POLITICAL USE OF IMMIGRANT STEREOTYPES 1 Schrodinger’s Immigrant: The Political and Strategic Use of (Contradictory) Stereotypical Traits about Immigrants

In two experimental studies, we explored the politicized use of immigrant stereotypes by members of the Portuguese host society. Our goal was to demonstrate that stereotypes strategically vary to support political arguments and mobilize others towards one’s political position. In Study 1, the way in which participants described immigrants varied as a function of (a) whether or not they were provided with the opportunity to mobilize an ingroup audience, (b) the direction in which they sought to mobilize that audience, and (c) the instrumentality of particular stereotypical traits in a given argumentative context. Study 2 examined the cognitive processes underlying the strategic expression of stereotypes, by measuring implicit stereotyping and manipulating participants’ ability to do cognitive work prior to and during stereotype expression. The findings showed that (a) variability in stereotype expression was matched by a consonant variability in implicit representations of immigrants, and that (b) the ability to do cognitive work was a necessary condition for variability in stereotype expression to emerge. Taken together, these findings suggest that strategic effects in stereotype expression are underpinned by strategic work on intra-psychic representations. We conclude by discussing the complementarity between strategic and cognitive explanations of stereotype expression, as well as by arguing the possibility of a more predictive approach to the politicized use of stereotypes. POLITICAL USE OF IMMIGRANT STEREOTYPES 3 I get tired of people trying to stereotype us because they never get it right. You get a group that says “Mexicans are lazy”. Then they say “Mexicans are taking away our jobs” (laughs and applause). Which one is it? Are we lazy or are we taking away your jobs? -Cristela Alonzo, Mexican-American stand-up comedian Schrodinger’s Immigrant: The Political and Strategic Use of (Contradictory) Stereotypical Traits about Immigrants In this paper, we provide the first experimental evidence that outgroup stereotypes can be used strategically to politically mobilize an audience. Specifically, we focus on the issue of immigration and show that individuals attempting to mobilize an audience about immigration restriction strategically adjust their stereotypical descriptions of immigrants to achieve mobilization towards their own position. We further show evidence suggesting that strategic variability in overt stereotyping need not be a purely public effect, but can be underpinned by strategic cognitive work leading to similar variability in the private, cognitive representations of stereotype users. Our general purpose, however, is less to demonstrate the flexibility of stereotypical descriptions than to illustrate their role as political tools. That is, stereotypical descriptions constitute one of the main bases on which political arguments about groups are deployed and legitimized. Coined by a parody news website (Adler, 2014), the expression “Schrodinger’s immigrant” illustrates this logic applied to immigrant stereotypes. The expression aims to mock anti-immigration discourse by pointing out the apparent contradiction between describing immigrants as simultaneously “stealing our jobs” and “lazing around on benefits”. The subtext of the parody is to discredit such descriptions as nothing more than convenient devices used to advance a political agenda. Like Schrodinger’s cat, famously both dead and alive at the same time, the immigrant is being attributed contradictory properties, but the contradiction can easily be explained if the political effect of the stereotypes matters more POLITICAL USE OF IMMIGRANT STEREOTYPES 4 than their consistency. By switching between the two contradictory representations, immigrants can be presented as a threat both to job availability for host community members and to the financial viability of the system that provides benefits to the unemployed. In both cases, this justifies taking anti-immigration measures. In the present studies, we rely on this very contradiction between the “lazy profiteer” and the “job-stealing” immigrant to investigate more broadly the strategic and political use of outgroup stereotypes. The Political Function of Stereotypes In his analysis of the social functions of stereotypes, Tajfel (1981; Tajfel & Forgas, 1982) stressed the role that group stereotypes can play in justifying a variety of actions directed at the outgroup. He argued that stereotypes act as “weapons in the struggle for power constantly being waged in society” (Tajfel, 1981, p. 144), and that they could not be understood without taking into account that political role (see also Oakes, Haslam, & Turner, 1994). This relationship between stereotypes and politics is also stressed by System Justification Theory (Jost & Banaji, 1994) and by Social Dominance Theory (Sidanius, & Pratto 1999), which both underline the role of stereotypes in maintaining the political status quo. One point that remains relatively unaddressed by these approaches, however, concerns the extent to which stereotype users can intentionally use stereotypes to achieve political outcomes—that is, the extent to which stereotypes can become politicized (Condor, 1990; Sindic & Condor, 2014). Only a few studies on collective mobilization, carried out within the context of the strategic component of the Social Identity model of Deindividuation Effects (SIDE; Barreto & Ellemers, 2003; Klein, Spears, & Reicher, 2007; Postmes, Spears, Lea, & Reicher, 2000; Reicher, Spears, & Postmes, 1995) have addressed this question directly. POLITICAL USE OF IMMIGRANT STEREOTYPES 5 SIDE was initially developed to account for deindividuation effects (e.g., Zimbardo, 1969) in social identity terms. It proposed that situational factors such as group immersion and anonymity could increase the salience of social identity, and thereby trigger behavior driven by social identity rather than antisocial behavior (the cognitive component of SIDE). However, the model was later extended to look at how situational factors affect the strategic expression of identity (the strategic component of SIDE), and in particular how that expression varies as a function of audiences characteristics (Barreto, Spear, Ellemers, & Shahinper, 2002; Klein et al., 2007). In that context, SIDE studies on political mobilization have shown that stereotypical descriptions can be strategically deployed in order to sway audiences and achieve mobilization towards the realization of group-based political goals (Reicher, Hopkins, & Condor, 1997; Reicher & Hopkins, 2001; Klein & Licata, 2003; see also Banton, 1988, Roosens, 1989). They show not only that individuals with different political goals rely on different stereotypes to suggest different course of actions, but also that the same individuals can rely on different stereotypes as a function of how their goal is best served given the particular audience. Such intra-individual variability strongly supports the existence of a purposeful political intent. Importantly, SIDE stresses that such behavior is an expression of social identity. Therefore, it occurs when social identity is salient and is led by goals relating to the perceived interest of the group as a whole rather than by personal self-interest. While this does not exclude stereotypes being used strategically for other (e.g., interpersonal) purposes, in line with SIDE our focus in this paper lies with their use in the pursuit of group-based political goals, and specifically in seeking to achieve group mobilization towards affecting the nature of intergroup relationships. POLITICAL USE OF IMMIGRANT STEREOTYPES 6 To be more precise, our main purpose is to extend the findings of the above SIDE studies in at least three ways. First, those studies have focused on the analysis of spontaneous stereotype expression and relied on qualitative (Reicher, Hopkins, & Condor, 1997) or quantitative content analysis (Klein & Licata, 2003). This provides them with a high degree of external validity, but also means that the triangulation of their findings with experimental data is still lacking (see Sindic & Reicher, 2008). In particular, while their findings provide evidence of the role of political goals in shaping spontaneous stereotype expression, the claim of a relation of causality between the two could be reinforced by using political goals as an independent variable and examining the subsequent impact on stereotype use. Second, SIDE does not offer an a priori model of stereotype contents and their potential political uses. Rather, the consonance between stereotypes and goals is established a posteriori, as a result of the empirical analyses themselves. This limits the ability to make predictions as to which contents are used in what contexts and for what purposes. Thus, our second goal in this paper is to contribute towards the development of a more predictive approach to the content of politicized stereotypes. While a comprehensive and systematic model is beyond what can be accomplished with a couple of experimental studies, our aim is to illustrate the type of approach and studies that would be necessary in order to progress towards that purpose, by demonstrating the very possibility of prediction. Third, although the SIDE model includes hypotheses about both cognitive and strategic stereotyping, the articulation between those two sides of the model remains relatively unclear. So far the cognitive side of SIDE has mainly focused on explaining variations in intra-psychic stereotypical representations, and the strategic side on explaining variations in the public expression of stereotypes, as if they were separate phenomena. Thus, our third goal is to contribute towards bridging this gap, by looking at what occurs to the POLITICAL USE OF IMMIGRANT STEREOTYPES 7 intra-psychic representations of stereotype users when they use stereotypes strategically and publicly. The Context and Content of Politicized Immigrant Stereotypes If predicting the content of politicized stereotype expression depends upon modelling stereotype contents and their potential political functions, then one possibility would be to combine SIDE with other approaches that seek precisely to do that. For reasons detailed below, in the current studies we rely on the semantic distinctions proposed by Osgood (1964) between the dimensions of Evaluation (good-bad), Potency (strong-weak), and Activity (active-passive). However, we also borrow from the Stereotype Content Model (SCM; Fiske, Cuddy, Glick, & Xu, 2002; Lee & Fiske, 2006) the notion that the function of outgroup stereotypes is to inform us as to whether the outgroup poses a threat to the ingroup (Cuddy et al., 2009; Kervyn, Fiske, & Yzerbyt, 2013). Combined with SIDE’s focus on political mobilization, it leads us to propose that would-be mobilizers should make use of those particular stereotypical traits that represent others as a political threat (or as politically benign) in order to shape actions towards the outgroup. There is, however, a significant difficulty when it comes to translating that general rationale into specific predictions. For the SCM, threat is associated with perceiving the other as motivated and capable of doing us harm (i.e., cold and competent, Kervyn et al., 2013). Yet, even the most cursory glance at political discourses shows that political threats can take various forms and rely on a variety of stereotypical traits. For instance, the threat of political domination can be presented as more serious if it is claimed that it will occur even in the absence of harmful intentions by the outgroup (e.g., Scott, 1985). Likewise, it is possible to claim that the weakness or passivity of others is harmful to the ingroup, such as for instance when welfare recipients (immigrants or not) are depicted as lazy “parasites” living off the taxpayers’ money. Drawing unconditional associations between specific stereotypical traits POLITICAL USE OF IMMIGRANT STEREOTYPES 8 and the establishment or denial of political threat therefore seems to represent too simplistic an approach to capture that flexibility. We contend, however, that predictions are possible if one also takes into account the context of strategic stereotype expression. As noted, one of our main goals in this paper is to illustrate this point in the context of immigrant stereotypes. Specifically, in our first study, we manipulated the argumentative context in which stereotypes were expressed, inducing participants to focus on the issue of whether immigrants pose a threat to job availability or to the resources of the social security system (by claiming unemployment benefits). As our opening example illustrates, a typical anti-immigration argument in relation to job availability consist in arguing that immigrants are “stealing our jobs”. This argument can be supported by stressing that immigrants are hungry for work (e.g., the immigrant as “cheap labour”, willing to do hard work for little pay), as this suggests they introduce a high degree of competition in the job market, and end up taking many jobs away from members of the host society. By contrast, a typical anti-immigration argument in the context of social security resources consists in arguing that immigrants are more likely to be unemployed and thereby to drain the resources of social security. This argument can be supported by depicting immigrants as “lazy profiteers” who do not actively look for work and are content to live off unemployment benefits, sustaining the notion that they are indeed more likely to be unemployed. Thus, in both cases, the key focus lies with the “hardworking” versus “lazy” quality of immigrants. Similarly, those who seek to promote a more pro-immigration stance can support it by using the same stereotypical traits but in a reverse pattern. Those particular stereotypical traits offer some basis for prediction not merely because they form the basis of common arguments in relation to the above issues, but also because the flexibility with which they can be used is limited. Indeed, the argument that immigrants are stealing jobs cannot easily be supported by describing them as lazy, and the POLITICAL USE OF IMMIGRANT STEREOTYPES 9 argument that they laze around on benefits by describing them as