Parenting Deater-Deckard, K. (in press). Parenting the genotype. In K. McCartney & R. Weinberg (Eds.) Festschrift for Sandra Scarr (title forthcoming). London: Taylor and Francis.

ACKNOWLEDGEMENTS I am grateful to the participants and research staff who contributed to this research, and for the generous collegiality of the principal- or co-investigators of these studies including Naama Atzaba-Poria and Alison Pike (British Family and Child Behavior Study), Stephen A. Petrill (Western Reserve Reading Project, Northeast-Northwest Collaborative Adoption Projects), Robert Plomin (TRACKS Twin Study, Colorado Adoption Project), and Lee Thompson (Western Reserve Reading Project). I also thank Paula Mullineaux and Jungmeen Kim for providing input on a prior draft of the chapter. AUTHOR NOTE Kirby Deater-Deckard, Ph.D., is Professor of Psychology at Virginia Polytechnic Institute and State University. Correspondence can be addressed to the author in the Department of Psychology, Virginia Tech, 109 Williams Hall (0436), Blacksburg, VA, 24061; Tel: (540) 2310793; Email: [email protected]

1

Parenting

2

Parenting the Genotype Research on parenting is thriving, having emerged over the past three decades as its own discipline within psychology and developmental science. Sandra Scarr’s work has been a catalyst. She has provided innovative and provocative ideas that challenge assumptions about the causes and consequences of parenting behavior, and the connections between parenting and individual differences in children’s developmental outcomes (Scarr, 1992, 1993). In particular, Scarr’s theory has prompted researchers to consider more seriously the role of child effects on parenting. Investigation of parents’ effects on their children and children’s effects on their parents has been crucial in child clinical and developmental psychopathology research, given that the identification of inter-personal mechanisms within families is necessary for developing new and more effective interventions. The literature shows that parents of children with persistent behavioral problems (e.g., inattentiveness, opposition/noncompliance, aggression) are more distressed and show difficulties in parenting that include hostility, insensitivity or nonresponsiveness—and in more extreme situations, child maltreatment or abuse. There remains little doubt that distressed parenting contributes to the development and maintenance of these problematic child behaviors. However, it also has become clear that the family processes underlying these connections also include child effects on parents (Deater-Deckard, 2004; O’Connor, 2002; Patterson, 1997). In the current chapter, I present findings from a series of studies in which my colleagues and I have tested one of Scarr’s ideas: that genetically influenced child behaviors causally elicit certain parenting behaviors, and that this explains the differential parenting of sibling children that occurs within families.

Parenting

3

Scarr’s Developmental Theory Scarr’s theory of genotype environment effects (Scarr, 1992, 1993; Scarr & McCartney, 1983) states that over the course of development in early and middle childhood, children increasingly select their own experiences that are consistent with their particular interests and attributes, while also consistently eliciting or evoking certain behaviors from their caregivers, siblings, and peers. Using Plomin, DeFries, and Loehlin’s (1977) taxonomy of geneenvironment correlations, Scarr and McCartney proposed that these developmental shifts in person environment transactions are the vehicle for genotype environment transactions. As a result, genetically influenced individual difference attributes, such as intellectual and interpersonal skills, serve to constrain environments and experiences. They proposed that shared environmental influences—non-genetic influences that cause sibling similarity in phenotypes— should be most prominent in early childhood, when the effects of the home environment are most likely to exert their influence. As development proceeds, children gain competencies and are exposed to more frequent and varied opportunities for selecting the people they spend time with and the experiences they encounter (active gene-environment correlation). Children also accrue more experiences in which they consistently elicit or evoke certain behaviors from other people (evocative or reactive gene-environment correlation). Accordingly, while shared environmental influences decrease in their influence on individual differences over childhood, heritable influences increase in prominence—in no small part because of the effects of these genetically influenced active and evocative person-environment processes. This theory has an important implication for how one thinks about the development of individual differences within families. Across several children who are born to the same parents and who are raised together, the parenting environment that each child experiences will differ. If

Parenting

4

parenting is systematically related to stable and genetically influenced child attributes (i.e., traits), these “child-specific” or “differential” parenting environments might arise as a direct result of individual children’s effects on their parents’ behavior. They also arise as a result of non-genetic influences—nonshared environmental factors that account for sibling differences in complex behaviors over and above genetic influences on sibling differentiation. The net effect is that in most families, siblings have distinct childrearing experiences that correspond with the influences of genetic as well as non-genetic factors (Dunn & Plomin, 1990; Plomin, 1994). The degree of sibling differentiation in attributes, and its connection with differential parenting environments, often is substantial. In our research, we have sought to identify potential child effects on parenting by focusing on two aspects of child behavior— inattentiveness/distractibility and conduct problems (e.g., noncompliance/opposition and aggression)—and three aspects of parenting: negativity/warmth, restrictive control (including physical punishment), and sensitivity/responsiveness. Figure 1 shows data that follow a general pattern—when a correlation is found between a complex child behavior (such as conduct problems) and a parenting behavior (such as negativity or hostility) in variation between families, a similar correlation typically is found within families (based on analysis of sibling differences). This figure includes correlations between child conduct problems and maternal negativity from three sibling behavioral genetic studies—the TRACKS twin study of 3.5-year-old same-sex twins in England, the Western Reserve Reading Project (WRRP) study of 4-to-8-year-old samesex twins in the US, and the Northeast-Northwest Collaborative Adoption Project (N2CAP) of 316-year-old adoptive siblings in the US (see Deater-Deckard & Petrill, 2004, and Petrill, DeaterDeckard, Thompson, DeThorne, & Schatschneider, 2006). These data represent covariation between conduct problems and maternal negativity in two ways. For the between-family

