Adverse childhood experiences (ACEs) refer to a broad range of stressful experiences that infants and children can be exposed to whilst growing up, such as emotional, physical and sexual abuse, emotional and physical neglect, or household dysfunctions including parental mental illness and domestic violence (Bernstein et al., 2003; Edwards et al., 2003; Zhang et al., 2023). It has been well-established that ACEs are robustly associated with an increased risk of major depressive disorder (MDD) in a non-linear dose-response relationship, which underscores the cumulative risk of ACEs (Tan and Mao, 2023; Wang et al., 2020). Beyond this cumulative risk, the dimensional model of adversity and psychopathology (DMAP) recently conceptualizes ACEs along two core dimensions with shared features: including threat (i.e., experiences of harm or violence; e.g., abuse) and deprivation (i.e., reduced social or cognitive stimulation, e.g., neglect) (Sheridan and McLaughlin, 2014; McLaughlin and Sheridan, 2016; McLaughlin et al., 2021).The dimensions of threat and deprivation have been argued to influence psychopathology in ways that are at least partially distinct (Sadikova et al., 2025). Unfortunately, the underlying cognitive mechanisms from ACEs as well as threat and deprivation dimensions on MDD remain poorly understood, especially in the key period of adolescence during which both prevalence and severity of MDD and/or depressive symptoms dramatically increase (Thapar et al., 2022; Kerestes et al., 2016).

Recently, attentional bias to facial expressions that contains abundant (social) information, has been suggested as a novel, promising cognitive bridge linking ACEs to MDD (McCrory et al., 2017; Mao et al., 2020). Relying on behavioral reaction time tasks (e.g., dot-probe or emotional Stroop tasks), extant research has provided more evidence for attentional bias to sad, rather than anger, faces in MDD individuals, with such biases typically emerging after long (>1000 ms) facial expression exposure durations (Mogg et al., 1995; Mathews and MacLeod, 2005; Peckham et al., 2010). However, the association between attentional bias to happy faces and depression remains inconsistent (e.g., Peckham et al., 2010; Hankin et al., 2010). Although these studies advance our knowledge about emotional processing mechanism underlying depression, reliance on traditional reaction time tasks may introduce reporting bias due to imprecise assessments of attentional biases (Yiend and Mathews, 2005).

Methodologically, early research using reaction time tasks ignored the poor internal consistency and reliability of derived indices, leading to mixed findings and inadequate replicability (Lavi et al., 2025; Sears et al., 2018). Compared to this traditional approach, eye-tracking has been validated as having superior psychometric properties for capturing attentional bias, as it provides direct and continuous records of attention over extended periods, and allows for a mechanistic distinction between early and late components of attention, i.e., orientation vs. maintenance bias (Calvo and Lang, 2004; Lazarov et al., 2018; Suslow et al., 2020). A recent meta-analysis using eye-tracking has revealed significant roles for both heightened and reduced maintenance bias to sad and happy faces, respectively, in depressed individuals (Suslow et al., 2020). This finding supports emotion theory (Clark, 2005), which proposes that depression is characterized not only by enhanced negative emotion but also by reduced positive emotion. An alternative interpretation suggests that the reduced maintenance bias to happiness reflects insensitivity to (social) reward, which constitutes a main cause of depressive-typical symptom of anhedonia (Jaffee, 2017). Meanwhile, another eye-tracking meta-analysis further highlights heightened maintenance bias to anger in depressed individuals (Huang et al., 2023), consistent with the concept of interpersonal theories of depression that those who are alert to faces signaling social threat and rejection are more susceptible to developing depression (Hames et al., 2013). However, when considering the early attentional component, no evidence of orientation bias was reported in meta-analyses (Huang et al., 2023; Suslow et al., 2020). The findings may suggest that depressed individuals are primarily accompanied by deficits and biases in elaborative information processing, and attention bias in depression is probably a controlled and motivated process occurring mainly at later, rather than early, stages of information processing.

