Interaction between 1/f spectral structure and nonlinear dynamics constrains brain-state exploration in psychosis

Psychosis has traditionally been studied through isolated neural features such as spectral structure or measures of dynamical complexity, yielding limited mechanistic insight. A central unresolved question is whether observed abnormalities primarily reflect alterations in aperiodic spectral organization or broader disruptions in large-scale neural dynamics. This preprint investigates whether psychosis is associated with a constrained neurodynamic regime emerging from the interaction between aperiodic spectral structure (1/f) and nonlinear temporal dynamics. Using 1, 932 EEG recordings from the publicly available ASZED dataset (153 subjects; 77 controls and 76 patients), the study evaluates spectral, dynamical, and interaction-based models within a falsification-driven framework combining spectral whitening, surrogate-data testing, delay-embedded state-space reconstruction, entropy analysis, effective dimensionality estimation, and subject-level leakage control. Models based solely on aperiodic spectral structure performed near chance (AUC ≈ 0. 58), whereas dynamical features alone achieved moderate discrimination (AUC ≈ 0. 61–0. 70). In contrast, combined spectral–dynamical models consistently achieved higher performance (AUC ≈ 0. 73–0. 80), indicating that neither spectral nor dynamical organization alone is sufficient to account for the observed group differences. Reconstructed state-space manifolds revealed substantially reduced large-scale trajectory exploration in psychosis, including marked reductions in trajectory radius and path length. Importantly, this global contraction coexisted with increased effective dimensionality (dₑff), suggesting altered local occupancy organization within a partially constrained dynamical manifold rather than a simple reduction in variability. Together, these findings support a systems-level interpretation in which psychosis is associated with constrained large-scale neural dynamics that cannot be fully explained by aperiodic spectral structure alone and are more consistent with partial interaction between scale-free spectral organization and nonlinear temporal dynamics. A minimal nonlinear generative model reproduced qualitative features compatible with the empirical observations and provides a phenomenological dynamical framework for interpreting constrained brain-state exploration in psychosis.