Childhood Threat Exposure and Poor Emotional Awareness Predict Neural Correlates of Emotion Regulation in Adolescent Girls

Abstract

Increasing evidence suggests that early life adversity characterized by threat disrupts amygdala connectivity to the prefrontal cortex (PFC), a neural circuit integral to adaptive emotional functioning. Yet, few empirical studies have investigated how adversity predicts amygdala connectivity while participants intentionally try to downregulate their emotions. Here, we examined (a) associations between dimensions of adversity (i.e., threat and deprivation) and amygdala connectivity during a cognitive reappraisal task and (b) whether threat exposure interacts with emotional awareness- a capacity that facilitates emotion regulation- to predict amygdala connectivity. We conducted secondary analyses of neuroimaging data from one hundred and twenty-five adolescent females aged 9– 17 years old (M_age = 12. 77 years, SD = 1. 94 years). Whole-brain activation analyses revealed that greater threat exposure predicted less activation of the anterior cingulate gyrus during cognitive reappraisal. Amygdala connectivity during reappraisal did not vary as a function of threat or deprivation exposure; however, threat exposure interacted with low emotional awareness to predict coupling between the amygdala and vmPFC during emotion regulation. Specifically, youth with relatively high threat exposure and poorer emotional awareness exhibited the strongest amygdala-vmPFC connectivity. The results of this interaction were similar after adjusting for co-occurring deprivation exposure. In contrast, deprivation exposure, when controlling for threat, was unrelated to neural activation and amygdala connectivity during reappraisal. These findings contribute to the growing evidence that threat exposure is uniquely associated with neural correlates of emotion functioning and suggest that poor emotional awareness may exacerbate threat-related differences in neural systems underlying effortful emotion regulation.

Emotion regulation, or processes related to monitoring, evaluating, and modifying one’s emotional reactions to accomplish one’s goals, is fundamental to healthy emotional and social development (Thompson, 1991). Children and adolescents who struggle to manage negative emotions are at heightened risk for several forms of psychopathology (e.g., Beauchaine, 2015; Compas et al., 2017; McLaughlin et al., 2011) and impaired social functioning (e.g., Eisenberg et al., 2006; English et al., 2012; Kim & Cicchetti, 2010), whereas those with stronger emotion regulation abilities report greater peer acceptance and family satisfaction (Chervonsky & Hunt, 2019; Perry-Parrish et al., 2017). A well-established risk factor for difficulties in emotion regulation is early life adversity (Gruhn & Compas, 2020; Lavi et al., 2019; Miu et al., 2022). Early life adversity (ELA) affects approximately half of adolescents in the United States (McLaughlin et al., 2012), and the majority of people experience not one but multiple, often co-occurring types of adversity (Kessler et al., 2010; McLaughlin et al., 2012). Early life adversity (ELA) is a robust predictor of adverse developmental outcomes (Evans et al., 2013; Wade et al., 2022; Lambert et al., 2017), including difficulties in emotion regulation. Increasingly, scholars argue that understanding the mechanisms through which adversity shapes emotion regulation requires accounting for the differential effects of distinct, co-occurring adversities (Milojevich et al., 2021). Building on the Dimensional Model of Adversity and Psychopathology (McLaughlin, Sheridan, et al., 2014; McLaughlin & Sheridan, 2016; Sheridan & McLaughlin, 2014), we examine how two dimensions of early life adversity—threat and deprivation—relate to brain activation and amygdala–prefrontal connectivity during cognitive reappraisal in adolescent girls, and test whether individual differences in emotional awareness help explain variability in these associations.

Dimensions of early life adversity and the amygdala

Meta-analytic evidence suggests that children exposed to adversity are more likely to report problems managing negative affect and to rely on less effective regulation strategies, such as disengagement, expressive suppression, and rumination, while using typically effective strategies like cognitive reappraisal less frequently than their non-exposed peers (Gruhn & Compas, 2020; Lavi et al., 2019). Much of this work has relied on a cumulative-risk approach, which assumes that distinct forms of adversity exert additive effects on development and are interchangeable in their consequences (e.g., Evans et al., 2013). Although this approach has been instrumental in documenting the breadth of adverse outcomes associated with ELA, it offers limited insight into the specific mechanisms through which different types of experiences shape emotion regulation.

Dimensional models of adversity address this limitation by positing that ELA can be organized into core dimensions that cut across specific experiences and are linked to partially distinct patterns of emotional, cognitive, and neurobiological development (Ellis et al., 2022; McLaughlin et al., 2021). One of these models, the Dimensional Model of Adversity and Psychopathology, identifies two distinct dimensions, deprivation and threat (McLaughlin et al., 2014, 2021; Sheridan & McLaughlin, 2014). Deprivation refers to experiences in which expected cognitive and social inputs are absent (e.g., neglect, institutional rearing, poverty-related reductions in stimulation). In contrast, threat refers to experiences in which harm or threat of harm to the child or close others is present (e.g., physical or sexual abuse, witnessing domestic or community violence). Converging evidence suggests that deprivation is associated with alterations to neural systems involved in higher order cognition including frontoparietal regions (Lurie et al., 2024; Rosen et al., 2018; Sheridan et al., 2017, 2022), whereas threat is associated with heightened responsivity and altered plasticity in neural systems involved in identifying and responding to potential dangers in the environment, including salience network regions (Hein et al., 2020; Machlin et al., 2019; McLaughlin et al., 2015; Puetz et al., 2020). The distributed “salience network” (consisting of the amygdala, dorsal and ventral anterior cingulate cortex, anterior insula, ventral striatum and thalamus) supports the brain’s ability to tune attention and behavior towards stimuli that may have affective relevance to a person (Lindquist & Barrett, 2012).

The amygdala, in particular, has received considerable attention in research on childhood adversity exposure due to its role in the detection of threat (Fareri & Tottenham, 2016; McLaughlin et al., 2019; Tottenham & Sheridan, 2010). More generally, the amygdala is responsive to motivationally relevant or salient cues in the environment (Cunningham & Brosch, 2012; LeDoux, 2007; Lindquist et al., 2012). As a result, exposure to threat is hypothesized to result in heightened amygdala activity to salient information (McLaughlin et al., 2014, 2021; Sheridan & McLaughlin, 2014). Indeed, prior work suggests that childhood threat exposure predicts differential amygdala activity in response to negative stimuli, although initial studies examined exposure to violence or abuse specifically, rather than treating threat as a continuous dimension consisting of various multiple exposure types (McLaughlin et al., 2019). These studies consistently find that threat exposure predicts greater amygdala activation in response to prototypically negative facial expressions (Marusak et al., 2015a; McCrory et al., 2011, 2013; McLaughlin, Busso, et al., 2014; Suarez et al., 2024; White et al., 2019) and more sustained amygdala activation (i.e., less habituation) in response to repeated aversive relative to non-aversive cues (DeCross et al., 2022; Hein et al., 2020; Stevens et al., 2023). While heightened sensitivity to salient environmental cues may be beneficial in threatening environments, it is maladaptive in non-threatening contexts (DeCross et al., 2022; Machlin et al., 2019). Indeed, it is thought that childhood threat exposure can result in long-term difficulty regulating emotional responses (Gruhn & Compas, 2020; Milojevich et al., 2021). Although these prior studies provide preliminary support for dimensional models, many studies examining threat exposure and amygdala functioning fail account for co-occurring experiences of other adversities, like deprivation (e.g., Marusek et al., 2015; McCrory et al., 2011, 2013; McLaughlin et al., 2014; Stevens et al., 2023). Consequently, it remains unclear whether heightened neural responsiveness to negative stimuli is associated with threat but not with deprivation-related experiences.

