Examining Latent Profiles of Psychopathy in a Mixed-Gender Sample of Juvenile Detainees

Abstract

There is growing support for the disaggregation of psychopathy into primary and secondary variants. The present study employed latent profile analysis to distinguish psychopathic variants in a sample of male and female adolescent detainees (N = 162). Youth were classified by their scores on the self-report Triarchic Psychopathy Measure, indexing trait Boldness, Meanness, and Disinhibition, as well as measures of anxiety and guilt. Four groups were found, two of which were non-psychopathic. Consistent with theory, however, two distinct classes of youth scoring high on psychopathic traits were identified: a primary variant with below-average levels of anxiety and guilt, and a secondary variant with markedly above-average levels. Youth in the latter category also presented with the highest levels of psychopathology within the detainee sample. The ratio of males to females was 2:1 within the low-anxious psychopathy variant, and 1:2 in the high-anxious psychopathy variant. Implications for identification and treatment of adolescent psychopathic variants are discussed.

Psychopathy describes a severe personality disorder, comprised of a constellation of socio-maladaptive traits. Its manifestations are interpersonal (e.g., deceitfulness, manipulation, superficial charm), affective (e.g., callousness, low emotionality, egocentricity), and behavioral (e.g., impulsivity, irresponsibility, recklessness) (Hare, 2006). Combined, these traits tend towards the violation of rules and norms, and situate individuals on the margins of society. Early identification of psychopathy has long been important to researchers and justice system stakeholders, alike, because while individuals high in psychopathic traits represent just 1% of the general population, they represent an estimated fifth of those in prison, commit a disproportionate percentage of violent crimes, and are at least twice as likely as other offenders to recidivate (Murrie & Cornell, 2002; Singh, Grann, & Fazel, 2011). While at the symptomlevel, adolescents with psychopathic traits may appear to be homogenous, researchers have long observed apparent etiological differences within the group. This study aims to further understanding of the etiological and symptomatic heterogeneity present in adolescent psychopathy.

Primary and Secondary Psychopathy Variants

The first influential conceptualization of psychopathy (Cleckley, 1941) described psychopaths, at first glance, to “embody a happy, well-adjusted person.” However, it was quickly clear that not all individuals with psychopathic traits were captured by Cleckley’s charismatic characterization, and so scientists began to parse that single phenotype into two heterogeneous groups. These have alternately been termed “idiopathic” vs “symptomatic,” (Karpman, 1941), “inherent” vs “acquired” (Porter, 1996), and, most commonly, “primary” vs “secondary” (Skeem, Poythress, Edens, Lilienfeld & Cale, 2003). Each dichotomy is distinguished by levels of trait anxiety, as well as, relatedly, indications of disparate etiologies (Skeem et al., 2007). Primary psychopathy is characterized by supernormal emotional detachment (including low trait anxiety), an inability to feel empathy or guilt, average to above-average cognitive abilities, and a disinclination to suicide (Fowles & Dindo 2006; Kimonis et al., 2011; Lilienfeld et al., 2012). Conversely, secondary psychopathy is associated with high anxiety, emotional and behavioral lability, lower cognitive ability, on average, as well as high risk for psychopathology (Hicks et al., 2004; Salekin, Neumann, Leistico, & Zalot, 2004).

One prominent etiological explanation for primary psychopathy, supported by laboratory biobehavioral tasks, is that constitutional deficits in reacting to fear-inducing stimuli (including punishment) leads to risk-taking, social fearlessness (expressed as dominance), and failure to internalize punishment (expressed as guiltlessness) (Berg et al., 2013; Lykken, 1995). By contrast, secondary psychopathy is thought to depend more on an environmental component, consistent with the observation that this highly neurotic psychopathic variant tends to present with high rates of early abuse and neglect. Here, it is proposed that adversity results in a “pathological adaptation” (Lansford et al., 2006) of “acquired callousness” (Kerig & Becker, 2010), designed to dissociate the individual from traumatic experience. Combined with high dispositional disinhibition, secondary psychopathic individuals may come to manifest the same callousness, prosocial impairment, and irresponsible risk-taking behavior as those with the primary variant, but without the same low-fear and blunted reactivity which forebear distress (Del Gaizo & Falkenbach 2008; Lykken, 1995; Porter, 1996). These distinctions in the etiology of psychopathy likewise implicate distinct deficits and possible motivations for change, and so may hold the key to developing effective interventions (Skeem et al., 2003).

