Canadian Journal of Counselling I Revue canadienne de counseling I 2002, Vol. 36:4 281 Evaluation of An Ecological Program to Reduce Bullying in Schools Leila Rahey Wendy M. Craig Queen's University ABSTRACT Recently researchers have made efforts to reduce the incidence of bullying in the schools using a systemic or Whole School Approach. Reports on the efficacy of antibullying programs have produced variable, but promising, results ( Elsea & Smith, 1998; Olweus, 1993). The present study examined the short-term outcome of an antibullying program by comparing children (grades 1-8) attending a school involved in the program to chil dren attending a school not receiving the program. Results indicated that bullying did not decrease four months after the antibullying program was introduced. There were, however, improvements reported in the level of victimization, peer isolation, and per ceptions of school safety in older children (grades 3-8). Contradictory findings were found for younger children (grades 1 and 2). Implications and limitations of the study are discussed from a systemic perspective. RÉSUME Dans les dernières années, les chercheurs ont essayé de réduire le phénomène de l'intimidation dans les écoles en utilisant une approche systèmique ("Whole School Ap proach"). Les résultats des démarches prouvant l'efficacité des programmes contre l'intimidation varient mais sont prometteurs (Elsea & Smith, 1998; Olweus, 1993). La présente étude a examiné le résultat à court terme d'un programme contre l'intimidation en comparant des enfants (de 1ère année à la 8e année) d'une école impliquée dans le programme à une école comparative. Les résultats on démontré que l'intimidation n'avait pas diminué 4 mois après que le programme fut amorcé. On a noté, cependant, chez les enfants plus âgés (3e à 8e année) des améliorations dans les niveaux de victimisation, dans l'isolement des pairs, et une perception de l'école comme étant plus sécuritaire. Les chercheurs ont trouvé des résultats contradictoires chez le groupe d'enfants plus jeunes (1ère et 2e année) et entre les garçons et les filles. Nous explorons les implications et limites de la présente étude en utilisant une perspective systèmique. Peer victimization is an area that recently has received significant attention. Studies indicate that 15-20% of children are involved in bullying as either the This research was supported by a grant from the Advisory Research Council at Queen's University and the Social Sciences and Humanities Research Council. An earlier version was presented as a poster at the annual meeting of the Society for Research in Child Development Albuquerque, NM, 1999. We would like to thank Carol Brule, Claire Crooks, Isabel Fearon, Kathy Henderson, Adrian Laxton, Corey MacKenzie, Nicole Rielly for their leadership in the intervention, and the students who participated in this research. 282 Leila Rahey and Wendy M. Craig aggressor or the victim (Craig, 1998; Olweus, 1993; Rigby & Slee, 1993). Re searchers have found both immediate and distal problems are associated with victimization and bullying. Repeated victimization is associated with social anxi ety, loneliness, depression, poor self-esteem, and unhappiness at school (Austin & Joseph, 1996; Craig, 1998). Moreovet, depression in adulthood, and in tare incidents suicide, are related to severe peer victimization (Olweus, 1993; Smith & Shatp, 1994). Children who bully others report elevated levels of depression, unhappiness at school, and family conflict (Oliver, Oaks, & Hoover, 1994). Olweus (1993) reported that childten involved in bullying in grades 6 through 9 were four times as likely to have criminal convictions by early adulthood. Bully ing also affects peers who may witness the aggression and the resulting distress experienced by the victim (Olweus, 1993). Bullying problems are pervasive and effect all members of the school environment. The negative prognosis for victims, bullies, and theit peets necessitates intet- vention. The ptevailing approach in bullying interventions utilizes a systemic design that tatgets the school environment and its community (Smith & Shatp, 1994). Whole School programs reduce bullying by effecting change at each level of the system. These programs operate at the levels of the community (parents), the whole school, the playground, each classroom, and individual students. Ob jectives of these programs include: increasing awareness among school members; engaging school personnel, parents, and peers; developing a school policy with explicit rules against bullying behaviout; and providing assistance to victims. The goal is to implement a climate of "positive peer pressure" within the school that condemns bullying and commends prosocial behaviour within a supportive at mosphere (Garrity, Jens, Porter, Sager, & Short-Camilli, 1996). The theoretical basis of Whole School programming derives from