Double-blind

Double-blind names two roles, not two people. The subjects do not know which treatment they received, and neither do the people who interact with them or score the response. The assignment has not vanished: it sits in a code held by someone who never meets a subject, and it is opened once the data are in. Blind exactly one of those two roles and the design is single-blind, whichever one it is. None of this describes what the control group receives.

A cold remedy trial randomly assigns 240 volunteers, 120 to the drug and 120 to an identical placebo, and the response is days until symptoms clear, judged by a nurse working from a checklist. Suppose the nurse knows the assignment and resolves borderline cases a quarter of a day early in the drug group and a quarter of a day late in the placebo group. The estimated difference between the groups then moves by $0.25 + 0.25 = 0.5$ days on its own, against a real drug effect of 0.8 days. Most of what got measured is the nurse. Coding the bottles so she cannot tell removes it and costs nothing.

The sentence to correct is "the study was double-blind because the control group got a placebo." A placebo is something the control group receives; blinding is about who knows. Identical pills from bottles marked A and B, with the nurse holding the key, is single-blind. A second version says the researchers must not know who got what. Somebody has to: the person who ran the randomization knows, and that is fine. The ones who must be kept in the dark are those in contact with subjects or scoring the outcome.

Plenty of experiments cannot be double-blind at all. A subject knows whether she spent twelve weeks lifting weights. Blinding whoever measures the outcome is still available, still worth doing, and still single-blind. Blinding protects a measurement and does nothing to balance the groups, so an unblinded randomized experiment still supports a causal claim while a blinded study without random assignment does not.