Group Processes &

Intergroup Relations
2007 Vol 10(3) 359–-372

Discrimination and
the Implicit Association Test
Laurie A. Rudman
Rutgers University
Richard D. Ashmore
Rutgers University

Prejudice researchers have been criticized for failing to assess behaviors that reflect overtly
hostile actions (i.e. racial animus; Arkes & Tetlock, 2004; Mackie & Smith, 1998). Two studies
sought to begin to fill this gap in the implicit literature by showing that scores on the Implicit
Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998) are linked to harmful intergroup
behaviors. In Study 1, the IAT predicted self-reported racial discrimination, including verbal
slurs, exclusion, and physical harm. In Study 2, the IAT predicted recommended budget cuts
for Jewish, Asian, and Black student organizations (i.e. economic discrimination). In each study,
evaluative stereotype (but not attitude) IATs predicted behaviors even after controlling for
explicit attitudes. In concert, the findings suggest that implicit stereotypes are more predictive
of overtly harmful actions than implicit attitudes in the intergroup relations domain.

keywords discrimination, implicit prejudice, implicit stereotypes, racial

stereotypes, social cognition, intergroup relations

Empirical observation distinguishes scientific The ingeniously simple concept underlying

psychologists from philosophers, novelists, the IAT is that tasks are performed well when
and priests, who are also purveyors of truths they rely on well-practiced associations between
about human nature. Because the quality of objects and attributes. In the attitude IAT,
our insights depends on the clarity of our tools, respondents categorize two classes of objects
methodological advances are the lifeblood of (e.g. dogs and cats) with both good and bad
psychology. When a method is introduced, it words (e.g. vacation vs. poison). An automatic
should rightly be the target of skepticism and preference for dogs is shown to the extent that
debate until its usefulness has been established.
This was the expected trajectory for the Implicit Author’s note
Association Test (IAT; Greenwald, McGhee, & Address correspondence to, Laurie A.
Schwartz, 1998), a response latency task that Rudman, Department of Psychology,
was developed to measure implicit attitudes, Tillett Hall, Rutgers, the State University of
but has since proved useful for assessing other New Jersey, 53 Avenue E, Piscataway,
constructs, including implicit stereotypes (e.g. New Jersey 08854-8040, USA
Greenwald, Pickrell, & Farnham, 2002). [email: rudman@rci.rutgers.edu]

Copyright © 2007 SAGE Publications

(Los Angeles, London, New Delhi and Singapore)
10:3; 359–372; DOI: 10.1177/1368430207078696
Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015
Group Processes & Intergroup Relations 10(3)

the pro-dog task (dogs + good/cats + bad) is & Banaji, 2001; Karpinski, Steinman, & Hilton,
performed faster and more accurately than 2005; Nosek, 2005).
the pro-cat task (dogs + bad/cats + good). For However, implicit researchers can be criticized
stereotype IATs, good and bad words are replaced —along with prejudice researchers in general—
with specific attributes associated with each for rarely assessing overtly hostile behaviors
object (e.g. loyal vs. aloof). If the stereotype (Mackie & Smith, 1998). This oversight affords
congruent task (dogs + loyal/cats + aloof) is an opening for implicit social cognition critics.
performed faster and more accurately than the In particular, the IAT’s predictive utility in the
stereotype incongruent task (dogs + aloof/cats + prejudice domain has been questioned by Arkes
loyal), an implicit stereotype is shown. and Tetlock (2004)—along with behavioral
Although still young, the IAT has been tested data for other implicit measures (e.g. Dovidio,
in over 100 studies—far more so than any other Kawakami, & Gaertner, 2002; Fazio, Jackson,
response latency technique. A recent meta- Dunton, and Williams, 1995)—on the basis that
analysis (Poehlman, Uhlmann, Greenwald, & some behaviors might reflect emotions other than
Banaji, 2004) supported the IAT’s temporal antipathy (e.g. nonverbal reactions, such as gaze
stability, internal consistency, and criterion or speech disfluencies, might indicate guilt or
validity (e.g. the IAT predicted voting, Scholastic anxiety). However, in Poehlman et al.’s (2004)
Aptitude Test [SAT] scores, and consumer meta-analysis of IAT findings, less than a third
choice). Most promisingly, the IAT predicted of the behaviors assessed were nonverbal (or
behaviors better than self-reports did when the otherwise ambiguous). Yet, as already noted, the
domain concerned prejudice and stereotypes. IAT was a better predictor of these behaviors,
The behaviors included target evaluations, compared with self-reports. Moreover, the role
hiring decisions, and pro-social indicators (both of anxiety in prejudicial responding has long
verbal and nonverbal), suggesting a wide range been recognized (e.g. Islam & Hewstone, 1993;
of utility for the IAT. Given that self-reported Stephan & Stephan, 1985). That is, people
prejudice was less useful when these behaviors may feel anxious in the presence of outgroup
were at stake, the IAT appears to be a promising members, but this does not mean they are
methodological advance. egalitarians. In fact, prejudice is typically defined
In addition, implicit associations behave as a negative orientation that can be expressed
in accord with classic attitude and intergroup as moving against (animus) or moving away from
theories (for a review, see Uhlmann & Poehlman, outgroup members (which can also reflect guilt
2005). For example, they are sensitive to context and anxiety; Ashmore, 1970).
and conditioning, just as attitudes and prejudice Nonetheless, although automatic biases have
are (Blair, 2002; Fazio & Olson, 2003). Further, been linked to negative judgments of Blacks
IAT scores have supported cognitive consistency (e.g. Jackson, 1997; Lambert, Payne, Ramsey, &
principles (Greenwald, Pickrell, & Farnham, 2002), Shaffer, 2005; Rudman & Lee, 2002), and
the contact hypothesis (Rudman, Ashmore, & female job applicants (Rudman & Glick, 2001),
Gary, 2001), aversive racism theory (Son Hing, Li, unambiguously harmful behaviors are seldom
& Zanna, 2002), social identity theory (Ashburn- investigated, whether explicit or implicit bias
Nardo, Voils, & Monteith, 2001), and system is assessed. The present research sought to
justification theory ( Jost, Pelham, & Carvallo, begin to fill this gap in the implicit literature.
2002; Rudman, Feinberg, & Fairchild, 2002). To do so, we focused on the IAT because it has
Finally, the relationship between implicit and borne the brunt of researchers’ criticisms. In
explicit attitudes can be characterized as hetero- Study 1, we assessed participants’ reports of
geneous (Blair, 2001; Fazio & Olson, 2003; their harmful actions toward Blacks in the past.
Nosek, Banaji, & Greenwald, 2002), but it is Behaviors consisted of both active harm (e.g.
moderated by theoretically expected variables, verbal insults and physical violence) and passive
including attitude strength, social desirability, harm (e.g. avoidance or exclusion). In Study 2,
and measurement error (Cunningham, Preacher, we measured people’s willingness to cut the

