a."

v . .2
.74. at...
r

.1

 

$4.

$4

 

 

«3....
\

 

54007i75/

This is to certify that the
dissertation entitled

SUBTYPING AND SUBGROUPING: ANTECEDENTS AND
CONSEQUENCES OF THE RECATEGORIZATION OF
COUNTERSTEREOTYPIC TARGETS

presented by
Jeanette M. Renaud

has been accepted towards fulﬁllment
of the requirements for the

PhD. degree in Social Psychology

 

 

or’s Signature

@13qu
We»

llvﬁrpabos

Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

-------—--.-o--n-.-.—.-.--

 

 

. LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c-JCIRCIDateDuepss-p. 15

SUBTYPING AND SUBGROUPING: ANTECEDENTS AND CONSEQUENCES
OF THE RECATEGORIZATION OF COUNTERSTEREOTYPIC TARGETS

By

Jeanette M. Renaud

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Psychology

2003

ABSTRACT

SUBTYPING AND SUBGROUPING: ANTECEDENTS AND CONSEQUENCES
OF THE RECATEGORIZATION OF COUNTERSTEREOTYPIC TARGETS

By

Jeanette M. Renaud

Two studies were conducted in an attempt to address some important
issues regarding the antecedents and consequences of categorizing target group
members who violate a stereotype about the overall group. In particular, Study 1
examined the cognitive consequences of subtyping versus subgrouping atypical
group members. It was expected that when an atypical group member was
perceived as a subtype of the overall group, the overall group category would be
inhibited. However, when an atypical group member was perceived as a
subgroup of the overall group, the overall category would be relatively more
activated. Study 2 addressed factors that might inﬂuence whether perceivers
would subtype or subgroup atypical counterstereotypic target group members. It
was predicted that perceivers’ categorization of such targets would vary in the
service of maintaining their overall group prejudice. Unfortunately, no evidence
consistent with these predictions was found in either study. However, some

unexpected results were found, and their implications are discussed.

Dedicated to Susanne Atherton, my strongest supporter

ACKNOWLEDGMENTS

First and foremost, I would like to thank my advisor and the chair of my
dissertation committee, Allen McConnell. Without your support and
understanding over these past couple of years, this dissertation would not have
materialized. Again, thank you so much for believing in me when I didn’t believe
in myself!

I would also like to thank Norb Kerr for “co-chairing" my dissertation. You
always seem to be able to push me a little further by questioning assumptions
that l was making without realizing l was doing so, and by continuously asking
me to clarify concepts and ideas. Not only did it improve my dissertation overall,
it helped me to understand it that much more. You have always been very
supportive of me, and I thank you for that.

In addition, I would like to thank Debby Kashy for helping me in the early
stages of my dissertation proposal by showing me how to set small goals so
things wouldn’t seem so overwhelming.

Thank you also to Tom Carr for providing an invaluable perspective to the
committee. The questions you posed both in the proposal defense meeting and
in the ﬁnal defense meeting challenged me to think more deeply about the issues
involved.

Finally, I would like to thank Tim Levene for quickly agreeing to serve as
the Dean’s Representative for the ﬁnal defense meeting and for being more than

“just a reader.”

TABLE OF CONTENTS

LIST OF TABLES .................................................................................. vi
LIST OF FIGURES ............................................................................... vii
INTRODUCTION ................................................................................... 1
Subtyping ................................................................................... 4
Subgrouping .............................................................................. 1 1
Motivated Subtyping versus Subgrouping ........................................ 15
Activation and Inhibition of Group Categories, Subtypes, and
Subgroups .............................................................................. 20
EXPERIMENT 1: ACTIVATION AND INHIBITION OF GROUP
CATEGORIES, SUBTYPES, AND SUBGROUPS ..................................... 26
Overview .................................................................................. 26
Method .................................................................................... 27
Participants and Design ...................................................... 27
Materials ......................................................................... 28
Procedure ........................................................................ 30
Results .................................................................................... 32
Discussion ................................................................................ 35

EXPERIMENT 2: MOTIVATED SUBTYPING VERSUS SUBGROUPING.........40

Overview .................................................................................. 41
Method .................................................................................... 44
Prescreening .................................................................... 48

Main Experimental Session ................................................. 50

Participants and Design ............................................. 50

Procedure ............................................................... 51

Results .................................................................................... 55
Discussion ................................................................................ 61
GENERAL DISCUSSION ....................................................................... 67
FOOTNOTES ...................................................................................... 75
APPENDICES ...................................................................................... 76
REFERENCES ................................................................................... 108

LIST OF TABLES

Table 1. Activation and Inhibition Expectances for Experiment 1 .................. 100

Table 2. Mean Facilitation Scores and Standard Deviations for
Experiment 1 (Person Versus Object Judgments) and Experiment 2
(Black Versus White Judgments) .......................................................... 101

Table 3. Group Level and Subcategory Level Predictions for
Experiment 2 .................................................................................... 102

Table 4. Categorization and Recategorization Expectations for
Experiment 2 .................................................................................... 1 03

Table 5. Actual Group Level and Subcategory Level Means and Standard
Deviations for Experiment 2 ................................................................ 104

Table 6. Two-Way Interaction between Valence of the Target's
Counterstereotypic Behavior and Target Prototypicality for Group
Stereotype Change ........................................................................... 105

Table 7. Two-Way Interaction between Perceiver’s Prejudice Level and
Target’s Prototypicality for Group Stereotype Change .............................. 106

vi

LIST OF FIGURES

Figure 1. Process of Categorizing a Counterstereotypic Group Member........107

vii

INTRODUCTION

A Black lawyer violates aspects of the stereotype that many people hold
about Blacks, namely that Blacks are poor and uneducated. When people
encounter someone who violates an expectation based on that person’s group
membership (Le, a stereotype about the group to which that person belongs)
such as a Black lawyer, how do they ultimately view that person and how does
that person’s behavior impact the perceiver’s group stereotype? For instance,
would a Black lawyer be viewed as “Black," as a “lawyer,” or as some
subcategory (e.g., a “Black lawyer”)? Further, if such a person can be
categorized in multiple ways, what factors inﬂuence the way in which that person
is ultimately viewed? For example, someone who has negative prejudice toward
Blacks may be less likely than someone who has positive prejudice toward
Blacks to incorporate positive traits, such as wealthy and intelligent, into their
representation of Blacks after encountering a Black lawyer. Moreover, once a
person is viewed as a member of a particular category, are categories to which
that person could have been, but were not ultimately, placed affected? For
instance, if the target is ultimately viewed as a “Black lawyer," rather than as
“Black” or as a “lawyer,” how might the categories of Blacks or lawyers be
affected? The current research addressed these questions. In particular, the
way in which targets who disconﬁrrn a stereotype are categorized and whether
such categorizations are inﬂuenced by perceiver motivations (i.e., the

maintenance of prejudiced attitudes toward a particular group) were examined.

In addition, this research examined how the category used inﬂuences other
potentially relevant categories that could have been used to view the person.

The continuum model of impression formation processes developed by
Fiske and Neuberg (1990) suggests that impressions of individuals can be
derived from category-based knowledge (i.e., impressions based on broad
categories, such as race, gender, or age) or more individuated impressions (i.e.,
impressions that take into account the speciﬁc individual features of the target
person). Impressions are initially category-based, and they progress toward
more individuated judgments as the perceiver attends to the target person more
effortfully. Thus, upon encountering a target person, perceivers initially
categorize that target using a basic category such as race, gender, or age. Such
categorizations are usually based on salient physical features of the target and
are relatively automatic, thus not requiring much cognitive processing. In fact, it
is believed that this initial process of categorization is primarily perceptual and
automatic in nature, not requiring explicit or conscious intent on the part of the
perceiver (Bruner, 1957). For example, upon encountering a Black lawyer, a
typical perceiver would likely initially categorize him as a member of the Black
racial category because skin color is typically a very salient feature. Of course,
categorization based on race is not the only possibility. For instance,
categorization could be based on gender, especially if gender is more accessible
to the perceiver because of recent or frequent use (Higgins, 1996).

If the perceiver has no further interest in the target because the target is

not relevant to the perceiver’s goals (e.g., the perceiver will not interact with the

target) or because the perceiver lacks the cognitive resources to process
additional information related to the target (e.g., the perceiver is distracted by
another demanding task), progress along the impression formation continuum
will stop and the target will be categorized as a member of the basic category
(e.g., “Black”).

If, however, the perceiver has the interest and the cognitive resources to
process information about the target and is faced with a target who contradicts
the stereotypic features associated with the initial basic category, Fiske and
Neuberg (1990) argue that the target will be recategorized. One recategorization
option is to use a subcategory of the basic category used in the initial
categorization. For instance, if perceivers notice that the Black target is a lawyer,
having characteristics that contradict at least two features of the “Black
stereotype” (i.e., that Blacks tend to be poor and uneducated), they may
recategorize him using a more reﬁned subcategory of the overall group, such as
“Black lawyer." Of course, it may also be possible that the perceiver would
recategorize the target using a different basic category (e.g., man) in lieu of the
original basic category (e.g., Black). However, if the initial categorization is
based on a highly accessible social cue such as skin color, it is unlikely that the
perceiver would, or even could, disregard this cue in an attempt to recategorize
the target in a fashion that would not include such a salient feature (Fiske &
Neuberg, 1990). Thus, the current discussion following Fiske and Neuberg
(1990) will assume that recategorization of targets who disconﬁnn one or more

aspects of a stereotype more often involves the use of a subcategory rather than

use of a completely different basic category. Figure 1 outlines the process of
categorization for a counterstereotypic target in the current work.

Two primary types of subcategorization have been suggested in the
literature: subtyping and subgrouping. Before continuing with the current
discussion, each of these processes will be discussed in more detail.

Subtyping

Subtyping refers to the process of cognitively excluding (or “fencing off,”
as described by Allport, 1954) group members who possess attributes that
contradict the stereotype for the overall group. Members of a subtype are viewed
as “exceptions to the rule” about a basic-level social group. Although this
conceptual deﬁnition is useful, it does not specify the conditions under which a
target who disconﬁrrns the group stereotype will be subtyped. For instance, will a
target person who simply violates one aspect of a stereotype be subtyped?
Further, does the extremity of the violation inﬂuence whether a target who
disconﬁrrns the group stereotype will be subtyped? A closer examination of the
literature on subtyping can provide some indication as to the factors that
inﬂuence the process of subtyping.

Much of the research on subtyping has provided participants with
information about a number of members of a particular group. Some of these
members are described as performing behaviors that are consistent with traits
related to the overall group stereotype, whereas others are described as
performing behaviors that are inconsistent with traits related to the overall group

stereotype. Based on the information provided, participants are then asked to

make judgments about the group members and about the group as a whole.
From these measures (to be discussed), researchers infer whether the group
members who performed behaviors inconsistent with the group stereotype were
“excluded” from the group members who performed behaviors consistent with the
group stereotype. This is primarily examined by assessing whether exposure to
the targets who performed behaviors inconsistent with the group stereotype had
an impact on the perceivers overall group stereotype (i.e., weakened the group
stereotype).

In one of the ﬁrst studies to examine the process of subtyping, Weber and
Crocker (1983, Experiment 1) provided participants with information about 30
corporate lawyers. Although all participants were presented with the same
overall proportion of stereotype-consistent (17%), stereotype-inconsistent (33%),
and stereotype-irrelevant information (50%), the way in which the stereotype-
inoonsistent information was dispersed among the 30 members of the group of
corporate lawyers was manipulated. For some participants, the stereotype-
inconsistent information was dispersed across all 30 members of the group such
that each member exhibited a single stereotype-inconsistent behavior. For other
participants, the stereotype-inconsistent information was concentrated such that
one-third of the members of the group each exhibited three stereotype-
inoonsistent behaviors. Results demonstrated that when the stereotype-
inconsistent information was concentrated within fewer members of the group
rather than spread among all members of the group, participants” overall group

stereotype did not become weaker, suggesting that the disconﬁrrning members

were viewed as separate from the overall group (i.e., as a subtype) rather than
as full-ﬂedged members of the overall group.

Because in their ﬁrst study the number of disconﬁrrning targets was
confounded with the extremity of disconﬁrrning targets, Weber and Crocker
(1983) conducted a follow-up experiment to examine whether extremity alone
might inﬂuence the process of subtyping by keeping the number of disconﬁrrning
members consistent while varying the number of stereotype-inconsistent
behaviors each of those members performed. Thus, some participants were
given information such that nine of the 30 corporate lawyers each exhibited a
single stereotype-inconsistent behavior (moderate disconﬁrrnation), whereas
other participants were provided with information such that nine of the corporate
lawyers each exhibited three stereotype-inconsistent behaviors (extreme
disconﬁnnation). Results demonstrated that participants tended to subtype both
groups of disconﬁnners, suggesting that members did not have to violate the
stereotype in an extreme manner in order to be subtyped. In other words, a
single stereotype-inconsistent behavior was enough for subtyping to occur.

Johnston and Hewstone (1992) conducted a study similar to Weber and
Crocker (1983) using a different target group, physics students. In particular,
they provided participants with information about eight physics students. The
same amount of stereotype-inconsistent information was either concentrated
within two members or dispersed across six members of the group. Results
replicated those of Weber and Crocker (1983), demonstrating that when the

stereotype-inconsistent information was concentrated within a couple members

of the group, that counterstereotypic information was less likely to change the
overall group stereotype than when the same stereotype-inconsistent information
was dispersed across most members of the group.

In addition, Johnston and Hewstone (1992) found that participants judged
targets who performed stereotype-inconsistent behaviors as less typical of the
overall group when the stereotype-inconsistent behaviors were concentrated
within a couple of group members than when they were dispersed across most
members. Mediational analyses indicated that the perceived typicality of the
disconﬁrrning group members inﬂuenced stereotype change. Targets who
disconﬁrrned the stereotype but were perceived as relatively typical group
members were more likely to weaken the perceiver’s overall group stereotype
than were targets who disconﬁrrned the stereotype but were perceived as
atypical group members. Thus, atypical disconﬁrrning targets were more likely to
be subtyped than were typical disconﬁrrning targets.

In contrast to providing participants with information about several group
members, some of whom disconﬁrrned the stereotype and others of whom
conﬁrmed the stereotype, more recent subtyping research has provided
participants with a single member of a particular group who violated a single
dimension of the stereotype associated with the overall group. For example,
Kunda and Oleson (1995) provided participants with information describing a
lawyer who was introverted, thus violating the stereotype that lawyers are
extraverted. Although the research by Johnston and Hewstone (1992) showed

that perceiving the target’s disconﬁrrning behaviors as atypical is a necessary

condition for subtyping to occur, information regarding a target’s typicality may
not always be readily available to perceivers.

Thus, Kunda and Oleson (1995) examined the possibility that participants
will use neutral information as the basis to perceive a disconﬁnning target as
atypical, and hence, justify subtyping the target. Accordingly, one-third of the
participants in their study were also given information that the introverted lawyer
worked for small law ﬁrm, another one-third were told that he worked for a large
law ﬁrm, and the rest of the participants were given no information regarding the
size of the law ﬁrm for which he worked. Results showed that participants who
were given the additional piece of information (small or large ﬁrm size) were
more likely to subtype the introverted lawyer than were participants who were not
given any additional information. Kunda and Oleson (1995) reasoned that
participants who were given the additional neutral information used it to perceive
the target as atypical of the overall group, resulting in the target being subtyped.
When the neutral information was available, the target’s stereotype-inconsistent
behavior did not weaken the overall stereotype of lawyers as being extraverted (a
consequence consistent with subtyping). The target’s stereotype-inconsistent
behavior did, however, inﬂuence judgments of the subtyped group. In particular,
participants came to believe that lawyers who work for small or large (depending
on the condition) law ﬁrms were more likely to be introverted than lawyers in
general.

