Introduction

Peer victimization or harassment—which includes excessive teasing, name calling, physical assault, and social rebuke—is now so endemic in our nation’s schools that it has been designated a public health concern by the American Medical Association. This concern is partly in response to a growing empirical literature over the past two decades that has documented both the pervasiveness of peer victimization from early childhood through adolescence and its negative consequences for mental and physical health (see Juvonen and Graham 2001; Elias and Zins 2003 for reviews). Youth who are chronic victims of harassment sometimes turn outward and aggress against their perpetrators. More often, they turn inward and suffer from depression, social anxiety, loneliness, and low self-esteem.

Although the negative consequences of peer victimization are now well documented, largely missing from that research is a focus on context, or the broader sociocultural milieu in which the experience of harassment unfolds. Context is defined as the physical and social settings in which individuals develop (Steinberg and Avenevoli 2000) and some contextual factors include peer groups, ethnic groups, classrooms, and schools. A good deal of peer relations research, including the study of victimization, is conducted in urban schools where multiple ethnic groups are represented, but very little of that research has systematically examined ethnicity-related context variables.

In the research reported here, we turn to ethnic context. We have a particular focus on the ethnic composition of classrooms and schools and our goal is to propose a framework for infusing ethnicity and ethnic diversity into the study of peer harassment. That framework draws on attribution theory, which is concerned with the perceived causes of events (see Weiner 1986). Here we examine how victims from different ethnic groups might construe the reasons for their plight when they are in classrooms with few or many same-ethnicity peers, and how particular causal construals influence adjustment. We hope to make a case for the importance of causal attributions as a theoretical framework and ethnic diversity as a central context variable, both of which can enrich our understanding of the dynamics of peer victimization.

Ethnicity as a Context for Peer Victimization

Picture two middle school students—one African American and one Latina/o—both attending the same middle school in an urban setting. The middle school, like many in large urban districts, is not very diverse. The population is 25% African American and 75% Latino, such that Latino students hold the clear numerical balance of power. For which student—the African American minority or the Latino majority group member—would the known consequences of harassment be more negative? It seems reasonable to think that our hypothetical African American student—the numerical ethnic minority victim—would be more vulnerable to maladjustment. That argument would be consistent with conventional wisdom about numerically less powerful groups and the reality that minority victims may have fewer same-ethnicity friends to either ward off potential harassers or buffer the consequences of harassment (e.g., Hodges et al. 1999).

Yet, consider what it must be like to be a victim and a member of a numerical majority group, like the Latino adolescent in our example. Being a victim when one’s ethnic group is numerically more powerful might be especially debilitating because that individual deviates from what is perceived as normative for their group (i.e., to be dominant). Social psychologists have used the term social misfit to describe the negative outcomes of an individual whose problem social behavior deviates from group norms (Wright et al. 1986).

In keeping with the social misfit analysis, in two prior studies we documented some of the negative outcomes associated with having both a reputation as a victim and membership in an ethnic majority group. In the first study with a multiethnic sample from one middle school, Graham and Juvonen (2002) found that African American youth who were victims and members of a numerical majority group were especially likely to feel lonely, have low self-esteem, and be rejected by their peers compared to victims from the numerical minority ethnic groups in the school. In a second study with a much larger multiethnic sample from 11 middle schools, Bellmore et al. (2004) documented that the relations between victim reputation and the outcomes of loneliness and social anxiety were heightened for students who resided in classrooms with more students from their own ethnic group (i.e., they were more likely to be members of the majority ethnic group). These studies suggest that students who are both victims and members of the majority ethnic group might be particularly vulnerable to the well-documented negative psychological consequences of victimization.

The Mediating Role of Self-Blaming Attributions

What processes explain this hypothesized vulnerability of majority group victims? One process might pertain to how victims interpret the reason for their plight. When someone is a member of the majority group, repeated encounters with peer hostility or even an isolated yet especially painful experience, might lead that victim to ask, “Why me?”. In the absence of disconfirming social cues, such an individual might come to blame their predicament on their personal shortcomings, by concluding, for example that “I’m someone who deserves to be picked on”. In the adult literature on causal explanations for rape (another form of victimization) it has been documented that attributions that imply personal deservingness, labeled characterological self-blame, are especially detrimental (Janoff-Bulman 1979; Anderson et al. 1994). Characterological self-blame describes an attribution that is internal (“it’s something about me”), stable (“things will always be that way”), and uncontrollable (“there is nothing I can do to change it”). From an attributional perspective, a self-ascription for failure to an internal, stable, and uncontrollable cause can be particularly debilitating inasmuch as individuals can feel both hopeless and helpless (Weiner 1986).

