Parental involvement as an etiological moderator of middle childhood oppositional defiant disorder

Abstract

The goal of this study was to investigate parental involvement as an etiologic moderator of oppositional defiant disorder (ODD) during middle childhood. Previous studies examining the influence of genetic and environmental factors on ODD have not considered whether and how these factors might vary by parental involvement. We thus conducted a series of “latent G by measured E” interaction analyses, in which measured parental involvement was allowed to moderate genetic, shared, and non-shared environmental influences on child ODD. Participants include 1027 twin pairs (age ranged from 6 to 11 years old) from the Michigan State University Twin Registry (MSUTR). Results did indeed suggest that the etiology of ODD varies with maternal involvement, such that genetic influence on ODD became more prominent as maternal involvement decreased. However, these results were specific to children’s perceptions of maternal involvement and did not extend to maternal perceptions of her involvement. There was no evidence that paternal involvement moderated the etiology of ODD, regardless of informant. The different results found in twins’ and parents’ data is consistent with previous research that children may have different perceptions from parents about their family relationships and this discrepancy needs to be taken into account in future research.

Oppositional Defiant Disorder (ODD) is a leading cause of referral to children’s mental health services (Loeber, Burke, Lahey, Winters, & Zera, 2000). As defined by the DSM-5, ODD consists of a recurrent pattern of negative, defiant, hostile, and disobedient behavior toward authority figures. ODD has often been considered a mild condition (Rey et al., 1988), likely because many of its symptoms approximate typical child behavior (e.g., easily losing temper and arguing with parents). Partly as a result of this perception, and the related idea that ODD is simply a milder form of conduct disorder (Burke, 2009; Loeber, Burke, & Pardini, 2009), oppositional behaviors are often only researched in the context of other behavioral problems, if they are even considered at all. However, ODD is distinct from both normal childhood behavior (Keenan & Shaw, 2003) and conduct disorder (Rowe et al., 2010), occurring in about 2% of girls and 5% of boys (Maughan, Rowe, Messer, Goodman, & Meltzer, 2004). As such, understanding the specific etiology (i.e., genetic and environmental origins) of ODD is vital, both for the basic understanding of human development, and the creation of interventions and policies that aim to reduce rates of child psychopathology.

Previous research on ODD has consistently pointed to important contributions of both genetic and environmental influences on individual differences in ODD symptoms. However, there are often discrepancies across studies regarding the relative magnitude of genetic and environmental effects (Burt, 2009; Hudziak, Derks, Althoff, Copeland, & Boomsma, 2005). These fluctuations are likely due in part to methodological differences between studies (e.g., differently aged participants). Yet another possibility is that the use of simple quantitative genetic models, which only compute average levels of genetic and environmental influences across the sample, may mask meaningful etiological differences as a function of environmental contexts, or gene-environment interaction (GxE).

Gene-envirioment interaction refers to a genetically modulated responsiveness to environmental factors, such that the importance of genotype varies across environmental contexts (Belsky & Pluess, 2009; Burt, 2011). Within this framework, the exposure to a given environment moderates the importance of genetic and/or environmental contributions in explaining variance for a given outcome (Plomin, DeFries, & Loehlin, 1977; Rutter, Moffitt, & Caspi, 2006). Multiple types of gene-environment interaction models have been proposed. For example, the diathesis-stress model suggests that genetic influences on psychopathology will be greatest within the context of riskier environments; in a neutral or enriching environment, however, both genetically vulnerable and resilient individuals will show equivalently low levels of psychopathology (Rutter, et al., 2006). In contrast, the bioecological model proposes that environmental influences are strongest in environments characterized by risk, whereas genetic influences are most prominent in the absence of environmental risk (Bronfenbrenner & Ceci, 1994; Burt, 2011; Burt & Klump, 2014; Nikolas, Klump, & Burt, 2015).

