Bullying in Different Contexts

3 Bullying in schools: thirty years of research

Bullying in school has become a topic of international concern over the last 30 years. Starting with research in Scandinavia, Japan and the UK, there is now active research in most European countries, in Australia and New Zealand, Canada and the USA and Japan and South Korea (Jimerson et al., 2010; Koo et al., 2008; McGrath and Noble, 2006; Smith et al., 1999b). In fact, the research on school bullying can be thought of as a research programme, in the sense of Lakatos (1970), with its core being the conception of bullying as a distinct category of aggressive behaviour. This programme has gone through four distinct waves or phases. In this chapter I will set the scene by outlining these four waves of research, and then follow the standard chapter format for this book in reviewing the research in more detail.

First wave of research: origins, 1970s–1988

Leaving aside one or two isolated earlier studies, the systematic study of bullying in schools can be dated from the 1970s, mainly in Scandinavia. A physician, Heinemann, published a book Mobbning – Gruppväld bland barn och vuxna(Mobbing – Group Aggression against Boys and Girls) in 1972, which Olweus credits as first seriously raising awareness of the issue. In 1973, Olweus published Forskning om skolmobbning, translated into English as Aggression in Schools: Bullies and Whipping Boys (1978). This book was the first important scientific work on the topic, and thus justifies the assertion that there is now a 30-year tradition of research.

In this book, Olweus defined bullying or ‘mobbing’ in terms of physical and verbal behaviours, although he explicitly rejected the connotations of the ‘mobbing’ label (which implies group bullying), since much bullying appeared to be by one person. Through the 1980s Olweus developed a self-report questionnaire to assess bullying, an important tool in subsequent work. Also, in parallel with the first Norwegian National Anti-Bullying campaign, launched in 1983, he developed a school-based intervention programme. His evaluation of the original version of the Olweus Bullying Prevention Program (1983–85), with reports of reductions in bullying of around 50 per cent, encouraged researchers and inspired the next wave of research. A meeting in Stavanger in 1987 brought together a number of European researchers.

Second wave of research: establishing a research programme, 1989–mid 1990s

From around 1989, books and journal articles started to appear, and surveys in other countries beyond Scandinavia were beginning to be carried out. Besides self-report surveys, some studies started to use peer-nomination methodology. Also, some intervention campaigns took place, partly inspired by the Norwegian campaign; early large-scale interventions were in England (Smith and Sharp, 1994), Canada (Pepler et al., 1994) and Flanders (Stevens and Van Oost, 1995).

An important change in this period was a broadening of researchers’ definitions of bullying, to include indirect and relational bullying (such as rumour-spreading, social exclusion); this followed the similar broadening of understanding of ‘aggression’, following the work of Bjorkqvist and colleagues in Finland and Crick and colleagues in the USA. Furthermore, work on bullying was becoming more international. Contacts were taking place with researchers in North America (for example, Pepler in Canada) and, towards the end of this period, with researchers in Japan. In fact, studies on ijime (the term closest to ‘bullying’) dated back at least to the 1980s, but this separate research tradition only made substantial contact with the Western research tradition following a Monbusho/UNESCO study in the latter 1990s (see Smith et al., 1999b).

Third wave of research: an established international research programme, mid 1990s–2004

During this period, research on ‘traditional’ bullying became an important international research programme. Many more publications appeared, and research on bullying featured substantially in European and international conferences. Surveys, and interventions, took place in many countries (see 21 country reports in Smith et al. (1999b) and 11 country reports on interventions in Smith et al. (2004a)). A notable methodological step was the introduction of participant roles in bullying, from Salmivalli’s work in Finland (Salmivalli et al., 1996). Researchers in the USA substantially developed research on victimisation and bullying during this period: see the edited collections by Juvonen and Graham (2001) and Espelage and Swearer (2004). Important work was also being undertaken in Australia and New Zealand: see Rigby (2002) and Sullivan et al. (2004).

