Know which conclusions hold.And why.

The papers

• Lyons 2026 ↗

  Human brain changes after first psilocybin use

• Daws 2022 ↗

  Increased global integration in the brain after psilocybin therapy for depression

• Jiang 2026 ↗

  Psilocybin triggers an activity-dependent rewiring of large-scale cortical networks

• Nardou 2023 ↗

  Psychedelics reopen the social reward learning critical period

• Novelli 2025 ↗

  PsiConnect: A Multimodal Neuroimaging Study of Psilocybin-Induced Changes in Brain and Behaviour

• Carhart-Harris 2021 ↗

  Trial of Psilocybin versus Escitalopram for Depression

• Goodwin 2022 ↗

  Single-Dose Psilocybin for a Treatment-Resistant Episode of Major Depression

• Shao 2021 ↗

  Psilocybin induces rapid and persistent growth of dendritic spines in frontal cortex in vivo

• Siegel 2024 ↗

  Psilocybin desynchronizes the human brain

• Johansen 2026 ↗

  Psilocybin’s effect on human brain synaptic plasticity

• Pasquini 2026 ↗

  Modeled Long-Term Effects of Psilocybin on Dynamic Activity and Effective Connectivity of Fronto-Striatal-Thalamic Circuits

• Bhagavan 2026 ↗

  Sustained Effects of Low-to-Moderate Doses of Psilocybin on Brain Connectivity

• Madsen 2024 ↗

  CCH attack frequency reduction after psilocybin correlates with hypothalamic functional connectivity

• Brendstrup-Brix 2026 ↗

  Effects of psilocybin on sleep quality and brain microstructure in chronic cluster headache

• Harding 2025 ↗

  Dissociable effects of psilocybin and escitalopram for depression on processing of musical surprises

• Farré-Colomés 2026 ↗

  Psilocybin effects on brain functional connectivity: a systematic review of fMRI studies

• Sabbah 2026 ↗

  Temporal dynamics in neuroimaging as correlates of therapeutic response to psilocybin in major depressive disorder: A systematic review and critical appraisal

• Kramer 2026 ↗

  Psychedelics produce enduring behavioral effects and functional plasticity through mechanisms independent of structural plasticity

• Gudmundsen 2026 ↗

  Serotonergic psychedelics induce distinct patterns of metabolic activity and covariance within biologically informed rat brain networks

• Keenan 2026 ↗

  Effects of a single dose of psilocybin on diet-induced weight loss in obese mice

• Sahoo 2026 ↗

  Sex-dependent developmental changes in behavior, brain structure, functional connectivity, and sensory perception following exposure to psilocybin during adolescence

• García-Cabrerizo 2025 ↗

  Psilocybin as a fast-acting and long-lasting antidepressant for adolescence: Proposing NeuroD1 as a biomarker of its long-term plasticity

• Raison 2023 ↗

  Single-Dose Psilocybin Treatment for Major Depressive Disorder: A Randomized Clinical Trial

• Erritzoe 2024 ↗

  Effect of psilocybin versus escitalopram on depression symptom severity in patients with moderate-to-severe major depressive disorder: observational 6-month follow-up of a phase 2, double-blind, randomised, controlled trial

The claims and their sources

1. Claim 1: 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. ↗

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

   • Lyons 2026 ↗Nature Communications
2. Claim 2: 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. ↗

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

   • Daws 2022 ↗Nature Medicine
3. Claim 3: 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. ↗

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

   • Daws 2022 ↗Nature Medicine
4. Claim 4: In mice, one dose drives activity-dependent structural rewiring: dendritic-spine remodeling in frontal cortex that strengthens some routes and weakens recurrent cortical loops, and that depends on the drug-evoked neural activity itself. ↗

   Jiang: C57BL/6 mice, rabies-tracing; silencing a presynaptic region during dosing disrupts the rewiring, evidence it is activity-dependent.

