Abstract

Purpose: The extant criminal career literature supports the assertion that risk factors for violent and non-violent offending are the same. However, such studies have not examined the role of psychopathic personality disturbance (PPD) in the development of persistent violence across the life course. A situational action theory perspective was used to help illustrate the utility of PPD in explaining persistent violent offending. Methods: Convictions for violent offense and non-violent offenses were measured for Canadian male (n = 262) and female (n = 64) offenders at each year between ages 12 and 28. Semi-parametric group-based modeling was used to identify joint trajectories of violent and non-violent offending. Symptoms of adolescent PPD and other criminogenic risk factors were also measured. Results: Through the joint trajectory model, five violent and five non-violent trajectories were identified. PPD emerged as a strong predictor of membership in the trajectory associated with chronic violent offending but lower levels of non-violent offending. Conclusions: Contrary to earlier criminal career research, the most persistent violent offenders were not also involved in the highest rate of general offending. Theories that help explain why individuals are involved in persistent violence are needed. Incorporating PPD into such a theoretical framework appears necessary.

Keywords

Criminal trajectories; group-based trajectory modeling, juvenile delinquency, PCL: YV, psychopathic personality disturbance; violent offending

Several criminal career studies have indicated that violent offending and general offending can be explained by the same risk factors (e.g., Capaldi & Patterson, 1996; Farrington, 1989). Not surprisingly, parsimonious theories that view violence as part of a general antisocial tendency have been predominant (Capaldi & Patterson, 1996; DeLisi & Vaughn, 2014; Farrington, 1991, 1998). However, as Hart (1998) argued earlier, explanations of violence that did not consider psychopathic personality disturbance (PPD) were incomplete. Given that measures of PPD have been notably absent in criminal career research (Farrington, 2005; McCuish, Corrado, Lussier, & Hart, 2014), it is premature to conclude that specific explanations of violent offending are unwarranted.

The importance of PPD in predicting violence outcomes is well recognized within the literature on risk assessment. Some have argued that PPD is the single best predictor of violent offending (e.g., Douglas, Vincent, & Edens, 2006; Harris, Rice, & Lalumière, 2001). Consequently, this construct has been included in several violence risk assessment tools, such as the SAVRY (Borum, Bartel, & Forth, 2002), HCR-20 (Webster, Douglas, Eaves, & Hart, 1997), and VRAG (Quinsey, Harris, Rice, & Cormier, 1998). Just as the risk assessment literature can help guide criminal career researchers’ incorporation of PPD as a key covariate of offending trajectories, the criminal career paradigm can help guide risk assessors’ measurement of offending outcomes. Specifically, there is a tendency within the violence risk assessment literature to focus only on the ‘next offense’ (i.e., recidivism outcomes), instead of on the development of violent offending over the life course. From both a theoretical (e.g., Blumstein, Cohen, Roth, & Visher, 1986; DeLisi & Piquero, 2011) and empirical (Lussier & Davies, 2011) perspective, the former approach is misleading as an indicator of the seriousness of an offender. In addition, focusing more narrowly on recidivism outcomes likely also underestimates the strength of the relationship between PPD and violence. Hart (1998) argued that more sophisticated analytic strategies that better accounted for the complexity of offending over time were necessary to adequately capture this relationship. Modeling violent offending trajectories is one method of capturing the complexity of patterns of violence over time (see Brame, Mulvey, and Piquero, 2001; MacDonald, Haviland, & Morral, 2009; Piquero, Brame, Mazerolle, & Haapanen, 2002). Thus far; however, the role of PPD in helping to explain the unfolding of violence trajectories has not been examined.

Retrospective and prospective longitudinal data from a sample of Canadian male (n = 262) and female (n = 64) adolescent offenders incarcerated between 1998 and 2001 were used to model joint trajectories of violent and non-violent offending. The use of an offender-based sample meant that the full range of violent offending and the full range of symptoms of PPD would be accounted for. Symptoms of PPD were measured using the using the Psychopathy Checklist: Youth Version (PCL: YV) to help explain variance in joint trajectories of violent and non-violent offending. The aim of the current study was to bring together one of the most important risk factors for violence according to the risk assessment literature (i.e., PPD), and one of the most comprehensive measures of an individual’s criminal career (i.e., measures of offending trajectories).

Evidence for the Relationship between PPD and Violence

Although not all individuals with PPD are violent, and not all violent offenders have high symptoms of PPD, individuals with PPD are disproportionately involved in violence (Hare & Neumann, 2008; Hart & Hare, 1997; Ribeiro da Silva, Rijo, & Salekin, 2012). The relationship between symptoms of PPD and an earlier time to recidivism has been demonstrated in both youth and adult incarcerated populations (Corrado, Vincent, Hart, & Cohen, 2004; Douglas et al., 2006; Harris; Rice, & Cormier, 1991; Serin, 1996; Vaughn & DeLisi, 2008; Vaughn, Howard, & DeLisi, 2008). Despite being one of the most important individual-level risk factors for violent offending, Vitacco et al. (2006) noted a clear lack of prospective longitudinal studies examining the relationship between PPD and persistent violence. Overall, there has been a general lack of research on the long-term predictive validity of PPD. Recently, McCuish et al. (2014) found that high scores on both the three and four factor models of the PCL: YV were indicative of involvement in chronic general offending from age 12 to 28. Using the same data, Corrado, McCuish, Hart, and DeLisi (2015) found that the influence of PPD symptoms on trajectory membership was maintained after controlling for several important criminogenic covariates. However, in both studies, contrary to expectation the affective and interpersonal facets of the PCL: YV were unrelated to chronic general offending. Corrado et al. (2015) proposed that these prototypical symptoms of PPD (see Cooke, Hart, Cohen, & Michie, 2012; Hoff, Rypdal, Mykletum, & Cooke, 2012; Kreis & Cooke, 2011) were more well-suited to explaining persistent violent offending. The specific mechanisms in which PPD appears to operate on the unfolding of a trajectory characterized by persistent violence is described below.

The Relationship between PPD and Persistent Violence: A Situational Action Perspective

The relatively few violence-specific criminological theories may be due to the assertion that general theories of serious criminality sufficiently explain violent offending too (Capaldi & Patterson, 1996; Farrington, 1991, 1998). Not surprisingly, there are even fewer criminological theories that specify the relationship between personality types, such as PPD, and persistent violence. If a relationship between PPD and persistent violence does exist, theories that help to explain the causal mechanisms responsible this relationship will be critical, as prediction alone cannot sufficiently explain the development of persistent violence (e.g., Laub, 2006). Wikström’s situational action theory of violence (Wikström, 2006; Wikström & Treiber, 2007, 2009), at least potentially, provides a framework for specifying the complex hypothesized relationship between PPD and violence. Although situational action theory is an event-based perspective, Wikström and Treiber’s (2009) description of the conditions that precipitate violent events are conditions that appear consistently present among individuals with PPD.

In situational action theory, the two main conditions facilitating violence are propensity and situational (e.g., environmental) context (Wikström & Treiber, 2009). In reference to propensity, Wikström and Treiber (2009) asserted that an individual’s set of moral rules combined with low levels of self-control increases their propensity to use violence as an action alternative (e.g., as an alternative to walking away or diffusing the conflict). They also argued that situational contexts such as intoxication, provocation, and peer-influence facilitated violent offenses by increasing an offender’s level of disinhibition. In a situational context not conducive to violence, an individual with a high propensity for violence will still offend, provided that external deterrent factors (e.g., presence of police, responsible adults) are absent or not recognized by the offender (Wikström & Treiber, 2009). A high-propensity individual may, therefore, be more likely to be involved in persistent violence than individuals with a low propensity for violence, because the latter are dependent upon specific situational contexts to occur consistently across the life course.

