Eric Garner, Michael Brown, Akai Gurley, Walter Scott, Samuel DuBose—the list of high-profile killings by police in the U.S. continues to grow.* When this story went to press in September, officers had already shot and killed 680 people in 2015, according to a database maintained by the Washington Post. By its count, only 6 percent of the white suspects were unarmed, compared with 14 percent of the black victims. If the numbers reflected U.S. demographics, unarmed white deaths should be twice as high—and unarmed black deaths should be three times lower.

That African-Americans bear the brunt of unjustified police violence is not news: data from the U.S. Centers for Disease Control and Prevention, the Department of Justice and other sources indicate that blacks are three times more likely to be killed by police than whites are. But in recent months the visibility of the issue has surged. Highly publicized protests in Ferguson, Mo., and elsewhere; damning smartphone footage shared far and wide via social media; and movements such as Black Lives Matter have ignited a call for serious reform.

The outcry and demand for change now stem from citizens and police forces alike. Before determining how best to curb police violence against unarmed black citizens, though, many law-enforcement experts and scientists are trying to understand the psychological origins of the problem. Research is yielding some clues about how bias, fear and a lack of sleep, among other factors, can give rise to deadly split-second errors in judgment and action.

The larger challenge will be figuring out how to harness these insights to create evidence-based training programs that can prepare officers to cope justly with the unpredictable and life-threatening circumstances that are intrinsic to the job. The fact is that the exact relation between police violence and racial bias remains an active and unsettled area of investigation. And not all the answers are in on whether training designed to reduce bias will be effective or if other approaches might offer greater benefits. But propelled by the current crisis, social psychologists, criminologists and neuroscientists are working on a range of remedies.

Bias in the brain

Modern society looks down on overt expressions of bigotry, but evidence from both the real world and the laboratory betrays a darker truth: even as racial discrimination has lessened, racial inequalities have not. Blacks in America continue to face higher levels of poverty, incarceration and unemployment, among a myriad of other inequities. “Prejudice and bigotry are in retreat,” says social psychologist Phillip Atiba Goff, currently a visiting scholar at Harvard University. “But I think we're a long way from having anything to celebrate.”

For decades researchers such as Goff have sought to explain the ongoing racial divide, tracing it in part to something called implicit bias. This form of bias is so subtle that scientists find it even among people who appear to harbor no obvious prejudices. Unlike blatant racism, implicit bias is not an individually held belief but is one generally shared by everyone in a society. Because our brain naturally makes sense of the world by grouping things into categories, we all generate unconscious stereotypes based on the generalizations we absorb through experiences that include movies, television, music and the news.

With time and reflection, most people can mentally correct for implicit biases. “If I'm asking you to take a long, hard look at a job candidate, your implicit biases are not in play,” Goff says. But in highly stressful situations, he adds, they can govern our actions. “If I get your heart rate up, get your adrenaline pumping, and say, ‘If you don't make the right decision immediately, there will be consequences for you and your family,’ then you may end up relying on implicit biases.” In other words, implicit biases come into play precisely in the kinds of situations that lead to police shootings of unarmed suspects.

Beginning in the early 2000s, social psychologist Joshua Correll of the University of Colorado Boulder and his colleagues began a series of experiments in which they asked people to play a fast-paced video game that involved opponents facing off with various armed and unarmed suspects appearing on the screen. “Technically, skin color is irrelevant in this simple task, which is to shoot the guys with guns and not the guys without guns,” Correll explains. “But what we find is that black targets, which society stereotypically associates with threat or danger, are more likely to lead to a shooting response.”

Indeed, Correll has observed that his study subjects are more likely to mistakenly fire at an unarmed black avatar than at a white one. Similarly, they are faster to shoot at an armed black target than at an armed white one. And they are quicker to deem an unarmed white figure nonthreatening, compared with an unarmed black one. These patterns hold up whether a shooter is outwardly racist or not—and even when the shooter is black.

Kurt Hugenberg and Galen V. Bodenhausen, both then at Northwestern University, further discovered that the more implicit bias a white person harbors, the more likely he or she is to perceive a black face as hostile. Again, this reaction reflects implicit prejudice, so people are unaware of the perceptual skew. “This means we can't just say, ‘Don't go shooting friendly black people,’” says David Amodio, a psychologist and neuroscientist at New York University. “Stereotyping is already causing a neutral black face to appear as more threatening.”

