Seta, J. J., Seta, C. E., & McElroy, T. (2003). Attributional biases in the service of stereotype maintenance: A
schema-maintenance through compensation analysis. Personality and Social Psychology Bulletin, 29: 151-163.
Published by SAGE (ISSN: 0146-1672).
Attributional Biases in the Service of Stereotype
Maintenance: A Schema-Maintenance Through
Compensation Analysis
John J. Seta, Catherine E. Seta, and Todd McElroy
ABSTRACT
Six experiments were conducted to test assumptions of a schema-
maintenance through compensation analysis. The results of
these experiments indicated that perceivers can compensate for
the inconsistent action of one individual (the target) by altering
their attribution concerning the action of a fellow group member.
When the target performed an inconsistent behavior, perceivers
compensated by making especially extreme stereotypically consistent
attributions concerning a fellow group member’s subsequent
action. In addition, in Experiment 5, perceivers compensated
via a fellow group member for a target’s inconsistent action while
maintaining their general view of group members. Experiments
also provided tests of the capability and motivation assumptions
of the schema-maintenance through compensation analyses.
A vast body of research on attributional processes has
generally assumed that causal analysis of a person’s
behavior is motivated by the person’s need to predict the
future (see Fiske & Taylor, 1991, for a review; see also
Heider, 1958; Jones & Davis, 1965; Kelley, 1973). Understanding
the dispositional or situational forces that are
associated with behavior enables people to feel that they
can control or at least predict events. Thus, attributions
are assumed to serve a functional role, contributing to a
person’s sense of self-efficacy. In addition, attributions
serve a role in interpersonal relations in influencing, for
example, the course of social interaction including close
relationships and intergroup behavior. The use of stereotypes
is also functional in this context in making the
social world more predictable (e.g., Allport, 1954; Fiske&
Taylor, 1991). However, functional utility comes with a
social cost; stereotypes often are maintained in the face
of disconfirming evidence that can lead to the perpetuation
of prejudice and bias.
Research has shown that the type of attribution a
perceiver makes can influence whether an aberrant
behavior has an impact on a perceiver’s representation
of the stereotyped group (e.g., Crocker, Hannah, &
Weber, 1983; Hewstone, 1989). For example, Crocker
et al. (1983) found that the type of attribution that was
made about an inconsistent action influenced the
impact of the inconsistent action on memory for the
action and representations of the behavior. The aberrant
act had a stronger influence on changing representations
of social groups when it was described as being
internally, rather than externally, motivated. Further
research has demonstrated that stereotypes are most
likely to change when counterstereotypic behaviors are
performed by group members thought to be typical
members of outgroups and when these behaviors are
attributed to stable, internal causes (e.g., Wilder, Simon, &
Faith, 1996).
There are several perspectives that implicate attributions
in the process of stereotype maintenance/change
(e.g., Crocker et al., 1983; Hewstone, 1989; Johnston,
Bristow,&Love, 2000; J. J. Seta&Seta, 1993; Wilder et al.,
1996; Yzerbyt, Rogier, & Fiske, 1998). Thus, exploring
the factors that influence the outcomes of attributional
search processes is an important area of investigation. In
our research, we focused on the role of category-based
expectancies on individuals’ attributions and perceptions
of a behavior that followed exposure to stereotypically
inconsistent acts. Specifically, we were interested in
determining if the nature of perceivers’ attributions was
influenced by their propensity to compensate for the
stereotypically inconsistent behaviors of a group member.
The present research follows from our previous
work on compensatory motives in social judgment. First,
we describe this compensatory motivation analysis followed
by a description of the current research.
SCHEMA-MAINTENANCE
THROUGH COMPENSATION
A schema includes goals, expectations, self-concepts,
stereotypes, moods, and other such organizing concepts
and motives. Compensation is not restricted to one type
of schema or one type of motivation. For example, people
can compensate in an attempt to maintain a goal, to
maintain control, or to maintain a particular view of
themselves or others (e.g., J. J. Seta, Seta,&Erber, 1993;
J. J. Seta & Seta, 1993). When individuals attempt to
maintain a schema, they are effectively trying to prevent
a schema from changing. Thus, this model has particular
relevance to situations in which individuals attempt to
preserve their stereotype in the face of disconfirming evidence.1
In this article, we are concerned with stereotype
maintenance. According to our view, exposure to
counterstereotypic behaviors can result in the generation
of hyperconsistent expectations that the group
member will conform to the stereotype in the future by
performing a behavior that is hyperconsistent with the
stereotype (J. J. Seta&Seta, 1993). For example, observations
of a member of a social group, typically thought to
be lacking in intelligence, performing an intellectual
task may lead individuals to expect that this group member
would be likely to perform in a manner especially
lacking in intellectual acuity in the future. This expectation
may effectively help maintain the negative group
stereotype. Similarly, individuals may attempt to maintain
positive stereotypes by generating compensatory
expectations, such as when observations of a minister
behaving in an immoral manner lead individuals to
expect the minister to behave in an ultrarighteous manner
at a future date. This expectation may or may not
effectively counter the influences of observed inconsistencies.
Regardless of whether the generation of
compensatory expectations effectively maintains the stereotype,
we should observe the influences of this motive
on subsequent social judgments under certain
conditions.
According to our schema-maintenance through compensation
approach, compensation can occur when an
event or experience is subjectively inconsistent with a
stereotype or schema (J. J. Seta & Seta, 1993; J. J. Seta,
Seta, & Erber, 1993). In this situation, compensation will
occur when a person has an implicit or explicit motivation
to maintain the stereotype or schema and the person
has the capability of generating information (or
behavior) that could potentially offset the inconsistency.
In support of this assertion, research has demonstrated
that cognitive capacity is necessary for stereotype maintenance
in the face of receiving inconsistent information
about a target (e.g., Moreno & Bodenhausen, 1999;
Yzerbyt, Coull, & Rocher, 1999). Individuals are especially
willing to expend cognitive effort in generating
compensatory information when, for example, important
or valued schema are threatened by inconsistent
information. Therefore, within limits, one’s capability of
generating information may vary as a function of schema
importance.
Perceivers’ motivation to compensate should be influenced
by factors such as the schema’s subjective importance
(C. E. Seta, Seta, & Goodman, 1998), schema
strength, and the consequences of maintaining this
schema. There are also several factors that constrain a
perceiver’s capability of accessing or generating potentially
compensating information. These include factors
such as the frequency and intensity of observed inconsistency.
As evidence mounts that disconfirms expectations,
it becomes increasingly difficult to disregard the
data (e.g., J. J. Seta & Seta, 1982, 1992; Wilson, Lisle,
Kraft, & Wetzel, 1989). At this point, expectations may
not be completely guided by schematic knowledge, and
predictions about the deviant target’s behavior may be
guided by bottom-up observations of the target’s behavior.
In this case, the perceiver is basing his or her expectations
on the observed behavior of the deviant target, and
perceivers will not expect compensation from the deviant
actor.