In parallel, accumulating evidence suggests that attentional bias to facial expressions may be one of important and embedded cognitive alterations underlying early adversity (Harb et al., 2024; Sharma et al., 2022). Seminal work by Pollak and colleagues demonstrated that children exposed to physical abuse show heightened perceptual sensitivity to angry facial expression, requiring less perceptual information to identify anger (Pollak and Kistler, 2002; Pollak and Sinha, 2002). Nevertheless, to date, few studies have employed the eye-tracking approach to test this cognitive alteration (e.g., Hoepfel et al., 2022; Bodenschatz et al., 2019). Even in those relevant behavioral studies, considerable controversy exists regarding whether ACEs induce biased attention toward or away from negative, especially threatening (e.g., anger), faces (Gibb et al., 2009; Fani et al., 2011; Bertó et al., 2017). Recently, a very interesting opinion proposed that, developmental stage may help explain this divergence: young children exposed to trauma often direct their attention away from threatening stimuli (“attention hysteresis”), whereas adolescents tend to exhibit heightened attentional bias toward threatening stimuli, particularly long after the trauma has occurred (“attention vigilance”; Fani et al., 2011; Bertó et al., 2017). Moreover, it appears that blunted attentional bias to happiness in adolescents exposed to ACEs might be more prominent during adolescence, a period during which reward-related neural systems (e.g., ventral striatum) significantly accelerate to be mature (Braams et al., 2015). That is, when considering developmental periods, adolescents exposed to ACEs appear to be especially characterized by both heightened attentional bias to anger and reduced attentional bias to happiness, which requires verification via the eye-tracking approach.

More importantly, the cognitive “bridge” effect of attentional bias to happiness and anger in the context of ACEs and MDD has been further emphasized by recent theories. For instance, the latent vulnerability theory proposes that maltreatment exposure may embed enduring vulnerability to later psychopathology by altering neurocognitive function, with threat and reward processing representing two crucial mechanisms (McCrory et al., 2017). Within this broader framework, the DMAP further posits that different dimensions of adversity may influence psychopathology through partially distinct cognitive mechanisms (McLaughlin and Sheridan, 2016). Fear learning and threat processing would be particularly disrupted by threat-related ACEs, whereas reward processing may be implicated in both threat- and deprivation-related ACEs (Sadikova et al., 2025). Unfortunately, to date, only one study has investigated the mediating role of attention bias in the ACEs-depression links (Mao et al., 2020). Surprisingly, by conducting a dot-probe task with four pairs of emotional (happy, sad, anger, fearful) vs. neutral facial expressions in young adults, this study revealed that reduced attentional bias to sad, rather than other faces, significantly mediated the effect of ACEs on depressive symptoms. In parallel, their fMRI data highlighted the crucial role of within-network connectivity of ventral attention network (Mao et al., 2020). This study contributes to existing knowledge by providing preliminary behavioral evidence for the “bridge” effect of attentional bias in the ACEs-depression link. However, further evidence is required before a robust conclusion can be drawn, especially considering the key period of adolescence and the adoption of more validated indicators of attentional bias (i.e., eye-tracking approach) and testing the remaining subtype effect of ACEs.

In a classic free-viewing task in which happy, anger, sad, and neutral faces were presented, we conducted the first eye-tracking study to test (i) the mediation effect of attentional bias (maintenance vs. orientation bias) in the ACEs-adolescent MDD link, and (ii) explored the specific driven effect of dimensions of ACEs (see Fig. 1). In light of latent vulnerability theory and adolescence imprint, ACEs may increase vulnerability to adolescent MDD through maintenance, but not orientation, bias to facial expressions, especially to happy and anger faces. According to the DMAP theory, both deprivation- and threat-related ACEs might exert their effect primarily via reduced maintenance attention to happiness, whereas threat-related ACEs might work on MDD only via heighten maintenance bias to anger.