hardworking. By contrast, other stereotypical traits such as intelligence potentially enjoy more flexibility in the way in which they can be used in the same argumentative contexts. Thus the “benefit-profiteering” immigrant might be portrayed as possessing little intelligence or professional qualifications, but might also be portrayed as cunning in terms of knowing how to take advantage of the system. Both versions can serve the same purpose. Our predictions therefore focused on variations along the “hardworking versus lazy” dimension, but where this dimension is to be distinguished from “intelligence” traits to which it is often associated (Fiske et al., 2002). For this reason, we rely on Osgood’s nomenclature, as the distinction can conveniently be mapped on to the dimensions of Activity and Potency. Similar distinctions proposed in the literature between motivation and ability (Carrier, Louvet, Chauvin, & Rohmer, 2014) and between brilliance and dedication (Bian, Leslie, Murphy, & Cimpian, 2018) lend further weight to the meaningfulness of the distinction, and could also be used to label the stereotypical traits at stake. Study 1 In Study 1, we looked at the way in which participants of Portuguese nationality described immigrants as a function of whether they sought to support or oppose increased restriction on the amount of immigrants allowed into the country. To maximize the external validity of our findings, we relied on participants’ pre-existing goals in relation to this issue. However, to avoid variability in stereotypical descriptions being potentially reducible to stable inter-individual differences in political views, we also introduced two manipulations to achieve distinct predictions of strategic effects in stereotype expression. Our first manipulation consisted in introducing the opportunity to mobilize an ingroup audience prior to, or subsequent to, stereotype expression. Simply put, only when a mobilization opportunity was provided prior to stereotype expression should participants be POLITICAL USE OF IMMIGRANT STEREOTYPES 10 driven by the goal of mobilizing others through stereotype expression. Thus, prior to stereotype expression, we instructed half of our participants that their responses would be shown to a group of undecided Portuguese to assess the extent to which they could be swayed to sign a petition in line with the participants’ position. Audience members were characterized as undecided to ensure that they would be perceived as highly susceptible to be mobilized and provide participants with sufficient motivation to use stereotypes for mobilization purposes. In addition, the use of undecided individuals served to neutralize the potential interference of complicating factors flowing from addressing audiences with existing positions on the issue (in such circumstances, additional strategic considerations may intervene). Our second manipulation aimed to affect the nature of the particular stereotypical traits seen as instrumental to mobilize the audience in a particular direction. To achieve this, we manipulated the argumentative context in which the question of whether or not immigrants represents a threat arose. Specifically, we introduced the study as focusing on the extent to which immigrants do or do not represent a threat to the financial resources of the social security system, or to job availability for host society members. Note that in Portugal, as in most European countries, all workers (immigrant or not) must pay a contribution to social security, and that the latter provides a variety of government benefits, from health cover and state pensions to unemployment benefits. However, the text of our manipulation made it clear that the issue was specifically about whether immigrants have a positive or negative impact on social security funds through providing contributions as workers versus claiming unemployment benefits. We chose these two particular issues for three reasons. First, in Portugal, as in most European countries, both issues are among the most commonly debated points of contention in the public arena about immigration (e.g., ACIDI, 2007, “David Cameron”, 2014). POLITICAL USE OF IMMIGRANT STEREOTYPES 11 Therefore, both issues could credibly be presented as key in determining positions on immigration restriction. Second, the “lazy” versus “hardworking” stereotypes of immigrant with which these issues are associated are familiar to the population at large, whether or not they are endorsed. This provided participants with the ability to rely on credible stereotypes whichever way they chose to argue. Third, the use of these issues allowed for predictions contrasting with expectations based on stable interpersonal differences in political goals or orientations. As noted, we predicted that participants with the same goals would use opposite ends of the “lazy versus hardworking” dimension as a function of which issue is at stake, insofar as it determines which stereotypical descriptions are instrumental in portraying immigrants as a threat. More specifically, we hypothesized a three-way interaction between the direction of political goals, the timing of the mobilization opportunity, and the nature of the potential threat at stake, whose pattern should be as follows: H1. When the potential threat represented by immigrants bears upon social security resources, participants whose goal is to increase restriction on immigration should stress that immigrants are less hardworking when they are provided with a mobilization opportunity prior to stereotype expression, compared to when they are not (H1a). The opposite pattern should apply for participants whose goal was to oppose increased restriction on immigration (H1b). H2. When the potential threat represented by immigrants bears upon job availability, participants whose goal is to increase restriction on immigration should stress that immigrants are more hardworking when they are provided with a mobilization opportunity prior to stereotype expression, compared to when they are not (H2a). The opposite pattern should apply for participants whose goal was to oppose increased restriction on immigration (H2b). POLITICAL USE OF IMMIGRANT STEREOTYPES 12 Asides from traits relating to the hardworking dimension, we also measured stereotypical ascriptions of intelligence and sociability (Potency and Evaluation). Although we had no predictions as to the impact of our manipulations on those traits, it was important to include them to verify that variability in the use of stereotypical content was indeed limited to the hardworking traits. This was necessary to confirm the validity of our rationale as to the importance of distinguishing those traits for making predictions in the contexts at stake. Method Participants. Two hundred and seventy-six participants took part on a voluntary basis. Three participants were discarded for missing values on the main dependent variable, leaving 273 participants (age range: 18-70, M = 25.2, SD = 8.07). Most participants were university students in various fields (70.3%), 20.1% were employed, 5.1% unemployed, and 4.4% were in other situation. Since we had no previous similar studies to rely on, and no expectations as to the size of the hypothesized effects, sample size resulted from collecting data for a fixed period of time (i.e., one month), after which the effect size that the achieved sample size allowed to detect was assessed. A sensitivity power analysis indicated that the sample size of 273 obtained in that time allowed the detection of effects as small as ƒ ≥ .029 with 80% power or ƒ ≥ .048 with 95% power for the main multiple regression analyses (which included seven regression terms). This was deemed sufficient so that no further data was collected. No data analysis was performed before data collection ended. Design. The study followed a Potential Threat (social security vs. job availability) × Mobilization Opportunity (opportunity vs. no opportunity) × Goal Direction (continuously measured, from opposing to supporting increased restriction on immigration) betweenparticipants design. To simplify, we refer to the levels of Mobilization Opportunity as opportunity versus no opportunity, with the understanding that this stands for the POLITICAL USE OF IMMIGRANT STEREOTYPES 13 mobilization opportunity being introduced prior to or after stereotype expression. Our main dependent variable was the use of hardworking stereotypical traits to describe immigrants (Hardworking). All measures, manipulations, and exclusions in this study are disclosed either in the main text or in supplementary materials. All means, standard deviations, and correlations between the main measures are reported in supplementary materials. Procedure. Individuals were approached in university premises and those interested in participating were asked to provide their email addresses. The experimenter subsequently sent an email with a hyperlink directing them to one of four versions of an online questionnaire. The experimenter randomly determined which hyperlink participants received and thereby which experimental condition they were assigned to. At the end of the questionnaire, participants were thanked and debriefed. All measures and manipulations were part of the questionnaire itself. However, as part of the manipulation of Mobilization Opportunity, the order in which they were introduced varied across conditions. The exact chronological sequences for all conditions are represented in Figure 1. Overall, only the relative position of stereotypical ratings vis-à-vis the manipulations was altered as a function of Mobilization Opportunity. Note that in all conditions we measured Goal Direction after the manipulation of Potential Threat, because it was possible for this manipulation to affect participants’ position on immigration restriction. Therefore, measuring Goal Direction beforehand might have yielded an inaccurate estimate of participants’ goals during stereotype expression. We address the possible issue of correlation between the measure and the manipulation in the results section. Note also that, conversely, in the control conditions, stereotypical ratings preceded the manipulation of Potential Threat (as well as Mobilization Opportunity). This is because mentioning any of the political debates before stereotypical ratings could have made the POLITICAL USE OF IMMIGRANT STEREOTYPES 14 political implications of stereotypes salient, and thereby potentially trigger a mobilization goal, even in the absence of a direct opportunity to mobilize a specific audience. We address the possible issue of interpretation this might create in the discussion. Salience of national identity. Since the political goals under scrutiny were groupbased (i.e., relating to the perceived interest of the national ingroup as a whole), it was important to ensure that all participants were first primed to act in terms of their ingroup (national) identity. Thus, prior to introducing both experimental manipulations and stereotypical measures, national identity was made salient for all participants by requesting them to indicate their nationality and complete a scale of identification with being Portuguese (Verkuyten & Hagerdoorn, 1998; translated to Portuguese). Manipulation of Potential Threat. Participants were informed that we were interested in their position on immigration restriction, and that the key issue in terms of determining people’s positions on this issue was the positive or negative impact of immigrants on the job market versus the resources of the social security system. For each issue, we were careful not to take a position and stressed that both sides of the debate have been argued in the public domain. Manipulation of Mobilization Opportunity. Prior to stereotypical ratings, participants in the “opportunity” conditions were given the opportunity to mobilize an ingroup audience towards their position. Specifically, they were provided with an open space to present their arguments in relation to immigration restriction, and were told that both their arguments and 1 Since both control conditions measured stereotypical ascriptions in the same conditions, it could be argued that they should be collapsed. We did run similar analyses to the ones presented below using a single collapsed condition and obtained similar results. For reasons of transparency, we present both design and analyses as they were originally intended. POLITICAL USE OF IMMIGRANT STEREOTYPES 15 their answers on the stereotype scales would be shown to a group of undecided Portuguese to assess whether their responses could sway those individuals to sign a petition in line with their own position. The other half of participants were assigned to one of two control conditions, in which stereotypical ratings were applied at the very beginning of the questionnaire, that is, before the opportunity to mobilize an audience was introduced. Measures. All items in the questionnaire were answered on 11-point Likert-type rating scales ranging from 0 to 10. We describe here the measures used in the analyses (additional measures are disclosed in supplementary materials). Goal Direction. Three items measured the direction of political goals toward immigration restriction (α =.88): “Currently, do you think that immigration in Portugal should be increased or decreased?” (0 = increase a lot to 10 = decrease a lot), “Portugal has more immigration than is fair” (0 = strongly disagree to 10 = strongly agree), “Portugal should take more severe measures to reduce immigration” (0 = strongly disagree to 10 = strongly agree). Scores above the scale midpoint reflected the goal to increase immigration restriction. Because it could be argued that, despite the high overall alpha, disagreeing with increased restriction in items 2 and 3 is not fully equivalent to supporting increased immigration in item 1, we refer to participants scoring below the scale midpoint as opposing increased restriction, as this is inclusive of low scores on all 3 items. Stereotypical traits. Stereotypical descriptions of immigrants were measured using ratings on the following traits (translated from Portuguese): a) “hardworking” and “resistant to hard working conditions” (Activity), b) “competent” and “intelligent” (Potency), and c) “friendly”, “loyal”, and “trustworthy” (Evaluation). We used only positive traits to limit the impact of social desirability considerations