LIBRARY Michigan State * University PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE ”you“ ‘L \r : LA ’1:r“‘1 {“4 J __ L— ”TV—'1 MSU Is An Affirmative Action/Equal Opportunity Institution TWO MODELS OF THE MANIFESTATION OF RACIAL AMBIVALENCE IN WHITE COLLEGE STUDENTS: AVERSIVE RACISM AND PREJUDICE WITH COMPUNCTION By Bradley J.M. Hack A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1993 ABSTRACT TWO MODELS OF THE MANIFESTATION OF RACIAL AMBIVALENCE IN WHITE COLLEGE STUDENTS: AVERSIVE RACISM AND PREJUDICE WITH COMPUNCTION By Bradley J.M. Hack Two perspectives of how Whites manifest racial ambivalence toward African-Americans were investigated. Aversive Racist theory (Gaertner & Dovidio, 1986) posits that Whites are not typically aware that they have conflicting egalitarian and prejudiced beliefs toward African—Americans and that they engage in self-deception to camouflage prejudiced responses. Consequently they are unaware that they respond in a prejudiced manner. Prejudice with Compunction theory (Devine, Monteith, Zuwerwink, & Elliott, 1991) posits that whites gm typically aware of their ambivalence toward African- Americans and experience compunction when they respond in a prejudiced way. In Study 1, subjects gave self-reports of their standards for how they should respond and how they would respond in contact situations with African-Americans or Whites. Half were given an opportunity for self-deception, while the other half were not. Discrepancies between their responses were used as a measure of subjects' racial ambivalence. Results did not support the Aversive Racism perspective but remained somewhat consistent with the Prejudice with Compunction model. In Study 2, subjects gave self- reports of should standards but based would standards on a live interaction with either an African-American or White confederate. Results from Study 2 replicated findings from Study 1. ACKNOWLEDGEMENTS I would like to thank my committee chairperson, Galen Bodenhausen for his wisdom and guidance throughout the research process. His help has been an invaluable asset to me. I would like to thank committee members Norman Abeles and Linda Jackson for their encouragement and. assistance during this project. I would also like to thank Jackie Davis and Abigail Patchan for their assistance in conducting the experiments. A special thanks goes to Cassandra Vogel who has supported and encouraged me on this project from the beginning. Finally, I would like to thank my parents, Stephen and Lorraine Hack, and my brother, Michail Hack, for the opportunities they have provided for me throughout my life. TABLE OF CONTENTS 1. INTRODUCTION Aversive Racism Prejudice with Compunction Goals of Present Research Low Self-Insight Hypotheses High Self-Insight Hypotheses Predictions 2. EXPERIMENT 1 Method Results Discussion 3. EXPERIMENT 2 Introduction Goals of the Second Study Prediction-s Method Results Discussion 4. REFERENCES 5. APPENDICES Appendix A: Should and Would Scenarios Appendix 8: Interview Questions NOLOO‘i-Ii-J —I—I Z1 31 35 37 38 39 48 57 64 67 67 LIST OF TABLES Name gf Table Bags TABLE 1: Should Scores 68 TABLE 2: Would Scores 68 TABLE 3: . Perceived Awareness of a Discrepancy 69 TABLE 4: Actual Discrepancy Scores 1 69 TABLE 5: Affect Scores 70 TABLE 6: Should and Would Scores 71 TABLE 7: Should and Would Scores by Level of Prejudice 72 TABLE 8: Perceived Awareness of a Discrepancy 72 TABLE 9: Affect Scores 73 1 Racial prejudice has been a domain of interest to psychologists dating back to the pioneering work of Katz and Braly in the early 1930's. Since this work, there has been quite a history in the prejudice literature of researchers acknowledging the coexistence, in the same individual, of the conflicting thought structures of ' stereotypical beliefs and non-prejudiced beliefs about African- American people (Allport, 1954; Gaertner & Dovidio, 1986; Higgins & King, 1981; Katz, Wackenhut, & Haas, 1986;. McConahay, 1986; Myrdal, 1944). Myrdal (1944) posited that low-prejudice persons experience internal conflict between the egalitarian notions of the "American Creed" and their more specific prejudices. Allport (1954) further asserted that even though some low-prejudice people have consciously rejected prejudiced beliefs, they still react to African- Americans in a prejudiced or intolerant manner. More recently, researchers have reiterated this notion by proposing that people are socialized in the American culture to have negative beliefs and feelings about African-Americans (Devine, 1989; Erhlich, 1973; Gaertner & Dovidio, 1986; P. Katz, 1976). At the same time, however, these people also strongly identify with the egalitarian values America's Constitution propounds and endorse non-prejudiced attitudes. Consequently, low-prejudice persons experience a dilemma about what to believe and how to behave toward African- Americans. WM One prominent, recent theory is the "Aversive Racist" theory" (Gaertner & Dovidio, 1986). Gaertner and Dovidio describe aversive 2 racists as people who sympathize with the victims of past injustices, support policies aimed at promoting racial equality, and regard themselves as non-prejudiced. They state that such people are typically subjects scoring low on conscious measures of prejudice but who may also reveal discriminatory behavior on unobtrusive measures of prejudice. In addition, aversive racists usually express their egalitarian views and not their prejudiced attitudes when interacting with others. If the norms prescribing non-prejudiced behavior in an interracial interaction are not salient to aversive racists, however, their negative ideas and feelings about African-Americans may be expressed. This expression of negative ideas mayalso occur in an interracial interaction even if the norms for how to behave in a non- prejudiced manner are salient. Here, aversive racists may use a non-racial justification factor (NRJF) to express negative attitudes. They may Channel their negative thoughts through the non-racial factor by making it the salient issue instead of race. This also allows them to maintain their egalitarian self-image because it keeps their negative ideas about African-American people out of awareness. A study by Dovidio and Gaertner (1981), highlights how aversive racists might use a NRJF to camouflage prejudiced behavior. In this study, the prosocial behavior of subjects toward either an African- American or White confederate was measured. Subjects were told that they would be working on an anagram task with. another student. The confederates were either introduced as supervisors or subordinates to the subjects. In addition, the subjects were told 3 that the confederates were either of higher or lower ability than the subjects. Prior to the anagram task, the confederates knocked over a can of pencils and the number of pencils that the subjects picked up were used as a measure of prosocial behavior. The results revealed that relative status and not relative ability was the primary determinant of subjects' helping behavior toward African- American confederates. African-American supervisors were helped less than African-American subordinates, regardless of ability level, while White supervisors were helped more than White subordinates. In postexperimental evaluations of their partners, however, subjects rated both low and high ability African-American confederates as less intelligent than themselves but rated high ability White partners as somewhat more intelligent than themselves. Here it seems White subjects rationalize their negative response to African-American supervisors by evaluating them as less competent or intelligent than themselves. In this case, Gaertner and Dovidio would label the subjects as aversive racists because they assume the African-American person possesses less ability for the task. A negative assumption such as this is, however, in conflict with aversive racists' non-prejudiced personal beliefs. To cope with this ambivalence toward African- American people, they have made ability, not race, the salient issue in an attempt to exclude this negative view from awareness. Lacking aWareness of this negative reaction, in turn, also allows aversive racists to maintain their egalitarian self-image. Consequently, aversive racists are unaware of their conflicted beliefs and express prejudice in subtle and covert ways by using 4 non-racial justification factors that protect their egalitarian self- image. Prejudigg with nggunctign Devine (1989) has taken an information-processing approach to sort out the existence of the internal conflict low-prejudice subjects experience. She has shown that both automatic and controlled processes of prejudice operate differently for Stereotypes and non-prejudiced personal beliefs. Specifically,she posits that prejudice-like responses are automatically activated in persons in the presence of a member of the stereotyped group. Non- prejudiced responses require the inhibition of these automatically activated negative responses and also the intentional activation of the non-prejudiced beliefs. When these thoughts and feelings are not inhibited, or the activation of non-prejudiced beliefs prevented, then prejudiced responses may be produced on the unobtrusive measures. She states clearly that the adoption of non-prejudiced beliefs does not lead to an all-or—none change from prejudice to non-prejudice. Rather it is a slow process, as the activation of stereotypes can continue to produce prejudiced thoughts and feelings which are in conflict with the non-prejudiced personal beliefs. In addition, Devine, Monteith, Zuwerink, & Elliott (1991) have shown that as long as controlled processes are allowed to operate, most people are aware of both their prejudiced and non-prejudiced thoughts and feelings. In addition, people showing a discrepancy between the way they feel they "should" respond to African- 5 Americans (non-prejudiced beliefs) and the way they actually "would" respond (stereotype-consistent behavior), reveal feelings of compunction. The low-prejudice individuals with this discrepancy express global discomfort, self-criticism, and guilt while high- prejudice subjects experience only global discomfort. This is consistent with Higgins' (1987) Self-Discrepancy theory in which peeple are said to experience self-contempt and guilt if they experience a discrepancy between their actual selves ("would" response) and their ought selves ("should" response). Devine's work shows that low-prejudice subjects are apparently aware of their conflicted feelings and attitudes, and many show feelings of ‘guilt over still possessing prejudiced attitudes (Devine et al., 1991). Upon examination of these two theories, it is apparent that there is a theoretical dispute in the prejudice literature. Gaertner and Dovidio (1986) and Devine et al. (1991) propose that different manifestations of ambivalence occur when low-prejudice persons are asked to think about or interact with African-Americans. Gaertner and Dovidio propose that the conflicting thoughts and feelings low-prejudice persons, or aversive racists, express are often excluded from their awareness with the help of non-racial justification factors. In this way, their egalitarian self-image can be maintained even when they behave in negative ways toward African-Americans. Devine et al. on the other hand, state that low- prejudice persons are often well aware of .both their prejudiced and non-prejudiced beliefs. In addition, if their negative behavior toward African-Americans contradicts their non-prejudiced personal beliefs, then they experience feelings of compunction. Goals of the Present Research The present research will attempt to resolve this apparent inconsistency based on a modified version of the Devine et al. (1991) paradigm. The set of hypotheses to be outlined below will be investigated in two experiments. The first will be a paper and pencil study that is based on a modified version of Devine et al.'s (1991) paradigm. The second will involve an actualinterracial interaction, unobtrusive measures of racial prejudice, as well as a paper and pencil section similar to that in the first experiment. Goals of Study 1 The current experiment builds on Devine et al.'s (1991) earlier study by including a non-racial justification factor in the "should" and "would" vignettes, by including ethnically neutral actors in the vignettes, and by asking directly whether subjects are aware of a discrepancy in their responses. It is assumed that if the ethnicity of the target in the should and would situations is not identified as specifically African-American or some other ethnicity, then subjects will assume that the "default ethnicity" of the actor is White. All subjects will first report their personal standards for how they feel they should respond to Whites or African-Americans in certain situations. They will then report how they actually would respond to Whites or African-Americans in the same situations. This will allow us to ~test for differences in should and would responses to White targets and African-American targets. For half of the subjects, a non-racial justification factor will be inserted 7 into the should and would vignettes. The other half of the subjects will not receive this information. This will allow the subjects to utilize this factor, if so desired, by making it the salient issue in their responses instead of race. It will also allow us to determine if there is a difference between subjects' responses to White targets with the NRJF and African-American targets with a NRJF in the situation. For example, subjects who reCeive a NRJF and also respond to African-American targets may give more negative responses than subjects who receive the same NRJF but respond to White targets. If subjects who did not receive a NRJF respond as negatively to African-American targets as to White targets, it is then apparent that the presence of a NRJF has led Subjects to respond to African American targets in a more negative manner than Whites. The non-racial factor may also be utilized to keep an actual discrepancy between should and would responses out of awareness. To assess this possibility, a “direct measure of subjects' perceived . awareness of a discrepancy will be included. They will simply be asked to estimate how well their should responsesmatched their would responses. It may be that subjects show an actual discrepancy but are unaware of it. Finally, the Need for Cognition measure (Caccioppo & Petty, 1982) will be included to examine the possibility that this variable acts as a moderator of the. impact of a NRJF and ethnicity on subjectS‘ should and would responses and their should-would discrepancies. The Need for Cognition.