critique

Stereotypic explanatory bias: Implicit stereotyping as a predictor of discrimination

Denise Sekaquaptewa, a,* Penelope Espinoza,

a Mischa Thompson,

a

Patrick Vargas, b and William von Hippel

c

a Department of Psychology, University of Michigan, Ann Arbor, MI 48109-1109, USA

b Department of Advertising, University of Illinois, IL, USA

c School of Psychology, University of New South Wales, Sydney, Australia

Received 7 March 2001; revised 11 February 2002

Abstract

Two experiments examined whether a measure of implicit stereotyping based on the tendency to explain Black stereotype-

incongruent events more often than Black stereotype-congruent events (Stereotypic Explanatory Bias or SEB) is predictive of be-

havior toward a partner in an interracial interaction. In Experiment 1 SEB predicted White males� choice to ask stereotypic questions of a Black female (but not a White male or White female) in an interview. In Experiment 2 the type of explanation

(internal or external attribution) made for stereotype-inconsistency was examined. Results showed that White participants who

made internal attributions for Black stereotype-incongruent behavior were rated more positively and those who made external

attributionswere ratedmore negatively by aBlackmale confederate. These results point to the potential of implicit stereotyping as

an important predictor of behavior in an interracial interaction.

� 2002 Elsevier Science (USA). All rights reserved.

Keywords: Stereotyping; Intergroup behavior; Prejudice; Interracial interaction

Stereotyping and prejudice are difficult to measure

because people are often unwilling to admit negative

attitudes and beliefs about social groups (Fazio, Jack-

son, Dunton, & Williams, 1995). Additionally, people

may sometimes be unable to accurately report on these

topics because how they think and feel about social groups may not be consciously accessible to them

(Greenwald&Banaji, 1995). Thus, researchers are faced

with a substantial ‘‘willing and able’’ problem when

attempting to measure prejudice and stereotyping.

In response to this ‘‘willing and able’’ problem, re-

searchers turned to measures of implicit prejudice and

stereotyping. Such measures are thought to tap con-

sciously inaccessible group-based attitudes and beliefs (Greenwald & Banaji, 1995). Most research attention

has focused on implicit prejudice measures, which are

intended to assess the degree of positivity or negativity

an individual implicitly associates with social groups

(e.g., Greenwald, McGhee, & Schwartz, 1998). Some-

what less research attention has focused on implicit

stereotype measures (e.g., Wittenbrink, Judd, & Park,

1997), and implicit stereotyping, which we define as the

unintended influence of stereotypes on information

processing (cf. Brewer, 1996). In part, this focus on prejudice rather than stereotypes/stereotyping probably

emerged because prejudice has traditionally been

thought to be more consequential than stereotyping for

behavioral outcomes such as discrimination (Brigham,

1971; Stangor, Sullivan, & Ford, 1991).

To the extent that measures of implicit prejudice and

stereotyping assess important processes relevant to in-

tergroup attitudes and perceptions (von Hippel, Se- kaquaptewa,&Vargas, 1995, 1997), it seems reasonable

to expect them to relate to intergroup behavior. Yet

such demonstrations are rare. In one study, White

participants who implicitly favored Whites over

African-Americans were rated by observers as having

more positive interactions with a White than a Black

experimenter (McConnell & Leibold, 2001; see also

Journal of Experimental Social Psychology 39 (2003) 75–82

www.elsevier.com/locate/jesp

Journal of Experimental Social Psychology

* Corresponding author. Fax: 1-734-647-9440.

E-mail address: [email protected] (D. Sekaquaptewa).

0022-1031/02/$ - see front matter � 2002 Elsevier Science (USA). All rights reserved. PII: S0022-1031(02)00512-7

Fazio et al., 1995). Similarly, Whites high in implicit prejudice showed greater indications of anxiety (e.g.,

eyeblinking) when interacting with a Black partner

(Dovidio, Kawakami, Johnson, Johnson, & Howard,

1997). The dearth of studies on the predictive utility of

such measures (particularly on implicit stereotyping,

concerning which no studies could be located) indicates

the need for more experimental investigations in this

area. If stereotyping is defined as the use of stereotypes in

information processing (Brewer, 1996), then there are

almost as many ways to operationalize implicit stereo-

typing as there are known stereotypic biases. Of these

biases, we focused on an explanatory bias adapted from

Hastie (1984) and introduced in our earlier work (von

Hippel et al., 1997,Experiment 3).This explanatorybias

emergeswhenone ismore likely to provide explanations for behaviors that are inconsistent with expectancies

than for behaviors that are consistentwith expectancies.

