Translating Models of Antisocial Behavioral Development Into Efficacious Intervention Policy to Prevent Adolescent Violence

Abstract

Adolescent chronic antisocial behavior is costly but concentrated in a relatively small number of individuals. The search for effective preventive interventions draws from empirical findings of three kinds of gene-by-environment interactions: (1) parenting behaviors mute the impact of genes; (2) genes alter the impact of traumatic environmental experiences such as physical abuse and peer social rejection; and (3) individuals and environments influence each other in a dynamic developmental cascade. Thus, environmental interventions that focus on high-risk youth may prove effective. The Fast Track intervention and randomized controlled trial are described. The intervention is a 10-year series of efforts to produce proximal change in parenting, peer relations, social cognition, and academic performance in order to lead to distal prevention of adolescent conduct disorder. Findings indicate that conduct disorder cases can be prevented, but only in the highest risk group of children. Implications for policy are discussed.

INTRODUCTION

The total annual burden of crime to American society now exceeds $1 trillion (Anderson, 1999). These costs include medical costs to victims, costs of prosecution, costs of incarceration and attempts at rehabilitation, and lost earnings, but they cannot begin to include the emotional pain to victims, perpetrators, and bystanders. It has long been known that the majority of crimes are committed by just a small percentage of the population, estimated at 7% by Wolfgang, Figlio, and Sellin (1972) and between 6% and 10% by Howell, Krisberg, and Jones (1995). These individuals have been called career criminals by criminologists (Blumstein & Cohen, 1987), early starters by Moffitt (1993), and antisocial personality disorder/conduct disorder by psychiatrists (American Psychiatric Association, 1994). Prospective studies indicate that many of these life-persistent antisocial individuals begin their aggressive behaviors in early childhood, peak during adolescence, and very gradually desist as they move into later adulthood (Moffitt, 1993). Cohen (1998) has estimated the cost of losing a single high-risk youth to a life of crime at $1.25 to $2 million.

The pernicious persistence of (Cohen, Rust, Steen, & Ridd, 2004). antisocial behavior in these individuals during their adolescence has led some observers to be very pessimistic about the prospects for prevention (Dilulio, 1996; Herrnstein & Murray, 1994). In the 1990s, they incited American policy makers to ostracize this group through longer prison terms, zero-tolerance rules in schools, increased adjudication of juveniles in adult courts, and labels such as “super-predators” (Dilulio, 1995). Americans have not completely given up on these adolescents, however, as “willingness-to-pay” studies in economics indicate that taxpayers would be willing to pay large sums to reduce or prevent crime, if the intervention policy can be proven effective.

The goal of the program of research described in this article is to move toward effective intervention policy to prevent the problem of life-persistent antisocial behavior, by conducting developmental studies to identify pathways to chronic violence, applying this knowledge to the design of innovative interventions to prevent antisocial behavior, and translating this knowledge to effective public policy.

GENES AND ENVIRONMENTS IN THE DEVELOPMENT OF ANTISOCIAL BEHAVIOR

The pessimism about intervention with early starters is partly based on behavior genetic studies that indicate that chronic antisocial behavior may be highly heritable (Rhee & Waldman, 2002). However, heritability does not imply inevitability, and a flurry of recent studies is revealing that environments interact with genes in complex ways that render both factors crucial in leading to antisocial development (Rutter, 2006). The authors have identified three types of gene-by-environment (G × E) interplay in the prospective Child Development Project (CDP) of 585 children followed from preschool through adulthood.

