Self-Affirmation and Prejudice Against Religious Groups: The Role of Ideological Malleability

“A given ideology can mean different things to different people (…) individuals exploit the malleability of sociopolitical ideologies by actively endorsing them in forms that promote their intergroup goals” (Knowles et al., 2009, p. 857). Ideological malleability implies a motivational component. If individuals are motivated to protect the status quo, they may interpret an available ideology in a way that reinforces intergroup hierarchy, even if that ideology at its core is constructed to challenge social inequality (Knowles et al., 2009). This can be accomplished through two complementary ways: (a) the change in the intensity of adherence to the ideology and (b) the modification of the ideology’s content. Individuals can show a greater adherence to the ideology in certain specific domains. However, this is possible because they modify the content of the ideology, its cognitive representation, to make it more consistent with the maintenance of social hierarchy.

Taking the example of the colorblindness ideology, Knowles and collaborators (Knowles et al., 2009) have shown how this principle that individuals should not be judged by the color of their skin, can be variably used to reduce versus to maintain the status quo of racial inequality in the United States. Specifically, Knowles and colleagues (2009) identified two meanings of the colorblindness ideology: a distributive-justice versus a procedural-justice dictate. Distributive justice refers to the equal treatment of people in resource allocation, regardless of the color of their skin. People with an egalitarian view of intergroup relations agree more with this principle of distributive justice than people with an anti-egalitarian, hierarchical view of social groups. However, to achieve equal treatment of Black people and White people, Martin Luther King’s dream, it might sometimes be necessary to consider people’s skin color through positive actions that help to equalize opportunities and reduce social hierarchy. Anti-egalitarian people can, in these circumstances, claim to be attached to the application of the colorblindness principle in procedures such as employment or university admission. Thus, individuals motivated by the protection of racial hierarchy interpret the ideology of colorblindness as a procedural justice dictate. Studies by Knowles and colleagues (2009) showed that, when faced with an intergroup threat, anti-egalitarian participants’ support for the ideology of colorblindness increased compared to a non-threat condition. However, this is explained by the fact that anti-egalitarian participants shifted their representation of colorblindness from an ideology of distributive justice to one of procedural justice.

Drawing on this work, other researchers have shown that secularism (i.e., “laïcité”) is a malleable ideology that can be adopted by individuals with contrasting attitudes towards immigrants in France, as is the case with colorblindness in the United States (E. Roebroeck & Guimond, 2018). More specifically, Roebroeck and Guimond (2018) showed that anti-egalitarian participants endorsed more the secularism ideology, when an outgroup was presented as a symbolic threat compared to the condition where the same outgroup was presented as a realistic threat or to the control condition. Egalitarian participants did not endorse secularism differently as a function of the condition.

Participants with a hierarchical view of intergroup relations may have a different representation of secularism than those less attached to social dominance (L. Roebroeck & Guimond, 2016). Indeed, in French society, two representations of secularism coexist: the “new” secularism according to which citizens are not allowed to practice their religion in public institutions, and the “historical” secularism, conforming to which citizens are free to practice their religion, without any restriction in public spaces. Unlike “historical” secularism, which is based on the 1905 law guaranteeing the neutrality of the State regarding religious practice, the “new” secularism stems from more recent laws enacted in 2004 and 2010, the first one prohibiting the wearing of religious symbols in public schools and the second banning face coverage in institutions and in the public space. More recently, as of September 2023, it is prohibited by law to wear an Abaya (a long loose dress worn by Muslim women) in public schools. Studies have shown that adhesion to the “new” representation of secularism is associated with a higher level of prejudice against immigrants (L. Roebroeck & Guimond, 2016). In addition, experimental evidence showed that the salience of a “new” secularism norm played a role in increasing prejudice and discrimination (Anier et al., 2019; Nugier et al., 2016).

In this research, we examine whether ideological malleability influences the effectiveness of the self-affirmation procedure in reducing prejudice towards religious groups. We take the example of affective and behavioral prejudice towards Muslims in France, and of the secularism ideology. We argue that self-affirmation on the value of secularism will increase its importance for participants, and their own support for it. Thus, based on evidence showing the differential associations between each secularism representation and prejudice against Muslims (L. Roebroeck & Guimond, 2016), the effect of the self-affirmation manipulation should differ depending on the specific representation of secularism affirmed. Consequently, we argue that self-affirmation decreases prejudice towards Muslims among participants with a “historical” representation of secularism and increases prejudice among participants with a “new” representation of secularism.

Previous research has shown that the effect of self-affirmation on intergroup attitudes can be moderated by values that are central to the individual (e.g., individualism). For example, in the study conducted in the aftermath of terrorist attacks in France, participants who scored higher on individualism were more responsive to the self-affirmation procedure and diminished their support for the government’s discriminative policy against immigrants, compared to those who scored low on this value (Badea et al., 2018). Indeed, the basic principle of self-affirmation is that the procedure taps into the values that are important to the individual (G. L. Cohen & Sherman, 2014). Ideologies governing intergroup relations and their representation in participants’ minds are part of these values that can be used in self-affirmation procedures designed to reduce prejudice. We argue that if the ideology’s representation is compatible with the principle of social equality, self-affirmation can reduce prejudice against minority groups, by increasing the importance of the ideology and support for it. Conversely, if the ideology’s representation enhances social hierarchy, self-affirmation can increase prejudice.

