Differential adolescent neurodevelopment of emotion processing across internalizing psychopathology and childhood adversity Authors: Galit Karpov, Brian W Kleinschmidt, Sara A Heyn, Justin D Russell, Ryan J Herringa Published: 2026 ABSTRACT AbstractBackgroundInternalizing disorders (IDs) such as anxiety and depression show a peak onset during adolescence, a period characterized by maturation of the frontal lobe. Impaired emotion regulation and elevated reactivity are prominent transdiagnostic symptoms that are further impacted by childhood adversity. However, the mechanisms through which IDs and adversity alter neurodevelopmental trajectories of emotion processing during adolescence are poorly understood, particularly across implicit and explicit regulation strategies.MethodsUtilizing an accelerated longitudinal design, youth (aged 10–17) with IDs (n = 65) and a typically-developing comparison group (CG; n = 74) completed an emotional N-Back (implicit) and cognitive reappraisal (explicit) task during fMRI at baseline and one-year follow-up. Linear mixed effects modeling was used to investigate group and age differences in regional frontal lobe activation, then explored how these effects were moderated by adversity.ResultsAs compared to CG-youth, ID-youth showed developmental differences in the left ventral anterior insula during the implicit task, and the left dorsal anterior insula during the explicit task. Across both tasks, ID-youth lacked the increasing differential activation between negative and neutral stimuli across age that CG-youth exhibit. In ID-youth, adversity moderated neural activity during the explicit task, including the developmental trajectory of the dorsal anterior insula, but not the implicit task.ConclusionsThe left anterior insula is a key developmental region with subregion-specificity. These findings highlight the importance of individual neural maturation and carry implications for emotion regulation-based pediatric clinical interventions. Further investigation is necessary to assess whether intervention effectiveness is dependent on adversity history.Keywords: Functional magnetic resonance imaging, Adolescence, Emotion, Internalizing disorders, Anterior insula, Childhood adversity Introduction Adolescents have seen a global increase in the prevalence of psychiatric conditions, including internalizing disorders (IDs) such as depression and anxiety (Piao et al., 2022). Adolescence coincides with both brain maturation and the peak onset of IDs, suggesting that alterations in neurodevelopmental trajectories in this period contribute to psychopathology (Blackford and Pine, 2012, Kessler et al., 2007, McLaughlin et al., 2011a, Park et al., 2014, Uhlhaas et al., 2023). Elevated emotional reactivity and difficulty with emotion regulation are particularly robust transdiagnostic symptoms in IDs (Compas et al., 2017, Dryman and Heimberg, 2018, Marano et al., 2025), although the developmental mechanisms underlying these symptoms remain understudied. Relatedly, exposure to chronic adversity is strongly associated with increased psychiatric symptom severity, including emotion regulation dysfunction, often resulting in a more detrimental psychiatric course and greater treatment resistance to conventional therapies (Teicher et al., 2022). Incorporating adversity in our understanding of IDs is imperative, as half of psychiatric patients have significant experiences of childhood adversity (Kuzminskaite et al., 2022; J. Nelson et al., 2017; Struck et al., 2020; Teismann et al., 2025). Given that successful interventions in youth with IDs are associated with improvements in regulation (Daros et al., 2021, Eadeh et al., 2021, Moltrecht et al., 2021), knowledge of the underlying neural trajectories can inform novel treatment approaches or fine-tune existing therapies to improve the effectiveness of clinical intervention. The seminal Extended Process Model (Gross, 2015, Sheppes et al., 2015) highlights three stages of emotion regulation: identification, selection, and implementation. The identification stage is concerned with whether there is a need to regulate emotion based on the activity of valuation systems that identify salient stimuli to generate emotional responses (i.e., reactivity). When the need for regulation is identified following sufficient emotional reactivity, the selection stage determines which strategy to use, from options including but not limited to implicit (unconscious and automatic) or explicit (conscious and voluntary) regulation. The last stage implements the chosen regulation strategy. Difficulties within this emotion process can occur due to increased emotional reactivity in valuation systems, decreased ability to implement regulation strategies, or both. The neural basis is generally understood to involve the prefrontal cortex (PFC), which implements regulation strategies to modulate activity in the valuation systems that control emotional reactivity in regions such as the amygdala and insula (Braunstein et al., 2017, Gyurak et al., 2011; Herringa, 2017; Marano et al., 2025; McRae et al., 2012; Silvers et al., 2015). The current study is specifically concerned with how IDs alter both the early stages (identification/emotional reactivity) and the later stages (implementation/regulation) of this emotion processing across development. During adolescence, the PFC undergoes significant maturation. This occurs primarily by increasing the efficiency of connections with subcortical and limbic regions via mechanisms such as synaptic pruning and myelination (Ahmed et al., 2015, Delevich et al., 2021, Larsen and Luna, 2018, Casey et al., 2019, Marano et al., 2025). Improvements in implementing regulation during this period is often associated with reduced emotional reactivity in the amygdala and insula, and increased top-down control of the PFC (Casey et al., 2019, Gee et al., 2013, Heller et al., 2016, Keding and Herringa, 2016, Pitskel et al., 2011, Silvers et al., 2015, Silvers et al., 2017a, Somerville et al., 2010, Stephanou et al., 2016). The valuation system assesses saliency differently across development, such that children and early adolescents perceive neutral stimuli more negatively than adults (particularly faces, e.g., (Marusak et al., 2017, Thomas et al., 2001, Tottenham et al., 2013). Accordingly, reductions in neural emotional reactivity occur for both negative and neutral stimuli throughout adolescence (Karpov et al., 2026, Silvers et al., 2017b). Our previous longitudinal work showed that changes in emotional reactivity decrease at a faster rate to neutral stimuli relative to negative stimuli (Karpov et al., 2026), suggesting a role for differentiation in valuation across valence in developmental improvements during emotion processing. Such decreases in reactivity can be observed even in the absence of active regulation, including during passive viewing (Silvers et al., 2015, Silvers et al., 2017b). In this way, emotional reactivity and regulation are inter-related processes (Sheppes et al., 2015), where adolescent maturation of regulation-related PFC circuits are mirrored by concurrent decreases in emotional reactivity (Casey et al., 2019). The