Prejudice and the Acceptance of Muslim Minority Practices: A Person-Centered Approach.

Growing Muslim minorities in Western societies has sparked debate about which Muslim practices should be accepted, with many people finding certain practices intolerable. Two competing perspectives on this intolerance argue that it represents either principled objections or prejudice. Using four large samples from the Netherlands, we apply latent profile analysis and find four groups of people: two groups that like and dislike Muslims and their practices respectively, but also two groups who are intolerant of some or most Muslim practices without necessarily displaying prejudice. A person-centered analysis of key demographic and psychological variables suggests that the two intolerant groups differ with one group's intolerance motivated more by anti-Muslim feelings, while the second group's intolerance is motivated more by principled objections.

Questions of the accommodation of specific Muslim practices and rights (e.g., headscarf, Mosques, Islamic education) within the limits of liberal societies are at the center of the polarized debate in Western Europe and other Western societies (e.g., Carol & Koopmans, 2013). Research has focused on understanding attitudes toward Muslim practices and rights in terms of group-based prejudice and Islamophobia (e.g., Kalkan, Layman, & Uslaner, 2009; Raiya, Pargament, Mahoney, & Trevino, 2008; Savelkoul, Scheepers, van der Veld, & Hagendoorn, 2012). Support for banning the headscarf, for example, reflects anti-Muslim feelings (Helbling, 2014; Saroglou, Lamkaddem, Van Pachterbeke, & Buxant, 2009). However, some studies suggest that the relationship between prejudice toward Muslims and acceptance of their religious practices is not straightforward (e.g., Sniderman & Hagendoorn, 2007; Van der Noll, 2014; Van der Noll, Poppe, & Verkuyten, 2010). The level of acceptance or rejection of Muslim practices may depend on the nature of the specific practice in question, and individuals can be intolerant because of more principled objections to that particular practice rather than due to prejudicial feelings.

Considering general feelings toward Muslims as well as the acceptance of a range of Muslim practices allows us to examine the possibility that people reject particular Muslim practices and the related rights (e.g., Islamic primary education) while thinking well of Muslims as a group, and vice versa. Further, if we were to investigate only one specific practice we risk missing the attitudes of those who would like to ban that practice but accept other Muslim practices, or conversely those who might accept the practice but want to forbid all else. Therefore, and in going beyond previous research on anti-Muslim attitudes, our primary goal is to jointly examine majority members’ general feelings toward Muslims and their acceptance of a range of controversial but legal Muslim practices. Using four large datasets from the Netherlands and a person-centered approach we examined whether there are distinct groups of individuals within the majority population with different combinations of feelings and acceptance, and with different demographic and social psychological characteristics.

Anti-Muslim Reactions

Research in Europe (e.g., Spruyt & Elchardus, 2012; Strabac & Listhaug, 2008) and in the United States (e.g., Kalkan et al., 2009) indicates that anti-Muslim feelings are more wide-spread than negative feelings toward other immigrant groups. Kalkan and colleagues (2009) found that an empirical distinction can be made between people’s attitudes toward categories that are defined by racial, ethnic, and religious background and their attitudes toward cultural groups that are defined by dissenting practices and behaviors. Anti-Muslim feelings were found to be connected to both attitudes, and most strongly to the latter ones.

Using survey data, several studies try to examine the extent to which these anti-Muslim feelings reflect group-based prejudice and the extent to which these reflect specific forms of (religious) critique (Breton & Eady, 2015; Hagendoorn & Poppe, 2012; Imhoff & Recker, 2012; Kalkan et al., 2009; Sniderman & Hagendoorn, 2007). For instance, research has demonstrated that Enlightenment values (e.g., Gustavsson, Van der Noll, & Sundberg, 2016; Imhoff & Recker, 2012), secularism (Van Bohemen, Kemmers, & De Koster, 2011), and universalistic notions (Elchardus & Spruyt, 2014; Saroglou et al., 2009) predict anti-Muslim feelings over and above the statistical effect of generalized prejudice. These findings indicate that criticism of Islam cannot be reduced to anti-Muslim feelings.

Another strategy that can be used to understand anti-Muslim feelings is to distinguish between group-based prejudice and (in)tolerance of specific practices. Research in Western Europe has argued that people with more prejudicial attitudes toward Muslim minorities also more strongly oppose, for example, the building of Mosques, the wearing of headscarves, and Islamic schools. Thus, opposition to and rejection of these specific practices would reflect a general dislike of Muslims. In support of this perspective, several studies have found that higher prejudice is indeed associated with higher intolerance of Muslim practices and that the rejection of dissenting practices is used to justify anti-Muslim feelings (e.g., Helbling, 2014; Saroglou et al., 2009; Van der Noll, 2014; Van der Noll & Saroglou, 2015).

However, there is also research that supports an alternative perspective, indicating that (political) intolerance and prejudicial attitudes are distinct phenomena (Crawford & Pilanski, 2014; Hagendoorn & Poppe, 2012; Klein & Zick, 2013; Van der Noll, Poppe, & Verkuyten, 2010; Wirtz, van der Pligt, & Doosje, 2016). For example, in the context of Quebec, Canada, while those who hold prejudicial views supported a ban on religious symbols, a majority of the people supporting the ban did so out of principled secularism rather than prejudice (Breton & Eady, 2015). In another study in Quebec it was found that feelings of cultural threat and generalized prejudice predicted support for banning minority religious symbols whereas holding liberal values predicted support to ban all religious symbols (Bilodeau, Turgeon, White, & Henderson, 2018). Similarly, analyzing data from six European countries, Helbling (2014) found that Europeans with secular liberal values felt positively toward Muslims as a group, but felt torn regarding the legislation of religious practices such as the wearing of the headscarf. In addition, among national samples in the UK, France, Germany, and the Netherlands, a substantial portion of people with a positive attitude toward Muslims supported a ban on headscarves (Van der Noll, 2010; see also Saroglou et al., 2009) and also rejected Islamic education and the building of Mosques (Van der Noll, 2014).

Drawing on these two competing perspectives, an objection to a specific practice may represent (1) an expression of one’s negative attitude toward Muslims as a group, (2) an expression of disapproval of that particular practice, or (3) a combination of these two. For example, majority members can reject the founding of Islamic schools because they dislike Muslims, or because they believe that religion in general has no place in education, or a combination of the two. In other words, the distinction between group-based attitudes toward Muslim minorities and practice-specific disapproval is important and requires greater attention (Bilodeau et al., 2018; Dangubic, Verkuyten, & Stark, 2019; Hurwitz & Mondak, 2002). One way to address this is by considering a range of Muslim practices in combination with general feelings toward Muslims. This allows us to develop a more detailed understanding of particular combinations of group-based feelings of Muslim minorities and the acceptance of specific Muslim practices.