Guided by attribution principles, Graham and Juvonen (1998) developed a measure of reactions to hypothetical experiences of peer harassment, including a set of items designed to capture characterological self-blame. Middle school students with reputations as victims in that study were more likely than nonvictims to endorse characterological self-blame as the cause of peer harassment and these same victims felt more lonely and anxious at school. We also hypothesized and found partial support for a victimization → self-blame → maladjustment sequence in which self-blame mediated the relation between peer harassment and adjustment outcomes. The more students with reputations as victims blamed themselves for their plight, the more likely they were to report psychological adjustment problems.

If characterological self-blame plays this mediational role, then it provides a plausible mechanism to explain why the negative consequences of peer harassment are heightened for youth who are both victims and members of the numerical majority ethnic group. We hypothesize that victims belonging to numerically more powerful groups will be particularly vulnerable to self-blaming attributions of the characterological type. As the number of same ethnicity peers increases in one’s social milieu, it becomes less plausible to make external attributions to the prejudice of others, which can protect self-esteem and buffer mental health (e.g., Crocker and Major 1989; Major et al. 2003). Hence, a victim of harassment is more likely to conclude “it’s something about me” when her reference group is many same-ethnicity peers who are not victimized. Once endorsed, self-blame for peer harassment should then predict psychological maladjustment.

The Present Study

In the research reported here, we build on earlier findings to examine (1) the mediating role of characterological self-blame on victimization-maladjustment relations across time; and (2) the moderating role of ethic diversity on the relations between victimization and characterological self-blame. Latino and African American 6th grade participants were recruited from classrooms in middle schools that ranged in ethnic diversity. Using carefully constructed measures of ethnic diversity at the classroom and individual student level, we created three non-overlapping ethnic context groups: students who were members of the numerical majority ethnic group in their classroom, the numerical minority, or who resided in classrooms where no one group held the numerical balance of power. Peer nomination data at the classroom level were gathered in Fall of 6th grade to determine which students had reputations as victims of harassment. In Spring of 6th grade, self-report data were gathered on characterological self-blame for imagined peer harassment and two esteem-related adjustment outcomes: depressive symptoms and feelings of self-worth. Because causal construals about the causes of harassment were expected to unfold over time, the mediating role of characterological self-blame was examined using this short-term longitudinal design.

For all three ethnic diversity groups, we expected that a stronger victim reputation in the Fall would predict more depression and low self-worth in the Spring. That would be consistent with the larger empirical literature on the psychological consequences of peer harassment. However, the mediating role of self-blame was expected to be different in the three ethnic diversity groups. We hypothesized that deviation from the norm (being both victim and a member of the numerical majority group) would promote attributions for harassment to the self (it must be me). Thus the victimization → self-blame temporal sequence was expected to be strong among ethnic majority group members. Because victim reputation and membership in a numerical minority ethnic group provides an opportunity for attributions to the prejudice of others (e.g., “it could be them”), we expected weak relations between victim status and self-blaming tendencies for minority group members (i.e., no mediation). Finally, an ethnically diverse context where no one group holds the numerical balance of power is likely to promote attributional uncertainty (e.g., “it could be me or it could be them”). For students in diverse classrooms we expected both indirect (mediated) and direct effects of victim reputation on maladjustment (i.e., partial mediation).

Our analyses focused on 6th grade students during their first year of middle school, a time when the frequency and prevalence of peer victimization may peak (Nylund et al. 2007). Early adolescence and the accompanying transition to middle school is also a time of heightened concern about peer approval, finding one’s niche, and otherwise “fitting in” with the local norms (Eccles and Midgley 1989). Ethnicity and the ethnic diversity of classrooms and schools may take on added significance for early adolescents of color as they attempt to find their niche through affiliations with similar others (e.g., Hamm et al. 2005). Such youth who both encounter peer disdain and deviate from the norm might be particularly vulnerable to the causal appraisals and adjustment difficulties examined in this research.

