A novel drug discrimination procedure was used to study the discriminability and subjective effects of caffeine in seven human volunteers who abstained from dietary sources of caffeine. Daily sessions involved p.o. ingestion of two capsules sequentially, one of which contained caffeine and the other placebo, under double-blind conditions. Each day subjects attempted to identify and were later informed which capsule contained caffeine and which contained placebo. All subjects acquired rapidly the initial discrimination (100 or 178 mg vs. placebo). Examination of progressively lower caffeine doses showed that accuracy and ratings of confidence in accuracy were increasing functions of dose. There were individual differences in the lowest discriminable dose: three subjects discriminated 56 mg, three discriminated 18 mg and one discriminated 10 mg. Discrimination accuracy was usually higher after the second capsule than after the first capsule. All subjects indicated that they believed that they made the discrimination predominantly on the basis of central effects of caffeine vs. placebo, such as changes in mood and socializing. Compared to placebo, 100 mg of caffeine increased ratings of alertness, well-being, social disposition, motivation for work, concentration, energy and self-confidence and decreased ratings of headache and sleepiness. This dose of caffeine also produced a large increase in a measure of "euphoria." The present study documents biological activity of caffeine at lower doses than heretofore recognized. The general approach to investigating the effects of low drug doses may have broad application in human psychopharmacology research for characterizing other subtle psychotropic effects.