Stereotype Susceptibility: Shifts in Quantitative Performance from Socio-Cultural Identification
Todd L. Pittinsky
Department of Psychology, Harvard University
Recent studies have documented that performance is hindered when individuals feel that a sociocultural group to which they belong is negatively stereotyped in a domain (Steele, 1997). We report that implicit activation of a social identity can facilitate as well as impede performance on a quantitative task. By making a particular social identity salient at an implicit level, performance was altered in the direction predicted by the stereotype associated with the identity. Common cultural stereotypes hold that Asians have superior quantitative skills compared to other ethnic groups and that women have inferior quantitative skills compared to men. We found that Asian-American women performed better on a mathematics test when their ethnic identity was activated, but worse when their gender identity was activated, compared to a control group who had neither identity activated. Cross-cultural investigation indicated that it is the stereotype, and not the identity per se, that influenced performance.
Stereotype Susceptibility: Shifts in Quantitative Performance from Socio-Cultural Identification
A series of recent studies have found that the implicit activation of sociocultural stereotypes can influence the performance of the stereotyped individual (Levy, 1996; Steele & Aronson, 1995). Steele and Aronson (1995) found that when told that a test was diagnostic of their abilities, African-American students, who are stereotyped to be poor students, under-performed relative to white students. In addition Levy (1996) found that elderly people performed worse on a memory task if they had previously been primed with a negative stereotype of the elderly, than if they had been primed with a positive stereotype of the elderly.
In this paper, we expand the work on the powerful effects of automatic and unconscious activation (see for example Banaji, Hardin & Rothman, 1993; Banaji & Greenwald, 1994; Bargh, Chen & Burrows, 1996) and self-application of stereotypes by examining whether the implicit activation of particular identities can facilitate as well as debilitate academic performance. We focus on two issues that social psychologists have largely overlooked but that merit investigation - the effects of the activation of different identities an individual may have and the positive effects of some stereotypes on performance. Past research on self-stereotyping has focused almost exclusively on only one dimension of participants identity. For example, both Levy's (1996) and Steele & Aronson's (1995) studies classified participants along only one dimension of social identity, age and race respectively. Yet people carry with them many rich dimensions of social identity (Hewstone, 1996) and different social situations and goals trigger different identities (Richeson & Ambady, 1998). For example, a white male Christian accountant may be identified as a male, as white, as a Christian or as an accountant. Because different social identities are associated with different stereotypes, individuals may be susceptible to different, and in some cases even conflicting stereotypes.
A second neglected issue in the research on self-stereotyping is the effect of positive stereotypes. Focusing only on the negative effects of stereotypes associated with particular social categories means that critical dynamics of how stereotypes affect individuals in the real world are ignored.
In the present paper, we considered both positive as well as negative stereotypes associated with different identities that can co-exist within an individual. Can implicit activation of different dimensions of identity lead individuals to access and behave in accordance with different stereotypes that might help or hinder their academic performance? We conducted two studies to examine whether subtly activating different self-stereotypes would affect the quantitative performance of Asian-American women. A common cultural stereotype of women, supported by test performance data, is that they have inferior quantitative skills compared to men (Benbow, 1988; Hedges & Nowell, 1995). A common stereotype about Asians, supported by test performance data, is that they have superior quantitative skills compared to other ethnic groups (Steen, 1995). These stereotypes suggests that the performance of Asian-American women in a quantitative domain is susceptible to two very different stereotypes.
In Study 1, we asked undergraduate Asian-American women to take a quantitative test. We hypothesized that their performance would be depressed when their female identity was made salient and enhanced when their Asian identity was made salient, compared with a control group of Asian-American women for whom no particular identity was made salient1. Identity salience was manipulated by having participants complete different versions of a questionnaire about residential life at their university. Participants randomly assigned to the female identity salient condition had to indicate their sex and answer questions related to their gender identity. Participants in the Asian identity salient condition were asked to indicate their ethnicity and to answer questions related to their ethnic identity. In the no identity salient control condition, participants were not asked to indicate their gender or their ethnicity, but were asked to answer questions unrelated to either identity. The questionnaires were constructed to make salient the identity of interest (ethnic or gender) implicitly, without directly priming the actual stereotype (superior or inferior quantitative skills).
