The childhood- versus adolescent-onset distinction

A number of reviews of the literature have summarized research supporting the distinction between children who begin showing severe conduct problems and antisocial behavior in childhood versus those whose onset of antisocial behavior does not emerge until adolescence (Moffitt, Reference Moffitt1993, Reference Moffitt, Lahey, Moffitt and Caspi2003; Patterson, Reference Patterson1996). Children in the childhood-onset group often begin showing mild conduct problems as early as preschool or early elementary school, and their behavioral problems tend to increase in rate and severity throughout childhood and into adolescence (Lahey & Loeber, Reference Lahey, Loeber and Routh1994). In contrast, the adolescent-onset group does not show significant behavioral problems in childhood, but they begin exhibiting significant antisocial and delinquent behavior coinciding with the onset of adolescence (Moffitt, Reference Moffitt, Lahey, Moffitt and Caspi2003). In addition to the different patterns of onset, the childhood-onset group is more likely to show aggressive behaviors in childhood and adolescence, and is more likely to continue to show antisocial and criminal behavior into adulthood (for a review, see Frick & Loney, Reference Frick, Loney, Quay and Hogan1999).

For example, Farrington, Gallagher, Morley, St. Ledger, and West (Reference Farrington, Gallagher, Morley, St. Ledger and West1988) reported that boys who were arrested prior to age 12 showed almost twice as many convictions at two later points in time (between the ages of 16 and 18 and between the ages of 22 and 24). Similarly, Robins (Reference Robins1966) reported that boys referred to a mental health clinic for antisocial behavior prior to age 11 were twice as likely to receive a diagnosis of an antisocial personality disorder as an adult, compared to boys who began showing antisocial behavior after age 11. As a final example, Moffitt, Caspi, Harrington, and Milne (Reference Moffitt, Caspi, Harrington and Milne2002) reported that, within a birth cohort of 539 males born in New Zealand and followed through age 26, the 45 men who had displayed significant conduct problems prior to adolescence were more likely to have had a criminal conviction (55%) and to have a greater number of convictions (M = 6.9, SD = 11.5) as adults than the 121 men who showed significant conduct problems starting in adolescence (34%; M = 3.5, SD = 10.8). Both groups, however, differed from controls without histories of conduct problems (17%; M = 0.6, SD = 3.1). The difference was more dramatic when the focus was on convictions for violent offenses, with the early-onset group being much more likely (38%) to be convicted as an adult than either the adolescent-onset (14%) or control (5%) groups.

Thus, the childhood and adolescent subtypes of antisocial youth show very different trajectories of antisocial behavior, both in terms of their patterns of onset and life-course trajectory. However, if these were the only differences between groups, this method for subtyping would not be much different from other approaches that have focused on the different behavioral trajectories of antisocial behavior over time. Of importance, there is now a rather extensive body of research to suggest that the two groups also differ on a number of dispositional and contextual risk factors that seem to implicate different developmental processes in the disruptive behaviors of the two groups (Moffitt, 2002). To summarize these findings, childhood-onset conduct problems seem to be more strongly related to neuropsychological (e.g., deficits in executive functioning) and cognitive (e.g., low intelligence) deficits (Fergusson, Lynsky, & Horwood, Reference Fergusson, Lynsky and Horwood1996; Kratzer & Hodgins, Reference Kratzer and Hodgins1999; Piquero, Reference Piquero2001; Raine, Yaralian, Reynolds, Venables, & Mednick, Reference Raine, Yaralian, Reynolds, Venables and Mednick2002). Further, the childhood-onset group has been reported to show more temperamental and personality risk factors, such as impulsivity (McCabe, Hough, Wood, & Yeh, Reference McCabe, Hough, Wood and Yeh2001; Silverthorn, Frick, & Reynolds, Reference Silverthorn, Frick and Reynolds2001), attention deficits (Fergusson et al., Reference Fergusson, Lynsky and Horwood1996), and problems in emotional regulation (Moffitt, Caspi, Dickson, Silva, & Stanton, Reference Moffitt, Caspi, Dickson, Silva and Stanton1996). This group has also been shown to come from homes with greater family instability, more family conflict, and with parents who use less effective parenting strategies (Aguilar, Sroufe, Egeland, & Carlson, Reference Aguilar, Sroufe, Egeland and Carlson2000; McCabe et al., Reference McCabe, Hough, Wood and Yeh2001; Patterson & Yoerger, Reference Patterson, Yoerger and Osgood1997; Woodward, Fergusson, & Horwood, Reference Woodward, Fergusson and Horwood2002). When children within the adolescent-onset group differ from control children without conduct problems, it is often in showing higher levels of rebelliousness and being more rejecting of conventional values (Dandreaux & Frick, Reference Dandreaux and Frick2009; Moffitt et al., Reference Moffitt, Caspi, Dickson, Silva and Stanton1996).

The different outcomes and risk factors for the two subtypes of antisocial individuals have led to theoretical models that propose very different causal mechanisms operating across the two groups. For example, Moffitt (Reference Moffitt, Lahey, Moffitt and Caspi2003) has proposed that children in the childhood-onset group develop their problem behavior through a transactional process involving a difficult and vulnerable child (e.g., impulsive, with verbal deficits) who experiences an inadequate rearing environment (e.g., poor parental supervision, poor quality schools). This dysfunctional transactional process disrupts the child's socialization leading to poor social relations with persons both inside (e.g., parents and siblings) and outside the family (e.g., peers and teachers). These disruptions lead to enduring vulnerabilities that can negatively affect the child's psychosocial adjustment across multiple developmental stages.

