Descriptive statistics for all study variables are presented in Table 1. Findings from Little’s MCAR test suggested that the data was missing completely at random (X²(1838)= 127.38, p > .99). Relatedly, independent samples t-tests suggested that none of our main predictors (i.e., maltreatment subtypes) nor outcomes (i.e., depressed mood and anhedonia) varied as a function of completion status (p > .01). Thus, data were estimated using FIML. We next tested whether the growth of depressed mood and anhedonia was linear, quadratic, or cubic. For depressed mood symptoms, a quadratic model corresponded to the lowest BIC (13157.34); however, this was only 0.41 less than the linear model (BIC=13157.75). Similarly, the cubic model provided the best fit for anhedonic symptoms (BIC=8646.80); however, this was only a 1.03 BIC improvement over the linear model (BIC=8647.83). Therefore, linear growth models were applied to both dependent variables.

LGC models with maltreatment subtypes entered as covariates were next tested. Our model exhibited acceptable fit (X²/df=1.83; CFI=.95; RMSEA=.04; SRMR=0.04). Of note, only emotional neglect positively predicted both the intercept (p<.0001) and slope (p<.001) for anhedonic symptoms. depressed mood, emotional abuse (p<.0001) and emotional neglect (p=.002) were the only predictors of elevated baseline levels, while none of the predictors related to the slope (p > .20). These findings are displayed in Figure 1. The covariance matrix (Table 1) and estimated/actual means (Figure 1), with sex and race entered as covariates, for this model are presented in the appendix along with the parameter estimates for significant and non-significant findings in our model (Table 2 in the supplemental material). As seen in Figure 1 in the appendix, there was little change in the sample’s average levels of anhedonia and depressed mood annually, which is consistent with the larger adolescent depression literature (Conway et al., 2017).

Finally, we examined the influence of sex and race/ethnicity on our outcomes. We first tested if symptoms varied overall in our sample, and found that girls reported elevated levels of depressed mood at baseline (β= −0.24; SE=0.04; t= −5.50, p < .001). We did not find any other direct relations between demographics and depressive outcomes (p > .01). Next, the moderating influence of race was tested on the relation between maltreatment and depression. African-Americans reported elevated anhedonic symptoms at baseline within the context of elevated emotional abuse (β= 0.23; SE=0.11; t= 2.91, p < .01) as well as physical abuse (β= 0.49; SE=0.14; t= 3.60, p < .001). However, the slope for the association between anhedonia and physical abuse for African-Americans in our sample decreased over time (β= −0.69; SE=0.22; t= −3.04, p = .002), suggesting that elevated symptoms at baseline may regress to the mean. All other interactions between sex/race and maltreatment subtypes were non-significant (p > .01).

One of the more robust findings in the depression literature is the deleterious consequences of maltreatment across the lifespan (Liu, 2017; Norman et al., 2012). Recent findings suggest that emotional maltreatment in particular may confer depressogenic risk, especially in adolescents (Infurna et al., 2016). Our study sought to extend these findings by distinguishing between forms of emotional maltreatment (i.e., abuse and neglect) as well as symptoms of depression (i.e., depressed mood and anhedonia). Overall, our findings are congruent with Infurna and colleagues (2016) such that emotional abuse and neglect were the only maltreatment subtypes that uniquely related to elevated depressive symptoms. Further, we extended these findings by showing that the consequences of emotional neglect may be especially problematic in adolescence due to its impact on both depressed mood and anhedonia. These results, among others from our study, reiterate the importance of distinguishing between childhood adversities and symptom clusters when characterizing the relation between maltreatment and depression.

