Sign up

A test of Dovidio and Gaertner's integrated model of racism.
Article Type:
Survey
Subject:
Americans (Surveys)
Racism (Surveys)
Racism (Research)
Authors:
Nail, Paul R.
Harton, Helen C.
Barnes, Anna
Pub Date:
03/01/2008
Publication:
Name: North American Journal of Psychology Publisher: North American Journal of Psychology Audience: Academic Format: Magazine/Journal Subject: Education; Psychology and mental health Copyright: COPYRIGHT 2008 North American Journal of Psychology ISSN: 1527-7143
Issue:
Date: March, 2008 Source Volume: 10 Source Issue: 1
Topic:
Event Code: 310 Science & research
Geographic:
Geographic Code: 100NA North America

Accession Number:
178452280
Full Text:
Politically liberal and conservative White Americans were asked to evaluate a legal case in which a police officer was acquitted of assault charges against a motorist in state court but then retried in federal court. When the initiator of the assault was unspecified, liberals perceived a violation of double jeopardy significantly more for a Black officer-White motorist pair than vice versa. When the officer was identified as the initiator, however, double jeopardy ratings decreased significantly only for the Black officer. These results support the hypothesis that liberals, as aversive racists, will often discriminate in favor of Blacks except when discrimination against Blacks can be justified as non-racist. Unexpectedly, conservatives showed no indication of racial bias under any of the conditions.

**********

Contemporary theories of racism propose that race-based prejudice in North America has not decreased as much in recent years as some surveys would seem to indicate (see Dovidio & Gaertner, 1998, 2005). Instead, the expressions of racism are simply less overt than they used to be. The theories of modern racism (McConahay, 1986) and symbolic racism (Sears, 1998; Sears & Henry, 2003) both posit that due to changing social norms against direct and overt expressions of racism many White Americans beginning in the 1960s have tended to express their racism in indirect, disguised, or symbolic ways. For example, such individuals might argue that they oppose social policies like affirmative action not because they are prejudiced as individuals but because it is against a proper interpretation of Civil Rights laws to use race as a factor in decisions such as hiring and university admissions. By taking such a position, modern-symbolic racists can continue to maintain or at least project a non-prejudiced self-image despite possessing negative, explicit, race-based beliefs and attitudes.

The theory of aversive racism (Dovidio & Gaertner, 1998, 2005) also proposes subtle, indirect expressions of prejudice by White Americans. However, aversive racists, unlike modern-symbolic racists, are hypothesized to have internalized some genuine egalitarian attitudes like fairness and equality and also to possess a sincere, non-prejudiced self-image. Yet, Dovidio and Gaertner still categorize such individuals as racist because research indicates that they have non-conscious negative, race-based feelings that sometimes bubble to the surface as prejudice (e.g., Dovidio, Evans, & Tyler, 1986; Dovidio & Gaertner, 1991; Son Hing, Chung-Yan, Grunfeld, Robichaud, & Zanna, 2005; Son Hing, Li, & Zanna, 2002). Thus, a non-conscious emotional conflict is hypothesized to exist in these individuals due to their need to maintain a genuinely non-prejudiced self-image while at the same time they experience involuntary, negative, race-based feelings. The label aversive racism describes this unpleasant internal conflict.

Dovidio and Gaertner's (1998, 2005) Integrated Model of Racism

Dovidio and Gaertner's (1998, 2005) integrated model suggests that the modern, symbolic, and aversive racism models all have validity but that they apply basically to different types of people. Specifically, the model links modern-symbolic racism to political conservatives and aversive racism to political liberals. The reasoning is that North American conservatives as a group still maintain many prejudiced beliefs and attitudes and have only learned that it is improper to express such beliefs in most contemporary public settings. North American liberals, in contrast, while strongly opposed to racism in both their public behaviors and private beliefs, still maintain negative but largely unconscious feelings/beliefs based on race.

The integrated model sometimes predicts different reactions from conservatives and liberals in specific inter-racial settings. For example, White North American conservatives should show bias against Blacks relative to Whites anytime they are not aware that their race-related behavior is under surveillance (e.g., Gaertner, 1973; Nail, Harton, & Decker, 2003). If participants do not know that their potential discrimination is being monitored, they should be less on guard to conceal their prejudice.

White North American liberals, on the other hand, should show favoritism toward Blacks relative to Whites in any situation where (a) race and (b) norms of fairness are both salient (Dovidio & Gaertner, 1998). Such favoritism is a way that liberals can support their non-prejudiced self-image while simultaneously defending against their automatic, negative, race-based feelings (e.g., Fein, Morgan, Norton, & Sommers, 1997; Frey & Gaertner, 1986). We refer to such favoritism as the reverse discrimination effect. It is reminiscent of Freud's reaction formation and of Shakespeare's line from Hamlet, "The Lady doth protest too much, methinks."

Research Evidence Concerning the Integrated Model

Evidence consistent with the integrated model has been reported in numerous studies (e.g., Gaertner, 1973; Fein et al., 1997; Frey & Gaertner, 1986; Son Hing et al., 2002, 2005). Research by Nail et al. (2003), however, is the only published account performed since the initial proposal of the integrated model. They tested the model directly by measuring political orientation and including both conservatives and liberals in the same study, inspired by the infamous Rodney King beating by Los Angeles police officers in 1991. White American participants in Study 1, Sample 1 were asked to evaluate a legal case in which a White male police officer physically assaulted a Black male motorist or vice versa. The officer was acquitted of assault and battery charges in state court but was later tried and convicted in federal court. The dependent variable was the extent to which participants perceived that the case had violated the U.S. Constitution's protection against trying a defendant twice for the same crime (viz., double jeopardy).

The findings supported the integrated model's predictions of bias on the part of both conservatives and liberals. Importantly, however, the bias shown by each group was different in form. Conservatives gave significantly higher double jeopardy ratings to the White officer (M = 4.62) than to the Black officer (M = 1.25). This finding supports the integrated model's prediction that many conservatives are still fundamentally prejudiced at heart and that, consequently, they will discriminate against minorities if they are unaware that their prejudice is under surveillance, as in between-participants designs.

For liberals, however, the results were reversed; they gave significantly higher double jeopardy ratings to the Black officer (M = 3.18) than to the White (M = 1.91). This finding supports the integrated model's prediction that when race and norms of fairness are both salient, liberals will show favoritism toward minorities (i.e., the reverse discrimination effect). Presumably this is a means of defending against their negative race-based feelings while at the same time providing strong evidence that they are not prejudiced. Race was salient because the confrontation was always between a Black and a White. Norms of fairness were salient because the study concerned a legal case and trial. The courtroom, with its standard of blind justice, is ideally the quintessential American fairness setting. The Nail et al. (2003) findings were replicated in two follow-up experiments (Study 1, Sample 2; Study 2).

