Bullying behaviour, intentions and classroom ecology
- First Online:
- Cite this article as:
- Pryce, S. & Frederickson, N. Learning Environ Res (2013) 16: 183. doi:10.1007/s10984-013-9137-7
- 1.5k Views
Anti-bullying commitment across school communities is seen as crucial to the effectiveness of interventions. This exploratory study used a mixed-methods design to investigate bullying behaviour, intentions and aspects of the classroom ecology within the context of an anti-bullying initiative that was launched with a declaration of commitment. Across the sample of 14 primary school classes, containing 338 children aged 8–11 years, changes over time in peer-assessed and self-reported bullying and victimisation were found to be associated with changes in pupils’ sense of school belonging and perceptions of their classroom climate. Using a newly-developed theory of planned behaviour measure, changes in bullying were found to be associated with pupils’ intentions and perceived control with regard to engagement in bullying behaviour. No differences were found between intervention and comparison classes on any of the pupil outcome measures. However teachers of intervention classes reported a relative increase in perceived control over undertaking anti-bullying work with their class. The role of the class as a meaningful unit of analysis in the investigation of ecological-systemic bullying interventions in primary schools is highlighted.
KeywordsBullyingClassroom climateIntentionsInterventionSchool belongingVictimisation
Bullying, defined as an abuse of power that is systematic, repeated and deliberate (Olweus 1991; Rigby 2002), is associated with a variety of negative adjustment outcomes. Children who are victimised report elevated levels of depression, somatic problems, school absenteeism, social anxiety and isolation (Hawker and Boulton 2000; Nishina et al. 2005). Children who bully others are at increased risk of conduct problems, delinquency and psychosocial maladjustment, and can experience increased levels of depression (Kumpulainen et al. 2001; Wolke et al. 2000).
Bullying has also been found to have negative associations with indicators of well-being, such as children’s sense of belonging to their school (Bosworth et al. 1999) and happiness at school (Rigby and Slee 1993). There could be reciprocal effects between the level of bullying in school and children’s perceptions of school climate (Astor et al. 2002). Drawing on ecological-systems theories (Song and Stoiber 2008; Swearer and Espelage 2004), anti-bullying interventions often have a significant environmental component, and it is widely recommended that bullying should be tackled through consistent implementation of strategies at organisational, group and individual levels (Olweus 1993; Swearer et al. 2009). However, to date, few evaluation studies have investigated the impact of anti-bullying interventions on the school climate and learning environment.
The present study addressed this gap in the literature. It was conducted within an ecological-systems theoretical framework with a focus on aspects of the social ecology which interact with aspects of individuals’ disposition to influence engagement in bullying behaviour (Song and Stoiber 2008). The key social ecology and individual disposition variables selected for investigation are outlined in this subsection and the empirical rationale for their selection is presented in the next subsection. Three aspects of the social ecology are investigated: classroom climate, pupil adjustment and school belonging. Classroom climate is a collated measure of each pupil’s perceptions of his/her classroom learning environment, including class cohesiveness (classmates know, help and are friendly towards each other), friction (tension and quarrelling among classmates) difficulty (with class work), satisfaction (liking for the class) and competition (perceived atmosphere of competition in the classroom). Pupil adjustment is inversely assessed through a collated measure of self-reported adjustment difficulties, namely, emotional symptoms, conduct problems, hyperactivity and attention and peer relationship problems. It provides an assessment of relevant characteristics of the classroom peer group. School belonging is a collated measure within a class group of pupils’ general affect towards their school, sense of membership in it and perceptions of support, help and acceptance from adults and peers at school.
Despite encouraging early reports of substantial decreases in victimisation rates following implementation of multilevel anti-bullying interventions (Olweus 1993), meta-analytic findings indicate that most whole-school anti-bullying programmes have not had a significant impact on self report measures of bullying or victimisation (Smith et al. 2004). It has been suggested that varying success rates across schools and studies are influenced by school commitment to implementation (Smith et al. 2004; Stassen Berger 2007), and the importance of school commitment to anti-bullying work been emphasised in guidance from national government in England (Department for Children Schools and Families (DCSF) 2007). Schools have been encouraged to sign up to a national anti-bullying charter, with a member of the school governing body, the principal and a pupil representative signing to state commitment to anti-bullying and an associated plan of work in the school. The Anti-Bullying Pledge Scheme (ABPS), which is described in this article, is one of a number of local schemes set up to support schools in implementing these government recommendations.
