Developmental Trajectories of Interpersonal Callousness from Childhood to Adolescence as Predictors of Antisocial Behavior and Psychopathic Features in Young Adulthood

Abstract

Although previous research has established a link between early interpersonal callousness (IC) from childhood to adolescence and later antisocial behavior and psychopathic features, the majority of these studies assess more proximal outcomes (e.g., assessed in adolescence). Thus, it is unclear whether youth with early-onset chronic levels of IC will continue to have negative outcomes into adulthood (i.e., roughly 14 years after IC was assessed). The current study used data from the youngest cohort (N = 503) of the Pittsburgh Youth Study to examine how latent classes of youth with different developmental patterns of IC across a 7-year period (~ages 8 to 15) differed in their official records of juvenile (~ages 16–17) and young adult (~ages 18–31) offending, as well as self-reported psychopathic features and aggression in young adulthood (~age 29). Results indicated that after adjusting for race, early offending, and externalizing behaviors in adolescence, youth with an early-onset chronic pattern of IC had substantially elevated risk for a serious and persistent pattern of offending, particularly violent offending. However, IC class generally no longer significantly predicted psychopathic features in adulthood once these covariates were included. Thus, it is possible that the stability from early patterns of IC to adult psychopathic features may have previously been overstated. Future work could examine whether interventions to reduce IC in childhood and adolescence could successfully result in improved outcomes into adulthood.

General Scientific Summary

This study found that youth with high and stable levels of interpersonal callousness from childhood to adolescence have greater levels of aggressive and violent behavior in adulthood. However, developmental patterns of interpersonal callousness did not predict adult psychopathic features after controlling for race and early externalizing and offending behaviors.

Youth with high interpersonal callousness (IC; e.g., manipulative, dishonest, lack of guilt) are at increased risk for psychopathic features, as well as more serious and violent offending, in adolescence and adulthood (Barry et al., 2000; Byrd, Loeber, & Pardini, 2012; Pardini & Loeber, 2008), even after controlling for early conduct problems (Byrd et al., 2012; Frick, Stickle, Dandreaux, Farrell, & Kimonis, 2005). Previous studies have indicated that there are different potential etiological pathways to IC, and that youth on these pathways might differ in their patterns of IC over time and their exposure to biopsychosocial risk and protective factors (Byrd, Hawes, Loeber, & Pardini, 2018; Fontaine, Rijsdijk, McCrory, & Viding, 2010). However, there is a general lack of research examining whether youth with different developmental trends of IC exhibit different outcomes decades later into adulthood while controlling for important confounds, such as early offending and problem behaviors. The current study aims to assess how youth with different developmental patterns of IC from childhood into adolescence have different levels of offending in adolescence, and aggressive behaviors, psychopathic personality features, and offending in adulthood.

Interpersonal Callousness

In extending the concept of adult psychopathy to youth, researchers have focused on the interpersonal (e.g., manipulative, deceitful) and affective (e.g., lack of empathy or guilt) features of psychopathy, as they overlap less with symptoms of other disruptive behavior disorders, such as conduct disorder and oppositional defiant disorder, relative to the impulsive and antisocial features of psychopathy. The combination of interpersonal and affective characteristics, or IC, in childhood and adolescence is associated with greater risk for developing a more stable, serious, and violent pattern of antisocial behavior (Byrd et al., 2012; Frick & White, 2008; Pardini, Byrd, Hawes, & Docherty, 2018; White & Frick, 2010).

IC is often studied from a developmental psychopathology framework and conceptualized as prosocial development gone awry; in other words, youth with elevated IC lack the personal and contextual resources necessary to develop an aversive response to engaging in problem behaviors, hurting others, and being punished (Kochanska, 1991; Pardini, 2006; White & Frick, 2010). Specifically, youth with a fearless and uninhibited temperament seem particularly impervious to experiencing anxiety after misbehaving or being punished, and these youth are at greater risk for exhibiting elevated levels of IC (Byrd et al., 2018; Kochanska, 1991; Kochanska & Thompson, 1997). Person by environment interactions seem to play an important role, as fearless, uninhibited youth may be more likely to have higher levels of IC in the presence of harsh, non-responsive parenting (Beaver, Hartman, & Belsky, 2015; Waller & Hyde, 2017, 2018), as this type of parenting impedes children’s ability to internalize social norms and moral imperatives through interactions with parents.

Thus, youth with high levels of IC in childhood may experience adverse consequences due to cumulative consequences of their early IC (Caspi & Bem, 1990; Caspi, Bem, & Elder, 1989; Moffitt, 1993). In other words, having high IC at a young age, during a crucial time in moral development and socialization, could hinder not only conscience development, but the development of prosocial behaviors, warm relationships with caring others, and social skills. These deficits could continue to snowball into a developmental cascade that leads to greater antisocial behavior later in life. In addition, if individuals remain elevated in IC throughout life, they may also be at risk for antisocial behavior through contemporary consequences, as IC is contemporaneously associated with problem behaviors (Pardini, Lochman, & Powell, 2007; van Baardewijk, Vermeiren, Stegge, & Doreleijers, 2011). Finally, previous research has indicated that high, stable levels of IC from an early age are more heritable (Fontaine et al., 2010), and thus youth with this pattern of IC may be at genetic risk for other personality features and problem behaviors. In contrast, youth with childhood-limited IC might experience the cumulative consequences associated with elevated IC early in life, but would not necessarily be subject to contemporaneous consequences later in adolescence and adulthood. Finally, youth with low levels of IC in childhood but who increase in their IC through adolescence may also be at risk for offending and aggression because of contemporary consequences, though to a lesser extent because they will not have the same levels of cumulative consequences as youth with high, stable levels of IC, and also may not have the same level of genetic risk.

