Meditation, psychedelics, and brain connectivity: A randomized controlled resting-state fMRI study of N,N-dimethyltryptamine and harmine in a meditation retreat

Serotonergic psychedelics such as psilocybin, LSD and N,N-dimethyltryptamine (DMT) produce profound alterations of consciousness largely via 5-HT2A receptor agonism. Ayahuasca combines DMT with the monoamine‑oxidase inhibitor harmine to permit oral activity and has been investigated for therapeutic effects in depression, PTSD and addiction. Parallel lines of research have shown that meditation practices also alter consciousness and can change brain functional connectivity (FC), with reported effects on networks such as the visual network (VIS), salience network (SAL), default mode network (DMN) and frontoparietal network (FPN). Prior fMRI work has examined acute and subacute connectivity changes after psychedelics and after meditation, and a small number of studies have begun to probe combined psychedelic‑meditation interventions, chiefly with psilocybin, but it remains unclear whether findings generalise across compounds or how DMT‑based formulations interact with meditation practice. Egger and colleagues designed the present study to test whether an ayahuasca‑inspired formulation (oromucosal DMT plus harmine) administered during a 3‑day meditation retreat would produce distinct or synergistic subacute changes in resting‑state FC compared with placebo. The investigators focused on four complementary connectivity domains: within‑ and between‑network FC of major resting‑state networks, network‑to‑whole‑brain connectivity (voxel-wise FC), parcel‑wise global connectivity, and cortical gradient structure and dispersion. Based on earlier subacute and acute psychedelic studies, they hypothesised increased within‑SAL FC and altered DMN integrity, plus greater VIS‑DMN and SAL‑DMN coupling and higher global FC in the DMT‑harmine group versus placebo, and they tested whether cortical gradient perturbations persisted two days after administration.

The study used a double‑blind, placebo‑controlled, mixed between‑within design conducted during two structurally identical 3‑day meditation retreats. Forty experienced meditation practitioners were randomised to receive either placebo or an oromucosal DMT‑harmine formulation on Day 2 of the retreat; 39 participants entered the final analyses (19 DMT‑harmine, 20 placebo; mean age 43.5 ± 10.3 years; mean meditation experience 2450 ± 1890 h). Pre‑retreat resting‑state fMRI was acquired one day before the retreat and post‑retreat scans were obtained the day after the retreat. Participants practised full days of sitting and walking meditation during the retreat; subjective measures including the Mystical Experience Questionnaire (MEQ) and several mindfulness, compassion and insight scales were collected during and after the retreat. The active treatment comprised four tablets administered at 30‑minute intervals, each tablet containing 30 mg DMT and 30 mg harmine (freebase), for a cumulative dose of 120 mg DMT and 120 mg harmine. Placebo tablets were matched for appearance and taste but contained unloaded template‑inverted particles; nevertheless, blinding efficacy was limited, with >80% of participants correctly identifying allocation at the end of dosing. Safety reporting noted one participant with headache after DMT‑harmine and no other major adverse events. MRI data were acquired at 3T with a standard T1 structural acquisition (on the second visit) and 7.4‑minute resting‑state EPI runs (eyes closed) pre‑ and post‑retreat. Preprocessing used fMRIPrep followed by denoising in CONN, including motion correction (Friston 24), aCompCor physiological regression (components explaining 50% variance), scrubbing (framewise displacement > 0.5 mm), smoothing (6 mm FWHM) and bandpass filtering (0.008–0.09 Hz). Global signal regression (GSR) was performed as a complementary analysis and results are reported both with and without GSR. Connectivity analyses comprised four levels. First, network connectivity maps were computed for six canonical resting‑state networks (VIS, auditory‑sensorimotor (ASM), dorsal attention (DAN), SAL, FPN and DMN) to estimate voxel‑wise FC from each network seed. Second, within‑network FC was derived by averaging connectivity within each network mask and between‑network FC by ROI‑to‑ROI matrices among the six networks. Third, global FC was estimated after Schaefer parcellation into 400 parcels by averaging each parcel's Fisher‑z correlations with all other parcels. Fourth, cortical gradient analysis used diffusion map embedding on thresholded FC matrices to extract the principal gradients (first three explained ~70% of variance), followed by Procrustes alignment and computation of within‑ and between‑network dispersion in 3D gradient space. Statistical contrasts tested (i) synergy (group × time interaction: DMT‑harmine > placebo, post > pre), (ii) group differences post‑retreat (DMT‑harmine > placebo, post), (iii) meditation effect (placebo post > pre), and (iv) psychedelic‑augmented meditation (DMT‑harmine post > pre). Voxelwise inferences used a cluster‑forming p < .001 voxel threshold and FDR q < .05 cluster‑size correction; surface‑based tests for global FC and gradients used established linear model tools. Associations between significant connectivity clusters and questionnaire scores were examined with ordinary least squares regression and FDR correction across questionnaires.

