What Influences Shooter Bias? The Effects of Suspect Race, Neighborhood, and Clothing on Decisions to Shoot.
Police shooting deaths of unarmed Blacks and African Americans led to psychological research on the influence of racial stereotypes on decisions to shoot, an effect called shooter bias. This article investigates how contextual cues signaling threat or safety interact with the race of the target to moderate shooter bias. Across two experimental studies using a first person shooter task, participants viewed Black or White male targets who held either a neutral (wallet or cellphone) or dangerous (gun) object. Study 1 manipulated the perceived safety or threat associated with the neighborhood context these shooting decisions occurred in, and Study 2 manipulated the perceived safety or threat associated with the targets’ clothing. Participants made quick decisions to “shoot” or “not shoot” the presented target, with error rates serving as the dependent variable. Across both studies, results confirmed that racial bias in shooting decisions against Blacks was present in perceived threatening neighborhoods and in perceived threatening clothing, and it was reduced in perceived safe neighborhoods and when wearing perceived safe clothing. Results help to identify contextual factors that may lead to mistaken shooting decisions, which can be used to improve police training and decision making to reduce bias.
The shooting of African American Trayvon Martin in 2012, resulting in his death, by neighborhood watch coordinator George Zimmerman relaunched societal discourse about the role of race in fatal shooting encounters. Martin's death was followed by a number of similar police shootings of unarmed racial minorities, including Michael Brown Jr. in Ferguson, Missouri, perpetuating racial discord and outrage. In 2014, the shooting of Tamir Rice, a 12‐year‐old African American boy, by Cleveland police officers exemplifies how race can potentially influence these encounters, as officers mistook a toy gun he was holding to be a real weapon. Demonstrating the extent of racial disparities in police shootings, young Black men were nine times more likely than other Americans to be killed by police officers in 2015 (Swaine, Laughland, Lartey, & McCarthy, [ 38] ), and disparities in police shootings persist even when controlling for racial differences in criminal activity (Scott, Ma, Sadler, & Correll, [ 34] ). Further, these racial biases are present when examining mistaken identity police‐on‐police shootings, with out of uniform Black police officers more likely to be perceived to be a civilian and killed than White officers (Charbonneau, Spencer, & Glaser, [ 4] ).
Examining this potentially fatal misperception from an experimental framework, social psychologists have directly investigated the causal role of racial stereotypes on decisions to shoot using shoot/do not shoot video simulations. Participants’ responses often reflect a stereotypically biased decision‐making pattern, termed shooter bias (Correll, Park, Judd, & Wittenbrink, [ 6] ). Because shooting decisions involve a complex myriad of situational factors that can influence decisions, additional research is needed to examine potential moderators of the shooter bias effect (Kahn & McMahon, [ 21] ). Drawn from discourse surrounding high profile shootings of racial minorities, the present experiments test what factors are associated with racially biased decisions to shoot racial minorities. Specifically, these studies examine how neighborhood context (Study 1) and clothing type (Study 2) interact with the race of suspects to influence decisions to shoot. In doing so, it provides evidence and implications for police training to reduce the likelihood of its occurrence.
How Suspect Race Impacts Decisions to Shoot
Science has been limited in its ability to understand the causal role of suspect race on police behavior, due to a lack of access to quality police data, the lack of ability to directly test police behavior using experimental designs, and the limited nature of nonpolice samples to draw causal conclusions about police behavior (Goff & Kahn, [ 17] ; Kahn & Martin, [ 20] ). One of the most direct examinations into the role of race on decisions to shoot is the study of shooter bias (Correll et al., [ 6] ). Applying knowledge of how stereotypes impact perceptions, participants complete a “shoot/do not shoot” videogame, which measures participants’ error rates in quick decision‐making situations. After varying backgrounds are seen, a White or Black male target appears holding either a neutral object (e.g., wallet, soda can) or a gun. The goal is to correctly “shoot” the suspects carrying guns and correctly “not shoot” the suspects who are carrying neutral objects. Results confirmed the phenomenon of shooter bias, as participants mistakenly “shot” unarmed Black targets more often than unarmed White targets, and erroneously “did not shoot” White targets with a gun more often than Black targets with a gun. This outcome represents a Black racial stereotype‐based response pattern, as participants’ errors reflect an overperception of Black targets associated with danger. Indicating the source of these errors, participants are more likely to adopt a lower decision criterion to shoot Black targets compared to White targets, which leads to an increased likelihood of mistakenly shooting unarmed Black targets (Correll et al., [ 6] ).