Heterogeneity in Development of Interpersonal Callousness

Consistent with different theorized causal pathways to IC, a growing body of research suggests that there is considerable heterogeneity in the developmental patterns of IC in the population from childhood to adolescence (Hawes, Byrd, Waller, Lynam, & Pardini, 2017; Muratori et al., 2016; Pardini & Loeber, 2008), suggesting that accounting for individual differences in IC trajectories may be warranted in examining later outcomes. Although these studies modeled individual-level heterogeneity in development of IC, they still followed assumptions – such as that individuals’ IC trajectories are normally distributed around a mean trajectory – that may not be entirely valid, given potential important etiological differences. Indeed, a number of studies have demonstrated that youth can be classified into latent groups based on their unique development of callousness over time. These studies have consistently identified four subgroups of youth: 1) early-onset chronic group; 2) late-onset increasing group; 3) childhood-onset decreasing; and 4) a low stable group (Byrd et al., 2018; Fanti, Colins, Andershed, & Sikki, 2017; Fontaine, McCrory, Boivin, Moffitt, & Viding, 2011; Fontaine et al., 2010; Klingzell et al., 2016).

Because youth can differ qualitatively on their initial levels and changes in callousness over time, they might also differ on their risk exposure and problem behavior outcomes. For example, youth with high, stable levels of IC tend to be more fearless and exposed to higher levels of negative parental discipline and maltreatment than youth with other developmental patterns of IC (Byrd et al., 2018; Fontaine et al., 2010). Thus, youth with different patterns of IC likely have different causal mechanisms underlying their development of IC. Youth with high, stable levels of IC might have high levels of both personal and contextual risk, while youth with decreasing levels might have high levels of personal risk but be exposed to greater environmental protective factors, and youth with increasing IC might be exposed to adverse environmental factors, such as harsh parenting and trauma, that could cause them to become more manipulative and less empathetic and remorseful.

In addition to differences in risk exposure, studies have found that youth with different patterns of IC exhibit different outcomes (Fanti et al., 2017; Fontaine et al., 2011). Specifically, youth with high, stable levels of callousness tend to have higher levels of peer problems (Fontaine et al., 2011) and lower levels of academic performance, peer social support, and parental involvement (Fanti et al., 2017) than youth in other IC groups. However, because outcomes were assessed simultaneously with the final assessment of callousness, it is unclear whether these groups would continue to exhibit different outcomes – especially costly outcomes, such as serious violence – years later. Although one study did examine developmental patterns of psychopathic features – a broader collection of personality characteristics that includes IC, as well as impulsive or erratic lifestyle and antisocial tendencies – this study failed to control for important confounds, such as early levels of problem behaviors, that could explain the association between early trajectories of psychopathic features and later outcomes (Hawes et al., 2018). This is particularly important, given that previous studies have found early levels of problem behaviors to be much higher among youth with high, stable levels of IC (Byrd et al., 2018; Fontaine et al., 2011).

Interpersonal Callousness and Outcomes in Adulthood

Previous studies that have examined adult outcomes of IC in childhood and adolescence have tended to use a variable-centered approach, and used callous-unemotional traits or IC at a single point in time to predict later outcomes (Byrd et al., 2012; McMahon, Witkiewitz, Kotler, & The Conduct Problems Prevention Research Group, 2010). For example, one study found that IC in childhood and adolescence predicted delinquency from adolescence to adulthood, including a more persistent pattern of offending (Byrd et al., 2012). However, IC was only measured once in childhood, at age 7, and once in adolescence, at age 15. Another study found that callous- unemotional traits in seventh grade predicted offending and antisocial personality disorder up to age 20 (McMahon et al., 2010).

Additional studies have assessed how levels of IC or changes in IC assessed at multiple time points in childhood and/or adolescence are associated with later outcomes (Burke, Loeber, & Lahey, 2007; Hawes et al., 2017; Muratori et al., 2016; Pardini & Loeber, 2008). For example, in one clinic-referred sample, IC from childhood to adolescence was associated with psychopathic features in young adulthood (Burke et al., 2007). However, some of these studies have assessed outcomes after a relatively shorter follow-up period (e.g., within one or two years; Burke et al., 2007; Muratori et al., 2016). Stronger evidence comes from studies that have a longer follow-up time frame and predict outcomes into adulthood. Some of these studies have indicated that IC in late childhood (Hawes et al., 2017) and adolescence (Pardini & Loeber, 2008) is associated with psychopathic features in adulthood, and that IC in childhood is associated with a pattern of persistent and violent adult offending (Pardini et al., 2018).

