Prejudice Under the Microscope: The Implicit Association Test (Part I)

Editor’s Note: This article is part of an ongoing symposium on white fragility and its related concepts. To view all of the essays in this series, click here.

A collective groan could be heard around the world as Stars Wars fans finished viewing the eagerly anticipated Episode I: The Phantom Menace (TPM) — released approximately 15 years after the much beloved original Star Wars trilogy changed movie making forever. Die-hard fans objected to numerous aspects of the franchise reboot — not the least of which was the manner in which ‘The Force’ was de-mystified for audiences.

In the first movie (Episode IV: A New Hope), the Force was depicted as a mysterious entity that is the source of a Jedi Knight’s power. It was simply described in the movie as an energy field created by all living things, which surrounds and penetrates beings and binds the galaxy together. Makes sense? Not really. But it was sufficient for fueling the affection and creative imagination of millions worldwide.

In TPM, however, the living cells of certain special individuals were blessed with huge amounts of microscopic life forms called ‘midichlorians’ that enabled them to be ‘strong in the force’ (whatever that means). Disappointed fans felt that this ‘technobabble’ (as one critic characterized it) was one among many problems with the film that robbed the original trilogy of its sense of fun, mystery, and magic.

Obviously, the Stars Wars movies — despite the obscene amounts of money and popularity they have generated over the decades — are a work of pure fiction. There are no such entities as Jedi Knights, light sabers, habitable ice planets, Wookies, or Sith Lords. However, there is one aspect of the storytelling that has relevance for today — the strong desire of psychologists to reduce relatively complex, mysterious, and enduring human psychological concepts down to something that can be oversimplified, put into a test tube, and studied. The technical term for this tendency of psychological science is ‘reductionism,’ which has been roundly criticized.

To illustrate, psychologists and mental health counselors who believe in microaggressions theory assure receptive audiences that brief verbalizations such as ‘all lives matter,’ ‘you sound so articulate,’ or ‘where are you from?’ reflect a form of racialized aggression (regardless of intent) against others, which inflicts ‘small trauma’ as well as ‘assaults on the mental and physical health of persons of color.’

Neuropsychologists affirm that passionate love can be understood, in part, as reflecting activation of the middle insula, anterior cingulate cortex, putamen, retrosplenial cortex, the ventral tegmental area, and right caudate of the human brain — along with deactivation of the amygdala. In addition, who knew that being perceived as attractive boils down to having a face that is 1.5 times longer than it is wide; equal distances between the forehead hairline to the spot between the eyes, the spot between the eyes to the bottom of the nose, and the bottom of the nose to the bottom of the chin; and an ear length that matches the length of the nose?

All human societies, to varying degrees, must grapple with issues related to racial discrimination and racial prejudice — whether such problems are indeed real or are merely perceived to be real. Within America, racial tensions between subpopulations lie at the root of vexing problems across a wide variety of areas as diverse as housing, employment, education, human health, state/national elections, and crime.

The public’s reaction to popular books such as Robin D’Angelo’s White Fragility: Why It’s So Hard for White People to Talk About Racism has uncovered a sharp divide between those who insist that whites are irredeemably racist, are woefully unaware of their white privilege, and are ultimately at fault for most of the problems persons of color face and those who fundamentally reject such ideas. The central thesis of D’Angelo’s book is that whites’ conscious denial of their racism and white privilege are symptomatic of their white fragility — which in and of itself is symptomatic of their unconscious racism.

Wouldn’t it indeed be groundbreaking if a method were discovered whereby the deep-seated evil of racism (however defined) could be objectively identified and accurately measured within persons — much like a midichlorian count in the Star Wars movies could measure how strong a person is with the Force? What if this method were able to bypass an individual’s attempts to consciously manipulate or misrepresent their responses — but instead could mine attitudes deeply lodged in the unconscious? More importantly, what if poor scores from this method could allow important stakeholders to predict future prejudicial and discriminatory actions of persons toward others on the basis of race? Some believe that they have found this in the research surrounding the Implicit Association Test (hereafter abbreviated as the IAT).