In another set of studies, Kunda and Oleson (1997) examined the

possibility that the extremity of the stereotype-inconsistent behavior that a target

person performed could be used as a means of perceiving the target as atypical.
Participants read about a single target described either as a moderately
unassertive feminist woman or as an extremely unassertive feminist women.
They found that participants were more likely to subtype the extremely
unassertive feminist woman than the moderately unassertive feminist woman.
Thus, greater extremity of deviance from the stereotype leads perceivers to view
the target as more atypical, increasing the likelihood of subtyping despite the
extreme behavior being especially at odds with the group stereotype.

The above ﬁndings suggest that the perception of atypicality is a
necessary condition for subtyping to occur. Perceptions of typicality can be
inﬂuenced by the way in which stereotype-inconsistent information is dispersed
among group members, by providing an additional piece of information that
perceivers can use to view a disconﬁnning target as an atypical group member,
or by providing perceivers with information that the target disconﬁrrns an aspect
of the group stereotype in an extreme fashion. In sum, it seems that providing
participants with a single target who violates a stereotype in a moderate way on
at least one dimension with additional information that can be used to perceive
the target as atypical of the overall group is sufﬁcient to induce subtyping.

Despite considerable efforts by researchers (e.g., Weber-Kollmann, 1985),
measures examining the process involved in subtyping are currently not
available. To determine whether subtyping has occurred, most researchers look
for changes in the overall group stereotype with respect to the disconﬁnning

behaviors of the target. For example, if information about a lawyer who performs

behaviors that disconﬁrrn the stereotype of lawyers being extraverted leads
perceivers to view lawyers in general as more introverted than they originally
believed, it is concluded that the behaviors of the disconﬁnning introverted lawyer
were incorporated into the overall group stereotype and that subtyping did not
occur. If, on the other hand, perceivers do not come to perceive lawyers in
general as more introverted than they originally believed, it is concluded that
perceivers subtyped the introverted lawyer. This is precisely how Kunda and
Oleson (1995, Experiment 1) inferred that subtyping did occur. In other words,
when a disconﬁrming target is subtyped, less change in the overall group
stereotype occurs than when a disconﬁn'ning target is included in the overall
group representation.

One measure that has begun to be used to assess subtyping is the
perceived typicality of disconﬁrming targets. As discussed previously, perceived
typicality is central to subtyping because perceiving disconﬁrrning targets as
atypical group members seems to be necessary for subtyping to occur.
Following an examination of the available subtyping measures, Park, Wolsko,
and Judd (2000) argued that perceived typicality is currently the best available
measure to determine whether subtyping did occur. Richards and Hewstone
(2001 ), in their review of the literature on subtyping, supported Park et al.’s
(2000) claim by suggesting that perceived typicality as a measure of subtyping
does represent the best approximation available. Therefore, although no direct
measure of subtyping exists, the current work will infer subtyping by examining

both perceived typicality of targets who disconﬁrrn a group stereotype and

10

changes in the overall group stereotype. Similar to Kunda and Olson (1995),
measures of the subtyped group (to be discussed) will also be used to assess
subtyping. It should also be noted that using measures such as perceived
typicality and changes in the overall group stereotype involve examining
continuums and thus, these measures cannot pinpoint a discrete point at which
subtyping has occurred.

Subgrouping

Subgrouping is another form of subcategorization that has only recently
been examined. Whereas subtyping involves excluding a counterstereotypic
group member from the overall group category, subgrouping involves the
inclusion of such a target in the overall group category. Thus, when a perceiver
subgroups a target, the counterstereotypic information related to that target has
an impact on the overall group representation. In particular, when a target is
subgrouped, the perceiver's overall group representation becomes more
heterogeneous, which is one reason subgrouping is viewed as a process by
which stereotypes can ultimately be changed.

Much of the early research on subgrouping focused on simply having
participants identify different subgroups within larger social groups, such as
gender (Clifton, McGrath, & Wick, 1976; Vonk & Olde-Monnikhof, 1998), race
(Judd, Park, Ryan, Brauer, & Kraus, 1995), and groups based on academic
majors (Park, Ryan, & Judd, 1992). The primary method in these studies was to
present participants with an overall group label and ask them to list as many

different “types” of the group as they could. Other research provided participants

11

with subgroup labels (e.g., Devine & Baker, 1991) or pictures (Brewer, Dull, &
Lui, 1981) and asked them to provide attributes related to each subgroup or to
sort target descriptions based on the subgroups provided.

In a more recent study, Maurer, Park, and Rothbart (1995) directly
compared the two processes of subtyping and subgrouping. They did so by
providing participants with information about 16 members of the Big Brother
group. The stimulus materials were designed such that the group could
conceptually consist of ﬁve subcategories, four of which conﬁrmed aspects of the
group stereotype, each in a unique way, and one of which disconﬁmed aspects
of the group stereotype. Participants were randomly assigned to one of three
conditions: nonsort, subtype, and subgroup. In the nonsort condition,
participants were simply told to use the information to form an impression of the
group. In the subtype condition, participants were asked to sort the members of
the group into two piles, one including those who ﬁt the group and the other
including those who did not ﬁt the group. In the subgroup condition, participants
were asked to sort the members into as many piles as they desired, “grouping
together members who seem similar in some way and different from members in
the other subgroups” (p. 814).

Maurer et al. (1995) expected that participants in the subtype condition
would perceive the subcategory of members who disconﬁrrned the stereotype of
the group as less typical than would either participants in the nonsort or subgroup
conditions. By subtyping these disconﬁrming members, participants would

perceive the overall group as more stereotypical and less heterogeneous than if

12

they had subgrouped the disconﬁrming members. Participants in the subgroup
condition, on the other hand, were expected to perceive the subcategory of
members who disconﬁnned the stereotype as more typical than either
participants in the nonsort or subgroup conditions. By including these members
as a subgroup of the overall group, the overall group would then be perceived as
less stereotypical and more heterogeneous than if the disconﬁnning members
had been subtyped.

Three main dependent measures were used to examine these
hypotheses. One was a measure of typicality in which participants were asked to
rate each target on how typical each was of the overall group of targets. A mean
and range estimation task was used to assess both stereotypic beliefs and
perceived variability of the overall group. This was measured by having
participants ﬁrst indicate the group average and then the points at which the most
extreme members of the group fall on the stereotypic traits. The group average
served as the measure of stereotypic beliefs and the range around that group
average served as the measure of perceived variability. The variability measure
is important because it has been proposed as the most important consequence
differentiating whether a target has been subtyped or subgrouped (Richards &
Hewstone, 2001 ). Perceived variability of the overall group is greater when a
counterstereotypic target has been subgrouped than when that target has been
subtyped.

Results indicated that participants in the subtype condition viewed the

disconﬁrming subcategory members as less typical of the group, which led them

13

to perceive the overall group as more stereotypical and less heterogeneous than
did participants in either the subgroup or nonsort (control) conditions.
Participants in the subgroup condition, on the other hand , viewed the
disconﬁrming subcategory members as more typical of the group and came to
view the overall group as less stereotypical and more heterogeneous than did
participants in either the subtype or nonsort conditions. Thus, subtyping and
subgrouping affected perceivers' group stereotypes, perceptions of group
typicality, and perceptions of group variability.

In a second study, Maurer et al. (1995) examined whether participants
would naturally subtype or subgroup participants without explicit instructions to
do so. The procedure was similar to that of their ﬁrst study except that a different
target group was used. In particular, participants were provided with information
about 15 members of a national gay-activist group. Again, the stimulus
information was designed such that the group could cenceptually consist of ﬁve
subcategories, four of which conﬁrmed the overall group stereotype each in a
unique way, and one of which disconﬁrrned the overall group stereotype.
Participants were asked to read through the information and form an impression
of the overall group, similar to the nonsort condition in Study 1.

Results replicated those of Study 1, demonstrating that when perceivers
viewed disconﬁrming members as atypical, they were more likely to perceive the
overall group as more stereotypic and less heterogeneous than when perceivers
viewed the disconﬂnning members as typical. This study is especially important

because it shows that natural variability in the extent to which perceivers see

14

disconﬁnning members as typical exists and that it is associated with whether
perceivers subtype or subgroup a target who violates a group stereotype even
without explicit instructions.

It is important to note that participants in Maurer et al.’s (1995) studies
were all given the same information. The way in which they approached social
categorization inﬂuenced how they ultimately subcategorized targets who
violated aspects of the group stereotype. Thus, it seems that a target who
violates a perceiver’s stereotype could ultimately be subtyped or subgrouped
based on the goals of the perceiver (i.e., whether a counterstereotypic target is
perceived as typical or atypical of the group). The current research examined
another set of goals that may inﬂuence whether a disconﬁnning target is
subtyped or subgrouped. In particular, the role of prejudice and the valence of
the counterstereotypic behavior of the target in inﬂuencing whether a
disconﬁnning target is subtyped or subgrouped were examined.

Motivated Subtyping versus Subgrouping

Because one of the main consequences of subtyping is that the overall

 

group stereotype remains unchanged, some researchers have argued that
subtyping may be inﬂuenced by the motivation to maintain group stereotypes or
group prejudices. Arguing one's motivation to maintain one’s prejudiced attitudes
is a primary function of subtyping, McConnell (1998) proposed the motivated
subtyping hypothesis. According to this hypothesis, the extent to which a
perceivers attitudes about a target group are consistent with the valence of

counterstereotypical behaviors performed by a disconﬁrming group member will

15

inﬂuence whether subtyping will occur. For example, if perceivers who hold
negative prejudices toward a group observe an atypical target performing a
positive counterstereotypical behavior, they are more likely than perceivers who
hold positive prejudice toward that group to subtype the target rather than to
perceive the target as part of the overall group in order to preserve their negative
attitudes. If, on the other hand, perceivers who hold positive prejudices toward
the group observe an atypical target performing a negative counterstereotypical
behavior, they are more likely than perceivers who hold negative prejudice
toward the group to subtype that target rather than perceive the target as part of
the overall group in order to preserve their positive attitudes. Motivation in this
case refers to the maintenance of one’s evaluation of the group. Indeed, a great
deal of research has shown that people work hard to maintain strong attitudes
and beliefs in the face of disconﬁnning information (e.g., Ditto & Lopez, 1992;
Kunda, 1990; Sinclair & Kunda, 1999). In addition, extreme attitudes have been
shown to be more resistant to change (e.g., Osgood & Tannenbaum, 1955;
Sherif & Hovland, 1961 ).

McConnell (1998) examined the motivated subtyping hypothesis by having
participants who varied in their feelings toward Blacks (i.e., those who held
relatively positive attitudes toward Blacks, those who held relatively negative
attitudes toward Blacks, and those who held relatively neutral attitudes toward
Blacks) read about six Black college male targets. Three of the six targets
conﬁrmed one of two stereotypes that were varied between participants. The

stereotype was either positive (i.e., athletic) or negative (i.e., hostile), which was

16

varied between subjects. Two of the six targets, on the other hand, disconﬁrrned
the relevant stereotype (i.e., in the positive stereotype condition the disconﬁrrners
violated the positive stereotype that Blacks are athletic, whereas in the negative
stereotype condition the disconﬁn'ners violated the negative stereotype that
Blacks are hostile). The extent to which the disconﬁnners were prototypic of
Black men in general was also varied between subjects. For some participants,
the disconﬁrrners were prototypic (i.e., they had prototypical names and came
from urban backgrounds). For other participants, the disconﬁrrners were
nonprototypic (i.e., they had nonprototypical names and came from mral
backgrounds).

When perceivers who held relatively negative attitudes toward Blacks
were presented with nonprototypic targets (i.e., rural Blacks) who disconﬁrrned a
negative stereotype (i.e., the targets’ benevolent behaviors were incongment with
the perceiver’s group attitudes), they should have been motivated to subtype that
target. Also, when participants who held relatively positive attitudes toward
Blacks were presented with nonprototypic targets who disconﬁrrned a positive
stereotype (i.e., the targets’ behaviors were incongruent with the perceiver's
group attitudes), they should have been motivated to subtype that target.
However, when participants who held relatively negative attitudes toward Blacks
were presented with nonprototypic targets that disconﬁn'ned a positive stereotype
(i.e., the targets’ behaviors were congruent with the perceiver’s group attitudes),
they should not have been motivated to subtype that target. And similarly, when

participants who held relatively positive attitudes toward Blacks were presented

17

with nonprototypic targets that disconﬁnned a negative stereotype (i.e., the
targets' benevolent behaviors were congruent with the perceiver's group
attitudes), they should also not have been motivated to subtype that target.

The pattern of ﬁndings for the negative stereotype condition was as
McConnell (1998) predicted. However, the pattern of ﬁndings for the positive
stereotype condition was not entirely consistent with expectations in that
participants with positive attitudes toward Blacks did not subtype atypical
nonathletic Blacks, as it was predicted they would. One reason offered to explain
why this was the case is that subtyping nonathletic Blacks may have been
undesirable in the sense that doing so would involve endorsing a stereotype,
something those with positive attitudes toward Blacks may wish to avoid.

The current research attempted to extend the motivated subtyping
hypothesis by examining processes that lead to both subtyping and subgrouping.
It is believed that either outcome is possible depending on the goals of the
perceiver (Maurer et al., 1995). It was expected that when perceivers did not
subtype a counterstereotypic target, they would subgroup that target instead.
More speciﬁcally, when perceivers who held relatively negative attitudes toward a
particular group were presented with atypical targets who disconﬁrrned a
negative stereotype (i.e., the targets’ behaviors were incongruent with the
perceiver's group attitudes), they should have been motivated to subtype that
target. When participants who held relatively positive attitudes toward the
particular group were presented with atypical targets who disconﬁrrned a positive

stereotype (i.e., the targets’ behaviors were incongruent with the perceiver's

18

group attitudes), they should have been motivated to subtype that target.
However, when participants who held relatively negative attitudes toward the
group were presented with typical targets who disconﬁnned a positive stereotype
(i.e., the targets’ behaviors were congruent with the perceiver’s group attitudes),
they should have been motivated to subgroup that target. And, when participants
who held relatively positive attitudes toward the group were presented with
typical targets who disconﬂrrned a negative stereotype (i.e., the targets’
behaviors were congruent with the perceiver’s group attitudes), they should have
been motivated to subgroup that target.

When a counterstereotypic target has been subgrouped, perceived
typicality of the target to the overall group and perceived variability of the
stereotype for the overall group should be greater than when the target has been
subtyped. In addition, stereotypic beliefs of the overall group should be
weakened when a counterstereotypic target has been subgrouped but not when
that target has been subtyped. In addition, perceived variability and stereotypic
beliefs for the subcategory itself should be greater when the target has been
either subtyped or subgrouped, indicating that the target was recategorized
rather than simply categorized as just a group member or as being ignored.