It is important to distinguish youth not just on their level of IC, but on their developmental patterns of IC, to account for potentially different etiological pathways, as noted above. One study with a juvenile justice system-involved sample used developmental patterns of callousness from adolescence to young adulthood to predict offending outcomes at age 24 (Baskin-Sommers, Waller, Fish, & Hyde, 2015), but this study suffered from several limitations. For example, the offending outcomes were contemporaneous with the final assessment of callousness, which complicates the interpretation of the temporal ordering; callousness was only measured from adolescence to young adulthood, which means emerging patterns of IC in childhood were not assessed; and all key variables (i.e., callousness, offending) were measured via youth self-report, which could lead to inflated associations due to shared method variance. Thus, we still do not know how developmental patterns of IC from childhood to adolescence are uniquely associated with outcomes in adulthood while controlling for early problem behaviors. This kind of research can better inform theoretical models of IC development and antisocial behavior, as well as point to potential targets for prevention and intervention efforts.

Limitations of Prior Work

Much of the work examining heterogeneity in IC development (Hawes et al., 2017; Pardini & Loeber, 2008), predictors of IC trajectories (Byrd et al., 2018), and adult outcomes of youth IC (Byrd et al., 2012; Hawes et al., 2018; Hawes et al., 2017) has been done with the Pittsburgh Youth Study (PYS). The PYS interviewed boys enrolled in Pittsburgh public schools, as well as their parents and teachers, from childhood into early adulthood, and includes official records of offending. Although this work has advanced the understanding of IC development, existing studies have yet to identify the degree to which developmental patterns of IC early in development are uniquely associated with behavioral outcomes and psychopathic features in adolescence and adulthood, net of other confounds such as early problem behaviors. In addition, many previous studies have used self-reports for both IC and the outcomes of interest, which may produce inflated associations because of shared method variance. Therefore, it is important to examine the research questions with other informants to clarify and determine the robustness of the associations.

The Current Study

The primary contribution of the current study is to examine how different developmental patterns (e.g., stable and high, low, or moderate; increasing; or decreasing) of teacher-reported IC from childhood to adolescence are associated with official records of offending in adolescence and adulthood, as well as self-reported psychopathic personality features and aggressive behavior in adulthood, while controlling for important confounds – race, early offending, and externalizing problems in adolescence. We examined this research question with the PYS data and used the IC trajectories reported in a previously published paper with these data (Byrd et al., 2018). Although the prior study identified the early childhood factors associated with the IC trajectory groups, they did not examine any outcomes associated with these trajectory groups (Byrd et al., 2018). We hypothesize that, compared to the low IC class, all other IC classes will have greater arrests, charges, and violence in adolescence, as well as higher psychopathic characteristics and aggression scores, more arrests and charges, and greater likelihood of engaging in serious violence and offending in adulthood. Additionally, the early- onset chronic class with high, stable levels of IC will have the highest risk for all outcomes. We hypothesize that the magnitude of these associations might be reduced after controlling for race, early offending, and externalizing problems in adolescence, but they will remain significant.

Although we do not have specific hypotheses regarding differences between other IC classes (e.g., increasing vs. decreasing), we included these comparisons to explore how different developmental trends of IC uniquely explain differences in outcomes. Each of these comparisons can inform the contemporaneous and enduring impact of elevated levels of IC, and whether elevations in IC at any particular developmental window might be particularly important for the development of psychopathic features and antisocial behavior. Worse outcomes for youth on a decreasing trajectory might indicate that elevated IC during childhood is particularly problematic, potentially due to the processes of moral and social development taking place during this time, and to the cumulative consequences (e.g., deficits in perspective taking and social skills) that may be the result of IC at this age. On the other hand, worse outcomes for youth on an increasing trajectory could indicate that the contemporary consequences (e.g., peer victimization, violence exposure) of high levels of IC in adolescence are especially detrimental.

Methods

Participants and Procedure

The sample for the current study comes from the youngest cohort (N = 503) of the PYS (Loeber, Farrington, & Stallings, 2011). In 1987–1988, a random sample of 856 boys enrolled in first grade in Pittsburgh public schools was selected to participate in a screener. Among the boys selected for screening, a multi-informant (parent, teacher, and youth) assessment of conduct problems was used to oversample boys who were perceived to have a higher risk of engaging in serious criminal behavior in adolescence. In particular, a random sample of boys who scored around or above the 30^th percentile on the screener (n = 249), and a random sample of boys below the 30^th percentile (n = 257), were selected to participate in the follow up study (total N = 503). The follow-up sample did not differ from the screening sample on race, family composition, parental education, or parental employment (Loeber, Farrington, Stouthamer- Loeber, & White, 2008). Boys selected for follow-up averaged 7 years of age at screening (SD = 0.55) and were predominantly Black (55.7%) or White (40.6%). The primary caretaker for most boys was their biological mother (89.5%), and more than half (57.7%) of the sample was on public assistance at the first follow-up assessment.

