Risk-Taking and Self-Regulation

Abstract

The purpose of this study was to investigate the juvenile justice intervention literature for the statistical consideration given the impact that age, as a marker of neuropsychosocial development, may have on delinquency outcomes. A systematic review of 117 studies published between 1996 and 2009 was conducted to assess the methods by which curvilinear and moderating effects of age were included in the analysis of delinquency outcomes. Ninety-one percent of studies may have underestimated intervention effects through the misspecification of the effect of age on delinquency outcomes. Of the 10 studies that did test for curvilinear and interaction effects, 80% had findings consistent with neuropsychosocial theories of age on delinquency. To account for age effects on delinquency, the regular use of multiple age groups in analysis may increase both the precision with which intervention effects are measured and the identification of specific age groups with whom individual interventions are most effective.

Keywords

delinquency intervention systematic review brain development risk-taking self regulation

The most recent data published by the Office of Juvenile Justice and Delinquency Prevention (OJJDP) estimated that the number of court cases heard in the juvenile justice system in 2008 was approximately 1.7 million, with almost 600,000 of those cases ultimately adjudicated delinquent (Knoll & Sickmund, 2010; Puzzanchera, Adams, & Sickmund, 2010). Among those youth adjudicated delinquent, approximately 56% recidivate at least once—a rate that increases with each additional referral (Synder & Sickmund, 2006).

This high recidivism rate along with results of meta-analyses suggests that juvenile justice interventions, particularly those in juvenile justice custody facilities, have met with limited success (Lipsey, 1994, 1999; Lipsey, Wilson, & Cothern, 2000). The most comprehensive meta-analysis to date of facility-based interventions, for instance, found effect sizes to range from −0.11 for shock incarceration to 0.11 for intensive aftercare (Lipsey, 1999). The purpose of the systematic review chronicled here was to explore possible methodological explanations for these findings.

Although there are no doubt multiple reasons for the small effect sizes found in meta-analysis, one possible cause may originate in the methods by which intervention effects are analyzed, rather than in the interventions themselves. For instance, previous delinquency research has shown a consistent relationship between age and delinquency, whereby delinquency increases sharply from about age 10 through age 16 or 17, after which it decreases sharply into late adolescence (Blumstein & Cohen, 1987; Gottfredson & Hirschi, 1990; Hirschi & Gottfredson, 1983; Moffitt, 1993; Scott & Steinberg, 2008; Steinberg, 2008; Steinberg & Scott, 2003). More recent neuroscience research has identified two predictable neuropsychosocial processes that may underlie this relationship. The first process is a neurologically determined spike in reward-seeking behavior that often expresses itself in an increase in sensation-seeking and risk-taking in early adolescence that peaks in middle adolescence and diminishes rapidly through late adolescence (Bava & Tapert, 2010; Doremus-Fitzwater, Varlinskaya, & Spear, 2010; Ernst, Nelson et al., 2005; Steinberg, 2009). The second process is the slow neural development of self-regulation, a process that does not appear to be fully matured until the early 20s (Casey, Tottenham, Liston, & Durston, 2005; Scott & Steinberg, 2008; Steinberg, 2008, 2009; Steinberg & Scott, 2003).

Assuming that risk-taking and self-regulation both play a part in delinquent behavior, the relationship between age (as a proxy for these processes) and delinquency should be an important one to consider when designing and evaluating interventions that target delinquency given that the neurological processes underlying this relationship would predict differential rates of intervention effectiveness depending on the age of the youth. For instance, youth who are in early to midadolescence during an intervention follow-up period would be predicted to experience greater developmental barriers to using intervention skills meant to reduce delinquency due to higher levels of normative risk-taking and lower levels of self-regulation compared to youth who are in late adolescence during the same follow-up period.

It is currently unclear whether attempts to measure the effectiveness of juvenile justice interventions are taking this relationship between age and delinquency into account. The most common use of age in the juvenile justice intervention literature appears to be the comparison of mean age across conditions as a measure of equivalence between treatment and control groups. This practice may be overlooking important variation in neuropsychosocial development that could impact comparisons made both within studies, in determining the effects of a single intervention, and across studies in meta-analysis. As an example, Lipsey (1999) found two seemingly contradictory effects of age on recidivism in his meta-analysis of juvenile justice interventions: (1) using hierarchical multiple regression analysis, age effects based on mean age were overall unrelated to effect size, and (2) intervention effectiveness (i.e., effect size) was increased when “the average age of the juveniles in the sample was greater than the median for all programs (fifteen and a half years)” (p. 634).

The first finding, in which age was unrelated to effect size, may be a result of using mean age in analysis without consideration of age range when mean age masks developmental variation—for instance, when a mean age of 16 represents an age range of 12 to 22 in one study and a range of 15 to 17 in another. The variation in neuropsychosocial development in each of the two age ranges would likely predict different effect sizes due to the differences in the variation within groups. Thus, age, which reflects general levels of neuropsychosocial development (among other things), would appear to have no consistent predictability on delinquency outcomes when the same mean age in different studies represents different overall levels of risk-taking and self-regulation, both of which greatly influence the degree to which intervention effects translate to changes in delinquency outcomes.

The second finding, that intervention effectiveness was greater in studies with higher average ages, is consistent with neuropsychosocial theories of age and delinquency such that higher levels of psychosocial maturity would be predicted in samples with an average age at pretest that is above 15½, particularly in studies with longer follow-up periods. This higher level of psychosocial maturity would predictably (1) influence the participant’s ability to use intervention skills and refrain from those behaviors that lead to rearrest/readjudication and (2) be accompanied by a normative, developmentally determined reduction in risk-taking behavior and increase in self-regulation that are separate from intervention effects.

To further explore the degree to which age effects are addressed in the analysis of delinquency outcomes, a systematic review of the juvenile justice intervention literature was conducted. One hundred and seventeen content-relevant studies were identified, and the methods by which age effects were addressed were explored.

Background

Previous research has consistently found adolescence to be a time of increased sensation-seeking, risk-taking, and antisocial behavior (Blumstein & Cohen, 1987; Gottfredson & Hirschi, 1990; Hirschi & Gottfredson, 1983; Moffitt, 1993; Scott & Steinberg, 2008; Steinberg, 2008; Steinberg & Scott, 2003). In her seminal 1993 article on the subject, Terrie Moffitt provided a developmental taxonomy of adolescent delinquency that described two divergent pathways: a developmentally normative curvilinear pathway in which these behaviors both increase and decrease sharply within the adolescent period (adolescence limited) and a second, atypical pathway (life-course persistent), which she described as arising from an interaction among neuropsychological deficits, intergenerational vulnerabilities, and family, school, and neighborhood disadvantage. In the life-course persistent pathway, antisocial/delinquent behavior begins in childhood and continues throughout the lifespan. More recent social science and developmental neuroscience research has provided mounting evidence for both the causal mechanisms and environmental correlates of these two distinct pathways.

Although the etiology of adolescent delinquency for the life-course-persistent pathway is not necessarily informative to the current discussion regarding the impact of normative developmental processes on interventions targeting delinquency, a brief description of this atypical pathway nevertheless helps to provide a more complete picture of the factors affecting adolescent delinquency and antisocial behavior. Therefore, the following section begins with a brief description of the research findings regarding atypical development followed by a more detailed account of the normative neuropsychosocial processes at work in the adolescent brain that impact delinquent behavior.

