Proposed Specifiers for Conduct Disorder-Short Version (PSCD-SV): Psychometric Properties, Concurrent Correlates and Parenting Predictors

The sample used in the current study is part of a three-year longitudinal study which was conducted in the Autonomous Community of Galicia (NW Spain) between 2017 and 2019. The initial sample included a total of 642 adolescents in the first grade of compulsory secondary education [1° ESO] from 11 state secondary schools. Participants in T1 ranged in age from 12 to 15 (M = 12.49; SD = 0.67), and 45.4% were females. The second wave that took place one-year after the first assessment (T2) included a total of 625 adolescents aged 13–17 (M = 13.43; SD = 0.68), of which 47.5% were females. Finally, a total of 627 adolescents took part in the third wave which was carried out 2 years after the first assessment (T3). Participants in T3 aged 14–18 (M = 14.42; SD = 0.68), 48.6% females, and were involved in the third grade of compulsory secondary education [3° ESO]. It should be noted that second and third waves included not only participants assessed at the first wave but also new participants that met the criteria for their inclusion in the study. For the purposes of the current study, only adolescents who were assessed at T1 and followed-up at T3 were included in the analysis. This gave rise to a total sample of 414 adolescents aged 12–15 at the first assessment (M = 12.32; SD = 0.50; 49.2% females), which was followed up 2 years later when participants were in third grade (M = 14.18; SD = 0.40). Most of participants lived with both parents at T1 (84.7%), whereas 12.1% lived only with their mother, 1.9% lived only with their father, and 1.2% lived with other relatives. More than 90% of the sample were Galician, white, and came from middle and low-middle socio-economic backgrounds.

The procedure followed throughout the investigation was in compliance with the standards of the University Ethics Committee and the Declaration of Helsinki. Following a convenience sampling approach, the heads of 24 secondary schools were initially contacted by phone to present the study and ask for potential collaboration. Schools located in the four provinces of the Autonomous Community, and representative of this area, were contacted. A brief report with more detailed information about the research project and the goals of the study were sent by mail to all those interested in taking part in. Specifically, the main goal of this longitudinal study was to delve into the role that certain parenting practices have on adolescent development, as well as the interactional effect with other factors such as peers or personality traits. Hence, to analyze the psychosocial development from early adolescence, the target population of study encompassed all students involved in the first degree of secondary education at the beginning of the project. After the first contact, a total of 11 secondary schools agreed to participate in the longitudinal study. Qualified psychologists from the research group visited all the schools to explain the procedures and solve any doubts that may arise regarding the prospective study. Parental consent was requested each year before the beginning of data collection and, subsequently, adolescent assent was obtained during questionnaire implementations. Given that informed consent was annually requested, no restrictions regarding new participations were made for their inclusion in the study. Data collection was conducted by qualified research assistants, who visited the centers during school hours in order to provide proper instructions to the adolescents and monitor the process. Adolescents filled out the questionnaires in group sessions of approximately 50 min in classroom. Adolescent participation was voluntary, and confidentiality was guaranteed following the ethical criteria. Personal but anonymous codes were used to link questionnaires’ information from different waves of study. A total of three waves of data collection were carried out during a three-year period, with intervals of approximately 12 months between follow-ups.

Descriptive statistics for main study variables were first examined in SPSS 25. Second, in order to replicate the factorial structure of the short version of the PSCD proposed by Luo et al. [15], a Confirmatory Factor Analysis (CFA) was conducted in Mplus 7.4 [58], with robust weighted least squares used as estimator (WLSMV). A four-factor interrelated and a four-factor superordinate model were specified, including the 13 items as observed variables and psychopathy as a second order factor. Model fit was assessed using root-mean-square error of approximation (RMSEA), comparative fit index (CFI), and the Tucker–Lewis index (TLI). According to Hu and Bentler [59] suggestions, RMSEA values lower or equal to 0.05, and TLI and CFI values of 0.95 or higher were considered indicators of good model fit, whereas a RMSEA smaller than 0.08, and TLI and CFI larger than 0.90 indicated adequate model fit. Internal consistency of the PSCD-SV total score was computed with Cronbach’s alpha (α). However, given the dependence of alpha on the number of items in a scale, as well as other concerns about its reliance (e.g., the normal distribution of items) [60], mean inter-item correlation (MIC) was computed as a more appropriate indicator of the internal consistency for the PSCD-SV subscales (i.e., 3 items per subscale), with values ranging from 0.15 to 0.50 being considered adequate [61]. Concurrent associations of the PSCD factors and Total score with external variables measuring behavioural, emotional and psychosocial problems, antisocial behaviour and parenting practices were examined by computing zero-order correlations in SPSS 25. Partial correlations controlling for age and gender were computed for PSCD Total. For each PSCD factor, we also computed partial correlations controlling for the effect of gender, age and the other three PSCD factors. In this way, we were able to examine the differential associations of each PSCD factor with external criteria, above and beyond the shared variance among dimensions. To counteract the issue of multiple testing, Bonferroni’s correction was applied, and the threshold levels of significance were settled at 0.005 (11 variables). Finally, a Structural Equation Model (SEM) was computed in Mplus 7.4 using WLSMV as estimator, to analyze the influence of parenting practices in early adolescence on psychopathic traits 2 years later, including both parental support and parental psychological control as latent variables predicting the four latent psychopathic factors. SEM controlled for age and gender of the participants, people they lived with (coexistence), and antisocial behavior in T1. All models adjusted statistically for the school-level clustering of data and used listwise to manage missing data.