Following screening, boys were interviewed every six months for four years, followed by nine annual assessments (17 assessments total). They were then re-interviewed in adulthood (Time 18; M age = 29.25, SD = 1.11). After screening, parents and teachers were also interviewed every six months for four years, followed by five annual assessments (12 assessments total). Data collection consisted of an interview with each boy, an interview with his primary caregiver, and questionnaires completed by the boy’s teacher. Official criminal records were also reviewed in order to obtain information on arrests, charges, and convictions up through age 31. Families were paid for their participation, and boys’ assent and a legal guardian’s informed consent (or the boys’ informed consent, once they turned 18) was obtained prior to each assessment. The University of Pittsburgh Institutional Review Board reviewed and approved all data collection procedures.

Consistent with prior work (Byrd et al., 2018), the current study used teacher-reported data from the spring assessments in 2^nd through 9^th grade, roughly corresponding to ages 8 to 15, to assess IC from childhood to adolescence. The adolescent outcome variables were measured with official criminal records from ages 16 to 17. The adult outcome variables were measured with self-reported psychopathic features and aggression from the age 29 interview (Time 18) and official criminal records from ages 18 to 31. Retention remained high across the study; 81% at Time 8 for teacher report, and 90% at Time 8 and 80% at Time 18 for youth self-report. More detailed information on participants, measures, and data collection procedures for the PYS can be obtained elsewhere (Loeber et al., 2011; Loeber, Farrington, Stouthamer-Loeber, & Van Kammen, 1998; Loeber et al., 2008).

Measures

Interpersonal callousness (IC).

IC was measured at the annual spring assessments from ages 8 to 15 (Times 2, 4, 6, 8–12) with teacher report on eight items (Pardini, Obradović, & Loeber, 2006) from the Teacher’s Report Form (TRF; Achenbach & Edelbrock, 1986). Teachers rated how well each item (e.g., “doesn’t seem to feel guilty after misbehaving”) described the child on a 3-point Likert scale from 0 (“Not true”) to 2 (“Very true”). Prior studies have found evidence that this scale exhibits moderate temporal stability from childhood to adolescence and predicts future offending among youth (Byrd et al., 2018; Pardini & Loeber, 2008; Pardini et al., 2006). The internal consistency of the scale was high across the assessments used in the current study (αs from .91–.94). Teacher-reported IC has been used in a number of studies and shows similar correlates both concurrently and prospectively compared to self-reported and parent- reported IC (for a review, see Pardini & Byrd, 2013).

Outcomes.

Psychopathic features.

Psychopathic features were measured in adulthood at the age 29 interview (Time 18) using the Self-Report Psychopathy scale (SRP-III; Paulhus, Neumann, & Hare, 2009). The SRP-III includes four facets, comprised of 15 items each, for a total of 60 items, and previous research using a subsample from the PYS demonstrated moderate to high correlations between the SRP-III and scores on an interviewer-rated psychopathy instrument (Neumann, Hare, & Pardini, 2015). Facets include interpersonal manipulation (e.g., “I can talk people into anything”; α = 0.78), callous affect (e.g., “I never feel guilty over hurting others”; α = 0.70), erratic lifestyle (e.g., “I easily get bored”; α = 0.76), and antisocial lifestyle (e.g., “I was convicted of a serious crime”; α = 0.76). Participants rated how much they agreed with each item on a 5-point Likert scale from 1 (“Disagree strongly”) to 5 (“Agree strongly”). After reverse scoring items worded in the opposite direction, items within each subscale were averaged to compute facet scores, and all 60 items were averaged to produce a total score (α = 0.92).

Physical aggression.

Physical aggression was measured at the age 29 interview (Time 18) using nine items from the physical aggression subscale of the Buss-Perry Aggression Questionnaire (Buss & Perry, 1992). Participants rated each item (e.g., “I get into fights a little more than the average person”) on a 5-point scale from 1 (“Not at all like me”) to 5 (“Completely like me”). After reverse scoring one item, the nine items were summed to produce a total score (α = 0.92).

Antisocial behavior.

Official criminal records from age 11 to age 31 were obtained via local, state, and federal sources. Antisocial behavior in adolescence was measured using the number of charges received and the number of times participants were arrested, excluding minor offenses (e.g., traffic violations, status offenses), as well as whether or not participants were arrested for serious violence, beginning from Time 13 (~age 16) and through age 17. Serious violence was measured using a binary variable indicating whether participants had been charged with homicide, aggravated assault, sexual assault, and/or robbery. Antisocial behavior in adulthood was measured using the number of charges received and the number of times participants were arrested, excluding minor offenses, and whether or not participants were arrested for a serious or violent charge, from age 18 to age 31. Additionally, active criminal offending was defined as having an arrest within the last two years of the most recent criminal record search at age 31.

Covariates.

Race.

Race was assessed by caretaker report at the screening assessment and was computed as a binary variable indicating whether or not a participant was identified as Black.

Early offending.

Official records of offending were used to create a variable indicating whether or not a participant had been arrested by the age 15 assessment (Time 13).

