A Behavioral Genetic Study of the Overlap Between Personality and Parenting
A Behavioral Genetic Study of the Overlap Between Personality and Parenting Frank M. Spinath University of Bielefeld Thomas G. O Connor Department of Child and Adolescent Psychiatry and Social, Genetic, and Developmental Psychiatry Research Centre, London ABSTRACT The current study had three aims. The first was to examine the covariation between personality of parents and parenting behaviors. The second aim was to examine the genetic and environmental influences on parenting behaviors. The third aim was to examine the extent to which the association between personality and parenting was mediated by genetic and environmental factors. Personality (Five Factor Model, NEO-FFI) and parenting data were collected as part of a larger German study of 300 adult twin pairs (GOSAT). The current paper analyzes data on a subset of the 300 twin pairs from the GOSAT sample who were concordant for having children (n 5 98 pairs or 196 Frank M. Spinath, Department of Psychology, University of Bielefeld, Bielefeld, Germany; Thomas G. O Connor, Department of Child & Adolescent Psychiatry and Social, Genetic, and Developmental Research Centre, London, United Kingdom. The research reported in this article was supported in part by a grant from the German Research Foundation (Deutsche Forschungsgemeinschaft). We are indebted to the twins for their participation. We d like to thank Alois Angleitner, Peter Borkenau and Rainer Riemann for their support and Robert Plomin for helpful comments on an earlier draft of this paper. The study is based on the German Observational Study of Adult Twins (GOSAT) directed by Angleitner, Borkenau, and Riemann. Correspondence concerning this article should be addressed to Frank M. Spinath, Department of Psychology, University of Bielefeld, P.O. Box 100131, 33501 Bielefeld, Germany. Fax: 149 521 106 6422. E-mail: fspin@uni-bielefeld.de Journal of Personality 71:5, October 2003. Blackwell Publishing 2003 786 Spinath & O Connor individuals). Results indicated modest overlap between personality and parenting. In addition, univariate behavioral genetic analyses indicated moderate genetic influence on select parenting dimensions. Results also indicated that the moderate phenotypic covariation between personality and parenting was attributed largely to nongenetic factors. Implications of the findings for research on parenting and personality are considered. A predominant view in clinical and developmental psychology research is that parental behavior is a relatively specific, learned pattern of behavior that reflects a history of relationship influences and is affected by current stresses and supports, as well as child characteristics (Belsky, 1984; Conger, Ge, Elder, Lorenz, & Simons, 1994; Cox, Owen, Lewis, & Henderson, 1989). An alternative view, derived from personality psychology, is that behavior in a parenting context may not be domain-specific, but instead, reflect general and stable personality characteristics (e.g., Kochanska, Clark, & Gold- man, 1997). Alongside environmental hypotheses of the determi- nants of parental behavior is the further question of the role of genetic influences. The current study, based on a community sample of German twins, was designed to test competing theories about the origins of individual differences of parenting. Specifically, we examine the covariation between parenting and personality and the degree to which covariation between personality and parenting is explained by overlapping genetic and environmental influences. Parenting and Personality Numerous investigations connect various forms of psychopathology with impaired parenting behavior (Downey & Coyne, 1990; Rutter & Quinton, 1984; Radke-Yarrow, 1991), but considerably less research has addressed the correlation between normal variation in personality and parenting. Available findings are somewhat difficult to interpret given the diverse ways in which both parenting and personality have been measured, but some similarities are apparent. For example, Kochanska et al. (1997) found that mothers high in the personality characteristics of negative emotionality and disagree- ableness were rated by observers as being more rejecting toward their children (see also Clark, Kochanska, & Ready, 2000). Similarly, Belsky, Crnic, and Woodworth (1995) reported that neuroticism was a robust predictor of a range of observed parenting behaviors, notably negativity/rejection. A third set of findings, based Personality and Parenting 787 on questionnaire measures, indicated that a parenting dimension termed positive support was moderately positively correlated with each of the five dimensions of personality measured; the overlap between negative affect and control dimensions of parenting with personality was less consistent (Losoya, Callor, Rowe, & Goldsmith, 1997). The phenotypic data reviewed above indicate that there is a link between parenting and personality, but that it may be modest in magnitude. A second lesson is that the degree of overlap depends on the personality and parenting dimensions considered. Specifically, the most robust connection appears to be between neuroticism or negative affect and negative/rejecting parenting. What has not received much attention are the origins of the overlap between parenting and personality. Genetic Influences on Parent-Child Relationships Quality of parent-child relationships is partly influenced by genetic factors. Evidence of this comes from twin and adoption studies of young children and adolescents across a range of methodologies (Lytton, 1977; Deater-Deckard & O Connor, 2000; O Connor, Hetherington, Reiss, & Plomin, 1995; Plomin, Reiss, Hetherington, & Howe, 1994). However, nearly all of the studies of genetic influences on parent-child relationships are child-based designs. That is, the genetically informative feature of the sample is the children being parented rather than the parents themselves. Child- based designs are useful for testing hypotheses concerning child effects models of parenting behavior, but a limitation is that genetic influences on parenting are measured only indirectly (i.e., via genetic influences on children s behavior that elicits parenting behaviors). The inverted design, in which the parents are the twins, is needed in order to identify direct genetic influences on parenting behavior. Far fewer investigations use a parent-based design to assess genetic influences on parent-child relationships. In one of the first studies of this kind, Plomin, McClearn, Pedersen, Nesselroade, and Bergeman (1988) reported that perceptions of the current family environment in a sample of elderly twins was partly heritable. A small number of studies focus more