Although testing situational action theory is not the purpose of the current study, through its concepts of propensity, situational context, and deterrence, this theory provides a framework for explaining why individuals with PPD are more likely to be involved in persistent violence. Regarding Wikström and Treiber’s (2009) concept of violence propensity, Gretton, Hare, and Catchpole (2004) noted that adolescent offenders with PPD were characterized by a strong and long-term risk for involvement in violence that distinguished them from other offenders. PPD may also increase the likelihood of situational contexts that are conducive to violence. For example, early research on PPD indicated that individuals with PPD tended to commit violence indiscriminately (e.g., against strangers and persons known to them, against both males and females), and due to both instrumental and reactive motivations. In contrast, individuals without PPD symptoms were more likely to require specific situational contexts, such as a victim previously known to them or an event that elicited a strong emotional response, to facilitate involvement in violence (Hart & Dempster, 1997; Serin, 1991; Williamson, Hare, & Wong, 1987). In effect, the conditions necessary for violence are set at a lower threshold for individuals with PPD. Furthermore, regarding Wikström and Treiber’s (2009) emphasis on factors that may act as deterrents to even high propensity individuals in situational contexts conducive to violence, it is noteworthy that several studies have found that individuals with PPD were less sensitive to the possibility of punishment (Lykken, 1995; Newman, MacCoon, Vaughn, & Sadeh, 2005) and thus factors known to deter other offenders may have less of an impact on individuals with PPD.

Finally, because symptoms of PPD are asserted to be at least moderately stable over time (Lynam, Caspi, Moffitt, Loeber, & Stouthamer-Loeber, 2007; Vachon, Lynam, Loeber, & Stouthamer-Loeber, 2012), from a developmental perspective on violence, it is likely that violence involvement will continue over the life course. In sum, although this is an event-based theory, the situational action theory perspective can help guide the specification of how individuals with PPD have a high risk for violence, have personality profiles that create situational contexts that keep them primed for violence, and personality symptoms associated with a lack of concern for consequences to themselves and others that limits the effectiveness of deterrence. However, there are several conceptual challenges associated with assessing the hypothesized association between PPD and persistent violence.

The Association between PPD and Persistent Violence: Some Conceptual Challenges

The specific risk factors associated with persistently violent offenders are relatively unknown, in part because this type of offender is rarely found within the types of samples typically examined within criminal career research (Farrington, 1997; Piquero et al., 2002). Given the low prevalence of both PPD and persistent violence in general population samples, and even in general delinquency/criminal samples, identifying risk factors for persistent violence likely requires research using adjudicated samples with sufficient base-rates of both PPD symptoms and violence (DeLisi, 2001; McCuish et al., 2014). By using a sample of formerly incarcerated serious and violent young offenders whose offending histories were coded from age 12 to 28, the current study was unaffected by low base-rate concerns.

As another conceptual challenge, many argue that violent offending occurs within the context of a versatile criminal career characterized primarily by non-violent offending (Barnes, 2014; Doherty & Ensminger, 2014; Farrington, Snyder, & Finnegan, 1988; Loeber et al., 2008; Weiner, 1989). If persistent violent offenders are simply chronic general offenders, then any relationship between PPD and persistent violence may simply reflect the relationship between PPD and general offending. Controlling for an offender’s involvement in non-violent offending is needed before making conclusions about the relationship between PPD and persistent violence. One way to control for involvement in non-violent offending is through joint trajectory modeling, which is an extension of the traditional semi-parametric group-based model, and can be used to facilitate the simultaneous modeling of violent and non-violent offending trajectories (Piquero, Jennings, & Barnes, 2012).

Using joint trajectory modeling, Brame et al. (2001) indicated that most persistent violent offenders were also chronic non-violent offenders. However, Brame et al. (2001) placed constraints on their model, which required individuals in a specific violence trajectory to also be assigned to a specific non-violent trajectory. This may have artificially inflated the degree of concordance between offenders belonging to high-violence/high-non-violence trajectories (see MacDonald et al., 2009). Piquero et al. (2002) allowed trajectories of violent and non-violent offending to be measured independently in a sample of serious young offenders and found an imperfect concordance between violent and non-violent trajectories. However, in this latter study, negative life circumstance and other covariates examined were not helpful in distinguishing offenders associated with different trajectory groups. This second validity issue was addressed by using joint trajectory modeling to simultaneously estimate violent and non-violent offending trajectories and by including PPD as a covariate expected to predict chronic violent but not chronic non-violent offending.

Methodology

Sample

Data for the current study were derived from the Incarcerated Serious and Violent Young Offender study conducted in British Columbia, Canada. As part of this study, adolescent offenders between the ages of twelve and nineteen were interviewed in open and secure custody facilities within the Greater Vancouver Regional District and surrounding areas. Focus within the current study was on the sub-sample of offenders (n = 326) that had been assessed using the PCL: YV as part of the Incarcerated Serious and Violent Young Offender study. With the exception of seven percent of the sample who were between age 25 and 27 at the time of data collection, violent and non-violent convictions for all offenders were coded until age 28. The sample is overwhelmingly composed of male (80.4%) and Caucasian (60.9%) offenders. On average, offenders were approximately 16 years old at the time of their assessment (see Table 1 for sample characteristics).

Procedure

The purpose of the Incarcerated Serious and Violent Young Offender study was to conduct interviews with juvenile offenders and collect file-based information on risk factors associated with adolescent criminal activity and the continuation of this activity into adulthood. To recruit research participants, informed consent was first provided by the British Columbia Ministry of Child and Family Development (MCFD). MCFD serves as the legal guardian to all youth in custody, and their consent allowed the research team to approach all youth in custody centers throughout the province of British Columbia. Youth were approached while they were incarcerated and asked whether to participate. Specific procedures involved in recruitment have been discussed at length elsewhere (see McCuish et al., 2014; Corrado, et al., 2015).

Measures

Ethnicity and gender were measured through self report interviews. Although some offenders in the current study were in their early thirties, criminal trajectories were only measured to age 28 and therefore it was unnecessary to control for age in subsequent analyses. The primary focus within the current study was on whether symptoms of adolescent psychopathic personality disturbance (PPD), controlling for other criminogenic factors, were associated with persistent violent offending, controlling for involvement in non-violent offending. In addition to PPD, seven domains of risk factors were examined: substance use, school behavior issues, abuse experiences, sexual activity, personality development, residential mobility, and aggression. PCL: YV scores and all criminogenic risk factors were measured at the time of the subject’s interview during their incarceration in adolescence (See Table 1).