One difficulty in testing the effect of secularism representation on the self-affirmation outcome is linked to methodological aspects of the self-affirmation manipulation. Indeed, in many self-affirmation procedures, participants are presented with a list of values (e.g., love, loyalty) that they must rank from the most to the least important to them personally (e.g., Badea et al., 2021; Cascio et al., 2016). In most cases, the list of values has no direct connection with a specific ideology (that can be the main interest of a study). Thus, it is not sure that a given representation of an ideology is cognitively salient when individuals undertake this kind of self-affirmation task. In this research, we propose to affirm participants on secularism and to examine the effect of self-affirmation on prejudice towards a minority religious group (e.g., Muslims).

Previous research suggests that manipulating the content of the list of values used in the self-affirmation procedure can have an impact on attitudes towards immigrants. For example, affirming values congruent with one’s political orientation increased positive attitudes towards Syrian refugees (Badea et al., 2017). In a different study, Lehmiller and collaborators (Lehmiller et al., 2010) showed that participants who were affirmed by valuing relationships with family and friends were significantly more prejudiced against sexual minorities than participants who were affirmed by valuing other self-relevant aspects. These findings demonstrate that the type of value affirmed is an important consideration to take in the design of self-affirmation procedures. Specifically, in Lehmiller and al.’s (2010) study, familial-based affirmations undermined the reduction of prejudice towards minority sexual groups because they reminded individuals of values that are seen as conflicting with expressing tolerant attitudes towards these groups.

In our case, the efficacy of affirming participants on the secularism value would depend on participants’ representation of this ideology. The representation of “historical” secularism in terms of freedom of religious practice is more compatible with positive attitudes towards minority religious groups. On the other hand, the representation of “new” secularism is more restrictive regarding religious practices and may, on the contrary, increase participants’ prejudice towards religious groups. Since new secularism places limits on religious practices and the display of religious affiliation, affirming this value could seem to conflict with expressing tolerant attitudes towards people that are concerned, religious groups. Being reminded of this value should lead to the perception of minority groups that display their religious affiliation as incompatible with French values, thus perceiving these individuals as a threat to French society, and expressing prejudice against them. Muslims in particular can be targeted under this norm, as wearing visible religious symbols (e.g., headscarf) is common.

In this study, we began by identifying French participants’ representation of secularism (“historical” vs. “new”). They were then randomly assigned to either a self-affirmation experimental condition or to a control condition. In one of the two self-affirmation conditions, participants explained why secularism was an important value for them and provided examples from their daily lives where this principle has guided their behavior. However, secularism can be considered as a threat-related value. The attachment to the principle of secularism is linked to different attitudes towards religious groups (e.g., L. Roebroeck & Guimond, 2016). These minority groups may represent a threat to national identity and therefore to individuals who may identify strongly with their country (e.g., Badea et al., 2020). Thus, we also included a self-affirmation task focused on a threat-unrelated value (humor; see also Lehmiller et al., 2010). Finally, in the control condition, participants were asked to write about an important value (physical endurance) for someone else. All participants completed a measure of prejudice against Muslims.

Nevertheless, members of the majority society may express this view either because they are prejudiced against this community in particular, or because they are committed to the principle of “new” secularism that prohibits religious practices for all religious groups. Indeed, in a recent survey, Dangubić et al. (2020) examined attitudes towards religious practices in many European countries. The authors simultaneously examined attitudes towards Muslims and Christians. They distinguished between participants who rejected Muslim but not Christian religious practices, and those who rejected practices regardless of the religious group. The authors argue that much of the previous research on anti-Muslim prejudice has only considered Muslim practices, potentially misidentifying some participants as prejudiced against Muslims. In our study, we also assessed participants’ attitudes towards the majority religious group (Christians), in order to specifically examine the impact of self-affirmation on attitudes towards Muslims.

We measured the affective dimension of prejudice towards religious groups using a feeling thermometer and the behavioral component of prejudice using distributive matrices (Anier et al., 2018; Rubin et al., 2014). The feeling thermometer measure allowed us to add several other groups, as a distraction to the target groups. Moreover, we analyzed relative negative feelings, comparing the target group (Muslims or Christians) to the mainstream ingroup (French people), instead of the simple feeling towards each group. This allowed us to use a measure less sensitive to social desirability concerns (see Hugenberg & Bodenhausen, 2003). For the behavioral component of prejudice, we asked participants to imagine that they were responsible for allocating subsidies at Paris City Hall, and that they had to decide how to divide a sum of money between a Muslim/Christian association vs. a French association with no religious affiliation. Participants had to indicate their intention to distribute the money between the two associations. Again, the relative behavioral prejudice against each religious group compared to the French association, was assessed.

We did not include the cognitive measure of prejudice (i.e. stereotyping) because it seems difficult to compare stereotypes about Christians (e.g., lower scientific competence, Rios et al., 2015) with stereotypes about Muslims (e.g. terrorist, Saleem & Anderson, 2013), see also Erentzen et al. (2022). In contrast, examining affective and behavioral prejudice allowed us to use the same measures for both groups.

According to previous work, we hypothesized a main effect of secularism representation, a main effect of the type of religious group and a main effect of self-affirmation such that (1) participants with a “historical” representation of secularism were expected to exhibit less prejudice towards religious groups than those with a “new” representation of secularism, (2) higher prejudice was expected towards Muslims than towards Christians, and (3) participants who self-affirmed on the threat-unrelated value were predicted to show lower prejudice towards religious groups compared to those in the other two conditions. Following this reasoning, we also expected an interaction effect between secularism representation and the type of religious group such that participants with a “new” representation of secularism would exhibit more prejudice (as compared to “historical” representation), and that this difference would be higher regarding prejudice towards Muslims as compared to Christians.