The use of a declaration of commitment has received some support as a successful strategy for bringing about transformation within educational contexts. MacKay (2006) found that positive declarations by pupils regarding their future reading achievement significantly improved early literacy skills, as well as attitudes and expectations. Such associations between attitudes, expectations and behaviour are predicted by the TPB. The TPB has been used to investigate changes in the behaviour of children and adolescents in a range of areas, including: smoking cessation (Higgins and Conner 2003) and fruit and vegetable consumption (Gratton et al. 2007). However no research has been located that has used the TPB to evaluate the effectiveness of an anti-bullying intervention.
Applying the TPB to bullying behaviour suggests that intentions to bully, together with perceived behavioural control over bullying behaviour, predict the likelihood that a student will bully someone in his/her school. Intentions to bully are determined by attitudes towards bullying, perceived social pressure to bully, and perceptions of control over bullying behaviour. Where constructs related to the TPB components have been investigated singly, their relevance to bullying behaviour and the evaluation of anti-bullying interventions has been supported. For example, Beran et al. (2004) found that pupils’ attitudes towards victims became significantly more positive following intervention. Intervention has also been found to produce less positive attitudes towards bullies (Frey et al. 2005).
Although the effect of intervention on pupils’ intentions to intervene in bullying incidents has been investigated (Andreou et al. 2008; Rigby and Johnson 2006), there has been little investigation of intentions to bully others. One study revealed that children’s intentions to bully were stronger when peer group norms emphasised rejection and dislike for children outside the immediate peer group (Nesdale et al. 2008). It is plausible that changes in TPB components (intentions, attitudes, perceptions of others’ expectations and personal control in relation to bullying behaviour) can be apparent at an earlier stage following the launch of an anti-bullying intervention than more-common evaluation measures. Indeed one meta-analysis concluded that school bullying interventions were more likely to influence attitudes and self-perceptions than actual bullying behaviours (Merrell et al. 2008).
The majority of anti-bullying intervention evaluations have focused on examining whether the rates of bullying and victimisation have reduced, and most have used self-report bullying inventories. As these self-report measures are typically collected anonymously, analysis of change over time can be difficult (Leff et al. 2004), unless data are aggregated to a class or school level, as would be appropriate in the evaluation of a systemic intervention. Peer assessment of bullying is also commonly used, and is well suited to a class-level analysis as children are typically asked to identify classmates who fit behavioural descriptions of bullying behaviour or victimisation (Cornell et al. 2006).
Following the implementation of anti-bullying interventions, initial increases, rather than decreases, in rates of bullying and victimisation are sometimes found and are often attributed to heightened awareness (Eslea and Smith 1998). It seems less likely that such an effect would be found in school systems currently, particularly where there is a legal requirement for anti-bullying work of some kind to be conducted. In addition, within ecological-systems approaches (Swearer and Espelage 2004), the inclusion of broader measures, such as classroom climate and school belonging that are theoretically and empirically associated with bullying, could provide a means of detecting effects of an intervention with less likelihood of either reactive effects or social desirability responding.
The widespread, legally-mandated implementation of anti-bullying work in schools has other implications for the evaluation of anti-bullying interventions. All schools, whether recruited to an intervention or comparison group, are engaged in anti-bullying work of some kind and intervention schools are adding to, or replacing, existing anti-bullying strategies. Considerations such as this have led to calls for more use of mixed-methods research in anti-bullying studies (Powell et al. 2008) in order to adequately contextualise quantitative findings.
Do classes that show positive changes over time in bullying and victimisation also show positive changes in (a) ecological variables (classroom climate, pupil adjustment and school belonging) and (b) TPB pupil and teacher variables (intention, attitude, subjective norm and perceived behavioural control)?
