Whistling Vivaldi

Wide-ranging, subconscious effects of stereotype threats---and how to combat them

Posted: 2016-06-05

Tags: book, psychology, society, stereotype

These are my notes on Whistling Vivaldi: And Other Clues to How Stereotypes Affect Us by Claude Steele.

The book is very revealing, showing through many careful experimental studies how stereotypes have wide-ranging effects beyond our conscious awareness—for example, how just simply being in a society where there are stereotypes can depress academic performance of black students or math scores of female students. (Incidentally, this makes a strong argument for affirmative action in certain cases.) Moreover, it proposes several steps we can take counteract these stereotypes.

1 An introduction: at the root of identity


Claude Steele first realized he was black when he learned blacks couldn’t swim at the neighborhood pool except on Wednesday afternoons. This is an example of an identity contingency, something a person deals with because of a given social identity. The book will show that identity contingencies have a strong effect on standardized test scores, memory, comfort level, etc. A central identity contingency is stereotype threat.


Brent Staples, an African American man, recounts how when he walked down the streets of Chicago Hyde Park, people were frightened of him (ex. moving to the other side of the street). When he whistled Vivaldi’s Four Seasons, however, he diffused the negative stereotype—people stopped being afraid of him. The stereotype of violent blacks was no longer applicable to him personally.


White students who were told that a golf task measures “natural athletic ability” did worse what white students in the control group, because there is a stereotype that compared with blacks, whites have less natural athletic ability.

Black students who were told that a golf task measures “sports strategic intelligence” did worse, because there is a stereotype of blacks being less intelligent.


Steele’s research shows the following general ideas:

  1. Even though we think of ourselves as autonomous individuals, contingencies tied to our social identities have an undeniable effect on our lives.
  2. Identity threats play an important role in society’s most important social problems.
  3. These threats impair a broad range of human functioning through predictable psychological effects.
  4. There are things we can do as individuals to reduce the impact of these threats.

Edward Jones and Richard Nisbett argued that when explaining people’s behavior, the “observer’s perspective” and “actor’s perspective” are very different. An observer emphasizes things about the actor (characteristics, traits) while an actor is aware of the circumstances.

During his visit to Ann Arbor, Steele talked to minority students. They were concerned about the university environment and talked about being a minority. They talked about how social life was organized by race, expressed a desire for a place where they weren’t made aware of being a minority, and worried about how others see them. They didn’t have low expectations or make excuses for academic performance. Steele saw that

among students with comparable academic skills, as measured by the SAT, black students got less of a return on those skills in college than other students.


Among their six closest friends, neither white nor black students averaged even one friend from the other racial group.


Underperformance didn’t seem to be caused primarily by racism by teachers and classmates, nor by motivational nor cultural defecits. Rather, it was caused by racialized aspects of campus life: racial marginalization, racial segregation of networks, group underrepresentation, etc. Social groups are important; for example, over 85% of Americans get jobs through acquaintances. Ideas of who and what was “cool” were dominated by whites.


Jane Elliott conducted an experiment where she treated blue-eyed and brown-eyed students differently, and found that the marginalized group performed significantly worse—they paid less attention, didn’t remember instructions, were slow, and got many answers wrong. Students’ environment seemed to be an actual component of ability.

How is this relevant outside of experiments? Schools inherit a social organization from the larger society and their own history that place black students under “downwardly constituting pressures” which can interfere directly with intellectual performance.


Steele hypothesizes that stigmatization is the primary cause of underperformance (Nancy Hewitt, Elaine Seymour).

Women report a “chilly climate” in advanced math classes, fearing their abilities are under suspicion, feminine characteristics discredit their seriousness, and that they have to prove themselves constantly.

Steele gave male and female students who performed equally well in math a section of the math GRE, and likewise for English literature. Taking a difficult math test puts a woman at risk for stigmatization because of the stereotype that women are worse at math.

Steele told one group

You may have heard that woman don’t do as well as men on difficult standardized math tests, but that’s not true for the particular standardized math test; on this particular test, women always do as well as men.

This group did much better, equally as well as men.

(“Sterotype threat” is a worry of confirming a negative stereotype of one’s group.)


At a conference on women in science, Larry Summers (former president of Harvard) hypothesized “differential availability of aptitude at the high end” as a cause for disparaties between male and female performance. This was blown up by the media and he was pressured to resign (this was not the only cause). Steele’s experiment differentiated between the hypothesis of stereotype threat and this “second hypothesis” (underperformance is caused by something about women).

3 Stereotype threat comes to light, and in more than one group


When the coach for the Seattle Supersonics was replaced, the team went from a losing to a winning streak. The sportscasters (from the viewpoint of observers) explained the losses with player deficiencies; when they started winning, their weaknesses were written into strengths.


