Legitimating Discrimination Against Students with Disability in School: The Role of Justifications of Discriminatory Behavior

have seen an increasing trend toward including children with disability in regular education settings in various education systems. In France, this tendency was triggered by legal requirements (law no2005-102 of February 11, 2005), making schooling in the mainstream education system compulsory for children with special needs. Empirical evidence has shown the benefits of inclusive education for all children, with and without disability, by developing openness to and tolerance of individual differences and improving academic performance (e.g., Baker et al., 1995; Ruijs & Peetsma, 2009). Acting in an egalitarian way thus became a prescriptive norm for teachers. Nevertheless, studies reported that although teachers positively evaluate this principle, they do not completely, on a voluntary or involuntary basis, adhere to the ‘zero reject’ idea (Avramidis & Norwich, 2002). Teachers may be reluctant to include children with disability in their classroom (De Boer et al., 2010) and may be prone to believe that these students would receive better education in specialized education settings. Thus, teachers often hold ambivalent feelings and beliefs toward the inclusion of students with disability, in the sense of the reluctance to openly express one’s own opinions and feelings on sensitive social issues. Indeed, negative attitudes toward children with disability are unacceptable in Western societies because of the very strong normative protection of the children group, in general, and of children with disability, in particular (Crandall et al., 2002; Lüke & Grosche, 2018; Rohmer & Louvet, 2018). Teachers may solve this contradiction between anti-discriminatory norms and perceived difficulties to include students with disability by overlooking discrimination when covered by benevolent justifications. The present paper seeks to specifically answer the following question: How do teachers react to prejudiced behaviors against children with disability exhibited by fellow ingroup members in their professional environment? More specifically, we aimed to show that the social acceptability of the justification provided when excluding a student with disability decreases the perception of discrimination and the social rejection of the perpetrator of the discrimination. Bastart, J., et al. (2021). Legitimating Discrimination Against Students with Disability in School: The Role of Justifications of Discriminatory Behavior. International Review of Social Psychology, 34(1): 1, 1–11. DOI: https://doi.org/10.5334/irsp.357