The first type of G × E interaction is that the environment may titrate gene expression, that is, sufficient structure in the environment may mitigate the degree to which genes are expressed in antisocial behavior. In the CDP, Bates, Pettit, Dodge, and Ridge (1998) found that the genetic proxy variable of infant temperament, measured by mother ratings of the infant’s “persistence in forbidden acts,” was predictive of teacher-rated externalizing behavior problems at ages 7–11 only under environmental circumstances of low parental restrictiveness (r = .22, P < .05). When parents implemented high restrictiveness in rules and structure, infant temperament did not predict externalizing problems at all (r = .01, n.s.). A more direct test of this form of G × E interaction came later when DNA in participants’ saliva was genotyped. Previous studies had identified a main-effect association between problems in disinhibition such as conduct disorder and alcoholism and single nucleotide polymorphisms (SNPs) in the larger haplotype block that extends downstream from intron 3 in the gene that codes for the major inhibitory neurotransmitter in the mammalian brain, gamma-aminobutyric acid (GABA) A receptor, alpha 2, known as GABRA2 (Dick et al., 2006). Using CDP data, Dick et al. (2009) found that the degree of parental monitoring and supervision in early adolescence dramatically buffered the impact of GABRA2 on externalizing disorder in later adolescence. Specifically, under conditions of low parental monitoring, the percentage of individuals reaching criteria for persistent externalizing disorder was 6.9%, 18.5%, and 27.7%, respectively, for individuals with 0, 1, or 2 risk alleles in GABRA2; however, these figures were 10.3%, 12.0%, and 16.3%, respectively, for individuals with 0, 1, or 2 risk alleles under conditions of high parental monitoring. Thus, parental behavior, especially in restrictiveness, monitoring, and supervision, can minimize the impact that genes have on a child’s antisocial behavioral development, opening a possibility for preventive intervention.

The second type of G × E interaction occurs when a child’s genes alter the impact of an environmental cause of antisocial behavior. It is known that harsh and rejecting environments often lead to growth in aggressive behavior in children (Dodge, Coie, & Lynam, 2006). One such environment is the experience of chronic social rejection by peers. Children who have been peer-rejected are likely to increase their aggressive behavior, even controlling for aggression that had preceded peer rejection (Dodge et al., 2003). However, a “calm” temperament during infancy as rated by the mother protects a child from this growth in aggression during elementary school (Dodge et al., 2003). That is, those children who had been rated as having a below-median (easy) temperament were not likely to grow in aggressive behavior after even three consecutive years of peer rejection (mean Teacher Rating Form Aggression scores of 1, 3, 3, and 1, for 0, 1, 2, and 3 years of peer rejection, respectively), whereas those children who had been rated as having an above-median (difficult) temperament were highly sensitive to the adverse effects of peer rejection (mean Teacher Rating Form Aggression scores of 6, 12, 17, and 23, for 0, 1, 2, and 3 years of peer rejection, respectively). Peer social rejection is a toxic factor in antisocial development, but calm temperament can protect a child from its effects.

The experience of physical maltreatment in early life has been identified as a major causal factor (perhaps the strongest factor) in the development of conduct disorder, even when genetic factors are controlled (Jaffee et al., 2005). However, heritable factors can protect a child from the harmful effects of maltreatment. In the E-Risk study of twins in Great Britain, among the group of children at highest heritable risk for conduct disorder (based on one’s twin’s diagnosis), the experience of physical maltreatment dramatically increased a child’s probability of receiving a conduct disorder diagnosis from 46.1% to 69.6%, whereas among the group of children at lowest heritable risk, the experience of maltreatment had only a small effect, increasing the probability of conduct disorder from 1.9% to 3.5% (Jaffee et al., 2005).

Caspi et al. (2002) were the first to find that a polymorphism in the gene that encodes the monoamine oxidase A (MAOA) enzyme exacerbates the impact of physical maltreatment on adolescent outcomes such as conduct disorder. MAOA degrades dopamine, norepinephrine, and serotonin in response to major stressors. Disruption of this process through a genetic polymorphism makes a child especially vulnerable to traumatic effects of an acute threat such as maltreatment. This interaction effect has been replicated in the CDP, where it was found that parent-to-child physical maltreatment in the first five years of life interacted with the MAOA polymorphism to predict teacher-, parent-, and self-reported levels of aggression and delinquency at ages 6–22 (Edwards et al., in press). The implications of this finding for preventive intervention are that harsh parenting behavior is a strong risk factor that should be targeted for change in parent training and that some children may be especially susceptible to environmental impact (and thus highly adversely affected by negative environments but also very positively responsive to favorable environments).

