Callous-unemotional traits in predicting the severity and stability of conduct problems and delinquency.

It is becoming increasingly clear that, within the group of children who show severe antisocial and aggressive behavior, there are important subgroups which differ on the severity and stability of their behavior problems (see Frick & Loney, 1999 for a review) and which exhibit distinct characteristics that could suggest different causal processes underlying their behavioral disturbance (see Frick & Ellis, 1999 for a review). Evidence for this heterogeneity within antisocial youth has led to a number of subtyping approaches in an attempt to capture these variations in severity and causes. Most of these approaches have focused on the severity or type of antisocial or aggressive behaviors manifested by the child (Crick & Dodge, 1996; Loeber et al., 1993; Stickle & Blechman, 2002) or on the timing of onset of these behaviors (American Psychiatric Association, 2000; Moffitt, 1993).

A notable exception is research that has attempted to extend the construct of psychopathy to understanding antisocial behavior in youth. The construct of psychopathy, rather than focusing on the severity or pattern of antisocial behavior, focuses on the person's affective (e.g., absence of guilt, constricted display of emotion), interpersonal (e.g., failure to show empathy, use of others for one's own gain), self-referential (e.g., views self as more important than others) and behavioral (e.g., acts in a careless and impulsive manner) style to designate a distinct subgroup of antisocial individuals (Hare 1998; Hart & Hare, 1997). Although much of the research on psychopathy has been conducted in adult samples, the presence of psychopathic features has proven to designate a subgroup of antisocial youth with more severe and more aggressive patterns of antisocial behavior in forensic (Caputo, Frick, & Brodsky, 1999; Kruh, Frick, & Clements, 2005) and mental health (Christian, Frick, Hill, Tyler, & Frazer, 1997) samples. Further, children with conduct problems who also show psychopathic traits show a number of distinct characteristics, such as showing a preference for novel, exciting, and dangerous activities (Frick, Lilienfeld, Ellis, Loney, & Silverthorn, 1999), being less sensitive to cues of punishment, especially when a reward-oriented response set is primed (Barry et al., 2000; Fisher & Blair, 1998), and being less reactive to threatening and emotionally distressing stimuli (Blair, 1999; Loney, Frick, Clements, Ellis, & Kerlin, 2003). In addition, their conduct problems appear to be less strongly associated with dysfunctional parenting practices (Wootton, Frick, Shelton, & Silverthorn, 1997) and with deficits in verbal intelligence (Loney, Frick, Ellis, & McCoy, 1998) compared to children with conduct problems who do not show psychopathic traits.

Unfortunately, this research extending the construct of psychopathy to youth has largely been cross-sectional in nature. As a result, there is limited evidence for the predictive utility of psychopathic features prior to adulthood (Edens, Skeem, Cruise, & Cauffman, 2001). This limitation is unfortunate because one of the most important facets of psychopathy in adult samples is its utility for predicting antisocial outcomes, such as violence and aggression (Hart, Kropp, & Hare, 1988; Serin, 1993; Serin, Peters, & Barbaree, 1990). There are several notable exceptions in which the predictive utility of psychopathic traits has been tested in samples of institutionalized adolescents. These studies have documented that psychopathic features predict subsequent delinquency, aggression, number of violent offenses, and a shorter length of time to violent reoffending in antisocial youth (Brandt, Kennedy, Patrick, & Curtin, 1997; Forth, Hart, & Hare, 1990; Toupin, Mercier, Dery, Cote, & Hodgins, 1995). In one of the only studies to test the predictive utility of psychopathic features in a non-referred sample of children, Frick, Cornell, Barry, Bodin, and Dane (2003) reported that children with conduct problems who also showed psychopathic features showed more severe and more instrumental aggression and had higher rates of self-reported delinquent acts one year later compared to children with conduct problems but without these features. The current study focuses on this same sample of children using additional follow-up data.

These results are promising in suggesting that psychopathic features may designate an especially severe, aggressive, and chronic type of disturbance in antisocial youth. However, there are a number of limitations in this body of research. First, most studies have relied on clinic-referred or forensic samples (except Frick et al., 2003; Lynam, 1997) and focused on the adolescent age group (except Christian et al., 1997; Frick et al., 2003). Second, even the studies that have used prospective designs have had limited follow-up periods, typically with follow-up periods of one (Frick et al., 2003) to two (Brandt et al., 1997; Forth et al., 1990) years. Therefore, the utility of these traits for predicting antisocial behavior over longer periods of time has not been established. Third, it is not clear from this research which dimension or dimensions of the construct of psychopathy might be most important for predicting later antisocial and aggressive behavior.

