In plain terms

The claim being assessed: Psilocybin causes a sustained reduction in human depressive symptoms by inducing a temporary increase in cortical plasticity.

Where the evidence stands: Γ — Plausible mechanism. There is a believable biological story and partial support, but the causal step has not been demonstrated.

What this mechanism does: The evidence reviewed does not show this mechanism causing the outcome. That does not show it has no effect.

Strongest evidence for it: SC1 protocol-level clinical response and SC3 adult experimental cortical effects reach Ε; approximately twelve-week persistence is a qualified Δ. The temporary human cortical-plasticity interval and its causal contribution to sustained symptoms each remain Γ under rubric 07's acute-pharmacodynamic and mediation-by-correlation rules. Weakest-link therefore returns Γ. The red-team pass changed neither the overall letter nor any subclaim level.

What decisive evidence would look like

In plain terms, researchers would change the proposed cause, confirm it changed as intended, and measure a meaningful health outcome while checking other explanations. The result should be able to support or count against the claim.

Relative explanatory fit: Alternative ranked ahead

What this means for psilocybin and depression treatment

In structured clinical protocols, psilocybin has shown symptom improvement for some adults with depression, and experimental work shows short-lived effects on the cortex. The attractive explanation that a brief, measurable period of increased cortical plasticity causes the longer-lasting clinical benefit remains unproven. Treatment effect and proposed mechanism are separate questions.

Implications for health and treatment: The evidence justifies continued, carefully controlled clinical research and nuanced discussion of protocol-based treatment—not claims that a brain 'reset' is established or that an unsupervised experience reproduces trial care. Trial results depend on selected participants, preparation, supervision, follow-up, the treatment setting, and concurrent support. A mechanism being uncertain does not mean symptom benefit is impossible; it means the pathway, durability, and who is likely to benefit still need direct tests.

Limits for individual care: This review cannot say whether psilocybin is safe, appropriate, or effective for a particular person. Psychiatric history, medicines, cardiovascular health, substance use, legal setting, acute adverse effects, and the risk of destabilisation can matter. This page is not medical advice: do not stop prescribed psychiatric treatment, self-medicate, or change care based on it; seek urgent help for suicidal thoughts, severe mood changes, or loss of contact with reality.

Medical information notice: This review provides general health information, not medical advice. It cannot diagnose a condition or replace an individual discussion with a qualified clinician.

This assessment rests on 50 sources and 50 individually checked claims. 50 of those claims were verified against the original paper.

A low grade is not an accusation. It means the causal work has not been done yet, which is the ordinary state of most actively studied mechanisms in biology.

At a glance

Sub-claims assessed separately

ID	Sub-claim	Type	Level	Pivotal card
SC1	A supervised psilocybin-containing protocol causes post-acute reduction of depressive symptoms.	sufficiency	Ε	Card ↗
SC2	Depressive-symptom reduction persists approximately twelve weeks after treatment.	generalization	Δ	—
SC3	Psilocybin induces cortical structural or functional plasticity in adult experimental systems.	mechanism	Ε	—
SC4	Psilocybin induces a temporary period of enhanced cortical plasticity in humans.	mechanism	Γ	Card ↗
SC5	That cortical-plasticity period causally mediates sustained human depressive-symptom reduction.	mechanism	Γ	Card ↗

Visual Evidence Cards · Pivotal Subclaims

Pivotal subclaim illustrations

Dark-theme simulated-micrograph renders and concise causal evidence summaries for the pivotal subclaims that govern this review's grade. Click any card to inspect the master 2400×2400 render.

Master 2400×2400 render ↗

Master 2400×2400 render ↗

Master 2400×2400 render ↗

Figures

Explanatory figures summarizing this review. They carry no evidential weight of their own; where a figure and the review differ, the review governs.

Full assessment

From here the review is written for a technical reader: caps and thresholds, competing models, the independence basis, and the citation ledger.

HSEF v5.4 assessment — Red-Team Synthesis Edition (2026-09-28)

Original fresh assessment 2026-09-27 (authoring model: Codex/GPT agent). Cross-model red team and synthesis by Claude (Anthropic), 2026-09-28. Findings, dispositions and every change are itemised in redteam/REDTEAM_REPORT.md and redteam/SYNTHESIS_CHANGE_LOG.md.

Adversarial review: Authoring model of record: Sol 6.1. (This edition's own provenance records a Codex/GPT original assessment of 2026-09-27 with a Claude red-team synthesis pass of 2026-09-28; that record is preserved above and is not altered here.) Adversarially reviewed by Claude Opus 5.5 (Anthropic) on 2026-10-14. 5 corrections applied (see ADVERSARIAL REVIEW AND APPROVAL). Status: Approved as revised.

AT A GLANCE

Hypothesis: Psilocybin produces durable reduction in depressive symptoms through a drug-induced period of enhanced cortical plasticity.

Overall: Γ — Mechanistically plausible, partial system support. Confidence: Moderate.

Randomized trials support a reduction in depressive symptoms after supervised psilocybin treatment that outlasts the dosing day, with important primary-endpoint nulls (psilocybin versus escitalopram; the 2026 EPISODE trial) and with near-ceiling allocation guessing wherever blinding integrity has actually been measured. Months-long improvement is reported, although the strongest long-term reports have no untreated comparison group, allow additional treatment, or rest on exploratory outcomes. Adult mouse experiments directly demonstrate cortical remodeling and identify the pyramidal cell type and receptor required for it. The remaining causal gap is whether a measured, temporary increase in human cortical plasticity carries the durable antidepressant effect — and, one step earlier, whether such an interval has been measured in a living human brain at all. No reviewed human study measures cortical inducibility with onset and closure, and none manipulates the proposed mediator while leaving the drug in place.

The next discriminating result: a human treatment study that measures a cortical-plasticity time course and changes the proposed plasticity-dependent process independently of drug exposure, acute experience and supportive care, with a prespecified sustained depressive-symptom endpoint. Two trials now running use exactly the design shape the psychiatry rubric names as threshold-crossing — psilocybin given with and without a 5-HT2A/D2 antagonist, with plasticity and symptom readouts (ClinicalTrials.gov NCT06512220, NCT07490353) — but both are small, neither is complete, and the first is explicitly a feasibility study.

Plain-language summary

The hypothesis says that psilocybin briefly makes the brain's outer layer more able to change, and that this period of change produces lasting relief from depression. Clinical benefit has stronger evidence than this explanation of how it happens. Randomized trials show benefit in several settings, while other trials missed their main target, and long-term advantages are less certain. [5, 6, 2, 9, 4]

Almost everyone in these trials can tell whether they received psilocybin, and two recent analyses of pooled trial data suggest this matters for the results. In mice, researchers can watch new or enlarged connections appear and can test which cortical cells and receptors are needed. Human scans, electrical recordings and cognitive tests measure different things, each with limitations. A recent human brain-imaging study found no overall increase in a synaptic protein marker one week after treatment. [47, 48, 24, 32, 37, 41]

Taken together, the evidence makes the proposed mechanism plausible while leaving its causal contribution to lasting human symptom relief unestablished. The complete explanation is graded Γ with Moderate confidence: plausible, with real experimental support in other systems, but no reviewed human study shows that a temporary period of increased brain plasticity is what carries the sustained antidepressant effect.

Scorecard detail and subclaim matrix

Aggregation recorded before scoring: weakest-link for the conjunction of sustained human benefit, cortical plasticity, and plasticity-mediated benefit. The derivation is Ε, Δ, Ε, Γ, Γ → Γ. A primary-subclaim rule centred on the mediator (SC5) also returns Γ; scope-matched grading of the complete human mediated claim returns Γ. Grading protocol-level treatment efficacy alone would return Ε and would answer a narrower question; grading drug-specific pharmacological efficacy would itself be Γ-capped by the psychiatry rubric's functional-unblinding rule.

Role at the claimed human scope: causal role unestablished. The framework's required evidentiary label is epiphenomenon, because human causal dependency has not been demonstrated in either direction. This label records the present evidence status; it is not the result of an experiment establishing biological inactivity, and it does not apply to the mouse cortex, where dependency on the 5-HT2A receptor and on pyramidal-tract neurons has been demonstrated. The demonstrated animal cortical effects are assessed separately as SC3.

Literature coverage: 50 primary and synthesis reports: 23 clinical trial reports from 17 original clinical cohorts, 14 experimental/cellular reports, 10 human brain/target/proxy reports, and 3 quantitative syntheses bearing on comparator and blinding behaviour. In this red-team pass, 26 reports were read at primary full text — 5 complete and 21 partial (open-access or PubMed Central deposits, plus the openly available author version of one in-press study) — and 24 were verified at abstract, publisher-record and metadata level, which is a stricter accounting than the original edition's claim of 41 reports read at full text (see Reviewer notes and provenance). Corrections, precursor preprints, reviews, editorials, posters and protocols do not inflate this count. Reports from shared cohorts and research programs are explicitly linked. Company and regulatory documents describing the phase 3 program are reported as context and are excluded from the graded evidence base and from the ledger.

