Journal of Personality and Social Psychology 2003, Vol. 84, No. 4, 697–721
Copyright 2003 by the American Psychological Association, Inc. 0022-3514/03/$12.00 DOI: 10.1037/0022-35184.108.40.2067
Does Social Dominance Generate Prejudice? Integrating Individual and Contextual Determinants of Intergroup Cognitions Serge Guimond, Michae¨l Dambrun, Nicolas Michinov, and Sandra Duarte Universite´ Blaise Pascal, CNRS Social dominance orientation (SDO) has been proposed as an important variable in the explanation of prejudice. We distinguish between three conceptualizations of SDO: SDO as a personality trait (personality model), SDO as a moderator of the effects of situational variables (Person ⫻ Situation model), and SDO as a mediator of the effect of social position on prejudice (group socialization model [GSM]). Four studies (N ⫽ 1,657) looking at the relations between social positions, SDO, and prejudice in a natural setting and in a laboratory setting provide strong support for the GSM. In contrast to previous correlational findings, there is evidence of a cause (dominant social position), an effect (prejudice increases), and a mediator (SDO). These results suggest new perspectives on the integration of individual and contextual determinants of prejudice.
society can be characterized by the existence of group-based hierarchy in which at least one group is dominant over others and enjoys a disproportionate share of privilege, and at least one group occupies a subordinate position. Three basic types of group-based social hierarchies are distinguished as follows: an age system, a gender system, and an arbitrary-set system. Arbitrary-set systems, with which we will be concerned here, consist of hierarchies of socially constructed groups based on any socially relevant group distinctions such as ethnicity, social class, or religion. SDT suggests that within these systems, there are groups and institutions that promote cognitions either reinforcing, or to the contrary, attenuating group inequality (Van Laar & Sidanius, 2001). These cognitions or ideologies are called “legitimizing myths” and a basic distinction is made between hierarchy-enhancing (H-E) legitimizing myths whose main function is to legitimize group inequality (e.g., racism, sexism, conservatism) and hierarchyattenuating (H-A) legitimizing myths and institutions seeking to legitimize group equality (e.g., socialism, feminism, universal rights of man). The theoretical and empirical contribution of this approach rests, in major aspects, in the construct of SDO, which is defined as “the degree to which individuals desire and support group-based hierarchy and the domination of ‘inferior’ groups by ‘superior’ groups” (Sidanius & Pratto, 1999, p. 48). SDO is proposed as the single most important variable to account for the acceptance or rejection of cognitions that promote or attenuate inequality. It is the motivation that is assumed to drive people’s tendency to adopt certain beliefs, attitudes, or values, and to seek out membership in groups or organizations that will reinforce (H-E) or attenuate (H-A) group-based inequality. In support of this conceptualization, Pratto et al. (1994) have developed a measure of SDO and shown that scores on this measure reliably predict an extremely wide range of ideologies and political beliefs. For example, those who score higher on SDO are more prejudiced, more conservative, more favorable toward the military, and more patriotic, whereas those who score lower are more favorable toward women’s rights, gay rights, and social programs in general. Other researchers have confirmed these findings, especially by showing that SDO often comes out as one of the primary predictors of both prejudice (see
Understanding the social, cognitive, and motivational processes that contribute to intergroup perceptions and intergroup conflict represents an important goal of research in social psychology. One of the most enduring issues that researchers in this area have dealt with concerns the relative importance of personality predispositions and situational factors in the explanation of prejudice (Brewer & Brown, 1998; Brown, 1995; Duckitt, 2001; Fiske, 1998, 2000; Turner, 1999a; Verkuyten & Hagendoorn, 1998). Is prejudice part of the personality make-up of the individual? Is it the result of social forces acting on the individual? Or, does a Person ⫻ Situation model provide a better theoretical framework for understanding prejudice? The relative importance, and possible integration, of an intraindividual analysis with a contextual analysis has remained up to this day one of the central themes of the theoretical analysis of prejudice (Duckitt, 2001; Fiske, 2000). The present research addresses this basic issue in relation to the construct of social dominance orientation (SDO). This construct has been proposed within social dominance theory (SDT), a major integrative framework seeking to understand and explain group-based prejudice and oppression (see Pratto, 1999; Pratto, Sidanius, Stallworth, & Malle , 1994; Sidanius, 1993; Sidanius & Pratto, 1999). SDT postulates that every complex
Serge Guimond, Michae¨l Dambrun, Nicolas Michinov, and Sandra Duarte, Laboratoire de Psychologie Sociale de la Cognition, Universite´ Blaise Pascal, CNRS, Clermont-Ferrand, France. This article was presented, in part, at the group meeting of the European Association of Experimental Social Psychology on “The psychology of domination,” Grenoble, France, March 2000, and at the general meeting of the European Association of Experimental Social Psychology, San Sebastian, Spain, July 2002. This research was supported by ACI Cognitique Grant FNS 2001. We are grateful to Susan Fiske and Michael Schmitt for their useful comments on an earlier version of this article. Correspondence concerning this article should be addressed to either Serge Guimond, Michae¨l Dambrun, or Sandra Duarte, Laboratoire de Psychologie Sociale de la Cognition, Universite´ Blaise Pascal, UMR 6024, CNRS, 34 Avenue Carnot, Clermont-Ferrand 63000, France, or to Nicolas Michinov, who is now at the Departement de Psychologie, Universite´ de Poitiers, 99 Avenue du Recteur Pineau, B.P. 632, Poitiers Cedex F-86022, France. E-mail: [email protected] 697
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
Altemeyer, 1998; Esses, Jackson, & Armstrong, 1998; McFarland, 1999) and support for extreme right-wing political parties (Dambrun, Maisonneuve, Duarte, & Guimond, 2002). These findings are important, but they leave several fundamental questions unanswered. As Schmitt, Branscombe, and Kappen (in press) have recently argued, there is at present little evidence supporting the idea of a causal relation between SDO and prejudice. Moreover, the question of where SDO comes from has not been thoroughly researched (Sidanius & Pratto, 1999). As a result, the theoretical status of SDO is open to various interpretations (see Duckitt, 2001). In the present research, we distinguish (and test) three different conceptualizations of the way SDO operates in the explanation of prejudice. We call the first conceptualization the personality model of prejudice. In this approach, researchers focus on SDO as a personality trait, a psychological characteristic independent of the individual’s position in the social structure. However, an alternative approach is offered by the Person ⫻ Situation model, which argues that SDO moderates the effects of situational variables. A third conceptualization that we call the group socialization model (GSM) suggests that SDO may function as a mediator of the effects of social position on prejudice. By social position, we mean the location of people within the social structure, with membership in groups at the top of the social hierarchy, within H-E organizations, characterized by a relatively large share of such things as wealth, political power, and high status being labeled dominant social positions (Sidanius & Pratto, 1999). Considering these three perspectives simultaneously is a challenge that has not been taken in the past but one that is likely to extend in significant ways previous theory and research on the role of social dominance in intergroup relations. After outlining these three approaches and showing that each imply different and specific predictions about the relations between SDO, social position, and prejudice, we present the results of four studies that examine their relative merits.
The Personality Model of Prejudice Reynolds, Turner, Haslam, and Ryan (2001) have referred to a recent upsurge in research on what they called the “personality approach to prejudice,” citing studies using the SDO and RightWing Authoritarianism (RWA) scales as prime examples of such a tendency. According to this personality approach, SDO is conceived, perhaps mistakenly, as will be argued below, as an individual characteristic, relatively immutable across situations, that can explain prejudice. As Reynolds et al. (2001) stated, in this approach, “the character of the individual person is used to explain prejudice independent of immediate social contextual factors” (p. 428). Research by Altemeyer (1998), McFarland (1999), and Whitley (1999) provided independent support for this model, which suggests that two main personality variables, SDO and RWA, can explain prejudice. Thus, after a series of research revealing that, together, SDO and RWA account for more than 50% of the variance in prejudice, Altemeyer (1998) concluded that “if you want to explain the many kinds of prejudice in this situation, they are largely matters of personality. And only two kinds of personality are basically involved: The social dominator and the right-wing authoritarian” (p. 60). Whitley offered a similar theoretical and empirical contribution, arguing that these two constructs are “causes of prejudice,” with SDO being the predominant one. He provided a clear answer to the question of personality and
prejudice in the following terms: “Are there people who ‘don’t like anybody very much?’ . . . Two personality-based perspectives— right-wing authoritarianism and social dominance orientation— suggest that the answer is yes” (Whitley, 1999, p. 126). McFarland (1999) has also made extensive research in this area giving general support to his view that “some persons are more prejudice prone than are others” (p. 4). More recently, Heaven and Quintin (in press) claimed to have found strong evidence “supporting the views of Sidanius (1993) and Altemeyer (1998) that prejudice is largely a matter of individual differences in RWA and SDO” (p. 8). In sum, research within this personality model suggests that SDO, much like RWA, is an individual characteristic that can explain prejudice. In theory, as in research, this approach is not concerned with possible effects of situational variables on prejudice. Although we will discuss other possibilities below, because Pratto et al. (1994) stated in the very title of their article that SDO is a “personality” variable and because it is proposed as a central factor in prejudice, it is understandable that such a personality approach has grown in recent years. Snyder and Cantor (1998) defined the “dispositional strategy” for the study of social behavior as seeking “to understand consistencies in social behavior in terms of stable traits, enduring dispositions and other properties that are thought to reside within individuals” (p. 636). They consider SDO as a recent example of such a strategy. Clearly then, there are reasons to believe that SDO is part of a personality approach to prejudice, and it is considered as such in recent reviews of the literature (see Fiske, 2000; Jones, 2002; Nelson, 2002). Historically, the fundamental goal of such a personality approach, personified and upheld in recent years in the Big Five tradition, has been to identify “stable dispositions that remain invariant across situations and that are distinctive for the individual” (Mischel & Shoda, 1998, p. 231). Applied to the explanation of prejudice, such a personality model has been subjected to several criticisms that question its validity on methodological, theoretical, and empirical grounds (see Billig, 1976; Brown, 1995; Duckitt, 2001; Pettigrew, 1958; Reynolds et al., 2001; Tajfel & Turner, 1986; Verkuyten & Hagendoorn, 1998). These criticisms are well known (see Jones, 2002; Nelson, 2002). Most of them, however, have been applied to other personality models, and the extent to which they are valid for SDO has been largely unexplored empirically. For example, if SDO is an enduring disposition, how stable and enduring is it? Sidanius and Pratto (1999) have found that SDO is “highly stable over time” (p. 45). But is it stable across situations? This is an empirical question that can only be decided on the basis of empirical evidence, and we will address this issue. It is important to note, however, that there are reasons to think that Sidanius and Pratto (1999) did not conceive SDO in line with the above personality model. They stated that “our present focus on SDO is not meant to imply that all phenomena related to prejudice and group conflict can be solely understood or reduced to individual differences. In fact, social dominance theory implies that SDO and other individual variables must be considered within their social context” (Pratto et al., 1994, p. 757). Indeed, SDT may be more properly considered as an attempt to integrate an individual-difference variable such as SDO within a contextual analysis of prejudice and intergroup relations. Still, social dominance theorists have not been very specific so far as to how such an integration can be conceptualized and empirically tested. We
SOCIAL DOMINANCE AND PREJUDICE
suggest that besides the personality model, there are at least two other approaches by which SDO may be integrated within a contextual analysis of prejudice.
The Person ⫻ Situation Model: SDO as a Moderator A second conceptualization of SDO corresponds to the classic interactionism approach (Magnusson & Endler, 1977). It suggests that personality can moderate the effect of situational factors. Renewed interests have been shown in recent years for such a model in the context of a wide variety of social–psychological phenomena (e.g., Britt, Boniecki, Vescio, Biernat, & Brown, 1996; Chen, Lee-Chai, & Bargh, 2001; Mendoza-Denton, Ayduk, Mischel, Shoda, & Testa, 2001; Mischel & Shoda, 1998). Blass (1991), for instance, asserted that research related to Milgram’s (1974) classic obedience experiments supports a Person ⫻ Situation explanation rather than a strictly situational thesis. This model has been applied, at least implicitly, in the case of SDO (see Chen et al., 2001; Danso & Esses, 2001; Pratto & Shih, 2000). For example, Chen et al. (2001) proposed a model of the effect of power as a situational variable in which an individual-difference variable, relationship orientation, which they showed to be related to SDO, acts as a moderator. For individuals having a communal relationship orientation, which, according to Chen et al.’s data, is negatively related to SDO, power would lead to socially responsible behaviors (i.e., less racism) whereas for those having an exchange relationship orientation, which is positively related to SDO, power would lead to more negative and self-interested behaviors. In this model, the effect of the situational variable (i.e., power) on racism depends on the level of a third variable, an individual-difference variable that is known to be related to SDO. Testing such a model directly with SDO, which was not the focus of Chen et al.’s studies, would suggest that the combination of a dominant social status with a high level of SDO would lead to the highest level of prejudice. The effect of social position (dominant vs. subordinate) on prejudice would depend on the level of SDO, but there is no assumption here that the situation has an impact on SDO (see Baron & Kenny, 1986). Support for this model has been found concerning other personality variables such as RWA. Both Reynolds et al. (2001) and Verkuyten and Hagendoorn (1998) have shown that RWA predicts prejudice in some situations (e.g., when personal identity is activated) but not in others (e.g., when national identity is activated). However, using a similar procedure but measuring SDO instead of RWA, Heaven and Quintin (in press) failed to find similar evidence: In support of the personality model and contrary to the interactionist model, SDO remained predictive of prejudice regardless of the situation. Still, in the explanation of prejudice, there are reasons to believe that SDO may also function as a mediator, and not simply as a moderator.
The Group Socialization Model (GSM): SDO as a Mediating Variable Research on group socialization suggests a third conceptualization of the role of SDO (Guimond, 1998, 2000; Harris, 1995; Levine, Moreland, & Ryan, 1998). This perspective emphasizes the fact that people change as a function of the group that they join. Consequently, in contrast to the personality model or to the Person ⫻ Situation model, the GSM suggests explicitly that SDO may vary according to the social context. Within this model, SDO is
assumed to function as a mediating variable. More specifically, for moderation to occur, the effect of the independent variable on another variable must depend on the level of a third variable (the moderator), whereas for mediation to occur, one requirement is that this third variable must itself be causally affected by the independent variable (see Baron & Kenny, 1986). Consistent with this view, Sidanius and Pratto (1999) have proposed that social variables may affect scores on the SDO. For example, they state that “as one possible socialization source, we expect that people from dominant groups (i.e., dominants) will adopt higher levels of SDO than people from subordinate groups (i.e., subordinates) will” (Sidanius & Pratto, 1999, p. 77). They report a number of studies showing that members of higher status groups (i.e., Whites) do have higher scores on the SDO than members of lower status groups (i.e., Blacks). More recently, Schmitt et al. (in press) have shown that manipulating ingroup status can change scores on the SDO, with those randomly allocated to a high-status position scoring significantly higher on the SDO. Sidanius and Pratto (1999) also reported a study of Levin (1996) in which the salience of the relative status of real social groups was experimentally manipulated and was shown to have the predicted impact on the SDO scores of members of high- and low-status groups. Finally, Sinclair, Sidanius, and Levin (1998) have shown that scores on the SDO can drop significantly as a function of the time spent in H-A environments. Thus, group socialization may affect scores on the SDO with people in socially dominant positions within H-E settings having higher scores than others. As noted above, research on the personality model has clearly established that higher scores on the SDO are related to higher levels of prejudice. However, surprisingly little work has been done, especially of an experimental nature, to try to integrate these two lines of evidence, that is, the evidence that status is related to SDO and that SDO is related to prejudice. The GSM suggests that it may be fruitful to do so and posits, as a central hypothesis, that SDO can mediate the effect of social position on prejudice. Given that little evidence exists to substantiate this hypothesis, one of the major goals of the present research was to address this issue. Our aim was to assess the validity of the GSM within “arbitrary-set stratification systems,”1 but also to test predictions derived from the personality model and the Person ⫻ Situation model. This was achieved by studying in a natural setting (Studies 1 and 2) and in a laboratory setting (Studies 3 and 4) the relations between social positions, SDO, and prejudice. Each model briefly described above 1 SDT distinguishes the gender stratification system, for which it proposes the invariance hypothesis, from other stratification systems such as the age system and the “arbitrary-set” system. The invariance hypothesis suggests that gender differences in SDO are not affected by situational or contextual variables (Sidanius, Levin, Liu, & Pratto, 2000; Sidanius & Pratto, 1999; Sidanius, Pratto, & Bobo, 1994). As such, this hypothesis is inconsistent with the GSM, which is assumed to apply equally to women and men. However, a detailed theoretical and empirical analysis of gender differences is beyond the scope of the present article, which focuses on arbitrary-set systems. In this series of studies, a test of the invariance hypothesis is not possible because of insufficient numbers of males and females in some studies. In this regard, we follow the practice of social dominance theorists of keeping distinct, and reporting in a distinct manner, findings concerning gender relations from those concerning other sets of intergroup relations.