According to our view, even though the perceiver is
unable to compensate for a target’s inconsistent actions,
it is possible that the perceiver will expect compensation
from a different group member. A fellow group member
and the target of the inconsistent action are related in a
perceiver’s knowledge structure (e.g., Fiske & Neuberg,
1990; Fiske & Taylor, 1991). Therefore, perceivers may
attempt to maintain their stereotype by expecting the
future action of a nondeviant group member to compensate
for the inconsistent actions of a deviant member.
There are several reasons why a perceiver can expect
compensation from a fellow group member under conditions
in which he or she does not expect compensation
from a deviant target. First, the assumed variability
among members of a group is often greater than the
assumed variability in a single individual’s behavior. For
example, we generally are not surprised when grades in
our class range from As to Fs, but we are surprised when
the grades of a single student range from As to Fs. People
often expect more behavioral deviance among a group
of people than within a single individual (e.g., Hamilton &
Sherman, 1996). This effect also can occur because the
implications of the inconsistent actions of the deviant
are more directly tied to the personality of the deviant
target than to the personality of a fellow group member.
For both reasons, it should take an objectively less severe
inconsistent action to constrain a perceiver’s capability
of generating compensatory expectations about the
target of the inconsistent actions than a fellow group
member’s future actions. Likewise, it should take an
objectively milder inconsistent action to generate compensatory
expectations for the target than a fellow group
member. It follows, then, that after a very inconsistent
action on the part of the target, we may observe compensation
for a fellow group member but not for the
target.2
These predictions were supported in a series of experiments
(C. E. Seta et al., 1998; J. J. Seta & Seta, 1993). In
one experiment, for example, we found that after a target
performed a very mild inconsistent action, participants
generated compensatory expectations about the
future action of the target; however, following a highly
inconsistent action, participants generated compensatory
expectations for the fellow group member but not
for the acting target. Just as perceivers demonstrate compensatory
motivation by elevating their future expectations
of a group member’s actions, the type of attribution
they make about a group member’s actual behavior also
may reflect this motive. In particular, individuals may be
biased toward making more extreme internal attributions
for a group member cohort’s stereotypically consistent
behavior in an effort to maintain their existing
group schema. Thus, an influence of compensatory
motives on patterns of attributions would be demonstrated
by finding more extreme dispositional attributions
for stereotypically consistent behaviors in the presence
versus the absence of inconsistency. The present
research explored this possibility.
OVERVIEW: EXPERIMENTS 1 AND 2
In these studies, we explored the viability and
generalizability of the schema-maintenance through
compensation analysis with respect to participants’
interpretations and attributions of potentially compensatory
behaviors. In all studies, participants were presented
with the aberrant behavior of a member of a stereotyped
group, and positive group stereotypes were
used that are typically associated with either women
(nurses) or men (ministers). Experimental condition
participants were exposed to stereotypically inconsistent
information about a member of one of these groups and
were asked to make judgments about the causes of a
stereotypically consistent behavior that was described as
emanating from the aberrant group member (same target
conditions) or a fellow member of the target’s social
group (fellow group member conditions). In addition,
control conditions were employed in which participants
were not exposed to stereotypically inconsistent behaviors
and were asked simply to make judgments about the
causes of the stereotypically consistent behaviors.
These controls allowed for an estimate of baseline
attributions used for reference points in assessing the
extent to which exposure to inconsistency changed patterns
of attributions. An influence of compensatory
motives on patterns of attributions would be demonstrated
by finding more extreme dispositional attributions
for stereotypically consistent behaviors in experimental
versus control conditions. We expected to see
more extreme attributions about the causes of fellow
group members’ stereotypic behavior, which are consistent
with the disposition of the actor, after being exposed
to the inconsistent behaviors of a target group member.
Participants would be indicating stronger stereotypically
consistent dispositional causes for the behavior than
baseline estimates. The primary purpose of these studies
was to demonstrate the generalizability of the hypothesized
effects across materials and stereotypes. Further
studies, reported in this article, were designed to rule out
potential alternative explanations, such as contrast
effects. These findings would dovetail with prior
research (e.g., C. E. Seta et al., 1998; J. J. Seta & Seta,
1993) that has found increased expectations for
stereotypically consistent behaviors on the part of fellow
group members after exposure to highly inconsistent
behaviors of an acting group member.
METHOD: EXPERIMENTS 1 AND 2
Stimulus Materials
In all experiments, inconsistent behaviors were chosen
from a normative set collected for prior research
(J. J. Seta&Seta, 1993) and were considered to be highly
inconsistent with the stereotyped group. The target compensatory
behavior used in Experiments 1 and 2 also was
selected from these norms. In addition, a separate sample
of participants was given the target compensatory
behavior used in these studies. They were asked to
assume that either a minister (Experiment 1) or a nurse
(Experiments 2-6) had performed a behavior and were
asked if this behavior was consistent or inconsistent with
the internal character of the person’s group. For ministers,
they were asked to assume that a minister volunteered,
in the presence of other ministers, to help the
March of Dimes (a charitable, philanthropic group)
over the weekend. Then, they were asked which of two
conclusions for this behavior (he did it because he was
totally concerned about the March of Dimes or he
wanted to make a good impression) was most consistent
and which was most inconsistent with the internal character
of ministers. Of the 30 participants that were sampled,
27 believed that the conclusion that suggested that
the minister helped because he was concerned about the
March of Dimes was most consistent with the character
of ministers, X
2
= 19.2, p < .001, whereas 28 believed that
the conclusion that he wanted to make a good impression
was most inconsistent with the character of ministers,
X
2
= 22.5, p < .001. For nurses, participants were
asked to assume that a nurse volunteered, in the presence
of other nurses, to help the March of Dimes over
the weekend. Then they were asked which of two conclusions
(she did it because she is a truly generous person or
because she simply wanted to impress) was most consistent
and which one was most inconsistent with the internal
character of nurses. Of the 30 participants, 28
believed that the conclusion that she was truly generous
was most consistent, X
2
= 22.5, p < .001, whereas 29
believed that the conclusion that she wanted to impress
was most inconsistent, X
2
= 26.1, p < .001.
Participants, Design, and
Procedure (Experiments 1 and 2)
Experiment 1. Ninety-one women and 21 men participated
in Study 1. No participant was included as an overlapping
participant in any of the studies reported in this
article. Participation was rewarded by the receipt of partial
credit for the research requirements of introductory
psychology classes at the University of North Carolina,
Greensboro.
In Experiment 1, participants were assigned randomly
to one of four between-subjects conditions
defined by the factorial crossing of two levels of actor
inconsistency (a high-inconsistency experimental condition
and no inconsistency control) with two levels of target
identity variable (aberrant group member and a fellow
member condition). The study was conducted in
small groups where the number per group typically
ranged from 8 to 12. Participants were told that, in the
present study, they would be asked their impressions of
other persons. They were then given written descriptions
comprising one of the four conditions used in the
study; all conditions were approximately equally represented
within each experimental session.
