Explicit And Implicit Attitudes That Caused Brexit Assignment

Description

Implicit and Explicit Attitudes

Historically, social psychology has focused on the measurement of explicit attitudes, that is, a person’s conscious attitudes about a person, object, or idea. In areas of the human experience that deal with socially sensitive topics, however, social psychologists have found the need to try and assess hidden, or implicit, attitudes that people may not be aware they hold, or may not wish to express. While at times social psychologists are still interested in the explicit attitudes that people choose to express, and why they choose to express them, the field is increasingly drawn to the implicit attitudes that are unexpressed and the reasons why people hide them.

For this Discussion, consider a current controversial social issue that sparks reasoned debate and think about explicit and implicit attitudes related to that issue. Also consider how implicit attitudes might influence intergroup behavior.


In the essay, please follow the instructions as below:

1. Select a current controversial social issue, and give a brief introduction. (For example, US-China Trade War, Brexit…)

2. Explain the distinction between explicit and implicit attitudes related to that issue.

3. Finally, explain possible impacts of implicit attitudes and beliefs on intergroup behavior in the issue you selected.


Be sure to support your postings and responses with specific references to the Learning Resources and the current literature.

Journal of Personality and Social Psychology
2006, Vol. 91, No. 4, 652– 661
Copyright 2006 by the American Psychological Association
0022-3514/06/$12.00 DOI: 10.1037/0022-3514.91.4.652
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Stereotyping and Evaluation in Implicit Race Bias: Evidence for
Independent Constructs and Unique Effects on Behavior
David M. Amodio
Patricia G. Devine
New York University
University of Wisconsin—Madison
Implicit stereotyping and prejudice often appear as a single process in behavior, yet functional neuroanatomy suggests that they arise from fundamentally distinct substrates associated with semantic versus
affective memory systems. On the basis of this research, the authors propose that implicit stereotyping
reflects cognitive processes and should predict instrumental behaviors such as judgments and impression
formation, whereas implicit evaluation reflects affective processes and should predict consummatory
behaviors, such as interpersonal preferences and social distance. Study 1 showed the independence of
participants’ levels of implicit stereotyping and evaluation. Studies 2 and 3 showed the unique effects of
implicit stereotyping and evaluation on self-reported and behavioral responses to African Americans
using double-dissociation designs. Implications for construct validity, theory development, and research
design are discussed.
Keywords: prejudice, stereotyping, implicit evaluation, affect, cognition
affective and semantic neural circuits are most pronounced in more
basic, implicit levels of processing, theories of implicit race bias
have much to gain by considering the alternative roles of affect and
cognition.
In the present research, we examined the roles of affect and
cognition in implicit race bias and their effects on behavior. On the
basis of theory and research from social psychology and neuroscience, we argue for a conceptual distinction between implicit
stereotyping and implicit evaluative race bias and propose that
these two forms of implicit race bias are predictive of different
types of discriminatory responses.
The distinction between affect and cognition in the human
psyche dates back to the earliest philosophers of the mind and
continues to be a major feature of modern psychology and neuroscience. Indeed, contemporary theorists have argued that the
affective– cognitive distinction is essential for understanding the
mind, brain, and behavior (Cacioppo, Gardner, & Berntson, 1999;
Damasio, 1994; Zajonc, 1980), and neuroscientists have delineated
distinct neural pathways for basic affective and cognitive systems
of learning and memory (Davis & Whalen, 2001; Squire & Zola,
1996). In the intergroup relations literature, affect and cognition
traditionally correspond to two key components of race bias:
prejudice and stereotyping (Allport, 1954; Devine, 1989; Dovidio,
Brigham, Johnson, & Gaertner, 1996; Fiske, 1998; Mackie &
Smith, 1998). Whereas the term prejudice refers to negative affective responses toward outgroup members (McConahay &
Hough, 1976), the term stereotype refers to cognitive representations of culturally held beliefs about outgroup members (Hamilton,
1981).
In research on traditional, explicit race biases, the conceptual
distinction between prejudice and stereotyping has provided a
useful framework for examining their respective contributions to
different forms of discrimination (Dovidio, Esses, Beach, & Gaertner, 2004; Park & Judd, 2005). By contrast, in research on more
automatic, or implicit, forms of race bias, little attention has been
given to the affective– cognitive distinction or the important implications it may have for understanding the relationship between
implicit biases and behavior. Because the distinction between
Relationship Between Implicit Stereotyping and
Evaluation
A survey of the implicit race bias literature reveals that very few
studies have directly examined the relation between affective and
cognitive aspects of implicit bias (for reviews, see Blair, 2001;
Fazio & Olson, 2003), and none have sought to obtain independent
measures of implicit stereotyping versus evaluation.1 Indeed, most
expressions of race bias reflect a combination of affective and
cognitive processes, and the most commonly reported African
American stereotypes are negative in valence (e.g., unintelligent,
hostile, poor, lazy, and dishonest; Devine & Elliot, 1995). Yet
despite the common concurrence of negative valence and stereotypes of stigmatized groups, underlying distinctions between affective and cognitive components may be important for understanding mechanisms of implicit race biases and their effects on
behavior.
David M. Amodio, Department of Psychology, New York University;
Patricia G. Devine, Department of Psychology, University of Wisconsin—
Madison.
We gratefully acknowledge Carolyn Stahlhut, Ryan Beld, and Marissa
Langhoff for their assistance in data collection.
Correspondence concerning this article should be addressed to David M.
Amodio, Department of Psychology, New York University, 6 Washington
Place, New York, NY 10003. E-mail: David.amodio@nyu.edu
1
Throughout this article, we use the term implicit evaluation rather than
implicit prejudice as a more precise label to refer to automatic evaluative
associations. By using implicit evaluation, we avoid invoking unintended
connotations associated with the complicated construct of prejudice, such
as consciously endorsed racist attitudes and beliefs (Amodio et al., 2003;
Devine et al., 2002).
652
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
IMPLICIT STEREOTYPING VS. EVALUATIVE RACE BIAS
Although the distinction is seldom made, past research has
featured measures that may be characterized as assessing either
implicit stereotyping (e.g., Lepore & Brown, 1997; Spencer, Fein,
Wolfe, Fong, & Dunn, 1998), implicit evaluation (e.g., Amodio,
Harmon-Jones, & Devine, 2003; Dovidio, Kawakami, Johnson,
Johnson, & Howard, 1997; Fazio, Jackson, Dunton, & Williams,
1995; Greenwald, McGhee, & Schwartz, 1998), or some combination of stereotyping and evaluation (e.g., Dovidio, Evans, &
Tyler, 1986; Kawakami, Dion, & Dovidio, 1998; Rudman, Ashmore, & Gary, 2001; Wittenbrink, Judd, & Park, 1997, 2001). The
use of such measures suggests that both are valid constructs that
have been studied somewhat independently and that both forms of
implicit bias are prevalent among European Americans, such that
African Americans are typically associated with negative evaluations and with the culturally defined stereotype content (Blair,
2001). However, although the distinction between implicit stereotyping and implicit evaluation has been acknowledged in past
work (e.g., Greenwald & Banaji, 1995; Greenwald et al., 2002),
theorizing has not been advanced to directly address the relation
between cognitive and affective mechanisms underlying these two
forms of implicit race bias.
Distinct Neural Substrates for Basic Affective and
Semantic Associations
In the neuroscience literature, neural substrates of affective
forms of learning and memory have been distinguished from
semantic forms, and this distinction has implications for the
present set of issues. Results from decades of research on animals
and humans suggest that the amygdala and its associated subcortical circuits are central to affective learning and memory (Lang,
Bradley, & Cuthbert, 1990; LeDoux, 2000). This body of work has
shown that affective associations are learned quickly, often after a
single presentation of an unconditioned stimulus in a fear-learning
paradigm. Once learned, such associations extinguish slowly, and
subsequent reconditioning to the stimulus is facilitated (Bouton,
1994). It is important to note that amygdala-based learning does
not depend on semantic associations; for example, mice easily
learn affective associations despite their inability to process semantic information. By comparison, semantic learning and memory (e.g., conceptual priming) are embedded in mechanisms for
language, believed to be supported by a phylogenetically newer
network of neocortical structures that are significantly expanded
among humans compared with those of other species (Gabrieli,
1998; Rissman, Eliassen, & Blumstein, 2003; Squire & Zola,
1996). Semantic associations may be learned in the absence of
affective content, such that patients with a damaged amygdala
retain normal semantic associations despite the loss of conditioned
physiological responses in a fear-conditioning paradigm (Bechara,
Damasio, & Damasio, 2003).
An examination of anatomical and neurochemical connectivity
of the amygdala and surrounding structures reveals strong direct
links with neural regions associated with modulating behavior on
the basis of reward and punishment cues (e.g., basal ganglia, motor
cortex, orbital frontal cortex; Davis & Whalen, 2001; Park & Judd,
2005) and for mobilizing fight or flight responses (e.g., via the
hypothalamic–pituitary–adrenal axis; Feldman, Conforti, & Weidenfeld, 1995). By contrast, neocortical regions associated with
semantic associations appear to have few, if any, direct connections to these systems. Rather, semantic associations are likely
653
embedded in distributed networks in association cortex and thus
may influence social cognition by biasing higher order information
processing, such as when inferring the beliefs and intentions of
another person (Amodio & Frith, 2006). Although systems for
affect- and semantic-based associations typically function in concert, and thus tend to appear blended in outward verbal and
behavioral responses, a consideration of their distinct operations is
critical for understanding the behavioral effects of implicit stereotyping and evaluation.
Relationship Between Implicit Stereotyping and Implicit
Evaluative Race Bias
On the basis of social psychological and neuroscientific theorizing, we proposed that implicit stereotyping and evaluation
should represent independent constructs. Although past theorizing
has pointed to this distinction (e.g., Greenwald & Banaji, 1995;
Greenwald et al., 2002), few studies have explored it directly (cf.
Dovidio et al., 1986; Kawakami et al., 1998; Rudman, Greenwald,
& McGhee, 2001; Wittenbrink et al., 1997, 2001), and none has
examined the respective implications of implicit stereotyping versus implicit evaluation for behavior. A limiting factor in this line
of inquiry is that in previous research, independent assessments of
implicit stereotyping and evaluative race bias have not been obtained, and thus it has not been possible to examine the conceptual
relationship of implicit stereotyping and evaluative race bias and
their potentially unique effects on behavior. Hence, the first major
goal in the present work was to obtain independent measures of
implicit stereotyping and evaluation that would permit a fair test of
the independence hypothesis.
