Journal of Adolescence 37 (2014) 681–689

Contents lists available at ScienceDirect

Journal of Adolescence

journal homepage: www.elsevier .com/locate/ jado

Under the surface of adolescent psychopathic traits: High-anxious and low-anxious subgroups in a community sampleof youths

Selma Salihovic*, Margaret Kerr, Håkan StattinÖrebro University, Sweden

a r t i c l e i n f o

Article history:Available online 27 March 2014

Keywords:AnxietyPsychopathic traitsPrimarySecondarySubgroupsAdolescents

* Corresponding author. Center for Developmentafax: þ46 (0) 19 301 421.

E-mail address: selma.salihovic@oru.se (S. Saliho

http://dx.doi.org/10.1016/j.adolescence.2014.03.0020140-1971/� 2014 The Foundation for Professionals

a b s t r a c t

In this study, we examined subgroups of adolescents based on their levels of psychopathictraits and anxiety. Participants were 914 youths from a community sample, with a meanage of 14.28 (SD ¼ .94) years. We used adolescents’ self-reports of psychopathic traits andtheir parents’ reports of the adolescent’s anxiety to identify distinct subgroups of youths.Using latent class analysis, we identified five groups that varied in levels of psychopathictraits and anxiety. Two groups were characterized by high levels of psychopathic traits andhigh or low scores on anxiety. Validation of these subgroups revealed that they differedsignificantly from each other in theoretically meaningful waysdthe low-anxious subgroupreported higher levels of psychopathic traits, lower levels of impulsivity and hyperactivity,and lower levels of aggression than the high-anxious group. These findings are in line withprevious empirical research and provide support that anxiety discriminates between twosubgroups of adolescents with psychopathic traits.� 2014 The Foundation for Professionals in Services for Adolescents. Published by Elsevier

Ltd. All rights reserved.

Psychopathy is a syndrome characterized by a combination of interpersonal, affective, and behavioral traits (Cooke &Michie, 2001). People high on psychopathic traits are described as callous, dishonest, manipulative, and impulsivedchar-acteristics that contribute to a lifestyle often characterized by antisocial behavior and violence. Over the past decade, scholarsinterested in understanding the etiology of serious antisocial behavior have focused on the expressions of the disorder inyounger populations in order to identify those at risk of becoming antisocial adults. There is today ample evidence thatpsychopathic traits can be identified in adolescents and that their presence is linked to antisocial behavior and violence bothconcurrently and prospectively (for reviews, see Forth & Book, 2010; Salekin & Frick, 2005; Salekin, Rosenbaum, Lee, & Lester,2009). There is also increasing support that youths with psychopathic traits are a heterogeneous group (e.g., Skeem,Poythress, Edens, Lilienfeld, & Cale, 2003). Hence, there is a need to identify, describe, and understand subgroups ofyouths with psychopathic traits.

Theoretically, it has been suggested that the development of psychopathy can originate from innate biological dispositionsas well as environmentally influenced factors (e.g., Cleckley, 1976; Lykken,1995; Karpman,1941). Karpman (1941) was amongthe first to propose a distinction between two subtypes of psychopathy. The primary subtype of psychopathy is hypothesized

l Research, JPS: Psychology, Örebro University, 701 82 Örebro, Sweden. Tel.: þ46 (0) 19 301 407;

vic).

in Services for Adolescents. Published by Elsevier Ltd. All rights reserved.

S. Salihovic et al. / Journal of Adolescence 37 (2014) 681–689682

to reflect individuals whose behavior is the result of a congenital emotional deficit, while the secondary subtype is believed tobe the result of an environmentally acquired emotional disturbance. Karpman (1941) argued that the difference in etiologyhad implication for both research and practice, because he believed that only the secondary type might be responsive totreatment efforts. Several researchers have further developed Karpman’s distinction and outlined models of primary andsecondary subtypes of psychopathy (Blackburn, 1998; Lykken, 1995; Mealey, 1995; Porter, 1996). These models generallyconverge with Karpman’s idea that both subtypes are similar in terms of showing high levels of antisocial behavior, but differin their etiological underpinnings. Thus, the body of theoretical literature suggests that psychopathy is a heterogeneousconstruct.

According to the abovementioned theories, the core difference between the primary and secondary type relates to anxiety.Compared to the primary type, secondary psychopaths are believed to be capable of experiencing and expressing emotionalreactionsdthey are anxious, impulsive, and they engage in reactive aggression and hostility (Skeem et al., 2003). Thisemotional instability is hypothesized to stem from environmental distress and is believed to result in an impulsive andaggressive behavioral style (e.g., Lykken, 1995; Skeem et al., 2003). The primary type, on the other hand, is described as theprototypical psychopath who is incapable of emotions, such as empathy and guilt, and so appears callous, fearless, andrelatively free of anxiety (e.g., Karpman, 1941). This combination of traits is believed to result in a stable personality stylecharacterized by proactive aggression and antisocial behavior (Skeem et al., 2003). In short, although equally prone toantisocial behavior, the key difference between the primary and the secondary subtype seems related to anxiety.

In keeping with this tradition, empirical findings have shown that primary and secondary subtypes can be identified inforensic settings among individuals scoring high on psychopathic traits (Hicks, Vaidyanathan, & Patrick, 2010; Newman,MacCoon, Vaughn, & Sadeh, 2005; Skeem, Johansson, Andershed, Kerr, & Eno Louden, 2007; Swogger & Kosson, 2007).Consistent with theoretical ideas, the high-anxious secondary subtype showed more aggression, reactive hostility, impul-sivity, and had more histories of violence and criminality than the low-anxious primary subtype (Hicks, Markon, Patrick,Krueger, & Newman, 2004). Moreover, there is some support to suggest that the primary subtype, compared to the sec-ondary, shows somewhat higher levels of core psychopathic traits (Newman et al., 2005; Skeem et al., 2007), which isconsistent with theoretical ideas. Given current knowledge then, it is possible to distinguish between variants of psychopathicoffenders that parallel theoretical descriptions of primary and secondary psychopathy.