A Typical Development and Delinquency

Moffitt’s (1993) description of the individual- and family-level factors affecting youth whose delinquency and antisocial behavior persists beyond adolescence has been supported by longitudinal, cross-sectional, and developmental research. Collectively, this research has focused on childhood aggression/behavior problems as the most reliable predictor of life-course persistence, with associated factors that include (1) neurological deficits, (2) individual and family risk factors, and (3) higher scores on measures of callous-unemotional (CU) and/or psychopathic personality (PP) traits.

Brain imaging studies of neurological activity in children and youth with varying rates of antisocial or problem behavior coupled with CU or PP traits has found differences in those brain areas associated with the recognition of fearful and distressed faces. Specifically, children and youth with problem behavior plus higher scores of CU or PP traits showed less activity within and between those brain areas used for facial recognition of fear/distress (i.e., the amygdala) and those areas that use that information to make behavioral choices (i.e., the ventromedial prefrontal cortex; Jones, Laurens, Herba, Barker, & Viding, 2009; Marsh et al., 2011; Marsh et al., 2008). This reduced activity suggests impairment in the ability to process social cues of fear and distress in others, social cues that “guide healthy children away from antisocial behavior” (Marsh et al., 2008, p. 717). Additionally, youth with higher scores on PP measures plus a disruptive behavior disorder were found to have reduced activity in brain areas associated with reinforcement learning (orbitofrontal cortex and caudate), which indicates impairment in the ability to learn from the emotional responses of others what behaviors will and will not be rewarded (Finger et al., 2011).

Longitudinal, twin, and cross-sectional studies have explored the relationships among family- and individual-level factors and aggressive, antisocial, and delinquent behavior in childhood, adolescence, and early adulthood. Higher scores on measures of CU traits in youth demonstrating antisocial behavior have been associated with a more stable pattern of antisocial behavior (Viding, Jones, Frick, Moffitt, & Plomin, 2008) as well as higher levels of anxiety, shyness, cognitive problems, loneliness, difficulty controlling emotions, and lower levels of stress management (McLoughlin, Rucklidge, Grace, & McLean, 2010) and higher rates of parental psychiatric illness (Viding et al., 2008) and parental criminal convictions (McLoughlin et al., 2010). Other factors associated with early onset antisocial behavior and conduct problems in childhood include deficits in executive and cognitive functioning, family instability, family conflict, and dysfunctional parenting (Dandreaux & Frick, 2009).

Overall, these findings support Moffitt’s description of a constellation of risk factors in the lives of individuals who go on to continue their antisocial behavior into adulthood. The reciprocal nature of these factors may be, as Moffitt suggests, a primary feature of the life-course persistent pathway, given that experienced separately, nurturing environments can be corrective in the case of neurological deficits and, similarly, neurological health can be protective in dysfunctional family environments (Moffitt, 1993).

Normative Risk-Taking in Adolescence

One of the emerging approaches to explaining the normative spike in adolescent risk-taking, with delinquent/antisocial behavior as a specific example, is based on recent advances in developmental neuroscience. Brain imaging studies have identified two main processes whose co-occurrence in the healthy adolescent brain directly impacts delinquent behavior: one that is associated with a sharp rise in risk-taking behavior at puberty and a second that is involved with a much more slowly developing ability to self-regulate behavior that continues to mature into the early 20s (Casey et al., 2005; Fareri, Martin, & Delgado, 2008; Luna, Padmanabhan, & O’Hearn, 2009; Steinberg, 2009; Steinberg et al., 2008). Changes in behavior in response to these two neurological processes have been termed psychosocial development (Steinberg, 2009), with psychosocial maturation reflecting a reduction in risk-taking and an increase in the ability to self-regulate emotions and behavior. For the purposes of this review, the neurological processes underlying psychosocial development will be referred to as neuropsychosocial development or neuropsychosocial processes.

Neuropsychosocial Development: Risk Taking

The first neuropsychosocial process implicated in heightened risk-taking involves sudden and dramatic changes in activity in the limbic system that coincides with puberty. These changes include an increase in dopamine activity in the nucleus accumbens, an area in the brain involved in motivating the approach response towards rewards, with increased dopamine activity making rewards seem more salient and rewarding (Bava & Tapert, 2010; Fareri et al., 2008). Accompanying this is reduced activity in the amygdala, which is involved in the avoidance response to danger or threat, with reduced activity making negative outcomes seem less aversive (Doremus-Fitzwater et al., 2010; Ernst, Nelson et al., 2005). These changes have the effect of both increasing the salience of wanting and reward seeking in adolescents while at the same time decreasing the impact or salience of threat (of negative consequences) on behavioral choices (Bava & Tapert, 2010; Doremus-Fitzwater et al., 2010; Ernst, Nelson et al., 2005).

It is this combination of increased reward salience and decreased threat salience that is hypothesized to be a primary factor in the increase in risk-taking (Bava & Tapert, 2010; Fareri et al., 2008; Scott & Steinberg, 2008; Steinberg, 2008, 2009) via its impact on sensation-seeking behavior, which is “the tendency to seek novel, varied or highly stimulating experiences and the willingness to take risks in order to attain them” (Steinberg et al., 2008, p. 1765). This neurologically determined and normative increase in sensation seeking during adolescence has been theorized to be adaptive in that it allows for the incorporation of experience into neural development while at the same time motivating adolescents to become more autonomous and to develop relationships outside of the family—both major tasks of the adolescent period (Somerville, Jones, & Casey, 2010; Steinberg, 2008). Thus, the risk-taking behavior that leads to increases in unwanted pregnancy, drug and alcohol addiction, and delinquency during adolescence is nested within the biologically predisposed behavior of sensation seeking.

Neuropsychosocial Development: Self-Regulation

Accompanying this sudden increase in sensation seeking is a second, slower neural process that increases self-regulation. This process includes an increase in white matter (myelination) and decrease in gray matter (synaptic pruning) in the prefrontal cortex and on neural pathways between the prefrontal cortex and the limbic system. As more axons are encased, the speed with which information is processed and communicated within and between cortical areas increases as well. Similarly, as unused connections between neurons are pruned, the diffuse activation of areas not needed for a given task is also decreased, thus increasing the focal activation of those areas that are needed. This process of increased myelination and synaptic pruning in the prefrontal cortex speeds the cellular communication among those areas of the brain that coordinate social and emotional responses and regulate behavior (Casey et al., 2005; Scott & Steinberg, 2008; Steinberg, 2008, 2009; Steinberg & Scott 2003).

Three of the brain areas that have been most directly implicated in self-regulation are the medial prefrontal cortex (MPFC), the ventrolateral prefrontal cortex (VLPFC), and the anterior cingulate cortex (ACC). These three areas (along with several others) reach maturity sometime in the late teens to early 20s (Casey et al., 2005; Fareri et al., 2008; Luna et al., 2009) and have been linked to the control of motivated behavior (Ernst, Pine, & Hardin, 2005) and the regulation of the brain’s approach (nucleus accumbens) and avoidant (amygdala) systems (Fareri et al., 2008).

The increase in efficiency resulting from maturation of these brain areas translates to improvements in abilities associated with psychosocial functioning. For instance, as efficiency in cellular communication is increased within the prefrontal cortex, so should the ability to inhibit responses, plan ahead, weigh risks and rewards, and simultaneously consider multiple sources of information (Steinberg, 2008). Similarly, as efficiency in cellular communication increases between the limbic system and the prefrontal cortex, so should the ability to self-regulate the wants of the limbic system with the cognitive considerations of the prefrontal cortex (Ernst, Pine et al., 2005; Kambam & Thompson, 2009; Steinberg, 2008, 2009).

