Does one dose of psilocybin rewire the brain?

In healthy volunteers, a single 25 mg dose left enduring functional brain changes “largely absent” one month later, though some diffusion-imaging and network-modularity signals persisted; what lasted was mainly psychological: well-being and insight.

Why this tierLyons: within-subjects, 28 enrolled, exploratory, not pre-registered. “Largely absent” is the authors' phrasing.

Would change ifA pre-registered, adequately powered study in healthy volunteers finds enduring functional brain change at one month or later. That would move this from preliminary to active debate; an independent replication of the null would move it toward established.

In depression patients given two doses with psychological support, psilocybin increased global brain integration (lower network modularity), and the size of that change tracked the antidepressant response.

Why this tierDaws: two trials; correlational; subscription article (facts cited, figures not reused).

Would change ifAn independent research program fails to find increased global integration in depression patients, or finds it without any link to antidepressant response. A second independent replication would move this toward established.

Those integration changes were seen up to three weeks after the second of two doses; the study did not establish how much longer they last.

Why this tierDaws: fMRI at baseline and ~3 weeks post-dose; longer-term durability not measured.

In mice, one dose reorganizes the brain-wide inputs to frontal cortex pyramidal neurons in a network-specific way: it strengthens the routing of inputs from perceptual and medial regions, a homolog of the default mode network, to subcortical targets, and weakens inputs that form cortico-cortical recurrent loops. Silencing one presynaptic region during dosing interfered with this rewiring, so the reorganization depends on drug-evoked spiking activity.

Why this tierJiang 2026: monosynaptic rabies tracing of the inputs to frontal cortex pyramidal neurons; per the abstract, silencing a presynaptic region during dosing interfered with the rewiring. The silencing result concerns these traced inputs, not dendritic spines. Spine growth after one dose is the same lab's earlier result (Shao 2021, c-mouse-spines-persist).

Would change ifAn independent lab fails to find the network-specific input reorganization with rabies tracing, or finds that it survives silencing of the presynaptic region during dosing. Either would move the mechanism from preliminary to active debate.

In mice, psychedelics reopen a “critical period” for social-reward learning, and the window's length is proportional to each drug's acute subjective duration: roughly two weeks for psilocybin, up to four for ibogaine.

Why this tierNardou/Dölen: shared across psychedelics; mechanism via oxytocin-mediated LTD and extracellular-matrix remodeling. Durations are the paper's mouse findings.

The popular “psilocybin opens a four-week window that rewires your brain” collapses different things: it appears to borrow ibogaine's four-week mouse figure (psilocybin's was ~2 weeks) or to stretch a one-month mouse spine result, and it merges a behavior-specific mouse finding with human brain imaging, none of which is the same claim.

Why this tierThe mouse critical-period result is about social-reward learning, not general brain rewiring; the four-week figure is ibogaine's, not psilocybin's; human durable functional change is modest at best (Lyons).

In seven healthy adults scanned repeatedly before, during and after a 25 mg dose, psilocybin acutely desynchronized brain networks far more than an active control (methylphenidate). Most changes faded; an exploratory analysis in six of them suggested reduced hippocampus–default-mode coupling during the three-week follow-up.

Why this tierSiegel: precision functional mapping (many repeated scans of a few people rather than one scan of many); n = 7; the only study here with a psychoactive control. Washington University group; Carhart-Harris is one co-author of 33, so it counts as a separate program tied to the Imperial one. Participants were scanned for three weeks after the dose. The persistence analysis used six participants and 186 repeated observations pooled over days 1 to 21: anterior hippocampus functional-connectivity change, linear mixed-effects estimate 0.095 (95% CI 0.032 to 0.168), P = 0.0033 uncorrected, in a region chosen by a screen at P < 0.05. That statistic tests the region's change metric, while the hippocampus–default-mode decrease is shown in the paper's Fig. 5c–d, and it does not establish persistence on each day or at day 21. At an open-label revisit 6 to 12 months later, values appeared near baseline in four people, a sample not powered for small changes.

In the largest trial here (233 people with treatment-resistant depression), a single 25 mg dose reduced depression scores more than a 1 mg control at three weeks; 10 mg did not separate from control, and a lasting benefit beyond three weeks was not established.

Why this tierGoodwin: phase 2b, randomized, double-blind, three doses (25, 10, 1 mg) with psychological support, after participants stopped their antidepressants; primary outcome MADRS change at week 3. Sponsor-funded (COMPASS Pathways). Adverse events, including suicidal ideation or self-injury, occurred in every dose group. A clinical outcome, not brain imaging.

