Psychopathy refers to a dysfunctional personality condition characterized by interpersonal, emotional, and behavioral deviancy features that reflect the presence of arrogance, cold-heartedness, and daring behavior (e.g., Hare 2003; Patrick et al. 2009). Research has consistently highlighted the association between psychopathy and a vast array of externalizing behavior such as violence and other forms of criminality (Douglas et al. 2006; Hare 2003), criminal recidivism (e.g., Leistico et al. 2008; Walters et al. 2008), substance use (e.g., Gustavson et al. 2007; Kennealy et al. 2007), and sexual offending (Caldwell et al. 2008).

During the last two decades, the study of psychopathic traits in children and adolescents has gradually become a major research topic (Salekin & Lynam 2010). This is because investigating psychopathy in these age groups may help clinicians and researchers to gain insights into the different pathways toward severe antisocial behavior, and to understand the etiology of this severe adult personality disorder, as well as to offer preventive interventions or early treatment programs (van Baardewijk et al. 2011). Indeed, Salekin et al. (2010) convincingly contended the notion that psychopathy is untreatable; rather, Salekin et al.' (2010) findings indicated that the earlier the psychopathic subjects are treated, the better the treatment gains.

To date, studies of psychopathic traits in youth have yielded results that are strikingly similar to those in adults in terms of latent structure, stability, and relations to cognitive and emotional functioning as well as conduct disorder and aggressive behavior (see, for a review, Andershed 2010; Lynam 2011). Research on the latent structure of psychopathy has shown that it is organized along a continuum (more or less psychopathy) rather than in discrete categories (psychopath vs. non-psychopath) in both adults and adolescents (e.g. Marcus et al. 2004; Murrie et al. 2007; Walters 2014) and across settings (community, clinical, and forensic). Recently, to better understand the developmental processes that lead to this deviant personality syndrome, studies on large-scale non-referred community samples have emerged (e.g., Andershed et al. (2002).

Measurement of Psychopathy

The Psychopathy Checklist: Youth Version (PCL:YV; Forth et al. 2003) is often considered to be the most reliable and valid measure of psychopathic traits among forensic youth (Andershed et al. 2007; Hare 2003). However, for research on non-incarcerated adolescents and community samples, self-report instruments are often deemed useful because they are easy to use and fast to administer (Andershed et al. 2007). Moreover, self-report instruments do not 1) rely on file-information for their use (Colins et al. 2012), 2) require parent or teacher ratings (when teachers or parents are not thought to be accurate, or easy to access), or 3) require extensive training on the part of the test administrator (Lilienfeld and Fowler 2006; Loney et al. 2003). Finally, self-report measures provide an important perspective on self rated traits, because adolescents, unlike parents or teachers, are in the unique position to report on their traits and behavior across a range of situations, including the home, the classroom, and during free time with peers (van Baardewijk et al. 2010).

A number of self-report instruments that have been designed to assess psychopathic features in children and adolescents are currently available (see, for a review, Kotler and McMahon 2010). However, the psychometric properties for the measures vary. Among the self-report questionnaires currently available for assessing psychopathic traits in adolescence, the Youth Psychopathic Traits Inventory has been considered particularly favorable by several reviewers (YPI; Andershed et al. 2002; Kotler and McMahon 2005; Vaughn and Howard 2005). Research has shown its psychometric properties to be particularly good in terms of factor structure and reliability (Colins et al. 2014b; Oshukova et al. 2015; Poythress et al. 2006).

Regardless of the measure, however, whether it be the YPI or another self report measure, such as the APSD, empirical evidence suggests that psychopathy may represent a hierarchical construct whose basic features (i.e., directly observable behaviors and attitudes) are clustered by first-order (or even second-order) correlated latent dimensions that represent facets of an over-arching general psychopathy factor. Although alternative models with differing numbers of first-order factors have been proposed, usually two broad factors (e.g., Hare 1991; Harpur et al. 1988; Levenson et al. 1995), or narrower three-factor (Patrick et al. 2009) and four-factor models (Hare and Neumann 2006) of psychopathy measures have been reported in empirical studies.