Primary and secondary variants have consistently been found in adult incarcerated populations (Hicks et al., 2004; Newman, MacCoon, Vaughn & Sadeh, 2005; Hicks, Vaidyanathan, & Patrick, 2010; Poythress et al., 2010), but research has only recently explored these variants in youth. Early studies first identified youth who scored high on psychopathy or on callous-unemotional traits; then, using cluster analysis, split youth into two categories of “primary” and “secondary” variants based on a measure of anxiety or worry, and/or a measure of depression, and/or a measure of trauma (Kimonis, Skeem, Cauffman, & Dmitrieva, 2011; Kimonis, Frick, Cauffman, Goldweber & Skeem, 2012; Tatar, Cauffman, Kimonis, & Skeem, 2012). For instance, Kimonis and colleagues (2011) examined a sample 200 male juvenile detainees between the ages of 14 and 17. Youth with scores in the psychopathic range on the Psychopathy Checklist-Youth Version (PCL-YV) were then split using model-based clustering (Banfield & Raftery, 1993) on each subdomain of the PCL-YV and the Revised Children’s Manifest Anxiety Scale (RCMAS; Reynolds & Richman, 1985). They found two clusters that scored comparably on the PCL-YV, but differed significantly on scores of anxiety and worry. These two clusters cohered with the hypothesized external correlates of each variant, in that “secondaries” reported more extensive abuse histories than “primaries,” and scored higher on indices of psychological immaturity, depression, interpersonal aggression, and global psychological distress. Such studies suggest that primary and secondary variants do exist in (male) juvenile detainee populations, and that their differentiation unmasks other salient psychosocial features which would otherwise be obscured if psychopathy were treated as a uniform diagnostic category.

Although robust, there are several limitations to much of the early work on primary and secondary variants of psychopathy in adolescents; these, in turn, underscore the need for additional research. First, consistent with the field’s increasing appreciation of psychopathy as a dimensional (rather than taxometric) phenomenon (Murrie, Marcus, Douglas, Zee, & Salekin, 2007), there is a need to examine samples that span the continuum of psychopathic traits-not just those pre-selected for scoring above a clinical cutoff on measures of psychopathy. Second, there has also been a move toward increasingly sophisticated analyses that require no a priori assumptions about the data, but rather seek to find the number of clusters that best fits the data (Mokros et al., 2015; Wareham et al, 2009). Using these enhanced methodologies, several studies have failed to find primary and secondary variants among youth, or else have proposed alternative subtyping structures (Colins, Fanti, Salekin, & Andershed, 2017; Wareham, Dembo, Poythress, Childs & Schmeidler, 2009; Lee, Salekin, Iselin, 2010). Further research is needed to build consensus around the number and contours of adolescent psychopathy profiles.

A final significant limitation of previous research is that nearly all psychopathic subtyping has occurred in all-male detainee samples. While this is consistent with the observation that psychopathy is more common in males than in females (Cale & Lilienfeld, 2002), the rate of female involvement in juvenile detention is rising faster than that of males, and has recently reached an all-time high (Teplin et al., 2002). Compared to adolescent male psychopathy, adolescent female psychopathy is poorly understood, and additional studies are needed to better document both the phenomenon, and the unique risk factors with which its etiology and trajectory are associated (Hicks et al., 2010; Schrum & Salekin, 2006).

Measuring Psychopathy in Adolescents

In 2009, Patrick and colleagues departed from the practice of “downwardly extending” adult assessments of psychopathy, and reconceptualized psychopathy in terms of more developmentally stable traits, “more elemental phenotypic constructs with clearer psychological and neurobiological referents.” This developmentally informed reconceptualization yielded the triarchic model of psychopathy: Meanness measuring callous, hostile, and exploitative traits,Boldness indexing fearlessness, social dominance, and low stress reactivity, and Disinhibition assaying behavioral control deficits and irresponsibility. These three factors are measured with the Triarchic Psychopathy Measure (TriPM) (Patrick, 2010). Disinhibition is thought to correspond to dysfunction in the regions of the brain important for behavioral control, including the orbitofrontal cortex, whereas Boldness is related to deficiencies in the paralimbic system, including reduced sensitivity in the amygdala to emotional stimuli (Patrick & Bernat, 2009). Meanness is thought to arise, like Boldness, from a fearless temperament, in conjunction with other genetic and environmental (e.g., rearing) influences that result in impaired, aggressive social relating (Larsson, Viding, & Plomin, 2008; Patrick, Drislane, & Strickland, 2012). TriPM Meanness is akin to, and correlates with, the affective and interpersonal facets of the PCL-YV, and callous-unemotional traits on the Antisocial Process Screening Device (Frick & Hare, 2001), whereas Disinhibition corresponds to the lifestyle factors on the PCL-YV (Patrick, 2010).