Only a few studies have examined whether similar subgroups can be identified among adolescents. In some of thesestudies, measures of anxiety were used to identify subgroups of adolescents with extreme levels of psychopathic traits(Kimonis, Frick, Cauffman, Goldweber, & Skeem, 2012; Kimonis, Skeem, Cauffman, & Dmitrieva, 2011; Lee, Salekin, & Iselin,2010; Vaughn, Edens, Howard, & Smith, 2009; Veen, Andershed, Stevens, Doreleijers, & Vollebergh, 2011; Wareham,Dembo, Poythress, Childs, & Schmeidler, 2009). In support of the distinction, studies have shown that juvenile offenderswith high scores on psychopathic traits and anxiety reported more negative emotionality, attention problems, impulsivity,anger, childhood abuse, and were more affected by distressing emotional stimuli than their low-anxious counterparts(Kimonis et al., 2011, 2012). Similarly, juvenile offenders characterized with high levels of psychopathic traits and anxietyreport more psychiatric symptoms, drug use, delinquent behavior, and trauma history than the low-anxious group (Vaughnet al., 2009; Veen et al., 2011). Other findings are less conclusive, however. For instance, Lee et al. (2010) found that anxiety didnot discriminate between a primary and secondary subtype in a sample of male adolescent offenders. There was only supportfor a high-anxious group, which most closely resembled the secondary variant, and no low-anxious (i.e., primary) groupemerged.Warenham et al. (2009), on the other hand, found a low-anxious group of adolescent offenders but did not identify ahigh-anxious subgroup. In general, then, there is some, albeit inconclusive, empirical support that high-anxious and low-anxious subgroups can be identified among adolescents in forensic settings.

To our knowledge, no previous studies have examined whether the primary and secondary subgroups can be identifiedamong youth in normative community samples. Instead, previous studies have focused on institutionalized youth pop-ulations who are at the high end of the distribution of psychopathic traits (e.g., Kimonis et al., 2011; Lee et al., 2010; Vaughnet al., 2009; Veen et al., 2011; Wareham et al., 2009). This strategy, although informative for initial exploration, has severallimitations. First, only selecting individuals who meet a certain cut-off score restricts the variation in psychopathic traits andexcludes potentially important subgroups that, despite lower mean scores, may have similar profiles. This knowledge couldhelp identify factors that protect youths with less extreme levels of psychopathic traits from negative development, andfacilitate effective prevention strategies (e.g., Poythress & Skeem, 2006). Second, it is increasingly understood that psycho-pathic traits are distributed across a continuum rather than constituting a discrete category (for review, see Poythress &Skeem, 2006). This, in itself, suggests that there is variation in the expression of the traits, and is an argument for study-ing the full range in samples with normative distributions. Finally, in their review of the literature on subtypes of psycho-pathic traits, Poythress and Skeem (2006) concluded that community samples are ideal populations for studying the natureand prevalence of subgroups because they provide important information about both normative and non-normativedevelopment. Hence, the purpose of the current study is to examine whether subgroups similar to primary and secondarysubtypes can be identified in a community sample of adolescents.

Consistent with theoretical conceptualizations, we examinedwhether anxiety discriminates between subgroups of youthswith high levels of psychopathic traits. Notably, to the extent these subgroups are identified, we will refer to them as low-anxious and high-anxious subgroups rather than primary and secondary subtypes, in order to avoid pejorative labeling andconflation with the clinical construct of psychopathy. Based on previous research and theory, we expected that these sub-groups would differ significantly from each other onmeasures assessing aggression, delinquent behavior, attention problems,

S. Salihovic et al. / Journal of Adolescence 37 (2014) 681–689 683

and impulsivity. Specifically, we expected that the adolescents who scored high on both psychopathic traits and anxiety (i.e.,the high-anxious subgroup) would be more impulsive, have more attention problems, and show more reactive aggressionand delinquency, than the low-anxious subgroup. In contrast, we expected that adolescents with high levels of psychopathictraits and low scores on anxiety (i.e., the low-anxious subgroup) would have higher scores on proactive aggression than thehigh-anxious subgroup and show higher levels of core psychopathic traits, such as callous–unemotional traits, grandiose–manipulative traits, and impulsive–irresponsible behavior. Given the normative nature of our sample, we anticipated thatother groups, with varying levels of psychopathic traits and anxiety, would likely emerge but theywill not be characterized byhigh levels of psychopathic traits, delinquent behavior, and aggression.

Method

Participants and procedure

Participants were adolescents and their parents from a community-based study that took place in a city in Sweden (totalpopulation of about 26,000). The unemployment rate in the community was similar to that in the country as a whole (6%),whereas the mean income was about 4.0% lower than in the rest of Sweden (Statistiska centralbyrån [SCB], 2001). Twelvepercent of the inhabitants in the community had a foreign background, which is comparable to the 13% at the national level.For this study, students in all 7 through 9th grade classes in the city (N¼ 982) composed the target sample. Of the 982 youths,914 (93% of the target sample; 51% boys) were present on the day of the data collection and answered the questionnaires.

Adolescents filled out the questionnaires during regular school hours in sessions administered by trained research as-sistants. They were given information about the study and were informed that their participationwas voluntary. Adolescentstook part in the study unless their parents returned a postage-paid postcard stating that they did not want their child toparticipate (less than 1% did so). Neither teachers nor school administrators were in the room during the time the adolescentscompleted the questionnaires. Questionnaires and postage-paid return envelopes were sent to the adolescent’s parents.Neither adolescents nor parents were paid for their participation. The study, including all measures and procedures, wasapproved by the University’s Ethics Review Board.