Intolerance and Multiple Practices

The combination of negativity toward a minority group with nevertheless accepting this minority group’s civil rights is central in the literature on political tolerance (Gibson, 2006; Sullivan, Piereson, & Marcus, 1982). However, negative out-group feelings are not a precondition for (political)intolerance because one can reject a specific practice (e.g., ritual slaughter of animals) of people or groups (Jews, Muslims) to whom one has neutral or even positive feelings (Hurwitz & Mondak, 2002; Sniderman, Tetlock, Glaser, Green, & Hout, 1989).

Considering various dissenting practices simultaneously makes it possible to take into account the distinction between majority members who consistently accept or rather reject various Muslim practices, and people who do not consistently accept or reject the different practices. Cross-practice consistency can indicate a general like or dislike of Muslim minorities and inconsistencies can indicate principle considerations or rather social normative concerns about specific practices (Dangubic et al., 2019; Sniderman et al., 1989). Specifically, some people may display positive feelings toward Muslims and are consistent in accepting the various practices (“liking”), while others have negative feelings and consistently reject all practices (“disliking”). The existence of these two groups is in agreement with the literature that links support for Muslim minority practices with group-based feelings. Additionally, however, there is the possibility that these positive or negative feelings go together with the acceptance of some Muslim practices but not of other practices. This practice-based inconsistency indicates that people do not only consider information about whom they are asked to tolerate but also on the nature of the dissenting practices (e.g., Gibson & Gouws, 2003; Petersen, Slothuus, Stubager, & Togeby, 2011). The consideration of multiple practices might demonstrate that people have objections to a particular practice (e.g., wearing of a headscarf) but not toward another practice (e.g., Islamic education). A particular practice might raise specific moral concerns, such as the wearing of a headscarf which might evoke the issue of gender equality and religious education in public schools that can evoke concerns about the secular nature of the state (Moss, Blodorn, Van Camp, & O’Brien, 2017; Sarrasin, 2016). However, the rejection of some Muslim practices might also result from social normative standards that make it socially acceptable to express one’s prejudiced feelings by rejecting these practices and not others (Crandall & Eshleman, 2003). Thus, some individuals may be practicing a form of principled intolerance, where their disapproval emerges from specific views held about specific practices, while other individuals may be practicing a prejudiced intolerance, where the disapproval of specific practices emerges from a prejudice toward the group that they may be unwilling to express. Using data from four representative samples of the Dutch majority and a person-centered approach, we will examine whether these four groups of individuals do indeed exist and how many people can be described as practicing, for example, a principled or prejudiced intolerance. Furthermore, as a matter of construct validity we consider whether some key demographic and social psychological constructs characterize the different groups.

Person-Centered Approach

In general, social psychological research typically investigates associations using a variable-centered approach in which the evaluation of an out-group and its practices is considered to have a common underlying dimension that ranges individuals from low to high prejudice (see Meeusen, Meuleman, Abts, & Bergh, 2018). This approach ignores the possibility that individuals are not equally negative toward distinct sets of practices and therefore that a single score does not accurately reflect the stances that majority members take. Further, it is possible that people combine their feelings and objections in different ways leading to groups of individuals with distinct constellations of ratings. In other words, a variable-centered approach ignores the complex constellation of characteristics that make up individuals and precludes the possibility that there are profiles of people which would provide a nuanced and more detailed understanding of how majority members perceive and evaluate Muslims as a group and the various practices they are engaged in.

Taking a person-centered approach makes it possible to consider these combinations of out-group feelings and the evaluation of different out-group practices simultaneously (e.g., Bergman & Magnusson, 1997; Meeusen et al., 2018; Muthén & Muthén, 2000). This type of analysis seeks to identify unobserved groups, or categories, of individuals who differ in the particular ways in which they combine their out-group feelings and objections to a range of out-group practices. For example, in a research on political tolerance, the best model for the data required four categories of individuals rather than a continuum of tolerance (McCutcheon, 1985). In addition to a group of individuals who were consistently positive toward various minority groups and three different practices and a group of people who were consistently negative, two other categories of people were found (see also Herson & Hofstetter, 1975; McIntosh, Mac Iver, Abele, & Nolle, 1995; Merton, 1976; Sniderman et al., 1989). Individuals in these two categories accepted some groups and some practices but rejected others. These four categories could not be readily placed on a unidimensional positive-negative continuum because there was no monotonic change across the four groups of individuals. Rather, they formed four latent classes of political tolerance.

Thus, a person-centered approach allows us to investigate whether feelings toward Muslims as a group of people and objections to various Muslim practices are combined in different ways by different groups of individuals. This makes it possible to examine whether, and how many, majority members have, for example, a liking-based, a disliking-based, or a principled or prejudiced intolerant profile concerning their attitudes toward Muslims and Muslim practices. One group of individuals might reject almost all Muslim practices while another group of individuals might reject the building of Mosques but accept the establishment of Islamic schools. Thus, rather than a unidimensional continuum, by considering different dissenting practices simultaneously it is possible to assess potentially different patterns among different groups of individuals that may speak to the underlying attitudes that affect which practices they accept. While it might be reasonable for someone to have a fair objection to a single specific practice, it seems unlikely that they would have such objections to all outgroup practices. Rather someone who is generally prejudiced toward Muslims as a group will be less likely to differentiate based on the specifics of individual practices but will tend to be broadly intolerant of a range of Muslim practices.

Profiles and Their Correlates

Beyond identifying groups of individuals based on their acceptance of outgroup practices and general out-group feelings, it is important to examine whether the different groups of individuals differ for some key characteristics that are typically considered in research on minority group prejudice. This is important because both principled and prejudiced intolerance involve practice-inconsistent rejection and therefore are largely indistinguishable in terms of observed responses. However, several factors should strengthen or weaken a tendency to principled tolerance and these can be examined as a matter of construct validity. We will look into the role of educational background and political orientation as two main demographic predictors of prejudicial attitudes and of important social psychological constructs.

