Examining a Dual-Process Model of Desensitization and Hypersensitization to Community Violence in African American Male Adolescents

The purpose of the current study was to examine a dual-process model of reactivity to community violence exposure in African American male adolescents from urban communities. The model focused on desensitization and hypersensitization effects as well as desensitization and hypersensitization as predictors of aggressive behavior. Participants were 133 African American male high school students, mean age = 15.17 years, SD = 0.96. Participants completed measures of exposure to community violence, depressive symptoms, hyperarousal symptoms, aggressive beliefs, and aggressive behaviors at two time points. Community violence exposure predicted changes in aggression, β = .25, p = .004, and physiological arousal, β = .22, p = .010, over time, but not aggressive beliefs. The curvilinear association between community violence exposure and changes in depression over time was not significant, β = .42, p = .083, but there was a significant linear association between the exposure to community violence (ECV) and changes in levels of depression over time, β = .21, p = .014. Results indicated a significant mediation effect for hyperarousal on the association between community violence exposure and aggressive behavior, B = 0.20, 95% CI = [0.04, 0.54]. Results showed support for physiological hypersensitization, with hypersensitization increasing the risk for aggressive behavior. National studies highlight the extensive reach of exposure to community violence (ECV) on the lives of adolescents (Finkelhor, Turner, Shattuck, & Hamby, 2015). African American male adolescents from urban communities report disproportionately high levels of ECV (Zimmerman & Messner, 2013). As ECV increases, the likelihood of engaging in aggressive behavior also increases (e.g., Forster, Grigsby, Unger, & Sussman, 2015). Therefore, understanding mechanisms that link ECV and aggressive behavior is critical. One potential mechanism is desensitization, where adolescents “adapt” to repeated ECV by numbing or blunting affective symptoms of depression (Ng-Mak, Salzinger, Feldman, & Stueve, 2002). When youth become desensitized, they may lose inhibitions about using violent behavior (Ng-Mak et al., 2002). One limitation of the literature on desensitization to ECV has been the almost exclusive focus on desensitization of emotional responses (Mrug, Madan, & Windle, 2016). However, ECV may also lead to changes in cognitive functioning that reflect cognitive desensitization and changes to physiological functioning that reflect Correspondence concerning this article should be addressed to Noni K. Gaylord-Harden, Department of Psychology, Loyola University Chicago, 1032 W. Sheridan Rd., Chicago, IL 60660. E-mail: [email protected] Copyright C 2017 International Society for Traumatic Stress Studies. View this article online at wileyonlinelibrary.com DOI: 10.1002/jts.22220 physiological hypersensitization (Boxer et al, 2008; Fowler, Tompsett, Braciszewski, Jacques-Tiura, & Baltes 2009). These findings support dual-process theories of learning (e.g., Groves & Thompson, 1970), which assert that both desensitization and hypersensitization processes should be considered together to understand behavioral response to a stimulus. Thus, the purpose of the current study was to examine a dual-process model of reactivity to ECV in African American male adolescents that examined emotional and cognitive desensitization and physiological hypersensitization effects as outcomes of ECV, as well as mediators of the association between ECV and aggressive behavior. Community Violence Exposure in African American Male Adolescents African American male adolescents from urban communities report higher levels of ECV than other groups (Zimmerman & Messner, 2013). Across studies, estimates of exposure range from 50% to 96% of African American boys, with many exposed more than once during adolescence (Fehon, Grilo, & Lipschitz, 2001; Gaylord-Harden, Cunningham, & Zelencik, 2011; Gorman-Smith, Henry, & Tolan, 2004). One in four African American boys report victimization (e.g., having been beaten or shot at), compared with 12% of African American females (Chen, 2010). Further, experiencing the victimization 463 464 Gaylord-Harden, Bai, & Simic and homicide deaths of friends and family members peaks during mid-adolescence for African American males (Smith, 2015). For those disproportionately exposed to ECV and without protective factors in place, the likelihood of engaging in aggressive behavior is heightened (Barroso et al., 2008; Forster et al., 2015). Even after controlling for previous symptoms and other stressors, ECV is associated with increases in aggression (McCabe, Lucchini, Hough, Yeh, & Hazen, 2005). Given the dangerous consequences of engaging in aggressive behavior in