Effects of classic psychedelic drugs on turbulent signatures in brain dynamics

Previous research has suggested that classic psychedelic drugs can produce therapeutic effects for conditions such as treatment-resistant depression and addiction, but the neural mechanisms underlying these effects remain incompletely understood. A prominent mechanistic hypothesis — articulated in frameworks such as REBUS (RElaxed Beliefs Under pSychedelics) and the entropic/anarchic brain account — proposes that psychedelics relax the brain's hierarchical processing by reducing the precision of high-level priors, thereby permitting increased bottom-up information flow. The authors frame this study in terms of that hypothesis and note that 5-HT2A receptor agonism, a common pharmacological action of classic psychedelics, plausibly alters large-scale neuronal synchrony and canonical rhythms in ways that could underpin such hierarchical relaxation and associated subjective phenomena (for example, ego dissolution).

Cruzat and colleagues analysed two independent, placebo-controlled fMRI datasets in healthy volunteers to characterise how LSD and psilocybin modulate the brain's functional hierarchy. The LSD dataset comprised sessions from 20 recruited participants, of whom 15 remained for analysis after exclusions for motion or other issues; LSD (75 μg i.v. bolus in 10 mL saline) and saline placebo were delivered in a balanced within-subjects design, with resting-state scans performed before and after a music block and eyes-closed instructions. The psilocybin dataset began with 15 participants and, after motion exclusion, included 9 participants; each subject underwent two 12-min eyes-closed resting-state scans on separate days and received either psilocybin (2 mg i.v. in 10 mL saline, infused over 60 s) or placebo in a counterbalanced order. Imaging parameters differed modestly between studies (LSD: TR = 2000 ms, 35 slices; psilocybin: TR = 3000 ms, 33 slices). Pre-processing and denoising used the CONN toolbox and SPM12 with standard steps including removal of initial volumes, realignment, slice-timing correction, grey-matter template normalisation and 6 mm smoothing. The analysis combined two complementary approaches inspired by turbulence theory. The model-free approach computed four measures derived from a local Kuramoto order parameter: amplitude turbulence (local synchronisation, Rλ(x,t)), information transfer (spatial decay of time correlations of Rλ as a function of Euclidean distance r), information cascade flow (time-lagged correlation of Rλ at adjacent scales), and the information cascade (average cascade flow across scales). Spatial scales were indexed by λ values from 0.01 (~100 mm) to 0.21 (~5 mm) in steps of 0.03; higher λ indicates shorter spatial distance. Node-level variability of R (standard deviation across time) and Kolmogorov-Smirnov distance (KSD) were used to compare the distributional similarity of node-wise turbulence between psychedelic and placebo states. The Schaefer 1,000-parcel atlas defined nodes and Euclidean parcel distances. The model-based approach constructed whole-brain dynamical models whose local node dynamics followed the normal form of a supercritical Hopf bifurcation (Stuart–Landau). Structural connectivity used tractography-based connectomes from the Human Connectome Project and an exponential distance rule. Models were fitted by minimising the Euclidean distance between empirical and simulated functional connectivity as a function of inter-node distance FC(r), identifying an optimal global coupling G. In silico perturbations were applied by randomly altering each node's bifurcation parameter a within [-0.02, 0]; model sensitivity (susceptibility) was quantified as the change in the mean modulus of the local order parameter between perturbed and unperturbed simulations, averaged across nodes and trials, while information encoding capability was defined as the trial-wise standard deviation of that perturbation-induced change. Statistical comparisons used permutation-based paired t-tests (alpha = 0.05) with FDR correction for the model-free measures, Wilcoxon rank-sum tests for susceptibility and information capability, and KSD for distributional comparisons at node level.