For example, if one expects an individual ‘‘James’’ to be

unintelligent, learning that ‘‘James received anAon the

test’’ may instigate explanatory processing, in an at-

tempt to make sense of the incongruity (‘‘. . .because it was an easy test’’). Learning that ‘‘James received a D

on the test,’’ on the other hand, is unlikely to instigate explanatory processing. Because such expectancies can

be based on stereotypes, an explanatory bias can emerge

in response to stereotype-inconsistency as well. To the

extent that this stereotypic explanatory bias (SEB) re-

flects the unintended influence of stereotypes on pro-

cessing, it is well-suited to the goal ofmeasuring implicit

stereotyping. 1

If implicit stereotyping as indicated by SEB reflects differences in the way perceivers process stereotype-

relevant information, then it seems likely that people

who vary in SEB would react differently during inter-

actions with stereotyped individuals. People who show

SEB should tend to selectively discount counter-stereo-

typic behaviors from Blacks, and thereby behave in a

more negative manner when interacting with a Black

person. According to this logic, we predicted that re- spondents who show SEB would display discriminatory

behavior towards aBlack but not aWhite individual. In

the lab, discriminatory behaviors are likely to be rela-

tively subtle, involving nonverbal behaviors (Dovidio et

al., 1997) or behavioral choices that are not clearly as-

sociatedwithdiscrimination.For example,Rudmanand

Borgida (1995) identified sexist behavior in men who

chose to ask subtly sexist/stereotypic questions in an interview context. Experiment 1 was conducted using a

similar procedure adapted for interracial interaction.

White male participants engaged in a mock job inter- viewwith either aWhite orBlack interviewee (actually a

research assistant). Participants were given a list of

Black stereotypic and neutral questions to select for use

in their interview. It was predicted that White partici-

pants who showed SEB would tend to choose stereo-

typic questions to ask of a Black but not a White

applicant.

Because prejudicial attitudes have been shown to be better predictors of behavior than endorsement of ste-

reotypes (e.g., Brigham, 1971), the Modern Racism

Scale (MRS: McConahay, Hardee, & Batts, 1981) was

also administered. We predicted that SEB would ac-

count for variance in behavioral choice beyond that

explained by the MRS. Because implicit measures are

frequently unrelated to more traditional, direct mea-

sures of racial attitudes (Dovidio et al., 1997;Kawakami & Dovidio, 2001; von Hippel et al., 1997), we expected

that the SEB measure would not be correlated with the

MRS.

Experiment 1

Method

Participants

Fifty-five White males participated in partial fulfill-

ment of psychology course requirements.

Materials

Stereotypic explanatory bias was assessed by pre-

senting participants with a series of 25 sentence beginnings, 16 of which were designed to measure re-

sponses to Black stereotype-consistent behaviors (e.g.,

easily made the team) and Black stereotype-inconsis-

tent behaviors (e.g., got a job at Microsoft). Behaviors

were paired with 50% male and 50% female African-

American (e.g., Marcellus, Lakisha) and White names

(e.g., Adam, Deborah). Nine race-neutral behaviors

(e.g., Linda ate a sandwich) were also included as filler items. It was necessary to include White targets in the

measure to ensure that participants were responding to

the com- bination of the target�s race and the race stereotypicality of the behavior, as opposed to only the

behavior itself. Additionally, the SEB items included

both positive and negative Black stereotypic behaviors

(i.e., easily made the team; blasted loud music in his

car) and positive and negative counter-stereotypic be- haviors (i.e., enrolled at Princeton; refused to dance).

Participants were asked to add words to the end of the

sentence stem in any manner that created a grammat-

ically correct sentence (see Hastie, 1984). SEB is evi-

denced by providing more explanations for Black

targets engaging in Black stereotype-inconsistent than -

consistent behaviors.

1 In a pretest SEB was demonstrated in students enrolled in a

course on stereotyping and was uncorrelated with explicit stereotype

endorsement, rð59Þ ¼ �:005, p ¼ :97, supporting the idea that the bias is implicit and distinct from conscious stereotype endorsement.

76 D. Sekaquaptewa et al. / Journal of Experimental Social Psychology 39 (2003) 75–82

The Modern Racism Scale (a seven-item self-report measure) was also administered.

Procedure

Under the guise of a study on interview skills, par-

ticipants were asked to engage in a mock job interview

with a research assistant (RA), said to be in training to

make a video on interviewing skills. It was explained

that it was the participants� role to help with this training by providing the opportunity for the RA to

interact with someone unfamiliar to him or her. It was

also explained that the interview skills sessionwouldnot

take long, so that after itwas over the participantwould

take part in an unrelated survey on social attitudes to

complete the hour. Participants were randomly assigned

to interact with one of three RAs.