In both child (Frick, Bodin, & Barry, 2000; Frick, O'Brien, Wootton, & McBurnett, 1994) and adult (Cooke & Michie, 2001; Hare, Hart, & Harpur, 1991) samples, factor analyses of psychopathic features result in multiple correlated dimensions. Further, there has been some debate as to which of these dimensions may be most important for distinguishing antisocial youth who fit more with traditional conceptualizations of psychopathy. For example, some studies have placed primary importance on the presence of impulsivity (Lynam, 1996), whereas others have emphasized the presence of callous and unemotional (e.g., lack of guilt and empathy) (CU) traits (Barry et al., 2000). In support of the latter conceptualization, Barry et al. (2000) reported that clinic-referred children with conduct problems and high levels of impulsivity only showed characteristics associated with the construct of psychopathy (e.g., fearlessness, a reward dominant response style) if they also showed high rates of CU traits. Further, in the sample that is the focus of the current study, Frick et al. (2003) reported that the presence of CU traits, but not impulsivity, predicted greater levels of aggression and particularly greater levels of instrumental and premeditated aggression at a one year follow-up in non-referred children with conduct problems. However, an important piece of information for informing this debate is to determine which of these dimensions more strongly predicts the stability of antisocial and aggressive behavior over a more extended period of time.

Based on these considerations, the main goal of the current study was to extend the findings of Frick et al. (2003) to determine the predictive utility of CU traits over a more extended (four year) follow-up period. Children with conduct problems, half of whom had high rates of CU traits and half of whom did not, were assessed at four yearly assessment points. The utility of CU traits for predicting the stability of conduct problems and delinquency was assessed. Further, the relative utility of these traits for predicting stability after controlling for impulsivity was investigated. Finally, the utility of CU traits was tested, not only for predicting the severity of antisocial behavior, but also for predicting different types of conduct problems (e.g., aggression, property destruction, and oppositional) and delinquent acts (e.g., violence, property offenses, status offenses, and drug offenses).

METHOD

Participants

There were several goals that guided sample recruitment. First, one goal was to obtain a community sample of youth with conduct problems to avoid potential referral biases that might be present in clinic-referred or forensic samples. Second, when obtaining a non-referred sample, it was also important to ensure that enough children with severe conduct problems were recruited and, even more importantly, that sufficient numbers of conduct problem youth with and without CU traits were recruited. Third, it was important to ensure that the oversampled groups were still representative of that group in the community from which they were sampled.

Based on these considerations, a two step stratified random sampling procedure was employed to recruit participants. In the first step, four thousand parents of children in the third, fourth, sixth, and seventh grades of two school systems in a moderate sized city in the southeastern United States received announcements about the study. The two school systems were chosen because one served the immediate urban area and the second served the surrounding region that was predominantly suburban and rural. Those parents who agreed to have their child participate in the study completed consent forms and a screening questionnaire used to assess the presence of DSM-IV symptoms of Oppositional Defiant Disorder (ODD) and Conduct Disorder (CD; American Psychiatric Association, 2000) and CU traits (Frick & Hare, 2001). Following receipt of the parents' consent forms and rating scales, the child's teacher completed analogous questionnaires. For each child who participated in this initial phase of screening, his or her teacher received $ 10 for educational supplies for the classroom. This first phase yielded a sample of 1136 children that was 53% female, 77% Caucasian, 19% African-American, and 21% receiving special education services, all of which closely matched the overall demographics of the two school systems. The range of Duncan's Socioeconomic Index (SEI; Hauser & Featherman, 1977) was 0-92.3, with a mean of 47.20 (SD = 23.8) and with scores of 24 and 64 at the 1st and 3rd quartiles of the sample. This distribution indicates that the sample showed generally a normative range of socioeconomic status, given that middle class categories generally fall between 35 and 65 (higher scores indicating higher status categories) on this index.

In the second phase of recruitment, the sample of 1136 children was divided into four groups based on combined parent and teacher ratings of conduct problem symptoms and CU traits in the full sample. The first group was below the mean on both dimensions (n = 225), a second group was at or above the upper quartile on the conduct problem measure but below the mean on the measure of CU traits (n = 66), one group was at or above the upper quartile on the measure of CU traits but below the mean on the measure of conduct problems (n = 77), and the last group was above the upper quartile on both dimensions (n = 128). These four groups were then stratified on gender, ethnicity, and socioeconomic status. Next, 25 children in each of the four groups were recruited to participate in the four yearly follow-up assessments. These children were selected through a stratified random sampling procedure to ensure that the four groups matched the group in the community sample from which they were sampled on the three stratification variables. Also, the sampling ensured that approximately equal numbers of children were included from the younger (3rd and 4th grade) and older (6th and 7th) grade cohorts. Errors in data collection resulted in the loss of two participants from the group high on conduct problems but low on the measure of CU traits. Therefore, participants in this study were 98 children that fell into the four groups described above.

The screening took place in the Fall and Spring of the 1997-1998 school year and these children were assessed at four yearly follow-up assessments, ending in the fall of the 2002 school year. The mean length of time between the initial screening and the first follow was 12.99 (SD = 4.67) months, with intervals of 12.65 (SD = 1.61), 12.63 (SD = 1.82), and 13.38 (SD = 2.82) months occurring between the subsequent follow-up assessments. This led to an average interval between completion of the screening measures to form the study groups and the fourth follow-up assessment of 50.91 months (SD = 4.4) and an interval of 38.60 months (SD = 2.90) between the first and fourth follow-up assessments. The demographic characteristics of this sample are described in Table I. As evident from this table, there were differences across groups on some demographic variables, which reflect the characteristics of that group in the larger community sample. Specifically, the presence of …