Commentary on Special Section on “Bidirectional Parent–Child Relationships”: The Continuing Evolution of Dynamic, Transactional Models of Parenting and Youth Behavior Problems
- First Online:
- Cite this article as:
- Pettit, G.S. & Arsiwalla, D.D. J Abnorm Child Psychol (2008) 36: 711. doi:10.1007/s10802-008-9242-8
- 1.1k Downloads
Creative and methodologically sophisticated longitudinal research on bidirectionality in parenting and child behavior can shed important new light on the dynamics of behavior-problem development. The articles assembled for this Special Section are at the forefront of efforts aimed at tracing the interplay of parenting behavior and child and adolescent adjustment problems over time. In this commentary, we provide a brief overview of thinking and research on bidirectionality and then highlight key themes and findings reported in these articles. We describe some of the challenges in research on bidirectional processes and offer some recommendations for future research in this area.
KeywordsParent–child relationships Parenting Child and adolescent behavior problems Bidirectionality Longitudinal research
It is axiomatic that socialization processes are bidirectional, with parents shaping as well as being shaped by their children’s adjustment characteristics. Understanding how these processes emerge and express themselves has become a topic of increasing interest to researchers, practitioners, and policy makers, constituent groups who should welcome the set of papers assembled for this Special Section. Using sophisticated longitudinal modeling techniques, the authors provide clear and convincing evidence that bidirectionality underpins the relation between parenting qualities and youth adjustment difficulties. The collection of papers stands as a testament to the creative new thinking that now characterizes the study of dynamic socialization processes. This commentary highlights key themes and findings from the articles and offers suggestions as to how future research may benefit from, and build upon, these lines of inquiry. Some of the history of the field’s embrace of bidirectional models is presented first to provide a context for the discussion that follows.
Bidirectionality in Historical Context
The reinterpretation of the putative parent-effects research findings by Bell (1968) served as a catalyst for inquiry into not only the impact of children’s behavior on parenting and the parent–child relationship but also on the interplay of parent and child behavior over time. Spurred by this reappraisal and early empirical findings demonstrating evidence of child effects, researchers sought to provide a more nuanced perspective on both the conditions under which child or parent effects may be more (or less) potent, and on how growth (i.e., change) over time in parent or child attributes may hinge on the early and continuing history of the parent–child relationship. Bates (e.g., Bates 1976) provided a number of useful demonstrations of how child behavior stimuli could be experimentally manipulated to induce particular kinds of adult response, and, importantly, how adult individual differences served as moderators of the effect of the child on the adult. Later studies (e.g., Bates and Pettit 1981) pointed to gender, personality, and adult perceptions as conditional influences on child effects. Subsequent applications of the experimental paradigm (e.g., Pelham et al. 1997) showed that experimental manipulations of child deviance were associated with parents’ feelings of inadequacy and self-reported depression.
Another stream of research sought to disentangle child and parent effects by contrasting the interactions of mothers with their own vs. other children who varied in terms of adjustment problems. Using this procedure, Anderson et al. (1986) found that conduct-problem children elicited more negatives, and were less compliant, compared to normal controls, irrespective of whether they were interacting with their own or another mother. Mothers of conduct problem children were more coercive with their own children. Dumas and LaFreniere (1993) showed that mothers of competent children were positive and reciprocal toward their own and others’ children, whereas mothers of aggressive children were only positive and reciprocal toward unfamiliar children. Competent and average children were positive and reciprocal with their own and an unfamiliar mother, but aggressive children were positive only with unfamiliar mothers. Findings from these kinds of studies provided important insights into the role of child, parent, and relationship effects in the observed interaction patterns of problem-behavior children and their mothers.
