PROTOCOL: Universal Preschool- and School-based Education Programmes for Reducing Ethnic Prejudice and Promoting Respect for Diversity among Children Aged 3-11: A Systematic Review and Meta-Analysis

Cognitive Explanations

Cognitive explanations for the development of prejudice go some way to increasing our understanding of how these attitudes can arise from an automatic and often universal habit of mind (Fiske & Russell, 2010). This approach studies the underlying thought processes that can instigate and maintain bias. Research in this field has suggested how some social behaviours are not malleable by conscious control and do not entail much cognitive reasoning. Lippmann (1922) introduced the term stereotype to describe this cognitive component of prejudice. Arguments that researchers provide concerning the persistence of prejudices are that although we can teach an understanding of the importance of diversity, it is cognitively much easier to stereotype and compare differences (Fiske, Cuddy & Glick, 2007). It is reported that the discrepancy between what people say (explicit) and how they behave (implicit) reflects simple cognitive processes that do not rely on much conscious thought. To put it simply, it is easier to rely on stereotypes to make quick decisions regarding the actions of another person (Fiske, Cuddy & Glick, 2007; Olson & Zanna, 1993) and this may go some way to explain why prejudice is still a stubborn and repetitive factor in todays advanced society. Allport (1979: 20) observed that: ‘The human mind must think with aid of categories … Categories are the basis for normal prejudgement. We cannot possibly avoid this process. Orderly living depends upon it.‘

Many researchers believe that stereotypes are born from one of two sources. The first source is through an acceptance of a real difference between groups and in being so it is actually a correct representation, such as men as a whole are physically stronger than women. It is this process of stimuli simplification in the context of categorisation which allows the perceiver to depend and fall back on previous information. It becomes a stereotype however when a person applies the group level characteristic to individual members of groups, thus assuming for example that a person is strong because they are a man or that they are invariably going to be weaker because they are female. In addition, the second source of stereotyping arises when beliefs are formed quite independent of observable group differences. This faulty generalisation will often lead to the perceiver making a quick judgement across many additional facets (Hilton & VonHippel, 1996).

It is important to note that humans vary immensely in how they exhibit and feel prejudices. Individual differences in how much someone will accept/endorse a prejudiced attitude suggests that being prejudiced is an inherent personality trait. It has been found that individuals who are likely to exhibit prejudiced tendencies towards an ethnic out-group are also more likely to develop and exhibit prejudices towards other social groups as well, whether defined in terms of disability or gender for example (Son-Hing & Zanna, 2010). In this regard, inner state theory of prejudice suggests that individuals with a very strict upbringing by critical parents were most likely to develop an authoritarian personality (Adorno, Frenkel-Brunswick, Levinson & Sanford, 1950). Entrenched authoritarian personality traits were consequently found to create a tendency for individuals to be prejudiced; being more likely to be hostile to those who hold, in their view, inferior status, but also more obedient to those with elevated status. Authoritarian personality types are believed to be fairly rigid in their opinions and beliefs and fairly traditional, upholding long-established values. This personality type was also more likely to categorise people into “us” and “them” groups while seeing their own group as superior. Finally, it is important to note that there is also considerable evidence that minority group members can also harbour prejudice towards majority group members. However, much of this prejudice could be described as reactive, reflecting anticipation of being discriminated against by majority group members (Johnson & Lecci, 2003).

As has been highlighted, inherent cognitive and social-cognitive abilities such as personality, classification skills, categorical structures and perspective taking are notable contributors to inter-group attitude development in children. An awareness of one's own ethnicity, a perception of their groups' elevated status and the ability to highlight stereotypical differences between themselves and others in turn could predispose that child to develop a negative attitude or prejudice towards those people who are in a different category (Allport, 1954; Quadflieg, Mason & Macrae, 2010).

Socialisation and Group Process Explanations

In contrast to the explanations outlined above that tend to focus on internal, cognitive processes, a second set of theories of prejudice tend to focus on the influence of externally-oriented inter-personal processes. For example, one strand of social psychology has been concerned with understanding the predictability of people in favouring their in-group while displaying negative bias towards the out-group (Brown, 1995); especially where mere membership of a group can influence equality and fairness, even in cases where the separation between groups is meaningless (Tajfel, Bilig, Bundy & Flament, 1971). In this respect, social reflection theory suggests that people adopt attitudes and stereotypes about groups that correspond to the relative power and status held by those groups. Any negative attitudes and thoughts about ethnic groups are thus thought to simply reflect the structure of society (Sayers, 1983).