Claim restatement and rubric routing

Graded sentence: Psilocybin causes a sustained reduction in human depressive symptoms by inducing a temporary increase in cortical plasticity.

The review concerns treatment response, rather than a claim about what initially causes depression. It distinguishes (a) assignment to a psilocybin-containing clinical protocol, (b) post-acute symptom reduction, (c) persistence of that reduction, (d) cortical change, and (e) mediation by an interval of increased plasticity. The clinical literature generally combines drug dosing with preparation, supervision, support and follow-up; randomization estimates the effect of assignment within that protocol. It does not allocate causal credit to each component.

Rubric routing (corrected in this edition). The original edition recorded that "no psychiatry-specific rubric exists in this packet" and applied the universal framework alone. The packet does contain references/evidence-rubrics/07-psychiatry-neuromodulation.md, added in framework v4.0 precisely because "v3.0 shipped six rubrics and a psychedelic-therapy hypothesis got scored under neurodegeneration, where placebo response, functional unblinding, and subjective-experience mediators have no analogue." Its scope names "psilocybin dual-phase models … rapid-plasticity antidepressant models," so it is the governing field rubric here. The hypothesis index's assignment of the neurodegeneration rubric is an acknowledged v3-era mismatch and is not applied; neurodegeneration-specific pathology, seeding and progression requirements are not used. Routing to rubric 07 does not move the overall letter — it makes the Γ/Δ boundary argument explicit rather than improvised, and it supplies the field-valid designs that SC4 and SC5 still lack. The pre-registered scope and aggregation plan is otherwise preserved unchanged; the factual error in its rubric-availability sentence is logged, not silently edited.

The packet's closest index entry, "Psilocybin dual-phase (drug plus integration therapy)," additionally proposes that formal integration and subjective experience are required. Those are different hypotheses and are not added to the user's sentence.

Durability is reported by interval. Six weeks is meaningful post-acute benefit; twelve weeks is the prespecified benchmark for this review's durability subclaim; six and twelve months are stronger persistence tests. This benchmark is an analytical choice, not a universally agreed definition. Using a six-week definition would strengthen the durability clause but would leave the missing human mediator unchanged. A persistent new spine, a persistent circuit state, and an interval of increased capacity for further change are different measurements. Demonstrating one does not automatically measure the others.

Competing mechanistic models

The evidence most securely supports a context-dependent treatment response with unresolved mediators. The nominal plasticity-window explanation remains a credible candidate. A lasting change in circuit function is another candidate, and it partly overlaps the nominal model when "plasticity" is used broadly. The distinguishing prediction is a measurable interval of increased inducibility whose manipulation changes subsequent clinical benefit. Tournament rank does not assign causal primacy, and the tournament did not move the letter.

Expectancy and functional unblinding are an inference audit, and in this edition they are an evidenced one. Wherever blinding integrity has been measured it has failed: 86% of EPISODE participants correctly identified 25 mg psilocybin; 13 of 13 psilocybin and 7 of 10 placebo participants guessed correctly in the NHS trial, whose authors state plainly that "blinding was not upheld"; every participant in the clinician trial knew and was correct by the end of the session; in the Swedish trial the blinded rater correctly guessed 12 of 17 psilocybin and 15 of 17 niacin allocations at day 8, and by day 365 16 of 17 psilocybin participants and 16 of 16 niacin participants guessed correctly. [2, 1, 20, 4]

Two quantitative syntheses turn that observation into a testable comparison. Comparing psychedelic-assisted therapy with open-label conventional antidepressant treatment — equalising the knowledge of allocation rather than the pill — Williams and colleagues found no advantage for psychedelic therapy (8 trials, 249 patients versus 16 open-label antidepressant trials, 7,921 patients; estimated difference 0.3 favouring open-label antidepressants, 95% CI −1.39 to 1.98, P=.73), while open-label antidepressant treatment outperformed blinded antidepressant treatment (difference 1.3, 95% CI 0.07 to 2.51, P=.04) and no equivalent blinding difference appeared for psychedelic therapy (0.67, 95% CI −3.08 to 1.73, P=.58). Hieronymus and colleagues approached the same problem from the control arm: across 17 trials, pre-to-post standardized mean change in control arms was 0.50 (SEM 0.15) in psilocybin trials versus 1.00 (0.08) in SSRI and 1.12 (0.17) in esketamine trials, with study population a significant moderator of control-arm change (QM 10.4, P=.005) but not of active-arm change (QM 1.21, P=.55), and control-arm response rates 14 and 23 percentage points lower than in SSRI and esketamine trials respectively. A third synthesis found the pooled psilocybin effect larger when depression was secondary, when outcomes were self-rated, and in participants with prior psychedelic use (Hedges' g=0.66, 95% CI 0.46 to 0.86 overall; 0.88, 0.42 to 1.33 for secondary depression and for self-report scales; metaregression coefficient 6.00, 95% CI 2.48 to 9.53 for previous psychedelic use). [47, 48, 49]

These are Tier B syntheses: they frame the inference and are not treated as primary experimental evidence, and none of them shows that psilocybin's antidepressant effect is only expectancy. Symmetry requires the counter-datum: EPISODE's own descriptive analysis, stratifying week-6 outcomes by unblinding status, suggested that functional unblinding of the first dose "played a negligible role in depression outcomes at week 6," while the same paper notes that raters' guesses and participants' expectations were never measured. The audit therefore constrains what repeated randomized trials can be counted for — they are not independent causal lines when a single systematic error could produce all of them — without resolving the question. [2]

HYPOTHESIS STATE EVALUATION FRAMEWORK (HSEF v5.4)

1. Human depressive-symptom benefit: several positive trials, with meaningful contrary results

The early open-label treatment-resistant depression study and its expanded six-month report established feasibility and documented prolonged improvement without a concurrent untreated comparator: QIDS improved by −11.8 points at one week (95% CI −9.15 to −14.35, p=0.002, Hedges' g=3.1) and −9.2 points at three months (95% CI −5.69 to −12.71, p=0.003, g=2), and in the expanded cohort nine of nineteen completers met response and four met remission criteria at five weeks, with within-cohort effect sizes of about 1.5 at three months and 1.4 at six months. Davis and colleagues then used immediate versus delayed treatment in major depression; in the corrected report, 71% of the overall treated sample met response criteria at week 1 and at week 4, with remission in 58% and 54%. Gukasyan's twelve-month follow-up concerns the same treated cohort: GRID-HAMD fell from 22.8 at baseline to 7.7 at twelve months, with 75% response and 58% remission, while eight of twenty-four participants (33%) began a new antidepressant during follow-up. These reports contribute different time points, not independent clinical replications. [13, 14, 11, 12]

The active-comparator trial against escitalopram randomized 59 patients. Its primary six-week QIDS-SR-16 change was −8.0±1.0 points with psilocybin and −6.0±1.0 with escitalopram, a between-group difference the trial reports as 2.0 points (95% CI −5.0 to 0.9), P=0.17 — a null on the trial's own prespecified primary outcome, and not an equivalence result. Response occurred in 70% versus 48% (difference 22 percentage points, 95% CI −3 to 48) and remission in 57% versus 28% (28 points, 95% CI 2 to 54), but these secondary outcomes were not corrected for multiplicity and remain exploratory. The six-month observational follow-up preserves information about later course while subsequent care and attrition complicate attribution: 37 of 59 participants (62.7%) received additional medication, psychotherapy or psychedelics, and missingness was unequal between arms (average 13% for psilocybin and 26% for escitalopram, reaching 38% in the escitalopram arm at six months). [9, 10]

Goodwin's 233-person treatment-resistant depression trial found a week-three MADRS advantage for 25 mg over 1 mg of −6.6 points (95% CI −10.2 to −2.9, P<0.001), while the 10 mg versus 1 mg comparison was not significant (−2.5, 95% CI −6.2 to 1.2, P=0.18); the formal testing hierarchy stopped there, and the trial reports that the incidences of response and remission at three weeks, "but not sustained response at 12 weeks," were generally supportive of the primary result. Raison's 104-person trial against niacin found a primary day-43 MADRS advantage of −12.3 points (95% CI −17.5 to −7.2, P<.001) with independent central blinded raters, a day-8 advantage of −12.0 points (95% CI −16.6 to −7.4), a Sheehan Disability Scale difference of −2.31 points (95% CI −3.50 to −1.11), and more participants meeting sustained response — but not sustained remission — with psilocybin. Neither trial reported a blinding-integrity check, which the psychiatry rubric treats as disqualifying for the drug-specific inference. Together they support post-acute clinical response at protocol scope without selecting a cortical mediator. [6, 5]

