Neural mechanisms of psychedelic visual imagery

Stoliker and colleagues situate the study within longstanding questions about how internally generated visual imagery is produced by the brain, and how classic serotonergic psychedelics such as psilocybin alter visual experience. The introduction contrasts externally driven perception with internally generated imagery and highlights that psychedelics can reliably produce vivid, eyes-closed visual hallucinations in healthy participants, providing an experimental model to probe the neuropharmacology and large-scale connectivity of the visual system. Preclinical findings implicating 5-HT2A receptor (5-HT2AR) agonism in reduced synaptic gain and altered sensory drive are presented as a mechanistic lead, alongside human observations of altered occipital-parietal alpha oscillations and changes in posterior versus frontal functional synchrony under psychedelics. The study set out to test how psilocybin alters directed (effective) connectivity within a small visual-imagery network comprising the early visual area (EVA), fusiform gyrus (FG), intraparietal sulcus (IPS) and inferior frontal gyrus (IFG). Based on preclinical work suggesting reduced synaptic gain under 5-HT2AR agonism, the investigators hypothesised increased self-inhibition (i.e. reduced synaptic gain) across these regions and reduced excitatory feedforward connectivity under psilocybin, with EVA self-connectivity linked to elementary imagery and top-down directed connectivity linked to complex imagery. The focus on eyes-closed resting-state scans emphasises mechanisms of internally generated imagery rather than stimulus-driven perception.

This paper analysed data from a previously registered, double-blind, randomised, placebo-controlled cross-over study. Twenty-four healthy adult volunteers (mean age 26.3 years, range 20–40; 12 males, 11 females reported prior to exclusions) were recruited and screened with structured psychiatric interviews and symptom questionnaires to exclude current or familial major psychiatric disorders. Each participant attended two sessions at least two weeks apart and received orally either psilocybin (0.2 mg/kg body weight) or an oral placebo (179 mg mannitol with 1 mg colloidal silicon dioxide). Resting-state fMRI scans (10 minutes) were acquired at 20, 40 and 70 minutes after administration, but only the 70-minute scans—corresponding to peak effects—were used for the present analyses. During scanning participants were instructed to keep their eyes closed and to avoid repetitive counting; a short in-scanner version of the 5-Dimensions of Altered States of Consciousness scale (5D-ASC) assessing elementary imagery, complex imagery, disembodiment and experience of unity was administered immediately after the scan. MRI data were collected on a 3 T Philips scanner and preprocessed in SPM12 (slice-timing correction, realignment, spatial normalisation to MNI space, smoothing with a 6 mm FWHM Gaussian kernel). Head-motion was assessed and three subjects were excluded for excessive motion while one additional subject lacked the 70-minute scan, leaving 20 subjects for the connectivity analyses. Regions of interest were chosen from prior imagery literature and comprised EVA, FG, IPS and IFG; time series were extracted as the first principal component of voxels within 8 mm spheres around specified coordinates. The main preprocessing pipeline did not include global signal regression (GSR), although results with GSR are reported in the Supplementary. Effective connectivity was estimated using spectral Dynamic Causal Modelling (DCM). A fully connected DCM (four ROIs, no exogenous inputs) was inverted per subject to fit the observed cross-spectral density; average DCM fits reported were 87.3% for placebo and 86.3% for psilocybin. Group-level inference used Parametric Empirical Bayes (PEB) with a Bayesian general linear model, and Bayesian model reduction was applied for efficient model selection. In DCM, directed connection strengths are reported in Hertz (Hz) and self-connections are modelled as inhibitory and log-scaled (reflecting synaptic gain or sensitivity to inputs). Behavioural–connectivity associations were tested within the PEB framework and the authors adopted a stringent posterior probability threshold (>0.99, approximately Bayes Factor >150) to identify very strong evidence for relationships between connectivity estimates and subjective scores.