In contrast, children in the adolescent-onset pathway have problems that are more likely to be limited to adolescence and show fewer risk factors. Thus, this group is conceptualized as showing an exaggeration of the normative process of adolescent rebellion (Moffitt, Reference Moffitt, Lahey, Moffitt and Caspi2003). That is, all adolescents show some level of rebelliousness to parents and other authority figures (Brezina & Piquero, Reference Brezina and Piquero2007). This rebelliousness is part of a process by which the adolescent begins to develop his or her autonomous sense of self and his or her unique identity. According to Moffitt (Reference Moffitt, Lahey, Moffitt and Caspi2003), the child in the adolescent-onset group engages in antisocial and delinquent behaviors as a misguided attempt to obtain a subjective sense of maturity and adult status in a way that is maladaptive (e.g., breaking societal norms) but encouraged by an antisocial peer group. Given that their behavior is viewed as an exaggeration of a process specific to adolescence, and not because of an enduring vulnerability, their antisocial behavior is less likely to persist beyond adolescence. However, they may still have impairments that persist into adulthood because of the consequences of their adolescent antisocial behavior (e.g., a criminal record, dropping out of school, substance abuse; Moffitt & Caspi, Reference Moffitt and Caspi2001).

Callous–unemotional (CU) traits and subgroups of antisocial youth

This distinction between the childhood and adolescent-onset patterns of antisocial behaviors illustrates a subgrouping approach for antisocial individuals that proposes different developmental mechanisms underlying the problem behavior across groups. Thus, this is an example of the important concept in developmental psychopathology of equifinality, which proposes that the same developmental outcome (e.g., antisocial behavior) can result from very different developmental processes (Cicchetti & Rogosch, Reference Cicchetti and Rogosch1996). It is important that research has begun extending the concept of equifinality for understanding antisocial youth by exploring whether additional distinctions can be made within groups who show a childhood onset to their antisocial behavior.

This distinction is based on the presence of a callous and unemotional interpersonal style characterized by a lack of guilt and empathy and a callous use of others. It is similar to the distinction made within samples of antisocial adults using the construct of psychopathy that was described previously. As noted previously, within incarcerated adults, the presence of psychopathic traits has proven to designate an important subgroup of antisocial individual who show a more severe, violent, and difficult to treat pattern of antisocial behavior (Hemphill, Reference Hemphill, Herve and Yuille2007; Leistico et al., Reference Leistico, Salekin, DeCoster and Rogers2008; Porter & Woodworth, Reference Porter, Woodworth and Patrick2006). They also show a number of different background and emotional characteristics that seem to suggest a unique etiology to their antisocial behavior (Blair et al., Reference Blair, Mitchell and Blair2005; Patrick, Reference Patrick, Herves and Yuille2007). There is growing evidence that the same may be true for antisocial youth who show CU traits.

Specifically, youth who show the childhood onset to their antisocial behavior tend to score higher on measures of CU traits than those who show an adolescent onset (Dandreaux & Frick, Reference Dandreaux and Frick2009; Moffitt et al., Reference Moffitt, Caspi, Dickson, Silva and Stanton1996; Silverthorn et al., Reference Silverthorn, Frick and Reynolds2001). Further, several recent qualitative (Frick & Dickens, Reference Frick and Dickens2006; Frick & White, Reference Frick and White2008) and quantitative (Edens, Campbell, & Weir, Reference Edens, Campbell and Weir2007; Leistico et al., Reference Leistico, Salekin, DeCoster and Rogers2008) reviews have been published showing that CU traits are predictive of a more severe, stable, and aggressive pattern of behavior in antisocial youth. For example, Edens et al. (Reference Edens, Campbell and Weir2007) conducted a quantitative meta-analyses of 21 nonoverlapping samples showing that measures that include CU traits were associated with general or violent recidivism with effect sizes of r = .24 or .25, respectively. Frick and Dickens (Reference Frick and Dickens2006) reported on a qualitative review of 24 published studies using 22 independent samples. Ten of these studies showed a concurrent association between CU traits and measures of aggressive, antisocial, or delinquent behavior and 14 studies showed a predictive relationship with follow-up intervals ranging from 6 months to 10 years. These authors further reported on 5 studies showing that CU traits were associated with poorer treatment outcomes. Of importance, these studies included community (n = 6), clinic-referred (n = 4), and forensic (n = 13) samples, and they included samples ranging in age from 4 to 20.

One important issue in interpreting this body of research is that CU traits seem to predict later antisocial behavior even after controlling for other risk factors (e.g., past criminal offenses, drug use, delinquent peers; Salekin, Reference Salekin2008). Further, the association between CU traits and aggressive behavior could explain some of the typologies proposed for understanding aggressive individuals. That is, youth with CU traits not only show a more severe and pervasive pattern of aggressive behavior but they also tend to show aggression that is both reactive and proactive in nature (Enebrink, Andershed, & Langstrom, Reference Enebrink, Anderson and Langstrom2005; Frick, Cornell, Barry, Bodin, & Dane, Reference Frick, Cornell, Barry, Bodin and Dane2003; Kruh, Frick, & Clements, Reference Kruh, Frick and Clements2005). In contrast, antisocial youth without CU traits tend to show less aggression overall and, when they do show aggressive behavior, it tends to be largely reactive in nature. Further, many of the social–cognitive and affective differences that have been found between reactive and proactive aggression may be due to the differences in their association with CU traits (Munoz, Frick, Kimonis, & Aucoin, Reference Munoz, Frick, Kimonis and Aucoin2008; Pardini, Lochman, & Frick, Reference Pardini, Lochman and Frick2003; Waschbusch, Walsh, Andrade, King, & Carrey, Reference Waschbusch, Walsh, Andrade, King and Carrey2007).

