Rats were trained to discriminate pentylenetetrazole (PTZ; 20 mg/kg) from midazolam (MDZ; 1 mg/kg) and from saline using a three-lever food-reinforced choice task. Using a cumulative dosing procedure, PTZ substituted for PTZ, and MDZ, chlordiazepoxide and diazepam (DZP) substituted for MDZ, in a dose-dependent manner. The animals were then treated with chronic DZP (20 mg/kg/8 hr for 7 days); 24 hr after the last dose of this regimen, the dose-effect curve of DZP was redetermined. The ED50 for the discrimination of DZP increased 4.8-fold after chronic DZP. In a second group of subjects trained on this discrimination and treated with DZP (20 mg/kg/8 hr for 7 days), the ED50 for the discrimination of MDZ was increased 2.2-fold. After 14 days of recovery, the MDZ dose-effect curve shifted back to the left and was not significantly different from the ED50 value obtained before chronic DZP treatment. When the benzodiazepine antagonist flumazenil was tested in a cumulative manner (1.0-32.0 mg/kg), the animals selected the saline lever. Analogous to the previous chronic dosing regimen, DZP (20 mg/kg/8 hr/7 days) was then administered, and a combination of three tests were then given at 1, 2, 4, 7, 10 and 14 days after the last DZP treatment. On each of these days, a saline test was given first; it was followed by a DZP (2.5 mg/kg) test; which was then followed by a flumazenil (32.0 mg/kg) test. On the 1st day of testing, tolerance was seen to DZP and precipitated withdrawal (PTZ lever selection) was seen with flumazenil.(ABSTRACT TRUNCATED AT 250 WORDS)