The four-factor model of the PSCD-SV showed an acceptable-to-good model fit, χ² (60) = 100.92, p ˂ 0.001, CFI = 0.94, TLI = 0.92, RMSEA = 0.05 [0.04, 0.07]. The four-factor model was better than the one-factor model, χ² (65) = 313.14, p ˂ 0.001, CFI = 0.78, TLI = 0.73, RMSEA = 0.10, and the three-factor model χ² (24) = 74.11, p ˂ 0.001, CFI = 0.93, TLI = 0.89, RMSEA = 0.07. Modification indices substantiated a residual covariance between items 15 and 14 as representing an acutely misspecified parameter in the model. After including this new parameter, a good model fit was observed for the four-factor model, χ² (60) = 100.92, p ˂ 0.001; CFI = 0.96, TLI = 0.95, RMSEA = 0.04 [0.03, 0.05]. Equal model fit indices were obtained for the two alternative four-factor interrelated and four-factor superordinate models (see Fig. 1 for a representation of the four-factor superordinate model). As can be observed in Table 2, all 13 items loaded significantly on the expected PSCD factor and the latent global construct, with all item and factor loadings being above 0.40.

In terms of internal consistency, Cronbach’s alpha was 0.73 for the PSCD Total score. Given the very low level of number of items for the subscales, we utilized the MIC to further examine scale homogeneity. All MIC values were indicative of an adequate internal consistency for all the PSCD factors (MIC = 0.30, 0.28, 0.47, 0.43 for GM, CU, DI, and CD respectively). The MIC for the Total score also showed good homogeneity (MIC = 0.35). Significant correlations (p < 0.001) were observed between the PSCD Total score and the four PSCD factors (r’s = 0.69^GM; 0.55^CU; 0.73^DI; 0.74^CD).

As displayed in Table 3, the PSCD Total score was significantly correlated with all analyzed variables. At the zero-order level, all PSCD factors showed significant correlations with conduct problems and all facets of antisocial behavior. DI and CD traits were significantly correlated with emotional problems and hyperactivity, which was also related with the GM factor. All GM, CU and CD factors were negatively correlated with prosocial behavior, whilst only CU traits showed a significant correlation with peer problems. With respect to parenting practices, GM, CU, and CD factors were negatively correlated with parental support, whilst only GM and CD significantly correlated with psychological control.

When controlling for the effect of the other psychopathic traits, the GM factor remained significantly correlated with conduct problems, hyperactivity, prosocial behavior, and parental support (both inversely) and psychological control (see Table 3). The CU factor, in contrast, only showed a unique significant (inverse) correlation with prosocial behavior. The DI factor remained significantly correlated with conduct problems, hyperactivity, and rule breaking. Finally, the CD factor remained significantly correlated with conduct problems, hyperactivity, all facets of antisocial behavior, and both parenting practices.