In sum, when a counterstereotypic target is encountered, that target can
be categorized in at least one of three ways. If the perceiver lacks interest or
cognitive resources or simply discounts or ignores the counterstereotypic
information, the target should be categorized using a basic category based on a

salient feature of the target (what I will refer to as “grouping”). If, on the other

19

hand, the perceiver has interest and resources to continue to process the
counterstereotypic information related to the target, a process of
subcategorization should be implemented. This subcategorization will be based
on how typical the counterstereotypic target is perceived to be of the overall
group. If the perceiver is motivated to maintain group prejudice and the target’s
counterstereotypic behavior is congruent with that prejudice, the target should be
perceived as relatively typical and thus be classiﬁed as a subgroup of the overall
group. If the perceiver is motivated to maintain group prejudice but the target’s
counterstereotypic behavior is incongruent with that prejudice, the target should
be perceived as relatively atypical and thus be subtyped.

Activation and Inhibition of Group Categories, Subgges, and Subgroups

In addition to examining the role of motivation in affecting whether

 

perceivers subtype or subgroup a deviant target, the current research also
investigated the cognitive representation and relations between subcategories
(i.e., subtypes and subgroups) and the overall group category by examining the
relative activation and inhibition of subcategories and of the overall group
category when a counterstereotypic person is encountered. Activation, in this
case, refers to the extent to which social categories become activated in memory
(or the relative degree to which they are brought to mind) upon encountering a
counterstereotypic target. For instance, when one encounters a Black lawyer,
most perceivers should, at least initially, categorize that target as Black. Thus,
the “Black” category perceivers have in memory would become activated. When

a category becomes active in memory, attributes associated with it also become

20

active in memory (Collins & Loftus, 1975). Thus, if perceivers categorize the
target as Black, attributes (including stereotypes and attitudes) associated with
the “Black" category will also become activated and inﬂuence perceptions and
judgment (Bruner, 1957; Higgins, Rholes, & Jones, 1977; Fazio, Jackson,
Dunton, & Williams, 1995; for a review, Smith, 1998).

I If perceivers proceed to subcategorize the Black lawyer by subtyping him,
the “Black” category may then become inhibited. Inhibition, in this case, refers to
a category that has become even less active than it was at its baseline level. For
instance, following subtyping, the “Black” category may actually be inhibited or be
at a lower level of activation potential than it was before the perceiver
encountered the target. Alternatively, if perceivers subcategorize the Black
lawyer by subgrouping him, the “Black” category should be relatively active in
memory.

Macrae, Bodenhausen, and Milne (1995) examined both activation and
inhibition of social categories for a target who could be categorized by more than
one superordinate category. Participants watched a video of an Asian woman
(i.e., a target who could be categorized as an Asian, as a woman, or as both).
Prior to viewing the video, some participants were primed with words related to
women to make the woman category more active in memory in order to
encourage categorizing the target as a woman, whereas others were primed with
words related to the Asian category in order to encourage categorizing the target
as Asian, and still others were not primed with either category (i.e., control

participants). After viewing the video, participants were asked to perform an

21

allegedly unrelated lexical decision task. This task involves having participants
judge as quickly and as accurately as possible whether letter strings presented
on a computer monitor are “words” or “nonwords” by pressing one of two keys on
a computer keyboard. The time between the presentation of the letter string and
the participant's response was recorded by the computer, and this response
latency constituted the main dependent measure. The stimuli presented
consisted of words that were stereotypic of women but not of Asians, words that
were stereotypic of Asians but not of women, and words that were not stereotypic
of women or of Asians, in addition to nonword letter strings.

If stereotypic concepts related to a category were already activated in
memory by the priming manipulation, participants should be faster to respond to
concepts associated with that category on the lexical decision task than
participants who were not primed with either category. The relative speed of a
response following priming (i.e., facilitation) reﬂects the extent to which the
concept and the primed category are associated in memory. Conversely, if
stereotypic concepts related to the other category are actively inhibited,
participants should be slower to respond to those concepts than participants who
were not primed with either category. This would demonstrate inhibition of an
alternative category. This is exactly what Macrae et al. (1995) observed. When
participants were primed with the “woman” category, they were signiﬁcantly
faster than control participants to respond to words that were stereotypic of
women, demonstrating activation of the “woman” category. In addition, these

same participants were signiﬁcantly slower than control participants at

22

responding to words that were stereotypic of Asians, demonstrating inhibition of
the “Asian” category. Conversely, when participants were primed with the
“Asian” category, they were signiﬁcantly faster than control participants to
respond to words that were stereotypic of Asians, demonstrating activation of the
“Asian” category. In addition, these same participants were signiﬁcantly slower
than control participants at responding to words that were stereotypic of women,
demonstrating inhibition of the “woman” category. Thus, when a perceiver can
categorize a target using more than one social category, the category that “wins"
becomes activated in memory and the category that “loses” is inhibited.

Rudman and Borgida (1995) examined activation and inhibition with
respect to subordinate categories of women. Male participants ﬁrst watched
either a video containing a number of sexist commercials (to prime them with the
category of sexual object women, a subcategory of women) or a control video
containing nonsexist commercials targeted for men. Both videos contained
commercials featuring beer, cologne, cars, and clothing. Those commercials
included in the sexist video condition were rated as relatively high in sexism and
eroticism, whereas those commercials included in the control video condition
were rated as relatively low in sexism and eroticism in a pretest study. After
viewing the video, participants performed an allegedly unrelated lexical decision
task similar to that used by Macrae et al. (1995). The stimuli presented consisted
of both sexist and nonsexist words, in addition to nonword letter strings. Results
demonstrated that when men were primed with the subcategory sexual object

women they were faster to respond to the sexist words than were control

23

participants. The authors inferred that this demonstrated activation of the
subcategory of sexual objects women. These participants were also slower to
respond to the nonsexist words than were controls, which the authors inferred as
demonstrating inhibition of a nonsexual objects women subcategory.

It is not clear, however, whether the results of Rudman and Borgida’s
(1995) study actually demonstrate activation and inhibition of “subcategories” of
women. The traits associated with broad subcategories such as sexual object
women and nonsexual object women are likely to be negatively correlated at the
onset. Thus, activating traits associated with one of these broad subcategories
may inhibit opposing traits simply because such traits are negatively correlated,
and not because activation of one of the subcategories inhibits activation of the
other subcategory.

Therefore, the current research examined activation of subcategories and
of the overall group category when a target is subtyped or subgrouped by
focusing on the activation level of categories rather than the activation level of
the traits associated with the categories. As previously mentioned, a
counterstereotypic target should be categorized with a basic category during
initial categorization. However, when a disconﬁrming target is subtyped, the
counterstereotypic information associated with that target is less likely to become
incorporated into the overall group representation. Thus, it was expected that
when a disconﬁrming target was subtyped, the overall group category would be
inhibited. Such inhibition would prevent the counterstereotypic information from

being incorporated into the overall group representation, thereby providing a

24

means by which overall group stereotypic beliefs and prejudice can be
maintained. On the other hand, when a disconﬁrming target is subgrouped, the
counterstereotypic information does become incorporated into the overall group
representation, ultimately reducing the overall group stereotype. Therefore, it
was expected that when a disconﬁrming target is subgrouped, the overall group
category would remain activated and the disconﬁn'ning target’s behavior would

be incorporated with the overall group knowledge.

25

EXPERIMENT 1: ACTIVATION AND INHIBITION OF GROUP CATEGORIES,
SUBTYPES, AND SUBGROUPS

The ﬁrst experiment examined the relative activation and inhibition of
subcategories and of the overall group category when a target is subtyped or
subgrouped. Unlike previous experiments that focused on the attributes
associated with the categories or subcategories involved (e.g., Macrae et al.,
1995; Rudman & Borgida, 1995) this study focused on the activation of the
superordinate categories and of the subcategories in the same experiment.
W

Participants were presented with images of social group members who ﬁt
different subcategories of Black men based on a pretest study. For the ﬁrst part
of the study, another image (the target) of either a person (i.e., a very typical
Black man or a very typical White man) or of an object (i.e., a typical sofa or a
typical chair) was presented after each initial image (the prime) presentation.
Participants responded as quickly as possible using one of two keys to indicate
whether the target image was a person or an object. For the second part of the
study, the target image that followed each prime was either of a typical Black
man or of a typical White man (i.e., no objects were presented). Participants
responded as quickly as possible using one of two keys to indicate whether the
person depicted in target image was Black or White. The predictions were the
same for both parts of the study. In particular, it was expected that when a group
member presented as a prime was subgrouped, the overall group category would

be activated, facilitating responses to members who ﬁt the overall group

26

category. For instance, when a typical Black male target image was preceded by
a prime image of a moderately typical subcategory member (i.e., a group
member who is likely to be subgrouped), responses to the typical Black male
target should be relatively faster than when that target is preceded by a neutral
image prime (i.e., a series of ﬁve ampersands). When a group member
presented as a prime was subtyped, on the other hand, it was expected that the
overall group category would be inhibited, slowing responses to members who ﬁt
the overall group category. Thus, when a typical Black male target image was
preceded by a prime image of an atypical subcategory member (i.e., a group
member who is likely to be subtyped), responses to the typical Black male target
should be relatively slower than when that target was preceded by a neutral
prime image.
METHOD

Participants and Design

The design of the study was a 5 (Prime: Black subgroup member, Black
subtype member, typical Black group member, typical White group member, or
control) X 3 (Target: typical Black group member, typical White group member, or
object) within-subjects design.

A power analysis was conducted to determine the number of participants
necessary to detect a 2-way interaction in the current study. At an alpha level of
.05, with two within-subject factors, a large effect size of .40, and a desired power

level of .80, it was determined that a total of 26 participants would be adequate

27

(Cohen, 1988). Ultimately, 47 White students participated in the primary
experiment for credit in their introductory psychology course.
Materials

Stimulus Development. In order to identify typical and atypical
subcategories college students perceive within the overall category of Blacks, a
pretest was conducted in which 49 participants were given instructions that
included an example of typical (e.g., birds that ﬂy, birds that chirp) and atypical
(e.g., birds that don’t ﬂy, birds that talk) birds. They were then provided with
various superordinate social groups (including Blacks) and asked to list typical
and atypical types for each group.

Images representing members from each of the subcategories that at
least 33% of participants listed (Gilbert & Hixon, 1991) combined with what prior
research (e.g., Devine & Baker, 1991) identiﬁed as different subcategories of
Blacks were found via the intemet. These subcategories included athletes,
businessmen, criminals, doctors, educated, poor, musicians, and religious. In
addition, images of Black men who ﬁt the overall category of Blacks (i.e., very
typical) were sought. These images were to be used as primes and as targets
for the ﬁrstexperiment. The goal was to obtain four images that represented
members from six different subcategories (three relatively typical subcategories,
or subgroups, and three relatively atypical subcategories, or subtypes), in
addition to 12 images that represent members who ﬁt the overall category of

Blacks (i.e., those who were unlikely to be subcategorized).

28

The target images for the ﬁrst part of the study involved two main types:
persons and objects. The person target images consisted of Black men who
were rated as very typical of the overall Black category and of White men who
were rated as very typical of the overall White category. The object targets
consisted of typical sofas, chairs, and tables. The target images for the second
part of the study consisted only of images of very typical Black men and of very
typical White men.

Once the images were collected, a second pretest was conducted. The
images were randomly presented on a computer to 46 participants who rated
how typical the target (e.g., a Black man) depicted in each was of the overall
category (e.g., Blacks) by typing a number based on a scale ranging from 1 (not
at all typical) to 9 (extremely typical). Of the 260 images presented, 95 depicted
Black men, 111 depicted White men, 24 depicted women, and 30 depicted
furniture. Images rated low in typicality depicting a Black man who belonged to
one of the subcategories identiﬁed by participants in the ﬁrst pretest were
classiﬁed of members of subcategories that were more likely to be subtyped than
subgrouped, which should be more likely to inhibit the overall group category
upon presentation. Images rated as moderately typical depicting a person who
belonged to one of the subcategories identiﬁed by participants in the ﬁrst pretest
were classiﬁed as members of subcategories that were more likely to be
subgrouped than subtyped, which should be more likely to activate the overall
group category upon presentation. And, images rated as extremely typical

depicting a person who did not appear to belong to a particular subcategory

29

identiﬁed by participants in the ﬁrst pretest were classiﬁed as members who ﬁt
the overall group category rather than a subcategory of the group. Thus, these
images should activate the overall group category only.

Based on these typicality ratings, four images were identiﬁed for each of
the following Black subtypes: businessmen (M = 5.91 ), doctors (M = 5.36), and
educated (M = 6.10). In addition, four images were identiﬁed for each of the
following Black subgroups: athletes (M = 6.81 ), musicians (M = 6.81), and
religious (M = 6.31 ). And, 12 images were identiﬁed for the overall Black
category (M = 7.21 ). For consistency, four images depicting White men in each
of the same subcategories were also selected for the main experimental session.
Thus, images for the following groups were selected: businessmen (M = 7.55),
doctors (M = 7.17), educated (M = 7.52), athletes (M = 6.53), musicians
(M = 5.52), religious (M = 5.25), and typical (M = 7.64). In addition to these
experimental primes, a neutral prime image consisting of a series of ﬁve
ampersands was used in the task (Fazio et al., 1995). Responses to targets
following this neutral prime were used as baseline comparisons to responses to
targets of the same type that followed the experimental primes. The images
selected and their mean typicality ratings are presented in Appendix A.
Procedure

Upon arrival to the lab, participants were told that the study was examining
how quickly and accurately people categorize people and objects. At private

workstations, they completed the categorization task.

30

Categorization Task. While seated in front of a computer monitor,
participants were presented with an orienting stimulus (i.e., a “+”) for 2500 ms
followed by a randomly selected prime (i.e., Black subtype, Black subgroup,
Black group, White group, or neutral image) presented for 315 ms (similar to the
procedure used by Fazio, et al., 1995). Participants were told that although they
did not have to respond to the ﬁrst image, they should attend to it because there
would allegedly be a memory test for the images they had seen later in the
experimental session. Immediately following the presentation of this prime
image, a randomly selected target image (either a person or an object) was
presented. The participants’ task was to decide as quickly and as accurately as
possible whether the target image presented was a person or an object by
pressing one of two buttons on the keyboard labeled “Person” and “Object.” This
task consisted of an initial 36 practice trials which included images not used in
later trials, 144 experimental trials in which either a Black typical image or a
White typical image was the target, and 144 ﬁller trials in which an image of a
typical piece of furniture was the target. For the experimental trials, each prime
was presented twice, once followed by a Black typical target image and once
followed by a White typical target image. The trials and images presented were
randomized differently for each participant. In addition, following the 36 practice
trials and at each 72 trial-interval, participants were given a one-minute rest
period during which a message was displayed on the screen to remind them to
pay attention to the ﬁrst image but to base their judgments on the second image.

Response latencies between the presentation of the target image and the key

31

press were recorded by the computer and were used to determine whether the
target category was activated or inhibited following activation of the primed
concept (the ﬁrst image).

Immediately following this task, participants performed a similar task in
which they categorized the persons depicted in the target images as either
“Black” or “White.” This task consisted of 144 experimental trials with a one-
minute rest period after the ﬁrst block of 72 trials to remind participants of the
instructions. No practice or ﬁller trials were used in this part of the study. For the
experimental trials, each prime was presented twice, once followed by a Black
typical target image and once followed by a White typical target image. The trials
and images presented were randomized differently for each participant. The
categorization task was implemented using DirectRT Precision Timing Software
(Jarvis, 2002).

