Introduction
Previous studies in twins and adoptees showed a substantial heritability in internalizing and externalizing problems over childhood [
6,
9,
16,
20,
21]. Bartels et al. [
6] showed that the heritability of internalizing problems decreased during childhood from 59% at age 3 to 37% at age 12, and the contribution of shared environmental influences increased over time from 13% at age 3 to 37% at age 12. For externalizing problems, small increases in heritability were observed for boys from age 3 (57%) to age 12 (64%), whereas for girls the influences of genetic (50%) and environmental factors remained stable over this period. Significant genetic and shared environmental influences have also been reported for anxious-depressive and withdrawn behavior at age 12 [
24].
In addition to differences in heritability as a function of age and sex, genetic effects may also be conditional on environmental exposures, a process often referred to as gene–environment interaction (G × E) [
15]. For the current study we were interested in the moderating role of parental divorce on the genetic architecture of internalizing and externalizing problems in children. The experience of parental divorce is often associated with higher mean levels of internalizing and externalizing problems in children [
3,
17,
34]. So far, little is known about a possible interaction effect of parental divorce on sources of variation in internalizing and externalizing problems.
In disadvantageous environments, heritabilities may differ from those in advantageous environments. So far, there is empirical support for both higher and lower heritabilities of child problem behavior in the context of disadvantageous environments. Lower heritabilities in disadvantageous environments can result from the suppression of genetic effects by environmental exposures that push or predispose a child to show behavioral problems [
10,
32,
35]. Several studies reported that heritabilities are lower in disadvantageous environments. For instance, Hicks et al. [
18] showed that the heritability of internalizing problems in 17-year-olds was lower in the context of greater environmental adversity, which included parental divorce. Also, it has been reported that the heritability of adolescent depressive symptoms was lower at higher levels of parental negativity (e.g., frequency of conflicts, punitiveness, and parent–child disagreement) [
13]. Another study showed that the heritability of conduct problems in children aged 5–17 years dropped from 77 to 0% at high levels of family dysfunction [
11]. The heritability of boys’ antisocial behavior was lower in adolescents from families with lower socioeconomic status [
37]. A lower heritability does, however, not necessarily imply less genetic variance. Heritability is a ratio (genetic variance divided by total phenotypic variance), so enlarged environmental variance may result in a lower heritability, even when the absolute amount of genetic variance remains the same. For instance, Feinberg et al. [
13] showed that the amount of non-shared environmental variance was higher at high levels of parental negativity, whereas the absolute amount of genetic variance remained stable.
Alternatively, when genetic effects are triggered or amplified by disadvantageous environments, heritabilities will increase [
27,
35]. South and Krueger [
36] reported a higher heritability of internalizing problems when marital quality was lower, suggesting that persons with a genetic predisposition to internalizing problems may be more likely to express this predisposition in the context of a problematic marriage. Similarly, the heritability of child and adolescent depressive symptoms was higher when the levels of family conflict were high [
33]. The heritabilities of adolescent depressive symptoms and externalizing problems were higher when adolescents had experienced more stressful life events [
19,
25], and Feinberg et al. [
13] reported that the heritability of antisocial behavior was greater at low levels of parental warmth. Regarding these previous studies that report higher as well as lower heritabilities in the context of disadvantageous environments, it is obvious that G × E findings with regard to child and adolescent psychopathology are mixed.
We previously reported in this journal that parental divorce was preceded and predicted by higher levels of externalizing problems in girls at age 3 [
34]. Furthermore, parental reports indicated that children at age 12 years with divorced parents showed more internalizing and externalizing problems than children with married parents. The aim of the current study was to investigate as to whether genetic and environmental influences on internalizing and externalizing problems are modified by parental divorce. Therefore, we compared the genetic architecture of internalizing and externalizing problems at ages 3 and 12 between children from divorced and non-divorced families. We expected to find the genetic architecture of pre- and post-divorce internalizing and externalizing problems to be influenced by parental divorce. However, since earlier G × E findings are mixed we could not formulate a specific hypothesis about the direction of effects.
A unique aspect of this study is that all families were still intact at age 3, and a subgroup of children experienced parental divorce between ages 3 and 12 years. This way, we were able to investigate variation in pre-divorce and post-divorce internalizing and externalizing problems in children. Another unique aspect is the large sample size of 4,592 twin pairs. As the genetic architecture of child problem behaviors may be different for boys and girls [
6], and because the moderating effect of parental divorce may depend on sex [
38], we took divorce, sex, age, and the interaction of sex by divorce into account as moderators.
Discussion
The present study contributes to the field of G × E interaction research by investigating whether the genetic architecture of internalizing and externalizing problems differed between children living in intact versus divorced families. Children’s behavioral problems were assessed pre-divorce (at age 3 years) as well as post-divorce (at age 12 years). Although age-specific genetic and environmental influences [
6] may have contributed to a somewhat different pattern of interaction effects at ages 3 and 12 years, we showed that at both ages heritabilities were slightly lower for children from divorced families than for children from intact families. These results support the hypothesis that genetic influences are less important in the context of parental divorce [
10,
32,
35], which is consistent with findings from several other studies [
11,
13,
18,
37]. However, as the absolute amounts of genetic variance were similar between children from divorced and intact families (except for boys’ internalizing problems at age 12), genetic effects were not suppressed by the experience of parental divorce. Heritabilities were lower for children from divorced families because for these children there was more environmental variance. This finding emphasizes the importance of considering both the relative and the absolute contributions of genes and environment (i.e., the standardized and unstandardized estimates of genetic and environmental variances) when studying G × E. A sole focus on standardized estimates (i.e., heritabilities) may provide a limited picture of what is going on, and can even lead to misinterpretation [
13].