Early life adversity and neural correlates of emotion regulation

Amygdala–prefrontal circuitry has been identified as a plausible pathway through which ELAs may confer risk for later difficulties in emotion regulation (Gee, 2021; Tottenham, 2020). The amygdala has rich structural and functional connections to the prefrontal cortex (PFC), including the ventromedial PFC (vmPFC). The vmPFC plays a key role in integrating information from affective sensory and social cues, long-term memory, and representations of the “self” (Roy et al., 2012; Shenhav et al., 2013). Among non-adversity exposed populations, connectivity with prefrontal regions is thought to modulate the amygdala’s reactivity to affective stimuli by the mid-to-late stages of puberty (Silvers, 2022). Specifically, it is believed that negative connectivity between the vmPFC and amygdala (i.e., greater vmPFC and lesser amygdala activity) reflects the vmPFC’s ability to adaptively inhibit continued processing of uncertain or salient stimuli (Motzkin et al., 2015; Tottenham & Gabard-Durnam, 2017). Prior research has suggested that amygdala hyperactivity observed in maltreated youth may be the result of altered amygdala-vmPFC connectivity when viewing negative stimuli (e.g., Callaghan & Tottenham, 2016), although the direction of effects and the consequences of this alteration are not consistent across the literature. For instance, there is some evidence that childhood adversity characterized by threat is associated with positive amygdala-vmPFC coupling (i.e., fluctuations of the blood-oxygenation-level dependent (BOLD) signal in both regions are positively correlated) (Herringa et al., 2016; Marusak et al., 2015b). Other studies have found that threat is associated with negative amygdala-vmPFC connectivity (Peverill et al., 2019), where spontaneous BOLD fluctuations in the two regions are inversely correlated (Liang et al., 2012). At least one recent study did not find task-related amygdala connectivity differed as a function of threat (Weissman et al., 2020). Collectively, this evidence suggests that early life threat exposure may be associated with atypical amygdala-vmPFC connectivity during emotion processing, but the specific pattern of altered amygdala-vmPFC coupling is less clear. Moreover, this work largely examines the effects of adversity on neural recruitment during tasks where there is no explicit goal to regulation emotion; less is known about the effects of adversity on amygdala activity and connectivity while participants intentionally regulate their emotional responses.

The ability to engage in goal-directed emotion regulation develops from early childhood through emerging adulthood (Fombouchet et al., 2023; Silvers et al., 2012) and is crucial to adaptive coping, social relationships, and well-being (Gross, 2002; Daniel, Abdel-Baki & Hall, 2020). Cognitive reappraisal, a goal-directed strategy which involves changing the meaning of a stimulus to change its emotional impact (Gross, 2015), is a particularly useful way of regulating one’s emotion. Indeed, adolescents and young adults who use reappraisal more often report lower negative affect, reduced internalizing symptoms, and have stronger social connections (English et al., 2012; Lennarz et al., 2019; McRae et al., 2012; Shapero et al., 2019). Cognitive reappraisal is linked closely to executive functions such as cognitive control and cognitive flexibility (Guassi Moreira et al., 2020; Malooly et al., 2013; McRae et al., 2012); unsurprisingly, cognitive reappraisal is supported by neural systems that partially overlap with executive functions and these systems continue to develop until adulthood (Buhle et al., 2014; Pozzi et al., 2021; Silvers, 2020). Evidence from neuroimaging meta-analyses suggests that cognitive reappraisal success is associated with attenuated amygdala activation and increased recruitment of areas of the PFC, including regions involved in cognitive control such as the dorsolateral prefrontal cortex (dlPFC), dorsomedial prefrontal cortex (dmPFC), and dorsal anterior cingulate cortex (dACC) in adults (Buhle et al., 2014; Morawetz et al., 2020). Although activation of the vmPFC is not consistently observed in studies of cognitive reappraisal in adults, some scholars have proposed that activation of cognitive control regions during reappraisal may attenuate amygdala activity by acting on the vmPFC (Buhle et al., 2014; Diekhof et al., 2011; Silvers & Guassi Moreira, 2019). Moreover, inverse amygdala-vmPFC connectivity appears to support reappraisal across adolescence (Silvers et al. 2017).

While meta-analytic evidence suggests that youth exposed to threats are less likely to spontaneously use cognitive reappraisal compared to their non-exposed peers (Gruhn & Compas, 2020; Miu et al., 2022), only a handful of studies to our knowledge investigate links between threat exposure and patterns of neural activity during cognitive reappraisal (Jenness et al., 2021; McLaughlin et al., 2015; Rodman et al., 2019). In one study, McLaughlin and colleagues (2015) found that threat-exposed youth exhibited greater activation in the superior frontal gyrus, mPFC, and bilateral dACC during cognitive reappraisal compared to non-exposed youth (McLaughlin et al., 2015). Two studies using a largely overlapping sample also found evidence to suggest that greater PFC recruitment in adversity-exposed youth may support reappraisal. Jenness and colleagues’ (2021) did not find a main effect of threat exposure on neural activation. However, older compared to younger threat-exposed youth exhibited increased recruitment of the left inferior frontal gyrus, left insula, and dACC; on the other hand, non-exposed youth displayed decreased activation in these regions as age increased. Rodman and colleagues (2019) also found no main effect of maltreatment history on neural activation; however, they did find that maltreated youth with greater, compared to less, superior frontal gyrus and right dACC activity during reappraisal reported lower depressive symptoms. These studies suggest that childhood adversity may be associated with differential recruitment of the prefrontal cortex during cognitive reappraisal. However, these studies do not address whether adversity exposure influences patterns of connectivity between the prefrontal cortex and other regions during reappraisal.

Emotional awareness

The degree to which adversity exposure affects neurobiological mechanisms underlying emotion regulation may also be influenced by an adolescent’s level of emotional awareness. Emotional awareness is a multidimensional construct that reflects the extent to which one understands and can describe their emotional experiences (Mankus et al., 2016). Those with low versus higher emotional awareness lack access to and understanding of their feelings across contexts (Boden & Thompson, 2015). Psychological constructionist models of emotion suggest that attending to and understanding one’s emotions allows a person to make more context-appropriate predictions about the nature of the situation at hand, resulting in more adaptive emotional responses (Hoemann et al., 2021). In non-adversity-exposed adolescents, greater emotional awareness predicts greater emotion regulation success (Vine & Aldao, 2014). Conversely, low emotional awareness is associated with increased difficulty regulating one’s own emotions (Boden & Thompson, 2015; Van Beveren et al., 2019) and appropriately responding to the emotions of others (Dickerson & Quas, 2021).

Early life threat adversity, emotional awareness, and neural connectivity

Prior research supports the possibility that emotional awareness moderates the effect of threat exposure on emotion regulation. Boden et al. (2012) found that the ability to attend to and understand one’s emotions buffered against the effects of trauma exposure in adults with PTSD, such that trauma-exposed adults with both high emotional awareness and high use of reappraisal reported lower PTSD-related symptoms and greater positive affect compared to those with poorer emotional awareness. Although there is very little neuroimaging work on emotional awareness, at least one study finds that emotional awareness is associated with increased functional connectivity between nodes of the salience network and between prefrontal regions during rest (Faulkner et al., 2022; although see Smith et al., 2017). Greater connectivity within these networks may reflect efficient information exchange to facilitate attention to incoming sensory information and conceptualization of that information.