RESULTS

Responses to the target images were ﬁrst checked for correctness. For
the ﬁrst part of the study (i.e., person versus object judgments), it was important
that participants pressed the key corresponding to “person” (as opposed to the
key corresponding to “object”) when a person was depicted in the target image.
For the second part of the study (i.e., Black versus White judgments), it was
important that participants pressed the key corresponding to “Black" when a
Black target image was presented and "White” when a White target image was

presented. The overall error rates for the ﬁrst and second parts of the study were

32

low (2% and 3%, respectively). Only correct responses were included in the
analyses (see Fazio, 1990).

A logic—transformation was performed on each of the correct response
latencies (Fazio, 1990), and facilitation scores were then calculated by
subtracting the average response latency for each prime-target combination from
the average response latency for the same target when preceded by the control
prime. Thus, greater positive facilitation scores indicated that a particular target
was more strongly associated with the primed concept relative to a neutral
concept (i.e., the control prime). In particular, mean facilitation scores were
calculated for each type of prime image (i.e., Black typical member, Black
subgroup member, Black subtype member, and White typical member) that was
followed by Black typical target images. Also, mean facilitation scores were
calculated for each type of prime image (i.e., Black typical member, Black
subgroup member, Black subtype member, and White typical member) that was
followed by White typical target images. Positive facilitation scores indicated
responses faster than baseline (suggesting activation of the target given its
prime), whereas negative facilitation scores indicated responses slower than
baseline (suggesting inhibition of the target given its prime).

A 4 (Prime type: Black subgroup member, Black subtype member, typical
Black group member, or typical White group member) X 2 (Target Type: typical
Black group member or typical White group member) repeated measures
ANOVA was conducted on the facilitation scores. Table 1 shows the expected

pattern of results. In particular, it was expected that responses to Black targets

33

would be fastest (suggesting strong activation of the Black category) when
preceded by Black typical primes as compared to the other three prime types.
Because it was assumed that Black subgroups are related to the overall Black
category, responses to Black targets should be relatively faster when preceded
by Black subgroup primes than when preceded by Black subtype primes or White
typical primes, demonstrating that subgrouping results in activation of both the
subgroup and overall group categories. Black subtypes, on the other hand, have
attributes that contradict the overall group stereotype. Thus, it was expected that
responses to Black targets would be moderately slower (suggesting inhibition of
the Black category) when preceded by Black subtypes than when preceded by
Black typical primes or by Black subgroup primes. Because beliefs about Blacks
and Whites are often contradictory, it was expected that responses to Black
targets will be slowest when preceded by White typical primes (reﬂecting
inhibition of the Black category) than when preceded by any of the other three
prime types.

For White targets, it was expected that responses would be slower
(suggesting inhibition) when preceded by Black typical primes or Black subgroup
primes than when preceded by Black subtype primes or White typical primes.
However, because subtypes have attributes that oppose the overall Black
category and may share attributes that may be more consistent with the overall
White category, responses to White targets may be relatively faster when
preceded by a Black subtype prime than when preceded by Black typical primes

or by Black subgroup primes. Finally, responses to White targets should be

34

fastest when preceded by White typical primes than when preceded by any of the
other three types of primes.

However, the actual mean facilitation scores for each judgment task,
which are presented in Table 2, did not reveal the predicted pattern. Contrary to
predictions, no signiﬁcant effects were found for the ﬁrst part of the study
(i.e., person versus object judgments, top panel of Table 2). However, there was
a marginally signiﬁcant interaction between prime type and target type,

F(3,138) = 3.76, p < .06. Planned contrasts showed a signiﬁcant difference in
response latencies to White versus Black targets when preceded by a Black
typical prime, F(1,46) = 4.14, p < .05. Surprisingly, responses to White targets
were faster than responses to Black targets following Black typical group primes.

Also contrary to predictions, no signiﬁcant effects were found for the
second part of the study (i.e., white versus black judgments; bottom panel of
Table 2).

Interestingly, the overall mean latency scores for both parts of the study
(M = .07 for the ﬁrst part and M = .02 for the second part) were signiﬁcantly
greater than baseline, t(46) = 6.70, p < .001 for the ﬁrst part and t(46) = 4.28,

p < .001 for the second part. Participants responded faster to Black and White
typical targets overall when preceded by any of the primes depicting a person
than when preceded by the neutral prime image.
DISCUSSION
The predicted pattern of results would have demonstrated that when an

atypical target is subtyped, the overall group category is subsequently inhibited.

35

Such an outcome would support a mechanism by which atypical group members’
counterstereotypic behaviors would have little to no impact in changing group
stereotypes. Because atypical targets are presumably categorized using basic
social categories (e.g., Blacks), at least initially (Fiske & Neuberg, 1990), the
inhibition of the overall social group category could explain why encountering
such targets does not affect overall group stereotypes. On the other hand, when
an atypical target is recategorized as a subgroup member, the overall group
category should remain activated. Various researchers have postulated relations
such as these (e.g., Allport, 1954; Kunda & Oleson, 1995, 1997; Maurer et al.,
1995; McConnell, 1998), but they have not studied category activation directly.
This experiment, on the other hand, was the ﬁrst to examine category activation
and inhibition for groups, subgroups, and subtypes. Unfortunately, it provided no
evidence consistent with the predicted pattern of results.

There was, however, an unexpected ﬁnding. It was found that participants
tended to respond faster to White typical targets that followed Black typical group
primes than they responded to Black typical targets that followed Black typical
group primes. Examination of the means for both judgment tasks of the study
suggest that in general, participants tended to respond slightly faster to target
images belonging to the White superordinate category than to target images
belonging to the Black superordinate category. Interestingly, participants were
especially fast to respond to target images belonging to the White superordinate
category than they were to respond to target images belonging to the Black

superordinate category when the preceding prime was an image belonging to the

36

Black superordinate category. One possibility for this finding is that White primes
in general may not evoke a social category for White participants. Black typical
group primes, on the other hand, would be especially likely to evoke racial
categories, which in turn would activate the Black category as well as activating
the White category. This heightened activation might then lead participants to
respond especially fast to White typical targets that followed Black typical group
primes.

It is important to note that this experiment differed from previous studies
that examined the activation and inhibition of social categories in a number of
ways. First, other studies (i.e., Macrae et al., 1995; Rudman & Borgida, 1995)
primed participants and measured the activation of social categories or
subcategories in two separate tasks. Participants were ﬁrst primed with one of
two superordinate (e.g., Asian or woman) or subordinate (e.g., women as sexual
objects or women as non-sexual objects) categories and then performed a task
to measure activation and inhibition of both of those superordinate or subordinate
social categories. The current study combined the priming of multiple
superordinate and subordinate categories and the measurement of activation
and inhibition of the broader superordinate categories in one task. More
speciﬁcally, participants were primed with various images depicting individuals
belonging to one of two superordinate categories (i.e., Black or White) or one of
six subordinate categories (e.g., Black businessmen, White musicians), while

measuring the activation and inhibition of the superordinate categories.

37

Second, whereas previous studies relied on stereotypic words associated
with a category or subcategory to measure activation and inhibition of the
broader social categories or subcategories, the current study used images
depicting actual social group members to measure activation and inhibition of the
broader social categories. As pointed out by Fazio et al. (1995), using images
should activate group membership more directly than the use of stereotypic
words, providing a more ecologically valid assessment of what happens when a
person actually encounters a member of a particular social group. However,
there has been some controversy in the literature regarding whether social
categories are automatically activated upon encounter with a group member. In
particular, some have argued that encountering social group members should
automatically activate group knowledge (e.g., Devine, 1989) especially for social
groups whose accessibility in memory is strong (e.g., Fazio et al., 1995). Others
have argued (e.g., Gilbert & Hixon, 1991; Spencer, Fein, Wolfe, Fong & Dunn,
1998) that merely encountering a social group member does not necessarily
result in the activation of the social group in memory. Instead, they have argued
that perceivers must adopt evaluation goals in order for social category activation
to have its full effect. The paradigm used by Fazio et al. (1995) relies on judging
the connotation of target words that follow image primes. Perhaps such
evaluative goals are necessary for full social category activation to occur.
Indeed, similar arguments have been made about when spontaneous trait
inferences will be rendered (e.g., Bassili & Smith, 1986) and for when priming

has behavioral consequences (e.g., Bargh & Chartrand, 1999). Thus, perhaps

38

the current judgment task did not lead to activation of the intended social
categories because the target judgment task was not evaluative in nature.

A ﬁnal possibility is that perhaps subtyping-induced inhibition was not
observed because, in fact, targets who have appeared to be subtyped were
actually ignored rather than “refenced” (Allport, 1954). Indeed, the primary
measure of subtyping used in most studies is the amount of change in stereotype
strength after encountering counterstereotypical information associated with
atypical group members. When there is little or no change in stereotype
strength, it is inferred that subtyping occurred. Perhaps these atypical group
members are actually ignored and not subtyped. Ignoring these targets would
produce little or no change in stereotype strength (the “result” reported to support
subtyping in numerous studies), and the overall social category would not need
to be inhibited to produce this outcome.

Although the lack of support for inhibition and activation following the
presentation of nonprototypic group members was disappointing, it is possible
that less-than-typical counterstereotypical group members may be treated
differently than typical counterstereotypical group members. As acknowledged
above, it is possible that “subtyping consistent” outcomes might be observed
even if nonprototypical group members who violate group stereotypes are
ignored. Thus, a second study was conducted to explore the implications of
counterstereotypical group members” behaviors for judgments of the overall
group, and for knowledge about subcategories related to the overall group, in

particular.

39

EXPERIMENT 2: MOTIVATED SUBTYPING VERSUS SUBGROUPING
Whereas Experiment 1 examined the consequences of subtyping and

subgrouping targets who belong to preexisting subcategories for the average
college student, Experiment 2 examined the recategorization of targets who
violate an overall group stereotype in a novel manner. Research (e.g., Johnston
& Hewstone, 1992; Weber & Crocker, 1983) has established that perceived
typicality of a counterstereotypic target to the overall group inﬂuences whether
that target will be subtyped. In particular, when a counterstereotypic target is
perceived as relatively typical of the overall group, the target seems to be
included in the overall group and the stereotype for the overall group is
weakened. However, when a counterstereotypic target is perceived as relatively
atypical of the overall group, the target will be subtyped and thus the stereotype
will remain unchanged.

Thus, the current research argued that differences in whether a target
would be subtyped or subgrouped based on perceiver motivations should only
occur when the target is perceived to possess an attribute that is nonprototypic of
the overall group. The way in which that nonprototypic information is used will be
based on the motivations of the perceiver. In particular, it was expected that
when perceivers were motivated to maintain group prejudice and the target’s
counterstereotypic behavior was congruent with that prejudice, the target would
be perceived as relatively typical and thus, the target be recategorized as a
subgroup member of the overall group even though information suggests the

target is atypical. On the other hand, when perceivers were motivated to

40

maintain group prejudice but the target's counterstereotypic behavior was
incongruent with their prejudice, the nonprototypic information would be used to
perceive the target as atypical and thus, the target would be subtyped. In the
former case, group stereotypes should be weakened by the incorporation of
stereotype inconsistent information with overall group knowledge. In the latter
case, however, group stereotypes should remain unchanged because the
target’s counterstereotypic behavior would be associated with a subtype category
instead of with the overall group.
W

Participants who varied in their level of prejudice (i.e., negative prejudice,
neutral prejudice, and positive prejudice) toward the target group were presented
with information about a target member who performed a counterstereotypic
behavior. For some participants, this counterstereotypic behavior was positive
(thus violating a negative group stereotype), whereas for other participants the
behavior was negative (thus violating a positive group stereotype). In addition,
the extent to which the target possessed a prototypic attribute was also varied.
In particular, for some participants the target was presented as being relatively
prototypic of the overall group, whereas for other participants the target was
presented as being relatively nonprototypic of the overall group.

In order to assess whether participants categorized the target as a group
member, or recategorized the target by subtyping or subgrouping, two classes of
dependent measures were collected from participants: those involving

perceptions of the overall group and those involving perceptions of the

41

subcategory. For perceptions of the overall group, perceived typicality of the
target to the overall group was assessed, and both the extremity and variability of
the stereotypic trait for the overall group were measured before and after the
presentation of the target information. For perceptions of the subcategory, the
extremity and variability of the stereotypic trait were assessed both before and
after the presentation of the target information. It was expected that the
counterstereotypic target would be grouped, subgrouped, or subtyped, and the
expected pattern of results is outlined in Table 3.

In particular, it was expected that when the target was grouped, perceived
typicality of the target to the overall group would be greater than when the target
was subtyped or subgrouped. In addition, the stereotype would be weaker and
perceived variability of the stereotypic attribute would be greater for the overall
group than when the target was subtyped or subgrouped. For example, if the
target was an unintelligent gay male police ofﬁcer (who violates the stereotype
that gay men are intelligent), participants should report that gay men in general
are less intelligent when they have grouped the target than when they have
subcategorized the target. In addition, they should perceive greater variability in
intelligence for gay men in general following grouping than following
subcategorization. There should be no change in stereotypic beliefs or perceived
variability when the target is grouped.

When the target was subgrouped, perceived typicality of the target to the
overall group should be greater than when the target was subtyped but less than

when the target was grouped. The stereotype for the overall group should also

42

be weakened moderately but to a lesser degree than when the target was
grouped. In addition, perceived variability of the stereotypic attribute with regard
to the overall group should be greater than when the target was subtyped but not
as great as when the target was grouped. With regard to subcategory
perceptions, the stereotype for the overall group should only be weakly
associated with the subgroup. The counterstereotypic attribute, on the other
hand, should be more strongly associated with the subgroup than with the overall
group. For example, the subcategory of gay male police ofﬁcers would be
perceived as extremely unintelligent. Thus, it was expected that the subgroup
would be perceived as even more counterstereotypic following subgrouping than
following grouping. In addition, perceived variability of the stereotype for the
subcategory should become less extreme following subgrouping than following
grouping, suggesting that the subgroup was perceived as more homogeneous
than before. For example, participants should report less variability in
intelligence for gay male police ofﬁcers when the target was subgrouped than
when the target is grouped.

When the target was subtyped, perceived typicality of the target to the
overall group should be the lowest compared to when the target is grouped or
subgrouped. Further, the stereotype and its perceived variability for the overall
group should remain unchanged with regard to the overall group following
subtyping. However, the stereotype should be weaker with regard to the
subcategory when the target was subtyped than when the target was grouped,

whereas the counterstereotypic attribute should be stronger with regard to the

43

subtype when the target was subtyped than when the target was grouped. Thus,
it was expected that the subtype would be perceived as more counterstereotypic
than initially perceived. In addition, perceived variability of the stereotype for the
subcategory should become less extreme following subtyping than following
grouping, suggesting that the subtype was perceived as even more
homogeneous than it was before.
METHOD

Stimulus Development

A number of pretests were conducted in order to determine the social
group to be used, the stimuli for the positive and negative counterstereotypic
behaviors that the target person from that group would perform, and the
prototypical or nonprototypical information to be associated with that target
person for the current study. For the ﬁrst pretest, 68 participants were presented
with a number of attributes, some of which have been shown to be stereotypic of
gay men and some of which are irrelevant to gay men. They ﬁrst estimated the
proportion (from 0 to 100) of people in the general population who they believe
possessed each attribute. They then rated the desirability of each of those
attributes on a scale ranging from 1 (not at all desirable) to 9 (extremely
desirable). Finally, they estimated the proportion (from 0 to 100) of gay men who
possessed each of the same attributes.