specifically on parenting, but report inconsistent results. For example, Kendler (1996) found moderate evidence of genetic effects on a dimension of parental 788 Spinath & O Connor warmth, although the influences of environmental factors, including the effects of a presumed common rearing environment in child- hood, were dominant for the other two dimensions assessed, protectiveness and authoritarianism. These findings were partly replicated, but also contradicted, by Perusse, Neale, Heath, and Eaves (1994) using the same measure of parenting. In the latter study, genetic influences were found for all parenting behaviors and there was no evidence of shared or common rearing environment. Somewhat different results were reported in a separate study by Losoya et al. (1997), who found moderate genetic influence on a measure of positive/supportive parenting; the relative impact of genetic and environmental variance on a negative control dimension of parenting was less clear. Although findings concerning the genetic influence on parents behavior differ across method and measure, there is some suggestion that genetic factors may be involved. Regardless of which specific parenting dimension may reflect genetic influence, it remains to be seen whether or not the nature of the connection is direct or instead mediated through some other genetically influenced behavior. The hypothesis tested in the current study is that genetic influence on parenting behavior is mediated via the genetic effects on personality. Framework of the Current Study We sought to integrate the lines of research discussed above in order to address key questions concerning the origins of parent-child relationships. Specifically, we reasoned that because there are robust genetic influences on personality (e.g., Plomin, DeFries, McClearn, & McGuffin, 2001) and, to a lesser extent, on at least some dimensions of parenting behavior, then the overlap between personality and parenting may be partly genetically mediated. The current study is one of the first to examine the genetic and environmental influences on the covariation between parenting and personality. The parent-based twin design used in this investigation not only provides a critical test of genetic influence, but also of environmental influence. Specifically, behavioral genetic designs distinguish be- tween two types of environmental variance, shared and non- shared . Shared or common environmental variance is an index of experiences that make siblings similar to one another, that is, Personality and Parenting 789 similarity for nongenetic reasons. In research on adult twins, shared environment is typically interpreted as the impact of influences carried over from the family of origin (Kendler, 1996). If this assumption is correct, then a significant shared environmental parameter for parenting would suggest that the transmission of parenting may be partly cultural or familial rather than, or in addition to, via genetic influences. Furthermore, shared environ- mental mediation of the correlation between personality and parenting would be consistent with the view that environmental influences, which may have been carried over from experiences in the family of origin, may mediate the connection between the two dimensions. Behavioral genetic research also examines nonshared environ- mental variance, which is attributable to experiences that make siblings different from one another. In parallel with the above hypotheses concerning shared environment, the current study is able to test the hypothesis that personality, parenting and their covariation are influenced by environmental factors that are unique to each sibling. The current study extends existing research in two important ways. First, we test hypotheses synthesized from two separate lines of research, namely, that the phenotypic connection between parenting and personality is explained by genetic as well as environmental factors. Second, whereas most of the pertinent investigations to date have been conducted in U.S. samples, the current study is based on a German community sample of twins. METHODS Sample The German Observational Study on Adult Twins (GOSAT) is a subset of twins from a larger community twin sample (Spinath, Angleitner, Borkenau, Riemann, & Wolf, 2002). Monozygotic (MZ) and same-sex dizygotic (DZ) twins selected to participate in GOSAT ranged in age from 18 70 years, with an average age of 34 years for the MZ pairs and 35 years for the DZ pairs. There was a predominance of female twin pairs in both the MZ (75%) and DZ samples (81%). The GOSAT sample was heterogeneous with respect to socioeconomic status. The current paper analyzes data on a subset of the 300 twin pairs from the GOSAT sample who were concordant for having children (n 5 98 pairs or 196 790 Spinath & O Connor individuals). The subsample with children differed from the sample only with respect to age (see preliminary analyses). The children of the parents ranged in age from early childhood through young adulthood. The number of children ranged from 1 5; approximately three-quarters of the parents had 1 2 children. Procedure GOSAT twins were invited to attend a day-long assessment in a university laboratory (Spinath et al., 1999). As part of day-long series of assessments, twins were asked to complete questionnaire measures of personality and parenting. An extensive battery of observational measures of personality and temperament were also collected, but are not part of this report. Measures Zygosity. Zygosity of MZ and DZ twins pairs was determined by genome mapping (Becker et al., 1997) and questionnaire measures (Oniszczenko, Angleitner, Strelau, & Angert, 1993), with a misclassifica- tion rate of less than 1%. Personality: NEO-FFI. Parental personality was measured using the NEO-FFI (Costa & McCrae, 1989, 1992), a widely used inventory of the five core personality dimensions known as the five-factor model of personality (Neuroticism, Extraversion, Openness to Experience, Agree- ableness, and Conscientiousness). Previous validity of the NEO-FFI on German samples has been reported (Angleitner, Ostendorf, & John, 1990). Previous analyses of the NEO-FFI on the community sample replicated prior research and suggested considerable reliability and validity for the five dimensions (Riemann, Angleitner, & Strelau, 1997). Parenting: FDEE-R. Parenting behavior was measured using the FDEE-R, a 31-item self-report measure of parenting with previously established reliability and validity (Lukesch & Tischler, 1975). The FDEE-R, which was developed for a German sample, is conceptually similar to widely used measures in U.S. samples. The measure has four subscales, Over-protective (14 items; internal consistency (alpha) 5 .76), Rejecting (6 