Does the launch of the Anti-Bullying Pledge Scheme lead to more positive initial change in the intervention schools than the comparison schools in (a) bullying, victimisation and ecological variables and (b) TPB pupil and teacher variables?
Do qualitative pupil perceptions of bullying and how it is addressed (a) relate to the quantitative data collected and (b) differ between the intervention and comparison schools.
Design and analysis
To address the research questions a partially-mixed sequential, quantitative-dominant design was adopted (Powell et al. 2008). Quantitative data were collected from classes at time 1, prior to the launch of the intervention, and at time 2, two to three months following the launch. Class-level change scores were used throughout the quantitative analyses. These were calculated by subtracting the pre-intervention from the post-intervention mean score on each variable for each class. Preliminary analyses confirmed normal score distributions except for one pupil planned behaviour scale score and for the teacher planned behaviour scores, which were positively skewed. Non-parametric tests were therefore used in the analyses of the planned behaviour scales.
Research questions 1(a) and 1(b) were addressed using correlational analyses. Spearman correlations were used in analyses involving the planned behaviour scales, and Pearson product-moment correlations were used otherwise. Research questions 2(a) and 2(b) were addressed using a between-groups design, with equal numbers of classes in the intervention and comparison groups. The Mann–Whitney U Test was used to analyse differences between the scores of intervention and comparison classes on the planned behaviour scales, whereas scores on all other variables were analysed using t tests.
Research questions 3(a) and 3(b) were addressed through the analysis of qualitative data collected through focus groups at time 2. The four anonymised focus-group transcripts were analysed using thematic analysis, following the staged procedure described by Braun and Clarke (2006). Coding was completed in accordance with the guidelines described by Miles and Huberman (1994). The codes identified were sorted into themes and the frequency of mention of each theme in each school was identified. A check on the reliability of the final coding by an independent researcher produced 86 % agreement.
Participants were 338 pupils (178 males and 160 females) aged 8–11 years in Year 4 (aged 8–9 years), Year 5 (aged 9–10 years) and Year 6 (aged 10–11 years). They attended 14 classes in four primary schools in a district in the West Midlands of England. The teachers of these classes also participated. The percentage of pupils from minority ethnic groups was between 1 and 2 % across the four schools. This is below the national average for primary schools but reflects the most recent UK population census figure of 2.4 % for the school district in which the study took place. Eligibility for free school meals, an index of socioeconomic status, was found to range from 11 to 25 %, which is broadly in line with national statistics in the UK for this age range. The percentage of pupils with special educational needs ranged from 10.2 to 21.2 %, which is close to the national average of 18.1 % for primary schools (Office of National Statistics 2008).
The ABPS intervention was introduced to schools in the district on a rolling programme. The two schools that expressed interest in joining the next wave agreed to take part in the research. Comparison schools were selected by the Educational Psychology Service as actively involved in anti-bullying work; these schools are similar to the intervention schools in terms of size and pupil characteristics. The comparison schools participated in the quantitative and qualitative data collection but continued with their existing anti-bullying work and did not receive any additional consultation or other support.
Initial meeting between senior management team and the pledge supporter when the pledge process is described and a decision made about engagement with the scheme.
Planning meeting between the school pledge coordinator or working group and the pledge supporter.
Representatives from the school make a declaration of their commitment to the pledge scheme and to anti-bullying.
Questionnaires are sent to pupils, parents, staff and governors to gain their views about anti-bullying work currently taking place in the school.
Information from pupils, parents, staff and governors is collated and used in drawing up an anti-bullying action plan, with review dates.
Visits by the pledge supporter to the school to review progress.
Annual visit by the pledge supporter to update the action plan.
The study was approved by the university ethics committee and the school district involved and carried out in accordance with the British Psychological Society’s Code of Conduct and Ethical Principles. Informed parental consent was sought via the schools by letter and no parent refused permission for participation. Assent was obtained from the pupils who were told that they could stop at any time if they did not want to take part in the questionnaire or focus-group activities.
