Does one dose of psilocybin rewire the brain?

The headlines can't agree. Some say a single dose rewires the brain and opens a weeks-long window of change; others, that the brain quietly returns to normal within a month. Both are describing real studies. Part of the disagreement comes from the science being fragmented: different labs, populations, and instruments, rarely read side by side.

Put side by side, a pattern appears. In depression patients given two doses with therapy support, psilocybin increased the brain's overall “integration,” and bigger changes went with bigger improvements in mood (Daws, 2022). In healthy volunteers given one dose, a newer study found that a month later the enduring changes in how the brain functions were “largely absent”: what lasted was mainly how people felt (Lyons, 2026). Same drug, opposite-sounding results, and one studied sick brains, the other healthy ones.

The healthy side isn't a clean zero, though. A study that scanned seven healthy adults over and over (before, during and for weeks after a dose) found the brain's networks thrown far out of sync during the experience, with most of that fading afterwards. But one link, between the hippocampus and the brain's “default-mode” network, stayed weaker for weeks (Siegel, 2024). So far, lasting change in healthy people looks narrower than in patients, but not absent, and the healthy-volunteer studies measured at different times.

And treatment is a separate question from rewiring. In 233 people with hard-to-treat depression, one 25 mg dose beat a tiny control dose at three weeks, but a lasting benefit beyond that was not established (Goodwin, 2022). In a head-to-head trial against a standard antidepressant, psilocybin did not clearly win on the main measure after six weeks (Carhart-Harris, 2021). In patients, both the brain changes and the symptom changes are established at about three weeks, not yet beyond.

The measure matters too. In mice, one dose visibly remodels the wiring: new connections sprout in the frontal cortex within a day and are still there a month later (Shao, 2021), and the remodeling depends on the brain's own activity while the drug is on board (Jiang, 2026). So structural “rewiring” and “function returned to baseline” can both be true at once: the scaffolding shifts while the day-to-day signal resets, though the clearest structural evidence here is in mice, and some functional changes in people last weeks. The most direct human test so far points the other way: in 15 healthy adults, synaptic density measured by PET one week after a dose did not rise significantly overall, though in a small exploratory comparison, people dosed in a therapy-like room showed more of an increase in the frontal cortex than people dosed in a scanner (Johansen, 2026).

And the famous “four-week window”? That comes from a mouse study showing psychedelics reopen a critical period for social learning, with the window scaled to how long the trip lasts: about two weeks for psilocybin, four for ibogaine (Nardou & Dölen, 2023). It's a striking result, but it's about a specific kind of learning in mice, not a proven four-week rewiring of the human brain. The popular version seems to borrow ibogaine's number and stretch a mouse finding onto people.

One more thing worth knowing: three of the human studies here come from the same research team, two more share a co-author with it, and both mouse rewiring studies come from one lab, so they aren't independent confirmations. A large new study following 62 people out to a year (PsiConnect) may help settle it, but it's still a preprint.

Our reading, offered as a question, not a verdict: across the nine studies in version 2.0, the “does it rewire the brain?” fight may partly dissolve once you separate patients from healthy people, structure from function, and mice from humans. Which leaves the real open question: is the broad change seen in patients something specific to the depressed brain, a fingerprint of getting better, rather than a general effect of the drug? It's also possible the brain change follows the mood improvement rather than drives it, and that dose, therapy support and timing explain part of the split. That's a question for the next round of research, not a claim we can bank today. This reading has not yet been re-tested against Johansen 2026.

Human, patients (Daws 2022, Nat Med). Two depression trials, two doses each with psychological support; psilocybin decreased fMRI network modularity (increased global integration), and integration change correlated with antidepressant response. Imaging to ~3 weeks; long-term durability not established. Subscription article: cited, not reproduced.

Human, healthy (Lyons 2026, Nat Commun). Within-subjects, 28 enrolled, 25 vs 1 mg. Acute EEG changes (LZc↑, alpha↓, p<0.001); at one month, enduring functional changes “largely absent,” with some persistent DTI/modularity signals; psychological gains endured (well-being d = 0.6, insight d = 1.2). Exploratory, not pre-registered.