and their possible interference with strategic influences (see General Discussion). Participants were asked to rate each trait twice, the first time in absolute terms (i.e., “do you consider immigrants in general to be X?”) and the POLITICAL USE OF IMMIGRANT STEREOTYPES 16 second in comparative terms (i.e., “do you consider immigrants in general to be more or less X than the Portuguese?”). Comparative ratings were included because, in principle, the political arguments at stake involve a comparison with host society members. For instance, arguing that immigrants are a threat to job availability due to being hardworking only works if it means they are more hardworking than the Portuguese, since it is about establishing them as competition for the Portuguese. In practice, however, absolute ratings were likely to trigger such comparisons implicitly—indeed, self-categorization theory (Turner, Hogg, Oakes, Reicher, & Wetherell, 1987) contends that all stereotyping is inherently comparative, either explicitly or implicitly. Since Cronbach’s alphas were satisfactory when using the comparative and absolute ratings together (for hardworking traits, α =.79; for potency traits, α =.84; for evaluation traits, α =.91), we used them as single scales to avoid repetition in the analyses. Results Impact of Potential Threat on Goal Direction. Prior to the test of our hypotheses, we first established that Potential Threat had a small but significant impact on the goal direction measure, F(1,271) = 4.24, p = .041, η = .015, with participants in the job availability conditions being overall slightly more supportive of increased immigration restriction (M = 5.49, SD = 2.37) than participants in the social security conditions (M = 4.81, SD = 2.54). However, this effect was not a cause for statistical concern, since it was quite small (η = .015), and VIF scores for all predictors in subsequent multiple regression analysis were all well below the most conservative criterion of acceptable range (i.e., all below 2.5, with the highest being 1.14). Furthermore, our hypotheses focused on interactions rather than on main effects, for which correlation between predictors is even less an issue both statistically (Friedrich, 1982) and in terms of interpretation (see Discussion). Nevertheless, to ensure no bias in interpreting the pattern of POLITICAL USE OF IMMIGRANT STEREOTYPES 17 those interactions, we gave equal weight to both sides of Goal Direction in plotting the interaction for all conditions. Specifically, we used one standard deviation (SD) below and above the scale midpoint as low and high conditional values of Goal Direction, as opposed to the more usual values of one SD below and above the mean. The resulting simple slopes, therefore, illustrate the trends for participants with the goal to support or oppose increased immigration restriction in absolute terms, irrespective of the condition, rather than in comparison to other participants sharing the same level of Potential Threat. This also allowed for a more meaningful representation of our data (Aiken & West, 1991). Stereotypical ratings. To test our hypotheses, we performed a multiple regression analysis with Hardworking as dependent variable, and Goal Direction (centered), Mobilization Opportunity (coded as -1/1), Potential Threat (coded as -1/1), and all interactions between these three variables as predictors (see Table 1). This analysis yielded a significant negative effect of Goal Direction: Overall, participants rated immigrants as less hardworking when their goal was opposing more restriction on immigration. The two-way interaction between Goal Direction and Potential Threat was also significant. Most importantly, however, both effects were qualified by the predicted three-way interaction between all regression terms. Since the test of our hypotheses implied examining the impact of Mobilization Opportunity on participants sharing similar goals, we then calculated the simple slopes for the conditional effect of Mobilization Opportunity at different values of Goal Direction, in the social security and job availability conditions respectively (see Table 1 & Figure 2). The results indicated that, in the social security conditions, there was a significant two-way interaction between Goal Direction and Mobilization Opportunity. At high values of Goal Direction (denoting support for increased restriction), participants described immigrants as significantly less hardworking when provided with a mobilization opportunity compared to POLITICAL USE OF IMMIGRANT STEREOTYPES 18 when they were not (supporting H1a). By contrast, at low values of Goals Direction (denoting opposition to increased restriction), participants described immigrants as more hardworking when provided with a mobilization opportunity compared to when they were not (supporting H1b). Note that the significant two-way interaction confirms that the simple slopes were significantly different from each other. In the job availability conditions, the two-way interaction between Goal Direction and Mobilization Opportunity was also significant, but displayed the opposite pattern. At high values of Goals Direction, participants described immigrants as significantly more hardworking when provided with a mobilization opportunity compared to when they were not (supporting H2a). By contrast, at low values of Goal Direction, participants described immigrants as less hardworking when provided with a mobilization opportunity compared to when they were not (supporting H2b). In sum, the three-way interaction found in the overall equation emerged due to lower-level two-way interactions with opposing patterns in the social security and job availability conditions. Furthermore, the directions of these two-ways interactions were fully consistent with our hypotheses H1 and H2. Finally, as regards stereotypical traits relating to Evaluation and Potency, using the same multiple regression procedure as above showed that the three-way interaction was not significant for those traits (for evaluation, b = -.06, SE = .04, t(265) = 1.59, p = .113; For potency, b = -.01, SE = .04, t(265) = .34, p = .738). The only significant effects were negative main effects of Goal Direction on ratings of evaluation and potency, similar to its main effect on Hardworking traits (for evaluation, b = -.27, SE = .04, t(265) = 7.24, p <.001, ƒ = .20; for potency, b = -.23, SE = .04, t(265) = 6.35, p < .001, ƒ = .15). Thus, our results show that the pattern of variability in stereotyping predicted for Hardworking traits did not extent to other stereotypical traits, confirming the importance of distinguishing between them in the context of the immigration issues at stake. POLITICAL USE OF IMMIGRANT STEREOTYPES 19 Discussion The main goal of Study 1 was to provide evidence that political intent affects stereotypical descriptions of the outgroup. To achieve that purpose, we provided our Portuguese participants with the opportunity to mobilize an audience before or after they rated immigrants on stereotypical traits. Our results show that this manipulation led to significant differences in the content of the expressed stereotypes, in a direction that was consonant with participants’ political goals. This is consistent with the claim that political intent played a causal role in shaping stereotype expression. Moreover, we aimed to show that the specific stereotypical traits used to defend a given position can strategically vary as a function of which stereotype is seen as most instrumental to achieve one’s goal in a given situation. We sought to illustrate this flexibility by manipulating the specific immigration issue presented to participants as being most relevant to positions on immigration restriction. Our results show that this manipulation led participants with similar goals to express opposite stereotypes, in a direction that was fully consistent with our predictions. This significant interaction between pre-existing goals and immigration issues is important because it means that we can reject alternative explanations based on stable individual differences in cognitive representations—that is, the fact that pre-existing political positions or orientations might be consistently associated with the use of certain stereotypes. Likewise, the interaction cannot be fully explained by invoking purely contextual effects, that is, that the use of different immigration issues triggered differences in the situational intrapsychic salience of stereotypes, which then led to differences in stereotype expression. Nonetheless, it would still be possible to combine the above rationales and argue that the two immigration issues made different stereotypes cognitively salient for different individuals. Since exposure to the immigration issues occurred in conjunction with the POLITICAL USE OF IMMIGRANT STEREOTYPES 20 mobilization opportunity, this could potentially account for the interaction. In the mobilization conditions, the issues of job availability might have made the “hardworking” stereotype salient and the issue of social security the “lazy” stereotype salient for participants supporting immigration restriction, and vice-versa for participants with the opposite position. By contrast, in the control conditions, both manipulations took place after participants expressed stereotypes, so that such effect could not occur. Moreover, in the mobilization conditions, participants were able to elaborate their opinion about immigration in open format before rating immigrants. This might have further reinforced the cognitive salience of those stereotypes activated by the particular combination of context and individual differences. Theoretically, the SIDE model does not deny the role of situational and/or cognitive factors on stereotype expression (Reicher et al., 1995). Rather, its strategic component predicts that, under conditions of mobilization, political goals should affect stereotype expression over and above the impact of those factors. Thus, our explanation does not deny that variations in the cognitive representations of immigrants might have occurred. Rather, it implies that if such variations did occur, they should not constitute the sole explanation for variations in stereotype expression. Therefore, the key difference between the two explanations does not lie with the presence or absence of variation in cognitive salience, but in whether or not political intent is also necessary to account for the expressed stereotypes. In addition, variations in cognitive salience themselves may be the product of cognitive work driven by political intent (see Study 2). Since both explanations can account for the results and for possible variations in cognitive salience, the question is which interpretation should then be preferred, based on other criteria. In this context, we note that the non-strategic interpretation possesses a number of significant disadvantages. First, it requires arguing that, compared to pro-immigration participants, anti-immigration participants were more predisposed (as a matter of stable POLITICAL USE OF IMMIGRANT STEREOTYPES 21 differences in cognitive representations) to endorse the representation of immigrants as more hardworking and as more lazy. While individuals may possibly hold ostensibly contradictory beliefs, this is certainly less plausible than claiming that political goals led each stereotype to become salient in the relevant context. As we shall argue shortly, the influence of goals and motivations on cognitive salience is a well-established phenomenon. Second, the alternative explanation is less parsimonious, since it requires postulating the existence of two distinct pre-existing cognitive structures to replace the role of political intent. Third, it treats the particular content of both cognitive representations and expressed stereotypes as mere happenstance, rather than as driven by a rationale, thereby leading to a loss of explanatory power. Finally, the logic is open to post-hoc rationalizations as to the relevant features of the context and pre-existing cognitive structures that lead to a specific pattern of results. By contrast, a strategic framework provided the general rationale as to which features should matter, on which basis we were able to predict the results a priori (see also General Discussion). Nevertheless, in Study 2, we sought to look at the cognitive processes underlying strategic stereotype expression, which provided an opportunity to test the alternative explanation more directly. In addition, we also sought to address the related issue of the overlap between the introduction of the specific immigration issues and of the mobilization opportunity, given that it helped to lend credence to this alternative. Study 2 In Study 2, our main goal was to look at the cognitive processes that underlie the political/strategic expression of stereotypes. Specifically, we aimed to clarify the relation between the content of publicly expressed stereotypes and the content of the private, intrapsychic, cognitive representations of the mobilizer. POLITICAL USE OF IMMIGRANT STEREOTYPES 22 Much research has shown the resilience of stereotypical beliefs (e.g., Snyder, 1981) and the fact that they may persist privately despite social and cultural pressures to suppress or moderate their expression (e.g., Dambrun & Guimond, 2004; Devine, 1989; Dovidio, Kawakami, & Beach, 2001). On that basis, one could conceive of political/strategic influences on stereotype expression in similar terms, that is, as affecting only the outward expression of stereotypes and leaving untouched the more stable, private cognitive representations of the stereotype user. However, there is also a considerable amount of research showing that information processing in general, and cognitive stereotyping in particular, can be influenced by motivations and goals (e.g., Kunda & Sinclair, 1999), even at the implicit level (Blair, 2002). For instance, the motivation to arrive at a particular representation may lead individuals to selectively scan their memory in search of examples that support the desired impression while discarding those that do not (Sanitioso, Kunda, & Fong, 1990), or may affect the way in which specific instances are seen as typical or else subtyped as unrepresentative exceptions (Carnaghi & Yzerbyt, 2007). Similarly, the existence of a mobilization goal might trigger selective intra-psychic