Journal of Experimental 48 (2012) 543–549

Contents lists available at SciVerse ScienceDirect

Journal of Experimental Social Psychology

journal homepage: www.elsevier.com/locate/jesp

Reports in the mind: The predictive power of social dominance orientation across social contexts and domains

Nour Kteily ⁎, Arnold K. Ho, Jim Sidanius

Harvard University, USA article info abstract

Article history: The question of whether social dominance orientation represents a generalized orientation towards group-based hi- Received 27 May 2011 erarchies continues to arouse heated debate. Some researchers maintain that rather than indexing support for hier- Revised 3 November 2011 archy across a variety of situations and social contexts, social dominance orientation scores simply reflect individuals' Available online 12 November 2011 attitudes towards whatever specific context individuals had in mind while completing the scale. We systematically examine the generality of SDO by investigating its pattern of relationships with a very wide range of variables across Keywords: a variety of disparate contexts, exploring inequality both as an ideal and as manifested in specific policies towards Social dominance orientation Individual differences particular groups. We also experimentally test an important question raised by Sibley and Liu (2010) about whether administration of modified instructions to think only of “groups in general” is required to ensure SDO's generality. Evidence that SDO functions as a generalized orientation only when administered with instructions to think of groups in general would be a cause for much concern among the many researchers who have used the unmodified scale to index such an orientation. As expected, our results are clear in suggesting (a) that SDO represents a gener- alized orientation towards group-based hierarchy, and (b) that this property is not dependent on specificinstruc- tions to participants to think only of groups in general. Theoretical and practical implications for the status of SDO are discussed. © 2011 Elsevier Inc. All rights reserved.

Introduction from some researchers (e.g., Schmitt, Branscombe, & Kappen, 2003; Lehmiller & Schmitt, 2007; Kreindler, 2005; Turner & Reynolds, 2003). As one of the most widely used variables in social and personality In particular, these authors have disputed the notion that SDO represents psychology, it is not surprising that there has been much debate about an enduring, generalized preference towards hierarchy and inequality, the nature of social dominance orientation (SDO; Pratto, Sidanius, and instead, propose that it simply reflects attitudes towards whatever Stallworth, & Malle, 1994). SDO has been defined as “a generalized ori- specific group categorizations people have in mind while completing entation towards and desire for unequal and dominant/subordinate re- the scale. This view, appropriately labeled a “contextualist” critique of lations among salient social groups, regardless of whether this implies SDO, takes issue with the idea that individuals “carry around”–across so- ingroup domination or subordination” (Pratto, Sidanius, & Levin, 2006, cial contexts – a generalized attitude or predis p. 282). SDO has been shown to relate – across a variety of samples, position towards group-based hierarchy. countries, and contexts – to a vast array of variables, including, for exam- The contextualist critique of SDO has centered around two related ple, , , support for wars of , generalized prejudice, points. First, these authors argue that SDO behaves differently as a func- career choice, physiological arousal to outgroup faces, discriminatory be- tion of the social context in which it is measured. It has been shown that havior in minimal group experiments, anti-miscegenation attitudes, em- correlations between SDO and outcome measures can in fact differ as a pathy, and activation within brain regions associated with the ability function of salient features of the social context (e.g., Lehmiller & to share and feel concern for other's pain (Altemeyer, 1998; Chiao, Schmitt, 2007; Dru, 2007; Duckitt & Sibley, 2010; but see Cohrs & Mathur,&Harada,2009;Green,Thomsen,Sidanius,Staerklé,& Asbrock, 2009,andHeaven & St. Quintin, 2003, who fail to find evidence Potanina, 2009; Ho et al., in press; McFarland, 2010; McFarland & for moderating effects of context on SDO). For example, Lehmiller and Adelson, 1996; Navarrete et al., 2009; Pratto et al., 1994, 2006; Sidanius Schmitt (2007) primed individuals to think of the increased equality & Pratto, 1999). that would be enjoyed by same-sex couples if they were to receive Despite its widespread application and usefulness in understanding legal recognition, and found that SDO predicted heterosexism. Howev- intergroup relations, SDO has been the subject of substantial criticism er, SDO did not predict heterosexism when people were primed to think about increased equality in a different domain: the implementa- tion of fairer regulations that would improve the odds of religious orga- ⁎ Corresponding author at: Department of Psychology, Harvard University, Cambridge, MA 02138, USA. nizations receiving federal funding. Similarly, Duckitt and Sibley (2010) E-mail address: [email protected] (N. Kteily). observed that SDO predicted prejudice against a fictitious immigrant

