Why are some people prejudiced while others are more tolerant? Answering this question has generated much theory and research in social psychology and the social sciences more generally. An argument is that by understanding a person's particular personality, genetic make-up (e.g. temperament) and/or childhood learning and socialization, it is possible to explain why he or she holds particular social and political attitudes including prejudice. Following the work by Adorno, Frenkel-Brunswik, Levinson, and Sanford (1950), initial interest was on the role of authoritarianism, dogmatism and cognitive style (i.e. metal rigidity, toughmindedness), but currently the focus is on social dominance orientation (SDO, e.g. Sidanius & Pratto, 1999) and right-wing authoritarianism (RWA, e.g. Altemeyer, 1988). The central thesis here is that, there is a reasonably direct relationship between people's psychology as individuals and prejudice.
RWA is understood as an individual factor, “a personality variable, a “trait” if you like” (Altemeyer, 1988, p. 3) that stems from social learning and cultural socialization and is relatively stable within a person by early adulthood. It characterizes those people who will submit to authority, who are aggressive towards socially sanctioned minority groups and who adhere to established social conventions and beliefs. Those high in authoritarianism have been found to hold punitive attitudes towards law-breakers to support the death penalty to endorse traditional roles for women and to be more prejudiced towards minority group members.
SDO is the psychological component of social dominance theory (SDT) and embodies a generalized orientation towards, and desire for, unequal relations between salient groups, regardless of whether one's own group is dominant or subordinate. SDO is argued to be multiply determined and relatively stable. It is influenced by temperament, personality and socialization experiences as well as social context, group position and gender, and as such is not only a product of certain hierarchical relationships among groups, but is also a partial cause of those relationships (Pratto, Sidanius, & Levin, 2006). In terms of its predictive power, SDO is considered to be a ‘ubiquitous motive driving most group-relevant social attitudes and allocative decisions’ (Sidanius & Pratto, 1999, p. 57). Accordingly, those high in SDO should express more prejudice towards low status group members and to some degree endorse measures that maintain dominant–subordinate hierarchical structures. A particular focus of SDT that differentiates it from the more traditional personality analysis is its emphasis on the interaction between the SDO personality and social structural factors (Sidanius, Pratto, van Laar, & Levin, 2004; but see Rubin & Hewstone, 2004).
Despite the volume of research suggesting a relationship between traditional personality variables (e.g. authoritarianism) and prejudice, a number of problems with the analysis have been identified. It has been argued that personality-based explanations: (a) tend to ignore the role that group memberships and intergroup relations play in shaping prejudiced attitudes (e.g. Billig, 1976; Duckitt, 1994; Duckitt, Wagner, du Plessis, & Birum, 2002; Pettigrew, 1958; Sherif, 1967; Tajfel & Turner, 1979); (b) cannot account for the large-scale uniformity in prejudice amongst members of certain groups and the way prejudice can increase or decrease within such groups over reasonably short time periods (e.g. Sherif, 1967) and (c) may redescribe the current reality of intergroup relations and collective understandings of those relationships rather than being measures of an individual psychological cause or drive for such structures (e.g. Billig, 1976; Duckitt, 1994; Duckitt et al., 2002; Lehmiller & Schmitt, in press; Reynolds, Turner, Haslam, & Ryan, 2001; Reynolds & Turner, 2006; Schmitt, Branscombe, & Kappen, 2003; Stellmacher & Petzel, 2005; Turner, Reynolds, Haslam, & Veenstra, 2006). A common theme, then, is that people's group memberships and associated social norms tend to be overlooked within the personality account of prejudice and that individual differences may reflect the nature of intergroup relations rather than the other way around.
The contrasting viewpoint is apparent in social identity theory and self-categorization theory (the social identity perspective, see Turner & Reynolds, 2001) where there is an explicit focus on the role played by group memberships and intergroup relations in shaping people's attitudes, beliefs and values. These theories argue that people can define themselves as either individuals in terms of their difference from others (i.e. ‘I’ and ‘me’ – personal identity) or group members where their similarities and commonalities with others become salient (i.e. ‘we’ and ‘us’ – social identity).
When a person's social identity becomes salient, there are increased perceptions of interchangeability between oneself and like-minded others – in-group members – which provides the basis for cooperation and coordinated action in achieving shared goals and interests. Moreover, sharing a social identity allows other in-group members to become a viable source of information about social reality, thus making possible processes of mutual social influence and persuasion (Turner, 1991; Turner & Oakes, 1989). Within self-categorization theory, social identity salience and processes of social influence are important because they locate the self within, and connect the self to, broader social forces. Through an analysis that recognises both the personal and social self it is possible to understand how societal factors come to impact on an individual's immediate attitudes and actions.