A diagnosticity ratio was calculated based on a procedure developed by
McCauley and Stitt (1978; McCauley, Stitt, & Segal, 1980), who suggested that

when the probability of an attribute given a particular social category is greater

than the probability of the same attribute given people in general, a stereotype
exists in the mind of the perceiver. The diagnosticity ratio quantiﬁes the
extremity of a stereotype belief, and is expressed as p(attribute | target group)
divided by p(attribute | people in general). To the extent that the ratio deviates
from 1.0, a stereotypic belief for the group exists.

In addition, to assess participants’ prejudice toward gay men, participants
indicated using a feeling thermometer how favorably they evaluate the typical
gay man on a scale ranging from 100 degrees (extremely favorable) to 0 degrees
(extremely unfavorable) with a midpoint of 50 degrees (neither favorable nor
unfavorable).

The most common positive and negative attributes that both positive and
negative prejudice participants held as a stereotype of gay men to the same
degree (based on the diagnosticity ratios) were selected to use as the stimuli for
the current study. “Intelligent” was selected as the positive attribute and
“promiscuous” was selected as the negative attribute. For the pretest
participants, the mean stereotype diagnosticity ratio for intelligent was 1.16
(SD = .62) and the mean stereotype diagnosticity ratio for promiscuous was
1.00 (SD = .89). The mean desirability for intelligent was 8.65 (SD = .73) and the
mean desirability for promiscuous was 2.06 (SD = 1.64). To be certain that these
stereotypes were held to a similar extent by participants who varied in their level
of prejudice toward gay men, the diagnosticity ratios for intelligence and
promiscuous were analyzed in one-way analyses of variance (ANOVAs) with

prejudice level (positive, neutral, or negative) as the independent variable.

45

Results showed no difference in stereotype strength across participants who
varied in their level of prejudice toward gay men (for intelligence, F(2,67) = .47,
ns, and for promiscuity, F(2,67) = 1.22, ns). Thus, participants differed in their
level of prejudice, but not in the strength with which they held the stereotypic
belief that gay men are intelligent and promiscuous.

An additional pretest was conducted on the attributes selected (i.e.,
intelligence and promiscuous) to conﬁrm that they were indeed considered as
stereotypic attributes for the group of interest. Thus, 28 participants estimated
the proportion of gay men in general (from 0 to 100) who possessed each
attribute. In addition, because prototypicality has been identiﬁed in previous
research as a factor inﬂuencing subtyping, another attribute (i.e., target’s
occupation) unrelated to the stereotypes selected for the group (i.e., intelligent
and promiscuous) was identiﬁed and used to manipulate the prototypicality of the
target person in the main experimental session. In order to ensure that the
positive and negative stereotype attributes selected as stimuli were unrelated to
this prototypicality factor, participants were also asked to estimate the proportion
of gay men (from 0 to 100) with different occupations who possessed each
attribute. These occupations included police ofﬁcer and interior designer. Paired
t-tests indicated participants did not perceive a difference in either intelligence or
promiscuity between police ofﬁcers and interior designers, t(27) = -1.39, ns and
t(27) = .84, ns, respectively. Thus, these two occupations were used to
manipulate the prototypicality of the target person in the main experimental

session. Identifying the target’s occupation as an interior designer should lead

46

participants to perceive him as prototypical of gay men in general, whereas
identifying the target’s occupation as a police ofﬁcer should lead participants to
perceive him as nonprototypical of gay men in general.

From the results of these pretest studies, four vignettes purportedly based
on interviews with gay men attending a “Gay Lifestyles of the New Millennium”
conference were developed. Speciﬁcally, participants in a third pretest were
presented with information about a homosexual male target who described his
behavior as being either positively counterstereotypic (e.g., non-promiscuous) or
negatively counterstereotypic (e.g., unintelligent). To ensure that ratings were
based on the information about the counterstereotypic attribute of interest and
not inﬂuenced by other factors, they were not given any information about his
occupation (i.e., the prototypicality factor). Thirty-four participants read one of
the two vignettes and then rated the target on how intelligent and promiscuous
they perceived him to be on separate scales ranging from 1 (not at all intelligent
or not at all promiscuous) to 7 (extremely intelligent or extremely promiscuous).
One-way ANOVAs with type of vignette (unintelligent target or non-promiscuous
target) as the independent variable were conducted on the ratings for intelligence
and promiscuous, separately. Results showed that the target in the
“unintelligent” vignette was perceived to be relatively low in intelligence
(M = 3.88, SD = 1.05), as intended, and somewhat high in promiscuity
(M = 5.41, SD = 1.54), as would be expected based on the stereotype of gay

men as promiscuous when no information about promiscuity is given,

47

F(1,33) = 33.29, p < .001. The target in the “non-promiscuous” vignette was
perceived as relatively low in promiscuity (M = 2.59, SD = 1.70), as intended, and
somewhat high in intelligence (M = 6.06, SD = 1.14), as would be expected
based on the stereotype of gay men as intelligent when no information about
intelligence is given, F(1,33) = 25.74, p < .001.

Prescreening

In a mass prescreening, 477 students in introductory psychology courses
at Michigan State University responded to two questionnaires prior to being
asked to participate in the main experimental session. One of these
questionnaires assessed their stereotypes about homosexual men and the other
assessed their attitudes toward homosexual men. The two questionnaires are
described below.

Stereotygs about Homosexual Men. Similar to the pretest, participants
ﬁrst estimated the proportion (from 0 to 100) of people in the general population
who they believed possessed a series of attributes, including the positive
stereotypic trait (i.e., intelligent) and the negative stereotypic trait (i.e., sexually
promiscuous) identiﬁed in the pretest described above. Next, they rated the
desirability of each of those attributes on a scale ranging from 1 (not at all
desirable) to 9 (extremely desirable). Then they estimated the proportion (from 0
to 100) of homosexual men who possess each of the same attributes.

Prejudice Toward Homosexual Men. In order to assess participants’
prejudice toward homosexual men, participants indicated how favorably they

view homosexual men in general on a scale ranging from 100 degrees

48

(extremely favorable) to 0 degrees (extremely unfavorable) with a midpoint of 50
degrees (neither favorable nor unfavorable).

Recruitment Criteria. To be selected to participate in the primary
experimental session, participants needed to have diagnosticity ratios for at least
one of the critical traits greater than 1.0, indicating that they believe intelligent or
promiscuous is a stereotype associated with gay men. In addition, participants
selected must have rated the desirability of each trait as intended (i.e., intelligent
as desirable and promiscuous as undesirable). Of the 477 prescreening
participants, 378 met these criteria and were invited via e-mail to participate in
the main experimental session. Among these participants, the mean stereotype
diagnosticity ratio for intelligent was 1.61 (SD = 1.37) and the mean stereotype
diagnosticity ratio for promiscuous was 1.89 (SD = 1.70). The mean desirability
for intelligent was 8.13 (SD = .96) and the mean desirability for promiscuous was
2.33 (SD = 1.53). These participants were then classiﬁed based on their
prejudice toward homosexual men. In particular, those who reported scores
above 50 on the feeling thermometer were classiﬁed as having positive prejudice
toward homosexual men, those who reported scores below 50 were classiﬁed as
having negative prejudice toward homosexual men, and those who reported
scores of 50 as having neutral prejudice toward homosexual men.

To be certain that these stereotypes were held to a similar extent by
participants who varied in their level of prejudice toward gay men, the
diagnosticity ratios for intelligence and promiscuous were analyzed in one-way

analyses of variance (ANOVAs) with prejudice level (positive, neutral, or

49

negative) as the independent variable. Results showed no difference in strength
of stereotype across participants who varied in their level of prejudice toward gay
men (for intelligence, F(2,377) = 1.31, ns, and for promiscuity, F(2,377) = 1.21,
ns). Thus, participants differed in their level of prejudice, but not in the strength
with which they held the stereotypic belief about the target group, ensuring that
strength of stereotype did not differ as a function of participant prejudice.

Main Experimental Session
Participants and Design.

The design of the study was a 3 (group prejudice: positive, negative, or
neutral) X 2 (valence of the stereotype violated: positive or negative) X 2
(prototypicality: prototypical or nonprototypical) between-subjects factorial
design. Participants were randomly assigned to one of the four following
conditions: prototypical target who violated a positive stereotype, prototypical
target who violated a negative stereotype, nonprototypical target who violated a
positive stereotype, or nonprototypical target who violated a negative stereotype.

A power analysis was conducted to determine the number of participants
necessary to detect a 3-way interaction in the current study. At an alpha level of
.05, with three between-subjects factors, a moderate effect size of .25, and a
desired power level of .80, it was determined that a total of 156 participants
would be adequate (Cohen, 1988). Thus, every attempt was made to obtain this
desired number of participants.

Six to ten weeks after the prescreening, participants meeting the

recruitment criteria outlined above were invited to participate in the main

50

experimental session via e-mail. If interested, they were to select a session
posted on the Michigan State University Department of Psychology’s Subject
Pool Website. Immediately after selecting a session, they were prompted to
complete a brief pre-experimental questionnaire on the website. Ultimately, 151
students participated in the main experiment for credit in their introductory
psychology courses, in addition to the opportunity to win one of two cash prizes
(one for $50.00 and the other for $75.00) in a drawing conducted on the last day
of the semester. These participants had a mean stereotype diagnosticity ratio for
intelligent of 1.66 (SD = 1.99) and a mean stereotype diagnosticity ratio for
promiscuity of 1.67 (SD = 1.37). The mean desirability for intelligent was 8.09
(SD = 1.09) and the mean desirability for promiscuous was 2.61
(SD = 1.85). Of these participants, 62 were classiﬁed as having relatively
positive affect toward gay men (M = 77.65, SD = 12.52), 41 were classiﬁed as
having relatively negative affect toward gay men (M = 29.49, SD = 13.29) and, 48
were classiﬁed as having relatively neutral affect toward gay men (M = 50.00,
SD = 0.00).
Procedure

Pre-Experimental Questionnaire. Stereotypicality and perceived variability
were measured both before and after the presentation of the target information in
order to examine changes in overall group stereotype strength and perceived
variability as a function of being exposed to the counterstereotypical target.
Thus, immediately after scheduling a session for the main experiment on the

Subject Pool website, participants were presented with a number of attributes

51

and behaviors, including the stereotypic positive and negative attributes selected
for the main study (i.e., intelligent and promiscuous). Similar to the procedure
used by other researchers (e.g., Jones, Wood, & Quattrone, 1981; Park & Judd,
1990; Park, et al., 1992), they indicated where, on average, homosexual men fall
with regard to each attribute or behavior on a scale ranging from 0 (e.g.,
extremely unintelligent) to 100 (e.g., extremely intelligent). They then indicated
the highest and lowest points at which homosexual men might fall on the
continuum with regard to each attribute. From this, a range score was calculated
by taking the absolute value of the difference between the lowest and highest
ratings for each participant (Maurer et al., 1995).

Main Experimental Session. Upon arrival to the lab, participants were told
that the study was exploring how people form impressions of others based on
varying degrees of information. In order to enhance privacy, participants
completed the study individually at private computer workstations. They were
told that the computer would randomly select and present an excerpt of an
interview from a database developed from interviews of people attending various
conferences around the country. It was emphasized that they should read the
information carefully because they would be required to answer questions about
the target later in the study. Speciﬁcally, they were presented with information
about a homosexual male target who described his behavior as being either
positively counterstereotypic (e.g., non-promiscuous) or negatively
counterstereotypic (e.g., unintelligent). In addition, participants were given

information indicating that the target was either prototypical of homosexual men

52

in general (e.g., is an interior designer) or not nonprototypical of homosexual
men in general (e.g., is a police ofﬁcer). The program used to present the
information about the target and obtain the measures was MediaLab2001 (Jarvis,
2002). The excerpts presented to participants are presented in Appendix B.

Following the presentation of this information, participants responded to
two classes of dependent measures: those applying to the overall group (i.e.,
homosexual men in general) and those applying to each of the subcategory
groups (i.e., homosexual male interior designers and homosexual male police
ofﬁcers). They then responded to some questions intended to ensure that they
had read the information presented in the excerpt thoroughly.

Dependent Measures

Group Perceptions

Perceived Typically. Participants indicated how typical the target was of
homosexual men in general on a scale ranging from 1 (not at all typical) to 9
(extremely typical), with a midpoint of 5 (moderately typical).

Stereogpe Strength and Perceived Variability (Time 2). Participants were

again presented with a number of attributes and behaviors, including the

 

stereotypic positive and negative behaviors associated with the target presented
to participants when they scheduled a session for the main experiment via the
Michigan State University Psychology Subject Pool website. They indicated
where the average homosexual man falls with regard to each attribute or
behavior on a scale ranging from 0 (e.g., not at all promiscuous) to 100 (e.g.,

extremely promiscuous), and both the highest and lowest points at which

53

homosexual men in general might fall. The ﬁrst score served as the measure of
stereotype strength. The range for each attribute was calculated by taking the
absolute value of the difference between the lowest and highest ratings for each
participant. The ranges for the critical traits provided measures of perceived
variability.
Subcategoty Perceptions

Stereogpe Strength and Perceived Lahability (Ttm_e_2_). Participants were
again presented with a number of attributes and behaviors, including the
stereotypic positive and negative behaviors associated with the target presented
to participants eartier. They indicated where the average homosexual male
interior designer and the average homosexual male police ofﬁcer (separately)
would fall with regard to each attribute or behavior on a scale ranging from 0
(e.g., not at all intelligent) to 100 (e.g., extremely intelligent), and both the highest
and lowest points at which homosexual male interior designers and police
ofﬁcers in general might fall. The ﬁrst score was the measure of stereotype
strength for the subcategory. The range for each attribute was calculated by
taking the absolute value of the difference between the lowest and highest
ratings for each participant. The ranges for the critical traits provided measures
of perceived variability for the subcategory.
Manipulation Checks

Participants were told that the study was also examining how accurately
people remember information presented in digital format. They were asked to

recall the target’s ﬁrst name, age, sexual orientation, occupation, and the name

54

of the conference in which the interview took place. These measures were used
to ensure that participants could recall the important information related to the
manipulations.
RESULTS

Manipulation Checks

Because the predictions depended on whether the target’s behavior was
perceived as counterstereotypical and whether the target himself was perceived
as prototypical, it was important that participants read the information provided in
the excerpts thoroughly. Thus, responses to the manipulation checks were
assessed for correctness. Participants who could not recall the target’s
occupation and could not recall either the target’s sexual orientation or the name
of the conference in which the interview took place were excluded from the
analyses (23 could not). The remaining 128 participants had a mean stereotype
diagnosticity ratio for intelligent of 1.64 (SD = 2.01) and a mean stereotype
diagnosticity ratio for promiscuous of 1.65 (SD = 1.46). The mean desirability for
intelligent was 8.10 (SD = 1.00) and the mean desirability for promiscuous was
2.55 (SD = 1.77). Of these participants, 52 were classiﬁed as having relatively
positive affect toward gay men (M = 78.54, SD = 13.24), 33 were classiﬁed as
having relatively negative affect toward gay men (M = 29.45, SD = 13.11) and,
43 were classiﬁed as having relatively neutral affect toward gay men (M = 50.00,

so = 0.00).

