Introduction

Callous-unemotional traits can be viewed as precursors to childhood psychopathy (Frick, 1998; Frick et al., 2000). The introduction of the specifier “with limited prosocial emotions” for conduct disorder (CD) in the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM.5; American Psychiatric Association, 2013) marked a notable addition and is commonly referred to as “callous-unemotional” (CU) traits. CU traits are prevalent in up to 50% of children and adolescents diagnosed with CD (Kahn et al., 2012). CU traits are understood to be a complex and multidimensional construct encompassing affective and interpersonal aspects (Frick & Ray, 2015). Children who display callous-unemotional traits typically lack regret and empathy, engage in manipulative behaviors for their own benefit, and express their emotions in a shallow way (Frick & White, 2008). The examination of callous-unemotional traits has gained significant attention in the field of developmental psychology and psychiatry due to its association with various negative outcomes, including antisocial behavior (Waller et al., 2017; Zych et al., 2019), impairments in moral development (Craig et al., 2021), and conduct disorder (Frick et al., 2003). Understanding the etiology and factors contributing to CU traits is crucial for effective intervention and prevention strategies (Hawes et al., 2014a).

Callous-unemotional traits can be defined by three distinct dimensions: callousness, uncaring, and unemotionality. Callousness represents a state characterized by an absence of remorse and concern for others (Cardinale & Marsh, 2020). Uncaring, on the other hand, denotes a general indifference towards both others and one’s own performance (Cardinale & Marsh, 2020). Lastly, unemotionality signifies a limited display or expression of emotions (Cardinale & Marsh, 2020). The three distinct dimensions of CU traits—callousness, uncaring, and unemotional—necessitate a comprehensive understanding for tailored interventions and preventive measures. To evaluate callous-unemotional traits in children and adolescents different, instruments exist (e.g., Youth Psychopathy Traits Inventory; Psychopathy Checklist: Youth Version; Child Problematic Traits Inventory; Andershed et al., 2007; Colins et al., 2014; Hare, 1991, 2003). However, researchers often utilize the Inventory of Callous-Unemotional Traits (ICU) questionnaire (Frick, 2004; Gao & Zhang, 2016; Houghton et al., 2013). The ICU has therefore significantly contributed to the development of the specifier introduced in the DSM-5 (American Psychiatric Association, 2013). The ICU has different versions (parents, teachers, and self-report versions and ICU-Parent Preschool and -Teacher Preschool versions). However, the different versions contain similar items in each version. Some items may have been formulated differently in the different versions, although the underlying meaning has remained the same (Frick, 2004). The ICU consists of 24 items and includes the three mentioned subscales: callousness, uncaring, and unemotional traits (Frick, 2004). In this study, the German translation of the self-report questionnaire was used (Essau et al., 2006). Studies have found that a three-factor bifactor structure, comprising an overarching callous-unemotional trait and its underlying dimensions of callousness, uncaring, and unemotionality, yields the most favorable model fit across different samples (Cardinale & Marsh, 2020; Essau et al., 2006). Cardinale and Marsh’s (2020) meta-analysis revealed satisfactory pooled internal consistency for the total score of ICU (α̅ =.83) and the three scales (unemotional: α̅ =.71, uncaring: α̅ =.80, and callousness: α̅ =.75). Nevertheless, the unemotionality dimension has methodological gaps, including concerns about its validity, limited ability to predict typical outcomes like externalizing behavior problems, and weak associations with psychopathic traits (Cardinale & Marsh, 2020).

Factor structure of callous-unemotional traits in middle childhood

Originally conceived as a unidimensional measure, the ICU is now widely recognized for its three-factor bifactor structure, revealed through subsequent exploratory factor analyses. This structure comprises an overarching CU dimension and three subfactors: callous, uncaring, and unemotional (Cardinale & Marsh, 2020). However, Tomlinson et al. (2022) discuss that most work analyzing the factor structure focused on adolescence (e.g., Essau et al., 2006; Pihet et al., 2015) or early childhood (e.g., Longman et al., 2016; Zumbach et al., 2021) rather than middle childhood (e.g., Gao & Zhang, 2016; Houghton et al., 2013). In this regard, middle childhood emerges as an interesting life stage because antisocial behavior is relatively more stable and less prevalent here than in adolescence (Moffitt, 2018; Tomlinson et al., 2022).

