Implicit Social Cognition and Law

Kristin A. Lane, 1,2 Jerry Kang,3 and Mahzarin R. Banaji2

1Department ofPsychology, Bard College, Annandale-on-Hudson, New York 12504; email: [email protected]

2Department ofPsychology, Harvard University, Cambridge, Massachusetts 02138; email: [email protected] JSchool ofLaw, University ofCalifornia, Los Angeles, California 90095; email: [email protected] [TJo be as intelligent as we can is a moral when the data reveal unappealing reflections obligation-that intelligence is one of the ofhuman behavior, including our own. talents for the use ofwhich we shall be called to account-that if we haven't exhausted ev­ ery opportunity to know whether what we THE SCIENCE are doing is right, it will be no excuse for us Imagine sitting at a computer. Your job ap­ to say that we meant well. pears simple: As words such as happy and an­ gry appear sequentially on the screen, indicate John Erskine, American Character whether each is good (happy is good) or bad and Other Essays, 1915 (angry is bad) by pressing marked keys on a In the early years of the twentieth century, keyboard. But more than the words appear on John Erskine, American educator and author, the screen. In fact, each word to be judged as worried about a lack of faith in intelligence good or bad is preceded by a black or white as a virtue, and contrasted it specifically with face (i.e., individuals with origins in Africa or the elevated status accorded to goodness as Europe) that you see but do not respond to. a virtue. To be sure, Erskine was not us­ You merely ignore the face and respond to ing the term intelligence to mean a narrowly the words. First presented in a psychology lab specified mental faculty of the sort that nine­ over a dozen years ago, this task represents teenth, and even twentieth, century psychol­ the basic method of sequential or repetition ogists called intelligence. Rather he was refer­ priming, designed to measure indirectly less ring to a broad set ofcompetencies, skills, and conscious racial attitudes (Fazio et al. 1995). knowledge. The computer records the time taken to The research reviewed here is offered in offer the easy answer that happy represents a an Erskinian spirit because, more now than positive or good concept and that angry rep­ ever, the mind sciences suggest unappealing resents a negative or bad concept. To the psy­ truths about the nature ofthe brain and mind chologists who performed this study, the data that originate from its bounded rationality of interest were the speed to respond (with and largely unconscious operation. Despite some attention to accuracy of responses) to the commonly held belief that the opposite eachword. Datawere sorted into four separate is true (i.e., that humans are savage rational­ types: Trials in which good words like happy ists and that consciousness is the default men­ were preceded by (a) a white face, (b) a black tal state), we have incontrovertible evidence face; and trials in which bad words like angry that thoughts, feelings, and actions are shaped were preceded by (c) a white face, (d) a black by factors residing largely outside conscious face. From the many studies that have used awareness, control, and intention (see Carney this procedure, we know that speed to judge & Banaji 2007). that happy is good is noticeably faster when Such evidence and their implications for that word is preceded by the mere flash of a human nature and human experience urge white (rather than a black) face. Likewise, it that we be newly intelligent about vari­ is mentally easier to respond that angry is bad ous matters of law. Experiments from social when it is preceded by the briefpresence of a cognition-afield concerned with the content black rather than white face. This differential and mechanisms of beliefs and preferences ease ofpairing white+good and black+bad is about oneself, other social beings, and social taken as an indirect measure of the strength groups-are this review's mainstay, with a fo­ of automatic relative preference for the two cus on ordinary beliefs and preferences that social groups. operate without conscious intention, aware­ If the interest is in knowing a person's ness, or control. We present the evidence racial attitude, why use such an odd mea­ first, urging readers to heed Erskine's message sure? Why not merely ask for reports of

428 Lane. Kong. Bona}; feelings toward black and white Americans? ofpatients (Green et al. 2007); and (d) despite The simple answer is that decades ofresearch their seemingly uncontrollable nature, are on the nature of perception, attention, mem­ malleable. ory, and decision making demonstrate that in­ Two concepts, attitude (or preference) and direct measures that bypass the mind's access stereotype (or belief), are central to the study to conscious cognition tell us something in­ of implicit social cognition and its applica­ teresting about mental states and the behav­ tion to the law. "Implicit attitudes are intro­ iors theyspawn. Specifically, thevirtue ofsuch spectively unidentified (or inaccurately iden­ methods is that they tell us something differ­ tified) traces of past experience that mediate ent from self-reported survey-type responses. favorable or unfavorable feeling, thought, or Moreover, they may potentiallypredictmean­ action toward social objects" (Greenwald & ingful behaviors ofthe sort that are central to Banaji 1995, p. 8). For example, an object any system oflaw, e.g., behaviors that help and is likely to be preferred as a result of prior harm. For instance, the strength ofblack+bad exposure to it, even without awareness that and white+good associations in white sub­ such exposure causes the increase in prefer­ jects predicted the quality oftheir interaction ence (Zajonc 1980). Likewise, "Implicit stereo­ with black experimenters (Fazio et al. 1995). types are the introspectively unidentified (or Participants with stronger antiblack bias on inaccurately identified) traces of past experi­ the computerized test were less likely to be ence that mediate attributions of qualities to friendly toward the black experimenter than members ofa social category" (Greenwald & those with more positive scores. Moreover, Banaji 1995, p. 15). An implicit stereotype such participants were more likely to report would be the (mistaken) identification that thatblacks, compared withwhites, had greater Dave Sebastian is famous, but Diane Sebas­ responsibility for the civil unrest, riots, and vi­ tian is not (Banaji & Greenwald 1995), based olence in Los Angeles following the 1992 ac­ on the belief(a correct belief, in this case) that quittal ofpolice officers in the case involving men are more likelyto be famous thanwomen. Rodney King. Such preferences (attitudes) as well as the Interestingly, the same participants' self­ ascription of specific qualities (stereotypes) reported attitudes toward the groups on a tra­ are jointly referred to here as implicit so­ ditional survey were uncorrelated with their cial cognitions (ISCs), and the biases observed friendliness; in other words, these tradition­ in studying them are referred to as implicit ally measured expressions ofattitude were not biases. as predictive ofbehavior. This experiment set As an example of how the science of im­ the stage for the subsequent explosion ofwork plicit social cognition provides an alternate on implicit social cognition-the main results lens for viewing human behavior, consider two ofFazio and colleagues (1995) would be repli­ police shootings inthe outer boroughs ofNew cated many times with differing measures of York City. Both Amadou Diallo (in 1999) and implicit attitudes and stereotypes. This body Michael Bell (in 2006) were unarmed black of research captured the essence of a new men, mistakenly shot and killed by police of­ generation of discoveries about automatic, ficers. Of central relevance is the probabil­ nonconscious, or implicit preferences and ity of the occurrence of such a response had beliefs, primarily that they (a) are both per­ the victims been white rather than black. In vasive (large numbers ofindividuals show ev­ the earlier case, officers reported mistaking idence of them) and large, statistically speak­ Diallo's reach into his pocket as an attempt to ing; (b) diverge from the consciously reported get a gun. During the opening statements in preferences and beliefs of the same individ­ the criminal trial ofthe four police officers­ ual; (c) appear to predict behavior, even con­ who were all acquitted-the prosecution ar­ sequential behavior such as doctors' treatment gued that "when they got out of the car in

www.annualreviews.org • Implicit Social Cognition and Law 429 front of Amadou Diallo's home in the early decisions and does so without conscious morning of February 4, they made the con­ awareness. scious decision to shoot him." Among the criticisms ofsuch studies is that Dataand theoryfrom implicitsocial cogni­ they reflect a reality far from that ofthe pro­ tion provide a different interpretation ofthese fessionals who rely on their gut to make the incidents, namely that the officers acted with­ right decisions in their real jobs-college stu­ out conscious racial animus but were still in­ dents, after all, don't spend their hours in tar­ fluenced by the victims' race. No satisfactory get practice. To test a different group ofpro­ understanding of the role of implicit bias in fessionals, those who take an oath to serve the the police officers' behavior can be obtained sick whether rich or poor, we studied the be­ by analyzing single lethal events, for we can­ havior of physicians making assessments of notknowwhether the same officers in an iden­ patients. Green and colleagues (2007) mea­ tical universe but for the victim's race would sured implicit bias among emergency room have acted differently. The laboratory offers physicians. Physicians recommended treat­ a way to replicate the essential conditions of ment based on vignettes that depicted pa­ such situations and to examine patterns ofbe­ tients with myocardial infarction who differed havior that arise in systematic ways. Several solely on race. Although doctors' reports of laboratories created controlled environments racial attitudes and beliefs did notpredicttheir analogous to the police officers' situation. In proposed treatment, their ISCs did: Physi­ one such study, black and white men appeared cians with stronger implicitantiblack attitudes one at a time on a computer screen, holding and stereotypes were less likely to prescribe either a gun or a harmless object (e.g., a soda thrombolysis (a blood-thinning procedure) can) (Correll et al. 2002). If the target held for African Americans compared with white a gun, participants were instructed to press Americans with identical medical profiles. Ex­ one key to shoot; if it was a harmless object, plicit race attitudes, held with complete sin­ they were told to press another key for don't cerity and showing no race bias, did not pre­ shoot. dict medical recommendations. The data revealed systematic racial bias in Consciously held attitudes and stereotypes shooting, with faster and more accurate re­ are also important predictors of behavior. sponses to unarmed white targets and armed They are simply not the only ones to contend black targets compared with armed white tar­ with as we understand human behavior and its gets and unarmed black targets. Neither par­ vicissitudes. We can study ISCs with physio­ ticipants' endorsement of racial stereotypes logical techniques that measure cardiovascu­ nor their reports of feelings toward blacks lar responses (e.g., Blascovich et al. 2001) or predicted shooter bias. Knowledge ofcultural micro facial movements (e.g., Vanman et aI. stereotypes, however, did predictshooterbias: 2004), neuroimaging techniques that measure Those with greater awareness of the stereo­ brain activation (e.g., Mitchell et al. 2006, type of African Americans as hostile were Phelps et al. 2000), and those based on re­ more likely to mistakenly shoot unarmed sponse time, such as the priming task de­ black suspects and not shoot armed white scribed earlier (Perdue etal. 1990), to measure suspects (see Greenwald et al. 2003b, Payne automatic attitudes and beliefs. This last class 2001, Plant & Peruche 2005 for similar re­ has thus far been the most reliable and widely sults). In light of these data, the New York used for measuring individual differences in City shootings can be explained without re­ ISCs and, consequently, represents the bulk liance on explicit bias. Simple exposure to the ofthe work reviewed here. stereotype that blacks are hostile, even with­ The logic underlying such measures is out endorsement of that stereotype, may be relatively straightforward. First, a very sim­ sufficient to create bias thatalters split-second ple definition of an attitude is assumed:

