (Fri,) studied this question.