0022-1031/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.jesp.2011.11.007 544 N. Kteily et al. / Journal of Experimental Social Psychology 48 (2012) 543–549 group when people were primed to see that group as competitive, but differences. Importantly, the overall measure of SDO (i.e., non- not when it was described as morally deviant. category-specific) predicted a substantial proportion of this between- An implication of this work is that priming participants with par- person variation, indicating that this overall measure of support for ticular contexts can make specific examples of relationships between group-based hierarchy was largely responsible for person-level consis- groups more salient in the minds of participants and increase the ex- tencies in attitudes toward several group-specific measures. In sum, the tent to which these exemplars drive responses to the SDO scale (and, authors concluded that, contrary to the strong version of the contextu- by extension, its pattern of relationships with other variables). Impor- alist critique, SDO was not amerereflection of prior attitudes but rather tantly, however, the fact that features of the salient social context can reflected a generalized, abstracted support for group-based hierarchy moderate the effects of SDO does not invalidate the notion that it (see also Kteily, Sidanius, & Levin, 2011). measures a generalized preference for hierarchy. Indeed, Sidanius An important caveat is that prior to assessing ‘overall-SDO’ using and Pratto (1999) explicitly state that “the groups most likely to be the unmodified SDO6 scale, Sibley and Liu (2010) gave the following the targets of social dominance drives will be those groups which novel instruction: are both the most salient and define the sharpest power differential with- in any given society at any given time (p.61, emphasis in original)”. The following statements examine general/overall opinions about Thus, social dominance theorists expect that, under certain circum- different major social groups in society. Major social groups in- stances, specific, highly accessible exemplars can exert increased in- clude groups based on relatively fixed characteristics. Please try fluence on participants' SDO. However, as long as SDO continues to to answer these questions with regards to major social groups in relate to a wide variety of intergroup attitudes and behaviors, the general, rather than thinking about just one group in particular fact that the strength of its relationship with particular variables can (p. 15, emphasis in original). be stronger or weaker as a function of salient aspects of the context in which it is measured is not inconsistent with its overall generality. These instructions were included in order to “clarify the intended A related criticism of SDO goes further still. According to this view, measurement of SDO as a measure of global or generalized attitudes not only does SDO behave differently as a function of different salient (p.15).” SDO's standard instructions simply ask participants to rate contexts, but its meaning is fully determined by context-specificatti- their level of agreement with the scale items without any further spec- tudes (e.g., Schmitt et al., 2003, studies 1 and 2). These authors argue ifications. As such, the authors were only able to conclude that when that, as people complete the SDO6 scale, they have certain group rela- assessed with the ‘general instructions’,SDOreflects a generalized and ab- tionships in mind, and SDO scores simply reflect those preexisting atti- stracted orientation to group-based hierarchy. This leaves open the im- tudes towards these specific groups. As such, rather than SDO predicting portant question of whether SDO continues to measure a generalized attitudes towards racism, for example, it simply measures them: where orientation in the absence of such an instruction. Indeed, although scale items refer to “some groups”, individuals are said to mentally re- Sibley and Liu (2010) prescribe the addition of this instruction prior to place the word “some” with the word “racial”, leading SDO to be noth- the measurement of SDO, they note that research is needed to compare ing more than a redundant index of racism. the effects of SDO when measured with the modified vs. standard in- Sibley and Liu (2010) investigated whether SDO measures a gen- structions. Given SDO's widespread use in research, evidence that it eralized orientation towards group-based hierarchy. While acknowl- does not function as intended without inclusion of a general instruction edging the insights into SDO generated by the contextualist critique, would be major cause for concern. Moreover, from a practical perspec- they state, tive, empirical evidence that SDO behaves more like a generalized ori- entation when participants are explicitly told to think of groups in Although it is possible to shape the measurement of SDO so that the general would be strong evidence in favor of modifying the standard in- scale assesses little more than a proxy measure of attitudes toward a structions of SDO along those lines. given form of group-based inequality, it is also possible to assess Our study systematically investigates the question of SDO's gener- global individual differences in the tendency to prefer hierarchically ality, contributing to an important ongoing theoretical debate regard- structured, dominance-based, intergroup relations across a range of ing the appropriate interpretation of one of the most widely-used social categories. Just because the interpretation and meaning that re- variables in psychology. Importantly, we assess SDO's generality as spondents give to items can be altered under certain conditions does not traditionally used, and compare it to SDO's generality when assessed mean that the construct it was designed to assess does not exist (p. 9, with Sibley and Liu's (2010) modified instructions. Moreover, where- emphasis added). as Sibley and Liu (2010) investigated the pattern of relationships be- tween SDO and attitudes towards three specific contexts, we examine These authors reasoned that if SDO is little more than a reflection of the generality of SDO as it applies to a wide range of consequential pre-existing attitudes towards specific, highly accessible examples of variables, specifically chosen to span a variety of intergroup contexts. intergroup relations, it should not be predicted by attitudes toward hi- The criterion variables we choose also allow us to explore attitudes erarchy specifictoseveral different social categories. In a stringent test towards inequality both as a principle, and as manifested in specific of the generality of SDO, they modified items of the SDO6 scale to assess policies towards particular groups. attitudes towards group-based hierarchy specific to each of ethnic, gen- der, and age stratification. Thus, for example, the SDO6 item “If certain Hypotheses groups stayed in their place, we would have fewer problems” was reworded to state “If certain ethnic groups stayed in their place, we The research that has found a contextual influence on SDO has ex- would have fewer problems”. The authors measured individuals' SDO plicitly attempted to prime participants to think of one particular con- levels using the original SDO6 scale, and subsequently measured text. For example, Schmitt et al. (2003, study 2) found that racism but ethnic-specific SDO, gender-specific SDO, and age-specific SDO. Critical- not sexism was a unique predictor of SDO when participants were spe- ly, they found that although these three category-specificmeasures cifically made to think about racial groups prior to completing the SDO6 were highly intercorrelated, they each predicted unique variance in scale. Indeed, it is specifically in these unique cases – where the exper- the overall SDO6 score, suggesting that SDO is not merely the reflection imenter has purposely attempted to modify the meaning of SDO – that of attitudes towards inequality in any one context. Moreover, using hi- its effects have sometimes been found to be moderated. When partici- erarchical linear modeling, they found that a substantial proportion pants are not forced to think about any specific context, however, SDO of the total shared variance in the group-specific SDO measures has been observed to relate to a multitude of quite different constructs, reflected between-person variation, indicative of systematic individual all measured in one experimental sitting (e.g., Ho et al., in press; Pratto N. Kteily et al. / Journal of Experimental Social Psychology 48 (2012) 543–549 545 et al., 1994; Sidanius, Levin, & Pratto, 1996a). The contextualist critics of Results SDO have – helpfully – shown that the scale can be made to measure something different than it was intended to. On the other hand, some- Our first analysis of interest investigated whether SDO represents a one who, after being made to smell a particularly strong perfume in a generalized orientation toward group-based hierarchy or, alternatively, department store, cannot properly smell the next, would not go so far asimplereflection of more specific intergroup attitudes. We first con- as to claim that the human nose is not a generalized scent detector. Per- ducted the same test used by Sibley and Liu (2010): that is, collapsing haps the general instructions suggested by Sibley and Liu (2010) could across experimental condition, we entered race-SDO, gender-SDO, and serve the role of the coffee beans sometimes provided at perfumeries: age-SDO into a simultaneous regression to see whether each of these washing out previous ‘scents’ and ‘restoring’ the SDO scale (the meta- three intergroup attitudes based on a specific stratification would predict phorical nose) to ‘normal function’. However, to the extent that partic- unique variance in overall-SDO. Consistent with Hypothesis1a,thethree ipants have not been purposely exposed to a specific prime to the variables together accounted for a significant proportion of the variance contrary (or one's nose to a salesman's assertive spray), we see no com- in overall SDO, R2 =.67, F(3, 337)=226.12, pb.001. Most importantly, pelling reason to expect that the SDO scale will require any re- although the three specific attitudes were highly intercorrelated (aver- calibration (or a whiff of those coffee beans!). As such, we hypothesized age r=.61),each of the three specific attitudes uniquely predicted signif- that – in the absence of any specific prime to think of one particular con- icant variance in overall SDO: brace-SDO =.54, t(340)=10.80, pb.001; text – (H1a) SDO would reflect a generalized orientation towards group- bage-SDO =.23, t(340)=7.03, pb.001; bgender-SDO =.11, t(340)=2.15, based hierarchy, and (H1b) that it would correlate with a very wide p=.03. Thus, similarly to Sibley and Liu (2010),weobservedthatSDO range of intergroup constructs. Moreover, we further expected (H2) could not be accounted for by attitudes based on any particular inter- that SDO measured with the standard instruction would exhibit the group context. In order to further investigate the generality of SDO same pattern of relationships with other variables as SDO measured (H1b), we also examined the pattern of correlations between SDO and a with the general instructions. wide variety of relevant outcome variables. As expected, collapsing