For example, Pardini et al. (Reference Pardini, Lochman and Frick2003) studied detained adolescents' (ages 11–18) responses to eight vignettes depicting peers involved in aggressive acts in various age-appropriate social contexts. Youth were asked to respond to questions asking how likely and how important various possible outcomes to the aggressive interpersonal situations were. In this ethnically diverse sample of boys and girls, CU traits were associated with responses indicating a tendency to emphasize the positive and rewarding aspects of aggression, to value the importance of being dominant in aggressive interactions, and to minimize the potential for punishment for being aggressive. All of these social cognitive factors have been associated with proactive aggression in past research (Dodge & Pettit, Reference Dodge and Pettit2003). Similarly, Munoz et al. (Reference Munoz, Frick, Kimonis and Aucoin2008) studied 85 boys (ages 13–18) who had been detained for delinquent behavior. They reported that those who showed high rates of both reactive and proactive aggression showed lower levels of emotional reactivity to provocation, consistent with past research on adolescents who show proactive aggression (e.g., Pitts, Reference Pitts, Raine, Brennan, Farrington and Mednick1997). However, this reduced reactivity to provocation was only found for those who were also high on CU traits. Finally, Frick, Cornell, Bodin, Dane, Barry, and Loney (Reference Frick, Cornell, Bodin, Dane, Barry and Loney2003) reported that a tendency to attribute hostile intent to the actions of others (i.e., a hostile attribution biase) that has been associated with reactive forms of aggression was only found in boys with conduct problems who did not show CU traits.

Frick and White (Reference Frick and White2008) provided a comprehensive review of this research on the social, cognitive, and emotional characteristics of antisocial youth with and without CU traits. First, they reviewed four studies showing that the conduct problems in youth without CU traits are more strongly related to dysfunctional parenting practices (see also Edens, Skopp, & Cahill, Reference Edens, Skopp and Cahill2008). Second, they reviewed 10 studies showing differences in how antisocial youth with and without CU traits process emotional stimuli, with youth high on CU traits showing deficits in the processing of negative emotional stimuli and, even more specifically, deficits to signs of fear and distress in others. Third, another 10 studies were reviewed showing distinct cognitive characteristics of antisocial youth with CU traits, such as being less sensitive to punishment cues, especially when a reward oriented response set is primed, showing more positive outcome expectancies in aggressive situations with peers, and being more likely to exhibit verbal deficits than other antisocial youth. Fourth, they reviewed seven studies showing that youth with CU traits have unique personality characteristics, such as showing more fearless or thrill seeking behaviors and less trait anxiety or neuroticism.

Thus, taken together, there is now a fairly substantial body of research to suggest that CU traits designate an important subgroup of antisocial youth, that differ both in the severity and stability of their behavior but also on important emotional, cognitive, and social characteristics. These latter findings could suggest distinct etiological mechanisms leading to their antisocial behavior. As a result, like the research reviewed on the childhood and adolescent-onset distinction, this research provides additional evidence supporting the concept of equifinality in understanding the development of antisocial behavior in youth.

Before discussing the different etiological mechanisms and their relationship with normal developmental processes, it is also important to note that research on CU traits also supports the related concept of multifinality, which recognizes that the same risk factors can have multiple developmental outcomes. Specifically, as noted previously, one characteristic of youth with CU traits is the presence of a fearless, thrill seeking, and behaviorally uninhibited temperament. Consistent with the importance of this temperament for the development of CU traits, Cornell and Frick (Reference Cornell and Frick2007) reported that preschool children who were rated by their teachers as being behaviorally uninhibited were more at risk for showing problems in empathy and guilt than other children. However, they also reported that uninhibited preschoolers showed enhanced conscience development if they experienced consistent discipline and a parenting style that emphasized a strong and obedience-oriented (i.e., authoritarian) approach to parenting. These authors suggested that the underarousal exhibited by fearless children may require parents to incorporate stronger methods of socialization that bring arousal levels into an optimal range for the child to internalize parental norms for prosocial behavior (Fowles & Kochanska, Reference Fowles and Kochanska2000; Kochanska, DeVet, Goldman, Murray, & Putnam, Reference Kochanska, DeVet, Goldman, Murray and Putnam1994). Thus, the same temperamental risk factor (i.e., fearless and uninhibited) could have different outcomes (i.e., normal or deficient conscience development) depending on the type of parenting the child experiences. A similar possibility has been proposed by Patrick (in press) who suggests that the same temperament of fearlessness can lead some children to become “bold” and others “mean” depending on the presence of other temperamental traits or depending on the presence of certain types of socializing environments.

Understanding developmental mechanisms

The research reviewed in the previous section clearly suggests that there are important subgroups of antisocial youth who are at differential risk for continuing their antisocial behavior into adulthood and who show different patterns of risk factors that could suggest etiologies involving distinct developmental mechanisms. As noted previously, based on the differences in the course and correlates to antisocial behavior for youth with adolescent-onset antisocial behavior, Moffitt (Reference Moffitt, Lahey, Moffitt and Caspi2003) proposed that these children show an exaggeration of the normative process of adolescent rebellion. In contrast, children in the childhood-onset group show significant problems in adjustment across multiple developmental stages. Further, as also noted above, they also show more numerous and more severe dispositional and environmental risk factors associated with their behavior problems. However, within this group, those who show CU traits seem to show different characteristics that could suggest some differences in which developmental mechanisms may be involved in their antisocial propensities.

