Juvenile delinquency is a complex and heterogeneous phenomenon that cannot be adequately explained by a single etiological factor or developmental trajectory. Historically, delinquent youth have often been conceptualized as a relatively homogeneous group characterized by antisocial tendencies. However, even among adolescents receiving identical legal dispositions, the underlying psychological and biological mechanisms may differ substantially. Reflecting this heterogeneity, the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition introduced the with limited prosocial emotions specifier for conduct disorder based on callous-unemotional traits [1,2]. From a developmental perspective, Moffitt’s taxonomy further distinguished between life-course persistent and adolescence-limited patterns of antisocial behavior, proposing distinct developmental pathways [3]. Despite these advances, categorical classification systems remain limited in their ability to capture the substantial functional heterogeneity observed within diagnostically similar groups of delinquent youth.

In response to these limitations, recent research has increasingly adopted person-centered approaches, such as latent class analysis and latent profile analysis, to identify subgroups characterized by distinct configurations of risk factors and behavioral trajectories [4,5]. Unlike variable-centered approaches that focus on average effects, person-centered methods allow for the identification of qualitatively different subgroups within a population, thereby providing a more nuanced understanding of individual differences. In South Korea, several studies have applied psychological assessment tools such as the Special Personality Inventory-III and the Personality Assessment Inventory to classify delinquent adolescents or juvenile offenders based on psychosocial or personality profiles, demonstrating meaningful differences in antisocial tendencies and aggres-sion across subgroups [6,7]. While these findings underscore the utility of subtype classification for refined assessment and intervention planning, prior studies have largely focused on contemporaneous behavioral symptoms or personality profiles, with limited attention to temperament as a more fundamental and developmentally rooted substrate of delinquent behavior.

High rates of psychiatric morbidity among delinquent youth have been consistently reported across countries, further highlighting the need for a more integrative framework. In a study of adolescents in French detention facilities, Bronsard et al. [8] reported that 90.2% met criteria for at least one psychiatric disorder, with a substantial proportion of youths diagnosed with conduct disorder exhibiting additional comorbid conditions. Moreover, a meta-analysis by Fazel et al. [9] demonstrated that adolescents involved in the juvenile justice system were approximately ten times more likely to experience psychosis compared to their peers in the general population, suggesting that psychopathology among delinquent youth extends beyond externalizing problems to include severe mental disorders. Consistent findings have been reported in South Korea, where Kim et al. [10] found that 90.8% of male juvenile detainees met criteria for at least one psychiatric diagnosis, and 75.1% presented with two or more comorbid disorders. Importantly, such psychiatric vulnerabilities are not confined to adolescence. A 15-year longitudinal study by Teplin et al. [11] demonstrated substantial continuity of mental disorders from adolescence into adulthood among justice-involved youth, underscoring the importance of early identification and intervention targeting underlying psychological vulnerabilities.

From this perspective, temperament represents a core construct linking biologically based predispositions to behavioral expression and provides a valuable theoretical framework for understanding heterogeneity in delinquent behavior. According to Cloninger’s psychobiological model, temperament dimensions— namely novelty seeking (NS), harm avoidance (HA), and reward dependence (RD)—are relatively stable, heritable traits associated with impulsivity, emotional regulation, and interpersonal functioning [12]. Previous studies have shown that high NS and low HA are significantly associated with conduct disorder, aggression, externalizing behaviors, and substance use problems in adolescents [13-16]. However, the independent effects of single temperament dimensions are insufficient to explain the diverse clinical presentations observed in delinquent youth. Rather, combinations of temperament traits within individuals may organize distinct patterns of psychopathology and behavioral trajectories.

Despite the theoretical relevance of temperament-based profiling, person-centered studies integrating Cloninger’s temperament model with comprehensive measures of psychopathology remain scarce. In particular, little research has examined temperament-based subtypes in high-risk clinical populations, such as adolescents referred for court-ordered psychiatric evaluations. Moreover, no known studies in South Korea have integrated temperament profiles with the Minnesota Multiphasic Personality Inventory-Adolescent-Restructured Form (MMPI-A-RF), a multidimensional measure of adolescent psychopathology.

