The Psychology of Bigotry 519consists of a symbolic representation of the attitude object,
such as digitized photos of stimulus persons from one or
more racial groups. Immediately following the prime, a target
in the form of a positive or negative evaluative adjective is
displayed, and the participant is required to indicate its con-
notation as either good or bad by pressing different computer
keys. When the prime and target are evaluatively congruent
for the participant, responding should be facilitated as mani-
fested in a faster, more efficient reaction time. By contrast,
when prime and target are evaluatively incongruent with one
another from the viewpoint of the participant, responding
should be slowed, as reflected by a longer reaction time.
Using this priming procedure, Fazio et al. (1995) showed
in several studies that White U.S. university students showed
greater facilitation when negative adjectives were preceded by
photos of Black people. By contrast, a small sample of Black
participants showed response facilitation on the priming task
when photos of Blacks preceded positive adjectives and when
White photos were preceded by negative adjectives. More-
over, scores on this unobtrusive measure of racial attitudes had
predictive validity for a Black experimenter’s ratings of the
participant’s friendliness and interest when interacting with
her, to which MRS scores were unrelated.
Along similar lines, Greenwald, McGhee, and Schwartz
(1998) suggested the Implicit Association Test (IAT) as a
related procedure for assessing implicit attitudes, defined as
behaviors, feelings, or thoughts elicited outside the partici-
pant’s awareness by automatically activated evaluation
procedures (see Greenwald & Banaji, 1995). The IAT con-
sists of a series of five discrimination tasks, conducted on
computer, in which the participants differentiate between two
categories of stimuli by responding as quickly as possible on
different computer keys.
If one were assessing White attitudes toward Black people
with the IAT, the first task would be an initial target-concept
discrimination in which they might be asked to differentiate
between White and Black American first names by pressing
different keys on the computer. The second task is an associ-
ated attribute discrimination in which the participant differ-
entiates pleasant from unpleasant words. The third step is the
initial combined task in which the two prior tasks are now su-
perimposed or mapped onto one another, such as using one
key for individual stimuli that are either White or pleasant
and another key for stimuli that are either Black or unpleas-
ant. In the fourth step, the response keys from the first
task are reversed. The fifth and final step, the reverse com-
bined task, reverses the response key contingencies from the
third step (e.g., one key for stimuli that are either White or
unpleasant or either Black or pleasant. The difference in
speed of responding to the two combined tasks on the IAT
provides the measure of implicit attitudes. Following the ear-
lier example, a latency shorter for the first combined task than
for the reverse combined task would suggest a less positive or
more negative implicit attitude toward Blacks by a White
participant.
Using the IAT, Greenwald and Banaji (1998) found evi-
dence that it may reveal the existence of prejudice that is not
evident on paper-and-pencil attitude measures such as the
semantic differential scale. Whereas a majority of a sample of
White American participants in one study indicated no
Black-White difference or even a pro-Black preference on
paper-and-pencil ratings, all but one had IAT scores indicat-
ing a White preference, presumably a nonconscious one.
Greenwald and his colleagues have also found modest posi-
tive correlations between IAT scores and some “explicit” at-
titude measures such as the feeling thermometer (in which
social groups are rated on a 100-point thermometer scale) and
a diversity index but not others, especially semantic differen-
tial scales. IAT scores, they suggested, do not merely reflect
greater familiarity with one’s in-group (e.g., naming prac-
tices, facial stimuli) compared to an out-group. The IAT pro-
cedure, they also proposed, yields stronger effect sizes and is
therefore more sensitive than the priming procedure devised
by Fazio et al. (1995) and by other investigators.
One would not necessarily expect implicit and explicit
measures of racial attitudes to correlate highly with one
another. Demonstrating this point, Dovidio, Kawakami,
Johnson, Johnson, and Howard (1997) showed that the predic-
tive validity of implicit (i.e., elicited by automatic processing
techniques, such as priming or the IAT) and explicit measures
of racial attitudes (i.e., elicited by self-report measures such as
scales of modern and old-fashioned racism) of White partici-
pants toward Black people diverges in a predictable manner.
Specifically, implicit prejudice measures predicted sponta-
neous cognitions and behaviors that are not easily monitored
but reflect automatic processing, such as performance on a
word-completion task in which answers may be racially tinged
or nonverbal behavior such as eye blinking or direct gaze
when interacting with a Black person. By contrast, explicit
prejudice measures possessed predictive validity for delibera-
tive thoughts and actions that reflect controlled processing,
such as judgments of a Black defendant’s guilt in a juridic
decision-making task and evaluations.
Fazio et al. (1995) had previously obtained a similar pat-
tern of findings. Their unobtrusive priming measure of preju-
dice in Whites had predictive validity for rated quality of
interaction with a Black experimenter, whereas explicit mea-
sures predicted deliberative acts such as attractiveness ratings
of photos and evaluations of the fairness of the Rodney King
verdict (in which White police officers in Los Angeles were