Back to Journal Cover Page


CURRENT RESEARCH IN SOCIAL PSYCHOLOGY


Volume 8, Number 19
Submitted: June 18, 2003
First Revision: June 24, 2003
Accepted: June 25, 2003
Publication Date: June 25, 2003

Aversive Bias Toward Gay Men?

Christopher L. Aberson
Humboldt State University

ABSTRACT

The aversive racism paradigm predicts that African Americans are evaluated unfavorably only when bias can be justified on non-category-based grounds. The current study applies the aversive racism to evaluations of gay men through examination of evaluations of either highly qualified, unqualified, or ambiguously qualified male job candidates who are gay or heterosexual. Analyses based on a sample of 195 U.S. college students found no differences in ratings of gay and heterosexual job candidates within any of the qualification conditions. Only candidate qualifications influenced ratings. A plausible interpretation of this result is that is aversive biases may not apply to evaluations of gay men in U.S. college student samples. Discussion addresses potential explanations for the differences between reactions to ethnicity and sexual orientation.

[266]
---------------
[267]

INTRODUCTION

The aversive racism perspective suggests that negative attitudes towards minority groups are acquired early in life, resulting from immersion in a society with a long history of racial bias (Dovidio & Gaertner, 1991). However, there are pressures against overt expression of negative attitudes towards minority group members. The conflict between negative attitudes and sanctions against the expression of these attitudes results in complex forms of biased expression (Dovidio & Gaertner, 1986). For example, negative evaluations of African Americans are unlikely in situations where a negative evaluation could bring about attributions of bigotry to the evaluator. In situations where judgments would appear based solely on race, such as evaluating an African-American less favorably than a White when all else was equal, bias is unlikely. However, when aspects of the situation allow the individual to rationalize evaluations based on non-racial characteristics, less favorable evaluations are more likely (Dovidio & Gaertner, 1991).

Dovidio and Gaertner (2000) provide a prototypical demonstration of the aversive racism phenomenon. Participants rated an African American or White job candidate who was either strongly, ambiguously, or not at all qualified. When rating unqualified and strongly qualified applicants, ratings were not influenced by ethnicity. In these instances, evaluation criteria are clear, either the candidate is obviously qualified or unqualified. However, ratings of candidates with ambiguous qualifications (i.e., unclear as to whether qualified or not) indicated stronger support for White applicants. This result suggests that individuals attempt to be egalitarian in their evaluations, however, when the situation was ambiguous, participants exhibited bias against African American applicants.

Applying Modern Forms of Racism to Sexual Prejudice

The present study applies the aversive racism paradigm to sexual prejudice, specifically discrimination against gay men. There are several parallels between attitudes towards gays and African-Americans. Sexual prejudice is similar in origin and form to prejudice directed against other groups (Ficarratto, 1990). Gay men are a group characterized by negative stereotypes and attitudes (D’Augelli, 1989). Similar to bias against ethnic minorities, expression of bias toward gay men has become less overt as society increasingly condemns such negative attitudes. Increased societal acceptance, combined with a long history of negative attitudes, characterizations, and institutionalized prejudice may produce reactions toward gay men similar to those found toward African-Americans. The roots of aversive racism toward African-Americans results from the conflict between egalitarian beliefs and a societal tradition of racism. Similarly, egalitarian beliefs conflicting with traditional societal values favoring heterosexual men may produce similar biases against gay men (Herek, 1984).

An earlier study investigated the role of aversive bias against gay men. Participants viewed videos of job candidates who were either gay or heterosexual and either rude or pleasant. Evaluations of the rude candidate (i.e., a reason to be biased) found no differences in ratings of gay and heterosexual candidates, failing to support an aversive bias perspective (Aberson, Swan, & Emerson, 1999). Though this study provides some test of an aversive bias hypothesis, the rude candidate condition could be considered negatively valued rather than ambiguous, thus confusing the interpretation of the result.

The current study asks whether reactions to gay men follow a pattern predicted by aversive bias. If evaluations do follow this pattern, gay men will be evaluated less favorably than heterosexual men when evaluation criteria are ambiguous. When evaluation criteria are clear, no differences in ratings between gay and heterosexual men are predicted.