360

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

 Rudman & Ashmore discrimination and the iat

budget for student organizations designed to Table 1. Stimuli for implicit prejudice and stereotype
support Jews, Asians, and Blacks. If attitude and measures (Study 1)
stereotype IAT scores predict these behaviors, Attitude IAT Stereotype IAT
results would lend confidence to the IAT’s ability
to tap implicit prejudice, as opposed to ‘mere Pleasant Unpleasant Negative Positive
associations’ (Arkes & Tetlock, 2004, p. 268). words words traits traits

sunshine filth lazy ambitious

Study 1 smile death shiftless industrious
angel devil unemployed successful
Our primary aim was to examine the relation- luck slime hostile calm
ship between IAT-assessed biases and non- rainbow cancer dangerous trustworthy
Black participants’ self-reported harmful actions paradise hell threaten ethical
toward Blacks. To encourage candid responding, fortune poison violent lawful
we asked participants whether they had been Note: IAT stimuli were adopted from past research
the victim of each behavior prior to their report (Rudman et al., 2001).
of being a perpetrator. A secondary aim was to
compare the attitude and stereotype IATs as associated with Blacks (e.g. lazy, hostile) and
predictors of harmful discrimination. To do positive attributes associated with Whites (e.g.
so, we used evaluative stereotypes because they ambitious, calm) shown in columns 3–4.
are conceptually akin to prejudice (see also The IATs were adopted from and administered
Rudman et al., 2001; Wittenbrink, Judd, & Park, exactly as in past research (Rudman et al., 2001).
1997). Because the stereotype IAT consists of The order in which participants performed
both evaluative and cognitive associations, they the critical blocks was counterbalanced across
might capture the implicit prejudice construct subjects, as was the order in which participants
more completely (Breckler, 1984). Finally, we performed the IATs (these procedural vari-
included self-report measures as a means of ables did not influence results). The IAT effect
testing whether implicit associations can predict was computed so that high scores reflect greater
discrimination above and beyond explicit tendency to associate Blacks with negative versus
attitudes. positive attributes, compared to when these
associations were reversed. Scoring for the IAT
Method followed recent recommendations (Greenwald,
Participants Sixty-four volunteers (21 male, Nosek, & Banaji, 2003). Specifically, we used
43 female) participated to partially fulfill an the D statistic because it has been shown to be
introductory psychology course requirement less influenced by procedural variables (e.g.
(M age = 20). Of these, 52 were White (81%), counterbalancing).
6 were Asian American (9%), and 6 were Latino
(9%). Data from six participants showing Explicit measures
high error rates (> 25%) on the computer tasks Explicit attitudes Participants completed a
were eliminated, as were data from 13 African feeling thermometer and the Modern Racism
American participants. Scale (MRS; McConahay, 1986). The feeling
thermometer asked participants to indicate,
IAT measures The attitude IAT and stereotype separately for African American men and White
IAT each used 7 White male names (e.g. John, American men, the extent to which they felt
Andrew, Peter, Brad) and 7 Black male names positively toward each group (0 = extremely cold,
(e.g. Lamar, Malik, Rashan, Leroy) as the or unfavorable ; 99 = extremely warm, or favorable).
target concepts. Table 1 shows the remaining The difference between these measures was com-
stimuli. The attitude IAT used the pleasant puted such that high scores represented more
and unpleasant words shown in columns 1–2. positive attitudes toward White than Black men
The stereotype IAT used negative attributes (to mirror the IAT). The MRS consists of seven

361

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

Group Processes & Intergroup Relations 10(3)

items (e.g. ‘Blacks are getting too demanding in Because they served no other purpose, they are
their push for equal rights’), scored on a scale not further discussed.1
ranging from 1 (strongly agree) to 5 (strongly agree).
MRS scores were averaged (α = .82) so that high Procedure Volunteers participated individually
scores reflect more anti-Black attitudes. in separate cubicles. Measures were administered
using a computer program that presented items
Discriminatory behaviors To assess harmful randomly, within each measure. To increase par-
behaviors, we used a slightly shorter version of a ticipants’ confidence in their anonymity, they
past measure (Contrada et al., 2001). Participants generated their own identification number. They
were asked to report how often, over the course first completed the attitude and stereotype IATs,
of their lifetime, they had engaged in specific in counterbalanced order. They then completed
actions on a scale ranging from 1 (never) to 7 the explicit attitude and behavior measures, in
(very often). The verbal index averaged two counterbalanced order. Consistent with past
items pertaining to making ethnically offensive research, the order of administrating the IAT
comments and jokes, either in the presence and self-reports did not affect scores on either
of targets or behind their backs (r(62) = .56, implicit or explicit measures (e.g. Greenwald
p < .001) (M = 3.65, SD = 1.30). The defensive index et al., 2003).
averaged two items pertaining to avoiding or ex-
cluding others from social gatherings and organ- Results and discussion
izations because of their ethnicity (r(62) = .69, Preliminary analyses To assess internal consist-
p < .001) (M = 3.72, SD = 1.35). The offensive index ency, we correlated practice trials with critical
averaged three items pertaining to nonverbal trials within each IAT. The coefficient for the
hostility (e.g. giving ‘the finger’), and physically attitude IAT was reliable (r(62) = .69, p < .001),
hurting targets or their property (or threaten- as it was for the stereotype IAT (r(62) = .71,
ing to do so) because of their ethnicity, (α = .89; p < .001). Table 2 (fourth row) displays mean
M = 2.65, SD = .80). For each item, participants latencies for the implicit attitude and stereotyping
first indicated the extent to which they had been measures.2 On average, participants favored
the target of ethnic discrimination (e.g. ‘How Whites over Blacks on both IATs, resulting in large
often have you been the target of offensive com- effect sizes for both measures (attitude d = .75,
ments because of your ethnicity?’). The purpose stereotype d = .76). 3 By contrast, negligible
of these items was to encourage reporting prejudice was reported on the MRS and the
discrimination toward others (i.e. to justify thermometer index (i.e. on average, Whites
participants’ own behavior). As expected, these were not evaluated more favorably than were
items covaried with reports of verbal, defensive, Blacks; d = .09).
and offensive behaviors (all rs > .31, ps < .05).

