In this double-blind, placebo-controlled 10-week trial, the anxiolytic properties of the nonbenzodiazepine buspirone were compared with the benzodiazepine lorazepam and placebo in 125 outpatients with generalized anxiety disorder according to DSM-III. After a 3-to 7-day wash-out period, patients were allocated at random to receive orally 3 × 5 mg buspirone (n = 58), 3 × 1 mg lorazepam (n = 57), or placebo (n = 10) over a 4-week period. The study also comprised a 2-week taper period and a 4-week placebocontrol period to assess the stability of clinical improvement. The patient´s clinical state was estimated on entry and at weekly intervals by general practitioners using the Hamilton Rating Scale for Anxiety (HAM-A) and Clinical Global Impression (CGI) assessment and by a self-rating scale (State Trait Anxiety Inventory X2 = STAI-X2). Lorazepam treatment resulted in descriptively, but not signiÞcantly, greater improvement on the Hamilton Rating Scale for Anxiety during the whole treatment (week 0 4) and taper period (week 5, 6) than did buspirone. After treatment with active drugs had been discontinued, the 4-week placebo control period showed buspirone-treated patients to display a stability of clinical improvement, while the symptoms of lorazepam-treated patients worsened at week 710. Both buspirone and lorazepam were more e¦cacious in reducing anxiety symptoms than placebo during the treatment and taper period; however, in contrast to the active drugs (buspirone, lorazepam), patients of the placebo group showed further clinical improvement during the control period, especially in the HAM-A score, so di¤erences between placebo and active drugs became smaller at the end of the study.