Forty-five Asian-American female undergraduate participants performed a quantitative test. Participants were run individually in a laboratory session and first were asked to fill out the appropriate manipulation questionnaire by an experimenter blind to the manipulation. In the female identity salient condition (n =14), participants were asked (a) whether they lived on or off campus; (b) whether they had a roommate; (c ) whether their floors were coed or single sex; (d) whether they preferred coed or single sex floors; (e) to list 3 reasons why they would prefer a coed floor; and (f) to list 3 reasons why they would prefer a single-sex floor. In the Asian-identity salient condition, participants (n = 16) were asked (a) whether their parents or grandparent spoke any languages other than English; (b) what languages they knew; (c) what languages they spoke at home; (d) what opportunities they had to speak other languages on campus; (e) what percentage of these opportunities were found in their residence halls; and (f) how many generations their family had lived in America. In the control condition participants (n = 16) were asked (a) whether they lived on or off campus; (b) whether they used the university telephone service; (c) to rate on a 7-point scale how satisfied they were with the service; (d) whether they would consider subscribing to cable television; (e) how much they would be willing to pay per month for cable TV; and (f) to list one or two reasons why they would or would not subscribe to cable TV.
After the questionnaire, participants were given a quantitative test that consisted of 12 math questions from the Canadian Math Competition2 and were told that they would have 20 minutes to complete this task. At the end participants completed a final set of questions, indicating their own score on the quantitative section of the Scholastic Aptitude Test, how much they enjoyed participating in the experiment, how talented they were at mathematics, and their thoughts about the goals of the research. Participants were then paid and debriefed.
Results and Discussion
The main dependent variable was accuracy, which was the number of mathematical questions a participant correctly answered divided by the number of questions that the participant attempted to answer (Steele & Aronson, 1995). As shown in Table 1, our
Insert Table 1 about here
predictions were supported: performance on the quantitative test was the best in the Asian identity salient condition, followed by the control condition, and was the worst in the female identity salient condition. As can also be seen in Table 1, the performance increment associated with activation of the Asian-American dimension of identity was similar in magnitude to the performance decrement associated with the activation of the female dimension of identity. These findings are particularly compelling because the effect was exhibited among students who possess excellent quantitative skills (the mean reported math SAT of the participants was 750.9, standard deviation = 46.53).
Participants in the Asian identity salient condition answered an average of 54% of the questions that they attempted correctly, participants in the control condition answered an average of 49% correctly, and participants in the female identity salient condition answered an average of 42% correctly. A linear contrast analysis testing our prediction that participants in the Asian salience condition scored the highest, participants in the control condition scored in the middle, and participants in the female identity salient condition scored the lowest revealed that this pattern was significant (t (43) = 1.86, p < .05, r = .27). Participants in the Asian identity salient condition performed significantly better than participants in the female identity salient condition (t (29)= 2.02, p < .05, r = .35). When only the number of questions answered correctly are considered, we find the same, albeit less statistically significant, pattern of results. Participants in the Asian identity salient condition answered the most questions correctly (M = 5.37) while participants in the female identity condition answered the fewest questions correctly (M = 4.71) and participants in the control condition were in the middle (M = 5.31) (t(43) = .89, p = .19, r = .13). Accuracy , however, is a more meaningful dependent variable than the number of questions answered correctly because it not only takes into account the number of questions answered correctly but it also accounts for the number of questions attempted.