some urban communities, understanding mechanisms of the link between ECV and aggression is warranted. Empirical Research on Desensitization to Community Violence One potential mechanism is desensitization to community violence. According to the pathologic adaptation model (PAM; Ng-Mak et al., 2002), desensitization is a process whereby adolescents adapt to repeated ECV by suppressing emotional distress. In this model, adolescents with high ECV show increased aggressive behavior over time, but fewer depressive symptoms. These patterns are expressed as a linear relationship between ECV and aggression, and a curvilinear (inverse U) relationship between ECV and depression (Ng-Mak et al., 2002). When youth become emotionally numb, they may lose inhibitions about using aggressive behavior (Ng-Mak et al., 2002). Male and ethnic minority youth report fewer depressive symptoms compared with females and white youth exposed to CV (e.g., Martinez, Polo, & Zelic, 2014) and may be more likely to show desensitization. Cross-sectional and longitudinal findings support a linear relation between violence exposure and externalizing behaviors, and a curvilinear effect on depressive symptoms (e.g., Mrug, Loosier, & Windle, 2008; Mrug et al., 2016). An Alternate Model of Reactivity to ECV A limitation of the literature on PAM and desensitization to ECV is the almost exclusive focus on emotional desensitization (Mrug et al., 2016), or numbing or blunting of affective symptoms of sadness. However, youth with high ECV also experience increases in normative beliefs about aggression (Boxer et al., 2008), which suggests cognitive desensitization. Because ECV activates cognitive schemas and scripts regarding violence, the more ECV that an adolescent experiences, the more scripts they develop regarding the acceptability of violence (McMahon, Felix, Halpert, & Petropoulos, 2009). Youth may also experience changes in levels of arousal that suggest physiological hypersensitization. For example, ECV is associated with alterations in salivary cortisol responses (Aiyer, Heinze, Miller, Stoddard, & Zimmerman, 2014; Kliewer, 2016; Peckins, Dockray, Eckenrode, Heaton, & Susman, 2012; Suglia, Staudenmayer, Cohen, & Wright, 2010) and more self-reported hyperarousal symptoms (Fowler et al., 2009; Margolin & Gordis, 2000; Mazza & Reynolds, 1999; Paxton, Robinson, Shah, & Schoeny, 2004; Singer, Anglin, Song, & Lunghofer, 1995) in African American adolescents. Further, the most commonly experienced symptom of PTSD in young adult African American men is physiological hyperararousal, exhibited in direct response to the threat of community violence (Rich & Grey, 2005; Smith & Patton, 2016). The current study proposed a more comprehensive model of reactivity to ECV that incorporates emotional, cognitive, and physiological responses to violence exposure. Prior research with African American adolescents found a curvilinear association between ECV and depression, but a linear association for anxiety (Gaylord-Harden et al., 2011), suggesting that youth may experience both emotional desensitization and physiological hypersensitization. According to the dual-process theory of learning (Groves & Thompson, 1970), two processes operate together to explain responses to a stimulus: habituation and sensitization. Habituation occurs when one’s response to a stimulus decreases over time due to repeated exposure to the stimulus. Sensitization occurs when one’s response to a stimulus increases due to the stimulus being strong or harmful. The combined effect of the two processes produces the behavioral response to a stimulus. Thus, the frequency of ECV may result in a habituation effect on emotions and cognitions, in which repeated exposure leads to emotional numbing (emotional desensitization) and normalization of violence (cognitive desensitization). Simultaneously, the unpredictable, personally harmful or severe nature of violence exposure may result in the sensitization of physiological reactions (physiological hypersensitization; McCarty & Gold, 1996; Smith & Patton, 2016). Further, these three reactivity processes may lead to a higher propensity of violence perpetration, such that they mediate the association between ECV and aggression. When youth become emotionally numb, begin to normalize violence, and are more physiologically aroused, they may be more likely to behave aggressively (Ng-Mak et al., 2002; Ng-Mak, Salzinger, Feldman & Stueve, 2004). The Current Study The purpose of the current study was to test a dual-process model by examining emotional and cognitive desensitization, and physiological hypersensitization to ECV over time. Although recent research on the PAM demonstrates that youth may become emotionally desensitized to ECV, dual-process theories of learning underscore the need to consider other forms of desensitization, as well as hypersensitization effects. Also, longitudinal