The present study

In sum, multiple lines of evidence suggest that youth exposed to adversity, particularly experiences of threat, demonstrate significantly different patterns of amygdala and PFC activation compared to their non-exposed peers when responding to emotionally evocative images. Although this work suggests that the neural correlates of cognitive reappraisal may vary as a function of early adversity exposure, two important gaps in the literature remain. First, many studies examining emotion regulation in the context of adversity focus on the effects of maltreatment exposure (i.e., abuse and neglect) compared to non-exposure, which may obscure the effects of other prevalent forms of adversity, such as other forms of threat (i.e., peer victimization) and deprivation (i.e., poverty) (Miu et al., 2022); even fewer studies consider the differential impact that threat and deprivation may have on emotion regulation (Milojevich et al., 2021). As such, we still lack robust evidence that threat exposure, above and beyond deprivation uniquely predicts neural correlates of emotion regulation.

Second, while some studies of deprivation and threat assess emotion regulation using task-based or self-report measures of reappraisal (McLaughlin et al., 2015; Milojevich et al., 2021), most empirical evidence supporting dimensional models of adversity use paradigms where there is no explicit goal to change one’s emotions (e.g., Lambert et al., 2017; Sadikova et al., 2025; Schäfer et al., 2023; Sheridan et al., 2020). Therefore, we still lack robust evidence that threat exposure, above and beyond deprivation, uniquely predicts reappraisal or neural mechanisms that support reappraisal.

The goal of the present study was to investigate the effects of threat and deprivation on brain activation and amygdala functional connectivity during reappraisal. To address this research question, we ran secondary analyses on a sample of 138 9- to 17-year-old females from the Southeastern United States. First, to replicate prior research (Jenness et al., 2021; McLaughlin et al., 2015; Rodman et al., 2019), we examined the effects of threat, controlling for deprivation, on whole-brain activation during cognitive reappraisal. Second, we tested whether childhood threat exposure has unique associations with amygdala connectivity during cognitive reappraisal. Based on the extant research summarized above, we hypothesized that experiences of threat, controlling for deprivation, would be associated with significant differences in whole-brain functional connectivity of the amygdala, specifically the amygdala’s connectivity to the vmPFC, during reappraisal. However, we did not make directional hypotheses about the effect of threat on amygdala connectivity because of mixed findings from the literature. Given the relatively few studies that use a dimensional approach to examine the effects of adversity on neural activation during emotion regulation, we also examined the effect of deprivation, controlling for threat, on amygdala connectivity. Our final preregistered aim was to examine whether emotional awareness modulates the effect of threat-related adversity on amygdala connectivity while participants downregulate their responses to negative images. We predicted that lower insight into one’s own emotions may exacerbate the effects of threatening experiences on amygdala connectivity during reappraisal. Additionally, we examined associations between the effects of threat, controlling for deprivation, and emotional awareness on participants’ self-reported emotional intensity during the reappraisal task; however, we did not register hypotheses specific to the behavioral analyses.

Method

Transparency and Openness

The data for the present analyses were drawn from a project exploring biological and social factors contributing to risk for internalizing psychopathology and self-injurious thoughts and behaviors in adolescents assigned female sex at birth (e.g., Gruhn et al., 2024; Miller et al., 2024; Patel et al., 2023; Pelletier-Baldelli et al., 2024; Rodriguez-Thompson et al., 2024). This study was approved by the University of North Carolina at Chapel Hill Institutional Review Board and data were collected from 2016 to 2020.

Two prior studies using this data have published findings using the cognitive reappraisal task reported herein (Gruhn et al., 2024; Miller et al., 2024). One study investigated the effects of threat exposure on the association between cortisol reactivity to an acute social stress task and neural activation during emotional reactivity (Gruhn et al., 2024). Another study investigated the effects of targeted social rejection on neural activation during emotional reactivity and emotion regulation (i.e., neural recruitment before and after targeted social rejection) and examined significant group differences between adolescent females with and without a history of suicidal thoughts and behaviors (Miller et al., 2024).

The present study was post-registered (https://osf.io/jse65), meaning that the research questions and analysis plan, including how we determined data exclusions, were registered after data had been collected but before analyses were conducted (Benning et al., 2019). Information about data manipulations and exclusions can also be found in the Measures and Analysis Plan sections. The first authors (AB and AC) had access to the neural data, specifically contrasts of cognitive reappraisal (Decrease Negative > Look Negative) at the subject level and whole brain level, prior to post-registration; however, these models did not include the effects of threat or deprivation as regressors. The post-registration deviations from the original the post-registration (https://osf.io/jse65/files/cnaxf), and analysis code are available at https://osf.io/jse65/overview. Requests for the MRI data can be sent via email to the senior author (MS). Data were analyzed using the Functional Magnetic Resonance Imaging of the Brain Software Library (FSL) version 6.00 (Woolrich et al., 2009) and R version 4.2.0 (R Core Team, 2024).

Participants

Participants were 125 adolescent females drawn from a larger study examining risk factors for self-injurious thoughts behaviors. Youth assigned female sex at birth were recruited between 2015 and 2019 from a range of community and clinical placements, including inpatient psychiatric hospital units, outpatient mental health service agencies, and the local community, using flyers and social media advertisements. Inclusion criteria included being age 9–15 years old, a history of mental health concerns (i.e., affective disorders, anxiety, substance use disorders, or disruptive behavior disorders) in the past two years or prior life stress as reported by the participant’s caregiver. Participants were excluded if they were diagnosed with an intellectual disability, a pervasive developmental disorder, or were experiencing active psychosis. Of the 229 participants who enrolled in the larger study, 219 completed all laboratory procedures and were invited to take part in an optional MRI session. The MRI visit was introduced as a supplemental aim after the parent study had begun and thus was not included in the original consent. Participants who declined any further follow-up after the larger study (n = 1), endorsed left-handedness (n = 13), or reported MRI contraindications such as braces (n = 4) were not eligible for the MRI session, leaving 201 MRI-eligible youth. Among these 201 eligible participants, 63 either declined participation in the optional MRI visit, moved out of the area, or could not be recontacted. The remaining 138 youth attended the MRI session (see Figure S1 for participant flow diagram). On average, the MRI session took place 4.6 months (SD = 6.86) after the initial baseline session. Out of these 138 participants, 125 youth aged 9–17 years old (M age = 12.77 years old, SD =1.94 years) provided useable MRI data for the present analyses. See Analysis Plan for details regarding data exclusions.

Caregivers reported that 57 participants (45.6% of the sample with useable MRI data) were regularly taking psychotropic medications. Thirty-eight participants were taking an anti-depressant medication (30.4% of the sample). Sixteen participants were taking psycho-stimulant medication (10 of whom reported taking both anti-depressant and stimulant medication and 6 of whom reported only taking stimulant medication). Of the 16 youths taking stimulant medication, four reported taking stimulant medication on the day of the scan (3.2% of the sample), while 12 completed a 24-hour medication wash. Given that medication use is associated with neural correlates of emotion regulation (Anand et al., 2007; McRae et al., 2014; Outhred et al., 2015; Saberi et al., 2025), all analyses included medication status (i.e., taking any medication versus not) as a covariate.

Race and ethnicity data were collected based on the National Institutes of Health policy and US census categories. Participating youth’s race was reported by their caregiver. Race was not reported for one participant. Caregivers identified participants as Asian (n = 3, 2.42%), Black or African American (n = 44, 35.48%), Hispanic or Latina/o/x (n = 7, 5.65%), White (n = 54, 43.54%), or more than one race/other (n = 16, 12.90%). If “more than one race/other” was chosen, caregivers were prompted to report their child’s race in their own words; see the Supplemental Materials for this free response data. Additional demographic characteristics of the final sample are provided in Table 1.

