Statistical models of implicit personality theory: A comparison.

Evaluated factor analytic (FA), cluster analytic (CA), and multidimensional scaling (MDS) models of implicit personality theory in a methodological study that compared statistically derived dimensions and categories of traits and stimulus persons with S-generated groups of traits and stimulus persons. 60 undergraduate and graduate students rated 10 familiar and 10 unfamiliar people on 20 diverse traits and, from these traits and persons, formed groups of related traits and groups of similar people. The trait ratings were subjected to FA, CA, and MDS procedures, computed for each S individually. Comparison of the direct groups with the derived FA and MDS dimensions and CA categories, using 4 different match indices, indicated that the CA model provided the best match. A tendency toward better matching for traits than persons was observed, as well as greater differentiation in cluster space for familiar than unfamiliar persons. It is concluded that when Ss categorize rather than dimensionalize stimulus objects, a model such as CA may be more appropriate. Problems associated with testing the complementary hypothesis—that a dimensional model is more appropriate when perceivers dimensionalize—are discussed. (33 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)