Table 2. Summary statistics for implicit and explicit measures (Study 1)

Thermometer
Measure Attitude IAT Stereotype IAT MRS index

Stereotype IAT .33**

MRS .37** .20
Thermometer
index .25* .16 .32*
Mean .43 .28 1.86 2.20
SD .56 .37 .64 24.02

*p < .05; **p < .01.

Note: IAT results are displayed using the D statistic (Greenwald et al., 2003). High scores on the IATs and the
thermometer index reflect more positive evaluation of Whites compared with Blacks.

362

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

 Rudman & Ashmore discrimination and the iat

As seen in Table 2, the attitude and stereotype Table 3. Relations among intergroup orientations
IATs covaried (see also Rudman et al., 2001). The and discrimination (Study 1)
two explicit measures were also related.The Behavior Attitude Stereotype
attitude IAT was reliably related to the MRS measure IAT IAT MRS Thermometer
and to the thermometer. The stereotype IAT
was positively but unreliably related to the Verbal .41** .34** .37** .17
Defensive .15 .30* .51* .44**
explicit measures.
Offensive .12 .25* .53** .31*

Predicting discrimination Study 1’s main *p < .05; **p < .01.
objective was to test whether the IAT predicts un- Note: Correlations were computed using the
ambiguously harmful behaviors. As shown in D statistic (Greenwald et al., 2003). Correlations
Table 3, the attitude IAT covaried with verbal using log transformed latencies were similar.
discrimination (e.g. ethnic slurs and jokes),
Table 4. Predicting harmful discrimination from
whereas the stereotype IAT was related to each explicit and implicit measures (Study 1)
behavioral index (verbal, defensive, and offen-
sive). Finally, the MRS reliably covaried with Hierarchical
all three behavioral indexes, whereas the therm- regression
ometer index reliably covaried with defensive model Step β t R2 F∆
and offensive, but not verbal, discrimination. MRS 1 .49 4.51** .35 16.71**
The behavioral measures were robustly Thermometer
related (α = .80), and were therefore combined. index 1 .22 2.00*
A hierarchical regression analysis was then MRS 2 .42 3.86**
conducted to examine whether the IAT pre- Thermometer
dicts unique variance in discrimination, after index 2 .31 2.85**
Attitude IAT 2 .07 .65
accounting for explicit measures. Table 4 shows
Stereotype IAT 2 .33 2.98** .44 4.53*
the results. The discrimination index was re-
liably predicted by the stereotype (but not the *p < .05; **p < .01.
attitude) IAT, even after accounting for the Notes: Standardized regression coefficients are
MRS and the thermometer index, which also shown. IAT effects used in these analyses were based
contributed unique variance. on the D statistic (Greenwald et al., 2003).
In sum, Study 1’s focal results were the link-
ages shown between implicit associations and the attitude IAT was entered first, followed by
participants’ history of anti-Black discrimin- the stereotype IAT. In Step 1, the attitude IAT
ation. The behaviors predicted by the stereotype was a reliable predictor (β = .27, p < .05), but
IAT ranged from active harm (e.g. verbal slurs it was reduced to nonsignificance in Step 2
and personal and property violations) to more (β = .17, p = .16), suggesting that the stereotype
passive harm (e.g. exclusion and avoidance), IAT is a more effective predictor of harmful
whereas the attitude IAT predicted offensive actions (β = .28, p < .05).
comments and jokes (using bivariate analyses).
Although the attitude IAT did not contribute
Study 2
unique variance to the discrimination index
after accounting for explicit measures, this may In Study 2, we extended our analysis to include
be due to the stronger correlations between it economic discrimination against Jews, Asians,
and the direct measures, compared with the and Blacks. Data concerning each group were
stereotype IAT. Alternatively, the stereotype collected over three phases, during a time
IAT, because it combines beliefs with evaluation, period of approximately three months. Each
may be a superior measure of implicit bias. For investigation examined predictive utility for
exploratory purposes, we conducted a hierarch- IAT-assessed evaluative stereotypes (e.g. negative
ical regression on the behavioral index in which attributes associated with Jews and positive

363

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

Group Processes & Intergroup Relations 10(3)

attributes associated with Christians) vis-a-vis Asian-White phase Each IAT used six Asian
budget cuts for minority student organizations. surnames (e.g. Chang, Kwan, Yamashita) and six
As in Study 1, the MRS and feeling thermometers White names (e.g. Miller, Taylor, Johnson) as the
were included for comparison purposes. The target concepts. The attitude IAT was otherwise
Asian-White and the Black-White phases also identical to Experiment 1’s. The stereotype IAT
included an attitude IAT. used six negative Asian attributes (e.g. reserved,
Because some investigators have argued that stiff, inhibited) and six positive White attributes
the IAT is a measure of environmental asso- (e.g. warm, friendly, outgoing).
ciations (Karpinski & Hilton, 2001; Olson &
Fazio, 2004), we also included a direct measure Black-White phase The attitude and stereotype
of cultural favoritism. If IAT scores are related to IATs were adopted from Experiment 1.
perceptions that cultural stereotypes are more
positive for majority than minority groups, sup- Explicit measures
port for this reasoning will be shown. However, Explicit attitudes The thermometer index was
Nosek and Hansen (in press), using over 50 atti- identical to Experiment 1’s, with participants
tude objects (including groups based on ethni- indicating their feelings toward the appropriate
city, religion, sexual orientation, and gender) and groups in each phase. A difference score was
thousands of Web site respondents, consistently computed such that high scores indicated more
found a negligible link between cultural fav- positive evaluation of Christians compared with
oritism and the IAT. Therefore, we expected to Jews, Whites compared with Asians, or Whites
find a similar pattern. compared with Blacks. When necessary, the MRS
was modified by replacing Blacks with either
Method Jews or Asians as the target group (all αs > .84).
Participants All participants volunteered in The measure was scored such that high scores
exchange for partial fulfillment of their Intro- reflected more symbolic prejudice.
ductory Psychology course research requirement.
Only data from group members represented Cultural knowledge Participants were asked to
in the IAT were used in the analyses. In the rate how positive the cultural stereotypes of
each group represented in the IATs were on
Jewish-Christian phase, there were 89 volunteers
scales ranging from 1 (not at all) to 10 (extremely).
(64 Christians, 25 Jews). Of these, 37 were men
A difference score was formed so that high
and 52 were women. In the Asian-White phase,
scores reflected judging stereotypes about
there were 89 volunteers (59 Whites, 30 Asians).
majority groups as more positive than stereo-
Of these, 38 were men and 51 were women. In
types about majority groups for each phase
the Black-White phase, there were 126 volunteers
(e.g. Christians higher than Jews in the Jewish-
(89 Whites, 37 Blacks). Of these, 34 were men
Christian phase).
and 92 were women.
Economic discrimination Following past research
IAT measures
(Haddock, Zanna, & Esses, 1993; Zanna, 2004),
Jewish-Christian phase The stereotype IAT
participants completed a budget measure that
used six negative Jewish attributes (e.g. cheap, was presented as a survey conducted on behalf
controlling, dominating) and six positive of the Psychology Department (i.e. separate
Christian attributes (e.g. generous, charitable, from the main study), and was prefaced by the
friendly). Following past research (Rudman, following statement:
Greenwald, Mellott, & Schwartz, 1999), target
We have been asked to administer this short survey
concepts consisted of six Jewish surnames
as part of all of our research protocols this year, as
(e.g. Shapiro, Cohen, Katz) and six Christian sur- a means of gathering student opinion. The student
names (e.g. Miller, Taylor, Johnson). The attitude government has been forced to cut funding to
IAT was not administered in this phase. student organizations by 20%. We ask that you help