exact circuits underlying the implementation of regulation differ across strategies. In adults, the ventromedial (vm)PFC and ventrolateral/dorsolateral (dl)PFC shown increased recruitment during implicit and explicit strategies, respectively (Braunstein et al., 2017, Etkin et al., 2015, Gyurak et al., 2011, Ochsner et al., 2012). In typically developing youth, these two regions show heterogeneous maturation trajectories. The vmPFC reaches maturity earlier than lateral regions (Shaw et al., 2008), with the dlPFC being the last PFC region to fully mature (Casey et al., 2000). These regional maturation differences likely underly changes in regulatory ability across strategies. It has been observed that explicit regulation strategies such as cognitive reappraisal are more effective in older adolescents (Silvers et al., 2017a, Stephanou et al., 2016, Theurel and Gentaz, 2018), whereas both implicit and explicit regulation are equally effective for early adolescents (Theurel and Gentaz, 2018). From such work, it has been proposed that implicit and explicit strategies have differing developmental trajectories (Silvers, 2020, Zou and Yuan, 2021), although the paucity of longitudinal studies preclude any strong conclusions (Silvers, 2020). Similarly, the ways in which IDs alter neurodevelopmental trajectories are not well understood. Pediatric studies have predominantly relied on static group differences between youth with and without psychopathology, with some using cross-section cohorts to assess development. Current evidence from such designs describe youth with IDs as having difficulties across multiple stages of emotion processing. ID adolescents show increased emotional reactivity to negative stimuli, particularly in the amygdala (Beesdo et al., 2009, Blackford and Pine, 2012; Herringa, 2017; Keding and Herringa, 2016; Kerestes et al., 2014; Padgaonkar et al., 2021; Perlman et al., 2012; Yang et al., 2010). Furthermore, ID individuals often evaluate neutral stimuli as more negative, a phenomena referred to as a negative interpretation bias (Chen et al., 2020, Hirsch et al., 2016, Klein et al., 2018, Orchard and Reynolds, 2018, Smith et al., 2018, Stuijfzand et al., 2018). Accordingly, higher emotional reactivity to neutral stimuli is observed in the amygdala and insula (Cooney et al., 2006, Filkowski and Haas, 2017, MacDonald et al., 2025, Padgaonkar et al., 2021), as well as lower neural differentiation between aversive and non-aversive stimuli (Denefrio et al., 2019, Glenn et al., 2020, Henje Blom et al., 2015). The association between negative interpretation bias and internalizing symptoms increase as youth age, suggesting an important role of development (Songco et al., 2020, Stuijfzand et al., 2018). Difficulties in implementing regulation strategies in response to these heightened reactions have largely been attributed to decreased PFC-amygdala coupling in youth (Connolly et al., 2017, Heyn et al., 2019, Keding and Herringa, 2016, Kerestes et al., 2014, Padgaonkar et al., 2021, Perlman et al., 2012, Swartz and Monk, 2014, Tang et al., 2018, Wolf and Herringa, 2016), posited to reflect the decreased ability of the PFC to modulate emotions. Despite this, the patterns of regulation-related PFC activity in ID youth has been mixed; evidence for decreased (Colich et al., 2016), no difference (Díaz et al., 2025, LeWinn et al., 2018, McLaughlin et al., 2014, Padgaonkar et al., 2021), and increased (Blackford and Pine, 2012, Padgaonkar et al., 2021, Stephanou et al., 2017) ventromedial/lateral PFC activity have been reported relative to typically developing and low-symptom severity youth. Few studies have directly investigated the moderating effect of adversity within clinical samples. However, it has been theorized that psychiatric patients with and without significant early life adversity differ in neural structure and function as a result of stress-induced alterations of developmental trajectories (Teicher et al., 2022, Teicher and Samson, 2013). One altered developmental pathway includes difficulties with emotion regulation, which shows a well-established association with childhood adversity (Dvir et al., 2014, Gruhn and Compas, 2020; Kim and Cicchetti, 2010; Kim et al., 2023; Lambert et al., 2017; Marusak et al., 2015; McCrory et al., 2012; McCrory et al., 2017; Tottenham et al., 2010; VanTieghem and Tottenham, 2017; Villalta et al., 2018), including mediating the relationship between adversity and psychopathology (Huang et al., 2025, Jennissen et al., 2016, Larkin and Shelleby, 2026, Miu et al., 2022, Sabzehparvar and Amiri, 2025, Weissman et al., 2019). Similar to symptoms observed in IDs, community youth show increased neural reactivity to negative stimuli in salience-related regions (Hein and Monk, 2017; Herringa, 2018; Herringa et al., 2016; Jenness et al., 2021; McCrory et al., 2011; McLaughlin et al., 2015; Rodman et al., 2019; Suarez et al., 2022; Suzuki et al., 2014; Tottenham et al., 2011; White et al., 2019; Wymbs et al., 2020) along with decreased neural and physiological discrimination between negative and neutral stimuli (Iffland et al., 2018, Qiao et al., 2025). There is some evidence for increased lateral PFC activity during implementation of regulation in adversity-exposed community youth (Eaton et al., 2025, Jenness et al., 2021, McLaughlin et al., 2015), which has been interpreted as a compensatory mechanism for inefficient PFC-limbic communication. Given that decreased recruitment of lateral PFC activity during reappraisal (Rodman et al., 2019) and emotional reactivity (Jenness et al., 2021) was associated with greater psychopathology symptoms in adversity-exposed youth, it seems plausible that adolescents with a diagnosed ID would show similar patterns. The overlap in symptoms between adversity-exposed community youth and ID youth conceivably contributes to the worse psychiatric outcomes commonly observed in those with both an ID and high adversity exposure. However, the dearth of longitudinal studies has limited our understanding of how IDs and adversity alter neurodevelopmental trajectories of emotion processing. In this preregistered study (Karpov et al., 2025), we aim to directly address this gap via two questions. First, how the neurodevelopmental trajectories of emotion processing across implicit and explicit regulation strategies differ between youth with and without IDs, and second, how adversity exposure moderates these trajectories within ID youth. To this end, an accelerated longitudinal design was chosen for the advantageous ability to assess within-subject change while covering the entire period of adolescence through overlapping cohorts (Russell et al., 2021). Adolescents with IDs (n = 65) and typically developing youth as the comparison group (CG; n = 74) completed both an implicit and explicit emotion regulation task while undergoing functional magnetic resonance imaging (fMRI) at baseline and 1-year follow up. We investigated group-related changes across task performance and brain activation in the frontal lobe, amygdala, and hippocampus. Where group differences were identified, an exploratory investigation on the potential moderating influence of adversity