items; internal consistency 5 .66), Supportive/Indulgent (6 items; internal consistency 5 .68), and Authoritarian (5 items; internal consistency 5 .68). Parents with high scores on the over-protective scale report that children ought to be watched closely and that parents need to make sure that they always know what their children do or think in order to save them from harm. High scores on the rejecting scale Personality and Parenting 791 reflect negative attitudes toward parenthood and children s behavior. Rejecting parents indicate that having children is occasionally regarded a burden they find hard to handle. High scorers on the mild-compliant scale propagate mild and lenient behavior towards children and state that parents should turn a blind eye more often. Finally, high scores on the authoritarian scale indicate the parent s acceptance of physical punishment (e.g., smacking a disobedient child). Items were rated on a 3-point Likert-type scale. The above factors were based on factor analysis using varimax rotation (Lukesch & Tischler, 1975). Parenting factor scores included in the analyses below are based on the items identified to load on each factor (unit weighted). Data analysis. The results are presented in three sections. First, we analyze the intraclass correlations and report univariate behavioral genetic analyses for the four parenting factors. Second, the phenotypic correlations between the personality and parenting dimensions are presented. Third, we report multivariate behavioral genetic analyses on the covariation between personality and parenting variables for those pairs of variables that are significantly correlated. Behavoral genetic analyses. The logic and analyses associated with partitioning variance of a phenotypic measure into genetic and environmental sources is described in detail elsewhere (Plomin et al., 2001). In brief, behavioral genetic models hypothesize three sources of individual differences. In the case of a twin study, additive genetic variance (Ga) is implied if the correlation among MZ twin pairs exceeds that found in DZ twin pairs. Shared environment (Es), or those environmental factors that function to make siblings similar to one another, is suggested if the correlation in DZ twins is greater than half that found in the MZ twin sample. Finally, nonshared environmental (En) influence indexes those environmental factors that function to make siblings different from one another. This parameter, which also includes measurement error, is most clearly seen as 1 the MZ correlation and is especially implicated if the similarity among twin pairs is small and nonsignificant, that is, the siblings differ substantially from one another. Estimates for the above parameters are based on model-fitting procedures. Following quantitative genetic theory, these models specify that MZ twins are genetically identical and are therefore correlated 1.0 for additive genetic variance; DZ twins share half of their genes and are therefore correlated .5. By definition, shared environment is correlated 1.0 because it represents those experiences that function to make siblings similar for environmental reasons; also by definition, nonshared environment correlation for all siblings pairs is 0 because it represents 792 Spinath & O Connor the influences that are unique to each sibling and make siblings different (see Plomin et al., 2001). In the bivariate model the sibling cross-covariances, or sibling A s score on measure 1 with sibling B s score on measure 2, are analyzed. The bivariate model includes genetic (Ga) as well as the shared (Es) and nonshared (En) environment estimates mediating the relationship between the two measured variables. The percentage of the phenotypic correlation that is due to each of the three sources can be computed by multiplying the common paths (i.e., the two paths connecting the measures) and dividing by the phenotypic correlation. The genetic (ga) and shared (es) and nonshared (en) environmental estimates unique to each measure are also examined. For example, a significant specific genetic estimate indicates that there are genetic factors influencing individual differences in one measure that are independent of the genetic influences on the second measure. Several methodological conditions for the genetic models should be noted. First, in these analyses, we examine only additive genetic variance; nonadditive (dominant and epistatic) genetic variance was not modeled because we have little power to discriminate additive and nonadditive genetic influences and because the twin similarities suggested that for the most part genetic influences were of the additive variety. Second, consistent with other studies, parental age and sex were partialed from the covariances prior to genetic modeling because these variables can artificially inflate similarity. Third, variance/covariance matrices (based on double-entered data) were analyzed because they are sensitive to, and allow tests of, variance differences among groups (Neale & Cardon, 1992). Fourth, both univariate and bivariate behavioral genetic models make certain assumptions about the nature of the processes being estimated. Specifically, the models assume no gene environment interac- tion or assortative mating but the appropriateness of the equal- environments assumption for the investigated characteristics. In the current study, we were unable to directly test the validity of these assumptions. Detailed discussions of the validity of these assumptions are reported elsewhere (Borkenau, Riemann, Angleitner, & Spinath, 2002; Loehlin, 1992; Rutter, Silberg, O Connor, & Simonoff, 1999). The significance of parameters in behavioral genetic models was examined by fixing the parameter to 0 and reestimating the fit, the conventional manner used in behavioral genetic research (Neale & Cardon, 1992). A significant change in the chi-square with 1 df indicates a significant worsening of the fit and that the parameter should be retained in the model. For purposes of presentation, however, all estimates are provided and a parameter is displayed as 0 only if the model estimated the parameter to be 0. There is wide recognition that the chi-square test Personality and Parenting 793 for the full model should be accompanied by alternative fit indices and that each alternative fit index has advantages and limitations. Because a primary concern in model-fitting is residual variance, the root mean square error of approximation was chosen to accompany the chi-square test and goodness of fit results ( Joreskog & Sorbom, 1993). Analyses were run using LISREL, 8th edition ( Joreskog & Sorbom, 1993). Model-fitting results for the bivariate models are given in the Figure captions. RESULTS Preliminary analyses. Prior