In the literature, various models of the factor structures of callous-unemotional traits and their associations with developmental outcomes in middle childhood are proposed. Willoughby et al. (2015) identified a two-factor model using the parent-report version of the ICU with first-grade children. This model differentiated empathic-prosocial (EP) behaviors from callous-unemotional (CU) behaviors and demonstrated their associations with oppositional defiant disorder and conduct disorder. Houghton et al. (2013) examined the ICU in children (self-report) aged 7.6 to 12.8 years and found a two-factor model with Uncaring and Callous Unemotional factors. This model was invariant across gender and age, with older children showing higher scores on the Uncaring dimension. Gao and Zhang (2016) explored the factor structure of the ICU in a community sample of 8- to 10-year-old boys and girls. They found a two-factor structure with Uncaring and Callousness factors for the self-report version. In contrast, the parent-report version supported a three-factor structure. CU traits were positively associated with conduct disorder and internalizing behavior symptoms. Ciucci et al. (2014) identified a three-factor model for CU traits (self-report) in children in grades 6 and 8, including callousness, uncaring, and unemotional subfactors. These factors contributed to an overarching general factor. CU traits correlated positively with school behavior problems, bullying, and reactive aggression, mainly influenced by the callousness and uncaring dimensions. It becomes clear that there are inconsistencies in the results of the factor structures. These inconsistencies highlight the need for further investigation as the factor structure influences the decision on which intervention or prevention strategies can be used. Previous research has shown that children with high CU traits respond less positively to standard interventions, especially to punishment (Frick & Ray, 2015; Hawes et al., 2014b). Thus, interventions that are tailored to the unique characteristics of children with CU traits are needed (Frick et al., 2014). To develop these interventions, an understanding of the structure of CU traits is needed, as different values in the underlying scales could further influence the responsivity to intervention strategies. Waller et al. (2015) identified a three-factor bifactor model and a revised two-factor model using parent- and teacher-reported ICU in high-risk 9-year-olds. The general CU traits and specific uncaring factor scores within the three-factor bifactor framework were associated with lower internalizing and higher externalizing behavior problems. The study emphasizes how important it is to refine the items. The authors also discuss the unemotional items, endorsing the suggestion made by Hawes et al. (2014a) to trim the unemotional items from the conceptualization of CU traits and concentrate on callous and uncaring traits. Additionally, some short versions were tested, such as Wang et al. (2020), who analyzed the factor structure of five short ICU versions and measurement invariance across self, teacher, and parent reports in primary school children (grades 1 to 6) in China. The shortened version of 11 items (ICU-11) assessing the callousness and uncaring factors had a good overall fit. Moreover, the ICU-11 demonstrated invariance across different informants.

Payot et al. (2022) examined the factor structure of the ICU parent version in a sample of children aged 3 to 9 years from Belgium. They identified a second-order model (18 Items) consisting of three first-order factors (unconcern of performance, lacking conscience, and lacking emotional expression), a second-order factor (general CU traits), and a methodological factor (items that were negatively worded), discussing this negatively worded factor as a methodological artifact. Additionally, results show measurement invariance for gender.

The literature reveals varying factor structures proposed for middle childhood, necessitating a more unified understanding of CU traits in this developmental stage (Ciucci et al., 2014; Gao & Zhang, 2016; Hawes et al., 2014b; Houghton et al., 2013; Payot et al., 2022; Waller et al., 2015; Wang et al., 2020; Willoughby et al., 2015).