Interestingly, the salience manipulations did not affect the level of motivation of the participants which was similar across all conditions: there were no significant differences across conditions in (a) the number of questions participants attempted to answerthe effort that they exhibited; (b) the number of questions for which participants reported guessing; ( c) participants liking of the test; (d) their assessment of how well they did on the test; (e) their assessment of test difficulty and (f) their assessment of their mathematical skills3. It is therefore unlikely that the negative effects of self-stereotyping in the female identity salient condition could have been decreased by simply encouraging participants to try harder. These results replicate those of Steele and Aronson (1995) who also did not find differences in the number of questions attempted. Further, individuals in the female identity salient condition attempted just as many questions and were equally confident in their performance. Yet their performance was significantly poorer compared to their counterparts who had their Asian-American identity implicitly made salient, or had no identity made salient. Finally, there were no differences across conditions in how well participants thought they did on the test. Thus, participants were not aware that their performance was being affected by the brief questionnaire implicitly activating different identities. Further, we found that participants were not aware that there was a target identity being made salient, and did not know the hypothesis of the study. This evidence indicates that the stereotypes and identity activation were operating at an implicit level. These results suggest that it is unlikely that anxiety over what the experimenter may be thinking is driving this effect.
There were also no significant differences in participants reported Quantitative Scholastic Aptitude Test (SAT) scores across conditions. The mean Math SAT score for the sample was 750.9 with a minimum score of 600 and a maximum score of 800, revealing an extremely restricted range of scores. The mean score for the general population on the quantitative scale of the SAT for 1996 was 508 with a standard deviation of 110 (SAT summary reporting service, 1996). The narrow range of scores and the high scores for this sample indicate that these were all highly gifted women. Obtaining significant changes in performance with such a highly gifted sample with an implicit identity activation task makes these results all the more striking. Correcting for restricted range makes the effect stronger (Snedecor & Cochran, 1980). Using the 1996 SAT standard deviation reported by the Educational Testing Service to correct for restriction of range yields, t (43) = 2.37, p = .01, r = .34 for the main hypothesis.
Study 2 examined the hypothesis that it is the stereotype associated with an identity (i.e. Asians are talented at math.) and not the identity itself (i.e. Asian) that drives this performance differential. In order to do this, we replicated the study in a second culture where different stereotypes were associated with these same identities: Vancouver, Canada, where the Asian community is largely a recently immigrated one. A questionnaire designed to examine prevailing ethnic and gender stereotypes confirmed that in Vancouver there is less of a stereotype that Asians have superior quantitative skills compared to other ethnic groups. We compared American and Canadian stereotypes for Asians being quantitatively gifted and found that the stereotype prevailed more in America than in Canada (t (81) = 2.07, p <.05, r = .22)4.
Using the same paradigm as in Study 1, we predicted that the quantitative test performance of Asian women from Vancouver, Canada would be depressed relative to a control group when their gender identity was made salient. We also predicted that, unlike the US sample, performance would not be enhanced relative to a control group when their Asian identity was made salient. We ran 19 Asian-Canadian female high school students using the procedure reported for Study 1. Participants in the female identity salient conditions in Canada performed worse than participants in both the Asian identity salient and the control condition.
Results and Discussion
Results support the hypothesis that stereotypes and not just identities influence performance. In this population, the stereotype that Asians possess superior quantitative skills is not prevalent. As predicted (see Table 1), participants in the female identity salient condition performed the worst (28% correct) of the three groups. Participants in the Asian identity salient condition (44% correct) also performed worse than a control group (59% correct), contrast (t (16) = 4.55, p < .0005, r = .75).
When considering only the number correct, we find a similar pattern. As predicted (see Table 1), participants in the female identity salient condition answered the fewest questions correctly (M = 3.00) of the three groups. Participants in the Asian identity salient condition also answered fewer questions correctly (M = 4.50) than the control group (M = 5.00). A contrast using lambda weights of 1, 1, -2, revealed that the pattern was significant (t (16) = 2.43, p = .01, r =.52).
Combination of the two samples
When we combine the two studies and run an ANOVA on the percentage correct, we find a significant country by condition interaction effect (F(2, 60) = 2.78, p = .069, eta = .30) 5. such that the pattern of performance across the three conditions was significantly different in the two countries. This suggests that the identity salience manipulations had different effects on performance of the participants in the United States and Canada.
