Journal of Personality and Social Psychology1994, Vol. 66, No. 4, 762-775Copyright 1994 by the American Psychological Association, Inc.0022-3514/94/$3.00Testing a Model for the Genetic Structure of Personality: A Comparisonof the Personality Systems of Cloninger and EysenckA. C. Heath, C. R. Cloninger, and N. G. MartinGenetic analysis of data from 2,680 adult Australian twin pairs demonstrated significant geneticcontributions to variation in scores on the Harm Avoidance, Novelty Seeking, and Reward Depen-dence scales of Cloninger's Tridimensional Personality Questionnaire (TPQ), accounting for be-tween 54% and 61 % of the stable variation in these traits. Multivariate genetic triangular decompo-sition models were fitted to determine the extent to which the TPQ assesses the same dimensions ofheritable variation as the revised Eysenck Personality Questionnaire. These analyses demonstratedthat the personality systems of Eysenck and Cloninger are not simply alternative descriptions of thesame dimensions of personality, but rather each provide incomplete descriptions of the structure ofheritable personality differences.Much research has been published on the factorial structureof personality, and many theories of personality structure havebeen propounded (e.g., Eysenck & Eysenck, 1969, 1976; Gray,1982; McCrae & Costa, 1989; Tellegen, 1985). Much of thiswork has used samples of unrelated individuals. However, agrowing body of data indicates an important contribution ofgenetic factors to personality differences (Eaves, Eysenck, &Martin, 1989; Loehlin, 1992). Support for the role of geneticfactors comes from studies using a variety of personality assess-ments, including studies of separated twins (Bouchard, Lykken,McGue, Segal, & Tellegen, 1990; Pedersen, Plomin, McClearn,& Friberg, 1988; Shields, 1962; Tellegen et al., 1988), of adopt-ees and their families (Loehlin, 1981; Loehlin, Willerman, &Horn, 1985; Scarr, Webber, Weinberg, & Wittig, 1981), of twinpairs reared together (Eaves et al., 1989; Rose & Kaprio, 1988;Rose, Koskenvuo, Kaprio, Sarna, & Langinvainio, 1988), andof twin pairs and their parents, siblings, and adult children(Eaves, Heath, Neale, Hewitt, & Martin, 1993), and using rat-ings by informants (Heath, Neale, Kessler, Eaves, & Kendler,1993) as well as self-report data. Although a genetic influenceon personality has been demonstrated for a wide range of per-sonality scales, attempts to formulate a parsimonious model forA. C. Heath, Department of Psychiatry, Department of Psychology,and Department of Genetics, Washington University School of Medi-cine, C. R. Cloninger, Department of Psychiatry and Department ofGenetics, Washington University School of Medicine; N. G. Martin,Queensland Institute of Medical Research, Brisbane, Queensland, Aus-tralia.Data collection was supported by Alcohol, Drug Abuse, and MentalHealth Administration (ADAMHA) Grant AA07728 and by a grantfrom the Australian National Health and Medical Research Council.Data analysis was also supported by ADAMHA Grants AA03539,AA07535, DA05588, MH31302, and MH40828.We acknowledge the assistance of Sue Mason, Ann Eldridge, OliviaZhang, and the late Ulrich Kehren and the helpful comments of Greg-ory Carey and Kay Phillips.Correspondence concerning this article should be addressed to A. C.Heath, Department of Psychiatry, Washington University School ofMedicine, 4940 Children's Place, St. Louis, Missouri 63110.the underlying genetic structure of personality have been muchrarer than descriptions of the phenotypic structure. Becausequite different genetic and environmental factor structures mayunderlie the observed phenotypic structure of personality (e.g.,Heath & Martin, 1990; Kendler, Heath, Martin, & Eaves,1987), performing separate genetic analyses of personality fac-tors defined by phenotypic, that is, within-person, correlationsmay be quite misleading about the underlying genetic structureof personality (Heath & Martin, 1990; Heath, Neale, Hewitt,Eaves, & Fulker, 1989).Cloninger (1986,1987,1988,1991) has proposed a three-fac-tor model for the structure of personality, which is hypothesizedto have a close correspondence with the underlying geneticstructure, and has developed a self-report questionnaire, theTridimensional Personality Questionnaire (TPQ; Cloninger,Przybeck, & Svrakic, 1991), to assess three postulated dimen-sions of personality: Harm Avoidance (HA), Novelty Seeking(NS), and Reward Dependence (RD). In this article, we use datafrom a mailed questionnaire survey of a large Australian twinsample to test whether scores on these personality dimensionsare indeed substantially influenced by genetic factors. We alsofit genetic models to self-report responses to the revised EysenckPersonality Questionnaire (EPQ-R; Eysenck, Eysenck, & Bar-rett, 1985) obtained from the same subjects in the same ques-tionnaire. The Eysenckian personality dimensions have beenextensively studied in genetic studies using twins and their fam-ily members (Eaves et al., 1989; Eaves et al., 1993), separatedtwin pairs (Pedersen et al., 1988), and adoptees (Scarr et al.,1981) and thus provide a basis for comparing the results fromthis study with results of other behavioral genetic surveys usinga variety of different research designs. We fit multivariate ge-netic models (Heath, Neale, et al., 1989; Martin & Eaves, 1977;Neale & Cardon, 1992) to determine the extent to which theTPQ assesses the same dimensions of genetic variation as theEPQ-R (Eysenck & Eysenck, 1976; Eysenck et al., 1985) andthe extent to which it assesses new dimensions of genetic vari-ability (cf. Martin, Eaves, & Fulker, 1979). The methods that weuse have broad applicability in research aimed at simplifyingour understanding of the structure of personality from an etio-logic perspective.762TESTING A MODEL OF GENETIC STRUCTURE763MethodSampleData for this report derive from the 1989 follow-up mailing ofthe Australian National Health and Medical Research Council(NH&MRC) volunteer adult twin panel. In 1980-1981, a 12-page ques-tionnaire was mailed to 5,967 adult twin pairs aged 18-88 (Eaves et al.,1989; Martin & Jardine, 1986). Completed questionnaires were re-turned by both members from 3,808 twin pairs (64% pairwise responserate) and by one twin only from 576 pairs (69% individual responserate). In 1988-1989, twin pairs in which both twins had cooperated inthe 1981 survey were mailed a follow-up questionnaire, with remailingsand telephone follow-up of nonrespondents.