Shifts in the salience of a particular identity or self-categorization changes the ‘lens’ so to speak, through which the perceiver sees the world and makes sense of it. As the vantage point of the perceiver varies (e.g. an individual, Australian, academic, employee) so to can his or her cognitions, evaluations, judgments and emotional responses. People's psychology (their attitudes, behaviour, cognition) when they are responding in terms of their personal identity and interpersonal differences from others (i.e. in terms of individual personality) can be qualitatively different from when they respond when a particular group-based social identity is salient. There is now a large body of work that demonstrates this psychological discontinuity between people acting as individuals and those acting as group members – the so called discontinuity hypothesis (Turner, Hogg, Oakes, Reicher, & Wetherell, 1987; see Onorato & Turner, 2002 for a review).
One relevant series of studies is the minimal group experiments by Tajfel, Flament, Billig, and Bundy (1971) where social categorization and identification produced intergroup behaviour independent of personality and interpersonal relations. Work by Verkuyten and Hagendoorn (1998) and Reynolds et al. (2001) provide other examples. Verkuyten and Hagendoorn assessed participants' in-group stereotypes (related to the treatment of the Turkish minority in The Netherlands) and their individual differences in authoritarian attitudes (10 items based on Altemeyer's RWA scale). The salience of participants' personal or national identity (Dutch) was then manipulated and prejudice towards Turkish immigrants was assessed. In line with predictions, authoritarianism only correlated with prejudice when personal identity was salient. In the national identity condition, prejudice correlated with in-group stereotypes, but not with authoritarianism. Reynolds et al. (2001) manipulated whether people identified as individuals compared with other individuals, as males vs. females, as Australians vs. Americans, and as young vs. older people. Making certain intergroup comparisons and their associated norms, values and beliefs salient affected the relationship between authoritarianism and prejudice towards Aboriginal Australians quite substantially (e.g. r =.89 in the gender condition and r =.01 in the national identity condition; see also Haslam & Wilson, 2000).
The social identity perspective follows in the intergroup tradition of Sherif (1967) and suggests that personality tends to become irrelevant to prejudice where social identity or group membership is salient. This is not to say that group-based categorizations of oneself and others will inevitably produce ethnocentrism, discrimination or out-group denigration (e.g. Reynolds, Turner, & Haslam, 2000; Turner & Reynolds, 2001). The degree to which social identification predicts discrimination and intergroup conflict is influenced by broader shared understandings of intergroup relations. For example, whether there is cooperation or conflict between groups depends on the perceived legitimacy, stability and permeability of intergroup relations within the social system (Tajfel & Turner, 1979; Turner, 2005).
It has been difficult to reconcile theory and research concerning the role of group membership in explaining prejudice and discrimination with the classic personality account. Some researchers have suggested that variables, such as authoritarianism and SDO, may affect processes of in-group identification itself (Perreault & Bourhis, 1999; Sidanius, Pratto, & Mitchell, 1994). In other words, an individual who has a certain personality or cognitive style may be predisposed to be effected by group factors in particular ways such that they identify more strongly with relevant in-groups. Along these lines, authoritarians are believed to be people who have an intolerance of ambiguity and perceive the world in ‘black’ and ‘white’, ‘right’ and ‘wrong’ terms (Brown, 1965). People with such personalities may be perceptually ready to define themselves and others as members of in- and out-groups and differentiate strongly between their own and other social groups (e.g. Duckitt, 1989). Likewise, people who are ethnocentric and believe their groups are superior (aspects of SDO) may be more likely to identify with the in-group and differentiate it more strongly from relevant out-groups.
To date, this possible relationship between personality and in-group identification has been examined using the MGP (minimal group paradigm) because the groups are ‘relatively meaningless’ and lack any ‘concrete reality’ outside the experimental setting (Sidanius et al., 1994, p. 153). In addition, the MGP provides an environment where if such orientations are going to play a role in in-group identification and/or patterns of discrimination, they will be readily observable because other factors are controlled.