Measures

Clustering variables

To explore subgroups of adolescents with psychopathic traits, we used adolescents’ self-reports of psychopathic traits anda parent reported measure of the adolescent’s anxiety as clustering variables.

Psychopathic traitsAdolescent psychopathic traits were measured with the Youth Psychopathic Traits Inventory (YPI: Andershed, Kerr,

Stattin, & Levander, 2002), a self-report instrument designed to capture psychopathic traits in community samples ofyouths 12 years and older. Reliability and construct validity of this instrument has been reported elsewhere (e.g., Andershedet al., 2002; Declercq, Markey, Vandist, & Verhaeghe, 2009; Poythress, Dembo, Wareham, & Greenbaum, 2006; Skeem &Cauffman, 2003). Participants responded to 50 items, which made up 10 internally consistent subscales. The four-pointresponse scale ranged from 1 (Does not apply at all) to 4 (Applies very well). Because psychopathy appears to be a multidi-mensional construct, the YPI items and its subscales can be summed to yield three factor scores that reflect three separate butinterrelated dimensionsdaffective, interpersonal and behavioraldconsistent with findings from previous studies (Cooke &Michie, 2001). In the current study, we analyzed both the total score and the factor scores. The Cronbach’s alphas were .90for the total YPI score, and .87, .90, and .86 for each factor score, respectively.

AnxietyThe measure of adolescent’s anxiety was created for the project. Parents were asked six questions about their youth’s

anxious behavior. Response options ranged from 1 (Does not apply at all) to 4 (Applies exactly). The items were: “Oftenworriesabout things without a reason,” “Often thinks about everything that can go wrong or bad before he/she does things,” and “Isoften more nervous than others in his/her age.” The alpha reliability for this scale was .82.

External criterion variables

To validate the hypothesized subgroups in our sample, we compared the groups on a set of theoretically relevant variablesthat were not used to generate the cluster solution.

DelinquencyAdolescent-reported delinquency was measured with 22 items similar to other commonly used, validated self-report

measures (e.g., Haynie, 2001) and has been validated in a Swedish sample in a comparison with official records (Magnusson,Dunér, & Zetterblom, 1975). Items referred to delinquent behaviors during the past year, ranging from acts of vandalism to

S. Salihovic et al. / Journal of Adolescence 37 (2014) 681–689684

shoplifting or breaking into a building. Items were rated on a 5-point scale ranging from 1 (No, it has not happened) to 5 (Morethan 10 times), and canbe summed toyield a total score for delinquency or subscale scores for violence, serious property offences,minor property offences, and substance use. In the present study, we analyzed both the total delinquency score and the subscalescores. The Cronbach’s alphas were .93 for total delinquency score, and .80, .87, .91, and .81 for each subscale, respectively.

Attention deficit/hyperactivity disorder symptoms (ADHD)To measure ADHD symptoms, we used the Swanson, Nolan and Pelham Rating Scale (SNAP-IV; Swanson et al., 2001)

which is a parent-reported measure with good reliability (Bussing et al., 2008). The measure was developed to assess ADHDsymptoms (i.e., hyperactivity/impulsivity and attention problems) according to DSM-IV, and can be used in community-basedpopulations. Parents reported howwell different statements applied to their adolescent’s behavior, and the response optionsranged from 1 (Does not apply at all) to 4 (Applies exactly). Hyperactivity/Impulsivitywas measured with eight statements, andexamples of these were: “Often leaves his/her place and has difficulties sitting still in situations that demand sitting still” and“Often has difficulties waiting for his/her turn.” Attention problems were measured with nine statements, and examples ofthese were: “Is often distracted by things that happen around him/her” and “Often have difficulties with maintainingattention during a longer period of time.” The scales showed good alpha reliabilities, .87 and .91 for hyperactivity/impulsivityand attention problems, respectively.

AggressionThe measure of adolescent-reported aggression was created for the project and closely modeled after existing scales (e.g.,

Raine et al., 2006). It comprised 12 items that reflect reactive and proactive types of aggression. The response format rangedfrom 1 (Does not apply at all) to 4 (Applies exactly). Reactive aggressionwasmeasuredwith six items, for example: “If I am reallyangry I usually throw, hit or kick things” and “If I ammad at someone, I don’t hesitate to hit him/her.” Proactive aggressionwasmeasured with six items, examples of items were: “If I am mad at someone I can say bad things to him/her, even though he/she hasn’t said anything bad tome” and “I pick a fight with people I don’t like, even though they haven’t done anything tome.”Both scales showed good alpha reliabilities, .82 and .86 for reactive and proactive aggression, respectively.

Strategy of analysis

To identify subgroups of adolescents with high levels of psychopathic traits, we used latent class analysis (LCA; Jung &Wickrama, 2008) in Mplus 6.0 (Muthén & Muthén, 1998–2010). LCA is a person-centered latent variable method thatidentifies subgroups of individuals within a heterogeneous sample. These subgroups are referred to as latent classes. Themodel defines the proportion of individuals belonging to each class and provides various indications of the goodness-of-fit foreach class enumeration. Hence, people who are classified as belonging to a single class can be considered to be homogenous.Because the analytical procedure is based on empirically defined rather than theoretically imposed classes, we estimatedlatent class models ranging from a two to a six-class solution for the combination of psychopathic traits and anxiety. Todetermine the adequate number of classes, we followed recommendations from previous studies and chose the model withthe lowest Bayesian information criterion (BIC; Kass & Raftery, 1995) the highest classification accuracy (Entropy; Muthén,2003), and a significant Lo Mendell Rubin Likelihood Ratio Test (LMR-LRT; Lo, Mendell, & Rubin, 2001) in which a signifi-cant p value suggests that the current number of classes represents the data better than a model with one less class (k � 1).

