Ego dissolution as a collapse of integration, not a surge of entropy

AbstractThe “entropic brain” hypothesis holds that psychedelic and other ego-dissolving states correspondto an increase in the entropy or signal diversity of spontaneous brain activity. Motivated by afirst-person report of a combined serotonergic–dissociative state (LSD/DMT together with nitrousoxide, N2O), we asked a sharper question: when a 5-HT2A agonist and an NMDA antagonist acttogether, is the resulting state dominated by a rise in local signal complexity, by a collapse of longrange functional integration, or by both? We address this in a validated whole-brain dynamicmean-field (DMF) model (reduced Wong–Wang, Deco et al. 2014) implemented in The VirtualBrain, on empirical human structural connectomes, with regional excitation/inhibition balancedby analytic Feedback Inhibition Control (FIC). The two drug classes are represented by separable,biophysically motivated parameters: 5-HT2A agonism as a density-weighted increase in excitatoryresponse gain, and sub-anaesthetic NMDA antagonism as interneuron-preferential disinhibition(reduced excitatory→inhibitory NMDA conductance) plus reduced long-range NMDA coupling.Crucially, because both perturbations shift mean firing rate — and both Lempel–Ziv complexity(LZc) and functional connectivity (FC) co-vary with rate — we rate-match every condition backto the ∼3 Hz physiological set-point before reading out the metrics, isolating mechanism from atrivial rate confound. At matched rate, the combination produces a robust, dose-dependent,topology-general collapse of functional integration (mean ΔFC = −56%±12% SD, negative in 6/6 noise seeds and reproduced on a second 192-region connectome) together with a smallbut statistically reproducible increase in signal complexity (ΔLZc = +0.9%±0.5% SD,positive in 6/6 seeds, 𝑡 ≈ 4). These multi-seed values are the canonical results; single-run figuresquoted in the development log are individual noise realisations. The 5-HT2A entropy increase(+5.4%) is largely cancelled by an N2O-driven hypersynchrony (−4.6%), so the net complexitychange is marginal while the integration collapse is large. We conclude that, in this model, thecombined ego-dissolving state is dissociation-dominated rather than entropy-dominated:the brain loses its capacity to integrate, not its local richness. We discuss this as a refinement ofthe entropic-brain framing for combined serotonergic–dissociative states, and report the analysestransparently including a working-point artefact we explicitly reject.