430 Lane· Kong • Bana}i "Essentially, then, an attitude can be viewed mind, but "DOOR" if it is not (e.g., Sinclair as an association between a given object and a & Kunda, 1999)]. Additionally, the extent to given evaluative category" (Fazio et al. 1982, which behaviors completed by black actors p. 341). Second, response speed to varying are described in abstract terms (von Rippel stimuli measures the association's strength. In et al. 1997) or to which stereotype­ the priming task described earlier, those with inconsistent behaviors are eXplained rather antiblack feelings were presumed to have a than just described (Sekaquaptewa et al. relatively stronger association between the so­ 2003) is used as a measure of implicit cial category black and the evaluative category stereotypes. bad (compared with the categories white and This report draws four key conclusions good). The difference in this response time about the current state ofknowledge ofISCs: is presumed to reflect a person's attitude to­ (a) Measures that assess ISCs provide distinct ward blacks (see Wittenbrink 2007 for further assessments from self-reported orexplicitver­ information on the priming task). sions; (b) preferences for ingroups as well as Other measures involve similar logic but sociallyvalued groups are widespread; (c) ISCs differ in implementation. The Implicit As­ relate systematically to behavior; and (d) ISCs sociation Test (!AT; Greenwald et al. 1998), are flexible and respond to experience and for example, requires participants to rapidly environmental cues. The practical and ethi­ classify individual stimuli into one of four cal implications ofsuch findings have already distinct categories using only two responses. been explored in preliminary ways in various As with priming, the assumption is that re­ domains [Banaji et al. 2003, Banaji & Bhaskar sponses will be facilitated-and thus faster 2000, Bazerman et al. 2005; see also the 2006 and more accurate-when categories that are California L{f/}) Review symposium on behav­ closely associated are paired than when they ioral realism (Bayern 2006)]. In this review, are not paired (see Lane et al. 2007, Nosek we focus on how these scientific findings have et al. 2006 for further information on the influenced legal scholarship and judicial opin­ lAT). For example, a person with a negative ions thus far. implicit attitude toward blacks would be ex­ pected to go more quickly when black and bad share one key and white and good the Implicit Social Cognition is Distinct other than when the pairings of good and from Explicit Social Cognition bad are switched (readers may try an !AT ISCs often reveal different levels of inten­ at http://www.projectimpliclt.com). These sity and patterns of attitudes and stereotypes measures-priming and the !AT-are the than those provided by explicit, self-report most commonly used tasks to measure ISCs measures. Evidence suggests that implicit and are emphasized in our review, although and explicit measures provide distinct, al­ newer tasks, such as the GolNo-Go Associ­ though sometimes related, assessments of ation Task (Nosek & Banaji 2001), the Eval­ biases. uative Movement Assessment (Brendl et al. For example, although people tend to re­ 2005), the extrinsic affective Simon task (De port only slight preference for white Ameri­ Houwer 2003), and the affect misattribution cans over black Americans, implicit measures procedure (Payne etal. 2005), have been used, show a quite different tendency: On several and likely more will be developed. implicit measures, strong and consistent pref­ Other tasks measure ISCs withoutmeasur­ erences emerge for white Americans relative ing response speed or physiological response. to black Americans (e.g., Nosek et al. 2002, Completion of word fragments may re­ 2007). Such data, in conjunction with small veal mentally active stereotypes [e.g., _OOR correlations between explicit and implicit mightbe "POOR"whensocial class is onone's measures of the same attitude or stereotype,

www.annualreviews.org • Implicit Social Cognition and Law 43 r support the idea that the two systems of im­ (Cunningham et al. 2001). This pattern is plicit and explicit social cognition exist as more general: Implicit and explicit attitudes separate mental spheres with communication were best represented by two factors for 56 channels that are present but don't always out of 57 different attitude objects (Nosek work. 2005). Relationships between implicit and ex­ Evidence from the neural basis of social plicitmeasures vary across study, target group, cognition converges with that from behavioral and participant characteristics (Bosson et al. tests. The focus has been on subcortical and 2000, delong et al. 2003, Egloff & Schmukle cortical structures, with the loose assumption 2002, Greenwald et al. 1998, Karpinski & that implicit evaluations that are less control­ Hilton 2001, Ottaway et al. 2001, Rudman & lable should map onto activations in the for­ Kilianski 2000). A meta-analysis-a compre­ mer and processes that rely on deliberative hensive quantitative analysis of experiments thought should engage the latter. The amyg­ on a particular topic allowing more general dala is a subcortical brain structure known to conclusions than any single study-of 126 be reliably engaged in processing emotional correlations between implicit (assessed with (especially fear-relevant) and novel stimuli the IAT) and explicit attitudes revealed con­ (phelps 2006). When faces were presented so siderable variabilityinthe strength ofthe rela­ rapidly as to be under the limen or sublimi­ tionship between implicit and explicit cogni­ nal threshold (30 milliseconds, or 3/100th sec­ tions, although on average they were related ond), black faces elicited greater amygdala ac­ to one another, mean population r = 0.24 tivation than did white faces. When faces were (Hofmann et al. 2005). exposed long enough to be visible to partici­ Statistical analyses of responses to both pants (525 milliseconds) no such difference in measures further support the idea thatimplicit amygdala activation was observed, and instead and explicit measures tap separate processes. greater activation was seen in regions assumed A construct represents an abstract concept to be associated with regulating thoughts and that, although measurable, cannot be directly exerting control (the right ventrolateral pre­ observed, such as self-esteem, IQ, or in this frontal cortex, right dorsolateral prefrontal case implicit and explicit social cognitions. cortex, and anterior cingulate). Moreover, Do implicit and explicit measures provide amygdala activation correlated with the lAT assessments ofa single ormultiple constructs? measure ofracial attitudes when the faces were A statistical procedure, confirmatory factor presented subliminally, but not so when pre­ analysis (CFA), tests whether measures tap the sented supraliminally, indicating that the lAT same underlying construct (in which case they reflects more automatic rather than controlled would, in CFA's language, load onto a sin­ reactions to social groups (Cunningham et al. gle factor) or multiple constructs (in which 2004a). case they would load onto multiple factors). Most experts (these authors included) do Consistent with the notion that implicit and not believe that measures of implicit social explicit attitudes are distinct (Greenwald & cognition reflect the "true" attitude any more Banaji 1995, Wilson et al. 2000), implicit and than do measures of explicit social cogni­ explicit attitudes are best fit as separate fac­ tion such as questionnaire responses (Fazio tors, even when they are correlated. Forexam­ & Olson 2003, Lane et al. 2007). Implicit ple, implicit self-esteem and gender identity and explicit measures appear to tap separate each loaded onto a separate factor than their constructs thatoperate differently: Theyboth explicit counterparts (Greenwald & Farnham predict behavior (which one predicts better 2000). Implicit racial attitudes (measured by appears to depend on the person and situa­ the IAT and priming) also represented a sep­ tion). Privileging one over the other would be arate construct from explicit racial attitudes scientifically misguided.

432 Lane. Ivmg. Banaji Implicit Social Cognitions black, light-skinned people over dark-skinned are Robust and Pervasive people, other people over Arab Muslim, abled One ofthe most consistent findings from the people over disabled people, thin people over large literature on intergroup relations is the obese people, and straight people over gay fact of ingroup favoritism (Tajfel & Turner people). Implicit stereotypes were also con­ 1986). This tendency is so strong that peo­ sistent with widespread cultural beliefs. For ple report liking ingroups even when they example, most participants (72 %) associated are randomly assigned to them (Cadinu & the concepts male with science and female Rothbart 1996, Gaertner et al. 1989, Tajfel with humanities. Similarly, participants found et al. 1971). This pattern also emerges on im­ it easier to categorize white, rather than Asian plicit measures: After random assignment to or Native American, faces with American, re­ the group "Quan" or "Xanthie," participants flecting an implicit stereotype that "Ameri­ demonstrated implicit preference for their as­ can equals white" (Devos & Banaji 2005; T. signed group (Ashburn-Nardo et al. 2001). It Devos, B.A. Nosek & M.R. Banaji, unpub­ is mirrored on measures of implicit prefer­ lished manuscript). Additionally, participants ence for known social groups. For example, exhibited stereotypes reflecting associations both Japanese Americans and Korean Amer­ between blacks and weapons (compared with icans preferred their own ethnic group rela­ whites and harmless objects), and between the tive to the other (Greenwald et al. 1998), as social group male and career (compared with did East and West Germans (Kiihnen et al. female and family). Explicit reports ofingroup 2001). preference and stereotypes were smaller in Positive implicit attitudes toward and magnitude in white Americans than those ob­ stereotypes about members of socially served on implicit measures. privileged groups are also pervasive. At Second, the standard deviations shown in the website mentioned above (http://www. Table 1 reveal that despite the prevalence of ~mplicit projectimplicit.com), visitors try one (or bias, there was substantial variability more) IATs and receive feedback about the m the extent to which people showed such magnitude and direction of their implicit bias-some people were much higher than attitude or stereotype. With over 5 million the average (reflecting strong bias), but others tests completed, this is the largest repository were lower than the average (reflecting min­ of data available to look at variability and imal bias or even biases in the opposite di­ frequency of ISCs (see Nosek et al. 2007 for rection than the majority of test takers). For a review). Ofcourse, visits to the website are example, although 68% of test takers implic­ optional, and thus the data do not reflect a itly preferred white compared with black, a representative sample ofthe population. Even nontrivial group-14%-showed the oppo­ with this caveat, the data provide a unique site, and preferred black over white, and 18% look at ISCs and allow comparisons across showed no preference. Such variability sug­ self-reported demographic characteristics gests that additional factors may moderate an (e.g., age, gender, race/ethnicity, region, individual's level ofimplicit bias. Group mem­ political orientation). bership attenuated implicit bias. Although Table 1 depicts results from 17 IATs avail­ people ofEuropean, Asian, and Hispanic de­ able at the website (Nosek et al. 2007). Two scent implicitly preferred white over black, features ofthe data are readily apparent. First, black participants did not, on average, show implicit preferences in either direction, away ingroup preference; equal numbers of black from the neutral position ofno bias, were typi­ participants preferred the outgroup white as cal. Onaverage, participants preferred socially preferred the ingroup black. The pattern of privileged groups (young over old, white over stronger ingroup preference among mem­ bers ofsocially privileged groups is predicted

www.annualrroiews.org • Implicit Social Cognition and Law 433 Table 1 Attitudes and stereotypes for 17 topics among visitors to a public website: overall sample (implicit and explicit attitudes and stereotypes), by gender (nnplicit), and by ethnicity (implicit)'