Externalizing in adolescence.

Externalizing problems in adolescence were indexed by the externalizing subscale of the Youth Self-Report at the age 15 assessment (Achenbach, 1991).

Missing Data

Retention remained relatively high throughout the study. Out of 503 youth, 501 had valid data on teacher-reported IC, 502 youth had valid data on official records, and 399 youth had valid data on psychopathic features and physical aggression at age 29. We conducted analyses using full information likelihood (FIML) in Mplus 7.2 (Muthén & Muthén, 2012) to handle missing data. FIML uses all available data to generate parameter estimates and provides unbiased and efficient parameter estimates under the assumption that the data are missing at random (Enders, 2001; Enders & Bandalos, 2001). Because continuous and count outcome variables were skewed, we used a maximum likelihood estimator robust to non-normality (MLR) for all models.

Plan of Analysis

A previous study using these data conducted latent class growth analysis of teacher- reported IC and found that a five-class solution provided the best fit to the data (Byrd et al., 2018). The largest class (52%) had consistently low levels of IC over time (“low”). There was also a small class (10%) that had high and stable levels of IC over time (“early-onset chronic”). There were also three classes that changed substantially in their level of IC over time: one group that started high but decreased to low levels (10%; “childhood-limited”); one group that started low but increased to high levels (12%; “adolescent-onset”); and one group that started relatively low, peaked at a moderate level around age 11, and then decreased again to low levels (16%; “moderate”). These trajectories are plotted in Figure 1. All of the observed differences in teacher-reported IC were significant at age 8, and each of the classes was again significantly different at age 15, with the exception of no significant difference between the early-onset chronic and adolescent-onset classes. In terms of mean level of IC across the study period, the early-onset chronic group had the highest, followed by the adolescent-onset and moderate classes, then the childhood-limited class, and finally the low class.

Figure 1.

Latent trajectories of teacher-reported callousness at annual assessments from ages 8 to 15. CL=childhood-limited; AO=adolescent-onset; EOC=early-onset chronic.

For the purposes of the current study, we classified youth into their most likely trajectory based on posterior predicted probabilities and used this classification as a nominal predictor variable in regression models. Posterior predicted probabilities of class membership were generally pretty high, with averages of .93 for the low class, .88 for the childhood-limited class, .82 for the moderate class, .85 for the adolescent-onset class, and .90 for the early-onset chronic class. We ran models predicting each outcome while controlling for race, early offending, and externalizing in adolescence as potential confounds. We tested for the main effect of IC class using an omnibus Wald χ² test, which indicated whether there were any significant group differences. In the case of a significant Wald test, we included pairwise comparisons between each of the IC classes. We also calculated Cramer’s V as a measure of effect size for the Wald χ² test, and Cohen’s d as a measure of effect size for pairwise comparisons (Cohen, 1988). We used Cohen’s (1988) guidelines for small, medium, and large effect sizes: 0.2, 0.5, and 0.8, respectively, for Cohen’s d, and 0.05, 0.15, and 0.25, respectively, for a Wald χ² with 4 degrees of freedom. We specifically highlight comparisons between the early-onset chronic group and all other groups, especially the childhood-limited and adolescent-onset groups, to examine how a high, stable level of IC from childhood to adolescence uniquely predicts outcomes in adulthood. However, we included all possible pairwise comparisons to examine how different trajectories of IC (e.g., increasing vs. decreasing) might confer different risk for future antisocial outcomes. Adolescent externalizing was standardized in order to aid in interpretation (a 1-unit increase in the predictor indicates an increase of 1 standard deviation). We used linear regressions to predict psychopathic features and aggression, negative binomial regressions to predict counts of arrests and charges, and logistic regressions to predict the binary measures of serious violence and active offending.

Results

Antisocial Behavior in Adolescence

In models using IC class to predict the number of arrests, number of charges, and whether or not youth were arrested for a serious violent offense prior to age 18, while controlling for race, early offending, and externalizing in adolescence, the only significant main effects of IC class were for number of charges (p = .040) and serious violence (p = .002). Results are displayed in Table 1. Results indicated that the adolescent-onset and early-onset chronic classes had significantly more charges in adolescence than the low, childhood-limited, and moderate classes, and that the early-onset chronic class was more likely to have an arrest for serious violence in adolescence than the low, childhood-limited, and moderate classes. Overall effect size was small for both serious violence (V = .09) and charges (V = .07), and pairwise effect sizes were small for charges (ds = 0.17–0.44) and small to medium for serious violence (ds = 0.41–0.58). Race was a significant predictor for arrests and serious violence, such that Black youth had more arrests and were more likely to have an arrest for serious violence in adolescence than White youth. Early offending was a significant positive predictor in each of the models, and externalizing only significantly predicted the number of charges in adolescence.

Table 1.