   • Jiang 2026 ↗Cell
5. Claim 5: 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. ↗

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

   • Nardou 2023 ↗Nature
6. Claim 6: 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. ↗

   The 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).

   • Nardou 2023 ↗Nature
   • Lyons 2026 ↗Nature Communications
   • Shao 2021 ↗Neuron
7. Claim 7: 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. ↗

   Siegel: 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.

   • Siegel 2024 ↗Nature
8. Claim 8: 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. ↗

   Goodwin: 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.

   • Goodwin 2022 ↗New England Journal of Medicine
9. Claim 9: 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. ↗

   Carhart-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.

   • Carhart-Harris 2021 ↗New England Journal of Medicine
10. Claim 10: 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. ↗

    Shao: 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.

    • Shao 2021 ↗Neuron
11. Claim 11: 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. ↗

    Lyons, 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.)

    • Lyons 2026 ↗Nature Communications
    • Daws 2022 ↗Nature Medicine
    • Carhart-Harris 2021 ↗New England Journal of Medicine
    • Pasquini 2026 ↗Human Brain Mapping
    • Harding 2025 ↗Molecular Psychiatry
    • Erritzoe 2024 ↗eClinicalMedicine
12. Claim 12: 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. ↗

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

    • Novelli 2025 ↗bioRxiv (preprint)
13. Claim 13: 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. ↗

    Johansen: [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.

    • Johansen 2026 ↗Translational Psychiatry
14. Claim 14: 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. ↗

    Pasquini: 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.

    • Pasquini 2026 ↗Human Brain Mapping
15. Claim 15: 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. ↗

    Bhagavan: 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.

    • Bhagavan 2026 ↗medRxiv (preprint)
16. Claim 16: 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). ↗

    Madsen 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).

    • Madsen 2024 ↗Headache
    • Brendstrup-Brix 2026 ↗Journal of Psychopharmacology
17. Claim 17: 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. ↗

    Harding: 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.

    • Harding 2025 ↗Molecular Psychiatry
18. Claim 18: 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. ↗

    Farré-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.

    • Farré-Colomés 2026 ↗Discover Mental Health
    • Sabbah 2026 ↗Journal of Affective Disorders
19. Claim 19: 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. ↗

    Kramer (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.

    • Kramer 2026 ↗Neuropsychopharmacology
20. Claim 20: 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. ↗

    Gudmundsen: 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.

    • Gudmundsen 2026 ↗Nature Communications
21. Claim 21: 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. ↗

    Keenan: 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.

    • Keenan 2026 ↗Translational Psychiatry
22. Claim 22: 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. ↗

    Sahoo (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.

    • Sahoo 2026 ↗Neuropsychopharmacology
    • García-Cabrerizo 2025 ↗Biomedicine & Pharmacotherapy
23. Claim 23: 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. ↗

    Raison (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.

    • Raison 2023 ↗JAMA
    • Erritzoe 2024 ↗eClinicalMedicine

Authored relationships

• population difference

  The headline conflict (“it rewires the brain” vs “changes are largely absent”) comes from studies of different people: broad network change lasting weeks has been reported in patients; in the healthy-volunteer studies here, lasting change has been narrower (Siegel, exploratory) or largely absent (Lyons). Whether population explains the difference is untested: dose count, psychological support, timing and analysis differ too, and a significant result in one group with a non-significant one in another is not evidence of a difference between groups.

  Two different populations and analyses; the field genuinely debates whether durable change is population-specific or a general drug effect.

  Claim 2 ↗ → Claim 1 ↗

• measures differently

  The studies also differ in what they measure, and structural and functional measures can diverge in both directions: one mouse lab found new spines lasting a month, while a rat study found neuron function changed months later with no structural difference detected. So “rewiring” and “function returned to baseline” need not contradict each other, but the structural evidence here is mainly in rodents, so this split overlaps species, and which pattern holds in people is open.

  Structural vs functional measures answer different questions on different timescales.