Detecting GxE effects can be challenging, however twin study designs represent a powerful tool in this regard. In twin designs, the differential genetic similarity of monozygotic and dizygotic twins is leveraged to model the extent to which latent genetic and environmental influences are moderated by some external variable. The direction of any observed moderation in these statistical models distinguishes between the different conceptual GXE models (e.g., if genetic influences are more pronounced at higher levels of a risk factor, there is evidence for a diathesis-stress interaction). Parenting is perhaps the most frequently examined moderator in GxE studies, with several studies published in the last decade (e.g., Burt, 2011; Burt, Klahr, Neale, & Klump,, 2013; Button, Lau, Maughan, & Eley, 2008). However, as there are no studies (to our knowledge) that directly examine GxE effects in ODD per se, we must look to studies of antisocial behavior for clues about etiologic moderation in ODD. The most commonly supported model of GxE in adolescence is the diathesis-stress model (Beach et al., 2010; Feinberg, Button, Neiderhiser, Reiss, & Hetherington, 2007; Hicks, South, Dirago, Lacono, & McGue, 2009). Specifically, there is evidence that increased parental negativity is associated with greater genetic influences on antisocial behavior.

However, in contrast to what is observed in adolescence, the diathesis-stress model has not emerged as the dominant GxE model for explaining the relation between antisocial behavior and parenting practices in childhood. Instead, the bioecological model tends to be more prominent, perhaps consistent with the greater role parents play in the lives of children versus adolescents (Burt & Klump, 2014). Such findings bolster prior speculation that G x E processes may change in meaningful ways across the lifespan (Burt, 2011).

Although there is thus evidence that parenting can moderate the etiology of disruptive behaviors, it is worth noting that most of the relevant literature has focused solely on parental risk factors, such as parental negativity and poor parenting practices. There has been comparitively little consideration of possible moderation by protective environmental factors, such as parental warmth and invovlement, which represents a shortcoming in this body of work. One recent exception examined parental involvement as an etiological moderator of ADHD in childhood and found that genetic influences increased, and non-shared enivronmental influences decreased, with increasing parental involvement, consistent with the bioecological models of G x E (Nikolas, et al., 2015).

The current study attempts to further fill this gap in the literature by investigating parental involvement as a moderator of genetic and environmental influences on ODD during middle childhood. This study contributes to the existing literature through a specific focus on ODD, and positive parenting behaviors.. In this study both child and parent reports of invovement were included, as considering the unique perceptions of individual family members is often of critical importance when predicting adjustment (Harold, Fincham, Osborne, & Conger, 1997; Powers, Welsh, & Wright, 1994). We also examined both maternal and paternal involvement as potential moderators. Although fathers have historically been underrepresented in the developmental psychopathology literature (Phares, Fields, Kamboukos, Lopez, 2005), there is some evidence for differential influences of maternal versus paternal involvement on child behavior (Hoeve et al., 2009; Winsler, Madigan, & Aquilino, 2005).

Importantly, as noted by Nikolas and colleagues (2015), when discussing GxE, it is crucial to consider genotype-environment correlations (rGE), or non-random exposure to certain environmental experiences (Plomin, et al., 1977; Scarr & McCartney, 1983). Because the association between parenting behavior and child adjustment is in part due to genetic overlap between parents and offspring (Marceau et al., 2013), it is possible that ostensible GxE actually reflects a genetic correlation between the outcome and the moderator. For example, oppositional children may exhibit more negative, defiant, disobedient, and hostile behaviors toward authority figures, making them more difficult to parent, which could subsequently result in lower levels of parental involvement (Anderson, Lytton, and Romney, 1986; Snarr, Strassberg & Slep, 2003). To address this, modeling techniques were used in the current study that control for concurrent rGE.

Method

Participants

Young twins and their parents (N=1,027 families) were recruited as part of the Twin Study of Behavioral and Emotional Development in Children (TBED-C), a study within the population-based Michigan State University Twin Registry (MSUTR) (Burt & Klump, 2013; Klump & Burt, 2006). The TBED-C includes two independent samples: a population-based sample of 1,054 twins from 527 families recruited from across Lower Michigan, and an “at-risk” sample of 1,000 twins from 500 families residing in modestly-to-severely disadvantaged neighborhoods in the same recruitment area. To be eligible for participation in the TBED-C, neither twin could have a cognitive or physical condition that would preclude completion of the assessment (as assessed via parental screen; e.g. a significant developmental delay). The twins were 48.7% female and ranged in age from six to 10 years (mean = 8.03, SD = 1.49; although 30 of the 1,027 pairs had turned 11 by the time the family participated). All data were obtained according to the MSU institutional review board (IRB) approved protocols (IRB# 04-887; title “Genotype-environment interactions in childhood antisocial behavior”)