work affecting the cognitive representation of immigrants. For instance, individuals might give different weights to individual instances or immigrant subgroups when forming the prototype of immigrant as a whole. From a political and strategic point of view, there are in fact good reasons to expect this process to occur, and to postulate that it constitutes one of the possible mechanisms underlying variability in stereotype expression. As Kunda pointed out, in relation to motivated reasoning, “people are more likely to arrive at conclusions that they want to arrive 2 In Study 2, we actually included a measure relating to immigrant subgroups to explore that possibility, and provide a brief overview of the results in supplementary materials. POLITICAL USE OF IMMIGRANT STEREOTYPES 23 at, but their ability to do so is constrained by their ability to construct seemingly reasonable justifications for these conclusions” (1990, p.480; see also W. Klein & Kunda 1992). Likewise, from a SIDE perspective, it is precisely because stereotypes provide the “reasonable justifications” of political arguments that they are used for political purposes. However, stereotypes can only play this role if they are deemed to reflect social reality. Thus, potential mobilizers have an interest in recruiting the particular cognitive resource (such as instances of behavior confirming the stereotype) that allows them to substantiate and legitimize their claims. Furthermore, when addressing an ingroup audience, mobilizers generally expect that the audience will agree with them in their assessment of social reality (Haslam, Turner, Oakes, McGarty, & Reynolds, 1998). This should lead them to assume a priori that the best arguments to convince that audience are those that persuade themselves. Therefore, the default option in terms of formulating arguments should be to build upon their private beliefs. Psychologically, this should lead to the strategic examination of those private beliefs, which in turn may lead to variability in immigrant representation at the cognitive level. To investigate this question, Study 2 used a design similar to Study 1, but focused only on the issue of social security, and introduced two important additions. First, we developed an adaptation of the Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998) to assess the effect of the manipulations on the intra-psychic cognitive representations of immigrants. Second, we added a manipulation of cognitive load to affect the ability of individuals to do cognitive work. We reasoned that the activation of mobilization goals should lead to motivated information processing (or strategic thinking) about immigrant stereotypes, resulting in variability in intra-psychic representations. In turn, the content of intra-psychic representations should shape the content of the stereotypes overtly expressed. However, cognitive load should impede that process and prevent POLITICAL USE OF IMMIGRANT STEREOTYPES 24 variability from appearing at both the explicit and implicit levels. More specifically, given that the issue at stake was the impact of immigrants on social security resources, we predicted that: H1: When provided with the opportunity to mobilize an audience, and when able to do cognitive work, participants supporting increased restriction on immigration would show a weaker implicit association between immigrants and hardworking traits compared to participants who had no mobilization opportunity, while participants opposing increased restriction would show a stronger implicit association. By contrast, we expected no difference among participants under cognitive load, whether or not they were provided with a mobilization opportunity. H2: The same pattern of three-way interaction was expected to occur for overt stereotypical ratings. Asides from testing our hypotheses, the manipulation of cognitive load also enabled different predictions for a strategic explanation of stereotype variability compared to one based purely on the combination of situational and cognitive effects. For reasons to which we shall return (see General Discussion), this alternative poses a more critical challenge to a strategic interpretation if those effects are deemed to reflect the action of purely reactive and automatic processes. This implies that, provided that participants are not under load when stereotypes are psychologically activated, variability in both implicit and explicit stereotyping should occur independently from the presence or absence of cognitive load. In addition, we also modified the design of the control conditions to assess the impact of receiving a mobilization opportunity independently from the impact of manipulating the focus of mobilization attempts.


Abstract
In two experimental studies, we explored the politicized use of immigrant stereotypes by members of the Portuguese host society. Our goal was to demonstrate that stereotypes strategically vary to support political arguments and mobilize others towards one's political position. In Study 1, the way in which participants described immigrants varied as a function of (a) whether or not they were provided with the opportunity to mobilize an ingroup audience, (b) the direction in which they sought to mobilize that audience, and (c) the instrumentality of particular stereotypical traits in a given argumentative context. Study 2 examined the cognitive processes underlying the strategic expression of stereotypes, by measuring implicit stereotyping and manipulating participants' ability to do cognitive work prior to and during stereotype expression. The findings showed that (a) variability in stereotype expression was matched by a consonant variability in implicit representations of immigrants, and that (b) the ability to do cognitive work was a necessary condition for variability in stereotype expression to emerge. Taken together, these findings suggest that strategic effects in stereotype expression are underpinned by strategic work on intra-psychic representations. We conclude by discussing the complementarity between strategic and cognitive explanations of stereotype expression, as well as by arguing the possibility of a more predictive approach to the politicized use of stereotypes.
I get tired of people trying to stereotype us because they never get it right. You get a group that says "Mexicans are lazy". Then they say "Mexicans are taking away our jobs" (laughs and applause). Which one is it? Are we lazy or are we taking away your jobs? -Cristela Alonzo, Mexican-American stand-up comedian Schrodinger's Immigrant: The Political and Strategic Use of (Contradictory) Stereotypical

Traits about Immigrants
In this paper, we provide the first experimental evidence that outgroup stereotypes can be used strategically to politically mobilize an audience. Specifically, we focus on the issue of immigration and show that individuals attempting to mobilize an audience about immigration restriction strategically adjust their stereotypical descriptions of immigrants to achieve mobilization towards their own position. We further show evidence suggesting that strategic variability in overt stereotyping need not be a purely public effect, but can be underpinned by strategic cognitive work leading to similar variability in the private, cognitive representations of stereotype users. Our general purpose, however, is less to demonstrate the flexibility of stereotypical descriptions than to illustrate their role as political tools. That is, stereotypical descriptions constitute one of the main bases on which political arguments about groups are deployed and legitimized.
Coined by a parody news website (Adler, 2014), the expression "Schrodinger's immigrant" illustrates this logic applied to immigrant stereotypes. The expression aims to mock anti-immigration discourse by pointing out the apparent contradiction between describing immigrants as simultaneously "stealing our jobs" and "lazing around on benefits".
The subtext of the parody is to discredit such descriptions as nothing more than convenient devices used to advance a political agenda. Like Schrodinger's cat, famously both dead and alive at the same time, the immigrant is being attributed contradictory properties, but the contradiction can easily be explained if the political effect of the stereotypes matters more than their consistency. By switching between the two contradictory representations, immigrants can be presented as a threat both to job availability for host community members and to the financial viability of the system that provides benefits to the unemployed. In both cases, this justifies taking anti-immigration measures. In the present studies, we rely on this very contradiction between the "lazy profiteer" and the "job-stealing" immigrant to investigate more broadly the strategic and political use of outgroup stereotypes.

The Political Function of Stereotypes
In his analysis of the social functions of stereotypes, Tajfel (1981;Tajfel & Forgas, 1982) stressed the role that group stereotypes can play in justifying a variety of actions directed at the outgroup. He argued that stereotypes act as "weapons in the struggle for power constantly being waged in society" (Tajfel, 1981, p. 144), and that they could not be understood without taking into account that political role (see also Oakes, Haslam, & Turner, 1994). This relationship between stereotypes and politics is also stressed by System Justification Theory (Jost & Banaji, 1994) and by Social Dominance Theory (Sidanius, & Pratto 1999), which both underline the role of stereotypes in maintaining the political status quo.
One point that remains relatively unaddressed by these approaches, however, concerns the extent to which stereotype users can intentionally use stereotypes to achieve political outcomes-that is, the extent to which stereotypes can become politicized (Condor, 1990;Sindic & Condor, 2014). Only a few studies on collective mobilization, carried out within the context of the strategic component of the Social Identity model of Deindividuation Effects (SIDE; Klein, Spears, & Reicher, 2007;Postmes, Spears, Lea, & Reicher, 2000;Reicher, Spears, & Postmes, 1995) have addressed this question directly. SIDE was initially developed to account for deindividuation effects (e.g., Zimbardo, 1969) in social identity terms. It proposed that situational factors such as group immersion and anonymity could increase the salience of social identity, and thereby trigger behavior driven by social identity rather than antisocial behavior (the cognitive component of SIDE). However, the model was later extended to look at how situational factors affect the strategic expression of identity (the strategic component of SIDE), and in particular how that expression varies as a function of audiences characteristics (Barreto, Spear, Ellemers, & Shahinper, 2002;Klein et al., 2007).
In that context, SIDE studies on political mobilization have shown that stereotypical descriptions can be strategically deployed in order to sway audiences and achieve mobilization towards the realization of group-based political goals (Reicher, Hopkins, & Condor, 1997;Reicher & Hopkins, 2001;Klein & Licata, 2003; see also Banton, 1988, Roosens, 1989. They show not only that individuals with different political goals rely on different stereotypes to suggest different course of actions, but also that the same individuals can rely on different stereotypes as a function of how their goal is best served given the particular audience. Such intra-individual variability strongly supports the existence of a purposeful political intent. Importantly, SIDE stresses that such behavior is an expression of social identity. Therefore, it occurs when social identity is salient and is led by goals relating to the perceived interest of the group as a whole rather than by personal self-interest. While this does not exclude stereotypes being used strategically for other (e.g., interpersonal) purposes, in line with SIDE our focus in this paper lies with their use in the pursuit of group-based political goals, and specifically in seeking to achieve group mobilization towards affecting the nature of intergroup relationships.
To be more precise, our main purpose is to extend the findings of the above SIDE studies in at least three ways. First, those studies have focused on the analysis of spontaneous stereotype expression and relied on qualitative (Reicher, Hopkins, & Condor, 1997) or quantitative content analysis (Klein & Licata, 2003). This provides them with a high degree of external validity, but also means that the triangulation of their findings with experimental data is still lacking (see Sindic & Reicher, 2008). In particular, while their findings provide evidence of the role of political goals in shaping spontaneous stereotype expression, the claim of a relation of causality between the two could be reinforced by using political goals as an independent variable and examining the subsequent impact on stereotype use.
Second, SIDE does not offer an a priori model of stereotype contents and their potential political uses. Rather, the consonance between stereotypes and goals is established a posteriori, as a result of the empirical analyses themselves. This limits the ability to make predictions as to which contents are used in what contexts and for what purposes. Thus, our second goal in this paper is to contribute towards the development of a more predictive approach to the content of politicized stereotypes. While a comprehensive and systematic model is beyond what can be accomplished with a couple of experimental studies, our aim is to illustrate the type of approach and studies that would be necessary in order to progress towards that purpose, by demonstrating the very possibility of prediction.
Third, although the SIDE model includes hypotheses about both cognitive and strategic stereotyping, the articulation between those two sides of the model remains relatively unclear. So far the cognitive side of SIDE has mainly focused on explaining variations in intra-psychic stereotypical representations, and the strategic side on explaining variations in the public expression of stereotypes, as if they were separate phenomena. Thus, our third goal is to contribute towards bridging this gap, by looking at what occurs to the intra-psychic representations of stereotype users when they use stereotypes strategically and publicly.