After a long segregation period, the last two decades have seen an increasing trend toward including children with disability in regular education settings in various education systems. In France, this tendency was triggered by legal requirements (law nº2005-102 of February 11, 2005, making schooling in the mainstream education system compulsory for children with special needs. Empirical evidence has shown the benefits of inclusive education for all children, with and without disability, by developing openness to and tolerance of individual differences and improving academic performance (e.g., Baker et al., 1995;Ruijs & Peetsma, 2009). Acting in an egalitarian way thus became a prescriptive norm for teachers. Nevertheless, studies reported that although teachers positively evaluate this principle, they do not completely, on a voluntary or involuntary basis, adhere to the 'zero reject' idea (Avramidis & Norwich, 2002). Teachers may be reluctant to include children with disability in their classroom (De Boer et al., 2010) and may be prone to believe that these students would receive better education in specialized education settings. Thus, teachers often hold ambivalent feelings and beliefs toward the inclusion of students with disability, in the sense of the reluctance to openly express one's own opinions and feelings on sensitive social issues. Indeed, negative attitudes toward children with disability are unacceptable in Western societies because of the very strong normative protection of the children group, in general, and of children with disability, in particular (Crandall et al., 2002;Lüke & Grosche, 2018;Rohmer & Louvet, 2018). Teachers may solve this contradiction between anti-discriminatory norms and perceived difficulties to include students with disability by overlooking discrimination when covered by benevolent justifications. The present paper seeks to specifically answer the following question: How do teachers react to prejudiced behaviors against children with disability exhibited by fellow ingroup members in their professional environment? More specifically, we aimed to show that the social acceptability of the justification provided when excluding a student with disability decreases the perception of discrimination and the social rejection of the perpetrator of the discrimination. Bastart, J., et al. (2021). Legitimating Discrimination Against Students with Disability in School: The Role of Justifications of Discriminatory Behavior. International Review of Social Psychology, 34(1): 1, 1-11. DOI: https://doi.org/10. 5334/irsp.357 The Role of Justification in Attribution to Prejudice Attribution to prejudice is about ascribing a differentialoften negative-group-based treatment to discrimination (Major & Sawyer, 2009). Social psychology literature conceptualizes prejudice and discrimination as negative intergroup behaviors. Accordingly, people attribute discrimination to perpetrators to the extent that they express a negative attitude or behavior toward an outgroup target (Jetten et al., 2013). Even if legitimate forms of discrimination may exist (e.g., child labor is forbidden by law in many countries), we are interested in forms of discrimination that are socially rejected, such as the discrimination of children with disability at school. According to the justification-suppression model of prejudice (Crandall & Eshleman, 2003), socially acceptable justifications of an unfair treatment enable the expression of genuine prejudice otherwise generally rejected. Moreover, most people believe in a just world and are reluctant to accept social injustice and thus discrimination (Hernandez et al., 2015;Ruggiero & Taylor, 1995;Sechrist et al., 2004). As a consequence, they tend to accept every legitimizing argument aimed at justifying discriminatory behaviors (Kappen & Branscombe, 2001). Indeed, a negative group-based behavior is more likely to be recognized as discrimination if rooted in unambiguous hostile intentions (Simon et al., 2019) than if perceived as acceptable in terms of social justice (Iyer et al., 2014). Research on ambivalent sexism has largely documented this effect of ambivalence: differential group-based treatments based on benevolent stereotypes go undetected more often than similar treatments based on hostile stereotypes (Glick & Fiske, 1996;Glick et al., 2000). For instance, a disparaging behavior toward a woman is less likely to be considered as sexist when perpetrated under the cover of a benevolent justification (i.e., the disparaging behavior appears as a positive solution for his/her own good) compared to a hostile one (i.e., this disparaging behavior is based on the willingness to segregate the target group; Swim et al., 2003). Benevolent justifications thus give the opportunity to legitimize discriminatory behaviors and disguise the unjust side of the unequal treatment. Therefore, discrimination may be less identified as such.
To sum up, when perpetrators legitimate their behaviors through a benevolent justification, they likely appear as good and moral persons, magnanimous toward social minorities. This kind of 'noblesse oblige' expression (Cambon & Yzerbyt, 2018;Vanbeselaere et al., 2010) implies that perpetrators comply with the anti-discrimination norm. Thus, disguising a counter-normative behavior like discrimination under praiseworthy motivations is a good way to make this behavior socially acceptable, to make a good impression, and to clear perpetrators from blame because of normative deviance. As a consequence, the legitimacy of the perpetrator's motivations appears to be a core determinant of attribution to prejudice: the more observers perceive the perpetrator's motivation as legitimate, the less they will attribute the perpetrator's behavior to prejudice.

On the Direct Consequences of Attributing Prejudice: Distancing Oneself From the Perpetrator
Discrimination is often a counter-normative behavior (Cambon & Yzerbyt, 2017;Crandall et al., 2002;Klonis et al., 2005;Plant & Devine, 1998) that is inconsistent with the prevalent egalitarian norm and thus largely socially disapproved (Czopp & Monteith, 2003;Monteith & Walters, 1998). In line with research about the black sheep effect, people are usually very sensitive to the wrongdoings of ingroup members (Marques & Yzerbyt, 1988). Because the counter-normative behavior of an ingroup member can damage the whole group's reputation and by extension their personal identity, people are motivated to symbolically distance themselves from the deviant by pushing him/her outside of the group's boundaries (Eidelman & Biernat, 2003;Eidelman et al., 2006). Witnessing an ingroup member exhibiting counter-normative behaviors elicits negative emotions like shame (Nugier et al., 2010), and witnessing discrimination elicits negative emotions, like anger (Dickter & Newton, 2013). One can expect people to experience more hostile emotions (e.g., anger, shame) after being exposed to hostile, rather than benevolent, justifications to exclude. We call this process 'symbolic distancing from the perpetrator', because it does not imply any behavioral intention toward the perpetrator.
Being exposed to counter-normative behaviors can have also tangible consequences. This was shown in both intergroup contexts and intragroup contexts (Rudman et al., 2012). People may be motivated to punish the deviant by excluding him/her from their social and professional life in a backlash effect (Rudman et al., 2012). Research on social judgments has consistently shown that, for guiding behavior, people organize social information about others following two fundamental questions: Is this person nice and trustworthy? And is this person capable and intelligent? Put it otherwise, is this person warm and/or competent? (For a synthesis, see Fiske, 2018). While one could expect a teacher to be competent, it is also expected from her/him to display relational qualities, such as being cooperative, maniagnimous, caring (Bartell, 1998), in other words warmth attributes (Fiske et al., 2007). This is even more the case when the teacher is facing vulnerable students such as those in inclusive settings. One could posit that counter-normative behaviors of a teacher (e.g., refusing the inclusion of students with disability) may lead to a backlash effect: counter-normative positions trigger less warmth attribution to this person. Attributing less warmth indicates larger social distance (Fiske et al., 2007). Accordingly, hostile justification will reduce attributions of warmth to the perpetrator, which in turn will lead to backlash against the perpetrator. We call this process 'pragmatic distancing from the perpetrator', because it implies behavioral intentions toward the perpetrator.
Although depending on different processes, the black sheep effect and the backlash effect cover quite similar functions related to the protection of social and personal selves by distancing from deviants. Because of the antidiscrimination norms, symbolic and pragmatic distancing from the deviant can be obvious when the discrimination is justified in a hostile manner. Of more interest, we advance that justifying discrimination in a benevolent manner protects to a certain extent the deviant from symbolic and pragmatic distancing. This is because the discrimination is less perceived as such. Accordingly, the more the perpetrator of an exclusion is perceived as prejudiced, the more he/she would elicit hostile emotions and the more witnesses would emphasize dissimilarities with the perpetrator in order to de-identify with him/her. In the same vein, the more the perpetrator of an exclusion is perceived as prejudiced, the less he/she would be perceived as warm and the more witnesses would express backlash in order to distance themselves from him/her. We thus expect discrimination perception to mediate the impact of justification on symbolic and pragmatic distancing.

