Journal of Personality and Social Psychology1990, Vol. 58, No. 1,111-121Copyright 1990 by the American Psychological Association, Inc.0022-3514/90/SO0.75Psychoticism as a Dimension of Personality: A Multivariate Genetic Testof Eysenck and Eysenck's Psychoticism ConstructA. C. HeathDepartment of Human GeneticsMedical College of VirginiaN. G. MartinQueensland Institute of Medical ResearchHerston, Queensland, AustraliaIn this study, we applied multivariate genetic analysis, a generalization of factor analysis and behaviorgenetic analysis, to responses to items of the Psychoticism scale of the Eysenck Personality Question-naire by 2,903 adult same-sex Australian twin pairs. Item loadings on genetic, shared environmental,and nonshared environmental common and specific factors were estimated. The genetic factor struc-ture differed considerably from the environmental structures, particularly in men. The genetic corre-lation between suspiciousness items and items reflecting unconventional or tough-minded attitudesor hostility to others was negative, but the environmental correlation was positive. Thus, conven-tional behavior genetic studies that have reported significant heritability of psychoticism, on thebasis of analyses of scale scores, are misleading as to what trait is being inherited.In the Eysenckian personality scheme (H. J. Eysenck, 1981;H. J. Eysenck & S. B. G. Eysenck, 1976; H. J. Eysenck & M. W.Eysenck, 1985), psychoticism constitutes the third personalitydimension, orthogonal to extraversion and neuroticism. It isconceptualized as a continuum of liability to psychosis (princi-pally schizophrenia and bipolar affective disorder) with "psy-chopathy" (i.e., anti-social behavior) defined as "a halfway stagetowards psychosis" (H. J. Eysenck & S. B. G. Eysenck, 1976, p.203). Thus, schizophrenics, bipolars, and psychopaths areviewed by H. J. Eysenck and S. B. G. Eysenck (1976) as beingdifferent only in degree, rather than qualitatively, from normals,with the single personality dimension of psychoticism differen-tiating normals from psychopaths (intermediate in psychoti-cism) and from schizophrenics and bipolars (extreme in psy-choticism). Self-report questionnaire scales have been devel-oped (H. J. Eysenck & S. B. G. Eysenck, 1975, 1976; S. B. G.Eysenck, H. J. Eysenck, & Barrett, 1985), including a children'sversion (H. J. Eysenck & S. B. G. Eysenck, 1975), that attemptto measure psychoticism. High scorers on the Psychoticism (P)scale are conceptualized as "cold, impersonal, lacking in sym-pathy, unfriendly, untrustful, odd, unemotional, unhelpful, an-tisocial, lacking in insight, strange, with paranoid ideas thatData analysis for this article was supported by National Institutes ofHealth Grants GM3O25O and AG04954 and by Alcohol, Drug Abuse,and Mental Health Administration Grants AA06781, AAO7535,AA07728, DAO5588, and MH40828. Data collection was funded bygrants to N. G. Martin, J. D. Mathews, and J. B. Gibson from the Aus-tralian National Health and Medical Research Council.We acknowledge the roles of J. D. Mathews in the establishment ofthe Australian Twin Register and of Marilyn Olsen and R. Jardine inthe preparation of the data. We are grateful to our colleagues LindonEaves, John Hewitt, and Mike Neale and to anonymous reviewers fortheir helpful comments.Correspondence concerning this article should be addressed to A. C.Heath, who is now at the Department of Psychiatry, Washington Uni-versity School of Medicine, 4940 Audubon Avenue, St. Louis, Missouri63110.people were against him" (H. J. Eysenck & S. B. G. Eysenck,1976, p. 47). Prisoners, schizophrenics, alcohol and drug abus-ers, and children reporting many antisocial behaviors have allbeen reported to show elevated P scores (reviewed in H. J. Eys-enck & S. B. G. Eysenck, 1976).Key evidence cited by H. J. Eysenck and S. G. B. Eysenck(1976) in support of the psychoticism construct is derived fromgenetic studies. Family and adoption studies (e.g., Heston, 1966;Kety, Rosenthal, Wender, & Schulsinger, 1968; Mednick, Schul-singer, Higgins, & Bell, 1974; Odegard, 1963) have found in-creased risk of psychopathy, criminality, alcoholism, variouspersonality disorders, not just schizophrenia in the biologicalrelatives of schizophrenics. Unfortunately, such data are opento alternative interpretations. H. J. Eysenck and S. B. G. Eys-enck's interpretation is that they provide evidence for a com-mon dimension of genetic predisposition to psychosis and psy-chopathic or antisocial behavior and that it is this dimension ofheritable liability that is assessed by the P scale. Equally plausi-ble is the interpretation that predispositions to schizophrenia,antisocial personality, and other disorders are biologically inde-pendent but that there is cross-assortative mating (e.g., an in-creased likelihood that a schizophrenic woman will becomepregnant by an antisocial man; Gottesman & Shields, 1976;Mednick, 1974).The assumption of a single personality continuum that deter-mines liability to psychopathy, schizophrenia, and bipolar dis-order; the heterogeneous nature of groups scoring high on theP scale; and the relatively weak evidence that schizophrenicsobtain higher scores have stimulated considerable criticism ofthe P scale (Bishop, 1977; Block, 1977a, 1977b). Nonetheless,when traditional factor analytic and behavioral genetic meth-ods, as commonly applied in personality research (e.g., Eaves,Eysenck, & Martin, 1989; Fulker, 1981), have been used withthe P scale, results have been broadly consistent with H. J. Eys-enck and S. B. G. Eysenck's predictions. Factor analyses of theEysenck Personality Questionnaire (EPQ) that estimate fourcommon factors are usually able to recover a fourth factor withHI112A. C. HEATH AND N. G. MARTINloadings consistent with the psychoticism construct (e.g., H. J.Eysenck&S.B.G.Eysenck, 1976;S.B.G.Eysencketal., 1985;Heath, Jardine, Eaves, & Martin, 1988): suspiciousness ("para-noid") items, tough-minded or hostile ("antisocial") items, anditems relating to unconventional attitudes and behavior all havepositive loadings on the Psychoticism factor. It is only whenmore than four factors are extracted that psychoticism breaksdown into several factors, which are only modestly correlatedin an oblique solution (Heath et al., 1988). The large-sampletwin studies that have obtained data on the P scale have re-ported that P scores are moderately heritable (Eaves & H. J.Eysenck, 1977; Eaves & Young, 1981; Eaves etal., 1989;