A large reliability study of DSM-III-R anxiety disorders is reported in which outpatients (n = 267) received two independent structured interviews (Anxiety Disorders Interview Schedule-Revised). It is the only reliability study to date in which the final DSM-III-R criteria are used throughout the study. Reliability was assessed for each diagnosis when it was assigned as a principal diagnosis and when it was assigned as either a principal or an additional diagnosis. Excellent reliability was obtained for current principal diagnoses of simple phobia, social phobia, and obsessive-compulsive disorder. Agreement was good for panic disorder when all severity levels of agoraphobic avoidance were combined. Reliability was fair for generalized anxiety disorder. Remaining diagnostic difficulties, particularly in identifying levels of agoraphobic avoidance and in reliably diagnosing generalized anxiety disorder, are discussed in the context of changes in diagnostic criteria that are under consideration for DSM-IV.