Using fiction to improve intergroup attitudes: Testing indirect contact interventions in a school context

The aim of the study was to evaluate the efficacy of an intervention using indirect intergroup contact for improving intergroup attitudes in schools, in particular anti-Roma prejudice. Vicarious contact—a form of indirect intergroup contact—can be experienced through fictional book characters, who can serve as positive role models in terms of intergroup attitude improvement. A vicarious contact experiment was conducted with sixth grade students (N = 177) from three Slovak elementary schools, using passages from the Harry Potter series. A three-group pretest-posttest design was adopted; experimental condition 1 involved reading passages without subsequent discussion, experimental condition 2 involved reading the same passages followed by a discussion, while the control group was not involved in any activities. There was a significant improvement of intergroup attitudes in experimental condition 2 compared to the control group, while condition 1 did not show any such improvement. The first contribution of the study is in tackling conceptually distinct mediators of the intervention’s effect – perspective taking and narrative transportation. The second contribution is in demonstrating the added value of discussion in interventions focused on prejudice reduction in schools.


Introduction
The aim of this study is to test the effectiveness of an indirect contact intervention based on reading literary fiction to improve intergroup attitudes in schools. Even though there has been a growing body of empirical research using literary fiction to reduce prejudice in schools (Vezzali et al., , 2014a, this study tested whether a subsequent discussion is essential for strengthening the interventions' effect. Therefore, the main contribution of the paper is in showing the added value of discussion, by separating the effect of reading with subsequent discussion from the effect of mere reading. Another contribution is in testing potential mediators -perspective taking and narrative transportation -specifically in an engaging story such as the Harry Potter series, which allows for immersion in the narration and characters.

Direct and indirect contact
Studies from the last 60 years show that direct contact between members of different groups is effective in reducing intergroup bias and prejudice, and helps to promote positive intergroup relations (Allport, 1954;Pettigrew, 1997). This effect has been demonstrated by a number of cross-sectional (Christ et al., 2014;Vezzali et al., 2014a) and longitudinal studies (Al Ramiah & Hewstone, 2012;Swart et al., 2011). The sustainability of this effect over time, in different places and with different participants, has been confirmed by a meta-analysis conducted by Pettigrew and Tropp (2008). Although the contact hypothesis was originally designed for ethnic and racial groups, the positive consequences of contact also apply to reduce prejudice towards other groups in society, such as homosexuals (Vonofakou et al., 2007), people with disabilities both physical (Cameron & Rutland, 2006) and mental (Couture & Penn, 2003), and people belonging to different religions (Scacco & Warren, 2018).
Contact between members of different groups leads to positive results, but works best if its facilitating conditions are met (Allport, 1954;Pettigrew & Tropp, 2008). These are: equal group status within the contact situation, common goals, intergroup cooperation, and support from the authority (Wright et al., 1997). In real life, direct contact is often not possible (e.g., due to segregation, deep inequality between groups, and disproportionate number of members in each group). Thus, indirect contact, which does not require face-to-face interaction, is often used in contemporary research, with newer forms still emerging (White et al., 2021). Direct and indirect contact both reduce prejudice (Pettigrew et al., 2007) and it is important to note that indirect contact should be viewed as a preparatory step towards direct contact since it reduces intergroup anxiety (Wölfer et al., 2019). The effect of indirect contact is typically stronger among people who have less direct contact (Christ et al., 2010). There are three main types of indirect contact -extended, imagined, and vicarious (Di Bernardo et al., 2017); the last one is examined in this study. Extended contact refers to a situation where a person knows that one or more members of their in-group have positive ongoing contact with a member of an out-group (Munniksma et al., 2013;Wright et al., 1997). Imagined contact is based on a mere mental simulation of such a social interaction between an out-group and in-group member (Jones & Rutland, 2018;Turner et al., 2007).
Vicarious contact may occur, for example, via reading books in which readers observe the main characters experiencing positive contact with different groups. The hero of the story acts as a fictional friend and role model for the reader, and because of the hero's friendly behaviour towards members of other groups, the story provides the reader with indirect contact. The desired outcome is the transfer of this positive vicarious experience with the representative of an out-group member, to the whole out-group. This is known as primary transfer (Boin et al., 2021). Liebkind and McAlister (1999) and Liebkind et al. (2014), used peer stories from the real world, where the protagonist's attitudes towards an out-group member changed due to positive contact experiences and friendships. The story's protagonists acted as role models for their peers (aged 13-19 years), who subsequently displayed more positive intergroup attitudes toward immigrants in Finland. The key message of the peer stories was not to deny the complexity of intercultural contact, but to focus on positive intergroup relationships. Cameron and Rutland (2006) and Cameron et al. (2011) conducted a series of studies focusing on British children aged 5-11 years. The children read short stories about intergroup friendship (over a period of 6 weeks), and reading these stories improved attitudes towards refugees and people with disabilities. The key was to emphasise the protagonist's membership in an out-group, and to achieve primary tramsfer.