Efforts increasingly were made in the 1990s to develop conceptual models and to systematically test for bidirectionality across a range of parent–child interaction and relationship contexts (e.g., Lollis and Kuczynski 1997; Kuczynski et al. 1997; Russell et al. 1998) and additional empirical evidence in support of such models began to accrue (e.g., Pettit and Lollis 1997). Theoretical models of social development became more explicitly transactional and contextual, with attention no longer focused solely on who affects whom but on the continuous dynamic (and bidirectional) interactions of the child with his or her family and social context. The seminal contribution of Sameroff (1975) was pivotal in drawing attention to these more complex forms of transactions, and more recent models (e.g., Dodge and Pettit 2003) extended the earlier formulations in useful ways by drawing attention to mediating mechanisms and cascading effects across key developmental periods. Although transactional models have currency in explaining a wide variety of interpersonal events and individual adjustment trajectories, it is in the realm of developmental psychopathology that the models have been applied and empirically tested most extensively. Among the most frequently cited and widely incorporated model of transactional effects in behavior-problem development is the coercion model, advanced by Patterson and colleagues (e.g., Patterson et al. 1992). The essential elements of the model are well-known (and summarized by many of the articles in this Special Section): A difficult, oppositional child coupled with an inept, stressed-out parent are the key ingredients in a series of transactions that fuel the development of antisocial behavior. The ingredients combine when aversive child behavior is reinforced by parents who relent and fail to provide negative consequences for the behavior. Such parental actions are themselves reinforced because the aversive child behavior is (temporarily at least) lessened. The parent–child relationship deteriorates, the child fails to acquire skills needed for relating to peers in constructive ways, deviant peers assume a prominent role in the child’s socialization, and parents withdraw (rather than confront) and fail to provide needed support and supervision. Serious antisocial behavior problems ensue.
The coercion model explains how a mismatch of child attributes (oppositional, defiant behavior) and parenting style (capitulation) engender a spiraling and destructive cycle of interpersonal dysfunction. The mismatch notion has been examined by other researchers in terms of the fit between child temperament and parenting behavior (see Bates and Pettit 2007). In our own work we have shown that children’s resistant-to-control temperament forecasts later externalizing behavior problems in the context of low but not high maternal restrictive control (Bates et al. 1998). This illustrates a pathway through which ineffective parental management and early difficult and demanding child characteristics foster the development of antisocial behavior.
The preceding overview is meant to provide a glimpse of some of the post-Bell (1968) history of theory and research on bidirectional linkages between parenting, and parent characteristics more broadly, and behavior problems in childhood and adolescence. The articles in this Special Section add importantly to the continuing evolution of thinking about bidirectionality. It is heartening that there now are several large longitudinal data sets that have the kinds of over-time measurements that enable more nuanced and developmentally-sensitive evaluations of parent→child and child→parent effects. These studies allow us to look more deeply into the workings of bidirectional/transactional effects to discern when, what, and for whom such effects are present. To foreshadow a later conclusion, the studies have only begun to explicate the “how” of such transactional effects, that is, the processes and mechanisms that explain how and why such effects operate.
Before turning to a discussion of the findings of the articles comprising this Special Issue, we note the studies differed substantially in sampling and breadth of measurement. Pardini et al. (2008) and Burke et al. (2008) provide the most extensive coverage of parenting practices, including parents’ use of physical punishment, parent–child communication, positive reinforcement, parental monitoring and supervision, timid parenting, and low positive involvement. Hipwell et al. (2008) focus on two parenting behaviors, harsh discipline and warmth, and Larsson et al. (2008) derive a single measure of parental negativity directed toward twin siblings. Gross et al. (2008), on the other hand, examine mothers’ depression, a known correlate of parenting ineffectiveness. In terms of child adjustment difficulties, most of the studies included at least one index of conduct problems, derived from parent, teacher, and/or youth report, capturing aspects of externalizing behavior problems. One study (Hipwell et al. 2008) also examined levels of child-reported depression. With respect to sampling, the Hipwell et al. (2008) sample consisted solely of girls, and Pardini et al. (2008), Burke et al. (2008), and Gross et al. (2008) sampled boys only. Three studies sampled from the Pittsburgh, Pennsylvania metropolitan area (Gross et al. 2008; Hipwell et al. 2008; Pardini et al 2008), another from Pittsburgh and Georgia (Burke et al. 2008), and another from the United Kingdom (Larsson et al. 2008). Samples also varied in terms of selection criteria, with some broadly representative of the community (Hipwell et al. 2008; Larsson et al. 2008), others oversampling conduct-disordered (high externalizing problems) children (Pardini et al. 2008), as well as a poverty sample (Gross et al. 2008) and a clinic-referred sample (Burke et al. 2008). To the extent that bidirectional processes show sensitivity to socio-demographic factors, these study differences are not inconsequential. We turn now to a summary of the major themes and findings of the five articles.
