Functional Unblinding: Why Psychedelic Trials Cannot Fully Blind
Participants in psilocybin trials correctly identify their arm far above chance because the experience is unmistakable, so expectancy inflates the drug arm and deflates placebo simultaneously — and no standard analysis separates the two. A meta-analysis restricted to equal-unblinding conditions found psychedelic therapy may be no more effective than open-label conventional antidepressants, and pre-treatment expectancy predicts outcome. The general test: could participants tell, and was the endpoint subjective?
**Functional unblinding** is the failure of a trial's blind in practice, even when the protocol is formally double-blind: participants and often assessors can tell which arm they are in. It is the central methodological problem in psychedelic research, and it is severe enough that it may account for a substantial share of the reported effect. ## Why psychedelic trials break the blind The intervention produces unmistakable subjective effects for several hours. In psilocybin trials, participants correctly identify their assignment at rates far above chance — the experience is not something a placebo can imitate. Therapists present during the session usually know too. Once both parties know, expectancy operates freely. In the active arm it inflates response; in the placebo arm, participants who realise they received nothing are disappointed, which **deflates** the control. The measured difference between arms is therefore the drug effect *plus* the expectancy gap, and no standard analysis can separate them. Depression outcomes are especially vulnerable because the primary endpoints are subjective rating scales rather than objective measurements. ## What the evidence suggests Two findings are hard to dismiss. A meta-analysis restricted to conditions of **equal unblinding** — comparing like with like on how much each arm knew — found psychedelic therapy may be no more effective than open-label conventional antidepressants. And **pre-treatment expectancy scores predict outcome**, meaning some of the measured antidepressant effect tracks what participants believed before dosing rather than what the receptors did. This does not show that psilocybin has no pharmacological antidepressant effect. It shows that the size of that effect is not established by the existing trials, because the design cannot isolate it. ## Attempted fixes, and why they are partial - **Active placebos** (niacin, low-dose psilocybin, midazolam) produce noticeable effects to muddy the guess. They rarely convince anyone who has had a full psychedelic dose, and a low dose of the study drug is not a true control. - **Dose-comparison designs** — 25 mg versus 1 mg — avoid an inert comparator but change the question: the result is a dose-response finding, not a drug-versus-nothing finding. - **Blinding assessors** to arm assignment removes one route, not the participant's own expectancy. - **Measuring and adjusting for expectancy** helps quantify the problem without eliminating it. ## The general lesson Functional unblinding is not unique to psychedelics. It affects any trial where the intervention is obvious — exercise, surgery, acupuncture, cold exposure, many device trials, and anything with unmistakable side effects (which is one reason SSRI trials have their own version of this problem). The transferable question when reading any trial is: **could participants tell which arm they were in, and was the outcome measure subjective?** If both, the reported effect size includes an expectancy component of unknown size. See Psilocybin Misconceptions: Case Reports, Blinding, and Headline Inflation.