Relationships Between Parental Negativity and Childhood ... - CiteSeerX

5 downloads 0 Views 304KB Size Report
Jan 8, 2007 - (see Burt et al. 2005 for a detailed description of the model) to assessments of parental negativity towards the child (i.e., anger, frustration ...
J Abnorm Child Psychol (2008) 36:633–645 DOI 10.1007/s10802-007-9151-2

Relationships Between Parental Negativity and Childhood Antisocial Behavior over Time: A Bidirectional Effects Model in a Longitudinal Genetically Informative Design Henrik Larsson & Essi Viding & Fruhling V. Rijsdijk & Robert Plomin

Received: 8 January 2007 / Accepted: 1 June 2007 / Published online: 30 June 2007 # Springer Science + Business Media, LLC 2007

Abstract This study examined the direction and etiology underlying the relationships between parental negativity and early childhood antisocial behavior using a bidirectional effects model in a longitudinal genetically informative design. We analyzed parent reports of parental negativity and early childhood antisocial behavior in 6,230 pairs of twins at 4 and 7 years of age. Results from a cross-lagged twin model contribute to the understanding of the mechanisms underlying the bidirectional processes involved in parental negativity and childhood antisocial behavior. Specifically, the findings of this study suggest that the association between parenting and child antisocial behavior is best explained by both parent-driven and childdriven effects. We found support for the notion that parent’s negative feelings towards their children environmentally mediate the risk for child antisocial behavior. We also found evidence of genetically mediated child effects; in which genetically influenced antisocial behavior evoke parental negativity towards the child. H. Larsson (*) Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, P.O. Box 281, 171 77 Stockholm, Sweden e-mail: [email protected] H. Larsson Department of Woman and Child Health, Karolinska Institutet, Stockholm, Sweden E. Viding : F. V. Rijsdijk : R. Plomin Social Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, London, UK E. Viding Department of Psychology, University College London, London, UK

Keywords Parenting . Antisocial behavior . Twins . Longitudinal . Genetics

Introduction A vast body of research has documented a robust relationship between different components of parenting (e.g., parental negative feelings, parenting discipline strategies and negative/conflictive parent-child relationships) and antisocial behavior in children and adolescents (Loeber and Stouthamer-Loeber 1986; Hill 2002; Rutter et al. 1998). This association is most likely explained by reciprocal processes; that is, by an explanatory model that emphasizes the importance of both parent-driven and childdriven effects (Patterson et al. 1992). Despite this, most studies have tended to examine parentdriven effects. For example, it has been demonstrated that parental harsh discipline, expressed negativity, and frustration, anger and criticism towards the child are associated with increased antisocial behavior in their off-spring (e.g., Beauchaine et al. 2005; Collins et al. 2000; Gardner et al. 1999; Kilgore et al. 2000; Vuchinich et al. 1992). Although fewer studies have examined child-driven processes, child effects have also been supported empirically (Anderson et al. 1986; Barkley et al. 1985; Danforth et al. 1991; Kerr and Stattin 2000; Lytton 1990). For example, Anderson et al. (1986) assessed how boys, with and without antisocial behavior, interacted with their own parents, the parents of other typically developing children, and the parents of other antisocial children. The results showed that antisocial children evoked most negative reactions from all groups of parents, indicating child-driven effects on parenting. The evocative effect was most pronounced in parents of antisocial children (i.e. the biological