Indirect contact through stories in the school environment
Similarly, Vezzali et al. (2012) showed how summer compulsory reading of fiction could be used to reduce prejudice against members of an out-group: in their stories, heroes from different cultures experienced common everyday interactions and problems. For Italian students aged 11-13 years, there was a change in attitudes and behavioural intentions, as well as a reduction in stereotyping of immigrants, when they read a book on the coexistence of different groups. Vezzali et al. (2021) also tested how stories about large social disparities between groups stimulate willingness to engage in collective action. Apart from reading stories over a period of six months, 15-year-old participants also engaged in meetings with collective discussion.
A common issue related to indirect contact effect is its durability. For example, the short-term effect of stories on a change in individuals' beliefs was confirmed by Aronson et al. (2016), Green and Brock (2000), and Wojcieszak and Kim (2016). In Appel and Richter's (2007) research, participants' beliefs were more intense after a time lag (two weeks) because of transportation into the story. Fictional stories, in contrast to scholarly texts, political speeches, or advertisements, do not present arguments to convince the reader, as readers do not necessarily have to deal with the truth or factuality of the story, which is written for the pleasure of reading. The individual thus focuses on the story and perceives the information, but later does not remember reading specific facts in a particular context of a book passage, and, after a while, accepts them as his/her own, which is called the sleeper effect (Appel & Richter, 2007;Bal & Veltkamp, 2013).

Vicarious contact via reading Harry Potter stories
Reading fiction elicits a stronger emotional effect in the reader than reading factual literature (Goldstein, 2009); similarly, reading popular fiction that depicts intergroup relations also has a greater effect, as opposed to a story written by researchers primarily for the sake of an intervention to reduce prejudice (Vezzali et al., 2014a. The Harry Potter stories are an example of such popular literature that is engaging for children and young people, while also depicting marginalised groups in the wizarding world that are mostly different from those in the real world. Vezzali and Giovanini (2012) found that the effect of secondary transfer of attitudes from one out-group (e.g., homosexuals, people with disabilities) also influenced attitudes towards another out-group (e.g., immigrants). Therefore, even through vicarious contact with fictional minorities in Harry Potter stories, individuals can change their attitudes towards minorities in the real world (Vezzali et al., 2014a). Brown (2008) examined social topics in the Harry Potter series, which describe in detail the social hierarchy of the wizarding world with descriptions of individual marginalised groups (e.g., giants, werewolves, elves), and an emphasis on explaining the efforts of wizards to maintain oppression of these groups. Gierzynski and Eddy (2013) used questionnaires and essays to find out that Harry Potter fans are more receptive to differences in other individuals, more politically tolerant, more supportive of equality, less authoritarian, more opposed to violence and torture, less cynical, and show higher political commitment than their peers. Identifying, sympathising, and feeling connected to the heroes of the book means having an understanding of the characters they accept and befriend -even creatures like a half-giant (Hagrid) or a werewolf (Lupin) who are considered inferior and dangerous in the wizarding world. The same effect has been shown for high school and university students who read the Harry Potter series while growing up, and later showed less prejudice towards stigmatised minorities (Vezzali et al., 2014b). Vezzali et al. (2014a) also demonstrated improvement in intergroup attitudes of majority participants towards stigmatised minorities after reading six literary extracts (within six weeks) from the Harry Potter series, focusing mainly on identification with the protagonist as a key factor. The experimental condition used examples from the books that were relevant to the topic of intergroup relations, focusing on prejudice and its consequences. In the control condition, the extracts were irrelevant to the topic of prejudice and discrimination. Subsequently, fifth grade students discussed with the research assistant whether this behaviour was appropriate or not and why, and in a broader context talked about prejudice against stigmatised minorities and its consequences for members of these groups. Pretest-posttest questionnaires showed that children in the experimental group had better attitudes towards out-group members than children in the control group. Thus, vicarious contact has the potential not only to evoke a more sensitive perception of a particular stigmatised group but also to create a secondary transfer, where the effect from one out-group (e.g., people with non-wizarding origin in the wizarding world) is transferred to another (e.g., immigrants in the participants' country), which was not included in the story. The main limitation of the research conducted by Vezzali et al. (2014a) was that the study of children (study 1) was conducted with a sample of only 34 children. Addition-ally, the mediator investigating the role of perspective taking was only tested with adults (study 3). Hence, we have decided to use this mediator to study its role in children, and increased the sample size. Further, in Vezzali et al. (2014aVezzali et al. ( , 2021, it is difficult to distinguish the effect of merely reading the passage from the effect of subsequently discussing it with the researcher. Therefore, the authors recommended testing an intervention based on reading the selected passages without a subsequent discussion. Accordingly, the present study compared the effects of two interventions that involved reading passages from the Harry Potter novels: one without discussion (experimental condition 1) and one with subsequent discussion (experimental condition 2), with a control group with no reading or discussion.

