Flexible Solidarity with Refugees: Integrating Minority Influence and Intergroup Communication

Integrating Minority Influence and Intergroup Communication More than one million migrants and refugees crossed into Europe in 2015 (International Organization for Migration, 2015), causing a crisis throughout the continent and dividing opinions over how to deal with resettling refugees. The massive arrival of people fleeing from persecution and war exhibited sharp cultural rifts, readily mobilized by right-wing parties in order to promote restrictive policies. As ideological boundaries have risen, public opinion in Europe has split into those who support solidarity with refugees and those who reject it. We believe that minority influence plays a role in the solidarity European citizens may express. Many volunteer movements and organizations are stepping up to assist in that role by bypassing official institutions, contesting authorities and asking for the participation of citizens (Gunter, 2015). In this paper, we study how a minority appeal can convince reticent targets to show solidarity with refugees. We thereby hope to advance the theoretical debate around minority influence by proposing a conceptual integration of two alternative approaches: self-categorization and conflict elaboration theories. Running an experimental study based on Gabriel Mugny’s and his colleagues’ research on minority influence, we will show how his ideas remain relevant today and how they relate to our work on intergroup communication and the social order representations model (SORM; see Staerklé, Delay, Gianettoni & Roux, 2007; Staerklé, 2009, 2016). Moving away from a static understanding of social categories, we propose a dynamic and communicative approach to intergroup relations and show that outgroup minorities can exert manifest influence under specific conditions. Based on early insights by Doise (1976), we suggest that intergroup and intragroup forms of differentiation interact during the influence process and that minority influence is enhanced through communication strategies that break the representation of groups as carrying alternative and incompatible normative positioning.

elicit identity-based resistance among the majority of nationals, since favouring refugees automatically means betraying national interests (Mugny, Kaiser, Papastamou & Perez, 1984;Sanchez-Mazas, Mugny & Falomir-Pichastor, 1997).In this respect, compliance with the authority and preference for strict legislations can be interpreted as a self-defensive form of ingroup favouritism.What matters is therefore the subjective experience of the targets: For them, the authority is legitimate so long as it represents the will and the interests of the people (Turner, 2005).
Minority influence thus takes place within a tripartite context of power relations involving a minority group in an antagonistic relationship with the authority and seeking to mobilize the support of the majority (Mugny, 1982).By breaking the link between the authority and the majority, minorities gain influence (Mugny, 1982;Mugny & Papastamou, 1976;Mugny & Pérez, 1991).In other words, minority influence is possible by way of a self-categorization process of the majority 'that ultimately redefines the authority as outgroup and the minority as ingroup' (Subašić, Reynolds & Turner, 2008, p. 331; see also Turner, 1991).
Whether the re-categorization of the social field is possible or not depends on the social distance between the source and target of influence, meaning whether or not the majority and minority share the same group membership (Mugny & Papastamou, 1982a).In this respect, theories of self-categorization (Turner et al., 1987) and conflict elaboration (Pérez & Mugny, 1993) lead to very different predictions (Gardikiotis, 2011;Pérez & Mugny, 1998).
Nevertheless, evidence is not always in line with these predictions, especially at the latent level (Pérez & Mugny, 1987;Souchet, Tafani, Codaccioni & Mugny, 2006).Furthermore, ingroup minorities are particularly effective in convincing the majority as long as they argue in favour of ingroup vested interest and provide arguments consistent with its norms.These so-called ' orthodox minorities' (Moscovici, 1976) polarize the audience and move targets towards more radical opinions (Abrams & Hogg, 1990;David & Turner, 1999).
What about an ingroup minority arguing in favour of outgroup vested interest, for example a social movement defending the rights of refugees?Together with their stance against authority, they propose a counter-norm-solidarity-implying a negative outcome for the ingroup (Subašić et al., 2008).These 'heterodox minorities' (Moscovici, 1976) introduce a new point of view and thus trigger social and cognitive conflicts, possibly stimulating divergent and creative thinking (Nemeth, 2009).Therefore, conflicting elaboration of alternative solutions is crucial for heterodox minorities (Mugny & Pérez, 1991;Pérez & Mugny, 1993).
The role and the nature of such conflicts differentiate conflict elaboration from self-categorization theories.Conflict elaboration theory argues that targets of influence wish to avoid common membership with the minority source.An ingroup minority source may therefore arouse greater resistance and differentiation than an outgroup source (Mugny & Pérez, 1989, 1998).This hypothesis puts conflict elaboration theory ' at odds with the self-categorization theory of influence, the basic tenet of which is that all positive influence is based on shared social identity' (David & Turner, 1999, p. 116).People expect to agree with ingroup members and to disagree with outgroup members.It is precisely the unexpected disagreement with ingroup members that provokes uncertainty and may elicit attitude change.Since no agreement is expected with outgroup members, no attitude change should occur (Turner, 1991;Turner & Oakes, 1986).
Quiamzade and colleagues (Quiamzade, Pérez, Mugny & Alonso, 2003) present a theoretical reconciliation of these two alternative points of view.Focusing on attitudes towards gypsies in Spain, they manipulated the number of characteristics the minority source had in common with participants.Several sociodemographic characteristics were provided, the source being presented as either a man or a woman, young or old, and from Valencia or from Catalonia.Different sets of commonalities were then made salient: Subjects shared either no characteristics with the source (' complete outgroup' condition), one characteristic ('hybrid outgroup' condition), two out of three characteristics ('hybrid-ingroup' condition), or all three (' complete ingroup' condition).The authors showed the hybrid outgroup minority to be the most effective in creating influence.Cross-categorization (Crisp, Ensari, Hewstone & Miller, 2003;Deschamps & Doise, 1978) was therefore found to be a necessary condition.