The knowledge of cultural stereotypes linking Blacks and criminality drives the shooter bias effect, and it is not related to affective or explicit prejudice (Correll et al., [ 6] ; Correll, Park, Judd, & Wittenbrink, [ 7] ). When individuals encounter a Black suspect, the Black racial stereotype is primed, which includes violence and criminality, influencing decision making to be in line with the stereotype. This results in the tendency to shoot Black suspects more than White suspects (Correll et al., [ 6] ; Correll, Park, Judd, & Wittenbrink, [ 7] ). Aspects that increase the saliency of these racial stereotypes, such as reading a newspaper article about crime committed by Black men before completing the shooter task, increase the amount of shooter bias (Correll , Park, Judd, & Wittenbrink, [ 7] ). Further, the more stereotypicality Black an individual phenotypically appears, the more likely shooter bias is to occur (Kahn & Davies, [ 19] ). This is because highly phenotypically stereotypical Blacks (e.g., darker skin tone, more racially representative facial features) are targeted by racial stereotypes more than low stereotypical Blacks (Blair, Judd, & Chapleau, [ 2] ; Eberhardt, Goff, Purdie, & Davies, [ 14] ). Similarly, Black men are targeted by shooter bias more compared to Black women, because Black men have stronger violence stereotypes associated with their group (Plant, Goplen, & Kunstman, [ 30] ). In another operationalization of shooter bias focusing on participants’ reaction times to make correct decisions (e.g., how fast participants are to correctly make the right decision), Black men with guns elicited faster response times, but Black men without guns elicited slower response times, compared to Latinos, Asians, and Whites (Sadler, Correll, Park, & Judd, [ 33] ). Reflecting changes in the prominence of societal group stereotypes, stereotypes linking Muslims to terrorism became culturally salient after 9/11 in the United States and shooter bias could be found against Muslims (Unkelbach, Forgas, & Denson, [ 40] ).
As most shooter bias research has been conducted with community members or undergraduate samples, translational research with police officers in a shooter bias paradigm has been limited (see Kahn & McMahon, [ 21] , for a review). Some of this research has shown that police officers make similar errors as other participant groups, such that they are more likely to shoot unarmed Blacks compared to unarmed Whites, although this bias can be reduced with extended practice (Peruche & Plant, 2006; Plant & Peruche, [31] ). Other investigations have demonstrated that officers are more likely to display racial bias in reactions times, such that they are faster to respond to Blacks with guns and slower to respond when Blacks do not have a gun, but are less likely to make errors in line with those responses (Correll , Park, Judd, Wittenbrink, Sadler, & Keesee, [ 7] ; Cox, Devine, Plant, & Schwarz, [ 10] ; Sim, Correll, & Sadler, [ 35] ). This pattern indicates that officers may still experience the same biases in perceptions, but may ultimately be better at overriding those biases. More in‐depth research is needed employing police officers to better understand this effect, and to connect it to the real‐world racial disparities in shooting deaths discussed earlier.
The lack of research directly involving police officers highlights the need for a Full‐Cycle Model (Cialdini, [ 5] ; Mortensen & Cialdini, [ 27] ) approach to the study of shooter bias and the role of race in police decision making (Kahn & McMahon, [ 21] ; Sadler et al., [33] ). The Full‐Cycle Model approach emphasizes the interplay between naturalistic observation of social phenomena, theory development, and lab experimentation to critically understand social phenomena. It is necessary to move cyclically between tightly controlled laboratory experiments and field observations to develop theory, design interventions, and improve outcomes. The study of shooter bias necessitates such an approach, as causal factors can only be isolated in laboratory experiments, both with and without police officers, which can then be combined with studies of police behavior in the field and noncausal observations (e.g., studies finding racial disparities in policing outcomes). Together, this approach can develop theory that can inform interventions to reduce shooter bias. The current studies provide a causal experimental laboratory approach as part of the larger Full‐Cycle Model paradigm.
More research is needed on the conditions under which shooter bias is more or less likely to occur, in order to better understand ways to reduce its occurrence in the real world (Kahn & McMahon, [ 21] ). This will help inform police training and ways to reduce shooter bias by more precisely understanding the potential risk factors for police to make mistaken shooting decisions. Individual difference variables have been shown to increase the likelihood that shooter bias occurs, including higher levels of in‐group identification (Kenworthy, Barden, Diamond, & del Carmen, [ 22] ), implicit associations between the self and prejudice (Glaser & Knowles, [ 16] ; Park, Glaser, & Knowles, [ 28] ), and higher levels of fatigue (Ma, Correll, Wittenbrink, Bar‐Anan, Sriram, & Nosek, [24] ). Here, we focus on relevant contextual variables in the shooting situation that may influence the expression of shooter bias. Specifically, the current studies provide a causal investigation into the role of perceived safe or threatening neighborhood context (Study 1) and suspect attire (Study 2) on decisions to shoot. Both variables are drawn from characteristics associated with high‐profile shootings of unarmed racial minorities, providing suggestive evidence of the causal factors that may have played a role. Across both studies, it is hypothesized that factors associated with perceived threat increase shooter bias, whereas factors associated with perceived safety will reduce the magnitude of shooter bias directed against Black, compared to White, suspects.
Study 1
Neighborhoods have varying associations with threat and safety, making environmental context an important variable when considering shooter bias effects. Initial investigations regarding the role of environmental context and suspect race have demonstrated its potential to influence policing outcomes. In one study, officers made more incorrect shooting decisions in neighborhoods associated with the opposite race from the officer and less errors in same race neighborhoods; presumably, this is due to differences in perceived threat (Cox et al., [ 10] ). Other investigations have found that dangerous backgrounds (dilapidated buildings, subway stations with graffiti), compared to neutral backgrounds, reduce levels of racial bias in shooting decisions by increasing error rates for all suspects; specifically, by increasing mistaken shooting decisions against unarmed Whites such that they are treated similarly to unarmed Blacks (Correll, Wittenbrink, Park, Judd, & Goyle, [ 9] ). Neighborhood socioeconomic status has also been shown to explain officer decision making regarding police use of force (Terrill & Reisig, [ 39] ). Further, officers who work in high‐crime urban neighborhoods with racial minority residents show a greater amount of shooter bias in their reaction times (i.e., how fast they make a correct decision; Correll , Park, Judd, Wittenbrink, Sadler, & Keesee, [ 7] ).