Attitude 1.25 1.17 1.25 Age Young+GoodlOld+Bad .49 .39 1.23 .39 .78 .51 1.17 1.'10 1.22 1.22 lAO 1.23 1.00 .56 .70 Race White+GoodIBlack+Bad .37 .43 .86 .26 .73 .36 .79 .93 .79 .88 -.05 .79 .'19 .66 Skin-tone Light Skin+GoodIDark .30 Al .73 .17 .67 .25 .71 .80 .59 .76 .22 .68 .88 Skin+Bad .'10 .64 Child-race White+GoodIBlack+Bad .33 .45 .73 .19 1.30 .15 .69 .80 .53 .76 -.16 .67 .87 .00 Atab- Other Peoples+GoodiAtab .14 .42 .33 .45 .77 .58 .24 .'18 .29 .17 .19 .36 .40 .19 Muslim Muslims+Bad -.39 -.30 -.36 Religion Other Religions + -.15 .44 -.34 .14 1.05 .13 -.36 -.34 -.20 -.14 -.18 -.20 GoodJ]ewish+Bad 1.07 .98 .93 Disability Abled+GoodlDisabled+Bad .45 .43 1.05 .38 .67 .57 .98 1.28 .98 1.09 1.14 1.02 .74 .66 .70 Sexuality Straight People+GoodiGay .35 .'17 .74 .49 .91 .54 .68 .94 .83 .91 1.11 .87 People+Bad .88 .76 .79 Weight Thin+GoodiObese+Bad .35 .42 .83 .64 .73 .88 .81 .93 .81 .76 .74 .81 -.16 -.27 -.24 Presidential Bush+GoodiOther -.07 .45 -.15 -.94 1.28 -.73 -.20 -.13 -.18 -.24 -.33 -.20 President+Bade -.27 -.37 -.39 Election George Bush+Goodi -.14 .51 -.27 -.69 1.64 -.42 -.37 -.20 -.25 -.37 -.61 -.31 2004 John Kerry+Bad -.11 -.25 -.25 Election George Bush+Goodi -.09 .56 -.16 -.32 1.60 -.20 -.20 -.07 -.11 -.32 -.70 -.23 2000 AI Gore+Bad Stereotypes .86 .89 Race- White+Harmless .37 .37 1.00 .34 1.10 .31 .97 1.03 .84 .92 .59 .92 1.05 Weapons ObjeetslBlack+Weapons .16 .30 Arnerican- White Arnerican/+Arnerican/ .23 .50 .46 -.76 1.79 -.42 .46 .'18 -.24 .48 .44 .46 .56 Native Native ArnericanIForeign .51 .61 Arnerican- White+Arnerican/Asian+Foreign .26 .41 .62 .57 1.27 .45 .61 .68 .59 .29 .83 .68 .85 Asian .88 .93 Gender- Male+ScienceIFemale+Liberal .37 .40 .93 .52 .66 .79 .98 .93 .90 .88 .85 .85 .98 Science Arts Gender- Male+CareerlFemale+Family .39 .36 1.10 .54 .60 .89 1.19 .94 1.06 .97 1.19 1.06 1.14 1.06 1.03 Career

'Adapted from Tables 1, 2, 3, and 4 ofNosek et al. (2007). bIAT scores are mean D scores (Greenwald et a1. 2003a). d = Cohen's d effect size, which is a standardized measure of the magnitude ofan effect; the following guidelines indicate the size of the effect: small d = 0.2, medium d = 0.5, I"'ge d = 0.8 (Cohen, 1988). cThe comparison category for George Bush varied (Ointan, Reagan, Nixon, Kennedy, FOR, Lincoln, Jefferson, or a collection ofrecent presidents). by system justification theory (Jost & Banaji A claim has been made thatISCs, in partic­ 1994), which suggests that the tendency to ular preferences for social groups, reflect not maintain the status quo will lead to reduced a person's own attitude but rather knowledge ingroup preference among members of dis­ that the person has acquired about the atti­ advantaged groups (see Jost et a1. 2004 for a tude present in the larger culture (Karpinski review ofsystem-justifying tendencies on im­ & Hilton 2001). Of many possible responses plicit and explicit measures). to this claim, we offer three observations Although the tendencytoward ingroup lik­ that suggest that ISCs do reflect the state ing appears strong, lab studies also demon­ of an individual's mind. First, implicit pref­ strate that a group's status moderates the erences emerge even without cultural knowl­ magnitude of implicit ingroup preference. edge about the group: Simply being told you Predominant cultural evaluations attenuate are a Quan is sufficient to generate implicit this tendency toward ingroup preference preference for Quans (Ashburn-Nardo et a1. (see Jost et a1. 2004 for a review). As in 2001). If implicit biases solely reflect knowl­ the web data summarized in Table 1, al­ edge of cultural hierarchies and stereotypes though most white Americans showed strong that exist separately from individual social preference for white over black on the cognition, preferences favoring these nomi­ lAT, black Americans, on average, did not nal groups should not appear. prefer either group (Ashburn-Nardo et a1. Second, as reviewed more extensively be­ 2003, Livingston 2002, Nosek et a1. 2007). low, the magnitude of implicit bias predicts Similarly, weight and socioeconomic status behavior (poeWman et a1. 2007); if implicit (Rudman et a1. 2002), as well as university biases reflect disjointed cultural knowledge, (Jost et a1. 2002) and college dorm status they should not be linked to a specific individ­ (Lane et a1. 2005) moderated the strength of ual's discriminatory behavior. Finally, we note ingroup liking, such that people from higher­ thatitis often explicit, rather than implicit, at­ status groups were more likely to exhibit titudes that systematically relate to reports of implicit ingroup preference. cultural knowledge (Nosek & Hansen 2007). In other words, the elderly, the poor, and The culture and the person are intricately in­ those associated with less prestigious institu­ tertwined, and itwould be amistake to assume tions all showed weaker preference for their that signals ofingroup-favoring attitudes, be­ own group on implicit measures. Although cause their contentmay have its origins in cul­ the tendency toward ingroup preference is ro­ turally shared knowledge, are hence not a part bust, a group's status in the larger social hier­ ofthe individual (Banaji 200 1).Just as itwould archy is a decisive influence on the implicit make little sense to ask what makes a rectan­ biases that are observed. Members of privi­ gle a rectangle, its length or its width, so too leged groups overwhelmingly show ingroup would disambiguating person from culture in preference (70% and up is not a rare find­ any simplistic way be nonsensical. ing), whereas this tendency is subdued among members of socially derogated groups who internalize the broader cultural evaluation of Implicit Social Cognitions Predict their group. The elderly, for example, show Behavior no ingroup-favoring implicit attitudes; gays The nature of ISCs should be of interest to and lesbians show weaker ingroup preference the law to the extent that they predict behav­ than do heterosexual participants (Noseketal. ior. Before such an exercise is initiated, itis im­ 2007); and blacks in the United States and portant to know ifimplicit measures differen­ in South Africa show substantially weaker in­ tiate groups known to differ a priori. Indeed, group preference than do whites in those na­ in several domains, the !AT detects expected tions (Shutts et a1. 2007). group differences. Several known-groups

W'WlJ}.annualreviews.org • Implicit Social Cognition and Law 435 validations exist from the clinical and so­ A priming measure assessing attitudes to­ cial domains. Although both criminals high ward the overweight also predicted how far and low in psychopathy [characterized by participants placed their chair from an over­ "grandiosity, callousness, manipulation, lack weight woman (Bessenoff & Sherman 2000). of empathy, and lack of guilt or remorse" Strikingly, white students' implicit attitudes (Snowden et al. 2004, p. 621)] implicitly pre­ toward blacks, measured at the beginning of ferred the concept "peaceful" compared with the semester, predicted the longevity of their "violent," this preference was muted among relationship with a randomly assigned black murderers diagnosed as psychopathic (Gray roommate. Those who showed the most im­ et al. 2003, Snowden et al. 2004); that is, they plicit negativity toward blacks at the school exhibited lower implicit dislike for violence year's start were more likely not to be living than did nonpsychopathic murderers. Like­ with their black roommate at the academic wise, clinically identified pedophile criminals year's end (Towles-Schwen & Fazio 2003). implicitly associated images of children with These results sit in line with laboratory find­ sex, whereas nonpedophile criminals associ­ ings that people with more negative implicit ated adult images with sex; 78% ofpedophiles racial attitudes were viewed as less friendly in this study were correctly classified by the by confederates or observers unaware oftheir task (Gray et al. 2005). An implicit measure racial attitudes (Dovidio et al. 2002, Fazio clearly distinguished between subjects who et al. 1995, McConnell & Leibold 2001). had phobias associated with either snakes or Nonverbal behaviors such as facial expres­ spiders (Teachman & Woody 2003) and be­ sions, eye contact, and bodyposture have been tween subjects who were smokers and non­ shown to leak implicit attitudes. Those who smokers (Swanson et al. 2001). The IAT also possess stronger negative attitudes toward a detected stronger self+injury associations in stigmatized group tend to exhibit more neg­ self-injurers, as well as suicide ideation in ative behaviors (e.g., blinking) and less pos­ those who are at risk (Nock & Banaji 2007a, itive behaviors (e.g., smiling) when interact­ Nock & Banaji 2007b). It also efficiently de­ ing with a member of that group (Lemm marcates social groups. Men and women are 2006, McConnell & Leibold 2001). Further, distinguishable with near-lOO% accuracy on ISCs relate to judgments of identical actions a gender identity measure. Other tests also or objective states, depending on the target's detect group differences such as Japanese ver­ group membership. People with higher im­ sus Korean, blackversus white, and gay versus plicit bias judged ambiguous actions bya black straight, among others (Nosek et al. 2007). (Rudman & Lee 2002) orTurkish (Gawronski Importantly, implicit bias also predicts in­ et al. 2003) target more negatively. Even the dividual discriminatory behaviors. In addi­ same facial expression appeared different de­ tion to predicting medical interns' treatment pending on implicit bias. More negative im­ decisions (Green et al. 2007), people with plicit racial attitudes were related to a lowered greater implicit negativity toward blacks were threshold for detecting hostility on black, but less likely to anticipate befriending an African not white, faces (Hugenberg & Bodenhausen American and tended to sit further away 2003). from an African American partner. Thosewho Otherlines ofresearch demonstrate the re­ more strongly associated black with the con­ lationship between ISCs and lower-level be­ cept "physical" (compared with the concept haviors that are not immediately observable, "mental") were also more likelyto predict that such as cognitive processing or neurological an African American partner would perform responses. Implicit racial attitudes assessed by poorly on a joint academic task and rate an the IATwere linked to greater amygdala acti­ individual African American partner in more vation when viewing unfamiliar black (versus stereotypic fashion (Amodio & Devine 2006). white) faces (Cunningham et al. 2003, Phelps