--Insert Table 1 about Here--

Psychopathy Checklist: Youth Version (PCL: YV; Forth, Kosson, & Hare, 2003).1 The PCL: YV is a symptom rating scale that ranges from 0-2 and is scored using information from a 60-90 minute semi-structured interview as well as a review of file-based collateral information. Access to file information, in addition to interviews, were used to score the PCL: YV. Inter-rater reliability was not conducted in this particular study; however, Vincent (2002) evaluated inter-rater reliability in a subsample of 30 randomly selected cases and the intraclass correlation coefficient was high (ICC₁ = 0.92). The 20 items comprising the PCL: YV were identified as the fundamental personality and behavioral traits believed to represent the construct of PPD in adolescence. These 20 items are asserted to represent different facets of the underlying psychopathy construct, though the appropriate number of facets has been debated. Forth et al. (2003) recommended using a four factor model that consists of an interpersonal factor, an affective factor, a lifestyle factor, and an antisocial factor. Cooke and Michie (2001) recommended a three-factor model that excludes Forth et al.’s (2003) antisocial factor to avoid using prior criminal behavior to predict future criminal behavior. Total scores, factor scores, and individual factors are presented in Table 1. Approximately 30% of males and 34% of females scored what could be considered ‘high’ on the PCL: YV (25 or higher). Independent sample t-tests indicated that total PCL: YV scores did not significantly differ between males and females (p > .05).

Criminogenic Risk Factors. Substance use included separate measures of the age of onset of alcohol and drug use as well as eight dichotomized items (alcohol, marijuana, hallucinogens, ecstasy, cocaine, heroin, crack cocaine, and crystal meth) used to create an aggregate scale of self-reported substance use. Based on tetrachoric ordinal alpha values, which are appropriate for scales comprised of dichotomous items (Gadermann, Guhn, & Zumbo, 2012), scale reliability was high (0.88). School behavior issues included the age at which they began getting into trouble at school, the age at which they started skipping school, the number of times that they changed schools, and whether they were attending school prior to their incarceration. Abuse experiences included dichotomous self-report measures of whether the youth had experienced physical abuse and sexual abuse. Sexual activity was measured using one item on the age of onset of consensual sexual activity. Personality development was measured using Schneider’s (1990) Good Citizen’s Scale, a self-report inventory of 15 identity traits coded on a 1-7 scale (Cronbach’s alpha = 0.74), with items reverse-coded so that lower scores indicated a negative identity. Aggression was assessed by asking participants about the frequency of their involvement in physical fights, whether the participant felt they got angry easily, and whether the participant reported that someone had told them they had a bad temper. To measure familial delinquency and disruption, participants were asked to report whether any members of their biological parents or biological siblings had trouble with alcohol or drugs, had experienced physical or sexual abuse, had a criminal record, or had mental illness. These six items were aggregated into a global scale (tetrachoric ordinal alpha = 0.78). Residential mobility measured whether the participant had left home willingly for more than a day to live somewhere else, whether the participant had been kicked out of their home for more than a day, whether the participant was raised by their biological parents, and whether the participant lived in foster care or other forms of ministry care.

Measures of Offending. All measures of offending were based on official data from British Columbia Corrections’ computerized system, Corrections Network (CORNET), which contains information on an offender’s movement in and out of custody as well as the exact criminal offense, date of conviction, and sentence type received. CORNET data includes only offenses committed within the province of British Columbia. The primary focus within the current study was on examining violent criminal careers. Using data from CORNET, every violent criminal charge that resulted in a conviction was coded for the entire sample from age 12, the age of criminal responsibility in Canada, to age 28. A violent offense was defined as any offense that involved physical contact or use of a weapon to threaten physical harm. Uttering threats was not included in the operationalization of a violent offense. Types of violent offenses in this study included assault, assault with a weapon, aggravated assault, and manslaughter/murder. As an indication of the degree of violence among this sample, 20 offenders (6.1%) had been charged with murder or manslaughter during the study period. For this sample, the average number of non-violent charges for which the individual was convicted for was 20.51 (SD = 16.55). The median number of non-violent convictions was 17.00, showing that the higher number of convictions was not an artifact of a small subgroup of individuals. The vast majority of the sample (84.0%) had been convicted of a violent offense and offenders averaged nearly three convictions. Total time spent in custody was also calculated in order to control for exposure time in the semi-parametric group-based model. On average, offenders spent 1,166 days in custody (SD = 1,167). The distribution of violent and non-violent offending at each age is presented in Figure 1.

--Insert Figure 1 about Here--

Analytic Strategy

Semi-parametric group-based modeling (SPGM) developed by Nagin and Land (1993) was used to identify the number and shape of violent and non-violent offending trajectories that best fit the data. Analyses were conducted in SAS 9.4 using the Proc TRAJ add-on developed by Jones, Nagin, and Roeder (2001). Separate trajectories of violent and non-violent offending were modeled simultaneously using the joint trajectory modeling extension (Nagin & Jones, 2007). Trajectories were measured using all violent and non-violent convictions incurred between age 12 and 28. Eleven offenders died (3.4%) and six (1.8%) moved outside the province. Convictions for these offenders after the age of death or move were coded as missing. Unlike cluster analysis and other grouping methods that identify groups ex ante, the SPGM method allows developmental trajectories to emerge from the data (Nagin, 2005). To control for time at risk, exposure time was built into the SPGM model by adapting Piquero et al.’s (2001) original formula. This adaptation adjusted for high standard errors and improbable rates of offending by inflating the minimum exposure time2 to a value of 0.2:

Exposure_ji = 1 - (Number of Days Incarcerated/455)

where j is the respondent and i is the year of observation.

In SPGM, the functional form of the trajectories is specified to estimate the distribution of offenses over age. Quadratic functional form specifies a more parsimonious distribution that captures one major change in the patterning of offending. Cubic functional form specifies a more complex distribution that captures two major changes in the patterning of offending (Bushway, Thornberry, & Krohn, 2003). To illustrate, if a trajectory was marked by a steep decline followed by a steep increase, a model that specified quadratic functional form would only adequately capture the steep decline. Cubic functional form seemed more appropriate for the current study. When modeling general offending trajectories, if an offender is involved in a period of chronic property offending followed by a period of chronic violent offending, both offending patterns are captured by the general offending trajectory. In contrast, if this scenario occurred when modeling violent offending trajectories, chronic property offending would be equivalent to non-offending. A model with quadratic functional form thus would not capture the increase in violent offending after a period of frequent non-violent offending. Although this scenario seems highly specific, offenders from the Philadelphia Birth Cohort typically transitioned from early versatility to greater specialization (Piquero, Paternoster, Mazerolle, Brame, & Dean, 1999). Ensuring that these transitions were captured by the SPGM was critical to accurately describing both violent and non-violent trajectories among the sample.

After identifying the number and shape of violent and non-violent offending trajectories that best fit the data, the association between violence trajectories, symptoms of PPD, and criminogenic factors were examined in a series of bivariate analyses. All significant criminogenic risk factors and measures of PPD were then included in a series of multinomial logistic regression analyses, controlling for the non-violent offending trajectories, to examine whether these factors helped predict a particular course of violent offending.

Results

Model Identification and Interpretation

The first stage of the SPGM analysis involved identifying the number and shape of violent and non-violent offending trajectories that best fit the data. A zero-inflated Poisson (ZIP) model with cubic functional form was used to estimate the distribution of violent and non-violent offending trajectories. Bayesian Information Criteria (BIC) values were used to identify the number of trajectories that best represented the data. Similar to prior studies examining joint trajectories, the same number of violent and non-violent trajectories were specified for each model (e.g., Brame, Mulvey, & Piquero, 2001). A five trajectory-group model resulted in a BIC value of -10233, which was closer to zero than both a four group model (BIC = -10255) and a six group model (BIC = -10421). BIC values for a five group solution with quadratic functional form were also not closer to zero (BIC = -10378), despite being the more parsimonious model. To further examine the fit of the five group cubic model, Jeffrey’s scale of evidence based on the Bayes factor approximation was used to determine whether there were substantive differences in BIC values between models specifying a different number of trajectories (see Nagin, 2005). The Bayes factor is calculated as e ^{BICi – BICj} where values of B_ij greater than ten indicate strong evidence for model ‘i’ according to the Jeffrey’s scale (Nagin, 2005). The five group model was retained as there was strong evidence for this model over both a four group model and a six group model (B_ij >10). The parameters of the five group model are outlined in Table 2. Classification accuracy, based on the average posterior probabilities of accurately assigning individuals to a particular trajectory, was good for each of the five violent trajectories (range 0.79-0.92) and non-violent trajectories (range 0.87-0.94). Odds of correct classification (OCC) values, which are a more conservative estimate of trajectory assignment compared to average posterior probabilities, were calculated as:

OCC_g = (AvePP_g/ (1-AvePP_g)) / (∏_g/ (1-∏_g))

where ∏_g is the estimated size of group g (see Skardhamar, 2010).

