Genetic and Environmental Influences onPersonality: A Study of Twins RearedTogether Using the Self- and Peer ReportNEO-FFI ScalesRainer Rlemann Alois AngleitnerUniversity of Bieleield, GermanyJan StrelauUniversity of Warsaw, PolandABSTRACT Previous behavior-genetic research on personality has beenalniosi exclusively based tin self-report questionnaire measures. The purposeof this research was to measure personahty constructs via self- and peerreports on tbe items of tbe NEO Five-Factor Inventory (Costa & McCrae.1989). The sample included 660 monozygotic and 200 same sex and 104opposite sex dizygoiic twin pairs. We collected self- and two independent peerreports for each of the twins. Our analysis of self-report data replicates earlierfmdings of a substantial genetic influence on the Big Five (/r = .42 to .56).We also found this influence for peer reports. Our results validate tindingsbased solely on self-reports. However, estimates of genetic contributions tophenotypic variance were substantially higher when based on peer reports (/i' =.57 to .81) or self- and peer reports (ir = .66 to .79) because these data allowedPreliminarj' analyses of the data were presented at the seventh meeting of the Interna-tional ScKiety for the Study of Individual Differences, Warsaw, Poland. July 15-19,1995, and in Angleitner, Riemann, Spinath. Hempel. Thiel. and Strelau (1995). Thisresearch was supported in part hy funds from the Max Planck Forschungspreis(awarded by the Alexaiider von Huinboldt Stiftung and the M;ix-Planck Gesellschaft)10 Alois Angleitner and Jan Strelau. We thank Peter BorkLMiau. Rainer Reisen/ein. andFrank M. Spinath for helpful comments, and Wolfgang Thiel for his assistance in ana-lyzing the data. CorrespKmdence conceming this article should he addressed to RainerRiemann. Department of Psychology. University of Bielefeld. Postfach 100131, 33501Bielefeld, Germany. E-mail may he sent to [email protected] of PersonaUty 65:3, September 1997.Copyright © 1997 by Duke University Press.Riemann et al.us to separate error variance from variance due to nonshared environmental in-fluences. Correlations between self- and peer reports reflected the same geneticinfluences to a much higher extent than identical environmental effects.Humati behavior-genelic research has provided important itisights intothe etiology of individual differences in personality traits (Loehlin,1992; Plotnin & McClearn, 1993; Rose, 1995; Rowe, 1994). This re-search has focused most strongly on Extraversion and Neuroticism asthe two major dimensions of personality. In recent years, however, thefive-factor model has gained a prominent position within personalityresearch (for reviews see Digman, 1990; John, 1990; John, Angleitner,& Ostendorf, 1988: McCrae, 1989). Ntimerous studies have shownthat five broad factors—Extraversion or Surgency, Agreeableness,Conscientiousness, Emotional Stability or Neuroticism, and Culture/Intellect or Openness to Experience—account for much of the com-mon variance in personality rating and questionnaire data. They aremeaningful dimensions for the description of individual differences inpersonality and robustly found in various sources of data. The NEOPersonality Inventory-Revised (NEO-PI-R; Costa & McCrae, 1992)and the short version of the NEO-PI, the NEO Eive-Factor Inventory(NEO-FFI; Costa & McCrae, 1989) are now standard questionnairemeasures of the five-factor model.Four major research designs that are used in human behavior-genetic studies can be distinguished: (a) studies of twins reared to-gether in the same family, (h) studies of twins who have been reared indifferent families, (r) studies of adoptive families (ideally includingmeasures of the adoptees' biological parents), and (d) family studiesanalyzing the data of participants who are genetically related to varyingdegrees. All of these designs have their particular strengths and weak-nesses, and firm conclusions regarding the etiology of individual dif-ferences are based on an integrative analysis of data based on all fourdesigns (e.g., Bouchard & McGue, 1981; Loehlin, 1992). The mostpopular design, the comparison of monozygotic (MZ) and dizygotic(DZ) twins reared together, requires several assumptions to estimategenetic and environmental effects on traits: (a) There are only twotypes of twins (MZ and DZ) and DZ twins share on average 50% oftheir genes (i.e., there are no effects of assortative mating),' (b) twins1. There is some evidence of assortative mating for Openness and Conscientiousness(see McCrae, 1996). A.ssortative mating increases the correlation of additive andTWin Study of NEO-FFI Scales 451are a representative sample of the population to which one wishes togeneralize, (c) the effects of environmental influences shared by twins isno greater for MZ than for DZ twins ("equal environments" assump-tion), and (d) gene-environment correlation and interaction have a neg-ligible effect on the trait under study. Finally, studies of twins rearedtogether do not allow simultaneous estimation of environmental eflectsshared by the twins and nonadditive genetic influences (e.g., due togenetic dominance). The validity of these assumptions is thoroughly dis-cussed in the behavior-genetic literature (see, e.g.. Eaves, Eysenck, &Martin. 1989; Loehlin, 1992). From this discussion it can be concludedthat studies of twins reared together provide a valuable tool for behavior-genetic research. The strength of this design is that large samples areavailable representing the whole range of genotypes and environments.Combining data from twin, adoption, and family studies into anextremely large data set, Loehlin (1992) thoroughly reviewed the thenavailable behavior-genetic research on the five-factor model of person-ality. All of the studies summarized by Loehlin used self-report question-naires that, with one exception (Bergeman et al., 1993), had not beenconstructed with the goal of measuring the Big Five. Instead, Extra-version and Neuroticism were frequently measured by the EysenckPersonality Inventory (EPI; Eysenck & Eysenck, 1964), the EysenckPersonality Questionnaire (EPQ; Eysenck & Eysenck, 1975), the EPQ-Revised (Eysenck & Eysenck, 1991), or short versions of one of thesequestionnaires. Measures related to Agreeableness comprised scalesdeveloped to measure masculinity versus femininity, aggression, altru-ism, empathy, and related constructs. Conscientiousness was indexedby measures of, for example,