Method

Participants

Participants were selected from a larger sample of 2003 6th grade students (909 boys and 1094 girls, M age = 11.5 years) who were taking part in a 3-year (6-wave) longitudinal study of peer relations during the middle school years. The data reported in this article were gathered during Fall (Wave 1) and Spring (Wave 2) of 6th grade. Students were recruited from 99 classrooms in 11 middle schools in greater Los Angeles that were carefully selected to yield an ethnically diverse sample, but within the constraints of a school district that is heavily Latino. Five schools were predominantly (more than 50%) Latino, three were predominantly African American, and three were ethnically diverse, with no single ethnic group constituting more than a 50% majority. Based on student self-report, the ethnic composition of the sample was 45% Latino (n = 910, primarily of Mexican origin); 26% African American (n = 511), 11% Asian (n = 212, predominantly Korean and Chinese), 9% White (n = 188), and 9% multiethnic (n = 182). There were approximately equal numbers of boys and girls within each ethnic group. To avoid confounding ethnicity with social class, all of the schools were located in predominantly low SES neighborhoods and all qualified for Title I compensatory education funding.

Both written parent consent and student assent were obtained prior to participation. For the larger longitudinal study, 75% of parents who were initially contacted returned signed consent forms. Of the forms returned, 89% of parents provided written consent for their child to participate.

Because only Latino and African American students were present in schools and classrooms that spanned the full range of diversity (i.e., neither White, Asian, or multiethnic youth were ever the majority ethnic group), the present analyses focused on a subsample of Latino and African American students. Students from these two ethnic groups were included if their classrooms had more than 50% participation (80 out of 99 classrooms) and if they had complete data at both waves. Of 1328 Latino and African American 6th grade students who met the 50% participation criteria, 1105 had complete data (491 boys and 614 girls), of which 707 (64%) were Latino and 398 (36%) were African American. Over 90% of Latino students were second generation (US born children of immigrants) or third generation and all were sufficiently proficient in English to complete written surveys.

Measures of Ethnic Diversity

Two distinct, but complementary, measures of ethnic diversity were developed and then utilized as a basis for creating the ethnic context groups used in the main analyses. We focused on the classroom rather than the school as the proximal context for measuring diversity for two reasons. First, the middle schools in our sample were organized as teams such that 6th graders spent much of their school day with the same classmates in only a few classrooms. And second, the measure of victim reputation (see below) was based on peer nominations from classroom student rosters rather than grade or school-level rosters.

Classroom Ethnic Diversity

The ethnic diversity (D C) of each classroom during Fall was computed using the following formula (Simpson 1949):

$$ D_{\text{C}}= 1 - \sum \limits_{i = 1}^{g} {p_{i}^{2} } $$

where p is the proportion of students in the classroom who are in ethnic group i. This proportion is squared \( \left( {p_{i}^{ 2} } \right), \) summed across g groups, and then subtracted from 1. Referred to as Simpson’s index of diversity, the measure gives the probability that any two students randomly selected from a classroom will be members of different ethnic groups. Possible values range from 0 to approximately 1, where higher values indicate greater diversity (i.e., more ethnic groups that are relatively evenly represented, or a higher probability that two randomly selected students will be from different ethnic groups). To illustrate how both number of groups and their relative proportion contribute to the ethnic diversity index, consider these examples: In a classroom where 75% of the students are Latino and 25% African–American, ethnic diversity = .375; in a classroom with two ethnic groups represented equally (e.g., Latino and African Americans each represent 50% of the class) ethnic diversity = .50; and in a classroom where three ethnic groups are represented equally (e.g., 33% Latino, 33% African American, 33% White) ethnic diversity = .66. In the current study, the classroom diversity index was calculated based on five groups: Latino, African American, Asian, White, and multiethnic (maximum value of D C using five groups is .80). Across the 80 participating classrooms, D C ranged from 0 (3 classes that were entirely ethnically homogenous) to .77 (M = .47, SD = .22), indicating substantial variability in diversity.

Percentage same ethnicity

While the classroom index captures the range of diversity across classrooms, it does not provide information about whether individuals are majority or minority ethnic group members in relatively low diverse classrooms. For each student, an index was therefore created that described the proportion of same-ethnicity peers in his or her classroom. The number of same-ethnicity participating classmates was divided by the total number of students (−1) in the classroom. The larger the proportion, the more likely an individual student was to be a member of the ethnic majority group.