Families were recruited directly from birth records, or from a population-based registry that was itself recruited via birth records, via anonymous recruitment mailings in conjunction with the Michigan Department of Health and Human Services (formerly known as the Michigan Department of Community Health). Recruitment procedures for the at-risk sample were identical except that mailings were restricted to those families residing in neighborhoods with more than 10.5% of households in poverty. Our cut-off of 10.5% was chosen based on Census data in 2008, which indicated that the mean level of neighborhood poverty in the state of Michigan was 10.5%. Because neighborhood poverty is positively skewed, 65% of neighborhoods did not meet this criterion. Put differently, our approach resulted in a sample comprised of families residing in the most disadvantaged 1/3 of neighborhoods in Michigan. Consistent with this, the mean family income of $72,027 in our population-based sample is nearly identical to that seen in recent Census data ($73,373; see Burt & Klump, 2012), and notably larger than that seen in our ‘at-risk’ sample ($57,281). In short, our at-risk sample evidences higher neighborhood poverty and lower family income than are seen when recruiting using a population-based strategy, and thus can be considered ‘at-risk’ relative to our population-based sample

Compared to the population-based sample, the at-risk sample was significantly more racially diverse (15% Black, 75% White), reported lower family incomes (Cohen’s d = −.38), higher paternal felony convictions (d = .30), and higher rates of twin conduct problems and hyperactivity (d = .34 and .27, respectively). As indexed via a brief questionnaire administered to ~85% of non-participating families, there were few differences between participating and non-participating twins and families (see Table 1 in Burt & Klump, 2013).

Measures

Oppositional-Defiant Behaviors

Parents completed the Achenbach Child Behavior Checklist (CBCL; Achenbach & Rescorla, 2001) for each twin, while the twins completed the Semistructured Clinical Interview for Children and Adolescents (SCICA; McConaughy & Achenbach, 2001). Twins were interviewed in separate rooms by different interviewers. The twins’ teacher(s) completed the Achenbach Teacher Report Form (TRF; Achenbach & Rescorla, 2001). In the current study, we specifically used the DSM-oriented Oppositional Defiant Problems or ODP scale (Achenbach & Rescorla, 2001; McConaughy & Achenbach, 2001), which can be found on the CBCL, SCICA, and the TRF. These scales are consistent with the DSM-IV diagnostic category of ODD (e.g. argues, defiant, temper etc.; Achenbach, Dumenci, & Rescorla, 2001). Internal consistency reliabilities for the CBCL scales were adequate (α = .80 and .76 for mother and father informant reports, respectively). Roughly 10% of SCICA interviews were videotaped to obtain inter-rater reliability, and the average intraclass correlation across raters was .88.

As expected based on prior meta-analytic evidence (Achenbach, McConaughy, & Howell, 1987), the various informant-reports of ODP were modestly to moderately intercorrelated (rs range from .16 to .29, all p<.01). ODP data were averaged across informants to create a composite measure of ODP, which provides a more complete assessment of twin symptoms (Achenbach, et al., 1987). Consistent with manual recommendations (Achenbach & Rescorla, 2001), analyses were conducted on the raw ODP composite score, which were log-transformed prior to analysis to adjust for positive skew.

Parental Involvement

The parental involvement scale of the Parental Environment Questionnaire (PEQ; Elkins, McGue, & Iacono, 1997) was administered to parents and twins to assess parental involvement. This scale includes12 items with 4 point scales that assess support, closeness, and communication. Mothers and fathers individually rated their relationships with each twin, while the twins individually rated their relationships with their mother and their father. Items were the same for parents and children, with alterations in wording appropriate for particular raters. The involvement scale displayed adequate internal consistency reliability, with alphas between .72 and .81 across individual informants (e.g. parents and children). The various informant-reports were modestly, though significantly, correlated (r = .13 to .18, all p<.01).