examinations are warranted, given that African American males in urban communities experience repeated ECV and are more likely than females to show desensitization to violence (Ng-Mak et al., 2002). Further, there is a need to extend existing research by examining indicators of desensitization and hypersensitization as mediators of the association between ECV and aggressive behavior. First, consistent with prior research, we predicted that ECV would predict increases in aggression over time. Second, we predicted that ECV would show a curvilinear or inverse U association to changes in depressive symptoms and aggressive beliefs over time, providing Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies. Dual-Process Model of Community Violence 465 Residual Score for Aggressive Behavior Time 2 Depression, Hyperarousal, & Aggressive Beliefs Time 1 Community Violence Exposure Time 1 Depression, Hyperarousal, & Aggressive Beliefs c/c' a b Figure 1. Proposed model of mediated pathways. Proposed model with mediating effects of depressive symptoms, hyperarousal symptoms, and aggressive beliefs on the association between community violence exposure and aggressive behavior. Paths a and b represent indirect effects. Path c represents the total effect. Path c represents the direct effect. evidence for emotional desensitization and cognitive desensitization. Third, we predicted that ECV would predict increases in hyperarousal symptoms over time, supporting a physiological hypersensitization effect. Finally, we predicted that Time 2 (T2) depression, T2 aggressive beliefs, and T2 hyperarousal would mediate the association between ECV and increases in aggression over time (Figure 1). Method Participants Participants were part of a larger study on academic achievement from a public charter school in a large metropolitan area. The school is located in a low-income, predominantly African American community, and 85% of students are eligible for free or reduced school lunch. The consent rate for the study was approximately 59%, with 136 students participating in data collection at the two time points. The current sample includes 133 African American male adolescents in Grades 9 to 11 with complete data on the current study variables at two time points (mean age = 15.17 years, SD = 0.96). There were no differences between participants included in the current analyses (n = 133) and excluded participants (n = 3) on participant age, family structure, or the Time 1 (T1) outcome variables. Procedure The current study was granted ethical approval by the institutional review board (IRB) at Loyola University Chicago (Chicago, Illinois, USA). A recruitment letter and consent form was sent home with all. Active parental consent and youth assent were obtained for all study participants. Youth completed a packet of paper-and-pencil questionnaires in a classroom setting at two time points approximately 12 to 15 months apart. Each participant received a movie pass as compensation. Measures My Exposure to Violence Scale (MEV). The MEV (Buka, Selner-O’Hagan, Kindlon, & Earls, 1996) scale assesses lifetime and past-year exposure to violent events that have either been witnessed or personally experienced by the subject as a direct victim. This measure obtained the frequency of exposure on a 6-point scale (0 = never, 1 = once, 2 = 2–3 times, 3 = 4–10 times, 4 = 11–50 times, 5 = more than 50 times). Ten items of witnessing violence (e.g., “seen someone get attacked with a weapon”) and eight items of victimization (e.g., “been attacked with a weapon”) were provided. To minimize recall bias, past year reports of violence were used, and mean score of these items was computed to obtain an overall exposure score. Cronbach’s alpha for the overall past year exposure score was .88 at T1 and .92 at T2. The Aggression Scale (TAS). TAS (Orpinas & Frankowski, 2001) is an 11-item scale that measures the frequency of occurrence of anger and aggressive behaviors between students. To minimize recall bias, the scale requests information about behaviors during the past 7 days. Responses to each item range from 0 times to 6 or more times. A composite score using the mean of the nine items assessing aggressive behavior were used in the current study. Internal consistency estimates for this score, using Cronbach’s alpha values, in the current sample were .88 at T1 and .85 at T2. Normative Beliefs About Aggression Scale (NOBAGS). The NOBAGS (Huesmann & Guerra, 1997) is a measure that assesses children’s beliefs about the acceptability of aggression under varying conditions of provocation. Participants responded to the eight items that assessed approval of aggression “in general” when no provocations are mentioned. Items were rated on a 4-point scale (1=really wrong, 2=sort of wrong, 3= sort of OK, 4 = perfectly OK). The items are summed to form Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies. 