  Claim 1 ↗ → Claim 4 ↗

• mechanism for

  The mouse rewiring is one candidate mechanism for the human network changes, but that is a cross-species inference, not a demonstration in humans.

  Mechanism shown in mice by one lab (Jiang, Shao); human evidence is imaging-level and correlational. Nardou's oxytocin-dependent critical-period mechanism is a rival candidate.

  Claim 4 ↗ → Claim 2 ↗

• measures differently

  Partly in tension, same population: in a dense-scanning study of seven healthy adults, an exploratory analysis of six suggested one functional change (weaker hippocampus–default-mode coupling) during three weeks of follow-up, where the group study found enduring functional change “largely absent” at a month. Different designs and timepoints; together they are consistent with narrow lasting change, or with changes that fade between a few weeks and a month.

  Siegel (n = 7 enrolled, 6 in the persistence analysis, uncorrected, many scans each, weeks) and Lyons (28 enrolled, group analysis, one month) ask the same question with different instruments. Separate programs, tied by one shared co-author (Carhart-Harris).

  Claim 7 ↗ → Claim 1 ↗

• extends

  A larger, sponsor-funded, single-dose trial from a separate program (one shared co-author) reports patient effects at the same three-week mark and did not establish a lasting benefit beyond it; Daws's imaging stops at three weeks. The shared horizon is partly by design.

  Goodwin measures symptoms and followed patients past week 3; Daws measures brain networks and did not image beyond three weeks. Agreement on timescale, not on mechanism.

  Claim 8 ↗ → Claim 3 ↗

• measures differently

  Same patients, two measures: part of Daws's integration finding comes from a trial where psilocybin did not clearly beat escitalopram on its primary depression score. The integration signal is evidence about the brain, not proof of a better treatment.

  Daws analyzed imaging from two trials, one of them Carhart-Harris 2021. One program, so this is context, not independent support.

  Claim 9 ↗ → Claim 2 ↗

• extends

  Jiang's network-scale rewiring builds on the same lab's finding that the new spines persist for a month: one lab refining its own result, not an independent replication.

  Shao and Jiang share the senior author (Kwan) and several co-authors.

  Claim 4 ↗ → Claim 10 ↗

• measures differently

  The most direct human test of the synapse-growth idea so far found no significant overall increase in synaptic density one week after a dose, while mice show more and larger dendritic spines within a day that last a month. Different species, measures and timepoints, and the human study is small; an exploratory setting comparison found a larger frontal change in a therapy-like room than in the scanner, though the room group's own change was not significant.

  Johansen (human SV2A PET, n = 15, one week) against Shao (mouse two-photon spine imaging, one month). SV2A PET and spine counts are different measures of synapses.

  Claim 13 ↗ → Claim 10 ↗

• measures differently

  Same healthy-volunteer sample (25 in the re-analysis; Lyons enrolled 28), a different analysis: where the primary report found enduring functional change “largely absent” at one month, a re-analysis finds activity in fronto-striatal-thalamic circuits more dynamic four weeks after 25 mg than four weeks after 1 mg, only a trend against baseline, and calls the evidence weak to moderate. One program re-reading its own data, not a replication.

  Pasquini reuses the Lyons sample and fMRI; Lyons and Carhart-Harris are authors on both.

  Claim 14 ↗ → Claim 1 ↗

• measures differently

  Different species, timepoint and method: in mice, more spines a month after one dose; in rats given psilocybin or 25CN-NBOH, several months after one dose, changed neuron function with no difference in synaptic density or spine types detected (the abstract does not say whether the structural result holds for psilocybin alone). The rat study's authors read this as “enduring function plasticity rather than structural plasticity”; a late structural nondetection cannot show that earlier structural growth never occurred.

  Kramer (rats, slice electrophysiology and microscopy, months; abstract only) against Shao (mice, two-photon imaging, one month). Separate labs. The structural null is reported for “psychedelic-treated” rats; the abstract does not separate the two drugs.