Table 1.

Sample characteristics and sample comparisons between final scanned sample and the sub-sample with emotional awareness data

	Scanned sample (n = 125)			Emotional awareness subset (n = 86)
Variable	% (n)	Mean (SD)	Range	% (n)	Mean (SD)	Range
Age, years		12.77 (1.94)	9 – 17		12.39 (2.11)	9 – 17	t(123) = 1.34, p = 0.18
Gender identity							χ2(2) = 3.83, p = 0.15
Girl	93.50 (115)			91.49 (86)
Prefer to self-describe	3.25 (4)			4.26 (4)
Boy	3.25 (4)			4.26 (4)
Annual Household Income, dollars		67514.90 (55358.46)	0 – 300,000		65976.37 (54797.26)	0 – 300,000	t(122) = 0.46, p = 0.64
Income-to-Needs		2.76 (2.33)	0.01 – 11.47		2.74 (2.33)	0.01 – 11.47	t(121) = 0.085, p = 0.93
Adversity Exposure
Threat composite		2.03 (1.32)	0 – 4		2.04 (1.34)	0 – 4	t (123) = −0.33, p = .74
Deprivation composite		0.77 (0.77)	0 – 2		1.10 (0.79)	0 – 2	t(123) =1.95, p = 0.053
Emotional Awareness		17.51 (7.06)	8 – 37		17.25 (7.09)	8 – 37

Note. SD refers to the standard deviation. The participant self-reported their gender identity; all participants were assigned female sex at birth. The original response options for gender were “Female,” “Male,” and “Self-Report,” but we amended the language here because Female/Male typically refers to binary sex categorization.

Procedure

Youth provided written assent, and primary caregivers provided written consent for all study procedures. Participants completed two sessions. As part of the parent study, the youth and a parent or legal guardian completed a baseline laboratory assessment where the youth self-reported experiences of adversity and caregivers reported on their child’s exposure to adversity using questionnaires and semi-structured clinical interviews. The MRI protocol included a structural T1-weighted scan and functional MRI scans during which participants completed a well-established cognitive reappraisal paradigm. Other procedures not relevant to the current analyses are reported elsewhere (Gruhn et al., 2024; Miller et al., 2024; Pelletier-Baldelli et al., 2024; Rodriguez-Thompson et al., 2024).

Measures

Threat and deprivation.

Early life adversity was assessed using a multi-method, multi-informant set of validated questionnaires and interviews (Berman et al., 2022). Following standard practices in dimensional models of adversity (Berman et al., 2022; Machlin et al., 2025), we constructed composite scores representing the number of distinct types of experiences within a dimension. This approach is widely used in studies of deprivation and threat (Garrisi et al., 2025; Machlin et al., 2019; Milojevich et al., 2019; Murgueitio et al., 2024; Peverill et al., 2023; Sumner et al., 2019; Weissman et al., 2022). Experiences were classified as “present” or “absent” based on established cutoffs for the relevant measures, and the number of present adverse experiences was summed to yield dimension-specific indices (threat range: 0–4; deprivation range: 0–2).

Threat exposure included endorsement of (1) sexual abuse, (2) physical abuse and/or harsh discipline such as spanking, (3) emotional abuse, or (4) physical danger outside of the home and/or peer victimization. These exposures were considered “present” if endorsed by the adolescent on select items of the Mini International Neuropsychiatric Interview for Children and Adolescents PTSD module (MINI-PTSD Sheehan et al., 1998), the Childhood Trauma Questionnaire (CTQ; Bernstein et al., 2003), Peer Victimization Questionnaire (Prinstein et al., 2001), the Stress and Adversity Inventory (Adolescent STRAIN; Slavich et al., 2019), and/or by the parent on select items of the Parenting Styles and Dimensions Questionnaire (Robinson et al., 2001). Deprivation exposure included endorsement of (1) neglect or lack of parental availability and (2) material deprivation. Exposure was counted as present if endorsed by the adolescent on select items of the Adolescent STRAIN, CTQ, Child Chronic Strain Questionnaire (CCSQ; Rudolph et al., 2001), and/or by their parent on select items of the Parent STRAIN. The full list of items within the threat and deprivation dimensions are provided in Supplementary Table 1.

Poor emotional awareness.

Participants completed the 8-item “Poor Awareness” subscale of the Emotion Expression Scale for Children (EESC; Penza-Clyve & Zeman, 2002). This subscale assesses difficulty identifying and understanding one’s emotions or the cause of one’s feelings (e.g., Sometimes I just don’t have words to describe how I feel; I have feelings I can’t figure out; I often do not know why I’m angry). The EESC was adapted from the Toronto Alexithymia Scale for adults (Bagby et al., 1994) and was initially validated on children 9–12 years old (Penza-Clyve & Zeman, 2002). Participants indicated the extent to which they agreed with each item on a scale ranging from 1 (not at all true) to 5 (extremely true), and responses were summed for a total score. As a result, higher scores reflect lack of emotional awareness, or poorer emotional awareness. The subscale has good psychometric properties (Penza-Clyve & Zeman, 2002) and has been used in prior studies of early to mid-adolescents (Sim & Zeman, 2005; McLaughlin et al., 2011). This subscale demonstrated good internal consistency in our sample (α = 0.90; ω = 0.91).

Stimuli and Procedure

Participants completed a well-established cognitive reappraisal paradigm (McLaughlin et al., 2015; Ochsner et al., 2004; Silvers et al., 2012). During the task, participants viewed pictures from the International Affective Picture System (Lang, Bradley & Cuthbert, 2005) and a set of similar images normed for children and adolescents (https://osf.io/43hfq/; Jenness et al., 2018). For each trial, participants were first shown a 2-second instructional cue to either Look or Decrease, followed by an image (Negative or Neutral). Participants were instructed that the Look cue meant that they should “allow [themselves] to have whatever feelings, thoughts, and emotion [they] have about the picture without trying to change them.” Neutral images were always preceded by the Look cue. Participants were instructed that the Decrease cue meant they should “try to decrease or reduce the emotion [they] were feeling.” The full task instructions, including example reappraisal strategies provided to participants, are provided in the Supplemental Materials. After each image, participants rated the intensity of their negative emotion on a 5-point scale from 0 (“you felt almost no emotion” or “neutral”) to 4 (“it would be hard for you to feel that emotion even more strongly”). The Decrease cue always preceded Negative images, so that participants were reappraising their responses to an aversive stimulus. The presentation of the images was randomized for 4, 6, or 8 second duration. The rating screen appeared for 4 seconds, followed by an inter-trial interval that lasted from 0.5 to 7.5 seconds.

The task consisted of six runs, with a targeted social rejection paradigm introduced between the third and fourth run. Each run was 24 trials and evenly divided across the three conditions, Look Neutral, Look Negative, and Decrease Negative, for a total of 144 trials across all six runs. The mean rating across Look Neutral trials was 1.56 (range = 1.00–4.38, SD = 0.55). The mean rating across Look Negative trials was 2.87 (range = 1.00–4.78 SD = 0.78). For negative images across Decrease Negative trials, the mean rating was 2.34 (range = 1.00–4.36, SD = 0.70). Here, we focus on the effects of threat exposure on neural activation and amygdala connectivity during the Decrease Negative relative to Look Negative trials. Whole brain neural activation during the main task effects, Look Negative > Look Negative and Decrease Negative > Look Negative, before the targeted rejection are reported in Miller et al. (2024).

