Risk of withdrawal was investigated in a prospective, double-blind comparison of clorazepate dipotassium, a benzodiazepine with a long half-life, and the nonbenzodiazepine buspirone hydrochloride in the long-term treatment of anxious outpatients. Patients were treated with therapeutic doses of clorazepate dipotassium (15 to 60 mg/d) or buspirone hydrochloride (10 to 40 mg/d) for six continuous months before their tranquilizer therapy was blindly and abruptly stopped. There was a significant increase in symptom severity consistent with a withdrawal reaction for the clorazepate group but not the buspirone group. For the clorazepate group, there was a suggestion that previous discontinuous exposure to benzodiazepines might sensitize patients to subsequent withdrawal effects. For the buspirone group, a higher dropout rate raised questions about patient satisfaction with therapy in this rather chronically anxious population.