Recently, prominent authors suggested that greater emphasis should be placed on individual psychopathy-related traits rather than on psychopathy level in predicting early delinquent behavior. On the one hand, Frick and colleagues (see Frick and White 2008 for a review) found that the presence of callous-unemotional (CU) traits represents a risk for conduct problems at all ages, including childhood and adolescence. On the other hand, Salekin (2016a, b, 2017) convincingly proposed that CU traits is only one of three dimensions of child psychopathy. Thus, one can argue that grandiose-manipulative (GM) and daring-impulsive (DI; or impulsive-irresponsible; IMP) traits also should be considered and potentially incorporated as specifiers for conduct disorder (CD) in future versions of the DSM and ICD. Interestingly, Salekin (2016a, b, 2017) suggested that future research should test the proposed specifiers interaction with CD, because it would make a larger contribution to our understanding of youth with CD allowing clinicians to better describe and treat individuals with conduct problems.

The various dimensions of psychopathy may indeed interact with each other in the relation to, and in the prediction of, maladaptive behaviors related to psychopathy (e.g., conduct problems, self-reports of delinquent behavior). In fact, psychopathy is often argued to be constituted by the combination of high levels of all dimensions of psychopathy rather than high levels of a single dimension (e.g., Andershed 2010). Thus, the combination of high levels on all dimensions of psychopathy should be expected to be highly associated with problem behavior, more so than the individual factors. However, the available evidence about this interaction effect between the three psychopathic traits is scarce (Orue and Andershed 2015; Salekin 2016b). To our knowledge, only a few studies (Colins et al. 2014a Orue and Andershed 2015) have been specifically designed to evaluate the presence of significant interaction effects among the psychopathy dimensions of GM, CU, and II traits in youth. Colins et al. (2014a) examined three-way interaction in a Swedish general population sample of 2,056 3-to5-year-old children. Using an observer-rated analogue of the YPI (i.e., the Child Problematic Traits Inventory; CPTI), Colins et al. (2014a) found evidence that the three-way interaction among GM, CU, and II dimensions was a strong predictor of concurrent conduct problems, suggesting the usefulness of the traits in terms of nomological network validity. Moreover, Orue and Andershed (2015) examined interaction effects in a sample of high school students, and showed that the interaction between the three factors of psychopathy was the best predictor of proactive aggression.

Starting from these considerations, the present study aimed at evaluating if the interaction among GM, CU, and II dimensions may add significant information in predicting self-reported delinquency in a sample of 558 community dwelling Italian adolescents who were administered the YPI. Our adolescent participants were asked to report deviant behavior that occurred during the last year using the Self-Report of Delinquency Scale (SRDS; Elliott and Ageton 1980). The SRDS assesses the participants’ self-report of 40 illegal acts developed from a list of all offenses reported in the Uniform Crime Report with a juvenile base rate greater than 1% (Elliott and Huizinga 1984). Based on previous findings (Colins et al. 2014a; Orue and Andershed 2015), we expected that the interaction among GM, CU, and II dimensions would add a significant amount of information in predicting self-reported juvenile delinquency, over and above the individual factors.

Method

Participants

A total of 623 adolescents attending four high schools in Sardinia Island, Italy, agreed to participate in the study; 146 adolescents (23.4%) were male, and 472 adolescents (75.8%) were female, while 5 adolescents (0.8%) did not disclose their gender. Participants’ mean age was 16.30 years, SD = 1.70 years. All participants signed a written informed consent after detailed presentation of the study and after the study was approved by each high school Educational Board. In the case of participants of minor age, parents (or legal tutors) signed the informed consent form in order to let their child to take part in the study.

Sixty-five (10.4%) participants yielded incomplete data (i.e., they did not answer more than 10% of YPI and/or Self-Report of Delinquency Scale (Elliott and Ageton 1980) items) and were excluded from the final sample. Among participants who yielded incomplete YPI data, 21 adolescents (32.3%) were male, 40 adolescents (61.5%) were female, and four adolescent (6.2%) refused to disclose their gender; participants’ mean age was 14.80 years, SD = 0.80 years. Little MCAR test showed that missing values were completely at random, χ2(26) = 1.51, p > .90. Participants who yielded incomplete data did not significantly differ from participants who yielded complete responses on gender, Yates-corrected χ2(1) = 3.74, p > .05, φ = .08; rather, participants who yielded complete data were significantly older than participants who reported incomplete data, separate-variance t(145.97) = 13.64, p < .001, Vargha and Delaney’s (2000) A = .78. The proportion of participants who refused to disclose their gender significantly differentiated adolescents who reported complete data from adolescents who yielded incomplete data, Fisher exact test 2-tailed p < .001, φ = .20.

Thus, the final sample included 558 participants; 125 adolescents (22.4%) were male, and 432 adolescents (77.4%) were female, while 1 adolescent (0.2%) refused to disclose his/her gender. Participants’ mean age was 16.47 years, SD = 1.69 years.