Morality in middle childhood and associations with callous-unemotional trait

In addition to the mentioned developmental outcomes in middle childhood, such as conduct disorder or internalizing problems (Gao & Zhang, 2016; Willoughby et al., 2015), callous-unemotional traits are also associated with impairments in moral development (Craig et al., 2021). Most analyses of associations between callous-unemotional traits and moral development were conducted in adolescence (Schipper & Koglin, 2021a; Vasconcelos et al., 2021) rather than in middle childhood (Thornberg & Jungert, 2017; Kingsford et al., 2018). Frick et al. (2014) suggested that callous-unemotional traits could be seen as a misdevelopment of conscience. The moral self refers to an individual’s internalized sense of right and wrong, which guides their moral decisions and behaviors (Blasi, 2004; Kochanska, 2002). Literature suggests that children’s moral self is an early form or precursor of adolescent’s moral identity (Hardy & Carlo, 2011; Kingsford et al., 2018; Krettenauer & Hertz, 2015; Lapsley, 2015). Blasi’s (1983) Self Model of Moral Functioning highlights the significance of the moral self for social-emotional development. The moral self guides children in fulfilling personal moral expectations, maintaining self-standards, and navigating social contexts (Aquino & Reed, 2002). In early childhood, children adhere rigidly to rules, learning social and moral norms (Smetana, 2006; Turiel, 1983, 1998), while in middle childhood, moral comprehension advances, conscience develops, and an awareness of the flexibility of morality emerges (Smetana, 2013). The moral self undergoes continuous development in middle childhood, influenced by individual differences and personal experiences (Hardy & Carlo, 2011). In addition, social and emotional skills also continue to develop in middle childhood, partly because children grapple with their morality or with moral conflicts and situations. This can have an impact on the development of their empathy and perspective-taking skills (Carr, 2017), which in turn contribute to overall social adjustment (Wilson et al., 2016). In summary, middle childhood is a crucial period for moral and social-emotional development (Carr, 2017; Smetana, 2013). Therefore, it can be presumed that children with high callous-unemotional traits might not perceive certain immoral decisions as wrong (Vasconcelos et al., 2021), as the internalization of moral values into the self is inhibited due to the lack of remorse (Cardinale & Marsh, 2020). CU traits have already been identified as predictors of the moral identity of adolescents (Schipper & Koglin, 2021a). Nevertheless, measuring morality and specifically the moral self or identity poses challenges for researchers due to the complexity of the constructs and is increasingly being discussed (Schütz & Koglin, 2023; Tissot et al., 2023). A differentiated perspective on morality is often advocated (Schütz & Koglin, 2023; Oser, 2013). Research on the moral self or moral identity frequently employs self-evaluation measures of moral characteristics or traits (e.g., fairness). This method has been effective in several studies (e.g., Aquino & Reed, 2002; Bäker, 2023; Barriga et al., 2001; Patrick & Gibbs, 2012; Schipper & Koglin, 2021b; Schütz & Bäker, 2023). Nevertheless, there is also a need for psychometrically validated instruments to measure the moral self of children in middle childhood.

Addressing the identified gaps in the literature concerning CU traits in middle childhood, understanding inconsistencies in factor structures, and developing a nuanced perspective on the association between CU traits and moral development is crucial for the advancement of effective interventions and prevention strategies.

Current study

The analysis of the psychometric properties of the ICU in middle childhood has increasingly been focused on (Gao & Zhang, 2016; Hawes et al., 2014b; Houghton et al., 2013; Willoughby et al., 2015). The studies have examined different models in different samples (Gao & Zhang, 2016; Hawes et al., 2014b; Houghton et al., 2013; Willoughby et al., 2015). However, no study has examined the factor structure in a German sample in middle childhood. Therefore, the study seeks to assess the psychometric structure of the ICU in a community sample of middle childhood participants from Germany. Ueno et al. (2021) investigated the psychometric structure of the ICU for students between the ages of six and 18, therefore including middle childhood. However, they did not differentiate the age groups but rather focused on differences between the parent, teacher, and self-report of the ICU (Ueno et al., 2021). Children under 13 did not use the self-report, and parents also answered the ICU for older students; thus, the results of Ueno et al. (2021) cannot be compared to the analysis proposed in this study.