Mouse, mechanism (Jiang 2026, Cell). Activity-dependent dendritic-spine remodeling in medial frontal cortex; network-specific (strengthens DMN-homolog→subcortical routing, weakens cortico-cortical recurrence); silencing presynaptic input during dosing disrupts rewiring. CC-BY-NC (cited).

Mouse, critical period (Nardou/Dölen 2023, Nature). Psychedelics reopen the social-reward-learning critical period; window duration ∝ acute subjective duration (psilocybin ~2 wk; ibogaine ~4 wk); mechanism via oxytocin-mediated LTD and ECM remodeling. CC-BY.

Human, preprint (PsiConnect 2025, bioRxiv). Multi-echo fMRI + structural/DWI + EEG, n = 62, 19 mg, resting + naturalistic paradigms, follow-up to one year. Not peer-reviewed; CC-BY-NC-ND.

Human, healthy, precision mapping (Siegel 2024, Nature). n = 7, dense repeated fMRI before, during and after 25 mg, methylphenidate as active control. Acute functional-connectivity desynchronization far larger than control; most changes normalized within days; decreased anterior hippocampus–default-mode connectivity persisted through the three weeks scanned and was back at baseline at a 6 to 12 month revisit (n = 4, underpowered). CC BY 4.0 (cited).

Human, TRD trial (Goodwin 2022, NEJM). Phase 2b, n = 233, single dose 25 / 10 / 1 mg with psychological support; MADRS change at week 3 favored 25 mg over 1 mg, 10 mg did not separate; durability beyond week 3 not established. Adverse events including suicidal ideation or self-injury in all groups. Sponsor: COMPASS Pathways.

Human, active-comparator trial (Carhart-Harris 2021, NEJM). Phase 2, n = 59, two 25 mg doses vs daily escitalopram (plus 1 mg psilocybin) over six weeks; primary QIDS-SR-16 difference not significant; secondaries favored psilocybin without multiplicity correction. Part of Daws's imaging comes from this trial.

Mouse, spines (Shao 2021, Neuron). Chronic two-photon imaging in medial frontal cortex; one dose increased spine density and size within 24 hours, still elevated a month later. Kwan lab, as Jiang 2026. License not yet checked (cited).

Human, healthy, synaptic density (Johansen 2026, Transl Psychiatry). 0.3 mg/kg, [11C]UCB-J PET (SV2A) in frontal cortex and hippocampus at baseline and one week; open-label, single arm, no placebo. No significant overall increase: VT at an interim analysis of the 12 with complete arterial data, mean change frontal cortex 0.3 mL/cm3 (SD 2.6) and hippocampus 0.3 mL/cm3 (SD 2.0), stopped for futility; BPND in all 15, mean change −0.01 (SD 0.35) frontal and 0.01 (SD 0.28) hippocampal. Exploratory setting contrast in the change in VT, therapy-like room against MRI scanner: 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 room group alone: 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). Accelerated article preview as of 2026-10-05. CC BY 4.0 (cited).

Integrity note. Lyons, Daws and Carhart-Harris 2021 share a lead or senior author (Carhart-Harris) and several co-authors, and Daws reuses that trial's patients: the Imperial human evidence is one program. Siegel and Goodwin each list Carhart-Harris as one mid-list co-author, and Johansen 2026 does not list him; the trust panel shows those as ties between programs, not merges. Shao and Jiang are one lab.

Posed synthesis (our interpretation, two automated stress tests, expert review pending). Tested against the nine studies in version 2.0, not yet against Johansen 2026. The divergence may be largely explained by population (global integration in patients after two doses, imaged ≤ 3 weeks; in healthy volunteers after one dose, so far only narrower persistent change, e.g. hippocampus–DMN coupling), measure (persistent structural change mainly in rodents, so overlapping species; some human functional change lasts weeks), and species (mechanistic clarity in rodents). Testable question: is the broad human signal a psilocybin-specific response of the depressed brain rather than a general drug effect? Alternatives not excluded: population is confounded with dose count, psychological support and timepoint; all human imaging samples are small; integration may follow mood improvement rather than drive it; in patients neither imaging nor clinical effects are established beyond about three weeks.