Subliminal

German captors, as he would later point out, if his identity were discovered, it would be his classification as a Polish Jew that would determine his fate.¹² But there was also danger in lying. So, from the menu of stigmatization, Tajfel chose the middle dish: he spent the next four years pretending to be a French Jew.¹³ He was liberated in 1945 and in May of that year, as he put it, was “disgorged with hundreds of others from a special train arriving at the Gare d’Orsay in Paris … [soon to discover] that hardly anyone I knew in 1939—including my family—was left alive.”¹⁴ Tajfel spent the next six years working with war refugees, especially children and adolescents, and mulling over the relationships between categorical thinking, stereotyping, and prejudice. According to the psychologist William Peter Robinson, today’s theoretical understanding of those subjects “can almost without exception be traced back to Tajfel’s theorizing and direct research intervention.”¹⁵

Unfortunately, as was the case with other pioneers, it took the field many years to catch up with Tajfel’s insights. Even well into the 1980s, many psychologists viewed discrimination as a conscious and intentional behavior, rather than one commonly arising from normal and unavoidable cognitive processes related to the brain’s vital propensity to categorize.¹⁶ In 1998, however, a trio of researchers at the University of Washington published a paper that many see as providing smoking-gun evidence that unconscious, or “implicit,” stereotyping is the rule rather than the exception.¹⁷ Their paper presented a computerized tool called the “Implicit Association Test,” or IAT, which has become one of social psychology’s standard tools for measuring the degree to which an individual unconsciously associates traits with social categories. The IAT has helped revolutionize the way social scientists look at stereotyping.

IN THEIR ARTICLE, the IAT pioneers asked their readers to “consider a thought experiment.” Suppose you are shown a series of words naming male and female relatives, such as “brother” or “aunt.” You are asked to say “hello” when presented with a male relative and “good-bye” when shown a female. (In the computerized version you see the words on a screen and respond by pressing letters on the keyboard.) The idea is to respond as quickly as possible while not making too many errors. Most people who try this find that it is easy and proceed rapidly. Next, the researchers ask that you repeat the game, only this time with male and female names, like “Dick” or “Jane” instead of relatives. The names are of unambiguous gender, and again, you can fly through them. But this is just an appetizer.

The real experiment starts now: in phase 1, you are shown a series of words that can be either a name or a relative. You are asked to say “hello” for the male names and relatives and “good-bye” for the female names and relatives. It’s a slightly more complex task than before, but still not taxing. What’s important is the time it takes you to make each selection. Try it with the following word list; you can say “hello” or “good-bye” to yourself if you are afraid of scaring away your own relatives who may be within earshot (hello = male name or relative; good-bye = female name or relative):

John, Joan, brother, granddaughter, Beth, daughter, Mike, niece, Richard, Leonard, son, aunt, grandfather, Brian, Donna, father, mother, grandson, Gary, Kathy.

Now for phase 2. In phase 2 you see a list of the names and relatives again, but this time you are asked to say “hello” when seeing a male name or female relative and “good-bye” when you see a female name or male relative. Again, what’s important is the time it takes you to make your selections. Try it (hello = male name or female relative; good-bye = female name or male relative):

John, Joan, brother, granddaughter, Beth, daughter, Mike, niece, Richard, Leonard, son, aunt, grandfather, Brian, Donna, father, mother, grandson, Gary, Kathy.

The phase 2 response times are typically far greater than those for phase 1: perhaps three-fourths of a second per word, as opposed to just half a second. To understand why, let’s look at this as a task in sorting. You are being asked to consider four categories of objects: male names, male relatives, female names, and female relatives. But these are not independent categories. The categories male names and male relatives are associated—they both refer to males. Likewise, the categories female names and female relatives are associated. In phase 1 you are asked to label the four categories in a manner consistent with that association—to label all males in the same manner, and all females in the same manner. In phase 2, however, you are asked to ignore your association, to label males one way if you see a name but the other way if you see a relative, and to also label female terms differently depending upon whether the term is a name or a relative. That is complicated, and the complexity eats up mental resources, slowing you down.

That is the crux of the IAT: when the labeling you are asked to do follows your mental associations, it speeds you up, but when it mixes across associations, it slows you down. As a result, by examining the difference in speed between the two ways you are asked to label, researchers can probe how strongly a person associates traits with a social category.

For example, suppose that instead of words denoting male and female relatives, I showed you terms related to either science or the arts. If you had no mental association linking men and science or women and the arts, it wouldn’t matter if you had to say “hello” for men’s names and science terms and “good-bye” for women’s names and arts terms, or “hello” for men’s names and arts terms and “good-bye” for women’s names and science terms. Hence there would be no difference between phase 1 and phase 2. But if you had strong associations linking women and the arts and linking men and science—as most people do—the exercise would be very similar to the original task, with male and female relatives and male and female names, and there would be a considerable difference in your response times in phase 1 and phase 2.

When researchers administer tests analogous to this, the results are stunning. For example, they find that about half the public shows a strong or moderate bias toward associating men with science and women with the arts, whether they are aware of such links or not. In fact, there is little correlation between the IAT results and measures of “explicit,” or conscious, gender bias, such as self-reports or attitude questionnaires. Similarly, researchers have shown subjects images of white faces, black faces, hostile words (awful, failure, evil, nasty, and so on), and positive words (peace, joy, love, happy, and so on). If you have pro-white and anti-black associations, it will take you longer to sort words and images when you have to connect positive words to the black category and hostile words to the white category than when black faces and hostile words go in the same bin. About 70 percent of those who have taken the test exhibit this pro-white association, including many who are (consciously) appalled at learning that they hold such attitudes. Even many black people, it turns out, exhibit an unconscious pro-white bias on the IAT. It is difficult not to when you live in a culture that embodies negative stereotypes about African Americans.

Though your evaluation of another person may feel rational and deliberate, it is heavily informed by automatic, unconscious processes—the kind of emotion-regulating processes carried out within the ventromedial prefrontal cortex. In fact, damage to the VMPC has been shown to eliminate unconscious gender stereotyping.¹⁸ As Walter Lippmann recognized, we can’t avoid mentally absorbing the categories defined by the society in which we live. They permeate the news, television programming, films, all aspects of our culture. And because our brains naturally categorize, we are vulnerable to acting on the attitudes those categories represent. But before you recommend incorporating VMPC obliteration into your company’s management training course, remember that the propensity to categorize, even to categorize people, is for the most part a blessing. It allows us to understand the difference between a bus driver and a bus passenger, a store clerk and a customer, a receptionist and a physician, a maitre d’ and a waiter, and all the other strangers we interact with, without our having to pause and consciously puzzle out everyone’s role anew during each encounter. The challenge is not how to stop categorizing but how to become aware of when we do it in ways that prevent us from being able to see individual people for who they really are.

THE PSYCHOLOGY PIONEER Gordon Allport wrote that categories saturate all that they contain with the same “ideational and emotional flavor.”¹⁹ As evidence of that, he cited a 1948 experiment in which a Canadian social scientist wrote to 100 different resorts that had advertised in newspapers around the holidays.²⁰ The scientist drafted two letters to each resort, requesting accommodations on the same date. He signed one letter with the name “Mr. Lockwood” and the other with the name “Mr. Greenberg.” Mr. Lockwood received a reply with an offer of accommodations from 95 of the resorts. Mr. Greenberg received such a reply from just 36. The decisions to spurn Mr. Greenberg were obviously not made on Mr. Greenberg’s own merits but on the religious category to which he presumably belonged.

Prejudging people according to a social category is a time-honored tradition, even among those who