These two kinds of interaction effects are important, but they do not attend to the facts that genes and environments are often empirically correlated and the child and environment may reciprocally influence each other across development. A “difficult” young child may elicit harsh parenting, which may lead to early conduct problems at school entry. In turn, conduct problems may lead to peer social rejection in elementary school, which exacerbates conduct problems and creates new parenting difficulties. Parenting difficulties such as problems in monitoring and supervision may give rise to new problems in peer relations, such as association with deviant peers, which leads to even worse deviant behavior in adolescence. These dynamic cascades have been empirically identified in several studies. In the CDP, Dodge et al. (2009) identified six child and environmental factors that cascaded across development from early childhood through adolescence to predict illicit substance use. Using prospective data from the Fast Track study, Dodge, Greenberg, Malone, and the Conduct Problems Prevention Research Group (2008) identified seven child and environmental factors that cascaded to predict violent behavior in adolescence.

The importance of these findings for preventive intervention is that even though environmental factors are partly the product of the child, they also exert independent incremental effects on antisocial development and can be targeted for early intervention. However, multiple environmental factors must be targeted, and intervention must be implemented at multiple time points across development.

IDENTIFYING AND TESTING DEVELOPMENTAL MODELS OF PREVENTIVE INTERVENTION

Empirically-based models of the development of antisocial behavior have been used to design preventive interventions for children who are at risk for antisocial outcomes. Several different intervention programs have been tested through randomized controlled trials and found to be efficacious in disrupting the antisocial developmental process at least temporarily (reviewed by Dodge et al., 2006). One such intervention is the Fast Track Program designed by the Conduct Problems Prevention Research Group (CPPRG, 1992). They hypothesized that positive long-term effects would require intervention with multiple environmental factors over multiple points in development.

Models of the cross-time effects of prevention as children grow up have not been described in much detail in the literature, but the Fast Track program developers identified four possible models (CPPRG, in press), depicted in Figure 1. The first model posits that a onetime intervention early in life will have an enduring impact on antisocial behavior across the lifespan, as in an inoculation. Although the dynamic cascade findings reviewed above suggest that an “inoculation model” will not hold because of emergent environmental risk factors across development, this model is still plausible. A second model, consistent with a dynamic cascade, posits that intervention effects at an early point in development could occur but will be small and can be supplemented by intervention effects at multiple later points in development. These intervention effects cumulate in a “cumulative dose” model, “if intervention is sustained.” A third model posits that intervention could temporarily keep a child from deviant behavior while the intervention itself remains in place but will have no preventive effect after the intervention is stopped because the child factors will resume to have an impact. This model, also not consistent with a dynamic cascade, is nonetheless plausible as a “temporary scaffold model.” The final “null effects model” is that intervention will have no effect at all. These four models offer contrasting hypotheses about the effects of an intervention across a child’s development and were tested in a randomized controlled trial of Fast Track.

FIGURE 1.

Models of development in intervention effects. (CPPRG, in press).

THE FAST TRACK STUDY

Sample

The Fast Track project was initiated in 1990 to evaluate the feasibility and efficacy of a long-term, comprehensive program designed to prevent the development of conduct disorders in adolescence (CPPRG, 1992). Fast Track was a 10-year, multicomponent intervention for children at high risk for developing conduct problems. The program was implemented in a randomized controlled trial embedded within a longitudinal study of normative and high-risk youth in four demographically diverse sites (Durham, NC; Seattle, WA; rural central Pennsylvania; and Nashville, TN) across 54 schools serving communities identified as high risk based on neighborhood crime and poverty levels.

Participants were recruited through a multistage screening procedure using both teacher and parent ratings of disruptive behavior (Lochman & CPPRG, 1995). All 9,594 kindergarten children in three successive cohorts (1991, 1992, and 1993) at participating schools were screened based on teacher ratings of disruptive and aggressive classroom behaviors. Parents of children who scored in the top 40% (n = 3,600) were contacted, and 91% of the parents contacted (n = 3,267) provided ratings of the child’s behavior problems at home. Teacher and parent ratings were standardized and combined to create a severity-of-risk screen score.

