The interrater reliability of the Structured Clinical Interview for DSM-III-R (SCID) was studied. Fifty-four audiotaped SCID interviews were rated independently by 3 raters. The highest interrater agreements were observed for schizophrenia (0.94), major depressive disorder (0.93), dysthymia (0.88), generalized anxiety disorder (0.95), panic disorder (0.88), alcohol use disorder (0.96) and other psychoactive substance use disorder (0.85). The remaining diagnoses of mood and anxiety disorders obtained acceptable interrater agreement (0.70-0.80), with an exception for obsessive-compulsive disorder (0.40). The poorest agreement was obtained for somatoform disorders ( -0.03). Lack of hierarchy in DSM-III-R allows for multiple Axis I diagnoses. Interrater reliability for multiple diagnoses was tested. Agreement was generally good for combinations of 2 diagnoses, and poorer when 3 diagnoses were combined. Our findings confirm that SCID yields highly reliable diagnoses. SCID is recommended for research on mental disorders.