Communication Strategies for Reducing Intergroup Differentiation
Much of the literature discussed so far has approached social categories uniquely in terms of stable and ascribed group memberships, which precede the influence setting and remain unchanged after minority intervention.However, the way we juxtapose people with social categories and the bases upon which we treat these categories are in constant motion (Reicher, 2004).Ascribed identities influence group processes, but in the same way, intergroup communication and social interaction influence identities and the way they are understood by people (Mugny & Papastamou, 1982b;Postmes, Haslam & Swaab, 2005; see also Doise, 1976).
The communicational style minorities strategically adopt during the process of influence determines whether they are willing or not to become ' entrepreneurs of identities' (Reicher & Hopkins, 2001).This means that the identification process is not complete with the mere categorization of the source as ingroup or outgroup.By informing the audience about the normative positioning of the source within the social field, the attributes and representations contextually made salient during social interaction equally determine if the target identifies with or rejects the minority (Butera & Pérez, 1995;Hogg & Turner, 1987;Mugny, Kaiser & Papastamou, 1983).
The unexpected disagreement with those categorized as ingroup members has been shown to stimulate uncertainty and induce influence (Turner, 1981(Turner, , 1991)).The unexpected agreement with those categorized as outgroup members should work in the same way.Since we expect groups to behave in line with our representations (Doise, 1976), minorities that disrupt these expectations and exploit their 'strangeness' manage to be particularly influential (Roux, 1995).This was the case in a study conducted by Roux (1991) on attitudes towards refugees in Switzerland.Only the radical left-wing source using accommodating arguments obtained influence.The mismatch between source and arguments led most xenophobic people to improve their general attitudes towards foreigners.
This hypothesis was also applied in a study on attitudes towards immigration policies in Switzerland: Mugny, Kaiser and Papastamou (1983) simultaneously manipulated a) the categorization of the minority source as Swiss (ingroup) or foreign (outgroup), and b) the arguments mobilized in the appeal as humanitarian or socialist.Results partially confirmed the hypothesis: Although the ingroup source exerted more influence compared to the outgroup source, they also found a significant interaction between source and argument, showing that foreigners benefitted from humanitarian arguments whereas Swiss benefitted from socialist ones.
Mugny and colleagues (Mugny et al., 1983;Perez & Mugny, 1985) argued that foreigners established a normative connection with Swiss respondents by invoking the humanitarian traditions prototypically associated with Switzerland.Reference to class inequalities, in turn, was considered to be associated with foreigners.Nevertheless, this hypothesis was not explicitly tested and, to the best of our knowledge, no empirical studies on minority influence in intergroup settings have ever orthogonally crossed source and message in order to yield a clear ingroup vs. outgroup comparison on both dimensions.
Thirty years later, we come back to these ideas, proposing that processes of intergroup influence originate from the strategic mobilization and communication of models of social order as suggested by the SORM model (Staerklé, 2009;Staerklé et al., 2007).The model was initially developed to account for the structure of ideological attitudes, hypothesised as reflecting different socio-cognitive processes of categorisation (normative vs. categorical) and different comparison dimension content (identity-based vs. material). 1In the present study we are concerned with the interaction between social influence and categorisation, and therefore focus only on one aspect of the model, namely the distinction between categorical and normative differentiation.Categorical differentiation operates at the intergroup level and represents individuals as members of distinct subgroups within the social setting.In the present study, this form of differentiation will be operationalised with group membership of the source of influence (ingroup vs. outgroup member).Normative differentiation operates at the intragroup level and classifies group members as a function of their proximity to (or conformity with) the group prototype.This factor will be manipulated through the influence message content (ingroup or outgroup prototypicality of the message).

