Factor Structure on the Semantic Differential as a Function of Method of Analysis

Publisher Website
Google Scholar
Abstract
Researchers and critics of the semantic differential technique have recently cautioned about the problem of concept-scale interaction. The present study compared the factor structures obtained when the same data were factor analyzed in several ways: (a) for the total sample and by sex for collapsed concepts and (b) by individual concept for the total sample, by sex, and for internal and external locus of control groups. Three concepts (HURRICANE, MY CITY, and MYSELF) were analyzed which yielded 18 factor matrices. The semantic differential instrument was composed of 19 scales purporting to measure the dimensions of Evaluation, Potency, and Activity. In 16 of the 18 factor matrices, four of the five scales (fair-unfair, friendly-unfriendly, good-bad, and pleasant-unpleasant) expected to measure Evaluation had loadings of 1.301 or higher on Factor I. There was moderate consistency in scales which comprised the Potency dimension for several of the 18 analyses. There was little, if any, consistency in scale composition for the Activity dimension. If these findings can be generalized to other populations and concepts, it appears that the method of analysis, whether by concept or by population, is relatively unimportant for the Evaluation dimension (Factor I), but very important for other "meaning" dimensions.