55

Tests of Main Hypotheses

Difference scores were calculated for stereotype change and perceived
variability with regard to both the overall group and the subcategory measures.
In particular, the mean of the stereotype measure at Time 1 was subtracted from
the mean of the stereotype measure at Time 2. Negative scores indicated that
the stereotype was weaker following the presentation of the target information.
For perceived variability, the absolute value of the difference between the lowest
and the highest point on the continuum was calculated for participants separately
at Time 1 and at Time 2. Perceived variability absolute differences at Time 1
were then subtracted from the perceived variability absolute differences at Time
2 to produce a change in perceived variability score. Greater change in
variability scores indicated that the stereotype became more variable (i.e., less
homogeneous) after the presentation of the target information.

To test the hypotheses, 3 (prejudice level: positive, negative, or neutral) X
2 (valence of the stereotype violated: positive or negative) X 2 (target
prototypicality: prototypical or nonprototypical) ANOVAs were conducted on the
group level perceived typicality scores, group level stereotype change scores,
group change in variability scores, subcategory level stereotype change scores,
and the subcategory change in variability scores, separately. Table 4 shows the
expected outcomes in terms of categorization and recategorization of the target
based on prototypicality, perceiver prejudice level, and valence of the stereotype
violated. A main effect for prototypicality was expected for group-level

judgments, such that when the target was prototypic, he would be perceived as a

56

more typical member of the overall group, stereotypes for the overall group would
become weaker, and perceived variability of the overall group would be greater
(no changes in the subcategory stereotype or perceived variability were
expected). Contrary to the predictions, no signiﬁcant main effects were found for
perceived typicality, F(1,127) = 2.00, ns, group stereotype change,

F(1,127) = .21, ns, or perceived variability of the overall group, F(1,127) = .00,

ns. Consistent with predictions, no main effect was found for stereotype change,
F(1,127) = .63, ns, or change in stereotype variability, F(1,127) = .90, ns, for the
subcategory ratings.

A 3-way interaction among prejudice level, valence of the stereotype, and
prototypicality was also predicted. In particular, it was expected that the
combination of prejudice level and valence of the target’s counterstereotypic
behavior would inﬂuence recategorization only when the target was
nonprototypic. Thus, when the target was nonprototypic, subtyping would be
possible, but its occurrence would depend on the perceiver’s prejudice level and
the congruence of the valence of the target’s counterstereotypic behavior with
the perceiver's prejudice (i.e., when the perceiver had positive prejudice toward
the target group and the target’s counterstereotypic behavior was negative, and
when the perceiver had negative prejudice toward the target group and the
target’s counterstereotypic behavior was positive), the target would be subtyped.
This would be demonstrated by ratings of the target as relatively atypical of the
overall group, no change in the stereotype for the overall group (i.e., the

stereotype was maintained or possibly strengthened), and no change in

57

perceived variability with regard to the overall group. In addition, perceptions of
the subcategory would become more stereotypic and perceived variability would
be lower (i.e., the subcategory would be perceived as more homogeneous) than
it was previously.

However, when the target is nonprototypic and the perceivers prejudice
level is consistent with the valence of the target's counterstereotypic behavior
(i.e., when the perceiver has positive prejudice toward the target group and the
target’s counterstereotypic behavior is positive and when the perceiver has
negative prejudice toward the target group and the target’s counterstereotypic
behavior is negative), the target will be subgrouped. This would be
demonstrated by ratings of the target as moderately typical of the overall group,
weakening of the stereotype for the overall group, and greater perceived
variability with regard to the overall group. In addition, perceptions of the
subcategory would become more stereotypic and perceived variability would be
more homogeneous.

Finally, when the target is prototypic, one would expect grouping to occur.
In other words, the target’s counterstereotypic behaviors would be incorporated
into group knowledge, leading to perceptions that the target was relatively typical
of the overall group, and producing a reduction in the strength of the stereotype
for the overall group, and greater perceived variability with regard to the overall
group. However, unlike the prediction for subgrouping, judgments of qualities

ascribed to the subgroups would remain unchanged.

58

As Table 5 reports, contrary to predictions, no 3-way interactions for
perceived typicality of the target to the overall group, F(2,126) = .61, ns, change
in stereotype of the overall group, F(2,126) = .52, ns, or change in perceived
variability of the stereotype for the overall group, F(2,126) = .45, ns, were found.
In addition, no 3-way interactions were found for change in the stereotype related

to the subcategories, F(2,126) = .51, ns, or for change in perceived variability of

the stereotype related to the subcategories, F(2,126) = 1.82, ns.1

There were, however, a number of unanticipated signiﬁcant ﬁndings. In
particular, there was a main effect of the valence of the target's
counterstereotypic behavior for perceived typicality, F(1,126) = 104.38, p < .001.
When the target’s counterstereotypic behavior was positive in nature (i.e.,
nonpromiscuous), participants perceived him as relatively more typical of gay
men in general (M = 6.13, SD = 1.21) than when the target’s counterstereotypic
behavior was negative (i.e., unintelligent) in nature (M = 3.64, SD = 1.42).

There was also a main effect for the valence of the target's
counterstereotypic behavior for group stereotype change, F(1,126) = 11.41,

p < .01, which was qualiﬁed by a 2-way interaction with target prototypicality,
F(1,126) = 4.83, p < .05. As shown in Table 6, the relation between the target’s
prototypicality and group stereotype change varied as a function of the valence of
the target’s counterstereotypic behavior. In particular, the overall ﬁnding of
stronger stereotypes following negative counterstereotypic behaviors and weaker
stereotypes following positive counterstereotypic behaviors was stronger when

the target was prototypic (i.e., an interior designer) than when he was

59

nonprototypic (i.e., a police ofﬁcer). Although this ﬁnding was as expected for the
positive counterstereotypic condition (i.e., weaker promiscuity stereotypes for gay
men in general resulting from encountering a prototypic nonpromiscuous gay
man), the reversal for the negative counterstereotypic behavior condition was not
consistent with expectations.

In addition, there was a signiﬁcant 2-way interaction between perceiver
prejudice level and target prototypicality for group stereotype change,

F(2,126) = 4.38, p < .05. Table 7 shows that perceivers with positive prejudice
toward gay men showed evidence of the expected typicality effect. In particular,
the group stereotypes for those with positive prejudice became weaker after
encountering prototypical counterstereotypic targets, but showed relatively little
change in stereotypes after encountering nonprototypical counterstereotypic
targets, t(50) = -12.02, p < .05. Those with negative and neutral prejudice toward
gay men, however, showed the opposite pattern (i.e., little or no change in
stereotypes after encountering prototypical counterstereotypic targets and
weaker stereotypes after encountering nonprototypical counterstereotypic
targets).

For subcategory stereotype change, there was a main effect for the
valence of the target’s counterstereotypic behavior, F(1,126) = 4.81, p < .05. The
stereotype of the subcategory became weaker when the target’s
counterstereotypic behavior was negative (M = -3.34, SD = 18.49) than when the
target’s counterstereotypic behavior was positive (M = 4.52, SD = 18.28). There

was also a main effect for perceiver’s prejudice level on subcategory stereotype

60

change, F(2,126) = 5.37, p < .01. The stereotype of the subcategory became
stronger for perceivers with negative attitudes toward gay men in general
(M = 8.94, SD = 19.84) than for perceivers with either positive attitudes
(M = -1.48, SD = 18.15) or neutral attitudes (M = -2.40, SD = 17.08) toward gay
men in general.
DISCUSSION

If results were as predicted, they would have suggested that the way in
which an atypical counterstereotypic group member target was recategorized
was inﬂuenced by the congruency between the perceiver’s attitudes and the
valence of the target’s counterstereotypic behavior. When the perceiver's
attitudes and the valence of the target’s behavior were incongruent with one
another, the target would be excluded from the perceiver’s overall group
representation (i.e., subtyped) if any additional information suggested that the
target could be considered as atypical. However, when the perceiver's attitudes
and the valence of the target’s counterstereotypic behavior were congruent, the
target would be included within a subset of the perceiver’s overall group
representation (i.e., subgrouped), even in cases where additional information
might suggest the target was potentially atypical. However, no evidence
consistent with the predicted results was found.

Yet, there were some unexpected ﬁndings with regard to both group
stereotype change and subcategory stereotype change. For group stereotype
change, the valence of the target’s counterstereotypic behavior inﬂuenced the

perception of typicality of that target to the overall group. In particular, positive

61

counterstereotypic targets (i.e., those who were nonpromiscuous) were
perceived as more typical of gay men in general than were negative
counterstereotypic targets (i.e., those who were unintelligent). Prototypicality of
the target was the only factor predicted to inﬂuence perceived typicality. The
valence of the target’s counterstereotypic behavior was expected to work in
combination with the target’s prejudice to affect the way in which the target was
categorized. Thus, the ﬁnding that the valence of the target’s counterstereotypic
behavior (and not the prototypicality of the target itself) had an effect on
participants’ perception of typicality is surprising. One explanation for this ﬁnding
may be based on the speciﬁc stereotypes used. For instance, promiscuity could
be perceived as a more malleable trait than intelligence. Thus, it may have been
easier for participants to perceive a nonpromiscuous gay man, as opposed to an
unintelligent gay man, as more typical of gay men in general because
promiscuity could be viewed as conduct that one chooses, whereas intelligence
may be viewed as a trait over which one does not have a choice. In addition,
closer examination of the promiscuous excerpt used in the current study reveals
that it may have unwittingly suggested to participants that the target actually was
promiscuous at one time, but changed his behavior when he met his current
pannen

Another ﬁnding with regard to group stereotype change was based on the
valence of the target’s counterstereotypic behavior in combination with the
target’s prototypicality. In particular, weaker promiscuity stereotypes for gay men

in general were found after encountering a prototypic nonpromiscuous gay man,

62

which is what would be expected if such a target was grouped. In addition,
relatively little change in intelligent stereotypes for gay men in general as a result
of encountering a nonprototypic unintelligent gay man was as expected if such a
target was subtyped. However, weaker promiscuity stereotypes for gay men in
general as a result of encountering a nonprototypic nonpromiscuous gay man
and stronger intelligent stereotypes for gay men in general as a result of
encountering a prototypic unintelligent gay man were not consistent with what
would be expected if these targets were subtyped and grouped, respectively. In
fact, it was predicted that the nonprototypic nonpromiscuous gay man would be
subtyped and thus, he should have no effect on the overall group stereotype. It
was also predicted that the prototypic unintelligent gay man would be grouped
and thus, the stereotype of gay men as intelligent would have become weaker
rather than stronger. The latter of these two ﬁndings seems to be consistent with
a boomerang effect, which suggests that stereotypes may become stronger after
encountering very counterstereotypic targets (Kunda & Oleson, 1997). As
previously discussed, participants perceived the unintelligent gay target as
relatively atypical of gay men in general, which may have led them to contrast
him from the group as a whole, leading them to view the overall group as even
more stereotypic than they did prior to encountering such a counterstereotypic
target. The reason for weaker stereotypes following a nonprototypic
nonpromiscuous gay man is less clear. One possibility is that because

participants tended to perceive the nonpromiscuous gay man as more typical of

63

gay men in general, they incorporated him into their overall group representation,
thus weakening their stereotype of gay men as promiscuous.

There was also some evidence to support the basic prototypicality effect
found in previous subtyping research. In particular, the interaction between
perceiver prejudice level and target prototypicality for group stereotype change
suggests that after encountering a prototypical counterstereotypic target, those
with positive prejudice toward gay men revealed weaker group stereotypes,
consistent with what would be expected if the target was categorized as a group
member. And similarly, after encountering a nonprototypical counterstereotypic
target, those with positive prejudice toward gay men revealed relatively little
change in their group stereotypes, consistent with what would be expected if the
target was subtyped. Those with negative prejudice toward gay men and those
with neutral prejudice toward gay men, however, showed the opposite pattern. In
particular, after encountering a prototypical counterstereotypic target, those with
negative or neutral prejudice revealed relatively little change in their group
stereotypes, consistent with what would be expected if the target was subtyped.
And, after encountering a nonprototypical counterstereotypic target, those with
negative or neutral prejudice reported relatively weaker group stereotypes, a
ﬁnding that would be expected if the target was grouped. Thus, only the positive
prejudice perceivers demonstrated the basic prototypicality effect. In many of the
studies exploring subtyping (e.g., Brewer et al., 1981; Johnston & Hewstone,
1992), the targets were members of groups for which participants should have

relatively positive evaluations (e.g., librarians, physics majors, grandmothers).

64

Thus, perhaps the current results for neutral and negative prejudice participants
are not entirely at odds with the literature because the majority of studies in the
literature have used relatively likable target groups for which most participants
would have relatively positive attitudes. Perhaps positive attitudes toward targets
induce relatively more positive affective states for the perceiver, which lead them
to process target information with less effort (Schwarz & Clore, 1996) and thus
make subtyping more likely (especially if subtyping really results from ignoring
atypical targets rather than actively inhibiting social categories as suggested by
Study 1).

There were also two ﬁndings with regard to subcategory stereotype
change. First, it was shown that those with negative attitudes toward gay men in
general revealed stronger subcategory stereotypes than did those with either
positive or neutral attitudes toward gay men. It was expected that subcategory
stereotypes would change only as a result of the target being subcategorized and
that this change would be in the form of weaker subcategory stereotypes rather
than stronger subcategory stereotypes. Stronger subcategory stereotypes
suggest that negative prejudice participants may have incorporated the
unintelligent or nonpromiscuous target into their overall group representation,
and then contrasted that group with the subcategories of gay male interior
designers and gay male police ofﬁcers, somewhat similar to the boomerang
effect described in relation to the ﬁnding that stronger group stereotypes were

revealed after encountering a prototypic unintelligent gay man.

65

A second ﬁnding with subcategory stereotype change was based on the
valence of the target’s counterstereotypic behavior. Subcategory stereotypes
became weaker after encountering a target who behaved in a negatively
counterstereotypic (i.e., unintelligent) manner than after encountering a target
who behaved in a positively counterstereotypic (i.e., nonpromiscuous) manner.
Weaker subcategory stereotypes are consistent with a subcategorization process
in which the subcategory is perceived as even more counterstereotypic than
before. Thus, it seems that unintelligent targets were more likely to be
subcategorized than were nonpromiscuous targets. This is consistent with the
previous ﬁnding that unintelligent targets were perceived as less typical of gay
men in general than were nonpromiscuous targets. Together these ﬁndings
suggest that unintelligent targets were perceived as relatively atypical of gay men
in general and thus, were subtyped, leading the subcategory to be viewed as

extremely counterstereotypic.

66

GENERAL DISCUSSION

The two studies were designed to address some important issues with
regard to the antecedents and consequences of categorizing target group
members who violate a stereotype related to the overall group. In particular, the
ﬁrst study examined the consequences of encountering typical and atypical
group members with regard to category activation and inhibition. If the predicted
results would have been borne out, a process by which counterstereotypic
information is blocked from incorporation into the stereotype for the overall group
when a target is subtyped would have been revealed. Heretofore, the process
I underlying subtyping has not been well understood, and unfortunately, Study 1
did not shed much light on it either.