One theory emerging from this is based on the widely accepted belief that humans possess a desire to increase self-esteem by viewing themselves in a positive light (Crisp & Turner, 2010). Since part of their personal identification is directly influenced by the groups to which they belong, it is believed that there is a general predisposition to make biased decisions which will value more highly the characteristics of the in-group. It is suggested that there is also a tendency to make comparisons related to superiority between groups often to the detriment of the out-group (Brown, 1995). However, the study by Connolly, Muldoon & Kehoe (2007) mentioned earlier is an example of how it would be erroneous to assume that in-group preferences will inevitably lead to out-group prejudice. While they showed that children will actively engage in in-group activities they found little evidence that this led to any notable intended prejudice towards the out-group. In this sense they found that in-group preference was not reflective or indicative of prejudiced attitudes on the part of the children involved. Crucially however, the unintended outcomes of in-group preference, in relation to social exclusion and divisions, can still have negative consequences for the out-group.

Maass & Schaller (1991) offer some explanations for this general trend. They propose that participants approach the task with an inherent belief that the group to which they belong is more favourable (in-group). However, and as highlighted above, this in-group preference does not automatically lead to out-group derogation. This belief of the in-group being better will however increase effort from the participant and elicit greater pride in their group. This finding is presented in a study by Rabbie & Horwitz (1969) where chance allocation meant that one arbitrary group would receive a prize while the other would not. Even in this random context, subjects displayed significant bias in positive trait evaluations even when their team lost'.

It is suggested that negative and competitive relationships between groups will be more likely to produce prejudice and discrimination whereas inter-group interdependence and cooperative interaction that leads to successful outcomes reduces intergroup biases (Sherif, Harvey, White, Hood & Sherif, 1954).

The most prominent theoretical concept to be developed from social learning explanations is the inter-group contact theory. This theory, also known as the ‘contact hypothesis‘, is based on the work of Gordon Allport (1954). The contact hypothesis is based on the idea that positive contact and relationships between children from differing ethnic groups will lead to reduction in prejudice and increased respect for diversity.

The traditional contact hypothesis outlines specific conditions that need to be in place between groups in order for inter-group contact to lead to prejudice reduction, such as the groups: sharing equal status; working towards common goals; working cooperatively; and having support for the initiative from those in authority. By strengthening bonds between groups through contact it is proposed that there should be an increase in appreciation and understanding that, in turn, should diminish prejudice and promote respectful attitudes through friendship (Pettigrew & Tropp, 2006).

Intergroup Conflict and Societal Explanations

The third set of theories of prejudice acquisition are those that tend to explain prejudice by reference to broader structural and sociological factors rather than individual psychological processes. Rather than focusing on the causes of prejudice being intrinsic – whether being due to the effects of an individual's cognitive functioning or perceptions emerging directly from inter-personal encounters – the focus here is on extrinsic factors. In this sense, such theories tend to emphasise structural inequalities and power relations as evident in employment, housing and education (Williams & Collins, 2001; Bertrand & Mullainathan, 2004) and also reflected and reinforced through the media and popular political discourses that can tend to demonise and develop moral panics surrounding particular minority ethnic groups (Bloch, Neal & Solomos, 2013).

Whilst prejudice may reflect particular cognitive processes and/or be the result of experiences gained through inter-personal encounters, such explanations tend to be unable to account for why it tends to be certain ethnic groups that are targeted and for the consistency of the negative attitudes that others tend to hold towards them. In this sense, it is contended that ethnic prejudice among children can only be fully understood against the wider sets of societal processes and structures – informed by economic, political and social changes – that tend to create inequalities between ethnic groups and thus generate and sustain ethnic divisions and conflicts (Troyna, 1993; Gillborn & Youdell, 2000).

