همسالان و خودسنجی قربانی و زورگویی: ارتباط افتراقی آنها با مشکلات درونی و سازگاری اجتماعی

Abstract Researchers typically employ either peer or self-reports to assess involvement in bullying. In this study, we examined the merits of each method for the identification of child characteristics related to victimization and bullying others. Accordingly, we investigated the difference between these two methods with regard to their relationship with social adjustment (i.e., perceived popularity, likeability, and self-perceived social acceptance) and internalizing problems (i.e., anxiety, depression, and self-worth) in 1192 Dutch school children, aged 9 to 12 years. Perceived popularity and likeability were more strongly correlated with peer reports than self-reports, for both victimization and for bullying others. Self-perceived social acceptance correlated equally strong with peer and self- reports of victimization. Furthermore, peer reports of bullying were also correlated with self-perceived social acceptance, whereas self-reports of bullying were not. All internalizing problems showed stronger relations with self-reports than peer reports; although only the relation between self-reported victimization and internalizing problems was of practical significance. Despite our findings indicating that using only one type of report could be efficient for examining the relation between bullying behaviors and separate child characteristics, both types of report are necessary for a complete understanding of the personal and social well-being of the children involved.

1. Introduction Bullying behavior is defined as the repeated exposure to negative actions by one or more individuals over time. During those negative actions someone (the bully) intentionally attempts to inflict injury or discomfort upon another individual (the victim). Moreover, an imbalance in perceived or actual power exists between victim and bully (Greene, 2006, Olweus, 1993, Olweus, 2000 and Salmivalli and Peets, 2009). Bullying behavior can be considered normative because it is “likely to be found in any relatively enduring human group that is difficult to leave” (Smith & Brain, 2000, p. 2). School classes are such enduring groups, and many studies about bullying focus on this particular group and corresponding developmental period. Although the scientific definition is straightforward, the operationalization of this definition to assess victimization and bullying is less clear. Investigators use different methods and different informants—peers, teachers, parents, or the children themselves—to measure children's involvement in bullying. The earliest, and still most widely used method, is self-report, followed by peer reports (Cook, Williams, Guerra, & Kim, 2010). However, the advantages and disadvantages of these two methods are still hotly debated. Whereas several researchers advocate self-reports (Olweus, 1993 and Solberg and Olweus, 2003), others prefer peer reports (Cole et al., 2006, Perry et al., 1988 and Salmivalli et al., 1996), and some have argued that both measures should be used together (Branson and Cornell, 2009, Graham and Juvonen, 1998, Juvonen et al., 2001 and Salmivalli and Peets, 2009). According to the last view, both measures have their own strengths and weaknesses and should therefore be combined to complement each other. An important additional argument is that peer and self-reports appear to relate to different problems in children, as described hereafter. Relying solely on one method, therefore, may potentially obscure problems of children involved in bullying. The present study was conducted to further investigate this differential relation between peer and self-reports and children's problems observed in the bullying literature.

3. Results 3.1. Preliminary analyses Self-reports of victimization (M = 1.28, SD = 1.23, Mdn = 1.00, Min = 0.00, Max = 5.00) and bullying (M = 0.91, SD = 1.05, Mdn = 1.00, Min = 0.00, Max = 4.00), as well as peer reports of victimization (M = .07, SD = .12, Mdn = .04, Min = .00, Max = .88) and bullying (M = .11, SD = .14, Mdn = .06, Min = .00, Max = .77), were positively skewed. These results indicate that most children were not considered as frequent bullies or victims by themselves or others. Five of the six response variables were also skewed. Likeability and social acceptance were negatively skewed, and anxiety, depression, and self-worth were positively skewed. Accordingly, we applied transformations to normalize these skewed distributed variables. Specifically, we used square root transformations for likeability; logarithm transformations for depression, social acceptance, self-worth, and self-reports of victimization and bullying, and inverse transformations for anxiety and peer-reports of victimization and bullying, depending on the severity of the skewness ( Tabachnick & Fidell, 2007). When necessary, we reflected the variable after transformation to keep the interpretation of the direction of the variable similar to the original scale. Only a few outlying scores were detected; nevertheless, due to the sample size and relative small influence of the outlying scores, no further actions were undertaken. 