Conduct problems encompass a range of disruptive childhood behaviours, including aggression, rule-breaking, oppositionality, and irritability. In their most severe form, conduct problems may lead to a diagnosis of conduct disorder, which is associated with lasting personal, social, and economic costs (Fairchild et al., 2019; Staginnus et al., 2026).
One of the greatest challenges facing the field is the heterogeneity of conduct problems and conduct disorder (CD). For example, there are over 32,000 possible symptom profiles for CD that differ in developmental trajectories, comorbidities, and aetiologies, complicating both research and treatment (Fairchild et al., 2019; Staginnus et al., 2023). One important source of this variation is callous-unemotional (CU) traits, which are present in a subset of individuals with conduct problems or CD, characterised by reduced empathy, guilt, and prosocial behaviour, and associated with a more severe presentation. Gender is also a factor, as symptoms of conduct problems can present differently in boys and girls (Konrad et al., 2022).
But how should researchers make sense of this complexity?
Traditionally, psychiatric disorders have often been understood using latent variable models, where an underlying disorder is assumed to produce observable symptoms. Network analysis offers a different perspective. Rather than treating symptoms simply as consequences of an underlying disorder, it examines how individual symptoms are connected to one another (Borsboom & Cramer, 2013). Researchers can then identify central symptoms, which are strongly connected within a network, and bridge symptoms, which connect different groups of symptoms.
Considering the heterogeneity of conduct problems, the condition represents an ideal testing ground for the network approach. In their paper, Álvarez-Voces and colleagues (2026) apply network analysis to four samples of children from the United States and Spain, examining how conduct problems and CU traits are organised and whether their symptom networks differ between boys and girls and across countries.
Methods
The authors analysed data from four large population-based samples (11,874, 2,426, 1,342 and 1,259 children) aged 8 to 14 from the United States (Adolescent Brain Cognitive Development (ABCD) study and ABCD Social Development substudy) and Spain (ELISA project, 2023 and 2024). They first compared levels of conduct problems and callous-unemotional (CU) traits between boys and girls, as well as the proportion of children screening as at risk for conduct disorder across countries and between boys and girls.
To map how individual symptoms related to one another, the authors created networks in which each symptom was represented as a point, with lines connecting symptoms that were related after accounting for the other symptoms in the network. They then examined which symptoms were most strongly connected to others, which symptoms linked conduct problems and CU traits, and whether these patterns differed between boys and girls and across countries.
Results
Boys had higher levels of conduct problems than girls across both countries, although the differences were small (d = 0.20 to 0.31). Rates of conduct problem risk were largely similar across boys and girls and between countries, except that boys aged 11 and younger had higher conduct problem risk than girls in the main US Adolescent Brain Cognitive Development sample.
Boys also consistently had higher callous-unemotional (CU) traits than girls across both countries, although differences were modest (d = 0.17 to 0.34). There was little evidence of meaningful differences in CU traits between the US and Spain.
Network analysis
- CU traits were more strongly connected to each other than conduct problem symptoms across samples, suggesting that CU traits formed a more tightly connected cluster.
- Disobedience and deceitfulness, including rule-breaking and lying/cheating, emerged as important and highly connected conduct problem symptoms across the networks, although the most central individual conduct problem symptoms varied between samples.
- Overall, network patterns were broadly similar between boys and girls and across countries. However, some differences emerged. Relational aggression was more central among girls, while property destruction and theft were more central in the US.
- Lack of guilt emerged as an important link between CU traits and conduct problems. When the same brief CU measure was used, lack of guilt was consistently the strongest bridge symptom, meaning it was particularly strongly connected across the conduct problem and CU clusters. Different bridge symptoms emerged when alternative measures were used.
- Findings were broadly consistent across different measures. However, the more comprehensive Inventory of Callous-Unemotional Traits captured additional features, with uncaring traits emerging as relatively central.
Conclusion
Overall, while the broad organisation of conduct problem and callous-unemotional trait symptom networks was similar across boys and girls, countries, and measurement instruments, there were differences in which individual symptoms were most central, further highlighting the heterogeneous nature of conduct problems. The authors conclude that network analysis offers a different way of understanding this heterogeneity and, by identifying central and bridge symptoms, may eventually help identify more precise targets for personalised interventions.