The key point for advocates of an ecological approach to understanding prejudice is that such negative attitudes and behaviours are unlikely to be addressed fully until the wider societal contexts that give rise to ethnic inequalities and divisions are also addressed (Tomlinson, 2008). As such, researchers adopting this more sociological perspective tend to emphasise the need to address broader structural factors that create and sustain inequalities and divisions, within the school and beyond, as well as dealing directly with children's attitudes (Troyna, 1987, 1993; Dixson & Rousseau, 2006).

Cognitive Development Approaches

The first type of approach aimed at reducing prejudice and promoting respect for diversity derive from the cognitive explanations of prejudice acquisition outlined above. As discussed, complex cognitive features such as stereotyping and prejudice can be derived from the innate human tendency to categorise. Some ways to counteract the rigid thinking pattern caused by categorisation are to: redefine the group; change the linguistic descriptors of the group; challenge the memory recall of what this group means; and/or change the group membership. An outline of some of these techniques will be discussed briefly.

Some researchers have suggested that introducing people with multiple dimensions rather than as a single category can lead to a reduction in intergroup bias (Crisp, Stone & Hall, 2006). This is known as the multiple categorisation model and in essence will describe a person with multiple labels rather than just one. Instead of simply belonging to a gender group female/male, for example, the person may be reclassified across many dimensions: female, Chinese, young, astrologist, golfer. Another popular model is known as the common in-group identity model. Within this people are reclassified as belonging to one larger, overarching group. For example, schoolchildren belonging to the categories ‘Catholic’ or ‘Protestant’ could be reclassified as being Christian.

It has also been suggested that the use of non-neutral group labels such as ‘us’ and ‘them’ are a linguistic catalyst to unconsciously instigate intergroup biases. In a study conducted by Perdue, Dovidio, Gurtman & Tyler (1990) participants had these words linked to previously neutral stimuli. The application of the words ‘us’ and ‘them’ lead to participants showing tendencies to bias and a propensity to elicit corresponding evaluative material from semantic memory, particularly negative trait memories when prompted by the word ‘them’ and positive trait memories when prompted by the word ‘us‘. A crossed categorisation model was developed by Doise (1978) that suggests that one categorisation tends to be primary and so adding another dimension should lead to intergroup differences weakening. Findings from further researchers demonstrated that bias was virtually eradicated if the participants shared membership of at least one group, however, bias increased against those who shared no common categories (Brown & Turner, 1979; Diehl, 1990; Vanbeselaere, 1991).

Brown (1995) suggested that these findings may be useful and significant to policy makers, as in theory the results show that the arrangement of social situations to contain two or more categories should encourage the persistent prejudices along any of these dimensions to be reduced. These briefly outlined studies also show that whilst the complexities of groups and the importance of these categories have real implications for understanding and interpreting the efficacy of crossed categorisation programmes, there is a need to test the models with a greater focus on the conditions in which crossed categorisation works best.

One major criticism surrounding interventions based on intentional category change is presented by Park & Judd (2005) who emphasise that it is neither feasible nor helpful to attempt to eliminate social categories. They point out that there is very little evidence to suggest that stronger group boundaries leads to an increase in out-group hostility and so weakening group boundaries would do little to alleviate the outcomes of prejudice. They state a more useful approach would be to promote inter-group harmony based on a respect for diversity and inclusion.

Socio-Emotional Development Approaches

A second approach to reducing prejudice that also derives from broadly intrinsic explanations of prejudice focuses on promoting healthy socio-emotional development. It is believed that through this, a child will be better able to understand the perspectives of others and thus empathise with them. It is believed that this, in turn, would make these children less likely to be prejudiced or to discriminate or exclude others (Bischof-Kohler, 1991).

Interventions that fall into this category are based on elements that attempt to increase children's own emotional self-awareness, their ability to recognise and regulate their own emotional states and a greater appreciation of, and thus empathy for others. By focusing on the development of empathy, children are expected to be more able to take on the viewpoint of the other person and react to the emotional behaviour of that person, thus reducing prejudices and encouraging a greater respect for diversity. In this regard, higher rates of empathy in children have been found to be indicative of pro-social attitudes in behaviour whereas low rates were associated with anti-social behaviour (Stephan & Finlay, 1999). Moreover, increasing empathy has been shown to lead to a promotion of more positive inter-group relations. One approach is the ‘jigsaw classroom‘, based on similar principles to the contact theory in that when children work closely together they will learn to take on the perspective of the other and foster friendships which will increase empathy scores in those children (Bridgeman, 1981).