In Contempt and Pity, Daryl Scott argues that people have been trying to explain the African American situation in terms of “psychic damage”, i.e., they internalize the bad depictions of themselves in society, hurting their confidence and expectation.


Followup questions after the research on women in math: Does stereotype threat generalize to other groups? What effect did intellectual strength and motivation have?


4 A Broader view of identity: in the lives of Anatole Broyard, Amin Maalouf, and the rest of us


Anatole Broyard was a book reviewer for the New York Times who essentially changed his racial identity from black to white, immersing himself in white culture and social circles. By changing his racial identity, he could live in different places, have access to different resources (bank loans, professional network), education (for children), etc.


Contingencies are conditions that are special to you because you have a given social identity.

For example, what are the identity contingencies for a white and black student in the cafeteria? A white student would be averse to sitting at a table of black students becuase of fear of being judged as inauthentic and racially insensitive, and because of cultural differences; similarly, a black student sitting at a white table would fear being met with disapproval and non-understanding, or being thought of as disloyal by other blacks.

To explain the lunchroom’s racial segregation, one needn’t postulate even an iota of group prejudice on the part of any student in the room. Its segregation could arise solely to avoid the bad contingencies of these two group identities in that place.


There are more identity contingencies beyond age, sex, sexual orientation, race, like mental health status.

Identity threat is diffuse: one knows that a variety of negative things could happen, not that anything specific will. A diffuse threat is preoccupying.


From In the Name of Identity: Violence and the Need to Belong, Amin Maalouf argues that people can commit crimes in the name of an identity under siege, that they would never do for themselves.

People often see themselves in terms of whichever one of their allegiances is most under attack. And sometimes, when a person doesn’t have the strength to defend that allegiance, he hides it… [until] it invades the person’s whole identity.

When we are threatened is when we are most aware of being a particular kind of person.1 We are not as aware of positive identity contingencies.

James Cormer, a successful school reformer, gives the following advice on facing discrimination. Ignore it the first two times, and then raise all hell. This addresses ambiguity, which is a constant concern (“was that meant to hurt me?”).


Students, after an estimation task, were told arbitrarily that they were over-estimators or under-estimators. Later they disciminated in favor of students in their own group, preferring a payout like ( + 2, +1) to ( + 3, +3) (not to scale). This is the minimal group effect.

Our need for self-regard is powerful enough to make us care about even trivial group identities.


If our social identities are rooted in local contingencies, they are malleable. (Ex. in the experiments, a single statement changed students’ performance on tests.)

Many African-Americans expatriate to Paris. Why? One woman says that even though there is prejudice against blacks in France (though more strongly against North African immigrants), her “blackness” is no longer the most central thing about her when she meets people, like it is in the US. She feels like she was “first met as a full person” in Paris. The French see her as an educated black person.

Contrast passing (what Broyard did)—changing race but keeping country—with expatriation—keeping race but changing country.

“If you want to change the behaviors and outcomes associated with social identity—say, too few women in computer science—don’t focus on changing the internal manifestations of the identity, such as values, and attitudes. Focus instead on changing the contingencies to which all of that internal stuff is an adaptation.

We don’t need to change something “disgraceful” about being a Negro; we need to change the “inconvenience” (identity contingencies) of being a Negro (Bert Williams).

5 The many experiences of stereotype threat


Stereotype threat can affect everyone, even people in the more advantaged group. In a primarily black class on African-American political science, Ted McDougal constantly worried about proving himself as a nonracist white person. He was hesitant to voice his thoughts, ask questions about what he didn’t understand, felt the other students weren’t including him, etc.

Would stereotype threat affect people who hadn’t been “hammered” with it?

To remedy stereotype threat effects, remind test takers of identities that counter the stereotype. Reminding women math students that they were “Stanford students” also helped.


Lower-class French students performed worse when told a test was diagnostic of language ability.

Questions to address: how does stereotype threat cause its effects, and what can be done to reduce its effect?

In particular, many people give the advice of “using the stereotype to motivate you to prove it wrong.” Is this good advice?

6 Identity threat and the efforting life

Philip Uri Treisman conducted innovative workshops teaching college math to students from groups whose math abilities are negatively stereotyped. He conducted an extensive anthropological study of his students’ work habits.

One student, Jeff, saw that two white students reading Playboy and drinking beer in class got a A while he got a C-. The TA suggested to Jeff that he was not adequately prepared and should transfer to a community college. This was only the first blow—he hadn’t correctly predicted which courses he would fail, and felt betrayed.