was assessed within ID youth. We hypothesized that in CG, age would be associated with greater differentiation between negative and neutral stimuli in salience-related regions, reflecting the normative decrease in emotional reactivity to neutral stimuli. In contrast, ID youth would show little to no changes in valence differentiation across age, reflecting the greater salience of neutral stimuli in this population. Moreover, we hypothesized that CG will show greater activation differences between emotional reactivity and regulation as youth age in regions associated with cognitive control (e.g. dlPFC), whereas this pattern would not be observed in ID. Lastly, we hypothesized that any group differences will be exaggerated in ID youth with greater adversity exposure. Methods and materials Data were drawn from a parent study examining cognitive-emotional development across adolescence. A comprehensive description of study methods have been published elsewhere (Karpov et al., 2026), and a selective overview of methods relevant to the hypotheses are provided here. Participant recruitment A sample of 139 youth (aged 10–17 years at baseline, inclusive) were recruited to complete this two-wave, accelerated longitudinal neuroimaging study of emotion processing at baseline and a one-year follow-up. Typically developing youth and youth with a current ID (e.g., major depressive disorder, generalized anxiety disorder, PTSD) were recruited. Noteworthy exclusion criteria included ongoing trauma exposure, past/present diagnosis of psychosis or bipolar disorder, current substance abuse disorder, and/or acute suicidality. Use of psychotropic medications was not exclusionary. Within the typically developing comparison group (CG), a history of trauma was exclusionary. Written informed consent from a parent or legal guardian along with youth assent were obtained from all participants. All study procedures were approved by the University of Wisconsin-Madison Health Sciences Institutional Review Board. Clinical and behavioral assessments Eligibility was determined via comprehensive trauma and diagnostic assessment performed at an initial intake visit. The presence of past or current mental health disorders was assessed by trained study staff using the clinician-administered Kiddie Schedule for Affective Disorders and Schizophrenia (Kaufman et al., 1997, Townsend et al., 2020) using DSM−5 criteria. Final diagnostic decisions were determined during a consensus interview and approval by a board-certified child and adolescent psychiatrist (RJH) or qualified designee. Additional at-home questionnaire assessments were completed by youth (via RedCAP; Harris et al., 2009, Harris et al., 2019) within 2 weeks of neuroimaging procedures. These included a measure of cumulative adversity exposure (Stressful Life Events Schedule [SLES]; Williamson et al., 2003), psychopathology symptom severity (Child Behavior Checklist – Youth Self Report [CBCL-YSR]; Achenbach, 1999), and the difficulty with emotion regulation scale (DERS; Gratz and Roemer, 2004). Emotion regulation tasks At each session, participants completed both an implicit and explicit emotion regulation task while undergoing fMRI (task order randomized) to assess emotion processing (i.e., emotional reactivity and regulation) across different regulation strategies. Implicit emotion regulation was indexed using the emotional N-Back (EN-Back; Casey et al., 2018) task (Fig. 1A). Angry and neutral faces (NimStim (Tottenham et al., 2009) and RADIATE (Conley et al., 2018) stimuli sets) were used as negative and neutral stimuli (2 s; ITI, 500 ms), respectively. Participants were instructed to identify whether the current stimuli matched a target face (0-Back), or matched the stimulus that was presented two trials prior (2-Back). This task assesses both reactivity and regulation via differences in neural activity between negative and neutral stimuli (Gaillard et al., 2025, Geckeler et al., 2022, O’Brien et al., 2020, Ojha et al., 2024). Regulation is commonly indexed with the 2-Back condition (e.g., Bertocci et al., 2012, Ojha et al., 2024), or by collapsing across cognitive loads (e.g., Gaillard et al., 2025, Geckeler et al., 2022, O’Brien et al., 2020). However, differences in cognitive loads can differentially engage emotion-related networks (Luo et al., 2014, Sun et al., 2020), and thus we use the 0-Back condition to assess a lower implicit regulation state. Behavioral measures include reaction time (RT) and percent accuracy. A total of two blocks (10 trials per block) for each stimulus/instruction combination was presented. Task design and Age Characteristics. (A) Emotional N-Back task, indexing implicit emotion regulation. Top row, example 0-Back instruction block with negative stimuli. Bottom row, example 2-Back instruction block with neutral stimuli. (B) Perspective task, indexing explicit emotion regulation. Top row, example neutral-far trial. Bottom row, example negative-close trial. (C) Age characteristics of the cohort. CG - comparison group; ID - internalizing disorder (group); ISI – inter-stimulus interval; ITI – inter-trial interval. Explicit emotion regulation was indexed using the Perspective Task (Fig. 1B), an adapted-for-youth cognitive reappraisal task (Silvers et al., 2012, Silvers et al., 2015, Silvers et al., 2017a). Participants were instructed to either down-regulate and emotionally disengage from the stimuli (cued with the word “Far”), or focus on experiencing their emotional reactions and thus emotionally engage with stimuli (cued with “Close”) (Braunstein et al., 2017). Here, close trials are used to assess emotional reactivity and far trials for regulation (Silvers et al., 2012, Silvers et al., 2015, Silvers et al., 2017a). Negative and neutral stimuli were taken from the International Affective Picture System [IAPS] (Lang et al., 2005). After each stimulus presentation (8 s; mean ITI, 3 s) in a given trial, youth rated their current emotional affect on a Likert scale of −5–5. A total of 16 trials per stimulus valence-instruction pairing (negative-close, neutral-close, negative-far, neutral-far) were presented across four blocks. Magnetic resonance imaging MRI acquisition and preprocessing is detailed elsewhere (Karpov et al., 2026). The average blood-oxygen-level-dependent (BOLD) activity within select a priori regions were extracted from participants first level models for each stimulus valence-instruction pairing. Cortical regions-of-interest (ROIs) were taken from the Schaefer Atlas (400 parcellations, 7 network (Schaefer et al., 2018)), including all frontal lobe regions except those associated with the somato-motor network (126 ROIs). Six additional regions (for a total of 132) were taken from the Tian subcortical atlas (Tian et al., 2020), including the left and right the amygdala, anterior and posterior hippocampus. As regulatory processes are known to engage regions outside of the frontal lobe, results of a whole-brain analysis are reported in the supplementary material (Tables S10-S14, and Figures S7,S8). In order to preserve both statistical power and the ability to inferentially interpret developmental differences between distinct