The TriPM most differs from other measures of psychopathy in its assessment of Boldness. This dimension highlights the adaptive correlates of psychopathy (including superficial charm and low anxiety), shifting the assessment away from an arguable over-reliance on antisocial behavior, and closer to Cleckley’s (1941) original conceptualization of psychopathy as a “mask of sanity” (Berg, Lilienfeld, & Selbom, 2017; Drislane, Patrick, & Arsal 2014). The TriPM is an attractive option for assessment of at-risk adolescent populations, both because of the developmentally-driven approach to its creation, as well as its ease to administer. While much research on youth psychopathy variants has favored the PCL-YV, this measure can be labor-intensive to administer: it requires both a semi-structured interview and records review (Murrie & Cornell, 2002). Also, as others have noted, adolescents have had less time than adults to amass records reflective of their predispositions; if a self-report measure like the TriPM can demonstrate solid construct or predictive validity within this population, it would be a feasible measure to incorporate into standard assessments of juvenile detainee risk.

The Present Study

The present study aims, firstly, to identify primary and secondary variants of psychopathy in a diverse northeastern sample of male and female adolescent detainees. Secondly, we aim to examine the associations between these psychopathic variants and psychopathology. Finally, this study aims to contribute to research on adolescent detainee psychopathy by examining, for the first time, the ability to detect primary and secondary variants using the TriPM.

Consistent with previous studies of psychopathy in juvenile detainees, we anticipate that a minority subset of youth will endorse high levels of psychopathic traits as indexed by the TriPM (Somma, Borroni, Drislane, & Fossati, 2016). Furthermore, we anticipate that by examining TriPM scores in conjunction with self-reported measures of anxiety and the experience of guilt, we will be able to distinguish youth displaying primary and secondary psychopathic traits both from one another, and from the rest of the sample. Youth with primary traits will expectantly score high on psychopathy, but presumably quite low on measures of anxiety and guilt. Youth with secondary traits, however, will present with high psychopathy scores, but also high scores on indices of anxiety and guilt.

We predict that, relative to youth with primary psychopathic traits, those with secondary traits will report more mental health impairments, as indexed by substance use, trauma, anger and irritability, and suicidal ideation, and thus will represent a group uniquely at risk within the juvenile justice system (Pennington, Cramer, Miller, & Anastasi 2015; Vaughn et al., 2009). By contrast, and consistent with their essential unemotionality, youth with primary traits are expected to be inured to internalizing mental health disturbances (Lilienfield et al., 2012). Consistent with the hypothesized etiology of secondary psychopathy, and in light of the higher rates of trauma and neuroticism generally observed among adolescent females, we anticipate that a greater proportion of youth with secondary traits will be females (Gill & Stickle, 2016; Ribeiro da Silva, Rijo, & Salekin, 2012).

Methods

Participants and Settings

Participants included 89 girls and 91 boys (including one transgender male) involved in the juvenile legal system in a large metropolitan area (12–18 years-old, M = 15.0, SD = 1.0). Using non-exclusive categories, 63.4% percent of youth identified as Black, 36.6% percent as Hispanic/Latino, 20.9% as Caribbean, 7.5% percent as White, and 5.2% percent as Asian or Asian-American. Approximately a third (32.6%) of youth identified as sexual minorities; all youth so-identifying were female. Because demographic information was not available from 46 youth (25.6% of the sample), examination of demographics in subcategories of youth was limited. While socioeconomic status information is not available for the present sample, elsewhere it has been documented that juvenile-justice involved youth come disproportionately from low-income communities (Birckhead, 2012).

Data were collected from youth housed in eight short-term detention facilities (STDFs) in a large American city. This city is one of several nationwide to have merged its child welfare and juvenile justice systems in response to the exigent need for inter-system collaboration. These STDFs function as short-term residential facilities that house adolescents while they await Family Court decisions regarding long-term housing (e.g., foster care) or sentences for violent and non-violent juvenile offenses; all youth had committed, at a minimum, a status offense. To ensure an adequate sample of females, all four STDFs serving girls were selected; four boys’STDFs matched based on locality were also selected for inclusion. All youth in participating facilities were invited to participate.

Procedures

The current study is a cross-sectional, de-identified, secondary analysis of baseline data originally collected to track individual youths’ progress in pilot group therapy treatments. Permission was granted by the local juvenile legal system for the team to invite youth at residential centers to participate in the group therapy intervention and assessments. All assessments were conducted in group settings, with assessment team members and facility staff present. To respect youth’s privacy and ensure their confidentiality, participants were asked to sit apart from one another. Each assessment session lasted between 30 and 90 minutes, depending on the number of youth present and their sustained attention. Snacks were provided to all youth in the facilities regardless of participation, and alternative materials (i.e., word searches) were offered to individuals who declined survey participation. The research protocol was reviewed by relevant Institutional Review Boards, and the collaborating government agency granted permission for the team to invite youth at participating facilities to enroll in the intervention and assessments.

