Phenomenal Transitions as Dissipative Events: Neural Entropy Production Rate Peaks at Moments of Conscious Reorganisation
Zenodo (CERN European Organization for Nuclear Research) March 29, 2026 preprint DOI: 10.5281/zenodo.19323431 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Observational analysis of three independent EEG datasets |
|---|---|
| Sample size | 58 |
| Population | Human EEG participants across three datasets: rule reversal learning (N=22), polysomnographic sleep (N=7), and binocular rivalry (N=29) |
| Measures | Spatial efficiency η, rate of change of spatial efficiency v_η, Lempel-Ziv complexity, perturbational complexity index |
| Key findings | The authors argue that phenomenal transitions are thermodynamic events marked by elevated entropy production, captured by v_η. They report v_η nearly doubled baseline during reversal learning (d=1.285), |Δη| was elevated at sleep stage boundaries (d=1.355), and η was lower before binocular rivalry switches (d=−0.932), suggesting v_η measures a dimension of phenomenal dynamics distinct from existing level-of-consciousness metrics. |
Abstract
Abstract Existing metrics of phenomenal consciousness — spatial efficiency η, Lempel-Ziv complexity, perturbational complexity index — characterise the level of conscious organisation. Here we propose and empirically validate a complementary metric: the rate of change of spatial efficiency, v_η = |dη/dt|, interpreted as a proxy for neural entropy production rate during phenomenal transitions. Drawing on non-equilibrium thermodynamics, we propose the Dissipative Events Hypothesis: phenomenal transitions — moments of conscious reorganisation — coincide with local maxima of entropy production, not with maxima of organisational level. We test this hypothesis across three independent EEG datasets spanning qualitatively distinct types of phenomenal transition: rule reversal learning (ds004295, N=22), polysomnographic sleep stage changes (ANPHY, N=7), and binocular rivalry perceptual switches (N=29). In reversal learning, v_η during the transition period [-1.5, +1.5s] is nearly double the baseline rate (d=1.285, p<0.001), with peak velocity at t=+1.2s — following, not preceding, the reversal event, consistent with post-transition reorganisation. In sleep, epoch-to-epoch |Δη| is significantly elevated at stage boundaries relative to within-stage continuations (d=1.355, p=0.031), and correlates with the magnitude of phenomenal rank difference between stages (Spearman r=0.107, p<0.001). In binocular rivalry, η is significantly lower during pre-switch epochs than during mid-stable control periods (d=−0.932, p<0.001), indicating that systems approaching a perceptual switch occupy a less organised — more dissipative — state. Together, these results support the view that phenomenal transitions are thermodynamic events characterised by elevated entropy production, and suggest that v_η captures a dimension of phenomenal dynamics orthogonal to existing level-of-consciousness metrics. Keywords: EEG; spatial efficiency; entropy production; phenomenal transitions; non-equilibrium thermodynamics; dissipation; binocular rivalry; sleep staging; reversal learning; Negative Space Encoding