Amodio and his colleagues have looked for what prompts some of these responses in the brain. In a series of experiments during the past decade, they have found that when white volunteers are presented with a black face, they appear to experience more fear than they do in response to a white face. Specifically when study participants look at black faces, they have stronger startle reflexes linked to activation in the amygdala, which is involved in emotional responses.

Our implicit biases can render black faces not only more threatening but less human, too. In 2008 Goff and his colleagues reported that people who had been subliminally “primed” by viewing photographs of black faces instead of white ones were faster to identify obscured, pixelated images of apes that slowly improved in visibility over time. The team also found that news articles written about black convicts are more likely to contain dehumanizing words (such as barbaric, beast, predator, stalk and savage) and that convicts portrayed as being more apelike have a greater chance of receiving a death sentence.

“If in the deep recesses of someone's mind, they perceive white people as being more human than black people, they'll respond to those groups differently,” says social psychologist Jack Glaser of the University of California, Berkeley. “Combine that with bias to see a weapon, and those two things go a fairly long way to explain what we're seeing with use of deadly force.”

“Man with gun”

Last November, just five days before Thanksgiving, 12-year-old Tamir Rice was playing with a toy gun in a park in Cleveland. According to a call from a concerned citizen, he was pointing it at people. Two police officers, 26-year-old Timothy Loehmann and his partner, 46-year-old Frank Garmback, responded. As soon as the dispatcher radioed over key descriptors—black, male, gun (she did not mention that the caller described it as “probably fake”)—they likely triggered subconscious impressions for those officers. “Once these stereotypes are activated in the mind, they start to guide the way we see a situation,” Amodio says. “You're tuned to see a hostile person who definitely has a gun.” Loehmann certainly did. As captured on tape, he jumped from the patrol car as it rolled to a stop and fired twice. The toy-wielding child died the next day.

These so-called threat-perception failures—mistaking a toy gun for a real one or a cell phone or some other object for a weapon—are not uncommon. An investigative report commissioned by the Department of Justice (DOJ) on the use of deadly force in Philadelphia found that these psychological blunders accounted for 49 percent of officer-involved shootings of unarmed citizens from 2007 to 2013. Both black and white officers were more likely to make errors when the suspect was black.

Goff and his colleagues have found similar misperceptions in the lab. They subliminally exposed both undergraduates and police officers to images of black and white faces and then asked them to identify mystery objects in deliberately blurred photographs. The subjects were faster to correctly label guns in the degraded images after “seeing” black faces. In reverse, they were also quicker to focus their visual attention on black faces after seeing split-second images of guns: the participants spotted a moving dot near a black face faster than they noticed one near a white face. “Thinking about black people makes people think about weapons, and thinking about weapons makes people think about blacks,” Glaser says. “So officers, when confronted with a black person, are more prone to see a weapon.” The fact that blacks are regarded as threatening, Goff points out, “is not endemic to the culture of policing exclusively—it's an American problem.”

Goff notes that stereotype-driven intuition can readily cause a cascade of erroneous suspicion. When a cop walks toward a person he or she believes looks suspicious, that person may begin to seem even more uneasy—precisely because the police officer is approaching. As Goff points out: “Black folks get nervous when they're worried about being stereotyped as criminals.” As the suspect becomes more uncomfortable, the officer's suspicions are reinforced. If the suspect tries to evade the situation, an altercation can ensue. “Implicit bias plays a role in every one of those steps,” Goff says. “It greases the wheel of disaster in terms of interpersonal interactions.”

Other emotions and circumstances can also lead to dangerous escalations. In a study published in 2013 in the journal Basic and Applied Social Psychology, Correll and his colleagues found that both undergraduates and police recruits who did not get enough sleep—even by just an hour—had less cognitive control. As a result, they were more likely to act on bias and erroneously shoot at unarmed black targets in a video game. Exhaustion may well play a role in some real-world shootings, too: a 2000 DOJ study found that between 19 and 41 percent of examined officers showed signs of severe sleep deprivation, depending on the test used to measure the deficit. In the test that found the former figure, the exhaustion caused slightly more than 6 percent of subjects to suffer impairments on par with being legally drunk.