We also tested the key hypothesis concerning the interaction between secularism representation and self-affirmation on prejudice. Specifically, we predicted that participants who self-affirmed on a threat-unrelated value would express lower prejudice compared to control, regardless of their secularism representation, as self-affirmation would buffer the psychological threat causing the expression of prejudice. On the other hand, we expected different results when self-affirming on secularism, as the affirmed threat-related value would moderate the buffering effect of self-affirmation. In this case, the content of the affirmed value is fundamental to the effect (unlike the affirmation of a threat-unrelated value, where the content of the value is irrelevant), and could increase the perceived threat when the value is intolerant or decrease it when it is tolerant. Therefore, we expected participants with a “historical” representation of secularism to exhibit lower prejudice in the condition of self-affirmation on the secularism value as compared to control. Conversely, we expected participants with a “new” representation of secularism to exhibit higher prejudice in the condition of self-affirmation on secularism compared to control. We examined whether these effects were particularly observed concerning prejudice towards Muslims.

We provided all materials, data, and code on: https://osf.io/63rnq/. This project received Peer Community in Registered Report Stage 1 in-principle acceptance (https://osf.io/smc5p) and was pre-registered (https://osf.io/f4wm6), then we proceeded to data collection. All measures, manipulations, exclusions conducted for this investigation are reported, and data collection was completed before analyses.

We calculated the required sample size using G*Power 3.1. We set the SESOI to f = .10 for an ANOVA repeated measures within-between interactions. We concluded that the minimum required sample size for 6 experimental groups, 2 repeated measures, a power of 0.95 and alpha of 0.05 is 504 participants (G*Power screenshot is presented in the Supplementary Figure S1). To minimize the risk of overestimating the true population effect size (Perugini et al., 2018), as well as to account for potential exclusion of participants, we planned to recruit 600 participants.

Based on these aforementioned calculations, we recruited a final sample of 602 native French speakers, born, raised and based in France, regardless of their religious identification, through a survey institution (Bilendi) to ensure high quality data collection and to have a representative sample of the French population in terms of regions, socio-economic status, age, and gender. Exclusion criteria were automatically applied during data collection, thus all participants that completed the study to the end were included in the final sample. Participants who composed the final sample were aged between 18 and 90 years old (M_age = 47.3, SD = 16.3, Mdn = 47). The sample included 299 females, 301 males and 2 who did not disclose their gender. Concerning religious identification, Christians composed the majority of the sample (53.8%) followed by atheists (23.4%). Participants were compensated with points that can be collected and used to buy gifts.

We summarize the experimental design in Table 1. The design is as follows: 2 (secularism representation: historical and new; between-subjects) by 3 (self-affirmation: threat-related value, threat-unrelated value and control; between-subjects) by 2 (religious group: Muslims and Christians; within-subjects). Display of the within-subjects factor (type of religious group) was counterbalanced. All translated materials are presented in the Supplementary.

[For review: The Qualtrics survey .QSF file and an exported DOCX file are provided on the OSF folder. The translated Qualtrics survey can be found in the Supplementary. A preview link of the Qualtrics survey is provided on: https://parisouestpsy.eu.qualtrics.com/jfe/form/SV_07BJ9ppKM7hWCGy]

The online survey was built using Qualtrics. Participants indicated their consent, with four questions confirming their eligibility, understanding, and agreement with study terms, which they must answer with a “yes” in order to proceed to the study. Participants then provided their demographic information. Next, after indicating their personal representation of secularism, they were randomly assigned to one of the three experimental conditions (self-affirmation on secularism, self-affirmation on humor or control condition). Finally, they completed the measurements of prejudice against Muslims and Christians. At the end of the experiment, they answered a few funneling questions.

Participants were first told that secularism is an important value in France and that two visions of it coexist. Vision 1 (corresponding to “historical” secularism) stated: “Individuals are free to practice their religion in private and in public. Citizens have the right to show their religious affiliation in the public sphere. The State must remain completely neutral regarding these religious practices”. Vision 2 (corresponding to “new” secularism) stated: “Individuals are free to practice their religion in private, but not in public. Citizens do not have the right to show their religious affiliation in the public sphere. The State must regulate these religious practices”. These visions of secularism were constructed on the basis of items used by Roebroeck and Guimond (2016) and De la Sablonnière et al. (2020). Participants were then asked to indicate which of the two visions best reflected their personal convictions.

As a manipulation check, participants then reported on a 7-point scale, ranging from 1 (strongly disagree) to 7 (strongly agree), how much they agreed with the following statement “Religious practices and symbols should not be allowed in public institutions”, representing the main difference between the two visions. Participants with a “new” secularism vision were expected to score higher on this question than participants with a “historical” secularism vision.

Following, participants were randomly assigned into one of the three conditions: self-affirmation on the threat-related value (i.e., secularism), self-affirmation on the threat-unrelated value (i.e., humor) or control condition. In many self-affirmation activities, participants are asked to rank a list of values in order of importance for them personally, and to answer questions about the importance of the top ranked value (e.g., Badea et al., 2018). In this study, we have adapted the procedure to ask them to self-affirm on the threat-related value (secularism), or a threat-unrelated value (humor, see also, Lehmiller et al., 2010, for a similar procedure). Participants in both self-affirmation conditions were asked to affirm themselves on the corresponding value (their representation of the secularism value or the humor value) by answering two questions. The first asked them to explain why this value (i.e., secularism or humor) would be important for them personally, and the second required giving examples of how this value guides their behavior in daily life (for example, in their interactions with others). In the control condition, participants explained why physical endurance would be an important value for another person and gave examples on how this value could guide this person’s behavior. In order to keep constant the salience of participants’ secularism representation across all conditions, all participants were reminded of their choice of secularism representation before they start the self-affirmation or the control activity.

