Cognitive and affective perspective taking amongst adolescent offenders with variants of callous–unemotional traits

The notion that antisocial individuals are characterised by deficits in empathy has a long history in the clinical and forensic literature; however, empirical evidence of this has often been inconsistent [1,2,3]. Recent advances in the conceptualization of empathy and antisocial behaviour have been informed extensively by models of heterogeneity amongst individuals with antisocial behaviour, particularly the distinction between those with high versus low levels of callous and unemotional (CU) traits. CU traits comprise features including a lack of remorse or guilt, deficient affect and lack of caring towards others, and have been conceptualised based on the affective component of psychopathy. A lack of empathy is considered core to psychopathy and CU traits, and accounts of this have emphasised a distinction between affective and cognitive constructs underlying empathy [4].

As conceptualised in current accounts (e.g. [5, 6]), affective components of empathy include processes such as affective sharing and physiological arousal to another’s emotion. Cognitive components include mentalising processes such as Theory of Mind (ToM), which refers to the capacity to understand others’ mental states; and perspective taking (PT), which refers to the recognition and understanding of another’s viewpoint based on situational cues. A further distinction has been made between cognitive PT and affective PT, the latter of which involves adopting another’s perspective to understand their emotional state. Current theory has further emphasised the hierarchical structure of these social-cognitive processes, differentiating predominantly cognitive mentalising processing, predominantly affective processes when witnessing others’ emotions, based on shared emotional, motor and somatosensory representations; and combined processes that engage cognitive and affective functions in parallel [7, 8].

Substantial neuroimaging research has implicated specific brain regions (e.g. medial prefrontal cortex, posterior cingulate cortex, precuneus and temporal parietal junction) in both cognitive and affective mentalising [9], and functional connectivity between these key brain regions is understood to develop across childhood to form a social network [10]. Evidence also suggests that specific brain regions are engaged during cognitive PT (e.g. dorsomedial and dorsolateral prefrontal cortex) as opposed to affective PT (e.g. ventromedial prefrontal cortex, amygdala and basal ganglia), and that cognitive PT appears more dependent on executive function [11]. Further, computational modeling has indicated that activation in the anterior cingulate cortex is associated with pro-social learning, and is stronger in individuals higher in empathy [12], who learn more quickly when benefitting others [13].

Youth with high levels of antisocial behaviour have been found to show disrupted connectivity between social network brain regions during social cognition tasks [14], as well as less efficient connectivity in frontoparietal regions supporting executive function [15]. Higher levels of conduct problems have also been found to be associated with delays in cognitive processes during pro-social decision making [16]. This evidence may explain why PT has been found to be significantly impaired amongst youth offenders compared to non-offenders [17], as well as amongst antisocial children [18].

Importantly, there is also extensive evidence that CU traits and psychopathy are associated with impairments in domains related to ToM and PT [19,20,21]. In terms of mechanisms that may account for this, youth with high levels of CU traits have been found to demonstrate irregular functional connectivity between social network brain regions including posterior cingulate cortex, precuneus, dorsomedial prefrontal cortex and amygdala (e.g. [22]). Higher CU traits have also been found to be associated with less functional connectivity between social network brain regions, which support perspective taking, and conflict network brain regions (e.g. anterior cingulate cortex, pre-supplementary motor area), linked to cognitive control [23], as well as information accumulation processes which support self-serving decisions during pro-social decision making [16].

Emerging evidence has, however, suggested that deficits in these cognitive components of empathy may be far more pronounced and complex than previously thought (for a review see [24]), both in adults with psychopathic traits (e.g. [25]), and children with CU traits (e.g. [26]). Evidence has further suggested that CU traits may be significantly associated with specific forms of PT in childhood and adolescence. In one key study, Anastassiou-Hadjicharalambos and Warden [27] found different associations for cognitive and affective PT in children. Their results revealed that, compared to typically social peers, antisocial children with low CU traits had deficits in both cognitive and affective PT, whilst those with high CU traits had selective deficits in affective PT. These findings are supported by evidence that conduct disordered children with high CU traits presented with deficits in affective PT [28] and that, amongst adolescents disengaged from school, CU traits were associated with a relative deficit in affective PT [29].

