Recently, short-form versions of structured psychiatric diagnostic interviews have been developed for epidemiologic and survey research. These short forms can reduce research costs in large-scale studies; however, their accuracy is likely to be less than that of a full diagnostic interview. We evaluated the positive and negative predictive values of a short-form interview derived from the Composite International Diagnostic Interview (CIDI). Samples of subjects who scored both positively (n = 277) and negatively (n = 136) on the CIDI Short Form for Major Depression (CIDI-SFMD) were administered the full depressive disorders section of the CIDI. Almost all subjects who were negative on the short form were similarly classified as not having major depression by the CIDI. Approximately 25% of subjects had false positive results; these subjects tended to be older and less educated than true positives. Approximately 75% of subjects scoring five or more on the CIDI-SFMD had major depression according to the full CIDI, and a proportion of the remainder had less severe depressive syndromes. Some CIDI- SFMD positive subjects may have had depressive symptoms attributable to organic or other etiologies excluded under the definition of major depression.