Research on the effect of education on prejudicial attitudes suggests a complex relationship. The ideological refinement perspective (Jackman & Muha, 1984) views education as endowing majority members with more advanced cognitive skills and ideological commitments to support abstract ideas of equality and justice, while simultaneously using their cognitive skills to protect the status quo by rejecting social policies designed to overcome group-based inequalities. Research also suggests that the association between higher education and a more positive attitude toward minority groups is not explained by a greater tendency of the higher educated to respond in a socially desirable way (Heerwig & McCabe, 2009; Ostapczuk, Musch, & Moshagen, 2009; Wagner & Zick, 1995). This does not mean that the higher educated have less spontaneous negative reactions toward ethnic minorities. Research on aversive racism (Dovidio & Gaertner, 2004) has shown that aversive racists show relatively strong prejudice on implicit but not on explicit measures (Son Hing, Chung-Yan, Hamilton, & Zanna, 2008), and the higher educated have been found to have lower explicit but not implicit prejudices (Kuppens & Spears, 2014). However, the association between higher education and a more positive self-reported attitude toward ethnic and cultural minority groups is one of the most replicated findings in the social sciences (Jenssen & Engesbak, 1994) and has been predominantly explained in terms of cognitive development and the learning of liberal values. Education is associated with cognitive ability and flexibility (Bobo & Licari, 1989; Ohlander, Batalova, & Treas, 2005) and is a strong correlate of political sophistication (Bennett, 1996; Delli Carpini & Keeter, 1996; Highton, 2009; Hillygus, 2005). When the educational level in the population increases, the ideological sophistication also increases (Tedin, 1987). Education implies political socialization which involves a better understanding of the values and beliefs that underlie political-ideological differences (Osborne & Sibley, 2012; Stenner, 2005) and making the higher educated better able to understand the importance of basic norms and values of equality and tolerance underlying the democratic culture (Vogt, 1997). The higher educated are not only more likely to be tolerant in general but also as a matter of principle (Sniderman et al., 1989). Thus, the literature leads us to expect that the “general liking” and “principled intolerant” groups are more highly educated than the “prejudiced intolerant” and “generally disliking” groups. In addition to education, we also considered a direct measure of cognitive sophistication as the tendency to consider how issues that one feels strong about can have multiple perspectives. Based on our reasoning for the role of education we can expect that the “general liking” and “principled intolerant” will display higher cognitive sophistication than the prejudiced intolerant and the disliking groups.

Social psychological research on the social cognition model has argued and demonstrated that two core aspects capture the most important differences between the political right and the left (Jost, Glaser, Kruglanski, & Sulloway, 2003). Political orientation is manifested in a specific ideological configuration in which respect for tradition and acceptance of inequality are central (Jost, 2006, 2017). The first dimension concerns attitudes toward cultural tradition and social deviance, while the second relates to (in)equality and egalitarianism. Individuals on the political right tend to prefer traditions and social conformity, and to accept inequalities. In contrast, those on the left more strongly embrace socio-cultural change and equality. Extensive empirical research in political psychology and in different Western countries (Piurko, Schwartz, & Davidov, 2011) confirms that these two core dimensions capture the most important ideological differences between right-wing and left-wing political orientations (Jost 2006, 2017; Schwartz, Caprara, & Vecchione, 2010). This means that we expected that the Muslim liking and principled intolerant groups have a stronger left-wing political orientation compared to the Muslim disliking and prejudiced intolerant groups.

We further investigated Social Dominance Orientation (SDO), Right Wing Authoritarianism (RWA), and perceived out-group threat as three key social psychological constructs that have been extensively linked to tolerance as well as prejudicial attitudes toward minority groups. Examining these constructs allows us to assess whether the expected groups of individuals do not only differ in their attitude toward Muslims and Muslim practices but also in these important correlates. Specifically, the “generally like” group and the “principled intolerance” group should have lower SDO, lower RWA, and lower perceived threat compared to the “prejudiced intolerance” and “generally dislike” groups. Furthermore, we also sought to determine whether the groups of individuals differ in their internal motivation to control and not express prejudices. This measure can give a further understanding of the underlying difference between principled and prejudiced intolerant individuals. Specifically, we can expect that the “general liking” and “principled intolerant” will display a stronger internal motivation to control prejudice than the prejudiced intolerant and the disliking groups.

Finally, we tried to extend beyond attitudes and ideologies to behavioral intentions. Specifically, we considered the willingness to engage in anti-discrimination activities. We selected this variable because someone who is intolerant of specific practices for more principled reasons should not accept discriminatory treatment of minority members or members of other minority groups. Therefore, support for anti-discrimination activities is expected to serve as a useful identifier for making a distinction between the principled and prejudiced intolerant groups of individuals.

Summing Up

In this article, we drew on data from four large representative datasets of majority Dutch participants collected in the Netherlands between 2014 and 2018 which allows us to investigate how well the data aligns over time and across sampling error. We examine feeling thermometer ratings toward the two most prominent and typical groups of Muslims in the Netherlands (of Turkish and Moroccan background), together with the acceptance of five different Muslim practices that are strongly debated in society (public expression of Muslim religion, wearing of the headscarf, celebration of Islamic holidays, building of Mosques, founding of Islamic schools). Our first aim is to investigate whether there are groups of individuals who are characterized by particular combinations of general feelings and acceptance of specific practices. More specifically, we expect to identify the four types of dislike-based (prejudiced feelings and rejection of all practices), like-based (non-prejudiced feelings and acceptance of all practices), principled intolerant (non-prejudiced feelings and differential rejection of practices), and prejudiced intolerant (non-prejudiced feelings and general rejection of practices).

Second, we expect that the groups of individuals differ in terms of several well-known correlates of prejudice toward minority groups. Based on existing variable-centered research we expect that the groups differ in level of education, political orientation, RWA, SDO, perceived out-group threat, the internal motivation to control prejudice, cognitive sophistication, and the tendency to be involved in anti-discrimination actions. Specifically, we expected that the liking group and the principled intolerance group differ on these correlates from the disliking group and the prejudiced intolerance group.

Method

Participants

For this paper, we analyzed data of four large representative datasets that contain various measures and that have been used for other purposes in previous research (e.g., Mepham & Verkuyten, 2017; Verkuyten, Martinovic, Smeeks, & Kros, 2016). In all four studies the data were collected online among probability samples drawn from nationally representative pools of the majority Dutch population. The response rate of the different studies was around 55% which is similar to other research in the Netherlands (Stoop, 2005). The samples covered various segments of the Dutch public in terms of age, gender, education, household size, and the region of residence. The samples were selected by research consultancy companies which maintain a database of majority Dutch people who regularly participate in surveys in return for remuneration.

In Study 1, 469 majority Dutch participants in the 2014 dataset (Study 1) completed all seven key measures and were retained for analysis. Participants were identified as majority Dutch based on self-identification and if both of their parents were born in the Netherlands. Participants who identified as Muslim were excluded in the analysis (two participants each in Studies 2 and 3). Similarly, 800 participants in the 2015 dataset (Study 2), 590 in the 2017 dataset (Study 3), and 563 participants in the 2018 dataset (Study 4) completed all key measures and were retained for analysis. The participants were split relatively evenly along gender lines (2014: 52.2% male; 2015: 50.0% male; 2017: 54.56% male: 2018: 50.3% female), and from a wide range of ages (2014: M = 50.26, SD = 16.98, range = 18–88; 2015: M = 50.65, SD = 17.16, range = 18–87; 2017: M = 55.46, SD = 14.60, range = 18–87;2018: M = 51.08, SD = 17.52, range = 18–91), education levels and political orientation (descriptive statistics reported below).