Parenting

5

analyses, one child per family was selected and Pearson correlations were estimated—this is the traditional approach taken in socialization studies. For the within-family analyses, sibling relative difference scores were computed for conduct problems and maternal negativity (one child’s score was subtracted from her sibling’s score), and Pearson correlations between these two difference scores were estimated. Analyses were conducted for observers’ ratings of mother-child interactions and for parents’ reports using questionnaires, including a self-report instrument for maternal negativity and the Strengths and Difficulties Questionnaire (Goodman, 1997) or Child Behavior Checklist (Achenbach, 1991) for ratings of child conduct/externalizing problems. In order to minimize rater effects, independent observers were used to rate each of the two mother-child dyads within each family, and analyses of parents’ ratings used mothers’ self-reported negativity and fathers’ reports of child conduct problems. Observers’ ratings were based on videotapes of 10 to 20 minutes of parent-child interactions that were filmed during visits to the home and in which the dyad completed several cooperative tasks. The dyads were given an Etch-A-Sketch drawing toy and pictures to copy together, with the rule that parent and child each could operate one dial but not touch each other’s dial. Most of the dyads also completed a second task in which the parent and child had to use their own dials to move a marble through a tilting Labyrinth Maze game. Trained and reliable coders rated each parent-child pair using global ratings. A different coder rated each mother-child dyad within any given family, to eliminate rater bias effects on estimates of sibling similarity. As Figure 1 shows, the association between more conduct problems and more maternal negativity was comparable between families (the usual method in the literature) and within families. Although not shown in this figure, the findings were very similar for the link between

Parenting

6

maternal negativity and child anxious/depressive symptoms (i.e., internalizing problems), and for associations between maternal differential warmth/positivity and child conduct or internalizing problems. Thus, a child who was receiving higher levels of maternal negativity or lower levels of maternal warmth also exhibited higher levels of conduct and internalizing problems, and this was true when that child was compared to children from other families, as well as when compared to her or his sibling within the same family. What is most striking is that the effect sizes are comparable even though the sources of variation are so different. In the between-family analyses, the variability is from genetically unrelated children in different households, but in the withinfamily analyses the variability is from genetically related siblings (typically). The point to be taken from these correlational analyses is that the statistical effect linking children’s differential parenting environments and sibling differences in behavioral and emotional problems is as large and stable an effect as is the widely studied effect found between families. Siblings are “receiving” different levels of parental hostility and warmth, and this is tied to sibling differences in behavioral and emotional problems in unsurprising ways. Child Effects Although useful in a descriptive sense, the correlations in Figure 1 do not permit inferences about whether and how child effects might explain the within-family covariation between behavioral/emotional problems and parental negativity/warmth. Scarr’s theory offers one testable explanation—that the within-family data in Figure 1 represent child effects on parenting, and more importantly, that these child effects are explained by genetically influenced child attributes and behaviors. Many of today’s parenting scientists look to Richard Q. Bell’s Psychological Review (1968) article as one of the most seminal papers for contemporary discourse on children’s effects

Parenting

7

on their caregivers’ behaviors. Bell encouraged developmental scientists to conceptualize socialization within families not as simply unidirectional (from parent  child), but as systematically bi-directional, with the child influencing the parent’s behavior as well. There are at least two ways to conceptualize bi-directionality. One approach is to view bi-directionality as representative of on-going and intertwined parent and child effects that operate in transactions (e.g., Sameroff, 1975). Another approach is to view bi-directionality as reflecting different types of processes—some that reflect parentchild effects primarily, and others that reflect childparent effects primarily (e.g., Lytton, 1977; 1991). Both approaches require methodologies that discern child and parent effects. However, it is not possible to disentangle the child and parent’s influences on each other. Therefore, experimental and quasi-experimental designs have been developed, accompanied by statistical tools for variance decomposition. These include “round robin” experimental designs and statistical tools such as the Social Relations Model or Actor-Partner Interdependence Model (Cook & Kenny, 2005). These types of methods have been applied to understand better the wellestablished connections between child behavioral problems (e.g., impulsivity, hyperactivity, inattention, conduct problems) and multiple dimensions of parenting behavior that are found in clinical and community studies. Children who show higher levels of behavioral and attention problems experience parenting environments that are more distressed, reactive, and hostile (e.g., Johnston & Mash, 2001; Patterson, 1997). The findings from experimental studies—some involving observations of social interactions between disordered children and adults (Anderson, Lytton, & Romney, 1986) or age-mate peers (Coie & Kupersmidt, 1983; Dodge, 1983), and some involving psychopharmacological interventions (Wells et al., 2000)— have provided some

