Youth delinquency represents a global public health challenge, leading to negative consequences for individual development (e.g. legal problems, health problems, low educational attainment, low socio-economic status), for family functioning, and for society in general (1). Although increased efforts have been invested to reduce delinquent behaviors in youth, current intervention and prevention strategies have proven to be limited in their efficiency (2). For this reason, a large body of research focused on the psychological factors underpinning youth delinquency. Delinquent behaviors have been consistently associated with high psychopathic traits, described by three main dimensions - emotional callousness (CU), impulsivity and interpersonal manipulative behaviors (3, 4). The impulsivity dimension of psychopathy has been linked to reactive forms of delinquency, probably underpinned by difficulties in emotion regulation and impulse control (5, 6). Furthermore, previous studies have suggested that individuals with high CU and interpersonal manipulative traits are associated with calculated, goal-oriented, proactive delinquent behaviors (7–9). Of note, impulsivity is not specific to delinquency but rather a shared trait in externalizing disorders [e.g. ADHD, conduct disorder, substance abuse (10)].

Youth delinquency has also been previously associated also with impairments in social-cognitive processes, such as reduced empathy, as in inability to feel and understand others’ emotions (11–13) and reduced reflective functioning (14, 15). Reflective functioning (RF) refers to the capacity to recognize that one’s own and others’ behaviors are motivated by internal states (e.g. emotions, thoughts, desires etc.) (16). Thus, reduced RF represents the difficulty to reflect on behaviors in terms of mental states, difficulties that can be assessed using the subscale of uncertainty about mental states of the Reflective Functioning Questionnaire (17). Furthermore, several epidemiological studies have indicated higher rates of psychological problems in adolescents presenting delinquent behaviors (18, 19), some of which (e.g. attention and hyperactivity problems and conduct problems) put youth at increased risk of committing delinquent acts (20). Although a large body of research has focused on the psychological factors, little is known about the neural correlates of youth delinquency. Investigating the neural processes and their correlates with psychological risk factors can provide an integrative understanding of the mechanisms sustaining delinquent behaviors in youth, which may stimulate further development of prevention and intervention strategies.

A promising approach to understand the neural correlates of behavioral and psychological processes is to conceptualize and study the brain as a network of interconnected regions (21), by using resting state functional magnetic resonance imaging (rs-fMRI). rs-fMRI measures the spontaneous fluctuations of blood oxygen level-dependent (BOLD) signals of various regions (22, 23) to assess the brain’s patterns of functional organization “at rest”, in the absence of any explicit task (24). Focusing on the resting state FC in delinquent, incarcerated adolescents (IA) compared to non-delinquent, community adolescents (CA) will help us understand the specific default functional connectivity (FC) associated with delinquent behaviors in youth, as well as less extreme maladaptive behaviors. Compared to the CA, IA previously presented both increased and decreased resting state FC between brain regions previously identified as part of different brain networks, implicated in self-referential thinking (e.g. default mode network DMN), impulse control (e.g. executive control network ECN) and integration of sensorial information (e.g. salience network- SN). Chen and colleagues showed dysfunction in a right medial prefrontal-caudate circuit (25), further Aghajani and colleagues found psychopathic trait-specific increases and decreases between the amygdala subnetwork and respectively cortical midline structures, and frontolimbic network (26, 27). Finally, Sun and colleagues observed default mode network alterations with increased FC in posterior cingulate and decreased FC in middle temporal, angular, precuneus and middle frontal cortex (28).

Furthermore, previous studies report that the differences in FC observed in IA could be partially explained by the presence of psychological and personality factors. For instance, IA presenting high scores on the emotional callousness dimension of psychopathy showed reduced FC between brain regions previously identified as part of DMN and SN (26, 27, 29). On the other hand, the presence of high impulsivity and interpersonal psychopathic traits have been associated with increased connectivity in executive and motor networks, but also in DMN and SN networks (30, 31). Although these studies provided some insights into the neural correlates of psychopathic traits in IA, including the lack of empathy present in CU traits, rs-FC associations with other psychological factors associated with delinquent behaviors in youth (e.g. RF) as well as with different psychological problems presented by IA have yet to be examined.