Forty participants were randomised and 39 were included in analyses (19 DMT‑harmine, 20 placebo). Motion and data quality were comparable across groups; one participant was excluded for excessive motion. Blinding was poor with >80% correctly identifying allocation; one participant reported a headache after DMT‑harmine and no other major adverse events were recorded. Network connectivity (voxel‑wise seed‑to‑voxel) analyses showed increased FC between VIS and bilateral insula (SAL) and reciprocal increases between SAL and bilateral calcarine/occipital regions in the Synergy (interaction) contrast, and a Group differences post‑retreat contrast identified increased SAL‑calcarine (VIS) FC in the DMT‑harmine group compared with placebo. These voxelwise cluster findings were the most consistent between analyses and remained when GSR was included. Within‑network FC and between‑network ROI‑to‑ROI results did not survive FDR correction, but several uncorrected trends emerged. The Synergy contrast showed a main effect of time in within‑VIS FC (F(1,74) = 6.16, p uncorrected = .015). In the Psychedelic‑augmented meditation contrast (DMT‑harmine post > pre), within‑VIS FC increased (t = 2.29, p uncorrected = .025). Between‑network repeated‑measures ANOVA interaction trends included VIS‑SAL (F(1,74) = 5.31, p uncorrected = .024), VIS‑FPN (F(1,74) = 4.54, p uncorrected = .037), and DAN‑FPN (F(1,74) = 6.62, p uncorrected = .012). The Group differences post‑retreat contrast showed increased VIS‑SAL (t = 2.51, p uncorrected = .014) and a trend for VIS‑FPN (t = 1.99, p uncorrected = .050) in the DMT‑harmine group. The Meditation contrast (placebo post > pre) revealed decreased FC between VIS‑FPN (t = 2.71, p uncorrected = .001) and DAN‑FPN (t = 2.72, p uncorrected = .001), and decreased VIS‑right insula voxelwise FC. Global FC analyses using the 400‑parcel Schaefer map found no significant group‑by‑time or group post‑retreat differences after FDR correction. However, the Meditation contrast showed three clusters in SAL/FPN with significantly decreased global FC after the retreat in the analysis without GSR (FDR‑corrected); these effects disappeared when GSR was applied. Cortical gradient analyses reproduced the expected principal axes (Gradient 1: sensorimotor to transmodal; Gradient 2: insular to visual; Gradient 3: visual to FPN) and the first three gradients explained approximately 69–71% of variance across scan conditions. No statistically significant differences between groups or time points were detected for any gradient or for within‑ and between‑network dispersion, regardless of inclusion of GSR. Exploratory regression analyses relating the SAL‑VIS cluster (from the Group differences post‑retreat contrast) to psychometric outcomes identified significant group × questionnaire interactions for MEQ Day 2 (b = 0.0065, 95% CI [0.0014, 0.0117], q FDR = .036) and TMS Day 2 (b = 0.0089, 95% CI [0.0019, 0.0159], q FDR = .036). Post hoc correlations showed a positive association between SAL‑VIS FC and MEQ Day 2 in the placebo group (r = 0.52, p = .020); TMS correlations were non‑significant. Repeating analyses with delta (post − baseline) questionnaire scores produced no significant group × questionnaire interactions, suggesting that some associations may reflect baseline differences rather than change induced by the intervention.

Egger and colleagues interpret the data as indicating distinct neural signatures for meditation alone versus psychedelic‑augmented meditation. The principal between‑group finding was increased VIS‑SAL connectivity in the DMT‑harmine group compared with placebo, and within‑group contrasts suggested that DMT‑harmine produced increased VIS integrity and greater VIS‑DAN coupling while the placebo/meditation condition tended to decrease FC between several networks. The authors argue these patterns point to different neurobiological pathways: meditation here chiefly reduced inter‑network coupling, whereas DMT‑harmine appeared to enhance visual and attention network integration. The increased VIS‑SAL FC is discussed as potentially reflecting heightened visual imagery and enhanced integration of visual and emotionally salient information following the psychedelic experience; the SAL is involved in detecting and integrating sensory, emotional and memory inputs, which could explain persistent VIS‑SAL coupling two days post‑administration. The authors note that their initial hypotheses (e.g. increased within‑SAL FC or stronger VIS‑DMN and SAL‑DMN coupling) were not confirmed and attribute divergences to differences in compound, dosing ratios, analytical focus and non‑pharmacological contextual factors such as the group retreat setting. They further emphasise that the absence of cortical gradient disruption two days after DMT‑harmine supports the idea that cortical hierarchy collapse is an acute, reversible phenomenon associated with the immediate psychedelic state rather than a lasting reorganisation. Limitations acknowledged by the investigators include the lack of an intervention‑free control group and a meditation‑free psychedelic arm, which restricts the ability to parse unique versus synergistic effects. The study captured only subacute time points and thus cannot describe acute dynamics or longer‑term persistence. Blinding efficacy was poor and may have influenced subjective and neural outcomes; the authors justify using an inert placebo to preserve neuroimaging contrast but note the trade‑off with expectancy. Generalisability is limited by the sample composition (predominantly white, highly experienced meditators) and by conducting the study in a group retreat setting where cross‑condition influences and two separate retreat sessions could have affected results. The authors also mention methodological considerations such as relatively low smoothing estimates and a relatively stringent denoising pipeline that might reduce sensitivity to subtle effects. For future work the authors recommend studies that include both acute and subacute scans, additional control arms (drug‑only, no‑intervention), investigations of effective (directional) connectivity to probe VIS–SAL interactions, and broader, more diverse samples to improve generalisability. They suggest that the observed VIS‑SAL changes indicate a potential post‑acute window during which therapeutic exercises (for example, imagery‑based interventions) could be coupled with mindfulness practices to leverage the integration of visual and emotionally salient material.

The investigators conclude that psychedelic‑augmented meditation produces neural effects that are distinct from meditation alone, with DMT‑harmine enhancing functional connectivity within visual areas and between visual and attention/salience networks, while meditation in the placebo condition chiefly reduced inter‑network connectivity. The absence of altered cortical gradient architecture two days after dosing suggests that the acute collapse of hierarchical cortical organisation reported with psychedelics is transient and returns to baseline within days. These results point to complementary mechanisms by which psychedelics and mindfulness practice might be combined in therapeutic contexts, and they motivate further controlled studies to clarify timing, specificity and clinical relevance.