No twin study has previously investigated the whole range of personality disorders (PDs) recorded by interviews. Based on twin and patient registries, 92 monozygotic (MZ) and 129 dizygotic (DZ) twin pairs were interviewed with the Structured Clinical Interview for DSM-III-R Personality Disorders (SCID-II). Observed prevalence rates from a normal population study of more than 2,000 individuals were used in combination with data from the present study to generate statistics assumed to be valid for a normal twin population, and these statistics were used for structural equation modeling. The best-fitting models had a heritability of .60 for PDs generally, .37 for the eccentric (A) cluster, .60 for the emotional (B) cluster, and .62 for the fearful (C) cluster. Among the specific PDs, the heritability appeared to be .79 for narcissistic, .78 for obsessive-compulsive, .69 for borderline, .67 for histrionic, .61 for schizotypal, .57 for dependent, .54 for self-defeating, .29 for schizoid, .28 for paranoid, and .28 for avoidant PDs. The best-fitting models never included shared-in-families environmental effects. However, a model with only shared familial and unique environmental effects could not be ruled out for dependent PD. Shared familial environmental effects may also influence the development of any PD and borderline PD. Passive-aggressive PD did not seem to be affected by genes or family environment at all. The low occurrence of antisocial PD in the twin sample precluded any model for this disorder. PDs seem to be more strongly influenced by genetic effects than almost any axis I disorder, and more than most broad personality dimensions. However, we observed a large variation in heritability among the different PDs, probably partly because of a moderate sample size and low prevalence of the specific disorders.