A study was conducted to compare two methodologies for establishing hierarchical relationships--the Dayton and Macready model and the White and Clark procedure--using criterion referenced test data. Several phonics and structural analysis skills were selected from a reading skills battery. The opinions of reading "experts" were solicited in order to establish the hypothesized hierarchy. Both methodologies were then applied to validate the connections. The hypothesized hierarchies were then compared to those produced using empirical data based on the administration of four criterion referenced test levels to approximately 14,000 elementary school children. Specification of the hierarchies based on empirical data via either probabilistic model produced more stable results than those hypothesized by the content specialists. (Author)