More recent evidence suggests that, for some individuals, CU traits could be associated with enhanced skills in cognitive PT. This evidence is based on the subtyping of primary and secondary variants of CU traits in terms of concurrent level of anxiety (primary variant, low anxiety; secondary variant, high anxiety) [30]. In novel research involving incarcerated youth offenders, Kahn et al. [31] investigated the moderating role of anxiety in associations between CU traits and empathy. Their results revealed no significant interactions between CU traits and anxiety in relation to either self-report or laboratory measures of affective empathy. However, findings demonstrated that CU traits and self-reported cognitive empathy were negatively correlated in individuals with high anxiety (i.e. secondary variant CU traits), but uncorrelated in individuals with low anxiety (i.e. primary variant CU traits). Further, these authors were able to assess different elements of cognitive empathy through a PT task, in which participants were required to interpret a character’s thought (cognitive PT) or feeling (affective PT). CU traits and cognitive PT were found to be positively correlated in individuals with low anxiety (primary variant CU traits), but not significantly correlated in individuals with high anxiety (secondary variant CU traits). In contrast, there were no significant interactions between CU traits and anxiety in relation to affective PT [31]. These potentially important yet unpredicted findings were highlighted as worthy of further investigation by Kahn et al. [31], as understanding the differences in empathy deficits between individuals with primary and secondary variants of CU traits could better inform individualised intervention for antisocial youth.

The major aim of the current study was to examine associations between PT, CU traits and youth offending. Moreover, attention was given to both cognitive and affective forms of PT, and to primary (high CU traits, low anxiety) and secondary (high CU traits, high anxiety) variants of CU traits. Although emerging evidence has been reported to suggest that youth with distinct variants of CU traits exhibit somewhat distinct deficits in cognitive PT (e.g. [31]), current conceptualisations of cognitive PT emphasise distinctions that have rarely been incorporated into such research to date. As such, the current study was designed to provide novel tests of these associations based on a more fine-grained examination of PT addressing both first- and second-order cognitive PT, as well as affective PT. In addition, there is a need for such research to index PT using ecologically valid measures that reflect real-world skills (as utilised by [18]), given the reliance on text, verbal and picture-based stimuli to date (e.g. [27, 31]). For this reason, in the current study, forms of PT were indexed using a video-based method involving conversational exchange in real-world social settings.

It was predicted, first, that antisocial behaviour would be associated with deficits in both cognitive and affective PT, such that youth offenders would have significantly poorer PT skills than non-offenders. The remaining hypotheses addressed associations between CU traits and PT based on CU variant and form of PT. Second, it was hypothesised that both primary and secondary variants of CU traits would be associated with poorer affective PT. Third, it was hypothesised that higher CU traits would be associated with better cognitive PT skills, but only amongst youth with primary variant CU traits. No such association was expected for youth with secondary variant CU traits or low CU traits.

Method

Participants

Participants were 130 male adolescents between the ages of 13 and 20 years (M = 16.32, SD = 1.35). Participants were included in the current research if they had undertaken most of their schooling in an English-speaking country, did not have a known diagnosis of intellectual impairment or hearing impairment, and were not known to be experiencing an acute episode of mental illness. Youth offenders (n = 81) were recruited through 19 youth justice centres (detention = 6; community service = 13), whilst non-offenders (n = 49) were recruited through five public secondary schools. All participants were resident in New South Wales, the most populous state in Australia. Most participants reported non-Indigenous Australian ethnicity (53.8%; majority Caucasian), but a substantial proportion reported Indigenous ethnicity (46.2%; majority Aboriginal). All participants reported that their primary language of communication was Standard Australian English. Using Socio-Economic Indexes for Areas [32], participants were assigned an Index of Relative Socio-Economic Advantage/Disadvantage (IRSAD; from one (lowest) to nine (highest)) based on postcode of usual residence. Participants had a mean IRSAD of 3.26, indicating relatively greater disadvantage and lack of advantage in general, consistent with lower socio-economic status (SES).