The aim of the study is to add to previous research and replicate findings of studies investigating the properties of the ICU in German samples (e.g., Essau et al., 2006; Zumbach et al., 2021) and of international studies focusing on children in middle childhood (e.g., Hawes et al., 2014b; Houghton et al., 2013; Willoughby et al., 2015). Children with CU traits may deviate from normal developmental pathways, affecting the internalization of moral judgments, reasoning and reducing guilt after transgressions (Frick et al., 2014; Malti & Krettenauer, 2013; Nunner-Winkler, 2007; Thornberg & Jungert, 2017; Vasconcelos et al., 2021). Social-emotional processing deficiencies linked to CU traits could impair children’s moral development (Craig et al., 2021). Schipper and Koglin (2021a) identified that callousness and uncaring dimension are associated with moral identity in adolescents. In elementary school children, callousness and unemotionality negatively impact moral reasoning. There is a research gap on associations between CU traits and children’s moral self. Therefore, additional path analyses are to be conducted between the identified factors and the moral self. We assume that the identified factors are negatively associated with the moral self.

Potential gender differences in callous-unemotional traits (Essau et al., 2006; Houghton et al., 2013; Northam & Dadds, 2020; Wilke & Goagoses, 2023) and morality (Gilligan, 1982; Jaffee & Hyde, 2000; Kochanska, 2002; Molchanov, 2013; White, 1999) are discussed in different studies. Inconsistent findings emerge regarding the presence of gender-specific differences in CU traits among children. For instance, some indicate that CU traits may vary depending on the child’s gender (e.g., Essau et al., 2006; Raschle et al., 2018), with boys exhibiting significantly more CU traits than girls (Essau et al., 2006). Conversely, other studies argue that there are no significant gender-specific differences in children (Houghton et al., 2013; Wilke & Goagoses, 2023). Although some inconsistencies can be identified, we included gender in our analysis to enhance the understanding of the nuanced relationship between gender, callous-unemotional traits, and morality.

Method

Participants

The data presented is derived from a larger study that the authors conducted that examined children’s and adolescents’ social-emotional development. Only instruments and data relevant to the stated research question are reported. A total of N = 824 children participated in the study. Age ranges from 7 to 12 (M = 10.34, SD = 1.65), with 430 boys (52.18%). The sample’s gender distribution is balanced. The Commission for Research Assessment and Ethics approved the study, and the authors have declared no conflicts of interest. The study was additionally approved by the relevant State Education Authority, and the Data Protection Authority of the university where it was conducted examined and approved the study. Schools were first enlisted. Organizational measures, such as sending out general participant information and obtaining consent forms for the implementation of data collection by the school administrators, were carried out in the case of interest to participate. The selection of schools requested to participate was done randomly. Participation was voluntary, and informed consent was required from both legal guardians and children. If all consent forms were not obtained, students were excluded from the study. This was due to data privacy protection reasons. Therefore, we were also unable to determine which students did not participate and why they did not participate. Additionally, for students from single parent families, the consent of the legal guardian was enough. The students received no compensation for their participation.

Instrument

To assess callous-unemotional traits in middle childhood, the German self-report version of the Inventory of Callous-Unemotional Traits (Essau et al., 2006; Frick, 2004) was utilized. Children respond to 24 items on a scale ranging from (1) not at all true to (4) definitely true. The questionnaire includes three subscales: callousness (e.g., The feelings of others are unimportant to me), unemotional (e.g., I hide my feelings from others), and uncaring (e.g., I always try my best; require reverse scoring; Essau et al., 2006). Sum scores for the overall CU measure were calculated, with higher scores indicating higher CU-traits levels.

