Read this article . Pay special attention to the methods used by the researchers. What are the strengths and limitations of the methods used?

this article

Social Dominance Orientation: A Personality Variable Predicting Social and Political Attitudes The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters

Citation Pratto, Felicia, James Sidanius, Lisa M. Stallworth, and Bertram F. Malle. 1994. Social dominance orientation: A personality variable predicting social and political attitudes. Journal of Personality and Social Psychology 67, no. 4: 741-763.

Published Version http://dx.doi.org/10.1037/0022-3514.67.4.741

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:3207711

Terms of Use This article was downloaded from Harvard University's DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-ofuse#LAA

Journal of Personality and Social Psychology 1994, Vol. 67, No. 4, 741-763

Copyright 1994 by the American Psychological Association, Inc. 0022-3514/94/S3.00

Social Dominance Orientation: A Personality Variable Predicting Social and Political Attitudes

Felicia Pratto, Jim Sidanius, Lisa M. Stallworth, and Bertram F. Malle

Social dominance orientation (SDO), one's degree of preference for inequality among social groups, is introduced. On the basis of social dominance theory, it is shown that (a) men are more social dominance-oriented than women, (b) high-SDO people seek hierarchy-enhancing professional roles and low-SDO people seek hierarchy-attenuating roles, (c) SDO was related to beliefs in a large number of social and political ideologies that support group-based hierarchy (e.g., meritocracy and racism) and to support for policies that have implications for intergroup relations (e.g., war, civil rights, and social programs), including new policies. SDO was distinguished from interpersonal dominance, conservatism, and authoritarianism. SDO was negatively correlated with empathy, tolerance, communality, and altruism. The ramifications of SDO in social context are discussed.

Group conflict and group-based inequality are pervasive in human existence. Currently, every continent is enduring some form of ethnic conflict, from the verbal debate over multiculturalism in the United States and Canada to civil war in Liberia and Bosnia. Other conflicts between groups are ancient: the European persecution of Jews, "Holy Wars" waged by Christians and Muslims around the Mediterranean, imperialism in South America, and anti-Black racism in northern Africa and elsewhere. Regardless of the intensity of the conflict, the participants justify their behavior to others by appealing to historical injustices, previous territorial boundaries, religious prohibitions, genetic and cultural theories of in-group superiority, or other such ideologies. Prompted by the ubiquitous nature of group-based prejudice and oppression, we developed social dominance theory (see Pratto, in press; Sidanius, 1993; Sidanius & Pratto, 1993a). The theory postulates that societies minimize group conflict by creating consensus on ideologies that promote the superiority of one group over others (see also Sidanius, Pratto, Martin, & Stallworth, 1991). Ideologies that promote or maintain group inequality are the tools that legitimize discrimination. To work smoothly, these ideologies must be widely accepted within a society, appearing as self-apparent truths; hence we call them hierarchy-legitimizing myths.' By contributing to consensual or

normalized group-based inequality, legitimizing myths help to stabilize oppression. That is, they minimize conflict among groups by indicating how individuals and social institutions should allocate things of positive or negative social value, such as jobs, gold, blankets, government appointments, prison terms, and disease. For example, the ideology of anti-Black racism has been instantiated in personal acts of discrimination, but also in institutional discrimination against African-Americans by banks, public transit authorities, schools, churches, marriage laws, and the penal system. Social Darwinism and meritocracy are examples of other ideologies that imply that some people are not as "good" as others and therefore should be allocated less positive social value than others. Thus far, we have given examples of legitimizing myths that enhance or maintain the degree of social inequality. Other ideologies may serve to attenuate the amount of inequality. For example, the "universal rights of man" and the view summarized by "all humans are God's children" are inclusive, egalitarian ideologies that explicitly do not divide persons into categories or groups. To the extent that such ideologies are widely shared, there should be less group inequality. There are, then, two varieties of legitimizing myths: hierarchy-enhancing legitimizing myths, which promote greater degrees of social inequality, and hierarchy-attenuating legitimizing myths, which promote greater social equality.

Felicia Pratto, Lisa M. Stallworth, and Bertram F. Malle, Department of Psychology, Stanford University; Jim Sidanius, Department of Psychology, University of California at Los Angeles. We are grateful to a number of people for their diligence and creativity in this research: Erron Al-Amin, Jill Andrassy, Sahr Conway-Lanz, Nick Clements, Magda Escobar, Jack Glaser, Louis Ibarra, Kent Harber, John Hetts, Amy Lee, Johanna Jensen, John Moore, Jenn Pearson, Holly Schaefer, Margaret Shih, Stacey Sinclair, Gayatri Taneja, Jack Wang, and Wes Williams. Bob Altemeyer, Monisha Pasupathi, Vernon Schabert, Michael Mitchell, Steve Gangestad, Corinne Kosmitzki, Ted Goertzel, and three anonymous reviewers provided useful comments on a draft of this article. Correspondence concerning this article should be addressed to Felicia Pratto, Department of Psychology, Jordan Hall, Stanford University, Stanford, California 94305-2130.

SOCIAL DOMINANCE ORIENTATION

Given our theoretical postulate that acceptance of legitimizing myths has significant influence on the degree of inequality in societies, it is quite important to understand the factors that lead to the acceptance or rejection of ideologies that promote or attenuate inequality. Social dominance theory postulates that a

1 The term myth is meant to imply that everyone in the society perceives these ideologies as explanations for how the world is—not that they are false (or true). Social dominance theory is meant only to describe the social and psychological processes that act on these ideologies, not to ascertain whether these ideologies are true, fair, moral, or reasonable.

741

742 PRATTO, SIDANIUS, STALLWORTH, AND MALLE

significant factor is an individual-difference variable called social dominance orientation (SDO), or the extent to which one desires that one's in-group dominate and be superior to outgroups. We consider SDO to be a general attitudinal orientation toward intergroup relations, reflecting whether one generally prefers such relations to be equal, versus hierarchical, that is, ordered along a superior-inferior dimension. The theory postulates that people who are more social-dominance oriented will tend to favor hierarchy-enhancing ideologies and policies, whereas those lower on SDO will tend to favor hierarchy-attenuating ideologies and policies. SDO is thus the central individual-difference variable that predicts a person's acceptance or rejection of numerous ideologies and policies relevant to group relations. Another way that individuals' levels of SDO may influence their contribution to social equality or inequality is in the kinds of social roles they take on, particularly, roles that either enhance or attenuate inequality. We thus predict that those who are higher on SDO will become members of institutions and choose roles that maintain or increase social inequality, whereas those who are lower on SDO will belong to institutions and choose roles that reduce inequality. The purpose of the present research was to demonstrate that individual variation in SDO exists and to show that this construct behaves according to the theory outlined above. Specifically, our goals were (a) to develop a measure of SDO that is internally and temporally reliable, (b) to show that SDO is related to the attitudinal and social role variables specified by social dominance theory (predictive validity), (c) to show that the measure is not redundant with other attitude predictors and standard personality variables (discriminant validity), and (d) to show that SDO serves as an orientation in shaping new attitudes.