Parallels between processing of fictional stories and the real world
A well-crafted story makes the reader feel like the protagonist who is experiencing the situations described in the story. Identifying or sympathising with characters helps readers "leave" their surroundings and enter the story's world (Green et al., 2004;Kaufman & Libby, 2012). This process is called narrative transportation (Green & Brock, 2000. Narrative transportation increases perspective taking, thereby allowing readers to consider the character's beliefs, values, and goals (van Krieken et al., 2017). There is a temporary acceptance of the values presented in the story (Krause & Appel, 2020), the effect of persuasion is increased in the reader, and reactance is suppressed (i.e., if the content contradicts the reader's existing opinions; Murrar & Brauer 2019). This process has strong emotional and cognitive consequences, and leaves readers more open and sensitive to issues touched on by the story (Gerrig, 1993;Green & Brock, 2002). Correlational data indicate that reading stories improves people's social ability to understand the rules of interpersonal interaction, increases empathy (Lodge, 2012;Mar & Oatley, 2008), and affects prosocial behaviour (Johnson, 2012). This creates parallels with the real world, where the reader uses the experience gained.
It is crucial to focus on the similarities between the protagonist and the reader (Slater & Rouner, 2002), because the reduced social distance between them affects the success of persuasion (So & Nabi, 2013). Therefore, peers are chosen as the main protagonists of the texts, as identification with the characters influences readers' sensitivity and experience of the story (Mäkinen et al., 2019;Moyer-Gusé & Nabi, 2010). Several studies have found significant correlations between the racial attitudes of adolescents and their peers, which are explained by the perceived importance of compliance with intragroup norms (Liebkind & McAlister, 1999;van Zalk et al., 2013). A peer's action is less authoritative than that of an adult, which contributes to the impression of less pressure to change attitudes and, thereby, making the recipient more susceptible to the story (McGrane et al., 1990). In addition to similarity, the protagonist's likability is also important -readers prefer to identify with a "good" and moral figure (van Krieken et al., 2017). Subjective perception of the reader's similarity with the protagonist could be, at least temporarily, accompanied by an effort to shift one's own qualities towards those manifested in the story (Sestir & Green, 2010). Repeated exposure to a favourite hero (Sestir & Green, 2010) or story of a similar type (Bal & Veltkamp, 2013) can lead to chronic activation of these traits.
Language processing in the form of stories involves perceptual simulations similar to real experiences and, therefore, activates similar neural pathways (Sanford & Emmott, 2012;Zwaan et al., 2002Zwaan et al., , 2004. For example, when reading about intergroup contact, similar nerve centers are activated as during actual involvement in intergroup activities (Mar, 2004;Speer et al., 2009). Emotional excitement of a person immersed in a story has been measured with galvanic skin reaction and heart rate variability (Sukalla et al., 2015); for emotionally tuned passages, neural signals associated with empathy have been activated (Hsu et al., 2014). According to Gabriel and Young (2011), experiencing a story leads to psychological assimilation with the group presented in the story. Implicit (Implicit Association Test) and explicit (Transportation scale, Narrative Collective Assimilation scale) methods confirm that reading a story about wizards (the Harry Potter series) and vampires (the Twilight saga) leads to psychological assimilation with the characters in the story. Furthermore, collective assimilation through the story is related to fulfilling the basic need to belong to a group (Gabriel & Young, 2011). It does not matter for the reader, if the story is fictional or based on facts -acceptance of the story and transportation are equally strong in both cases (Carpenter & Green, 2012).

Present research
In Slovakia, a small country in Central Europe, direct intergroup contact with members of minority groups is problematic or limited, especially for young people growing up in an environment with low cultural and ethnic diversity (International Organization for Migration, 2021; Vašečka, 2009). Attitudes towards minorities are largely negative and, in many cases, based on a shared history of conflict. The Hungarian minority has been traditionally accused of endangering the integrity of Slovakia and the Slovak language, the Roma minority has been constructed as abusing the social system and blamed for their own poverty (Kende et al., 2017), and Muslims have been viewed as a threat to Slovak culture and values (Vašečka, 2009). Slovaks have minimal interest in dealing with and resolving the issues of ethnic as well as sexual minorities (Bianchi & Luha, 2010). The conservative attitudes of 20% of the Slovak population towards sexual minorities' rights to marry and adopt, were also reflected in the (invalid) referendum on the "traditional family" (Statistical Office of the Slovak Republic, 2015). This hostile normative context is similar to many other Central and Eastern European countries (see e.g. Kende et al., 2020), which makes Slovakia a good case study for the CEE region.
In Slovakia, as elsewhere in Europe, anti-Roma prejudice is one of the most widespread and normatively accepted biases (Kende et al., 2017Findor et al., 2020), which results in socio-economic deprivation of the Roma minority, as well as difficulty accessing basic resources to meet their needs (European Union Agency for Fundamental Rights, 2014). Segregation or marginalisation of children at school is common (Balážová, 2015). The Roma are disadvantaged in housing and employment, and suffer from a general lack of integration (United Nations Development Program, 2012). Programs to reduce prejudice implemented so far have had minimal results or even volatile effects (Hargašová et al., 2021). The media share information with negative framing and generalise unfavourable news reports about individuals to the entire Roma population (Kroon et al., 2016). Poor relations with the Roma minority are reported by those members of the majority who are in direct (Kende et al., 2017;Rosinský, 2009) as well as indirect contact with the Roma (Vašečka, 2001). A generally negative perception of the Roma minority in society (European Commission, 2015) is also present among students of pedagogy, active teachers, and teaching assistants (Kövérová, 2016;Rosinský, 2009). Yet, it is crucial to emphasize that despite negative attitudes and discrimination, the Roma minority is relatively under-researched in prejudice reduction literature in mainstream social psychology .
Previous research (Paluck et al., 2018) has shown that contact interventions work better for some societal out-groups than for others, with lower effects of interventions to reduce prejudice towards racial and ethnic minorities. In the present study, we expected to find more negative attitudes towards strongly stigmatised minorities, such as Muslims, the Roma, homosexuals, and refugees (Kalmárová et al., 2017;European Commission, 2015;Vašečka, 2009). At the same time, we expected more positive attitudes towards conventionally more accepted minorities, such as wheelchair users, the Hungarian minority, and Black people (Paluck et al., 2018;Velšic, 2017). Since attitudes and recognition towards different subgroups differ considerably in the Slovak context, we chose to focus specifically on attitudes only towards homosexuals instead of the entire LGBTQ + community (Gyárfášová, 2008) and only wheelchair users among people with disabilities.
In order to better understand the position of the Roma minority in the Slovak intergroup setting and to verify the effectiveness of our intervention, we decided to use a feeling thermometer for six more stigmatised minorities. Feeling thermometer is a single item measure, which is fast and easy to use. We expected that even if the intervention effect would be nonexistent for the intergroup attitudes towards the Roma, who are deeply stigmatised, it could help us to see the intervention effect on intergroup attitudes towards other minority groups. The subsequent measures focused solely on the Roma minority that was the focal group in our study, since adding more items concerning six other minorities would have been tiring for our participants.
To further explore intergroup attitudes towards our out-group of main interestthe Roma minority-we also focused on social distance, intergroup anxiety, and intergroup trust. Even though intergroup trust is difficult to establish and maintain, positive experiences through direct and indirect contact can help in building it (Boin et al., 2021;Tam et al., 2009). This is hindered by the fact that direct negative contact has a stronger effect on intergroup attitudes than positive contact, even if it is less frequent than positive contact (Graf et al., 2014). Intergroup anxiety has been defined as a negative feeling experienced during interaction with an out-group member (Turner et al., 2007); reduced anxiety can increase a sense of security in intergroup contact and, therefore, lead one to seek and engage in intergroup interactions (Page-Gould et al., 2008). Social distance indicates the extent of people's acceptance or rejection of others from different groups. Despite "immutability" of relatively large social distance towards the Roma in Slovakia, this is an important component of intergroup attitudes and needs to be systematically verified. Thus, these three important components of intergroup attitudes frequently used in the literature (Lášticová & Findor, 2016) were chosen to evaluate to what extent our intervention affects intergroup attitudes of participants.
We chose the age group of 6 th grade students of elementary school (11-12 years) for two reasons: first, Raabe and Beelmann's (2011) meta-analysis indicated that contact interventions with minorities are needed after there has been a renewed increase in prejudice among children aged 10 years and older. Second, we chose participants who were similar in age to the main characters in the Harry Potter series to make it easier for them to identify with the characters.
The aim of the present research was to determine whether reading literary stories without subsequent discussion (experimental condition 1) and reading literary stories with subsequent discussion (experimental condition 2) would improve the attitudes of students towards stigmatised minorities in Slovakia, with particular focus on the Roma, as compared to the control condition.
We hypothesised that the interventions would have an effect on improving intergroup attitudes; with the effect shown in condition 2 being stronger than the effect shown in condition 1, and that this effect would be mediated by higher narrative transportation (Appel et al., 2015) and higher perspective taking (Boin et al., 2021;Vezzali et al., 2014a) 1 .