Classifying Students into Ethnic Diversity Groups

The classroom diversity score distinguishes between students residing in high or low diverse classrooms, whereas percentage same ethnicity scores distinguish between students who are in the numerical majority or minority in their classroom. Both of these indices are important for measuring ethnic diversity because in low diverse classrooms students can be in either the majority or minority ethnic group. For example, in a classroom that is 75% Latino and 25% African American, Latino students would be classified as residing in a classroom where they are the numerical majority group whereas African American students would be labeled as residing in a classroom where they are the ethnic minority. Combining the two diversity indices allowed us to create a hybrid between an individual and a classroom-level variable that best captured the person in context. These indices were then used to assign participating students to one of three ethnic diversity groups: a high diversity group, a low diversity group in which they were a member of the numerical majority ethnic group, or a low diversity group in which they were a member of a numerical minority ethnic group.

Students were classified into the High Diversity group (n = 250) if they were in classrooms (1) with an ethnic diversity index (D C) greater than .60 (i.e., a 60% chance that two randomly selected students from the classroom would be from different ethnic groups); and (2) their percentage same ethnicity score was less than or equal to .50 (i.e., they were not a member of the numerical majority ethnic group). Students were classified as Low Diversity Majority Group (n = 681) if they were in classrooms with D C less than or equal to .60 and they had a percentage same ethnicity score greater than .50. Students were classified as Low Diversity Minority Group (n = 174) if they resided in classrooms with D C less than or equal to .60 and they had a percentage same ethnicity score less than or equal to .50.

We used selection criteria that were responsive to the range of classroom ethnic diversity in our sample, allowed us to have an adequate sample size in each group, and minimized the likelihood that students could be classified as residing in a diverse classroom, but still be in the numerical ethnic majority (i.e., diversity scores tend to increase with the presence of more ethnic groups even when one group is a numerical majority). The .60 cutoff for classroom diversity was close to one half of a standard deviation above the sample mean [.47 × 1/2 (.22)] and it approximated the difference between an evenly balanced classroom with two groups (D C = .50) and an evenly balanced classroom with five groups (D C = .80). For students in the High Diversity Group, D C ranged from .61 to .77 and percentage same ethnicity ranged from .03 to .50. For students in the Low Diversity Majority Group, D C ranged from .00 to .57 and percentage same ethnicity ranged from .52 to 1.0. For students in the Low Diversity Minority Group, D C ranged from .07 to .58 and percentage same ethnicity ranged from .04 to .46.

Variables in the Model

Victim Reputation

Peer nomination procedures were used to determine which students had reputations as victims. During Fall, participants were given a roster that contained the names of all the students in their homeroom, arranged alphabetically and by gender. Using that roster, participants were asked to nominate up to four of their classmates of either gender who fit each of three behavioral descriptions of victimization: physical victimization (“gets pushed around”), verbal victimization (“gets put down or made fun of by others”), and indirect or relational victimization (“other kids spread nasty rumors about them”). This limited nomination methodology is utilized widely by other peer relations researchers and it has been shown to reliably identify victims of peer harassment in both elementary and middle school (see Pellegrini 2001). Across the sample, participating class size ranged from 10 to 43 students. To control for variations in class size, a proportion score was created for each of the three items by totaling the number of nominations that each student received for the item and dividing by the number of nominators in their classroom.

Characterological Self-Blame

We used the measure developed by Graham and Juvonen (1998) to assess self-blame attributions for hypothetical peer victimization in the Spring of 6th grade. Participants were presented with the following scenario where they imagined that they were the target of peer harassment at school:

Imagine that you just bought your lunch after waiting in line for a long time. As you are walking away, someone in the line sticks out their foot and trips you. You’re not hurt, but most of your food spills on your clothes. Other kids in line start laughing at you.

Following the vignette, respondents rated how much they agreed with 32 statements that captured what they might think, feel, and do if the incident actually happened to them. The thoughts included six causal appraisals designed to tap characterological self-blame for peer harassment (e.g., “I know this will happen to me again” and “If I were a cooler kid, I wouldn’t get picked on”). Items were rated on 5-point scales (1 = “definitely NOT” to 5 “definitely YES”) and averaged such that higher scores reflected higher levels of characterological self-blame (for this sample, α = .80). The instrument was designed to measure causal appraisals following peer victimization as opposed to the frequency or intensity of actual victimization experiences. In previous research (Graham and Juvonen 1998), characterological self-blame was moderately correlated with a self-report measure of victimization (r = .26, p < .01), suggesting that the two instruments are measuring independent, albeit related, constructs.