One of the unique aspects of this study is the assessment of pre-divorce as well as post-divorce problems. Although we did not find evidence for higher mean levels of pre-divorce internalizing problems at age 3 in this study as well as in our earlier report [
35], our present results show a larger non-shared environmental variance for pre-divorce internalizing problems. Thus, pre-divorce environmental factors increase heterogeneity in children’s internalizing problems without significantly increasing mean levels of internalizing problems.
Our findings are in agreement with previous studies that showed increasing environmental variances of internalizing problems with age [
6,
9], which may likely be explained by mounting environmental influences as a result of children’s transition to school, during which they develop social relations with other children, and must cope with several new demands. Our study shows that shared environmental influences become even more important for children of divorced families, which may be an expression of the impact of divorce on family dynamics, such as family relationships. It may also partly reflect children’s understanding of the family situation or twins’ shared access to possible support in relationships outside the family. More research is needed to define which specific environmental factors are of interest in explaining the effects of parental divorce on variation in children’s problem behaviors.
The significant moderation of the shared and non-shared environment by divorce status indicates a greater heterogeneity in internalizing and externalizing problems among children with divorced parents. This underlines the importance of focusing on etiologies of individual differences in the effects of parental divorce instead of focusing on general mean effects.
The ways in which children cope with stressors associated with parental divorce, such as spending less time with the non-custodial parent, parents beginning new romantic relationships, and the intensity, frequency, and style of parental conflict [
11,
21], may be important mediators and moderators of the relation between parental divorce and behavioral problems [
12]. One could hypothesize that exposure to parental conflict could enhance shared environmental effects, while coping strategies would be more likely to enhance genetic effects.
There seems to be a stronger G × E effect for internalizing problems than for externalizing problems, since divorce moderates genetic effects on internalizing problems, but not those on externalizing problems. One possible explanation for this difference is that additive genetic factors are the main source of stability in externalizing problems over time, whereas most of the stability in internalizing problems can be explained by shared environmental factors [
6]. The high genetic contribution to stability in externalizing problems results from the fact that a subset of genes is active at multiple ages throughout childhood. Also, heritability estimates of externalizing problems are generally more similar at different ages during childhood than heritability estimates of internalizing problems [
6].
G × E approaches to human behavior are still at the initial stages of inquiry [
27]. Research to the role of the environment in genetic expression may be complicated if the population is heterogeneous in the type of G × E interactions [
35]. Some interactions may apply only to subgroups of the population, that differ for instance in their social experiences or that have specific genetic variants. For researchers interested in identifying specific genes associated with externalizing or internalizing problems, our results suggest that greater efficiency in detecting associations with specific genes could be obtained by limiting samples to children who have not experienced parental divorce, because then the relative influence of genetics is the largest.
Besides some powerful aspects of this study, such as the large number of participating families, there are some important limitations. Firstly, the observed rate of parental divorce in our twin sample (i.e., 10%) is not representative for the general population. In a large prospective study of Dutch adolescents, about 20% had experienced parental divorce before the age of 12 years [
30]. Non-response could be related to divorce status. Non-responders are not necessarily unwilling to return the questionnaire, but as parental divorce often involves change of address, surveys could have been sent to wrong addresses when the families moved without informing the NTR. Secondly, divorced parents may over-report children’s problems or they may be more sensitive to them. Over-reporting could be due to parental characteristics such as anxiety or depression [
14], to parents’ concern about their children’s mental health, or to their own distress associated with the divorce [
26]. Part of the increased environmental variance found in children from divorced families could therefore be due to over-reporting. However, earlier studies to the effects of rater bias and unreliability showed that these effects distort the estimates of the shared and non-shared environmental factors to a small degree only. For both the internalizing and the externalizing scale, rater bias accounted for at most 13% of the variance, and measurement error and unreliability accounted for less than 11% of the variance [
4,
5]. Thirdly, as younger children may be differently affected by parental divorce than older children and adolescents [
17], our results cannot be generalized to children of other ages. The heritability estimates obtained in our moderation analyses control for common variance between the moderators and child behavioral problems. Therefore, our estimates may not be directly comparable to studies that only estimate heritability.
In conclusion, environmental influences become more important in explaining variation in children’s problem behaviors in the context of parental divorce. Replication of these G × E effects is necessary. It would be interesting to extend our study with adult data of these children, in order to investigate if the experience of parental divorce in childhood also affects variability in problem behaviors in adulthood.
Acknowledgments
Data collection was funded by grants “Genetics of externalizing disorders in children” (NWO 904-57-94), “Spinozapremie” (NWO/SPI 56-464-14192), “Twin-family database for behavior genetics and genomics studies” (NWO 480-04-004), “Twin research focusing on behavior and depression” (NWO 400-05-717), “Developmental Study of Attention Problems in Young Twins” (NIMH, RO1 MH58799-03), and “Bridge Award” (NIMH R56). Meike Bartels is financially supported by a senior fellowship of the EMGO+ Institute for Health and Care. Sylvana Robbers is supported by the Sophia Children’s Hospital Foundation, project number 483.