Placebos, Blinding, and Confounding

People given a fake pill often improve anyway, just because they believe they're being treated. This is the placebo effect: expectation itself changes the body, through things like lowered stress or changed perception of pain. Because of this, every real drug trial needs a placebo group, something that looks and feels identical to the treatment but has no active ingredient. If the treatment group and placebo group both improve, the drug gets credit only for the improvement beyond the placebo effect.

Blinding handles a second leak of expectation: the researcher's own hopes. If a doctor knows who got the real drug, they might unconsciously rate those patients as doing better, ask them more encouraging questions, or interpret vague symptoms more generously. Double blinding means neither the patient nor the person evaluating them knows who got the treatment, so neither side's expectations can bias the outcome. Single blinding only hides the assignment from the patient, leaving the researcher's bias in play.

Confounding is a different problem, and a control group alone does not fix it. A control group only helps if the groups are otherwise alike. If a trial's treatment group also happened to exercise more than the control group, you cannot tell whether improvement came from the drug, the exercise, or both working together. Exercise is confounded with the treatment. This is exactly why random assignment matters elsewhere in the unit: it spreads lurking variables like exercise habits evenly across groups, so comparisons mean something.