Reduced Social Differentiation Generates Influence
According to the SORM, categorical and normative differentiation may act in parallel and bolster each other: Prototypes are used to define categories as different from each other (Staerklé, 2016).Hence, by crossing the two levels of differentiation, we can describe the most favourable conditions for minority influence to occur.As mentioned above, the best representative of the ingroup prototype is the authority (Mugny, 1982;Turner, 2005).An ingroup minority using prototypical ingroup arguments fails to distinguish itself from the authority and cannot propose a divergent point of view.An ingroup minority using prototypical outgroup arguments, instead, breaks connections with the authority, reduces intergroup differentiation and provokes an unexpected conflict among fellow ingroup members (Turner, 1981(Turner, , 1991)).We expect ingroup minorities to induce manifest influence only under the second condition.
The same line of reasoning applies to outgroup minorities.The theoretical distinction between 'representative outgroup minorities' and ' dissident outgroup minorities' proposed by Volpato et al. (1990) supports our predictions.Representative outgroup minorities use prototypical outgroup arguments.Despite their minority position in the intergroup setting, they represent the majority of their fellow members.Moreover, these minorities are perceived as having the strongest vested interest since they argue for a positive outcome for their group and they use arguments that are in line with these same interests.For this reason, representative outgroup minorities are often rejected (Maass & Clark, 1984;Maass, Clark & Haberkorn, 1982).
Dissident outgroup minorities use prototypical ingroup arguments.While seeking a positive outcome for their group, they use arguments that are perceived to go against their own vested interest.Indeed, these minorities are ' double' minorities that try to create connections with their targets all the while occupying a peripheral position in their own group.We therefore expect only dissident outgroup minorities to induce manifest influence, and not representative outgroup minorities.