The RAs were a White male, a White female, and a Black female. Both a White and Black female RA were

used to assess whether the SEB measure predicts be-

haviors specifically toward members of the targeted

stereotyped group (African-Americans) as opposed to

other stereotyped outgroup members (women).

After being introduced, the participantwas separated

from theRA.Theparticipantwas told hewould serve in

the role of interviewer and the RA would be the inter- viewee, and that the job the RA was interviewing for

was a restaurant supervisor position. Participants were

given a series of interview questions from which they

were to choose one option from each of 14 pairs. In-

terleaved among eight filler question pairs were six

question pairs designed so that the participant would be

forced to choose between mildly stereotypic or nonste-

reotypic question wording (see Rudman & Borgida, 1995).

2 For example, participants were asked to choose

between the questions ‘‘. . .some people think they can get awaywith stealing food, silverware, even cash.Have

you ever had any trouble like this?’’ (stereotypic), and

‘‘. . . some people think they can get away with taking work supplies home. Have you ever experienced it, and

what did you do about it?’’ (nonstereotypic). Partici-

pants were allowed to look over the question pairs, and then the interview began.

As the participant chose a question and asked it of

the RA, he made a checkmark on a sheet indicating

which question in each pair he selected. The RA re-

sponded by giving answers that were scripted to be

pleasant but neutral in tone, sufficient to answer either

the stereotypic or nonstereotypic questions, and were

identical across the race and gender of the RA. After this interview, theRA left the roomwhile theparticipant

rated the RA on interview skills (skilled interviewer, capable, and confident).

Next, the experimenter explained that the RA would

now serve as the interviewer and the participant as the

interviewee, in order for the RA to practice asking as

well as answering questions. It was explained to the

participant that he should ‘‘role play’’ and answer the

questions as if it were an actual job interview, and that

his responses were not important to the exercise. The RA then asked the subject a series of open-ended

questions related to a restaurant job. The RA then left

the roomagainwhile theparticipant completed a second

questionnaire asking about the RA�s skills at asking questions. During this time, the RA completed a ques-

tionnaire rating his/her overall impression of the par-

ticipant (liking of the participant, friendliness of the

participant), in order to test whether the RA�s impres- sion of the participant would be predicted by SEB

(similar to Fazio et al., 1995).

After the participant finished rating the RA, the ex-

perimenter explained that the interview skills study was

over, and the experimenter and theRA thanked the par-

ticipant. The experimenter then escorted the participant

toanother roomtotakepart inan ‘‘intercollegiate survey

onattitudesabout socialgroups.’’Within this surveywas the SEBmeasure followed by theMRS.Finally, partici-

pantswere probed for suspicion and fully debriefed.

Results

Scoring SEB

The sentence completions that participants provided

on the SEB measure were coded by two independent judges as to whether they explained the behavior in the

sentence beginning or simply continued the sentence

without explaining. Because the judgments of the two

coders were highly correlated (r ¼ :91, p < :001Þ, these ratings were collapsed across judges.

An SEB scorewas derived by subtracting the number

of explanations provided forBlack stereotype-consistent

events (Black actors engaging in stereotypically Black behaviors) fromthenumberofexplanationsprovidedfor

Black stereotype-inconsistent events (Black actors en-

gaging in counter-stereotypic or stereotypically White,

behaviors; i.e., BW–BB, where the first letter indicates

the race of the target and the second letter indicates the

whether the behavior is Black or White stereotypic).

Higher positive scores indicated greater Black SEB (B-

SEB).B-SEBscores ranged from )2.50 to2.50, M ¼ :77, SD ¼ 1:06, and differed significantly from zero, tð54Þ ¼ 5:40, p < :001. A second SEB score was calculated in order to test

whether Whites� behaviors toward a Black interaction partner is predicted by the processing of Black as op-

posed to White targets engaged in stereotype-incongru-

ency. This second SEB score was derived by subtracting

2 An independent sample of 20 White undergraduates rated each

questionon stereotypicality, ona scale from1 (notat all stereotypic) to

10 (very stereotypic). Stereotypic questions ðM ¼ 4:44Þ were rated significantly higher than neutral questions (M ¼ 2:15Þ, tð19Þ ¼ 6:24, p < :001.