Explanations for the success of empathy as a method for reducing prejudice and encouraging respect for diversity has been suggested to be due in part to its tendency to reduce perceived differences and threat. Stephen & Finlay (1999) suggest that it may also be due to the adoption of cognitive dissonance. Festinger (1962) hypothesised that an individual who behaves in ways that are opposed to their own pre-held beliefs will experience feelings of discontent and negativity. Inherently individuals will then strive to solve the negative feeling of dissonance and may alter their attitudes to better suit their behaviours. This decision to change will be more likely when the behaviour is counter to a pre-existing belief or belief structure. Individuals hold the cognitive ability to recognise this imbalance and alter their behaviour to match their attitudes even at preschool age (Aronson & Carlsmith, 1963). Children's early ability to empathise with peers will directly influence their behaviours through the experience of cognitive dissonance.

Social Learning Approaches

A third area which has provided the basis for various interventions arises broadly from socialisation and social learning explanations of prejudice acquisition and suggests that prejudice is learned in the same way any other attitude or value is acquired, through modelling of behaviour, association of the group with a negative stereotype, and reinforcement of the occurrence of this behaviour (Bandura, 1977). In earlier studies Bandura, Ross & Ross (1961, 1963) found that children who witnessed models acting verbally and physically aggressive towards a doll displayed twice as many aggressive behaviours as those children who witnessed a subdued model or no model at all. Authors also direct the reader to other factors involved in this type of learning. Children when observing the aggression being punished were less likely to show these same negative tendencies. This theory would suggest that a child who observes adults or peers who show discriminatory behaviours are likely to imitate these behaviours, especially when these behaviours go unchecked.

Social learning approaches to reduce ethnic prejudice therefore encourage and rely on positive role models. This responsibility can fall to the individual, such as the school teacher, or to a whole group, such as a group of peers. Cooperative learning techniques suggest that functional relations between groups are critical in determining inter-group attitudes.

The contact hypothesis, outlined earlier, has in turn, led to a number of interventions based on promoting positive inter-group contact. Prejudice-reduction interventions that focus on group contact can be viewed as being informed by social learning approaches given that they focus on modelling out positive relationships that it is hoped that children learn. This theory however also incorporates cognitive development elements as it is believed that through supporting exposure to out-groups in a proactive way children's existing rigid categorisations and the stereotypes which often accompany these will diminish.

The contact theory has proven useful in certain settings, most notably schools (Pettigrew, 1998). Examples of structured contact interventions include cooperative learning techniques, integrated schooling and public campaigns such as the use of high profile individual/s from out-groups to promote inclusion messages. The contact hypothesis remains one of the best-supported theories in social psychology with hundreds of studies showing that the positive effects apparently hold for a variety of different situations, settings and samples (Dovido, Gaertner & Kawakami, 2003).

The contact theory has gained some criticism in recent research, in part due to the lack of evidence associated with the need to meet the strict criteria originally associated with positive inter-group contact and as specified by Airport (1954) and as outlined earlier. In particular, a meta-analysis carried out by Pettigrew & Tropp (2006) provided evidence that the traditional ‘optimal’ conditions were not always necessary to achieve positive outcomes. The authors individually tested each of the four optimal conditions and found that individually each of these outcomes did not emerge as a significant or independent indicator of contact success. This finding was supported by 94% of the included samples showed an inverse relationship between inter-group contact and prejudice when only 19% involved using the traditional conditions in the structure encouraged by Allport. The authors conclude that when these conditions are used collectively they do generally enhance the success of the intervention, however, a true knowledge of the conditions essential to reduce prejudice require rigorous testing in order for a more comprehensive theory of contact to emerge. Researchers also point out that the processes involved in making contact successful is a cognitive issue, and that the underlying cognitive issues that will make contact successful is under-addressed and misunderstood within the literature (Rothbart & John, 1985).