The data analysis proceeded in two steps. In the first step, we estimated LCA models and chose the model that bestrepresented the data. Then, in the second step, we used One-way Analysis of Variance (ANOVA) with post-hoc comparison tovalidate the latent classes by examining mean-level differences on external criterion variables.

Results

Correlations between the clustering variables (i.e., psychopathic traits and anxiety) and external criterion variables areshown in Table 1. Psychopathic traits were positively associated with delinquent behavior, reactive and proactive aggression,

Table 1Correlations between clustering variables and external validation variables.

Variables YPI Anxiety

Delinquency (total score) .33*** .03Serious property offenses .20*** .08Minor property offenses .37*** �.01Violence .21*** �.01Substance use .13*** .03Reactive aggression .45*** .02Proactive aggression .49*** �.01Hyperactivity/Impulsivity .23*** .45***Attention problems .29*** .46***

Note. YPI ¼ Psychopathic traits (total score).***p < .001.

S. Salihovic et al. / Journal of Adolescence 37 (2014) 681–689 685

and ADHD symptoms. Anxiety was uncorrelated with delinquent behavior and aggression, but positively related to ADHDsymptoms. It should be noted that the correlation between psychopathic traits and anxiety was not significant (r ¼ .01,p ¼ .86).

Identifying subgroups

Are there unique subgroups of youthswith high levels of psychopathic traits and different levels of anxiety? To answer thisquestion, we tested LCA models with two through six classes. As shown in Table 2, a five-class solution provided the bestoverall fit to the data in terms of the lowest BIC and AIC values, satisfactory entropy, and significant LMR-LRT test, whichsuggests that five classes represent the data better than four or six-classes. The posterior probability that the youths werecorrectly assigned to a class was high (ranged between .81 and .91), which further suggests that five groups adequatelyrepresent the data. According to the five-class solution, the groups were characterized by relatively low, average, or highlevels of psychopathic traits and low or high scores on anxiety. As expected, two relatively small classes (n ¼ 34) had highscores on psychopathic traits. One was characterized by high levels of psychopathic traits and anxiety (n ¼ 12), the other byhigh levels of psychopathic traits and low anxiety (n ¼ 22). For descriptive purposes, we labeled these groups as low-anxiousand high-anxious psychopathic traits, respectively, based on their relative level of psychopathic traits and anxiety. Theremaining three groups comprised the majority of the sample (n ¼ 880), and were labeled low, elevated psychopathic traits,and anxious subgroups because they were characterized by relatively lower or elevated scores on psychopathic traits and lowor high scores on anxiety. For descriptive purposes, we will refer to these groups as “non-psychopathic” because they werenot characterized by high levels of psychopathic traits. Thus, consistent with our hypothesis, there were two relatively smallgroups of adolescents who were high on psychopathic traits but differed in levels of anxiety.

Cluster validation

To address the theoretical meaningfulness and validity of the subgroups, we compared the groups on a set of externalcriterion variables. We hypothesized that the high-anxious psychopathic traits group would score higher on ADHD-symptoms, reactive aggression, and delinquent behavior than the low-anxious group. The low-anxious group, on the otherhand, would score higher on proactive aggression. For measures where the homogeneity of variance assumptionwas not met(i.e., psychopathic traits, delinquent behavior, and aggression), we used Welch’s F test and the Games-Howell post hocprocedure for follow-up comparisons, and for all other measures (i.e., inattention and hyper-impulsivity) we used Fisher’s Fand Tukey’s HSD post hoc comparison.

Table 3 shows the standardized means for the five subgroups across the variables used in the class derivation (i.e., psy-chopathic traits and anxiety) and the external criterion variables (i.e., delinquent behavior, aggression, ADHD-symptoms). Fordescriptive purposes, we also examined differences in levels of core psychopathic traits (i.e., callousness, grandiose–manipulative traits, and impulsive–irresponsible behavior) because they provide additional information about the profiles ofthe subgroups. As shown in the upper part of Table 3, the low-anxious psychopathic traits group had the highest scores on allmeasures of psychopathic traits, and did not differ from the non-psychopathic groups on anxiety. The high-anxious psy-chopathic traits group, in contrast, had somewhat lower scores on psychopathic traits than the low-anxious psychopathictraits group, and higher scores on anxiety than all other groups. Regarding external criterion variables, we found significantmain effects for delinquent behavior, aggression, and ADHD-symptoms. As shown in the lower part of Table 3, there wereclear differences between the groups characterized by high levels of psychopathic traits and the three non-psychopathicgroups on delinquency, with the non-psychopathic groups scoring significantly lower on all aspects of delinquentbehavior. The high-psychopathic trait groups also differed from one another: the low-anxious psychopathic traits subgroupshowed higher levels of minor offences, whereas the high-anxious psychopathic traits group scored significantly higher onserious offences. With respect to aggression, the high-anxious psychopathic traits group reported the highest levels of bothreactive and proactive aggression. Adolescents with high levels of psychopathic traits and low anxiety, in contrast, were notsignificantly different from the average group on either aspects of aggression. Finally, regarding ADHD-symptoms, the high-anxious psychopathic traits group scored significantly higher on hyperactivity/impulsivity and attention problems than anyother group. As expected, adolescents with high levels of psychopathic traits and low anxiety were not characterized byADHD-symptomsdthey did not differ significantly from the non-psychopathic groups.

Table 2Class solutions of the LCA models.