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
makes some specific and testable predictions concerning the relations between these variables, as shown below.
Theoretical Predictions The personality model is a self-selection model when it comes to understanding the role of social position. It suggests that people holding top positions in the hierarchical structure may display greater prejudice but this would be a by-product of the personality predispositions which led them to select such a structural position. In fact, SDT suggests that people with high SDO tend to seek H-E social roles (see Sidanius & Pratto, 1999). Several studies have confirmed the hypothesis that SDO drives the selection of social roles or careers (see Pratto, Stallworth, Sidanius, & Siers, 1997). Thus, people who intend to do a career in one of the “power professions,” such as business administration or law, score significantly higher on the SDO than others (see Pratto et al., 1994; Sidanius, Liu, Shaw, & Pratto, 1994). As a result, one major and unique prediction of the personality model is that SDO accounts for the type of social position that people seek out. The personality model also suggests of course that SDO will predict prejudice, and because SDO leads people to select certain situations, it follows that SDO should predict prejudice regardless of the situation. However, the Person ⫻ Situation model clearly argues against the latter prediction. It suggests instead that SDO may manifest itself differently in different situations (see Pratto, 1999). In particular, predictions from this model that posit SDO as a moderator can be derived from the notion of degree of fit between the person and the situation (Van Laar & Sidanius, 2001; Van Laar, Sidanius, Rabinowitz, & Sinclair, 1999). A position of high status and power would “fit” with a high level of SDO and, therefore, should lead to the highest level of prejudice. In other words, people with high SDO may feel more at ease in expressing their prejudice toward stigmatized social groups if they are in a dominant social position, a context in which such expression may be accepted, than if they are in a lower status position. In agreement with this proposition, Pratto and Shih (2000) suggested that differences between high and low SDO on prejudice are larger rather than smaller in a more contextualized condition. This means that the manifestation of SDO in the form of prejudicial attitudes varies as a function of the social context. The “ideological asymmetry hypothesis” of SDT can be related to this model of SDO as a moderator.2 According to this hypothesis, “social attitudes and policy preferences of dominants are more strongly driven by social dominance values than is the case among subordinates” (Sidanius & Pratto, 1999, p. 45). In other words, this hypothesis predicts, in contrast to the prediction of the personality model noted above, that the correlation between SDO and prejudice can vary according to the situation: it should be stronger among dominants than among subordinates. Consistent with this view, Jost and Burgess (2000) found that the correlation between SDO and (ambivalent) attitude toward a female victim of discrimination is significantly higher among men (dominants) than among women (subordinates). The GSM goes one step further than the Person ⫻ Situation model and makes different predictions. The fundamental distinction between a moderator and a mediator is that in the former case, the intervening variable is assumed to stay the same across situations but not in the latter. In the study of Chen et al. (2001), testing the role of a moderator, “communals” were assumed to behave
differently from “exchangers” in certain situations but the “communals” were not assumed to become more like “exchangers” or vice versa. In contrast, a mediator model makes exactly such an assumption. It suggests that in a given situation, anybody can change and become more dominant, that the “personality” variable itself is transformed under the impact of a situation. In short, the moderator model assumes that personality is more or less immutable across conditions, but the mediator model does not. Consequently, unlike the two previous models, the GSM predicts an effect of social position on SDO, and also on prejudice, with the latter being mediated by the former. These are the three central predictions of the GSM. The GSM is the only model to predict that SDO will change as a function of the situation. As such, it is quite distinct from the Person ⫻ Situation model. Indeed, to test the latter model, one may select participants who are high or low on a given individual-difference variable and put them in different situations (e.g., Britt et al., 1996; Chen et al., 2001; Pratto, Tatar, & Conway-Lanz, 1999). This procedure makes it impossible to test whether scores on this individual-difference measure can change as a function of the situation, revealing the tacit assumption of immutability across situations. Similarly, consider the research referred to above as supporting a Person ⫻ Situation explanation in the case of RWA (i.e., Reynolds et al., 2001; Verkuyten & Hagendoorn, 1998). In these experiments, the salience of personal or social identity was manipulated and both RWA and prejudice were measured as dependent variables. Because the results revealed no effect whatsoever of the experimental manipulation on RWA, nor on prejudice, they do not support a mediator model, although they are consistent with a Person ⫻ Situation model. There are then clear empirical and statistical criteria by which to distinguish the Person ⫻ Situation model from the GSM. In the case of SDO, there is research, as noted above, that supports the first prediction of the GSM: members of groups holding higher status positions display higher levels of SDO. Why this might be the case has not been clearly articulated by social dominance theorists. Noting that little research has been done so far on this issue, Sidanius and Pratto (1999) simply suggested that one reason may be that a high SDO is consistent with the need for a positive self-esteem among members of dominant groups. A more obvious reason is that it is in the interest of the dominant group to argue that they should be at the top and others at the bottom. If we assume for a moment, as Duckitt (2001) has convincingly argued, that SDO is not a measure of personality but a measure of ideological beliefs, then it makes sense for dominant groups to disseminate the belief that it is only proper and natural that some people rule and dominate over others. As Bobo (1988) stated, “a dominant group seeks to articulate a set of beliefs that persuade themselves as well as others, that their privileged status is for the general good” (p. 99). Thus, the effect of group status on SDO may simply indicate that people are more likely to believe that certain groups should be at the top and others at the bottom when they are at the top (see Jost & Burgess, 2000; Jost & Thompson, 2000). The GSM also proposes, as a second prediction, that holding a dominant social position leads to higher levels of prejudice. Why? This prediction follows from the conceptualization of prejudice as 2
We express our appreciation to Michael Schmitt for suggesting this point.
SOCIAL DOMINANCE AND PREJUDICE
an H-E legitimizing myth put forward by social dominance theorists. Such a conceptualization is consistent with a wide range of theories about the social function of prejudice and discrimination (see Crocker, Major, & Steele, 1998; Glick & Fiske, 2001; Guimond, 2000; Guimond & Dambrun, 2002; Guimond, Dif, & Aupy, 2002; Jost & Banaji, 1994; Kluegel & Smith, 1986; Tajfel, 1981). It implies that people will use prejudice to justify their social dominance. Consequently, those having higher social status and power (i.e., dominants), because they are likely to experience a greater need for such a justification, should display higher levels of prejudice than others. As Crocker et al. (1998) have argued, “people of higher status may stigmatize those of lower status to justify their advantages” (p. 509). Considerable evidence supports this hypothesis that social position can affect prejudice (Guimond & Dambrun, 2002). First, minimal group experiments have showed that high-status group members discriminate more than low-status group members (see Bettencourt, Dorr, Charlton, & Hume, 2001; Mullen, Brown, & Smith, 1992). Second, Sidanius and Pratto (1999) have demonstrated that within the United States and elsewhere, the observed level of ingroup bias follows the position of the social groups in the hierarchy. Groups holding the highest position (e.g., Whites) display the highest level of ingroup bias. Finally, longitudinal studies on the process of group socialization suggest that people internalize the attitudes and values, including negative intergroup attitudes and beliefs, of the group that they join (Guimond, 2000). For example, research indicates that students and faculty differ widely in their social and political views as a function of their academic discipline (see Bereiter & Freedman, 1962; Guimond, 1998; Guimond, Palmer, & Be´ gin, 1989; Ladd & Lipset, 1975; McClintock, Spaulding, & Turner, 1964; Sidanius, Pratto, Martin, & Stallworth, 1991). Accordingly, using a sample of 5,655 students from the University of Texas, Sidanius et al. (1991) have shown that there are striking differences between career tracks on racism. Students in the “power professions,” such as commerce and law, emerged as being the most racist whereas people in arts or social work are among the least racist. Moreover, there is both cross-sectional and longitudinal evidence to suggest that these attitudinal differences result, at least in part, from a socialization process in the sense that differences are minimal at first but grow larger over time spent in an academic field of study (Feldman & Newcomb, 1969; Guimond, 1992, 1995, 1998, 1999; Guimond & Palmer, 1990, 1996a, 1996b; Newcomb, 1943; Sidanius et al., 1991; Van Laar, et al., 1999). Thus, students in business administration, a socially dominant or H-E academic major, become more negative toward immigrants from their first to their third year of university whereas students in the social sciences, a less dominant academic major, maintain a positive attitude toward immigrants during the same time period (Guimond & Palmer, 1996b). Similarly, a 4-year longitudinal study (i.e., Guimond, 2000) revealed that English–Canadian prospective military officers become more negative toward immigrants and other outgroups over the course of their training program. This suggests that, perhaps, socialization into any dominant social position or H-E role fosters the acquisition of H-E legitimizing myths whereas socialization into any H-A position would foster the acquisition of H-A legitimizing myths. To explain these relations between social position and prejudice, previous research has documented the role of self-categorization and processes of normative as well as informational social influ-
ence (Dambrun, Guimond, & Duarte, 2002; Guimond, 1998, 1999, 2000, 2001). Thus, people can change and become more prejudiced because they learn to conform to certain normative standards. However, they may also change in order to justify their new social position. Consequently, and complementing previous research, in this series of studies, we examine how important the acquisition of a dominant social position can be in the socialization process. More specifically, the fundamental question raised here is: Can these effects of social position on prejudice be mediated by SDO? Is it possible that when people find themselves in a dominant social position and when they categorize themselves as belonging with those who are “superior/dominant/more competent,” they suddenly see group-based inequality in a different light and, as a result, come to denigrate those who are “inferior/incompetent/ weak”? If those who are dominant do express higher levels of prejudice, and if prejudice serves to justify the economic and social superiority of those who are dominant, then a variable such as SDO that measures “ideological tendencies to legitimize social arrangements” (Jost & Burgess, 2000, p. 303) should mediate the effect of a dominant position on prejudice. We report the results of four studies that, collectively, provide a conclusive test of this hypothesis. More specifically, Studies 1 and 2 seek to establish, in a natural setting involving differences in status and dominance, that SDO can function as a mediator of the effect of social position on prejudice even though in such settings, SDO can also play a role in the selection of social positions, as shown in past research. Studies 3 and 4 involve an experimental manipulation of social dominance that is required to establish that the results observed in natural settings are not caused by self-selection. It is important to note that although each of the three models discussed above make distinctive predictions, this does not necessarily imply that they are mutually exclusive. The three models can also be combined in various ways such that, in principle, all three could be valid under the appropriate condition.
Study 1 In the first study, our aim was to provide preliminary evidence for the hypothesis that social dominance leads to prejudice through its effect on SDO by capitalizing on naturally occurring differences in status and dominance between upper-level students in law and in psychology. Past research has shown that most teachers, parents, and the students themselves tend to agree with the hierarchical structure of academic disciplines, with some disciplines being perceived as having much higher prestige and status than others (see Chambon, 1990; Monteil & Huguet, 1999). Law and psychology were selected as the focus of our research because within this hierarchical structure, the field of law, which is associated with more power, wealth, and status, can be considered as more H-E than the field of psychology (Pratto et al., 1994). Furthermore, Study 1 examines upper-level university students because unlike first-year students, they have been well socialized into their social position. We predict that upper-level law students, because they have been socialized into a dominant social position, will display higher levels of prejudice and higher levels of SDO than upper-level psychology students. Furthermore, on the basis of the GSM, we expect that scores on SDO will mediate the effect of social position (law vs. psychology) on prejudice. In contrast, the Person ⫻
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
Figure 1. Social dominance orientation (SDO) as a mediator of the effect of social position (law vs. psychology) on prejudice (Study 1). Path weights are standardized. The path weights in parentheses do not control for the effect of the mediator. **p ⬍ .01.
Situation model suggests that a dominant social position will be related to higher levels of prejudice especially if it is associated with a high level of SDO. This model predicts an interaction of social position and personality (SDO) on prejudice with people in law being more likely to display higher levels of prejudice than those in psychology when SDO is high rather than low. Such an interaction would be contrary to the personality model, which expects that the correlation between SDO and prejudice should be the same in law and in psychology.
Method Participants. Seventy-five university students in their third or fourth year of study in psychology (n ⫽ 45) or law (n ⫽ 30) served as participants. All were students in a university from a midsize city in France and volunteered to take part in a study on “social perception and decision making.” The average age of the total sample is 22.6 years old. Particular attention was given to the gender composition of the sample so that effects of academic major would not be confused with effects of gender. In law, we succeeded in getting an equal number of males and females, that is 15, whereas in psychology there are 11 males and 34 females. Questionnaire. Participants were tested individually in their own academic area and were requested to fill out a questionnaire containing measures of SDO and prejudice, in that order. The main dependent measure was a 15-item scale of prejudice developed on the basis of previous research (Dambrun, 2001; Dambrun & Guimond, 2001; Guimond & Dambrun, 2002). Eight items of this scale are worded in a way that agreement reflects a positive attitude toward other ethnic groups (e.g., “It is pure nonsense to blame Algerians or Moroccans for the economic problems in France”). Seven items are worded in a way that agreement reflects a negative attitude toward outgroups (e.g., “Those immigrants who do not have immigration documents should be sent back to their countries”). The content of the items is fairly similar to that of other measures of prejudice, such as Lepore and Brown’s (1997) prejudice scale, although it is adapted to the intergroup context in France where the prime targets of prejudice and discrimination are North Africans (e.g., “Arabs”; see Lambert, Moghaddam, Sorin, & Sorin, 1990; Pettigrew et al., 1998; Sabatier & Berry, 1994). Participants indicate their responses for this and subsequent measures using a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). Previous research has documented the reliability and validity of this scale by showing that it is strongly correlated with other measures of prejudice (e.g., Adorno, Frenkel-Brunswick, Levinson, & Sanford’s, 1950, Ethnocentrism Scale) and that it can predict discriminatory behaviors (see Dambrun & Guimond, 2001; Guimond & Dambrun, 2002; Michinov, Dambrun, Guimond, & Meot, 2002). In the present study, ␣ ⫽ .88, reflecting adequate reliability. Higher scores on this scale indicate greater prejudice. On the basis of the research of Pratto et al. (1994), the questionnaire also contained a 10-item version of SDO. As the study was carried out in France, the items were back-translated into French. Five items indicate a favorable orientation toward group dominance and inequality, and 5 items indicate a favorable orientation toward social equality
(reverse coded). The reliability of this scale is satisfactory (␣ ⫽ .81) and similar to that reported by Pratto et al. (1994).3
Results Given a typical gender imbalance between fields of study, with more females being in psychology and more males being in law, effects of academic major could in fact be mistaken for gender effects and vice versa (Guimond & Palmer, 1990). To prevent such problems, we tested the effects of academic major with, and without, gender of participants as a covariate, and compared them. The results were the same suggesting that effects of fields of study are not due to gender. To underline the fact that the effects reported are not a function of the gender composition of the samples, we report the results obtained using gender as a covariate. Consistent with expectations, students in law express a significantly higher level of prejudice (M ⫽ 3.07, SD ⫽ 1.10) than students in psychology (M ⫽ 2.40, SD ⫽ 0.88), F(1, 74) ⫽ 8.77, p ⬍ .004. Moreover, consistent with our socialization hypothesis and with SDT, the results show that students in law have a significantly higher score on SDO (M ⫽ 2.59, SD ⫽ 1.01) than students in psychology (M ⫽ 1.72, SD ⫽ 0.65), F(1, 74) ⫽ 15.34, p ⬍ .001. To test the mediating role of SDO in the relation between academic major and prejudice, the regression procedure advocated by Baron and Kenny (1986) was followed. Figure 1 displays the results. It first requires that the mediating variable (e.g., SDO) be related to the independent variable (e.g., academic major) and the dependent variable (e.g., prejudice). As Figure 1 shows, the social position or academic major (coded as ⫹1 for law and ⫺1 for psychology) has a significant effect on SDO (␤ ⫽ .48, p ⬍ .001). SDO, in turn, is strongly predictive of prejudice (␤ ⫽ .61, p ⬍ .001). This evidence means that SDO fulfils two initial require3 Jost and Thompson (2000) have shown that two correlated factors can be distinguished within the SDO scale. The results of several factor analyses on this and other samples (see Dambrun, 2001) reveal the same two factors on the French version of the scale, a factor dealing with group-based dominance and a factor concerning opposition to equality (OEQ). The importance of this distinction has been shown by Jost and Thompson. Consequently, in this series of research, analyses were performed on the total score of SDO and on the two factors. However, we did not find any systematic pattern suggesting that the results differ reliably when using the factor scores rather than the total score. This of course does not mean that these differences do not exist but that we were unable to observe them. One reason may be that we focus on dominants whereas this distinction may be more important when considering subordinate groups. Jost and Thompson reported that the two factors of SDO are more strongly correlated among high than low-status groups.