Measures

Psychopathic Traits.

Psychopathic traits were assessed using the self-report Triarchic Psychopathy Measure (TriPM; Patrick, 2010), which is comprised of three subscales measuring conceptually distinct components of psychopathy: Boldness (“I’m a born leader”), Meanness (“I don’t mind if someone I dislike gets hurt”), and Disinhibition (“I jump into things without thinking”) (Patrick, 2010). Each of the measure’s 58 items is rated by the respondent as true, somewhat true, somewhat false, or false (Patrick, 2010b). The three subscales are scored independently by summing all coded responses for each (Boldness M = 24.36, SD = 10.49, Min/Max = 0/57; Meanness M = 22.20, SD = 11.57, Min/Max =0/57; Disinhibition M = 26.40, SD = 12.44, Min/Max = 0/60). A total psychopathy score is achieved by summing the three subscale totals (M = 72.97, SD=32.26, Min/Max = 0/174.0) (Patrick, 2010). In the present study, the subscales Boldness, Meanness, and Disinhibition showed strong internal consistency (Cronbach’s alpha = 0.83, 0.86, and 0.87, respectively).

Inconsistent Responding.

Inattentive and careless responding was assessed using the Triarchic Assessment Procedure for Inconsistent Responding (TAPIR; Mowle et al., 2017), which involves comparing responses from 13 pairs of similar TriPM items. We calculated a total score by summing the absolute difference between scores on paired items, such that higher scores indicate more inconsistent responding. Based on the correctional sample findings of Mowle et al. (2017), we adopted a cut score of 17 (corresponding to 90.7% specificity) for inclusion in analyses. A total of 18 participants were excluded for inconsistent responding (final N =162).

Mental Health.

Mental health indicators were assessed by the Massachusetts Youth Screening Instrument-Version 2 (MAYSI-2), a standard, self-report survey designed to identify emotional distress in adolescents during entry or transition within the juvenile justice system (Grisso & Barnum, 2006). The instrument is written at a fifth-grade reading level, and consists of 48 yes-or-no items regarding “the past few months.” The following five subscales were used in this study to assess mental health symptoms: Substance Use (e.g., “Have you done anything you wish you hadn’t when you were drunk or high?”); Angry/Irritability (e.g., “Have you lost your temper easily, or had a ‘short fuse’?”); Anxious/Depressed (e.g., “Have nervous or worried feelings kept you from doing things you want to do?”); Suicide Ideation (e.g., “Have you wished you were dead?”); and Traumatic Experiences (e.g., “Have you ever been badly hurt or in danger of getting badly hurt or killed?”). The Traumatic Experiences subscale traditionally includes one gender-specific item, but due to concerns about construct validity for the male item (“Have people talked about you a lot when you’re not there?”), the gender-specific items were removed. Items rated yes are summed to yield subscale scores, which serve to illuminate areas of risk (Substance Abuse M = 2.09, SD = 2.13; Anxious/Depressed M = 1.84, SD = 2.07; Anger/Irritability: M =3.47, SD = 2.84; Suicidal Ideation M = .44, SD = 1.10; Trauma M = 1.34, SD = 1.50) (Grisso & Barnum, 2006). Scores are non-diagnostic, but rather are used by the juvenile justice system to identify youth requiring further assessment; specifically, developers established a “caution” cutoff based on the clinical cutoff for the Millon Adolescent Clinical Inventory and the Youth Self-Report, as well as a “warning” cutoff, set at the scores of the highest 10% of responders. In the present study, 22.2% of youth hit the caution or warning further assessment cutoff on the Substance Abuse scale, 40.1% on Anger/Irritability, and 9.9% on Suicidal Ideation. MAYSI subscales’ alpha coefficients ranged from .75 (Trauma) to .87 (Anger/Irritability).

Guilt.

The State Shame and Guilt Scale – Revised (SSGS-R) is a 15-item self-report scale (Marschall, Sanftner, & Tangney, 1994). We used the Guilt subscale (e.g., “I feel remorse, regret”), because we expected it to distinguish primary and secondary psychopathic variants (Lilienfeld et al., 2012). Items are rated on a 5-point Likert scale from 1 (“not feeling this way at all”) to 5 (“feeling this way strongly”), yielding scores on a scale from 5 to 25 (Guilt: M=11.17 SD=5.85). The Guilt subscale had good internal consistency (Cronbach’s alpha of .84).

Data Analyses

Latent Profile Analysis.