types of emotion regulation (implicit vs. explicit), we chose to utilize similar contrasts across tasks. We extracted patterns of neural activation to emotionally salient negative [Neg] stimuli controlling for how they react to neutral images [Neu] (e.g., Geckeler et al., 2022, McRae et al., 2012, Ojha et al., 2024, Silvers et al., 2015) for each task instruction (implicit regulation task: 0-Back, 2-Back; explicit regulation task: Close, Far). Reactivity was assessed using second-order contrasts to detect differences in BOLD activation between task instructions in our statistical models (see Section 2.5.), using 0-Back[Neg-Neu] > 2-Back[Neg-Neu] during the implicit regulation task and Close[Neg-Neu] > Far[Neg-Neu] during the explicit regulation task. Regulation was interrogated using 2-Back[Neg-Neu] > 0-Back[Neg-Neu] for implicit regulation and Far[Neg-Neu] > Close[Neg-Neu] for explicit regulation. Analyses using the standard approach to characterize explicit regulation (Neg[Far-Close], e.g. McRae et al., 2012, Silvers et al., 2015) are also reported in Supplementary Table S9. Statistical analysis Statistical analyses were performed using Mplus (v.8.11; Muthén and Muthén, 2017) and batch processed in R (v.4.4.0; R Core Team. 2024) with the package MplusAutomation (v1.1.1; Hallquist and Wiley, 2018). For each task, a series of linear mixed-effects models predicting BOLD activation were run across ROIs, and three additional models predicting behavior (implicit regulation task: RT, accuracy; explicit regulation task: affective rating). To facilitate interpretation across modalities and tasks, all models contained a three-way interaction between group (CG, ID), age, and task instruction (implicit regulation task: 0-Back, 2-Back; explicit regulation task: Close, Far), with subject included as a random effect and biological sex as a covariate. The lower bound of the age distribution (i.e., the youngest participant) was set to zero to provide a meaningful intercept (Hoffman, 2015). Follow-up decomposition analyses described the differential effects of negative and neutral stimuli, modelled by interacting a stimulus valence variable to the effect of interest. Correction for multiple comparisons used the False Discovery Rate (FDR) method and was applied separately to implicit regulation task BOLD, explicit regulation task BOLD, and behavioral outcomes. Group main effects and group-age interactions that survive correction are reported here (padj <.05). Age effects have been previously reported (Karpov et al., 2026), and thus results from the current study are provided in the supplementary materials along with any additional statistical outcomes not reported in the primary text. Standardized parameter estimates (β) are reported with their 95% confidence intervals (CI). Sensitivity analyses with psychiatric medication status as a covariate did not show meaningful differences (supplementary Table S8). Post hoc analyses. To explore the potential moderating effect of adversity exposure in ID youth, post hoc analyses were run within the dependent variables that were found to have significant group-related effects. As a history of trauma was exclusionary for the CG group, only ID youth were included in this analysis. The same variables were included as in the linear mixed-effect models used in the primary analysis, except the variable of ‘group’ was replaced by cumulative adversity exposure (SLES score), with the added fixed effect of psychopathology severity (CBCL-YSR total score) to control for psychiatric symptoms. CBCL-YSR and SLES scores were mean-centered. Missing values in the questionnaire data were minimal and accounted for by using Full Information Maximum Likelihood estimation. To preserve statistical power, the model only probed the most complex significant interaction and retained the remaining variables as fixed effects. The predicted slopes of low adversity (−2SD SLES score) ID youth were compared to those with high adversity (+2 SD score) to interpret adversity by age interactions. Open science statement Hypotheses and analysis plans were preregistered on OSF (Karpov et al., 2025). Analysis scripts are available at the following Github repository: https://github.com/BRAVEResearchCenter/CEDA/tree/main/KarpovEtAl2026 Results Participant characteristics Group demographics at baseline, including age (Fig. 1C), are summarized in Table 1. There were no significant differences between the groups for sex, age, race, ethnicity, socioeconomic status, nor for the length of time between sessions. As expected, ID youth exhibited higher self-reported psychopathology symptom severity, cumulative adversity exposure, and difficulty with emotion regulation than CG youth. Within ID youth, 35% (n = 23) were on psychiatric medication. Longitudinal participant retention between sessions was inhibited by pandemic-related difficulties and funding constraints, resulting in a large proportion of youth who were not invited to participate in the follow-up session (follow-up nCG = 66 [89%]; nID = 33 [51%]). There were no demographic differences between retained and non-retained ID youth (supplementary Table S1). Table 1Sociodemographic characteristics at baseline.Internalizing Disorder GroupaComparisonGroupBetween-Group StatisticsTest-statistic, p-valueN6574Sex (Female/Male)n = 43/22n = 37/37χ(1) = 3.06.08Age(Years)13.25 ± 2.02(10.2 – 17.2)13.35 ± 1.89(10.2 – 17.2)t(131.78) = −0.3.76RaceAsian (n = 1)Black (n = 4)Multiracial (n = 5)White (n = 52)Unknown (n = 3)Asian (n = 3)Black (n = 0)Multiracial (n = 4)White (n = 67)Unknown (n = 0)χ(2) = 0.83.36EthnicityHispanic (n = 9)Not Hispanic (n = 55)Unknown (n = 1)Hispanic (n = 4)Not Hispanic (n = 70)χ(2) = 2.08.15Socioeconomic status(Annual family income)$0 - $24,999 (n = 1)$25,000 - $49,999 (n = 2)$50,000-$99,999 (n = 21)$100,000-$199,999 (n = 18)$200,000 or higher (n = 7)Missing (n = 16)$0 - $24,999 (n = 0)$25,000 - $49,999 (n = 2)$50,000-$99,999 (n = 10)$100,000-$199,999 (n = 29)$200,000 or higher (n = 11)Missing (n = 21)χ(5) = 9.29.16Psychopathology Symptom Severity(CBCL-YSR)59.25 ± 25.29(0 – 114)18.06 ± 13.92(0 – 66)t(81.1) = 10.9< .001Adversity Exposure(SLES)37.58 ± 23.07(0 – 111)18.51 ± 12.86(0 – 49)t(68.4) = 5.2< .001Difficulty with ER(DERS)97.16 ± 22.09(55 – 160)62.86 ± 17.13(41 – 131)t(87.9) = 9.22< .001Time between sessions (years)1.47 ± 0.59(0.8 – 3.1)1.28 ± 0.47(0.9 – 3.1)t(55.7) = 1.66.10Continuous variables are listed as mean ± SD (range). Note that Hawaiian/Pacific Islander and Native American/Alaskan Native were provided as options for race, however all individuals who identified with one of those labels also identified with another race and were thus categorized as ‘multiracial’. a Details regarding clinical diagnoses are provided in the supplement (Figure S1). CBCL-YSR: Child Behavioral Checklist - Youth Self Report; DERS: Difficulty with Emotion Regulation Scale; SLES: Stressful Life Events Schedule. Sociodemographic characteristics at baseline. Continuous variables are listed as mean ± SD (range). Note that Hawaiian/Pacific Islander and Native American/Alaskan Native were provided as options for race, however all individuals who identified with one of those