Overview and Hypothesis
We conducted this study in an educational setting with a population of teachers facing the application of the inclusive education principle. Whereas school inclusion is considered a crucial social issue, the inclusion policy in the French educational context has mixed results. Therefore, it seemed relevant to conduct the research among a population of teachers. In accordance with actual legislation (LOI nº2005-102), unfair treatment of children on the ground of disability is banned. Law and social pressure condemn discriminatory behavior, especially toward children with disability. However, this application of the inclusion principle remains partial, and teachers do not wholly commit to their role of educators for children with disability (Avramidis & Norwich, 2002). When facing this dilemma, teachers may try to justify their reluctance to school inclusion. For this reason, we advanced that the study of benevolent justification to exclude students with disability is particularly relevant in this context. As far as we know, this was the first study to investigate these questions.
In the current French educational context, the inclusion of students with disability in mainstream classes is a normative injunction. We assumed that teachers who deviate from this norm may be perceived as prejudiced and be rejected by their fellow ingroup members. Furthermore, this would be more obvious when teachers expressed hostile justifications to exclude children with disability from their mainstream class compared to benevolent ones. First, to conceptually replicate previous findings (e.g., Swim et al., 2003), we hypothesized that teachers who exclude a student with disability will be perceived as less prejudiced when they provide a benevolent justification compared to a hostile justification (H1). Of greater importance, our second and third hypotheses concerned the participants' willingness to distance themselves from the perpetrator. Teachers will express less willingness to symbolically and pragmatically distance themselves from an excluding peer who provides a benevolent rather than a hostile justification. Accordingly, participants will report more hostile emotions (H2a) and willingness to de-identify (H2b) when exposed to a hostile perpetrator than to a benevolent perpetrator. Similarly, they will attribute less warmth (H3a) and express more backlash (H3b) when exposed to a hostile perpetrator than to a benevolent one.
Finally, to provide a test of the mechanism underlying the impact of the justifications to exclude on distancing, we hypothesized that the relation between justification and distancing will be mediated by the attribution to prejudice. More precisely, in the case of symbolic distancing, this mediation will be explained through hostile emotions toward the perpetrator (H4a). In the case of pragmatic distancing, this mediation will be explained by warmth judgements of the perpetrator (H4b). The tests of these hypotheses imply comparing an exclusion situation under the cover of a benevolent justification to an exclusion situation under the cover of a hostile justification. To check for effectiveness of the exclusion induction, we also added an inclusion situation as a third experimental condition.