HYPOTHESES

The first set of hypotheses we tested was derived from social dominance theory and concerned those variables to which SDO should strongly relate, termed predictive validity. The second set of hypotheses, termed discriminant validity, states either that SDO should be independent of other variables or that SDO should have predictive value in addition to the effects of these other variables. We also hypothesized that SDO should relate moderately to certain other personality variables, from which SDO is conceptually distinct. The third set of hypotheses we tested concerns SDO's power to predict new social attitudes.

Predictive Validity

Gender

The world over, men and women hold different roles with regard to the maintenance of hierarchy. Ubiquitously, men serve as military leaders and hold leadership roles in religious, social, political, and cultural spheres (e.g., Brown, 1991, pp. 110, 137). Moreover, men hold more hierarchy-enhancing attitudes, such as support for ethnic prejudice, racism, capitalism, and rightwing political parties, than do women (e.g., Avery, 1988; Eisler & Loye, 1983; Ekehammar & Sidanius, 1982; Shapiro & Mahajan, 1986; Sidanius & Ekehammar, 1980; see review by Si

danius, Cling, & Pratto, 1991). On the basis of these general societal patterns, we have predicted and shown that, on average, men are more social dominance-oriented than women (see Pratto, Sidanius, & Stallworth, 1993; Sidanius, Pratto, & Bobo, in press). We tested this hypothesis with the measure of SDO developed in the present research.

Legitimizing Myths

Ethnic Prejudice

One of the major kinds of ideology concerning relative group status is ethnic prejudice. In the United States, the most longstanding and widely disseminated version of ethnic prejudice is anti-Black racism. Therefore, we predicted that SDO would be strongly related to anti-Black racism in the present U.S. samples. In the United States, a theoretical and empirical debate about how best to measure anti-Black racism has been conducted for some time (e.g., see Bobo, 1983; McConahay, 1986; Sears, 1988;Sniderman&Tetlock, 1986a, 1986b). Social dominance theory merely postulates that SDO should predict whatever ideologies are potent within the culture at the time of measurement. From our theoretical viewpoint, it does not matter whether the basis for racism is fairness (e.g., Kluegel & Smith, 1986), genetic or biblical racial inferiority theories, symbolic racism (e.g., Sears, 1988), or family pathology (e.g., Moynihan, 1965). Any potent ideology that describes groups as unequal and has policy implications is a legitimizing myth and should, therefore, correlate with SDO. During the period the present research was conducted, our subjects' country was engaged in a war against Iraq, so we also measured anti-Arab racism and expected it to correlate with SDO.

Nationalism

A more general kind of in-group prejudice that can occur in nation-states is nationalism, chauvinism, or patriotism. Kosterman and Feshbach (1989) suggested that procountry feelings (patriotism) can be distinguished from comparative prejudice, that is, that one's country is better than other countries (nationalism), and as such should dominate other countries (chauvinism). Even so, all three reflect attitudinal bias in favor of the national in-group, and thus we postulated that patriotism, nationalism, and chauvinism would all be significantly related to SDO.

Cultural Elitism

All societies share the idea that one of the defining features of those who belong to their society (are part of the in-group, or are considered by them to be human) is that they are "cultured." In some societies, including English and American society, an elitist ideology built on the cultured-not cultured distinction postulates that the elite class has "culture" not shared by middle- and working-class people and is therefore more deserving of the "finer things in life." We term this legitimizing myth cultural elitism, and we expected it to correlate with SDO as well.

Sexism

We believe that antifemale sexism is a ubiquitous legitimizing myth, although, as with ethnic prejudice, the content basis of

SOCIAL DOMINANCE ORIENTATION 743

sexist ideology varies widely with religion, cultural history, and technology. In the present U.S. samples, we used scales that assess sexism as the extent to which people believe men and women are "naturally" different and should have different work roles outside and inside the home (Benson & Vincent, 1980; Rombough & Ventimiglia, 1981) and the extent to which people believe that women rather than men can be blamed for unwanted sexual advances such as rape and sexual harassment (Burt, 1980). We predicted that all of these would be positively correlated with SDO, even controlling for subject sex.

Political-Economic Conservatism

Political-economic conservatism is associated with support for capitalism versus socialism (e.g., Eysenck, 1971). Given that capitalism implies that some people and businesses should thrive, while those who are less "competitive" should not, we consider political-economic conservatism to be a hierarchy-enhancing legitimizing myth that should positively correlate with SDO (see also Sidanius & Pratto, 1993b). Other policies supported by conservatives, such as that women should stay home with children and that the USSR must be kept in its place, divide people into groups "deserving" different treatment, so we feel conservatism generally can be viewed as a legitimizing myth. In fact, Wilson's extensive work on the body of attitudes that make up conservatism shows that a preference for hierarchical social relationships is one of conservatism's many dimensions (Wilson, 1973, p. 22).

Noblesse Oblige

A hierarchy-attenuating ideology that exists in many cultures is that those with more resources should share them with those who have fewer resources (e.g., the Marxist maxim, "From each according to his [sic] ability, to each according to his need," and the potlatch custom of the Kwakiutl). The English-American version is called noblesse oblige, which we expected to be negatively correlated with SDO.

Meritocracy

Another hierarchy-enhancing ideology is that wealth and other social values are already distributed appropriately, based on the deservingness of the recipients. The Protestant work ethic and just world theory are examples of meritocratic ideologies, so we administered standard measures of belief in the Protestant work ethic and belief in a just world and predicted that they would be positively correlated with SDO. In the United States, attributions for poverty due to laziness or to some other inherent fault in the poor are predicated on the idea that equal opportunity is available to all (Kluegel & Smith, 1986), so we wrote an equal opportunity scale and predicted that it would correlate positively with SDO.

Social Policy Attitudes

According to social dominance theory, individuals who are social dominance oriented will favor social practices that maintain or exacerbate inequality among groups and will oppose social practices that reduce group inequality. The particular social policies that correlate with SDO may vary from society to soci

ety, but we predicted that SDO would relate to support for, or opposition to, the following policies in U.S. samples.

Social Welfare, Civil Rights, and Environmental Policies

We expected SDO to correlate with opposition to social policies that would reduce inequality between U.S. nationals and foreigners or immigrants, rich and middle class or poor, men and women, ethnic groups, heterosexuals and homosexuals, and humans versus other species. As such, we measured our subjects' attitudes toward a variety of government social programs, racial and sexual discrimination laws, gay and lesbian rights, domination of foreigners, and environmental policies. In several samples we also assessed attitudes toward "interracial dating" and "interracial marriage," because miscegenation has been central to the U.S. racial policy debate.

Military Policy

Because the military is a symbol of nationalism and can be one of the chief means of domination of one nation over others, we expected SDO to correlate positively with expressed support for military programs and actions.