D. Sekaquaptewa et al. / Journal of Experimental Social Psychology 39 (2003) 75–82 77

the number of explanations provided for White stereo- type-consistent events (Whites engaging in stereo-

typically White behaviors) from the number of

explanations provided forWhite stereotype-inconsistent

events (Whites engaging in counter-stereotypic behav-

iors; i.e., WB–WW). Higher positive scores indicated

greater White SEB (W-SEB). W-SEB scores ranged

from )2.50 to 1.50, M ¼ �:67, SD ¼ :98, and differed significantly from zero, tð54Þ ¼ �5:08, p < :001.

Scoring MRS

Response choices on the MRS ranged from 1

(strongly disagree) to 7 (strongly agree), and were

summed such that higher scores indicated higher mod-

ern racism. Scores on the MRS ranged from 7.00 to

36.00, M ¼ 18:84, SD ¼ 5:92.

Question choice

The number of stereotypic questions asked ranged

from0 to 6. In order to test the primary hypothesis that

B-SEB would predict question choice with a Black but

not aWhite interviewee, question choicewas analyzed in

a simultaneous multiple regression, using B-SEB, MRS

and type of RA (White male, White female, and Black

female) and interactions as predictor variables. Results showed that the B-SEB by RA type interaction term

accounted for significant variance in question choice,

b ¼ :34, p < :03, whereas the MRS by RA type inter- action term did not, b ¼ :01, p ¼ :96. Within cell re- gression analyses showed that B-SEB significantly

predicted question choice for the Black female RA, but

not for the White male or White female RA conditions

(see Table 1). In order to compare B-SEB toW-SEB, an additional

multiple regression analysis was conducted including

W-SEB in themodel.Results showed that theB-SEBby

RA type interaction term remained a significant pre-

dictor of question choice, b ¼ :32, p < :05. The W-SEB by RA type interaction term was not significant,

b ¼ :17, p ¼ :24.3 No significant effects of B-SEB, W-SEB, or MRS

emerged in participants� ratings of the RA or the RA�s impression of participants and none of these variables

were reliably correlated with each other.

Discussion

Consistent with predictions, Experiment 1 revealed

that the tendency to engage inSEB (i.e., toprovidemore explanations for Black stereotype-inconsistency than

Black stereotype-consistency) predicted the number of

stereotypic questions White male participants chose to

askwhile interviewing aBlack femaleRA. SEBwas not

predictive of this question choice for interviews with

either a White male or White female RA. Scores on the

Modern Racism Scale did not predict this question

choice with any of the RAs, nor did it predict the

competence ratings of the RA by the participant, and

the RA�s final impression of the participant. SEB also did not predict these evaluations.

In light of previous research showing that White participants high in implicit prejudicewere disliked by a

Black female experimenter (Fazio et al., 1995; see also

McConnell & Leibold, 2001), it was somewhat surpris-

ing that SEB failed to predict the RA�s evaluations of the participant. It is possible, however, that the inter-

action in the current experiment may have been suffi-

ciently constrained by forcing participants to play the

relatively scripted roles of interviewer and interviewee that there was insufficient latitude in behavior for the

predicted effect to emerge. Impression ratings might be

predicted by SEB in a more spontaneous, unscripted

social interaction.

Experiment 2

AlthoughExperiment 1 provided initial evidence that

SEB can predict behavioral outcomes, there are several

questions left unanswered. Because SEB did not predict

impressions of the participants, perhaps because the

interactionwas too scripted, the first goal ofExperiment

2 was to provide a less constrained interaction setting.

Second, in Experiment 1 SEB was found to predict be-

havior beyond that predicted by theMRS, but theMRS was not itself an effective predictor. It may be that SEB

predicted unique variance in behavior because it was

paired with an explicit rather than an implicit prejudice

measure. Therefore, Experiment 2 included an implicit

prejudice measure (the Implicit Associations Test, or

IAT; Greenwald et al., 1998).

Finally, and perhaps most importantly, in Experi-

ment 1 responses on the SEB measure were coded sim- ply as whether they provided an explanation for the

Table 1

QuestionchoicepredictedbyB-SEBandMRSforBlack female,White

male, and White female interviewees (Experiment 1)

b p r p

(a) Black female RA

SEB .521 .05 .53 .02

MRS .216 .42 .21 .23

(b) White male RA

SEB ).257 .36 ).20 .22 MRS .167 .54 .12 .32

(c) White female RA

SEB ).309 .20 ).38 .04 MRS ).094 .69 ).22 .16

3 An analysis without B-SEB also showed that W-SEB was not a

significant predictor.