364

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

 Rudman & Ashmore discrimination and the iat

out by recommending which organizations listed for the Jewish-Christian phase (r(87) = .61,
below should have their funds decreased. Current p < .01); the Asian-White phase (rs(87) = .59
funding for each organization is listed in column 1. and .60 for the attitude and stereotype IATs,
Place your recommended funding in column 2. respectively, ps < .01); and the Black-White phase
Keep in mind that your suggestions should result in
(rs(124) = .57 and .63) for the attitude and
an approximately 20% decrease in funding. Please
place this survey in the box when you are through.
stereotype IATS, respectively, ps < .01. Table 5
The results of this survey will be presented to the shows descriptive statistics for each phase of
student government. Study 2’s data collection, as a function of group
membership (Christians compared with Jews,
Eight student organizations were then listed, Whites compared with Asians, and Whites
including the focal groups (Chabad Jewish compared with Blacks).
Student Organization, Japanese Cultural Asso- As seen in Table 5, the IATs showed the expected
ciation, and Blacks United to Save Themselves) pattern of known groups validity, accompanied
and five fillers (e.g. the PIRG organization, the by reasonably large group difference effect
drama club, and the marching band). Current sizes (all ds > .73), as did the explicit attitude
funding for the focal group in each phase was measures (all ds > .74). As in past research, Jews
listed as US$11, 500. The difference between and Asians showed reliable ingroup bias on
this and participants’ recommended funding both sets of measures (Rudman et al., 2002).
for that group was computed so that high scores Blacks demonstrated the typical pattern of
indicated greater budget cuts (i.e. economic showing weak implicit, but robust explicit,
discrimination). ingroup bias (e.g. Nosek et al., 2002). Not
surprisingly, majority groups (Christians and
Procedure Upon entering the lab, participants Whites) showed greater economic discrimin-
were escorted to a separate room and asked to ation, compared with minority groups (Jews,
complete the budget recommendation meas- Asians, and Blacks; all ds > .62). Finally, the
ure before participating in the ‘main study’. results of the cultural stereotype index (the
Participants placed their completed survey perceived tendency for society to view majority
(subtly coded with their identification number) group members more positively than minority
in a box marked ‘Psychology Department Survey’ group members) revealed that minority group
to enhance the cover story. Participants were then members tended to report a greater discrepancy
led to a private cubicle where they performed the than did majority group members. However,
attitude and stereotype IATs in counterbalanced this difference was reliable only in the Asian-
order (except in the Jewish-Christian phase, White phase (t(87) = 2.86, p < .01). In sum, there
when only stereotypes were assessed), as well as was general agreement among majority and
the explicit measures (in the order described minority group members concerning how the
above). The implicit and explicit measures were culture viewed their groups, but little agreement
administered in counterbalanced order. The concerning how the groups should be evaluated,
IATs were administered exactly as in Experiment 1 either implicitly or explicitly.
(e.g. with task order counterbalanced). The Table 6 shows the relationships among Study
effects of these procedural variables were non- 2’s variables, for each phase. As expected, the IAT
significant in each phase. Upon completion was unrelated to cultural stereotype index in each
of the measures, participants returned to the phase (Nosek & Hansen, in press). By contrast,
main room for a process debriefing. No subject the IATs and thermometer indexes were positively
expressed suspicion that the budget measure correlated (rs ranged from .27 to .53). The MRS
was part of the protocol. covaried with the Jewish-Christian stereotype IAT
and with attitude IAT scores in the Black-White
Results and discussion phase (echoing Study 1’s results). Finally, the
Preliminary analyses Internal consistency MRS tended to be negatively linked to cultural
analyses of the IATs revealed reliable coefficients stereotypes. That is, people who thought the

365

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

Group Processes & Intergroup Relations 10(3)

Table 5. Summary statistics for implicit and explicit measures (Study 2)

IAT Attitude Stereotype Therm Budget Cultural

measure IAT IAT MRS index index stereotypes
Jewish-Christian
Christians (n = 64) – .37 2.45 11.25 $2480 2.35
Jews (n = 25) – –.35 1.62 –12.91 $1200 3.04
Pooled SD – .44 .60 16.42 $1890 2.59
Group difference d – 1.64 1.38 1.47 .68 –.27

Asian-White
Whites (n = 59) .42 .32 2.31 7.97 $1895 2.24
Asians (n = 30) –.47 –.28 1.20 –14.17 $632 4.07
Pooled SD .63 .47 .64 17.58 $1981 2.94
Group difference d 1.41 1.28 1.74 1.25 .63 –.62

Black-White
Whites (n = 89) .39 .28 1.70 3.18 $1669 5.44
Blacks (n = 37) .04 –.17 1.37 –16.35 $517 6.03
Pooled SD .47 .41 .44 19.61 $1711 2.38
Group difference d .74 1.09 .75 1.02 .67 –.25

Notes: IAT results are displayed using the D statistic (Greenwald et al., 2003). For each measure, high scores
reflect greater bias against minority groups ( Jews, Asians, or Blacks) or greater perceived bias in society
(cultural stereotypes). The effect size (Cohen’s d) represents group differences. Conventional small, medium,
and large effect sizes are .20, .50, and .80, respectively (Cohen, 1988).

