In a 8-week double-blind placebo controlled study, 48 outpatients with generalized anxiety disorder were randomized to diazepam, buspirone, a non-benzodiazepine anxiolytic, or placebo. During the treatment phase of 4 weeks duration diazepam was found to be significantly better than placebo and buspirone. Following abrupt withdrawal by placebo substitution the diazepam group showed a gradual relapse maximal after two weeks while the buspirone and the placebo groups did not differ. There were more cases of rebound anxiety with diazepam as compared to buspirone or placebo. In addition, there were three early terminations related to rebound anxiety in the diazepam group while there were none in the placebo and buspirone groups. There were significantly more new symptoms in the diazepam group than in the placebo or buspirone group.