Parenting

8

evidence that children’s behavioral problems often do consistently evoke or elicit more hostile and harsh treatment from others. However, those studies have not addressed whether and how genetically influenced child effects may be operating. For that, quasi-experimental behavioral genetic designs and structural modeling techniques are needed, that permit the decomposition of genetic and non-genetic sources of variance and covariance between child and parent behavior (Neale & Cardon, 1992). Theories and methods for testing hypotheses about how genetic influences might operate through child effects on parenting were developed in the 1970s and 1980s (Dunn, 1993; Lytton, 1977; Plomin, 1994; Rowe, 1981; 1983). By the 1990s, a number of large-scale longitudinal studies were being conducted—studies that integrated behavioral genetic and traditional family socialization designs such as the Colorado Adoption Project, CAP (Petrill, Plomin, DeFries, & Hewitt, 2003) and the Nonshared Environment in Adolescent Development Project, NEAD (Reiss, Hetherington, Neiderhiser, & Plomin, 2000). These studies set the stage for genetically informative socialization studies being conducted today, including those presented in the current chapter. They also provided clear evidence that the genetic variance in children’s behavioral adjustment was accounting for some (and sometimes most) of the child genetic variance in measures of child-specific parenting within families (e.g., Ge et al., 1996; O’Connor et al., 1998)—that is, evidence of evocative gene-environment correlation effects (see also Plomin’s chapter in the current volume). Child Attention Span/Persistence and Maternal Warmth/Negativity In our work, we have tested some of these ideas using observational and questionnaire methods with samples spanning early childhood through early adolescence. In a series of studies, we have examined the correlation between maternal affect and child attention span. The data

Parenting

9

seem to suggest a developmental shift from early to middle childhood, with the emergence of genetically influenced child effects on maternal warmth and hostility operating through poor attention span. In the TRACKS twin study of preschoolers, we developed an attention span composite that included a behavioral performance task (number of completed trials in a dull touch-screen task), a tester’s rating of the child’s attention span and persistence during a lengthy cognitive assessment, and a separate observer’s rating of the child’s “on task” behavior during a parentchild interaction task in the home. In the subsequent WRRP twin study of 4-8 year old twins, we adapted the composite to discard the time consuming behavioral task and to include three items from the Teacher Report Form (Achenbach, 1991) tapping problems with concentration, attention span, and task completion. Although the composite score is based on multiple informants and methods, it shows very good internal and external validity, as well as the expected moderate stability of individual differences and modest mean level increases over one year in middle childhood (Deater-Deckard, Petrill, Thompson, & DeThorne, 2006). Behavioral genetic analyses showed evidence of a developmental shift, with an increase in heritable variance and decrease in shared environmental variance over the transition to middle childhood (see Figure 2). In the data for preschoolers (the 3-year-olds in TRACKS and the 4year-olds in WRRP; the first set of bars in the figure), MZ and DZ twin intra-class correlations were comparable, meaning that genetic variance was negligible but shared and nonshared environmental variance were moderate in magnitude at these younger ages. In contrast, as shown in the second and third set of bars in this figure, MZ twin similarity appeared to increase and DZ twin similarity to decrease as a function of age. Although these are cross-sectional data, they suggest that genetic variance in attention span/persistence increases from a modest to substantial

Parenting

10

level and shared environmental variance dissipates over the transition to middle childhood. Consistent with this pattern was evidence that much of the stability of individual differences in attention span over one year was accounted for by stable genetic influences over time (DeaterDeckard et al., 2006). There is a statistical association between attention span/persistence and maternal warm responsive behavior that also appears to shift with development. Among preschoolers, most of the moderate correlation between child attention span and observed maternal warmth was accounted for by shared environment effects (Petrill & Deater-Deckard, 2004). However, as the shared environmental variance decreased to zero and heritable variance increased substantially (see Figure 2), an evocative child effect seemed to emerge. For the 6-8-year-olds in the WRRP twin study, the correlation between maternal warmth and child attention span was accounted for in part by child genetic covariance. The twin cross-correlations (i.e., twin 1’s attention span and twin 2’s maternal warmth) were .23 for MZ twins and .07 for DZ twins. The same effect was found for self-reported maternal negativity. The phenotypic correlation of -.27 with child attention span included a genetic correlation of .40—an effect that very likely reflects a genetically influenced child effect on maternal hostility that is operating through child attention problems (Mullineaux, Deater-Deckard, Petrill, & Thompson, 2007). Although these data are cross-sectional, they are consistent with Scarr and McCartney’s (1983) developmental theory of geneenvironment effects, demonstrating their proposed shift from shared environmental to child genetic influences. Mother-Child “Mutuality” and Child Conduct Problems In a closely related line of research we have been examining maternal sensitivity and responsiveness, operationalized as mother-child “dyadic mutuality”. From early in life, a child

Parenting

11

and her mother establish mutuality—a stable pattern of reciprocal affective and behavioral reciprocity. Differences between mother-child dyads emerge in toddlerhood, and these dyad differences between families vary widely. Kochanska’s work has shown that this developmental process and the variation between families is associated with a variety of child attributes including better self-regulation, fewer conduct problems, and more prosocial cognitions and behaviors (Kochanska, 2002). Kochanska’s research converges with other studies of parent-child reciprocity and children’s developmental outcomes (Harrist & Waugh, 2002). However, that entire literature is limited, in that it is based on studies of only between-family variation in child and parent behavior. Using Kochanska’s construct, we have been examining within-family differences in dyadic mutuality using observers’ ratings (based on the videotaped Etch-A-Sketch and Labyrinth games described previously) of parent responsiveness to child, child responsiveness to parent, parent-child verbalized cooperation, and parent-child emotional reciprocity (i.e., matched or shared positive affect; a “conversation like” quality to the interaction). Previously, we tested evocative gene-environment effects in mutuality in the TRACKS and CAP studies (DeaterDeckard & O’Connor, 2000). However, those analyses were based on fairly small samples and were limited to early childhood. For the analyses that follow, five datasets containing observers’ ratings of mutuality spanning early childhood to early adolescence were compiled. The resulting dataset included families from the TRACKS, N2CAP, and WRRP studies (described above) as well as data on full and adoptive same-sex siblings in the US from the CAP study (Petrill et al., 2003), and male and female singletons from Caucasian and Indian/Pakistani English families in London from the British Family and Child Behavior Study or BritFam (Deater-Deckard, AtzabaPoria, & Pike, 2004).