Strengths and limitations
Strengths
- Large, population-based samples: Four population-based US and Spanish samples provided high statistical power and allowed the authors to explore differences across gender and countries.
- Methodological rigour: Methodological thoroughness is a key strength of this study. The authors compared two different measures of callous-unemotional (CU) traits within Sample 2 to distinguish findings related to the underlying construct from those specific to the questionnaire used, an approach they note they were the first to adopt (Álvarez-Voces et al., 2026). They also removed an item that overlapped between the conduct problem and CU measures, and used partial correlations and graphical LASSO (least absolute shrinkage and selection operator), which, in simple terms, allowed them to examine the relationship between two symptoms while accounting for all other symptoms in the network, while filtering out weaker or potentially spurious connections (McNally, 2021). Finally, they assessed several network metrics and tested the stability of their findings, rather than relying on visual interpretation alone.
- A novel conceptual framework: Network analysis moves beyond viewing conduct problems as one broad construct and instead examines how individual symptoms relate to each other. Identifying highly connected symptoms and those linking conduct problems and CU traits generates hypotheses about symptoms that might eventually become personalised intervention targets.
Limitations
- Sample characteristics and generalisability: The study examined children from the general population rather than those diagnosed with conduct disorder (CD), meaning the same symptom patterns may not emerge in clinical or forensic populations who have more severe conduct problem profiles. The samples were also predominantly white, limiting generalisability, and relying on parent reports may introduce reporting bias and inflate relationships between symptoms (De Los Reyes & Kazdin, 2005). The authors also identified children at risk for CD using screening cut-offs rather than clinical assessments, so these should not be interpreted as diagnosed cases.
- Interpreting network findings: Due to the correlational nature of this study, it cannot be concluded that highly connected symptoms are necessarily causing other symptoms (Bringmann & Eronen, 2018). While the authors discuss central and bridge symptoms as potential targets for personalised interventions, these implications should be interpreted with greater caution. Some networks also approached the authors’ minimum acceptable threshold for stability, adding further caution when interpreting which individual symptoms appeared most central. The findings are better viewed as generating hypotheses for future longitudinal and intervention studies than as identifying causal mechanisms underlying conduct problems.
- Conceptual clarity: A further consideration concerns the rationale for combining conduct problem symptoms and CU traits within a single network. The network approach was originally proposed to conceptualise disorders as systems of interacting symptoms, whereas CU traits form the basis of a DSM-5 specifier rather than symptoms of CD themselves (American Psychiatric Association, 2022; Borsboom, 2017). Examining these constructs together is entirely appropriate, particularly for identifying links between them, but it addresses a somewhat different conceptual question from mapping relationships among CD symptoms themselves. This distinction is important when interpreting potential intervention targets.
- Measurement and bridge findings: Finally, although lack of guilt consistently emerged as the strongest bridge when the same brief CU measure was used, the strongest bridge symptoms differed when alternative measures were used, sometimes emerging from conduct problems and other times from CU traits. This may reflect differences in age, country, measurement or sampling, making the generalisability of these potential targets less clear.
Implications for practice
This study has several valuable implications for both research and clinical practice. First, it provides insight into how conduct problem symptoms and callous-unemotional (CU) traits are organised across different groups, countries, and measurement instruments. In doing so, it further dissects the well-established heterogeneity of conduct problems and highlights that the way these behaviours and traits manifest may not be identical across all children. This reinforces the growing appreciation that a “one-size-fits-all” conceptualisation of conduct problems is unlikely to be sufficient and adds support to efforts to understand heterogeneity at the symptom level.
Perhaps more broadly, the authors’ use of the network approach aligns with the current climate in psychopathology research. The field is increasingly embracing frameworks that seek to move beyond rigid diagnostic categories towards more nuanced conceptualisations of psychopathology. In this context, the present study represents one of relatively few attempts to dip its toe into the pond of network psychopathology for conduct problems and CU traits, contributing to a broader movement seeking to understand childhood psychopathology at the symptom level.