across experimental conditions, we found that the SDO6 scale was signif- icantly related to all 19 variables examined (see Table 1), inconsistent Method with the notion that SDO only captures beliefs about people's views in aspecific intergroup context.

We recruited 363 participants (Mage =32.5, 60% female) on Ama- One important question remains unaddressed, however. Given zon's mTurk platform, a reliable and high-quality platform for the re- that the results presented above collapsed across experimental condi- cruitment of diverse participant samples (Buhrmester, Kwang, & tions, it remains possible that the findings were driven by participants Gosling, 2011). All participants were residents of the U.S., with 77% in the ‘general instruction’ condition. Conversely, however, we of participants White Americans, 7.7% Asian Americans, 6.2% African expected (H2) that, in the absence of a context-specific prime, SDO Americans, 3.5% Latino Americans, 3.2% biracial/mixed race, .6% Arab would exhibit generality regardless of whether participants were Americans, .6% Native Americans, and 1.2% participants indicating asked to “think of groups in general”. To test whether the ‘general in- ‘other’. We followed Sibley and Liu's (2010) method as closely as pos- structions’ made a difference to SDO's predictive validity, we first ex- sible. Thus, we assessed our constructs in the following order: first, amined our finding that overall-SDO was uniquely predicted by each participants were asked to complete the traditional SDO6 scale of the context-specific SDO measures. Specifically, we conducted a (Pratto et al., 1994). Participants randomly received either (a) the manifest-variables group analysis in LISREL 8.8, entering overall- standard instructions typically accompanying the SDO6 scale, or SDO as the endogenous variable, and each of race-SDO, gender-SDO, (b) Sibley and Liu's (2010) ‘general instructions’. Participants complet- and age-SDO as exogenous variables, and assessing this model sepa- ed all 16 items of the SDO6 scale (see Table 1 for all scale means, inter- rately for participants in each of the experimental conditions. Evi- correlations and reliabilities). Subsequently, participants completed dence that overall-SDO is only uniquely predicted by the variety of each of the three group-specificmodifications of the SDO6 scale created the context-specific SDO measures when participants received the by Sibley and Liu (2010):age-specific SDO, gender-specificSDO,and ‘general instructions’ would suggest that SDO requires such modified race-specific SDO. We then assessed a very wide range of relevant vari- instructions in order to exhibit generality, thus shedding doubt on ables against which to test the generality of SDO: we measured hostile previous interpretations of SDO (as traditionally assessed). and benevolent sexism (Glick & Fiske, 1996), pro-war attitudes As can be seen in Fig. 1a, however, each of race-SDO, gender-SDO, (Sidanius & Liu, 1992), support for the persecution of immigrants and age-SDO was a significant contributor to overall SDO in the ‘stan- (POSSE; Altemeyer, 1996;adaptedbyThomsen, Green, & Sidanius, dard instructions’ condition. On the other hand, as can be seen in 2008), support for the death penalty (Sidanius, Mitchell, Haley, & Fig. 1b, only race-SDO and age-SDO were significant contributors in Navarrete, 2006), punitiveness (Sidanius et al., 2006), support for racial the ‘general instructions’ condition, whereas gender-SDO was not. policy, support for affirmative action (Haley & Sidanius, 2006), support Thus, SDO did not seem to exhibit greater generality in the “general in- for welfare, support for hierarchy-enhancing jobs (e.g., FBI agent; structions” condition. Indeed, we had expected to find that the “general Sidanius, Pratto, Sinclair, & Van Laar, 1996b), support for hierarchy- instructions” made little difference to SDO's generality. In order to test attenuating jobs (e.g., civil rights lawyer; Sidanius et al., 1996b), support whether these two models were statistically equivalent, we set param- for the principle of legacy admissions at university, and political conser- eter constraints between all analogous paths in each model (i.e., setting vatism. We also included a scale developed for the purposes of this the Φ, Ψ,andΓ matrices to equality). We subsequently observed no sig- study, intended to measure support for unequal distribution of re- nificant deterioration in model fit, χ2 (10)=13.7, p>.05, suggesting sources in a new institution (see Appendix A for previously unpublished that, overall, these two models were statistically equivalent. To examine measures). This item should provide a particularly good test of whether the possibility that specific paths of interest might significantly differ SDO measures general attitudes toward hierarchy, as respondents between the two models, we also considered more specificconstraint should not have pre-existing attitudes about groups that have not yet analyses. First, we examined model deterioration when only the Γ ma- been formed. Moreover, given that the scale measures attitudes to- trices were constrained to equality. This test yielded no significant dete- wards hierarchical distribution of resources at a fictional institution rioration in model fit, χ2 (3)=2.33, p>.05, suggesting that there was with no real-world implications relevant to the participant, this item no significant difference in the overall pattern of relationships between should be a good measure of support for the principle (or ideal) of hier- each of the group-specific SDO measures and overall-SDO across exper- archy. Finally, participants were asked to indicate demographic infor- imental conditions. Secondly, seeing as gender-SDO predicted overall mation, including their age, gender, ethnicity, and education level. SDO in the ‘standard instructions’ condition but not in the “general 546 N. Kteily et al. / Journal of Experimental Social Psychology 48 (2012) 543–549