634

parents reacted to their antisocial children with particularly bad parenting). This finding may be explained by either a previous history of transactional negative parent–child interactions or by underlying genetic factors that predispose both parent and child to antisocial behavior or, most likely, by some combination of the two. There have also been a few attempts to more explicitly test the “bidirectional effect model” of parenting and antisocial behavior (Bates et al. 1998; Burt et al. 2005; Danforth et al. 1991; Kerr and Stattin 2000; O’Connor 2002). For example, Bates et al. (1998) found that children’s difficult temperament was more strongly associated with later behavior problems when the mother had been observed to be relatively low in control actions than when she had been high in control actions. This pattern of results highlights the importance of both parent-driven and childdriven effects, which is consistent with predictions from the “bidirectional effect model”. Taken together, a growing body of research suggests that the association between parenting and child antisocial behavior is explained by both parent-driven and childdriven effects that most likely operate via reciprocal processes. Experimental intervention research appears to operate from the assumption of bidirectional processes taking place. Results from recent child antisocial behavior interventions suggest improvements in intervention effects by combining “parent training” with “child training” (Beauchaine et al. 2005; Webster-Stratton and Hammond 1997; Webster-Stratton et al. 2004). Nevertheless, the most successful interventions are still only effective for about two thirds of antisocial children (Webster-Stratton and Hammond 1997). This clearly highlights the importance of further clarification of the mechanisms underlying the bidirectional processes involved in parenting and childhood antisocial behavior. Influential reviews in recent years suggest that genetically informative longitudinal research designs, that make use of “natural experiments” to disentangle the normally inseparable contribution of genes and environments, may provide a much needed perspective on parenting and child antisocial behavior (Hinshaw 2002; Moffitt 2005; O’Connor 2002). One key feature of such studies is that they seek to examine the etiology of reciprocal effects by testing for both environmentally mediated parent effects and genetically mediated child effects (Burt et al. 2005). Although the literature is still limited, results from studies using genetically informative designs support the existence of both environmentally mediated parent effects on children’s antisocial behavior and genetically influenced child effects on parenting (Caspi et al. 2004; Ge et al. 1996; O’Connor et al. 1998). Ge et al. (1996) collected data on biological and adoptive parents of children adopted at birth. The adopted children were classified as being at genetic risk or

J Abnorm Child Psychol (2008) 36:633–645

not at genetic risk for antisocial behavior based on diagnosed drug, alcohol, or antisocial personality disorder in the biological parent. Antisocial behavior in the adopted children and the adoptive parents’ harsh/inconsistent parenting were assessed via multiple sources information (i.e., adoptive parents, the adoptee, and observers). O’Connor et al. (1998) used a sample of adopted children who were classified as being at genetic risk or not at genetic risk for antisocial behavior based on their biological mothers’ selfreport history of antisocial behavior. Antisocial behavior in the adopted children and the adoptive parents’ negativity towards the child were assessed via adoptive parents’ raings. The results from these two studies showed that adoptive children at high genetic risk for antisocial behavior received more negative parenting from their adoptive parents compared to adopted children at low genetic risk. It was also demonstrated that the association between the child’s genetic risk and negative parenting was partly mediated via the child’s genetically influenced antisocial behavior (Ge et al. 1996; O’Connor et al. 1998). In addition, a recent study by Burt et al. (2005) used a large sample of same-sexed twins who participated in the Minnesota Twin Family Study. A cross-lagged twin model was applied to parent-child conflict and children’s externalizing problems assessed by maternal reports as well as children’s reports. Both parent–child conflict and externalizing problems at age 11 were found to independently predict the other at age 14, providing support for a bidirectional effect model that allows for both parent-driven and child-driven processes. In addition, the results indicated that parent-driven effects were a function of both genetic and environmental effects (i.e., parent–child conflict at age 11 explained 3.7% of the heritability, 3.4% of the shared environment and 2.0% of the non-shared environmental variance in externalizing at age 14), whereas child-driven effects was largely a function of genetic factors (externalizing at age 11 explained 1.7% of the heritability, 0.9% of the shared environment and 0.9% of the non-shared environmental variance in parent-child conflict at age 14). Thus, this study showed, on the one hand, that parent–child conflict contributed to childhood externalizing problems via environmental mechanisms, but also that genetically influenced externalizing problems evoke parent-child conflict (Burt et al. 2005). Results from another twin study, that used multivariate genetic analysis of assessments of parental negativity and antisocial behavior in a sample of twin and adoptive sibling pairs (from 10 to 18 years old), provided further support for genetically influenced child effects on parental negativity (Neiderhiser et al. 1999). However, Neiderhiser and colleagues applied two different biometrical models that examine the genetic and environmental contribution to the parent-driven and child-driven effects, separately. Thus, this