METHOD

Participants

Two hundred and twenty-five undergraduates at a U.S. public university participated in a study of hiring decisions. Analyses excluded 30 participants who identified themselves as gay/lesbian/bisexual, leaving 195 cases. Participants were primarily White (76.4%) and women (68.7%), with a median age of 20. Based on an average effect size of d = .65 found in Dovidio and Gaertner (2000), I chose a sample of 30 participants per cell to ensure power of .80 for mean comparisons (cf. Cohen, 1988). As I planned to examine only results for heterosexual participants, several cells included more than 30 participants to ensure adequate sample size.

[267]
---------------
[268]

Procedure and Measures

Participants received a questionnaire with instructions informing them that the study examined hiring decisions and the types of information people use to make decisions. Instructions indicated responses were anonymous and the participant could withdraw at any time without penalty, and that they must be 18 or older to participate. Participants read a hiring scenario, a summary of an applicant’s qualifications, rated the applicant on several items, and indicated their own and the applicant’s sexual orientation (manipulation check).

Scenario

Participants read the following instructions: "Imagine that you are acting as the student representative for your school’s HIV/AIDS awareness campaign. One of the tasks facing the committee is to hire an individual to run the program. The job involves organizing educational events and safe sex campaigns, giving lectures on safe sex, and working directly with student organizations." Participants read a summary of a job candidate’s qualifications and a personal statement. All candidates were men.

Qualification manipulation

Questionnaires varied candidate qualifications through manipulation of job experience and answers to interview questions. The highly qualified candidate’s resume included experience as an assistant in a position with responsibilities similar to those required of the current position. For the ambiguously qualified candidate, no information about experience was presented. For the unqualified candidate, experience included only passing out free condoms as a volunteer for a campus AIDS education program. The questionnaire further established qualifications through presentation of answers to interview questions. Each candidate responded to the following question: "If a student came to you because she recently learned that an ex-boyfriend had tested HIV-positive, what would you do?" Qualified candidates answered that they would "explain her options to her and ask her if she would like the telephone number of the health center." Ambiguously qualified candidates responded, "ask her if she would like the telephone number of the health center." Unqualified candidates answered, "tell her that it is too personal and that she must talk with her parents." These manipulations are modified from Dovidio and Gaertner (2000).

Sexual orientation manipulation

Candidate personal statements manipulated sexual orientation. The candidate indicated that he became interested in HIV/AIDS education when a former sexual partner became infected with HIV. The candidate was tested and learned that he was not infected. Personal statements manipulated sexual orientation by referring to the former partner as an ex-girlfriend (heterosexual condition) or an ex-boyfriend (gay condition).

Candidate ratings

Participants rated the candidate on three items taken from Dovidio and Gaertner (2000). These items included "how qualified is the candidate for this position?" (0 = not at all to 10 = completely), "would you recommend the applicant for this position" (0 = no, 1 = yes), and "how strongly would you recommend this applicant for the position?" (0 = not at all to 10 = very strongly). An eight-item evaluative rating scale, developed by the author, including "I believe this candidate would be able to relate to college students" and "I think that this person could have a negative impact on the education program" comprised an evaluation scale (0 = disagree completely to 10 = agree completely). Reliability for the 8-item scale was good (alpha = .91).

The evaluative rating scale, and the recommendation and strength of recommendation questions serve as three dependent measures used in hypothesis tests. The qualification-rating item comprises a manipulation check.

[268]
---------------
[269]

RESULTS

Assumptions

The evaluation scale (responses to the 8-items) was negatively skewed. A reflected logarithmic transformation normalized the distribution (cf. Tabachnick & Fidell, 2001). Following transformation, variances between cells were adequate to meet homogeneity of variance assumptions.