labels also identified with another race and were thus categorized as ‘multiracial’. a Details regarding clinical diagnoses are provided in the supplement (Figure S1). CBCL-YSR: Child Behavioral Checklist - Youth Self Report; DERS: Difficulty with Emotion Regulation Scale; SLES: Stressful Life Events Schedule. Groups differences during the implicit emotion regulation task Behavior A main effect of group was observed for accuracy (β = −0.951, CI95% = [−1.811, −0.092], z = −2.169, padj =.045), such that the ID group showed a larger difference in accuracy between negative and neutral stimuli (meanΔ = −3.00%, SE = 0.65) relative to CG (meanΔ = −1.80%, SE = 0.42; Fig. 2A). No significant effects in RT were detected (padj >.05). EN-Back task group results. (A) Accuracy (% difference, negative – neutral stimuli) across groups. Means are denoted with a square point in the boxplot, the dotted line highlights the difference in means. (B) ROIs with a significant main effect of group, plotted on an inflated brain parcellated with the Schaefer Atlas. Effect size is represented by the absolute z-score, blue gradient. (C) Left ventral anterior insula (Schaefer 143; red region) three-way interaction between group, age, and task instruction. Shaded region encompases the 95% confidence interval around the predicted slope. Significance markers denote simple slope effects, indicating whether the predicted slope is significantly different from zero. * p < .05. CG: comparison group; ID: internalizing disorder (group); ns: non-significant. Differential patterns in neural activation Eleven regions showed a main effect of group (Table 2, Fig. 2B), including the left insula, bilateral cingulate cortex, dorsomedial PFC, and lateral PFC. Within each region, CG showed greater differential activation between negative and neutral stimuli relative to ID. Table 2ROIs with a group main effect during the EN-Back task.Schaefer LabelRegionβz-score95% CIp.adjIDPredictedΔCGPredicted Δ143LH Insula−0.989−3.472[−1.547, −0.431].0060.0830.557347RH Triangular IFG−0.987−3.617[−1.522, −0.452].0040.1410.311311RH ACC−0.982−5.605[−1.326, −0.639]< .00010.0340.200107LH Cingulate cortex−0.974−8.095[−1.21, −0.739]< .0001−0.0280.200293RH Opercular IFG−0.965−14.109[−1.099, −0.831]< .00010.1220.455314RH Superior motor area−0.926−3.866[−1.395, −0.456].002−0.0390.107350RH Opercular IFG−0.915−5.683[−1.23, −0.599]< .00010.3090.545141LH Precentral sulcus−0.866−4.258[−1.265, −0.467].00030.2340.476307RH Rolandic Operculum0.9279.92[0.744, 1.11]< .0001−0.140−0.396175LH Opercular IFG0.9317.658[0.692, 1.169]< .0001−0.085−0.305318RH Superior frontal gyrus0.95811.959[0.801, 1.115]< .0001−0.204−0.942317RH SMA1.0025.801[0.663, 1.34]< .0001−0.060−0.231Region labels taken from the AAL3 v1 Atlas. Reported estimates (β) are standardized, while the predicted Δ (negative – neutral) are unstandardized. ACC – anterior cingulate cortex; CG – comparison group; FEF – frontal eye fields; ID – internalizing disorders (group); IFG – Inferior Frontal Gyrus; LH – Left Hemisphere; RH – Right Hemisphere; SMA – Supplementary Motor Area. ROIs with a group main effect during the EN-Back task. Region labels taken from the AAL3 v1 Atlas. Reported estimates (β) are standardized, while the predicted Δ (negative – neutral) are unstandardized. ACC – anterior cingulate cortex; CG – comparison group; FEF – frontal eye fields; ID – internalizing disorders (group); IFG – Inferior Frontal Gyrus; LH – Left Hemisphere; RH – Right Hemisphere; SMA – Supplementary Motor Area. Differences in neurodevelopmental trajectories The left ventral anterior insula (Schaefer 143, Brodmann Area [BA] 47/48) showed a significant group by instruction by age interaction (β = −1.156, CI95% = [−1.683, −0.628], z = −4.294, padj. =.002). Simple slope analysis confirms that within CG, insula activity increasingly differentiates between negative and neutral stimuli across age during low cognitive loads (Fig. 2C). In contrast, the ID group does not show differential activity between stimuli valences, with no changes across age for either instruction (p > .05, supplementary Table S3). Group differences during the explicit emotion regulation task Behavior Affective rating showed an expected main effect of instruction (β = 0.699, CI95% = [0.414, 0.983], z = 4.818, padj <.001), indicating that the difference in ratings between negative and neutral stimuli was smaller when instructed to emotionally disengage (Far; meanΔ = −1.70, SE = 0.08) compared to emotionally engage (Close; meanΔ = −2.56, SE = 0.07). A significant main effect of group was detected (β = 0.525, CI95% = [0.158, 0.892], z = 2.802, padj =.015), where ID youth showed a smaller difference (meanΔ = −2.00, SE = 0.09) relative to CG (meanΔ = −2.22, SE = 0.07; Fig. 3A). Perspective task group results. (A) Affective rating difference (negative – neutral) across groups. Means are denoted with a square point in the boxplot, the dotted line highlighting the difference in means. (B) ROIs with a significant main effect of group, plotted on an inflated brain parcellated with the Schaefer Atlas. Effect size is represented by the absolute z-score, blue gradient. (C) Left dorsal anterior insula (Schaefer 101; yellow region) three-way interaction between group, age, and task instruction. Shaded region reflects the 95% confidence interval around the predicted slope. Significant markers denote simple slope effects, indicating whether the predicted slope is significantly different from zero. *p < .05, **p < .01. CG – comparison group; ID – internalizing disorders (group). Differential patterns in neural activation A total of 28 regions showed a main effect of group (Table 3, Fig. 3B), in which CG showed greater differential activation across stimuli valence. This included the left OFC, right frontal pole, left insula, right dorsomedial PFC, and bilateral cingulate cortex. Table 3ROIs with a group main effect during the Perspective Task.Schaefer LabelRegionβz-score95% CIp.adjIDPredictedΔCGPredictedΔ134LH medial OFC−0.859−4.07[−1.273, −0.446].00040.4470.802167LH lateral OFC−1.015.629[−1.361, −0.658]< .00010.0560.225186LH MFG−0.8424.687[−1.194, −0.49]< .00010.0080.162185LH MFG−0.8393.619[−1.294, −0.385].0020.0380.152353RH Precentral Gyrus−0.8335.455[−1.132, −0.534]< .00010.1280.32886LH FEF−0.8245.701[−1.107, −0.541]< .00010.1180.269170LH Lateral OFC−0.8223.416[−1.293, −0.35].0040.0930.307324RH Frontal pole−0.7723.818[−1.168, −0.376].0010.3280.985314RH SMA0.7052.677[0.189, 1.222].037−0.083−0.224308RH Rolandic Operculum0.7063.235[0.278, 1.134].007−0.131−0.297147LH Cingulate cortex0.7392.711[0.205, 1.274].035−0.100−0.276183LH medial Superior Frontal Gyrus0.742.618[0.186, 1.294].042−0.041−0.224317RH SMA0.7723.391[0.326, 1.218].004−0.092−0.225316RH Paracentral Lobule0.7753.363[0.323, 1.226].005−0.176−0.278113LH SMA0.8083.595[0.368, 1.249].002−0.061−0.245189LH SMA0.8394.602[0.481, 1.196]< .0001−0.085−0.261115LH medial OFC0.8522.958[0.288, 1.417].0160.016−0.239391RH medial Superior Frontal Gyrus0.8542.597[0.21, 1.498].0430.030−0.08198LH Insula0.8574.819[0.508, 1.206]< .0001−0.062−0.162345RH MFG0.8727.281[0.637, 1.106]< .0001−0.198−0.414358RH Cingulate cortex0.8766.467[0.61, 