The second study examined possible antecedents that might lead one to
subtype or subgroup an atypical target and to observe its implications on group-
relevant judgments. If the results would have been aS‘predicted, they would
have supported previous ﬁndings with regard to the inﬂuence of perceived
typicality on the process of subtyping as well as the motivated subtyping
hypothesis, which suggests that the social categorization can serve to reafﬁrrn
one’s strongly held prejudices. In addition, the ﬁndings would have contributed to
previous work (the motivated subtyping hypothesis in particular) by highlighting
another subcategorization process perceivers may use when encountering an
atypical group member, namely subgrouping. Unfortunately, Study 2 neither
replicated previous ﬁndings nor extended our understanding of subcategorization

processes very much.

67

Still, there were some interesting ﬁndings with regard to both studies that
deserve further discussion. Although the ﬁrst study was not successful in
demonstrating activation and inhibition of categories and subcategories following
exposure to group targets, the design of the study was novel. As discussed, the
study used images to both prime and measure activation and inhibition of group
categories and subcategories. Previous studies examining activation and
inhibition of group categories and subcategories used stereotypic words related
to the group categories and subcategories instead. Although the study’s
predictions were not supported, it does highlight a potential alternative method
for priming and measuring activation and inhibition of group categories and
subcategories. As noted previously, it may be the case that such methods must
be incorporated with additional goals to fully activate the social category (e.g.,
Gilbert 8 Hixon, 1991; Bargh & Chartrand, 1999). Clearly, future research should
explore what sort of processing objectives are necessary to maximize the
likelihood of observing social category activation and inhibition.

The ﬁndings of Study 1 seem to suggest that the White typical group
category tended to be more active overall than did the Black typical group
category. This was especially apparent when a Black typical group member
image preceded it as the prime. As mentioned previously, it is possible that
White primes in general may not evoke a social category for White participants in
the same way Black typical group primes would. In particular, Black typical
group primes should be especially likely to evoke race, which would in turn

activate the Black category as well as the White category. This heightened

68

activation would then lead participants to respond especially fast to White typical
targets that followed Black typical group primes.

Study 2 attempted to identify conditions under which perceivers would
group, subtype, or subgroup counterstereotypic targets, and it sought to
demonstrate how such categorizations would affect judgments of the overall
group and of subcategories related to the overall group in particular. Group
stereotype change effects provided some evidence that perceivers did group or
subtype targets under certain conditions. However, because there were no
effects for perceived group variability, there is no evidence to suggest that
perceivers subgrouped atypical targets. Although the prototypicality factor did
not lead participants in general to categorize typical targets as group members
and to categorize atypical targets as subcategory members, it did inﬂuence those
with positive attitudes toward gay men to do so. In particular, positive prejudice
participants demonstrated the basic typicality effect found in previous research,
which has demonstrated that when a target was typical of the overall group, that
target is more likely to be incorporated into the overall group representation,
whereas when a target was atypical of the overall group, the target is more likely
to be excluded (i.e., subtyped) from the overall group representation (Johnson &
Hewstone, 1992). Positive prejudice participants demonstrated weaker group
stereotypes after encountering a prototypical counterstereotypic target
suggesting they grouped these targets. After encountering a nonprototypical
counterstereotypic target, however, positive prejudice participants demonstrated

very little change in their group stereotypes, which is consistent with subtyping

69

these targets. Thus, the positive prejudice participants showed the anticipated
prototypicality effect.

Interestingly, negative and neutral prejudice participants showed the
opposite effect, suggesting that they grouped the nonprototypical
counterstereotypic target and subtyped the prototypical counterstereotypic target.
As noted earlier, most studies in the subtyping literature have presented target
groups that are relatively positive in nature (e.g., librarians, grandmothers).
Perhaps “typicality effects” observed in these studies (e.g., Brewer et al., 1981 ;
Johnston & Hewstone, 1992) occur for perceivers processing information about
“likable” target groups. It is possible that processing information about positive
targets reduces the cognitive expenditures that perceivers devote to targets,
which could result in their ignoring atypical counterstereotypical targets rather
than actively subtyping them (Schwarz & Clore, 1996). This explanation is
consistent with the current ﬁndings for positive prejudice participants, and
suggests that subtyping might occur because atypical targets are ignored rather
than “refenced” through active inhibition (which is consistent with the ﬁndings of
Study 1).

It is also interesting to note that some of the ﬁndings also support a
boomerang effect. Although it was predicted that group stereotypes or
subcategory stereotypes would become weaker following either grouping or
subcategorization, there were some instances in which group stereotypes and
subcategory stereotypes actually became stronger. This ﬁnding is consistent

with a boomerang effect, which occurs when a counterstereotypic target leads to

70

contrast effects with the overall group or subcategory to which the target was not
included (Kunda & Oleson, 1997). There are three instances in which
stereotypes became stronger. When a prototypical negative counterstereotypic
(i.e., unintelligent) target was encountered, group stereotypes became stronger,
suggesting that the target was subtyped and then that subcategory was used as
a contrast for the overall group. This would lead participants to perceive the
overall group as even more stereotypic than before. Again, because participants
perceived the unintelligent gay target as relatively atypical of gay men in general,
it may have been relatively easy for them to subtype him and thus use the
subcategory within which he was placed (i.e., gay male interior designers) as a
contrast to the overall group, leading to the boomerang effect. Interestingly,
however, this subtyping did not lead to weaker subcategory stereotypes, as
would be expected.

The other two instances of stronger stereotypes following encounters with
target group members were found for subcategory stereotypes. In particular,
after encountering a positive counterstereotypic target, subcategory stereotypes
became stronger, suggesting that such targets were grouped and then the
overall group was used as a contrast for the subcategory. Also, negative
prejudice participants demonstrated stronger subcategory stereotypes overall,
suggesting they tended to group targets, regardless of their prototypicality or the
valence of their counterstereotypic behavior, and then used the group as a
contrast for the subcategory, leading them to perceive the subcategories as more

stereotypic than before.

71

These boomerang effects, although not the focus of the primary
predictions in the current study, are still important. In many cases, boomerang
effects that result from contrast effects suggest to perceivers that the stereotypes
should be even more strongly endorsed upon encountering targets who violate
them. That Is, when a counterstereotypical target is used as a contrast standard,
it can become “the exception that proves the rule,” allowing participants to
maintain strong (or stronger) stereotypes in the face of disconﬁrming evidence.
Although it seems evident that encountering counterstereotypic targets should
weaken stereotypes, the current results (and those reported by others, such as
' Kunda & Oleson, 1997) indicate that this is not always case the case. Although
these processes are not fully understood, it is clear that additional research is
needed to understand when and why counterstereotypical cases lead to stronger
group stereotypes.

One limitation of Experiment 2 is that a number of participants had to be
dropped from the analyses because they either did not initially read the
information presented thoroughly or did not retain the information they did
process, which ultimately reduced the ability to detect a moderate effect for the
predicted 3-way interaction. It was emphasized to participants in the instructions
that they would be asked questions about the excerpt later in the study, so they
should have been induced to thoroughly read and retain the relatively small
amount of information given in the excerpt. However, it is clear that not all
participants took the task as seriously as hoped. It should be noted, however,

that power analyses indicated that the current sample size in Study 2 was

72

reasonable for observing signiﬁcant effects, thus it seems that low power was not
a critical problem in Study 2. Moreover, the sample size in Study 1 was more
than sufﬁcient to observe the Intended results. Thus, sample size is probably not
a major culprit in the lack of success observed in the current work.

As mentioned earlier, it is possible that evaluative goals are required to
observe full social category activation and inhibition effects (eg. Gilbert & Hixon,
1991 ). It is interesting to note that no systematic exploration of how processing
goals affect priming has been conducted in the social cognitive literature.
Although some have speculated about the importance of instruction set in
assimilation and priming effects (e.g., Bargh & Chartrand, 1999; Bassili & Smith,
1986), no programmatic work has tackled these issues directly. Clearly,
additional work is needed to explore these issues more fully in order to
understand when social category activation and inhibition should be observed.

Stereotypes of social groups have been shown to be quite resilient to
change, even when information that runs counter to those stereotypes is
encountered. Thus, future research should continue to focus on factors that
might inﬂuence people to change their negative group stereotypes, and
ultimately, negative group prejudice. In addition to considering how maintaining
important self-relevant beliefs affects social categorization (e.g., Kunda &
Oleson, 1997; Sinclair & Kunda, 1999), research should consider how affect
inﬂuences social categorization processes and stereotyping as well (e.g.,
Bodenhausen, Kramer, & Suesser, 1994; Schwarz & Clore, 1996). Affect may

not only be important as a motivator, but affect can guide the mechanisms by

73

which social information is processed as well. Although the current work may not
represent a major advance in our understanding of these important issues, it is
hoped that the issues identiﬁed in the current work will be addressed by future

researchers to better understand how stereotypes are maintained and modiﬁed.

74

FOOTNOTES

1In addition to examining the hypotheses with an ANOVA, a 3 (prejudice level:

positive, neutral, or negative) X 2 (valence of the stereotype violated: positive or
negative) X 2 (target prototypicality: prototypical or nonprototypical) multivariate
analysis of variance (MANOVA) was conducted on the group level judgments
(i.e., typicality of target, group stereotype change, and group stereotype
variability change) and on the subcategory level judgments (i.e., subcategory
stereotype change and subcategory variability change), separately. Neither of
these 3-way interactions obtained. Multiple regression analyses with prejudice
level as a continuous variable, and valence of stereotype violated and target
prototypicality as dummy coded variables and all 2-way interactions and the
3-way interaction were also conducted on each of the dependent measures.
None of the predicted 3-way interaction terms was signiﬁcant: beta=.25,
t(126)=1.40, ns for target typicality, beta=.01, t(126)=.06, ns for group stereotype
change, beta=-.40, t(126)=-1.63, ns for group stereotype variability change,
beta=.12, t(126)=.51, ns for subcategory stereotype change, and beta=-.21,
t(126)=-.89, ns. Thus, these additional analyses did not provide any evidence in

support of the primary hypotheses.

75

APPENDICES

76

APPENDIX A

Images used in Experiment 1

Black Athlete Images

 

M=6.65 (SD=1.57) M=6.85 (SD=1.66)

 

M=7.25 (SD=1.62) M=6.70 (SD=1.59)

77

Black Musician Images

   

M=6.30 (SD=2.03)

 

M=7.25 (SD=1.33) M=6.90 (SD=1.74)

78

Black Preacher/Religious Images

 

M=6.30 (SD=2.03)

   

M=6.45 (SD=1.57) M=6.15 (SD=2.03)

79

Black Businessmen Ima e

 

 

M=5.65 (SD=1.90)

S

 

 

M=5.60 (SD=2.06)

M=5.35 (SD=1.42)

80

Black Doctor/Medical Images

   

M=4.65 (SD=1.84)

 

M=4.80 (SD=2.17) M=5.05 (SD=1.90)

81

Black Educated Images

 

M=4.90 (SD=1.94) M=5.45 (SD=2.01)

82

Black Typical Images

 

M=6.90 (SD=1.71)

   

M=7.50 (SD=1.40) M=7.55 (SD=1.23)

83

 

M=7.75 (SD=1.37)

M=7.45 (SD=1.39)

 

84

 

M=7.05 (SD=1.28)

 

M=7.45 (SD=1.43) M=6.80 (SD=2.09)

85

White Athlete Images

 

M=6.52 (SD=1.67) M=6.50 (SD=1.56)

86

White Musician Ima es

 

5.63 (SD=1.81)

M

5.35 (SD=1.73)

M

 

 

5.30 (SD=1.59)

M

5.80 (SD=1.77)

M

87

White Preacher/Religious Images

     

M=4.13 (SD=2.09)

 

M=5.63 (SD=2.11) M=5.91 (SD=2.13)

88

White Businessmen Images

   

M=7.50 (SD=1.33)

 

   

M=7.41 (SD=1.63) M=7.65 (SD=1.32)

89

White Doctor/Medical Images

 

> "E
{egg-{1:1 as: 1;.

M=6.67 (SD=1.81) M=7.33 (SD=1.35)

   

M=7.39 (SD=1.37) M=7.28 (SD=1.49)

90

White Educated Ima es

 

   

M=7.65 (SD=1.32) M=6.89 (SD=1.39)

   

M=7.80 (SD=1.22) M=7.72 (SD=1.19)

91

White Typical Images

 

=8.09 (SD=1.07)

   

M=7.67 (SD=1.28) M=7.67 (SD=1.37)

92

 

M=7.65 (SD=1.39) M=7.65 (SD=1.29)

 

M=7.57 (SD=1.17) M=7.57 (SD=1.36)

93

M=7.54 (SD=1.35)

 

M=7.13 (SD=1.26)

 

 

. J’s-L

M=7.59 (SD=1.38)

94

Neutral Prime Image

   

95

APPENDIX B

Stimuli Used in Experiment 2

Negative Stereotype (Promiscuous) Violation by Prototypical Tagrget Condition

 

Exceﬂof Interview from “Gav Lifestyles in the New Millennipm”§onference

Date: January 28, 2002

Name: Phil IV—

Gender: Male

Age: 32

Sexual orientation: Homosexual
Occupation: Interior Designer

Interviewer: Can you tell me a bit about the intimate relationships in your life?
Subject: My partner and l have been together for over two years now. I
pretty much consider myself married at this point.

Interviewer: When you’re not attending conferences such as this one, what do
you do with your spare time?

Subject: I enjoy watching movies... in fact, my partner and l are both movie
buffs, so we spend a lot of time going to the theatre and having
“Blockbuster” movie nights.

Interviewer: Are you enjoying the conference?

Subject: Yes, I am. It’s been interesting to see what other people are up to.
We’ve run into some old friends... it’s been nice catching up with
them.

Interviewer: Have you had a chance to go to any clubs while you’ve been in
town for the conference?

Subject: That’s not really something I have an interest in anymore. I’ve
really just been hanging out with my partner, enjoying one another’s
company.

 

 

96

Negative Stereotype (Promiscuous) Violation by Non-prototypical Target

Condition

 

ExceMof Interview from “Gay Lifestyles in the New Millennium” Conference

Date: January 28, 2002

Name: Frank G—
Gender: Male

Age: 32

Sexual orientation: Homosexual
Occupation: Police Officer

Interviewer:
Subject:

Interviewer:

Subject:

Interviewer:
Subject:

Interviewer:

Subject:

Can you tell me a bit about the intimate relationships in your life?
My partner and l have been together for over two years now. I
pretty much consider myself married at this point.

When you’re not attending conferences such as this one, what do
you do with your spare time?

I enjoy watching movies... in fact, my partner and l are both movie
buffs, so we spend a lot of time going to the theatre and having
“Blockbuster” movie nights.

Are you enjoying the conference?

Yes, I am. It’s been interesting to see what other people are up to.
We’ve run into some old friends... it’s been nice catching up with
them.

Have you had a chance to go to any clubs while you’ve been in
town for the conference?

That’s not really something I have an interest in anymore. I’ve
really just been hanging out with my partner, enjoying one another’s
company.

 

97

Positive Stereotype (Intelligent) Violation by Prototypical Taget Condition

 

Excerpt of Interview from “Gay Lifestyles in the New Millennium” Conference
Date: January 28, 2002

Name: Phil IV-

Gender: Male

Age: 32

Sexual orientation: Homosexual

Occupation: Interior Designer I

Interviewer: Can you tell me a bit about your educational background?

Subject: Well, I dropped out of high school in the 10th grade. I ended up
getting my GED a few years back, because I wanted to take some
college classes. I took a couple of math classes at night at a
community college, but didn’t do very good, so I decided not to
waste my time.