J Abnorm Child Psychol (2008) 36:633–645

study was not able to simultaneously estimate the strength of the cross-lagged associations between parenting and adolescent antisocial behavior from middle to late adolescence and thereby test for the potential importance of reciprocal effects. To extend the literature on the etiology of the reciprocal effects in parenting and antisocial behavior, we analyzed data at 4 and 7 years from a large population-based twin study, the Twins Early Development Study (TEDS; Trouton et al. 2002). We applied a biometric cross-lagged model (see Burt et al. 2005 for a detailed description of the model) to assessments of parental negativity towards the child (i.e., anger, frustration, distance) and early childhood antisocial behavior. This model has previously been used to clarify the phenotypic-driven processes by which parenting and antisocial behavior impacts each other over time (Burt et al. 2005). Specifically, by constraining all associations across age to take the form of phenotypic partial regression coefficients, this model has the advantage of simultaneously estimating the strength of the longitudinal phenotypic associations (from age 4 to age 7) between any two measures when controlling for preexisting association between parental negativity and antisocial behavior at age 4. In addition, the use of twin data allows the model to decompose the phenotypic cross-age coefficients into their genetic and environmental components (Burt et al. 2005). Thus, by using this model we assume that the genetic and environmental factors of parental negativity and antisocial behavior at age 4 influences parental negativity and antisocial behavior at age 7 indirectly through phenotypic-driven processes. The present study differs from the previous cross-lagged study (Burt et al. 2005) in its focus on preschool-aged twins. Research on preschool children may be particularly important for three reasons: The origins of life-course persistent antisocial behavior can be traced to the preschool years (Loeber and Farrington 2000; Moffitt 2003), the strength of the parent-driven and child-driven effects may shift during development (Moffitt 2005), and the genetic and environmental contribution to parent- and child-driven effects may change over development (Lyons et al. 1995; Scarr and McCartney 1983). Two main research questions were asked: How do parental negativity and antisocial behavior at age 4 contribute to parental negativity and antisocial behavior at age 7? How do genetic and environmental factors in parental negativity and antisocial behavior at age 4 contribute to parental negativity and antisocial behavior at age 7? In addition, we also asked: Are there sex differences in parental negativity and antisocial behavior? How much of the variance in parental negativity and antisocial behavior is due to genetic and environmental factors? How do genetic and environmental factors influence the age-specific overlap between parental negativity and antisocial behavior? Given the findings of Burt et al.

635

(2005), we hypothesized that we would find evidence of bidirectional effects between parental negativity and antisocial behavior. We predicted that parental negativity would contribute to childhood antisocial behavior via environmental mechanisms. We also hypothesized that we would find evidence of a genetic component contributing to stable, early-appearing antisocial behavior. This reasoning was based on prior research indicating that genetic influences on antisocial behavior are more important at younger ages (Arseneault et al. 2003; Eley et al. 2003; Rhee and Waldman 2002; Taylor et al. 2000). Based on the same reasoning, we predicted that childhood antisocial behavior would influence parental negativity, mainly via genetic mechanisms.

Method Participants Participants are members of the Twins Early Development Study (TEDS), a large population-based longitudinal study based on all twins born in England and Wales 1994–1996 (Trouton et al. 2002). Background information regarding pregnancy, birth, and family demographics was obtained when the twins were 18 months old. The parents completed behavioral rating scales for both twins at ages 2, 3, 4 and 7. Zygosity was determined using a standard zygosity questionnaire, which has been shown to have 95% accuracy (Price et al. 2000); 75% of same-sex pairs have subsequently been confirmed with DNA markers (Freeman et al. 2003). The sampling frame for the present study was 6,763 twin pairs from the 1994 and 1995 birth cohorts for whom data are available at the 4- or 7-year assessments. We excluded twin pairs from current analyses if the twin had parental reports of medical or neurological conditions or if zygosity could not be assigned with confidence (533 twin pairs). Thus, the total number of twin pairs included in the model fitting was 6230 twin pairs: 1,033 monozygotic male twin pairs (MZM); 1,020 dizygotic male twin pairs (DZM); 1,141 monozygotic female twin pairs (MZF); 1,020 dizygotic female twin pairs (DZF); 2,016 dizygotic opposite sex twins (DZO). Despite attrition, the TEDS sample that provided data at age 7 is closely matched to the UK population in terms of ethnicity and maternal education (Harlaar et al. 2005). Measures Antisocial Behavior Information about symptoms of antisocial behavior was obtained when the twins were 4 and 7 years old using parent reports of the Strengths and Difficulties Questionnaire (SDQ; Goodman 1997). The SDQ is a widely used (see http://www.sdqinfo.com) 25-