SOCIAL DOMINANCE AND PREJUDICE
ments of a mediating variable. The final and most basic requirement specified by Baron and Kenny is that a mediating variable should predict the dependent variable (prejudice) even when the independent variable (academic major) is statistically controlled, while the effect of the independent variable on the dependent measure should be substantially reduced when the mediating variable is statistically controlled. Figure 1 indicates that these requirements are fulfilled in the present case. The effect of academic major on prejudice (␤ ⫽ .34, p ⬍ .01) becomes nonsignificant when SDO is statistically controlled (␤ ⫽ .05), but the effect of SDO on prejudice remains significant even when academic major is statistically controlled (␤ ⫽ .59, p ⬍ .001). To test whether this pattern of results reflects a significant reduction in the variance accounted for by academic major, a z-score test was performed (see Kenny, 1998). As would be expected, this test is significant (z ⫽ 3.63, p ⬍ .01). Testing the reverse model, proposing that prejudice mediates the relation between social position and SDO, did not provide confirming evidence. When controlling for prejudice, the effect of academic major on SDO remains significant (␤ ⫽ .31, p ⬍ .001). The Person ⫻ Situation model was tested using a multiple regression procedure. As specified by Baron and Kenny (1986), and in contrast to the analyses needed to identify a mediator, the statistical criteria for moderation is a significant interaction between the independent variable and the moderator. The variables entered into the equation were academic major followed by SDO used as a continuous variable and the interaction between the latter two variables (Baron & Kenny, 1986). The continuous predictor was centered for this and subsequent regressions involving the test of interaction effects (Aiken & West, 1991). The interaction term was created by multiplying academic major by SDO. This analysis revealed a marginally significant interaction of SDO and academic major on prejudice (␤ ⫽ ⫺.17), t(71) ⫽ 1.76, p ⬍ .10, providing only weak support for the hypothesis that SDO moderates the effect of social position on prejudice. The correlations between SDO and prejudice are significant in law (r ⫽ .46, p ⬍ .01) and in psychology (r ⫽ .67, p ⬍ .001). These two correlations do not differ significantly from each other (z ⫽ 1.3, ns).
Discussion As expected, Study 1 indicates that university students differ significantly in their level of prejudice as a function of their academic area. Upper-level students in law, an H-E academic major, display more negative attitudes toward outgroups than upper-level students in psychology, a field which may be defined as H-A, consistent with the findings of Sidanius et al. (1991). Moreover, as predicted on the basis of the GSM, and replicating the results of Sidanius, Liu, Shaw, and Pratto (1994), students in law score significantly higher on SDO compared with students in psychology. More importantly, Study 1 confirms the hypothesis that SDO can account for the relation between academic major and prejudice. The data provide a clear empirical indication that SDO can mediate the relation between social position and prejudicial attitudes in the context of what social dominance theorists refer to as the “arbitrary-set stratification system.” Once SDO was covaried out, the relation between social position and prejudice disappeared, but the reverse was not the case. Upper-level students in law still have significantly higher scores on the SDO even when scores of the 15-item scale of prejudice
were statistically controlled. This evidence is important because the correlations that are obtained between SDO and prejudice, being relatively strong, could lead to the argument that SDO is itself a measure of prejudice. If this was the case, the effect of social position on SDO should also disappear when controlling for prejudice, something that is not observed. This pattern of evidence suggests that SDO has some explanatory power and does not merely represent a descriptive relabeling of the variables. This is one of the basic criticisms put forward against a personality approach—that it is merely descriptive and does not really explain anything (Jones, 2002). This criticism can be applied to previous research but not to the data presented here. We show that SDO can account for the fact that law students display more prejudice than psychology students. Because students in law have been socialized in a more dominant social position, their score on the SDO has probably increased over time whereas the score of psychology students may have decreased (see Sinclair et al., 1998). This effect of social position on SDO would be responsible for the difference in prejudice between students in psychology and those in law. However, this is only one possible interpretation. The personality model can also account for these results, unlike the Person ⫻ Situation model. In fact, overall, the results show the least support for the Person ⫻ Situation model. First, the interaction between SDO and academic major on prejudice is only marginal and second, the correlation between SDO and prejudice does not differ reliably in law compared with psychology. This is inconsistent with the ideological asymmetry hypothesis, which predicts higher correlations between SDO and prejudice among dominants (law students), whereas the trend observed is in fact going in the opposite direction (if anything, the correlation between SDO and prejudice is higher in psychology than in law). Thus, the evidence is more supportive of the personality model than the Personality ⫻ Situation model: SDO predicts prejudice in essentially the same manner, regardless of the situation. Furthermore, although the personality model does not predict an effect of social position on prejudice, it can argue that this effect is spurious and derives only from the fact that SDO shares some common variance with both social position and prejudice. In other words, because high SDO students are more likely to select the field of law and because high SDO students have greater prejudice, an association between academic major and prejudice is observed. However, according to the personality model, this association may reflect a self-selection process rather than a socialization process. This interpretation however is not consistent with previous longitudinal research carried out to distinguish between self-selection and socialization (Guimond, 1998). These studies revealed that in addition to selfselection, there are socialization effects with attitudes and beliefs changing significantly and in a different manner depending on the academic major. If the interpretation proposed on the basis of the GSM is the right one, then it means that the results of Study 1 were obtained because the participants were upper-level students who have been socialized into their social position. The GSM would not expect similar results to hold for first-year students who have not been socialized into their position, but the personality model would. Consequently, a second study with a larger sample was conducted, identical to the first one, except that in this case, first-year students in psychology and in law were tested in addition to upper-level students. This allowed us to obtain preliminary evidence concern-
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
ing possible socialization effects on the SDO. The GSM predicts an interaction of academic major and academic year such that SDO increases with academic year in law but decreases in psychology (see Guimond, Palmer, & Be´ gin, 1989). More importantly, this allowed us to test the prediction that the pattern of results observed in Study 1 would be replicated only among upper-level students, not among first-year students. However, the personality model, because it explains the results of Study 1 by a process of selfselection, predicts the same results among first-year students. Those who have a high SDO are more likely to select the field of law, and because high SDO students have more prejudice, an association between academic major and prejudice should be observed even among first-year students.
Study 2 Method Participants. A total of 1,603 students participated in Study 2, of which 125 were excluded because their academic major was missing or was not psychology or law, leaving a sample of 1,478 students. Of these, 1,124 are first-year students with 686 in psychology (61 males, 625 females) and 438 in law (141 males, 297 females). The remaining 354 are students in their third or fourth year, of which 128 are in psychology (18 males, 110 females) and 226 are in law (59 males, 167 females). The mean age of first-year and upper-year students is, respectively, 19.3 and 21.5 years old. In contrast to Study 1, participants were asked to answer a questionnaire in groups, during one of their classes being delivered in a large lecture hall. Exclusion of missing data accounts for the slight variation in number of participants from one analysis to the other. Questionnaire. SDO was measured with the full 16-item scale developed by Pratto et al. (1994), ␣ ⫽ .86. Because of time constraints, the complete 15-item scale of prejudice could not be used in Study 2. Rather, among first-year students, four items from this scale were used (two positive and two negative items). Among upper-level students, the questionnaire included these same four items plus an additional four other items, giving an eight-item scale of prejudice that can be used in analyses restricted to upper-level students. The alpha of the four-item scale of prejudice is .71; the alpha of the eight-item scale is .84. Among an independent sample of 38 students, the correlations of these two scales with the 15-item measure of prejudice used in Study 1 are, respectively, .85
( p ⬍ .001) and .93 ( p ⬍ .001). SDO was always measured before prejudice, as it is expected to mediate the effect on prejudice.
Results As in Study 1, analyses conducted with, and without, gender of participants as a covariate revealed the same results. A 2 ⫻ 2 analysis of covariance (ANCOVA), with fields of study as a first factor (law vs. psychology), academic year as a second factor (first year vs. third and fourth year), and gender as a covariate was performed using scores on the SDO as a dependent variable. This analysis revealed a main effect of academic major, F(1, 1451) ⫽ 39.85, p ⬍ .001, with students in law displaying higher levels of SDO (M ⫽ 2.61, SD ⫽ 1.05) than those in psychology (M ⫽ 2.20, SD ⫽ 0.82). The main effect of academic year was not reliable (F ⬍ 1), but the predicted interaction between academic major and academic year was significant, F(1, 1451) ⫽ 21.41, p ⬍ .001. As Figure 2 indicates, upper-level students in law (M ⫽ 2.77, SD ⫽ 1.04) have significantly higher scores on the SDO compared with their first-year counterparts (M ⫽ 2.54, SD ⫽ 1.04), F(1, 661) ⫽ 9.87, p ⬍ .005. The reverse is observed in psychology, with upper-level students (M ⫽ 1.97, SD ⫽ 0.77) displaying lower scores on the SDO than first-year students in psychology (M ⫽ 2.23, SD ⫽ 0.82), F(1, 811) ⫽ 13.19, p ⬍ .001. It should be noted that even in first year, the difference between psychology and law is significant, F(1, 1121) ⫽ 11.92, p ⬍ .001. Using the four-item scale of prejudice common to all participants, the 2 ⫻ 2 ANCOVA indicates a significant main effect of academic major, F(1, 1451) ⫽ 3.75, p ⫽ .053, and a marginal effect of academic year, F(1, 1451) ⫽ 2.75, p ⬍ .10. These main effects are qualified by a significant interaction of academic major and academic year, F(1, 1451) ⫽ 3.99, p ⬍ .05. This interaction reveals that in psychology, upper-level students display a significantly lower level of prejudice (M ⫽ 2.24, SD ⫽ 1.22) than their first-year counterparts (M ⫽ 2.55, SD ⫽ 1.19), F(1, 798) ⫽ 6.13, p ⬍ .01. However, in law, first-year students (M ⫽ 2.60, SD ⫽ 1.29) do not differ from upper-level students (M ⫽ 2.61,
Figure 2. Significant interaction effect of academic major and academic year on social dominance orientation (Study 2).
SOCIAL DOMINANCE AND PREJUDICE
SD ⫽ 1.11, F ⬍ 1).4 Finally, one should also note that in their first-year, students in psychology and law do not differ on prejudice (F ⬍ 1), whereas among upper-level students, with a smaller sample size, the difference is highly reliable, F(1, 347) ⫽ 7.10, p ⬍ .005. These results are consistent with the socialization model but not with the personality model. Because first-year students do not differ on prejudice, SDO cannot possibly be a mediator of prejudice. Accordingly, the significant difference on the SDO among first-year students may be taken as a self-selection effect, consistent with the personality model. However, the fact that upper-level students differ on the SDO as a function of their major may be partly due to self-selection but also partly due to socialization. Consequently, the GSM predicts that, among upper-level students, SDO should mediate the effects of academic major on prejudice, as in Study 1. Mediation analysis. Consistent with mediation, the above noted difference on the four-item measure of prejudice between upper-level students in law and psychology, F(1, 347) ⫽ 7.10, p ⬍ .005, becomes nonsignificant (F ⬍ 1) when SDO is used as a covariate, F(1, 346) ⫽ 97.97, p ⬍ .001. However, the reverse is not the case. Controlling for prejudice, F(1, 346) ⫽ 109.20, p ⬍ .001, does not change the fact that law students have a stronger score on SDO than psychology students, F(1, 346) ⫽ 39.91, p ⬍ .001. These results support the hypothesis that the effect of academic major on prejudice is mediated by SDO, but they do not support the hypothesis that the effect of academic major on SDO is mediated by prejudice. Identical results are observed when using the more reliable eight-item measure of prejudice available for upper-level students. Regression analyses using this eight-item measure indicate that the significant effect of academic major on prejudice (␤ ⫽ .24, p ⬍ .001) becomes unreliable when SDO is statistically controlled (␤ ⫽ .04, ns) and this reduction in the variance accounted for by academic major is significant (z ⫽ 6.42, p ⬍ .001). At the same time, and as required to argue that SDO is a mediating variable, the effect of SDO on prejudice remains reliable even when the effect of academic major is controlled (␤ ⫽ .53, p ⬍ .001). Testing the reverse causal model suggesting that prejudice mediates the effect of academic major on SDO did not provide confirming evidence. When controlling for prejudice, the effect of academic major on SDO remains significant (␤ ⫽ .25, p ⬍ .001, when using the eight-item scale; ␤ ⫽ .30, p ⬍ .001, when using the four-item scale). The above analysis replicates the results of Study 1. Because in Study 2, first-year students are also involved, further analyses were conducted including academic year as a factor. We find that the significant Academic Year ⫻ Academic Major interaction on prejudice (␤ ⫽ .17, p ⬍ .05) becomes nonsignificant when SDO is statistically controlled (␤ ⫽ .006, ns). Because the effect of academic year on prejudice is significant among psychology students (␤ ⫽ .09, p ⬍ .01), we tested more directly whether this effect can be accounted for by SDO. This is indeed the case. When SDO is covaried out, the effect of academic year on prejudice among psychology students is no longer reliable (␤ ⫽ .04, ns). Moderation analysis. As in Study 1, further analyses were performed to test the Person ⫻ Situation model. With the total sample, the variables entered into the regression equation were academic major (law coded as ⫹1 and psychology coded as ⫺1) followed by academic year, SDO used as a continuous variable and centered, the two-way interactions between the previous variables, and the three-way interaction between academic major, academic
year, and SDO. This analysis revealed, using the four-item scale of prejudice as the dependent variable, that the predicted interaction between SDO and academic major was significant, ␤ ⫽ ⫺.25, t(1462) ⫽ ⫺3.36, p ⬍ .001. This Person ⫻ Situation interaction was however qualified by a significant interaction between SDO, academic major, and academic year, ␤ ⫽ .23, t(1462) ⫽ 3.04, p ⬍ .001. This means that the Person ⫻ Situation model does not apply in the same manner across academic years. Separate regression analyses performed within each levels of academic year revealed, among first-year students, a significant Person ⫻ Situation interaction, ␤ ⫽ .07, t(1108) ⫽ ⫺2.71, p ⬍ .005. This indicates that considering the Person ⫻ Situation interaction adds significantly to the prediction of prejudice over and above SDO, although the amount of variance explained by the interaction term is relatively small (R2 change ⫽ .005). As the ideological asymmetry hypothesis would predict, the correlation between SDO and prejudice among first-year students is higher in law (r ⫽ .51, p ⬍ .001) than in psychology (r ⫽ .30, p ⬍ .001, z ⫽ 4.11, p ⬍ .01). Thus, consistent with the Person ⫻ Situation model, high SDO is associated with more prejudice in law (dominant) compared with psychology (subordinate). Among upperlevel students, the regression analysis performed using the fouritem scale of prejudice also reveals a significant Person ⫻ Situation interaction, ␤ ⫽ .13, t(353) ⫽ 2.46, p ⬍ .01. However, the shape of this interaction is opposite to the prediction of the ideological asymmetry hypothesis (and to what is observed among first-year students). The correlation between SDO and prejudice, as measured by the four-item scale, tends to be higher among upper-level students in psychology (r ⫽ .49, p ⬍ .001) rather than law (r ⫽ .44, p ⬍ .001, z ⫽ 0.69, ns). In other words, this interaction means that as the level of SDO increases, it is associated with relatively more prejudice in psychology rather than in law, the reverse of the results among first-year students. Furthermore, using the more reliable eight-item measure of prejudice available among upper-level students does not improve things for the Person ⫻ Situation model. In this case, the Academic Major ⫻ SDO interaction effect is not significant, ␤ ⫽ .08, t(353)⫽ 1.58, p ⫽ .11, as in Study 1, where only a marginal effect was observed. Contrary to the ideological asymmetry hypothesis, the correlation between SDO and the eight-item measure of prejudice is not significantly higher in law (r ⫽ .52, p ⬍ .001) than in psychology (r ⫽ .48, p ⬍ .001, z ⫽ 0.48, ns). Overall then, little consistent support is found for the role of SDO as a moderator among upper-level students.