The quantitative measures were administered in a single whole-class session. Children were assured that their questionnaires would be kept confidential and were asked not to discuss them. Following questionnaire completion, a buffer activity was administered to minimise the likelihood of discussion of the measures, which was not reported by any of the teachers.
Qualitative data were collected through a focus group of six children in each school. Focus groups have a number of advantages in the investigation of an issue such as bullying, which is sensitive and likely to be subject to social desirability effects, but on which children hold a range of views (Vaughn et al. 1996). Through the group process, the facilitator can communicate that a range of views are valued and allow respondents to react to, and build upon, what others have said, without creating any requirement for individuals to respond to any particular issue (Stewart et al. 2006; Vaughn et al. 1996).
Two pupils from each of the three year groups were randomly selected and invited to take part in focus groups conducted in a quiet room in their school by the first author ‘to find out what the pupils think about how their school deals with bullying’. Pupils were assured about individual anonymity and encouraged to talk freely and openly. They were asked a series of questions about: what happened in their school if someone was being bullied; how they feel about what happens in their school; and whether anything had changed recently in how their school deals with bullying.
Bully. This person is a bully and often picks on other people or hits them, or teases them or does other nasty things to them for no good reason.
Victim. Someone who often gets picked on, or hit, or teased or has nasty things done to them by other children for no good reason.
The score for each child was the proportion of classroom peers who nominated him/her for each descriptor. Nabuzoka and Smith (1993) reported concurrent validity data for their peer-assessed bullying and victimisation scores from measures of sociometric status and other social behaviours.
Self-reported bullying and victimisation was assessed by the Peer Relations Questionnaire–short version (PRQ; Rigby and Slee 1993). This is a 20-item self-report measure of a person’s tendency to bully (four items), to be victimised (four items) and to be pro-social (four items). The remaining eight items act as filler items. Pupils rate how often statements are true for them on a 4-point scale ranging from 1 (Never) to 4 (Very often). Pre-intervention Cronbach’s alpha reliability was 0.72. In support of the construct validity of this measure, Rigby and Slee (1993) reported significant correlations between the subscales and measures of self-esteem, happiness and liking for school.
The 25-item short form of the My Class Inventory (MCI–SF; Fraser 1982) assesses primary-aged pupils’ perceptions of their classroom learning environment The MCI–SF is composed of five scales, each containing five items to which pupils respond either Yes or No. The scales are Cohesiveness (classmates know, help and are friendly towards each other), Friction (tension and quarrelling among classmates), Difficulty (with class work), Satisfaction (liking for the class) and Competition (perceived atmosphere of competition in the classroom). A composite Classroom Climate score, calculated by subtracting mean scores on the three negative climate scales from those on the two positive scales, had a Cronbach’s alpha coefficient of 0.76. The validity of the MCI has been supported by a range of studies (reviewed by Fraser 1998) which report discrimination between classrooms and a positive association with better student outcomes and attitudes.
School belonging was assessed by the Belonging Scale (Frederickson et al. 2007). Pupils rate how true statements are for them on a 3-point scale, with a higher mean rating reflecting a higher sense of belonging to school. Cronbach’s alpha for the 12-item scale was 0.82. In support of the scale’s validity, Frederickson et al. (2007) found that scores for children aged 7–11 years were positively correlated with peer group acceptance and negatively correlated with peer rejection.
Adjustment Difficulties were assessed using the Total Difficulties score on the self-report version of the Strengths and Difficulties Questionnaire (SDQ; Goodman et al. 1998). The 20 items, covering emotional symptoms, conduct problems, hyperactivity and attention, and peer relationship problems, are rated on a 3-point scale of Not true, Somewhat true or Certainly true. Cronbach’s alpha in the present study was 0.74. Muris et al. (2004) reported acceptable criterion validity in relation to teacher ratings of problem behaviour for children of this age.