Punitive Policies

Despite its stated creed to enact equality before the law, the U.S. criminal justice system shows class and ethnic bias at all levels from arrest to plea bargaining to sentencing (e.g., Bienen, Alan, Denno, Allison, & Mills, 1988; General Accounting Office, 1990; Kleck, 1981; Nickerson, Mayo, & Smith, 1986; Paternoster, 1983; Radelet & Pierce, 1985; Reiman, 1990; Sidanius, 1988). As one example, in a review of 1,804 homicide cases in South Carolina, Paternoster (1983) found that in cases where Blacks killed Whites, rather than other Blacks, prosecutors were 40 times more likely to request the death penalty. For this reason, we expected support for "law and order" or punitive policies, particularly the death penalty, to be positively related to SDO (see also Mitchell, 1993; Sidanius, Liu, Pratto, & Shaw, 1994).

Discriminant Validity

Interpersonal Dominance

SDO, or preference for unequal relationships among categories of people, is conceptually distinguishable from the common personality conception of interpersonal dominance, which concerns the extent to which individuals like to be in charge and are efficacious. For example, people who score high on the California Personality Inventory (CPI) Dominance scale are confident, assertive, dominant, and task oriented, whereas people who score low are unassuming and nonforceful (Gough, 1987, p. 6). People who score high on the Jackson Personality Research Form (JPRF) Dominance scale attempt to control their environments and influence or direct other people; they are forceful, decisive, authoritative, and domineering (Jackson, 1965). We tested this theoretical distinction between social and task or interpersonal dominance by using the CPI and JPRF Dominance subscales in several samples reported here. We predicted that SDO would not correlate with these two measures.

744 PRATTO, SIDANIUS, STALLWORTH, AND MALLE

Authoritarianism

There is clearly some theoretical similarity in the effects of social dominance theory's SDO construct and authoritarian personality theory's authoritarian construct (see Adorno, Frenkel-Brunswik, Levinson, & Sanford, 1950). High-SDO people and authoritarian personalities are theorized to be relatively conservative, racist, ethnocentric, and prejudiced, and they should show little empathy for lower status others. Our conception of SDO, however, differs from classical authoritarianism in several respects. First, classical authoritarian theorists viewed authoritarianism as an aberrant and pathological condition and as a form of ego-defense against feelings of inadequacy and vulnerability (see also Frenkel-Brunswik, 1948, 1949). SDO, however, is not conceived of in clinical terms, as an aberrant personality type, or as a form of ego-defense. Rather, SDO is conceived of as a "normal" human propensity on which people vary. Second, authoritarian personality theory emphasized the sources of authoritarianism as springing from psychodynamic processes. Specifically, Adorno et al. (1950) postulated that strict and harsh parental styles would provoke conflicts between the child and parents that would be "unresolved." As a way of resolving these, the child as an adult would submit to authorities and be intolerant of those who would not. In contrast, we theorize that such a personal history is unnecessary to developing a relatively high SDO tendency. Rather, both temperament and socialization probably influence one's level of SDO. Third and most important, whereas authoritarianism is primarily conceived as a desire for individual dominance resulting from experiences with authority figures, SDO is regarded as the desire that some categories of people dominate others. Because the two constructs are defined differently, measurements of each should not be highly correlated. Given that authoritarianism should predict many of the same variables we postulate SDO should predict, it is important for us to show that SDO has explanatory value in addition to authoritarianism. We tested the "marginal utility" of the SDO construct by testing whether correlations between SDO and support for legitimizing myths and policies are significant after partialing out authoritarianism.

Conservatism

Political-economic conservatism serves as a legitimizing myth in our theory, and thus we expect it to correlate positively with SDO. Conservatism is also a well-known robust predictor of social and political attitudes (e.g., Eysenck & Wilson, 1978; Wilson, 1973). To show that SDO has utility in addition to political-economic conservatism, we tested whether SDO substantially correlated with social attitudes after partialing out conservatism.

Standard Personality Variables

Because we think our concept of SDO is a yet unstudied personality dimension, we expected it to be independent of other standard personality variables such as self-esteem and the BigFive personality dimensions: Extraversion, Agreeableness, Openness, Neuroticism, and Conscientiousness (see Costa & MacRae, 1985; John, 1990, for reviews).

Empathy, Altruism, Communality, and Tolerance

People who are highly empathic with others would seem to be less prejudiced and discriminatory against out-groups. Thus, it is reasonable to expect a general concern for other people to be negatively correlated with SDO. Similarly, any general prosocial orientation might mitigate prejudiced feelings and behaviors toward out-group members, so altruism should be negatively correlated with SDO. Furthermore, people who are quite inclusive in their definitions of what constitutes an in-group should be less able to discriminate against out-groups, so we expected communality to be negatively correlated with SDO. And finally, because tolerance is the antithesis of prejudice, we might expect that a general measure of tolerance would be negatively correlated with a general desire for in-group superiority. We used Davis' (1983) multidimensional empathy scale, Super and Nevill's (1985) altruism subscale, the Personal Attribute Questionnaire (PAQ) Communality scale (Spence, Helmreich, & Stapp, 1974), and the Jackson Personality Inventory (JPI) Tolerance scale (Jackson, 1976) to test these hypotheses. If SDO has merit as a new personality variable, none of these correlations should be very high.

PRESENT RESEARCH

Overview

We examined data from 13 samples to test the predictive and discriminant validity and reliability of our measure of SDO. Our logic in using this large number of samples is to examine statistically significant results that are reliable across samples. We organized the results by topic, but we report the results in each sample so that the reader can see the magnitude of effects in each sample and the stability of the results across samples. At the end of the Results section, we provide a summary of the results in the form of meta-analyses.

Data Collection

Generally, subjects were college students who participated in a study called "Social Attitudes" for partial course credit. All of their responses were anonymous and confidential, and they completed batteries of self-administered questionnaires. Subjects in Samples 2, 3b, 5,6,8, 9, and 13 spent about 1 hr in our laboratory completing the questionnaires. The experimenter described the study as designed to measure students' social attitudes and personal preferences. Subjects in Samples 1 and 13 completed the SDO scale after participating in unrelated experiments, and subjects in the remaining samples completed the SDO scale and follow-up scales in two consecutive mass-testing sessions normally conducted on subject pool participants. All subjects completed a demographic background sheet and our 14-item SDO scale intermixed with related items, a Nationalism scale based on Kosterman and Feshbach's (1989) measure, along with other attitude or experience measures, each having their own instructions and response scales. We also administered some standard personality or attitude scales according to the instructions of their authors. In several samples we also administered ideological (legitimizing myths) or policy attitude items on a questionnaire entitled "Policy Issues Questionnaire."