436 Lane. Kong. Bana}i et al. 2000). Additionally, white participants and behavior (Fazio & Towles-Schwen 1999) with stronger antiblack bias performed poorly suggests that both motivation to be egalitarian on a measure ofcognitive control after inter­ and the opportunity to control one's behavior acting with a black partner, suggesting that affect whether implicit bias is manifested be­ those with stronger bias used more cognitive haviorally. Consider two people with identical resources during the interaction (Richeson & negative implicit outgroup attitudes. If per­ Shelton 2003). In another study, participants son A is motivated to be nonbiased and able viewed unfamiliar black faces while activation to control the influence of his bias, then the in key brain regions was assessed. Those with MODEmodel suggests that the ISC-behavior stronger antiblack bias exhibited more acti­ link will be disrupted. Alternatively, ifperson vation in brain regions associated with cog­ B lacks such motivation, her ISCs are more nitive control, suggesting they were engaging likely to influence behavior. in attempts to control automatic reactions to Datasupport this idea. Among participants the faces. Activation in one ofthese regions­ low in motivation to control bias, implicit bias the dorsolateral medial prefrontal cortex­ (measured via priming) predicted trait rat­ mediated, or accounted for, the relationship ings ofblack targets, relative to whites. Those between implicit bias and amount of cogni­ highly motivated to control bias showed the tive interference on a task that followed an reverse pattern, indicatingtheymay have been interaction with a black individual (Richeson overcorrecting for their implicit bias (Olson & Shelton 2003). These data suggest that the & Fazio 2004). Similarly, for white partici­ depletion of cognitive resources while inter­ pants low in motivation to control prejudicial acting with a black partner was due to an at­ responses, implicit bias predicted anticipated tempt to exercise control over one's biases. comfortlevel duringan unscripted interaction The most systematic exploration of the with a black partner. However, itwas not pre­ connection between implicit bias and behav­ dictive for those strongly motivated to control ior is a meta-analysis of studies that included prejudice (Towles-Schwen & Fazio 2003). Im­ the IAT and participant behaviors (Poehlman plicit bias toward gay people predicted non­ et al. 2007). Across studies, ISCs predicted a verbal behavior during an interaction with a range of criterion variables, including non­ gay partner only for participants low in mo­ verbal behavior, social judgments, physiolog­ tivation and in the tendency to control their ical responses, and social action. Both im­ behavior (Dasgupta & Rivera 2006). plicit (average r = 0.27) and explicit (average The path from implicit bias to negative be­ r = 0.34) measures were significantly related havior does not appear immutable. At least to behavior across topics. In the domain of under certain conditions, when people are stereotyping and prejudice, ISCs better pre­ motivated to behave in a nonprejudiced man­ dicted criterion behavior (average r = 0.22) ner, they may be able to override the effects than explicit cognitions (average r = 0.12). ofimplicit bias on behavior. People are most likely to be able to perform this correction Moderators ofthe bias-behavior relation­ when they can recognize the potential for so­ ship. Implicit bias does not inevitably lead to cial group membership to influence their be­ behavior. With the relationship between bias havior and to control the behavior itself. and behavior fairly well established, research is turning to when and how implicit bias is likely to be linked to action. In particular, mo­ Implicit Social Cognitions are Malleable tivation may moderate the link between ISCs and behavior. The MODE (Motivation and Despite their prevalence and magnitude, ISCs Opportunity as DEterminants of behavior) are not impervious to change. Like their ex­ model of the relationship between attitude plicit counterparts, the immediate situation's

www.annua!reviews.org • Implicit Soda! Cognition and Law 437 demands and a person's personality affect the types or "yes" to counterstereotypes, implicit strength of ISCs. Situational features affect stereotyping was eliminated (Kawakami et al. implicit biases, either by altering some fea­ 2000). Similarly, repeated pairings of black ture ofthe target social group or by changing faces with positive words during an ostensibly the participant's momentary goals or motiva­ unrelated exercise resulted in more egalitarian tions (see Blair 2002 for a review). Implicit implicit racial attitudes, even though partici­ bias not only changes as a function of the pants were unaware of any systematic pairing social situation but also varies with personal between positive words and black faces. This characteristics. reduction persisted for two days following ex­ Implicit biases are sensitive to features posure to the black-positive pairing (Olson & of the local situation. Exposure to counter­ Fazio 2006). Racial shooting bias in a police stereotypical outgroup members often re­ simulation was decreased after repeated expo­ duces implicit bias. For example, implicit sure to pairs ofstimuli in which ethnicity was race bias was reduced both immediately and unrelated to criminality (Plant et al. 2005). 24 hours following exposure to well-liked Besides these interventions, other tech­ African Americans and disliked white Ameri­ niques have capitalized on the extent to which cans (Dasgupta & Greenwald 2001), whereas implicit bias depends on salient social cat­ imagining a female leader reduced implicit egories. Michael Jordan, like mere mortals, gender stereotyping (Blair et al. 2001). Ad­ belongs to multiple social categories: gender ditionally, when the group "black" was repre­ (male), occupation (athlete), and race (African sented by well-liked African Americans such American). Classifying prominent black ath­ as Michael Jordan, implicit attitudes toward letes and white politicians according to their blacks became more positive (Mitchell et al. occupation reduced antiblack implicit bias 2003). compared with classification by race (Mitchell Benefits of exposure to (or thoughts et al. 2003; see Barden et al. 2004 for a similar about) counterstereotypical individuals ex­ result). In a striking example ofhow multiple tend beyond the lab: Dasgupta & Asgari identities can shape performance, a subtle re­ (2004) measured implicit gender stereotypes minder ofAsian women's gender or ethnicity (male+leader) among female students at a co­ lowered or raised their performance, respec­ educational and an all-women's college. Al­ tively, on a math task, compared with a con­ though women at both schools held similar trol condition (Shih et al. 1999). Even abilities implicit beliefs at the start of college, after expected to be stable and impervious to such one year women at the single-sex college did minor suggestions are affected by factors not not display any implicit bias, whereas women easily identifiable. at the coeducational school had, on average, Goals vary by situation, and immediate stronger implicit stereotypes ofmale+leader. goals help to determine ISCs. Implicit stereo­ Greater exposure to female teachers at the types were reduced after receiving positive single-sex school was responsible for the de­ feedback from a black manager (presumably crease in bias. Presumably, observing and in­ increasing motivation to value him) but in­ teractingwith female professors strengthened creased after receivingnegative feedback (pre­ the association offemale+leader. sumably increasing motivation to derogate Conscious exertion to be unbiased may­ him) (Sinclair & Kunda 1999). Desire to affili­ at least temporarily-reduce implicit bias. In ate with others lowers bias: Implicit racial bias one study, participants responded "no" when was lower after interacting with a black supe­ theysaw a social group (skinheads) paired with rior than with a black subordinate (Richeson a stereotypic trait, and "yes" when they saw & Ambady 2001) or after interacting with the group paired with a counterstereotypic an experimenter who wore a shirt that said trait. After 480 trials ofsaying "no" to stereo- "Eracism" (implying egalitarian beliefs) when

438 Lane. Knng • Banaji the experimenter was well liked (Sinclair et al. fluence behavior, and to be sensitive to in­ 2005). tervention. What are the legal implications? Motivation, of course, is not solely situ­ Broadly framed, this question is not novel. ational: Some people are more disposition­ Legal scholars have long analyzed the sig­ ally motivated to be nonprejudiced, a ten- nificance ofvarious states of mind, including . dency that moderates ISCs. People motivated ones that are notentirelypurposeful; the topic to be nonprejudiced for personal (or internal) of unconscious prejudice has been discussed reasons, but not social (or external) reasons for decades (Lawrence 1987). Recent findings showed reduced implicit racial bias on a phys­ in implicit social cognition sharpen the ques­ iological measure (Amodio et al. 2003) and tion because it is now supported by replicable a reaction-time task (Devine et al. 2002; see evidence from numerous laboratories rather Vanman et al. 2004 for a case in which motiva­ than by anecdote, hunch, or psychoanalytic tion was related to a reaction time, butnot to a theory. physiological, measure ofbias). Implicitbias is Indeed, assuming that the scientific re­ also related to more general cognitive styles, search continues along its current trajectory, such that people with highly rigid thinking implicit social cognition has the potential to styles or strongly right-wing ideologies ex­ influence the understanding of intent in all hibitstrongerimplicitbias (Cunninghametal. bodies oflaw. For instance, constitutional and 2004b). statutory law governing civil rights and the Personal and situational factors do not ex­ equal treatment of individuals is clearly sub­ ist in a vacuum, and the individual traits that ject to revision because implicit social cog­ a person brings to a situation often interact nition destabilizes conventional understand­ with the immediate context. Social dominance ings ofdisparate treatment, disparate impact, orientation (SDO), an attitudinal system re­ hostile environments, and color or gender ferring to whether people prefer hierarchi­ consciousness. Observed disparities between cally organized groups, was related to implicit social groups will likely be examined for con­ ingroup preference only when the students' nection to implicit bias as an aggravating fac­ university was threatened. Without social tor. In criminal law, the data are relevant in threat, SDO was unrelated to preference for a debating policies and laws surrounding racial generic ingroup. Under a condition ofgroup profiling, self-defense, community policing, threat, participants who preferred hierarchi­ juryselection, and penalty setting. Finally, the cal arrangements ofsocial groups (high-SDO) various media that transmit and sustain our showed much larger implicit ingroup prefer­ culture will increasingly become a source for ence than did low-SDO students (Pratto & investigation ofimplicit bias because they are Shih 2000). Peruche & Plant (2006) found the obvious source of what we see and hear. that training participants not to associate race Already, such matters have begun to be exam­ with athleticism (by repeated exposure to pair­ ined in law reviews and cases. ings of black and white faces with athletic or nonathletic objects, in which race and athletic Legal Literature features were independent) was particularly successful in reducing this race-athlete stereo­ The modernhistoryoflegal scholarlyengage­ type among participants highly motivated for ment with implicit social cognition began in personal reasons to be nonbiased. the mid-1990s. In 1995 and 1998, Krieger wrote the seminal articles applying cognitive psychology to questions of, first, employment THE LAW discrimination and, next, affirmative action. Implicit biases appear to be widespread, to She argued that employment discrimination be discrepant from self-reported bias, to in- may be caused not by racist villains who

