Several lines of evidence have established that performance during the initial steps of acquisition on a shuttle-box avoidance task is an anxiety-mediated behavior (i.e., the differences between strains selectivity bred for emotionality; the effects of postnatal handling; the course of the corticosterone response and behavioral measures of fear during acquisition). The present study was carried out to add pharmacological evidence to that view by testing the action of anxiogenic and anxiolytic drugs. Single 40-trial sessions with mild shocks (0.4 mA-0.6 mA) were used. In the first experiments the action of sodium pentobarbital (1.25, 2.5 and 5 mg/kg) and three benzodiazepines (diazepam, 2 and 4 mg/kg; alprazolam, 1, 1.25 and 1.5 mg/kg and adinazolam, 1, 2, 4 and 6 mg/kg) were tested. The last two experiments tested a possible proanxiety action of Ro 15-4513 (2, 5 and 10 mg/kg) and FG 7142 (5, 10 and 15 mg/kg), two partial inverse agonists of benzodiazepine receptors, which previous data had suggested to be anxiogenic. The results showed that the measure of acquisition of a two-way active avoidance is a sensitive mean for detecting either anxiolytic or anxiogenic effects of drugs, independently of their effects on locomotor activity, thus suggesting that such test could be a valid model of anxiety in animals.