636

item questionnaire that is designed to assess behavioral competencies as well as behavior problems in children ages 4 to 16 years. The present analyses used the SDQ five-item conduct problem scale (e.g., ‘Often fights with other children or bullies them’, ‘Often has temper tantrums or hot tempers’), which has been shown to be reliable and valid for parents as well as teachers (Goodman 1997). The raters were asked to indicate on a three-point response scale (ranging from not true to certainly true) how well each item described the child’s behavior over the past six months. The alphas for the parent-reported SDQ conduct problem scales were 0.56 and 0.62 at ages 4 and 7 years respectively. Although not particularly high, these alpha coefficients are in agreement with those found in other large-scale studies using SDQ with children and adolescents (Goodman 2001). Parental Negativity Parental negativity was assessed at age 4 using the parent negativity scale of the Parent Feelings Questionnaire (Deater-Deckard 2000; Knafo and Plomin 2006). This scale has four items that were rated on a fivepoint scale (where 1=definitely untrue and 5=definitely true) for the first-born twin, including the following statements: “Sometimes I feel very impatient with him/ her”; “Sometimes I wish s/he would go away”; “Sometimes s/he makes me angry”; “Sometimes I am frustrated by him/ her”. In an attempt to avoid parents answering similarly about their twins, after answering about the first-born twin, parents were then asked, “Do you feel this way more or less with your second-born twin?” These questions were rated on a five-point scale from “a lot more” to “a lot less.” The parental negativity scale shows good internal consistency in the TEDS sample (alpha=0.80). For first-born twins, the sum of the respective items was standardized for the whole population to zero mean and unit variance. For second-born twins, this standardized sum was added to the standardized sum of the differential items indicating more or less parental negative feelings; this composite score was then standardized (Knafo and Plomin 2006). At age 7, slightly modified items were used to assess parental negativity towards the firstborn twin (“Do you ever feel impatient with the elder twin?” “Does the elder twin make you feel frustrated?” “Do you ever wish the elder twin would leave you alone?” Does the elder twin ever make you angry?”). The four items were rated on a four-point scale (where 1= never and 4=often) for the first-born twin. After answering about the first-born twin, parents were then asked, “Do you feel this way more or less often with the younger twin”? These questions were rated on a three-point scale ranging from “more” to less”. Treatment of the second-born twin scores at age 7 was similar to that at age 4 and in line with previously published analyses of this measure (Knafo and Plomin 2006). The parental negativity scale shows good internal consistency in the TEDS sample (age 7: alpha=.80).

J Abnorm Child Psychol (2008) 36:633–645

Statistical Analysis The structural equation modeling program Mx (Neale et al. 2003) was used to calculate correlations (phenotypic correlations and intra-class correlations) and to test for mean differences across sex and age by likelihood-ratio χ2 tests (described below). These tests control for the dependence that exists between the scores within a twin pair. We also used MX to perform the genetic analyses. Because mean effects of age and sex can spuriously inflate twin resemblance, all models were fitted to age- and sexadjusted residual scores from multivariate linear regression modeling (McGue and Bouchard 1984). We used the method of raw maximum likelihood estimation to handle missing data and retain twins who have missing data at one or more assessments. Goodness of fit of the models was assessed by a likelihood-ratio χ2 test, which is the difference between −2 log likelihood (−2 ll) of the full model and that of the restricted model. This difference is distributed as a χ2 with degrees of freedom (df) equal to the difference between the number of estimated parameters in the full model and that in a restricted model. Akaike’s information criterion (AIC=χ2 −2×df) was also computed, in which a lower AIC value indicates better fit of the model to the observed data. Twin studies make use of the difference in the proportion of genes shared between MZ twins, who share 100% of their genes, and DZ twins, who share on average 50% of their segregating genes. In the basic twin model, total phenotypic variance of a measured trait can be divided into additive genetic factors (A), shared environmental factors (C), and non-shared environmental factors (E). Shared environmental factors refer to non-genetic influences that contribute to similarity within pairs of twins, whereas nonshared environmental factors are those experiences that make siblings dissimilar and include measurement error (Plomin et al. 2001). In the present study we examined the associations between parental negativity and antisocial behavior at ages 4 and 7 by using a cross-lagged twin model (see Fig. 1). More specifically, this model constrains all associations across age (i.e., b11, b22, b21, b12) to take the form of phenotypic partial regression coefficients. The cross-age stability paths (i.e., b11, b22) estimate the 3-year stability for parental negativity and antisocial behavior, when controlling for the preexisting association between the two phenotypes. The cross-lagged paths (i.e., b21, b12) estimate the independent contribution of parental negativity at age 4 on antisocial behavior at age 7 (i.e., b21) and similarly the independent contribution of antisocial behavior at age 4 on parental negativity at age 7 (i.e., b12), when controlling for the stability in the two phenotypes. The age-specific variance and covariance for parental negativity and antiso-