(nCog) variable is presumed to measure the need peeple have to engage in elaborative thought. It may be that When subjects with a low Need 8 for Cognition are presented with a NRJF, they may be more likely to engage in and be successful at self-deception. Subjects low in nCog are likely to be less thoughtful and complex than subjects high in nCog and therefore may be more likely to use a NRJF without thinking about its camouflaging effect on prejudiced responses. The Devine et al. (1991) theory of Prejudice with Compunction and the Gaertner and Dovidio theory of Aversive Racism provide convenient starting points for the current study. These theories, however, do not directly address the behavior which may result from the methods used in this experiment. For example, Devine et al. (1991) does not directly address the role a non-racial justification factor plays in subjects' awareness of should-would discrepancies. Similarly, Gaertner and Dovidio do not directly address the role of NRJF's in a comparison of subjects' "should" and "would" standards, or the role of compunction in the processes of prejudice. For this reason, the current study will construct a "Low Self-Insight" hypothesis (Low SI) based on Gaertner and Dovidio's work, and a "High Self-Insight" hypothesis (High SI) based on Devine et al.'s work. This will provide specific predictions for the current experiment. These researchers' theories will be used as starting points for the Low SI and High SI hypotheses, but the hypotheses will also include extrapolations from these theories to fit the current methods. Wham In general, it is expected that subjects higher in prejudice will give more negative responses to African-American targets in the 9 vignettes than subjects lower in prejudice. In addition, it is anticipated that, overall, the responses to African-American targets will be more negative than the responses to ethnically nondescript targets. (em Self-Insight eretheses As implied by the Aversive Racist theory, the Low SI position posits that even when norms for how to behave in a situation are clear, subjectsmay use a NRJF to camouflage their prejudiced ‘ behavior. Simultaneously, the subjects' use of the NRJF keeps the prejudiced thoughts and feelings that led to this behavior out of awareness. In terms of the present study, the Low SI hypothesis asserts that since NRJF's provide a camouflage for prejudice, we would expect more negative reactions to African-Americans when the NRJF has been presented compared to when it has not. In order to attribute this pattern to prejudice, the NRJF must result in a significantly more negative response when applied to an African- American target than when the same NRJF is applied to a White target. When a NRJF is not presented, reactions to African- Americans will be as positive as reactions to Whites. This will be due to social norms dictating non-prejudiced responses when a NRJF is not available. The Low Self-Insight hypothesis also asserts that because NRJF's provide a camouflage for prejudice, subjects themselves may be unaware that they are behaving in a prejudiced manner. Subjects presented with a NRJF may indicate less awareness of a discrepancy 10 between their should and would responses than subjects who are not presented with a NRJF. _ The Low Self-Insight hypothesis finally asserts that if the NRJF does preclude awareness of a discrepancy, then the presence of a NRJF may also affect the intensity of feelings of compunction that arise from a discrepancy. That is, subjects who have a discrepancy and are presented with a NRJF may have weaker feelings of compunction compared with subjects who also have a discrepancy but are not presented with a NRJF. Simply stated, subjects who are unaware of a discrepancy should not have any reason to feel compunction about a discrepancy. High Self-Ineighg Hyeetheees Similar to what is stated in Devine et al.'s Prejudice with Compunction theory, the High Self-Insight position posits that low- prejudice subjects are struggling to overcome their prejudices. In doing so, they are fully aware that they have conflicting thoughts . and feelings about African-American people and feel guilty and self- critical when they behave in ways inconsistent with their non- prejudiced personal standards. In terms of the present study, the High SI position asserts that people are sensitive to the discrepancies between how they should and would react to minorities. With this in mind, NRJF's may have no particular effect on subjects' reactions to African-American targets when compared to White targets (other than a possible effect for~NRJF's that is not contingent on the race of the target). Subjects will not "use" the NRJF in their responses any differently for African-American 11 targets than they would for White targets. If this proves to be the case, then NRJF's would not be viewed as providing camouflage for prejudiced responses. The High Self-Insight hypothesis also asserts that if people are sensitive to discrepancies between their behavior and their personal standards, they should also be aware of these discrepancies. This awareness may not be affected by the presence of a NRJF. I If this is f accurate, then subjects receiving a NRJF will not be significantly different in their awareness of a discrepancy from subjects who do not receive a NRJF. Finally, the High Self-Insight hypothesis asserts that since people are sensitive to and aware of their discrepancies toward African-Americans, they will feel self-critical and guilty when they have them. Specifically, low-prejudice subjects will experience compunction while high-prejudice subjects will experience global discomfort. Overall, subjects should not experience discrepancies toward White targets, however, and thus should not experience compunction or discomfort when responding to White targets. The presence of a NRJF-may not particularly affect these feelings of compunction and discomfort when responding to African-American targets. For this to be accurate, low-prejudice and high-prejudice subjects responding to African-Americans who have a discrepancy and receive a NRJF will not differ significantly in their feelings of compunction and global discomfort respectively, from subjects who also have a discrepancy but do not receive a NRJF. Summary of Main Predictions The Imeact ef a NRJF on Sheuld end Would Respenses (1a) The Low Self-Insight hypothesis predicts an interaction between the presence of a NRJF and the ethnicity of target on both should and would responses. Specifically, it is expected that subjects who are given a NRJF when responding to an African- American target will give significantly more negative responses on both the should and would measures than subjects who are not given a NRJF when responding to African-American targets,_ subjects who are given a NRJF when responding to White targets, and subjects who are not given a NRJF when responding to White targets. (1b) The High Self-Insight hypothesis does not expect an interaction between the presence of a NRJF and the ethnicity of target on should and would scores. The Imeeet ef e NRJF on the Awarenese ef a Diecreeeney (2a) The Low Self-Insight hypothesis predicts an interaction between the presence of a NRJF and the ethnicity of the target, on subjects' awareness of a discrepancy between their should and would responses. Specifically, it is expected that subjects who receive a NRJF when responding to an African-American target will show significantly less awareness of a discrepancy than subjects receiving a NRJF when responding to a White target, subjects not receiving a NRJF when responding to a White target, and subjects not ’ receiving a NRJF when responding to an African-American target. 12 13 (2b) The High Self-Insight hypothesis does not expect an interaction between the presence of a NRJF and the ethnicity of the target on subjects' awareness of a discrepancy between their should and would responses. The Imeeet ef e NRJF on Meeeuree ef Affect (3a) The Low Self-Insight hypothesis predicts an interaction between the presence of a NRJF and the ethnicity of the target on subjects' feelings of compunction and discomfort. Specifically, for subjects with a discrepancy between their should and would responses, it is predicted that these subjects who receive a NRJF when responding to an African-American target will experience significantly less compunction than those subjects who receive a NRJF when responding to a White target, and who do not receive a NRJF when responding to either an African-American or White target. (3b) The High Self-Insight hypothesis does not expect an . interaction between the presence of a NRJF and the ethnicity of the target on the feelings of compunction and discomfort for subjects who have a discrepancy between their should and would scores. Ancillary Predictions (I) It is predicted that there will be interaction between subjects' level of racism and the ethnicity of target on should and would responses. Specifically, it is expected that high-prejudice subjects will give significantly more negative should and would reSponses to African-American targets than low-prejudice subjects 14 will. It is also expected that high-prejudice subjects responding to an African-American will give significantly more negative responses on the should and would measures than either high- or low-prejudice subjects who respond to White targets. (2) It is predicted that there will be a main. effect for ethnicity of target on subjects' should and would scores. Specifically, it is expected that the responses to African-American targets Will be more negative overall, than the responses to ethnically non-descript targets. 3 ' (3) It is predicted that there will be a main effect for ethnicity of target on subjects' actual discrepancies. Specifically, it is expected that most of the subjects who respond to an African- American target will have an actual discrepancy between their should and would responses, while the subjects who respond to ethnically nondescript targets will have fewer and smaller discrepancies between their should and would responses. (4) It is anticipated that there will be a main effect for NRJF on should and would responses to both African-American and White targets. Specifically, we anticipate that subjects who receive a NRJF will give more negative should and would responses to both types of targets than subjects who do not receive a NRJF. (5) It is anticipated that there will be a main effect for ethnicity of target on affect scores. Specifically, it is expected that subjects who respond to African-American targets will have a should-would discrepancy (see #3 above) and consequently, will experience significantly greater Compunction than subjects who respond to White targets. 15 (6) It is anticipated that there will be a main effect for level of prejudice on affect scores. Specifically, it is expected that low- prejudice subjects with a discrepancy will experience significantly greater compunction than high-prejudice subjects but both groups will experience about the same level of global discomfort. EXPERIMENT 1 Method Seet'eete Q Design Subjects were 114 White male and female Introductory Psychology students who participated in the experiment to receive partial credit for course requirements. There were unequal numbers of women (N=71) and men (N=43) due to the greater number of women who enroll in Introductory Psychology. Gender differences were examined and the sex of subject variable had no systematic effects. Consequently, the gender variable will not be discussed further. _ Subjects were selected on the basis of their scores on the Modern Racism Scale (McConahay, Hardee, & Batts, 1981). This scale is a relatively non-reactive measure of racially prejudiced attitudes toward African-Americans. Subjects rated each item on a 9-point scale ranging from disagree strongly (-4) to agree strongly (4). Composite MRS scores were Computed by summing subjects' ratings, after reverse scoring when necessary. Scores on the MRS can range from -28 (low prejudiced) to 28 (high prejudiced). A median split (median = -9) was employed to dichotomize the level of prejudice variable. Subjects with scores less than -9 were classified as having low prejudice while subjects with scores greater than -9 were classified as having high prejudice. The design of the study is a Z (ethnicity of interviewer) X 2 (non-racial justification factor condition) factorial design. 