Differential Effects of Implicit Evaluative Race Bias and
Stereotyping on Behavior
If implicit stereotyping and evaluation reflect independent
cognitive and affective systems, then they may be uniquely
associated with different types of discriminatory responses.
Consistent with this hypothesis, Millar and Tesser (1986, 1989)
argued that instrumental behaviors (e.g., forming judgments
and goals) are driven primarily by cognitive processes, whereas
consummatory behaviors (e.g., appetitive or aversive behaviors) are driven primarily by affective– evaluative processes. On
the basis of this theorizing, Dovidio and his colleagues (1996,
2004; Esses & Dovidio, 2002; see also Stangor, Sullivan, &
Ford, 1991) proposed that by considering the match between the
affective versus cognitive nature of race-bias measures and
forms of discriminatory outcomes, greater correspondences between assessments of race bias and behavior may be attained. In
a meta-analysis focusing on explicit self-reports of stereotyping
and prejudice, Dovidio et al. (2004) showed that affect-based
measures of race bias tended to be predictive of basic approach/
avoidance responses (e.g., nonverbal behaviors and affective
responses) toward African Americans, whereas cognition-based
measures of race bias tended to be predictive of the endorsement of stereotypes and support for policies that disadvantage
African Americans.
Although Dovidio et al.’s meta-analysis focused on explicit
measures of race bias, extant findings from the implicit race
bias literature are generally consistent with these predictions
(Ashburn-Nardo, Knowles, & Monteith, 2003; Dovidio et al.,
AMODIO AND DEVINE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
654
1997; Dovidio, Kawakami, & Gaertner, 2002; Fazio et al.,
1995; McConnell & Leibold, 2001; Wilson, Lindsey, &
Schooler, 2000). For example, Fazio et al. (1995) found that
implicit evaluative bias was predictive of less friendly behavior
toward a Black experimenter but was not associated with participants’ views on the Rodney King verdict and ensuing riots.
In other research, greater implicit evaluative bias was associated with more uncomfortable interactions (e.g., less eye contact, more blinking) with a Black confederate compared with
those involving a White confederate (Dovidio et al., 1997,
2002) and more negative interactions with a Black experimenter
on a host of indicators, including speech hesitations and errors
and behavior judged to be abrupt, unfriendly, and uncomfortable (McConnell & Leibold, 2001). By contrast, researchers
have not examined the unique effects of implicit stereotyping
on behavior, although some previous findings bear on the topic.
For example, Kawakami et al. (1998) found that higher levels of
implicit stereotyping were predictive of the attribution of stereotypic traits to a larger proportion of African Americans (in
addition to reporting more prejudiced attitudes). It is important
to note that in previous research, the hypothesis that implicit
stereotyping and implicit evaluation are uniquely predictive of
alternative forms of race-biased behavior has not been directly
tested. Hence, the second main goal of the present work was to
test this hypothesis directly.
Overview of Present Research
In the present research, we examined the relationship between
implicit stereotyping and implicit evaluative race bias and their
respective effects on instrumental versus consummatory forms of
race-biased behavior. Although stereotyping and evaluation processes typically operate in concert, it was necessary for us to obtain
relatively pure measures of implicit stereotyping and evaluation to
examine their unique effects on behavior. To this end, we designed
separate implicit association tests (IATs) to assess implicit stereotyping and implicit evaluative race bias. The IAT was chosen
because it has been shown to be reliable (Greenwald, Nosek, &
Banaji, 2003), and it has been widely used in the implicit race bias
literature (Devine, 2001). In Study 1, we examined the degree to
which measures of implicit stereotyping and evaluative race bias
were independent (i.e., uncorrelated). In Studies 2 and 3, we
examined the unique effects of implicit stereotyping and evaluation on instrumental and consummatory forms of behavior.
Study 1
Method
Black or White by pressing one of two keys on the computer keyboard.
Stimuli consisted of pleasant and unpleasant words as used by Greenwald
et al. (1998) and pictures of White and Black male faces displaying neutral
expressions (Malpass, Lavigueur, & Weldon, 1973) as used by Devine,
Plant, Amodio, Harmon-Jones, and Vance (2002; Study 3). Pleasant words
included honor, lucky, diamond, loyal, freedom, rainbow, love, honest,
peace, and heaven. Unpleasant words included evil, cancer, sickness,
disaster, poverty, vomit, bomb, rotten, abuse, and murder.
The IAT procedure comprised five blocks of trials (Greenwald et al.,
1998). Stimuli were presented individually in the center of the computer
monitor in randomized order. In Block 1, participants viewed 10 Black
and 10 White faces and categorized Black faces by pressing the left
response key (“a” on the alphabetic keyboard) and White faces by
pressing the right response key (“5” on numeric keypad). In Block 2,
participants viewed 10 pleasant and 10 unpleasant words, categorizing
unpleasant words with the left response key and pleasant words with the
right response key. In Block 3, stimuli included White faces, Black
faces, pleasant words, and unpleasant words, and response mappings
were combined such that participants categorized Black faces and
unpleasant words by pressing the left response key and White faces and
pleasant words by pressing the right response key. This block consisted
of 40 trials and was referred to as the compatible block (Greenwald et
al., 1998), given that response pairings of White with good and Black
with bad are compatible with Whites’ tendency to prefer White faces
over Black faces. In Block 4, participants viewed 10 Black and 10
White faces but this time categorized White faces with the left response
key and Black faces with the right response key to counterbalance
response mappings. In Block 5, categorizations were again combined
such that participants categorized White faces and unpleasant words by
pressing the left response key and Black faces and pleasant words by
pressing the right response key. This block included 40 trials and was
referred to as the incompatible block. Half of the participants completed
the IAT as described above; half completed a version with reversed
response mappings.
Stereotyping IAT. We designed a new IAT in which participants
viewed two classes of words associated with the positive characteristics of
intelligence and athleticism/rhythmicity, and categorized them as mental or
physical, respectively, in addition to the Black versus White face categorizations. Intelligence and athleticism/rhythmicity are central to the African
American stereotype, such that African Americans are stereotyped as more
athletic/rhythmic and less intelligent than European Americans (Devine &
Elliot, 1995). Because the mental and physical categories were relatively
neutral, the categorization of words relating to athleticism/rhythmicity and
intelligence as mental or physical did not involve evaluative judgments.2
Target word stimuli used in the stereotyping IAT were selected on the
basis of pretesting.3 Ten target words were selected for each category on
the basis of category fit and stereotypicality. Mental words included math,
brainy, aptitude, educated, scientist, smart, college, genius, book, and read.
Physical words included athletic, boxing, basketball, run, agile, dance,
jump, rhythmic, track, and football. The procedure for the stereotyping IAT
was identical to that of the evaluative IAT, except that the pleasant and
Participants and Procedure
One hundred fifty-one European American introductory psychology
students (82 women, 69 men) participated in exchange for extra course
credit. After providing informed consent, participants received instructions
on completing separate IAT measures of stereotyping and prejudice administered on a PC using Inquisit software (Millisecond Software, Seattle,
WA). IAT order was counterbalanced across participants. After completing
the measures, participants were debriefed, thanked, and dismissed.
Materials
Evaluative IAT. The IAT is a dual categorization task in which participants categorize words as pleasant or unpleasant and faces as either
2
We developed additional IATs for other common African American
stereotypes. Using the method by which Rudman, Greenwald, and McGhee
(2001) measured implicit gender stereotyping, we pretested sets of target
words related to poor (vs. wealthy), hostile (vs. friendly), and lazy (vs.
motivated). In each case, however, the stereotype was strongly related to
evaluation (e.g., poor is negative and wealthy is positive), and therefore
these were not suitable for examining the independence of implicit evaluation and implicit stereotyping.
3
Our lab group first generated separate lists of 22 words corresponding
to the physical and mental categories. Sixty-one introductory psychology
students then rated the fit of each word with its respective category and its
degree of association with White and Black Americans on a scale of 1 (not
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
IMPLICIT STEREOTYPING VS. EVALUATIVE RACE BIAS
unpleasant target words and category labels were replaced with
intelligence- and athletics-related target words and the mental and physical
category labels. Hence, the compatible block included Black/physical and
White/mental categorizations and the incompatible block included Black/
mental and White/physical categorizations.
IAT scoring. Responses to the evaluative and stereotyping IATs were
scored using the “improved algorithm,” outlined by Greenwald et al. (2003,
p. 214), which produced the D statistic.4 However, because the IAT used
in Study 1 consisted of the original five-block version (Greenwald et al.,
1998), steps involving practice blocks were omitted. Following the algorithm, responses with latencies greater than 10,000 ms were removed.
Separate means were computed for correct raw response latencies on
compatible and incompatible blocks. Error responses within each block
were replaced by the mean correct reaction time for that block, plus a
600-ms error penalty. D was quantified as the difference between incompatible and compatible mean reaction times divided by the pooled standard
deviation of reaction times on compatible and incompatible blocks. Data
from two participants were excluded because of outlying scores (Student’s
t scores differed significantly from mean, p ⬍ .05), and data from one
participant were excluded because a high percentage of his responses
(18%) on the stereotyping IAT were faster than 300 ms (Greenwald et al.,
2003); results did not differ when outliers were included.
Results
Evidence for implicit bias was examined using one-sample t
tests of D scores (effect size r is presented for each t value). All
tests were two-tailed. Evaluative IAT scores were significantly
greater than zero (M ⫽ .51, SD ⫽ .42), t(147) ⫽ 14.60, p ⬍ .001,
r ⫽ .77, suggesting a negative evaluative association with Black
faces relative to White faces, replicating past work. Stereotyping
IAT scores were also significantly greater than zero (M ⫽ .17,
at all) to 9 (extremely). Pretest ratings of category fit for physical and
mental target words exceeded the scale midpoint, ps ⬍ .001, indicating that
target words were good exemplars of their respective categories, and fit
scores for the mental and physical target words did not differ, t(60) ⫽ .40,
p ⫽ .69, r ⫽ .05. Physical target words were rated as more stereotypical of
Black people (M ⫽ 7.68, SD ⫽ 0.98) than of White people (M ⫽ 5.19,
SD ⫽ 1.26), t(60) ⫽ 13.12, p ⬍ .001, r ⫽ .86, whereas mental target words
were rated as more stereotypical of White people (M ⫽ 7.23, SD ⫽1.42)
than of Black people (M ⫽ 4.30, SD ⫽ 1.31), t(60) ⫽ 13.31, p ⬍.001,
r ⫽ .86.
A separate sample of 39 participants rated the favorability of words
associated with the mental/physical and pleasant/unpleasant IATs on a
scale from 1 (extremely unfavorable) to 9 (extremely favorable). Pleasant
words (M ⫽ 7.93, SD ⫽ 0.54) were rated much more favorably than