OCC values for the five violent trajectories ranged from 3.7-11.4 and the values for the five non-violent trajectories ranged from 6.6-15.4 (see Table 2). Nine of the ten OCC values were higher than both Nagin (2005) and Skardhamar’s (2010) recommendation that values of at least five be interpreted as an indicator of excellent classification accuracy.

--Insert Table 2 about Here--

Figure 2a presents the violent trajectory model and Figure 2b presents the non-violent trajectory model. Beginning with the violent trajectories, a bell-shaped trajectory (23.6% of the sample) represented a group of offenders who were involved in violent offending at a very low rate and for only a short period between age 12 and 28. For this group, violence peaked at 16 and reached a near-zero rate of offending by age 20. A group of early-onset fast desisters (EOFD; 30.7% of the sample) had the second highest rate of violent offending between age 12-14. However, by age 17, this group had the lowest rate of violent offending of the five trajectories. A stable persister group (23.0% of the sample) peaked at age 15, albeit at a relatively low rate of violent offending, but maintained this rate through age 28. A high-rate chronic group (HRC; 5.8% of the sample) averaged a higher rate of violent offending at age 12 than the bell-shaped, EOFD, and stable persister trajectories at their highest rate of offending from age 12-28. However, a sharp decline in violent offending was observed for the HRC group as they entered late adolescence and early adulthood. Through the mid-twenties; however, fluctuations in rate of violent offending were observed. Finally, the high-rate slow desister (HRSD) trajectory (16.9% of the sample) had the highest rate of offending for any particular year compared to all other trajectories (see age 16 in Figure 2a). By early adulthood, the rate of violence of this group was relatively low, but remained stable through age 28. The HRC and HRSD trajectories represented two chronically violent offending patterns, but the trajectories took on different shapes.

--Insert Figure 2a about Here--

The shapes of the non-violent trajectories (Figure 2b) were highly similar to the shapes of the general offending trajectories identified in the study by Corrado et al. (2015) that used the same sample of offenders. A low-rate group (24.8% of the sample) never averaged more than one non-violent conviction through the study period. By age 18-19, this group reached a near-zero rate of offending. A bell-shaped non-violent trajectory (27.3% of the sample) also reached a near-zero rate of offending by age 18-19, but unlike the low-rate group, this group averaged at least one non-violent conviction between age 13 and 17. A stable persister group (21.5% of the sample) mirrored the shape of the violent stable persister trajectory, but non-violent offenses were committed at a higher rate. Similar findings were observed for the high-rate chronic (HRC) non-violent trajectory (12.6% of the sample), which mirrored the HRC violent trajectory, but non-violent convictions were committed at a higher rate over each year. Finally, a slow-rising chronic (SRC) trajectory was observed (13.8% of the sample) that continued to offend through adulthood at a high rate. As the primary interest in the current study was on violent offending trajectories, subsequent analyses were focused on this portion of the joint trajectory analysis.

--Insert Figure 2b about Here--

Association between Violent and Non-violent Trajectories

Of the trajectories in Figures 2a and 2b, four were interpreted as indicative of ‘chronic’ offending: the HRC and HRSD trajectories in Figure 2a, and the HRC and SRC trajectories in Figure 2b. Chi-square analyses were used to examine whether chronic violent offenders were simply chronic general offenders. As can be seen at the bottom of Table 2, individuals assigned to the two chronic violent trajectories were not necessarily also assigned to the two chronic non-violent trajectories. Indeed, less than 50 percent of violent HRC and HRSD offenders were in the non-violent HRC and SRC trajectories. The majority of chronically violent offenders were assigned to the stable persister non-violent trajectory. It was also clear that many offenders associated with chronic non-violent trajectories were not also associated with either of the two chronic violent trajectories.

To reflect this finding in subsequent analyses, the trajectories of violent and non-violent offending were combined to create four joint trajectories. The first group, referred to as the Low Violence/Low Non-Violence (Low-V/Low-NV) trajectory (60.7% of the sample), included offenders associated with both a low rate violent offending trajectory (i.e., one of the bell-shaped, EOFD, and stable persister trajectories) and a low rate non-violent offending trajectory (i.e., one of the low-rate, bell-shaped, and stable persister trajectories). The second group, referred to as the Low Violence/High Non-Violence (Low-V/High-NV) trajectory (16.6% of the sample), included offenders associated with one of the low rate violent offending trajectories as well as one of the high rate non-violent offending trajectories (i.e., one of the HRC or SRC trajectories). The third group, referred to as the High Violence/Low Non-Violence (High-V/Low-NV) trajectory (12.9% of the sample), included offenders associated with one of the high rate violent trajectories (i.e., one of the HRC and HRSD trajectories) as well as one of the low rate non-violent offending trajectories. The fourth group, referred to as the High Violence/High Non-Violence (High-V/High-NV) trajectory (9.8% of the sample), included offenders associated with both a high rate violent trajectory and a high rate non-violent trajectory.

Several criminal career parameters differentiated the four joint trajectories (see Table 3). The three groups comprised of at least one chronically violent or non-violent trajectory (Low-V/High-NV, High-V/Low-NV, or High-V/High-NV) all averaged a significantly greater number of days in custody compared to the trajectory group without any chronically violent/non-violent offenders (Low-V/Low-NV). The High-V/Low-NV and High-V/High-NV trajectories also averaged a significantly earlier age of onset of offending compared to the Low-V/Low-NV trajectory. The Low-V/High-NV trajectory averaged the most non-violent convictions, significantly more than the Low-V/Low-NV and High-V/Low-NV trajectories. In contrast, the High-V/Low-NV trajectory averaged the most violent convictions, significantly more than the Low-V/Low-NV and Low-V/High-NV trajectories. This group also had the highest prevalence of continuity of violent offending from adolescence to adulthood. Very importantly, the Low-V/High-NV trajectory averaged a significantly greater number of general convictions compared to the High-V/Low-NV trajectory, showing that high rate violent offenders were not necessarily the most frequent general offenders. Significant differences between joint trajectories also emerged regarding demographic characteristics, symptoms of PPD, and criminogenic risk factors.