Awareness-Raising Approaches

A fourth type of approach is based on the belief that prejudice is caused by a limited understanding of other ethnic groups and so attempts to reduce prejudice need to include a focus on increasing their awareness and understanding of different cultures, histories, traditions and religions. In this regard, there is now a substantial body of work first summarised by Stephen & Stephen (1984) that suggests that prejudice can arise from ignorance. This theory is supported by empirical findings including an early study that suggested that increased knowledge surrounding the history and demography of black students led to positive intergroup relationships towards black college students (Reckless & Bringen, 1933) and later studies, such as that by Stephen & Stephen (1984), indicating that contact between white and Chinese students promoted a transfer of knowledge about different cultures and actively reduced prejudice

Critiques of the awareness-raising approach tend to point out that mere exposure to diversity is insufficient to produce respect and nor does it predict a decrease in ethnic prejudice. Studies have found that promoting positive information about an out-group is simply not sufficient for change and these styles of intervention require advanced development (Bigler & Liben, 2006).

Multicultural education is perhaps the most common type of intervention in schools aimed at reducing prejudice and encouraging respect for diversity. Informed largely by the types of awareness-raising approaches described above, multicultural education has sought to reduce ignorance and thus prejudice by exposing children to, and educate them about, other cultures. This can be done through directly teaching about other religions and cultures alongside promoting the sharing of cultural idiosyncrasies throughout the classroom. Whilst placing an emphasis on awareness-raising, multicultural education also includes social learning elements as it tends to approach prejudice reduction indirectly through increasing awareness and modelling out good relations. In addition, it can include a socio-emotional learning component as well in encouraging an identification with, and thus empathy for, other groups.

Multiculturalism has been criticised for being based on a simplistic view of racism as being the result of cultural misunderstandings. Moreover, it has been suggested that, ironically, teaching about other ethnic cultures, especially if done tokenistically, can risk actually increasing children's stereotyping of and prejudices towards others (Bennett, 1995; Kincheloe & Steinberg, 1997). This, it has been suggested, can be further exacerbated by the reliance upon teachers to deliver the teaching who may well hold prejudices themselves and thus (albeit unintentionally and unwittingly) pass on misunderstanding and ignorance to their students (Lawrence, 1997). It is with this in mind that the importance of training that seeks to encourage teachers to explore their own attitudes has been stressed. In particular, Chou (2007) has suggested that a higher standard of teacher training prior to qualification will ultimately lead to better practices within the classroom which will have a direct impact on student learning.

Anti-Bias Approaches

The fifth, and final, type of approach to prejudice reduction arises from the more ecological and sociological explanations of ethnic prejudice outlined earlier and places emphasis on directly recognising and addressing prejudice and discrimination (Gillborn, 1995). Given the focus on wider societal processes and practices, these approaches that have been labelled as ‘anti-racist education’ and/or ‘critical multiculturalism‘, tend to occur at a number of different levels. Firstly, they requires teachers not to ignore instances of prejudice among children but instead to challenge these in a constructive way; through getting the children to think through what they have said, understand that it does not make sense, and to reflect upon how prejudiced/racist attitudes make others feel.

Secondly, and at another level, they require a focus on the curriculum, with an emphasis on using opportunities available to increase children's awareness of racism and ethnic prejudices and discrimination and to encourage them to develop a value base that promotes social justice and inclusion and respects diversity (May, 1999). This can be done in English, for example, through stories that encourage children to appreciate other people's experiences of prejudice and discrimination. In subjects such as geography and history there are opportunities to increase children's understanding of the slavery, migration and how inequalities and divisions arose and have been sustained. Similarly, citizenship education provides clear opportunities to explore issues relating to inequality and injustice and also to examine the efficacy of differing approaches to addressing these.

Thirdly, and at further level, an anti-bias approach would also tend to include a focus on wider structures and practices within the school with the aim of identifying and addresses inequalities that may exist (Troyna, 1987; Gillborn, 1995) . This could include monitoring the placement of children into sets and streams and working proactively with children and parents from particular ethnic groups should they be found to be over-represented in low attainment classes. It is also likely to involve explicit training and support for teachers to help them reflect upon their own attitudes and beliefs and be equipped with the skills to address instances of prejudice in a constructive manner with children when they arise.