Multivariate Models Using Callousness Classes to Predict Psychopathic Features and Antisocial Behavior, Controlling for Covariates

	Low		Childhood-limited		Moderate		Adolescent-onset		Early-onset chronic		Wald
Outcome	Pr/M	(S.E.)	Pr/M	(S.E.)	Pr/M	(S.E.)	Pr/M	(S.E.)	Pr/M	(S.E.)	χ2(4)
Psychopathic features in adulthood
Total	2.44	(0.03)	2.51	(0.07)	2.52	(0.05)	2.49	(0.07)	2.55	(0.07)	3.38
Interpersonal	2.44	(0.03)	2.46	(0.07)	2.48	(0.06)	2.48	(0.07)	2.41	(0.08)	0.89
Affective	2.62	(0.03)	2.67	(0.07)	2.67	(0.06)	2.66	(0.07)	2.66	(0.08)	1.08
Erratic	2.73	(0.04)	2.77	(0.08)	2.79	(0.06)	2.69	(0.09)	2.83	(0.09)	2.01
Antisocial	1.97a	(0.04)	2.13a,b	(0.10)	2.15b	(0.08)	2.13a,b	(0.10)	2.34b	(0.10)	11.72*
Antisocial behavior in adolescence
Arrests	0.44	(0.07)	0.50	(0.11)	0.60	(0.12)	0.75	(0.16)	0.89	(0.18)	8.88
Charges	1.04a	(0.17)	1.39a	(0.35)	1.53a	(0.44)	2.24b	(0.59)	2.41b	(0.63)	10.04*
Serious violence	0.07a	(0.02)	0.08a	(0.04)	0.05a	(0.02)	0.13a,b	(0.04)	0.25b	(0.07)	16.75**
Antisocial behavior in adulthood
Aggression	20.84a	(0.39)	22.60a,b	(0.89)	21.21a	(0.63)	21.91a	(0.95)	24.82b	(1.11)	12.79*
Arrests	1.84a	(0.18)	3.30b,c	(0.56)	2.98b	(0.34)	4.14b,c	(0.60)	4.38c	(0.60)	34.04***
Charges	5.80a	(0.70)	11.62b,c	(2.30)	8.80b	(1.20)	12.49b,c	(2.27)	14.18c	(2.14)	26.26***
Arrest past 2 years	0.09a	(0.02)	0.22b	(0.06)	0.16a,b	(0.04)	0.25b	(0.06)	0.28b	(0.07)	13.30**
Serious violence	0.16a	(0.03)	0.25a	(0.07)	0.27a	(0.05)	0.24a	(0.06)	0.47b	(0.08)	15.62**

Note.

^*p < .05

^**p < .01

^***p < .001.

Parameters reflect estimated means (M) or predicted counts or probabilities (Pr) and standard errors (S.E.) from maximum likelihood (MLR) models.

Wald χ²(4) tests the main effect of callousness class in each model.

Subscripts denote statistically significant group differences.

Models adjust for race (White/Black), early offending (yes/no), and externalizing in adolescence, each of which was standardized.

Antisocial Behavior in Adulthood

We next used IC class to predict aggression, number of arrests and number of charges, and arrest for serious violence in adulthood, as well as whether there was an arrest in the past two years of the final criminal record search at age 31, while controlling for race, early offending, and externalizing in adolescence. Results are displayed in Table 1, and models generally indicated a main effect for IC class (p < .05 for all Wald tests; see Table 1 for results). The early-onset chronic class had significantly higher levels of aggression in adulthood than the low, moderate, and adolescent-onset classes, and had a greater proportion arrested for serious violence in adulthood compared to all other classes. In the models predicting number of arrests and number of charges, the childhood-limited, moderate, adolescent-onset, and early-onset chronic classes had more arrests and charges than the low class, and the early-onset chronic class had more arrests and charges than the moderate class.

Finally, the childhood-limited, adolescent-onset, and early-onset chronic classes were more likely than the low class to be active offenders; i.e., arrested within the past two years at age 31. Thus, elevated callousness in either childhood or adolescence could be associated with greater risk for continued offending into adulthood. Overall effect size was small for all outcomes: arrests (V = .13), charges (V = .11), serious violence (V = .09), arrest in the past two years (V = .08), and aggression (V = .08). Pairwise effect sizes were medium for comparing the early-onset chronic and low classes (ds = 0.53–0.80), small to medium for the comparisons between adolescent-onset and low classes (ds = 0.46–0.70) and between early-onset chronic and moderate classes (ds = 0.39–0.54). All other pairwise comparisons had small effect sizes (ds = 0.27–0.48). Race was a significant predictor of aggression, number of arrests, number of charges, and arrest for serious violence in adulthood, such that Black youth had higher levels of antisocial behavior than White youth. Early offending significantly predicted number of arrests, number of charges, and arrest for serious violence in adulthood. Externalizing was only a significant predictor for aggression.