Study 1 adds to this developing literature by experimentally investigating whether real‐world neighborhoods (perceived to be safe vs. dangerous by the decision maker), at an implicit level, can influence shooting decisions against Black and White suspects. We directly examine specific neighborhood names, rather than manipulating contextual cues in the environment. Police may be subtly primed before encountering a suspect as they are driving through different neighborhoods or by hearing the neighborhood names stated over dispatch on calls. The current study tests the effects of a perceived “threatening” and “safe” context (to the perceiver) through subliminal neighborhood name primes (i.e., Beverly Hills vs. South Central). The South Central name should prime a perceived threatening context due to its association with violence, crime, lower SES, and the Black stereotype, whereas the Beverly Hills name should prime a perceived safe context as it is not associated with threat, counters the Black stereotype, is associated with Whites, higher SES, and lacks violence‐related associations.
We hypothesize that shooter bias will be more pronounced in perceived dangerous neighborhoods that are stereotype consistent with Black racial stereotypes, and reduced in Black stereotype inconsistent, perceived safe neighborhoods. Specifically, we expect that neighborhoods perceived to be threatening will prime danger consistent with the Black stereotype and increase threat, leading to more biased decision making against Black suspects. Neighborhoods perceived to be safe by the perceiver, which are inconsistent with Black stereotypes associated with danger, should instead prime safety and serve to reduce the application of racial stereotypes toward Black suspects. That is, when primed with perceived safety (e.g., Beverly Hills), participants should be more likely to view all suspects as being less threatening and reduce, but maybe not eliminate, the overall bias in decision making based on race. Because White suspects are already viewed as less dangerous than Black suspects (Devine & Elliot, [ 12] ; Eberhardt et al., [ 14] ), priming a safe context should most strongly affect decisions toward Black suspects. Therefore, when primed with perceived safe environmental contexts, we expect that Black targets will be treated more similar to White targets, resulting in reduced overall racial bias in shooting error rates.
Support for these hypotheses is drawn from literature on the expression of implicit bias and priming effects. Previous shooter bias studies have demonstrated that increasing the saliency of Black stereotypes exacerbates shooter bias (Correll, Park, Judd, & Wittenbrink, [ 7] ), which should occur in the threatening context. Further, in priming research, bias is often increased in negative contexts, and negative contexts can increase the impact of negative stereotypes on responses (Gawronski, Deutsch, & Seidel, [ 15] ; Wittenbrink, Judd, & Park, [ 42] ). However, priming a perceived safe context may reduce the saliency of the Black stereotype for Black targets, and they may be assimilated into the safe context and viewed as less threatening. It is possible that the bias will not be entirely eliminated, but reduced as Black suspects will be treated more similarly to White suspects than in the perceived threatening context. Indeed, Correll et al. ([ 9] ) suggest a similar hypothesis regarding the potential for safe environments to reduce racial bias, arguing that positive contextual cues may reduce racial bias against Black targets by promoting more positive responses toward them. This is the first shooter bias study to examine a perceived “safe” context and how it affects the expression of bias compared to a more threatening environment, as well as the first examination to use real‐world neighborhood names, increasing external validity.
Method
Participants and Design
Eighty undergraduate students (28 males and 52 females)[ 1] at the University of California, Los Angeles, participated in the experiment. The racial composition of the sample was 50 Asian/Asian Americans, 19 Caucasians, 8 Hispanics/Latinos, and 3 individuals who identified as “Other.” Three participants were excluded from the final analyses because they recognized part of the subliminal prime, although their inclusion does not affect the results. Participants were given partial course credit in exchange for participation. The experiment uses a within‐subjects 2 neighborhood prime (threatening vs. safe) × 2 object (gun vs. neutral) × 2 target race (Black vs. White) design.
Materials
The shooter bias program was run through PsyScope and used the original design by Correll et al. ([ 6] ). The shooter bias game consisted of a total of 115 trials. The first 15 trials were practice, while the following 100 trials broke down into 25 trials by condition: object (gun vs. no gun) and target race (Black vs. White). All targets were male. The neutral objects held by the suspects were a black cell phone, a black wallet, a silver camera, and a silver soda can. The Correll shooter bias game consisted of 25 neutral backgrounds, involving scenes such as parks, city sidewalks, and building entrances.
The current experiment altered the shooter bias game to include a subliminal prime that occurs prior to the target presentation. The neighborhood names chosen for this experiment were “Beverly Hills” to represent the safe neighborhood and “South Central” to represent the dangerous neighborhood, which were highly relevant neighborhoods for the Los Angeles, CA, sample. Pretesting confirmed that Beverly Hills is considered a safe neighborhood, while South Central is considered a dangerous neighborhood, and both were equally familiar to the sample population. Both names also consist of two words and are composed of 12 letters each. Two shooter bias sets were created: one with the South Central prime and one with the Beverly Hills prime. Order of prime game was counterbalanced. Participants completed both sets of stimuli, for a total of 230 trials, following the neighborhood prime (threatening vs. safe) × 2 object (gun vs. neutral) × 2 target race (Black vs. White) within‐subject design.
