Exploring group polarisation in the classroom: manifestations, frequency and determinants

Abstract

In recent times, many scholars have highlighted the emergence of group-based polarisation within Western societies. Research has demonstrated that this phenomenon has also reached educational settings, where groups of adolescents are opposed to each other. Surprisingly little has been written, though, on group-based polarisation in the classroom. This study examines along which lines (e.g. ethnic, socio-economic, or religious), how often and under which conditions group polarisation manifests itself in the context of secondary school classrooms. It does so by theorising this phenomenon, hypothesising its most common manifestation as well as the factors that may favour group-based polarisation. Following an exploratory sequential mixed-methods research design, consisting of a qualitative pre-study (N = 31) and an original survey that was fielded among Dutch secondary school teachers (N = 1034), it then demonstrates that: group polarisation, as perceived by teachers, is relatively rare in Dutch secondary education, generally occurring once or a few times per year; that it manifests itself predominantly along ethnic lines, and that the ethnic composition of educational contexts, tracking, and different educational stages statistically predict this form of perceived group polarisation. Based on our findings we propose that teacher training, focusing on how to cope with group-based polarisation in the classroom should be context-specific and tailor-made.

Keywords:

• Polarisation
• intergroup relations
• school ethnic diversity
• tracking
• secondary education

Reviewing Editor:

• Reza Gholami, Education and Social Justice, University of Birmingham, United Kingdom

Subjects:

• Secondary Education
• Sociology of Education
• Education – Social Sciences

1. Introduction

In recent years, many authors have signalled trends of group-based polarisation in western societies, notably along ethnic, political and religious lines—lines that sometimes overlap with one another (e.g. Abramowitz, 2013; Iyengar & Westwood, 2015; Kriesi et al., 2008). To be sure, group polarisation is inescapably part of life in democratic societies (Mouffe, 2005; Schattschneider, 1960) and can actually be beneficial for its members, as contrasts with out-groups make individuals conscious of their in-group bonds and allow them to develop meaningful identities, whether they are based on socio-economic, religious, ethnic, political, or other affiliations (Coser, 1964; Levine & Campbell, 1972; Tajfel & Turner, 1979). However, polarisation can also foster inter-group segregation and lead to overt conflicts between groups, thus becoming threatening rather than enriching for peaceful democratic societies (Kuper, 1977; Levitsky & Ziblatt, 2018).

Several studies have shown that polarisation has also reached the classroom, where groups of adolescents are opposed to one another (e.g. de Haan, 2023; Kleijwegt, 2016; Lozano Parra et al., 2021; Tyler & Iyengar, 2023). This trend is potentially problematic in educational settings, as it can lead to unsafe and hostile learning environments for pupils and teachers (Wansink et al., 2019). At the same time, schools are often seen by policy and opinion makers as the ideal context to foster inter-group friendships and to promote desired citizenship behaviour for children in society (e.g. Schofield, 1991; Thijs & Verkuyten, 2014), which, in turn, may mitigate radically antagonistic forms of polarisation.

Existing studies have paid much attention to controversial topics and the ways in which teachers deal with such controversies (Hand, 2008; Hess, 2009; Pace, 2021; Sijbers et al., 2015; Thijs et al., 2021). Yet surprisingly little has been written on (group-based) polarisation in the classroom. This is all the more remarkable, since controversies in classrooms are often rooted in opposed groups who feel threatened in their identities (Damhuis & Wansink, 2023; Goldberg, 2013; Hess, 2009; King, 2009). The aim of this paper is to assess how (i.e. along which lines), how often and under which conditions group polarisation manifests itself in the context of secondary school classrooms, where adolescents (and, thus, future generations) spend a substantive amount of their time, and where durable processes of socialisation take place (Biesta et al., 2009; Bubritzki et al., 2018; Öhrn, 2011). This approach, we argue, can help us to work out ways to deal with controversies in the classroom, for instance by developing tailored training for teachers.

As will be discussed below, we focus empirically on the perspective of these educational professionals, since they are the ones who have regular social relationships with individual students and supervise how (groups of) students behave with each other in the classroom (Endedijk et al., 2022), thus providing a privileged empirical access to the relations between and identities invoked by pupils. We thereby ask two research questions. First: With respect to which groups and how often do secondary school teachers observe polarisation in the classroom? Second, we focus on the theoretically and empirically most frequent form of group polarisation, i.e. ethnic group polarisation, and ask, which factors account for this group polarisation? Empirically speaking, this study follows a sequential mixed methods approach, consisting of a qualitative pre-study (N = 31) and an original survey that was fielded among Dutch secondary school teachers (N = 1034). Yet, before presenting and discussing this empirical strategy as well as our findings in further detail, we will first turn to the paper’s theoretical framework.

2. Theory

2.1. What is polarisation?

The central concept of this paper, polarisation, has been theorised in different ways by different researchers (see e.g. DiMaggio et al., 1996; Iyengar & Westwood, 2015). To better situate our conceptualization of polarisation within the existing literature, at least three relevant dimensions should be distinguished, regarding the (1) content, (2) temporality, and (3) expressions of polarisation (see Table 1 for an overview). When it comes to its content, a distinction should be drawn between identity or group-based polarisation on the one hand and issue-based polarisation on the other (Klein, 2020). Whereas the former—overlapping with so-called affective polarisation (e.g. Iyengar & Westwood, 2015)—is rooted in social identity theory (Tajfel & Turner, 1979) and based on contrasting liked ingroup and disliked outgroups; the latter (also known as belief polarisation) refers to the degree to which people have contrasting viewpoints on a specific topic (Lord et al., 1979). In this paper, we focus on group-based polarisation, by following the conceptualization of Boutellier (2021, p. 21), who defined polarisation as ‘tense relations between groups, whose identity determines the stakes’ (see also Esteban & Ray, 1994).

