Hierarchical multi-sample confirmatory factor analysis (CFA) was used to address several unresolved issues concerning the structure of the Cognitive Assessment System (Das & Naglieri, 1997), a new test of intelligence based upon the planning, attention, and simultaneous-successive (PASS) processes theory of human cognition. First, we investigated whether the CAS measures the same constructs from ages 5 to 17 years. Second, we examined the adequacy of the fit provided by the PASS model and the correlations among the PASS factors. Third, and finally, we compared the fit provided by the PASS model with several models that reflect plausible rival hypotheses of the CAS structure. Results of these analyses suggest that the CAS measures the same constructs across its 12-year age span. Nevertheless, although the correlated PASS model provided a better fit to the data than did several competing nonhierarchical models, it did not provide a particularly good fit. Moreover, an hierarchical model reflecting the implied theoretical structure of the CAS provided a marginal fit to the data and a statistically worse fit than a competing hierarchical model. The best-fitting model was a third-order factor model with psychometric g at the apex of the factor hierarchy, an intermediate combined Planning/Attention factor, and four first-order PASS factors. Based on these results, the CAS tests are better understood within the framework of Carroll's (1993) three-stratum theory as measures of processing speed (rather than planning and attention processes), memory span (rather than successive coding), and a mixture of fluid intelligence and broad visualization (rather than simultaneous coding). Results of our CFA of the standardization data do not support use of the CAS for differential diagnosis or for planning treatments based upon the PASS Scales.