Sidanius et al. (1994) measured participants' gender, self-esteem and SDO, and then, supposedly based on their responses to a dot estimation task, classified them as being underestimators or overestimators. The in-group identification measure, comprising one item that assessed the degree to which participants agreed with their classification, was then completed along with a number of measures assessing the perceived relationship between the groups. The main discrimination measure was the differential evaluation of underestimators and overestimators on two positive and two negative traits. For this variable, in-group identification was a significant predictor of discrimination (over gender and self-esteem) as was the interaction between in-group identification and SDO. Those who agreed more with their classification to the minimal group and had high SDO expressed more in-group favouritism. Thus, identification was moderated by participants' endorsement of SDO.1
Along these same lines, Perreault and Bourhis (1999) designed an experiment to explore the role of personality variables in predicting in-group identification and associated discrimination between minimal groups. Personality measures, such as authoritarianism (based on a French Canadian abbreviated version of Adorno et al., 1950, F-scale), ethnocentrism and personal need for structure (PNS), were included as possible individual orientations that influence how people construe their membership of minimal groups and associated discrimination against the out-group. An additional feature of the research was that participants were ascribed to minimal groups (Group K or Group W) or could choose their group membership.
Perreault and Bourhis (1999) found that in-group identification was the only variable significantly correlated with discriminatory behaviour. In addition, while none of the personality variables explained on their own a significant amount of the variance in discriminatory behaviour, ethnocentrism predicted discrimination through in-group identification. Given evidence of a strong correlation between ethnocentrism and authoritarianism, Perreault and Bourhis also argued that if a different measure of authoritarianism had been used, a relationship with discrimination through in-group identification may have emerged.
In summary, Sidanius et al. (1994) demonstrated that in-group identification and the interaction between SDO and in-group identification were important while Perreault and Bourhis (1999) found evidence that ethnocentrism was related to discrimination only insofar as individuals did identify with their minimal in-group. They suggested that if a different measure had been used, the same pattern may have been evident for authoritarianism. There is clearly more work to be done. In particular there is a need to assess systematically whether personality variables do explain why some people identify more strongly with the in-group than others and are therefore more prejudiced.
We conducted two experiments to investigate the independent and interactive relationships between authoritarianism, other personality variables and in-group identification in explaining prejudice. The first aim was to address some of the methodological issues that previous researchers themselves identified in order to ascertain whether personality, in-group identification, or in-group identification in interaction with personality dimensions, would be predictive of discrimination. For this reason, we included more contemporary measures of the personality variables of interest (e.g. authoritarianism), a conventional measure of in-group identification (e.g. Doosje, Ellemers, & Spears, 1995) and an assessment of discrimination through use of the Tajfel matrices rather than evaluative ratings alone (Sidanius et al., 1994).
The second aim was to explore the discontinuity hypothesis. On the basis of self-categorization theory it would be expected that to the degree that personality is related to patterns of discrimination, it should become less so the more one's social rather than personal identity becomes contextually meaningful. For this reason, we adapted Perreault and Bourhis' (1999) design to include a condition where like the original minimal group studies people were allocated to groups (supposedly) on the basis of a meaningful dimension of comparison (e.g. preferences for Klee or Kandinsky, performance on a dot estimation task). In this criterion condition, social categorization is meaningfully correlated with attitudinal similarity. As such, identification should be stronger and behaviour towards the out-group should be explained more by one's identification with the in-group rather than particular personality dimensions (Bourhis, Turner, & Gagnon, 1997; Turner & Bourhis, 1996).
By way of summary, in line with the above arguments, it is possible to identify three main hypotheses. In line with previous work on the MGP, in-group identification should be the most important predictor of discrimination between minimal groups (Hypothesis 1 – H1). Following Sidanius et al. (1994) and Perreault and Bourhis (1999), to the degree that personality variables impact on the pattern of discrimination between minimal groups they should do so through or in interaction with in-group identification (Hypothesis 2 – H2). In relation to the discontinuity hypothesis, if there is a direct relationship between certain personality variables and discrimination between minimal groups, these should be attenuated the more one's social identity becomes salient. As such, while personality variables may predict discrimination in the random and voluntary conditions, in the criterion condition in-group identification should be the strongest predictor and personality less so (Hypothesis 3 – H3).
The rationale and predictions for this experiment were similar to Experiment 1. However, the order in which the various measures were completed was changed so that the impact of SDO on in-group identification and discrimination could be assessed. The individual difference measures, including SDO, were completed 3 months prior to other aspects of the experiment.
Through the inclusion of an additional ‘power’ condition, we also used this experiment to investigate in an exploratory way the role of SDO in processes of discrimination. More specifically, the aim was to investigate: (a) one of the original predictions of SDT that there should be a universal human predisposition to prefer hierarchical over egalitarian and more cooperative systems (Sidanius & Pratto, 1999, p. 38) and (b) whether those with higher levels of SDO are more likely to endorse hierarchical systems irrespective of their own group holding a dominant or subordinate position in that system (Sidanius, 1993). Most of the work in this area has examined people's attitudinal preferences with respect to an existing hierarchy. We are not aware of any other study that has looked at the relationship between SDO and actual choices in favour of the establishment of a hierarchical system.