Materials

The materials used are discussed below and the English (translated) versions of all items can be found in the Electronic Supplementary Material (ESM 1). Complete data and analytic scripts can be found on the Open Science Framework at https://osf.io/7j2zm/.

Out-Group Feelings

Respondents were presented with feeling thermometers to indicate their coldness or warmth to members of the two main Muslim minority groups in the Netherlands: Turks and Moroccans. Both groups are over 95% Muslim, and are widely recognized as the two prototypical Muslim groups in the Netherlands. Using feeling thermometers with wider ranges of responses than Likert-type scales generates a more reliable measure (Alwin, 1997), and these explicit measures tend to correlate with subtler methods of assessing prejudice (Dovidio, Kawakami, & Beach, 2001). Overall and reflecting the ethnic hierarchy in the Netherlands (Schalk-Soekar, van de Vijver, & Hoogsteder, 2004), feelings toward Turks hovered around the midpoint of the 1–11 scale, neither warm nor cold (MStudy1 = 6.05, SDStudy1 = 2.29; MStudy2 = 6.20, SDStudy2 = 2.20; MStudy3 = 6.45, SDStudy3 = 2.27; MStudy4 = 5.13, SDStudy4 = 2.21), while feelings toward Moroccans were consistently lower (MStudy1 = 4.36, SDStudy1 = 2.23; MStudy2 = 4.28, SDStudy2 = 2.12; MStudy3 = 4.66, SDStudy3 = 2.47; MStudy4 = 4.54, SDStudy4 = 2.24). More negative attitudes toward Turks in 2018 (Study 4) may be due to a 2017 political and diplomatic crisis between Turkey and the Netherlands. Across all three studies, the two measures were positively correlated (r ranging from .49 to .61) and in a recent study among a Dutch representative sample, the feelings toward Turks and Moroccans correlate strongly with the feeling toward Muslims as a category (.73 and .77, respectively).

Acceptance of Muslim Practices

The specific practices presented to the respondents were partially adapted from previous research (e.g., Gieling, Thijs, & Verkuyten, 2010; Smeekes, Verkuyten, & Poppe, 2011). These relate to different types of civil liberties that, however, are subject to much debate in Dutch society. Using 7-point scales, one item asked participants’ agreement that Muslims can express their faith in public (“Muslims in the Netherlands must be able to show and experience their own faith in public life”; MStudy1 = 4.31, SDStudy1 = 1.60; MStudy2 = 4.44, SDStudy2 = 1.53; MStudy3 = 4.16, SDStudy3 = 1.77; MStudy4 = 4.72, SDStudy4 = 1.54), the second referred to Muslim women’s ability to wear the headscarf (“Muslim women should have the opportunity to wear a headscarf anywhere in the Netherlands”; MStudy1 = 3.34, SDStudy1 = 1.77; MStudy2 = 3.78, SDStudy2 = 1.80; MStudy3 = 3.63, SDStudy3 = 1.92; MStudy4 = 3.87, SDStudy4 = 1.77), the third asked about Muslim’s rights to celebrate their festivals in public (“Muslims in the Netherlands should not only be able to celebrate their Islamic holidays at home, but also in public”; MStudy1 = 4.04, SDStudy1 = 1.64; MStudy2 = 4.13, SDStudy2 = 1.60; MStudy3 = 3.99, SDStudy3 = 1.82; MStudy4 = 4.46, SDStudy4 = 1.62), the fourth about the right to build mosques (“Muslims must have the right to build mosques in the Netherlands”; MStudy1 = 3.72, SDStudy1 = 1.76; MStudy2 = 3.94, SDStudy2 = 1.71; MStudy3 = 3.91, SDStudy3 = 1.87; MStudy4 = 4.29, SDStudy4 = 1.72), and the fifth about the right to establish Islamic schools (“Muslims must have the right to establish Islamic schools”; MStudy1 = 3.09, SDStudy1 = 1.65; MStudy2 = 3.08, SDStudy2 = 1.65; MStudy3 = 2.84SDStudy3 = 1.71; MStudy4 = 3.35, SDStudy4 = 1.75). Thus, these items range from the less objectionable (showing and experiencing faith in public) to the more objectionable (establishing Islamic schools that may perceived to prevent social and cultural integration).

Predictor Variables

Educational background was measured using a single-item in which participants indicated their highest educational achievement on a scale ranging from 1 (= no higher education) to 8 (= doctorate or advanced masters) in Studies 1 and 3, and 1 (= no higher education) to 7 (= doctorate or advanced masters)in Study 2 and 4 (Study 1: 23.4% low, 48.9% middle, 27.7% high, Mstudy1 = 5.14, SDstudy1 = 1.68; Study 2: 25.4% low, 49.7% middle, 25.0% high, Mstudy2 = 5.03, SDstudy2 = 1.68; Study 3: 23.5% low, 45.1% middle, 31.4% high,Mstudy3 = 4.24, SDstudy3 = 1.75; Study 4: 17.1% low, 47.4% middle, 35.5% high, Mstudy4 = 4.45, SDstudy4 = 1.72). The distinction between these levels of achieved education is comparable to the International Standard Classification of Education (ISCED)-measure that is used, for example, in the European Social Survey. Similar to other research in the Netherlands (e.g., De Graaf, De Graaf, & Kraaykamp, 2000; Van de Werfhorst & Van Tubergen, 2007), education was treated in the analysis as a continuous variable which allows us to investigate the difference between lower and higher educated participants.

Political orientation was measured with the well-known self-placement question (Jost, 2006). A 5-point scale was used ranging from politically left, center-left, center, center-right to right. In all four samples, the overall mean for political ideology fell in the center (Mstudy1 = 3.00, SDstudy1 = 1.09; Mstudy2 = 2.92, SDstudy2 = 1.05; Mstudy3 = 3.02, SDstudy3 = 1.18; Mstudy4 = 2.93, SDstudy4 = 1.27) with an equal distribution to the political right and to the political left.

A measure of social dominance orientation was available in two datasets. Eight (in Study 2; M = 3.24, SD = 0.86; α = .76), and six (in Study 3; M = 3.20, SD = 1.02; α = .72) items of a short version of SDO were used that was validated and translated to Dutch by Duriez and Van Hiel (2002) and that has been used in other research (Duriez, Soenens, & Vansteenkiste, 2007; Meeus, Duriez, Vanbeselaere, Phalet, & Kuppens, 2009).

Right-wing authoritarianism (RWA) was measured using a short measure previously utilized in the Netherlands (Sniderman & Hagendoorn, 2007) and that focuses on the conformity aspect of RWA. This measure was used in Study 2 (3 items, M = 4.61, SD = 1.12, α = .70) and Study 3 (4 items; M = 5.09, SD = 0.98; α = .70).