www.annualreviews.o'-g • Implicit Social Cognition and Law 439 relish inflicting harm on minorities; rather, it ing scientists of the lAT, including its cre­ may be a natural byproduct of banal cogni­ ators, view as responsible use, Saujani raised tive sorting (Krieger 1995). Affirmative action thoughtful, provocative questions. may exacerbate negative stereotypes ofminor­ In 2005, Kang provided arguably the first ity beneficiaries. However, a strategy ofcolor­ systematic synthesis in the law reviews of the blindness is cognitively naive (Krieger 1998). implicit bias research, in support of a model Written largely before the empirical focus on of racial mechanics. In addition to provid­ individual differences in ISCs, these articles ing this synthesis, Kang (2005) questioned relied on the antecedent psychological litera­ the Federal Communications Commission's ture on bounded rationality and emphasized strong preference for local news as the way the role of schematic thinking. Nonetheless, that broadcasters should satisfy the Commu­ they created the foundation for subsequent le­ nication Act's public interest standard. Given gal analyses. Other significant contributions the violent crime stories disproportionately drawing on similar science include Pollard's featured on local news, he queried whether (1999) call for an evidentiary privilege for news programs functioned as Trojan horse employers who test for unconscious bias and viruses that increased implicit bias against mi­ Armour's (1995) call to break the prejudice norities. He also suggested a broad legal re­ habit. search agenda and was the first to encourage a By 2002, legal scholars began focused en­ behavioral realist incorporation ofthe science gagement with implicit social cognition sci­ ofimplicit social cognition. ence, including the lAT, which by then had In 2006, the California Law Review pub­ gained scientific prominence. Thatyear, Blasi lished a special symposium volume on be­ (2002) described various psychological the­ havioral realism (Bayern 2006). In simplified ories with substantial attention to automatic terms, behavioral realism involves athree-step categorization, motivated cognition, and im­ process. First, identify advances in the mind plicit bias, and analyzed why advocacy based and behavioral sciences that provide a more on folk theories ofprejudice would fail. Inpar­ accurate model of human cognition and be­ ticular, he rejected traditional advocacystrate­ havior. Second, compare that new model with gies that understood stereotypes as either the latent theories of human behavior and empirical mistakes to be corrected or moral decision making embedded within the law. failures to be redeemed. These latent theories typically reflect com­ In2003, Saujani provided extensive discus­ mon sense based on naive psychological the­ sion of the lAT and suggested that it could ories. Third, when the new model and the playa useful role in adjudicating equal pro­ latent theories are discrepant, ask lawmak­ tection claims. In washington v. Davis (1976), ers and legal institutions to account for this the Supreme Courtclarified that an equal pro­ disparity. An accounting requires either alter­ tection violation required a finding of dis­ ing the law to comport with more accurate criminatoryintent, notmere disparate impact. models ofthinking and behavior or providing Because such intent may have to be estab­ a transparent explanation of "the prudential, lished through circumstantial evidence, the economic, political, or religious reasons for Court identified various relevant factors in a retaining a less accurate and outdated view" subsequent case, Village ofArlington Heights (Kang & Banaji 2006, p. 1065). The extent v. Metropolitan Housing Development Authority of the pressure to be more behaviorally re­ (1977). Saujani (2003) suggested that legisla­ alistic depends on numerous factors, such as tors should take thelATand have those results the strength of the scientific consensus re­ admitted as relevant evidence in this search garding the emergent model, the size of the for discriminatory intent. Although such a gap between the new model and old assump­ recommendation far exceeds what the lead- tions, and the consequences of both action

440 Lane • Kong • Banaji and omission. This call for behavioral real­ The aforementioned works provide a ism is consistent with the increased attention sound introduction to the emergent legal lit­ to behavioral economics in legal discourse and erature on implicit bias and the law. In addi­ calls for a more comprehensive and situation­ tion, dozens of articles have referred to the sensitive psychological portrait ofhuman ac­ idea ofimplicit bias generally and the IAT in tion, e.g., Hanson's "critical realism" (Hanson particular. Most make only passing reference, & Yosifon 2003). but some have engaged more substantially Within the symposium, Greenwald & with the science and its legal implications. Krieger (2006) succinctly summarized the sci­ Noteworthy examples include discussions of ence underlying implicit bias. Krieger &Fiske petit (page 2005) and grand jury (Teshima (2006) discussed how disparate treatment doc­ 2006) selection, implicit bias of capital de­ trine under Title VII ofthe 1964 Civil Rights fense attorneys (Eisenberg &]ohnson 2004), Act might be made more behaviorally re­ health-care delivery (Shin 2002), and employ­ alistic, specifically by challenging the "hon­ ment discrimination (Bagenstos 2006, Poirier est belief" rule (that an employer should es­ 2003). cape liability if it honestly believed in the The articles mentioned so far largely em­ nondiscriminatory reasons it provided) and brace the implicit bias research as credible, the "same actor" inference (that ifthe person reliable, and illuminating. Others have been who fired an employee was also the person more skeptical. For instance, Rachlinski and who had hired that employee, there would be colleagues (2007) question whether labora­ an inference ofno discrimination). tory findings ofimplicit bias predictbehaviors Kang & Banaji (2006) revised various affir­ in the real world that surpass some mini­ mative action arguments into new "fair mea­ mum threshold ofmoral or legal significance. sures." In particular, they reframed certain af­ Judges completed an IAT measure of racial firmative action programs, not as reparations attitudes and then sentenced fictional defen­ for the past or ways to achieve general future dants based on a paper profile. In this study, social benefits, but as specific techniques to judges' ISCs were uncorrelated with their counter present implicit bias. One such tech­ judgments. Null results such as this one can nique would bc to deploy "debiasing agents" appear for two reasons-either because no ef­ who are specifically selected for their coun­ fect of implicit bias predicting behavior ac­ tertypical attributes that decrease implicit tually exists or because such an effect exists bias instead ofstandard "role models," which but fails to be detected because of an un­ have a dubious constitutional and Title VII derpowered or otherwise flawed test (the lat­ status. ter is referred to as a Type 2 error in statis­ Blasi & Jost (2006) provided the first tics). The relatively small sample of judges systematic law review introduction to sys­ (N = 70) suggests that their null result may tem justification theory, which analyzes the be a Type 2 error, as does the diverse na­ motivation to defend and justify the extant ture of their sample (37 white, 30 black, social order, and examined its implications for and 3 Latino), the different distributions of effective legal and social justice advocacy.]olls bias among black and white Americans, and & Sunstein (2006) supplied a theoretical de­ subjects' different priming conditions, which scription ofways that the law might respond could have added systematic variability that to implicit bias, for example through alterna­ masked a bias-behavior relationship. tive methods ofdebiasing. Finally, Banks et al. The Poehlman et al. (2007) meta-analysis (2006) turned to criminal law and examined is­ provides the best response to this predictive sues ofprofiling, sentencing, and shooting in validity concern. Rachlinski et al. (2007) are light ofimplicit bias to expose the lack ofcon­ correctinnoting that the Poehlman paper de­ sensus on what racial equality actually entails. fined behavior expansively as "any measure of

www.annualreviews.org • Implicit Social Cognition and Low 441 a physical action, judgment, decision orphysi­ opinions make some mention ofunconscious ological reaction." Awkward body language is or implicit bias that it would be impossible to hardly in and of itself actionable. That said, try to list them all. such intermediate behaviors should not be In both concurring and dissenting opin­ dismissed as unimportant, for they can con­ ions, the Supreme Court has acknowledged tribute to the final decision-making process the potential for implicit bias to impede jus­ and influence marginal cases. Summed over tice.For example, in her dissent in Adarand large populations engaged in daily interac­ Constructors, Inc. v. Pefia (1995), Justice tions and evaluations, the aggregate impacton Ginsburg noted that "[b]ias both conscious individuals and groups maybe substantial. Ad­ and unconscious, reflecting traditional and ditionally, studies in the meta-analysis specif­ unexamined habits of thought, keeps up bar­ ically measured final or ultimate decisions, riers that must come down if equal opportu· including hiring recommendations (Rudman nityand nondiscrimination are ever genuinely & Glick 2001) and doctors' treatment rec­ to become this country's law and practice." ommendations (Green et al. 2007). Finally, She echoed these sentiments in the University other studies have provided further support ofMichigan affirmative action cases, quoting in additional contexts, such as job evaluations these precise words in her dissent in Gratz (Ziegert & Ranges 2005). Of course, further v. Bollinger (2003), and writing in her concur­ research will help clarify the behavioral con­ ring opinion in Grutterv. Bollinger (2003), that sequences ofimplicit bias. "[i]t is well documented that conscious and In contrast to Rachlinski's reasonable calls unconscious race bias, even rank discrimina­ for caution, Mitchell & Tetlock (2007) la­ tion based on race, remain alive in our land, bel the large body of work on implicit bias impeding realization ofour highest values and to be pseudoscience created by ideologues ideals." in order to warp public policy. In our own More specifically, the Courthas speculated work (J. Kang, K.A. Lane & M.R. Banaji, that implicit bias may affect the perceptions manuscript in preparation), we suggest that of participants in the legal system: In Batson this accusation of junk science should be v. Kentucky (1989), Justice Marshall suggested understood as predictable political backlash, in his concurring opinion that "[a] prosecu­ regrettablylaced with ad hominem and sttaw­ tor's own conscious or unconscious racism person excess. Indeed, Bagenstos (2007) char­ may lead him easily to the conclusion that acterized their appraisal as "based not on any a prospective black juror is 'sullen,' or 'dis­ 'scientific' ground, but on normative assump­ tant,' a characterization that would not have tions ... rest[ed] on a very narrow view, based come to his mind if a white juror had acted on notions of individual fault, that the law identically."Justice O'Connorvoiced a similar should prohibit only discrimination that re­ concern in her dissent in Georgia v. McCollum sults from irrational animus." (1992), noting "[i]t is by now clear that con­ scious and unconscious racism can affect the way white jurors perceive minority defen­ Judicial Opinions dants and the facts presented at their trials, We have reviewed how the science has moved perhaps determining the verdict of guilt or from psychology journals into law reviews, innocence." She then offered a possible rem­ but what about the next move into the case edy, suggestingthat"[u]singperemptorychal­ law? Just as various theories of unconscious lenges to secure minority representation on bias have long been pondered in law reviews, the jurymayhelp to overcome such racial bias, many judicial opinions have noted the exis­ for there is substantial reason to believe that tence and potential legal significance of un­ the distorting influence of race is minimized conscious forms ofbias. Indeed, so many such on a racially mixed jury."