In a six-week double-blind trial of 59 people with moderate-to-severe depression, two 25 mg doses of psilocybin did not significantly beat daily escitalopram on the primary depression score; secondary measures leaned toward psilocybin but were not corrected for multiple comparisons.

Why this tierCarhart-Harris 2021: phase 2, active comparator (escitalopram plus 1 mg psilocybin); primary outcome QIDS-SR-16 at week 6, which may be short for escitalopram to reach full effect. A clinical outcome, not imaging. Included because part of Daws's imaging comes from this trial's patients.

In mice, a single dose increased the density and size of dendritic spines in frontal cortex within a day, and the higher spine density was still measurable about a month later.

Why this tierShao: chronic two-photon imaging of the same dendrites in living mice before and after one dose, in medial frontal cortex (Cg1/M2); the month follow-up (day 34) imaged a subset of four mice. Kwan lab: the same lab as Jiang 2026, so the two mouse rewiring papers are one program.

Six of the human papers here come from one research program (Carhart-Harris and colleagues) and draw on three cohorts, with explicit dataset reuse and shared investigators, so they are not six independent replications; two other human studies (Siegel, Goodwin) each list him as one co-author among many.

Why this tierLyons, Daws and the 2021 escitalopram trial share a lead or senior author (Carhart-Harris) and several co-authors (Erritzoe, Nutt, and others), and part of Daws's imaging comes from that trial's patients. Pasquini re-analyses the Lyons sample, and Harding re-analyses the escitalopram trial; both list Carhart-Harris and Nutt. Erritzoe 2024 follows up the escitalopram trial's patients for six months, with Carhart-Harris as senior author and Nutt among the authors. The six papers rest on three samples (the Lyons volunteers, the escitalopram trial, and Daws's open-label trial). Siegel (33 authors) and Goodwin (74 authors) each list Carhart-Harris mid-list. Verifiable authorship and data facts. Shared authorship alone does not show shared samples, and separately recruited cohorts within one program still add participant data. (Version 2.2 changed “Three” to “Six” when Pasquini, Harding and Erritzoe were added.)

A large multimodal human study (PsiConnect, 62 participants, follow-up to one year) is mapping psilocybin's effects across imaging methods, but it is a preprint, and not yet a settled result.

Why this tierbioRxiv preprint (not peer-reviewed), CC-BY-NC-ND; a dataset/methods contribution rather than a specific finding.

In 15 healthy adults, one dose did not significantly increase synaptic density one week later overall, measured by SV2A PET in frontal cortex and hippocampus. In an exploratory comparison, participants dosed in a therapy-like room showed a greater frontal increase than those dosed in an MRI scanner; the hippocampal difference was smaller, with an interval that included zero.

Why this tierJohansen: [11C]UCB-J PET at baseline and one week after 0.3 mg/kg psilocybin, in an open-label, single-arm study with no placebo. The VT analysis, run at an interim look on the 12 participants with complete arterial blood data, found no significant increase (mean change: frontal cortex 0.3 mL/cm3, SD 2.6; hippocampus 0.3 mL/cm3, SD 2.0; one-sided p = 0.29 and 0.33), and the study was stopped for futility; BPND, computed for all 15, showed no significant increase either (mean change: frontal cortex −0.01, SD 0.35; hippocampus 0.01, SD 0.28). Setting difference in the change in VT: frontal cortex 3.0 mL/cm3 (95% CI 0.4 to 5.5), hippocampus 1.9 mL/cm3 (95% CI −0.3 to 3.9); the authors call it preliminary and hypothesis-generating. With no placebo arm, the authors note they may have underestimated an increase, because other SV2A studies saw drops at a second scan. The room group's own change in VT: frontal cortex 1.1 mL/cm3 (95% CI −0.5 to 2.7), hippocampus 0.8 mL/cm3 (95% CI −0.5 to 2.1). The Methods list participants 1 to 5 as dosed in the scanner and 6 to 15 in the room; the main text does not say how setting was allocated. SV2A density is a marker of synaptic density in living people, not a count of dendritic spines.

Would change ifA larger or independently replicated PET study finding a significant overall increase in synaptic density after one dose would move the human question to active debate. A preregistered test of the setting effect would show whether the subgroup difference holds.

Re-analysing the same healthy volunteers as Lyons, a study that also used computational modelling found activity in fronto-striatal-thalamic circuits more dynamic four weeks after 25 mg than four weeks after the 1 mg control dose, but only a trend against the pre-dose baseline; the authors call the evidence weak to moderate.