Timing of Neuropsychosocial Development

These two neuropsychosocial processes are initiated and completed on different timetables, with the changes in reward orientation and sensation seeking in the limbic system occurring with the onset of puberty in early adolescence (ages 9 to 13 in girls, 10 to 14 in boys; Graber, Nichols, & Brooks-Gunn, 2009) and peaking in middle adolescence (ages 15 to 17; Cauffman & Steinberg, 2000; Steinberg, 2009), while self-regulation and the ability to effectively use the executive processes of the prefrontal cortex to regulate emotion and control impulses increases gradually with the ongoing growth of myelination into the early 20s (Casey et al., 2005; Fareri et al., 2008; Luna et al., 2009; Steinberg, 2009; Steinberg et al., 2008). This protracted development of self-regulation is particularly important to risk-taking behavior given the sudden increased activity in the nucleus accumbens and decreased activity in the amygdala that occurs in early adolescence prior to the full maturation of self-regulatory systems. Steinberg (2005, p. 70) summarized these events on the following timeline: puberty heightens emotional arousability, sensation seeking, and reward orientation in early adolescence, followed by a period of heightened vulnerability to risk-taking and problems in regulation of affect and behavior in middle adolescence, and finally the maturation of the frontal lobes facilitates regulatory competence in late adolescence.

Age as a Proxy for Normative Neuropsychosocial Development

The two neurological processes that underlie normative neuropsychosocial development parallel one of the most enduring relationships found in the social sciences: the age–crime curve (Scott & Steinberg, 2008). For decades, criminological, sociological, and psychological research on delinquency has consistently found the same pattern among both juvenile-justice-involved and general population youth across multiple Western nations, whereby delinquent and antisocial behavior increases sharply from age 10 until approximately age 16/17, followed by an equally sharp drop that levels off in late adolescence and early adulthood (Blumstein & Cohen, 1987; Gottfredson & Hirschi, 1990; Hirschi & Gottfredson, 1983; Moffitt, 1993; Scott & Steinberg, 2008; Steinberg, 2009; Steinberg & Scott, 2003). This trajectory clearly mirrors that found in neurological research on the two processes involved in neuropsychosocial development: the increase in risk-taking at puberty, the peaking, and subsequent decrease in risk-taking in mid-late adolescence accompanied by the slow increase in self-regulation into late adolescence/early adulthood.

Taken together, the breadth of research replicating this trajectory along with brain imaging research that has found consistent developmental patterns that correlate with age (for instance, Bava & Tapert, 2010; Doremus-Fitzwater et al., 2010; Dumontheil, Houlton, Christoff, & Blakemore, 2010; Ernst, Nelson et al., 2005; Forbes & Dahl, 2010; Luna et al., 2009; Stroud et al., 2009) suggests that age is an appropriate indication of or proxy for level of neuropsychosocial development. The use of age as a proxy for different levels of development within adolescence is, in fact, already an accepted practice in the education and mental health literatures (for instance, Anderson, Sabatelli, & Kosutic, 2007; Dyk & Adams, 1987; Froh et al., 2010; Gardner & Steinberg, 2005; McLean, Breen, & Fournier, 2010; Smetana, Campione-Barr, & Metzger, 2006).

Age Effects and Delinquency Outcomes

As reflected in both the age–crime curve and research on neuropsychosocial development, the relationship between age and delinquency is of a curvilinear nature, with the delinquent behaviors associated with risk-taking increasing at puberty, peeking in midadolescence, and dropping through late adolescence into early adulthood. This relationship, which has been found across national boundaries and historical eras, continues to hold true for the current juvenile justice population in the United States. Displaying the available data from The Census of Juveniles in Residential Placement 1997–2010 graphically, it is clear that this relationship exists in the current age pattern of youth in custody (see Figure 1). Specifically, the percentage of youth in custody (i.e., those committed to a facility, those detained awaiting adjudication or placement, or those placed in a youth facility as part of a diversion program) in the present era also follows a curvilinear pattern in which the percentage of youth in each age group increases linearly through midadolescence, then drops in the later-adolescent group (Sickmund, Sladky, Kang, & Puzzanchera, 2011).

Figure 1:

The Percentage of Youth in Each Age Category Across Offenses Based on Data Available From The Census of Juveniles in Residential Placement 1997–2010

Looking at offense types separately, this pattern can also be seen across all data collection years for 14 of the 17 listed offenses, with some small (robbery) to moderate (criminal homicide) variation. The only notable exception to this pattern is sexual assault, where the percentage of youth in each age range continues to increase past the midadolescence range into the late-adolescent age group in 6 of the 7 data collection years.¹ Beyond this, however, the curvilinear age pattern is clearly present across years and offense types. Figure 2 shows the age profile for the most recent year (2010) for each of the 17 offense types.

Figure 2:

Percentage of Youth in Each Age Category Separated by Offense for the Year 2010 Based on Data Available From The Census of Juveniles in Residential Placement 1997–2010

This curvilinear relationship between age and delinquency has implications for the methods by which age is included as a covariate in analyses of intervention effects on delinquency outcomes, particularly in samples with a wide range of ages. For instance, in samples with age ranges that include both those in early adolescence who are experiencing a normative increase in delinquency and those in later adolescence who are experiencing a normative decrease in delinquency, the relationship between the age and delinquency variables will not be linear.

Additionally, the age range of a sample has important implications for analysis given that age can be considered a moderator of the relationship between intervention effects and delinquency. Given that neurological processes involved in self-regulation develop on a predictable timeline, increasing through late adolescence into early adulthood, the ability of youth to delay gratification, consider the consequences of their actions, and/or control their emotional responses to situations (all of which are targets of juvenile justice interventions) and therefore refrain from delinquent behavior should be limited by the degree to which neurological changes (i.e., myelination and synaptic pruning) and accompanying self-regulation have matured. For instance, supervisory interventions such as probation assume that the threat of violation will influence the decisions made by youth in the moment when the opportunity for delinquent behavior presents itself. The developmental neuroscience evidence discussed here indicates that the ability to inhibit responses based on weighing risks and rewards and simultaneously considering multiple sources of information is not fully developed until the early 20s, suggesting that a 14-, 15-, or 16-year-old would not have the same neurodevelopmental capability that an 18-, 19-, or 20-year-old would have given identical situations. Similarly, interventions that attempt to reduce delinquency by increasing emotional control (e.g., controlling aggression or reducing the use of drugs or alcohol) may also be limited in early-mid adolescence by the protracted maturation of efficient coordination of the limbic system and the prefrontal cortex. Specifically, as coordination between these two areas matures, the ability to self-regulate the wants of the limbic system (e.g., to get high, to react physically when angry) with the cognitive considerations of the prefrontal cortex (e.g., the consequences of getting high or assaulting someone) increases as well. Thus, the age or age group of a participant may predict an increase or decrease in the effectiveness of interventions that rely on these self-regulatory abilities.

Taken together, this evidence supports the proposition that the evaluation of intervention effects on delinquency outcomes should take into account the stage of neuropsychosocial development of those for whom the intervention is designed. Doing so may make such evaluations more precise in separating out intervention effects from the normative increase in risk-taking associated with changes in neural activity in early and midadolescence and/or the normative decline in delinquency associated with maturation in self-regulatory areas of the brain in late adolescence. To explore whether these effects of age, as a proxy for neuropsychosocial development, have been addressed in the analyses of juvenile justice interventions, a systematic review of the juvenile justice intervention literature was conducted.