Immediately after the manipulation, participants were required to indicate the importance of the affirmed value in each condition (secularism, humor and physical endurance accordingly) to them personally on a 7-point scale from 1 “not at all important” to 7 “very important”.

Affective prejudice against the two targeted religious groups (Muslims and Christians) was measured using a feeling thermometer. Participants were asked to indicate their feelings towards 10 groups including Muslims, Christians, and French people on a scale from 0 “Very negative feelings” to 100 “Very positive feelings”. Then, a relative affective prejudice score was computed (see also Hugenberg & Bodenhausen, 2003) by calculating the difference between feelings towards the ingroup (French people) and the target group (Muslims or Christians). A higher score indicated higher relative affective prejudice (see supplementary for more details on the score calculation). The presentation order of the ten evaluated groups was randomized.

Two distributive matrices were used to measure behavioral prejudice against each target group. Participants were asked to imagine that they are responsible for allocating subsidies at Paris City Hall, and that they have to indicate their intention to divide a sum of money between two high impact associations that need support. Four fictitious student associations were used. One matrix included a Muslim vs. a French association, while the other one included a Christian vs. another French association. Participants were given 7 options representing different ways of attributing money to the French association on the one hand, and to the Muslim or Christian association on the other. They had to choose between 1 and 7, with a higher score representing more money allocated to the French association and less money allocated to the Muslim or Christian association. A higher score indicated higher relative behavioral prejudice towards the religious group in question (see supplementary for more details). These matrices allowed us to measure the relative de-favoritism of Muslims and that of Christians in comparison with French people. Relative prejudice was used to minimize the sensitivity of our measures to social desirability bias.

Participants were asked to indicate on a 7-point scale “To what extent would you say you practice a religion?” from 1 “Not at all” to 7 “Absolutely”.

One attention check item was included in the feeling thermometer to detect careless responding: “Please put the cursor on ten”.

Firstly, we expected a significant main effect of both the secularism representation and the type of religious group on prejudice, as well as an interaction between the two variables. It was hypothesized that (1) affective and behavioral prejudice will be higher among participants who have a “new” representation of secularism (vs. “historical”) and that (2) there will be more affective and behavioral prejudice towards Muslims than towards Christians. We expected secularism representation to interact with the type of religious group in that participants who have a “new” representation of secularism will exhibit more affective and behavioral prejudice (as compared to “historical” representation), and that this difference will be higher regarding prejudice towards Muslims as compared to Christians. Secondly, we expected a main effect of self-affirmation, such that participants who self-affirm on the threat-unrelated value will show lower prejudice towards religious groups compared to those in the other two conditions. Central to this work, we aimed to test the interaction hypothesis between self-affirmation and secularism representation on prejudice towards religious groups. We hypothesized that self-affirmation on secularism (the threat-related value) will decrease affective and behavioral prejudice towards religious groups, among participants who have a “historical” representation of secularism compared to control. The reverse pattern is expected (self-affirmation on secularism increasing prejudice against religious groups) among participants who have a “new” representation of secularism. We expected self-affirmation on humor (the threat-unrelated value) to decrease affective and behavioral prejudice against religious groups compared to control, regardless of the chosen secularism representation. Finally, we examine whether the interaction between self-affirmation and secularism representation is more strongly observed when considering prejudice towards Muslims than prejudice towards Christians.

To test these hypotheses, we ran a three-way mixed ANOVA : 2 (secularism representation: historical vs. new) x 3 (self-affirmation: threat-related value vs. threat-unrelated value vs. control) x 2 (religious group: Christians vs. Muslims). For all analysis, significance threshold α was set at .05.

Participants who failed the attention check item (i.e., selecting another response than 10 to this question) and participants who reported less than 4 out of 5 on the question asking them to what extent they completed the study seriously were automatically excluded from the study and could not complete the study to the end. We did not classify outliers.

Firstly, we checked if our identification of participants’ representation of secularism successfully distinguished different beliefs about secularism. We conducted an independent t-test and found a significant difference between participants endorsing a “new” vision of secularism (M = 5.89, SD = 1.53) and those endorsing a “historical” vision of secularism (M = 3.72, SD = 2.02) in their opposition to public religious display, t(600) = 13.60, p < .001, d = 1.31. This indicates that participants who chose “new” secularism displayed significantly distinctive beliefs concerning secularism from those who chose “historical” secularism.

Secondly, to verify that the affirmed values were considered important to participants, we ran an ANOVA 3 (Value: Secularism vs. Humor vs. Physical endurance) between-subjects with the importance of the value as the dependent variable. Results indicated a significant difference in value-importance between conditions, F(2, 599) = 44.10, p < .001, η_p² = .128. The secularism value (M = 5.75, SD = 1.46) was perceived as significantly more important than the physical endurance value (M = 4.87, SD = 1.52), p < .001, d = .645. The humor value (M = 6.14, SD = 1.09) was significantly more important to participants than the physical endurance value, p < .001, d = .930. Finally, the difference between the importance of the secularism value and that of the humor value was significant, p = .010, d = .286. Overall, these results suggest that participants rated the affirmed values (secularism and humor) as more important than the control value (physical endurance).

Descriptive statistics are presented in Table 2.

In order to test each hypothesis, we ran a 2 (secularism representation: historical vs. new) by 3 (self-affirmation: threat-related value vs. threat-unrelated value vs. control) by 2 (religious group: Muslims vs. Christians) mixed ANOVA for each of the prejudice measures (affective measure and behavioral measure).