The 2026 EPISODE trial is an important boundary. Among 144 randomized patients (143 dosed; 142 in the primary analysis), the primary week-six HAMD-17 response endpoint was 17.0% with 25 mg versus 12.5% with 5 mg and 10.6% with nicotinamide; adjusted OR 1.73, 95% CI 0.53–6.23, P=.19, so confirmatory hierarchical testing stopped and all later comparisons are exploratory. The key secondary continuous outcome favoured 25 mg (HAMD-17 difference −4.60, 95% CI −7.01 to −2.18, P<.001). After the second administration at week six all groups had received at least one 25 mg dose, so later follow-up has a different estimand from the initial placebo-controlled comparison. The NHS feasibility trial in 60 patients met its recruitment and retention targets and reported a week-three adjusted MADRS difference of −10.41 points (95% CI −14.86 to −5.95, Cohen's d=−1.70) sustained at week six (−12.92, 95% CI −17.37 to −8.47, d=−2.11) — but its primary outcomes were feasibility parameters, not confirmatory efficacy, and its authors attribute part of the difference to expectancy in both arms given the failed blind. Both findings belong in the review. [2, 3, 1]

Other primary reports add dose, medication, setting and population boundaries: the Swiss placebo-controlled trial (MADRS remission in 14/26, 54%, versus 3/26 on placebo at day 14, P=0.0023), a fixed-order exploratory trial in which depression improved after both placebo and psilocybin with no significant difference between conditions (response 66.7% and remission 46.7% after psilocybin; within-condition effect sizes larger after psilocybin, g′ 1.02–2.27, than after placebo, 0.65–0.99; mystical-experience intensity uncorrelated with subsequent antidepressant effect), an open-label concomitant-SSRI study (MADRS −14.9 points, 95% CI −20.7 to −9.2, with 8 of 19 participants, 42.1%, both responding and remitting), a trial in frontline clinicians (day-28 MADRS −12.00, 95% CI −17.67 to −6.33, P<.001, with 100% correct allocation identification by session end), a life-threatening-illness RCT (HADS depression B=−2.49, P=.02, d=1.12; BDI-II B=−7.56, P=.004; open-label extension for all participants afterwards), and a psilocybin-plus-CBT pilot (10 of 16 responders, 62.5%, and 9 of 16 remitters, 56.3%, maintained three months after a four-month combined treatment). No clinical paper in this evidence set selectively tested cortical plasticity as the mediator of depressive-symptom relief. [15, 22, 8, 20, 23, 21]

Phase 3 context, reported and excluded from the graded base. The largest randomized psilocybin datasets in depression are the two Compass Pathways phase 3 trials, described so far only in company investor releases, a poster and a regulatory filing: COMP005 (258 dosed; 25 mg n=171 versus placebo n=87) reported a week-six MADRS difference of −3.6 points (95% CI −5.7 to −1.5, p<0.001), and COMP006 (581 dosed; 25 mg n=296, 10 mg n=142, 1 mg n=143) reported −3.8 points for 25 mg versus 1 mg (95% CI −5.8 to −1.8, p<0.001), with continued separation described through week 26 in a period in which 58% of the 25 mg arm was re-treated and protocol-allowed antidepressants were permitted; a rolling new-drug application was under way in 2026. Under the framework's source-tier rule, press releases and investor materials are inadmissible as evidence items, so these results carry no part of the grade and appear in no ledger row. They are reported here because omitting the largest randomized datasets would misinform a reader, because their week-six point estimates are markedly smaller than the peer-reviewed phase 2 estimates above, and because their eventual peer-reviewed publication is a named trigger that could move SC1 and SC2 in either direction. Company documents cited for context only: Compass Pathways investor releases of 17 February 2026 and 7 July 2026, the 7 July 2026 SEC 8-K exhibit, and the September 2026 52-week Part C topline report.

2. Durability: evidence becomes less controlled as follow-up lengthens

Cancer-associated depression/anxiety trials by Grob, Ross and Griffiths provide independent clinical contexts. Crossover designs allowed controlled early contrasts — 83% versus 14% BDI response at seven weeks before crossover in the NYU trial, and 92% versus 32% GRID-HAMD response after the first session in the Johns Hopkins dose-sequence trial — but later measurements generally followed active treatment in both sequences, and the Johns Hopkins six-month figure (78% response) has no untreated comparison behind it. The UCLA pilot reported reduced trait anxiety at one and three months and a BDI improvement reaching significance at six months in unblinded follow-up. Agin-Liebes' multi-year follow-up assessed 15 of the 16 surviving participants from the NYU trial at a mean of 3.2 and 4.5 years, with approximately 60–80% meeting criteria for clinically significant antidepressant or anxiolytic response; selective follow-up, self-report and the intervening clinical course constrain causal attribution. These studies establish important observations about persistence rather than years-long randomized proof of a drug-specific effect. [19, 16, 18, 17]

The 2026 Swedish randomized trial followed 35 people for a year. Clinician-rated MADRS favoured psilocybin on day 8 (−7.27, 95% CI −12.89 to −1.65, P=.01), day 15 (−11.03, 95% CI −16.65 to −5.42, P<.001) and day 42 (−8.33, 95% CI −13.94 to −2.71, P=.004). On day 365 the contrast was not significant (−3.68, 95% CI −9.30 to 1.94, P=.20). Monthly self-ratings favoured treatment through day 102 (−6.60, 95% CI −13.01 to −0.19, P=.04), while a sensitivity analysis censoring participants at antidepressant initiation supported separation only through day 72. These are qualified, exploratory signals around the twelve-week benchmark; they do not provide a confirmatory year-long advantage. [4]

Longer observational extensions include the Johns Hopkins cohort, the escitalopram-comparator trial, the Compass observational extension and EPISODE. The Compass extension (COMP004) enrolled 66 participants from two earlier trials and reported median times to a prespecified depressive event of 92, 83 and 62 days for the original 25 mg, 10 mg and 1 mg cohorts, with most COMP001 participants having experienced a depressive event by week 12 (25 mg 37/75; 10 mg 38/79; 1 mg 44/75). EPISODE's naturalistic follow-up retained 126 of 144 participants for at least one visit and found HAMD-17 still below baseline at six months (−7.93, 95% CI −9.17 to −6.70) and twelve months (−7.74, 95% CI −9.04 to −6.43) with no significant group differences, while sustained response across groups fell from 28/126 (22.2%) at six months to 20/126 (15.9%) at twelve months and re-initiating antidepressant pharmacotherapy was strongly associated with higher HAMD-17 scores (β=3.79, 95% CI 1.98 to 5.60). Covariate adjustment for later treatment does not recreate randomization of that treatment, and the psychiatry rubric is explicit that "a positive open-label result never crosses Δ, however large." The exploratory, randomized self-report separation through day 102 in the Swedish trial is what carries SC2; its rescue-sensitive analysis and the unconfirmed twelve-week sustained-response endpoint in the largest peer-reviewed trial make that support fragile. Thus SC2 reaches Δ with Low confidence, with stronger support for several weeks than for six months or a year. [4, 6, 12, 10, 7, 3]

3. Experimental cortical plasticity: strongest in a defined adult mouse system, with an independent functional line in rats

Shao's 2021 study followed cortical dendritic spines before and after psilocybin in adult mice, finding increased spine density (+7±2% at day 1, +12±3% at day 7), larger spine heads (+11±2% at day 1), a higher formation rather than lower elimination rate, and increased miniature excitatory current frequency at 24 hours (P=0.0002). About half of the new day-1 spines were still present at day 7, and 34±10% (females) and 37±12% (males) persisted to day 34. The same paper contains the most-cited receptor result in this literature and one the original edition omitted: pretreatment with ketanserin at a dose that completely abolished the head-twitch response did not abolish structural remodeling — spine density increases were no longer statistically significant (+5±2% day 1, +8±2% day 7, main effect P=0.09) while spine head width (+12±1%, P=0.01), protrusion length and formation rate remained elevated. The authors are careful about what this licenses: 1 mg/kg ketanserin blocks only about 30% of rodent cortical 5-HT2A receptors, so the experiment shows that a partial knockdown sufficient to remove head-twitch does not remove remodeling, not that remodeling is 5-HT2A-independent. [24]

The 2025 study separates pyramidal tract (PT) and intratelencephalic (IT) cell types, deletes 5-HT2A receptors in adult cortex, and manipulates cell activity during dosing. Psilocybin increased spine density in both cell types and the elevation remained significant at the final imaging session on day 65; silencing PT neurons eliminated psilocybin's effect on stress-related behaviour while silencing IT neurons did not; and targeted adult knockout of cortical 5-HT2A receptors abolished both the behavioural effect and PT structural remodeling. This is the strongest causal chain in the corpus — and it is a mouse chain. Increased spine density at day 65 measures persistence of remodeled structures; it does not establish a 65-day period of enhanced inducibility. The receptor-control learned-helplessness arm had a floor effect that the authors flag explicitly. [32, 24]

A receptor deletion affecting both remodeling and behaviour can act through parallel downstream routes; selective interruption of spine formation after dosing, with the acute drug action preserved, would discriminate the proposed mediator more directly. Head-twitch behaviour is a rodent pharmacological readout and cannot establish human subjective-experience requirements — the rubric classes it as pharmacodynamic, not therapeutic. Jiang's activity-dependent rewiring shows that the pattern of input reorganisation depends on spiking during drug exposure: silencing retrosplenial cortex during dosing prevented the subsequent gain in retrosplenial input to PT neurons, and the brain-wide pattern contained losses as well as gains. That work is from the same research program. [32, 33]