Participants and design
In total, 267 first-year psychology students participated in the study as part of their course requirements during scheduled laboratory sessions. The study had four conditions: random, voluntary, criterion and power, which were manipulated between laboratory sessions. Only responses from participants who completed all scales correctly (six participants' responses were incomplete) and could be matched up with their Phase 1 responses were included in the analyses. Based on these criteria, the sample included 222 participants (65 males, 157 females).
Procedure and materials
Three months prior to the main body of the study, participants completed a questionnaire similar to Experiment 1 except that, instead of the Ethnocentrism scale, a 16-item SDO scale (α=.81; SDO6 scale from Pratto, 1999, p. 206) was included. In addition, the ‘dummy’ engagement variable included three additional items (α=.63) that focused more on personal attributes (i.e. ‘I like to get out of bed early and get on with things’, ‘When I do something I like to do it properly’, ‘I am enthusiastic about most things that I do’) and the identification scale was varied (see below). The procedure in the random, voluntary and criterion conditions was the same as in Experiment 1. In the power condition, participants were allocated to groups on the basis of supposed attitudinal similarity as in the criterion condition. After receiving the instructions for distributing resources using the Tajfel matrices, in this condition, participants received further instructions which emphasized that decisions in the matrix task could be made in more or less fair or unfair ways. The participants in this condition were told that the experimenter was interested in their decisions and their pattern of distributions would be used to determine how the rest of the time in the laboratory would be organized.
The participants were informed that, later in the class, there would be some additional tasks to complete, some of which were boring, some of which were more fun and involved supervising other class members (Platow, Reid, & Andrew, 1998). Along the lines of the classic prisoner's dilemma (Scodel & Minas, 1960), they were told that if both groups (X and Y) were fair in the decision-making task, then both groups could decide how to allocate the additional tasks; whereas if both, or just one, of the groups took an unfair approach, then one group (chosen through the toss of a coin) would be given the power to decide how to allocate the tasks. The coin toss to decide which group would dominate the hierarchy was designed to unconfound the actual desire for hierarchy from the desire for one's own group to be advantaged by being in the dominant position.3
Participants in the power condition then completed manipulation checks to ensure that they understood the instructions. Following the instructions for the matrices, participants completed an in-group identification measure which incorporated four rather than five items (α=.78; ‘I identify with other members of Group ___’, ‘I see myself as a member of Group ___’, ‘I feel strong ties with other members of Group ___’, ‘I am glad to be a member of Group ___’). The participants then completed the same matrices as used in Experiment 1. In addition, in this experiment, participants were asked how much they would like to cooperate with members from the other group and how much they would like to compete with members of the other group (on a 7-point scale from 1, not at all, to 7, very much). Other manipulation checks related to their group assignment also were completed: ‘To what extent did you think other members of your group answered the initial social beliefs scale in the same way as you?’; ‘To what extent did you think that your group membership was a matter of chance?’; ‘How much control did you feel you had in determining your group membership?’ and ‘To what extent do you feel that your responses on the decision-making task have an impact on the rest of the lab?’ (on a 7-point scale from 1, not at all, to 7, very much). Following the completion of all items participants were debriefed extensively.
In order to investigate whether the different group assignment methods had the desired effect, separate one-way ANOVAs were performed on the manipulation checks. Results confirmed a significant effect for group assignment on perception of chance, F(3, 217)=5.78, p <.01, η2=.07, with simple contrasts revealing that those participants in the random condition did indeed feel that their assignment was significantly more a matter of chance (M =5.68) than those in the other conditions (M values: V =4.83, t(217)=2.25, p <.05; C =4.38, t(217)=3.96, p <.001; Power (P)=4.77, t(217)=2.77, p <.01). Similarly, there was a significant effect for group assignment on perceptions of control, F(3, 217)=45.97, p <.001, η2=.39, with those participants in the voluntary condition reporting that their assignment was significantly more under their control (M =5.33) than those in the other conditions (M values: R =1.35, C =3.16, P =2.80). All the mean values here were significantly different from each other except for the comparison between the criterion and power conditions. However, there was no significant effect for group assignment on perceptions of similarity in initial social beliefs F(3, 217)<1, ns. This finding may indicate that towards the end of the study (where the manipulation checks were completed), participants in all conditions imbued their category membership with social meaning. In addition, there was a significant effect for group assignment on the belief that responses would impact on later tasks, F(3, 216)=3.45, p <.05, η2=.05, with those participants in the power condition (M =4.05) indicating greater perceived impact than those participants in the random (M =3.11, t(216)=3.08, p <.01) or voluntary (M =3.31, t(216)=2.04, p <.05) conditions. Although the criterion mean (M =3.64) was lower than that obtained in the power condition the difference was not statistically significant.
