|Home | About | Journals | Submit | Contact Us | Français|
The current study examined pathways in a model of desensitization, the Pathologic Adaptation Model, in adolescent males of color. Specifically, the current study examined depressive symptoms and deviant beliefs as mediators of the association between community violence exposure and subsequent violent behavior.
The current study included 250 African American (67%) and Latino (33%) male adolescents (T1 mean age = 15.32) from the Chicago Youth Development Study.
Consistent with the Pathologic Adaptation Model, results demonstrated that depressive symptoms mediated the association between the quadratic violence exposure term in middle adolescence and violent behaviors in late adolescence, but the direction of the mediation effect was dependent upon the levels of violence exposure in middle adolescence. However, deviant beliefs were not found to be a significant mediator.
Emotional desensitization effects may increase the likelihood of violence perpetration in adolescent males exposed to community violence, and the implications for future research and intervention efforts are discussed.
Exposure to community violence (ECV) has been linked with a number of maladaptive psychosocial outcomes for male adolescents of color in urban communities (Salzinger, Feldman, Stockhammer, & Hood, 2002), and a burgeoning body of research examines desensitization as an outcome of community violence exposure. Desensitization to ECV posits that youth who experience repeated exposure to violence “adapt” by suppressing emotional distress and viewing violence as normal (Cooley-Quille & Lorion, 1999; McCart et al., 2007). However, when youth become emotionally numb and normalize violence, they may lose inhibitions about using aggressive behavior, which may lead to violence perpetration and additional violence exposure (Garbarino, Kostelny, & Dubrow, 1991; Ng-Mak, Salzinger, Feldman, & Stueve, 2002; 2004). Given the dangerous consequences of engaging in aggressive behavior in some urban communities, understanding mechanisms of the link between ECV and violence perpetration is warranted for adolescent males of color.
Desensitization is consistent with the concept of habituation—when an organism’s response to a stimulus decreases over time due to repeated or prolonged exposure to the stimulus (Groves & Thompson, 1970). Exposure to a stressor will produce an initial increase in the organism’s response, followed by decrements in the organism’s response when repeatedly exposed to the same stimulus (Groves & Thompson, 1970). One model of desensitization to ECV is the Pathologic Adaptation Model (PAM; Ng-Mak et al., 2002), which states that youth begin to respond to high levels of violence with emotional numbing, evidenced by lower than expected levels of emotional symptoms, but higher levels of violent behavior (Ng-Mak et al., 2002). The PAM focuses on emotional desensitization, which occurs when an individual experiences a numbing or blunting of emotional reactions to ECV, rather than the expected strong emotional response (Funk, Baldacci, Pasold, & Baumgardner, 2004). Depressive symptoms are often used as a proxy for emotional symptoms, and the PAM suggests a curvilinear (inverted U-shaped) relationship between ECV and depression. For youth exposed to community violence, affective symptoms of sadness, emptiness, or worthlessness may be blunted as a way to cope with the traumatic nature of violence exposure (Porter, 1996). Empirical studies show support for emotional desensitization to ECV, with a curvilinear association of ECV to depression and anxiety symptoms (Gaylord-Harden et al., 2011; Mrug et al., 2008).
This emotional desensitization process appears adaptive in the short-term for decreasing depressive symptoms, but may place youth at elevated risk for callousness, violence perpetration, and additional violence exposure if it becomes habitual (Kerig, Bennett, Thompson & Becker, 2012; Ng-Mak et al., 2002, 2004). In other words, the curvilinear association between ECV and depressive symptoms may predict higher levels of aggressive behavior over time. Specifically, youth who experience a numbing of emotional reactions to violence exposure may be more likely to view violent and aggressive behavior as normative, lose inhibitions about using violent behavior, and use more dangerous self-protective behaviors (e.g., carrying a weapon; Jenkins & Bell, 1994; Gorman-Smith & Tolan, 1998; Guerra, Huesmann, Spindler, 2003; Reese, Vera, Thompson &Reyes, 2001; Spano, Rivera, & Bolland, 2010). Indeed, cross-sectional research demonstrates positive associations between exposure to violence across multiple contexts (home, neighborhood, and school), numbing of sadness, and aggressive behavior (Allwood, Bell & Horan, 2011).
Given these proposed associations, there is a need to extend existing research on desensitization by examining depressive symptoms as a mediator of the association between ECV and violent behavior. As noted above, the PAM posits that ECV results in lower levels of depressive symptoms, which in turn, lead to more violent behavior; however, the cross-sectional nature of research in this area precludes direct tests of this assertion. An exception within existing research is a recent longitudinal investigation of desensitization, which demonstrated that high levels of violence exposure across multiple contexts (home, school, and neighborhood) in pre-adolescence showed a curvilinear association to depression and anxiety symptoms at age 13 (Mrug, Madan, & Windle, 2015). In turn, these lower levels of internalizing problems predicted more violent behaviors at age 18. Though the findings were not specific to community violence, they lend critical evidence for the role of blunted affective or emotional responses to violence exposure, as strong support for desensitization is indicated only if these blunted emotional responses are directly linked to increases in aggressive behavior (Bartholow et al., 2006).