Table 6. Correlations among implicit and explicit measures (Study 2)

IAT measures Explicit measures

Cultural
Measure Attitude Stereotype Thermometer MRS stereotypes

Jewish-Christian
Thermometer index – .53**
Modified MRS – .39** .53**
Cultural stereotypes – .11 –.05 –.31**
Budget index – .38* .47** .11 .11

Asian-White
Stereotype IAT .28**
Thermometer index .43** .28**
Modified MRS .15 .05 .33**
Cultural stereotypes .02 –.07 –.08 –.12
Budget index .25* .30** .28** .16 .10

Black-White
Stereotype IAT .47**
Thermometer index .42** .27**
MRS .42** .14 .30**
Cultural stereotypes .02 .01 –.13 –.19*
Budget index .23* .18* .08 .05 –.03

*p < .05; **p < .01.

Note: IAT correlations were computed using the D statistic (Greenwald et al., 2003). Correlations using log
transformed latencies were similar.

366

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

 Rudman & Ashmore discrimination and the iat

culture favored majority groups with more reliably related to the budget measure in the
positive stereotypes than minority groups tended Jewish-Christian and Asian-White phases, but
to show less symbolic prejudice (especially toward not in the Black-White phase. By contrast, the
Jews and Blacks). This suggests that minority MRS did not covary with the budget index,
groups’ justice-seeking behaviors are supported irrespective of the attitude object (all rs < .17, ns).
when society is perceived as biased. Finally, the cultural stereotype index was not
a predictor of economic discrimination (all
Predicting discrimination Our primary aim rs < .12, ns).
was to examine the relationship between IAT A hierarchical regression analysis was con-
measures and economic discrimination. Table 6 ducted for each phase to examine whether the
shows that the stereotype IAT was reliably linked IAT predicts recommended budget cuts after
to the budget index in each phase. That is, people accounting for the thermometer index (the
who associated minority group members with MRS and cultural stereotypes were not included
negative attributes and majority group members because they showed weak predictive utility).
with positive attributes were also likely to re- Moreover, we controlled for group identity
commend budget cuts for the target minority (coded as 0 = majority, 1 = minority) to provide
group’s student organization. The attitude IAT a more conservative test (cf. Karpinski et al.,
performed similarly in the Asian-White and 2005). Table 7 shows the results. For the Jewish-
Black-White phases (it was not administered Christian phase, discrimination was reliably
in the Jewish-Christian phase). Thus, both the predicted by group identity, the thermometer
stereotype and attitude IAT predicted economic index, and the stereotype IAT. For the Asian-
discrimination. The thermometer index was White phase, discrimination was marginally

Table 7. Predicting economic discrimination from explicit and implicit measures (Study 2)

Hierarchical regression model Step β t R2 F∆

Jewish-Christian
Group identity 1 –.23 2.40* .34 11.53**
Thermometer index 1 .48 4.44**
Group identity 2 –.27 2.95**
Thermometer index 2 .33 2.58*
Stereotype IAT 2 .39 3.03** .41 9.21**

Asian-White
Group identity 1 –.14 1.17 .10 4.32*
Thermometer index 1 .21 1.73
Group identity 2 –.01 .10
Thermometer index 2 .22 1.83
Attitude IAT 2 .15 1.38
Stereotype IAT 2 .27 2.41* .17 3.61*

Black-White
Group identity 1 –.24 2.38* .09 5.93**
Thermometer index 1 .10 1.00
Group identity 2 –.17 1.55
Thermometer index 2 .08 .82
Attitude IAT 2 .01 .10
Stereotype IAT 2 .24 2.51* .14 3.33*

*p < .05; **p < .01.

Notes: Standardized regression coefficients are shown. IAT effects used in these analyses were based on the
D statistic (Greenwald et al., 2003). Group identity was dummy coded (0 = majority, 1 = minority).

367

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

Group Processes & Intergroup Relations 10(3)

predicted by the thermometer index (p = .07), in severity from racial jokes to budget cuts to
and reliably predicted by the stereotype IAT. For blatant aggression, but each is overtly discrimin-
the Black-White phase, only the stereotype IAT atory and therefore of consequence. Moreover,
contributed uniquely to discrimination. Thus, we included behaviors that can be characterized
the stereotype IAT remained a predictor even as unambiguously hostile (e.g. giving the finger
after controlling for explicit attitudes and group and physically harming an individual or their
identity in each phase. By contrast, the attitude property). Although we cannot rule out the pos-
IAT did not account for unique variance, and sibility that these behaviors refl ected other
the thermometer index was only significant in emotions (such as fear or guilt), they undoubtedly
the Jewish-Christian phase. also stem from antipathy.
As in Study 1, we hierarchically regressed the Taken together, the results support concep-
behavioral index on the attitude IAT, followed tualizing the IAT as a measure of individual
by the stereotype IAT. For the Asian-White differences in automatic biases. Study 2 directly
phase, the attitude IAT was a reliable predictor tested whether cultural favoritism influences the
in Step 1 (β = .25, p < .05), but it was reduced IAT, but found no evidence to support the hypo-
to nonsignificance in Step 2 (β = .16, p = .11) thesis (see also Nosek & Hansen, in press). Major-
after the stereotype IAT was accounted for ity and minority group members recognized the
(β = .30, p < .01). For the Black-White phase, the latter’s lower status, but this did not influence
attitude IAT was a reliable predictor in Step 1 their implicit biases. Thus, our results are not in
(β = .23, p < .05), but in Step 2 it was dramatically line with the strong form of the environmental
reduced (β = .05, ns); by contrast, the stereotype associations hypothesis, in which it is argued
IAT was significant (β = .24, p < .05). that IAT scores are attributable primarily to cul-
In sum, Study 2 showed that implicit biases tural, rather than personal, attitudes (Arkes &
predicted economic discrimination toward Tetlock, 2004; Olson & Fazio, 2004; Karpinski
Jews, Asians, and Blacks, and that the stereotype & Hilton, 2001; cf. Karpinski et al., 2005 for a
IAT was either an equal or superior predictor, more moderate version).
compared with explicit attitudes. When con- This is not to imply that cultural milieu has no
ditions afforded a comparison of the attitude influence on implicit biases, but rather to stress
and stereotype IATs, the latter was more effective that there is no clear boundary between self and
vis-a-vis contributing unique variance. In concert society—and this may be particularly true at the
with Study 1, the pattern suggests that evaluative automatic level (Banaji, 2001; Devine, 1989).
stereotypes reflect implicit biases better than Indeed, there are theoretical reasons to suspect
evaluative associations alone. Finally, in each that culture can condition people’s attitudes,
phase, the cultural stereotype index was un- with or without their consent (e.g. Banaji, 2001;
related to either attitude or stereotype IATs, as in Devine, 1989; Gaertner & Dovidio, 1986;
past research (Nosek & Hansen, in press). It was Greenwald & Banaji, 1995). Moreover, cultural
also unrelated to economic discrimination. biases may be internalized for many reasons,
including self-esteem, system justification,
and social adjustment functions. Thus, the
General discussion
relationship between self and society is likely
Across two studies, the stereotype IAT predicted to be interdependent, even for individuals who
harmful actions toward outgroup members, resist being prejudiced—a fact that leads to the
even after accounting for explicit prejudice necessity of becoming aware of automatic biases
measures. In Study 1, behaviors included non- in order to combat them.
Black participants’ reported history of verbal,
defensive, and offensive racial discrimination. The normativeness of implicit bias
In Study 2, stereotype IATs predicted budget The observation that IAT scores predict a range
reductions for Jewish, Asian, and Black student of discriminatory actions suggests that they
organizations. The behaviors assessed ranged are person-centered and somewhat reflective