Theory of planned behaviour anti-bullying scales
Pupil and teacher questionnaires were designed for the present study following the guidelines provided by Ajzen (2002). Responses to all items were rated on 4-point Likert-type scales ranging from 1 (agree in a big way) to 4 (disagree in a big way). The 14-item pupil questionnaire assesses pupils’ intentions to participate in bullying behaviour (two items, including “I don’t intend to bully someone in my school in the next month”), attitudes to participating in bullying behaviour (three items, including “Bullying others in school in the next month would be fun”), subjective norm about bullying based on perceptions of parent, teacher and peer expectations (six items, including “My friends don’t think I should bully others in school in the next month”) and pupils’ perceived control over bullying behaviour (three items, including “It is my decision whether I bully others in school in the next month”). The teacher TPB measure focused on undertaking anti-bullying work and assessed intentions (two items, including “I plan to do anti-bullying work with my class in the next month”), attitudes (five items, including “Doing anti-bullying work with my class in the next month is important to me”), subjective norm (five items, including “Parents of children in my class think I should do anti-bullying work with my class in the next month”) and perceived control (two items, including “I have lots of control over whether I do anti-bullying work with pupils in my class”). Cronbach alpha values were 0.85 (intention), 0.68 (attitude), 0.81 (subjective norm) and 0.63 (perceived control) for the pupil questionnaire; and 0.93 (intention), 0.92 (attitude), 0.76 (subjective norm) and 0.87 (perceived control) for the teacher questionnaire.
Analysis of changes over time across the sample
Pearson Correlations between measures of change in bullying, victimisation, belonging, class ethos and adjustment difficulties
Spearman correlations between measures of change in bullying, victimisation and planned behaviour model components for classes and teachers
Pupil subjective norm
Pupil perceived control
Teacher subjective norm
Teacher perceived control
Analysis of differences between the intervention and comparison classes
Intervention and comparison class means and standard deviations of pre-post change scores on bullying, victimisation, climate, belonging, adjustment, pupil planned behaviour and teacher planned behaviour scores
Pupil subjective norm
Pupil perceived control
Teacher subjective norm
Teacher perceived control
No significant differences were found between the classes in the intervention and comparison schools for any of the components of the pupil planned behaviour scale (intention U = 18, n1 = n2 = 7, p = 0.41; attitude U = 17, n1 = n2 = 7, p = 0.34; subjective norm U = 22, n1 = n2 = 7, p = 0.75; and perceived behavioural control U = 14, n1 = n2 = 7, p = 0.18). In the analysis of changes in the teacher planned behaviour scale, a significant difference in the predicted direction was found for teachers’ perceived control over undertaking anti-bullying work (U = 9, n1 = n2 = 7, p = 0.04), which increased more over time in the intervention than the comparison schools. Change scores on the other components of the teacher planned behaviour scale were not significant (intention U = 19, n1 = n2 = 7, p = 0.45; attitude U = 19.5, n1 = n2 = 7, p = 0.50; and subjective norm U = 13.5, n1 = n2 = 7, p = 0.14).
Frequency of positive and negative perception themes in intervention and comparison schools
Frequency of theme
Positive pupil perception of school effectiveness
Staff willing to listen
Staff take action
Talk to others (friends, parent, lunch supervisor) about it
Negative pupil perception of school effectiveness
Staff not willing to listen
Staff not taking action
Response not consistent
It wouldn’t matter who you’d go to cus they’d all make you feel better and do something about it (Staff take action, Comp. 1).
Mrs x (Principal) would ask the people who saw the fight or bullying to go into her office and talk to her and tell her what happened (Staff take action, I1).
The school would always do something to that person and it probably won’t happen again (Staff take action, I2).
Say if someone has been bullying you, you go and tell the teacher and then she sends for them and she puts them on the behaviour chart and things (Staff take action, Comp. 2).
Sometimes the teachers are busy so you don’t get chance to speak to them… (Staff not willing to listen, I1).
I mean like they do talk to them but they don’t actually like go proper serious they just say you’ll miss your break and don’t do it again. They never do anything too bad to make them not do it again (Staff not taking action, I1).
Well I think that there have been more bullies coming since when I was in year 3. Since last year there has been quite a lot of bullying in school. They just don’t stop (Amount of bullying is high/has increased, I1).
The schools are a better place since the worry box came in because the teachers will speak to you and the person who did it individually and them they’ll sort it out to make sure it won’t happen again (Amount of bullying is low/has decreased, I2).