The Context and Content of Politicized Immigrant Stereotypes
If predicting the content of politicized stereotype expression depends upon modelling stereotype contents and their potential political functions, then one possibility would be to combine SIDE with other approaches that seek precisely to do that. For reasons detailed below, in the current studies we rely on the semantic distinctions proposed by Osgood (1964) between the dimensions of Evaluation (good-bad), Potency (strong-weak), and Activity (active-passive). However, we also borrow from the Stereotype Content Model (SCM; Fiske, Cuddy, Glick, & Xu, 2002;Lee & Fiske, 2006) the notion that the function of outgroup stereotypes is to inform us as to whether the outgroup poses a threat to the ingroup (Cuddy et al., 2009;Kervyn, Fiske, & Yzerbyt, 2013). Combined with SIDE's focus on political mobilization, it leads us to propose that would-be mobilizers should make use of those particular stereotypical traits that represent others as a political threat (or as politically benign) in order to shape actions towards the outgroup.
There is, however, a significant difficulty when it comes to translating that general rationale into specific predictions. For the SCM, threat is associated with perceiving the other as motivated and capable of doing us harm (i.e., cold and competent, Kervyn et al., 2013).
Yet, even the most cursory glance at political discourses shows that political threats can take various forms and rely on a variety of stereotypical traits. For instance, the threat of political domination can be presented as more serious if it is claimed that it will occur even in the absence of harmful intentions by the outgroup (e.g., Scott, 1985). Likewise, it is possible to claim that the weakness or passivity of others is harmful to the ingroup, such as for instance when welfare recipients (immigrants or not) are depicted as lazy "parasites" living off the taxpayers' money. Drawing unconditional associations between specific stereotypical traits and the establishment or denial of political threat therefore seems to represent too simplistic an approach to capture that flexibility.
We contend, however, that predictions are possible if one also takes into account the context of strategic stereotype expression. As noted, one of our main goals in this paper is to illustrate this point in the context of immigrant stereotypes. Specifically, in our first study, we manipulated the argumentative context in which stereotypes were expressed, inducing participants to focus on the issue of whether immigrants pose a threat to job availability or to the resources of the social security system (by claiming unemployment benefits). As our opening example illustrates, a typical anti-immigration argument in relation to job availability consist in arguing that immigrants are "stealing our jobs". This argument can be supported by stressing that immigrants are hungry for work (e.g., the immigrant as "cheap labour", willing to do hard work for little pay), as this suggests they introduce a high degree of competition in the job market, and end up taking many jobs away from members of the host society. By contrast, a typical anti-immigration argument in the context of social security resources consists in arguing that immigrants are more likely to be unemployed and thereby to drain the resources of social security. This argument can be supported by depicting immigrants as "lazy profiteers" who do not actively look for work and are content to live off unemployment benefits, sustaining the notion that they are indeed more likely to be unemployed. Thus, in both cases, the key focus lies with the "hardworking" versus "lazy" quality of immigrants. Similarly, those who seek to promote a more pro-immigration stance can support it by using the same stereotypical traits but in a reverse pattern.
Those particular stereotypical traits offer some basis for prediction not merely because they form the basis of common arguments in relation to the above issues, but also because the flexibility with which they can be used is limited. Indeed, the argument that immigrants are stealing jobs cannot easily be supported by describing them as lazy, and the argument that they laze around on benefits by describing them as hardworking. By contrast, other stereotypical traits such as intelligence potentially enjoy more flexibility in the way in which they can be used in the same argumentative contexts. Thus the "benefit-profiteering" immigrant might be portrayed as possessing little intelligence or professional qualifications, but might also be portrayed as cunning in terms of knowing how to take advantage of the system. Both versions can serve the same purpose.
Our predictions therefore focused on variations along the "hardworking versus lazy" dimension, but where this dimension is to be distinguished from "intelligence" traits to which it is often associated (Fiske et al., 2002). For this reason, we rely on Osgood's nomenclature, as the distinction can conveniently be mapped on to the dimensions of Activity and Potency.
Similar distinctions proposed in the literature between motivation and ability (Carrier, Louvet, Chauvin, & Rohmer, 2014) and between brilliance and dedication (Bian, Leslie, Murphy, & Cimpian, 2018) lend further weight to the meaningfulness of the distinction, and could also be used to label the stereotypical traits at stake.

Study 1
In Study 1, we looked at the way in which participants of Portuguese nationality described immigrants as a function of whether they sought to support or oppose increased restriction on the amount of immigrants allowed into the country. To maximize the external validity of our findings, we relied on participants' pre-existing goals in relation to this issue.
However, to avoid variability in stereotypical descriptions being potentially reducible to stable inter-individual differences in political views, we also introduced two manipulations to achieve distinct predictions of strategic effects in stereotype expression.
Our first manipulation consisted in introducing the opportunity to mobilize an ingroup audience prior to, or subsequent to, stereotype expression. Simply put, only when a mobilization opportunity was provided prior to stereotype expression should participants be driven by the goal of mobilizing others through stereotype expression. Thus, prior to stereotype expression, we instructed half of our participants that their responses would be shown to a group of undecided Portuguese to assess the extent to which they could be swayed to sign a petition in line with the participants' position. Audience members were characterized as undecided to ensure that they would be perceived as highly susceptible to be mobilized and provide participants with sufficient motivation to use stereotypes for mobilization purposes. In addition, the use of undecided individuals served to neutralize the potential interference of complicating factors flowing from addressing audiences with existing positions on the issue (in such circumstances, additional strategic considerations may intervene).
Our second manipulation aimed to affect the nature of the particular stereotypical traits seen as instrumental to mobilize the audience in a particular direction. To achieve this, we manipulated the argumentative context in which the question of whether or not immigrants represents a threat arose. Specifically, we introduced the study as focusing on the extent to which immigrants do or do not represent a threat to the financial resources of the social security system, or to job availability for host society members. Note that in Portugal, as in most European countries, all workers (immigrant or not) must pay a contribution to social security, and that the latter provides a variety of government benefits, from health cover and state pensions to unemployment benefits. However, the text of our manipulation made it clear that the issue was specifically about whether immigrants have a positive or negative impact on social security funds through providing contributions as workers versus claiming unemployment benefits.
We chose these two particular issues for three reasons. First, in Portugal, as in most European countries, both issues are among the most commonly debated points of contention in the public arena about immigration (e.g., ACIDI, 2007, "David Cameron", 2014. Therefore, both issues could credibly be presented as key in determining positions on immigration restriction. Second, the "lazy" versus "hardworking" stereotypes of immigrant with which these issues are associated are familiar to the population at large, whether or not they are endorsed. This provided participants with the ability to rely on credible stereotypes whichever way they chose to argue. Third, the use of these issues allowed for predictions contrasting with expectations based on stable interpersonal differences in political goals or orientations. As noted, we predicted that participants with the same goals would use opposite ends of the "lazy versus hardworking" dimension as a function of which issue is at stake, insofar as it determines which stereotypical descriptions are instrumental in portraying immigrants as a threat.
More specifically, we hypothesized a three-way interaction between the direction of political goals, the timing of the mobilization opportunity, and the nature of the potential threat at stake, whose pattern should be as follows: H1. When the potential threat represented by immigrants bears upon social security resources, participants whose goal is to increase restriction on immigration should stress that immigrants are less hardworking when they are provided with a mobilization opportunity prior to stereotype expression, compared to when they are not (H1a). The opposite pattern should apply for participants whose goal was to oppose increased restriction on immigration (H1b).
H2. When the potential threat represented by immigrants bears upon job availability, participants whose goal is to increase restriction on immigration should stress that immigrants are more hardworking when they are provided with a mobilization opportunity prior to stereotype expression, compared to when they are not (H2a). The opposite pattern should apply for participants whose goal was to oppose increased restriction on immigration (H2b).
Asides from traits relating to the hardworking dimension, we also measured stereotypical ascriptions of intelligence and sociability (Potency and Evaluation). Although we had no predictions as to the impact of our manipulations on those traits, it was important to include them to verify that variability in the use of stereotypical content was indeed limited to the hardworking traits. This was necessary to confirm the validity of our rationale as to the importance of distinguishing those traits for making predictions in the contexts at stake.

Method
Participants. Two hundred and seventy-six participants took part on a voluntary basis. Three participants were discarded for missing values on the main dependent variable, leaving 273 participants (age range: 18-70, M = 25.2, SD = 8.07). Most participants were university students in various fields (70.3%), 20.1% were employed, 5.1% unemployed, and 4.4% were in other situation.
Since we had no previous similar studies to rely on, and no expectations as to the size of the hypothesized effects, sample size resulted from collecting data for a fixed period of time (i.e., one month), after which the effect size that the achieved sample size allowed to detect was assessed. A sensitivity power analysis indicated that the sample size of 273 obtained in that time allowed the detection of effects as small as ƒ 2 ≥ .029 with 80% power or ƒ 2 ≥ .048 with 95% power for the main multiple regression analyses (which included seven regression terms). This was deemed sufficient so that no further data was collected. No data analysis was performed before data collection ended.
Design. The study followed a Potential Threat (social security vs. job availability) × Mobilization Opportunity (opportunity vs. no opportunity) × Goal Direction (continuously measured, from opposing to supporting increased restriction on immigration) betweenparticipants design. To simplify, we refer to the levels of Mobilization Opportunity as opportunity versus no opportunity, with the understanding that this stands for the mobilization opportunity being introduced prior to or after stereotype expression. Our main dependent variable was the use of hardworking stereotypical traits to describe immigrants (Hardworking). All measures, manipulations, and exclusions in this study are disclosed either in the main text or in supplementary materials. All means, standard deviations, and correlations between the main measures are reported in supplementary materials.
Procedure. Individuals were approached in university premises and those interested in participating were asked to provide their email addresses. The experimenter subsequently sent an email with a hyperlink directing them to one of four versions of an online questionnaire. The experimenter randomly determined which hyperlink participants received and thereby which experimental condition they were assigned to. At the end of the questionnaire, participants were thanked and debriefed.
All measures and manipulations were part of the questionnaire itself. However, as part of the manipulation of Mobilization Opportunity, the order in which they were introduced varied across conditions. The exact chronological sequences for all conditions are represented in Figure 1. Overall, only the relative position of stereotypical ratings vis-à-vis the manipulations was altered as a function of Mobilization Opportunity.
Note that in all conditions we measured Goal Direction after the manipulation of Potential Threat, because it was possible for this manipulation to affect participants' position on immigration restriction. Therefore, measuring Goal Direction beforehand might have yielded an inaccurate estimate of participants' goals during stereotype expression. We address the possible issue of correlation between the measure and the manipulation in the results section.
Note also that, conversely, in the control conditions, stereotypical ratings preceded the manipulation of Potential Threat (as well as Mobilization Opportunity). This is because mentioning any of the political debates before stereotypical ratings could have made the political implications of stereotypes salient, and thereby potentially trigger a mobilization goal, even in the absence of a direct opportunity to mobilize a specific audience. We address the possible issue of interpretation this might create in the discussion. 1 Salience of national identity. Since the political goals under scrutiny were groupbased (i.e., relating to the perceived interest of the national ingroup as a whole), it was important to ensure that all participants were first primed to act in terms of their ingroup (national) identity. Thus, prior to introducing both experimental manipulations and stereotypical measures, national identity was made salient for all participants by requesting them to indicate their nationality and complete a scale of identification with being Portuguese (Verkuyten & Hagerdoorn, 1998; translated to Portuguese).