Measures

Offender status was operationalised as a two-level categorical variable (youth offender; non-offender). A youth offender was categorised as such based on officially documented contact with a youth justice agency at the time of participation in research. This official contact could be in the form of either supervision through a youth justice community service or detention in a secure youth justice centre. Non-offender categorisation was based on self-report of no current or historical official contact with a youth justice agency.

CU traits were measured using the Inventory of Callous-Unemotional Traits (ICU; [33]). This self-report scale, based on restructuring of the Callous-Unemotional subscale of the Antisocial Process Screening Device (APSD; [34]), was designed to comprehensively assess the presence and magnitude of CU traits in youth. Participants were asked to respond to 24 items that were each rated on a 4-point scale (0 = not at all true, 1 = sometimes true, 2 = very true, and 3 = definitely true). However, given that items 2 and 10 have not shown consistently strong correlations with total scores in the self-report format [35], these items were excluded in calculating a 22-item-based total score. The validity of the 22-item ICU has been demonstrated in a range of research involving community (e.g. [36, 37]) and forensic (e.g. [35, 38]) samples of adolescents. In the current study, analysis revealed a Cronbach’s α score of 0.82 for the total 22-item scale, which was comparable to findings from previously referenced research (a range of 0.78–0.83).

Anxiety was measured using the Anxious-Depressed subscale of the Youth Self Report (YSR; [39]). The YSR is a questionnaire designed to assess adaptive and maladaptive functioning in adolescents, forms part of the Achenbach System of Empirically Based Assessment (ASEBA) and has been normed for ages 11–18 years. The Anxious-Depressed subscale consists of 16 items that are scored on a 3-point scale (0 = not true, 1 = somewhat true, 2 = very or often true). The YSR has demonstrated reliability and validity [39], and has been used extensively in research investigating psychosocial functioning in adolescents. Cronbach’s α reliabilities for the Anxious-Depressed subscale of the YSR in samples of adolescents, including samples of youth offenders, have been demonstrated to be high, ranging from 0.80 to 0.86 (e.g. [40,41,42]). In the current study, analysis revealed a Cronbach’s α 0.81 for the Anxious-Depressed subscale.

The Awareness of Social Inference Test (TASIT; [43]) was developed as an ecologically valid means to systematically assess different facets of social perception. TASIT has been designed for use with ages 13–60 years, and to differentiate between neurologically typical individuals and those with compromised skills. Participants were assessed using the Social Inference-Minimal Task, which involved viewing a series of 15 short, videotaped vignettes of actors interacting in everyday conversational exchanges. Five of these scenes represented sincere exchanges, where words and meaning were consistent, and ten represented sarcastic exchanges, in which paralinguistic cues indicate inconsistency between words and meaning. After watching each scene, participants were asked four questions, each capturing a distinct facet of the inferential process [44]. Two questions were representative of cognitive PT. ‘Belief’ questions examined the participants’ capacity to accurately construe what the speaker knew or believed, and represented first-order cognitive PT. ‘Intent’ questions examined the participants’ capacity to accurately construe what the speaker intended (including what they intended the listener to believe) and represented second-order cognitive PT. ‘Feel’ questions, representative of affective PT, examined the participants’ capacity to accurately construe what the speaker was feeling. ‘Say’ questions examined the participants’ capacity to accurately construe meaning from the conversational exchange but were not the focus of analysis in the current study. Participants were allocated one point for each correct response. Question category scores were then summed to produce total belief (first-order cognitive PT), intent (second-order cognitive PT) and feel (affective PT) scores.