SOCIAL DOMINANCE ORIENTATION 745

Measures

SDO In previous archival studies, we measured proxies for SDO using items dealing with equality from the National Election Study or the S6 Conservatism scale (see Sidanius, 1976). In developing the present measure of SDO, we tested over 70 items whose content we felt related to SDO or to constructs one can define as separate but that might be considered adjacent to SDO (e.g., nationalism and prestige-striving), following Loevinger's (1957) suggestion about scale construction. However, on the basis of our desire to develop a simple, unidimensional scale that is balanced, we selected 14 items from this extensive questionnaire as the SDO scale. The selected items concerned the belief that some people are inherently superior or inferior to others and approval of unequal group relationships (see items in Appendix A). The 14-item SDO scale was balanced in that half the items indicated approval of inequality and half indicated approval of equality (see items in Appendix A). We assume that these items tap a latent construct and so we are interested in the relationships between the scale mean and other measures rather than relationships between individual SDO items and other measures. SDO is an attitudinal orientation, so instructions read, "Which of the following objects or statements do you have a positive or negative feeling towards? Beside each object or statement, place a number from' 1' to '7' which represents the degree of your positive or negative feeling." The scale was labeled very positive (7), positive (6), slightly positive (5), neither positive nor negative (4), slightly negative (3), negative (2), and very negative (1). The order of the SDO items and the filler items differed among Form A, completed by Samples 1, 2, 3, and 4; Form B, completed by Samples 5, 6, 7, 8, and 12; and Form C, completed by Samples 9, 10, and 11. The format and instructions for the three forms were identical, and we saw no evidence that results pertinent to reliability or validity issues differed across the questionnaire form. Subsequent to the present research, we have used just the 14 items on a questionnaire and found reliability coefficients of .90 and predictive validity results similar to those reported below.

Political-Economic Conservatism

Some of the standard scales assessing political-economic conservatism actually measure individuals' support for particular social policies (e.g., the C-scale, Wilson & Patterson, 1968). Because we wished to measure political-economic conservatism separately from policy attitudes, and because we wanted to use a measure that should not vary with time and place, we used a self-identified liberal-conservative measure in all samples. On the demographic background sheet, the political-economic conservatism question read, "Use one of the following numbers to indicate your political views in the accompanying categories." Below these instructions was a scale labeled very liberal (1), liberal (2), slightly liberal (3), middle of the road (4), slightly conservative (5), conservative (6), and very conservative (7) and a blank next to each type of issue: "foreign policy issues," "economic issues," and "social issues." Political-economic conservatism was the mean of self-ratings on these three items.

Authoritarianism Authoritarianism research has been fraught with measurement difficulties. After surveying the authoritarianism measurement literature, we decided to administer two rather different measures of authoritarianism, both of which are balanced: the Right Wing Authoritarian (RWA) scale by Altemeyer (1981) and Goertzel's (1987) bipolar personality measure. Goertzel (1987) intended his adjective checklist to measure the personality rather than the ideological aspect of authoritarianism, but did show that it correlates with attitudes toward policies falling along toughness and consistency dimensions. Altemeyer's (1981) scale is the only other internally reliable measure of authoritarianism that is close to the original conception of authoritarianism, including conventionalism, authoritarian submission, and authoritarian aggression (see Duckitt, 1989, for a review).

Original Legitimizing Myths and Policy Attitudes

The consent form and instructions informed subjects that their opinions and preferences toward a variety of ideas, kinds of people, events, and so forth would be measured. On our "Policy Issues Questionnaire" we included items from various legitimizing myth or policy attitude scales. Items from each scale were interspersed throughout the questionnaire. Next to each item was a 1-7 scale, and the instructions read, "Which of the following objects, events, or statements do you have a positive or negative feeling towards? Please indicate your feelings by circling the appropriate number alongside each item. Use one of the following responses. Remember, your first reaction is best. Work as quickly as you can." The scale points were labeled very negative (1), negative (2), slightly negative (3), uncertain or neutral (A), slightly positive (5), positive (6), and very positive (7). Items from the original legitimizing myths and policy attitude scales were selected for their content and for their internal reliability across samples. These scales are shown in Appendix B. Several personality measures were used as well; these are described in the Method section.

Method

Subjects

Although our 1,952 subjects were college students, they represent some diversity in terms of sex, ethnicity, and income groups, coming from public and private universities in California. Demographic information about the samples is shown in Table 1.

Samples and Procedures

Sample 1 (spring 1990) consisted of 98 University of California at Berkeley undergraduates who completed the CPI Dominance, Flexibility, and Capacity for Status subscales (Gough, 1987), the JPRF Dominance subscale (Jackson, 1965), the JPI Tolerance subscale (Jackson, 1976), and the Rosenberg (1965) Self-Esteem Scale (RSE). Sample 2 (fall and winter 1990-1991) consisted of 463 San Jose State University (SJSU) undergraduates who completed the CPI and JPRF Dominance subscales; Mirels and Garrett's (1971) Protestant Work Ethic Scale; the Just World Scale (Rubin & Peplau, 1975); the fourfactor Interpersonal Reactivity Index (IRI), which measures empathy (Davis, 1983); a number of policy attitude measures; and some demographic descriptors.

746 PRATTO, SIDANIUS, STALLWORTH, AND MALLE

Table 1 Description of Samples

Measure

n Age range % men % women

% Euro-American % Asian-American % Hispanic % Black % Arab-American

Under 20K 20-30K 30-40K 40-55K 55-7OK 70-100K 100-150K 15O-2OOK 200K.+

1

98 17-34 50 50

48 23 13 15 1

2

463 15-56 47 53

38 40 8 5 2

12 9 11 17 20 14 8 5 5

3a

81 17-21

3b

57 17-21 51 49

58 16 4 14 6

4 5

Sample

6

Age and gender breakdown

190

47 53

144 17-35 49 51

49 17-23 69 31

Ethnic breakdown

38 40 8 5 2

53 24 10 8 0

59 24 15 2 0

Family income

10 8 5 10 10 21 15 13 8

21 16 12 8 19 14 6 3 1

6 6 8 10 10 19 19 11 11

7

224

50 50

49 25 10 6 1

8

115 17-59 40 60

29 51 14 2 3

17 13 13 15 17 13 5 2 5

9

97 17-36 33 67

19 45 17 10 8

19 15 17 12 15 9 2 3 6

10

231

54 46

67 22 4 4 1

11

100

12

135

59 41

50 33 10 4 1

13

46

100 0

52 33 11 0 4

Note. Missing numbers indicate that information was not available. Samples 4, 7, 10-13 are probably similar in age distribution and range to Sample 3. Income was self-reported annual family income in thousands of dollars.