Table 1. Three dimensions of polarisation.

Dimension	Aspects	Key authors
1. Content	Identity-based vs. issue-based	Klein (2020); Iyengar and Westwood (2015); Lord et al. (1979)
2. Temporality	State vs. process	DiMaggio et al. (1996)
3. Expressions	Prejudice, exclusion and violence	Allport (1954); Mullen and Leader (2005)

Second, from a temporal perspective, polarisation implies both a state and a process (DiMaggio et al., 1996, p. 693). Polarisation as a state refers to the degree of intergroup hostility with respect to some theoretical maximum. Polarisation as a process concerns the increase of such hostility over time. This second approach was famously advocated by Leo Kuper (1977, p. 128), when he defined polarisation as ‘a process of increasing aggregation of the members of society into exclusive and mutually hostile groups, entailing the elimination of the middle ground and of mediating relationships.’ In this paper, though, we focus on group-based polarisation as a state rather than a process, as our aim is to find out which groups polarise in the classroom and which factors foster polarisation, rather than to assess how it evolves over time.

Third, polarisation might lead to overt conflict, but it may also remain at a level of articulated group differences (Schinkel et al., 2009, p. 173), for instance through stereotypes and prejudice. In other words, different expressions of polarisation can be distinguished. Such distinctions can already be found in Gordon Allport’s classic study on The Nature of Prejudice (1954), in which he identified five different forms of intergroup enmity: antilocution, avoidance, discrimination, physical attack and extermination. These five categories, in turn, as Mullen and Leader (2005, p. 202) argued, can be subsumed under three categories: (1) prejudice towards; (2) exclusion of; and (3) violence vis-à-vis outgroups. In line with Esteban and Ray (1994) and Boutellier (2021), focusing broadly on ‘tense social relations’, we do not, a priori, exclude verbal, exclusive, or physical manifestations of polarisation. Accordingly, in this paper, we conceptualise polarisation as a state of tense relations, manifesting itself through antilocution, exclusion or violence, between social groups, whose identity determines the stakes.

2.2. Which groups do we expect to polarise in Dutch classrooms?

Although many studies have been written on cultural and ethnic diversity in the classroom (e.g. Banks, 2015; Loader & Hughes, 2017; Thijs & Verkuyten, 2014), we are not aware of any publication that systematically deals with polarising groups of students in educational settings. Despite this absence, there are two strands of literature suggesting that group polarisation along ethnic lines is most salient in classrooms in the Netherlands as well as in other European countries (Damhuis et al., 2022; King, 2009; Kleijwegt, 2016; Mlinar & Peček, 2023; Pasque et al., 2013; Thijs & Verkuyten, 2014). First, a good number of publications exist, mainly in the field of social psychology, investigating intergroup relations in the classroom. These studies notably show that classmates with different ethnic group backgrounds are less likely to bond than classmates of the same ethnic group (e.g. Clark & Ayers, 1992; Knifsend & Juvonen, 2014; Rícan, 1996; Van Caudenberg et al., 2020). Teenagers living in the Netherlands are no exception to this ‘rule’ (Baerveldt et al., 2004; Thijs & Verkuyten, 2014; Vermeij et al., 2009).

Second, the existing literature on controversial topics contains important clues. For instance, Vogels et al. (2021) observed that conflict situations in Dutch classrooms mainly occur ‘between pupils with and without migration backgrounds’ (Ibid: 139) as well as ‘between groups of pupils who are from different migrant backgrounds.’ (Ibid: 174). Similar situations would mainly be evoked when cultural differences and political standpoints are discussed. These findings of Vogels et al. are in line with other studies, pointing at the saliency of sensitive topics with a prominent ethnic component, including the holocaust, right-wing radicalism, integration of ethnic of minorities and anti-muslimism (Damhuis et al., 2022, Damhuis & Wansink, 2023; Goldberg, 2013; Hossain, 2017; Sijbers et al., 2015).

Importantly, as these studies suggest, the ethnic background of polarising groups in the classroom generally cannot be understood in an isolated way, as it is often interwoven with socio-economic, political and religious elements. In that light, several authors emphasise the role played by Islam, as well as—though to a lesser extent—the rhetoric of nationalist anti-immigrant parties (in the Netherlands, notably the Party for Freedom, or PVV). Kleijwegt (2016), for instance, observed the strongest oppositions among students in the Netherlands between those ‘with a Muslim background’ (i.e. religion) on the one hand and ‘autochthonous’ pupils supporting the radical right on the other (i.e. ethnic and political). Quite similarly, in Flandres, Van Praag et al. (2015) showed that religion, while outlining in- and out-groups in a visible way, forms an important aspect of students’ ethnic identity. This interconnectedness between ethnicity (of students with and without a bicultural background) and religiosity (regarding Islam in particular) should not be surprising in the Dutch case, since the two largest ethnic minority groups in the Netherlands—having a Turkish and Moroccan background—are overwhelmingly Islamic (Huijink, 2018).¹ Moreover, public debate concerning integration of migrants in schools and beyond is strongly focused on religion, thus reinforcing the relationship between these two dimensions (Agirdag et al., 2011).