Method

To build a pool of content-relevant studies, electronic databases (n = 32) were searched for studies published from the beginning of January 1996 through the end of December 2009, in which programs or interventions for pre-adjudicated or adjudicated juveniles were assessed. Three basic search terms were applied to the electronic searches. The first included population terms delimiting the demographic focus on youth and adolescents, the second included law enforcement terms specifying the range of legal institutions and infractions of interest to the study, and the third included program intervention terms specifying the range of program content of interest to the study. Articles identified through this process were then manually screened using a predetermined protocol to assess whether they fit the purpose of the larger systematic review. Four inclusion criteria were applied to the articles: (1) studies that were located in the United States, (2) studies reporting program/intervention effectiveness, (3) studies reporting quantitative findings, and (4) studies using a comparison group. Overall, studies were included if they employed an experimental (randomized controlled trial) or quasi-experimental design with a control group that was not exposed to the intervention but was measured and compared on the same outcome variables as the treatment group (see Evans-Chase & Zhou, in press, for a more comprehensive description of review methodology).

Using the above criteria, a total of 141 studies were identified as content relevant, with 117 of those studies passing inclusion criteria for the current review. Individual studies recorded in the original review as multiple studies due to the inclusion of multiple samples, treatments, or control groups were collapsed and considered as a single study when age was treated similarly across samples, treatments, or control groups. Studies that had no delinquency outcome measure were also dropped from the review. Additional data were then collected from the 117 articles on the following age variables: (1) mean age of the sample, (2) age range of the sample, (3) the use (yes or no) of an analytical method that accounts for the curvilinear relationship between age and delinquency and/or the moderating effect of age on the relationship between intervention effects and delinquency (sufficiency variable), and (4) the method by which age effects were addressed in analysis (methods variable). All 117 articles were double coded on both the sufficiency and methods variables, with interrater reliability of 91% for the methods variable and 99% for the sufficiency variable. Discrepant results were reconciled through discussion with all three authors.

Addressing Age Effects in Analysis

The primary methods used to correctly test for curvilinear relationships and moderating effects in analysis were identified and used to operationally define and code the sufficiency variable for each study. Studies were coded as having sufficiently tested for the curvilinear relationship between age and delinquency if a quadratic equation in polynomial regression (Y = a + b₁ * TREATMENT + b₂ * AGE + b₃ * AGE² + ϵ) was used in analysis. This method not only correctly models curvilinear relationships between age and delinquency outcomes but also helps to avoid or reduce bias on the coefficient of treatment, which might be produced by misspecification of the age variable (Wooldridge, 2006).

Although using a quadratic term of age helps to avoid misspecification based on assumptions of linearity, the coefficient of treatment in such analyses is still estimated based on the assumption that the effect of treatment does not vary across age groups. As discussed previously, however, age, as a proxy for neuropsychosocial development, may also moderate the relationship between treatment and delinquency outcomes. One method by which moderation effects can be estimated is the inclusion of a product term that represents the interaction effect (Aguinis, 2004; Aiken & West, 1991; Baron & Kenny, 1986). Therefore, articles were coded as having sufficiently tested for the potential moderating effect of age on the relationship between intervention effects and delinquency outcomes in analysis that used either (1) an interaction term in regression (Y = a + b₁ * TREATMENT + b₂ * AGE + b₃ * AGE * TREATMENT + ϵ) or (2) an interaction term in ANOVA (Treatment × Age).

Finally, articles that used multiple group analysis were also coded as having sufficiently tested for moderating effects, given that it is also designed to test whether values of model parameters vary across groups (Kline, 2005). Additionally, given that multiple group analysis controls for any effects of age, both linear and curvilinear, by eliminating or minimizing within-group variation, these articles were also coded as having appropriately accounted for the curvilinear relationship between age and delinquency outcomes.

What About Randomization?

The influence of age (as a proxy for neuropsychosocial development) on delinquency outcomes may not be ameliorated with random assignment to groups. Assuring the equality of age and thus psychosocial range across groups using randomization does not necessarily protect against a biased (suppressed) effect size when the overall sample has a wide age range and group equivalence is measured by mean age. Using mean age with a wide age range may hide a wide range of psychosocial capabilities that could differentially influence delinquency outcomes separate from intervention effects. For instance, if a sample’s age range is 12 to 19, some participants would be in the low risk-taking/low self-regulation stage, some in the high risk-taking/low self-regulation stage and some in the diminishing risk-taking/increasing self-regulation stage, and thus, there would be greater variance in delinquency outcomes separate from intervention effects. This increased variance would reduce the measured effectiveness of the intervention given the use of pooled variance in determining effect size (e.g., Cohen’s d = M_t – M_c / SD_pooled; Cohen, Cohen, West, & Aiken, 2003). Therefore, the use of random assignment was not included in this review as an indicator of having sufficiently addressed age effects on delinquency outcomes.

Results

Participant ages ranged widely across studies, with some studies reporting participants as young as age 7 and some as old as age 22. Mean age, which ranged from 13 to 21, did not necessarily reflect the same range of ages across studies, nor did it give a complete picture of the sample. For instance, a mean age of 15 represented a range of ages in one sample that spanned from age 10 to age 17, while in a second sample a mean age of 15 represented a range that spanned from age 12 to age 22. Figure 3 depicts the age ranges and means found in the reviewed studies that reported both means and ranges, without duplication (i.e., the number of studies represented in Figure 3 do not add up to 117 due to missing information and duplicate means/ranges).

Figure 3:

Sample Age Ranges Centered Around the Reported Sample Mean

Overall, the majority of studies (91%) addressed neither the curvilinear nor the moderating effects of age on delinquency outcomes. The most common method of including age in analysis, employed by 50 studies (43%), was the use of age as a variable to establish or check for equivalence between treatment and control groups. The second most common method, employed by 30 studies (25%), was the use of age as a covariate in linear regression analyses such as logistic regression, multiple regression, or Cox regression, without the use of quadratic terms, interaction terms, or multiple age groups. Finally, 20 of the 117 studies (17%) made no mention of age anywhere in the analysis.

Studies that Addressed Age Effects

As depicted in Table 1, 10 out of the 117 studies (9%) used at least one of the operationally defined methods of addressing possible age effects in analysis. Of those 10 studies, 7 found evidence of age effects influencing baseline or posttest measures of delinquency, 6 of which were consistent with neuropsychosocial theories of age and delinquency.

Table 1:

Studies That Addressed Age Effects in Analysis

Author	Effect	Method	Mean Age	Age Range	Age Differences	Outcome	Follow-Up
Butts, Buck, & Coggeshall (2002)	Both	Multiple age groups	14	7 to 17	No	Recidivism	6 months
Cammack-Dawson (1998)	Both	Multiple age groups	15	11 to 18	Yes^a	Recidivism	1 year
Chernoff & Watson (2000)	Both	Multiple age groups	14T, 16C	10 to 17	Yes^a	Recidivism	3 years
Hanlon et al. (2002)	Mod^b	Interaction term	13	9 to 17	Yes^a	Substance abuse self-report delinquent recidivism	1 year
Henggeler et al. (2002)	Mod^b	Interaction term	15	No info	No^d	Substance use recidivism	4 years
Pealer (2004)	Both	Multiple age groups	16	12 to 20	Yes^a	Recidivism	3 years
Rodriguez (2007)	Curv^c	Quadratic Equation	14	No info	Yes^a	Recidivism	2 years
Schaeffer (2000)	Mod^b	Interaction Term	14	9 to 17	No^a	Juvenile + adult recidivism	10, 13 years
Shelden (1997)	Both	Multiple age groups	No info	No info	Yes^a	Recidivism	No info
Wells, Minor, Angel, & Stearman (2006)	Both	Multiple age groups	17	14 to 19	Yes	Recidivism	1 year

Note. a. Findings are consistent with neuropsychosocial theories of age and delinquency.