With respect to the degree to which participants formed in-group–out-group categorizations, the in- and out-group homogeneity measures revealed significant differences between responses across conditions (in-group: F(3, 217)=11.75, p <.001, η2=.14; out-group: F(3, 217)=10.74, p <.001, η2=.13). There was greater in-group homogeneity in the criterion (M =4.38) compared to the random (t(217)=4.18, p <.001) and voluntary (t(217)=4.33, p <.001) conditions and between the power condition and the random (t(217)=4.03, p <.001) and voluntary (t(217)=4.21, p <.001) conditions. The same pattern was observed with the out-group homogeneity variable (M values: R =3.32, V =3.25, C =4.20, P =4.39). Again, participants perceived the out-group to be more homogeneous in the criterion condition when compared with both the random (t(217)=3.75, p <.001) and the voluntary (t(217)=3.35, p <.001) conditions and in the power condition when compared with both the random (t(217)=4.58, p <.001) and the voluntary (t(217)=4.04, p <.001) conditions. Similarly, there was a main effect for the group assignment variable on the perceived difference between the groups (M values: R =3.26, V =2.69, C =4.09, P =4.00; F(3, 217)=7.86, p <.001, η2=.10). Participants in the criterion condition judged the groups to be more different than those in the random (M =3.26, t(217)=2.93, p <.01) or voluntary (M =2.69, t(217)=4.09, p <.01) conditions. In the power (M =4.00) condition, there also was greater perceived difference than in the random or voluntary condition (t(217)=2.61, p <.01 and t(217)=3.83, p <.01, respectively).
A one-way ANOVA also revealed a significant effect for group assignment on the degree of in-group identification, F(3, 215)=5.27, p <.01, η2=.07. As expected, identification was stronger where allocation to groups was based on attitudinal similarity. Simple contrasts revealed that identification was significantly higher in the criterion (M =3.14) and power (M =3.22) conditions compared with the random condition (M =2.81; t(215)=2.96, p <.01 and t(215)=3.68, p <.01, respectively). Although there were lower levels of identification in the voluntary condition (M =3.02) compared with the criterion and the power conditions, the differences were not statistically significant.
Regression analysis was used to assess the impact of individual difference variables in predicting levels of in-group identification. The responses on the SDO, authoritarianism, PNS and the ‘dummy’ variable (centred) were entered as predictors, and in-group identification (centred) was the dependent variable. Neither the model (F(4, 205)=1.72, p =.15) nor the contribution of authoritarianism (β=−0.04), SDO (β=−0.02) or engagement (β=0.11) was significant. There was a non-significant tendency for personal need for structure to predict in-group identification (β=0.14, p =.052).
As in Experiment 1, scores were calculated for each of the Tajfel matrices using the standard procedure. To assess whether there was discrimination between groups, each pull score for FAV on P, MD on MIP + MJP and FAV on MJP was compared to 0 (a score that reflects parity) in each condition separately. The analysis revealed that there was discrimination on the matrices tasks. In addition, an analysis of a composite measure of discrimination based on an average of these three discrimination measures revealed that the overall grand mean was significantly different from 0 (t(205)=8.17, p <.001).
In order to investigate the effect of group assignment on distribution strategy, six separate one-way ANOVAs were conducted with the matrix scores as dependent measures (Table 3). Results show that the method used for group assignment had no significant effect on whether participants used a joint profit strategy as opposed to maximizing difference (MIP + MJP on MD) or displaying in-group favouritism (MJP on FAV), F values <1.5, ns.
Experiment 2: mean scores for in-group identification, distribution strategies, relative desire for co-operation and personality scales as a function of group assignment
Significant effects did emerge, however, for the use of parity as opposed to in-group favouritism (P on FAV, F(3, 213)=10.08, p <.001, η2=.12) and for each of the discrimination distribution strategies (FAV on P, F(3, 213)=5.16, p <.01, η2=.08; MD on MIP + MJP, F(3, 207)=3.39, p <.05, η2=.05; FAV on MJP; F(3, 213)=3.99, p <.01, η2=.06). From Table 3, it appears that there is more discrimination in the criterion compared with the random and voluntary conditions but that levels of discrimination are significantly lower in the power condition.