--Insert Table 3 about Here--

Joint Trajectories, PPD, and Criminogenic Risk Factors

Initial bivariate comparisons were made between the joint trajectories (see Table 4). Males were significantly more likely than females to be in the High-V/Low-NV, Low-V/High-NV, and High-V/High-NV trajectories. Non-Caucasian/non-Aboriginal offenders were the least likely to be involved in the High-V/Low-NV, Low-V/High-NV, and High-V/High-NV trajectories. Most importantly, symptoms of PPD, based on the four factor model of the PCL: YV, were significantly higher among the three joint trajectories comprised of at least one chronically violent or non-violent trajectory compared to the Low-V/Low-NV trajectory. However, only the High-V/Low-NV trajectory had significantly higher scores on the three factor PCL: YV model compared to the Low-V/Low-NV joint trajectory. Equally important, the High-V/Low-NV trajectory also had significantly higher scores on the affective factor of the PCL: YV compared to the Low-V/Low-NV trajectory. Scores on the antisocial factor of the PCL: YV were significantly higher among the three joint trajectories comprised of at least one chronically violent or non-violent trajectory compared to the Low-V/Low-NV trajectory

Different domains of criminogenic risk factors emerged as important for different joint trajectories. Risky lifestyles and local life circumstances seemed to best characterize the High-V/High-NV trajectory. This group, compared to the Low-V/Low-NV trajectory, had an earlier onset of both skipping school and sexual activity, as well as the highest prevalence of fighting on a weekly basis. The High-V/High-NV trajectory also had the highest scores on the family disruption scale, though differences between the other groups only trended toward significance (p < .10). The Low-V/Low-NV and Low-V/High-NV trajectories did not differ on any risk factor measures. Scores on the scale measuring positive identity were significantly higher for the Low-V/Low-NV trajectory compared to the High-V/Low-NV trajectory. Interestingly, the trajectories that included chronically violent or chronically non-violent offenders were not necessarily characterized by a greater number of risk factors or a higher intensity of certain risk factors. For example, the Low-V/Low-NV group had the highest prevalence of sexual abuse experiences compared to other groups, although differences between groups only trended towards significant.

--Insert Table 4 about Here--

Covariates of Joint Violent and Non-Violent Trajectories

Demographic characteristics, measures of PPD, and all significant criminogenic factors from Table 3 were entered into a series of multinomial logistic regression (MLR) analyses. This allowed for an examination of whether PPD predicted involvement in the two chronically violent joint trajectories, controlling for other important demographic and criminogenic factors (see Table 5). Given the controversy surrounding the theoretically appropriate number of factors underlying the PCL: YV (Cooke & Michie, 2001; Forth et al., 2003), three separate models were conducted to examine the predictive utility of the four factor structure, three factor structure, and the four individual factors: interpersonal, affective, lifestyle, and antisocial. Multicollinearity was not an issue as all correlations between the covariates included in the models were lower than 0.400. Moreover, when all covariates were entered into a linear regression model predicting frequency of violence, variance inflation factor values were all less than two. Gender was removed from the three models because of low base-rates of females in the chronically violent trajectories. Ethnicity was dichotomized as Caucasian or non-Caucasian due to the low base rate of non-Aboriginal and non-Caucasian offenders in the High-V/High-NV trajectory.

With the Low-V/Low-NV trajectory as the reference category, all three models were statistically significant (see Table 5). Regardless of whether a four factor model, three factor model, or individual factors were examined, scores did not differ between the Low-V/Low-NV trajectory and the Low-V/High-NV trajectory. In effect, symptoms of PPD were unrelated to being a chronic offender if this type of offender was not associated with a trajectory characterized by chronic violent offending. In Model 1, controlling for other criminogenic risk factors, the PCL: YV four factor model significantly increased the odds of membership in the High-V/Low-NV (OR = 1.16) and High-V/High-NV (OR = 1.17) trajectories compared to the Low-V/Low-NV trajectory. In Model 2, controlling for other criminogenic risk factors, the PCL: YV three factor model significantly increased the odds of membership in the High-V/Low-NV (OR = 1.18), but not the High-V/High-NV trajectory, compared to the Low-V/Low-NV trajectory. In Model 3, when other criminogenic risk factors were controlled for, only the antisocial factor was significant. Scores on this factor significantly increased the odds of being in the High-V/High-NV trajectory compared to the Low-V/Low-NV trajectory (OR = 2.36). The lack of a relationship between other PCL: YV factors and combined trajectories may have been due to shared variance between the four factors. When the interpersonal, affective, lifestyle, and antisocial factors were entered separately, controlling for the other criminogenic factors in the model, multiple differences emerged (not shown). Specifically, higher scores on the interpersonal and affective factors increased the odds of membership in the High-V/Low-NV trajectory compared to the Low-V/Low-NV trajectory (OR = 1.34, 1.37, and 1.37, respectively). In addition to symptoms of PPD, two criminogenic risk factors remained significant in the three models. In each of the three models, a one unit increase in scores on the positive self-identity scale were associated with a six percent decrease in the odds of membership in the High-V/Low-NV trajectory (OR = 0.94). An earlier onset of skipping school was also associated with increased odds of being in the Low-V/High-NV trajectory and the High-V/High-NV trajectory relative to the Low-V/Low-NV trajectory.

--Insert Table 5 about Here--

Discussion

Research on offending trajectories is quite common (Jennings & Reingle, 2012; Piquero, 2008), as is research on the relationship between psychopathic personality disturbance (PPD) and offending (DeLisi, 2005, 2009; Edens, Skeem, Cruise, & Cauffman, 2001; Gretton et al., 2004; Hare, McPherson, & Forth, 1988; Salekin, 2008). Yet, the former research often does not examine risk factors underlying trajectories, and the latter research is typically concerned with violent re-offending rather than long term patterns of violence. The purpose of the current study was to help merge these two lines of empirical study by examining whether PPD and other covariates were associated with long term patterns of violence. Using the joint trajectory modeling extension for Proc TRAJ (Nagin & Jones, 2007), five violent and five non-violent trajectories were identified, and then these trajectories were combined to create joint trajectories of violent and non-violent offending. Of the two violence trajectories characterized by a high level of violence, one was also associated with a high rate of non-violent offending (the High-V/High-NV trajectory), whereas the other high rate violence trajectory was associated with a lower rate of non-violent offending (the High-V/Low-NV trajectory). Thus, in contrast to some earlier assertions (Capaldi & Patterson, 1996; Farrington, 1991, 1998), being a chronic violent offender did not necessarily imply chronic general offending.

For offenders associated with a chronic violent trajectory, distinguishing this group on the basis of their involvement in non-violent offending was very important for understanding the relationship between symptoms of PPD, measured using the PCL: YV, and violent offending. As another indication of the need for specific explanations of persistent violent offending, the association between PPD and persistent violence was not simply due to violent offenders having a tendency to be involved in a high rate of general offending. Results from the multinomial logistic regression analysis indicated that scores on the PCL: YV three factor model were significantly higher for the High Violence/Low Non-Violence trajectory compared to the Low Violence/Low Non-Violence trajectory. Equally important, scores on the PCL: YV three factor model, which excludes the antisocial factor, did not differentiate offenders in the High-V/High-NV trajectory from offenders in the Low-V/Low-NV trajectory. This finding was consistent with Corrado et al.’s (2015) postulation that PPD is a better indicator of persistent violence than of a high rate of general offending. In effect, the high overall PCL: YV scores that were observed for the High-V/High-NV trajectory were likely driven by the antisocial factor, implying a tautological concern, where past involvement in violence (measured by the antisocial factor) predicted involvement in future violence. Thus, offenders with higher symptoms of PPD seemed to have a specific proclivity for involvement in violent but not non-violent offending (i.e., the High-V/Low-NV trajectory), which is further support for the need for specific theories, models, and explanations of persistent violent offenders. To help move the field beyond prediction of persistent violence and toward explanation, situational action theory is revisited in the next section, and specific symptoms of PPD are linked to key concepts from this theory.