Professors Anthony G. Greenwald, Debbie E. McGhee, and Jordan L.K. Schwartz of the University of Washington introduced the IAT to the world in a 1998 article published in the Journal of Personality and Social Psychology (JPSP), “Measuring Individual Differences in Implicit Cognition: The Implicit Association Test.” To date, the article has achieved the status of being among the top five most cited articles from the JPSP, and since its publication has generated hundreds of peer-reviewed journal articles.

In the underlying science supporting the IAT, psychologists theorize that the brain contains a vast interconnected web of knowledge called a ‘semantic network.’ Each concept in this network is represented as a ‘node,’ which is a collection of brain cells that code and store a particular concept. According to the theory, each concept shares connections with other concepts in the network that are related or similar in some way. Any perceived stimuli automatically and involuntarily trigger within the brain associated biases, preferences, and emotions.

The measurement tools used in IAT research fit conceptually within the broader discipline of mental chronometry — which is manifested today in the use of computer technology to measure an individual’s responses, or reaction times (abbreviated RT) to a visual or auditory stimulus that calls for a particular response, choice, or decision. The use of computers allows reaction/response times to be measured in precise milliseconds (thousandths of a second). Hence, although a particular subject involved in an RT experiment may consciously feel that his reaction/response time was the same in one computerized simulation compared to another, more precise chronometric measurement may actually reveal significant differences in average reaction/response times that lie beyond the subject’s conscious awareness across the two simulations.

RT methodology seeks to identify which perceived concepts are either strongly or weakly related to other perceived concepts. RT methodology is used in market research, for example, to discover what consumers automatically and subconsciously associate with brand names. To illustrate, when a consumer is shown a picture of the Apple computer brand logo, it is theorized to trigger what is known as ‘spreading activation,’ or the automatic and involuntary activation of other concepts in the brain’s semantic network. If the Apple logo is strongly associated with the words/phrases ‘Steve Jobs,’ ‘computer,’ ‘fruit,’ or ‘cool,’ then persons will react faster to concepts perceived to be closely related to the Apple logo than concepts perceived to be unrelated.

The main attraction of computerized RT methodology is that it frees the researcher from relying on paper-and-pencil self-report measures, as subjects may be unwilling to report their true personal thoughts or feelings toward a particular subject matter, or may consciously distort/misrepresent their true responses. The low correlations that research studies have found between ‘implicit’ versus ‘explicit’ measures of the same subject matter appear to reinforce the idea that certain biases lie beyond a person’s conscious awareness.

How the Race IAT Works

In Greenwald et al.’s original 1998 experiment, subjects are instructed to sort word stimuli presented on the computer screen into two categories (i.e., pleasant vs. unpleasant), then indicate their responses by pressing a key with their left hand on the left side of the keyboard (to signal one category), or their right hand on the right side of the keyboard (to signal the other category). In early studies, using the IAT to explore racial issues involved the use of black- versus white-sounding names as computer stimuli. Here, the subject participates in the same exercise using the same computer keys, but must sort pleasant versus unpleasant words that are neutral and not associated with race (e.g., ‘gift’, ‘grief’, ‘poison’, ‘lucky’). The same key must be pressed for a black-sounding name and an unpleasant word, and the other key must be pressed for a white-sounding name and a pleasant word. For subsequent trials, the racial and nonracial word stimuli are then presented in an intermingled fashion, and the subject presses either the left or right keyboard key as quickly as possible after each stimulus appears on the computer screen.

Next, the instructions to the subject are reversed for the random stimuli presentation, in that one key must be pressed to indicate a black-sounding name and a pleasant word, and the other key must be pressed to indicate a white-sounding name and an unpleasant word. If a subject takes longer sorting black-sounding names using the computer key associated with a “pleasant” word than he does sorting white-sounding names using the computer key associated with a “pleasant” word, the IAT deems the subject as a bearer of “implicit bias.” The IAT ranks the subject’s degree of implicit bias based on the differences in milliseconds with which he accomplishes the different sorting tasks. At the conclusion of the test, each subject finds out whether he has a strong, moderate, or weak preference for blacks or for whites. More recent applications of the race IAT use as stimuli actual pictures of faces representing ‘European’ vs. ‘African’ ancestry.