unpleasant words (M ⫽ 1.75, SD ⫽ 0.58), t(38) ⫽ 38.83, p ⬍ .001, r ⫽ .99.
Unexpectedly, mental words (M ⫽ 7.06, SD ⫽ 0.78) were rated as more
favorable than were physical words (M ⫽ 6.22, SD ⫽ .77), t(38) ⫽ 7.07,
p ⬍ .001, r ⫽ .75, although both mental and physical word lists were rated
significantly above the neutral midpoint of the scale, ps ⬍ .001, and both
were rated as more favorable than unpleasant words, ps ⬍ .001, and less
favorable than pleasant words, ps ⬍ .001. Although mental words were
rated more favorably than physical words, this difference was much
smaller than the difference in ratings between pleasant and unpleasant
words, t(38) ⫽ 28.77, p ⬍ .001, r ⫽ .98. We used covariate analyses in our
hierarchical regressions to ensure that effects of stereotyping IAT scores
were not driven by evaluative associations (and vice versa) because any
shared variance was statistically controlled. If anything, the valence effect
found among the stereotyping IAT words would enhance the relationship
between implicit stereotyping and evaluation, thereby working against our
hypotheses and rendering more conservative tests.
655
SD ⫽ .43), t(147) ⫽ 4.72, p ⬍ .001, r ⫽ .36, such that participants
exhibited a pattern of stereotypic trait associations with Black and
White faces. No effects were found for sex or IAT order, Fs ⬍ 1.
Next, we tested our primary hypothesis that levels of implicit
stereotyping and implicit evaluation should be independent by
examining their correlation. Participants’ evaluative and stereotyping IAT scores were not significantly correlated, r(147) ⫽ .06, p ⫽
.47, supporting our hypothesis.
Discussion
The results of Study 1 showed that participants possessed significant levels of implicit evaluative and stereotyping biases but
that their levels of each bias were uncorrelated, suggesting conceptual independence. It is noteworthy that although athleticism,
rhythmicity, and (un)intelligence represent a subset of commonly
observed African American stereotypes, they are among the most
central to the stereotype. Indeed, these three attributes were the
most frequently cited by participants instructed to freely list traits
associated with African Americans (Devine & Elliot, 1995). Because our stereotyping IAT focused on the three most central traits
of the African American stereotype, and given previous findings
that the activation of a central stereotype typically activates the
constellation of African American stereotypes (Devine, 1989;
Lepore & Brown, 1997), it is likely that our measure of implicit
stereotyping reflected associations with the general African American stereotype. Nevertheless, it would be important to show that
stereotyping IAT scores were predictive of responses to an African
American target, reflecting stereotypic content that reached beyond traits of (un)intelligence, athleticism, and rhythmicity.
Studies 2 and 3 were designed with two goals in mind: to
replicate Study 1 findings and to test the hypothesis that implicit
stereotyping and evaluation are uniquely predictive of different
forms of race-biased behavioral outcomes. The behavioral effects
of implicit stereotyping and evaluation in Studies 2 and 3 were
examined using double-dissociation designs constructed to isolate
unique effects of predictors on specific outcome variables. Here,
we tested the hypothesis that implicit stereotyping would be associated with instrumental but not with consummatory forms of
race-biased behavior, whereas implicit evaluative race bias would
be associated with consummatory but not with instrumental forms
of race-biased behavior.
Study 2
In Study 2, we examined the degree to which participants’ levels
of implicit stereotyping and evaluation influenced their impressions of an African American student. To measure instrumental
forms of behavior, we assessed participants’ use of stereotypes as
they formed an impression of the African American student on the
basis of the student’s writing sample (Moreno & Bodenhausen,
2001). To measure basic approach/avoidance responses associated
with consummatory behaviors, we examined participants’ preference for the writer as a potential friend. We also collected participants’ affective ratings of various ethnic groups, including African Americans, using a feelings thermometer measure. We
4
IAT effects based on difference scores (e.g., Greenwald et al., 1998)
replicated results reported for the D statistic in all studies. Analyses of
difference scores are available from the authors.
AMODIO AND DEVINE
656
hypothesized that implicit stereotyping but not implicit evaluation
would be related to more stereotypic trait ratings of the African
American student, whereas implicit evaluation but not stereotyping
would relate to a greater desire to befriend the writer and more
negative affective responses toward African Americans.
Method
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Participants and Procedure
Thirty-six European American introductory psychology students (15
men, 21 women) participated in exchange for extra course credit. After
providing consent, participants were told that the study consisted of two
parts. The experimenter explained that the first part examined people’s
ability to form impressions of others on the basis of short writing samples.
Participants were shown a set of 10 file folders containing different writing
samples and were asked to choose one at random (although all folders
contained identical materials). Participants were given the chosen folder,
which contained the writer’s demographic information, a copy of the essay,
and a set of forms to record their ratings. The demographic information
included the writer’s name, age, sex, ethnicity, year in college, and hometown, indicating that he was a 19-year-old male African American sophomore from Milwaukee, WI. Participants transferred this demographic
information onto the evaluation form; read the essay, which contained
some grammatical and spelling errors; and then provided their ratings of
the essay and the writer. As the second part of the study, participants
completed the evaluative and stereotyping IATs, in counterbalanced order,
and the feelings thermometer measure. The essay ratings, IATs, and the
feelings thermometer measure were administered in this order to prioritize
the more covert measures as a means of minimizing participants’ suspicions. Lastly, participants were probed for suspicion regarding the cover
story and hypotheses, debriefed, thanked, and dismissed. Five participants’
data were excluded because their scores on one or more measures differed
significantly from the mean ( p ⬍ .05) in a Student’s t distribution and were
considered outliers. Although inclusion of outliers inflated standard errors
and thus weakened effect sizes, it did not change the pattern of effects.
Materials
Evaluative and stereotyping IATs. The evaluative and stereotyping
IATs consisted of the same stimuli described in Study 1 but were administered using DirectRT software (Empirisoft, New York) and included sets
of 20 practice trials before the compatible and incompatible blocks. The D
statistic was computed as in Study 1, with the additional incorporation of
responses from the practice blocks (Greenwald et al., 2003).
Essay evaluation materials. The essay evaluation form included items
for rating (a) the general quality and style of the essay (included to bolster
the cover story), (b) the trait attributes of the writer, and (c) participants’
liking of and perceived similarity with the writer. Trait ratings of the writer
were made using a scale ranging from 1 (not at all) to 10 (very much) on
a list of adjectives known to be highly associated with the Black stereotype
(lazy, dishonest, unintelligent, and trustworthy; Devine & Elliot, 1995)
intermixed with filler traits that were relatively neutral and not typically
associated with the stereotype (modest, assertive, and thoughtful). Ratings
were averaged to form separate indices of stereotypic ratings (␣ ⫽ .68,
with trustworthy reverse-scored) and neutral filler ratings (␣ ⫽ .53). Liking
ratings were made for five items (e.g., “The writer seems like the type of
person I would like to get to know better”; “The writer and I have a lot of
things in common”) on a scale of 1 (strongly disagree) to 10 (strongly
agree; mean ratings: ␣ ⫽ .73).
Feelings thermometer. The feelings thermometer questionnaire
consisted of a scale along which a range of “degrees” were depicted,
from 0° (extremely unfavorable) to 100° (extremely favorable), with
50° labeled neither favorable nor unfavorable. Ratings were provided
for African Americans, European Americans, Asian Americans, and
Latino Americans.
Results
IAT Effects
As in Study 1, participants exhibited significant levels of implicit evaluation (M ⫽ 0.32, SD ⫽ 0.17), t(31) ⫽ 10.96, p ⬍ .001,
r ⫽ .89, and implicit stereotyping (M ⫽ 0.29, SD ⫽ 0.23), t(31) ⫽
7.24, p ⬍ .001, r ⫽ .79, yet IAT scores were not significantly
correlated, r(30) ⫽ .16, p ⫽ .37. No significant effects emerged for
sex or IAT order, Fs ⬍ 2.04, ps ⬎ .16.
IAT Effects on Behavioral Responses
To test our main hypotheses regarding double dissociations of
the stereotyping and evaluation IATs, we used hierarchical linear
regressions. First, the D score for the IAT that was not hypothesized to predict the outcome was entered as a covariate in Step 1.
In Step 2, D for the hypothesized predictor was added to the
regression model. We could then obtain evidence for a double
dissociation by examining the simultaneous effects of the two
predictors in Step 2. The semipartial r (sr) is reported as an effect
size estimate.
Stereotype Ratings
Evaluative IAT scores, entered in Step 1, did not predict stereotype ratings of the African American essay writer, ␤ ⫽ ⫺.17,
t(29) ⫽ ⫺0.90, p ⫽ .37, sr ⫽ ⫺.17. However, higher stereotyping
IAT scores were associated with more stereotypic ratings of the
African American essay writer in Step 2, ␤ ⫽ .39, t(28) ⫽ 2.70,
p ⫽ .03, sr ⫽ .39, whereas the effect of evaluative IAT scores
remained nonsignificant, ␤ ⫽ ⫺.23, t(28) ⫽ ⫺1.33, p ⫽ .20, sr ⫽
⫺.23. Ratings of nonstereotypic traits were not associated with
scores on the stereotyping IAT, ␤ ⫽ ⫺.01, t(29) ⫽ ⫺0.04, p ⫽ .97,
sr ⫽ ⫺.01, or the evaluative IAT, ␤ ⫽ ⫺.02, t(28) ⫽ ⫺0.11, p ⫽
.92, sr ⫽ ⫺.02. Finally, when nonstereotypic ratings were included
as a covariate in Step 1, stereotyping IAT scores continued to
predict stereotypic ratings, ␤ ⫽ .39, t(27) ⫽ 2.73, p ⬍ .01, sr ⫽
.38, whereas evaluative IAT scores did not, ␤ ⫽ ⫺.18, t(28) ⫽
⫺1.13, p ⫽ .26, sr ⫽ ⫺.18.
Affective Responses
In analyses of preference for the writer, stereotyping IAT scores
entered in Step 1 were not predictive of preferences, ␤ ⫽ .06,
t(29) ⫽ 0.33, p ⫽ .75, sr ⫽ .06. In Step 2, higher evaluative IAT
scores were associated with less desire to befriend the essay writer,
␤ ⫽ ⫺.32, t(28) ⫽ ⫺1.79, p ⫽ .08, sr ⫽ ⫺.32, whereas the effect
for stereotyping IAT scores remained nonsignificant, ␤ ⫽ .01,
t(28) ⫽ .04, p ⫽ .97, sr ⫽ .01, supporting our hypothesis. Participants’ feelings thermometer ratings provided an additional index
of consummatory response toward African Americans that could
be used to corroborate the marginally significant effect on preference for the writer. Average thermometer ratings for Whites,
Asians, and Latinos were entered in the first regression step as a
baseline covariate, followed by stereotyping IAT scores in Step 2
and evaluative IAT scores in Step 3. The effect for baseline
thermometer ratings was significant, ␤ ⫽ .90, t(29) ⫽ 11.40, p ⬍
IMPLICIT STEREOTYPING VS. EVALUATIVE RACE BIAS
.001, sr ⫽ .90, which reflected individual differences in scale
usage, but the effect for stereotyping IAT scores was not significant, ␤ ⫽ ⫺.10, t(28) ⫽ ⫺1.25, p ⫽ .22, sr ⫽ ⫺.10. Notably,
higher evaluative IAT scores were predictive of more negative
feelings toward African Americans, ␤ ⫽ ⫺.18, t(27) ⫽ ⫺2.04,
p ⫽ .05, sr ⫽ ⫺.16, consistent with effects for writer preference.
Additional analyses examining IAT effects on thermometer ratings
of Whites, Asians, and Latinos produced no significant effects.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Discussion
The results of Study 2 further supported the independence of
implicit stereotyping and implicit evaluation, such that scores on