Manipulation Checks

A 3 (High, Ambiguous, Low Qualifications) x 2 (Gay vs. Heterosexual Candidates) ANOVA examined the impact of qualifications and sexual orientation on ratings of candidate qualifications. Supporting the qualification manipulation, participants rated the qualifications of unqualified (M = 4.9, SD = 2.3), ambiguously qualified (M = 6.7, SD = 2.0), and highly qualified (M = 8.0, SD = 1.6) differently, F(2, 188) = 39.5, p < .001, eta-squared = .30. Tukey comparisons indicated that all pairs differed significantly and in the expected direction (all p’s < .01).

The research scenario portrayed individuals who applied for an HIV/AIDS education position. It would be reasonable to expect that participants viewed gay applicants as more qualified because they are more likely to have personal or social experiences related to HIV/AIDS. Qualification ratings however revealed no differences between gay (M = 6.4, SD = 2.5) and heterosexual (M = 6.6, SD = 2.3) candidates, F(1, 188) < 1, p = .60, eta-squared = .00, nor was there an interaction between sexual orientation and qualifications, F(2, 188) < 1, p = .67, eta-squared = .00. This result suggests sexual orientation was not related to perceptions of qualification and supports the use of this scenario in the study.

With regard to the sexual orientation manipulation, only 85.6% of participants correctly identified the candidate as heterosexual or homosexual. Participants did not differ in misidentification of heterosexual (9.3% incorrect) and gay (16.8%) candidates, chi-square (1) = 2.4, p = .12. Nor were differences between classifications of low (9.4% incorrect), ambiguous (15.9%), or highly qualified (13.8%) candidates, chi-square (2) = 1.2, p = .54. To address the potential for differences across conditions not uncovered by these analyses (i.e., an interaction), I examined classification across the six conditions. This analysis also revealed no differences, chi-square (5) = 4.4, p = .49.

For each analysis reported below, I examined all heterosexual participants and only those heterosexual participants who correctly identified the applicant’s sexual orientation. In all cases, analyses produced conceptually identical results. Only results for the full set of heterosexual participants appear below.

Main Analyses

Table 1 presents descriptive statistics for the three dependent measures by condition. Most relevant to aversive bias hypotheses are qualification by sexual orientation interactions. For the strength of recommendation measure, there was no interaction, F(2, 189) = 1.0, p = .36, eta-squared = .01, suggesting that there was no aversive bias effect. There was a main effect for qualifications, F(2, 189) = 51.8, p < .001, eta-squared = .35, revealing that stronger qualifications yielded stronger hiring recommendations. Sexual orientation did not influence ratings, F(1, 189) < 1, p = .94, eta-squared = .00.

[269]
---------------
[270]

Table 1: Target Ratings by Qualifications and Sexual Orientation

 

Strength of Recommendation

Percentage Recommended

Evaluative Scale

Condition

     

Strong Qualifications

     

Heterosexual (n = 35)

7.49 (1.90)

88

8.15 (1.59)

Gay (n = 31)

7.65 (1.74)

97

8.06 (1.50)

Ambiguous Qualifications

     

Heterosexual (n = 32)

6.78 (1.66)

84

7.61 (1.44)

Gay (n = 31)

6.13 (2.66)

77

7.17 (1.99)

Weak Qualifications

     

Heterosexual (n = 33)

3.52 (2.64)

27

4.84 (2.01)

Gay (n = 33)

3.94 (2.52)

27

5.30 (2.13)

Note: Means for evaluative scale presented as untransformed to retain interpretability, analyses use reflected log transformed dependent variable. Standard deviations are in parentheses. Higher values on each item represent evaluations that are more positive.

Scores on the evaluation scale revealed similar results. There was no orientation by qualification interaction, F(1, 189) <1, p = .48, eta-squared = .01, and no effect for sexual orientation, F(1, 189) < 1, p = .87, eta-squared = .00. Again, qualifications significantly influenced ratings, F(2, 189) = 43.3, p < .001, eta-squared = .31.

Logistic regression analyses examined recommendation ratings as this dependent measure comprised a dichotomous outcome (yes/no). Again, strength of qualification significantly influenced ratings, Chi-square (2) = 73.5, p < .001. Neither the interaction, Chi-square (2) = 2.3, p = .32, nor sexual orientation, Chi-square (1) = 0.01, p = .91, significantly predicted recommendation.

