Many children meet criteria for multiple Diagnostic and Statistical Manual (DSM) categories, such as oppositional defiant disorder (ODD) and conduct disorder (CD). If each disorder has been well validated as a separate entity, statistically significant co-occurrence of different disorders may be highly informative. However, ODD and CD have not been well validated as separate entities. The very high rate of overlap between diagnoses of ODD and CD may therefore arise artifactually from the lack of a valid distinction between them, rather than from potentially informative comorbidity between two different disorders. Empirical research strongly supports a distinction between two syndromes that correspond to subsets of the DSM-III-R criteria for CD. Designated as delinquent behavior and aggressive behavior, these syndromes have been found to differ in biological correlates, heritability, developmental stability, course, response to interventions, and long-term outcomes. At this stage of our knowledge, empirically based assessment and taxonomic methods can be especially useful for distinguishing between syndromes, deriving norms, doing longitudinal studies, and detecting patterns of comorbidity. These methods do not preclude categorical taxa, which can be formed by imposing cutpoints on the distributions of syndrome scores. Categorical taxa can also be formed by cluster analyzing profiles of syndrome scores. Accelerated longitudinal designs can bring multiple analyses to bear on empirically derived syndromes to test complex developmental relations more quickly and powerfully than can traditional longitudinal designs. Follow-ups of high-risk groups can identify variables that predict good versus poor outcomes if standardized baseline and outcome measures are used. Interventions for conduct problems should be designed to counteract the risk factors found to predict poor long-term outcomes.