the stereotyping and evaluative IATs were not significantly correlated. Furthermore, our regression analyses revealed the hypothesized double dissociation between implicit stereotyping and implicit evaluation effects. These results indicated that cognitive and
affective forms of implicit race bias are uniquely associated with
instrumental versus consummatory forms of race-biased behavior,
respectively, and hence showed discriminant and predictive validity of the stereotyping and evaluative IATs. It is notable that
although the stereotyping IAT focused on a subset of the African
American stereotype (e.g., athleticism, rhythmicity, and lack of
intelligence), it predicted a broader instantiation of the stereotype,
including the descriptors of lazy, dishonest, and (un)trustworthy,
consistent with research evidencing strong links between subcomponents of the stereotype (e.g., Devine, 1989).
Although Study 2 provided good support for our hypotheses
using conventional social psychological measures, it may have
been limited in some respects. For instance, the procedure of Study
2 did not provide a good model of how implicit race biases would
predict a White person’s responses in anticipation of a real-life
interaction with an African American. A second potential limitation was that the predictor and outcome variables were collected in
the same experimental session, precluding causal inference and
raising the possibility that the outcome measures might have
influenced IAT scores. These limitations were addressed in Study
3, in which participants completed measures of implicit stereotyping and evaluation several weeks before being recruited for a
purportedly separate experiment in which they expected to interact
with an African American participant.
Method
Participants
In the first phase of this study, participants were 43 introductory psychology students, 23 of whom were successfully recruited later in the
semester for what they believed was an unrelated study. Evaluative IAT
data from 2 participants were missing because of a computer malfunction,
leaving 21 participants (13 women, 8 men) with valid data from both
sessions. IAT scores of participants who did versus those who did not
return for Session 2 did not differ, ps ⬎ .23.
Procedure
Session 1. Participants completed stereotyping and evaluative IATs in
one of two counterbalanced orders, and the IATs were scored to yield D
scores, as in Study 2.
Session 2. Participants were told the study would involve pairs of
participants. At the scheduled experiment time, the experimenter entered
the waiting room and called out the names of the participant and the
(imaginary) partner. The partner’s name alternated between “Darnell Stewart” and “Tyrone Washington” to suggest African American ethnicity
(Greenwald et al., 1998). Noting that the partner had not yet arrived, the
experimenter escorted the participant to the experiment room to get started.
After providing consent, the participant was told the following:
We’re studying peoples’ ability to cooperate with another person on
some tasks assessing different types of general knowledge. You and a
partner are going to complete a set of tasks, and then your combined
score on these tasks will be compared with other teams who are in this
study. You should try your best on these tasks, because the teams with
the top five combined scores will be entered into a drawing for $40.
Participants were then asked to rate their abilities in various subject
areas, including their mathematic and verbal skills and their knowledge of
sports and cultural trivia. The experimenter then left momentarily, purportedly to check for the arrival of the partner. After a few minutes, the
experimenter returned to explain that the other participant had arrived and
was filling out initial questionnaires in another room. The participant was
then shown the one-page participant information form identical to that used
in Study 2. The top half was already filled in by the partner so that the
participant would see he was African American. The participant completed
the bottom half of the form.
Next, the experimenter noted they were running behind schedule and
gave the following explanation:
To save time, I’m going to have you decide which tasks you’ll do and
which your partner will do. Then we’ll all go to the main testing room.
Remember, you want to choose tasks for yourself and your partner
that will give you the best combined score, not just so that only you
or he will do well. There are four different tests: one has questions
from the math SAT, another has questions from the verbal SAT, and
the other two have questions about sports and popular culture.
Study 3
Study 3 comprised two sessions. In the first session, participants
completed IAT measures of stereotyping and evaluative race bias.
In the second, ostensibly unrelated session, participants were led to
believe that they would interact with an African American partner
on various tasks involving tests of academic (verbal and mathematic) and nonacademic (sports and popular culture) knowledge.
Participants rated how well they thought that they and their partner
would perform on each of these tasks (Ashburn-Nardo et al., 2003)
as an index of instrumental behavior. To assess consummatory
behavior, we measured the distance participants chose to sit from
the partner’s belongings in a row of chairs just prior to their
interactions (Macrae, Bodenhausen, Milne, & Jetten, 1994). We
hypothesized that implicit stereotyping but not evaluation would
predict stereotype-consistent performance expectations, whereas
implicit evaluation but not stereotyping would predict seating
distance from the partner.
657
Participants indicated which tasks they chose for themselves and their
partners and then rated their perceptions of how well they and their partners
would perform on each of the tasks.5 After leaving briefly to check up on
the supposed partner, the experimenter explained that the participant and
the partner would now meet together in another room to complete their
tasks. The experimenter led the participant out of the experiment room and,
5
Participants were rather egalitarian in their assignments of the academic tasks, with 20 of 21 participants assigning one of the SAT tasks to
themselves and the other to the partner. This pattern restricted the variance
of task assignments, and thus it was an insensitive measure of stereotypeconsistent behavior.
AMODIO AND DEVINE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
658
explaining that the partner had left momentarily to use the bathroom,
directed the participant to sit in one of a row of chairs to wait. Eight
identical chairs were arranged in a line, equally spaced approximately 4 in.
(10.16 cm) apart along the hallway. A coat and backpack putatively
belonging to the partner were placed on the chair nearest to the experiment
room doorway. After the participant chose a seat, the experimenter surreptitiously recorded the participant’s seating position and then left momentarily to photocopy the participant’s information sheet.
After returning, the experimenter explained that the session would have
to end early and led the participant back into the experiment room. The
experimenter then probed the participant for suspicion regarding the cover
story and the connection between Sessions 1 and 2, provided a debriefing
and full explanation of the procedures, and then thanked and dismissed the
participant. Two participants expressed some suspicion but were unable to
identify key aspects of the cover story, the connection between Sessions 1
and 2, or the hypotheses.
Materials
Participants rated how well they thought that they would perform on the
tests of SAT mathematic and verbal skills, sports trivia, and popular culture
on a scale ranging from 1 (very poorly) to 9 (very well). Ratings of
expected enjoyment on each task were also made on a scale ranging from
1 (not at all) to 9 (very much). Next, participants rated their expectations
of their partner’s performance and enjoyment on the same tasks, using the
same scales.
Results
IAT Scores
Participants exhibited significant levels of implicit evaluation
(M ⫽ 0.38, SD ⫽ 0.29), t(20) ⫽ 5.93, p ⬍ .001, r ⫽ .80, and
implicit stereotyping (M ⫽ 0.15, SD ⫽ 0.18), t(20) ⫽ 3.70, p ⬍
.001, r ⫽ .64. Evaluative and stereotyping IAT scores were uncorrelated, r(19) ⫽ .02, p ⫽ .93, replicating the findings of Studies
1 and 2.
Ratings of Partner Abilities and Enjoyment
An index was created to represent the extent to which the
partner was expected to perform poorly on academic tasks but to
excel on nonacademic tasks, relative to participants’ own expected
performance. Participants’ self-expectation ratings on each task
were subtracted from their partner-expectation ratings. These
scores were standardized, with ratings of counter-stereotype skills
(mathematic and verbal) reverse-scored, and averaged, such that
higher scores represented more stereotype-consistent expectations
of the partner’s performance, relative to expectations of the self.
The hypothesized double-dissociation effects were tested using
hierarchical regressions as in Study 2. In Step 1, evaluative IAT
scores were not significantly associated with expectations of the
partner’s performance, ␤ ⫽ ⫺.24, t(19) ⫽ ⫺1.08, p ⫽ .29, sr ⫽
–.24. However, in Step 2, higher stereotyping IAT scores significantly predicted more stereotype-consistent expectations for the
partner’s performance, ␤ ⫽ .47, t(18) ⫽ 2.32, p ⫽ .03, sr ⫽ .47,
whereas the effect of evaluative IAT scores remained nonsignificant, ␤ ⫽ ⫺.25, t(28) ⫽ ⫺1.24, p ⫽ .23, sr ⫽ ⫺.25.6 When
participant sex was included in Step 1 as a covariate, effects for
sex, ␤ ⫽ ⫺.18, t(18) ⫽ ⫺0.76, p ⫽ .46, sr ⫽ ⫺.17, and evaluative
IAT scores, ␤ ⫽ ⫺.19, t(18) ⫽ ⫺0.79, p ⫽ .44, sr ⫽ ⫺.18, were
not significant, whereas the effect for stereotyping IAT scores
remained significant, ␤ ⫽ .50, t(17) ⫽ 2.48, p ⫽ .02, sr ⫽ .49.
Similarly, ratings of expected partner enjoyment on more
stereotype-consistent tasks were not associated with evaluative
IAT scores in Step 1, ␤ ⫽ ⫺.06, t(19) ⫽ ⫺0.25, p ⫽ .81, sr ⫽
⫺.06, but were significantly associated with stereotyping IAT
scores in Step 2, ␤ ⫽ .44, t(18) ⫽ 2.07, p ⫽ .05, sr ⫽ .44.
Seating Distance From Partner
On average, participants sat 1.7 (SD ⫽ .78) chairs away from the
partner’s belongings. Stereotyping IAT scores, included in Step 1,
were not associated with seating distance, ␤ ⫽ ⫺.09, t(19) ⫽
⫺0.37, p ⫽ .71, sr ⫽ ⫺.09. However, as revealed in Step 2,
participants with higher evaluative IAT scores chose to sit further
from the partner’s belongings, ␤ ⫽ .44, t(18) ⫽ 2.10, p ⫽ .05, sr ⫽
.44, whereas the effect of stereotyping IAT scores remained nonsignificant, ␤ ⫽ ⫺.09, t(28) ⫽ ⫺0.45, p ⫽ .66, sr ⫽ ⫺.09,
supporting our hypothesis.
Discussion
The results of Study 3 corroborated and extended the findings of
Study 2. Greater implicit stereotyping scores uniquely predicted
more stereotype-consistent expectations for the partner’s performance, whereas greater implicit evaluation scores uniquely predicted greater seating distance from the African American partner’s belongings. These findings provided additional support for
our double-dissociation hypothesis of implicit stereotyping versus
evaluation, whereby implicit stereotyping is rooted in semantic
processes and is uniquely predictive of discrimination associated
with instrumental responses, whereas implicit evaluation is rooted
in affective processes and is uniquely predictive of discrimination
associated with consummatory responses.
The Study 3 findings allayed concerns over some potential
limitations of Study 2. First, the differential effects of implicit
stereotyping and evaluation of Study 2 were replicated in a more
realistic, ecologically valid context. Second, the two-session procedure used in Study 3 alleviated concerns regarding the order in
which measures were administered in Study 2. Moreover, because
IAT scores collected in the initial session were predictive of
behaviors weeks later, our results suggest the effects of implicit
bias are stable over time.
General Discussion
The present research produced two major findings. First, results
suggest that implicit stereotyping and evaluative race biases represent conceptually independent constructs. Despite exhibiting significant levels of bias on both implicit measures across studies,
participants’ scores on these two measures were not significantly
correlated, consistent with evidence for independent mechanisms
