This post has been updated.

Most white Americans demonstrate bias against blacks, even if they're not aware of or able to control it. It's a surprisingly little-discussed factor in the anguishing debates over race and law enforcement that followed the shootings of unarmed black men by white police officers. Such implicit biases -- which, if they were to influence split-second law enforcement decisions, could have life or death consequences -- are measured by psychological tests, most prominently the computerized Implicit Association Test, which has been taken by over two million people online at the website Project Implicit.

Based on this data, it appears that whites in some states may exhibit higher levels of implicit bias than those in other states. The following map, courtesy of Project Implicit, shows the states with the highest level of implicit bias (high number, red) and lowest level of implicit bias (low number, blue). Gray represents states with a middle amount of implicit bias; Michigan is the median state. Overall, the map reflects the scores of 1.51 million individuals, ranging from a high of 99,660 test takers from California to a low of 1,722 test takers from Hawaii.

Courtesy of Project Implicit.
Courtesy of Project Implicit.

A cautionary note: The people who have taken the IAT at the Project Implicit website are not a random sample of Americans, either nationally or on a state-by-state basis. Rather, they're people who, for some reason, chose to take an online test measuring their implicit biases -- which may actually mean they are less biased than average. (After all, at least they wanted to know how biased they are.)

"Please keep in mind that this map describes volunteers for the online IAT," says psychologist Anthony Greenwald of the University of Washington,  who created the Implicit Association Test in 1995. "These volunteers are younger, more educated, more politically liberal, and more female than the U.S. population as a whole."

So, let's take some time to unpack further what this image means -- and doesn't mean.

The Implicit Association Test comes in many versions, but in a version that detects uncontrolled racial biases or preferences -- as opposed to, say, gender bias or age bias -- your task is to rapidly sort a series of faces as either "African American" or "European American," even as you also sort a variety of words (like "agony," "joy," "happy," "anger") as either "good" or "bad."

Bias in the test occurs when people are faster at categorizing negative words when they are paired with African American faces, or faster at sorting positive words when they're paired with white faces -- suggesting an uncontrolled mental association between negative things or concepts and African Americans.

(The "pairing" occurs because you might be asked to sort both "African American" faces and "bad" words to the left side of the computer screen, and "European American" and "good" to the right. And since the differences in sorting speed are measured in milliseconds, this is not something people can consciously control.)

Bias in the IAT is measured in terms of a "D" score, which signifies how much faster people are in their categorizing when African American faces are paired with "bad" concepts (and white faces with "good" concepts or words) as opposed to vice versa. Any score over 0 means that  "when white and good are on the same side, the participants are responding faster than when black and good are on the same side," explains Kaiyuan Xu, a data analyst for Project Implicit.

It is very important to note that implicit racial bias is not the same thing as conscious racism. People who harbor implicit biases may not think of themselves as prejudiced, and in fact, might consider prejudice to be abhorrent. They also may not know they even have these biases.

With this background in place, one key thing to notice about the map above is that white people in every U.S. state are biased. Their mean scores vary by state, but participants from the median state, Michigan, show an average, positive IAT score of 0.402. According to Xu, a score of .35 is the "cutoff point between 'moderately prefer white' and 'strongly prefer white.'" [Update 12/19/2014: Kaiyuan Xu contacted us to let us know that this prior statement was incorrect, and gave us a new statement: "A score of .35 is the cutoff point between 'slightly prefer white' and 'moderately prefer white.'" A score of .65 would actually be the cutoff point between ‘moderately prefer white’ and ‘strongly prefer white.’]

Xu says that at this point, researchers from Project Implicit cannot fully explain the most striking feature of the map, which is the fact that implicit bias levels appear stronger in the U.S. southeast and east.

It is also worth noting that there are reasons to think that Americans as a whole may be more biased than the map suggests. After all, as Greenwald noted above, the Project Implicit test takers trend younger than average, as well as towards liberal political beliefs and higher levels of education. But other research has suggested that older Americans past the age of 65, in particular, tend to have higher IAT D scores -- suggesting that those included in this sample may be less biased than Americans as a whole.

Overall, looking at a map like this one tells us something pretty crucial to our understanding of racial bias: It is everywhere, from north to south, from Maine to California. It is present among liberals and conservatives, men and women, young and old.

We have a huge amount of work to do.