The Implicit Association Test (IAT) is a computer-based tool that measures the strength of automatic mental associations between concepts, like race and positive or negative qualities. Created in 1998 by psychologists Anthony Greenwald, Debbie McGhee, and Jordan Schwartz, it works by timing how quickly you sort words and images into paired categories. The core idea: if two concepts are strongly linked in your mind, you’ll pair them faster than concepts you don’t automatically connect.
Since its launch, the IAT has become one of the most widely used tools in social psychology. Millions of people have taken it through Harvard’s Project Implicit website, where versions cover race, gender, sexual orientation, age, weight, anxiety, and other topics. It has also sparked significant debate about what it actually reveals and how useful those revelations are.
How the Test Works
The IAT is built on a simple principle: sorting tasks get easier when the two categories sharing a response key feel “natural” together in your mind, and harder when they don’t. During the test, words and images flash on screen one at a time, and you press one of two keys to sort each item into its correct category as fast as you can.
A typical race IAT has you sort faces of Black and White people alongside positive words (like “joy” or “wonderful”) and negative words (like “terrible” or “awful”). In some rounds, “White faces” and “good words” share one key while “Black faces” and “bad words” share the other. In later rounds, the pairings flip. If you’re consistently faster when White faces are paired with good words than when Black faces are paired with good words, the test interprets that speed gap as an implicit preference for White over Black.
The test typically includes practice blocks of about 20 trials each, where you first sort just the words, then just the faces, before combining them. These practice rounds let you learn the key assignments before the critical combined rounds that generate your score.
What the D-Score Means
Your result isn’t a raw reaction time. It’s a calculated value called a D-score, which accounts for both the speed difference between the two pairings and the natural variability in your response times. The algorithm takes the average reaction time from the “compatible” pairing (the one that felt easier) and subtracts it from the “incompatible” pairing, then divides by the overall standard deviation of all your responses. This standardization prevents people who are generally slow or fast typists from getting artificially inflated or deflated scores.
The final D-score typically falls somewhere between -2 and +2, with scores near zero indicating little or no automatic preference. Scores are usually reported as slight, moderate, or strong preference in one direction. Trials where you took longer than 10 seconds are thrown out, and if more than 10% of your responses are faster than 300 milliseconds (essentially random button-mashing), your data is excluded entirely.
Why Implicit and Explicit Attitudes Differ
Many people get IAT results that surprise them. Someone who genuinely believes in racial equality may still show an implicit preference favoring one group. This discrepancy is central to dual-process theories of cognition, which propose that the brain handles information on two tracks. Implicit processes operate rapidly and outside conscious awareness. Explicit processes are slower, deliberate, and accessible to introspection.
The divergence happens because explicit thinking can override or refine initial automatic reactions. When you consciously reflect on a social situation, you draw on past experiences, consider context, and correct for snap judgments. Brain imaging research supports this: initial automatic responses, like heightened activity in threat-processing areas when viewing outgroup faces, are often reevaluated through later, more deliberate processing that produces a non-biased response. The IAT captures that first, fast layer before the deliberate correction kicks in.
How Well It Predicts Behavior
This is where the IAT gets controversial. The central question is whether the automatic associations it detects translate into real-world discrimination, and the honest answer is: somewhat, but not strongly.
Two major analyses have examined this. One found that race IAT scores correlated with discriminatory behavior at about r = .24, while a competing analysis put the figure lower, at roughly r = .15. Both groups ultimately agreed that the predictive validity of the race IAT lands around r = .20. In practical terms, that means IAT scores explain about 4% of the variation in discriminatory behavior. It’s a real statistical relationship, but a modest one, leaving the vast majority of behavioral differences unexplained by the test alone.
The test’s reliability over time also raises questions. A meta-analysis of children’s IAT scores found test-retest correlations ranging from .34 to .48 depending on the scoring method. For adults, reliability estimates are generally higher but still moderate. Your score on Monday may differ meaningfully from your score on Thursday, which makes it difficult to treat any single IAT result as a stable measure of who you are.
The IAT in Healthcare
One area where implicit bias research has gained traction is medicine. A review of 42 studies found a consistent positive relationship between healthcare providers’ levels of implicit bias and lower quality of care for minority patients. In orthopedic surgery specifically, a study of 119 pediatric orthopedic surgeons found that over 85% demonstrated some degree of implicit racial bias on the IAT. Twenty-nine percent showed strong pro-White implicit bias, compared to 19% of the general U.S. population.
These findings don’t mean that a high IAT score causes a doctor to deliver worse care in any individual case. But across large numbers of clinical encounters, the pattern is consistent enough to raise concern about systemic effects on treatment decisions, pain management, and referral patterns.
Workplace Bias Training and the IAT
Many organizations now use the IAT as part of unconscious bias training, hoping that awareness of hidden preferences will change behavior. The evidence for this is thin. A comprehensive evaluation of bias training programs found that out of ten studies designed to change behavior, only two actually measured whether behavior changed, and even those used methods with limited validity because they didn’t observe real-world actions.
The effects of bias awareness also tend to fade quickly. One study of 292 university students found that the impact of an unconscious bias intervention waned within two weeks. A follow-up two years later offered a more hopeful finding: participants who had gone through the training were more likely to publicly challenge an essay endorsing racial stereotypes. But overall, the research on whether taking the IAT and learning about implicit bias leads to lasting behavioral change remains inconclusive.
Training programs do appear more effective at raising awareness than at changing outcomes. One intervention with a multidisciplinary school team showed meaningful improvements in staff members’ recognition of their own vulnerability to bias and in their expectations for individual African American students. Awareness is not nothing, but it’s also not the same as measurable reductions in discrimination.
What a Score Does and Doesn’t Tell You
The IAT captures something real: the automatic associations your brain has absorbed from a lifetime of cultural exposure. Growing up in a society where certain groups are consistently portrayed in particular ways leaves traces in your mental architecture, whether or not those associations match your conscious values. The test reliably detects these traces at the group level, which is why it remains a valuable research tool for studying how bias operates across populations.
What it cannot do reliably is diagnose an individual as biased or predict how any one person will behave in a specific situation. The modest predictive validity, the score instability over time, and the gap between implicit associations and conscious decision-making all limit what a single IAT score can tell you about yourself. If you take the test and get a result you didn’t expect, it’s worth reflecting on, but it’s not a verdict.