of basic cognitive and affective processes (Cacioppo et al., 1999;
Squire & Zola, 1996; Zajonc, 1980). Second, the results showed
that implicit stereotyping and implicit evaluation have unique
effects on alternative forms of race-biased behavior. The results of
Study 2 showed that implicit stereotyping but not evaluation was
6
IAT scores were not associated with absolute ratings of expected
performance for the self, ps ⬎ .37, or the partner, ps ⬎ .30, indicating that
implicit stereotyping effects were observable only when partner ratings
were anchored by participants’ self-reference.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
IMPLICIT STEREOTYPING VS. EVALUATIVE RACE BIAS
predictive of stereotype-consistent trait ratings of a Black student
that were based on a short writing sample. In contrast, implicit
evaluative race bias but not stereotyping was predictive of participants’ belief that they would get along with the student as a friend.
Study 3 extended these findings by focusing on participants’
behavior as they prepared to interact with an African American
partner. In this study, implicit stereotyping, but not evaluation,
predicted stereotype-consistent expectations of how well the African American partner would perform on a series of tasks. On the
other hand, implicit evaluative race bias, but not stereotyping,
predicted how far participants chose to sit from the African American partner’s belongings in a row of chairs. Although the samples
used in Studies 2 and 3 were relatively small, the use of doubledissociation designs ensured that null effects were always interpreted in the context of a complementary significant effect, and
therefore low statistical power cannot account for the pattern of
results. Taken together, these findings support the overarching
hypothesis that implicit stereotyping processes are predictive of
instrumental forms of race-biased behavior, whereas implicit evaluative processes are predictive of consummatory forms of racebiased behavior.
Implications for Theory and Research on Implicit
Race Bias
Clarifying the Construct of Implicit Race Bias
In recent years, social psychologists have grappled with the
meaning of implicit race biases in an effort to understand what
they represent, how they function, and what they may predict (cf.
Devine, 2001; Fazio & Olson, 2003). Against a backdrop of mixed
findings regarding the effects of implicit race bias on behavior
(Blair, 2001), our theorizing and results suggest that significant
effects of implicit race bias on behavior may be observed when
their underlying affective versus cognitive processes are taken into
consideration and are matched with classes of behavior associated
with consummatory versus instrumental responses (cf. Ajzen &
Fishbein, 1977). On the basis of neuroscientific research, implicit
evaluation is supported by subcortical mechanisms and is most
directly expressed in basic approach/withdrawal behaviors. By
contrast, implicit stereotyping is supported by neocortical networks and is most directly expressed in biased cognitive processing. This analysis provides a theoretical basis for conceiving of
implicit stereotyping and evaluation as independent constructs and
suggests refined definitions of these constructs that are rooted in
neural mechanisms of learning and memory. Furthermore, it suggests that the effects of implicit stereotyping versus evaluation are
likely to be expressed to different degrees in different situations
and on different assessments (Livingston & Brewer, 2002; Macrae,
Bodenhausen, Milne, Thorn, & Castelli, 1997). Although findings
to date regarding the effect of implicit race bias on behavior are
notoriously mixed, many null effects reported in the literature may
have resulted from a mismatch between forms of implicit bias with
outcome measures of discrimination.
Implications for Theory
To date, theories of implicit race bias have not addressed the
possibility that implicit forms of stereotyping and evaluation may
arise from distinct underlying processes and may affect behavior
659
via alternative routes of processing (cf. Greenwald et al., 2002).
Granted, stereotypes and affective responses are typically congruent and work together to facilitate a coordinated response (e.g.,
racial discrimination). Nevertheless, the predictive utility of a
theory depends on whether it can be used to discern underlying
processes and their respective effects on behavior. Future models
of implicit race bias will benefit from the conceptual distinction
presented here in several ways. First, a consideration of alternative
forms of implicit bias will enhance predictive validity by permitting more refined hypotheses for how different forms of implicit
bias should affect behavior. Second, our analysis links implicit
stereotyping and evaluative bias to physiological models of the
brain and behavior, permitting integration with other theoretical
approaches and suggesting appropriate physiological indicators for
different forms of bias. Indeed, previous research has associated
indices of amygdala activity with implicit evaluation (e.g., Amodio
et al., 2003; Phelps et al., 2000). Although neural correlates of
implicit racial stereotyping have not yet been determined, eventrelated potential research on stereotype-based expectancy violation
is consistent with a neocortical (versus subcortical) substrate (e.g.,
Bartholow, Fabiana, Gratton, & Bettencourt, 2001).
If implicit stereotyping and evaluation arise from distinct neural
substrates, as we proposed, it follows that they are learned and
unlearned via different mechanisms. One may refine theories of
implicit race bias malleability and change by considering the
respective dynamics of classical (fear) conditioning versus semantic associative learning. For example, human and animal models of
learning and memory suggest that implicit evaluations may be
learned more quickly and unlearned more slowly than implicit
stereotypes. They also suggest that claims that implicit prejudice
can be extinguished following a single experimental manipulation
may be implausible and that other interpretations should be considered (e.g., the manipulation inhibited the initial activation of
bias or elicited preconscious forms of regulation).
Implications for Study Design
It follows from the theoretical implications listed above that
future research will benefit from a careful selection of measures
and response contexts when examining the effects of implicit bias
on behavior. The results of the present work suggest that implicit
evaluation corresponds most directly with consummatory responses involving basic behavioral approach/withdrawal and that
these effects should be strongest when behaviors involve minimal
controlled processing. By contrast, implicit stereotyping affects
behavior by biasing cognitive processing and thus should be most
evident on measures that involve a higher degree of cognitive
processing, provided that participants are unaware of the potentially biasing effects.
Future Directions
Joint Effects of Implicit and Explicit Race Bias
Although we went to great lengths to distinguish effects of implicit
stereotyping from those of implicit evaluation in the present work,
these two forms of bias typically operate in concert. An important new
theoretical issue concerns the interplay of implicit stereotyping and
implicit evaluation: When and how do they operate in concert? For
example, behaviors that combine elements of consummatory and
AMODIO AND DEVINE
660
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
instrumental responses may be best predicted by the joint effects of
implicit stereotyping and prejudice. Additionally, there are many
situations in which explicit measures of prejudice and stereotyping
may be better predictors of behavior. Finally, although levels of
implicit stereotyping and evaluation were not correlated in our samples, these two forms of implicit bias may be more strongly correlated
among some groups of individuals (e.g., highly biased individuals)
than others. Future research is needed to explore how the full range of
discriminatory behavior may be explained by complex interactions
among implicit and explicit forms of prejudice and stereotyping for
different groups of people.
Regulatory Mechanisms for Implicit Stereotyping Versus
Implicit Evaluative Race Bias
Our findings raise new questions as to whether the behavioral
effects of implicit stereotyping and evaluation may be regulated via
different processes and whether either form of implicit bias is more
difficult to regulate. It is likely that regulation occurs at several
different levels. For example, the spreading activation of automatic
stereotypes within a semantic network could be inhibited via lateral
inhibition (Bodenhausen & Macrae, 1998). Alternatively, the effects
of implicit stereotypes may be inhibited in behavioral channels, such
that a stereotype-congruent response tendency is overridden by a
deliberative unbiased response (Amodio et al., 2004). Implicit evaluation associated with amygdala activation may be inhibited by the
countervailing activation of reward structures in the brain, or its effect
on behavior may be overridden via controlled processes as a behavioral response is formed. The inhibition of implicit stereotyping and
evaluation effects at the response-formation level likely rely on the
same frontal cortical mechanisms of control (Amodio et al., 2004;
Amodio, Kubota, Harmon-Jones, & Devine, 2006). On the other
hand, the inhibition of stereotypes within a neocortical semantic
network and evaluations within a subcortical affective network rely
on different mechanisms, and thus the parameters of regulation may
vary considerably. The present theoretical analysis highlights some
previously unexplored complexities regarding mechanisms for regulating the effects of implicit race bias.
Conclusion
Affect and cognition represent two fundamental processes of the
human mind, and the distinction between affective and cognitive
processes is critical for the understanding of a wide range of
psychological functions (Cacioppo et al., 1999). On the basis of
past social psychological and neuroscientific theories, we showed
that cognitive and affective components of implicit race bias are
conceptually independent and are uniquely predictive of instrumental and consummatory forms of race-biased behaviors, respectively. The present work is also notable in that we took a social
neuroscientific approach: We applied neurocognitive models of
learning and memory to elucidate social psychological conceptions
of implicit processes that had been poorly defined. Our findings
suggest that greater conceptual clarity in implicit race bias research
may be achieved by considering the differential effects of implicit
stereotyping and evaluation when interpreting extant findings,
developing new theories, and designing future research.
References
Ajzen, I., & Fishbein, M. (1977). Attitude– behavior relations: A theoretical analysis and review of empirical research. Psychological Bulletin,
84, 888 –918.
Allport, G. W. (1954). The nature of prejudice. Reading, MA: AddisonWesley.
Amodio, D. M., & Frith, C. D. (2006). Meeting of minds: The role of the
medial frontal cortex in social cognition. Nature Reviews Neuroscience,
1, 268 –277.
Amodio, D. M., Harmon-Jones, E., & Devine, P. G. (2003). Individual
differences in the activation and control of affective race bias as assessed
by startle eye blink responses and self-report. Journal of Personality and
Social Psychology, 84, 738 –753.
Amodio, D. M., Harmon-Jones, E., Devine, P. G., Curtin, J. J., Hartley,
S. L., & Covert, A. E. (2004). Neural signals for the detection of
unintentional race bias. Psychological Science, 15, 88 –93.
Amodio, D. M., Kubota, J. T., Harmon-Jones, E., & Devine, P. G. (2006).
Alternative mechanisms for regulating racial responses according to
internal vs. external cues. Social Cognitive and Affective Neuroscience,
1, 26 –36.
Ashburn-Nardo, L., Knowles, M. L., & Monteith, M. J. (2003). Black
Americans’ implicit racial associations and their implications for intergroup judgment. Social Cognition, 21, 61– 87.
Bartholow, B. D., Fabiani, M., Gratton, G., & Bettencourt, B. A. (2001). A
psychophysiological analysis of cognitive processing of and affective responses to social expectancy violations. Psychological Science, 12, 197–204.
