How to Create a Personality Scale

Abstract

In this chapter, a general discussion challenges the way in which personality scales are commonly created and proposes an extended psycho–lexical approach. Standard procedures use statistical methods (e.g., factor analysis) at a very early stage and determine the final composition of the scale ex post by eliminating traits that do not “behave” well. The widespread practice of the post hoc reduction of traits is frequently criticized. Exploratory factor analyses are primarily mathematical approaches that are not based on a substantive theory. The extended approach tries to close this theoretical gap with a more intensive use of the dictionary. The rationale behind the demonstrated method is the insight that the factor loadings that determine the scale structure are based on a correlation among personality traits that can only occur when these traits are synonyms or near-synonyms. This chapter is divided into a theoretical part and an empirical part. In the theoretical part, factor structures were determined a priori by maximizing the cross ratio between the average number of items’ common synonyms within a factor and the average number of common synonyms outside this factor. This procedure translated the rules of convergent and discriminant validity into rules for semantic factor analysis in the extended lexical approach. In the empirical part of this chapter, the theoretically developed scales were analyzed through factor analyses using Aaker’s (1997) and Grohmann’s (2009) personality models. As could be expected from the a priori scales, Grohmann’s (2009) model had a much better fit than Aaker’s (1997).