Parenting

12

The compiled dataset included 1453 observed child-parent dyads with children ranging from 3 to 12.9 years of age. Of these parent-child dyads, 1101 were biological parent-child dyads and 352 were adoptive parent-child dyads. Most were mother-child dyads (1275), and slightly more than half were parent-daughter dyads (775 girls). For the sibling analyses, the dataset included 509 dyads with complete data: 138 monozygotic (MZ) twin pairs, 169 dizygotic (DZ) twin pairs, 56 full sibling pairs, and 146 adoptive sibling pairs. The four indicators of mutuality (mother responsiveness, child responsiveness, reciprocity, and cooperation) converged well into a composite score (with loadings on the first principal component in the .6 to .7 range). This composite showed substantial stability in its internal validity and factor structure across studies, sibling types (i.e., MZ twin, DZ twin, full sibling, and adoptive sibling) and parent sex. Mutuality was normally distributed, with a mean of 4 on the seven-point scale and the entire range of possible scores represented. Sibling intra-class correlations (with child sex and age controlled statistically, given the variations in these child factors across the five studies) are shown in Figure 3. Sibling similarity in mutuality with the same parent was accounted for entirely by the genetic similarity of the siblings, with the MZ twin correlation near .50, the full sibling and DZ twin correlations at about .25, and the adoptive sibling correlation near 0. The presence of an evocative gene-environment child effect was evident in these data, with genetic influences on child behavior being reflected in levels of mutuality in the mother-child dyad. Perhaps the most striking aspect of these findings is the near-zero correlation for the adoptive sibling pairs (the CAP and N2CAP studies). A mother’s mutuality with one child could not be predicted from the same mother’s mutuality with her other child. Interestingly, this was true in both adoption studies, even though the interactions between the two parent-child dyads were

Parenting

13

filmed two years apart on average in the CAP study, and only about 30 minutes apart in the N2CAP study. If all of the sibling similarity in mutuality with the same mother is genetically variable, is representative of an evocative child effect on maternal behavior, and reaches a maximum sibling correlation of .50 (i.e., the MZ twins in Figure 3), then a child’s mutuality with her mother should be correlated .50 with that same child’s mutuality with her father. In BritFam (the only study in which we examined children’s mutuality with both parents), we found a correlation of .47 between child-mother mutuality and child-father mutuality (Deater-Deckard et al., 2004). This is noteworthy because it extends the evidence for evocative child effects based on sibling comparisons to evidence based on parent-offspring comparisons within families. In regards to a connection between mutuality and child problem behaviors, we found that children with more conduct problems had less mutuality with their mothers in the N2CAP adoption study (r = -.30; Deater-Deckard & Petrill, 2004) and the BritFam study (r = -.28; Deater-Deckard et al., 2004). This replication was interesting, because the two samples of families were so different in their socio-demographic characteristics. N2CAP was comprised of middle-to-high-income white adoptive parents and adoptees (most of whom were internationally adopted, typically from Asia or Eastern Europe) living in US suburbs. BritFam consisted of lowto-middle-income white and Asian (Indian or Pakistani) biological parents and children living in central London, England. The statistical association between child conduct problems and lower mutuality was the same within families as it was between families—mirroring the pattern found for conduct problems and maternal negativity shown in Figure 1. In the N2CAP adoptive sibling study, there was a .38 correlation between the sibling relative difference score for mutuality with the same

Parenting

14

parent, and the sibling relative difference score for conduct problems. Thus, the child who was higher in conduct problems also was the child who experienced less mutuality with her mother, compared to her sibling. Recall that all of the sibling similarity in dyadic mutuality was due to genetic variance. In addition, and consistent with the large behavioral genetic literature on antisocial behavior problems (Rhee & Waldman, 2002), most of the sibling similarity in conduct problems in these studies was attributable to genetic variance (Deater-Deckard, Petrill, & Thompson, 2007; Deater-Deckard & Petrill, 2004). Among the school-age children in the WRRP twin study and N2CAP adoption study, sibling intra-class correlations for conduct problems were .72 for MZ twins, .48 for DZ twins, and .08 for genetically unrelated adoptive siblings (i.e., heritability of .4-.5, and shared environmental variance from .1-.2). Taken together, these findings suggest that genetically-influenced child conduct problems are implicated in an evocative gene-environment child effect on maternal sensitive responsiveness. Shared Environment Our research on child attention and conduct problems and maternal warmth/negativity and dyadic mutuality points to the potential role of genetically-influenced child effects. However, shared environmental influences cannot be disregarded. Shared environmental variance typically is modest compared to genetic and nonshared environmental variance, but the examination of potential shared environment effects is as illuminating as tests for evocative child effects. Consider again the connections between parenting and children’s conduct problems. Child externalizing behavior problems include some shared environmental variance. Rhee and Waldman’s (2002) meta-analysis showed that about one-sixth of the variance in antisocial behavior was due to shared environmental influences—substantially less than the variance due to