Perceived out-group threat was measured in three datasets in a reliable but not identical way. In Studies 1 and 2, symbolic threat was measured using four and three items, respectively, that asked about Muslims in the Netherlands undermining the Dutch identity and way of life (2014: M = 4.00, SD = 1.49, α = .90; 2015: M = 3.99, SD = 1.64, α = .96). In Study 3, in light of the refugee crisis, the four threat items were expanded to refer to refugees who are from predominantly Muslim nations rather than Muslims as a group (M = 3.97, SD = 1.89, α = .97).

Study 3 also included a measure of internal motivation to control prejudice (Plant & Devine, 1998), with the goal of better understanding the motivations of people who might not explicitly declare prejudice but nonetheless be unaccepting of Muslim practices. The scale was comprised of four items measured on a 1–7 scale (M = 5.01, SD = 1.09, α = .97).

Cognitive sophistication was also measured in Study 3 using a 4-item scale (M = 5.18, SD = 0.99, α = .83) that focused on whether and how frequently participants sought to understand alternative perspectives on issues that they felt strongly about.

Behavior intentions. Following previous research in the Netherlands (Verkuyten, 2017), in Study 3 respondents were asked how likely it is that they would engage in a set of actions in response to discrimination against immigrants in the Netherlands (M = 2.34, SD = 0.95, α = .85).

Analyses

For the analysis we used a latent profile analysis approach (e.g., Oberski, 2016) which we conducted using R software. Latent profile analysis is analogous to latent class analysis (LCA; Muthén & Muthén, 2000) and identifies patterns across a set of continuous variables that can be identified as different profiles of individuals. We included the two feeling thermometers measuring attitudes toward Muslims (Turks and Moroccans) and the five items about Muslim practices into latent profile analysis to determine how general feelings and acceptance of specific practices combine to best identify different profiles of individuals in Dutch society. We conducted this same analysis across four separate datasets to triangulate in on a set of profiles that best represent Dutch society.

The LPA analyses performed identify the best models of profiles using different sources of empirical information about the appropriate number of profiles. First, the Bayesian information criteria (BIC) and Akaike information criteria (AIC) indicate how well a model with the selected number of profiles fits the data, with the lowest numbers indicating the best fit. Secondly, the Bootstrappped Likelihood Ratio Test (BLRT) conducts significance tests comparing the selected model with a model including one fewer profile (comparison between k and k − 1 profiles). Lastly, entropy scores indicate how uniquely the datapoints belong to one profile and not others. Low entropy scores indicate that datapoints (participants) could be classed into more than one profile, whereas high entropy scores indicate that participants are uniquely classed into one profile and not others. A final component of determining profiles is based on interpretability. One important nuance of interpretability is the need not to overfit the data by generating a model with profiles that represent specific variances unique to the dataset rather than generalizable profiles of participants. The suggested ways to avoid this is to ignore small classes and to inspect the profiles that emerge for interpretability. In our analysis, we used all of these approaches to arrive at the best profile fit for the data.

Results

Latent Profiles

Table 1 shows that across four independent datasets collected over a 4-year period, the combination of respondents’ general feelings toward Muslims minorities and their support for specific Muslim practices reveal four groups of individuals. Investigation of the best type of model across the four datasets revealed that a model that allowed for varying means while holding variance and covariance equal best fit the data. As can be seen in Table 1, while the BIC criteria in Studies 2–4 indicate that six profiles provide a slightly better fit, in both cases the improvement over the previous iteration is relatively small, and the criterion of theoretical interpretability suggests a four profile solution across the four datasets. Specifically, when we investigated profiles generated by the six-profile solution, the resultant profiles were inconsistent across the studies, generating profiles of less than 50 people (n ranging from 31 to 45 in Studies 2 and 4), and somewhat larger, though still inconsistent profiles in Study 3 (n = 61–78). In all studies, the six-profile solution returned groups consistent with the Like, Dislike, and Prejudiced Intolerant groups (see below), and returned fractured sections of what we term the Principled Intolerant group, usually in which some participants displayed greater or lesser concern about individual practices like wearing the headscarf or opening Islamic schools. These differences, while likely reflecting a range of concerns among that group, did not consistently return distinct profiles across the datasets, so we elected to use the more interpretable four profile solution. Thus, while the six profile solution performed better statistically, there was no consistent pattern across the six-profile solutions that improved interpretability above the four-profile model.

Table 1

Model fit indices across three studies

Study #	Profile #	BIC	AIC	BLRT	Entropy
Note. Bootstrapped Likelihood Ratio Test (BLRT) tests whether each number of profiles represents an improvement over identifying one fewer profile. BIC = Bayesian Information Criteria; AIC = Akaike Information Criteria. Bolded rows indicate the four-profile solution that was indicated across all four datasets.
Study 1 (2014)	3	11,644.66	11,432.98	65.56, p < .001	0.93
	4	11,563.52	11,318.633	130.35, p < .001	0.92
	5	11,697.80	11,419.71	−85.07, p = .999	0.87
	6	11,650.53	11,339.23	NA	0.87
Study 2 (2015)	3	19,845.43	19,606.42	57.85, p < .001	0.88
	4	19,689.68	19,413.29	209.23, p < .001	0.86
	5	19,698.98	19,385.11	44.18, p < .001	0.84
	6	19,663.19	19,311.85	89.26, p < .001	0.89
Study 3 (2017)	3	14,842.74	14,619.35	142.92, p < .001	0.89
	4	14,719.83	14,461.41	173.95, p < .001	0.88
	5	14,712.99	14,419.52	57.89, p = .001	0.90
	6	14,712.11	14,383.61	51.91, p = .002	0.86
Study 4 (2018)	3	13,773.84	13,552.84	63.82, p < .001	0.90
	4	13,615.86	13,360.20	208.64, p < .001	0.92
	5	13,625.82	13,335.49	40.71, p < .001	0.91
	6	13,628.64	13,303.65	47.84, p < .001	0.90

Figure 1 (panels A–D) presents the mean levels of Muslim group feelings and acceptance of the specific practices across the four studies for the four profiles identified. Note that for ease of interpretation these scores were subtracted from the neutral midpoints of the scales of the items used. Thus, positive scores indicate attitudes that are higher than the midpoint and negative scores indicate attitudes below the midpoint. This allows us to differentiate between relative differences in attitudes that indicate negativity as compared to neutrality.

Figure 1

Panels A–D: Mean scores for the attitude and tolerance of practices variables for the four profiles in each study.