Participants in the high-inconsistency experimental
condition read about a minister who had sexually
molested a 7-year-old and sold heroin to a teenager. In
the same aberrant group member condition, they then
read that this minister volunteered in the presence of
other ministers to help the March of Dimes in the forthcoming
months. They were then asked to consider why
this minister had volunteered; they rated their attribution
along a 7-point scale anchored by the endpoints 1
(to make a good impression) and 7 (he is totally concerned
about the March of Dimes).
Participants in the fellow group member condition
read about the inconsistent behaviors of the aberrant
minister but were told that another minister, unrelated
to the inconsistent actor, volunteered to help the March
of Dimes. They were then asked to rate the reason why he
volunteered along the 7-point scale described above.
Participants in the two control groups were not
exposed to inconsistent information about ministers
and were simply asked to rate the reasons why a minister
would volunteer or why a different minister would volunteer.
All participants were asked a probe question serving
as a manipulation check in which they were asked to indicate
the identity of the rating target. Following this question,
participants were thanked and debriefed.3
Experiment 2. Similar procedures were used in Experiment
2. However, this study employed 73 female participants.
In the high-inconsistency conditions, participants
read a scenario in which they were asked to imagine finding
a nursing student described as red-faced and grasping
a frail elderly man by the collar while slapping him in
the face. A broken glass was described as lying close-by
on the floor in a puddle of milk. They then read that later
that week, a member of the community asked Mary to
donate a great deal of time on a Saturday to help a charitable
organization. This request was described as made
in the presence of other persons.
At this point, our manipulation of the attributional
target was made. Participants in a high-inconsistency–
same target condition read that Mary was the nursing
student that they had just read about in the preceding
scenario (i.e., the aberrant nurse), whereas participants
in a high-inconsistency–different target condition read
that Mary was not the nursing student described in the
preceding scenario but was a different nursing student
from the one who performed the aberrant actions. Participants
were then asked to rate their attributions for
why Mary volunteered along a similar 7-point scale used
in Experiment 1.
To obtain a baseline estimate of the extent to which
volunteering was considered to be a dispositional/situational
motive for nursing students, an additional group
of participants was asked simply to assume that Mary was
a nursing student who was asked to volunteer without
exposure to any stereotypically inconsistent information.
Thus, the design of Experiment 2 included three
between-subject conditions (high-inconsistency–same
target; high-inconsistency–different target; no inconsistency
control). In addition to the probe question used
in Experiment 1, participants in this study were asked
whether they were nursing majors.4
RESULTS AND DISCUSSION:
EXPERIMENTS 1 AND 2
Participants’ rating scores in Experiment 1 were
entered into a 2 (group member identity) x 2 (inconsistency)
between-subjects ANOVA. The analysis revealed a
main effect of the target identity variable, F(1, 101) =
25.97, p < .001, and a marginally significant inconsistency
main effect, F(1, 101) = 2.94, p < .09. These effects were
qualified by a Target Identity × Inconsistency interaction,
F(1, 101) = 39.80, p < .001. This interaction was
interpreted via planned comparisons. As expected, participants
who read about the aberrant behaviors of the
minister attributed his generosity in volunteering to the
dimension that was more inconsistent with the
dispositional attribution of ministers—impressing others
(M = 2.26)—in contrast to the same target control
participants (M= 4.62), F(1, 101) = 34.06, p < .01. On the
other hand, participants in the fellow group member
condition attributed the other minister’s generosity to
the dimension that was more consistent with the disposition
of ministers (M = 5.61) in contrast to comparable
control participants who were not exposed to the inconsistent
actions (M= 4.26), F(1, 101) = 10.31, p < .01. Thus,
exposure to the previous aberrant behavior of a minister
produced stronger dispositional attributions for
another group member’s stereotypically consistent
behavior.5
A conceptually similar pattern of results was obtained
in Experiment 2. The results of the one-factor between subjects
ANOVA revealed significant differences
between conditions, F(2, 58) = 39.23, p < .0001. Planned
comparisons among the cell means indicated that participants
weighted factors that were inconsistent with
the dispositional character of nurses more heavily in
their attributions of the nursing student’s charity work
when she was described previously as performing
stereotypically inconsistent behaviors (M = 1.75) than
did control participants (M = 3.81), F(1, 58) = 33.59, p <
.001. In contrast, participants who had read about the
aberrant behavior of the nursing student rated the charity
work of another nursing student as more consistent
with nurses’ dispositional character (M = 4.89) than did
control participants (M = 3.81), F(1, 58) = 9.23, p < .01.
Thus, Experiment 2 provided a conceptual replication
of Experiment 1 in demonstrating a role of compensatory
motives on attributions of causality.
INTRODUCTION: EXPERIMENT 3
The results of Experiments 1 and 2 dovetail with our
previous research (e.g., C. E. Seta et al., 1998; J. J. Seta &
Seta, 1993) in suggesting that exposure to stereotypically
inconsistent behavior often invokes a motive to compensate
for this deviance by expecting hyperconsistent
behavior on the part of another group member, thereby
increasing the weight of attributions that are consistent
with the disposition of the actor. In these studies, both
the aberrant actor and the fellow group members were
related via their common group membership. The theoretical
rationale for the obtained predictions of extreme
attributions vis-à-vis a fellow group member was based on
the assumption that the attributional process was biased
by motives to maintain the integrity of the group concept
in the face of observed inconsistency. In an attempt to
accomplish this, perceivers generated attributions that
were consistent with those of the target’s group.
It follows from this perspective that such results
should only be obtained when the attributional target
(e.g., fellow group member) shares a perceived unit relationship
with the deviant actor. That is, if patterns of
attributions are colored by motives to maintain existing
concepts, then these motives would only be operative
when such attributions are potentially functional in
meeting the maintenance goal. Please note, we are not
assuming that such attributions result in the successful
maintenance of extant stereotypes; they may or may not
result in stereotype maintenance. Rather, the assumption
in this and our prior work is that attributional and
expectation-generation processes (and even behavior)
are biased in the direction of stereotype/concept maintenance.
If this reasoning is correct, we should observe
these biases only in contexts in which attributional targets
are related to the aberrant targets; for example,
when they are members of the stereotyped group “under
threat.”
However, if some sort of perceptual contrast is operative
in the setting, then the attributional target would be
expected to appear more extreme in contrast to the
aberrant actor regardless of his or her relationship to the
actor. That is, if exposure to a deviant group member
leads individuals to perceive a subsequent target in a
more extreme light in contrast to the actor, any distinct
target may be subject to this process (e.g., Martin & Seta,
1983; Stapel & Spears, 1996).
Study 3 assessed this prediction by varying whether
the attributional target did or did not share a group unit
relationship with an aberrantly behaving group member.
The target was either a member of the actor’s group
or was not a fellow group member. If perceptual contrast
effects underlie observed attributional patterns, then
both the fellow group and different group member target
conditions should show more extreme attributions
than controls. However, if compensatory motivations are
operative, then extreme attributions should be evidenced
in the fellow group member versus the different
group member condition.