78 D. Sekaquaptewa et al. / Journal of Experimental Social Psychology 39 (2003) 75–82

behaviors. Not all explanations are equally potent dis- missals of the implications of a behavior, however, as

some types of explanations are likely to be more ste-

reotype-maintaining than others. In particular, attribu-

tions to the situation rather than the actor may be

particularly stereotype-maintaining. For example, re-

sponding to, ‘‘Shaniqua scored high on the SAT. . .’’ by adding ‘‘because she took preparation courses’’ main-

tains the stereotype thatBlacks areunintelligentmore so than the explanation ‘‘because she is smart.’’When SEB

involves primarily internal attributions for stereotype-

inconsistency, SEB may actually promote stereotype

change or reduction, whereas when SEB involves pri-

marily external attributions, it is likely to lead to ste-

reotype maintenance. If SEB predicts discriminatory

behavior due to its stereotype-maintaining properties, a

formof SEB that instead promotes stereotype reduction may actually predict positive interracial interactions.

Therefore, in Experiment 2, the coding of the SEB

measure included categorizationof internal and external

explanations, with the prediction that external SEBmay

be more related to negative behavioral outcomes than

internal SEB.

Method

Participants

Seventy-nine White participants (27 male and 52 fe-

male) participated in partial fulfillment of psychology

course requirements.

Procedure

Upon reporting to the lab, the participant was shown to a cubicle and led to believe other students

occupied the remaining cubicles. It was explained that

participants were to be paired with another participant

for an experiment on game-playing, but that the task

would not take long, so they would also take part in

an unrelated survey on social attitudes to complete the

hour. Participants were randomly assigned to interact

with either a Black male or a White male confederate. The participant and confederate were escorted to a

larger room for the game-playing task. The participant

and confederate played three rounds of a paper-and-

pencil game similar to an extended version of tic-tac-

toe. Each round was terminated after 3min regardless

of whether one person had won. The confederate was

instructed to not try to win, nor initiate conversation,

but to respond cordially if the participant initiated conversation. The experimenter recorded the outcome

of each round of the game. Between the second and the

third round, the experimenter left the room for 2min,

presumably to attend to other students. This provided

an opportunity for the participant to have an un-

structured interaction and possibly initiate conversa-

tion with the confederate.

After the final round of the game, the participant returned to the cubicle to complete a filler questionnaire

regarding strategies used during the game. During this

time, the confederate completed the primary dependent

measures. The confederate rated his impression of the

participant, using the following items: I liked the par-

ticipant; the participant was nice; was friendly; was

selfish; was cold towards me; didn�t want to play the game with me (the final three were reverse scored). The confederate also rated the participant on the positive

behaviors he saw the participant display during the in-

teraction, using the following items: the participant

looked me in the eye; spoke to me before or during the

game; maintained a closed posture by crossing his/her

arms (reverse scored).

After completing the filler questionnaire, participants

were instructed to begin the second study as in Experi- ment 1. The survey packet instructed the participant to

complete the SEB measure, then use the computer

located in the cubicle for the IAT (the race versionof the

IAT was given as in Greenwald et al., 1998), and when

finished to complete the rest of the survey packet, which

contained the MRS. 4 Participants were then probed for

suspicion, debriefed, and thanked.

Results

The IAT computer program recorded reaction times

in a categorization task wherein participants responded

to Black andWhite names and pleasant and unpleasant

words. Outliers in the IAT data were trimmed, and the

data log transformed prior to analysis, according to the

procedure described by Greenwald et al. (1998). Aver- aged reaction times to unpleasant words paired on the

same response key as Black names, and pleasant words

paired on the same response key as White names, were

subtracted from averaged reaction times to unpleasant

words paired on the same response key asWhite names,

and pleasant words paired on the same response key as

Black names.Higher positive difference scores indicated

more negative associations to Blacks and/or more po- sitive associations to Whites (see Greenwald et al.,

1998). IAT difference scores ranged from )174.00 to 879.00, M ¼ 315:14, SD ¼ 184:37, and differed signifi- cantly from zero, tð78Þ ¼ 15:29, p < :001. The SEB measure was identical to that used in Ex-

periment 1 and responses were again scored as to whe-

ther they explained the behavior in the sentence stem.

(As in Experiment 1, W-SEB was computed and found

4 Because responding on the SEB measure may be altered when

participants report racial attitudes first, whereas responding on the

IAT seems largely uncontrollable (Kim & Greenwald, 1998), the SEB

was given first, followedby the IAT, and theMRSwas given last, as it

is apparent to participants that it measures racial prejudice (Fazio

et al., 1995).

