LSD alters eyes-closed functional connectivity within the early visual cortex in a retinotopic fashion

Psychedelic effects of lysergic acid diethylamide (LSD) have been linked to activation of the serotonin 2A receptor (5-HT2A R), which is highly expressed in occipital cortex, particularly primary visual cortex (V1). Earlier work has reported occipital alpha suppression, increased blood oxygen level dependent (BOLD) signal and cerebral blood flow (CBF) in visual areas, and associations between these physiological changes and reports of visual hallucinations under psychedelics. Models of psychedelic geometric imagery have hypothesised aberrant dynamics within the retinotopically organised visual cortex, but the behaviour of spatially organised (retinotopic) spontaneous activity during eyes-closed LSD-induced imagery had not been directly tested. This study set out to test whether eyes-closed LSD alters resting-state functional connectivity (RSFC) within early visual cortex in a manner consistent with retinotopic organisation. Specifically, Roseman and colleagues hypothesised that, under LSD but not placebo, RSFC between V1 and V3 patches that represent the same polar meridian (horizontal–horizontal or vertical–vertical, i.e. retinotopically congruent) would be stronger than RSFC between patches representing different meridians (horizontal–vertical or vertical–horizontal, i.e. incongruent). The prediction reflects the idea that the early visual system under LSD may behave “as if” it were processing spatially localised visual input even with the eyes closed.

This analysis used data from a within-subjects, balanced-order design in which 20 healthy volunteers attended two scanning days (LSD and placebo) at least two weeks apart. Each session comprised an intravenous infusion over 2 minutes of either LSD (reported in the text as 75 mg administered in 10 ml) or placebo saline, followed by an acclimatisation period and imaging. MRI scanning began at about 70 minutes postdosing and included structural scans, arterial spin labelling, and BOLD fMRI; two eyes-closed resting-state BOLD scans (each 7 min 20 s) were acquired per session and used for the present analyses. A BOLD retinotopic localizer was performed after the resting-state scans to identify retinotopically selective patches in V1 and V3. Participants were screened for physical and psychiatric health, had prior experience with psychedelic drugs, and provided written informed consent. Key exclusion criteria included age under 21, personal or immediate family history of psychotic disorder, recent psychedelic use, pregnancy, problematic alcohol use, or other medical contraindications. Imaging was performed on a 3T scanner. One participant withdrew during scanning due to anxiety; additional participants were excluded during preprocessing for excessive head motion or imperfect retinotopic patch identification, leaving 10 subjects (3 females) in the final sample for the reported analyses. Preprocessing combined tools from FSL, AFNI and FreeSurfer. Steps included removal of the first three volumes, despiking, slice-timing correction, motion correction, brain extraction, registration to anatomy, scrubbing using a frame-wise displacement (FD) threshold of 0.4 mm (volumes exceeding threshold were replaced with the mean of surrounding volumes), band-pass filtering (0.01–0.08 Hz), linear and quadratic detrending, and regression of nine nuisance regressors (three anatomical regressors: ventricles, draining veins, local white matter; and six motion parameters). The authors used a local white-matter regressor (25-mm radius) rather than a global regressor to mitigate motion-related artefacts. Retinotopic mapping used a 4 min 24 s stimulus alternating vertical and horizontal polar angles (8 cycles) presenting a coloured checkerboard wedge with moving faces and dots. Fourier analysis of the placebo-session localizer identified vertical and horizontal meridian activations, and borders between V1, V2 and V3 were delineated manually for each subject. From these maps four patches were defined per hemisphere: V1_hor, V1_ver, V3_hor, and V3_ver. Because the LSD-session localizer data were noisier due to motion, the retinotopic patches were defined based on the placebo localizer only. For connectivity, the mean time series from each patch were extracted from each resting-state scan. Rather than simple Pearson correlations, linear regressions were performed between patch pairs to produce parameter estimates (b values) representing RSFC (b_hor_hor, b_hor_ver, b_ver_hor, b_ver_ver). The primary outcome, termed retinotopic coordination, was computed as (b_hor_hor + b_ver_ver - b_hor_ver - b_ver_hor)/2, capturing the relative strength of congruent versus incongruent retinotopic couplings. Subjective measures included in-scanner visual analogue scale (VAS) ratings after each scan for simple and complex eyes-closed visual imagery and the 11-factor Altered States of Consciousness (ASC) questionnaire administered retrospectively at the session peak.