A first group consists of individuals with relatively positive feelings toward the two Muslim minority groups combined with relatively high levels of acceptance for all the Muslim practices (“liking”; between 27.8% and 36.5% across the datasets).The second (“disliking”) consists of individuals with negative feelings toward Muslim minorities combined with a tendency to reject the different Muslim practices (between 13.8% and 22.0%). In addition, the latent profile analysis indicates that almost half of the participants do not appear to be less prejudiced against Muslim minority groups than the disliking group but nonetheless are unwilling to tolerate Muslim practices. This category of individuals emerges as two profiles: one in which all or almost all Muslim practices are rejected without apparent distinctions made between them, and another in which some practices are rejected but not others. The first of these subgroups we label as “prejudiced intolerance” (between 28.5% and 34.5%). While they do not appear to have particularly negative feelings toward Muslims, displaying neutral to slightly positive attitudes toward the Turkish minority group for example, their rejection of all Muslim practices (most evident in Studies 2 and 3) without differentiating much between them suggests that this rejection may be driven by a general dislike of Muslims. In contrast, members of the second of these subgroups show substantial differences in attitudes toward distinct practices. For example, in Study 1 they are opposed toward the headscarf, and across all of the studies they are especially opposed to Islamic schools. At the same time, their support for religious freedom (i.e., building Mosques) and public expression is usually very similar to that of the “liking” group. Thus, this fourth group is distinguished both from those that are generally positive across the board and those that are neutral or negative across the board, and we label this group “principled intolerant” (between 17.6% and 29.2%). Their intolerance of some Muslim practices but not others suggests a rejection based on specific objections rather than a generalized dislike. Thus, while responses on some of the less controversial practices are broadly similar in pattern to those of other groups, it is precisely the differences in one or two practices that differentiate the groups. Across all four studies, the group we identify as principled intolerant deviates from the pattern found among the other three groups, specifically by displaying positivity or neutrality to most practices, but a strong and consistent objection to specific other practices. Thus, these findings support our expectation of the existence of four specific profiles, with two subgroups of intolerance without prejudice emerging: one which appears to disapprove of all practices and thus seems to more closely fit the description of prejudiced intolerance, and another which is intolerant of some practices but not others and thus appears to resemble a more principled intolerance.

Predictors of the Groups of Individuals

Further evidence for the construct validity of these groups is provided by investigation of differences in important correlates. Therefore, we next looked at whether the four groups of people identified across these datasets differ on key demographic, social psychological, and behavioral variables. To do this, we generated multinomial logistic regression models using SAS software to predict membership in these categories. We created the model using three steps. In the first step, we looked at education and, the related concept of cognitive sophistication, as well as political orientation as key characteristics. In the second step, we included SDO, RWA, and internal motivation to control prejudice to see whether and how ideological world view differences predicted membership in these four profiles. Then we added in out-group threat and the behavioral measure of willingness to act regarding discrimination against immigrants in the Netherlands, to see how those added to our understanding of how the types of people who fall into these profiles differ. We used the three-step model since we expect there to be a substantial overlap between demographic and world view predictors and with threat and behavioral predictors. By conducting the analysis in three steps we are able to see the effects of demographic variables independent of the other predictors, as well as demographic and world view predictors independent of threat and behavioral predictors. In the third step, we can also see which predictors play a meaningful unique role when included with all other predictors. For the analyses and following our predictions, the principled intolerant group was set as the referent. All variables included in these analyses were rescaled from 0 to 1 which means that the log odds beta coefficients and the odds ratios indicate the relative change in likelihood of belonging to each of the groups as a function of a full scale increase in a given predictor.

Table 2 shows how, while there is variation across the four datasets, there also appears to be a distinction between unique predictors for the four profiles. Overall the principled intolerant group appears to be distinct from the prejudiced intolerant and disliking groups, and in some cases appears to be slightly more similar to the liking profile than the prejudiced intolerant profile. First, across the four studies and as expected, people in the principled intolerance group tended to be among the best educated: the disliking and prejudiced intolerant groups are less well educated than the principled intolerant group and the liking group. Second, consistent with the education variable, the principled intolerant group was more likely to engage in cognitively sophisticated thought than the prejudiced intolerant or disliking groups. Third, members of the principled intolerant group are politically more left-leaning than people in the prejudiced intolerance and disliking groups. Fourth, the principled intolerant group has consistently RWA and had lower SDO tendencies than the disliking and prejudiced intolerant groups, with mixed differences from the liking group. Lastly, the principled intolerant group has marginally less motivation to control prejudice than the liking group and (marginally) more than the prejudiced intolerance group and the disliking group.

Table 2

Multinomial logistic regression results of a demographics-only and demographics and personality predictors models with the principled intolerant group as the referent

	Liking		Prejudiced intolerant		Disliking
Reference = principled intolerant	Log odds β (SE)	Odds ratios	Log odds β (SE)	Odds ratios	Log odds β (SE)	Odds ratios
Notes. All variables were feature scaled (0–1) before inclusion for ease of comparison. Coefficients and odds ratios therefore represent changes due to movement from the lowest to the highest value on the scale. Coefficient betas represent log odds with standard errors in parentheses. Odds ratios represent odds of category belonging compared to the referent group. Only Studies 2 and 3 were included in the second step, as Studies 1 and 4 did not include any ideological predictors. SDO = Social Dominance Orientation; RWA = Right Wing Authoritarianism; SE = Standard Error. †p < .09–.05; *p < .05; **p < .01; ***p < .001.
Panel A: Demographic predictors
Education
Study 1	−0.19 (0.50)	0.83	−0.69 (0.49)	0.50	−1.99*** (0.54)	0.14
Study 2	1.20** (0.38)	3.31	−1.26** (0.40)	0.29	−1.58*** (0.48)	0.21
Study 3	0.39 (0.38)	1.48	−1.87*** (0.42)	0.15	−1.86*** (0.52)	0.16
Study 4	0.64 (0.50)	1.89	−0.80† (0.48)	0.45	−2.16*** (0.59)	0.12
Cognitive sophistication
Study 3	1.54† (0.83)	4.67	−2.08** (0.79)	0.13	−4.79* (0.91)	0.01
Political orientation
Study 1	−1.57** (0.51)	0.21	−0.44 (0.50)	0.65	1.13* (0.56)	3.10
Study 2	−0.27 (0.39)	0.76	1.30** (0.41)	3.68	0.83† (0.50)	2.30
Study 3	0.44 (0.41)	1.55	2.56*** (0.46)	12.90	2.66*** (0.57)	14.36
Study 4	−0.84† (0.45)	0.43	1.09* (0.44)	2.97	2.94*** (0.57)	18.97
Panel B: Demographic and personality predictors
Education
Study 2	1.14** (0.39)	3.13	−0.89* (0.41)	0.41	−0.83 (0.51)	0.44
Study 3	0.23 (0.41)	1.26	−1.40** (0.45)	0.25	−1.02† (0.58)	0.36
Cognitive sophistication
Study 3	1.21 (0.93)	3.35	−1.30 (0.93)	0.27	2.02† (1.10)	0.13
Political orientation
Study 2	−0.22 (0.43)	0.80	0.71 (0.45)	2.04	0.14 (0.54)	0.87
Study 3	0.57 (0.43)	1.77	2.33*** (0.47)	10.30	1.81** (0.61)	6.08
SDO
Study 2	0.92 (0.76)	2.52	2.88** (0.80)	17.78	4.61*** (1.01)	100.20
Study 3	0.82 (0.77)	2.28	1.18 (0.83)	3.25	2.96** (1.09)	19.22
RWA
Study 2	−0.82 (0.62)	0.44	1.01 (0.67)	2.75	3.18*** (0.87)	24.07
Study 3	−1.18† (0.70)	0.31	1.98* (0.80)	7.22	3.81*** (1.05)	45.11
Internal motivation to control prejudice
Study 3	1.21 (0.91)	3.37	−0.71 (0.88)	0.49	−3.19*** (1.06)	0.04