Although empirical evaluations of programmes that include anti-bias approaches are generally positive, showing that a reduction in biased judgement processes and prejudice can be achieved (Aboud & Fenwick, 1999), it should be pointed out that this type of intervention can only be effective if tailored to the specific cognitive development and emotional maturity of the child, so that a programme aimed at a 3 year old, for example, would have a different approach and content than that aimed at an eleven year old.

Whilst anti-racist/critical multiculturalism interventions tend to be largely informed by the types of anti-bias approach outline above, they can also include other elements such as social learning where teachers model out positive behaviour in relation to the unacceptability of prejudice and racism and also multiculturalism, where children are encouraged to appreciate the cultures and identities of others. As regards this latter point, for example, Ukpokodu (2003: p. 19) describes critical multiculturalism as:

learning paradigm in which teachers and students consciously engage in the construction of knowledge, critique the various forms of inequities and injustices embedded in the educational system, and strive to gain the empowerment needed to engage in culturally responsive and responsible practice.

Advocates of critical multiculturalism feel it is necessary to confront bias and promote empathy within groups of students, in order to make it possible to change attitudes of prejudice within the school system. Aboud & Fenwick (1999) found through their study that a peer's negative reaction to a derogatory slur or racial joke has a significant impact on the future occurrence of discriminative behaviours. This type of intergroup dialogue is an emergent intervention being used within classrooms that promote critical multiculturalism. This dialogue is designed to facilitate participants with a group environment that is safe, structured and designed to promote optimal understanding between groups. Students are able to freely discuss differences between ethnic groups based on openness, trust, authenticity and the removal of any preconceived ideas or assumptions of the group. Nagda (2006) states that this type of intervention is important in promoting appreciations of diversity while improving critical self-reflection and alliance building.

Those researchers who evaluate the use of anti-racist techniques within the classroom suggest that racial harassment, the under- recruitment of minority staff in schools, inability or refusal to tackle stereotypes and the inappropriateness of the curriculum for a multiethnic society are the underlying catalysts for prejudice persistence (Short & Carrington, 1996). Fullan (1991) additionally warns that a main component often overlooked is the inability of the teacher to fully internalise the problem. The author suggests that if the teacher does not fully understand and believe in the reasons behind the changes then anti-racist education will not be effective in that context.

However, Fullinwider (2008) suggests that teachers lack confidence that they will be able to approach the issues in a way which will encourage a positive intergroup student experience. Moreover, Aboud & Fenwick, (1999) point out that difficulty may arise when implementing anti-racist strategies as parents and teachers are often reluctant to discuss such sensitive issues. There is thus a need to see an anti-racist approach as including engagement with parents and wider communities, reflecting the ecological approach to seeing prejudice as having a structural/societal cause. Some curriculum interventions based on anti-bias literature have produced promising outcomes including a respect for diversity in children (Adams & Ebbeck, 1997). In addition, there has been some research into how the family can influence children's levels of respect towards other ethnic groups although these findings remain inconclusive. While one paper reports that children's ability to handle racism is directly influenced from family discussion (Coll et al., 1996) another reports that children over five take more cues from their social surroundings when making preferences rather than basing their decisions on parental values (Aboud & Doyle, 1996).

Types of study designs

The study must report programme effects for at least one outcome variable, measured on the subjects, representing either: 1) a reduction in ethnic prejudice; and/or 2) an increase in respect for ethnic diversity.

Results must be reported in a manner that permits a reliable calculation of effect sizes. No date restrictions will be applied to studies.

Only those studies that use a control group design will be eligible. The study must contain an intervention group and at least one untrained control group. These groups (intervention and control) can be assigned randomly or non-randomly. Non-random studies will be coded in relation to how matched the two groups are at pre-test and also whether attempts were made to control for pre-test differences in the analysis. The potential impact of study design on effect sizes will be explored as part of the subgroup analyses and any significant influences will be controlled for as part of the meta-analysis through meta-regression (see below).