1.141]< .0001−0.097−0.351104LH Rolandic Operculum0.88110.564[0.717, 1.044]< .0001−0.130−0.367303RH Insula0.8987.523[0.664, 1.133]< .0001−0.109−0.26089LH FEF0.9085.161[0.563, 1.253]< .00010.041−0.083102LH Insula0.9084.176[0.482, 1.335].0002−0.145−0.244103LH Opercular IFG0.91512.54[0.772, 1.058]< .0001−0.108−0.330101LH Insula0.99130.542[0.928, 1.055]< .0001−0.194−0.42899LH Insula0.9984.226[0.535, 1.461].0002−0.069−0.157Region labels taken from the AAL3 v1 atlas, except for the FEF which was taken from the Schaefer Atlas. Reported estimates (β) are standardized, while the predicted Δ (negative – neutral) are unstandardized. CG – comparison group; FEF – frontal eye fields; ID – internalizing disorders (group); IFG – Inferior Frontal Gyrus; LH – Left Hemisphere; MFG - Middle Frontal Gyrus; OFC – Orbital frontal cortex; RH – Right Hemisphere; SMA -Supplementary Motor Area. ROIs with a group main effect during the Perspective Task. Region labels taken from the AAL3 v1 atlas, except for the FEF which was taken from the Schaefer Atlas. Reported estimates (β) are standardized, while the predicted Δ (negative – neutral) are unstandardized. CG – comparison group; FEF – frontal eye fields; ID – internalizing disorders (group); IFG – Inferior Frontal Gyrus; LH – Left Hemisphere; MFG - Middle Frontal Gyrus; OFC – Orbital frontal cortex; RH – Right Hemisphere; SMA -Supplementary Motor Area. Differences in neurodevelopmental trajectories Two regions showed a 3-way interaction between group, age, and instruction. This was the left dorsal anterior insula (Schaefer 101, BA 48; β = 1.236, CI95% = [0.527, 1.945], z = 3.417, padj. =.040; Fig. 3C) and left Rolandic Operculum (Schaefer 104, BA 48; β = 1.342, CI95% = [0.747, 1.938], z = 4.418, padj. =.001). As confirmed by a simple slope analysis, both regions demonstrated a similar pattern (supplementary Figure S3, Table S3). While activity increasingly differentiated valence with age for all youth and task instructions, this increase was greater both in CG youth and during the emotional engagement instruction. Moderating effect of adversity exposure in ID youth Moderating behavior Controlling for psychopathology symptom severity, there were no significant effects of adversity on accuracy during the implicit regulation task (p > .05). For affective rating during the explicit regulation task, a significant main effect of adversity exposure was observed (β = 0.314, CI95% = [0.060, 0.569], z = 2.422, p = .015). As adversity exposure increased, the difference in affective rating between negative and neutral stimuli decreased (Fig. 4A). Valence decomposition analysis suggests that increasing adversity exposure decreases affective rating to neutral stimuli, but does not moderate rating of negative stimuli (supplementary Figure S4, Table S6). Moderating effect of adversity in youth with internalizing disorders during the Perspective task. (A) Affective rating with a main effect of adversity exposure. Individual raw difference scores are plotted with the predicted slope, where the shaded region reflects the 95% confidence interval. (B) ROIs with a main effect of adversity, plotted on an inflated brain parcellated with the Schaefer Atlas. Effect size is plotted as the z-score, blue colorbar. (C) Three-way interaction between adversity exposure, age, and task instruction in the dorsal anterior insula (Schaefer 101, yellow region). Predicted slope is plotted with the shaded region reflecting the 95% confidence interval. Significant markers denote simple slope effects, indicating whether the predicted slope is significantly different from zero. *p < .05; ns – non-significant. Moderating main effects of group in BOLD activation Of the 12 regions that displayed a main effect of group during the implicit regulation task (Table 2), none showed a significant main effect of adversity exposure. For the explicit regulation task, eight of the 25 regions with a group main effect (Table 3) had a significant main effect of adversity exposure (Fig. 4B; Table 4). Regions that displayed increasing valence discrimination with increasing adversity included the left insula (Schaefer 98), right supplementary motor area (Schaefer 317), left supplementary motor area (Schaefer 113), and right paracentral lobule (Schaefer 316). A more complex pattern was observed in the left frontal eye fields (FEF; Schaefer 86, 89), right medial superior frontal gyrus (Schaefer 391), and right middle frontal gyrus (Schaefer 345), where low and high adversity showed opposing patterns (supplementary Figure S5). Table 4ROIs with adversity main effects during the Perspective task in youth with internalizing disorders.Schaefer LabelRegionβz-score95% CIp86LH FEF−0.281−4.352[−0.408, −0.155]< .001317RH SMA−0.312−3.690[−0.477, −0.146]< .001316RH Paracentral Lobule−0.316−3.332[−0.502, −0.13]< .001113LH SMA−0.220−2.419[−0.398, −0.042].016391RH Superior Frontal Gyrus−0.247−3.427[−0.388, −0.106]< .00198LH Insula−0.222−2.563[−0.393, −0.052].010345RH MFG−0.217−2.639[−0.378, −0.056].00889LH FEF−0.251−2.885[−0.421, −0.08].004Region labels taken from the AAL3 v1 atlas, except for the FEF which was taken from the Schaefer Atlas. FEF – frontal eye fields; LH – Left Hemisphere; MFG - Middle Frontal Gyrus; RH – Right Hemisphere; SMA -Supplementary Motor Area. ROIs with adversity main effects during the Perspective task in youth with internalizing disorders. Region labels taken from the AAL3 v1 atlas, except for the FEF which was taken from the Schaefer Atlas. FEF – frontal eye fields; LH – Left Hemisphere; MFG - Middle Frontal Gyrus; RH – Right Hemisphere; SMA -Supplementary Motor Area. Moderating changes in developmental trajectories of BOLD activation During the explicit regulation task, a three-way interaction between instruction, age, and adversity exposure was observed the left dorsal anterior insula (Schaefer 101, BA 48; β = 1.902, CI95% = [0.832, 2.972], z = 3.484, p = .0005). This indicated that ID youth with high adversity exposure showed increasing differential activation with age during the emotional engagement instruction, whereas youth with lower adversity showed no significant changes (Fig. 4C, center). Decomposition analyses of stimulus valence revealed that in ID youth with low adversity exposure, dorsal anterior insula activity decreased across time for both negative and neutral stimuli, while youth with high adversity showed faster increases in activation with age to neutral as compared to negative stimuli (Fig. 4C, right). For the implicit regulation task, a two-way interaction between age and adversity exposure trended towards significance in the left ventral anterior insula (Schaefer 143, BA 47/48; β = −0.997, CI95% = [−2.019, 0.025], z = −1.913, p = .056). Here, ID youth with low adversity exposure seemed to exhibit greater discrimination with age, while youth with high adversity showed no-age related changes in discrimination (supplementary Figure S6). Discussion Elucidating the neural mechanisms underlying the development of emotion processes can facilitate improvements in treatment approaches for youth struggling with internalizing disorders. To this end, the current longitudinal study sought to characterize how