In the shooter game, the appearance of a target constituted a trial. Trials were presented in random order. Each trial began with a plus sign (+) appearing in the center of the screen to use as a fixation point. The program then flashed between one and four different backgrounds for between 500 ms and 1 s. After the last background, the neighborhood name prime was presented foveally for 15 ms (see Bargh & Chartrand, [ 1] , for review of priming techniques), and was approximately 3″ by 3″ in size. It was then replaced by the target, appearing for 630 ms. The 630 ms presentation of the target was the response time window in which participants must make a decision to “shoot” or “not shoot” the target, and is established as the response window for error rates as the dependent variable (Correll et al., [ 6] ). Not making a decision during the response window was recorded as a timeout, and participants then received a message that they were too slow in responding. The goal of the game is to shoot suspects who are carrying guns by pressing the letter Z on the keyboard, and not shoot suspects who are carrying neutral objects by pressing the letter M on the keyboard. Errors represent an incorrect shooting decision.
Procedure
Participants entered the laboratory and were seated in front of one of three computers by a White male research assistant. After agreeing to participate and giving consent, they received information about the study and how to complete the computer game. Each participant then completed either the Beverly Hills or South Central neighborhood prime version of the shooter bias game first; again, order was counterbalanced. Participants then took a 5‐minute break prior to completing the other prime version of the shooter bias game. Finally, the participants completed a questionnaire including demographics, questions about the subliminal prime, and a probe for suspicion. They were then debriefed and given credit for their participation.
Results
Following past shooter bias study procedures (e.g., Correll et al., [ 6] ; Kahn & Davies, [ 19] ), errors were defined as incorrectly deciding to “shoot” an unarmed suspect or “not shoot” an armed suspect, error rate scores were the number of errors divided by the number of valid trials, and timeouts (i.e., when the participant did not respond in the 630 ms window) were considered invalid trials.[ 2] To test the study hypotheses, a neighborhood prime (threatening vs. safe) × 2 object (gun vs. neutral) × 2 target race (Black vs. White) within‐subjects ANOVA revealed a significant main effect of target race, F( 1, 76) = 8.30, p <.005, partial η2 =.10, and a main effect of object, F( 1, 76) = 34.59, p <.001, partial η2 =.31. These two main effects, however, were qualified by the predicted object x target race interaction, F( 1, 76) = 14.90, p <.001, partial η2 =.16. This object × race interaction replicates the Correll et al. ([ 6] ) shooter bias findings; however, in the present research, all of the above findings were qualified by a three‐way interaction between neighborhood prime, object, and target race, F( 1, 76) = 4.18, p =.04, partial η2 =.05. This three‐way interaction indicates that the presence of the object × race interaction (i.e., the traditional shooter bias effect) depends on the neighborhood prime, see Figure [NaN] .
We broke down the three‐way interaction by neighborhood prime to examine the shooting decisions in each neighborhoodcontext. When primed with the threatening neighborhood (i.e., South Central), a main effect of object, F( 1, 76) = 30.49, p <.001, partial η2 =.29, is evident, but this main effect is qualified by an object × race interaction, F( 1, 76) = 18.21, p <.001, partial η2 =.19, which replicates the traditional shooter bias pattern. In contrast to the threatening neighborhood, when primed with a safe neighborhood (i.e., Beverly Hills), there is a main effect of object, F( 1, 76) = 23.82, p <.001, partial η2 =.24, and a main effect for target race, F( 1, 76) = 6.31, p =.01, partial η2 =.08, but the object × race interaction is no longer significant, F( 1, 76) = 2.0, p =.15. That is, the traditional shooter bias effect was attenuated by the safe neighborhood prime.
To facilitate error rates being directly compared, we separated the data based on object type (i.e., gun vs. neutral object), which ensures that errors compared have the same meaning and conducted paired sample t tests. In the gun condition, a paired sample t test conducted in the threatening neighborhood condition found that more errors were made with armed Whites (M = 0.118, SD = 0.08) than armed Blacks (M = 0.070, SD = 0.06), t(76) = ‒4.89, p <.001. This represents racial disparities in shooting rates (i.e., shooter bias) because White targets with a gun were mistakenly perceived as safe and thus subjected to more incorrect decisions than Black targets with a gun.
When comparing across neighborhood primes, participants who were primed with the safe neighborhood (M = 0.089, SD = 0.07), compared to the threatening neighborhood, made more errors with armed Black suspects, t(76) = ‒2.19, p =.03. That is, the safe neighborhood prime made the armed Black target appear to be safer than the threatening neighborhood prime, leading to more errors in the gun condition. For White suspects with a gun, however, the difference between neighborhood primes was not significant (safe neighborhood: M = 0.121, SD = 0.08), t(76) = ‒0.29, p =.78. The remaining difference between armed Blacks and armed Whites in the safe neighborhood was still significant, t(76) = ‒3.25, p =.002, but it was reduced in size. Therefore, White suspects remained unaffected by the primes, while the Black suspects were perceived to be “safer” when considered in a safe context, which reduced, but did not eliminate, overall racial bias in shooting error rates.
In contrast, in the neutral object condition, when primed with a perceived threatening neighborhood, unarmed Black suspects (M = 0.182, SD = 0.17) were marginally more likely to be shot than unarmed White suspects (M = 0.160, SD = 0.14), t(76) = 1.71, p =.09, representing shooter bias. However, when subliminally primed with a perceived safe neighborhood, this race effect disappeared (unarmed Black: M = 0.171, SD = 0.15, unarmed White: M = 0.180, SD = 0.14), t(76) = ‒0.64, p =.52. The safe neighborhood prime eliminated racial bias in the neutral object condition, and Black and White targets were treated similarly.