Specifically, the research reviewed previously suggests that children with CU traits designate a group of children with severe conduct problems who show a distinct temperamental style, characterized by a preference for dangerous and novel stimuli, a reward-oriented response style, and a lack of reactivity to emotional stimuli that signify distress in others. The bold and fearless temperamental style has been variously labeled as low fearfulness (Rothbart & Bates, Reference Rothbart, Bates, Damon and Eisenberg1998), low behavioral inhibition (Kagan & Snidman, Reference Kagan and Snidman1991), low harm avoidance (Cloninger, Reference Cloninger1987), or high daring (Lahey & Waldman, Reference Lahey, Waldman, Lahey, Moffitt and Caspi2003). Of importance, there have been a number of studies of normally developing children documenting both concurrent (e.g., Fowles & Kochanska, Reference Fowles and Kochanska2000; Kochanska, Gross, Lin, & Nichols, Reference Kochanska, Gross, Lin and Nichols2002) and predictive (Rothbart, Ahadi, & Hershey, Reference Rothbart, Ahadi and Hershey1994) associations between a fearless temperament and lower scores on measures of conscience development. Further, this research has led to a number of theories to explain this link. For example, fearless children may be less likely to experience transgression-related arousal in response to behavior that has been punished by others (Kochanska, Reference Kochanska1993; Newman, Reference Newman1987). In addition, children with this temperament may be less likely to experience empathic arousal linked to the distress in others (Blair, Reference Blair1999). In short, the temperamental deficits in different aspects of emotional reactivity could make it more difficult for a child to develop appropriate levels of guilt, empathy, and other dimensions of conscience that, at its extreme, could result in CU traits and severe patterns of antisocial behavior. Consistent with this theoretical model, Pardini (Reference Pardini2006) reported that the association between fearlessness and violent delinquency was mediated by the presence of CU traits in a sample of adjudicated adolescents.

As noted previously, children with childhood-onset antisocial behavior but without CU traits show very different dispositional (e.g., impulsivity, low verbal intelligence, poor emotional regulation) and contextual (e.g., higher rates of family dysfunction) risk factors. Most importantly, the children in this group show high rates of anxiety (Andershed, Gustafson, Kerr & Stattin, Reference Andershed, Gustafson, Kerr and Stattin2002; Frick, Lilienfeld, Ellis, Loney, & Silverthorn, Reference Frick, Lilienfeld, Ellis, Loney and Silverthorn1999; Pardini, Lochman, & Powell, Reference Pardini, Lochman and Powell2007), they do not typically show problems in empathy and guilt (Pardini et al., Reference Pardini, Lochman and Frick2003), and they appear to be distressed by the effects of their behavior on others (Loney, Frick, Clemens, Ellis, & Kerlin, Reference Loney, Frick, Clements, Ellis and Kerlin2003; Pardini et al., Reference Pardini, Lochman and Frick2003). Thus, the antisocial behavior in this group does not seem to be easily explained by deficits in conscience development.

Given the many different types of risk factors that are found in this group, it is likely that there are a number of different causal processes that could lead to the impulsive and antisocial behavior exhibited by these youth (Frick & Morris, Reference Frick and Morris2004). For example, in this group, there is a strong association between ineffective parenting practices and their antisocial behavior. Thus, it is possible that children in this group are not socialized adequately and, as a result, do not learn to appropriately regulate their behavior in response to environmental contingencies (Kochanska et al., Reference Kochanska, Gross, Lin and Nichols2002). Another consistent finding for antisocial youth without CU traits is that they often have problems regulating their emotions. That is, children in this group appear to show a temperament characterized by strong emotional reactivity, a deficit in the skills needed to adequately regulate their emotional reactivity, or both (Frick & Morris, Reference Frick and Morris2004). These problems in emotional regulation can result in the child committing impulsive and unplanned aggressive and antisocial acts for which he or she may be remorseful afterward but may still have difficulty controlling in the future (Pardini et al., Reference Pardini, Lochman and Frick2003). Such problems in regulating emotion would also explain the findings that the aggressive behavior displayed by this group tends to be confined to reactive forms of aggression (Frick, Kimonis, Dandreaux, & Farrell, Reference Frick, Kimonis, Dandreaux and Farrell2003; Kruh et al., Reference Kruh, Frick and Clements2005).

Different types of deviations from normal development

Thus, this developmental model of antisocial behavior specifies three subgroups of antisocial individuals that differ on the developmental trajectory of their antisocial behavior and on the developmental mechanisms that seem to be involved in their problem behavior. Like symptoms of most forms of psychopathology, some level of antisocial and aggressive behavior is normal in children and adolescents (Moffitt et al., Reference Moffitt, Caspi, Dickson, Silva and Stanton1996). Therefore, it is critical to determine the relationship between normal and abnormal patterns of antisocial behavior. The developmental models outlined above illustrate two ways that abnormal antisocial behavior can differ from normal development. Specifically, the problems in adjustment experienced by the adolescent-onset group seem to be more specific to a single developmental stage (i.e., adolescence) and result from a failure to adequately adjust to the developmental demands (e.g., separation and individuation from parents) of that stage. In contrast, both groups who show a childhood onset to their antisocial behavior show risk factors (e.g., emotional and cognitive deficits) that negatively affect development across multiple stages, albeit in different ways for the two groups.

Potential gender differences

In general, the majority of research on antisocial behavior in youth has focused on the samples of boys. Thus, the applicability of these developmental models to girls is open to question. One consistent finding is that a childhood onset to severe antisocial and aggressive behavior is much rarer in girls than in boys (Hipwell et al., Reference Hipwell, Loeber, Stouthamer-Loeber, Keenan, White and Kroneman2002; Moffitt & Capsi, Reference Moffitt and Caspi2001; White & Piquero, Reference White and Piquero2004). However, despite the predominance of adolescent onset in antisocial girls, there is evidence that girls with severe conduct problems show poor outcomes in adulthood and show a large number of the dispositional and contextual risk factors that are more characteristic of childhood-onset antisocial behavior in boys (Frick & Dickens, Reference Frick and Dickens2006).