Procedure

After obtaining Institutional Review Board approval from the university and the principals of the schools, researchers recruited from classrooms. Parental written consent and/or participant written assent was obtained after the study had been explained to participants. Participants were all volunteers who received no incentive for taking part in the research. Participants were selected from public high schools where data from the National Institute of Statistics of Italy showed the adolescents to be representative of the Italian adolescent population (ISTAT 2015). Graduate research assistants administered questionnaires during class time. The graduate research assistants were provided a time to conduct the study when teachers were not present in the classrooms to avoid potential concerns regarding candid reporting. The students were informed that their participation would be anonymous and questionnaires were administered in random order to reduce or eliminate any order effects. Internal consistency reliability coefficients for YPI and SRDS scores are exhibited in Table 1.

Table 1 Youth Psychopathic traits Inventory and Self Reported Delinquency Scale: Descriptive statistics in the full sample (N = 558) and broken down by gender (male adolescent high school students, n = 125; female adolescent high school students, n = 432)
Full size table

Measures

Youth Psychopathic Traits Inventory (YPI; Andershed et al. 2002). All participants were administered the Italian translation of the YPI (Fossati et al. 2016). The YPI is a 50-item self-report questionnaire designed to measure the core traits of psychopathic personality in adolescents (Andershed et al. 2002). The YPI measures psychopathic traits in terms of ten different scales, each containing five items (Andershed et al. 2002). In line with Cooke and Michie’s (2001) conceptualization of psychopathy, these scales manifest in a three factor structure consisting of: (1) a Grandiose-Manipulative dimension, (2) a Callous-Unemotional dimension, and (3) an Impulsive-Irresponsible dimension. The YPI total score is thought to assess the general level of psychopathic traits in adolescence. Each item of the YPI is scored on a four-point Likert scale ranging from Does not apply at all to Applies very well. The internal consistency reliability and scale-level factor validity of the Italian translation of the YPI in adolescent high school students were previously assessed (Fossati et al. 2016). In the present study, the YPI total score ranged from 52 to 167.

Self-Report of Delinquency Scale (SRDS; Elliott and Ageton 1980). The SRDS assesses the participants’ self-report of 40 illegal acts. The illegal acts are drawn from a list of all offenses reported in the Uniform Crime Report where the juvenile base rate greater than 1% (Elliott and Huizinga 1984). Each SRDS item is measured on a 6-point Likert scale from “Never” to “20 Times or More”. Consistent with previous use of this scale (Krueger et al. 1994), a composite measure was created by summing and then averaging the scores of all SRDS items. The psychometric properties of the Italian translation of the SRDS have been recently published (Fossati et al. 2016; Somma et al. 2014). Somma et al. (2016) formally assessed the factor validity of the Italian translation of the SRDS; exploratory structural equation modeling (ESEM) analysis supported the a priori bifactor model of the SRDS, in which each item loaded on a specific subscale. However, all items measured a general factor of self-reported delinquency. The fit statistics for the model were generally good: χ2 (626) = 787.38, p < .001, TLI = .98, CFI = .98, RMSEA = .02 (90% CI: 0.02, 0.03). In the present study, the SRDS total score ranged from 40 to 160.

Measure Translation Procedures

The YPI was independently translated into Italian by one of the authors (A.F.). Accuracy and equivaence were the primary aims and guiding principle to maintain the scale’s original meaning (Denissen et al. 2008). Two additional clinical psychologists who were fluent in English read the translation and made comments. After reaching a consensus on all items, a professional translator (English, first language), translated the Italian version back into English, and this English back-translation (Cha et al. 2007; Van de Vijver and Hambleton 1996; Geisinger 1994) was then sent to the author of the YPI for comments. The translators came to an agreement on the definitive Italian translation. The authors followed the same procedure of translation concerning the SRDS.

Data Analyses

Comparisons between two independent means were tested using Student t-test, while Cohen d was used to assess the magnitude of the effect (effect size) of the mean difference. In the case of non-homogeneous variances, a separate-variance t-test was used. Vargha and Delaney’s (2000) A was used as an effect size measure for separate-variance t-test. Vargha and Delaney (2000) provide suggested thresholds for interpreting the effect size, .5 means no difference at all; up to 0.56 indicates a small difference; up to 0.64 indicates medium; values over 0.71 are considered large (the same intervals apply below 0.5). Pearson r coefficients were used to examine the association between continuous variables. In the case of multiple comparisons, the nominal significance level (i.e., p < .05) was corrected according to the Bonferroni procedure.