To detect youth endorsing patterns consistent with primary and secondary psychopathy, a latent profile analysis (LPA) was performed using TriPM Boldness, Meanness, and Disinhibition scales; the MAYSI-2 Depression/Anxiety scale; and the SSGS-R Guilt scale. We evaluated latent class models using the Bayesian Information Criterion (BIC) (smaller values indicate better model fit), discrimination among the classes (entropy values), and the Lo-Mendell-Rubin–adjusted likelihood ratio test (Lo, Mendell, & Rubin, 2001), where a significant p-value indicates that a model with k number of classes is preferred over a model with k-1 classes. Latent profile analyses were performed in Mplus 8 using the maximum likelihood robust estimator (Muthén & Muthén, 2013). The scales used in the LPA were not excessively skewed or kurtotic (all values were between 1 and −1).

After identifying the classes, they were examined to assess hypothesized external correlates using chi-square tests and analysis of variance (ANOVA) with Bonferroni-corrected p-values. All analyses were conducted using SPSS Version 26. The maximum item-level missingness on the self-report scales was 5.0%, and these missing data were replaced using multiple imputation methods. Because demographic data (apart from gender) were missing for 48 subjects, analyses other than gender were not pursued for these variables.

Results

Latent Profile Analysis: Identifying Variants of Psychopathy in Youth

LPA was used to examine whether the TriPM subscales, along with measures of internalizing symptoms and guilt (MAYSI-2 Depression/Anxiety and SSGS-R Guilt subscales), could be used to identify discrete psychopathy profiles among juvenile detainees. Model fit for solutions with two to five latent classes were examined, and the results are presented in Table 1.¹ The Lo-Mendell-Rubin–adjusted LRT indicated that models with 2 and 4 classes showed improved fit over those with one fewer class. The four-class solution was ultimately selected as the best fitting model, because it had a lower BIC value than the two-class solution. It also conformed with hypotheses, consistent with previous research, that some youth with psychopathic traits are characterized by markedly low levels of anxiety and social emotions, whereas others tend to experience these acutely (Kimonis, et al., 2012; Newman et al., 2005).

Table 1.

Model Fit of the Latent Profile Analyses for Juvenile Detainees.

No. latent classes	BIC	Adjusted BIC	Adjusted LMR LRT p	Bootstrap LRT p	Entropy
2-Class	5307.10	5256.44	0.004	<0.001	0.92
3-Class	5194.10	5124.45	0.067	<0.001	0.86
4-Class	5158.95	5070.31	0.045	<0.001	0.89
5-Class	5152.37	5044.73	0.821	<0.001	0.83

Note. BIC = Bayesian Information Criterion. LMRA-LRT p = Lo-Mendell-Rubin–adjusted Likelihood Ratio Test p-value; LRT = Likelihood Ratio Test.

As illustrated in Figure 1, the first class was characterized by low ratings on all scales (“Low/No Psychopathic Traits,” 12.9% of the sample, n = 21); the second class was characterized by high TriPM scores, but low internalizing symptoms and guilt (“Primary Psychopathic Traits”, 17.2% of sample, n = 28); the third class was marked by the most common response pattern (i.e., moderate levels) (“Average Psychopathic Traits”, 51.2% of the sample, n = 83); and the fourth class was marked by high TriPM scores as well as high endorsement of internalizing symptoms and guilt (“Secondary Psychopathic Traits”, 18.5% of the sample, n = 30). Average latent class probabilities ranged from .93 (Primary Traits) to .99 (Low), suggesting that individuals were well classified by this model.

Figure 1. Latent Profiles of Psychopathy among Juvenile Detainees.

Note. Disinhibition, Boldness, and Meanness were measured by the TriPM. Guilt was assessed by the SSGS. Depression/Anxiety was measured by the MAYSI-2. Letters denote the results of post-hoc pairwise comparisons. A= Low and Average groups differ significantly. B = Low and Primary groups differ significantly. C = Low and Secondary groups differ significantly. D = Average and Primary groups differ significantly. E = Average and Secondary groups differ significantly. F = Primary and Secondary groups differ significantly. Z-score values were calculated with the sample means and standard deviations.

Examination of Class Differences

Primary versus Secondary Variants.

First, we examined differences between youth exhibiting primary and secondary trait profiles. Chi-Square tests revealed that youth with secondary psychopathic traits (63.3%) were more likely to be female than those with primary psychopathic traits (32.2%) (χ²=5.64, p =.018). The primary and secondary variants did not differ from each other on overall levels of psychopathic traits, though youth in the primary group scored higher on Meanness (M = 36.59, SD = 6.85), but not Boldness or Disinhibition, than those in the secondary group (M= 30.78, SD = 5.90) (F_(1,56) = 12.02, p =.014).

