Vous êtes sur la page 1sur 16

Journal of Personality and Social Psychology

1999, Vol. 77. No. 1, 78-93


Copyright 1999 by the American Psychological Association, Inc.
0022-3514/99/53.00
Images and Affect: A Functional Analysis of Out-Group Stereotypes
Michele G. Alexander, Marilynn B. Brewer, and Richard K. Herrmann
The Ohio State University
Drawing from research on inter-nation images, the authors proposed and tested a functional theory of
out-group stereotypes in 3 experiments. In the theory, it is hypothesized that behavioral orientations
elicited by specific patterns of intergroup relationships (goal compatibility, relative power, and relative
status) give rise to unique schematic representations of an out-group. The representations specified in the
theory include 1 positive image (i.e., ally) and 3 differentiated negative images (i.e., enemy, dependent,
barbarian). In all 3 experiments, participants read and imagined scenarios describing an intergroup
situation in which the structure of relationships between in-group and out-group was varied. Results from
Experiments 1 and 2 indicated that relationship cues were sufficient to elicit the predicted images, and
some of the images were more likely to be activated under high incidental arousal. In a 3rd experiment
an implicit measure was used to demonstrate that the images are activated spontaneously. Overall, results
implicate the role of affective state and behavioral intent in shaping the content of social stereotypes.
With the rise of social cognition as a dominant perspective in
social psychology, the focus of research on social stereotypes and
stereotyping shifted from the study of the content of social stereo-
types embedded in intergroup relations to the study of structure
and process. Questions about why groups are stereotyped in certain
ways gave way to questions about how stereotypes are formed and
what role they play in processing social information. Although
earlier theories of intergroup perception explicitly acknowledged
the functions that stereotypes serve in maintaining and justifying
the structural relationships between groups (e.g., Campbell, 1967;
M. B. Smith, Bruner, & White, 1956; Tajfel, 1981), this functional
perspective was, for the most part, ignored in the social psycho-
logical literature on social stereotypes in the 1970s and 1980s.
More recently, there has been evidence of renewed interest in
understanding the function that group stereotypes serve in specific
intergroup contexts. Questions are again being raised about the
content of category stereotypes: How is content shaped by the
status and roles assigned to groups in a society, and how do
stereotypes function to maintain and justify the structure of inter-
group relations and intergroup behavior (e.g., Eagly, 1987; Fiske,
1998; Jost & Banaji, 1994)? As Bouris, Turner, and Gagnon
(1997) put it, "Stereotypes function to represent intergroup reali-
ties . . . creating images of the out-group (and the in-group) that
explain, rationalize and justify the intergroup relationship and
one's past, present and future behaviour within it" (p. 273). From
Michele G. Alexander and Marilynn B. Brewer, Department of Psychol-
ogy, The Ohio State University; Richard K. Herrmann, Department of
Political Science, The Ohio State University.
We thank Jane Baki, Elaine Bednar, Heidi Parsons, Cindy Plew, and
Toby Pollock for their invaluable assistance with data collection and M.
Atiq Rahman for his advice on the statistical analyses in portions of this
article.
Correspondence concerning this article should be addressed to Michele
G. Alexander, who is now at the Psychology Department, 5742 Little Hall,
University of Maine, Orono, Maine 04469-5742. Electronic mail may be
sent to michelea@maine.edu.
this perspective, understanding group cognition requires under-
standing the context of intergroup relationships from which beliefs
and stereotypes are derived.
Image Theory in International Relations
Social psychological interest in the origins and functions of
group stereotypes has its parallel in the international relations field
of political science, where scholars have been studying the origins
and consequences of the images that nation-states hold of each
other as political entities, particularly in the context of interna-
tional conflict. Boulding (1956, 1959) defined an image in this
context as a cognitive, affective, and evaluative structure and
argued that "the images which are important in international sys-
tems are those which a nation has of itself and of those other
bodies in the system which constitute its international environ-
ment" (1959, pp. 120-121). Boulding felt that perceived hostility
or friendliness and the perceived strength or weakness of a unit
were central features of a nation's image of that unit. In subsequent
work, particular attention was paid to the enemy images that form
when the other unit is perceived to be very hostile and strong (R.
Cottam, 1977; Holsti, 1967; Shimko, 1991; Silverstein, 1989;
White, 1965, 1968).
Enemy images have been treated in political science in terms
analogous to how psychologists conceive of a stereotype. Features
of the enemy image include a very simple picture of the other
group's motivations in both substantive and normative terms. The
enemy is seen as motivated by a very few self-serving interests, all
of which are judged to be evil and immoral. The enemy image also
features a characterization of the adversary as being conspiratorial
and led by a monolithic and hierarchical decision-making system
(Jervis, 1976). The enemy image in its ideal typical form also
describes the adversary as a "paper tiger," meaning that it aims to
intimidate others with its threats and noises, but in fact it is quite
hollow and weak. The stereotype describes the enemy as pursuing
its evil ambitions in an opportunistic way, advancing when it
senses weakness on other people's part and retreating quickly
78
IMAGES AND AFFECT 79
when it senses firm will and resolve on the part of others (R.
Cottam, 1977; Jervis & Snyder, 1991).
International relations scholars have not only identified the
attributes of an enemy image, they have also demonstrated its
effects on the perception and interpretation of new information,
memory, and policy choice (Herrmann, Voss, Schooler, & Ciar-
rochi, 1997). A number of studies have shown that the stereotype
is immune to empirical disconfirmation (Stuart & Starr, 1982).
Aggressive acts on the part of the adversary fulfill expectations,
whereas nonaggressive acts are attributed to situational constraints
or conspiratorial tricks. In either case, no evidence falsifies the
basic image. When two adversaries have mirror enemy images of
each other, a spiral escalation in tension is expected and is not
easily defused (Bronfenbrenner, 1961; Osgood, 1962).
Although the enemy image and spiral models of conflict have
been the subjects of a substantial body of research, international
relations researchers have explored other images as well. White
(1968) paired the enemy image with a virile self-image. R. Cottam
(1977) identified an ally image and more complex imperialist and
colonialist images. Herrmann (1985; Herrmann & Fischerkeller,
1995) extended the analysis by examining degenerate and barbar-
ian images, and M. L. Cottam (1986, 1994) identified and refined
a set of dependent images. As additional images were cataloged,
component parts of images were organized into sets of attributes,
such as attributes describing the target country's motivations,
attributes describing the decision-making process in the target
country, and attributes describing the willpower and abilities of the
target country. Herrmann et al. (1997) demonstrated experimen-
tally that these component attributes were connected to each other
in systematic ways, and that enemy, ally, colony, and degenerate
images had schematic properties. When research participants were
given information about a hypothetical country that contained one
attribute of an image, they assumed that the country had additional
attributes consistent with the same image. For instance, when
given information about the motives of a foreign country that
conformed to the enemy image, participants who were then asked
about the country's decision process and power structure selected
attributes consistent with the enemy image.
Going beyond the mere description of images as an organized
stereotype, the theory of images advanced by R. Cottam (1977)
and Herrmann (1985) argued that images form as a consequence of
strategic relationships between nations and serve a functional
purpose. Working from Heider's (1958) balance theory, Herrmann
(1985) suggested that images in international relations serve to
balance a positive self-image and behavioral inclinations. Herr-
mann argued that the nature of relationships between nations is
represented in terms of perceived threats and perceived opportu-
nities that are analogous to Heider's notion of sentiment. Images
are a set of unit relations (cognitive attributes) that are attached to
a foreign country. The cognitive system is balanced when the
evaluative and behavioral implications of the image match the
behavioral inclination generated from the threat or opportunity in
a way that maintains consistency with a positive, moral self-image
(see also Sande, Goethals, Ferrari, & Worth, 1989).
Herrmann (1985) identified three critical dimensions of inter-
group relationships that give rise to sentiments and associated
images. Like Boulding, Herrmann assigned particular importance
to (a) evaluations of competitive versus cooperative goal interde-
pendence and (b) assessments of relative power and added (c)
evaluations of relative cultural status as the central determinants of
strategic relationships. When relationships are defined by extreme
values on these three dimensions, attributions about the other
nation's motivations, decision-making process, willpower, and
abilities follow from the need for psychological balance or behav-
ioral justification.
According to image theory, then, the structure of intergroup
relationships (cooperation-competition, relative strength, and rel-
ative status) gives rise to sentiments and behavioral inclinations
that must be balanced with the in-group's positive and moral
self-image. Any tension between the behavioral inclination and the
restraints of a moral self-image elicits a balancing process that
results in a construction of the situation in which the tension is
resolved by the in-group providing a morally acceptable account
for acting in line with the behavioral inclination. Stereotypes or
images of the out-group are constructed to serve this balancing
function.
In this theoretical account, the enemy image arises when the
relationship between a person's in-group and an out-group is
characterized by intense competition between two groups that are
similar in cultural status and in power or strength. These relation-
ship characteristics generate a sentiment of threat and a behavioral
inclination to eliminate the threat by attack, which is in tension
with normative rules about behavior toward similar others and by
prospects of retaliation. The enemy stereotype balances the tension
between the behavioral inclination and the moral constraints. It
portrays the adversary as being so hostile in its motivational
attributes that destroying it, even through the use of force, is
justified. To protect self-image from qualms about mistreating
innocent.members of the out-group, the in-group will adopt an
enemy image that describes the out-group as monolithic in its
decision making. The image describes the adversary's strength as
essentially derivative of the in-group's weakness. In other words,
it constructs a picture of the adversary in which the adversary is
described as a paper tiger only capable of mischief if the in-group
is timid and lacks resolve. The result of the image is a construction
in which attacking the enemy is the instrumentally reasonable