Study design
Our experiment investigated the effectiveness of an intervention based on working with literary passages from the Harry Potter book series to improve intergroup attitudes in primary school students, based on a three-group pretest-posttest design. Whole classes were randomly assigned to the control group and experimental conditions 1 and 2. The experiment took place during three recurring meetings, since interventions using recurring meetings have been more effective than those using one-off meetings (Ülger et al., 2018).

Participants
The research sample consisted of 177 students(79 boys, 76 girls, 22 not specified) from three public primary schools with an average age of 11.88 years (SD = 0.51). Participants in each school were randomly assigned to three conditions: control (n = 63), experimental 1 (n = 52), and experimental 2 (n = 62).
All schools were selected from localities with a low concentration of the Roma minority (based on Mušinka et al., 2014), since the effect of indirect contact interventions would be negligible in places with high levels of direct contact.

Procedures
The main study was inspired by an existing research design (Vezzali et al., 2014a) in which selected passages from the Harry Potter book series were read and discussed. The control condition in Vezzali et al. (2014a) was based on reading Harry Potter passages irrelevant to intergroup relations. In the present research, experimental conditions 1 and 2 used passages that explicitly addressed the issue of prejudice and stereotyping; for example, passages where Harry encounters marginalised groups in the wizarding world. To account for the possibility that the effect of the experiment is in part caused by the discussion (Meleady et al., 2012), and not by reading a passage from Harry Potter itself, we created two experimental conditions. Experimental condition 1 involved only reading extracts from Harry Potter stories, while in experimental condition 2, the passages were also discussed after the participants had read them. The control condition did not include any intervention, and these participants were only administered questionnaires for measuring the level of intergroup attitudes -this was preferred instead of reading passages unrelated to intergroup relations, as reading in general can have some effect on students. Moreover, as the Harry Potter saga is well known in Slovakia thanks to the Harry Potter movies, we would not be able to rule out that, even if reading neutral passages from the Harry Potter books, the participants were aware of the inequalities in the wizarding world based on their previous knowledge of the story (which would have an effect on their intergroup attitudes).
The structure of experimental condition 2-meeting, discussion, and subsequent activities-was inspired by the teacher manual "Stories of Friendship" (Solares et al., 2012). At the beginning of each meeting, theoretical information was provided, or an activity was conducted to explain the key terms related to intergroup relations, which were also used later during the discussion. This was followed by reading the passages aloud en masse, which is a realistic process in Slovak schools (not in small groups as in Vezzali et al., 2014a), and then completing a short questionnaire regarding the passage that was read, which was later used to evaluate the students' perspective taking and transportation. The meeting ended with a content-oriented structured discussion, where collective participation was stimulated. The discussion structure focused on the main characters of the story and their identity and membership in specific groups within the hierarchy of the wizarding world -highlighting disparities in rights and obligations. In each meeting, different groups were mentioned in the stories, and students critically evaluated the quality of life of different groups and other important aspects of intergroup relationships depicted in the stories. Specifically, an identifiable peer (in our case, Harry Potter), was always mentioned as a positive example of tolerant intergroup views and actions (as in Vezzali et al., 2021). No other stories, besides those read at the meeting (or previous meetings), were discussed. The first rationale of this structure was to help students to achieve perspective taking by engaging with different characters. The second rationale was to discuss the same topics in different classrooms that participated in experimental condition 2. A structured discussion with a discussion guide could also help teachers to be more ready to use materials in the classroom in the future.