Discussion The explanation of the results of Study 1 that was proposed on the basis of the personality model implied that similar results should be observed among first-year students and among upperlevel students. In contrast, the GSM was predicting different results as a function of academic year such that only the data from upper-years students would replicate the results of 4
As Feldman and Newcomb (1969) have observed, finding no difference between first-year and upper-years students, as in the case of law students here, “does not necessarily mean that nothing has happened” (p. 55). Given that students typically become more tolerant with education, the fact this is not the case among law students may indicate that a significant countervailing influence is at work (see Guimond, 1999).
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
Study 1. This is exactly what Study 2 reveals: SDO mediates the relation between academic major and prejudice only among upper-level students, not among first-year students. Why? This is difficult to explain on the basis of the personality model. If SDO drives the selection of a social position, and if SDO predicts prejudice, then why do we find a relation between social position and prejudice only among upper-level students and not among first-year students? To put it bluntly, these results contradict the self-selection thesis as an explanation of the relation between social position and prejudice: Selfselection cannot, at one and the same time, account for the fact that there are no differences in prejudice between first-year law and psychology students, despite a sample of over 1,000 participants, although there is a significant difference in third-year students, with a much smaller sample size. Neither can it be suggested that self-selection is “really” operative only among upper-level students—this is contradicted by the highly reliable difference in SDO among first-year students, a finding that confirms the self-selection prediction of the personality model. In contrast, the GSM can easily account for these results as they conform to predictions. First-year students have been tested at the beginning of their first year. Consequently, these students, contrary to upper-level students, have not been socialized into their position. Because in first year nobody is in a dominant social position, there is no effect of position on prejudice to be expected according to the GSM. In contrast, after 3 or 4 years of study in law or in psychology, there are reasons to believe that these two groups of students occupy different positions in the social structure. Given that the field of law is socially dominant (H-E) compared with psychology (H-A), upper-level law students were expected to express higher levels of prejudice than psychology students. The results supported this prediction, replicating the findings of Study 1 obtained with a slightly different methodology and confirming similar results obtained in longitudinal and cross-sectional studies among university students from various other academic institutions (see Guimond, 1999, 2000; Guimond & Palmer, 1990, 1996b; Sidanius et al., 1991). Of course, the most important question is why upper-level students in law display outgroup prejudice to a greater extent than those in psychology, or why in previous longitudinal studies, commerce students did not differ from social science students at the beginning of their training (contrary to a selfselection explanation) but became more prejudiced against immigrants after 3 years in commerce? Previous research has identified a number of relevant factors (e.g., normative influence), but we have argued here that the acquisition of a dominant position in the social structure may also be important in the socialization process. It has been known for some time that promoting an individual to a higher position in the social hierarchy represents a powerful means to socialize that individual to a particular culture or ideology (Guimond, 1995; Schein, 1984). Indeed, individuals can undergo a prolonged period of intense socialization without changing their basic values until they reach a point where they are promoted to a particular position (see Guimond, 1995). At this point, change can occur suddenly and be quite dramatic (Schein, 1984). Consistent with this view, Levine et al. (1998) have argued that “role transition” is one of the fundamental processes in group socialization. Similarly, the first postulate of the theory of organizational socialization of Van Maanen and Schein (1979) states that “so-
cialization, although continuous . . . is no doubt more intense . . . for a member (and others) just before and just after a particular boundary passage” (p. 224). Furthermore, the reason why change at this point can be sudden is familiar to cognitive dissonance theorists, especially to the radical view of dissonance theory (Joule & Beauvois, 1998). When people reach a particular boundary passage, they need to rationalize or justify their position in the social system. In this process, SDO as a system of beliefs that justify social dominance may be a key variable. We have argued that it may represent the mediating process linking social position to prejudice. The results of Study 2 clearly support this view, and replicate on a much larger sample, the results of Study 1: SDO can explain why there is a highly reliable difference in prejudice between upper-level law students and psychology students. The reverse causal model, that prejudice explains why there is a difference between fields of study on SDO, receives no support. It may be pointed out that these results are correlational and that “obviously,” these students selected their major because of preexisting preferences for social dominance between groups. This however is totally missing the point of this study. The significance of the present findings derives precisely from the fact that they are obtained in the real world among real students who will be real actors in society. In such a context, where SDO does indeed drive the selection of academic major, as the results among first-year students indicate, we precisely show that SDO is not a mediating variable accounting for greater prejudice (in first year). Thus, the theoretical importance of Study 2 is that it was carried out in a setting where SDO can and does play a role in self-selection, as the personality model predicts, so that it can be demonstrated that even then, such a model is insufficient and does not provide a full account of the results. The fact that SDO is mediating the effect of academic major on prejudice only among upper-year students is not consistent with the personality model but with the GSM. To establish the cause of an event, it is necessary to show that when the cause is present, the event is present. But equally important, if not more, is to show that when the cause is absent, the event is also absent (Kenny, 1979). This is what Study 2 demonstrates: When there is socialization (upper-years), there is mediation, but when there is no socialization (first year), there is no mediation. The Academic Major ⫻ Academic Year interaction observed on SDO further supports the hypothesis that social position has an impact on SDO. Among psychology students, the fact that academic year is negatively related to SDO is consistent with the longitudinal results of Sinclair et al. (1998), showing that scores on the SDO decrease significantly over time in an H-A environment. To this, however, our results add the fact that it may be quite different in an H-E environment. For law students, we find exactly the reverse: academic year is related positively to SDO. Finally, some support was also found for the Person ⫻ Situation model. SDO moderates the effect of social position on prejudice such that a combination of being in law and having a high score on the SDO gives the highest level of prejudice. This finding is consistent with previous research (see Chen et al., 2001; Pratto & Shih, 2000). As Pratto (1999) has argued, “the way SDO is expressed depends on social context” (p. 247). Thus, students high in SDO are more likely to express prejudicial attitudes when they are in a relatively dominant academic major such as law. However,
SOCIAL DOMINANCE AND PREJUDICE
support is found for this model and for the ideological asymmetry hypothesis only among first-year students. Among upper-level students, the moderating role of SDO was not consistent with the ideological asymmetry hypothesis, nor very reliable, being nonsignificant with the eight-item scale, as in Study 1. In contrast, the results of the mediation analysis are clearly supportive regardless of the measure of prejudice used. In sum, Study 2 replicates and extends the results of Study 1. The evidence suggests that for newcomers (first-year students), SDO is involved in the selection of a social position and moderates the effect of social context. But for more established group members (upper-level students), SDO also mediates the effect of social context and becomes a generating mechanism for prejudice. Nevertheless, to establish the existence of a mediator, it is necessary to demonstrate a causal impact of the independent variable on the mediator and on prejudice, something that is still lacking. Studies 3 and 4 were designed to address this limitation.
Study 3 One of the key hypotheses tested in Study 3 is that the mere fact of placing an individual in a dominant social position can generate prejudice against stigmatized social groups. Whereas in Studies 1 and 2, we show that students who occupy a more dominant position also express more prejudice, in Study 3, we attempt to replicate this finding by experimentally inducing social dominance. Because previous studies were carried out in the field, many different factors were necessarily operating at the same time. In Study 3, we experimentally manipulate one of the components of group socialization assumed to be responsible for the results of Studies 1 and 2, namely the promotion of individuals to a dominant social position. The participants are asked to take part in a study of leadership. In the social dominance condition, they are given (false) feedback suggesting that they have the ability to lead and to hold positions of high responsibility. In the control condition, they are not provided with such feedback. Subsequently, they are asked to fill out a questionnaire (measuring prejudice) ostensibly for an unrelated study. We predict that participants in the social dominance condition will show evidence of greater prejudice than those in the control condition. Furthermore, SDO is assessed among these participants, after they received the (false) feedback. If SDO is a basic personality factor, it should be unaffected by the experimental manipulation but may account for an additional source of variance in prejudice. If SDO is a mediating variable, a generating mechanism for prejudice, then it may be expected to be responsive to the situational context. Accordingly, this leads to the prediction that participants will score higher on the SDO in the dominance condition than in the control condition and that scores on the SDO will mediate the effect of experimental conditions on prejudice, consistent with Studies 1 and 2.
Method Participants. Seventy-four first-year psychology students at the Universite´ Blaise Pascal, Clermont-Ferrand, France, served as participants. The average age of the sample is 19.6 years old and there are 68 women and 6 men. An additional 8 participants of foreign origin (e.g., NorthAfricans, Portuguese) were allowed to take part in the study but their data were not considered.
Procedure. First-year students who signed up for an experiment on “leadership” were greeted by a female experimenter upon their arrival at the laboratory. While seated in an experimental room in front of a computer terminal, the students were told that they would participate in two short experiments. In reality, the “first experiment” used an elaborate procedure to induce social dominance under the cover of a study of leadership. What was for the participants a second unrelated experiment involved answering a questionnaire on “social perception.” The first experiment was divided into several tasks that all dealt with leadership ability. The “second experiment” involved a second female experimenter who was introduced at the end of the first experiment and asked participants to answer a questionnaire for an entirely different study. Thus, to make sure that participants would see no connection between the two experiments, two different experimenters were used, the two experiments were run in different locations, and whereas Experiment 1 was entirely computer-administered, Experiment 2 was a questionnaire-based study. The second experimenter was blind as to the hypothesis of the study. The computer task used in the first segment of the study was developed by Michinov et al. (2002). It consists of two basic parts, which make clear to the participants that the study concerns behaviors within a business-like organization. In the first part, participants are invited to answer a series of 60 items presented on the computer screen one after the other (e.g., “I am not anxious when I speak with others”). These items are ostensibly part of Gordon’s Personality Inventory, a scientific test designed to assess leadership skills. After each item, a 5-point rating scale is presented and participants must use the mouse to indicate their response. At the end of this part, participants were requested to pause while the computer calculated their score on the test. Those randomly assigned to the social dominance condition learned that they had a very high leadership score (31 on a maximum score of 40) . In presenting this score, the computer screen displayed a rating scale from 0 (low leadership) to 40 (high leadership). The participants’ score (31 out of 40) was indicated on this scale with a written message stating that “you clearly have the profile of a person who is able to lead and to hold position of high responsibility.” This is the social dominance condition. Participants randomly assigned to the control condition learned, in contrast, that they had an average score (20 out of 40). In presenting this score, the computer screen displayed the message that “you have the profile of a person that has an average ability to lead and to hold position of average responsibility.” This is the only systematic difference between the two conditions. Participants were then invited to go on to the second part of the computer task, involving a number of filler tasks, one of which was used to assess the effectiveness of the manipulation. With the instructions displayed on the computer screen, participants were asked to assume that they work in an office composed of six people and themselves. The task of the participants is to make decisions about how to best organize this office. They are then presented with a hierarchical structure composed of four levels. There are eight places in this structure and the task of the participants is to decide where to put each of the six employees and themselves in this structure. The label “group leader” at the top of the hierarchical structure was intended to make sure that the positions in this structure would be perceived as involving different levels of responsibility. This task, which reinforces the fact that this is an experiment on “leadership,” is used as a measure of the effectiveness of the manipulation, as participants in the social dominance condition are expected to place themselves higher in the hierarchy, if the manipulation was successful. When the computer task was finished, Experiment 1 was ostensibly terminated and participants were requested by the second experimenter to complete a questionnaire on social perception that contained the dependent measures. Dependent measures. The questionnaire included the 10-item version of the SDO used in Study 1 (␣ ⫽ .76) and the 15-item scale of prejudice also used in Study 1 (␣ ⫽ .77). Moreover, additional measures of prejudice were included in Study 3, namely a six-item scale of neosexism developed by Tougas, Brown, Beaton, and Joly (1995), and
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
Table 1 Correlations Among Dependent Measures (Study 3) Measure 1. 2. 3. 4. 5. 6. 7.
SDO Prejudice scale Bias against Arabs Bias against Blacks Neosexism Scale Bias against Americans Bias against Asians
1 — .37** .30** .23* .31** ⫺.16 .21†
— .62** .36** .11 .0 .25*
— .64** .14 .21† .48**
— .10 .37** .55**
— .05 .05
Note. N ⫽ 74. SDO ⫽ Social dominance orientation. † p ⬍ .10. * p ⬍ .05. ** p ⬍ .01.
a series of items designed to assess ingroup bias. To measure ingroup bias, participants were asked to indicate their attitudes toward different social groups—French (ingroup), Arabs, Black Africans, Americans, and Asians— using a scale that ranged from 1 (very unfavorable) to 7 (very favorable). Difference scores were computed to indicate the degree of bias in favor of the ingroup following the method used in past research testing SDT and other theories (Castano, Yzerbyt, Paladino, & Sacchi, 2002; Guimond & Dambrun, 2002; Levin, Federico, Sidanius, & Rabinowitz, 2002; Sidanius & Pratto, 1999). Previous research has provided evidence for the validity of such measures of ingroup bias (see Dambrun & Guimond, 2001; Guimond & Dambrun, 2002; Guimond & Palmer, 1993). Typically, the difference between the rating of the ingroup and the rating of the outgroup is a more sensitive measure than the rating of the outgroup alone.5 Thus, the dependent measures in Study 3 include a much broader range of measures of prejudice. This is important because one of the key features of SDT is that by conceptualizing prejudice as H-E legitimizing myths, it provides a common way to explain many different types of prejudice including racism and sexism. Thus, we expect that social dominance will increase prejudice toward many different stigmatized social groups such as Arabs or Black Africans, as measured by the ingroup bias measures, and women or immigrants in general, as measured by the Neosexism Scale and the prejudice scale. At the same time however, attitudes toward groups that are perceived as holding relatively higher social status, such as Americans or Asians, should be affected differently. In fact, SDT expects that dominants will be more positive toward high-status groups than to others (see Sidanius & Pratto, 1999, p. 81). Correlations among these various measures of prejudice (and SDO) are presented in Table 1. Consistent with past research (e.g., Dambrun & Guimond, 2001), the measures of bias against Arabs and Blacks are positively correlated with the scale of generalized prejudice. Table 1 also shows that SDO is positively and significantly correlated with the generalized prejudice scale, the Neosexism Scale and the measures of bias against Arabs and Blacks, but not with the measures of bias against Americans and Asians. These correlations between SDO and measures of ingroup bias are similar to those obtained when using the ratings of the outgroup alone. In this case, SDO is significantly and negatively correlated with the evaluation of Arabs (r ⫽ ⫺.29, p ⬍ .01) and Blacks (r ⫽ ⫺.23, p ⬍ .05), but the correlations with Americans (r ⫽ .10), Asians (r ⫽ ⫺.17), or the ingroup, the French (r ⫽ ⫺.02), are not statistically reliable. Finally, to rule out an alternative interpretation of the results in terms of the effect of mood (e.g., Bodenhausen, 1993), we included in Study 3 three items measuring mood states. Participants were requested to indicate, using a 7-point scale, if they felt happy or sad, and if they experienced mainly positive or negative feelings. These were the first items in the questionnaire followed by the SDO scale and the measures of prejudice. At the end of the experiment, participants were debriefed and requested not to discuss the experiment with other students before its completion.