Because organic chemistry is so important to getting into medical school, advises tell poorly performing students to drop the course and take it another semester. White and Asian students heed the advice, but black students reject it and persist.

To test “over-efforting”, Steele gave difficult anagram problems to black and white students, then gave them a choice of how many anagram/analogy problems they wanted to do in the next iteration. (Analogy problems are easier.) When told that anagrams were a measure of cognitive ability, blacks chose to do an average of 4 more anagram problems.


Stereotype threat hurts performance at the frontier of one’s skills, but boosts performance on easier tasks. Examples:

In these situations, people are not just learning or performing, but refuting a stereotype—which can hurt their learning and performance.

Treisman started a program to get black students to study in groups like Asian students, and found large improvements.

7 The mind on stereotype threat: racing and overloaded


In an experiment, men mistook the lingering anxiety they had after crossing a wobby bridge over a gorge (as opposted to crossing a less-dangerous bridge) for attraction towards the woman interviewer on the other side (as measured by how many times they called her “for more information”).


Groups under stereotype threat didn’t report being more anxious. People can’t judge their own anxiety levels.



Worrying about confirming a stereotype and the consequences takes up attention and distracts us from the task at hand. Researchers measured this by heartbeats (stable heartbeats indicate cognitive load).

What does long-term exposure to identity threats do to people? People may be paying a constant tax without realizing it.


Black Americans have higher rates of hypertension than white Americans. Sherman James developed a test to measure “John Henryism”—tendency to work harder when faced with problems (“persist with effortful active coping under difficult conditions”)—and found that it correlated with hypertension.

8 The strength of stereotype threat: the role of cues


Women in the Supreme Court:

Pressure decreased when there were two women on the Supreme Court because they had a critical mass. Critical mass is when there are enough minorities in a setting that individuals no longer feel uncomfortable because they are minorities, and thus no longer feel an interfering level of identity threat. Critical mass is different in different settings; for example, empirically speaking, conditions for women improve significantly in an orchestra with 20%/40% women.


How much identity threat a person feels in a setting is determined by cues. Examples:

Note identity threat is not the threat of prejudice alone; it is the threat of contingencies.


One cue can shape the interpretation of another. If enough cues can lead members to feel “identity safe”, then it can neutralize the impact of other cues. Changing a few critical cues may be enogh.

9 Reducing identity and stereotype threat: A new hope


When Steele started grad school, he was the only black student in the psychology program. Many incidental features of being white academics were implicitly associated with excellence; assimilation was difficult. He lacked a narrative of the situation that could inspire trust.


Bill Bowen was a president of Princeton who strongly believed that major policy issues in higher education should be based on empirical research. He and Bok reported in The Shape of the River that students admitted to several schools under affirmative action made stronger-than-average contributions downriver (later in their lives).

Douglas Massey found that early differences in grades between minorities and other groups is explained by different susceptibilities to stereotype threat and different levels of college preparation. There are many factors (“a thousand bites”) like family instability, segregated backgrounds, low income, networks less based on academics, etc.

Black professors virtually eliminated stereotype threats in their classroom.


Steele’s relationship with his advisor Thomas Ostrom helped nullify the negative cues not because Ostrom discussed these things with him. Rather, Tom didn’t focus on Steele at all, but on research, sending him the signal that he was a worthy research partner. A work-focused relationship acted as a positive cue comparable to the critical mass cue.


How does a white teacher give cricial feedback to a black student so that the feedback is trusted and motivating?

The researchers assigned essays to students, gave them certain types of feedback, and measured how much they trusted their feedback and were motivated to improve the essay.

A possible objection:

Is it a good idea to talk someone out of worrying about something that could be a real threat to her or him?

Response: It’s better to err more in the direction of greater trust than greater vigilence, because it is hard to learn without trust, and reactions to threat are costly.


Being shown a narrative of upperclassmen experiences (saying that when they were freshmen, they felt deep alienation, but they gradually gained a sense of belonging) increased grades of black students by $\rc3$ of a letter grade. Similarly, a late-night bull session program where groups of <15 students talked about personally relevant topics increased grades of black students by $\rc3$ of a letter grade. The talks revealed that many stresses of college life were universally shared regardless of race.

Students who believed in incremental theory (i.e., having a growth mindset)—the idea that abilities are leranable and incrementally expandable—improved more than those who believed in fixed theory—ability is a fixed capacity. Black students who were given information on the expandability of human intelligence (including scientific evidence and explanations) and asked to write a letter about it improved grades by $\rc3$ of a letter grade. The task had no effect on white students.