The Race IAT Catches Fire

This initial research resulted in the founding in 1998 of Project Implicit, which is described by the Atlas of Public Management as a non-profit organization of implicit social cognition researchers whose goal is to ‘educate the public about hidden biases and to provide a virtual laboratory for collecting data on the internet.’ In addition, Project Implicit provides ‘consulting, education, and training services on implicit bias, diversity and inclusion, leadership, the application of science to practice, and innovation.’ To date, as many as 17 million persons have taken the race IAT (along with additional IAT protocols that focus on other social groups) on Harvard University’s Project Implicit website.

The IAT is a veritable godsend for the racism industry in America and abroad, as it appears to affirm the useful narrative that explicit racism — rather than being rendered ineffective by antidiscrimination laws and modern social pressures — is nevertheless alive and well as manifested in ‘hidden racism,’ ‘aversive racism,’ ‘reasonable racism,’ ‘unconscious bias,’ ‘benevolent discrimination,’  ‘color-blind racism,’ ‘benign bigotry,’ ‘symbolic racism,’ ‘silent racism,’ and the like. According to one writer, getting to know one’s implicit biases (via taking the race IAT) is ‘step one’ of becoming an ‘anti-racist ally’ to the Black community — as it ‘unravels the workings of racist thinking’ and allows race IAT users to ‘confront their own racist thoughts and beliefs’.

As seeming validation for the social importance of the race IAT, the Project Implicit website reports that three implicit bias researchers (Mahzarin Banaji, Tony Greenwald, and Brian Nosek) were given the prestigious Golden Goose Award from the American Association for the Advancement of Science during a ceremony at the Library of Congressin recognition of their work “as a major breakthrough with exceptional social impact”.

At this juncture, a distinction needs to be made between the philosophical assumptions of America’s ‘racism industry’ and the long-established principles of patient, careful, and dispassionate science. In this latter world, before any instrument (that purports to be a good measure of a psychological construct) is ready for real-world use, it must first provide solid and consistent empirical evidence that answers questions in four basic categories:

  1. Does the instrument actually measure what it purports to measure? Or, are there plausible competing explanations as to what is being measured? How much of whatever is being measured is useless ‘noise’?
  2. Whatever the instrument purports to measure, does it do so reliably? That is, can whatever is measured provide stable scores over time?
  3. Is whatever is being measured related in meaningful ways to other behaviors measured in laboratory settings?
  4. Is whatever is being measured related in meaningful ways to other behaviors deemed to be significant and important in real-world contexts?

In Part II of this series, I will examine the evidence for the race IAT under the first three categories of questions. In Part III of this series, I will discuss evidence for the fourth question in considerable detail, concluding with an overall assessment of the utility of the race IAT in light of its numerous claims.

Image: Michael Longmire, Public Domain


12 thoughts on “Prejudice Under the Microscope: The Implicit Association Test (Part I)

  1. Researchers across the political spectrum have questioned the accuracy of the IAT test. One study [] found the test unreliable, with participants likely to receive a different outcome each time they took the test.

    New York Magazine published a column [] detailing the shortcomings of the “implicit association” test.

    “The IAT, it turns out, has serious issues on both the reliability and validity fronts, which is surprising given its popularity and the very exciting claims that have been made about its potential to address racism,” columnist Jesse Singal wrote. “That’s what the research says, at least, and it raises serious questions about how the IAT became such a social-science darling in the first place.”

    Academics are grasping at straws in an effort to convince everyone that all whites are racist, even if they have never had a racist thought. The University of Pittsburgh expresses a desire to end racism while simultaneously discriminating against whites in all areas of academic life. Alumni have to think long and hard before donating to their alma mater, which is becoming a source of shame and embarrassment rather than pride.

  2. I have taken the IAT about a dozen times. In every single case, the first iteration of the test required me to press a particular key if a saw a “positive” word or a white face and to press a different key if I saw a “negative” word or a black face.
    In other words, I was TRAINED to associate “positive” words with white faces and “negative” words with black faces.
    Then, on the second iteration of the test they switch things around. Is it any wonder that it takes an extra millisecond or two to remember to associate “positive” with black? But of course that extra millisecond is all it takes to be branded a racist.
    Social psychology is not a scientific discipline — it is a left-wing social and political movement that masquerades as science.