Accordingly, the present study aimed to identify temperament-based subtypes among Korean delinquent adolescents referred for court-ordered psychiatric evaluation using cluster analysis and to examine differences in MMPI-A-RF psychopathology profiles across the identified subgroups. By elucidating temperament-level heterogeneity within a high-risk delinquent population, this study seeks to provide empirical evidence to inform more precise assessment practices and individualized intervention strategies.

The present study sought to identify latent heterogeneity within a clinically referred population of juvenile offenders undergoing pre-trial forensic psychiatric evaluation, based on temperament-related characteristics, and to systematically examine the psychopathological profiles associated with each identified subtype. Cluster analysis yielded three distinct subgroups: the Relationship-dependent conformist group (Cluster 1), the isolated internalizing risk group (Cluster 2), and the impulsive externalizing risk group (Cluster 3).

Cluster 1 demonstrated overall psychopathological indices within the non-clinical range. Despite their involvement in delinquent behavior, individuals in this cluster showed relatively high levels of SD and CO, suggesting preserved psychological resources related to emotional regulation and social problem-solving. Notably, this group exhibited particularly high RD, indicating a temperament characterized by heightened sensitivity to interpersonal approval and relationship maintenance. While such social sensitivity may function adaptively under supportive conditions, it may also increase vulnerability to conformity when affiliation with delinquent peer groups is reinforced. This pattern is consistent with the concept of “social imitation” or adolescence-limited delinquency, in which antisocial behavior is primarily driven by peer context and environmental influences rather than enduring antisocial traits [3]. Accordingly, for this cluster, relationship-oriented temperament should not be viewed solely as a target for correction, but rather as a potential protective resource that can be leveraged when connected to appropriate social models. Interventions emphasizing mentoring, decision-making skills, and assertiveness training—alongside core emotion regulation modules—may help redirect social sensitivity toward adaptive interpersonal functioning, making punitive approaches less suitable for this subgroup.

Cluster 2 was characterized by a temperament profile dominated by behavioral inhibition, with pronounced HA and elevated INTR-r. This emotional profile suggests a core vulnerability centered on emotional distress and social withdrawal, rather than overt antisocial motivation. The tendency to internalize negative affect and to respond to interpersonal situations with avoidance and withdrawal indicates that delinquent behavior in this group may emerge as a maladaptive response to unmanaged emotional pain rather than as goal-directed aggression. Consistent with prior findings that social withdrawal and loneliness are associated not only with depressive affect but also with hostility and maladaptive behaviors [25,26], the delinquency observed in this cluster may reflect failures in emotion regulation within socially isolating contexts. While both high-risk clusters exhibited elevated RC6, its co-occurrence with high HA and social withdrawal in Cluster 2 suggests that their delinquency may stem from a defensive, reactive aggression rooted in a distorted perception, rather than proactive antisocial intent. Although NS remained at an average level, SD was relatively low (M=40.29), suggesting insufficient self-regulatory capacity to cope effectively with emotional distress. In particular, INTR-r functioned as the most salient indicator distinguishing this cluster, linking maladaptive behavior to emotional isolation, school maladjustment, and restricted social engagement. For this group, interventions focused primarily on suppressing problem behavior may be inadequate. Instead, integrated approaches that enhance basic emotional competencies, incorporate Cognitive-Behavioral Therapy to restructure cognitive distortions, and provide gradual, small-scale opportunities for safe interpersonal experiences are likely to be more appropriate [27].

