Callous-unemotional (CU) traits are a youth antisocial phenotype associated with persistent violent and criminal (i.e., antisocial) behavior (Blair et al., 2014) that relate to affective (i.e., emotional) deficits of adult psychopathy involving poor affective response and empathy for others (Barry et al., 2000). Central CU traits impairments involve perspective taking (i.e., understanding another by adopting their perspective [Davis, 1980; 1983]; Anastassiou-Hadjicharalambous & Warden, 2008; Lui et al., 2016; O'Kearney et al., 2017) and cognitive control (also called executive control; Baskin-Sommers et al., 2015; Gluckman et al., 2016; Winters & Sakai, 2021). Cognitive control impairments relate to antisociality (Zeier et al., 2012) and is necessary to modulate socio-cognitive processes (for reviews see: Mahy et al., 2014; Wade et al., 2018) such as perspective taking (Lamm et al., 2010; Qureshi & Monk, 2018). Decrements in perspective taking associated with CU traits is exacerbated under increased cognitive control demands (Winters & Sakai, 2021). These impairments appear to partially result from differences observed in brain regions associated with perspective taking (e.g., Thijssen & Kiehl, 2017; Umbach & Tottenham, 2020) and cognitive control (e.g., Pu et al., 2017; Szabó et al., 2017; Yoder et al., 2016), but these neuroimaging results fail to adequately account for parametric brain pattern heterogeneity in youth with CU traits (e.g., Winters, Sakai, et al., 2021). Overall, cognitive control impairments robustly associate with antisocial behavior, and CU traits are a developmentally useful and clinically meaningful precursor to antisocial behavior. Given the inherent harm to others and society, it is imperative to fully understand what aspects of behavior and biology might explain this association and contribute to later violent and criminal behavior. While much has been learned about the neural underpinnings of antisocial and delinquent behavior among adults, the need for generalizing these findings to adolescents, accounting for individual heterogeneity of brain patterns (e.g., Winters, Sakai, et al., 2021), and pinpointing malleable intervention targets remains (Herpers et al., 2014). As such, the present study examines functional properties of adolescent brains and perspective taking – an important and impoverished ability among adolescents with elevated CU (Lui et al., 2016) – as a plausible explanation for the association between CU traits and cognitive control.

Cognitive control is differentiated from the umbrella of executive functions in its intentional relation toward goal directed behavior as opposed to habitual cognitive processes (Friedman & Robbins, 2022) that involve monitoring conflict to gage the demand for control and then using information from the present context to regulate goal directed behaviors (Botvinick et al., 2001). Impairments in cognitive control, and damage to the anterior cingulate cortex (ACC) in particular, reduces the capacity of resolving conflict necessary to respond to other's emotions and subsequent response to other's emotions (Maier & di Pellegrino, 2012). A lack of response to others’ emotions is considered a fundamental pathway for psychopathy (Blair, 2008; Blair & Mitchell, 2009). Interpreting this lack of emotional response has largely centered on the existence of profound affective impairments (Blair, 2008; Blair & Mitchell, 2009), but there is substantial evidence it can be explained by a “bottleneck” in the early stages of attention processing (i.e., cognitive control; Braver, 2012) blocking peripheral information processing (i.e., the response modulation hypothesis; Lorenz & Newman, 2002).This effect has been demonstrated using conflict paradigms measuring cognitive control, such as the Stroop or Flanker task, that are modified to include a spatial component for peripheral stimuli (e.g., Gluckman et al., 2016; Vitale et al., 2005). These studies suggest that CU traits are associated with decrements in cognitive control (Gluckman et al., 2016) that are especially prominent during high conflict, incongruent task conditions (Botvinick et al., 2001; Yeung, 2013).

Brain regions recruited during conflict paradigms measuring cognitive control center around the anterior cingulate cortex (ACC) and pre-supplementary motor area (pre-SMA) forming a conflict network. The ACC is a central region for detecting conflicts in information processing and signaling top-down control (e.g., Brown, 2013; Shenhav et al., 2013; Yeung, 2013). The pre-SMA is thought to play a leading role in response-based conflict (Egner et al., 2007; Nachev et al., 2007) by evaluating outcomes of actions (Bonini et al., 2014). The ACC and pre-SMA are particularly sensitive to conflict during the Stroop task (Banich, 2019; Milham & Banich, 2005) and have been used to evaluate conflict sensitivity (e.g., Yang et al., 2021). These specific ROIs are used to investigate cognitive control during flanker tasks with spatial stimuli (e.g., Iannaccone, Hauser, Ball, et al., 2015; Iannaccone, Hauser, Staempfli, et al., 2015). The dorsal medial prefrontal cortex has also been implicated (e.g., Alexander & Brown, 2011). However, others have argued that this activation reflects a longer task duration rather than a conflict response (Grinband et al., 2011a, 2011b), which has been supported empirically using both EEG and fMRI (Iannaccone, Hauser, Staempfli, et al., 2015). As such, many studies have not included the dorsal medial prefrontal cortex when investigating conflict paradigms to assess cognitive control (e.g., Iannaccone, Hauser, Ball, et al., 2015; Iannaccone, Hauser, Staempfli, et al., 2015).

