PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN Reisenzein, Schimmack / AFFECT STRUCTURE RESEARCH Similarity Judgments and Covariations of Affects: Findings and Implications for Affect Structure Research

Results from four studies document very high agreement between the raw data matrices of affect similarity judgments and correla-tions computed from self-ratings. This was found both for the global structural level, where the correlation-similarity corre-spondence was as high as the average correspondence of correla-tion matrices from different studies, and for local subgroups of affects, provided that reliable measures were used. The findings suggest that self-rating correlations and similarity judgments of affects contain essentially the same structural information. As a consequence, the two data types are equally valid starting points for structural analyses of affect, and differences between the models of affect that have been derived from them may reflect pri-marily method artifacts, such as the use of different scaling meth-ods, rather than differences in the original data. Possible expla-nations for the high correspondence between self-rating correlations and similarity judgments of affect are discussed. During the past decades, a large body of research has accumulated whose aim is to clarify the structure—that is, the constitutional makeup and/or the interrela-tions—of affects.1 Two of the major empirical approaches that have been used to address these prob-lems are covariance structure analysis and the analysis of similarity judgments. The first approach consists of the analysis of the intercorrelations among self-ratings of affective states, typically by means of exploratory factor analysis (for reviews, see, e.g., Larsen & Diener, 1992; Lorr, 1989; Watson & Tellegen, 1985). The second approach consists of the analysis of laypeople’s judg-ments of the similarities of affects, typically by means of multidimensional scaling or hierarchical clusterin