Why this tierPasquini: secondary analysis of the Lyons trial's resting fMRI, 25 psychedelic-naive adults in a fixed order (1 mg, then 25 mg four weeks later; no placebo, no randomized order). Dynamic functional homogeneity across 28 fronto-striatal-thalamic regions was lower four weeks after 25 mg than four weeks after 1 mg (t(24) = 2.87, p = 0.008) and trended lower than baseline (t(24) = 1.92, p = 0.06); the paper reads lower homogeneity as more dynamic, more variable activity across a scan, and its Methods say FDR correction “was used to adjust p-values, unless specified otherwise”, without saying which applies to these t-tests. Whole-brain homogeneity fell at both later scans, which the authors attribute to order effects. The modelled drop in structure-function coupling was stable across repeated simulations, which the authors say “do not represent independent biological samples”; its group-level permutation test was a trend (one-sided p = 0.064; the observed difference was smaller than 90% of permuted differences). The effective-connectivity results rest only on stability statistics across simulations. Same sample and program as Lyons (Lyons and Carhart-Harris are co-authors), so not independent. Full text read; CC BY-NC 4.0, cited only.

In ten healthy volunteers who each received three low-to-moderate doses (5 to 20 mg) in random order (a preprint, read from its abstract), brain connectivity one week after a dose showed one dose-dependent increase within a network that survived correction, and none of the 77 nominal dose-dependent changes in single connections did; dose-adjusted changes in conscious state were linked to wider connectivity changes, many surviving correction at the network level.

Why this tierBhagavan: medRxiv preprint, not peer reviewed. Ten healthy volunteers each received three doses between 5 and 20 mg in randomized, blinded order at least a week apart; resting fMRI at baseline and one week after a single dose (registered ACTRN12621000560897). Per the abstract: 77 dose-dependent edge changes (76 increases, 1 decrease, of 1275) with none surviving correction; one dose-dependent between-network increase (of 21) and one within-network increase (of 7), the latter surviving correction; dose-adjusted changes in conscious state were associated with connectivity change, many surviving correction. Abstract only: the full text was unreachable, so which dose in the sequence each scan followed, possible carry-over from earlier doses, the correction procedure, the network involved and the effect sizes are not checked.

In a small open-label trial in chronic cluster headache, with three doses of 0.14 mg/kg a week apart, a change in hypothalamic connectivity tracked how much attacks fell (10 patients in the trial), and a later analysis of the same trial detected no statistically significant change in brain microstructure one week after the last dose (11 patients at baseline, 8 with complete follow-up).

Why this tierMadsen 2024 (abstract only; full text unreachable): 10 patients, 0.14 mg/kg in weeks five, six and seven, fMRI the day before the first dose and one week after the last; attacks fell 31% on average (SD 31), and the change in hypothalamic-diencephalic connectivity correlated with the change in attack frequency (R = -0.81, pFWER = 0.03); how many patients entered that correlation is not checked. Brendstrup-Brix 2026 (publisher full text checked during the 2026-10-11 review; this session could not reach it): a post hoc analysis of the same EPOCH trial (NCT04280055), diffusion MRI before and one week after, 11 patients at baseline and 8 with complete follow-up, analysed with mixed models; all six Holm-adjusted microstructure comparisons had p = 1.00, a nondetection rather than evidence of no change; psilocybin “was not associated with statistically significant changes in brain microstructure or water diffusivity”, though seven of the eight showed descriptive decreases in white-matter diffusivity and intra-neurite volume fraction. Exactly which patients are in both analyses is not checked. No placebo. One Copenhagen program, the same as Johansen 2026 (Knudsen and Madsen are authors on all three).

In depression patients from the escitalopram trial, task fMRI three weeks after the second psilocybin session showed a greater pre/post decrease in surprise-related ventromedial prefrontal and right angular gyrus responses than escitalopram, and, within the psilocybin group, more activity in sensory regions, an increase that did not differ significantly from escitalopram.

Why this tierHarding: secondary analysis of the phase II psilocybin versus escitalopram trial (NCT03429075), 22 psilocybin and 19 escitalopram patients with usable fMRI, scanned the day before the first dose and six weeks later, three weeks after the second psilocybin session. Ventromedial prefrontal treatment-by-time interaction F(1,39) = 7.074, p = 0.0113; the within-group decrease after psilocybin t(21) = 2.195, p = 0.0395, with no correction across the three regions of interest described. The right angular gyrus difference comes from the whole-brain contrast of pre/post change between groups; the within-psilocybin decrease was in the left angular gyrus. No significant change in the nucleus accumbens or superior temporal gyrus, and no significant correlations were detected between the three prespecified ROI responses and anhedonia or music ratings; whole-brain maps cluster-corrected; baseline vmPFC differed between groups at trend level (p = 0.081). Surprise-related affective ratings showed no significant change after psilocybin (they fell after escitalopram). A published correction changed the abstract to “decreased activation in the ventromedial prefrontal cortex”; this claim follows the correction. One song, one task. Same trial and program as Carhart-Harris 2021 and part of Daws 2022. Full text read; CC BY 4.0, cited only.