Sensitivity to Intergroup Differentiation for Strong Identifiers
Our rationale assumes targets to be personally concerned by the differentiation between ingroup and outgroup sources.Since the intergroup context of the present study involves a national ingroup vs. a refugee outgroup, high national identification should increase normativity of ingroup norms and perceived threat from outgroup sources.Indeed, strong identifiers are more likely to be loyal to their group and attentive to possible threats coming from outgroup members (Verkuyten, 2009).Not only do they attribute more importance to categorical differentiation, but they are also more sensitive to the normative content and the prototypical behaviour ascribed to each category (Hogg & Turner, 1987).Since strong identifiers endorse ingroup norms to a greater extent than weak identifiers, they should therefore be particularly sensitive to dissident outgroup minorities endorsing ingroup norms (Sanchez-Mazas et al., 1997;Volpato et al., 1990).
To summarize, our main hypothesis is that categorical differentiation (i.e. the minority source categorized as ingroup or outgroup) and normative differentiation (i.e. the argument used as prototypical of one or the other category) must interact in order to generate influence.Our second complementary hypothesis is that the interactive effect of categorical and normative differentiation will be stronger for strong identifiers.

The Present Study
In order to test these hypotheses, we present a study based on the experimental paradigm developed by Mugny, Kaiser and Papastamou (1983) in which we also integrate insights from the SORM (Staerklé, 2016).We simultaneously manipulated a) the categorization of the minority source as Swiss (ingroup) or refugee (outgroup), and b) the arguments mobilized in the appeal as assimilationist (ingroup normative) or multicultural (outgroup normative).Assimilationist arguments propose that foreigners, migrants and refugees should let go of their cultural heritage in order to favour the way of life in the host country.Alternatively, multicultural arguments encourage maintenance of cultural differences.
We turned to these two lines of arguments since strategies for integrating refugees are currently at the centre of controversy with alternative ideologies supporting very different solutions.Nationals are generally more in favour of assimilation and foreigners prefer multiculturalism (Callens, Valentová & Meuleman, 2014;Wolsko, Park & Judd, 2006).The first is therefore normative for ingroup members and is most often associated with maintenance of the status quo.The second is normative for outgroup members and generally supports social change (Green & Staerklé, 2013;Politi & Staerklé, 2017).Contrary to previous findings, we predict that neither of the two norms made salient in the pro-welcoming appeal will predict solidarity directly (Guimond et al. 2013).Instead, we expect each norm to show differential effects depending on the source of the message.

Participants
Participants included 125 students in an introductory social psychology course at the University of Lausanne.Only participants with Swiss nationality were retained for analyses (90% of the original sample, N = 112).Age ranged from 18 to 44 (M = 21.00,SD = 3.72), although 97% of participants were under 26 years old.A majority of participants were women (73%, N = 82).

Procedure and Materials
During class time, participants filled out a questionnaire comprised of an experimental manipulation and the measures described below.

Pre-test measures
Prior to reading the experimental manipulation, participants responded to items concerning their opinion towards refugee and immigration policies, their opinion towards the European Union, as well as their degree of identification with Switzerland.Questions were answered on a six-point Likert scale ranging from not at all in agreement (1) to completely in agreement (6).
Solidarity with refugees was assessed with two items (extracted from the European Social Survey 2002) and was used as a pre-test measure: 'The government should be generous in judging people's applications for refugee status', and 'While their cases are being considered, the Swiss government should give financial support to asylum seekers'.Items were strongly correlated (r = 0.51) and were averaged to create a single score.Questions regarding opinions towards immigration policies and the European Union were included as distractors from the pre-test, as the same two items would appear a second time following the manipulation.
Identification with Switzerland was measured with four items, for example 'I identify strongly with Switzerland'.Internal consistency of the scale was good (α = 0.85).

Experimental conditions
Each participant was then randomly attributed to one of five conditions, according to a 2 (source) × 2 (message) experimental design, plus a control condition.In each of the four main conditions, participants read a statement about the current Syrian refugee situation in Europe and in Switzerland more specifically.No statement was included in the control condition.
According to the statement, Switzerland would soon need to take position and declare if and how it will participate in the allocation of refugees throughout Europe.A political appeal in favour of receiving refugees in Switzerland was then introduced.For the first manipulation (source), it was stated that the appeal was made either by a Swiss minority activist or by a refugee minority activist currently residing in Switzerland.For the remainder of the manipulation, in order to make the intergroup setting more salient, the Swiss activist referred to the audience using inclusive pronouns (e.g., we, our country, etc.), while the refugee activist used exclusive pronouns (e.g., you, your country, etc.).
The source then stated that Switzerland must respond to this call for humanitarian aid, following which a description was provided regarding how refugees should be accepted and integrated into the country.For the second manipulation (message), the proposed integration approach was either multicultural, advocating maintenance of traditions, customs and languages of origin, or assimilationist, advocating adoption of Swiss traditions, customs and languages.Both appeals concluded by indicating that such an approach would contribute to the harmonious development of Switzerland.Table 1 shows the distribution of participants according to the five conditions.