Class AIC BIC Entropy LMR-LRT

2 Classes 2320.681 2354.405 .69 .0023 Classes 2301.962 2350.140 .57 .4514 Classes 2281.680 2344.311 .59 .1295 Classes 2246.189 2323.274 .68 .0016 Classes 2246.810 2338.349 .57 .238

Note. N ¼ 914. AIC ¼ Akaike Information Criterion; BIC ¼ Bayesian Information Criterion; LMR-LRT ¼ Lo Mendell Rubin Likelihood Ration Test.

Table 3Standardized mean level differences between the subgroups.

Class 1High-anxiouspsychopathic traits(n ¼ 12)

Class 2Low-anxious psychopathictraits (n ¼ 22)

Class 3Anxious(n ¼ 101)

Class 4Elevated psychopathictraits (n ¼ 386)

Class 5Low(n ¼ 393)

F

M M M M M

Clustering variablesYPITotal score .84a 1.31b �.31c .38d �.40c 630.98***Callous, unemotional traits .84a 1.37b �.39c .42d �.42c 298.26***Grandiose, manipulative traits .79a 1.56b �.33c .48d �.46c 496.36***Impulsive, irresponsible behavior .76a 1.16b �.28c .28d �.38c 203.53***Anxiety 1.97a �.17b 1.01c �.18b �.40b 263.28***

External criterion variablesDelinquent behaviorTotal score .71a .64a �.13b .18b �.12b 14.97***Serious property offences .87a .36ab �.07b .12b �.11b 6.52***Minor property offences .82a .92a �.14b .30b �.12b 18.70***Substance use .26a .23a �.13b .09ab �.09b 4.95*Violent behavior .50a .51a �.13b .09ab �.10b 6.10***AggressionProactive aggression .72a .16bc �.17bc .30ab �.23c 30.87***Reactive aggression .89a .18bc �.15c .30b �.18c 36.99***ADHD-symptomsHyperactivity/Impulsivity 1.28a .45b .15c .03c �.23c 21.49***Attention problems 1.56a .34b .30b .12c �.29c 3.45***

Note. YPI ¼ Psychopathic traits. Across rows, means with different subscripts differ significantly whereas means sharing one of the subscripts do not differ.*p < .05; ***p < .001.

S. Salihovic et al. / Journal of Adolescence 37 (2014) 681–689686

In sum, our findings show that adolescents with high levels of psychopathic traits can be classified into a high and a low-anxious subgroup. Consistent with our expectations and previous empirical findings, the high-anxious group showed moreattention problems, impulsivity, and aggression, and reported more serious offences than the low-anxious subgroup.

Discussion

Early theoretical portrayals of psychopathy differentiate between two types of psychopaths: a primary type whosebehavior is believed to originate from a congenital affective deficit and a secondary type whose behavior is thought to be theresult of disadvantageous rearing in early childhood (e.g., Karpman, 1941). It is also suggested that the main distinctionbetween the types is in the level of anxiety: the primary, low-anxious type is defined by a lack of anxiety, guilt and fear, andthe secondary, high-anxious type is described as experiencing more neuroticism, impulsivity, anger, and distress (Cleckley,1976; Karpman, 1941; Lykken, 1995; Skeem et al., 2003). We used a normative sample of adolescents to test whether anxi-ety distinguishes between subgroups of adolescents with high levels of psychopathic traits. We identified five groups ofadolescents, two of which had high scores on psychopathic traits and low or high scores on anxiety. Although these groupswere relatively small, they differed from each other in several theoretically meaningful waysdhigh-anxious adolescents withpsychopathic traits were more impulsive, aggressive, and engaged in serious offending to a higher extent than low-anxiousyouths with psychopathic traits. Thus, there appears to be meaningful variation among youths with high levels of psycho-pathic traits.

The link between adolescent psychopathic traits and anxiety has been considered previously, and there is some empiricalsupport that primary and secondary variants can be identified among juvenile offenders (e.g., Kimonis et al., 2011, 2012; Leeet al., 2010; Vaughn et al., 2009; Veen et al., 2011). We extend this line of research by showing that similar subgroups can beidentified and validated within a normative community sample. Consistent with our hypothesis, the groups showed differ-ences in ADHD-symptoms, anger, and violence, which is further in line with the results from recent studies (e.g., Kimoniset al., 2011, 2012). Our prediction that the low-anxious psychopathic traits group would show more pronounced proactiveaggressionwas not supported, however. Instead, it was the high-anxious psychopathic traits group that had the highest levelsof both reactive and proactive aggression than all other groups. Although inconsistent with our initial hypotheses, previousstudies have found that the high-anxious psychopathic traits group typically displays more hostility and violence than thelow-anxious psychopathic traits group (e.g., Vaughn et al., 2009). A similar pattern was also observed in the presentstudydalthough the groups generally did not differ in the extent of problem behaviors, the high-anxious psychopathic traitsgroup expressed more serious offending, which could explain the higher levels of both proactive and reactive aggression incomparison with the low-anxious psychopathic traits group. Taken together, our results contribute to the growing body ofliterature on low-anxious and high-anxious subgroups of youths with high levels of psychopathic traits.