Depressive Symptoms and Self-Worth

The 10-item short form of the Children’s Depression Inventory (CDI; Kovacs 1992) was used to assess depressive symptoms during Spring of 6th grade. For each item, respondents were presented with three sentences that describe “how kids might feel” and they chose the sentence that best described how they had been feeling during the past 2 weeks (e.g., “I do most things right ”; “I do many things wrong”; “I do everything wrong”). Items were scored on a 0–2 scale and averaged such that higher scores indicated more depressive symptoms (α = .82 for this sample).

The 6-item global self-worth subscale of Harter’s (1985) Self-Perception Profile for Children (SPPC) was used to assess students’ global self-worth in the Spring of 6th grade. For each item, respondents were presented with two sentences separated by the word “But,” with each statement reflecting high or low self-worth. An example item is: “Some kids are happy with themselves as a person BUT other kids are often not happy with themselves.” Students chose one of the two alternatives and then indicated whether the selected alternative is “really true for me” or “sort of true for me.” That creates a 4-point scale for each item that was summed and averaged across items, with higher scores representing higher levels of self-worth (α = .83).

Procedure

Sixth grade students were recruited from their homeroom. Because students spent several periods per day with the same classmates and a small number of teachers under the team structure, by the time of data collection in the Fall semester (October and November), students knew one other well enough to complete the peer nomination measure. Questionnaires containing the student self-report and peer nomination measures were assembled in booklet form and administered to participating students in their homerooms by a trained undergraduate and graduate student. The survey was re-administered approximately 6 months later (April and May) during the Spring semester of 6th grade (average time between waves: M = 5.89 months, SD = .95). Each classroom received $5 per participating student during each wave of data collection to be used for purchasing academic enrichment materials.

Results

Analytic Strategy

We used structural equation modeling (SEM) to test the hypothesized moderating effect of the three ethnic diversity groups on the associations between victimization, self-blame, and maladjustment. The models were tested with AMOS version 4.0 (Arbuckle and Wothke 1999) using maximum likelihood estimation.Footnote 1 As illustrated in Fig. 1, victimization during Fall was represented as a latent variable derived from the three peer nomination items. Self-blame in Spring was represented as a measured variable. Psychological maladjustment in Spring was represented as a latent variable derived from depressive symptoms and self-worth scores. Because causal appraisals of experiences with harassment are likely to unfold over time, we chose to test a longitudinal mediational model with the predictor variable assessed in the Fall and the mediator and outcome variables assessed in Spring. We did not assess Fall to Spring change in any of the variables in the model because of their relatively high stability over time (e.g., T1–T2 stability for physical victimization, self-blame, and depressive symptoms = .71, .44, and .62 respectively).

Fig. 1
figure 1

Theoretical mediational model with characterological self-blame mediating the association between victimization and maladjustment

Full size image

To evaluate our hypotheses, we examined the fit of the overall model (see Fig. 1) which included a direct path from victimization to maladjustment as well as a path between victimization and self-blame, a path between self-blame and maladjustment, and an indirect path between victimization and maladjustment. We used three different indicators of model fit: (1) chi-square, which measures “badness of fit”; (2) comparative fit index (CFI; Bentler 1990), which compares the hypothesized model to a completely uncorrelated model; and (3) root mean squared error of approximation (RMSEA; Steiger 1990), which assesses model fit adjusting for model complexity. A non-significant chi-square, a CFI above .95 (Bentler 1990), and the RMSEA below .05 (MacCallum et al. 1996) indicate good model fit. If the fit of the overall model was good, we next evaluated the strength of the victimization → self-blame and self-blame → maladjustment path coefficients as well as the strength of the indirect effect.

Preliminary Analyses

First we examined whether ethnic group membership per se (i.e., being Latino or African American, independent of ethnic diversity context) influenced the measurement or hypothesized structural model. We tested the measurement and mediational model for the full sample (n = 1,105) and then conducted multi-group SEM with Latino (n = 707) and African American (n = 398) participants as the two groups. For the full sample, the model depicted in Fig. 1 fit the data well, χ 2(7) = 10.25, p = .175; CFI = 1.0; RMSEA = .02 The measured victimization and maladjustment variables loaded significantly onto their respective latent variables, all path coefficients were significant, and the test of indirect effects was also significant (B = .176, β = .058, p < .01), suggesting that the effect of victimization on maladjustment was partially mediated by self-blame.