3.2. Regression analyses First, assumptions for regression analyses were checked. The assumption of independent observations requires that there is no association between cases due to stratified sampling procedures (e.g., clusters). Because our data were collected within classes, we calculated intraclass correlations (Snijders & Bosker, 2011). Following the guidelines described by Peugh (2010), we computed intraclass correlations and corresponding design effects on the response variables. For none of the six response variables3 did the design effect exceed the criterion described in Peugh (i.e., 2.0; the design effect is computed from the intraclass correlation and the mean number of participants within each higher-level unit), indicating that the sampling procedure did not result in violation of the independence assumption. We therefore used hierarchical multiple regression modeling instead of multilevel modeling (Peugh, 2010 and Snijders and Bosker, 2011). The assumptions of linearity, homoscedasticity, and normally distributed residuals were examined with residual plots. After the skewed variables were transformed, no serious deviations from the assumed patterns were observed; without transformations, the assumptions of homoscedasticity and normally distributed residuals were violated. Importantly, the outcomes and conclusions based on the transformed and untransformed data were similar (see the diagonal in Table 1 for the correlations between transformed and untransformed variables). Therefore, we present here the outcomes of the untransformed data as these outcomes can be more easily interpreted. Table 1. Bivariate correlational statistics for bullying reports, victimization reports, social response variables, and internalizing response variables. Variable 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 1. Peer report victimization .78* .44* .03 .12* − .44* − .42* − .40* .11* .17* − .19* 2. Self-report victimization .35* .98* .01 .25* − .26* − .25* − .36* .24* .26* − .23* 3. Peer report bullying .05 − .01 .81* .41* .41* − .42* .11* .00 .01 − .03 4. Self-report bullying .09 .25* .32* .98* .14* − .20* − .05 .11* .09 − .11* 5. Popularity − .38* − .24* .44* .15* 1.00* .23* .39* − .05 − .13* .11* 6. Likeability − .45* − .24* − .37* − .18* .24* .99* .26* − .04 − .07 .14* 7. Social acceptance − .32* − .32* .18* − .03 .38* .23* .98* − .25* − .35* .46* 8. Anxiety .08 .24* − .04 .10* − .04 − .02 − .22* .94* .49* − .37* 9. Depression .15* .24* − .04 .09 − .12* − .06 − .33* .46* .98* − .43* 10. Self-worth − .17* − .21* .02 − .10* .10* .13* .47* − .35* − .39* .98* Note. N = 1192. Correlations presented above the diagonal are based on untransformed variables, correlations below the diagonal are based on transformed variables, and correlations on the diagonal and bolded are between the untransformed and transformed variables. *p < .001. Table options Regression analyses were performed separately for victimization and bullying. The goal of these analyses was to investigate whether social adjustment would be more strongly related to peer reports of victimization and bullying and whether internalizing problems would be more strongly related to self-reports of victimization and bullying. For this purpose, we first calculated the bivariate correlations between each predictor variable (reports of victimization and bullying) and each response variable. After that, we computed Williams's t for dependent correlations (i.e., Williams's modification of Hotelling's T; Howell, 2009, Steiger, 1980 and Williams, 1959) to compare the association between type of report (peer or self-report) and the response variables. After these initial correlational computations, hierarchical multiple regression analyses were performed. Because several of the previously reviewed studies reported gender differences in the occurrence of bullying behaviors and on the relation between bullying behaviors and child characteristics, we also included gender as a predictor (in addition to peer and self-reports of victimization or bullying). Each hierarchical multiple regression analysis was performed in four steps. Following standard hierarchical regression procedures (Aiken & West, 1991), we first computed a model in which only the covariate gender was included (in model 1); thereafter, we added the main effects of peer and self-reports (in model 2), the three two-way interactions (between peer and self-report, peer report and gender, and self-report and gender in model 3), and the three-way interaction (between peer report, self-report, and gender in model 4). At each step, we examined whether the new, more complex model was an improvement relative to the previous model. Due to our relatively large sample size, increases in explained variance easily reached statistical significance. Moreover, the large number of tests performed increased the risk of capitalization on chance. Therefore, we chose a relatively stringent alpha level (i.e., α = .001) to test each model's significance. Moreover, because for none of the response variables did model 3 or 4 add more than 1.5% explained variance (i.e., ∆R2 < .015), we first present the models solely containing the variables' main effects (models 1 and 2). In the case of significant interaction effects (models 3 and 4), we subsequently present the meaning of that interaction effect for the prediction of the response variables. 