466 Gaylord-Harden, Bai, & Simic a total score. The internal consistency estimate for the T1 total score, using Cronbach’s alpha values, in the current sample was .80, and T2 total score was .88. Child PTSD Symptom Scale (CPSS). The CPSS (Foa, Johnson, Feeny, & Treadwell, 2001) assesses posttraumatic stress disorder (PTSD) symptoms in youth who have experienced a traumatic event. Participants were asked to respond to the CPSS based on a recent experience with community violence or traumatic loss. The measure includes 17 symptom items that correspond to the DSM-IV-TR reexperiencing, avoidance, and hyperarousal symptom clusters of PTSD and ascertain the frequency of these items in the past two weeks on a 4-point scale (0 = not at all, 1 = once a week or less, 2 = 2–4 times a week, 3 = 5 or more times a week). The 17 symptom items yield severity scores for each of the three symptom clusters. The Hyperarousal subscale score, created by summing the 5 hyperarousal items and using Cronbach’s alpha, yielded internal consistency estimates of .76 at T1 and .82 at T2 in the current sample. Depression Anxiety Stress Scale (DASS). The DASS (Lovibond & Lovibond, 1995) is designed to assess symptoms of depression, anxiety, and chronic nonspecific arousal in nonclinical samples. The DASS includes seven items to assess the frequency of depressive symptoms in the past week on a 4-point Likert scale (0 = does not apply at all, 1 = applies some of the time, 2 = applies a good part of the time, 3 = applies most or all of the time). The current study utilized the depression score (e.g., “I felt that I had nothing to look forward to”) using a sum score of the seven items. Internal consistency estimates, using Cronbach’s alpha values in the current sample for depression, were .86 at T1 and .85 at T2. Data Analysis Regression analyses were used to examine Hypothesis 1 (that ECV would predict increases in aggression over time) and Hypothesis 2 (that ECV would show a curvilinear or inverse U association to changes in depression and aggressive beliefs over time). Each T2 outcome variable was regressed on its corresponding T1 variable (e.g., T2 depressive symptoms regressed on T1 depressive symptoms), and the unstandardized residual scores were saved from these analyses. The residual scores were used as dependent variables in the regression analyses and reflected change over time in the outcome variables. For the predicted curvilinear associations between ECV and depressive symptoms and aggressive beliefs in Hypothesis 2, the linear T1 ECV variable and the quadratic T1 ECV variable were entered in the model in the prediction of each outcome variable’s residual score. To test Hypothesis 3, that the longitudinal association between T1 ECV and changes in aggressive behavior would be mediated by T2 depressive symptoms, T2 hyperarousal symptoms, and T2 aggressive beliefs, bias-corrected bootstrapping procedure macros in SPSS 24 were used (Preacher & Hayes, 2008). Three separate bootstrapping analyses were conducted. For significant nonlinear associations between T1 ECV and depression and aggressive beliefs (Hypothesis 2), we proposed to use the MEDCURVE macro (Hayes & Preacher, 2010) for SPSS to test Hypothesis 3. MEDCURVE does not impose the constraint that the paths between variables are linear (Hayes & Preacher, 2010). Results Descriptive Analyses Means, standard deviations, and correlations for the study variables are listed in Table 1. For the longitudinal analyses, Time 1 (T1) is baseline data collection and Time 2 (T2) is 12–15 months later. At Wave 1, 96.2% of the sample reported ECV, with 96.2% of the sample reporting having witnessed violence and 69.9% reporting direct victimization in the past year. The average number of events experienced by participants was one event (range of 0 events to 4 to10 events), with 8% exposed to more than one event. At Time 2, 96.9% of the sample reported ECV, with 96.2% of the sample reporting having witnessed violence and 80.0% reporting direct victimization in the past year. The average number of events experienced by participants was approximately two events (range of 0 events to 4 to 10 events), with 16% exposed to more than one event. A power analysis was conducted to determine the smallest R2 value where the sample size of 133 provides sufficient (i.e., 80%) power to detect at two-tailed p < .05 in multiple regression. Consistent with our regression models, this is based on the assumption that there are two other independent variables in the model that explain 15% of the variance in the dependent variable. A sample size of 133 achieves 80% power to detect an R2 of .048 attributed to one independent variable(s) using an F test with a significance level (alpha) of .05 (Hintze, 2011). Hypotheses ECV as a predictor of aggression over time. As predicted by Hypothesis 1, T1 ECV predicted