Participants and Design
We determined the sample size based on small to medium effect sizes of the impact of justification on discrimination perception observed in previous research (Bastart et al., Study 3, in press;Simon et al., 2019), a between subject design with 3 conditions and 80% power. The more conservative estimation (small effect size) indicated 146 participants; whereas, the less conservative (medium effect size) indicated 93 participants. On this basis, we recruited 135 active teachers (91 women; mean age = 40.96, SD = 9.68) to participate in the study. Their mean seniority in the educational field was 15.57 years (SD = 9.53). Seventyseven percent of them were primary school teachers, and 23% were secondary school teachers. We excluded participants with over 50% of missing values (n = 6). Moreover, three participants were outliers according to the Cook's Distance on one or more of the dependent variables and were thus excluded from the sample. The final sample was composed of 126 participants who were randomly distributed across the three experimental conditions: inclusion, hostile exclusion, and benevolent exclusion.

Procedure and Materials
Participants were approached at their workplace, before or after teaching. They were told the survey concerned inclusive education and that researchers were interested in their opinions on the subject. They were asked to imagine that the situation they were going to read about was taking place in their school. After agreeing to participate, they read a vignette describing a teachers' meeting before the summer holidays. The head of school informs the team that two students with disability will be welcomed to the school the next year. Then, the team discusses the inclusion of the students with specific needs. At this point, we manipulated the nature of the teacher's opinion (Inclusion, Hostile exclusion, Benevolent exclusion).

Construction of the Material
To ensure the credibility of the scripts and to provide realistic content to the participants, we elaborated the experimental material based on the spontaneous reactions of teachers toward disability collected prior to the main study. Three graduate students participating in an internship in teaching and education reported verbatim phrases collected during informal discussions with their teaching advisors. Students were asked to scrupulously report sentences they could hear during conversations about students with disability. Sixteen sentences were reported. Considering that hostile prejudice consists in dominance, contempt, and antipathy (e.g., Cary et al., 2016 for ambivalent ageism; Glick et al., 2000 for ambivalent sexism), we used the more disparaging comments to elaborate the vignette corresponding to the hostile justification condition. In this condition, the teacher refused to include these students in his/her class justifying as it follows: 'The disabled, they couldn't progress like normal students. I think they need to be readapted before they can be considered like other students'. Taking into account that benevolent prejudice mostly consists of paternalistic attitudes (e.g., Cary et al., 2016 for ambivalent ageism; Glick et al., 2000 for ambivalent sexism), we selected the comments of teachers reluctant to inclusion for the student's own good to elaborate the vignette corresponding to the benevolent justification condition. In this condition, the teacher refused to include these students in his/her class justifying as it follows: 'I am afraid that the time and attention I will have to give to this student will be at the expense of the rest of the class. I think this student will be the only one in the class who is different and I am afraid that I will expose him/her to the mockery of others'. We used the comment expressing agreement with inclusion to elaborate the inclusion condition. In this condition, the teacher agreed to include these students in his/her class in the following manner: 'The inclusion of students with disability must be seen as any adventure in our profession. In my opinion, their inclusion can even help other students to progress'. It's worth mentioning that no information concerning gender or type of disability of the student was given. Finally, because one of our hypotheses dealt with the participant's identification with the teacher depicted in the vignette, female participants read a vignette depicting a female teacher and male participants read a vignette depicting a male teacher. We did so to avoid the control of the gender variability in the statistical analysis.

Measures
We measured attribution to prejudice on three items (e.g., Marie's/Jules' opinion is discriminatory; α = 0.88; M = 4.66, SD = 1.99). We measured hostile emotions with four items (e.g., anger, shame; α = 0.80; M = 2.77, SD = 1.59; Matheson & Anisman, 2009), and also added four filler items (e.g., surprise, amusement). To measure the participant's distance from the perpetrator, we used the inclusion of other in the self-scale (Aron et al., 1992). This pictorial single item measures the extent to which participants identify with the perpetrator. We reversed the score; thus, the higher the score, the higher the de-identification to the target (M = 5.26, SD = 1.64). Concerning the social judgment toward the perpetrator, participants rated to what extent the teacher appeared as warm on six items (e.g., likeable, honest; α = 0.81; M = 4.72, SD = 1.05; Brambilla et al., 2012), and competent on six items (e.g., competent, efficient; α = 0.84; M = 4.57, SD = 1. 19;Carrier et al., 2014). We measured the extent to which the target teacher was treated with backlash with four items adapted from Phelan et al., 2008 (e.g., it would be useful to have Marie/Jules in a work team; α = 0.89). We reversed the score for consistency reasons. Thus, the higher the score, the higher the backlash toward the target (M = 3.91, SD = 1.49). All these items were rated on a 7-point Likert scale from 1, 'Not at all', to 5, 'Totally', except for the inclusion of other in the self-item. Finally, participants answered the demographic questions, including questions regarding their experience with individuals with disability (e.g., 'Have you attended learning disability trainings?') and their proximity to disability (e.g., 'Do you have people with disability among your relatives?') and were thanked and debriefed.