Fourth wave of research: cyberbullying, 2004–present

A substantial shift and new impetus in the bullying research programme has been the advent of cyberbullying (see Rivers et al., Chapter 10). Significant awareness of cyberbullying, and press reports, appear to date from around 2000–01. In England, for example, the DfES pack Don’t Suffer in Silence (2000) did not mention cyberbullying; but a revision published in 2002 mentions ‘sending malicious emails or text messages on mobile phones’ (p. 9). Indeed, early forms of cyberbullying seem to have been mainly text messages or emails, at least judging from the extant research; for example, Noret and Rivers (2006) surveyed 11,000 English pupils from 2002 to 2005, asking about who had ‘received nasty or threatening text messages or emails’. Starting from a small number of articles in the first few years of this century (although rather more on websites), academic publications in the cyberbullying area have increased rapidly in the last few years. Notable publications in the area are the book by Willard (2006), a special issue of the Journal of Adolescent Health (2007), the books by Shariff (2008), Kowalski et al. (2008) and Hinduja and Patchin (2008), a special issue of the Zeitschrift für Psychologie/Journal of Psychology (2009) and a book by Li et al. (in press).

Of course, work on traditional bullying continues in this latest wave of studies, but methodologies (such as questionnaires) have had to be adapted to include new forms of bullying. Cyberbullying also brings some new challenges and opportunities to the bullying research programme, notably greater disciplinary breadth, and a chance to extend the focus beyond just the school context, and beyond just the school age range (Smith, 2010).

Definitions of bullying

Although there is no universally agreed definition, there is some consensus in the Western research tradition that bullying refers to repeated aggressive acts against someone who cannot easily defend themselves (see Olweus, 1999; Ross, 2002). A similar definition, though perhaps with broader connotations, is that bullying is a ‘systematic abuse of power’ (Rigby, 2002; Smith and Sharp, 1994). Although the two criteria of repetition and power imbalance are not universally accepted, they are now widely used. It should be noted however that there are problems with extending this definition in a simple way to cyberbullying; both as regards repetition (a single act by an aggressor, such as posting a nasty website comment, may be seen, commented on and forwarded by many others) and as regards imbalance of power (clearly it does not refer to physical strength or numerical strength in this context). The distinction of ‘bullying’ from more general ‘aggression’ has been an important and widely accepted aspect of traditional bullying research, but may be on much less firm ground as regards cyberbullying and cyber-aggression (Dooley et al., 2009).

Taking the traditional definition of bullying, then the repetition and imbalance of power are likely to lead to particular outcomes, such as fear of telling by the victim, low self-esteem and depression. The relative defencelessness of the victim implies an obligation on others to intervene, if we take their democratic rights seriously. Olweus (1993) has argued that it is a ‘fundamental democratic right’ not to be bullied. The increase in international concern about school bullying appears to reflect an increase in concern for rights issues throughout the twentieth century (Greene, 2006), which has been evidenced by an awareness of, and legislation against, forms of discrimination on the basis of (for example) sex, race, age, religion, disability and sexual orientation – a process that is still continuing. Although bullying or victimisation generally refers to discrimination on a more individual basis, the term bias bullying has been used to refer to victimisation based on such group characteristics.

A number of European countries have developed legal requirements concerning bullying in schools, including Finland, France, Ireland, Luxembourg, Malta, Sweden and the UK; others (e.g. Belgium and Germany) have legal requirements concerning violence in schools (see Ananiadou and Smith, 2002^).

What we know about school bullying

Given this history of research, as briefly sketched previously, what have we learnt? Here I focus on pupil–pupil bullying, which is by far the most extensively studied. In school, we can consider teacher–teacher (although this comes under workplace bullying), teacher–pupil, pupil–teacher, as well as pupil–pupil bullying (Terry, 1998). However, it is mainly pupil–pupil bullying which has been the focus of research up until now. I review the following topics: methods of study; incidence figures; types of bullying; roles in bullying; causes of bullying and correlates of bully and victim roles; structural features of bullying; contributions of theory; implications for intervention; and proposals for future research.

Methods of study

It is not immediately obvious how to study school bullying, as it generally is not something that bullies would want to be discovered doing. However, several methods have been developed and used to study the phenomenon.

A straightforward method is to get teacher or parent reports. Teachers, or parents, can usually give some information on whether their pupil, or son/daughter, is or has been involved as a bully or victim. Such information is easy to get, and teacher opinions are important in their own right. But it is generally of limited value in understanding pupils’ experiences, as other data suggest that adults often know only a fraction of what is going on. As far as pupil–pupil bullying is concerned, it is the pupils themselves who are the expert witnesses. Teacher-reports may nevertheless be most useful for young children around four to five years of age (Monks et al., 2003).