For the multi-group SEM, we analyzed a model in which all factor loadings and paths were constrained to be equal for Latino and African American students. The resulting model fit was then compared to the fit of the unconstrained model presented above via a chi-square difference test. The chi-square difference test was nonsignificant [χ 2 diff (13) = 5.88, p = .950], suggesting that there were no differences as a function of ethnic group membership per se in either the measurement or structural model.Footnote 2

These preliminary analyses showed that the model was similar for the two ethnic groups. That is, the psychological processes by which victimization was associated with maladjustment were similar for Latino and African American students when their data were modeled independent of ethnic diversity group. We now turn to the main analyses that test hypotheses about differences in the strength of the model in three ethnic diversity groups: African American and Latino students who were either (1) the numerical ethnic majority in low diverse classrooms, (2) a numerical ethnic minority in low diverse classrooms, or (3) in ethnically diverse classrooms.

Main Analyses: Low Diversity Majority Group

For students who were in the numerical ethnic majority group, we hypothesized that victimization would predict characterological self-blame and that self-blame, in turn, would mediate the relation between victimization and maladjustment. Table 1 presents the means and standard deviations of all measured variables in the Majority model along with the correlation coefficients among the measured variables. The top panel of Fig. 2 displays the SEM results, including the standardized and unstandardized estimates for the path coefficients and factor loadings in the model. The model fit the data well, χ 2 (8) = 9.16, p < .329, CFI = 1.0, RMSEA = .02, the direct path coefficients were each significant, and the test of indirect effects was significant (B = .204, β = .064, p < .01), suggesting that mediation by self-blame occurred. For both the Majority and Minority group models, the error variance for depressive symptoms was fixed at .01 (see Loehlin 2004 for a discussion of plausible values of error variances).

Table 1 Means, standard deviations, and correlations among measured variables for students in low diverse classrooms and in the numerical ethnic majority (n = 681)
Full size table
Fig. 2
figure 2

Standardized path coefficients and factor loadings for students in Low Diverse Majority, Low Diverse Minority, and High Diverse classrooms. Unstandardized values followed by their corresponding standard errors are presented in italics

Full size image

Low Diversity Minority Group

For students who were in the numerical minority group, we hypothesized a weak effect of victim reputation on self-blame (see Table 2 for means, standard deviations, and correlations among the measured variables). The model yielded good overall fit: χ 2(8) = 6.63, p = .578; CFI = 1.0, RMSEA = 0. However, as illustrated in the middle panel of Fig. 2 where the standardized and unstandardized coefficients are presented, although self-blame predicted maladjustment (β = .24, p < .01), victimization did not significantly predict self-blame (β = .07, p > .05). Consistent with this pattern, the test of indirect effects was not significant (B = .043, β = .017, ns). Thus, as hypothesized, for students in the numerical ethnic minority in low diverse classrooms, self-blame did not mediate the relationship between victimization and maladjustment.

Table 2 Means, standard deviations, and correlations among measured variables for students in low diverse classrooms and in the numerical ethnic minority (n = 174)
Full size table

High Diversity Group

For students in classrooms with high ethnic diversity (i.e., many different ethnic groups with relatively equal representation), we predicted that students might experience more attributional uncertainty and that there would be both direct and indirect (mediated) effects of victim reputation on maladjustment. Table 3 shows the means and standard deviations of the measured variables along with the correlations among those variables. The overall model fit the data well: χ 2(7) = 12.35, p < .089; CFI = .99; RMSEA = .06. As illustrated by the third panel in Fig. 2, where both standardized and unstandardized coefficients are presented, Fall victim reputation significantly predicted Spring self-blame (β = .16, p < .05) and Spring self-blame significantly predicted Spring maladjustment (β = .32, p < .001). Moreover, the test of indirect effects was significant (B = .171, β = .05, p < .01). Thus, like students in the numerical ethnic majority, self-blame mediated the association between victimization and maladjustment for students in diverse classrooms, although the magnitudes of the indirect effect and the path from victimization to self-blame were smaller.