Results Manipulation check. To assess the effectiveness of the experimental manipulation, we examined the position that participants selected for themselves in the hierarchical structure. Because in the social dominance condition, participants were told that they have higher leadership ability than in the control condition, they would be expected to select a higher position in the hierarchy. The structure had four levels that were coded from 1 (top level) to 4 (bottom level). A comparison of the two conditions on this measure indicates that participants in the social dominance condition (M ⫽ 1.95, SD ⫽ 0.88) selected for themselves a higher position in the hierarchy than those in the control condition (M ⫽ 2.53, SD ⫽ 0.95), F(1, 71) ⫽ 7.38, p ⬍ .008. This evidence suggests that we succeeded in creating two conditions, only one of which led participants to see themselves as occupying a more dominant social position. Prejudice. Table 2 displays the mean scores on the various measures of prejudice and ingroup bias as a function of experimental conditions, significance tests, and percent of variance accounted for. As predicted, and regardless of how prejudice is measured, the mean scores in the social dominance condition are higher than those in the control condition (or average dominance condition). These differences are statistically significant on all instances where attitudes toward low-status groups are involved (e.g., Arabs, women, and Blacks). When attitudes toward more dominant or high-status groups are involved (e.g., Americans and Asians), the effects of experimental conditions are not significant or marginally so. Contrary to SDT, there is no evidence that participants in the social dominance condition rate high-status groups more favorably. SDO. Analysis of scores on the SDO reveals a significant difference, F(1, 73) ⫽ 6.25, p ⬍ .015. Participants in the social dominance condition show a higher score on the SDO (M ⫽ 2.46, 5
As in previous research, the data from Study 3 indicate that the 15-item scale of prejudice correlates less strongly with the single-item measure of the evaluation of Blacks (r ⫽ .15, ns) or Arabs (r ⫽ .43, p ⬍ .001) than with ingroup bias measures involving the comparative rating of the ingroup (French) and Blacks (r ⫽ .36, p ⬍ .001), or the ingroup and Arabs (r ⫽ .62, p ⬍ .001). Of course, a repeated-measures ANOVA using target group as a factor (ingroup vs. outgroup) produces the same results as when a difference score (ingroup minus outgroup) is directly used.
SOCIAL DOMINANCE AND PREJUDICE
Table 2 The Effect of Experimentally Induced Social Dominance on Measures of Prejudice and Ingroup Bias (Study 3) Measure
Social dominance condition
3.0† 1.6 (ns)
Prejudice scales Prejudice scale Neosexism
3.68 (0.85) 2.1 (0.70)
3.25 (0.81) 1.75 (0.66)
Ingroup bias against low-status groups Bias against Arabs Bias against Blacks
1.15 (1.38) 0.37 (1.10)
0.47 (1.26) ⫺0.21 (1.07)
Ingroup bias against high-status groups Bias against Americans Bias against Asians
0.87 (1.28) 0.65 (1.18)
0.42 (0.89) 0.33 (1.03)
Note. Standard deviations are reported in parentheses. † p ⬍ .10. * p ⬍ .05.
SD ⫽ 0.92) than those in the control condition (M ⫽ 1.95, SD ⫽ 0.79).6 Mood state. Comparison of the two experimental conditions reveals no differences in terms of mood, F(1, 73) ⫽ 0.42, ns. The average score is 5.3 (SD ⫽ 1.33) on a 7-point scale, indicating mildly positive feelings or mood. Mediation analysis. A series of regression analyses were performed testing the extent to which the effects of experimental conditions on prejudice were reduced when SDO is part of the regression equation compared with when it is not. As Figure 3A shows, when the dependent variable is the 15-item scale of prejudice used in previous studies, there is evidence that SDO is a mediating variable. The effect of experimental condition on prejudice (␤ ⫽ .25, p ⬍ .01) becomes unreliable (i.e., ␤ ⫽ .15, ns) when SDO is statistically controlled, whereas the effect of SDO on prejudice remains significant even when experimental condition is taken into account (␤ ⫽ .33, p ⬍ .01). The z test confirmed that this pattern of results reflects a significant reduction in the amount of variance accounted for by experimental condition on the level of generalized prejudice (z ⫽ 2.12, p ⬍ .05). These results replicate the findings of Study 1 and Study 2. They confirm the hypothesis that SDO is a mediator, a generating mechanism, for the effect of the experimental induction of social dominance on prejudice. Figure 3B and 3C show similar results when using neosexism and bias against Arabs, respectively, as dependent variables. In both cases, the effects of conditions on these measures become nonsignificant when SDO is controlled. The only exception is for the measure of bias against Black Africans (Figure 3D). In this case, the effect of condition remains marginally significant when controlling SDO. One reason for this may be that, in this case only, participants in the control condition display slightly more favorable attitudes toward Blacks (M ⫽ 5.82, SD ⫽ 1.24) than toward their own group (M ⫽ 5.60, SD ⫽ 1.47), t(33) ⫽ ⫺1.12, ns, perhaps because of social desirability. Participants in the social dominance condition display more favorable attitudes toward their own group (M ⫽ 6.00, SD ⫽ 0.97) than toward Blacks (M ⫽ 5.61, SD ⫽ 1.23), t(39) ⫽ 2.15, p ⬍ .05. The reverse causal model in which prejudice mediates the effect of the manipulation on SDO is
generally not supported. As Figure 4 indicates, the effects of conditions on SDO tend to remain significant even when prejudice is covaried out. This is clearly the case for three out of four measures of prejudice. For this model, none of the z tests designed to appreciate the significance of mediation are statistically significant at the .05 level. Similarly, controlling for the effect of positive mood does not account for the experimental effects.
Discussion This study indicates that participants led to believe that they have the profile of a person able to occupy a dominant social position express more prejudice on a variety of measures than those led to believe otherwise. They score higher on our scale of generalized prejudice but they are also more sexist, and more biased against Arabs and Black Africans. These data provide strong support for our hypothesis. The mere fact of being in a dominant social position is sufficient to generate prejudice and ingroup bias. Four features of these data are particularly significant. First, if prejudice is the result of deep-seated personality predispositions, then people should display essentially the same level of prejudice regardless of whether they are told that they have 6 It is important to consider that the validity of this experimental procedure as a means to change SDO is supported by the results of a further study conducted by Dambrun (2001). That study among a sample of 89 first-year students involved two important modifications compared with Study 3. First, scores on the SDO were assessed 3 months prior to the experiment itself to establish that scores on the SDO change significantly over time as a result of this manipulation. This was the case, as a significant Time ⫻ Experimental Condition interaction was observed on SDO. Second, in contrast to Study 3, participants assigned to the control condition in this further study were not given any feedback whatsoever about their performance on the test of their leadership potential. The results of that study also showed significantly higher scores on the SDO among participants in the social dominance condition compared with those in the control condition. Details of that study can be obtained from the authors upon request.
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
Figure 3. Tests of social dominance orientation (SDO) as a mediator of the effect of experimental conditions on prejudice (Study 3). Path weights are standardized. The path weights in parentheses do not control for the effect of the mediator. ⫹p ⬍ .10. *p ⬍ .05. **p ⬍ .01.
strong or average leadership ability. Thus, the mere fact that we observe reliable change in the level of prejudice of our participants is noteworthy as it challenges the personality model. Second, because our participants are mainly female students, the fact that an increase in sexism is observed in the dominance condition suggests that women who succeed in reaching a powerful position may turn their back on their gender and become more sexist. This fits with a number of important theoretical perspectives in social psychology (see Dambrun, 2001; Ellemers, 2001). Third, showing that scores on the SDO were significantly affected by our experimental manipulation raises questions about any assumption that SDO is a measure of a basic and stable personality trait. Fourth, given the nature of our experimental manipulation, it might be argued that the results reflect the role of positive mood. Participants in the dominance condition who learn that they did well on the leadership test may be said to be in a more positive mood than those in the average dominance condition. Because research has shown that positive mood is associated with greater stereotyping and discrimination (Bodenhausen, 1993; Forgas & Fiedler, 1996),
it may be claimed that the higher level of prejudice observed in the high dominance condition was caused by mood. However, because we found no difference between experimental conditions on reported mood, this explanation can be ruled out. But the role of SDO cannot. Indeed, the theoretical significance of our results rests in the fact that not only do we observe greater prejudice in the high dominance condition, but we also find that scores on the SDO are significantly higher. The results from our mediation analyses are clear: For three measures of prejudice on which a significant effect of condition is observed out of four, these effects become nonsignificant when SDO is controlled. In contrast, the effects of condition on SDO remain significant or marginally so when prejudice is controlled. This provides evidence that extends existing research in a significant manner. The bulk of existing research suggests that SDO is correlated with prejudice. Although this is often interpreted to suggest that SDO is a cause of prejudice (e.g., Whitley, 1999), this is obviously going well beyond what the data suggest. In fact, the results of Study 3 provide what is probably the strongest
SOCIAL DOMINANCE AND PREJUDICE
Figure 4. Tests of the reverse model positing prejudice as a mediator of the effect of experimental conditions on social dominance orientation (SDO; Study 3). Path weights are standardized. The path weights in parentheses do not control for the effect of the mediator. ⫹p ⬍ .10. *p ⬍ .05. **p ⬍ .01.
evidence to date in favor of the idea of a causal relation between social dominance and prejudice. More specifically, the SDO scale may successfully capture a set of beliefs that are strongly related to prejudice. However, this set of beliefs appears to derive from the individual’s location in the social system. Given a dominant position, people are more likely to subscribe to the idea that inequality is not that bad and, the more they subscribe to this idea, the more prejudiced they become. If, as argued above, the data challenge the claim that SDO is a personality trait in the traditional sense, the question arises as to how these results relate to a socialization explanation of prejudice. As noted in the Discussion section of Study 2, one may relate the findings of Study 3 to some well-known principles of socialization. According to Schein (1984), one of the main ways in which organizations socialize their members is by promoting them to a position of responsibility. As Schein (1984) explains, “the same values which the new member may have criticized or jeered at
from his [sic] position at the bottom of the hierarchy suddenly look different when he [sic] has subordinates of his own . . . Many of my panel members . . . reported with considerable shock that some of the practices that they had condemned in their bosses were quickly adopted [italics added] by them once they had themselves been promoted” (p. 14). This principle received strong confirmation in a series of longitudinal studies investigating the effects of military socialization (Guimond, 1995). It was demonstrated that the main factor that accounts for change in values during the socialization process is simply “being assigned to a position of leadership” (Guimond, 1995, p. 252). Specifically, to practice their leadership skills, some military recruits are offered the opportunity to be a leader during the last year of their military program. It was found that those who are promoted to such leadership positions within the military are more likely than others to internalize military values. Their values did not differ from others before being promoted, but they differed reliably after. Schwarzwald,
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
Koslowsky, and Shalit (1992) have reported similar evidence. They found that members of a large corporation who were promoted to the position of Division Head displayed significantly higher levels of organizational commitment afterward, compared with their nonpromoted peers. Thus, it can be argued that Study 3 provides an experimental demonstration of such principles of organizational socialization in showing that promoting students to a position of responsibility is sufficient to change some of their fundamental social attitudes. Such a demonstration however is more telling than any previous field research. In field research, the socialization agents controlling the situation are military officers or top managers. But in our study, the experimenter assumed this role and gave the promotion randomly. The only systematic difference that existed between our two groups of participants is their fictitious scores on the test of their leadership ability. We know that half of them, at random, found out that they have very high leadership ability. Consequently, if their scores on SDO are significantly higher, then we know that this is not a reflection of their basic personality predispositions—it is the result of learning that they have what it takes to occupy a dominant social position. Seeing themselves at the top leads them to perceive inequality in a different manner. They are now ready to accept that certain groups should be at the top and others at the bottom. Finally, the fact that they express more prejudice is shown to be mediated by this set of beliefs about group dominance, confirming the hypothesis that group dominance is intimately related to the very nature of prejudice.
Study 4 A final study was conducted to provide further experimental evidence for the effects of social position on prejudice and to rule out a number of alternative interpretations. First, in Study 3, the sample is largely composed of female students, which raises the question as to whether these effects can be obtained among males also. Thus, Study 4 uses a male sample. Second, the feedback on the leadership potential of the participant can be criticized for being overly individualistic. One may wonder if, as our theoretical arguments suggest, the simple placement of participants in a dominant position can produce effects, perhaps of a more modest nature, but in line with the results of Study 3. Consequently, in Study 4, we went to the other extreme and allocated participants in a purely random manner to different social positions, without false feedback. No existing studies, to our knowledge, have ever tested the effects of such a minimal social position on prejudice toward socially significant outgroups. Being randomly put in a high- or low-status position has, to be sure, been shown to have psychological impact on other variables such as social perception and cognitive performance (see Humphrey, 1985; Langer & Benevento, 1978). These findings lead us to predict that such effects can also be obtained on measures of prejudice and group dominance. Finally, whereas in Study 3 we have ruled out the role of positive mood, an alternative explanation in terms of increased self-esteem may also be suggested. Perhaps participants in the social dominance condition experience a boost in their self-esteem, and perhaps this is an additional factor, apart from SDO, that can explain the results. For this reason, the role of personal self-esteem was examined in Study 4.
Method Participants. The participants were 30 male students, aged between 19 and 30 years old (M ⫽ 22.57). All were recruited on the campus of the Universite´ Blaise Pascal in a random manner and accepted to take part in a short experiment. Exclusion of missing data accounts for the slight variation in the number of participants in the analysis. Procedure. The participants were told that there were in fact two short experiments, one on behaviors within a business organization, and another on social perception that would involve answering a questionnaire. For the first “experiment,” the experimenter instructed the participants to follow the instructions on the computer screen of a portable Macintosh computer. The entire first part of the study, including the experimental manipulation, was a modified version of the computer task used in Study 3. The first screen indicated that this was a “study of behaviors within business-like organizations” and that the computer was going to assign randomly to the participant one of three possible positions in the organization: Director, Assistant Director, or Receptionist. The participants were then instructed to click on a dice to find out which position was allocated to themselves. After spinning for a while, the computer indicated to half of the participants that they had the position of Director, with a message stating that “as Director, you will be in a position of high responsibility.” The other half of the participants, at random, received the message stating that they were a Receptionist and that “as Receptionist, you will be in a position of low responsibility.” These statements are similar to those used in Study 3 (nobody was assigned to the position of Assistant Director). However, the participants in Study 4 are aware that the position was allocated in a purely random manner. In this sense, Study 4 examines the effects of a minimal social position. The remaining procedure (and debriefing) is essentially the same as in Study 3, with the exception of some changes in the dependent variables. Dependent measures. Personal self-esteem was measured with the 10-item scale of Rosenberg (1965). This scale has been widely used in previous research as a reliable and valid indicator of personal self-esteem (see Crocker et al., 1998). In the present study, the reliability is acceptable (␣ ⫽ .91). SDO was measured with the 16-item scale (␣ ⫽ .92). These two measures were administered as parts of Gordon’s Personality Inventory. In other words, 26 filler items that were used in Study 3 were replaced by items measuring personal self-esteem and SDO, respectively. No participants were given any feedback after completing Gordon’s Personality Inventory. They went on to do the next task involving decisions about how to organize an office. Prejudice was measured, as in Study 3, as part of a “second and unrelated study on social perception.” The 15-item scale of prejudice, our main measure of prejudice in previous studies, was not used in Study 4. There are two important reasons for this. First, because this is a scale of generalized prejudice and because SDO is also a general orientation toward intergroup relations, one might infer that our findings merely stem from the fact that these two scales have similar “generalized” content. But this is not so, as Study 3 shows that the findings hold for measures of neosexism and ingroup bias. We believe that the reason for the strong results that are obtained using the 15-item scale of prejudice is that this scale provides a good assessment of attitudes toward North African immigrants who represent the prime target of prejudice and discrimination in France. To reinforce this claim, we used a measure of ingroup bias in Study 4 involving the ratings on a 7-point scale (1⫽ very unfavorable; 7⫽ very favorable) of 20 different groups: the French, German, Swiss, Austrian, Italian, Spanish, Portuguese, Europeans, Moroccan, Algerian, Tunisian, North African (Maghre´bins), Turkish, Chinese, Japanese, Russian, British, American, women, and men. We expect the effect of social dominance to be particularly evident on bias against the people from Maghreb—Moroccans, Algerians, Tunisians—who collectively represent North Africans (also called Maghre´bins in French, the actual word that was used as one of the target groups).
SOCIAL DOMINANCE AND PREJUDICE
Figure 5. Effect of placement in a dominant social position (Director vs. Receptionist) on bias against North Africans (NA, outgroup) and in favor of the French (Fr, ingroup; Study 4).
