Serotonergic psychedelic drugs LSD and psilocybin reduce the hierarchical differentiation of unimodal and transmodal cortex

Research on serotonergic psychedelics such as LSD and psilocybin has reported substantial changes in whole-brain functional organisation and dynamics, including reduced segregation of large-scale networks, increased global functional connectivity, and greater complexity of regional activity. A recent theoretical account—the RElaxed Beliefs Under Psychedelics (REBUS) model—posits that psychedelics increase the influence of low-level sensory and limbic inputs on high-level representations encoded in transmodal cortex, implying a reduction in hierarchical differentiation between unimodal and transmodal regions. However, whether psychedelics acutely alter macroscale hierarchical organisation has not been directly tested. Girn and colleagues set out to test this idea by applying gradient-mapping techniques to pharmacological resting-state fMRI datasets acquired under LSD and psilocybin. Their primary hypothesis was that both drugs would produce a contraction of the principal cortical connectivity gradient—an axis running from unimodal sensorimotor regions to transmodal association cortex—reflecting reduced hierarchical differentiation and increased unimodal–transmodal crosstalk. Secondary aims included examining drug effects on the second and third gradients, which capture sensory and executive differentiation respectively, and relating gradient changes to self-reported subjective effects such as ego-dissolution and complex imagery. The study therefore investigates whether macroscale connectivity gradients are sensitive to acute serotonergic pharmacological manipulation and whether such changes relate to phenomenology.

The study analysed previously published within-subject resting-state BOLD fMRI datasets for LSD and psilocybin. For LSD, 20 subjects were scanned on two occasions (LSD and placebo) separated by two weeks, with 75 μg LSD or saline administered intravenously; two non-music resting scans (scans 1 and 3) were used. For psilocybin, 15 subjects underwent two scanning sessions (psilocybin and placebo) with a 2 mL psilocybin infusion in 10 mL saline or saline alone; a single 12-minute scan was used and the post-infusion half of each session was analysed. Subjects were blind to session order; experimenters and analysts were not. Ego-dissolution ratings were collected using intra-scanner visual analogue scales and complex imagery was assessed from the 11-factor altered states of consciousness (ASC) questionnaire completed after each day. Preprocessing and denoising followed an identical pipeline for both datasets. Excessive motion (>15% volumes with FD ≥ 0.5) led to exclusions: four participants in the LSD dataset and six in the psilocybin dataset were removed, and one additional LSD participant left the scanner due to anxiety, yielding final samples of 15 (LSD) and 9 (psilocybin). Standard steps included removal of initial volumes, de-spiking, slice-timing and motion correction, brain extraction, rigid and non-linear registration to MNI space, scrubbing (FD = 0.4), 6 mm spatial smoothing, band-pass filtering (0.01–0.08 Hz), linear and quadratic detrending, regression of six motion parameters and three anatomical nuisance signals; global signal regression was not performed. Quality control indicated no distance-dependent motion confounds after denoising. Gradient-mapping analyses were conducted using the BrainSpace toolbox. Surfaces were downsampled to 10,000 vertices, and for each subject a 10,000 × 10,000 Pearson correlation matrix was computed, z-transformed and thresholded row-wise at 90% sparsity to retain strongest connections. Cosine similarity was computed on the thresholded matrix and diffusion map embedding (α = 0.5) was applied to obtain low-dimensional embedding components (gradients). Individual gradients were aligned to a group template using Procrustes rotation. Group contrasts and behavioural associations were evaluated with surface-based linear models implemented in SurfStat. To probe hierarchical specificity, vertex-wise gradient scores were binned into ten percentile ROIs per subject; combined unimodal (lowest three bins) and transmodal (highest three bins) ROIs were created and bin-wise functional connectivity (Pearson r) with these combined ROIs was compared between drug and placebo. Seedmaps for each bin were also computed and compared at vertex-wise and network-wise levels.