Table 3 shows how the addition of threat and behavioral predictors adds to our understanding of the different groups. Individuals in the principled intolerant group tend to perceive somewhat less threat from Muslims than the prejudiced intolerant group, further differentiating them, although those in the principled intolerant group nonetheless perceived greater threat than those in the liking group. Importantly, the distinctions between the two intolerant groups also emerged for behavioral intentions. A measure of willingness to engage in anti-discrimination activities showed that while people in the liking group were equally likely to engage in anti-discrimination activities as those in the principled intolerant group, those in the prejudiced intolerant and disliking groups were less likely to engage in such activism compared to the principled intolerant group (see ESM 1 for additional tables).

Table 3

Multinomial logistic regression results of a full predictor model with principled intolerant as the referent group

	Liking		Prejudiced intolerant		Disliking
Reference = Principled intolerant	Log odds β (SE)	Odds ratios	Log odds β (SE)	Odds ratios	Log odds β (SE)	Odds ratios
Notes. All variables were feature scaled (0–1) before inclusion for ease of comparison. Coefficients and odds ratios therefore represent changes due to movement from the lowest to the highest value on the scale. Coefficient betas represent log odds with standard errors in parentheses. Odds ratios represent odds of category belonging compared to the referent group. SDO = Social Dominance Orientation; RWA = Right Wing Authoritarianism; SE = Standard Error. †p < .09–.05; *p < .05; **p < .01; ***p < .001.
Education
Study 1	−1.06† (0.55)	0.35	−0.93† (0.51)	0.40	−1.33* (0.60)	0.26
Study 2	0.97* (0.39)	2.64	−0.82* (0.41)	0.44	−0.63 (0.53)	0.53
Study 3	0.14 (0.42)	1.16	−1.08* (0.46)	0.34	−0.38 (0.63)	0.69
Study 4	0.64 (0.50)	1.89	−0.80† (0.48)	0.45	−2.16*** (0.59)	0.12
Cognitive sophistication
Study 3	1.23 (0.95)	3.43	−1.18 (0.96)	0.31	−1.67(1.16)	0.19
Political orientation
Study 1	−0.56 (0.58)	0.57	−0.23 (0.53)	0.80	0.16 (0.61)	1.17
Study 2	0.11 (0.45)	1.12	0.51 (0.46)	1.67	−0.78 (0.56)	0.46
Study 3	0.68 (0.47)	1.97	1.73*** (0.50)	5.66	0.81 (0.65)	2.25
Study 4	−0.84† (0.45)	0.43	1.09* (0.44)	2.97	2.94*** (0.57)	18.97
SDO
Study 2	1.78* (0.82)	5.94	2.44** (0.84)	11.47	3.05** (1.05)	21.22
Study 3	0.91 (0.78)	2.48	0.66 (0.87)	1.94	1.82 (1.17)	6.15
RWA
Study 2	−0.27 (0.65)	0.76	0.54 (0.71)	1.71	0.99 (0.95)	2.68
Study 3	−0.98 (0.72)	0.38	1.06 (0.83)	2.88	2.09† (1.15)	8.11
Motivation to control prejudice
Study 3	0.99 (0.89)	2.69	0.36 (0.94)	1.43	−1.26 (1.14)	0.28
Threat
Study 1	−4.69*** (0.76)	0.01	−1.14† (0.69)	0.32	5.76*** (0.96)	317.23
Study 2	−1.67*** (0.51)	0.19	0.94† (0.53)	2.56	4.27*** (0.77)	71.59
Study 3	−0.50 (0.51)	0.61	1.55** (0.50)	4.72	3.80*** (0.74)	44.81
Anti-discrimination
Study 3	0.00 (0.60)	1.00	−1.49* (0.67)	0.23	−3.36*** (1.01)	0.04

Discussion

The accommodation of Muslim practices in Western societies tends to evoke much political and public debate whereby some sections of the population argue for the acceptance of these practices and others are in favor of banning them (Carol & Koopmans, 2013; Morin & Horowitz, 2006). While previous research has examined how the public evaluates these sorts of practices, this research tends to consider the rejection of these practices as an expression of anti-Muslim feelings (e.g., Helbling, 2014; Saroglou et al., 2009; Van der Noll, 2014). However, while people can reject certain practices because of their prejudicial feelings toward Muslims as a group, they might also be opposed to these practices because of more principled objections (e.g., Gustavsson et al., 2016; Imhoff & Recker, 2012; Van Bohemen et al., 2011). Individuals can be intolerant of specific practices while having either negative or positive feelings toward a group (Hagendoorn & Poppe, 2012; Hurwitz & Mondak, 2002; Sniderman et al., 1989).

We examined majority Dutch reactions to the Muslim minority group and different Muslim practices. Drawing on data from four large datasets covering 4 years, we used latent profile analysis to identify groups of individuals across the datasets, and then tested the construct validity of those groups by considering important demographic, psychological, and behavioral correlates of these groups. The advantage of these analyses is that the varying levels of anti-Muslim feelings and rejection to a range of Muslim practices are taken into account and that the heterogeneity of the population is identified.

Across the four datasets and similar to research on political tolerance in different national contexts (McCutcheon, 1985; McIntosh et al., 1995), we identified four latent profiles. In addition to those who generally like and generally dislike Muslims and their practices, we found evidence of two distinct groups of intolerant people that do not explicitly appear motivated by strong negative feelings toward both Muslim groups (but less negative toward Turks compared to Moroccans). One group was intolerant of all or almost all Muslim practices and we labeled this group “prejudiced intolerant.” The second one was intolerant of some but not all practices and was labeled “principled intolerance.” The principled and prejudiced intolerant groups appeared to represent large sections of society with up to half of the participants in our samples falling into these two groups, indicating the importance of investigating (in)tolerance when seeking to understand intergroup attitudes (Verkuyten & Yogeeswaran, 2017).

