Forty-four patients with DSM-III-R generalized anxiety disorder participated in this double-blind, randomized study. Patients were on a benzodiazepine before the study and were stabilized on 3 to 5 mg/day lorazepam for 5 weeks (weeks 0 to 5). Thereafter, they were randomized to 15 mg/day buspirone or placebo for the following 6 weeks (weeks 6 to 11). During the first 2 weeks of double-blind, randomized treatment (weeks 6 to 7), lorazepam was tapered off. During weeks 12 to 13, patients received single-blind placebo. Assessment included the Hamilton Rating Scale for Anxiety, the State-Trait Anxiety Inventory, the Zung and Eddy Self-Rating Scale of Anxiety Symptoms, the Hamilton Rating Scale for Depression, and the Rome Depression Inventory, completed at weeks 0, 5, 6, 7, 8, 9, 11, and 13. Side effects were assessed through the Dosage Treatment Emergent Symptoms at the same times. The benzodiazepine-withdrawal syndrome was evaluated through a 27-symptom checklist (Clinical-Rated Benzodiazepine Withdrawal Symptom Schedule) at weeks 0, 5, 6, 7, 11, and 13. The results showed that buspirone was more effective than placebo and comparable to lorazepam. Buspirone-treated patients showed no rebound anxiety or benzodiazepine-withdrawal syndrome compared with placebo. Buspirone caused fewer side effects than lorazepam and was not different from placebo in this respect. Finally, buspirone maintained its anxiolytic effect for at least 2 weeks after the discontinuation of treatment.