16 Proeeeure Subjects participated in groups of 10-15 people. The subjects were either told the study involved their reactions to "African- Americans" or their reactions to "people" in certain situations. They were assured that their identity would remain completely confidential. They were encouraged to be as open and honest as possible. The questionnaire consisted of 5 sections: Should, Need for Cognition, Would, Awareness, and Affect measures. Shegle Measure , Subjects first reported how they felt they should respond to either White or African-American people in different situations. The instructions read as follows: Often times we set up personal standards or guidelines for evaluating our own behavior or responses to various groups of people. We usually phrase these guidelines in terms of how we believe we ehould respond or behave in various situations. Based on your own personal standards for how you eheelg respond, consider the following situations. For each situation, circle the number between 0 (strongly disagree) and 10 (strongly agree) that best reflects your personal standard for how you sheuld respond in the situation. (Devine et al., 1991, p. 819) For subjects in the African-American and non-racial factor absent condition, one situation read as follows: Imagine that an African-American person boarded a bus and sat next to you. You ehould feel uncomfortable that an African- American is sitting next to you. For subjects in the white and non-racial factor absent condition, one situation read as follows: 17 18 Imagine that someone boarded a bus and sat next to you. You eheeld feel uncomfortable that someone is- sitting next to you. Subjects in the African-American and non-racial factor present condition read the identical instructions but their situation read as follows: Imagine that an African-American boarded a bus and sat down next to you. This person were dark sunglasses, wore an earring in their nose, and was unkempt in their general appearance. You eheulg feel uncomfortable that an African-American is sitting next to you. Subjects in the white and non-racial factor present condition read the following situation: Imagine that someone boarded a bus and sat down next to you. This person were dark sunglasses, wore an earring in their nose, and was unkempt in their general appearance. You sheuld feel uncomfortable that this person is sitting next to you. Two of the remaining four situations also focused on feelings. One situation involved either a White or an African-American couple moving in next door. The other, the situation the subject read last, involved feeling uncomfortable that a job interviewer is either White or African-American. The other two situations involve stereotypic thoughts that subjects might have in contact situations with White or African-American people. One thought situation focused on seeing three middle-aged White or African-American men 19 on the street corner and thinking "Why don't they get a job?" The other involves either a White or an African-American woman with several small children and thinking "How typical." (For a listing of the five scenarios, see Appendix A) One thought and two feeling situations were phrased using how the subject should not feel. These items were later reversed scored. In each of the four situations, the non-racial justification factors appeared as the second sentence in the situation. N f r nii n M r Following this, subjects completed a shortened version of Cacioppo & Petty's (1982) Need for Cognition measure consisting of 25 items (Tanaka, Panter, & Winborne, 1988). These 25 items have been shown by Tanaka et al. to load onto 3 primary dimensions of Cognitive Persistence, Cognitive Complexity, and Cognitive Confidence. This measure was included to serve as a distracter task for subjects between completing the should and would measures. It also provides a personality variable that may or may not act as a moderator of subjects' should and would responses or their awareness of a disCrepancy between personal standards and actual responses. Although no precedent for this exists in the prejudice literature, it may be that subjects low or high on these three dimensions systematically differ in their perceptions of discrepancies between their personal standards and actual behavior. It may also be that subjects low or high in prejudice are either low or high on these three dimensions of the Need for Cognition. Wedld Meeedre The subjects then responded to the same 5 should situations, only they reported how they "would" actually respond in these situations. The instructions read as follows: Although we set up personal standards for how we should respond, our actual responses may or may not be consistent with these standards or guidelines. Consider the five sitUations you responded to previously. You will now be presented with these same situations. But this time, report on the O to 10 scale how you believe youw weuld aetdally respond in the situations. Base these responses on your personal thoughts and feelings. It is important to keep in mind that there are no right or wrong answers. Your responses may or may not be the same as the ones you gave earlier. It is also important that you be as honest and open as possible. Remember that your answers will be completely anonymous. The would situations, in both non-racial justification factor conditions, were identical to those in the should situations except the war "would" replaced theword "should." Dieere dandy Index . The discrepancy index was calculated by subtracting subjects' should ratings from their would ratings for each scenario. A discrepancy score was computed for each situation. Awareness Meeedre The fourth measure asked subjects to report how well they felt their actual responses matched. their earlier reported should standards. They rated this degree of matching on an 11-point scale, ranging from "0" (did not match at all) to "10" (matched perfectly). 20 21 They were reminded that those subjects showing a good match between should and would responses were not necessarily any "better" or "worse" than those not showing a good match. Affeet Measure The last measure asked subjects to report how they were feeling, at that moment, about how well their should responses matched their would responses (Devine et al., 1991). They completed 35 affect items, rating each one for the degree to which it described their feelings on scales ranging from "1" (does not apply at all) to"7" (applies very much). R esu I ts Shed/d and Wedld Respenses According to the Low SI hypothesis, when a NRJF provides a basis for justifying a negative response to an African-American target, subjects should openly report more negative should and would responses to the target. If ethnicity is the true source of these negative responses, the negative impact of the NRJF should be significantly greater for African-American than for White targets. This ethnicity of target by NRJF interaction, however, is not predicted by the High SI hypothesis. Means relevant to this prediction are presented in Tables 1 and 2, which show subjects' should and would responses as a function of target ethnicity and the ' presence of a NRJF. A mixed-model analysis of variance was conducted on the five should items, with NRJF and target ethnicity serving as between-subjects factors, and scenario serving as a 22 within-subject factor. Looking first at should scores, it is clear that overall, the race by NRJF interaction was not significant, F(1,110) < 1.0, ns, nor was there a significant ethnicity by NRJF by scenario interaction, F(4, 440) = 1.71, p >.15. The absence of these interactions casts doubt on the validity of the Low SI hypothesis. Inspection of the means in Table 1 suggests that subjects' should scores were strongly affected by the NRJF, independently of the race variable. Overall, the presence of a NRJF produced more negative should ratings (M2382) than when the NRJF was absent (M=2.11), F(1, 110)=92.84, p< .001. The magnitude of this maineffect was marginally contingent upon the particular scenario in question, F(4,440) =2.72, p< .06, with Scenario 2 showing the largest effect, F(1,110)=S4.46, p< .001 and Scenario 4 showing the smallest one F(1,110)=9.87, p< .01 (see Table 1). Nevertheless, the main effect of NRJF on should scores was significant for each of the five individual scenarios, all p's < .01. It is apparent from these means that the NRJF has the effect of leading subjects to respond in a more negative manner to whomever is the target of their response, regardless of ethnicity. It is also clear from Table 1 that there was a main effect of target ethnicity on subjects' should scores, F(1,110)=12.18, p< .001. Specifically, subjects responded more negatively to ethnically nondescript targets (M=3.38) than to African-American targets (M=2.65) independent of NRJF condition. This main effect, however, was qualified by a significant ethnicity by scenario interaction, ‘ F(4,440)=6.10, p< .001. Scenario 5 showed the largest effect, F(1,110)--26.12, p< .001, and Scenario 4 the next largest, 23 F(1,110)=5.79, p< .05, while the other three scenarios showed no effect, p's > .10 (see Table 1 for means). Two things are apparent from this result. First, the overall main effect for ethnicity of target on should scores may largely be due to one highly statistically significant item (item 5). Second, the main effect for race of target is in the opposite direction from what was predicted. 0n items 4 and 5, subjects were more negative toward White targets than African-American targets. It appears that this pattern of responses on items 4 and 5 reflect a phenomenon particular to these items. MW To test the Low Self-Insight hypothesis prediction of an interaction between NRJF and ethnicity of target, a mixed-model analysis of variance was conducted on the 5 would items with NRJF and ethnicity of target serving as between-subjects factors. Means relevant to this prediction are presented in Table 2. Looking at the would scores, it is apparent that this interaction was significant F(1,110)=5.04, p< .05. This interaction, however, was qualified by a higher-order interaction between scenario, ethnicity, and NRJF F(4,440)-2.57, p< .05, indicating that scenario is an important factor in how the NRJF by ethnicity interaction was manifested in the different would scenarios. To investigate the specific pattern of the interaction, univariate analyses of variance were conducted on each of the five would items. The results revealed a significant ethnicity by NRJF interaction only for item 5, F(1,110)=14.24, p< .001, while items 1- 24 4 were not significant F(1,110)=3.28, p>.07; F(1,110)<1.0, ns; F(1,110)=1.02, p> .10; F(1,110)<1.0, ns, respectively. Inspection of the means for scenario 5 in Table 2 suggests that the pattern of scores is inconsistent with the specific Low SI hypothesis prediction. Subjects in the African-American and NRJF present condition did not exhibit significantly more negative would scores (higher) than in the other conditions. Rather, would responses were significantly more favorable toward African-American targets than toward White targets when the NRJF was absent but treatment was about the same when the NRJF was present. In addition to this, the significance of the interaction is apparently due to a single, highly statistically significant item (#5). With both of these points in mind, we cannot assume that the Low SI hypothesis' predicition of an interaction between NRJF and ethnicity of target has been corroborated. According to the ancillary predictions for Study 1, it was anticipated that high-prejudice subjects would give significantly more negative should and would responses to African-American targets than low-prejudice subjects who respond to African- American targets and either low- or high-prejudice subjects who respond to White targets. A multivariate analysis of variance was conducted on the 5 should and 5 would items with ethnicity, NRJF, and level of prejudice serving as independent variables. As stated above, to separate subjects into dichotomous Classes of either high- prejudice or low-prejudice, a median split was performed on subjects' Modern Racism Scale scores (median score= -9).' The results revealed that the predicted interaction between ethnicity of 25 target and level of prejudice on should scores was not significant F(5,102) <1.0, ns. The results also revealed that the predicted interaction between ethnicity of target and level of prejudice on would scores was not significant F(S,102)<1.0, ns. Despite a nonsignificant interaction, inspection of cell means revealed a very interesting pattern of should responses. On each should item, and on would items 1,4, and 5, low-prejudice subjects were more positive toward African-American targets than White targets regardless of NRJF condition. This pattern of responses by low-prejudice subjects may be due to what could be called a social desirability bias in which White subjects respond more positively to a minority group than to their own ethnic group in an attempt to appear extremely non-prejudiced. In summary, the results of the analyses of should and would responses fail to support the Low SI hypothesis but are consistent with the High SI hypothesis. Despite an ethnicity by NRJF interaction on the would responses, which was predicted by the Low . SI hypothesis, the pattern of means were inconsistent with the specific pattern predicted by this hypothesis. In addition, this interaction was qualified by a significant scenario by ethnicity by NRJF interaction, which was apparently due to a single, highly significant item. Finally, the