Table 3 Means, standard deviations, and correlations among measured variables for students in classrooms with high ethnic diversity (n = 250)
Full size table

Discussion

This research explored the usefulness of causal attributions as a theoretical framework and ethnic diversity as a meaningful context variable for studying peer victimization as it unfolds in multiethnic urban schools. Specifically, we used principles from attribution theory to examine one process by which experiences with victimization predict psychological maladjustment and we then examined how that process was influenced by whether or not victims shared their social context with many rather than few same-ethnicity peers.

To our knowledge, this is the first study in the peer harassment literature to test a mediational model of the role of self-blaming attributions over time with both peer informants and self-reports. The use of multiple informants addressed some of the problems associated with exclusive reliance on self-report data. SEM results suggested that part of the relationship between having a reputation as a victim among one’s peers and subsequent adjustment difficulties is explained by the degree to which victims blame themselves for their plight. It is as if the victim is saying to him- or herself: “It’s something about me, things may always be that way, and there is nothing that I can do to change it”. In adult research on coping with stigma, there is a growing empirical literature guided by attribution theory on the negative mental and physical health consequences of blaming social predicaments on internal and uncontrollable causes (Major and O’Brien 2005). The time seems right to bring that literature to developmental social psychology and to studies of coping with the stigma of chronic harassment by peers.

Most unique to our approach were tests of whether a particular ethnic context variable moderated the linkages between victim reputation, characterological self-blame, and psychological maladjustment. As hypothesized, the clearest evidence of mediation emerged in the analysis of majority group members and the weakest evidence emerged for minority group members. Taken together, these ethnic context findings offer new insights into the dynamics of peer harassment during early adolescence.

Self-blame and the Ethnic Diversity Context

Being a victim and holding the numerical balance of power has its own particular vulnerability. We suggested that numerical majority group victims resemble social misfits (Wright et al. 1986) who deviate from what is perceived as normative for their group. Residing in nondiverse classrooms with many same ethnicity peers, these youth are more susceptible to causal interpretations of non-normative behavior that implicate the self (“it must be me”). Such interpretations, in turn, explain the process by which deviant behavior (victim reputation) results in psychological maladjustment. We suspect that the process of vulnerability to self-blame for non-normative behavior will get activated in many settings where social cues make it difficult to discount self-attributions as the cause of failure. With the increasing salience of ethnicity during early adolescence and the heightened importance of conforming to the local norms, causal construals about peer harassment by ethnic majority group members are especially likely to activate that process.

For minority group members who do not hold the numerical balance of power, the pathway from victim reputation to maladjustment was not explained by self-blaming attributions. For these youth in nondiverse classrooms with few same ethnicity peers, other attributions for peer harassment may have been more related to adjustment difficulties. For example, numerical minority group members—compared to their numerical majority counterparts—are in a better position to discount their own personal shortcomings as a cause of social failure in favor of external attributions such as the prejudice of majority group classmates. On the one hand, an attribution to others’ prejudice may protect self-esteem (Crocker and Major 1989). Yet, an attribution to prejudice can also take its psychological toll if it leads to loss of perceived control, anxiety, and worry about what members of the majority group think (e.g., Major et al. 2003; Schmitt and Branscombe 2002). In other words, there are trade-offs to external attributions for social predicaments that have implications for mental health.

For victims in ethnically diverse classrooms where no group held the numerical balance of power, we found evidence of both indirect (mediated) and direct effects of victim reputation on maladjustment. It is unclear at this point what these data tell us about the role of self-blame or other attributions for harassment when multiple ethnic groups are represented and the distinction between numerical majority and minority status is less salient (i.e., the most diverse classrooms). In other analyses with this same sample we found that African American and Latino students felt less vulnerable and better about themselves as school and classroom ethnic diversity increased (Juvonen et al. 2006). We suggest that ethnic diversity creates enough attributional uncertainty to ward off self-blaming tendencies. Greater diversity among ethnic groups who share the numerical balance of power discourages attributions for social failure to the self, while allowing for attributions to external factors or other causes that have fewer psychological costs. In social contexts where multiple social cues are present, attributional uncertainty can be particularly adaptive if it allows the perceiver to draw from a larger repertoire of causal schemes.

Limitations of the Research

We acknowledge the need for caution when interpreting our findings since we did not study attributions other than those related to characterological self-blame; nor did we examine mean differences between ethnic diversity groups in the magnitude of those attributions. For theoretical reasons we focused on the attributional pattern most likely to mediate victimization-maladjustment relations and for which we had hypotheses about the moderating role of ethnic diversity context. Future research will need to assess a full range of causal explanations for peer harassment and how specific attributions map on to particular adjustment consequences.

