Sixty-five patients with social phobia were treated in a study that compared a cognitive-behavioral group treatment program with pharmacotherapy with alprazolam, phenelzine sulfate, or pill-placebo plus instructions for self-directed exposure to phobic stimuli. Statistically significant repeated-measures effects were shown on all measures, indicating that the treatments studied were associated with substantial improvements in patients with severe and chronic social phobia. Patients who were treated with phenelzine were rated by clinicians as more improved on a measure of work and social disability than patients who were treated with alprazolam or placebo (patients in the cognitive-behavior therapy group were not rated on this measure). Subjects showed positive cognitive changes from before to after treatment, and there were no differences between treatment groups on the cognitive measure. We discuss the implications of these findings within the context of demographic and clinical predictors of response.