METHOD: EXPERIMENT 3
Participants and Design
Seventy-two female students from introductory psychology
classes served as participants and received partial
credit for their research requirements. The design
factorially crossed two levels of inconsistency (no inconsistency
control and an inconsistent experimental condition)
with two levels of fellow group member target identity
(same group, different group).
Procedure
Participants were oriented to the experiment in a
manner identical to the previously described methods.
Inconsistent information was given about a nurse who
was “red faced, grasping a small child by the collar and
slapping him painfully in the face”: the inconsistent condition.
This information was not presented to control
condition participants. Participants read that a member
of the community asked a different nurse (in the same
group identity condition) or asked a salesperson (in the
inconsistent different group identity condition) to
donate the better part of a Saturday to help an organization
while in the presence of others. All participants were
asked the likelihood that this person donated time
because she was truly a generous person or because she
simply wanted to impress. Participants rated their attributions
along a 7-point scale in which a 1 indicated that
the person volunteered because she simply wanted to
impress, whereas a 7 indicated that the person volunteered
because she was truly generous. Control nurse
and control salesperson participants were asked the
same question. The only difference was that they were
not exposed to the inconsistent information concerning
a nurse’s inconsistent behavior.6
RESULTS AND DISCUSSION: EXPERIMENT 3
Participants’ rating scores were entered into a 2 x 2
between-subjects analysis of variance. Table 1 contains
the means of the four experimental conditions. The
analysis revealed a significant Fellow Group Member
Identity (same group or different group) x Inconsistency
interaction, F(1, 60) = 5.55, p < .03. Neither main
effect reached significance.
To interpret this interaction, and to test our specific
predictions, we used the error term of the overall analysis
to make specific comparisons. Planned comparisons
indicated that, in the same group condition, participants
in the inconsistent condition had ratings that were significantly
more related to the internal characteristics of
nurses than control participants, F(1, 60) = 7.97, p < .01.
In the different group condition, participants’ ratings in
the inconsistent condition did not statistically differ from
those of control group participants, F(1, 60) = 0.21, ns.
These results add to those obtained in Experiments 1
and 2 in supporting our assumption that patterns of attributions
result from perceivers attempt to maintain stereotypic
views in the face of inconsistent information.
Perceivers made more extreme internal attributions of a
related group member. They did not, however, change
their attribution of a person who was not associated with
the target group. This was expected because a person
who is not associated with the target’s group does not
help maintain the target’s stereotype; salespersons do
not share the trait of “generosity” with the nurse stereotype.
Thus, as expected, compensatory attributions were
seen in the different nurse condition but not in the different
group identity condition.
EXPERIMENT 4
Kunda and Oleson (1997) measured the strength of
perceivers’ stereotypes after providing them with information
concerning a deviant group member. In one
study, they found that the strength of a perceiver’s stereotype
was enhanced following information about an
extremely deviant group member.
Although Kunda and Oleson (1997) did not provide a
direct test of their perspective, they offered two potential
reasons for their effects. One reason involved perceptual
contrast. Following exposure to an extremely deviant
group member, they reasoned that perceivers contrasted
the prototypical group member with the deviant, resulting
in an especially strong stereotype (perceptual contrast
also was discussed by J. J. Seta & Seta, 1993). Their
second reason involved surprise; because perceivers
were surprised by the extremely deviant group member,
they searched their knowledge structure for information
that confirmed their original stereotype, and because
they could easily construe information that confirmed
their stereotype, their search resulted in an especially
strong stereotype.
Although Study 3 suggests that a simple form of perceptual
contrast does not account for our findings, it
does not rule out the operation of active search processes
for stereotype-confirming information. Indeed,
similar to Kunda and Oleson (1997), our model also
involves active search processes. However, we assume
that active search processes follow from an implicit or
explicit attempt to maintain a schema (mood, goal, or
expectation). From our view, this search is initiated by
perceivers’ motivation to maintain a stereotype, not
because they are simply surprised by the deviant’s abnormal
actions or characteristics.
One way to distinguish our view from that of Kunda
and Oleson (1997) would be to demonstrate that
perceivers who are motivated to maintain a stereotype
engage in more compensation than those who have a
lesser degree of motivation to maintain a stereotype. In
contrast to the view of Kunda and Oleson, our analysis
specifies specific compensatory motives assumed to
underlie social judgments. To test this assumption, we
can compare the responses of individuals who differ in
their motivation to maintain their schema in the face of
inconsistent information. Individuals who are especially
motivated to maintain their schema also should be especially
inclined to demonstrate the influence of compensatory
motives following observed inconsistency.
An individual difference methodology, using individuals
who were differentially motivated to maintain a
group concept, has provided evidence in support of our
interpretation (C. E. Seta et al., 1998). Following information
about a deviant group member, highly motivated
individuals demonstrated compensatory expectations,
whereas individuals low in motivation did not. Although
this study provided support for our motivation interpretation,
it is not entirely clear whether these results were
due to differences in perceivers’ motivation or to differences
in their knowledge structure. In Experiment 4, we
attempted to control for these types of effects by altering
participants’ motivation within the experimental session.
To vary motivation, some participants were given
information that we expected would increase their motivation
to maintain a stereotype, whereas others were
given information that would decrease motivation. If
perceivers’ motivation to maintain their schema is a critical
dimension, then individuals who were given information
to bolster their motivation should demonstrate
effects of compensatory motives in their social judgments
in contrast to those given information that weakened
this motivation.
In this experiment, we manipulated participants’ levels
of motivation by describing belief-maintenance as a
process either that had been found or had not been
found to lead individuals to be successful in life and
relationships. For example, in the high-motivation condition,
we told participants that research showed that
persons who hold on to their beliefs in the face of
disconfirming evidence are found to be successful in
their relationships. In the low-maintenance condition,
participants were told that holding on to beliefs in the
face of disconfirming information was nonfunctional
and led individuals to be unsuccessful in relationships.
Onthe surface, these manipulations may appear to be
rather heavy handed and artificial. One response to the
criticism of artificiality is that this is an advantage of laboratory
research, which allows us to construct contexts
appropriate for theory testing that do not occur in the
real world. But, regardless of this advantage, it is clear
that differential motivations to maintain preexisting
beliefs of the type used in this study do occur naturally.
For example, many women believe that holding on to
previous attitudes about their partners in the face of
disconfirming evidence, such as abuse, will ultimately
lead to a successful relationship, whereas other women
more quickly revise their beliefs in the context of similar
information. Although there are individual differences
that underlie these orientations, contextual variables are
also potent determinants of these motivations (e.g.,
interdependence theory).
We formulated our hypotheses so that they would not
conform to potential demands that on the surface may
seem to be apparent in this context. For example, in the
high-motivation condition, participants were exposed to
an actor who responded in a manner inconsistent with
their extant stereotypes. If participants simply conform
to instructions to maintain their preexisting views in the
face of disconfirming evidence, then their responses
should be similar to those of control condition participants.