Children were selected to participate in the study based on this screen score, moving from the highest score downward until the desired sample sizes were reached within each site, cohort, and condition. Ninety-five percent of the selected sample scored in the top 20% on both the teacher and parent screening measures. In all, 92% of eligible families whose child enrolled in first grade at a participating school (n = 891) agreed to take part in the study. Schools within each site were assigned to matched sets according to demographic characteristics (e.g., size, ethnic composition, and poverty level) and then sets were randomly assigned to intervention or control conditions. The intervention group consisted of 445 children in 191 first-grade classrooms, and the control group consisted of 446 children in 210 first-grade classrooms.

Participants were 6.5 years old on average (SD = .48) at the time of identification. The sample was ethnically diverse across sites (51% African American, 47% European American, and 2% of other ethnicity (e.g., Pacific Islander, Hispanic), and 69% of the participants were boys. The level of socioeconomic disadvantage in the sample was high: 58% of the families were headed by a single parent, 29% of parents were high school dropouts, and 35% of the families were in the lowest socioeconomic class as scored by Hollingshead (1975).

Intervention Components

The intervention was multifaceted and occurred in both home and school settings, utilizing multiple agents who naturally and significantly influence the child’s socialization (e.g., parents, teachers, and peers). Intervention components targeted several risk factors for the development of persistent antisocial behavior across the domains of parenting, peer relations, social information processing, and academics, including: poor parental behavior management, poor parental monitoring and supervision, poor peer relations, deficient child social-cognitive and emotional coping skills, weak academic skills, disruptive and rejecting classroom environments, and poor home-school relations. The intervention was implemented across a 10-year time period, from 1st through 10th grade, to address specific risk factors at the point in development when they were most operative.

Fast Track included both universal and targeted approaches to intervention. The universal component of the intervention involved delivery by classroom teachers of a newly adapted version of the PATHS curriculum (Kusche & Greenberg, 1993) to promote social competence and problem-solving, emotion regulation, and self-control skills. Lessons in these areas were given two to three times per week during first through fifth grade. Targeted interventions were implemented through academic tutoring, social skills training (Bierman & CPPRG, 1996), and parent groups (McMahon, Slough, & CPPRG, 1996) with home visiting. Program content was designed to change from year to year to correspond with developmental changes in the needs of the children and families in the study. During the first year of the intervention, all participants received a standard level of tutoring, peer pairing and coaching, and home visiting services. In subsequent years, the amount and type of services provided to each child and family were adjusted based on an assessment of risk and level of functioning.

During first and second grades, child participants received academic tutoring two to three times a week for 30 min to improve reading skills. They were also paired with popular peers and received weekly tutoring sessions in enhancing friendships. For the first 5 years of the program, parents and children participated in regular, 2-hr group enrichment sessions, meeting weekly in first grade, bi-weekly in second grade, and monthly in third through fifth grade. During these sessions, children met in “friendship groups” of five or six children led by educational coordinators for social skills training for 60–90 min, while parents met in a group to discuss positive parenting strategies with family coordinators. In the last 30 min, parents and children were reunited to practice parenting skills with staff support and engage in positive cooperative activities. Several group sessions and workshops were included in fifth through eighth grade that involved coping with developmentally relevant challenges and transitions. Individualized plans for intervention services were created for each family in 7th through tenth grade. Home visits were conducted regularly to help parents generalize the positive parenting skills learned through the group sessions and provide individualized assistance. These home visits also were used to promote general parental functioning by enhancing problem-solving skills that could generalize to other life stressors faced by parents, such as marital conflict and housing issues.

Implementation and Participation

The Fast Track intervention spanned 10 years, beginning with cohorts in 1991, 1992, and 1993. Multiple mediators and outcome measures were assessed throughout the course of the intervention and during a follow-up period that continues today.

