The reliability of DSM-III anxiety disorder diagnoses was determined using a new structured interview, the Anxiety Disorders Interview Schedule (ADIS). Two interviewers examined 60 consecutive outpatients at an anxiety disorders clinic and assigned primary and secondary diagnoses based on the ADIS. The kappa statistic, calculated on the basis of perfect matches on primary diagnoses, indicated good agreement for anxiety, affective, and adjustment disorders, as well as for the specific anxiety disorder categories of agoraphobia, panic, social phobia, and obsessive-compulsive disorder, but not for generalized anxiety disorder. We evaluated the causes for diagnostic disagreement, particularly in relation to the difficult differentiation between generalized anxiety disorder and other anxiety disorders.