Despite this support for the integrated model, the Nail et al. (2003) research is limited in at least one important way: It did not examine what is perhaps the integrated model's most intriguing (and risky) prediction--that liberals, as aversive racists, will show discrimination against Blacks relative to Whites if such discrimination can be reasonably justified as due to factors other than race. According to Dovidio and Gaertner's (1998, 2005) theorizing, in interracial settings liberals/aversive racists spend a good deal of time and energy suppressing their automatic negative feelings toward Blacks and other minorities. Thus, the theory goes, they should be on edge and anxious to vent these negative feelings in the form of direct discrimination against Blacks. However, they need a good excuse so they can discriminate without challenging their own non-prejudiced self-image. Such a condition should exist any time there are strong cues from the environment that might serve as a justification for a negative evaluation of a Black (Frey & Gaertner, 1986; Son Hing, Chung-Yan, Hamilton, & Zanna, in press).

Employing the Nail et al. (2003) double jeopardy paradigm, what if some liberals are told that it was the Black officer who had initiated the assault by throwing the first punch? Liberals should give significantly lower double jeopardy ratings here relative to liberal participants not told who initiated the assault. This prediction holds because only participants told that it was the Black officer who had thrown the first punch would be able to clearly attribute their low rating to the Black officer's behavior--initiating the assault--rather than to his race. Based on the theory of aversive racism, liberals can not easily attribute the officer's behavior to his race because it would threaten their non-prejudiced self image. Even stronger support for the integrated model would be obtained if, knowing that the officer threw the first punch, liberal Black officer-White motorist participants were to give significantly lower double jeopardy ratings than liberal White officer-White motorist control participants.

There is surprisingly little direct evidence that liberals will discriminate against Blacks if it can be justified as non-racist. Little direct evidence exists because few studies have assessed political orientation while at the same time providing a strong excuse to discriminate against Blacks. Studies by Gaertner (1973) and Sniderman, Piazza, Tetlock, and Kendrick (1991) are two such studies. Yet, the evidence they supply regarding the integrated model is mixed.

Gaertner (1973) had confederates telephone predominately White liberals and conservatives at home with a request for help. The caller was unknown to the resident and was either Black or White. Black callers spoke with "a modified 'Southern Negro' dialect," White callers "with a speech pattern common to whites in New York" (p. 336). If the resident did not hang up before the request for help was made, conservatives were significantly more likely to help a White (92%) than a Black caller (65%). Liberals showed no significant difference in helping as a function of race, White (76%), Black (64%). Note that these results for conservatives support the integrated model. Because the phone call was presumably a private one, conservatives' racial bias was not under surveillance, at least to their knowledge. As a result, conservatives apparently felt free to let their true colors show (i.e., discriminating against the Black by not helping). Support for the integrated model would have been stronger, however, had liberals shown the reverse discrimination effect, that is, significantly more helping for the Black caller than for the White.

What happened if the resident hung up before the request for help could be made? Here liberals showed bias against a Black caller (19%) relative to a White (3%), whereas the difference among conservatives was not significant statistically, Black (8.3%), White (4.7%). Note that the liberals' discriminatory behavior against Blacks in this case supports the integrated model in that liberals could attribute their abrupt ending of the call to a strong external cue other than race, namely, getting an uninvited, probably unwanted phone call at home from a stranger. Under these conditions, liberals could hang up on the Black caller without threatening their self-image as being non-prejudiced.

Sniderman et al. (1991) conducted telephone interviews of White American liberal and conservative participants regarding the degree to which a White or Black laid-off worker should be entitled to governmental assistance. In one condition, a strong external cue for discrimination was provided in that the worker was described as being either dependable or undependable. None of the results, however, were very supportive of the integrated model. In the dependable condition, a significantly greater percentage of conservatives supported at least some assistance for the Black worker (81%) as compared to the White worker (59%), whereas in the undependable condition there was no difference among conservatives in the percentage favoring support as a function of race. The percentage of liberal participants endorsing help in the undependable/Black worker condition (84%) was not less than that in the undependable/White worker condition (81%), this despite the fact that liberals here might have attributed any lack of support to a strong external cue--the worker's lack of dependability. In the dependable condition, liberals also showed no difference in the percentage endorsing help between Blacks (83%) and Whites (83%).

Further studies are needed because of (a) the lack of studies specifically designed to test Dovidio and Gaertner's (1998, 2005) integrated model by including a measure of participants' political orientation, (b) the failure of the Nail et al. (2003) studies to include a strong external cue that might justify discrimination by liberals against Blacks, and (c) the lack of consistency in findings between studies that have included such a cue (viz., Gaertner, 1973; Sniderman et al., 1991). The purpose of the present research is to provide a critical test of the aversive racism/integrated models. We pursued this goal by extending the Nail et al. double jeopardy paradigm to include in some conditions a strong situational cue that might serve to justify discrimination against Blacks.

The Present Research

We employed a 2 X 3 X 2 design. One independent variable was a measured individual difference variable, political orientation (liberal vs. conservative). A second was a manipulated, between-participants variable, race combination (White officer-White motorist [control], Black officer-White motorist, or White officer-Black motorist). The third independent variable was also manipulated between participants: Who threw the first punch? (unspecified vs. the officer).

Hypotheses Regarding Liberals. Based on Dovidio and Gaertner's (1998) integrated model and the Nail et al. (2003) research, the results for liberals in the "first punch unspecified" cells should replicate those of Nail et al., where the thrower of the first punch was also unspecified. Accordingly, Hypothesis 1 states that liberals in the Black officer-White motorist/unspecified cell should show favoritism to the Black officer; specifically, these participants should perceive double jeopardy significantly more than liberals in the White officer-Black motorist/unspecified cell. Such favoritism can supply evidence to liberals/aversive racists that they really must be non-prejudiced after all.

Notwithstanding this prediction, if the integrated model is correct in its proposition that liberals/aversive racists will reveal their prejudice by discriminating against Blacks provided that such discrimination can be justified, Hypothesis 2 states that liberals should significantly decrease their double jeopardy ratings for the Black officer once it is revealed that he initiated the assault; specifically, liberals in the Black officer-White motorist/officer first punch cell should perceive double jeopardy significantly less than liberals in the Black officer-White motorist/unspecified cell. This prediction follows the logic of the integrated model because here a lower rating by liberals can be attributed to the Black officer throwing the first punch rather than to his race.