Explaining why Persistent Violence Occurs: The Role of Specific Symptoms of PPD

Although it is clear that situational action theory is an event-based theory of violence, as Cullen (2011) asserted, more research is needed regarding the nexus between propensity and opportunity. PPD seems to be a particularly useful covariate for illustrating this connection. Specifically, the three core conditions of this event-based theory, propensity, low deterrence, and situational context, (Wikström & Treiber, 2009), are all seemingly influenced by individual-level characteristics (i.e., PPD). In Figure 3, the symptoms of PPD that are hypothesized to influence the presence and magnitude of these three conditions are outlined. Symptoms were identified from prototypicality studies aimed at identifying the core features of PPD (Cooke et al., 2012; Hoff et al., 2012; Kreis & Cooke, 2011). Regarding individual propensity, a sense of entitlement and intolerance towards others may provide conceptual grounds for the moralistic component of propensity as described by Wikström and Treiber (2009). Behavioral styles associated with impulsivity, disruption, recklessness, and aggression were identified, conceptually, as symptoms that may increase the low self-control component of propensity (Wikström & Treiber, 2009). In terms of situational context, it was hypothesized that an interpersonal style characterized by domineering, manipulative, and antagonistic symptoms would have such an effect on an individual’s environment as to create conditions conducive to violence. Cognitive deficits such as inflexibility and suspiciousness, and emotional deficits such as a lack of emotional stability were also included, as these symptoms may be associated with poor coping strategies even when the degree of conflict in an individual’s environment is low. Finally, individuals with sensation seeking tendencies, a sense of invulnerability, and a lack of forethought are unlikely to recognize external forms of deterrence. Instead, these symptoms allow offenders to act on offending opportunities without hesitation. Moreover, deterrence is likely to be low for individuals that are unconcerned with consequences. Therefore, affective deficits (e.g., being detached, uncommitted uncaring, unempathic, and uncommitted) and emotional deficits (e.g., lack of anxiety and lack of emotional depth) were specified as being conceptually related to low levels of deterrence.

--Insert Figure 3 about Here--

It is important to re-emphasize that the specification of these symptoms is purely conceptual and also should not be considered an exhaustive list. To test situational action theory, event-based data is needed. This means focusing on why a specific offense occurred, as opposed to examining an individual’s frequency of violent offending over the life course. Thus, nothing in the current data could have been used to test this theory. The purpose here was simply to demonstrate how symptoms of PPD appear particularly conducive to involvement in persistent violence given their association with the propensity, deterrence, and situational contexts described in situational action theory (Wikström & Treiber, 2009). If this conceptual specification of the relationship between PPD, situational action theory conditions, and persistent violence is to be successful examined in the empirical literature, some methodological issues associated with the current study need to be highlighted and addressed in future research.

Limitations and Future Research

The current study as well as prior research (e.g., Corrado et al., 2004; Gretton et al., 2004; McCuish et al., 2014; Walters, 2003) failed to identify a relationship between the core interpersonal and affective symptoms of PPD as measured by the PCL: YV and different offending outcomes. Although the PCL: YV is considered the ‘gold standard’ of the assessment tools, perhaps the standard is not set high enough and thus it is necessary to move towards an improved measure of PPD. Results of several prototypicality studies have indicated a much broader conceptualization of PPD than the PCL: YV’s 20 items (Cooke et al., 2012; Hoff, et al., 2012; Kreis & Cooke, 2011). Being over-inclusive of the symptoms of PPD, rather than assuming that just 20 items comprise this construct, may lead to a fuller understanding of the relationship between PPD and offending. Future research in this area may also benefit from analytic strategies other than semi-parametric group-based modeling (SPGM).

Although SPGM can account for exposure time, in offender samples, the combination of a high rate of offending with lengthy/frequent periods of incarceration leads to inordinate estimations of the rate of offending over each year. As such, it is necessary to artificially inflate the amount of time that individuals are exposed to the community to avoid model estimation issues (see van der Geest, Blokland, & Bijleveld, 2009). Although this procedure is unlikely to impact trajectory membership of high rate offenders that enter and exit custody with great frequency, the rate of offending for individuals involved in the most serious offenses is likely under-estimated. For example, a homicide offender may be sentenced to ten years in custody, but through procedures that inflate exposure time, is actually treated as spending several months in the community each year without committing a new offense. Unless the homicide offense occurred at the end of this offender’s criminal career, they will most likely be associated with a lower-rate trajectory. Individuals with PPD and other high-risk individuals are perhaps most likely to be identified as this type of ‘false desister’ since they are most likely to be involved in serious offenses and spend more time in custody. Using PPD to predict involvement in chronic offending trajectories will likely result in lower than expected effect sizes. In effect, a limitation of SPGM is that it is incorrect to assume that the analysis will classify the most serious offenders as chronic offenders. Some of these serious offenders, regardless of their level of risk, will be associated with low-rate trajectories because of an inability to account for the full period of time that they are incarcerated. Additional analytic strategies that involve the construction of ‘seriousness’ metrics may be helpful in this regard. For example, the length of time an offender spends incarcerated at each year could be measured, not simply to control for inopportunity, but as an indication of the severity of an individual’s criminal career.

This sample of Canadian incarcerated adolescent offenders is very specific, and likely cannot be generalized to non-incarcerated populations. At the same time, the type of research questions addressed by this study likely could not be accomplished with a population-based sample. A continuing problem in trajectory studies is the exclusion of female offenders, or a failure to identify chronic female offenders. Perhaps it is time to treat female offenders as their own group and analyze their trajectories separate from males. Despite the limitations of the current study, this line of analysis seemed important given that very little is known about chronic violent offenders despite the serious harm they cause. It is not sufficient to simply understand that individuals with PPD are involved in violence. A deeper understanding of the symptoms of PPD contributing to different mechanisms associated with violence (e.g., propensity, situational context, deterrence). The CAPP model of PPD may provide the complexity needed to specify the symptoms associated with different mechanisms involved in persistent serious and violent offending over the life course (see Corrado, DeLisi, Hart, & McCuish, 2015).

References

Barnes, J. C. (2014). Catching the really bad guys: An assessment of the efficacy of the US criminal justice system. Journal of Criminal Justice, 42(4), 338-346.

Blumstein A., Cohen J., Roth J., Visher C.A. (1986) Criminal careers and ‘‘career criminals’’. Report of the National Academy of Sciences Panel on Research on Criminal Careers; Washington. Washington, DC: National Academy Press.

Borum, R., Bartel, P., & Forth, A. (2002). Manual for the Structured Assessment for Violence Risk in Youth (SAVRY): Consultation version. Tampa: University of South Florida: Florida Mental Health Institute.

Brame, R., Mulvey, E. P., & Piquero, A. R. (2001). On the development of different kinds of criminal activity. Sociological Methods & Research, 29(3), 319-341.

Bushway, S. D., Thornberry, T. P., & Krohn, M. D. (2003). Desistance as a developmental process: A comparison of static and dynamic approaches. Journal of Quantitative Criminology, 19(2), 129-153.

Capaldi, D. M., & Patterson, G. R. (1996). Can violent offenders be distinguished from frequent offenders: Prediction from childhood to adolescence. Journal of Research in Crime and Delinquency, 33(2), 206-231.

Cooke, D. J., & Michie, C. (2001). Refining the construct of psychopathy: towards a hierarchical model. Psychological Assessment, 13(2), 171–188.

