Benzodiazepine receptor (BZR) ligands previously characterized as differing in intrinsic efficacy were evaluated first for potency in antagonizing flunitrazepam-induced sleep in monkeys. Data from these experiments were used to define approximately equieffective doses for subsequent use in precipitating withdrawal in diazepam-treated monkeys. It was shown that partial agonists with intermediate intrinsic efficacy (bretazenil, Ro 41-7812) were relatively ineffective in precipitating withdrawal reactions in diazepam-treated squirrel monkeys. The potent and specific BZR antagonist flumazenil, which possesses weak intrinsic efficacy, was more effective in precipitating a withdrawal reaction in diazepam-treated monkeys. In contrast, the highest dose of the BZR antagonist ZK 93426 that could be administered failed to precipitate withdrawal under the same experimental conditions. Finally, the BZR partial inverse agonist sarmazenil was the most effective of these BZR ligands in eliciting a precipitated withdrawal reaction. Thus, the results of the present investigation strongly suggest that BZR ligands differing in intrinsic efficacy differentially precipitate withdrawal in squirrel monkeys treated chronically with diazepam.