A limitation of the PAM and the literature on desensitization to ECV is that this body of work does not distinguish between various forms of desensitization (Gaylord-Harden, Bai & Simic, 2016; Mrug et al., 2015). However, youth with repeated ECV also experience increases in normative beliefs about aggression (Boxer et al., 2008), which suggests cognitive desensitization. Cognitive desensitization refers to a change in one’s belief that violence is uncommon to the belief that violence is mundane and inevitable, as a result of prolonged exposure to violence (Funk, Baldacci, Pasold, & Baumgardner, 2004). It is consistent with the concept of “normalization of violence,” which is described as a possible mediator of the relationship between community violence exposure and affective and behavioral outcomes (Ng-Mak et al., 2002). ECV activates cognitive schemas and scripts regarding violence and the more ECV that an adolescent experiences, the more scripts they develop regarding the acceptability of violence (McMahon et al., 2009).
With one exception, cognitive desensitization has not been explicitly examined in the existing research on desensitization to community violence. This cross-sectional study showed that ECV predicted more aggressive beliefs and in turn more aggressive behavior in African American youth (Boxer et al., 2008). It is important to extend this research with longitudinal designs that provide more information about the cumulative effects of ECV and the development of behavior problems over time, as it has been proposed that desensitization to violence is a gradual process occurring over time (e.g., Ferguson, 2014). Consistent with habituation, initial exposure to violence may predict cognitive aversion, but after repeated exposure, youth may develop aggression-supporting cognitions or beliefs. Also, the associations between ECV and aggression may be masking the effect of aggression that occurs before measurement of ECV (Bartholow, Bushman & Sestir, 2006), and research controlling for baseline rates of violent behavior is warranted.
The purpose of the current study was to examine emotional and cognitive desensitization effects of community violence exposure in male adolescents of color. To this aim, the current study extended prior research by examining depressive symptoms and deviant beliefs as mediators of the association between ECV and violent behavior over time. It was predicted that at high levels of ECV in middle adolescence, lower levels of depressive symptoms, and higher levels of deviant beliefs would predict more violent behavior in late adolescence. Testing longitudinal mediation models of desensitization will allow us to disentangle the complex effects of ECV on the psychological functioning of male adolescents of color, thus providing a clearer roadmap for intervention for this population.
Although there have been some successful efforts to curtail violence and crime in urban communities, male adolescents of color who reside in these communities remain vulnerable to high levels of ECV. Desensitization may be one of the most dangerous consequences of violence exposure because it is believed to lead to violence perpetration and additional violence exposure as youth begin to experience emotional numbing, view violence as normative, and lose inhibitions about using violent behavior (Garbarino et al., 1991). However, little is known about various forms of desensitization and how the phenomenon is linked to subsequent violent behavior. Further, existing studies of desensitization are either limited by cross-sectional designs, have not isolated the effects of community violence from other forms of violence, or do not focus on those with the highest rates of ECV (McCart et al., 2007; Mrug et al., 2008). Given these limitations, it is imperative that researchers devote continued efforts to advancing this construct empirically.
Adolescence is an ideal developmental stage to examine such patterns in ECV in this population, as behavioral patterns related to violence and aggression, which may emerge during childhood or early adolescence, reach their peak prevalence during this time (e.g., Tolan et al., 2003). Further, developmental neuroscience research details the changes in structural and functional aspects of brain development during adolescence, and demonstrates that areas of the brain involved in emotion regulation are still developing during adolescence (Ahmed, Bittencourt-Hewitt & Sebastian, 2015). As such, the higher order cognitive processes that impact emotion-regulation are not fully coordinated and controlled during adolescence, resulting in difficulties with mood (Steinberg, 2005). The current study builds on existing research in this area in a number of ways. First, the study focuses specifically on ECV in male adolescents of color, a group at heightened vulnerability for desensitization to ECV (Ng-Mak et al., 2002). Also, implicit in existing models is the notion that desensitization is the result of repeated exposure over time. The current study used data from multiple time points to examine the cumulative effect of ECV on desensitization outcomes. Building on prior research, it was predicted that the longitudinal association between community violence exposure and violent behavior would be mediated by depressive symptoms and deviant beliefs, providing a more direct test of proposed paths for the effects of desensitization. Specifically, hypothesis 1 predicted that at high levels of ECV in middle adolescence, lower levels of depressive symptoms would predict violent behavior in late adolescence. Hypothesis 2 predicted that at high levels of at high levels of ECV in middle adolescence, higher levels of deviant beliefs would predict more violent behavior in late adolescence.