Along these lines, for P on FAV, where a lower score reveals greater in-group favouritism (relative to parity) in the distribution of points between the in- and out-groups, simple contrasts revealed that there was more favouritism in the criterion (M =4.39) compared with the random condition (M =6.68; t(213)=2.59, p <.001). There was also significantly less in-group favouritism in the power condition (M =9.19) compared with the random, voluntary (M =5.00) and criterion conditions (t(213)=2.92, p <.001, t(213)=3.97, p <.001 and t(213)=5.15, p <.001, respectively). For FAV on P, where a higher score is evidence of greater in-group favouritism (relative to parity), there was more in-group favouritism in the criterion condition (M =4.28) compared with the random (M =1.61; (t(213)=3.09, p <.01) and power (M =1.17; t(213)=3.36, p <.01) conditions. There also was more in-group favouritism in the voluntary condition (M =3.45) compared with the power condition (t(217)=2.21, p <.05). For MD on MIP + MJP, there was greater differentiation between the groups (relative to maximum in-group and joint profit) in the criterion condition (M =3.55) than the power condition (M =0.89; t(207)=3.06, p <.01). For FAV on MJP, there was significantly less in-group favouritism (relative to maximum joint profit) in the power condition (M =0.89) compared with the other three conditions (M values: R =2.49, t(213)=1.96, p =.05), V =2.92, t(213)=2.02, p <.05; C =3.85, t(213)=3.41, p <.01). In summary, these results reflect higher levels of favouritism and discrimination in the criterion condition with levels dropping off significantly in the power condition.
With respect to participants' willingness to cooperate and compete with the members of the other group, a relative cooperation score was computed by subtracting desire to compete from desire to cooperate. On the resulting measure, a higher score reflects a greater willingness to cooperate. There was a significant main effect for group assignment (F(3, 217)=8.27, p <.001, η2=.10). Simple contrasts revealed that the level of cooperation expressed in the power condition (M =2.14) was significantly different from each of the other conditions (M values: R =0.23, t(217)=4.82, p <.001; V =0.53, t(217)=3.39, p <.01; C =0.95, t(217)=2.84, p <.01). Strong endorsement of cooperation over competition in the power condition is evidence of participants wanting to prevent the introduction of a hierarchical power structure between groups.
Individual difference variables, in-group identification and discrimination
A one-way ANOVA was performed on each of the personality variables (e.g. authoritarianism, SDO) and the measure of engagement. The results revealed no significant effect for group assignment on any of these variables, all F values <1.
Correlations were calculated to examine the relationship between the individual difference variables, the degree of in-group identification and discriminatory behaviour (the composite measure formed by averaging FAV on P, MD on MIP + MJP and FAV on MJP). As can be seen from Table 4, overall, none of the individual difference variables were significantly correlated with discrimination. In fact, the only significant predictor of discriminatory behaviour was the degree of in-group identification (r =.20, p <.01). However, the pattern of correlations varied quite substantially across the different group assignment conditions, hence, correlations were also calculated as a function of each of the experimental conditions (see Table 4). As in Experiment 1, the correlation between in-group identification and discrimination was the strongest in the criterion condition (r =.42, p <.01). There was also a significant correlation between authoritarianism and discrimination in the random condition (r =.24, p <.05).
Experiment 2: correlations with discrimination overall and as a function of group assignment
Perhaps the findings most of interest given predictions based on SDT are that while none of the individual difference variables were significant in the criterion condition, when a potential power hierarchy was introduced there was a significant correlation between SDO and discrimination (r =.27, p <.05) and a marginally significant correlation between authoritarianism and discrimination (r =.26, p <.10). Given that a similar pattern was observed with both SDO and authoritarianism, these findings were investigated further.
One possible explanation is that in the power condition where there was a very clear shift towards fairness or parity, those participants who scored higher on these scales did not move as far in this general direction. In line with this interpretation, Table 5 shows levels of discrimination as a function of group assignment and whether participants scored below (low) or above (high) the median on the SDO scale and the authoritarianism scale, respectively. As the results described above in Table 3 show, there was less discrimination in the power condition which is reflected in a significant main effect for condition for both SDO (F(3, 195)=5.90, p <.001, η2=.08) and authoritarianism (F(3, 194)=6.15, p <.01, η2=.09). While the interactions were not significant for either variable (F values <1.5), a comparison between mean values in each of the conditions separately does help explain the pattern of correlations. It is only in the power condition that there were significant differences in the levels of discrimination amongst those participants who scored low (M =0.11) and high (M =1.47) for SDO (t(53)=2.01, p <.05) and low (M =0.20) and high (M =1.53) for authoritarianism (t(53)=2.01, p =.05). In the power condition, both low and high scores are adopting a fairer, less discriminatory strategy in their allocation of points between the groups, but within this general shift those participants with higher scores are being relatively more discriminatory.