Although it may seem that the natural focus of research on psychopathic traits should be on youths within forensicsettings, there are advantages to studying youths before their problems have escalated to warrant institutional care. With

S. Salihovic et al. / Journal of Adolescence 37 (2014) 681–689 687

respect to theory development, our results are in line with the notion that there are two groups of youths with high levels ofpsychopathic traits who, despite similarities in overt behavior, show differences in temperament. Thus, observing thesegroups in a normal sample provides further support for the theoretical idea. Furthermore, the community approach isrelevant given the empirical evidence suggesting that psychopathy is a dimensional configuration of traits rather than adiscrete category (Edens, Marcus, Lilienfeld, & Poythress, 2006; Marcus, John, & Edens, 2004; Marcus, Lilienfeld, Edens, &Poythress, 2006). Studying the expression of psychopathic traits only among youths who are at the extreme end of thecontinuum does not allow the study of development, and precludes identification of other potentially interesting groups. Inthe current study, three groups emerged that were not characterized by high levels of psychopathic traits. One group, inparticular, was not anticipated because it comprised adolescents who scored somewhat above average on psychopathic traitsbut displayed above average scores on delinquent behavior and aggression. The problematic behavior of these youths clearlywarrants further investigation, even though their level of psychopathic traits was moderate. Thus, the primary strength of ourfindings lies in identifying subgroups within a normative sample.

Although anxiety differentiated between two theoretically meaningful subgroups, the link between psychopathic traitsand anxiety is far from clear. Theoretically, anxiety has been conceptualized to be negatively associated with psychopathicpersonality. Cleckley (1976), for example, placed particular emphasis on absence of anxiety as one of the defining features ofpsychopathic individuals. Previous empirical work has generally found that adult psychopathy and anxiety are either un-associated or negatively correlated (Lilienfeld & Penna, 2001; Schmitt & Newman, 1999). However, findings on psychopathictraits in adolescence suggest that anxiety is positively associated with measures of adolescent psychopathic traits andoffending (Kubak & Salekin, 2009; Vitale et al., 2005). These contrasting findings have been interpreted to suggest devel-opmental differences in the expression of psychopathic traits (Kubak & Salekin, 2009; for a review, see Salekin & Frick, 2005).The extent of these differences, however, is still relatively unknown and it will be important for future research to untanglethe relationship between anxiety and psychopathic traits.

We believe that our findings may have implications for research and practice. First, our results show that there seems to bemeaningful variation in the expression of psychopathic traits in adolescence. Future consideration of the heterogeneity in thisgroup will likely result in greater reliability and validity of results (Hicks et al., 2004), as well as more tailored interventionsand treatments. Second, although both groups with high levels of psychopathic traits were similar with respect to their levelof problem behavior, youths with high levels on both psychopathic traits and anxiety showed a more problematic profile. Interms of treatment, it seems that the aggressive and impulsive behaviors of this group need particular consideration. Somepreliminary, albeit promising, findings suggest that youths who most closely resembled the secondary variant showed re-ductions in violence over two years (Kimonis et al., 2011). Still, more studies are needed to evaluate the practical relevance ofthese subgroups.

Some limitations with the design of the current study need mention. First, we recognize that the cross-sectional design ofthe study precludes us from inferring whether the observed subgroups represent distinct etiological mechanisms in devel-opment.We have only described the variation in psychopathic traits and shown that there appears to be groups that resemblethe descriptions of the primary and secondary subtypedwhether or not these groups represent discrete subtypes is aquestion for future studies to answer. Second, as the youths in our sample were recruited from schools during regular schoolhours, wemay have failed to assess youthswith extreme levels of psychopathic traits because theymay have not been presentfor the data collection. Nevertheless, because we used a validated measure of psychopathic traits designed for communitysamples (i.e., the YPI; Andershed et al., 2002), we can be relatively confident that the high-psychopathic groups compriseadolescents with the highest scores within our sample. Another limitation that deserved mention is the small group sizes ofyouths with high levels of psychopathic traits. However, we assessed youths from a community setting where the base rate ofhigh-scorers is generally low. Although the groups were distinguishable on external variables, our results need replication.Finally, adolescent’s anxiety was assessed with parent reports. As an internal state, anxiety is typically assessed with self-reports or other physiological measures. Although our measure showed good internal consistency and distinguished thesubgroups as hypothesized, we recommend that future studies replicate these findings with validated physiological measuresof anxiety or self-report anxiety scales.

Despite these limitations, this study has several strengths. One strength is the large representative sample of ado-lescents. Previous studies have examined these subgroups in institutionalized samples among youths scoring above acut-off (e.g., Kimonis et al., 2011, 2012; Veen et al., 2011). Given that there is limited support for a categorical distinctionbetween psychopaths and non-psychopaths, this method imposes differences between youths that may not be mean-ingful or that may result in classification errors. In this study, we used the whole variation to classify youths based onsimilarities and differences, providing further support for the validity of the distinction. Another strength is the use ofLCA which, as a model-based approach, estimates latent variables that are free from measurement error, and providesmore reliable cluster solutions than standard clustering techniques (Magidson & Vermunt, 2002). Given that our mainobjective was to identify theoretically meaningful subgroups without a priori restrictions on the number of classes, theuse of LCA was appropriate.

At the broadest level, our results suggest that a small proportion of adolescents characterized by high levels of psycho-pathic traits engage in antisocial behavior and aggression to a much higher extent than adolescents low on psychopathictraits. However, within this small group, there are some youths who seem more at-risk. In this study, these youths werecharacterized by high scores on both psychopathic traits and anxiety. These findings are in line with theory and empiricalliterature showing that people with psychopathic traits can be disaggregated into groups that are behaviorally alike but differ

S. Salihovic et al. / Journal of Adolescence 37 (2014) 681–689688

in temperamental features. The etiological underpinning of these groups remains unknown, but there seems to be mean-ingful variation among youths with high levels of psychopathic traits that future research needs to consider.

