Caffeine engenders qualitatively different subjective effects in humans at low and high doses. Low doses of caffeine are mildly reinforcing and produce psychomotor stimulation. High doses of caffeine can produce subjective feelings of anxiety, dysphoria and depression. The present study was designed to model these different subjective states in rats using a discrete trial shock avoidance/escape drug discrimination paradigm. Rats were trained to discriminate between i.p. injections of saline and either 10 or 56 mg/kg of caffeine. Rats trained at 10 mg/kg of caffeine acquired the discrimination in an average of 93 sessions and generalized completely to a variety of xanthine and nonxanthine behavioral stimulants including: d-amphetamine, apomorphine, 7-(beta-chloroethyl)theophylline, 9-chloro-2-(2-furanyl)-5,6-dihydro-1, 2,4-triazolo[1,5-c]quinazolin-5-imine (CGS 15943), cocaine, 1,7-dimethylxanthine, diethylpropion, beta-hydroxyethyltheophylline, methylphenidate, phenidimetrazine and theophylline. Rats trained at 56 mg/kg of caffeine acquired the discrimination in an average of 43 sessions and generalized completely only to theophylline. A variety of drugs representing diverse pharmacologic classifications including: benzodiazepine inverse agonists, pentylenetetrazol, yohimbine, ethylketocyclazocine and phencyclidine, were not generalized from either training dose, demonstrating the pharmacologic specificity of the discrimination. The discriminative effects of 10 mg/kg of caffeine appear to derive from a state of behavioral arousal, possibly mediated by catecholamines, and parallel the subjective effects produced by low doses of caffeine in humans. The discriminative effects of 56 mg/kg of caffeine are qualitatively different from those of 10 mg/kg but cannot be defined further at this time.