Two adult-rat reports from other groups now supply an independent line, and it points to function rather than persistent structure. Kramer and colleagues report that single doses of psilocybin or of the selective 5-HT2A agonist 25CN-NBOH reduced forced-swim immobility with no decline in effect size for at least three months, with changes in resting membrane potential, firing rate and synaptic excitation in layer 5 medial prefrontal neurons months later, and no differences in spine density, spine classification, or pre- and postsynaptic marker gene expression. Purple and colleagues, recording with chronically implanted probes in freely behaving adult rats, found acute high-frequency oscillations most pronounced in infralimbic cortex lasting about an hour with decreased firing rates and reduced spike-train complexity, followed by gradually emerging increases in beta and low-gamma local field potential power at one, two and six days. Neither study measured depressive symptoms in humans, and forced-swim immobility is a coping proxy rather than a model of human depression. [36, 50]

Hesselgrave reports behavioural and hippocampal synaptic effects under ketanserin in stressed mice; Cameron's constitutive receptor knockout reports that 5-HT2A activation is required for tryptamine psychedelics' antidepressant-like effects; Sekssaoui reports that psilocybin retained behavioural efficacy for up to 15 days in constitutive 5-HT2A knockout mice while DOI and lisuride did not. Doses, receptor occupancy, developmental compensation, brain region, stress preparation and assays differ, and absence of a head-twitch response is insufficient evidence that all cortical 5-HT2A signaling was blocked. These are informative mechanistic contrasts rather than a clean resolution of one necessary pathway; the adult, region- and cell-type-restricted deletion is the more interpretable design. [25, 27, 31, 32]

Raval's independent pig study measured SV2A and receptor binding in cortex and hippocampus (hippocampal SV2A +4.42% at day 1; +9.24% hippocampus and +6.10% prefrontal cortex at day 7, with acutely reduced 5-HT2A density), but used juvenile animals, a presynaptic protein surrogate and ex vivo autoradiography. Nardou measures reopening of social-reward learning in adult mice, prominently involving nucleus-accumbens metaplasticity: the open state was present one and two weeks after psilocybin and closed by three weeks, with the duration across drugs tracking the duration of their acute subjective effects in humans. That is the closest thing in the corpus to a measured plasticity window — in a non-cortical reward-learning assay, in mice, with no depressive endpoint. Schmidt's eLife research article provides a human-cellular bridge: psilocin increased BDNF abundance, neuronal complexity, synaptic-marker colocalisation and functional excitability in iPSC-derived human cortical neurons, with BDNF induction reproduced in four independent control lines — at 10 µM, a concentration and duration far from clinical exposure, in a developmental culture system. Vargas' intracellular-receptor experiments inform cellular mechanisms; their compounds and preparations must be kept distinct from direct psilocybin evidence in living humans. [26, 30, 34, 28]

The adult experimental cortical response therefore receives Ε, Moderate confidence. Multiple modalities within one laboratory program strengthen the structural claim, the two independent rat reports strengthen the functional claim while contradicting persistent structural excess at three months, and juvenile pig marker data, developmental cultures and assay-noncomparable models constrain an audited Ζ claim across independently verified adult systems. Under the framework's independence test, the direct adult structural imaging line and its receptor/cell manipulations share a program and a technique: a single systematic error could produce both, so they are one line, however many papers they span.

4. Molecular route and persistent physiology: plausible branches, with unresolved specificity

Moliner reports that LSD and psilocin bind TrkB with affinities about 1,000-fold higher than those of conventional antidepressants, at a transmembrane site partly overlapping the antidepressant site, and that the plasticity and antidepressant-like effects of psychedelics in mice depend on TrkB binding and endogenous BDNF signaling while LSD-induced head twitching depends on 5-HT2A. The in vivo genotype dissociation and the prolonged behavioural benefits in that paper concern LSD, so the psilocybin-specific evidence is binding and cell-level. Jain's broad profiling study screened 41 psychedelics against 318 human GPCRs and LSD against more than 450 human kinases, documenting extensive polypharmacology; the specific claim that the kinase panel showed no potent LSD–TrkB interaction could not be verified at the access depth available for this pass and is therefore not used as counter-evidence to Moliner. In any case a kinase-domain panel is not an assay-comparable test of a proposed transmembrane allosteric interaction with the full-length receptor, and neither study establishes a mediated clinical pathway in depressed humans. [29, 35]

The independent adult-rat dissociation described above — months-long behavioural and cortical electrophysiological change with no excess spine density or synaptic-marker expression at twelve weeks — constrains a model that requires persistent excess spines, while remaining compatible with an earlier transient growth phase and with functional plasticity. [36, 50]

5. Human brain measurements: distinguish target engagement, state change, structural markers, and inducibility

Madsen's PET study in eight healthy volunteers relates psilocin exposure to neocortical 5-HT2A occupancy (up to 72%) and to subjective intensity, verifying human target engagement during the acute state. Carhart-Harris' post-treatment fMRI in the open-label TRD cohort related reduced amygdala blood flow to immediate symptom improvement and some post-acute connectivity measures to five-week response. Daws' later analysis reported that reduced fMRI network modularity correlated with symptom improvement in both psilocybin cohorts and that no such change was seen with escitalopram; the paper's inference rests on within-arm change and change–symptom correlation, and it reports no significant psilocybin-versus-escitalopram interaction for modularity. The specific interaction statistic cited in the original edition could not be verified from a primary source at this pass's access depth and has been removed rather than repeated. A significant within-arm result alongside a non-significant result in another arm does not establish a difference between arms, and none of these measures counts synapses. [43, 40, 39]

Doss found cognitive flexibility improved for at least four weeks after treatment in 24 depressed patients, with no correlation to antidepressant change (all |r| < 0.2, all p > 0.250), reduced anterior cingulate glutamate (t(19)=2.54, p=0.020) and N-acetylaspartate (t(19)=3.05, p=0.007) at one week, and — counter-directionally — greater increases in anterior-to-posterior cingulate dynamic connectivity associated with less cognitive improvement. This does not show that plasticity is irrelevant; it shows that this cognitive measure cannot be assumed to carry the antidepressant effect. Siegel's precision-imaging study in seven healthy participants found acute desynchronization more than threefold larger than methylphenidate's, with whole-brain architecture largely returning to baseline while anterior hippocampus–default-network connectivity remained reduced for about three weeks; the authors state that the proposed antidepressant mechanism requires studies in patients. Repeated scanning increases within-person information, not the number of independent participants. Barrett similarly documents a mixture of persisting and returning-to-baseline affective and imaging outcomes in twelve healthy volunteers. [41, 42, 44]

Lyons' 2026 multimodal study in 28 psychedelic-naive healthy participants found acute EEG signal-entropy increases and, one month after 25 mg, reduced axial diffusivity in prefrontal–subcortical tracts (prefrontal–striatal p=0.006; prefrontal–thalamic p=0.005), together with a significant reduction in network modularity relative to baseline (p=0.02) and a correlation between modularity change and well-being change in the 25 mg versus 1 mg contrast (r=−0.4, p=0.04, two-tailed); a tract-based spatial statistics analysis was null, and enduring functional changes were otherwise described by the authors as largely absent. Its design was fixed order, with 1 mg always preceding 25 mg rather than randomized counterbalancing, so practice, habituation and expectation cannot be excluded — the authors say so. Axial diffusivity has several possible biological determinants, and EEG–insight–well-being statistical mediation is not a selective test of cortical plasticity carrying durable relief of depression. [38]

The 2026 Johansen study is the closest structural-marker test in this set and the most informative about SC4. Fifteen healthy participants received 0.3 mg/kg psilocybin and underwent [11C]UCB-J SV2A PET at baseline and one week. In the preregistered primary analysis — the twelve participants with complete arterial-input data — frontal-cortex binding rose by 2.3% (one-sided p=0.29) and hippocampal binding by 3.0% (p=0.33); in the fifteen-participant reference-region analysis the changes were −0.03% (p=0.54) and +1.01% (p=0.44). Recruitment stopped at an interim futility analysis (conditional power 24% at the hypothesised effect size, 2% at the observed one). Dosing setting was assigned sequentially and not randomly: participants 1–5 were dosed inside the scanner and 6–15 in a therapeutic-like room. In exploratory analyses the room subgroup showed 5–6% increases whose own confidence intervals included zero (frontal ΔV_T 1.1 mL/cm³, 95% CI −0.5 to 2.7; hippocampal 0.8, 95% CI −0.5 to 2.1), while the between-setting difference reached nominal significance frontally (3.0 mL/cm³, 95% CI 0.4 to 5.5; likelihood-ratio p=0.03) but not in hippocampus (1.9, 95% CI −0.3 to 3.9; p=0.09). Change in SV2A binding was not significantly related to three-month persisting effects (p>0.37). The setting interaction warrants a randomized test; it does not replace the overall primary null. One time point in healthy participants cannot exclude an earlier transient window, other forms of plasticity, or a different effect in depression — a limitation the authors state themselves. [37]

Peripheral BDNF is an incomplete bridge. Holze's randomized, placebo-controlled crossover in 28 healthy participants found that neither psilocybin dose nor either LSD dose altered plasma BDNF relative to placebo. Becker's crossover reported that psilocybin significantly raised peak plasma BDNF from baseline under both escitalopram (t22=5.4, P<0.001) and placebo pretreatment (t22=3.6, P=0.002) — but both conditions received psilocybin, so a pretreatment placebo is not a psilocybin placebo. These results do not provide a reliable substitute for cortical measurement. [45, 46]

What is missing is not a detail. A targeted search of the human literature and of trial registries for psilocybin studies using validated inducibility measures — paired associative stimulation, TMS-EEG plasticity protocols, visual or somatosensory LTP-like potentiation, or any assay that probes the capacity of cortex to change — returned no completed study. SC4 is not held at Γ because the human window is small or contested; it is held at Γ because it has not yet been measured.