Data Analyses

Twin studies leverage the difference in the proportion of genes shared between monozygotic or MZ twins (who share 100% of their segregating genes) and dizygotic or DZ twins (who share roughly 50% of their segregating genes) to estimate additive genetic (A), shared environmental (e.g. environmental factors that make twins similar to each other; C), and non-shared environmental (e.g. factors that make twins different from each other, including measurement error; E) contributions to a given phenotype. More information on twin studies is provided elsewhere (Plomin, DeFries, Knopik, & Neiderhiser, 2012).

For our primary analyses, we evaluated whether parental involvement moderated the etiology of ODD using the ‘extended univariate GxE’ model (Purcell, 2002; van der Sluis, Posthuma, & Dolan, 2012). As shown in Figure 1(a), the A, C, and E paths of ODD were modeled as functions of parental involvement (the moderator, M). To circumvent possible rGE confounds, the parental involvment values of both twins were entered in a means model of ODD behaviors (van der Sluis, et al., 2012). Linear moderation was then modeled on the residual ODD variance (i.e., the variance that does not overlap with parental involvement), separately for each component of variance (e.g. β_xM, β_yM, and β_zM for A, C, and E paths, respectively). Non-linear moderators are not shown in Figure 1(a), but were examined. For ease of presentation, co-twin paths and variables are also not shown in Figure 1(a).

Fig. 1.

(a) The extended univariate gene-environment interation (G x E) model. (b) The bivariate G x E model. A, C, and E represent genetic, shared environmental, and non-shared environmental influences, respectively.

The extended univariate GxE model is quite flexible. Twins are not required to be concordant on the value of the moderator (although they can be), and the moderator can be either continuous or categorical, although it should include zero. As such, the parental involvement variables were floored at zero and collapsed into tertiles to ensure sufficient numbers of pairs per cell to estimate ACE. Although the interpretation of standardized or proportional ACE estimates may be useful in some cases, it generally recommended that unstandardized or absolute ACE estimates be presented (Purcell, 2002). We thus standardized our log-transformed ODD score to have a mean of zero and a standard deviation of one to facilitate interpretation of the unstandardized values.

A series of models were then tested to determine the nature of the moderation, if any. The first and least restrictive of these models allowed for linear and non-linear moderation. At each parental involvement level, linear coefficients and quadratic moderation coeffiicients were added to the genetic and environmental paths. More restrictive models in which certain moderation paths were constrained to be zero were then estimated and compared to the less restrictive models.

Mx, a structural-equation modeling program (Neale, Boker, & Xie, 2003), was used to fit models to the raw data using Full-Information Maximum-Likelihood estimation. The competing GxE models were compared using four information theoretic fit indices that balance overall fit (via minus twice the log-likelihood; −2lnL) with model parsimony: the Akaike’s Information Criterion (AIC; Akaike, 1987), the Bayesian Information Criteria (BIC; Raftery, 1995), the sample-size adjusted Bayesian Information Criterion (SABIC; Sclove, 1987), and the Deviance Information Criterion (DIC; Spiegelhalter, Best, Carlin, & Van Der Linde, 2002). For these four indices, the lowest value among a series of nested models denotes the best model according to a particular index. Because fit indices do not always agree (they place different values on parsimony, among other things), we reasoned that the best fitting model should yield lower or more negative values for at least 3 of the 4 fit indices (Hicks, et al., 2009).

Importantly, van der Sluis and colleagues (2012) recommended that researchers confirm positive findings of etiological moderation using the bivariate GxE model (see Figure 1b; (Purcell, 2002), since the extended univariate GxE model is unable to distinguish between moderation of the covariance path and moderation of the residual path (only the latter of which represents “true” GxE). Although useful for confirming GxE in this way, the bivariate GxE model otherwise suffers from a number of problems, including issues of identifiability (Rathouz, Van Hulle, Rodgers, & Waldman, 2008). Given these problems, we restricted our core GxE analyses to the extended univariate model, and made use of the bivariate model only to confirm those results. Should results support moderation of the unique paths, it would suggest that any findings of moderation in our primary analyses reflect “true” GxE rather than moderation of the covariance terms.