to behavioral genetic model-fitting, a series of preliminary analyses were run on the total sample to examine the connection between zygosity, parent age and sex, and child age and sex with the parenting variables. Only one significant correlation (of the 20) was found: child age with Authoritarian parenting (r 5 .15, po.05). However, this association was not significant once parental age was partialed. Thus, partialing parental age and sex prior to fitting behavioral genetic models eliminated the significant correlation between child age and Authoritarian parenting. Although all twin pairs completed the parenting questionnaire, we were primarily interested in those pairs in which both twins were parents. From a total sample of 300 pairs, 98 pairs were concordant for having children, 157 were concordant for not having children and 40 were discordant (data were missing in 5 cases). Subsequent analyses are based on the twins who were concordant for children (n 5 196). Pairs who were concordant for having children differed from those who were concordant for not having children in one main respect: age. The average age in the former group was 47 years (range 29 69), whereas the average age in the latter group was 25 years (range 18 67). As suggested by the ranges, the difference was accounted for by the absence of relatively young parents in the sample. Importantly, there were no significant differences between the two groups in any of the 4 parenting or 5 personality measures once age was covaried. Univariate genetic analyses. Prior to examining the genetic and environmental sources of covariance between personality and parenting, it was first necessary to examine the univariate behavioral genetic results for each dimension. Behavioral genetic analyses of the personality data in GOSAT have been previously reported 794 Spinath & O Connor (Borkenau, Riemann, Angleitner, & Spinath, 2001). As is typically found for self-reported personality traits, additive genetic factors (h2) explained about a third of the phenotypic variance, whereas shared environmental factors (c2) across the dimensions of the five- factor model were only modest (average c2 across traits 5 .13). Average peer reports which were available for the GOSAT participants indicated genetic influences on personality of similar magnitude as self-reports (average h2 across traits 5 .39), yet indicated virtually no environmental effect of the shared variety. Interestingly, aggregated video-based trait ratings of actual behavior suggested somewhat stronger shared environmental influences than typically found in questionnaire studies. However, it is important to note that the data on the current study are a subset of twins from the larger sample, that is, those twin pairs with children. The correlations between twins for the personality and parenting dimensions are displayed in Table 1. For the parenting variables, consistent with the hypothesis of genetic influence, the correlation between MZ twins was generally greater than that for DZ twins. The magnitude of differences was substantial e.g., for two scales the MZ correlation was greater than twice the DZ correlation with the single exception of Rejecting. In addition, shared environment is implicated for all factors in which the DZ correlation is moderate and similar to that found in the MZ sample. Finally, given that the MZ correlation is substantially lower than unity, nonshared environment is implicated. The personality correlations for the twin pairs concordant for children are also displayed in Table 1. The correlations are generally similar to the larger sample from which they were drawn, although there is a tendency for the MZ twin correlation to be somewhat lower in this subsample. The same pattern was observed in the peer- reported data on personality for this subsample (see Riemann et al., 1997, for details on peer-reported personality). The model-fitting for the measures of parenting, displayed in Table 2, confirm the impressions from the correlational data. Specifically, significant and moderate genetic influence was found for Over-protectiveness, Supportive/Indulgent and Authoritarian parenting; no genetic effect was found for Rejecting. Conversely, sibling similarity could be attributed to environmental factors only for Rejecting. For each parenting dimension, nonshared environ- mental variance (plus measurement error) was substantial. Using Personality and Parenting 795 Table 1 Intra-class Correlations in MZ and DZ Twins: 5 Personality Factors and 4 Dimensions of Parenting Measure MZ DZ Personality Neuroticism .26 .17 Extraversion .35 .27 Openness .46 .42 Agreeableness .22 .12 Conscientiousness .41 .37 Parenting Behavior Over-protective .48 .27 Rejecting .26 .22 Supportive/Indulgent .33 .12 Authoritarian .43 .16 Note: For MZ twins, the N is 94 (i.e., 47 twin pairs); for DZ twins, the N is 80 (i.e., 40 pairs) for the personality measures and 86 (i.e., 43 pairs) for the parenting measures. The above correlations are based on intra-class correlations using the double-entry procedure (see text). 1 Cronbach s alpha to estimate the magnitude of the error component for each scale, nonshared environmental influences (e2) explained between 21 per cent (Authoritarian) and 42 per cent (Rejecting) of the reliable phenotypic variance. The model-fitting information provided in Table 2 indicates that each model provided a satisfactory fit to the data. Overlap between personality and parenting behavior. Correlations between the five personality factors and the four dimensions of parenting are displayed in Table 3. The correlations presented are those based on double-entered data among the twin pairs concordant for having children; the significance of the correlations was adjusted for the correct (i.e., non-double-entered) sample size. The patterns of correlations did not differ significantly according to zygosity, parent or child sex, or parent or child age. In addition, the pattern of correlations was very similar to, and not significantly different from, the pattern observed in the total sample (i.e., the sample that included twins with and without children). Overall, although there were modest to minimal correlations between 796 Spinath & O Connor Table 2 Model Fit Results, Estimates (Standard Errors), and Standardized Variance Estimates from Univariate Genetic Analyses of Parenting Variables LISREL Standardized Model Fitting Results Estimates Estimates w2/df GF RMSEA Ga Es En h2 c2 e2 Over-Protectiven 0/0 1 0 .73 .28 .81 .42 .06 .52 (.34) (.70) (.09) Rejecting 2.79/4 .95 0 .53 .94 .24 .76 (.12) (.07) Supportive/Indulgent .36/2 .99 0 .58 .84 .32 .67 (.12) (.08) Authoritarian 3.7/2 .97 .1 .63 .67 .48 .53 (.09) (.07) Note: GF 5 Goodness of Fit index; RMSEA 5 root mean