changes in levels of aggression over time, β = .25, p = .004. Also as predicted, T1 ECV predicted changes in levels of hyperarousal symptoms over time, β = .22, p = .010. See Table 2. ECV as a predictor of depression and aggressive beliefs over time. Hypothesis 2 predicted that ECV would show a curvilinear negative (inverse U) association to depressive symptoms and aggressive beliefs over time. Initial results indicated that a suppressor situation may be occurring between the linear and quadratic ECV variables in the prediction of depressive symptoms, but the test for a suppression effect (i.e., the increase of the path coefficient for the quadratic term for ECV) was not significant, z = 1.50, p = .132. Results indicated that the curvilinear association between T1 ECV and changes in depressive symptoms over time was not significant, as evidenced by a nonsignificant increase in the variance accounted for by the Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies. Dual-Process Model of Community Violence 467 Table 1 Means, Standard Deviations, and Pearson Correlations Among the Study Variables Variable Rangea 1 2 3 4 5 6 7 8 9 10 11 12 13 1. T1 ECV 0–6 – 2. T2 ECV 0–6 .57** – 3. T1 Aggression 0–6 .18 .30** – 4. T2 Aggression 0–6 .36** .41** .40** – 5. T1 Beliefs 0–32 .13 .18* .31* .23* – 6. T2 Beliefs 0–32 .07 .18 .35** .39** .46** – 7. T1 Hyperarousal 0–15 .21* .07 .40** .17* .11 .31** – 8. T2 Hyperarousal 0–15 .28** .35** .15 .33** .13 .14 .24* – 9. T1 Depression 0–21 .19** .27** .41** .30** .08 .30* .52** .29** – 10. T2 Depression 0–21 .26** .31** .23** .23** .27** .15 .24** .41** .30** – 11. Age 14–18 −.01 .04 −.06 −.11 .03 −.14 −.00 −.06 .04 .10 – 12. Grade 9–11 .03 .09 .00 −.04 .12 −.15 −.03 −.02 .04 .08 .75** – 13. Family Statusb 0–2 .02 −.05 .09 .02 −.02 −.09 −.07 −.08 −.13 −.13 .04 .03 – M 0.94 1.19 3.53 3.91 13.32 13.71 5.27 5.74 6.15 6.24 15.17 9.58 1.37 SD 0.71 0.78 3.11 2.81 4.00 4.69 3.80 4.03 5.36 5.05 0.96 0.73 0.60 Note. T1 = Time 1; T2 = Time 2; ECV = exposure to community violence; Beliefs = aggressive beliefs. aRange of possible scores. bFamily status refers to single-parent vs. two-parent family status. *p < .05. **p < .01. ***p < .001. Table 2 Linear and Quadratic Effects of Exposure to Community Violence (ECV) on Outcome Variables Outcome B SE β R2a Depression Step 1 Intercept −1.34 0.68 .04* T1 ECV 1.14 0.58 .21* Step 2 Intercept −0.32 0.89 .06* T1 ECV −1.25 1.64 −.18 T1 ECV quadratic term 1.08 0.62 .42 Hyperarousal Step 1 Intercept −1.11 0.59 .05* T1 ECV 1.21 0.47 .22* Aggressive Beliefsc Step 1 Intercept −0.18 0.63 .001 T1 ECV 0.14 0.54 .02 Step 2 Intercept −0.85 0.84 .01 T1 ECV 1.89 1.52 .27 T1 ECV quadratic term −0.70 0.58 −.32 Aggressiond Step 1 Intercept −0.81 0.35 .06** T1 ECV 0.86 0.29 .25** Note. All outcome variables represented as unstandardized residual scores. T1 = Time 1. aR2 refers to the total variance explained at each step. *p < .05. **p < .01. ***p < .001. quadratic violence term, R2 = .02, p = .083. However, there was a significant linear association between ECV and changes in levels of depression over time, β = .21, p = .014. See Table 2. Psychological symptoms as mediators of the association between ECV and aggression. Hypothesis 3 predicted that the association between T1 ECV and changes in aggressive behavior over time would be mediated by T2 depressive symptoms, T2 hyperarousal symptoms, and T2 aggressive beliefs. Given that the nonlinear associations between T1 ECV and T2 changes in depression and aggressive beliefs were not significant, the MEDCURVE macro was not used to examine the mediation effects of depression and aggressive beliefs. Instead, only bias-corrected bootstrapping procedure macros in SPSS were used for all tests of Hypothesis 3 with the linear terms for hyperarousal, depression, and aggressive beliefs (Preacher & Hayes, 2008). As predicted, T2 hyperarousal symptoms significantly mediated the relationship between T1 ECV and changes in aggressive behavior. Specifically, T1 ECV was significantly related to changes in aggression, B = 0.91, p = .003, and T2 hyperarousal, B = 1.26, p = .009, and the association between T2 hyperarousal and changes in aggression was also significant, B = 0.16, p = .003) The direct effect of T1 ECV on changes in aggression was reduced when T2 hyperarousal symptoms were in the model, B = 0.70, p = .023. Bootstrapping procedures demonstrated that this mediation effect for T2 hyperarousal symptoms was significant, B = 0.20, R2 = .12, 95% CI [0.04, 0.54]. The mediational effect for the linear term for aggressive beliefs was not statistically significant. T1 ECV significantly predicted changes in aggressive behavior, B = 0.85, p = .010, Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies. 468 Gaylord-Harden, Bai, & Simic and T2 aggressive beliefs significantly predicted changes in aggressive behavior, B = 0.14, p = .008. However, T1 