To reconcile these findings, Silverthorn and Frick (Reference Silverthorn and Frick1999) proposed a modification of the developmental model outlined above which they labeled as a delayed-onset pathway to antisocial behavior for girls. These authors proposed that antisocial and aggressive behavior in girls show the same causal mechanisms as those outlined previously for childhood-onset boys. However, their severe antisocial behavior is often delayed until adolescence coinciding with biological (e.g., hormonal changes associated with puberty) and psychosocial (e.g., less parental monitoring and supervision; greater contact with deviant peers) changes that encourage antisocial behavior in girls with predisposing vulnerabilities (e.g., CU; problems in emotional regulation). In an initial test of this theory, adjudicated adolescent girls who largely showed an adolescent onset to their antisocial behavior also showed high levels of CU traits, problems with impulse control, and a number of other social and temperamental vulnerabilities that were more similar to childhood-onset boys than to adolescent-onset boys (Silverthorn et al., Reference Silverthorn, Frick and Reynolds2001). Despite this initial positive finding, additional tests of this model have been more mixed (Lahey et al., Reference Lahey, Van Hulle, Waldman, Rodgers, D'Onofrio and Pedlow2006; McCabe et al., Reference McCabe, Hough, Wood and Yeh2001; Moffitt & Caspi, Reference Moffitt and Caspi2001; White & Piquero, Reference White and Piquero2004).

As a result of these conflicting findings, the predictions made from the delayed-onset theory require further testing. Also, it is possible that, although many girls may not show the overt antisocial and aggressive behavior prior to adolescence, they may show other forms of antisocial behavior such as relational aggression. Relational aggression can be defined as behaviors that seek to harm another child's social relationships (e.g., telling lies about them, excluding them from social events) rather than physically harming the child (Crick & Grotpeter, Reference Crick1996) Several studies have shown that when girls behave aggressively, they are more likely to choose relational aggression (rather than physical or overt aggression) as a strategy for use within the peer group (Crick, Reference Crick1996; Crick, Casas, & Mosher, Reference Crick, Casas and Mosher1997; Crick & Grotpeter, Reference Crick and Grotpeter1995; Lagerspetz, Björkqvist, & Peltonen, Reference Lagerspetz, Björkqvist and Peltonen1988; Ostrov & Keating, Reference Ostrov and Keating2004). Of importance, relational aggression shares some of the same risk factors as those reviewed previously as being associated with antisocial behavior, including impulsivity (Zalecki & Hinshaw, Reference Zalecki and Hinshaw2004) and CU traits (Marsee & Frick, Reference Marsee and Frick2007). Therefore, it may be that girls with risk factors for antisocial behavior display relational aggressive behaviors prior to adolescence, giving them the appearance of showing an adolescent onset, if such behaviors are not assessed.

Stable personality traits versus developmental transitions

As noted previously, there is some debate as to whether antisocial personality should be defined solely by a consistent pattern of antisocial behavior or by the underlying personality traits that can lead to this pattern of behavior. In either case, there is some assumption of stability across development. In fact, many definitions of antisocial personality disorder require the presence of antisocial behavior starting in childhood or adolescence (American Psychiatric Association, 2000). Prospective studies of either clinic referred (Lahey, Loeber, Burke, & Applegate, Reference Lahey, Loeber, Burke and Applegate2005; Robins, Reference Robins1966) or community participants (Moffitt et al., Reference Moffitt, Caspi, Harrington and Milne2002) have suggested that between 14 and 54% of children with severe conduct problems will show an antisocial personality disorder as adults. Further, as noted previously, children with a childhood onset to their conduct problems (Moffitt, Reference Moffitt, Lahey, Moffitt and Caspi2003) and those with CU traits (Frick & Dickens, Reference Frick and Dickens2006; Frick & White, Reference Frick and White2008) seem to show a more stable pattern of antisocial behavior. Thus, this research suggests that some patterns of antisocial behavior are more stable than others. Unfortunately, there are no clear guidelines to suggest what level of stability in antisocial behavior would be sufficient to warrant designation as a personality disorder. Further, even if a pattern of behavior is not highly stable, such as adolescent-onset conduct problems, it may still lead to significant impairments and warrant treatment (Dandreaux & Frick, Reference Dandreaux and Frick2009; Moffitt et al., Reference Moffitt, Caspi, Dickson, Silva and Stanton1996).

The issue of stability may be even more important if one focuses on the personality traits that may lead to antisocial behavior. That is, an important consideration is whether CU traits are stable enough in children or adolescents to warrant the designation of “traits” that implies some level of continuity across development (Edens, Skeem, Cruise, & Caufmann, Reference Edens, Skeem, Cruise and Cauffman2001; Seagrave & Grisso, Reference Seagrave and Grisso2002). There are now a number of studies showing that these traits are relatively stable from late childhood to early adolescence both when assessed by self-report (Munoz & Frick, Reference Munoz and Frick2007) or by parent report (Frick, Kimonis, et al., Reference Frick, Kimonis, Dandreaux and Farrell2003). For example, Frick, Kimonis, et al. (Reference Frick, Kimonis, Dandreaux and Farrell2003) reported a stability estimate of .71 using an intraclass correlation coefficient across 4 years for CU traits in a sample of children with an average age of 10.65 at the initial assessment. This level of stability for parent report is much higher than is typically reported for parent ratings of others aspects of children's adjustment (Verhulst, Koot, & Berden, Reference Verhulst, Koot and Berden1990). With respect to younger children, Dadds, Frazer, Frost, and Hawes (Reference Dadds, Fraser, Frost and Hawes2005) found moderate 1-year stability estimates for parent-reported CU traits (r = .55) in a community sample of Australian children who were 4 to 9 years of age. Using a more extended follow-up period (i.e., 9 years), Obradovic, Pardini, Long, and Loeber (Reference Obradović, Pardini, Long and Loeber2007) reported relatively high rates of stability for parent (r = .50) but modest levels of stability for teacher ratings (r = .27) ratings of CU traits for boys between the ages of 8 and 16. Blonigen, Hicks, Kruger, Patrick, and Iacono (Reference Blonigen, Hicks, Kruger, Patrick and Iacono2006) reported that CU traits were relatively stable (r = .60) from late adolescence (age 17) into early adulthood (age 24). Further, Loney, Taylor, Butler, and Iacono (Reference Loney, Taylor, Butler and Iacono2007) reported that CU traits in adolescence (ages 16–18) were significantly stable (intraclass correlation coefficient = .40) over a 6-year follow-up period. Finally, two studies have shown that measures of CU traits assessed prior to adolescence were significantly associated with similar measures in adulthood, even after controlling for other childhood risk factors (Burke, Loeber, & Lahey, Reference Burke, Loeber and Lahey2007; Lynam, Caspi, Moffitt, Loeber, & Stouthamer-Loeber, Reference Lynam, Caspi, Moffitt, Loeber and Stouthamer-Loeber2007).