Discussion
Results from Study 1 illustrate that neighborhood context can influence racial bias in decisions to shoot. Shooter bias was more likely to occur in the perceived threatening neighborhood compared to the perceived safe neighborhood. Relative to South Central, subliminally priming Beverly Hills created a perceived safe environment to the decision maker that reduced, but did not eliminate, the racial discrepancy in shooter bias error rates. In the gun conditions, the bias was attenuated; while in the no‐gun condition, the bias was eliminated. Black targets were treated more positively after the safe prime, reducing overall levels of racial bias compared to treatment of Whites. In both neighborhood conditions, White suspect error rates were unaffected by priming, and their race was more influential on shooting behavior than contextual cues. This may be because shooter bias is driven by racial stereotypes of Blacks and African Americans that Whites are seen in contrast to, relatively protecting them from stereotype‐driven decisions to shoot.
Research has demonstrated that predominately Black physical spaces are stereotyped as unsafe, are devalued, and people are less likely to protect these spaces from harm (Bonam, Bergsieker, & Eberhardt, [ 3] ). The current data extend this notion such that Black individuals in these spaces are also less likely to be protected from harm—and more likely to be mistakenly shot—compared to White individuals. The interaction of individuals embedded within stereotyped physical spaces may further contribute to racial disparities in these locations across a variety of outcomes, including health, employment, and educational achievement.
The neighborhood primes may have also had their effect on racial bias in shooting decisions toward Blacks through cognitive subtyping processes. The safe neighborhood prime may have caused participants to cognitively subtype the Black targets as a positive subtype, and subsequently treat them safer. When individuals violate a common applicable stereotype, people maintain the stereotype by subtyping the deviant individual by finding additional attributes to which to attribute the behavior violations (Kunda & Oleson, [ 23] ). Behavior is often directed at the subtype, rather than the global stereotype level (Devine & Baker, [ 11] ). The safe environment may have served as a cue that the Black suspect is stereotype inconsistent, and should be responded to differently than the global Black stereotype. They may have been categorized, for example, as a “Black businessman,” a positive Black subtype, rather than a negative subtype, such as “Black criminal” (Devine & Baker, [ 11] ). Hence, the Black suspect is treated more like the White suspect, leading to a reduction in the racial bias in shooting error rates. Study 2 further tests this hypothesis by examining the role of clothing worn by the suspect.
Study 2
Study 2 extends this investigation into the moderators of shooter bias by examining another aspect of decisions to shoot: suspect clothing. After the shooting death of Trayvon Martin, the gray hooded sweatshirt that he was wearing when he was shot became a representation of racial bias in policing (Yankah, [ 43] ). Hoodies are perceived to be associated with criminality, danger, and “thugs,” and are viewed as threatening, particularly when worn by Black males (Minchillo, [ 26] ). In addition, baggy sweatshirts, bandanas, baseball caps, and baggy jeans are similarly associated with threat and a “criminal stereotype” (MacLin & Herrera, [ 25] ). Some individuals in the media blamed the hoodie—rather than Martin's race—for the reason that he was shot and killed, going so far to say that Martin was in part to blame for his shooting because he chose to wear a hoodie (Wemple, [ 41] ). To counter this perception, protests around the country saw individuals wear a hoodie in solidarity with Martin to support the fight against racial bias in policing (Joseph & Somaiya, [ 18] ).
Study 2 tests whether different types of clothing exacerbate or reduce racially disparate shooting decisions. Mimicking the results of Study 1, it is hypothesized that racial bias in shooting decisions against Blacks will be greater when they are wearing racially consistent perceived “threatening” attire, which will serve to prime danger in line with the Black stereotype and increase perceptions of threat. In contrast, we predict that racial bias against Blacks will be attenuated when they are wearing Black stereotype inconsistent and perceived safe attire (i.e., a business suit). The perceived safe attire condition will prime safety and potentially lead participants to subtype Black targets as safe and treat them more positively, reducing overall levels of racial bias in decisions to shoot.
Method
Participants and Design
Fifty‐six undergraduate students (20 males and 36 females) at the University of California, Los Angeles, participated in the experiment. The racial breakdown of the participants was 28 Asian/Asian Americans, 16 Caucasians, 9 Hispanics/Latinos, and 3 who chose to identify as “Other.” The experimental design was a 2 clothing (threatening vs. safe) × 2 object (gun vs. neutral) × 2 target race (Black vs. White) within‐subject design. Participants received partial course credit for participating in the experiment.
Materials
In order to manipulate the clothing of targets, a new shooter bias game was developed in DirectRT based on specifications from past shooter bias studies (e.g., Correll et al., [ 6] ; Kahn & Davies, [ 19] ). In the program, each trial begins with a blank screen of between 1–3 s, after which a photograph of the target appeared. As in Study 1, participants were instructed to “shoot” targets carrying guns and “not shoot” targets carrying neutral objects, which included equal number of wallets and cellphones. Participants had a 630 ms response window to make a decision or a timeout was indicated on the screen.
There were six targets in the game, which were three Black and three White males. The male targets were pretested and matched in terms of attractiveness, build, and stereotypicality for their race, and have been used in previous shooter bias studies (see Kahn & Davies, [ 19] ). Each target appeared in an individual photograph against a white background holding one of two neutral objects (one black cell phone and one black wallet) and one of two guns (both black guns). Each object was held in three different poses: object to the side, object near chest, and object straight ahead. To manipulate clothing type, each target appeared in either perceived threatening attire or perceived safe clothing. The threatening attire consisted of a baggy gray sweatshirt, a gray headband, and a black baseball cap, worn to the side (e.g., see MacLin & Herrera, [ 25] ). The safe attire consisted of a light button up shirt and a tie. Outfits were pretested and confirmed the stereotypical safe and threatening attire perceptions with the participant population.