ECV was unrelated to T2 aggressive beliefs, B = 0.16, p = .766. The mediational effect for depressive symptoms on the association between ECV and aggression was not statistically significant. T1 ECV significantly predicted changes in aggressive behavior, B = 0.81, p = .008, and T1 ECV significantly predicted T2 depressive symptoms, B = 1.50, p = .011, but the association between T2 depressive symptoms and changes in aggressive behavior was not significant, B = 0.02, p = .530; see Table 3). Discussion Given that the reactivity pattern to ECV may be more complex than is accounted for in prior desensitization research, the current study examined a dual-process model of reactivity to ECV that includes both desensitization and hypersensitization effects. Consistent with predictions, higher levels of ECV predicted changes in aggressive behavior over time. Inconsistent with predictions, the curvilinear association between ECV and changes in depression was not significant, but positive and linear over time. Also consistent with predictions, higher levels of ECV predicted changes in hyperarousal. Inconsistent with predictions, ECV did not predict changes in aggressive beliefs. Results also demonstrated that hyperarousal mediated the relationship between ECV and aggressive behavior. Inconsistent with predictions, neither depressive symptoms nor aggressive beliefs were significant mediators. Inconsistent with prior research, the findings do not support emotional desensitization (Gaylord-Harden et al., 2011; Mrug et al., 2008; Mrug et al., 2016; Ng-Mak et al., 2004). However, previous desensitization research focused on early adolescence or assessed the initial wave of ECV in early adolescence. In contrast, the current study focused on middle to late adolescence, with the initial wave of ECV assessed at ninth grade. Adolescents may show more depression in response to ECV than children (Fowler et al., 2009), suggesting that ECV assessed in older samples may be an unintentional proxy for greater duration of exposure to violence (Fowler et al., 2009). Thus, the emotional desensitization effects observed in prior research may be unique to late childhood or early adolescence as the frequency of ECV begins to increase. Also, this linear effect might also be due to grief and loss experiences as ECV begins to claim the lives of more significant others during late adolescence (Smith, 2015). Finally, it should be noted that the past-year rates of exposure to violence in the current sample may have been lower than in other samples of youth of the same age. Thus, the levels of ECV may not have been high enough to produce an emotional desensitization effect, and the results may not generalize to youth with higher rates of ECV. The participants were recruited from an all-male, public charter high school, and whereas the racial and socioeconomic demographics of this charter school is analogous to the demographics of the public school system, the self-selection into a charter school that focuses on acceptance into a 4-year university may suggest that the participants in the current sample possess higher levels of positive future orientation, which may lead to less engagement in behaviors that increase the risk for ECV (So, Gaylord-Harden, Voisin & Scott, in press). The results further suggest that youth in the current study reported physiological hypersensitization. The findings are consistent with prior ECV research showing that greater ECV is associated with more hyperarousal and PTSD symptoms (Margolin & Gordis, 2000; Singer et al., 1995) and extend this research by demonstrating this relationship over time. The likelihood of physiological hypersensitization in response to chronic levels of community violence may reflect the threatening, traumatic, and uncontrollable nature of community violence. Youth living with a constant threat of community violence may develop a hypervigilant neurological structure ready to manage threat at any time (Perry, 2001). It is unclear why ECV was unrelated to aggressive beliefs in the current study. According to the social information processing model (Dodge & Crick, 1990), there are multiple domains of cognition that may be associated with ECV, and different forms of ECV are associated with different types of social cognition processes (Shahinfar, Kupersmidt, & Matza, 2001). Future research on ECV may benefit from assessing multiple components of social information processing. Findings also demonstrated that hyperarousal symptoms mediated the association between ECV and aggression. Hyperarousal-induced aggression is impulsive or reactive aggressiveness that results from an individual’s difficulty in using cues to attend to incoming information and overreacting in Table 3 Mediation of the Effect of Exposure to Community Violence (ECV) on Changes in Aggressive Behavior Through Depression, Hyperarousal, and Aggressive