A total of 12 chronically anxious psychiatric outpatients were treated with racemic propranolol (Inderal), diazepam (Valium), and placebo for one week each, using a balanced cross-over experimental design. Six patients had predominantly somatic anxiety, complaining mostly of bodily symptoms, and six had mainly psychic anxiety, complaining primarily of psychological symptoms. Clinical ratings of anxiety were made by patient and psychiatrist after each treatment. Though diazepam was in general more effective than propranolol or placebo in relieving anxiety, propranolol was more effective than placebo in patients with somatic anxiety but not in those with psychic anxiety. We suggest that propranolol should be reserved for patients whose anxiety symptoms are mainly somatic.