Two confirmatory maximum likelihood factor analytic models of the Personality Research Form-E are presented and discussed for the men's and women's standardization data (Jackson, 1974) (N = 215). The two models investigated were: (1) a six-factor model based on Jackson's conceptual framework for the Personality Research Form-E; and (2) the five-factor empirical model advanced by Fowler (1985). Overall, the patterns of correlations within each data matrix are highly similar, but the factor patterns, factor intercorrelations, as well as the error variances, do differ. The fivefactor model was found to be more adequate generally, but produced no absolute fit to the data matrices, while the six-factor model invariably produced inaccurate estimates of the parameters. Comparisons of the completely estimated, Varimax-rotated factor loading matrices to that reported by Fowler found, on the whole, stronger similarity between matching factors for the women than for the men. The implications of this finding for the modelling of the PRF-E and its application in clinical contexts are considered at length.
Jackson (1 97 l), among others, has highlighted the roles of theory-based construction, multivariate (componential) analysis, and the need for minimizing response bias as ways of clarifying the definition of personality variables. This continuing lack of precise definition, which has been commented upon by any number of investigators in the field (e.g., Wolman, 1978), generally is recognized as a distinct impediment to the development of personality theory as well as to the construction of a personality language, which then could be used for mapping aspects of personality to other phenomena (e.g., family dynamics, cognitive and behavioral states).
Requests for reprints should be addressed to the author, at the Division of Child and Family Psychiatry,