Manipulation checks
Following the experimental manipulation, four sets of questions were presented for manipulation checks: Evaluation of the text, understanding of the text, and perceived Swiss and refugee agreement with the appeal. 2 Evaluation of the text was measured using five items on a five-point scale ranging from 1 (easy to follow, obscure, consensual, extreme, interesting) to 5 (difficult to follow, clear, controversial, moderate, uninteresting).
Understanding of the text was verified using six items.Participants were asked to indicate the degree to which the six statements regarding the text were true, on a sixpoint scale ranging from 1 (not at all true) to 6 ( completely true).The first two suggested the appeal stated that Switzerland should or should not help the refugees (one worded negatively; no significant correlation; items kept separate).Two others suggested the appeal supported multicultural integration (r = 0.55; averaged to create a single score), and the final two suggested the appeal supported assimilationist integration of refugees (r = 0.79; averaged to create a single score).
Perceived Swiss agreement with the appeal was assessed by three items (α = 0.84), for example, 'in your opinion, how many Swiss citizens would agree with the appeal?'.All items were coded on a five-point scale ranging from 1 (very few/not at all) to 5 (most/completely).
Perceived refugee agreement was measured with three items (α = 0.83) worded similarly to the previous three, but referring to refugees living in Switzerland, instead of Swiss citizens.

Dependent measures
Participants then responded to three sets of questions on assimilationist, multicultural and refugee policy attitudes.These measures were also present in the control condition.
Agreement with assimilationist principles was assessed using seven items (α = 0.79), three of which were translated from Wolsko et al. (2006), for example, 'In order to have a smoothly functioning society, members of cultural minorities must adapt better to Swiss norms', and four of which were extracted from Badea (2012), for example, 'Immigrants should adopt Swiss customs and traditions'.Items were coded on a six-point scale ranging from 1 (not at all in agreement) to 6 (completely in agreement).
Agreement with multicultural principles was assessed using six items (α = 0.82) translated from Wolsko et al. (2006), for example, 'We must appreciate the unique characteristics of different ethnic groups in order to have a harmonious Swiss society'.Responses were coded on the same six-point scale described above.
Solidarity with refugees was measured using four items (α = 0.76), all extracted from the European Social Survey (2002).Two of these items were also present in the pre-test.The same response scale was used.
Finally, participants indicated basic sociodemographic information.

Results
Data analysis was carried out with Statistical Package for the Social Science (SPSS), version 22. Descriptive statistics and correlations were assessed first, followed by manipulation checks.Finally, hypotheses were tested.Gender was excluded from analyses as no differences were found between men and women on any measures in the study.

Descriptive Statistics and Correlations
Table 2 shows means and standard deviations of main measures in the questionnaire which did not make reference to the appeal, as well as binary correlations for all participants regardless of experimental conditions.In general, participants had reasonably high levels of national identity.They also tended to be in support of refugee policies and multicultural principles.Their support for assimilationist principles was also positive, although less so.
According to correlations, the more participants identified with Switzerland, the more they supported assimilationist principles, and the less they supported multicultural ones as well as solidarity with refugees.Indeed, greater support for assimilationist principles was associated with less solidarity with refugees, while greater support for multicultural principles was associated with more solidarity.Finally, solidarity with refugees measured in the pre-test was significantly associated with similar solidarity measured later on.
In sum and of central importance to the present study, participants were generally quite supportive of principles and policies in favour of refugees, especially when they had lower levels of national identity.When they had higher levels of national identity, this support decreased.