Results
All analyses were conducted with OLS regressions. The independent variable had three levels and was thus broken down into two orthogonal contrasts (Brauer & McClelland, 2005). The first one compares the two exclusion conditions to the inclusion condition (C1 = hostile and benevolent condition coded -1 and the inclusion condition coded 2) and therefore tests the effectiveness of the exclusion manipulation. This test constitutes a preliminary step to ensure that the exclusion is perceived as a differential treatment compared to the inclusion condition. The second contrast compares the hostile exclusion condition to the benevolent exclusion condition (C2 = hostile exclusion coded 1; benevolent exclusion coded -1; inclusion coded 0) and therefore tests the hypotheses opposing the reactions to the two justifications for exclusions on the different measures. For each dependent variable, we reported the omnibus effect of the nature of the teacher's decision, as well as the effect of the two contrasts. Means and standard deviations corresponding to the first contrast are reported in Table 1, means and standard deviations corresponding to the second contrast are reported in Table 2.

Attribution to Prejudice
The analyses revealed an effect of the nature of the teacher's decision, F(2, 123) = 155.47, p = 0.001, h p 2 = 0.72. The first contrast checks for the exclusion manipulation, and results showed that participants perceived the teacher's behavior as more discriminatory when he/she decided to exclude the students with disability than when he/she decided to include them, F Contrast1 (1, 124) = 292.38, p = 0.001, h p 2 = 0.70. The contrast testing H1 showed that participants perceived the teacher's behavior as more discriminatory when he/she used a hostile justification to exclude the students with disability rather than a benevolent one, F Contrast2 (1, 124) = 20.45, p = 0.001, h p 2 = 0.14. These results replicated previous findings concerning the role of justification on discrimination attribution (Bastart et al., in press;Simon et al., 2019;Swim et al., 2003).

Hostile Emotions Toward the Perpetrator
Similarly, the analyses revealed an effect of the nature of the teacher's decision, F(2, 123) = 29.71, p = 0.001, h p 2 = 0.33. The first contrast checked for the exclusion manipulation, and results showed that participants expressed more hostile emotions toward the teacher when he/she decided to exclude the students with disability than when he/she decided to include them, F Contrast1 (1, 124) = 55.05, p = 0.001, h p 2 = 0.31. Concerning the test of H2a, participants reported more hostile emotions toward the teacher when he/she used a hostile justification to exclude the students with disability rather than a benevolent one, F Contrast2 (1, 124) = 4.76, p = 0.031, h p 2 = 0.04.

De-identification
The analyses revealed an effect of the nature of the teacher's decision, F(2, 123) = 62.25, p = 0.001, h p 2 = 0.50. The first contrast checked for the exclusion manipulation, and results showed that participants distanced themselves from the teacher less when he/she decided to include the students with disability than when he/she decided to exclude them (F Contrast1 (1, 124) = 117.69, p = 0.001, h p 2 = 0.49). Concerning the test of H2b, participants distanced themselves from the teacher less when he/she used a benevolent justification to exclude the students with disability rather than a hostile one, F Contrast2 (1, 124) = 7.53, p = 0.007, h p 2 = 0.06.