In other situations, tempers suddenly flare out of an urge to exert power and dominance. Police can be readily caught up in a “good guy/bad guy” mind-set, charged, as they are, with maintaining law and order and keeping the peace. In an infamous 2014 incident in Staten Island, N.Y., Daniel Pantaleo, a 29-year-old officer, accidentally suffocated 43-year-old Eric Garner, an unarmed black man he suspected of illegally selling loose cigarettes, restraining the asthmatic Garner in a choke hold when he resisted officers' attempts to handcuff him. Patrick J. Lynch, president of the Patrolmen's Benevolent Association of the City of New York, was quick to offer an explanation that reflected this line-in-the-sand mentality. Too many people attempt to protest police decisions on the street, he said in a press conference, and “resisting arrest is a real and dangerous crime.”

In yet other cases, impulsivity triggered by fear lays the groundwork for deadly force. This past summer 25-year-old officer Ray Tensing killed 43-year-old Samuel DuBose, another unarmed black man, during a routine traffic stop in Cincinnati. Tensing made a snap decision that DuBose would run him over and panicked, shooting him in the head. “Clearly, that officer was incompetent and should never have been a policeman, but I don't think his intention was to kill anyone that day,” Glaser says. “Police are human, and there will be variability in their maturity, self-control and courage in a situation like that” [see “Who Should Be a Cop?” on page 50].

Can better training help?

Some studies hint that it may be possible to reduce implicit bias through exposure. In 2005 Ashby Plant, a social psychologist at Florida State University, and her colleagues described an experiment using a shooter video game similar to Correll's. They recruited 50 primarily white officers and found that the more they played the game—in which black and white “suspects” were equally likely to be unarmed—the more accurate and less biased they became in their targeting. In Correll's experiments, too, cops generally made better targeting decisions than undergraduates did—they were no more likely to shoot unarmed blacks than they were unarmed whites—but they were still faster to pull the trigger at black avatars than white ones when they did fire.

Whether such improvements through training can translate from the lab to the streets, Correll says, “is the $10,000 question.” Other research also indicates that increasing the number of positive interactions with members of stereotyped groups might help blunt subconscious bias. Ken Paller, director of the cognitive neuroscience program at Northwestern, and his colleagues have demonstrated the potential merits of so-called counterstereotype training—seeing faces paired with words that run counter to cultural stereotypes, for instance. They are even investigating whether pairing sounds with these antistereotypes—and then playing those sounds when subjects sleep after the training—might concentrate their effects.

But no matter how promising any of these results may seem, “the real world is not a safe, controlled little video game with pictures popping up on a screen,” Correll says. “In the real world, it's chaotic and messy, and you're scared for your life.” Plant adds: “We don't want to make large generalizations about things we see in undergraduates in the lab and then suggest that the New York City Police Department do that.”

Although statistics on cop shootings of unarmed blacks suggest that implicit bias plays a key role, Correll points out that the association, no matter how compelling, is only correlational. Are officers really homing in on race, or is skin color confounding the actual variable (or variables) driving the disparity? “We know that a black person is much more likely per capita to die at the hands of an officer—just squint at a DOJ report, and that's painfully obvious,” Correll says. “But in the exact same situation, if a white guy and a black guy are in the same neighborhood, wearing the same clothing and behaving in exactly the same fashion, would an officer treat them differently? That, we don't know.”

In fact, no official national data exist on police behavior or officer-involved shootings. “They say you can't hide a body, but it turns out that criminal-justice-statistics folks miss a fair amount of homicides, partly because they categorize them as accidental deaths, and the Bureau of Justice Statistics does not track officer-involved shootings directly,” Glaser says. “There's no one clearinghouse for reliable data.”

To amend that situation, Goff—with the support of Glaser and other researchers—has co-founded the Center for Policing Equity. The consortium will serve as the first national justice database and collect information about pedestrian stops, complaints filed against officers, and those officers' discipline records and their use of force. So far departments responsible for policing more than one third of the U.S. by population have agreed to participate.