J Abnorm Child Psychol (2008) 36:633–645 E C A re

rc

637 e3

e1 c1

Parental Negativity Age 4

b 11 b 21

Parental Negativity Age 7

c3

Es Cs

a1

a3

As

a2

a4

As

ra

ra A C E

c2

Antisocial Behavior Age 4

b 12 b 22

Behavior vior Antisocial Beha Age 7

c4 e4

e2

rc

re

Cs Es

Fig. 1 A path diagram of the cross-lagged model. The latent variables A (additive genetic factor), C (shared environmental factor), and E (non-shared environmental factor) are depicted in circles. As (additive genetic) is residual variance specific to age 7, likewise for Cs (shared environment), and Es (non-shared environment). Measured variables

are depicted in rectangles. Standardized path estimates for these factors (i.e., a1, c1, e1/a2, c2, e2/a3, c3, e3/a4, c4, e4), genetic and environmental correlations (i.e., ra, rc, re), cross-age stability paths (i.e., b11, b22) and cross-lagged paths (i.e., b21, b12) are also represented in the diagram

cial behavior at age 4 and for parental negativity and antisocial behavior at age 7 was Cholesky decomposed and standardized to genetic, shared environmental and nonshared environmental path estimates (i.e., a1, c1, e1/a2, c2, e2/a3, c3, e3/a4, c4, e4) and genetic, shared environmental and non-shared environmental correlations (i.e., ra, rc, and re). One main set of calculations, using the parameter estimates from the best fitting cross-lagged model, were carried out to examine the most central research question of our study: How do genetic and environmental factors in parental negativity and antisocial behavior at age 4 contribute to parental negativity and antisocial behavior at age 7? Specifically, we examined the actual contribution of genetic and environmental variation in parental negativity and antisocial behavior at age 4 on the genetic and environmental variation in parental negativity and antisocial behavior at age 7. In other words, we decomposed the total genetic, shared environmental and non-shared environmental variance in parental negativity and antisocial behavior at age 7 into effects due to: (1) cross-age effects, (2) crosslagged effects, (3) common effects from age 4 and (4) specific effects from age 7. For example, the total genetic variance in antisocial behavior at age 7 was decomposed in the following way: (1) cross-age effects: genetic influences unique to antisocial behavior at age 4 (calculation; b222  a22 ), (2) cross-lagged effects: genetic influences unique to parental negativity at age 4 (calculation; b221  a12 ), (3) common effects from age 4: genetic effects common to parental negativity and antisocial behavior at age 4 (calculation; 2×[b22 ×a2×ra ×a1×b21]), and (4) specific genetic effects from age 7 (calculation; a42). The total shared environmental and non-shared environmental variation in antisocial behavior at age 7 was also decomposed into the same four components using the same logic. We fitted a series of sex-limitation models to test for qualitative sex-differences, quantitative sex-differences and phenotypic variance differences between the sexes. Signif-

icant qualitative sex-differences, indicated by genetic correlations between the opposite-sex twins of less than 0.5, suggest that different genes are influencing phenotypic variation in the sexes. Opposite-sex twins may also have fewer shared environmental experiences than same-sex twins, indicated by shared environmental correlations between opposite-sex twins of less than 1. Quantitative sex differences refer to differences between boys and girls in the magnitude of additive genetic, shared environmental and non-shared environmental influences on the measured phenotypes. The significance of such quantitative sex differences can be tested by allowing the magnitude of the parameter estimates to differ between boys and girls. Potential phenotypic variance differences between the sexes can be tested using a scalar model. Such a model allow the phenotypic variances to differ between boys and girls, whereas the genetic and environmental parameter estimates are equated across sexes and the genetic correlations for opposite-sex twins are constrained to be equal to the genetic correlation for same-sex twins.

Results Descriptive Statistics Means, standard deviations (SD) and number of respondents for age-regressed scores of parental negativity and antisocial behavior at age 4 and 7 are presented in Table 1. At both ages, there were significant differences in mean scores between boys and girls for parental negativity (age 4: Δχ2 =60.40, df=1, p