Data for the current study were derived from an archival dataset, the Chicago Youth Development Study (CYDS), which investigated developmental pathways leading to the onset of and participation in antisocial behaviors during early adolescence. Data collection for the original longitudinal study started in 1991, and participants and their caregivers were surveyed approximately annually over five years with multi-modal assessments.
Participants were initially recruited from the fifth and seventh grades of 17 Chicago public schools. After obtaining parental consent and youth assent for teacher ratings, 1,105 males (92% of fifth- and seventh-grade males in the school) were screened with the Achenbach Teacher Rating Form (Achenbach, 1991). Potential participants were then selected for participation in the longitudinal study so that 50% of them were considered at “high risk” for the development of serious aggression on the basis of teacher ratings indicating that they were already engaging in high levels of aggressive behavior (above the 90th percentile using national norms). After this categorization, participants were randomly selected from the remainder of those screened. Additional parental consent and child assent were obtained prior to the first wave of assessment.
During the first wave of interviews, there were a total of 341 fifth and seventh grade males (range = 11–14 years, mean age = 12.41 years). Subsequent waves were collected when participants were, on average 13, 14, 15, 18 and 19 years of age (Prince et al., in press). Specifically, these youth included 208 African American and 133 Latino males residing in urban neighborhoods characterized by high violence and high poverty. Among these individuals, 62% lived in single-parent homes, 47.6% had a total family income below $10,000 per year, and 73.5% had incomes below $20,000 per year1.
Participants had been interviewed in their homes or in a mutually agreed-upon location by trained interviewers when the youth were in sixth or eighth grades. These interviews lasted between 3 to 3½ hours, and the same information was collected across participants during each wave. After a joint family interaction task, individual interviews were conducted separately with the participant and his caregiver(s). The study was successful in tracking and assessing study participants from Waves 2 to 6. The first cohort was assessed from the sixth to eleventh grades, while the second cohort was assessed from the eighth to twelfth grades. More than 73% of the youth participants completed self-report data for each of the waves of data collection. Specifically, a majority (86%) of participants completed self-report data for at least 3 out of the possible 5 assessments.
In the current study, analyses included 250 participants who completed at least one interview during Waves 2–4 and at least one interview during Waves 5 – 6 of the CYDS. Of the 250 participants included in the current analyses, 67% were African American and 33% were Latino. Consistent with other studies using the CYDS data (e.g., Brady, Gorman-Smith, Henry & Tolan, 2008), in the current analyses, Waves 2–4 from the original study were combined to represent Time 1 (average age = 15.31, SD = 1.19) and Waves 5–6 from the original study were combined to represent Time 2 (average age = 19.11, SD = 1.19).
For the current study, data were extracted from the original study for participants who completed measures assessing exposure to community violence, deviant behaviors, depressive symptoms and violent behavior at least once during Waves 2–4 (84% of those interviewed during Wave 1) and participated in at least one interview during Waves 5 and 6 (73% of those interviewed during Wave 1; 88% of those interviewed during Waves 2–4). Comparison of data from the current sample with data from those not continuing after the initial interview yielded no significant differences on initial teacher ratings of aggression, F(1, 524) = 0.57, p = .45, or official arrest records (ever arrested) as of Wave 4, X2(1, N = 298) = 0.37, p = .54. Comparisons were also made for general purposes between data from those continuing to participate and data from those not continuing for each wave on the predictors and behavioral measures. There were no significant differences in the current study variables.
Demographic information, including ethnicity, age, parental marital status, income, household size, and relationship to the target youth had been collected from the caregiver(s).
Prior to the first wave of data collection, teachers completed the Achenbach Teacher Rating Form (TRF; Achenbach, 1991) to assess participants’ aggressive behaviors. The TRF is a teacher-report measure that includes an aggressive behavior scale. Scores on the aggressive behavior scale were used in the CYDS study to determine pre-test aggression status. Participants with a score above the 90th percentile using national norms were coded as high aggression and other youth were coded as low aggression (Gorman-Smith & Tolan, 1998). In the current sample, 54% of youth were categorized as low aggression, and 46% were categorized as high aggression.
Exposure to community violence had been measured at Waves 2–5 using the Exposure to Violence Interview, a section of the CYDS Stress and Coping Interview (Tolan & Gorman-Smith, 1991). During each annual interview, participants were asked to indicate the number of times they had witnessed the following five events within the past year: 1) anyone in your family was robbed or attacked (or otherwise hurt intentionally by someone), 2) someone else you know, other than a member of your family, was beaten, attacked, or really hurt by others, 3) you saw anyone beaten up, 4) you saw anyone shot or killed, 5) you witnessed any violent crime (not counting what you have already told me about). Reported frequencies were recoded using the scale: 0 – none, 1 – once, 2 – more than once. The mean score of these items was computed to create a score for the overall frequency of witnessing community violence averaging across Waves 2–4 (Time 1) and Waves 5–6 (Time 2) separately.
