An explosion of studies over the last couple of decades have claimed that almost everyone is subconsciously a racist, a sexist, or a bigot of some other sort.

These studies often rely on a new method of studying bias, called the Implicit Association Test. But now it seems this research method may not have been telling us what we thought after all.

Unconscious Biases

Among the less controversial insights of Sigmund Freud retained by modern, scientific psychology is the assumption that many important human motivations lie in the subconscious: inaccessible, beneath our awareness, and — perhaps most importantly — unavailable to external observers (such as, say, psychological scientists).

One subject of particular interest is bias against racial groups, women, and religious minorities. Unfortunately, these topics are notoriously difficult to study. Most of us would be reluctant to state outright that we believe our race is morally superior to all others, or that we prefer the company of members of our own race over members of other races. But this doesn’t mean we don’t harbor those beliefs. In fact, we might not even be aware of our own biases in the first place.

This problem has bedeviled psychology for decades. How do you study what you cannot measure directly? Imagine if physicists had no instrument capable of measuring weight or energy, or if biologists were forced to labor without the aid of microscopes.

To a large extent, the state of a science is defined by the tools that aid its progress. Today, much of psychological research is performed with rather blunt instruments: structured interviews and paper-and-pencil questionnaires that can do little to help us understand attitudes that the subjects of the study are unable (or unwilling) to admit to themselves, let alone others.

The Implicit Association Test

The Implicit Association Test (IAT) was developed to overcome those limitations and tap into underlying biases that would otherwise be difficult to measure. The assumption behind the IAT is that if certain concepts have positive connotations — even on an unconscious level — then they will be easily associated with each other. In a typical IAT session (you can try it out for yourself), a participant will be seated in front of a computer and instructed to press keys in response to different pairs of stimuli on opposite sides of the screen, combined with an image in the middle.

In the classic test of racial bias, a participant might be instructed to press the “E” key to indicate when an African American face or word with a positive connotation (e.g., “happy”) was displayed in the center of the screen, and the “I” key whenever a white face or word with a negative connotation (e.g., “harmful”) was displayed. After completing one session with a set of pairings, the task would reverse: white faces would be paired with positive words, and black faces would be paired with negative words.

The key is that the stimuli being displayed are constantly bouncing back and forth, forcing the participant to decide which button to press each time; if a person harbors bias against African Americans, they will respond faster when photos of African Americans are paired with negative words, because they make those associations with relative ease.

Voila! A measure of heretofore inaccessible unconscious bias.

Many psychological scientists (particularly in my own sub-discipline, social psychology) embraced this method and applied it with gusto to the study of implicit bias toward racial minorities. Anti-black bias was a common target, and the idea that nearly all white individuals (and even some black individuals) reliably exhibit implicit, negative associations with black faces quickly became part of the social psychological canon.

My own undergraduate social psychology textbook, published just four years after the IAT was formally introduced, explained the test’s promise:

In clever experiments by Anthony Greenwald and his colleagues (1998, 2000), nine in ten White people took longer to identify pleasant words (such as peace and paradise) as “good” when associated with Black rather than White faces. The subjects, mind you, typically expressed little or no prejudice, only an unconscious, unintended response.

The IAT also garnered attention outside the field, due in no small part to careful marketing by the tool’s creators. You have almost certainly heard about the IAT, even if you don’t explicitly (pun intended) know it. Today, media accounts routinely sport confident headlines (“Implicit Bias Is Real. Don’t Be So Defensive”) and treat results from the test as establishing, without a doubt, that we are all harboring latent racist tendencies (e.g., “The science of your racist brain”).

The lesson many were taking from this body of findings was that most people (if not everyone) — no matter how progressive they might think they are — harbors some bias. At its most problematic, implicit bias has been said to explain everything from redlining to police shootings.

A Troubled Instrument

But as Tom Bartlett and Jesse Singal point out in their recent pieces on the IAT, there are numerous problems with the instrument itself. First and foremost, we must recognize that if it assesses anything at all, the IAT is a proxy (i.e., indirect) measure of bias, and a rather crude one at that. Let’s say it takes me longer to press a key in response to a positive word paired with a black face than a positive word paired with a white face.

Does this mean that I harbor anti-black bias? Possibly. Or, perhaps I just have more exposure to white faces than black faces? That’s possible, too. The literature is by no means definitive on which explanation best fits the data.

The IAT also appears to be a poor predictor of actual behavior (which is, after all, what we’re really concerned with if we are trying to explain phenomena such as hiring discrimination and police violence). A recent meta-analysis (a large scale “study of studies,” where the results from multiple papers are statistically analyzed together) indicated that the multiple versions of the IAT failed to provide results that were any better than simple, explicit questionnaires (i.e., sophisticated ways of asking, in a straightforward manner, “are you racist?”).

Finally, the IAT exhibits extraordinarily low reliability: you take it in the morning and you exhibit bias; take it in the afternoon and you don’t. We don’t normally expect separate administrations of a psychological test to be identical, but as Singal points out, even the most generous estimates in this area show the IAT to be woefully inadequate.

An Instrument of Bias

Of course, it is indeed possible — if not probable — that most people harbor some form of preference for their own racial, ethnic, or religious group (take your pick — the IAT has been used to measure a variety of bias types). But what the media, advocates, and many scholars have claimed is that the effect of such preferences can explain a whole host of social ills.

The reality is certainly more complex: in the United States. For example, income disparities by race exist due to a variety of factors, starting with explicit, state-sponsored oppression from legalized slavery through the Jim Crow era, and including emergent cultural norms that reified bigoted assumptions and further marginalized racial and ethnic minorities.

State-sponsored housing programs concentrated poverty in urban areas, and state-run school systems —  free of any market influence — failed our most vulnerable populations. In short, our modern problems regarding race are sociological, educational, and economic in nature. To whatever extent psychological factors play a role in modern race relations, they do so within a much broader context. A focus on unconscious bias ignores that simple fact.

But what is perhaps most fascinating about the IAT is its meteoric rise, despite failing to meet the most basic criteria for a psychological measure. The introduction of psychometric instruments usually requires years of study, scrutiny and revision before they are broadly accepted by researchers. Had the IAT been a measure of a comparatively mundane construct (say, extraversion or anxiety), I doubt it would have gained widespread acceptance with such velocity.

Ironically, the popularity of the IAT may be due to bias — specifically, the desire of an entire field to explain the complex phenomenon of racial disparities and bigotry in a manner compatible with leftist ideology. And while the test’s designers state explicitly that they do not yet have a cure for the implicit bias that ails us, that is surely their implication.