Based on these findings, the stability of CU ratings across development appears to be better or equivalent to other trait indices across childhood to adolescence and early adulthood. However, there is no agreed on metric regarding what level of stability would warrant the classification of a “stable trait.” Critically, even reasonably high levels of stability would not imply that CU traits are unchangeable. For example, Frick, Kimonis, et al. (Reference Frick, Kimonis, Dandreaux and Farrell2003) reported that, despite the high level of stability in these traits across their 4-year study period, there were a significant number of youth who decreased in their level of CU traits over the course of the study (for a similar pattern of change over a longer period of development, see also Lynam et al., Reference Lynam, Caspi, Moffitt, Loeber and Stouthamer-Loeber2007). Further, this decrease in the level of CU traits was related to the level of conduct problems displayed by the child, the socioeconomic status of the child's parents, and the quality of parenting the child received. Thus, CU traits do appear to be at least somewhat malleable and changes in their level across development seem to be influenced by factors in the child's psychosocial environment.

Behavioral genetic studies

There have been a large number of studies showing moderate heritability and nonshared environmental influence, as well as modest shared environmental influence on antisocial behavior (e.g., Mason & Frick, Reference Mason and Frick1994; Rhee & Waldman, Reference Rhee and Waldman2002). In other words, individual differences in susceptibility to antisocial behavior have both a heritable and environmental component. In recent years, there have been several lines of research that have attempted to go beyond this basic partitioning of genetic and environmental influences to antisocial behavior that could be especially important for understanding the different developmental pathways that may lead to severe patterns of antisocial behavior.

First, twin and adoption studies have demonstrated several important types of gene–environment correlations and gene–environment interactions. For example, risk factors that have been traditionally conceptualized as environmental (e.g., parenting reactions) may actually be evoked partly by the heritable temperamental features of the child (gene–environment correlation; Larsson, Viding, Rijsdijk, & Plomin, Reference Larsson, Viding, Rijsdijk and Plomin2008). Further, genetically influenced individual differences in the sensitivity to environmental risk factors such as maltreatment (gene–environment interaction) are also important in explaining variance in antisocial behavior. Specifically, research has suggested that genes regulating serotonergic neurotransmission, in particular monoamine oxidase A (MAOA), are important for explaining individual differences in antisocial behavior (Buckholtz & Meyer-Lindenberg, Reference Buckholtz and Meyer-Lindenberg2008). However, the genetic vulnerability to antisocial behavior conferred by the MAOA low-activity allele (MAOA-L) may only become evident in the presence of an environmental trigger, such as maltreatment (Caspi et al., Reference Caspi, McClay, Moffitt, Mill, Martin and Craig2002; Kim-Cohen et al., Reference Kim-Cohen, Caspi, Taylor, Williams, Newcombe and Craig2006).

Second, and more directly related to the different developmental pathways to antisocial behavior, several studies have begun studying differences in the heritability of antisocial behavior across subtypes. For example, Taylor, Iacono, and McGue (Reference Taylor, Iacono and McGue2000) reported greater genetic influence for childhood-onset delinquency, compared to adolescence-onset delinquency. A recent study by Silberg, Rutter, Tracy, Maes, and Eaves (Reference Silberg, Rutter, Tracy, Maes and Eaves2007) extended these findings with a larger sample of twins. They reported that a single genetic factor was important in influencing antisocial behavior that started in childhood and was persistent into adulthood. They also found that there were adolescent-onset shared environmental factors that were not influencing individual differences in antisocial behavior before adolescence. Finally, they reported a transient genetic effect at puberty and argued that this finding is consistent with a notion of genetically mediated influence on the timing of puberty affecting the expression of genetic differences in antisocial outcomes.

Directly related to the model of the CU traits differentiating subtypes within the childhood-onset group, antisocial behavior in children with CU traits has been found to be under strong genetic influence (heritability of .81) with little influence of shared environment (Viding, Blair, Moffitt, & Plomin, Reference Viding, Blair, Moffitt and Plomin2005). In contrast, antisocial behavior in children without elevated levels of CU shows more modest genetic influence (heritability of .30) and substantial environmental influence (shared environmental influence = .34, nonshared environmental influence = .26). It was important that the differences in heritability could not be attributed to differences in the severity of conduct problems in this sample of 7-year-old twins. These findings were replicated when the children were 9 years old, and this latter study also demonstrated that the difference in heritability magnitude holds even after hyperactivity scores of the children were controlled (Viding, Jones, Frick, Moffitt, & Plomin, Reference Viding, Jones, Frick, Moffitt and Plomin2008).