A widely used approach has been pupil self-report data in questionnaires, usually anonymous. Olweus (1996) developed this methodology, and other variants have been used, for example the Life in School questionnaire in the UK (Arora, 1994). Such a method is most suitable for large surveys, and indeed has no substitute in this respect. Of course, the information is limited to what is in a structured questionnaire, and if it is anonymous, individual bullies or victims cannot be identified. Generally, correlates of bully or victim (or bully/victim) status can only be made with one individual’s questionnaire response – that is to say, it relies on one informant.

Peer nominations provide an alternative pupil-based approach. Here, pupils are asked to nominate classmates for involvement in roles such as bully or victim. Multiple informants can provide good reliability in identifying bullies or victims, and this method is suitable for class-based work. It is more time-consuming (involving individual interviews), and ethical issues are raised by asking pupils to identify others in this way. Two common instruments are those by Rigby and Slee (1991) and the Salmivalli Participant Role Scale (1996).

Direct observations can be made, for example by watching in the playground and making notes or discrete audio-taped observations (Boulton, 1995), or by using video and radio microphones (Pepler and Craig, 1995; Pepler et al., 1998). Such methods are particularly difficult and time-consuming, and are not used widely. It may also be biased towards detecting physical bullying, rather than more subtle verbal and indirect forms. Nevertheless direct observations have a special validity, as all other methods are indirect, filtered through perceptions of individuals and the nature of reporting to the investigator.

Table 3.1. Incidence of bullying and being bullied in US students (from Nansel et al., 2001)

Table 3.2. Incidences of being a victim of bullying, with gender and age differences (10 to 14 years), in four countries (from Morita, 2001)

More qualitative methods can be employed, such as focus groups, ethnographic observations and interviewing key informants. Important examples in this tradition have demonstrated the range of girls’ bullying (Owens et al., 2000) and of sexual bullying (Duncan, 1999). These do not exhaust the range of methodologies, as records kept by schools, such as incident reports of bullying, may also be useful.

Some studies have reported on correlations across different methodologies, typically for identifying bullies and victims. These are usually reasonable but no means high correlations, and reinforce the advantage of multi-method approaches (Monks et al., 2003; Pellegrini and Bartini, 2000).

Incidence figures

All the above methods can provide information on the incidence of bullying and victimisation, although some are better suited to this, especially survey questionnaires. However the actual incidence figures obtained can vary very greatly, independent of the actual phenomenon. Two examples from large-scale surveys are given in Tables 3.1 and 3.2.

Nansel et al. (2001) reported a survey of over 15,000 students in US schools, in Grades 6 to 10 (ages 12 to 16). Their figures, summarised in Table 3.1, are based on a self-report questionnaire, and show frequency of being bullied/bullying others, in the last term.

Morita (2001) provided a report to Monbusho (Ministry of Education in Japan) on a cross-national study of bullying (or ijime). This used the same self-report questionnaire with 10- to 14-year-old school pupils in Japan, England, the Netherlands and Norway, with samples of several thousand in each country. Table 3.2 shows percentages of pupils who reported being bullied (victim) more than just once or twice in the last six months, based on a self-report questionnaire. Differences by sex and age are also shown.

Even considering just questionnaires, incidence figures will be influenced by: what time span is being asked about (e.g. last month, last term, last year, ever at school); what frequency is regarded as bullying (e.g. once/twice a term, once a month, once a week or more); what definition is used (e.g. whether it includes indirect as well as direct forms, and also cyberbullying). When the questionnaire is administered in the school or calendar year can also be important, if a short time span (last month or term) is taken.

All these issues often make it difficult to compare across different studies. It also means that absolute incidence figures are rather meaningless in isolation. Some broad generalisations can be made. It is clear that victims of bullying are a substantial minority (maybe around 5–20 per cent of pupils), and bullies (who take part in bullying others) are usually a smaller minority (maybe around 2–20 per cent). Those who are both bullies and victims, the bully/victims, vary greatly in incidence according to methodology and criteria.

