Problems in Assuming the Comparability of Pretest and Posttest in Autoregressive and Growth Models

In my relationship with Lee Cronbach I have never completely gotten over my initial perception, in which I was the young outsider building upon the work of a highly respected and already established scholar. Although the calendar of birth makes me only 7 months his junior, the 7 years between our Ph.D.'s (his 1940, mine 1947) better expresses my perception (the accuracy of which would be supported by citation analysis). Here is a chronology of examples:My own dissertation (1947) was on social attitude measurement, and, while independent, was done in the Berkeley environment which was producing the "authoritarian personality" studies (Adorno, Frenkel-Brunswik, Levinson, & Sanford, 1950; Frenkel-Brunswik, Levinson, & Sanford, 1947). When I became aware of Cronbach's (1946; see also 1950) work on response sets, I immediately regretted that each of my 25 scales had 5 items (two pro and three anti) and thus were unbalanced for direction-of-wording effects. The implications for the Frenkel-Brunswik, Levinson, and Sanford studies were more serious. Response sets exaggerated the relationship between the F scale of authoritarian personality trends and the Ethnocentrism (E) scale, since all items in both scales were in the anti direction. This artificially raised the correlation between prejudice and personality, and artificially lowered the correlation with political and economic conservativism (the PEC scale) which contained a mixture of both liberal and conservative items (in one widely used form, 9 liberal and 5 conservative).