Sample sizes after exclusions were 15 participants for LSD and 9 for psilocybin. Using the model-free turbulence measures, both psychedelics produced increases in turbulent signatures compared with placebo, but with drug-specific spatial profiles. LSD increased turbulence across all spatial scales examined, with the largest effects at longer distances (smaller λ values corresponding to long-range scales). Psilocybin showed significant turbulence increases primarily at longer-range scales (reported as λ < 0.06). The authors explored the slope of mean turbulence across λ and observed that both drugs produced a monotonic decrease in that slope at longer spatial scales. Information transfer — the spatial decay slope of time correlations of the local synchronisation — was significantly increased under both psychedelics across all spatial scales, indicating flatter decay and therefore greater long-range information transmission. Information cascade flow (correlation of scale λ with λ − ∆λ at consecutive times) was significantly elevated by LSD at all λ values, whereas psilocybin increased cascade flow primarily at higher spatial scales (i.e. long distances/low λ). Averaged across scales, the information cascade measure was significantly increased under both drugs. For LSD, increases in turbulence, information transfer and information cascade did not correlate with subjective report measures; for psilocybin, only increases in information cascade correlated with ego-dissolution ratings (details reported in supplementary figures according to the text). At the node level, computation of absolute differences in node-level turbulence (example shown at λ = 0.12) and counting of nodes in the upper 15% quantile per resting-state network revealed different network signatures: LSD-related changes were most prominent in the somatomotor (SOM) and dorsal attention network (DAN), while psilocybin-related changes were concentrated in the default mode network (DMN), frontoparietal network (FPN), and ventral network (VEN). In the model-based perturbational analyses, whole-brain Hopf models were fitted separately to each empirical brain state (LSD vs placebo; psilocybin vs placebo) using structural connectivity and FC(r). Systematic in silico perturbations produced two principal effects assessed at λ = 0.18: susceptibility — the model's sensitivity to perturbation — was significantly decreased under both LSD and psilocybin compared with their respective placebos; conversely, information encoding capability — the trial-wise variability in perturbation response — was significantly increased under the psychedelic conditions (reported p < 0.001, two-sided Wilcoxon rank-sum test). The extracted text indicates the susceptibility decrease and information capability increase were consistent across compounds.

Using two independent, placebo-controlled fMRI datasets, the researchers applied turbulence-inspired, model-free and model-based analyses to test the hypothesis that classic psychedelics relax hierarchical processing in the brain. Across methods and compounds, the psychedelic state was characterised by generally increased turbulent dynamics and enhanced information transmission across spatial and temporal scales. The authors interpret these results as consistent with prior observations that psychedelics increase entropy of spontaneous brain activity, broaden the repertoire of connectivity states and enhance global connectivity between high-level networks and the rest of the brain. The turbulence framework is presented as offering mechanistic insight into how information transfers across space and time in whole-brain dynamics; increased turbulence and flatter spatial-decay slopes were taken to indicate enhanced long-range information transfer, which the authors link to recruitment of long-distance cortical connections thought important for conscious processing. Network-level patterns differed by drug: psilocybin affected high-level networks (DMN, FPN, VEN), while LSD produced more pronounced changes in somatomotor and dorsal attention networks, suggesting that different psychedelics perturb hierarchical processing in partly distinct ways. The authors note that changes in setting (for example, music) might interact with turbulence but that this was outside the present study's scope. From the perturbational modelling, the investigators report that whole-brain dynamics under psychedelics are less susceptible to external perturbations yet display greater information encoding capacity. They caution that susceptibility is a measure of a system's ability to be perturbed rather than a direct index of complexity; by contrast, information capability reflects how external inputs are encoded and relates more closely to established complexity metrics. The authors situate their perturbative approach as complementary to observational analyses, arguing that data-constrained whole-brain models permit systematic exploration of mechanistic responses to targeted perturbations. They further relate their findings to previous perturbation work in the LSD state that reported increased variability in perturbational responses and slower recovery to baseline, noting their results advance understanding by focusing on information processing characteristics. Overall, the study concludes that classic psychedelics produce a patterned increase in turbulence-related metrics that may be mechanistically linked to their effects on conscious experience and information processing. The authors present these findings as both enriching the understanding of the psychedelic state and demonstrating the utility of turbulence-inspired metrics for characterising brain function more broadly.