This work demonstrates the powerful influence of sociocultural stereotypes on individual performance. In the present study, participants were not explicitly primed with stereotype content, but simply had a sociocultural category to which they belong subtly activated. Perhaps, most significant, we find evidence that when an identity is made salient at an implicit level, performance can be facilitated as well as debilitated. Previous work has shown that the performance of women on quantitative tasks is impeded when they are told that the task generally showed gender differences but not when they are told that the task was insensitive to gender differences (Aronson, Quinn & Spencer, in press; Steele, 1997). The present research indicates that womens quantitative performance can be affected both positively as well as negatively without any explicit instructions. While it remains disturbing that implicitly activating a female gender identity can inhibit performance, it is encouraging that implicitly activating certain ethnic identities can help some individuals perform.
Our results speak to a number of applied areas of great concern to the scientific community as well as the general public, including womens performance in quantitative fields such as math and science. They suggest that more attention needs to be paid to sociocultural influences on the performance of women (Barinaga, 1994), particularly in the light of recent findings regarding sex differences in mathematics and science (Hedges & Nowell, 1995) in a number of different cultures (Lummis & Stevenson, 1990; Stevenson, Chen, & Lee, 1993).
The results of this study also suggest that when the fact that people have multiple identities is addressed, self-stereotyping effects may be approached far more strategically than previously recognized. To date, individuals have been largely cast as victims of self-stereotyping. But the possibility exists that interventions, in this case an experimental manipulation that made one dimension of identification salient over others, could be used to strategically impact performance. The present finding also hints at intriguing unexplored questions about how individuals may experience identity salience and self-stereotyping processes. Brewer's (1991) theory of "optimal distinctiveness" posits that individuals identify along particular dimensions to achieve a state of optimal distinctiveness. It seems plausible that individuals may also identify themselves along dimensions in order to achieve "optimal adaptiveness."
Finally, findings that academic performance can be helped as well as hindered through implicit shifts in identification raises important challenges to notions of academic performance and intelligence. Although there is considerable debate about the nature of intelligence (Fraser, 1995; Neisser, Boodoo, Bouchard, Boykin, Brody, Ceci, Halpern, Loehlin, Perloff, Sternberg & Urbina, 1996), strong supporters of genetic differences in IQ assume that ability is fixed and can be quantified through testing (Herrnstein & Murray, 1994). The results presented here clearly indicate that test performance is both malleable and surprisingly susceptible to implicit sociocultural pressures.
TABLE. 1. Accuracy by condition for Study 1 and Study 2
Asian Identity Salient
No Identity Salient
Female Identity Salient
Mean = 0.54
St. Dev. = 0.17
Mean = 0.49
St. Dev. = 0.20
Mean = 0.43
St. Dev. = 0.16
Mean = 0.44
St. Dev. = 0.17
Mean = 0.59
St. Dev. = 0.20
Mean = 0.28
St. Dev. = 0.16
Aronson, J., Quinn, D. M., & Spencer, S. J. (in press). Stereotype threat and the academic underperformance of minorities and women. In J. Swim & C. Stangor (Eds.). Prejudice: The targets perspective. Academic Press.
Banaji, M. R., Hardin, C. & Rothman, A. J. (1993). Implicit stereotyping in person judgment. Journal of Personality & Social Psychology, 65, 272-281.
Banaji, M. & Greenwald, A. G. (1994). Implicit stereotyping and prejudice. In Zanna, M. P. & Olson, J. M. (Eds.), The psychology of prejudice: The Ontario symposium, (Vol. 7., pp. 55-76). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
Bargh, J. A., Chen, M. & Burrows, L. (1996). Automaticity of social behavior: Direct effects of trait construct and stereotype activation on action. Journal of Personality & Social Psychology, 71, 230-244.
Barinaga, M. (1994). Surprises across the cultural divide. Science, 263, 1468-1472.