Despite the difficulties associated with incidence figures, such information is important, for two main purposes. First, reports of incidence can be vitally important in awareness raising, and associated publicity, when concern about the topic of school bullying is lacking. In England, some early incidence figures led to wide publicity and nationally funded action, and this scenario has been repeated in various forms in several other countries (Smith et al., 1999b). Second, incidence figures are necessary for monitoring and evaluating the effects of school-based interventions. When the same instrument and criteria are used at pre- and post-test (and by the same research team), then there can be reasonable confidence that the figures obtained are meaningful, at least in terms of registering a comparative increase or decrease in the phenomenon (although, even then, problems raised by increased awareness, and changed pupil definitions of bullying following the intervention, may contaminate the straightforward interpretation of the findings).

The media often portray bullying as a problem that is on the increase. In fact (with the probable exception of cyberbullying; see Rivers et al., Chapter 10), it appears to be decreasing in England. For example, over the last few years, data gathered from some 16,000 pupils in Leicestershire schools (Pupil Attitude Survey 2005/2006) show that the proportion who say that they have ever been bullied in school this year more than just once or twice was 16.3 per cent in 2002/2003, 14.9 per cent in 2003/2004, 14.4 per cent in 2004/2005 and 13.9 per cent in 2005/2006. This slow but steady decline does suggest that anti-bullying work is having an effect, but also that much remains to be done.

Types of bullying

While a number of typologies of aggression and of bullying exist, the main types include:

• Physical – hitting, kicking, punching, taking or damaging belongings.

• Verbal – teasing, taunting, threatening.

• Social exclusion – systematically excluding someone from social groups (‘You can’t play with us’).

• Indirect and other relational – spreading nasty rumours, telling others not to play with someone.

• Cyber – using new forms such as text message, email, website bullying.

A 14-country cross-national study of words for ‘bullying’ (Smith et al., 2002), and the situations they are applied to, found considerable similarities across various countries in these types (though with some variations, particularly related to social exclusion situations). Boys and girls generally described bullying in the same way, even though their experiences may differ (see below). However there were noticeable age changes. Young pupils (e.g. up to about eight years) seem to mainly distinguish ‘nice’ and ‘nasty’ behaviours, so that bullying is not distinguished readily from other aggressive or unpleasant behaviours. Older pupils however distinguish bullying from fighting, and also distinguish the main forms as outlined above (Monks and Smith, 2006).

Bias bullying

Some bullying is based on (or justified by) the victim being a member of a particular group, often a marginalised or disadvantaged one, rather than on individual characteristics. Children can experience racist teasing and name-calling, and those of non-white ethnic origin have been shown to experience more racist name-calling (though not necessarily other forms of bullying) than white children of the same age and gender. Bullying can take place on the basis of a child’s sex, or involve sexual harassment (Duncan, 1999). Mainly in secondary schools, children may be teased about their sexual orientation, and even physically assaulted or ridiculed about this by other pupils or teachers (Rivers, 1995).

Roles in bullying

The obvious roles obtained from questionnaires and from nomination data are those of bully and victim. It is usually these incidence figures that are reported. It is easy to calculate data for two other roles. Bully/victims are those who score on both bully and victim, and non-involved or control pupils are those who score on neither. However, this is only the bare bones of a more complicated dynamic.

Pikas (2002) and others had long distinguished two types of victim: the so-called passive victim, who has not directly provoked the bullying; and the provocative victim, who can be thought to have contributed to their being bullied by having acted in an annoying, provocative way to peers (thus, this group may overlap considerably with the bully/victims).

Salimivalli et al. (1996) took things further by splitting up the roles involved in bullying. Through peer nomination procedures she identified roles of ringleader bullies (who take the initiative), follower bullies (who then join in) and reinforcers (who encourage the bully or laugh at the victim). She also identified defenders (who help the victim) and bystanders (who stay out of things), as well as the victims themselves. Subsequently some investigators have distinguished between bystanders, who see the bullying but do not act in any way, and outsiders, who just have not seen what is happening.

Salmivalli’s work was with young adolescents. Sutton and Smith (1999) adapted the participant roles methodology for 8–11-year-olds in England, and Monks et al. (2003) adapted it further for use with young children (four to six years), using cartoon pictures. The roles of bully (or aggressor), victim and defender can be identified with some reliability even at four to six years, but the other roles are not clearly understood until middle childhood (Sutton and Smith, 1999).