Bechara, A., Damasio, H., & Damasio, A. R. (2003). Role of the amygdala
in decision-making. Annual Review of Neuroscience, 985, 356 –369.
Blair, I. (2001). Implicit stereotypes and prejudice. In G. Moskowitz (Ed.),
Cognitive social psychology: On the tenure and future of social cognition (pp. 359 –374). Mahwah, NJ: Erlbaum.
Bodenhausen, G. V., & Macrae, C. N. (1998). Stereotype activation and
inhibition. In R. S. Wyer, Jr. (Ed.), Advances in social cognition (Vol.
11, pp. 1–52). Mahwah, NJ: Erlbaum.
Bouton, M. E. (1994). Conditioning, remembering, and forgetting. Journal
of Experimental Psychology: Animal Behavior Processes, 20, 219 –231.
Cacioppo, J. T., Gardner, W. L., & Berntson, G. G. (1999). The affect system
has parallel and integrative processing components: Form follows function.
Journal of Personality and Social Psychology, 76, 839 – 855.
Damasio, A. D. (1994). Descartes’ error: Emotion, reason, and the human
brain. New York: Avon.
Davis, M., & Whalen, P. J. (2001). The amygdala: Vigilance and emotion.
Molecular Psychiatry, 6, 13–34.
Devine, P. G. (1989). Prejudice and stereotypes: Their automatic and
controlled components. Journal of Personality and Social Psychology,
56, 5–18.
Devine, P. G. (2001). Implicit prejudice and stereotyping: How automatic
are they? Introduction to the special section. Journal of Personality and
Social Psychology, 81, 757–759.
Devine, P. G., & Elliot, A. J. (1995). Are racial stereotypes really fading?
The Princeton Trilogy revisited. Personality and Social Psychology
Bulletin, 21, 1139 –1150.
Devine, P. G., Plant, E. A., Amodio, D. M., Harmon-Jones, E., & Vance,
S. L. (2002). The regulation of explicit and implicit race bias: The role
of motivations to respond without prejudice. Journal of Personality and
Social Psychology, 82, 835– 848.
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.
Dovidio, J. F., Brigham, J. C., Johnson, B. T., & Gaertner, S. L. (1996).
Stereotyping, prejudice and discrimination: Another look. In C. N.
McCrae, C. Stangor, & M. Hewstone (Eds.), Stereotypes and stereotyping (pp. 276 –319). New York: Guilford.
Dovidio, J. F., Esses, V. M., Beach, K. R., & Gaertner, S. L. (2004). The
role of affect in determining intergroup behavior: The case of willing-
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
IMPLICIT STEREOTYPING VS. EVALUATIVE RACE BIAS
ness to engage in intergroup affect. In D. M. Mackie & E. R. Smith
(Eds.), From prejudice to intergroup emotions: Differentiated reactions
to social groups (pp. 153–171). Philadelphia: Psychology Press.
Dovidio, J. F., Evans, N., & Tyler, R. B. (1986). Racial stereotypes: The
contents of their cognitive representations. Journal of Experimental
Social Psychology, 22, 22–37.
Dovidio, J. F., Kawakami, K., & Gaertner, S. L. (2002). Implicit and
explicit prejudice and interracial interaction. Journal of Personality and
Social Psychology, 82, 62– 68.
Esses, V. M., & Dovidio, J. F. (2002). The role of emotions in determining
willingness to engage in intergroup contact. Personality and Social
Psychology Bulletin, 28, 1202–1214.
Fazio, R., Jackson, J., Dunton, B., & Williams, C. (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.
Fazio, R. H., & Olson, M. A. (2003). Implicit measures in social cognition
research: Their meaning and uses. Annual Review of Psychology, 54,
297–327.
Feldman, S., Conforti, N., & Weidenfeld, J. (1995). Limbic pathways and
hypothalamic neurotransmitters mediating adrenocortical responses to
neural stimuli. Neuroscience and Biobehavioral Reviews, 19, 235–240.
Fiske, S. T. (1998). Stereotyping, prejudice, and discrimination. In D. T.
Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social
psychology (Vol. 2, pp. 357– 411). New York: McGraw-Hill.
Gabrieli, J. D. (1998). Cognitive neuroscience of human memory. Annual
Review of Psychology, 49, 87–115.
Greenwald, A., McGhee, D., & Schwartz, J. (1998). Measuring individual
differences in implicit cognition: The Implicit Association Test. Journal
of Personality and Social Psychology, 74, 1464 –1480.
Greenwald, A. G., & Banaji, M. R. (1995). Implicit social cognition.
Psychological Review, 102, 4 –27.
Greenwald, A. G., Banaji, M. R., Rudman, L. A., Farnham, S. D., Nosek,
B. A., & Mellott, D. S. (2002). A unified theory of implicit attitudes,
stereotypes, self-esteem, and self-concept. Psychological Review, 109, 3–25.
Greenwald, A. G., Nosek, B. A., & Banaji, M. R. (2003). Understanding
and using the Implicit Association Test: I. An improved scoring algorithm. Journal of Personality and Social Psychology, 85, 197–216.
Hamilton, D. L. (1981). Stereotyping and intergroup behavior: Some
thoughts on the cognitive approach. In D. L. Hamilton (Ed.), Cognitive
processes in stereotyping and intergroup behavior (pp. 333–353). Hillsdale, NJ: Erlbaum.
Kawakami, K., Dion, K. L., & Dovidio, J. F. (1998). Racial prejudice and
stereotype activation. Personality and Social Psychology Bulletin, 24,
407– 416.
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1990). Emotion, attention,
and the startle reflex. Psychological Review, 97, 377–395.
LeDoux, J. E. (2000). Emotion circuits in the brain. Annual Review of
Neuroscience, 23, 155–184.
Lepore, L., & Brown, R. (1997). Category and stereotype activation: Is
prejudice inevitable? Journal of Personality and Social Psychology, 72,
275–287.
Livingston, R. W., & Brewer, M. B. (2002). What are we really priming?
Cue-based versus category-based processing of facial stimuli. Journal of
Personality and Social Psychology, 82, 5–18.
Mackie, D. M., & Smith, E. R. (1998). Intergroup relations: Insights from a
theoretically integrative approach. Psychological Review, 105, 499 –529.
Macrae, C. N., Bodenhausen, G. V., Milne, A. B., & Jetten, J. (1994). Out
of mind but back in sight: Stereotypes on the rebound. Journal of
Personality and Social Psychology, 67, 808 – 817.
661
Macrae, C. N., Bodenhausen, G. V., Milne, A. B., Thorn, T. M. J., &
Castelli, L. (1997). On the activation of social stereotypes: The moderating role of processing objectives. Journal of Experimental Social
Psychology, 33, 471– 489.
Malpass, R. S., Lavigueur, H., & Weldon, D. E. (1973). Verbal and visual
training in face recognition. Perception and Psychophysics, 14, 285–292.
McConahay, J. B., & Hough, J. C. (1976). Symbolic racism. Journal of
Social Issues, 32, 23– 45.
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.
Millar, M. G., & Tesser, A. (1986). Effects of affective and cognitive focus
on the attitude-behavior relation. Journal of Personality and Social
Psychology, 51, 270 –276.
Millar, M. G., & Tesser, A. (1989). The effects of affective-cognitive
consistency and thought on the attitude-behavior relation. Journal of
Experimental Social Psychology, 25, 189 –202.
Moreno, K. N., & Bodenhausen, G. V. (2001). Intergroup affect and social
judgment: Feelings as inadmissible information. Group Processes and
Intergroup Relations, 4, 21–29.
Park, B. & Judd, C. M. (2005). Rethinking the link between categorization
and prejudice within the social cognition perspective. Personality and
Social Psychology Review, 9, 108 –130.
Phelps, E. A., O’Connor, K. J., Cunningham, W. A., Funayama, S.,
Gatenby, J. C., Gore, J. C., & Banaji, M. R. (2000). Performance on
indirect measures of race evaluation predicts amygdala activation. Journal of Cognitive Neuroscience, 12, 729 –738.
Rissman, J., Eliassen, J. C., & Blumstein, S. E. (2003). An event-related
fMRI investigation of implicit semantic priming. Journal of Cognitive
Neuroscience, 15, 1160 –1175.
Rudman, L. A., Ashmore, R. D., & Gary, M. L. (2001). “Unlearning”
automatic biases: The malleability of implicit stereotypes and prejudice.
Journal of Personality and Social Psychology, 81, 856 – 868.
Rudman, L. A., Greenwald, A. G., & McGhee, D. E. (2001). Implicit
self-concept and evaluative implicit gender stereotypes: Self and ingroup
share desirable traits. Personality and Social Psychology Bulletin, 27,
1164 –1178.
Spencer, S. J., Fein, S., Wolfe, C. T., Fong, C., & Dunn, M. A. (1998).
Automatic activation of stereotypes: The role of self-image threat. Personality and Social Psychology Bulletin, 24, 1139 –1152.
Squire, L. R., & Zola, S. M. (1996). Structure and function of declarative
and nondeclarative memory systems. Proceedings of the National Academy of Sciences of the USA, 93, 13515–13522.
Stangor, C., Sullivan, L. A., & Ford, T. E. (1991). Affective and cognitive
determinants of prejudice. Social Cognition, 9, 359 –380.
Wilson, T., Lindsey, S., & Schooler, T. (2000). A model of dual attitudes.
Psychological Review, 107, 101–126.
Wittenbrink, B., Judd, C. M., & Park, B. (1997). Evidence for racial
prejudice at the implicit level and its relationship with questionnaire
measures. Journal of Personality and Social Psychology, 72, 262–274.
Wittenbrink, B., Judd, C. M., & Park, B. (2001). Evaluative versus conceptual judgments in automatic attitude activation. Journal of Experimental Social Psychology, 37, 244 –252.
Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences.
American Psychologist, 35, 151–175.
Received April 29, 2004
Revision received January 9, 2006
Accepted January 27, 2006 䡲
Journal of Experimental Social Psychology 43 (2007) 399–409
www.elsevier.com/locate/jesp
Plausible assumptions, questionable assumptions and post
hoc rationalizations: Will the real IAT, please stand up?
Hart Blanton
a,*
, James Jaccard b, Charlene Christie c, Patricia M. Gonzales
d
a
Texas A&M University, Department of Psychology, 4235 TAMU, College Station, TX 77843-4235, USA
Florida International University, Department of Psychology, University Park, Miami, Florida 33199, USA
Indiana University—Purdue University, Department of Psychology, 4601 Central Avenue, Columbus, Indiana 47203, USA
d
New York State Energy Research and Development Authority, 17 Columbia Circle, Albany, NY 12203-6399, USA
b
c
Received 10 October 2006
Available online 27 December 2006
Abstract
In a recent article, we described psychometric limitations to the Implicit Association Test (IAT). These limitations restrict the utility of
this measure and render it problematic for testing many psychological theories that posit a causal role for implicit attitudes. Past failures
to recognize this may have promoted faulty conclusions in the literature. In a critique of our article, Nosek and Sriram rejected our entire
analysis. They asserted that our original article was based on faulty assumptions and argued that the IAT performs nicely when these
assumptions are replaced by other, more plausible assumptions. We show that these plausible assumptions have all the hallmarks of post
hoc rationalizations. They make little theoretical sense, are buttressed by deceptive statistical practices, contradict statements these same
researchers have made in the past and do little to advance research and theory on implicit attitudes. We close by considering the vigor
with which IAT researchers have dismissed meaningful criticisms of their measure.
Ó 2006 Elsevier Inc. All rights reserved.
Keywords: Implicit association test; Psychometrics; Theory testing; Validity; Prejudice; Discrimination
In a recent article (Blanton, Jaccard, Gonzales, & Christie, 2006), we pointed out flaws in the Implicit Association
Test (IAT), noting that the measure rests on a dubious psychometric foundation and that its design prevents it from
testing many psychological theories that are of interest to
researchers. In their critique of our article, Nosek and Sriram (2007) stated that our analysis was flawed. They suggested that we invoked faulty assumptions and asserted
that the IAT becomes psychometrically and theoretically
sound if one replaces our faulty assumptions with assumptions that they deem to be plausible. We show here, however, that these plausible assumptions have properties
suggesting they are post hoc rationalizations: They make
little theoretical sense and contradict statements these same
*
Corresponding author.
E-mail address: hblanton@psych.tamu.edu (H. Blanton).
0022-1031/$ – see front matter Ó 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.jesp.2006.10.019
researchers have made in other contexts. We also show
how Nosek and Sriram seem to lose sight of the purpose
of measurement by elevating the IAT above its potential
utility. Instead of discussing ways that the IAT might be
used as a tool to answer important questions, these