We considered several demographic and social psychological variables to examine the meaningfulness of the distinction between the groups. The findings supports the expectation that there are individuals whose rejection of specific practices seems to be guided more by principled considerations than a general dislike toward Muslims. For example, the principled intolerant group is higher educated and more likely to engage in cognitively sophisticated thought than the prejudiced intolerant group. Interestingly, an additional analysis which looked only at the predictive power of cognitive sophistication above and beyond education indicated that adding cognitive sophistication as a predictor had no effect on the role of education in predicting category membership (see ESM 1). In light of past research suggesting that education improves intergroup attitudes through increased cognitive sophistication, this might suggest that education may improve intergroup attitudes by conveying liberal and accepting values or by decreasing feelings of intergroup competition. Level of education is also known as a strong correlate of political sophistication (Highton, 2009; Hillygus, 2005; Tedin, 1987) and of being (in)tolerant as a matter of principle (Sniderman et al., 1989). Similarly, compared to the prejudiced intolerant group, the principled intolerant group is more left-wing politically and research in different Western countries has found that political orientation organizes people’s values and beliefs about equality and social deviance (Jost, 2006, 2017; Piurko et al., 2011). Moreover, the principled intolerant group did not only endorse social dominance and authoritarianism less than the prejudiced intolerant group but was also more willing to address the unjust treatment of minority groups.

Our research offers greater nuance than the common distinction between more or less prejudice that is typically used as an underlying continuum for understanding people’s attitude toward Muslim minority groups and the different practices they engage in. By making a distinction between people’s feelings toward the group of people and toward a range of out-group practices, it is possible to identify a more complex constellation of evaluations. For some individuals, their (un)acceptance of Muslim practices corresponds to their anti-Muslim feelings, but for others it does not. For the principled intolerant group, generally positive group feelings are associated with positivity toward some Muslim practices with disapproval of other practices (i.e., founding Islamic schools and, to a lesser extent, wearing the headscarf). Moreover, not all Muslim practices are rejected to the same extent which indicates that a relative interpretation of rejection is more appropriate than an interpretation in terms of generalized rejection. These findings indicate that rejection of a particular practice (headscarf or Islamic schools) cannot simply be taken to indicate prejudice toward Muslims, and that acceptance of a particular practice does not have to indicate non-prejudicial feelings. Research on anti-Muslim attitudes has examined the extent to which these attitudes reflect prejudice or specific forms of critique based on the endorsement of secularism, and Enlightenment and universalistic values (e.g., Breton & Eady, 2015; Elchardus & Spruyt, 2014; Van Bohemen et al., 2011). Indeed, our analyses indicate that while some people may not openly express prejudice, their objection appears to be guided by hidden prejudice, while others do not express prejudice and appear to be rejecting specific practices as a function of principled objections. Following research on political tolerance, we have tried to argue and demonstrate that it is also useful and important to consider group-based attitudes together with the acceptance of group-specific practices. One can tolerate certain practices of a disliked minority group and when the practice itself is controversial one can be intolerant toward the group one likes or dislikes (Hurwitz & Mondak, 2002; Sniderman et al., 1989).

By also evaluating a range of predictors, we were able to identify differences in attitudes, background, and behavioral intentions between the principled and prejudiced intolerant groups. Thus, despite the group identified as prejudiced intolerant showing a relative absence of prejudice in explicit attitudes toward the Muslim outgroup, their consistent rejection of Muslim practices coupled with differences in predictive attitudes suggests that they may harbor implicit prejudice or have prejudice that they are aware of but are unwilling to express (e.g., Pearson, Dovidio, & Gaertner, 2009).

However, although we found similar patterns in four different datasets we need to be careful about generalizing the specific content and the size of the different profiles. The findings of latent profiled analyses and person-centered approaches are context-specific and sensitive to the practices that are considered. The practices used in this research reflect issues that are broadly debated in Western Europe and the profiles indicate how people tend to group these issues. However, different profiles might emerge if different practices were considered. For instance, the consideration of more demanding issues (e.g., arranged marriage, Sharia ruling in the Netherlands) could result in very skewed distributions of answers with different profiles as a consequence, such as the “liking” and the “principled intolerant” groups not sustain their overall acceptance of various practices (Gibson, 2005). Further research is needed to test whether this pattern is broadly replicated across a range of other Muslim minority practices.

Furthermore, although we considered familiar predictors of prejudice and tolerance, it is important to note that our indicators were not explicitly developed for the current analysis. Additionally, while our analysis of predictors of membership in the different groups drew on a wide range of predictors, these predictors did not always use identical items and were not present in all of the four datasets. Therefore, while some predictors have evidence from across multiple datasets, others are present in only one dataset and thus provide weaker evidence. Further research using similar items in new datasets may allow us to further update these findings. In addition, future research could consider other important predictors such as intergroup contact, need for closure, and cultural diversity beliefs as these might provide a further understanding of the differences between the groups of individuals.

Lastly, it is important to note that we focused our discussion on the more stable patterns we found in our datasets and not on the differences between the datasets. While differences are relevant and may be informative, it can be difficult to know with any degree of certainty whether they are due to random noise or whether they reflect socio-political factors (see, e.g., our discussion on attitudes toward Turks in Study 4 in the Out-Group Feelings section). Therefore, we focus on between-sample similarities that we find across multiple datasets collected over a 4-year period. This allows us to assess the probability that the results we discuss represent more general patterns of population characteristics.

Conclusion

In light of the important academic and societal debates on the acceptance and accommodation of Muslims in Western societies, it is critical to parse between forms of intolerance of particular practices that represent more principled positions on complex matters of policy and those which tend to justify the disliking of Muslim minorities. In this research, we found support for both sides of competing perspectives on the source of intolerance and we have explained in more detail how these groups differ. Many majority members are struggling with questions around immigration and Muslim minorities, and the acceptance of dissenting minority practices in particular. Psychologically various types of feelings, beliefs, norms, and values come into play and the weighting and balancing of these considerations against each other is not easy (Verkuyten & Yogeeswaran, 2017). A social psychological perspective that tries to understand the rejection of specific minority practices only in terms of prejudicial attitudes is limited, as is a perspective that ignores the justification of prejudicial feelings and negative beliefs (Crandall & Eshleman, 2003). Ordinary people are influenced by their group-based likes and dislikes but are also capable of considering different principles and values, including the importance of tolerance. Using a person-centered approach makes it possible to identify unobserved groups of individuals who differ in the particular ways in which they try to combine their out-group feelings and evaluations of a range of out-group practices (McCutcheon, 1985; Meeusen et al., 2018). These groups cannot be placed on a unidimensional prejudice continuum but rather form latent classes of majority group members who differently combine their general feelings toward Muslim groups and their acceptance of Muslim practices. In this way a more nuanced understanding of majority members’ evaluation of minority groups and minority practices can be provided which is critical for the continuing social and theoretical debates.

The electronic supplementary material is available with the online version of the article at https://doi.org/10.1027/1864-9335/a000380