In sum, we expect that ethnicity constitutes the main dimension along which polarisation manifests itself in Dutch classrooms. Yet, based on the existing literature, we also suppose that this dimension is interconnected with political elements (notably nationalism) and particularly, and religiosity (notably Islam).

2.3. Which factors favour ethnic polarisation in the classroom?

The following question, then, is under which conditions ethnic group polarisation in the classroom is most likely to take place. We hereby expect three factors to be theoretically relevant: (1) the ethnic composition of pupils, in terms of their bi-culturality; (2) different tracks; and (3) educational stages (lower v. upper secondary). To be sure, in view of the scarcely available literature on the topic, these theoretical expectations are largely exploratory.

2.3.1. Ethnic composition

Many studies on interethnic group relations of adolescents in school contexts use the theoretical lens of contact theory (Allport, 1954; Pettigrew, 1998) and ethnic competition theory (Blalock, 1967; Quillian, 1995). The objective of similar investigations is to find out whether there is a positive (contact) or negative (conflict) relationship between ethnic composition of schools or classrooms on the one hand and the interethnic relations among pupils therein (e.g. Agirdag et al., 2011; Juvonen et al., 2006; Thijs & Verkuyten, 2014). Interestingly, the findings of existing studies are rather mixed: ethnically integrated schools could lead to both interethnic contact and to interethnic conflict (e.g. Goldsmith, 2004; Schofield, 1991; Moody, 2001).

One of the possible explanations in that respect is that little attention has thus far been paid to the potentially curvilinear relationship between intergroup diversity on the one hand and group oppositions on the other. Following this line of thought, we expect that high levels of homogeneity among pupils would foster an absence of conflict (as there is no identity threat or other source of enmity); yet high levels of heterogeneity would do so too, since the setting in high-diversity contexts resembles that of low-diversity contexts (Knowles, 2020; Moody, 2001, pp. 708–709). Empirically, this argument reflects findings from research on super-diversity in ethnically heterogeneous neighbourhoods, characterized by certain relaxedness about difference’ and ‘a focus on commonalities’ (Wessendorf, 2014, p. 170). Theoretically, it is reminiscent of insights in conflict sociology (e.g. Ross, 1920; Coser, 1964; Kuper, 1977), according to which social tensions are likely to be absent in a situation of high heterogeneity. In similar situations, the argument goes a multiplicity of crisscrossing, potentially antagonistic, groups ‘prevents the formation of a united front’ (Coser, 1964, p. 77), by cancelling each other out. Contrarily, when only one single line of -potential- conflict cuts through a group (or a classroom), e.g. along the lines of ethnicity or religion, it may divide ‘the members into two hostile camps’ (Ibid.). Therefore, as Horowitz (1985), in his seminal book on the issue of ethnic group conflict, also argued, social conflict is more likely in societies (or classrooms) in which there are two large (ethnic) groups, while being less likely in highly homogeneous and highly heterogeneous contexts (see also Dincer, 2011; Esteban & Ray, 1994; Montalvo & Reynal-Querol, 2005). Along similar lines, James Moody (2001, pp. 708–709) suggested that: ‘When there are only two races in the school, there is a greater likelihood for ‘us vs. them’ social dynamics’, compared to contexts with a higher diversity. According to Moody, these dynamics are likely to disappear in high-diversity contexts, as each group becomes a small minority in relation to the rest of the school, leading to different mixing patterns that create bridges between groups.² In view of these considerations, we expect (ethnic) group polarisation to reach its peak in a 50/50 situation, while being lowest in strongly homogeneous and strongly diverse classrooms.

2.3.2. Tracking

The findings of studies on interethnic group relations of adolescents in school contexts may also be mixed due to another oft-ignored factor: tracking. Not only because tracks are important in structuring intergroup relations and attitudes (Oakes, 2005; Van Praag et al., 2015). More importantly, the composition of tracks is heavily skewed in ethnic and socio-economic terms (Turcatti, 2018). In the Netherlands, for instance, as in other west-European countries, adolescents having a non-western migration background are strongly over-represented in pre-vocational tracks (Vmbo) and strongly under-represented in the tracks preparing for higher education (Havo and Vwo) (CBS, 2022). As Van Praag et al. (2015) indicated, regarding the case of Flanders, differences between tracks are crucially important to understand open or conflictual interethnic interactions in the classroom. In pre-academic fields of study, where the proportion of pupils with a migrant background is relatively low, they observed a ‘separation between students of different ethnic descent’, without ‘any visible conflicts’. In pre-vocational education, by contrast, where the percentage of pupils having a non-western immigrant background is highest, and where experiences of symbolic and realistic group threats (Stephan et al., 2005) are most dominant,³ ‘interethnic tensions and conflict were obvious.’

Interestingly, from a different angle, Sijbers et al. (2015) came to similar conclusions, when showing that classroom controversies concerning issues with ethnic components are quasi-inexistent in the Dutch pre-academic track, while being most prominent in the pre-vocational tracks. This tendency may be fostered by the observation that students in pre-vocational tracks are more vocal in the classroom when sharing their viewpoints on controversial topics, compared to students in pre-academic tracks (Damhuis et al., 2022; Vogels et al., 2021). Another relevant finding in this respect is that students in pre-academic education are more tolerant (in the sense of putting up with something one objects to) than students in pre-vocational tracks, which may be related to their more strongly multiculturalist beliefs (Thijs et al., 2021). Taken together, we thus expect ethnic group polarisation to be strongest in pre-vocational tracks (Vmbo) compared to tracks preparing for higher education (Havo and Vwo).