course of action and a moral duty.
If the enemy image facilitates behavior in a threatening situation
with a coequal other, then the ally image serves a comparable
function in facilitating the pursuit of cooperation. In a relationship
defined by positive goal interdependence, similar capabilities, and
equal status, there is a behavioral inclination to cooperate. This is
expected to produce an image of the partner that is characterized
by positive attributes. The partner's motives will be seen as benign
and, in the ideal form, mostly altruistic. The out-group will also be
seen as enjoying the benefits of gifted and popular leaders. The
stereotypical image eliminates reservations that might inhibit full
cooperation or cast any moral doubt on the partnership. By de-
scribing the out-group in entirely positive terms, the ally image,
like its enemy image counterpart, constructs a picture in which
acting on the behavioral inclination induced by the relationship is
not just acceptable but a moral duty.
The enemy and ally images are symmetric; as long as the two
groups involved perceive the intergroup relationships in the same
way, their mutual stereotypes should be mirror images of each
other (Bronfenbrenner, 1961). Two other relationships that have
received substantial attention in image theory are those generated
by mutually incompatible goal interdependence between groups
80 ALEXANDER, BREWER, AND HERRMANN
that differ in power and status. The first of these arises when the
in-group nation is much stronger than the target out-group nation
and also has higher cultural status. In this relationship in-group
members have an opportunity to eliminate the goal incompatibility
by exploiting the asymmetries in their favor. Thus, this pattern of
intergroup relationships generates a behavioral inclination to seize
the opportunity, assert control over the target, and exploit the
situation to maximize the likelihood of achieving the in-group's
own goals. Exploitation, however, is typically constrained by
moral norms that do not equate might with right, in a political
setting where such exploitation is seen as imperialism. This tension
generates a cognitive representation of the out-group that permits
exploitative behavior to be reinterpreted as helping.
In the resulting dependent image, the target group is described
as being deeply divided between moderate, responsible, but in-
competent good leaders, who are motivated by their concern with
advancing the welfare of the people, and radical, agitating bad
leaders, who are uninterested in the welfare of their own people.
For in-group nations, this image of a divided and vulnerable nation
justifies using asymmetries in power and status to take control of
the target out-group; in fact, it can now be construed as a moral
duty. Taking control is seen not as exploiting the group but as
protecting it from both the evil internal radicals and their evil
external patrons. In the 19th century, metaphors like "white man's
burden" captured the essence of the stereotype as it was applied to
Asians and Africans by Europeans. Like the other images, it
allowed people to seize the opportunity and act on the behavioral
inclination without creating doubts about self-image. In extreme
cases, the interlopers expect gratitude and thanks from the target
group. They are mystified and offended by complaints and resis-
tance from the target group, especially if these complaints suggest
the in-group is simply advancing its own self-interest.
The second asymmetric relationship image theorists have exam-
ined is one in which the in-group country enjoys high cultural
status but is vastly weaker than the out-group country and has its
goals severely threatened by the out-group's agenda. This relation-
ship pattern is said to produce a behavioral inclination that in-
volves insulating the in-group by appeasing the stronger out-group.
By appeasing the out-group in the short run, the in-group hopes
that its essence can be preserved until some future time when the
balance of capabilities changes and a viable chance for effective
resistance exists. Appeasing the out-group can produce serious
moral tension when it involves sacrificing important goals and
values of the in-group. Passivity in this situation would threaten a
positive self-image if it was understood to be cowardly and un-
principled behavior.
The barbarian image relieves the moral tension by constructing
a picture in which restraint is not only the prudent course but also
the normatively correct choice. The image does this by portraying
the out-group as evil and as enjoying wanton destruction for its
own sake. The out-group is also seen as incredibly strong, immune
to rational deterrents because its leaders enjoy destruction, willing
to endure horrific losses, and unable to process costs and benefits
in a rational way. The barbarians are led by highly emotional and
not terribly intelligent leaders who will destroy anything at the
slightest provocation, but they can be manipulated by cunning and
intelligence. A barbarian image leads one to conclude that the both
moral and prudent course of action is to avoid provoking or
antagonizing the out-group. By using the in-group's superior in-
telligence and avoiding what would be a futile test of material
strength, the in-group in this construction can best ensure its own
future. Appeasement can then be seen as the best strategy to
advance the future of the group and to protect its values rather than
as a capitulation and cowardly abandonment of principle.
A Generalized Model of Out-Group Images
Many of the components of image theory have counterparts in
social psychological research and theory. The idea that perceptions
of out-groups are shaped by the structural relationships between
in-group interests and out-group interests is essential to functional
theories of intergroup relations (Sherif & Sherif, 1953). Sherifs
classic Robbers Cave summer camp experiments (Sherif, Harvey,
White, Hood, & Sherif, 1961) provided an empirical test of the
notion that goal compatibility determines intergroup attitudes and
perceptions. Under competitive conditions, groups developed de-
rogatory stereotypes of the out-group that changed in both content
and valence when common goals were introduced. In more recent
work, Insko and Schopler (1987) hypothesized that the competi-
tiveness characterizing intergroup encounters gives rise to a ge-
neric "group schema," in which "own-group members are ex-
pected to be cooperative, friendly, loyal, helpful, courteous, and so
on, while other-group members are expected to demonstrate such
characteristics as competitiveness, unfriendliness, aggressiveness,
and boastfulness" (p. 244). In addition to goal compatibility, status
and power relations between groups are also implicated in recent
theories of stereotype formation. Fiske (1998; Fiske, Xu, Cuddy, &
Glick, in press), for instance, speculated that interdependence and
status differentials give rise to two different types of ambivalent
out-group stereotypes: the respected-but-disliked out-group and
the liked-but-disrespected (incompetent) out-group. Fiske's con-
cepts are similar to image theory in making distinctions between
specific negative schemas that are elicited by different patterns of
intergroup relationships.
International relations image theory, then, provides a basis for a
more general theory of the functional origins of out-group stereo-
types. At a somewhat more abstract level, the components of the
ally, enemy, dependent, and barbarian images may constitute ge-
neric schema that are evoked by particular intergroup contexts or
patterns of intergroup relationships. More specifically, a general-
ized image theory leads to the following hypotheses:
1. In-group-out-group relationships characterized by perceived
equal status, similar capacity, and compatible goals that give rise
to inclinations for cooperative intergroup behavior will produce a
cognitive representation of the out-group that parallels the ally
image in the political arena (peaceful, trustworthy, democratically
organized, and intelligent).
2. In-group-out-group relationships characterized by equal sta-
tus, equal capacity, and incompatible goals with behavioral incli-
nations to attack the out-group will evoke a cognitive representa-
tion of the out-group parallel to the enemy image (hostile,
untrustworthy, monolithic, and opportunistic).
3. Intergroup relationships characterized by the incompatible
goals of an in-group and a weaker, lower status out-group and
associated with exploitative behavioral inclinations will elicit the
dependent image of the out-group (childish, incompetent, and
divided).
4. Relationships in which the out-group is stronger than the
IMAGES AND AFFECT 81
in-group but lower in status and is perceived to have incompatible
goals that evoke defensive appeasement will give rise to the
barbarian image of the out-group (ruthless, evil, and irrational).
To test these hypotheses, we conducted three experiments
wherein participants were presented with scenarios describing an
intergroup situation in which information about the pattern of
status, power, and goal relationships between the groups was
varied. The question to be addressed in these studies was whether
these relationship cues alone are sufficient to elicit detailed images
of an out-group's characteristics that correspond to the hypothe-
sized generic schemas.
Experiment 1
To assess the generic images associated with specific intergroup
relations, we gave research participants a realistic scenario that
included information about an in-group to which they belonged
and the in-group's relationships with a particular out-group. Par-
ticipants read the scenario while imagining themselves being the
leader of a group of university students who are attempting to
secure the future of their financially threatened campus. Informa-
tion about goal interdependence, relative capability, and relative
status of the out-group defined four different patterns of intergroup
relationships. After reading one of the scenarios, participants re-
sponded to a questionnaire that first assessed their strategic re-
sponses to the out-group and then measured their beliefs about the
out-group's motivational and organizational characteristics.
Method
Participants
A total of 93 (63 female and 30 male) undergraduate students from The
Ohio State University-Mansfield participated in the study as part of a
requirement in their introductory psychology course.
Procedure
Participants believed that the experiment was designed to assess person
perception and cognition. Individuals were run in groups of approxi-
mately 20, and each participant was randomly assigned one of the four
intergroup relationship scenarios as materials were distributed. Each indi-
vidual participant was given a booklet, with instructions to read the
intergroup scenario on the first page and imagine being involved in the
situation presented. After reading the scenario, participants completed
questionnaires assessing their perceptions of the out-group on a series of
measures designed to determine whether the pattern of intergroup relation-
ships evoked the intended out-group images. Participants were then de-
briefed and excused.
Materials and Measures
Construction of intergroup scenarios. A general intergroup scenario
was developed as a basis from which to manipulate the specific intergroup
schemas tested in the present study. The participants were students attend-
ing a branch campus of a large university who were accustomed to sharing
a campus and campus resources (e.g., library, bookstore, parking lot) with
students from a neighboring community technical college. Given this
arrangement, students attending the technical college composed a salient
and meaningful out-group to the students at the branch campus who
participated in this study.
Included in the general scenario was a brief description of the respon-