Measures
Intergroup attitudes were measured both before and after the intervention, using selected parts of the INTERMIN questionnaire (Lášticová & Findor, 2016) and questions focused on the Harry Potter series. Items from the INTERMIN questionnaire were used to examine attitudes towards all stigmatised minorities: the Roma, black people, the Hungarian minority, wheelchair users, refugees, homosexuals, and Muslims (feeling thermometer 0-100). For the Roma minority, we also focused on social distance (3 items, α pretest = 0.89, α posttest = 0.92; e.g., "To what extent would it be acceptable or unacceptable for you if you had a Roma classmate?"), anxiety (4 items, α pretest = 0.74, α posttest = 0.44; e.g., "If a new Roma classmate came to your class, would you feel nervous?"), trust (3 items, α pretest = 0.79, α posttest = 0.76; e.g., "In general, I trust Roma people."). After reading a short passage, students were asked to complete an additional questionnaire comprising 22 items (based on Appel et al., 2015): six statements focused on perspective taking (α = 080; e.g., "I was able to identify with at least one character in the passage."), thirteen statements examined components of transportation (cognitive α = 050, emotional α = 0.54, general α = 0.72, and imaginative α = 0.73; e.g., "I was wondering how the story would end."), and three statements focused on story properties (α = 0.89; e.g., "The passage we just read was very engaging."). Apart from the feeling thermometer, all other items were rated on Likert scales ranging from 1 (completely disagree) to 5 (completely agree).
The pretest questionnaire was completed one week before the first intervention session, while the posttest questionnaire was completed one week after the last session (total experiment duration: 5 weeks). The questionnaires were pretested in the first pilot study that was conducted with nine children-from the same school grade and county as those in the main study-using cognitive interviews (Lavrakas, 2008). Three groups were formed comprising three children each, and approximately half an hour was spent per group to thoroughly explore the cognitive processes that individuals used to answer the questionnaire (e.g., "How did you come to understand this word/question?"). Based on these interviews, changes were made to vocabulary and visual aspects (e.g., smiley faces scale) of the questionnaire, in order to improve readability and comprehension.
The second pilot study was conducted-with a sample of 55 children, aged 11-14 years-to gain an understanding of passages from Harry Potter that depict prejudiced intergroup attitudes towards different characters. Six passages were selected from across the series that met the following criteria: • relevance -depicts intergroup relations. • understanding -does not require knowledge of the rest of the story. • representativeness -different minorities are shown as out-groups. • identification -Harry Potter is one of the characters present. • length -not longer than two pages.
These six pre-selected passages were read in class during the second pilot study, and followed by a short questionnaire assessing understanding, likeability, and transportation (similar to the questionnaire used in the experimental conditions in the main study). Responses for these factors were ranked on Likert scales ranging from 1 (completely disagree) to 7 (completely agree). Based on the highest scores, three final passages that met the aforementioned criteria were selected: Harry Potter meeting house-elf Dobby for the first time and acquiring understanding of his position within the Malfoy family; Harry defending and choosing to stay with his indigent friend Ron during open verbal conflict with Draco Malfoy when first traveling to Hogwarts; Harry understanding the meaning of swearword "mudblood" that Draco Malfoy used to insult Hermione 2 .

Results
The data 3 were analysed using various statistical methods depending on the hypothesis being examined. The first method used the Brunner-Munzel test, a non-parametric statistical test for stochastic equality of two samples, which allows for the testing of hypotheses regarding differences between groups, differences between items, and pretest-posttest differences. This method was used to examine attitudes toward minorities. Additionally, we applied the structural equation modelling with MLR estimator, which is widely used in behavioural sciences. This method represents a technique for investigating relationships between latent variables that are measured by multiple manifest variables. Therefore, this tool was ideal to study the mediation effects, the overall effect, and intergroup attitudes toward the Roma minority.