Cooke, D. J., Hart, S. D., Logan, C., & Michie, C. (2012). Explicating the construct of psychopathy: Development and validation of a conceptual model, the Comprehensive Assessment of Psychopathic Personality (CAPP). International Journal of Forensic Mental Health, 11(4), 242-252.

Cooke, D. J., Michie, C., Hart, S. D., & Clark, D. A. (2004). Reconstructing psychopathy: Clarifying the significance of antisocial and socially deviant behavior in the diagnosis of psychopathic personality disorder. Journal of Personality Disorders, 18(4), 337–357.

Corrado, R.R., DeLisi, M., Hart, S.D., & McCuish, E.C. (2015). Can the causal mechanisms underlying chronic, serious, and violent offending trajectories be elucidated using the psychopathy construct? Journal of Criminal Justice, 43(3).

Corrado, R.R., McCuish, E.C., Hart, S.D., & DeLisi, M. (2015). The role of psychopathic traits and developmental risk factors on offending trajectories from early adolescence to adulthood: A prospective study of incarcerated youth. Journal of Criminal Justice, 43(3).

Corrado, R. R., Vincent, G. M., Hart, S. D., & Cohen, I. M. (2004). Predictive validity of the Psychopathy Checklist: Youth Version for general and violent recidivism. Behavioral Sciences & the Law, 22(1), 5-22.

Cullen, F. T. (2011). Beyond adolescence-limited criminology: Choosing our future - The American Society of Criminology 2010 Sutherland Address. Criminology, 49(2), 287-330.

DeLisi, M. (2001). Extreme career criminals. American Journal of Criminal Justice, 25(2), 239- 252.

DeLisi, M. (2005). Career criminals in society. Thousand Oaks, CA: SAGE

DeLisi, M. (2009). Psychopathy is the unified theory of crime. Youth Violence and Juvenile Justice, 7(3), 256-273.

DeLisi, M., & Piquero, A. R. (2011). New frontiers in criminal careers research, 2000–2011: A

state-of-the-art review. Journal of Criminal Justice, 39(4), 289-301.

DeLisi, M., & Vaughn, M. G. (2014). Foundation for a temperament-based theory of antisocial

behavior and criminal justice system involvement. Journal of Criminal Justice, 42(1), 10-25.

Doherty, E. E., & Ensminger, M. E. (2014). Do the adult criminal careers of African Americans fit the “facts”? Journal of Criminal Justice, 42(6), 517-526.

Douglas, K. S., Vincent, G. M., & Edens, J. F. (2006). Risk for criminal recidivism: The role of psychopathy. In C.J. Patrick (Ed.), The handbook of psychopathy (pp. 533-554). New York, NY: Guilford Press.

Edens, J. F., Skeem, J. L., Cruise, K. R., & Cauffman, E. (2001). Assessment of “juvenile psychopathy” and its association with violence: A critical review. Behavioral Sciences & the Law, 19(1), 53-80.

Farrington, D. P. (1989). Early predictors of adolescent aggression and adult violence. Violence and Victims, 4(2), 79-100.

Farrington, D. P. (1991). Childhood aggression and adult violence: Early precursors and later-life outcomes. In D.J. Pepler and K.H. Rubin (Eds.), The development and treatment of childhood aggression (pp. 5-29). Hillsdale, NJ: Lawrence-Erlbaum.

Farrington, D. P. (1997). Early prediction of violent and non-violent youthful offending. European Journal on Criminal Policy and Research, 5(2), 51-66.

Farrington, D. P. (1998). Predictors, causes, and correlates of male youth violence. Crime & Justice, 24, 421.

Farrington, D. P. (2005). The importance of child and adolescent psychopathy. Journal of Abnormal Child Psychology, 33(4), 489-497.

Farrington, D. P., Snyder, H. N., & Finnegan, T. A. (1988). Specialization in juvenile court careers. Criminology, 26(3), 461-488.

Forth, A.E. (1995). Psychopathy and young offenders: Prevalence, family background, and violence. Unpublished manuscript. Carelton University, Ottawa, Canada.

Forth, A.E., Kosson, D.S., & Hare, R. (2003). The Hare Psychopathy Checklist: Youth Version. New York, NY: Multi-Health Systems.

Gadermann, A. M., Guhn, M., & Zumbo, B. D. (2012). Estimating ordinal reliability for Likert- type and ordinal item response data: A conceptual, empirical, and practical guide. Practical Assessment, Research & Evaluation, 17(3), 1-13.

Gretton, H. M., Hare, R. D., & Catchpole, R. E. (2004). Psychopathy and offending from adolescence to adulthood: a 10-year follow-up. Journal of Consulting and Clinical Psychology, 72(4), 636-645.

Hare, R. D., McPherson, L. M., & Forth, A. E. (1988). Male psychopaths and their criminal careers. Journal of Consulting and Clinical Psychology, 56(5), 710-714.

Hare, R. D., & Neumann, C. S. (2008). Psychopathy as a clinical and empirical construct. Annual

Review of Clinical Psychology, 4(1), 217-246.

Harris, G. T., Rice, M. E., & Cormier, C. A. (1991). Psychopathy and violent recidivism. Law and Human Behavior, 15(6), 625.

Harris, G. T., Rice, M. E., & Lalumière, M. (2001). Criminal violence: The roles of psychopathy, neurodevelopmental insults, and antisocial parenting. Criminal Justice and Behavior, 28(4), 402-426.

Hart, S. D. & Hare, R. D. (1997). Psychopathy: Assessment and association with criminal conduct. In D. M. Stoff, J. Maser, & J. Brieling (Eds.), Handbook of antisocial behavior (pp. 22-35). New York, NY: Wiley.

Hart, S. D. (1998). The role of psychopathy in assessing risk for violence: Conceptual and methodological issues. Legal and Criminological Psychology, 3(1), 121-137.

Hart, S. D., & Dempster, R. J. (1997). Impulsivity and psychopathy. In C. D. Webster & M. A. Jackson (Eds.), Impulsivity: Theory, assessment, and treatment (pp. 212-232). New York: Guilford Press.

Hoff, H. A., Rypdal, K., Mykletun, A., & Cooke, D. J. (2012). A prototypicality validation of the Comprehensive Assessment of Psychopathic Personality model (CAPP). Journal of Personality Disorders, 26(3), 414-427.

Jennings, W. G., & Reingle, J. M. (2012). On the number and shape of developmental/life-course violence, aggression, and delinquency trajectories: A state-of-the-art review. Journal of Criminal Justice, 40(6), 472-489.

Jones, B. L., & Nagin, D. S. (2007). Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociological Methods & Research, 35(4), 542-571.

Jones, B. L., Nagin, D. S., & Roeder, K. (2001). A SAS procedure based on mixture models for estimating developmental trajectories. Sociological Methods & Research, 29(3), 374-393.

Kreis, M. K., & Cooke, D. J. (2011). Capturing the psychopathic female: A prototypicality analysis of the Comprehensive Assessment of Psychopathic Personality (CAPP) across gender. Behavioral Sciences & the Law, 29(5), 634-648.

Laub, J. H. (2006). Edwin H. Sutherland and the Michael‐Adler Report: Searching for the soul of criminology seventy years later. Criminology, 44(2), 235-258.

Loeber, R., Farrington, D.P., Stouthamer-Loeber, M., White, H.R. (2008). Introduction and key questions. In R. Loeber, D.P. Farrington, M. Stouthamer-Loeber, & H. R. White (Eds.), Violence and serious theft: Development and prediction from childhood to adulthood (pp. 3-24). New York, NY: Routledge.