research on the mechanisms that reinforce, maintain, and terminate bullying behaviours (Olweus, 1993; Smith & Sharp, 1994). Bullying is embedded in a larger context beyond the bully-victim dyad (Craig &c Pepler, 1995; Salmivalli, Lagerspetz, Bjorkqvist, Osterman, & Kaukiainen, 1996; Twemlow, Sacco, & Williams, 1996). Peers are often involved in bullying situations as active participants or as passive on-lookers (Craig & Pepler, 1995). Bullying may be ttansmitted through the peer group when children with insecure peer status "model" the behaviours of a "powerful," aggressive child (Olweus, 1993). Thus, bullying may act as a social contagion whereby children observing these actions later engage in bully ing themselves. Twemlow et al. (1996) suggest that witnessing bullying may vi cariously victimize other children, sending the message that certain behaviours result in negative consequences, and thus decreasing the possibility of peer intet- vention. Bullying can permeate through the peer network and leave children feeling unsafe and insecure in their daily surroundings. The implication for in tervention is to change the behaviours and actions of the peers. School climate influences negative peer interactions and consequently, it needs to be addressed in an antibullying program. Olweus (1993) found that school factors such as teachers' attitudes toward bullying, the amount of supervi sion, and school policies regarding violence contributed to decreasing bullying. Reduce Bullying in the Schools 283 When teachers view bullying as a serious problem, encourage positive behaviour, and intervene in a systematic fashion, reports of bullying decreased. By creating an atmosphere of adult concern, and improving communication between teach ers and students, students can communicate when someone is being victimized (Smith & Sharp, 1994). Improving the frequency of intervention by teachers also can be accomplished by increasing teacher availability. Olweus (1993) found that increasing teacher density on the play ground led to decreased bullying. The issue of supervision may be two fold: low levels of supervision decrease the chances that children who are bullying others will be identified; and low supervision may be interpreted by students that teachers are not concerned about bullying. Teacher attitudes and behaviours may be altered through an implementation model that targets the school policy (Pepler et al., 1993; Smith & Sharp, 1994). Schools with the best long-term outcomes were those with consistent routines of responding to bully ing. The administration needs to set the example that bullying is taken seriously and will not be tolerated. Principals play a significant role in influencing the efficacy of antibullying programs (Farrington, 1993). To date, researchers using systemic models in bullying prevention programs report positive results in reducing incidents of bullying (Pepler et al., 1993; Sharp & Smith, 1991). Eight months after introducing a systemic antibullying program, Pepler et al. (1993) reported that victimization decreased in children in grades 3 through 8. Olweus (1993) reported a 50% decrease in bullying after the introduction of antibullying programs. The present study extended previous re search by evaluating the effectiveness of an antibullying program implemented at an elementary school (i.e., grades 1-8). Researchers have not examined children who are not able to read, yet they have included these children in their interven tions. A second goal was to examine the intervention effects, with a considera tion of sex and age differences. Earlier intervention efforts have been systemic in their design; however, re searchers have examined only the individual characteristics of bullies and victims. We examined (a) individual problems with bullying and victimization, and (b) peer group and school environment processes utilizing a systemic interactional model with an evaluation at each level of the intervention. The program aimed to create a supportive and safe environment within the school and establish firm limits and consistent consequences for bullying. Specific goals included increas ing awareness, empathy and supervision, invoking the peers to speak against bul lying and support others when they are victimized, and formulating a clear set of rules against bullying. METHOD Participants Participants at one Intervention and one Comparison school were children in grades one through eight. Questionnaires were completed by 240 children (114 284 Leila Rahey and Wendy M. Craig boys and 126 girls) at the Intervention school and 251 children (123 boys and 128 boys) at the Comparison school. Drop outs in the study from Time 1 to Time 2 were due to children missing school. A two-way analysis of variance re vealed no significant differences on scores of bullying and victimization between schools on those that dropped out of the study from Time 