predicted level of prejudice by ethnicity of target interaction was not significant, indicating that the median split may not have produced meaningfully different groups of low- and high-prejudice subjects. Should end Weeld Discreeancies Perceived Diecreeancies According to the Low SI hypothesis, when a NRJF provides a camouflage for prejudiced responses to African-Americans, it also serves to keep these prejudiced responses out of the subjects' awareness. Because of this, subjects will not perceive a discrepancy between should and would responses but will instead perceive that their responses were non-prejudiced and consistent. If this process occurs, subjects who receive a NRJF and respond to an African-American target should show significantly less awareness of a discrepanCy than subjects in the other conditions. This ethnicity of target by NRJF interaction, however, is not predicted by the High SI hypothesis. Means relevant to this prediction are presented in Table 3. A univariate analysis of variance was conducted on the variable which measured subjects' perceived awareness of a discrepancy, with ethnicity of target and NRJF serving as independent variables. It is clear that the ethnicity by NRJF interaction was not Significant F(1,110)<1.0, ns. Cell means indicate that subjects in each condition perceived about the same level of a discrepancy between their should and would reponses. Aetdel Dieereeeneiee As stated earlier, findings in the. prejudice literature have shown that subjects can be ambivalent about how to respond to African- ‘American targets. Based on this evidence, it was anticipated that subjects responding to African-American targets would have significantly greater actual discrepancies between their should and 26 27 would responses than subjects responding to ethnically nondescript targets. Discrepancy scores were computed by subtracting subjects' would score from their should score for each scenario. Cell means 3 relevant to this predicted main effect for ethnicity of target on actual discrepancies are presented in Table 4. A mixed-model analysis of variance was conducted on the discrepancy scores for each scenario with NRJF and ethnicity of target serving as between- subjects factors and scenario serving as a within-subject factor. The results indicate that the ethnicity of target main effect was not significant F(5,106)< 1.0, ns. The results also indicate that the ethnicity by scenario interaction was not significant F(4,440)=1.41, p> .10. Inspection of cell means suggests that subjects responding to White targets were just as likely as subjects responding to African-American targets to have a discrepancy. The pattern of cell means in Table 4 do suggest, however, an ethnicity of target by NRJF interaction on the discrepancy scores F(1,110)=6.37, p< .05, which is particularly evident in scenario 5 F(1,110)=12.38, p< .001. Overall, however, the magnitude of this interaction was not-contingent upon the particular scenario in question F(4,440)< 1.0, ns. Examination of cell means in Table 4 indicates that the form of the interaction varies dramatically in each scenario, and thus, there is no clear interpretation of the overall effect. These results from analyses of both perceived and actual discrepancies casts doubt on the Low SI hypothesis but is consistent with the High SI notion that a NRJF would not impact ’ subjects'. perceived discrepancy scores differently in the African- American and White target conditions. Affeet Remnees As stated earlier, according to the Low SI hypothesis, when a NRJF provides a camouflage for prejudiced responses to African- Americans, it also serves to keep these prejudiced responses out of the subjects' awareness. Consequently, the Low SI hypothesis posits that because subjects will be unaware of a discrepancy, they will not report feelings of compunction or discomfort about a discrepancy. For this process to occur, subjects who receive a NRJF and respond to an African-American target should experience significantly Iess'compunction and discomfort than subjects in the other 3 conditions. This ethnicity of target by NRJF interaction, however, is not predicted by the High SI hypothesis. Means relevant to this prediction are presented in Table 5. A univariate analysis of variance was conducted on subjects' self-reports of their level of compunction and discomfort with ethnicity and NRJF serving as independent variables. The ethnicity of target by NRJF interaction was not significant on the compunction variable F(1,110)= 2.29, p> . .10. Inspection of cell means indicate that subjects who received a NRJF and responded to African-American targets actually had the highest levels of compunction when compared to the other three conditions. This result is the opposite of what the Low SI predicted and casts doubt on the validity of this hypothesis. It is clear that the ethnicity by NRJF interction was also not significant for the discomfort variable F(1,110)< 1.0, ns. Cell means indicate that subjects in each condition experienced about the same level of discomfort about their discrepancies. 28 29 Overall, this pattern of results on the measures of affect fail to support the Low SI hypothesis. The results are also inconsistent with the High SI hypothesis, which predicted main effects for ethnicity of target and level of prejudice on compunction scores Cell means seem to indicate an overall trend that subjects responding to African-American targets experienced greater compunction, but this trend was not significant F(1,110)< 1.0, ns. One reason may be that the predicted main effect for ethnicity of target on actual discrepancies was not observed. If subjects were not significantly more likely to have a should-would discrepancy when responding to an African-American than a White target, they also would not experience significanlty greater compunction than subjects who responded to White targets. The results also did not reveal, however, the High SI hypothesis' predicted main effect for level of prejudice on compunction F(1,110)=2.96, p> .08, or on global discomfort F(1,110)=1.49, p>.10. Consequently, the pattern of results indicate that neither the Low or High SI hypotheses were accurate depictions of subjects' affective reactions to the manifestation of racial ambivalence. Need Fer Qegnitien Scores .The Need for Cognition Scale (nCog) (Cacioppo & Petty, 1982) presumably measures the need people have to engage in elaborative thought. Tanaka, Panter, and Winborne (1988) Conducted a factor analysis on this 34-item measure and found 3 significant factors: ‘ Cognitive Persistence, which measures the degree to which an individual enjoys engaging in cognitive tasks, Cognitive Confidence, 30 which measures the degree to which an individual is confident about engaging in cognitive activities, and Cognitive Competence, which measures a preference for complex, relative to simple information- processing demands. The nCog scale was included in Study 1 to examine the possibility that nCog acts as a moderator of should and would responses and should-would discrepancies. Prior to testing this hypothesized relationship, Pearson product-moment correlations were computed between the 3 factors of the nCog scale and the 5 should items, the 5 would items, and the S discrepancies as an initial test for a relationship between these variables. Inspection of the correlations revealed that only the Cognitive Confidence factor had any significant correlations. These correlations were with should item 2, = -.184, p< .05, (two-tailed), would item 1, r= -.21, p< .05, (two-tailed), and with would item 5, r= - .229, p< .01 (two-tailed). It appears that a strong relationship does not exist between the nCog factors and the should, would, and discrepancy scores. For this reason, further tests of nCog as a moderator variable were not conducted. Additional PearSon product-moment correlations computed between the 3 nCog factors and subjects' level of prejudice and perceived discrepancy scores revealed significant relationships. All 3 nCog factors were significantly negatively correlated with subjects' level of prejudice: Cognitive Confidence, r= -.228, p< .05; Cognitive Persistence, r= -.279, p< .01; Cognitive Competence, r= -.237, p< .05, all tests of significance two-tailed. Cognitive Confidence was also significantly correlated with subjects' perceived awareness of a discrepancy, r= .221, p< .05, (two-tailed) 31 as was Cognitive Complexity, r= .226, p < .05, (two-tailed). These results indicate that subjects who are more likely to engage in and enjoy elaborative thought have less prejudiced personal standards and expect greater consistency between their responses than do subjects who are unlikely to engage in elaborative thought. Discussion The goals of Study 1 were to examine the impact of a non-racial justification factor, ethnicity of target, and level of prejudice on discrepanciesbetween how subjects felt they should respond and how they actually would respond to White or African-American targets in 5 different situations. 0f the main predictions for StudyI, those made by the High Self-Insight hypothesis were corroborated while those made by the Low Self-Insight hypothesis were not corroborated. 0f the ancillary predictions, only the main effect for NRJF on should and would scores was corroborated. The results from Study 1 have theoretical implications for the Low and High Self-Insight hypotheses as well as for the Aversive Racist theory and Prejudice with Compunction theory. Based on our results, the Low 51 hypothesis was not an accurate explanation for how subjects cope with having both racially prejudiced and non- prejudiced beliefs. With the specific NRJF's and scenarios investigated here, subjects did not appear to capitalize on the presence of a NRJF to camouflage a prejudiced response. Subjects who received a NRJF did not give significantly more negative responses to African-American targets than to White targets. In 32 addition, the NRJF did not camouflage subjects' discrepancies from their own awareness of them. Subjects receiving a NRJF did not show significantly less awareness of a discrepancy between their should and would standards toward African-Americans than toward Whites. The NRJF did not "protect" subjects from feeling badly about a discrepancy. Subjects who received a NRJF did not experience significantly lesscompunction. . I This indirectly casts doubt on the Aversive Racist theory of prejudice, on which the Low SI hypothesis is based. The theory's hypothesized role of a NRJF in the processes of prejudice did not hold. Given that a number of the findings were contingent upon the specific scenarios and NRJF's employed, it will be important for future work to determine exactly what features of a given scenario or NRJF affect its impact on subjects' ratings of appropriate or actual behavior in the scenario. In fairness to the Low SI hypothesis, it may be possible to find certain scenarios and or NRJF's that will produce the sorts of subtle prejudice described by Gaertner and Dovidio. It will be particularly important to understand what the fundamental characteristics of such scenarios are. The results also weakly corroborate Devine et al.'s notion that a NRJF would not impact subjects' responses to African-American targets. These results are only weak support for Prejudice with Compunction theory, however, because it predicted the null hypothesis. Finding in favor of the null hypothesis is not strong evidence to corroborate a theory. In addition, the predictions concerning subjects' affective reactions were also not corroborated, 33 which casts doubt on the consequences of transgressing personal standards that Devine et al.