Functional abnormalities in the ACC and pre-SMA among youth with CU traits further supports the presence of cognitive control impairments. For example, the ACC demonstrates less activity during facial emotion recognition tasks (Szabó et al., 2017) and pain responses (Marsh et al., 2013) as well as less functional connectivity seeded in the ACC (Yoder et al., 2016). Similarly, the pre-SMA demonstrates decreased activation when responding to others’ emotions (O'Nions et al., 2017) and viewing another person in pain (Decety et al., 2013), as well as aberrant functional connectivity of the pre-SMA (Pu et al., 2017) at higher levels of CU traits. It is therefore pertinent to focus on the ACC and pre-SMA to understand cognitive control in relation to CU traits.

Defining neurophysiological substrates of cognitive controls influence on perspective taking in relation to CU traits is critical for understanding mechanisms of antisocial behavior. Given that the above literature suggests functional coupling between the ACC and pre-SMA supports cognitive control during conflict paradigms, we hypothesize that those with elevated CU traits would have disrupted connectivity whereas greater connectivity would positively associate with cognitive control and indirectly impact CU traits. Such evidence could reveal brain function underlying cognitive control impairments associated with CU traits; and this investigation could be further extended by accounting for perspective taking.

Perspective taking, or capacity to adopt another's point of view and attributing their thoughts and feelings (Decety, 2011), is supported by distinct brain regions and is critical for healthy social (Decety, 2005), moral (Decety & Cowell, 2014), and prosocial behavior (Decety et al., 2016; Tamnes et al., 2018). CU trait impairments in perspective taking (Anastassiou-Hadjicharalambous & Warden, 2008; Lui et al., 2016; O'Kearney et al., 2017) predict antisocial behavior above clinical ratings of CU traits (Gillespie et al., 2018; Song et al., 2016). CU trait impairments in perspective taking have been observed in the brain. For example, brain regions supporting perspective taking consist of the temporal parietal junction (TPJ), medial prefrontal cortex (mPFC) and the posterior cingulate cortex (PCC) (for meta-analysis: Fehlbaum et al., 2021) that form a social network (Blakemore, 2012; Klapwijk et al., 2013; McCormick et al., 2018); and, youth with elevated CU traits demonstrate aberrantly decreased functional connectivity among these regions (Thijssen & Kiehl, 2017; Umbach & Tottenham, 2020; Winters & Hyde, 2022). Although underlying multiple cognitive functions, these regions’ associations with perspective taking are thought to reflect cognitive processes of reflection and understanding mental states of oneself and others’ (Buckner et al., 2008; Buckner & Carroll, 2007; Uddin et al., 2009). Less connectivity in this network may indicate difficulty inferring others’ cognitive and emotional states; whereas greater connectivity in the social network is associated with greater perspective taking in adolescents (Winters, Pruitt, et al., 2021). Thus, greater connectivity in the social network could be associated with lower CU traits indirectly via perspective taking. Such evidence could reveal brain patterns underlying social processing impairments associated with CU traits. Moreover, core decrements associated with CU traits could be revealed by examining these brain association with cognitive control.

Substantial evidence supports that perspective taking is supported by cognitive control in typically developing samples. For example, those with higher inhibitory control report higher perspective taking (Carlson, 2005; Lamm et al., 2010; Wade et al., 2018); and selecting between different stimuli to perform perspective taking is reliant on cognitive control (Qureshi & Monk, 2018; Qureshi, Monk, et al., 2020). Numerous studies converge on these links suggesting that cognitive control modulates perspective taking (for reviews see: Mahy et al., 2014; Wade et al., 2018). Perspective taking is thought to be modulated by (1) processing the emotional context and (2) resolving conflicts between one's own and others emotions (Deschrijver & Palmer, 2020).

Neuroimaging studies support links between cognitive control and perspective taking. Processing conflict during perspective taking has shown causal associations in transcranial magnetic stimulation of the dorsolateral prefrontal cortex (dlPFC) in typical samples (Kalbe et al., 2010; Qureshi, Bretherton, et al., 2020) and in those with CU traits (Konikkou et al., 2020), which can be influenced by exerting cognitive control (e.g., Seymour et al., 2018). Thus, it is plausible that social and conflict network connectivity with the dlPFC may underlie cognitive control and perspective taking impairments amongst youth with CU traits.

Youth with CU traits demonstrate connections between social and conflict networks that may account for impairments in the cognitive control and perspective taking link. In typically developing samples we expect anticorrelation between the social and conflict networks because regions of the conflict network are active when engaged in a task (i.e., task-positive network) whereas social network regions are active when not engaged in an external task (i.e., task-negative network; Uddin et al., 2009). However, youth with CU traits demonstrate less anticorrelation between task-positive and task-negative networks (Pu et al., 2017; Winters, Sakai, et al., 2021), which may represent developmental immaturity (Richardson et al., 2018). Given the importance of cognitive control for perspective taking, it is plausible that functional coupling observed with greater anticorrelation between these networks may be important for cognitive control and perspective taking impairments in CU traits.

