7. PRELIMINARES COMPUTACIONALES
8.2. Progresión de un flujo newtoniano en una cánula
Despite not recruiting a large enough sample of schools to conduct a MLM adequately, possibly due to the doubly sensitive nature of the research (bullying and SEND), there was a good variety of types of schools in the study. There were both rural and inner-city schools in the sample, from several different Local Education Authorities, including one private all-girls school. Additionally, there were schools with a good range of Ofsted ratings, ranging from ‘requires improvement’ to ‘outstanding’, and a good range of percentage of children with SEND at the schools (8 – 35%). This variety of schools was recruited to ensure a good representation of schools and to ensure bullying/victimisation was investigated in schools with differing climates. Most of the scales used in the main study had been rigorously checked in terms of reliability in the pilot study, which meant that we could have confidence that these scales were appropriate for use with the
participants. This meant that very few changes had to be made in the data preparation stage of the main study because all of the items within each scale were reliable based on the Cronbach’s alphas they were retained.
It is important to consider the strengths and limitations of the research that
contributes to this thesis, in conjunction with the findings, to ensure that any conclusions are made within the appropriate context of the research.
Creating tools to measure inclusion was a challenge, as inclusion is a relatively vague concept. The commonly used definition is that “all students attend and are welcomed by their neighbourhood schools in age-appropriate, regular classes and are supported to learn, contribute and participate in all aspects of the life of the school” (InclusionBC, 2017). Although schools must be inclusive, it is still quite broadly defined. When creating the Ofsted report coding tool, it was suspected that Ofsted judges are reasonably diplomatic in their reports, as they do not often say a school does something badly and instead often leave it out of the report. This then means the report does not accurately represent the school – areas may have been left out for several reasons, potentially because there was no evidence of it, because it was done badly, or because it was done acceptably and was not worth mentioning. This then meant that it was difficult to accurately score schools in some areas if details were missing from the Ofsted report. Perhaps this indicates an issue with Ofsted reports – a lack of consistency across schools may cause problems, such as the one identified by this research, i.e. what is mentioned in one Ofsted report is not necessarily mentioned in another and the reasons behind omissions is unclear. Very little information is available on the Ofsted website about their
inspections, however, it states that a framework will be followed in order for inspectors to report on particular aspects of the provision (Ofsted, 2017). Perhaps a more rigorous survey needs to be employed by Ofsted, measuring key aspects of interest, such as bullying and inclusion, so that Ofsted reports appear to be consistent across schools and could improve any future research that utilises their reports. This could potentially be
accompanied by a report written in full prose elaborating on any aspects in the survey that schools have done well or could improve upon. This would allow Ofsted reports to be used more reliably in research in the future.
Previous work had been conducted in the review of anti-bullying policies (Smith, Smith, Osborn & Samara, 2008), however, no research has been carried out using inclusion and/or Special Educational Needs policies. When creating the policy coding tool, it became clear that there were quite a few schools that had used similar templates for their policies, particularly schools that are part of an academy group, as they all had the same policy. This meant it was important to identify the specific items that would allow us to identify more and less inclusive schools. With no previous evaluation of school inclusion policies, this was challenging and time consuming but resulted in a new coding scheme for measuring school policies. Creating the policy coding schemes was an iterative process, involving several revisions before the final version was created with accompanying information regarding how to best answer the questions. This information was key to ensuring that the raters gave similar scores to the content of the policies, as it is good practice when
applying content analysis.
Additionally, there is very little evidence to suggest that policies necessarily reflect what happens in practice. However, schools are required by law to have an inclusion policy and put reasonable adjustments in place in order to be inclusive (Gov.uk, 2018). This indicates that if they do not follow through with their policies, they may be in breach of the law, implying that what they say they do in their policy is put into practice in order to be within the law. While the policies were useful in predicting defender behaviour, they did not appear to be associated with the more negative experiences (victimisation, bullying and follower behaviours) and so other ways to measure inclusion may need to be considered, such as data from the children and teachers on actual inclusion practices. This will be expanded upon further in a subsequent section, outlining the future of research in this area.
For the child questionnaire, a limitation, previously touched upon in Chapter 4 section 4.4.6, was the method used to measure bullying and victimisation. Self-report has been used extensively in previous bullying research, however, it does have some
limitations, which were considered before deciding to use this approach. Children may be less forthcoming about negative behaviours and attitudes towards SEND when they are reporting on themselves. Additionally, the children may not realise they are carrying out the negative behaviours, but their peers may have reported it if peer-nomination methods were used. For example, Brandon and Cornell (2009) found that twice as many students were categorised as bullies using peer nomination compared to self-report (11% and 5% respectively). Peer nominations of bullying roles would have potentially given us a more accurate representation of bullying and victimisation, as children are not reporting on their own behaviours, however, following discussions with the university ethics committee it was agreed that peer nominations were not appropriate in this study, due to the additional sensitivities around asking questions about disability. In addition to using self-report, this research required children to put their names on their questionnaires, which may have led to them not feeling comfortable about being honest about negative behaviours. In order to measure reciprocal friendships, however, it was necessary to have the students’ names on the questionnaires. Teacher reports of bullying could have been collected, and previous research has used teachers as a source of information on student bullying (see Farrington & Baldry, 2010, for review), however there are limitations with this. For example, teachers may find it time consuming and not complete the survey properly and, more importantly, in high schools, the teachers only see the students for one lesson a day and so may not be aware of what the students are doing outside of those lessons, in the hallways, or in breaks. Therefore, self-reports were the preferred option.