The 27-item Child Depression Inventory (CDI; Kovacs, 1992) and the 21-item Beck Depression Inventory (BDI; Beck et al., 1961) were used as measures of depressive symptoms. Participants completed the CDI at Waves 2–4. The CDI is a widely used measure of youth depression consisting of 27 items. Each item contains three sentences and respondents are asked to choose the sentence within each item that most closely describes him over the past two weeks. The 27 items can be combined to provide a score that measures overall depression with higher scores indicating higher levels of depression. An overall score for Time 1 was generated by combining scores from Waves 2–4. Cronbach’s α in the current sample was .86 at wave 2. Given that the average age for participants at Waves 5–6 was 19 years old and beyond the age limit for the CDI, participants at those waves completed the Beck Depression Inventory (BDI; Beck et al., 1961). The BDI is a validated measurement scale consisting of 21 items to assess the intensity of symptoms of depression. Each item is a list of four statements arranged in increasing severity about a symptom of depression. Respondents were asked to choose the sentence within each item that most closely described him over the past two weeks. The 21 items combine to provide a score that measures overall depression with higher scores indicating more depression. An overall score for Time 2 was generated by combining scores from Waves 5–6. Cronbach’s α in the current sample was .87 at wave 5. The correlation between the CDI at Waves 2–4 and the BDI at Waves 5–6 is .21 (p < .001).
Deviant beliefs/cognitions had been assessed using the 4-item deviant beliefs subscale of the Family Relationship Scale (Tolan, Gorman-Smith, Huesmann, & Zelli, 1997). Shared deviant beliefs represents values shared by family members with regards to deviant behavior, and includes 4 items: “It’s okay to fight if the other guys says bad things about your family,” “It’s O.K. to steal something from someone who is rich and can easily replace it,” “It’s O.K. to skip school every once in awhile,” and “It’s okay to lie to someone if it will keep you out of trouble with them.” The measure computes a product score of the belief items, combining mother and adolescent report, to obtain an index of agreement among family members. This score captured the extent of endorsement by each individual, as well as the agreement across sources (Tolan et al., 1997; Gorman-Smith et al., 1998). For the present analyses, the Shared Deviant Beliefs subscale (α = .68) was used with higher scores indicating higher levels of deviant beliefs.
At each wave, youth participants completed a modified version of the Self-Report of Delinquency (SRD; Elliott, Dunford, & Huizinga, 1987). Youth participants indicated the number of times they had engaged in any of 8 violent behaviors within the past year: 1) threw objects at people, 2) became involved in gang fights, 3) hit someone with intent to harm, 4) carried a hidden weapon, 5) hurt or threatened someone for sex, 6) tried to force someone to have sex, 7) used a weapon or force to get something, 8) attacked someone with a weapon. Reported frequencies for each event were recoded (0 – none, 1 – once, 2 – more than once) and weighted for legal seriousness, ranging from an index of 4 (Class A misdemeanors, e.g., Item 1) to an index of 8 (Class 1, corresponding to serious felonies, e.g., Item 8) (Tolan, Gorman-Smith, & Henry, 2003). After summing across weighted scores for each item within wave, composite scores were created for violent behavior by averaging across Waves 2–4 (Time 1) and Waves 5–6 (Time 2) separately. Due to positive skewness of the composite scores, natural log transformations were used in the analyses.
To test hypotheses 1 and 2, that the longitudinal association between Time 1 ECV and Time 2 violent behavior would be mediated by Time 2 depressive symptoms and deviant beliefs, bias-corrected bootstrapping procedure macros in SPSS were used. Bootstrapping provides the most powerful and reasonable method of obtaining confidence limits for specific indirect effects (Preacher & Hayes, 2008). Bootstrapping methods are preferred over the Sobel test because the Sobel test makes unrealistic assumptions about the shape of the sampling distribution of the indirect effect. Bootstrapping yields k estimates of the total and specific indirect effects of X on Y with resampled data, and the bootstrap confidence interval (CI) for the population-specific indirect effect through the mediator is derived by sorting the k values from low to high (Preacher & Hayes, 2008). If the 95% CI does not contain zero, the indirect effect is significantly different from zero.