D. Sekaquaptewa et al. / Journal of Experimental Social Psychology 39 (2003) 75–82 79

not to relate to any dependent measures, so results are reported for B-SEBonly.) In addition, responses judged

to be explanations were further scored as internal (i.e.,

dispositional) or external (i.e., situational) explanations.

As such, two separate SEB scoreswere derived from the

measure.An external SEB score (E-SEB)wasderivedby

subtracting the number of external explanations pro-

vided for Black stereotype-consistent events from the

number of external explanations provided for Black stereotype-inconsistent events. An internal SEB score

(I-SEB) was derived by subtracting the number of in-

ternal explanations provided for Black stereotype-con-

sistent events from the number of internal explanations

provided for Black stereotype-inconsistent events.

E-SEB scores ranged from )2.00 to 1.00, M ¼ �:11, SD ¼ :49, and differed significantly from zero, tð78Þ ¼ �2:05, p < :05. I-SEB scores ranged from )3.00 to 1.50, M ¼ �:25, SD ¼ :79, and differed significantly from zero, tð78Þ ¼ �2:78, p < :01.5 Higher scores on E-SEB and I-SEB indicate that more explanations (ex-

ternal or internal) were generated for Black stereotype-

inconsistency than consistency. Additionally, for some

analyses, a difference score was created by subtracting

E-SEB from I-SEB. Positive values on this SEB differ-

ence score indicated that the participant showedmore I- SEB than E-SEB.

TheMRSwas scoredas inExperiment 1.MRSscores

ranged from 7.00 to 34.00, M ¼ 13:22, SD ¼ 5:00.

Ratings of the interaction by the Black and White

interaction partners

Theconfederates� responses to the six items indicating his feelings toward the participant (I liked the partici- pant, the participant was friendly, etc.) and the three

items indicating his observation of the nonverbal be-

haviors of the participant (eye contact, initiating con-

versation, and closed posture) were combined to form a

social interaction score (Cronbach�s a ¼ .86). Confeder- ates� ratings on thismeasure didnot differ byparticipant gender, F 0s < 1, so data were collapsed across gender. The social interaction scores were analyzed in a

multiple regression analysis, using confederate race,

SEB difference score, IAT, MRS, and interactions as

predictor variables. Results showed that the SEB dif-

ference score by confederate race interaction termwas a

significant predictor, suggesting that amongparticipants

who interacted with a Black confederate, those who

showed more E-SEB received lower social interaction

scores, whereas those who showed more I-SEB received higher social interaction scores.The IAT by confederate

race and MRS by confederate race interaction terms were not significant (see Table 2).

In order to further examine how I-SEB and E-SEB

predicted the nature of the interaction differentially for

Black and White confederates, two separate regression

analyseswere conducted including the I-SEBandE-SEB

scores separately.These analyses showed that theE-SEB

by confederate race interaction term was significant,

b ¼ �:23, p < :05, and the I-SEB by confederate race interaction was also significant, but in the opposite di-

rection, b ¼ :28, p < :05.Within cell regression analyses showed that in the Black confederate condition, E-SEB

predicted more negative social interaction scores,

b ¼ �:67, p < :001, and I-SEB predicted more positive social interaction scores, b ¼ :33, p < :05. These scores were not significant predictors of social interaction

scores in the White confederate condition, b ¼ �:13, p ¼ :43, and b ¼ �:24, p ¼ :17, for E-SEB and I-SEB, respectively.

Correlations between SEB, IAT, and MRS

I-SEB and E-SEB were not significantly correlated

with the IAT, r ¼ :01, p ¼ :96, and r ¼ :15, p ¼ :19, respectively. E-SEBwas significantly correlatedwith the

MRS, r ¼ :22, p < :05, and I-SEB was not, r ¼ :19, p ¼ :10. The MRS was not significantly correlated with the IAT, r ¼ :16, p ¼ :15. I-SEB and E-SEB were sig- nificantly negatively correlated, r ¼ �:23, p < 05.6

Discussion

The results of Experiment 2 support the idea that

stereotypic explanatory bias canpredict the nature of an interracial interaction. Regression analyses showed that

E-SEB and I-SEB were significant predictors of the

nature of an interaction betweenWhite participants and

their Black partner, in opposite directions. Black con-

federates rated their social interaction with White par-

ticipants more negatively when the participants

provided external attributions for Black stereotype-

inconsistency,whereas they rated their social interaction with White participants more positively when the par-

ticipants provided internal attributions for Black ste-

reotype-inconsistency. White confederates did not differ

in their social interaction ratings depending on I-SEBor

E-SEB. IAT and MRS scores did not predict social in-

teraction ratings.