Participation in the group sessions was high in the first year of the intervention, with 96% of parents and 98% of children attending one or more sessions and 79% of parents and 90% of children attending more than 50% of all sessions. Participation remained relatively strong but declined somewhat in subsequent years, so that by fourth grade, 73% of parents and 82% of children attended one or more group sessions, and 68% of parents and 80% of children attended more than 50% of all sessions. In later phases, the intervention included increasingly more individualized activities, but participation remained at relatively constant levels. For example, in seventh and eighth grades, 63% of participants attended at least one scheduled family meeting, 80% of participants attended at least one parent meeting, 78% of participants attended at least one individual meeting with youth, and 73% of participants attended at least one group meeting with youth.

Effects on Proximal Processes

Findings have been published in multiple reports (e.g., CPPRG, 1999, 2002a, 2004, 2007 e.g., CPPRG, in press). The Fast Track researchers hypothesized that the multicomponent intervention would have a direct impact on parenting behaviors, competency with peers, social-cognitive skills, and academic skills, and that improvements in these proximal processes would in turn lead to reductions in problem behaviors. Consistent with this hypothesis, the intervention had a significant main-effect impact on several measures of parenting behaviors across multiple reporting sources in early elementary school. At the end of first grade, parents in the intervention group were less likely to endorse physical punishment solutions to parenting challenges and reported greater improvements in their parenting skills, displayed more warmth and positive involvement, exhibited more appropriate and consistent discipline practices, and were more involved at school than parents in the control group (CPPRG, 1999). Some of these positive effects on parenting continued later in elementary school. At the end of third grade, the intervention-group parents reported less use of physical punishment and greater improvements in their parenting skills than did control-group parents (CPPRG, 2002a).

Similarly, the Fast Track intervention had a significantly positive impact on peer relations during elementary school. At the end of first grade, children in the intervention group displayed more positive peer interactions and received higher social preference scores than children in the control group (CPPRG, 1999). These effects on peer relations also occurred in later years of elementary school. At the end of fourth grade, intervention-group children scored higher than control-group children on sociometric ratings of social preference and were less likely to associate with deviant peers (CPPRG, 2002b).

The intervention also significantly affected children’s social-cognitive patterns in the elementary school years. At the end of first grade, children in the intervention group demonstrated greater social-cognitive skills such as emotion recognition, emotion coping, and social problem-solving than children in the control group, and less aggressive retaliation (CPPRG, 1999). The Fast Track children continued to show differences from control children in social-cognitive skills at the end of third grade, with marginally less hostile interpersonal attributions and greater social problem-solving skills (CPPRG, 2002a). Similar differences in social cognition were found in the fourth and fifth grades (CPPRG, 2004).

The children in the intervention group also demonstrated differences from control children in academic skills and performance. At the end of first grade, the intervention group had significantly higher language arts grades than the control group (CPPRG, 1999). The Fast Track children also scored higher on the Spache word attack reading test, though the mean difference between groups on the word attack test was statistically significant only at one site (Durham, NC) and significant at a trend level at another site (Seattle, WA). Some improvements in the academic domain continued into later elementary school. At the end of fourth grade, teacher ratings of improvements in academic competence were significantly higher for the intervention group (CPPRG, 2004).

Effects on Conduct Disorder

As discussed above, random assignment to the intervention had positive effects on the domains of parenting, peer relations, social cognition, and academics in the elementary school years. In later years of the intervention, the impact of Fast Track on the primary outcome of interest, prevention of conduct disorder, was assessed. Psychiatric diagnoses of conduct disorder were made at the end of third grade, sixth grade, ninth grade, and twelfth grade. The assessment in 12th grade took place 2 years after the intervention ceased, allowing the researchers to examine whether intervention effects (if any) were maintained beyond the delivery of services.