A related notion is that the hypothesized decrease in double jeopardy ratings among liberals for a Black officer who throws the first punch should be greater than that for a White Officer who does so. Thus among liberals, Hypothesis 3 states that the independent variables, race combination and first punch, should interact. Such an interaction would indicate that knowledge of the officer throwing the first punch adversely affected liberals' perceptions of the Black officer more so than their perceptions of the White officers.

If it is true that liberals/aversive racists will reveal their prejudice by discriminating against Blacks when such discrimination can be justified, Hypothesis 4 states that when the officer throws the first punch, liberal/Black officer-White motorist participants should give lower double jeopardy ratings than liberal/White officer-White motorist controls. Once more, such a finding would indicate that knowledge of the officer throwing the first punch adversely affected liberals' perceptions of the Black officer more so than their perceptions of the White officer. Theoretically, such discrimination allows liberals/aversive racists the opportunity to vent their intra-psychic conflicts over race, all the while maintaining an egalitarian, non-prejudiced self-image.

Hypotheses Regarding Conservatives. As with liberals, the findings for conservatives in the first punch unspecified cells should replicate those of Nail et al. (2003). Accordingly, Hypothesis 5 states that conservatives in the White officer-Black motorist/unspecified cell should show favoritism to the White officer. Specifically, they should perceive double jeopardy significantly more than conservatives in the Black officer-White motorist/unspecified cell. Since race is being manipulated between participants, conservatives should not know that their prejudice is under surveillance. Thus, they should not be on guard to conceal their prejudice; they should show greater sympathy for a White officer's predicament of being tried twice than for a Black officer's identical predicament.

When it is known that the officer threw the first punch, the integrated model's predictions for liberals and conservatives are similar. That is, both liberals and conservatives should use the Black officer's throwing the first punch as an excuse to discriminate against him (cf. Son Hing et al., in press). Further, it seems reasonable that conservatives, like liberals, should decrease their sympathy for any officer who throws the first punch in a confrontation with a citizen. Hypothesis 6 states, therefore, that conservatives will give lower double jeopardy ratings in the first punch conditions, in general, than in the unspecified conditions. Because conservatives/modern-symbolic racists are hypothesized to be biased against Blacks, however, this decrease should be greater in the Black officer cell than in the White officer cells. Thus, Hypothesis 7 states that among conservatives the independent variables race combination and first punch should interact.

METHOD

Participants

Participants (N = 193) were recruited from a southwestern U.S. college town and its surrounding area. For comparability with the Nail et al. (2003, Studies 1 and 2) groups, which were composed of non-college aged adults, four undergraduate student/experimenters, three females and one male, recruited volunteers from their non-college-aged friends and acquaintances. Because our focus was on the hypothesized racism among White political liberals and conservatives, we eliminated all non-White participants, as well as all politically moderate participants, leaving a sample of 86 for the analyses (41 liberals, 45 conservatives; 51 females, 35 males; Range: 24 to 87 years, Md = 46 years). With the exception of political orientation, participants were randomly assigned to the six between-participants cells of our design: 3 (race combination: White officer-White motorist [control], Black officer-White motorist, or White officer-Black motorist) X 2 (first punch: unspecified vs. the officer). We obtained no main effects or interactions involving gender; thus, gender will not be given further consideration.

Materials

Underneath a snapshot of either a Black or White police officer in uniform, a paragraph described how Officer Bradley Williams had been put on trial twice, first in a state court and acquitted, then in federal court and convicted, for the same offense--assaulting a motorist. Pre-testing demonstrated that the photos were rated equal in handsomeness. The assault had occurred following an argument over a minor traffic violation. The scenario indicated the race of both the officer and the motorist. Each scenario was identical for both the Black and White officer except that an extra sentence was added or not to indicate in the former case that the argument had escalated as a result of Officer Williams throwing the first punch. On the next page, we quoted the double jeopardy clause of the Fifth Amendment, followed by a question assessing the primary dependent variable, double jeopardy: "In your opinion, does the Bradley Williams case violate the spirit or letter of the double jeopardy clause?" (1 = Not at all to 7 = Yes, definitely). A second question was used to as a manipulation check regarding who threw the first punch: "Who do you think was most likely responsible for initiating the assault? (1 = The motorist to 7 = Officer Williams)

The third page was made up of an 8-item political attitudes scale adapted from a similar scale used by Nail et al. (2003, Study 1, Sample 2). Scores on this scale were used to support the validity of a separate indicator of political orientation: participants' self-reported political orientation (liberal, moderate, or conservative), which served as our operational definition for liberal or conservative and appeared at the top of the fourth page. (1)

The eight-item political attitudes scale from page 3 contained a list of four stereotypically conservative targets (Republicans, President George W. Bush, conservatives, and increasing military spending) and four stereotypically liberal targets (The American Civil Liberties Union, liberals, Senator Ted Kennedy, and The National Organization of Women). Participants indicated their attitude toward each target using a scale ranging from -5 (Very unfavorable) to +5 (Very favorable). To create a single composite measure, liberal targets were recoded such that across all eight items negative numbers indicated liberal leanings and positive numbers, conservative leanings. Overall, the political attitudes scale showed good internal consistency ([alpha] = .92). A composite political attitudes score was computed by averaging across the eight items.

Following the self-report item of political orientation on the fourth page, participants reported their age, gender, and ethnicity. Because participants' political attitudes and political orientations were assessed only after they had responded to Officer Williams' case, participants were blind to the fact that political orientation would be a factor in this research. Further, the ordering of the questionnaires assured that the experimenters were also blind as to a participant's political orientation. The fifth and final page was made up of the Modern Racism Scale (MRS; McConahay, 1986) as modified slightly by Nail et al. (2003). The scale revealed good internal consistency with the present sample ([alpha] = .87). It was used to help validate our operational definition of liberals/aversive racists and conservatives/modern racists.

Consistent with previous theory and research (e.g., Lambert & Chasteen, 1997; Nail et al., 2003; Weigel & Howes, 1985), we expected that liberals would score lower on modern racism than conservatives. Since liberals are hypothesized to have a genuinely non-prejudiced self-image, the logic is that they will not endorse any item that might indicate that they are even the least bit racist, even the relatively subtle items of the MRS (e.g., "Discrimination against Blacks is no longer a problem in the United States;" 1 = Strongly disagree to 5 = Strongly agree). Conservatives, in contrast, have less difficulty agreeing with such an item, at least to a moderate degree, as they have been shown to have generally rationalized such beliefs as non-racist (Emerson, Smith, & Sikkink, 1999; Kinder, 1986).