A number of comments expressed satisfaction with these reactive strategies, with at least one such comment being made in each focus group, for example:
Like in our class there are lots of people who miss their breaks (Take away privileges, I1).
Umm, usually, the person who was the bully would be sent to the head teacher and she would tell them off, and if it went too far then she would get their parents in (Sent to Principal & Involve Parents, Comp. 1).
Comments expressing dissatisfaction were also recorded in each school, except Intervention school 2, for example:
If people miss breaks and that, they umm, they sometimes, if they aren’t usually very naughty, they usually stop (I1).
In Comparison schools 1 and 2, expressions of satisfaction and dissatisfaction with reactive strategies were both small in number. In Intervention school 1, the number of expressions of satisfaction was comparable to the other schools, but the number of expressions of dissatisfaction was much greater.
They get lots of punishments but they keep doing it (I1).
X got excluded and, when he came back, he said ‘it was really good, and it was so much fun, and I got to play on my X Box’. And everyone is now gonna try and be bad (I1).
Participants expressed only satisfaction with these preventative initiatives. Comments included:
Buddies, they have a yellow cap with playground friend on it and you can go to them (Interventions involving peers, Comp. 1).
The schools have been a better place since the worry box came in (I2).
Since the buddies came, there have been less fights, which is good.
Reports of new initiatives were confined to Intervention schools 1 and 2, with three instances being recorded in each intervention school.
We do have a worry box in the classroom. That’s quite new.
The playground pals are quite new and the friendship stop.
Class means for Peer-assessed bullying at Time 2
Peer-assessed bullying mean
The results of this study offer substantial support for the hypothesised associations between changes in bullying behaviour in school and changes in variables derived both from ecological-systems theory and the TPB. However the evaluation results failed to support the hypothesised effects of the ABPS intervention-launch on bullying, victimisation, ecological-systemic indicators or TPB components. As hypothesised, decreases in bullying and victimisation were associated with positive changes in sense of school belonging and classroom climate. The relationships held for peer-assessment, as well as self-report measures, indicating that the findings cannot be explained by common method variance. However changes in bullying or victimisation within a class were not strongly associated with changes in class members’ reported adjustment problems.
The influence of the teacher on classroom climate (Burnett 2002) and belonging (Solomon et al. 1997) is well established. It is consistent that bullying and victimisation in classes were found to decrease as sense of belonging to school and, to some extent, the positivity of the classroom climate increased. It is perhaps not surprising that changes in bullying and victimisation in classes are more strongly associated with changes in feelings of belonging towards school and classroom climate than with change in broader adjustment. However it is possible that a different measure of adjustment, possibly one more sensitive to change within normal score ranges, might have detected some changes. It should also be acknowledged that 2–3 months is a short period in which to expect changes in reliable measures of adjustment.
The results substantially supported the TPB model and indicated the potential value for future research and practice of the measure developed for this study. Changes in bullying behaviour were found to be significantly associated with changes in the two components of the model most closely related to the performance of the behaviour (Ajzen 1991). In classes where pupils reported reduced intentions to engage in bullying and increased sense of control over their bullying behaviour, reductions in bullying were reported. The positive correlation found between changes in the perception of strong anti-bullying norms and self-reported victimisation appears to suggest that in classes where pupils perceive the expectations of teachers, peers and parents to have become increasingly ‘anti-bullying’, they become more willing to report victimisation.
Neither changes over time in teachers’ intentions to engage in anti-bullying work, nor their perceived control over doing so, were associated with changes in bullying or victimisation in their classes. However teachers who perceived an increase in the expectations of the school principal, other teachers or parents that anti-bullying work would be carried out, had classes that reported a reduction in peer-assessed victimisation. It is reasonable to suppose that such teachers are more likely to be seen by children in their classes to take action in supporting pupils and dealing with bullying, although as no measure was taken of teachers’ actual engagement in anti-bullying work, this hypothesis cannot be tested with the current data. It should also be noted that this finding was obtained with peer-assessed, but not self-reported victimisation. This is broadly consistent with previous findings that peer-assessed victimisation is more strongly associated with systemic variables such as low social acceptance, while self-report measures of victimisation are more strongly associated with individual variables, such self-worth (Juvonen et al. 2001).