368

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

 Rudman & Ashmore discrimination and the iat

of antipathy (see also Poehlman et al., 2004). weak performance of the attitude IAT (after
In the same breath, we do not believe that IAT accounting for the stereotype IAT, as well as
scores indicate explicit bigotry. Implicit biases explicit measures) suggests that evaluative be-
are simply automatic (over-learned) evaluations; liefs may capture the implicit prejudice construct
while they may reflect hostility, they also stem better than simple good-bad associations. This
from many other influences (Rudman, 2004), advantage may stem from the fact that negative
including a natural proclivity for partisanship outgroup stereotypes (and positive ingroup
(Greenwald et al., 2002). In this respect, we are stereotypes) afford more justification for
reminded of Allport’s (1954) defense of the discrimination than associating ingroup and
normality of social categorization and pre- outgroup members with pleasant and unpleasant
judgment, which softened the moral sting of words. However, considerably more research that
prejudice without removing responsibility for affords a comparison between implicit attitudes
it (Fiske, 2004). and stereotypes is needed before we can have
Indeed, a recurrent insight from response confidence in the stereotype IAT’s superior
latency measures is that people tend to automatic- predictive utility.
ally react with preference for similar others (as In addition, it would be interesting to uncover
they do for themselves; Greenwald & Farnham, moderators of reactions to IAT scores. To date,
2000). Although this bias is condoned for many there are indications that people who are motiv-
preferences (e.g. for our own children), it raises ated not to appear racist, or who are anxious
the specter of bigotry when applied to groups about their scores, tend to react defensively
who do not share our genetic makeup or cultural (Frantz, Cuddy, Burnett, Ray, & Hart, 2004;
background. Perhaps this is why some authors Monteith, Voils, & Ashburn-Nardo, 2001).
have argued (prematurely, in our view) that But some people respond to their IAT scores
cultural biases are primarily responsible for IAT with greater equanimity. For example, the first
scores. But if we can view automatic biases as author demonstrates her automatic biases in
reflective of the normal human condition, we the classroom, to create an atmosphere of trust.
will be less likely to stiff-arm the messenger and Placed in this context, implicit orientations
hopefully, more open to becoming aware of them become more normative and less threatening—
in order to better combat their consequences. not because society is to be blamed for them, but
because growing up in a culture where some
Limitations and future directions people are valued more than others is likely to
In Study 1, asking participants to report past permeate our private orientations, no matter how
hostile behaviors was likely to evoke social discomfiting the fact (Banaji, 2001; Devine, 1989).
desirability concerns. To counter this, we allowed For this reason, the IAT is a powerful educational
participants to first report the extent to which tool, as it opens people up to discussions about
they had suffered each behavior (Contrada social justice that might otherwise be dismissed
et al., 2001). The goal was to afford justification as antiquated (Bombardieri, 2005). Even people
for respondents’ own actions and encourage who argue that biases can be rational (e.g. Arkes &
honesty. Future research should compare results Tetlock, 2004) can appreciate the disconcerting
with and without justification items to examine fact that Blacks as a group are automatically
their effect. If they are advantageous, the pro- associated with negative attributes for many
cedure may provide a template for assessing Whites.
overt hostility—an important research agenda Future research should also continue to com-
(Mackie & Smith, 1998). pare the predictive utility of implicit and explicit
In two studies, the stereotype IAT predicted attitudes. Although it has been argued that
a range of discriminatory behaviors, in support implicit biases best predict spontaneous behaviors
of its construct validity. This was true when we (Fazio & Olson, 2003), self-reports (a controlled
controlled for explicit attitudes in both studies, behavior) often covary with them. In fact, validity
and for group identity in Study 2. The relatively for response latency measures has often relied

369

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

Group Processes & Intergroup Relations 10(3)