Understanding neurophysiological underpinnings of how cognitive control and perspective taking are linked to CU traits would have tremendous value in the public health domain. We previously tested how affective theory of mind (a socio-cognitive process related to perspective taking) is affected when placing additional demands are placed on cognitive control. This study revealed that those higher in CU traits had greater difficulty inferring others’ emotions after taxing cognitive control (Winters & Sakai, 2021). Our study demonstrates a vulnerability in cognitive control that impacts theory of mind accuracy; however, it does not answer whether this vulnerability is because of cognitive control and perspective taking directly interacting nor does it identify the neurophysiological underpinnings of this relationship. Further investigation into these unanswered questions can help clarify the core behavioral and neural processes underlying CU traits, which may in turn provide valuable predictive and intervention-related information.

Previous studies characterizing the brain in relation to constructs of interest could be improved by modeling heterogeneity of individual connectivity in adolescent brains. For example, functional brain patterns are as unique as fingerprints (Damoiseaux et al., 2021). Ignoring this heterogeneity when modeling brain features can cause spurious associations, whereas modelling the heterogeneity of individual brains can more accurately characterize network patterns (Gates & Molenaar, 2012). Additionally, lagged connections, or the temporal relationship such that a prior timepoint of one brain node predicts the future timepoint of another brain node, are often overlooked in traditional approaches but are an important feature characterizing brain function (Mitra et al., 2014) and help improve estimation of contemporaneous connections. Thus, the present study uses a method that accounts for heterogeneity and both lagged and contemporaneous connections called subgrouping group iterative multiple model estimation (S-GIMME; Gates et al., 2017). S-GIMME and GIMME outperform other network modeling approaches (e.g., Bayes nets and Granger causality; Gates et al., 2017; Gates & Molenaar, 2012), but here we use S-GIMME because it improves estimation of individual level connections over GIMME by providing the model with more known priors (Beltz & Gates, 2017). These methods have been used to demonstrate brain heterogeneity in psychopathy (Dotterer et al., 2020) as well as CU traits (Winters, Sakai, et al., 2021). Thus, it is appropriate to leverage S-GIMME to characterize functional brain properties of CU traits and underlying psychological processes supporting these symptoms.

In addition, prior work investigating the brain and CU traits primarily relies on task-based activations and our use of S-GIMME could meaningfully build on this prior work by incorporating a contemporary understanding of brain function. As opposed to the modular view of the brain espoused by task-based studies making up the majority of investigations on CU traits and the brain, contemporary views on the brain recognize that human cognition requires the integration of multiple brain regions (McIntosh, 2000), forming networks supporting brain function and that understanding the functional connectivity across these regions and networks can provide a deeper insight into human behavior (Bassett & Sporns, 2017). This functional connectivity represents the distributed function among brain regions (Zhang et al., 2021), that capture critical developmental features of adolescent brains (Ernst et al., 2015; Uddin et al., 2011). Investigating adolescent brains using S-GIMME will quantitate individual level functional connectivity while allowing us to derive network properties of node centrality and network density. Network density quantifies the number of connections that exist between the set of nodes making up a network with higher values indicating more paths where information can be transferred between nodes of that network and lower values indicating fewer paths for transferring information between nodes. The density of these paths can represent positive or negative connections. Higher positive density indicates regions activating together whereas higher negative density indicates regions activating opposite to each other, or more functional coupling. Node centrality measures the number of connections a node has relative to and potential number of connections it could have, which represents the importance of that node. Positive node centrality indicates the number of positive connections and negative node centrality indicated the number of negative connections.

The present study aims to 1) examine the extent to which cognitive control's association with perspective taking accounts for CU traits and 2) characterize the related functional brain properties in a community sample of early-to-mid adolescents (ages 13-17). We took an iterative approach by first examining the behavioral model, then the associated functional brain properties, before constructing a final model. To characterize the functional brain properties, we used S-GIMME to generate person specific connectivity maps to derive network density and node centrality for all participants. We hypothesized that cognitive control's positive association with perspective taking would partially account for CU traits. For functional brain properties, we hypothesized the following: a) elevated CU traits would associate with less positive connection density in the conflict and social networks; b) cognitive control would be positively associated with connection density in the conflict network; and c) perspective taking would positively associate with positive connection density in the social network. For the dlPFC association with social and conflict networks, we hypothesized that negative connections would correlate positively with both perspective taking and cognitive control. For node centrality, the right TPJ is a central node for perspective taking (Martin et al., 2020), so we hypothesized that perspective taking would associate with node centrality in the social network; and, given the ACC is central for detecting conflict, we hypothesized that cognitive control would associate with centrality in the ACC. Finally, we anticipated that less negative density between the social and conflict networks would associate with CU traits and greater negative density would associate with cognitive control and perspective taking.