References

Andershed, H., Kerr, M., Stattin, H., & Levander, S. (Eds.). (2002). Psychopathic traits in non-referred youths: A new assessment tool. The Hague: Elsevier.Blackburn, R. (1998). Psychopathy and personality disorder: implications of interpersonal theory. In D. J. Cooke, A. E. Forth, & R. D. Hare (Eds.), Psychopathy:

Theory, research and implications for society (pp. 269–301). Dordrecht, The Netherlands: Kluwer.Bussing, R., Fernandez, M., Harwood, M., Hou, W., Garvan, C. W., Eyberg, S. M., et al. (2008). Parent and teacher SNAP-IV ratings of Attention Deficit Hy-

peractivity Disorder symptoms: psychometric properties and normative ratings from a school district sample. Assessment, 15(3), 317–328. http://dx.doi.org/10.1177/1073191107313888.

Cleckley, H. (1976). The mask of sanity (5th ed.). St. Louis, MO: Mosby.Cooke, D. J., & Michie, C. (2001). Refining the construct of psychopathy: towards a hierarchical model. Psychological Assessment, 13(2), 171–189. http://dx.doi.

org/10.1037/1040-3590.13.2.171.Declercq, F., Markey, S., Vandist, K., & Verhaeghe, P. (2009). The youth psychopathic trait inventory: factor structure and antisocial behaviour in non-referred

12-17-year-olds. Journal of Forensic Psychiatry & Psychology, 20(4), 577–594. http://dx.doi.org/10.1080/14789940802651757.Edens, J. F., Marcus, D. K., Lilienfeld, S. O., & Poythress, N. G. (2006). Psychopathic, not psychopath: taxometric evidence for the dimensional structure of

psychopathy. Journal of Abnormal Psychology, 115(1), 131. http://dx.doi.org/10.1037/0021-843X.115.1.131.Forth, A. E., & Book, A. S. (2010). Psychopathic traits in children and adolescents: the relationship with antisocial behaviors and aggression. In R. T. Salekin, &

D. R. Lynam (Eds.), Handbook of child and adolescent psychopathy (pp. 251–283). New York: The Guilford Press.Haynie, D. L. (2001). Delinquent peers revisited: does network structure matter? American Journal of Sociology, 106(4), 1013–1057. http://dx.doi.org/10.1086/

320298.Hicks, B. M., Markon, K. E., Patrick, C. J., Krueger, R. F., & Newman, J. P. (2004). Identifying psychopathy subtypes on the basis of personality structure.

Psychological Assessment, 16(3), 276–288. http://dx.doi.org/10.1037/1040-3590.16.3.276.Hicks, B. M., Vaidyanathan, U., & Patrick, C. J. (2010). Validating female psychopathy subtypes. Personality Disorders: Theory, Research, and Treatment, 1(1),

38–57. http://dx.doi.org/10.1037/a0018135.Jung, T., & Wickrama, K. A. S. (2008). An introduction to latent class growth analysis and growth mixture modeling. Social and Personality Psychology

Compass, 2(1), 302–317. http://dx.doi.org/10.1111/j.1751-9004.2007.00054.x.Karpman, B. (1941). On the need of separating psychopathy into two distinct clinical types: the symptomatic and the idiopathic. Journal of Criminal Psy-

chopathology, 3, 112–137.Kass, R. E., & Raftery, A. E. (1995). Bayes factors. Journal of the American Statistical Association, 90(430), 773–795.Kimonis, E. R., Frick, P. J., Cauffman, E., Goldweber, A., & Skeem, J. L. (2012). Primary and secondary variants of juvenile psychopathy differ in emotional

processing. Development and Psychopathology, 24(3), 1091. http://dx.doi.org/10.1017/S0954579412000557.Kimonis, E. R., Skeem, J. L., Cauffman, E., & Dmitrieva, J. (2011). Are secondary variants of juvenile psychopathy more reactively violent and less psycho-

socially mature than primary variants? Law and Human Behavior, 35(5), 381–391. http://dx.doi.org/10.1007/s10979-010-9243-3.Kubak, F. A., & Salekin, R. T. (2009). Psychopathy and anxiety in children and adolescents: new insights on developmental pathways to offending. Journal of

Psychopathology and Behavioral Assessment, 31(4), 271–284. http://dx.doi.org/10.1007/s10862-009-9144-2.Lee, Z., Salekin, R. T., & Iselin, A.-M. R. (2010). Psychopathic traits in youth: is there evidence for primary and secondary subtypes? Journal of Abnormal Child

Psychology, 38(3), 381–393. http://dx.doi.org/10.1007/s10802-009-9372-7.Lilienfeld, S. O., & Penna, S. (2001). Anxiety sensitivity: relations to psychopathy, DSM-IV personality disorder features, and personality traits. Journal of

Anxiety Disorders, 15(5), 367–393. http://dx.doi.org/10.1016/S0887-6185(01)00070-6.Lo, Y., Mendell, N. R., & Rubin, D. B. (2001). Testing the number of components in a normal mixture. Biometrika, 88(3), 767–778. http://dx.doi.org/10.1093/

biomet/88.3.767.Lykken, D. T. (1995). The antisocial personalities. Hillsdale: NJ: Erlbaum.Magidson, J., & Vermunt, J. (2002). Latent class models for clustering: a comparison with K-means. Canadian Journal of Marketing Research, 20(1), 36–43.Magnusson, D., Dunér, A., & Zetterblom, G. (1975). Adjustment: A longitudinal study. New York: Wiley.Marcus, D. K., John, S. L., & Edens, J. F. (2004). A taxometric analysis of psychopathic personality. Journal of Abnormal Psychology, 113, 626–635. http://dx.doi.

org/10.1037/0021-843X.113.4.626.Marcus, D. K., Lilienfeld, S. O., Edens, J. F., & Poythress, N. G. (2006). Is antisocial personality disorder continuous or categorical? A taxometric analysis.