We expected a significant main effect of both the secularism representation and the type of religious group on prejudice, as well as an interaction between secularism representation and type of religious group.

Affective Prejudice. We found a significant main effect of secularism representation, F(1, 596) = 14.69, p < .001, η_p² = .024, and main effect of religious groups, F(1, 596) = 98.59, p < .001, η_p² = .142. Participants reported significantly more prejudice towards religious groups when they displayed a “new” vision (M = 19.50, SD = 25.12) of secularism as compared to a “historical” vision (M = 10.03, SD = 23.40), d = .383. Affective prejudice scores were significantly higher towards Muslims (M = 26.7, SD = 32.3) than towards Christians (M = 7.92, SD = 26.4), d = .601.

Finally, we found support for the two-way interaction between secularism representation and religious group, F(1, 596) = 26.84, p < .001, η_p² = .043, see Figure 1. Pairwise comparisons indicated that participants displaying a “new” representation of secularism exhibited significantly higher affective prejudice against Muslims (M = 30.68, SD = 32.03) than towards Christians (M = 8.32, SD = 26.76), t(596) = 15.71, p < .001, d = .722. Participants displaying a “historical” representation of secularism exhibited significantly higher affective prejudice against Muslims (M = 13.45, SD = 29.55) than against Christians but the effect was smaller, (M = 6.61, SD = 25.24), t(596) = 2.71, p = .035, d = .237.

Behavioral Prejudice. As for affective prejudice, there was a significant main effect of secularism representation, F(1, 596) = 14.12, p < .001, η_p² = .023, and a main effect of the religious group, F(1, 596) = 45.76, p < .001, η_p² = .071, on behavioral prejudice. Participants having a “new” representation of secularism (M = 5.23, SD = 1.37) exhibited significantly more behavioral prejudice against religious groups than those reporting a “historical” vision (M = 4.75, SD = 1.29), d = .361. Behavioral prejudice scores were significantly higher towards Muslims (M = 5.42, SD = 1.54) as compared to Christians (M = 4.82, SD = 1.55), d = .409.

The two-way interaction between secularism representation and religious group was significant, F(1, 596) = 10.77, p = .001, η_p² = .018, see Figure 2. Participants with a “new” representation (M = 5.58, SD = 1.52) exhibited significantly more prejudice towards Muslims than towards Christians (M = 4.88, SD = 1.60), p < .001, d = .461. Among participants with a “historical” representation of secularism, behavioral prejudice against Muslims (M = 4.87, SD = 1.45) was not significantly higher than that towards Christians (M = 4.62, SD = 1.38), p = .194, d = .214.

To begin, we tested whether participants who self-affirmed on the threat-unrelated value showed lower prejudice towards religious groups compared to those in the other two conditions. Then, we tested the key interaction hypothesis between self-affirmation and secularism representation on prejudice towards religious groups. We expected self-affirmation on humor (the threat-unrelated value) to decrease affective and behavioral prejudice against religious groups compared to control, regardless of the chosen secularism representation, since the process of self-affirmation should protect from any perceived threat leading to the expression of prejudice. On the other hand, we expected self-affirming on the threat-related value to be affected by the content of the value. In other terms, affirming one’s representation of secularism should moderate the protecting effect of the mere self-affirmation process. We hypothesized that self-affirmation on secularism (the threat-related value) will decrease affective and behavioral prejudice towards religious groups, among participants who have a “historical” representation of secularism compared to control. The reverse pattern is expected (self-affirmation on secularism increasing prejudice against religious groups) among participants who have a “new” representation of secularism.

We also examined the interaction between the three factors (self-affirmation, secularism representation and religious group). We expected self-affirmation and secularism representation to interact with the type of religious group, such that the key interaction effect between self-affirmation and secularism representation would be observed particularly for prejudice against Muslims compared to prejudice against Christians.

For each prejudice score (affective and behavioral), we begin by presenting the results concerning the main effect of self-affirmation on prejudice. We then test the interaction between self-affirmation and secularism representation on prejudice (general score). Finally, we test whether this interaction depends on the religious group (Christians and Muslims), reporting the results concerning the three-way interaction between self-affirmation, secularism representation and religious groups.

Affective Prejudice. We found no significant main effect of self-affirmation on affective prejudice, F(2, 596) = 2.26, p = .106, η_p² = .008. Participants did not significantly report more prejudice towards religious groups in the threat-unrelated condition (M = 18.91, SD = 25.61) as compared to the threat-related condition (M = 14.95, SD = 23.64), and the control condition (M = 18.09, SD = 25.81), respectively, t(596) = 2.03 , p = .107, d = .158, and t(596) = 0.44, p = .900, d = .033. The difference between the threat-related condition and the control condition was not significant, t(596) = 1.51, p = .287, d = .125.

We then tested the interaction hypothesis regarding self-affirmation and secularism representation on affective prejudice towards religious groups. The interaction between self-affirmation and secularism representation was not significant, F(2, 596) = 1.08, p = .342, η_p² = .004.

In order to test whether the interaction between self-affirmation and secularism representation on prejudice is moderated by the religious group (Muslims or Christians), we examined the three-way interaction between these variables. The three-way interaction was not significant, F(2, 596) = 0.202, p = .818, η_p² = .001.

Behavioral Prejudice. We ran the same analysis on behavioral prejudice. We found no significant main effect of self-affirmation on behavioral prejudice, F(2, 596) = 0.780, p = .459, η_p² = .003. Participants did not significantly report more prejudice towards religious groups in the threat-unrelated condition (M = 5.14, SD = 1.43) as compared to the threat-related condition (M = 5.04, SD = 1.29), and the control condition (M = 5.19, SD = 1.37), respectively, t(596) = 1.25, p = .425, d = .073, and t(596) = 0.62, p = .808, d = .034. The difference between the threat-related condition and the control condition was not significant, t(596) = 0.58, p = .834, d = .107.