MRI data acquisition and preprocessing

Imaging data were collected on a 3.0-T Siemens Prisma Scanner using a 32-channel head coil. Anatomical scans (T1-weighted MPRAGE volumes; TR=2,530ms, TE=1670–7,250ms, flip angle=7°, FOV=192×192mm, 176 slices, in-plane voxel size=1mm) were acquired for co-registration with functional magnetic resonance imaging data. Functional data were acquired using a gradient-echo T2-weighted EPI sequence of 44 2.4mm-thick slices acquired parallel to the AC-PC line (repetition time = 2500 msec, echo time = 28 msec, flip angle = 90°, bandwidth = 2,312 Hz/Px, echo spacing = 0.52 msec, field of view = 230 × 230 mm). Six runs of approximately 6.5 min each (155 TRs) were collected.

Preprocessing was conducted through an in-house pipeline with standard analysis preparation. For consistency, we adapt this section from previous publications using this sample (Miller et al., 2024, Rodriguez-Thompson et al., 2024). Preprocessing steps included skull-stripping using the FSL’s Brain Extraction Tool (Smith, 2002), motion outlier detection using the fsl_motion_outliers program to identify framewise displacement exceeding 0.9mm in any direction (Jenkinson, Beckmann et al., 2012), motion correction and slice-timing correction using NiPy (Roche, 2011). Anatomical coregistration of T2*weighted functional images with each participant’s T1-weighted images was performed using bbregister. Images were then normalized into standard space (Montreal Neurological Institute, 2 mm 152 Atlas Brain) using Advanced Normalization Tools software (Avants et al., 2011), spatially smoothed in FSL with a Gaussian kernel of 5mm full width at half maximum.

Analysis Plan

Reasons for exclusion

Participants needed valid data for half of the MRI task (i.e., three out of six runs) to be included in our analyses. Out of the 138 participants who attended the MRI session, seven participants were excluded prior to data analysis due to (a) ending the session before completing three runs of the task (n = 5), (b) an incidental finding (n =1), or (c) not following task instructions (n =1). Six additional participants were excluded after MRI preprocessing due to excessive motion observed during four or more runs of the task, leaving a final analytic sample of 125 youth.

Individual runs were excluded if more than 40% of timepoints exceeded 0.9mm framewise displacement and/or there was a single motion spike greater than 5mm. A minimum threshold of 40% useable volumes would ensure participants have a minimum of ~186 usable volumes across all runs (7.75 minutes of data). Among the included participants, 28 individual runs were excluded based on these criteria. Among the included runs, the average framewise displacement was 0.31 mm (SD = 0.27; Supplemental Figure S4) and 7.14% of volumes exceed 0.9mm framewise displacement

Self-reported ratings analysis

We conducted a set of post-registered analyses to test the effects of threat exposure, emotional awareness, condition, and their interactions on participant ratings during the task. Hierarchical linear models were run using the lme4 package in R (version 1.1–33; Bates et al., 2015) with participant ID included as a random intercept (unconditional ICC: 0.27) ¹. Trial-level emotional intensity ratings were modeled as the outcome variable and trial condition was included as a level 1 fixed effect with Look Negative as the referent. Threat, deprivation, and poor emotional awareness were modeled as level 2 fixed effects and between-person-mean centered. Income-to-needs ratio, age, and medication use (yes/no) were included as level 2 covariates given their correlations with adversity exposure (e.g., Weissman et al., 2022).

Two separate models were run: one in which threat exposure was the focal predictor (controlling for deprivation), and one in which deprivation was the focal predictor (controlling for threat), each including the three-way interaction between condition, adversity, and emotional awareness. Fixed effects were evaluated using Type III tests, and unique effect sizes (partial η²) were estimated with the effectsize package (Ben-Shachar et al., 2020). To interpret significant condition by adversity interactions, we used the emmeans package (Lenth, 2022). We first estimated the simple slopes of deprivation or threat within each condition then compared these slopes across conditions using Tukey-adjusted pairwise tests.

fMRI Analysis

General Linear Modelling.

A general linear model was constructed in the FSL’s Expert Analysis Tool (version 6.00; Woolrich et al., 2009). Equally weighted event onsets and durations for the task conditions were convolved with a double-gamma canonical hemodynamic response function. High-motion volumes (framewise displacement > 0.9mm) as well as 12 motion parameters (six rigid-body motion parameters and their temporal derivatives) were included as nuisance regressors. To remove low-frequency drifts, data were high-pass filtered with a cutoff period of 100 seconds. FILM pre-whitening was used to estimate and remove temporal autocorrelation (Woolrich et al., 2001). The main contrast of interest compares neural responses during the Decrease Negative trials to responses during the Look Negative trials (i.e. Decrease Negative > Look Negative), reflecting cognitive reappraisal of negative stimuli. We also conducted post-hoc analyses to examine activation during emotional reactivity, Look Negative > Look Neutral across useable runs; these results can be found in the Supplemental Materials. The decision to investigate this question was made after post-registration and after gPPI analyses were complete. Consequently, analyses concerning Look Negative > Look Neutral should be considered post-hoc because they were not described in the post-registration (Hollenbeck & Wright, 2017).

Within-subject runs were combined in a fixed-effects model for each participant, which averaged the contrast estimates in FMRIB’s local analysis of mixed effects (FLAME 1; Beckmann, Jenkinson, & Smith, 2003). Individual-level estimates were submitted to a group-level mixed-effects model using FLAME 1. At the group level, whole-brain mixed-effects analyses tested associations between threat and deprivation exposures and the Decrease Negative > Look Negative contrast, with each exposure examined while controlling for the other. Significant results are reported after cluster-level correction in FSL with a voxel-level threshold of z > 3.1, p < 0.01 (Eklund et al., 2016).

Psychophysiological interaction.

We used generalized psychophysiological interaction analyses to identify threat exposure-related differences in functional connectivity of the amygdala during emotional reactivity and emotion regulation (gPPI; McLaren et al., 2012). The amygdala seed was created by combining across right and left anatomical regions defined using the Harvard-Oxford probabilistic subcortical atlas at 50% a probability threshold (e.g., Colich et al., 2023; DeCross et al., 2022; Weissman et al., 2022). We then extracted the mean time series from the bilateral amygdala seed region using the FSL means function. The included regressors for each task condition, the bilateral amygdala time series, the interaction between demeaned task regressor and the amygdala timeseries, and the nuisance regressors (volumes with framewise displacement > 0.9mm and six motion parameters). The gPPI models were run in FSL’s Expert Analysis Tool (version 6.00; Woolrich et al., 2009), which by default does not deconvolve the extracted amygdala timeseries prior to the interaction (O’Reilly et al., 2012) (although see Bloom et al., 2022).² Other covariates added to the models are described below. Within-subject runs were combined in a fixed effects model for each participant in FLAME and participants were then combined in a higher-level mixed effects model using FLAME Stage 1. We applied the same cluster-level correction as in whole brain analysis, z > 3.1, p > 0.01.