However, predictions made in the study called for
participants in the high-motivation condition to be different
from controls; specifically, these participants were
expected to respond with relatively more extreme internal
attributions (consistent with the character of the
actor) concerning the stereotypically consistent behavior
of a group member.
METHOD: EXPERIMENT 4
Participants, Design, and Procedures
Eighty-seven female students from introductory psychology
classes served as participants. The general procedures
were identical to those described in the other
studies. The design factorially crossed two levels of motivation
to maintain a stereotype (low or high) with two
levels of inconsistency (no inconsistency control and an
inconsistency experimental condition).
Participants were first given the motivation manipulation.
In the high-motivation condition, they read that
prominent researchers believed that individuals who
were motivated to maintain their stereotypes were likely
to be successful in life (e.g., in relationships). They were
then given an example of a woman who “believes that
men who belong to a particular group—a group that she
has known for a long time—are generally positive.” They
were then told,
These researchers believe that this woman should maintain
her attitude toward this group even if one of the
members behaves in a negative way. They believe that
she should try to maintain her attitude toward this group
even if she finds that a particular group member has
acted in a negative way.
In the low-motivation condition, participants read
that prominent researchers believed that individuals
who were motivated to maintain their stereotype were
likely to be unsuccessful in life (e.g., in relationships).
They were then given an example of a woman who “believes
that men who belong to a particular group—a
group that she has recently been acquainted with—are
generally positive.” They were then told that,
These researchers believe that this woman should not
maintain her attitude toward the group if one of the
members behaves in a negative way. They believe that
she should not try to maintain her attitude toward this
group when she finds out that a particular group member
has acted in a negative way.
In both motivation conditions, participants were told
that the experimenter wanted them to think about how
they felt about the information they just read and that
the experimenter would ask them their opinion at the
end of the session.
Participants in the inconsistent condition were given
information identical to that described earlier about a
female nursing student who slapped a child. They were
then given the information about a different nursing student
(Mary) who was asked to donate time to charity in
the presence of other nursing students.
All participants were then asked the likelihood that
this female nursing student donated her time because
she is a truly generous person or because she simply
wanted to impress, using the 7-point scale in which a 1
indicated that the nurse volunteered because she simply
wanted to impress and 7 indicated that she volunteered
because she was truly generous. No-inconsistency control
participants were treated identically with the exception
that they were not exposed to the inconsistent information.
As in the prior studies, participants were asked
the probe question concerning the identity of the rating
target.7 Following these procedures, participants were
debriefed and thanked for their participation.
Results and Discussion:
Experiment 4
Participants’ rating scores were entered into a
between-subjects ANOVA. The analysis revealed a significant
Motivation × Inconsistency interaction, F(1, 71) =
4.33, p < .05. Table 2 contains the means for each of our
four conditions.8 To interpret this effect, and to test our
specific predictions, we used the error term of the overall
analysis to make specific comparisons. Planned comparisons
indicated that participants in the high-motivation
inconsistent condition had more extreme attributions
(consistent with the internal character of nurses) than
those in the low-motivation inconsistent condition, F(1,
71) = 6.99, p < .02, whereas the ratings of low- and high-motivation
control participants did not differ, F(1, 71) =
.01, ns. Furthermore, participants in the high-motivation
inconsistent condition had more extreme internal attributions
than those of control participants, F(1, 71) =
4.16, p < .05, whereas the ratings of participants in the
low-motivation inconsistent condition did not differ
from that of control participants, F(1, 71) = 1.22, ns. In
fact, they were somewhat lower than those of control
participants.
The results of Experiment 4 are entirely consistent
with our expectations. As expected, an effect of compensation
on attributions was obtained in the high-motivation
condition. And, as expected, this effect was
not obtained in the low-motivation condition. These
results argue for an interpretation in which perceivers’
motivation to maintain stereotypic beliefs is a critical
dimension for the generation of attributions.
INTRODUCTION: EXPERIMENT 5
Our view assumes that perceivers attempt to maintain
a stereotype in the face of inconsistent information by
engaging in compensation and that compensation typically
occurs at the component level. That is, in an
attempt to maintain a stereotype, perceivers alter their
judgment of an individual who comprises membership
of the group. In so doing, the overall evaluation of the
target group may be maintained. However, as previously
noted, this model does not assume that compensatory
processes always lead to the successful maintenance of
extant group concepts. Rather, the model posits that
maintenance motives influence judgments about group
members in a direction that is compensatory in nature.
In contrast, Kunda and Oleson’s (1997) research was
concerned with how the stereotype itself changes as a
result of receiving inconsistent information about a tar-
get group member. In their studies, participants were
first given information about a deviant group member.
After evaluating the deviant group member, participants
were asked about their stereotype—they were asked to
judge members of the group, in general. Their assumption
was that perceivers’ judgment of a group member is
equivalent to their judgment of the stereotype. Our view
does not make this assumption. Rather, we assume that
perceivers can attempt to compensate at the individual
level while attempting to maintain their stereotype. Individuals
can have a hyperconsistent view of a fellow group
member while maintaining a different view of members
in general (their stereotype). Consequently, judgments
of a fellow group member can be significantly affected by
the inconsistent actions of the target, whereas stereotypes
may be largely unaffected by this information.
Experiment 5 was designed to test this idea. We asked
some participants about a fellow group member,
whereas others were given a stereotype question about
members of the group in general. This question was conceptually
similar to the one used by Kunda and Oleson
(1997).
METHOD: EXPERIMENT 5
Participants, Design, and Procedure
Seventy-eight female students from introductory psychology
classes served as participants. The design
included three group conditions: no inconsistency control
(measuring judgments of nurses, in general) and
two inconsistent target conditions—one that measured
judgments of a fellow group member (inconsistent fellow
group member condition) and one that measured
judgments of nurses, in general (inconsistent stereotype
condition).
Participants were oriented to the study in a manner
identical to the previously described procedures. In this
study, all participants read that there are approximately
65,000 female student nurses in the country. In the no
inconsistency control condition, participants were asked
to consider that “if all of the student nurses were asked to
donate the better part of a Saturday to help an organization,
what would be the likelihood that they would help
because they are, in general, truly generous?” They
responded on a 7-point scale where 1 represented not at
all generous and 7 represented truly generous. Participants
in the inconsistent stereotype target condition were also
presented this information. However, before being
asked the likelihood question concerning the entire
population of student nurses, participants in this condition
first read information about a female student nurse
who slapped the child.
In the inconsistent fellow group target condition, participants
were exposed to the procedures described
above; however, they were not asked about their opinion
of the entire population of student nurses. Rather, they
were asked the likelihood that a different female nursing
student would help because she is truly generous on the
7-point scale described above. In this study, our probe
question simply asked whether our noncontrol participants
evaluated an entire population (65,000 student
nurses) or a different individual. All participants gave
the correct answer.