Causes of bullying and correlates of the bully and victim roles

Aggressive behaviour and inequalities of power are commonplace in human groups, including peer-groups in school, so bullying can be a temptation. We can envisage many levels of causation:

• Society level – tolerance of violence, bullying and abuse of power in society; portrayals in the mass media.

• Community level – neighbourhood levels of violence and safety; socioeconomic conditions.

• School level – school climate and quality of teacher and pupil relationships; school policies on and sanctions against bullying; school physical environment.

• Family level – nature of parent–child and sibling relationships.

• Interpersonal level – attitudes of the main peer-groups in the school; nature and quality of friendships.

• Individual level – temperament and personality.

Most of the research on school bullying has been at the family, interpersonal and individual level. However, class-level factors can be important (e.g. Salmivalli and Voeten, 2004), as can community-level variables (Benbenishty and Astor, 2005). In Colombia, Chaux et al. (2009) found that high levels of community violence as well as some school variables were significant predictors of bullying rates.

Correlates of the victim role

Many studies have examined correlates of the victim role. Victims of bullying often experience anxiety and depression, low self-esteem, physical and psychosomatic complaints (Williams et al., 1996). In extreme cases, they may commit suicide (Kaltiala-Heino et al., 1999). Hawker and Boulton (2000), carrying out a meta-analysis of many studies, found that victimisation was most strongly related to depression, moderately associated for social and global self-esteem and less strongly associated with anxiety.

Hodges et al. (1997) found that having few friends, or friends who can not be trusted or who are of low status, and sociometric rejection (dislike by peers) are risk factors for being a victim. Some victims come from over-protective or enmeshed families (Smith and Myron-Wilson, 1998). Children who are both bullies and victims (aggressive victims) may come from particularly troubled or abusive families (Schwartz et al., 1997).

Having a disability or special educational needs is another risk factor for being a victim. Children with special needs are two to three times more at risk of being bullied; they are also more at risk of taking part in bullying others (Knox and Conti-Ramsden, 2003; Nabuzoka, 2000). Possible reasons for this include: particular characteristics that may make them an obvious ‘target’; in mainstream settings these children are usually less well integrated socially and lack the protection against bullying which friendship gives; and those with behavioural problems may act out in an aggressive way and become ‘provocative victims’.

Correlates of the bully role

Personal correlates of the bullying role include temperamental factors (such as being hot-tempered; Olweus, 1993), readily attributing hostile motives and having defensive egotism (Salmivalli et al., 1999); however ringleader bullies at least may have high social intelligence and theory of mind skills, although used for antisocial ends – they can be ‘skilled manipulators’ (Kaukiainen et al., 1999; Sutton et al., 1999).

At the interpersonal level, bullying children tend to be peer-rejected in infant/junior school but less so in secondary school; towards adolescence, some aggressive and bullying children can have quite high status in peer-groups (Pellegrini and Bartini, 2001).

Family factors have been commonly implicated as risk factors for children who persistently bully others. They are more likely to come from families lacking warmth, in which violence is common and discipline inconsistent. Fathers who were bullies at school are likely to have sons who were bullying at school (Farrington, 1993; Olweus, 1993).

Correlates of the defender role

Although less studied, the characteristics of those children who are prepared to actively defend victims, rather than be passive or even colluding bystanders, are of interest and importance. Some research suggests that defenders are more empathic (Nickerson et al., 2008); however, other research suggests that empathy alone is not enough. In addition, defending is predicted by feelings of self-efficacy, and high sociometric standing, such that defenders can feel confident and empowered to defend, despite the strength and popularity of some bullying children (Caravita et al., 2009).

Structural features of bullying

Characteristic age differences emerge from the self-report surveys. Self-reports of being a victim decline over 8 to 16 years (see Table 3.2 as an example), whereas self-reports of bullying others do not decline (Smith et al., 1999a). There is also a shift with age away from physical bullying and towards indirect and relational bullying, as for aggression generally (Björkqvist et al., 1992). Typical gender differences are also found. Boys are more numerous in the bully category, but the sexes are more equal in the victim category. Boys practise/experience more physical bullying; girls more indirect and relational bullying.