Understanding the neural correlates of socio-cognitive processes and psychological problems related to delinquent behaviors is especially important during adolescence, a critical period in the development of socio-cognitive processes (32) and for the onset of a wide range of psychological problems (33). Furthermore, most previous studies focused on a priori defined brain regions and networks, overlooking the FC of more distributed network of brain regions. Using a whole-brain approach widens our understanding about functional architecture in IA compared to a group of CA.

The present study aims to map the potential differences between incarcerated (IA) and community (CA) male adolescents in rs-FC among a comprehensive set of 142 distinct cortical and subcortical brain regions, without defining a priori brain networks. For this purpose, we used a whole-brain connectivity approach, namely the Network Based Statistics (NBS) (34, 35). NBS maps regional brain networks that show a significant between-group differences in FC, by identifying inter-regional brain connections that are associated with a contrast of interest (in this case, IA vs. CA). The idea behind NBS is to utilize the presence of connections associated with such contrast of interest to potentially yield greater power than is possible by correcting the p-values computed for each connection comprised in the graph. We expect to identify differences in FC within specific subnetworks (e.g. DMN, SN, ECN). Given the mixed results of previous studies regarding the nature of resting state FC differences (increased or decreased connectivity) between IA and CA, and that the present study is the first to use a whole-brain NBS approach to study resting state FC in delinquent adolescents, we do not hypothesize any specific directions of FC differences between the two groups. Resting-state functional MRI (rs-fMRI) studies have employed various analytical approaches to investigate neural connectivity in populations with behavioral and cognitive impairments. Seed-based analyses, which focus on pre-defined regions of interest, have been instrumental in identifying focal alterations in connectivity, particularly in regions like the amygdala and prefrontal cortex that are implicated in emotion regulation and impulse control. For instance, studies of delinquent juveniles with high callous-unemotional (CU) traits have reported reduced connectivity between the amygdala and prefrontal regions, aligning with deficits in emotional processing and moral reasoning (26). However, this method is constrained by its reliance on a priori localized, single-seed hypothesized pathways of connectivity, potentially overlooking broader network-level disruptions. Independent Component Analysis (ICA), another commonly used approach, enables the decomposition of functional networks, such as the default mode network (DMN) and salience network (SN). Studies using ICA have observed reduced DMN connectivity in adolescents with conduct disorder and externalizing behaviors, implicating deficits in self-referential processing (36). While powerful, ICA results can vary depending on methodological choices, such as the number of components extracted, and may not capture inter-network interactions.

In contrast, the Network-Based Statistic (NBS) method used in this study addresses these limitations by identifying significant subnetworks of connectivity differences without the need for predefining regions or networks. This data-driven approach is particularly suited to capturing distributed connectivity patterns across the brain, which may reflect the heterogeneity of delinquent behaviors and their neurobiological correlates. For example, adolescents with externalizing behaviors often exhibit impairments across multiple functional networks, including the DMN, SN, and executive control networks (ECN), suggesting that a whole-brain approach like NBS is critical for detecting these complex and interconnected patterns.

A secondary aim of the present study is to investigate the associations among resting state FC and psychopathic dimensions (CU, impulsivity, and interpersonal problems), socio-cognitive processes (empathy and reflective functioning), and different psychological problems (externalizing, internalizing, attention, and hyperactivity and thought problems). We expect distinct associations among psychopathy dimensions and resting state connectivity. Based on previous evidence, we hypothesized that the callousness dimension would be associated with reduced resting state FC, whereas the impulsivity and interpersonal dimensions would correlate with increased FC. This hypothesis stems from previous literature on the neurobiological underpinnings of psychopathy, where callous-unemotional traits have been linked to reduced connectivity in brain networks related to emotional regulation (e.g. DMN), suggesting a deficit in affective processing. In contrast, impulsivity, a common feature across externalizing disorders (e.g. ADHD, conduct disorder), has been associated with hyperconnectivity in areas implicated in reward processing and executive function (37, 38). Given the novelty of our study, investigating the relations among resting state connectivity and socio-cognitive processes and psychological problems, we do not hypothesize any specific direction of these associations.