The moral self of the children was assessed using a questionnaire adapted for children based on Koglin’s (2017) work, which measures moral identity in adolescence. The adapted version for children is adjusted by including standardized definitions to ensure comprehension of the specified items (e.g., Item 10: being fair, definition: Being fair means being just and following rules). Children with high scores on the scale increasingly perceive moral characteristics as meaningful to themselves. Previous studies demonstrate satisfactory reliability (internal consistency) of the questionnaire in adolescents (e.g., α =.79 see Schütz-Wilke et al., 2023, and e.g., α =.71 to.78 Bäker, 2023) and middle childhood (e.g., α =.71 see Schütz & Bäker, 2023 and α =.77 see Wilke et al., 2024). Furthermore, Bäker (2023) assessed the stability of moral identity in adolescents over one year (ß =.36, p <.001) and conducted an exploratory factor analysis of the moral identity scale, revealing satisfactory factor loadings on one factor (ranging from.34 to.78). Both questionnaires were designed with reference to the Self-Importance of Moral Identity Measure (Aquino & Reed, 2002) and the adapted version of the Good Self-Assessment (Barriga et al., 2001). Additionally, both questionnaires (measuring moral identity in adolescents or moral self of children) consisted of 18 items: nine moral characteristics (e.g., being selfless or fair) and nine non-moral characteristics (e.g., being funny or intelligent). The nine non-moral items function as distractors and are not taken into account in the scale formation. Children rated these characteristics on a four-point rating scale ranging from (1) not at all important to me to (4) very important to me.

Data analytic strategy

Confirmatory factor analyses

Our study’s aim is to analyze the factor structure of the ICU (Frick, 2004) for a German sample of children. Therefore, we test different models based on the literature on the ICU. (a) Model 1 is an undifferentiated one-factor model with a universal CU-factor. (b) Model 2 is conducted as a two-factor model, including a callousness and an uncaring factor, with a total of 12 items following the procedure of Hawes et al. (2014b). (c) Model 3 is a two-factor model with the factors callous-unemotional and empathic-prosocial derived from Willoughby et al. (2015). (d) Model 4 represents a two-factor model (callousness and uncaring), including 16 items (Houghton et al., 2013). (e) Model 5 describes a three-factor model with callousness, uncaring, and unemotional according to Ezpeleta et al. (2013) and Frick (2004). (f) Model 6 follows Gao and Zang (2016) and is conducted as a two-factor model (callousness and uncaring) without the unemotional factor. (g) Model 7 is conducted as a three-factor-higher-order model (Ciucci et al., 2014), and (h) model 8 is a three-factor-bi-factor model (Fanti et al., 2009; Kimonis et al., 2008; Roose et al., 2010). A description of the models, the items in- and excluded from the analysis can be found in Supplementary Table 2.

The models’ quality is assessed using the Comparative Fit Index (CFI), and the Root Mean Square Error of Approximation (RMSEA). A good model fit is defined as having RMSEA values <.08 and CFI values >.90 (Hu & Bentler, 1999). Additionally, Chi2 values are presented; however, it is important to take into account the sensitivity of Chi2 for larger samples (N > 200; see Vandenberg, 2006). Due to missing values, the Standardized Root Mean Squared Residual (SRMR) cannot be calculated. Data with entirely missing values are omitted from the analysis (N = 144). The students started filling out the questionnaire but choose not to answer certain scales thus they were excluded from the analysis and imputation. Using Full Information Maximum Likelihood (FIML; Enders & Bandalos, 2001) individual missing values for single items are estimated as part of the model estimation (10% of the remaining children miss individual items). In the presence of missing data, FIML maximizes data efficiency, minimizes biases, and ensures consistent parameter estimates making it well-suited for complex models (Graham, 2009). P-values are FDR-corrected (false discovery rate; Benjamini & Hochberg, 1995) in order to account for the probability of alpha-errors (Cribbie, 2007).

Path model

After identifying the best fitting model for the present data, a path model is used to calculate the relations between CU traits and the moral self. In this context, a factor-analytical examination of the morality items is included. The estimation procedure and description of model fit is used as described above. All analyses are conducted using STATA 18.