Evidence of Bidirectionality: Forms, Timing, and Conditions
Symmetric and Asymmetric Forms
In thinking about bidirectional influences across development, one may ask whether the relative impact of child behavior on parent behavior and of parent behavior on child behavior is generally comparable in magnitude (in a sense, bidirectional equivalence or symmetrical bidirectionality), whether one partner (or partner’s characteristics) predominates (unequal or asymmetric bidirectionality), or whether, across development, there is a shift in relative degree of influence (non-linear or shifting bidirectionality). Findings generally consistent with both asymmetric and symmetric bidirectionality were evident in these studies in terms of number and/or magnitude of cross-time links. Hipwell et al. (2008) found that girls’ conduct problems and depression were associated somewhat more strongly with subsequent parental depression and warmth than vice versa, though in neither instance were the effects very large. Burke et al. (2008) tested cross-time associations between conduct disorder (CD) and oppositional-defiant disorder (ODD) and five aspects of parenting (timid discipline, low involvement, poor supervision, poor communication, and harsh discipline). After controlling for a variety of structural-demographic variables and the corresponding child (i.e., controlling for CD when examining relations with ODD) and parent (i.e., controlling for other forms of parenting when examining a particular parenting measure) variables, child disruptive behavior problems were more consistently predictive of subsequent parenting than parenting was predictive of subsequent behavior problems.
A somewhat different pattern of results are reported by Pardini et al. (2008). Conduct problems (CP) as rated by parents and teachers were associated with each of several parenting problem scores (assessed in a manner similar to that in Burke et al. 2008) and each problematic parenting score was associated with subsequent CP. Larsson et al. (2008) also found comparable effect sizes for child→parent and parent→child effects in their genetically-informed twin study of parental negativity and child antisocial behavior between ages 4 and 7. Interestingly, analyses showed that, as predicted, parent effects on subsequent child aggression were environmentally mediated, whereas child aggression effects on subsequent parental negativity were genetically mediated (i.e., genetically influenced child aggression elicited parental negativity).
Gross et al. (2008) examined bidirectional links between boys’ externalizing problems and mothers’ depression from ages 5 to 10 and ages 11 to 15. Mothers reported on their own depression at all ages; mothers and alternate caregivers rated child externalizing behavior at the earlier ages and youths self-reported antisocial behavior at the later ages. Separate parallel processing models (correlated growth models) were run for the three sets of externalizing scores and one was significant: Linear growth in maternal depression was positively correlated with mother-reported child externalizing behavior from ages 5 to 10. Cross-lag paths revealed that, across all three informants’ (i.e., mothers’, alternate caregivers’, youths’) scores, mother→child links were significant in three of nine instances and child→mother links were significant in two of nine instances.
There also is some evidence of shifting (non-linear) patterns in the sense that the magnitude of effects varied as a function of age of child. Timing effects were examined in two studies. Importantly, specific predictions were tested. In our own research on bidirectionality in parental monitoring knowledge and adolescent delinquent behavior across grades 9 to 12 (Laird et al. 2003), cross-lag paths were found to be similar in magnitude, but we had advanced no hypotheses regarding age-related changes. In contrast, Gross et al. (2008), in testing for timing (age) effects in the links between maternal depression and child externalizing problems, make two age-related predictions. Considerable prior research has suggested that physical maturation and changes in the child’s social environment heighten child and family stress levels during the transition to adolescence and the transition to formal schooling. Gross et al. (2008) therefore hypothesized that both child and parent effects would be more prevalent during these time periods (i.e., from age 5 to 6 and from age 11 to 12). Their findings generally were consistent with these predictions.
Pardini et al. (2008) also tested for age (timing) effects across six-month intervals from ages 6 to 16, using an accelerated longitudinal design. Systematic linear and quadratic developmental timing effects were probed at representative ages that were two years apart across this larger time period (i.e., ages 7, 9, 11, 13, and 15). Pardini et al. hypothesized that the strength of cross-lag links between monitoring and boys’ conduct problems would increase across childhood and adolescence and that child→parent effects would increase in magnitude during the transition to adolescence as youths assert more autonomy in family decision-making. They found evidence of bidirectional age effects for monitoring and conduct problems and child→parent age effects for low positive reinforcement and timid parenting. Poor parental monitoring was significantly associated with increases in teacher-rated child conduct problems at ages 9, 11, and 13 but not at ages 7 and 15. Teacher-rated conduct problems were in turn associated with increases in poor monitoring at ages 11 and older. Parent-reported conduct problems were associated with increases in low positive reinforcement at ages 9, 11, and 12, but not at ages 7 and 15. The same pattern of effect was found for teacher-reported problems in relation to increases in timid parenting. The lack of effect for/on monitoring at age 7 may be due to the lower need for monitoring at younger ages; the effects for timid parenting may suggest a tendency on the part of parents of youth with conduct problems to be reluctant to confront them about their behavior. The Gross et al. and Pardini et al. model testing approaches are a welcome addition to the research literature on bidirectionality. By articulating and testing timing effects they help to advance an understanding of processes of bidirectionality within a developmental framework.
