Two groups of 12 outpatients each (six men and six women) with generalized anxiety disorder, participated in this study. Each patient was treated single-blind with placebo during the first 7 days (baseline), followed by a double-blind drug treatment period of 4 consecutive weeks (active) and ending again with 7 days single-blind placebo treatment (washout). One group received buspirone 5 mg three times a day in the first week and continued with 10 mg in the morning, 5 mg in the afternoon, and 5 mg in the evening during the second, third, and fourth weeks. The other group received diazepam 5 mg three times a day in all 4 weeks. On the evening of the seventh day of each treatment week the Hamilton Rating Scale for Anxiety and the Symptom Check List (90 items) were applied to assess the therapeutic effects, followed by an on-the-road driving test that started 1.5 hours after the last drug or placebo intake. The test consisted of operating an instrumented vehicle over a 100 kilometer highway circuit while attempting to maintain a constant speed and a steady lateral position within the right traffic lane. Two patients in the diazepam group were unable to complete their test after the first and second treatment week, respectively, because of serious sedative reactions. Both buspirone and diazepam were equally effective in reducing overall anxiety symptoms. The specific profiles showed that buspirone also reduced concomitant depressive symptoms and symptoms of interpersonal sensitivity and anger-hostility. In contrast, diazepam was found to be slightly more effective in reducing somatic symptoms and to positively affect sleep disturbances. Moreover, abrupt discontinuation of diazepam resulted in a relapse of psychic anxiety symptoms comparable with the placebo-baseline level and a partial relapse of somatic anxiety symptoms. Chronic treatment with buspirone had no significant effects on lateral position and speed control. In contrast, diazepam significantly impaired control of lateral position in the first 3 weeks of treatment. There was no significant impairment in the fourth treatment week and the placebo-washout week. Speed control was significantly impaired only in the first week. The relevance of the trend toward decreasing performance impairment during chronic treatment remains to be established.