Manipulation of Potential Threat. Participants were informed that we were interested in their position on immigration restriction, and that the key issue in terms of determining people's positions on this issue was the positive or negative impact of immigrants on the job market versus the resources of the social security system. For each issue, we were careful not to take a position and stressed that both sides of the debate have been argued in the public domain.
Manipulation of Mobilization Opportunity. Prior to stereotypical ratings, participants in the "opportunity" conditions were given the opportunity to mobilize an ingroup audience towards their position. Specifically, they were provided with an open space to present their arguments in relation to immigration restriction, and were told that both their arguments and 1 Since both control conditions measured stereotypical ascriptions in the same conditions, it could be argued that they should be collapsed. We did run similar analyses to the ones presented below using a single collapsed condition and obtained similar results. For reasons of transparency, we present both design and analyses as they were originally intended. their answers on the stereotype scales would be shown to a group of undecided Portuguese to assess whether their responses could sway those individuals to sign a petition in line with their own position. The other half of participants were assigned to one of two control conditions, in which stereotypical ratings were applied at the very beginning of the questionnaire, that is, before the opportunity to mobilize an audience was introduced.
Measures. All items in the questionnaire were answered on 11-point Likert-type rating scales ranging from 0 to 10. We describe here the measures used in the analyses (additional measures are disclosed in supplementary materials).
Goal Direction. Three items measured the direction of political goals toward immigration restriction (α =.88): "Currently, do you think that immigration in Portugal should be increased or decreased?" (0 = increase a lot to 10 = decrease a lot), "Portugal has more immigration than is fair" (0 = strongly disagree to 10 = strongly agree), "Portugal should take more severe measures to reduce immigration" (0 = strongly disagree to 10 = strongly agree). Scores above the scale midpoint reflected the goal to increase immigration restriction. Because it could be argued that, despite the high overall alpha, disagreeing with increased restriction in items 2 and 3 is not fully equivalent to supporting increased immigration in item 1, we refer to participants scoring below the scale midpoint as opposing increased restriction, as this is inclusive of low scores on all 3 items.
Stereotypical traits. Stereotypical descriptions of immigrants were measured using ratings on the following traits (translated from Portuguese): a) "hardworking" and "resistant to hard working conditions" (Activity), b) "competent" and "intelligent" (Potency), and c) "friendly", "loyal", and "trustworthy" (Evaluation). We used only positive traits to limit the impact of social desirability considerations and their possible interference with strategic influences (see General Discussion). Participants were asked to rate each trait twice, the first time in absolute terms (i.e., "do you consider immigrants in general to be X?") and the second in comparative terms (i.e., "do you consider immigrants in general to be more or less X than the Portuguese?"). Comparative ratings were included because, in principle, the political arguments at stake involve a comparison with host society members. For instance, arguing that immigrants are a threat to job availability due to being hardworking only works if it means they are more hardworking than the Portuguese, since it is about establishing them as competition for the Portuguese. In practice, however, absolute ratings were likely to trigger such comparisons implicitly-indeed, self-categorization theory (Turner, Hogg, Oakes, Reicher, & Wetherell, 1987) contends that all stereotyping is inherently comparative, either explicitly or implicitly. Since Cronbach's alphas were satisfactory when using the comparative and absolute ratings together (for hardworking traits, α =.79; for potency traits, α =.84; for evaluation traits, α =.91), we used them as single scales to avoid repetition in the analyses.

Impact of Potential Threat on Goal Direction.
Prior to the test of our hypotheses, we first established that Potential Threat had a small but significant impact on the goal direction measure, F(1,271) = 4.24, p = .041, η 2 = .015, with participants in the job availability conditions being overall slightly more supportive of increased immigration restriction (M = 5.49, SD = 2.37) than participants in the social security conditions (M = 4.81, SD = 2.54). However, this effect was not a cause for statistical concern, since it was quite small (η 2 = .015), and VIF scores for all predictors in subsequent multiple regression analysis were all well below the most conservative criterion of acceptable range (i.e., all below 2.5, with the highest being 1.14). Furthermore, our hypotheses focused on interactions rather than on main effects, for which correlation between predictors is even less an issue both statistically (Friedrich, 1982) and in terms of interpretation (see Discussion). Nevertheless, to ensure no bias in interpreting the pattern of those interactions, we gave equal weight to both sides of Goal Direction in plotting the interaction for all conditions. Specifically, we used one standard deviation (SD) below and above the scale midpoint as low and high conditional values of Goal Direction, as opposed to the more usual values of one SD below and above the mean. The resulting simple slopes, therefore, illustrate the trends for participants with the goal to support or oppose increased immigration restriction in absolute terms, irrespective of the condition, rather than in comparison to other participants sharing the same level of Potential Threat. This also allowed for a more meaningful representation of our data (Aiken & West, 1991).
Stereotypical ratings. To test our hypotheses, we performed a multiple regression analysis with Hardworking as dependent variable, and Goal Direction (centered), Mobilization Opportunity (coded as -1/1), Potential Threat (coded as -1/1), and all interactions between these three variables as predictors (see Table 1). This analysis yielded a significant negative effect of Goal Direction: Overall, participants rated immigrants as less hardworking when their goal was opposing more restriction on immigration. The two-way interaction between Goal Direction and Potential Threat was also significant. Most importantly, however, both effects were qualified by the predicted three-way interaction between all regression terms.
Since the test of our hypotheses implied examining the impact of Mobilization Opportunity on participants sharing similar goals, we then calculated the simple slopes for the conditional effect of Mobilization Opportunity at different values of Goal Direction, in the social security and job availability conditions respectively (see Table 1 & Figure 2). The results indicated that, in the social security conditions, there was a significant two-way interaction between Goal Direction and Mobilization Opportunity. At high values of Goal Direction (denoting support for increased restriction), participants described immigrants as significantly less hardworking when provided with a mobilization opportunity compared to when they were not (supporting H1a). By contrast, at low values of Goals Direction (denoting opposition to increased restriction), participants described immigrants as more hardworking when provided with a mobilization opportunity compared to when they were not (supporting H1b). Note that the significant two-way interaction confirms that the simple slopes were significantly different from each other.
In the job availability conditions, the two-way interaction between Goal Direction and Mobilization Opportunity was also significant, but displayed the opposite pattern. At high values of Goals Direction, participants described immigrants as significantly more hardworking when provided with a mobilization opportunity compared to when they were not (supporting H2a). By contrast, at low values of Goal Direction, participants described immigrants as less hardworking when provided with a mobilization opportunity compared to when they were not (supporting H2b). In sum, the three-way interaction found in the overall equation emerged due to lower-level two-way interactions with opposing patterns in the social security and job availability conditions. Furthermore, the directions of these two-ways interactions were fully consistent with our hypotheses H1 and H2.
Finally, as regards stereotypical traits relating to Evaluation and Potency, using the same multiple regression procedure as above showed that the three-way interaction was not

Discussion
The main goal of Study 1 was to provide evidence that political intent affects stereotypical descriptions of the outgroup. To achieve that purpose, we provided our Portuguese participants with the opportunity to mobilize an audience before or after they rated immigrants on stereotypical traits. Our results show that this manipulation led to significant differences in the content of the expressed stereotypes, in a direction that was consonant with participants' political goals. This is consistent with the claim that political intent played a causal role in shaping stereotype expression.
Moreover, we aimed to show that the specific stereotypical traits used to defend a given position can strategically vary as a function of which stereotype is seen as most instrumental to achieve one's goal in a given situation. We sought to illustrate this flexibility by manipulating the specific immigration issue presented to participants as being most relevant to positions on immigration restriction. Our results show that this manipulation led participants with similar goals to express opposite stereotypes, in a direction that was fully consistent with our predictions. This significant interaction between pre-existing goals and immigration issues is important because it means that we can reject alternative explanations based on stable individual differences in cognitive representations-that is, the fact that pre-existing political positions or orientations might be consistently associated with the use of certain stereotypes.
Likewise, the interaction cannot be fully explained by invoking purely contextual effects, that is, that the use of different immigration issues triggered differences in the situational intrapsychic salience of stereotypes, which then led to differences in stereotype expression.
Nonetheless, it would still be possible to combine the above rationales and argue that the two immigration issues made different stereotypes cognitively salient for different individuals. Since exposure to the immigration issues occurred in conjunction with the mobilization opportunity, this could potentially account for the interaction. In the mobilization conditions, the issues of job availability might have made the "hardworking" stereotype salient and the issue of social security the "lazy" stereotype salient for participants supporting immigration restriction, and vice-versa for participants with the opposite position.
By contrast, in the control conditions, both manipulations took place after participants expressed stereotypes, so that such effect could not occur. Moreover, in the mobilization conditions, participants were able to elaborate their opinion about immigration in open format before rating immigrants. This might have further reinforced the cognitive salience of those stereotypes activated by the particular combination of context and individual differences.
Theoretically, the SIDE model does not deny the role of situational and/or cognitive factors on stereotype expression (Reicher et al., 1995). Rather, its strategic component predicts that, under conditions of mobilization, political goals should affect stereotype expression over and above the impact of those factors. Thus, our explanation does not deny that variations in the cognitive representations of immigrants might have occurred. Rather, it implies that if such variations did occur, they should not constitute the sole explanation for variations in stereotype expression. Therefore, the key difference between the two explanations does not lie with the presence or absence of variation in cognitive salience, but in whether or not political intent is also necessary to account for the expressed stereotypes. In addition, variations in cognitive salience themselves may be the product of cognitive work driven by political intent (see Study 2).
Since both explanations can account for the results and for possible variations in cognitive salience, the question is which interpretation should then be preferred, based on other criteria. In this context, we note that the non-strategic interpretation possesses a number of significant disadvantages. First, it requires arguing that, compared to pro-immigration participants, anti-immigration participants were more predisposed (as a matter of stable differences in cognitive representations) to endorse the representation of immigrants as more hardworking and as more lazy. While individuals may possibly hold ostensibly contradictory beliefs, this is certainly less plausible than claiming that political goals led each stereotype to become salient in the relevant context. As we shall argue shortly, the influence of goals and motivations on cognitive salience is a well-established phenomenon. Second, the alternative explanation is less parsimonious, since it requires postulating the existence of two distinct pre-existing cognitive structures to replace the role of political intent. Third, it treats the particular content of both cognitive representations and expressed stereotypes as mere happenstance, rather than as driven by a rationale, thereby leading to a loss of explanatory power. Finally, the logic is open to post-hoc rationalizations as to the relevant features of the context and pre-existing cognitive structures that lead to a specific pattern of results. By contrast, a strategic framework provided the general rationale as to which features should matter, on which basis we were able to predict the results a priori (see also General Discussion).
Nevertheless, in Study 2, we sought to look at the cognitive processes underlying strategic stereotype expression, which provided an opportunity to test the alternative explanation more directly. In addition, we also sought to address the related issue of the overlap between the introduction of the specific immigration issues and of the mobilization opportunity, given that it helped to lend credence to this alternative.

Study 2
In Study 2, our main goal was to look at the cognitive processes that underlie the political/strategic expression of stereotypes. Specifically, we aimed to clarify the relation between the content of publicly expressed stereotypes and the content of the private, intrapsychic, cognitive representations of the mobilizer.