    1. Yep. It took about 5 seconds to recognize that doing white= pleasant first before reversing it introduces bias into the test. Do it the other way first or randomly for a better result.

      But we know that’s why they don’t run the test that way… because that training destroys the tests results and thus its usefulness to the race hustlers.

    2. Yes, exactly. The sequencing of the tests matters. It’s not clear to me whether any of the IAT-ish studies control for this. And it’s not even clear to me that it is possible to control for it. For any individual, the first test (i.e. measurement) perturbs all subsequent measurements—even ones days or weeks later.

      Too, I see a problem with the choices of specific words or images in the test. How have they established that the set of test words is calibrated perfectly such that there are no other possible confounding connotations or associations that bias the results? I mean even seemingly silly things like the word bad—what if the test subject has a fondness for referring to things that please him as ‘bad’ antiphrastically? What if the subject is the #1 fan of Michael Jackson’s song “Bad”? (Or indeed the fact that most Americans know Michael Jackson was black and sang the song “Bad”? Can it be demonstrated that these connections don’t confound the result? What about differences in personal color preferences? Maybe there’s some subtle difference in the concepts of white vs black as racial groupings—maybe a test subject thinks of himself as Spanish or Greek and not as white, and doesn’t care for the concept of a ‘white’ race. Does this affect his association speeds positively or negatively? And would the affects of any be symmetrical with respect to the associations times of good versus bad, and relative to white and black? (Remember this is the crowd that now insists on ‘Black’ and ‘white’ to indicate that Blacks have a unique history that deserves a capital letter whereas whites do not.) Or what if because of the extreme taboo in American society against racism and any taint of it, IAT subjects’ association times are influenced by some fear pathway in the brain, such that the path and neural ‘distance’ traveled when recognizing white/good, white/bad, black/good, black/bad are not symmetrical and this accounts for time differentials. This examples seem absurd to consider, but then why can’t IAT proponents give a simple scientific description of the test that shows these factors as ruled out? Again, these researchers are justifying the IAT with the quasi-technical idea of a sort of neural topography of clusters of nodes in the brain and that the network distance between two concepts can be precisely quantified via the IAT (as though the mind were running OSPF protocol) and that this network distance reflects precisely the strength of their association and that this association entails particular higher-order thought processes with respect to racial groups that can be constructed from the simple time difference data. The claim is that the IAT reveals the underlying mental structures within which all thoughts are born and thus the shape and connections of those thoughts. They claim milliseconds of difference are significant, and that we can extract meaning out of these tiny differences. (I am not aware if claims are advanced regarding mapping the magnitude of time difference to the degree of racial bias.) So we should understand exactly any factors that may influence recognition speed and whether this speed proves anything. I don’t believe IAT proponents have good answers to how the test accounts for confounding variables like this, or why we can ignore them, and that this belies how weakly supported are some of the logical leaps they take and how under the covers there is no rigor to the correlation of milliseconds to real-world racial bias. It’s all so murky to me and I don’t see that just simply repeating a test round with white and black swapped accounts for and eliminates confounding variables.

      And the technical issues with the test design of course tell us nothing about the validity of the IAT as a proxy for the concept of racial bias—which itself is not rigorously defined—and how it manifests in the real world.

      I also see a problem linking association to morality, as some interpretations of the IAT by lay people do. An example that goes in this direction: let’s say we designed an IAT with images of Americans and Soviets, along with food items like hot dogs, apple pie, Coca Cola, borscht, and vodka. Conceding temporarily for the sake of argument that we can extract meaning out of quicker associations made on the IAT, would it indicate some defect in my thinking or corruption of my morality if I more quickly associated apple pie, hot dogs, and Coke with Americans? No, because in fact these are American food items. To drill a little closer: the human mind perceives differences, it notices associations. If the reality is that Americans tend to like hot dogs, the recognition of that reality is nothing more than the proper functioning of the mind. To wit, most professional basketball players in the US are black. Is it be racist if an individual’s IAT shows a quicker response associating ‘black’ with ‘NBA’? Can we speak of this in terms of wrong or right thinking? Assuming again for the sake of argument that the association indicates higher-level thought. What do we say to the person who says that it is entirely logical—let’s say evolutionarily fit / highly conserved—for a person’s brain to have a quicker link between ‘black’ and ‘NBA’ for it is nothing more than the brain noticing reality—that is, noting the prevailing population distribution in the NBA?

