Ego Development: Questions of Method and Theory

Beginning where my odyssey began, Jackson raises some issues that had been concerned about but which are not adequately covered in my article. He repeats some of my list of possible target traits as set down prior to gathering data with the Family Problems Scale (see Loevinger, Sweet, Ossorio, & LaPerriere, 1962, pp. 67-68). Those putative traits guided our invention of items. They were not our names for traits that were measuring, however, but rather for hypotheses that hoped to test. He is correct that the names are biased, but they were biased in the direction favored by child guidance clinicians of that time. For such reasons, said, we propose to seek scoring keys in the responses of our subjects rather than in our own (Loevinger et al., 1962, p. 59). No one begins with a tabula rasa. As it happened, none of our initial hypotheses were supported by our data. In place of punitiveness, found Authoritarian Family Ideology (AFI). The reason that the prominence in AFI of inability to conceptualize inner life is interesting is that were not looking for it, nor had hypothesized it before setting up the study and writing items. The authors of The Authoritarian Personality (Adomo, Frenkel-Brunswik, Levinson, & Sanford, 1950) also were turned toward the anti-intraception conception by their data, not by their preconceptions or research priorities. Jackson's suggestion that one should know whether the two statements in our paired-choice items are rigorously opposite, if he means by that contradictory, is incompatible with our aim to state each choice in the way that those holding the opinion would recognize or state it (i.e., incorporating their own rationalizations). The statements often cannot be logically opposite and psychologically opposite at the same time. Because maximal discriminating power is yielded by median items (items that cut the group as near to the middle as possible), items where one alternative is clearly wrong are almost always less valuable. The question of the robustness of our findings with other samples-in particular, other subcultural groups-is a good one. It was investigated by Ernhart (Ernhart & Loevinger, 1969). Although our original homogeneous keying project had found several other clusters of items, only AFI was clearly cross-validated. As her sample was large and diverse, Ernhart was able to study the homogeneity of AFI within cells of her analysis-of-variance study. The fact that the within-cell value of the homogeneity coefficient was almost as high as the homogeneity coefficient for the entire sample shows the robustness of this variable. How many objective personality tests can meet such a criterion? I do not retract the statement that factor analysis of dichotomous items is psychologically meaningless. The fundamental equation of factor analysis states that each variable (here, item) can be expressed as the weighted sum of a set of factors. The weights characterize the item; the factor scores characterize the person. How can the score on a dichotomous item (i.e., 0 or 1) be decomposed into the weighted sum of factor scores? It is mathematically absurd, besides depicting a process hard even to imagine. Sorting items into homogeneous clusters, as did, is more meaningful, or at least less absurd.