Results did not indicate a significant two-way interaction between self-affirmation and secularism representation, F(2, 596) = 1.47, p = .231, η_p² = .005. The three-way interaction between self-affirmation, secularism representation and religious group was not significant, F(2, 596) = 0.32, p = .728, η_p² = .001.

As religion may be related to prejudice towards religious groups, it could be argued that the differences in participants’ religiosity may have accounted for our findings. We therefore examined whether participants’ religiosity had an impact on the effects of self-affirmation and of secularism representation on affective prejudice and behavioral prejudice. To test this possibility, we ran a multiple-regression analysis on prejudice towards religious groups, including participants’ religiosity (i.e., the extent to which they report practicing a religion; centered), two orthogonal contrasts corresponding to self-affirmation conditions (C1: self-affirmation on secularism = 2, self-affirmation on humor = -1, control = -1; C2: self-affirmation on secularism = 0, self-affirmation on humor = -1, control = 1), secularism representation (“New” secularism = -0,5; “Historical” secularism = 0.5), the interactions between religiosity and self-affirmation’s contrasts, and the interaction between religiosity and secularism representation.

Affective Prejudice. There was no significant interaction between C1 and religiosity, b = 0.02, 95% CI [-0.78, 0.82], t(594) = 0.05, p = .962, η_p² = .000, nor between C2 and religiosity, b = -0.57, 95% CI [-2.01, 0.88], t(594) = 0.77, p = .441, η_p² =.001. The interaction between religiosity and secularism representation was not significant, b = .82, 95% CI [-0.40, 2.03], t(594) = 1.32, p = .186, η_p² = .001.

Behavioral Prejudice. The interactions between C1 and religiosity and C2 and religiosity were not significant, respectively, b = 0.01, 95% CI [-0.04, 0.05], t(594) = 0.32, p = .752, η_p² = .000, and b = -0.05, 95% CI [-0.13, 0.03], t(594) = 1.21, p = .227, η_p² = .002. There was no significant interaction between religiosity and secularism representation, b = 0.04, 95% CI [-0.03, 0.11], t(594) = 1.24, p = .217, η_p² = .003.

While no significant main effect of the self-affirmation condition was observed, the mean patterns suggest that affective prejudice was lower in the threat-related condition as compared to the other two self-affirmation conditions. In order to test whether participants expressed lower prejudice in the self-affirmation on secularism condition as compared to the other two conditions together, we ran a supplementary analysis including C1(self-affirmation on secularism = 2, self-affirmation on humor = -1, control = -1), C2 (self-affirmation on secularism = 0, self-affirmation on humor = -1, control = 1), secularism representation (“New” secularism = -0,5; “Historical” secularism = 0.5) and their interactions, on both affective and behavioral prejudice.

Affective Prejudice. The contrast opposing self-affirmation on secularism against the two other conditions was significant, b = -1.70, 95% CI [-3.33, -0.08], t(596) = 2.06, p = .040, η_p² = .007. Participants were less prejudiced against religious groups in the self-affirmation on secularism condition (M = 14.95, SD = 23.64) as compared to the other two conditions combined (M = 18.52, SD = 25.68). The contrast opposing self-affirmation on humor (M = 18.91, SD = 25.61) and the control condition (M = 18.09, SD = 25.81) was not significant, b = -0.66, 95% CI [-3.60, 2.29], t(596) = 0.44, p = .661, η_p² = .000. The interactions between C1 and secularism representation, and C2 and secularism representation were not significant, respectively, b = 2.36, 95% CI [-0.89, 5.61], t(596) = 1.43, p = .155, η_p² = .003 and b = 0.87, 95% CI [-5.02, 6.75], t(596) = 0.29, p = .773, η_p² = .000.

Behavioral Prejudice. We did not find a significant effect of C1, b = -0.05, 95% CI [-0.14, 0.04], t(596) = 1.06, p = .291, η_p² = .002, nor of C2, b = -0.05, 95% CI [-0.21, 0.11], t(596) = 0.62, p = .535, η_p² = .001 on behavioral prejudice. There were no significant interactions between C1 and secularism representation, nor between C2 and secularism representation, b = 0.05, 95% CI [-0.13, 0.22], t(596) = 0.50, p = .616, η_p² = .000 and b = 0.26, 95% CI [-0.06, 0.59], t(596) = 1.62, p = .106, η_p² = .004.

The contrast analysis showed that participants who self-affirmed on secularism were less prejudiced against religious groups as compared to those in the other two conditions combined. This contradicts our hypothesis (H1), as we expected self-affirmation on the threat-unrelated value (i.e., humor) to reduce prejudice against religious groups compared to the other two conditions. Although the effect size η_p² = .007 which corresponds to f = 0.084 is lower than the SESOI we defined (f = .14), we wanted to understand this pattern, and we decided to conduct an exploratory content analysis (Stemler, 2000) of the open-ended responses in the affirmation task. We analyzed the responses of the 602 participants in the three experimental conditions. First, two coders looked at the data to get an idea of the emerging codes and establish an initial version of the codebook. Then, two authors coded a subset of the responses. The codebook was developed iteratively, after several discussions, and the final version was established. Intercoder reliability was evaluated (following the guidelines of O’Connor & Joffe, 2020) and Cohen’s kappa was computed for each code individually and across all codes (De Vries et al., 2008). Cohen’s pooled kappa indicated an almost perfect agreement, k_pooled = .97 (Stemler, 2000). All data were then coded using the final version of the codebook (for a detailed description of the codes see the codebook at https://osf.io/63rnq/) and frequencies of the codes were calculated (see Table S1 in the Supplementary).