6. Caps, precedence, and the Β/Γ/Δ boundary under the psychiatry rubric

The universal constraints are applied to the inference being made, and the psychiatry rubric supplies the field-specific hard rules:

• Universal: "In vitro evidence alone cannot exceed Γ." Human-derived cultured neurons provide a cellular bridge, not endogenous clinical mediation.
• Universal: "Pure association evidence — however large, however replicated — cannot exceed Β." An imaging–symptom association remains associational on its own; the combined mechanistic case can reach Γ through additional experimental evidence.
• Universal: "Biomarker change without functional/phenotypic change caps at Γ." A marker, peripheral protein or network metric cannot establish a sustained clinical mechanism merely by moving after dosing.
• Rubric 07, mediation-by-correlation rule: "Candidate mediator correlated with outcome, never manipulated." This fires on every human mediation analysis in the set — modularity, EEG entropy and insight, cognitive flexibility, SV2A change — and it is the rule that holds SC5 at Γ.
• Rubric 07, functional-unblinding rule: "Inert-placebo trial of an intervention with unmistakable acute subjective effects — psychedelics, ketamine, entactogens, suprathreshold stimulation, ECT — with no blinding-integrity check reported. Such a trial did not control what it claims to control." This fires on the largest peer-reviewed positive trials (Raison; Goodwin 2022; von Rotz; Griffiths; Ross; Davis) and caps the drug-specific pharmacological inference at Γ even where protocol-level response reaches Ε.
• Rubric 07, open-label rule: "A positive open-label result never crosses Δ, however large." This governs the long follow-ups and extensions that carry most of the twelve-month narrative.
• Rubric 07, acute-pharmacodynamic rule: "Acute molecular, occupancy, or imaging change with no link to symptom trajectory." This governs occupancy and acute desynchronization results offered for SC4.
• Rubric 07, rodent-proxy rule: "Forced-swim, tail-suspension, sucrose-preference, or head-twitch alone; head-twitch is pharmacodynamic, not therapeutic." This governs the rodent behavioural endpoints offered as antidepressant evidence, including the three-month forced-swim persistence.

Four of these rules bind an inference class rather than a subclaim as scored — the functional-unblinding rule on drug-specific attribution within SC1, the open-label rule on the long-term evidence class within SC2, the rodent-proxy rule on behavioural endpoints within SC3, and the in vitro and association caps on their respective components. They are recorded in the scorecard as component caps precisely so that a component ceiling is not mistaken for a subclaim-wide ceiling, which is the error the framework's cap-precedence rule exists to prevent.

A clinical randomization can cross Δ for protocol-level treatment response, and an adult mouse drug/receptor experiment can cross Δ for its cortical response. The governing scope rule is: "A perturbation crosses a threshold within the system it was actually run in"; a universal cap governs the inference from that system to the claimed scope. The two rules bind at different steps, and the collisions are recorded rather than resolved silently: the mouse cell-type and receptor manipulations cross Δ within mouse cortex while the cap blocks their export as an established human symptom mediator, and the association cap binds the isolated imaging–symptom inference without being re-applied to the convergent mechanistic set (framework pitfall 21).

For SC4, the reviewed human measurements do not positively demonstrate a temporary increase in cortical inducibility with onset and closure; the one preregistered human synaptic-marker test returned a null and stopped for futility, which under framework pitfall 17 leaves the threshold uncrossed rather than crossing it by virtue of design. For SC5, no reviewed study independently changes the proposed cortical-plasticity process and demonstrates a sustained depressive-symptom consequence. Their first unmet threshold is Δ at the claimed human scope, where rubric 07 defines the qualifying designs as an acute-effect-matched comparator with measured blinding integrity, blinded on/off or randomized-withdrawal neuromodulation, mediator manipulation with the intervention intact, a human depletion or challenge experiment with symptom trajectory as the outcome, or psychotherapy-component dismantling with blinded assessors. A statistically positive clinical treatment result supplies none of these. Necessity plus sufficiency would be an Ω-level demand; demonstrating some causal contribution is the immediate Δ target.

7. Upper ladder, falsification, and interpretive pitfalls

The upper ladder was considered using the field-valid equivalents rubric 07 names — blinding-integrity-reported randomized trials, discontinuation and rechallenge designs, and genetic instruments where a plausible one exists — plus selective adult animal genetics, dose-response, longitudinal brain measurement, and natural-experiment or instrumental-variable designs. Blinding-integrity-reported trials exist and have positive primary endpoints (the Swedish trial; the clinician trial), but their comparators are not matched for acute subjective effects and the measured guessing is near-ceiling, so they do not separate the drug from its package; discontinuation and rechallenge cannot be applied to a single-dose intervention whose proposed mediator is a transient window; no plausible genetic instrument for drug-induced cortical inducibility was identified. A direct randomized experiment on human synapses is not the only possible route upward — but at present no independent human genetic, natural-experiment, pharmacological or timing instrument in the reviewed evidence separates the plasticity process from the other treatment paths convincingly enough to establish the mediated claim.

The important contrary patterns are narrow and specific: an overall one-week human SV2A null that stopped for futility; cognitive improvement dissociated from depressive improvement; an unconfirmed twelve-week sustained-response endpoint in the largest peer-reviewed trial; no clear one-year clinician-rated advantage in the Swedish trial; a null primary endpoint against escitalopram and a null primary response endpoint in EPISODE; persistent rat cortical physiology with no persistent excess spines or synaptic-marker expression; and pooled analyses showing that psychedelic therapy does not outperform open-label antidepressant treatment while psilocybin-trial control arms improve far less than control arms in SSRI and esketamine trials. Each constrains a particular prediction. None, by itself, eliminates all cortical functional plasticity or all clinical benefit. Rubric 07's falsification pattern "allocation guessing is near-perfect and outcome tracks guessed allocation" is half-met: the guessing is near-perfect, and whether outcome tracks it is contested — EPISODE's own stratified description suggests it did not at week six. The pattern "blocking the acute experience leaves the effect intact" has been tested only in rodents, with mixed results, and is precisely what the two ongoing human antagonist trials will address. [37, 41, 6, 4, 9, 2, 36, 47, 48]

The principal pitfalls flagged here are therapeutic effect treated as mechanistic proof; a biomarker substituted for the mediator; persistence substituted for a plasticity window; "neuroplasticity" used as an unfalsifiable bridge from an acute molecular event to a distant clinical outcome; subgroup or secondary results replacing the primary outcome; nonrandom setting differences interpreted causally; rodent behavioural-despair assays treated as models of human depression; a conjunctive claim scored on its easier clause; and multiple reports from one cohort or one laboratory program counted as independent replication.

8. Steelman, causal tally, and final adjudication

The case one rung higher, Δ: randomized treatment benefits in several populations, direct adult cortical remodeling with cell-type and receptor specificity, an adult-mouse demonstration that the behavioural effect requires cortical 5-HT2A receptors and pyramidal-tract neurons, a measured two-week reopening of a learning-sensitive state in adult mice, human target engagement at known occupancies, and human imaging changes that correlate with symptom improvement. Such triangulation makes a causal contribution by plasticity credible and gives a clear experimental target. Taken at its strongest, the hypothesis is not speculative: it names a molecule, a receptor, a cell type, a time course and a measurable readout, and each of those has been confirmed somewhere.

Why Δ is not yet assigned to the full claim: the links are established in different systems with different outcomes. The animal manipulations can affect several downstream routes; human symptom-associated scans measure proxies; the one human synaptic-marker test was null; and no human study manipulates the proposed mediator while leaving the drug in place. The missing threshold is a positive, sufficiently selective human test connecting a demonstrated cortical-plasticity interval to sustained symptom benefit. A necessity experiment is one powerful option, not the only allowed causal design.