Warmth and Competence
Prior to the test of the hypotheses, we conducted a mixed ANOVA with the teacher's decision as the between subject factor and the judgements of competence and warmth as within subject factors. The judgements of warmth competence differed from each other, F(1, 125) = 5.08, p = 0.030, h p 2 = 0.04; warmth judgements (M = 4.72, SD = 1.05) were higher than competence judgements (M = 4.57, SD = 1.19). More important, and as expected, this main effect was qualified by an interaction with the teacher's decision, F(2, 123) =12.15, p = 0.001, h p 2 = 0.16. Analyses conducted on the warmth dimension showed an effect of the teacher's decision, F(2, 123) = 23.74, p = 0.001, h p 2 = 0.28. The first contrast checked for the exclusion manipulation: participants attribute more warmth to the teacher when he/she included the students with disability than when he/she excluded them, F Contrast1 (1, 124) = 41.74, p = 0.001, h p 2 = 0.25. Finally, the test of H3a showed that participants attributed more warmth to the teacher when he/she gave a benevolent justification to exclude the students with  disability rather than a hostile one, F Contrast2 (1, 124) = 6.14, p = 0.015, h p 2 = 0.05. The analyses conducted on the competence dimension also showed an effect of the experimental condition, F(2, 123) = 33.45, p = 0.001, h p 2 = 0.35. Participants attributed more competence to the teacher in the inclusion condition than in the exclusion conditions, F Contrast1 (1, 124) = 66.88, p = 0.001, h p 2 = 0.35, but the two exclusion conditions did not differ, F Contrast2 (1, 124) = 0.002, p = 0.965.

Backlash
Finally, the analyses revealed an effect of nature of the teacher's decision, F(2, 123) = 64.92, p = 0.001, h p 2 = 0.51. The first contrast checked for the exclusion manipulation and showed that participants backlashed against the teacher more in the exclusion conditions than in the inclusion condition, F Contrast1 (1, 124) = 104.22, p = 0.001, h p 2 = 0.46. The test of H3b showed that participants backlashed the teacher more when he/she used a hostile justification to exclude the students with disability rather than a benevolent one, F Contrast2 (1, 124) = 4.08, p = 0.046, h p 2 = 0.03.

Mediation Analyses
The aim of this research was to study the consequences of excluding a student with disability depending on the justification provided by the teacher. We predicted that exclusion will lead to more prejudice attribution when justified in a hostile manner rather than a benevolent one and thus to more distancing from the teacher, both pragmatically and symbolically. In accordance with these hypotheses, we conducted two separated mediation analyses in which we compared the hostile justification condition to the benevolent justification condition (i.e., C2 = hostile justification coded 1, benevolent justification coded -1, and inclusion coded 0). In these mediation analyses, the second contrast was entered as a covariate (C1 = hostile and benevolent condition coded -1 and the inclusion condition coded 2).

Symbolic Distancing
We reasoned that the teacher who provided hostile justification to exclude the student with disability will be perceived as more deviant than the teacher who provided benevolent justification. This perception of deviance will lead to a greater willingness to symbolically distance from the teacher, that is, to greater willingness to de-identify with the teacher. According to research dealing with the black sheep effect, ingroup deviants elicit hostile emotions and a desire to de-identify with the black sheep (Nugier et al., 2010). To test the hypothesis that attribution to prejudice and hostile emotions toward the teacher act as serial mediators of the relationship between the justification provided by the teacher and the willingness to distance oneself from the teacher, we used the Process macro (Model 6, 5000 percentile bootstrap; Hayes, 2014). This indirect effect was significant, b = 0.050, 95% CI [0.007, 0.120]. The indirect effect of the justification on distancing from the teacher through attribution to prejudice alone was significant, b = 0.231, 95% CI [0.057, 0.426]; whereas, the indirect effect of the justification on distancing through hostile emotions alone was not, b = 0.006, 95% CI [-0.058, 0.068]. Accordingly, this mediation supported our hypothesis of symbolic distancing. The full mediation model is displayed in Figure 1.