2.3.3. Educational stages

The third factor we expect to be theoretically relevant relates to educational stages. Just as in many other countries, secondary education in the Netherlands is divided into two stages: lower and upper secondary education. Dutch lower secondary education, roughly comparable to junior high school in the United States, covers the first and second year of pre-vocational education (i.e. students are generally between 12 and 14 years old), and the first three years of tracks preparing for higher education (with students between 12 and 15 years old). Upper secondary education refers to the third and fourth year of pre-vocational education (students are around 15 to 16 years old), and the fourth, fifth and sixth year of tracks preparing for higher education (with students between 16 and 18 years old). Although several studies show that prejudice against ethnic outgroups is relatively stable in adolescence (Crocetti et al., 2021), other research points out that similar can decrease when pupils grow older, especially when they have more contact opportunities, which might lead to less polarisation in the classroom (Mitchell, 2019; Raabe & Beelmann, 2011). The explanatory mechanism behind this link is that, as pupils grow older, they gain an increased capacity to think about conflict and other groups in a more nuanced way (Rekker et al., 2015). Following these studies, we expect polarisation along ethnic lines to be more likely in lower secondary education compared to upper secondary education.

3. Material and methods

3.1. Data collection

The research design of this study follows a sequential mixed-methods approach. During the first step (a pre-study for the subsequent survey), in-depth interviews were held, by the first author, with 31 secondary school teachers in all educational fields and tracks, all across the Netherlands. This fieldwork took place between September 2019 and August 2020.⁴ In line with our definition of polarisation, teachers were asked, during the interviews, if they experienced any tensions (spanningen) between pupils, and if so, when (i.e. under which circumstances) and how (i.e. in what way) these tensions manifested themselves. The interview data were subsequently used to develop the survey questionnaire, notably when it comes to the selection of respondents qua subject (biology, Dutch, geography, history, and social science⁵) and the construction of the item on forms of polarisation (ethnic, political, religious, and socio-economic), which turned out to be most salient during the interviews.

The survey was fielded online by a research institute (DUO onderwijs en onderzoek) with a large panel of Dutch teachers in November and December 2021. In total, 1040 secondary school teachers within this panel participated through voluntary sampling.⁶ The sample consists of teachers aged between 23 and 70 years (M = 47.5, SD = 11.3), with an almost equal distribution of 49% women and 49% men (a few respondents chose not to disclose their gender or selected the ‘other’ category). Six teachers were removed from the analysis, as they did not teach any of the five selected subjects. All teachers provided consent for the anonymous use of their questionnaire responses, which were completed in Dutch.

The questionnaire’s content validity was assessed through evaluation by educational experts and teacher judgments. Ten interviews were held in which experts provided feedback on the survey and teachers were asked to think aloud while completing the questionnaire. This process led to revisions in the introductory text, enhancement of reader-friendliness, and a reordering of the questions in order to mitigate bias (personal questions were placed at the end of the survey). The experts validated the way in which the term polarisation was used throughout the questionnaire.

3.2. Data analysis

To answer our research questions, we combine descriptive statistics (presented below) with binary logistic regression analyses (King, 2009). Regarding the logistic regression analyses, we use two outcome variables. One variable to measure the presence of ethnic polarisation. And one variable covering the saliency of ethnic polarisation.⁷ As to the first outcome (presence of ethnic polarisation), we used a dummy variable pertaining to the following survey question: ‘Based on your own experiences during the past 12 months: to what extent does polarisation occur in your classroom along [ethnic] lines? (Annually, Quarterly, Monthly, Weekly, Never, Don’t know).’⁸ Thereby, the presence of perceived ethnic polarisation (i.e. annually, quarterly, monthly and weekly) was coded as 1, while the absence of perceived polarisation as well as don’t know answers were coded as 0. As to the second outcome (saliency), we created a dummy variable based on the answers given to the following open question on polarisation in the classroom: ‘Between which groups of pupils (‘us’ versus ‘them’) do you experience that polarisation is strongest?’⁹ Ethnic groups were coded as 1. Non-ethnic groups as well as the absence of any form of polarisation was coded as 0. Intercoder reliability checks were carried out to measure the agreement between two different coders (Lombard et al., 2002). The results showed a strong level of agreement (86.7%).

When it comes to the explanatory variables, the ethnic composition of the student population was measured via a question in which teachers were asked to indicate the proportion of bi-cultural students in their school on a scale ranging from 0 to 100.¹⁰ Bi-cultural students were thereby defined as ‘students who, in addition to their Dutch background, also have a link to at least one other culture from the country of (one of) the (grand) parents, for instance Poland, Syria, Turkey or Morocco.’¹¹ Importantly, this explanatory variable (ranging from 1% to 97%) was not used in a linear way, for two key reasons. First, because we are substantially interested in the non-linearity of the relationship between the diversity of pupils. Second, because an important assumption underlying logistic regression modelling was not met. That is, the assumption of a linear relationship between the independent variable (bicultural background of pupils) and our dependent variables did not hold when performing Box-Tidwell tests (Box & Tidwell, 1962), yielding p = 0.025 pertaining to the first outcome variable (presence) and p = 0.011 when it comes to the second outcome variable (saliency).