Parenting

15

genes, but certainly not a trivial effect either. This shared environmental variance is a potential clue about a different set of parenting processes that may be related to sibling similarity (rather than sibling differentiation) in conduct problems within families, even after genetic effects are controlled. These processes probably reflect parent effects on children, including the use of harsh, strict discipline. We have studied harsh physical discipline within families in much the same way that we have studied mutuality and negativity/positivity, but the results are markedly different. Harsh discipline tends to be used with all children in the family if it is used at all. Although one child may be punished physically more often or more severely than her or his sibling, overall sibling similarity in “exposure” to physical punishment at home is substantial. Harsh physical discipline is more difficult to measure than mutuality or maternal warmth and negativity. Hours of observations in the home rarely yield opportunities to observe a single physical punishment event, let alone to understand how frequently and severely it occurs for each child in the family. Therefore, we have used an in-depth interview methodology with parents to assess their attitudes and practices regarding physical punishment and other discipline methods. This interview permits the parents to describe how recently and how frequently they currently use or previously used physical punishment with each child in the family. This produces an overall “harshness of discipline” score for each child in the family, based on a coding system developed by Dodge, Bates, and Pettit (1990). In data from the TRACKS twin study of preschoolers and the N2CAP adoptive sibling study of school-age children, sibling intra-class correlations for harsh discipline were .77 for MZ twins, .66 for DZ twins, and .65 for adoptive siblings (Deater-Deckard, Ivy, & Petrill, 2006). The correlation between harsh discipline and child conduct problems was .25. Thus, parental strict, harsh discipline included ample shared environmental variance and was correlated with child

Parenting

16

conduct problems—which, as discussed above, often include at least some shared environmental variance. Instead of a shared environmental effect, these findings could reflect passive geneenvironment correlation, whereby the statistical association between harsh discipline and child conduct problems reflects shared genes between parents and children that influence harsh, aggressive types of behavior in both generations of family members. We have found no evidence for this alternative explanation. The correlation between discipline and conduct problems was .26 for biological parent-child pairs and .21 for genetically unrelated (adoptive) parent child pairs. This finding replicates what we have seen in our other studies. For example, the parent and child behaviors that comprise dyadic mutuality are moderately to substantially inter-correlated, regardless of whether the parent and child are genetically related. Furthermore, recall that the correlation between child conduct problems and mutuality was -.30 for adoptive parent-child pairs in the N2CAP adoption study, and -.28 for the biological parent-child pairs in the BritFam study—no difference as a function of parent-child genetic relatedness. An even more stringent test within the N2CAP study was possible, because it included a subset of 47 families who had one adopted child and one biological child living in the same household. The correlation between mutuality and child conduct problems for the biological parent-child pairs in N2CAP was -.37, an effect that was the same for the adoptive parent-child pairs. Simply put, in all of our studies and across many analyses of different sets of variables, we have yet to see any evidence of passive gene-environment correlation effects in our measures of parenting behavior. Although parent-child genetic similarity may not moderate the link between harsh discipline and child conduct problems (i.e., there is no passive gene-environment correlation), maternal warmth/negativity does operate as a moderator. Using a composite measure of maternal

Parenting

17

warmth that combined self-reported and observer-rated parental behavior, conduct problems and harsh discipline were moderately correlated in the low-warmth dyads (.40), but only negligibly associated in the high-warmth dyads (.09). Again, this pattern was very consistent across sibling types (i.e., adoptive or biological). It suggests that the connection between harsh discipline and child conduct problems arises when physical punishment occurs within the context of a parentchild relationship that lacks warmth or is imbued with negativity. Interestingly, harsh discipline and maternal warmth were only modestly correlated in our studies (r = -.17); these represent distinct aspects of parental behavior and of each parent-child relationship within the family. The pattern in these maternal warmth/negativity and discipline data reinforce an important point made by Plomin and Bergeman (1991) and Lytton (1991)—that the affective or emotion-laden aspects of parenting behavior appear to be much more sensitive to genetically variable child effects, in contrast to parental control and discipline strategies that appear to be most strongly related to parents’ own attributes (e.g., childrearing attitudes, parenting stress). While it may be true that a parent who frequently uses physical punishment with one child will do so with all of her children at one time or another, the affective quality of each of these parentchild relationships within the family will differ and be influenced powerfully by each child’s specific attributes. The more general point is that these data remind us not to over-generalize parenting behaviors as being inherently and primarily influenced by child effects, parent effects, or both. Certain aspects of parental behavior may be far more reactive while other aspects may be very stable over time and across children in the family. Parent Factors There is an obvious competing explanation for the data presented in Figure 1—that parents behave differentially toward their sibling children in ways that cause sibling