No significant differences between the intervention and comparison schools were found in changes over time on any of the outcome measures. There was no evidence that the launch of the intervention had any impact on pupil intentions to bully or on perceived behavioural control over bullying. However, relative to teachers in the comparison schools, teachers in the intervention schools showed a greater increase over time in their perceived control over engagement in anti-bullying work. The launch of the intervention appears to have had an empowering effect on teachers, which is consistent with conclusions of a meta-analysis that school bullying interventions enhanced teachers’ feelings of efficacy, their knowledge of effective practices and their actual behaviour in responding to incidents (Merrell et al. 2008). It would appear to support the suggestion that, given other pressures in schools, such as a crowded curriculum and national attainment targets, a strongly expressed commitment at a whole-school level might be needed if teachers are to feel empowered to engage in anti-bullying work (Limber et al. 2004).
The qualitative data provided evidence of intervention implementation in that pupils within the intervention schools had noticed recent changes in the way in which their school dealt with bullying, whereas pupils in comparison schools had not. The new preventative initiatives identified were positively received by pupils, but the qualitative data suggest that a stronger focus on staff listening to pupils, taking responsive action and following it through could be required in addition.
A number of limitations of this study should be acknowledged. The measures of bullying and victimisation focused exclusively on direct bullying, which involves physical and verbal confrontation, and did not assess indirect bullying or relational bullying involving behaviours such as social exclusion, rumour spreading or cyber bullying (Wolke et al. 2000; Stassen Berger 2007). There is no reason to believe that the launch of the intervention investigated in this study would have been any more likely to impact on indirect bullying. Indeed school interventions typically place more emphasis on direct forms of bullying (Slee and Mohyla 2007). However, it must be acknowledged that different factors have been associated with different types of bullying (Viding et al. 2009), and generalisation of findings across types of bullying cannot be assumed. Generalization of findings across schools, even in the same district, cannot be assumed either. All of the schools were volunteers, they were not randomly assigned, and the number of schools able to engage with the intervention at the time of the study was very small. While these schools were demographically representative of the district, it is likely that they were more motivated to engage in anti-bullying work than primary schools generally.
In conclusion, this study offers a number of implications for practice and future research when investigating the launch of anti-bullying and other whole-school initiatives. The study suggests the potential of the theory of planned behaviour as a framework for examining bases for behavioural change by different members of the school community, and as well as providing instruments, with initial data, to support the development of this work. The value of using a mixed-methods design in exploratory studies of this kind is also highlighted as the combination of quantitative and qualitative methodologies provided a cross validation of the between-class differences identified. Such studies are likely to have increasing relevance as calls increase for the implementation of evidence-based practice models in determining the effectiveness of bullying and other interventions in schools (Song and Stoiber 2008). The role of the class as a meaningful unit of analysis in the evaluation of ecological-systemic bullying interventions in primary schools, where the class teacher typically has responsibility both for delivering anti-bullying curriculum components and implementing first-stage reactive strategies, is an important issue that is highlighted by the findings but which all too often is ignored in the existing literature.
The findings of the study indicate that it could be beneficial for systemic anti-bullying interventions to incorporate a declaration of commitment at the whole-school level if teachers are to feel more empowered to engage in anti-bullying work. The importance of communicating strong anti-bullying norms that encourage pupils to report victimisation was also demonstrated. Links were identified between reductions in bullying and victimisation and increases in feelings of belonging and positive perceptions of classroom climate. While it is acknowledged that these findings are correlational and further intervention studies are required to test the suggestion, it might be beneficial for anti-bullying interventions to include preventative strategies such as creating a positive class learning environment and promoting a high level of school belonging. Certainly the qualitative findings indicate that strategies, such as creating a school ethos in which staff listen to pupils, take responsive action and follow it through, are valued by pupils and strongly influence their perception of the school’s effectiveness in dealing with bullying.