Results and Discussion:
Experiment 5
Participants’ scores were entered into a three-factor
between-subjects ANOVA. The analysis revealed a significant
condition main effect, F(2, 75) = 5.60, p < .005. To
interpret this effect, and to test our specific predictions,
we used the error term of the overall analysis to make
specific comparisons. Planned comparisons indicated
that participants in the inconsistent fellow group member
target condition (M = 5.01) had more extreme attributions
that were consistent with the internal character
of nurses than those in the no inconsistency control (M=
4.09), F(1, 75) = 10.2, p < .01. The results conceptually
replicate those obtained in Experiments 1 through 4. In
addition, the judgments of participants in the fellow
group member target condition were significantly more
consistent with the internal character of nurses than
those in the inconsistent stereotype target condition (M=
4.25), F(1, 75) = 6.56, p < .05. Furthermore, judgments in
the inconsistent stereotype condition did not differ from
those in the no inconsistency control (M = 4.09), F(1,
75) = .31, p > .60.
These results are entirely consistent with our predictions.
Perceivers had more extreme attributions that
were consistent with the character of nurses following
exposure to the inconsistent actions of a deviant nurse.
They did not, however, change their view of nurses, in
general. Their judgments in the inconsistent stereotype
condition did not differ from those in the control condition.
Thus, these results suggest that judgments about a
particular group member can reflect compensatory processes,
whereas judgments about the stereotyped group
per se remain relatively equivalent to conditions in
which no inconsistency was encountered. Therefore, it
appears that the overall evaluation of the group stereotype
was maintained in the face of inconsistency. This
pattern of results is quite different from that of Kunda
and Oleson (1997), who found hyperconsistent judgments
of the group stereotype in their research. It may
be the case that Kunda and Oleson’s results were
obtained because participants in their study rated the
person who performed the inconsistent action immediately
before being asked about the group stereotype.
This procedure may have produced response scale shifts
such that participants’ group ratings were inflated following
their previous negative evaluations of the inconsistent
actor.
The present results also provide evidence that
perceivers did not compare and contrast the valence of
the target’s actions to that of other group members. If
this was the case, the judgments of participants in the
inconsistent stereotype target condition should have
been more extreme than that of those in the control
condition.
INTRODUCTION: EXPERIMENT 6
Experiment 6 was designed to provide a test of our
model’s capability component. In Experiment 4 and in
prior research (C. E. Seta et al., 1998), we have shown
that perceivers compensate if they are motivated to
maintain their stereotype. However, this does not mean
that they will always be capable of accessing or generating
information that can potentially compensate for all
types of inconsistent actions. For example, perceivers
with a certain level of motivation may be capable of compensating
for the inconsistent action of one target but
not those of multiple actors. Experiment 6 was designed
to determine if perceivers are capable of compensating
for the inconsistent actions of one target nursing student,
but not those of multiple nursing students.
In addition to supporting our capability assumption,
the expected results also would provide additional evidence
against an interpretation in which perceivers compare
the valence of the target’s behavior to that of a fellow
group member. If this type of contrast is occurring,
perceivers should see the action of a fellow group member
as especially positive after observing the inconsistent
actions of both multiple and individual targets.
METHOD: EXPERIMENT 6
Participants, Design, and Procedures
Seventy-seven female students from introductory psychology
classes served as participants and received partial
credit for their research requirements.9 The design
included three group conditions: a no inconsistency
control and two inconsistent group conditions. In the
two inconsistent group conditions, participants read
either about the inconsistent actions of a single group
member (single actor condition) or about multiple
actors (multiple actor condition). The orientation of
participants in this study was identical to that of the previous
studies. However, participants in the multiactor
inconsistent condition read about a nursing student who
slapped a child and a different nursing student who
taunted an old man. Participants in the single actor
inconsistent condition were presented these inconsistent
actions as performed by one nursing student. Participants
in both inconsistency conditions were asked
about the likelihood of a different nurse donating her
time, in the manner described in the previous studies.
Control participants were asked the same question but
were not exposed to inconsistent information.
RESULTS: EXPERIMENT 6
Participants’ rating scores were entered into a three-factor
between-subjects analysis of variance. The analysis
revealed a significant group main effect, F(2, 56) = 6.01,
p < .01. To interpret this effect, and to test our specific
predictions, we used the error term of the overall analysis
to make specific comparisons. Planned comparisons
indicated that participants in the inconsistent single
actor condition (M = 5.53) had more extreme attributions
that were consistent with the disposition of nurses
than those in the control condition (M= 4.24), F(1, 56) =
9.32, p < .01, whereas the attributions of participants in
the multiple actor condition (M = 4.26) did not differ
from that of control participants, F(1, 56) = .0001, ns.
DISCUSSION: EXPERIMENT 6
The results of Experiment 6 are entirely consistent
with our expectations. First, we found that participants
in the single group member actor condition had more
extreme attributions that were consistent with the internal
character of nurses than those in the control condition.
The novel finding in this study involved the ratings
of participants in the multiple group member actor’s
condition. The attributions of these participants did not
differ from those of controls, and they were significantly
less consistent with the internal character of the actors
than those of participants in the singular group member
actor condition. Thus, when participants were exposed
to the inconsistent actions of two different nurses, they
did not interpret a fellow group member’s behavior in a
compensatory manner. Therefore, just as participants in
Experiments 1 and 2 did not interpret the target’s behavior
in a compensatory manner following the target’s own
inconsistent actions, participants in Experiment 6 did
not interpret the actions of a fellow group member in a
compensatory manner after observing the inconsistent
actions of two different targets.
These data suggest that it may take a larger number of
inconsistent actions to constrain compensation in the
context of a fellow group member (e.g., it took the
inconsistent actions of two target persons) than to constrain
compensation in the context of the person performing
the inconsistent actions (e.g., it only took the
inconsistent actions of the target to constrain compensation
in Experiments 1 and 2). As discussed in the introduction,
there are several reasons why a perceiver can
expect compensation from a fellow group member but
not from the target of the inconsistent action. First, the
assumed variability among members is often greater
than the assumed variability in a single individual’s
behavior (e.g., Hamilton & Sherman, 1996). In addition,
the implications of inconsistent actions are more directly
tied to the personality of the target of the inconsistent
action than to the personality of a fellow group member.
Thus, it follows that it should take an objectively less
severe inconsistent action (or fewer inconsistent
actions) to constrain a perceiver’s capability of generating
compensatory expectations and attributions about
the acting target versus a fellow group member.
In addition, these results provide evidence that participants
are not comparing the valence of the target to a
prototype member of the group. If this interpretation
accounted for our effects, participants should have
viewed the behavior of a fellow group member as especially
consistent with the character of nurses after observing
the highly inconsistent actions of multiple targets.
This was not the case.