3.2.1. Gender Main effects of gender (model 1; dummy coded: boys = 0, girls = 1) were found for the response variables likeability (r = .15, p < .001), anxiety (r = .14, p < .001), and self-worth (r = − .11, p < .001). Girls were generally more liked and had more symptoms of anxiety, whereas boys had higher self-worth. Additionally, boys were more often nominated as bullies (r = − .29, p < .001) and reported themselves to bully others more often (r = − .18, p < .001). For the occurrence of peer- and self-reported victimization, perceived popularity, self-perceived social acceptance, and depression, no gender differences were found. 3.2.2. Victimization The correlation between peer and self-reports of victimization amounted to .44 (p < .001). Examined separately, both peer and self-reports were correlated with the social response variables perceived popularity, likeability, and social acceptance (see Table 1). These findings indicate that children who received more peer reports of being victimized, or more often reported being victimized, were less popular, less liked by peers, and lower in self-perceived social acceptance. A comparison of these associations with the social response variables revealed that peer reports were more strongly associated than self-reports with perceived popularity (Williams's t = 6.50, p < .001) and likeability (Williams's t = 5.73, p < .001), whereas peer and self-reports' association with social acceptance did not significantly differ from each other (Williams's t = 1.44, ns). Adding both peer and self-reports to the model in which only gender was included as a predictor, substantially improved the predictive power of the model (perceived popularity: ∆R2 = .21, ∆F = 155.23, p < .001; likeability: ∆R2 = .21, ∆F = 158.42, p < .001; social acceptance: ∆R2 = .20, ∆F = 149.24, p < .001). Inspection of the squared semi-partial correlations (i.e., sr2s see Table 2) indicated that this improvement was mainly due to peer reports. Only for social acceptance, self-reports also uniquely explained part of the variance. Table 2. Regression statistics for the models containing gender (0 = male, 1 = female) and reports of victimization as predictors. Response variable Predictor B SE sr2 t Popularitya Gender − .10 .05 .00 − 2.05 Peer report − .41 .03 .14 − 14.47* Self-report − .08 .02 .01 − 3.55* Likeabilitya Gender .31 .05 .02 6.11* Peer report − .41 .03 .14 − 14.65* Self-report − .08 .02 .01 − 3.54* Social-acceptanceb Gender − .07 .03 .00 − 1.99 Peer report − .19 .02 .07 − 10.15* Self-report − .13 .02 .05 − 8.87* Anxietyb Gender .15 .03 .02 4.64* Peer report .01 .02 .00 0.63 Self-report .11 .01 .04 7.43* Depressionb Gender .07 .02 .01 2.95 Peer report .04 .01 .01 2.78 Self-report .08 .01 .04 7.38* Self-worthb Gender − .13 .03 .01 − 3.73* Peer report − .07 .02 .01 − 3.62* Self-report − .09 .02 .03 − 6.12* Note. sr2 = squared semi-partial correlation. aPeer-reported variables. bSelf-reported variables. *p < .001. Table options Only for perceived popularity, the inclusion of the three-way interaction term significantly increased the explained variance (∆R2 = .01, ∆F = 21.61, p < .001). The inclusion of the three-way interaction also yielded a significant two-way interaction between peer report and gender (t = − 4.65, p < .001, sr2 = .01) and reduced the magnitude of the main effect of both peer reports (t = 11.43, p < .001, sr2 = .09) and self-reports (t = − 2.80, p = .005, sr2 < .01). Simple slopes analyses were performed for peer-reported victimization on perceived popularity, when self-reports (one standard deviation below the mean [low] versus one standard deviation above the mean [high]) and gender were taken into account. In general, peer-reported victimization was associated with lower perceived popularity (all bs < − .35, ts > 8.47, ps < .001), which was in line with the previously described main effect. Only for girls who did not report themselves as victims, this relation changed. For these girls, the level of peer-reported victimization was not significantly associated with perceived popularity (b = − .07, t = − 0.88, ns). Peer and self-reports also correlated with the internalizing problems anxiety, depression, and self-worth (see Table 1). Children who received more peer nominations of victimization, and children who reported being more frequently victimized, experienced more symptoms of anxiety, depression, and had a lower