‘ have proposed. Results from the ancillary predictions showed that subjects are working perhaps a bit too hard to portray themselves as non-prejudiced. Subjects evaluated members of their own ethnic group more negatively than members of a non-dominant ethnic group. Perhaps these subjects have gone beyond being non-prejudiced, in which case members of both groups would be evaluated similarly, to what some might call "reverse prejudice or reverse discrimination." It may be that the cost of being seen as prejudiced in today's social Climate outweighs the cost of evaluating one's own ethnic group negatively. In other words, perhaps it is okay for Whites to derogate other Whites but it is not okay for Whites to derogate African-Americans. In addition, there may be situations in which this type of behavior (ingroup derogation) is more likely. The results revealed a scenario effect whereby situations that seem to be more salient to our sample population (job interview), had a greater likelihood of drawing upon this ingroup derogation. Likewise, although this was not demonstrated in Study 1, there may be scenarios in which outgroup derogation Is more likely in the should-would paradigm. This possibility is an interesting avenue for future research to determine in what situations ingroup and outgroup derogation might occur and if a NRJF could play a role in the processes of outgroup derogation. ' One shortcoming of Study 1 involves the nature of the stimulus materials. Subjects were told at the outset that this study involved their reactions to African-American people. In addition, in each 34 scenario, cues were present that repeatedly reminded them that they were responding to African-American targets. This seems to "stack the deck" against the Low SI hypothesis because of the difficulty in engaging in self-deception under these conditions. A . second shortcoming is the lack of aforementioned specificity in determining which scenarios pull more strongly for certain responses. This could have been ameliorated with thorough pretesting of different should and would scenarios. A subset of scenarios in which subjects respond equally positively to the targets could have been selected by using this process. A third shortcoming that may explain the ingroup derogation is that our distinction between low- and high-prejudiced subjects was not meaningful. A median split of a unimodal distribution does not provide a strong basis for such a conceptual distinction. In addition, a good portion of our high-prejudice group had negative (low in prejudice) scores. This implies that a sampling bias is present that compromises the generalizability of our findings. Based on our sample, it seems white college students on a large midwestern campus tend to be low in prejudice on attitude surveys. Devine et al., however, were able to obtain significant differences between high- and low-prejudice subjects on a similar campus. A fourth and final shortcoming is the self-report nature of the questionnaires, which many researchers have shown, can draw upon a social desirability bias for non-prejudiced responses. In an attempt to address this last issue, a second study was designed to involve an unobtrusive behavioral measure of racial prejucice. EXPERIMENT 2 Introduction Study 1 utilized subjects' self-reports of their personal standards (should response) and their actual responses (would response) to African-Americans. It would seem that it might be possible to obtain a report of subjects' "would" responses in a more . realistic manner. Based on past criticisms in the prejudice literature of self-report bias and the results of Study 1, we conducted a second study to more closely approximate how subjects "would" respond in an actual interaction with an African-American. Criticisms of the self-report method of measuringracial prejudice stem from the inconsistency between subjects' responses on surveys and unobtrusive measures of prejudice. Since the late . 1960's, there has been a marked decrease in endorsement of negative attitudes about African-American people on surveys (Campbell, 1971; Lipset & Schneider, 1978; Taylor, Greeley, & Sheatsley, 1978). Many have taken this evidence to mean prejudiced attitudes toward African-Americans have declined significantly since the Civil Rights Movement. Some might even go so far as to conclude from this evidence that racial prejudice is no longer pervasive in the US. Other research shows, however, that many of the same people endorsing non-prejudiced attitudes on surveys behave in discriminatory ways to African-Americans on unobtrusive measures . of racial prejudice (Crosby, Bromley, & Saxe, 1980; Gaertner, 1976; Gaertner & Dovidio, 1986). To explain this inconsistency, some "researchers have stated that verbal reports are not trustworthy . measures of peoples' attitudes and beliefs (Crosby et al. 1980; 35 36 McConahay, 1986; Sigall 8: Page, 1971). Instead, these researchers assert that the survey results represent how people believe they "should" be responding and the unobtrusive measures get at how people actually "would" respond. Since these unobtrusive measures reveal prevalent discriminatory behavior toward African-American people, some conclude that racial prejudice is still pervasive in the US. Yet, still other researchers have proposed a different explanation to resolve the inconsistency in results between self-report and unobtrusive measures of prejudice. As stated earlier, rather than asserting that the unobtrusive measures reflect White people's "actual" feelings, they hypothesize that Caucasians' beliefs about African-Americans contain both prejudiced (stereotypes) and non- prejudiced (egalitarian values) components (Allport, 1954; Gaertner et al. 1986; Higgins & King, 1981; Katz, Wackenhut, & Hass, 1986). That is, their feelings and attitudes toward African-Americans are conflicted. Sometimes their prejudiced views are expressed and ‘ other times their non-prejudiced views are expressed. As we have seen, Devine et al. (.1991) and Gaertner and Dovidio (1986) propose different ideas concerning the manifestations of the two sets of beliefs. They also utilized different methods in the studies that led to their theories. Gaertner & Dovidio used primarily unobtrusive measures of prejudice in prosocial behavior toward African- Americans while Devine et al. used self-report. It should therefore be illuminating to examine both groups of researchers' theories in a study utilizing both measures of prejudice. flea/5 Qf the Seednd Study The goal of the present research is to determine whether the conclusions of Study 1 hold true when the method of measuring prejudice is altered. Study 2 will utilize both self-report and unobtrusive measures of prejudice. Instead of having subjects imagine how they "would" respond in certain situations, subjects will participate in a live interaction with an African-AmeriCan (an interview situation) and use this experience to reporttheir feelings and behavior concerning the interaction. This, in essence, will be their would response. The goal is to provide a real situation they can draw upon in answering how they would feel during such an interaction. The live job interview situation will be adapted from a study by Word, Zanna, & Cooper (1974) in which confederates served as job applicants and subjects served as job interviewers. The applicants were trained to behave in a standard way. The authors found that the subjects behaved in a less immediate manner toward the African-American confederates than toward the White confederates. Mehrabian (1969) has shown that less immediate non- verbal behaviors are. a consequence of negative attitudes toward a target. . The interview situation in the current study will be reversed as the subjects will serve as the applicants and the White and African- American confederates will serve as the interviewers. The non- racial justification factor condition will again be manipulated in the should and would situations. The subjects will also complete the should situations as in Study 1. .The discrepancy scores will be computed by directly comparing subjects' self-reports of their 37 38 behavior and their self-reported should scores from the last vignette involving the interviewer. They will also be asked to assess the degree to which their feelings and behavior during the interaction (an interview) matched the les_t_ should situation involving the interviewer. This will provide the subjects' awareness of this discrepancy based on a more realistic interaction. The rest of the experiment will be identical to Study 1. Main Predictions The main and ancillary predictions for Study 2 are identical to those for Study 1. These predictions can be found on pages 12-14 of this text. In addition, there are two predictions that are particular to Study 2 which are outlined below. General Predictions (1) Based on the findings of Word et al. (1974), a main effect is predicted for ethnicity of interviewer on the distance the subjects position themselves from the interviewer. Specifically, it is expected that those subjects (both Iow- and high-prejudice) with an African-American interviewer will position themselves farther away from the interviewer. , (2) A main effect is also expected for ethnicity of interviewer on the time the subjects spend answering the standard list of questions asked by the experimenter. Specifically, it is anticipated that subjects interviewed by an African-American confederate will take 39 less time answering the questions than subjects interviewed by a White confederate. Method QED/£11 Following a modified version of Word, Zanna, & Cooper (1974), subjects were ledto believe they would be participating in an unrelated two part study. In the first part they were told they would be helping Career Planning and Placement Services, in conjunction with the Psychology Department, to train interviewers for an upcoming summer job operation. Subjects served as the interviewees for a brief session with an African-American or White, male confederate interviewer. The confederates were trained to speak and behave in a standard way. Subjects were placed in one of two conditions. In the non-racial justification factor present condition, subjects were told that the interviewer was a new member on the summer job project and had missed earlier training. In the non-racial justification factor absent condition, subjects were not given this. information. In the second part of the experiment, subjects were told they would be filling out a few brief questionnaires. Throughout the two- part experiment, dependent measures were administered to assess how subjects felt they “should” (Devine et al., 1991). respond to African-American people and also to assess ‘how they actually “would” respond to the White and African-American interviewers, both unobtrusively and through self-report. After completing these questionnaires, subjects were thanked and debriefed. S ' c D i n Subjects consisted of 62 Introductory Psychology students. Subjects were White male (N=31) and female (N=31) undergraduates ' who were low or high in prejudice. Gender differences were examined and the sex of subject variable had no systematic effects. Consequently, the gender variable will not be discussed further. Subjects were selected on the basis of their scores on the Modern Racism Scale (McConahay, Hardee, & Batts, 1981). This scale is a relatively non-reactive measure of racially prejudiced attitudes toward African-Americans. Subjects rated each item on a 9-point scale ranging from disagree strongly (-4) to agree strongly (4). Composite MRS scores were computed by summing subjects' ratings, after reverse scoring when necessary. Scores on the MRS can range from -28 (low prejudice) to 28 (high prejudice). Scores in our sample ranged from -28 to 19. A median split was employed to dichotomize the level of prejudice variable. Subjects with scores from -15 to -28 were classified as having low prejudice while subjects with scores from -13 to 19 were classified as having high prejudice. The design of the study is a 2 (Ethnicity of interviewer) X 2 (Non-racial justification factor condition) factorial design. Broader: Subjects were run two at a time. Upon their arrival in the lab, the experimenter handed out written instructions and answered any questions. The instructions explained that the experiment would be in two separate and unrelated parts. The first part would provide practice 40 41 for student interviewers from Career Planning & Placement. The second part would be an unrelated study conducted by the Psychology Department on feelings. . In the first part, subjects were told that they will be acting as interviewees for a student from Career Placement who needs ' practice interviewing others for a summer job operation. Subjects were told that they would be given a background sheet to answer from that protected their anonymity by providing a fictitious name and employment history. They were also told that the interview would last about‘10-15 minutes. The experimenter explained to the subjects that the job they would be interviewing for was a fictitious “marketing position with a cutlery company”. They were assured that the purpose of the interview was really to serve as practice for the student interviewers and were reminded that their responses would be anonymous. Subjects were told that to give the interviewer the best possible practice, they should try to present themselves similarly to how they would respond in an actual interview. At this time, the interviewer handed the subjects a background sheet of information that they used to answer interview questions from. The interviewer then left the room for about 5 minutes, giving the subjects a chance to read the information. When the experimenter returned, he brought with him the two confederate interviewers (one White, one African-American). Subjects were then randomly assigned to one of the interviewers based on their drawing of one of the interviewer's names out of a hat. This was done in the presence of the interviewers so that 42 subjects would know that they had an equally likely chance to be interviewed by an African-American as by a White. All subjects were interviewed by either one of two African-American male confederates or one of two White male confederates, all of whom were trained to speak and behave in a standard manner. This included speech error rate, forward lean, shoulder orientation, and eye-contact (Word et al., 1974). * Nen-Reeiel Juetifieetion Feeter _ The background sheet that the experimenter provided for the subjects included a fictional name, residence, major, and previous job experience. It also contained a few interviewing tips, such as how to respond if asked the question “why should you get this job over other applicants?” (See Appendix B). This was done to prevent any unusual stress for the subjects during the actual interview. The subjects were allowed to refer to this sheet during the interview. At this point half the subjects read that their particular interviewer was a new recruit in the Career Planning & Placement summer program and had not had the opportunity to practice interviews yet. Because the regular. training program had already passed, he would be getting some practice with the psychology subject pool to bring his interview skills up to par. This “lack of skills” information served as the non-racial justification factor. The other half of the subjects were not given this information. lmm ia M a r Two immediacy factors, the physical distance between the subject and experimenter and the length of the interview were measured. Word et al.