This study represents the first investigation of whole-brain resting state functional connectivity (FC) in a group of incarcerated adolescents (IA) compared to a group of community adolescents (CA) and the correlates with psychopathic traits together with socio-cognitive processes (empathy and reflective functioning RF) and psychological problems. Compared to the CA, IA showed altered FC in a distributed subnetwork including several cortical regions; temporo-parietal regions and mPFC (as part of default mode network- DMN), dACC (as a major node of the salience network -SN), occipital cortex (part of the visual network), and the frontal network, but also subcortical regions, such as basal ganglia and thalamus. We hypothesized that callousness would be associated with reduced resting-state FC, while impulsivity and interpersonal dimensions would correlate with increased FC. These claims were supported by the data; in the IA group, the mean connectivity in the subnetwork was positively correlated with lower reflective functioning levels (e.g. higher uncertainty about mental states-RFQu), whereas in the CA it was negatively correlated with the interpersonal dimension of psychopathy. These results were further supported by correlations in the whole sample of adolescents: the mean connectivity was positively correlated with impulsivity, externalizing behaviors, and RFQu. Finally, linear regression models revealed that the mean connectivity was predicting higher scores on the RFQu, while considering the incarceration status and all the psychological variables.

Our results are in line with previous literature that suggested atypical resting state FC in IA compared to CA (25–27), but also with previous studies on adult and adolescent delinquents (30, 56, 57) and other externalizing behaviors in adolescent and adult populations (58, 59). Using the whole brain NBS analysis, the results of our study add evidence to existing research, suggesting that atypical FC in IA extends to several brain regions, including part of the SN (e.g. dACC), DMN (e.g. parietal, temporal cortex, vmPFC), frontal network (e.g. supfrontal), visual network (e.g. occipital cortex), and subcortical regions (e.g. basal ganglia, thalamus). These brain regions have been previously associated with impairments in different psychological processes that also characterize IA profiles and other populations presenting behavioral problems (58, 60–63), and encompass impairments in self and other processing (vmPFC, mPFC), aberrant reward processing (basal ganglia), lack of impulse control and emotion regulation (dACC, basal ganglia), lack of moral reasoning (mPFC), and weaker integration of interoceptive signals (thalamus and parietal regions). While our findings indicate that these distributed brain regions are involved, it is important to note that the observed relationships between FC alterations and behavioral/psychological impairments are correlational and do not imply direct causality. The NBS approach provides unique insights into the neural correlates of delinquent behaviors by identifying subnetworks that are not easily detected with other methods. For example, the increased connectivity observed in this study between regions of the DMN (e.g., medial prefrontal cortex, posterior cingulate cortex), SN (e.g., dorsal anterior cingulate cortex), and visual and frontal networks suggests widespread alterations in resting-state functional architecture. These findings align with prior research linking DMN dysfunction to impaired self-reflection and moral reasoning in antisocial individuals (61) and SN alterations to deficits in emotion regulation and impulse control (60). By identifying these interconnected subnetworks, NBS highlights how distributed connectivity changes may contribute to core deficits, such as low RF and externalizing behaviors, that characterize delinquent adolescents. Furthermore, the distributed subnetwork identified in this study suggests potential compensatory mechanisms, where increased connectivity in some regions might reflect attempts to offset deficits in others. This is consistent with studies showing that adolescents with high impulsivity may exhibit hyperconnectivity in frontal regions to compensate for reduced top-down control (29). Unlike traditional methods, NBS captures these complex, distributed patterns, offering a more comprehensive understanding of the neural mechanisms underlying delinquent behaviors.

The observed distributed regions affected in the subnetwork suggest that the heterogeneity of IA profiles might only be captured by whole brain analyses. While it is possible for highly focal effects to exist in one given network, most networks - being interconnected, by definition, are likely to show secondary consequences of any focal effects that can propagate along interconnected pathways. Therefore, most effects are likely to influence interconnected subnetworks that the NBS is well suited to detect. In the present study, it seems that IA adolescents present a differential resting state architecture involving regions from several “independent” networks. The distinct interactions between subregions of the networks are correlated with behavioral and psychological impairments observable in IA, more specifically reduced RF, impulsivity and externalizing behaviors. Additionally, these associations might reflect compensatory mechanisms aimed at enhancing the cognitive, motor, and emotional control, despite high levels of impulsivity at the behavioral level (64). Thus, previous studies investigating only a subnetwork might only look at its association with individual variables characterizing IA impairments. However, we acknowledge that the alterations in the observed subnetwork do not necessarily sustain or cause impairments in reflexive functioning (RF). Rather, they may reflect a complex interplay of factors, including neurodevelopmental trajectories and environmental influences, that contribute to these socio-cognitive processes. We could also consider the hypothesis of delayed maturation in the delinquent group, particularly given that we observed a negative correlation between age and mean connectivity. The maturational lag has been proposed in the literature on antisocial behaviors and psychopathy (65), and some studies have linked maturational delays in brain development, such as slow-wave activity to antisocial traits (66). Further, brain development during adolescence is influenced by different developmental factors, such as childhood trauma, prenatal factors, and environmental factors (67). Interestingly, early alcohol abuse has been shown to impair the maturation of brain regions responsible for cognitive control, such as the prefrontal cortex, which is crucial for regulating emotional impulses (68). This disruption can result in long-lasting effects on executive functioning, impulse control, and decision-making, potentially increasing susceptibility to antisocial behavior in later adolescence and early adulthood (69).