A SEM model was conducted to predict PSCD psychopathic traits two years following the initial assessment, including both parental support and psychological control as predictors. For methodological issues, the SEM model was based on the four-factor interrelated model of the PSCD-SV, allowing to test the unique association between parenting predictors and later psychopathic traits. This model, statistically equivalent than the four-factor superordinate model, provided a more parsimonious solution to be accommodated in SEM. In order to preliminary test the independent contribution of each parenting practice, two independent models were also tested. One included parental support as independent predictor of later psychopathic traits, χ² (241) = 264.41, p = 0.144, CFI = 0.96, TLI = 0.95, RMSEA = 0.02 [0.00, 0.03], showing that support significantly predicted lower levels of GM (β = − 0.27), CU (β = − 0.27), and CD (β = − 0.14). An alternative model included psychological control as T1 predictor χ² (242) = 269.63, p = 0.107, CFI = 0.96, TLI = 0.95, RMSEA = 0.02 [0.00, 0.03], showing that it significantly predicted higher levels of CU (β = 0.28) and CD (β = 0.18) (further details are available upon request). The integrated model, including both support and control, exhibited an adequate fit, χ² (452) = 486.14, p = 0.129, CFI = 0.97, TLI = 0.96, RMSEA = 0.01 [0.00, 0.02]. with all psychopathic traits being significantly correlated (r’s = 0.72^GM−CU; 0.59^GM−DI; 0.75^GM−CD; 0.41^CU−DI; 0.60^CU−CD; 0.66^DI−CD; ps < 0.001). Table 4 displays the unstandardized and standardized results of this model, as well as the explained variance. Gender (being male) significantly predicted GM, CU, and CD, and antisocial behavior at T1 significantly predicted higher levels of DI and CD. However, including both parenting practices (r = − 0.58, p < 0.001) did not reveal a significant effect on later psychopathic traits, as observed for the non-significant standardized effects of parenting practices on psychopathic traits. Only the unstandardized effect of psychological control on CU was statistically significant (Fig. 2).

The present study aimed to provide new evidence on the psychometric properties of the PSCD [6] and, more specifically, the self-reported short version (PSCD-SV) first examined by Luo et al. [15]. The PSCD-SV addresses interpersonal (i.e., GD), affective (i.e., CU) and behavioral/lifestyle (i.e., DI) dimensions of psychopathy, along with the four key CD symptom categories of CD and one ODD criteria [9, 23, 64]. According to the original proposal [9]; see also [15], the four-factor structure of the PSCD-SV was supported, representing the three psychopathy dimensions (GM, CU, and DI) along with CD, with an overarching latent hierarchical factor. The standardized factor loadings were overall higher than 0.60, with only four in the range of 0.40–0.60, being also supportive of the PSCD as a four-factor measure that distinctively assesses four different but interrelated dimensions of the psychopathic construct [1, 9]. Due to the low number of items per scale, we relied on the MICs as an estimate of reliability. The MICs for the total and subscale scores were good and therefore supported scale homogeneity.

At the zero-order level, both the total score and all the PSCD-SV factors were related with conduct problems and all facets of antisocial behavior [65], and all but one were significantly correlated with hyperactive behaviors (except CU traits) and—inversely—with prosocial behavior (except DI). Only CU traits showed a significant correlation with peer problems [66]. In regards to parenting, GM, CD and the Total score showed significant correlations with both support (inverse) and psychological control, whilst CU traits were negatively correlated with parental support, similar to what has been found in past research [35, 37, 40, 67]. Of note, there was a positive correlation between the Total score and emotional problems, which was essentially driven by DI and CD traits. The co-occurrence between anxiety and CD symptoms have been previously shown in some studies in childhood and adolescence [68], and may partially explain this result. Also, even considering that individuals with psychopathic traits have been traditionally defined as low anxious [69], anxiety and other related emotional problems have been considered as differential indicators of the primary (i.e., low anxious) and secondary variants (i.e., high anxious) of psychopathy, or at least component parts of psychopathy [70], a result that should be further explored in future research.

These results tend to be substantially weaker when controlling for other PSCD-SV factors, reinforcing the assumption that psychopathy dimensions depend at least in part on each other in their relation to external correlates [7]. In this regard, all psychopathic dimensions but one (i.e., CU traits) remained significantly correlated with conduct problems and hyperactivity. Significant correlations with prosocial behavior held for GM and CU traits. Nevertheless, with regard to antisocial behavior, current results suggest that the CD factor primarily accounted for most of the correlations between the PSCD factors and external criteria that are antisocial in nature. This possibility has been an issue in some conceptions of the psychopathic construct, in which conduct problems, antisocial behavior, or CD, have been treated as correlates of psychopathic personality instead as fundamental features [71]. However, evidence for the key role of antisocial behaviors in understanding and assessing the psychopathy construct is extensive [1, 9, 72‐75]. Moreover, several studies have shown that psychopathic traits are important for predicting several problematic outcomes, even after controlling for severity and the onset of conduct problems [76], or in the absence of concurrent conduct problems [77, 78].