Our analysis implies that victims of peer harassment may engage in more self-blame when their perpetrators are members of their own ethnic group, a higher probability occurrence for students in the majority ethnic group. However, we did not gather data on the ethnicity of perceived perpetrators in particular school contexts nor did we manipulate ethnicity information in our harassment vignette. Because ingroup (same-ethnicity) victimization implies that a person’s membership in the group is questioned, it should have a greater emotional impact than victimization by an outgroup member (see Juvonen and Galvan 2008). At present, there are no studies of the relative rates of ingroup versus outgroup victimization in different ethnic contexts or the psychological impact of each.

We tested a particular victimization-to attribution-to adjustment sequence because it is consistent with attribution theory: when people experience negative outcomes, they often ask “why” and particular attributions have well-documented cognitive and affective consequences (see Weiner 1986). Other sequences surely are plausible, particularly if the cyclical nature of victimization is addressed. For example, although attributions are causal explanations for victimization at one point in time, it could certainly be that self-blame and its consequent depressive affect can put one at risk for subsequent victimization. Because the processes that we study are cumulative over time, new cycles of victimization-self blame-maladjustment are likely predicted by those that precede them. Longitudinal analyses with repeated assessments of victimization experiences, attributions, and adjustment, as well as the changing ethnic diversity context, will be needed to test alternative models.

Finally, we used the terms majority and minority ethnic groups to refer to relative group size, with no link to social status differences by ethnicity (e.g., Whites compared to African Americans) as is often done in the social science literature. As African American and Latino youth, all of our participants were members of historically low status groups. Because of continued housing segregation and the growing presence of Latino youth in the public schools of Los Angeles, it would have been impossible to recruit middle schools with large enough samples of White and Asian youth who comprised the numerical majority without confounding ethnic diversity context with socioeconomic status. Thus it remains to be seen whether our findings can be replicated with different (higher status) ethnic groups who hold the numerical balance of power. We also do not know how ethnic diversity shapes the experiences of more recent immigrant Latino and Asian adolescents with limited English proficiency. For these less acculturated youth, the hypothesized benefits of residing in diverse classrooms (e.g., warding off self-blaming attributions) may be less apparent. Our approach, and our measurement of ethnic diversity, underscore the importance of multiple ethnic groups that vary in their representation. Today’s multi-ethnic urban schools that vary in ethnic configurations and generational status of students provide ideal conditions for further explorations of the effects of diversity.

Implications for the Study of Ethnicity

Early in this article, we stated that the peer relations literature, including the study of victimization, has not given adequate attention to ethnicity. There are many ways that developmental psychologists can incorporate ethnicity in their analyses and the strengths and weaknesses of various approaches have been discussed (e.g.,Garcia Coll et al. 1996; Graham 1992; McLoyd 1998). For example, comparative studies that simply examine mean differences between two ethnic groups on one or more psychological variables have been criticized on numerous grounds, including likely confounds with SES or other third variables, the focus on deviance rather than difference, and ignoring within-group variability. More preferred if one’s goal is theory development are approaches that link ethnicity with the study of process, such as whether a particular set of relations is different in particular ethnic groups (see Rowe et al. 1994). Ethnic differences in the relations between harsh parenting and childhood aggression (Deater-Deckard et al. 1996) or between peer group influence and academic achievement (Steinberg et al. 1992) are examples of studying ethnicity as a moderator of psychological process in the social domain.

In preliminary analyses reported in the present research, we began with tests of whether ethnicity per se moderated the hypothesized mediational model and those analyses produced no ethnicity effects. Had we stopped there, we would have incorrectly concluded that ethnicity does not moderate the relations between victimization, characterological self-blame, and psychological adjustment. We selected ethnic diversity as our contextual variable because of its theoretical relevance to hypotheses about how individuals in contexts of varying diversity might differentially interpret the reasons for peer harassment. Surely there are many other developmental analyses in which the ethnic composition of the social setting is a relevant contextual variable, just as there are numerous other ways to situate ethnicity within a meaningful social context. We hope that the conceptual analysis and research presented here will stimulate new ways to think about ethnicity in context, ethnic diversity, and a full range of developmental outcomes during adolescence.