One option to analyse such nonlinearity is to categorise the continuous variable (Altman, 2005). A similar approach, though, comes at significant costs, notably in terms of the reduction of available information. Moreover, it creates abrupt discontinuity in the response while assuming homogeneity in different categories (Bennette & Vickers, 2012). For a more realistic estimation and the ability to test directly for non-linearity, we thus performed analyses using restricted cubic splines (Bennette & Vickers, 2012), with three equally spaced knots at 10%, 50%, and 90% (following Gauthier et al., 2020; Harrell, 2015, pp. 27–28), i.e. including one knot at the 50% point, where we theoretically expect polarisation to be its peak. To check the robustness of our findings, we also divided the continuous explanatory variable into five categories of equal percentages: 0–20, 20–40, 40–60, 60–80, and 80–100. These categories satisfied the following four criteria: (1) to have equal percentual proportions; (2) that are odd in number, in order for one category to surround 50% (i.e. our 40–60 category); (3) to have groups that are not too large, to allow for refined analysis; (4) while being not too small in number of concerned respondents, for the solidity of the analyses. As Table A5, in the appendix, shows, this approach yields findings that are highly similar to the ones presented in the results section below (compare Tables A2 and A3).

The other two explanatory variables were categorical from the start. Educational tracks—pre-vocational on the one hand and those preparing for higher education on the other—were divided into three categories, rather than two, in order to avoid losing respondents teaching in both phases during the analysis, thus leading to the following grouping: (1) pre-vocational only (in Dutch: Vmbo); (2) respondents teaching both in pre-vocational tracks and in Havo/Vwo (preparing for higher education); and (3) Havo and/or Vwo only.¹² Following the same logic, educational stages (lower vs. upper secondary) too were coded into three categories: (1) mainly/only lower secondary; (2) equally lower and upper secondary; (3) mainly/only upper secondary.¹³

In terms of control variables, we accounted for factors on both the school level and the teacher level, in order to control for potentially relevant confounding variables, as well as individual biases among our respondents (notably regarding ethnicity). At the school level, these variables cover the proportion of pupils with a low SES; the denomination of the school (public, protestant, roman catholic or other); the region and the size of the municipality in which the school is located.¹⁴ At the individual level, we controlled for the subject respondents teach, their experience (i.e. number of years) in secondary education, their presence or absence of a migration background, and political views (i.e. party voted for: radical right and right-wing conservative or not). All the explanatory and control variables are described in the online appendix (Table A1).

4. Results

4.1. Which groups polarise in Dutch classrooms according to teachers?

In line with our expectations, Figure 1 shows that polarisation in Dutch classrooms most frequently manifests itself along ethnic lines, followed by religious, political and socio-economic forms of polarisation, respectively. At the same time, the occurrence of polarisation in the classroom is relatively limited: all forms of polarisation (i.e. along different lines) primarily manifest themselves on an annual basis. Only very few teachers observe polarisation between their pupils (along ethnic or other lines) on a weekly, monthly or even quarterly basis.

Figure 1. Presence of different forms of group polarisation in Dutch classrooms as perceived by teachers (in %).

Note: This figure presents answers to the following question: ‘Based on your own experiences during the past 12 months: to what extent does polarisation occur in your classroom along the following lines?’ (Ethnic background, Political preference, Religious conviction, Socio-economic background). For clarity’s sake, Figure 1 does not present ‘never’ and ‘don’t know’ answers. Total n = 1034, including 98 teachers who did not respond to this question, as they stated, previously in the survey, that they did not observe any form of polarisation in their classroom over the past 12 months.

When it comes to the manifestations of polarisation (i.e. along which lines), the answers given to the open-ended question: ‘between which groups of pupils (‘us’ versus ‘them’) do you experience that polarisation is strongest?’, presented in Figure 2, are similar to closed-ended question presented in Figure 1. In these open questions too, ethnic groups are mentioned, by far, most frequently. When looking more closely, these groups mainly concern ‘pupils with’ and ‘pupils without a migration background’, as the following answers illustrate: ‘immigration background and not’; ‘native Dutch students and students with a different ethnic background’; ‘NL vs Foreigners’; ‘young people (jongeren) with a migration background—without a migration background’ and ‘Dutch vs Non-Western Migrants’.

Figure 2. Categories of most strongly polarising groups in Dutch classrooms, as perceived by teachers (in %).

Note: Group categories are based on the answers to the following open question on polarisation in the classroom: ‘Between which groups of pupils (‘us’ versus ‘them’) do you experience that polarisation is strongest?’ N = 690.

Besides ethnicity, teachers also referred to other group categories, first of all along religious lines. Reflecting our theoretical expectations, the most frequently observed group opposition in that respect concerned ‘Islamic’ versus ‘non-Islamic pupils’. For instance, when teachers described polarisation between their pupils in terms of ‘Muslims and others’ or ‘Girls with hijab and without’. To be sure, this religious group distinction significantly overlaps with the previous category, based on ethnicity, since the two largest ethnic minority groups in the Netherlands (with a Turkish and Moroccan background) are overwhelmingly Islamic (Huijink, 2018). It should be emphasised, though, that Muslim pupils and religious pupils more broadly, are far from monolithic, while sometimes being internally opposed. For instance, several teachers observed polarisation between students within the same religious group, notably between ‘more strict believers’ and ‘less strict believers’.