Sample 3a (September, 1990) consisted of 81 Stanford University undergraduates who completed the SDO scale as part of a mass-testing session. Sample 3b included 57 subjects from the same population who participated in a study in our lab in December, 1990, during which they completed the SDO scale again and a number of attitude and personality measures. The overlap of these two samples (N = 25 with complete data) was used to assess the cross-time reliability of SDO. Sample 4 (January, 1991) consisted of 190 Stanford University undergraduates who completed the SDO scale and an attitude scale about the Iraq war assessing environmental concerns in the war, anti-Arab racism, willingness to sacrifice for the war, willingness to restrict civil liberties for the war effort, and support for the use of military force by the United States against Iraq. Sample 5 (fall 1991) consisted of 144 SJSU undergraduates who completed the RSE (Rosenberg, 1965), the Rombough and Ventimiglia (1981) Tri-Dimensional Sexism Scale, the Sexist Attitudes Toward Women Scale (Benson & Vincent, 1980), the Rape Myths Scale (Burt, 1980), the Altruism subscale from the Values Scale (Super & Nevill, 1985), and the IRI (Davis, 1983). We also measured policy attitudes toward gay rights, women's equality policies, militarism, punitiveness, racial policies, and environmental policies. In addition, we measured ideologies such as anti-Black racism, elitism, patriotism, belief in equal opportunity, and opposition to miscegenation. Sample 6 (September, 1991) consisted of 49 Stanford undergraduates who completed the same measures as subjects in Sample 5. Sample 7 (September, 1991) consisted of 224 Stanford undergraduates who completed a battery of personality questions, including Malle and Horowitz's (1994) bipolar descriptions of Factors I (Extraversion), II (Agreeableness), IV (Neuroticism), and V (Conscientiousness) of the Big-Five personality dimensions (see John, 1990, for a review). A few weeks later, in the 3 days including and following the day Clarence Thomas was confirmed to the Supreme Court, those subjects who had

given their prior permission were telephoned and asked four questions about their opinions regarding this Supreme Court nomination. In all, 149 subjects were reached by telephone, and the response rate was 100%. Sample 8 (February, 1992) consisted of 115 Stanford undergraduates who completed the PAQ (Spence et al., 1974), CPI Dominance scale (Gough, 1987), JPRF Dominance scale (Jackson, 1965), JPI Tolerance scale (Jackson, 1976), IRI (Davis, 1983), RSE (Rosenberg, 1965), a post-Iraq war attitude survey, a general war attitude survey, and a number of other policy attitude measures similar to those in Sample 5. Sample 9 (April, 1992) consisted of 97 SJSU undergraduates. They completed the CPI and JPRF Dominance subscales; the JPI Tolerance subscale; the IRI; the Protestant Work Ethic Scale; all 19 of the authoritarian bipolar adjective choices (Goertzel, 1987); Altemeyer's (1981) 30-item RWA Scale; John, Donahue, and Kentle's (1992) Big-Five Personality Inventory; the PAQ; McConahay's (1986) Modern Racism Scale; and Katz and Hass' (1988) Pro-Black, Anti-Black, and Humanitarian-Egalitarian Scales. They also completed a number of policy attitude items similar to those for Sample 5. Sample 10 (March, 1992) consisted of 231 Stanford undergraduates who completed the SDO scale. Two weeks later, 176 of these subjects completed a comprehensive survey about their ideologies and general attitudes about the death penalty and their attitude about the execution of Robert Alton Harris, who was executed by the state of California the day before the survey was administered. Sample 11 (March, 1991) consisted of 100 Stanford University undergraduates who completed the SDO scale and a battery of other questionnaires including Snyder's (1974) self-monitoring scales; Fenigstein, Scheier, and Buss' (1975) Self-Consciousness scales; and Malle and Horowitz' (1994) bipolar adjective versions of Factors I and IV of the BigFive personality dimensions. Sample 12 (January, 1992) included 139 Stanford undergraduates

SOCIAL DOMINANCE ORIENTATION 747

who completed the SDO scale in a mass-testing session. Of these, 70 also completed Malle and Horowitz' (1994) measures of Factors I and IV. Sample 13 included 46 undergraduate men at Stanford during 19901991 who were selected to be in an experiment on the basis of having either extremely high or low SDO scores in Samples 3,4, and 11. They participated in the experiment between 6 weeks and 8 months after their first testing and completed the SDO scale again.

Results

We first present the internal and temporal reliability of our SDO scale. We then examine whether this measure related to the ideological, policy attitude, and hierarchy role variables predicted by social dominance theory. We show that SDO was either independent of other personality variables with which it might be confused or that it predicted the attitudinal outcomes over and above the effects of these other variables. We also show that it was not redundant with other personality measures. Finally, we show that SDO predicted new social and political attitudes. To summarize the results across samples, we report simple averages of the internal reliability coefficients across samples and averaged correlations across samples using Fisher's z-to-r transformation.

Reliability of the SDO Measure

Unidimensionality

We conducted two kinds of analyses to confirm that the 14 SDO items assessed a single construct. First, within each sample, principal-components analyses of the 14 SDO items showed that a single dimension captured the bulk of the variance in these items. That is, there was a precipitous drop between the values of the first and second eigenvalues in every sample. Second, we subjected our largest sample, Sample 2 {N = 446 with complete data on all SDO items) to confirmatory factor analysis. Using maximum-likelihood estimation, we tested a model in which all 14 items were driven by a single latent construct. Each item had a statistically significant relationship to the latent factor (ps < .0001). By freeing only 3 of 91 possible off-diagonal elements of the 66 matrix,2 we obtained a satisfactory x2/dfratio of 2.89 (e.g., Carmines & Mclver, 1981), suggesting that our data are consistent with a model in which a single dimension underlies responses to all the items. Thus, the 14 items appear to measure a unitary construct.

Internal Reliability

Item statistics showed that the 14-item SDO scale showed good internal reliability across all samples, averaging a = .83 (see internal reliability coefficients and item statistics by sample in Table 2). Item analyses also showed that all items were highly correlated with the remainder of the scale in every sample. The average lowest item-total correlation across samples was .31 and the average highest item-total correlation across samples was .63. Item 7 had the lowest item-total correlation in 4 of 12 independent samples (Z = 3.52, p < .001). Item 9 had the highest item-total correlation in 3 samples (Z = 2.40, p < .01). No other items were either the most or least correlated across samples in numbers that differed from chance using a binomial test.

Stability of SDO Measure Over Time

We measured the stability of scores on our scale over time in two samples. Twenty-five of the subjects in Sample 3 were tested on SDO twice at a 3-month interval. Their SDO scores substantially correlated from Time 1 to Time 2 (r = .81, p < .01). The mean difference from Time 1 to Time 2 was 0.09 on a 7-point scale, which did not differ reliably from zero {t < 1). In contrast, the Time 1-Time 2 correlation for the 10-item RSE was .50. Sample 13 consisted of 46 of the highest and lowest scoring men on the SDO scale from Samples 3, 4, and 11, who completed that scale again some months later. The correlation in this sample from Time 1 to Time 2 was .84 (p < .001), and the mean difference in scores from Time 1 to Time 2 was essentially zero (M = 0.03, t < 1; for the high group, M = -0.03 and for the low group, M = 0.09). All of the subjects first classified as "high" or "low" on SDO met this criterion again in the second testing. The near-zero mean changes within both groups are particularly telling because one could have expected at least some regression toward the mean. Thus, even in different testing contexts, our SDO measure appears highly stable in the short term.

Predictive Measures

Gender Differences

The gender difference we expected showed in all but two samples; men were higher on SDO than women (see point-biserial correlations in Table 2).