Much research has shown the resilience of stereotypical beliefs (e.g., Snyder, 1981) and the fact that they may persist privately despite social and cultural pressures to suppress or moderate their expression (e.g., Dambrun & Guimond, 2004;Devine, 1989;Dovidio, Kawakami, & Beach, 2001). On that basis, one could conceive of political/strategic influences on stereotype expression in similar terms, that is, as affecting only the outward expression of stereotypes and leaving untouched the more stable, private cognitive representations of the stereotype user. However, there is also a considerable amount of research showing that information processing in general, and cognitive stereotyping in particular, can be influenced by motivations and goals (e.g., Kunda & Sinclair, 1999), even at the implicit level (Blair, 2002). For instance, the motivation to arrive at a particular representation may lead individuals to selectively scan their memory in search of examples that support the desired impression while discarding those that do not (Sanitioso, Kunda, & Fong, 1990), or may affect the way in which specific instances are seen as typical or else subtyped as unrepresentative exceptions (Carnaghi & Yzerbyt, 2007). Similarly, the existence of a mobilization goal might trigger selective intra-psychic work affecting the cognitive representation of immigrants. For instance, individuals might give different weights to individual instances or immigrant subgroups when forming the prototype of immigrant as a whole. 2 From a political and strategic point of view, there are in fact good reasons to expect this process to occur, and to postulate that it constitutes one of the possible mechanisms underlying variability in stereotype expression. As Kunda pointed out, in relation to motivated reasoning, "people are more likely to arrive at conclusions that they want to arrive at, but their ability to do so is constrained by their ability to construct seemingly reasonable justifications for these conclusions" (1990, p.480; see also W. Klein & Kunda 1992).
Likewise, from a SIDE perspective, it is precisely because stereotypes provide the "reasonable justifications" of political arguments that they are used for political purposes.
However, stereotypes can only play this role if they are deemed to reflect social reality. Thus, potential mobilizers have an interest in recruiting the particular cognitive resource (such as instances of behavior confirming the stereotype) that allows them to substantiate and legitimize their claims. Furthermore, when addressing an ingroup audience, mobilizers generally expect that the audience will agree with them in their assessment of social reality (Haslam, Turner, Oakes, McGarty, & Reynolds, 1998). This should lead them to assume a priori that the best arguments to convince that audience are those that persuade themselves.
Therefore, the default option in terms of formulating arguments should be to build upon their private beliefs. Psychologically, this should lead to the strategic examination of those private beliefs, which in turn may lead to variability in immigrant representation at the cognitive level.
To investigate this question, Study 2 used a design similar to Study 1, but focused only on the issue of social security, and introduced two important additions. First, we developed an adaptation of the Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998) to assess the effect of the manipulations on the intra-psychic cognitive representations of immigrants. Second, we added a manipulation of cognitive load to affect the ability of individuals to do cognitive work. We reasoned that the activation of mobilization goals should lead to motivated information processing (or strategic thinking) about immigrant stereotypes, resulting in variability in intra-psychic representations. In turn, the content of intra-psychic representations should shape the content of the stereotypes overtly expressed. However, cognitive load should impede that process and prevent variability from appearing at both the explicit and implicit levels. More specifically, given that the issue at stake was the impact of immigrants on social security resources, we predicted that: H1: When provided with the opportunity to mobilize an audience, and when able to do cognitive work, participants supporting increased restriction on immigration would show a weaker implicit association between immigrants and hardworking traits compared to participants who had no mobilization opportunity, while participants opposing increased restriction would show a stronger implicit association. By contrast, we expected no difference among participants under cognitive load, whether or not they were provided with a mobilization opportunity.
H2: The same pattern of three-way interaction was expected to occur for overt stereotypical ratings.
Asides from testing our hypotheses, the manipulation of cognitive load also enabled different predictions for a strategic explanation of stereotype variability compared to one based purely on the combination of situational and cognitive effects. For reasons to which we shall return (see General Discussion), this alternative poses a more critical challenge to a strategic interpretation if those effects are deemed to reflect the action of purely reactive and automatic processes. This implies that, provided that participants are not under load when stereotypes are psychologically activated, variability in both implicit and explicit stereotyping should occur independently from the presence or absence of cognitive load. In addition, we also modified the design of the control conditions to assess the impact of receiving a mobilization opportunity independently from the impact of manipulating the focus of mobilization attempts.
Sample size was determined by power analysis using G-POWER. Since the closest effect to the hypothesized three-way interaction in this study was the two-way interaction in the social security conditions of Study 1 (ƒ 2 = .09), we used a commensurate but more conservative (for a higher order effect) value of ƒ 2 = .05 as assumption. The analysis indicated that a minimum sample size of 159 was required to detect such effect with a power of 0.80. We then added 8 participants (2 per condition) to cover for potential missing values. Procedure. The study took place in a laboratory with participants randomly assigned to conditions. All measures and manipulations were applied through the questionnaire itself, except for the IAT. As in Study 1, however, their order varied as a function of the manipulation of Mobilization Opportunity. The exact chronological sequences for all conditions are represented in Figure 3.
Regarding the cognitive load manipulation, our reasoning implied that (in the relevant conditions) cognitive load should be applied and maintained throughout the phase during which participants would otherwise be able to think strategically about what stereotypes to express. This included the time during which participants could openly present their arguments (as part of the manipulation of Mobilization Opportunity), as well as the completion of the explicit measure of stereotyping themselves, Thus, for the "opportunity/load" condition, we started cognitive load after the mobilization opportunity had been explained (so that the mobilization goal was activated), but before participants could expose their arguments openly, and maintained it until after the explicit measure of stereotyping. In the "no opportunity/load" condition, we applied cognitive load during the explicit stereotypical ratings only, as the mobilization opportunity was introduced after stereotypical ratings and the IAT, so that any strategic cognitive work that might take place at that point could not affect our dependent variables. In both conditions, cognitive load ended right after the completion of the explicit stereotypical ratings, so that no participant was under cognitive load during the IAT or the rest of the questionnaire.
Note that, although our reasoning was that variability in implicit associations should precede variability in overt expression, in all conditions the IAT was applied after the explicit measure of stereotyping, because it might otherwise have interfered with and/or been affected by the cognitive load manipulation. Admittedly, this made it possible for implicit associations to be influenced by the act of filling out the explicit measure, rather than vice-versa (see General Discussion). However, since this was constant in all conditions, any differences in IAT scores between conditions can still be attributed to the manipulations.
Salience of national identity. National identity was made salient at the beginning for all participants through the same means as in Study 1.

1.
Manipulation of Cognitive Load. We used a classic manipulation of cognitive load (e.g., Gilbert & Osborne, 1989) to affect participants' ability to do cognitive work prior to and during stereotype expression. Participants in the load conditions were presented with an 8-digit number for 30 seconds and asked to rehearse the number without attempting to memorize it, until the experimenter asked them to write it down.
Measures. All items in the questionnaire were answered on 11-point Likert-type rating-scales ranging from 0 to 10.
Goal Direction. The same three items as in Study 1 were used (α = .91).
Implicit measure of stereotyping. Our IAT used as its two sets of categories Portuguese versus Immigrant and Hardworking versus Lazy. The first set of stimuli was composed of 12 (6x2) first names that were either recognizably of Portuguese origin (e.g., Pedro, Leonor) or immigrant origin (e.g., Svetlana, Xing: different origins were used, consistent with the most prominent immigrant subgroups in Portugal). The second set of stimuli was composed of 12 (6x2) words semantically associated with hardworking (e.g., persistent, active) and lazy (e.g., procrastinating, inactive), selected for their degree of semantic association and familiarity based on pre-test results.
The test was organized in seven trial blocks as in the standard IAT, and contained a total of 216 trials. Following the recommended procedure (Greenwald, Nosek, & Banaji, 2003), D scores were computed from the raw response times. Since we had no a priori assumptions as to which associations between categories and stimuli should be considered as congruent or incongruent, we computed D scores in such a way that they followed the same direction as explicit stereotypical rating. That is, the higher the D score, the more immigrants were associated with the hardworking category and the fewer with the lazy category (and vice-versa for the Portuguese).

Results
Explicit stereotypical ratings. We begin by reporting the results on the explicit ratings, because it was important first to check that the pattern found in Study 1 in the social security conditions was replicated for participants under no cognitive load. Thus, to test H2, we performed a multiple regression analysis with Hardworking as dependent variable, and Goal Direction (centered), Mobilization Opportunity (coded as -1/1), Cognitive Load (coded as -1/1), and all interactions between these three variables as predictors (see Table 2). The analysis yielded a significant negative effect for Goal Direction, a significant negative effect for Mobilization Opportunity, and a marginally significant two-way interaction between Goal Direction and Mobilization Opportunity. As predicted, these effects were qualified by a significant three-way interaction between all predictors.
Note that, as in Study 1, this three-way interaction was not significant for Evaluation  Table 2 & Figure 4). The analysis showed that, under cognitive load, the two-way interaction between Goal Direction and Mobilization Opportunity was not significant. The simple slope at low values of Goal Direction did come out as marginally significant (p = .069), but the lack of interaction means that this slope was not significantly different from the slope at low levels of Goal Direction. However, when participants were not under cognitive load, the two-way interaction between Goal Direction and Mobilization Opportunity was significant. At low values of Goal Direction (i.e., opposition to increased restriction), participants described immigrants as more hardworking when provided with a mobilization opportunity compared to when they were not, although the simple slope was not significant. By contrast, at high values of Goals Direction (i.e., support for increased restriction), participants described immigrants as significantly less hardworking when provided with a mobilization opportunity compared to when they were not.
Overall, then, this pattern provides support for H2, and replicates the results of the social security conditions of Study 1, albeit with one caveat. Unlike in Study 1, the simple slope representing the effect of Mobilization Opportunity for participants opposing increased immigration restriction was not significant. Of course, conditional values are merely points on a continuum chosen to illustrate the direction of the interaction, and the simple slope for participants became significant when using even lower conditional values of Goal Direction (the region of significance began at -1.30 SD below the scale midpoint, still well within the range of actual answers). However, this result indicates that the two-way interaction under no load conditions was driven more by the ratings of participants supporting more immigration restriction than by their counterparts. Possibly, this may be because participants who opposed increased restriction included those who merely supported the status quo as well as those supporting a decrease in restriction. The former might have used more "neutral" stereotypical ratings to deny rather than truly contradict the idea that immigrants are a threat. At any rate, since this is a slight difference in pattern with Study 1, it suggests that strategic effects for those participants were less stable than for their counterparts.
Implicit associations. To test H1, we performed the same analysis described above on IAT D-scores, with highly similar results (see Table 3). The overall regression equation revealed a significant negative effect for Goal Direction, a marginally significant two-way interaction between Goal Direction and Mobilization Opportunity (p = .060), and the expected, significant three-way interaction between all predictors.