The third most frequently mentioned group categories relate to the socio-economic background of pupils. Respondents thereby notably referred to ‘pupils with a high SES and pupils with a low SES’, or ‘rich’ versus ‘poor’ pupils. As shown in Figure 2, other categories were mentioned less frequently (<10%). These include political polarisation (mainly between ‘right-wing’ or ‘conservative’ pupils on the one hand and ‘leftist’ or ‘progressive’ pupils on the other); Covid-19 (notably between ‘vaccinated’ and ‘unvaccinated’ pupils); sexual diversity (e.g. ‘LHBTIQ + versus hetero’). And other even less frequently mentioned categories, such ‘Students from the city and from outside the city’ or ‘street culture and no street culture’. Finally, more than 10% of all invoked oppositions can be labelled as ‘composite polarisation’, as they concern tense relations between students along a combination of analytically different dimensions. In line with our expectations, these combinations mainly relate to ethnicity and politics. For instance, when respondents observe tensions between ‘pupils with a non-Western migration background and pupils that are politically right-wing’.

4.2. Which factors favour perceived ethnic polarisation in the classroom?

We now shift our attention to the factors favouring perceived ethnic group polarisation in Dutch classrooms. In our theoretical framework we tentatively discerned and discussed three potentially relevant factors: (1) the ethnic heterogeneity among pupils; (2) educational stages; and (3) tracking. In the following three subsections, we will discuss our empirical findings pertaining to these three factors, respectively.

4.3. Ethnic heterogeneity among pupils

When it comes to ethnic heterogeneity, our findings, presented in Figure 3, largely confirm our theoretical expectations, in that we see a clear curvilinear relationship when it comes to the presence of polarisation along ethnic lines and the relative saliency thereof.¹⁵ These results remain statistically significant when including control variables and when using a categorised continuous variable instead of restricted cubic splines (see Tables A2–A4 in Appendix B). As Figure 3 shows (Table A2–A4), it should be noted though, that the results concerning those respondents who teach in educational contexts with more than 50% of the pupils having a bicultural background are less univocal compared to the findings concerning those respondents who teach in educational contexts with less than 50% of the pupils having a bicultural background.

Figure 3. Adjusted predictions of proportion of bicultural pupils on perceived ethnic polarisation in the classroom, with 95% CIs.

Note: In both graphs, knots were placed at 10, 50, and 90% of the continuous explanatory variable. N = 916 for the graph on the left and 847 for the graph on the right.

Interestingly, the curvilinear pattern depicted in Figure 3 also showed up when looking at the answers to the open survey-question that was presented to respondents who reported that no polarisation between pupils took place in their classroom (‘Can you explain in a few sentences why there is no polarisation between pupils in your classroom?’, see Figure A1 in Appendix B for frequencies of the provided answers). In that respect, homogeneity among pupils turned out to be the factor these respondents invoked most frequently to explain why there is no polarisation in their classroom. For example, by stating that: ‘Almost all my students are from the same background. A single student is 'different’ but is completely accepted’, or that ‘There are simply too few students of different ethnicities there [in the classroom] to see anything of group formation among themselves or group formation against them developing’. At the same time, teachers referred to the strong heterogeneity of their pupils to account for a low degree of polarisation in the classroom. For instance, when stating that ‘there are so many different backgrounds that the differences are no longer noticeable’; ‘We have 70 different nationalities. Everyone is different.’ Or that ‘we have a lot of different cultures together at school, not per ce 1 or 2 big groups of the same culture.’

4.4. Tracking and educational stages

Besides ethnic composition, we also expected that tracking plays a role in favouring perceived ethnic polarisation in the classroom. As shown in Figure 4, this theoretical expectation is confirmed by our empirical analyses, as teachers in tracks preparing for higher education (Havo and Vwo) observe significantly less and less salient polarisation along ethnic lines compared to teachers in pre-vocational tracks. As shown in Tables A2 and A3, these findings remain statistically significant when using control variables.

Figure 4. Adjusted predictions of tracks on perceived ethnic polarisation in the classroom, with 95% CIs.

Note: N = 916 for the graph on the left and 847 for the graph on the right.

Finally, we expected polarisation along ethnic lines to be more frequent and more salient in lower secondary education compared to upper secondary education. As Figure 5 shows, this expectation, too, was confirmed by our regression analyses, albeit less clearly than the previous two factors. Teachers teaching mainly or only in upper secondary education are significantly less likely to observe ethnic polarisation in the classroom (see graph on the left), yet this relationship becomes statistically non-significant when using control variables (Table A2). Similarly, teachers teaching mainly or only in lower secondary education are significantly more likely to perceive ethnic polarisation in the classroom as the strongest form of polarisation between their pupils (see graph on the right). Yet, this relationship too becomes statistically non-significant when using control variables (Table A3).

Figure 5. Adjusted predictions of educational stage on perceived ethnic polarisation in the classroom, with 95% Cis.

Note: In both graphs, the categories ‘Lower’ and ‘Upper’ include respondents that teach ‘only’ and ‘mainly’ in lower or upper secondary education. The category ‘Lower and upper’ refers to respondents teaching equally in lower and upper secondary education. N = 916 for the graph on the left and 847 for the graph on the right.