In total, for each of six targets, there were 24 photographs (4 objects × 3 poses × 2 clothing types). Following the 2 clothing (threatening vs. safe) × 2 object (gun vs. neutral) × 2 target race (Black vs. White) design, there was 144 trials, equally distributed across conditions. Trials were randomly presented by the DirectRT program to participants, such that they did not know which target was coming next. Trials ended with participants pressing a designated key to “shoot” or “not shoot” the target, or when the 630 ms response window was surpassed and a timeout was marked.
Procedure
After arriving to the laboratory, participants were greeted by a research assistant and seated in front of a computer. After giving informed consent to participate, participants received instructions and began the shooter bias game on the computer screen, lasting approximately 10 minutes. After completing the game, participants answered a questionnaire, in which they entered their demographic information and general comments about the study. They were then thanked, debriefed, and given credit for participating.
Results
Replicating the analytical procedure in Study 1 and prior shooter bias studies (e.g., Correll et al., [ 6] ; Kahn & Davies, [ 19] ), errors were defined as incorrect shooting decisions, such that participants mistakenly decided to “shoot” a target without a gun, or “not shoot” a target with a gun. Error ratios were calculated as the number of errors divided by the number of valid trials (e.g., trials in which a decision was made during the 630 ms response window). A 2 clothing type (threatening vs. safe) × 2 object (gun vs. neutral) × 2 target race (Black vs. White) within‐subjects ANOVA was run on error ratios. As hypothesized, ANOVA results found a main effect of object, F( 1, 55) = 10.54, p =.002, partial η2 =.16, and a marginal race × object interaction, F( 1, 55) = 3.72, p =.059, partial η2 =.06. These were qualified by the predicted three‐way interaction between clothing, target race, and object, F( 1, 55) = 4.62, p =.04, partial η2 =.08, indicating that the race × object interaction (i.e., the shooter bias effect) depends on clothing, see Figure [NaN] .
To analyze the three‐way interaction, we broke down the data by clothing type. In the perceived threatening clothing condition, there was a main effect of object: F( 1, 55) = 5.37, p =.02, partial η2 =.09, and a significant race × object interaction, F( 1, 55) = 9.00, p =.004, partial η2 =.14. This interaction replicates the shooter bias effect. In the safe attire condition, the race x object interaction was no longer significant, F < 1, and only a main effect of object remained, F( 1, 55) = 11.46, p =.001, partial η2 =.17. This indicates that more errors were made in the gun condition than the no‐gun condition, as the safe attire led participants to mistakenly perceive safety. Importantly, this error pattern did not differ based on target race.
Paired sample t tests were run by object type to further break down the effect. In the gun condition, errors represent incorrectly “not shooting” a target with a gun, such that participants were mistakenly missing a real threat. Racial bias in shooting decisions was found in the perceived threatening clothing condition, but not in the perceived safe clothing condition. Specifically, fewer errors were made with armed Black targets in threatening attire (M = 0.154, SD = 0.13) than armed Whites in threatening attire (M = 0.188, SD = 0.16), t(55) = ‒2.06, p =.04. That is, participants were less likely to correctly perceive the armed White target in threatening attire as dangerous and correctly shoot him compared to armed Blacks in threatening attire. Further, participants made marginally fewer errors against the armed Black target when he was wearing threatening attire compared to safe attire (M = 0.189, SD = 0.18), t(55) = ‒1.68, p =.10. This indicates that the threatening attire made the Black target appear to be more dangerous than the safe attire, leading to fewer errors in the gun condition. In the safe attire condition, however, racial bias was eliminated, as there were no differences between armed Blacks and armed Whites (M = 0.202, SD = 0.18) in shooting decisions, t(55) = ‒0.81, p =.43. Finally, the clothing worn by the White targets who had guns produced no difference in participants’ shooting response, t(55) = ‒0.83, p =.41.
An analysis of simple effects in the neutral object condition shows a similar pattern. In the unarmed condition, errors represent mistakenly “shooting” an unarmed target, such that participants misperceived a threat when it was absent. Again consistent with the above patterns, racial bias was evident in shooting decisions in the perceived threatening clothing condition, but less so in the perceived safe clothing condition. Unarmed Black targets in threatening attire (M = 0.135, SD = 0.14) were marginally more likely to be mistakenly shot compared to unarmed White targets in threatening attire (M = 0.110, SD = 0.13), t(55) = 1.72, p =.09, such that they were perceived to be more dangerous than unarmed Whites in threatening attire. Marginally more errors were also made against unarmed Blacks in threatening attire compared to Blacks in safe attire (M = 0.112, SD = 0.13), t(55) = 1.67, p =.10. In safe attire, racial bias was reduced, as unarmed Black and unarmed White (M = 0.128, SD = 0.11) targets did not differ in error rates, t(55) = ‒1.15, p =.25. Clothing did not affect shooting decisions for Whites without a gun, t(55) = ‒0.94, p =.35.