Diffusion map embedding revealed multiple orthogonal gradients per subject; the report focuses on the first three gradients that explained the largest variance. The principal gradient replicated prior work by spanning unimodal somatomotor regions at one extreme to transmodal default-network and superior frontal regions at the other. Variance explained by this gradient was: LSD placebo mean 12.9%, LSD mean 12.6%; psilocybin placebo mean 13.1%, psilocybin mean 11.9%. Quantitatively, the principal gradient contracted under both drugs: LSD showed a significant reduction in the difference between maximum and minimum principal gradient values (t28 = -2.5, p = 0.01) and psilocybin showed a larger contraction (t16 = -3.9, p = 0.001). Vertex-wise and network-wise contrasts indicated that both unimodal-proximal and transmodal-proximal regions became less differentiated along this axis. Specific topographical changes varied somewhat by drug but displayed broad convergence. Under LSD, vertex-wise increases in principal gradient scores appeared in somatomotor cortex and medial/lateral visual regions, while decreases were observed in precuneus, premotor cortex, superior and inferior frontal gyri, superior parietal lobule, and Wernicke's area (FWE p < 0.05, cdt < 0.01). Psilocybin produced vertex-wise increases in somatomotor, auditory, and lateral visual cortex, and decreases in superior frontal gyrus and inferior parietal lobule (FWE p < 0.05, cdt < 0.01). Network-wise, LSD increased gradient scores in visual and somatomotor networks and decreased scores in default, frontoparietal control, and dorsal attention networks (p < 0.05 FDR-corrected). Psilocybin showed a network-wise increase in the somatomotor network and decrease in the default network (p < 0.05 FDR-corrected). Control analyses found no significant correlation between principal gradient differences and motion. Gradient-based connectivity mapping supported the hypothesis of increased unimodal–transmodal crosstalk. Lower percentile bins (unimodal side) exhibited significantly reduced connectivity with the combined unimodal ROI and less negative connectivity with the combined transmodal ROI, while higher percentile bins (transmodal side) showed increased connectivity with the unimodal ROI but not with the transmodal ROI (p < 0.05, FDR-corrected or uncorrected). Seedmap analyses indicated that lower bins decreased FC with somatosensory and visual networks and increased FC with frontoparietal control regions, notably lateral prefrontal and lateral parietal cortex (FWE p < 0.05, cdt < 0.05 vertex-wise; network-wise p < 0.05 FDR or uncorrected). Higher bins tended toward increased FC with visual, dorsal attention, and salience networks. The second gradient separated visual cortex from somatomotor/auditory cortex and explained: LSD placebo mean 8.8%, LSD mean 8.6%; psilocybin placebo mean 9.8%, psilocybin mean 10.3%. The third gradient separated executive/task-positive regions from other cortex and explained: LSD placebo mean 7.7%, LSD mean 6.0%; psilocybin placebo mean 8.4%, psilocybin mean 7.5%. LSD produced vertex-wise increases in lateral/ventral visual cortex and middle/superior frontal gyri and decreases in somatomotor/auditory cortex for the second gradient (FWE p < 0.05, cdt < 0.01), with network-wise increases in frontoparietal control and visual networks and decreases in the somatomotor network (FDR p < 0.05). Psilocybin showed no significant second-gradient effects, which the authors attribute to limited statistical power. For the third gradient, both drugs showed patterns consistent with a contraction reflecting reduced differentiation of executive regions, though reported effects were stronger for LSD. Finally, behavioural associations identified one significant relationship: in the LSD dataset, greater principal gradient values in a left medial prefrontal/anterior cingulate cluster extending into superior frontal gyrus were positively associated with increases in ego-dissolution (FWE p < 0.05, cdt < 0.01). No significant associations were observed for psilocybin or for complex imagery in either dataset.

Girn and colleagues interpret their results as evidence that serotonergic psychedelics acutely attenuate macroscale cortical hierarchy, consistent with predictions derived from the REBUS model. The principal gradient contracted under both LSD and psilocybin, indicating reduced differentiation between unimodal and transmodal cortex. Gradient-based connectivity analyses suggest this contraction reflects structured increases in unimodal–transmodal crosstalk: unimodal regions became less internally cohesive and more connected to transmodal hubs, while transmodal regions preserved intra-transmodal connectivity and increased connections to unimodal regions. The authors propose that this pattern corresponds to fewer intervening processing steps between sensorimotor and association cortex, which could underlie the experiential blurring of internal and external boundaries reported during psychedelic states. The discussion places these findings in the context of prior literature showing altered static and dynamic functional connectivity and increased signal complexity with psychedelics, arguing that the gradient approach offers a low-dimensional organisational framework that can reconcile heterogeneous network-level findings. Differences between LSD and psilocybin were noted for the second gradient—LSD showed reduced visual differentiation not evident with psilocybin—which aligns with prior reports of stronger visual-cortex whole-brain connectivity changes with LSD. The positive association between principal gradient changes in left mPFC/ACC and ego-dissolution under LSD is highlighted as a potentially meaningful link between hierarchical cortical organisation and a clinically-relevant subjective effect, though the authors explicitly caution that replication is needed given the small samples and the absence of the effect in the psilocybin data. Key limitations acknowledged by the investigators include relatively small final sample sizes, which constrain statistical power and generalisability; difficulties in collecting larger psychedelic datasets due to logistical, funding, and ethical hurdles are noted. The authors also emphasise that, while LSD and psilocybin have complex pharmacology, the findings are consistent with past work implicating 5-HT2A receptor agonism in psychedelic effects, suggesting a possible serotonergic modulation of cortical functional differentiation. They conclude that cortical gradients are sensitive to acute pharmacological manipulation and recommend replication and further research to determine whether acute reductions in cortical hierarchy relate to therapeutic outcomes observed in psychotherapeutically mediated psychedelic treatments. The authors present their results as a foundational contribution to understanding how serotonergic psychedelics reorganise macroscale brain organisation and its relation to conscious experience.