Evaluating the Big Five

Abstract

Widiger presents cogent evidence in support of dimensional over categorical models of the personality disorders, and we agree with him that a trait-based approach to clinical assessment offers attractive diagnostic advantages. We do not agree, however, that the five-factor model largely sufficient for characterizing normal and abnormal personality functioning, and we hope that the DSM-IV Personality Disorders Work Group will consider further Tellegen and Waller's (in press) reinterpretation and extension of the Big Five. To encourage consideration of this latter model-respectfully known as the Big Seven-we demonstrate, using data from the newly established Califomia Twin Registry, that Positive Valance and Negative Valence (Tellegen & Waller, in press), two dimensions of intrapersonal and interpersonal evaluation that are particularly relevant to Axis II disorders, are not accommodated by the five-factor model. First, we comment on why previous researchers have failed to find Positive Valence and Negative Valence dimensions in earlier lexical analyses. The English language includes thousands of adjectival descriptors of individual differences in expressive behavior. Thus, it is not surprising that numerous attempts have been made to winnow down this massive corpus into a manageable set of terms. Allport and Odbert (1936) made an early effort toward this end more than 50 years ago when, using Webster's Unabridged New International Dictionary, they separated the biophysical trait terms from the remaining lexical chaff. Almost 18,000 descriptors were classified as (a) personality traits, (b) temporary states, (c) social evaluations, or (d) metaphorical and doubtful terms. Future personologists were advised to focus on Category I terms, as these are the most objective, affectively neutral and more appropriate for science (Allport & Odbert, 1936, p. v). On the other hand, Category III terms, the emotionally charged evaluative terms, be avoided by psychologists (p. vii). Henceforth, thousands of descriptors of human behavior were banished from the field of personality. The rationally based lexical taxonomy of Allport and Odbert (1936) was unavoidably influenced by the subjective predilections of the two authors. Fortunately, some of the exiled evaluative terms have found their way back into the accepted parlance of our field. For example, Agreeable, a commonly used label for one of the Big Five dimensions, was originally considered too evaluative to qualify as a useful trait designator. Other illegal immigrants from Category III include ambitious, amiable, and amoral. What other useful terms, or dimensions, might be found if we took an unbiased look at the evaluative chaff of the early lexographers? Tellegen and Waller (in press) proposed that two large dimensions of Self-Evaluation and Other Evaluation can be recovered from the personality lexicon if unnecessarily restrictive exclusion criteria are not employed in the initial culling of trait terms. These dimensions are labeled Positive Valence and Negative Valence, and they cannot be located within the dimensional hyperspace of the five-factor model. Positive Valence is defined by terms such as excellent, outstanding, impressive, and superior versus ordinary, whereas Negative Valence is defined by terms such as cruel, mean, evil, and awful versus decent. Tellegen, Grove, and Waller (1991) recently developed a provisional 161-item measure of the Big Seven that includes measures of Positive Valence, Negative Valence, and five other higher order dimensions that are similar to, although not isomorphic with, the Big Five. We have collected data on this inventory from a large sample of twins and twin family members who are participants of the California Twin Registry. These data have allowed us to evaluate directly the veracity of several key statements made in the target article. Widiger suggests that Positive Valence and Negative Valence maladaptive variants of the existing five factors, that Positive Valence suggest excessively low neuroticism, and that Negative Valence may suggest extreme antagonism. We believe that the five-factor model cannot accommodate these newly identified dimensions. Table 1 lists the NEO Personality Inventory facet labels (from Widiger's Table 2) and the best match marker items from our Big Seven inventory. We could not find ideal items for all facets, although most facets were well represented in our inventory. We have also included several marker items of Positive Valence and Negative Valence, and we have filled out the Conscientiousness factor with several terms marking unconventionality. We do not believe that these markers represent the definitive set of either the Big Five or the Big Seven. The 30 items selected, however, should allow us to discriminate provisionally between the two competing models. As part of a larger study, we collected responses to the items in Table 1 from 542 female and 210 male participants of the Califomia Twin Registry. The individuals of this community-dwelling sam-