Table 1 Descriptive statistics and correlations between variables (N=363; Listwise N=313).

Variable 1 2 3 4 5 6 7 8 9 10

1. Overall SDO – ⁎⁎⁎ 2. Race-SDO .78 – ⁎⁎⁎ ⁎⁎⁎ 3. Gender-SDO .67 .77 – ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 4. Age-SDO .62 .54 .52 – ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 5. Hostile sexism .49 .46 .48 .46 – ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 6. Benevolent sexism .23 .26 .23 .17 .44 – ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 7. Pro-war attitudes .40 .33 .23 .22 .19 .26 – ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 8. .58 .49 .36 .37 .37 .27 .38 – ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 9. Anti-immigrant Posse scale .53 .53 .42 .33 .37 .34 .38 .46 – ⁎⁎⁎ ⁎⁎⁎ ⁎⁎ ⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 10. Death penalty support .30 .22 .15 .14 .19 .23 .38 .44 .29 – M 2.50 2.00 2.75 1.84 3.15 3.62 2.94 2.30 2.06 4.22 SD 1.10 1.13 1.30 1.11 1.49 1.43 1.29 .61 1.33 1.89 α .93 .86 .89 .85 .87 .80 .84 .82 .91 .91 Number of items 16 6 6 6 5 5 6 7 6 4

Variable 11 12 13 14 15 16 17 18 19 20 ⁎⁎⁎ ⁎⁎⁎ ⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ † ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 1. Overall SDO .38 .48 .15 .54 .24 .35 .10 −.29 .36 .38 ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ † ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 2. Race-SDO .30 .44 −.07 .47 .25 .35 .09 −.21 .30 .31 ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎ ⁎⁎⁎ ⁎⁎⁎ 3. Gender-SDO .21 .31 .04 .37 .21 .28 .07 −.12 .25 .21 ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 4. Age-SDO .24 .26 −.03 .34 .27 .24 .03 −.20 .25 .20 ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎ † ⁎⁎⁎ ⁎⁎⁎ 5. Hostile sexism .29 .25 −.01 .35 .23 .27 .14 −.10 .43 .31 ⁎⁎⁎ † ⁎⁎ ⁎⁎ ⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 6. Benevolent sexism .26 .09 −.01 .16 .15 .13 .22 .06 .45 .23 ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 7. Pro-war attitudes .44 .28 −.18 .34 .15 .17 .21 −.16 .47 .51 ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎ ⁎⁎⁎ ⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 8. Symbolic racism .43 .56 −.43 .49 .13 .26 .12 −.33 .44 .48 ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎ ⁎⁎⁎ ⁎ ⁎⁎⁎ ⁎⁎⁎ 9. Anti-immigrant Posse scale .43 .26 .05 .33 .25 .15 .19 −.11 .38 .29 ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎ ⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 10. Death penalty support .50 .25 −.28 .26 .04 .14 .16 −.24 .44 .40 11. Punitiveness – – – – –– –– –– ⁎⁎⁎ 12. Racial policy .22 – – – –– –– –– ⁎⁎⁎ ⁎⁎⁎ 13. Affirmative action opposition −.17 −.54 – – –– –– –– ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ 14. Welfare opposition .38 .54 −.30 ––––––– ⁎⁎ ⁎⁎⁎ ⁎⁎ 15. Legacy admissions support .15 −.02 .19 .16 –– –– –– ⁎ ⁎⁎⁎ ⁎ ⁎⁎⁎ † 16. New hierarchy .14 .37 −.12 .42 .10 ––––– ⁎⁎⁎ † ⁎⁎ † 17. Hierarchy enhancing jobs .33 .10 −.03 .15 .01 .10 –– –– ⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎ ⁎⁎ ⁎⁎⁎ 18. Hierarchy attenuating jobs −.17 −.35 .25 −.38 −.12 −.17 .37 ––– ⁎⁎⁎ ⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎ 19. RWA .58 .16 −.17 .28 .19 .18 .24 −.13 –– ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎ ⁎⁎⁎ ⁎ ⁎⁎⁎ ⁎⁎⁎ 20. Political conservatism .44 .42 −.30 .43 .15 .22 .14 −.26 .53 – M 3.22 3.35 3.36 2.82 3.12 3.52 3.72 4.36 3.80 3.45 SD 1.55 1.34 1.32 1.39 1.57 1.43 1.70 1.63 1.10 1.62 α .82 .82 .84 .78 .91 .80 .89 .89 .84 .87 Number of items 3 6 6 3 5 4 4 5 12 3