      This is uncomfortable to talk about not least because people can be fired and shunned for stating patent truths if they are related, even however distally, to race. But now let’s go further and consider what to say if the IAT shows a person was quicker to associate a negative concept with a particular racial group? Like it or not, there are many differences between racial groups that are starkly borne out in population-level statistics. Just so I don’t have to worry about anyone wanting to get me tossed out of society, let me give an example involving white males. Let’s say the IAT shows a bias toward associating white males with opiate overdoses versus, say, Asian females: is it any defense to refer to statistics which show that indeed white males are much more likely to overdose? The human mind discriminates, that is to say, mathematically speaking, it makes sets—evolution gave it this sublime power. If an IAT shows people associate white males more strongly with overdose, are those people immoral? Does it tell us anything about what the people actually think, in their ‘rational’ minds, about white males? About how they view white males’ inherent human worth and whether that value is any different from other groups? Does it probe to touch on questions of human feeling, moral reasoning, government policy, civilizational architecture? No, it doesn’t. It clearly does not demonstrate anything like ‘racism’ which is the belief that a racial group is inferior, with all the subjective value judgment the term ‘inferior’ implies.

      1. We should just throw people in the water and see if they float.

        If they float they are racist, if they drown they aren’t.

  3. As to Star Wars, the original 1977 movie always struck me as a patriotic depiction of the Second World War, albeit being fought in space.

    Nazi-like bad guys intent on genocide (killing everyone on a planet would constitute “genocide”) and the under-equipped good guys succeeding in their heroic struggle to defeat them. The cockpit of the Millennium Falcon was that of a B-29 and the anti-spacecraft guns with their alternating barrels resembled the anti-aircraft guns of the WW-II era.

    “Luke, I am your father” and everything else came later….

    1. Lucas had no original thoughts here. He was simply copying the serials of the 30’s adding in some WW2 Dambuster like attacks. If he hadn’t developed a new camera technique, Star wars would be forgotten 1970s pulp.

  4. What I’ve yet to see is results of these bias tests broken down by race. Do the responses from blacks, Hispanics, Asians, etc. differ from those of whites? And by how much? What happens if blacks have the same results as those of whites? What’s the research on that?

    1. And then there are the distinctions within the cadre of Blacks — “light skinned” Blacks versus “dark skinned” Blacks, along with “American-Born” Blacks versus Immigrant Blacks. Some of the worst racism I have ever witnessed was between Blacks born in Africa and Blacks born in DC.

      As to “Asians” — I’ve never met one. Instead, I have met people from China, Japan, Vietnam, Cambodia, The Philippines, etc. These are peoples with cultural hatreds (i.e. racist attitudes) that date back THOUSANDS of years — yes, some of it dates back to what the Japanese did during WW-II but even that was predated by things that had happened thousands of years earlier.

      An example, on a *very* cold day, a Chinese friend refused to accept a ride in my Toyota until I showed him the “1” on the VIN that indicated that the car was built in Canada. Talk about “microaggressions”….

      And as to Africa itself, let us not forget the tragedy of Rwanda…

      1. I lived on Oahu, Hawaii for seven years. I’m white and a Haole. See:

        I remember reading books aimed at tourists that talked about the beautiful “harmony” between the many cultures in Hawaii.

        Once you live there and get a feel for the real attitudes of the “locals” you realize that this is a preposterous joke.

        Under the surface, you find the various cultures despise each other.

        The one attitude where they seem to be united is that they hate Haoles more than they hate each other.

      2. Look up the brown paperbag test. I’m married to a black woman and some of the discoveries I have had about black culture… how light skinned blacks discriminate and look down on their darker skinned colleagues was an eye opener for me.

        Anyone who tells you blacks can’t discriminate just needs to be asked about the brown paper bag test.

Leave a Reply

Your email address will not be published. Required fields are marked *