Control groups can include placebo, no treatment, waitlist, or treatments vs ‘treatment as usual‘. Any study which includes one group pre-test/post-test or in which a treatment group is only compared to another treatment group will not be eligible for inclusion.

Types of interventions

Interventions that will be included within this systematic review will be any universal, standardised intervention programme delivered on a whole-class basis in preschool or primary/elementary school to children aged 3-11 and that includes an explicit objective of reducing ethnic prejudice and/or promoting respect for ethnic diversity.

Types of participants

The interventions must be delivered to children aged 3-11. This age range has been selected to be inclusive of the first stage of compulsory schooling i.e. primary school stages in the UK (Years 1 to 6) and will include elementary school stage in the USA (Grade 1 – Grade 6). Also, as ethnic prejudices are first visible in children in early education settings at age three (Aboud, 1989; Aboud & Levy, 2000), and an increasing number of interventions have been designed for this initial phase in preschool and kindergatern, it is logical to include this age-range also.

If a study includes children older than 11 then the authors will first attempt to extract only the results associated with those in the eligible age range. If the results cannot be isolated due to either the author pooling the results or not being explicit with the age group then this study will be excluded.

Types of outcome measures

Primary outcomes of interest within this review are those that relate to ethnic prejudice and also respect for ethnic diversity. Those that relate to ethnic prejudice may include outcomes such as: reductions of in-group favouritism and out-group derogation; a decrease in discriminatory behaviour; and increased inter-ethnic friendships. Those that are related specifically to a respect for ethnic diversity may include: an increased knowledge of other ethnic groups; improved awareness of ethnic inequalities and discrimination; and increased commitment to social justice and inclusion.

Tools which may be used to report changing attitudes include, but are not limited to: Preschool Racial Attitude Measure 2 (PRAM2) (Williams, Best, Boswell, Mattson & Graves, 1975); Multi-Response Racial Attitude Test (MRA) (Doyle & Aboud, 1995); and Reported and Intended Behaviour Scale (RIBS) (Evans-Lacko et al., 2011). There will be a list of the most popular outcome measures provided within the final coding manual.

As there are less known standardised measures for the outcome ‘respect for diversity,‘ it is expected that the team will find a number of bespoke measures used in studies. Acceptable outcome measures that will be included in the meta-analysis will be those that have details concerning their reliability and validity included within the empirical paper or enough detail so this could be reported by the coder.

If the measures of effect are not reported within the study and cannot be ascertained by calculating reported outcome measures then the review team will contact the author directly to request this raw data. In the circumstance where this is not possible then the paper will be excluded from the meta-analysis on intervention effects but may be included for subgroup analysis. In addition, a relevant study will also be excluded from the meta-analysis if the outcomes are not measured at all.

Once a final list of outcome measures has been identified from the eligible studies identified, these will be organised into a number of coherent outcome domains for the purpose of the meta-analyses.

Types of settings

The interventions must be delivered in a preschool or primary (elementary) school setting during normal school hours. Afterschool programmes will not be included as interventions outside of school hours are vulnerable to a range of uncontrolled factors which can introduce extraneous variability which may influence outcomes and validity.

Creation of Search Terms

The number of relevant articles is finite and so keywords should be selected to make the literature search as inclusive as required, but specific enough to make the study feasible. One search strategy which has been proven effective in locating the most relevant and inclusive keywords uses the ‘pearl harvesting’ method (Sandieson, 2006; Sandieson, Kirkpatrick, Sandieson & Zimmerman, 2010). This method follows exact guidelines in order to find all relevant keywords that will locate relevant articles and has been used successfully in a previous Campbell Review (Waddington, Snilstveit, Hombrados, Vojtkova, Anderson & White, 2012). The first author has received extensive training in the method directly from the developer. The process comprises a number of steps that are explained more specifically in Appendix A but outlined briefly below.

STEP 1: The first step involves choosing some relevant studies that meet the pre-determined inclusion criteria (these are referred to as ‘the pearls‘). These studies are usually found from a relevant systematic review/meta-analysis or through using obvious search terms in a relevant journal.

STEP 2: Then the researcher should extract the relevant keywords from the ‘pearls‘. By surveying a wide number of articles in this way it is possible to gain an insight into the terminology used by authors and the terminology used by indexers for database input. This will ensure a range of keyword terms that will produce the most comprehensive and relevant return on the topic.