SDO and Hierarchy Role

A question on the demographic background questionnaire asked subjects in what sector of the economy they intended to work after graduation. There were 20 career choices provided. Theoretically, we define those whose work is primarily aimed at protecting, serving, or benefiting elite members of society more than oppressed members of society "hierarchy-enhancing." Those whose work benefits the oppressed more than elites we define as "hierarchy-attenuating." As such, we classified subjects as (a) hierarchy enhancers (those intending careers in law, law enforcement, politics, and business); (b) "middlers" who would not obviously attenuate or enhance inequality through their professional work, such as science and sales; or (c) hierarchy attenuators (those intending to be in such professions as social work or counseling; see also Sidanius, Pratto, Martin, & Stallworth, 1991). We predicted that hierarchy enhancers would have higher SDO levels than hierarchy attenuators, and that middlers' SDO levels would fall somewhere between the other two. Sample 2 was large enough to test this hypothesis; we also combined Samples 5, 6, 8, and 9 to replicate the test. Because more women tend to go into hierarchy-attenuating careers, and because we know that SDO exhibits a gender difference, we also included subject sex as an independent variable along with hierarchy role. SDO was the outcome variable in simultaneous regression-style analyses of variance (ANOVAs)

2 The freed elements of the matrix corresponded to Items 8 and 9, Items 2 and 4, and Items 10 and 11 in Apendix A.

748 PRATTO, SIDANIUS, STALLWORTH, AND MALLE

Table 2 Coefficient Alphas, Correlation With Subject Gender, and Average Item Means and Variances by Sample for 14-Item Social Dominance Orientation Scale

Measure

a

M Variance

1

.85 .29** 2.44 0.14

2

.83 .27** 2.74 0.22

3a

.84 .32** 2.55 0.18

3b

.85 .31* 2.31 0.17

4

.84 .32** 2.59 0.21

5

.81 .11 2.97 0.40

Sample

6

.84 .36* 2.50 0.24

7

.89 .28* 2.59 0.23

8

.82 .27** 3.02 0.18

9

.80 .03 3.12 0.36

10

.83 .30** 3.13 0.66

11

.81

2.91 0.27

12

.83 .26** 2.60 0.23

Note. Positive correlations with gender indicate that men were higher than women *p<.05. **p<.0l.

with planned contrasts. In Sample 2, the results were as expected: Those who intended to work in hierarchy-attenuating professions had lower SDO levels (M = 2.28) than did middlers {M = 2.72), F( 1,432) = 5.49, p < .05, and also lower levels than those intending to work in hierarchy-enhancing professions (M = 2.88), F(l, 432) = 10.21, p < .01. Men also had higher SDO levels (M = 3.03) than women (M = 2.51), F( 1,432) = 36.86, p < .001. In the merged sample, hierarchy attenuators again had lower SEX) levels (M = 2.64) than hierarchy enhancers (M = 3.09), F(l, 378) = 5.01, p < .05. Middlers' SDO levels were in the middle (M = 2.94) and were not distinguishable from those of either enhancers or attenuators. Again, men (M = 3.07) had higher SDO levels than women (M = 2.90), F(, 378) = 3.72, p = .05. Results from both these large samples indicate that intended hierarchy attenuators did indeed have lower SDO levels than intended hierarchy enhancers, even after controlling for subjects' sex.

SDO and Hierarchy-Legitimizing Myths

We hypothesized that SDO should be related to any social or political ideology that helps legitimize group-based inequality. Ideologies. The three-item index of self-described political ideology had good internal reliability, averaging a = .78 across samples (see Table 3). SDO correlated positively and significantly with political-economic conservatism in 7 of 8 samples, averaging r = .38; conservatives were higher on SDO than liberals (see Table 3). The scales measuring meritocratic ideologies, the Protestant Work Ethic and Just World Scales, had fairly low internal reliabilities in all samples, considering that they are 19and 20-item scales, respectively (see Table 3). In a Stanford sample (Sample 3b), but not in two samples from SJSU (Samples 2 and 9), the Protestant Work Ethic Scale and Just World Scale had significant positive correlations with SDO (see Table 3). This suggested to us that variations in the cultural background of these samples may affect the ideologies known to and accepted by them. Results from our demographic questionnaire showed that compared with Stanford, SJSU tends to have more first-generation American, more Catholic, and fewer EuroAmerican students. The other legitimizing myth scales that we constructed fared better (see items for all other scales Appendix B). The nationalism, patriotism, cultural elitism, and equal opportunity measures all had good internal reliability and were positively correlated with SDO (rs ranged from .22 to .67), with only one ex

ception (see Table 3). These correlations showed that the more subjects tended to prefer group dominance in general, the more nationalistic and patriotic they were (average rs = .51 and .45, respectively) and the more they subscribed to cultural elitism (average r = .40) and equal opportunity ideologies (average r = .46). As predicted, the noblesse oblige scale was strongly negatively correlated with SDO in every sample, ranging from -.39 to -.69 (see Table 3). In Samples 1, 3a, and 7 (not shown in Table 3), SDO correlated -.47, -.56, and -.67 with noblesse oblige (ps < .01), for an overall average correlation of-.54. Ethnic prejudice. SDO was strongly correlated with our anti-Black racism measure in every sample, ranging from .42 to .65 and averaging .55 (see Table 3). In Sample 4, we also measured anti-Arab racism (a = .73), which correlated with SDO (r = .22, p < .05). In Sample 9, we administered McConahay's (1986) seven-item Modern Racism Scale (a = .79), which correlated .53 with SDO. Katzand Hass' (1988) 10-item Pro-Black Scale (a = .68) was negatively correlated with SDO (r = —.38, p < .01), and their 10-item Anti-Black Scale (a = .62) was positively correlated with SDO (r = .30, p < .01). These results, using rather different racism measures, are consistent with the idea that generalized preference for group dominance drives belief in culturally specific forms of ethnic prejudice. Sexism. We assessed antifemale sexism in Samples 5 and 6 with several measures, all of which proved to be internally reliable. These measures were highly correlated with SDO (rs ranging from .34 to .63; see Table 4). Across both samples and all sexism measures, the average correlation was .47. Partial correlations controlling for gender with SDO were also reliable and of about the same magnitude. For this reason, the large correlations between SDO and sexism cannot be attributed to gender differences on SDO or sexism measures. In summary, all of the measured ideologies (hierarchy-legitimizing myths) except the Protestant Work Ethic Scale and Belief in a Just World Scale were reliably correlated with SDO in the expected directions across virtually all samples. SDO was most strongly related with ideologies concerning group prejudice against other nations, ethnic groups, and women.

