The objective of the present study was to determine the reliability of psychiatric diagnoses using a translation and adaptation of Portuguese of the "Structured Clinical Interview for DSM-III-R-patient version" (SCID-P) and the "Structured Clinical Interview for DSM-III-R Personality Disorders" (SCID-II), using the joint interviews methodology. Thirty-nine subjects were evaluated using the SCID-P and 20 of them using the SCID-II. Interrater reliability was analyzed statistically by means of the Kappa Coefficient. Agreement between results obtained with SCID-P was statistically significant for the major diagnostic categories of DSM-III-R and for 10 of the 12 specific diagnostic categories studied (a minimum of 4 subjects per diagnosis). Agreement was not statistically significant for Psychotic Disorder Not Otherwise Specified (NOS) and for Other Bipolar Disorder. The Weighted Kappa for the main diagnoses and the Overall Kappa for the entire set of 25 specific diagnostic categories proposed by the SCID-P were statistically significant. The general agreement for Personality Disorders with SCID-II was statistically significant. The Kappa Coefficient was determined for the Avoidant, Paranoid, Histrionic and Borderline Personality Disorders and for the Conduct Disorder. The remaining Personality Disorders were not analyzed statistically because of their low prevalence in the sample. Agreement was not significant only for the Histrionic Personality Disorder. These data suggest that the translation and adaptation of the SCID-P and SCID-II to Portuguese presents, in general, good reliability indices, and thus its use is recommended.