Simple slopes computation showed that, when participants were under cognitive load prior to and during stereotype expression, the two-way interaction between Goal Direction and Mobilization Opportunity was not significant. However, this interaction was significant when participants were not under load. At low values of Goals Direction (i.e., opposition to increased restriction), participants associated immigrants more (and Portuguese less) with the hardworking category when provided with a mobilization opportunity compared to when they were not, though the simple slope was not significant (p = .102, region of significance beginning at -1.17 SD). By contrast, at high values of Goal Direction (i.e., support for increased restriction), participants associated immigrants significantly less (and Portuguese more) with the hardworking category when provided with a mobilization opportunity compared to when they were not.
In sum, as predicted, participants in the critical mobilization/no load condition showed variability in their implicit stereotypical associations consistent with their overt stereotypical expression. As with explicit stereotyping, it is not clear how much participants opposing more restriction contributed to this effect. However, this fact is in itself consistent with the pattern found on the explicit ratings, thus supporting our hypothesis of a consonance between the two. It should, however, be noted that results on the IAT did differ significantly from explicit ratings in absolute terms. That is, in all conditions, D scores were significantly below the midpoint, indicating a general tendency for weaker associations between being immigrant and hardworking than between being Portuguese and hardworking. In contrast, explicit ratings of immigrants as being hardworking were significantly above the midpoint, even when considering only the comparative items (a series of t tests easily confirmed this).
Differences in the tools might perhaps partly account for this disparity. However, since this finding is consistent with existing research on implicit stereotyping (Dovidio et al. 2001), in all likelihood it indicates that explicit ratings were also affected by social norms against negative stereotyping. We come back briefly to the issue of social desirability in the General Discussion.

General Discussion
While the results of Study 1 provided evidence of variability in overt stereotyping for participants given the opportunity to mobilize an audience, Study 2 showed that, under the same conditions, a consonant variability occurred in the implicit associations of mobilizers when they were able to make full use of their cognitive resources. Our interpretation of this finding is that individuals motivated by a political goal of mobilization, and in possession of all their cognitive resources, engaged in active cognitive work that resulted in the formation of a psychological representation of immigrants consonant with their goal. Furthermore, since variability in overt expression also disappeared under cognitive load, it suggests that strategic effects in overt expression may be dependent upon such cognitive work, that is, upon the ability to construct private representations that can successfully shape and sustain the stereotypes one wishes to express publicly.
This last conclusion should be qualified in two respects, however. First, the fact that we had to apply the IAT after explicit stereotyping means that it is also possible that variability in cognitive representations resulted from the strategic expression of stereotypes (similar to a "saying is believing" effect, Higgins & Rholes, 1978). Theoretically, our position is that the relation between internal thinking and external arguing is bidirectional (Billig, 1987). Therefore, we do not dismiss the possibility that such a process might also occur or even co-exist with its counterpart. Second, the necessity of performing strategic cognitive work in order for strategic expression to occur may also be relative to the specific setting of our studies and background of our participants. Specifically, we relied on immigration issues familiar to participants, but it is unlikely that many participants were familiar with the experience of mobilizing an anonymous audience on the issue of immigration restriction. Thus, while participants could rely on existing resources (i.e., common immigrant stereotypes) to form a cognitive representation adequate to their purposes, it is likely that most of them required doing this online. In contrast, those who are more experienced in mobilizing on a specific issue (such as dedicated activists) might well rely on the results of previous cognitive work, and thus be less sensitive to pressures on cognitive resources. Also, when the motivation to mobilize an audience on an issue is paramount, it might possibly override the need to construct private representations consonant with the political goal. Again, this was unlikely to be the case for the majority of our participants, but when those conditions are present, strategic variability in stereotype expression might well occur despite a lack of cognitive resources and/or lead to the stereotypes expressed being at odds with the private beliefs of the stereotype user.
Our interpretation, therefore, represents only one of the possible cognitive paths leading to the strategic expression of stereotypes. As discussed previously, there are good reasons to believe that forming or recruiting internal representations consonant with one's political goal is the preferable option when it can be successfully accomplished. Thus, in all likelihood, it is a path commonly taken in spontaneous contexts. Nevertheless, further research should clarify the conditions under which this process is or is not necessary for strategic expression to occur, and the part it plays in relation to other possible processes.
The results of Study 2 also allow rejecting more categorically the alternative explanation left open by Study 1. As a reminder, this alternative posited that exposition to a particular immigration issue, coupled with pre-existing individual differences in political positions, might have automatically triggered differences in the cognitive representations of immigrants-differences which in turn would be fully responsible for the variations in overt stereotyping. This predicted that variability in both implicit and explicit stereotyping should occur even when participants were under cognitive load, and regardless of the manipulation of mobilization opportunity. That prediction was not supported by the results.
Of course, one might still argue that the cognitive effects of exposition to a specific immigration issue, and of the opportunity for opinion elaboration that ensued, need not be automatic. However, once non-automatic processes are invoked, the possibility that participants conducted motivated reasoning during opinion elaboration cannot be rejected a priori. Participants had the control and cognitive resources required, as well as a relevant reason, to elaborate in a way that supported their goals. Therefore, to maintain the alternative explanation in those conditions requires arguing why their thought process nevertheless remained impervious to their political intent while still yielding results consistent with it.
That is, not only does it necessitate making auxiliary assumptions about pre-existing individual differences in cognitive structures (see Discussion of Study 1), it also requires the addition of a rationale justifying the separation of participants' cognitive and motivational processes, without the help of the notion of automaticity. In other words, it involves rejecting the most direct and obvious candidate in favor of more convoluted alternatives, without any apparent need to do so, and without any theoretical or empirical gain, since political intent does not only account for, but was able to predict, the results.
Overall, then, and in line with the three goals we defined at the outset, our studies contribute to extend existing research on the SIDE model in three ways. First, to our knowledge, they represent the first experimental demonstration of SIDE's claim that people tailor their representations of the outgroup to mobilize their audience. For this reason, they provide stronger evidence for the causal role of political intent compared to previous SIDE studies based on interviews or archival data (Klein & Licata, 2003;Reicher et al., 1997).
Second, we showed the possibility of predicting the content of strategic stereotype expression by combining SIDE with stereotype content models as well as with an analysis of the context of stereotype expression. This also complements the aforementioned SIDE studies, which relied on a posteriori content analysis. Third, our studies and general approach pave the way towards a better integration of the cognitive and strategic sides of SIDE (Reicher et al., 1995).
The third point deserves further comment, given that this connection between cognitive and strategic factors occurs at several levels. At the most immediate empirical level, the results of Study 2 suggest that strategic goals affect not only the overt expression of stereotypes, but can also facilitate their cognitive activation at an early level of processing.
That is, although we reject an explanation based purely on contextual cues and individual differences, a strategic explanation is otherwise perfectly compatible with the idea that these factors might have been associated with differences in the cognitive salience of stereotypes.
As noted, the key difference merely lies with the claim that political intent is also necessary to account for the results.
However, this implies a broader difference in terms of how we conceive of the impact of contextual cues and existing cognitive structures on stereotype expression. Rather than seeing participants as passively guided by those features, a strategic approach claims that context and pre-existing cognitive structures provide both the resources and the constraints within which participants' goals actively shape both their thinking and stereotype expression.
Otherwise put, it proposes to conceive of context and cognitive structures as constraints of action rather than as deterministic triggers of reactions.
In this perspective, strategic and situational/cognitive factors do not necessarily contradict each other, but can even actively interact. For instance, pre-existing cognitive structures may affect the ease with which mobilizers are able to construct a strategically adequate internal representation possessing "reasonable justifications", and thereby the extent to which they are confident to express certain stereotypes for strategic purposes. Likewise, pursuing political mobilization can also interact with other motivational influences in shaping stereotype expression. Thus, expressing stereotypes that go too strongly against accepted social norms of desirability can be avoided not only because of the negative self-image it projects, but also because it might affect success at mobilization.
Finally, at an even broader theoretical level, cognitive and strategic factors also intertwine because mobilization can only be achieved if stereotypes possess the cognitive reality necessary to shape people's attitudes and actions. In the specific context of our studies, our argument implies that: (a) it is because individuals are cognitively dependent on stereotypes to determine the extent of outgroup threat that immigrant stereotypes can be used to mobilize audiences towards immigration restriction, and (b) it is because mobilizers are aware of this cognitive dependency that they seek to capitalize on it. Moreover, mobilization will potentially be more successful if it takes into account the nature of the existing beliefs in the audience. Once again, this puts constraints on the flexibility of strategic stereotype expression. This is why mobilizers typically seek to build upon existing socio-cognitive realities in making their case (Klein, Clark, & Lyons, 2010).
Certainly, whereas previous SIDE studies focused on the content of stereotypes spontaneously expressed in naturalistic settings, our use of a laboratory setting means that caution should be exercised in terms of making direct empirical generalizations about the content of strategic stereotype expression. If the specifics of politicized stereotypes depend on the resources and constraints of a particular context, they should vary when those constraints and resources change. There is little doubt that, outside our particular experimental setup, other stereotypes can be used to support the goals of reinforcing or relaxing constraints on immigration. Yet, generalizations are possible based on our theoretical interpretation of the processes involved. That is, one can acknowledge the creativity and rhetorical flexibility in which stereotypical contents can be deployed in naturalistic settings, while maintaining that it is possible to predict general trends based on how the context shapes and constrains both the arguments that can be used, and the specific stereotypical contents that support them. In our view, it is certainly not a coincidence, for instance, that both the images of the "lazy" immigrant and of the immigrant as a "cheap hardworker" are widespread across societies characterized by a significant amount of economic immigration (e.g., Enesco, Navarro, Paradela, & Guerrero, 2005;Van Dijk, 2000). We suggest that the reason for this must be related with the fact that immigration triggers similar political debates across those countries, creates similar political goals, and provides similar constraints on the nature of the tools deployed to accomplish them. The next step should be to work towards a broader formalization of the relationships between political goals, context of expression, and stereotypical contents. For instance, passive stereotypical traits may be used to depict the outgroup as a threat when sharing a common superordinate group membership, space, and institutions with that outgroup (such as is the case with immigration), but in other circumstances a pro-active response against the outgroup may indeed require depicting it as actively harmful. On its own, a political and strategic framework is but one step towards accomplishing that task, but it does provide a general rationale that should help identifying the relevant features of contextual and socio-cognitive realities and predicting the way in which these are likely to affect the content of politicized stereotype expression.
Finally, our approach also offers the possibility to integrate evidence of stability and of variability in stereotype expression. Indeed, SIDE studies showing the flexibility in stereotyping (Klein & Licata, 2003;Reicher et al., 1997) contrast with many others that have found great commonality and stability in stereotypical contents across time and cultures (e.g., Cuddy et al., 2009). In part, the apparent contradiction might result from the fact that the former focus on spontaneous stereotype expression in politicized contexts, while the latter actively elicit stereotyping, in conditions abstracted from any specific political implications, and without a clearly defined audience. In other words, both approaches are selective in the type of empirical phenomenon they base their claims on. Theoretically, however, a political/strategic framework is not limited to explaining variability in stereotype expression.
Rather, it also suggests an alternative mechanism through which stereotypes can be maintained and perpetuated. That is, stability in the contents of spontaneous expression may reflect stability in cognitive structures, or be the product of more or less universal motivational biases, but it can also emerge from a continuity in the political tasks (Geertz, 1995) that stereotypes are called upon to serve.       Oppose more restriction Support more restriction