Two 2026 systematic reviews flag limits in human follow-up imaging: one, covering studies published to July 2024, found follow-up scans lacking in most psilocybin fMRI studies, with lasting effects reported up to four weeks in those that had them; the other, in depression, grouped some network changes as “Longer-term changes (5+ weeks)” without saying in its abstract whether that is when scans were taken, and warned that many studies reuse overlapping datasets.

Why this tierFarré-Colomés (full text read): 36 articles from 20 datasets, published between February 2012 and July 2024; “follow-up scanning timepoints were lacking in most of the studies”, and studies with follow-up scans suggest effects “lasting up to 4 weeks after intake”; PRISMA, not preregistered, no named risk-of-bias tool. Sabbah (abstract only, full text unreachable): 11 studies in depression, PROSPERO-registered; early (0 to 4 weeks) changes included lower modularity and higher integration, and “Longer-term changes (5+ weeks)” included higher prefronto-parietal and salience-network connectivity; “Many studies reused overlapping datasets with high exploratory flexibility and risk of bias.” Which studies carry Sabbah's 5+ week finding, and whether it refers to scan or outcome timing, cannot be checked without its full text. Farré-Colomés's search ended in March 2025 and covered publications to July 2024, so its finding is a coverage statement, not a ceiling on the current literature; its use of “follow-up” is ambiguous, since Daws reports scans three weeks after the second dose. Two separate teams.

In rats, a single dose of psilocybin or of the 5-HT2A agonist 25CN-NBOH reduced immobility in a forced swim test for at least three months; several months after treatment, prefrontal layer 5 neurons still showed changed resting potential, firing and synaptic excitation after either drug, while no difference in synaptic density or spine types was detected between control and psychedelic-treated rats; the abstract does not say whether that structural result holds for psilocybin on its own.

Why this tierKramer (Nichols lab): per the abstract, there were “no overt behavioral differences between psilocybin and 25CN-NBOH treated animals” and functional plasticity “was evident for both psychedelics”; the microscopy and gene-expression nulls compare “control and psychedelic-treated” animals, and the abstract does not say whether psilocybin was tested on its own for them. Brain-slice electrophysiology found functional plasticity “several months after treatment”; microscopy of the recorded neurons found no differences in synaptic density or spine classification, and gene expression for several synaptic markers did not differ. Abstract only: the full text was unreachable, so animal and cell numbers, sampling dates, multiplicity and any drug-specific structural tests are not checked. The longest-lasting change after a single dose reported in this synthesis, from a lab separate from the mouse spine studies.

In rats, one week after a single dose, compared with the same rats before dosing, psilocybin on its own changed relative brain glucose uptake in few measures: uptake rose in the insula and the across-rat correlation between prefrontal and insula uptake strengthened, and no whole-brain cluster survived correction for psilocybin alone; the paper's headline one-week hypometabolism pools psilocybin with LSD and 2C-B.

Why this tierGudmundsen: 12 Long Evans rats of both sexes per drug group (11 analysed for LSD), [18F]FDG-PET at baseline, acutely and one week after 1 mg/kg psilocybin given under the skin, each group compared with its own baseline (no drug-versus-saline contrast described; the saline cohort showed no change between repeated scans). Uptake is normalized to the whole-brain mean, so it is relative. Insula mean change in normalized uptake 0.16 (adjusted p < 0.05); prefrontal-insula covariance (a correlation across rats, not a within-brain connection) change ΔZ = 1.53 (adjusted p = 0.04); thalamus-prefrontal change not significant after correction (adjusted p = 0.61). “Psilocybin produced no acute or persistent clusters surviving FDR correction.” Correction families differ: region-of-interest adjustments cover the specified comparisons, while voxel analyses use separate FDR and FWE procedures. The paper's Results (“despite not reaching significance in individual drug contrasts”) and Discussion (“significantly reduced for psilocybin”) disagree about psilocybin's own change in the pooled one-week clusters. Full text read; CC BY-NC-ND 4.0, cited only.

In diet-induced obese male mice given one dose, no statistically significant treatment difference was detected in synaptic protein puncta counts (PSD-95, synapsin-1) in prefrontal cortex and hypothalamus 30 days later.