5. Discussion

This paper aimed at assessing how (i.e. along which lines), how often and under which conditions group polarisation manifests itself in the context of Dutch secondary school classrooms. Given the very limited amount of existing literature on the topic and the study’s focus on one country, this paper unavoidably has an exploratory character. Nonetheless, based on descriptive statistics and binary logistic regression analyses, it yields at least five important insights. First, our findings indicate that group polarisation in Dutch classrooms is relatively rare and generally manifests itself only on an annual basis, irrespective of its expressions along ethnic, religious, socio-economic or political lines. This finding seems to contrast with sometimes quite alarming statements regarding polarisation in western politics in society (e.g. Klein, 2020). It should be noted though that teachers in pre-academic tracks are over-represented in our survey, which may reduce the reported levels of perceived polarisation.¹⁶ At the same time, this potential underrepresentation mainly relates to ethnic group polarisation. Polarisation between pupils along political lines, for instance, is actually observed more frequently by teachers teaching in tracks that prepare for higher education.¹⁷ It should also be emphasised that the relatively low frequency of polarisation that we observed, does not forcibly correspond to the intensity of polarisation as it is experienced by pupils and teachers. The latter can, in fact, be quite strong, even if manifestations of polarisation are rare (Wansink et al., 2019). Examining the intensity and impact of polarisation would be an interesting avenue for further research.

Second, we find that group polarisation in Dutch classrooms is most frequently perceived along ethnic lines, notably among pupils with and without a recent migration background, whereas socio-economic tensions were least pregnant. Besides well-attested social-psychological processes of ethnic in-group versus out-group dynamics (e.g. Baerveldt et al., 2004; Clark & Ayers, 1992; Knifsend & Juvonen, 2014; Rícan, 1996; Thijs & Verkuyten, 2014; Van Caudenberg et al., 2020) and the contemporary socio-political salience of issues containing a strong ethnic component (e.g. migration, Islam, nationalism, etc.) that have made their way into the classroom (e.g. Kleijwegt, 2016), this finding could also be related to our focus on group polarisation through the eyes of teachers, whereby tense relations between pupils along ethnic and religious lines may be more directly visible compared to other forms of polarisation (e.g. based on social class¹⁸). At the same time, the relative absence of polarisation along socio-economic lines may also have to do with the fact that Dutch classes, notably in urban areas, tend to be homogeneous in socio-economic terms (Boterman, 2018), thus hampering similar tensions. Moreover, given their age, students do not have a crystallised socio-economic position in society (yet), with corresponding interests and potentially conflicting political preferences (e.g. as CEO’s, business owners or workers). For a better understanding of these findings, though, we propose to complement the perspective of teachers with an empirical investigation of the views of pupils concerning polarisation in the classroom.

Third, when focusing on factors favouring the most frequently observed form of group polarisation in Dutch classrooms (i.e. along ethnic lines), we find a curvilinear relationship with respect to the proportion of students having a bicultural background. This empirical finding confirms our expectations based on conflict theory (e.g. Agirdag et al., 2011; Juvonen et al., 2006; Thijs & Verkuyten, 2014), and more specifically, Moody’s (2001) ideas on ‘us/them’ dynamics in the classroom. On a practical level, these tensions in highly diverse educational contexts point at the potential role of teachers, not to neglect polarisation, but to use it in view of so-called ‘learning democracy’ in which friction is part of the education of pupils (Lozano Parra et al., 2021). Empirically, we should note, though, that our findings concerning school contexts with a high proportion of pupils having a bicultural background are less univocal compared to school contexts with a low proportion of students with a similar background. While being congruent with observations in the super diversity literature, indicating that positive encounters in highly heterogeneous settings ‘do not preclude negative attitudes towards specific groups’ (Wessendorf, 2014, p. 170), we see two methodological reasons for this outcome. On the one hand, there are relatively few respondents teaching in highly bicultural contexts, which goes hand in hand with wider confidence intervals. On the other, we were unable to measure the degree of actual ethnic diversity in classrooms. After all, an educational context with a high percentage of pupils having a bicultural background can both be very homogeneous (i.e. all students sharing a similar bicultural background) and very heterogeneous (i.e. students having a wide variety of backgrounds). For future research, we therefore suggest measuring the cultural diversity of pupils at the classroom level (cf. Dincer, 2011; Esteban & Ray, 1994; Montalvo & Reynal-Querol, 2005; Thijs & Verkuyten, 2014).

Fourth, we observe that perceived ethnic polarisation is significantly more present and more salient in pre-vocational tracks, compared to tracks preparing for higher education. This finding is congruent with earlier studies indicating that pupils in pre-vocational tracks can be more vocal when sharing their viewpoints on controversial topics (Damhuis & Wansink, 2023; Vogels et al., 2021) and less ‘tolerant’, in the sense of putting up with something one objects to Thijs et al. (2021), compared to students in educational tracks preparing for higher education. We should note though, that this finding, in our view, does not forcibly reflect something negative. After all, polarisation can also have positive effects, as contrasts with out-groups can heighten individuals’ awareness of their in-group connections, thus enabling the formation of significant identities (Coser, 1964; Levine & Campbell, 1972; Tajfel & Turner, 1979). Future research could tease out more precisely why polarisation plays out differently in different tracks, and what the consequences of ethnic group polarisation are for different groups of pupils.

Fifth, we observed that ethnic group polarisation is more present and more salient in lower secondary education, compared to upper secondary education. This, we theorised, may have to do with the psychological development of students, and their increased capacity, over time, to think about conflict and other groups in a more nuanced way (Rekker et al., 2015). Our results, however, do not remain significant when using control variables. Future research could therefore focus more specifically on group polarisation in different educational stages.

On a more general level, it would be interesting to examine how group polarisation in educational contexts plays out beyond the Dutch borders, in order to assess the extent to which our findings can travel. For instance, to countries where schools are socio-economically and culturally less segregated. Similarly, it would be interesting to use longitudinal data, to study polarisation in the classroom not only as a state, but also as a process, and to find out under which circumstances tensions between social groups increase or decrease over time.