1 to Time 2. Parents and teachers of the students completed questionnaires on bullying, victimiza tion, and internalizing and externalizing behaviours of their children prior to the program and at the end. Questionnaires were completed by 23 teachers and 184 parents. Intervention and Comparison schools were selected based on an expressed interest in reducing bullying. The two schools were comparable on the school programs, number of students, family composition (78% living with both natu ral parents, 5% living with one natural parent and a step-parent, 15% living with one natural parent, and 0.9% living in an "other" family configuration), race of the children (97% White, 0.3% Black, 1% First Nations and, 1% Asian), and age of children (M^ = 9.4 years). t INSTRUMENTS Child Questionnaires Bully/victim questionnaire. A shortened version of the Olweus (1993) ques tionnaire was employed to assess perceptions of being bullied and bullying oth ers. This questionnaire is the most commonly employed measure that assesses bullying. The scales included: 1. Standardized bullying/victimization self-report. A sum total of the items on bullying and victimization in the last five days and in the last six weeks was calcu lated. Scores were standardized and higher scores indicated higher levels of bully ing and victimization. A similar scale was given to parents; hence, bullying and victimization reports for parents were computed. For the self and parent scales, Cronbach's • ranged from .70 - .79. 2. Frequency and severity of victimization, bullying and group bullying. Items included the frequency that each child bullied, bullied with a group or was bul lied in six different school locations (e.g., on the bus). Each item was answered on a five-point scale, with high scores indicating more severity and frequency. Utilizing the same scale format, severity of seven types of victimization (e.g., "Being teased in an unpleasant way") was summed. 3. Content of victimization and the number oflocations where bullying occurred. Items examined the frequency of bullying, group bullying, victimization, and the severity of victimization. The total number of places where bullying, victimiza tion, and group bullying occurred, and the number of ways children were victim ized were calculated. The total possible score was six for number of places of bullying and victimization scales, and seven for the content of victimization scale. (Cronbach's a for all items ranged from .68 - .99). Reduce Bullying in the Schools 285 Peer Questionnaires Sociometría scale. (Coie & Dodge, 1988) Nomination scores involved solicit ing children's friendship choices. Students also nominated children in their class who they believed bullied others and were bullied by others. Proportions and standardized scores were calculated within each class for bullying, victimization and number of best friends. Parent and Teacher Questionnaires Child Behavior Checklist - Parent Rating Form (CBCL-PRF) (Achenbach & Edelbrock, 1983) and Teacher Rating Form of the Child Behavior Checklist (CBCL-TRF) (Achenbach & Edelbrock, 1991). These standardized instruments are the most widely used measures of child psychopathology. Only the anxious (14 items), social withdrawal (9 items), somatization (12 items), delinquency (11 items), and aggression (20 items) subscales of the parent form were used. Alpha levels ranged from .62 - .88 on all subscales for Time 1 and 2. Only the anxious (18 items), delinquency (9 items), and aggression (25 items) subscales for the teacher form were used. Alpha values for anxious, delinquency, and ag gression subscales ranged from .70 - .97 for Time 1 and 2. Two items were added to the Teacher Report Form: "is bullied by others" and "bullies others." Program Integrity Measures Administrative data. Program leaders completed weekly checklists to deter mine program adherence. Approximately 80-85% of the material was covered in each class. Material that was missed was included in following sessions increasing program integrity to 90%. There was missing data for two classes. Homework assignments were completed 54% of the time. PROCEDURE Program The twelve-week program was based on the Bully Proofing Your School pro gram (Garrity et al., 1996) which is designed to increase understanding of bullying and decrease the incidence of bullying. Program components included: school- wide programs including a psycho-educational program implemented within each classroom, a Peer Mediation program, and groups for children referred for involve ment in bullying and victimization. The program was modified for children in grades 1 and 2 so that the terminology, techniques, and tools were age appropriate. Testing was done at Time 1, two weeks before the program began, and at Time 2, one week after the program ended. School-wide programs. The psycho-educational program was implemented by seven graduate psychology