Selecting Representative Trait Adjectives

Abstract

A strong argument has been made for specifying the possible characteristics of personality by using trait adjectives from the natural language. This argument implies a rationale for the selection of variables to measure personality characteristics: One should select variables to be representative of trait adjectives in the natural language. In this study I argue that this rationale has never been carried through, and I make an attempt to do so. This requires defining a pool of trait adjectives, classifying them according to similarity of meaning, and selecting variables to represent this classification. Previous classifications have either led to no selection of variables or to a selection that is not representative. A pool of 571 terms was taken from three existing lists. These terms were classified into categories, and scales were selected to represent the categories. A first application of these scales was then made. Judges made ratings of similarity of meaning on 57 scales-each defined by two opposite trait-adjectives-for the corresponding 114 single terms. Factor analysis of the scales produced six factors. These showed some resemblance to the well-known Big Five factors, but with major differences. Instead of five factors of comparable size, the first three factors were predominant, representing 70% of the total variance. The Culture factor appeared as Intelligence; Emotional Stability as a small, only partly differentiated, factor. These differences may reflect both (a) differences between judgments about traits (internal structure) and judgments about people (external structure), and (b) representativeness (e.g., the previous nonrepresentation of intelligence and overrepresentation of Emotional Stability). © 1987 American Psychological Association.