442 Lone. Kang • Banaji In Price Waterhouse v. Hopkins (1989), Jus­ academic literature were to early, important tice Brennan, writing for the plurality, sug­ critical race theory literature (e.g., Lawrence gested that lack ofawareness ofbias does not 1987), other references were made to the excuse the bias or outcomes resulting from it: more social cognitive literature (citations "unwitting or ingrained bias is no less injuri­ made to Bargh 1989, Krieger 1995). ous or worthy of eradication than blatant or United States v. Taveras (2006), provides calculated discrimination"; "thefact thatsome dicta that go the other way-minimizing the or all ofthe partners at Price Waterhouse may significance of implicit bias. At issue was the have been unaware of that motivation, even admissibility of potentially prejudicial evi­ within themselves, neither alters the fact of dence inthe penalty phase ofa murder convic­ its existence nor excuses it." tion. In dictum, JudgeJack Weinstein opined Among appellate court opinions that ad­ that dress implicit or unconscious bias, the most Judges, who are often exposed to the nas­ interesting ones consider whether Title VII tier elements of human behavior on a reg­ can account for such cognitive processes. ular basis, are likely to be better able than Take, for example, Thomas v. Eastman Kodak laypersons to control the effects of prej­ Co. (1999), which clarified that "[t]he ultimate udicial information on their decisions. cf. question is whether the employee has been Jeffrey J. Rachlinski, et al. Does Uncon­ treated disparately 'because of race.' This is scious Bias Affect Trial Judges? (2006) (un­ so regardless of whether the employer con­ published manuscript, on file with court) sciously intended to base the evaluations on (empirical study ofimplicit associations sug­ race, or simply did so because of unthinking gests that judges, while still susceptible to stereotypes or bias." unconscious biases, are able to set them Ifwe raise the threshold of relevance and aside when rendering judgment even when discuss only those cases that directly rely on primed with information designed to elicit implicit social cognition research for a factual negative reactions). finding or legal conclusion, few cases warrant mention. Two cases discuss implicit bias The Rachlinski article cited above cau­ research specifically, but in dicta and toward tioned aboutpredictive validity. From the lack opposite conclusions about its significance. of correlation between judges' ISCs and sen­ In Chinn v. Runnels (2004), a habeas corpus tencing decisions, Judge Weinstein inferred petitioner challenged the San Francisco a certain judicial objectivity, driven by prior Superior Court's grand jury selection process and regular exposure to the "nastier elements for never having selected a foreperson of of human behavior." As anybody who works Chinese, Filipino, or Latino descent for 36 with correlational data knows, it is danger­ years. Given the very deferential standard ous to interpret a null result as showing a lack of review, the federal court accepted the ofrelationship, especially given the large de­ California appellate court's ruling that the mand characteristics cueing the judges as to government had successfully rebutted the how they should respond on the sentencing prima facie case of discrimination. However, exercise and given the low statistical power in the conclusion, the federal court wrote to detect an effect. This null result should be extensively to explain why under a de novo read in light of the Poehlman et al. (2007) standard of review there could have been a meta-analysis showing a systematic relation­ different result. The court specifically cited ship between implicit bias and behavior. Of a "growing body of social science [that] course,Judge Weinstein did notfollow such a recognizes the pervasiveness of unconscious reading in Taveras. It is noteworthy, and per­ racial and ethnic stereotyping and group haps not entirely surprising, that one of the bias." Although many of the citations to the first references to the literature in a published

'11Y1JJW.annualreviews.org • Implicit Social Cognition and Law 443 judicial opinion proposed that judges are im­ ties alone. As of this writing, the case is on mune to implicit bias. appeal. Finally, in Farrakhan v. Gregoire (2006) implicit bias expert testimony apparently in­ fluenced an important judicial finding of CONCLUSION discrimination-although that finding was In a recent commentary, Banaji (2007) wrote nevertheless insufficient for plaintiff's victory. that "[t]o be intelligent means many things The case addressed a §2 Voting Rights Act of course .... [I]ntelligence is knowing how (VRA) challenge to Washington state's felon to weigh the evidence that flies in the face disenfranchisement statute. Surprisingly, the of steadfast assumptions. It means to know court found "compelling evidence of racial when causality can be inferred and not, to discrimination and bias in Washington's crim­ know when the weight of correlational evi­ inal justice system." This finding was not dence must be taken seriously, to know that based "solely on statistics" that showed racial a replication is worth much more than a sin­ disparities in criminal law enforcement. In­ gle demonstration, to know that when new stead, it was based on expert testimony that methods divulge strange truths about us and included substantial discussion ofboth struc­ our brethren, it may be the theory that has to tural/institutional factors and, more impor­ go. The moral obligation to be intelligent re­ tant to this discussion, the potential impact of quires that we keep abreast ofdiscoveries that implicit bias. One ofthe two principal expert require old views to be bagged and put out on reports relied upon by the court contained an the curb for recycling-every week." extensive discussion of implicit bias. In ad­ The eager engagement of legal scholars dition, another source described as helping and lawyers in the discoveries about implicit to "bolster the Court's conclusion" included social cognition comes from the challenge the expert testimony ofAnthony Greenwald, they pose to existing assumptions about hu­ inventor of the IAT, who attached a draft man nature that currently reside in the law. It of his submission to the Behavioral Realism is not for this review to take a position on how symposium. such evidence should inform the law. How­ This finding of racial discrimination was, ever, this review should reveal the robustness however, insufficient to find a VRA vio­ of the evidence that much of human cogni­ lation under a "totality of circumstances" tion can and does occur without introspective test. The court heavily emphasized that the access, that such processes nevertheless influ­ state of Washington did not have any his­ ence and guide decision making, that the costs tory of official discrimination that under­ incurred by individuals and social groups are mined minority voting rights. Emphasizing less at the hands ofthe malign and more likely this factor seems to miss the point of how to come from the unaware and uncontrolled implicit bias functions. Nonetheless, Far­ mental acts of ordinary people. The law will rakhan demonstrates how implicit social cog­ need to include these discoveries about how nitive explanations can produce judicial find­ the mind (really) works to be true to Erksine's ings of discrimination that probably would idea that intelligence about such matters, not not have been found by statistical dispari- just meaning well, is the virtue.

DISCLOSURE STATEMENT Mahzarin R. Banaji is an officeholder in Project Implicit, a nonprofit organization that seeks to disseminate and educate about the science ofimplicit social cognition.

444 Lone· Kang • Banaji ACKNOWLEDGMENTS We thank the National Institutes ofHealth, the Mind Science Institute, the Third Millennium Foundation, Harvard University, UCLASchool ofLaw, and the UCLAAsian American Studies Center for their support ofthis work, and Tiffany Meites, Christopher Dial, Megan Magleby, and Leigh Rich for help in the preparation ofthe manuscript.

LITERATURE CITED Adarand Constructors, Inc. v. Pena, 515 U.S. 200 (1995) Amodio DM, Devine PG. 2006. Stereotyping and evaluation in implicit race bias: evidence for independent constructs and unique effects on behavior. J. Pers. Soc. Psychol. 91 :652-61 Amodio DM, Harmon-Jones E, Devine PG. 2003. Individual differences in the activation and control of affective race bias as assessed by startle eyeblink response and self-report. J. Pers. Soc. Psychol. 84:738-53 Armour J. 1995. Stereotypes and prejudice: helping legal decisionmakers break the prejudice habit. Calif. Law Rev. 83:733-72 Ashburn-Nardo L, Knowles ML, Monteith MJ. 2003. Black Americans' implicit racial associ­ ations and their implications for intergroup judgment. Soc. Cogn. 21:61-87 Ashburn-Nardo L, Voils CI, MonteithMJ. 2001. Implicitassociations as the seeds ofintergroup bias: How easily do they take root? J. Pers. Soc. Psychol. 81:789-99 Bagenstos SR. 2006. The structural turn and the limits of antidiscrimination law. Calif. Law Rev. 94:1--47 Bagenstos SR. 2007. Implicit bias, "science," and antidiscrimination law. Harvard Law Policy Rev. 1:477-93 Banaji MR. 2001. Implicit attitudes can be measured. In The Nature ofRemembering: Essays in Honor ofRobert G. Crowder, ed. HL Roediger, JS Nairne, I Neath, A Surprenant, pp. 117-50. Washington, DC: Am. Psychol. Assoc. Banaji MR. 2007. Foreword. In Beyond Common Sense: Psychological Science in the Courtroom, ed. E Borgida, ST Fiske. Oxford: Blackwell. In press Banaji MR, Bazerman MH, Chugh D. 2003. How (un)ethical are you? Harvard Bus. Rev. 81:56-64 Banaji MR, Bhaskar R. 2000. Implicitstereotypes and memory: the bounded rationalityofsocial beliefs. In Memory, Brain, and Belief, ed. DL Schacter, E Scarry, pp. 139-75. Cambridge, MA: Harvard Univ. Press Banaji MR, Greenwald AG. 1995. Implicit gender stereotyping in judgments offame. J. Pers. Soc. Psychol. 68:181-98 Banks RR, Eberhardt ]L, Ross L. 2006. Discrimination and implicit racial bias in a racially unequal society. Calif. Law Rev. 94:1169-90 BardenJ, Maddux WW, Petty RE, Brewer ME. 2004. Contextual moderation of racial bias: the impact ofsocial roles on controlled and automatically activated attitudes. J. Pers. Soc. Psychol.87:5-22 Bargh]A 1989. Conditional automaticity: varieties ofautomatic influence in social perception and cognition. In Unintended Thought, ed. JS Uleman, ]A Bargh, pp. 3-51. New York: Guilford Batson v. Kentucky, 476 U.S. 79 (1989) Bayern SJ, ed. 2006. Symposium on behavioral realism. Calif. Law Rev. Vol. 94, No.4 Bazerman MR, Chugh D, Banaji MR. 2005. When good people (seem to) negotiate in bad faith. Negotiation 8(10):3-5