(1974) found these immediacy factors to be 43 significantly different for African-American and White applicants when being interviewed by a White interviewer. The interview set- up in the current study is reversed, but these immediacy behaviors are still assumed to apply. Because the interviewers followed a standard script, the length of interview measure was expected to be affected mostly by the subject. Each interviewer covertly started a stopwatch as he and the subject entered the room and stopped the timing as the subject passed through the doorway at the conclusion of the interviewer. The longer each subject spent on giving her/his answer, the more immediate the behavior. The more immediate-the behavior, the less prejudiced it is (Mehrabian, 1967, 1968; Word et al., 1974). To measure the distance variable, the subject was asked to “pull up a chair” from a stack in the corner. The distance s/he positioned her/himself from the interviewer was measured after the subject left the room. The greater the distance, the less immediate the behavior (Word et al., 1974). The less immediate the behavior is, the less positive the subject's attitude is toward the interviewer (Mehrabian, 1967; 1968; Word et al. 1974). The immediacy behaviors were measured to serve as unobtrusive indicators of racial prejudice. Each measure wasincluded to compare differences in time and distance between subjects interviewing with an African- American and those interviewing with a White. It was also included to provide a test of whether lbw—prejudice subjects show more immediate behaviors than high-prejudice subjects and to provide a check on subjects' self-reports of their behavior during the interview. 44 Feet Interview Qdeetienneire After the interview, the subject was instructed to return to the waiting room. Ostensibly to provide Career Planning and Placement with additional feedback, the subject was given a brief questionnaire that was designed to measure her/his mood (Word et al., 1974) and behavior toward the interviewer, and to assess the subjects' "general. ratings of the interviewer." This questionnaire served as an initial mood measure before the subjects were made aware of the relation between the two parts in the experiment and its purposes. This was done in anticipation of a possible difference between the mood of the Subject before and after awareness of a discrepancy in their interracial behavior and personal standards. The mood measure, the PANAS, consisted of a list of 20 different feelings and emotions. Subjects were asked to rate how much they were feeling each emotion on a scale of 1 ("very slightly or not at all") to S ("extremely") (Watson, Clark, & Tellegen, 1988). The behavior measure served as the "would" response. Subjects Were asked to simply report how positively they behaved toward the interviewer during the interview on a scale of "0" (not at all positive) to "10" (extremely positive). This would response, although still a self-report, may be a more realistic representation of subjects' actual behaviorbecause they were able to draw upon the interracial interaction that immediately preceded the questionnaire. The general ratings of the interviewer asked subjects to rate the interviewer on 4 dimensions [eye contact (Word et al., 1974), ability to communicate, friendliness and overall perceived skill] on an 11- point scale from "0" (very low on the dimension) to "10" (very high 45 on the dimension). This measure was included to test for differences due to race of interviewer or presence of non-racial factor on the ratings. £33.11 . After completing this questionnaire, the experimenter thanked the subject and said the experimenter for the Psychology Department's study would be in shortly. After approximately 3 minutes, a different experimenter introduced himself and led the subject into a different room. The experimenter handed the subject instructions that explained the study would be concerning her/his feelings about a socially sensitive topic. Subjects were reminded of the importance of being open and honest and that their responses would be kept confidential. This was accomplished by instructing I subjects to seal their unsigned questionnaires in an envelope and were then asked to place this envelope in a large box containing a number of identical envelopes. For the purposes of later comparisons, all the questionnaires (including the PANAS and behavioral measure) were pre-marked with a three digit number and randomly assigned to the subjects. The questionnaire consisted of 3 sections: the "should" vignettes, an awareness measure, and an affect measure. W This measure was adapted from Devine et al., (1991). Subjects in each condition were asked to report their personal standards for how they should respond in five different situations involving - African-American or ethnically neutral people. The instructions read as follows: 46 Often times we set up personal standards or guidelines for evaluating our own behavior or responses to various groups of people. We usually phrase these guidelines in terms of how we believe we should respond or behave in various situations. Based on your own personal standards for how you should respond consider the following situations. For each situation, Circle the number between 0 (strongly disagree) and 10 (strongly agree) that best reflects your personal standard for how you should respond in the situation. (Devine et al., 1991, p.819) For example, one situation in the African-American condition read as follows: Imagine that you arrive at a job interview and the person who is going to be interviewing you is an African-American. You eho'uld behave in a positive manner because an African-American is interviewing you. One situation for subjects in the white interviewer condition read as follows; Imagine that you are in a job interview situation. You eheuld behave in a positive manner toward the interviewer. Subjects in the non-racial justification factor condition, regardless of interviewer's ethnicity, received an additional sentence that served as the non-racial factor. These factors were identical to those in Study 1. The scenarios were also identical to those in Study 1. When subjects read the last situation, which concerned the job interview, they read additional instructions that noted the similarity between their earlier interaction and their current situation. It was stressed that the subjects answer the should 47 situation based on their own personal standards, not based on their earlier interaction. Awereneee meeedre ' The instructions for this section acknowledged that subjects may have noticed the similarity between the last should situation and their earlier interaction with the interviewer. The instructions then revealed that the two parts of the experiment were indeed related. Subjects in both ethnicity conditions were then asked to rate how closely their personal standards for responding in the situation matched how they actually responded. For example, "Were you thinking and feeling things and behaving in ways that your personal standards dictate you should not think, feel or behave? Or was what you were thinking, feeling, and behaving during the interview the same as what your personal standards would dictate?" Subjects rated this match on an 11-point scale ranging from "0" (no match at all between my personal standards and my actual feelings and behavior in the interview) to "10" (perfect match between my personal standards and actual feelings and behavior during the interview). This scale served as the subjects' level of perceived awareness of a discrepancy between their should and would scores. Affeet Meeedre ‘ After reporting whether they showed a discrepancy in their responses, subjects completed a mood evaluation concerning how they felt about how well their actual feelings and behavior in the interview matched their personal standards. They indicated the degree to which each of 35 affect items described their feelings on 48 scales ranging from does not apply at all (1) to applies very much (7) (Devine et al., 1991). After completing this final section subjects were probed for suspicions concerning the purpose of the study and were then fully debriefed. They were told that the experiment had been designed to deceive them and that they should not feel badly if they did not see the connection. They were also reassured that any awareness of a discrepancy was not an indication that they were or were not prejudiced. Results She eld end Would Reseense; According to the Low SI hypothesis, when a NRJF provides a basis for justifying a negative response to an African-American target, subjects should openly report more negative should and would responses to the target. If race is the true source of these negative responses, the negative impact of the NRJF should be significantly greater for African-American than for White targets. This ethnicity of target by NRJF interaction is not predicted by theHigh SI hypothesis, however. Means relevant to this prediction are presented in Table 6, which shows subjects' should and would responses as a function of target ethnicity and the presence of a NRJF. To test this hypothesis, 3 univariate analysis of variance was conducted on the last should item, the job interview situation, with NRJF and ethnicity of target serving as independent variables. Looking first at the should scores, it is clear that the race by NRJF interaction was not significant (F(1,58)<1.0, p>.10). This result 49 casts doubt on the Low SI prediction that subjects who receive a NRJF and respond to an African-American target will give more negative responses than subjects in the other three conditions. Inspection of cell means in Table 6 suggests that subjects' should scores were strongly affected by the NRJF, independently of the ethnicity variable. Overall, the presence of a NRJF produced more negative should ratings (M=4.58) than when the NRJF was absent (M=2.97), F(1,58)=15.07, p< .001. It is also clear from Table 6 that subjects' should scores were affected by the ethnicity of the target they responded to, independent of the presenCe or absence of a NRJF. Specifically, subjects repsonded more negatively to ethnically nondescript targets (M=4.66) than to African-American targets (M=2.83), F(1,58)=18.19, p< .001. This pattern of results for subjects' should scores replicates the findings in Study 1. MW To test the Low SI hypothesis prediction of an interaction between NRJF and ethnicity of target, a univariate analysis of variance was conducted on the would variable with NRJF and ethnicity of target serving as independent variables. Means relevant to this prediction are presented in Table 6. The results indicate that the NRJF by ethnicity interaction was not significant F(1,58)< 1.0, ns. In addition, the presence or absence of a NRJF did not significantly affect subjects' would responses F(1,58)=1.27, p > .10. Subjects who received a NRJF had would scores (M=8.32) similar to subjects who did not receive a NRJF (M=7.84). The ethnicity of the SO interviewer also did not significantly affect subjects would scores F(1,58)< 1.0, ns. Subjects interviewing with an African-American had about the same mean would score (M=8.10) as subjects interviewing with a White confederate (M=8.06). According to the ancillary predictions for Study 1, it was anticipated that high-prejudice subjects would give significantly > more negative responses to African-American targets than both low prejudiced subjects who respond to African-American targets and either Iow- or high-prejudice subjects who respond to White targets. The means relevant to this prediction are presented in Table 7. A univariate analysis of variance was conductEd on the should item and the would item with ethnicity, NRJF, and level of prejudice serving as independent variables. The results revealed ' that the predicted interaction between ethnicity of target and level of prejudice on should scores was not significant F(1,54)=1.11, p> .10. The results also revealed that the predicted interaction on would scores was not significant F(1,54)< 1.0, ns. Inspection of cell means for both the should and would item indicates the direction of the predicted pattern did indeed emerge, although not significant. In summary, the results of the analyses of should and would responses fail to support the Low Self-Insight hypothesis but are consistent with the High Self-Insight hypothesis. The predicted ethnicity of target by NRJF did not emerge but main effects for NRJF and ethnicity did emerge on the should scores. These effects were not present, however, on the would scores. In addition, the predicted ‘level of prejudice by ethnicity of target interaction was not 51 significant, indicating that the median split may not have produced meaningfully different groups of low- and high-prejudice subjects. It is interesting that the main effect on the Should item was in the direction opposite of what was predicted, but that no significant main effect for race of interviewer was present for the would item. It may be that subjects believed that they should behave toward African-Americans in a manner that is more positive than they would behave toward ethnically nondescript targets. When it came down to actually behaving as their should standards prescribed, however, they behaved equally positively toward White and African- American interviewers. Thus, it appears that subjects' behavior was consistent toward African-Americans but discrepant toward Whites. h