dent's own in-group (i.e., students attending a regional campus of a large
state university seeking 4-year bachelor's degrees) and a description of the
targeted out-group (i.e.. students attending the community technical col-
lege). All participants read that the future of their campus and its resources
was uncertain because of reductions in state funding. As the leader of his
or her in-group, the participant's goal was to acquire funding to ensure the
campus continued to provide educational opportunities for those seeking
4-year degrees in the region and to obtain more money and resources for
the branch campus. The participant's in-group was given some leverage to
reach its goal. Specifically, the in-group knew some members of the board
of regents, and the group had the campus dean's support. The students at
the neighboring community technical college, however, were also con-
cerned and were working on securing funding for their own college.
The general scenario was then tailored with specific information de-
signed to manipulate the four patterns of intergroup relationships being
studied. Information about the out-group was varied to generate different
configurations of relative strength, prestige, and goal compatibility asso-
ciated with the four different behavioral orientations as specified by image
theory. The specific manipulations corresponding to each intended image
are presented in Appendix A.
Response strategies. The extent to which the scenario information
elicited the intended behavioral orientation was assessed by one multiple-
choice item. Participants selected one among four alternative response
strategies that their group could use in dealing with the out-group. The four
strategies were intended to represent the four different behavioral inclina-
tions specified by image theory (cooperation, attack, exploitation, OT de-
fensive appeasement) and were adapted to conform with the specific
context presented in the scenario. The options provided were as follows:
1. Sharing their resources and cooperating with the out-group
(cooperation/ally)
2. Competing with the out-group and strengthening their own position
while preventing the out-group from achieving its goal (attack/enemy)
3. Using the out-group's resources and information to help their own
in-group achieve its goals (exploitation/dependent)
4. Defending their own position and resources by compromising with
the out-group (appeasement/barbarian)
Image measures. For each hypothesized image, five descriptive state-
ments were generated that instantiated different component aspects of that
image. These components were those that according to international rela-
tions image theory differentiated the images in terms of perception of
out-group goals, motivations, intentions, leadership, and decision making.
(See Appendix B for the component descriptions associated with each
out-group image.) Image components were then decomposed and distrib-
uted as response options to five multiple-choice questions about charac-
terizations of the out-group. Four responses were provided (in randomized
order) for each of the five questions, with each response representing an
attribute of one of the four images. For each question, participants were
asked to select the one description among the four options that came closest
to reflecting their perception of or expectations about the out-group. For
each respondent, we computed four image scores based on the number of
ally, enemy, barbarian, and dependent descriptors the respondent chose
across the five questions.
At the end of the image questionnaire, a composite image measure was
added. The five components of each image were clustered together into
separate sets, or image categories. Participants were asked to select the one
of these four descriptive categories that best represented their overall
impression of the out-group.
Manipulation checks. To assess the effectiveness of the intergroup
relationship manipulation, we had participants rate the degree of goal
interdependence between their group and the out-group and the relative
capability and status of the out-group on 5-point rating scales (with scale
endpoints labeled not at all and extremely). Participants rated the extent to
which the out-group's goals threatened their own goals, the extent to which
in-group and out-group goals were compatible, how strong the out-group
82 ALEXANDER. BREWER, AND HERRMANN
Table 1
Means and Standard Deviations for Manipulation Checks by Scenario: Experiment I
Measure
Perceived threat
Perceived compatibility
Comparative strength
Comparative status
(n
M
1.88
3.72
2.88
4.00
Ally
= 25)
SD
1.24
1.10
0.60
1.24
Scenario
Enemy
in = 24)
M
3.92
2.50
2.92
3.72
SD
1.10
1.28
0.71
1.20
Dependent
(n = 21)
M
2.76
2.86
2.62
2.60
SD
0.99
1.23
0.86
1.02
Barbarian
(n = 23)
M
3.91
2.39
3.78
2.57
SD
0.95
1.23
1.04
0.78
was compared with their in-group, and how similar the out-group was to
the in-group. Also, to determine the extent to which the hypothetical
scenario was self-relevant to participants, we had them rate how strongly
they identified with their role in the story, how important The Ohio State
UniversityMansfield is to them, and how bad they would feel if the
campus were closed down.
Results
Manipulation Checks
Mean ratings of the scenarios were analyzed in one-way anal-
yses of variance (ANOVAs) to determine whether the four sce-
narios differed significantly on the expected dimensions. A signif-
icant main effect of scenario was found on the measures of
perceived threat, F(3, 92) = 20.18, p = .0001; perceived compat-
ibility, F(3, 92) = 6.02, p = .001; comparative strength, F(3,
92) = 8.70, p = .0001; and comparative status, F(3, 92) = 5.89,
p = .001. Means and standard deviations of these measures for
each scenario are represented in Table 1. Overall, the differences
indicated that participants correctly interpreted the experimental
variations as intended. No differences across conditions emerged
on the extent to which participants identified with the scenario,
considered the campus important to them, and would be upset if
the campus they were currently attending closed.
Response Strategies
We conducted a chi-square test to determine whether each
scenario elicited the intended behavioral orientation as indicated
by participants' response strategy selections. The overall 4 (sce-
nario) X 4 (response strategy) association was significant, )f{9,
N =93) = 53.59, p = .00001, indicating that response choices did
differ as a function of scenario content. The proportion of partic-
ipants who selected each response strategy by scenario is presented
in Table 2.
By comparing the proportions that fall in the diagonals in
Table 2 with those in the respective rows for each response option,
we found it clear that although the base rate of selecting the
various responses differed, the probability of selecting a particular
behavioral response was consistently highest in the corresponding
scenario condition. Although the cooperative response strategy
was chosen to some extent across all scenarios, the proportion in
the ally condition differed significantly from the enemy, depen-
dent, and barbarian conditions, ^ ( 3 , TV = 93) = 15.76, p =
.00007; 9.24, p = .002; and 13.41, p = .0003, respectively.
Using a separate 3 X 4 chi-square we determined that the three
negative scenarios produced significantly different patterns of re-
sponse strategy selection, ^ ( 6, N = 6$) = 29.18, p = .001. The
attack response was most frequently elicited by the enemy scenario
and the appeasement response by the barbarian scenario. However,
in the case of the dependent scenario participants showed a pref-
erence for the cooperative (ally) response compared with the
exploitative (dependent) strategy, possibly reflecting the low threat
posed by the low strength and low status out-group and the social
undesirability of the dependent response strategy (which had the
lowest overall probability of selection).
Image Scores
As an initial test of the coherence of the four images, statistical
analyses were conducted to determine whether the probability of
selecting multiple items from the same image (across the five
image questions) was greater than chance. Image scores, ranging
from 0 to 5, were computed for each participant for all four
images. Then, the frequency distribution of these obtained image
scores was compared with the expected values for each image
derived from a binomial distribution (where p = overall proportion
of choices of responses of that image type).
1
Chi-squares were computed for the difference between obtained
and expected frequency distributions for each image. Response
patterns for the ally, ^ ( 5 , N = 93) = 185.24, p = .001; enemy,
5, N = 93) = 17.58, p = .005; and dependent image scores,
5, N = 93) = 19.90, p = .005, all differed significantly from
chance, indicating that the proportion of scores of 0 or 4-5 were
higher than expected on the basis of random choice. The chi-
square for the barbarian image scores did not differ significantly
from chance, ^ ( 5 , N = 93) = 8.96, p = .20. Thus, for at least
three of the four images, the selection of attributes reflected
schematic connections among the component features of the
image.
Scenario X Image. To test whether the image scores varied as
a function of the intergroup scenario, we undertook an analysis
appropriate to the ipsative nature of the multiple-choice scores. An
ANOVA involving one between- (ally scenario vs. enemy scenario
vs. dependent scenario vs. barbarian scenario) and one within-
1
These base-rate proportions were .29 for components of the ally image,
.32 for the enemy image, .27 for the dependent image, and .12 for the
barbarian image.
IMAGES AND AFFECT 83
Table 2
Proportion of Response Strategy as a Function of Scenario:
Experiment I
Response strategy
Ally
Enemy
Dependent
Barbarian
Ally
.88
.04
.00
.08
Enemy
.25
.42
.08
.25
Scenario
Dependent
.62
.00
.33
.05
Barbarian
.30
.17
.04
.49
Note. Values in boldface indicate the proportion of correctly matched
response strategies selected for each scenario.
(ally score vs. enemy score vs. dependent score vs. barbarian
score) subjects factor was first conducted, with score nested in
scenario. A significant overall interaction emerged, F(9,
92) = 5.87, p = .0001, indicating that out-group scores were
affected by the intergroup manipulations represented in the differ-
ent scenarios. To determine whether the image patterns were
distinct for the three negative scenarios, we conducted a separate
3 X 4 mixed ANOVA on the image scores with the ally scenario
removed. Results yielded a significant two-way interaction, F(6,
68) = 4.21, p = .0005.
Table 3 reports the mean image component scores by scenario.
One-way ANOVAs were conducted on each image score by
scenario, using the interaction error term, to determine if the mean
score varied by scenario in the predicted pattern. The ANOVA
conducted on the ally image scores revealed a main effect for
scenario, F(3, 92) = 4.51, p = .006. Planned comparisons indi-
cated that ally scores were greatest for those participants exposed
to an ally scenario, compared with those in the enemy condition,
F(3, 92) = 6.62, p = .001; the dependent condition, F(3,
92) = 13.39, p = .0001; or the barbarian condition, F(3,
92) = 11.54,p = .0001.
The ANOVA conducted on the enemy image scores also
yielded a main effect for scenario, F(3, 92) = 3.68, p = .02.
Simple effects tests indicated that those exposed to the enemy
scenario had significantly higher scores than those exposed to the
ally, F(3, 92) = 6.77, p = .001, or the dependent out-group, F(3,
92) = 5.20, p = .005, but not the barbarian out-group, F(3,
92) = 2.34, p = .10.
The analysis of the dependent image scores revealed a signifi-
cant main effect for scenario, F(3, 92) = 5.66, p = .001, as well,
reflecting the finding that dependent scores were greater for the
dependent condition than for the ally, F(3, 92) = 7.00, p = .001;