 

IFTT

Interviewer: When you’re not attending conferences such as this one, what do
you do with your spare time?

Subject: I watch a lot of TV. Reality shows are my favorite. When there is
nothing on TV, I try to read to keep up to date on current events,
but I get kind of bored with that real fast.

Interviewer: Are you enjoying the conference?
Subject: Yeah. It’s been interesting to see what other people are up to. We
ran into some old friends... it’s been nice catching up with them.

Interviewer: Have you had a chance to go to any clubs while you’ve been in
town for the conference?

Subject: Oh, every night so far. It’s the best thing about coming to these
things. But, it kind of makes it hard to get up in the morning, with a
hangover and all.

 

98

Positive Stereotype (Intelligent) Violation m Non-prototypical Tagget Condition

 

Excerpt of Interview from “Gay Lifestyles in the New Millennium” Conference

Date: January 28, 2002

Name: Frank G—
Gender: Male

Age: 32

Sexual orientation: Homosexual
Occupation: Police Ofﬁcer

Interviewer:

Subject:

Interviewer:

Subject:

Interviewer:

Subject:

Interviewer:

Subject:

Can you tell me a bit about your educational background?

Well, I dropped out of high school in the 10th grade. I ended up
getting my GED a few years back, because I wanted to take some
college classes. I took a couple of math classes at night at a
community college, but didn’t do very good, so I decided not to
waste my time.

When you’re not attending conferences such as this one, what do
you do with your spare time?

I watch a lot of TV. Reality shows are my favorite. When there is
nothing on TV, I try to read to keep up to date on current events,
but I get kind of bored with that real fast.

Are you enjoying the conference?
Yeah. It’s been interesting to see what other people are up to. We
ran into some old friends... it’s been nice catching up with them.

Have you had a chance to go to any clubs while you’ve been in
town for the conference?

Oh, every night so far. It’s the best thing about coming to these
things. But, it kind of makes it hard to get up in the morning, with a
hangover and all.

 

99

 

APPENDIX C

Table 1. Activation and Inhibition Expectancies for Experiment 1

 

 

 

Tarqet
Prime Black White
Black typical Strong Activation Strong Inhibition
Black subgroup Moderate Activation Strong Inhibition
Black subtype Moderate Inhibition Weak Activation
White typical Strong Inhibition Strong Activation

 

100

 

APPENDIX D

Table 2. Mean Facilitation Scores and Standard Deviations for Experiment 1
(Person Versus Object Judgments) and Experiment 2 (Black Versus White
Judgments)

Person versus Object Judgments

 

 

 

 

 

 

 

Taroet
Prime Black White
Black typical 1.13 (SD=1.32) 1.30 (SD=1.39)
Black subgroup 1.15 (SD=1.24) 1.18 (SD=1.26)
Black subtype 1.15 (SD=1.30) 1.20 (SD=1.26)
White typical 1.13 (SD=1.24) 1.18 (SD=1.30)
Black versus White Judgments

Target
Prime Black White
Black typical 1.07 (SD=1.18) 1.07 (SD=1.24)
Black subgroup 1.07 (SD=1.20) 1.05 (SD=1.15)
Black subtype 1.05 (SD=1.15) 1.03 (SD=1.15)
White typical 1.20 (SD=1.15) 1.05 (SD=1.18)

 

101

APPENDIX E

Table 3. Group Level and Subcategory Level Predictions for Experiment 2

 

 

 

Catemrigtion
Dependent Measures Grouping Subgrouping Subtyping
Group Perceptions
Perceived typicality Greatest Moderate Lowest
Stereotype change Least Less No change

 

stereotypic stereotypic

Perceived variability Most Moderate No change
heterogeneity heterogeneity

Subcategory Perceptions

Stereotype change No change More More
counterstereotypic counterstereotypic

Perceived variability No change More More
homogeneous homogeneous

 

102

APPENDIX F

Table 4. Categorization and Recategorization Expectations for Experiment 2

 

Perceiver Preiuﬂze
Valence of
Stereotype Violated Positive Negative Neutral

 

Prototypical Targets I
Positive Grouping Grouping Grouping

Negative Grouping Grouping Grouping

 

 

Nonprototyical Targets
Positive Subtyping Subgrouping Subtyping

Negative Subgrouping Subtyping Subtyping

 

103

APPENDIX G

Table 5. Actual Group Level and Subcategory Level Means and Standard

Deviations for Experiment 2

 

Perceiver Preiudicg

 

 

 

 

Targﬁet Prototypicality Positive Negative Neutral
Prototypical Targets
Group Perceptions:
Perceived typicality 5.46 (1.89) 5.41 (1.77) 4.88 (1.96)
Stereotype change -9.17 (19.36) 1.12 (19.95) 2.60 (15.64)
Perceived variability -6.79 (56.14) -4.18 (18.50) 15.24 (27.35)
Subcategory Perceptions:
Stereotype change -2.96 (17.92) 13.71 (17.50) -.88 (19.37)
Perceived variability 2.13 (29.97) 26.41 (39.15) 17.04 (42.72)
Nonprototypical Targe_ts
Group Perceptions:
Perceived typicality 4.61 (1.73) 4.88 (1.63) 4.78 (1.80)
Stereotype change 2.86 (12.50) -3.19 (11.55) -5.67 (16.94)
Perceived variability -2.82 (24.31) 1.44 (27.53) 7.39 (19.33)
Subcategory Perceptions:
Stereotype change -.21 (18.57) 3.88 (21.44) -4.50 (13.50)
Perceived variability 10.68 (26.65) 12.63 (23.29) 4.72 (25.50)

 

104

 

APPENDIX H

Table 6. Two-Way Interaction between Valence of the Target’s
Counterstereotypic Behavior and Target Prototypicality for Group Stereotype
Change

 

Valence of Target’s Counterstereogpic Behavior

 

Target Prototypicality Positive (Nonpromiscuous) Negative (Unintelligent)
Prototypic

(Interior designer) -8.84 (SD=17.83) 7.14 (SD=16.00)
Nonprototypic

(Police Ofﬁcer) -3.35 (SD=13.50) 1.00 (SD=14.40)

 

105

 

APPENDIX I

Table 7. Two-Way Interaction between Perceiver’s Prejudice Level and Target’s
Prototypicality for Group Stereotype Change

 

Perceiver's Prejudice Level

 

Target Prototypicality Positive Neutral Negative
l
Prototypic
(Interior designer) -9.17 (80:19.36) 2.60 (SD=15.64) 1.12 (SD=19.95)
Nonprototypic
(Police Ofﬁcer) 2.86 (SD=12.50) -5.67 (SD=16.94) -3.19 (SD=11.55)

 

 

106

APPENDIX J

Figure 1. Process of Categorizing a Counterstereotypic Group Member

 

Group Inclusion Group Exclusion

 

Encounter
group member

 

 

 

 

I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Categorize
based on group
membership
Emma
minimally
relevant?
No / §Yes
Does target ﬁt
my group
stereotype?
Yes / \No
magma
typical group
member?
Yes / \No
Is target’s behavior
consistent with my
prejudice?
Yes / \No
Grouping Subgrouping Subtyping

107

REFERENCES

Allport, G. (1954). The nature of prejudice. Reading, MA: Addison-Wesley.

Bargh, J. A., & Chartrand, T. L. (1999). The unbearable automaticity of being.
American Psychologist, 54, 462-479.

Bassili, J. N., & Smith, M. C. (1986). On the spontaneity of trait attribution:
Converging evidence for the role of cognitive strategy. Journal of
Personality and Social Psychology, 50, 239-245.

Bodenhausen, G. V., Kramer, G., & Suesser, K. (1994). Happiness and
stereotypic thinking in social judgment. Jgtimal of Personality and Social

Psychology, 66, 621-632.

Brewer, M. B., Dull, V., & Lui, L. (1981 ). Perceptions of the elderIy: Stereotypes
as prototypes. Journal of Personalig and Social Psychology, 41, 656-670.

Bruner, J. (1957). On perceptual readiness. Psychological Review. 64, 123-152.

Clifton, A. K., McGrath, D., & Wick, B. (1976). Stereotypes of women: A single
category? Sex Roles. 2, 135-148.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Erlbaum:
Hillsdale, NJ.

Collins, A. M., & Loftus, E. F. (1975). A spreading activation theory of semantic
processing. Psychological Review, 82, 407-428.

Devine, P. G. (1989). Stereotypes and prejudice: Their automatic and controlled
components. Jgumal pf Personality and Social Psychology, 56, 5-18.

Devine, P. G., & Baker, S. M. (1991). Measure of racial stereotype subtyping.
Personality and Social Psychology Bulletin, 17, 44-50.

Ditto, P. H., & Lopez, D. F. (1992). Motivated skepticism: Use of differential
decision criteria for preferred and nonpreferred conclusions. Journal of

Personal'm and Social Psychology, 63, 568-584.

Fazio, R. H. (1990). A practical guide to the use of response latency in social
psychological research. In C. Hendrick & M. S. Clark (Eds.), Review of
personality and sociaLpsychology, (Vol. 11, pp. 74-97). Newbury Park,
CA: Sage Publications.

108

Fazio, R. H., Jackson, J. R., Dunton, B. C., & Williams, C. J. (1995). Variability in
automatic activation as an unobtrusive measure of racial attitudes: A bona
ﬁde pipeline? J9_umal of Personality and §ocial Psychology. 69, 1013-
1027.

Fiske, S. T., & Neuberg, S. L. (1990). A continuum of impression formation, from
category-based to individuated processes: Inﬂuences of information and
motivation on attention and interpretation. In M. P. Zanna (Ed.), Advances

in experimental social psychology (Vol. 23, pp. 1-74). New York:
Academic Press.

Gilbert, D. T., & Hixon, J. G. (1991). The trouble of thinking: Activation and
application of stereotypic beliefs. Jo_umal of Personality and Social

Psychology, 60, 509-517.

Higgins, E. T. (1996). Knowledge activation: Accessibility, applicability, and
salience. In E. T. Higgins & A. W. Kruglanski (Eds.), Social psycholpgy:
Hangbook of pasic principles, (pp. 138-168). New York: Guilford.

Higgins, E. T., Rholes, W. S., & Jones, C. R. (1977). Category accessibility and
impression formation. Journal of Experimental Spcial Psychology, 13, 141-
154.

Jarvis, B. (2002). QrectRT Precision Timing Software. Empirisoft Corporation,
Version 2002.308.

Jarvis, B. (2002). MediaLab2001. Empirisoft Corporation, Version 2002.4.1.

Johnston, L., & Hewstone, M. (1992). Cognitive models of stereotype change: 3.
Subtyping and the perceived typicality of disconﬁrming group members.

Journal of Experimental Social Psychology, 28, 360-386.

Jones, E. E., Wood, G. C., & Quattrone, G. A. (1981). Perceived variability of
personal characteristics in in-groups and out-groups: The role of

knowledge and evaluation. Personalﬂ' and Social Psychology Bulletin, 7,
523-528.

Judd, C. M., Park, B., Ryan, C. S., Brauer, M., & Kraus, S. (1995). Stereotypes
and ethnocentrism: Diverging interethnic perceptions of African American

and white American youth. Journal of Personality and Social Psychology,
ﬂ, 460-481. '

Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108,
480-498.

109

Kunda, Z., & Oleson, K. C. (1995). Maintaining stereotypes in the face of
disconﬁrrnation: Constructing grounds for subtyping deviants. Journal of
Personality and Social Psychology, 68, 565-579.

Kunda, Z., & Oleson, K. C. (1997). When exceptions prove the rule: How
extremity of deviance determines deviants’ impact on stereotypes. Journal
of Personality and Social Psychology, 72, 965-979.

Macrae, C. N., Bodenhausen, G. V., & Milne, A. B. (1995). The dissection of
selection in person perception: Inhibitory processes in social stereotyping.
Journal of Personality and Social Psychology, 69, 397-407.

Maurer, K. L., Park, 8., & Rothbart, M. (1995). Subtyping versus subgrouping
processes in stereotype representation. Journal of Personality and Social
Psychology, 69, 812-824.

McCauley, C., & Stitt, C. L. (1978). An individual and quantitative measure of
stereotypes. Journal of Personally and Social lisychology, 36, 929-940.

McCauley, C., Stitt, C. L., & Segal, M. (1980). Stereotyping: From prejudice to
prediction. Psychological Bulletin, 87, 195-208.

McConnell, A. R. (1998). The role of affect in stereo in and sub in African-
American individuals. Paper presented at the 106t annual meeting of the
American Psychological Association, San Francisco, CA.

Osgood, C. E., & Tannenbaum, P. H. (1955). The principle of congruity in the
prediction of attitude change. Psychological Reviewp62, 42-55.

Park, 3., & Judd, C. M. (1990). Measures and models of perceived group
variability. ﬂimal pf Personality ang_SociaI Psychology, 59, 173-191.

Park, 3., Ryan, C. S., & Judd, C. M. (1992). Role of meaningful subgroups in
explaining differences in perceived variability for in-groups and out-groups.
Journal of Personally and Social Psychology, 6;, 553-567.

Park, 8., Wolsko, C., & Judd, C. M. (2000). Measurement of subtyping in
stereotype change. Journal of Experimental Social Psychology, 37, 325-
332.

Richards, Z., & Hewstone, M. (2001). Subtyping and subgrouping: Processes for

the prevention and promotion of stereotype change. Personality and
Social Psychology Review, 5, 52-73.

110

 

Rudman, L. A., 8 Borgida, E. (1995). The afterglow of construct accessibility: The
behavioral consequences of priming men to view women as sexual
objects. Journal oLExpenmental Social Psychology. 31, 493-517.

Schwarz, N., 8 Clore, G. L. (1996). Feelings and phenomenal experiences. In E.
T. Higgins 8 A. W. Kruglanski (Eds.), Social Psychology: Handbook of
Basic Principles (pp. 433-465), New York, NY: Guilford Press.

Sherif, M., 8 Hovland, C. I. (1961). Social judgment: Assimilation and contrast

effects in communication and attitude change. New Haven, CT: Yale
University Press.

Sinclair, L. 8 Kunda, Z. (1999). Reactions to a Black professional: Motivated
inhibition and activation of conﬂicting stereotypes. ippmal of Personality
and Social Psychology, 77, 885-904.

Smith, E. R. (1998). Mental representation and memory. In D. T. Gilbert 8 S. T.
Fiske (Eds.), The Handbook of Social Psychology (Vol. 2, pp. 391-445).
New York, NY: McGraw-Hill.

 

Spencer, S. J., Fein, S., Wolfe, C. T., Fong, C., 8 Dunn, M. A. (1998). Automatic
activation of stereotypes: The role of self-image threat. Personalig and
_S_opial Psyphology Bulletin, 24, 1139-1152.

Vonk, R., 8 OIde-Monnikhof, M. (1998). Gender subgroups: lntergroup bias
within the sexes. European Joumal of Social Psychology, 28, 37-47.

Weber, R., 8 Crocker, J. (1983). Cognitive processes in the revision of
stereotypic beliefs. ﬂmal of Personality and Social Psychology, 45, 961-
977.

Weber-Kollman, R. (1985). Subtyping: The development and conseguences of
differentiated categories for stereotyped groups. Unpublished doctoral

dissertation, Northwestem University, Evanston, IL.

111