Consistent with a desensitization effect, it was expected that the association of Time 1 ECV to Time 2 depressive symptoms and deviant beliefs would be curvilinear. Therefore, the MEDCURVE macro for SPSS was used to test the proposed mediation model. MEDCURVE does not impose the constraint that the paths between variables are linear in nature and estimates instantaneous indirect effects of a causal variable (X) on the outcome variable (Y) through a proposed mediator (M) (Hayes & Preacher, 2010). Instantaneous indirect effects of X on Y through M quantifies the change in Y through M as X is changing to account for nonlinear associations in the model, such as between X and M (Hayes & Preacher, 2010). MEDCURVE generates bias corrected bootstrap confidence intervals for instantaneous indirect effects on specific values of X, typically 1 standard deviation above the mean, the sample mean, and one standard deviation below the mean (Hayes & Preacher, 2010). By utilizing the MEDCURVE macro in SPSS, the model specified that Time 2 depressive symptoms (M) were a quadratic effect of Time 1 ECV (X), controlling for average participant age at Time 1, ethnicity, and pre-test aggression status, Time 1 violent behaviors (W1), Time 1 depressive symptoms (W2). Further, Time 2 violent behavior (Y) was modeled as exponentially related to Time 2 depressive symptoms (M) controlling for average participant age at Time 1, ethnicity, and pre-test aggression status, the curvilinear effect of time 1 ECV (X), as well as the linear effects of Time 1 violent behaviors (W1) and Time 1 depressive symptoms (W2) (Hayes & Preacher, 2010).
Means and standard deviations for each of the study variables are listed in Table 1. Correlations among the study variables are also presented in Table 1. During Time 1 (Waves 2–4) of the study, 87% of youth reported exposure to violence in the past year and 64% reported exposure to more than one event.
The results indicated that the quadratic term for Time 1 ECV was significantly related to Time 2 depressive symptoms (p =.03) and that there was a significant mediational effect. However, due to the curvilinear association between ECV and depressive symptoms (Figure 1), the direction of the mediation effect depended on the levels of violence exposure at Time 1. Bootstrapping analysis revealed that at low levels of community violence exposure at Time 1, the quadratic term for ECV significantly predicted increases in Time 2 violent behaviors through its effects on Time 2 depressive symptoms (θ[.13] = .33; 95% CI: .02 to .75). Bootstrapping analysis also revealed that at high levels of community violence exposure at Time 1, the quadratic term for ECV significantly predicted decreases in Time 2 violent behaviors through its effects on Time 2 depressive symptoms (θ[.68] = −.17; 95% CI: −.42 to −.01). At moderate (θ[.40] = .08; 95% CI: −.10 to .30) levels of ECV at Time 1, increases in ECV had no effect on violent behaviors through depressive symptoms. In sum, depressive symptoms mediated the relationship between ECV and violent behavior at high and low levels of ECV at Time 1.
The model specified that Time 2 deviant beliefs (M) were a mediator between Time 1 ECV and Time 2 violent behaviors, while controlling for average participant age at Time 1, ethnicity, and pre-test aggression status, Time 1 violent behaviors, and Time 1 deviant beliefs. The results indicated that the quadratic term for Time 1 ECV was not significantly related to Time 2 deviant beliefs (p=.23) and that the indirect effect of Time 1 ECV (X) and Time 2 violent behaviors (Y) was not statistically different from zero at high (unstandardized indirect effect = .07, CI: −.05 to .28) or low (unstandardized indirect effect = −.15, CI: −.51 to .06) levels of ECV. In other words, Time 2 deviant beliefs were not found to be a significant mediator in the longitudinal association between Time 1 ECV and Time 2 violent behavior. See Figure 2.
The aim of the current study was to examine emotional and cognitive desensitization effects of community violence exposure in male adolescents of color residing in low socioeconomic urban neighborhoods. To this aim, the current study extended prior research by examining depressive symptoms and deviant beliefs as mediators of the association between ECV and violent behaviors over time. In partial support of predictions for hypothesis 1, results indicated that depressive symptoms mediated the association between ECV in middle adolescence and violent behavior in late adolescence. The mediation effect was significant at both high and low levels of ECV in middle adolescence, but it should be noted that the direction of effects was inconsistent with predictions. Specifically, at high levels of ECV in middle adolescence (1 SD above the mean), the nonlinear association between ECV and depressive symptoms predicted fewer violent behaviors in late adolescence. At low levels of violence exposure in middle adolescence (1 SD below the mean), the nonlinear association between ECV and depressive symptoms predicted more violent behaviors in late adolescence. Inconsistent with hypothesis 2, deviant beliefs did not significantly mediate the association between community violence exposure and subsequent violent behavior.