Intergroup attitudes
First, we were interested in whether intergroup attitudes towards several minorities would improve as a result of participating in the intervention. To answer this question, we measured the emotional component of attitudes, using the feeling thermometer (scale 0-100) in the pretest and posttest.
The assumption of normality was tested by the Shapiro-Wilk test, homogeneity of variance was tested by Levene's test, and the absence of outliers was tested using boxplots (i.e., the interquartile range). As no variable had a normal distribution, many of them contained extreme values, and homogeneity of variance was violated several times, it was not appropriate to use parametric tests. Thus, the nonparametric Brunner-Munzel test in the R program (Brunner et al., 2002), specifically, the "ParLD" library (Noguchi et al., 2012), was used. This test allows for testing of hypotheses regarding differences between groups, differences between items, and pretest-posttest differences. We were mostly interested in interactions: specifically, whether there was a difference between the pretest and the posttest in the experimental groups (i.e., group × time interaction) and whether there were differences in individual items (i.e., group × item × time interaction). It is possible that the intervention is not effective for each of these groups, but only for some; for example, it may be effective for the Roma minority, but not for refugees. A significant group × time interaction would indicate that the intervention was effective for each group. Even if this interaction was not significant, the intervention may still be effective for individual groups (though not for all) and the group × item × time interaction would be significant.
The dependent variable was the participants' answers to individual items. Independent variables (predictors) were differences between groups (three groups -experimental 1, 2 and control), differences between items (seven items probing attitudes towards different minorities), and differences in pretest/posttest measurements (two time points). ANOVA-type statistics of the nonparametric Brunner-Munzel test showed that the group × time interaction was not significant, F(1,94) = 1.73, p = .179, although the group × item × time interaction was significant, F(10.09) = 2.06, p = .023. This means that at least one of the conditions was effective; however, only for some items. To accurately interpret which items changed significantly and in which condition, the relative effects for each item and each group needed to be examined. Relative effects are a good measure of the magnitude of effects, as they express the probability that the randomly selected value from the data will be lower than the value from the group being compared (and thus have easily interpretable values ranging from 0 to 1). Another advantage is the ability to calculate confidence intervals. Figure 1 shows the relative effects for the control group, experimental condition 1, and experimental condition 2; in each group, the results of measurements were measured by time (1pretest, 2 -posttest) for individual items (colour-coded, items 1-7).
Apart from showing that there are differences in intergroup attitudes toward various minorities in Slovakia, a detailed look at the relative effects in Fig. 1 reveals the following results. First, intergroup attitudes remained unchanged towards Black people, the Hungarian minority, and wheelchair users following the intervention. Second, intergroup attitudes toward any minority did not worsen because of the experiment. Third, intervention 1 (reading only) had some effect on attitudes towards the Roma minority, refugees, homosexuals, and Muslims, but this effect was small and not statistically significant. Fourth, intervention 2 (reading with subsequent discussion) had a relatively large and significant effect on attitudes towards the Roma Fig. 1 Relative effects based on the feeling thermometer Note: Item 1 = Roma minority, Item 2 = Black people, Item 3 = Hungarian minority, Item 4 = Wheelchair users, Item 5 = Refugees, Item 6 = Homosexuals, Item 7 = Muslims minority, refugees, and Muslims, and large but not statistically significant effect on attitudes towards homosexuals. Table 1 shows that the confidence intervals for items 1, 5, and 7 (Roma minority, refugees, and Muslims, respectively) do not overlap, so the effect is significant.

Social distance, anxiety, and intergroup trust regarding the Roma minority
Given the arguments in the previous section and possible measurement errors, we did not test the gross scores but calculated the structural model directly. The structural model for social distance had excellent fit with the data, CFI = 0.97 and TLI = 0.96, RMSEA = 0.067. We could therefore compare latent averages for individual groups. The difference in the latent mean on the pretest between experimental group 1 and the control group was not significant, M = -0.01, p = .934, so this assumption was met. However, the difference in latent means between the control group and experimental group 1 on the posttest was also not significant, M = 0.10, p = .434. Therefore, the intervention was not effective for experimental group 1 regarding social distance. The difference in the latent mean on the pretest between experimental group 2 and the control group was also not significant, M = -0.18, p = .105. The difference in latent means between the control group and experimental group 2 on the posttest was significant, M = 0.33, p = .009. Thus, the intervention was effective for experimental group 2, as social distance to the Roma minority was reduced.
The structural model for anxiety had acceptable fit with the data, CFI = 0.92 and TLI = 0.91, RMSEA = 0.053. We could therefore compare latent means for individual groups. The difference in the latent mean on the pretest between experimental group 1 and the control group was not significant, M = -0.03, p = .370, so this assumption was met. The difference in the latent mean between the control group and the experimental group 1 on the posttest, however, was also not significant, M = 0.01, p = .901. Therefore, the intervention was not effective for experimental group 1 regarding anxiety. The difference in the latent mean on the pretest between experimental group 2 and the control group was also not significant, M = 0.01, p = .713. The difference in latent averages between the control group and experimental group 2 on the posttest was significant, M = -0.15, p = .016. Thus, anxiety towards the Roma minority was reduced after the intervention, but only for experimental group 2.
Furthermore, for intergroup trust, we calculated a structural model that had excellent fit with the data, CFI = 0.99 and TLI = 0.99, RMSEA = 0.029. Therefore, we could compare the latent averages for individual groups. The difference in the latent mean on the pretest between experimental group 1 and the control group was not significant, M = -0.02, p = .904, so this assumption was met. However, the difference in  Table 1 Relative effects for items 1, 5, and 7 in experimental group 2 Note. RE = relative effect, CI = confidence interval latent averages between the control group and experimental group 1 on the posttest was also not significant, M = 0.21, p = .141. Therefore, the intervention was not effective for experimental group 1 regarding intergroup trust. The difference in the latent mean on the pretest between experimental group 2 and the control group was not significant, M = -0.04, p = .792. The difference in latent averages between the control group and experimental group 2 on the posttest was significant, M = 0.51, p = .001. Therefore, there was an increase in the trust in the Roma minority after the intervention, but only for experimental group 2.