STEP 3: Once an extensive list of keywords has been complied, the next step is to decide those terms that can be replaced with a truncation (*) e.g. cognition and cognitive could be replaced with cogniti*

These terms will then be searched within a database to look for those that produce high precision (relevant articles). This will be done over two databases, namely ERIC and psycINFO, as a type of pilot test to decide on the final keyword list, it may be that some terms ‘prejudice’ produce a high recall into the tens of thousands (psycINFO) but low precision of those which are actually relevant to the study.

Once the list of specific terms has been determined, this is called the ‘synonym ring’ and those final keywords included on this list are defined by those keywords which uniquely produce relevant articles and which cannot be retrieved by any other term

STEP 4: The fourth step is then to validate the search keywords in the synonym ring. This will be achieved by comparing to the key terms presented in the protocol by Curry, De Amicis & Gilligan (2010) and in the systematic review by Aboud et ah, (2012) and measuring the difference in relevant articles achieved. A further step will be facilitated by comparing with a pre-existing list of terms available on database thesauri.

For the proposed review there will be three synonym rings:

Population targeted (children aged 3-11)

It is expected that literature will be widely scattered due to the multidisciplinary nature of the title and the many differences within terminology. There is also a lack of similarity used within bibliographic databases for indexing keywords (Dixon-Woods et ah, 2006). The indexes of the database alongside the authors' keywords must then be used in a standalone technique as described in step two of the pearl harvesting method (Sandieson, 2006) in order to find those words which match the description.

Key words for the proposed study may include many combinations and permutations of the following types of keywords: ‘prejudic*, ethnic*’, ‘Grade 1, Grade one, pre-school’ or ‘contact theory, cooperat*, collaborat*‘.

The full procedure for the generation of the three synonym rings, and the final synonym rings themselves, will be described in detail in the full review so that it can be fully and precisely replicated.

Databases

Comprehensive searches will be carried out on several specialist databases. ERIC and PsycINFO will be searched for educational and psychological interventions respectively. Psychological abstracts, PsycINFO, PsycLIT, Social Science Citation Index and Sociofile will also be searched to ensure inclusivity of all subject-specific indexing.

Due to the universal approach of the review, several regional databases will also be included. The British Education Index, The Australian Education Index and The CBCA education database (which includes Canadian research) will be searched to ensure maximum inclusion.

Every effort will be expended to retrieve all empirical studies that meet the predetermined and explicit inclusion and exclusion criteria. This includes studies that appear in non-published as well as published literature. Searches for unpublished studies will be carried out in the search engine Google, ProQuest Dissertation and Theses and Dissertation Abstracts International. Non-journal publications such as privately and publically funded research may be identified by searching relevant conference proceedings. The United Nations educational, scientific and cultural organisation (UNESCO) will be searched for relevant conference presentations and international ethnicity specific conferences such as ‘world conference against racism’ and ‘overcoming racism’ will be extensively searched. Government reports will be searched via the online portal such as GOV.uk for United Kingdom and the gao.gov for the US government accountability office.

To ensure consistent comprehensiveness of the search and to meet the requirement to use retrieval methods other than simple database searching, the bibliographies of previous reviews and meta-analyses in the area will be reviewed for studies that meet the eligibility criteria. Hand searches of all the relevant included studies will be conducted as well as specific journals such as psycEXTRA and ISI Index to social science specifically.

Finally, follow up searches of first and second authors of eligible studies will be conducted. These authors will then be approached via email to enquire whether there are ongoing studies or unpublished research. This will ensure that unpublished studies can be included for analysis and ongoing research can be included in the review update.

Studies identified will be retrieved online and via libraries at Queen's University Belfast (UK) and Peabody Research Institute, Tennessee (USA). In the event that a relevant paper is not accessible due to institution restrictions, then the team will request an inter-library loan or email the author directly.

All searches on electronic databases, journals, conference and government proceedings will be carried out before April 2016. Those studies which are not reported in English must be excluded due to language restrictions. The review team will include all relevant research papers from January 1980 through to May 2016.