As expected, youth with secondary traits presented with significantly higher scores on the MAYSI-2 subscales of Trauma, Suicidal Ideation, Anger and Irritability, and Substance Abuse than youth with primary traits (Table 2). As previously noted, there are established “caution” and “warning” cutoffs on the MAYSI-2 in all areas except for Trauma, to designate youth with particularly high mental health needs. A significantly higher percentage of youth displaying secondary psychopathic traits reported clinically-significant levels of suicidal ideation (36.7% vs. 3.6%, χ²=9.66, p =.002) and anger and irritability (90.0% vs. 35.7%, χ²=18.47, p <.001) than youth characterized by the primary variant. In contrast, the two psychopathy variants did not differ in terms of the percentage requiring follow-up for substance abuse (50.0% secondary vs.35.7% primary, χ²=1.21, p =.27) (Figure 2).

Table 2.

Mental Health Indicators by Latent Profile

	Non-Psychopathic Variant	Primary Psychopathic Variant	Secondary Psychopathic Variant	Pairwise Comparisons between Classes Bonferroni-corrected p-values
	(n = 104)	(n = 28)	(n = 30)	Non-Psychopathic vs Primary	Non-Psychopathic vs Secondary	Primary vs Secondary
Total Psychopathy (M/SD)	56.15 (26.25)	107.23 (16.69)	99.30 (14.28)	<.001	<.001	.577
Trauma Exposure (M/SD)	0.86 (1.24)	1.28 (1.27)	3.03 (1.35)	.369	<.001	<.001
Anger/Irritability (M/SD)	2.56 (2.58)	3.65 (2.70)	6.47 (1.40)	.111	<.001	<.001
Substance Abuse (M/SD)	1.30 (1.61)	2.88 (2.13)	4.09 (2.21)	<.001	<.001	.039
Suicidal Ideation (M/SD)	0.18 (0.61)	0.21 (0.95)	1.55 (1.74)	1.00	<.001	<.001

Note. Absent Bonferroni correction, the Non-Psychopathic vs. Primary Traits comparison for Anger/Irritability, and the Primary vs. Secondary Traits comparison for Total Psychopathy showed trends toward significance (p = .052, p = .056, respectively). Results do not change if the “Low/Non-Psychopathic” and “Average” classes are analyzed separately rather than combined.

Figure 2. Percentage of Youth Requiring Further Psychological Assessment by Class.

Note. The above figure reflects the percentages of youth scoring above the “Caution” or “Warning” cutoff on the MAYSI-2 subscales (Trauma not shown, because it does not have corresponding cutoffs). Non-psychopathic group includes both “Low” and “Average” classes. * denotes a significant difference (p<.05).

Non-Psychopathic vs. Psychopathic Variants.

Because we were primarily interested in examining differences in the youth displaying the two psychopathic-trait variants, first from one another and then from the broader detainee population, we collapsed the first two classes (“Low” and “Average”) into a single, non-psychopathic comparison group.

Gender differences were not observed between youth classified as non-psychopathic (female: 48.1%) and those with secondary psychopathic or primary psychopathic traits (χ²s < 3.46, ps >.063). As expected, youth displaying primary and secondary traits scored higher overall on the TriPM than the detainees in the non-psychopathic comparison group (Table 2). Youth with secondary traits also displayed more mental health impairments than non-psychopathic youth, on every assessed dimension (Table 2). Youth with psychopathic traits of either type scored higher than non-psychopathic youth on substance abuse, but in every other psychopathological domain (i.e., suicidal ideation, anger and irritability, and trauma exposure), youth with primary traits were not different from the non-psychopathic detainee population.

Discussion

In this study, we aimed to demonstrate that juvenile psychopathy is a heterogeneous phenomenon that can be differentiated by the capacity to experience trait anxiety and guilt. Using latent profile analysis (LPA), we identified four groups that varied on these emotions, as well as on a three-factor measure of psychopathy. Two of these groups were non-psychopathic, while two groups fit the profiles of primary and secondary psychopathic variants. Upon examining external correlates, we found, consistent with previous literature, that the high-anxious secondary variant presented with high levels of psychopathological impairment (Lee et al., 2010; Kimonis et al., 2011; Gill & Stickle, 2016). Meanwhile, the primary variant was psychiatrically indistinguishable from the non-psychopathic groups, except on the index of substance abuse. This study represents an important elaboration of previous work, extending the findings of adolescent psychopathic variants to a sample that includes an equal proportion of female juvenile detainees. Moreover, we verified the existence of these variants using a relatively new, developmentally-informed triarchic measure of psychopathy (i.e., the TriPM).