square error of approximation; Ga 5 additive genetic estimate; Es 5 shared environment estimate; c2 5 shared environmental variance; e2 5 nonshared environmental variance. *Model run allowing variances to differ between MZ and DZ pairs. personality and parenting measures, the overlap that did exist is consistent with prior studies. Specifically, the effect size of the association between Neuroticism and Rejection (r 5 .22) was small, as was the association between Openness and Over-protectiveness (r 5 .23). The complete personality X parenting dimension matrix is reported in Table 3, although specific hypotheses were not offered for each correlation. In fact, of the parenting dimensions assessed, only Rejection was at least modestly correlated with more than one personality dimension. Given the correlation results, there is reason to examine genetic and environmental mediation more closely in the case of Neuroticism and Rejecting, and Openness and Over-protectiveness. Multivariate behavioral genetic analyses. Given the phenotypic correlations, multivariate behavioral genetic analyses were needed to explain genetic and environmental sources of (significant) covar- iance between personality and parenting in two cases: Neuroticism and Rejecting, and Openness and Over-protectiveness. The covar- iance matrices used in the model-fitting are reported in Table 4. Personality and Parenting 797 Table 3 Phenotypic Correlations Between Personality and Parenting Personality: NEO-FFI Agree- Conscien- Extra- Neuro- Open- ableness tiousness version ticism ness Parenting Over-Protective .02 .07 .09 .00 .23n Rejecting .11 .15 .17 .22n .04 Supportive/Indulgent .02 .04 .08 .09 .10 Authoritarian .10 .05 .01 .02 .05 n Note: The above correlations are based on double-entered data. po.05. Estimates from the bivariate model fitting results are displayed in Figures 1 and 2. For Neuroticism and Rejecting, the results indicate that virtually all of the covariance was mediated by nonshared environmental influences. Because the estimates in the figures are standardized, multiplying the paths from the same genetic or environmental factor and summing across Ga, Es, and En reproduces, within rounding error, the observed phenotypic correlation. Thus, (.17n.20)1(.39n.47) 5 .21 (the observed phenotypic correlation is .22). Furthermore, 86% (i.e., .18/.21) of the phenotypic correlation is explained by nonshared environmental influences. The absence of significant genetic mediation between the two dimensions is explained by the relative absence of genetic influence on either construct. Figure 1 also illustrates the genetic and environmental influences unique to personality and parenting. As expected from the correlations and univariate analyses, there was no genetic influence unique to Neuroticism or Rejecting parenting, that is, what little genetic influence there was on rejecting parenting was shared with Neuroticism, and vice versa. There were, however, significant shared and nonshared environmental influences unique to each construct. It is particularly interesting that both dimensions were strongly influenced by shared environment, but that the shared environ- mental influences were unique to each dimension. In other words, the environmental factors that made siblings similar in Neuroticism 798 Spinath & O Connor Table 4 Cross-Sibling Correlations Between Overlapping Dimensions of Personality and Parenting Rejecting and Neuroticism Rejec- Neuro- Rejec- Neuro- ting ticism ting ticism TwinA TwinA TwinB TwinB MZ Twins Rejecting TwinA 1.40 Neuroticism TwinA .27 .78 Rejecting TwinB .30 .12 1.40 Neuroticism TwinB .12 .15 .27 .78 DZ Twins Rejecting TwinA 1.05 Neuroticism TwinA .15 .93 Rejecting TwinB .25 .14 1.05 Neuroticism TwinB .14 .16 .15 .93 Over-protectiveness and Openness Over-pro- Open- Over- Openness tective ness protective TwinB TwinA TwinA TwinB MZ Twins Over-protective TwinA 1.29 Openness TwinA .24 .86 Over-protective TwinB .70 .25 1.29 Openness TwinB .25 .40 .24 .86 DZ Twins Over-protective TwinA .84 Openness TwinA .14 .65 Over-protective TwinB .19 .15 .84 Openness TwinB .15 .27 .14 .65 Note: Diagonal elements are variances Variances and correlations are based on double-entry procedure. The N s for the univariate analyses of parenting and bivariate analyses differ slightly because of missing data (o10%). were different from those that made siblings similar in their rejecting parenting. Finally, the model statistics indicate that the estimates provided a reasonable fit to the data. Personality and Parenting 799 Ga Es En .20 .39 0 0 .47 .17 Rejecting Neuroticism 0 0 .46 .78 .40 .76 ga es en ga es en Figure 1 Standardized estimates from the bivariate behavioral genetic analyses of parental Rejection and Neuroticism, w2(6) 5 5.04, ns; RMSEA 5 0, GFI 5 .95. For reasons of simplicity, the data were recoded so that phenotypic correlations across twin groups were positive. Although the paths to a common factor (i.e., Ga, Es, or En) are fixed to be equal in the model, the standardized estimates reported in the Figure differ slightly because of variance differences between the personality and parenting variables. Although the phenotypic correlation between Openness and Over-protectiveness is essentially identical to that found for Neuroticism and Rejecting parenting, the sources of this covariance are different. Figure 2 illustrates that the only significant source of covariance between the two constructs was shared environment (.43n.51 re-creates, within rounding error, the observed phenotypic correlation of .23). That is, experiences that made siblings similar in Openness also made siblings similar in their tendency to be Over- protective in their parenting. The experiences did not overlap entirely, however, as suggested by the significant unique shared environmental parameter for Openness. Interestingly, the absence of 800 Spinath & O Connor Ga Es En 0 0 .43 .51 0 0 Overprotective Openness 0 .55 0 .71 .44 .74 ga es en ga es en Figure 2 Standardized estimates from the bivariate behavioral genetic analyses of parental Over-protectiveness and Openness, w2(7) 5 4.12, ns; RMSEA 5 0, GFI 5 .96. For reasons of simplicity, the data were recoded so that phenotypic correlations across twin groups were positive. Although the paths to a common factor (i.e., Ga, Es, or En) are fixed to be equal in the model, the standardized estimates reported in the Figure differ slightly because of variance differences between the personality and parenting variables. significant nonshared environmental mediation indicates that experiences that made siblings different in Openness were not the same as those that made siblings different in Over-protectiveness. The genetic and environmental influences specific to Openness and Over-protectiveness (reflected in the Ga, Es, and En parameters) were also anticipated by the correlations and univariate results. Specifically, the significant genetic influence on Over-protectiveness was not shared with Openness, as the latter was little influenced by genetic influences. Nonshared environmental factors did influence each dimension to a substantial degree and, as noted above, were Personality and Parenting 801 unique to each construct. Finally, as suggested by the fit statistics, the estimates provided an acceptable fit the to the data. DISCUSSION The influence of personality and, in particular, the theory of personality included in the current measurement approach, is credited with explaining a wide range of behaviors in interpersonal situations. Indeed, the larger GOSAT study illustrates how many subtle interpersonal behaviors are influenced by personality variables (Borkenau et al., 2001; Borkenau, Riemann, Spinath, & Angleitner, 2000). In a parallel manner, considerable importance is placed on parenting behavior, and particularly the dimensions assessed in the current study. To date, few researchers have investigated the intersection of personality and parenting. We applied a twin study method to assess the degree to which personality and parenting behavior covaried, and the extent to which this overlap could be explained by shared genetic and environmental factors. The central findings were a) individual differences in parenting behaviors were influenced by both genetic and environmental factors, with the exception of Rejecting parenting, which showed no genetic influence; b) there were relatively isolated but predictable connections between personality and parenting behaviors; and c) the covariance between personality and parenting dimensions was mediated by environmental influences which were of the shared variety in the analysis of over-protective parenting and Openness and of the unshared variety in the analysis of rejecting parenting and Neuroticism. Genetic Influences on Parenting There is an increasing appreciation that genetic influence extends beyond traditional trait-like measures of behavior, such as personality, to more dynamic, interpersonal behaviors, such as family interactions (O Connor, Deater-Deckard, & Plomin, 1998). The current study extends research of this type to a community sample of adult German twins. Moderate genetic influence was observed for Over-protectiveness, Authoritarianism, and Suppor- tive/Indulgent parenting; no genetic effect was found for rejecting parenting. 802 Spinath & O Connor Linking the current findings with previous studies is not straightforward, as there is considerable variation in the kinds of parenting behaviors assessed and results reported. Indeed, even studies using identical measures of parenting have reported conflicting findings (cf. Kendler, 1996 and Perusse et al., 1994). Nonetheless, the findings that the three parenting dimensions noted above did show moderate genetic influence is generally consistent with existing data. It is not possible to determine the degree to which the discrepancies observed between the current findings and those from previous studies can be attributed to a range of plausible methodological reasons, including measures, design, and culture. Further research and a greater homogenization among methodol- ogies is clearly needed to resolve outstanding divergences in this relatively new line of investigation. However, it is noteworthy that each of the studies conducted to date suggests that genetic influences play some role in parenting behavior. This is an important finding with substantial implications for research on the mechanisms explaining the environmental roots of individual differences in children s behavior and psychopathology. Alongside the genetic findings are the equally important findings concerning environmental influence. A particularly striking finding in this regard is the absence of any genetic influence and the presence of shared environmental influence on parental Rejecting. To the extent that shared environmental influence can be interpreted as the impact of environmental influences carried over from childhood the period in which environmental influences were strongly shared by siblings these data provide evidence supporting a cultural/ familial transmission of harsh and rejecting parenting and the long- term consequences of early experiences on parenting behavior in adulthood. A unique feature of the current study is that, by accounting for genetic factors directly, we were able to demonstrate that individual differences in parenting are at least partly mediated by nongenetic influences. These results support recent findings suggesting how intergenerational links in parent-child relationship quality might be environmentally mediated (Belsky, Jaffee, Hsieh, & Silva, 2001; Waters, Merrick, Treboux, Crowell, & Albersheim, 2000) and that key features of the parent-child relationship, such as attachment, are largely influenced by shared environment rather than by genetic factors (O Connor & Croft, 2001). Personality and Parenting 803 Overlap Between Personality and Parenting The degree of separation between personality and parenting found in this study is noteworthy and generally consistent with previous reports. Where there was overlap, it was between negative, irritable, or hostile parenting behaviors and personality features. Parallel findings were noted by U.S. investigators using very different methods and measures (Belsky et al., 1995; Kochanska et al., 1997). One connection not found previously was that between over- protective parenting and the Openness personality factor. The inverse association between these two constructs is readily inter- pretable as a tendency for individuals who are themselves not open to experiences to restrict the behaviors or intrude on the activities of their children. Although the Openness personality factor has not received attention in the parenting literature, there are obvious conceptual similarities with family environmental measures used in developmental research, such as in the HOME inventory (Plomin et al., 2001). For example, high scorers on Openness to Experience typically pursue intellectual interests for their own sake, are open- minded about unconventional ideas, and enjoy philosophical arguments. They are willing to try different activities or go new places. In addition to the readiness to reexamine social, political, and religious values, they have an active fantasy life and a deep appreciation for art and beauty (Costa & McCrae, 1992). It has been demonstrated empirically that Openness to Experience can be judged by unacquainted observers on the basis of people s living or working environments with high accuracy (Gosling, Ko, Mannarelli, & Morris, 2002). Parenting behavior therefore stands out as an important behavioral variable that is only modestly associated with person- ality, at least in non-clinic samples and within the normal range of functioning. A modest overlap between personality and quality of parent-child relationships is also consistent with dynamic, bidirec- tional formulations of parent-child relationships that underscore the interactive effects of parental personality and child characteristics in eliciting parental behavior (Clark et al., 2000). A corollary of the distinctiveness of personality and parenting behavior is that there may be substantial differences in their etiology, although this is yet to be systematically examined. Our understanding of the develop- ment of personality is well advanced (Caspi, 1997); further research 804 Spinath & O Connor on the developmental origins of parenting is needed in order to test this hypothesis more directly. To our knowledge, this is the first investigation of the genetic and environmental mediation of personality and parenting. Although the findings must be considered preliminary and in need of replication, they raise important questions about the nature of parenting risks and the intergenerational transmission of personality and psychopathology from parent to child. Several specific findings deserve attention. First, the covariance between rejecting parenting and Neuroticism was mediated almost entirely via nonshared environmental experiences. That is, experiences associated with (subsequent) sibling differences in the twins rejecting parenting overlapped with those that gave rise to differences in Neuroticism. This finding may be especially important for understanding how parental personality problems and poor parenting jointly influence children s psychological development. Rejecting, hostile parenting, which is implicated as a feature of antisocial personality disorder (American Psychiatric Association, 1994; see also Capaldi & Patterson, 1991), has long been associated with a wide range of behavioral and emotional problems in children (Patterson, Reid, & Dishion, 1992). These results suggest that personality problems and poor parenting may co-occur for similar reasons and, furthermore, that their effects (inter- and intragener- ationally) may be environmentally mediated. A second example of the overlap between personality and parenting behavior also suggested a primarily environmental mediation, Open- ness and Over-protectiveness. In this case, experiences that made siblings similar to one another in personality also made them similar in parenting. The magnitude of the shared environmental mediation between the two constructs is relatively uncommon in multivariate genetic studies of adult samples and suggests that the particular shared experiences may pre-date parenthood. Experiences that may make siblings similar on (lack of) Openness to Experience and Over- protective in their stance toward their own children may derive from the family of origin or other shared settings, such as peer culture or other social settings (e.g., neighborhood). Limitations Interpretation of the findings must be viewed in the context of its limitations. Perhaps the most important limitation is the small Personality and Parenting 805 sample size. Although the sample of twins is slightly larger than used in parallel studies (e.g., Losoya et al., 1997), we had limited power to detect anything but relatively large effects. The limitation was offset, to some degree, by the novelty and importance of the research questions. Moreover, given that the subsample of twins who were concordant for having children did not differ from the larger sample from which they were drawn with the exception of being slightly older it would appear that the findings obtained may not be substantially biased. A further limitation, and one that is common to many behavioral genetic investigations, is that we are unable to identify which environmental influences are implicated in the models above. By covarying genetic effects, behavioral genetic designs provide the most powerful evidence of environmental factors, but additional research is needed to identify specific environmental factors and their mode of influence (i.e., by making siblings similar to or different from one another). Third, the same individuals provided data on personality and parent-child relationship quality. As a consequence, it is possible that shared method variance may explain some of the results, such as the overlap of nonshared environmental variance connecting rejecting parenting and Neuroticism. Finally, the meaning or source of shared environment cannot be identified from this study. Shared environmental variance in studies of adults is usually interpreted as a carrying forward of effects from childhood because siblings share environments to a substantially greater extent as children than they do as adults. Nevertheless, we do not have direct evidence supporting a family of origin interpretation to the shared environmental variance components. The current study highlights the need to test hypotheses spawned from crossing personality and developmental psychology and genetic and environmental explanations of individual differences in behavior. The findings reveal important gaps in research and theory in separate but complementary fields within psychology, and may foster further collaborative research approaches. REFERENCES American Psychiatric Association (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: American Psychiatric Associa- tion. 806 Spinath & O Connor Angleitner, A., Ostendorf, F., & John, O. (1990). Towards a taxonomy of personality descriptors in German: A psycho-lexical study. European Journal of Personality, 4, 89 118. Becker, A., Busjahn, A., Faulhaber, H-D., Bahring, S., Robertson, J., Schuster, H., & Luft, F. C. (1997). Twin zygosity: automated determination with microsatellites. Journal of Reproductive Medicine, 42, 260 266. Belsky, J. (1984). The determinants of parenting: A process model. Child Development, 55, 83 96. Belsky, J., Crnic, K., & Woodworth, S. (1995). Personality and parenting: Exploring the mediating role of transient mood and daily hassles. Journal of Personality, 63, 905 929. Belsky, J., Jaffee, S., Hsieh, K-H., & Silva, P. A. (2001). Child-rearing antecedents of intergenerational relations in young adulthood: A prospective study. Developmental Psychology, 37, 801 813. Borkenau, P., Riemann, R., Angleitner, A., & Spinath, F. M. (2001). Genetic and environmental influences on observed personality: Evidence from the German Observational Study of Adult Twins. Journal of Personality and Social Psychology, 80, 655 668. Borkenau, P., Riemann, R., Angleitner, A., & Spinath, F. M. (2002). Similarity