Planned Contrasts

Despite failing to obtain significant qualification by sexual orientation interactions, I performed planned contrasts on the two conditions most relevant to aversive bias, those comparing the gay and heterosexual candidates with ambiguous qualifications. These tests provide a liberal examination of the effects. Also presented are effect sizes and confidence intervals around each effect size (e.g., Smithson, 2002; Thompson, 2002). These confidence intervals are based on non-central t and chi-square distributions and provide estimates of the population effect size. For strength of recommendation and overall rating, contrasts found no differences between gay and heterosexual candidates, t(189) = -1.0, p = .31, d = -.15, 95% CI = -.14 delta .44, and t(189) = 0.7, p = .48, d = -.10, 95% CI = -.18 delta .39, respectively. Similarly, percentage recommendation did not differ between these conditions, Chi-square (1) = 0.5, p = .48, Cramer’s V = .09, 95% CI = .00 V .36. The Cramer’s V measure ranges from 0 to 1.0. The magnitude of the statistic may be interpreted like a correlation coefficient but it does not produce negative values. Converted to a d statistic, the Cramer’s V values reflect a 95% CI = .00 delta .77.

[271]
---------------
[272]

I found no differences in our contrast conditions using ANOVA or a liberal contrast procedure. For each test, the largest plausible value of the effect size would yield only a moderate size effect. The average of the upper limits of the three estimates is d = .51. This value is smaller than the average effect size of d = .65, found in Dovidio and Gaertner (2000). This result suggests that the aversive bias effect, if present at all, is weaker for evaluations of gay men than for evaluations of African Americans.

DISCUSSION

The primary finding of this study is that aversive biases were not present in evaluations of gay men. As the aversive racism paradigm suggests that aversive biases result from deeply ingrained biases against African Americans, this type of bias may not apply to groups with different histories of discrimination. Aversive racism is rooted in fundamental categorization processes (e.g., us vs. them; Gaertner & Dovidio, 2000). A category, such as sexual orientation, lacks the visual cues found in ethnic categories, possibly producing less automatic categorization. If this form of categorization is less prevalent, it may reduce the category-based classifications that lead to aversive bias.

Though there are similarities between heterosexist attitudes and racism, there are also important differences. For example, attitudes towards gay men are often characterized by beliefs that gay men are mentally ill and that homosexuality is pathological (Herek, 2002). These types of beliefs are not associated with attitudes towards African Americans. Given these differences, applying racially derived theories to attitudes towards gay men may not be appropriate.

Several limitations and alternative explanations temper results. It may be the case that aversive bias does apply to evaluations of gay men but this effect is smaller than that found for evaluations of African Americans. For this experiment, the average effect size was d = .22, a much smaller value than that found in previous research (e.g., Dovidio & Gaertner, 2000; d = .65). Another limitation may be the research scenario. Our scenario described applicants for an HIV/AIDS education program. It is possible that participants more closely associate gay men with HIV/AIDS, resulting in more positive evaluations of gay candidates. Similarly, the research scenario portrayed reactions to a woman who feared she had contracted HIV from an ex-boyfriend. As this situation is explicitly heterosexual, participants may have recommended the heterosexual candidate more strongly in these conditions. Examination of qualification ratings somewhat nullifies these concerns. There were no differences in perceived qualifications of gay and heterosexual applicants, suggesting that participants did not view either as more qualified for the position. Still, future research could utilize a more neutral scenario. Finally, all findings are qualified by the sample of U.S. college students.

Despite the limitations, the study sample does afford enough sensitivity to detect effects of the magnitude found in previous studies. As such, given the lack of differences found on three separate dependent measures, it may be the case that aversive bias effects are not present in evaluations of gay men or that these effects are smaller than those found in reactions to African Americans. Given these findings, the applicability of the aversive racism framework to attitudes towards gay men may be limited.

References

Aberson, C. L., Swan, D. J., & Emerson, E. P. (1999). Covert discrimination against gay men by U.S. college students. The Journal of Social Psychology, 139, 323-334.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). New York: Academic Press.