SDO and Policy Attitudes

We hypothesized that SDO would predict support for social policies with implications for the distribution of social value among groups. We assessed attitudes toward chauvinist policies (United States dominating other national groups), law and or

SOCIAL DOMINANCE ORIENTATION 749

Table 3 Coefficient Alphas of Legitimizing Myth Scales and Correlations With Social Dominance Orientation Sample

No of Sample 2 Sample 3b Sample 5 Sample 6 Sample 8 Sample 9 Sample 10 Measure items (n = 408) (n = 57) Sample 4 (n = 144) (n = 49) («=115) (n = 95) (n = 156)

Coefficient a

Political-economic conservatism 3 .69 .83 .89* Protestant Work Ethic 19 .68 .75 Just World 20 .55 .42 Nationalism 6 .75 .88 .80" Patriotism 12 Cultural elitism 7 Equal opportunity 6 Noblesse oblige 6 .58 .80 .69" Anti-Black racism 5 .68 .77

.80

.68 .83 .67 .65 .68 .70

.71

.86 .89 .78 .76 .73 .74

.80

.78

.56 .60

.72 .73

.66 .80 .59 .49 .72 .77

.78

.62

.54

Political-economic conservatism Protestant Work Ethic Just World Nationalism Patriotism Cultural elitism Equal opportunity Noblesse oblige Anti-Black racism

.26** -.03 .09 .52**

-.39" .57**

Correlations

.28* .33* .43** .41**

-.54** .42**

44»*b

.53**"

_43*.b

.11

.43** .43** .51** .51** -.60** .49**

.72**

.67** .65** .23 .51**

.61**

.24*

.51**

-.69** .65**

.17* .03

.47** .22* .44** .34** -.50** .52**

.55**

.72**

-.72**

•n=180. »« = 90. *p<.05. **p<.0l.

der policies, military programs, gay rights, women's rights, social programs generally, racial policies, and environmental policies in most of our samples (see coefficient alphas in Table 5). Support for chauvinist policies and law and order policies were positively correlated with SDO in almost all samples, averaging .34 and .28, respectively. Support for military programs was positively correlated with SDO in all samples, averaging .44. Support for gay rights, women's rights, social welfare programs, ameliorative racial policy, miscegenation, and environmental policy were significantly negatively related to SDO in all but three cases (see Table 5). These relationships were of about the same magnitudes as the policy attitudes described above. We assessed political party preference by having subjects rate themselves from strong Democrat (1) through independent (4) to strong Republican (7) and others. Excluding "others," Republican political party preference correlated positively and sig

nificantly with SDO in six out of six samples, averaging .28 (see Table 5). In addition to support for military programs, we expected support for military action including war to be positively related to SDO. We tested this hypothesis by surveying attitudes toward war in general and specific attitudes toward the war against Iraq fought by the United States and other nations at the time of data collection. In fall 1990, while Iraq was occupying Kuwait and the United States was amassing troops near Iraq, data from Sample 2 were collected, including a single war policy item, "Going to war to maintain low oil prices." This item correlated .30 with SDO (p < .01). In January, 1991, when the United States and allies had just begun bombing Iraq, we administered a balanced scale concerning war and related attitudes to Sample 4. A reliable (a = .85) eight-item pro-war scale correlated .51 with SDO (p < .01). One year later, we asked Sample 8 about

Table 4 Coefficient Alphas of Sexism Scales and Correlations With Social Dominance Orientation Within Samples

Measure

Rombough & Ventimiglia sexism Sex differences Internal (household) labor External (paid) labor Sexist Attitudes Toward Women Rape Myths

No. of items

20 5 10 6 40 10

Coefficient a

Sample 5

.90 .68 .89 .78 .91 .84

Sample 6

.94 .65 .94 .85 .94 .75

Correlations

Sample 5

.44** .38** .34" .45** .46" .46"

Sample 6

.54" .56" .63" .36* .55" .40**

*p<.05. "/x.Ol.

750 PRATTO, SIDANIUS, STALLWORTH, AND MALLE

Table 5 Coefficient Alphas of Policy Scales, Correlations With Social Dominance Orientation, and Partial Correlations Controlling for Conservatism, Across Samples

Policy scale

Chauvinism Law and order Military programs Gay & lesbian rights Women's rights Social programs Racial policy Miscegeny Environmental policies

Chauvinism Law and order Military programs Gay & lesbian rights Women's rights Social programs Racial policy Miscegeny Environmental policies Republican party preference

Chauvinism Law and order Military programs Gay & lesbian rights Women's rights Social programs Racial policy Miscegeny Environmental policies

No. of items

8 4 3 2 4 10 7 2

5

2 (« = 455)

.64 .67" .82 .63 .78 .71 .96

3b (n = 50) 4

Coefficient a

.73

.71 .75 .91 .72 .77 .81 .97

Sample

5 (n = 129)

.59 .73 .85 .69 .79 .68 .93

.71

6 (n = 37)

.73 .IT .67 .86 .63 .86 .72 .94

.80

Correlations of social dominance orientation and policy items

.08 .33** -.32** -.42** -.50** -.42** -.31**

.15**

.23* .27* -.50** -.32** -.31** -.46** -.15

.25*

.37** .30** .33** -.29** -.39** -.29** -.23** -.30**

-.27**

.24**

.49** .59** .70** -.55** -.34* -.70** -.62** -.31*

-.40**

.45**

Partial correlations removing political-economic conservatism

-.02 .16*** -.28*** -.38** -.30*** -.33*** -.28***

.15 .18 -.32** -.31** -.27* -.30** -.19

.40** .29*** .31*** -.29*** -.35*** -.30*** -.22*** -.31***

-.27***

.16 .31* .40** -.14 -.27* -.49** -.38** -.08

-.31*

8 (n = 100)

.67

.80° .66C .60c .91

.24*

-.52** -.55** -.54** -.25*

.33**

.25***

-.46*** -.50*** -.49*** -.23**

9 (n = 89)

.58 .77" .59 .83 .74 .81" .77 .87

.76

.14 .19 .47** -.17 -.42** -.39** -.34** -.18

-.47**

.27*

.06 .15 .46*** -.15 -.40*** -.37*** -.31*** -.17*

-.46***

* Three items. b Two items. c Six items. d Seven items. *p<.05. **p<.0. ***/><.001.

their attitudes toward the Iraq war. The resulting Iraq War Attitudes scale was reliable (a = .85) and correlated .29 with SDO

Does SDO, then, unconditionally predict support for war, or only war for certain purposes? We attempted to answer this question by designing a General War Attitudes scale including two kinds of items, namely, Wars of Dominance, which we expected to relate positively to SDO, and Wars for Humanitarian Reasons, which we did not expect to relate positively to SDO. This scale was administered to Sample 8 in January, 1992. Factor analysis confirmed that these were two independent dimensions. The Wars of Dominance scale (eight items) was reliable (a = .82) and correlated positively with SDO (r= .31,p< .01). The Humanitarian Wars scale (six items)3 was also reliable (a = .73) and correlated negatively with SDO (r = -.41, p < .01), so

SDO is not merely antipacifism. The Wars of Dominance scale was positively correlated with support for the Iraq war (r = .63, p < .001), but the Humanitarian Wars scale was uncorrelated with support for the Iraq war (r = .07). These results suggest that SDO does not predict support for war unconditionally; rather, SDO predisposes people to endorse group dominance ideologies, thus facilitating support for wars of dominance.

Discriminant Validity

We expected SDO to correlate with political-economic conservatism, and indeed it did. However, to show that SDO has

! Two unreliable items were eliminated from the scale.