Results

Prior to confirmatory testing, descriptive statistics and intercorrelations of the ICU items were calculated and presented in Tables 1 and S1 in the supplementary material.

Table 1 Descriptive statistics of the ICU items
Full size table

Confirmatory factor analyses

In the first step, the results of the confirmatory factor analysis, to examine the factor structure of the ICU, are presented. We tested seven alternate CFA models based on previous research studies. The model fit indices for all models tested are summarized in Table 2. The results indicate that none of the confirmatory models show a good model fit with RMSEA values <.08 and CFI values >.90. Two of the models (three-factor-higher-order model and three-factor-bi-factor model) did not find a fit for the present sample.

Table 2 Model-fit indices of CFA models of the ICU
Full size table

Modified confirmatory testing (post-hoc analyses)

Since none of the tested models showed a good model fit, post-hoc analyses were performed in the next step by calculating modified CFA models. For this purpose, the item-total correlations across the ICU items are considered in order to identify poorly discriminating items for each model. This procedure is chosen following the approach of Gao and Zhang (2016). Items with item-total correlations <.3 are excluded for the following repeated confirmatory testing. For model 5, 6, 7 and 8 four items of the callousness factor are deleted: Item 10 (r =.228), item 18 (r =.236). item 20 (r =.287), and item 21 (r =.243). Model 7 and 8 continue to fail to achieve convergence, while model 5 (∆RMSEA = −.006, ∆CFI =.011) and model 6 (∆RMSEA =.006, ∆CFI =.013) slightly improve, at least the CFI.

Since models 2, 3, and 4 have reassigned the items to different factors, the item-total correlations are looked at individually here. For model 2 and 4 all item-total correlations are above.3, so no changes are made based on item-total correlations. For model 3 three items of the callous-unemotional factor are deleted: Item 21 (r =.187), item 6 (r =.140), and item 22 (r =.111), and six items of the empathic-prosocial factor: Item 14 (r =.074), item 1 (r =.154). item 13 (r =.174). item 19 (r =.119), item 3 (r =.183), and item 10 (r =.139). Afterward, the model fit improves, but does not yet reach desired values (∆RMSEA = −.008, ∆CFI =.154). Even with the elimination of the poorly discriminating items, none of the models yield a good model fit. Therefore, we choose the best unmodified model for the further procedure, which is the two-factor model according to Hawes et al. (2014b).

Similar to the procedure in the study by Gao and Zhang (2016), we correlate single error terms for the following four items “I do not care who I hurt to get what I want, item 4” and “I apologize (“say I am sorry”) to persons I hurt, recoded, item 16”, “I do not feel remorseful when I do something wrong, item 18” and “I feel bad or guilty when I do something wrong, recoded, item 5”, “The feelings of others are unimportant to me, item 21” and “I am concerned about the feelings of others, recoded, item 8”, and “The feelings of others are unimportant to me, item 21” and “I try not to hurt others’ feelings, recoded, item 17”. The model reaches a good model fit (RMSEA =.051, CFI =.941) and factor loadings and Cronbach’s Alpha values are presented in Table 3.

Table 3 Factor loadings and reliability of the two-factor model according to Hawes et al. (2014b)
Full size table

For the following analyses, we have therefore retained the most appropriate model, namely Model 2 according to Hawes et al. (2014b) with a callousness and an uncaring factor (see Table 3). The path model shows a stronger overarching association between the uncaring factor and moral self than with callousness. The factor structure of moral self was also tested as part of the path model. The moral self-scale shows good reliability (Cronbach’s Alpha =.787). The overall path model shows good model fit (see Fig. 1). The model explains R2 = 52.09% of the variance of morality. In the next step, the relations between the factors callousness and uncaring and the moral self are reported separately for the total sample as well as by gender (see Table 4). For boys, the model explains 56.70% of the variance of morality and for girls R2 = 34.96% is reported.