Results

Boys evidenced significantly higher rates of ODD than did girls (mean (SD) = 3.59 (1.93) for boys and 3.09 (1.65) for girls; t (2031) = 6.28, p<.01). Mean levels of parental involvement were higher in girls as compared to boys; t (2023) = 2.77, p<.05. Involvement was moderately associated with twin age (r = .23, p<.01), and ODD demonstrated a small association with age (r = .05, p<.05). Sex and age were regressed out of ODD prior to analyses in order to control for these variables (McGue & Bouchard, 1984). As expected, ODD was negatively associated with parental involvement according to all four informants (rs range from −.09 to −.14, p<.01).

Primary Analyses

Formal tests of etiologic moderation were conducted next. As seen in Table 1, there was no evidence of moderation by maternal or paternal reported involvement, or by twin reported paternal involvement. However, there was evidence of moderation as a function of twin reported maternal involvement. In this case, the best-fitting model was the linear ACE moderation model. Estimated paths and moderators from the best-fitting linear models are presented in Table 2. Unstandardized or absolute genetic and environmental variance contributions to ODD at each level of twin-reported maternal involvement are plotted in Figure 2. As can be seen, A and E contributions to ODD were significantly greater than zero and moderate in magnitude at low levels of maternal involvement. Shared environmental contributions, by contrast, were small and non-significant in the presence of low maternal involvement. As mothers’ involvement with their twins increased, however, genetic influences decreased and environmental influences (both shared and non-shared) increased significantly. Indeed, in the most protective environments (e.g. those with the highest levels of maternal involvement), ODD was largely environmental in origin, with only minimal contributions of genetic influence.

Table 1.

Univariate G x E model fit statistics

Moderator models	−2lnL	Df	AIC	BIC	SABIC	DIC
Maternal involvement data collected from twins as the moderator
(1a) Non-linear ACE moderation	5282.3	1947	1388.3	−4064.9	−973.01	-
(1b) Linear ACE moderation	5286.2	1950	1386.2	−4073.3	−976.66	-
(1i) No moderation	5301.5	1953	1395.5	−4075.9	−974.56	-
Paternal involvement data collected from twins as the moderator
(2a) Non-linear ACE moderation	5133.4	1883	1367.4	−3887.7	−897.51	-
(2b) Linear ACE moderation	5137.3	1886	1365.3	−3896	−901.07	-
(2i) No moderation	5142.3	1889	1364.3	−3903.8	−904.12	-
Maternal involvement data collected from mothers as the moderator
(3a) Non-linear ACE moderation	5328.2	1953	1422.2	−4065.6	−964.16	-
(3b) Linear ACE moderation	5333.3	1956	1421.3	−4073.4	−967.22	-
(3i) No moderation	5335.1	1959	1417.1	−4082.8	−971.86	-
Paternal involvement data collected from fathers as the moderator
(4a) Non-linear ACE moderation	4448.5	1643	1162.5	−3296.4	−687.21	-
(4b) Linear ACE moderation	4451.7	1646	1159.7	−3304.9	−691.33	-
(4i) No moderation	4455.3	1649	1157.3	−3313.2	−694.83	-

Note. The best fitting model is in bold.

Table 2.

Unstandardized path and moderator estimates for the G x E Model

Moderator models	Path coefficients for main effects			Coefficients for moderation of linear paths
	A	C	E	A1	C1	E1
Maternal involvement data collected from twins as the moderator
Full ACE model (as the best fitting model)	.71*(.55, .82)	0.22(−.06, .43)	58*(.51, .68)	−.16*(−.28, −.04)	.21*(.09, .33)	.08*(.01, .14)
Paternal involvement data collected from twins as the moderator
No moderation (as the best fitting model)	.49* (.22, .65)	.50* (.32, .63)	.70* (.65, .75)	---	---	---
Maternal involvement data collected from mothers as the moderator
No moderation (as the best fitting model)	.58*(.39, .72)	.42*(.17, .57)	.68*(.64, .73)	---	---	---
Paternal involvement data collected from fathers as the moderator
No moderation (as the best fitting model)	.59* (.27, .77)	.41* (.12, .57)	.66* (.62, .71)	---	---	---

Note. A, C, and E represent genetic, shared, and non-shared environmental parameters, respectively.

The 95% confidence intervals are presented in parentheses.

Fig 2.