researchers suggested ways that research programs might
be shaped so that the IAT can be used.
Theory testing
Consider some psychological theories that might be of
interest to a psychologist. Fig. 1 presents examples using
the traditional schematic for representing causal models.
The top panel presents the causal model that would be tested by a researcher who wants to predict college students’
identification with math from their attitudes towards math
and arts. Based on the path coefficients that this researcher
400
H. Blanton et al. / Journal of Experimental Social Psychology 43 (2007) 399–409
Fig. 1. Causal models using component parts. Note 0 indicates path is set
to zero. + Indicates path is a direct relationship.
Indicates path is an
inverse relationship.
has hypothesized, math identification should be predicted
from the attitude towards math, not the attitude towards
arts. This pattern would be consistent with research on attitude-behavior correspondence (Jaccard, 1975; Jaccard,
King, & Pomazal, 1977).1
The second panel presents a causal model of interest to a
researcher who wishes to predict the tendency for white
participants to act in a hostile manner towards a white versus a black confederate. The hypothesized path coefficients
suggest that this researcher believes that hostility towards a
black confederate will be driven by attitudes towards
1
This data pattern would also be consistent with the results that
Blanton et al. (2006) reported in their replication and extension of (Nosek,
Banaji, & Greenwald, 2002).
blacks, not attitudes towards whites and that hostility
towards whites will be unaffected by attitudes towards
blacks. These predictions are consistent with research on
attitude-behavior correspondence. It is further hypothesized, however, that correspondence will not be supported
in the causal relationship between attitude towards whites
and hostility towards whites. Based on research suggesting
that members of majority groups do not have strong or
consequential attitudes towards their in-groups (Blanton
& Christie, 2003; Mullen, 1991), the researcher predicts
that attitudes towards whites will not exert significant influence on whites’ treatment of a white confederate.
The third panel presents a causal model of interest to a
researcher trying to predict adolescent condom use from
their attitudes towards condoms and their attitudes
towards other methods of pregnancy protection, in this
case, the contraceptive pill. The hypothesized path coefficients indicate that a more positive attitude towards condoms will predict greater use of condoms. However,
based on past research suggesting that attitudes towards
other protection methods are important to consider, the
researcher predicts that attitudes toward the pill will moderate the relationship between condom attitudes and condom use: The more positive the attitude toward the pill,
the weaker the impact of condom attitudes on condom
use behavior.
Each of these theories predicts a reasonably complex
causal dynamic based in sound psychological theory. Let’s
now compare these models with the models that researchers would have to test if they oriented their theories around
implicit attitudes and used the IAT as their research tool.
These models appear in the three panels in Fig. 2. To avoid
the accusation that we have built implausible assumptions
into our analyses, we were careful to use the causal form
that Nosek and Sriram promoted in their article. Nosek
and Sriram proposed that, when testing the causal influence of implicit math and arts attitudes on math identification, a researcher using the IAT would test for the influence
of a single IAT factor on this outcome. They labeled this
single factor the ‘‘latent math-arts attitude’’ (p. 397). This
new attitude construct would replace the two math and arts
attitude constructs that were originally of interest to the
researcher (see Nosek and Sriram’s Fig. 2 for comparison
with the first panel of our Fig. 2).
What is a ‘‘math-arts attitude’’ and where did it come
from? And, more to the point, why is psychological theory
being wrapped around this new construct? On the face of
it, the ‘‘math-arts attitude’’ is a construct of questionable
value. Metaphorically speaking, it mixes apples with oranges. In the original causal models that we presented in Fig. 1,
the causal influence of math attitudes (apples) was assumed
to be distinct from the causal influence of arts attitudes
(oranges). This makes both common and theoretic sense.
But Nosek and Sriram would have psychologists ignore
these facts and mix these two psychological constructs
together into one big theoretical fruit bowl. This also
would occur in the other causal models. In trying to predict
H. Blanton et al. / Journal of Experimental Social Psychology 43 (2007) 399–409
Fig. 2. Causal model using preferences.
the nonverbal treatment of white and black confederates,
for instance, a researcher who uses the black-white IAT
would not be able to examine the independent influences
of black attitudes versus white attitudes. Instead, a
researcher would have to adopt a psychological theory that
focused attention on a single ‘‘black-white attitude.’’ The
resulting causal model is represented in the second panel
of Fig. 2. Similarly, instead of testing a causal model in
which attitudes towards condoms interacts with attitudes
towards the pill in the prediction of condom use, a
researcher who uses the condom-pill IAT must orient theory around a single ‘‘condom-pill attitude.’’ The model
that results is shown in the third panel of Fig. 2.
By comparing the models in Fig. 1 with the models that
the IAT can test in Fig. 2, it is evident that the IAT is a limited research tool. One goal of our original article was to
make this evident to researchers so that they would not
apply the IAT in situations where it lacks utility. Instead
of exploring the ramifications of the limited causal model
built into the IAT, Nosek and Sriram embraced it.
The irrelevance of a one factor result
Whether one is most interested in testing the causal
models in Fig. 1 or the causal models in Fig. 2, the theory
one ultimately chooses will have consequences for the type
of attitude measures one must use. Consider the causal
401
model in the first panel of Fig. 1. This model requires the
measurement of two distinct attitudes or, stated another
way, it requires attitude measures that have a two-factor
structure. One set of items should provide a valid estimate
of each participant’s ‘‘math attitude’’ and the other set of
items should provide a valid estimate of each participant’s
‘‘arts attitude.’’ If this two-factor structure is not observed
for the attitude measures, then they cannot be used to test
the model of interest. Now consider the causal model in the
top panel of Fig. 2. This model requires an attitude measure that has a single-factor structure. If a single-factor
structure is not observed, then the measure is reflecting
more than the single construct of interest and this is potentially problematic.
In Blanton et al. (2006), we argued that IAT researchers
have ignored the psychometric structure of the component
parts of the IAT and instead have moved directly to testing
the relatively crude causal model that was built into it.
With this in mind, we evaluated the psychometric properties of the component parts of the IAT to determine if
the factor structure was consistent with the psychometric
model we thought should underlie it. We found that the
two components had a multidimensional factor structure,
which was inconsistent with the Nosek and Sriram causal
model.
Nosek and Sriram objected our analysis and argued that
the psychometric model we invoked was not appropriate
for the IAT. Ostensibly, this was because the act of computing a difference score from the IAT components creates
a meaningful and somewhat mystical aspect of relative attitudes (or preference) that cannot be captured from the
component parts of the IAT. They re-scored our data using
their D method (as we discuss later), then formed multiple
instantiations of the difference scores, and showed that
these multiple instantiations mapped onto a one-factor
model. In essence, they turned a multi-factor solution into
a single-factor solution. In the process, they also argued
that something special was being captured by the differenced IAT indicators. Although we are heartened that
our article has forced IAT theorists to be somewhat more
explicit about their psychometric assumptions, we show
below that their algebraic manipulations reveal nothing
special and that all they have succeeded in doing is mixing
apples and oranges.
Nosek and Sriram’s justification: The whole is greater than
the sum of its parts
Nosek and Sriram’s justification for treating their data
in the way they did revolved around a hypothesized property of a theoretical construct they called a relative preference. They argued that, in many instances, a relative
preference cannot be understood by studying the
component attitudes that make up this preference. As they
said, ‘‘relative preferences do not necessarily reduce to
component attitudes.’’ (p. 393) In other words, the relative-preference whole is greater than the sum of its attitu-
402
H. Blanton et al. / Journal of Experimental Social Psychology 43 (2007) 399–409
dinal parts. With this justification in hand, Nosek and Sriram performed the data manipulations that mixed apples
with oranges and found what they were looking for—that
math-arts IAT has a unidimensional factor structure. As
noted, they named this factor a ‘‘latent math-arts
attitude.’’
One of the more troubling features of this logic is that it
is in direct contradiction to the way IAT researchers have
treated relative preferences in the past. This suggests that
this response is but a post hoc justification. To begin, suppose that global relative preference measures do have some
special quality about them. As a result, there will be times
in which relative preferences cannot be detected by measurement strategies that focus on the component parts that
make up these relative preferences. Assuming this is true—
and assuming that awareness of this fact is one reason that
the IAT originally was designed to measure relative preferences—then IAT researchers certainly would have worked
under this philosophy. But they have not. In fact, IAT
researchers routinely estimate relative attitudes by focusing
their analyses on the component parts that make up a relative preference. And there is no greater offender than
Nosek.
We examined every empirical article that Nosek has
published (as retrieved by PsychInfo). In every article that
had a measure of explicit attitudes, he pursued a measurement strategy (in one or more instances) whereby the component attitudes underlying explicit relative preferences
were measured separately and then differenced to create a
single estimate of explicit relative preference. Moreover,
these computed difference scores were presented by Nosek
as the conceptual equivalent of the relative preference
assessed by the IAT. As one example, Nosek (2005)
described his strategy for assessing relative explicit attitudes as follows:
‘‘Explicit preferences were assessed by calculating the
difference between feelings of warmth ratings for the
two social objects to conceptually parallel the relative
measurement feature inherent in the IAT.’’ (p. 570; italics
added)
This statement could not be more clear. In fact, it mirrors statements in Nosek et al. (2002), the article that we
critiqued in our original article. Recall that our analyses
focused on the math-arts IAT. And recall that Nosek
and Sriram argued that this attitude could not be understood if it is reduced to its component parts. Now consider
Nosek, Greenwald and Banaji’s original statements about
their strategy for measuring the explicit math-arts
preference:
‘‘To assess explicit attitudes toward math and arts, we
had participants complete paper-and-pencil questionnaires. Specifically, we used feeling thermometers (preference ratings based on a 0–100 scale from cold/
unfavorable to warm/favorable) to assess participants’
feelings of warmth toward math and arts as academic
domains. By taking the difference between the math and
arts temperature ratings, we made the explicit attitude
measures comparable to the implicit measures.’’ (p. 48;
italics added)