Cluster 3 followed a prototypical externalizing pathway marked by strong behavioral activation and relatively weak behavioral inhibition. Temperamentally, this cluster was characterized by high NS and poor impulse control, resulting in a tendency to react immediately to external stimuli rather than engaging in reflective or deliberative processes. Consequently, delinquent behavior in this group may reflect a tendency toward direct behavioral discharge in response to situational triggers, rather than avoidance-driven or internalized distress. This is empirically supported by the significantly elevated scores on RC4 and BXD observed in this cluster. Additionally, this cluster exhibited low RD, indicating limited responsiveness to social cues, emotional feedback, or external reinforcement contingencies. As a result, insight-oriented or empathyfocused counseling alone may have limited effectiveness. Given these characteristics, intervention strategies for Cluster 3 may benefit from prioritizing structured environmental control, consistent behavioral regulation, and externally mediated contingencies over approaches focused solely on internal change. Beyond shared emotion regulation modules, clear behavioral rules, impulse control skills training, and predictable systems of reinforcement and consequence are critical. In this regard, multisystemic therapy, which simultaneously targets the home, school, and judicial contexts, may represent a particularly suitable intervention model aligned with the temperamental profile of this subgroup [28].

A key implication of the present findings is that the two high-risk clusters—Cluster 2 and Cluster 3—exhibited markedly different behavioral expressions yet shared substantial vulnerabilities in cognitive and emotional domains. Both clusters showed significantly elevated levels of psychopathology compared to Cluster 1, particularly in THD and RC6, highlighting a common core of psychological vulnerability. This pattern suggests that risk assessment based solely on the visibility or severity of externalizing behaviors may underestimate clinical severity, especially among youth whose difficulties manifest primarily through internalizing pathways. Although the two high-risk clusters shared common vulnerabilities in THD and persecutory ideation, they diverged clearly in socioemotional orientation, with Cluster 2 characterized by internalized withdrawal and Cluster 3 by impulsive externalization. INTR-r emerged as the most discriminative indicator, sharply distinguishing Cluster 2 from Cluster 3 and underscoring that similar levels of psychopathology may be expressed through either social withdrawal and emotional isolation or impulsive behavioral dysregulation, depending on underlying temperamental dispositions.

Taken together, these findings indicate that juvenile delinquency may not be fully captured by a simple internalizing- externalizing dichotomy. Instead, delinquent behavior may arise from shared core psychopathological vulnerabilities that are differentially shaped by temperament into distinct socioemotional and behavioral manifestations. Importantly, significant between-cluster differences in THD and RC6—dimensions not fully captured by the JTCI alone—underscore the incremental value of integrating the MMPI-A-RF into forensic evaluations. A multidimensional clustering approach that combines temperament and psychopathology thus offers a more precise, transdiagnostic framework for risk stratification and for prioritizing intervention targets based on clusterspecific vulnerabilities. While this study provides meaningful evidence of heterogeneity among juvenile offenders, several limitations should be noted.

First, the exclusive reliance on self-report measures may introduce response biases commonly observed in forensic contexts. Although validity scales were monitored, the absence of multi-informant data (e.g., from parents or teachers) remains a limitation for comprehensive clinical profiling. Second, the relatively small sample size (n=72) and recruitment from a single geographic region (Northern Gyeonggi Province) limit statistical power and generalizability. In particular, regional characteristics may be associated with differences in developmental contexts and access to psychosocial resources, which could not be systematically controlled in the present study. Nevertheless, the observed cluster stability (Jaccard coefficients >0.75) suggests acceptable internal reliability of the identified profiles within this clinical sample. Third, the cross-sectional design precludes causal inferences and limits conclusions regarding temporal development and long-term outcomes such as recidivism or progression to adult antisocial behavior. Longitudinal studies are needed to clarify the stability and predictive validity of these profiles. Fourth, due to sample size constraints within specific offense categories, systematic analysis of the relationship between temperament clusters and index offense types could not be conducted. Future multi-center studies with larger samples are needed to examine these associations and enhance forensic applicability. Finally, given the exploratory nature of this study, the identified subtypes should be considered preliminary and hypothesis-generating. Future research should incorporate multi-informant data and additional biological or cognitive markers to further elucidate the mechanisms underlying temperament-psychopathology associations in juvenile populations.

Despite these limitations, the present study suggests that juvenile offenders referred for forensic psychiatric evaluation may not represent a homogeneous risk group, but rather may exhibit distinct patterns of temperament-related risk and psychopathology. These findings provide preliminary support for moving toward more differentiated intervention approaches tailored to subgroup-specific vulnerabilities.