This experiment tested whether benzodiazepine withdrawal could be detected in an animal model of anxiety. Rats were trained in operant chambers using food reward to press one lever after pentylenetetrazol (PTZ), 20 mg/kg, injection and the other lever after saline injection. Previously, the PTZ cue has been shown to be simulated by anxiogenic drugs and blocked by anxiolytic drugs. After rats reliably performed this discrimination, they were injected with diazepam, 20 mg/kg, from 1 to 4 times a day for six days. For one group of subjects, on the third, fourth and sixth days, they were also injected with 40 mg/kg of RO 15-1788, a benzodiazepine receptor antagonist, and tested for lever selection: 50-80% of the subjects selected the PTZ lever; these results are in contrast to those obtained prior to chronic diazepam treatment in which RO 15-1788 did not generalize to PTZ. A second group of subjects was also injected for six days with diazepam and then allowed to withdraw spontaneously for eight days: PTZ lever selection over this period varied from 20 to 60% of rats. These data indicate that animals trained to discriminate a PTZ cue: 1) generalize the benzodiazepine withdrawal state to the PTZ cue, and 2) discriminate the withdrawal state for long periods of time, agreeing with clinical observations of long-lasting anxiety signs during benzodiazepine withdrawal.