l n W l i r n i P r iv Di r n i According to the Low SI hypothesis, when a NRJF provides a camouflage for prejudiced responses to African-Americans, it also serves to keep these prejudiced responses out of the subjects' awareness. Because of this, subjects will not perceivea discrepancy between should and would responses but will instead perceive that their responses were non-prejudiced and consistent. If this process occurs, subjects who receive a NRJF and respond to an African-American target should show significantly less awareness of a discrepancy than subjects in. the other conditions. This ethnicity of target by NRJF interaction, however, is not predicted by the High SI hypothesis. Means relevant to this prediction are presented in Table 8. A univariate analysis of 52 variance was conducted on the variable which measured subjects' perceived awareness of a discrepancy with ethnicity of target and NRJF serving as independent variables. It is clear that the ethnicity by NRJF interaction was not significant F(1,58)< 1.0, ns. Subjects perceived about the same level of a discrepancy between their should and would responses. This casts doubt on the Low SI hypothesis but is Consistent with the High SI notion that a NRJF would not impact subjects' perceived discrepancy, scores differently in the African-American and White target conditions. Aeteel Dieereeeneiee It was anticipated that subjects responding to African-American targets would have significantly greater actual discrepancies between their should and would responses than subjects responding to ethnically nondescript targets. Discrepancy scores were computed by subtracting subjects' would score from their should score for each item. The would score was measured on an 11-point scale and consequently the magnitude of the actual discrepancy scores are not directly comparable to the actual discrepancies from Study 1. This, however, does not preclude an analysis of the direction of the actual discrepancy scores. A univariate analysis of variance was conducted on the discrepancy scores. It is clear that subjects' discrepancy scores were affected by the ethnicity of target F(1,58)=18.70, p< .001. ' Inspection of cell means, however, indicate that the direction of this main effect is inconsistent with the predictions. Subjects were more consistent in their responses to African-American S3 targets than in their responses to White targets. Subjects responding to African-American targets actually had lower mean discrepancies (M= -1.48) than subjects who responded to White targets M= -3.30). It appears from the results that, contrary to the findings in. much of the prejudice literature, White subjects (both low and high prejudice) are M conflicted about how they believe they should and would behave toward African-Americans. Instead, they appear conflicted about how to behave toward White targets. One possible explanation for this consistency in responding to African-American targets may be a social desirability response bias in which subjects' report that they should and would behave in a positive manner in order to appear non-prejudiced to the reseracher. Another possible explanation for this unexpected finding might be that subjects were unsure how to respond to the ethnically nondescript targets in the should vignettes. In these vignettes there was very little description of what the target in the situation was like. With such little information, perhaps the tendency to respond in a neutral manner was a way of suspending their prescription for behavior (i.e. their should response) until they received more information. Perhaps if the vignettes described the target as a "White person" instead of just a "person", maybe their responses would have been as consistent as subjects’ responses to targets described as an "African-American person." Affeet Reseoneee . According to the Low SI hypothesis, when a NRJF provides a camoflague for prejudiced responses to African-Americans, it also 54 serves to keep these prejudiced responses out of the subjects' awareness. Consequently, the Low SI hypothesis posits that because subjects will be unaware of a discrepancy, they will not report feelings of compunction or discomfort about a discrepancy. For this process to occur, subjects who receive a NRJF and respond to an African-American target should experience significantly less compunction and discomfort than subjects in the other 3 conditions. This ethnicity of target by NRJF interaction, however, is not predicted by the High SI hypothesis. Means relevant to this prediction are presented in Table 9. A univariate analysis of variance was conducted on subjects' self-reports of their level of compunction and discomfort with ethnicity and NRJF serving as independent variables. The ethnicity of target by NRJF interaction was not significant on the compunction variable F(1,58)< 1.0, ns. This interaction was also not significant on the discomfort variable F(1,58)=1.05, p> .10. These results cast doubt on the validity of the Low SI hypothesis but are also not entirely consistent with the High SI hypothesis. Neither of the'predicted main effects for level of prejudice on compunction F(1,58) < 1.0, ns, or global discomfort F(1,58) < 1.0, ns, were corroborated. The cell means in Table 9 also do not seem to indicate that subjects responding to an African- American target experienced significantly greater compunction (M=12.60) than subjects responding to a White target .(M=15.63), F(1,58)=2.99, p< .09, as the High SI hypothesis had predicted. Rather, there seems to be a trend in the opposite direction, although it is not significant. Subjects seem to experience greater compunction overall, when responding to White targets, which is consistent with 55 the finding that subjects responding to White targets had greater discrepancies than subjects responding to African-American targets. lmmediacz Meaeeree Pizza's: Based on the findings of Word et al. (1974), it was anticipated ’ that subjects interviewing with an African-American target would be less immediate in physical distance (i.e. sit farther away) than subjects interviewing with a White target. To test this prediction, a univariate analysis of variance was conducted on the distance variable with NRJF and ethnicity of target serving as independent variables. It is clear that there is a significant main effect for ethnicity F(1,58)=4.81, p< .05. Inspection of the cell means ‘ indicates that overall, subjects sat farther away from African- American interviewers (M=27.26 inches) than White interviewers (M=22.7S inches). These findings corroborate the findings of Word et al., (1974). This suggests that subjects' differential expression of immediacy through physical distance toward African-American and White interviewers has not changed significantly in the past 20 years. lime Based on the findings of Word et al. (1974), it was anticipated that subjects interviewing with an African-American target would be less immediate in time spent with the interviewer than subjects interviewing with a White target. To test this prediction, 3 univariate analysis of variance was conducted on the time variable 56 with NRJF and ethnicity of target serving as independent variables. It is clear that there is a significant main effect for ethnicity on the time variable F(1,58)=6.41, p< .05. Inspection of the cell means indicates, however, that the direction of this pattern is inconsistent with the prediction. Subjects interviewing with an African- American target spent significantly greater amount of time in the interview (M=274.60 seconds) than did subjects who interviewed with a White target (M=239.91 seconds). This is the opposite of what was predicted. It would appear that something about interviewing with the African-American targets led subjects to take a greater amount of time in their interview. This unexpected finding may be due to an interviewer effect for the two African-American interviewers. That is, perhaps they took a greater amount of time asking the standard list of questions than the white interviewers did. This is unlikely, however, as each interviewer was given at least three separate practice interviews prior to conducting the study, to test for differences on the immediacy variables between interviewers. Interviewers continued to practice until their performance showed that there were no differences on time spent asking questions, distance interviewer sat from subject, eye contact, and speech- error rate (see Word, et al., 1974) between the different interviewers. The direction of the difference becomes interesting in searching for reasons why subjects were less immediate In distance toward African-American interviewers but more immediate in time spent with them. Perhaps spending more time in the interview is a way of 57 “making up” for not sitting close to the African-American interviewers in an attempt to appear non-prejudiced. Or perhaps spending more time was the "lesser of two evils" during the interview. It is also possible that subjects may be more likely to - consciously monitor the amount of time they spend in the interview but not to think about how closely they sit from the interviewer. If this is the case, time spent is subject to intentional modification by subjects in order to appear non-prejudiced whereas physical distance is not. This possibility is consistent with the findings in the prejudice literature of racial ambivalence on the part of White subjects. They consciously attempt to appear non-prejudiced (greater time spent with interviewer) but still behave in prejudiced ways (greater distance sat from interviewer). Further research on measures of physical immediacy is needed to more fully explore these possibilities. Discussion , The goals of Study 2 were to examine whether the results found in Study 1 held true if the methods of measuring prejudice were altered. Unobtrusive measures of immediacy were added in Study 2 and were taken during a mock interview that the subjects participated in. As in StudyI, we were concerned with the impact of a non-racial justification factor, ethnicity of'target, and level of prejudice on subjects' discrepancies between how they reported they should respond and how they actually did respond to an interviewer in a job interview situation. Of the main predictions for Study 2, 58 those made by the High Self-Insight hypothesis were corroborated while those made by the Low Self-Insight hypothesis were not corroborated. Of the ancillary predictions, only the main effect for . NRJF on should scores was corroborated. Of the general predictions, only the effect of ethnicity of interviewer on distance the subjects sat from the interviewer was corroborated. The results from Study 2 have theoretical implications for the Aversive Racist theory and the Prejudice with Compunction theory. The results of Study 2 found no support for the Low Self-Insight hypothesis, which was derived from the Aversive Racist theory. This indirectly casts doubt on this theory. It does not appear to be an accurate explanation for how subjects cape with prejudiced beliefs that conflict with non-prejudiced attitudes. The subjects in ' our study did not use NRJF's as a way of camouflaging prejudiced attitudes about African-American targets. They also did not use NRJF's as a way of keeping their discrepancies between their should and would responses out of awareness. In addition, the presence of a NRJF did not protect subjects from feeling badly about having discrepancies. The results from Study 2 did weakly corroborate the predictions made by the High Self-Insight hypothesis, which was derived from the Prejudice with Compunction theory. This indirectly provides weak support for Devine et al.'s theory. As with Study 1, the support is weak because the predictions made by this theory were the same as the predictions made by the null hypothesis. Finding in favor of the null hypothesis is not strong evidence to corroborate a theory. Inconsistent with Devine's findings, however, was that low- S9 prejudiced subjects in Study 2, (also in Study 1) did not experience greater compunction than high-prejudice subjects. Also inconsistent was that overall, White subjects were more consistent about how they should and would respond toward African Americans than about how they should and would respond to Whites. This could be due to a social desirability response bias, or a tendency on the part of subjects to suspend their prescription for behavior (i.e. how they should behave) when information'about a target is lacking. This could also mean that, contrary to what the literature indicates, subjects in our sample are dot conflicted about how to behave toward African-Americans. This latter explanation may account for the fact that the High SI hypothesis prediction that subjects responding to an African-American target would experience greater compunction was not supported. This claim in the literature that subjects are conflicted about how to behave toward African-Americans is based upon subjects who endorse non-prejudiced attitudes on self-reports but who behave in prejudiced ways when measured unobtrusively. Results from our unobtrusive measures of immediacy indicate that subjects are more immediate (i.e. less negative) toward African-American interviewers than white interviewers on time spent answering questions. Although subjects were less immediate (i.e. more negative) toward African-Americans than White interviewers on the distance variable, the unexpected finding in regard to the time variable may indicate that both attitudes and behavior toward African-Americans