Psychopathic Features in Adulthood

The models predicting psychopathic features in adulthood from IC class, while controlling for race, early offending, and externalizing in adolescence, generally did not find a significant main effect for IC class, for the total score or for many of the facet scores (see Table 1 for results). Only one Wald test was significant (χ²(4) = 11.72, p = .020), for predicting the antisocial behavior scale, and there were only two significant pairwise comparisons. The moderate class (standardized β = 0.10, S.E. = .05, p = .049) and the early-onset chronic class (standardized β = .17, S.E. = .06, p = .002) each scored significantly higher than the low class. Cramer’s V indicated a small effect size (V = .08) for the Wald test, and Cohen’s d indicated small to medium effect sizes for the pairwise comparisons (d = 0.27–0.57). Race was a significant predictor in all models except the one predicting erratic lifestyle, such that Black youth reported higher levels of psychopathic features than White youth. Externalizing was a significant positive predictor in each of the models. Early offending was significantly associated with higher psychopathic feature scores, except for in the models predicting the interpersonal and erratic subscale scores.

To explore the relationship between IC class, early problem behaviors, and adult psychopathic features, we added the IC class predictor and covariates into the model separately in a stepwise fashion (step 1a = IC class; step 1b = race, early offending, and adolescent externalizing; step 2 = step 1a + step 1b) for each of the psychopathic feature outcomes. In step 1a, the R² value indicated that IC class alone was a significant predictor (all ps < .01) and accounted for 5–25% of the variability in outcomes. However, step 1b indicated that the covariates were also significant predictors on their own, and explained a moderate amount of the variance in the outcomes (R² = .11 – .28). In step 2 of each model, R² was elevated relative to step 1a, but not step 1b, suggesting that race, early offending, and adolescent externalizing are generally able to account for all of the variability in psychopathic features explained by IC class, and also uniquely account for an additional 3–9% of variability.

Discussion

The current study examined adolescent and adult outcomes of IC trajectories from ages 8 to 15 in a community sample of 503 boys. Results indicated that IC class generally was not associated with self-reported psychopathic features in adulthood after controlling for symptoms of early conduct problems. However, there were small to medium associations between IC class and aggression and antisocial behavior in adolescence and adulthood, even after controlling for early offending, adolescent externalizing problems, and race. These findings are important because this was the first study to include critical confounding factors while examining the extent to which IC trajectories from childhood to adolescence predict antisocial outcomes in adulthood.

There are many reasons why IC class might not be associated with adult psychopathic features after including covariates; for example, because they are developmentally distinct phenomena, because of measurement limitations, or because the association is accounted for through confounding variables, such as early externalizing problems. Indeed our data suggest that potentially co-morbid early problem behaviors may completely explain any observed stability between childhood/adolescent IC and adult psychopathic features. Future research can examine the extent to which early onset problem behaviors account for the association between IC during childhood and adolescence and psychopathic features in adulthood. These results suggest that the stability from IC early in life to psychopathic features in adulthood might be over-stated in prior studies that do not control for early problem behaviors. This is particularly interesting in light of the prior study using these trajectories, which found that the early-onset chronic class had higher levels of psychosocial risk factors, such as fearlessness, conduct problems, and psychosocial adversity, which are also theorized to be associated with psychopathic features in adulthood (Byrd et al., 2018).

Youth with high, stable levels of IC from childhood to adolescence and youth who had increasing levels of IC into adolescence had a greater number of charges in adolescence compared to the other IC classes. On average, these youth had more than twice the number of charges relative to youth who displayed low, stable levels of IC during childhood and adolescence. Youth with chronic IC were uniquely at risk for serious violence in adolescence, and were 2 to 5 times more likely to engage in serious violence relative to youth in the other IC classes, even while controlling for prior offending and adolescent externalizing problems. Thus, youth with adolescent onset IC appeared to experience some contemporary consequences of their elevated IC, as they were similar in adolescent outcomes relative to the early-onset chronic youth.

There were small to medium associations between IC class and behavioral outcomes (e.g., aggression, number of arrests) 14 years later in adulthood, even while controlling for model covariates. These findings were particularly pronounced for aggression and serious violence, in which the group of youth who had high and stable levels of IC from childhood through adolescence had the highest levels of these outcomes, and were roughly two to three times more likely to be arrested for serious violence than youth in any other trajectory class. This is in line with prior work, which has found callousness to be more effective in predicting more severe problem behaviors (Barry et al., 2000; Frick & Ellis, 1999; Frick & White, 2008), and extends prior work by examining long-term outcomes for individuals who exhibit different developmental patterns of IC.

The results of this study justify the need to examine developmental changes in IC, and to use person-level approaches to investigate differences between latent groups that differ in their trends over time. These results also point to early patterns of IC as a potential target for violence prevention programs. Although the childhood-limited and moderate classes had elevated IC during childhood that decreased substantially by adolescence, these youth had more arrests and charges than the low IC class. Further, although the adolescent-onset group had low levels of IC at younger ages, they were more similar to the early-onset chronic group than to the low group in their levels of psychopathy and antisocial behavior in adulthood, with the exceptions of aggression and serious violence. Thus, elevated IC at any point in youth might be associated with similarly elevated risk for offending behaviors in adulthood. It might be that elevated IC at any point in early development could confer risk for adverse outcomes through either cumulative or contemporary continuities. Specifically, the childhood-limited group may have greater antisocial behavior in adulthood because of cumulative consequences: early levels of elevated IC could contribute to deficits in moral development, positive parent-child relationships, and peer socialization, that might continue to lead to other escalating problems in a developmental cascade that ends with antisocial behavior, even though IC may have decreased by that point. In contrast, the adolescent-onset group is most likely impacted by contemporary consequences of elevated IC, such as increases in negative responses from parents and peers, which could lead to increases in antisocial behavior. This is supported by their elevated number of charges in adolescence relative to the low, childhood-limited, and moderate IC classes. Future research could examine more proximal differences between the childhood-limited and adolescent-onset groups, to determine whether different intermediary processes explain outcomes for these groups (e.g., deficits in moral development for childhood-limited, and increases in peer victimization for adolescent-onset). The combination of cumulative and contemporary consequences, however, likely experienced by the early-onset chronic class is associated with the most adverse outcomes.