Pragmatic Distancing
The perception of deviance will lead to greater willingness to pragmatically distance from the teacher, that is, to a lesser willingness to share professional responsibilities and social bonds with the perpetrator. Research dealing with the backlash effect suggests that social and economic penalties go along with lesser attribution of warmth to the target (Fiske et al., 2007). We thus hypothesized that the relationship between the justification provided and backlash against the teacher is serially mediated by attribution to prejudice and warmth attribution to the teacher. The indirect effect of the justification on backlash through discrimination attribution and warmth attribution to the teacher was significant, b = 0.08, 95% CI [0.015, 0.164]. We also tested the simple mediations. The indirect effect of the justification on backlash though attribution to prejudice alone was significant, b = 0.192, 95% CI [0.056, 0.334]; whereas, the indirect effect of the justification on backlash through attribution of warmth alone was not, b = 0.026, 95% CI [-0.047, 0.115]. Results of this mediation analysis also supported our hypothesis concerning pragmatic distancing. The full mediation model is displayed in Figure 2.
To confirm that the sample size allows detecting the individual paths in the mediation models, we conducted sensitivity analysis with 80% power (i.e., a path from C2 to Mediator 1 controlled for C1, d path from Mediator 1 to Mediator 2 controlled for C1 and C2, and b path from Mediator 2 to the DV controlled for C1, C2, and for Mediator 1). For the a path, the analysis revealed a critical h 2 of 0.13. The a path is identical for the two mediation models and is above this critical value (h p 2 = 0.14) For the d paths, the analysis revealed a critical h 2 of 0.06. Both d paths are above this critical value (h p 2 = 0.18 and 0.27 for symbolic and pragmatic distancing, respectively). For the b paths, the analysis revealed a critical h 2 of 0.06. The b path of the pragmatic distancing mediation analysis is above this critical value (h p 2 = 0.14); whereas, the b path of the symbolic distancing mediation analysis is slightly below this critical value (h p 2 = 0.04). The mediation analysis concerning the symbolic distancing should be interpreted with caution. However, the replication of the distancing process on pragmatic distancing is consistent with our hypotheses.

Additional Analyses
We also asked participants if they were a disabled individual or if one of their relatives was, if they had already taught disabled students, and if they had attended training on disability in the classroom.
Only six participants reported being disabled. Consequently, we conducted the same analyses without these participants. All the effects remained significant (all p < 0.001).
Concerning the second measure, 40 participants indicated having a relative with disability. We thus chose to include this factor as a moderator in our models. No main effect of this factor, or interaction with the nature of the teacher's decision, emerged (all p > 0.343). Moreover, the main effects of the nature of the teacher's decision remained significant (all p < 0.001).
Regarding the third measure, 101 participants reported having taught disabled students. Given this high proportion, we did not conduct any supplemental analyses including this factor.
On the last measure, 59 participants reported having attended disability training. We thus choose to include this factor as a moderator in our models. No main effects or interactions emerged (all p > 0.112). Moreover, the main effects of the nature of the teacher's decision remained significant (all p < 0.001).

Discussion
In this research, we investigated the teachers' reactions to discriminatory behaviors against children with disability in their professional environment. We argued that teachers will be less likely to overlook discrimination and to symbolically and pragmatically distance themselves from the perpetrator when he/she provides a benevolent justification to exclude a student with disability rather than a hostile justification. The first hypothesis dealt with the conceptual replication of the effect of justification on prejudice attribution to ableism, while the second and third hypotheses dealt with the distancing reactions toward a prejudiced ingroup member. The fourth hypothesis concerned mediation processes. We expected that the relation between justification and symbolic and pragmatic distancing will be mediated by the attribution to prejudice. To test these hypotheses, we conducted an experiment in a school setting, which usually protects against discrimination (Lüke & Grosche, 2018).
Concerning the first hypothesis, our results replicated previous findings. Swim et al. (2003) originally showed that benevolent justifications to exclude women lead to less attribution to sexism compared to hostile justifications. These findings were recently replicated on sexism and racism (Bastart et al., in press;Simon et al., 2019). Indeed, legitimizing unequal treatment is likely to reduce discrimination attribution (Jetten et al., 2013).
Concerning the second and third hypotheses, we investigated the different reactions of the participants toward a fellow ingroup member depending on his/her compliance to accept a student with disability in his/her class. Because the rejection of counter-normative individuals can take several forms, we addressed this question by considering both the symbolic aspect and the pragmatic aspect of distancing. We hypothesized that benevolent justifications would lead to less symbolic and pragmatic distancing than hostile ones. Indeed, concerning symbolic distancing, the results showed that participants expressed fewer hostile emotions and de-identified with the perpetrator more when he/she expressed a benevolent rather than a hostile justification to exclude the student with disability. The results on pragmatic distancing mirror those obtained on symbolic distancing: participants attributed more warmth to the perpetrator and less backlash when he/she used a benevolent, rather than a hostile justification to exclude the student with disability. The fact that we replicated the expected pattern of results on both aspects of distancing sustains the reliability of our findings. These results indicated that participants perceived the two exclusion situations as discriminatory (both means are above the middle of the scale: 6.25 for hostile justification, 5.22 for benevolent justification) but granted the benevolent perpetrator the benefit of the doubt. Moreover, results showed unexpected valorization of competence in the case of inclusive teachers. Past research indicated that exclusion behaviors of teachers toward students with disabilities are related to their feelings of professional incompetence for managing diversity in the classroom (Bukvic, 2014;Subban & Sharma, 2006). Consequently, they could recognize specific skills of their peers who meet this challenge.
Finally, concerning the fourth hypothesis, results of the mediation analyses revealed that attribution to prejudice account for the relationship between the justification manipulation and both symbolic and pragmatic distancing. Taken together, the results suggest that participants expressed more willingness to distance from the hostile perpetrator compared to the benevolent one because they perceived the hostile perpetrator as more prejudiced than the benevolent one.