The association with RF impairments is not surprising, as the identified altered subnetwork includes regions evidenced to play a role in mental state reasoning, encompassing cortical midline structures, medial PFC, anterior cingulate cortex, and medial posterior parietal cortices (70–72), and the temporoparietal junction (73, 74). Mentalizing, especially the usage of reflective functioning in the context of attachment, has previously been reported as mediating the association between psychopathic traits and proactive aggression in a community sample of typically developing adolescents (75, 76). In our study, the association between low RF and FC in extended brain regions related to different theoretical frameworks, suggests that social-cognitive impairments in delinquent populations might be associated with atypical whole brain connectivity and integration, not limited to the regions previously associated with social cognition, such as the DMN (61). This result comes in the continuity of previous studies suggesting that RF and executive functions, such as impulse control are distinct, but interdependent processes (77). However, it remains unclear which neural mechanisms are shared, and which ones are specific to each of these higher-order processes. Future research should further explore these associations to better understand these mechanisms.

Our results might provide new insights on the mechanisms and neural basis potentially linked to low RF, not only in the IA, but also in CA. Increased resting state FC identified in our study included brain regions previously associated with self-reflection, self-other differentiation, and inference of others’ beliefs, which might explain the self-centered, egocentric bias and the impairments in perspective taking previously observed in IA (39, 78). Additionally, increased connectivity between occipital cortex and different brain regions from SN (dACC) and frontal regions might indicate that IA rely to a greater extent on sensorial, external stimuli, in the detriment of internal, abstract cognitions, which might lead to a bottom-up social cognitive style (79) and ultimately resulting in an external orientated, concrete thinking about mental states and their impact on behaviors (15, 76, 80–82). Further, we could explore the possibility that reduced RF may reflect a broader developmental lag, influenced by both biological and environmental factors. Biologically, this lag could be linked to maturation of brain networks involved in social cognition, such as PFC and temporo-parietal junction, essential for mentalizing and understanding others’ emotions and perspectives. These areas gradually mature throughout adolescence and early adulthood, suggesting that impairments in RF may result in part from delayed cerebral maturation (83). Environmentally, factors such as neglect, early adverse events, or insufficient social interactions during critical developmental periods can further disrupt the maturation of these brain networks. Limited opportunities for reciprocal friendships, group affiliation and social learning might also inhibit the capacity to reflect on mental states, leading to reduced RF (84). Thus, a combination of biological maturation delays and reduced social interactions may contribute to deficits in RF, particularly in populations presenting high levels of antisocial behavior and/or emotional dysregulations. Nevertheless, these interpretations need confirmation from future studies focusing on the neural basis of RF and their role in the development of delinquent behaviors during adolescence. In addition, most previous studies used task-based functional MRI to assess the association between FC and behavioral or socio-cognitive impairments related to IA. Contrasting with previous studies observing a decrease in FC in IA (25, 26) and in community adolescents presenting antisocial behaviors (25, 36, 59, 85–87), our results indicate that, compared to the CA, the IA presented only increased resting state FC. Notably, previous studies showed that the impulsivity dimension of psychopathy was associated with increased resting state FC (29, 88–92), whereas the emotional callousness dimension was associated with decreased resting state FC in adolescents and adults presenting antisocial behaviors (30, 58, 93), which could explain these discrepancies. This result might be explained by the characteristics of our subgroups. IA adolescents presented significantly higher impulsivity and externalizing behaviors and lower verbal IQ and RF than the CA. However, the two subgroups did not differ in the emotional callousness, interpersonal dimensions, empathy scores or any other psychopathology problems scores. These group characteristics might suggest that IA presented mainly reactive, impulsive forms of delinquent behavior. Furthermore, the CA subgroup might also be a heterogenous group in terms of psychopathology profiles, since the two subgroups do not differ in the ADHD scores, which is consistent with previous studies suggesting that ADHD symptoms is also prevalent in the general population, especially in boys (94). This is further sustained by our findings, suggesting that in the whole sample (IA and CA combined), higher levels of impulsivity and externalizing behaviors were positively correlated with the increase in connectivity in the subnetwork identified by the NBS. These results might suggest that the subnetwork identified by the NBS is not specific to delinquency, but connects regions involved in emotional and behavior dysregulations, present in both normative and clinical populations presenting externalizing behaviors. Future studies should replicate our findings using the NBS approach in IA groups presenting high levels of CU traits.