Manipulation Checks
Evaluation and understanding of the text Manipulation checks showed that participants evaluated and understood the respective appeals as expected.For the questions on evaluation of the text, a two-way Multivariate Analysis of the Variance (MANOVA) showed only one significant difference between conditions: The refugee source (M = 1.88,SE = 0.14) was considered easier to follow than the Swiss source (M = 1.46,SE = 0.14), F(1, 88) = 4.51, p < 0.05, although both evaluations were on the ' easy' side of the scale.The four texts were otherwise perceived to be clear, extreme, controversial and interesting to similar degrees.Understanding of the text was verified using another two-way MANOVA.For the first two questions, regarding Switzerland's responsibility to help refugees, no significant differences between conditions were found: In line with our expectations, all four appeals were interpreted to call for Swiss aid for refugees to a similar degree.For items indicating the appeal had multicultural content, univariate tests with estimated marginal means showed a significant effect of the message, F(1, 88) = 183.50,p < 0.001: As expected, the multicultural message (M = 5.30, SE = 0.15) had significantly higher scores than the assimilationist message (M = 2.35, SE = 0.16).For the assimilationist questions, univariate tests again showed the expected effect of the message, F(1, 88) = 200.70,p < 0.001: The assimilationist message (M = 5.29, SE = 0.17) had significantly higher scores than the multicultural message (M = 2.05, SE = 0.16).Participants therefore successfully recognized when the message was multicultural or assimilationist.

Perceived outgroup and ingroup normativity of the message
Before testing our hypotheses, the normativity of the two messages was verified to ensure that assimilationism was perceived to be more in line with Swiss (ingroup) norms and that multiculturalism was perceived to be more in line with refugee (outgroup) ones.A 2 (message) × 2 (source) MANOVA, predicting the belief that Swiss (R adj 2 = 0.03) and refugees (R adj 2 = 0.10) would agree with the appeal, was carried out to test normativity of each message.Results showed a significant main effect of the message, F(2, 86) = 6.42, p < 0.01, a marginal main effect of the source, F(2, 86) = 2.55, p = 0.08, and a non-significant interaction between the two conditions, F(2, 86) = 0.64, p = 0.53.
With regards to the main effect of the message, between-subject effects showed that Swiss citizens were perceived to agree with the appeal significantly more when the source endorsed assimilationism (M = 2.42, SE = 0.12) than when he/she endorsed multiculturalism (M = 2.01, SE = 0.11), F(1, 87) = 6.17, p < 0.05.Conversely, refugees were perceived to agree with the appeal significantly more when the source endorsed multiculturalism (M = 3.82, SE = 0.11) than when he/she endorsed assimilationism (M = 3.45, SE = 0.12), F(1, 87) = 5.30, p < 0.05.These results confirm that the assimilationist message was perceived to be more normative for Swiss citizens (the ingroup), whereas the multicultural message was perceived to be more normative for refugees (the outgroup).
With regards to the marginal main effect of the source, between-subject effects showed that refugees were perceived to agree with the appeal significantly more when the source was a refugee (M = 3.82, SE = 0.12) than when the source was Swiss (M = 3.45, SE = 0.11), F(1, 87) = 5.16, p < 0.05.Swiss agreement with the appeal, however, was perceived to be similar regardless of the source of the message, F(1, 87) = 0.10, p = 0.76.