⁎⁎⁎ pb.001. ⁎⁎ pb.01. ⁎ pb.05. † pb.10. instructions” condition, we investigated the possibility of a significant We further extended this investigation to the relationships we ob- difference between this path across models. However, constraining served between SDO and the variety of dependent variables we exam- the paths from gender-SDO to overall-SDO in each model to equality ined (see Table 1). Once again, we wanted to assess whether the did not significantly deteriorate model fit, χ2 (1)=1.98, p>.05. As generality we observed in SDO (i.e., its significant pattern of relationships such, we were able to conclude not only that SDO was, on average, pre- with a wide variety of relevant variables) depended on the “general in- dicted by a variety of context-specific intergroup attitudes, but that this structions”.WeusedHayes and Matthes' (2009) MODPROBE macro to pattern of relationships supporting the generalized nature of SDO was assess whether the relationship between SDO and relevant outcome vari- statistically equivalent across both experimental conditions. ables was moderated by experimental condition. If Sibley and Liu's

Fig. 1a. Manifest-variables structural equation model of the relationship between Fig. 1b. Manifest-variables structural equation model of the relationship between Race-SDO, Age-SDO, and Gender-SDO and Overall-SDO in the “standard instructions Race-SDO, Age-SDO, and Gender-SDO and Overall-SDO in the “general instructions condition”. Overall model fit is not included as this model is fully saturated. All param- condition”. Overall model fit is not included as this model is fully saturated. All param- eters represent standardized coefficients, and are statistically significant. eters represent standardized coefficients. Dashed lines indicate nonsignificant paths. N. Kteily et al. / Journal of Experimental Social Psychology 48 (2012) 543–549 547