The current study extended prior research by applying a longitudinal research design to the examination of desensitization to ECV, as the existing investigations of the PAM (e.g., Boxer et al., 2008; Gaylord-Harden et al., 2011; Ng-Mak et al., 2004) are cross-sectional in nature. While these prior studies provide support for emotional desensitization to violence, some research finds that the effects of violence exposure on youth developmental outcomes are time-limited (Sharkey, 2010). By including assessments of violence at multiple time points during middle adolescence, the current findings demonstrate that middle adolescence is a particularly vulnerable time for increases in ECV, with effects on emotional and behavioral outcomes that extend into late adolescence. Given the increasing shifts in roles, responsibilities, and expectations that are occurring during this developmental period (Tolan, 2014), increases in a non-normative stressor, such as ECV, may be especially detrimental for optimal developmental outcomes of males of color during middle adolescence (Kuther & Wallace, 2003).
In particular, the current study sought to determine whether desensitization effects would mediate the association between ECV and violent behavior. Discussions of the PAM suggest that the decrease in depressive symptoms at high levels of ECV account for the subsequent increases in violent or aggressive behavior (Ng-Mak et al., 2002), but very little research has explored the role of emotional desensitization as a predictor of violent behavior in adolescence and no research has explored this association specifically for male adolescents of color. Overall, the current study demonstrated that depressive symptoms mediated the effects of ECV on violent behavior, but interestingly, when examining these associations at high and low levels of ECV, a more complex pattern emerged from the data.
Specifically, ECV during middle adolescence for youth with high levels of ECV predicted fewer violent behaviors in late adolescence through its curvilinear effect on depressive symptoms. Contrary to predictions, these findings suggest that emotional desensitization may be protective for adolescent males with high levels of ECV. Perhaps males with high levels of ECV and experience emotional desensitization do indeed “adapt” emotionally to the exposure with lower levels of depressive symptoms, as posited in the PAM (Ng-Mak et al., 2002). As a result of repeated ECV, these youth may have developed coping strategies (e.g., cognitive avoidance, distraction, substance use) that shield them from symptoms of depression that may lead to more violent behavior, such as hopelessness for the future (McMahon et al., 2009). However, due to the strong links between ECV and violent behavior in the literature, it should be acknowledged that there may be different pathways to violent behavior for youth with initially high levels of ECV that were not assessed in the current study. For example, normative beliefs about aggression (e.g., Boxer et al., 2008) have been identified as mediators of the relationship between violence exposure and violent behavior for youth exposed to high levels of violence. High levels of violence may lead to a “normalization of violence” or cognitive desensitization (Boxer et al., 2008; Colder, Mott, Levy, & Flay, 2000; McMahon et al., 2009). Perhaps this cognitive desensitization process is more of a vulnerability factor than emotional desensitization for youth with high ECV.
In contrast to high exposure, ECV for youth low in violence exposure at middle adolescence increased violent behavior in late adolescence through its curvilinear effect on depressive symptoms. While the PAM focuses on the risk of violent behavior for youth exposed to the high rates of ECV, our findings suggest that the likelihood of violent behavior is heightened for youth with initially low levels of ECV. Perhaps male adolescents of color who have maintained low levels of ECV up to middle adolescence found ways to stay protected in risky contexts during early adolescence. ECV during middle adolescence may be a novel experience for boys with initially low levels of violence, and their emotional desensitization response may reflect a trauma reaction to ECV (Porter, 1996). As noted above, when youth experience avoidance or numbing symptoms associated with PTSD, they may underreport their affective symptoms of depression (McCart et al., 2007; Mrug et al., 2008). Further, these youth may not have developed the coping or emotion regulation skills to manage their emotional desensitization reaction to increases in ECV, and as a result, they may “act out” by exhibiting externalizing behaviors.
Also, some research with African American youth suggests that ECV is associated with an attenuated salivary cortisol response (Aiyer, Heinze, Miller, Stoddard, & Zimmerman, 2014; Kleiwer, 2006; Kleiwer, 2016). Meta-analytic findings show that uncontrollable, physically threatening, and traumatic stressors, such as community violence, produce lower than-normal cortisol levels in the morning and higher-than-normal cortisol levels in the evening (Miller, Chen, Zhou, 2007). Abnormal levels of cortisol, in tandem with abnormalities in levels of norepinephrine, are associated with hyperarousal in children exposed to trauma (Kendall-Tackett, 2000). Indeed, exposure to community violence is consistently associated with more self-reported hyperarousal symptoms in African American adolescents (Fowler et al., 2009; Margolin & Gordis, 2000; Mazza & Reynolds, 1999; Paxton et al., 2004; Singer et al., 1995), and hyperarousal symptoms have been found to lead to emotional numbing or desensitization (Weems, Saltzman, Reiss, & Carrion, 2003). Future research on this topic may benefit from the inclusion of the measurement of cortisol and other biological indicators to more adequately capture the role of PTSD symptoms in the emotional desensitization process.