SOCIAL DOMINANCE ORIENTATION 751

utility as a predictor of policy attitudes over and above politicaleconomic conservatism, we computed the correlations between SDO and the policy attitudes reported above after partialing out political-economic conservatism. Of the 41 significant zero-order correlations between SDO and policy attitudes in Table 5, only 5 become nonsignificant when political-economic conservatism is partialed out.4 A few of the very high zero-order correlations were reduced substantially, but many more partial correlations were almost the same as the zero-order correlations (see Table 5). Across all the samples, then, there was no consistent evidence that political-economic conservatism could replace SDO as a predictor of the policy attitudes we assessed. In Sample 9, we assessed another rival predictor of policy attitudes, namely authoritarianism, using two measures. Altemeyer's 30-item RWA scale had good internal reliability {a = .78); Goertzel's measure was adequate for a bipolar scale (a = .53). Both measures of authoritarianism correlated with political-economic conservatism (r = .31 for RWA, r = .29 for the Goertzel measure, both ps < .01), confirming their validity. Neither, however, correlated strongly with SDO. RWA correlated .14 (ns) with SDO, and the Goertzel measure correlated .18 with SDO (p < . 10). Correcting these correlations for attenuation yielded slightly higher correlations (r* = .18, p < .05 for RWA; r* = .28, p < .01 for the Goertzel measure). We also computed partial correlations between SDO and the policy attitudes, partialing the two authoritarianism measures and political-economic conservatism. In Sample 9, all the policies that showed significant zero-order correlations with SDO also had significant correlations with SDO, partialing out the effects of political-economic conservatism, RWA, and the Goertzel measure. Both authoritarianism measures showed substantial zero-order correlations with attitudes that were not as highly correlated with SDO in this sample: gay rights (r = -.51 for RWA, r = -.31 for the Goertzel measure, ps <.01) and chauvinistic policies (r = .38 for RWA, p < .01, r = .25 for the Goertzel measure, p < .05). As Peterson, Doty, and Winter (1993) showed recently, authoritarianism still predicts social attitudes, particularly those relevant to untraditional sexual practices and prejudice against foreigners. Because authoritarian personality theory (Adorno et al., 1950) also postulates that authoritarianism should predict ethnocentrism, racism, nationalism, and conservatism, we tested whether SDO would still predict belief in these legitimizing myths, controlling for authoritarianism. All the reliable zeroorder correlations between SDO and ideological measures were reliable after controlling for RWA and the Goertzel measure, except for the correlation with political-economic conservatism. The correlation between SDO and political-economic conservatism, partialing RWA, was . 13 (p = . 11). Partialing the Goertzel measure, the correlation between SDO and politicaleconomic conservatism was .16 (p = .07), and partialing both measures, the correlation was .13 (jo = .11). Although the relationship between SDO and conservatism may be explained by their joint relationship to authoritarianism, the relationships between SDO and racism and nationalism cannot.

Dominance and Self-Esteem

Conservatism and authoritarianism were the only rival variables we identified as predictors of social and political attitudes,

and the analyses above show that SDO substantially related to such attitudes, even when controlling for political-economic conservatism and for authoritarianism. To show a different kind of discriminant validity, we tested whether SDO correlated with other personality measures. If any of these were large, we would then be obliged to test the partial correlations with the social and political attitudes discussed above. Only once did SDO correlate with the CPI and JPRF Dominance subscales across fives amples (see Table 6). On average, CPI Dominance correlated .03 with SDO, and JPRF Dominance correlated -.006. These results clearly indicate that SDO is independent of interpersonal dominance. In Sample 1, SDO was also unrelated to CPI Flexibility (r = .06) and Capacity for Status (r = .05). For the most part, SDO was also uncorrelated with self-esteem in Samples 1 through 9, averaging -.08 (see Table 6).

Other Personality Measures

We used data collected by other researchers at Stanford during mass testing sessions to further investigate the discriminant validity of SDO. SDO was uncorrelated with all the self-monitoring and self-consciousness scales in Sample 11. In Samples 7, 9, 11, and 12, SDO correlated -.06, -.11, .08, and -.19, respectively, with Extraversion; none of these correlations differed reliably from zero. SDO correlated -.02,. 13, -.08, and .21 in those samples, respectively, with Neuroticism; none of these differed reliably from zero. SDO correlated -.03 with Agreeableness in Sample 7 and -.41 (p < .01) in Sample 9. SDO correlated -.04 and -. 14 with Conscientiousness in Samples 7 and 9, neither of which differed reliably from zero. SDO correlated -.28 with Openness (p < .01) in Sample 9. These data do not suggest that SDO is redundant with any of the BigFive dimensions and strongly imply that SDO is independent of Extraversion and Neuroticism. Across this set of correlations, there was also no evidence that SDO is related to the positively valued personality dimensions (e.g., Extraversion and Calmness) as opposed to the negative dimensions (e.g., Introversion and Neuroticism).

Convergent Validity

Empathy, Altruism, and Communality

We expected that feelings of closeness and kindness toward others should mitigate desire to dominate other groups, so empathy, altruism, and communality should correlate negatively with SDO. We first tested whether different varieties of empathy were negatively related to SDO using Davis' (1983) IRI. The Concern for Others subscale was significantly negatively correlated with SDO in every sample (rs ranged from —.40 to —.53 and averaged -.46; see Table 7). High dominance-oriented people expressed less concern for others than did low dominanceoriented people. The patterns of correlations between SDO and the other subscales were not as consistent across samples, but when they were significant, all correlations were negative (see Table 7). The total Empathy scale was negatively correlated

4 One other correlation actually became significant because partial correlations use one-tailed tests.

752 PRATTO, SIDANIUS, STALLWORTH, AND MALLE

Table 6 Coefficient Alphas of Interpersonal Dominance and Self-Esteem and Correlations With Social Dominance Orientation Within Samples

Measure

CPI Dominance JPRF Dominance Rosenberg Self-Esteem

CPI Dominance JPRF Dominance Rosenberg Self-Esteem

No. of items

35 19 10

1 (n = 98)

.82 .81 .87

-.11 -.04 -.09

2 (« = 403)

.79 .81 .87

-.03 .13** -.18

3a 3b (n = 80) (n = 57)

Coefficient a

.79 .74 .88 .88

Correlations

-.17 -.17 .09 .01

Sample

4 (n = 90)

.90

.16

5 (« = 144)

.87

-.23**

6 (« = 56)

.85

-.01

8 (« = 115)

.71 .73 .84

.24** .01 -.29**

9 (n = 95)

.71 .69 .83

.20 .04 -.14*

Note. CPI = California Psychological Inventory; JPRF = Jackson Personality Research Form. *p<.05. **p<.0l.

with SDO in 5 of 6 samples (averaging r = -.31), but not as highly as the Concern for Others subscale. We conclude that concern for others is the form of empathy that precludes the desire to dominate other groups. As one might expect, altruism was correlated with the Concern for Others subscale in Sam