Abstract
This study examined the problems in using product-moment correlations for the 2 general functions of factor analysis, identification (or verification) and use of psychological constructs embedded in bipolar ratings. For illustration, a set of fictitious semantic differential ratings, with known structural similarities and differences between concept profiles, were factored. The results were assessed against 12 criteria for evaluation derived from simultaneous considerations of issues in statistics, measurement, and psychosemantics. It was concluded that solutions from factor analysis of r coefficients do not identify or verify underlying attributional relationships among concepts implicit in subjects' cognitive processes and manifest in ratings. An alternative index, Beta Coefficient, was then used to illustrate its advantage in analysis of the hypothetical bipolar ratings. Implications of the present study are discussed.
Get full access to this article
View all access options for this article.