After exclusions for withdrawal, head motion and imperfect retinotopic identification, 10 subjects (3 females) remained for analysis. Retinotopic patches (V1_hor, V1_ver, V3_hor, V3_ver) were identified for each subject and their surface areas reported in the paper's figures. The principal finding was a significant increase in retinotopic coordination under LSD relative to placebo. Mean retinotopic coordination was 0.068 ± 0.058 under LSD and -0.005 ± 0.02 under placebo; a one-tailed paired t-test yielded t = 3.93, p = 0.0018, with Cohen's d = 1.6. Nine of the 10 subjects showed a change in the predicted direction. The effect was significant when computed on the mean of the two resting-state scans and remained significant when each resting-state scan was analysed separately (p = 0.0288 and p = 0.0057). Hemisphere-wise trends were reported (right p = 0.022; left p = 0.078), and the result held when using Pearson correlation instead of regression to compute RSFC (p = 0.033). Subjective ratings showed increased visual imagery under LSD: within-scanner VAS differences (LSD–placebo) averaged 9.73 ± 6.43 for simple hallucinations and 7.37 ± 6.87 for complex hallucinations (scale range 0–20). Retrospective ASC scores at peak reported mean increases of 65 ± 34 and 42 ± 28 for elementary and complex imagery factors, respectively, with these changes reaching p < 0.05 after Bonferroni correction. Importantly, the increase in retinotopic coordination did not correlate with the between-condition difference in head motion (reported as r = -0.83, p = 0.999, one-tailed in the extracted text) nor with subjective imagery ratings (e.g., ASC imagery factors or VAS simple and complex hallucinations). Additional analyses showed that other indices of visual-area activity in the same dataset (increased CBF, decreased MEG alpha power) did not correlate with the retinotopic coordination change (reported p = 0.98 and p = 0.97, one-tailed, respectively).

Roseman and colleagues interpret the data as evidence that LSD modulates spontaneous activity within early visual cortex such that functional coupling reflects the intrinsic retinotopic architecture: regions of V1 and V3 that represent the same visual meridian become more tightly coupled under LSD than regions representing different meridians. This pattern is consistent with the proposition that, during eyes-closed psychedelic imagery, the early visual system may behave "as if" it were processing spatially localised visual inputs. The authors situate these findings with prior observations of occipital alpha suppression and increased BOLD/CBF in visual areas under psychedelics, and they note that this study addresses a specific unresolved question about lower-level visual involvement in eyes-closed psychedelic imagery. Several possible explanations and limitations are discussed. One explanation for the absence of correlation between retinotopic coordination and overall subjective imagery intensity is that the metric may capture the spatial properties or spatial acuity of hallucinations rather than their gross intensity; the authors suggest that measures targeted at spatial quality or localisation of imagery might better relate to the retinotopic effect. Head motion is acknowledged as a major confound in psychedelic neuroimaging: the investigators report that subjects with greater LSD–placebo motion differences tended to show smaller retinotopic coordination differences, implying motion diluted rather than produced the effect. They also consider and reject, on the basis of non-significant correlations, the idea that retinotopic coordination merely reflects a general increase in activity (CBF or decreased alpha) or a global-arousal driven increase in BOLD variance. Other limitations highlighted include the small final sample size (N = 10) after data loss and the consequence that future studies should collect extra data to allow for motion-related exclusion. The authors propose methodological improvements for follow-up work, such as including objective and subjective arousal measures to include as covariates and acquiring task conditions that provide benchmarked visual processing (for example, movie viewing) to compare retinotopic coordination during stimulus-driven and stimulus-free imagery. Overall, the paper frames the finding as a novel indication that early visual cortical organisation is recruited or reorganised during LSD-induced eyes-closed imagery, while emphasising the need for further studies to localise sources and clarify relationships to subjective phenomenology.

The study concludes that under LSD the visual cortex appears to act as if processing spatially localised visual information during eyes-closed rest. The authors call for additional research to determine whether eyes-closed psychedelic imagery arises from changes confined to early visual cortex or whether upstream or downstream regions are also implicated, and to identify more precise associations between the subjective qualities of psychedelic imagery and underlying brain activity.