Fig. 1
figure 1

Path model to analyze the relation between CU traits and moral self. ICU Model according to Hawes et al. (2014b); M = Items of moral self; Confidence intervals in brackets; *p < .05; **p < .01; ***p < .001; χ2(182) = 465.08, p < .05; RMSEA = .048; CFI = .912

Full size image
Table 4 Relations between callousness and uncaring and moral self separately for boys and girls
Full size table

Discussion

The results of this study provide valuable insights into the psychometric structure of the Inventory of Callous-Unemotional Traits in a German community sample of middle childhood participants and its associations to morality. Confirmatory factor analyses were conducted to evaluate different models, and it was found that none of the tested models showed a good fit according to established criteria. However, post-hoc analyses revealed that the best-fitting model was a two-factor model, comprising callousness and uncaring factors, which aligns with the findings by Hawes et al. (2014b) in a similar age group and Zumbach et al. (2021) in a sample of preschoolers. This similarity suggests that the underlying structure of callous-unemotional traits might remain consistent during the transition from kindergarten to middle childhood. Past research predominantly favored the three-factor-bi-factor model in older children and adolescent samples (Cardinale & Marsh, 2020; Fanti et al., 2009; Kimonis et al., 2008; Roose et al., 2010), implying that established models can’t be directly applied to diverse age groups. Particularly in childhood, accurately gauging CU traits becomes challenging without considering other developmental factors (Zumbach et al., 2021).

The dominance of callousness and uncaring factors in the final model of this study implies that these two dimensions might be the core components that differentiate callous-unemotional traits in middle childhood. It can be hypothesized that these two dimensions play a more substantial role in shaping callous-unemotional traits and psychopathic tendencies during this developmental stage. Moreover, the exclusion of the unemotional items from the final model suggests that these traits might be more closely related to the broader social-emotional development at this age rather than specifically characterizing callous-unemotional traits (Kahn et al., 2012; Zumbach et al., 2021). Additionally, the exclusion of the unemotional items from the final model is in line with literature highlighting methodological shortcomings in the unemotionality dimension (Cardinale & Marsh, 2020).

It is noteworthy that in the final two-factor model, the uncaring factor was completely reverse-scored, while the callousness factor was not. This arrangement, where one factor comprises positively-valenced items and the other negatively-valenced items, may potentially result in multiple factors. A number of studies questioned that the two factors are just the result of a methodological artifact (e.g., Cardinale & Marsh, 2020; Payot et al., 2022; Ray et al., 2016). To investigate whether the negative formulation of the items contributes to these two factors, further analysis is necessary. Future studies should explore alternative item formulations to avoid reverse coding. Moreover, formulations involving negations, commonly used in the callousness scale (e.g., “I do not care about doing things well”), may pose comprehension challenges for children, potentially leading to misunderstandings. Additionally, observed disparity in scoring between the uncaring and callousness factors raises the possibility that the measure’s current structure may not accurately capture the intended constructs. This signals that, rather than a genuine representation of distinct underlying factors, the observed factors may be influenced by the valence of the items. This raises questions about the overall validity of the measure. Consequently, consideration should be given to the exploration of alternative measurement approaches.

Additionally, middle childhood is characterized by significant changes in social cognition, empathy, and moral development (Malti et al., 2018; Raffaelli et al., 2005), and the emergence of callous-unemotional traits may reflect a deviation in the typical development of moral understanding and emotional processing. The literature reveals discrepancies in defining CU traits using the ICU among child and adolescent samples (Gao & Zhang, 2016; Hawes et al., 2014b; Houghton et al., 2013; Willoughby et al., 2015). Many studies do not consider the influence of children’s social-emotional (Zumbach et al., 2021) and moral development on their analyses. Neglecting these aspects, particularly during middle childhood, constrains the depth of result interpretation. Thus, we examined the associations between moral self and CU traits. In line with results of Schipper and Koglin (2021a) this study indicates that uncaring shows greater associations with the moral self than callousness. Children who value the concerns and feelings of others regard being a moral person as crucial (Schipper & Koglin, 2021a). Accordingly, the moral self can be highlighted as a possible construct that is impaired in development to the detriment of children and, in turn, is associated with behavioral problems (Gini et al., 2021). Additionally, the highest association between uncaring and the moral self in this study is reported for the boys. Gender differences in CU traits are increasingly being discussed in studies (Essau et al., 2006; Houghton et al., 2013; Northam & Dadds, 2020; Wilke & Goagoses, 2023). Nevertheless, further studies are needed to make specific statements about gender differences and effects.