Hypothesis Testing
Normative and categorical differentiation interact to generate influence Table 3 shows results from a two-step hierarchical regression with solidarity with refugees as the outcome variable.
Step One tests our first hypothesis, introducing the four main experimental conditions (source × message) as predictors and controlling for solidarity with refugees prior to the experimental condition.Step Two incorporates national identification as well as the remaining interactions, and tests our second hypothesis.
Neither categorical differentiation (source condition), nor normative differentiation (message condition) showed significant main effects.However, a significant interaction between the two conditions was found, in line with our first hypothesis.Simple effects for the interaction in Step One showed that when the source of the appeal was an ingroup member (i.e.Swiss), the multicultural message was significantly more effective in increasing solidarity with refugees than the assimilationist message, t(83) = 2.12, p < 0.05.When the source was an outgroup member (i.e.refugee), no significant differences were found, although means were in the expected direction, t(83) = -1.22,p = 0.22.Similarly, when analysing simple effects for each message, the multicultural message had significantly more influence in the ingroup (i.e.Swiss) source condition, t(83) = -2.10,p < 0.05, while no significant differences were found for the assimilationist message, even though means were, again, in the expected direction, t(83) = 1.25, p = 0.21.In Step Two of the model, no main effect of national identity was found.Results did however show a significant interaction effect between normativity of the message and national identity.When national identity was lower, a multicultural message resulted in marginally more solidarity with refugees than an assimilationist one, t(79) = 1.82, p = 0.07.When national identity was higher, no significant difference between messages was found, t(79) = -1.12,p = 0.27.These effects were qualified by a significant three-way interaction between national identity, normativity of the message and categorization of the source (see Figure 1).The two-way interaction between source and message described above was present for strong identifiers with Switzerland, t(79) = -3.36,p < 0.01, but not for weak identifiers, t(79) = 0.16, p = 0.87.
Comparison with the control condition In order to complete previous findings, we tested the effectiveness of the four experimental conditions compared to the control condition.Four dummy variables were created in order to capture the difference between each experimental condition and the control condition, which was used as the reference category.The four dummy variables were then multiplied with national identity in order to construct interaction terms. 3This procedure enabled testing differential effects of each appeal contrasted with the control condition for both high levels (+1 SD) and low levels (-1 SD) of national identification.
The full model regressing solidarity with refugees against all previous parameters was significant, F(9, 100) = 2.57, p = 0.01, R adj 2 = 0.12.The overall test of the omnibus interaction effect between national identity and conditions was also significant, ΔF = 2.57, p = 0.04.For weak identifiers (-1 SD), simple effects showed no significant mean differences between conditions, meaning that none of the four appeals resulted in significantly more solidarity than the control condition.For strong identifiers (+1 SD), both the Swiss source endorsing multiculturalism and the refugee source endorsing assimilationism resulted in significantly more solidarity than the control condition (see Table 4).