We relied on hierarchical multiple regression models to test for the significance of adding interaction terms among YPI GM, CU, and II dimensions to the simple linear combination of YPI second-order dimensions in explaining SRDS total scores. Because in the present study participant’s gender was significantly associated with both SRDS total score and YPI dimension scores, participant gender was entered in Step 1 as a covariate. All predictors were mean centered before testing interaction effects. The significance of main effects of the YPI GM, CU, and II dimensions, controlling for the effect of gender was assessed in Step 2. Although we were interested in testing the significance of the GM-by-CU-by-II three-way interaction in predicting the SRDS total score (Step 4) of the hierarchical regression model, we relied on a full-factorial approach in which all two-way interactions among YPI GM, CU, and II dimensions were entered in Step 3 of the hierarchical model.

In order to identify the best fitting regression model in the hierarchy, we relied on the following indices: Akaike information criterion (Akaike 1973), Amemiya (1980) prediction criterion, Mallow’s (1973) C p , and Schwartz information criterion (Schwartz 1978). We considered as best fitting model the step in hierarchical regression analyses that corresponded to the minimum value that was reached by the majority of these criteria. Significance of change in R2 value was then used as an additional index to identify the best fitting model. The amount of variance explained in the SRDS total score by the independent variables was evaluated by computing R2 and adjusted R2 indices. Standardized regression coefficients were used to evaluate the effect of each YPI predictor. Variance inflation factor (VIF) was used to evaluate multi-collinearity. Usually, VIF values greater than 5 or even 10 are considered to suggest problems with multi-collinearity (Kutner et al. 2004).

Finally, we performed hierarchical regression analyses in order to evaluate if adding the GM-by-CU-by-II three-way interaction among the YPI dimensions significantly improved the regression model in predicting the SRDS total score over a model that included only the YPI total score as independent variable, even after controlling for participant’s gender. In these hierarchical regression analyses, we relied on the same model selection criteria, multi-collinearity indices, and effect size estimates described above.

Results

Descriptive statistics and gender comparisons for YPI dimensions and SRDS scores are listed in Table 1. The sum of male adolescents and female adolescents does not equal the total number of participants because one participant did not disclose his/her gender. No significant association was observed between YPI scores and participant’s age; Pearson r values for the association between participant’s age and YPI GM, CU, II, and total scores were .02, −.02, −.05, and −.01, all ps > .25, respectively. Similarly, no significant correlation was observed between participant’s age and SRDS total score, r = .05, p > .25.

In our study, YPI dimensions were significantly inter-correlated; YPI GM dimension correlated with YPI CU and II dimensions (r .49 and .50, all ps < .001, respectively), whereas YPI CU dimension correlated with YPI II dimension (r .36, p < .001). In this investigation, the YPI GM, CU, and II dimension correlated with the SRDS (r values of .41, .35, .44, all ps < .001, respectively. Finally, the association between YPI total score and SRDS total score was .50, p < .001.

Considering two-way interactions (i.e., product terms) between YPI dimensions, the SRDS total score correlated .45, .50, and .50 (all ps < .001) with GM-by-CU, GM-by-II, and CU-by-II product terms. Finally, GM-by-CU-by-II three-way interaction (i.e., product term) among YPI dimensions showed a significant bivariate correlation with SRDS total score, r = .52, p < .001. None of the r coefficient values were significantly different in male adolescents and female adolescents, as it was indicated by z values for the difference between independent r coefficients ranging from 0.44 (association between SRDS total score and YPI CU dimension) to 1.79 (association between SRDS total score and GM-by-CU-by-II three-way interaction), all ps > .05.