While the findings of the current study were inconsistent with Mrug et al.’s (2015) recent findings, the current study focused on ECV, while Mrug et al. focused on violence exposure across multiple contexts. The current results may be specific to ECV, but there is a need for additional research to fully understand whether the findings are unique to one context. Further, due to methodological constraints in the measurement of ECV, the current study focused on witnessing violence and not victimization. Although witnessing violence produced desensitization effects in the current study, victimization is more likely than witnessing to lead to violent behavior because the greater proximity to violence leads to a stronger emotional numbing response that facilitates violent reactions and provides stronger reinforcement of the effectiveness of violence as a way to solve problems (Fowler et al., 2009; Kerig et al., 2012). Future research with male adolescents of color should assess desensitization effects separately for witnessing violence and victimization.
It is unclear why community violence exposure was unrelated to deviant beliefs in the current study. Social-cognitive models of exposure to violence suggest that the activation of cognitive schemas plays an important role in the relationship between violence exposure and subsequent violent behavior (Guerra, Huesmann & Spindler, 2003; Tolan, Guerra, & Kendall, 1995). In this theoretical model, exposure to violence activates schemas that increase the normative beliefs and acceptance of violence as a form of conflict resolution (Huesmann, 1998; McMahon et al., 2009). While research exploring the links between ECV and social information processing is less prevalent than research linking other forms of violence exposure to social information processing, there are some studies that demonstrate positive associations between community violence exposure and normative beliefs about aggression (e.g., Guerra, Huesmann & Spindler, 2003; Malik, Sorenson & Aneshensel, 1997; Shahinfar, Kupersmidt, & Matza, 2001). Further, one study showed that beliefs in retaliation influence the association between violence exposure and violence perpetration (McMahon et al., 2009). It should be noted that some of the items on the deviant beliefs measure were not specific to beliefs about aggression, which may have made it more difficult to detect a possible association between ECV and deviant beliefs.
However, findings across these studies suggest that the effects of ECV on beliefs about aggression are not clear-cut. For example, one study found that violence exposure predicted increases in aggressive beliefs, but only during later elementary school years, compared to early elementary school years (Guerra et al., 2003). The authors speculate that the effects of community violence exposure on aggressive beliefs are the result of a habituation process that occurs over an extended period of time and show the strongest effects during later elementary school years. Perhaps for youth exposed to community violence since childhood, the effect on aggressive beliefs stabilizes by late adolescence. Further, deviant beliefs were assessed as the level of agreement between participants and their parents on the acceptability of deviant behaviors. Research demonstrates that this subscale is positively associated with family cohesion, and thus deviant beliefs may be inadvertently representing more closeness and communication in the family (Tolan et al., 1997), rather than a maladaptive aspect of family functioning. Thus, the responses to the deviant beliefs measure may be related to the various types of various exposure. For example, victimization to a family member may have a differential impact on deviant beliefs than exposure to someone not in the family because these questions ask about values shared by family members and family members may also be more likely to endorse these beliefs if they were also victimized.
Finally, according to the Social Information Processing Model (Dodge & Crick, 1990), there are multiple domains of cognition that may be associated with ECV. Research assessing multiple domains of social cognitions shows that different forms of community violence are associated with different types of social cognition processes (Shanhinfar et al., 2001). The inclusion of multiple types of cognitive variables is similar to media violence research that distinguishes between multiple social cognition variables, such as individuals’ perceptions of violence as the norm, decreased negative attitudes towards violence, and decreased empathy (Carnagey et al., 2007). In addition to distinguishing between various forms of violence exposure, future research on cognitive desensitization may benefit from assessing multiple components of social information processing.
The results of the current study should be considered in light of potential limitations. First, we do not have data on the countries of origin for the Latino youth and their families. Second, because the current sample is a school-based sample, the most violence-exposed or aggressive youth may have dropped out or have been expelled from the school system, and thus may have been inadvertently excluded from participation in the study. Likewise, ECV was represented by witnessing violence only and not victimization, which may have impacted the validity of high and low levels of ECV in the current study by not adequately representing the most violence-exposed youth. Also, the measure used to assess deviant beliefs was not specific to youth report and, instead, assessed the agreement among youth and parents regarding their beliefs about delinquent behavior. While this method of assessing family functioning has been validated in the current sample (Tolan et al., 1997; Gorman-Smith et al., 1998), future research in this area should consider youth self-report of deviant beliefs. Further, in order to examine the cumulative effects of ECV and consistent with other studies using the CYDS dataset (e.g., Brady et al., 2008), the current study combined data across time points, resulting in a two-wave study or a half-longitudinal design (Cole & Maxwell, 2003). While two waves of data are acceptable for examining mediation effects (Cole & Maxwell, 2003), and half-longitudinal designs have been used in prior meditational research on violence exposure in youth (e.g., McMahon et al., 2009), mediation models are best tested with at least three time points (Cole & Maxwell, 2003). In addition, while the average ages of participants at the Time 1 (15 years old) and Time 2 (19 years old) data points correspond with the accepted ages for middle (ages 15–17) and late adolescence (ages 18 and above) (Ashford & LeCroy, 2009), this approach of combining data to create time points may have masked some characteristics of the data, such as the stability over time. However, it should be noted that the design in the current study represents a clear improvement over simple cross-sectional designs that predominate the literature on desensitization to ECV. Specifically, the current study’s longitudinal, prospective design assesses violence exposure over multiple time points, which is important given that male youth of color in urban communities experience repeated violence exposure over time (Miller et al., 1999).