The results of Experiment 2 extend the results of

Experiment 1, by showing that the type of explanations generated for Black stereotype-inconsistency is a critical

5 In contrast toExperiment 1, average SEB scoreswere negative in

Experiment 2, indicating thatparticipants showed lowerSEBoverall in

this study. This unexpected result may be limited to this particular

sample; however, the possible diminishing effect of a spontaneous

interracial interaction on SEB should be addressed in future research.

6 Because IAT and MRS were administered after the social

interaction took place, it is possible that they failed to predict social

interaction scores because prior interaction with a Black individual

altered responding on these measures. While this is more likely

regarding the MRS than the IAT (see 4 ), this is still a possibility.

80 D. Sekaquaptewa et al. / Journal of Experimental Social Psychology 39 (2003) 75–82

factor in predicting the nature of an interracial interac-

tion. Attributing Black counter-stereotypic behavior to

external forces (e.g., ‘‘Marcellus got a job at Micro-

soft. . . because he knew someone there’’) predicts neg- ative interactions, whereas attributing such behavior to

internal traits or abilities (e.g., ‘‘. . . because he�s good with computers’’) predicts positive interactions. These

results suggest that providing external attributions for stereotype-inconsistency has greater stereotype-main-

taining properties than providing internal attributions

(cf. Crocker, Hannah, & Weber, 1983; Deaux, 1976).

General discussion

These experiments explored the idea that implicit stereotyping can predict behaviors in interracial inter-

actions. A paper-and-pencil measure of implicit stereo-

typingwas developed, based on the tendency to provide

explanations for Black stereotype-inconsistency (the

stereotypic explanatory bias, or SEB). In Experiment 1,

SEB predicted whether White male participants asked

stereotypic questions of a Black (but not a White) in-

terviewee in amock job interview. InExperiment 2, SEB derived from external attributions predicted negative

social interactions between White participants and

Black confederates, as indicated by the ratings of the

interaction by the confederate. SEB derived from in-

ternal attributions, on the other hand, predicted positive

social interactions between White participants and

Black confederates. The IAT and MRS did not predict

these outcomes. These experiments suggest that the tendency to spontaneously explain stereotype-inconsis-

tency can have a significant relationship with how an

individual behaves during interactions with members of

the stereotyped group, and that this relationshipmay be

independent of level of racial prejudice.

An interesting feature of SEB is its ability to predict

not only negative behavioral outcomes, but positive

ones as well. Examining the locus of the attribution generated in response to stereotype-inconsistency (in-

ternal vs. external) allowed the directional prediction of

behavior in relatively unstructured social interactions in

Experiment 2. However, in Experiment 1, stereotypic

question choice was predicted by an assessment of SEB

that did not differentiate between internal and external

explanations. Does locus of explanation differentially

predict outcomes such as stereotypic question choice?

To address this question, we conducted supple-

mentary analyses of the SEB data from Experiment 1.

Explanations were re-coded as internal or external and

I-SEB and E-SEB were tested as predictors of ste-

reotypic question choice. Results showed that neither

the I-SEB nor the E-SEB by confederate type inter- action terms were significant, although the direction of

the betas indicated that I-SEB and E-SEB were both

related to asking more stereotypic questions of a

Black than a White confederate (I-SEB b ¼ :21, p ¼ :24 and E-SEB b ¼ :25, p ¼ :14Þ. Although our explanation of this is entirely post hoc,

it seems possible that I-SEB and E-SEB predict in the

same direction in Experiment 1 but in different direc- tions in Experiment 2, because of the difference in the

dependent variable in the two studies. I-SEBandE-SEB

seem to predict behavioral outcomes differently when

the outcome is based on impressions that are presum-

ably heavily dependent on interpretations of another�s behavior (for example, liking, friendliness of partner;

Experiment 2). But for outcomes that are not, or are

more ‘‘cold’’ (e.g., choosingmildly stereotypic questions in Experiment 1), I-SEB and E-SEB may work in the

same manner. This hypothesis should be further ex-

plored in future research.

In sum, the current findings provide a demonstra-

tion that measures of stereotyping can predict inter-

group behavior. To our knowledge, this is the only

report documenting the relationship between implicit

stereotyping and behavior. Thus, these results may help establish the predictive validity of implicit ste-

reotyping measures. In addition, these findings add to

the literature on the relationship between stereotypes

and discrimination, by providing evidence that mea-

sures of stereotyping can predict discriminatory be-

havior when the measures tap implicit stereotypic

processing.