Understanding the distinct features of callousness and uncaring traits in middle childhood is crucial for identifying potential risk factors and developing targeted interventions. Early identification and intervention in this critical period might be beneficial for preventing the escalation of callous-unemotional traits and associated psychopathic tendencies. Frick and Ray (2015); Hawes et al. (2014b) were able to show that children with high CU traits do not respond to standard interventions the same way as children with low CU traits. Children with high CU respond negatively to punishment (Frick & Ray, 2015; Hawes et al., 2014b). Thus, interventions that are tailored to the unique characteristics of children with CU traits are needed (Frick et al., 2014). These strategies can be employed by teachers in schools (especially with a focus on positive reinforcement and not punishment) additionally, intervention strategies for psychological settings are needed that focus on the particular needs of children with high CU traits.

Limitations and further research

While this study provides valuable insights, it is essential to acknowledge some limitations. Firstly, the study relied only on self-report measures, which might be subject to bias or social desirability effects. Moreover, other studies have already identified different factor structures in self-report and parent-report within the same sample (e.g., Gao & Zhang, 2016). Future research could incorporate multiple reports, such as self-, parent-, and teacher-reports, to enhance the validity of the findings. Secondly, the study focused on a German community sample, and cross-cultural comparisons could provide a more comprehensive understanding of the universality of the identified factor structure. Another limitation pertains to the examination of cross-sectional data. During childhood, numerous behavioral aspects undergo continuous developmental changes. To ensure valuable and dependable findings, studies should account for this progression in their analysis. It’s crucial not only to evaluate CU variables across multiple measurement points but also to track the evolution of potential outcomes like morality or aggressive behavior. It should also be noted that the sample is an ad hoc sample of German participants. Some of the children have a different racial/ethnic background than white Germans, but unfortunately, we do not have consistent information on this for the sample. Other studies have already shown that cultural aspects can have an impact on CU traits (Deng et al., 2024) and moral development (Malti & Keller, 2010). Racial and ethnic aspects should be considered in greater depth in further research.

It is noteworthy to highlight a critical aspect related to the name of the construct known as callous-unemotional traits. Since the best fit is the two-factor model with the “callous-uncaring traits” there seems to be inconsistency in terminology. The question arises about the appropriateness of labeling the instrument or the construct itself as “callous-unemotional traits”. This incongruence prompts a reevaluation of the alignment between the conceptualization of the construct and its operationalization. As suggested by Waller et al. (2015) and Hawes et al. (2014a), it could be helpful to trim the unemotional items from the conceptualization of CU traits and concentrate on callous and uncaring traits. Additionally, some factor loadings emerge as low (ranging from.307 to.686). However, according to Stevens’ (2002) criteria for items in factor analysis, the loadings for this sample size are satisfactory. Although the ICU is frequently used in research (e.g., Frick, 2004; Gao & Zhang, 2016; Houghton et al., 2013), these limitations should be taken into account when utilizing it with children.

Conclusion

In conclusion, this study delves into the psychometric structure of the Inventory of Callous-Unemotional Traits in middle childhood, revealing a two-factor model with callousness and uncaring factors. Notably, the reverse-scoring of the uncaring factor and the potential influence of item valence call for further exploration and alternative item formulations. Associations between CU traits and the moral self highlight the relevance of moral development. Neglecting developmental factors limits result interpretation, emphasizing the need for a comprehensive approach. Recognizing distinct features in middle childhood is vital for measuring and for targeted interventions of callous-unemotional traits and associated psychopathic tendencies. This study contributes important insights to understanding and addressing these traits during this critical developmental period.