Parenting

18

differentiation in behaviors and attributes—a “parent effect”. If that is the case, then correlates or statistical predictors of differential parenting should be identifiable—predictors of both the degree and direction (i.e., why one child is treated differently from her sibling) of differential parenting that can account for the link between child-specific parenting and sibling differences in behavior. Are the most differential parents behaving that way because there is something different about them as parents, compared to their less differential counterparts? Or are they more differential because their sibling children simply are more different from each other (in part due to differences in genes) when compared to the sibling children of less differential parents? In attempting to answer this question, we have found a very different picture from the findings in Figures 2 and 3 when we have examined parent factors (e.g., parent age, gender, childrearing attitudes, stress, and socioeconomic status) that might explain why some parents are more differential than others. Considering first the results from the TRACKS twin study of 3.5-yearolds, we computed absolute difference scores for maternal negativity toward siblings, to estimate the overall magnitude of differential negativity in each family. We then estimated correlations between these scores and variables representing maternal and paternal education levels and occupational prestige scores. These correlations ranged from ± .12 to .27. A few were significant, but the correlations were in the opposite direction for parent- versus observer-rated maternal negativity. Thus, the analyses yielded no discernible pattern of covariation between parental education and occupational prestige and within-family differential negativity. In the WRRP and N2CAP studies of older sibling pairs in the US, we found negligible and nonsignificant correlations (± .00 to .16) between the overall magnitude of differential negativity and parent factors, based on analyses of a broader set of indicators including mother/father years

Parenting

19

of education, occupational prestige scores, number of stressful life events in the past year, and traditional childrearing attitudes (from the Parental Modernity Scale, Schaefer & Edgerton, 1985). Furthermore, we found the same result in all three studies when we analyzed differential maternal positivity. Also, in the N2CAP study—our only study that included non-twin and opposite-sex sibling pairs—the magnitude of differential negativity and positivity was uncorrelated with sibling age or sex differences. Thus, in all three studies, the only variables that seem to explain any variance in the degree of differential maternal negativity and positivity are those that represent the degree of sibling differences in behavioral and emotional problems (see Figure 1). Nonshared Environment Even though it may not be possible to predict which parents are more or less differential in their parenting based on the parents’ own attributes or circumstances, this does not mean that genetically influenced child effects are the only explanation for the emergence of child-specific parenting environments within families. Differential parenting also can arise through nonshared environmental mechanisms—non-genetic influences that do not lead to sibling similarity in behavior. Sibling differences in genotypes can help explain why a parent behaves differently toward each child in the family, but nonshared environment effects are pervasive and sometimes account for large proportions of the variance in differential parenting as well as complex behavioral phenotypes (Plomin & Daniels, 1987). The importance of nonshared environment is illustrated in our data on attention span/persistence described above. In addition to showing emergence of heritable variance (see Figure 2) and evidence of an evocative child effect on maternal warmth and negativity, the data showed a developmentally stable nonshared environmental connection between observed

Parenting

20

maternal warmth and child attention. In longitudinal analyses of identical twin differences in the degree of change in attention span over one year, there was a correlation of .28 with differential maternal warm supportive behavior (Deater-Deckard et al., 2007). The identical twin who showed a larger increase in attention span over one year also was receiving more maternal warmth, compared to her co-twin. Mullineaux and DiLalla (2007) found the same pattern in their brief (2.5 minutes) observations of differential maternal warmth and identical twin differences in increasing attention span from one to three years of age. This suggests that from early in life, sibling differences in attention span are associated with differential maternal warm supportive behavior—above and beyond any genetic child effects that may be operating. Clearly, nonshared environment is central to any discussion of child-specific parenting within families, and identifying nonshared environmental mechanisms is important because so much of the variance in individual difference attributes appears to derive from nonshared environmental sources. However, there are some limitations with respect to understanding whether and how children’s attributes evoke or elicit parental behavior when seeking to identify nonshared environment processes (Turkheimer & Waldron, 2000). First, the nonshared environment component of variance and covariance that is derived using a behavioral genetic design includes random measurement error. This can make it difficult to identify nonshared environmental factors (such as those that might arise from differential parenting) that actually account for nontrivial amounts of the nonshared environmental variance in children’s developmental outcomes. Second, identifying a nonshared environmental process (e.g., the attention span—differential warmth link just described above) implies nothing about parenting effects on children or child effects on the parent. Differential parenting that reflects nonshared environmental processes could be causing sibling differences in behavior, be in response to

Parenting

21

sibling differences in behavior, or be both. These caveats aside, child-specific parenting environments within families comprise nonshared environmental as well as genetically influenced child effects, and these different processes need not show similar patterns over the course of development. Conclusions Scarr’s work and the research that it has inspired have broad implications for research on parenting. Studying the ways in which parents behave toward and actively socialize their children provides a foundation upon which rests the study of within-family as well as betweenfamily differences in parenting. These different levels of analysis (within- and between-families) are not in competition; findings based on one level of analysis inform the other. That is, the causes and consequences of child-specific parenting within families can be interpreted within the context of all that we know from the literature on between-family differences in parenting. At the same time, our understanding of between-family differences in parenting can be enhanced by the study of differential parenting within families. Over time as the family grows, each parent-child dyad develops a relationship that is distinct from that same parent’s relationships with other children in the family. These withinfamily differences emerge as stable patterns of dyadic interaction and relationship quality across parent-child dyads. Genetic differences between siblings account for at least some of the withinfamily variation in parenting environments, most likely as a result of evocative child effects. Although it is striking how markedly different each of the parent-child relationships within a family can be, there is wide variation between parents in the degree of differential parenting that they show. For instance, among children with serious conduct or emotional problems, there is wide variation in parenting. The parents of some of these children show elevated levels of

Parenting

22

distress and harsh reactive parenting whereas others do not, due in part to between-family differences in parental self-regulation and psychopathology. Parental reactions or responses to child behavior among the most troubled parents can be powerfully modified through education and intervention, a testament to the flexibility of parental responses to children’s evocative behavior. Studying whether and how children’s individual differences elicit particular parenting behaviors within a population of families is important. It provides the knowledge that parenting scientists and practitioners need as they strive to develop more effective ways to support parents in their efforts to promote optimal developmental outcomes for their children.

