Implicit attitudes: How children develop biases about race

Twelve years ago, Greta and I were awakened by a rattling on the door of our Bronx apartment. It was about three A.M.; our children were asleep in the next room. "What should I do?" Greta whispered to me. She had woken first and was holding the deadbolt on the door locked so the intruder couldn't get in.

"Call the police," I whispered, and took hold of the lock. I ventured a peek through our peephole. I could see only the grizzled razor stubble of a man who was clearly shorter than I was. He continued to struggle with the door. He was making progress picking our lock -- I had to forcefully resist to keep the lock from turning. As I heard Greta talking with the 911 operator on the phone in the other room, I grew bolder. "Who's there?" I asked, in as gruff and aggressive a voice as I could manage. As soon as he realized there was someone in the apartment, he was gone.

About 30 seconds later, the police appeared at our door. They had been less than a block away when they received the dispatcher's call, had already searched the stairwells, and found no one. We told them our story, and they asked for a description. I told them about the man's height, and the razor stubble. "Did you notice anything else," the officer asked.

"No," I responded.

"What about race -- was he black?"

As I gave my reply: "no," I could see an expression of perplexed astonishment crossing the officer's face. Was it possible that the burglar had walked right past the police, and they had assumed this couldn't be the guy, since he was white? I'll never know the answer, because the police didn't say anything more to me about race. They never caught the intruder.

Let's suppose he actually had walked right by the offender: could we then say that this police officer was racist? After all, he was probably playing the odds -- it's likely that there were more black criminals than white criminals in our area at that time. He never expressed an explicit racial bias to me: he was simply trying to obtain an accurate description of the perpetrator.

A team of researchers at Harvard University have developed another measure of bias, which has been widely reported in the popular press: Project Implicit. As the name of the project suggests, it seeks to measure "implicit attitudes" -- biases that people don't express overtly. The method has been used to measure all sorts of bias: gender, age, even science versus humanities classes. But not surprisingly, racial biases have caught the lion's share of the attention: a recent article in U.S. News and World Report, for example, suggests ways to counter implicit racial biases. The libertarian blogger "Winterspeak" takes a dim view of implicit bias research, suggesting that such biases are merely the result of rational decisions based on knowledge such as "blacks are more likely to commit crimes."

But what about when our "knowledge" about racial differences isn't true? I had a humbling moment a few months ago when I admitted on Cognitive Daily that I was surprised that African American kids are less likely to do drugs or consume alcohol than white kids. How much real knowledge do most people have about racial differences? Could knowledge really be the sole motivator of implicit bias?

Andrew Scott Baron and Mahzarin Banaji recently conducted a study that offers some tentative answers. They gave their racial implicit bias test to white middle class kids aged 6, 10, as well as adults. You can try the test for yourself at Project Implicit, but here's a quick summary of how it works. First, you're shown pictures of black faces or white faces: the task is to press a button as quickly as possible when you you see each face (E for a black face or I for a white face). Next you're shown a set of words, some good and some bad (love, joy, friend, hate, vomit, bomb), and again, you're asked to press a designated key for each type. Finally, the tasks are combined: "When you see a black face or a good word, press the E key" and "When you see a white face or a bad word, press the I key." Then the tasks are reversed, so good words are associated with white faces and bad words are associated with black faces. Reaction times are measured, and when a particular association results in a faster response time, then participants are said to have an implicit attitude prefering that association.

In this case, the test was modified for the smallest children so that instead of words appearing on the screen, recorded words were played for them. Here are the results:


For every age group, the association of white faces with good words was stronger than the association of black faces with good words: an implicit bias for white faces over black faces. The bias must have formed before the age of six, and is undiminished in adulthood. To make sure everyone understood the task, a similar test was given to measure preference for insects versus flowers. Everyone except six-year-old boys said they preferred the flowers, but when the preference was measured with the implicit task, even the boys showed an implicit bias for flowers.