Procedure

The University of Sydney Human Research Ethics Committee, the Department of Communities and Justice, NSW, and the Department of Education, NSW, approved this research. Having received these approvals, inclusion criteria, participant information and consent forms were distributed to participating youth justice centres and schools. The first author visited all participating youth justice centres and schools to administer assessments and questionnaires on site. Testing commenced with a semi-structured interview followed by assessment of non-verbal ability with the Matrices subtest of the Kaufman Brief Intelligence Test, 2nd edition (KBIT-2) [45]. All other measures were presented in a random order. All items on the TASIT were read to participants, and demonstrations and practice opportunities were provided. Participants could choose to have items on the ICU and YSR read to them and have their responses recorded for them by the first author.

Data analytic plan

All analyses were conducted using SPSS, Version 27 [46]. An a priori power analysis was conducted using G*Power 3.1 [47], based on recommendations by Dattalo [48], and considering use of a separate hierarchical regression analysis for each of three dependent variables. Based on the assumptions of an alpha of 0.05, a power of 0.8, and a medium effect size (Cohens f2 = 0.15), it was determined that the minimum desired sample size was 99.

Preliminary analyses revealed no violation of the assumptions of normality, linearity, multicollinearity and homoscedasticity. Hypotheses were tested using a set of three hierarchical regression analyses consisting of a separate model for each of the dependent variables of cognitive (first order; second order) and affective PT. All models included the same independent variables, which were entered in two blocks. Block one consisted of the continuous variables of age, SES, and non-verbal ability, and the recoded weighted categorical variable of ethnicity. Also entered in this block were the mean centred variables of CU traits and anxiety and the categorical variable of offender status (youth offender; non-offender). To test for associations with variants of CU traits based on high or low anxiety, the two-way interaction term for CU x anxiety was entered in a second block. To increase confidence in the statistical significance of coefficient terms, each hierarchical regression analysis was reanalysed on 2000 wild bootstrap samples [49] and the bootstrap distribution was corrected for bias and acceleration [50]. Significant interactions were probed using simple slope analyses involving tests of conditional effects of CU traits on each PT variable at high (1 standard deviation above the mean), medium (mean) and low (1 standard deviation below the mean) levels of anxiety, as described in Bauer and Curran [51].

Results

Descriptive statistics

Means, standard deviations and zero order correlations for relevant study variables are shown in Table 1. Status as a youth offender was associated with higher levels of CU traits (youth offender M = 25.41; non-offender M = 20.49) and anxiety (youth offender M = 6.23; non-offender M = 4.67). Status as a youth offender was also associated with poorer first-order cognitive PT (youth offender M = 11.70; non-offender M = 12.57), second-order cognitive PT (youth offender M = 10.98; non-offender M = 12.10) and affective PT (youth offender M = 12.79; non-offender M = 13.57). There were no significant correlations between CU traits or anxiety and either cognitive or affective PT. Age was significantly correlated with offence status and anxiety, with older age associated with status as a youth offender and slightly higher anxiety. Ethnicity was significantly correlated with both first- and second-order cognitive PT, but not affective PT. Non-Indigenous identification was associated with better cognitive PT. In contrast, higher SES was associated with better affective PT, but SES was not significantly correlated with cognitive PT. Non-verbal ability (NVA) was significantly correlated with offence status and PT, with higher NVA associated with status as a non-offender, and better cognitive and affective PT.

Table 1 Descriptive statistics and zero-order correlations
Full size table

Tests of main study hypotheses

The results of hierarchical regression analyses testing main effects of offender status, CU traits and anxiety, as well as interactions between CU traits and anxiety, on cognitive and affective perspective taking appear in Table 2.