GENERAL DISCUSSION
Attributions play an important role in theoretical
accounts of social behavior, and attributional processes
have been the subject of considerable research (see
Fiske & Taylor, 1991, for a review). The results of the
present research add to this body of knowledge by demonstrating
an influence of compensatory motivation on
patterns of attribution. In particular, results show that
perceivers are biased toward relatively extreme stereotypical
internal attributions of group members’ behaviors
when previously exposed to behaviors that are
inconsistent with stereotypic expectations. These attributions
could potentially compensate for the inconsistency.
Thus, attributional processes were shown to be
biased by motives to maintain stereotype concepts.
The present research also complements existing
research on attributional mechanisms of stereotype
maintenance. For example, the type of attribution a
perceiver makes about an inconsistent action can influence
whether a person who behaves in an aberrant manner
is subtyped, that is, seen as an “exception to the rule”
(e.g., Crocker et al., 1983; Johnston & Hewstone, 1992;
Weber & Crocker, 1983). Subtyping and compensation
can work in tandem. It should be pointed out, however,
that whereas subtyping and compensation can work in
tandem, compensation does not require subtyping—as
has been shown, for example, by the finding that
perceivers can compensate via the target himself or herself
(J. J. Seta & Seta, 1993).
The results of the present studies demonstrate that
interpretations of the behavior of group members who
are related to targets only by virtue of category membership
also are influenced by the inconsistent behavior of
the deviant. Thus, this research demonstrates that the
aberrant behavior of a deviant group member not only
affects attributions about that person’s behavior but also
affects attributions about other members of the stereotyped
group. This finding has important implications for
the resistance of stereotypes to change, which will be
elaborated on further in this article. In brief, because the
assumed variability among members of a group is often
greater than the assumed variability in a single individual’s
behavior, the generation of compensation at the
group level is possible when it may not be possible at the
individual level. Furthermore, the implications of inconsistent
actions are tied more directly to the personality of
the target of the inconsistent actions than to the personality
of fellow group members. Thus, these processes
may constitute a strong force against stereotype change.
Study 5 demonstrated that judgments about individual
group members may not be identical to stereotypic
judgments about members of the group in general. In
this study, we used a stereotype measure that was conceptually
similar to the one used by Kunda and Oleson
(1997). From our model, individuals may exhibit
extreme compensatory judgments about members of a
group while maintaining their extant group stereotype.
That is, attributions and expectations concerning a particular
group member may reflect compensatory processes,
whereas judgments about the stereotyped group
per se remain equivalent to conditions in which no
inconsistency is observed. In addition to ruling out simple
contrast explanations for these findings, this result
suggests that compensatory processes do potentially
serve the function of maintaining, while not necessarily
strengthening, existing stereotypes.
The results of Experiment 6 demonstrated that inconsistent
actions on the part of several group members constrain
perceivers’ capability of compensating via different
members of stereotyped groups. In Experiment 6,
the ratings of participants in the multiple actor inconsistent
condition were not more compensatory than control
group participants, and they were less compensatory
than those of participants in the inconsistent single actor
condition.
This research also demonstrated that the generation
of compensatory attributions involves motivation. For
example, in Study 4, priming or framing participants
with arguments pro or con the utility of maintaining/
changing extant beliefs had the effect of influencing the
generation of compensatory attributions.
Contrast and Compensation
These experiments rule out various alternative interpretations.
They, for example, rule out a contrast interpretation
in which perceivers compare the target to a
prototypic member of the group. It should be pointed
out, however, that contrast (and assimilation) can be
used as descriptive terms. Contrast refers to the fact that
one stimulus moves away from another, whereas assimilation
refers to the fact that one stimulus moves toward
another. If contrast is used in a descriptive manner, then
a person may want to substitute the term motivated contrast
for compensation in our analysis while maintaining
our key assumptions. For example, perceivers, to maintain
their stereotype, can be viewed as contrasting the fellow
group member to the target. In this case, people are
motivated to move their evaluations and valuations of a
fellow group member away from the group mean to
maintain the stereotype. This will occur if perceivers are
motivated to maintain their stereotype and if they can
generate or access information that is compensatory in
nature. In this case, motivated contrast is being used in
place of compensation. It is being used to portray a process
by which perceivers compensate for one inconsistency
by altering another action. If contrast is used in this
way (i.e., by incorporating our key assumptions), then it
is not necessary to tease apart a contrast interpretation
(or other interpretations with similar meanings) from
our schema maintenance through compensation analysis.
We have chosen the word compensation (e.g., J. J. Seta
& Seta, 1992, 1993) instead of motivated contrast
because compensation implies that a person is motivated
to counterbalance an observed inconsistency.
Implications for the resistance of stereotypes and attitudes.
Not only do the results of our experiments support the
idea that perceivers compensated in an attempt to maintain
their stereotype, but the results of Experiment 5, in
particular, suggest that compensation may result in successful
stereotype maintenance. Following an inconsistent
action on the part of a group member, perceivers
evaluated a different behavior of a different group member
in a hyperconsistent manner. However, using a stereotype
measure that was conceptually similar to one
used by Kunda and Oleson (1997), we found that a separate
group of participants did not change their stereotype
following observed inconsistency. Assuming that
these participants also would have generated
compensatory attributions if assigned to the different
group member target condition (e.g., that there were no
failures of random assignment), this result supports the
idea that compensation can result in successful stereotype
maintenance. It also implies that compensation can
help prevent different types of schema from changing,
such as goals and attitudes.
After exposure to the inconsistent actions of a target
group member, perceivers can maintain their overall
group stereotype in a number of different ways. One way
is to modify their expectations and attributions about
how and why a different group member performed the
same behavior as the target; to balance the inconsistent
actions of a target, the different group member’s actions
are perceived to be hyperconsistent. In this way,
perceivers are maintaining the content as well as the
overall evaluative nature of their stereotype. The overall
evaluative nature of a nurse stereotype, for example, is
maintained if one component of the stereotype balances
a different component. Although the overall evaluative
nature of nurses may be maintained, exposure to inconsistency
and the generation of compensatory judgments
also might lead to changes in stereotype content or
variability.
Summary
The results of all six experiments provide support for
the key assumptions of our schema-maintenance
through compensation analysis. In addition, they add to
the body of knowledge concerning attributions in demonstrating
the role of compensatory motivation on patterns
of attributions.
Caveats
In this article, we have discussed situations in which
individuals were attempting to keep their view of a particular
group from changing. However, a schema can be
dynamic, such as when a person is interested in becoming
progressively better and better at a particular skill
(e.g., Wood, 1989). In this situation, because the goal is
improvement, we are attempting to maintain a situation
in which our outcomes and activities are progressively
more and more positive. So, for example, if we are
attempting to maintain the goal of increasing our net
worth each year, we will attempt to generate activities
that will satisfy this upwardly mobile goal.
Although research has not explicitly dealt with a
schema of the sort we have just discussed, it has dealt with
situations in which the schema was related to the activities
of the perceiver rather than to the activities of other
individuals (e.g., J. J. Seta et al., 1993; J. J. Seta & Seta,
1982). Therefore, our analysis is applicable to situations
in which the perceiver is the actor as well as the observer.