Acknowledgments

The authors wish to thank Oscar Ybarra, James

Hilton, and James Jackson for helpful comments on

earlier drafts of this manuscript. Experiment 2 was

Table 2

Social interaction score predicted by SEB difference score, IAT, and MRS for White and Black confederate (Experiment 2)

b p r p

SEB difference score�confederate race .32 .005 .32 .002 IAT�confederate race .05 .65 .03 .41 MRS�confederate race ).005 .96 .01 .48 Note. Positive values on the SEBdifference scores indicate that the respondent showedmore internal than external SEB. Therefore, participants

who showedE-SEB had more negative and those who showed I-SEB had more positive interactions with a Black confederate. Either form of SEB

was not related to social interactions with a White confederate.

D. Sekaquaptewa et al. / Journal of Experimental Social Psychology 39 (2003) 75–82 81

conducted as part of a master�s thesis written by the second author.

References

Brewer, M. B. (1996). When stereotypes lead to stereotyping: The use

of stereotypes in person perception. In C. N. Macrae, C. Stangor,

& M. Hewstone (Eds.), Stereotypes and stereotyping. New York:

Guilford Press.

Brigham, J.C. (1971).Racial stereotypes, attitudes, and evaluations of

and behavioral intentions towards Negroes and Whites. Sociom-

etry, 34, 360–380.

Crocker, J., Hannah, D. B., & Weber, R. (1983). Person memory and

causal attributions. Journal of Personality and Social Psychology,

44, 55–66.

Deaux, K. (1976). Sex: A perspective on the attribution process. In J.

H. Harvey, W. I. Ickes, & R. F. Kidd (Eds.), New directions in

attribution research (Vol. 1). Hillsdale, NJ: Erlbaum.

Dovidio, J., Kawakami, K., Johnson, C., Johnson, B., & Howard, A.

(1997). On the nature of prejudice: Automatic and controlled

processes. Journal of Experimental Social Psychology, 33, 510–540.

Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995).

Variability in automatic activation as an unobtrusive measure of

racial attitudes: A bona fide pipeline? Journal of Personality and

Social Psychology, 69, 1013–1027.

Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition:

Attitudes, self-esteem, and stereotypes. Psychological Review, 102,

4–27.

Greenwald, A. G., McGhee, D. E., & Schwartz, J. L. K. (1998).

Measuring individual differences in implicit cognition: The implicit

association test. Journal of Personality and Social Psychology, 74,

1464–1480.

Hastie, R. (1984). Causes and effects of causal attribution. Journal of

Personality and Social Psychology, 46, 44–56.

Kawakami, K., & Dovidio, J. F. (2001). The reliability of implicit

stereotyping. Personality and Social Psychology Bulletin, 27,

212–225.

Kim, D., & Greenwald, A. G. (1998). Voluntary controllability of

implicit cognition:Can implicit attitudes be faked?Paper presented

at the 70th annual meeting of the Midwestern Psychological

Association, Chicago, IL.

McConahay, J. B., Hardee, B. B., & Batts, V. (1981). Has racism

declined? It depends onwho�s asking andwhat is asked. Journal of Conflict Resolution, 25, 563–579.

McConnell, A. R., & Leibold, J. M. (2001). Relations among the

Implicit Association Test, discriminatory behavior, and explicit

measures of racial attitudes. Journal of Experimental Social

Psychology, 37, 435–442.

Rudman, L., & Borgida, E. N. (1995). The afterglow of construct

accessibility: The behavioral consequence of priming men to view

women as sexual objects. Journal of Experimental Social Psychol-

ogy, 6, 493–517.

Stangor, C., Sullivan, L. A., & Ford, T. E. (1991). Affective and

cognitive determinants of prejudice. Social Cognition, 9,

359–380.

vonHippel,W., Sekaquaptewa,D.,&Vargas, P. (1995).On the role of

encoding processes in stereotype maintenance. Advances in Exper-

imental Social Psychology, 27, 177–254.

von Hippel, W., Sekaquaptewa, D., & Vargas, P. (1997).

The Linguistic Intergroup Bias as an implicit indicator of

prejudice. Journal of Experimental Social Psychology, 33,

490–509.

Wittenbrink, B., Judd, C. M., & Park, B. (1997). Evidence for racial

prejudice at the implicit level and its relationship with question-

naire measures. Journal of Personality and Social Psychology, 72,

262–274.

82 D. Sekaquaptewa et al. / Journal of Experimental Social Psychology 39 (2003) 75–82