neurodevelopmental trajectories of emotional reactivity and regulation are impacted by internalizing disorders and childhood adversity in adolescents. We report patterns of frontal lobe activation that differed between groups, where youth with internalizing disorders consistently showed less differential activation between negative and neutral stimuli. Our hypotheses regarding group differences in developmental trajectories were partially supported. Although regulation-related PFC activity did not change with age, we did observe differences in trajectories in the left anterior insular cortex (AIC) more closely associated with emotional reactivity. Specifically, trajectories differed in the left ventral AIC during the implicit regulation task, whereas the left dorsal AIC and Rolandic Operculum trajectories differed during the explicit regulation task. Lastly, exploratory analyses of adversity exposure identified a moderating effect primarily during the explicit regulation task in ID youth. These findings highlight the AIC as a developmentally sensitive region and potential biomarker for clinical interventions. The left AIC and Rolandic Operculum were identified as key regions whose development is altered in youth with IDs. Consistent with our prior work (Karpov et al., 2026), age appears to drive increasingly differentiated patterns of activation in response to negative and neutral stimuli, likely underlying a greater capacity for valence discrimination. In contrast, such differentiation was not observed among ID youth, suggesting alterations in systems relating to salience detection, diminished neural specificity, or affective blunting. Our finding that AIC trajectories between CG and ID youth increasingly diverge with age seems to mirror the strengthening association between negative interpretation bias and internalizing symptoms in the same developmental period (Songco et al., 2020, Stuijfzand et al., 2018), and may be one neural mechanism contributing to the negative interpretation bias observed in ID individuals. More broadly, these findings emphasize psychopathology as a developmental processes (Casey et al., 2014, Uhlhaas et al., 2023). Differences in slope or timing are crucial for characterizing neurodevelopmental patterns (Russell et al., 2021), and deviations from these trajectories are more likely to manifest as psychopathology than are categorical deficits (Casey et al., 2014, Schnack et al., 2015). The AIC is generally understood to be an emotionally reactive region which functions to integrate internal states for saliency and motivational processes (Craig, 2009, Etkin et al., 2015, Fermin et al., 2022, Uddin et al., 2017). Due to similarity in developmental trajectories and the robust connectivity with the insula (Mălîia et al., 2018), our Rolandic Operculum interpretations will be synonymous with the AIC for the purposes of this discussion. The lack of developmental change among ID youth could stem from an indiscriminate AIC resulting from impaired connectivity to the PFC. Typically developing youth display age-related decreases in AIC cortical thickness, decreases in the fiber density of neural tracts linking AIC-PFC, and increased functional connectivity between the two (Churchwell and Yurgelun-Todd, 2013, Dennis et al., 2014, Fair et al., 2007; J. Kim et al., 2025; LeWinn et al., 2018; Shaw et al., 2008), each a hallmark of improved communication efficiency. Conversely, there is evidence of increased gray matter density in the AIC of youth with IDs (Adams et al., 2024, Li et al., 2025, Mueller et al., 2013) and decreased AIC-PFC coupling (LeWinn et al., 2018). To note, studies reporting the opposite observation in ID youth (i.e. decreased gray matter volume (Snyder et al., 2017; Suffren et al., 2019; Townend et al., 2025)) often parcellate the entire insula as one ROI, which could wash out subregion-specific patterns. Insular subregions have been shown to be differentially sensitive to adversity, psychopathology, and other variables such as sex in youth (Heyn et al., 2022, Keding et al., 2021, Klabunde et al., 2017). Developmental studies would benefit from analyzing the insula by subregion to account for the heterogeneity of structure, connectivity, function, and developmental trajectories (Chang et al., 2013, Fermin et al., 2022, Menon, 2025, Schaefer et al., 2018, Shaw et al., 2008, Uddin et al., 2017), which may clarify inconsistencies. Differentiation across AIC subregions was observed in the current work, where developmental changes occurred in the ventral AIC during the EN-Back task and the dorsal AIC during the Perspective task. Despite their spatial proximity, the dorsal and ventral AIC have distinct structural and functional profiles. The agranular ventral AIC has unique connections with limbic regions such as the amygdala and vmPFC and is implicated in emotional motivational processes (Chang et al., 2013, Deen et al., 2011, Kelly et al., 2012, Menon, 2025, Molnar-Szakacs and Uddin, 2022, Nomi et al., 2018, Wager and Barrett, 2017). Conversely, the dysgranular dorsal AIC has unique connectivity with the dlPFC and is often associated with cognitive control related processes (Chang et al., 2013, Deen et al., 2011, Kelly et al., 2012, Menon, 2025, Molnar-Szakacs and Uddin, 2022, Nomi et al., 2018, Wager and Barrett, 2017). Our results indicate opposing trajectories in dorsal and ventral AIC, which likely stems from this heterogeneity and points to unique functions within emotion processing. Specifically, we report an increase in ventral AIC activity with age for CG youth during the low regulation condition (0-Back) of the implicit regulation task (supplementary Figure S2). Activity increased at a faster rate to neutral relative to negative stimuli, resulting in greater differentiation across stimuli valence (Fig. 2). Given the unique connectivity of the ventral AIC with the vmPFC, a region implicated in implicit regulation (Braunstein et al., 2017; Gyurak et al., 2011), this pattern of activity may reflect a pseudo-regulatory function whereby relatively greater regulation of neutral stimuli decreases the emotional reactivity during emotional perception. It is possible that the ventral AIC primarily engaged when there is little concurrent PFC activity related to cognitive control, evidenced by the lack of developmental change during the high cognitive load condition (2-Back). The AIC has been reported to show lower activation during the 2-Back condition relative to 0-Back in adults, posited to reflect the decreased neural resources available for emotional perception when there is a need for substantial executive control (Luo et al., 2014). In contrast, dorsal AIC activity decreased to neutral stimuli with age during the explicit regulation task (supplementary Figure S3), seemingly driving the increasing differentiation of valences (Fig. 3). This was most robust during the emotional engagement instruction, when youth were instructed to focus on their emotional experience. Such a pattern aligns with a function relating to salience or attention to interceptive states, likely indicating decreasing emotional reactivity with age. An alternative, non-mutually exclusive explanation