Subjects were given varying doses of a placebo, consisting of decaffeinated coffee, with double-blind or deceptive instructions. Deceptive administration simulated clinical situations in that subjects were led to believe that they were receiving an active drug. In contrast, subjects in double-blind conditions were aware that they might receive a placebo. Double-blind and deceptive administration of the placebo produced different, and in some instances, opposite effects on pulse rate, systolic blood pressure, and subjective mood. Deceptive administration produced an increase in pulse rate, whereas double-blind administration did not. A theoretically predicted curvilinear effect on systolic blood pressure, alertness, tension, and certainty of having consumed caffeine was confirmed with deceptive administration, but not with double-blind administration. Double-blind administration produced curves in the opposite direction on each of these variables. The effects of the placebo on motor performance varied as a function of subject's beliefs about the effects of caffeine. These data challenge the validity of double-blind experimental designs and suggest that this common method of drug assessment may lead to spurious conclusions.