Examining callousness at only one point in childhood or adolescence may miss these important developmental differences. For example, assessing IC at only age 8 might assume that the childhood-limited class would experience the worst outcomes, as these youth had the highest levels of IC in childhood; however, they were more similar in their outcomes to the moderate and adolescent-limited classes, which had much lower levels of IC at age 8. A single assessment in adolescence would also miss important information, as assessing IC at only age 15 would likely fail to capture differences between the early-onset chronic and adolescent-onset groups, and the low and childhood-limited groups. Thus, single assessments of IC at one point in development are likely not adequately capturing youth’s developmental patterns of IC. Even looking at mean level over time might assume worse outcomes for the adolescent-onset and moderate classes compared to the childhood-limited class, but this was not the case. Although clinicians are unlikely to have enough information to know a youth’s specific trajectory of IC, further research on early risk factors and indicators of trajectory membership could help to identify which youth are at greatest risk for early-onset stable levels of IC.

Future research could focus on examining the etiology of different IC trajectories – for example, not just early risk and protective factors, but also exposure to risk and protective factors at important turning points. It could be the case that the low and adolescent-onset groups might be comparable at baseline, but different in their exposure to socio-contextual risk factors in late childhood. This would be in line with previous work on shorter trajectories of callous- unemotional traits in childhood, which found that youth with an increasing or low stable trajectory had similar levels of maternal involvement at baseline, but the increasing group had decreasing levels of maternal involvement over the study period (Fanti et al., 2017). It may also be important to examine whether treatment could alter the course of a youth’s development of IC, and if so, whether this would be related to decreases in aggressive and violent behavior in adulthood, as the current study suggests it might. The childhood-limited group might have had high levels of personal risk for IC, as theory would suggest, but could have been exposed to protective factors, such as warm and responsive parenting, that influenced their decreases in IC. If this was indeed the case, then increasing youth’s exposure to protective factors like positive parenting could potentially reduce their risk for serious violence in adulthood. However, this study did not explicitly test these theoretical models; future work might explore how exposure to protective factors in childhood and adolescence could protect against committing severe violence in adulthood for youth with a fearless, uninhibited temperament.

Limitations

It is important to consider a number of limitations when interpreting results from the current study. For example, these results are based on a community sample of boys and may not generalize to girls, clinical, or offender samples. Future research should also strive to create predictors that are composites of multiple informant reports (e.g., parent, teacher, self-report), in order to provide the most accurate estimates possible. Although the latent class growth analysis allowed us to examine outcomes of youth who exhibit different developmental patterns of IC, these classes represent approximations of youths’ developmental trajectories of IC and should not be reified (Nagin & Tremblay, 2005). Because the classes are relatively small, some of the analyses may be underpowered. Because IC class did not significantly predict psychopathic features while early offending behaviors and externalizing in adolescence were included as covariates, it is possible that these intermediary behaviors may mediate the association between IC in childhood and adolescence and psychopathic features in adulthood; future research can investigate these and other potential intermediary processes. Finally, the measure of IC used in this study is an ad hoc measure adapted from the Teacher Report Form, and does not cover all of the facets of IC – such as shallow affect and lack of empathy – assessed using other common instruments in the literature (e.g., the Inventory of Callous-Unemotional Traits; Frick, 2004; Kimonis et al., 2008).

Conclusion

The current study indicates that IC trajectories assessed in childhood and adolescence uniquely predict offending outcomes, even into adulthood 14 years later. Even though results were not significant for psychopathic features once race and early problem behaviors were controlled for, the classes still significantly differed on costly outcomes, such as aggression and serious violence, demonstrating the long-term impact of these developmental patterns. This work expands on previous research by demonstrating that youth with relatively low levels of IC are at lowest risk for antisocial behavior in adulthood, while those with high stable IC are at the greatest risk, even after accounting for early levels of problem behaviors. However, even youth with elevated levels of IC in only childhood or only adolescence had more arrests and charges in adulthood than youth whose IC remained low and stable over the study period. It also demonstrates that the stability between early levels of IC and adult psychopathic features may not be as robust as previously thought, as this association was no longer significant when controlling for other factors. Future work can identify whether there are more proximal mechanisms underlying the associations between early IC trajectories and outcomes in adulthood, or whether interventions aimed at reducing IC in childhood and adolescence can improve adult outcomes.