To our knowledge, this is the first study to apply LPA to the detection of psychopathic variants and their psychopathological correlates in a diverse, equally male and female sample of juvenile detainees. Consistent with theory and research which have proposed that psychopathy is a dimensional construct, we identified a model best fit by four distinct groups that varied in severity of psychopathic traits (e.g., Colins et al., 2017; Docherty et al., 2016; Waller & Hicks, 2019). These groups included two non-psychopathic groups that showed low and average levels of psychopathic traits, as well as a group high in psychopathic traits but low in anxiety and guilt (i.e., a primary psychopathic variant), and a group high in psychopathic traits, but also with the highest levels of anxiety and guilt (i.e., a secondary psychopathic variant).

After collapsing the low and average clusters into a single non-psychopathic comparison group, we examined the groups’ mental health correlates. Among the high psychopathy variants, youth in the secondary group were more susceptible to mental health disturbances, having on average the highest levels of psychopathology on every assessed dimension compared to their detained peers. Not only did individuals with the secondary variant score dimensionally higher on measures of trauma, anger and irritability, substance abuse, and suicidal ideation, but they also more frequently scored above clinically meaningful cutoffs (Figure 2). These youths’ scores on the MAYSI-2 qualified all of them for further risk-assessment. However, their scores on a measure of psychopathy significantly situates their mental health impairments in the biopsychosocial context of affective-interpersonal and impulse-control deficits, which could have important ramifications for treatment strategies, over and above those for psychopathological symptomatology, alone. That is, understanding that these youth are not only at high psychiatric risk, but are also presenting with psychopathic characteristics that carry their own set of etiological and lifetime-risk implications, could lead to better targeted interventions that take these additional, durable vulnerabilities into account. This finding furthers previous research calling for multi-dimensional treatment targets for justice-system involved youth (and girls, in particular) (Leve, Chamberlain, & Reid, 2005).

Notably, while youth with secondary traits had higher average overall substance abuse scores than those with primary traits, both groups met the clinical cutoff requiring further assessment with equal frequency. This indicates that, even if there are differences in the motivation (e.g., approach vs. avoidance) or acuity of substance use between variants, they may appear diagnostically indistinct. Previous studies have found mixed evidence regarding substance use severity in primary and secondary psychopathy. Most adult studies have supported higher rates and riskier patterns of use in secondary than primary variants (Del Gaizo & Falkenbach, 2008; Vassileva, Kosson, Abramowitz, & Conrod, 2005), but less work has been done in adolescent samples. For instance, correlations between PCL-YV scores and substance use levels in adolescents are inconsistent, with some studies supporting the association among male detainees (e.g., Mailloux, Forth, & Kroner, 1997), and others finding no relation (O’Neill, Lidz, & Heilbrun, 2003). Moreover, the association between PCL-YV scores and substance use indices has been weaker for females than for males, with one study of 80 female offenders finding relatively weak correlations with alcohol abuse (r = .22), and no correlation with drug use (Bauer, Whitman & Kosson, 2008). Our findings, then, are responsive to a perennial debate about whether adolescent psychopathy presents differently from adult psychopathy; these results regarding clinically equivalent rates of substance use between primary and secondary variants suggest a developmental difference, and replicate another recent finding (Waller & Hicks, 2019). This difference may reflect the simple distinction that adolescence is marked by higher rates of substance use experimentation, overall, or it may point to the greater control exerted by environmental forces in the life of the adolescent. That is, adolescents’ substance use may be more likely than adults’ to be influenced by environmental factors, either protective (e.g., close parental monitoring) or harmful (e.g., deviant peer associations, parental substance use) (Petraitis, Flay & Miller, 1995; McGue, Elkins, & Iacono, 2000). Future research should assess substance use longitudinally into adulthood, to determine when and why substance use patterns between primary and secondary variants diverge.

This study also replicates the finding that secondary psychopathic traits are more prevalent in females than in males (Sevecke, Lemkuhl, & Krischer, 2009). Our findings are likewise consistent with the observation that female adolescents in contact with the justice system present with higher rates of trauma and severe psychological distress than males (Abram, Teplin, & McClelland 2003; Dixon, Howie, & Starling 2008; Falkenbach, Reinhard, & Larson, 2017). However, unlike a recent study which incorporated adolescent female detainees from a predominantly white, rural area (Gill & Stickle, 2016), we did not find that the secondary, majority-female variant scored significantly lower on all indices of psychopathy. Additional studies incorporating adolescent females and using multiple measures of psychopathy are needed to resolve this inconsistency. Moreover, in this study, within an evenly split sample, females made up one-third of those with primary traits, and males made up one-third of those with secondary traits. This finding both cautions against gendered overgeneralizations regarding the composition of psychopathic variants, as well as highlights the need for larger samples, which will facilitate further study of the features of the underrepresented gender within each variant.