Beliefs Model 1 Point Estimate SE BC CI (95%) Effects of depression as a mediator 0.0429 0.0723 [−0.0689, 0.2378] Effects of hyperarousal as a mediator 0.2062 0.1185 [0.0450, 0.5460] Effects of aggressive beliefs as a mediator 0.0240 0.0855 [−0.1238, 0.2246] Note. All variables represented as unstandardized residual scores. BC = bias corrected; ECV = exposure to community violence. Journal of Traumatic Stress DOI 10.1002/jts. Published on behalf of the International Society for Traumatic Stress Studies. Dual-Process Model of Community Violence 469 benign, ambiguous, or mild situations (Kerig, Becker, & Egan, 2010). Youth experiencing ECV may show a dysregulation of the hypothalamic–pituitary–adrenal (HPA) axis (Margolin & Gordis, 2004), which can lead to chronic hyperarousal and hypervigilance (Lynch, 2003; Perry, 2001). The hyperarousal associated with a dysregulated HPA axis makes it difficult to use social and environmental cues to appropriately regulate aggressive behavior. As a consequence, youth with chronic ECV may be more likely to misinterpret benign or ambiguous situations as threatening and possibly respond with impulsive or reactive aggression. Dysregulation of the HPA axis also heightens responsivity to stressors that are uncontrollable and characterized by social evaluative threat (situations in which the self may be negatively judged by others; Dickerson & Kemeny, 2004). A high social evaluative threat responsivity to victimization combined with poor self-regulation during adolescence can lead to dangerous behaviors (Dahl, 2008). The current study is not without limitations. First, physiological hyperarousal was assessed with self-report. The inclusion of physiological measures will provide more objective information on biological indicators of hypersensitization. Moreover, self-reported past-year reports of ECV were used, which may have diminished the effects of lifetime ECV. The current study utilized the specified time range, rather than lifetime exposure, consistent with the recommendation to use past-year reports to minimize recall for trauma exposure (Finkelhor, Ormrod, & Turner, 2009). Still, future research should consider delineating the effects of recent and lifetime exposure on adolescents. Further, the current study was a half-longitudinal design (Cole & Maxwell, 2003). Although two waves of data are acceptable for examining mediation effects (Cole & Maxwell, 2003), and have been used in prior meditational research on ECV in youth (e.g., McMahon et al., 2009), mediation models are best tested with at least three time points (Cole & Maxwell, 2003). However, the design in the current study represents an improvement over cross-sectional designs that predominate in the literature on desensitization to ECV. Finally, the scales utilized in the study assessed behaviors or experiences over differing time points. Specifically, violence exposure was assessed over the past year, and the outcome variables were assessed over a more recent time frame (e.g., past week). Thus, there may be more recent experiences that impacted self-reports of outcome variables. However, we chose to use the measures as they were originally designed, as the psychometric properties of the measures are based on their original format. In light of the limitations, the current study extends prior research by examining different domains of desensitization and hypersensitization that may occur in response to ECV. Similarly, this study examines the indicators of desensitization and hypersensitization as mediators of the association between ECV and aggression to delineate the complex association between ECV, symptoms, and behavior. Third, we examined the impact of ECV over time while controlling for baseline rates of outcome variables. Finally, by focusing on males, who have higher ECV than females, this study provides essential information about the mechanisms of the impact of ECV that are unique to adolescent males of color. Although policy efforts are needed to minimize crime and violence in communities to reduce adolescents’ opportunities for exposure, it is critical to devote research to understanding the deleterious effects of ECV on multiple domains of adolescent males’ lives, particularly aggression, and develop strategies to ameliorate these effects. Our results suggest that, when targeting aggressive behavior in preventive and clinical intervention efforts with adolescent males of color in under-resourced communities, it is imperative to recognize the impact of ECV in their lives and assess their history of traumatic stressors and trauma symptoms. Given the influence of hyperarousal on aggression, maximizing the effectiveness of interventions for aggressive and violent behavior will require the implementation of trauma-responsive approaches that build upon the strengths and resources of males of color to manage traumatic stress.