A within-subject design was used to characterize the effects of dose manipulations on discriminative and self-reported effects of oral diazepam and buspirone. Subjects were trained to discriminate diazepam (10 mg) versus placebo (n = 10), or buspirone (10 or 15 mg) versus placebo (n = 9). The compounds were identified to subjects by letter code before discrimination training began. In later sessions, correct identifications at 2 hr after the oral administration of drug earned money. All subjects showed accurate discrimination performance during the test-of-acquisition phase. In a low-dose generalization phase, diazepam and buspirone produced dose-related increases in drug identifications across a four-fold range of doses. In a subsequent low-dose training phase, in which subjects were trained to discriminate progressively lower drug doses, the median lowest discriminable dose of diazepam and buspirone was 2.5 and 7.5 mg, respectively. Dose-response functions for drug identifications were shifted leftward in the low-dose training phase relative to the low-dose generalization phase, suggesting that reinforcement of progressively lower doses enhances drug discriminability. The self-reported effects of diazepam and buspirone were similar (e.g., both drugs increased ratings of drug strength and clumsy/uncoordinated) and different (e.g., diazepam but not buspirone increased ratings of drowsy/sleepy; buspirone but not diazepam increased ratings of tense/nervous). This study demonstrates discriminative and self-reported effects of diazepam and buspirone at doses lower than previously shown to be behaviorally active, and suggests that at commonly used clinical doses, diazepam is relatively more discriminable than buspirone.