There has been considerable work on coping strategies used by pupils to deal with attempts to bully them. A consistent finding is that rates of telling a teacher are lower in older pupils and boys (Hunter and Boyle, 2004; Naylor et al., 2001). Another study of 406 pupils aged 13–16 years in schools with peer-support systems found the five most frequent coping strategies were to talk to someone, ignore it, stick up for yourself, avoid/stay away from bullies and make more/different friends (Smith et al., 2004b). Over a two-year period, those who had stopped being victims more often had talked to someone about it (67 per cent) than those who had stayed victims (46 per cent) or become victims (41 per cent). Coping strategies can be complex with outcomes dependent on many factors; telling teachers can be successful but needs a consistent and effective response from teaching staff.

Another finding relates to attitudes about bullying in the peer-group as a whole. Although most pupils say they do not like bullying, a significant minority do say they could join in bullying. Perhaps surprisingly, these ‘pro-bullying’ or ‘anti-victim’ attitudes increase with age up to 14–15 years (after which they start to decline). Such anti-victim attitudes are more marked in boys than girls – and especially for boys as regards boy victims (Olweus and Endresen, 1998).

Contribution of theory

One criticism sometimes made of the school bullying research programme is that there has not been much use made of theory (Smith, 2010). The majority of studies have been descriptive, making use of available instruments such as the Olweus questionnaire and the Salmivalli Participant Role Scale. There is certainly no one dominant theoretical approach, but there are a number of openings for theory.

Perhaps the most traditional approach has been from social-cognitive theory, and Crick and Dodge’s (1994) Social Information Processing Model has been influential. They propose that distorted or deviant processing of social information may result in aggression, and specifically bullying (Crick and Dodge, 1999), and that these are maladaptive behaviours. However, Sutton et al. (1999) found that ringleader bullies (those who ‘organise’ the bullying) scored highly on tasks assessing theory of mind. They argued that some bullies, rather than lacking social skills, are socially skilled in the way they organise and carry out bullying behaviour.

An evolutionary approach to bullying behaviour has begun to be developed (Kolbert and Crothers, 2003). This too would oppose a simple ‘maladaptive’ view of bullying behaviour, as it would see bullying, like other common forms of aggressive behaviour (Hawley et al., 2007), as having costs and benefits and as being in some circumstances adaptive for an individual doing the bullying (e.g. by gaining resources or defending sub-group identity), even if not beneficial for the victim or the wider community. This is consistent with, for example, Olweus’ (1993) discussion of the opportunity structure, or costs and benefits of bullying, in particular settings, and the possible adaptiveness of bullying behaviour as a strategy to gain peer-group status (e.g. Pellegrini and Long, 2002). Gender differences in bullying (as in aggression generally) can be explained in terms of sexual selection theory and effective strategies for damaging the status or reputation of males or females, respectively (Pellegrini and Archer, 2005). This perspective does not in any sense defend the morality of bullying, and there are also evolutionary roots to reciprocity and fairness (Trivers, 1971).

One example of a specific theoretical approach was the proposal of a scapegoating theory of victimisation by Schuster (1999). This supposed that there was a group need to have a victim (as ‘scapegoat’) so that one could expect one victim in most school class groups. This was a good theoretical proposal, which could be falsified, and indeed has been falsified by studies in both English and Austrian schools; see Mahdavi and Smith (2007) and other articles in the special issue of the European Journal of Developmental Psychology, 4 (3), 2007.

A more mainstream social psychological approach has come from social identity theory. Social identity is a person’s knowledge that he or she belongs to a social category or group, defined by dimensions such as ethnicity, race, religion, age and gender. From the perspective of social identity, people generally divide the social world into distinct categories, viewing other persons as belonging either to the in-group (their own group) or to the out-group (another group). Individuals in the ‘in-group’ are viewed in favourable terms whereas those in the ‘out-group’ are perceived more negatively (Nesdale and Flesser, 2001). There is also some pressure to conform to in-group norms of behaviour. For example Ojala and Nesdale (2004) found that children’s attitudes towards bullying could be moderated by in-group norms and perceived threat to group distinctiveness. This and further studies in a similar vein (e.g. Nesdale et al., 2008) have generally used artificial scenarios or group assignments, so can be criticised for not getting evidence from actual bullying behaviours, but they do throw some light on determinants of attitudes and the importance of group pressures.

