We have developed a model of benzodiazepine-type physical dependence in which mice were injected subcutaneously with the test compound on a fixed schedule (0800 and 1600 for 3 days, the PM dose = AM dose x 2). If tolerated, then a starting dose of 150 mg/kg/day was generally used initially and the dose was lowered to 15 and 1.5 mg/kg/day in subsequent assays if the higher doses were active in the test. Twenty-four hours after the last dose, the mice received an intravenous injection of flumazenil (2.5 mg/kg), and 5 min later they were tested for electroshock seizure thresholds by an up-down titration method. Flumazenil-precipitated withdrawal was manifested by a lowering of the mA seizure threshold. We have found that compounds with benzodiazepine agonist properties significantly lower these thresholds in a dose-related fashion. For example, the following compounds (lowest effective mg/kg/day dose) were active in this regard, chlordiazepoxide (150), diazepam (15), flurazepam (15), alprazolam (15), triazolam (15), midazolam (15), zopiclone (150), Ro 16-6028 (150), and Ro 17-1812 (150). In contrast, zolpidem (150), tracazolate (15), and CL 218872 (15) did not cause physical dependence by this criterion. This rapid and simple screening test may be readily used to predict the physical-dependence-inducing properties of compounds that act at the benzodiazepine receptor.