Procedure

An experimenter approached a potential participant to inquire about the possibility of filling out a "survey" to help with a psychology class research project. Individuals were informed that they would be simply listening and responding to a legal case that was real but that "had not received much media attention in our region of the country." The case was modeled on the Rodney King case but was actually fictional. Participants were informed that their responses would be purely opinion-based, that there would be no right or wrong answers, and that they would also be asked to answer some attitudinal and demographic questions. If a person agreed, the experimenter read a sheet of instructions and had the person sign a consent form. The experimenter then read the legal case scenario to the participant, and following, handed him or her a copy of it. Participants were informed that they could refer back to the scenario as many times as they liked while answering any of the remaining questions. Upon completion, participants sealed the entire survey packet in a provided envelope and returned it to the experimenter. Lastly, the researcher debriefed the participant, answered any questions, and thanked him or her for helping.

RESULTS AND DISCUSSION

Operationally Defining Political Orientation

In order to assess the validity of our self report measure of political orientation, the mean composite political attitudes scores of participants who indicated that they were politically liberal (M = -1.8, SD = 1.71) were compared with those who indicated that they were politically conservative (M = 2.4, SD = 1.38). The comparison yielded t(84) = - 12.67, p < .0001, [[eta].sup.2] = .66. Clearly, self-reported liberals differed in their political attitudes from self-reported conservatives. Moreover, liberals scored well below the midpoint of our -5 to +5 scale, and conservatives well above it. The correlation between political attitudes and self-reported political orientation for the entire sample yielded r(192) = .78, p < .0001. Taken as a whole, these findings replicate those of Nail et al. (2003) and support the validity of our operational definition of political orientation--self-reported political orientation.

Do Liberals and Conservatives Differ in Modern Racism?

If there is a link between liberalism and aversive racism on the one hand, and between conservatism and modern racism on the other, as the integrated model proposes, we should find that liberals score lower in modern racism than conservatives. We entered the mean modern racism scores into the 12 cells of our 2 (political orientation: liberal or conservative) X 3 (race combination: White officer-White motorist [control], Black officer-White motorist, or White officer-Black motorist) X 2 (first punch: unspecified vs. the officer) design and conducted an ANOVA. The results revealed a significant effect only for the political orientation main effect, F(1, 73) = 45.05, p < .0001, [[eta].sup.2] = .38. As expected, liberals scored lower in modern racism (M = 1.9, SD = .72) than conservatives (M = 2.93, SD = .70).

Manipulation Check for First Punch

If the manipulation of who threw the first punch was effective, officer-first-punch participants should perceive that Officer Williams threw the first punch relative to the motorist (M = 4.67, SD = 1.83) significantly more so than unspecified-first-punch participants (M = 2.8, SD = 2.03). A 2 X 3 X 2 ANOVA supported this prediction, yielding a significant effect only for the first punch main effect, F(1, 73) = 17.99, p < .0001, [[eta].sup.2] = .20.

[FIGURE 1 OMITTED]

Mean double jeopardy ratings for liberal participants are presented in Figure 1. First, considering only the unspecified first punch cells, did the data replicate the previous findings of Nail et al. (2003, Study 1)? Yes. Consistent with Hypothesis 1, liberal/Black officer-White motorist participants perceived double jeopardy significantly more (M = 6.12, SD = 1.73) than liberal/White officer-Black motorist participants (M = 2.83, SD = 2.23), t(35) = 3.52, p < .003, [[eta].sup.2] = .26. This finding is consistent with the hypothesis from Dovidio and Gaertner's (1998, 2005) integrated model that liberals sometimes show favoritism to minorities as a means of supporting their own non-prejudiced self-image.

We wondered, however, whether the liberal bias in favor of the Black officer would hold not only relative to White officer-Black motorist/unspecified participants but also relative to White officer-White motorist/unspecified control participants. To find out, we tested between liberal/Black officer-White motorist/unspecified participants (M = 6.12, SD = 1.73) and liberal/White officer-White motorist/unspecified participants (M = 4.0, SD = 2.0). The contrast yielded t(35) = 2.36, p < .03, [[eta].sup.2] = 14. This finding indicates that the liberals' bias in favor of the Black officer was significant even when race and racial conflict were not salient in the comparison group (viz., the White officer-White motorist control group). Accordingly, this bias can be attributed to the Black officer alone and not to a combination of bias (a) in favor of a Black officer and (b) against a White officer given a Black motorist/victim.

The focal hypotheses for liberals concern whether their bias in favor of the Black officer will reverse once it is known that the Black officer threw the first punch. Hypothesis 2 states that liberals in the Black officer-White motorist/officer first punch cell should perceive double jeopardy (M = 2.57, SD = 1.62) significantly less than liberals in the Black officer-White motorist/unspecified cell (M = 6.12, SD = 1.73). Hypothesis 2 was supported, t(35) = 3.96, p < .0001, [[eta].sup.2] = .31.

In the same vein, Hypothesis 3 states that the independent variables, race combination and first punch, should interact. Hypothesis 3 was supported, F(2, 35) = 3.52, p < .05, [[eta].sup.2] = .17. We explored this interaction by conducting tests for simple main effects. Considering the simple main effects for first punch, the analyses revealed that the interaction between race combination and first punch was due almost totally to the difference between the Black officer-White motorist/unspecified and the officer-first-punch conditions. Consistent with the a priori planned contrast reported above in connection with Hypothesis 2, Black officer-White motorist/officer first punch participants (M = 2.57, SD = 1.62) reported lower double jeopardy ratings than Black officer-White motorist/unspecified participants (M = 6.12, SD = 1.73), F(1, 35) = 15.71, p < .001, [[eta].sup.2] = .31. However, the difference in the ratings of White officer-Black motorist/officer first punch participants (M = 1.17, SD = .41) as compared to White officer-Black motorist/unspecified participants was only marginally significant (M = 2.83, SD = 2.23), F(1, 35) = 2.78, p < .11, [[eta].sup.2] = .07. Further, there was no difference between White officer-White motorist/officer first punch participants (M = 3.86, SD = 1.77) and White officer-White motorist/unspecified participants (M = 4.0, SD = 2.0), F < 1.

Taken together, the simple main effects for first punch reveal among liberals that it was only the double jeopardy ratings of the Black officer that were significantly lowered by knowledge that the first punch had been thrown by the officer. This is precisely the overall pattern that would be expected, consistent with Dovidio and Gaertner's (1998) integrated model. It predicts that liberals/aversive racists (a) sometimes show favoritism to minorities in order to support their non-prejudiced self-image but (b) sometimes discriminate against minorities as a direct expression of their automatic, negative race-based feelings. (2)

Hypothesis 4 states that when the officer throws the first punch, liberal/Black officer-White motorist participants should give lower double jeopardy ratings (M = 2.57, SD = 1.62) than liberal/White officer-White motorist controls (M = 3.86, SD = 1.77). Although in the direction predicted by the integrated model, this contrast failed to reach the traditional .05 level, t(35) = -1.39, p < .09, [[eta].sup.2] = .05.