D’Augelli, A. R. (1989). Homophobia in a university community: Views of prospective resident assistants. Journal of College Student Development, 30, 546-552.

Dovidio, J. F., & Gaertner, S. L. (1986). Prejudice, discrimination, and racism: Historical trends and contemporary approaches. In J. F. Dovidio & S. L. Gaertner (Eds.), Prejudice, discrimination, and racism (pp. 1-34). Orlando, FL: Academic Press.

[272]
---------------
[273]

Dovidio, J. F., & Gaertner, S. L. (1991). Changes in the expression and assessment of racial prejudice. In H. J. Knopke, R. J. Norrell, & R. W. Rogers (Eds.), Opening doors: Perspectives on race relations in contemporary America (pp.119-148). Tuscaloosa, AL: University of Alabama Press.

Dovidio, J. F., & Gaertner, S. L. (2000). Aversive racism in selection decisions: 1989 and 1999. Psychological Science, 11, 315-319.

Ficarrotto, T. J. (1990). Racism, sexism, and erotophobia: Attitudes of heterosexuals toward homosexuals. Journal of Homosexuality, 19, 111-117.

Gaertner, S. L., & Dovidio, J. F. (2000). Reducing ingroup bias: The common ingroup identity model. Philadelphia, PA: The Psychology Press.

Herek, G. M. (1984). Beyond "homophobia": A social psychological perspective on attitudes toward lesbians and gay men. Journal of Homosexuality, 10, 1-21.

Herek, G. M. (2002). Gender comparisons in public opinion about lesbians and gay men. Public Opinion Quarterly, 66, 40-66.

Smithson, M. J. (2002). Confidence intervals. Thousand Oaks, CA: Sage.

Thompson, B. (2002). What future quantitative social science research could look like: Confidence intervals for effect sizes. Educational Researcher, 31(3), 24-31.

Tabachnick, B., & Fidell, L. (2001). Using multivariate statistics (4th ed). Boston: Allyn and Bacon.

APPENDIX A: Descriptive Statistics and Correlations Among the Dependent Variables

 

M

SD

n

1

2

3

1

Strength of recommendation

5.9

2.7

195

     

2

Would you recommend this applicant?

1.3

0.5

194

-.86***

3

Evaluative Rating

6.8

2.2

195

-.84**

.73***

 

Note: Untransformed overall means presented. Correlations represent the reflected, transformed variable so direction is reversed. For "would you recommend" 1 = yes, 2 = no. *** p < .001.

[273]
---------------
[274]

APPENDIX B: Scale Items for Dependent Measures

Please rate your perceptions of the candidate in terms of the following questions. Use the 0 to 10 scale below where 0 would indicate that you disagree completely with the statement and 10 indicates that you agree completely with the statement. Please only use whole numbers, do not indicate a range:

0

1

2

3

4

5

6

7

8

9

10

Disagree Completely

         

Agree Completely

1. I believe this candidate would be able to relate to college students

___________

2. I think that this candidate would not be the right person to hire.

___________

3. I believe that this candidate would be reliable.

___________

4. I believe that this candidate isn’t suited to the job.

___________

5. I believe this candidate would be effective.

___________

6. I’m concerned about the candidate’s past problems.

___________

7. I believe this candidate would be a good addition to an HIV/AIDS education program.

___________

8. I think that this person could have a negative impact on the education program.

___________

9. Would you recommend this applicant for the position? Yes No

10. How strongly would you recommend this candidate for the position?

0

1

2

3

4

5

6

7

8

9

10

Not at all

         

Very Strongly

Note. Items 1 through 8 comprise the evaluative rating scale. Items 2, 4, 6, and 8 are reverse coded. Item 9 is the "would you recommend" item. Item 10 is the strength of recommendation item.

Author Biography

Chris Aberson (cla18@humboldt.edu) is an Assistant Professor of Psychology at Humboldt State University. Chris completed his Ph.D. in Social Psychology at Claremont Graduate University in 1999. His current research interests include intergroup relations and interactive tutorials for teaching core statistics.

[274]
---------------
[275]