Mediation effects
We also used more accurate structural models to test mediation, since based on previous analyses, the gross scores do not adequately capture the actual variance of the variables. The mediators were perspective taking and transportation into the story (used as latent factors). Figure 2 shows such a structural mediation model as an example of the feeling thermometer items and perspective taking (to save space, we do not display the individual measured variables of perspective taking, but only the latent factor). The figure shows that the direct relationship between the pretest and the posttest (c) was tested, along with the indirect relationships (mediation) between the pretest and perspective taking (a) and perspective taking and the posttest (b).
Since we know that the intervention was at least partially successful, we tested the hypothesis concerning whether and to what extent the intervention effectiveness was mediated by taking another's perspective and by narrative transportation. Since all mediation structural models are very similar in format, we do not list them one by one, but state in the tables the fit of individual models with the data and subsequently the parameters of the mediation models. Table 2 contains information on the fit of the mediation structural models with the data.
All mediation structural models had at least an acceptable fit with the data, except for the anxiety model and the mediator of narrative transportation. Therefore, we cannot further interpret these two models' parameters. Table 3 shows the regression coefficients of the remaining mediation models.
As can be seen in Table 5, all mediation effects were significant, except for anxiety. The magnitude of the mediation effect can be quantified as the ratio of how much of the total effect is caused by mediation. For the feeling thermometer, perspective taking mediated 33% of the total effect, while transportation mediated 34% of the total effect. For social distance, perspective taking mediated 19% of the total effect, while transportation mediated 15% of the total effect. For trust, perspective taking mediated 40% of the total effect, while transportation mediated 29% of the total effect. This shows that the degree of ability to take another's perspective was a mediator of    the effect of the intervention on all dependent variables, that is, that the intervention affected the dependent variables by increasing the ability to take another's perspective. Transportation was also a mediator of the effect of the intervention, for all variables except anxiety.

Discussion
The present research aimed to extend findings in the area of improving intergroup attitudes towards stigmatised minorities. To date, there are numerous studies working with vicarious intergroup contact via written stories, but the main contribution of this study is in separating the effect of reading with subsequent discussion from the effect of mere reading. Other existing studies combined reading and discussion in one condition, which makes understanding of their separate effects impossible. Another contribution is in the study of mediators -perspective taking and narrative transportation -in children.
Diverging patterns in intergroup attitudes towards various minorities are country-specific, and even region-specific. The students in this study's pretest showed more positive intergroup attitudes towards Black people, the Hungarian minority, and wheelchair users as compared to the other studied minorities. The interventions did not produce a desired positive shift in attitudes toward those minorities, likely due to a ceiling effect; that is, intergroup attitudes towards these groups are already positive and there is simply no room for improvement. Students showed more negative attitudes towards the Roma minority, refugees, homosexuals, and Muslims in the pretest, which corresponds to the strong stigmatization of these groups in Slovak society.
In our study, reading passages from the Harry Potter book series without subsequently discussing them did not have the desired effect -it did not affect intergroup attitudes toward stigmatised societal groups. Reading with subsequent discussion produced a large and significant effect on intergroup attittudes towards highly stigmatised minorities -the Roma, refugees, and Muslims, but a non-significant effect on intergroup attittudes towards homosexuals. One potential explanation could be that this last topic did not resonate with students and is not as significant for them as the other three topics, which are frequently addressed by the media and probably discussed in the students' families at the time of conducting the study. There is still a stigma about homosexuality in Slovakia, and it is difficult for many parents and teachers to have an enlightening discussion with students on this topic, in part due to their own personal or religious beliefs (European Commission, 2019).
The success of reading with subsequent discussion (experimental condition 2) can also be explained by the fact that children encounter stories in the form of books at home or in school and then talk about them, either among themselves or with adults. It is normal for them to have a parent or a teacher ask them questions about a story and focus on their feelings, understanding, or knowledge (Fountas & Pinnell, 2012). In our discussion, we talked about the characters and situations from the story that had implicit connections to real life. This made the participants more engaged with the situation depicted in the story. It is possible that students at this age are still waiting for a discussion after reading a passage from a story, which would allow them to better understand it -this was not done in experimental condition 1.
As reported in the results section, all mediation effects were significant, except for the anxiety model. Furthermore, narrative transportation and perspective taking are key factors in the effectiveness of interventions. If a person is not transported by the story, they only become an observer of the sequence of events who is not involved in the story. Despite the many positive effects of reading (Bavishi et al., 2016;Mar & Oatley, 2008), studies have also found that reading had negative consequences when readers were not immersed in the story. For example, Bal and Veltkamp (2013) found that the less the readers were transported by the story, the more their empathy decreased over time. This is particularly interesting, as increasing empathy, specifically towards the Roma minority, was highlighted as a central factor of the interventions to reduce anti-Roma prejudice (Nariman et al., 2020). Perspective taking is one of the most studied mediating mechanisms of contact research that improves attitudes towards the out-group (Boin et al., 2021). We have supported this claim in our study and broadened this result to the Slovak context where vicarious contact using literary fiction has not been previously investigated.
In spite of substantial empirical support for the effect of reading literary fiction on improving intergroup attitudes, the relationship between reading books and intergroup attitudes is not straightforward. Much depends on the particular story, the books previously read, reading literacy and intergroup attitudes previously held by the reader and their social referents such as parents, peers and teachers (Miklikowska, 2016). It has been argued that with avid readers, there is a ceiling effect in the tolerance rate (Gierzynski & Eddy, 2013) and that book reading may be considered hard work requiring sufficient time and psychological preparation in the case of less avid readers (Ross, 1999). Moreover, with transportation, there could be individual dispositions predetermining the degree to which an individual can experience immersion in the story (Green et al., 2004). It cannot be ruled out that the reader already has a predetermined opinion on the topic before reading, which may affect immersion in the story and lead to disagreement with the opinions read or even to a strong opposition to the ideas presented in the text (Mazzocco et al., 2010). By focusing on a fantasy saga such as Harry Potter, the students had a chance to witness intergroup relations presented in a story without being explicitly confronted with real life experiences. Furthermore, reading in class during school hours may be a different experience for the reader compared to reading with full immersion in the story in a home environment. Another factor that we are missing is the level of education of the parents, which was not measured.
As previous research (Brown, 1995) has shown, interventions are most successful when participants are only in the initial phase of getting to know members of another group and their intergroup attitudes are not sufficiently established. In our case, places with a low concentration of the Roma minority were chosen for carrying out the intervention and participants did not report being in frequent contact with the Roma. However, how attitudes are consolidated through experiences in the environment, as well as the influence of the media, is an open question. As stated by Kende et al. (2017), it is unrealistic to expect a positive effect of intergroup contact if there is a lack of minority support movements or legal and institutional support in society, and where the idea of multiculturalism does not exist or does not apply to the Roma minority.