Discussion
Clothing can influence the likelihood of shooter bias, representing a racially stereotyped response pattern in decisions to shoot. Racialized clothing perceived to be threatening, specifically a “hoodie,” played a prominent role in the societal discourse after the shooting death of African American Trayvon Martin, with some arguing that the stereotypically Black attire made it more likely that Martin was shot. Directly speaking to this issue, Study 2 finds evidence consistent with this social discourse. In a laboratory experimental study in which causal factors can be isolated, participants were more likely to display racial bias in their shooting decisions when targets wore stereotypically threatening attire associated with criminality than when they wore the perceived safe attire of a business suit. Specifically, participants tended to shoot unarmed Blacks in stereotypically threatening criminal attire more than similarly dressed Whites, and also more than when they wore safer business attire. The stereotypically dangerous clothing likely primed threat and the Black racial stereotype, which made it more accessible to influence shooting decisions. Oppositely, the perceived safer business clothing counteracted racial stereotypes for Blacks by priming safety, and they were treated more similarly to Whites, reducing, but not erasing, racial disparities. Again replicating Study 1, Whites, on the other hand, were less affected by clothing differences. Because they are less targeted by racial stereotypes regarding criminality and violence, which drives the shooter bias effect, Whites were relatively protected from its influence on shooting decisions.
This pattern of results is consistent with the findings from Study 1, despite a smaller sample size (although similar to other shooter bias study sample sizes), such that shooter bias was more pronounced in the stereotypically dangerous neighborhood than in the stereotypically safer neighborhood. In a similar process, the perceived threatening clothing may have primed danger stereotypes that interacted with Black racial stereotypes to increase perceptions of threat or danger, culminating in decisions to mistakenly shoot unarmed Blacks. The safe clothing condition may have facilitated the perception of the positive “Black businessman” subtype, leading participants to treat them differently than the global negative Black stereotype (Devine & Baker, [ 11] ). This subtyping could have led participants to perceive the Black targets as “safe” and more similar to the White targets.
General Discussion
Connecting to high profile shootings of unarmed minorities, this research identifies under what circumstances and conditions perceivers are more likely engage in racially biased decision making and when shooter bias is more likely to result. It experimentally tested factors present in prominent real‐world shooting deaths of racial minorities, and how those factors interact with target race to influence decisions to shoot. Study 1 experimentally tested the role of a perceived safe, compared to a perceived threatening, context on decisions to shoot using neighborhood names, and Study 2 examined clothing type on decisions to shoot. Across both studies, participants displayed more racial bias in their behavior when factors were present that increased perceived threat and were in line with the Black stereotype. This is consistent with past work indicating that it is the saliency of racial stereotypes that increase shooter bias, rather than explicit prejudice (Correll et al., [ 6] ; Correll, Park, Judd, & Wittenbrink, [ 7] ). Shooter bias was more pronounced in Black stereotype consistent threatening neighborhoods and threatening clothing, and reduced in Black stereotype inconsistent safe neighborhoods and clothing. Mirroring real‐world shooting incidents, this suggests that the hoodie in Trayvon Martin's death or the perceived dangerousness of predominately Black neighborhoods may play a role in differential shooting outcomes by race. Perceivers can be influenced by neighborhoods or clothing, which may exacerbate racial bias and contribute to disparate outcomes for Blacks and African Americans.
In order to lessen biased shooting decisions, efforts need to be made to reduce the association between context cues (clothing, location), race, and decisions to shoot among officers and members of the public. The primary emphasis must be on changing perceivers’ decision making. This may entail a long‐term process through training, education, and other intervention routes focused on decision makers and perceivers. In light of the current pervasive nature of these biases, racial minority group members may find themselves forced to adopt strategies to protect themselves from being the target of biases, which limits agency and can in itself impart harm to the individual. Consistent with research on targets’ reactions to stereotype threat (Steele, [ 36] ), stereotyped individuals may be forced to consciously adjust their behavior to signal the inapplicability of stereotypes to protect themselves. The famous anecdote about Black males being required to “whistle Vivaldi” late at night in order to reduce Whites’ perception of threat makes this point (Steele, [ 37] ). Similarly, African Americans have reported not wanting to wear a hoodie or other clothing because of the dangers associated with potential misperceptions. Feeling the need to counter these stereotypic perceptions, they are forced to “dress up” in business casual attire when going outside to remain safe (Yi, [ 44] ). That stigmatized individuals feel the pressing need to change their behavior in the short term to avoid potentially deadly consequences represents a troubling societal phenomenon. Again, long‐term change must target the biases in perpetrators themselves in order to produce equitable outcomes.
These biases are not only applicable to officers’ decisions to shoot but also can affect how the public views these shooting incidents. Societal discourse that blamed Trayvon Martin for his shooting because he wore a hoodie demonstrates the tendency to misguidedly blame victims, and particularly Black victims, for these shooting incidents. Indeed, when negative information is highlighted about shooting victims, which is commonly done by the media for Black victims, individuals are more likely to blame the victim and see less fault in the perpetrator (Dukes & Gaither, [ 13] ). The public may use racialized environmental contexts and clothing worn by victims to perpetuate this societal blame and further harm victims and racial minority communities.