involves the difference in stimuli used between the two tasks. In keeping with the original validated designs, emotional faces were used during the EN-Back task (Casey et al., 2018) whereas IAPS images were used during the Perspective Task (Silvers et al., 2012). Normative changes in sensitivity to social information occurs during adolescence (Andrews et al., 2021; Nelson et al., 2016; Somerville, 2013), and the social vs. non-social nature of the stimuli can lead to differences in neural activity unrelated to emotion processing (Quarmley et al., 2019). Future studies using consistent stimuli across tasks will be necessary to fully disentangle this regional specificity. Lastly, exploratory analyses indicated that adversity moderated ID youths’ emotional reactivity of the dorsal AIC during the explicit regulation task. Only a trending effect was seen during the implicit regulation task in the ventral AIC. While adversity has been shown to decrease implicit regulation function in youth (Kim et al., 2023; Lambert et al., 2017; Marusak et al., 2015), it is highly probable that this is associated with greater symptom severity (Compas et al., 2017, McLaughlin et al., 2011b). Given that the current study statistically controlled for psychopathology severity, our findings instead suggest that adversity exposure primarily increases emotional reactivity (Fig. 4). Here, ID youth exposed to lower levels of adversity show similar dorsal AIC developmental patterns to CG youth, as both show decreasing dorsal AIC activity with age. CG and low-adversity ID youth vary only in the ability of dorsal AIC activity to differentiate between stimuli valences. Unlike low-adversity ID and CG youth, high adversity ID youth showed increasing dorsal AIC activity with age. Notably, activity to neutral stimuli increased at a faster rate relative to negative stimuli. These results support the characterization of adversity-exposed ID youth as a clinically distinct phenotype (Teicher et al., 2022, Teicher and Samson, 2013), and build on literature showing subregion-specificity regarding the effects of adversity and trauma (He et al., 2022, Heyn et al., 2022, Hu et al., 2024, Keding et al., 2021, Klabunde et al., 2017).Overall, these post-hoc analyses provide preliminary evidence of the complex sequelae following adversity, underscoring the importance of investigating this variable in the context of pediatric psychiatry. Given the insula’s role in IDs and psychopathology more broadly (Goodkind et al., 2015, Namkung et al., 2017, Royer et al., 2024, Sha et al., 2019, Shanmugan et al., 2016), it has been proposed that interventions should focus on modulating insular activity (Downar et al., 2016, Ibrahim et al., 2019). In adults, conventional treatments such as pharmaceuticals, behavioral and exposure therapies alter AIC function (Gorka et al., 2019, Karch et al., 2024, Lee et al., 2024, Leroy et al., 2022, You et al., 2024), while direct neuro-feedback and stimulation targeting the insula have demonstrated promising clinical outcomes (Moraga-Amaro et al., 2023, Struckmann et al., 2022, Zhang et al., 2025, Zilverstand et al., 2015). Our results suggest that treatments aiming to alter emotion processing would benefit from considering individual trajectories of AIC maturation in adolescents. While preliminary evidence of neurofeedback training modulating dorsal AIC activity has been reported in youth during explicit regulation (Cohen Kadosh et al., 2016), age may moderate the long-term effectiveness of such interventions; this is exemplified by the increased benefits of cognitive reappraisal in older adolescents relative to younger adolescents (Silvers et al., 2015). Furthermore, our exploratory finding relating adversity to altered development of the dorsal AIC in ID youth may act as an additional confounding factor. Integrating individual variation in neural maturation (Russell et al., 2021) by assessing neurodevelopmental progress holds potential for personalizing pediatric clinical treatments. While promising, this work should be carefully interpreted within the context of its limitations. First, the moderate sample size may have limited our statistical precision, while also restricting available statistical approaches. For example, we were unable to model interactions of biological sex, adversity subtypes, or timing of adversity experiences that may elucidate sensitive periods of development. Data collection across two timepoints precludes modeling slope random-effects, further decreasing sensitivity to individual developmental trajectories. Second, attrition can bias longitudinal neuroimaging estimates (Matta et al., 2018), which is of particular relevance for our ID sample. While the majority of follow-up data was missing completely at random (supplementary Table S1), nonetheless the smaller sample decreases the precision of age-related estimates of within-subject developmental trajectories. Lastly, the heterogeneity of stimuli across tasks presents an important confound, as described in detail earlier. Altogether, future studies with larger samples, greater number of follow-up sessions, and consistent stimuli is warranted to replicate and expand on our results. In summary, this study is the first (to our knowledge) to use an accelerated longitudinal design probing the neurodevelopmental trajectories of emotion processes across both implicit and explicit regulation strategies in a clinical sample of adolescents. We provide evidence of functional differentiation within the AIC, and preliminary evidence that adversity moderates emotional reactivity during adolescence. These findings highlight the importance of incorporating longitudinal trajectories, adversity exposure, and different components of emotion processing in neurodevelopmental research. Such advances can inform clinical interventions, allowing for more effective treatments in youth. CRediT authorship contribution statement Ryan J. Herringa: Writing – review & editing, Supervision, Funding acquisition, Conceptualization. Sara A. Heyn: Writing – review & editing, Conceptualization. Justin D. Russell: Writing – review & editing, Methodology. Galit Karpov: Writing – review & editing, Writing – original draft, Visualization, Formal analysis. Brian W. Kleinschmidt: Writing – original draft. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements This work was supported by the National Institutes of Health/National Institute of Mental Health (Grant No. R01 MH115910 [to RJH]), as well as the University of Wisconsin School of Medicine and Public Health and Department of Psychiatry. We would like to thank Samantha DiMaio for organizing, managing, and maintaining the data used in the manuscript. We owe our sincerest gratitude to the youths and families who have given their time for this study. Footnotes Supplementary data associated with this article can be found in the online version at doi:10.1016/j.dcn.2026.101769. Supplementary material Supplementary material Data availability The data that has been used is confidential. References Associated Data Supplementary Materials Supplementary material Data Availability Statement The data that has been used is confidential.