Moderation of age on relationship between intervention effects and delinquency.

Curvilinear relationship between age and delinquency.

Not enough information to determine whether findings were consistent with theory.

Evidence of Age Effects

Of the studies that found evidence for an age effect on delinquency, one study tested for a curvilinear relationship among variables using a quadratic equation (age and age²) in logistic regression (Rodriguez, 2007). The author found that “the significant positive effect of age and negative effect of age² indicates that the likelihood of recidivating increases with age at lower ages and then tapers off at higher ages” (Rodriguez, 2007, p. 367). This inverted U-shaped relationship between age and delinquency is consistent with neuropsychosocial theories that predict a rise in risk-taking through middle adolescence, followed by a drop in delinquency in late adolescence.

Another study that found evidence for age effects on delinquency used statistical methods that test for an interaction effect of age on the relationship between intervention effects and delinquency outcomes. The author found that older participants had higher rates of delinquency at baseline and that younger participants had greater changes in outcomes after treatment (Hanlon, Bateman, Simon, O’Grady, & Carswel, 2002). Although self-regulation, the age effect that predicts an interaction between age and treatment effects, would predict reductions in delinquency with older participants, it is important to consider both the age range of the sample and the follow-up period when interpreting the outcomes of this study. The age range for this study was age 9 to 17, with 2% of participants age 17 (n = 7 of 428), two-thirds of participants falling between ages 12 to 15, and a follow-up at 1 year. Thus, the 14- to 15-year-old participants would have made up the bulk of the older participants, all of whom would be close to or at the apex of risk-taking during that year, coupled with still low levels of self-regulation. Alternatively, the younger group (9 to 13), which would also be predicted to have low levels of self-regulation, would additionally be predicted to have comparatively lower levels of risk-taking behavior during this period. Neuropsychosocial theory would therefore predict that intervention effects would have less to compete with in attempting to reduce delinquent behavior in the younger age group as compared to the risk-taking effects at full force in the older age group. Additionally, because 98% of the sample would be predicted to be on the same side of the risk-taking curve, making the relationship between age and delinquency linear, these findings would not be biased by the nonlinear relationship described between age and delinquency. Thus, the findings of this study are consistent with neuropsychosocial theories of a moderating effect of age on the relationship between intervention effects and delinquency outcomes.

Four studies used multiple group analysis, thus correctly testing for or modeling both the curvilinear and moderating effects of age on delinquency. Three studies provided enough evidence to say with confidence that the findings were consistent with neuropsychosocial theory (Cammack-Dawson, 1998; Chernoff & Watson, 2000; Pealer, 2004), while the remaining study provided only enough information to suggest consistency (Shelden, 1997).

Of the studies that did provide clear evidence, Pealer (2004) found that the younger age group (ages 13 to 14 at baseline) had the highest likelihood of incarceration during the 36-month follow-up period, compared to the 16 to 17 and 18 to 19 age groups. This is consistent with neuropsychosocial theory such that during the 3-year follow up, the 13- to 14-year-olds would age to 16 to 17, just at or past the apex of risk-taking behavior, whereas the older groups would have spent almost the entire follow-up period in the diminishing risk/increasing self-regulation period of late adolescence/early adulthood. Similarly, at 1-year follow-up, Cammack-Dawson (1998) found that the 11 to 14 age group (who would be 12 to 15 at follow-up) had crime rates that were double that of the 17 to 18 age group (who would be 18 to 19 at follow-up), separate from intervention effects. The third study, by Chernoff and Watson (2000), reported that at the end of a 36-month follow-up period, older participants (who were age 17 to 20 at follow-up) had lower recidivism as compared to younger participants (who were age 13 to 16 at follow-up) in both the treatment and control groups. Additionally, they found that the intervention effects were greater in the older group, with significantly lowered recidivism in the treatment compared to the control group, than in the younger group, where no significant difference between groups was found.

Finally, Shelden (1997) reported higher recidivism for youth who were age 14 and younger at baseline as compared to those who where age 15 and older but provided no information about mean age, age range, or length of follow-up, which makes interpretation of the findings difficult. If follow-up was 12 months or less, and the mean age was 15 with a small age range, then these findings would not be consistent with neuropsychosocial theory given that risk-taking (on average) is predicted to peak at age 16. However, if follow-up was longer or the age range was wide (making, for example, 12- to 14-year olds 14 to 16 at follow-up and 15- to 17-year-olds 17 to 19), as suggested by information provided in the article (i.e., that many of the youth turned 18 during the follow-up period), then these findings could be interpreted as being consistent with neuropsychosocial theory.

No Evidence of Age Effects

Of the two studies that found no differences in delinquency outcomes based on age, and that provided enough information to consider the context of those findings, one of the studies had an outcome that was nevertheless consistent with neuropsychosocial theory. In this study, the protracted follow-up period, which averaged 10 years, placed almost all of the participants in the same lowered risk-taking/maturing self-regulation period of older adolescence and early adulthood at follow up (Schaeffer, 2000). Because of the homogeneity of the sample with regard to stage of neuropsychosocial development, differences based on age effects would be predicted to not appear in this study.

This is in contrast to samples such as those found in the remaining eight studies, which would be predicted to show age effects on delinquency outcomes due to the range of ages that included youth on both sides of the inverted-U risk curve (i.e., the rising risk period of early adolescence and the dropping risk period of older adolescence) who would also range from low to high levels of self-regulation. Given that the findings of no age effects in the study with an unusually long follow-up period is also consistent with neuropsychosocial theories of adolescence, it appears that of the 9 studies for which there is enough information to make a determination, 78% tested for age effects had findings consistent with neuropsychosocial theories of age effects on delinquency.

Discussion

Juvenile justice system involvement provides jurisdictions with the opportunity to intervene in the lives of youth involved in delinquent and criminal behavior. Although there are certainly societal and community-level factors that interact with developmental and individual-level factors to increase the likelihood of delinquency, taking advantage of opportunities to address individual behavior with effective juvenile justice programming is one operative approach available to reducing these unwanted behaviors until those societal- and community-level factors can also be addressed.

To that end, it would seem vital to consider the impact that developmental processes have on the ability of youth to implement intervention skills aimed at reducing delinquent behavior. The neuropsychosocial theories and juvenile justice intervention studies discussed here support the need for such consideration. Specifically, brain imaging and social science research have consistently found evidence supporting two effects of age that are important to the evaluation of interventions designed to reduce delinquency: a curvilinear relationship between age and delinquency and a moderating effect of age on the relationship between intervention effects and delinquency outcomes.

Currently, only a small percentage (9%) of intervention studies conducted with juvenile-justice-involved youth have taken such age effects into account when evaluating delinquency outcomes. Of those studies that included analytical methods that correctly tested for age effects, 80% found evidence that supports the existence of these effects on delinquency outcomes as measured by recidivism rates (contact with the juvenile justice system, new juvenile court petitions, new referrals, new incarceration), substance abuse, and self-reported delinquency. These effects were also seen in both pretest and posttest levels of delinquency.

Implications for Research

Based on this evidence, there are several points in the research process at which a more informed use of age could help to increase the ability of researchers to identify interventions that are effective with juvenile-justice-involved youth.