Why this tierKeenan: male C57BL/6J mice made obese on a high-fat diet, one 1 mg/kg intraperitoneal dose, then either kept on the high-fat diet or switched to low-fat chow; brains taken 30 days later; 5 to 6 mice per group, between groups rather than repeated imaging. “Psilocybin had no influence on positive puncta of synapsin-1, PSD-95 or co-localisation of both proteins (i.e., synapses)” in the three regions. The authors note that spine density peaks around day 7, so by day 30 a difference “may not have been substantial enough to be detected with the methods used here”. Puncta counts are not spine counts, and a nondetection does not show equivalence or rule out smaller effects. Full text read; CC BY-NC-ND 4.0, cited only.

In two studies, repeated dosing in adolescent rodents was followed by brain differences weeks to months later whose meaning is unclear: mice dosed five times in adolescence showed exploratory differences in normalized regional brain volumes, diffusion and connectivity on MRI as adults, and in rats dosed daily for a week, the higher dose (1 mg/kg) raised a hippocampal progenitor-cell marker (NeuroD1) 16 days after treatment, which was associated with forced-swim escape behaviour measured the day before.

Why this tierSahoo (full text read): 28 mice, an approximate dose of 3.0 mg/kg by gavage every other day on postnatal days 40 to 50; structural and diffusion MRI at postnatal day 85, resting-state connectivity at day 107 and odour-evoked BOLD at day 176, one measurement per modality rather than repeated tests of persistence. Exploratory differences: normalized regional volumes lower in both sexes (males p < 0.001, females p < 0.05) and pooled diffusion measures rose (p < 0.0001); no area-level sex-by-treatment effect survived FDR correction, no standardized effect sizes are given, the regional tests state no correction and appear to pair brain areas rather than animals, the resting-state sample (10 versus 11) pooled the sexes, and the abstract and protein-method timeline conflict with the imaging dates. Funded by Ekam Imaging, in which two authors hold a partnership interest. García-Cabrerizo (abstract; Methods and Fig. 5 checked during the 2026-10-11 review, complete methods not checked): adolescent Sprague-Dawley rats of both sexes, 0.3 or 1 mg/kg by mouth daily for 7 days; on day 16, NeuroD1 was higher after 1 mg/kg (p = .002) but not significantly after 0.3 mg/kg (p = .123); forced-swim behaviour was measured on day 15; across 39 animals pooled over sexes, NeuroD1 correlated with escape behaviour (r = .418, p = .008). NeuroD1 marks progenitor cells: this is not evidence of new mature neurons or of mediation, and multiplicity is not checked. Repeated dosing in development, not one adult dose; two separate labs.

In patients, symptoms have been followed for longer than the brain scans in the patient imaging studies here. In a randomized trial of 104 adults with major depression, one 25 mg dose with psychological support lowered depression scores more than an active placebo (niacin) six weeks later. A six-month follow-up of the escitalopram trial found both groups still improved, with no significant difference between them, but it was observational, with missing data and further treatment allowed, so it cannot show how long a benefit caused by psilocybin lasts.

Why this tierRaison (abstract only; full text unreachable): phase 2, 11 US sites, 104 adults randomized 1:1 (51 psilocybin 25 mg, 53 niacin 100 mg), both with psychological support, after any psychotropic medication was tapered; primary outcome the change in central rater-assessed MADRS from baseline to day 43, mean difference −12.3 (95% CI −17.5 to −7.2; P < .001); at day 8, −12.0 (95% CI −16.6 to −7.4). More participants on psilocybin had a sustained response, but not remission. Psilocybin had higher rates of overall and severe adverse events and no serious treatment-emergent ones. Assessments ended at day 43; the sponsor is not named in the abstract and was not checked. Erritzoe (full text read; CC BY-NC 4.0, cited only): observational six-month follow-up of the Carhart-Harris 2021 trial (two 25 mg doses against six weeks of escitalopram, both with support); 25 of 30 and 21 of 29 patients completed; QIDS-SR-16 between-condition difference at six months 1.51 (95% CI −1.35 to 4.38; p = 0.311); functioning, connectedness and meaning in life favoured psilocybin. Patients had no restrictions on psychiatric care during follow-up, and the authors call the results descriptive, citing limited power, missing data, possible further treatment and self-report. Erritzoe was excluded from the 2026-10-10 screen, whose rules required a brain measure; it is added here under the synthesis's own criteria, which include clinical outcomes. Clinical outcomes, not brain imaging; same trial and program as Carhart-Harris 2021, Daws and Harding.