Acute effects of psilocybin on the dynamics of gaze fixations during visual aesthetic perception

Twenty‑three adults (four females; mean age 31 ± 4 years; mean weight 72 ± 15 kg) with prior experience of serotonergic psychedelics were recruited. Eligibility required at least two prior experiences with doses ≥ 3 g dried psilocybin mushrooms and excluded recent psychedelic or other psychoactive use, current psychiatric disorders by DSM‑5 criteria, pregnancy and ongoing psychiatric medication. After psychiatric screening and informed consent, participants scheduled two sessions separated by ≈ one month. Due to poor eye‑tracking data quality six participants were excluded from eye‑tracking analyses and two participants stopped the task during the high dose session, yielding a final sample of 15 participants for eye‑movement analysis. A priori power calculations had targeted 23 participants for an expected effect size of d = 0.6. The experiment used a self‑blinded, within‑subject design with an active control. Participants prepared and randomised two capsule types themselves: a high dose condition (~3 g ground dried Psilocybe cubensis in gel capsules) and an active control (~0.5 g ground active mushroom mixed with edible mushrooms to match capsule weight). Experimenters and participants were kept unaware of condition identity prior to analysis; sessions took place in homes chosen by participants with researchers present. The eye‑tracking task began about one hour after ingestion, following EEG setup and a 30–45 min acclimation. Each of 30 paintings (selected randomly from a curated BBC set) was presented for 30 s, preceded by a 2 s fixation cross; after each image participants provided two trial‑wise visual analogue scale (VAS) ratings for emotional valence and perceived beauty. Calibration was repeated every five blocks. The full experimental session included additional tasks and lasted approximately 3.5–5 h. Eye movements were recorded with a Gazepoint GP3 HD portable eye‑tracker (150 Hz temporal resolution, visual angle accuracy 0.5–1.0°). Fixations were detected using the device's velocity‑based algorithm with a minimum duration of 150 ms. Fixation coordinates were standardised to a common 1280 × 768 pixel frame. For each image, participant and condition the investigators computed: total number of fixations (N), mean distance between consecutive fixations (ds), mean time between fixations (dt), mean distance between all pairs of fixations (ds2) and their standard deviations. Spatial entropy of the 2D fixation probability distribution was estimated using Shannon entropy on discretised fixation histograms with four bin sizes (40×40, 60×60, 80×80 and 100×100 pixels) to assess robustness across spatial resolutions. Subjective measures comprised repeated VAS items capturing perceptual effects during the acute window (one hour after ingestion and then hourly), the Aesthetic Experience Questionnaire (AEQ) administered immediately after the eye‑tracking task during dosing, and other trait/state scales collected before and after dosing (e.g. STAI, PANAS, MEQ, 5D‑ASC). Statistical comparisons used paired Student's t‑tests for questionnaire totals, Wilcoxon signed‑rank tests for eye‑tracking and trial‑wise VAS data, linear mixed effects models to examine order effects, and Bayesian factors (BF10) to quantify evidence for alternative vs null hypotheses. Multiple comparisons were controlled using Benjamini‑Hochberg FDR where appropriate; effect sizes were reported (Cohen's d, rank‑biserial correlation) and supplementary tables contained additional model results.

Participants correctly identified condition (high vs low dose) on 42 of 46 measurement days (91% accuracy), indicating a substantial unblinding rate. As expected, the high dose produced robust subjective effects: all four total VAS scores (summed perceptual intensity items) were significantly higher under the high dose than the active control, with peak effects between 2 and 3 hours after ingestion. The high dose also yielded significant increases on the 5D‑ASC and MEQ30 scales. Many comparisons that were null showed BF10 values < 1/3, interpreted by the authors as evidence favouring the absence of an effect in those measures. On the AEQ, two subscales differed between conditions: the emotional response factor and the flow of experience factor scored significantly higher during the high dose. By contrast, trial‑wise VAS ratings of emotional valence and perceived beauty averaged across images did not differ between dosing conditions. The authors note that the AEQ captures a global emotional engagement across the session, whereas the VAS reflected immediate reactions to individual paintings. Eye‑tracking analyses (n = 15) revealed that the high dose led to more local visual exploration. Specifically, mean distance between consecutive fixations (ds) and mean distance between all fixation pairs (ds2) were significantly lower in the high dose condition, indicating shorter saccadic spans and a narrower spatial spread of fixations. The standard deviations of both ds and ds2 were also significantly reduced under the high dose, showing decreased variability in fixation distances. Shannon entropy computed for the 2D fixation probability distribution was significantly lower for the high dose across all tested bin widths, consistent with a less entropic, more spatially confined fixation distribution. No significant differences were reported for the number of removed fixations between conditions. Correlation analyses between total VAS scores and eye‑tracking metrics or aesthetic ratings did not yield significant associations after FDR correction. Repeated measures and Pearson correlations between fixation metrics and AEQ dimensions revealed four significant correlations in the low dose (control) condition and none in the high dose; the only significant difference between conditions in correlation strength was for ds versus the AEQ emotional response subscale (low dose r = 0.71, high dose r ≈ 0.08), indicating a decorrelation under the high dose. Linear mixed models including order as a factor found no significant order × condition interactions after FDR correction.

Muller and colleagues interpret their findings as indicating that acute high‑dose psilocybin alters the spatial dynamics of visual exploration during aesthetic perception, producing shorter, less variable fixation distances and a lower entropy of fixation distributions. Behaviourally, this pattern suggests a shift toward more local, spatially confined attention under the high dose. At the same time, participants reported greater overall emotional engagement and flow on the AEQ, while immediate trial‑by‑trial beauty and valence ratings were unchanged, pointing to a dissociation between global affective engagement and momentary evaluative judgements. These results contrast with a simple prediction derived from REBUS that relaxing high‑level priors should increase entropy of visual sampling. The authors propose alternative reconciliations: increased bottom‑up information flow or relaxed priors might heighten processing demands on low‑level features, provoking slower or more local inspection of stimuli, or cause attention to dwell on texture and form at the expense of global semantic sampling. Such a mode of perception aligns with historical anecdotal reports of intensified attention to low‑level detail. The observed loss of correlation between global emotional engagement and global sampling (ds) under the high dose suggests that increased emotional experience under psilocybin does not arise from broader semantic sampling of the artwork. The authors acknowledge several limitations. The self‑blinded, semi‑naturalistic design improved ecological validity but allowed a high unblinding rate and variable implementation of the blinding procedure. Natural settings and longer, diverse task batteries increased the risk of missing data, which led to the exclusion of six participants from eye‑tracking analyses and two participants voluntarily stopping the task in the high dose. Stimulus familiarity was not systematically assessed, stimuli were relatively short in presentation (30 s) and exclusively classical paintings were used, so it remains unclear whether effects generalise to non‑artistic scenes. The sample was predominantly male, and the study was exploratory and not pre‑registered, increasing degrees of analytical freedom. The authors also note that placebo or expectation effects cannot be fully ruled out, though they suggest behavioural measures may be less sensitive to placebo than self‑reports. For future research the investigators recommend pre‑registered, hypothesis‑driven studies, inclusion of non‑artistic control stimuli, varied stimulus durations, and replication under tighter lab control to disentangle low‑level perceptual from top‑down attentional contributions. They consider this study a first quantitative characterisation of how psilocybin modulates fixation statistics during aesthetic perception and a basis for hypothesis generation in subsequent confirmatory work.