First, attention to age range versus mean age could provide the needed information regarding the variability in risk-taking and self-regulation associated with differing levels of neuropsychosocial development, a variability that is missed with a focus on mean age. Consideration of variability in neuropsychosocial stage of maturation could be an indicator of the degree to which interventions can increase self-regulation and/or override the motivation to take risks, particularly in the peer-group settings in which most delinquent behaviors occur (Steinberg, 2008). What this means in practice is that interventions focused on increasing self-regulatory skills may work very well with 18- to 21-year-olds compared to 14- to 16-year-olds given that the ability to self-regulate would predictably be higher in older youth, while at the same time motivation to seek rewarding experiences (the basis of risk-taking) would be on the decline. However, if the outcomes of all age groups are analyzed together, the benefit gained by the older group may be lost in the averaging of delinquency outcome scores with those of the younger groups, whose changes in delinquent behavior would predictably be smaller.

Second, addressing possible differences in neuropsychosocial development by using multiple age groups in analysis would allow researchers to (1) compare delinquency outcomes across treatment/control groups within each age group, which would control for the curvilinear impact of developmentally normative risk-taking that increases in early adolescence and diminishes in late adolescence and (2) compare intervention effects across age groups, which would indicate whether intervention effects on delinquency outcomes are conditional based on age. Adopting this type of analysis may provide a more precise measure of intervention effects and better information about which interventions work for which age group(s) of adolescents.

Implications for Practice

Attention to the neuropsychosocial processes described here can also inform the design of interventions. For instance, the increase in reward-saliency and peer centrality in early and midadolescence suggests the application of immediate rewards and the opportunity for peer approval in individual-level interventions, both of which may increase youth engagement, and/or the implementation of structural barriers to risk-taking opportunities at a community or family level, which may be a more effective method of reducing risk-taking when self-regulation is low (Steinberg, 2009). Another approach would be the provision of structured opportunities to take risks in new and exciting activities under the supervision of adults. Approaching risk-taking in this way may be an opportunity for adults to engage and connect with youth through the acceptance of their desire to seek out new experiences and the provision of supervised opportunities to do so. If youth involved in the juvenile justice system can be introduced to activities that satisfy the desire to take risks, assuming that they can continue to participate in the activities once they are out of custody, then perhaps youth in early and midadolescence will be more likely to forgo some of the risk-taking activities that expose them to repeated contact with the juvenile justice system.

Additionally, there is ample evidence in the literature to indicate that the maturation of connections in and between those areas of the prefrontal cortex involved in self-regulation (the medial prefrontal cortex, the VLPFC, and the ACC) occurs on a predictable developmental timeline, while at the same time being responsive to experience, particularly during the adolescent period (Casey, Getz, & Galvan, 2008, p. 67; Cauffman & Watson, 2005). Both this developmentally determined trajectory and sensitivity to experience can also inform intervention design. For instance, the results of brain imaging studies suggest that there are interventions that activate those parts of the brain that are involved in self-regulation, which may encourage the development of more efficient neural connections during the adolescent period, thus increasing the ability of youth to self-regulate emotions. Studies of mindfulness meditation with adults have found increased activity in those areas of the prefrontal cortex that are MPFC still developing in the adolescent brain and involved in self-regulation (Creswell, Way, Eisenberger, & Lieberman, 2007; Davidson et al., 2003; Holzel et al., 2007). These findings suggest that mindfulness meditation could conceivably enhance self-regulation in adolescents through the experiential aspect of self-regulatory maturation (i.e., increased activity resulting in increased neural connectivity), while at the same time being limited by age group due to developmentally determined trajectories of neural growth.

Limitations

One caveat to the discussion of age effects on delinquency is that all delinquency outcomes may not be equal. There are moderators of the relationship between age and increased delinquency, one of which is opportunity—the greater the opportunity to participate in risk-taking through delinquent or criminal behavior, the greater the likelihood of that behavior (Steinberg, 2008). One example is the tendency for substance abuse to increase with age past the age 16 to 17 dropoff described for delinquency in general. This has been explained as a likely artifact of increased opportunity to obtain drugs and alcohol as youth age into their late teens (Steinberg, 2008), an opportunity not as ubiquitous in early adolescence. However, the age profiles for individual offense types discussed here suggests that this particular example, which is true for the general population of youth, is not the case for juvenile-justice-involved youth—at least not for those that are in custody. In this population of youth, drug offenses have followed the same trajectory as that described for delinquency in general. This is not surprising given that juvenile-justice-involved youth report higher levels of drug and alcohol use at younger ages as compared to youth in the general population (Sedlak & McPherson, 2010).

There is one type of offense that does not fit the more normative age/crime curve for juvenile-justice-involved youth: sexual assault. The age profiles of youth in custody suggest that sexual assault may be qualitatively different than other offenses given the pattern across years whereby the percentage of youth in custody for sexual assault consistently increases past midadolescence into the late adolescent age group. Future research may identify youth who commit sexual assault offenses as a subgroup that, similar to life-course-persistent offenders, have a more atypical developmental/environmental causal process at work than that found for other offenses.

Another caveat is the possible limitation of using age as a proxy for level of neuropsychosocial development. Although the use of age and age ranges to represent differing stages of development is supported by brain imaging studies in the developmental neuroscience literature (for instance, Bava & Tapert, 2010; Doremus-Fitzwater et al., 2010; Dumontheil et al., 2010; Ernst, Nelson et al., 2005; Forbes & Dahl, 2010; Luna et al., 2009; Stroud et al., 2009), research findings regarding the correlation between puberty, the more precise indicator of that part of neuropsychosocial development involving risk-taking, and age has been mixed. For instance, although Gunnar, Wewerka, Frenn, Long, and Griggs (2009) found puberty and age to be highly correlated when distinguishing among adolescents between the ages of 11 and 15, Quevedo, Benning, Gunnar, and Dahl (2009) found no correlation between stage of puberty and age among 13-year-olds (i.e., the number of months since the 13th birthday was not predictive of early, mid-, or late puberty). Thus, age may not necessarily provide precise information regarding stages of puberty among youth in early adolescence. However, this is not necessarily problematic under the current argument whereby age effects are discussed across age spans, such as ages 10 to 14 (when risk-taking begins to increase), ages 15 to 17 (when risk-taking peaks), and 18 to 22 (when risk-taking declines sharply and self-regulation is nearing maturity).

Conclusion

Youth in the juvenile justice system are just that: youth. They are boys and girls who are developing within the system into men and women. The normative neurodevelopmental processes at work in their brains can have profound implications for the ability of juvenile justice interventions to impact behavior. Unfortunately, of the 117 intervention studies reviewed here, 91% may have underestimated intervention effects through the misspecification of these developmental effects on delinquency outcomes. The results of this systematic review indicate that this lack of attention does not reflect the absence of this relationship (between age, as a proxy for neuropsychosocial development, and delinquency) in the juvenile justice population. The age profiles of youth in custody and the differences found for pre- and posttest scores of a variety of delinquency measures in studies that did test for age effects are both clear indicators that these processes do influence delinquent behavior in juvenile-justice-involved youth.