“The first goal is just to figure out how often this stuff happens, how severe it is, under what conditions it happens and how racially disparate the treatments are,” Goff says. “This database will allow us to begin—emphasis on begin—to make distinctions between law enforcement's responsibility for racial disparities in policing and the responsibility of the rest of us for creating a racially disparate criminal justice system.”

If we currently lack basic understanding about policing and the role that implicit bias plays in it, we know even less about how to counter the problem. “We're still a long way off from understanding why, exactly, a controversial shooting incident might occur, much less knowing how to implement a large-scale intervention to prevent such things,” Amodio says. “My guess is that any current proposals for reducing such incidents are, at the present time, based more on speculation and opinion than evidence.”

Premature interventions, Correll warns, could cause more harm that good. In one telling study from 2001, Keith Payne, now at the University of North Carolina at Chapel Hill, and his colleagues investigated how conscious goals can be used to control the unconscious influence of stereotypes. As expected, participants initially misidentified harmless objects as weapons more often after seeing an image of a black face; after seeing white faces, they were more likely to misidentify a weapon as a harmless object. But when the researchers admonished the participants to remain unbiased, their performance actually got worse. “It boomeranged,” Correll says. “By trying not to think in racial terms, they have to think in racial terms, which leads to hypersensitivity to race and more bias.”

By extension, training designed to neutralize officers' feelings about race could wind up unintentionally having the opposite effect, putting suspects in even graver danger. Conversely, if cops are taught to be overly self-conscious of their implicit biases, Correll says, their ability to accurately judge a situation could be compromised, putting their own lives at risk. And if legislators, mayors and police chiefs invest time and money in half-baked interventions that ultimately do not work, he adds, they may just throw up their hands and write off any future reform efforts.

Reality check

While science incrementally chips away at these problems, police forces and the public are clamoring for immediate action. A number of training programs are currently being deployed [see “Camden, N.J.,” on preceding page]. Lorie Fridell, a criminologist at the University of South Florida, says she and her colleagues are overwhelmed with requests from police departments, and they are trying to build a greater awareness of implicit bias. “Many in policing reject the allegations of widespread biased policing because they look at themselves and those around them, and they don't see any racists,” she says.

There are no empirical evaluations of the new training programs in place, but a few have produced results anecdotally. Las Vegas, for example, hit an annual high of 25 officer-involved shootings in 2010. But after an aggressive reform program—including training in how to de-escalate confrontations and specialized courses on fair and impartial policing, the latter taught by Fridell and her colleagues—the stats improved dramatically. Through the end of August, Las Vegas officers had been involved in just nine shootings for the year to date.

Some of that progress can probably be attributed to the lessons in managing tense situations. “Sometimes officers may be justified at the last split second in using deadly force,” Fridell says, “but if you look at the earlier decisions they made, they put themselves in that situation.” To that end, Dennis Rosenbaum, a professor of criminology, law and justice at the University of Illinois at Chicago and executive director of the National Police Research Platform, calls for more “emotional intelligence and social skills” so that officers can more adeptly deal with someone who is upset and recognize if the person poses a genuine threat. He thinks officers need active training in these areas: “When police officers go to the shooting range, they are required to hit the target x percent of the time to get certified. Where's the proficiency requirement for social skills and conflict de-escalation?”

Jonathan Wender, a sociologist at the University of Washington with more than 20 years of experience as a police officer and a sergeant, and Brian Lande, a sociologist who currently serves as a patrol officer in Richmond, Calif., are attempting to fill this gap. In 2010 the Defense Advanced Research Projects Agency offered them a $40-million budget to design an evidence-based training course for teaching military personnel how to handle people of different backgrounds in high-risk situations. Now their company, Polis Solutions, is providing similar coaching to local police forces across the country. Antibias training is part of their program, but it primarily focuses on practical ways to reduce negative interactions between the police and public.