Afterward, we looked for emerging categories in the codes. We noticed that the participants explained the importance of each affirmed value on different levels; they mentioned reasons that demonstrate why the value was important sometimes for them as individuals, sometimes for the value’s role in social interactions and other times for its relation to national values and identity. We therefore identified three main categories, inspired by (Doise, 1982), that distinguish the reasons given for the importance of each affirmed value according to the level concerned: individual, social and national. In the individual level, we included codes such as coping, wellbeing, and health. In the social level, we included respecting others, facilitating social interactions and avoiding conflict. Finally, the national level consisted of topics like freedom, equality and French identity. Next, we computed the observed frequencies of each level in the three self-affirmation conditions (Table S2 in the Supplementary). To test the differences in level frequencies in each condition, we used the X² test and found a significant difference in level frequencies X²(4, N = 717) = 495, p < .001, V = .59 (Table S3 in the Supplementary). We found that participants in the secularism condition have mostly explained the importance of this value at the national level (138 times compared to 88 social, and 6 individual), whereas those in the humor (166 individual, 88 social, and 8 national) and in the physical endurance (209 individual, 14 social, and 0 national) conditions mostly mentioned reasons at the individual levels.

One possible interpretation is that self-affirmation on secularism may have played a protective role for national identity, whereas this would not be the case in the other two conditions. Indeed, if we look more closely at the codes identified in the secularism condition, we can find French national values (e.g., liberty and equality), national norms carried by laws, (e.g., neutrality in the public sphere and limiting personal convictions to the private domain), and codes that can refer to the third value of the French motto, fraternity (e.g., avoiding conflicts and facilitating social interactions). For example, some participants explained that secularism is an important value because it is a part of the French identity :“The values of our parents and great-grandparents and our identity”, others said that it guarantees equality between citizens : “to maintain equality for all in the public sphere”. Freedom was particularly emphasized by participants with a historical representation : “Individuals have the right to live and express their beliefs”. Moreover, some respondents stated that secularism can help avoid conflicts, mostly in intergroup relations : “There would be less conflict if everyone practiced their religion in the private sphere without exposing themselves in the public arena”.

However, this interpretation may imply attributing to the participants an intention to protect their national identity in the self-affirmation on secularism, compared to the two other conditions. An alternative interpretation would be that the values used in the self-affirmation task may inherently encourage thinking in either national or individual terms, secularism being a collective value while humor and physical endurance are more personal ones. We develop this aspect in the discussion section.

The purpose of this study was to examine whether the ideological malleability of the principle of secularism influences the effectiveness of the self-affirmation procedure in reducing affective and behavioral prejudice towards religious groups (Muslims vs. Christians) in France. First, the results provide some evidence that the malleability of the understanding of secularism as an ideology is associated with attitudes towards religious groups in France, particularly attitudes towards Muslims. In a representative sample of the French population, we find that participants who report a “historical” representation of secularism exhibit less affective and behavioral prejudice towards religious groups than those who have a “new” representation of secularism, and that prejudice is higher towards Muslims than towards Christians. More specifically, participants with a “new” representation show significantly higher affective and behavioral prejudice towards Muslims than towards Christians, whereas those with a “historical” representation show a slightly higher affective, but not behavioral, prejudice towards Muslims than towards Christians. It is worth noting that participants’ religiosity had no impact on these effects.

These results are consistent with previous work that show similar relationships between adherence to “new” secularism norms and prejudice and discrimination against immigrants (Anier et al., 2019; Nugier et al., 2016; L. Roebroeck & Guimond, 2016). However, previous research assessed participants’ adherence to the principle of new secularism without considering the historical representation of secularism and examined how this adherence was associated with prejudice against North African immigrants, without specifying their religious affiliation. To our knowledge, this study is the first to clearly distinguish between participants with new and historical personal representations of secularism, and to demonstrate how these contrasting representations are related to prejudice against different religious groups. Participants in our study predominantly reported a personal “new” representation of secularism, which is considered to be a new and more restrictive interpretation of French secularism, driven by policies and laws banning religious symbols in the public sphere (Baubérot, 2012). This highlights the important role played by policies in shaping attitudes towards minority groups. In line with these findings, recent work by Kende et al. (2024) shows how integration policies impact majority perceptions of minority groups and their threat reactions in different contexts in Europe and the United States.

This study was a first attempt to test the interaction between secularism representations and self-affirmation interventions on prejudice. Contrary to our first hypothesis, self-affirmation had no significant effect on either affective or behavioral prejudice towards religious groups. Similarly, no significant interaction effect was found between self-affirmation and either of the other two variables (i.e., secularism representation and religious group). Our exploratory contrast analysis shows that participants in the condition of self-affirmation on secularism report lower affective (but not behavioral) prejudice towards religious groups as compared to the other two conditions combined (i.e., humor and physical endurance). This pattern is contrary to our prediction, based on previous research (e.g., Lehmiller et al., 2010), that self-affirmation on the threat-unrelated value should reduce prejudice against religious groups relative to the other two conditions, but the effect size is lower than the defined SESOI.