Covariates

We controlled for age at the time of the MRI scan, income-to-needs ratio, and medication use for a psychiatric condition (i.e., yes/no) in all self-report, activation and gPPI analyses. Income-to-needs ratio was calculated by dividing family income as reported by the caregiver by the poverty threshold for their family size based on the US Census Bureau. Medication use was coded as a dichotomous variable, with participants taking any medication coded as 1 and participants not taking any medication coded as 0. Predictors were added into the models in a stepwise fashion, such that the focal predictor (i.e., threat or low emotional awareness) and deprivation were added first, followed by the interaction terms in step two, age in step three, age and income-to-needs ratio in step four, and medication use in step five. Continuous covariates–age and income-to-needs ratio–were group mean-centered. Two participants were missing information necessary to compute the income-to-needs ratio; these values were replaced with the group mean, 2.76. The results of our main analyses adjusted for these covariates are presented in the Supplementary Materials. Our postregistered analysis plan proposed running sensitivity analyses excluding (1) five participants with dental implants and (2) nine participants who experienced minor technical issues with the head coil. The exclusion of these participants did not change our results; thus, results are presented with all 125 participants included.

Results

Descriptive statistics

Demographic characteristics of the final sample are displayed in Table 1. Most participants experienced early life adversity characterized by threat; 88.00% (n =110) reported at least one threat exposure, and 76.80% of the sample (n = 96) reported at least one deprivation exposure (see Table 2 and Supplementary Figure S3). In line with other studies of threat and deprivation (Miller et al., 2018; Sumner et al., 2019), threat and deprivation scores in our sample were positively correlated, r(125) = 0.33, p < 0.01. Threat and deprivation scores were also positively correlated with poor emotional awareness, r(125) = 0.36 and r(125) = 0.27, respectively, p’s < 0.01. Pearson correlations among variables of interest and covariates are presented in Supplementary Table 2.

Table 2.

Frequency of participants reporting exposure to each threat and deprivation category

Variable	% (n) of participants reporting presence	% (n) of participants reporting absence
Threat
Sexual Abuse	21.77% (27)	78.22% (97)
Physical Abuse or Harsh Discipline	65.6% (82)	34.4% (43)
Physical Danger Outside of the Home	55.2% (69)	44.8% (56)
Emotional Abuse	55.65% (69)	44.35% (55)
Deprivation
Neglect or Lack or Parental Availability	54.84% (68)	45.16% (56)
Material Deprivation	61.6% (77)	38.4% (48)

Note. n =125

Self-reported emotional intensity ratings

Manipulation check.

Participants self-reported emotional intensity ratings significantly varied across conditions F(2, 1385.85) = 581.69, p < .001, partial η² = .46. The task ratings varied in the expected directions; participants rated their emotional responses to the Look Negative trials as significantly more intense than Look Neutral trials (mean difference = 1.27, 95% CI [1.13, 1.42], t(124) = 17.25, p < .001; Cohen’s d =1.54, 95% CI [1.28, 1.80]) and more intense than the Decrease Negative trials (mean difference = −0.49, 95% CI [−0.60, −0.39], t(124) = −9.12, p < .001; Cohen’s d = −0.82, 95% CI [−1.02, −0.61]).

Early life adversity exposure.

A significant two-way interaction emerged between condition and threat exposure, F(2, 1385.85) = 5.06, p = 0.006, partial η² = 0.007, controlling for the effects of deprivation, age, income-to-needs ratio, and medication use. Follow-up simple slopes analysis within condition indicated that, for Look Neutral trials, higher threat exposure predicted higher ratings or more intense self-reported emotion b = 0.07, 95% CI [0.008, 0.12], p = .002, β = 0.09. In contrast, threat was not associated with ratings in the Look Negative or Decrease Negative conditions. Direct comparison of simple slopes indicated there were no significant differences in the effect of threat on Decrease Negative compared to Look Negative trials; there was also no significant differences in the effect of threat on Look Neutral relative to Look Negative trials.

Controlling for the effects of threat and the additional covariates, there was a significant condition by deprivation interaction F(2, 1385.80) = 32.35, p < 0.001, partial η² = 0.04. In the Look Neutral condition, higher deprivation predicted more intense self-reported emotion b = 0.38, 95% CI [0.29, 0.47], p < 0.001, β = 0.31. Similarly, higher deprivation exposure was associated with higher ratings in the Decrease Negative condition, b = 0.23 95% CI [0.14, 0.32], p < 0.001, β = 0.19. Post-hoc pairwise comparisons confirmed significant differences in the effect of deprivation between Look Negative compared to Look Neutral trials b = −0.38, SE = 0.048, p < 0.001 but not between the Decrease Negative and Look Negative trials. In response to a reviewer’s question, we probed condition effects at −1 SD, mean, and +1 SD levels of deprivation using estimated marginal means. Across all deprivation levels, participants rated Negative Look trials as more negative than Neutral Look trials and reported lower negative affect during Decrease Negative compared to Look Negative trials (all ps < .01). However, condition differences were smaller at higher deprivation (Look Negative > Look Neutral: −1 SD = 1.58, Mean = 1.29, +1 SD = 0.99; Look Negative > Decrease Negative: −1 SD = 0.55, Mean = 0.37, +1 SD = 0.19).

Relations between poor awareness and adversity exposure.

There was also a significant interaction between threat and emotional awareness, F (1, 77.02) = 5.39, p = 0.023, partial η² = 0.007 after accounting for deprivation and other covariates. Specifically, the association between threat and emotional intensity ratings strengthened across conditions as poor awareness increased. However, there was not a significant interaction between deprivation and emotional awareness (See Supplemental Figures S5 and S6).

fMRI results

Early life adversity exposure and brain activation during emotion regulation.

When examining the effect of threat exposure, controlling for deprivation, threat exposure was negatively associated with Decrease Negative > Look Negative contrast estimates in the right angular gyrus (Figure 1b; Table 3). In other words, the extent to which neural responses during Decrease Negative exceeded those during Look Negative was attenuated in the angular gyrus at higher levels of threat exposure. This association remained unchanged after adjusting for age, income-to-needs ratio, and medication use (see Supplementary Table 3 for sensitivity analyses). No significant associations with deprivation were observed for the Decrease Negative > Look Negative contrast.

Fig. 1.

Whole brain maps of (a) activation for the main effect of the Decrease Negative > Look Negative contrast, and where (b) activation for Decrease Negative > Look Negative is negatively associated with threat exposure, controlling for deprivation exposure. Areas in blue/green in Panel B signify regions where the extent to which responses during Decrease Negative > Look Negative was attenuated at higher levels of threat exposure.

Table 3.

Brain regions showing significant activation during emotion regulation

Contrast Label	Region of Peak Activation	BA	k	Z value	x	y	z
Decrease Negative > Look Negative
	Right lingual gyrus	18	22538	8.96	0	−86	−4
	mPFC	47	4169	6.18	48	28	−8
	Left middle frontal gyrus	8	1230	5.43	−36	22	48
	mPFC	10	572	4.84	−38	58	−4
	Middle temporal gyrus (posterior division)	21	364	6.81	56	−8	−26
	Middle temporal gyrus (anterior division)	21	330	5.74	−54	−10	−26
Threat
	Right angular gyrus *	39	194	5.33	58	−54	44

Note. n =125. The results reported here are adjusted for the effect of deprivation exposure. BA refers to the Broadmann’s area of the peak activation. k refers to the number of voxels in each cluster. z value refers to the peak activation level in each cluster. x,y and z refers to Montreal Neurological Institute coordinates.

^*denotes that threat regressor was modelled as negative contrast weight (threat = −1) to identify regions where threat exposure was negatively associated with the Decrease > Look contrast estimate; thus the reported effect indicate that higher threat exposure is associated with a reduced magnitude of the Decrease > Look contrast in this region.

Main effects of task condition on amygdala connectivity.

A whole brain gPPI analysis was used to examine connectivity between the amygdala and other regions of the brain during the Decrease Negative > Look Negative contrast. We did not find significant differences in amygdala connectivity between the Decrease Negative and Look Negative conditions.