Table 2 Hierarchical regression analyses for cognitive and affective perspective taking with predictors offence status, CU traits and anxiety
Full size table

In the model testing predictors of first-order cognitive PT, there was a significant main effect for offender status (ß = −0.33, p = 0.003), in which status as an offender was associated with poorer first-order cognitive PT. There were no significant main effects for CU traits or anxiety. However, there was a significant interaction between CU traits and anxiety in this model (ß = −0.27, p < 0.001). This significant interaction was probed by testing the conditional effects of CU traits on first-order cognitive PT at high and low anxiety (as illustrated in Fig. 1a). CU traits were significantly related to first-order cognitive PT when anxiety was low (ß = 0.35, p = 0.002), but not when anxiety was high. That is, higher scores on CU traits were associated with better first-order cognitive PT, but only for youth low in anxiety (i.e. primary variant CU traits). For youth high in anxiety (i.e. secondary variant CU traits), there were no significant associations between CU traits and first-order cognitive PT.

Fig. 1
figure 1

Associations between perspective taking and CU traits amongst youth with high versus low levels of anxiety. (a) First-order cognitive PT. (b) Second-order cognitive PT. (c) Affective PT

Full size image

In the model testing predictors of second-order cognitive PT, there was a significant main effect for offender status (ß = −0.23, p = 0.029), in which status as an offender was associated with poorer second-order cognitive PT. There were no significant main effects for CU traits or anxiety. However, there was a significant interaction between CU traits and anxiety in this model (ß = −0.35, p < 0.001). This significant interaction was probed by testing the conditional effects of CU traits on second-order cognitive PT at high and low anxiety (as illustrated in Fig. 1b). CU traits were significantly related to second-order cognitive PT when anxiety was low (ß = 0.38, p < 0.001), as well as when anxiety was high (ß = −0.33, p = 0.003). That is, higher scores on CU traits were associated with better second-order cognitive PT for youth low in anxiety (i.e. primary variant CU traits), but with poorer second-order cognitive PT for youth high in anxiety (i.e. secondary variant CU traits).

In the model testing predictors of affective PT, there was a significant main effect for offender status (ß = −0.33, p = 0.004), in which status as an offender was associated with poorer affective PT. There were no significant main effects for CU traits or anxiety. However, there was a significant interaction between CU traits and anxiety in this model (ß = −0.24, p = 0.004). This significant interaction was probed by testing the conditional effects of CU traits on affective PT at high and low anxiety (as illustrated in Fig. 1c). CU traits were significantly related to affective PT when anxiety was low (ß = 0.38, p = 0.001), but not when anxiety was high. That is, higher scores on CU traits were associated with better affective PT, but only for youth low in anxiety (i.e. primary variant CU traits). For youth high in anxiety (i.e. secondary variant CU traits), there were no significant associations between CU traits and affective PT.

Discussion

The current study investigated associations between antisocial behaviour, CU traits and cognitive and affective perspective taking. Findings support the hypothesis that associations between CU traits and cognitive PT are moderated by anxiety and may, therefore, differ between putative variants of CU traits. Specifically, amongst low-anxiety youth, higher levels of CU traits were associated with better cognitive PT skills (both first and second order). This finding is consistent with Kahn et al. [31] who found that CU traits were associated with better cognitive PT, but only amongst adolescents with lower levels of anxiety. Amongst high-anxiety youth, there was no association between these CU traits and first-order cognitive PT. Amongst high-anxiety youth, higher CU traits were, however, associated with poorer second-order cognitive PT. Considerable research has previously tested, and supported the associations between CU traits and multiple components of empathy (e.g. PT, ToM), independent of anxiety-based variants. Our findings build on this previous work by suggesting that these associations may further vary according to type of variant, as well as type of PT. They suggest that associations between CU traits and cognitive PT vary both according to level of anxiety, and according to type of cognitive PT (first- and second-order forms). This may explain why our results diverge somewhat from research based on more global indices of PT, such as previous findings of a non-significant negative association between CU traits and overall cognitive PT in individuals with high levels of anxiety [31].

