REEXAMINATION OF THE EVIDENCE FOR THE DSM-III PERSONALITY-DISORDER CLUSTERS

The 11 personality disorders are organized clinically into three categories, or clusters. Results from recent factor analytic studies have been interpreted as providing empirical support for the three cluster model. These studies, however, have typically subjected only one measure obtained from a single sample to factor analysis, weakening the reliability and generalizability of the results. Different studies have also used different factor analytic procedures, which renders comparison across studies difficult. None of the studies have formally tested their results against a hypothetical three cluster model. In this study, published correlational matrices of the 11 personality disorders generated from differing samples as well as scales were subjected to a common set of factor analytic procedures. Each factor solution was then compared statistically to a hypothetical factor structure matrix representing the three cluster model using congruence coefficients. All five samples revealed statistically significant congruence with the hypothetical matrix, although there were several instances where individual scales did not correspond well. This was especially so for Cluster C.