on controlled judgments (e.g. Livingston, 2002; et al. (2003), we computed the D statistic for
Rudman & Lee, 2002). Although some implicit use in correlational analyses. Results using log
measures appear to be better at predicting auto- transformed IAT scores were similar.
matic versus controlled behavior (Dovidio et al., 3. By convention, small, medium, and large effect
sizes correspond to .20, .50, and .80, respectively
2002), in other cases the behaviors have been a
(Cohen, 1988).
mix of automatic and controlled actions (Fazio
et al., 1995; McConnell & Leibold, 2001). The
present research used behavioral measures that Acknowledgments
were likely more controlled than automatic This research was partially supported by Grants
(reporting past discrimination in Study 1 and BCS-0109997 and BCS-0417335 from the National
recommending budget cuts in Study 2). Thus, Science Foundation. Preparation of this article was
there is no clean, process-driven divide by which partially supported by Grant BCS-0417335 from the
to define the predictive utility of implicit and National Science Foundation to the first author.
explicit responses.
Finally, the link between explicit attitudes and References
behavior is moderated by numerous factors; it Allport, G. W. (1954). The nature of prejudice.
cannot be otherwise for implicit attitudes (Fazio & New York: Addison-Wesley.
Olson, 2003; Karpinski et al., 2005; Nosek, 2005). Arkes, H. R., & Tetlock, P. E. (2004). Attributions
Moreover, behavior directed toward individuals of implicit prejudice, or ‘would Jesse Jackson
may be dramatically different from group-related “fail” the Implicit Association Test?’ Psychological
actions; our brains react differently when we Inquiry, 15, 257–278.
individuate (Wheeler & Fiske, 2005), and so Ashburn-Nardo, L., Voils, C. I., & Monteith, M. J.
do we (Brewer, 1988; Fiske & Neuberg, 1990). (2001). Implicit associations as the seeds of
Thus, we caution against assuming that IAT intergroup bias: How easily do they take root?
Journal of Personality and Social Psychology,
scores reflect how people will generally behave
81, 789–799.
toward outgroup members. Ashmore, R. D. (1970). The problem of intergroup
prejudice. In B. E. Collins (Ed.), Social psychology:
Notes Social influence, attitude change, group processes,
and prejudice (pp. 245–339). Reading, MA:
1. Readers might be interested in the relationships Addison-Wesley.
among being victimized and the implicit and Banaji, M. R. (2001). Implicit attitudes can be
explicit measures. Participants who reported measured. In H. L. Roediger, J. S. Nairne,
being excluded by outgroup members tended I. Neath, & A. Surprenant (Eds.), The nature of
to show high scores on the stereotype IAT remembering: Essays in honor of Robert G. Crowder
(r(62) = .32, p < .01), and marginally, on the (pp. 117–150). Washington D.C.: American
MRS and the thermometer (both rs(62) = .22, Psychological Association.
ps < .08). Targets of physical (or threatened) Blair, I. V. (2001). Implicit stereotypes and
aggression, and those who received the finger, prejudice. In G. B. Moskowitz (Ed.), Cognitive
also scored higher on the stereotype IAT (both social psychology: The Princeton symposium on the
rs(62) = .24, ps = .05). No other relationships legacy and future of social cognition (pp. 359–374).
approached significance (all rs(62) < .15, ns). Mahwah, NJ: Erlbaum.
2. The data for each trial block included mean Blair, I. V. (2002). The malleability of automatic
response latency (in ms) and error rates. stereotypes and prejudice. Personality and Social
Response latencies greater than 3000 ms and Psychology Review, 6, 242–261.
less than 300 ms were recoded as 3000 and Bombardieri, M. (2005, April 8). Summers
300 ms, respectively; the first two trials of each displays new understanding of women’s careers.
block were dropped because of their typically The Boston Globe, p. B3.
lengthened latencies. Analyses of participants’ Breckler, S. J. (1984). Empirical validation of
accuracy revealed low error rates on critical trials affect, behavior, and cognition as distinct
(an average of 5%); error trials were included in components of attitude. Journal of Personality
IAT effect scores. Following Greenwald and Social Psychology, 6, 1191–1205.

370

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

 Rudman & Ashmore discrimination and the iat

Brewer, M. B. (1988). A dual-process model of Greenwald, A. G., & Banaji, M. R. (1995). Implicit
impression formation. In T. K. Srull & social cognition: Attitudes, self-esteem, and
R. W. Wyer, Jr. (Eds.), Advances in social stereotypes. Psychological Review, 102, 4–27.
cognition (Vol. 1, pp. 1–36). Hillsdale: Greenwald, A. G., & Farnham, S. D. (2000). Using
NJ: Erlbaum. the Implicit Association Test to measure self-
Cohen, J. (1988). Statistical power for the behavioral esteem and self-concept. Journal of Personality and
sciences. Hillsdale, NJ: Erlbaum. Social Psychology, 79, 1022–1038.
Contrada, R. J., Ashmore, R. D., Gary, M. L., Greenwald, A. G., McGhee, D., & Schwartz, J. L. K.
Coups, E., Egeth, J. D., Sewell, A. et al. (1998). Measuring individual differences in
(2001). Measures of ethnicity-related stress: implicit cognition: The Implicit Association Test.
Psychometric properties, ethnic group Journal of Personality and Social Psychology, 74,
differences, and associations with well-being. 1464–1480.
Journal of Applied Social Psychology, Greenwald, A. G., Nosek, B. A., & Banaji, M. R.
31, 1775–1820. (2003). Understanding and using the Implicit
Cunningham, W. A., Preacher, K. J., & Banaji, M. R. Association Test: I. An improved scoring
(2001). Implicit attitude measures: Consistency, algorithm. Journal of Personality and Social
stability, and convergent validity. Psychological Psychology, 85, 197–216.
Science, 12, 163–170. Greenwald A. G., Pickrell, J. E., & Farnham, S. D.
Devine, P. G. (1989). Stereotypes and prejudice: (2002). Implicit partisanship: Taking sides for no
Their automatic and controlled components. reason. Journal of Personality and Social Psychology,
Journal of Personality and Social Psychology, 56, 83, 367–379.
5–18. Haddock, G., Zanna, M. P., & Esses, V. M. (1993).
Dovidio, J. F., Kawakami, K., & Gaertner, S. L. Assessing the structure of prejudicial attitudes:
(2002). Implicit and explicit prejudice and The case of attitudes toward homosexuals,
interracial interactions. Journal of Personality Journal of Personality and Social Psychology,
and Social Psychology, 82, 62–68. 24, 189–205.
Fazio, R. H., Jackson, J. R., Dunton, B. C., & Islam, M. R., & Hewstone, M. (1993). Dimensions
Williams, C. J. (1995). Variability in automatic of contact as predictors of intergroup anxiety,
activation as an unobtrusive measure of racial perceived out-group variability, and out-group
attitudes. A bona fide pipeline? Journal of attitude: An integrative model. Personality and
Personality and Social Psychology, 69, 1013–1027. Social Psychology Bulletin, 19, 700–710.
Fazio, R. H., & Olson, M. A. (2003). Implicit Jackson, J. R. (1997). Automatically activated racial
measures in social cognition research: Their attitudes. Unpublished doctoral dissertation,
meaning and use. Annual Review of Psychology, Indiana University.
54, 297–327. Jost, J. T., Pelham, B. W., & Carvallo, M. R. (2002).
Fiske, S. T. (2004). Intent and ordinary bias: Non-conscious forms of system justification:
Unintended thought and social motivation Implicit and behavioral preferences for higher
create casual prejudice. Social Justice Research, status groups. Journal of Experimental and Social
17, 117–127. Psychology, 38, 586–602.
Fiske, S. T., & Neuberg, S. L. (1990). A continuum Karpinski, A., & Hilton, J. L. (2001). Attitudes and
of impression formation, from category-based the Implicit Association Test. Journal of Personality
to individuating processes: Influences of and Social Psychology, 81, 774–788.
information and motivation on attention and Karpinski, A., Steinman, R. B., & Hilton, J. L.
interpretation. In M. P. Zanna (Ed.), Advances in (2005). Attitude importance as a moderator of
experimental social psychology (Vol. 23, pp. 1–74). the relationship between implicit and explicit
New York: Academic Press. attitude measures. Personality and Social Psychology
Frantz, C. M., Cuddy, A. J. C., Burnett, M., Ray, H., Bulletin, 31, 949–962.
& Hart, A. (2004). A threat in the computer: Lambert, A. J., Payne, B. K., Ramsey, S., & Shaffer,
The race Implicit Association Test as a stereotype L. M. (2005). On the predictive validity of
threat experience. Personality and Social Psychology implicit attitude measures: The moderating
Bulletin, 30, 1611–1624. effect of perceived group variability. Journal of
Gaertner, S. L., & Dovidio, J. F. (1986). The aversive Experimental Social Psychology, 41, 114–128.
form of racism. In J. F. Dovidio & S. L. Gaertner Livingston, R.W. (2002). The role of perceived
(Eds.), Prejudice, discrimination, and racism negativity in the moderation of African
(pp. 61–90). Orlando, FL: Academic Press. Americans’ implicit and explicit racial attitudes.