Adversity exposure and functional connectivity.

Next, we assessed whether adversity exposure was associated with amygdala connectivity during emotion regulation. Contrary to our hypotheses, threat exposure was not related to amygdala connectivity during the Decrease Negative relative to Look Negative condition. Similarly, deprivation was also unrelated to amygdala connectivity during the Decrease Negative > Look Negative contrast.

Poor awareness and functional connectivity.

We next investigated whether low emotional awareness was associated with amygdala connectivity with the rest of the brain during Decrease Negative relative to Look Negative trials. Poorer emotional awareness was associated with stronger connectivity between the amygdala and one large cluster of voxels, with a peak activation in the left thalamus. This cluster of voxels also wrapped into areas of the posterior cingulate cortex, left hippocampus, posterior insula, brainstem and cerebellum. On the other hand, poorer emotional awareness was associated with weaker connectivity between the amygdala and vmPFC (see Figure 2a, Table 4). These results remained largely unchanged when adjusting for age and income-to-needs; however, when additionally adjusting for medication use, emotional awareness was not associated with connectivity between the amygdala and any other region of the brain (see Supplemental Table 4).

Fig. 2.

Whole brain functional connectivity of bilateral amygdala with other regions where (a) emotional awareness interacted with task condition (Decrease Negative > Look Negative) and where (b) the interaction of emotional awareness and threat exposure was modulated by task condition, after controlling for deprivation exposure. Areas in blue/green in Panel A signify regions where the extent to which responses during Decrease Negative > Look Negative was attenuated at higher levels of poor awareness.

Table 4.

Brain regions showing significant coactivation with the bilateral amygdala during Decrease Negative > Look Negative

Contrast Label	Region of Peak Activation	BA	k	Z value	x	y	z
(Amygdala x Decrease Negative) > (Amygdala x Look Negative)
Emotional awareness
	Left thalamus		23442	5.00	−14	−34	−2
	vmPFC*		754	3.91	−36	54	−4
	vmPFC*		303	4.18	34	60	0
Emotional awareness x threat, controlling for deprivation
	vmPFC	10	129	4.02	−32	44	2
		10	119	3.97	−18	58	4

Note. n = 88. BA refers to the Broadmann’s area of the peak coordinate. k refers to the number of voxels in each cluster. z value refers to the peak activation level in each cluster. x,y and z refer to Montreal Neurological Institute coordinates.

^*denotes that threat regressor was modelled as negative contrast weight (threat = −1) to identify regions where threat exposure was negatively associated with the Decrease > Look contrast estimate; thus the reported effect indicate that higher threat exposure is associated with a reduced magnitude of the Decrease > Look contrast in this region.

Relations between threat, poor awareness and amygdala connectivity.

Finally, we tested whether threat exposure and poor emotional awareness interacted to predict amygdala connectivity for the contrast Decrease Negative > Look Negative varied by adversity exposure. Here, we found a significant three-way interaction between threat exposure, emotional awareness, and connectivity between the amygdala and two clusters within the vmPFC (Figure 2b, Table 4). These results were largely unchanged when adjusting for deprivation exposure but did not persist when adjusting for income-to-needs and medication use (see Supplemental Table 5). For descriptive purposes, we extracted parameter estimates of functional connectivity and plotted them against threat exposure and emotional awareness (See Figure 3).

Fig 3.

Descriptive plot of parameter estimates from interaction of threat exposure and low emotional awareness on amygdala-vmPFC connectivity. Results suggest that participants high in poor emotional awareness and high in threat exposure demonstrated stronger positive amygdala-vmPFC connectivity

We also investigated whether there was a significant interaction between deprivation-related adversity and low emotional awareness on amygdala connectivity. Deprivation exposure interacted with low emotional awareness to predict increased connectivity between the amygdala and five clusters of voxels with peak activations in the posterior insula, brainstem, left thalamus, precuneus, and vmPFC. However, the interaction between deprivation exposure and low emotional awareness was no longer significant when adjusting for the effect of threat exposure.

Discussion

The current study investigated how threat exposure affects brain activation and amygdala connectivity with the prefrontal cortex (PFC) circuitry during explicit emotion regulation (i.e., when youth explicitly had the goal to downregulate their responses to negative images using cognitive reappraisal). We also tested whether these associations were moderated by poor emotional awareness. Our results suggest that reappraisal-related amygdala connectivity varied as a function of poor awareness and childhood threat exposure. Given that the association between threat and emotional intensity ratings strengthened across conditions as poor awareness increased, it may be that case that poor awareness exacerbates difficulties in emotion regulation for youth with higher levels of threat exposure. Notably, the association between poor awareness, high threat exposure, and positive amygdala-vmPFC coupling persisted even after adjusting for another dimension of adversity, experiences of deprivation (i.e., lack of social, emotional, and cognitive stimulation). To our knowledge, these data are the first to examine how emotional awareness and dimensions of adversity relate to neural mechanisms underlying emotional regulation. In doing so, the results extend prior self-report evidence that poor emotional awareness contributes to emotion regulation difficulties in youth (Riley et al., 2019; Van Beveren et al., 2019) and add preliminary neuroimaging evidence to a growing body of work that examines the effects of differential effects of threat and deprivation on cognitive reappraisal (e.g., McLaughlin et al., 2015; Milojevich et al., 2019).

Threat adversity and neural activation during cognitive reappraisal

Our first aim examined associations between childhood threat exposure and neural recruitment during cognitive reappraisal. Prior work from McLaughlin and colleagues (2015) found that youth with greater threat exposure exhibited greater recruitment of the dACC, superior frontal gyrus, and mPFC during reappraisal. We did not replicate that pattern here but observed that greater threat exposure was associated with less activity in the right angular gyrus. The right angular gyrus is part of the right temporal parietal junction, a neural region frequently activated when reorienting one’s attention to unexpected external stimuli (Geng & Vossel, 2013; Krall et al., 2015). Reduced activation of the right angular gyrus may reflect the filtering of irrelevant sensory input, ensuring that motivationally relevant stimuli are noticed and unimportant stimuli do not divert attention away from targets (Shulman et al., 2007). Given that we did not observe threat-related differences in self-reported emotional intensity across trial conditions, decreased activation of the angular gyrus may serve a compensatory function, allowing threat-exposed youth to modulate their emotional responses when they explicitly engage in cognitive reappraisal.

Threat adversity and vmPFC-amygdala connectivity during cognitive reappraisal

The primary aim of the present study was to examine how threat exposure predicted reappraisal-related amygdala connectivity and whether this association was moderated by participants’ degree of emotional awareness. Looking first at the main effect of threat, we did not find any association between threat exposure and differences in amygdala connectivity during effortful regulation. Several studies document adversity-related differences in connectivity between the amygdala and PFC (Callaghan & Richardson, 2011, 2012; Honeycutt et al., 2020; Gard et al., 2022; Gee et al., 2013; Peverill et al., 2019; Jenness 2021). However, our study is among the first to examine this question during participants’ explicit attempts to regulate their emotions, underscoring the need for more research to disentangle the impact of threat on amygdala connectivity across a wide variety of task demands.

Emotional awareness and vmPFC-amygdala connectivity during cognitive reappraisal

Looking next at the main effect of emotional awareness, participants with poorer emotional awareness demonstrated greater connectivity between the amygdala and left thalamus when regulating their emotions. We also found that poorer emotional awareness predicted stronger connectivity between the amygdala and areas of the posterior insula, cerebellum, precuneus, and brainstem. One of the amygdala’s primary functions is determining the motivational salience of exteroceptive sensory information (Lindquist et al., 2012); this function is supported by reciprocal interactions between the amygdala and regions involved in perceptual processing, including the thalamus and primary sensory cortices, and default mode regions involved in conceptual knowledge, like the medial prefrontal cortex.