students who received training sessions and program manuals. Weekly meetings assessed the program progress and addressed any problems. Prior to the psycho-educational program, a school assembly 286 Leila Rahey and Wendy M. Craig introduced students to the program. Classroom programs consisted of education with experiential opportunities, such as role plays and puppet techniques. Cen tral topics included: bullying and victimization, conflict resolution, empathy, listening skills, and individual differences. The peer mediation program was based on the Peacemakers program (Alexander & McConnell, 1993) and in volved 16 students in grades 5 through 8. Individual programs for children involved in bullying. These sessions consisted of social skills, listening, and empathy training as well as supportive counselling. Children were referred to the bully or victim groups by teachers and the princi pal. These groups met separately for 45 minutes each week. Ten students were involved and met for eight sessions during the program. Teacher programs. Three meetings were held with the teachers to discuss bully ing, intervention approaches, and student support for those involved in bullying. Preliminary feedback was given to the teachers on the Time 1 data to increase awareness and sensitivity to bullying, the locations of bullying, and the level that students felt teachers intervened. During the intervention, the program coordi nators met with the principal and teachers. RESULTS In order to assess changes related to the program implementation, scores on bullying, victimization, internalizing and externalizing behaviours at Time 1 needed to be equivalent for both schools. Scores on measures that differed at Time 1 were covaried using a MANCOVA. A Repeated Measures MANOVA was used to examine changes across time for scales that did not have a significant difference across schools at Time 1. Peer and Self Child Reports Nominations of bullying and victimization. A 2 (school) x 2 (sex) x 2 (level) x 2 (time) repeated MANOVA examined the effect of the antibullying intervention on self-reports of bullying and victimization. The two levels of grade were Junior (grades 1-4) and Senior (grades 5-8). Only relevant interactions are described. A school by sex by level interaction (multivariate .F(5,433) = 2.45, p < .05) was found for the number of places bullying occurred (univariate F{ 1, 437) = 5.20,/? < .05) and number of places that group bullying occurred (univariate ^( 1,437) = 7.65,p < 0.01). Post hoc tests revealed that older girls at the Comparison school reported more places where both individual and group bullying occurred (M = 1.5, M = 1.6, respectively) than older girls at the Intervention school (M =.21, M=35 ). Older boys at the Comparison school reported fewer places where both individual and group bullying occurred (M = .81, M = .66 ) than at the Interven tion school (M = 1.5, M = 1.3). No differences were found for younger boys or younger girls on the locations of bullying and group bullying. A time by sex by level interaction (multivariate F(5,433) = 2.34, p < .05) was found for content of victimization (univariate ^(1,437) = 8.35, p < .01). Older girls perceived that the number of ways they were victimized increased from Reduce Bullying in the Schools 287 Time 1 to Time 2 at both schools (AÍ = 2.4, and 3.3, respectively). No differences were found for younger girls, younger boys or older boys. Self-nominations of bullying, number offriends, and severity of victimization. A MANCOVA indicated a significant school by level interaction (multivariate ^(3,429) = 3.36, p < .05) for severity of victimization (univariate 1,429) = 8.77, p < .01). Post hoc tests revealed that older children at the Comparison school reported higher severity of victimization at Time 2 than older children at the Intervention school. Younger children, at the Comparison school reported lower severity of victimization at Time 2 when compared to younger children at the Intervention School. Figure 1 presents the means for severity of victimization. FIGURE 1 Severity of victimization at Time 1 and Time 2. 4.5 -i 3.6- 288 Leila Rahey and Wendy M. Craig Frequency of bullying and victimization. A significant sex by level interaction (multivariate /^(3,436) = 3.57, p < .05) was found for frequency of victimization (univariate /-(1,438) = 5.01, p < .05). No significant differences were found for the frequency of bullying and group bullying. Post hoc tests revealed that younger boys reported higher frequencies of victimization (M = 2.6) than older boys (M = 1.1). There was no significant difference between older and younger girls on frequency of victimization. Peer group and school environment. A repeated MAN OVA examined the effect of the antibullying intervention on perceptions of safety, peer acceptance, and peer and teacher involvement in intervening when bullying occurred. Only the significant interactions are discussed. A time by school by level interaction (multivariate F(7,372) = 4.02,/) < .05) was found for safety (univariate F{ 1,378) = 11.18,/» < .01), being alone (univariate F{ 1,378) = 4.56,/>< .05), being left out FIGURE 2 Mean scores on school environment factors for Junior students at both schools at Time 1 and Time 2. 