Discussion
Overall, results confirmed our main hypotheses.The simple categorization of a minority source as an ingroup member was neither a necessary nor a sufficient condition of influence.Nonetheless, we demonstrated that categorical differentiation as ingroup or outgroup can set limits on the message content of a source: An ingroup source (i.e.Swiss) benefited from using outgroup normative (i.e.multicultural) arguments, whereas an outgroup source (i.e.refugee) benefited from using ingroup normative (i.e.assimilationist) arguments.In this respect, the representations conveyed during the communication process disclosed an anticipatory function: 'Representations, by improving a certain image of the other group, attributing certain kinds of motive to it, paved the way for action with respect to that group' (Doise, 1976).
The second set of hypotheses concerning national identification was also confirmed.Binary correlations showed that those who identified more strongly with Switzerland also tended to be more strongly opposed to pro-refugee policies.Furthermore, appeals resulted in different reactions depending on the level of national identification.For weak identifiers, results revealed no interaction between source and content of the appeal: The multicultural message was marginally more effective than the assimilationist one regardless of the source.For strong identifiers, results did reveal the expected interaction between message and source: Those who were more sensitive to intergroup differentiation were more influenced by communication strategies that increased similarities between the ingroup and the outgroup.
A main limitation of the present study is that influence was measured only at the manifest level, where solidarity with refugees was directly connected with the minority appeal.Research on minority influence has amply demonstrated the different processes underlying manifest and latent levels of influence (Mugny & Pérez, 1991;Pérez & Mugny, 1993).Our arguments are appropriate specifically in the case of manifest and public influence, in which social comparison and identity conflicts between the source and the target are salient (Mackie & Wright, 2003;Sanchez-Mazas & Falomir-Pichastor, 1995).
The ambivalent role of assimilation norms in promoting solidarity is also worth further investigation.Assimilationism is normally associated with ethnocentrism and the preservation of structural inequalities (Green & Staerklé, 2013;Guimond et al., 2013;Politi & Staerklé, in press).Although assimilationist arguments were particularly effective for a refugee source, they Further research should also provide finer empirical evidence about the underlying mechanisms of our findings.Additional measures (e.g.quality of the information processing, representations of the source, evaluation of vested interest and perceived similarity between the source and the target) could reveal the factors that are most responsible for the increased solidarity with refugees after the minority appeal.
For instance, increased or decreased vested interest attributed to the source of influence may provide another explanation of our results.Preliminary results revealed that refugees were perceived to endorse multicultural arguments to a greater extent than assimilationist ones.A refugee source using assimilationist arguments (i.e.arguments with which few other refugees would agree) could have therefore been perceived as demonstrating less group-based self-interest.Conversely, Swiss people were perceived to endorse assimilationist arguments more strongly than multicultural ones.A Swiss source using multicultural arguments (i.e.arguments with which few other Swiss people would agree) could have therefore been perceived as demonstrating less group-based self-interest.In other words, both ingroup and outgroup sources who displayed a minority opinion in their own group violated intergroup expectancies and opened the door for a more positive consideration of their arguments.Future investigations should make a more direct link between minority influence and vested interest for attitude-behaviour consistency (Crano, 1997).

Conclusion
We believe that our findings bear practical implications for movements trying to spread a message of solidarity, in contrast with the rising of the far right in Europe.There is a consensus in minority influence research suggesting that only ingroup sources are effective at the manifest level (Gardikiotis, 2011;Souchet et al., 2006).However, to assume that the situation of refugees in Switzerland can only be improved through the intervention of Swiss activists would be highly discouraging and even incorrect.Often voiceless, refugees are constantly excluded from public debate.Being the object of dispute, they are merely depicted as the problem to be solved instead of the victims to be heard.
Carrying a revolutionary point of view, Moscovici (1976) proposed that ' every group member, irrespective of his rank, is a potential source and receiver of influence' (p.67).His paradigmatic upheaval revealed that disadvantaged group members are not only objects, but also subjects of debate.In this respect, the present study showed that outgroup sources might be depositary of influence.In other words, we demonstrated that allowing refugees to speak for themselves is a valid strategy for promoting solidarity among the majority.Ultimately, we provide strategies of communication that can give voice back to those who have been deprived of such a right.

Notes
1 Detailed descriptions of the model can be found in Staerklé et al. (2007), Staerklé (2009Staerklé ( , 2016). 2 Two additional measures were included in the questionnaire: Evaluation of the author and personal agreement with the appeal.The former was not associated with hypotheses, nor used for analyses.The latter was used as an additional dependent variable to test our model.For more information, please contact the authors. 3For more information about this procedure see Jaccard and Turrisi (2003, pp. 39-43).

Figure 1 :
Figure 1: Conditions (source × message) interacting with national identity to predict solidarity with refugees.

Table 1 :
Means and standard deviations of the dependent variable (solidarity with refugees) as well as distribution of participants across the 5 experimental conditions.

Table 2 :
Correlations between variables regardless of experimental condition, including means and standard deviations.

Table 3 :
Hierarchical regression results: Conditions with national identity predict solidarity with refugees.

Table 4 :
Estimated means and standard errors of dependent variable (solidarity with refugees) according to five experimental conditions.result in paradoxical effects in the long term: Refugees may experience and even perpetuate structural disadvantages because of the arguments used to increase support.