Specificity and Moderators of Linkages
Only one of the studies directly compared boys and girls (Larsson et al. 2008), and model fit statistics showed that links between parental negativity and childhood antisocial behavior, and the genetic and environmental contribution to the two, were comparable for boys and girls. As noted earlier, boys only were participants in three of the studies, and girls only in one study. Given the dissimilarity in constructs of interest, direct gender comparisons are difficult. Hipwell et al. (2008), in their study of parenting and girls’ adjustment difficulties, found that low warmth and harsh discipline uniquely (additively) predicted both conduct problems and depression across time. This finding is consistent with prior research showing that positive parenting and negative parenting represent separate dimensions of parenting rather than a single continuum and that low levels of positive parenting can be as strongly associated with behavior problems as high levels of negative parenting (e.g., Pettit et al. 1997). Hipwell et al. also found that increases in harsh punishment were predicted only by conduct problems and that increases in low parental warmth were predicted only by girl’s depression. This kind of specificity in predictive links is relatively rare in socialization research, but seems sensible: high levels of anxiety and depression, especially over extended periods of time, may lead parents to withdraw and show less affection; high levels of externalizing problems tend to elicit a harsher reaction, often in the form of physical punishment.
Burke et al. (2008) also tested for specificity of links in their sample of boys. ODD and ADHD were uniquely predicted only by timid discipline; CD was uniquely predicted by poor communication. In a conservative analysis in which other forms of behavior problems and a host of socio-demographic variables were used as controls, only the timid discipline→ODD effect remained significant. Timid discipline, parental involvement, and poor communication were, in turn, uniquely predicted by ODD, and poor supervision was uniquely predicted by CD. Harsh punishment was predicted by CD, but the effect became nonsignificant once the socio-demographic controls were taken into account. Again, the specificity of effects seem reasonable and converges somewhat with Hipwell et al., except that harsh discipline seems to play a lesser role in the Burke et al. study compared to the other parenting behaviors. The harsh discipline measure, however, was a single dichotomous item rather than a multi-item (and presumably more reliable) scale. Because Burke et al. used a clinic sample, in which behavior problems likely are more severe than in non-referred samples, child→parent effects might be expected to be more prominent than parent→child effects.
Clearly there is a need in future research to consider gender dynamics in greater detail by including boys and girls in the same study. At the same time, it is important to study bidirectional processes with fathers and their children as well as mothers and their children. It is possible that there are gendered dyad-specific patterns, e.g., that girls’ depression has stronger impact on mothers’ warmth, or that boys’ conduct problems have a greater impact on fathers’ physical punishment. There have been repeated calls to include fathers in developmental psychopathology research, but only limited progress has been made (Phares et al. 2005). Inclusion of fathers often presents a challenge in practical terms (e.g., difficulties recruiting fathers) and in terms of the population of interest. High-risk populations may contain a higher number of father-absent homes, creating analytic quandaries because of the disproportionate numbers of single-parent and two-parent families. Still, this is a significant issue for the field and one in which such obstacles must be overcome.
Socio-demographic moderators of parent and child effects were examined by Hipwell et al. (race/ethnicity and poverty) and by Pardini et al. (race/ethnicity). One might also construe zygosity as a moderator in the Larsson et al. genetically-informed cross-lag analyses. Evidence of moderation was limited. Hipwell et al. found one significant interaction out of eight tested, between household poverty (defined as receiving vs. not receiving public assistance) and low warmth for CD symptoms: Among those not receiving public assistance a decrease in warmth was associated with an increase in CD. Among those girls living in poverty, there was no increase in CD as a function of warmth. This finding fits with some prior research suggesting that ineffective parenting is less likely to be predictive of externalizing problems among families living in poverty (e.g., Shaw et al. 2000), presumably because the poverty effects swamped the parenting effects.