enemy, F(3, 92) = 9.93, p = .001; and barbarian conditions, F(3,
92) = 5.56, p = .005.
Finally, analyses on the barbarian image scores also yielded a
significant main effect for scenario, F(3, 92) = 3.24, p = .03, with
participants in the barbarian condition showing significantly
higher barbarian scores than those in the ally, F(3, 92) = 3.65, p =
.025, or enemy conditions, F(3, 92) = 2.02, p = .05, and margin-
ally higher than those in the dependent condition, F(3, 92) = 1.91,
p = .10.
Composite images. A second test of the relationship between
intergroup scenario content and out-group images was provided
through the analysis of participants' selections of overall, summary
descriptions of the out-group at the end of the image questionnaire.
A 4 X 4 chi-square test revealed that the predicted association
between scenario and image category was significant, x*(9, N =
93) = 39.80, p = .0001. A 3 X 4 chi-square was conducted
separately for negative scenarios only to determine whether the
three differed significantly in participants' selection of a composite
image category, ^ ( 6, N = 68) = 12.12, p = .05.
Proportions of category choices by scenario condition are pre-
sented in Table 4. By examining the table, we found it clear that
participants were most likely to select the appropriate ally, enemy,
and dependent image categories when given the respective inter-
group relationship manipulation. The barbarian scenario, however,
produced no coherent pattern of composite image selection. Al-
though participants in the barbarian condition were more likely to
choose some barbarian component responses in the multiple-
choice items (as indicated by image score analyses), these re-
sponses did not cohere as well as components did for the other
three images, and respondents did not appear to accept the barbar-
ian image as a whole.
Correlations Among Measures
To examine the correspondence between the different measures
of image activation, we first computed a Cohen's kappa coefficient
between response strategy and composite image category identi-
fication. Across all participants, response strategy and category
choice were positively related (K = .33, p = .05). The associations
between image component scores with image categorization and
response strategy were calculated separately using Pearson corre-
Table 3
Mean Image Scores by Image Scenario: Experiment 1
Image score
Ally
Enemy
Dependent
Barbarian
(n
M
2.42
1.16
1.00
0.32
Ally
= 25)
SD
1.50
0.99
0.85
0.78
Scenario
Enemy
(n = 24)
M
1.46
2.13
0.92
0.50
SD
1.61
1.23
1.02
0.66
Dependent
in = 21)
M
1.00
1.24
2.15
0.50
SD
1.15
1.00
1.31
0.69
Barbarian
in = 23)
M
1.14
1.55
1.22
1.04
SD
1.32
1.22
1.17
1.11
Note. Means in boldface indicate the average of correctly matched images selected for each scenario.
84 ALEXANDER. BREWER, AND HERRMANN
Table 4
Proportion of Image Category Choices as a Function of
Scenario: Experiment 1
Category choice
Ally
Enemy
Dependent
Barbarian
Ally
.84
.08
.08
.00
Enemy
.33
.42
.17
.08
Scenario
Dependent
.29
.14
.52
.05
Barbarian
.13
.39
.30
.18
Note. Values in boldface indicate the proportion of correctly matched
image categories selected for each scenario.
lation coefficients.
2
Ally, enemy, dependent, and barbarian image
component scores strongly corresponded with category choice
( r = .52, p = .01; r = .52, p = .01; r = .49, p = .01; and r = .32,
p = .01, respectively), indicating that a higher image score was
associated with greater likelihood of selecting the corresponding
image category. Most importantly, the ally, enemy, dependent, and
barbarian image scores were significantly correlated with response
strategy (r = .29, p = .01; r = .35, p = .01; r = .25, p = .05; and
r = .21, p = .05, respectively). Overall, these significant correla-
tions lend support to the functional relationship between behavior
responses and cognitive representations of the out-group.
Discussion of Experiment 1
Overall, the pattern of results indicates that the manipulated
intergroup relationships in Experiment 1 were effective in eliciting
cognitive schemas consistent with the predicted generic out-group
images. Behaviorally, in all scenarios, with the exception of the
dependent condition, the dominant response strategy selection was
that expected on the basis of the information provided in the
scenario about the context of intergroup relationships. At the
cognitive level, analyses of image scores suggest that the pattern of
image coherence was strong overall and that image scores varied
systematically as a function of the content of the scenario pre-
sented and the associated behavioral response.
Most importantly, analyses of both response strategies and im-
age scores showed evidence of differentiation among the three
negative images. However, these differences were somewhat di-
luted by the fact that the more socially desirable positive response
alternatives had higher base rates of selection than did components
from the negative image sets (see Footnote 1); this was also true
for the cooperative behavioral response strategy (see Table 2).
Further, the response options associated with the barbarian image
had an especially low rate of endorsement.
One reason that the socially desirable positive response alterna-
tives were selected frequently, even in the negative intergroup
conditions, may be that the level of emotional arousal in this
scenario study was fairly low. Although the general scenario
constructed for this first experiment was intended to be highly
involving to students on this campus, it was clearly hypothetical,
and level of arousal was not systematically manipulated or con-
trolled. Any emotional arousal created was integral to the inter-
group situation and varied in unknown ways across individual
participants and scenarios.
The level of arousal experienced could influence whether the
scenario information produced the corresponding behavioral incli-
nation. Behaviors such as attack or flight may depend on high
levels of arousal, whereas behaviors involving trust or sympathy
do not require such intensity. The perception of threat underlying
the barbarian image, in particular, may require a high level of
arousal associated with fear. To explore the role of physiological
arousal more systematically, we designed a second experiment to
examine the effect of incidental arousal on image activation.
Experiment 2
In the second experiment, we manipulated the content of the
intergroup scenario and arousal independently. Arousal was ma-
nipulated prior to exposure to the content of the scenario so we
could create different levels of incidental affect (Bodenhausen,
1993). Incidental affect was expected to influence the activation of
out-group images in two ways. First, in complex social situations,
arousal increases selective attention to salient cues and produces
greater reliance on stereotypes (Kim & Baron, 1988). This effect
of arousal would be expected to facilitate all images to the same
extent.
A second effect of incidental arousal would depend on excita-
tion transfer and misattribution processes (Zillmann, 1971) inter-
acting with the content of the scenario. More specifically, the
presence of arousal combined with goal incompatibility and dif-
ferential strength should increase the likelihood of experiencing
the perceived threat that is necessary to elicit behaviors associated
with barbarian and enemy images. This aspect of arousal would
lead to an interaction between arousal level and scenario content in
determining degree of activation of associated images.
To test these effects, we designed a new scenario that placed
participants in a hypothetical sales-team situation. Unlike the sce-
nario used in Experiment 1, the second scenario did not contain
any self-relevant content that would elicit involvement or arousal
on its own. To create different levels of arousal, we had partici-
pants complete one of two incidental arousal-inducing tasks before
reading any scenarios.
3
Four versions of the sales-team scenario
were created to produce a 2 (low arousal vs. high arousal) X 4 (ally
vs. enemy vs. dependent vs. barbarian intergroup scenario) facto-
rial design. Measures similar to those used in Experiment 1 to
assess behavioral and cognitive image activation were again used
in Experiment 2.
2
For purposes of these analyses, category and response options were
assigned an ordinal score reflecting their similarity to the most appropriate
response for the particular image. So, for correlations with the ally image,
the cooperation response was scored as 4, appeasement as 3, exploitation
as 2, and attack as 1, and the enemy image response scoring was the
reverse. For the barbarian image, appeasement was scored as 4, attack as 3,
cooperation as 2, and exploitation as 1, and the dependent image scoring
was the reverse.
3
A third arousal condition had been included to allow us to compare the
effects of different types of incidental arousal. This condition was not
successful in producing any evidence of physiological arousal, so it was
dropped from the experimental design.
IMAGES AND AFFECT
85
Method
Participants
A total of 63 (41 female and 22 male) introductory psychology students
at The Ohio State University-Mansfield participated in the experiment to
fulfill a course requirement.
Procedure
Individual participants signed up for an investigation of activity and task
performance. A female experimenter, blind to the hypotheses, greeted and
read a description of the study to each participant, informing the participant
that the study would be conducted in two phases. After describing the
possible tasks to be completed in each phase (i.e., reading a short story or
riding an exercise bicycle and responding to questionnaires), the experi-
menter acquired the participant's consent and then measured the partici-
pant's resting blood pressure and heart rate. In phase 1, participants were
randomly assigned to one of the two arousal conditions, and specific
instructions were given to the participant for completing the assigned task.
Heart rate was taken every 2 min, and blood pressure and self-reports of
arousal were assessed immediately following completion of the arousal
task. Participants were then given the intergroup scenario and asked to read
and imagine being a part of the situation described. They then answered
questions based on the scenario they read. Following completion of the
image-activation measures, participants were debriefed and thanked for
their participation.
Materials and Measures
Construction of intergroup scenarios. A generic intergroup scenario
was developed in which participants imagined being salespersons for a
company that produced baseball caps. Participants read that the company
was divided into two separate sales teams, their own and another, whose
goal was to sell as many caps as possible every year. As an incentive, the
CEO of the company recently offered a 5-day trip to Hawaii for members
of the sales team that sells 10,000 more caps by the end of the year. As the
leader of one of the sales teams, the participant's goal was to have the
in-group team sell enough caps to win the trip. The participant's sales team
was given some leverage to reach its goal. Specifically, their group was an
experienced, high-prestige sales force that had a strong history of produc-
tive selling. The participant's sales group also had a large market size, and
they had already started working on strategies to increase their productivity
to achieve the 10,000-cap selling goal. The competing sales team was also
described as very interested in winning the trip.
As in the first experiment, this general scenario was tailored with
specific information to manipulate four patterns of intergroup relations.
Information about the relative strength, prestige, and goals of the out-group
was included in each scenario. The four scenarios generated for this
experiment are presented in Appendix C.
Physiological arousal manipulation. Prior to reading scenarios, partic-