Given the effects of neuropsychosocial development discussed here, attention to age should be an important part of designing and evaluating interventions for juvenile-justice-involved youth. One practical way of doing this is to adopt a basic approach to intervention design and evaluation that distinguishes between at least two groups of youth: those whose risk-taking is still on a developmentally normative rise and those whose risk-taking is on the decline at the same time that self-regulation is maturing. In evaluating intervention effects, separating youth into two age groups for analysis may reduce the variance of neuropsychosocial development within study groups and address the curvilinear relationship between age and delinquency. This should provide a more precise, unbiased intervention effect size and thus more accurate information regarding the age groups for whom the intervention is most effective. In designing interventions, recognizing the limited ability to self-regulate emotions in early/midadolescence would inform an approach to intervention design that is different for youth in early and midadolescence than for those in late adolescence in whom self-regulatory areas of the brain are more mature. Interventions that are focused on providing structural supports that limit risk-taking opportunities and/or provide safe opportunities to take risks may be a more successful approach to intervening in the lives of early and midadolescent youth. Alternatively, the increase in self-regulation in late adolescence suggests a focus for interventions that encourage and move youth toward the identification and accomplishment of long-term goals (Steinberg, 2009) with less need for structural barriers to risk-taking opportunities.

Footnotes

Authors’ Note

Michelle Evans-Chase, School of Social Policy & Practice, University of Pennsylvania; Minseop Kim, School of Social Policy & Practice, University of Pennsylvania; Huiquan Zhou, School of Social Policy & Practice, University of Pennsylvania. Huiquan Zhou is now at Department of Social Work, The Chinese University of Hong Kong, Shatin, N.T. Hong Kong.

This research was supported in part by grants from the Provost Office of the University of Pennsylvania. The authors would also like to acknowledge financial support from the Association of Private Correctional & Treatment Organizations (APCTO) to complete this study.

Notes

Michelle Evans-Chase is a PhD candidate in the School of Social Policy & Practice at the University of Pennsylvania and holds an MA in psychology from San Francisco State University. Her research focuses on the design and evaluation of interventions for adolescents in and out of the juvenile justice system informed by adolescent neural development and the influence of social policy and other macrolevel factors on individual-level behavior.

Minseop Kim is a PhD candidate in the School of Social Policy & Practice at the University of Pennsylvania and holds an MA in social welfare from Seoul National University in Korea. His research area includes child/youth development and social policy. He is also interested in advanced quantitative methods, including meta-analysis.

Huiquan Zhou received her PhD in social welfare from the University of Pennsylvania in May 2011. She holds an MSW–M.Ed. joint degree from Boston University and also an MA in biology from Boston University. Her research interests are social policy and social development, particularly service provision to the vulnerable and marginalized population.

References

Aguinis

(2004). Regression analysis for categorical moderators. New York: Guilford.

Aiken

West

(1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage.

Anderson

Sabatelli

Kosutic

(2007). Families, urban neighborhood youth centers, and peers as contexts for development. Family Relations, 56, 346-357.

Baron

Kenny

(1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182.

Bava

Tapert

(2010). Adolescent brain development and risk for alcohol and other drug problems. Neuropsychology Review, 20, 398-413.

Blumstein

Cohen

(1987). Characterizing criminal careers. Science, 28, 985-991.

Butts

Buck

Coggeshall

(2002). The impact of teen court on young offenders (Report of the Urban Institute Justice Policy Center, Washington DC). Retrieved from http://www.urban.org/uploadedpdf/410457.pdf

Cammack-Dawson

(1998). The effects of the Green Brigade program on horticultural knowledge, attitudes, and behavioral changes of juvenile offenders (Doctoral dissertation). Retrieved from Proquest Dissertations and Theses Fulltext. (9830892)

Casey

Getz

Galvan

(2008). The adolescent brain. Developmental Review, 28, 62-77.

10.

Casey

Tottenham

Liston

Durston

(2005). Imaging the developing brain: What have we learned about cognitive development? Trends in Cognitive Sciences, 9, 104-110.

11.

Cauffman

Steinberg

(2000). (Im)maturity of judgment in adolescence: Why adolescents may be less culpable than adults. Behavioral Sciences & the Law, 18, 741-760.

12.

Chernoff

Watson

(2000). An investigation of Philadelphia’s Youth Aid Panel: Community-based diversion program for first-time youthful offenders (report of Public/Private Ventures, Philadelphia). Retrieved from http://www.ppv.org/ppv/publications/assets/61_publication.pdf

13.

Cohen

West

Aiken

(2003). Applied multiple regression/correlation analysis for the behavioral sciences. Mahwah, NJ: Lawrence Erlbaum Associates.

14.

Creswell

Way

Eisenberger

Lieberman

(2007). Neural correlates of dispositional mindfulness during affect labeling. Psychosomatic Medicine, 69, 560-565.

15.

Dandreaux

Frick

(2009). Developmental pathways to conduct problems: A further test of the childhood and adolescent-onset distinction. Journal of Abnormal Child Psychology, 37, 375-385.

16.

Davidson

Kabat-Zinn

Schumacher

Rosendranz

Muller

Santorelli

Urbanowski

. (2003). Alterations in brain and immune function produced by Mindfulness Meditation. Psychosomatic Medicine, 65, 564–70.

17.

Doremus-Fitzwater

Varlinskaya

Spear

(2010). Motivational systems in adolescence: Possible implications for age differences in substance abuse and other risk-taking behaviors. Brain and Cognition, 72, 114-123.

18.

Dumontheil

Houlton

Christoff

Blakemore

(2010). Development of relational reasoning during adolescence. Developmental Science, 13, F15-F24.

19.

Dyk

Adams

(1987). The association between identity development and intimacy during adolescence: A theoretical treatise. Journal of Adolescent Research, 2, 223-235.

20.

Ernst

Nelson

Jazbec

McClure

Monk

Leibenluft

Blair

Pine

(2005). Amygdala and nucleus accumbens in response to receipt and omission of gains in adults and adolescents. NeuroImage, 25, 1279-1291.

21.

Ernst

Pine

Hardin

(2005). Triadic model of the neurobiology of motivated behavior in adolescence. Psychological Medicine, 36, 299-312.

22.

Evans-Chase

Zhou

(in press). A systematic review of the juvenile justice intervention literature: What it can (and cannot) tell us about what works with delinquent youth. Crime & Delinquency.

23.

Fareri

Martin

Delgado

(2008). Reward-related processing in the human brain: Developmental considerations. Development and Psychopathology, 20, 1191-1211.

24.

Finger

Marsh

Blair

Reid

Sims

Ng.

Blair

(2011). Disrupted reinforcement signaling in the orbitofrontal cortex and caudate in youths with conduct disorder or oppositional defiant disorder and a high level of psychopathic traits. American Journal of Psychiatry, 168, 152-162.

25.

Forbes

Dahl

(2010). Pubertal development and behavior: Hormonal activation of social and motivational tendencies. Brain and Cognition, 72, 66-72.

26.

Froh

Kashdan

Yurkewicz

Fan

Allen

Glowacki

(2010). The benefits of passion and absorption in activities: Engaged living in adolescents and its role in psychological well-being. The Journal of Positive Psychology, 5, 311-332.

27.

Gardner

Steinberg

(2005). Peer influence on risk taking, risk preference, and risky decision making in adolescence and adulthood: An experimental study. Developmental Psychology, 41, 625-635.

28.

Gottfredson

Hirschi

(1990). A general theory of crime. Stanford, CA: Stanford University Press.

29.

Graber

Nichols

Brooks-Gunn

(2009). Putting pubertal timing in developmental context: Implications for prevention. Developmental Psychobiology, 52, 254-262.

30.

Gunnar

Wewerka

Frenn

Long

Griggs

(2009). Developmental changes in hypothalamus-pituitary-adrenal activity over the transition to adolescence: Normative changes and associations with puberty. Development and Psychopathology, 21, 69-85.

31.

Hanlon

Bateman

Simon

O’Grady

Carswel

(2002). An early community-based intervention for the prevention of substance abuse and other delinquent behavior. Journal of Youth and Adolescence, 12, 459-471.