Pardini et al. (2008) tested for ethnicity as a moderator in the links between parenting and conduct problems (separately for mother-reported CP and teacher-reported CP, for a total of 12 tests) and found one significant interaction, between ethnicity and physical punishment: For African–Americans, physical punishment was associated with increases in teacher-reported CP from age 7 to 11; for European–Americans, physical punishment was associated with increases in CP at age 7 but not at later ages. None of the 12 tests of ethnicity as a moderator of links between CP and subsequent parenting was significant. The overall pattern, then, is one in which ethnic group similarity in patterns of bidirectionality was the rule. The one exception—physical punishment and its relation to CP—is an important one, however. An accumulating body of longitudinal research has tended to show a positive correlation between harsh punishment and teacher-reported conduct problems in European–American youth, but either no significant correlation or a slight negative correlation in African–American youth (Gunnoe and Mariner 1997; Lansford et al. 2004). Various explanations have been offered for this ethnic group difference, including the role of cultural normativeness and the child’s interpretation of, and reaction to, parents’ disciplinary strategies (Lansford et al. 2005). Also, when more extreme forms of discipline, such as physical maltreatment, are considered, significant positive associations with conduct problems are found in both African–American and European–American samples (e.g., Deater-Deckard et al. 1996). The index of physical punishment used by Pardini et al. did not differentiate between abusive and non-abusive forms of discipline, which may have contributed to the observed ethnicity differences.
As noted earlier, and as is apparent in this discussion of findings, the studies differed substantially in terms of sample characteristics and breadth of measurement, making it difficult to draw strong conclusions about specificity and moderators of patterns of relations. But at a more general level, the studies cohere in providing clear evidence of bidirectionality: Child and youth adjustment difficulties contribute to declines in parent–child relationship quality over time just as negative parenting (and parental depression) contributes to increases in child and youth adjustment problems over time.
Challenges, Questions, and Implications
Whither research on bidirectionality, then, in light of the approaches and contributions of the articles in this Special Section? Significant progress has been made, but challenges remain. Here we discuss some of those challenges and the questions still in need of answers. We conclude with a brief discussion of the implications of this research for model development and for intervention science.
Levels of Bidirectionality
The approach used in each of the studies in this Special Section to assess bidirectionality was to measure parenting, or parental depression, via questionnaire, in successive time periods, and to measure child adjustment, via questionnaire or diagnostic interview, in the same time periods. These periods may span several months or several years. One might construe these as macro-level indicators of parenting and child functioning. What can they actually reveal about how patterns of bidirectionality and reciprocity develop? What can these data tell us about spiraling processes of coerciveness in family interactions? As suggested by Gross et al. (2008), we must extrapolate from the more macro findings of bidirectional links between questionnaire-based measures of parenting (or parent depression) and child adjustment to relatively more micro models of mutuality and reciprocity observed during interactional exchanges. Examination of the moment-to-moment interactions between caregivers and their children may reveal in a more fine-grained way the manner in which parenting characteristics and child behavior characteristics become interconnected. It also would be fruitful to link experimental manipulations, such as that conducted by Pelham et al. (1997), with descriptive longitudinal research designs.
Modeling Change across Development
An examination of growth (change) in parenting and youth adjustment, and their interrelation, has been made possible because longitudinal research projects that have insured that key constructs are assessed in identical or near-identical ways. (But we note that recent advances in latent factor growth modeling allow for an examination of growth in parenting even when somewhat different items constitute the relevant measures at different ages; see Pettit et al. 2007.) Measuring constructs in the same way across developmental periods is good for modeling growth and correlated changes in parenting and child behavior, but it necessarily limits what can be said about transformations in forms of parenting (e.g., a move from proximal to more distal forms of parental supervision) and youth adjustment (e.g., the emergence of more serious forms of rule breaking). What is needed are complimentary approaches that track changes in forms of parenting and youth adjustment difficulties as well as patterns of growth in development in the parenting and adjustment characteristics that are similarly-defined indexes of these constructs.
The field has come a long way since cross-lagged panel analysis was the method of choice for examining the relation between early parenting and later child behavior problems (controlling for earlier child behavior problems) and the relation between early behavior problems and later parenting (controlling for earlier parenting). The Laird et al. (2003) study, cited by most of the articles in this Special Section, used latent growth modeling to test for correlated rates of change in monitoring knowledge and adolescent delinquent behavior, and followed up with cross-lagged path models to test for reciprocal associations among measures at specific time points. A similar strategy was used by Gross et al. (2008), who first conducted latent growth curve modeling and then estimated an autoregressive path model with cross-lagged associations between maternal depression and boys’ aggression. Larsson et al. (2008) expanded the two-wave panel analysis in a study of twins varying in zygosity (termed a biometric cross-lagged model) to isolate genetic and environmental factors contributing to parental negative and child antisocial behavior. Three of the studies, with lead authors affiliated with the University of Pittsburgh, used generalized estimating equation (GEE) transitional regression models to test for bidirectionality. The GEE technique appears to have a number of advantages, including accounting for non-independence in repeated measures of the same constructs in the same individuals over time. GEE analyses also allow for a straightforward testing of interactions (e.g., with time, with ethnicity) in child→parent and parent→child models.