The current findings are also consistent with broader conceptualisations of primary and secondary variant CU traits. Amongst high-anxiety youth in the current sample, CU traits were associated with poorer skills for inferring what a speaker intended another to believe (second-order cognitive PT), suggesting deficits in more complex cognitive PT skills. This is consistent with previous findings for antisocial children [18], and may be indicative of reduced connectivity amongst neural social networks found in youth with elevated CU traits [22]. It is also noteworthy that the secondary variant of CU traits is associated with more severe childhood maltreatment [52,53,54], which has been linked to delayed development of higher level theory of mind [55,56,57], as well as decreased functional connectivity between key brain regions [58]. In contrast, amongst low-anxiety individuals in the current sample, CU traits were associated with enhanced ability to infer both what a speaker believed and what a speaker intended another to believe.

As noted by Kahn et al. [31], the notion that primary variant CU traits may be associated with better cognitive perspective taking can be seen as consistent with theories and past research suggesting that individuals with these traits have an enhanced ability to notice when others are vulnerable, which may facilitate self-serving manipulative behaviour towards others [59,60,61,62]. Not only was this form of cognitive empathy specifically associated with CU traits amongst low-anxiety youth in our sample, but so too was affective PT—the other major form of cognitive empathy indexed. It is noteworthy that the relationship between affective PT and CU traits was previously not found to be moderated by anxiety in the only study that has investigated this specific association to date [31]. It is unclear whether methodological differences in sampling or measurement in the current study may account for the mixed findings between these studies. It is important to note that TASIT places demands on complex social inference by requiring participants to process cues in social interactions that specifically involve sarcasm. Recent evidence suggests that high levels of psychopathic traits are more strongly associated with use of sarcasm and associated aggressive humour styles than high levels of other “dark triad” personality traits [63]. Also, it has been suggested that individuals high in psychopathy exhibit difficulties in affective ToM which are relative to the complexity of evaluation required and associated demand on cognitive resources [62]. In light of this, it is conceivable that greater attunement to sarcastic communication forms may have limited the cognitive demands involved in processing affective PT amongst those higher in primary CU traits in the current study.

The method used to measure PT in the current study may also potentially account for there being no significant relationship found between overall CU traits (i.e. not distinguished by level of anxiety) and the PT variables. The instructions for TASIT serve to cue participants to attend to the information in the PT vignettes that pertain to the follow-up questions, and inform participants that these questions are presented in a set structure for each vignette. There is some evidence that individuals high in CU or psychopathic traits are capable of engaging in both cognitive PT [64] and affective PT [65] when provided with such goal directed attentional instructions. In addition, TASIT involves a particularly real-world test of PT comprising video-based interpersonal cues [43]. There is evidence that similar video-based social cognition tasks co-activate brain regions associated with both cognitive and affective processes (e.g. [66]), which may facilitate engagement of more neurocognitive resources [8]. In contrast, previous research has often indexed PT using tasks that activate brain regions associated with cognitive processes specifically [8]. Further, whilst TASIT vignettes include representations of both basic (e.g. happy, angry) and complex (e.g. annoyed, impressed) emotions, the response format for affective PT involved naming a single emotion and nominating the presence or absence of that emotion. Higher CU traits have been shown to be associated with a greater deficit in recognising complex emotions than basic emotions [67], which is amplified when increased demands are placed on cognitive control [68]. However, in these studies, the task (i.e. Reading the Mind in the Eyes) featured a potentially more complex response format than TASIT, requiring participants to select the appropriate emotion from a list of four options. Various elements of TASIT, including instructions, stimuli and response format, may have, therefore, inadvertently supported high-CU participants to engage effectively in PT through provision of attentional cues, prompting engagement of neurocognitive resources and minimising task complexity.