These connections enable the amygdala to integrate rich sensory inputs and, in turn, modulate and enhance neural representations of emotional stimuli (Cunningham & Brosch, 2012; LeDoux, 2007). For example, studies have shown that the amygdala’s connections with sensory regions allow rapid processing of basic sensory information, facilitating early detection of emotionally salient stimuli, particularly under conditions of uncertainty or threat (Herry et al., 2007; Vuilleumier et al., 2003); on the other hand, conscious awareness of salient stimuli is associated with coupling between the amygdala and mPFC (Lapate et al., 2016) Speculatively, in participants with low emotional awareness, this subcortical connectivity may dominate. Such connectivity may reflect a neural mechanism that prioritizes rapid detection and response to salient stimuli at the expense of relevant conceptual knowledge. In contrast, individuals with higher emotional awareness may be more likely to recruit both subcortical and prefrontal regions.

We also observed that poorer emotional awareness was associated with weaker connectivity between the amygdala and the vmPFC during reappraisal. Cognitive reappraisal is typically associated with correlated amygdala-vmPFC connectivity in adults, with some studies finding significant negative coupling between the amygdala and vmPFC (Banks et al., 2007; Kanske et al., 2011; Urry et al., 2006) On the other hand, in childhood, cognitive reappraisal is often associated with positive coupling between the amygdala and vmPFC (Dougherty et al., 2015; Silvers et al., 2016, 2017; c.f. McRae et al., 2012). As a result, some argue that negative coupling between the amygdala and vmPFC reflects a more “mature” neurodevelopmental pattern (e.g., Kopala-Sibley et al., 2020). One interpretation of our findings is that adolescents with poor emotional awareness have disruptions to developmentally “normative” reappraisal-related amygdala-vmPFC connectivity, showing a pattern more commonly observed in younger children. This is consistent with observations that understanding one’s emotions facilitates emotion regulation (Kashdan et al., 2015), and poor emotional awareness predicts emotion regulation difficulties (Boden & Thompson, 2015; Van Beveren et al., 2019). Ultimately, these findings should be interpreted with caution, given that there are very few studies on the association between emotional awareness and neural connectivity, and our results were no longer significant when accounting for the effects of for socioeconomic status (as measured by income-to-needs ratio) and medication use. Much more targeted research to replicate and extend these findings is needed to understand the neural mechanisms of emotional awareness.

Interaction between threat and emotional awareness on vmPFC-amygdala connectivity during cognitive reappraisal

We observed that the association between low emotional awareness and task-related functional amygdala-vmPFC connectivity during reappraisal was driven primarily by youth with relatively high exposure to threat. Specifically, youth reporting poorer emotional awareness and relatively high threat exposure demonstrated significantly more positive amygdala-vmPFC connectivity during reappraisal, but this association was not observed at average and low levels of threat exposure. These findings suggest that youth with low emotional awareness might be at particular risk for emotion regulation difficulty after experiencing threat-related adversity. Such individual differences may explain why amygdala-mPFC connectivity patterns are mixed across the early adversity literature (McLaughlin et al., 2019). Some studies find that youth with greater threat exposure exhibit negative coupling between the amygdala and mPFC when attending to or labeling negative stimuli (Colich et al., 2017; Peverill et al., 2019). In contrast, others find positive coupling (Herringa et al., 2016), and still others find null effects (Weissman et al., 2020). Our findings suggest that emotional awareness could be a third variable moderating the effects of threat on amygdala connectivity during cognitive reappraisal. Ultimately, studies using prospective longitudinal designs are needed to determine whether neural patterns associated with childhood threat exposure are a risk factor for, or result of, low emotional awareness.

Limitations

Several limitations of this study are worth noting. Although we have suggested that the interaction of threat exposure and low emotional awareness may influence the neurodevelopmental mechanisms underlying emotion, our cross-sectional design limits inferences about the temporal association between threat exposure and emotional awareness. Similarly, our analyses cannot account for the effects of timing and duration of adversity exposures on the brain. Adverse experiences that occur during developmentally sensitive windows, when the brain is undergoing drastic structural and functional changes, are more likely to alter neural development and have downstream consequences for responding to stress (Sheridan et al., 2022). Identifying the effects of timing and duration of adverse experiences would allow future researchers to better characterize the causal effects that unique environmental exposure may have on neural development.

In addition, our findings should be interpreted with caution given recognized limitations of gPPI, particularly in event-related designs. First, PPI analysis reveals undirected functional connectivity and, therefore, cannot tell us whether the connectivity we observed was primarily from the amygdala to the vmPFC or vice versa. Second, limited temporal separation between trial types in event-related designs, may limit unique variance attributable to the interaction term, reducing statistical power and increasing the probability of false null results (O’Reilly et al., 2012). Future work should explore alternative approaches such as beta-series correlation, which may may offer greater sensitivity in event-related designs (Bloom et al., 2022; Cisler et al., 2014). More broadly, our sample size (n = 125), while typical for neuroimaging studies (Szucs & Ioannidis, 2020), remains underpowered for detecting reliable brain–behavior associations. Emerging evidence indicates that brain-wide association studies often yield small effects (frequently r < .15) and may be susceptible to overestimation in samples of this size (Marek et al., 2022). These concerns are especially salient for interaction effects, which are typically less stable than main effects. Given these limitations, our findings should be considered an initial step that motivates replication. Ultimately, complementary analytic approaches such as model validation across multiple independent samples (e.g., Rosenberg & Finn, 2022; Spisak et al., 2023) and pooled inferences across multi-site datasets (e.g., Adhikari et al., 2019; van Velzen et al., 2022), will be critical for evaluating the robustness of these findings.

Implications and Conclusion

Despite these limitations, our study has several methodological strengths, including a relatively large sample size and multi-method assessments of adversity exposure that account for the distinct consequences of different exposures. Cognitive reappraisal is a useful way to cope with uncontrollable, acute stressors, especially those perceived as beyond one’s control (Troy et al., 2013). However, cognitive reappraisal is used less frequently by youth growing up with the threat of physical or emotional harm (Gruhn & Compas, 2020; Milojevich et al., 2018). Some scholars have posited that because maltreated youth are more likely to experience unsupportive emotion socialization contingencies (e.g., a caregiver punishing a crying child to alleviate their own stress) and are less likely to engage in positive emotion socialization (e.g., teaching emotional awareness and effective emotion regulation skills), maltreated youth are less likely to learn coping strategies that facilitate healthy regulatory responses (Cabecinha-Alati et al., 2022). Threat-exposed youth may in turn rely on disengagement coping strategies, which involve orienting away from the arousing aspects of a situation to avoid confrontation (Milojevich et al., 2018). Promisingly, there is evidence to suggest that youth exposed to threats can be taught to engage in cognitive reappraisal; when they do, they exhibit similar modulation of amygdala responses to negative stimuli to those youth not exposed to threats (McLaughlin et al., 2015; Rodman et al., 2019). Our findings further suggest that among youth exposed to threat, those with better emotional awareness have neural responses during cognitive reappraisal that are more like those youth with less threat exposure. In sum, this study provides novel insight into how emotional awareness may be associated with emotion regulation in youth exposed to threat adversity. Given the profound impact of early life adversity on risk for psychopathology, understanding factors that drive heterogeneity in developmental outcomes is key to creating targets for effective intervention.