In addition to reporting fewer mental health problems than youth in the secondary psychopathic group, adolescents with primary psychopathic traits were indistinguishable from non-psychopathic youth on these outcomes. Specifically, they were not different from non-psychopathic youth on measures of trauma, anger and irritability, depression, or suicidal ideation, which is consistent with predictions and previous research (Patrick, 2010; Sevecke Lehmkuhl, & Krischer, 2009). This indistinguishableness also coheres with the “Mask of Sanity” originally proffered by Cleckley (1941) as a demarcating characteristic of psychopathy. However, the identification of this psychological profile bears implications for intervention, as well. Although youth in the primary group would not be singled out for treatment in juvenile detentions centers by the standard administration of the MAYSI-2, a large body of literature suggests that they are at high risk for poor long-term social outcomes (Loney, Taylor, Butler, & Iacono, 2007; Moffitt, 1993). Their apparently intact mental health points to the need for the justice system to assess personality disturbances in addition to traditional internalizing and externalizing psychopathology. However, a great deal more research will be needed to assess which interventions may hold promise for reducing recidivism among primary psychopathic adolescents.

Notably, this study also represents the first use of the TriPM to detect primary and secondary adolescent psychopathic variants. As previously stated, the TriPM is a particularly promising tool for the identification of youth psychopathy (Patrick, 2010; Somma et al., 2016), both for its theoretical focus on developmentally stable constructs of Meanness, Boldness, and Disinhibition, and for its ease of administration compared to the PCL-YV. Indeed, the TriPM did successfully facilitate the identification of these psychopathic variants in a diverse sample of male and female adolescent detainees. Given the lack of previous research examining the psychopathy subtypes using the TriPM, we did not make specific predictions about potential differences between the primary and secondary variants on the individual TriPM subscales. Though studies have found that Boldness is moderately, negatively correlated with negative emotionality and positively with adaptive functioning (Miller & Lynam, 2012; Miller, Sleep, Crow, & Lynam, 2019), we did not observe significant mean differences between primary and secondary subtypes of psychopathy on this subscale. This finding is somewhat surprising given the elevated levels of mental health problems observed in the secondary psychopathy cluster, but is consistent with prior work that has failed to show differences between the two psychopathy variants on measures that index boldness/fearless dominance (Poythress et al., 2010). Differences were observed between the psychopathy subtypes on the Meanness subscale, with primary variant scoring higher than the secondary variantan interesting finding to explore in future studies.

There are several strengths of this study—primarily its use of a diverse sample, and its inclusion of an understudied population—adolescent females— into research on psychopathic variants. While its self-report mono-methodology represents a weakness in one sense, particularly in a sample in which deception and reliability are at issue — the study also has several methodological strengths. We have already noted the ease of administration of the TriPM. It is also notable that the MAYSI-2, from which we derived our index of anxiety, is already a standard instrument in juvenile detention centers nationwide, and our measure of guilt was derived from a brief (15-item) measure. In short, the measures of psychopathy-variant detection used herein could easily be deployed within the infrastructure already established within the juvenile justice system. While replication would be needed to establish reliable quantitative clustering criteria, such a possibility holds potential promise for those who would pursue early, targeted interventions with psychopathic youth (Lynam, 2002; Salekin, 2002; Salekin, Worley, & Grimes, 2010).

Because we lacked detailed detainee demographic data and information about specific reasons for detention, we could not explore the construct validity of clusters according to their offenses. Moreover, due to the cross-sectional design, we were unable to assess the predictive validity of clusters in relation recidivist trajectories, nor can we draw definitive causal inferences. Because this is a juvenile justice context, we might have hypothesized differences between psychopathic variants in terms of the nature or seriousness of their violations. For example, we might have anticipated more instrumental offenses by primary psychopathic youth, and more impulsive acts among secondary psychopathic youth; moreover, it would have been interesting to probe gender differences on this dimension. Sample size limitations prevented us from performing LPAs on boys and girls separately. To extend these findings, it would be instructive to expand male and female sample sizes, and to examine the offense patterns committed by these psychopathic variants, their developmental trajectories over time, as well as how they may respond differentially to treatment.

In sum, we provide evidence for the distinction between primary and secondary psychopathy among a diverse sample of male and female adolescents. We thereby contribute to a growing body of literature acknowledging that adolescent psychopathy (i) is an early detectable, heterogeneous phenomenon, and (ii) that it has meaningful external correlates with likely implications for mental health intervention. Because we included girls, who are often excluded from similar psychopathy research, we were further able to show that they represent a group at particular risk, as they are most likely to populate the intersection of juvenile delinquency and secondary psychopathy. Present findings contribute to the literature on psychopathic variants in youth and delineate novel relationships between these variants and mental health outcomes. Moreover, they suggest that treatment strategies for adjudicated youth would be aided by brief assessment tools to reliably differentiate primary and secondary psychopathy variants.