But Baron and Banaji didn't stop with measuring implicit preferences. They also performed an explicit preference task, in which participants were asked overtly whether they preferred a white face or a black face. Here are the results:


Unlike the implicit task, these results do change over time, with each age group showing a significant difference from the other groups, and adults showing an equal preference for black and white faces. Though the implicit biases remain until adulthood, explicit biases appear to have been extinguished.

This data certainly is compatible with the idea that people can claim they are "not racist," when their actions appear to contradict that notion. But what of Winterspeak's criticism: "The Implicit Project implicitly assumes that any differentiation between blacks and whites is racist"? That's a difficult notion to defend. Winterspeak offers no data in support of her/his claim, while Project Implicit can demonstrate that people's actions differ from their words. There's no mention of racism at all in Baron and Banaji's report, or in Banaji's quotes in the U.S. News article.

Worst of all, when people justify racial discrimination based on "knowledge" of racial differences, they are making two assumptions: that their knowledge of the stereotype is correct, and that the individual in question conforms to the stereotype. In three cases, (inaccurate stereotype, individual doesn't conform to stereotype, or both) their judgement is not only incorrect, but immoral. In the last case -- assuming the person doesn't know for certain whether the individual he or she discriminates against conforms to the stereotype -- it is merely immoral.

Isn't it better to accurately know what your implicit biases are, and to try to adjust your behavior accordingly?

Baron, A.S., & Banaji, M.R. (2006). The development of implicit attitudes: Evidence of race evaluations from ages 6 and 10 and adulthood. Psychological Science, 17(1), 53-58.

More like this

[originally posted on February 2, 2006] In connection to Monday's posting, Other-race faces: Why do they seem different?, I thought readers would be interested in a post from early last year concerning implicit attitudes on race. The link to the original post is above if you would like to see…
Very few of us can avoid stereotyping others. When we're actively trying to avoid racial stereotyping, we often end up looking ridiculous. But the very fact that we can try to avoid it suggests that there's something more to racial stereotypes than a "stereotype center" in the brain. If…
"Outing" gays and lesbians has always been a controversial practice, especially when done without the outed person's consent. But even when an individual outs him or herself, some people argue that outing is inappropriate because of the negative stereotypes that are evoked. But there's a subtler…
Keith Payne's work on racial stereotyping brings up an intriguing possibility. During the weapons identification task, viewers are more likely to erroneously identify a harmless object as a weapon if it was preceded by a black face compared to a white face. They are also more accurate identifying…

It's tough, I think, to do this kind of research, because inevitably, someone accuses you of bringing value judgements (e.g., racism, sexism, etc.) into the research. If I were to play devil's advocate, I guess I do see Winterspeak's point -- why were the only test stimuli good/bad traits? Why did they not pair white and black faces with slightly more neutral words like "atheletic" or "organized" or "aloof"?

Subjectively, it seems that black individuals are often perceived to be more athletic (Whether this is a positive thing or not is a whole other discussion!). If people are faster to respond to a black-athletic pairing, then it might support the view that the differences found in the study are influenced by "the odds", so to speak.

I think the bottom line to take away here is that regardless of where these associations come from or why they are maintained, they do exist, and they exist from early on. Though this is admittedly debatable, I do not believe we are slaves to our subconscious selves -- sure, we may *have* these implicit associations. But, we also have the ability to think critically and control our behaviours. Even if racism was ingrained implicitly at an early age, that is certainly no excuse to behave in ways that are hurtful to others.

The trick is to separate evidence from racism. If I lived in an area with 50% white and 50% black residents, but 70% of crimes in the area were committed by black residents, then if I was asked who was more likely to commit a crime in that area, I would say a black person. But this isn't racism, just statistics. To me, racism is the idea that the reason for the 70% is that blacks, because of their genetics, have a predisposition to crime.


I think what you're saying is true, from a logical point of view. But I'm not sure it explains the behavior in the experiment.

What's interesting to me about the Baron and Banaji study is that while children and adults have the same levels of implicit bias, their explicit bias decreases as they age -- presumably, as they gain more knowledge about racial differences. So as we age, increasingly our behavior does not match our knowledge of evidence and statistics.