Psychological Medicine, 36(11), 1571–1582. http://dx.doi.org/10.1017/S003329170600824.Mealey, L. (1995). Primary sociopathy (psychopathy) is a type, secondary is not. Behavioral and Brain Sciences, 18(03), 579–599. http://dx.doi.org/10.1017/

S0140525X00040024.Muthén, B. (2003). Statistical and substantive checking in growth mixture modeling: comment on Bauer and Curran (2003). Psychological Methods, 8(3),

369–377. http://dx.doi.org/10.1037/1082-989X.8.3.369.Muthén, L. K., & Muthén, B. (1998–2010). Mplus user’s guide (6th ed.). Los Angeles, CA: Muthén & Muthén.Newman, J. P., MacCoon, D. G., Vaughn, L. J., & Sadeh, N. (2005). Validating a distinction between primary and secondary psychopathy with measures of

Gray’s BIS and BAS constructs. Journal of Abnormal Psychology, 114(2), 319–323. http://dx.doi.org/10.1037/0021-843X.114.2.319.Porter, S. (1996). Without conscience or without active conscience? The etiology of psychopathy revisited. Aggression and Violent Behavior, 1(2), 179–189.

http://dx.doi.org/10.1016/1359-1789(95)00010-0.Poythress, N. G., Dembo, R., Wareham, J., & Greenbaum, P. E. (2006). Construct validity of the youth psychopathic traits inventory (YPI) and the antisocial

process screening device (APSD) with justice-involved adolescents. Criminal Justice and Behavior, 33(1), 26–55. http://dx.doi.org/10.1177/0093854805282518.

Poythress, N. G., & Skeem, J. (2006). Disaggregating psychopathy: where and how to look for variants. In C. Patrick (Ed.), Handbook of psychopathy (pp. 172–192). New York: Guilford Press.

Raine, A., Dodge, K., Loeber, R., Gatzke-Kopp, Lynam, D., Reynolds, C., et al. (2006). The reactive-proactive aggression questionnaire: differential correlates ofreactive and proactive aggression in adolescent boys. Aggressive Behavior, 32(2), 159–171. http://dx.doi.org/10.1002/ab.20115.

Salekin, R. T., & Frick, P. J. (2005). Psychopathy in children and adolescents: the need for a developmental perspective. Journal of Abnormal Child Psychology,33(4), 403–409. http://dx.doi.org/10.1007/s10802-005-5722-2.

Salekin, R. T., Rosenbaum, J., Lee, Z., & Lester, W. S. (2009). Child and adolescent psychopathy: like a painting by Monet. Youth Violence and Juvenile Justice,7(3), 239–255. http://dx.doi.org/10.1177/1541204009333832.

Schmitt, W. A., & Newman, J. P. (1999). Are all psychopathic individuals low-anxious? Journal of Abnormal Psychology, 108, 353–358.Skeem, J. L., & Cauffman, E. (2003). Views of the downward extension: comparing the youth version of the psychopathy checklist with the youth psy-

chopathic traits inventory. Behavioral Sciences & the Law, 21(6), 737–770. http://dx.doi.org/10.1002/bsl.563.Skeem, J. L., Johansson, P., Andershed, H., Kerr, M., & Eno Louden, J. (2007). Two subtypes of psychopathic violent offenders that parallel primary and

secondary variants. Journal of Abnormal Psychology, 116(2), 395. http://dx.doi.org/10.1037/0021-843X.116.2.395.Skeem, J. L., Poythress, N., Edens, J. F., Lilienfeld, S. O., & Cale, E. M. (2003). Psychopathic personality or personalities? Exploring potential variants of

psychopathy and their implications for risk assessment. Aggression and Violent Behavior, 8(5), 513–546. http://dx.doi.org/10.1016/S1359-1789(02)00098-8.

Statistiska Centralbyrån. (2001).

S. Salihovic et al. / Journal of Adolescence 37 (2014) 681–689 689

Swanson, J. M., Kraemer, H. C., Hinshaw, S. P., Arnold, L. E., Conners, C. K., Abikoff, H. B., et al. (2001). Clinical relevance of the primary findings of the MTA:success rates based on severity of ADHD and ODD symptoms at the end of treatment. Journal of the American Academy of Child & Adolescent Psychiatry,40(2), 168–179. http://dx.doi.org/10.1097/00004583-200102000-00011.

Swogger, M. T., & Kosson, D. S. (2007). Identifying subtypes of criminal psychopaths: a replication and extension. Criminal Justice and Behavior, 34(8), 953–970. http://dx.doi.org/10.1177/0093854807300758.

Vaughn, M. G., Edens, J. F., Howard, M. O., & Smith, S. T. (2009). An investigation of primary and secondary psychopathy in a statewide sample of incar-cerated youth. Youth Violence and Juvenile Justice, 7(3), 172–188. http://dx.doi.org/10.1177/1541204009333792.

Veen, V. C., Andershed, H., Stevens, G. W. J. M., Doreleijers, T. A. H., & Vollebergh, W. A. M. (2011). Psychopathic subtypes and associations with mentalhealth problems in an incarcerated sample of adolescent boys. International Journal of Forensic Mental Health, 10(4), 295–304. http://dx.doi.org/10.1080/14999013.2011.62971.

Vitale, J. E., Newman, J. P., Bates, J. E., Goodnight, J., Dodge, K. A., & Pettit, G. S. (2005). Deficient behavioral inhibition and anomalous selective attention in acommunity sample of adolescents with psychopathic traits and low-anxiety traits. Journal of Abnormal Child Psychology, 33(4), 461–470. http://dx.doi.org/10.1007/s10802-005-5727-X.

Wareham, J., Dembo, R., Poythress, N. G., Childs, K., & Schmeidler, J. (2009). A latent class factor approach to identifying subtypes of juvenile diversionyouths based on psychopathic features. Behavioral Sciences & the Law, 27(1), 71–95. http://dx.doi.org/10.1002/bsl.844.