W'WW.annualreviews.org • Implicit Social Cognition and Law 445 BessenoffGR, ShermanJW. 2000. Automatic and controlled components ofprejudice toward fat people: evaluation vs stereotype activation. Soc. Cogn. 18:329-53 Blair N. 2002. The malleability ofautomatic stereotypes and prejudice. Pers. Soc. Psychol. Rev. 6:242-61 Blair N, Ma JE, Lenton AP. 2001. Imagining stereotypes away: the moderation of implicit stereotypes through mental imagery. J. Pers. Soc. Psychol. 81 :828-41 BlascovichJ, Mendes WB, Hunter SB, Lickel B, Kowai-Bell N. 2001. Perceiver threat in social interactions with stigmatized others. J. Perro Soc. Psychol. 80:253-67 Blasi G. 2002. Advocacy against the stereotype: lessons from cognitive social psychology. UCLA Law Rev. 49:1241-81 Blasi G, Jost JT. 2006. System justification theory and research: implications for law, legal advocacy, and social justice. Calif. Law Rev. 94: 1119-68 BossonJK, Swann WBJ, Pennebaker JW. 2000. Stalking the perfect measure ofimplicit self­ esteem: the blind men and the elephant revisited? J. Pers. Soc. Psychol. 79:631-43 Brendl CM, Markman AB, Messner C. 2005. Indirectly measuring evaluations ofseveral atti­ tude objects in relation to a neutral reference point. J. Exp. Soc. Psychol. 41 :346-68 Cadinu MR, Rothbart M. 1996. Self-anchoring and differentiation processes in the minimal group setting.]. Pers. Soc. Psychol. 70:661-77 Carney DR, Banaji MR. 2007. Social cognition and social neuroscience. In Cognition, ed. M Tarr. MA Sinauer. In press Chinn v. Runnels, 343 F.Supp.2d 891 (N.D. Ca. 2004) Cohen J. 1988. Statistical Power Analysis for the Behavioral Sciences. Hillsdale, NJ: Earlbaum. 2nd ed. CorrellJ, Park B, Judd CM, Wittenbrink B. 2002. The police officer's dilemma: using ethnicity to disambiguate potentially threatening individuals.]. Pers. Soc. Psychol. 83: 1314-29 Cunningham WA,JohnsonMK, GatenbyJC, GoreJC, BanajiMR. 2003. Neural components ofsocial evaluation. J. Pers. Soc. Psychol. 85 :63 9-49 Cunningham WA,JohnsonMK, Raye CL, GatenbyJC, GoreJC, Banaji MR. 2004a. Separable neural components in the processing ofblack and white faces. Psychol. Sci. 15:806-13 Cunningham WA, NezlekJB, BanajiMR. 2004b. Implicit and explicit ethnocentrism: revisiting the ideologies ofprejudice. Pers. Soc. Psycho!. Bull. 30:1332-46 Cunningham WA, Preacher KJ, Banaji MR. 2001. Implicit attitude measures: consistency, stability, and convergent validity. Psychol. Sci. 121:163-70 Dasgupta N, Asgari S. 2004. Seeing is believing: exposure to counterstereotypicwomen leaders and its effect on the malleability of automatic gender stereotyping. J. Exp. Soc. Psychol. 40:642-58 Dasgupta N, Greenwald AG. 2001. On the malleability of automatic attitudes: combating automatic prejudice with images of admired and disliked individuals. J. Pers. Soc. Psychol. 81:800-14 Dasgupta N, Rivera LM. 2006. From automatic antigay prejudice to behavior: the moderating role ofconscious beliefs about gender and behavioral control.]. Pers. Soc. Psychol. 91:268­ 80 De HouwerJ. 2003. The extrinsic affective Simon task. Exp. Psychol. 50:77-85 de Jong PJ, van den HoutMA, Rietbroek H, HuijdingJ. 2003. Dissociations between implicit and explicit attitudes toward phobic stimuli. Cogn. Emot. 17:521-45 Devine PG, Plant EA, Amodio DM, Harmon-Jones E, Vance SL. 2002. The regulation of explicit and implicit race bias: the role of motivations to respond without prejudice. J. Pers. Soc. Psychol. 82:835-48

446 Lane. Knng • Banaji ------

Devos T, Banaji MR. 2005. American = white? J. Pers. Soc. Psychol. 88:447-66 Dovidio ]F, Kawakami K, Gaertner SL. 2002. Implicit and explicit prejudice and interracial interaction.]. Pen Soc. Psychol. 82 :62--68 Egloff B, Schmukle SC. 2002. Predictive validity of an Implicit Association Test for assessing anxiety.]. Pers. Soc. Psychol. 83:1441-55 Eisenberg T,]ohnson SL. 2004. Implicit racial attitudes ofdeath penalty lawyers. DePaul Law Rev. 53:1539-56 Farrakhan v. Gregoire, 2006 U.S. Dist. LEXIS 45987 (E.D. Wash 2006) Fazio RH, Chen], McDonel EC, Sherman S]. 1982. Attitude accessibility, attitude-behavior consistency, and the strength of the object-evaluation association. J. Exp. Soc. Psychol. 18:339-57 Fazio RH, Jackson ]R, Dunton BC, Wtlliams C]. 1995. Variability in automatic activation as an unobtrusive measure of racial attitudes: a bona fide pipeline? J. Pers. Soc. Psychol. 69:1013-27 Fazio RH, Olson MA. 2003. Implicit measures in social cognition research: their meaning and use. Annu. Rev. Psychol. 54:297-327 Fazio RH, Towles-Schwen T. 1999. The MODE model of attitude-behavior processes. In Dual-Process Theories in Social Psychology, ed. S Chaiken, Y Trope, pp. 97-116. New York: Guilford Gaertner SL, Mann], Murrell A, Dovidio]F. 1989. Reducing intergroup bias: the benefits of recategorization.]. Pers. Soc. Psychol. 57:239-49 Gawronski B, Geschke D, Banse R. 2003. Implicit bias in impression formation: associations influence the construal ofindividuating information. Eur. J. Soc. Psychol. 33:573-89 Georgia v. McCollum, 505 U.S. 42 (1992) Gratz V. Bollinger, 539 U.S. 244 (2003) Gray NS, Brown AS, MacCulloch M], Smith], Snowden R]. 2005. An implicit test of the associations between children and sex in pedophiles.]. Abnorm. Psychol. 114:304-8 Gray NS, MacCulloch M], Smith], Morris M, Snowden R]. 2003. Violence viewed by psy­ chopathic murderers. Nature 423:497-98 Green AR, Carney DR, Pallin D], Ngo LH, Raymond KL, et al. 2007. Implicit bias among physicians and its prediction of thrombolysis decisions for black and white patients. J. Gen. Intern. Med. 22:1231-38 Greenwald AG, Banaji MR. 1995. Implicit social cognition: attitudes, self-esteem, and stereo­ types. Psychol. Rev. 102:4-27 Greenwald AG, Farnham SD. 2000. Using the ImplicitAssociation Test to measure self-esteem and self-concept.J. Pers. Soc. Psycho/. 79:1022-38 Greenwald AG, Krieger LH. 2006. Implicitbias: scientific foundations. Calif. Law Rev. 94:945­ 67 Greenwald AG, McGhee DE, Schwartz]LK. 1998. Measuringindividual differences inimplicit cognition: the Implicit Association Test.]. Pen Soc. Psychol. 74: 1464-80 Greenwald AG, Nosek BA, Banaji MR. 2003a. Understanding and using the Implicit Associa­ tion Test: I: An improved scoring algorithm.]. Pers. Soc. Psycho/. 85:197-216 Greenwald AG, Oakes MA, Hoffman HG. 2003b. Targets ofdiscrimination: effects ofrace on responses to weapons holders. J. Exp. Soc. Psychol. 39:399-405 Grutter V. Bollinger, 539 U.S. 306 (2003) Hanson], Yosifon D. 2003. The situation: an introduction to the situational character, critical realism, power economics, and deep capture. Univ. Penn. Law Rev. 152:129-346

7ffIJnJ).annualreviews.org • Implicit Social Cognition and Law 447 Hofmann W, Gawronski B, Gschwendner T, Le H, Schmitt M. 2005. A meta-analysis on the correlation between the Implicit Association Test and explicit self-report measures. Pers. Soc. Psychol. Bull. 31:1369-85 HugenbergK, Bodenhausen GV. 2003. Facingprejudice: implicitprejudice and the perception offacial threat. Psychol. Sci. 14:640-43 Jolls C, Sunstein C. 2006. The law ofimplicit bias. Calif. Law Rev. 94:969-96 JostJT, Banaji MR. 1994. The role ofstereotyping in system-justification and the production offalse consciousness. Br:]. Soc. Psychol. 33:1-27 Jost JT, Banaji MR, Nosek BA. 2004. A decade of system justification theory: accumulated evidence ofconscious and unconscious bolstering ofthe status quo. Polito Psychol. 25:881­ 920 JostJT, Pelham BW, CarvalloMR. 2002. Non-conscious forms ofsystem justification: implicit and behavioral preferences for higher status groups.]. Exp. Soc. Psychol. 38:586-602 Kang]. 2005. Trojan horses ofrace. Harvard Law Rev. 118:1491-593 Kang J, Banaji MR. 2006. Fair measures: a behavioral realist revision of affirmative action. Calif. Law Rev. 94:1063-118 Karpinski A, HiltonJL. 2001. Attitudes and the Implicit Association Test. J. Pers. Soc. Psychol. 81:774-88 Kawakami K, Dovidio JF, Moll J, Hermsen S, Russin A. 2000. Just say no (to stereotyping): effects of training in the negation of stereotypic associations on stereotype activation. J. Pers. Soc. Psychol. 78:871-88 Krieger LH. 1995. The content ofour categories: a cognitive bias approach to discrimination and equal employment opportunity. Stanford Law Rev. 47:1161-248 Krieger LH. 1998. Civil rights perestroika: intergroup relations after affirmative action. Calif. Law Rev. 86:1276-329 Krieger LB, Fiske ST. 2006. Behavioral realism in employment discrimination law: implicit bias and disparate treatment. Calif. Law Rev. 94:997-1062 Kuhnen U, Schiessl M, Bauer N, Paulig N, Pohlmann C, Schmidthals K. 2001. How robust is the IAT? Measuring and manipulating implicit attitudes of East- and West-Germans. Z. Exp. Psychol. 48:135-44 Lane KA, Banaji MR, Nosek BA, Greenwald AG. 2007. Understanding and using the Implicit Association Test: IV: what we know (so far) about the method. In Implicit Measures of Attitudes, ed. B Wittenbrink, N Schwarz, pp. 59-102. New York: Guilford Lane KA, MitchellJP, Banaji MR. 2005. Me and mygroup: cultural status can disrupt cognitive consistency. Soc. Cogn. 23:353-86 Lawrence CRI. 1987. Theid, the ego, and equal protection: reckoningwithunconscious racism. Stanford Law Rev. 39:317-88 Lernm KM. 2006. Positive associations among interpersonal contact, motivation, and implicit and explicit attitudes toward gay men.]. Homosex. 51:79-99 LivingstonRW. 2002. The role ofperceived negativity in the moderation ofAfrican Americans' implicit and explicit racial attitudes. J. Exp. Soc. Psycho!. 38:405-13 McConnell AR, LeiboldJM. 2001. Relations among the Implicit Association Test, discrimi­ natory behavior, and explicit measures ofracial attitudes.]. Exp. Soc. Psychol. 37:435-42 Mitchell G, Tetlock PE. 2007. Antidiscrimination law and the perils ofmindreading. Ohio State Law J. 67:1023 Mitchell JP, Macrae CN, Banaji MR. 2006. Dissociable medial prefrontal contributions to judgments ofsimilar and dissimilar others. Neuron 50:655-63 MitchellJP, Nosek BA, Banaji MR. 2003. Contextual variations in implicit evaluation.]. Exp. Psychol. Gen. 132:455-69

