The University of Chicago Magazine

Social psychologist Joshua Correll uses a video game to test whether racial bias triggers the decision to shoot a suspect.

In Correll's video game, images of suspects--both armed and unarmed, black and white—flash onto a monitor. Within a split-second, subjects must decide whether to shoot.

On the day U.S. attorneys decided not to file federal charges against four New York City police officers for killing Amadou Diallo—a West African immigrant whom the officers fired upon 41 times before discovering that the object he’d been reaching for was not a gun after all—social psychologist Joshua Correll was in Denver watching the news with his father. It was January 31, 2001. Eleven months earlier, jurors had acquitted the officers of second-degree murder in a New York State criminal trial, and this latest announcement sparked fresh protests from those who saw Diallo’s death as evidence of police brutality and racial profiling.

Correll, set to begin PhD studies that fall at the University of Colorado, found himself wondering about “all the trouble that had come from trying to interpret what happened on that night in 1999.” Would different circumstances—race, in particular—have yielded a different outcome? “What if the officers had approached a white guy and he had run into the vestibule of his apartment building and reached for a wallet?” as Diallo did, Correll asks. “What would have happened—in that neighborhood in the Bronx in the wee hours of the night?” The fact is, he says, “we don’t know.”

Nevertheless, he has tried since then to wend his way toward an answer. In four years of doctoral work and two years of Chicago research as an assistant professor of psychology, Correll has examined how racial bias plays into an officer’s decision to shoot a suspect. Using images of white and black men, each gripping a cell phone, a wallet, or a handgun, Correll and his Colorado colleagues devised a video-game experiment that requires split-second judgments. One after another, images flash onto a monitor and participants must assess whether the man in each picture is carrying a gun. Within 850 milliseconds (or fewer, depending on “how much we want to push people,” Correll says), they must press one key to shoot or another to leave the figure unharmed. The “targets,” as Correll calls them, stand in different poses—kneeling, striding, arms crossed, hands near their pockets—and they’re placed before mostly urban backgrounds: a public fountain, an apartment-building courtyard, a construction site, a leafy park, a parking lot.

In experiment after experiment—Correll has tested undergraduates, DMV customers, mall food-court patrons, and police officers—people’s mistakes, although rare, follow a pattern: they shoot more unarmed blacks than unarmed whites, and they fail to shoot more whites than blacks who turn out to be holding weapons. Recounting the results of four separate studies in a 2002 Journal of Personality and Social Psychology article, Correll and fellow researchers wrote, “In the case of African American targets, participants simply set a lower threshold for the decision to shoot.” That trend held true even when the participants themselves were African American.

The trend is less likely a function of active prejudice than of ambient social stereotypes, Correll says. These cultural biases “come not from what you personally believe or want to believe, but from long-standing associations drilled into our heads every time we go to the movies or pick up a newspaper or hear a joke.” In surveying the participants, he has found keen awareness of stereotypes more reliable than racial prejudice at predicting performance. “So those who report that in America black people are more often seen as violent—not those who actually consider black people to be more violent—are the ones most likely to show bias.”

A study in the June 2006 Journal of Experimental Social Psychology offers proof of how deeply lodged those stereotypes are. In that test, Correll hooked electrodes to participants’ scalps to monitor the electrical activity of neurons firing as the video game played out. “Surprise, surprise,” he says, “the P200s”—a neuronal voltage jump associated with threat responses—“tended to be bigger for black faces than for white faces.” Especially strong P200s translated to more pronounced bias in the video game. “This fluctuation is happening just 200 milliseconds after the stimulus appears on the screen,” Correll says. “We’re talking very, very quick—preconscious. This is your first gut response.”

Correll’s latest experiments involve urban police officers. Overall, they’ve proved quicker and more accurate than ordinary citizens. “They make very few mistakes,” Correll says, “which is reassuring.” But they aren’t free of bias. For a study published in June’s Journal of Personality and Social Psychology, Correll recruited Denver police officers, Denver residents, and cops from 14 states to play the video game. On one key measure—the correlation between race and reaction time—officers showed the same susceptibility to stereotypes as civilians. “When they see a target that contradicts the stereotype,” a black person without a gun or an armed white person, “they hesitate,” he says. “They wait a couple of extra milliseconds, but they don’t make the wrong decision.”

Two other studies in the same JPSP article offer evidence that bias can be trained out of people. When participants—whether police, civilians, or students—played the game four times over successive days, “they got better.” Yet reaction-time bias persisted. “What changes is the number of mistakes.” Identifying a tiny object like a gun against a complex and changing background requires control and discipline, Correll says, especially when the image doesn’t conform to ingrained expectations. Police training teaches control and discipline, making officers’ mistakes rarer. But reducing errors is “as good as it gets,” he says, “unless we can change all the cultural stereotypes in the country.”

Plenty of research lies ahead, Correll says. He plans to investigate a finding from his experiments involving police: that race-influenced delay proved more pronounced in big-city officers than their small-city counterparts. He’d like to conduct more tests measuring neuronal fluctuations, and Correll is also beginning to use images of Hispanic and Asian men to examine how subjects from different regions react to those ethnicities. With every test, he says, he’s closing in on what makes police pull the trigger. “It’s still a very messy question.”