(2010) intuition that SDO behaves more like a generalized orientation the principle of legacy admissions at universities, support for hierarchy- when the instruction to think about groups in general is provided is accu- enhancing jobs and opposition to hierarchy-attenuating ones, and sup- rate, one would expect SDO to relate strongly to a wider variety of inter- port for inegalitarian distribution of resources in the administration of a group variables than when it is assessed without such instructions. fictional new university campus. We examined nineteen potential interaction effects between SDO and That SDO relates to such an impressively diverse array of variables experimental condition, predicting each of the variables listed in Table 1. makes it difficult to sustain the view that SDO is a mere reflection of atti- A significant interaction between SDO and experimental condition would tudes towards specific group relationships held in mind while complet- indicate that the relationship between SDO and the outcome measure dif- ing the scale. Although Schmitt et al. (2003, study 1) have shown some fered as a function of the instructions provided. As seen in Table 2,across evidence that individuals can think of several different groups while com- all outcome variables assessed, the relationship between SDO and the rel- pleting the SDO scale, we consider it quite unlikely that participants will evant variable was never significantly stronger in the ‘general instructions’ be simultaneously thinking of gender relations, their attitudes toward condition. In fact, in the only case where the relationship between SDO war, the desirability of being an FBI agent, and the desirability of legacy and a relevant outcome variable differed as a function of condition, the re- admission systems at universities as they complete the SDO6 scale, for ex- lationship between SDO and the variable was stronger in the ‘standard in- ample. Rather, we think this evidence is much more consistent with the structions’ condition. Critically, the nonsignificance of the interaction view that the SDO scale taps into an abstracted, generalized, orientation effects between condition and SDO was not due to SDO being unrelated towards hierarchy, representing individual consistency in support for to the constructs examined, given that SDO was significantly related to hierarchy-enhancing attitudes and ideologies across contexts. all nineteen variables we measured: it was simply that this pattern of re- Importantly, this does not preclude the possibility that people ac- lationships was not influenced by whether the ‘general instructions’ or cess certain exemplars when answering some items of the SDO scale. the standard instructions were provided prior to completion of the scale. It is likely that – as with other scales assessing abstract concepts – when individuals consider attitudes towards social hierarchy and intergroup relations, certain highly accessible exemplars come to Discussion mind and influence responses to some degree. Indeed, this is consis- tent with Sidanius and Pratto's (1999) stipulation that especially sa- Our results are clear in suggesting that SDO is best conceptualized as lient groups in society are likely to be the strongest targets of social a general measure of individual differences in the preference for group dominance drives. However, this does not mean that, over and based dominance and inequality. In the context of our study, SDO (a) re- above the influence of particular context-specific attitudes, SDO can- lated to a very wide variety of intergroup attitudes and ideologies, not continue to reflect a general orientation towards hierarchy. (b) could not be predicted by attitudes towards any one specificcon- Indeed, that SDO was uniquely predicted by each of race-SDO, gender- text, and, importantly, (c) maintained these properties regardless of SDO, and age-SDO among American participants– for whom race is likely whether participants were asked to “think of groups in general” before to have been a particularly salient social categorization – is further evi- its completion. Thus, we observed significant correlations between dence in support of this assertion. Thus, although further analyses found SDO and attitudes towards age-based (age-SDO), overall-SDO was significantly more closely related to race-SDO than gender-based discrimination (gender-SDO; hostile sexism; benevolent gender-SDO or age-SDO, each of the latter two variables continued to sexism), race-based discrimination (race-SDO; symbolic racism; affir- make important contributions to overall-SDO, suggesting that even in mative action opposition; opposition to racial policy), attitudes towards the presence of a highly accessible exemplar, SDO continues to exhibit immigrants, right-wing authoritarianism, political conservatism, and the generality suggestive of a generalized orientation. In fact, our test, support for war. Moreover, SDO also related to hierarchy-enhancing which replicated the observations of Sibley and Liu (2010) in another so- constructs seemingly less rooted in experience with specific groups or ciety and using a non-student sample, is a particularly conservative one. intergroup contexts and closer to personal ideals or principles: SDO re- That SDO should be uniquely predicted by more specificattitudesacross lated to general punitiveness, support for the death penalty, support for social stratifications all based on precisely the same item word-stem sug- gests that SDO cannot be reduced to attitudes towards any one group, within any one context. Table 2 Not only did the context-specific SDO measures each uniquely pre- Regression coefficients for the interaction effect between SDO and experimental condi- dict overall SDO, but they were also highly intercorrelated with one an- tion (general vs. standard instructions) predicting each outcome variable. other, and exhibited a pattern of associations with the set of dependent Dependent variable BSEB β variables examined that was similar to that found with overall-SDO (see Table 1). Thus, individuals who were high in SDO in one specificcontext Race-SDO .03 .07 .04 Gender-SDO −.00 .08 −.01 (e.g., age-SDO) also tended to be high in SDO in the other contexts Age-SDO .15 .10 .17 assessed (gender-SDO and race-SDO). This finding is consistent with Hostile sexism .02 .13 .02 the work of Levin (1996; as cited in Sidanius & Pratto, 1999), who Benevolent sexism .03 .13 .03 found that although the absolute level of SDO was influenced by contex- Pro-war attitudes .00 .11 .00 Symbolic racism −.03 .05 −.07 tual primes, individual SDO levels exhibited relative stability across con- ⁎ Anti-immigrant posse scale −.25 .11 −.28 texts (r=.50). This work converges to suggest that individuals have a Death penalty support .07 .18 .06 generalized orientation towards group-based inequality that they Punitiveness support −.04 .14 −.04 ‘carry’ with them from one context to another and informs their atti- Racial policy .05 .12 .06 tudes towards specific instances of inequality.1 Affirmative action opposition −.13 .13 −.14 Welfare opposition .05 .12 .05 Legacy admissions support −.19 .15 −.18 New hierarchy .20 .13 .20 1 Indeed, the high intercorrelation between the three context-specific forms of SDO Hierarchy enhancing jobs .21 .17 .18 suggests that attitudes in each specific context may be informed by a higher-order ab- Hierarchy attenuating jobs .10 .15 .09 stracted orientation towards group-based dominance. We tested this using confirma- RWA −.06 .10 −.09 tory factor analysis in LISREL, examining the extent to which each of race-SDO, Political conservatism .11 .15 .10 gender-SDO, and age-SDO could be predicted by a higher-order factor. Indeed, in both Note. For the purposes of this analysis, the “standard instructions” condition was coded experimental conditions, each of the three context-specific forms of SDO was predicted as ‘0’ and the “general instructions” condition was coded as ‘1’. very strongly by a higher-order factor. Moreover, parameter constraint analyses ⁎ pb.05. showed that this was equally the case in both conditions. 548 N. Kteily et al. / Journal of Experimental Social Psychology 48 (2012) 543–549