of childhood experiences and personality resemblance in monozygotic and dizygotic twins: A test of the Equal Environments Assumption. Personality and Individual Differences, 33, 261 269. Borkenau, P., Riemann, R., Spinath, F. M., & Angleitner, A. (2000). Behaviour genetics of personality: the case of observational studies. In S. E. Hampson (Ed.), Advances in Personality Psychology (vol. 1, pp. 107 137). Philadelphia: Psychology Press. Capaldi, D. M., & Patterson, G. R. (1991). Relation of parental transitions to boys adjustment problems: 1. A linear hypothesis; 2. Mothers at risk for transitions and unskilled parenting. Developmental Psychology, 27, 489 504. Caspi, A. (1997). Personality development across the life course. In W. Damon & N. Eisenberg (Eds.), Handbook of child psychology: Social, emotional and personality development. (5th ed., pp. 311 388). New York: Wiley. Clark, L. A., Kochanska, G., & Ready, R. (2000). Mothers personality and its interaction with child temperament as predictors of parenting behavior. Journal of Personality and Social Psychology, 79, 274 285. Conger, R. D., Ge, X., Elder, G. H., Lorenz, F., & Simons, R. (1994). Economic stress, coercive family process and developmental problems of adolescents. Child Development, 65, 541 561. Costa, P. T., & McCrae, R. R. (1989). NEO PI/FFI manual supplement. Odessa, FL: Psychological Assessment Resources. Costa, P. T., & McCrae, R. R. (1992). Revised NEO Personality Inventory (NEO PI-R) and Neo Five Factor Inventory. Professional Manual. Odessa, FL: Psychological Assessment Resources. Cox, M., Owen, M. T., Lewis, J., & Henderson, V. K. (1989). Marriage, adult adjustment and early parenting. Child Development, 60, 1015 1024. Personality and Parenting 807 Deater-Deckard, K., & O Connor, T. G. (2000). Parent-child mutuality in early child- hood: Two behavioral genetic studies. Developmental Psychology, 36, 561 570. Downey, G., & Coyne, J. C. (1990). Children of depressed parents: An integrative review. Psychological Bulletin, 108, 50 76. Gosling, S. D., Ko, S. J., Mannarelli, T., & Morris, M. E. (2002). A room with a cue: Personality judgments based on offices and bedrooms. Journal of Personality and Social Psychology, 82, 379 398. Joreskog, K. G., & Sorbom, D. (1993). LISREL VIII: User s guide. Mooresville, IN: Scientific Software. Kendler, K. S. (1996). Parenting: A genetic-epidemiological perspective. American Journal of Psychiatry, 153, 11 20. Kochanska, G., Clark L, . A., & Goldman, M. S. (1997). Implications of mothers personality for their parenting and their young children s developmental outcomes. Journal of Personality, 65, 387 420. Loehlin, J. L. (1992). Genes and environment in personality development. Newbury Park, CA: Sage. Losoya, S. H., Callor, S., Rowe, D. C., & Goldsmith, H. H. (1997). Origins of familial similarity in parenting: a study of twins and adoptive siblings. Developmental Psychology, 33, 1012 1023. Lukesch, H., & Tischler, A. (1975). Uberprufung und Revision des Fragebogens zur Diagnostik elterlicher Erziehungseinstellungen von E. Littman und E. Kasielke (Examination and revision of the questionnaire for the diagnosis of parental attitudes on education by E. Littman and E. Kasielke). Probleme und Ergebnisse der Psychologie, 51, 19 54. Lytten, H. (1977). Do parents create, or respond to, differences in twins? Developmental Psychology, 13, 456 459. Neale, M. C., & Cardon, L. R. (1992). Methodology for genetic studies of twins and families. Boston, MA: Kluwer Academic Publishers. O Connor, T. G., & Croft, C. M. (2001). A twin study of attachment in pre- school children. Child Development, 72, 1501 1511. O Connor, T. G., Deater-Deckard, K., & Plomin, R. (1998). Foundations of behavioral genetics research: Implications for clinical psychology. C. E. Walker (volume editor), Comprehensive Clinical Psychology, Vol. 1 Founda- tions, pp. 87 114). Oxford: Elsevier Sciences. O Connor, T. G., Hetherington, E. M., Reiss, D., & Plomin, R. (1995). A twin- sibling study of observed parent-adolescent interactions. Child Development, 66, 812 829. Oniszczenko, W., Angleitner, A., Strelau, J., & Angert, T. (1993). The questionnaire of twins physical resemblance. Unpublished manuscript, Department of Psycho- logy, University of Bielefeld/University of Warsaw. Patterson, G. R., Reid, J. B., & Dishion, T. J. (1992). Antisocial boys. Eugene, OR: Castalia. Perusse, D., Neale, M. C., Heath, A. C., & Eaves, L. J. (1994). Human parental behavior: Evidence for genetic influence and potential implication for gene- culture transmission. Behavior Genetics, 24, 327 335. 808 Spinath & O Connor Plomin, R., DeFries, J. C., McClearn, G. E., & McGuffin, P. (2001). Behavioral genetics (4th ed.). NY: Worth. Plomin, R., McClearn, G. E., Pedersen, N., Nesselroade, J. R., & Bergeman, C. S. (1988). Genetic influence on childhood family environment perceived retro- spectively from the last half of the life span. Developmental Psychology, 24, 738 745. Plomin, R., Reiss, D., Hetherington, E. M., & Howe, G. (1994). Nature and nurture: Genetic contributions to measures of the family environment. Developmental Psychology, 30, 32 43. Riemann, R., Angleitner, A., & Strelau, J. (1997). Genetic and environmental influences on personality: A study of twins reared together using the self- and peer reported NEO-FFI scales. Journal of Personality, 65, 449 475. Radke-Yarrow, M. (1991). Attachment patterns in children of depressed mothers. In C. M. Parkes, J. Stevenson-Hinde, & P. Marris (Eds.), Attachment across the life cycle (pp. 115 126). London: Tavistock/Routledge. Rutter, M., & Quinton, D. (1984). Parental psychiatric disorder: Effects on children. Psychological Medicine, 14, 853 880. Rutter, M., Silberg, J., O Connor, T., & Simonoff, E. (1999). Genetics and child psychiatry: I. Advances in quantitative and molecular genetics. Journal of Child Psychology and Psychiatry, 40, 3 18. Spinath, F. M., Angleitner, A., Borkenau, P., Riemann, R., & Wolf, H. (2002). German Observational Study of Adult Twins: A multimodal investigation of personality, temperament, and cognitive ability. Twin Research, 5, 372 375. Spinath, F. M., Riemann, R., Hempel, S., Schlangen, B., Weiss, R., Borkenau, P., & Angleitner, A. (1999). A day in the life: Description of the German Observational Study on Adult Twins (GOSAT) assessing twin similarity in controlled laboratory settings. In I. Mervielde, I. Deary, F. de Fruyt, & F. Ostendorf (Eds.), Personality Psychology in Europe (vol. 1, pp. 311 328). Tilburg, NL: Tilburg University Press. Waters, E., Merrick, S., Treboux, D., Crowell, J., & Albersheim, L. (2000). Attachment security in infancy and early adulthood: A twenty-year long- itudinal study. Child Development, 71, 684 689.