سطح و تغییر رفتار زورگویی در دوران دبیرستان: تجزیه و تحلیل منحنی رشد چند سطحی

Abstract The development of bullying behavior was examined across three years in a sample of 515 adolescents (46% females) from 41 classrooms. At time 1, the students were in grades 9 and 10 (mean age = 14.5 years; SD = .54). Results of a multilevel growth model showed that both baseline level and change of bullying varied significantly across individuals as well as across classrooms. At the individual level, gender, aggression and competition for social dominance were related with baseline level of bullying. Competition for social dominance and class change were additionally associated with increases in bullying over time. At the classroom level, pro-bullying behaviors were associated with higher baseline level of bullying, whereas anti-bullying behaviors with decreases in bullying over time. Finally, a cross-level interaction underlined that the link between aggression and bullying was moderated by the pro-bullying behaviors within each class. Results are discussed according to the child by environment perspective.

Introduction School bullying is a complex behavior usually influenced by multiple individual as well as social factors: it is defined as a subtype of aggressive behavior, in which an individual or a group of individuals repeatedly attacks, humiliates, and/or excludes a relatively powerless person (Olweus, 1993). Although it has been a frequent topic in literature, its development across time has not yet received the attention it deserves, especially in relation to possible individual and social predictors and their interaction. Although very few studies used longitudinal data to analyze bullying growth across time and their predictors at an individual level (Barker, Arseneault, Brendgen, Fontaine, & Maughan, 2008; Long & Pellegrini, 2003; Pepler, Jiang, Craig, & Connolly, 2008), it is not known whether changes in behavior vary between classrooms and which classroom characteristics might contribute to such variation. Through the child by environment perspective, the present study was designed to address these gaps in the literature by investigating changes in bullying behavior during high school both at the individual and classroom levels. The development of bullying across time The development of bullying across time has been mainly studied through cross-sectional data of different cohorts (Baldry & Farrington, 2000; Klomek, Marrocco, Kleinman, Schonfeld, & Gould, 2007; Langdon & Preble, 2008; Nansel et al., 2001; Olweus, 1993; Solberg & Olweus, 2003) and short-term longitudinal studies (Kim, Leventhal, Koh, Hubbard, & Boyce, 2006; Pellegrini & Bartini, 2000). Despite a relatively high stability of bullying involvement (Boulton & Smith, 1994; Kumpulainen, Rasanen, & Henttonen, 1999; Sourander, Helstela, Helenius, & Piha, 2000), studies indicate that bullying tends to increase during childhood, peak during early adolescence (6th–8th grade; Nansel et al., 2001) or middle adolescence (9th–10th grade; Langdon & Preble, 2008; Marsh, Parada, Craven, & Finger, 2004), and decline slightly during late adolescence (Guerra, Williams, & Sadek, 2011).

Results Growth curve model of bullying The model fit the data well (χ2 = 7.965; 3 df, p = .05; RMSEA = .052; CFI = .96). Results showed a significant mean of bullying intercept (t = .28, p < .001) and a not significant, although positive, bullying slope (t = .03, p = .10). Significant variance of both bullying intercept (t = .10, p < .001) and slope (t = .03, p < .05) was found. Multilevel growth curve model of bullying: unconditional model The model fit the data well (χ2 = 12.500; 5 df, p = .03; RMSEA = .050; CFI = .93). Significant within-level variance and between-level variance around the intercept mean (respectively: t = .07, p < .001; t = .03, p < .01) and around the slope mean (respectively: t = .02, p < .05; t = .01, p = .06) indicated that adolescents and classes differed in their bullying baseline levels and change over time. About 73% and 80% of the variance in initial bullying and change over time were due to systematic differences among students, whereas the rest was due to between classes differences. Multilevel growth curve model of bullying: conditional model The final model (Fig. 1) fit the data well (χ2 = 29.902; 21 df, p = .09; RMSEA = .029; CFI = .96). Final multilevel growth curve model. Unstandardized estimates. Fig. 1. Final multilevel growth curve model. Unstandardized estimates. Figure options At the individual level, gender, trait aggression, and competition for social dominance were associated with higher baseline level of bullying. A marginally significant interaction between gender and competition for social dominance showed that competition for social dominance was related to bullying behavior more strongly in males (b = .20***) than in females (b = .11*). Individual change in bullying was positively predicted by competition for social dominance and by class change. Adolescents who were high on competition for social dominance, and those who changed classes, were the ones whose bullying behavior increased more over time. Furthermore, we found a significant interaction between class change and gender. As we can see from Fig. 2 males who changed class showed an average increasing bullying trend but those who maintained the same class showed a stable trend. On the contrary, females who changed class showed an average decreasing bullying trend and those in the same class showed an average increasing trend. Individual-level interaction between class change and gender on bullying slope ... Fig. 2. Individual-level interaction between class change and gender on bullying slope (standardized estimates). Figure options At classroom level, pro-bullying behaviors showed a positive effect on the baseline level of bullying. Furthermore, grade level and anti-bullying behaviors were negatively related to change in bullying: Grade 10 classrooms showed less increase than grade 9 classrooms, and so did classrooms where anti-bullying behaviors were common. Multilevel growth curve model of bullying: cross-level interactions Only the interaction between individual trait aggression and classroom-level pro-bullying behaviors was significantly related to individual bullying intercept (t = .76, p < .001) (see Fig. 3). Results suggested that although trait aggression was predictive of bullying even in classrooms where the level of pro-bullying behaviors was low (t = .10, p < .001), this association was much stronger in classrooms where the level of pro-bullying behaviors was very high (t = .28, p < .001). Cross-level interaction between individual trait aggression and class ... Fig. 3. Cross-level interaction between individual trait aggression and class pro-bullying roles on bullying intercept (standardized estimates).