I think a possible criticism of the study is that implicit bias doesn't reflect meaningful behavior -- that in realistic scenarios such as how we behave when we meet someone on the street (whether, for example, we take a defensive posture because we believe the person might be a criminal), we behave more based on explicit biases than implicit ones.

On the other hand, a lot of anecdotal evidence (such as my police story) suggests that implicit biases do impact real-world behavior. I'd be interested if readers could point me to studies which test this notion. I'd also be interested in hearing additional anecdotes -- either in support of the idea that implicit biases affect behavior, or contradicting it.

It is only natural (from an evolutionary standpoint) for children to take on implicit biases for whatever is "different" from their immediate environment. This includes a natural xenophobia. what is necessary in society is to actively train ourselves to recognize these implicit biases. I'm dubious about the statistics on the diminishment of these biases in this data since I don't see that much attempt by most adults to face up to this need for introspection. As a guilty white liberal aware of this dynamic, with my own skewed childhood experiences, I try to be aware of my reformulation of biases as an adult, but I'm never quite sure how well I'm doing. I had an experience with a wrong number once that still haunts me, even though it was a minor offense. I had gotten into the habit of rudely hanging up on callers I didn't immediately recognize since they were most likely trying to sell me something. I responded this way once to a woman by hastily saying "I'm not interested, sorry" and starting to hang up. The last thing I heard was her calling me a "honkey". It rattled me. I'm not sure any of us can ever completely recover from our implicit biases, and they will sometimes be exposed, but we must do our best. As for the police officer, he was clearly exhibiting prejudice, not just bias. Regardless of the statistics of experience, a police officer must consciously maintain equal vigilence or else his/her biases become institutionalized.

I'm not sure how much weight to place on the explicit data... don't demand characteristics & ideas about social acceptability confound those findings? The study seemed like it would be pretty transparent to participants. Could it not be that as we age, we realize more that it's inappropriate to judge based on race, and thus consciously try to eliminate ("hide") any inital preference biases? It is less likely that the implicit measure would be influenced by these strategies.

Possible violation of construct validity. Did the researchers simply assume that their test was implicit or did they provide evidence that subjects were unaware of the experiment's purpose?

When you require different responses based on white or black faces, you're not disguising the manipulation. A better design would have elicited the same response action for both black and white faces, with a preceding variable biasing response times in one direction or the other.

By Mark Aveyard (not verified) on 04 Feb 2006 #permalink

"The trick is to separate evidence from racism. If I lived in an area with 50% white and 50% black residents, but 70% of crimes in the area were committed by black residents, then if I was asked who was more likely to commit a crime in that area, I would say a black person."

While this reasoning is logical, the problem with it is that no one really knows the percentage of crimes committed by white vs. black people. We only know the percentages of people who have been caught. Think of Dave Munger's story and this part: "Was it possible that the burglar had walked right past the police, and they had assumed this couldn't be the guy, since he was white? I'll never know the answer, because the police didn't say anything more to me about race. They never caught the intruder." Perhaps had he been black they would have caught him, and thus added another notch to the "black people commit more crimes" statistic.


To the best of my knowledge, no studies specifically examine the link between implicit attitudes and overt behavior (e.g. helping behaviors). A few studies examine implicit attitudes and "microbehaviors" (e.g. eye blinks, stuttering, etc.). Check out:

McConnell, A.R., & Leibold, J.M. (2001). Relations among the implicit association test, discriminatory behavior, and explicit measures of racial attitudes. Journal of Experimental Social Psychology, 37, 435-42;


Dovidio, J.F., Kawakami, K., & Gaertner, S.L. (2002). Implicit and explicit prejudice and interracial interaction. Journal of Personality and Social Psychology, 82, 62-68.

A classic pre-IAT study that examines how microbehaviors reciprocally influence interracial interactions would be:

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, 109-20.

By Ken Smith (not verified) on 04 Feb 2006 #permalink

In addition, the newest edition of American Psychologist (Vol 61, No 1) has an exchange between some of the critics of the IAT (e.g. it uses an arbitrary metric) and Anthony Greenwald, Brian Nosek, and N. Sriram (a few of the major proponents of the IAT).