Finally, what do our results mean for teacher practices? First, we propose that teacher training focusing on group-based polarization in the classroom should be made context and track specific. There should be special attention for the pre-vocational tracks, as teachers experience more polarisation in these classes. There is an impressive literature on multicultural and intercultural education that could inform teachers how to create inclusive classrooms (e.g. Banks, 2008). Second, in the Netherlands polarisation is not as binary as in the United States, but much more cross-cutting (Harteveld, 2021). Teachers might use this cross-cutting polarisation to show students that they can polarise on some aspects but are more unified on other aspects. Such an approach might actually help creating social cohesion in the classroom. However, more research regarding similar pedagogies is needed.

Acknowledgments

First of all, we would like to express our gratitude to all the teachers and experts who participated in this study. Moreover, we would like to thank Bob Timmer for his research assistance as well as the participants in a research workshop at the Utrecht University School of Governance. Their helpful comments helped us to improve the paper. Institutional Review Board Statement: This study is approved by the Ethics Committee of the Faculty of Law, Economics & Governance, Utrecht University, the Netherlands.

Disclosure statement

The authors report there are no competing interests to declare.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

We were able to write this article thanks to the financial support of the strategic theme Dynamics of Youth and TerInfo, both at Utrecht University.

Notes on contributors

Koen Damhuis

Koen Damhuis is assistant professor at the Utrecht University School of Governance. His research interests include intergroup conflict in society and schools, political polarization, and the [appeal of] electoral radical right-wing parties.

Bjorn Wansink

Bjorn Wansink is associate professor at the Department of Education and Pedagogy at Utrecht University. His research interests center on civics and history education, multicultural education, critical thinking, sensitive topics and peace education. He is also a teacher trainer for the European Association of History Educators in (post)-conflict countries.

Notes

1 As Huijnk (2018, p. 28) showed, 85% of Dutch citizens with a Turkish migration background, aged 15–24, consider themselves to be Muslim. This applies to 94% of Dutch citizens with a Moroccan migration background, aged 15–24.

2 At its core, this line of thought was already captured in Voltaire’s famous dictum, Lewis Coser (1964, p. 170) also alluded to: ‘Were there but one religion in England, despotism would threaten; were there two, one would see bloody civil war; but in fact there are dozens, and they live happily in peace’.

3 Accordingly, besides perceiving cultural threats and ethnic differences, these students also referred to financial and economic threats, which, according to Van Praag et al. (2015) can be explained by the fact that pre-vocational tracks are directly geared toward the labour market and specific future jobs.

4 Due to the Covid pandemic, the interviews were held online from mid-March 2020 onward. This did not significantly impact the quality of the collected data, as the teachers were themselves highly experienced users of online platforms, due to the mandatory character of online teaching in the Netherlands during this period.

5 Contrasting with subjects like chemistry or French, where polarization is less likely to manifest itself.

6 Unfortunately there are no data regarding the number of Dutch teachers qua subject. Based on data from the Dutch Ministry of Education, Culture and Science (2022), we do know that, at the time the survey was fielded, 76 thousand teachers were working in Dutch secondary education. Given our focus on teachers in five subjects—that cover roughly 30% of the amount of total hours within an averaged Dutch secondary education curriculum—, we can estimate that the population consists of approximately 22 thousands teachers. Accordingly, our sample covers about 4.7% of the entire population.

7 98 teachers did not respond to this question, as they stated (earlier in the survey) that they did not observe any form of polarisation in their classroom over the past 12 months.

8 During the first months of this period (14 December 2020-1 March 2021), teaching took place online due to the Covid pandemic. This may have slightly reduced the occurrence of observed polarization, yet rather limitedly

9 We are aware of the fact that similar (ethnic) group categorisations are not ‘neutral’. Instead, they should rather be considered as ways of making sense of the social world, i.e. ‘of recognizing, identifying, and classifying other people, of construing sameness and difference, and of ‘coding’ and making sense of their actions.’ (Brubaker, 2009, p. 33) In other words: ‘They are templates for representing and organizing social knowledge, frames for articulating social comparisons and explanations, and filters that govern what is noticed or unnoticed.’ (Ibid. 33).

10 We had to ask this question on the school level, instead of the classroom level, since respondents generally teach in different classes.

11 To cover the increasing heterogeneity of migration backgrounds in the Netherlands (Jennissen et al., 2018), these four examples include both European and extra-European countries, as well as ‘classic’ countries of origin in Dutch migration history (Morocco and Turkey) and countries of origin related to recent migration (Poland and Syria).

12 The tracks Havo and Vwo were merged as both tracks prepare for higher education, and because the number of respondents teaching only in Havo or only in Vwo classes is very low.

13 For more robust analyses, we merged these categories, as relatively few respondents teach only in lower secondary education or only in upper secondary education.

14 Following Sijbers et al. (2015) the data, for all analyses presented below, was weighted in terms of school denomination and region in order to align the sample with the national distribution in the study population.

15 Please note that the Y-scales differ between the graph on the left (presence of ethnic polarisation in the classroom) and on the right (polarisation strongest among ethnic groups); since the mean of the former is higher (0.67) than the mean of the latter (0.30). This also applies to Figures 4 and 5.

16 As far as we know, there are unfortunately no data regarding the number of teachers in different tracks in the Netherlands.

17 See for more details: Figure A2 in the online appendix.

18 Class-based differences between pupils would be particularly obscured in educational contexts where school uniforms are obligatory. Yet, this is not the case in the Netherlands.