“We think it's important to pull back from these high-level, abstract explorations about politics, culture and morality and instead look at the dynamics of interactions,” Wender says. “We ask, ‘What makes a good stranger? What enables people, especially when things are tense, to get things right?’” Using evidence from psychology, anthropology, linguistics and cognitive science and supported by the DOJ, the philosopher cops, as they call themselves, train officers to build empathy and trust, taking cues from someone's words, body language, facial expressions and other traits. Trainees currently practice with actors but may in the future have access to “social encounter” simulators. “Social interaction must be taught like any kind of skill,” Wender says. “You can't teach someone to swim by telling them to read a book about it.”

Their training also uses case studies. When 30-year-old state trooper Brian Encinia pulled over 28-year-old Sandra Bland in July for failing to signal a lane change in Waller County, Texas, she griped, and he responded with bullying, threatening to “light [her] up” with a Taser. He arrested Bland, who was later found dead in her jail cell. Wender says her death likely could have been avoided if Encinia had taken a different approach: asking Bland why she was upset and expressing some understanding of her frustration. “There are identifiable errors that he shouldn't have made and could have been taught not to make,” Wender says. “These skills are eminently trainable, but they have to be delivered in a realistic and sustainable way that goes beyond just checking a box.”

Better social skills may help officers avoid scenarios that typically lead to force, but it is impossible to eliminate high-stakes situations altogether. To prepare, Glaser suggests that officers practice in sessions that closely mimic real-world crises, preferably with black actors wielding something like a paintball gun. Some deputies shy away from this idea for fear of being accused of training cops to shoot black people. But without such training, Glaser says, officers will be less able to avoid that very scenario. “Anecdotally, I've heard from officers that once a gun is drawn or you think your life is in danger, you'd better hope you're well trained because you're no longer under control,” he says. “Some say they've even soiled themselves—it's absolutely terrifying.”

If Wender and others were to develop empirically tested, effective and long-lasting training programs, they would face one final problem: how to implement them. Unlike many other countries, the U.S. does not have a single police force but instead has 18,000 individual departments that operate on state, city or county levels. For this reason, issuing a federal decree that every department must undertake some specified reform program would not work, Fridell says. To try to address that challenge, in late August the organizers behind Black Lives Matter launched Campaign Zero, a series of policy-reform recommendations tailored to the federal, state and local level.

Their recommendations regarding training largely echo what social psychologists are promoting—teaching better social and coping skills and raising awareness about implicit bias. Aiming too broadly at reducing implicit bias, however, is a moving target: “The black-danger and black-bad associations are malleable,” Correll says. “That means they can be reduced, but by the very same token, that also means they can come right back.” The only way to ever rid ourselves of implicit biases is to fix the underlying societal inequalities that create them. “Concerning ourselves with eliminating implicit bias is chasing our tails,” Goff emphasizes. “We should actually concern ourselves with eliminating inequality—the bias will follow.

Who Should Be a Cop?

Recruitment presents an enormous challenge in modern policing. In hopes of finding those people best suited to the role— interested in helping citizens and not just locking them up— many forces administer psychological examinations during the hiring process. There are no clear guidelines, but some studies hint at traits to look for

Paul Detrick of Florissant Psychological Services in Florissant, Mo., and John Chibnall of Saint Louis University reviewed personality test data from 288 would-be police officers in the Midwest. They found that applicants who scored low on measures of neuroticism and high on extraversion and conscientiousness later proved to be the most successful officers.

In a 2013 study, Peter Weiss, then at the University of Hartford, and his colleagues linked the performance reviews of 4,348 police officers to their scores on the Minnesota Multiphasic Personality Inventory, a test used by many departments to prescreen job candidates for psychological disorders. As one might expect, they found that those with high scores—and potentially more psychological issues—had more problems later on. High scorers were more likely to receive citizen complaints, to be terminated for cause and to engage in problematic weapon use, among other undesirable behaviors.

Higher education is also beneficial. Many departments require only a high school diploma, but a 2011 review in the Journal of Criminal Justice Education found that officers with college degrees were associated with fewer violent incidents. Similarly, a 2008 study showed that college-educated patrol officers were significantly less likely to be involved in shootings.—Cara Tabachnick

*Editor's Note: This sentence from the print article was edited before it was posted online to remove the names of Trayvon Martin, Eric Garner and Freddie Gray, who were killed, but not as a result of being shot by police, as stated in the next sentence.