In a double-blind, placebo-controlled study, 48 anxious outpatients with a primary diagnosis of generalized anxiety disorder were randomly assigned to 4 weeks of treatment with bromazepam (18 mg/day), diazepam (15 mg/day), or placebo, after a 1-week washout period. From week 1 onward both active drugs were superior to placebo in relieving anxiety symptoms. Bromazepam was found to be significantly more effective than diazepam with respect to the somatic anxiety factor and the total score for the Hamilton Anxiety Rating Scale and the fear/anxiety factor of the Patient's Self-Rating Symptom Scale. Plasma concentrations of diazepam plus active metabolites were correlated significantly (r = 0.60, p less than 0.05) with the percentage reduction in self-rating anxiety scores. Bromazepam plasma concentration measurements showed greater variability than those of diazepam and were not found to be correlated significantly with clinical response. It is suggested that the use of strict diagnostic criteria (1978 draft of the third edition of Diagnostic and Statistical Manual of Mental Disorders), adequate sample sizes, and a 4-week study period gave increased sensitivity for the detection of significant differences between the two benzodiazepines.