Due to the different measures used in this research, the best possible approach was used in terms of having names on the questionnaires and using self-report. Peer
nominations were not used in this study due to the ethical sensitivities raised by the ethical committee. The questionnaire was not anonymous despite being self-report because of the friendship nomination questions, which meant that the children’s questionnaires had to be analysed for reciprocal friendships. Using anonymous self-report questionnaires or peer nominations, may have resulted in more bullies and followers in the data, however, there was no way to predict that the self-report measure would yield such low means for bullying and follower behaviour, when previous studies have used self-report measures successfully.
A potential reason for this may be the measure used; the PRS is usually a peer nomination scale (Salmivalli et al., 1996) and was adapted for this research in order for it to be used in a self-report method. In this research, the assistant and outsider roles had low Cronbach’s alphas, which indicated that the items in these subscales do not show
consistency when measuring the behaviours, indicating that the together the items are not appropriate for measuring the behaviours. A factor analysis was carried out on a shortened PRS (of 21 items) developed by Sutton and Smith (1999). They found that there was considerable overlap between the bully, assistant, and reinforcer subscales, concluding that a pro-bullying role should be adopted instead of the three subscales (Sutton & Smith, 1999). In this research, however, the difference between the follower and the bully roles was important to investigate for children with SEND, thus, the bullying and follower roles were kept as separate. Additionally, as the PRS does not measure physical bullying
specifically, it was not possible to break the bully role down into physical, verbal and relational bullying behaviours, which could have potentially yielded more individuals within these roles, but who do not carry out all of these behaviours. Future research could build upon the issues found with the PRS and the different roles children can take. Rather than combining the leader bully and the follower roles into one overall role, it may be important to consider them separately, as there may be differences between children who are the leader bullies or the followers, such as a diagnosis of SEND and emotional
symptoms. In this research, it was not possible to investigate which specific type of SEND diagnosis was most associated with certain roles. Due to the low frequencies for some types of SEND, the specific types of need were grouped under the four different types of need and conclusions cannot be made about whether, for example, children with dyslexia are more likely to victimised than children with dyspraxia (two types of need that were grouped under cognition and learning difficulties). Further research could be conducted, recruiting larger sample sizes, with the aim to investigate the risk of specific types of SEND on the different bullying roles.
Research could also be conducted into whether the roles are consistent across a variety of contexts, for example, a playground compared to a classroom; people they know, compared to people they do not know; and whether they are consistent across time using a longitudinal approach; is it possible for a victim to become a defender, for example, with the appropriate support and intervention? Throughout the reporting of the findings and the discussion of the findings in this thesis, the different ‘roles’ within bullying have been referred to as ‘behaviours’. This was because the researcher found that students did not predominantly fit one ‘role’ but carried out different behaviours. This is why prevalence rates have not been reported in this thesis; children did not just fit one role. The term ‘role’ may need to be altered as this implies a relatively fixed identity, which may not always be the case. For example, a child may consistently show outsider behaviours until their best friend is bullied and then this child may demonstrate defender behaviours. Huitsing, Snijders, Van Duijn and Veenstra (2014) investigated the social networks between bullies, victims and defenders in their research, for example, and found that victims of the same
bullies often defend one another, proving social support, indicating that they are not ‘pure’ victims, but can also display other behaviours. Future research could investigate the exclusivity of these ‘roles’ and investigate the contexts in which a child may deviate from their typical ‘role’.
As an alternative to the reciprocal friendship measure, sociometric status/ratings could have been used to measure popularity and/or liking. This is a different concept to reciprocal friendships and measures how visible/dominant a child is in the classroom. Bullies, by their definition, are potentially more dominant children and so tend to be rated as more popular, while victims are perhaps less visible/dominant and so are less popular (de Bruyn et al., 2010). However, measuring sociometric status would involve peer
nomination methods, which, as stated above, would not have been approved by the ethical committee. This was because of the sensitive nature of the research and the increased ‘spotlight’ that nomination puts on children in the classroom. Future research could investigate sociometric status and children with SEND compared to children without SEND, as it is a fairly under-research area.
Overall, there was a limitation with the methodology used and the analysis technique that had been planned. For multilevel modelling, there needs to be a larger number of groups than could be practically recruited in the confines of a PhD. Over 200 schools were contacted via email and telephone; 2 schools from this sample agreed to participate, 2 agreed and then dropped out, and the final 7 in the sample were recruited through the researcher’s personal contacts. Future studies may need to work closely with schools in order to be able to conduct this type of research. Deeper relationships and collaborations could be formed between universities and schools, to ensure that the schools feel they are benefiting from participating, such as providing a psychology talk for their students or inviting them to a psychology ‘away day’ at the university. The multilevel model analysis was therefore unable to find significant predictors in the majority of the school level measures, due to the lack of variance, potentially due to the small number of schools. In this research, the class level was not investigated due to the approach taken when data collecting; some schools wanted all the children to take part at the same time in one large group during assembly and so were not separated into classes. Future research could investigate the class-level factors involved in bullying, such as classroom discipline or classroom behaviours. This could potentially allow for higher numbers of the level 2 factor, as researchers could recruit many classes from just 1 school. However, the
practicality of this will need to be explored; in high schools, students do not always spend a lot of time with the same class during the day, for example, they are in one class for registration and then move into different classes for different subjects and have different teachers. There was a significant positive correlation between the policy analysis and the Ofsted analysis, indicating that they were both measuring the same concept in each school; if one school had a positive policy score, they also had a positive Ofsted score. This
suggested that these measures would be useful in future work when investigating school level variables, but with the inclusion of other measures.