This study investigates the extent to which the higher order factors of symbolic and semantic content are differentiated. The main concern is an expository description of the use of confirmatory factor analysis in establishing factorial validity. Reanalyzed data originally collected in 1968 involved a sample of 197 tenth, eleventh and twelfth grade gifted students from a middle-class urban school in Los Angeles. Confirmatory maximum likelihood factor analysis with the LISREL V computer program was used to test the goodness-of-fit to the data of three alternative models of higher order intellectual ability factors: a model of two higher order factors of symbolic and semantic content; a model of one general factor; and a null model of no factor structure. Three different matrices were specified: lambda pattern, phi, and a theta delta. Five procedures for assessing the fit of the proposed models included: (1) examination of LISREL V analysis results; (2) goodness-of-fit index; (3) root mean square residual; (4) Tucker-Lewis reliability coefficient, rho; and (5) a comparison of nested models. The two factor model appears to provide a better data fit than both the one factor or null model. The hypothesis that a previous differential correlation was inflated by transformations tests was supported. (PN)