To summarize briefly, 3 of 4 hypotheses regarding liberals that were derived from Dovidio and Gaertner's (1998, 2005) integrated model of racism were supported in the present study. The difference predicted in Hypothesis 4 was in the expected direction but fell short of the conventional .05 level.

Double Jeopardy Findings for Conservatives

Mean double jeopardy ratings for conservative participants are presented in Figure 2. First, considering only the unspecified first punch cells, did the data replicate the previous findings of Nail et al. (2003, Study 1)? No. In contrast to Hypothesis 5 and the integrated model, conservatives in the White officer-Black motorist/unspecified cell did not perceive double jeopardy significantly more (M = 5.0, SD = 2.27) than conservatives in the Black officer-White motorist/unspecified cell (M = 4.29, SD = 2.56), t < 1. The difference was in the predicted direction but fell far short of statistical significance. Given that (a) the participants were recruited from the same general population as the Nail et al. (2003) participants, (b) in these conditions we replicated the Nail et al. methodology exactly, and (c) we did replicate the Nail et al. results with regard to liberals, this failure to replicate among conservatives is puzzling.

Given this failure, we reexamined Hypothesis 5, but this time only with respect to those in our original sample who scored in the top 1/3 on the MRS. Our reasoning was that conservatism is only an indirect indication of modern racism relative to actual scores on the MRS. Thus, employing MRS scores should provide a stronger, more direct test of Hypothesis 5. Nevertheless, we still found only marginal support for the hypothesis that White modern racists will discriminate against Blacks if race is manipulated between participants. Among those scoring in the top 1/3 of the MRS, White officer-Black motorist/unspecified participants perceived double jeopardy only somewhat higher (M = 5.57, SD = 1.4) than Black officer-White motorist/unspecified participants (M = 3.86, SD = 2.12), t(30) = 1.64, p < .06, [[eta].sup.2] = .08.

Hypothesis 6 states that conservatives will give lower double jeopardy ratings in the officer-first-punch conditions, in general, than in the unspecified conditions. Yet, because conservatives/modern-symbolic racists are hypothesized to be generally biased against Blacks, we expected that this decrease would be greater in the Black officer cell than in the White officer cells. Thus, Hypothesis 7 states that the independent variables race combination and first punch should interact. Inspection of Figure 2 reveals, however, that neither of these hypotheses was supported. Indeed, knowledge that the officer had thrown the first punch, if anything, appears to have increased the double jeopardy ratings among conservatives (see Figure 2).

[FIGURE 2 OMITTED]

To explore for any non-hypothesized effects, we examined the data in Figure 2 in a 3 (race combination: White officer-White motorist [control], Black officer-White motorist, or White officer-Black motorist) X 2 (first punch: unspecified vs. the officer) ANOVA. The analysis revealed no significant main effects or interactions (all F's < 1). Given these findings, we repeated this analysis, but only among those participants who scored in the top 1/3 of our sample in MRS. Again, the analysis yielded no significant main effects or interactions (all F's < 1). Given the supportive evidence for the hypothesis that White conservatives will discriminate against Blacks when race is manipulated between-participants that has been reported in past research by Gaertner (1973) and Nail et al. (2003) and other evidence linking conservatism with modern-symbolic racism (e.g., Lambert & Chasteen, 1997; Sears & Henry, 2003; Son Hing et al., in press; Weigel & Howes, 1985), the present overall results for conservatives are puzzling.

In this regard it is important to reiterate, however, that Sniderman et al. (1991) also found that White conservatives did not generally discriminate against Blacks. Specifically, it will be recalled that slightly more conservatives reported that an undependable laid-off Black worker was entitled to at least some government assistance (69%) as compared to an undependable laid-off White worker (63%), this despite the fact that the label "undependable" should have primed negative minority group stereotypes and given conservatives a good excuse to discriminate. What is more, the percentage of conservatives endorsing help for a dependable laid-off worker was significantly greater for a Black (81%) than for a White (59%). Similar results were reported by McConahay (1983) who found that those scoring high on the MRS also discriminated in favor of a Black target over a White under some conditions.

Considering everything, it appears that the simple notion that White North American conservatives, because of their fundamental prejudice, will generally discriminate against Blacks when they do not know that their potential racism is under surveillance is not supported by the weight of the evidence. It is possible that any particular sample of conservatives might be non-prejudiced or at least low in measurable prejudice. For example, the Sniderman et al. (1991) study was based on a random sample from the San Francisco--Oakland Bay Area in 1986. The researchers acknowledged that "... one cannot leave out the South and still tell the full story of race as an issue in U.S. politics" (p. 443). Further, it appears that situational factors can come into play that can derail or even reverse any prejudice that any particular sample of conservatives might possess. For example, Blacks who seem to be conforming to White, middleclass values such as being a dependable worker may be evaluated more positively than matched Whites, as Sniderman et al. (1991) found. Similar to Dovidio and Gaertner's (1998, 2005) theorizing among liberals, conservatives might be able to use such positive evaluations of Blacks to prove to themselves that they are not really prejudiced at all. This issue could be fertile ground for future research.

Double Jeopardy Findings Considering Liberals and Conservatives Together

Given our predictions, in the analyses to this point we have considered liberals and conservatives separately. Nevertheless, there are two contrasts between liberals and conservatives that would be expected to be different based on Dovidio and Gaertner's (1998, 2005) integrated model. The expected differences in both cases follow from the integrated model's hypothesized biases either for or against Blacks by liberals and conservatives, respectively. Both contrasts apply to unspecified participants only. The first expectation is that liberal/Black officer-White motorist participants should have higher double jeopardy ratings (M = 6.12, SD = 1.73) than conservative/Black officer-White motorist participants (M = 4.29, SD = 2.56). This prediction was supported, t(74) = 1.83, p < .04, [[eta].sup.2] = .04. Yet, the means in the unspecified/White officer-White motorist control groups for liberals (M = 4.0, SD = 2.0) and conservatives (M = 4.4, SD = 2.3) both indicate that this effect was driven completely by the bias among liberals in favor of the Black officer to the exclusion of the hypothesized bias among conservatives in favor of the White officer.

The second expectation is that liberal/White officer-Black motorist participants should have lower double jeopardy ratings (M = 2.83, SD = 2.23) than conservative/White officer-Black motorist participants (M = 5.0, SD = 2.27). This hypothesis, too, was supported, t(74) = 2.07, p < .03, [[eta].sup.2] = .05. Again, comparison with the unspecified/White officer-White motorist control groups for liberals (M = 4.0, SD = 2.0) and conservatives (M = 4.4, SD = 2.3) reveals that this effect was primarily driven by bias among liberals against a White officer who assaults a Black motorist.