Results and Discussion Manipulation check. To assess the effectiveness of the experimental manipulation, we examined, as in Study 3, the position that participants selected for themselves in the hierarchical structure. It may be recalled that the structure has four levels, coded from 1 (top level) to 4 (bottom level). A comparison of the two conditions on this measure indicates that participants in the Director condition (M ⫽ 1.25, SD ⫽ 0.62) selected for themselves a higher position in the hierarchy than those in the Receptionist condition (M ⫽ 2.88, SD ⫽ 1.15), F(1, 27) ⫽ 19.61, p ⬍ .001. This evidence suggests that we succeeded in creating two conditions, only one of which led participants to see themselves as occupying a more dominant social position. Prejudice. To measure prejudice, attitudes toward 20 different target groups were obtained of which three can be considered as ingroups (French, men, and Europeans) and 17 as outgroups for our participants. We do not expect that prejudice will be expressed equally toward all these outgroups. As noted above, and considering the results of Study 3, we predict an effect of condition on bias against North Africans. Nevertheless, as a preliminary analysis, we simply constructed two scales, an ingroup scale averaging scores for the three ingroups (␣ ⫽ .87) and an outgroup scale averaging scores of all 17 outgroups (␣ ⫽ .94). Using these scores, a 2 (condition) ⫻ 2 (target group: ingroup vs. outgroup) analysis of variance (ANOVA), with repeated measure on the last factor, revealed a main effect of target, F(1, 22) ⫽ 7.93, p ⬍ .01. Ingroups are rated more favorably (M ⫽ 5.25, SD ⫽ 1.19) than outgroups (M ⫽ 4.90, SD ⫽ 0.94). The effect of condition is not reliable (F ⬍ 1). However, the predicted interaction is significant, F(1, 22) ⫽ 7.50, p ⬍ .01. Participants in the Director condition show clear evidence of a bias in favor of their ingroups (M ⫽ 5.50, SD ⫽ 0.97) against outgroups (M ⫽ 4.66, SD ⫽ 0.77). In contrast, participants in the Receptionist condition rate outgroups (M ⫽ 5.09, SD ⫽ 1.34) as favorably as ingroups (M ⫽ 5.08, SD ⫽ 1.05). Although this is consistent with our prediction, we expect such results to stem in large part from a bias against North
Africans, not from a bias against high-status groups (e.g., German, American, or Japanese). Indeed, SDT suggests that dominants should be more favorable toward high-status groups. If we select as outgroups only these three target groups (German, American, and Japanese), and use the same three ingroups, the ANOVA reveals a main effect of target group, F(1, 23) ⫽ 13.67, p ⬍ .001, with ingroups (M ⫽ 5.26, SD ⫽ 1.17) being evaluated more favorably than these high-status outgroups (M ⫽ 4.60, SD ⫽ 1.09). However, there are no other significant effects: Directors do not display more bias than Receptionists, F(1, 23) ⫽ 1.95, ns. However, when using French as the ingroup and North Africans as the outgroup (e.g., Moroccan, Algerian, Tunisian, and North African), the main effect of condition is not reliable (F ⬍ 1); but there is a main effect of target group, F(1, 27) ⫽ 6.47, p ⬍ .017, and the Condition ⫻ Target Group interaction is clearly significant, F(1, 27) ⫽ 6.75, p ⬍ .01. As Figure 5 shows, Directors are biased against North Africans but Receptionists are not. SDO and self-esteem. Participants in the Director condition score marginally higher on SDO (M ⫽ 3.02, SD ⫽ 1.19) compared with those in the Receptionist condition (M ⫽ 2.29, SD ⫽ 0.98), F(1, 27) ⫽ 3.12, p ⫽ .089.7 In contrast, Directors do not display reliably higher self-esteem (M ⫽ 5.17, SD ⫽ 0.99) compared with Receptionists (M ⫽ 4.63, SD ⫽ 1.36), F(1, 27) ⫽ 1.29, ns. 7
The effect of the experimental manipulation is somewhat stronger and significant on the OEQ factor (Jost & Thompson, 2000) of the SDO (␤ ⫽ .38, p ⬍ .05). The mean in the Director condition is 3.02 (SD ⫽ 1.17) and the mean in the Receptionist condition is 2.14 (SD ⫽ 0.95), F(1, 27) ⫽ 4.81, p ⬍ .05. However, the results on the group-based dominance factor (Jost & Thompson, 2000) of the SDO are virtually identical with Directors having a mean of 2.97 (SD ⫽ 1.35) and Receptionists having a mean of 2.35 (SD ⫽ 1.09), F(1, 26) ⫽ 1.62, ns. The experimental effect obtained on the total score of the SDO in Study 4 (␤ ⫽ .33) is similar in size to the one observed in Study 3 (␤ ⫽ .28). Thus, differences in sample size, and statistical power, are involved here rather than any substantial differences.
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
Table 3 Correlations Between Self-Esteem, SDO, and Intergroup Attitudes (Study 4) Measure
Overall ingroup bias
North Africans (outgroup)
Bias against North Africans
Note. SDO ⫽ Social dominance orientation. ** p ⬍ .01.
Mediation analysis. To be considered as a possible mediator of the effects of condition on prejudice, SDO or self-esteem must be predictive of the dependent variable. Table 3 displays the relevant correlations between SDO, self-esteem, and intergroup attitudes. It can be seen first that self-esteem does not correlate significantly with any measure of intergroup attitudes. Consequently, self-esteem cannot possibly explain why participants in the Director condition display greater ingroup bias than those in the Receptionist condition. However, SDO is very strongly and significantly related to the tendency to express negative attitude toward North Africans and to display a bias in favor of the French against North Africans. Table 3 also shows that SDO is not significantly correlated with a generalized negative attitude toward all outgroups, neither with a generalized ingroup bias, nor with attitude toward the ingroup. These results suggest that SDO may account for the effect of condition on bias against North Africans but not for the effect on bias against all outgroups. As Figure 6A indicates, a series of regression analyses show that this is borne out. The effect of the experimental condition, which is to increase bias against North Africans (␤ ⫽ .42, p ⬍ .05), becomes nonsignificant when SDO is statistically controlled (␤ ⫽ .25, ns; z ⫽ 1.65, p ⬍ .10). However, the predictive power of SDO is maintained even when condition is statistically controlled (␤ ⫽ .51, p ⬍ .01). This confirms Studies 1, 2, and 3 in showing that SDO mediates the effect of a dominant social position on prejudice. In contrast, as noted above, the effect of condition on bias against all outgroups and in favor of all ingroups is also significant (␤ ⫽ .49, p ⬍ .017). But this effect remains significant (␤ ⫽ .48, p ⬍ .027) even when SDO is statistically controlled. In other words, although SDO is a mediator when intergroup bias against North Africans is involved, SDO is not a mediator when a generalized intergroup bias against all outgroups is involved. In fact, in the case of the measure of overall bias, once the effect of experimental condition is considered, SDO does not add anything at all to the prediction of this overall measure of prejudice (␤ ⫽ .05, ns). A final mediation analysis was performed to test the reverse model, predicting that bias against North Africans mediates the effect of condition on SDO. As Figure 6B indicates, support for this model is also found. The effect of conditions on SDO (␤ ⫽ .33, p ⬍ .10) becomes unreliable when bias against North Africans is statistically controlled (␤ ⫽ .10, ns, z ⫽ 1.95, p ⬍ .10).8 This suggests that there may be some feedback loop operating with ingroup bias and social dominance reinforcing each other.
General Discussion The present series of studies was designed to investigate the relations between social position, SDO, and prejudice within arbitrary-set stratification systems. Hypotheses derived from three
distinct conceptualizations of the way SDO operates were tested in both field studies and laboratory experiments. The personality model was predicting that SDO, as a personality disposition, drives the selection of social positions. This hypothesis has received considerable support in previous research (e.g., Pratto et al., 1997). Our findings add to this body of evidence by showing that even at the beginning of their first year, university students in law, an H-E academic major, display significantly higher scores on the SDO than psychology students. The second prediction of the personality model is that SDO should predict prejudice regardless of the position of individuals in the social structure. This prediction can be contrasted with that of a Person ⫻ Situation model, which specifically suggests that SDO should be more strongly related to prejudice in certain situations than in others (e.g., the ideological asymmetry hypothesis). Our results do not provide unequivocal support for either proposition. First, Studies 1 and 2 both show that among upper-level students, the relation between SDO and prejudice is similar in law and in psychology. This evidence supports the personality model and contradicts the ideological asymmetry hypothesis. Second, however, among first-year students, we find that the relation between SDO and prejudice is reliably higher in law than in psychology. This is consistent with the Person ⫻ Situation model. However, neither the personality model nor the interactionist model can explain why the results would differ as a function of the academic year of the students. In contrast, the GSM was predicting different results as a function of academic year. As such, this model can be used to integrate in a single framework hypotheses from the other two models. In a first stage, SDO would be involved in the selection of a social position (personality model) and would function as a moderator of the effect of social context in the explanation of prejudice (Personality ⫻ Situation model). At this stage, and consistent with the definition of a moderator, SDO is not assumed to be affected by social position. In a second stage, however, SDO would change as a function of social position and become a mediating variable linking social position to prejudice, consistent with the GSM. Although this is undoubtedly what happens in many social and 8
Using the OEQ factor of the SDO, on which a statistically significant effect of condition is observed, reveals the same results as those reported for the total scale. The effect of condition on bias against North Africans (␤ ⫽ .45, p ⬍ .05) becomes nonsignificant when controlling for OEQ (␤ ⫽ .26, ns), and this is a statistically reliable reduction in the variance accounted for by experimental condition (z ⫽ 1.97, p ⬍ .05). Similarly, the reverse model is also supported when using the OEQ factor ( z ⫽ 2.01, p ⬍ .05). The effect of condition on OEQ (␤ ⫽ .38, p ⬍ .05) becomes unreliable when controlling for the level of bias against North Africans (␤ ⫽ .15, ns).
SOCIAL DOMINANCE AND PREJUDICE
Figure 6. Test of (A) social dominance orientation (SDO) as a mediator of the effect of experimental conditions on bias against North Africans, and (B) test of the reverse model positing ingroup bias as a mediator of the effect of conditions on SDO (Study 4). Path weights are standardized. The path weights in parentheses do not control for the effect of the mediator. ⫹p ⬍ .10. *p ⬍ .05. **p ⬍ .01.
organizational settings, considering the results of Studies 3 and 4, we are able to suggest that the first stage noted above (involving self-selection) is not an absolute requirement: Even when people are randomly promoted to a dominant social position, they become more likely than others to score high on the SDO. In other words, regardless of whether self-selection is potentially operating (Studies 1 and 2) or not (Studies 3 and 4), the evidence is consistent with the hypothesis that being promoted to a dominant social position has a definite impact on SDO. Indeed, the three main predictions of the GSM were confirmed in all four studies. We will discuss our findings concerning these three hypotheses in turn.
Social Position and SDO: How People Acquire Different Levels of SDO The first prediction of the GSM is that people in a dominant social position will score higher on the SDO than others. All four studies confirm this prediction, consistent with previous research (see Guimond & Dambrun, 2002; Schmitt et al., in press; Sidanius & Pratto, 1999). To these recent findings, our data add several new elements. We show for the first time that exposure to higher education is negatively related to SDO in psychology but positively related to SDO in law. As such, these results question what one is led to believe by studying the global effect of formal education (e.g., Sinclair et al., 1998; Wagner & Zick, 1995). More specifically, the finding that upper-level law students score higher on the SDO than first-year law students does not fit well with the view reinstated recently by Hewstone, Rubin, and Willis (2002) that education reduces intergroup bias. It does fit well with our argument that being socialized in a dominant social position increases scores on the SDO. This argument is buttressed by our experimental findings in Studies 3 and 4, which cannot be explained by self-selection, and which also reveal that promoting individuals to a dominant social position leads to an increase in SDO. This evidence has important implications for current views about the nature of what the SDO scale is measuring. Scores on the SDO do not appear to be independent of the position of individuals
in the social structure. This has been acknowledged by social dominance theorists. However, it must also be acknowledged as Schmitt et al. (in press) pointed out, that to define SDO as a general orientation toward intergroup relations, regardless of one’s position in the intergroup structure, without any qualifications, cannot be similarly maintained. The present research also means that the conceptualization of SDO as a personality trait, as a stable and enduring psychological disposition, is seriously challenged. In our experiments, when participants randomly allocated to a dominant social position display higher scores on the SDO, we know that this is not because they have an enduring psychological disposition toward group-based inequality. Of course, it can still be maintained that although the total score on the SDO reliably changes according to the situation, such change is peripheral, and there remains a deeper aspect that does not change in such sudden manner. One may assert that the rank order of participants has not changed (Van Laar & Sidanius, 2001). Still, the fact remains that we find significant and consistent shifts in SDO as a function of social position, something that seems clearly in line with Duckitt’s (2001) view that SDO is measuring ideological beliefs.
Social Position and Prejudice: Challenging the Personality Model The second basic hypothesis of the GSM is that those in a dominant social position will display higher levels of prejudice than others. Repeated testing of this hypothesis, across different populations, in different contexts, and with different measures, also provides converging evidence that strongly supports this prediction. These results have implications for any explanation that asserts that prejudice results from deep-seated personality predispositions. Assuming that prejudice is largely a matter of personality (e.g., Altemeyer, 1998; Heaven & Quintin, in press) leads to the hypothesis that whatever social position is ascribed to an individual, he or she should remain with the same level of prejudice. Any relation between social position and prejudice should be accounted for by the personality predispositions that lead to the selection of a particular position. Overall, our results are inconsis-
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
tent with this view. Change in prejudice is observed as a function of social position and regardless of any psychological predispositions of the individual. These findings are significant because recent research on similar issues by Reynolds et al. (2001) and Verkuyten and Hagendoorn (1998) have not succeeded in experimentally altering prejudice. Their findings suggested some limitations to a personality approach by revealing that personality predicts prejudice only in certain context, but there was no experimental evidence for the view that social factors can change prejudice. Our findings are very clear in this respect and cannot be accounted for by self-selection. That is, it cannot be argued that the effects of a dominant social position on prejudice that we have shown are accounted for by a preexisting personality predisposition. This is demonstrated most clearly perhaps in Study 4 where, on a purely random basis, participants assigned to the position of Director display more intergroup bias than those assigned to the position of Receptionist. This evidence brings to mind the findings of minimal group experiments (Tajfel & Turner, 1986). In these experiments, intergroup discrimination is observed even when participants are categorized into distinct social groups on a random basis. However, the major difference is that in minimal group experiments, intergroup bias is measured in relation with the minimal groups that have no meaning outside of the laboratory. In contrast, our results concern bias against socially significant outgroups. Thus, although questions have been raised about the relevance of findings from minimal group experiments to explain real-world prejudice (e.g., Mummendey & Wenzel, 1999), these questions cannot be raised against our findings. The evidence related to this second hypothesis of the GSM also challenges the Person ⫻ Situation model. As Chen et al. (2001) have recently argued, such a model suggests that a position of power (situational factor) does not inevitably lead to greater racism and prejudice. It can also have the opposite effect. To account for the effect of the situation, this model suggests that one must consider characteristics of the individual. Thus, for Chen et al., power leads to more prejudice only among individuals with an exchange relationship orientation, because these individuals are more likely to mentally associate power with such self-interested goals. Because we are dealing with slightly different concepts and measures, our results do not have a direct bearing on such claims. They do, however, have direct implications for the theoretical explanation that is at the basis of this model. We find clear support for this theoretical explanation only among first-year students in Study 2, not among upper-level students. This suggests that the validity of this model may be limited to certain circumscribed conditions. Indeed, the results of our field and laboratory studies indicate that the situation itself can lead to more prejudice, regardless of the predispositions of the individuals involved. However, and this is an important point, this is not to suggest that leadership as such, or any position of power in itself will produce an increase in prejudice. For example, would the leader of a civil rights movement display more racism than other members of this movement? One would think not. Yet, does this not follow from the theoretical position that we have put forward? Not at all. Such a prediction would follow from a model based on a generic concept of power and it is to solve this problem that Chen et al. proposed a Person ⫻ Situation model. Drawing from SDT, the GSM suggests an alternative solution. As stated in the introduction, SDT distinguishes between social positions and ideologies that reinforce group-based inequality
(H-E) and those that attenuate group-based inequality (H-A). The leader of a civil rights movement would clearly fall within the latter category. Furthermore, it is only socialization in a position in an H-E environment that is expected to increase H-E legitimizing myths (e.g., racism, sexism, conservatism). Holding a top position in an H-A environment is expected to increase H-A legitimizing myths (e.g., socialism, feminism). Research on group socialization has produced evidence clearly consistent with such a theoretical framework. For instance, Guimond, Palmer, and Be´ gin (1989) have shown, in a large cross-sectional study, that social science students (H-A) display significantly more positive attitudes toward “socialists” (H-A legitimizing myth) with an increasing amount of education, whereas exactly the opposite is observed among commerce students (H-E). These findings were replicated in a longitudinal study among students from a different institution (see Guimond & Palmer, 1996b). No reliable differences were observed at the beginning of the first year of university, but, because of field-specific change over time, reliable differences were observed among the same students in their third year, with those in the social sciences being significantly more favorable toward socialists than those in commerce. There is also considerable evidence showing that commerce students display an increasing tendency to legitimize socioeconomic inequality with an increasing amount of education, whereas the opposite is the case among social science students (see Guimond, 1998, 1999; Guimond, Be´ gin, & Palmer, 1989; Guimond & Palmer, 1990, 1996a, 1996b). Clearly, our predictions and our findings do not follow from a generic concept of power or leadership. It remains to be seen whether future research can also document experimentally that power and leadership in an H-A setting can influence H-A legitimizing myths. Furthermore, it should be noted that by making appropriate distinctions within psychology and even within law, one could make different predictions than those examined here. For example, within law, it is possible to distinguish between corporate lawyers (H-E) and civil rights lawyers (H-A). There is evidence that these different orientations within the field of law are associated with corresponding differences in SDO (see Sidanius, Liu, Shaw, & Pratto, 1994). Further research could examine the extent to which different levels of prejudice are also observed by making these distinctions.