Chronic stress, including childhood adversities, is a well-documented risk factor for anhedonia (e.g., Berenbaum & Connelly, 1993). Pizzagalli (2014) offers a biopsychological explanation for this association in which early exposure to childhood adversities blunts dopaminergic pathways, subsequently leading to anhedonic responses. Coupled with past research (Hanson et al., 2015; Van Veen et al., 2013), our findings have interesting implications for clarifying this pathway within Pizzagalli’s (2014) model. It may be that neglect specifically relates to anhedonia because it is more chronic in nature and stable over time compared to other forms of maltreatment (Proctor & Dubowitz, 2014). Within this theoretical framework, chronic stress would mediate the relation between neglect-exposure and reward processing by leading to more persistent patterns of stress-exposure, which can tax the dopaminergic system related to approach behaviors (Trew, 2011). Confounding this explanation, however, is emerging research that deficits in the stress response system may not adequately explain the consequences of deprivation exposure (Busso, McLaughlin, & Sheridan, 2017). Instead, it may be important for other mediating variables to be explored in explaining the effects of neglect. For instance, behavioral (Wismer Fries & Pollak, 2017) and neural (Mehta, Gore-Langton, Golembo, Colvert, Williams, & Sonuga-Barke, 2010) indicators of reward learning, a key component of anhedonia (Pizzagalli, 2014), are negatively associated with neglect-exposure. Thus, emotional neglect-exposure may directly impact reward processing via impaired learning, exacerbating levels of anhedonic symptoms in these vulnerable youth.

Emotional neglect not only predicted levels of anhedonia, but also predicted increases over time. This suggests that the harmful effects of emotional neglect may increase during the transition from adolescence to emerging adulthood. The adolescent-adult transition is marked by new, challenging developmental tasks related to independence and interpersonal relationships that are important to master in order to maintain and/or promote emotional well-being (Menon et al., 2018). Implicit and associative learning deficits in neglect-exposed youth may make it challenging for these adolescents to adjust to these new social roles (Geoffroy, Pereira, Li, & Power, 2016). Consistent with an organizational approach to psychopathology (Cicchetti & Toth, 2009), the inability to navigate developmentally-salient challenges leads to increased distress in subsequent developmental periods. The increasing trajectory of anhedonic symptoms associated with neglect-exposure in our study, therefore, may reflect the effects of not being able to engage positively with new developmental challenges in adolescence (e.g., building positive peer relationships), cascading into emerging adulthood.

Our findings replicated Van Veen and colleagues’ (2013) results with adults, in that emotional abuse and neglect were independently associated with depressed mood, while sexual abuse and physical abuse were not. In a recent integrative model proposed by Luyten and Fonagy (2017) which extends motivational models of depression (e.g., approach-avoidance; Trew, 2011), the authors suggested that three biobehavioral systems underlie depression symptomatology: stress, reward, and social cognition (sometimes referred to as mentalization). We propose that similar to how emotional neglect may uniquely impact the reward system that underlies anhedonic presentations, emotional abuse and neglect may directly impact social cognition which leads to elevated depressed mood. Extant theory suggests that emotional maltreatment uniquely predicts depression because different than other maltreatment experiences, attributions about oneself are supplied directly by the perpetrator (Rose & Abramson, 1992). Within the context of emotional abuse, statements from the perpetrator (e.g., “you’re stupid”) can lead to a myriad of negative cognitive biases (Gibb, 2002) such as poor self-awareness-a key aspect of social cognition (Luyten & Fonagy, 2017). While patterns of emotional unavailability and unresponsiveness linked with emotional neglect may also indirectly communicate messages about one’s self-worth, it is important to note that both subtypes of emotional maltreatment predicted depressed mood when tested simultaneously. This suggests that the reasons emotional abuse and emotional neglect impact social cognition, and subsequent depressed mood, may be distinct.

A possible reason that emotional neglect’s impact on this risk pathway would be distinct from emotional abuse is that learning deficits that stem from emotional neglect can contribute towards impairment in social cognition as well as reward processing. Recent research suggests that cognitive deficits stemming from exposure to neglect are unique, and continuous through adolescence and adulthood (Geoffroy, Pereira, Li, & Power, 2016). Importantly, the structural and functional individual brain differences that underlie one’s cognitive capacity also underpin individual differences in emotional intelligence (Hogeveen, Salvi, & Grafman, 2016). Thus, while emotional abuse may be associated with social cognition due to exposure to negative attributions, emotional neglect may be linked with social cognition via broader cognitive deficits. Support for distinct pathways for emotional abuse and neglect contributing towards impaired social cognition could have significant implications in tailoring empirically supported youth social-emotional learning interventions aimed at reducing depressed mood (see Taylor, Oberle, Durlak, & Weissberg, 2017) to become more trauma-informed.