To explore the overall data for any additional effects, we submitted the double jeopardy scores to the full 2 X 3 X 2 ANOVA of our design. We restrict our presentation here only to effects not redundant with the analyses reported above. The ANOVA yielded a significant main effect for political orientation, F(1, 74) = 7.72, p < .007, [[eta].sup.2] = .09. Consistent with stereotypic views, conservatives in general showed significantly greater support for the police officer in terms of higher double jeopardy ratings (M = 4.67, SD = 2.02) than liberals (M = 3.56, SD = 2.26). The ANOVA also yielded a significant main effect for first punch, F(1, 74) = 4.0, p < .05, [[eta].sup.2] = .05. Unspecified first punch participants in general gave significantly higher double jeopardy ratings (M = 4.53, SD = 2.28) than officer first punch participants (M = 3.78, SD = 2.08). Thus, quite reasonably enough, participants were more sympathetic to the police officer if it was not specified who threw the first punch than if it was known that the officer had done so.

Both the political orientation and officer first punch main effects, however, must be interpreted in light of a significant political orientation X first punch interaction, F(1, 74) = 4.87, p < .03, [[eta].sup.2] = .06 (see Figure 3). Tests for simple main effects revealed that this interaction was due primarily to the low double jeopardy ratings given by participants in the liberal/officer first punch condition (M = 2.53). These participants gave lower ratings than both liberal/unspecified first punch participants (M = 4.32), F(1, 74) = 8.65, p < .005, [[eta].sup.2] = .10, and conservative/officer first punch participants (M = 4.65), F(1, 74) = 12.98, p < .001, [[eta].sup.2] = .15. Neither of these groups differed, however from conservative/unspecified first punch participants (M = 4.56), both F's < 1.3

[FIGURE 3 OMITTED]

From the perspective of the integrated model, we did not anticipate that conservatives would be so uninfluenced by information that the police officer had thrown the first punch, be he White or Black (see Figure 2). We did not anticipate that conservative/officer first punch participants (M = 4.65) would give double jeopardy ratings so much higher on average than those of liberal/officer first punch participants (M = 2.53). Yet, considering the association between conservatism and rightwing authoritarianism that has been reported in numerous studies (e.g., Altemeyer, 1988; 1998; Christie, 1991), we might have predicted that conservatives would be more sympathetic than liberals to the plight of a twice-tried police officer.

A number of studies have shown that participants high in right-wing authoritarianism tend to support members of official government agencies and the police even when they violate the law (e.g., Altemeyer, 1996; Feather, 1998; Heaven & Bucci, 2001). If we assume that the present conservatives were also relatively high in right-wing authoritarianism, this might help explain the present overall main effect for political orientation and especially the significant political orientation X first punch interaction presented above (see Figure 3). In retrospect, it makes a good deal of sense that conservatives/high right-wing authoritarians relative to liberals/low right-wing authoritarians might be quick to defend a police officer tried twice for a single offense, assaulting a motorist, even if the officer threw the first punch. The link between conservatism and right-wing authoritarianism alone, however, can not explain our failure to replicate the Nail et al. (2003) results in our conservative/unspecified first punch cells.

It might be important in this regard to note that the Nail et al. (2003) data were collected before the Al Qaeda attacks on the United States on 9/11/01, whereas the present data were collected after 9/11/01, about one year after the attacks. Yet, the difference in double jeopardy ratings among conservatives in the Black officer-White motorist cells between Nail et al. (2003, Study 1, Sample 1) and the present study is striking (M = 1.25 vs. M = 4.29, respectively). Might the 9/11/01 attacks have induced a sense of threat in the present conservatives, a threat both to the security of the United States as a whole and to its individual citizens? If so, might this threat have triggered support among conservatives in response for any entity that symbolizes traditional American values, for any legitimate authority figure such as a police officer? Might this support have extended to police officers, in general, even to a Black police officer who assaulted a White motorist, even when he threw the first punch (see Figure 2)?

Yet, if the hypothesized threat induced in the nation in the aftermath of 9/11/01 caused the present conservatives to be hypersensitive to threat and to support a police officer in response, and if these events caused us, in turn, to fail to replicate the Nail et al. (2003) findings with respect to conservatives, why did this threat not have similar consequences with respect to the present liberals? Why were we able to replicate the Nail et al. (2003) findings with respect to liberals but not conservatives? Perhaps it is because conservatives as a group may be more sensitive to threats to their worldview and therefore more defensive than liberals (e.g., Greenberg et al., 1990, Study 2; Greenberg, Simon, Pyszczynski, Solomon, & Chatel, 1990, Study 1).

Pyszczynski et al. (2006, Study 2) found that conservatives responded with greater defensiveness than liberals to the specific threat of being reminded of 9/11/01. Future research that measures political orientation and right-wing authoritarianism and manipulates such variables as terrorism salience, race salience, fairness salience, and excuse-to-discriminate should be valuable in determining the conditions under which White conservatives and liberals will show bias with respect to Blacks and other minorities, be it positive or negative.

GENERAL DISCUSSION

The present research obtained support for Dovidio and Gaertner's (1998, 2005) integrated model of racism with respect to political liberals but not conservatives. We replicated the findings of Nail et al. (2003) that liberals will show bias in favor of a twice-prosecuted Black officer given a White victim, relative to a White officer given a Black victim. Furthermore, this bias in favor of a Black officer essentially reversed once liberals learned that the Black officer had initiated the assault on the motorist by throwing the first punch (Figure 1). Clearly, liberals were influenced by both the race of the officer and victim and by the circumstances under which the officer was tried twice for the same offense. Contrary to predictions, conservatives, in contrast, were remarkably uninfluenced by these same variables.

The overall pattern of results for liberals and conservatives does offer support for the integrated model in one way. The model holds that liberals/aversive racists, unlike conservatives/modern racists, experience significant intra-psychic conflict over race. If true, then one might expect that the behavior of liberals/aversive racists would be more highly variable and capricious across different race-related situations than that of conservatives/modern-symbolic racists (see also Katz, 1981; Katz, Cohen, & Glass, 1975).

Given greater underlying conflict and uncertainty, it follows that subsequent related behaviors would be more subject to change and variability (Brockner, 1983; Steele, Spencer, & Lynch, 1993). Visual inspection of Figures 1 and 2 regarding differences in liberals and conservatives, respectively, provides support for this proposition. As can be seen, there are wide differences in the behavior of liberals as a function of race combination and first punch, but essentially no differences for conservatives. The finding of a significant race combination X first punch interaction among liberals but not conservatives quantifies and gives statistical support to the differences between liberals and conservatives depicted in Figures 1 and 2 that are so apparent to the eye.