In summary, while the interpretation of case-control differences in resting-state functional MRI data may not always be straightforward (95), the NBS identified a core network of increased FC that is consistent with psychopathological alterations reported in the literature. Furthermore, findings of the present study could be explained as a cascade of associated processes. Our results suggest that the increased connectivity in the subnetwork identified by the NBS might be associated with processes involved on one hand in low socio-cognitive abilities, namely low RF, and in behavioral manifestations such as impulsivity and externalizing behaviors. We could hypothesize that, the increased FC observable in IA, driven by impairments in RF associated with a bottom-up cognitive style, will translate to externally oriented, concrete thinking about mental states, which in turn could lead to impairments in self-regulation manifesting as impulsivity and externalizing behaviors. Contemporary developments in mentalization-based therapy are currently evaluating the value of focusing on such processes in psychological therapy for adolescents with conduct disorder (96, 97), a youth group sharing some characteristics with the IA group in this study. Future studies should also address individual factors, such as early adverse events, which may further contribute to the understanding of RF impairments in juvenile offenders (15).

There are limitations to this study that need to be acknowledged. First, some of the brain regions identified as part of the subnetwork are relatively large, with substructures that have been previously related to different functions (e.g. basal ganglia). Further, while the identified subnetwork encompasses widely distributed regions, it reflects a significant and relevant pattern of altered connectivity that warrants further investigation. Future studies might use alternative connectivity methods following hypothesis-driven analyses, or more targeted network analyses to refine the understanding of these connectivity changes. Secondly, the cross-sectional design limits our developmental interpretation; future longitudinal studies conducted on IA are needed to confirm the implication of RF and the neural mechanisms at stake in the emergence of antisocial behaviors. We acknowledge that while the increased connectivity observed in the subnetwork correlates with behavioral and socio-cognitive measures, the nature of these relationships requires further exploration. Future longitudinal studies, ideally incorporating both resting-state and task-based fMRI paradigms, could help elucidate the causal pathways and mechanisms underlying these associations. Additionally, social-cognitive processes and psychopathic traits were investigated using self-reported questionnaires, which represent screeners requiring further in-depth exploration of these psychological processes. Future studies need to focus on these variables using other types of measures, such as experimental tasks or interviews. Furthermore, the sample size of the study is relatively small, and the results must be replicated on larger samples. The sample was also composed only of males, it would be interesting to investigate if the results can be replicated in female samples. Finally, although the number of days since admission to the juvenile detention center and the MRI testing was not significantly correlated with the resting state connectivity in the subnetwork identified by NBS, other factors associated with incarceration (e.g. social isolation, stress etc.) that we did not consider in the present study could influence the NBS results. All the variables entered in the GLM model (including group membership) explained 58% of the variance in the resting state connectivity, thus other variables (e.g. low socio-economic status, low IQ, childhood trauma, alcohol and drug use) might explain the resting state connectivity differences. Future studies must continue investigating the psychological variables associated with whole brain resting state connectivity in IA, ideally using matched samples on these possible confounding variables. Of note, the present study aimed to investigate the differences in resting state FC between two groups of IA and CA, and their correlations with psychological factors, it is therefore limited to these objectives and cannot be generalized to index the risk for delinquent behaviors or to qualify as biomarkers of delinquency.

Overall, our study extends previous research by identifying a unique subnetwork of brain regions that presented increased resting state FC in a group of IA, compared to CA. The increased connectivity in this subnetwork was uniquely associated with socio-cognitive impairments, namely low reflective functioning. While providing understanding of whole brain resting state FC in IA, our results might also provide some insights about the impaired mechanisms implicated in the socio-cognitive processes during adolescence.