The authors report group-level effective connectivity results derived from spectral DCM comparing placebo and psilocybin conditions using the eyes-closed resting-state scans at 70 minutes. They emphasise that self-connections in DCM are modelled as inhibitory and log-scaled, so positive changes reflect relatively increased inhibition (reduced synaptic gain or sensitivity) whereas negative changes reflect relative disinhibition (increased gain). Under psilocybin, the group mean effective connectivity pattern showed increased self-inhibition across the modelled regions, particularly in the EVA and inferior temporal cortex (FG and IFG). Between-region changes included reduced inhibition (i.e. a shift towards excitation) from the IFG to the IPS, FG and EVA, and reductions in inhibition for most connections into the EVA, with the exception of the FG→EVA connection which remained excitatory. The authors present these effects as indicating stronger recurrent inhibition of individual regions together with selective enhancement of top-down influences toward earlier visual cortex. Behavioural associations were examined using in-scanner ratings of elementary imagery, complex imagery, disembodiment and experience of unity. A shift toward excitation in the IFG→FG connection under psilocybin was positively associated with elemental imagery, disembodiment and experience of unity. The FG→EVA connection related to all subjective effects examined, but that particular connection did not show strong evidence of change between placebo and psilocybin. The extracted Results text truncates before listing all associations; consequently the summary cannot report further specific associations that may have been described in the missing portion of the extraction. The authors note that the reported results are the mean effective connectivity estimates without global signal regression, and supplementary analyses with alternative design matrices including GSR are referenced but not detailed in the extracted text.

The authors interpret their findings as evidence that psilocybin produces a pattern of increased self-inhibition (reduced synaptic gain or sensitivity) across visual and associative regions while concurrently reducing inhibitory top-down influences from associative/frontal areas onto posterior regions. They argue this combination would decrease the regions' responsiveness to external inputs and bias the visual system toward internally generated, top-down signalling, which could facilitate vivid eyes-closed visual imagery. This account is aligned with preclinical work showing 5-HT2AR agonism can reduce stimulus-evoked responses in the visual pathway and with prior human observations of dampened parieto-occipital alpha oscillations under psychedelics. The discussion contrasts the current results with alternative theoretical models. In particular, the authors note that their findings do not support the bottom-up dominance posited by the REBUS (RElaxed Beliefs Under pSychedelics) model for the visual–associative connections examined here; rather, they observe reduced feedforward and relatively enhanced feedback influences within the visual network studied. The investigators caution that some connectivity–behaviour associations were not specific to visual measures, because changes in the IFG→FG and FG→EVA connections were also associated with non-visual experiences such as disembodiment and unity. They therefore describe the behavioural associations as preliminary and stress the co-occurrence and correlation among different subjective effects as a challenge for precise statistical separation. Key limitations acknowledged include the modest final sample size after motion exclusion (n = 20), the relatively low–moderate psilocybin dose used (0.2 mg/kg, less than standard clinical doses often used in therapeutic trials), potential influence of ROI selection and coordinate choice, the timing and uniformity of imagery onset across participants, and differences between human and some preclinical paradigms (for example anaesthetised or eyes-open recordings in animals). The authors recommend replication, exploration of alternative region sets (including thalamic or spatial-processing regions), higher doses and task-based designs to probe reality-monitoring and the determinants of imagery content and complexity. They also propose that investigating populations with impaired voluntary imagery (for example aphantasia) could be informative. Overall, the authors present their results as evidence that psilocybin reduces sensitivity to external visual input and shifts hierarchical balance toward feedback signalling, which may underlie the emergence of eyes-closed visual imagery during the psychedelic state.

The authors conclude that psilocybin appears to weaken the sensitivity of visual brain regions to external inputs while altering effective connectivity among regions involved in visual processing, consistent with a hierarchical shift favouring feedback (top-down) signalling. They suggest these changes offer a mechanistic account for the occurrence of vivid eyes-closed visual imagery under psychedelics, extend preclinical findings implicating 5-HT2AR-mediated suppression of stimulus-evoked visual activity, and highlight opportunities for future research with varied dosages, expanded region sets and task designs to further clarify how psychedelics reshape perceptual inference and consciousness.