ipants were exposed to one of two arousal conditions, low or high arousal.
Both preliminary tasks took approximately 10 min to complete.
1. In the low arousal, or control, condition, minimal arousal was in-
duced. Participants read a short story, "The Mark on the Wall" by Virginia
Woolfe (1921/1979). They were told to take their time and to read it
thoroughly. Participants' average reading time for this short story was 10
min 45 s.
2. In the high arousal condition, participants rode a Schwinn stationary
bicycle for 10 min at a minimum of 20 mph (32 kph). The resistance level
on the bike was set at the same level for all participants. This manipulation
has been shown to produce reliable effects on blood pressure, heart rate,
and self-reported arousal (cf. Kim & Baron, 1988; Cantor, Zillmann, &
Bryant, 1975; Zillmann, Katcher, & Milavsky, 1972).
Arousal measures: Physiological. To ensure the arousal manipulations
worked as anticipated, we assessed heart rate and systolic and diastolic
blood pressure. Participants' blood pressure was measured at rest and
immediately following the arousal task, using an Omron blood pressure
monitor (Omron Corporation, Kyoto, Japan), while the participant was
seated. Heart rate was measured using a Labtron digital pulse meter
(Graham-Field, Bayshore, New York) attached to the index finger of the
nondominant hand. Heart rate measures were taken during a baseline rest
period, every 2 min during the arousal task, and while reading the scenario.
Arousal measures: Self-report. Participants rated, on 5-point scales
ranging from 1 {not at all) to 5 (extremely), the extent to which the task was
physically arousing, stressful, and relaxing. Relaxing scores were reversed,
and analyses were conducted on the aggregate of the three questions.
Image activation measures. Measures similar to those used in Exper-
iment 1 were used in Experiment 2. The image-related response strategies
were modified to fit the cap-selling scenario. As in the first experiment,
participants chose one from among four response strategies that their group
could use in dealing with the out-group. The four strategies were written to
reflect the behavioral orientations associated with each of the out-group
images. The options provided were as follows:
1. Cooperating with the out-group and combining their resources by
sharing their markets with the out-group (cooperation/ally)
2. Using their own market and best offense to actively go after the
out-group's market (attack/enemy)
3. Entering the out-group's market and getting control over it to help
their own in-group reach their quota (exploitation/dependent)
4. Protecting their corner of the market and selling as many caps as
possible, given what they have to work with (appeasement/barbarian)
The same image component questions and responses developed for
Experiment 1 were also used in Experiment 2. The categorical measure was
not included in this experiment.
Results
Arousal Manipulation Checks
Physiological measures. We conducted analyses on difference
scores between baseline and posttask measures of systolic blood
pressure. Analyses revealed the anticipated significant main effect,
F(l, 62) = 39.18, p = .0001. As expected, participants in the
bicycle condition (At = +14.62) had higher blood pressure in-
creases than those in the story condition (M = 5.06).
We computed heart rate difference scores by subtracting partic-
ipants' mean baseline responses from their aggregate heart rate
during the arousal task performance. One-way ANOVAs con-
ducted on heart rate difference scores yielded a significant main
effect for arousal condition, F(l, 62) = 100.07, p = .0001,
indicating that the stationary bicycle induced greater heart rate
increases (At = +51.21) than did the story (At = +.30). Com-
parisons with baseline were also made for heart rate measured
during the first 2 min of reading the scenario and were still
significantly different at that point, F(l, 62) = 56.81, p = .001
(Ms = +29.00 and 1.18 for high and low arousal conditions,
respectively).
Self-reported arousal. Participants' mean ratings of self-
perceived arousal were analyzed with a one-way ANOVA. Results
yielded the expected main effect between the arousal conditions,
F(l, 62) = 56.35, p = .0001, with participants in the exercise
bicycle condition reporting more arousal (At = 3.80) than partic-
ipants who read the story (At = 2.42).
86 ALEXANDER. BREWER, AND HERRMANN
Response Strategies
The frequency scores of participants' response selections were
analyzed with a logit analysis. A three-way frequency analysis
(Response Strategy X Scenario X Arousal Level) was performed
to develop a logit model of response selection. Predictors were
scenario (ally, enemy, dependent, or barbarian) and arousal level
(high or low). A nonhierarchical logit analysis produced a model
with two two-way associations (Scenario X Response Strategy and
Arousal X Response Strategy), the three-way association (Scenar-
io X Arousal X Response Strategy), and the first-order effect of
the dependent variable (Response Strategy). The model had an
adequate fit between observed and expected frequencies with a
likelihood ratio, G
2
(2) = 7.67, p = .10, and - 2 log x*(2, N =
63) = 10.68, p = .001.
The three-way association of response strategy as a joint func-
tion of scenario and arousal is presented in Table 5. In general,
high arousal produced a closer fit between scenario and response
strategy selection. The ally, enemy, and barbarian conditions were
clearly differentiated under high arousal. However, participants in
the dependent condition under high arousal tended to select the
barbarian strategy rather than the dependent (exploitation)
response.
Looking at each scenario separately, we found the selection of
response strategies for the ally and enemy scenarios was not
significantly affected by arousal: for ally, ^ ( 2, N = 18) = 0.96,
p = .62; for enemy, ^ ( 2, N = 15) = 4.04, p = .13. For the
barbarian scenario, the likelihood of selecting the barbarian (ap-
peasement) response strategy was significantly increased under
high arousal compared with low arousal, as expected, x
2
(2, N =
15) = 10.38, p = .005. By contrast, selection of the exploitation
response strategy in association with the dependent scenario was
significantly greater under low arousal than under high arousal,
)f{2, N = 15) = 10.42, p = .005.
Image Scores
As in Experiment 1, we computed image scores by summing the
number of component attributes of each image type the participant
Table 5
Response Strategy as a Function of Scenario and
Arousal Conditions: Experiment 2
Response
strategy
Ally
Enemy
Dependent
Barbarian
Ally
Enemy
Dependent
Barbarian
Ally
.60
.10
.20
.10
.88
.13
.00
.00
<
Enemy
Low arousal
.25
.38
.00
.38
High arousa
.14
.43
.14
.29
Scenario
Dependent
.13
.00
.63
.25
1
.29
.14
.14
.43
Barbarian
.25
.38
.25
.13
.00
.43
.00
.57
Note. Values in boldface indicate the proportion of correctly matched
response strategies selected for each scenario.
selected. To test whether mean image scores varied as a function
of out-group scenario and arousal condition, we conducted a 2
(high arousal vs. low arousal) X 4 (ally scenario vs. enemy
scenario vs. dependent scenario vs. barbarian scenario) X 4 (ally
score vs. enemy score vs. dependent score vs. barbarian score)
mixed ANOVA, with the last factor nested within the other two.
The two-way interaction between scenario and image score was
significant, F(9, 62) = 5.90, p = .0001, qualified by the significant
three-way interaction with arousal, F(9, 62) = 2.85, p = .001,
indicating that the image component scores were a joint function
of the intergroup manipulations and arousal (see Table 6). As in
Experiment 1, a subanalysis was conducted for the negative sce-
narios with the ally scenario removed. The results of a 2 X 3 X 4
ANOVA revealed a significant three-way interaction between
arousal, scenario, and image score, F(6, 44) = 3.03, p = .0001.
We conducted separate Arousal X Scenario ANOVAs on each
image score. The interaction was not significant for the ally scores,
F(3, 62) = 0.28, p = .80; the enemy scores, F(3, 62) = 0.11,/? =
.97; or the dependent scores, F(3, 62) = 0.82, p = .50. The
Scenario X Arousal interaction was significant for the barbarian
scores, F(3, 62) = 2.02, p = .05. Apparently, the interactive effect
of arousal was limited to activation of the barbarian image, which
was greater under high arousal than under low arousal.
Correlations Among Measures
We again computed correlations across conditions to examine
the correspondence between image component scores and re-
sponse choices. Collapsing across arousal conditions, we found
that ally, enemy, dependent, and barbarian image component
scores were significantly correlated with the response strategy
selected (r = 36, p = .01; r = .43,/? = .05; r = .27,p = .05; and
r = .43, p = .01, respectively).
Discussion of Experiment 2
By separating the cognitive and arousal mechanisms of the
generic out-group stereotypes examined here, we could study the
effects of incidental arousal on generic images. Under low arousal
conditions, the image score results from Experiment 2 essentially
replicated the patterns found in Experiment 1. Images were sys-
tematically related to the scenario information for the ally, enemy,
and dependent conditions, but this was less so for the barbarian
image. The relationships between ally, enemy, and dependent
scenarios and corresponding images were not altered under the
high arousal condition. However, high arousal was a significant
factor in eliciting the barbarian image. Apparently, information
about having goals incompatible with those of a high power and
low status out-group was not sufficient alone to evoke the barbar-
ian image strongly. Some level of arousal was required to induce
the level of defensive fear associated with the barbarian image.
The effect of arousal in this experiment, then, matched expecta-
tions based on excitation transfer effects interacting with specific
information content.
As expected from image theory, the activation of specific im-
ages correlated with the response orientation elicited by the sce-
nario information. In general, the effects of arousal on response
strategy selection mirrored the effects on image scores. In partic-
ular, the barbarian scenario was more effective in eliciting the
IMAGES AND AFFECT 87
Table 6
Mean Image Scores by Scenario and Arousal: Experiment 2
Image score
Ally
Enemy
Dependent
Barbarian
Ally
Enemy
Dependent
Barbarian
M
2.10
1.70
1.00
0.20
2.63
1.75
0.63
0.00
Ally
SD
1.44
1.15
0.89
0.42
1.68
1.03
0.74
0.00
M
Scenario
Enemy
SD
Low arousal
a
1.63
2.50
0.25
0.63
]
1.29
2.71
0.00
1.00
1.51
1.41
0.46
0.51
riigh arousal
13
1.38
1.11
0.00
0.32
Dependent
M
1.88
0.77
2.25
0.13
1.86
0.57
2.00
0.57
SD
1.80
0.66
1.58
0.39
1.58
0.53
1.29
0.53
Barbarian
M
1.38
1.00
1.71
0.88
0.86
1.14
0.86
2.14
SD
1.76
0.92
1.38
1.71
0.89
0.69
1.16
1.77
Note. Means in boldface indicate the average of correctly matched images selected for each scenario.
a
The number of participants in each scenario condition was as follows: ally, n = 10; enemy, n = 8; dependent,
n = 8; barbarian, n = 8.
b
The number of participants in each scenario condition was as follows: ally, n = 8;
enemy, n = 7; dependent, n = 7; barbarian, n = 1.
appeasement response strategy under high arousal compared with
low arousal, and the same was true for image scores. The one
exception to this pattern was the response strategy selection for the
dependent scenario. The exploitation strategy was not systemati-
cally selected in the high arousal condition even though it was by
far the response of choice under low arousal. Nonetheless, the
dependent image activation was strong for the dependent scenario
regardless of arousal. Thus, the response strategy results for the
dependent scenario in the high arousal condition are something of
an anomaly.
The results from these first two experiments effectively dem-
onstrate that rather richly detailed cognitive representations of
