Subjects with panic disorder (N = 23) were randomized into a crossover design involving diazepam, placebo, relaxation therapy, or control. Anxiety was measured by a 3-day hourly diary, psychological tests, and assessment of heart rate and skin conductance level during baseline, stress test, and interview. The greatest and only significant physiologic changes occurred with diazepam. The largest pre/posttreatment changes on the psychological tests and self-reported anxiety and depression occurred with relaxation, although the differences were not statistically significant. Relaxation therapy may be useful for treating the general anxiety associated with panic disorder.