Experiment 2: discrimination as a function of SDO, authoritarianism and group assignment
A regression analysis was conducted to examine the independent relationships between personality variables, and the interaction between these variables and in-group identification, and discrimination. For comparison with Experiment 1 and previous work, an initial analysis included only responses from the random, voluntary and criterion conditions. At Step 1, the personality variables and in-group identification were entered and at Step 2, the interactions were entered. Both models were significant (F(5, 138)=3.37, p <.01 and F(9, 134)=2.37, p <.05) with in-group identification emerging as the only significant predictor of discrimination (β=0.31, p <.001 at both the steps). With responses from all the four conditions, both models were significant. While there was a non-significant tendency (β=0.14, p <.10) for the interaction between personal need for structure and in-group identification to predict discrimination, in-group identification emerged as the strongest and only significant predictor (Step 1: β=0.21, p <.01; Step 2: β=0.23, p <.01).
As in Experiment 1, the overwhelming finding is that in-group identification, rather than personality variables alone or in interaction with in-group identification, is the best predictor of discrimination between minimal groups (H1). Regression analyses revealed that: (a) authoritarianism, SDO, PNS and our ‘dummy’ variable were not significant predictors of in-group identification and (b) these individual difference variables did not directly, or in interaction with in-group identification, significantly predict discrimination between minimal groups (H2). The findings suggest that in-group identification is a central factor in understanding discrimination between minimal groups and personality variables are not that informative in explaining such processes.
As in Experiment 1, it was difficult to assess the hypothesis that relationships between personality and prejudice would be attenuated under conditions where in-group–out-group categorizations became more salient (H3). Strong correlations between personality variables and discrimination did not emerge in the random or voluntary conditions, making it difficult to contrast the pattern of responses across conditions. Having said this, as predicted, the in-group–out-group categorization, assessed through judgements of group homogeneity and in-group identification, was stronger in conditions where people were allocated to groups on the basis of attitudinal similarity (in the criterion and power conditions).
In this experiment, an additional condition was included to assess predictions associated with SDO, more directly. In particular, the study was designed to assess: (a) the existence of a general preference for dominant–subordinate hierarchical relations between groups and (b) whether those high in SDO were more likely to endorse such a hierarchy. The results revealed little evidence of a general preference or drive for hierarchical relations between groups. In a context where discrimination on the matrices task would lead to the introduction of dominant–subordinate hierarchy, there was a strong shift towards parity and fairness in allocations and an increased desire for cooperative rather than competitive relations between the groups. In the power condition, despite strong levels of identification, discrimination decreased as participants acted to eliminate the possibility of the establishment of hierarchical relations between groups. In a context where it was uncertain whether one's own or the other group would be in the dominant position, participants chose egalitarianism over the risk of subordination.
This general shift towards fairness or parity was evident for all participants, but it was also the case that in the power condition a significant correlation between SDO and discrimination emerged. Those participants who scored higher in SDO did not move as far towards parity or equality in their distribution of points. Importantly, such action was taken when those participants were aware that discriminatory behaviour would lead to the establishment of a dominant–subordinate hierarchy where their own group would not necessarily be in the dominant position. These results do provide some support for SDT. It is also the case that this pattern of results was only evident in the power condition where participants were informed that their choices were relevant to the introduction or not, of hierarchical relations between groups.
In this paper, two studies have been reported which aimed to investigate the direct and interactive roles of personality variables and in-group identification in explaining discrimination between minimal groups. One of the central implications of the minimal group studies (Tajfel et al., 1971; Turner, 1980) was the idea that social identity was a central psychological factor in whether people discriminated or not. These findings, and subsequent theory and research associated with them, focused attention on group processes and intergroup relations rather than personality factors alone in understanding prejudice.
Building on this work, some researchers proposed an alternative route through which personality variables may impact on patterns of discrimination. They suggested that certain personality variables or cognitive styles may predispose people to be more or less affected by group factors (Perreault & Bourhis, 1999; Sidanius et al., 1994). For example, because people high in authoritarianism are more likely to form discrete cognitive boundaries, it is argued that they may be more ready to define themselves and others as members of contrasting groups. In this way, personality variables may explain discrimination through or in interaction with processes of in-group identification.
