Double-blind

Double-blind is a design feature in which neither the participants in a study nor the researchers assessing outcomes know who received the active substance and who received a control, typically a placebo. It is not a study type on its own but a layer applied to a randomised trial, a crossover design, or a challenge study.

The single-blind design hides assignment from participants but not from researchers. The open-label design hides nothing. Double-blind targets expectation from both sides simultaneously — the participant's belief about what they took, and the assessor's knowledge of what they gave.

On these pages, double-blind appears most often as a qualifier on a cited study — a signal that expectation bias was at least designed to be controlled. Whether it succeeded in practice is a different question, and the most important one.

What this design can establish

When the blind holds, a double-blind study can isolate the pharmacological signal from the placebo response. For substances where subjective state is the primary outcome — mood, perception, pain — this separation matters most, because expectation powerfully shapes how effects are experienced and reported.

Combined with randomisation, a double-blind design can support causal claims: that the substance, not belief or selection bias, produced the observed difference. It can establish dose-response relationships within the studied population and, when two active treatments are compared, distinguish their relative effects from each other.

What it cannot

A double-blind study cannot establish that the blind held. With psychoactive substances, blinding breaks often: a participant who notices strong perceptual effects guesses they received the active drug; a researcher who observes altered behaviour does the same. Studies routinely carry the double-blind label in publication without reporting whether blinding integrity was ever assessed.

It cannot establish how results generalise beyond the trial population. Controlled studies typically exclude people with psychiatric histories, medical comorbidities, or concurrent drug use — the populations most likely to encounter the substance in real conditions. A well-blinded result from that narrow sample does not transfer automatically.

It also cannot establish long-term safety from a short study, nor rule out rare events. A trial that observed no serious adverse events shows they did not occur in that sample over that duration — not that they do not happen.