Moderation of ODD by level of maternal involvement as A

Note. The three lines represent the shift of unstandardized estimates of the genetic (A), shared (C), and nonshared environmental (E) variance components for ODD, based on the estimate of the proportional contributions of A, C, and E at three levels of parental involvement.

We sought to further confirm the above results using the bivariate GxE model (Purcell, 2002), as recommended by van der Sluis et al. (2012). Analyses were run only for twin reports of maternal involvement. As seen in Table 3, results revealed little evidence of moderation of the common covariance terms. There was significant moderation of the unique covariance terms however, which indicates that the above findings are likely to represent “true” GxE that is not confounded with rGE.

Table 3.

Bivariate G x E model fit statistics—Maternal involvement rated by twins as the moderator

Moderator Model	−2lnL	Df	AIC	BIC	SABIC	DIC
Full ACE model with both unique & common covariance pathways	9998.66	3907	2184.66	−8457.49	−2253.15	−4867.20
Full ACE model with unique pathways only	10000.73	3910	2180.73	−8466.73	−2257.68	−4873.75

Note. The best fitting model is in bold.

Discussion

The goal of this study was to examine parental involvement as an etiologic moderator of ODD during middle childhood. Previous research has typically only considered ODD in conjunction with other disorders, if at all (Burke, 2009). Moreover, the few studies that have specifically examined genetic and environmental influences on ODD have typically restricted their analyses to the main effects of A, C, and E (Hudziak, et al., 2005), and have not considered whether and how these influences might vary by parental involvement.

The current study offered some support for the notion that the etiology of ODD varies as a function of parental involvement. However, these findings were restricted to maternal involvement as perceived by twins. Results revealed that genetic influences decreased and environmental influences increased with increasing levels of maternal involvement as perceived by twins. Such findings are most consistent with a diathesis stress model, which posits that risky enivornmental contexts (which include low levels of protective factors) “activate” genetic influences on psychopathology. Thus, maternal involvement would appear to act as a protective factor for ODD in that heightened levels of maternal involvement may be able to counteract genetic predispositions towards ODD, or reduce the role of genetic variability in accounting for individual differences between children.

The search for protective factors within the family context that can potentially counteract a genetic predisposition for behavior problems in children is important for both understanding and treating/preventing various disorders. As Marceau and colleagues (2013) have pointed out, behavioral genetic studies investigating the role of rGE and G x E effects in the development of disruptive behavior (including ODD) have shown that although a child’s genes may influence his or her behavior, the family helps to control how genetic influences operate. The present findings add to this literature both by highlighting the importance and potential role of maternal involvement in the development of ODD behaviors, as well as by broadly contributing to the understanding of the mechanisms by which the interplay between genetic and environmental forces contribute to the risk for psychopathology in childhood (Belsky & Pluess, 2009).

Indeed, knowing how and why parental involvement positively affects development can in turn further inform and facilitate the creation of various family level interventions and policies to both prevent and address behavioral disorders. For example, these findings add further support and justification for “parent training” programs that encourage and provide tools for parents in “at risk” families that help them be more involved with their children. Additionally, these results further emphasize the importance of pro-family policies, such as those providing paid family leave, that acknowledge and support parents who want to be involved with their children early on, but are constrained by their circumstances (e.g., low SES and working multiple jobs). Although very involved parents may of course still have children with ODD behaviors, knowing that a genetic predisposition for such behaviors is, to be sure, “not destiny” and can potentially be muted by a positive environment (i.e., maternal involvement), helps to support development on the broad scale, benefiting families, children, and society as a whole.

One important caveat to our study’s results is that child-perceived maternal involvement moderated the genetic variance of ODD behaviors, but paternal involvement did not. One possible explanation for the lack of a significant moderation effect for paternal involvement may be that fathers and mothers assume different parenting roles. Mothers have historically been assumed to be the primary caretakers, and although paternal involvement is higher than previous generations, mothers remain largely accountable for child care responsibilities (Doucet, 2013). Thus the effects of fathers’ involvement may be drowned out by that of mothers. Another possible explanation is that our measures of parental involvement may not have been sensitive enough to capture the types of paternal behaviors that could have a positive impact on reducing children’s ODD behaviors. Future research is needed to explore these possible explanations in more detail. That said, there is a need for cautious interpretation about sex differences in the effects of parenting. Most critically, our findings do not suggest father involvement is unimportant to child ODD or child development. Recent meta-analytic evidence (Hoeve, et al., 2009) for example suggests that fathers’ supportive behavior is more strongly related to delinquency than mothers’ support. Although paternal involvement did not moderate the etiology of ODD in middle childhood, it is still likely relevant to ODD through other mechanisms (e.g., direct modeling, etc.).