have changed since Word et al.'s findings 20 years ago. Perhaps White, midwestern college students are acting in 60 ways consistent with their attitudes about African-Americans. It may also be, however, that the amount of time spent with African- American's is subject to conscious monitoring by subjects and thus is not a good unobtrusive measure of prejudice. Inthis case, subjects would appear to still be conflicted about how to behave toward African-Americans. It could also be that, as postulated in the discussion of Study 1, the current social climate's .standard for "political correctness" involves behaving more positively toward outgroup members than toward ingroup members. In any case, the findings from Study 2 are not conclusive enough to support any one of these possibilities. Further research regarding subjects' attitude-behavior consistency in response to African-American targets is needed. As in Study 1, one shortcoming of the current study is that our distinction between high and low prejudice subjects was apparently not meaningful. A median split of a unimodal distribution is not a strong basis for dichotomizing low and high prejudice categories. Subjects in our sample had a mean MRS score of (M: -15) which indicates attitudes that are low in prejudice. Thus, it would appear that the majority of our subjects were low in prejudice. A second shortcoming with the current study involves the self-report nature of both the should and. would measures. The inclusion of unobtrusive measures of prejudice and a live interaction that subjects could draw their self-reports directly from still doesnot entirely eliminate the problem. It is very difficult to obtain subjects' should standards unobtrusively. The development of a different paradigm may be needed to alleviate this potential shortcoming. Conclusions and Future Directions The two experiments described indicate little support for the Low SI hypothesis and some support for the High SI hypothesis. This implies that Devine et al.'s hypothesized manifestation of racial ambivalence may be a more accurate explanation than Gaertner and Dovidio‘s Aversive Racist Theory. It remains for future research to determine whether other scenarios or circumstances can produce responses more in line with the expectations of this theory. In addition, we cannot conclude with confidence that Devine et al.'s hypothesis provides a better explanation, as it was not strongly corroborated. Subjects responding to African-American targets did not experience significantly greater feelings of compunction than subjects responding to White targets. Future directions of research in this area should include a systematic assessment of various non-racial justification factors and scenarios. This effort would provide the boundary conditions of these two theories. By determining the circumstances that lead people to be more likely to engage in self-deception (i.e. use a NRJF) and which ones lead people to feel compunction if their behavior transgresses their non-prejudiced personal standards, it may be possible to better understand the barriers that exist to the elimination of racism. The Aversive Racist Theory and Prejudice with Compunction Theory propose different prospects for someday eliminating racism. The Aversive Racist Theory states that because aversive racists engage in self-deception and are unaware of their » prejudiced behavior, interventions aimed at eliminating this subtle form of racism will meet with defensive reactions and will be quite 61 62 difficult. The Prejudice with Compunction Theory, however, states that people have internalized non-prejudiced standards but occasionally find themselves falling back into the "prejudice habit" by acting in prejudiced ways. Consequently, because people feel guilty about this transgression, they are more likely to avoid such transgressions in the future and hence, will try very hard not to behave in prejudiced ways. By knowing the characteristics of situations that elicit subtle, aversive racism and the Characteristics that elicit prejudice with compunction, interventions focused at eliminating prejudiced behavior can be tailored to the form of raCism that tends to occur in a particular situation. I Self-Affirmation Theory (Steele, 1988) may also play a role in the long range prospects of eliminating racism. Cognitive Dissonance Theory (Festinger, 1957) proposes that inconsistencies between attitude and behavior lead to aversive states, such as a loss of self-integrity, that motivate individuals to resolve the provoking inconsistencies. Steele has proposed that the reaffirmation of self- integrity on a dimension unrelated to the provoking inconsistency leads to the tolerance of the inconsistency between attitude and behavior. If we view a transgression of non-prejudiced personal standards by behaving in a prejudiced manner as this type of attitude-behavior inconsistency, Steele's theory indicates that self- affirmation Could reduce the feelings of compunction that Devine et al. propose and consequently eliminate the motivation to avoid such transgressions in the future. Thus, if this self-affirmation process applies to the transgression of non-prejudiced personal standards, 63 the prospect for eliminating racism may not be as optimistic as Devine et al. propose. Testing this idea might help provide a better understanding of how to tackle the problem of racial prejudice in the. future. In any event, it seems clear that in order for prejudice reduction and the elimination of discrimination to occur, we will need a better understanding of the manifestations of racial ambivalence in contemporary society. References Allport, G. W. (1954). The nature of prejudice. Reading, MA: Addison-Wesley. Ashmore, R.D., 8r Del Boca, F.K. (1976). Psychological approaches to understanding intergroup conflict. In P. Katz (Ed.), Toward the eliniation of racism (pp.73-123). New York: Pergamon Press. Cacioppo, J. & Petty, R. (1982). The need for cognition. Journal of Personaltiy and Social Psychology, 42, 116-131. Campbell, A. (1971). White attitudes toward black people. Ann Arbor: Institute for Social Research. Crosby, F., Bromley, 5., & Saxe, L. (1980). Recent unobtrusive studies of black and white discrinimation and prejudice: A literature review. Psychological Bulletin, 87, 546-563. Devine, PG. (1989). Stereotypes and prejudice: Their automatic and controlled components. Journal of Personality and Social Psychology, 5 6, 5-18. Devine, P.G., Monteith, M.J., Zuwerink, J.R., & Elliot, A.J. (1991). Prejudice with and without compunction. Journal of Personality and Social Psychology. 60, 817-830. Dovidio, J.F., & Gaertner, S.L. (1981). The effects of race, status, and ability on helping behavior. Social Psychological Quarterly, 44, 192-203. - Erlich, H.J. (1973). The social psychology of prejudice. New York: Wiley. Katz, I., Wackenhut, J., & Haas, R. (1986). Racial ambivalence, value duality, and behavior. In J. Dovidio & S. Gaertner (Eds) Prejudice, discrimination, and racism (pp. 35-60). San Diego, CA: Academic Press. 64 65 Katz, P. (1976). The acquisition of racial attitudes in children. In P. Katz (Ed.), Toward the elimination of racism (pp. 125-154). New York: Pergamon Press. Gaertner, S.L. (1976). Nonreactive measures in racial attitude research: A focus on "Liberals". In P. Katz (Ed.), Toward the elimination of racism (pp. 183-211). New York: Pergamon Press. Gaertner, S.L., & Dovidio, J.F. (1986). The aversive form of racism. In J.F. Dovidio & S.L. Gaertner (Eds.) Prejudice, discrimination, and racism (pp. 61-89). San Diego, CA: Academic Press. ’ Hamilton, D.L. (Ed.) (1981 ). Cognitive Processes in stereotyping and intergroup behavior. Hillsdale, NJzErlbaum. Higgins, ET. (1987). Self-discrepancy theory: A theory relating self and affect. Psychological Review, 94, 319-340. Higgins, E.T,. & King, G. (1981). Accessibility of social constructs: Information processing consequences of individual and contextual variability. In N. Cantor & J.F. Kihlstrom (Eds.) Personality and social interaction (pp. 69-121). Hillsdale, NJ: Erlbaum. Jones, R.E., & Cooper, J. (1971). Mediation of experimenter effects. Journal of Personality and Social Psychology, 20, 70-74. Lipset, S.M., & Schneider, W. (1978). The Bakke case: How would it be decided at the bar of public Opinion. Public Opinion, 1, 38-44. McConahay, JG. (1986). Modern Racism, ambivalence, and the modern racism scale. In J.F. Dovidio & S.L. Gaertner (Eds.) Prejudice, discrinination, and racism (pp. 91-125). San Diego, CAzAcademic Press. McConahay, J.G., Hardee, B.B., & Batts, V. (1981). Has racism delined? It depends on who's asking and what is asked. Journal of Conflict Resolution, 25, 563-579. Mehrabian, A. (1967). Orientation behaviors and nonverbal attitude communication. Journal of Communication, 17, 324-332. 66 Mehrabian, A. (1968). Inference of attitudes from the posture, orientation, and distance of a communicator. Journal of Consulting and Clinical dechology. 32, 296-308. Myrdal, G. (1944). An American Dilemma, New York: Harper & Row. Sigall, H., 8: Page, R. (1971). Current stereotypes: A little fading, a little faking. Jorunal of Personality and Social Psychology, 18, 247-255. Tanaka, J., Panter, A., & Winborne, W. (1988). Dimensions of the Need for Cognition: Subscales and gender differences. Multivariate Behavioral Research, 23, 35-50. Taylor, D., Sheatsley, P., & Greeley A. (1978). Attitudes toward racial integration. Scientific American, 238 (6), 42-49. Watson, 0., Clark, L., & Tellegen, A (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology. 54, 1063-1070. Word, C.O., Zanna, M.P., & Cooper, J. (1974). The nonverbal mediation of self-fulfilling prophecies in interracial interaction. Journal of Experimental Social Psychology, 10, 109-120. APPENDIX A Should and Weuld Scenariee 1. Imagine a person boarded a bus and sat next to you. You , should/would feel uncomfortable someone is sitting next to you. 2. Imagine someone moved in next door to you. You should/would feel uncomfortable someone moved in next to you. 3. Imagine seeing three middle-aged men on the street corner. You should/would respond by thinking "how typical." 4. Imagine seeing a woman with three small children. You should/would respond by thinking "how typical." 5. Imagine arriving at a job interview. You should/ would be uncomfortable that someone was interviewing you. APPENDIX B In rviw in 1. What do you feel your qualifications are for this job? 2. What are your strengths? 3. What are your weaknesses? 4. Why should you get this job over other applicants? 5. Why should we hire you? 67 TABLE 1 Should Scores (Experiment 1)* Non-Reciel =leetificatien Faeter Absent Present Black White Black White ScerLatLQ One 1.97 2.48 3.52 3.63 Two 2.09 1.76 4.32 4.00 Three 1.84 2.36 3.48 3.75 Four 2.09 2.22 2.44 3.90 Five 1.78 3.72 3.72 5.09 *Scores on this measure ranged from 1 "positive response" to 7 "negative response." TABLE 2 Would Scores (Experiment 1)* Non-Racial Justification Fector Aeeent Present Blaek White Bleek White Seenefio One 2.40 3.68 4.32 4.44 Two 3.19 2.56 4.92 3.97 Three 2.56 2.76 4.32 3.78 Four 2.90 3.44 2.96 3.72 Five 1.69 4.44 4.80 5.09 . *Scores on this measure ranged from 1 "positive response" to 7 "negative response." 68 TABLE 3 Perceived Awareness of a Discrepancy (Experiment 1)* Bleelg, White NRJF Absent 6.44 7.09 NRJF Present 6.84 6.92 *Scores on this measure ranged from 1 "aware of a discrepancy" to 10 "unaware of a discrepancy." TABLE4 Actual Discrepancy Scores (Experiment 1) NRJF Aeeent Present Black White Bleek White Seem One 0.43 1.20 0.80 0.81 Two 1.09 ' 0.80 , 0.60 031 Three 0.71 ‘ 0.40 0.84 ' 0.03 Four 0.81 1.20 0.52 '-0.18 Five -0.09 0.72 1.08 0.0 69 NRJF Absent NRJF Present TABLE 5 Affect Scores (Experiment 1) gemeunetiom ' Diseomfort“ Bleek White Blagk White 14.76 15.62 14.09 15.52 18.56 14.48 14.28 ' 14.47 *Scores on this scale ranged from 8 "low compunction" to 42 "high compunction." ** Scores on this scale ranged from 6 "low discomfort" to 34 "high discomfort." 7O ‘1 TABLE 6 Should and Would Scores (Experiment 2) h l r Woul ores (Scenario 5)* (Behavior Rating)" NRJF Absent 2.97 7.84 NRJF Preeent 4.58 8.32 Bleek Target 2.83 8.10 White Target 4.66 8.06 Should Seeres Weuld Sceres (Scenario 5)* (BehaVior Rating)“ Blaek White Bleek White NRJF Absent 1.54 4.00 7.85 7.83 NRJF Present 3.82 5.50 8.29 8.36 *Scores on this measure ranged from 1 "positive response" to 7 "negative response." ** Would Scores were based on an 11-p0int scale where 0 = "not at all positive" and 10 = "totally positive." Scores on this scale ranged from 1 to 10. 71 TABLE 7 Should and Would Scores by Level of Prejudice (Experiment 2) Sh l r s Woul S r 5 (Scenario 5)* (Behavior Rating)** Bleek White Blaek White Low 2.47 4.71 8.18 7.71 Prejudice High 3.31 4.61 8.00 8.33 Prejudice * Scores on this variable ranged from 1 "postive response" to 7 "negative response." ** Would Scores were based on an 11-p0int scale where "0"=n0t at all positive and "10"=t0tally positive. * Scores on this variable ranged from 1 to 10. TABLE 8 Perceived Awareness of a Discrepancy (Experiment 2)* 3|ng White NRJF Absent 7.62 7.28 NRJF Present 7.53 h 6.50“ *Scores on this variable ranged from 0 "awarene of a discrepancy" to 10 "unaware of a discrepancy." 72 TABLE 9 Affect Scores (Experiment 2) gemeuncti0n* Discomfort“ Bleek White Black White NRJF - Absent 13.08 16.06 12.69 16.00 NRJF Present 12.24 15.07 14.41 13.86 *Scores on this scale ranged from 9 "low compunction" to 43 "high compunction." ' ** Scores on this scale ranged from 6 "low discomfort" to 37 "high discomfort." 73 "lllllllllllllllllllllli