Benbow, C. P. (1995). Sex differences in mathematical reasoning ability in intellectually talented preadolescents: Their nature, effects, and possible causes. Behavioral and Brain Sciences, 11, 169-232.
Brewer, M.B. (1991). The social self: On being the same and different at the same time. Personality and Social Psychology Bulletin, 17, 475-482.
Fraser, S. (Ed.). (1995). The Bell Curve Wars. New York: Basic Books.
Greenwald, A.G. & Banaji, M.R. (1995). Implicit social cognition: Attitudes, self-esteem, and stereotypes. Psychological Review, 102 , 4-27.
Hedges, L. V. & Nowell, A. (1995). Sex differences in mental test scores, variability, and numbers of high-scoring individuals. Science, 269, 41-45.
Herrnstein, R. J., & Murray, C. (1994). The Bell Curve . New York: Free Press.
Hewstone, M. (1996). Contact and categorization: Social Psychological interventions to change intergroup relations. In C.N. Macrae, Stangor, C., and Hewstone, M. (Eds.), Stereotypes and Stereotyping. New York: Guildford.
Levy, B. (1996). Improving memory in old age through implicit self-stereotyping. Journal of Personality and Social Psychology, 71, 1092-1107.
Lummis, M., & Stevenson, H. W. (1990). Gender differences in beliefs and achievement: A cross-cultural study. Developmental Psychology, 26, 254-263.
Neisser, U., Boodoo, G., Bouchard, T. J., Boykin, A. W., Brody, N., Ceci, S. J., Halpern, D. F., Loehlin, J. C., Perloff, R., Sternberg, R. J., & Urbina, S. (1996). Intelligence: Knowns and unknowns. American Psychologist, 51, 77-101.
Richeson, J.A. & Ambady, N. (1998). Roles vs. Goals: When race matters in dyadic interactions. Manuscript submitted for publication.
SAT Summary Reporting Service (1996). 1996 Profile of College-Bound Seniors National Report, College Board Home Page : ETS.
Snedecor, G. W. & Cochran, W. G. (1980). Statistical Methods. (7th. Ed.). Ames, Iowa: Iowa State University Press.
Steele, C. M. (1997) . A threat in the air: How stereotypes shape intellectual identity and performance. American Psychologist, 52 , 613-629.
Steele, C.M. & Aronson, J. (1995). Stereotype threat and the intellectual test performance of African-Americans. Journal of Personality and Social Psychology, 69, 797-811.
Steen, L. A. (1987). Mathematics education: A predictor of scientific competitiveness. Science, 237, 251-253.
Stevenson, H. W., Chen, C. & Lee, S-Y. (1993). Mathematics achievement of Chinese, Japanese and American children: Ten years later. Science, 259, 53.
We gratefully thank J. Richard Hackman, Bob Rosenthal, and Karen Ruggiero for their comments and guidance. We also thank Tina Chen, Stacey DiCicco, Andrea Lewis and Karen Shih for assistance with data collection. This work was supported, in part, by a NIH grant # R03 MH05H549-01 (USA) and the Social Sciences and Humanities Research Council (Canada).
1. One can conceive of a fourth condition in which both gender and ethnicity are made salient in order to see if one identity would predominate. This is a provocative questions, but one beyond the scope of the present study in which we test whether we can alter performance by making one identity more salient than another rather than simultaneously activating multiple identities.
2. The Canadian Math Competition is a prestigious annual math competition for high school students sponsored by the University of Waterloo.
3. A one-way ANOVA was conducted to test the differences.
4. Randomly selected people (n=50) in Vancouver, B.C., Canada and Cambridge, Massachusetts, U.S.A. were asked to rate on a scale of 1-7 how common the stereotype "Asians are good at math" is in their society.
5. There was no significant country by condition interaction when we only considered the number of questions answered correctly. This is not surprising because of the difference in education level of the two samples. The Canadian participants were high-school students while the American participants were University level students. The American participants answered significantly more questions correctly (M = 5.12) than the Canadian participants (M = 4.11) ( F (1, 60) = 3.84, p = .05, eta = .25).