Implications
Our results showed that providing a benevolent justification to exclude a student with disability from mainstream class leads to less attribution to discrimination than providing a hostile justification. Therefore, it can be argued that ' ableism' mirrors sexism and racism (Goodley, 2014). However, while extending the justification effect to ableism, these findings go beyond previous research by showing that the perception of less prototypical instances of discrimination (i.e., on the grounds of disability) can also be impacted by the perpetrator's motivations. Individuals with disability are a strongly protected group (Crandall et al., 2002), especially when individuals with disability are children (Lüke & Grosche, 2018). Besides, one could expect teachers to be specifically vigilant to discriminating behaviors. However, it appears here that, in educational settings, justifying the exclusion of students with disability with a benevolent justification partially legitimates discrimination, making resistance to inclusion less questionable. This is plausible because of the real context experimental design based on a set of teachers' speech samples collected in their workplace and inserted in realistic situations. Participants therefore might have felt safe to overtly express their reactions to the situation depicted in the script.
This research also has implications with respect to the application of diversity policies on disability, especially regarding children with disability at school. Whereas in many countries these children are increasingly educated in mainstream settings through inclusive practices meant to prevent social exclusion, the effective application of this policy faces barriers (Ebersold et al., 2016). Despite consistent evidence of benefits induced by inclusive education, many students with disability struggle to access effective inclusive practices. Being less attentive to exclusion of these children because of benevolent justification can help to explain why protective legislation is only partially applied in schools. Moreover, teachers, socially perceived as humanistic and protective figures, may also be blamed if they exclude children without socially acceptable reasons. Nevertheless, a benevolent justification from the teachers' side condones discrimination even when teachers differentiate this specific situation from inclusion. Consequently, it makes discrimination subtle, more acceptable and the victims' complaints less obvious. In sum, even in environments that are highly normative and protective against exclusion, such as schools, the inclusion of children with disability cannot be taken for granted.

Limitations
Despite these promising findings, some limitations are to be addressed in future research. This single study research is an original work. We conceptually replicated the effect of the justifications provided on attribution to discrimination, but this replication concerned a less studied instance of discrimination: ableism. Direct replications of our results are needed to ensure the reliability of the effect of justifications on attribution to ableism. Besides, to investigate the social acceptability of the discrimination's justification, we relied on the participants' willingness to distance themselves from the target. Further research is needed to capture social acceptability with a different operationalization. Moreover, the vignettes we used in the study depicted the inclusion of a student with disability, without specifying the type of disability in question. Specifying the nature of the impairment may moderate the impact of the justifications. Indeed, more reluctance is expressed when including students with mental or psychological disabilities in mainstream settings compared to physical disabilities (Avramidis & Norwich, 2002). One can expect people to justify the exclusion of students with mental disability to a greater extent than those with physical disability.

Conclusion
The present paper constitutes a valuable contribution to the growing literature on the legitimation of discrimination of social minorities, and a better understanding of how these justifications might be used to excuse unfair treatments at school. More specifically, applied to inclu-sive education, the under-detection of subtle forms of discrimination can constitute a significant obstacle to the laudable efforts for inclusion policies. Making teachers aware of the possibility that benevolent intentions may hide unfair treatment is a relevant path to fight ideological barriers to a full application of the inclusion principle and its success.

Data Accessibility Statement
The data and the material can be found at https://osf. io/argcn/?view_only=5c4107ec89af4e3da799cd125 10f5809.