Rigid and extreme: a geometric representation of personality disorders in five-factor model space.

Personality disorder rigidity and extremity can be geometrically defined and operationalized within the 5-factor model (FFM) of personality. A series of geometric and substantive assumptions were derived and then tested in samples of college students (N = 1,323) and psychiatric patients (N = 86). Normal and disordered personalities were found to coexist in a variety of regions of the FFM multivariate space. Within regions, the profiles of normal and disordered personalities were very similar in characteristic configuration but notably different in profile variability. Personality-disordered individuals tended to be located in the perimeters or outer regions of the FFM space, as indicated by their longer vector lengths. These findings generalized across 2 measures of personality disorders and across 2 measures of normal personality traits.

Personality disorders and the five-factor model: a test of facet-level predictions.

We tested predicted relationships (Widiger, 1993; Widiger, Trull, Clarkin, Sanderson, & Costa, 1994) between personality disorder scores and facets of the five-factor model, and evaluated the relative benefits of facet-level analyses over domain-level analyses. Data from 614 undergraduates indicated: (a) 63% of the predicted facet relationships were significant, although many unpredicted relationships also emerged; (b) facet-level analyses did not yield substantially stronger effect sizes than domain-level analyses; but (c) facet-level analyses provided much better discrimination between personality disorders than domain-level analyses. Facets of the openness to experience domain also helped discriminate between personality disorders, which is in contrast to previous domain-level findings that openness is not important.

A test of models of personality disorder configuration.

In this study, the authors examined the degrees to which various models of personality disorder (PD) configuration are consistent with the primary data sets from clinical and community samples reported in the recent literature. Factor analyses were conducted on PD intercorrelation matrices, and the loading matrices were rotated to maximum possible fit with target matrices representing the PD configuration models. There was little support for the interpersonal circle or other circular orderings of PDs, or for T. Millon's (1990, 1996) biosocial learning theory. There was moderate support for the Diagnostic and Statistical Manual of Mental Disorders (4th ed., DSM-IV; American Psychiatric Association, 1994) dimensions, for C. R. Cloninger's (1987) tridimensional theory, and for S. Torgersen and R. Alnaes's (1989) decision tree. There was consistent, stronger support for the 5-factor model (T. A. Widiger, T. J. Trull, J. F. Clarkin, C. Sanderson, & P. T. Costa, 1994) and for an empirically derived 7-factor model by C. R. Cloninger and D. M. Svrakic (1994).

The big five factors of personality and their relationship to personality disorders.

Articles examining the relationship between the Big Five factors of personality and personality disorders (PDs) are reviewed. A survey of these studies indicates that there is some agreement regarding the relationship between the Big Five and PDs. However, the level of agreement varies and may be a function of instrumentation, the method of report, or how data have been analyzed. Future research should consider the role of peer-ratings, examine the relationship between PDs and the first-order factors of the Big Five, consider dimensions over and above the Big Five as predictors of PDs.

Correlational structure of the MCMI-III personality disorder scales and comparisons with other data sets.

We examine the correlational structure of the personality disorder (PD) scales from the MCMI-III (Millon, 1994) among 614 college students. The correlation matrix from our nonclinical sample was highly similar to the clinical sample matrices reported by Millon (1994). Further analyses revealed that the correlation matrices from a variety of MCMI data sets are generally similar to one another, but are only moderately similar to PD correlation matrices based on other assessment techniques. PD correlation matrices based on different assessment techniques are generally not very similar to one another. Two-, 3-, and 4-factor solutions for the MCMI-III PD scales are reported and provide a framework for integrating apparently conflicting findings from previous work. The 4-factor solution was most meaningful and was consistent with the 5-factor model of PD.

Factor structure of the Beck Hopelessness Scale.

The factor structure of the Beck Hopelessness Scale (BHS; Beck & Steer, 1988) was examined among 411 outpatients (male = 132, female = 272, unknown = 7). Varimax-rotated principal-components analysis extracted three factors greater than unity that accounted for 40.4%, 6.9%, and 5.6% of the variance. Because the structure was invariant across factor rotations and levels of hopelessness severity, it is concluded that the BHS consists of three factors (Expectations of Success, Expectations of Failure, and Future Uncertainty).