out-groups can be generated from relatively minimal informa-
tion about the structural relationships between in-groups and
out-groups. Given the forced-choice measures used in these
studies, however, we cannot say that these images were spon-
taneously activated by the intergroup context. Components of
the images were provided by the experimenter for endorsement
rather than freely generated. However, we can conclude from
this methodology that specific intergroup situations lead par-
ticipants to be ready to believe particular allegations about
out-group characteristics and to make attributions about inten-
tions and motivations. Further, the characterizations of the
out-group that were endorsed were consistent with the preferred
behavioral response toward that out-group, implicating the
functional role that such images play in explaining and justify-
ing intergroup behavior on the part of the in-group. We are
aware, however, that the evidence for generic out-group images
is limited without examining the spontaneous activation of the
stereotypes. To further substantiate our proposed generic ste-
reotypes, we conducted a third experiment that measured spon-
taneous endorsement of the stereotypes examined in Experi-
ments 1 and 2.
Experiment 3
An implicit measure of stereotype activation was designed and
used to assess participants' activation of the four images measured
in the preceding experiments. The same campus intergroup sce-
narios that were used in Experiment 1 were given to participants in
the third experiment. After reading one of the four scenarios,
participants completed what they believed was a questionnaire for
a separate study involving open-ended responses to 19 sentences.
This task provided the implicit measure of image activation.
Method
Participants
Ninety-four (53 female and 41 male) undergraduate students from The
Ohio State University-Mansfield participated in this study as partial ful-
fillment for a requirement in their introductory psychology course.
Procedure
As in Experiment 1, participants believed that the study was designed to
assess person perception and cognition. Individuals were run in groups of
approximately 4, and each participant was randomly assigned one of the
four intergroup relationship scenarios as the materials were distributed.
Participants were told that they were to read a hypothetical scenario,
imagine being in the scenario, and think about how they would respond to
the other group in the scenario. Participants were also asked to imagine
being the leader of their student group and to think about how they would
justify their response to the other members of their in-group, who were
those other participants being run in the same session. To enhance the level
of involvement and potential arousal further, participants expected that
there would be a discussion later in the session during which each would
explain his or her responses to the scenario to other group members.
After all participants finished reading the scenario, the experimenter
explained that another professor had requested time during this experimen-
tal session for participants to complete a brief questionnaire on impression
ALEXANDER. BREWER. AND HERRMANN
formation. They were further told that to avoid interfering with later
discussion, the questionnaire would be completed at this point in the
session, prior to the group discussion. Participants were then given the
sentence completion measure. After completing this task, participants
responded to manipulation checks and the response selection items used in
Experiment 1. Participants were then told they would not have the group
discussion after all; they were debriefed and then excused.
Materials and Measures
The same four scenarios used in Experiment 1 were used in Experi-
ment 3 (see Appendix A).
Implicit image activation measure. To assess whether the intergroup
scenarios had primed the respective images, we devised an indirect measure of
activation.
4
To construct this measure, we wrote 19 ambiguous sentence stems.
Each sentence described a social interaction of some kind involving the
behavior of one actor toward some other person or persons.
5
Participants were
asked to complete each sentence by providing a brief reason that described
why the main actor behaved the way he or she did. They were asked to write
the first description that came to mind for each sentence.
The sentences were sufficiently ambiguous as to the motive or intention
of the actor that respondents had to project an explanation for the behavior.
Each sentence allowed for at least two different interpretations, depending
on what assumptions were made about the actor's character. More specif-
ically, character attributions could have been generated that corresponded
to the stereotype content associated with one or more of the generic images.
For example, the sentence "Mike ran up to a resident in his dorm and
tackled him to the floor because . . . " could have generated an attribution
that Mike was crude and aggressive (barbarian image), that Mike and his
dormmate were engaged in a physical fight (enemy image), that Mike was
acting immature and childish (dependent), or that he was simply being
friendly and playful (ally image). Not all of the 19 sentences could have
generated logical responses corresponding to each of the four images, but
every sentence had at least two possible negative interpretations as well as
either positive or neutral explanations. Collectively, the set of 19 sentences
provided equal opportunity for all of the images to be expressed.
Coding of sentence completions. Two judges, blind to the hypotheses
and conditions, coded respondents' explanations for each of the sentences.
We developed coding instructions to specify categories of characterolog-
ical responses that were deemed to be representative of the ally, enemy,
dependent, and barbarian images. Coders were trained to evaluate the
entire description generated for each sentence and to place it in one and
only one of the response categories provided.
Four of the coding categories were defined by motivational and disposi-
tional characteristics uniquely associated with the four images under study. To
avoid any effects of category labels, we arranged for the response categories to
be identified by a symbol rather than a category name.
6
Coders were instructed
to classify a response as N; if it conveyed a character judgment of the actor as
crude, uncontrolled, ruthless, aggressive, irrational, destructive, or thought-
lessly careless. The N
2
category was to be used for responses that implied a
character assessment consisting of competitive, controlled but hostile, calcu-
lating, or deceitful traits. The N
3
category was for descriptions implying the
actor was naive, dependent, needy, indecisive, or childlike. Category P was
used for positive descriptors that conveyed the idea that the actor was coop-
erative, trustworthy, dependable, considerate, helpful, friendly, or morally
principled. Finally, a neutral category (X) was used to classify explanations
that had no characterological importance (such as situational explanations or
benign attributions).
Ratings of all sentences were done by both judges independently, and
their codings were compared for interjudge reliability. The codings proved
to be reasonably consistent (KS = .74, .71, .76, .70, and .69, ps = .01, for
ally, enemy, dependent, barbarian, and neutral descriptions, respectively).
Cases in which the judges disagreed were examined and reconciled so that
only one code was assigned to every sentence description. We computed
four image scores for each respondent by summing the number of ally,
enemy, barbarian, and dependent descriptors given by the respondent
across the 19 sentences.
7
Results
Responses to the manipulation check and the response strategy
measures replicated the findings from Experiment 1, indicating
that the scenarios were interpreted as intended.
We analyzed image scores from the implicit measure in a 4 (ally
scenario vs. enemy scenario vs. dependent scenario vs. barbarian
scenario) X 4 (ally score vs. enemy score vs. dependent score vs.
barbarian score) ANOVA, with score nested in scenario. A signif-
icant overall interaction emerged, F(9, 93) = 19.58, p = .0001,
indicating that respondents' sentence completions were systemat-
ically affected by the intergroup manipulations represented in the
different scenarios. We conducted a significant 3 X 4 mixed
ANOVA with the ally scenario removed, F(6, 73) = 19.26, p =
.0001, and confirmed that the three negative scenarios were sig-
nificantly differentiated. The mean implicit image scores by sce-
nario are presented in Table 7.
We conducted one-way ANOVAs on each Image Score X
Scenario using the omnibus interaction error term to determine if
the mean score varied by scenario in the predicted pattern. The
ANOVA conducted on the ally image scores revealed a main
effect for scenario, F(3,93) = 9.60, p = .0001. Simple effects tests
indicated that ally scores were greatest for those participants
exposed to an ally scenario, compared with those in the enemy
condition, F(3, 93) = 21.61, p = .0001; the dependent condition,
F(3, 93) = 24.72, p = .0001; or the barbarian condition, F(3,
93) = 25.85, p = .0001.
The ANOVA conducted on the enemy image scores also yielded
a main effect for scenario, F(3, 93) = 17.97, p = .0001. Simple
effects tests indicated that those exposed to the enemy scenario had
significantly higher scores than those exposed to the ally, F(3,
93) = 42.82, p = .0001; dependent, F(3, 93) = 27.55, p = .0001;
and barbarian scenarios, F(3, 93) = 46.42, p = .0001.
The analysis of the dependent image scores revealed a signifi-
cant main effect for scenario, F(3, 93) = 13.68, p = .0001, as well,
4
The test of priming we used here was a conservative one because the
content of the priming stimuli (intergroup scenarios) and the content of the
implicit measures were not commensurate. Experiments reported by Abel-
son, Dasgupta, Park, and Banaji (1998) indicate that priming effects are
weaker when the prime and the judgment task are not matched with respect
to group versus individual level of representation. Nonetheless, we chose to
have an implicit measure based on forming impressions of individual
persons to avoid any likely demand characteristics or suspicion about the
relationship between the two tasks.
5
Sample sentences used were, "Carla makes dinner for her best friend's
boyfriend when her friend is out of town because . . . ," "Jeff called his
parents from college because . . . , " "Spence bought a gun at the local pawn
shop because . . . "
6
N
l
corresponded to the barbarian category, N
2
to the enemy category,
and N
3
to the dependent category. P was used to refer to the ally category,
and X was used for the neutral category.
7
Because both positive and neutral responses implied putting the best
possible construction on the actor's behavior, neutral ratings were aggre-
gated with ratings of positive responses indicative of the ally image.
Therefore, respondents' ally scores on this measure include neutral re-
sponses.
IMAGES AND AFFECT 89
Table 7
Mean Implicit Image Activation Scores by Scenario: Experiment 3
Image score
Ally
Enemy
Dependent
Barbarian
(n
M
10.81
2.38
4.29
1.52
Ally
= 21)
SD
3.22
1.77
1.68
1.40
Scenario
Enemy
O = 26)
M
7.31
7.31
3.34
1.77
SD
2.92
3.45
2.15
3.02
Dependent
(n = 24)
M
7.00
3.50
7.17
1.33
SD
2.59
2.65
2.76
1.37
Barbarian
{n = 23)
M
6.87
2.30
3.96
5.87
SD
2.58
2.67
2.27
3.18
Note. Means in boldface indicate the average matching images generated for each scenario.
reflecting the finding that dependent scores were greater for the
dependent condition than for the ally, F(3, 93) = 13.88,p = .0001,
enemy, F(3, 93) = 27.54, p = .0001, and barbarian conditions,
F(3, 93) = 18.16, p = .0001.
Finally, analyses on the barbarian image scores also yielded a
significant main effect for scenario, F(3, 93) = 18.36, p = .0001,