One limitation of this research and of prejudice theory and research in general is that there is no generally accepted self-report measure of aversive racism. To elaborate, modern racism can be assessed with the MRS (McConahay, 1986), and herein we defined (a) a modern-symbolic racist as a self-reported conservative who scored relatively high on the MRS and (b) an aversive racist as self-reported liberal who scored relatively low on the MRS. Still, there is no scale that measures aversive racism more directly. Gaertner and Dovidio (1986) have even questioned whether the development of a conventional aversive racism scale is possible. Research by Son Hing et al. (2002, 2005), however, has made progress toward addressing this issue.

Given the position of Dovidio and Gaertner (1998, 2005) that aversive racists hold conscious positive beliefs and attitudes toward Blacks and other minorities but at the same time negative non-conscious feelings, Son Hing et al. have proposed that aversive racism might be accurately defined operationally by low explicit racism coupled with high implicit racism. Explicit racism can be assessed by self-report measures such as the MRS, implicit racism by race-related word completions (Son Hing et al., 2002; Steele & Aronson, 1995) or response latency associations to race-related stimuli (e.g., faces, positive or negative labels; e.g., Dovidio & Gaertner, 1991; Fazio, Jackson, Dunton, & Williams, 1995; Rudman, 2004; Son Hing et al., 2005). By whatever means implicit attitudes are measured, the assumption is that implicit racial attitudes represent fragments in memory from past experiences with minorities. Further, these fragments can influence behavior quite apart from one's conscious awareness or self-reports.

Supporting the validity of this conceptualization, Son Hing et al. (2002) reported that low explicit/high implicit racism individuals (i.e., aversive racists) were very likely to compensate by showing favoritism to a minority student organization once memories regarding their own past undesired racist behavior toward members of this minority group had been primed. In contrast, those scoring low in both explicit and implicit racism (i.e., the truly non-prejudiced) showed no such tendency. More recently, Son Hing et al. (in press) have extended the Son Hing et al. (2002) logic regarding explicit and implicit attitudes to modern racists--defined as those scoring high in both explicit and implicit racism. Using response latencies as the measure of implicit racism, Son Hing et al. (in press) found that modern racists and aversive racists alike were more likely to discriminate against a minority job applicant whose qualifications were mixed and therefore ambiguous relative to truly non-prejudiced individuals (Studies 2 and 3, respectively). Especially pertinent to the present investigation and analysis, in Study 1 Son Hing et al. (in press) showed that aversive racists scored very low on a measure of political conservatism, whereas modern racists scored relatively high.

Researchers interested in testing tenets of Dovidio and Gaertner's (1998, 2005) integrated model, as well as other contemporary models of racism, would probably be well-advised to consider in their research design and methodology the work of Son Hing and colleagues. Such research could be pivotal in establishing limit and boundary conditions of the models.

REFERENCES

Altemeyer, B. (1988). Enemies of freedom: Understanding right-wing authoritarianism. San Francisco, CA: Jossey-Bass.

Altemeyer, B. (1996). The authoritarian specter. Cambridge, MA: Harvard University Press.

Altemeyer, B. (1998). The other "authoritarian personality". In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 30, pp. 47-92). San Diego: Academic Press.

Brockner, J. (1983). Low self-esteem and behavioral plasticity: Some implications. In L. Wheeler & P. Shaver (Eds.), Review of personality and social psychology (Vol. 4, pp. 237-271). Beverly Hills, CA: Sage.

Christie, R. (1991). Authoritarianism and related constructs. In J. P. Robinson, P. R. Shaver, and L. S. Wrightsman (Eds.), Measures of personality and social psychological attitudes (pp. 501-571). San Diego, CA: Academic Press.

Dovidio, J. F., Evans, N., & Tyler, R. B. (1986). Racial stereotypes: The contents of their cognitive representations. Journal of Experimental Social Psychology, 22, 22-37.

Dovidio, J. F., & Gaertner, S. L. (1991). Changes in the expression of racial prejudice. In H. Knopke, J. Norrell, & R. Rogers (Eds.) Opening doors: An appraisal of race relations in contemporary America (pp. 201-241). Tuscaloosa: University of Alabama Press.

Dovidio, J. F., & Gaertner, S. L. (1998). On the nature of contemporary prejudice: The causes, consequences, and challenges of aversive racism. In J. L. Eberhardt & S. T. Fiske (Eds.), Confronting racism: The problem and the response (pp. 3-32). Thousand Oaks, CA: Sage.

Dovidio, J. F., & Gaertner, S. L. (2005). Aversive racism. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 36, pp. 1-52). San Diego, CA: Academic Press.

Emerson, M. O., Smith, C., & Sikkink, D. (1999). Equal in Christ, but not in the world: White conservative Protestants and explanations of black-white inequity. Social Problems, 46, 398-417.

Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995). Variability in automatic activation as an unobtrusive measure of racial attitudes: A bona fide pipeline? Journal of Personality and Social Psychology, 69, 1013-1027.

Feather, N. T. (1998). Reaction to penalties for offenses committed by the police and public citizens: Testing a social-cognitive process model of retributive justice. Journal of Personality and Social Psychology, 75, 528-544.

Fein, S., Morgan, S. J., Norton, M. I., & Sommers, S. R. (1997). Hype and suspicion: The effects of pretrial publicity, race, and suspicion on jurors' verdicts. Journal of Social Issues, 53, 487-502.

Frey, D. L., & Gaertner, S. L. (1986). Helping and the avoidance of inappropriate interracial behavior: A strategy that perpetuates a nonprejudiced self-image. Journal of Personality and Social Psychology, 50, 1083-1090.

Gaertner, S. L. (1973). Helping behavior and racial discrimination among liberals and conservatives. Journal of Personality and Social Psychology, 25, 335-341.

Greenberg, J., Pyszcaynski, T., Solomon, S., Rosenblatt, A., Veeder, M., Kirkland, S., & Lyon, D. (1990). Evidence for terror management theory II: The effects of mortality salience on reactions to those who threaten or bolster the cultural worldview. Journal of Personality and Social Psychology, 58, 308-318.

Greenberg, J., Simon, L., Pyszcaynski, T., Solomon, S., & Chatel, D. (1992). Terror management and tolerance: Does mortality salience always intensify negative reactions to others who threaten one's worldview? Journal of Personality and Social Psychology, 63, 212-220.