In both of these articles, Nosek conceptualized an
explicit relative preference as something that could be
understood by examination of it component parts.2 But
then Nosek and Sriram took us to task for doing exactly
the same thing. In fact, this objection was the primary
reason Nosek and Sriram rejected the results of our second
study. This study measured the component parts of
implicit attitudes and showed a case in which implicit attitude structure does not support the IAT measurement
strategy.
If relative attitudinal preferences have some mysterious
aspect to them that cannot be assessed by measuring the
component parts, and if Nosek knew this to be true, then
his own work on explicit measures is flawed. By his own
logic, psychologists must now call into question his past
conclusions regarding the properties of explicit attitudinal
preferences. They must also call into question past conclusions he has made regarding the IAT. This is because the
component explicit attitude measures were used by Nosek
and colleagues as a criterion when they validated a new
scoring algorithm for the IAT (Greenwald, Nosek, &
Banaji, 2003). But if the criterion measures were invalid,
then this casts doubt on the new scoring algorithm.
Intransitive preferences
It is not uncommon for IAT advocates to make sweeping
statements about precedent in other areas of psychology and
then to suggest that IAT critics lack knowledge or appreciation of these findings (see Banaji, Nosek, & Greenwald,
2004; Greenwald, Nosek, Banaji, & Klauer, 2005; Greenwald, Nosek, & Sriram, 2006; Greenwald, Rudman,
Nosek, & Zayas, 2006). There were multiple instances of
this in Nosek and Sriram. To justify their view of relative
preferences, for instance, they cited phenomena in other
domains that seem consistent with the assertion that a general preference often is not predictable from its component
attitudes. Space constraints limit our discussion of this
issue, but consider the literature they cite on intransitive
preferences. This literature is generally concerned with
how changes in the contexts in which people rate their preferences can alter their preferences (sometimes dramatically
so). Finding such contextual effects says nothing about
whether a global preference is a function of its component
attitudes. In fact, not one of the studies Nosek and Sriram
cite focused on traditional attitudinal constructs, nor did a
single study test if a relative preference rating could be predicted from its component parts.
2
The IAT demonstration webpages also assess the component attitudes
separately.
H. Blanton et al. / Journal of Experimental Social Psychology 43 (2007) 399–409
403
An empirical evaluation
Table 1
Factor loadings for factor analysis of apple–orange items
We now present data that explores issues central to
Nosek and Sriram’s critique. In a study using a convenience sample of 132 male and female college students,
we administered an attitude inventory that measured two
attitudes that a consumer psychologist might target with
an IAT task. These were (1) attitudes towards apples and
(2) attitudes towards oranges. The attitude towards apples
was measured with three (11-point) items and the attitude
towards oranges was measured with three separate (11point) items (see the Appendix for all measures used). An
attitude towards apples was operationalized as the sum
of the three apple items (alpha coefficient = 0.91) and an
attitude towards oranges was operationalized as the sum
of the three orange items (alpha coefficient = 0.91). In
addition, we obtained a direct global preference rating
for apples relative to oranges, in which participants rated
on one scale how much they preferred apples relative to
oranges or vice versa. The goal of the study was to predict
liking of fruit products (which might be of interest to
applied researchers who wish to increase fruit consumption). The psychological criteria we sought to predict from
attitudes focused on the liking of three different fruit-related
products. These were, (1) liking of orange marmalade, (2)
liking of applesauce, and (3) liking of fruit salad.
Question
Predicting preference from component attitudes
According to the whole-is-greater-than-the-sum-of-itsparts logic of Nosek and Sriram, the one-item preference
rating for apples relative to oranges should not be predictable from (or at least it should not be highly correlated
with) the component attitude towards apples and attitude
toward oranges. We tested this by regressing the preference
rating for apples relative to oranges onto the two individual
attitude measures. We observed a multiple correlation of
0.86 (p < .05). By any conventional standard, this represents a high degree of association. The regression coefficient for the apple attitude was 0.46 and the regression
coefficient for the orange attitude was 0.49, with both
being highly statistically significant. This calls into question
Nosek and Sriram (2007) assertion that the relative preference judgment captures something special that cannot be
addressed with the component attitudes themselves.3 We
present additional data below that further question this
assertion.
The factor structure of the six attitudinal items
To explore the factor structure of the six attitude items
(three for apples and three for oranges), we conducted a
traditional exploratory factor analysis. The factors were
3
Interestingly, however, these results also support the procedures in
Nosek (2005), where relative preferences were estimated by computing a
difference between two component attitudes.
Liking the taste of oranges
Favorable feeling towards oranges
Pleasure in orange eating experience
Liking the taste of apples
Favorable feeling towards apples
Pleasure in apple eating experience
Factor 1
Factor 2
.859
.948
.824
.005
.003
.008
.036
.092
.036
.894
.937
.821
extracted using a maximum likelihood algorithm with an
oblimin rotation to allow for correlated factors. Examination of the eigenvalues indicated a two-factor solution
accounting for 85% of the input variance-covariance. The
two extracted factors were correlated 0.16.4 Table 1 presents the pattern matrix representing the factor loadings.
It can be seen that the items segregate well along the lines
of the two latent attitudes. This finding is consistent with
the two-factor structure shown in Fig. 1 in Nosek and Sriram. The factor loadings for items measuring the same attitude are large on one factor and negligible on the other.
Consistent with the difference score approach in Nosek
and Sriram, we next formed three difference scores using
the six items so as to reflect the preference for one attitude
object over another. Specifically, the preference was operationalized as the difference in the attitude item for one
object (apples) minus the attitude item for the other object
(oranges). We thus created three difference scores (the taste
item for apples minus the taste item for oranges, the favorable item for apples minus the favorable item for oranges,
and the eating experience item for apples minus the eating
experience item for oranges). We factor analyzed these
three difference scores along with the global preference item
that directly assessed the preference.
As expected, the two-factor structure that was built into
the questionnaire was eliminated and a single factor that
accounted for 87% of the input variance-covariance
emerged. Table 2 presents the factor loadings for this solution. It can be seen that the factor loadings are large and
that the items seem to reflect well a single underlying preference. These analyses show that there is nothing special
about obtaining a one-factor solution via the Nosek and
Sriram scoring method. Even though common sense dictates that attitudes towards apples and attitudes towards
oranges are psychologically and psychometrically distinct
from one another, the Nosek and Sriram approach effectively mixes apples and oranges. Note also that the single-item global preference rating loaded highly on the
same factor. Contrary to Nosek and Sriram, the one factor
solution in no way points to the existence of some unique
and unspecified aspect of attitudinal preference.
4
Further probes of the data uncovered some evidence for small amounts
of method variance in the pairs of items (i.e., the two taste items, the two
favorability items, and the two eating experience items) but this method
variance was inconsequential and affects none of the conclusions reached
from these data.
404
H. Blanton et al. / Journal of Experimental Social Psychology 43 (2007) 399–409
Table 2
Factor loadings for factor analysis of apple–orange differenced items
Question
Loadings
Difference in liking
Difference in favorability
Difference in pleasure
Relative preference
.930
.976
.863
.873
Predicting criteria from preference versus component
attitudes
We next conducted analyses to predict our three criteria
(liking of applesauce, liking of orange marmalade, and liking of fruit salad) from the attitude measures. We first tested the causal model that Nosek and Sriram advocated in
their Fig. 2 (and presented in our Fig. 2). We then tested
a model that respected the two-factor structure of our questionnaire (consistent with our Fig. 1).
The Nosek–Sriram causal model
We first fit a path model predicting the criteria from the
relative preference, operationalized by the difference of the
two attitude measures.5 As shown in Fig. 3, the path coefficient from the relative preference to the liking of orange
marmalade was statistically significant (unstandardized
coefficient = .065,
standardized
coefficient = .24,
p < .005).6 The parameter for this effect was in a direction
that suggested that, as preference moved more towards preferring oranges over apples, people were more apt to like
orange marmalade. The path coefficient from the preference variable to liking of applesauce was not statistically
significant, nor was the path coefficient linking the preference variable to the liking of fruit salad.
Before turning to the causal model that used the two
attitudes separately, it is informative to consider how psychologists might react to the results of the previous analysis. The results in Fig. 3 might lead psychologists to
develop complex psychological theories that explain why
the tendency to prefer oranges over apples was significantly
related to the liking of orange marmalade but not significantly related to the liking of applesauce. Researchers
might further develop theories to account for the irrelevance of relative preference factor in the liking of fruit
salad. Perhaps people do not think of apples or oranges
when they think of fruit salad. Instead their minds wander
to more exotic fruits, such as kiwis. In short, the mere fact
that the IAT invokes a crude causal model in research
investigations in no way prevents researchers from developing nuanced theories that seem to account for their
findings.
5
We allowed for correlated errors among the criteria to reflect the
obvious fact that the correlations between the criteria are impacted by
factors other than simply the IAT inspired preference variable (e.g., liking
of fruit in general).
6
This is a just-identified model, hence fit indices are irrelevant.
The two-factor model
For this analysis, we fit a causal model tha…
Purchase answer to see full
attachment

We offer the bestcustom writing paper services. We have done this question before, we can also do it for you.

Why Choose Us

  • 100% non-plagiarized Papers
  • 24/7 /365 Service Available
  • Affordable Prices
  • Any Paper, Urgency, and Subject
  • Will complete your papers in 6 hours
  • On-time Delivery
  • Money-back and Privacy guarantees
  • Unlimited Amendments upon request
  • Satisfaction guarantee

How it Works

  • Click on the “Place Order” tab at the top menu or “Order Now” icon at the bottom and a new page will appear with an order form to be filled.
  • Fill in your paper’s requirements in the "PAPER DETAILS" section.
  • Fill in your paper’s academic level, deadline, and the required number of pages from the drop-down menus.
  • Click “CREATE ACCOUNT & SIGN IN” to enter your registration details and get an account with us for record-keeping and then, click on “PROCEED TO CHECKOUT” at the bottom of the page.
  • From there, the payment sections will show, follow the guided payment process and your order will be available for our writing team to work on it.