  Claim 19 ↗ → Claim 10 ↗

• measures differently

  Around the one-month mark, two-photon imaging found higher spine density in healthy mice, while synaptic protein puncta counts in obese mice detected no significant difference. Different measures, regions (Cg1/M2 for the spine imaging; prelimbic and infralimbic cortex and hypothalamus for the puncta), populations and labs, so this is not a failed replication, and the second study's authors say that, because spine density peaks around day 7, by day 30 a difference may have been too small for their methods to detect.

  Keenan (obese male mice, PSD-95 and synapsin-1 puncta, 30 days, 5 to 6 mice per group, between groups) against Shao (mice, repeated spine imaging of the same dendrites, about a month, in a subset of four mice). Puncta counts are not spine counts.

  Claim 21 ↗ → Claim 10 ↗

• measures differently

  Same trial, a different task: alongside the resting-state integration finding, task fMRI in the escitalopram trial's patients shows changed brain responses to musical surprise three weeks after the second session; unlike the integration finding, no significant correlations were detected between the three prespecified ROI responses and anhedonia or music ratings. It adds a measure at the same three-week horizon, not an independent confirmation or a longer follow-up.

  Harding and part of Daws come from the same trial (NCT03429075) and program; in that trial both scanned three weeks after the second psilocybin session (Daws also used an open-label trial). Harding found no significant correlation between these regions and anhedonia or music ratings.

  Claim 17 ↗ → Claim 2 ↗

• population difference (question)

  Posed as a question, not tested as an explanation: could who was dosed (depression patients or healthy volunteers), what was measured (structure or function) and which species account for why these twenty-four sources disagree about lasting change? The last two overlap, because synapse-level structure has been measured almost only in rodents. The sources report heterogeneous, mostly exploratory findings across populations, regimens, species, assays and follow-up intervals, and they do not establish whether persistent human network changes are specific to depression. What prompts the question: global network change in depression patients, imaged up to three weeks after the second of two doses; narrower or weaker lasting change in the healthy-volunteer studies here, which mostly measured at one week to one month (an exploratory, uncorrected analysis in Siegel, with values near baseline in four people revisited 6 to 12 months later; a change significant against the 1 mg scan but only a trend against baseline in the Pasquini re-analysis; functional change “largely absent” in Lyons, apart from some diffusion and modularity signals); and clear lasting synapse-level structural change after one dose shown by one mouse lab, while a rat study detected changed neuron function months after one dose of psilocybin or 25CN-NBOH with no structural difference (not split by drug in the abstract), an obese-mouse study detected no puncta difference at a month, the one human synaptic-density study detected no significant rise at a week, and a diffusion study in cluster headache detected no significant change at a week. Structure and function have come apart in both directions, and single-dose structural results disagree within rodents too. Two questions for the field: is the broad human signal a response of the depressed brain rather than a general effect of the drug, and does it hold outside the one program it comes from, when six of the human papers, resting on three cohorts, come from that program, including both sides of the population contrast (Daws, Lyons), and the two single-dose depression trials count as one program? What would test it: studies that dose patients and healthy volunteers alike and scan both at the same times, and animal studies that measure structure and function in the same animals over months. (Caveats: population is entangled with dose count, psychological support, age, timing, assay and analysis; a significant result in one population and a non-significant one in another is not evidence of a population difference; expectancy is poorly controlled, since only Siegel used a psychoactive control; every human imaging result here comes from a small sample, and several pre-post designs have no placebo arm; brain integration may follow mood improvement rather than drive it; several version 2.2 sources were read from abstracts only; studies before 2024 were not searched for this update; and in the depression-patient imaging studies included here, scans stop at about three weeks after the second dose, while symptoms improved more than with placebo at six weeks in one trial here, and an observational six-month follow-up found both the psilocybin and escitalopram groups still improved, with no significant difference between them, so how long a brain or symptom effect caused by psilocybin lasts is not settled.)

  PsychDoc's cross-source reading, offered to the field as a question, not asserted as fact.

  Claim 2 ↗ → Claim 1 ↗