Crucially, we found that SDO behaved as a generalized orientation previous findings using the construct would have had to contend with regardless of whether or not it was administered with specific instruc- the claim that SDO's use reflects near-tautology (e.g., Lehmiller & tions to answer with regard to groups in general. Thus, among those par- Schmitt, 2007). Nevertheless, we concur with Sibley and Liu's (2010) ticipants receiving only the standard instruction, overall SDO was still recommendation that care be taken not to systematically prime cogni- uniquely predicted by each of age-SDO, gender-SDO, and race-SDO. In- tions about specific groups prior to administration of the SDO6 scale. deed, although the difference in the models was nonsignificant, gender- As such, it would probably also be wise to heed their advice that SDO6 SDO was not a unique predictor of overall SDO for those in the ‘general scale be administered before measures of other intergroup attitudes. instructions’ condition. Moreover, SDO correlated significantly with all nineteen assessed constructs, and its pattern of relationships differed Conclusion only once between experimental conditions, with the relationship actu- ally stronger in the ‘standard instructions’ condition.2 The remarkable Converging evidence suggests that SDO represents a generalized, similarity between the findings in the two experimental conditions abstracted orientation toward group-based inequality and hierarchy was not for lack of statistical power: we specifically collected data that cannot be reduced to attitudes within any one specific context. from a sample more than large enough to detect statistical significance, While SDO is most certainly sensitive to social context, it also most cer- and ran – without any alpha-correction – over nineteen tests in which tainly relates to the support for hierarchy across social stratifications. such a difference could have emerged. 3 Our findings are important for a number of reasons. Firstly, they Appendix A speak to an important and ongoing debate about the nature of SDO, a variable that has been used very widely across psychology, ranging Previously unpublished scales are shown in their entirety here. Unless from research on generalized prejudice and discrimination (e.g., otherwise noted, all measures used a 1 (strongly disagree/disapprove) McFarland, 2010), to research on workplace aggression (e.g., Parkins, to 7 (strongly agree/favor) scale. Fishbein, & Ritchey, 2006), to work on career development among urban youth (Diemer & Blustein, 2006). The large majority of this re- Support for racial policy search has conceptualized SDO as it was originally developed: as a gen- eralized orientation towards hierarchy and group-based inequality. On 1. Government should see to it that minorities get fair treatment in the other hand, an alternative theoretical position (see Schmitt et al., jobs. 2003; Turner & Reynolds, 2003) has challenged this view, instead argu- 2. Government should not pass laws concerning the hiring of ethnic ing that these investigators have been redundantly assessing pre- minorities. existing attitudes specific to particular contexts participants may have 3. Government should ensure that Whites and minorities go to the held in mind at the time. Our work stringently tests the assumption same school. that SDO represents a generalized orientation and, in its pattern of re- 4. Government has no business trying to ensure racial integration in sults, provides clear evidence in the affirmative. Rather than invalidate schools. the general nature of the SDO scale, the contextualist critique of SDO 5. Government should do what it can to improve the economic con- has shown, importantly, that the scale can be ‘manipulated’ to behave dition of poor ethnic minorities. in ways it was not intended to (e.g., Dru, 2007; Lehmiller & Schmitt, 6. Government has no business trying to improve the economic con- 2007; Schmitt et al., 2003). When participants are explicitly made to dition of poor ethnic minorities. think of one particular social context, SDO correlates more strongly to variables related to that stratification and less strongly to unrelated var- Support for welfare iables. While this implies that care need be taken to avoid inadvertently fi priming participants strongly with speci c contexts before completing 1. Greater assistance to the poor the SDO6 scale, the conclusion that SDO does not typically measure a 2. Reduced public support for the homeless generalized orientation is inconsistent with our data. 3. Reduced benefits for the unemployed Moreover, our systematic comparison of SDO's generality when fi assessed with its typical instructions and modi ed instructions to Support for legacy admissions “think of groups in general” has important implications, both theoret- ical and practical. That SDO equally relates to a wide range of vari- The following statements concern “legacy” college admissions ables whether or not participants are asked to think of groups in (that is, considering family relationships to alumni in college admis- fi general implies that, in the absence of speci c primes forcing individ- sions decisions). Some universities favor applicants who have rela- uals' attention to particular contexts, they independently access an tives that currently attend or previously attended the university, abstracted orientation towards dominance. This suggests that al- and give them an advantage in making admission decisions. Please in- though individuals may sometimes bring to mind particular accessi- dicate how you feel about such admissions policies. ble exemplars when answering SDO scale items, these are not typically apt to “wash out” SDO's generality. 1. I support legacy college admissions. As it is, Sibley and Liu's (2010) practical recommendation that the 2. I oppose legacy college admissions. scale “be administered with instructions that clarify measurement of a 3. The potential benefits of legacy admissions outweigh the potential global or generalized construct (p. 31)” seems not quite as pressing as harm. it might have been. This should be a cause for some relief among the 4. Legacy college admissions are unjust — family connections to the many researchers who have long been using SDO as an index of individ- university should not be considered. ual differences in their research without any such instruction. Had we 5. It is sometimes OK to consider family connections to the university observed that SDO – in the absence of ‘general instructions’–simply in making admission decisions. reflected pre-existing attitudes towards one group, in one context, Support for unequal distribution of resources in a new institution