The current research has implications for policing, although it is important to note that these data come from nonpolice participants. The real‐world environment and social contexts in which shooting incidents occur are not neutral, but rather racialized with varying degrees of perceived threat. Though a reduction, but not elimination, of racial bias in safe neighborhoods was observed, the threatening neighborhood still illustrates pronounced racial bias. While South Central may be a dangerous place in general, our findings suggest that it holds differential hazards for Blacks and Whites. The environment itself can create and propel life‐threatening consequences for the Black individuals in them. Conversely, in Beverly Hills, the safety associated with the city buffers these racial bias shooting effects. Although it reduces racial disparities in shooting rates, it does not mean that individuals cannot or will not be mistakenly shot in perceived safer contexts, but rather that racial disparities are somewhat less likely (e.g., Blacks and Whites will be treated more similar to each other). Further, it is important to note that the original shooter bias studies that find racial bias in decisions to shoot use neutral backgrounds and clothing types. In the current studies, we find shooter bias present in the perceived threatening neighborhood and clothing types. Thus, in neutral environments or devoid of any other contextual cues, racial stereotypes are likely to operate to produce shooter bias, as it also does in the threatening contexts. It is only in the safe context that those stereotypes are directly countered when overall racial bias may be reduced. Finally, while the results of the current studies are most directly related to decisions to shoot, similar biases may be evident when considering the entire police use of force continuum, from low levels of force to the highest fatal force. Therefore, their impact may be seen in more common police–suspect interactions that do not end in fatal force, expanding the potential influence.
Officers should be aware of the effect the surrounding context and social category cues have on their shooting behavior, and work to counteract their influence. Police training, for example, should focus on dynamic and varying types of environments and operationalizations of context. Police should be exposed to multiple types of targets, from varying races and attires, to most closely mimic real‐world environments. This will provide practice on accurately assessing threat and engaging in correct decision making. Training programs have been shown to help reduce shooter bias through repeated practice and focusing on trials in which race is nondiagnostic of shooting decisions (Plant & Peruche, [ 31] ; Plant, Peruche, & Butz, [ 32] ). These training simulations should also include various environments and other social cues that might impart threat. Departmental policies can also play a role in reducing mistaken shooting decisions and promoting equitable policing more broadly, such as by emphasizing that police use de‐escalation techniques. One recent policy initiative in response to police shootings of unarmed minorities has touted body worn cameras as a possible tool to reduce biased policing (see Kahn & Martin, [ 20] , for a discussion). Future research can examine how shooter bias is influenced by the presence or absence of a body worn camera.
More research is needed to examine other types of contextual effects and their influence on shooting behavior. The current study hypothesizes that it is the neighborhoods’ association with relative safety that is driving the effects. Both neighborhoods also have strong racial ties, in addition to socioeconomic status ties, which could also be imparting influence. The racial makeup of each area, being predominately more White or Black/African American, also plays into the perceived safety of neighborhoods. Although these factors are often correlated in real life, future studies can attempt to disentangle these effects. In addition, it is important to understand how these contextual variables interact with each other—that is, how do clothing, neighborhood race, and target race interact together to influence shooting patterns? It may be that racial minorities wearing perceived threatening clothing in perceived safe contexts are more likely to experience bias, creating a contrast effect. Further, when considering neighborhood context, how might perceptions of individuals being a resident in a neighborhood compared to a nonresident visitor passing through influence decisions to shoot? One may be reminded of the Dr. Henry Louis Gates incident, where neighbors incorrectly perceived him to not be a resident and called police, and Trayvon Martin's shooting occurring in a neighborhood that he was visiting. Bias may increase, even in perceived safe neighborhoods if residents perceive Black individuals to be outsiders and nonresidents of the neighborhood.
Limitations should be noted with these studies. Notably, our sample used college student participants and not police officers. However, the shooting of Trayvon Martin by citizen George Zimmerman demonstrates that nonpolice citizens can and do sometimes act on these perceived biases. Again highlighting the Full‐Cycle Model, these laboratory results should now be translated back into the field with police officers and used to help explain real‐world disparities in shooting outcomes (e.g., see Scott et al., [ 34] ). The psychological processes may be similar, or hold some important differences. Therefore, it is important before generalizing these results to police to test this highly relevant population directly. A second limitation is that our sample only included male targets. Black females are viewed as less violent compared to Black males, which reduces shooter bias (Plant et al., [ 30] ). However, the shooting deaths of Rekia Boyd in 2012 and other females provide evidence that these shootings can and do affect Black females. And, the infamous 1979 case of Eulia May Love, the 39‐year‐old Black widow who was shot eight times in her front yard in South Central by Los Angeles Police in a dispute over an overdue gas bill, which resulted in widespread civil unrest, shows that this is not a new phenomenon. It is important to study how these contextual variables then interact with race and gender, and influence treatment of Black women in particular. Similarly, extensions beyond Black and White racial groups to other races are also worthwhile (e.g., see Sadler et al., [ 33] ). Larger sample sizes, particularly in Study 2, can also help to replicate the effect, and be used in a meta‐analysis of shooting decisions across different safety contexts. Finally, further studies are needed to test if cognitive subtyping is driving the reduction in racial bias error rates.
A continued Full‐Cycle Model approach, in which tightly controlled experiments are combined with field observations and intervention work, is necessary to best understand the effects of race on police decisions to shoot. Based on this continued collection of evidence, individualized police training and higher level public policy responses can be developed to mollify any biased effects.
icons on the left to see existing comments.
0 General Document comments
0 Sentence and Paragraph comments
0 Image and Video comments
New Conversation
New Conversation
New Conversation
New Conversation
New Conversation
New Conversation
New Conversation
New Conversation
New Conversation
New Conversation
New Conversation
New Conversation
New Conversation
New Conversation
General Document Comments 0