Mechanisms of Bidirectionality
Related to the notion of “levels” is the need to focus greater attention on developmental process. The articles that make up this Special Section provide some speculations about the evolution of parenting effects, child effects, and their interplay across development. Patterson’s coercion model is a favored explanation. But there is a need to incorporate models that consider intervening—and alternative—mechanisms of influence. Dunn (1997), in her commentary on a Special Issue of the Journal of Social and Personal Relationships devoted to bidirectionality and reciprocity in parent–child relationships (see Pettit and Lollis 1997) recommends “pitting” alternative process explanations against one another, a suggestion that has great relevance for contemporary studies of bidirectionality. As an example, consider the now-replicated finding that adolescent antisocial behavior and parental monitoring are reciprocally related across the adolescent years, i.e., as youths become more antisocial parents monitor them less closely and as levels of monitoring decline antisocial behavior increases (Burke et al. 2008; Laird et al. 2003; Pardini et al. 2008). Through what processes might these changes in monitoring and antisocial behavior occur? How might the efficacy of alternative process models be contrasted? Laird et al. (2003) lay out a number of possibilities. One explanatory model might be that as youths become more antisocial they tend to gravitate to deviant peer groups who provide training in, and models of, antisocial behavior, including ways of skirting parental supervision and engaging in covert (and not easily detected) forms of rule-breaking. The monitoring-antisocial behavior link also might be explained by changes in the parent–child relationship (communication, affective bond) that both lessen parents’ interest in and willingness to engage in supervision and monitoring, and that renders teens more susceptible to antisocial peer pressure. Additional explanatory mechanisms have been suggested by Crouter et al. (2005) and by Stattin and Kerr (2001), among others. Stewart et al. (2002) proposed and tested a specific mediator (mechanism) that may account for the bidirectional relationship between delinquency and parenting. Structural equation modeling analyses of the longitudinal Iowa Youth and Families Project dataset revealed that legal sanctions (e.g., being picked up by the police, placed in juvenile detention or jail, gone to court) fully mediated the impact of delinquent behavior on parenting and partially mediated the impact of parenting on delinquent behavior. The point here is that there is value in developing, testing, and contrasting mediating models (which likely are not mutually exclusive) to determine which might serve as the more salient process(es) and whether their salience as a linking mechanism varies across type of parenting, type of child outcome, and developmental period.
Clinical and Theoretical Implications
The most obvious implication of this collective set of findings, and one stressed in almost every article, is that effective interventions for youth displaying adjustment difficulties must necessarily take into account the parent’s role and the child’s role within a comprehensive developmental framework. The evidence presented here suggests that parent effects are perhaps not as pervasive or overwhelming as might have been presumed in the past, especially when examined with stringent methods that control for a variety of relevant socio-demographic controls, along with earlier child behavior characteristics. Child effects, on the other hand, are modest in magnitude but fairly widespread. But little is gained from either/or kinds of argument. Interventions that are attuned to specificity in the links between parenting behavior and developmental outcome and that target key normative (transitions to elementary and junior high school, puberty) and non-normative (change in neighborhood, divorce) switch points may prove to have the great success in altering trajectories of problem behavior (Dodge and Pettit 2003).
For the sake of both intervention science and developmental science it is important that we continue to map underlying processes of bidirectionality. Multi-method, multi-informant approaches to construct creation can help allay concerns about method bias and measurement reliability. Linking process-oriented research, whether in the context of experimental manipulations in laboratory settings or in clinical trials, and focusing on different levels of bidirectionality and reciprocity, would help to pinpoint what matters most in sustaining (positive reciprocity) or breaking (negative reciprocity) parent→child and child→parent chains of influence. This must be done within the broader social context of family, neighborhood, community, and culture. As the articles in this Special Section make clear, substantial progress has been made. The promise of continuing gains in understanding is encouraging and exciting.