Thus, these behavioral genetic studies provide strong support for differences in the etiology for the different patterns of antisocial behavior (childhood vs. adolescence onset; CU vs. non-CU), again supporting the importance of recognizing equifinality in causal models. Further, these findings are also consistent with some of the findings on behavioral, cognitive, affective, and social correlates reported previously. For example, the finding of persistent genetic influences being associated with childhood-onset antisocial behavior and new environmental influences coming to play in adolescence is consistent with the emphasis on biological and temperamental vulnerabilities for the early starters and greater emphasis on social causes for the adolescent-onset group. The finding of differences in etiological factors for antisocial behavior in children high and low on CU traits also is consistent with previous findings. For example, stronger environmental influences on antisocial behavior in children who are low on CU traits support those studies reviewed previously showing that parenting factors were more strongly associated with conduct problems in these children (Frick & White, Reference Frick and White2008).

Third, although no molecular genetic studies have directly compared different subtypes of children with conduct problems, some imaging studies have suggested that the MAOA-L risk polymorphism may relate specifically to children who show primarily impulsive and reactive types of conduct problems (for a review, see Buckholtz & Meyer-Lindenberg, Reference Buckholtz and Meyer-Lindenberg2008). Buckholtz and Meyer-Lindenberg (Reference Buckholtz and Meyer-Lindenberg2008) propose that MAOA-L causes an ontogentic excess of serotonin and influences the brain's affective circuitry in a way that amplifies the effects of environmental risk, such as maltreatment. It is interesting to note that some studies have reported an increased vulnerability to antisocial behavior in the presence of the MAOA high (as opposed to low) activity allele (e.g., Manuck et al., Reference Manuck, Flory, Ferrell, Mann and Muldoon2000). Thus, it is possible that different alleles of the same gene may predispose to different types of conduct problems by having opposite effects on the affective lability of an individual. This would be consistent with our developmental model specifying different patterns of emotional responding across subgroups of antisocial youth. However, this possibility is highly speculative at present. As for any psychiatric outcome, the genetic influences on any subtype of conduct problems will not be limited to a single candidate gene. However, these findings also nicely illustrate the concept of multifinality. That is, it is unlikely that specific genes that increase the likelihood for antisocial behavior would conform to diagnostic boundaries. For example, different levels of MAOA activity have been linked to a wide range of psychiatric disorders such as depression, autism, bipolar disorder, and alcoholism (Cicchetti, Rogosch, & Sturge-Apple, Reference Cicchetti, Rogosch and Sturge-Apple2007; Rende & Waldman, Reference Rende, Waldman, Cicchetti and Cohen2006; Shin, Chen, & Ridd, Reference Shin, Chen and Ridd1999). The multifinality of outcomes suggests that any genetic risk conferred by MAOA-L is likely to be moderated by both distinct environmental risk factors and other genes, as already outlined above.

Fourth, twin studies have not only begun to compare genetic and environmental influences in subgroups of antisocial youth, they have also begun to explore whether antisocial behavior, CU traits, and other related traits (e.g., substance abuse) co-occur because of common genetic and/or environmental factors (Krueger et al., Reference Krueger, Hicks, Patrick, Carlson, Iacono and McGue2002; Taylor, Loney, Bobadilla, Iacono, & McGue, Reference Taylor, Loney, Bobadilla, lacono and McGue2003; Viding et al., Reference Viding, Frick and Plomin2007). In much the same way as a comparison of identical and fraternal twins enables the partitioning of variance into genetic, shared, and nonshared environmental components, the twin design can also be used to partition covariance between traits into different etiological components. Such multivariate behavioral genetic analyses are ideal for investigating different phenotypic manifestations of shared genetic or environmental risk. In general, these types of analyses have suggested that the overlap among antisocial behavior and other externalizing symptoms or personality traits (e.g., CU) has a substantial genetic component (e.g., Viding et al., Reference Viding, Frick and Plomin2007). In addition, these studies have found that the nonshared environmental factors (i.e., those environmental risk factors that are not shared by two family members) are often phenotype specific. Such findings suggest that nonshared, phenotype specific, environmental risk factors are critical for differentiating the eventual behavioral outcome, even if the same genetic risk factors are important for two different behavioral patterns. It is also possible that some differentiation of behavioral outcome is because of those risk genes that are not common between two distinct outcomes.

Fifth, behavior genetic studies have begun to address the issue of whether severe antisocial behavior is better thought of as being on a continuum with normal behavior or whether it is better considered a qualitatively distinct taxon (Beauchaine, Reference Beauchaine2003; Vasey, Kotov, Frick, & Loney, Reference Vasay, Kotov, Frick and Loney2005). One possibility is that common behavioral disorders are the quantitative extreme of the same genetic effects that operate throughout the distribution of the trait (Plomin, Owen, & McGuffin, Reference Plomin, Owen and McGuffin1994). In this quantitative trait loci model, several genes (rather than a few rogue mutations) are hypothesized to act in a probabilistic manner to increase risk for the development of maladaptive behavioral outcome. Each gene is thought to account for only a miniscule proportion of variance for any risk outcome (Plomin, DeFries, McClearn, & McGuffin, Reference Plomin, DeFries, McClearn and McGuffin2008). It is thus unsurprising that even when environmental risk factors increase the “penetrance” of genetic risk, the interaction between any specific gene and an environmental risk factor does not account for all of the variance in a maladaptive behavioral outcome (Caspi et al., Reference Caspi, McClay, Moffitt, Mill, Martin and Craig2002). Twin studies have generally documented similar heritability (and environmental) estimates across the continuum of conduct problem scores, as well as at the extreme trait level (e.g., Slutske et al., Reference Slutske, Heath, Dinwiddie, Madden, Bucholz and Dunne1997; Viding et al., Reference Viding, Blair, Moffitt and Plomin2005, Reference Viding, Frick and Plomin2007). For example, Slutske et al. (Reference Slutske, Heath, Dinwiddie, Madden, Bucholz and Dunne1997) found that a multiple-threshold model of severe antisocial behavior fit their data well, suggesting that antisocial behavior seems to be an extreme of the normal variation in conduct problems, rather than a discrete entity. Similar heritability estimates for normal/subclinical and extreme forms of behavior are in line with the quantitative trait loci model.