32.

Henggeler

Clingempeel

Brondino

Pickrel

(2002). Four-year follow-up of Multisystemic Therapy with substance-abusing and substance-dependent juvenile offenders. Journal of the Academy of Child and Adolescent Psychiatry, 41, 868-874.

33.

Hirschi

Gottredson

(1983). Age and the explanation of crime. American Journal of Sociology, 89, 552-561.

34.

Holzel

Ott

Hempel

Hackl

Wolf

Stark

Vaitl

(2007). Differential engagement of anterior cingulated and adjacent medial frontal cortex in adept meditators and non-meditators. Neuroscience Letters, 421, 16-21.

35.

Jones

Laurens

Herba

Barker

Viding

(2009). Amygdala hypoactivity to fearful faces in boys with conduct problems and callous-unemotional traits. American Journal of Psychiatry, 166, 95-102.

36.

Kambam

Thompson

(2009). The development of decision-making capacities in children and adolescents: Psychological and neurological perspectives and their implications for juvenile defendants. Behavioral Science and the Law, 27, 173-190.

37.

Kline

R. B.

(2005). Principles and practice of structural equation modeling (2nd ed.). New York: Guilford.

38.

Knoll

Sickmund

(2010). Delinquency cases in juvenile court, 2007. OJJDP Fact Sheet. Retrieved from http://www.ncjrs.gov/pdffiles1/ojjdp/230168.pdf

39.

Lipsey

(1994). Juvenile delinquency treatment: A meta-analytic inquiry into the variability of effects. In Cook

Cooper

Cordray

Hartmann

Hedges

Light

Louis

Mosteller

(Eds.), Meta-analysis for explanation (pp. 83-127). New York: Russell Sage Foundation.

40.

Lipsey

(1999). Can rehabilitative programs reduce the recidivism of juvenile offenders? An inquiry into the effectiveness of practical programs. Journal of Social Policy & Law, 6, 611-641.

41.

Lipsey

Wilson

Cothern

(2000, April). Effective interventions for serious juvenile offenders. OJJDP Juvenile Justice Bulletin. Retrieved from http://www.ncjrs.gov/pdffiles1/ojjdp/181201.pdf

42.

Luna

Padmanabhan

O’Hearn

(2009). What has fMRI told us about the development of cognitive control through adolescence? Brain and Cognition, 72, 101-113.

43.

Marsh

Finger

Fowler

Jurkowitz

Schechter

Blair

(2011). Reduced amygdala-orbitofrontal connectivity during moral judgments in youths with disruptive behavior disorders and psychopathic traits. Psychiatry Research: Neuroimaging, 194, 279-286.

44.

Marsh

Finger

Mitchell

Reid

Sims

Kosson

Blair

(2008). Reduced amygdala response to fearful expressions in children and adolescents with callous-unemotional traits and disruptive behavior disorders. American Journal of Psychiatry, 165, 712-720.

45.

McLean

Breen

Fournier

(2010). Constructing the self in early, middle, and late adolescent boys: Narrative, identity, individuation, and wellbeing. Journal of Research on Adolescence, 20, 166-187.

46.

McLoughlin

Rucklidge

Grace

McLean

(2010). Can callous-unemotional traits and aggression identify children at high-risk of anti-social behavior in a low socioeconomic group? Journal of Family Violence, 25, 701-712.

47.

Moffitt

(1993). Adolescent-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674-701.

48.

Pealer

(2004). A community of peers: Promoting behavior change. The effectiveness of a therapeutic community for juvenile male offenders in reducing recidivism (Doctoral dissertation). Retrieved from Proquest Dissertations and Theses Fulltext. (3146517).

49.

Puzzanchera

Adams

Sickmund

(2010). Juvenile court statistics 2006–2007. Pittsburgh, PA: National Center for Juvenile Justice.

50.

Quevedo

Benning

Gunnar

Dahl

(2009). The onset of puberty: Effects on the psychophysiology of defensive and appetitive motivation. Development and Psychopathology, 21, 27-45.

51.

Rodriguez

(2007). Restorative justice at work: Examining the impact of restorative justice resolutions on juvenile recidivism. Crime & Delinquency, 53, 355-379.

52.

Schaeffer

(2000). Moderators and mediators of therapeutic change in multisystemic treatment of serious juvenile offenders (Doctoral dissertation). Retrieved from Proquest Dissertations and Theses Fulltext. (9999313)

53.

Scott

Steinberg

(2008). Rethinking juvenile justice. Cambridge, MA: Harvard University Press.

54.

Sedlak

McPherson

(2010). Conditions of confinement: Findings from the survey of youth in residential placement (OJJDP Juvenile Justice Bulletin). Washington, DC: US Department of Justice, Office of Juvenile Justice and Delinquency Prevention (OJJDP). Retrieved from http://www.ncjrs.gov/pdffiles1/ojjdp/227729.pdf

55.

Shelden

(1997). Detention diversion advocacy: An evaluation (Juvenile Justice Bulletin). Washington, DC: U.S. Department of Justice, Office of Juvenile Justice and Delinquency Prevention. Retrieved from www.ncjrs.gov/pdffiles1/ojjdp/171155.pdf

56.

Sickmund

Sladky

T. J.

Kang

Puzzanchera

(2011). Easy access to the census of juveniles in residential placement. Retrieved from http://www.ojjdp.gov/ojstatbb/ezacjrp/

57.

Smetana

Campione-Barr

Metzger

(2006). Adolescent development in interpersonal and societal contexts. Annual Review of Psychology, 57, 255-284.

58.

Snyder

Sickmund

(2006). Juvenile offenders & victims: 2006 national report. Washington, DC: U.S. Department of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention. Retrieved from http://www.ojjdp.gov/ojstatbb/nr2006/downloads/nr2006.pdf

59.

Somerville

Jones

Casey

(2010). A time of change: Behavioral and neural correlates of adolescent sensitivity to appetitive and aversive environmental cues. Brain and Cognition, 72, 124-133.

60.

Steinberg

(2005). Cognitive and affective development in adolescence. TRENDS in Cognitive Sciences, 9, 69-74.

61.

Steinberg

(2008). A social neuroscience perspective on adolescent risk-taking. Developmental Review, 28, 78-106.

62.

Steinberg

(2009). Should the science of adolescent brain development inform public policy? American Psychologist, 64, 739-750.

63.

Steinberg

Albert

Cauffman

Banuch

Graham

Woolard

(2008). Age differences in sensation seeking and impulsivity as indexed by behavior and self-report: Evidence for a dual systems model. Developmental Psychology, 44, 1764-1778.

64.

Steinberg

Scott

(2003). Less guilty by reason of adolescence: Developmental immaturity, diminished responsibility and the juvenile death penalty. American Psychologist, 58, 1009-1018.

65.

Stroud

Foster

Papandonatos

Handwerger

Granger

Kivlighan

Niaura

(2009). Stress response and the adolescent transition: Performance versus peer rejection stressors. Development and Pscyhopathology, 21, 47-68.

66.

Viding

Jones

Frick

Moffitt

Plomin

(2008). Heritability of antisocial behavior at 9: Do callous-unemotional traits matter? Developmental Science, 11, 17-22.

67.

Wooldridge

J. M.

(2006). Introductory econometrics: A modern approach (3rd ed.). Cincinnati, OH: South-Western College Publishing,

68.

Wells

Minor

Angel

Stearman

(2006). A quasi-experimental evaluation of a shock incarceration and aftercare program for juvenile offenders. Youth Violence and Juvenile Justice, 4, 219-233.