That said, enthusiasm for new approaches should be accompanied by careful scientific rigour. As promising as network analysis may be, we should remain clear about the questions our studies are answering and avoid extending conclusions beyond what the methods can support. Cross-sectional symptom networks generate valuable hypotheses, but they do not establish causal mechanisms or identify definitive treatment targets. Rather, they identify statistically important symptoms whose clinical utility requires validation in longitudinal and intervention studies. As the field navigates these novel conceptual frameworks, it is perhaps not unwise to refer to nautical wisdom: “Steer by the stars, but watch for the shoals.”
Statement of interests
Mona Nour Eddine has no conflicts of interest to declare.
Edited by
Dr Nina Higson-Sweeney
Links
Primary paper
María Álvarez-Voces, Natalie Goulter, Yael Paz, Beatriz Díaz-Vázquez, Laura López-Romero, Paula Villar, Amy L. Byrd, Samuel W. Hawes, Estrella Romero, & Rebecca Waller (2026). Applying a Network Approach To Characterize Gender Differences in Conduct Problems and Callous-Unemotional Traits among Children from Two Countries. Research on Child and Adolescent Psychopathology, 54(1), 10. https://doi.org/10.1007/s10802-025-01399-7
Other references
American Psychiatric Association. (2022). Diagnostic and statistical manual of mental disorders (5th ed., text rev.; DSM-5-TR). American Psychiatric Association Publishing.
Borsboom, D. (2017). A network theory of mental disorders. World Psychiatry, 16(1), 5–13. https://doi.org/10.1002/wps.20375
Borsboom, D., & Cramer, A. O. J. (2013). Network analysis: An integrative approach to the structure of psychopathology. Annual Review of Clinical Psychology, 9(1), 91–121. https://doi.org/10.1146/annurev-clinpsy-050212-185608
Bringmann, L. F., & Eronen, M. I. (2018). Don’t blame the model: Reconsidering the network approach to psychopathology. Psychological Review, 125(4), 606–615. https://doi.org/10.1037/rev0000108
De Los Reyes, A., & Kazdin, A. E. (2005). Informant discrepancies in the assessment of childhood psychopathology: A critical review, theoretical framework, and recommendations for further study. Psychological Bulletin, 131(4), 483–509. https://doi.org/10.1037/0033-2909.131.4.483
Fairchild, G., Hawes, D. J., Frick, P. J., Copeland, W. E., Odgers, C. L., Franke, B., Freitag, C. M., & De Brito, S. A. (2019). Conduct disorder. Nature Reviews Disease Primers, 5(1), 43. https://doi.org/10.1038/s41572-019-0095-y
Konrad, K., Kohls, G., Baumann, S., Bernhard, A., Martinelli, A., Ackermann, K., Smaragdi, A., Gonzalez‐Madruga, K., Wells, A., Rogers, J. C., Pauli, R., Clanton, R., Baker, R., Kersten, L., Prätzlich, M., Oldenhof, H., Jansen, L., Kleeven, A., Bigorra, A., … Freitag, C. M. (2022). Sex differences in psychiatric comorbidity and clinical presentation in youths with conduct disorder. Journal of Child Psychology and Psychiatry, 63(2), 218–228. https://doi.org/10.1111/jcpp.13428
McNally, R. J. (2021). Network analysis of psychopathology: Controversies and challenges. Annual Review of Clinical Psychology, 17(1), 31–53. https://doi.org/10.1146/annurev-clinpsy-081219-092850
Staginnus, M., Cornwell, H., Toschi, N., Oosterling, M., Paradysz, M., Smaragdi, A., González-Madruga, K., Pauli, R., Rogers, J. C., Bernhard, A., Martinelli, A., Kohls, G., Raschle, N. M., Konrad, K., Stadler, C., Freitag, C. M., De Brito, S. A., & Fairchild, G. (2023). Testing the ecophenotype model: Cortical structure alterations in conduct disorder with versus without childhood maltreatment. Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, 8(6), 609–619. https://doi.org/10.1016/j.bpsc.2022.12.012
Staginnus, M., Gao, Y., Townend, S., Zugman, A., Zhang, J., Wu, Q., … & Fairchild, G. (2026). Dimensional Associations Between Conduct Problems and Brain Structure Across 18 International Cohorts in ENIGMA. Journal of the American Academy of Child & Adolescent Psychiatry. https://doi.org/10.1016/j.jaac.2026.03.002