It is important to note that such a model does necessarily contradict the possibility that there may be different subgroups of antisocial youth with qualitatively different causal processes. That is, heritability and environmental estimates obtained at different cutoffs of any univariate distribution could be based on data from a heterogeneous group of individuals who could have a distinct set of genetic and environmental risk factors. In addition, there may be different normal to abnormal continua in terms of the relative importance of certain genes and environments for different subtypes. Thus, it is still possible that different subtypes of antisocial individuals could have distinct etiologies, even if their vulnerability profile on a single indicator (e.g., specific behavior or neurocognitive measure) is not qualitatively different from the rest of the population.

fMRI Brain imaging studies

Despite the demonstration of genetic influences on individual differences in antisocial behavior, it is important to emphasize that risk genes for antisocial behavior do not act deterministically. Instead, genes code for neurocognitive vulnerabilities that may, in turn, increase a child's risk for antisocial behavior in certain environmental contexts. For example, individual differences in several brain areas and cognitive functions associated with perception and regulation of emotions have been found to correlate with antisocial and violent behavior and could also mediate genetic influences (Davidson, Putnam, & Larson, Reference Davidson, Putnam and Larson2000). In particular, the orbitofrontal/ventromedial prefrontal cortex, anterior cingulate, amygdala, and interconnected regions have shown both structural and functional abnormalities in antisocial populations (Blair et al., Reference Blair, Mitchell and Blair2005).

fMRI is a relatively new tool for studying antisocial behavior in children and, as a result, only a handful of studies exist to date. Hence, to date, there are no fMRI studies contrasting childhood-onset with adolescent-onset subtypes or antisocial youth with and without CU traits. Instead, published studies have either focused on an undifferentiated group of children with antisocial behavior or on the CU subtype alone. However, these studies do provide some at least speculative support for the developmental model that has been outlined in this paper. Specifically, reduced anterior cingulate responsivity to threatening emotional scenes under passive viewing conditions has been reported for antisocial youth, possibly reflecting poor emotional regulation (Stadler et al., Reference Stadler, Sterzer, Schmeck, Krebs, Kleinschmidt and Poustka2007; Sterzer, Stadler, Krebs, Kleinschmidt, & Poutska, Reference Sterzer, Stadler, Krebs, Kleinschmidt and Poutska2005). The same studies also reported amygdala hypoactivation to negative and threatening pictures for antisocial youth, but this result emerged only when anxiety scores were used as a covariate. A more recent study using an almost identical passive viewing paradigm found increased amygdala activation, partly related to comorbid anxiety, in children with antisocial behavior (Herpertz et al., Reference Herpertz, Huebner, Marx, Vloet, Fink and Stoecker2008). These functional deficits would be consistent with some of the abnormalities in how children with CU traits process emotional stimuli.

In support of this possibility, two studies have directly measured CU traits and explicitly tested the hypothesis that this group would show amygdala hyporeactivity to others' distress (Jones, Laurens, Herbs, Baker, & Viding, Reference Jones, Laurens, Herba, Barker and Viding2009; Marsh et al., Reference Marsh, Finger, Mitchell, Reid, Sims and Kosson2008). Both studies employed an implicit emotion processing task (gender recognition) and found amygdala hyporeactivity to fearful faces in antisocial youth with CU traits compared to typically developing children and children with attention-deficit/hyperactivity disorder. Another study documented an abnormal ventromedial prefrontal cortex response to punishment in children with CU traits (Finger et al., Reference Finger, Marsh, Mitchell, Reid, Sims, Budhani and Kosson2008). This study employed a task in which the participants had to learn that a stimulus that used to be rewarding was subsequently associated with loss and that they should stop responding to that stimulus. This finding is consistent with previously reviewed findings showing abnormalities in how youth with CU traits respond to reward and punishment contingencies.

Thus, although much more work in this area is needed, these initial findings from brain imaging studies are promising in providing possible neurological bases to some of the cognitive and affective differences found across the different developmental pathways to antisocial behaviors, and could help in uncovering some of the neurocognitive mechanisms through which inherited predispositions can place a child at risk for the development antisocial traits (Viding & Jones, Reference Viding and Jones2008). Of importance, these promising studies demonstrate the importance of integrating findings across multiple levels of analyses. That is, the unique functional neurological deficits were documented when CU traits were used to form experimental groups and when tasks tapping the specific emotional and cognitive deficits that have been associated with these traits were used to test functional brain differences.

It is important to bear in mind that although our selective review focused on fMRI studies, such studies do not by any means represent the only avenue of investigation for charting neurobiological vulnerabilities that can mediate the genetic influences on antisocial behavior. For example, several studies have shown that youth with CU traits show differences in their autonomic reactivity to certain types of stimuli (Blair, Reference Blair1999; Kimonis et al., Reference Kimonis, Frick, Skeem, Marsee, Cruise and Munoz2008) and they show differences in both their resting (Loney, Butler, Lima, Counts, & Eckel, Reference Loney, Butler, Lima, Counts and Eckel2006) and stress-induced (O'Leary, Loney, & Eckel, Reference O'Leary, Loney and Eckel2007) cortisol levels. Thus, some of the genetic risk for this group of antisocial youth could be due to inherited individual differences in the functioning of their autonomic nervous system or in the reactivity of their hypothalamic–pituitary–adrenal axis as indexed by the hormone cortisol. Combining across all of these levels provides the greatest likelihood of advancing our knowledge of the different pathways to the equifinal outcome of antisocial behavior (Viding & Frith, Reference Viding and Frith2006).