Buspirone is an antianxiety compound that has been extensively evaluated in clinical trials: it has proved superior to placebo and comparable to diazepam in the treatment of patients with generalized anxiety disorder. In this study, 33 outpatients with generalized anxiety disorder were entered into a crossover study of 3 weeks each of placebo, buspirone 10 to 30 mg daily, and diazepam 10 to 30 mg daily. Psychiatrist and patient ratings were made, together with psychological tests and EEG and skin conductance measures before and after each treatment. Of the nine dropouts, six were on buspirone at the time of dropout. For the remaining 24 patients, the mean daily doses attained of buspirone and diazepam were both 20 mg. On most clinical ratings diazepam was superior to buspirone and placebo, which did not differ. Diazepam produced minor psychomotor changes and the expected major effects on the EEG. Buspirone was without effect. Side effects on buspirone were mainly nausea and giddiness and on diazepam, drowsiness. The lack of efficacy of buspirone is discussed in terms of the previous benzodiazepine exposure--23/24 patients had had previous exposure and only 10 were able to tolerate a pretrial placebo washout period. The implications are considerable for the introduction of any new antianxiety agent not cross-tolerant with the benzodiazepines into a chronically anxious group of patients with previous long-term benzodiazepine therapy.