Some theoretical approaches concerning individual development within the family may help to explain why some children become more likely to engage in bullying behaviour, or to be bullied. Such approaches include attachment theory, social learning theory, parenting styles, family systems theory, family relationship schemas, shame management theory and cognitive contextual theory (see Duncan, 2004; Smith, 2008).

For example, Ahmed and Braithwaite (2004) developed a shame management theory, in which shame acknowledgement plays a central role in maintaining adaptive personal relationships through effective acknowledgement, making amends and discharge of shame. Shame displacement is seen as maladaptive: debilitating feelings of persistent shame may lead to externalised hostility and scapegoating; and unidentified shame can lead to distancing from others and other-directed anger. They found that bullying others was predicted by shame displacement, lack of shame acknowledgment and parents stigmatising shaming, as well as family disharmony. Ttofi and Farrington (2008a) tested what they labelled as reintegrative shaming theory (RST), and found that mother (but not father) bonding to child related to reintegrative shaming and shame acknowledgement, and thus to reduced sibling bullying, though with some significant gender effects. Relatedly, they (Ttofi and Farrington, 2008b) suggest that these findings also fit with defiance theory (Sherman, 1993), whereby unfair or stigmatising sanctions (and failure to acknowledge shame from such sanctions) may actually increase rather than decrease subsequent misbehaviours (in this case, bullying).

Finally, some recent studies have found substantial heritability for bully and victim roles (e.g. Ball et al., 2008). This is basically an empirical finding rather than a theoretical approach, but it does suggest some vulnerability for getting into these roles, perhaps depending on temperament and/or neurological factors, and perhaps interacting with environmental factors.

Implications for intervention

Over the last 20 years, there have been an appreciable number of large-scale intervention projects in schools. Before reviewing these, it is important to state clearly that some interventions would be beyond the scope of the school. These include parent training, dealing with parental stress, reducing community violence and moderating levels of gratuitous violence in the mass media: all of these could be important if not necessary steps in reducing bullying in schools. The intervention projects to be considered did not attempt such ambitious targets but limited themselves mainly or entirely to work within the school.

The main types of intervention have included whole-school policies or class-based rules; work with pupils in classrooms, through the curriculum, to raise awareness and discuss school policy or class rules; improving the school environment, especially design and supervision of play areas and play times, where much bullying can happen; introducing peer-support systems, to help encourage outsiders and bystanders to be defenders, helping victims and challenging bullies; providing assertiveness training and social skills training to those pupils who may be victims; and working with bullying children, perhaps using Pikas or Support Group Method (Pikas, 2002; Robinson and Maines, 2007) approaches. In addition, it is important for schools to liaise with and involve parents in anti-bullying policies and work (Smith et al., 2003).

Interventions in Norway

The first large-scale, school-based intervention campaign (launched in 1983) was carried out at a nationwide level in Norway. There was a survey in schools, materials and videos for teachers, advice for parents and mass publicity, and two assessments of its impact have been reported. Olweus (1993) monitored 42 schools in Bergen. Using his self-report questionnaire, and comparing equivalent age groups, he found that, from 1983 to 1985, reported bullying fell by 50 per cent, for both boys and girls. There were also falls in reported antisocial behaviour and there was no increase in reported bullying outside school. This encouraging finding has been widely reported and has inspired much subsequent work. Roland (1993) monitored 37 schools in Stavanger. He found that, from 1983 to 1986, there was no clear decrease in victimisation, although there was a modest correlation of positive outcomes with active use of materials by the schools. The difference between these two reports may well be due to the difference in help given to schools – more support during the intervention was given in the Bergen study, while in Stavanger the researchers just returned after three years to administer the post-test questionnaires.

More recent work in Norway directed by Roland (2000) focuses on class climate and makes more use of pupils. Recent work by Olweus (2004) in the second Bergen project (1997–98) and the Oslo project (1999–2000) finds reductions in the range of 21–50 per cent in Grades 5 to 7 (though possibly less with older pupils).