Results from partial correlations also contrast with a large body of research based on CU traits as the potential hallmark of the construct of psychopathy, or for being considered as an identifier of a specific group of problematic children showing most serious patterns of conduct problems [79]. Also, once controlling for the effect of the other PSCD-SV psychopathic traits, CU traits only held significantly correlated with prosocial behavior [80], but not with conduct problems, suggesting that when examining CU traits, the effect of the other psychopathy factors should be also taken into account [7, 9]. In sum, regardless of whether using basic correlations or partial correlations, the correlations offer support for the validity of the PSCD and the multidimensionality of the psychopathy construct. Yet, results from partial correlations should be interpreted with caution since partialling may change variables, being difficult to know “what construct an independent variable represents once the variance shared with other independent variables is removed.” [81], p. 329]. Considering psychopathic personality as a constellation of co-occurring interpersonal, affective, and behavioural/lifestyle traits [1, 7, 10, 16], results based on their global and shared contribution should be, therefore, of primary interest.

Results from correlations showed a concurrent link between psychopathic traits as measured by the PSCD-SV and parenting practices, reinforcing some results observed in previous research [82, 83]. Overall, these findings lend further support to the validity of the PSCD-SV. However, the role of parenting practices could be understood beyond the concurrent association with psychopathic traits and, therefore, should be examined from a developmental perspective that may lead to better understanding of the potential mechanisms of change in psychopathy development or single component development. As was observed in the present study, parental support, defined as the perception of parental warmth, responsiveness, and closeness to parents, negatively predicted GM and CU traits 2 years later. Also, psychological control, based on psychological coercion and emotional manipulation, significantly predicted later CU traits and CD. These findings largely converge with previous research supporting the association between psychopathic traits—and more specifically CU traits—and an infrequent use of practices based on warmth, acceptance and involvement [36, 84], as well as deficient parent–child communication patterns [40], low autonomy transfer, and high harsh parenting [43]. However, it should be noted that the effect of both parenting practices was no longer significant when they were included together in the model, with just a marginal association between psychological control and CU traits. This points out the importance of taking into account the shared effects between different parenting practices when testing their influence on child’s development, and raises the need of further examining the role of parenting in the development of psychopathic traits.

The current results from the PSCD-SV, as a wider set of psychopathic traits and CD, may add to the literature by including the full array of psychopathic traits rather than only focusing in on one of them (i.e., CU traits). Therefore, results showed that not only CU traits, but also GM, could be affected by low parental support, warmth and affection, even after accounting for initial levels of antisocial behavior [35]. Also, the inclusion of psychological control as a predictor variable constitutes a novelty since it has rarely been addressed in previous research on this topic, although it has proved relevant for better understanding adolescent maladjustment [30, 33]. Interestingly, when the two parenting variables are included in the model, psychological control remained significantly related with CU traits, yet marginally, raising its importance as a potential developmental precursor of the affective component of psychopathy that should be also considered as a potential target for prevention and intervention purposes.

Notwithstanding these results, it should be noted that the pattern of relationship between parenting practices and psychopathic traits tend to be complex, with psychopathic traits involved in two-way effects with parent–child interactions [85‐87], and potential additional moderators affecting those associations [26]. Thereby, it has been suggested that the possible role of psychopathic traits in conferring greater risk for later developing conduct problems and antisocial behavior, might be derived by uniquely shaping dimensions of parenting practices [88], an issue that should be further examined in future research. In this regard, the association between parenting practices and psychopathic traits should be interpreted in the context of potential gene-environment interactions [89], where heritable patterns would potentially underpin the development of psychopathic traits [90], whilst parenting practices would play a role as a potential environmental-mediated predictor of later development of psychopathic traits [91]. To this point, it would be interesting to explore to what extent psychopathic traits in parents may led to parenting practices that exhibit greater levels of harshness or psychological control (as well as reduced parental support and involvement) which, in turn, will influence the development of overall psychopathic and more specific CU traits [92, 93]. Contrastingly, if the parenting seems somewhat “normal,” that is free of harshness, harshness does not affect the development of GM or DI then research will need to further explore if there are certain parenting practices that lead to those traits or if the traits are present regardless of the environment. This may mean that the traits come about naturally and are independent of parenting practices. It overall highlights the importance of include environmental factors in the study of psychopathic traits, in order to accurately depict their role in the development of this personality profile and, therefore, their potential usefulness for prevention.