The purpose of this study was to test the construct validity of the Behavior Assessment System for Children (BASC), Parent and Teacher Rating Scales (PRS and TRS). Six samples were considered, including the Normative General and Clinical Samples for each measure (Reynolds & Kamphaus, 1992). Another pair of samples were taken from a database of a Georgia hospital (PRS n = 130, TRS n = 108). The Normative Clinical Sample of TRS scores was multicollinear, and was not used.Five models were designed for each measure: a single factor solution, the theoretical model of the BASC, and three adaptations of the scoring system. Using AMOS, these models were fit to the samples. Only the theoretical model met minimum standards for adequate fit. Multi-sample analyses with different combinations of parameter restrictions were conducted to determine which aspects of the theoretical model's factor structure accounted for the most sample variance. When fit to both normative samples of PRS scores, all aspects of the factor solution were found to contribute. For all other runs, it was found that error, unique, and factor variances contributed the most to the factor solution. This suggests that the relationship of variables/scales to the factors/composites in this model could be improved. It is suggested that practitioners disregard composite scores, and that the authors/publishers of the BASC consider using regression weights to formulate composite scores in the scoring program.