By Ken Smith (not verified) on 04 Feb 2006 #permalink

There's a study that is discussed in the book "Blink" that found that if people were basically passively exposed to reading words like ""retire" and "prunes" (I can't remember the actual words)... words rather weakly related to old age, in a way that was hard to see them as being related to old age...
They walked slower leaving the room.

I think the point of implicit biases is that the tests are measuring how strongly topics or themes are linked in the brain. I took the Harvard website's test on disabled people, where they linked symbols of handicapping conditions like wheelchairs and crutches with positive or negative words. To me the result was really confounded by the fact that most people think that a wheelchair is an inherently sad thing, not an inherently bad thing (like a gun might be inherently bad).

I am a research assistant at a univeristy. I administer IATs to undergrads for research.

I think there is something very powerful that is being measured in IAT's but it's not so powerful that it can't be overridden given a few seconds of time and a brain that isn't distracted with another task.

Everyone is prejudiced, with the possible exception of mentally handicapped and/or autistic people, in my opinion. Even "high functioning autistics" tend to be naturally egalitarian, it's one of the very fine qualities that is overlooked in autistics because of the heaps of prejudice and bigotry piled on them, no matter what "race" they belong to.
See the work of Michelle Dawson at "no autistics allowed" online if you'd like to learn more about that.

I think this research is a great way to measure otherwise hidden variables, things in inside people's belief structures. But it certainly can't be directly interpreted as a "racism" measure. We have to distinguish between "good" prejudice and "harmful" prejudice. "Racial profiling" by police officers is a good example. In a way, profiling is all police officers do. That's the only way to catch criminals. You aren't going to catch criminals if you sit around a nursing home, for example. You go where the crimes take place. I don't think geographic profiling is any different from other types of profiling. The idea is simply to increase the likelihood of finding a criminal.

The law enforcement and justice systems are not perfect, they rely on a lot of probability calculations to work. Prejudice is really just a-priori probability estimation. It doesn't even really matter how accurate these probability estimates are, people are still going to use them. I think we need to clarify what really constitutes racism in a way that simple probability calculations don't count. I don't think bias or predjudice constitute racism. I don't buy all this "soft bigotry of low expectations" talk by W. Racism is not when you hire 10% more whites than blacks, racism is when you don't hire blacks, period. Or when you treat them like second-class employee. Racism is just plain dislike of people based on their race.

I don't think calculations like "he is 5 times more likely to be a criminal because he is black" by themselves constitute racism. I mean, young people are more likely to be criminals too, males are more likely to be criminals, people from Chicago are more likely to be criminals. So what? Just because someone is likely to commit a crime doesn't make them a bad or dislikable person. Neither do low test scores, low paying jobs, house size, or any of that. The real problem in this country is too many people think those things ARE all that matters. Materialistic values in a democracy are always going to lead to a kind of racism because minorities are always going to have less power and wealth, so if wealth is used to judge personal worth then minorities will be judged lower.

By James Gambrell (not verified) on 06 Feb 2006 #permalink

As I Black person who has experienced prejudice from people who say they are not prejudice, and are participants in studies against prejudice I support the results of the test. The notion of Black inferiority is promoted in the media to be equated with bad, is very much a part of the fabric of north American society - something I don't believe is accidental, but deliberately planned. Not so long ago in Canada, CTV news station broadcasted news about a serial killer who killed about 21 prostitues and a black offender who shot at a police and missed. I was angered how the two stories was dealt with. A sketch of the white serial killer was barely visible, yet a full blown picture of the black offender that fully covered the television screen was projected for for about five seconds. As white adults grow older they learn how to make racial overtures less explicit. And if you think I'm biased then I ask you a question, why is it that when a black person is featured in ads with with people the black person 95% of the time is off on the side, rarely the center subject? Or, why do most shows feature more white people than other races? Why aren't all races represented equally?

By P. Samuals (not verified) on 08 Nov 2006 #permalink