Dogs, surgically implanted with a gastric fistula, were chronically dosed with diazepam or lorazepam. Diazepam (60 mg/kg/day) or lorazepam (100 mg/kg/day) was administered intragastrically in four divided daily doses. Beginning no less than 2 weeks after the attainment of stabilization doses, dogs underwent withdrawal experiments, repeated at 2-week intervals. At a time of withdrawal determined by a Latin square crossover design, dogs were observed for 8 hr for signs of abstinence. Both diazepam and lorazepam caused a withdrawal abstinence syndrome to appear upon abrupt discontinuation of the drug. The two abstinence syndromes had many signs in common, including tremor, hot foot walking, rigidity and decreased food intake, but the lorazepam withdrawal abstinence syndrome was much less intense and had a shorter latency to onset than the diazepam abstinence syndrome, which also included clonic and tonic-clonic convulsions and was lethal in two dogs. Furthermore, the diazepam withdrawal abstinence syndrome was biphasic, the first phase apparent by 24 hr and a second phase beginning at 48 hr, whereas the lorazepam syndrome was not. Diazepam suppressed the major signs of diazepam abstinence in a dose-related manner, but failed to completely suppress all signs of abstinence. CGS-8216, a pyrazoloquinoline benzodiazepine antagonist, precipitated abstinence in the diazepam-dependent dog, but did not precipitate tonic-clonic seizures. No abstinence syndrome was precipitated in the lorazepam-dependent dog. These results would suggest that whereas diazepam and lorazepam both cause physical dependence the two syndromes are not the same and, furthermore, that physical dependence on, and withdrawal from, diazepam involves at least two separate mechanisms with different selectivity for benzodiazepine agonists and antagonists.