self-worth. Self-reports were more predictive than peer reports for anxiety (Williams's t = 4.36, p < .001) and depression (Williams's t = 3.03, p = .002), but both reports did not significantly differ in their association with self-worth (Williams's t = 1.34, ns). Adding both reports to the model in which only gender was included, significantly improved the prediction of anxiety (∆R2 = .05, ∆F = 35.07, p < .001), depression (∆R2 = .07, ∆F = 46.31, p < .001), and self-worth (∆R2 = .06, ∆F = 40.21, p < .001). This improvement was mainly due to self-reports; the unique contribution of peer reports to the explanation of internalizing problems was negligible (see sr2s Table 2). Model 3 only slightly improved the prediction of depression (∆R2 = .01, F = 5.96, p < .001). Although the inclusion of the two-way interactions modestly improved the model, none of the separate two-way interactions was significant at the .001 level. Moreover, inspection of the simple slopes indicated that—at most—the positive relation between self-reports and depression was slightly stronger for peer-reported victimized girls. 6.2.3. Bullying For bullying, the correlation between peer and self-reports was .41 (p < .001). Both peer and self-reports of bullying were correlated with the social response variables perceived popularity and likeability (see Table 1). The more a child was nominated by peers as a bully, and the higher the self-reported frequency of bullying, the higher the child's perceived popularity, and the lower the child's rating on likability. Consistent with these findings, peer and self-reported bullies are generally considered as relatively popular but not very much liked. Furthermore, peer reports were correlated with self-perceived social acceptance; more peer reports as a bully were associated with children feeling themselves less socially accepted. Self-reports of bullying, however, did not correlate with self-perceived social acceptance. Comparing the associations with the social response variables showed that peer reports of bullying correlated more strongly with all social response variables than did self-reports (perceived popularity: Williams's t = 9.35, p < .001; likeability: Williams's t = 7.64, p < .001; social acceptance: Williams's t = 5.13, p < .001). When both types of reports were added to the model originally containing only gender as a predictor, the new model (i.e., model 2) explained significantly more variance in perceived popularity (∆R2 = .19, ∆F = 144.04, p < .001), likeability (∆R2 = .17, ∆F = 126.11, p < .001), and social acceptance (∆R2 = .03, ∆F = 17.41, p < .001). Peer reports of bullying were independently related to all three social response variables and explained substantial amounts of variance in popularity and likeability, and a marginal amount of variance in social acceptance (see sr2s Table 3). The unique association between self-reports of bullying and social adjustment was negligible. Inclusion of the interaction effects did not substantially improve the prediction of perceived popularity, likeability, or social acceptance (all ∆R2s ≤ .01, ps > .001). Table 3. Regression statistics for the models containing gender (0 = male, 1 = female) and reports of bullying as predictors. Response variable Predictor B SE sr2 t Popularitya Gender .13 .05 .00 2.40 Peer report .46 .03 .18 16.26* Self-report − .01 .03 .00 − 0.31 Likeabilitya Gender .04 .05 .00 0.73 Peer report − .41 .03 .14 − 14.38* Self-report − .05 .03 .00 − 2.05 Social-acceptanceb Gender − .05 .04 .00 − 1.22 Peer report .12 .02 .02 5.50* Self-report − .07 .02 .01 − 3.71* Anxietyb Gender .17 .03 .02 5.11* Peer report − .02 .02 .00 − 1.24 Self-report .08 .02 .02 4.81* Depressionb Gender .08 .03 .01 3.37* Peer report − .02 .01 .00 − 1.38 Self-report .05 .01 .01 4.11* Self-worthb Gender − .17 .04 .02 − 4.54* Peer report − .00 .02 .00 − 0.07 Self-report − .08 .02 .02 − 4.27* Note. sr2 = squared semi-partial correlation. aPeer-reported variables. bSelf-reported variables. * p < .001. Table options The correlation between peer and self-reports of bullying and the internalizing problems anxiety, depression, and self-worth was low in magnitude. Nevertheless, the correlation between self-reported bullying and the three internalizing problems was statistically significant. Moreover, the correlation coefficients between self-reported bullying and internalizing problems significantly deviated from those same coefficients for peer reports (anxiety: Williams's t = 3.52, p < .001; depression: Williams's t = 2.55, p = .01; self-worth: Williams's t = 2.55, p = .01). Although the addition of both reports of bullying to the model containing only gender as a predictor improved the prediction of the internalizing response variables slightly (all ∆R2s < .02, ps < .001), the relation between reports of bullying and internalizing response variables was weak (see Table 3). The addition of the interaction effects (model 3 and 4) did not improve the model (all ∆R2s ≤ .01, ps > .001).