Parenting

23

References Achenbach, T. M. (1991). Manual for the Child Behavior Checklist and 1991 Profile. Burlington, VT: Department of Psychiatry, University of Vermont. Anderson, K. E., Lytton, H., & Romney, D. M. (1986). Mothers' interactions with normal and conduct-disordered boys: Who affects whom? Developmental Psychology, 22, 604-609. Bell, R. Q. (1968). A reinterpretation of the direction of effects in studies of socialization. Psychological Review, 75, 81-95. Coie, J. D., & Kupersmidt, J. B. (1983). A behavioral analysis of emerging social status in boys’ groups. Child Development, 54, 1400-1416. Cook, W. L., & Kenny, D. A. (2005). The Actor-Partner Interdependence Model: A model of bidirectional effects in developmental studies. International Journal of Behavioral Development, 29, 101-109. Deater-Deckard, K. (2004). Parenting Stress. New Haven, CT: Yale University Press. Deater-Deckard, K., Atzaba-Poria, N., & Pike, A. (2004). Mother- and father-child mutuality in Anglo and Indian British families: A link with lower externalizing problems. Journal of Abnormal Child Psychology, 32, 609-620. Deater-Deckard, K., Ivy, L., & Petrill, S. A. (2006). Maternal warmth moderates the link between physical punishment and child externalizing problems: A parent-offspring behavioral genetic analysis. Parenting: Science and Practice, 6, 59-78. Deater-Deckard, K., & O’Connor, T. G. (2000). Parent-child mutuality in early childhood: Two behavioral genetic studies. Developmental Psychology, 36, 561-570. Deater-Deckard, K., & Petrill, S. A. (2004). Parent-child dyadic mutuality and child behavior problems: Gene-environment processes. Journal of Child Psychology and

Parenting

24

Psychiatry, 45, 1171-1179. Deater-Deckard, K., Petrill, S. A., & Thompson, L. (2007). Task persistence, anger/ frustration, and conduct problems in childhood: A behavioral genetic analysis. Journal of Child Psychology & Psychiatry, 48, 80-87. Deater-Deckard, K., Petrill, S. A., Thompson, L., & DeThorne, L. (2005). A cross-sectional behavioral genetic analysis of task persistence in the transition to middle childhood. Developmental Science, 8, F21-F26. Deater-Deckard, K., Petrill, S. A., Thompson, L., & DeThorne, L. S. (2006). A longitudinal behavioral genetic analysis of task persistence. Developmental Science, 9, 498-504. Dodge, K. A. (1983). Behavioral antecedents of peer social status. Child Development, 54, 13861399. Dodge, K.A., Bates, J.E., & Pettit, G.S. (1990). Mechanisms in the cycle of violence. Science, 250, 1678-1683. Dunn, J. (1993). Young children’s close relationships: Beyond attachment. Newbury Park, CA: Sage. Dunn, J., & Plomin, R. (1990). Separate lives: Why siblings are so different. New York: Basic Books. Ge, X., Conger, R., Cadoret, R., Neiderhiser, J., Yates, W., Troughton, E., & Stewart, M. (1996). The developmental interface between nature and nurture: A mutual influence model of child antisocial behavior and parent behaviors. Developmental Psychology, 32, 574-589. Goodman R (1997) The Strengths and Difficulties Questionnaire: A research note. Journal of Child Psychology and Psychiatry, 38, 581-586. Harrist, A. W., & Waugh, R. M. (2002). Dyadic synchrony: Its structure and function in

Parenting

25

children’s development. Developmental Review, 22, 555-592. Johnston, C., & Mash, E. J. (2001). Families of children with Attention-Deficit/Hyperactivity Disorder: Review and recommendations for future research. Clinical Child and Family Psychology Review, 4, 183–187. Kochanska, G. (2002). Mutually responsive orientation between mothers and their young children: A context for the early development of conscience. Current Directions in Psychological Science, 11, 191-195. Lytton, H. (1977). Do parents create or respond to differences in twins? Developmental Psychology, 13, 456-459. Lytton, H. (1991). Different parental practices—different sources of influence. Behavioral and Brain Sciences, 14, 399-400. Mullineaux, P. Y., Deater-Deckard, K., Petrill, S. A., & Thompson, L. A. (2007). Maternal negativity and child attention span/persistence: A behavioral genetic analysis. Submitted for publication. Mullineaux, P. Y., & DiLalla, L. F. (May, 2007). Differential maternal scaffolding: A potential nonshared environmental influence. Poster presented at the Association for Psychological Science, Washington DC. Neale, M., & Cardon, L. (1992). Methodology for genetic studies in twins and families. Dordrecht, The Netherlands: Kluwer Academic Publishers. O’Connor, T. G. (2002). The ‘effects’ of parenting reconsidered: Findings, challenges, and applications. Journal of Child Psychology and Psychiatry, 43, 555-572. O'Connor, T. G., Deater-Deckard, K., Fulker, D. W., Rutter, M., & Plomin, R. (1998). Geneenvironment correlations in late childhood and early adolescence. Developmental