There was no main effect of the intervention on psychiatric diagnoses of conduct disorder at the end of third, sixth, or ninth grade, but there was a significant interaction between intervention and severity of initial risk at each time point (CPPRG, 2007). To interpret these interaction effects, intervention and control group means were contrasted within each of two levels of severity of initial risk. The highest risk group consisted of the top 3% of the normative population on the screening measure, and the moderate-risk group consisted of those scoring lower than the top 3%. The 3% cut-off was selected because it represents a common standard for defining psychiatric caseness in externalizing problems in youth.

Investigation of these interaction effects revealed that assignment to intervention had a significant impact on conduct disorder diagnoses in the highest risk group. Among the highest risk youth, assignment to intervention was responsible for preventing almost 50% of conduct disorder cases in third grade (prevalence of .11 vs. .20 in the control group), more than 50% of conduct disorder cases in sixth grade (prevalence of .10 vs. .23 in the control group), and more than 75% of conduct disorder cases in ninth grade (prevalence of .05 vs. .21 in the control group). In contrast, the intervention did not prevent conduct disorder cases among youth who were initially at only moderate levels of risk.

The Fast Track team next evaluated the impact of the intervention on the lifetime prevalence of conduct disorder through the end of 12th grade. Consistent with the analyses conducted at earlier time points, there was no significant main effect of intervention, but the interaction effect between the intervention and severity of initial risk was significant (CPPRG, in press). Specifically, within the highest risk group, random assignment to intervention was associated with a significantly lower lifetime prevalence of conduct disorder than assignment to control, with the intervention preventing more than 50% of lifetime conduct disorder cases (.20 compared to .41 in the control group). However, the intervention did not reduce the lifetime prevalence of conduct disorder among initially moderate-risk youth.

Discrete time survival analyses were conducted to estimate the time of onset of conduct disorder and examine the cumulative rates of conduct disorder from 3rd through 12th grade. Discrete time survival analysis takes into account the within-subject variable of time of measurement in a single analysis by estimating the probability of receiving a diagnosis given no previous diagnosis. Survival analyses also were used to determine whether the magnitude of intervention effects on conduct disorder varied by year. The time-by-intervention interaction effect tests whether the intervention effect differs at different time points. No effects of the interaction between intervention and time were significant, and no three-way interaction effects among intervention, risk, and time were significant, indicating that the intervention-by-risk effects described above held across each succeeding year (CPPRG, in press). That is, among the highest risk youth, new positive effects of the intervention on conduct disorder occurred each year. These effects are illustrated in Figure 2, which depicts the cumulative rates of psychiatric diagnoses of conduct disorder for the highest risk group of youth in the intervention versus the control group at each year of assessment. As shown, the difference between the intervention and control groups increased over each time period between 3rd and 9th grades and then leveled off between 9th and 12th grades, when very few new conduct disorder cases occurred in either group. This pattern of results for the highest risk group most closely resembles the “cumulative dose model” of intervention effects depicted in Figure 1.

FIGURE 2.

Cumulative rates of psychiatric conduct disorder diagnoses as a function of intervention among the highest risk group (CPPRG, in press).

Economic Analysis

The findings from the Fast Track program demonstrate that a long-term, comprehensive intervention can prevent the development of conduct disorder among those at highest risk, a group that has proven significantly difficult to treat. However, when evaluating an intervention’s impact for the purpose of wider dissemination and implementation, it is important to consider not just the program’s efficacy but also the relative costs and benefits involved in the intervention. Accordingly, Foster, Jones, and CPPRG (2006) evaluated the cost-effectiveness of the Fast Track program. The costs of delivering the 10-year intervention were estimated using principles of economic evaluation from the perspective of a potential payer, such as a state mental health agency. The estimated average cost per child of the intervention across sites and cohorts was $58,283. Incremental cost-effectiveness ratios (ICERs) reflecting the ratio of program costs to intervention effects were calculated for the full Fast Track sample as well as for subsamples defined by initial risk level.