NOTES
1. There are circumstances in which schema change is desirable.
For example, at times we have a goal to self-improve (e.g.,Wood, 1989).
In these situations, because the goal is improvement, we are attempting
to maintain change rather than the status quo.We will discuss situations
like this in the General Discussion.
2. It should be pointed out that our model makes relative and not
absolute predictions. For example, we cannot predict the exact level of
an inconsistent event (or the exact number of inconsistent events) that
will constrain perceivers’ capability of compensating.We can, however,
make relative predictions about the conditions that make it more or
less likely to compensate.
3. The data of 7 participants were not included because they did not
properly identify the person they were evaluating. The materials presented
in this and subsequent studies were complex, and if participants
did not pay careful attention they misread or confused the information.
4. Six participants indicated that they majored in nursing and were
excluded from the study. Also, 6 people were excluded from our study
for giving the incorrect answer on our probe question.
5. Perceivers may have believed that the actions of the inconsistently
behaving nurse who was asked to volunteer were known to the
other fellow group members. If the actions of the inconsistently behaving
nurse were known, the fellow group member might be especially
inclined to want to impress (e.g., have self-presentational motives). To
rule out this possibility, we gave the scenarios used in this research to a
separate group of participants. These participants (n = 18) were asked
whether the fellow group member nurse would want to impress a
group of other nurses—after they witnessed the inconsistent target’s
behavior—in contrast to a situation in which the information about the
inconsistently behaving nurse was not known.Ona 5-point scale where
1 represented wants to impress most when information was known, 5 represented
wants to impress most when information was not known, and 3 represented
no difference, results revealed no difference in estimates of “volunteering
to impress” across these two situations (M = 3.2, where 3
represented no difference), F(1, 17) = .25, ns. Thus, explanations based
on the idea that the manipulation either increased or decreased
assumed self-presentational motives are not viable.
6. Eight participants were excluded from the study for giving the
incorrect answer on our probe question, which was conceptually similar
to the one used in Experiment 1.
7. Twelve people were excluded from our study for giving the incorrect
answer on our probe question, which was conceptually similar to
the one used in Experiment 2.
8. Methodological changes incorporated into Study 4, including
group salience and name changes, most probably accounted for discrepancies
between the control groups in this study and the controls in
our previous nursing studies.
9. Four participants indicated that they majored in nursing and
were excluded from the study. Also, 14 participants gave an incongruent
answer for the question asking them about the person that they
evaluated and were excluded from the study.
REFERENCES
Allport, G. W. (1954). The nature of prejudice. Reading, MA: Addison
Wesley.
Crocker, J., Hannah, D. B., & Weber, R. (1983). Person memory and
causal attributions. Journal of Personality and Social Psychology, 44,
55-66.
Fiske, S. T., & Neuberg, S. L. (1990). A continuum of impression formation,
from category-based to individuating process: 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.
Fiske, S. T., & Taylor, S. E. (1991). Social cognition (2nd ed.). New York:
McGraw-Hill.
Hamilton, D. L., & Sherman, S. J. (1996). Perceiving persons and
groups. Psychological Review, 103, 336-355.
Heider, F. (1958). The psychology of interpersonal relations. New York:
John Wiley.
Hewstone, M. (1989). Causal attribution. Cambridge, MA: Basil
Blackwell.
Johnston, L., Bristow, M., & Love, N. (2000). An investigation of the
link between attributional judgments and stereotype-based judgments.
European Journal of Social Psychology, 30, 551-568.
Johnston, L., & Hewstone, M. (1992). Cognitive models of stereotype
change. Subtyping and the perceived typicality of disconfirming
group members. Journal of Experimental Social Psychology, 28, 360-
386.
Jones, E. E., & Davis, K. E. (1965). From acts to dispositions: The attribution
process in person perception. In L. Berkowitz (Ed.),
Advances in experimental social psychology (Vol. 2, pp. 220-266). New
York: Academic Press.
Kelley, H. H. (1973). The processes of causal attribution. American Psychologist,
28, 107-128.
Kunda, Z., & Oleson, K. C. (1997). When exceptions prove the rule.
How extremity of deviance determines the impact of deviant
examples on stereotypes. Journal of Personality and Social Psychology,
72, 965-979.
Martin, L. L., & Seta, J. J. (1983). Perceptions of unity and distinctiveness
as determinants of attraction. Journal of Personality and Social
Psychology, 44, 755-764.
Moreno, K., & Bodenhausen, G. (1999). Resisting stereotype change:
The role of motivation and attentional capacity in defending
social beliefs. Group Processes and Intergroup Relations, 2, 5-16.
Seta, J. J.,&Seta, C. E. (1982). Personal equity: An intrapersonal comparator
system analysis of reward value. Journal of Personality and
Social Psychology, 43, 222-235.
Seta, J. J.,&Seta, C. E. (1992). Personal equity-comparison theory: An
analysis of value and the generation of compensatory and
noncompensatory expectancies. Basic and Applied Social Psychology,
13, 47-66.
Seta, J. J.,&Seta, C. E. (1993). Stereotypes and the generation of compensatory
and noncompensatory expectancies of group members.
Personality and Social Psychology Bulletin, 19, 722-731.
Seta, J. J., Seta, C. E., & Erber, M.W. (1993). The role of cost in generating
expectations and value: A personal equity comparison theory
analysis. Basic and Applied Social Psychology, 14, 103-119.
Seta, C. E., Seta, J. J.,&Goodman, R.C. (1998). Social identity orientation
and the generation of compensatory expectancies. Basic and
Applied Social Psychology, 20, 280-295.
Stapel, D. A., & Spears, R. (1996). Event accessibility and context
effects in causal inference: Judgment of a different order. Personality
and Social Psychology Bulletin, 22(10), 979-992.
Weber, R., & Crocker, J. (1983). Cognitive processes in the revision of
stereotypic beliefs. Journal of Personality and Social Psychology, 45,
961-977.
Wilder, D. A., Simon, A. F., & Faith, M. (1996). Enhancing the impact
of counterstereotype information: Dispositional attributions for
deviance. Journal of Personality and Social Psychology, 71, 276-287.
Wilson,T. D., Lisle, D. J., Kraft, D.,&Wetzel, C. G. (1989). Preferences
as expectation-driven inferences: Effects of affective expectations
on affective experiences. Journal of Personality and Social Psychology,
56, 519-530.
Wood, J. V. (1989). Theory and research concerning social comparisons
of personal attributes. Psychological Bulletin, 106, 231-248.
Yzerbyt, V. Y., Coull, A., & Rocher, S. (1999). The role of cognitive
resources in the maintenance of stereotypes. Journal of Personality
and Social Psychology, 77, 449-462.
Yzerbyt, V. Y., Rogier, A., & Fiske, S. (1998). Groups entitativity and
social attributions: On translating situational constraints into stereotypes.
Personality and Social Psychology Bulletin, 24, 1089-1103.