Parenting

26

Psychology, 34, 970–981. Patterson, G. R. (1997). Performance models for parenting: A social interactional perspective. In J. E. Grusec & L. Kuczynski (Eds.), Parenting and children's internalization of values: A handbook of contemporary theory (pp. 193-235). New York: John Wiley & Sons. Petrill, S. A., & Deater-Deckard, K. (2004). Task orientation, parental warmth and SES account for a significant proportion of the shared environmental variance in general cognitive ability in early childhood: Evidence from a twin study. Developmental Science, 7, 25-32. Petrill, S. A., Deater-Deckard, K., Thompson, L., DeThorne, L., & Schatschneider, C. (2006). Genetic and environmental effects of serial naming and phonological awareness on early reading outcomes. Journal of Educational Psychology, 98, 112-121. Petrill, S., Plomin, R., DeFries, J.C., & Hewitt, J.K. (2003). Nature, nurture, and the transition to adolescence. New York: Oxford University Press. Plomin, R. (1994). Genetics and experience: The interplay between nature and nurture. Thousand Oaks, CA: Sage. Plomin, R., & Bergeman, C. S. (1991). The nature of nurture: Genetic influences on environmental measures. Behavioral and Brain Sciences, 14, 373-428. Plomin, R., & Daniels, D. (1987). Why are children in the same family so different from one another? Behavioral and Brain Sciences, 10, 1-59. Plomin, R., DeFries, J. C., & Loehlin, J. C. (1977). Genotype-environment interaction and correlation in the analysis of human behavior. Psychological Bulletin, 84, 309-322. Reiss, D., Neiderhiser, J., Hetherington, E. M., & Plomin, R. (2000). The relationship code: Deciphering genetic and social influences on adolescent development. Cambridge, MA:

Parenting

27

Harvard University Press. Rhee, S., & Waldman, I. (2002). Genetic and environmental influences on antisocial behavior: A meta-analysis of twin and adoption studies. Psychological Bulletin, 128, 490-529. Rowe, D. C. (1981). Environmental and genetic influences on dimensions of perceived parenting: A twin study. Developmental Psychology, 17, 203-208. Rowe, D. C. (1983). A biometrical analysis of perceptions of family environments: A study of twin and singleton sibling kinships. Child Development, 54, 416-423. Sameroff, A. (1975). Transactional models in early social relations. Human Development, 18, 65-79. Scarr, S. (1992). Developmental theories for the 1990s: Development and individual differences. Child Development, 63, 1-19. Scarr, S. (1993). Biological and cultural diversity: The legacy of Darwin for development. Child Development, 64, 1333-1353. Scarr, S., & McCartney, K. (1983). How people make their own environments: A theory of genotypeenvironment effects. Child Development, 54, 424-435. Turkheimer, E., & Waldron, M. (2000). Nonshared environment: A theoretical, methodological, and quantitative review. Psychological Bulletin, 126, 78-108. Wells, K., Epstein, J. N., Hinshaw, S. P., Conners, C. K., Klaric, J., Abikoff, H. B., Abramowitz, A., Arnold, L. E., Elliott, G., Greenhill, L. L., Hechtman, L., Hoza, B., Jensen, P. S., March, J. S., Pelham, W., Pfiffner, L., Severe, J., Swanson, J. M., Vitiello, B., & Wigal, T. (2000). Parenting and family stress treatment outcomes in attention deficit hyperactivity disorder (ADHD): An empirical analysis in the MTA study. Journal of Abnormal Child Psychology, 28, 543-553.

Parenting

28

Figures.

Figure 1. Correlations between measures of maternal negativity and child conduct problems in three sibling studies (TRACKS, 3.5-year-olds; WRRP, 4-to-8-year-olds; N2CAP, 3-to-16-yearolds). The dark bars show correlations for the “between family” correlation, based on selection of one child-mother pair per family. The light bars show correlations for the “within family” correlation, based on analysis of sibling relative difference scores on child behavior and maternal behavior. Figure 1a shows data for independent observers’ ratings of maternal negativity and child difficult behavior (i.e., off-task; noncompliant; frustrated or annoyed; non-responsive to mother) during structured dyadic cooperation tasks in the home. Figure 1b shows data for mothers’ self-reported negativity and fathers’ reports of child conduct problems.

Figure 2. Sibling similarity in attention span/persistence in the combined TRACKS (3.5-yearolds) and WRRP (4-to-8-year-olds) twin studies, estimated using intra-class correlations. The intra-class correlations are shown as a function of child age groups and sibling average genetic similarity: monozygotic or MZ twins (100%), and dizygotic or DZ twins (50%). Figure adapted from data in Petrill and Deater-Deckard (2004) and Deater-Deckard et al. (2005).

Figure 3. Sibling similarity in dyadic mutuality in 509 sibling pairs, estimated using intra-class correlations (controlling for child age and sex). The intra-class correlations are shown as a function of sibling average genetic similarity: monozygotic or MZ twins (100%), dizygotic or DZ twins, full siblings (50%), and adoptive siblings (0%).

Parenting Figure 1.

1a. Observer ratings 1 0.8 0.6

Between Within

0.4 0.2 0 TRACKS

WRRP

N2CAP

1b. Parent ratings 1 0.8 0.6

Between Within

0.4 0.2 0 TRACKS

WRRP

N2CAP

29

Parenting Figure 2.

1.00 0.80 0.60

MZ DZ

0.40 0.20 0.00 3 to 4

5 to 6

7 to 8

30

Parenting Figure 3.

1

0.8

0.6

0.4

0.2

0

MZ twins (138) DZ twins (169) Full sibs (56) Adoptive sibs (146)

31