Results from the hierarchical regression analysis of YPI scores as predictors of SRDS total score in our adolescent sample, are summarized in Table 2. All predictors were mean centered before testing interaction effects in order to reduce multi-collinearity. Participant’s gender was coded as a dummy variable, male gender = 0, female gender = 1. As seen in Table 2, Step 1 in the hierarchical regression analysis included only the covariate effect, that is gender as independent variables. Akaike information criterion, Amemiya prediction criterion, Mallow’s C p , and Schwartz information criterion were −1290.19, .92, 208.82, and −1281.54, respectively. Step 2 showed that the II, GM and CU factor were significantly related to self-reported delinquency. Akaike information criterion value was −1428.29, Amemiya prediction criterion value was .72, Mallow’s C p value was 41.17, and Schwartz information criterion value was −1406.68. Two-way interaction effects (i.e., Step 3) showed that GM-by-II interaction effect significantly predicted the SRDS total score. The corresponding Akaike information criterion, Amemiya prediction criterion value, Mallow’s C p , and Schwartz information criterion values were −1451.37, .69, 17.24 and −1416.79, respectively. Finally, Step 4 of the hierarchical regression analysis showed that the GM-by-CU-by-II interaction effect was the better predictor of self- reported delinquency as indicated by the higher standardized beta; this step was the best fitting model according to Akaike information criterion = −1459.69, Amemiya prediction criterion = .68, Mallow’s C p  = 9.00, and Schwartz information criterion = −1420.78. Figure 1 shows the three-way interaction effect (GM-by-CU-by-II) in predicting self-reported delinquency (i.e., SRDS total score). As shown in Fig. 1, when participants obtained high scores on all the three dimension of psychopathy, the scores on the SRDS total score were higher. In contrast, when GM, CU and II dimension scores are low, levels of self-reported delinquency were much lower.

Table 2 Youth Psychopathic trait Inventory scores as predictors of Self-Reported Delinquency Scale total score: Hierarchical regression analysis summary table (N = 558)
Full size table
Fig. 1
figure 1

Grandiose-Manipulative (GM), Callous Unemotional (CU) and Impulsive-Irresponsible (II) dimensions of psychopathy in predicting Self-Reported Delinquency Scale (SRDS) Total Score (dependent variable): Three-way interaction effect among adolescent high school students (N = 558)

Full size image

Table 3 summarizes the hierarchical regression model results based on YPI total score and YPI GM-by-CU-by-II three-way interaction effect as predictors of SRDS total score. Here again, predictors were mean centered before testing interaction effects in order to reduce multi-collinearity; participant’s gender was coded as a dummy variable (male gender = 0, female gender = 1). In Step 1 of the hierarchical regression model, Akaike information criterion, Amemiya prediction criterion, Mallow’s C p , and Schwartz information criterion were −1290.19, .92, 189.16, and −1281.54, respectively. Step 2 shows that the YPI total score was significantly related to self-reported delinquency; Akaike information criterion value was −1422.15, Amemiya prediction criterion value was .73, Mallow’s C p value was 32.52, and Schwartz information criterion value was −1409.18. The best fitting model according to Akaike information criterion = −1450.06, Amemiya prediction criterion = .69, Mallow’s C p  = 4.00, and Schwartz information criterion = −1432.77 was Step 3, in which GM-by-CU-by-II significantly predicted the SRDS total score.

Table 3 Youth Psychopathic trait Inventory total score and Youth Psychopathic trait Inventory scale three-way interaction as predictors of Self-Reported Delinquency Scale total score: Hierarchical regression analysis summary table (N = 558)
Full size table

Discussion

As a whole, our findings confirm and extend previous findings documenting that the interaction among GM, CU, and II dimensions may add a significant information in predicting self-reported delinquency (Colins et al. 2014a, b; Orue and Andershed 2015). That is, considering the product of the three sets of traits improves the predictive performance of the construct over and above the information that is provided by a single underlying dimension of psychopathic traits (e.g., CU dimension). In our study, the three-way interaction term accounted for 1% of the variance in self-reported delinquency according to the R2 change statistic. Thus, it could be argued that the three-way interaction term effect size was small according to conventional standard (Cohen 1988). However, it is noteworthy that we relied on a full-factorial approach in the present study and consequently, a small increment in variance was expected and can be meaningful (Jaccard et al. 1990).

According to our findings, the significant effect observed in our analyses for a three-way interaction among YPI dimensions suggests that the relationship between adolescents’ self-reports of deviant behavior on the SRDS and each YPI psychopathy trait is moderated by the joint presence of the other two YPI psychopathy traits. As displayed in Fig. 1, the positive relationship between CU traits and adolescents’ self-reports of delinquency was moderated by the presence of GM and II. That is, a stronger relationship between YPI CU dimension and SRDS total score is likely to be observed in subjects scoring high on both YPI GM and II dimensions than in subjects with low scores on YPI GM and II dimensions.

Interestingly, the lowest scores on the SRDS where observed when subject showed high scores on both GM and CU dimensions, but low scores on II dimension. This finding seems to underscore the importance of assessing II traits when the relationship between self-reported delinquency and psychopathy is at issue. Similar considerations held for GM dimension of psychopathy. Indeed, according to our findings, adolescents who scored high on GM dimension but low on CU and II dimensions of psychopathy, obtained higher scores on self-reported delinquency than adolescents who scored low on II dimension while scoring high on GM and CU dimensions. Thus, according to our findings, it may be useful to assess all three dimension of psychopathy in adolescence (e.g., Colins et al. 2014a, b; Salekin 2016a, b).