3-1 2.5- 2 Time 1 c 1.5- Time 2 o 0.5- Safety Alone Left Out Disliked Safety Alone Left Out Disliked Intervention School Comparison School Reduce Bullying in the Schools 289 (univariate F(l,378) = 9.28, p < .01), and being well liked by peers (univariate /•(1,378) = 11.93, p < .001). Younger children at the Intervention school re ported a decrease in safety, a decreased perception of being well liked, and an increase in exclusion from Time 1 to Time 2. Older children at the Intervention school reported an increased perception of school safety and being well liked by their peers and a decrease in exclusion by peers. School environment and peer factors are presented in Figure 2 and Figure 3. No significant effects were found for older or younger children at the Comparison school on these variables. A significant sex by school by level interaction (multivariate /-(7,372) = 2.08, p < .05) was found for perceptions of peer intervention (univariate /-(1,378) = 4.84, p < .05). Older boys at the Comparison school reported higher levels of peer intervention (M = 1.9) than did older boys at the Intervention school (M = 1.2). There was no significant differences between the older girls at both schools. FIGURE 3 Mean scores on school environment factors for Senior students at both schools at Time 1 and Time 2. 2.5- a> Time 1 Ho 1.5- Time 2 E 2 0.5- 1 Safety Alone Left Out Disliked Safety Alone Left Out Disliked Intervention School Comparison School 290 Leila Rahey and Wendy M. Craig A time by school interaction (multivariate /(7,372) = 2.33,/» < .05) was found for being left out by peets (univariate /(1,378) = 7.58, p < .05) and perceptions of teacher concern in reducing bullying (univariate /-"(1,378) = 5.94,/) < .05). Children at the Intervention school teported an increase in teacher concern (M = 1.6 Time 1, M = 2.4 Time 2) when compared to children at the Comparison school who did not report any differences on teacher concern (M = 2.0 Time 1 and Time 2). A time by level interaction (multivariate /(7,372) = 10.04,/) < .01) was found for safety (univariate F( 1,378) = 8.07,/>< .01), being alone (univariate/( 1,378) = 20.80, p < .001), being left out (univatiate F{ 1,378) = 6.91,/)< .01), teacher intervention in bullying (univariate /(1,378) = 7.01,/) < .01), teacher concern about bullying (univariate F( 1,378) = 21.66, p < .001) and being well liked by peers (univariate /(1,378) = 11.03,/)< .001). Post hoc tests revealed that younger children reported an increase in being alone and being left out by others (A/ = .6 Time 1, M = 1.3 Time 2), an increase in teacher intervention and concern {M = 1.2 Time l,M= 2.6 Time 2), and a increase in being well liked by their peers from Time 1 to Time 2 (M = .7 Time 1, M = 1.2 Time 2). Older children reported an increase in perception of school safety (M =2.1 Time 1, M = 2.8 Time 2) and a decrease in being alone at school from Time 1 to Time 2 (M = .6 Time 1, M = .4 Time 2). Parent Reports Only the significant interactions are reported. A time by sex by level by school interaction (multivariate /(4,171) = 4.07,/) < .01) was found for parents' reports of externalizing behaviour (univariate /(1,174) = 4.83,/) < .05). Parents reported increases in externalizing behaviours in older boys at the Intervention school (M = 5.7 Time 1, M = 8.5 Time 2) and a decrease in older girls (M = 6.2 Time 1, M = 5.2 Time 2). A time by sex by level interaction (multivariate /"(4,171) = 2.99, p < .05) was found for parents' reports of internalizing behaviour (univariate/(1,174) = 10.95, p = .001). Parents reported decreases in internalizing behaviour in older boys from Time 1 to Time 2 (M = 7.4 Time 1, M = 5.2 Time 2), whereas they reported an increase in older girls (M = 6.1 Time 1, M = 7.5 Time 2). Younger boys and younger girls did not change over time. A repeated measures MANOVA examined the effect of the antibullying inter vention on parent reports of aggression, delinquency, withdrawal, somatization, and anxiety. A significant main effect was found for sex (multivariate /(5,177) = 2.49,/) < .05) on delinquency (univariate /(1,181) = 7.58,p< .05) and aggression (univariate /(1,181) = 3.80, p < .05). Parents reported higher scores on delin quency {M= 6.3, M= 1.2, respectively) and aggression for boys compared to girls (Af=6.6, Af=4.7Time2). Teacher Reports A MANCOVA assessed the effect of the antibullying intervention on teacher reports of internalizing and externalizing behaviour and nominations. A signifi cant sex by school interaction (multivariate /(4,417) = 6.44,/) < .05) was found