Exploratory content analysis of the open-ended answers in the self-affirmation task reveals that when asked to reflect on reasons that make secularism an important value, participants mostly reflected on national-level values, norms and on intergroup relations, all of which make the collective aspect of the self salient, whereas, participants in the remaining two conditions mostly reflected on individual-level benefits. More specifically, while reflecting on “historical” and “new” secularism, participants were reminded of French values such as liberty, equality, and fraternity by indicating that secularism promotes positive interpersonal and intergroup relations. This primarily leads us to think that self-affirmation on secularism may have played a protective role for national identity, which may have led to decreased prejudice in this experimental condition. However, we should be cautious regarding this interpretation, as the individual vs. collective aspects evoked here can be inherent to the values themselves. In addition, the failure of self-affirmation on humor in reducing prejudice could be attributed to participants placing insufficient importance on humor as a value, thus it being incapable of protecting self-integrity from perceived threats coming from religious groups.

It is possible that participants in our study associated historical and new secularism with tolerant values that promote positive intergroup relations. However, our results in addition to a growing body of evidence suggest that new secularism is related to negative attitudes towards ethnic and religious minority groups. Even if hiding one’s religious beliefs or affiliations in public could have the potential to help people live more harmoniously, the fact that it is mandatory by law (in public institutions), and has become the social norm in public or even a personal belief, can turn people who do not comply to this norm into deviants, and can lead to their perception negatively, which defies the purpose. In a study conducted by Nugier et al. (2016), highly prejudiced individuals evaluated more negatively and exerted more normative pressure on a North African target who had attitudes that did not conform to the “new” secularism norm. In their second study, highly prejudiced individuals reported more negative evaluations and practiced more pressure against a Muslim target compared to a Catholic target who did not adhere to the “new” secularism norm. In line with Dangubić et al. (2020) work, adhesion to secularism principle may hide negative attitudes towards Muslims in particular rather than a rejection of religious practices regardless of the religious group.

A first limitation that can be highlighted is that our self-affirmation procedure imposed the value to be affirmed, which is different from many self-affirmation procedures where participants are asked to rank a list of values, then reflect on the highest ranked one. However, this procedure was inspired by Lehmiller et al. (2010), who still found more consistent results with the self-affirmation theory. Although, our value importance check showed that humor received the highest rating of importance to the self, and that physical endurance received the lowest, as intended, our manipulation does not guarantee that humor is sufficiently important to participants to offset the threat to self-integrity. Despite this limitation, our results raise questions about the efficacy of self-affirmation as a prejudice reduction intervention, and its ability to protect self-image in a way that would reduce perceived threat and defensiveness, especially in an increasingly tense national and global context.

A second limitation pertains to the behavioral prejudice measure and its sensitivity to the self-affirmation manipulation. Distributive matrices, being cognitively complex and demanding, may not be sufficiently sensitive to the subtle and brief self-affirmation activities. This may explain why some effects were not replicated on both affective and behavioral prejudice measures. Future research may benefit from using more sensitive measures of behavioral prejudice.

Finally, another limitation can be that we did not allow participants to choose neither of the two secularism representations. Nonetheless, our check item revealed that the dichotomous options successfully identified distinctive beliefs about secularism. Furthermore, the fact that the vast majority of participants in our study have a “new” representation of secularism could be seen as a limitation, but this seems to be representative of the French population as shown by survey data. IFOP (2023b) found that 81% of French people approved of the 2023 law banning the Abaya and the Qamis in schools, demonstrating an even higher support than that of the 2004 law banning religious symbols (73%). Our study confirms that high adherence to “new” secularism is related to negative attitudes against religious minorities. The idea that people should hide their religious identity in order to be accepted in society and to reduce conflicts reveals rejection of diversity and inclusion. We argue that this reasoning goes against what French historical secularism stands for, which is freedom and equality for all citizens, regardless of their beliefs or non-belief. It is worth noting that 78% of Muslims in France believe that secularism, in its currently applied form by the authorities, discriminates against them as Muslims (IFOP, 2023a). Indeed, pushing and promoting similar laws and norms could have the potential to exacerbate intergroup tensions and conflicts.

To extend the current findings, future research could examine the effect of such policies and ideologies on religious and ethnic minority groups’ attitudes towards the majority group and the State. Several outcomes, such as perceived injustice and perceived threat, would be interesting to study as they can be crucial for understanding and predicting intergroup relations (e.g., Doosje et al., 2013). Furthermore, Muslims and North Africans have been the main focus of studies on secularism in France. While this is legitimate given the existing data, it would be interesting to explore these effects in other ethnic and religious minority groups. Finally, self-affirmation, in its usual or more recent adapted procedures (e.g., Tavitian-Elmadjian et al., 2019), could also be studied in minority groups as a potential intervention to face adversity and increase wellbeing, in the context of less tolerant norms and policies.

Contributed to conception and design: AY, BC, SB

Contributed to acquisition of data: AY, SB

Contributed to analysis and interpretation of data: SB, AY

Drafted and/or revised the article: AY, BC, SB

Approved the submitted version for publication: AY, BC, SB

We thank Gilad Feldman (ORCID : 0000-0003-2812-6599) for his helpful feedback and support in writing this manuscript.

We thank the reviewer Pete Harris (ORCID: 0000-0003-4599-4929) for his important suggestions for ameliorating this manuscript at each stage of its conception.

This research was funded by the National Agency of Research in France (AFFIRMATIF ANR-21-CE28-0018-01), grant obtained by Constantina Badea.

IRB approval for this study was obtained (N° 2022-06-03) by the University of Paris Nanterre.

The authors declare no potential conflicts of interests with respect to the authorship and/or publication of this article.

All materials, data, and analysis scripts can be found on this paper’s project page on OSF: https://osf.io/63rnq/.