2 Given the number of tests conducted, this difference may be due to chance. Imagine the University of Massachusetts is building a new campus. It 3 The lack of a correction for multiple tests represents a conservative test in this con- text. Given the sample size and standard probability (.05) value we used to test for sig- has to decide how to fund its various schools (e.g., the law school, the nificance, we had more than sufficient power to detect any interactions. medical school, the engineering school etc.). Specifically, one option N. Kteily et al. / Journal of Experimental Social Psychology 48 (2012) 543–549 549 would be to adopt a model in which each school would be responsible Heaven, P. C. L., & St. Quintin, D. (2003). Personality factors predict racial prejudice. Personality and Individual Differences, 34, 625–634. for its own fundraising and expenses. While such a model may give Ho, A. K., Sidanius, J., Pratto, F., Levin, S., Thomsen, L., Kteily, N. S., & Sheehy-Skeffington, schools more freedom, it is likely that the schools would be unevenly J. (in press). Social dominance orientation: Revisiting the structure and function of funded, resulting in some schools with large operating budgets and a variable predicting social and political attitudes. Personality and Social Psycholo- gy Bulletin. many resources, and other schools with minimal budgets and limited Kreindler, S. (2005). A dual group processes model of individual differences in preju- resources. Another model would entail fundraising at the level of the dice. Personality and Social Psychology Review, 9(2), 90. university, and distributing resources equally between schools. Please Kteily, N. S., Sidanius, J., & Levin, S. (2011). Social dominance orientation: Cause or ‘ ’ indicate which of these two models you would prefer by using the mere effect ?: Evidence for SDO as a causal predictor of prejudice and discrimina- tion against ethnic and racial outgroups. Journal of Experimental Social Psychology, scale below for each of the following statements. 1, 208–214. Lehmiller, J., & Schmitt, M. (2007). Group domination and inequality in context: Evidence 1. I would prefer schools to be responsible for their own funding. for the unstable meanings of social dominance and authoritarianism. European Journal of Social Psychology, 37(4), 704. 2. I would prefer the university to distribute resources equally rather McFarland, S. G. (2010). Authoritarianism, social dominance, and other roots of gener- than have each school fund itself. alized prejudice. , 31(3), 453–477. 3. It would be unfair if schools had unequal budgets. McFarland, S. G., & Adelson, S. (1996). An omnibus study of personality, values and . Paper presented at the 19th annual convention of the International Society 4. It would be fair for each school to get the budget it earns. for Political Psychology, Vancouver, British Columbia, July 1996. Navarrete, C. D., Olsson, A., Ho, A. K., Mendes, W. B., Thomsen, L. T., & Sidanius, J. (2009). Fear extinction to an outgroup face: The role of target gender. Psychological References Science, 20(2), 155–158. Parkins, I. S., Fishbein, H. D., & Ritchey, P. N. (2006). The influence of personality on Altemeyer, B. (1996). The authoritarian specter. Cambridge, MA US: Harvard University workplace bullying and discrimination. Journal of Applied Social Psychology, 36, Press. 2544–2577. Altemeyer, B. (1998). The other “authoritarian personality”. In M. P. Zanna (Ed.), Ad- Pratto, F., Sidanius, J., & Levin, S. (2006). Social dominance theory and the dynamics of vances in experimental social psychology, Vol. 30. (pp. 47–92)San Diego, CA: Aca- intergroup relations: Taking stock and looking forward. European review of social demic Press. psychology, 17(1), 271–320. Buhrmester, M., Kwang, T., & Gosling, S. D. (2011). Amazon's Mechnical Turk: A new source of Pratto, F., Sidanius, J., Stallworth, L., & Malle, B. (1994). Social dominance orientation: A inexpensive, yet high-quality data? Perspectives on Psychological Science, 6,3–5. personality variable predicting social and political attitudes. Journal of Personality Chiao, J. Y., Mathur, V. A., & Harada, T. (2009). Values, empathy, and fairness across social and Social Psychology, 67, 741. barriers: Neural basis of preference for human social hierarcy versus . Schmitt, M., Branscombe, N., & Kappen, D. (2003). Attitudes toward group-based in- New York Academy of Sciences, 1167,174–181, doi:10.1111/j.1749-6632.2009.04508.x. equality: Social dominance or social identity? British Journal of Social Psychology, Cohrs, J. C., & Asbrock, F. (2009). Right-wing authoritarianism, social dominance orientation 42(2), 161–186. and prejudice against threatening and competitive ethnic groups. European Journal of Sibley, C. G., & Liu, J. H. (2010). Social dominance orientation: Testing a global in- Social Psychology, 39,270–289. dividual difference perspective. Political Psychology, 31(2), 175–207. Diemer, M. A., & Blustein, D. L. (2006). Critical consciousness and career development Sidanius, J., Levin, S., & Pratto, F. (1996). Consensual social dominance orientation and among urban youth. Journal of vocational behavior, 68, 220–232. its correlates within the hierarchical structure of American society. International Dru, V. (2007). Authoritarianism, social dominance orientation, and prejudice: Effects Journal of Intercultural Relations, 20, 385–408. of various self-categorization conditions. Journal of Experimental Social Psychology, Sidanius, J., & Liu, J. (1992). Racism, support for the Persian Gulf War, and the police 43, 877–883. beating of Rodney King: A social dominance perspective. Journal of Social Psychol- Duckitt, J., & Sibley, C. G. (2010). Intergroup and individual difference determinants ogy, 132, 685–700. of prejudice: Testing a differential moderation hypothesis. European Journal of Sidanius, J., Mitchell, M., Haley, H., & Navarrete, C. (2006). Support for harsh criminal Personality, 24,583–601. sanctions and criminal justice beliefs: A social dominance perspective. Social Justice Glick, P., & Fiske, S. T. (1996). The Ambivalent Sexism Inventory: Differentiating hostile Research, 19(4), 433–449. and benevolent sexism. Journal of Personality and Social Psychology, 70(3), 491–512. Sidanius, J., & Pratto, F. (1999). Social dominance: An intergroup theory ocbf social hier- Green, E., Thomsen, L., Sidanius, J., Staerklé, C., & Potanina, P. (2009). Reactions to crime archy and oppression. : Cambridge University Press. as a hierarchy regulating strategy: The moderating role of social dominance orien- Sidanius, J., Pratto, F., Sinclair, S., & Van Laar, C. (1996). Mother Teresa meets Genghis tation. Social Justice Research, 22(4), 416–436. Khan: The dialectics of hierarchy-enhancing and hierarchy-attentuating career Haley, H., & Sidanius, J. (2006). The Positive and Negative Framing of Affirmative Ac- choices. Social Justice Research, 9, 145–170. tion: A Group Dominance Perspective. Personality and Social Psychology Bulletin, Thomsen, L., Green, E. G. T., & Sidanius, J. (2008). We will hunt them down: How SDO 32, 656–668. and RWA fuels ethnic persecution of immigrants in fundamentally different ways. Hayes, A. F., & Matthes, J. (2009). Computational procedures for probing interactions in Journal of Experimental Social Psychology, 44, 1455–1464. OLS and logistic regression: SPSS and SAS implementations. Behavior Research Turner, J., & Reynolds, K. (2003). Why social dominance theory has been falsified. British Methods, 41, 924–936. Journal of Social Psychology, 42(2), 199–206.