371

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015

Group Processes & Intergroup Relations 10(3)

Journal of Experimental Social Psychology, 38, Rudman, L. A., Greenwald, A. G., Mellott, D. S.,
405–413. & Schwartz, J. L. K. (1999). Measuring the
Mackie, D. M., & Smith, E. R. (1998). Intergroup automatic components of prejudice: Flexibility
relations: Insights from a theoretically and generality of the Implicit Association Test.
integrative approach. Psychological Review, 105, Social Cognition, 17(4), 1–29.
499–529. Rudman, L. A., & Lee, M. R. (2002). Implicit and
McConahay, J. B. (1986). Modern racism, explicit consequences of exposure to violent
ambivalence, and the Modern Racism Scale. and misogynous rap music. Group Processes &
In J. F. Dovidio & S. L. Gaertner (Eds.), Prejudice, Intergroup Relations, 5, 133–150.
discrimination, and racism (pp. 91–126). Orlando, Sears, D. (1988). Symbolic racism. In P. Katz &
FL: Academic Press. D. Taylor (Eds.), Eliminating racism: Profiles in
McConnell, A. R., & Leibold, J. M. (2001). Relations controversy (pp. 53–84). New York: Plenum.
among the Implicit Association Test, explicit Son Hing, L. S., Li, W., & Zanna, M. P. (2002).
attitudes, and discriminatory behavior. Journal of Inducing hypocrisy to reduce prejudicial
Experimental Social Psychology, 37, 435–442. responses among aversive racists. Journal of
Monteith, M. J., Voils, C. I., & Ashburn-Nardo, L. Experimental Social Psychology, 38, 71–78.
(2001). Taking a look underground: Detecting, Stephan, W. G., & Stephan, C. W. (1985).
interpreting, and reacting to implicit biases. Intergroup anxiety. Journal of Social Issues,
Social Cognition, 19, 395–417. 41, 157–175.
Nosek, B. A. (2005). Moderators of the relationship Uhlmann, E., & Poehlman, T. A. (2005). Automatic
between implicit and explicit evaluation. Journal associations: Personal attitudes or cultural knowledge?
of Experimental Psychology: General, 134, 565–584. Unpublished manuscript, Yale University.
Nosek, B. A., Banaji, M. R., & Greenwald, A. G. Wheeler, M. E., & Fiske, S. T. (2005). Controlling
(2002). Harvesting implicit group attitudes and racial prejudice: Social cognitive goals affect
beliefs from a demonstration web site. Group amygdala and stereotype activation. Psychological
Dynamics, 6, 101–115. Science, 16, 56–63.
Nosek, B. A., & Hansen, J. J. (in press). The Wittenbrink, B., Judd, C. M., & Park, B. (1997).
associations in our heads belong to us: Searching Evidence for racial prejudice at the implicit
for attitudes and knowledge in implicit level and its relationship with questionnaire
evaluation. Cognition and Emotion. measures. Journal of Personality and Social
Olson, M. A., & Fazio, R. H. (2004). Reducing the Psychology, 72, 262–274.
influence of extra-personal associations on the Zanna, M. P. (2004). The naive epistemology of
Implicit Association Test: Personalizing the IAT. a working social psychologist (or the working
Journal of Personality and Social Psychology, 86, epistemology of a naive social psychologist):
653–667. The value of taking ‘temporary givens’ seriously.
Poehlman, T. A., Uhlmann, E., Greenwald, A. G., Personality and Social Psychology Review, 8,
& Banaji, M. R. (2004). Measuring and using 210–218.
the Implicit Association Test: III. Meta-analysis of
predictive validity. Manuscript submitted for Paper received 17 June 2005; revised version accepted
publication. 6 June 2006.
Rudman, L. A. (2004). Sources of implicit attitudes.
Current Directions in Psychological Science, 13(2),
79–83. Biographical notes
Rudman, L. A., Ashmore, R. D., & Gary, M. L. laurie rudman is an associate professor in the
(2001). ‘Unlearning’ automatic biases: The Psychology Department at Rutgers University.
malleability of implicit stereotypes and
Her primary research interests are prejudice,
prejudice. Journal of Personality and Social
stereotypes, and implicit social cognition.
Psychology, 81, 856–868.
Rudman, L. A., Feinberg, J. M., & Fairchild, K.
richard ashmore is a professor in the Psychology
(2002). Minority members’ implicit attitudes:
Department at Rutgers University. His primary
Ingroup bias as a function of group status. Social
research interests are self and identity and
Cognition, 20, 294–320.
intergroup relations.
Rudman, L. A., & Glick, P. (2001). Prescriptive
gender stereotypes and backlash toward agentic
women. Journal of Social Issues, 57, 743–762.

372

Downloaded from gpi.sagepub.com at University of Birmingham on May 28, 2015