The case one rung lower, Β: human mediator attribution is entirely correlational, the clinically relevant proxy results are mixed or null, two primary endpoints in well-powered trials were missed, and pooled analyses are consistent with a large expectancy contribution. Why Β is too low for the complete evidence base: experimental drug-induced cortical remodeling with adult, cell-type-restricted receptor deletion is a causal perturbation in a living mammalian cortex, and pharmacological/cellular intervention findings provide partial system support well beyond association. Thus the aggregate is Γ — Mechanistically plausible, partial system support, with Moderate confidence. Confidence is Moderate rather than Low because the evidence base is substantial, multi-modal and independently replicated across clinical and experimental lines, and rather than High because human measurement is heterogeneous, mediator identification is absent, and the largest randomized datasets are not yet peer reviewed. Confidence reflects evidence density and reproducibility; this reviewer's access depth is reported separately in the reading-depth and provenance notes and is not a basis for the band. The grade rests on peer-reviewed work; the phase 3 company reports carry no part of it.

Audited tally: zero reviewed studies demonstrate the complete human chain of an experimentally established cortical-plasticity interval and its causal contribution to sustained depressive-symptom relief. This is a tally of the exact mediated claim, not of all positive clinical or animal experiments, of which there are many. Recounted against the cap rules in this pass, the tally remains zero: the clinical trials do not randomize the mediator, the animal manipulations are not human, the human imaging and marker studies do not manipulate anything, and the one human study that measured a synaptic marker under a preregistered primary analysis returned a null. The ledger retains each report's system, result, access depth and overlap.

9. Role, Momentum, Stability, and Divergence

At human clinical scope, the causal role of the proposed plasticity period is unestablished. The framework's direction procedure stops at step 2, co-occurrence without demonstrated dependency in either direction, giving epiphenomenon. It cannot presently justify "primary driver" or "permissive factor" at that scope. Animal cortical effects — where dependency has been demonstrated — and the clinical protocol's symptom effects receive their own scope-specific scores.

Momentum: Decisive work underway. Two registered human trials administer psilocybin with and without a 5-HT2A/D2 antagonist while measuring plasticity and symptoms, which is the "mediator manipulation with the intervention intact" that rubric 07 lists as threshold-crossing. SYNVEST (NCT06512220, Centre for Addiction and Mental Health, recruiting since April 2025, n=12, primary completion April 2027) gives 25 mg psilocybin with and without risperidone 1 mg to adults with treatment-resistant depression, with [18F]SynVesT-1 SV2A PET before and four weeks after dosing, TMS-EEG cortical-plasticity measures, diffusion MRI and MADRS; its declared primary outcome is feasibility, and n=12 will not settle mediation. A Vienna trial (NCT07490353, recruiting since November 2025, n=85, primary completion May 2028) uses a psilocybin-with-and-without-risperidone crossover in depression with anhedonia plus healthy controls, with fMRI complexity measures, BDNF and clinician- and self-rated depression scales. Both are prospective and neither has results. An earlier SV2A study inside a psilocybin depression trial (NCT05601648) was withdrawn with zero enrolment, which is itself informative about how hard this measurement is. Named mechanistic artifacts continue to accumulate: adult cell-type and receptor manipulations, activity-dependent rewiring, human iPSC cortical neurons, human SV2A PET, and independent adult-rat electrophysiology.

Stability: Uncertain. Named triggers point in both directions. Upward: a positive, adequately powered human inducibility or blockade result linking a measured plasticity interval to sustained symptoms (SYNVEST, NCT07490353, or a successor) would cross Δ for SC4 and potentially SC5; peer-reviewed publication of the phase 3 program with blinding-integrity and durability analyses would firm up SC1 and SC2. Downward: adequately powered nulls in those designs, or phase 3 publications whose modest week-six effects and re-treatment-dependent durability are confirmed, would weaken the durability clause and leave the mediator unsupported. General efficacy trial programs alone do not decide the plasticity attribution in either direction.

Divergence: Contested. Primary papers support different receptor and plasticity routes, and recent human authors explicitly frame plasticity as a hypothesis while reporting substantial measurement limits. There is no defensible single field-implied HSEF level from these artifacts, so no arithmetic gap is manufactured. Read literally as settled human causation, the hypothesis sentence would overstate what has been demonstrated; read as a mechanistic candidate with strong animal support, it is fairly stated. M/S/D do not change the scientific score.

Proposed decisive experiments

1. Separate structural growth from acute cortical drug action in adult animals. Interrupt a validated plasticity effector after acute dosing, preserving drug exposure, acute cortical activation, general learning and motor function. Measure direct spine formation and cortical physiology alongside multiple sustained stress-related behavioural outcomes, and include the functional readouts that the independent rat work found to persist without structural excess. Add a selective rescue and independent laboratory replication. Behavioural benefit persisting despite verified loss of the proposed plasticity process would weaken its mediator role; selective loss and rescue of both would strengthen it within that model. A general receptor deletion or a broad mTOR inhibitor alone would retain alternative pathways.

2. Map a human window with a validated inducibility assay, then randomize its use. In a prespecified depressed population, compare psilocybin with a credible active control under matched support, measuring cortical plasticity with a validated inducibility protocol (for example paired associative stimulation or TMS-EEG potentiation) and complementary imaging before dosing and over hours, days and weeks. Independently randomize a learning or stimulation intervention to a measured candidate window versus a later interval. Prespecify clinician-rated outcomes at six and twelve weeks and six months, allocation guesses and confidence in participants and raters, expectancy ratings, participant-level analysis, multiplicity, attrition, and treatment-policy plus rescue-sensitive estimands. A drug-by-timing interaction supported by measured plasticity would add causal leverage; timing effects alone would remain open to expectancy and ordinary learning alternatives. A verified window with no related sustained symptom advantage would weaken the clinical mediator claim.

3. Test the mediator with the drug left in place, at adequate power. Give psilocybin with and without a 5-HT2A antagonist (or under conditions that suppress the acute experience) in depressed participants, with a preregistered plasticity readout and a prespecified sustained symptom endpoint, powered for the mediation contrast rather than for feasibility. This is the design shape the two ongoing trials adopt at small scale; the informative version needs enough participants to interpret a null, randomized rather than sequential assignment of setting and order, and a plasticity measure that indexes inducibility rather than a static marker.

Knowledge gaps

• Whether a drug-induced increase in human cortical inducibility exists at all, and, if so, its opening and closure times.
• Whether the relevant process is structural growth, functional adaptation, altered learning sensitivity, or a combination — the independent adult-rat dissociation makes this a live empirical question rather than a definitional one.
• Whether the proposed process causally carries sustained symptom benefit after accounting for experience, support, expectancy and later care.
• How much of the measured clinical benefit survives comparators matched for acute subjective effects, and what an acceptable comparator in this field would be.
• Which baseline clinical, medication, receptor, cell and circuit contexts reproduce the effect; and whether the dosing-environment effect reported in exploratory human imaging survives randomization.
• How much benefit remains attributable to initial treatment after twelve weeks, six months and a year, once re-treatment and rescue medication are handled prospectively.

Red-team synthesis note

This edition is the published output of a cross-model red-team pass: the review was authored by one model family and adversarially audited and rewritten by another. Sixteen findings were raised; fifteen were accepted or partially accepted and one — a grade change in either direction — was rejected on the evidence. The overall grade is unchanged at Γ with Moderate confidence, and all five subclaim levels are unchanged. The substantive changes are: routing the review to the psychiatry and neuromodulation rubric that the packet actually contains, and re-deriving the caps and thresholds from its text; adding the functional-unblinding evidence base — two 2025–2026 quantitative syntheses — so that expectancy is an evidenced constraint rather than an assertion; adding the omitted ketanserin arm of the 2021 mouse spine study and an independent adult-rat functional-plasticity line; reporting the phase 3 program as explicitly non-admissible context; recording two previously unlogged journal corrections and re-characterising a third; removing statistics that could not be verified from a primary source at this pass's access depth; and correcting the Momentum and Stability artifacts, since two registered human trials are in fact running the mediator-manipulation design the review said no one was attempting. Every finding, its evidence, and its disposition are recorded in redteam/REDTEAM_REPORT.md; every prose and scorecard change is itemised in redteam/SYNTHESIS_CHANGE_LOG.md.

Reviewer notes and provenance

This is a targeted primary-literature assessment through 2026-09-28, not an exhaustive systematic review or a pooled effect estimate. The supplied framework packet provides methodology and schema and no scientific evidence.

Reading depth is reported as verified in this pass, which is stricter than the original edition's accounting. full_text_complete means the main article was reviewed in this pass; full_text_partial means relevant primary Methods/Results/Discussion were reviewed, with limitations logged; abstract_metadata means full Methods were not available for audit at this pass's access depth, and visible publisher snippets or figures do not upgrade it. Several records the original edition marked as read at full text have no freely accessible full text (the two New England Journal of Medicine trials, the 2022 Nature Medicine imaging analysis, the 2011 pilot, the 2016 Lancet Psychiatry feasibility study, the 4.5-year cancer follow-up); they are recorded here at the depth this pass could verify, with their reported numbers taken from publisher abstracts. One in-press study's statistics were verified against the openly available author version of the same study, with a version check. Where a specific statistic could not be verified from a primary record, it has been removed and the gap is stated rather than repeated. Access availability is not an assertion of a redistribution licence, and no article PDFs are redistributed in this package.