with participants in the barbarian condition showing significantly
higher barbarian scores than those in the ally, F(3, 93) = 31.50,
p = .0001; enemy, F(3, 93) = 31.13, p = .0001; and dependent
conditions, F(3, 93) = 35.40, p = .0001.
In sum, results from this implicit measure of spontaneous image
activation proved sensitive to the implications of the intergroup
scenarios in all cases. Despite the fact that the differences between
response codings classified as enemy or barbarian were quite
subtle, the differentiation between enemy and barbarian conditions
was highly significant. If anything, these results were even stron-
ger than when image activation was assessed with responses
provided by the experimenters. These findings provide strong
evidence that the predicted images are spontaneously generated
when participants are asked to think about how they would re-
spond to specific intergroup situations.
General Discussion
The results of all three experiments strongly support a functional
theory of stereotype formation. Participants generating mental
representations of an out-group were responsive to some basic
structural characteristics of the given intergroup context in a man-
ner consistent with the self-justification hypotheses underlying
image theory. It is also important to note that the components of
the group images studied here involve beliefs about the motiva-
tions, intentions, and internal structure of a group qua group. This
is consistent with the idea that collectives can be represented as
active entities (e.g., Abelson et al., 1998) and reflects the origin of
image theory in the study of international relations, where nation-
states are treated as actors. It is not difficult to imagine, however,
that representations of the traits and dispositions of individual
group members could also be derived from the image of the group
as a whole. Thus, the content of social stereotypes may reflect the
intergroup context in which they are initially formed.
The findings from these studies are particularly important in
demonstrating systematic differentiation among negative images
of the out-group. Although three of the four intergroup scenarios
used in these experiments were expected to induce negative atti-
tudes and evaluations of the out-group, it is clear that the associ-
ated representations were not indiscriminately negative but differ-
entiated as a function of out-group power and status. These results
reinforce the suggestion that we ought to pay more attention to
prejudice as a product of social emotions and to the extent to which
stereotypes reflect the emotions that are evoked in particular
intergroup situations (E. R. Smith, 1993; Vanman & Miller, 1993).
Just as the different negative emotions of fear, disgust, and anger
give rise to different action tendencies, they also give rise to
different stereotypic representations of the out-group that evoke
that particular emotional response.
The primary purpose of the research reported here was to
demonstrate that the theory of images developed in the context of
international relations can be generalized to other intergroup con-
texts as well. Further, we wanted to test the idea that these images
have the properties of generic schemas that can be activated by a
few relational cuesan idea that has never before been tested
experimentally. For these purposes, we focused on four specific
images among the many that have been identified in the interna-
tional relations literature. These particular four images were se-
lected to represent both symmetric (ally and enemy) and asym-
metric (dependent and barbarian) intergroup relationships. Further,
these images represent pairs of intergroup contexts in which only
one relational dimension is changed (i.e., the ally and enemy
relationships are the same except for the goal compatibility dimen-
sion; the barbarian and dependent situations vary only on the
relative strength dimension). In most cases, changing only one cue
resulted in significant differences in image scores.
The asymmetric relationships that give rise to the dependent and
barbarian images have complementary images (imperialist and de-
generate) that were not studied here but would be of interest in their
own right. Further understanding of these generic images could make
sense of the divergent stereotypes that members of dominant and
minority groups in a particular society hold of each other (e.g., Judd,
Park, Ryan, Brauer, & Kraus, 1995). Whether White Americans
perceive African Americans in terms of the dependent or the barbar-
ian image would be expected to reflect the degree of perceived threat
from the out-group, and perceptions of an out-group as enemy or
imperialist reflect beliefs about relative status and power. A functional
perspective on intergroup stereotypes can both elucidate the implicit
90 ALEXANDER, BREWER, AND HERRMANN
relationships that underlie the content of group beliefs and suggest the
conditions under which stereotypes can be changed.
References
Abelson, R., Dasgupta. N., Park, J., & Banaji, M. (1998). Perceptions of
the collective other. Personality and Social Psychology Review, 2,
243-250.
Bodenhausen, G. V. (1993). Emotions, arousal, and stereotypic judgments:
A heuristic model of affect and stereotyping. In D. M. Mackie & D. L.
Hamilton (Eds.), Affect, cognition, and stereotyping: Interactive pro-
cesses in group perception (pp. 13-37). San Diego, CA: Academic
Press.
Boulding, K. (1956). The image. Ann Arbor: University of Michigan Press.
Boulding, K. (1959). National images and international systems. Journal of
Conflict Resolution, 3, 120-131.
Bouris, R. Y., Turner, J. C, & Gagnon, A. (1997). Interdependence, social
identity, and discrimination. In R. Spears, P. Oakes, N. Ellemers, & S. A.
Haslam (Eds.), The social psychology of stereotyping and group life (pp.
273-295). Oxford, UK: Blackwell.
Bronfenbrenner, U. (1961). The mirror image in Soviet-American rela-
tions: A social psychologist's report. Journal of Social Issues, 17, 45-56.
Campbell, D. T. (1967). Stereotypes and the perception of out-group
differences. American Psychologist, 22, 812-829.
Cantor, J. R., Zillmann, D., & Bryant, J. (1975). Enhancement of experi-
enced sexual arousal in response to erotic stimuli through misattribution
of unrelated residual excitation. Journal of Personality and Social Psy-
chology, 35, 69-75.
Cottam, M. L. (1986). Foreign policy decision making: The influence of
cognition. Boulder, CO: Westview Press.
Cottam, M. L. (1994). Images and intervention: U.S. policies in Latin
America. Pittsburgh, PA: University of Pittsburgh Press.
Cottam, R. (1977). Foreign policy motivation: A general theory and a case
study. Pittsburgh, PA: University of Pittsburgh Press.
Eagly, A. H. (1987). Sex differences in social behavior: A social-role
interpretation. Hillsdale, NJ: Erlbaum.
Fiske, S. T. (1998). Stereotyping, prejudice, and discrimination. In D. T.
Gilbert, S. T. Fiske, & G. Lindzey (Eds.), The handbook of social
psychology. (Vol. 2, 4th ed., pp. 357-411). New York: McGraw-Hill.
Fiske, S. T., Xu, J., Cuddy, A., & Glick, P. (in press). Respect versus
liking: Status and interdependence underlie ambivalent stereotypes.
Journal of Social Issues.
Heider, F. (1958). The psychology of interpersonal relations. New York:
Wiley.
Herrmann, R. K. (1985). Perceptions and behavior in Soviet foreign policy.
Pittsburgh, PA: University of Pittsburgh Press.
Herrmann, R. K., & Fischerkeller, M. (1995). Beyond the enemy image
and spiral model: Cognitive-strategic research after the cold war. Inter-
national Organization, 49, 415-450.
Herrmann, R. K., Voss, J., Schooler, T., & Ciarrochi, J. (1997). Images in
international relations: An experimental test of cognitive schemata.
International Studies Quarterly, 41, 403-433.
Holsti, O. (1967). Cognitive dynamics and images of the enemy. In D.
Finley, O. Holsti, & R. Fagen (Eds.), Enemies in politics (pp. 25-96).
Chicago: Rand McNally.
Insko, C. A., & Schopler, J. (1987). Categorization, competition, and collec-
tivity. In C. Hendrick (Ed.), Group processes: Review of personality and
social psychology (Vol. 8, pp. 213-251). Newbury Park, CA: Sage.
Jervis, R. (1976). Perception and misperception in international politics.
Princeton, NJ: Princeton University Press.
Jervis, R., & Snyder, J. (1991). Dominoes and bandwagons: Strategic
beliefs and great power competition in the Eurasian rimland. Oxford.
UK: Oxford University Press.
Jost, J. T., & Banaji, M. R. (1994). The role of stereotyping in system-
justification and the production of false consciousness. British Journal of
Social Psychology, 33, 1-27.
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, 69, 460-481.
Kim, H. S., & Baron. R. S. (1988). Exercise and the illusory correlation:
Does arousal heighten stereotypic processing? Journal of Experimental
Social Psychology, 24, 366-380.
Osgood, C. (1962). An alternative to war or surrender. Urbana: University
of Illinois Press.
Sande, G. N., Goethals, G. R., Ferrari, L., & Worth, L. T. (1989). Value-
guided attributions: Maintaining the moral self-image and the diabolical
enemy-image. Journal of Social Issues, 45, 91-118.
Sherif, M., Harvey, O. J., White, B. J., Hood, W. R., & Sherif, C. W.
(1961). Intergroup conflict and cooperation: The Robbers Cave exper-
iment. Norman: University of Oklahoma, Institute of Intergroup Rela-
tions.
Sherif, M., & Sherif, C. W. (1953). Groups in harmony and tension: An
integration of studies on intergroup relations. New York: Harper.
Shimko, K. (1991). Images and arms control: Perceptions of the Soviet
Union in the Reagan administration. Ann Arbor: University of Michigan
Press.
Silverstein, B. (1989). Enemy images: The psychology of U.S. attitudes
and cognition regarding the Soviet Union. American Psychologist, 44,
903-913.
Smith, E. R. (1993). Social identity and social emotions: Toward new
conceptualizations of prejudice. In D. M. Mackie & D. L. Hamilton
(Eds.), Affect, cognition, and stereotyping: Interactive processes in
group perception (pp. 297-325). San Diego, CA: Academic Press.
Smith, M. B., Bruner, J. S., & White, R. W. (1956). Opinions and
personality. New York: Wiley.
Stuart, D., & Starr, H. (1982). Inherent bad faith reconsidered: Dulles,
Kennedy, and Kissinger. Political Psychology. 3, 1-33.
Tajfel, H. (1981). Social stereotypes and social groups. In J. C. Turner &
H. Giles (Eds.), Intergroup behavior (pp. 144-162). Oxford, UK: Basil
Blackwell.
Vanman, E. J., & Miller, N. (1993). Applications of emotion theory and
research to stereotyping and intergroup relations. In D. M. Mackie & D. L.
Hamilton (Eds.), Affect, cognition, and stereotyping: Interactive processes
in group perception (pp. 213-238). San Diego, CA: Academic Press.
White, R. K. (1965). Images in the context of international conflict. In
H. C. Kelman (Ed.), International behavior: A social-psychological
analysis (pp. 236-276). New York: Holt, Rinehart & Winston.
White, R. K. (1968). Nobody wanted war: Misperception in Vietnam and
other wars. Garden City, NY: Doubleday.
Woolfe, V. (1979). The mark on the wall. In M. Abrams (Ed.), Norton
anthology of English literature (4th ed., Vol. 2, pp. 2025-2031). New
York: Norton. (Original work published 1921)
Zillmann, D. (1971). Excitation transfer in communication-mediated aggres-
sive behavior. Journal of Experimental Social Psychology, 7, 419-434.
Zillmann, D., Katcher, A. H., & Milavsky, B. (1972). Excitation transfer
from physical exercise to subsequent aggressive behavior. Journal of
Experimental Social Psychology, 8, 247-259.
IMAGES AND AFFECT 91
-g
m
"S
s
o
Q
c
<D
O H
<
O
p
Si
c
Hsiil:
; ~ "S 8 CJ ^ i
; ; l l
OJ T3 CJ
== o
*""" D. S ,_,
1 5 1 1 ^ 8
S 5 = 3
- D o u '^ o ? o
S SS
* ! E o
lip
o '3 a &
o
c
Q
c
o
5 5^
z;
G
QJ
too
H
o
a
r
d
o
o
l
o
u
"re
l
i
t
i
c
o
4)
CJ
h
a
n
CJ
CJ
1
V3
n
o
i
4>
re
e
r
e
H
re
j =
o
o
ab
o
w
n
.
o .
n
<u
_c
c
o
-2 E
w O M 0>
4) C ^j ^_, c
, 73 Ut
tL
3
~
4> >. ^O W "
o
O H
O
r
c
e
s
e
r
s
m
e
m
b
s
o
m
a
i
n
v
e