Lussier, P., & Davies, G. (2011). A person-oriented perspective on sexual offenders, offending trajectories, and risk of recidivism: A new challenge for policymakers, risk assessors, and actuarial prediction? Psychology, Public Policy, and Law, 17(4), 530-561.

Lykken, D. (1995). The antisocial personalities. New Jersey: Lawrence Erlbaum Associates.

Lynam, D.R., Caspi, A., Moffitt, T.E., Loeber, R., & Stouthamer-Loeber, M. (2007). Longitudinal evidence that psychopathy scores in early adolescence predict adult psychopathy. Journal of Abnormal Psychology, 116(1), 155-165.

MacDonald, J. M., Haviland, A., & Morral, A. R. (2009). Assessing the relationship between violent and nonviolent criminal activity among serious adolescent offenders. Journal of Research in Crime and Delinquency, 46(4), 553-580.

McCuish, E. C., Corrado, R., Lussier, P., & Hart, S. D. (2014). Psychopathic traits and offending trajectories from early adolescence to adulthood. Journal of Criminal Justice, 42(1), 66-76.

Nagin, D.S. (2005). Group-based modeling of development. Cambridge, MA: Harvard University Press.

Nagin, D.S. & Land, K. C. (1993). Age, criminal careers and population heterogeneity: Specification and estimation of nonparametric mixed poisson model. Criminology, 31(3), 327-354.

Newman, J. P., MacCoon, D. G., Vaughn, L. J., & Sadeh, N. (2005). Validating a distinction between primary and secondary psychopathy with measures of Gray's BIS and BAS constructs. Journal of Abnormal Psychology, 114(2), 319-323.

Piquero, A. R. (2008). Taking stock of developmental trajectories of criminal activity over the life course. In A.M. Liberman, (Ed.), The long view of crime: a synthesis of longitudinal research (pp. 23-53). New York, NY: Springer.

Piquero, A. R., Blumstein, A., Brame, R., Haapanen, R., Mulvey, E. P., & Nagin, D. S. (2001). Assessing the impact of exposure time and incapacitation on longitudinal trajectories of criminal offending. Journal of Adolescent Research, 16(1), 54-74.

Piquero, A. R., Brame, R., Mazerolle, P., & Haapanen, R. (2002). Crime in emerging adulthood. Criminology, 40(1), 137-170.

Piquero, A. R., Jennings, W. G., & Barnes, J. C. (2012). Violence in criminal careers: A review of the literature from a developmental life-course perspective. Aggression and Violent Behavior, 17(3), 171-179.

Piquero, A., Paternoster, R. P., Mazerolle, P., Brame, R., & Dean, C. W. (1999). Onset age and offense specialization. Journal of Research in Crime and Delinquency, 36(3), 275-299.

Quinsey, V. L., Harris, G. T., Rice, M. E., & Cormier, C. A. (1998). Violent offenders: Appraising and managing risk. Washington, DC: American Psychological Association.

Ribeiro da Silva, D., Rijo, D., & Salekin, R. T. (2012). Child and adolescent psychopathy: A

state-of-the-art reflection on the construct and etiological theories. Journal of Criminal Justice, 40(4), 269-277.

Salekin, R. T. (2008). Psychopathy and recidivism from mid-adolescence to young adulthood: Cumulating legal problems and limiting life opportunities. Journal of Abnormal Psychology, 117(2), 386-395.

Schneider, A. L. (1990). Deterrence and juvenile crime. New York, NY: Springer.

Serin, R. C. (1991). Psychopathy and violence in criminals. Journal of Interpersonal Violence, 6(4), 423-431.

Serin, R. C. (1996). Violent recidivism in criminal psychopaths. Law and Human Behavior, 20(2), 207.

Skardhamar, T. (2010). Distinguishing facts and artifacts in group-based modeling. Criminology, 48(1), 295-320.

Vachon, D.D., Lynam, D.R., Loeber, R., Stouthamer-Loeber, M. (2012). Generalizing the nomological network of psychopathy across populations differing on race and conviction status. Journal of Abnormal Psychology, 121(1), 263-269.

van der Geest, V., Blokland, A., & Bijleveld, C. (2009). Delinquent development in a sample of high-risk youth: Shape, content, and predictors of delinquent trajectories from age 12 to 32. Journal of Research in Crime and Delinquency, 46(2), 111-143.

Vaughn, M. G., & DeLisi, M. (2008). Were Wolfgang's chronic offenders psychopaths? On the

convergent validity between psychopathy and career criminality. Journal of Criminal Justice, 36(1), 33-42.

Vaughn, M. G., Howard, M. O., & DeLisi, M. (2008). Psychopathic personality traits and

delinquent careers: An empirical examination. International Journal of Law and Psychiatry, 31(5), 407-416.

Vincent, G.M. (2002). Investigating the legitimacy of adolescent psychopathy assessments: Contributions of item response theory. Unpublished doctoral dissertation.

Vincent, G. M., Odgers, C. L., McCormick, A. V., & Corrado, R. R. (2008). The PCL: YV and recidivism in male and female juveniles: A follow-up into young adulthood. International Journal of Law and Psychiatry, 31(3), 287-296.

Vitacco, M. J., Neumann, C. S., Caldwell, M. F., Leistico, A. M., & Van Rybroek, G. J. (2006). Testing factor models of the Psychopathy Checklist: Youth Version and their association with instrumental aggression. Journal of Personality Assessment, 87(1), 74-83.

Walters, G. D. (2003). Predicting institutional adjustment and recidivism with the psychopathy checklist factor scores: a meta-analysis. Law and Human Behavior, 27(5), 541-558.

Webster, C. D., Douglas, K. S., Eaves, D., & Hart, S. D. (1997). The HCR-20 scheme: Assessing risk for violence. Vancouver, Canada: Simon Fraser University.

Weiner, N.A. (1989). Violent criminal careers and “violent career criminals”: An overview of the research literature. In N.A. Weiner & M. Wolfgang (Eds.), Violent crime, violent criminals (pp. 35-158). Newbury Park, CA: Sage.

Wikström, P.-O.H. (2006). Individuals, settings, and acts of crime: Situational mechanisms and the explanation of crime. In P.O.H. Wikström and R.J. Sampson (Eds.), The explanation of crime: Context, mechanism, and development (pp. 61-107). Cambridge, UK: Cambridge University Press.

Wikström, P.-O.H., & Treiber, K. (2007). The role of self-control in crime causation: Beyond Gottfredson and Hirschi’s general theory of crime. European Journal of Criminology, 4(2), 237-264.

Wikström, P.-O. H., & Treiber, K. H. (2009). Violence as situational action. International Journal of Conflict and Violence, 3(1), 75-96.

Williamson, S., Hare, R. D., & Wong, S. (1987). Violence: Criminal psychopaths and their victims. Canadian Journal of Behavioural Science, 19(4), 454-462.

Tables and Figures

Table 1. Descriptive information of the sample (n = 326)

[CHART]

Figure 1. Distribution of violent and non-violent offending convictions from age 12 to 28

[CHART]

Figure 2a. Trajectories of violent offending from age 12 to 28

[CHART]

Figure 2b. Trajectories of non-violent offending from age 12 to 28

Figure 3. Conceptual outline of CAPP symptoms mapping onto situational action theory conditions for involvement in violence.