One, perhaps troublesome, detail is that evidence of moderation was specific to twin perceptions of maternal involvement. This could reflect one of two possibilities. First, the finding for twin perceptions could be spurious, and there is no real moderation. We would argue against this interpretation, in part because the evidence of moderation for twin perceptions of maternal involvement was rather strong and persisted across all three variance components. Moreover, there is a second possibile interpretation for this discrepancy. Namely, the informant discrepancies may actually be meaningful.

It is not rare to find rater differences when examining correlates of child behavior (Clark, Durbin, Hicks, Lacono, & McGue, 2016). This is perhaps unsuprising given that different raters of child behavior tend to only show modest agreement with one another (e.g. Achenbach, et al., 1987; Duhig, Renk, Epstein, & Phares, 2000). Importantly, these trends extend beyond ratings of child behavior to reports of parenting behavior as well. It is difficult to say with certainty whether one set of reports is more “correct” than the other. Parents, for instance, may feel pressed into reporting socially desireable parenting behaviors. To be sure, parental reports of involvement demonstrated less variability compared to twins’ reports (SD = 3.05, 4.21, 5.78, 6.99 for mothers, fathers, and twins 1 and 2, respectively). On the other hand, the self-reports of children at this age may be less reliable than those of adults, and may also be more prone to an acquiesence bias (Soto, John, Gosling, & Potter, 2008).

However, Reiss and colleagues (2000) found that third party observers’ ratings of parenting behvior corroborated the reports of young siblings much more so than the reports of the parents. Moreover, children’s subjective perceptions of some parenting behaviors may be more meaningful than the realtiy (or parents’ own perceptions) in certain circumstances. For example, Clark and colleagues (2015) found that adolescents’ perceptions of parental monitoring were more strongly related to substance use than their parents’ reports of monitoring efforts.

Thus, we would argue that the discrepancy across parent and child reports is likely substantively meaningful. Regardless of what parents’ report (and may actually do), children’s subjective perceptions and experiences of their parents may be the biggest factors in driving the exaggeration or attentuaion of genetic effects. This further opens the door to the possibility that children who are dispositionally predisposed towards negativity may view even involved parenting as bereft, and thus may not receive the protective benefits of such parenting. These results also further reinforce the importance of collecting data from multiple sources, as children and parents’ perceptives each provide unique information for untangling developmental phenomena.

Limitations and Future Directions

There are some limtations of this study that need to be acknowledged. First, although previous research has indicated that ODD heritability estimates do not vary significantly across sex (Burt, 2009; Hudziak, et al., 2005), it may nevertheless be worthwhile for future studies to examine whether the effects identified here differ across boys and girls using larger samples where a model with this degree of moderation is more analytically tractable and reliable. Similarly, future work can jointly examine the influence of maternal and paternal involvement in a single model, which we were unable to do given our sample size (which is moderate, by twin study standards), and the lack of appropriate vetted models. Our results should also be considered specific to the developmental period of middle childhood, and do not extend to either early childhood or adolescence. Moreover, as the overall sample remains predominantly white, the findings of present study may not extend to all ethnic groups. Indeed, the present results do not necessarily imply that all GxE effects underlying the etiology of ODD are diathetic in nature. In addition to generalizing the findings here to different groups and age ranges, future research needs to take other parenting behaviors into account, esepcially more protective behaviors.

Conclusions

Overall, two main findings emerged from the current study on G x E effects in the context of parental involvement and ODD in middle childhood. First, children’s perceptions of maternal invovement moderated the etiology of ODD, but parents reports of involvement did not. Second, the nature of the observed moderation supported a diathesis-stress model of gene-enviroment interaction. Our results support the notion that the early family environment is important for understanding the development of ODD, moreover, it may be the case that children’s idiosyncartic impressions of that environment are especially important.