The Diagnostic Interview Schedule for Children Version 2.1c (DISC-2.1c) and the YSR were administered to 289 homeless adolescents. Stepwise discriminant analysis identified YSR scales that contributed to predictions of DSM-III-R disorders. Paper-and-pencil prediction rules based on YSR "borderline" or "clinical" scores were evaluated. Statistically significant discriminant functions for disruptive disorders, depressive disorders, manic disorders, attention-deficit hyperactivity disorder, schizophrenia, and posttraumatic stress disorder, each based on a unique pair of YSR scales, produced overall hit rates of 0.66 to 0.90. Paper-and-pencil predictions produced comparable results. The weakest overall predictions were for the disruptive behaviors; the best rule ("If Aggressive or Delinquent is at least borderline, then predict oppositional defiant disorder or conduct disorder") produced a 0.72 hit rate. The strongest overall predictions were for schizophrenia; the best prediction rule ("If Thought Problems and Delinquent are at least borderline and at least one is clinical, then predict schizophrenia") produced a 0.87 hit rate. The study concludes that although the success rates reported in this study are specific to this sample, apparently the YSR has good ability to predict DSM-III-R diagnoses as determined by the DISC. Furthermore, the study shows that categorical diagnoses can be treated as locations or cluster sectors in a multidimensional space. 1 figure, 2 tables, and 16 references