Stereotype threats affect children. 5-7 year old Asian girls did worse on a math test after they colored a picture of a girl holding a doll, vs. a gender-neutral picture. Stereotype threats may be deflecting girls from an interest in math before they’ve even had a chance to engage it.

Note: stereotype threat may actually be weaker in identity-segregated schools. (See p. 171 for discussion.)


Self-affirmation theory posits a basic human motive to perceive oneself as morally and adaptively adequate. For example, after a self-image threat (ex. someone shows how you have contradicted yourself), you restore your image better if you are given a chance to step back and reaffirm yourself.

Students were asked to write down their values and either why the values were important to them (self-affirmation), or important to other people, in 15 minutes. Black students in the self-affirmation group (but not the other group) closed the gap between themselves and white students by 40%.

  1. With the self-affirmation, cues became less important.
  2. Better performance interrupts a vicious cycle of worry and bad performance.

In conclusion, psychological interventions can significantly improve minority performance.

In an experiment, high school students were assigned a college student mentor who either emphasized the expandability of intelligence or drug abuse prevention. Girls in the first group but not the second closed the gender gap on the state math test.


A constellation of teacher practices and classroom practices foster identity safety, for example, positive relationships with students, use of diversity as resource, child-centered teaching, etc.


Reducing identity threat is necessary but not sufficent; we also need to overcome real skill and knowledge defecits. But even though background socioeconomic factors are difficult to change, remedying immediate causes (ex. having a mentor) can have a dramatic effect.

Although studies claim that families and not schools are major sources of inequality in student performance, perhaps the measures for school quality are not the right ones.


Students who underperformed on the SAT and received intervention in college not only didn’t do worse than students at the same score level, they did better than nonstereotyped students with the same prior scores. This suggests that both SAT performance and later test performance were depressed by stereotype threat.

Female students taking the AP Calculus exam did better when asked their gender at the end of the exam than at the beginning.

10 The distance between us: The role of identity threat


In The Failures of Integration, Sheryll Cashin found through personal experience that whites avoid sitting down to blacks on planes, even when the seats are better. (This allows her to get “Southwest Airlines First Class.”)

The hypothesis of identity threat explains this without ascribing prejudice. Instead, white passengers could worry about being seen as racist in interacting with the black passenger, and avoid these situations completely.

In On Paradise Drive, David Brooks says that Americans are becoming more segregated into “cultural zones”. Segregation actually increased between 1986 and 2000. (The average white lives in a neighborhood 80% white; the average black lives in a neighborhood 51% black.)

60-90% of people got their job using some form of social capital; however most people claim that they earned the job on their own. Thus a social capital network is important. Glenn Loury (in The Anatomy of Racial Inequality) makes the point that

the everyday associational preferences that contribute to racially organized networks and locations in American life… may now be more important causes of racial inequality than direct discrimination against blacks.

Isolation is so great that black speech converges over great geographic distances and becomes more dissimilar from white speech a few miles away.


In a study, white students were told they would have a conversation with two black students on either “love and relationships” or “racial profiling” and asked to arrange the chairs. Students whose topic was “racial profiling” placed their seats further away. When the conversation partners were white, there was no difference.

The more sentence fragments the participants completed with stereotype words, the farther they sat from their partners.

Moreover, this was due to fear of prejudice rather than prejudice, as the distance was not correlated with scores on the Modern Racism Questionnaire and the Implicit Attitude Test.

In conclusion, having an interracial conversation on a racially sensitive topic made whites more mindful of the whites-as-racists stereotype, and the more mindful they were, the farther they sat.


Having a growth mindset nudges people away from avoidance: when the experimenter told the participant that tension was natural in discussion of racial profiling and that they should treat it as a learning experience, participants sat closer and no longer worried about being seen as racist.

Assuring participants that they wouldn’t be judged, or that differences in perspectives were valued, didn’t work. In fact, the more the researchers assured participants they wouldn’t judge their words, the more afraid they became!

When interactions between people from different backgrounds have learning from each other as a goal, it eases the potential tension between them, giving missteps less significance.

11 Conclusion: Identity as a bridge between us

Eliminating prejudice is not enough in building a fair society. Social and psychological contingencies have diminished but are still very powerful, and we can’t deny past history.

The protective side of human character can strongly affect our entire brain functioning without our conscious knowledge.

Unless you make people feel safe from the risk of these identity predicaments in identity-integrated settings, you won’t succeed in reducing group achievement gaps or in enabling people from different backgrounds to work comfortably and well together.

Actions we can take include

In Obama’s race speech, he embraces his racial identities. We should talk about our identities because

a lot of the experience of having one identity is like the experience of having another identity.

  1. cf. “something to protect”