The following studies examined the dose and time dependence, site specificity, and reversibility of chlordiazepoxide (CDP)-induced working memory impairments in adult male Sprague-Dawley rats. The rats were tested in a delayed non-match-to-sample radial-arm maze task in which a 1-h delay was imposed between the first four (predelay) and all subsequent (postdelay) arm choices. Intraperitoneal (ip) injection of 2.5 or 5.0 but not 1.25 mg/kg CDP immediately following the predelay session impaired performance in the task. CDP increased the number of errors and decreased the number of correct choices during the postdelay session. The observed working memory impairments also appeared to be site specific since injection of CDP into the medial septum, but not into the anterior amygdala nuclei, immediately following the predelay session also impaired working memory in a dose-related manner. Furthermore, there was a time window for CDP-induced working memory impairments since intraseptal injection of the drug immediately but not 15 min following the predelay session disrupted memory. This observation suggests that the performance deficits reflect disrupted working memory and not proactive effects on performance or the induction of state-dependent learning. In the final experiment, rats were injected ip with either saline or an amnestic dose of CDP (5.0 mg/kg) following the predelay session and then were immediately infused with 10 nmol flumazenil (RO15-1788), a benzodiazepine receptor antagonist or vehicle, into either the medial septum or anterior nuclei of the amygdala. Intraseptal injection of flumazenil prevented the working memory impairments produced by ip injection of CDP. In contrast, intra-amygdala injection of flumazenil did not attenuate, enhance, or modify the CDP-induced working memory impairment. These observations suggest that CDP disrupts working memory by interacting with benzodiazepine receptors in the medial septum.