Interventions in the UK

The largest intervention programme in the UK has been the DFE Sheffield project 1991–94 (Smith and Sharp, 1994). The team worked with 23 schools over four terms. Each school developed a whole-school policy and chose from a range of other interventions. There was approximately a 17 per cent reduction in being bullied for primary schools, and small reductions (around 3–5 per cent) in five of the seven secondary schools. In addition, there was a strong positive correlation between amount of effort (as assessed by both the research team and pupils) and outcomes achieved.

Other interventions

Other wide-scale interventions have taken place in many countries, including Australia, Belgium, Canada, Finland, Germany, Ireland, Italy, Spain, Switzerland and the USA. For an edited collection of intervention studies, see Smith et al. (2004a). The Olweus programme, although very successful in Norway, has not yet achieved the same levels of success in the USA (Bauer et al., 2007; Limber et al., 2004).

Meta-analyses of interventions

Amongst a number of reviews and meta-analyses of such studies, Farrington and Baldry (2007) found that, of 16 studies reviewed, outcomes were desirable in eight cases, mixed in two cases, non-significant in four cases and undesirable in two cases. In general, most interventions have some positive impact but with more modest effects than the 50 per cent reduction found in Olweus’ work in Bergen. One that does report this level of success is a project in Andalucía (Ortega et al., 2004), where a broad-based intervention was sustained over a four-year period. Reviews by Smith et al. (2003) and Ttofi et al. (2008) identified some factors related to success:

• Nature of intervention – an obvious possibility but in fact the evidence so far is that it is the effort and commitment that matters, rather than the details of what is done.

• Support by researchers – this may have explained the greater success in the Bergen compared to the Stavanger evaluations in Norway, but the evidence from the Flanders project in Belgium (Stevens et al., 2004) is that its more general impact may be small.

• Length of intervention – probably important and may help explain the relative success of the SAVE intervention in Andalucía.

• Ownership by school and effective implementation – probably very important, as suggested by the school variations in the Sheffield project (Smith and Sharp, 1994).

• Age of pupils – several projects find reductions easier to obtain in primary than in secondary schools (although Ttofi et al. (2008) reported an opposite finding, this appears to be due to their omission of studies such as those of Stevens et al. (2004) and is contrary to my own experience and that of Olweus).

• Neighbourhood, community and societal context – may be important but difficult to estimate (see Benbenishty and Astor (2005) for recent work on school violence, considering these levels).

Possible ways forward in research and practice

A number of suggestions can be made to try to improve the outcomes of anti-bullying interventions in schools. One is that we need to further design intervention materials and activities to target indirect bullying, and cyberbullying, as well as physical and direct bullying. Not all teachers and pupils recognise social exclusion and rumour-spreading as bullying (Boulton, 1997). Girls specialise more in these kinds of bullying, and there is some evidence that current anti-bullying interventions target boys more than girls (Eslea and Smith, 1998). We need to tackle girls’ bullying too, and this presents its own challenges (Owens et al., 2000). The recent rise of cyberbullying and its particular characteristics have presented a new set of challenges to anti-bullying work.

We need to consider the dynamics of bully–victim relationships and the different roles within them. We need to evaluate further the range of peer-support schemes that aim to change bystanders and outsiders to defenders (Cowie et al., 2002). We also need to understand that some bullying children are socially skilled: they may need empathy training, but simple-minded social skills training could be misguided or even counterproductive.

It is probably important to start interventions early, in infant/primary schools, before bully and victim roles become more stable. Also, it is necessary to ensure maintenance of interventions over time. Anti-bullying work cannot be ‘done’ and then dropped; it is a continuing process that needs maintenance and periodic renewal. There needs to be an understanding of this and incentives (legal requirements, and parent pressure, can contribute to this). Training of teachers in anti-bullying work should be placed on a much more regular footing, preferably included in initial teacher training courses much more systematically and thoroughly than at present (Nicolaides et al., 2002).

We need to monitor effects of different interventions carefully, using multi-methods where possible. Research has been very predominately quantitative so far, and there is a need for more qualitative studies too, perhaps detailed case studies of individual pupils or individual schools, to throw more light on perceptions of bullying, group dynamics, changes over time and processes contributing to change.