In terms of associations between antisocial behaviour and perspective taking, participant status as an offender was associated with both significantly poorer cognitive PT (first and second order) and significantly poorer affective PT. This is consistent with predictions, and with previous research that has found adolescent and adult offenders to have poor perspective taking skills in general [17, 69, 70]. Current findings are also consistent with evidence that individuals high in antisocial behaviour show disrupted neural connectivity in regions supporting executive function when engaged in socio-emotional processing, which may impede information processing and decision making [15].

The current findings have two key implications for intervention and service provision for antisocial youth. First, the current findings highlight the importance of examining risk and protective factors for antisocial behaviour at a fine-grained level. Specifically, current evidence underscores the contribution assessment of anxiety and elements of empathy could bring to understanding both the developmental histories and prospects of antisocial individuals with high levels of CU traits. This is particularly important for youth engaged with the justice system, where the risk-need-responsivity model [71] and risk assessment inventories (e.g. [72]) influence both the level of supervision and type of intervention an individual is likely to receive in response to their criminal behaviour [73]. Second, it is possible that by adolescence, individuals with primary variant CU traits may have already begun to adapt PT skills for antisocial purposes, whilst for individuals with secondary variant CU traits, poorer PT skills may have increased vulnerability to engagement in antisocial behaviour. Earlier identification could, at the very least, provide opportunities for development of core skills (such as perspective taking) that might facilitate more appropriate socialisation (e.g. [52, 74]), with specific strategies for intervention informed by evidence of patterns of neurocognitive activation during pro-social learning [68].

When interpreting the results reported for this study, several limitations must be taken into consideration. Given that analyses did not control for the severity of conduct problems, which is known to covary with CU traits, we cannot infer that the results are attributable to CU traits alone. Moreover, although age was included as a covariate, we did not control for pubertal development, which has previously been associated with both CU traits and ToM. The current sample included only male participants. It is important that findings are replicated with female participants because there is evidence that different associations exist between CU traits and anxiety [26], empathy [75] and antisocial behaviour [76] in females, as well as profiles of cognitive processes that are distinct for males [16]. In addition, as this was a cross-sectional study, it is essential that the results are not interpreted in terms of causal mechanisms. Ideally, the findings from the current and related (e.g. [29, 31]) studies will inform future longitudinal investigations of the temporal development of different facets of cognitive and affective PT amongst youth at high risk for developing primary or secondary variants of CU traits.

Although multiple forms of PT were examined in the current study, these were nonetheless indexed by a single test. In future research, assessment made up of several tests of these capacities could provide more detailed evidence of the differences in PT skill profiles (e.g. [18, 77]) between individuals exhibiting primary and secondary variants of CU traits. Also, analysis of composite measures of multiple tests may potentially provide more robust support for the emerging evidence that primary variant CU traits may be associated with better PT ability. Finally, it should be noted also that whilst our analytic plan was informed by that of Kahn et al. [31], person-centred clustering approaches such as latent class analysis may offer further advantages, and should be considered in future research.

The findings of the current study highlight the value of applying a fine-grained conceptualisation to several developmental factors implicated in current models of antisocial behaviour. Our findings support the distinction between primary and secondary variants of adolescent CU traits based on level of anxiety, which in our sample were associated with significantly different skill profiles for perspective taking. This may in turn indicate that these subgroups of youth may benefit from distinct intervention. Our findings may also inform models of the developmental mechanisms that contribute to antisocial outcomes and CU traits across childhood and adolescence. That anxiety moderated associations between CU traits and higher levels of cognitive and affective PT skills in the manner seen here suggests that perception of the beliefs, intentions and feelings of others may facilitate manipulative antisocial behaviour in individuals with primary variant CU traits. Whilst literature on psychopathy and CU traits has generally emphasised deficits in empathy, it has become apparent that this is not universal to all manifestations of psychopathy or CU traits, or all forms of empathy. As has been demonstrated in the current study, a key strategy in developing a greater understanding of these associations is to employ a nuanced perspective.