o
b
t
cc
a
n
d

1
55 g c
a 5 -
s as
* = "s"I
till
o re o P,
4) .y " 53
t re 4)
CD - a
K= 4J
o ^ u
CJ ^-. >-.
4>
g l
. 3 re en
^ re "i
a| s
i
ao
So - S
^ 7 WJ
o "3 o
> u .
o u
>> 3
2 S
C o
3 -a
o
k
i
r
w
o
r
S
1

o
s
0 1
oo-a
o
t
h
e
o
1
o
r
'
1
b
a
c
. _
-
y
e
a
o
00
3
A
3
1
i
c
h
i
&
o
o
o
u
r
s
w
o
t
a
r
e
CL
3
O
u
t
-
g
re
re
i
e
g
r
e
e
a
g
o
a
t
h
i
s
i
e
v
e

i
a
c
h
g a.
2 '- v
in
o g a
3
a
m
i
o
CJ
a)
:
i
a
l
C
4)
m
o
i
>
3
S
3
:
S
O
U
!
13
u
s
t
a
t
00
c
o
l
t
y
t
G

o
CJ
t
h
e
re
-o
4)
o
a
p
u
s

C3
a
l
e
i
o
n
r
e
g
:
W)
3
O
>-,
_ r
8
CJ
w5
CO
I
f

t
l
C
re
c
o
"2
CO
3
n,
re
Q
e
g
e
o
i
l
c
a
l
1
u
n
i
t
S
3
:
h
a
t
3
O
D
d
>
n
l
y
i .
re
O
-a
4)
?J 61
. 3 (D
u S tS S
o j W ) c 3 i - ' O j 4 > a j c / ; C
-is -Q aj * ^* OJ o
*3 4J * O '
:
- _ c ^ o o
S *- 3 O O 4J ^ 4) 4>
3 -
S
- o"J ^ "2 1? ^ '
E -S 2 S 4 < 8
3 <
o
3
O
4J
s
re
O H

o
o
re
o
D
U
I
O
<D
CD
o
u
r
c
! "
c3
re
JZ
CJ
re
1
re
CD
&0
4)
c
o
l
l
>\
'2

g
o
op OJ - '
3 ^
a S
5
rt 0> CO ^
J 3 J 3 <U ^
o
^ 3
5
<S <H
c
^- *^ o
O
s
3 o a
1 ^
11
a 3
I
92 ALEXANDER, BREWER, AND HERRMANN
Appendix B
Components of Each Out-Group Image
Ally
The group values cooperative solutions to problems and tries to avoid
conflict.
The group will not exploit our trust in them but instead reciprocate and
contribute their fair share.
The group is motivated by legitimate and reasonable concerns and
aspirations.
Members in the group like their leaders because they think of them as noble
people.
The group has an effective democratic decision-making structure.
Enemy
The group's objectives are self-centered and harmful to others.
The group would take advantage of any efforts on our part to cooperate,
and they would even try to exploit us.
The group is extremely competitive and wants to dominate but will play by
the rules.
Most people in the group don't question their leaders, not because they like
them, but because they assume they are strong and intelligent.
The group has a strict, well-organized authority structure for decision
makin?.
Dependent
In the other group, people who are interested in the group's welfare, as
opposed to interested in only their own personal gain, will cooperate
with our group.
People in the group are quite naive; they mean well but need guidance and
leadership from outside their group.
Most people in the group want to have things better for the group, but they
lack discipline and are not likely to work very hard.
Moderate and responsible leaders in the group are struggling to prevail
over potential radical leaders in the group.
Those making decisions for the group are weak and inefficient.
Barbarian
The group enjoys getting its way even if it spoils things for others.
The group takes whatever it wants from other groups.
The group is crude, unsophisticated, and willing to cheat to get its way.
The leaders of the group are ruthless and stay in power only as long as they
are the biggest and the strongest.
There is no clear decision-making structure within the group, so the leaders
can get away with anything they want.
IMAGES AND AFFECT 93
u
3
c
C
I
a.
x
ffl
o
3
O
0>
Pi
2
3
a
&
00
CU
>. , . H o i
S " 3 =^"g |
g | , 5 g o
u 5 3 s

ra
u- S O "S Q,
03 O ^
rt
3
J 8 u u t g
"5 o c
- o
= " c " 1
sill
0)
1 S
1 i
Iflif
!
JS >> ~ O
! - _ C
u JS * cu
3 HI
* ' C *re
* " CO
! > NU
CU 2
> w CQ U
I I a
c
re
cu
re
CO
C
t
e
a
m

n
*
j _
c
o
<u
re
1 ?
H
S
r

t
e
a
3
s

a
s

y
o
8
s
o
u
r
ca
O
CO
4)
C9
a
n
d
a
c
CU
(D
00
t.re
CU
c
re
>,
xi
o
n
,

t
I
n

a
d
d
i
a
r
k
e

.a
o
re
a
o
:
a
r
y
,
i
f

n
e
c
e
;
a
r
k
e
a
o
i
t
:
p
e
t
Q
r
o
n
g
e
y

a
r
e
J:
T
E
re
s S "2 " >;
a JS E
oo S ^ =
2 g , . 2
:
s >
1 -s 1 g
A w _ -f- !:
t i. re y^ ^* oo
S ^ 2 " u
^ <u ,43 x ; c
8
w
'Z
Si a
o E o o
03
S
6
5b
111 I
o o
=
s M
o S o
7 ) 1
:
r
o
u
00
3
O
1
o
0)
o
<D
p
e
o
l
e
s
re
(U 0
U
9
J
J
i
n
e
x
p
e
:
i
g
e
"S
ex
XJ
1
h
a
v
e

a
:
o
o
c
_
o
r
o
u
p
y
o
u
r

g
i
n
p
o
a.
o
>,
S
s
t
r
o
n
g
p

a
r
e
3
2
00
"re
- a
g
o
00
re
oo
r
o
n
g

:
re
E
o
t
h
e

c
II Si
O CO *- - n
^ I <u re
1 g a.E
o
O
. P e g
S
:
3 o S ,
S
U 00
i
S
S d !
S< a. I
P ja
S -5 -S 2
o 3
o
a
o l
- cu.S -
g '^ > u
re ** ^ 3
oj U . 2 efl
- > 4J X)
<D H ^ 2
"3 8 I t
ra
,S -
1 1 1 .s
VH
3

" So
1
"
S o c o
S -8 =
!
sa g s
00 cu 00 2:
3 O 3 S
.5 -S 3 2 8
UJ ^ ^^ ^^
. 2 H
"S S E |
S O ca
j ail
re _
- o - re
fill
w
e
c
- '
Jill
8 ti
re *
S ^
E ^
>^:= _ =
o
3 3 3 _2 ?i O
rtlll
.. " S" u
' S S ? 5
' 5 O
.*Z Q, ,
^ _g =
O
1
<y OD
C 4J '
00 00
ON ON
ON ON
g
CO
CD
Q
8
o
CD
0>
1CJ
u
Q
1
s. u
o

Vous aimerez peut-être aussi