Based on these arguments, two minimal group studies were conducted to investigate whether it was in-group identification, or personality variables directly, or in interaction with in-group identification, which explained patterns of discrimination. Broadly following Perreault and Bourhis (1999), participants completed a range of personality measures (authoritarianism, SDO, ethnocentrism, PNS, a ‘dummy’ engagement variable); they then were allocated to minimal groups (Group X or Group Y) either on an explicitly random basis by their own choice preference or (supposedly) on the basis of attitudinal similarity. The main dependent variable was the way in which points that represented money were allocated to members of the in- and out-group (Tajfel et al., 1971).
In the second experiment, the minimal group paradigm was adapted to include a condition where the role of SDO and central predictions of SDT could be examined (Sidanius & Pratto, 1999). In this condition, participants were informed that their behaviour on the matrices could affect how subsequent positive and negative tasks would be allocated between the groups. If their own and the other group behaved fairly in their point allocations, then both groups could decide how the tasks should be assigned; but if one or both groups behaved unfairly, then one group would be given the power to decide (determined by the toss of a coin). This instruction was designed to separate the desire for hierarchical relations from the desire for one's own group to be in a dominant position.
In both the experiments, the findings are consistent and clear. There was no evidence that traditional personality variables significantly explained the degree to which a person may or may not identify with their group. A ‘dummy’ personality variable, however, that was included to provide a comparison with the more traditional variables (i.e. authoritarianism, ethnocentrism, PNS), was a significant predictor of in-group identification in Experiment 1. This variable assessed engagement with psychology. The finding suggests that those participants who were more interested in psychology (and therefore perhaps paying most attention) identified more strongly with their in-group. With respect to explaining discriminatory behaviour, in both the experiments, personality variables did not directly or in interaction with in-group identification, significantly predict patterns of discrimination. On the basis of these findings, there is no evidence that in-group identification reflects a particular personality or cognitive style or that individual differences are of central concern in explaining discrimination in the minimal group paradigm.
In the power condition, introduced in Experiment 2, the minimal group paradigm was adapted to investigate the central aspects of SDT. Here, we found that there was a general shift away from discrimination indicating participants' preference for egalitarian rather than hierarchical relations between groups. With respect to this movement towards parity, it also was the case that compared with those who had a lower SDO score, those higher in SDO tended not to shift as far resulting in a significant correlation between SDO and discrimination. In line with SDT, those higher in SDO do appear to support more hierarchical group relations and this was the case even though those participants were informed explicitly that their own group would not necessarily be in the dominant position. This finding is important because it helps clarify that SDO does assess support for group-based dominance and is not necessarily tied to one's own group's self-interest (e.g. being in an advantaged or dominant position, see e.g. Sidanius, 1993).
An interesting aspect of findings in Experiment 2 is that there was no general relationship between SDO and discrimination. The relationship was evident only in the power condition where a comparison between egalitarian and hierarchical social structures was introduced. For example, it was only in this condition that those above and below the median in SDO and authoritarianism revealed significant differences in the levels of discrimination. Pratto and Shih (2000) also found a relationship between SDO and implicit group prejudice only under restricted conditions, in their case where there was intergroup threat.
These findings while not supporting the portrayal of SDO as a ‘ubiquitous motive’ or a ‘general desire for group-based dominance’ (Sidanius et al., 2004, p. 848) suggest that, under certain conditions SDO, and perhaps other related variables, are predictive of discrimination. Evidence that SDO and authoritarianism shape behaviour in some situations but not in others, makes it difficult to characterize these factors as personality dimensions in the traditional sense. There is no evidence of a general psychological orientation that shapes social and political attitudes. At the very least, such factors do not seem to predispose individuals in any way to being susceptible to group factors, such as increased levels of in-group identification (e.g. Perreault & Bourhis, 1999; Sidanius et al., 1994).
Following on from this point, and in line with recent work, SDO and authoritarianism may be better understood as representing a set of ideological beliefs that do not affect individual and group behaviour in a generic way, but rather do so most when dimensions related to such ideological beliefs (e.g. egalitarianism) become contextually meaningful or relevant for the perceiver (Guimond, Dambrun, Michinov, & Duarte, 2003; Lehmiller & Schmitt, in press; Reynolds et al., 2001; Reynolds & Turner, 2006; Schmitt et al., 2003; Sidanius et al., 2004; Turner & Reynolds, 2003). Certain social contextual factors (e.g. awareness or salience of group-based hierarchies) can increase the relevance of certain ‘personalities’ which then play a determining role in shaping behaviour. This research suggests that the path forward is to continue to move away from a dichotomy between the personality and the intergroup traditions, and to further investigate the interdependencies between group and personality processes.