Heaven, P. C. L., & Bucci, S. (2001). Right-wing authoritarianism, social dominance orientation and personality: An analysis using the IPIP measure. European Journal of Personality, 15, 49-56.

Katz, I. (1981). Stigma: A social psychological analysis. Hillsdale, NJ: Erlbaum.

Katz, I., Cohen, S., & Glass, D. (1975). Some determinants of cross-racial helping behavior. Journal of Personality and Social Psychology, 32, 964-970.

Kinder, D. R. (1986). The continuing American dilemma: White resistance to racial change 40 years after Myrdal. Journal of Social Issues, 42, 151-171.

Lambert, A. J., & Chasteen, A. L. (1997). Perceptions of disadvantage versus conventionality: Political values and attitudes toward the elderly versus African- Americans. Personality and Social Psychology Bulletin, 23, 469-481.

McConahay, J. B. (1983). Modern racism and modern discrimination: The effects of race, racial attitudes, and context on simulated hiring decisions. Personality and Social Psychology Bulletin, 9, 551-558.

McConahay, J. B. (1986). Modern racism, ambivalence, and the modern racism scale. In J. F. Dovidio & S. L. Gaertner (Eds.), Prejudice, discrimination and racism (pp. 91-125). Orlando, FL: Academic Press.

Nail, P. R., Harton, H. C., & Decker, B. P. (2003). Political orientation and modern versus aversive racism: Tests of Dovidio and Gaertner's (1998) integrated model. Journal of Personality and Social Psychology, 84, 754-770.

Pyszczynski, T., Abdollah, A., Solomon, S., Greenberg, J., Cohen, F., & Weise, D. (2006). Mortality salience, martyrdom and military might: The great Satan versus the axis of evil. Personality and Social Psychology Bulletin, 32, 525-537.

Rudman, L. A. (2004). Social justice in our minds, homes, and society: The nature, causes, and consequences of implicit bias. Journal of Social Issues, 17, 129-142.

Sears, D. O. (1998). Racism and politics in the United States. In J. L. Eberhardt & S. T. Fiske (Eds.), Confronting racism: The problem and the response (pp. 76-100). Thousand Oaks, CA: Sage.

Sears, D. O., & Henry, P. J. (2003). The origins of symbolic racism. Journal of Personality and Social Psychology, 85, 259-275.

Sniderman, P. M., Piazza, T., Tetlock, P. E., & Kendrick, A. (1991). The new racism. American Journal of Political Science, 35, 423-447.

Son Hing, L. S., Chung-Yan, G. A., Grunfeld R., Robichaud, L., & Zanna, M. P. (2005). Exploring the discrepancy between implicit and explicit prejudice: A test of aversive racism theory. In J. P. Forgas, K. Williams, & S. Latham (Eds.), Social Motivation, Conscious and Unconscious Processes (pp. 275-293). New York, NY: Psychology Press.

Son Hing, L. S., Chung-Yan, G. A., Hamilton, L. K., & Zanna, M. P. (in press). A two-dimensional model that employs explicit and implicit attitudes to characterize prejudice. Journal of Personality and Social Psychology.

Son Hing, L. S., Li, W., & Zanna, M. P. (2002). Inducing hypocrisy to reduce prejudicial responses among aversive racists. Journal of Experimental Social Psychology, 38, 71-78.

Steele, C. M., & Aronson, J. (1995). Stereotype threat and the intellectual test performance of African Americans. Journal of Personality and Social Psychology, 69, 797-811.

Steele, C. M., Spencer, S. J., & Lynch, M. (1993). Self-image resilience and dissonance: The role of affirmational resources. Journal of Personality and Social Psychology, 64, 885-896.

Weigel, R. H., & Howes, P. W. (1985). Conceptions of racial prejudice: Symbolic racism reconsidered. Journal of Social Issues, 41, 117-138.

Footnotes

(1) We employed the single-item self-report measure as our operational definition of political orientation for comparability with past research (e.g., Gaertner, 1973; Nail et al., 2003; Sniderman et al., 1991). Defining liberal versus conservative with a tripartite split of our political attitudes scale yielded results that were almost identical with those reported herein and would have required no changes in our interpretations or conclusions.

(2) Considering the simple main effects for race combination, exploration of the interaction revealed a significant effect when considering only unspecified participants, F(2, 35) = 6.59, p < .01, [[eta].sup.2] = .27. Post hoc analysis with the Games-Howell procedure revealed once more, consistent with the a priori planned contrast reported above in relation to Hypothesis 1, that Black officer-White motorist/unspecified participants perceived double jeopardy significantly more (M = 6.12, SD = 1.73) than White officer-Black motorist/unspecified participants (M = 2.83, SD = 2.23), p < .04. White officer-White motorist/unspecified control participants (M = 4.0, SD = 2.0), however, did not differ significantly from either of these two more extreme groups (both p's > .11). Together, these findings indicate that liberal/unspecified participants were neither biased in favor of a Black officer, nor biased against a White officer, per se. Rather, their bias was only apparent when the comparison was between officers and motorists both of different races (viz., the difference between Black officer-White motorist participants and White officer-Black motorist participants), a condition that presumably increases the saliency of race as compared to the White officer-White motorist condition. This finding supports the Dovidio and Gaertner (1998, 2005) hypothesis that liberals/aversive racists may discriminate in favor of or against Blacks only when race is salient.

Further exploration of the simple main effects for race combination also revealed a significant effect when considering only officer first punch participants, F(2, 35) = 3.9, p < .05, [[eta].sup.2] = .18. Post hoc analysis with the Games-Howell procedure revealed that it was only White officer-Black motorist participants (M = 1.17, SD = .41) who gave significantly lower ratings than White officer-White motorist control participants (M = 3.86, SD = 1.77), p < .02. Black officer-White motorist participants (M = 2.57, SD = 1.62) did not differ significantly from either of these two more extreme groups (both p's > .14).

(3) The three-way interaction, 2 (political orientation) X 3 (race combination) X 2 (first punch), yielded: F(2, 74) = 2.18, p = .12, [[eta].sup.2] = .06.

Author Note We thank Seth Miller, Courtney Phillips, and Katrina Bedell for helping with data collection and Seth Miller for helping with data entry. Special appreciation goes to Leanne Son Hing for her comments on a previous draft.

Paul R. Nail

University of Central Arkansas

Helen C. Harton

University of Northern Iowa

Anna Barnes

Southwestern Oklahoma State University

Correspondence should be sent to: Dr. Paul Nail, Dept. of Psychology & Counseling, University of Central Arkansas, 201 Donaghey, Conway, AR 72035.
Gale Copyright:
Copyright 2008 Gale, Cengage Learning. All rights reserved.