Is It Simply “Individual” Dominance? Before considering the mediating role of SDO, it is important to clarify another, more methodological, aspect of the present research. One question that might be raised about our experimental manipulation is whether it is dealing with individual dominance rather than group dominance. Despite the cover story, which is clearly individualistic, our manipulation and conceptual framework are dealing with group dominance. As social identity and self-categorization theorists have argued, a point made also within relative deprivation theory, the distinction between interindividual behavior and intergroup behavior does not rest on the number of individuals involved in a particular setting (Oakes, Haslam, & Turner, 1994, Turner, 1999b). The minimal group paradigm, which is well established as a paradigm to study intergroup behaviors, does not involve any real groups of participants. There is no interaction whatsoever between group members within a typical minimal group experiment (see Tajfel & Turner, 1986). Each participant, as an individual, is simply classified as a member of a
SOCIAL DOMINANCE AND PREJUDICE
given social category, and an attempt is made to study if that individual will be acting in terms of this membership in a social category. Similarly, in Study 3, our procedure is basically a way to categorize the participants as “leaders” as opposed to “followers” within a business organization. Furthermore, it is known that the SDO scale is not a measure of interpersonal dominance, but a measure of intergroup dominance (see Pratto et al., 1994; Sidanius & Pratto, 1999). The fact that our manipulation has a significant effect on this scale, and on prejudice against particular outgroups, suggests that our manipulation led participants to categorize themselves as members of a dominant social category. In this sense, our manipulation is not to be confused with the cover story, which stated that we were interested in studying leadership. The Chief Executive Officer (CEO) of a large corporation can be a well-known individual. But when he or she is acting as CEO, they are in a dominant social position, and are not simply acting as an individual. The classic study of Lieberman (1956) provides a clear illustration of this point. This was a longitudinal study among workers who, as individuals, were promoted as foremen and others who were made stewards. It is clear from the results that the workers who became foremen also became members of a new social category. Lieberman found that “workers who were made foremen tended to become more favorable toward management, and the workers who were made stewards tended to become more favorable toward the union” (p. 493). In fact, consistent with the experimental effect of social position demonstrated in the present research, when some workers who were made foremen were made workers again, their attitudes reverted back to their initial position.
SDO as a Mediator: Why Social Positions Can Increase or Decrease Prejudice Our third and most central theoretical and empirical contribution concerns the role of SDO as a mediator. Our goal in this research was to examine one critical feature of the process of group socialization that relates to the position in the social structure that is acquired as a result of this socialization. Simply put, one of the compelling dimensions of group socialization is that it involves people who are in the process of changing social positions (Levine et al., 1998). We have argued that the extent to which this acquisition of a position in the social structure generates higher levels of prejudice may have a lot to do with the extent to which this new position increases or decreases SDO. More specifically, we have proposed that the finding that SDO predicts prejudice on the one hand, and those suggesting that the position in the social structure has an effect on SDO, on the other, might be profitably integrated. The theoretical integration of these relatively separate lines of research that we have formulated in relation to the GSM suggests that SDO can be seen as the mediating variable linking social position to prejudice. Although consistent with SDT under certain conditions, such a theoretical proposition has not been articulated as such within SDT and no existing experimental research has, to our knowledge, ever tested such a mediating function of SDO. In four studies, we found consistent evidence in support of such a theoretical model. In Studies 1 and 2, students socialized in law (H-E), who should be well-versed in terms of issues of fairness and justice, display higher levels of bias against outgroups, and these effects of academic major on prejudice are mediated by SDO. In Studies 3 and 4, students randomly promoted to a dominant social position within an H-E setting display higher levels of prejudice,
and these experimental effects are also mediated by SDO. Thus, SDO represents a generative mechanism through which individuals located in different positions within the social structure come to express higher or lower levels of prejudice against stigmatized social groups. That is, not only does the evidence indicate very clearly that dominants are more prejudiced than others, but, in addition, there is strong empirical confirmation for the fact that these effects of a dominant social position emerge because of the intermediate role of SDO: When we control for SDO, the effect of social position on prejudice is no longer reliable. We have also tested the reverse model holding that prejudice mediates the effects of social position on SDO. Although support for this model is generally weaker, it is not totally absent. In Study 4, we found that both models appear to be valid. This suggests that a more dynamic conception may be in order with some feedback loop operating between intergroup bias and social dominance. Thus, Whitley (1999) has reported that SDO mediates gender differences in prejudice whereas Schmitt et al. (in press) found support for the reverse: Prejudice also mediates gender differences in SDO. These studies do not however include an experimental manipulation that could sustain the causal ordering that is assumed among the variables. This is the case also in Study 1 of the present research. However, Study 2 involves a quasi-experimental design by including academic year as a factor, and Studies 3 and 4 both involve an experimental design. These studies can be ordered as a function of the psychological significance of social positions for the participants from high (Study 2) to medium (Study 3) and low (Study 4). That is, in Study 2, upperlevel students have spent at least 3 years in their position; in Study 3, participants had to complete a fairly elaborate test before knowing their position; in Study 4, participants merely click on a dice to find out. With this ordering in mind, our results show increasing support for the exclusive mediating function of SDO with increasing psychological significance of social position. They also show increasing additional support for the mediating role of prejudice with decreasing psychological significance of social position. There is no evidence that supports an exclusive mediating function of prejudice. In other words, the message that comes out is that the mediating function of SDO observed under controlled (but artificial) conditions is likely to be stronger under real-life conditions when being in a subordinate or in a dominant social position is fully experienced. The fact that it is more likely that SDO leads to prejudice rather than the other way around is also the position taken by Duckitt (2001). In a series of structural equation modeling analyses, Duckitt (2001) found support for his dual-process model of prejudice in which SDO is considered as a measure of ideological beliefs, not as a measure of personality. He concluded that “personality does not have direct effects on prejudice, but has substantial impacts on ideological beliefs . . . and it is these ideological attitudes that influence prejudice” (p. 90). In other words, consistent with the present thesis, Duckitt (2001) also challenges the personality model of prejudice. He argued that scales such as the SDO or RWA are measures of ideological beliefs, not personality. He suggested that the causal direction is from general ideological beliefs (i.e., SDO) to prejudice, not the other way around—a point that is generally reinforced by our correlational and experimental findings. Duckitt (2001) pointed out that experimental evidence from Katz and Hass (1988) supported such a causal argument. One reason for the importance of abstract, general ideological beliefs is
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
that they can serve to hide some specific group interests that are pursued. As Gouldner (1976) noted, “being more generalized, an ideology may appeal to larger and more diversified groups” (p. 221) and “may solve the problem of how the pursuit of private interests may, nonetheless, generate public support” (p. 219).
Conclusion and Future Prospects Our results support the GSM. We show that a dominant social position has an effect on prejudice and we show that SDO as a measure of ideological beliefs is the mechanism accounting for this effect. These findings raise serious questions about SDT. As Popper (1987) has argued, any theory can be considered a scientific theory only to the extent that it is empirically testable. When there is no way to invalidate a theory, that theory is not scientific. From that point of view, the fact that predictions derived from three different conceptualizations of the way SDO operates in the explanation of prejudice may be compatible with SDT is highly problematic. Indeed, to claim that our findings are supportive of SDT would be close to suggesting that SDT cannot be invalidated. The implication then is that unless some clarifications are offered, the finding that SDO predicts prejudice may not be held as unambiguous support for SDT because that finding can be consistent with widely different theoretical explanations of prejudice. As we have argued, the three models that we have identified are not necessarily mutually exclusive. They are nevertheless distinct and cannot all be simultaneously valid. For example, the ideological asymmetry hypothesis of SDT, which received mixed support at best from the present research, leads to predictions that are contradictory with other aspects of SDT, a point acknowledged recently by Levin et al. (2002). SDT also suggests that gender differences in SDO, having a sociobiological origin, are invariant across all major cultural, social, or situational factors. In contrast, the GSM is expected to apply equally to women and men. Because of insufficient numbers of male and female participants, we were not able to directly test the invariance hypothesis in this series of research. However, comparing the behavior of women and men in experiments such as the one presented here would allow for a critical experimental test of the invariance hypothesis, which is now based only on correlational data. This is one important direction for further research. Finally, one significant feature of the present series of studies is that the experimental findings are in harmony with those obtained in the “real world” among psychology and law students. Nevertheless, the fact that randomly assigning people to a dominant position can increase bias against socially significant outgroups is striking. Recent advances in implicit social cognition and automatic behaviors suggest an alternative theoretical perspective that may also be useful to understand such effects (see Chen et al., 2001; Dijksterhuis & Bargh, 2001). According to the auto-motive model (Bargh, 1990), mental associations between a given concept and a given goal can provide an explanation for a wide range of behaviors. Research suggests that priming the concept of power, for example, may be sufficient to lead people to act in terms of goals that they mentally associate with power. Concerning the present studies, it could be argued that our experimental manipulations activated the concept of social dominance among some of our participants. To the extent that this concept is mentally associated with prejudice and discrimination against particular outgroups, then this would explain the intergroup bias observed in the
social dominance condition compared with the control condition. Such a theoretical possibility is consistent with a number of studies in the area (see Dijksterhuis & Bargh, 2001) and undoubtedly raises interesting prospects for future research. Nevertheless, additional assumptions may be needed for such a priming interpretation to account for the results of Studies 1 and 2, which do not involve any experimental manipulation and that strongly support the GSM. It may very well be that group socialization can account for the particular mental association that participants have formed between social dominance and intergroup discrimination, something that is treated as a given within current models of the effects of automatic activation. In a way, our results reveal the dark side of human psychology in that people seem able to rationalize any social arrangements that suit their purposes. But showing that the scores on the SDO can be experimentally increased gives some signs of hope by revealing that change is possible. We look forward to other research that may show how it is possible to decrease the SDO of those who are high in social dominance.
References Adorno, T. W., Frenkel-Brunswick, E., Levinson, D. J., & Sanford, R. N. (1950). The authoritarian personality. New York: Harper. Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage. Altemeyer, B. (1998). The other authoritarian personality. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 30, pp. 47–92). New York: Academic Press. Bargh, J. A. (1990). Auto-motives: Preconscious determinants of social interaction. In E. T. Higgins & R. M. Sorrentino (Eds.), Handbook of motivation and cognition (Vol. 2, pp. 93–130). New York: Guilford Press. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Bereiter, C., & Freedman, M. B. (1962). Fields of study and the people in them. In N. Sanford (Eds.), The American college: A psychological and social interpretation of higher learning (pp. 563–596). New York: John Wiley. Bettencourt, B. A., Dorr, N., Charlton, K., & Hume, D. L. (2001). Status differences and in-group bias: A meta-analytic examination of the effects of status stability, status legitimacy, and group permeability. Psychological Bulletin, 127, 520 –542. Billig, M. (1976). Social psychology and intergroup relations. London: Academic Press. Blass, T. (1991). Understanding behavior in the Milgram obedience experiments: The role of personality, situations, and their interactions. Journal of Personality and Social Psychology, 60, 398 – 413. Bobo, L. (1988). Group conflict, prejudice, and the paradox of contemporary racial attitudes. In P. A. Katz & D. A. Taylor (Eds.), Eliminating racism: Profiles in controversy (pp. 85–114). New York: Plenum Press. Bodenhausen, G. (1993). Emotions, arousal, and stereotypic judgments: A heuristic model of affect and stereotyping. In D. M. Mackie & D. L. Hamilton (Eds.), Affect, cognition, and stereotyping: Interactive processes in group perception (pp. 13–37). New York: Academic Press. Brewer, M. B., & Brown, R. J. (1998). Intergroup relations. In D. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (4th ed., pp. 554 –594). Boston: McGraw-Hill. Britt, T. W., Boniecki, K. A., Vescio, T. K., Biernat, M., & Brown, L. M. (1996). Intergroup anxiety: A Person ⫻ Situation approach. Personality and Social Psychology Bulletin, 22, 1177–1188.