A significant strength of our study was the ability to test our aims across demographic differences. With regard to sex, we found that the effects of maltreatment were invariant for boys and girls. These findings replicate past findings on the moderating role of sex on maltreatment subtypes and depressive symptoms (e.g., Arnow et al., 2011). This furthers the argument that depression risk in females may be due to increased levels of interpersonal stressors experienced during adolescence (Hankin & Abramson, 2001) as opposed to sensitivity to maltreatment. As for race/ethnicity, we found some support for the “double jeopardy” hypothesis within the context of emotional maltreatment and anhedonia. Specifically, African-Americans reported elevated levels of anhedonic symptoms within the context of emotional abuse. Significant findings for anhedonia compared to depressed mood may be due to findings that African-Americans tend to express more anhedonic responses compared to their peers (Lu, Lindsey, Irsheid, & Nebbitt, 2017). However, we recommend using caution when interpreting these findings as the significant findings faded over time. This could mean that adolescence is a sensitive period for anhedonic responses in African-Americans exposed to emotional abuse or, and potentially more likely, a regression to the mean statistical effect was observed. Regardless of one’s interpretation of this finding, overall our study supports a “racial invariance” perspective (Widom et al., 2013) as race/ethnicity did not moderate other findings. Future research would benefit from focusing more on malleable individual differences that may moderate the maltreatment-depression relation as opposed to invariant demographic constructs.

While the diverse, large adolescent sample, and multi-wave longitudinal study design were significant strengths of our study, there are some notable limitations. First, we used a retrospective childhood maltreatment assessment. While retrospective reports of maltreatment are common and valid methods for indexing maltreatment-exposure (Kendall-Tacket & Becker-Blease, 2004), there are biases associated with this approach (Hardt & Rutter, 2004). Relatedly, we assessed maltreatment at age 18, following some of our assessments of depressive symptoms. Therefore, statements concerning the predictive validity of maltreatment is based on theoretical models positing depressive symptoms as a consequence of maltreatment. Second, our study used an unselected community sample to investigate the effects of maltreatment. While using child welfare samples have significant limitations (e.g., underestimates of reports for neglect; Proctor & Dubowitz, 2014), others caution against using community samples when investigating the consequences of emotional maltreatment (McCrory et al., 2017). Therefore, this study’s aims should be replicated within the context of a child welfare sample with substantiated experiences of child maltreatment. Finally, we only used a single, self-report measure of depression. Best practices recommend using a multi-informant approach when assessing depression to fully capture the breadth of symptoms (Klein, Dougherty, & Olino, 2005). This issue may be especially relevant with regard to our two-item assessment of anhedonia as evidenced by lower levels of internal consistency. While short, self-report measures of anhedonia are commonly used in the literature (e.g., Bennik et al., 2014), a multi-method assessment can provide insight into which aspects of anhedonia may be particularly impaired within neglect-exposed youth.

Understanding the unique effects of specific childhood adversities holds great translational promise. Currently, empirically-based assessment and intervention strategies use a cumulative approach to childhood adversities. Inherently, these models treat neglect and emotional abuse as “stressors” within an overall victimization profile, and target consequences stemming from these experiences in a similar fashion (Lawson & Quinn, 2013). However, the current study is congruent with emerging developmental traumatology models (e.g., McCrory et al., 2017; McLaughlin et al., 2014) which suggest that the consequences stemming from specific forms of childhood adversities may be distinct. Specifically, the risk and type of depression secondary to maltreatment-exposure may vary as a function of whether one experiences abuse or neglect. These developmental models are consistent with recent depression models that suggest distinct pathways leading to the onset and maintenance of depression (e.g., Luyten & Fonagy, 2017; Trew, 2011) even at the symptom level (Hasler et al., 2004). Collectively, these models show how two different forms of maltreatment may differentially or simultaneously impact certain depressive presentations in adolescents. Ultimately, understanding which traumatic events contribute to specific depressive risk pathways and manifest symptoms can pave the way for a personalized medicine approach to trauma-informed mental health prevention and intervention approaches.