448 Lalle. [vl1lg • Ballaji NockMK, Banaji MR. 2007a. Prediction ofsuicide ideation and attempts among adolescents using a brief performance-based test. J. Consult. Clin. Psychol. In press Nock MK, Banaji MR. 2007b. Assessment of self-injurious thoughts using a behavioral test. Am.]. Psychiatry 164:820-32 Nosek BA. 2005. Moderators of the relationship between implicit and explicit evaluation. ]. &po Psychol. Gen. 134:565-84 Nosek BA, Banaji MR. 2001. The go/no-go association task. Soc. Cogn. 19:625-66 Nosek BA, Banaji M, Greenwald AG. 2002. Harvesting implicit group attitudes and beliefs from a demonstration web site. Group Dyn. 6:101-15 Nosek BA, Greenwald AG, Banaji MR. 2006. The ImplicitAssociation Test at age 7: a method­ ological and conceptual review. In Social Psycholog;y and the Unconscious: The Automaticity ofHigher Mental Processes, ed.]A Bargh, pp. 265-92. New York: Psychology Press Nosek BA, Hansen]]. 2007. The associations in our heads belong to us: searching for attitudes and knowledge in implicit evaluation. Cogn. Emot. In press Nosek BA, Smyth FL, Hansen]], DevosT, LinderNM, etal. 2007. Pervasiveness and correlates ofimplicit attitudes and stereotypes. Eur. Rev. Soc. Psychol. In press Olson MA, Fazio RH. 2004. Trait inferences as a function of automatically activated racial attitudes and motivation to control prejudiced reactions. Basic Appl. Soc. Psychol. 26: 1-11 OlsonMA, Fazio RH. 2006. Reducing automatically activated racial prejudice through implicit evaluative conditioning. Pers. Soc. Psychol. Bull. 32:421-33 Ottaway SA, Hayden DC, Oakes MA. 2001. Implicit attitudes and racism: effects of word familiarity and frequency on the Implicit Association Test. Soc. Cogn. 19:97-144 Page A. 2005. Batson's blind-spot: unconscious stereotyping and the peremptory challenge. Boston Univ. Law Rev. 85:155-262 Payne BK. 2001. Prejudice and perception: the role of automatic and controlled processes in misperceiving a weapon.]. Pers. Soc. Psychol. 81: 181-92 Payne BK, Cheng CM, Govorun 0, Stewart BD. 2005. An inkblot for attitudes: affect misat­ tribution as implicit measurement.]. Pers. Soc. Psychol. 89:277-93 Perdue CW, Dovidio ]F, Gurtman MB, Tyler RB. 1990. Us and them: social categorization and the process ofintergroup bias. J. Perr. Soc. Psychol. 59:475-86 Peruche BM, Plant EA. 2006. Racial bias in perceptions ofathleticism: the role ofmotivation in the elimination ofbias. Soc. Cogn. 24:438-52 Phelps EA. 2006. Emotion and cognition: insights from studies ofthe human amygdala. Annu. Rev. Psychol. 57:27-53 Phelps EA, O'Connor K], Cunningham WA, Funayama ES, Gatenby]C, et al. 2000. Per­ formance on indirect measures of race evaluation predicts amygdala activation. J. Cogn. Neurosd. 12:729-38 PlantEA, Peruche BM. 2005.Theconsequences ofrace for police officers' responses to criminal suspects. Psychol. Sci. 16:180-83 Plant EA, Peruche BM, Butz DA. 2005. Eliminating automatic racial bias: making race non­ diagnostic for responses to criminal suspects.]. &po Soc. Psychol. 41:141-56 Poehlman TA, Uhlmann EL, Greenwald AG, Banaji MR. 2007. Understanding and using the Implicit Association Test: III. Meta-analysis of predictive validity. Manuscript under review PoirierMR. 2003. Is cognitive bias at work a dangerous condition on land? Empl. Rights Employ. Policy]. 7:459-94 Pollard DA. 1999. Unconscious bias and self-critical analysis: the case for a qualified evidentiary equal employment opportunity privilege. Wash. Law Rev. 74:913-1031

www.annualreviews.org • Implicit Social Cognition and Law 449 Pratto F, Shih M. 2000. Social dominance orientation and group context in implicit group prejudice. Psychol. Sci. 11:515-18 Price Waterhouse v. Hopkins, 490 U5. 228 (1989) Rachlinski JJ, Johnson S, Wistrich AJ, Guthrie C. 2007. Does unconscious bias affect trial judges? Work. Pap., Cornell Univ., Ithaca, NY Richeson JA, Ambady N. 2001. Who's in charge? Effects of situational roles on automatic gender bias. Sex Roles 44:493-512 RichesonJA, SheltonJN. 2003. When prejudice does not pay: effects ofinterracial contact on executive function. Psychol. Sci. 14:287-90 Rudman LA, Feinberg], Fairchild K. 2002. Minority members' implicit attitudes: automatic ingroup bias as a function ofgroup status. Soc. Cogn. 20:294-320 Rudman LA, Glick P. 2001. Prescriptive gender stereotypes and backlash toward agentic women.J. Soc. Issues 57:743-62 Rudman LA, Kilianski SE. 2000. Implicit and explicit attitudes toward female authority. Pers. Soc. Psychol. Bull. 26:1315-28 Rudman LA, Lee MR. 2002. Implicit and explicit consequences of exposure to violent and misogynous rap music. Group Process. Intergroup Relat. 5:133-50 Saujani RM. 2003. The ImplicitAssociation Test: a measure ofunconscious racism in legislative decision-making. Mich.]. Race Law 8:395--424 Sekaquaptewa D, Espinoza P, Thompson M, Vargas P, von Hippel W. 2003. Stereotypic explanatory bias: implicit stereotyping as a predictor ofdiscrimination.]. Exp. Soc. Psychol. 39:75-82 Shih M, Pittinsky TL, Ambady N. 1999. Stereotype susceptibility: identity salience and shifts in quantitative performance. Psychol. Sci. 10:80-83 Shin MS. 2002. Redressing wounds: finding a legal framework to remedy racial disparities in medical care. Calif. Law Rev. 90:2047-100 Shutts K, Olson KA, Spelke ES, Banaji MR. 2007. Implicit attitudes in postapartheid South Africa. Work. Pap., Harvard Univ., Cambridge, MA SinclairL, Kunda Z. 1999. Reactions to a blackprofessional: motivated inhibition and activation ofconflicting stereotypes.]. Pers. Soc. Psychol. 77:885-904 Sinclair S, Lowery BS, Hardin CD, Colangelo A. 2005. Social tuning of automatic racial attitudes: the role ofaffiliative motivation.]. Pers. Soc. Psychol. 89:583-92 SnowdenR], GrayNS, SmithJ, MorrisM,MacCullochMJ. 2004. Implicitaffective associations to violence in psychopathic murderers. J. Forensic Psychiatry Psychol. 15:620-41 SwansonJE, Rudman LA, Greenwald AG. 2001. Using the Implicit Association Test to inves­ tigate attitude-behaviour consistency for stigmatised behaviour. Cogn. Emot. 15:207-30 Tajfel H, Billig MG, Bundy RP, Flament C. 1971. Social categorization and intergroup be­ haviour. Eur. J. Soc. Psychol. 1: 149-78 Tajfel H, Turner Jc. 1986. The social identity theory of intergroup behavior. In Psychology ofIntergroup Relations, ed. S Worchel, LWAustin, pp. 7-24. Chicago: Nelson-Hall Teachman BA, Woody SR. 2003. Automatic processing in spider phobia: implicit fear associa­ tions over the course oftreatment. J. Abnorm. Psychol. 112:100-9 Teshima DS. 2006. A "hardy handshake sort of guy": the model minority and implicit bias about Asian Americans in Chin v. Runnels. Asian Pac. Am. Law J. 11:122--41 Thomas v. Eastman Kodak Co., 183 F.3d 38 (1st Cir. 1999) Towles-Schwen T, Fazio RH. 2003. Choosing social situations: the relation between automati­ cally activated racial attitudes and anticipated comfort interactingwith African Americans. Pers. Soc. Psychol. Bull. 29:170-82

450 Lane· Kong. Banaji United States v. 'Taveras, 424 RSupp.2d 446 (E.D. N.Y. 2006) f Vanman E], SaItz]L, Nathan LR, Warren]A. 2004. Racial discrimination by low-prejudiced whites. FadaI movements as implicit measures ofattitudes related to behavior. Psycho!. Sci. I 15:711-14 Village ofArlington Heights v. Metropolitan Housing Development Authority, 429 U.S. 252 (1977) I von Hippel W, Sekaquaptewa D, Vargas P. 1997. The linguistic intergroup bias as an implicit I indicator ofprejudice.]. Exp. Soc. Psychol. 33:490-509 Washington v. Davis, 426 U.S. 229 (1976) Wilson TD, Lindsey S, SchoolerTY. 2000. A model ofdual attitudes. Psychol. Rev. 107:101-26 Wittenbrink B. 2007. Measuring attitudes through priming. In Implicit Measures ofAttitudes, ed. B Wittenbrink, N Schwarz, pp. 17-58. New York: Guilford Zajonc R. 1980. Feeling and thinking: preferences need no inferences. Am. Psychol. 35:151-75 Ziegert]C, Hanges PJ. 2005. Employment discrimination: the role ofimpIicit attitudes, mo­ tivation, and a climate for racial bias.]. Appl. Psychol. 90:553-62

I•

'WWW.annuolreviews.org • Implicit Social Cognition and Law 451