Source-status screen. All fifty ledger records were screened for retractions, expressions of concern and corrections against PubMed and Europe PMC comment/correction links on 2026-09-28. No retractions or expressions of concern were found. Six records carry errata, all now attached to their parent rows: the 2023 JAMA trial (byline plus minor typographical corrections; the primary MADRS estimate is unchanged), the 2021 JAMA Psychiatry waitlist trial (errors in a response rate, effect sizes and confidence intervals; the values used here are the corrected ones), the 2023 Swiss trial (figure colour labels), the 2024 clinician trial (byline; author affiliations), the 2025 Nature cell-type study (calcium-figure axis signs; spine and behaviour conclusions not withdrawn) and the 2019 occupancy PET study (figure numbering). Corrections are recorded as status flags on the parent source and are never given their own numbered row. The original edition's unsourced statement that a retracted rat-activity paper had been identified and excluded could not be verified and has been removed.

Shared cohorts and programs. Imperial TRD clinical and imaging reports; the psilocybin–escitalopram clinical, imaging and follow-up reports; Johns Hopkins clinical, cognitive and twelve-month follow-up; NYU cancer trial and multi-year follow-up; EPISODE and its naturalistic follow-up; COMP001/COMP003 and their observational extension. Independent cohorts are not automatically independent causal lines: functional unblinding can recur across trials, and the strongest direct adult cortical structural experiments share one laboratory program and one imaging technique.

The structural validator checks schema and internal consistency, not scientific truth or reproducibility. Its output for this edition is saved alongside the package.

Primary evidence and access ledger

ADVERSARIAL REVIEW AND APPROVAL

Authoring model: Sol 6.1 (the document's own provenance line records a Codex/GPT original assessment, 2026-09-27, and a Claude red-team synthesis, 2026-09-28; recorded, not edited). Adversarial reviewer: Claude Opus 5.5 (Anthropic). Date: 2026-10-14. Round: 2 (new-review inventory, group 1).

Scope of checks. All 50 REFERENCES PMIDs verified against PubMed get_article_metadata for title, first author, journal and year: 50/50 resolve correctly, 0 failures. The four attached erratum PMIDs (38277159, 33566065, 39804652, 40638123) were also verified and each is a Published Erratum in the journal named. No retraction or expression of concern appears in the article-type fields of any of the 54 records checked. Twelve load-bearing claims were spot-checked against PubMed abstracts, concentrating on the numbers that carry SC1, SC2, SC4 and the functional-unblinding audit: EPISODE's primary response endpoint and odds ratio; the Swedish trial's day-8/15/42/365 MADRS contrasts and the day-102 MADRS-S separation; Goodwin 2022's dose contrasts and its explicit "but not sustained response at 12 weeks" wording; the psilocybin-escitalopram primary QIDS-SR-16 null and its two secondary contrasts; Raison's day-43 and day-8 MADRS differences and Sheehan Disability Scale difference; the NHS feasibility trial's week-3 adjusted MADRS difference; Shao 2021's spine-growth result; Shao 2025's PT/IT silencing and adult cortical 5-HT2A knockout; Johansen's overall SV2A null; Kramer's functional-without-structural-plasticity dissociation; and all three Tier B syntheses (Williams, Hieronymus, Metaxa), including the direction of Hieronymus's 14- and 23-percentage-point control-arm gaps. Every spot-checked figure matched the source. The seven rubric-07 rules quoted in section 6 were checked verbatim against references/evidence-rubrics/07-psychiatry-neuromodulation.md and all seven are quoted exactly. Scoring logic was re-derived: the six universal caps, cap precedence, the pre-registered weakest-link aggregation, the independence basis required above Zeta, the Beta/Gamma/Delta boundary, the Momentum/Stability/Divergence artifacts, and the role-derivation procedure. The validate_review.py 5.4 structural validator was re-run.

Findings

No scientific finding required a change: the headline letter, the confidence band, all five subclaim levels, the role label and the M/S/D values are unchanged.

Rejected findings (checked, no change)

Lead policy applied — confidence bands and momentum (2026-10-14)

Confidence, two-step test. Step 1: access depth is struck as a determinant. How much of a paper this reviewer could read is neither evidence density nor reproducibility and never justifies a band; the rationale was rewritten accordingly (finding A-5). Step 2: the band was re-derived from what remains. Access depth was not the only thing holding the band down. Three substantive evidence-quality determinants survive and each is documented in the review's own text: human plasticity measurement is heterogeneous across PET, fMRI, EEG and cognitive proxies; the proposed mediator has never been manipulated, so no study measures the quantity the claim is about; and the largest randomized datasets (COMP005, COMP006) are company and regulatory documents rather than peer-reviewed reports. The review further records a null preregistered synaptic-marker test stopped for futility, a fixed-order healthy-volunteer design, and near-ceiling allocation guessing wherever blinding was measured. These are reproducibility and evidence-quality concerns of exactly the kind the policy names as valid. Moderate stands.

Momentum, stricter rule. "Decisive work underway" requires work actually being run that could cross the next threshold; a completed publication or trial never satisfies it, however real the study. Re-checked and the value holds. Both artifacts are trials actually recruiting, not completed reports: SYNVEST (NCT06512220, recruiting since 2025-04-21, primary completion 2027-04) and NCT07490353 (recruiting since 2025-11-01, primary completion 2028-05), and the review states that neither has results. Both run the psilocybin-with-and-without-antagonist design that rubric 07 lists as threshold-crossing. The review already discloses that SYNVEST's declared primary outcome is feasibility and that n=12 will not settle mediation. No completed publication is counted toward this value.

Residual limitations

• Twenty-four of fifty records are recorded at abstract depth, so figures taken from them were verified against publisher abstracts rather than full Methods. Three figures this reviewer could not verify at abstract depth are ones the review itself attributes to full text it did read: the NHS trial's week-6 MADRS difference and allocation-guessing counts (row [1]), Shao 2025's day-65 imaging session (row [32]), and Johansen's interim-futility conditional-power figures (row [37], verified in that pass against the author version). All three are consistent with the abstracts and are not flagged as errors.
• The largest randomized datasets in the field (COMP005, COMP006) remain company and regulatory documents. The review correctly excludes them from the graded base and from the ledger, so the grade cannot be checked against them.
• This pass did not re-derive the review's own earlier red-team dispositions, which live in redteam/REDTEAM_REPORT.md.

Approval: Following adversarial review and the corrections listed above, the revised document is approved. - Claude Opus 5.5 (Anthropic), 2026-10-14.

References

1. Rucker JJ et al. (2026). Psilocybin-assisted therapy for treatment-resistant major depressive disorder in a public healthcare setting: a randomized controlled trial. Nature medicine. DOI; PMID 42562964.

2. Mertens LJ et al. (2026). Efficacy and Safety of Psilocybin in Treatment-Resistant Major Depression: The EPISODE Randomized Clinical Trial. JAMA psychiatry. DOI; PMID 41848690.

3. Mertens LJ et al. (2026). Long-Term Efficacy of Psilocybin with Adjunct Psychotherapy in Treatment-Resistant Major Depression: 6- and 12-Month Naturalistic Follow-Up of a Phase 2b Trial. Psychotherapy and psychosomatics. DOI; PMID 42201843.

4. Yngwe H et al. (2026). Short-Term and Late-Term Effects of Psilocybin on Symptoms in Major Depression: A Randomized Clinical Trial. JAMA network open. DOI; PMID 42138922.

5. Raison CL et al. (2023). Single-Dose Psilocybin Treatment for Major Depressive Disorder: A Randomized Clinical Trial. JAMA. DOI; PMID 37651119. Linked correction: article corrected 26 January 2024 to add an author to the byline and to correct minor typographical errors; the primary MADRS estimate is unchanged (PMID 38277159).

6. Goodwin GM et al. (2022). Single-Dose Psilocybin for a Treatment-Resistant Episode of Major Depression. The New England journal of medicine. DOI; PMID 36322843.

7. Goodwin GM et al. (2025). Results From a Long-Term Observational Follow-Up Study of a Single Dose of Psilocybin for a Treatment-Resistant Episode of Major Depressive Disorder. The Journal of clinical psychiatry. DOI; PMID 40047545.

8. Goodwin GM et al. (2023). Psilocybin for treatment resistant depression in patients taking a concomitant SSRI medication. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. DOI; PMID 37443386.

9. Carhart-Harris R et al. (2021). Trial of Psilocybin versus Escitalopram for Depression. The New England journal of medicine. DOI; PMID 33852780.

10. Erritzoe D et al. (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. EClinicalMedicine. DOI; PMID 39764567.

11. Davis AK et al. (2021). Effects of Psilocybin-Assisted Therapy on Major Depressive Disorder: A Randomized Clinical Trial. JAMA psychiatry. DOI; PMID 33146667. Linked correction: errors in a response rate, effect sizes and confidence intervals corrected 10 February 2021 (PMID 33566065); the corrected values are used here.