Interestingly, from a different perspective, Patrick et al. (2009) explicitly hypothesized that subject’s level of psychopathy represents his/her simultaneous position on the latent distribution of three major dimensions that are supposed to represent developmental pathways, as well as personality traits leading to psychopathy. In other terms, according to the so-called triarchic model of psychopathy (Patrick et al. 2009) psychopathy requires a high-meanness-by-high-boldness-by-high-disinhibition configuration (i.e., interaction), rather than the simple sum of scale scores measuring meanness, boldness, and disinhibition, respectively, in order to be diagnosed (Patrick & Drislane 2015). Although the YPI was not based on the triarchic model of psychopathy, our findings were consistent with previous data as well as Patrick et al.’s general contention indicating the need for taking into account subject’s response pattern on GM, CU, and II dimensions in psychopathy assessments (e.g., Colins et al. 2014a, b; Salekin 2016a, 2017).

Previous studies (e.g., Frick et al. 2005; Frick and White 2008) have shown that CU traits (e.g., lack of guilt, absence of empathy, callous use of others) may be relatively stable across childhood and adolescence and designate a group of youth with a particularly severe, aggressive, and stable pattern of antisocial behavior. Our findings support the hypothesis (e.g., Salekin 2016a) that it may be important to take into account also GM and II dimensions of psychopathy in addition to CU traits. Even though child psychopathy is becoming increasingly recognized, only the CU dimension has been incorporated in the Diagnostic and Statistical Manual of Mental Disorders, fifth edition (DSM–5). The fact that the limited prosocial emotion has been added as a conduct disorder specifier seemed to suggest that child psychopathy is relatively underrepresented in this major diagnostic system (Salekin 2016b; Salekin et al. 2018). From this perspective, the results of the present study may be useful in providing empirical data on the dimensions of psychopathy and their interaction in predicting self-reported delinquency. Indeed, our data seem to underline that relying on a three-factor model of psychopathy may help clinicians and researchers in identifying and treating the wide variety of youth with conduct problems (Salekin 2017).

We would also like to argue that our findings may have implications also for YPI construct validity. Indeed, considerations based on YPI dimensions internal consistency may suggest that it is possible to rely on YPI total score but also its individual dimensions (e.g., Zwaanswijk et al. 2016). Considerations based on the YPI dimensions relation to theoretically relevant external correlates adds to its nomological network (namely, adolescent’s delinquency). The findings strongly support retaining the YPI as a measure of psychopathy and its dimensions as underlying elements of the condition.

Although our study had a number of strengths (large sample, non-overlapping scales), our findings should be interpreted in the light of several limitations. Our adolescent sample was not a random sample. Although the sample is similar in many ways to the Italian adolescents, we cannot be sure that there are not geographical differences in the broader Italian population. Our study did not include identifiable clinic-referred, court-involved, or adjudicated delinquent adolescents, who might have more elevated levels of psychopathic traits, or in other ways shed light on the constructs applicability to those specific groups. Thus, additional work is needed with these samples. Our sample was a mixed gender sample and additional research is needed on both males and females to determine if there are any potential gender effects (Verona, Sadeh & Javdani 2010). Although we conducted a large number of tests to examine gender effects (entered in first step of regression as well as created subsamples to separately examine if there were gender differences), and no such effects emerged, we would nonetheless encourage other researchers to test for potential gender effects in their studies.

Of course, relying on self-report to assess both independent (i.e., YPI dimension scores) and dependent variables (i.e., SRDS total score) may have yielded regression coefficient estimates that were positively inflated by shared method variance. This method limitation indicates the need for further studies based on a multi-method approach. Testing the effect of interactions among the YPI dimensions with respect to SRDS total score may be considered limited evidence in terms of a broader nomological network validity; however, it should be observed that criminal behavior is considered a major problem in psychopathy (e.g., Cooke and Michie 2001; Hare 2003). Finally, although we observed significant interaction effects among YPI dimensions in predicting the SRDS total score, the relevance of three-way interaction among psychopathy dimensions in predicting other relevant external variables - e.g., school/occupational failures, criminal recidivism, substance misuse, etc., should be evaluated in future studies. This will give us critical information regarding what factors might be important for treatment of those youth with conduct problems and psychopathic traits.