Limitations
The first limitation was the structure of experimental condition 2, where during each meeting, key terms related to intergroup relations were explained prior to the discussion. Hence, we cannot distinguish whether it was the discussion itself or the explanation of key terms that had contributed more to the shift in the students' attitudes. Yet in schools, discussion is often accompanied by explaining and lecturing and the fact that the intervention had an effect can inspire designing educational activities to promote tolerance in schools.
Second, it is reasonable to assume that some students understood the link between the questionnaires and the intervention, and that some degree of positive effect on their intergroup attitudes was simply a consequence of the perceived social desirability of answering more tolerantly. In addition, we only examined the answers provided by students in self-assessment questionnaires, and thus only measured explicit prejudice, which is more subject to social desirability than implicit prejudice.
Third, the choice of the interval in which the posttest was administered after the last activity is also of great importance. We chose to measure the outcomes one week after the last session. We can only speculate about what the results would be over a longer time period, for example, one month after the end of the intervention.
Fourth, vicarious contact interventions, including the one presented in this study, do not usually make a clear distinction between intergroup and interpersonal relationships presented in stories. With regard to the passages from Harry Potter, we worked with specific characters and the relationships between them; the group membership of characters may have sometimes seemed secondary. That is why the subsequent discussions focused on intergroup relations.
Fifth, in the case of unreal groups, it may be more difficult to take the perspective of the other and thus, the intervention effect may not be as strong (Vezzali et al., 2014b). Unfortunately, as in Vezzali et al. (2014a) we did not assess attitudes toward fantasy characters, so we lack direct evidence on the processes underlying this specific type of secondary transfer effect -that is, we could not confirm secondary transfer, it can be only assumed.

Future research
Firstly, there is an important question of the possible generalisation of our findings to other societal and intergroup contexts. As the previous experiments based on group reading (Vezzali et al., 2014a took place in Italy and in the UK, the method can be considered effective in different contexts. Of course, these assumptions need to be verified in other contexts and with different groups of participants. Second, the effect of teacher-led vs. researcher-led interventions, culturally mixed classrooms, as well as stories with minority vs. majority main characters (as suggested in Mäkinen et al., 2019;Ülger et al., 2018) should be examined in more detail. These interventions can serve as practicable instruments to reduce prejudice in classrooms. As Mäkinen et al. (2021) pointed out that a key factor is the perceived engagement of the teacher, it would be fruitful to expand research on the facilitator effect and discover other facilitator-related elements that could be crucial for interventions.
Third, since discussion itself might have an effect, even without any prior reading of passages, it can be explored in future studies. However, the issue of the topic for the discussion would need to be resolved, since the discussion would not be based on a prior story reading. Future studies could also examine the effect of discussion in which the parallels between the text and the real world are stated explicitly (as in Vezzali et al., 2021).
Finally, as Bigler and Hughes (2010) pointed out, it is difficult to reduce stereotypes and prejudice, as interventions often only work for a part of the sample, not for all participants. Another issue is the a priori assumption (for all activities or interventions in schools that work with the topic of prejudice) that students have prejudices that need to be reduced. In some cases, it may happen that negative opinions only begin to form in the classroom as a result of negatively framed questionnaire items (for example, attracting attention to stereotypical attributes of a minority) or group discussions (negative opinions voiced by the classmates).

Conclusions
This study has extended the findings of previous research showing that indirect contact via literary fiction reduces prejudice, and applied it in a hostile normative context with strong anti-egalitarian norms (Lášticová et al., 2021;Kende et al., 2020). In addition, our study focused on a target group which is relatively underrepresented in prejudice reduction research -the Roma (Kende et al., 2017). Moreover, it brought novel insights into the role of discussion in this process by disentangling the effect of mere reading and reading with subsequent discussion, and helped to further explain these effects by examining previously untested mediators.
Education itself (Hello et al., 2004), as well as the relationship with teachers (Geerlings et al., 2017) are important sources of attitude formation in students. School is a place where children have an opportunity to experience diversity and intergroup contact, which facilitates prejudice reduction. However, in schools, many topics (such as intergroup relations) are only discussed in abstract terms. Instead of working only with theoretical concepts, it could be reasonable to connect these topics with an engaging story such as the Harry Potter series, or any other powerful piece of fiction. Our results contribute to the growing body of research showing that the use of literary fiction in schools, informed by psychological research, can foster more tolerant intergroup attitudes in future generations.