Finally, the generalizability of the findings may be limited to African American and Latino youth from families and communities with high levels of poverty, violence, and associated stressors. Although youth from all racial/ethnic backgrounds and communities experience violence exposure (e.g., Finkelhor, Turner, Ormrod, & Hamby, 2009), community violence occurs with more frequency in low-income communities, and thus, males of color in these communities may be more likely to experience emotional numbing than males of color in more affluent communities with isolated incidents of violence. Further, there may be resources available for African American and Latino youth in higher socioeconomic communities that mitigate the impact of ECV on outcomes.
While the current study’s focus on male adolescents of color from low-income communities limits the generalizability of the findings, the emphasis on this population is also a strength as it supports understanding the unique experiences of a group that is exposed to higher rates of community violence than other groups of youth (Zimmerman & Messner, 2013). Future research should also examine the associations between the study variables in female adolescents of color. Future research should determine whether the decreased rates of depressive symptoms in this sample are due to posttrauma emotional numbing (Allwood, Bell & Horan, 2011) or due to active attempts to suppress depressive symptoms (Flynn, Hollenstein & Mackey, 2010). Further, males may suppress depressive symptoms to facilitate their ability to navigate dangerous neighborhoods (Cassidy & Stevenson, 2005). Also, existing studies of desensitization to community violence in youth of color from urban communities are cross-sectional (Boxer et al., 2008; Gaylord-Harden et al., 2011), and there is a need for longitudinal investigations. The current study’s longitudinal, prospective design assesses ECV over multiple time points, which is important given that male youth of color in urban communities experience repeated ECV over time (Miller et al., 1999). Finally, the current study focuses on community violence, and not other forms of violence. It is well documented that ECV is one risk factor that is heavily linked to other risk factors, including other forms of violence exposure (e.g., Finkelhor, Ormrod & Turner, 2009); however, research suggests that ECV shows unique effects on youth psychosocial functioning (Overstreet & Mazza, 2003). Future research should also consider the role of relationship proximity to victims in desensitization to community violence. Existing research demonstrates that witnessing violence against family members and close friends predicted more depressive symptoms; but there were no associations between depressive symptoms and witnessing violence against an acquaintance or a stranger (Lambert, Boyd, Cammack, & Ialongo, 2012).
Overall, the examination of desensitization in the current sample potentially helps to identify specific periods of vulnerability, critical levels of ecological risk, and vulnerability factors to provide information regarding when, where, and how to intervene in the lives of young males of color exposed to community violence. Specifically, the findings suggest that the implications of emotional desensitization on subsequent perpetration of violence found in other research (Mrug et al., 2015) may be specific to youth with initially low levels of ECV. The current data point to the importance of early detection of levels of ECV during early and middle adolescence to identify youth who may be more likely to exhibit violent behavior in late adolescence and early adulthood. These long-term effects of emotional desensitization may highlight a need for intervention efforts to include emotion regulation strategies for young males of color who witness community violence. The violent behavior outcomes that were linked to emotional desensitization in the current study carry legal ramifications and efforts to reduce the likelihood of these behaviors may, in turn, reduce rates of school suspension, expulsion, and incarceration in males of color. Understanding how individual (i.e., deviant beliefs, depressive symptoms) factors may increase the risk of maladaptive outcomes in response to exposure to community violence yields information about malleable processes that can be targeted to maximize intervention efficacy/effectiveness for this population.
Although there have been some successful efforts to curtail violence and crime in low-income, urban communities, ethnic minority male adolescents who reside in these communities remain vulnerable to very high levels of exposure to community violence and violence-related outcomes. The current study elucidates the role of community violence exposure in the development of emotional and behavioral problems in low-income, ethnic minority male adolescents and demonstrates that desensitization to violence may lead to increases in violent behavior over time.
This study was supported by a grant from the National Institute for Child Health and Human Development.
1The poverty threshold for a family of three with two related children was $10,973 in 1991 (https://www.census.gov/hhes/www.poverty/data/threshld/).