SOCIAL DOMINANCE AND PREJUDICE Brown, R. (1995). Prejudice: Its social psychology. Oxford, England: Blackwell. Castano, E., Yzerbyt, V., Paladino, M. -P., & Sacchi, S. (2002). I belong, therefore, I exist: Ingroup identification, ingroup entitativity, and ingroup bias. Personality and Social Psychology Bulletin, 28, 135–143. Chambon, M. (1990). La perception d’une discipline scolaire par les e´ le`ves: Repre´ sentation et effets identitaires [The representation of an academic discipline by the students: Effects on social identity]. European Journal of Psychology of Education, 6, 337–354. Chen, S., Lee-Chai, A. Y., & Bargh, J. A. (2001). Relationship orientation as a moderator of the effects of social power. Journal of Personality and Social Psychology, 80, 173–187. Crocker, J., Major, B., & Steele, C. (1998). Social stigma. In D. Gilbert, S. Fiske, & G. Lindzey (Eds.), Handbook of social psychology (4th ed., pp. 504 –543). Boston: McGraw-Hill. Dambrun, M. (2001). Dominance sociale et pre´ juge´ s: La re´ gulation sociale des cognitions intergroupes [Social dominance and prejudice: The social regulation of intergroup cognitions]. Unpublished doctoral dissertation, Laboratoire de psychologie sociale de la cognition, Universite´ Blaise Pascal, Clermont-Ferrand, France. Dambrun, M., & Guimond, S. (2001). La the´ orie de la privation relative et l’hostilite´ envers les Nord-Africains [Relative deprivation theory and hostility toward North Africans]. International Review of Social Psychology, 14, 57– 89. Dambrun, M., Guimond, S., & Duarte, S. (2002). The impact of hierarchyenhancing vs. attenuating academic major on stereotyping: The mediating role of perceived social norm. Current Research in Social Psychology, 7, 114 –136. Dambrun, M., Maisonneuve, C., Duarte, S., & Guimond, S. (2002). Mode´ lisation de quelques de´ terminants psychosociaux de l’attitude envers l’extreˆ me droite [Modeling the social-psychological determinants of support for extreme right-wing movements]. Cahiers Internationaux de Psychologie Sociale, 55, 49 – 63. Danso, H. A., & Esses, V. M. (2001). Black experimenters and the intellectual test performance of white participants: The tables are turned. Journal of Experimental Social Psychology, 37, 158 –165. Dijksterhuis, A., & Bargh, J. A. (2001). The perception-behavior expressway: Automatic effects of social perception on social behavior. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 33, pp. 1– 40). San Diego, CA: Academic Press. Duckitt, J. (2001). A dual-process cognitive-motivational theory of ideology and prejudice. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 33, pp. 41–112). San Diego, CA: Academic Press. Ellemers, N. (2001). Individual upward mobility and the perceived legitimacy of intergroup relations. In J. T. Jost & B. Major (Eds.), The psychology of legitimacy: Emerging perspectives on ideology, justice, and intergroup relations (pp. 205–222). New York: Cambridge University Press. Esses, V. M., Jackson, L. M., & Armstrong, T. L. (1998). Intergroup competition and attitudes toward immigrants and immigration: An instrumental model of group conflict. Journal of Social Issues, 54, 699 – 724. Feldman, K. A., & Newcomb, T. M. (1969). The impact of college on students. San Francisco: Jossey-Bass. Fiske, S. T. (1998). Stereotyping, prejudice and discrimination. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (4th ed., pp. 357– 411). Boston: McGraw-Hill. Fiske, S. T. (2000). Stereotyping, prejudice, and discrimination at the seam between the centuries: Evolution, culture, mind and brain. European Journal of Social Psychology, 30, 299 –322. Forgas, J. P., & Fiedler, K. (1996). Us and them: Mood effects on intergroup discrimination. Journal of Personality and Social Psychology, 70, 28 – 40. Glick, P., & Fiske, S. T. (2001). An ambivalent alliance: Hostile and
benevolent sexism as complementary justifications for gender inequality. American Psychologist, 56, 109 –118. Gouldner, A. W. (1976). The dialectic of ideology and technology: The origin, grammar and future of ideology. New York: Seabury Press. Guimond, S. (1992). Les effets de l’e´ ducation post-secondaire sur les attitudes intergroupes: L’importance du domaine d’e´ tudes [The effects of higher education on intergroup attitudes: The importance of academic major]. Revue Que´ be´ coise de Psychologie, 13, 74 –93. Guimond, S. (1995). Encounter and metamorphosis: The impact of military socialization on professional values. Applied Psychology: An International Review, 44, 251–275. Guimond, S. (1998). Processus de socialisation dans l’enseignement supe´ rieur: le pouvoir de la connaissance [Socialization processes in higher education: The power of knowledge]. In J. -L. Beauvois, R. V. Joule, & J.-M. Monteil (Eds.), 20 ans de psychologie sociale expe´ rimentale Francophone [Twenty years of francophone experimental social psychology] (pp. 231–272). Grenoble, France: Presses Universitaires de Grenoble. Guimond, S. (1999). Attitude change during college: Normative or informational social influence? Social Psychology of Education, 2, 237–261. Guimond, S. (2000). Group socialization and prejudice: The social transmission of intergroup attitudes and beliefs. European Journal of Social Psychology, 30, 335–354. Guimond, S. (2001). Epistemic authorities in higher education: The relative influence of peers, faculty and courses on attitude formation and change. In F. Butera & G. Mugny (Eds.), Social influence in social reality (pp. 211–223). Go¨ ttingen, Germany: Hogrefe & Huber. Guimond, S., Be´ gin, G., & Palmer, D. L. (1989). Education and causal attributions: The development of “person-blame” and “system-blame” ideology. Social Psychology Quarterly, 52, 126 –140. Guimond, S., & Dambrun, M. (2002). When prosperity breeds intergroup hostility: The effects of relative deprivation and relative gratification on prejudice. Personality and Social Psychology Bulletin, 28, 900 –912. Guimond, S., Dif, S., & Aupy, A. (2002). Social identity, relative group status and intergroup attitudes: When favourable outcomes change intergroup relations . . . for the worse. European Journal of Social Psychology, 32, 739 –760. Guimond, S., & Palmer, D. L. (1990). Type of academic training and causal attributions for social problems. European Journal of Social Psychology, 20, 61–75. Guimond, S., & Palmer, D. L. (1993). Developmental changes in ingroup favouritism among bilingual and unilingual Francophone and Anglophone students. Journal of Language and Social Psychology, 12, 318 – 351. Guimond, S., & Palmer, D. L. (1996a). Liberal reformers or militant radicals? What are the effects of education in the social sciences? Social Psychology of Education, 1, 95–115. Guimond, S., & Palmer, D. L. (1996b). The political socialization of commerce and social science students: Epistemic authority and attitude change. Journal of Applied Social Psychology, 26, 1985–2013. Guimond, S., Palmer, D. L., & Be´ gin, G. (1989). Education, academic program and intergroup attitudes. Canadian Journal of Sociology and Anthropology, 26, 61–75. Harris, J. R. (1995). Where is the child’s environment? A group socialization theory of development. Psychological Review, 102, 458 – 489. Heaven, P. C. L., & Quintin, D. (in press). Personality factors predict racial prejudice. Personality and Individual Differences. Hewstone, M., Rubin, M., & Willis, H. (2002). Intergroup bias. Annual Review of Psychology, 53, 575– 604. Humphrey, R. (1985). How work roles influence perception: Structuralcognitive processes and organizational behavior. American Sociological Review, 50, 242–252. Jones, M. (2002). Social psychology of prejudice. Upper Saddle River, NJ: Prentice Hall. Jost, J. T., & Banaji, M. R. (1994). The role of stereotyping in system-
GUIMOND, DAMBRUN, MICHINOV, AND DUARTE
justification and the production of false consciousness. British Journal of Social Psychology, 33, 1–27. Jost, J. T., & Burgess, D. (2000). Attitudinal ambivalence and the conflict between group and system justification motives in low status groups. Personality and Social Psychology Bulletin, 26, 293–305. Jost, J. T., & Thompson, E. P. (2000). Group-based dominance and opposition to equality as independent predictors of self-esteem, ethnocentrism, and social policy attitudes among African Americans and European Americans. Journal of Experimental Social Psychology, 36, 209 –232. Joule, R. V., & Beauvois, J.-L. (1998). Cognitive dissonance theory: A radical view. In W. Stroebe & M. Hewstone (Eds.), European review of social psychology (Vol. 8, pp. 1–32). West Sussex, United Kingdom: Wiley. Katz, I., & Hass, R. (1988). Racial ambivalence and American value conflict: Correlational and priming studies of dual cognitive structure. Journal of Personality and Social Psychology, 55, 893–905. Kenny, D. A. (1979). Correlation and causality. New York: Wiley. Kenny, D. A. (1998). Mediation analysis. Retrieved from http://users .rcn.com/dakenny/mediate.htm Kluegel, J. R., & Smith, E. R. (1986). Beliefs about inequality. New York: Aldine de Gruyter. Ladd, E. C., & Lipset, S. M. (1975). The divided academy: Professors and politics. New York: Norton. Lambert, W. E., Moghaddam, F. M., Sorin, J., & Sorin, S. (1990). Assimilation vs. multiculturalism: Views from a community in France. Sociological Forum, 5, 387– 411. Langer, E. J., & Benevento, A. (1978). Self-induced dependence. Journal of Personality and Social Psychology, 36, 886 – 893. Lepore, L., & Brown, R. (1997). Category and stereotype activation: Is prejudice inevitable? Journal of Personality and Social Psychology, 72, 275–287. Levin, S. (1996). A social-psychological approach to understanding intergroup attitudes in the United States and Israel. Unpublished doctoral dissertation, Department of Psychology, University of California, Los Angeles. Levin, S., Federico, C. M., Sidanius, J., & Rabinowitz, J. L. (2002). Social dominance orientation and intergroup bias: The legitimation of favoritism for high-status groups. Personality and Social Psychology Bulletin, 28, 144 –157. Levine, J. M., Moreland, R. L., & Ryan, C. S. (1998). Group socialization and intergroup relations. In C. Sedikides, C. A. Schopler, & C. A. Insko (Eds.), Intergroup cognition and intergroup behavior (pp. 283–308). London: Erlbaum. Lieberman, S. (1956). The effects of changes in roles on the attitudes of role occupants. In H. Proshansky & B. Seidenberg (Eds.), Basic studies in social psychology (pp. 485– 494). New York: Holt, Rinehart & Winston. Magnusson, D., & Endler, N. S. (Eds.). (1977). Personality at the crossroads: Current issues in interactional psychology. Hillsdale, NJ: Erlbaum. McClintock, C. G., Spaulding, C. B., & Turner, H. A. (1964). Political orientations of academically affiliated psychologists. American Psychologist, 20, 211–221. McFarland, S. (1999, July). Is authoritarianism sufficient to explain individual differences in prejudice? Paper presented at the general meeting of the European Association of Experimental Social Psychology, Oxford, England. Mendoza-Denton, R., Ayduk, O., Mischel, W., Shoda, Y., & Testa, A. (2001). Person ⫻ Situation interactionism in self-encoding (I am . . . when. . .): Implications for affect regulation and social information processing. Journal of Personality and Social Psychology, 80, 533–544. Michinov, N., Dambrun, M., Guimond, S., & Meot, A. (2002). Social dominance orientation, prejudice and discrimination. Manuscript submitted for publication.
Milgram, S. (1974). Obedience to authority: An experimental view. New York: Harper & Row. Mischel, W., & Shoda, Y. (1998). Reconciling processing dynamics and personality dispositions. Annual Review of Psychology, 49, 229 –258. Monteil, J. -M., & Huguet, P. (1999). Social context and cognitive performance: Towards a social psychology of cognition. East Sussex, United Kingdom: Psychology Press. Mullen, B., Brown, R., & Smith, C. (1992). Ingroup bias as a function of salience, relevance, and status: An integration. European Journal of Social Psychology, 22, 103–122. Mummendey, A., & Wenzel, M. (1999). Social discrimination and tolerance in intergroup relations: Reactions to intergroup difference. Personality and Social Psychology Review, 3, 158 –174. Nelson, T. D. (2002). The psychology of prejudice. Boston: Allyn & Bacon. Newcomb, T. M. (1943). Personality and social change. New York: Dryden Press. Oakes, P. J., Haslam, S. A., & Turner, J. C. (1994). Stereotyping and social reality. Oxford, England: Blackwell. Pettigrew, T. W. (1958). Personality and sociocultural factors in intergroup attitudes: A cross-national comparison. Journal of Conflict Resolution, 2, 29 – 42. Pettigrew, T. F., Jackson, J. S., Ben Bricka, J., Lemaine, G., Meertens, R. W., Wagner, U., & Zick, A. (1998). Outgroup prejudice in Western Europe. European Review of Social Psychology, 8, 241–273. Popper, K. (1987). Science: Conjectures and refutations. In J. A. Kourany (Ed.), Scientific knowledge: Basic issues in the philosophy of science (pp. 139 –157). Belmont, CA: Wadsworth. Pratto, F. (1999). The puzzle of continuing group inequality: Piecing together psychological, social, and cultural forces in social dominance theory. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 31, pp. 191–263). New York: Academic Press. Pratto, F., & Shih, M. (2000). Social dominance orientation and group context in implicit group prejudice. Psychological Science, 11, 515–518. Pratto, F., Sidanius, J., Stallworth, L. M., & Malle, B. F. (1994). Social dominance orientation: A personality variable predicting social and political attitudes. Journal of Personality and Social Psychology, 67, 741–763. Pratto, F., Stallworth, L. M., Sidanius, J., & Siers, B. (1997). The gender gap in occupational role attainment: A social dominance approach. Journal of Personality and Social Psychology, 72, 37–53. Pratto, F., Tatar, D., & Conway-Lanz, S. (1999). Who gets what and why? Determinants of social allocations. Political Psychology, 20, 127–150. Reynolds, K. J., Turner, J. C., Haslam, S. A., & Ryan, M. K. (2001). The role of personality and group factors in explaining prejudice. Journal of Experimental Social Psychology, 37, 427– 434. Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press. Sabatier, C., & Berry, J. W. (1994). Immigration et acculturation [Immigration and acculturation]. In R. Y. Bourhis & J. P. Leyens (Eds.), Ste´ re´ otypes, discrimination et relations intergroupes [Stereotypes, discrimination and intergroup relations] (pp. 261–291). Lie`ge, Belgium: Mardaga. Schein, E. H. (1984). Organizational socialization and the profession of management. In D. A. Kolb, I. M. Rubin, & J. M. McIntyre (Eds.), Organizational psychology: Readings on human behaviour in organization (pp. 7–21). Englewood Cliffs, NJ: Prentice-Hall. Schmitt, M. T., Branscombe, N. R., & Kappen, D. K. (in press). Attitudes toward group-based inequality: Social dominance or social identity? British Journal of Social Psychology. Schwarzwald, J., Koslowsky, M., & Shalit, B. (1992). A field study of employees’ attitudes and behaviors after promotion decisions. Journal of Applied Psychology, 77, 511–514. Sidanius, J. (1993). The psychology of group conflict and the dynamics of oppression: A social dominance perspective. In S. Iyengar, & W. J.
SOCIAL DOMINANCE AND PREJUDICE McGuire (Eds.), Explorations in political psychology (pp. 183–219). Durham, NC: Duke University Press. Sidanius, J., Levin, S., Liu, J., & Pratto, F. (2000). Social dominance orientation, anti-egalitarianism and the political psychology of gender: An extension and cross-cultural replication. European Journal of Social Psychology, 30, 41– 68. Sidanius, J., Liu, J. H., Shaw, J. S., & Pratto, F. (1994). Social dominance orientation, hierarchy attenuators and hierarchy enhancers: Social dominance theory and the criminal justice system. Journal of Applied Social Psychology, 24, 338 –366. Sidanius, J., & Pratto, F. (1999). Social dominance: An intergroup theory of social hierarchy and oppression. New York: Cambridge University Press. Sidanius, J., Pratto, F., & Bobo, L. (1994). Social dominance orientation and the political psychology of gender: A case of invariance? Journal of Personality and Social Psychology, 67, 998 –1011. Sidanius, J., Pratto, F., Martin, M., & Stallworth, L. (1991). Consensual racism and career track: Some implications of social dominance theory. Political Psychology, 12, 691–721. Sinclair, S., Sidanius, J., & Levin, S. (1998). The interface between ethnic and social system attachment: The differential effects of hierarchyenhancing and hierarchy-attenuating environments. Journal of Social Issues, 54, 741–757. Snyder, M., & Cantor, N. (1998). Understanding personality and social behavior: A functional analysis. In D. T. Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social psychology (4th ed., pp. 635– 679). Boston: McGraw-Hill. Tajfel, H. (1981). Social stereotypes and social groups. In J. C. Turner & H. Giles (Eds.), Intergroup behaviour (pp. 144 –167). Oxford, England: Blackwell. Tajfel, H., & Turner, J. C. (1986). The social identity theory of intergroup behaviour. In S. Worchel & W. G. Austin (Eds.), Psychology of intergroup relations (pp. 7–24). Chicago: Nelson-Hall.
Tougas, F., Brown, R., Beaton, A., & Joly, S. (1995). Neosexism: Plus c¸ a` change, plus c’est pareil. Personality and Social Psychology Bulletin, 21, 842– 849. Turner, J. (1999a, July). The prejudiced personality and social change: A self-categorization perspective. The Tajfel lecture: The 12th general meeting of the European Association of Experimental Social Psychology, Oxford, England. Turner, J. (1999b). Some current issues in research on social identity and self-categorization theories. In N. Ellemers, R. Spears & B. Doosje (Eds.), Social identity (pp. 6 –34). Oxford, England: Blackwell. Van Laar, C., & Sidanius, J. (2001). Social status and the academic achievement gap: A social dominance perspective. Social Psychology of Education, 4, 235–258. Van Laar, C., Sidanius, J., Rabinowitz, J. L., & Sinclair, S. (1999). The three Rs of academic achievement: Reading, ‘riting, and racism. Personality and Social Psychology Bulletin, 25, 139 –151. Van Maanen, J., & Schein, E. H. (1979). Toward a theory of organizational socialization. In B. Staw (Ed.), Research in organizational behavior (Vol. 1, pp. 209 –264). Greenwich, CT: JAI Press. Verkuyten, M., & Hagendoorn, L. (1998). Prejudice and self-categorization: The variable role of authoritarianism and in-group stereotypes. Personality and Social Psychology Bulletin, 24, 99 –110. Wagner, U., & Zick, A. (1995). The relation of formal education to ethnic prejudice: Its reliability, validity and explanation. European Journal of Social Psychology, 25, 41–56. Whitley, B. E. (1999). Right-wing authoritarianism, social dominance orientation, and prejudice. Journal of Personality and Social Psychology, 77, 126 –134.
Received October 15, 2001 Revision received November 4, 2002 Accepted November 12, 2002 䡲