12. Gukasyan N et al. (2022). Efficacy and safety of psilocybin-assisted treatment for major depressive disorder: Prospective 12-month follow-up. Journal of psychopharmacology (Oxford, England). DOI; PMID 35166158.

13. Carhart-Harris RL et al. (2016). Psilocybin with psychological support for treatment-resistant depression: an open-label feasibility study. The lancet. Psychiatry. DOI; PMID 27210031.

14. Carhart-Harris RL et al. (2018). Psilocybin with psychological support for treatment-resistant depression: six-month follow-up. Psychopharmacology. DOI; PMID 29119217.

15. von Rotz R et al. (2023). Single-dose psilocybin-assisted therapy in major depressive disorder: A placebo-controlled, double-blind, randomised clinical trial. EClinicalMedicine. DOI; PMID 36636296. Linked correction: Figure 2 color labels harmonized: blue psilocybin, yellow placebo; numerical findings unchanged.

16. Ross S et al. (2016). Rapid and sustained symptom reduction following psilocybin treatment for anxiety and depression in patients with life-threatening cancer: a randomized controlled trial. Journal of psychopharmacology (Oxford, England). DOI; PMID 27909164.

17. Agin-Liebes GI et al. (2020). Long-term follow-up of psilocybin-assisted psychotherapy for psychiatric and existential distress in patients with life-threatening cancer. Journal of psychopharmacology (Oxford, England). DOI; PMID 31916890.

18. Griffiths RR et al. (2016). Psilocybin produces substantial and sustained decreases in depression and anxiety in patients with life-threatening cancer: A randomized double-blind trial. Journal of psychopharmacology (Oxford, England). DOI; PMID 27909165.

19. Grob CS et al. (2011). Pilot study of psilocybin treatment for anxiety in patients with advanced-stage cancer. Archives of general psychiatry. DOI; PMID 20819978.

20. Back AL et al. (2024). Psilocybin Therapy for Clinicians With Symptoms of Depression From Frontline Care During the COVID-19 Pandemic: A Randomized Clinical Trial. JAMA network open. DOI; PMID 39636638. Linked corrections: byline (PMID 39804652) and author affiliations (DOI 10.1001/jamanetworkopen.2025.25303; PMID 40638123); numerical results unaffected.

21. Weintraub MJ et al. (2026). Psilocybin-assisted cognitive behavioral therapy for major depressive disorder: A pilot trial. Journal of affective disorders. DOI; PMID 41707717.

22. Sloshower J et al. (2023). Psilocybin-assisted therapy for major depressive disorder: An exploratory placebo-controlled, fixed-order trial. Journal of psychopharmacology (Oxford, England). DOI; PMID 36938991.

23. Ross ML et al. (2025). Psilocybin-assisted psychotherapy for depression and anxiety associated with life threatening illness: A phase 2b randomized controlled trial. General hospital psychiatry. DOI; PMID 40858059.

24. Shao LX et al. (2021). Psilocybin induces rapid and persistent growth of dendritic spines in frontal cortex in vivo. Neuron. DOI; PMID 34228959.

25. Hesselgrave N et al. (2021). Harnessing psilocybin: antidepressant-like behavioral and synaptic actions of psilocybin are independent of 5-HT2R activation in mice. Proceedings of the National Academy of Sciences of the United States of America. DOI; PMID 33850049.

26. Raval NR et al. (2021). A Single Dose of Psilocybin Increases Synaptic Density and Decreases 5-HT2A Receptor Density in the Pig Brain. International journal of molecular sciences. DOI; PMID 33467676.

27. Cameron LP et al. (2023). 5-HT2ARs Mediate Therapeutic Behavioral Effects of Psychedelic Tryptamines. ACS chemical neuroscience. DOI; PMID 36630260.

28. Vargas MV et al. (2023). Psychedelics promote neuroplasticity through the activation of intracellular 5-HT2A receptors. Science (New York, N.Y.). DOI; PMID 36795823.

29. Moliner R et al. (2023). Psychedelics promote plasticity by directly binding to BDNF receptor TrkB. Nature neuroscience. DOI; PMID 37280397.

30. Nardou R et al. (2023). Psychedelics reopen the social reward learning critical period. Nature. DOI; PMID 37316665.

31. Sekssaoui M et al. (2024). Antidepressant-like effects of psychedelics in a chronic despair mouse model: is the 5-HT2A receptor the unique player?. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. DOI; PMID 38212441.

32. Shao LX et al. (2025). Psilocybin's lasting action requires pyramidal cell types and 5-HT2A receptors. Nature. DOI; PMID 40175553. Linked correction: Calcium-figure axis signs corrected to (after−before)/before in Fig. 3e,h,k,n and Extended Data Fig. 7b–e; spine/behavior conclusions not withdrawn.

33. Jiang Q et al. (2026). Psilocybin triggers an activity-dependent rewiring of large-scale cortical networks. Cell. DOI; PMID 41352354.

34. Schmidt M et al. (2026). Psilocin fosters neuroplasticity in iPSC-derived human cortical neurons. eLife. DOI; PMID 41891829. eLife research article RP104006; human developmental culture.

35. Jain MK et al. (2025). The polypharmacology of psychedelics reveals multiple targets for potential therapeutics. Neuron. DOI; PMID 40683247.

36. Kramer HM et al. (2026). Psychedelics produce enduring behavioral effects and functional plasticity through mechanisms independent of structural plasticity. Neuropsychopharmacology : official publication of the American College of Neuropsychopharmacology. DOI; PMID 41224969.

37. Johansen A et al. (2026). Psilocybin's effect on human brain synaptic plasticity. Transl Psychiatry. DOI; PMID 42457666. Accepted publisher manuscript in press; precursor preprint not separately counted.

38. Lyons T et al. (2026). Human brain changes after first psilocybin use. Nat Commun. DOI; PMID 42086570.

39. Daws RE et al. (2022). Increased global integration in the brain after psilocybin therapy for depression. Nat Med. DOI; PMID 35411074.

40. Carhart-Harris RL et al. (2017). Psilocybin for treatment-resistant depression: fMRI-measured brain mechanisms. Sci Rep. DOI; PMID 29030624.

41. Doss MK et al. (2021). Psilocybin therapy increases cognitive and neural flexibility in patients with major depressive disorder. Transl Psychiatry. DOI; PMID 34750350.

42. Siegel JS et al. (2024). Psilocybin desynchronizes the human brain. Nature. DOI; PMID 39020167.

43. Madsen MK et al. (2019). Psychedelic effects of psilocybin correlate with serotonin 2A receptor occupancy and plasma psilocin levels. Neuropsychopharmacology. DOI; PMID 30685771. Linked correction: Figure numbering correction; no reported numerical change.

44. Barrett FS et al. (2020). Emotions and brain function are altered up to one month after a single high dose of psilocybin. Sci Rep. DOI; PMID 32042038.

45. Holze F et al. (2022). Direct comparison of the acute effects of lysergic acid diethylamide and psilocybin in a double-blind placebo-controlled study in healthy subjects. Neuropsychopharmacology. DOI; PMID 35217796.

46. Becker AM et al. (2022). Acute Effects of Psilocybin After Escitalopram or Placebo Pretreatment in a Randomized, Double-Blind, Placebo-Controlled, Crossover Study in Healthy Subjects. Clin Pharmacol Ther. DOI; PMID 34743319.

47. Williams ZJ et al. (2026). Psychedelic Therapy vs Antidepressants for the Treatment of Depression Under Equal Unblinding Conditions: A Systematic Review and Meta-Analysis. JAMA psychiatry. DOI; PMID 41848744. Tier B synthesis; added in the red-team pass.

48. Hieronymus F et al. (2025). Control Group Outcomes in Trials of Psilocybin, SSRIs, or Esketamine for Depression: A Meta-Analysis. JAMA network open. DOI; PMID 40736734. Tier B synthesis; added in the red-team pass.

49. Metaxa AM, Clarke M. (2024). Efficacy of psilocybin for treating symptoms of depression: systematic review and meta-analysis. BMJ. DOI; PMID 38692686. Tier B synthesis; added in the red-team pass.

50. Purple RJ et al. (2025). Short- and long-term modulation of rat prefrontal cortical activity following single doses of psilocybin. Molecular psychiatry. DOI; PMID 40858779. Added in the red-team pass.

Documentary record

Every grade on this site is reproducible from its record. The files below are the posted edition exactly as assessed.

• ADVERSARIAL_REVIEW.md
• SHA256.txt
• adversarial-findings.json
• codex-correction-provenance.json
• consensus-addendum.md
• consensus-validation.json
• edition-approval-audit.json
• psilocybin-therapy.review.md
• psilocybin-therapy.scorecard.json
• research-update-2026-10-05.bib
• research-update-2026-10-05.html
• research-update-2026-10-05.json
• research-update-2026-10-05.md
• research-update-2026-10-05.ris
• validation-report.json
• validation-round2-2026-10-14.json
• validation.txt
• video-insert-2026-10-05.md

Sources: 50 · load-bearing claims: 50 · verified: 50 · unverified: 0 · contradicted: 0 · schema 5.4