In evaluating a program’s cost-effectiveness, the primary question to be addressed is whether the intervention’s ICER is lower than a policymaker’s or society’s willingness to pay for improvements in the outcome of interest. Based on estimates that the costs to society of a career criminal exceed $1 million (Cohen, 2005), Foster et al. (2006) calculated society’s willingness to pay for preventing a case of conduct disorder to be $1 million for the purposes of these analyses. The estimated ICER (cost of the program per case of conduct disorder prevented) for the entire intervention sample was $3.48 million, with an estimated cost-effectiveness probability of 1% but an effectiveness probability of 82%, indicating that the program is not cost-effective for the entire sample but is likely to be effective at preventing conduct disorder.

However, in light of the fact that previous analyses had found that intervention effects on conduct disorder diagnoses varied by initial risk level, Foster et al. (2006) also examined whether cost-effectiveness similarly differed across risk groups, defining higher risk youth as those scoring above the 90th percentile on the initial screening measure and lower risk youth as those scoring below the 90th percentile. The lower risk group had a negative ICER and an effectiveness probability of 6%, indicating that the program is not cost-effective or likely to be effective at preventing conduct disorder for this group. In contrast, the higher risk group had an estimated ICER of $752,103 and an effectiveness probability of 99%, indicating that the program is both cost-effective (less than the $1 million threshold) and highly likely to be effective for the higher risk youth. Note that these analyses were based on the top 10% of the sample at initial assessment, which is a larger group than the top 3% used in evaluating intervention effects in conduct disorder in previous analyses.

CONCLUSION

The findings across these studies indicate that individual characteristics (genetically based or otherwise) interact with environmental factors to produce problem behavior outcomes during adolescence. Descriptive studies demonstrate that genetic factors exacerbate or mitigate the impact of traumatic environmental experiences such as early physical abuse and chronic peer social rejection. Likewise, parenting behaviors mute the impact of genetic dispositions. One problem with descriptive studies is that environmental variables are not randomly assigned. Thus, even with statistical controls in place, it remains plausible that individual characteristics elicit certain environments and render their impact nil. Random-assignment intervention studies overcome this problem by manipulating environmental experiences. The findings of the Fast Track intervention experiment indicate that, again, individual characteristics interact with the environmental intervention in leading to conduct disorder outcomes during adolescence. As with the descriptive studies, the environment had the strongest impact on the group of children who were at highest individual-level risk. It remains for future studies to examine whether the “high-risk” variable in the Fast Track study can be related to a genetic factor. These studies have important implications for science, practice, and public policy.

Scientific Implications

The G × E interaction reconciles the age-old debate about nature versus nurture by highlighting the essential importance of both factors in adolescent development. Furthermore, the genetic and environmental variables that have been identified in interaction effects appear not to be randomly matched but instead provide insight into the processes involved in the development of adolescent problem behavior. The genetic factors that interact with life experiences appear to code for proteins that are implicated in the brain’s response to threat or stress, thus highlighting the role of traumatic stress in the development of conduct disorder. The broader scientific implication is that theories of adolescent development must take account of how the growing youth transacts with the environment. We have barely scratched the surface of understanding transactive development, however, and thus the path of future scientific inquiry is wide open.

Prevention Implications

The repeated finding that environments alter the impact of genes on problem behavior outcomes gives rise to the hope for prevention. Preventive intervention might target: (1) traumatic and toxic environments (such as physical abuse, chronic peer rejection, and deviant peer groups) that have insidious effects on youth behavior and (2) children who are at high risk for problem outcomes. Successful interventions are not likely to be simple or inexpensive, but they might well prove to be cost-beneficial.

Public Policy Implications

The findings reported here have at least two implications for public policy. First, they stand in sharp contrast to the public framing of high-risk youth as “super-predators” who are resistant to environmental intervention and who must be imprisoned or quarantined away from the rest of society (Dodge, 2008). Indeed, the highest risk youth appear to be the most susceptible to environmental influence, be it positive or negative. More apt metaphors are needed to describe these youths and the task of prevention. The second implication is that preventive intervention that is based in developmental theory and empirical findings may be a wise and cost-beneficial investment as public policy.